开源软件构建与测试

Changes of Revision 2

_service:extract_file:libjpeg-turbo.spec Deleted

@@ -1,179 +0,0 @@
-Name:           libjpeg-turbo
-Version:        2.1.1
-Release:        2
-Summary:        MMX/SSE2/SIMD accelerated libjpeg-compatible JPEG codec library
-License:        IJG
-URL:            http://sourceforge.net/projects/libjpeg-turbo
-Source0:        http://downloads.sourceforge.net/libjpeg-turbo/libjpeg-turbo-%{version}.tar.gz
-
-BuildRequires:  gcc cmake libtool nasm
-
-Obsoletes:      libjpeg < 6b-47 turbojpeg < %{version}-%{release}  
-Provides:       libjpeg = 6b-47 turbojpeg = %{version}-%{release}  
-
-Provides:       %{name}-utils = %{version}-%{release}
-Obsoletes:      %{name}-utils < %{version}-%{release}
-
-%description
-libjpeg-turbo is a JPEG image codec that uses SIMD instructions (MMX, SSE2, NEON, AltiVec)
-to accelerate baseline JPEG compression and decompression on x86, x86-64, and ARM systems.
-
-%package        devel
-Summary:        Development files for the libjpeg-turbo library
-Requires:       libjpeg-turbo = %{version}-%{release}
-Provides:       libjpeg-turbo-static = 1.3.1 turbojpeg-devel = %{version}-%{release}
-Obsoletes:      libjpeg-turbo-static < 1.3.1 turbojpeg-devel < %{version}-%{release}
-Provides:       libjpeg-devel = 6b-47 libjpeg-devel%{?_isa} = %{version}-%{release}
-Obsoletes:      libjpeg-devel < 6b-47
-
-
-%description devel
-Development files for the libjpeg-turbo library.
-
-%package utils
-Summary:        Utilities for manipulating JPEG images
-Requires:       libjpeg-turbo%{?_isa} = %{version}-%{release}
- 
-%description utils
-The libjpeg-turbo-utils package contains simple client programs for accessing
-the libjpeg functions. It contains cjpeg, djpeg, jpegtran, rdjpgcom and
-wrjpgcom. Cjpeg compresses an image file into JPEG format. Djpeg decompresses a
-JPEG file into a regular image file. Jpegtran can perform various useful
-transformations on JPEG files. Rdjpgcom displays any text comments included in a
-JPEG file. Wrjpgcom inserts text comments into a JPEG file.
-
-%package -n turbojpeg
-Summary:        TurboJPEG library
- 
-%description -n turbojpeg
-The turbojpeg package contains the TurboJPEG shared library.
-
-%package -n turbojpeg-devel
-Summary:        Headers for the TurboJPEG library
-Requires:       turbojpeg%{?_isa} = %{version}-%{release}
- 
-%description -n turbojpeg-devel
-This package contains header files necessary for developing programs which will
-manipulate JPEG files using the TurboJPEG library.
-
-%package_help
-
-%prep
-%autosetup -n %{name}-%{version} -p1
-
-%build
-%{cmake} -DCMAKE_SKIP_RPATH:BOOL=YES -DCMAKE_SKIP_INSTALL_RPATH:BOOL=YES -DENABLE_STATIC:BOOL=NO .
-
-%make_build V=1
-
-%install
-%make_install
-%delete_la
-
-chmod -x README.md
-
-%ifarch x86_64
-  %global wordsize "64"
-%else
-  %global wordsize ""
-%endif
-
-if test -n "$wordsize"
-then
-  pushd $RPM_BUILD_ROOT%{_includedir}
-    mv jconfig.h jconfig-$wordsize.h
-    cat > jconfig.h <<EOF
-#ifndef JCONFIG_H_MULTILIB
-#define JCONFIG_H_MULTILIB
-
-#include <bits/wordsize.h>
-
-#if __WORDSIZE == 32
-# include "jconfig-32.h"
-#elif __WORDSIZE == 64
-# include "jconfig-64.h"
-#else
-# error "unexpected value for __WORDSIZE macro"
-#endif
-
-#endif
-EOF
-fi
-    
-%check
-LD_LIBRARY_PATH=%{buildroot}%{_libdir} make test %{?_smp_mflags}
-
-%ldconfig_scriptlets
-
-%files
-%defattr(-,root,root)
-%doc README.*
-%license LICENSE.md
-%{_libdir}/libjpeg.so.62*
-%exclude /usr/share/doc/libjpeg-turbo/*
-
-%files devel
-%doc coderules.txt jconfig.txt libjpeg.txt structure.txt example.txt 
-%exclude %{_includedir}/turbojpeg.h
-%{_includedir}/*.h
-%{_libdir}/libjpeg.so
-%{_libdir}/pkgconfig/libjpeg.pc
-%{_libdir}/cmake/%{name}/%{name}*.cmake
-
-%files utils
-%doc usage.txt wizard.txt
-%{_bindir}/cjpeg
-%{_bindir}/djpeg
-%{_bindir}/jpegtran
-%{_bindir}/rdjpgcom
-%{_bindir}/wrjpgcom
-%{_bindir}/tjbench
-
-%files -n turbojpeg
-%license LICENSE.md
-%{_libdir}/libturbojpeg.so.0*
-
-%files -n turbojpeg-devel
-%doc tjexample.c
-%{_includedir}/turbojpeg.h
-%{_libdir}/libturbojpeg.so
-%{_libdir}/pkgconfig/libturbojpeg.pc
-
-%files help
-%doc ChangeLog.md
-%{_mandir}/man1/*.1*
-
-%changelog
-* Wed Mar 30 2022 liuyumeng <liuyumeng5@h-partners.com> - 2.1.1-2
-- enable tests
-
-* Sat Nov 27 2021 wangkerong <wangkerong@huawei.com> - 2.1.1-1
-- update to 2.1.1
-- splite utils turbojpeg turbojpeg-devel subpackages
-
-* Wed Jan 27 2021 hanhui <hanhui15@huawei.com> - 2.0.6-1
-- Type: enhancement
-- ID:   NA
-- SUG:  NA
-- DESC: update to 2.0.6
-
-* Wed Jul 22 2020 openEuler Buildteam <buildteam@openeuler.org> - 2.0.5-1
-- Type:enhancement
-- Id:NA
-- SUG:NA
-- DESC:upgrade to 2.0.5
-
-* Tue Jan 14 2020 openEuler Buildteam <buildteam@openeuler.org> - 2.0.0-4
-- Type:bugfix
-- Id:NA
-- SUG:NA
-- DESC:remove the libjpeg-devel with arch in obsoletes
-
-* Fri Nov 1 2019 openEuler Buildteam <buildteam@openeuler.org> - 2.0.0-3
-- Type:bugfix
-- Id:NA
-- SUG:NA
-- DESC:add the libjpeg-devel and jconfig.h
-
-* Sat Sep 21 2019 Lijin Yang <yanglijin@huawei.com> - 2.0.0-2
-- Package init

_service:tar_scm:libjpeg-turbo.spec Added

@@ -0,0 +1,179 @@
+Name:           libjpeg-turbo
+Version:        2.1.1
+Release:        3
+Summary:        MMX/SSE2/SIMD accelerated libjpeg-compatible JPEG codec library
+License:        IJG
+URL:            http://sourceforge.net/projects/libjpeg-turbo
+Source0:        http://downloads.sourceforge.net/libjpeg-turbo/libjpeg-turbo-%{version}.tar.gz
+
+BuildRequires:  gcc cmake libtool nasm
+
+Obsoletes:      libjpeg < 6b-47 turbojpeg < %{version}-%{release}  
+Provides:       libjpeg = 6b-47 turbojpeg = %{version}-%{release}  
+
+%description
+libjpeg-turbo is a JPEG image codec that uses SIMD instructions (MMX, SSE2, NEON, AltiVec)
+to accelerate baseline JPEG compression and decompression on x86, x86-64, and ARM systems.
+
+%package        devel
+Summary:        Development files for the libjpeg-turbo library
+Requires:       libjpeg-turbo = %{version}-%{release}
+Provides:       libjpeg-turbo-static = 1.3.1 turbojpeg-devel = %{version}-%{release}
+Obsoletes:      libjpeg-turbo-static < 1.3.1 turbojpeg-devel < %{version}-%{release}
+Provides:       libjpeg-devel = 6b-47 libjpeg-devel%{?_isa} = %{version}-%{release}
+Obsoletes:      libjpeg-devel < 6b-47
+
+
+%description devel
+Development files for the libjpeg-turbo library.
+
+%package utils
+Summary:        Utilities for manipulating JPEG images
+Requires:       libjpeg-turbo%{?_isa} = %{version}-%{release}
+ 
+%description utils
+The libjpeg-turbo-utils package contains simple client programs for accessing
+the libjpeg functions. It contains cjpeg, djpeg, jpegtran, rdjpgcom and
+wrjpgcom. Cjpeg compresses an image file into JPEG format. Djpeg decompresses a
+JPEG file into a regular image file. Jpegtran can perform various useful
+transformations on JPEG files. Rdjpgcom displays any text comments included in a
+JPEG file. Wrjpgcom inserts text comments into a JPEG file.
+
+%package -n turbojpeg
+Summary:        TurboJPEG library
+ 
+%description -n turbojpeg
+The turbojpeg package contains the TurboJPEG shared library.
+
+%package -n turbojpeg-devel
+Summary:        Headers for the TurboJPEG library
+Requires:       turbojpeg%{?_isa} = %{version}-%{release}
+ 
+%description -n turbojpeg-devel
+This package contains header files necessary for developing programs which will
+manipulate JPEG files using the TurboJPEG library.
+
+%package_help
+
+%prep
+%autosetup -n %{name}-%{version} -p1
+
+%build
+%{cmake} -DCMAKE_SKIP_RPATH:BOOL=YES -DCMAKE_SKIP_INSTALL_RPATH:BOOL=YES -DENABLE_STATIC:BOOL=NO .
+
+%make_build V=1
+
+%install
+%make_install
+%delete_la
+
+chmod -x README.md
+
+%ifarch x86_64
+  %global wordsize "64"
+%else
+  %global wordsize ""
+%endif
+
+if test -n "$wordsize"
+then
+  pushd $RPM_BUILD_ROOT%{_includedir}
+    mv jconfig.h jconfig-$wordsize.h
+    cat > jconfig.h <<EOF
+#ifndef JCONFIG_H_MULTILIB
+#define JCONFIG_H_MULTILIB
+
+#include <bits/wordsize.h>
+
+#if __WORDSIZE == 32
+# include "jconfig-32.h"
+#elif __WORDSIZE == 64
+# include "jconfig-64.h"
+#else
+# error "unexpected value for __WORDSIZE macro"
+#endif
+
+#endif
+EOF
+fi
+    
+%check
+LD_LIBRARY_PATH=%{buildroot}%{_libdir} make test %{?_smp_mflags}
+
+%ldconfig_scriptlets
+
+%files
+%defattr(-,root,root)
+%doc README.*
+%license LICENSE.md
+%{_libdir}/libjpeg.so.62*
+%exclude /usr/share/doc/libjpeg-turbo/*
+
+%files devel
+%doc coderules.txt jconfig.txt libjpeg.txt structure.txt example.txt 
+%exclude %{_includedir}/turbojpeg.h
+%{_includedir}/*.h
+%{_libdir}/libjpeg.so
+%{_libdir}/pkgconfig/libjpeg.pc
+%{_libdir}/cmake/%{name}/%{name}*.cmake
+
+%files utils
+%doc usage.txt wizard.txt
+%{_bindir}/cjpeg
+%{_bindir}/djpeg
+%{_bindir}/jpegtran
+%{_bindir}/rdjpgcom
+%{_bindir}/wrjpgcom
+%{_bindir}/tjbench
+
+%files -n turbojpeg
+%license LICENSE.md
+%{_libdir}/libturbojpeg.so.0*
+
+%files -n turbojpeg-devel
+%doc tjexample.c
+%{_includedir}/turbojpeg.h
+%{_libdir}/libturbojpeg.so
+%{_libdir}/pkgconfig/libturbojpeg.pc
+
+%files help
+%doc ChangeLog.md
+%{_mandir}/man1/*.1*
+
+%changelog
+* Wed Jul 13 2022 wangkerong <wangkerong@h-partners.com> - 2.1.1-3
+- utils provide by libjpeg-turbo-utils subpackage,the main package shouldn't provide it
+
+* Wed Mar 30 2022 liuyumeng <liuyumeng5@h-partners.com> - 2.1.1-2
+- enable tests
+
+* Sat Nov 27 2021 wangkerong <wangkerong@huawei.com> - 2.1.1-1
+- update to 2.1.1
+- splite utils turbojpeg turbojpeg-devel subpackages
+
+* Wed Jan 27 2021 hanhui <hanhui15@huawei.com> - 2.0.6-1
+- Type: enhancement
+- ID:   NA
+- SUG:  NA
+- DESC: update to 2.0.6
+
+* Wed Jul 22 2020 openEuler Buildteam <buildteam@openeuler.org> - 2.0.5-1
+- Type:enhancement
+- Id:NA
+- SUG:NA
+- DESC:upgrade to 2.0.5
+
+* Tue Jan 14 2020 openEuler Buildteam <buildteam@openeuler.org> - 2.0.0-4
+- Type:bugfix
+- Id:NA
+- SUG:NA
+- DESC:remove the libjpeg-devel with arch in obsoletes
+
+* Fri Nov 1 2019 openEuler Buildteam <buildteam@openeuler.org> - 2.0.0-3
+- Type:bugfix
+- Id:NA
+- SUG:NA
+- DESC:add the libjpeg-devel and jconfig.h
+
+* Sat Sep 21 2019 Lijin Yang <yanglijin@huawei.com> - 2.0.0-2
+- Package init

_service Changed

_service:extract_file:libjpeg-turbo-2.1.1.tar.gz Deleted

_service:extract_file:libjpeg-turbo.yaml Deleted

_service:tar_scm:libjpeg-turbo-1649227308.dea8c6d.tar/libjpeg-turbo-2.1.1.tar.gz Deleted

_service:tar_scm:libjpeg-turbo-1649227308.dea8c6d.tar/libjpeg-turbo.spec Deleted

_service:tar_scm:libjpeg-turbo-1649227308.dea8c6d.tar/libjpeg-turbo.yaml Deleted

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/BUILDING.md Added

@@ -0,0 +1,716 @@
+Building libjpeg-turbo
+======================
+
+
+Build Requirements
+------------------
+
+
+### All Systems
+
+- CMake(http://www.cmake.org) v2.8.12 or later
+
+- NASM(http://www.nasm.us) or YASM(http://yasm.tortall.net)
+  (if building x86 or x86-64 SIMD extensions)
+  * If using NASM, 2.13 or later is required.
+  * If using YASM, 1.2.0 or later is required.
+  * If building on macOS, NASM or YASM can be obtained from
+    MacPorts(http://www.macports.org/) or Homebrew(http://brew.sh/).
+     - NOTE: Currently, if it is desirable to hide the SIMD function symbols in
+       Mac executables or shared libraries that statically link with
+       libjpeg-turbo, then NASM 2.14 or later or YASM must be used when
+       building libjpeg-turbo.
+  * If building on Windows, **nasm.exe**/**yasm.exe** should be in your `PATH`.
+  * NASM and YASM are located in the CRB (Code Ready Builder) repository on
+    Red Hat Enterprise Linux 8 and in the PowerTools repository on CentOS 8,
+    which is not enabled by default.
+
+  The binary RPMs released by the NASM project do not work on older Linux
+  systems, such as Red Hat Enterprise Linux 5.  On such systems, you can easily
+  build and install NASM from a source RPM by downloading one of the SRPMs from
+
+  <http://www.nasm.us/pub/nasm/releasebuilds>
+
+  and executing the following as root:
+
+        ARCH=`uname -m`
+        rpmbuild --rebuild nasm-{version}.src.rpm
+        rpm -Uvh /usr/src/redhat/RPMS/$ARCH/nasm-{version}.$ARCH.rpm
+
+  NOTE: the NASM build will fail if texinfo is not installed.
+
+
+### Un*x Platforms (including Linux, Mac, FreeBSD, Solaris, and Cygwin)
+
+- GCC v4.1 (or later) or Clang recommended for best performance
+
+- If building the TurboJPEG Java wrapper, JDK or OpenJDK 1.5 or later is
+  required.  Most modern Linux distributions, as well as Solaris 10 and later,
+  include JDK or OpenJDK.  For other systems, you can obtain the Oracle Java
+  Development Kit from
+  <http://www.oracle.com/technetwork/java/javase/downloads>.
+
+  * If using JDK 11 or later, CMake 3.10.x or later must also be used.
+
+### Windows
+
+- Microsoft Visual C++ 2005 or later
+
+  If you don't already have Visual C++, then the easiest way to get it is by
+  installing
+  Visual Studio Community Edition(https://visualstudio.microsoft.com),
+  which includes everything necessary to build libjpeg-turbo.
+
+  * You can also download and install the standalone Windows SDK (for Windows 7
+    or later), which includes command-line versions of the 32-bit and 64-bit
+    Visual C++ compilers.
+  * If you intend to build libjpeg-turbo from the command line, then add the
+    appropriate compiler and SDK directories to the `INCLUDE`, `LIB`, and
+    `PATH` environment variables.  This is generally accomplished by
+    executing `vcvars32.bat` or `vcvars64.bat`, which are located in the same
+    directory as the compiler.
+  * If built with Visual C++ 2015 or later, the libjpeg-turbo static libraries
+    cannot be used with earlier versions of Visual C++, and vice versa.
+  * The libjpeg API DLL (**jpeg{version}.dll**) will depend on the C run-time
+    DLLs corresponding to the version of Visual C++ that was used to build it.
+
+   ... OR ...
+
+- MinGW
+
+  MSYS2(http://msys2.github.io/) or tdm-gcc(http://tdm-gcc.tdragon.net/)
+  recommended if building on a Windows machine.  Both distributions install a
+  Start Menu link that can be used to launch a command prompt with the
+  appropriate compiler paths automatically set.
+
+- If building the TurboJPEG Java wrapper, JDK 1.5 or later is required.  This
+  can be downloaded from
+  <http://www.oracle.com/technetwork/java/javase/downloads>.
+
+  * If using JDK 11 or later, CMake 3.10.x or later must also be used.
+
+
+Out-of-Tree Builds
+------------------
+
+Binary objects, libraries, and executables are generated in the directory from
+which CMake is executed (the "binary directory"), and this directory need not
+necessarily be the same as the libjpeg-turbo source directory.  You can create
+multiple independent binary directories, in which different versions of
+libjpeg-turbo can be built from the same source tree using different compilers
+or settings.  In the sections below, *{build_directory}* refers to the binary
+directory, whereas *{source_directory}* refers to the libjpeg-turbo source
+directory.  For in-tree builds, these directories are the same.
+
+
+Ninja
+-----
+
+In all of the procedures and recipes below, replace `make` with `ninja` and
+`Unix Makefiles` with `Ninja` if using Ninja.
+
+
+Build Procedure
+---------------
+
+NOTE: The build procedures below assume that CMake is invoked from the command
+line, but all of these procedures can be adapted to the CMake GUI as
+well.
+
+
+### Un*x
+
+The following procedure will build libjpeg-turbo on Unix and Unix-like systems.
+(On Solaris, this generates a 32-bit build.  See "Build Recipes" below for
+64-bit build instructions.)
+
+    cd {build_directory}
+    cmake -G"Unix Makefiles" additional CMake flags {source_directory}
+    make
+
+This will generate the following files under *{build_directory}*:
+
+**libjpeg.a**<br>
+Static link library for the libjpeg API
+
+**libjpeg.so.{version}** (Linux, Unix)<br>
+**libjpeg.{version}.dylib** (Mac)<br>
+**cygjpeg-{version}.dll** (Cygwin)<br>
+Shared library for the libjpeg API
+
+By default, *{version}* is 62.2.0, 7.2.0, or 8.1.2, depending on whether
+libjpeg v6b (default), v7, or v8 emulation is enabled.  If using Cygwin,
+*{version}* is 62, 7, or 8.
+
+**libjpeg.so** (Linux, Unix)<br>
+**libjpeg.dylib** (Mac)<br>
+Development symlink for the libjpeg API
+
+**libjpeg.dll.a** (Cygwin)<br>
+Import library for the libjpeg API
+
+**libturbojpeg.a**<br>
+Static link library for the TurboJPEG API
+
+**libturbojpeg.so.0.2.0** (Linux, Unix)<br>
+**libturbojpeg.0.2.0.dylib** (Mac)<br>
+**cygturbojpeg-0.dll** (Cygwin)<br>
+Shared library for the TurboJPEG API
+
+**libturbojpeg.so** (Linux, Unix)<br>
+**libturbojpeg.dylib** (Mac)<br>
+Development symlink for the TurboJPEG API
+
+**libturbojpeg.dll.a** (Cygwin)<br>
+Import library for the TurboJPEG API
+
+
+### Visual C++ (Command Line)
+
+    cd {build_directory}
+    cmake -G"NMake Makefiles" -DCMAKE_BUILD_TYPE=Release additional CMake flags {source_directory}
+    nmake
+
+This will build either a 32-bit or a 64-bit version of libjpeg-turbo, depending
+on which version of **cl.exe** is in the `PATH`.
+
+The following files will be generated under *{build_directory}*:
+
+**jpeg-static.lib**<br>
+Static link library for the libjpeg API
+
+**jpeg{version}.dll**<br>
+DLL for the libjpeg API
+
+**jpeg.lib**<br>
+Import library for the libjpeg API
+
+**turbojpeg-static.lib**<br>
+Static link library for the TurboJPEG API
+
+**turbojpeg.dll**<br>
+DLL for the TurboJPEG API
+
+**turbojpeg.lib**<br>
+Import library for the TurboJPEG API
+
+*{version}* is 62, 7, or 8, depending on whether libjpeg v6b (default), v7, or
+v8 emulation is enabled.
+
+
+### Visual C++ (IDE)
+
+Choose the appropriate CMake generator option for your version of Visual Studio
+(run `cmake` with no arguments for a list of available generators.)  For
+instance:
+
+    cd {build_directory}
+    cmake -G"Visual Studio 10" additional CMake flags {source_directory}
+
+NOTE: Add "Win64" to the generator name (for example, "Visual Studio 10 Win64")
+to build a 64-bit version of libjpeg-turbo.  A separate build directory must be
+used for 32-bit and 64-bit builds.
+
+You can then open **ALL_BUILD.vcproj** in Visual Studio and build one of the
+configurations in that project ("Debug", "Release", etc.) to generate a full
+build of libjpeg-turbo.
+
+This will generate the following files under *{build_directory}*:
+
+**{configuration}/jpeg-static.lib**<br>
+Static link library for the libjpeg API
+
+**{configuration}/jpeg{version}.dll**<br>
+DLL for the libjpeg API
+
+**{configuration}/jpeg.lib**<br>
+Import library for the libjpeg API
+
+**{configuration}/turbojpeg-static.lib**<br>
+Static link library for the TurboJPEG API
+
+**{configuration}/turbojpeg.dll**<br>
+DLL for the TurboJPEG API
+
+**{configuration}/turbojpeg.lib**<br>
+Import library for the TurboJPEG API
+
+*{configuration}* is Debug, Release, RelWithDebInfo, or MinSizeRel, depending
+on the configuration you built in the IDE, and *{version}* is 62, 7, or 8,
+depending on whether libjpeg v6b (default), v7, or v8 emulation is enabled.
+
+
+### MinGW
+
+NOTE: This assumes that you are building on a Windows machine using the MSYS
+environment.  If you are cross-compiling on a Un*x platform (including Mac and
+Cygwin), then see "Build Recipes" below.
+
+    cd {build_directory}
+    cmake -G"MSYS Makefiles" additional CMake flags {source_directory}
+    make
+
+This will generate the following files under *{build_directory}*:
+
+**libjpeg.a**<br>
+Static link library for the libjpeg API
+
+**libjpeg-{version}.dll**<br>
+DLL for the libjpeg API
+
+**libjpeg.dll.a**<br>
+Import library for the libjpeg API
+
+**libturbojpeg.a**<br>
+Static link library for the TurboJPEG API
+
+**libturbojpeg.dll**<br>
+DLL for the TurboJPEG API
+
+**libturbojpeg.dll.a**<br>
+Import library for the TurboJPEG API
+
+*{version}* is 62, 7, or 8, depending on whether libjpeg v6b (default), v7, or
+v8 emulation is enabled.
+
+
+### Debug Build
+
+Add `-DCMAKE_BUILD_TYPE=Debug` to the CMake command line.  Or, if building
+with NMake, remove `-DCMAKE_BUILD_TYPE=Release` (Debug builds are the default
+with NMake.)
+
+
+### libjpeg v7 or v8 API/ABI Emulation
+
+Add `-DWITH_JPEG7=1` to the CMake command line to build a version of
+libjpeg-turbo that is API/ABI-compatible with libjpeg v7.  Add `-DWITH_JPEG8=1`
+to the CMake command line to build a version of libjpeg-turbo that is
+API/ABI-compatible with libjpeg v8.  See README.md(README.md) for more
+information about libjpeg v7 and v8 emulation.
+
+
+### In-Memory Source/Destination Managers
+
+When using libjpeg v6b or v7 API/ABI emulation, add `-DWITH_MEM_SRCDST=0` to
+the CMake command line to build a version of libjpeg-turbo that lacks the
+`jpeg_mem_src()` and `jpeg_mem_dest()` functions.  These functions were not
+part of the original libjpeg v6b and v7 APIs, so removing them ensures strict
+conformance with those APIs.  See README.md(README.md) for more information.
+
+
+### Arithmetic Coding Support
+
+Since the patent on arithmetic coding has expired, this functionality has been
+included in this release of libjpeg-turbo.  libjpeg-turbo's implementation is
+based on the implementation in libjpeg v8, but it works when emulating libjpeg
+v7 or v6b as well.  The default is to enable both arithmetic encoding and
+decoding, but those who have philosophical objections to arithmetic coding can
+add `-DWITH_ARITH_ENC=0` or `-DWITH_ARITH_DEC=0` to the CMake command line to
+disable encoding or decoding (respectively.)
+
+
+### TurboJPEG Java Wrapper
+
+Add `-DWITH_JAVA=1` to the CMake command line to incorporate an optional Java
+Native Interface (JNI) wrapper into the TurboJPEG shared library and build the
+Java front-end classes to support it.  This allows the TurboJPEG shared library
+to be used directly from Java applications.  See java/README(java/README) for
+more details.
+
+If Java is not in your `PATH`, or if you wish to use an alternate JDK to
+build/test libjpeg-turbo, then (prior to running CMake) set the `JAVA_HOME`
+environment variable to the location of the JDK that you wish to use.  The
+`Java_JAVAC_EXECUTABLE`, `Java_JAVA_EXECUTABLE`, and `Java_JAR_EXECUTABLE`
+CMake variables can also be used to specify alternate commands or locations for
+javac, jar, and java (respectively.)  You can also set the
+`CMAKE_JAVA_COMPILE_FLAGS` CMake variable or the `JAVAFLAGS` environment
+variable to specify arguments that should be passed to the Java compiler when
+building the TurboJPEG classes, and the `JAVAARGS` CMake variable to specify
+arguments that should be passed to the JRE when running the TurboJPEG Java unit
+tests.
+
+
+Build Recipes
+-------------
+
+
+### 32-bit Build on 64-bit Linux/Unix
+
+Use export/setenv to set the following environment variables before running
+CMake:
+
+    CFLAGS=-m32
+    LDFLAGS=-m32
+
+
+### 64-bit Build on Solaris
+
+Use export/setenv to set the following environment variables before running
+CMake:
+
+    CFLAGS=-m64
+    LDFLAGS=-m64
+
+
+### Other Compilers
+
+On Un*x systems, prior to running CMake, you can set the `CC` environment
+variable to the command used to invoke the C compiler.
+
+
+### 32-bit MinGW Build on Un*x (including Mac and Cygwin)
+
+Create a file called **toolchain.cmake** under *{build_directory}*, with the
+following contents:
+
+    set(CMAKE_SYSTEM_NAME Windows)
+    set(CMAKE_SYSTEM_PROCESSOR X86)
+    set(CMAKE_C_COMPILER {mingw_binary_path}/i686-w64-mingw32-gcc)
+    set(CMAKE_RC_COMPILER {mingw_binary_path}/i686-w64-mingw32-windres)
+
+*{mingw\_binary\_path}* is the directory under which the MinGW binaries are
+located (usually **/usr/bin**.)  Next, execute the following commands:
+
+    cd {build_directory}
+    cmake -G"Unix Makefiles" -DCMAKE_TOOLCHAIN_FILE=toolchain.cmake \
+      additional CMake flags {source_directory}
+    make
+
+
+### 64-bit MinGW Build on Un*x (including Mac and Cygwin)
+
+Create a file called **toolchain.cmake** under *{build_directory}*, with the
+following contents:
+
+    set(CMAKE_SYSTEM_NAME Windows)
+    set(CMAKE_SYSTEM_PROCESSOR AMD64)
+    set(CMAKE_C_COMPILER {mingw_binary_path}/x86_64-w64-mingw32-gcc)
+    set(CMAKE_RC_COMPILER {mingw_binary_path}/x86_64-w64-mingw32-windres)
+
+*{mingw\_binary\_path}* is the directory under which the MinGW binaries are
+located (usually **/usr/bin**.)  Next, execute the following commands:
+
+    cd {build_directory}
+    cmake -G"Unix Makefiles" -DCMAKE_TOOLCHAIN_FILE=toolchain.cmake \
+      additional CMake flags {source_directory}
+    make
+
+
+Building libjpeg-turbo for iOS
+------------------------------
+
+iOS platforms, such as the iPhone and iPad, use Arm processors, and all
+currently supported models include Neon instructions.  Thus, they can take
+advantage of libjpeg-turbo's SIMD extensions to significantly accelerate JPEG
+compression/decompression.  This section describes how to build libjpeg-turbo
+for these platforms.
+
+
+### Armv8 (64-bit)
+
+**Xcode 5 or later required, Xcode 6.3.x or later recommended**
+
+The following script demonstrates how to build libjpeg-turbo to run on the
+iPhone 5S/iPad Mini 2/iPad Air and newer.
+
+    IOS_PLATFORMDIR=/Applications/Xcode.app/Contents/Developer/Platforms/iPhoneOS.platform
+    IOS_SYSROOT=($IOS_PLATFORMDIR/Developer/SDKs/iPhoneOS*.sdk)
+    export CFLAGS="-Wall -arch arm64 -miphoneos-version-min=8.0 -funwind-tables"
+
+    cd {build_directory}
+
+    cat <<EOF >toolchain.cmake
+    set(CMAKE_SYSTEM_NAME Darwin)
+    set(CMAKE_SYSTEM_PROCESSOR aarch64)
+    set(CMAKE_C_COMPILER /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang)
+    EOF
+
+    cmake -G"Unix Makefiles" -DCMAKE_TOOLCHAIN_FILE=toolchain.cmake \
+      -DCMAKE_OSX_SYSROOT=${IOS_SYSROOT0} \
+      additional CMake flags {source_directory}
+    make
+
+
+Building libjpeg-turbo for Android
+----------------------------------
+
+Building libjpeg-turbo for Android platforms requires v13b or later of the
+Android NDK(https://developer.android.com/tools/sdk/ndk).
+
+
+### Armv7 (32-bit)
+
+**NDK r19 or later with Clang recommended**
+
+The following is a general recipe script that can be modified for your specific
+needs.
+
+    # Set these variables to suit your needs
+    NDK_PATH={full path to the NDK directory-- for example,
+      /opt/android/android-ndk-r16b}
+    TOOLCHAIN={"gcc" or "clang"-- "gcc" must be used with NDK r16b and earlier,
+      and "clang" must be used with NDK r17c and later}
+    ANDROID_VERSION={the minimum version of Android to support-- for example,
+      "16", "19", etc.}
+
+    cd {build_directory}
+    cmake -G"Unix Makefiles" \
+      -DANDROID_ABI=armeabi-v7a \
+      -DANDROID_ARM_MODE=arm \
+      -DANDROID_PLATFORM=android-${ANDROID_VERSION} \
+      -DANDROID_TOOLCHAIN=${TOOLCHAIN} \
+      -DCMAKE_ASM_FLAGS="--target=arm-linux-androideabi${ANDROID_VERSION}" \
+      -DCMAKE_TOOLCHAIN_FILE=${NDK_PATH}/build/cmake/android.toolchain.cmake \
+      additional CMake flags {source_directory}
+    make
+
+
+### Armv8 (64-bit)
+
+**Clang recommended**
+
+The following is a general recipe script that can be modified for your specific
+needs.
+
+    # Set these variables to suit your needs
+    NDK_PATH={full path to the NDK directory-- for example,
+      /opt/android/android-ndk-r16b}
+    TOOLCHAIN={"gcc" or "clang"-- "gcc" must be used with NDK r14b and earlier,
+      and "clang" must be used with NDK r17c and later}
+    ANDROID_VERSION={the minimum version of Android to support.  "21" or later
+      is required for a 64-bit build.}
+
+    cd {build_directory}
+    cmake -G"Unix Makefiles" \
+      -DANDROID_ABI=arm64-v8a \
+      -DANDROID_ARM_MODE=arm \
+      -DANDROID_PLATFORM=android-${ANDROID_VERSION} \
+      -DANDROID_TOOLCHAIN=${TOOLCHAIN} \
+      -DCMAKE_ASM_FLAGS="--target=aarch64-linux-android${ANDROID_VERSION}" \
+      -DCMAKE_TOOLCHAIN_FILE=${NDK_PATH}/build/cmake/android.toolchain.cmake \
+      additional CMake flags {source_directory}
+    make
+
+
+### x86 (32-bit)
+
+The following is a general recipe script that can be modified for your specific
+needs.
+
+    # Set these variables to suit your needs
+    NDK_PATH={full path to the NDK directory-- for example,
+      /opt/android/android-ndk-r16b}
+    TOOLCHAIN={"gcc" or "clang"-- "gcc" must be used with NDK r14b and earlier,
+      and "clang" must be used with NDK r17c and later}
+    ANDROID_VERSION={The minimum version of Android to support-- for example,
+      "16", "19", etc.}
+
+    cd {build_directory}
+    cmake -G"Unix Makefiles" \
+      -DANDROID_ABI=x86 \
+      -DANDROID_PLATFORM=android-${ANDROID_VERSION} \
+      -DANDROID_TOOLCHAIN=${TOOLCHAIN} \
+      -DCMAKE_TOOLCHAIN_FILE=${NDK_PATH}/build/cmake/android.toolchain.cmake \
+      additional CMake flags {source_directory}
+    make
+
+
+### x86-64 (64-bit)
+
+The following is a general recipe script that can be modified for your specific
+needs.
+
+    # Set these variables to suit your needs
+    NDK_PATH={full path to the NDK directory-- for example,
+      /opt/android/android-ndk-r16b}
+    TOOLCHAIN={"gcc" or "clang"-- "gcc" must be used with NDK r14b and earlier,
+      and "clang" must be used with NDK r17c and later}
+    ANDROID_VERSION={the minimum version of Android to support.  "21" or later
+      is required for a 64-bit build.}
+
+    cd {build_directory}
+    cmake -G"Unix Makefiles" \
+      -DANDROID_ABI=x86_64 \
+      -DANDROID_PLATFORM=android-${ANDROID_VERSION} \
+      -DANDROID_TOOLCHAIN=${TOOLCHAIN} \
+      -DCMAKE_TOOLCHAIN_FILE=${NDK_PATH}/build/cmake/android.toolchain.cmake \
+      additional CMake flags {source_directory}
+    make
+
+
+Advanced CMake Options
+----------------------
+
+To list and configure other CMake options not specifically mentioned in this
+guide, run
+
+    ccmake {source_directory}
+
+or
+
+    cmake-gui {source_directory}
+
+from the build directory after initially configuring the build.  CCMake is a
+text-based interactive version of CMake, and CMake-GUI is a GUI version.  Both
+will display all variables that are relevant to the libjpeg-turbo build, their
+current values, and a help string describing what they do.
+
+
+Installing libjpeg-turbo
+========================
+
+You can use the build system to install libjpeg-turbo (as opposed to creating
+an installer package.)  To do this, run `make install` or `nmake install`
+(or build the "install" target in the Visual Studio IDE.)  Running
+`make uninstall` or `nmake uninstall` (or building the "uninstall" target in
+the Visual Studio IDE) will uninstall libjpeg-turbo.
+
+The `CMAKE_INSTALL_PREFIX` CMake variable can be modified in order to install
+libjpeg-turbo into a directory of your choosing.  If you don't specify
+`CMAKE_INSTALL_PREFIX`, then the default is:
+
+**c:\libjpeg-turbo**<br>
+Visual Studio 32-bit build
+
+**c:\libjpeg-turbo64**<br>
+Visual Studio 64-bit build
+
+**c:\libjpeg-turbo-gcc**<br>
+MinGW 32-bit build
+
+**c:\libjpeg-turbo-gcc64**<br>
+MinGW 64-bit build
+
+**/opt/libjpeg-turbo**<br>
+Un*x
+
+The default value of `CMAKE_INSTALL_PREFIX` causes the libjpeg-turbo files to
+be installed with a directory structure resembling that of the official
+libjpeg-turbo binary packages.  Changing the value of `CMAKE_INSTALL_PREFIX`
+(for instance, to **/usr/local**) causes the libjpeg-turbo files to be
+installed with a directory structure that conforms to GNU standards.
+
+The `CMAKE_INSTALL_BINDIR`, `CMAKE_INSTALL_DATAROOTDIR`,
+`CMAKE_INSTALL_DOCDIR`, `CMAKE_INSTALL_INCLUDEDIR`, `CMAKE_INSTALL_JAVADIR`,
+`CMAKE_INSTALL_LIBDIR`, and `CMAKE_INSTALL_MANDIR` CMake variables allow a
+finer degree of control over where specific files in the libjpeg-turbo
+distribution should be installed.  These directory variables can either be
+specified as absolute paths or as paths relative to `CMAKE_INSTALL_PREFIX` (for
+instance, setting `CMAKE_INSTALL_DOCDIR` to **doc** would cause the
+documentation to be installed in **${CMAKE\_INSTALL\_PREFIX}/doc**.)  If a
+directory variable contains the name of another directory variable in angle
+brackets, then its final value will depend on the final value of that other
+variable.  For instance, the default value of `CMAKE_INSTALL_MANDIR` is
+**\<CMAKE\_INSTALL\_DATAROOTDIR\>/man**.
+
+NOTE: If setting one of these directory variables to a relative path using the
+CMake command line, you must specify that the variable is of type `PATH`.
+For example:
+
+    cmake -G"{generator type}" -DCMAKE_INSTALL_LIBDIR:PATH=lib {source_directory}
+
+Otherwise, CMake will assume that the path is relative to the build directory
+rather than the install directory.
+
+
+Creating Distribution Packages
+==============================
+
+The following commands can be used to create various types of distribution
+packages:
+
+
+Linux
+-----
+
+    make rpm
+
+Create Red Hat-style binary RPM package.  Requires RPM v4 or later.
+
+    make srpm
+
+This runs `make dist` to create a pristine source tarball, then creates a
+Red Hat-style source RPM package from the tarball.  Requires RPM v4 or later.
+
+    make deb
+
+Create Debian-style binary package.  Requires dpkg.
+
+
+Mac
+---
+
+    make dmg
+
+Create Mac package/disk image.  This requires pkgbuild and productbuild, which
+are installed by default on OS X/macOS 10.7 and later.
+
+In order to create a Mac package/disk image that contains universal
+x86-64/Arm binaries, set the following CMake variable:
+
+* `ARMV8_BUILD`: Directory containing an Armv8 (64-bit) iOS or macOS build of
+  libjpeg-turbo to include in the universal binaries
+
+You should first use CMake to configure an Armv8 sub-build of libjpeg-turbo
+(see "Building libjpeg-turbo for iOS" above, if applicable) in a build
+directory that matches the one specified in the aforementioned CMake variable.
+Next, configure the primary (x86-64) build of libjpeg-turbo as an out-of-tree
+build, specifying the aforementioned CMake variable, and build it.  Once the
+primary build has been built, run `make dmg` from the build directory.  The
+packaging system will build the sub-build, use lipo to combine it with the
+primary build into a single set of universal binaries, then package the
+universal binaries.
+
+
+Windows
+-------
+
+If using NMake:
+
+    cd {build_directory}
+    nmake installer
+
+If using MinGW:
+
+    cd {build_directory}
+    make installer
+
+If using the Visual Studio IDE, build the "installer" target.
+
+The installer package (libjpeg-turbo-*{version}*-gcc|-vc64.exe) will be
+located under *{build_directory}*.  If building using the Visual Studio IDE,
+then the installer package will be located in a subdirectory with the same name
+as the configuration you built (such as *{build_directory}*\Debug\ or
+*{build_directory}*\Release\).
+
+Building a Windows installer requires the
+Nullsoft Install System(http://nsis.sourceforge.net/).  makensis.exe should
+be in your `PATH`.
+
+
+Regression testing
+==================
+
+The most common way to test libjpeg-turbo is by invoking `make test` (Un*x) or
+`nmake test` (Windows command line) or by building the "RUN_TESTS" target
+(Visual Studio IDE), once the build has completed.  This runs a series of tests
+to ensure that mathematical compatibility has been maintained between
+libjpeg-turbo and libjpeg v6b.  This also invokes the TurboJPEG unit tests,
+which ensure that the colorspace extensions, YUV encoding, decompression
+scaling, and other features of the TurboJPEG C and Java APIs are working
+properly (and, by extension, that the equivalent features of the underlying
+libjpeg API are also working.)
+
+Invoking `make testclean` (Un*x) or `nmake testclean` (Windows command line) or
+building the "testclean" target (Visual Studio IDE) will clean up the output
+images generated by the tests.
+
+On Un*x platforms, more extensive tests of the TurboJPEG C and Java wrappers
+can be run by invoking `make tjtest`.  These extended TurboJPEG tests
+essentially iterate through all of the available features of the TurboJPEG APIs
+that are not covered by the TurboJPEG unit tests (including the lossless
+transform options) and compare the images generated by each feature to images
+generated using the equivalent feature in the libjpeg API.  The extended
+TurboJPEG tests are meant to test for regressions in the TurboJPEG wrappers,
+not in the underlying libjpeg API library.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/CMakeLists.txt Added

@@ -0,0 +1,1536 @@
+cmake_minimum_required(VERSION 2.8.12)
+
+if(CMAKE_EXECUTABLE_SUFFIX)
+  set(CMAKE_EXECUTABLE_SUFFIX_TMP ${CMAKE_EXECUTABLE_SUFFIX})
+endif()
+
+project(libjpeg-turbo C)
+set(VERSION 2.1.1)
+string(REPLACE "." ";" VERSION_TRIPLET ${VERSION})
+list(GET VERSION_TRIPLET 0 VERSION_MAJOR)
+list(GET VERSION_TRIPLET 1 VERSION_MINOR)
+list(GET VERSION_TRIPLET 2 VERSION_REVISION)
+function(pad_number NUMBER OUTPUT_LEN)
+  string(LENGTH "${${NUMBER}}" INPUT_LEN)
+  if(INPUT_LEN LESS OUTPUT_LEN)
+    math(EXPR ZEROES "${OUTPUT_LEN} - ${INPUT_LEN} - 1")
+    set(NUM ${${NUMBER}})
+    foreach(C RANGE ${ZEROES})
+      set(NUM "0${NUM}")
+    endforeach()
+    set(${NUMBER} ${NUM} PARENT_SCOPE)
+  endif()
+endfunction()
+pad_number(VERSION_MINOR 3)
+pad_number(VERSION_REVISION 3)
+set(LIBJPEG_TURBO_VERSION_NUMBER ${VERSION_MAJOR}${VERSION_MINOR}${VERSION_REVISION})
+
+# CMake 3.14 and later sets CMAKE_MACOSX_BUNDLE to TRUE by default when
+# CMAKE_SYSTEM_NAME is iOS, tvOS, or watchOS, which breaks the libjpeg-turbo
+# build.  (Specifically, when CMAKE_MACOSX_BUNDLE is TRUE, executables for
+# Apple platforms are built as application bundles, which causes CMake to
+# complain that our install() directives for executables do not specify a
+# BUNDLE DESTINATION.  Even if CMake did not complain, building executables as
+# application bundles would break our iOS packages.)
+set(CMAKE_MACOSX_BUNDLE FALSE)
+
+string(TIMESTAMP DEFAULT_BUILD "%Y%m%d")
+set(BUILD ${DEFAULT_BUILD} CACHE STRING "Build string (default: ${DEFAULT_BUILD})")
+
+# NOTE: On Windows, this does nothing except when using MinGW or Cygwin.
+# CMAKE_BUILD_TYPE has no meaning in Visual Studio, and it always defaults to
+# Debug when using NMake.
+if(NOT CMAKE_BUILD_TYPE)
+  set(CMAKE_BUILD_TYPE Release)
+endif()
+message(STATUS "CMAKE_BUILD_TYPE = ${CMAKE_BUILD_TYPE}")
+
+message(STATUS "VERSION = ${VERSION}, BUILD = ${BUILD}")
+
+# Detect CPU type and whether we're building 64-bit or 32-bit code
+math(EXPR BITS "${CMAKE_SIZEOF_VOID_P} * 8")
+string(TOLOWER ${CMAKE_SYSTEM_PROCESSOR} CMAKE_SYSTEM_PROCESSOR_LC)
+set(COUNT 1)
+foreach(ARCH ${CMAKE_OSX_ARCHITECTURES})
+  if(COUNT GREATER 1)
+    message(FATAL_ERROR "The libjpeg-turbo build system does not support multiple values in CMAKE_OSX_ARCHITECTURES.")
+  endif()
+  math(EXPR COUNT "${COUNT}+1")
+endforeach()
+if(CMAKE_SYSTEM_PROCESSOR_LC MATCHES "x86_64" OR
+  CMAKE_SYSTEM_PROCESSOR_LC MATCHES "amd64" OR
+  CMAKE_SYSTEM_PROCESSOR_LC MATCHES "i0-986" OR
+  CMAKE_SYSTEM_PROCESSOR_LC MATCHES "x86" OR
+  CMAKE_SYSTEM_PROCESSOR_LC MATCHES "ia32")
+  if(BITS EQUAL 64 OR CMAKE_C_COMPILER_ABI MATCHES "ELF X32")
+    set(CPU_TYPE x86_64)
+  else()
+    set(CPU_TYPE i386)
+  endif()
+  if(NOT CMAKE_SYSTEM_PROCESSOR STREQUAL ${CPU_TYPE})
+    set(CMAKE_SYSTEM_PROCESSOR ${CPU_TYPE})
+  endif()
+elseif(CMAKE_SYSTEM_PROCESSOR_LC STREQUAL "aarch64" OR
+  CMAKE_SYSTEM_PROCESSOR_LC MATCHES "^arm")
+  if(BITS EQUAL 64)
+    set(CPU_TYPE arm64)
+  else()
+    set(CPU_TYPE arm)
+  endif()
+elseif(CMAKE_SYSTEM_PROCESSOR_LC MATCHES "^ppc" OR
+  CMAKE_SYSTEM_PROCESSOR_LC MATCHES "^powerpc")
+  set(CPU_TYPE powerpc)
+else()
+  set(CPU_TYPE ${CMAKE_SYSTEM_PROCESSOR_LC})
+endif()
+if(CMAKE_OSX_ARCHITECTURES MATCHES "x86_64" OR
+  CMAKE_OSX_ARCHITECTURES MATCHES "arm64" OR
+  CMAKE_OSX_ARCHITECTURES MATCHES "i386")
+  set(CPU_TYPE ${CMAKE_OSX_ARCHITECTURES})
+endif()
+if(CMAKE_OSX_ARCHITECTURES MATCHES "ppc")
+  set(CPU_TYPE powerpc)
+endif()
+if(MSVC_IDE AND CMAKE_GENERATOR_PLATFORM MATCHES "arm64")
+  set(CPU_TYPE arm64)
+endif()
+
+message(STATUS "${BITS}-bit build (${CPU_TYPE})")
+
+
+###############################################################################
+# INSTALL DIRECTORIES
+###############################################################################
+
+if(WIN32)
+  if(MSVC)
+    set(CMAKE_INSTALL_DEFAULT_PREFIX "c:/${CMAKE_PROJECT_NAME}")
+  else()
+    set(CMAKE_INSTALL_DEFAULT_PREFIX "c:/${CMAKE_PROJECT_NAME}-gcc")
+  endif()
+  if(BITS EQUAL 64)
+    set(CMAKE_INSTALL_DEFAULT_PREFIX "${CMAKE_INSTALL_DEFAULT_PREFIX}64")
+  endif()
+else()
+  if(NOT CMAKE_INSTALL_DEFAULT_PREFIX)
+    set(CMAKE_INSTALL_DEFAULT_PREFIX /opt/${CMAKE_PROJECT_NAME})
+  endif()
+endif()
+if(CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT)
+  set(CMAKE_INSTALL_PREFIX "${CMAKE_INSTALL_DEFAULT_PREFIX}" CACHE PATH
+    "Directory into which to install ${CMAKE_PROJECT_NAME} (default: ${CMAKE_INSTALL_DEFAULT_PREFIX})"
+    FORCE)
+endif()
+message(STATUS "CMAKE_INSTALL_PREFIX = ${CMAKE_INSTALL_PREFIX}")
+
+# When the prefix is /opt/${CMAKE_PROJECT_NAME}, we assume that an "official"
+# build is being created, and thus we install things into specific locations.
+
+if(CMAKE_INSTALL_PREFIX STREQUAL "${CMAKE_INSTALL_DEFAULT_PREFIX}")
+  set(CMAKE_INSTALL_DEFAULT_DATAROOTDIR "")
+  set(CMAKE_INSTALL_DEFAULT_DOCDIR "<CMAKE_INSTALL_DATAROOTDIR>/doc")
+  set(CMAKE_INSTALL_DEFAULT_JAVADIR "<CMAKE_INSTALL_DATAROOTDIR>/classes")
+  if(UNIX AND NOT APPLE)
+    if(BITS EQUAL 64)
+      set(CMAKE_INSTALL_DEFAULT_LIBDIR "lib64")
+    elseif(CMAKE_C_COMPILER_ABI MATCHES "ELF X32")
+      set(CMAKE_INSTALL_DEFAULT_LIBDIR "libx32")
+    else()
+      set(CMAKE_INSTALL_DEFAULT_LIBDIR "lib32")
+    endif()
+  endif()
+endif()
+
+include(cmakescripts/GNUInstallDirs.cmake)
+
+macro(report_directory var)
+  if(CMAKE_INSTALL_${var} STREQUAL CMAKE_INSTALL_FULL_${var})
+    message(STATUS "CMAKE_INSTALL_${var} = ${CMAKE_INSTALL_${var}}")
+  else()
+    message(STATUS "CMAKE_INSTALL_${var} = ${CMAKE_INSTALL_${var}} (${CMAKE_INSTALL_FULL_${var}})")
+  endif()
+  mark_as_advanced(CLEAR CMAKE_INSTALL_${var})
+endmacro()
+
+set(DIRLIST "BINDIR;DATAROOTDIR;DOCDIR;INCLUDEDIR;LIBDIR")
+if(UNIX)
+  list(APPEND DIRLIST "MANDIR")
+endif()
+foreach(dir ${DIRLIST})
+  report_directory(${dir})
+endforeach()
+
+
+###############################################################################
+# CONFIGURATION OPTIONS
+###############################################################################
+
+macro(boolean_number var)
+  if(${var})
+    set(${var} 1 ${ARGN})
+  else()
+    set(${var} 0 ${ARGN})
+  endif()
+endmacro()
+
+option(ENABLE_SHARED "Build shared libraries" TRUE)
+boolean_number(ENABLE_SHARED)
+option(ENABLE_STATIC "Build static libraries" TRUE)
+boolean_number(ENABLE_STATIC)
+option(REQUIRE_SIMD "Generate a fatal error if SIMD extensions are not available for this platform (default is to fall back to a non-SIMD build)" FALSE)
+boolean_number(REQUIRE_SIMD)
+option(WITH_12BIT "Encode/decode JPEG images with 12-bit samples (implies WITH_ARITH_DEC=0 WITH_ARITH_ENC=0 WITH_JAVA=0 WITH_SIMD=0 WITH_TURBOJPEG=0 )" FALSE)
+boolean_number(WITH_12BIT)
+option(WITH_ARITH_DEC "Include arithmetic decoding support when emulating the libjpeg v6b API/ABI" TRUE)
+boolean_number(WITH_ARITH_DEC)
+option(WITH_ARITH_ENC "Include arithmetic encoding support when emulating the libjpeg v6b API/ABI" TRUE)
+boolean_number(WITH_ARITH_ENC)
+if(CMAKE_C_COMPILER_ABI MATCHES "ELF X32")
+  set(WITH_JAVA 0)
+else()
+  option(WITH_JAVA "Build Java wrapper for the TurboJPEG API library (implies ENABLE_SHARED=1)" FALSE)
+  boolean_number(WITH_JAVA)
+endif()
+option(WITH_JPEG7 "Emulate libjpeg v7 API/ABI (this makes ${CMAKE_PROJECT_NAME} backward-incompatible with libjpeg v6b)" FALSE)
+boolean_number(WITH_JPEG7)
+option(WITH_JPEG8 "Emulate libjpeg v8 API/ABI (this makes ${CMAKE_PROJECT_NAME} backward-incompatible with libjpeg v6b)" FALSE)
+boolean_number(WITH_JPEG8)
+option(WITH_MEM_SRCDST "Include in-memory source/destination manager functions when emulating the libjpeg v6b or v7 API/ABI" TRUE)
+boolean_number(WITH_MEM_SRCDST)
+option(WITH_SIMD "Include SIMD extensions, if available for this platform" TRUE)
+boolean_number(WITH_SIMD)
+option(WITH_TURBOJPEG "Include the TurboJPEG API library and associated test programs" TRUE)
+boolean_number(WITH_TURBOJPEG)
+option(WITH_FUZZ "Build fuzz targets" FALSE)
+
+macro(report_option var desc)
+  if(${var})
+    message(STATUS "${desc} enabled (${var} = ${${var}})")
+  else()
+    message(STATUS "${desc} disabled (${var} = ${${var}})")
+  endif()
+endmacro()
+
+if(WITH_JAVA)
+  set(ENABLE_SHARED 1)
+endif()
+
+# Explicitly setting CMAKE_POSITION_INDEPENDENT_CODE=FALSE disables PIC for all
+# targets, which will cause the shared library builds to fail.  Thus, if shared
+# libraries are enabled and CMAKE_POSITION_INDEPENDENT_CODE is explicitly set
+# to FALSE, we need to unset it, thus restoring the default behavior
+# (automatically using PIC for shared library targets.)
+if(DEFINED CMAKE_POSITION_INDEPENDENT_CODE AND
+  NOT CMAKE_POSITION_INDEPENDENT_CODE AND ENABLE_SHARED)
+  unset(CMAKE_POSITION_INDEPENDENT_CODE CACHE)
+endif()
+
+report_option(ENABLE_SHARED "Shared libraries")
+report_option(ENABLE_STATIC "Static libraries")
+
+if(ENABLE_SHARED)
+  set(CMAKE_INSTALL_RPATH ${CMAKE_INSTALL_FULL_LIBDIR})
+endif()
+
+if(WITH_JPEG8 OR WITH_JPEG7)
+  set(WITH_ARITH_ENC 1)
+  set(WITH_ARITH_DEC 1)
+endif()
+if(WITH_JPEG8)
+  set(WITH_MEM_SRCDST 0)
+endif()
+
+if(WITH_12BIT)
+  set(WITH_ARITH_DEC 0)
+  set(WITH_ARITH_ENC 0)
+  set(WITH_JAVA 0)
+  set(WITH_SIMD 0)
+  set(WITH_TURBOJPEG 0)
+  set(BITS_IN_JSAMPLE 12)
+else()
+  set(BITS_IN_JSAMPLE 8)
+endif()
+report_option(WITH_12BIT "12-bit JPEG support")
+
+if(WITH_ARITH_DEC)
+  set(D_ARITH_CODING_SUPPORTED 1)
+endif()
+if(NOT WITH_12BIT)
+  report_option(WITH_ARITH_DEC "Arithmetic decoding support")
+endif()
+
+if(WITH_ARITH_ENC)
+  set(C_ARITH_CODING_SUPPORTED 1)
+endif()
+if(NOT WITH_12BIT)
+  report_option(WITH_ARITH_ENC "Arithmetic encoding support")
+endif()
+
+if(NOT WITH_12BIT)
+  report_option(WITH_TURBOJPEG "TurboJPEG API library")
+  report_option(WITH_JAVA "TurboJPEG Java wrapper")
+endif()
+
+if(WITH_MEM_SRCDST)
+  set(MEM_SRCDST_SUPPORTED 1)
+  set(MEM_SRCDST_FUNCTIONS "global:  jpeg_mem_dest;  jpeg_mem_src;")
+endif()
+if(NOT WITH_JPEG8)
+  report_option(WITH_MEM_SRCDST "In-memory source/destination managers")
+endif()
+
+set(SO_AGE 2)
+if(WITH_MEM_SRCDST)
+  set(SO_AGE 3)
+endif()
+
+if(WITH_JPEG8)
+  set(JPEG_LIB_VERSION 80)
+elseif(WITH_JPEG7)
+  set(JPEG_LIB_VERSION 70)
+else()
+  set(JPEG_LIB_VERSION 62)
+endif()
+
+math(EXPR JPEG_LIB_VERSION_DIV10 "${JPEG_LIB_VERSION} / 10")
+math(EXPR JPEG_LIB_VERSION_MOD10 "${JPEG_LIB_VERSION} % 10")
+if(JPEG_LIB_VERSION STREQUAL "62")
+  set(DEFAULT_SO_MAJOR_VERSION ${JPEG_LIB_VERSION})
+else()
+  set(DEFAULT_SO_MAJOR_VERSION ${JPEG_LIB_VERSION_DIV10})
+endif()
+if(JPEG_LIB_VERSION STREQUAL "80")
+  set(DEFAULT_SO_MINOR_VERSION 2)
+else()
+  set(DEFAULT_SO_MINOR_VERSION 0)
+endif()
+
+# This causes SO_MAJOR_VERSION/SO_MINOR_VERSION to reset to defaults if
+# WITH_JPEG7 or WITH_JPEG8 has changed.
+if((DEFINED WITH_JPEG7_INT AND NOT WITH_JPEG7 EQUAL WITH_JPEG7_INT) OR
+  (DEFINED WITH_JPEG8_INT AND NOT WITH_JPEG8 EQUAL WITH_JPEG8_INT))
+  set(FORCE_SO_VERSION "FORCE")
+endif()
+set(WITH_JPEG7_INT ${WITH_JPEG7} CACHE INTERNAL "")
+set(WITH_JPEG8_INT ${WITH_JPEG8} CACHE INTERNAL "")
+
+set(SO_MAJOR_VERSION ${DEFAULT_SO_MAJOR_VERSION} CACHE STRING
+  "Major version of the libjpeg API shared library (default: ${DEFAULT_SO_MAJOR_VERSION})"
+  ${FORCE_SO_VERSION})
+set(SO_MINOR_VERSION ${DEFAULT_SO_MINOR_VERSION} CACHE STRING
+  "Minor version of the libjpeg API shared library (default: ${DEFAULT_SO_MINOR_VERSION})"
+  ${FORCE_SO_VERSION})
+
+set(JPEG_LIB_VERSION_DECIMAL "${JPEG_LIB_VERSION_DIV10}.${JPEG_LIB_VERSION_MOD10}")
+message(STATUS "Emulating libjpeg API/ABI v${JPEG_LIB_VERSION_DECIMAL} (WITH_JPEG7 = ${WITH_JPEG7}, WITH_JPEG8 = ${WITH_JPEG8})")
+message(STATUS "libjpeg API shared library version = ${SO_MAJOR_VERSION}.${SO_AGE}.${SO_MINOR_VERSION}")
+
+# Because the TurboJPEG API library uses versioned symbols and changes the
+# names of functions whenever they are modified in a backward-incompatible
+# manner, it is always backward-ABI-compatible with itself, so the major and
+# minor SO versions don't change.  However, we increase the middle number (the
+# SO "age") whenever functions are added to the API.
+set(TURBOJPEG_SO_MAJOR_VERSION 0)
+set(TURBOJPEG_SO_VERSION 0.2.0)
+
+
+###############################################################################
+# COMPILER SETTINGS
+###############################################################################
+
+if(MSVC)
+  option(WITH_CRT_DLL
+    "Link all ${CMAKE_PROJECT_NAME} libraries and executables with the C run-time DLL (msvcr*.dll) instead of the static C run-time library (libcmt*.lib.)  The default is to use the C run-time DLL only with the libraries and executables that need it."
+    FALSE)
+  if(NOT WITH_CRT_DLL)
+    # Use the static C library for all build types
+    foreach(var CMAKE_C_FLAGS CMAKE_C_FLAGS_DEBUG CMAKE_C_FLAGS_RELEASE
+      CMAKE_C_FLAGS_MINSIZEREL CMAKE_C_FLAGS_RELWITHDEBINFO)
+      if(${var} MATCHES "/MD")
+        string(REGEX REPLACE "/MD" "/MT" ${var} "${${var}}")
+      endif()
+    endforeach()
+  endif()
+  set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} /W3 /wd4996")
+endif()
+
+if(CMAKE_COMPILER_IS_GNUCC OR CMAKE_C_COMPILER_ID STREQUAL "Clang")
+  # Use the maximum optimization level for release builds
+  foreach(var CMAKE_C_FLAGS_RELEASE CMAKE_C_FLAGS_RELWITHDEBINFO)
+    if(${var} MATCHES "-O2")
+      string(REGEX REPLACE "-O2" "-O3" ${var} "${${var}}")
+    endif()
+  endforeach()
+endif()
+
+if(CMAKE_SYSTEM_NAME STREQUAL "SunOS")
+  if(CMAKE_C_COMPILER_ID MATCHES "SunPro")
+    # Use the maximum optimization level for release builds
+    foreach(var CMAKE_C_FLAGS_RELEASE CMAKE_C_FLAGS_RELWITHDEBINFO)
+      if(${var} MATCHES "-xO3")
+        string(REGEX REPLACE "-xO3" "-xO5" ${var} "${${var}}")
+      endif()
+      if(${var} MATCHES "-xO2")
+        string(REGEX REPLACE "-xO2" "-xO5" ${var} "${${var}}")
+      endif()
+    endforeach()
+  endif()
+endif()
+
+string(TOUPPER ${CMAKE_BUILD_TYPE} CMAKE_BUILD_TYPE_UC)
+
+set(EFFECTIVE_C_FLAGS "${CMAKE_C_FLAGS} ${CMAKE_C_FLAGS_${CMAKE_BUILD_TYPE_UC}}")
+message(STATUS "Compiler flags = ${EFFECTIVE_C_FLAGS}")
+
+set(EFFECTIVE_LD_FLAGS "${CMAKE_EXE_LINKER_FLAGS} ${CMAKE_EXE_LINKER_FLAGS_${CMAKE_BUILD_TYPE_UC}}")
+message(STATUS "Linker flags = ${EFFECTIVE_LD_FLAGS}")
+
+include(CheckCSourceCompiles)
+include(CheckIncludeFiles)
+include(CheckTypeSize)
+
+check_type_size("size_t" SIZE_T)
+check_type_size("unsigned long" UNSIGNED_LONG)
+
+if(SIZE_T EQUAL UNSIGNED_LONG)
+  check_c_source_compiles("int main(int argc, char **argv) { unsigned long a = argc;  return __builtin_ctzl(a); }"
+    HAVE_BUILTIN_CTZL)
+endif()
+if(MSVC)
+  check_include_files("intrin.h" HAVE_INTRIN_H)
+endif()
+
+if(UNIX)
+  # Check for headers
+  check_include_files(locale.h HAVE_LOCALE_H)
+  check_include_files(stddef.h HAVE_STDDEF_H)
+  check_include_files(stdlib.h HAVE_STDLIB_H)
+  check_include_files(sys/types.h NEED_SYS_TYPES_H)
+
+  # Check for functions
+  include(CheckSymbolExists)
+  check_symbol_exists(memset string.h HAVE_MEMSET)
+  check_symbol_exists(memcpy string.h HAVE_MEMCPY)
+  if(NOT HAVE_MEMSET AND NOT HAVE_MEMCPY)
+    set(NEED_BSD_STRINGS 1)
+  endif()
+
+  # Check for types
+  check_type_size("unsigned char" UNSIGNED_CHAR)
+  check_type_size("unsigned short" UNSIGNED_SHORT)
+
+  # Check for compiler features
+  check_c_source_compiles("int main(void) { typedef struct undefined_structure *undef_struct_ptr;  undef_struct_ptr ptr = 0;  return ptr != 0; }"
+    INCOMPLETE_TYPES)
+  if(INCOMPLETE_TYPES)
+    message(STATUS "Compiler supports pointers to undefined structures.")
+  else()
+    set(INCOMPLETE_TYPES_BROKEN 1)
+    message(STATUS "Compiler does not support pointers to undefined structures.")
+  endif()
+
+  if(CMAKE_CROSSCOMPILING)
+    set(RIGHT_SHIFT_IS_UNSIGNED 0)
+  else()
+    include(CheckCSourceRuns)
+    check_c_source_runs("
+      #include <stdio.h>
+      #include <stdlib.h>
+      int is_shifting_signed (long arg) {
+        long res = arg >> 4;
+        if (res == -0x7F7E80CL)
+          return 1; /* right shift is signed */
+        /* see if unsigned-shift hack will fix it. */
+        /* we can't just test exact value since it depends on width of long... */
+        res |= (~0L) << (32-4);
+        if (res == -0x7F7E80CL)
+          return 0; /* right shift is unsigned */
+        printf(\"Right shift isn't acting as I expect it to.\\\\n\");
+        printf(\"I fear the JPEG software will not work at all.\\\\n\\\\n\");
+        return 0; /* try it with unsigned anyway */
+      }
+      int main (void) {
+        exit(is_shifting_signed(-0x7F7E80B1L));
+      }" RIGHT_SHIFT_IS_UNSIGNED)
+  endif()
+endif()
+
+if(MSVC)
+  set(INLINE_OPTIONS "__inline;inline")
+else()
+  set(INLINE_OPTIONS "__inline__;inline")
+endif()
+option(FORCE_INLINE "Force function inlining" TRUE)
+boolean_number(FORCE_INLINE)
+if(FORCE_INLINE)
+  if(MSVC)
+    list(INSERT INLINE_OPTIONS 0 "__forceinline")
+  else()
+    list(INSERT INLINE_OPTIONS 0 "inline __attribute__((always_inline))")
+    list(INSERT INLINE_OPTIONS 0 "__inline__ __attribute__((always_inline))")
+  endif()
+endif()
+foreach(inline ${INLINE_OPTIONS})
+  check_c_source_compiles("${inline} static int foo(void) { return 0; } int main(void) { return foo(); }"
+    INLINE_WORKS)
+  if(INLINE_WORKS)
+    set(INLINE ${inline})
+    break()
+  endif()
+endforeach()
+if(NOT INLINE_WORKS)
+  message(FATAL_ERROR "Could not determine how to inline functions.")
+endif()
+message(STATUS "INLINE = ${INLINE} (FORCE_INLINE = ${FORCE_INLINE})")
+
+if(WITH_TURBOJPEG)
+  if(MSVC)
+    set(THREAD_LOCAL "__declspec(thread)")
+  else()
+    set(THREAD_LOCAL "__thread")
+  endif()
+  check_c_source_compiles("${THREAD_LOCAL} int i;  int main(void) { i = 0;  return i; }" HAVE_THREAD_LOCAL)
+  if(HAVE_THREAD_LOCAL)
+    message(STATUS "THREAD_LOCAL = ${THREAD_LOCAL}")
+  else()
+    message(WARNING "Thread-local storage is not available.  The TurboJPEG API library's global error handler will not be thread-safe.")
+    unset(THREAD_LOCAL)
+  endif()
+endif()
+
+if(UNIX AND NOT APPLE)
+  file(WRITE ${CMAKE_CURRENT_BINARY_DIR}/conftest.map "VERS_1 { global: *; };")
+  set(CMAKE_REQUIRED_FLAGS
+    "-Wl,--version-script,${CMAKE_CURRENT_BINARY_DIR}/conftest.map")
+  check_c_source_compiles("int main(void) { return 0; }" HAVE_VERSION_SCRIPT)
+  set(CMAKE_REQUIRED_FLAGS)
+  file(REMOVE ${CMAKE_CURRENT_BINARY_DIR}/conftest.map)
+  if(HAVE_VERSION_SCRIPT)
+    message(STATUS "Linker supports GNU-style version scripts")
+    set(MAPFLAG "-Wl,--version-script,")
+    set(TJMAPFLAG "-Wl,--version-script,")
+  else()
+    message(STATUS "Linker does not support GNU-style version scripts")
+    if(CMAKE_SYSTEM_NAME STREQUAL "SunOS")
+      # The Solaris linker doesn't like our version script for the libjpeg API
+      # library, but the version script for the TurboJPEG API library should
+      # still work.
+      file(WRITE ${CMAKE_CURRENT_BINARY_DIR}/conftest.map
+        "VERS_1 { global: foo;  local: *; }; VERS_2 { global: foo2; } VERS_1;")
+      set(CMAKE_REQUIRED_FLAGS "-Wl,-M,${CMAKE_CURRENT_BINARY_DIR}/conftest.map -shared")
+      check_c_source_compiles("int foo() { return 0; } int foo2() { return 2; }"
+        HAVE_MAPFILE)
+      set(CMAKE_REQUIRED_FLAGS)
+      file(REMOVE ${CMAKE_CURRENT_BINARY_DIR}/conftest.map)
+      if(HAVE_MAPFILE)
+        message(STATUS "Linker supports mapfiles")
+        set(TJMAPFLAG "-Wl,-M,")
+      else()
+        message(STATUS "Linker does not support mapfiles")
+      endif()
+    endif()
+  endif()
+endif()
+
+# Generate files
+if(WIN32)
+  configure_file(win/jconfig.h.in jconfig.h)
+else()
+  configure_file(jconfig.h.in jconfig.h)
+endif()
+configure_file(jconfigint.h.in jconfigint.h)
+if(UNIX)
+  configure_file(libjpeg.map.in libjpeg.map)
+endif()
+
+# Include directories and compiler definitions
+include_directories(${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR})
+
+
+###############################################################################
+# TARGETS
+###############################################################################
+
+if(CMAKE_EXECUTABLE_SUFFIX_TMP)
+  set(CMAKE_EXECUTABLE_SUFFIX ${CMAKE_EXECUTABLE_SUFFIX_TMP})
+endif()
+message(STATUS "CMAKE_EXECUTABLE_SUFFIX = ${CMAKE_EXECUTABLE_SUFFIX}")
+
+set(JPEG_SOURCES jcapimin.c jcapistd.c jccoefct.c jccolor.c jcdctmgr.c jchuff.c
+  jcicc.c jcinit.c jcmainct.c jcmarker.c jcmaster.c jcomapi.c jcparam.c
+  jcphuff.c jcprepct.c jcsample.c jctrans.c jdapimin.c jdapistd.c jdatadst.c
+  jdatasrc.c jdcoefct.c jdcolor.c jddctmgr.c jdhuff.c jdicc.c jdinput.c
+  jdmainct.c jdmarker.c jdmaster.c jdmerge.c jdphuff.c jdpostct.c jdsample.c
+  jdtrans.c jerror.c jfdctflt.c jfdctfst.c jfdctint.c jidctflt.c jidctfst.c
+  jidctint.c jidctred.c jquant1.c jquant2.c jutils.c jmemmgr.c jmemnobs.c)
+
+if(WITH_ARITH_ENC OR WITH_ARITH_DEC)
+  set(JPEG_SOURCES ${JPEG_SOURCES} jaricom.c)
+endif()
+
+if(WITH_ARITH_ENC)
+  set(JPEG_SOURCES ${JPEG_SOURCES} jcarith.c)
+endif()
+
+if(WITH_ARITH_DEC)
+  set(JPEG_SOURCES ${JPEG_SOURCES} jdarith.c)
+endif()
+
+if(WITH_SIMD)
+  add_subdirectory(simd)
+  if(NEON_INTRINSICS)
+    add_definitions(-DNEON_INTRINSICS)
+  endif()
+elseif(NOT WITH_12BIT)
+  message(STATUS "SIMD extensions: None (WITH_SIMD = ${WITH_SIMD})")
+endif()
+if(WITH_SIMD)
+  message(STATUS "SIMD extensions: ${CPU_TYPE} (WITH_SIMD = ${WITH_SIMD})")
+  if(MSVC_IDE OR XCODE)
+    set_source_files_properties(${SIMD_OBJS} PROPERTIES GENERATED 1)
+  endif()
+else()
+  add_library(simd OBJECT jsimd_none.c)
+  if(NOT WIN32 AND (CMAKE_POSITION_INDEPENDENT_CODE OR ENABLE_SHARED))
+    set_target_properties(simd PROPERTIES POSITION_INDEPENDENT_CODE 1)
+  endif()
+endif()
+
+if(WITH_JAVA)
+  add_subdirectory(java)
+endif()
+
+if(ENABLE_SHARED)
+  add_subdirectory(sharedlib)
+endif()
+
+if(ENABLE_STATIC)
+  add_library(jpeg-static STATIC ${JPEG_SOURCES} $<TARGET_OBJECTS:simd>
+    ${SIMD_OBJS})
+  if(NOT MSVC)
+    set_target_properties(jpeg-static PROPERTIES OUTPUT_NAME jpeg)
+  endif()
+endif()
+
+if(WITH_TURBOJPEG)
+  if(ENABLE_SHARED)
+    set(TURBOJPEG_SOURCES ${JPEG_SOURCES} $<TARGET_OBJECTS:simd> ${SIMD_OBJS}
+      turbojpeg.c transupp.c jdatadst-tj.c jdatasrc-tj.c rdbmp.c rdppm.c
+      wrbmp.c wrppm.c)
+    set(TJMAPFILE ${CMAKE_CURRENT_SOURCE_DIR}/turbojpeg-mapfile)
+    if(WITH_JAVA)
+      set(TURBOJPEG_SOURCES ${TURBOJPEG_SOURCES} turbojpeg-jni.c)
+      include_directories(${JAVA_INCLUDE_PATH} ${JAVA_INCLUDE_PATH2})
+      set(TJMAPFILE ${CMAKE_CURRENT_SOURCE_DIR}/turbojpeg-mapfile.jni)
+    endif()
+    add_library(turbojpeg SHARED ${TURBOJPEG_SOURCES})
+    set_property(TARGET turbojpeg PROPERTY COMPILE_FLAGS
+      "-DBMP_SUPPORTED -DPPM_SUPPORTED")
+    if(WIN32)
+      set_target_properties(turbojpeg PROPERTIES DEFINE_SYMBOL DLLDEFINE)
+    endif()
+    if(MINGW)
+      set_target_properties(turbojpeg PROPERTIES LINK_FLAGS -Wl,--kill-at)
+    endif()
+    if(APPLE AND (NOT CMAKE_OSX_DEPLOYMENT_TARGET OR
+                  CMAKE_OSX_DEPLOYMENT_TARGET VERSION_GREATER 10.4))
+      if(NOT CMAKE_SHARED_LIBRARY_RUNTIME_C_FLAG)
+        set(CMAKE_SHARED_LIBRARY_RUNTIME_C_FLAG "-Wl,-rpath,")
+      endif()
+      set_target_properties(turbojpeg PROPERTIES MACOSX_RPATH 1)
+    endif()
+    set_target_properties(turbojpeg PROPERTIES
+      SOVERSION ${TURBOJPEG_SO_MAJOR_VERSION} VERSION ${TURBOJPEG_SO_VERSION})
+    if(TJMAPFLAG)
+      set_target_properties(turbojpeg PROPERTIES
+        LINK_FLAGS "${TJMAPFLAG}${TJMAPFILE}")
+    endif()
+
+    add_executable(tjunittest tjunittest.c tjutil.c md5/md5.c md5/md5hl.c)
+    target_link_libraries(tjunittest turbojpeg)
+
+    add_executable(tjbench tjbench.c tjutil.c)
+    target_link_libraries(tjbench turbojpeg)
+    if(UNIX)
+      target_link_libraries(tjbench m)
+    endif()
+
+    add_executable(tjexample tjexample.c)
+    target_link_libraries(tjexample turbojpeg)
+  endif()
+
+  if(ENABLE_STATIC)
+    add_library(turbojpeg-static STATIC ${JPEG_SOURCES} $<TARGET_OBJECTS:simd>
+      ${SIMD_OBJS} turbojpeg.c transupp.c jdatadst-tj.c jdatasrc-tj.c rdbmp.c
+      rdppm.c wrbmp.c wrppm.c)
+    set_property(TARGET turbojpeg-static PROPERTY COMPILE_FLAGS
+      "-DBMP_SUPPORTED -DPPM_SUPPORTED")
+    if(NOT MSVC)
+      set_target_properties(turbojpeg-static PROPERTIES OUTPUT_NAME turbojpeg)
+    endif()
+
+    add_executable(tjunittest-static tjunittest.c tjutil.c md5/md5.c
+      md5/md5hl.c)
+    target_link_libraries(tjunittest-static turbojpeg-static)
+
+    add_executable(tjbench-static tjbench.c tjutil.c)
+    target_link_libraries(tjbench-static turbojpeg-static)
+    if(UNIX)
+      target_link_libraries(tjbench-static m)
+    endif()
+  endif()
+endif()
+
+if(WIN32)
+  set(USE_SETMODE "-DUSE_SETMODE")
+endif()
+if(WITH_12BIT)
+  set(COMPILE_FLAGS "-DGIF_SUPPORTED -DPPM_SUPPORTED ${USE_SETMODE}")
+else()
+  set(COMPILE_FLAGS "-DBMP_SUPPORTED -DGIF_SUPPORTED -DPPM_SUPPORTED -DTARGA_SUPPORTED ${USE_SETMODE}")
+  set(CJPEG_BMP_SOURCES rdbmp.c rdtarga.c)
+  set(DJPEG_BMP_SOURCES wrbmp.c wrtarga.c)
+endif()
+
+if(ENABLE_STATIC)
+  add_executable(cjpeg-static cjpeg.c cdjpeg.c rdgif.c rdppm.c rdswitch.c
+    ${CJPEG_BMP_SOURCES})
+  set_property(TARGET cjpeg-static PROPERTY COMPILE_FLAGS ${COMPILE_FLAGS})
+  target_link_libraries(cjpeg-static jpeg-static)
+
+  add_executable(djpeg-static djpeg.c cdjpeg.c rdcolmap.c rdswitch.c wrgif.c
+    wrppm.c ${DJPEG_BMP_SOURCES})
+  set_property(TARGET djpeg-static PROPERTY COMPILE_FLAGS ${COMPILE_FLAGS})
+  target_link_libraries(djpeg-static jpeg-static)
+
+  add_executable(jpegtran-static jpegtran.c cdjpeg.c rdswitch.c transupp.c)
+  target_link_libraries(jpegtran-static jpeg-static)
+  set_property(TARGET jpegtran-static PROPERTY COMPILE_FLAGS "${USE_SETMODE}")
+endif()
+
+add_executable(rdjpgcom rdjpgcom.c)
+
+add_executable(wrjpgcom wrjpgcom.c)
+
+
+###############################################################################
+# TESTS
+###############################################################################
+
+if(WITH_FUZZ)
+  add_subdirectory(fuzz)
+endif()
+
+add_subdirectory(md5)
+
+if(MSVC_IDE OR XCODE)
+  set(OBJDIR "\${CTEST_CONFIGURATION_TYPE}/")
+else()
+  set(OBJDIR "")
+endif()
+
+enable_testing()
+
+if(WITH_12BIT)
+  set(TESTORIG testorig12.jpg)
+  set(MD5_JPEG_RGB_ISLOW 9d7369207c520d37f2c1cbfcb82b2964)
+  set(MD5_JPEG_RGB_ISLOW2 a00bd20d8ae49684640ef7177d2e0b64)
+  set(MD5_PPM_RGB_ISLOW f3301d2219783b8b3d942b7239fa50c0)
+  set(MD5_JPEG_422_IFAST_OPT 7322e3bd2f127f7de4b40d4480ce60e4)
+  set(MD5_PPM_422_IFAST 79807fa552899e66a04708f533e16950)
+  set(MD5_JPEG_440_ISLOW e25c1912e38367be505a89c410c1c2d2)
+  set(MD5_PPM_440_ISLOW e7d2e26288870cfcb30f3114ad01e380)
+  set(MD5_PPM_422M_IFAST 07737bfe8a7c1c87aaa393a0098d16b0)
+  set(MD5_JPEG_420_IFAST_Q100_PROG 9447cef4803d9b0f74bcf333cc710a29)
+  set(MD5_PPM_420_Q100_IFAST 1b3730122709f53d007255e8dfd3305e)
+  set(MD5_PPM_420M_Q100_IFAST 980a1a3c5bf9510022869d30b7d26566)
+  set(MD5_JPEG_GRAY_ISLOW 235c90707b16e2e069f37c888b2636d9)
+  set(MD5_PPM_GRAY_ISLOW 7213c10af507ad467da5578ca5ee1fca)
+  set(MD5_PPM_GRAY_ISLOW_RGB e96ee81c30a6ed422d466338bd3de65d)
+  set(MD5_JPEG_420S_IFAST_OPT 7af8e60be4d9c227ec63ac9b6630855e)
+
+  set(MD5_JPEG_3x2_FLOAT_PROG_SSE a8c17daf77b457725ec929e215b603f8)
+  set(MD5_PPM_3x2_FLOAT_SSE 42876ab9e5c2f76a87d08db5fbd57956)
+  set(MD5_JPEG_3x2_FLOAT_PROG_NO_FP_CONTRACT a8c17daf77b457725ec929e215b603f8)
+  set(MD5_PPM_3x2_FLOAT_NO_FP_CONTRACT ${MD5_PPM_3x2_FLOAT_SSE})
+  set(MD5_JPEG_3x2_FLOAT_PROG_FP_CONTRACT
+    ${MD5_JPEG_3x2_FLOAT_PROG_NO_FP_CONTRACT})
+  set(MD5_PPM_3x2_FLOAT_FP_CONTRACT ${MD5_PPM_3x2_FLOAT_SSE})
+  set(MD5_JPEG_3x2_FLOAT_PROG_387 bc6dbbefac2872f6b9d6c4a0ae60c3c0)
+  set(MD5_PPM_3x2_FLOAT_387 bcc5723c61560463ac60f772e742d092)
+  set(MD5_JPEG_3x2_FLOAT_PROG_MSVC e27840755870fa849872e58aa0cd1400)
+  set(MD5_PPM_3x2_FLOAT_MSVC 6c2880b83bb1aa41dfe330e7a9768690)
+
+  set(MD5_JPEG_3x2_IFAST_PROG 1396cc2b7185cfe943d408c9d305339e)
+  set(MD5_PPM_3x2_IFAST 3975985ef6eeb0a2cdc58daa651ccc00)
+  set(MD5_PPM_420M_ISLOW_2_1 4ca6be2a6f326ff9eaab63e70a8259c0)
+  set(MD5_PPM_420M_ISLOW_15_8 12aa9f9534c1b3d7ba047322226365eb)
+  set(MD5_PPM_420M_ISLOW_13_8 f7e22817c7b25e1393e4ec101e9d4e96)
+  set(MD5_PPM_420M_ISLOW_11_8 800a16f9f4dc9b293197bfe11be10a82)
+  set(MD5_PPM_420M_ISLOW_9_8 06b7a92a9bc69f4dc36ec40f1937d55c)
+  set(MD5_PPM_420M_ISLOW_7_8 3ec444a14a4ab4eab88ffc49c48eca43)
+  set(MD5_PPM_420M_ISLOW_3_4 3e726b7ea872445b19437d1c1d4f0d93)
+  set(MD5_PPM_420M_ISLOW_5_8 a8a771abdc94301d20ffac119b2caccd)
+  set(MD5_PPM_420M_ISLOW_1_2 b419124dd5568b085787234866102866)
+  set(MD5_PPM_420M_ISLOW_3_8 343d19015531b7bbe746124127244fa8)
+  set(MD5_PPM_420M_ISLOW_1_4 35fd59d866e44659edfa3c18db2a3edb)
+  set(MD5_PPM_420M_ISLOW_1_8 ccaed48ac0aedefda5d4abe4013f4ad7)
+  set(MD5_PPM_420_ISLOW_SKIP15_31 86664cd9dc956536409e44e244d20a97)
+  set(MD5_PPM_420_ISLOW_PROG_CROP62x62_71_71 452a21656115a163029cfba5c04fa76a)
+  set(MD5_PPM_444_ISLOW_SKIP1_6 ef63901f71ef7a75cd78253fc0914f84)
+  set(MD5_PPM_444_ISLOW_PROG_CROP98x98_13_13 15b173fb5872d9575572fbcc1b05956f)
+  set(MD5_JPEG_CROP cdb35ff4b4519392690ea040c56ea99c)
+else()
+  set(TESTORIG testorig.jpg)
+  set(MD5_JPEG_RGB_ISLOW 1d44a406f61da743b5fd31c0a9abdca3)
+  set(MD5_JPEG_RGB_ISLOW2 31d121e57b6c2934c890a7fc7763bcd4)
+  set(MD5_PPM_RGB_ISLOW 00a257f5393fef8821f2b88ac7421291)
+  set(MD5_BMP_RGB_ISLOW_565 f07d2e75073e4bb10f6c6f4d36e2e3be)
+  set(MD5_BMP_RGB_ISLOW_565D 4cfa0928ef3e6bb626d7728c924cfda4)
+  set(MD5_JPEG_422_IFAST_OPT 2540287b79d913f91665e660303ab2c8)
+  set(MD5_PPM_422_IFAST 35bd6b3f833bad23de82acea847129fa)
+  set(MD5_JPEG_440_ISLOW 538bc02bd4b4658fd85de6ece6cbeda6)
+  set(MD5_PPM_440_ISLOW 11e7eab7ef7ef3276934bb7e7b6bb377)
+  set(MD5_PPM_422M_IFAST 8dbc65323d62cca7c91ba02dd1cfa81d)
+  set(MD5_BMP_422M_IFAST_565 3294bd4d9a1f2b3d08ea6020d0db7065)
+  set(MD5_BMP_422M_IFAST_565D da98c9c7b6039511be4a79a878a9abc1)
+  set(MD5_JPEG_420_IFAST_Q100_PROG 0ba15f9dab81a703505f835f9dbbac6d)
+  set(MD5_PPM_420_Q100_IFAST 5a732542015c278ff43635e473a8a294)
+  set(MD5_PPM_420M_Q100_IFAST ff692ee9323a3b424894862557c092f1)
+  set(MD5_JPEG_GRAY_ISLOW 72b51f894b8f4a10b3ee3066770aa38d)
+  set(MD5_PPM_GRAY_ISLOW 8d3596c56eace32f205deccc229aa5ed)
+  set(MD5_PPM_GRAY_ISLOW_RGB 116424ac07b79e5e801f00508eab48ec)
+  set(MD5_BMP_GRAY_ISLOW_565 12f78118e56a2f48b966f792fedf23cc)
+  set(MD5_BMP_GRAY_ISLOW_565D bdbbd616441a24354c98553df5dc82db)
+  set(MD5_JPEG_420S_IFAST_OPT 388708217ac46273ca33086b22827ed8)
+
+  set(MD5_JPEG_3x2_FLOAT_PROG_SSE 343e3f8caf8af5986ebaf0bdc13b5c71)
+  set(MD5_PPM_3x2_FLOAT_SSE 1a75f36e5904d6fc3a85a43da9ad89bb)
+  set(MD5_JPEG_3x2_FLOAT_PROG_NO_FP_CONTRACT 9bca803d2042bd1eb03819e2bf92b3e5)
+  set(MD5_PPM_3x2_FLOAT_NO_FP_CONTRACT f6bfab038438ed8f5522fbd33595dcdc)
+  set(MD5_JPEG_3x2_FLOAT_PROG_FP_CONTRACT
+    ${MD5_JPEG_3x2_FLOAT_PROG_NO_FP_CONTRACT})
+  set(MD5_PPM_3x2_FLOAT_FP_CONTRACT 0e917a34193ef976b679a6b069b1be26)
+  set(MD5_JPEG_3x2_FLOAT_PROG_387 1657664a410e0822c924b54f6f65e6e9)
+  set(MD5_PPM_3x2_FLOAT_387 cb0a1f027f3d2917c902b5640214e025)
+  set(MD5_JPEG_3x2_FLOAT_PROG_MSVC 7999ce9cd0ee9b6c7043b7351ab7639d)
+  set(MD5_PPM_3x2_FLOAT_MSVC 28cdc448a6b75e97892f0e0f8d4b21f3)
+
+  set(MD5_JPEG_3x2_IFAST_PROG 1ee5d2c1a77f2da495f993c8c7cceca5)
+  set(MD5_PPM_3x2_IFAST fd283664b3b49127984af0a7f118fccd)
+  set(MD5_JPEG_420_ISLOW_ARI e986fb0a637a8d833d96e8a6d6d84ea1)
+  set(MD5_JPEG_444_ISLOW_PROGARI 0a8f1c8f66e113c3cf635df0a475a617)
+  set(MD5_PPM_420M_IFAST_ARI 57251da28a35b46eecb7177d82d10e0e)
+  set(MD5_JPEG_420_ISLOW 9a68f56bc76e466aa7e52f415d0f4a5f)
+  set(MD5_PPM_420M_ISLOW_2_1 9f9de8c0612f8d06869b960b05abf9c9)
+  set(MD5_PPM_420M_ISLOW_15_8 b6875bc070720b899566cc06459b63b7)
+  set(MD5_PPM_420M_ISLOW_13_8 bc3452573c8152f6ae552939ee19f82f)
+  set(MD5_PPM_420M_ISLOW_11_8 d8cc73c0aaacd4556569b59437ba00a5)
+  set(MD5_PPM_420M_ISLOW_9_8 d25e61bc7eac0002f5b393aa223747b6)
+  set(MD5_PPM_420M_ISLOW_7_8 ddb564b7c74a09494016d6cd7502a946)
+  set(MD5_PPM_420M_ISLOW_3_4 8ed8e68808c3fbc4ea764fc9d2968646)
+  set(MD5_PPM_420M_ISLOW_5_8 a3363274999da2366a024efae6d16c9b)
+  set(MD5_PPM_420M_ISLOW_1_2 e692a315cea26b988c8e8b29a5dbcd81)
+  set(MD5_PPM_420M_ISLOW_3_8 79eca9175652ced755155c90e785a996)
+  set(MD5_PPM_420M_ISLOW_1_4 79cd778f8bf1a117690052cacdd54eca)
+  set(MD5_PPM_420M_ISLOW_1_8 391b3d4aca640c8567d6f8745eb2142f)
+  set(MD5_BMP_420_ISLOW_256 4980185e3776e89bd931736e1cddeee6)
+  set(MD5_BMP_420_ISLOW_565 bf9d13e16c4923b92e1faa604d7922cb)
+  set(MD5_BMP_420_ISLOW_565D 6bde71526acc44bcff76f696df8638d2)
+  set(MD5_BMP_420M_ISLOW_565 8dc0185245353cfa32ad97027342216f)
+  set(MD5_BMP_420M_ISLOW_565D ce034037d212bc403330df6f915c161b)
+  set(MD5_PPM_420_ISLOW_SKIP15_31 c4c65c1e43d7275cd50328a61e6534f0)
+  set(MD5_PPM_420_ISLOW_ARI_SKIP16_139 087c6b123db16ac00cb88c5b590bb74a)
+  set(MD5_PPM_420_ISLOW_PROG_CROP62x62_71_71 26eb36ccc7d1f0cb80cdabb0ac8b5d99)
+  set(MD5_PPM_420_ISLOW_ARI_CROP53x53_4_4 886c6775af22370257122f8b16207e6d)
+  set(MD5_PPM_444_ISLOW_SKIP1_6 5606f86874cf26b8fcee1117a0a436a6)
+  set(MD5_PPM_444_ISLOW_PROG_CROP98x98_13_13 db87dc7ce26bcdc7a6b56239ce2b9d6c)
+  set(MD5_PPM_444_ISLOW_ARI_CROP37x37_0_0 cb57b32bd6d03e35432362f7bf184b6d)
+  set(MD5_JPEG_CROP b4197f377e621c4e9b1d20471432610d)
+endif()
+
+if(WITH_JAVA)
+  add_test(TJUnitTest
+    ${Java_JAVA_EXECUTABLE} ${JAVAARGS} -cp java/turbojpeg.jar
+      -Djava.library.path=${CMAKE_CURRENT_BINARY_DIR}/${OBJDIR}
+      TJUnitTest)
+  add_test(TJUnitTest-yuv
+    ${Java_JAVA_EXECUTABLE} ${JAVAARGS} -cp java/turbojpeg.jar
+      -Djava.library.path=${CMAKE_CURRENT_BINARY_DIR}/${OBJDIR}
+      TJUnitTest -yuv)
+  add_test(TJUnitTest-yuv-nopad
+    ${Java_JAVA_EXECUTABLE} ${JAVAARGS} -cp java/turbojpeg.jar
+      -Djava.library.path=${CMAKE_CURRENT_BINARY_DIR}/${OBJDIR}
+      TJUnitTest -yuv -noyuvpad)
+  add_test(TJUnitTest-bi
+    ${Java_JAVA_EXECUTABLE} ${JAVAARGS} -cp java/turbojpeg.jar
+      -Djava.library.path=${CMAKE_CURRENT_BINARY_DIR}/${OBJDIR}
+      TJUnitTest -bi)
+  add_test(TJUnitTest-bi-yuv
+    ${Java_JAVA_EXECUTABLE} ${JAVAARGS} -cp java/turbojpeg.jar
+      -Djava.library.path=${CMAKE_CURRENT_BINARY_DIR}/${OBJDIR}
+      TJUnitTest -bi -yuv)
+  add_test(TJUnitTest-bi-yuv-nopad
+    ${Java_JAVA_EXECUTABLE} ${JAVAARGS} -cp java/turbojpeg.jar
+      -Djava.library.path=${CMAKE_CURRENT_BINARY_DIR}/${OBJDIR}
+      TJUnitTest -bi -yuv -noyuvpad)
+endif()
+
+set(TEST_LIBTYPES "")
+if(ENABLE_SHARED)
+  set(TEST_LIBTYPES ${TEST_LIBTYPES} shared)
+endif()
+if(ENABLE_STATIC)
+  set(TEST_LIBTYPES ${TEST_LIBTYPES} static)
+endif()
+
+set(TESTIMAGES ${CMAKE_CURRENT_SOURCE_DIR}/testimages)
+set(MD5CMP ${CMAKE_CURRENT_BINARY_DIR}/md5/md5cmp)
+if(CMAKE_CROSSCOMPILING)
+  file(RELATIVE_PATH TESTIMAGES ${CMAKE_CURRENT_BINARY_DIR} ${TESTIMAGES})
+  file(RELATIVE_PATH MD5CMP ${CMAKE_CURRENT_BINARY_DIR} ${MD5CMP})
+endif()
+
+# The output of the floating point DCT/IDCT algorithms differs depending on the
+# type of floating point math used, so the FLOATTEST CMake variable must be
+# set in order to tell the testing system which floating point results it
+# should expect:
+#
+# sse = validate against the expected results from the libjpeg-turbo SSE SIMD
+#       extensions
+# no-fp-contract = validate against the expected results from the C code when
+#                  floating point expression contraction is disabled (the
+#                  default with Clang, with GCC when building for platforms
+#                  that lack fused multiply-add FMA instructions, or when
+#                  passing -ffp-contract=off to the compiler)
+# fp-contract = validate against the expected results from the C code when
+#               floating point expression contraction is enabled (the default
+#               with GCC when building for platforms that have fused multiply-
+#               add FMA instructions or when passing -ffp-contract=fast to
+#               the compiler)
+# 387 = validate against the expected results from the C code when the 387 FPU
+#       is being used for floating point math (which is generally the default
+#       with x86 compilers)
+# msvc = validate against the expected results from the C code when compiled
+#        with a 32-bit version of Visual C++
+
+if(CPU_TYPE STREQUAL "x86_64" OR CPU_TYPE STREQUAL "i386")
+  if(WITH_SIMD)
+    set(DEFAULT_FLOATTEST sse)
+  elseif(CPU_TYPE STREQUAL "x86_64")
+    set(DEFAULT_FLOATTEST no-fp-contract)
+  elseif(CPU_TYPE STREQUAL "i386" AND MSVC)
+    set(DEFAULT_FLOATTEST msvc)
+  # else we can't really set an intelligent default for i386.  The appropriate
+  # value could be 387, no-fp-contract, or fp-contract, depending on the
+  # compiler and compiler options.  We leave it to the user to set FLOATTEST
+  # manually.
+  endif()
+else()
+  if((CPU_TYPE STREQUAL "powerpc" OR CPU_TYPE STREQUAL "arm64") AND
+    NOT CMAKE_C_COMPILER_ID STREQUAL "Clang" AND NOT MSVC)
+    set(DEFAULT_FLOATTEST fp-contract)
+  else()
+    set(DEFAULT_FLOATTEST no-fp-contract)
+  endif()
+endif()
+
+# This causes FLOATTEST to reset to the default value if WITH_SIMD has
+# changed.
+if(DEFINED WITH_SIMD_INT AND NOT WITH_SIMD EQUAL WITH_SIMD_INT)
+  set(FORCE_FLOATTEST "FORCE")
+endif()
+set(WITH_SIMD_INT ${WITH_SIMD} CACHE INTERNAL "")
+set(FLOATTEST ${DEFAULT_FLOATTEST} CACHE STRING
+  "The type of floating point math used by the floating point DCT/IDCT algorithms.  This tells the testing system which numerical results it should expect from those tests.  sse = libjpeg-turbo x86/x86-64 SIMD extensions, no-fp-contract = generic FPU with floating point expression contraction disabled, fp-contract = generic FPU with floating point expression contraction enabled, 387 = 387 FPU, msvc = 32-bit Visual Studio (default = ${DEFAULT_FLOATTEST})"
+  ${FORCE_FLOATTEST})
+message(STATUS "FLOATTEST = ${FLOATTEST}")
+
+if(FLOATTEST)
+  string(TOUPPER ${FLOATTEST} FLOATTEST_UC)
+  string(REGEX REPLACE "-" "_" FLOATTEST_UC ${FLOATTEST_UC})
+  string(TOLOWER ${FLOATTEST} FLOATTEST)
+  if(NOT FLOATTEST STREQUAL "sse" AND
+    NOT FLOATTEST STREQUAL "no-fp-contract" AND
+    NOT FLOATTEST STREQUAL "fp-contract" AND NOT FLOATTEST STREQUAL "387" AND
+    NOT FLOATTEST STREQUAL "msvc")
+    message(FATAL_ERROR "\"${FLOATTEST}\" is not a valid value for FLOATTEST.")
+  endif()
+endif()
+
+foreach(libtype ${TEST_LIBTYPES})
+  if(libtype STREQUAL "static")
+    set(suffix -static)
+  endif()
+  if(WITH_TURBOJPEG)
+    add_test(tjunittest-${libtype}
+      ${CMAKE_CROSSCOMPILING_EMULATOR} tjunittest${suffix})
+    add_test(tjunittest-${libtype}-alloc
+      ${CMAKE_CROSSCOMPILING_EMULATOR} tjunittest${suffix} -alloc)
+    add_test(tjunittest-${libtype}-yuv
+      ${CMAKE_CROSSCOMPILING_EMULATOR} tjunittest${suffix} -yuv)
+    add_test(tjunittest-${libtype}-yuv-alloc
+      ${CMAKE_CROSSCOMPILING_EMULATOR} tjunittest${suffix} -yuv -alloc)
+    add_test(tjunittest-${libtype}-yuv-nopad
+      ${CMAKE_CROSSCOMPILING_EMULATOR} tjunittest${suffix} -yuv -noyuvpad)
+    add_test(tjunittest-${libtype}-bmp
+      ${CMAKE_CROSSCOMPILING_EMULATOR} tjunittest${suffix} -bmp)
+
+    set(MD5_PPM_GRAY_TILE 89d3ca21213d9d864b50b4e4e7de4ca6)
+    set(MD5_PPM_420_8x8_TILE 847fceab15c5b7b911cb986cf0f71de3)
+    set(MD5_PPM_420_16x16_TILE ca45552a93687e078f7137cc4126a7b0)
+    set(MD5_PPM_420_32x32_TILE d8676f1d6b68df358353bba9844f4a00)
+    set(MD5_PPM_420_64x64_TILE 4e4c1a3d7ea4bace4f868bcbe83b7050)
+    set(MD5_PPM_420_128x128_TILE f24c3429c52265832beab9df72a0ceae)
+    set(MD5_PPM_420M_8x8_TILE bc25320e1f4c31ce2e610e43e9fd173c)
+    set(MD5_PPM_420M_TILE 75ffdf14602258c5c189522af57fa605)
+    set(MD5_PPM_422_8x8_TILE d83dacd9fc73b0a6f10c09acad64eb1e)
+    set(MD5_PPM_422_16x16_TILE 35077fb610d72dd743b1eb0cbcfe10fb)
+    set(MD5_PPM_422_32x32_TILE e6902ed8a449ecc0f0d6f2bf945f65f7)
+    set(MD5_PPM_422_64x64_TILE 2b4502a8f316cedbde1da7bce3d2231e)
+    set(MD5_PPM_422_128x128_TILE f0b5617d578f5e13c8eee215d64d4877)
+    set(MD5_PPM_422M_8x8_TILE 828941d7f41cd6283abd6beffb7fd51d)
+    set(MD5_PPM_422M_TILE e877ae1324c4a280b95376f7f018172f)
+    set(MD5_PPM_444_TILE 7964e41e67cfb8d0a587c0aa4798f9c3)
+
+    # Test compressing from/decompressing to an arbitrary subregion of a larger
+    # image buffer
+    add_test(tjbench-${libtype}-tile-cp
+      ${CMAKE_COMMAND} -E copy_if_different ${TESTIMAGES}/testorig.ppm
+        testout_tile.ppm)
+    add_test(tjbench-${libtype}-tile
+      ${CMAKE_CROSSCOMPILING_EMULATOR} tjbench${suffix} testout_tile.ppm 95
+        -rgb -quiet -tile -benchtime 0.01 -warmup 0)
+    set_tests_properties(tjbench-${libtype}-tile
+      PROPERTIES DEPENDS tjbench-${libtype}-tile-cp)
+
+    foreach(tile 8 16 32 64 128)
+      add_test(tjbench-${libtype}-tile-gray-${tile}x${tile}-cmp
+        ${CMAKE_CROSSCOMPILING_EMULATOR} ${MD5CMP} ${MD5_PPM_GRAY_TILE}
+          testout_tile_GRAY_Q95_${tile}x${tile}.ppm)
+      foreach(subsamp 420 422)
+        add_test(tjbench-${libtype}-tile-${subsamp}-${tile}x${tile}-cmp
+          ${CMAKE_CROSSCOMPILING_EMULATOR} ${MD5CMP}
+            ${MD5_PPM_${subsamp}_${tile}x${tile}_TILE}
+            testout_tile_${subsamp}_Q95_${tile}x${tile}.ppm)
+      endforeach()
+      add_test(tjbench-${libtype}-tile-444-${tile}x${tile}-cmp
+        ${CMAKE_CROSSCOMPILING_EMULATOR} ${MD5CMP} ${MD5_PPM_444_TILE}
+          testout_tile_444_Q95_${tile}x${tile}.ppm)
+      foreach(subsamp gray 420 422 444)
+        set_tests_properties(tjbench-${libtype}-tile-${subsamp}-${tile}x${tile}-cmp
+          PROPERTIES DEPENDS tjbench-${libtype}-tile)
+      endforeach()
+    endforeach()
+
+    add_test(tjbench-${libtype}-tilem-cp
+      ${CMAKE_COMMAND} -E copy_if_different ${TESTIMAGES}/testorig.ppm
+        testout_tilem.ppm)
+    add_test(tjbench-${libtype}-tilem
+      ${CMAKE_CROSSCOMPILING_EMULATOR} tjbench${suffix} testout_tilem.ppm 95
+        -rgb -fastupsample -quiet -tile -benchtime 0.01 -warmup 0)
+    set_tests_properties(tjbench-${libtype}-tilem
+      PROPERTIES DEPENDS tjbench-${libtype}-tilem-cp)
+
+    add_test(tjbench-${libtype}-tile-420m-8x8-cmp
+      ${CMAKE_CROSSCOMPILING_EMULATOR} ${MD5CMP} ${MD5_PPM_420M_8x8_TILE}
+        testout_tilem_420_Q95_8x8.ppm)
+    add_test(tjbench-${libtype}-tile-422m-8x8-cmp
+      ${CMAKE_CROSSCOMPILING_EMULATOR} ${MD5CMP} ${MD5_PPM_422M_8x8_TILE}
+        testout_tilem_422_Q95_8x8.ppm)
+    foreach(tile 16 32 64 128)
+      foreach(subsamp 420 422)
+        add_test(tjbench-${libtype}-tile-${subsamp}m-${tile}x${tile}-cmp
+          ${CMAKE_CROSSCOMPILING_EMULATOR} ${MD5CMP}
+            ${MD5_PPM_${subsamp}M_TILE}
+            testout_tilem_${subsamp}_Q95_${tile}x${tile}.ppm)
+      endforeach()
+    endforeach()
+    foreach(tile 8 16 32 64 128)
+      foreach(subsamp 420 422)
+        set_tests_properties(tjbench-${libtype}-tile-${subsamp}m-${tile}x${tile}-cmp
+          PROPERTIES DEPENDS tjbench-${libtype}-tilem)
+      endforeach()
+    endforeach()
+  endif()
+
+  # These tests are carefully crafted to provide full coverage of as many of
+  # the underlying algorithms as possible (including all of the
+  # SIMD-accelerated ones.)
+
+  macro(add_bittest PROG NAME ARGS OUTFILE INFILE MD5SUM)
+    add_test(${PROG}-${libtype}-${NAME}
+      ${CMAKE_CROSSCOMPILING_EMULATOR} ${PROG}${suffix} ${ARGS}
+        -outfile ${OUTFILE} ${INFILE})
+    add_test(${PROG}-${libtype}-${NAME}-cmp
+      ${CMAKE_CROSSCOMPILING_EMULATOR} ${MD5CMP} ${MD5SUM} ${OUTFILE})
+    set_tests_properties(${PROG}-${libtype}-${NAME}-cmp PROPERTIES
+      DEPENDS ${PROG}-${libtype}-${NAME})
+    if(${ARGC} GREATER 6)
+      set(DEPENDS ${ARGN})
+      set_tests_properties(${PROG}-${libtype}-${NAME} PROPERTIES
+        DEPENDS ${DEPENDS})
+    endif()
+  endmacro()
+
+  # CC: null  SAMP: fullsize  FDCT: islow  ENT: huff
+  add_bittest(cjpeg rgb-islow "-rgb;-dct;int;-icc;${TESTIMAGES}/test1.icc"
+    testout_rgb_islow.jpg ${TESTIMAGES}/testorig.ppm
+    ${MD5_JPEG_RGB_ISLOW})
+
+  # CC: null  SAMP: fullsize  IDCT: islow  ENT: huff
+  add_bittest(djpeg rgb-islow "-dct;int;-ppm;-icc;testout_rgb_islow.icc"
+    testout_rgb_islow.ppm testout_rgb_islow.jpg
+    ${MD5_PPM_RGB_ISLOW} cjpeg-${libtype}-rgb-islow)
+
+  add_test(djpeg-${libtype}-rgb-islow-icc-cmp
+    ${CMAKE_CROSSCOMPILING_EMULATOR} ${MD5CMP}
+      b06a39d730129122e85c1363ed1bbc9e testout_rgb_islow.icc)
+  set_tests_properties(djpeg-${libtype}-rgb-islow-icc-cmp PROPERTIES
+    DEPENDS djpeg-${libtype}-rgb-islow)
+
+  add_bittest(jpegtran icc "-copy;all;-icc;${TESTIMAGES}/test2.icc"
+    testout_rgb_islow2.jpg testout_rgb_islow.jpg
+    ${MD5_JPEG_RGB_ISLOW2} cjpeg-${libtype}-rgb-islow)
+
+  if(NOT WITH_12BIT)
+    # CC: RGB->RGB565  SAMP: fullsize  IDCT: islow  ENT: huff
+    add_bittest(djpeg rgb-islow-565 "-dct;int;-rgb565;-dither;none;-bmp"
+      testout_rgb_islow_565.bmp testout_rgb_islow.jpg
+      ${MD5_BMP_RGB_ISLOW_565} cjpeg-${libtype}-rgb-islow)
+
+    # CC: RGB->RGB565 (dithered)  SAMP: fullsize  IDCT: islow  ENT: huff
+    add_bittest(djpeg rgb-islow-565D "-dct;int;-rgb565;-bmp"
+      testout_rgb_islow_565D.bmp testout_rgb_islow.jpg
+      ${MD5_BMP_RGB_ISLOW_565D} cjpeg-${libtype}-rgb-islow)
+  endif()
+
+  # CC: RGB->YCC  SAMP: fullsize/h2v1  FDCT: ifast  ENT: 2-pass huff
+  add_bittest(cjpeg 422-ifast-opt "-sample;2x1;-dct;fast;-opt"
+    testout_422_ifast_opt.jpg ${TESTIMAGES}/testorig.ppm
+    ${MD5_JPEG_422_IFAST_OPT})
+
+  # CC: YCC->RGB  SAMP: fullsize/h2v1 fancy  IDCT: ifast  ENT: huff
+  add_bittest(djpeg 422-ifast "-dct;fast"
+    testout_422_ifast.ppm testout_422_ifast_opt.jpg
+    ${MD5_PPM_422_IFAST} cjpeg-${libtype}-422-ifast-opt)
+
+  # CC: RGB->YCC  SAMP: fullsize/h1v2  FDCT: islow  ENT: huff
+  add_bittest(cjpeg 440-islow "-sample;1x2;-dct;int"
+    testout_440_islow.jpg ${TESTIMAGES}/testorig.ppm
+    ${MD5_JPEG_440_ISLOW})
+
+  # CC: YCC->RGB  SAMP: fullsize/h1v2 fancy  IDCT: islow  ENT: huff
+  add_bittest(djpeg 440-islow "-dct;int"
+    testout_440_islow.ppm testout_440_islow.jpg
+    ${MD5_PPM_440_ISLOW} cjpeg-${libtype}-440-islow)
+
+  # CC: YCC->RGB  SAMP: h2v1 merged  IDCT: ifast  ENT: huff
+  add_bittest(djpeg 422m-ifast "-dct;fast;-nosmooth"
+    testout_422m_ifast.ppm testout_422_ifast_opt.jpg
+    ${MD5_PPM_422M_IFAST} cjpeg-${libtype}-422-ifast-opt)
+
+  if(NOT WITH_12BIT)
+    # CC: YCC->RGB565  SAMP: h2v1 merged  IDCT: ifast  ENT: huff
+    add_bittest(djpeg 422m-ifast-565
+      "-dct;int;-nosmooth;-rgb565;-dither;none;-bmp"
+      testout_422m_ifast_565.bmp testout_422_ifast_opt.jpg
+      ${MD5_BMP_422M_IFAST_565} cjpeg-${libtype}-422-ifast-opt)
+
+    # CC: YCC->RGB565 (dithered)  SAMP: h2v1 merged  IDCT: ifast  ENT: huff
+    add_bittest(djpeg 422m-ifast-565D "-dct;int;-nosmooth;-rgb565;-bmp"
+      testout_422m_ifast_565D.bmp testout_422_ifast_opt.jpg
+      ${MD5_BMP_422M_IFAST_565D} cjpeg-${libtype}-422-ifast-opt)
+  endif()
+
+  # CC: RGB->YCC  SAMP: fullsize/h2v2  FDCT: ifast  ENT: prog huff
+  add_bittest(cjpeg 420-q100-ifast-prog
+    "-sample;2x2;-quality;100;-dct;fast;-scans;${TESTIMAGES}/test.scan"
+    testout_420_q100_ifast_prog.jpg ${TESTIMAGES}/testorig.ppm
+    ${MD5_JPEG_420_IFAST_Q100_PROG})
+
+  # CC: YCC->RGB  SAMP: fullsize/h2v2 fancy  IDCT: ifast  ENT: prog huff
+  add_bittest(djpeg 420-q100-ifast-prog "-dct;fast"
+    testout_420_q100_ifast.ppm testout_420_q100_ifast_prog.jpg
+    ${MD5_PPM_420_Q100_IFAST} cjpeg-${libtype}-420-q100-ifast-prog)
+
+  # CC: YCC->RGB  SAMP: h2v2 merged  IDCT: ifast  ENT: prog huff
+  add_bittest(djpeg 420m-q100-ifast-prog "-dct;fast;-nosmooth"
+    testout_420m_q100_ifast.ppm testout_420_q100_ifast_prog.jpg
+    ${MD5_PPM_420M_Q100_IFAST} cjpeg-${libtype}-420-q100-ifast-prog)
+
+  # CC: RGB->Gray  SAMP: fullsize  FDCT: islow  ENT: huff
+  add_bittest(cjpeg gray-islow "-gray;-dct;int"
+    testout_gray_islow.jpg ${TESTIMAGES}/testorig.ppm
+    ${MD5_JPEG_GRAY_ISLOW})
+
+  # CC: Gray->Gray  SAMP: fullsize  IDCT: islow  ENT: huff
+  add_bittest(djpeg gray-islow "-dct;int"
+    testout_gray_islow.ppm testout_gray_islow.jpg
+    ${MD5_PPM_GRAY_ISLOW} cjpeg-${libtype}-gray-islow)
+
+  # CC: Gray->RGB  SAMP: fullsize  IDCT: islow  ENT: huff
+  add_bittest(djpeg gray-islow-rgb "-dct;int;-rgb"
+    testout_gray_islow_rgb.ppm testout_gray_islow.jpg
+    ${MD5_PPM_GRAY_ISLOW_RGB} cjpeg-${libtype}-gray-islow)
+
+  if(NOT WITH_12BIT)
+    # CC: Gray->RGB565  SAMP: fullsize  IDCT: islow  ENT: huff
+    add_bittest(djpeg gray-islow-565 "-dct;int;-rgb565;-dither;none;-bmp"
+      testout_gray_islow_565.bmp testout_gray_islow.jpg
+      ${MD5_BMP_GRAY_ISLOW_565} cjpeg-${libtype}-gray-islow)
+
+    # CC: Gray->RGB565 (dithered)  SAMP: fullsize  IDCT: islow  ENT: huff
+    add_bittest(djpeg gray-islow-565D "-dct;int;-rgb565;-bmp"
+      testout_gray_islow_565D.bmp testout_gray_islow.jpg
+      ${MD5_BMP_GRAY_ISLOW_565D} cjpeg-${libtype}-gray-islow)
+  endif()
+
+  # CC: RGB->YCC  SAMP: fullsize smooth/h2v2 smooth  FDCT: islow
+  # ENT: 2-pass huff
+  add_bittest(cjpeg 420s-ifast-opt "-sample;2x2;-smooth;1;-dct;int;-opt"
+    testout_420s_ifast_opt.jpg ${TESTIMAGES}/testorig.ppm
+    ${MD5_JPEG_420S_IFAST_OPT})
+
+  if(FLOATTEST)
+    # CC: RGB->YCC  SAMP: fullsize/int  FDCT: float  ENT: prog huff
+    add_bittest(cjpeg 3x2-float-prog "-sample;3x2;-dct;float;-prog"
+      testout_3x2_float_prog.jpg ${TESTIMAGES}/testorig.ppm
+      ${MD5_JPEG_3x2_FLOAT_PROG_${FLOATTEST_UC}})
+
+    # CC: YCC->RGB  SAMP: fullsize/int  IDCT: float  ENT: prog huff
+    add_bittest(djpeg 3x2-float-prog "-dct;float"
+      testout_3x2_float.ppm testout_3x2_float_prog.jpg
+      ${MD5_PPM_3x2_FLOAT_${FLOATTEST_UC}} cjpeg-${libtype}-3x2-float-prog)
+  endif()
+
+    # CC: RGB->YCC  SAMP: fullsize/int  FDCT: ifast  ENT: prog huff
+  add_bittest(cjpeg 3x2-ifast-prog "-sample;3x2;-dct;fast;-prog"
+    testout_3x2_ifast_prog.jpg ${TESTIMAGES}/testorig.ppm
+    ${MD5_JPEG_3x2_IFAST_PROG})
+
+  # CC: YCC->RGB  SAMP: fullsize/int  IDCT: ifast  ENT: prog huff
+  add_bittest(djpeg 3x2-ifast-prog "-dct;fast"
+    testout_3x2_ifast.ppm testout_3x2_ifast_prog.jpg
+    ${MD5_PPM_3x2_IFAST} cjpeg-${libtype}-3x2-ifast-prog)
+
+  if(WITH_ARITH_ENC)
+    # CC: YCC->RGB  SAMP: fullsize/h2v2  FDCT: islow  ENT: arith
+    add_bittest(cjpeg 420-islow-ari "-dct;int;-arithmetic"
+      testout_420_islow_ari.jpg ${TESTIMAGES}/testorig.ppm
+      ${MD5_JPEG_420_ISLOW_ARI})
+
+    add_bittest(jpegtran 420-islow-ari "-arithmetic"
+      testout_420_islow_ari2.jpg ${TESTIMAGES}/testimgint.jpg
+      ${MD5_JPEG_420_ISLOW_ARI})
+
+    # CC: YCC->RGB  SAMP: fullsize  FDCT: islow  ENT: prog arith
+    add_bittest(cjpeg 444-islow-progari
+      "-sample;1x1;-dct;int;-prog;-arithmetic"
+      testout_444_islow_progari.jpg ${TESTIMAGES}/testorig.ppm
+      ${MD5_JPEG_444_ISLOW_PROGARI})
+  endif()
+
+  if(WITH_ARITH_DEC)
+    # CC: RGB->YCC  SAMP: h2v2 merged  IDCT: ifast  ENT: arith
+    add_bittest(djpeg 420m-ifast-ari "-fast;-skip;1,20;-ppm"
+      testout_420m_ifast_ari.ppm ${TESTIMAGES}/testimgari.jpg
+      ${MD5_PPM_420M_IFAST_ARI})
+
+    add_bittest(jpegtran 420-islow ""
+      testout_420_islow.jpg ${TESTIMAGES}/testimgari.jpg
+      ${MD5_JPEG_420_ISLOW})
+  endif()
+
+  # 2/1--   CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 16x16 islow  ENT: huff
+  # 15/8--  CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 15x15 islow  ENT: huff
+  # 13/8--  CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 13x13 islow  ENT: huff
+  # 11/8--  CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 11x11 islow  ENT: huff
+  # 9/8--   CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 9x9 islow  ENT: huff
+  # 7/8--   CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 7x7 islow/14x14 islow
+  #         ENT: huff
+  # 3/4--   CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 6x6 islow/12x12 islow
+  #         ENT: huff
+  # 5/8--   CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 5x5 islow/10x10 islow
+  #         ENT: huff
+  # 1/2--   CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 4x4 islow/8x8 islow
+  #         ENT: huff
+  # 3/8--   CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 3x3 islow/6x6 islow
+  #         ENT: huff
+  # 1/4--   CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 2x2 islow/4x4 islow
+  #         ENT: huff
+  # 1/8--   CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 1x1 islow/2x2 islow
+  #         ENT: huff
+  foreach(scale 2_1 15_8 13_8 11_8 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8)
+    string(REGEX REPLACE "_" "/" scalearg ${scale})
+    add_bittest(djpeg 420m-islow-${scale}
+      "-dct;int;-scale;${scalearg};-nosmooth;-ppm"
+      testout_420m_islow_${scale}.ppm ${TESTIMAGES}/${TESTORIG}
+      ${MD5_PPM_420M_ISLOW_${scale}})
+  endforeach()
+
+  if(NOT WITH_12BIT)
+    # CC: YCC->RGB (dithered)  SAMP: h2v2 fancy  IDCT: islow  ENT: huff
+    add_bittest(djpeg 420-islow-256 "-dct;int;-colors;256;-bmp"
+      testout_420_islow_256.bmp ${TESTIMAGES}/${TESTORIG}
+      ${MD5_BMP_420_ISLOW_256})
+
+    # CC: YCC->RGB565  SAMP: h2v2 fancy  IDCT: islow  ENT: huff
+    add_bittest(djpeg 420-islow-565 "-dct;int;-rgb565;-dither;none;-bmp"
+      testout_420_islow_565.bmp ${TESTIMAGES}/${TESTORIG}
+      ${MD5_BMP_420_ISLOW_565})
+
+    # CC: YCC->RGB565 (dithered)  SAMP: h2v2 fancy  IDCT: islow  ENT: huff
+    add_bittest(djpeg 420-islow-565D "-dct;int;-rgb565;-bmp"
+      testout_420_islow_565D.bmp ${TESTIMAGES}/${TESTORIG}
+      ${MD5_BMP_420_ISLOW_565D})
+
+    # CC: YCC->RGB565  SAMP: h2v2 merged  IDCT: islow  ENT: huff
+    add_bittest(djpeg 420m-islow-565
+      "-dct;int;-nosmooth;-rgb565;-dither;none;-bmp"
+      testout_420m_islow_565.bmp ${TESTIMAGES}/${TESTORIG}
+      ${MD5_BMP_420M_ISLOW_565})
+
+    # CC: YCC->RGB565 (dithered)  SAMP: h2v2 merged  IDCT: islow  ENT: huff
+    add_bittest(djpeg 420m-islow-565D "-dct;int;-nosmooth;-rgb565;-bmp"
+      testout_420m_islow_565D.bmp ${TESTIMAGES}/${TESTORIG}
+      ${MD5_BMP_420M_ISLOW_565D})
+  endif()
+
+  # Partial decode tests.  These tests are designed to cover all of the
+  # possible code paths in jpeg_skip_scanlines().
+
+  # Context rows: Yes  Intra-iMCU row: Yes  iMCU row prefetch: No   ENT: huff
+  add_bittest(djpeg 420-islow-skip15_31 "-dct;int;-skip;15,31;-ppm"
+    testout_420_islow_skip15,31.ppm ${TESTIMAGES}/${TESTORIG}
+    ${MD5_PPM_420_ISLOW_SKIP15_31})
+
+  # Context rows: Yes  Intra-iMCU row: No   iMCU row prefetch: Yes  ENT: arith
+  if(WITH_ARITH_DEC)
+    add_bittest(djpeg 420-islow-ari-skip16_139 "-dct;int;-skip;16,139;-ppm"
+      testout_420_islow_ari_skip16,139.ppm ${TESTIMAGES}/testimgari.jpg
+      ${MD5_PPM_420_ISLOW_ARI_SKIP16_139})
+  endif()
+
+  # Context rows: Yes  Intra-iMCU row: No   iMCU row prefetch: No   ENT: prog huff
+  add_test(cjpeg-${libtype}-420-islow-prog
+    ${CMAKE_CROSSCOMPILING_EMULATOR} cjpeg${suffix} -dct int -prog
+      -outfile testout_420_islow_prog.jpg ${TESTIMAGES}/testorig.ppm)
+  add_bittest(djpeg 420-islow-prog-crop62x62_71_71
+    "-dct;int;-crop;62x62+71+71;-ppm"
+    testout_420_islow_prog_crop62x62,71,71.ppm testout_420_islow_prog.jpg
+    ${MD5_PPM_420_ISLOW_PROG_CROP62x62_71_71} cjpeg-${libtype}-420-islow-prog)
+
+  # Context rows: Yes  Intra-iMCU row: No   iMCU row prefetch: No   ENT: arith
+  if(WITH_ARITH_DEC)
+    add_bittest(djpeg 420-islow-ari-crop53x53_4_4
+      "-dct;int;-crop;53x53+4+4;-ppm"
+      testout_420_islow_ari_crop53x53,4,4.ppm ${TESTIMAGES}/testimgari.jpg
+      ${MD5_PPM_420_ISLOW_ARI_CROP53x53_4_4})
+  endif()
+
+  # Context rows: No   Intra-iMCU row: Yes  ENT: huff
+  add_test(cjpeg-${libtype}-444-islow
+    ${CMAKE_CROSSCOMPILING_EMULATOR} cjpeg${suffix} -dct int -sample 1x1
+      -outfile testout_444_islow.jpg ${TESTIMAGES}/testorig.ppm)
+  add_bittest(djpeg 444-islow-skip1_6 "-dct;int;-skip;1,6;-ppm"
+    testout_444_islow_skip1,6.ppm testout_444_islow.jpg
+    ${MD5_PPM_444_ISLOW_SKIP1_6} cjpeg-${libtype}-444-islow)
+
+  # Context rows: No   Intra-iMCU row: No   ENT: prog huff
+  add_test(cjpeg-${libtype}-444-islow-prog
+    ${CMAKE_CROSSCOMPILING_EMULATOR} cjpeg${suffix} -dct int -prog -sample 1x1
+      -outfile testout_444_islow_prog.jpg ${TESTIMAGES}/testorig.ppm)
+  add_bittest(djpeg 444-islow-prog-crop98x98_13_13
+    "-dct;int;-crop;98x98+13+13;-ppm"
+    testout_444_islow_prog_crop98x98,13,13.ppm testout_444_islow_prog.jpg
+    ${MD5_PPM_444_ISLOW_PROG_CROP98x98_13_13} cjpeg-${libtype}-444-islow-prog)
+
+  # Context rows: No   Intra-iMCU row: No   ENT: arith
+  if(WITH_ARITH_ENC)
+    add_test(cjpeg-${libtype}-444-islow-ari
+      ${CMAKE_CROSSCOMPILING_EMULATOR} cjpeg${suffix} -dct int -arithmetic
+        -sample 1x1 -outfile testout_444_islow_ari.jpg
+        ${TESTIMAGES}/testorig.ppm)
+    if(WITH_ARITH_DEC)
+      add_bittest(djpeg 444-islow-ari-crop37x37_0_0
+        "-dct;int;-crop;37x37+0+0;-ppm"
+        testout_444_islow_ari_crop37x37,0,0.ppm testout_444_islow_ari.jpg
+        ${MD5_PPM_444_ISLOW_ARI_CROP37x37_0_0} cjpeg-${libtype}-444-islow-ari)
+    endif()
+  endif()
+
+  add_bittest(jpegtran crop "-crop;120x90+20+50;-transpose;-perfect"
+    testout_crop.jpg ${TESTIMAGES}/${TESTORIG}
+    ${MD5_JPEG_CROP})
+
+endforeach()
+
+add_custom_target(testclean COMMAND ${CMAKE_COMMAND} -P
+  ${CMAKE_CURRENT_SOURCE_DIR}/cmakescripts/testclean.cmake)
+
+configure_file(croptest.in croptest @ONLY)
+add_custom_target(croptest
+  COMMAND echo croptest
+  COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/croptest)
+
+if(WITH_TURBOJPEG)
+  configure_file(tjbenchtest.in tjbenchtest @ONLY)
+  configure_file(tjexampletest.in tjexampletest @ONLY)
+  if(WIN32)
+    set(BASH bash)
+  endif()
+  if(WITH_JAVA)
+    configure_file(tjbenchtest.java.in tjbenchtest.java @ONLY)
+    configure_file(tjexampletest.java.in tjexampletest.java @ONLY)
+    add_custom_target(tjtest
+      COMMAND echo tjbenchtest
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest
+      COMMAND echo tjbenchtest -alloc
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -alloc
+      COMMAND echo tjbenchtest -yuv
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -yuv
+      COMMAND echo tjbenchtest -yuv -alloc
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -yuv -alloc
+      COMMAND echo tjbenchtest -progressive
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -progressive
+      COMMAND echo tjbenchtest -progressive -yuv
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -progressive -yuv
+      COMMAND echo tjexampletest
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjexampletest
+      COMMAND echo tjbenchtest.java
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest.java
+      COMMAND echo tjbenchtest.java -yuv
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest.java -yuv
+      COMMAND echo tjbenchtest.java -progressive
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest.java -progressive
+      COMMAND echo tjexampletest.java -progressive -yuv
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest.java
+        -progressive -yuv
+      COMMAND echo tjexampletest.java
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjexampletest.java
+      DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest
+        ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest.java
+        ${CMAKE_CURRENT_BINARY_DIR}/tjexampletest)
+  else()
+    add_custom_target(tjtest
+      COMMAND echo tjbenchtest
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest
+      COMMAND echo tjbenchtest -alloc
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -alloc
+      COMMAND echo tjbenchtest -yuv
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -yuv
+      COMMAND echo tjbenchtest -yuv -alloc
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -yuv -alloc
+      COMMAND echo tjbenchtest -progressive
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -progressive
+      COMMAND echo tjbenchtest -progressive -yuv
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest -progressive -yuv
+      COMMAND echo tjexampletest
+      COMMAND ${BASH} ${CMAKE_CURRENT_BINARY_DIR}/tjexampletest
+      DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/tjbenchtest)
+  endif()
+endif()
+
+
+###############################################################################
+# INSTALLATION
+###############################################################################
+
+set(EXE ${CMAKE_EXECUTABLE_SUFFIX})
+
+if(WITH_TURBOJPEG)
+  if(ENABLE_SHARED)
+    install(TARGETS turbojpeg EXPORT ${CMAKE_PROJECT_NAME}Targets
+      INCLUDES DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
+      ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
+      LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
+      RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+    install(TARGETS tjbench
+      RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+    if(NOT CMAKE_VERSION VERSION_LESS "3.1" AND MSVC AND
+      CMAKE_C_LINKER_SUPPORTS_PDB)
+      install(FILES "$<TARGET_PDB_FILE:turbojpeg>"
+        DESTINATION ${CMAKE_INSTALL_BINDIR} OPTIONAL)
+    endif()
+  endif()
+  if(ENABLE_STATIC)
+    install(TARGETS turbojpeg-static EXPORT ${CMAKE_PROJECT_NAME}Targets
+      INCLUDES DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
+      ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR})
+    if(NOT ENABLE_SHARED)
+      if(MSVC_IDE OR XCODE)
+        set(DIR "${CMAKE_CURRENT_BINARY_DIR}/\${CMAKE_INSTALL_CONFIG_NAME}")
+      else()
+        set(DIR ${CMAKE_CURRENT_BINARY_DIR})
+      endif()
+      install(PROGRAMS ${DIR}/tjbench-static${EXE}
+        DESTINATION ${CMAKE_INSTALL_BINDIR} RENAME tjbench${EXE})
+    endif()
+  endif()
+  install(FILES ${CMAKE_CURRENT_SOURCE_DIR}/turbojpeg.h
+    DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
+endif()
+
+if(ENABLE_STATIC)
+  install(TARGETS jpeg-static EXPORT ${CMAKE_PROJECT_NAME}Targets
+    INCLUDES DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
+    ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR})
+  if(NOT ENABLE_SHARED)
+    if(MSVC_IDE OR XCODE)
+      set(DIR "${CMAKE_CURRENT_BINARY_DIR}/\${CMAKE_INSTALL_CONFIG_NAME}")
+    else()
+      set(DIR ${CMAKE_CURRENT_BINARY_DIR})
+    endif()
+    install(PROGRAMS ${DIR}/cjpeg-static${EXE}
+      DESTINATION ${CMAKE_INSTALL_BINDIR} RENAME cjpeg${EXE})
+    install(PROGRAMS ${DIR}/djpeg-static${EXE}
+      DESTINATION ${CMAKE_INSTALL_BINDIR} RENAME djpeg${EXE})
+    install(PROGRAMS ${DIR}/jpegtran-static${EXE}
+      DESTINATION ${CMAKE_INSTALL_BINDIR} RENAME jpegtran${EXE})
+  endif()
+endif()
+
+install(TARGETS rdjpgcom wrjpgcom RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+
+install(FILES ${CMAKE_CURRENT_SOURCE_DIR}/README.ijg
+  ${CMAKE_CURRENT_SOURCE_DIR}/README.md ${CMAKE_CURRENT_SOURCE_DIR}/example.txt
+  ${CMAKE_CURRENT_SOURCE_DIR}/tjexample.c
+  ${CMAKE_CURRENT_SOURCE_DIR}/libjpeg.txt
+  ${CMAKE_CURRENT_SOURCE_DIR}/structure.txt
+  ${CMAKE_CURRENT_SOURCE_DIR}/usage.txt ${CMAKE_CURRENT_SOURCE_DIR}/wizard.txt
+  ${CMAKE_CURRENT_SOURCE_DIR}/LICENSE.md DESTINATION ${CMAKE_INSTALL_DOCDIR})
+if(WITH_JAVA)
+  install(FILES ${CMAKE_CURRENT_SOURCE_DIR}/java/TJExample.java
+    DESTINATION ${CMAKE_INSTALL_DOCDIR})
+endif()
+
+if(UNIX OR MINGW)
+  install(FILES ${CMAKE_CURRENT_SOURCE_DIR}/cjpeg.1
+    ${CMAKE_CURRENT_SOURCE_DIR}/djpeg.1 ${CMAKE_CURRENT_SOURCE_DIR}/jpegtran.1
+    ${CMAKE_CURRENT_SOURCE_DIR}/rdjpgcom.1
+    ${CMAKE_CURRENT_SOURCE_DIR}/wrjpgcom.1
+    DESTINATION ${CMAKE_INSTALL_MANDIR}/man1)
+endif()
+install(FILES ${CMAKE_CURRENT_BINARY_DIR}/pkgscripts/libjpeg.pc
+  ${CMAKE_CURRENT_BINARY_DIR}/pkgscripts/libturbojpeg.pc
+  DESTINATION ${CMAKE_INSTALL_LIBDIR}/pkgconfig)
+install(FILES
+  ${CMAKE_CURRENT_BINARY_DIR}/pkgscripts/${CMAKE_PROJECT_NAME}Config.cmake
+  ${CMAKE_CURRENT_BINARY_DIR}/pkgscripts/${CMAKE_PROJECT_NAME}ConfigVersion.cmake
+  DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${CMAKE_PROJECT_NAME})
+install(EXPORT ${CMAKE_PROJECT_NAME}Targets
+  NAMESPACE ${CMAKE_PROJECT_NAME}::
+  DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${CMAKE_PROJECT_NAME})
+
+install(FILES ${CMAKE_CURRENT_BINARY_DIR}/jconfig.h
+  ${CMAKE_CURRENT_SOURCE_DIR}/jerror.h ${CMAKE_CURRENT_SOURCE_DIR}/jmorecfg.h
+  ${CMAKE_CURRENT_SOURCE_DIR}/jpeglib.h
+  DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
+
+include(cmakescripts/BuildPackages.cmake)
+
+configure_file("${CMAKE_CURRENT_SOURCE_DIR}/cmakescripts/cmake_uninstall.cmake.in"
+  "cmake_uninstall.cmake" IMMEDIATE @ONLY)
+
+add_custom_target(uninstall COMMAND ${CMAKE_COMMAND} -P cmake_uninstall.cmake)

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/ChangeLog.md Added

@@ -0,0 +1,1783 @@
+2.1.1
+=====
+
+### Significant changes relative to 2.1.0
+
+1. Fixed a regression introduced in 2.1.0 that caused build failures with
+non-GCC-compatible compilers for Un*x/Arm platforms.
+
+2. Fixed a regression introduced by 2.1 beta113 that prevented the Arm 32-bit
+(AArch32) Neon SIMD extensions from building unless the C compiler flags
+included `-mfloat-abi=softfp` or `-mfloat-abi=hard`.
+
+3. Fixed an issue in the AArch32 Neon SIMD Huffman encoder whereby reliance on
+undefined C compiler behavior led to crashes ("SIGBUS: illegal alignment") on
+Android systems when running AArch32/Thumb builds of libjpeg-turbo built with
+recent versions of Clang.
+
+4. Added a command-line argument (`-copy icc`) to jpegtran that causes it to
+copy only the ICC profile markers from the source file and discard any other
+metadata.
+
+5. libjpeg-turbo should now build and run on CHERI-enabled architectures, which
+use capability pointers that are larger than the size of `size_t`.
+
+6. Fixed a regression introduced by 2.1 beta15 that caused a segfault in the
+64-bit SSE2 Huffman encoder when attempting to losslessly transform a
+specially-crafted malformed JPEG image.
+
+
+2.1.0
+=====
+
+### Significant changes relative to 2.1 beta1
+
+1. Fixed a regression introduced by 2.1 beta16(b) whereby attempting to
+decompress certain progressive JPEG images with one or more component planes of
+width 8 or less caused a buffer overrun.
+
+2. Fixed a regression introduced by 2.1 beta16(b) whereby attempting to
+decompress a specially-crafted malformed progressive JPEG image caused the
+block smoothing algorithm to read from uninitialized memory.
+
+3. Fixed an issue in the Arm Neon SIMD Huffman encoders that caused the
+encoders to generate incorrect results when using the Clang compiler with
+Visual Studio.
+
+4. Fixed a floating point exception (CVE-2021-20205) that occurred when
+attempting to compress a specially-crafted malformed GIF image with a specified
+image width of 0 using cjpeg.
+
+5. Fixed a regression introduced by 2.0 beta115 whereby attempting to
+generate a progressive JPEG image on an SSE2-capable CPU using a scan script
+containing one or more scans with lengths divisible by 32 and non-zero
+successive approximation low bit positions would, under certain circumstances,
+result in an error ("Missing Huffman code table entry") and an invalid JPEG
+image.
+
+6. Introduced a new flag (`TJFLAG_LIMITSCANS` in the TurboJPEG C API and
+`TJ.FLAG_LIMIT_SCANS` in the TurboJPEG Java API) and a corresponding TJBench
+command-line argument (`-limitscans`) that causes the TurboJPEG decompression
+and transform functions/operations to return/throw an error if a progressive
+JPEG image contains an unreasonably large number of scans.  This allows
+applications that use the TurboJPEG API to guard against an exploit of the
+progressive JPEG format described in the report
+"Two Issues with the JPEG Standard"(https://libjpeg-turbo.org/pmwiki/uploads/About/TwoIssueswiththeJPEGStandard.pdf).
+
+7. The PPM reader now throws an error, rather than segfaulting (due to a buffer
+overrun) or generating incorrect pixels, if an application attempts to use the
+`tjLoadImage()` function to load a 16-bit binary PPM file (a binary PPM file
+with a maximum value greater than 255) into a grayscale image buffer or to load
+a 16-bit binary PGM file into an RGB image buffer.
+
+8. Fixed an issue in the PPM reader that caused incorrect pixels to be
+generated when using the `tjLoadImage()` function to load a 16-bit binary PPM
+file into an extended RGB image buffer.
+
+9. Fixed an issue whereby, if a JPEG buffer was automatically re-allocated by
+one of the TurboJPEG compression or transform functions and an error
+subsequently occurred during compression or transformation, the JPEG buffer
+pointer passed by the application was not updated when the function returned.
+
+
+2.0.90 (2.1 beta1)
+==================
+
+### Significant changes relative to 2.0.6:
+
+1. The build system, x86-64 SIMD extensions, and accelerated Huffman codec now
+support the x32 ABI on Linux, which allows for using x86-64 instructions with
+32-bit pointers.  The x32 ABI is generally enabled by adding `-mx32` to the
+compiler flags.
+
+     Caveats:
+     - CMake 3.9.0 or later is required in order for the build system to
+automatically detect an x32 build.
+     - Java does not support the x32 ABI, and thus the TurboJPEG Java API will
+automatically be disabled with x32 builds.
+
+2. Added Loongson MMI SIMD implementations of the RGB-to-grayscale, 4:2:2 fancy
+chroma upsampling, 4:2:2 and 4:2:0 merged chroma upsampling/color conversion,
+and fast integer DCT/IDCT algorithms.  Relative to libjpeg-turbo 2.0.x, this
+speeds up:
+
+     - the compression of RGB source images into grayscale JPEG images by
+approximately 20%
+     - the decompression of 4:2:2 JPEG images by approximately 40-60% when
+using fancy upsampling
+     - the decompression of 4:2:2 and 4:2:0 JPEG images by approximately
+15-20% when using merged upsampling
+     - the compression of RGB source images by approximately 30-45% when using
+the fast integer DCT
+     - the decompression of JPEG images into RGB destination images by
+approximately 2x when using the fast integer IDCT
+
+    The overall decompression speedup for RGB images is now approximately
+2.3-3.7x (compared to 2-3.5x with libjpeg-turbo 2.0.x.)
+
+3. 32-bit (Armv7 or Armv7s) iOS builds of libjpeg-turbo are no longer
+supported, and the libjpeg-turbo build system can no longer be used to package
+such builds.  32-bit iOS apps cannot run in iOS 11 and later, and the App Store
+no longer allows them.
+
+4. 32-bit (i386) OS X/macOS builds of libjpeg-turbo are no longer supported,
+and the libjpeg-turbo build system can no longer be used to package such
+builds.  32-bit Mac applications cannot run in macOS 10.15 "Catalina" and
+later, and the App Store no longer allows them.
+
+5. The SSE2 (x86 SIMD) and C Huffman encoding algorithms have been
+significantly optimized, resulting in a measured average overall compression
+speedup of 12-28% for 64-bit code and 22-52% for 32-bit code on various Intel
+and AMD CPUs, as well as a measured average overall compression speedup of
+0-23% on platforms that do not have a SIMD-accelerated Huffman encoding
+implementation.
+
+6. The block smoothing algorithm that is applied by default when decompressing
+progressive Huffman-encoded JPEG images has been improved in the following
+ways:
+
+     - The algorithm is now more fault-tolerant.  Previously, if a particular
+scan was incomplete, then the smoothing parameters for the incomplete scan
+would be applied to the entire output image, including the parts of the image
+that were generated by the prior (complete) scan.  Visually, this had the
+effect of removing block smoothing from lower-frequency scans if they were
+followed by an incomplete higher-frequency scan.  libjpeg-turbo now applies
+block smoothing parameters to each iMCU row based on which scan generated the
+pixels in that row, rather than always using the block smoothing parameters for
+the most recent scan.
+     - When applying block smoothing to DC scans, a Gaussian-like kernel with a
+5x5 window is used to reduce the "blocky" appearance.
+
+7. Added SIMD acceleration for progressive Huffman encoding on Arm platforms.
+This speeds up the compression of full-color progressive JPEGs by about 30-40%
+on average (relative to libjpeg-turbo 2.0.x) when using modern Arm CPUs.
+
+8. Added configure-time and run-time auto-detection of Loongson MMI SIMD
+instructions, so that the Loongson MMI SIMD extensions can be included in any
+MIPS64 libjpeg-turbo build.
+
+9. Added fault tolerance features to djpeg and jpegtran, mainly to demonstrate
+methods by which applications can guard against the exploits of the JPEG format
+described in the report
+"Two Issues with the JPEG Standard"(https://libjpeg-turbo.org/pmwiki/uploads/About/TwoIssueswiththeJPEGStandard.pdf).
+
+     - Both programs now accept a `-maxscans` argument, which can be used to
+limit the number of allowable scans in the input file.
+     - Both programs now accept a `-strict` argument, which can be used to
+treat all warnings as fatal.
+
+10. CMake package config files are now included for both the libjpeg and
+TurboJPEG API libraries.  This facilitates using libjpeg-turbo with CMake's
+`find_package()` function.  For example:
+
+        find_package(libjpeg-turbo CONFIG REQUIRED)
+
+        add_executable(libjpeg_program libjpeg_program.c)
+        target_link_libraries(libjpeg_program PUBLIC libjpeg-turbo::jpeg)
+
+        add_executable(libjpeg_program_static libjpeg_program.c)
+        target_link_libraries(libjpeg_program_static PUBLIC
+          libjpeg-turbo::jpeg-static)
+
+        add_executable(turbojpeg_program turbojpeg_program.c)
+        target_link_libraries(turbojpeg_program PUBLIC
+          libjpeg-turbo::turbojpeg)
+
+        add_executable(turbojpeg_program_static turbojpeg_program.c)
+        target_link_libraries(turbojpeg_program_static PUBLIC
+          libjpeg-turbo::turbojpeg-static)
+
+11. Since the Unisys LZW patent has long expired, cjpeg and djpeg can now
+read/write both LZW-compressed and uncompressed GIF files (feature ported from
+jpeg-6a and jpeg-9d.)
+
+12. jpegtran now includes the `-wipe` and `-drop` options from jpeg-9a and
+jpeg-9d, as well as the ability to expand the image size using the `-crop`
+option.  Refer to jpegtran.1 or usage.txt for more details.
+
+13. Added a complete intrinsics implementation of the Arm Neon SIMD extensions,
+thus providing SIMD acceleration on Arm platforms for all of the algorithms
+that are SIMD-accelerated on x86 platforms.  This new implementation is
+significantly faster in some cases than the old GAS implementation--
+depending on the algorithms used, the type of CPU core, and the compiler.  GCC,
+as of this writing, does not provide a full or optimal set of Neon intrinsics,
+so for performance reasons, the default when building libjpeg-turbo with GCC is
+to continue using the GAS implementation of the following algorithms:
+
+     - 32-bit RGB-to-YCbCr color conversion
+     - 32-bit fast and accurate inverse DCT
+     - 64-bit RGB-to-YCbCr and YCbCr-to-RGB color conversion
+     - 64-bit accurate forward and inverse DCT
+     - 64-bit Huffman encoding
+
+    A new CMake variable (`NEON_INTRINSICS`) can be used to override this
+default.
+
+    Since the new intrinsics implementation includes SIMD acceleration
+for merged upsampling/color conversion, 1.5.15 is no longer necessary and has
+been reverted.
+
+14. The Arm Neon SIMD extensions can now be built using Visual Studio.
+
+15. The build system can now be used to generate a universal x86-64 + Armv8
+libjpeg-turbo SDK package for both iOS and macOS.
+
+
+2.0.6
+=====
+
+### Significant changes relative to 2.0.5:
+
+1. Fixed "using JNI after critical get" errors that occurred on Android
+platforms when using any of the YUV encoding/compression/decompression/decoding
+methods in the TurboJPEG Java API.
+
+2. Fixed or worked around multiple issues with `jpeg_skip_scanlines()`:
+
+     - Fixed segfaults or "Corrupt JPEG data: premature end of data segment"
+errors in `jpeg_skip_scanlines()` that occurred when decompressing 4:2:2 or
+4:2:0 JPEG images using merged (non-fancy) upsampling/color conversion (that
+is, when setting `cinfo.do_fancy_upsampling` to `FALSE`.)  2.0.06 was a
+similar fix, but it did not cover all cases.
+     - `jpeg_skip_scanlines()` now throws an error if two-pass color
+quantization is enabled.  Two-pass color quantization never worked properly
+with `jpeg_skip_scanlines()`, and the issues could not readily be fixed.
+     - Fixed an issue whereby `jpeg_skip_scanlines()` always returned 0 when
+skipping past the end of an image.
+
+3. The Arm 64-bit (Armv8) Neon SIMD extensions can now be built using MinGW
+toolchains targetting Arm64 (AArch64) Windows binaries.
+
+4. Fixed unexpected visual artifacts that occurred when using
+`jpeg_crop_scanline()` and interblock smoothing while decompressing only the DC
+scan of a progressive JPEG image.
+
+5. Fixed an issue whereby libjpeg-turbo would not build if 12-bit-per-component
+JPEG support (`WITH_12BIT`) was enabled along with libjpeg v7 or libjpeg v8
+API/ABI emulation (`WITH_JPEG7` or `WITH_JPEG8`.)
+
+
+2.0.5
+=====
+
+### Significant changes relative to 2.0.4:
+
+1. Worked around issues in the MIPS DSPr2 SIMD extensions that caused failures
+in the libjpeg-turbo regression tests.  Specifically, the
+`jsimd_h2v1_downsample_dspr2()` and `jsimd_h2v2_downsample_dspr2()` functions
+in the MIPS DSPr2 SIMD extensions are now disabled until/unless they can be
+fixed, and other functions that are incompatible with big endian MIPS CPUs are
+disabled when building libjpeg-turbo for such CPUs.
+
+2. Fixed an oversight in the `TJCompressor.compress(int)` method in the
+TurboJPEG Java API that caused an error ("java.lang.IllegalStateException: No
+source image is associated with this instance") when attempting to use that
+method to compress a YUV image.
+
+3. Fixed an issue (CVE-2020-13790) in the PPM reader that caused a buffer
+overrun in cjpeg, TJBench, or the `tjLoadImage()` function if one of the values
+in a binary PPM/PGM input file exceeded the maximum value defined in the file's
+header and that maximum value was less than 255.  libjpeg-turbo 1.5.0 already
+included a similar fix for binary PPM/PGM files with maximum values greater
+than 255.
+
+4. The TurboJPEG API library's global error handler, which is used in functions
+such as `tjBufSize()` and `tjLoadImage()` that do not require a TurboJPEG
+instance handle, is now thread-safe on platforms that support thread-local
+storage.
+
+
+2.0.4
+=====
+
+### Significant changes relative to 2.0.3:
+
+1. Fixed a regression in the Windows packaging system (introduced by
+2.0 beta12) whereby, if both the 64-bit libjpeg-turbo SDK for GCC and the
+64-bit libjpeg-turbo SDK for Visual C++ were installed on the same system, only
+one of them could be uninstalled.
+
+2. Fixed a signed integer overflow and subsequent segfault that occurred when
+attempting to decompress images with more than 715827882 pixels using the
+64-bit C version of TJBench.
+
+3. Fixed out-of-bounds write in `tjDecompressToYUV2()` and
+`tjDecompressToYUVPlanes()` (sometimes manifesting as a double free) that
+occurred when attempting to decompress grayscale JPEG images that were
+compressed with a sampling factor other than 1 (for instance, with
+`cjpeg -grayscale -sample 2x2`).
+
+4. Fixed a regression introduced by 2.0.25 that caused the TurboJPEG API to
+incorrectly identify some JPEG images with unusual sampling factors as 4:4:4
+JPEG images.  This was known to cause a buffer overflow when attempting to
+decompress some such images using `tjDecompressToYUV2()` or
+`tjDecompressToYUVPlanes()`.
+
+5. Fixed an issue (CVE-2020-17541), detected by ASan, whereby attempting to
+losslessly transform a specially-crafted malformed JPEG image containing an
+extremely-high-frequency coefficient block (junk image data that could never be
+generated by a legitimate JPEG compressor) could cause the Huffman encoder's
+local buffer to be overrun. (Refer to 1.4.09 and 1.4beta115.)  Given that
+the buffer overrun was fully contained within the stack and did not cause a
+segfault or other user-visible errant behavior, and given that the lossless
+transformer (unlike the decompressor) is not generally exposed to arbitrary
+data exploits, this issue did not likely pose a security risk.
+
+6. The Arm 64-bit (Armv8) Neon SIMD assembly code now stores constants in a
+separate read-only data section rather than in the text section, to support
+execute-only memory layouts.
+
+
+2.0.3
+=====
+
+### Significant changes relative to 2.0.2:
+
+1. Fixed "using JNI after critical get" errors that occurred on Android
+platforms when passing invalid arguments to certain methods in the TurboJPEG
+Java API.
+
+2. Fixed a regression in the SIMD feature detection code, introduced by
+the AVX2 SIMD extensions (2.0 beta11), that was known to cause an illegal
+instruction exception, in rare cases, on CPUs that lack support for CPUID leaf
+07H (or on which the maximum CPUID leaf has been limited by way of a BIOS
+setting.)
+
+3. The 4:4:0 (h1v2) fancy (smooth) chroma upsampling algorithm in the
+decompressor now uses a similar bias pattern to that of the 4:2:2 (h2v1) fancy
+chroma upsampling algorithm, rounding up or down the upsampled result for
+alternate pixels rather than always rounding down.  This ensures that,
+regardless of whether a 4:2:2 JPEG image is rotated or transposed prior to
+decompression (in the frequency domain) or after decompression (in the spatial
+domain), the final image will be similar.
+
+4. Fixed an integer overflow and subsequent segfault that occurred when
+attempting to compress or decompress images with more than 1 billion pixels
+using the TurboJPEG API.
+
+5. Fixed a regression introduced by 2.0 beta115 whereby attempting to
+generate a progressive JPEG image on an SSE2-capable CPU using a scan script
+containing one or more scans with lengths divisible by 16 would result in an
+error ("Missing Huffman code table entry") and an invalid JPEG image.
+
+6. Fixed an issue whereby `tjDecodeYUV()` and `tjDecodeYUVPlanes()` would throw
+an error ("Invalid progressive parameters") or a warning ("Inconsistent
+progression sequence") if passed a TurboJPEG instance that was previously used
+to decompress a progressive JPEG image.
+
+
+2.0.2
+=====
+
+### Significant changes relative to 2.0.1:
+
+1. Fixed a regression introduced by 2.0.15 that prevented a runtime search
+path (rpath) from being embedded in the libjpeg-turbo shared libraries and
+executables for macOS and iOS.  This caused a fatal error of the form
+"dyld: Library not loaded" when attempting to use one of the executables,
+unless `DYLD_LIBRARY_PATH` was explicitly set to the location of the
+libjpeg-turbo shared libraries.
+
+2. Fixed an integer overflow and subsequent segfault (CVE-2018-20330) that
+occurred when attempting to load a BMP file with more than 1 billion pixels
+using the `tjLoadImage()` function.
+
+3. Fixed a buffer overrun (CVE-2018-19664) that occurred when attempting to
+decompress a specially-crafted malformed JPEG image to a 256-color BMP using
+djpeg.
+
+4. Fixed a floating point exception that occurred when attempting to
+decompress a specially-crafted malformed JPEG image with a specified image
+width or height of 0 using the C version of TJBench.
+
+5. The TurboJPEG API will now decompress 4:4:4 JPEG images with 2x1, 1x2, 3x1,
+or 1x3 luminance and chrominance sampling factors.  This is a non-standard way
+of specifying 1x subsampling (normally 4:4:4 JPEGs have 1x1 luminance and
+chrominance sampling factors), but the JPEG format and the libjpeg API both
+allow it.
+
+6. Fixed a regression introduced by 2.0 beta17 that caused djpeg to generate
+incorrect PPM images when used with the `-colors` option.
+
+7. Fixed an issue whereby a static build of libjpeg-turbo (a build in which
+`ENABLE_SHARED` is `0`) could not be installed using the Visual Studio IDE.
+
+8. Fixed a severe performance issue in the Loongson MMI SIMD extensions that
+occurred when compressing RGB images whose image rows were not 64-bit-aligned.
+
+
+2.0.1
+=====
+
+### Significant changes relative to 2.0.0:
+
+1. Fixed a regression introduced with the new CMake-based Un*x build system,
+whereby jconfig.h could cause compiler warnings of the form
+`"HAVE_*_H" redefined` if it was included by downstream Autotools-based
+projects that used `AC_CHECK_HEADERS()` to check for the existence of locale.h,
+stddef.h, or stdlib.h.
+
+2. The `jsimd_quantize_float_dspr2()` and `jsimd_convsamp_float_dspr2()`
+functions in the MIPS DSPr2 SIMD extensions are now disabled at compile time
+if the soft float ABI is enabled.  Those functions use instructions that are
+incompatible with the soft float ABI.
+
+3. Fixed a regression in the SIMD feature detection code, introduced by
+the AVX2 SIMD extensions (2.0 beta11), that caused libjpeg-turbo to crash on
+Windows 7 if Service Pack 1 was not installed.
+
+4. Fixed out-of-bounds read in cjpeg that occurred when attempting to compress
+a specially-crafted malformed color-index (8-bit-per-sample) Targa file in
+which some of the samples (color indices) exceeded the bounds of the Targa
+file's color table.
+
+5. Fixed an issue whereby installing a fully static build of libjpeg-turbo
+(a build in which `CFLAGS` contains `-static` and `ENABLE_SHARED` is `0`) would
+fail with "No valid ELF RPATH or RUNPATH entry exists in the file."
+
+
+2.0.0
+=====
+
+### Significant changes relative to 2.0 beta1:
+
+1. The TurboJPEG API can now decompress CMYK JPEG images that have subsampled M
+and Y components (not to be confused with YCCK JPEG images, in which the C/M/Y
+components have been transformed into luma and chroma.)   Previously, an error
+was generated ("Could not determine subsampling type for JPEG image") when such
+an image was passed to `tjDecompressHeader3()`, `tjTransform()`,
+`tjDecompressToYUVPlanes()`, `tjDecompressToYUV2()`, or the equivalent Java
+methods.
+
+2. Fixed an issue (CVE-2018-11813) whereby a specially-crafted malformed input
+file (specifically, a file with a valid Targa header but incomplete pixel data)
+would cause cjpeg to generate a JPEG file that was potentially thousands of
+times larger than the input file.  The Targa reader in cjpeg was not properly
+detecting that the end of the input file had been reached prematurely, so after
+all valid pixels had been read from the input, the reader injected dummy pixels
+with values of 255 into the JPEG compressor until the number of pixels
+specified in the Targa header had been compressed.  The Targa reader in cjpeg
+now behaves like the PPM reader and aborts compression if the end of the input
+file is reached prematurely.  Because this issue only affected cjpeg and not
+the underlying library, and because it did not involve any out-of-bounds reads
+or other exploitable behaviors, it was not believed to represent a security
+threat.
+
+3. Fixed an issue whereby the `tjLoadImage()` and `tjSaveImage()` functions
+would produce a "Bogus message code" error message if the underlying bitmap and
+PPM readers/writers threw an error that was specific to the readers/writers
+(as opposed to a general libjpeg API error.)
+
+4. Fixed an issue (CVE-2018-1152) whereby a specially-crafted malformed BMP
+file, one in which the header specified an image width of 1073741824 pixels,
+would trigger a floating point exception (division by zero) in the
+`tjLoadImage()` function when attempting to load the BMP file into a
+4-component image buffer.
+
+5. Fixed an issue whereby certain combinations of calls to
+`jpeg_skip_scanlines()` and `jpeg_read_scanlines()` could trigger an infinite
+loop when decompressing progressive JPEG images that use vertical chroma
+subsampling (for instance, 4:2:0 or 4:4:0.)
+
+6. Fixed a segfault in `jpeg_skip_scanlines()` that occurred when decompressing
+a 4:2:2 or 4:2:0 JPEG image using the merged (non-fancy) upsampling algorithms
+(that is, when setting `cinfo.do_fancy_upsampling` to `FALSE`.)
+
+7. The new CMake-based build system will now disable the MIPS DSPr2 SIMD
+extensions if it detects that the compiler does not support DSPr2 instructions.
+
+8. Fixed out-of-bounds read in cjpeg (CVE-2018-14498) that occurred when
+attempting to compress a specially-crafted malformed color-index
+(8-bit-per-sample) BMP file in which some of the samples (color indices)
+exceeded the bounds of the BMP file's color table.
+
+9. Fixed a signed integer overflow in the progressive Huffman decoder, detected
+by the Clang and GCC undefined behavior sanitizers, that could be triggered by
+attempting to decompress a specially-crafted malformed JPEG image.  This issue
+did not pose a security threat, but removing the warning made it easier to
+detect actual security issues, should they arise in the future.
+
+
+1.5.90 (2.0 beta1)
+==================
+
+### Significant changes relative to 1.5.3:
+
+1. Added AVX2 SIMD implementations of the colorspace conversion, chroma
+downsampling and upsampling, integer quantization and sample conversion, and
+accurate integer DCT/IDCT algorithms.  When using the accurate integer DCT/IDCT
+algorithms on AVX2-equipped CPUs, the compression of RGB images is
+approximately 13-36% (avg. 22%) faster (relative to libjpeg-turbo 1.5.x) with
+64-bit code and 11-21% (avg. 17%) faster with 32-bit code, and the
+decompression of RGB images is approximately 9-35% (avg. 17%) faster with
+64-bit code and 7-17% (avg. 12%) faster with 32-bit code.  (As tested on a
+3 GHz Intel Core i7.  Actual mileage may vary.)
+
+2. Overhauled the build system to use CMake on all platforms, and removed the
+autotools-based build system.  This decision resulted from extensive
+discussions within the libjpeg-turbo community.  libjpeg-turbo traditionally
+used CMake only for Windows builds, but there was an increasing amount of
+demand to extend CMake support to other platforms.  However, because of the
+unique nature of our code base (the need to support different assemblers on
+each platform, the need for Java support, etc.), providing dual build systems
+as other OSS imaging libraries do (including libpng and libtiff) would have
+created a maintenance burden.  The use of CMake greatly simplifies some aspects
+of our build system, owing to CMake's built-in support for various assemblers,
+Java, and unit testing, as well as generally fewer quirks that have to be
+worked around in order to implement our packaging system.  Eliminating
+autotools puts our project slightly at odds with the traditional practices of
+the OSS community, since most "system libraries" tend to be built with
+autotools, but it is believed that the benefits of this move outweigh the
+risks.  In addition to providing a unified build environment, switching to
+CMake allows for the use of various build tools and IDEs that aren't supported
+under autotools, including XCode, Ninja, and Eclipse.  It also eliminates the
+need to install autotools via MacPorts/Homebrew on OS X and allows
+libjpeg-turbo to be configured without the use of a terminal/command prompt.
+Extensive testing was conducted to ensure that all features provided by the
+autotools-based build system are provided by the new build system.
+
+3. The libjpeg API in this version of libjpeg-turbo now includes two additional
+functions, `jpeg_read_icc_profile()` and `jpeg_write_icc_profile()`, that can
+be used to extract ICC profile data from a JPEG file while decompressing or to
+embed ICC profile data in a JPEG file while compressing or transforming.  This
+eliminates the need for downstream projects, such as color management libraries
+and browsers, to include their own glueware for accomplishing this.
+
+4. Improved error handling in the TurboJPEG API library:
+
+     - Introduced a new function (`tjGetErrorStr2()`) in the TurboJPEG C API
+that allows compression/decompression/transform error messages to be retrieved
+in a thread-safe manner.  Retrieving error messages from global functions, such
+as `tjInitCompress()` or `tjBufSize()`, is still thread-unsafe, but since those
+functions will only throw errors if passed an invalid argument or if a memory
+allocation failure occurs, thread safety is not as much of a concern.
+     - Introduced a new function (`tjGetErrorCode()`) in the TurboJPEG C API
+and a new method (`TJException.getErrorCode()`) in the TurboJPEG Java API that
+can be used to determine the severity of the last
+compression/decompression/transform error.  This allows applications to
+choose whether to ignore warnings (non-fatal errors) from the underlying
+libjpeg API or to treat them as fatal.
+     - Introduced a new flag (`TJFLAG_STOPONWARNING` in the TurboJPEG C API and
+`TJ.FLAG_STOPONWARNING` in the TurboJPEG Java API) that causes the library to
+immediately halt a compression/decompression/transform operation if it
+encounters a warning from the underlying libjpeg API (the default behavior is
+to allow the operation to complete unless a fatal error is encountered.)
+
+5. Introduced a new flag in the TurboJPEG C and Java APIs (`TJFLAG_PROGRESSIVE`
+and `TJ.FLAG_PROGRESSIVE`, respectively) that causes the library to use
+progressive entropy coding in JPEG images generated by compression and
+transform operations.  Additionally, a new transform option
+(`TJXOPT_PROGRESSIVE` in the C API and `TJTransform.OPT_PROGRESSIVE` in the
+Java API) has been introduced, allowing progressive entropy coding to be
+enabled for selected transforms in a multi-transform operation.
+
+6. Introduced a new transform option in the TurboJPEG API (`TJXOPT_COPYNONE` in
+the C API and `TJTransform.OPT_COPYNONE` in the Java API) that allows the
+copying of markers (including EXIF and ICC profile data) to be disabled for a
+particular transform.
+
+7. Added two functions to the TurboJPEG C API (`tjLoadImage()` and
+`tjSaveImage()`) that can be used to load/save a BMP or PPM/PGM image to/from a
+memory buffer with a specified pixel format and layout.  These functions
+replace the project-private (and slow) bmp API, which was previously used by
+TJBench, and they also provide a convenient way for first-time users of
+libjpeg-turbo to quickly develop a complete JPEG compression/decompression
+program.
+
+8. The TurboJPEG C API now includes a new convenience array (`tjAlphaOffset`)
+that contains the alpha component index for each pixel format (or -1 if the
+pixel format lacks an alpha component.)  The TurboJPEG Java API now includes a
+new method (`TJ.getAlphaOffset()`) that returns the same value.  In addition,
+the `tjRedOffset`, `tjGreenOffset`, and `tjBlueOffset` arrays-- and the
+corresponding `TJ.getRedOffset()`, `TJ.getGreenOffset()`, and
+`TJ.getBlueOffset()` methods-- now return -1 for `TJPF_GRAY`/`TJ.PF_GRAY`
+rather than 0.  This allows programs to easily determine whether a pixel format
+has red, green, blue, and alpha components.
+
+9. Added a new example (tjexample.c) that demonstrates the basic usage of the
+TurboJPEG C API.  This example mirrors the functionality of TJExample.java.
+Both files are now included in the libjpeg-turbo documentation.
+
+10. Fixed two signed integer overflows in the arithmetic decoder, detected by
+the Clang undefined behavior sanitizer, that could be triggered by attempting
+to decompress a specially-crafted malformed JPEG image.  These issues did not
+pose a security threat, but removing the warnings makes it easier to detect
+actual security issues, should they arise in the future.
+
+11. Fixed a bug in the merged 4:2:0 upsampling/dithered RGB565 color conversion
+algorithm that caused incorrect dithering in the output image.  This algorithm
+now produces bitwise-identical results to the unmerged algorithms.
+
+12. The SIMD function symbols for x86-64/ELF, MIPS/ELF, macOS/x86-64 (if
+libjpeg-turbo is built with YASM), and iOS/Arm64 builds are now private.
+This prevents those symbols from being exposed in applications or shared
+libraries that link statically with libjpeg-turbo.
+
+13. Added Loongson MMI SIMD implementations of the RGB-to-YCbCr and
+YCbCr-to-RGB colorspace conversion, 4:2:0 chroma downsampling, 4:2:0 fancy
+chroma upsampling, integer quantization, and accurate integer DCT/IDCT
+algorithms.  When using the accurate integer DCT/IDCT, this speeds up the
+compression of RGB images by approximately 70-100% and the decompression of RGB
+images by approximately 2-3.5x.
+
+14. Fixed a build error when building with older MinGW releases (regression
+caused by 1.5.17.)
+
+15. Added SIMD acceleration for progressive Huffman encoding on SSE2-capable
+x86 and x86-64 platforms.  This speeds up the compression of full-color
+progressive JPEGs by about 85-90% on average (relative to libjpeg-turbo 1.5.x)
+when using modern Intel and AMD CPUs.
+
+
+1.5.3
+=====
+
+### Significant changes relative to 1.5.2:
+
+1. Fixed a NullPointerException in the TurboJPEG Java wrapper that occurred
+when using the YUVImage constructor that creates an instance backed by separate
+image planes and allocates memory for the image planes.
+
+2. Fixed an issue whereby the Java version of TJUnitTest would fail when
+testing BufferedImage encoding/decoding on big endian systems.
+
+3. Fixed a segfault in djpeg that would occur if an output format other than
+PPM/PGM was selected along with the `-crop` option.  The `-crop` option now
+works with the GIF and Targa formats as well (unfortunately, it cannot be made
+to work with the BMP and RLE formats due to the fact that those output engines
+write scanlines in bottom-up order.)  djpeg will now exit gracefully if an
+output format other than PPM/PGM, GIF, or Targa is selected along with the
+`-crop` option.
+
+4. Fixed an issue (CVE-2017-15232) whereby `jpeg_skip_scanlines()` would
+segfault if color quantization was enabled.
+
+5. TJBench (both C and Java versions) will now display usage information if any
+command-line argument is unrecognized.  This prevents the program from silently
+ignoring typos.
+
+6. Fixed an access violation in tjbench.exe (Windows) that occurred when the
+program was used to decompress an existing JPEG image.
+
+7. Fixed an ArrayIndexOutOfBoundsException in the TJExample Java program that
+occurred when attempting to decompress a JPEG image that had been compressed
+with 4:1:1 chrominance subsampling.
+
+8. Fixed an issue whereby, when using `jpeg_skip_scanlines()` to skip to the
+end of a single-scan (non-progressive) image, subsequent calls to
+`jpeg_consume_input()` would return `JPEG_SUSPENDED` rather than
+`JPEG_REACHED_EOI`.
+
+9. `jpeg_crop_scanline()` now works correctly when decompressing grayscale JPEG
+images that were compressed with a sampling factor other than 1 (for instance,
+with `cjpeg -grayscale -sample 2x2`).
+
+
+1.5.2
+=====
+
+### Significant changes relative to 1.5.1:
+
+1. Fixed a regression introduced by 1.5.17 that prevented libjpeg-turbo from
+building with Android NDK platforms prior to android-21 (5.0).
+
+2. Fixed a regression introduced by 1.5.11 that prevented the MIPS DSPR2 SIMD
+code in libjpeg-turbo from building.
+
+3. Fixed a regression introduced by 1.5 beta111 that prevented the Java
+version of TJBench from outputting any reference images (the `-nowrite` switch
+was accidentally enabled by default.)
+
+4. libjpeg-turbo should now build and run with full AltiVec SIMD acceleration
+on PowerPC-based AmigaOS 4 and OpenBSD systems.
+
+5. Fixed build and runtime errors on Windows that occurred when building
+libjpeg-turbo with libjpeg v7 API/ABI emulation and the in-memory
+source/destination managers.  Due to an oversight, the `jpeg_skip_scanlines()`
+and `jpeg_crop_scanline()` functions were not being included in jpeg7.dll when
+libjpeg-turbo was built with `-DWITH_JPEG7=1` and `-DWITH_MEMSRCDST=1`.
+
+6. Fixed "Bogus virtual array access" error that occurred when using the
+lossless crop feature in jpegtran or the TurboJPEG API, if libjpeg-turbo was
+built with libjpeg v7 API/ABI emulation.  This was apparently a long-standing
+bug that has existed since the introduction of libjpeg v7/v8 API/ABI emulation
+in libjpeg-turbo v1.1.
+
+7. The lossless transform features in jpegtran and the TurboJPEG API will now
+always attempt to adjust the EXIF image width and height tags if the image size
+changed as a result of the transform.  This behavior has always existed when
+using libjpeg v8 API/ABI emulation.  It was supposed to be available with
+libjpeg v7 API/ABI emulation as well but did not work properly due to a bug.
+Furthermore, there was never any good reason not to enable it with libjpeg v6b
+API/ABI emulation, since the behavior is entirely internal.  Note that
+`-copy all` must be passed to jpegtran in order to transfer the EXIF tags from
+the source image to the destination image.
+
+8. Fixed several memory leaks in the TurboJPEG API library that could occur
+if the library was built with certain compilers and optimization levels
+(known to occur with GCC 4.x and clang with `-O1` and higher but not with
+GCC 5.x or 6.x) and one of the underlying libjpeg API functions threw an error
+after a TurboJPEG API function allocated a local buffer.
+
+9. The libjpeg-turbo memory manager will now honor the `max_memory_to_use`
+structure member in jpeg\_memory\_mgr, which can be set to the maximum amount
+of memory (in bytes) that libjpeg-turbo should use during decompression or
+multi-pass (including progressive) compression.  This limit can also be set
+using the `JPEGMEM` environment variable or using the `-maxmemory` switch in
+cjpeg/djpeg/jpegtran (refer to the respective man pages for more details.)
+This has been a documented feature of libjpeg since v5, but the
+`malloc()`/`free()` implementation of the memory manager (jmemnobs.c) never
+implemented the feature.  Restricting libjpeg-turbo's memory usage is useful
+for two reasons:  it allows testers to more easily work around the 2 GB limit
+in libFuzzer, and it allows developers of security-sensitive applications to
+more easily defend against one of the progressive JPEG exploits (LJT-01-004)
+identified in
+this report(http://www.libjpeg-turbo.org/pmwiki/uploads/About/TwoIssueswiththeJPEGStandard.pdf).
+
+10. TJBench will now run each benchmark for 1 second prior to starting the
+timer, in order to improve the consistency of the results.  Furthermore, the
+`-warmup` option is now used to specify the amount of warmup time rather than
+the number of warmup iterations.
+
+11. Fixed an error (`short jump is out of range`) that occurred when assembling
+the 32-bit x86 SIMD extensions with NASM versions prior to 2.04.  This was a
+regression introduced by 1.5 beta112.
+
+
+1.5.1
+=====
+
+### Significant changes relative to 1.5.0:
+
+1. Previously, the undocumented `JSIMD_FORCE*` environment variables could be
+used to force-enable a particular SIMD instruction set if multiple instruction
+sets were available on a particular platform.  On x86 platforms, where CPU
+feature detection is bulletproof and multiple SIMD instruction sets are
+available, it makes sense for those environment variables to allow forcing the
+use of an instruction set only if that instruction set is available.  However,
+since the ARM implementations of libjpeg-turbo can only use one SIMD
+instruction set, and since their feature detection code is less bulletproof
+(parsing /proc/cpuinfo), it makes sense for the `JSIMD_FORCENEON` environment
+variable to bypass the feature detection code and really force the use of NEON
+instructions.  A new environment variable (`JSIMD_FORCEDSPR2`) was introduced
+in the MIPS implementation for the same reasons, and the existing
+`JSIMD_FORCENONE` environment variable was extended to that implementation.
+These environment variables provide a workaround for those attempting to test
+ARM and MIPS builds of libjpeg-turbo in QEMU, which passes through
+/proc/cpuinfo from the host system.
+
+2. libjpeg-turbo previously assumed that AltiVec instructions were always
+available on PowerPC platforms, which led to "illegal instruction" errors when
+running on PowerPC chips that lack AltiVec support (such as the older 7xx/G3
+and newer e5500 series.)  libjpeg-turbo now examines /proc/cpuinfo on
+Linux/Android systems and enables AltiVec instructions only if the CPU supports
+them.  It also now provides two environment variables, `JSIMD_FORCEALTIVEC` and
+`JSIMD_FORCENONE`, to force-enable and force-disable AltiVec instructions in
+environments where /proc/cpuinfo is an unreliable means of CPU feature
+detection (such as when running in QEMU.)  On OS X, libjpeg-turbo continues to
+assume that AltiVec support is always available, which means that libjpeg-turbo
+cannot be used with G3 Macs unless you set the environment variable
+`JSIMD_FORCENONE` to `1`.
+
+3. Fixed an issue whereby 64-bit ARM (AArch64) builds of libjpeg-turbo would
+crash when built with recent releases of the Clang/LLVM compiler.  This was
+caused by an ABI conformance issue in some of libjpeg-turbo's 64-bit NEON SIMD
+routines.  Those routines were incorrectly using 64-bit instructions to
+transfer a 32-bit JDIMENSION argument, whereas the ABI allows the upper
+(unused) 32 bits of a 32-bit argument's register to be undefined.  The new
+Clang/LLVM optimizer uses load combining to transfer multiple adjacent 32-bit
+structure members into a single 64-bit register, and this exposed the ABI
+conformance issue.
+
+4. Fancy upsampling is now supported when decompressing JPEG images that use
+4:4:0 (h1v2) chroma subsampling.  These images are generated when losslessly
+rotating or transposing JPEG images that use 4:2:2 (h2v1) chroma subsampling.
+The h1v2 fancy upsampling algorithm is not currently SIMD-accelerated.
+
+5. If merged upsampling isn't SIMD-accelerated but YCbCr-to-RGB conversion is,
+then libjpeg-turbo will now disable merged upsampling when decompressing YCbCr
+JPEG images into RGB or extended RGB output images.  This significantly speeds
+up the decompression of 4:2:0 and 4:2:2 JPEGs on ARM platforms if fancy
+upsampling is not used (for example, if the `-nosmooth` option to djpeg is
+specified.)
+
+6. The TurboJPEG API will now decompress 4:2:2 and 4:4:0 JPEG images with
+2x2 luminance sampling factors and 2x1 or 1x2 chrominance sampling factors.
+This is a non-standard way of specifying 2x subsampling (normally 4:2:2 JPEGs
+have 2x1 luminance and 1x1 chrominance sampling factors, and 4:4:0 JPEGs have
+1x2 luminance and 1x1 chrominance sampling factors), but the JPEG format and
+the libjpeg API both allow it.
+
+7. Fixed an unsigned integer overflow in the libjpeg memory manager, detected
+by the Clang undefined behavior sanitizer, that could be triggered by
+attempting to decompress a specially-crafted malformed JPEG image.  This issue
+affected only 32-bit code and did not pose a security threat, but removing the
+warning makes it easier to detect actual security issues, should they arise in
+the future.
+
+8. Fixed additional negative left shifts and other issues reported by the GCC
+and Clang undefined behavior sanitizers when attempting to decompress
+specially-crafted malformed JPEG images.  None of these issues posed a security
+threat, but removing the warnings makes it easier to detect actual security
+issues, should they arise in the future.
+
+9. Fixed an out-of-bounds array reference, introduced by 1.4.902 (partial
+image decompression) and detected by the Clang undefined behavior sanitizer,
+that could be triggered by a specially-crafted malformed JPEG image with more
+than four components.  Because the out-of-bounds reference was still within the
+same structure, it was not known to pose a security threat, but removing the
+warning makes it easier to detect actual security issues, should they arise in
+the future.
+
+10. Fixed another ABI conformance issue in the 64-bit ARM (AArch64) NEON SIMD
+code.  Some of the routines were incorrectly reading and storing data below the
+stack pointer, which caused segfaults in certain applications under specific
+circumstances.
+
+
+1.5.0
+=====
+
+### Significant changes relative to 1.5 beta1:
+
+1. Fixed an issue whereby a malformed motion-JPEG frame could cause the "fast
+path" of libjpeg-turbo's Huffman decoder to read from uninitialized memory.
+
+2. Added libjpeg-turbo version and build information to the global string table
+of the libjpeg and TurboJPEG API libraries.  This is a common practice in other
+infrastructure libraries, such as OpenSSL and libpng, because it makes it easy
+to examine an application binary and determine which version of the library the
+application was linked against.
+
+3. Fixed a couple of issues in the PPM reader that would cause buffer overruns
+in cjpeg if one of the values in a binary PPM/PGM input file exceeded the
+maximum value defined in the file's header and that maximum value was greater
+than 255.  libjpeg-turbo 1.4.2 already included a similar fix for ASCII PPM/PGM
+files.  Note that these issues were not security bugs, since they were confined
+to the cjpeg program and did not affect any of the libjpeg-turbo libraries.
+
+4. Fixed an issue whereby attempting to decompress a JPEG file with a corrupt
+header using the `tjDecompressToYUV2()` function would cause the function to
+abort without returning an error and, under certain circumstances, corrupt the
+stack.  This only occurred if `tjDecompressToYUV2()` was called prior to
+calling `tjDecompressHeader3()`, or if the return value from
+`tjDecompressHeader3()` was ignored (both cases represent incorrect usage of
+the TurboJPEG API.)
+
+5. Fixed an issue in the ARM 32-bit SIMD-accelerated Huffman encoder that
+prevented the code from assembling properly with clang.
+
+6. The `jpeg_stdio_src()`, `jpeg_mem_src()`, `jpeg_stdio_dest()`, and
+`jpeg_mem_dest()` functions in the libjpeg API will now throw an error if a
+source/destination manager has already been assigned to the compress or
+decompress object by a different function or by the calling program.  This
+prevents these functions from attempting to reuse a source/destination manager
+structure that was allocated elsewhere, because there is no way to ensure that
+it would be big enough to accommodate the new source/destination manager.
+
+
+1.4.90 (1.5 beta1)
+==================
+
+### Significant changes relative to 1.4.2:
+
+1. Added full SIMD acceleration for PowerPC platforms using AltiVec VMX
+(128-bit SIMD) instructions.  Although the performance of libjpeg-turbo on
+PowerPC was already good, due to the increased number of registers available
+to the compiler vs. x86, it was still possible to speed up compression by about
+3-4x and decompression by about 2-2.5x (relative to libjpeg v6b) through the
+use of AltiVec instructions.
+
+2. Added two new libjpeg API functions (`jpeg_skip_scanlines()` and
+`jpeg_crop_scanline()`) that can be used to partially decode a JPEG image.  See
+libjpeg.txt(libjpeg.txt) for more details.
+
+3. The TJCompressor and TJDecompressor classes in the TurboJPEG Java API now
+implement the Closeable interface, so those classes can be used with a
+try-with-resources statement.
+
+4. The TurboJPEG Java classes now throw unchecked idiomatic exceptions
+(IllegalArgumentException, IllegalStateException) for unrecoverable errors
+caused by incorrect API usage, and those classes throw a new checked exception
+type (TJException) for errors that are passed through from the C library.
+
+5. Source buffers for the TurboJPEG C API functions, as well as the
+`jpeg_mem_src()` function in the libjpeg API, are now declared as const
+pointers.  This facilitates passing read-only buffers to those functions and
+ensures the caller that the source buffer will not be modified.  This should
+not create any backward API or ABI incompatibilities with prior libjpeg-turbo
+releases.
+
+6. The MIPS DSPr2 SIMD code can now be compiled to support either FR=0 or FR=1
+FPUs.
+
+7. Fixed additional negative left shifts and other issues reported by the GCC
+and Clang undefined behavior sanitizers.  Most of these issues affected only
+32-bit code, and none of them was known to pose a security threat, but removing
+the warnings makes it easier to detect actual security issues, should they
+arise in the future.
+
+8. Removed the unnecessary `.arch` directive from the ARM64 NEON SIMD code.
+This directive was preventing the code from assembling using the clang
+integrated assembler.
+
+9. Fixed a regression caused by 1.4.16 that prevented 32-bit and 64-bit
+libjpeg-turbo RPMs from being installed simultaneously on recent Red Hat/Fedora
+distributions.  This was due to the addition of a macro in jconfig.h that
+allows the Huffman codec to determine the word size at compile time.  Since
+that macro differs between 32-bit and 64-bit builds, this caused a conflict
+between the i386 and x86_64 RPMs (any differing files, other than executables,
+are not allowed when 32-bit and 64-bit RPMs are installed simultaneously.)
+Since the macro is used only internally, it has been moved into jconfigint.h.
+
+10. The x86-64 SIMD code can now be disabled at run time by setting the
+`JSIMD_FORCENONE` environment variable to `1` (the other SIMD implementations
+already had this capability.)
+
+11. Added a new command-line argument to TJBench (`-nowrite`) that prevents the
+benchmark from outputting any images.  This removes any potential operating
+system overhead that might be caused by lazy writes to disk and thus improves
+the consistency of the performance measurements.
+
+12. Added SIMD acceleration for Huffman encoding on SSE2-capable x86 and x86-64
+platforms.  This speeds up the compression of full-color JPEGs by about 10-15%
+on average (relative to libjpeg-turbo 1.4.x) when using modern Intel and AMD
+CPUs.  Additionally, this works around an issue in the clang optimizer that
+prevents it (as of this writing) from achieving the same performance as GCC
+when compiling the C version of the Huffman encoder
+(<https://llvm.org/bugs/show_bug.cgi?id=16035>).  For the purposes of
+benchmarking or regression testing, SIMD-accelerated Huffman encoding can be
+disabled by setting the `JSIMD_NOHUFFENC` environment variable to `1`.
+
+13. Added ARM 64-bit (ARMv8) NEON SIMD implementations of the commonly-used
+compression algorithms (including the accurate integer forward DCT and h2v2 &
+h2v1 downsampling algorithms, which are not accelerated in the 32-bit NEON
+implementation.)  This speeds up the compression of full-color JPEGs by about
+75% on average on a Cavium ThunderX processor and by about 2-2.5x on average on
+Cortex-A53 and Cortex-A57 cores.
+
+14. Added SIMD acceleration for Huffman encoding on NEON-capable ARM 32-bit
+and 64-bit platforms.
+
+    For 32-bit code, this speeds up the compression of full-color JPEGs by
+about 30% on average on a typical iOS device (iPhone 4S, Cortex-A9) and by
+about 6-7% on average on a typical Android device (Nexus 5X, Cortex-A53 and
+Cortex-A57), relative to libjpeg-turbo 1.4.x.  Note that the larger speedup
+under iOS is due to the fact that iOS builds use LLVM, which does not optimize
+the C Huffman encoder as well as GCC does.
+
+    For 64-bit code, NEON-accelerated Huffman encoding speeds up the
+compression of full-color JPEGs by about 40% on average on a typical iOS device
+(iPhone 5S, Apple A7) and by about 7-8% on average on a typical Android device
+(Nexus 5X, Cortex-A53 and Cortex-A57), in addition to the speedup described in
+13 above.
+
+    For the purposes of benchmarking or regression testing, SIMD-accelerated
+Huffman encoding can be disabled by setting the `JSIMD_NOHUFFENC` environment
+variable to `1`.
+
+15. pkg-config (.pc) scripts are now included for both the libjpeg and
+TurboJPEG API libraries on Un*x systems.  Note that if a project's build system
+relies on these scripts, then it will not be possible to build that project
+with libjpeg or with a prior version of libjpeg-turbo.
+
+16. Optimized the ARM 64-bit (ARMv8) NEON SIMD decompression routines to
+improve performance on CPUs with in-order pipelines.  This speeds up the
+decompression of full-color JPEGs by nearly 2x on average on a Cavium ThunderX
+processor and by about 15% on average on a Cortex-A53 core.
+
+17. Fixed an issue in the accelerated Huffman decoder that could have caused
+the decoder to read past the end of the input buffer when a malformed,
+specially-crafted JPEG image was being decompressed.  In prior versions of
+libjpeg-turbo, the accelerated Huffman decoder was invoked (in most cases) only
+if there were > 128 bytes of data in the input buffer.  However, it is possible
+to construct a JPEG image in which a single Huffman block is over 430 bytes
+long, so this version of libjpeg-turbo activates the accelerated Huffman
+decoder only if there are > 512 bytes of data in the input buffer.
+
+18. Fixed a memory leak in tjunittest encountered when running the program
+with the `-yuv` option.
+
+
+1.4.2
+=====
+
+### Significant changes relative to 1.4.1:
+
+1. Fixed an issue whereby cjpeg would segfault if a Windows bitmap with a
+negative width or height was used as an input image (Windows bitmaps can have
+a negative height if they are stored in top-down order, but such files are
+rare and not supported by libjpeg-turbo.)
+
+2. Fixed an issue whereby, under certain circumstances, libjpeg-turbo would
+incorrectly encode certain JPEG images when quality=100 and the fast integer
+forward DCT were used.  This was known to cause `make test` to fail when the
+library was built with `-march=haswell` on x86 systems.
+
+3. Fixed an issue whereby libjpeg-turbo would crash when built with the latest
+& greatest development version of the Clang/LLVM compiler.  This was caused by
+an x86-64 ABI conformance issue in some of libjpeg-turbo's 64-bit SSE2 SIMD
+routines.  Those routines were incorrectly using a 64-bit `mov` instruction to
+transfer a 32-bit JDIMENSION argument, whereas the x86-64 ABI allows the upper
+(unused) 32 bits of a 32-bit argument's register to be undefined.  The new
+Clang/LLVM optimizer uses load combining to transfer multiple adjacent 32-bit
+structure members into a single 64-bit register, and this exposed the ABI
+conformance issue.
+
+4. Fixed a bug in the MIPS DSPr2 4:2:0 "plain" (non-fancy and non-merged)
+upsampling routine that caused a buffer overflow (and subsequent segfault) when
+decompressing a 4:2:0 JPEG image whose scaled output width was less than 16
+pixels.  The "plain" upsampling routines are normally only used when
+decompressing a non-YCbCr JPEG image, but they are also used when decompressing
+a JPEG image whose scaled output height is 1.
+
+5. Fixed various negative left shifts and other issues reported by the GCC and
+Clang undefined behavior sanitizers.  None of these was known to pose a
+security threat, but removing the warnings makes it easier to detect actual
+security issues, should they arise in the future.
+
+
+1.4.1
+=====
+
+### Significant changes relative to 1.4.0:
+
+1. tjbench now properly handles CMYK/YCCK JPEG files.  Passing an argument of
+`-cmyk` (instead of, for instance, `-rgb`) will cause tjbench to internally
+convert the source bitmap to CMYK prior to compression, to generate YCCK JPEG
+files, and to internally convert the decompressed CMYK pixels back to RGB after
+decompression (the latter is done automatically if a CMYK or YCCK JPEG is
+passed to tjbench as a source image.)  The CMYK<->RGB conversion operation is
+not benchmarked.  NOTE: The quick & dirty CMYK<->RGB conversions that tjbench
+uses are suitable for testing only.  Proper conversion between CMYK and RGB
+requires a color management system.
+
+2. `make test` now performs additional bitwise regression tests using tjbench,
+mainly for the purpose of testing compression from/decompression to a subregion
+of a larger image buffer.
+
+3. `make test` no longer tests the regression of the floating point DCT/IDCT
+by default, since the results of those tests can vary if the algorithms in
+question are not implemented using SIMD instructions on a particular platform.
+See the comments in Makefile.am(Makefile.am) for information on how to
+re-enable the tests and to specify an expected result for them based on the
+particulars of your platform.
+
+4. The NULL color conversion routines have been significantly optimized,
+which speeds up the compression of RGB and CMYK JPEGs by 5-20% when using
+64-bit code and 0-3% when using 32-bit code, and the decompression of those
+images by 10-30% when using 64-bit code and 3-12% when using 32-bit code.
+
+5. Fixed an "illegal instruction" error that occurred when djpeg from a
+SIMD-enabled libjpeg-turbo MIPS build was executed with the `-nosmooth` option
+on a MIPS machine that lacked DSPr2 support.  The MIPS SIMD routines for h2v1
+and h2v2 merged upsampling were not properly checking for the existence of
+DSPr2.
+
+6. Performance has been improved significantly on 64-bit non-Linux and
+non-Windows platforms (generally 10-20% faster compression and 5-10% faster
+decompression.)  Due to an oversight, the 64-bit version of the accelerated
+Huffman codec was not being compiled in when libjpeg-turbo was built on
+platforms other than Windows or Linux.  Oops.
+
+7. Fixed an extremely rare bug in the Huffman encoder that caused 64-bit
+builds of libjpeg-turbo to incorrectly encode a few specific test images when
+quality=98, an optimized Huffman table, and the accurate integer forward DCT
+were used.
+
+8. The Windows (CMake) build system now supports building only static or only
+shared libraries.  This is accomplished by adding either `-DENABLE_STATIC=0` or
+`-DENABLE_SHARED=0` to the CMake command line.
+
+9. TurboJPEG API functions will now return an error code if a warning is
+triggered in the underlying libjpeg API.  For instance, if a JPEG file is
+corrupt, the TurboJPEG decompression functions will attempt to decompress
+as much of the image as possible, but those functions will now return -1 to
+indicate that the decompression was not entirely successful.
+
+10. Fixed a bug in the MIPS DSPr2 4:2:2 fancy upsampling routine that caused a
+buffer overflow (and subsequent segfault) when decompressing a 4:2:2 JPEG image
+in which the right-most MCU was 5 or 6 pixels wide.
+
+
+1.4.0
+=====
+
+### Significant changes relative to 1.4 beta1:
+
+1. Fixed a build issue on OS X PowerPC platforms (md5cmp failed to build
+because OS X does not provide the `le32toh()` and `htole32()` functions.)
+
+2. The non-SIMD RGB565 color conversion code did not work correctly on big
+endian machines.  This has been fixed.
+
+3. Fixed an issue in `tjPlaneSizeYUV()` whereby it would erroneously return 1
+instead of -1 if `componentID` was > 0 and `subsamp` was `TJSAMP_GRAY`.
+
+3. Fixed an issue in `tjBufSizeYUV2()` whereby it would erroneously return 0
+instead of -1 if `width` was < 1.
+
+5. The Huffman encoder now uses `clz` and `bsr` instructions for bit counting
+on ARM64 platforms (see 1.4 beta15.)
+
+6. The `close()` method in the TJCompressor and TJDecompressor Java classes is
+now idempotent.  Previously, that method would call the native `tjDestroy()`
+function even if the TurboJPEG instance had already been destroyed.  This
+caused an exception to be thrown during finalization, if the `close()` method
+had already been called.  The exception was caught, but it was still an
+expensive operation.
+
+7. The TurboJPEG API previously generated an error (`Could not determine
+subsampling type for JPEG image`) when attempting to decompress grayscale JPEG
+images that were compressed with a sampling factor other than 1 (for instance,
+with `cjpeg -grayscale -sample 2x2`).  Subsampling technically has no meaning
+with grayscale JPEGs, and thus the horizontal and vertical sampling factors
+for such images are ignored by the decompressor.  However, the TurboJPEG API
+was being too rigid and was expecting the sampling factors to be equal to 1
+before it treated the image as a grayscale JPEG.
+
+8. cjpeg, djpeg, and jpegtran now accept an argument of `-version`, which will
+print the library version and exit.
+
+9. Referring to 1.4 beta115, another extremely rare circumstance was
+discovered under which the Huffman encoder's local buffer can be overrun
+when a buffered destination manager is being used and an
+extremely-high-frequency block (basically junk image data) is being encoded.
+Even though the Huffman local buffer was increased from 128 bytes to 136 bytes
+to address the previous issue, the new issue caused even the larger buffer to
+be overrun.  Further analysis reveals that, in the absolute worst case (such as
+setting alternating AC coefficients to 32767 and -32768 in the JPEG scanning
+order), the Huffman encoder can produce encoded blocks that approach double the
+size of the unencoded blocks.  Thus, the Huffman local buffer was increased to
+256 bytes, which should prevent any such issue from re-occurring in the future.
+
+10. The new `tjPlaneSizeYUV()`, `tjPlaneWidth()`, and `tjPlaneHeight()`
+functions were not actually usable on any platform except OS X and Windows,
+because those functions were not included in the libturbojpeg mapfile.  This
+has been fixed.
+
+11. Restored the `JPP()`, `JMETHOD()`, and `FAR` macros in the libjpeg-turbo
+header files.  The `JPP()` and `JMETHOD()` macros were originally implemented
+in libjpeg as a way of supporting non-ANSI compilers that lacked support for
+prototype parameters.  libjpeg-turbo has never supported such compilers, but
+some software packages still use the macros to define their own prototypes.
+Similarly, libjpeg-turbo has never supported MS-DOS and other platforms that
+have far symbols, but some software packages still use the `FAR` macro.  A
+pretty good argument can be made that this is a bad practice on the part of the
+software in question, but since this affects more than one package, it's just
+easier to fix it here.
+
+12. Fixed issues that were preventing the ARM 64-bit SIMD code from compiling
+for iOS, and included an ARMv8 architecture in all of the binaries installed by
+the "official" libjpeg-turbo SDK for OS X.
+
+
+1.3.90 (1.4 beta1)
+==================
+
+### Significant changes relative to 1.3.1:
+
+1. New features in the TurboJPEG API:
+
+     - YUV planar images can now be generated with an arbitrary line padding
+(previously only 4-byte padding, which was compatible with X Video, was
+supported.)
+     - The decompress-to-YUV function has been extended to support image
+scaling.
+     - JPEG images can now be compressed from YUV planar source images.
+     - YUV planar images can now be decoded into RGB or grayscale images.
+     - 4:1:1 subsampling is now supported.  This is mainly included for
+compatibility, since 4:1:1 is not fully accelerated in libjpeg-turbo and has no
+significant advantages relative to 4:2:0.
+     - CMYK images are now supported.  This feature allows CMYK source images
+to be compressed to YCCK JPEGs and YCCK or CMYK JPEGs to be decompressed to
+CMYK destination images.  Conversion between CMYK/YCCK and RGB or YUV images is
+not supported.  Such conversion requires a color management system and is thus
+out of scope for a codec library.
+     - The handling of YUV images in the Java API has been significantly
+refactored and should now be much more intuitive.
+     - The Java API now supports encoding a YUV image from an arbitrary
+position in a large image buffer.
+     - All of the YUV functions now have a corresponding function that operates
+on separate image planes instead of a unified image buffer.  This allows for
+compressing/decoding from or decompressing/encoding to a subregion of a larger
+YUV image.  It also allows for handling YUV formats that swap the order of the
+U and V planes.
+
+2. Added SIMD acceleration for DSPr2-capable MIPS platforms.  This speeds up
+the compression of full-color JPEGs by 70-80% on such platforms and
+decompression by 25-35%.
+
+3. If an application attempts to decompress a Huffman-coded JPEG image whose
+header does not contain Huffman tables, libjpeg-turbo will now insert the
+default Huffman tables.  In order to save space, many motion JPEG video frames
+are encoded without the default Huffman tables, so these frames can now be
+successfully decompressed by libjpeg-turbo without additional work on the part
+of the application.  An application can still override the Huffman tables, for
+instance to re-use tables from a previous frame of the same video.
+
+4. The Mac packaging system now uses pkgbuild and productbuild rather than
+PackageMaker (which is obsolete and no longer supported.)  This means that
+OS X 10.6 "Snow Leopard" or later must be used when packaging libjpeg-turbo,
+although the packages produced can be installed on OS X 10.5 "Leopard" or
+later.  OS X 10.4 "Tiger" is no longer supported.
+
+5. The Huffman encoder now uses `clz` and `bsr` instructions for bit counting
+on ARM platforms rather than a lookup table.  This reduces the memory footprint
+by 64k, which may be important for some mobile applications.  Out of four
+Android devices that were tested, two demonstrated a small overall performance
+loss (~3-4% on average) with ARMv6 code and a small gain (also ~3-4%) with
+ARMv7 code when enabling this new feature, but the other two devices
+demonstrated a significant overall performance gain with both ARMv6 and ARMv7
+code (~10-20%) when enabling the feature.  Actual mileage may vary.
+
+6. Worked around an issue with Visual C++ 2010 and later that caused incorrect
+pixels to be generated when decompressing a JPEG image to a 256-color bitmap,
+if compiler optimization was enabled when libjpeg-turbo was built.  This caused
+the regression tests to fail when doing a release build under Visual C++ 2010
+and later.
+
+7. Improved the accuracy and performance of the non-SIMD implementation of the
+floating point inverse DCT (using code borrowed from libjpeg v8a and later.)
+The accuracy of this implementation now matches the accuracy of the SSE/SSE2
+implementation.  Note, however, that the floating point DCT/IDCT algorithms are
+mainly a legacy feature.  They generally do not produce significantly better
+accuracy than the accurate integer DCT/IDCT algorithms, and they are quite a
+bit slower.
+
+8. Added a new output colorspace (`JCS_RGB565`) to the libjpeg API that allows
+for decompressing JPEG images into RGB565 (16-bit) pixels.  If dithering is not
+used, then this code path is SIMD-accelerated on ARM platforms.
+
+9. Numerous obsolete features, such as support for non-ANSI compilers and
+support for the MS-DOS memory model, were removed from the libjpeg code,
+greatly improving its readability and making it easier to maintain and extend.
+
+10. Fixed a segfault that occurred when calling `output_message()` with
+`msg_code` set to `JMSG_COPYRIGHT`.
+
+11. Fixed an issue whereby wrjpgcom was allowing comments longer than 65k
+characters to be passed on the command line, which was causing it to generate
+incorrect JPEG files.
+
+12. Fixed a bug in the build system that was causing the Windows version of
+wrjpgcom to be built using the rdjpgcom source code.
+
+13. Restored 12-bit-per-component JPEG support.  A 12-bit version of
+libjpeg-turbo can now be built by passing an argument of `--with-12bit` to
+configure (Unix) or `-DWITH_12BIT=1` to cmake (Windows.)  12-bit JPEG support
+is included only for convenience.  Enabling this feature disables all of the
+performance features in libjpeg-turbo, as well as arithmetic coding and the
+TurboJPEG API.  The resulting library still contains the other libjpeg-turbo
+features (such as the colorspace extensions), but in general, it performs no
+faster than libjpeg v6b.
+
+14. Added ARM 64-bit SIMD acceleration for the YCC-to-RGB color conversion
+and IDCT algorithms (both are used during JPEG decompression.)  For unknown
+reasons (probably related to clang), this code cannot currently be compiled for
+iOS.
+
+15. Fixed an extremely rare bug (CVE-2014-9092) that could cause the Huffman
+encoder's local buffer to overrun when a very high-frequency MCU is compressed
+using quality 100 and no subsampling, and when the JPEG output buffer is being
+dynamically resized by the destination manager.  This issue was so rare that,
+even with a test program specifically designed to make the bug occur (by
+injecting random high-frequency YUV data into the compressor), it was
+reproducible only once in about every 25 million iterations.
+
+16. Fixed an oversight in the TurboJPEG C wrapper:  if any of the JPEG
+compression functions was called repeatedly with the same
+automatically-allocated destination buffer, then TurboJPEG would erroneously
+assume that the `jpegSize` parameter was equal to the size of the buffer, when
+in fact that parameter was probably equal to the size of the most recently
+compressed JPEG image.  If the size of the previous JPEG image was not as large
+as the current JPEG image, then TurboJPEG would unnecessarily reallocate the
+destination buffer.
+
+
+1.3.1
+=====
+
+### Significant changes relative to 1.3.0:
+
+1. On Un*x systems, `make install` now installs the libjpeg-turbo libraries
+into /opt/libjpeg-turbo/lib32 by default on any 32-bit system, not just x86,
+and into /opt/libjpeg-turbo/lib64 by default on any 64-bit system, not just
+x86-64.  You can override this by overriding either the `prefix` or `libdir`
+configure variables.
+
+2. The Windows installer now places a copy of the TurboJPEG DLLs in the same
+directory as the rest of the libjpeg-turbo binaries.  This was mainly done
+to support TurboVNC 1.3, which bundles the DLLs in its Windows installation.
+When using a 32-bit version of CMake on 64-bit Windows, it is impossible to
+access the c:\WINDOWS\system32 directory, which made it impossible for the
+TurboVNC build scripts to bundle the 64-bit TurboJPEG DLL.
+
+3. Fixed a bug whereby attempting to encode a progressive JPEG with arithmetic
+entropy coding (by passing arguments of `-progressive -arithmetic` to cjpeg or
+jpegtran, for instance) would result in an error, `Requested feature was
+omitted at compile time`.
+
+4. Fixed a couple of issues (CVE-2013-6629 and CVE-2013-6630) whereby malformed
+JPEG images would cause libjpeg-turbo to use uninitialized memory during
+decompression.
+
+5. Fixed an error (`Buffer passed to JPEG library is too small`) that occurred
+when calling the TurboJPEG YUV encoding function with a very small (< 5x5)
+source image, and added a unit test to check for this error.
+
+6. The Java classes should now build properly under Visual Studio 2010 and
+later.
+
+7. Fixed an issue that prevented SRPMs generated using the in-tree packaging
+tools from being rebuilt on certain newer Linux distributions.
+
+8. Numerous minor fixes to eliminate compilation and build/packaging system
+warnings, fix cosmetic issues, improve documentation clarity, and other general
+source cleanup.
+
+
+1.3.0
+=====
+
+### Significant changes relative to 1.3 beta1:
+
+1. `make test` now works properly on FreeBSD, and it no longer requires the
+md5sum executable to be present on other Un*x platforms.
+
+2. Overhauled the packaging system:
+
+     - To avoid conflict with vendor-supplied libjpeg-turbo packages, the
+official RPMs and DEBs for libjpeg-turbo have been renamed to
+"libjpeg-turbo-official".
+     - The TurboJPEG libraries are now located under /opt/libjpeg-turbo in the
+official Linux and Mac packages, to avoid conflict with vendor-supplied
+packages and also to streamline the packaging system.
+     - Release packages are now created with the directory structure defined
+by the configure variables `prefix`, `bindir`, `libdir`, etc. (Un\*x) or by the
+`CMAKE_INSTALL_PREFIX` variable (Windows.)  The exception is that the docs are
+always located under the system default documentation directory on Un\*x and
+Mac systems, and on Windows, the TurboJPEG DLL is always located in the Windows
+system directory.
+     - To avoid confusion, official libjpeg-turbo packages on Linux/Unix
+platforms (except for Mac) will always install the 32-bit libraries in
+/opt/libjpeg-turbo/lib32 and the 64-bit libraries in /opt/libjpeg-turbo/lib64.
+     - Fixed an issue whereby, in some cases, the libjpeg-turbo executables on
+Un*x systems were not properly linking with the shared libraries installed by
+the same package.
+     - Fixed an issue whereby building the "installer" target on Windows when
+`WITH_JAVA=1` would fail if the TurboJPEG JAR had not been previously built.
+     - Building the "install" target on Windows now installs files into the
+same places that the installer does.
+
+3. Fixed a Huffman encoder bug that prevented I/O suspension from working
+properly.
+
+
+1.2.90 (1.3 beta1)
+==================
+
+### Significant changes relative to 1.2.1:
+
+1. Added support for additional scaling factors (3/8, 5/8, 3/4, 7/8, 9/8, 5/4,
+11/8, 3/2, 13/8, 7/4, 15/8, and 2) when decompressing.  Note that the IDCT will
+not be SIMD-accelerated when using any of these new scaling factors.
+
+2. The TurboJPEG dynamic library is now versioned.  It was not strictly
+necessary to do so, because TurboJPEG uses versioned symbols, and if a function
+changes in an ABI-incompatible way, that function is renamed and a legacy
+function is provided to maintain backward compatibility.  However, certain
+Linux distro maintainers have a policy against accepting any library that isn't
+versioned.
+
+3. Extended the TurboJPEG Java API so that it can be used to compress a JPEG
+image from and decompress a JPEG image to an arbitrary position in a large
+image buffer.
+
+4. The `tjDecompressToYUV()` function now supports the `TJFLAG_FASTDCT` flag.
+
+5. The 32-bit supplementary package for amd64 Debian systems now provides
+symlinks in /usr/lib/i386-linux-gnu for the TurboJPEG libraries in /usr/lib32.
+This allows those libraries to be used on MultiArch-compatible systems (such as
+Ubuntu 11 and later) without setting the linker path.
+
+6. The TurboJPEG Java wrapper should now find the JNI library on Mac systems
+without having to pass `-Djava.library.path=/usr/lib` to java.
+
+7. TJBench has been ported to Java to provide a convenient way of validating
+the performance of the TurboJPEG Java API.  It can be run with
+`java -cp turbojpeg.jar TJBench`.
+
+8. cjpeg can now be used to generate JPEG files with the RGB colorspace
+(feature ported from jpeg-8d.)
+
+9. The width and height in the `-crop` argument passed to jpegtran can now be
+suffixed with `f` to indicate that, when the upper left corner of the cropping
+region is automatically moved to the nearest iMCU boundary, the bottom right
+corner should be moved by the same amount.  In other words, this feature causes
+jpegtran to strictly honor the specified width/height rather than the specified
+bottom right corner (feature ported from jpeg-8d.)
+
+10. JPEG files using the RGB colorspace can now be decompressed into grayscale
+images (feature ported from jpeg-8d.)
+
+11. Fixed a regression caused by 1.2.17 whereby the build would fail with
+multiple "Mismatch in operand sizes" errors when attempting to build the x86
+SIMD code with NASM 0.98.
+
+12. The in-memory source/destination managers (`jpeg_mem_src()` and
+`jpeg_mem_dest()`) are now included by default when building libjpeg-turbo with
+libjpeg v6b or v7 emulation, so that programs can take advantage of these
+functions without requiring the use of the backward-incompatible libjpeg v8
+ABI.  The "age number" of the libjpeg-turbo library on Un*x systems has been
+incremented by 1 to reflect this.  You can disable this feature with a
+configure/CMake switch in order to retain strict API/ABI compatibility with the
+libjpeg v6b or v7 API/ABI (or with previous versions of libjpeg-turbo.)  See
+README.md(README.md) for more details.
+
+13. Added ARMv7s architecture to libjpeg.a and libturbojpeg.a in the official
+libjpeg-turbo binary package for OS X, so that those libraries can be used to
+build applications that leverage the faster CPUs in the iPhone 5 and iPad 4.
+
+
+1.2.1
+=====
+
+### Significant changes relative to 1.2.0:
+
+1. Creating or decoding a JPEG file that uses the RGB colorspace should now
+properly work when the input or output colorspace is one of the libjpeg-turbo
+colorspace extensions.
+
+2. When libjpeg-turbo was built without SIMD support and merged (non-fancy)
+upsampling was used along with an alpha-enabled colorspace during
+decompression, the unused byte of the decompressed pixels was not being set to
+0xFF.  This has been fixed.  TJUnitTest has also been extended to test for the
+correct behavior of the colorspace extensions when merged upsampling is used.
+
+3. Fixed a bug whereby the libjpeg-turbo SSE2 SIMD code would not preserve the
+upper 64 bits of xmm6 and xmm7 on Win64 platforms, which violated the Win64
+calling conventions.
+
+4. Fixed a regression (CVE-2012-2806) caused by 1.2.06 whereby decompressing
+corrupt JPEG images (specifically, images in which the component count was
+erroneously set to a large value) would cause libjpeg-turbo to segfault.
+
+5. Worked around a severe performance issue with "Bobcat" (AMD Embedded APU)
+processors.  The `MASKMOVDQU` instruction, which was used by the libjpeg-turbo
+SSE2 SIMD code, is apparently implemented in microcode on AMD processors, and
+it is painfully slow on Bobcat processors in particular.  Eliminating the use
+of this instruction improved performance by an order of magnitude on Bobcat
+processors and by a small amount (typically 5%) on AMD desktop processors.
+
+6. Added SIMD acceleration for performing 4:2:2 upsampling on NEON-capable ARM
+platforms.  This speeds up the decompression of 4:2:2 JPEGs by 20-25% on such
+platforms.
+
+7. Fixed a regression caused by 1.2.02 whereby, on Linux/x86 platforms
+running the 32-bit SSE2 SIMD code in libjpeg-turbo, decompressing a 4:2:0 or
+4:2:2 JPEG image into a 32-bit (RGBX, BGRX, etc.) buffer without using fancy
+upsampling would produce several incorrect columns of pixels at the right-hand
+side of the output image if each row in the output image was not evenly
+divisible by 16 bytes.
+
+8. Fixed an issue whereby attempting to build the SIMD extensions with Xcode
+4.3 on OS X platforms would cause NASM to return numerous errors of the form
+"'%define' expects a macro identifier".
+
+9. Added flags to the TurboJPEG API that allow the caller to force the use of
+either the fast or the accurate DCT/IDCT algorithms in the underlying codec.
+
+
+1.2.0
+=====
+
+### Significant changes relative to 1.2 beta1:
+
+1. Fixed build issue with YASM on Unix systems (the libjpeg-turbo build system
+was not adding the current directory to the assembler include path, so YASM
+was not able to find jsimdcfg.inc.)
+
+2. Fixed out-of-bounds read in SSE2 SIMD code that occurred when decompressing
+a JPEG image to a bitmap buffer whose size was not a multiple of 16 bytes.
+This was more of an annoyance than an actual bug, since it did not cause any
+actual run-time problems, but the issue showed up when running libjpeg-turbo in
+valgrind.  See <http://crbug.com/72399> for more information.
+
+3. Added a compile-time macro (`LIBJPEG_TURBO_VERSION`) that can be used to
+check the version of libjpeg-turbo against which an application was compiled.
+
+4. Added new RGBA/BGRA/ABGR/ARGB colorspace extension constants (libjpeg API)
+and pixel formats (TurboJPEG API), which allow applications to specify that,
+when decompressing to a 4-component RGB buffer, the unused byte should be set
+to 0xFF so that it can be interpreted as an opaque alpha channel.
+
+5. Fixed regression issue whereby DevIL failed to build against libjpeg-turbo
+because libjpeg-turbo's distributed version of jconfig.h contained an `INLINE`
+macro, which conflicted with a similar macro in DevIL.  This macro is used only
+internally when building libjpeg-turbo, so it was moved into config.h.
+
+6. libjpeg-turbo will now correctly decompress erroneous CMYK/YCCK JPEGs whose
+K component is assigned a component ID of 1 instead of 4.  Although these files
+are in violation of the spec, other JPEG implementations handle them
+correctly.
+
+7. Added ARMv6 and ARMv7 architectures to libjpeg.a and libturbojpeg.a in
+the official libjpeg-turbo binary package for OS X, so that those libraries can
+be used to build both OS X and iOS applications.
+
+
+1.1.90 (1.2 beta1)
+==================
+
+### Significant changes relative to 1.1.1:
+
+1. Added a Java wrapper for the TurboJPEG API.  See java/README(java/README)
+for more details.
+
+2. The TurboJPEG API can now be used to scale down images during
+decompression.
+
+3. Added SIMD routines for RGB-to-grayscale color conversion, which
+significantly improves the performance of grayscale JPEG compression from an
+RGB source image.
+
+4. Improved the performance of the C color conversion routines, which are used
+on platforms for which SIMD acceleration is not available.
+
+5. Added a function to the TurboJPEG API that performs lossless transforms.
+This function is implemented using the same back end as jpegtran, but it
+performs transcoding entirely in memory and allows multiple transforms and/or
+crop operations to be batched together, so the source coefficients only need to
+be read once.  This is useful when generating image tiles from a single source
+JPEG.
+
+6. Added tests for the new TurboJPEG scaled decompression and lossless
+transform features to tjbench (the TurboJPEG benchmark, formerly called
+"jpgtest".)
+
+7. Added support for 4:4:0 (transposed 4:2:2) subsampling in TurboJPEG, which
+was necessary in order for it to read 4:2:2 JPEG files that had been losslessly
+transposed or rotated 90 degrees.
+
+8. All legacy VirtualGL code has been re-factored, and this has allowed
+libjpeg-turbo, in its entirety, to be re-licensed under a BSD-style license.
+
+9. libjpeg-turbo can now be built with YASM.
+
+10. Added SIMD acceleration for ARM Linux and iOS platforms that support
+NEON instructions.
+
+11. Refactored the TurboJPEG C API and documented it using Doxygen.  The
+TurboJPEG 1.2 API uses pixel formats to define the size and component order of
+the uncompressed source/destination images, and it includes a more efficient
+version of `TJBUFSIZE()` that computes a worst-case JPEG size based on the
+level of chrominance subsampling.  The refactored implementation of the
+TurboJPEG API now uses the libjpeg memory source and destination managers,
+which allows the TurboJPEG compressor to grow the JPEG buffer as necessary.
+
+12. Eliminated errors in the output of jpegtran on Windows that occurred when
+the application was invoked using I/O redirection
+(`jpegtran <input.jpg >output.jpg`.)
+
+13. The inclusion of libjpeg v7 and v8 emulation as well as arithmetic coding
+support in libjpeg-turbo v1.1.0 introduced several new error constants in
+jerror.h, and these were mistakenly enabled for all emulation modes, causing
+the error enum in libjpeg-turbo to sometimes have different values than the
+same enum in libjpeg.  This represents an ABI incompatibility, and it caused
+problems with rare applications that took specific action based on a particular
+error value.  The fix was to include the new error constants conditionally
+based on whether libjpeg v7 or v8 emulation was enabled.
+
+14. Fixed an issue whereby Windows applications that used libjpeg-turbo would
+fail to compile if the Windows system headers were included before jpeglib.h.
+This issue was caused by a conflict in the definition of the INT32 type.
+
+15. Fixed 32-bit supplementary package for amd64 Debian systems, which was
+broken by enhancements to the packaging system in 1.1.
+
+16. When decompressing a JPEG image using an output colorspace of
+`JCS_EXT_RGBX`, `JCS_EXT_BGRX`, `JCS_EXT_XBGR`, or `JCS_EXT_XRGB`,
+libjpeg-turbo will now set the unused byte to 0xFF, which allows applications
+to interpret that byte as an alpha channel (0xFF = opaque).
+
+
+1.1.1
+=====
+
+### Significant changes relative to 1.1.0:
+
+1. Fixed a 1-pixel error in row 0, column 21 of the luminance plane generated
+by `tjEncodeYUV()`.
+
+2. libjpeg-turbo's accelerated Huffman decoder previously ignored unexpected
+markers found in the middle of the JPEG data stream during decompression.  It
+will now hand off decoding of a particular block to the unaccelerated Huffman
+decoder if an unexpected marker is found, so that the unaccelerated Huffman
+decoder can generate an appropriate warning.
+
+3. Older versions of MinGW64 prefixed symbol names with underscores by
+default, which differed from the behavior of 64-bit Visual C++.  MinGW64 1.0
+has adopted the behavior of 64-bit Visual C++ as the default, so to accommodate
+this, the libjpeg-turbo SIMD function names are no longer prefixed with an
+underscore when building with MinGW64.  This means that, when building
+libjpeg-turbo with older versions of MinGW64, you will now have to add
+`-fno-leading-underscore` to the `CFLAGS`.
+
+4. Fixed a regression bug in the NSIS script that caused the Windows installer
+build to fail when using the Visual Studio IDE.
+
+5. Fixed a bug in `jpeg_read_coefficients()` whereby it would not initialize
+`cinfo->image_width` and `cinfo->image_height` if libjpeg v7 or v8 emulation
+was enabled.  This specifically caused the jpegoptim program to fail if it was
+linked against a version of libjpeg-turbo that was built with libjpeg v7 or v8
+emulation.
+
+6. Eliminated excessive I/O overhead that occurred when reading BMP files in
+cjpeg.
+
+7. Eliminated errors in the output of cjpeg on Windows that occurred when the
+application was invoked using I/O redirection (`cjpeg <inputfile >output.jpg`.)
+
+
+1.1.0
+=====
+
+### Significant changes relative to 1.1 beta1:
+
+1. The algorithm used by the SIMD quantization function cannot produce correct
+results when the JPEG quality is >= 98 and the fast integer forward DCT is
+used.  Thus, the non-SIMD quantization function is now used for those cases,
+and libjpeg-turbo should now produce identical output to libjpeg v6b in all
+cases.
+
+2. Despite the above, the fast integer forward DCT still degrades somewhat for
+JPEG qualities greater than 95, so the TurboJPEG wrapper will now automatically
+use the accurate integer forward DCT when generating JPEG images of quality 96
+or greater.  This reduces compression performance by as much as 15% for these
+high-quality images but is necessary to ensure that the images are perceptually
+lossless.  It also ensures that the library can avoid the performance pitfall
+created by 1.
+
+3. Ported jpgtest.cxx to pure C to avoid the need for a C++ compiler.
+
+4. Fixed visual artifacts in grayscale JPEG compression caused by a typo in
+the RGB-to-luminance lookup tables.
+
+5. The Windows distribution packages now include the libjpeg run-time programs
+(cjpeg, etc.)
+
+6. All packages now include jpgtest.
+
+7. The TurboJPEG dynamic library now uses versioned symbols.
+
+8. Added two new TurboJPEG API functions, `tjEncodeYUV()` and
+`tjDecompressToYUV()`, to replace the somewhat hackish `TJ_YUV` flag.
+
+
+1.0.90 (1.1 beta1)
+==================
+
+### Significant changes relative to 1.0.1:
+
+1. Added emulation of the libjpeg v7 and v8 APIs and ABIs.  See
+README.md(README.md) for more details.  This feature was sponsored by
+CamTrace SAS.
+
+2. Created a new CMake-based build system for the Visual C++ and MinGW builds.
+
+3. Grayscale bitmaps can now be compressed from/decompressed to using the
+TurboJPEG API.
+
+4. jpgtest can now be used to test decompression performance with existing
+JPEG images.
+
+5. If the default install prefix (/opt/libjpeg-turbo) is used, then
+`make install` now creates /opt/libjpeg-turbo/lib32 and
+/opt/libjpeg-turbo/lib64 sym links to duplicate the behavior of the binary
+packages.
+
+6. All symbols in the libjpeg-turbo dynamic library are now versioned, even
+when the library is built with libjpeg v6b emulation.
+
+7. Added arithmetic encoding and decoding support (can be disabled with
+configure or CMake options)
+
+8. Added a `TJ_YUV` flag to the TurboJPEG API, which causes both the compressor
+and decompressor to output planar YUV images.
+
+9. Added an extended version of `tjDecompressHeader()` to the TurboJPEG API,
+which allows the caller to determine the type of subsampling used in a JPEG
+image.
+
+10. Added further protections against invalid Huffman codes.
+
+
+1.0.1
+=====
+
+### Significant changes relative to 1.0.0:
+
+1. The Huffman decoder will now handle erroneous Huffman codes (for instance,
+from a corrupt JPEG image.)  Previously, these would cause libjpeg-turbo to
+crash under certain circumstances.
+
+2. Fixed typo in SIMD dispatch routines that was causing 4:2:2 upsampling to
+be used instead of 4:2:0 when decompressing JPEG images using SSE2 code.
+
+3. The configure script will now automatically determine whether the
+`INCOMPLETE_TYPES_BROKEN` macro should be defined.
+
+
+1.0.0
+=====
+
+### Significant changes relative to 0.0.93:
+
+1. 2983700: Further FreeBSD build tweaks (no longer necessary to specify
+`--host` when configuring on a 64-bit system)
+
+2. Created symlinks in the Unix/Linux packages so that the TurboJPEG
+include file can always be found in /opt/libjpeg-turbo/include, the 32-bit
+static libraries can always be found in /opt/libjpeg-turbo/lib32, and the
+64-bit static libraries can always be found in /opt/libjpeg-turbo/lib64.
+
+3. The Unix/Linux distribution packages now include the libjpeg run-time
+programs (cjpeg, etc.) and man pages.
+
+4. Created a 32-bit supplementary package for amd64 Debian systems, which
+contains just the 32-bit libjpeg-turbo libraries.
+
+5. Moved the libraries from */lib32 to */lib in the i386 Debian package.
+
+6. Include distribution package for Cygwin
+
+7. No longer necessary to specify `--without-simd` on non-x86 architectures,
+and unit tests now work on those architectures.
+
+
+0.0.93
+======
+
+### Significant changes since 0.0.91:
+
+1. 2982659: Fixed x86-64 build on FreeBSD systems
+
+2. 2988188: Added support for Windows 64-bit systems
+
+
+0.0.91
+======
+
+### Significant changes relative to 0.0.90:
+
+1. Added documentation to .deb packages
+
+2. 2968313: Fixed data corruption issues when decompressing large JPEG images
+and/or using buffered I/O with the libjpeg-turbo decompressor
+
+
+0.0.90
+======
+
+Initial release

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/LICENSE.md Added

@@ -0,0 +1,132 @@
+libjpeg-turbo Licenses
+======================
+
+libjpeg-turbo is covered by three compatible BSD-style open source licenses:
+
+- The IJG (Independent JPEG Group) License, which is listed in
+  README.ijg(README.ijg)
+
+  This license applies to the libjpeg API library and associated programs
+  (any code inherited from libjpeg, and any modifications to that code.)
+
+- The Modified (3-clause) BSD License, which is listed below
+
+  This license covers the TurboJPEG API library and associated programs, as
+  well as the build system.
+
+- The zlib License(https://opensource.org/licenses/Zlib)
+
+  This license is a subset of the other two, and it covers the libjpeg-turbo
+  SIMD extensions.
+
+
+Complying with the libjpeg-turbo Licenses
+=========================================
+
+This section provides a roll-up of the libjpeg-turbo licensing terms, to the
+best of our understanding.
+
+1.  If you are distributing a modified version of the libjpeg-turbo source,
+    then:
+
+    1.  You cannot alter or remove any existing copyright or license notices
+        from the source.
+
+        **Origin**
+        - Clause 1 of the IJG License
+        - Clause 1 of the Modified BSD License
+        - Clauses 1 and 3 of the zlib License
+
+    2.  You must add your own copyright notice to the header of each source
+        file you modified, so others can tell that you modified that file (if
+        there is not an existing copyright header in that file, then you can
+        simply add a notice stating that you modified the file.)
+
+        **Origin**
+        - Clause 1 of the IJG License
+        - Clause 2 of the zlib License
+
+    3.  You must include the IJG README file, and you must not alter any of the
+        copyright or license text in that file.
+
+        **Origin**
+        - Clause 1 of the IJG License
+
+2.  If you are distributing only libjpeg-turbo binaries without the source, or
+    if you are distributing an application that statically links with
+    libjpeg-turbo, then:
+
+    1.  Your product documentation must include a message stating:
+
+        This software is based in part on the work of the Independent JPEG
+        Group.
+
+        **Origin**
+        - Clause 2 of the IJG license
+
+    2.  If your binary distribution includes or uses the TurboJPEG API, then
+        your product documentation must include the text of the Modified BSD
+        License (see below.)
+
+        **Origin**
+        - Clause 2 of the Modified BSD License
+
+3.  You cannot use the name of the IJG or The libjpeg-turbo Project or the
+    contributors thereof in advertising, publicity, etc.
+
+    **Origin**
+    - IJG License
+    - Clause 3 of the Modified BSD License
+
+4.  The IJG and The libjpeg-turbo Project do not warrant libjpeg-turbo to be
+    free of defects, nor do we accept any liability for undesirable
+    consequences resulting from your use of the software.
+
+    **Origin**
+    - IJG License
+    - Modified BSD License
+    - zlib License
+
+
+The Modified (3-clause) BSD License
+===================================
+
+Copyright (C)2009-2021 D. R. Commander.  All Rights Reserved.<br>
+Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+- Redistributions of source code must retain the above copyright notice,
+  this list of conditions and the following disclaimer.
+- Redistributions in binary form must reproduce the above copyright notice,
+  this list of conditions and the following disclaimer in the documentation
+  and/or other materials provided with the distribution.
+- Neither the name of the libjpeg-turbo Project nor the names of its
+  contributors may be used to endorse or promote products derived from this
+  software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+POSSIBILITY OF SUCH DAMAGE.
+
+
+Why Three Licenses?
+===================
+
+The zlib License could have been used instead of the Modified (3-clause) BSD
+License, and since the IJG License effectively subsumes the distribution
+conditions of the zlib License, this would have effectively placed
+libjpeg-turbo binary distributions under the IJG License.  However, the IJG
+License specifically refers to the Independent JPEG Group and does not extend
+attribution and endorsement protections to other entities.  Thus, it was
+desirable to choose a license that granted us the same protections for new code
+that were granted to the IJG for code derived from their software.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/README.ijg Added

@@ -0,0 +1,258 @@
+libjpeg-turbo note:  This file has been modified by The libjpeg-turbo Project
+to include only information relevant to libjpeg-turbo, to wordsmith certain
+sections, and to remove impolitic language that existed in the libjpeg v8
+README.  It is included only for reference.  Please see README.md for
+information specific to libjpeg-turbo.
+
+
+The Independent JPEG Group's JPEG software
+==========================================
+
+This distribution contains a release of the Independent JPEG Group's free JPEG
+software.  You are welcome to redistribute this software and to use it for any
+purpose, subject to the conditions under LEGAL ISSUES, below.
+
+This software is the work of Tom Lane, Guido Vollbeding, Philip Gladstone,
+Bill Allombert, Jim Boucher, Lee Crocker, Bob Friesenhahn, Ben Jackson,
+Julian Minguillon, Luis Ortiz, George Phillips, Davide Rossi, Ge' Weijers,
+and other members of the Independent JPEG Group.
+
+IJG is not affiliated with the ISO/IEC JTC1/SC29/WG1 standards committee
+(also known as JPEG, together with ITU-T SG16).
+
+
+DOCUMENTATION ROADMAP
+=====================
+
+This file contains the following sections:
+
+OVERVIEW            General description of JPEG and the IJG software.
+LEGAL ISSUES        Copyright, lack of warranty, terms of distribution.
+REFERENCES          Where to learn more about JPEG.
+ARCHIVE LOCATIONS   Where to find newer versions of this software.
+FILE FORMAT WARS    Software *not* to get.
+TO DO               Plans for future IJG releases.
+
+Other documentation files in the distribution are:
+
+User documentation:
+  usage.txt         Usage instructions for cjpeg, djpeg, jpegtran,
+                    rdjpgcom, and wrjpgcom.
+  *.1               Unix-style man pages for programs (same info as usage.txt).
+  wizard.txt        Advanced usage instructions for JPEG wizards only.
+  change.log        Version-to-version change highlights.
+Programmer and internal documentation:
+  libjpeg.txt       How to use the JPEG library in your own programs.
+  example.txt       Sample code for calling the JPEG library.
+  structure.txt     Overview of the JPEG library's internal structure.
+  coderules.txt     Coding style rules --- please read if you contribute code.
+
+Please read at least usage.txt.  Some information can also be found in the JPEG
+FAQ (Frequently Asked Questions) article.  See ARCHIVE LOCATIONS below to find
+out where to obtain the FAQ article.
+
+If you want to understand how the JPEG code works, we suggest reading one or
+more of the REFERENCES, then looking at the documentation files (in roughly
+the order listed) before diving into the code.
+
+
+OVERVIEW
+========
+
+This package contains C software to implement JPEG image encoding, decoding,
+and transcoding.  JPEG (pronounced "jay-peg") is a standardized compression
+method for full-color and grayscale images.  JPEG's strong suit is compressing
+photographic images or other types of images that have smooth color and
+brightness transitions between neighboring pixels.  Images with sharp lines or
+other abrupt features may not compress well with JPEG, and a higher JPEG
+quality may have to be used to avoid visible compression artifacts with such
+images.
+
+JPEG is lossy, meaning that the output pixels are not necessarily identical to
+the input pixels.  However, on photographic content and other "smooth" images,
+very good compression ratios can be obtained with no visible compression
+artifacts, and extremely high compression ratios are possible if you are
+willing to sacrifice image quality (by reducing the "quality" setting in the
+compressor.)
+
+This software implements JPEG baseline, extended-sequential, and progressive
+compression processes.  Provision is made for supporting all variants of these
+processes, although some uncommon parameter settings aren't implemented yet.
+We have made no provision for supporting the hierarchical or lossless
+processes defined in the standard.
+
+We provide a set of library routines for reading and writing JPEG image files,
+plus two sample applications "cjpeg" and "djpeg", which use the library to
+perform conversion between JPEG and some other popular image file formats.
+The library is intended to be reused in other applications.
+
+In order to support file conversion and viewing software, we have included
+considerable functionality beyond the bare JPEG coding/decoding capability;
+for example, the color quantization modules are not strictly part of JPEG
+decoding, but they are essential for output to colormapped file formats or
+colormapped displays.  These extra functions can be compiled out of the
+library if not required for a particular application.
+
+We have also included "jpegtran", a utility for lossless transcoding between
+different JPEG processes, and "rdjpgcom" and "wrjpgcom", two simple
+applications for inserting and extracting textual comments in JFIF files.
+
+The emphasis in designing this software has been on achieving portability and
+flexibility, while also making it fast enough to be useful.  In particular,
+the software is not intended to be read as a tutorial on JPEG.  (See the
+REFERENCES section for introductory material.)  Rather, it is intended to
+be reliable, portable, industrial-strength code.  We do not claim to have
+achieved that goal in every aspect of the software, but we strive for it.
+
+We welcome the use of this software as a component of commercial products.
+No royalty is required, but we do ask for an acknowledgement in product
+documentation, as described under LEGAL ISSUES.
+
+
+LEGAL ISSUES
+============
+
+In plain English:
+
+1. We don't promise that this software works.  (But if you find any bugs,
+   please let us know!)
+2. You can use this software for whatever you want.  You don't have to pay us.
+3. You may not pretend that you wrote this software.  If you use it in a
+   program, you must acknowledge somewhere in your documentation that
+   you've used the IJG code.
+
+In legalese:
+
+The authors make NO WARRANTY or representation, either express or implied,
+with respect to this software, its quality, accuracy, merchantability, or
+fitness for a particular purpose.  This software is provided "AS IS", and you,
+its user, assume the entire risk as to its quality and accuracy.
+
+This software is copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
+All Rights Reserved except as specified below.
+
+Permission is hereby granted to use, copy, modify, and distribute this
+software (or portions thereof) for any purpose, without fee, subject to these
+conditions:
+(1) If any part of the source code for this software is distributed, then this
+README file must be included, with this copyright and no-warranty notice
+unaltered; and any additions, deletions, or changes to the original files
+must be clearly indicated in accompanying documentation.
+(2) If only executable code is distributed, then the accompanying
+documentation must state that "this software is based in part on the work of
+the Independent JPEG Group".
+(3) Permission for use of this software is granted only if the user accepts
+full responsibility for any undesirable consequences; the authors accept
+NO LIABILITY for damages of any kind.
+
+These conditions apply to any software derived from or based on the IJG code,
+not just to the unmodified library.  If you use our work, you ought to
+acknowledge us.
+
+Permission is NOT granted for the use of any IJG author's name or company name
+in advertising or publicity relating to this software or products derived from
+it.  This software may be referred to only as "the Independent JPEG Group's
+software".
+
+We specifically permit and encourage the use of this software as the basis of
+commercial products, provided that all warranty or liability claims are
+assumed by the product vendor.
+
+
+REFERENCES
+==========
+
+We recommend reading one or more of these references before trying to
+understand the innards of the JPEG software.
+
+The best short technical introduction to the JPEG compression algorithm is
+        Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+        Communications of the ACM, April 1991 (vol. 34 no. 4), pp. 30-44.
+(Adjacent articles in that issue discuss MPEG motion picture compression,
+applications of JPEG, and related topics.)  If you don't have the CACM issue
+handy, a PDF file containing a revised version of Wallace's article is
+available at http://www.ijg.org/files/Wallace.JPEG.pdf.  The file (actually
+a preprint for an article that appeared in IEEE Trans. Consumer Electronics)
+omits the sample images that appeared in CACM, but it includes corrections
+and some added material.  Note: the Wallace article is copyright ACM and IEEE,
+and it may not be used for commercial purposes.
+
+A somewhat less technical, more leisurely introduction to JPEG can be found in
+"The Data Compression Book" by Mark Nelson and Jean-loup Gailly, published by
+M&T Books (New York), 2nd ed. 1996, ISBN 1-55851-434-1.  This book provides
+good explanations and example C code for a multitude of compression methods
+including JPEG.  It is an excellent source if you are comfortable reading C
+code but don't know much about data compression in general.  The book's JPEG
+sample code is far from industrial-strength, but when you are ready to look
+at a full implementation, you've got one here...
+
+The best currently available description of JPEG is the textbook "JPEG Still
+Image Data Compression Standard" by William B. Pennebaker and Joan L.
+Mitchell, published by Van Nostrand Reinhold, 1993, ISBN 0-442-01272-1.
+Price US$59.95, 638 pp.  The book includes the complete text of the ISO JPEG
+standards (DIS 10918-1 and draft DIS 10918-2).
+
+The original JPEG standard is divided into two parts, Part 1 being the actual
+specification, while Part 2 covers compliance testing methods.  Part 1 is
+titled "Digital Compression and Coding of Continuous-tone Still Images,
+Part 1: Requirements and guidelines" and has document numbers ISO/IEC IS
+10918-1, ITU-T T.81.  Part 2 is titled "Digital Compression and Coding of
+Continuous-tone Still Images, Part 2: Compliance testing" and has document
+numbers ISO/IEC IS 10918-2, ITU-T T.83.
+
+The JPEG standard does not specify all details of an interchangeable file
+format.  For the omitted details, we follow the "JFIF" conventions, revision
+1.02.  JFIF version 1 has been adopted as ISO/IEC 10918-5 (05/2013) and
+Recommendation ITU-T T.871 (05/2011): Information technology - Digital
+compression and coding of continuous-tone still images: JPEG File Interchange
+Format (JFIF).  It is available as a free download in PDF file format from
+https://www.iso.org/standard/54989.html and http://www.itu.int/rec/T-REC-T.871.
+A PDF file of the older JFIF 1.02 specification is available at
+http://www.w3.org/Graphics/JPEG/jfif3.pdf.
+
+The TIFF 6.0 file format specification can be obtained from
+http://mirrors.ctan.org/graphics/tiff/TIFF6.ps.gz.  The JPEG incorporation
+scheme found in the TIFF 6.0 spec of 3-June-92 has a number of serious
+problems.  IJG does not recommend use of the TIFF 6.0 design (TIFF Compression
+tag 6).  Instead, we recommend the JPEG design proposed by TIFF Technical Note
+#2 (Compression tag 7).  Copies of this Note can be obtained from
+http://www.ijg.org/files/.  It is expected that the next revision
+of the TIFF spec will replace the 6.0 JPEG design with the Note's design.
+Although IJG's own code does not support TIFF/JPEG, the free libtiff library
+uses our library to implement TIFF/JPEG per the Note.
+
+
+ARCHIVE LOCATIONS
+=================
+
+The "official" archive site for this software is www.ijg.org.
+The most recent released version can always be found there in
+directory "files".
+
+The JPEG FAQ (Frequently Asked Questions) article is a source of some
+general information about JPEG.  It is available at
+http://www.faqs.org/faqs/jpeg-faq.
+
+
+FILE FORMAT COMPATIBILITY
+=========================
+
+This software implements ITU T.81 | ISO/IEC 10918 with some extensions from
+ITU T.871 | ISO/IEC 10918-5 (JPEG File Interchange Format-- see REFERENCES).
+Informally, the term "JPEG image" or "JPEG file" most often refers to JFIF or
+a subset thereof, but there are other formats containing the name "JPEG" that
+are incompatible with the DCT-based JPEG standard or with JFIF (for instance,
+JPEG 2000 and JPEG XR).  This software therefore does not support these
+formats.  Indeed, one of the original reasons for developing this free software
+was to help force convergence on a common, interoperable format standard for
+JPEG files.
+
+JFIF is a minimal or "low end" representation.  TIFF/JPEG (TIFF revision 6.0 as
+modified by TIFF Technical Note #2) can be used for "high end" applications
+that need to record a lot of additional data about an image.
+
+
+TO DO
+=====
+
+Please send bug reports, offers of help, etc. to jpeg-info@jpegclub.org.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/README.md Added

@@ -0,0 +1,357 @@
+Background
+==========
+
+libjpeg-turbo is a JPEG image codec that uses SIMD instructions to accelerate
+baseline JPEG compression and decompression on x86, x86-64, Arm, PowerPC, and
+MIPS systems, as well as progressive JPEG compression on x86, x86-64, and Arm
+systems.  On such systems, libjpeg-turbo is generally 2-6x as fast as libjpeg,
+all else being equal.  On other types of systems, libjpeg-turbo can still
+outperform libjpeg by a significant amount, by virtue of its highly-optimized
+Huffman coding routines.  In many cases, the performance of libjpeg-turbo
+rivals that of proprietary high-speed JPEG codecs.
+
+libjpeg-turbo implements both the traditional libjpeg API as well as the less
+powerful but more straightforward TurboJPEG API.  libjpeg-turbo also features
+colorspace extensions that allow it to compress from/decompress to 32-bit and
+big-endian pixel buffers (RGBX, XBGR, etc.), as well as a full-featured Java
+interface.
+
+libjpeg-turbo was originally based on libjpeg/SIMD, an MMX-accelerated
+derivative of libjpeg v6b developed by Miyasaka Masaru.  The TigerVNC and
+VirtualGL projects made numerous enhancements to the codec in 2009, and in
+early 2010, libjpeg-turbo spun off into an independent project, with the goal
+of making high-speed JPEG compression/decompression technology available to a
+broader range of users and developers.
+
+
+License
+=======
+
+libjpeg-turbo is covered by three compatible BSD-style open source licenses.
+Refer to LICENSE.md(LICENSE.md) for a roll-up of license terms.
+
+
+Building libjpeg-turbo
+======================
+
+Refer to BUILDING.md(BUILDING.md) for complete instructions.
+
+
+Using libjpeg-turbo
+===================
+
+libjpeg-turbo includes two APIs that can be used to compress and decompress
+JPEG images:
+
+- **TurboJPEG API**<br>
+  This API provides an easy-to-use interface for compressing and decompressing
+  JPEG images in memory.  It also provides some functionality that would not be
+  straightforward to achieve using the underlying libjpeg API, such as
+  generating planar YUV images and performing multiple simultaneous lossless
+  transforms on an image.  The Java interface for libjpeg-turbo is written on
+  top of the TurboJPEG API.  The TurboJPEG API is recommended for first-time
+  users of libjpeg-turbo.  Refer to tjexample.c(tjexample.c) and
+  TJExample.java(java/TJExample.java) for examples of its usage and to
+  <http://libjpeg-turbo.org/Documentation/Documentation> for API documentation.
+
+- **libjpeg API**<br>
+  This is the de facto industry-standard API for compressing and decompressing
+  JPEG images.  It is more difficult to use than the TurboJPEG API but also
+  more powerful.  The libjpeg API implementation in libjpeg-turbo is both
+  API/ABI-compatible and mathematically compatible with libjpeg v6b.  It can
+  also optionally be configured to be API/ABI-compatible with libjpeg v7 and v8
+  (see below.)  Refer to cjpeg.c(cjpeg.c) and djpeg.c(djpeg.c) for examples
+  of its usage and to libjpeg.txt(libjpeg.txt) for API documentation.
+
+There is no significant performance advantage to either API when both are used
+to perform similar operations.
+
+Colorspace Extensions
+---------------------
+
+libjpeg-turbo includes extensions that allow JPEG images to be compressed
+directly from (and decompressed directly to) buffers that use BGR, BGRX,
+RGBX, XBGR, and XRGB pixel ordering.  This is implemented with ten new
+colorspace constants:
+
+    JCS_EXT_RGB   /* red/green/blue */
+    JCS_EXT_RGBX  /* red/green/blue/x */
+    JCS_EXT_BGR   /* blue/green/red */
+    JCS_EXT_BGRX  /* blue/green/red/x */
+    JCS_EXT_XBGR  /* x/blue/green/red */
+    JCS_EXT_XRGB  /* x/red/green/blue */
+    JCS_EXT_RGBA  /* red/green/blue/alpha */
+    JCS_EXT_BGRA  /* blue/green/red/alpha */
+    JCS_EXT_ABGR  /* alpha/blue/green/red */
+    JCS_EXT_ARGB  /* alpha/red/green/blue */
+
+Setting `cinfo.in_color_space` (compression) or `cinfo.out_color_space`
+(decompression) to one of these values will cause libjpeg-turbo to read the
+red, green, and blue values from (or write them to) the appropriate position in
+the pixel when compressing from/decompressing to an RGB buffer.
+
+Your application can check for the existence of these extensions at compile
+time with:
+
+    #ifdef JCS_EXTENSIONS
+
+At run time, attempting to use these extensions with a libjpeg implementation
+that does not support them will result in a "Bogus input colorspace" error.
+Applications can trap this error in order to test whether run-time support is
+available for the colorspace extensions.
+
+When using the RGBX, BGRX, XBGR, and XRGB colorspaces during decompression, the
+X byte is undefined, and in order to ensure the best performance, libjpeg-turbo
+can set that byte to whatever value it wishes.  If an application expects the X
+byte to be used as an alpha channel, then it should specify `JCS_EXT_RGBA`,
+`JCS_EXT_BGRA`, `JCS_EXT_ABGR`, or `JCS_EXT_ARGB`.  When these colorspace
+constants are used, the X byte is guaranteed to be 0xFF, which is interpreted
+as opaque.
+
+Your application can check for the existence of the alpha channel colorspace
+extensions at compile time with:
+
+    #ifdef JCS_ALPHA_EXTENSIONS
+
+jcstest.c(jcstest.c), located in the libjpeg-turbo source tree, demonstrates
+how to check for the existence of the colorspace extensions at compile time and
+run time.
+
+libjpeg v7 and v8 API/ABI Emulation
+-----------------------------------
+
+With libjpeg v7 and v8, new features were added that necessitated extending the
+compression and decompression structures.  Unfortunately, due to the exposed
+nature of those structures, extending them also necessitated breaking backward
+ABI compatibility with previous libjpeg releases.  Thus, programs that were
+built to use libjpeg v7 or v8 did not work with libjpeg-turbo, since it is
+based on the libjpeg v6b code base.  Although libjpeg v7 and v8 are not
+as widely used as v6b, enough programs (including a few Linux distros) made
+the switch that there was a demand to emulate the libjpeg v7 and v8 ABIs
+in libjpeg-turbo.  It should be noted, however, that this feature was added
+primarily so that applications that had already been compiled to use libjpeg
+v7+ could take advantage of accelerated baseline JPEG encoding/decoding
+without recompiling.  libjpeg-turbo does not claim to support all of the
+libjpeg v7+ features, nor to produce identical output to libjpeg v7+ in all
+cases (see below.)
+
+By passing an argument of `-DWITH_JPEG7=1` or `-DWITH_JPEG8=1` to `cmake`, you
+can build a version of libjpeg-turbo that emulates the libjpeg v7 or v8 ABI, so
+that programs that are built against libjpeg v7 or v8 can be run with
+libjpeg-turbo.  The following section describes which libjpeg v7+ features are
+supported and which aren't.
+
+### Support for libjpeg v7 and v8 Features
+
+#### Fully supported
+
+- **libjpeg API: IDCT scaling extensions in decompressor**<br>
+  libjpeg-turbo supports IDCT scaling with scaling factors of 1/8, 1/4, 3/8,
+  1/2, 5/8, 3/4, 7/8, 9/8, 5/4, 11/8, 3/2, 13/8, 7/4, 15/8, and 2/1 (only 1/4
+  and 1/2 are SIMD-accelerated.)
+
+- **libjpeg API: Arithmetic coding**
+
+- **libjpeg API: In-memory source and destination managers**<br>
+  See notes below.
+
+- **cjpeg: Separate quality settings for luminance and chrominance**<br>
+  Note that the libpjeg v7+ API was extended to accommodate this feature only
+  for convenience purposes.  It has always been possible to implement this
+  feature with libjpeg v6b (see rdswitch.c for an example.)
+
+- **cjpeg: 32-bit BMP support**
+
+- **cjpeg: `-rgb` option**
+
+- **jpegtran: Lossless cropping**
+
+- **jpegtran: `-perfect` option**
+
+- **jpegtran: Forcing width/height when performing lossless crop**
+
+- **rdjpgcom: `-raw` option**
+
+- **rdjpgcom: Locale awareness**
+
+
+#### Not supported
+
+NOTE:  As of this writing, extensive research has been conducted into the
+usefulness of DCT scaling as a means of data reduction and SmartScale as a
+means of quality improvement.  Readers are invited to peruse the research at
+<http://www.libjpeg-turbo.org/About/SmartScale> and draw their own conclusions,
+but it is the general belief of our project that these features have not
+demonstrated sufficient usefulness to justify inclusion in libjpeg-turbo.
+
+- **libjpeg API: DCT scaling in compressor**<br>
+  `cinfo.scale_num` and `cinfo.scale_denom` are silently ignored.
+  There is no technical reason why DCT scaling could not be supported when
+  emulating the libjpeg v7+ API/ABI, but without the SmartScale extension (see
+  below), only scaling factors of 1/2, 8/15, 4/7, 8/13, 2/3, 8/11, 4/5, and
+  8/9 would be available, which is of limited usefulness.
+
+- **libjpeg API: SmartScale**<br>
+  `cinfo.block_size` is silently ignored.
+  SmartScale is an extension to the JPEG format that allows for DCT block
+  sizes other than 8x8.  Providing support for this new format would be
+  feasible (particularly without full acceleration.)  However, until/unless
+  the format becomes either an official industry standard or, at minimum, an
+  accepted solution in the community, we are hesitant to implement it, as
+  there is no sense of whether or how it might change in the future.  It is
+  our belief that SmartScale has not demonstrated sufficient usefulness as a
+  lossless format nor as a means of quality enhancement, and thus our primary
+  interest in providing this feature would be as a means of supporting
+  additional DCT scaling factors.
+
+- **libjpeg API: Fancy downsampling in compressor**<br>
+  `cinfo.do_fancy_downsampling` is silently ignored.
+  This requires the DCT scaling feature, which is not supported.
+
+- **jpegtran: Scaling**<br>
+  This requires both the DCT scaling and SmartScale features, which are not
+  supported.
+
+- **Lossless RGB JPEG files**<br>
+  This requires the SmartScale feature, which is not supported.
+
+### What About libjpeg v9?
+
+libjpeg v9 introduced yet another field to the JPEG compression structure
+(`color_transform`), thus making the ABI backward incompatible with that of
+libjpeg v8.  This new field was introduced solely for the purpose of supporting
+lossless SmartScale encoding.  Furthermore, there was actually no reason to
+extend the API in this manner, as the color transform could have just as easily
+been activated by way of a new JPEG colorspace constant, thus preserving
+backward ABI compatibility.
+
+Our research (see link above) has shown that lossless SmartScale does not
+generally accomplish anything that can't already be accomplished better with
+existing, standard lossless formats.  Therefore, at this time it is our belief
+that there is not sufficient technical justification for software projects to
+upgrade from libjpeg v8 to libjpeg v9, and thus there is not sufficient
+technical justification for us to emulate the libjpeg v9 ABI.
+
+In-Memory Source/Destination Managers
+-------------------------------------
+
+By default, libjpeg-turbo 1.3 and later includes the `jpeg_mem_src()` and
+`jpeg_mem_dest()` functions, even when not emulating the libjpeg v8 API/ABI.
+Previously, it was necessary to build libjpeg-turbo from source with libjpeg v8
+API/ABI emulation in order to use the in-memory source/destination managers,
+but several projects requested that those functions be included when emulating
+the libjpeg v6b API/ABI as well.  This allows the use of those functions by
+programs that need them, without breaking ABI compatibility for programs that
+don't, and it allows those functions to be provided in the "official"
+libjpeg-turbo binaries.
+
+Those who are concerned about maintaining strict conformance with the libjpeg
+v6b or v7 API can pass an argument of `-DWITH_MEM_SRCDST=0` to `cmake` prior to
+building libjpeg-turbo.  This will restore the pre-1.3 behavior, in which
+`jpeg_mem_src()` and `jpeg_mem_dest()` are only included when emulating the
+libjpeg v8 API/ABI.
+
+On Un*x systems, including the in-memory source/destination managers changes
+the dynamic library version from 62.2.0 to 62.3.0 if using libjpeg v6b API/ABI
+emulation and from 7.2.0 to 7.3.0 if using libjpeg v7 API/ABI emulation.
+
+Note that, on most Un*x systems, the dynamic linker will not look for a
+function in a library until that function is actually used.  Thus, if a program
+is built against libjpeg-turbo 1.3+ and uses `jpeg_mem_src()` or
+`jpeg_mem_dest()`, that program will not fail if run against an older version
+of libjpeg-turbo or against libjpeg v7- until the program actually tries to
+call `jpeg_mem_src()` or `jpeg_mem_dest()`.  Such is not the case on Windows.
+If a program is built against the libjpeg-turbo 1.3+ DLL and uses
+`jpeg_mem_src()` or `jpeg_mem_dest()`, then it must use the libjpeg-turbo 1.3+
+DLL at run time.
+
+Both cjpeg and djpeg have been extended to allow testing the in-memory
+source/destination manager functions.  See their respective man pages for more
+details.
+
+
+Mathematical Compatibility
+==========================
+
+For the most part, libjpeg-turbo should produce identical output to libjpeg
+v6b.  The one exception to this is when using the floating point DCT/IDCT, in
+which case the outputs of libjpeg v6b and libjpeg-turbo can differ for the
+following reasons:
+
+- The SSE/SSE2 floating point DCT implementation in libjpeg-turbo is ever so
+  slightly more accurate than the implementation in libjpeg v6b, but not by
+  any amount perceptible to human vision (generally in the range of 0.01 to
+  0.08 dB gain in PNSR.)
+
+- When not using the SIMD extensions, libjpeg-turbo uses the more accurate
+  (and slightly faster) floating point IDCT algorithm introduced in libjpeg
+  v8a as opposed to the algorithm used in libjpeg v6b.  It should be noted,
+  however, that this algorithm basically brings the accuracy of the floating
+  point IDCT in line with the accuracy of the accurate integer IDCT.  The
+  floating point DCT/IDCT algorithms are mainly a legacy feature, and they do
+  not produce significantly more accuracy than the accurate integer algorithms
+  (to put numbers on this, the typical difference in PNSR between the two
+  algorithms is less than 0.10 dB, whereas changing the quality level by 1 in
+  the upper range of the quality scale is typically more like a 1.0 dB
+  difference.)
+
+- If the floating point algorithms in libjpeg-turbo are not implemented using
+  SIMD instructions on a particular platform, then the accuracy of the
+  floating point DCT/IDCT can depend on the compiler settings.
+
+While libjpeg-turbo does emulate the libjpeg v8 API/ABI, under the hood it is
+still using the same algorithms as libjpeg v6b, so there are several specific
+cases in which libjpeg-turbo cannot be expected to produce the same output as
+libjpeg v8:
+
+- When decompressing using scaling factors of 1/2 and 1/4, because libjpeg v8
+  implements those scaling algorithms differently than libjpeg v6b does, and
+  libjpeg-turbo's SIMD extensions are based on the libjpeg v6b behavior.
+
+- When using chrominance subsampling, because libjpeg v8 implements this
+  with its DCT/IDCT scaling algorithms rather than with a separate
+  downsampling/upsampling algorithm.  In our testing, the subsampled/upsampled
+  output of libjpeg v8 is less accurate than that of libjpeg v6b for this
+  reason.
+
+- When decompressing using a scaling factor > 1 and merged (AKA "non-fancy" or
+  "non-smooth") chrominance upsampling, because libjpeg v8 does not support
+  merged upsampling with scaling factors > 1.
+
+
+Performance Pitfalls
+====================
+
+Restart Markers
+---------------
+
+The optimized Huffman decoder in libjpeg-turbo does not handle restart markers
+in a way that makes the rest of the libjpeg infrastructure happy, so it is
+necessary to use the slow Huffman decoder when decompressing a JPEG image that
+has restart markers.  This can cause the decompression performance to drop by
+as much as 20%, but the performance will still be much greater than that of
+libjpeg.  Many consumer packages, such as Photoshop, use restart markers when
+generating JPEG images, so images generated by those programs will experience
+this issue.
+
+Fast Integer Forward DCT at High Quality Levels
+-----------------------------------------------
+
+The algorithm used by the SIMD-accelerated quantization function cannot produce
+correct results whenever the fast integer forward DCT is used along with a JPEG
+quality of 98-100.  Thus, libjpeg-turbo must use the non-SIMD quantization
+function in those cases.  This causes performance to drop by as much as 40%.
+It is therefore strongly advised that you use the accurate integer forward DCT
+whenever encoding images with a JPEG quality of 98 or higher.
+
+
+Memory Debugger Pitfalls
+========================
+
+Valgrind and Memory Sanitizer (MSan) can generate false positives
+(specifically, incorrect reports of uninitialized memory accesses) when used
+with libjpeg-turbo's SIMD extensions.  It is generally recommended that the
+SIMD extensions be disabled, either by passing an argument of `-DWITH_SIMD=0`
+to `cmake` when configuring the build or by setting the environment variable
+`JSIMD_FORCENONE` to `1` at run time, when testing libjpeg-turbo with Valgrind,
+MSan, or other memory debuggers.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cderror.h Added

@@ -0,0 +1,123 @@
+/*
+ * cderror.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * Modified 2009-2017 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file defines the error and message codes for the cjpeg/djpeg
+ * applications.  These strings are not needed as part of the JPEG library
+ * proper.
+ * Edit this file to add new codes, or to translate the message strings to
+ * some other language.
+ */
+
+/*
+ * To define the enum list of message codes, include this file without
+ * defining macro JMESSAGE.  To create a message string table, include it
+ * again with a suitable JMESSAGE definition (see jerror.c for an example).
+ */
+#ifndef JMESSAGE
+#ifndef CDERROR_H
+#define CDERROR_H
+/* First time through, define the enum list */
+#define JMAKE_ENUM_LIST
+#else
+/* Repeated inclusions of this file are no-ops unless JMESSAGE is defined */
+#define JMESSAGE(code, string)
+#endif /* CDERROR_H */
+#endif /* JMESSAGE */
+
+#ifdef JMAKE_ENUM_LIST
+
+typedef enum {
+
+#define JMESSAGE(code, string)  code,
+
+#endif /* JMAKE_ENUM_LIST */
+
+JMESSAGE(JMSG_FIRSTADDONCODE = 1000, NULL) /* Must be first entry! */
+
+#ifdef BMP_SUPPORTED
+JMESSAGE(JERR_BMP_BADCMAP, "Unsupported BMP colormap format")
+JMESSAGE(JERR_BMP_BADDEPTH, "Only 8-, 24-, and 32-bit BMP files are supported")
+JMESSAGE(JERR_BMP_BADHEADER, "Invalid BMP file: bad header length")
+JMESSAGE(JERR_BMP_BADPLANES, "Invalid BMP file: biPlanes not equal to 1")
+JMESSAGE(JERR_BMP_COLORSPACE, "BMP output must be grayscale or RGB")
+JMESSAGE(JERR_BMP_COMPRESSED, "Sorry, compressed BMPs not yet supported")
+JMESSAGE(JERR_BMP_EMPTY, "Empty BMP image")
+JMESSAGE(JERR_BMP_NOT, "Not a BMP file - does not start with BM")
+JMESSAGE(JERR_BMP_OUTOFRANGE, "Numeric value out of range in BMP file")
+JMESSAGE(JTRC_BMP, "%ux%u %d-bit BMP image")
+JMESSAGE(JTRC_BMP_MAPPED, "%ux%u 8-bit colormapped BMP image")
+JMESSAGE(JTRC_BMP_OS2, "%ux%u %d-bit OS2 BMP image")
+JMESSAGE(JTRC_BMP_OS2_MAPPED, "%ux%u 8-bit colormapped OS2 BMP image")
+#endif /* BMP_SUPPORTED */
+
+#ifdef GIF_SUPPORTED
+JMESSAGE(JERR_GIF_BUG, "GIF output got confused")
+JMESSAGE(JERR_GIF_CODESIZE, "Bogus GIF codesize %d")
+JMESSAGE(JERR_GIF_COLORSPACE, "GIF output must be grayscale or RGB")
+JMESSAGE(JERR_GIF_EMPTY, "Empty GIF image")
+JMESSAGE(JERR_GIF_IMAGENOTFOUND, "Too few images in GIF file")
+JMESSAGE(JERR_GIF_NOT, "Not a GIF file")
+JMESSAGE(JTRC_GIF, "%ux%ux%d GIF image")
+JMESSAGE(JTRC_GIF_BADVERSION,
+         "Warning: unexpected GIF version number '%c%c%c'")
+JMESSAGE(JTRC_GIF_EXTENSION, "Ignoring GIF extension block of type 0x%02x")
+JMESSAGE(JTRC_GIF_NONSQUARE, "Caution: nonsquare pixels in input")
+JMESSAGE(JWRN_GIF_BADDATA, "Corrupt data in GIF file")
+JMESSAGE(JWRN_GIF_CHAR, "Bogus char 0x%02x in GIF file, ignoring")
+JMESSAGE(JWRN_GIF_ENDCODE, "Premature end of GIF image")
+JMESSAGE(JWRN_GIF_NOMOREDATA, "Ran out of GIF bits")
+#endif /* GIF_SUPPORTED */
+
+#ifdef PPM_SUPPORTED
+JMESSAGE(JERR_PPM_COLORSPACE, "PPM output must be grayscale or RGB")
+JMESSAGE(JERR_PPM_NONNUMERIC, "Nonnumeric data in PPM file")
+JMESSAGE(JERR_PPM_NOT, "Not a PPM/PGM file")
+JMESSAGE(JERR_PPM_OUTOFRANGE, "Numeric value out of range in PPM file")
+JMESSAGE(JTRC_PGM, "%ux%u PGM image")
+JMESSAGE(JTRC_PGM_TEXT, "%ux%u text PGM image")
+JMESSAGE(JTRC_PPM, "%ux%u PPM image")
+JMESSAGE(JTRC_PPM_TEXT, "%ux%u text PPM image")
+#endif /* PPM_SUPPORTED */
+
+#ifdef TARGA_SUPPORTED
+JMESSAGE(JERR_TGA_BADCMAP, "Unsupported Targa colormap format")
+JMESSAGE(JERR_TGA_BADPARMS, "Invalid or unsupported Targa file")
+JMESSAGE(JERR_TGA_COLORSPACE, "Targa output must be grayscale or RGB")
+JMESSAGE(JTRC_TGA, "%ux%u RGB Targa image")
+JMESSAGE(JTRC_TGA_GRAY, "%ux%u grayscale Targa image")
+JMESSAGE(JTRC_TGA_MAPPED, "%ux%u colormapped Targa image")
+#else
+JMESSAGE(JERR_TGA_NOTCOMP, "Targa support was not compiled")
+#endif /* TARGA_SUPPORTED */
+
+JMESSAGE(JERR_BAD_CMAP_FILE,
+         "Color map file is invalid or of unsupported format")
+JMESSAGE(JERR_TOO_MANY_COLORS,
+         "Output file format cannot handle %d colormap entries")
+JMESSAGE(JERR_UNGETC_FAILED, "ungetc failed")
+#ifdef TARGA_SUPPORTED
+JMESSAGE(JERR_UNKNOWN_FORMAT,
+         "Unrecognized input file format --- perhaps you need -targa")
+#else
+JMESSAGE(JERR_UNKNOWN_FORMAT, "Unrecognized input file format")
+#endif
+JMESSAGE(JERR_UNSUPPORTED_FORMAT, "Unsupported output file format")
+
+#ifdef JMAKE_ENUM_LIST
+
+  JMSG_LASTADDONCODE
+} ADDON_MESSAGE_CODE;
+
+#undef JMAKE_ENUM_LIST
+#endif /* JMAKE_ENUM_LIST */
+
+/* Zap JMESSAGE macro so that future re-inclusions do nothing by default */
+#undef JMESSAGE

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cdjpeg.c Added

@@ -0,0 +1,156 @@
+/*
+ * cdjpeg.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2019, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains common support routines used by the IJG application
+ * programs (cjpeg, djpeg, jpegtran).
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include <ctype.h>              /* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>              /* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>                 /* to declare setmode() */
+#endif
+
+
+/*
+ * Optional progress monitor: display a percent-done figure on stderr.
+ */
+
+METHODDEF(void)
+progress_monitor(j_common_ptr cinfo)
+{
+  cd_progress_ptr prog = (cd_progress_ptr)cinfo->progress;
+
+  if (prog->max_scans != 0 && cinfo->is_decompressor) {
+    int scan_no = ((j_decompress_ptr)cinfo)->input_scan_number;
+
+    if (scan_no > (int)prog->max_scans) {
+      fprintf(stderr, "Scan number %d exceeds maximum scans (%d)\n", scan_no,
+              prog->max_scans);
+      exit(EXIT_FAILURE);
+    }
+  }
+
+  if (prog->report) {
+    int total_passes = prog->pub.total_passes + prog->total_extra_passes;
+    int percent_done =
+      (int)(prog->pub.pass_counter * 100L / prog->pub.pass_limit);
+
+    if (percent_done != prog->percent_done) {
+      prog->percent_done = percent_done;
+      if (total_passes > 1) {
+        fprintf(stderr, "\rPass %d/%d: %3d%% ",
+                prog->pub.completed_passes + prog->completed_extra_passes + 1,
+                total_passes, percent_done);
+      } else {
+        fprintf(stderr, "\r %3d%% ", percent_done);
+      }
+      fflush(stderr);
+    }
+  }
+}
+
+
+GLOBAL(void)
+start_progress_monitor(j_common_ptr cinfo, cd_progress_ptr progress)
+{
+  /* Enable progress display, unless trace output is on */
+  if (cinfo->err->trace_level == 0) {
+    progress->pub.progress_monitor = progress_monitor;
+    progress->completed_extra_passes = 0;
+    progress->total_extra_passes = 0;
+    progress->max_scans = 0;
+    progress->report = FALSE;
+    progress->percent_done = -1;
+    cinfo->progress = &progress->pub;
+  }
+}
+
+
+GLOBAL(void)
+end_progress_monitor(j_common_ptr cinfo)
+{
+  /* Clear away progress display */
+  if (cinfo->err->trace_level == 0) {
+    fprintf(stderr, "\r                \r");
+    fflush(stderr);
+  }
+}
+
+
+/*
+ * Case-insensitive matching of possibly-abbreviated keyword switches.
+ * keyword is the constant keyword (must be lower case already),
+ * minchars is length of minimum legal abbreviation.
+ */
+
+GLOBAL(boolean)
+keymatch(char *arg, const char *keyword, int minchars)
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return FALSE;             /* arg longer than keyword, no good */
+    if (isupper(ca))            /* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return FALSE;             /* no good */
+    nmatched++;                 /* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return FALSE;
+  return TRUE;                  /* A-OK */
+}
+
+
+/*
+ * Routines to establish binary I/O mode for stdin and stdout.
+ * Non-Unix systems often require some hacking to get out of text mode.
+ */
+
+GLOBAL(FILE *)
+read_stdin(void)
+{
+  FILE *input_file = stdin;
+
+#ifdef USE_SETMODE              /* need to hack file mode? */
+  setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+  if ((input_file = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+    fprintf(stderr, "Cannot reopen stdin\n");
+    exit(EXIT_FAILURE);
+  }
+#endif
+  return input_file;
+}
+
+
+GLOBAL(FILE *)
+write_stdout(void)
+{
+  FILE *output_file = stdout;
+
+#ifdef USE_SETMODE              /* need to hack file mode? */
+  setmode(fileno(stdout), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+  if ((output_file = fdopen(fileno(stdout), WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "Cannot reopen stdout\n");
+    exit(EXIT_FAILURE);
+  }
+#endif
+  return output_file;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cdjpeg.h Added

@@ -0,0 +1,161 @@
+/*
+ * cdjpeg.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * Modified 2019 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2017, 2019, 2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains common declarations for the sample applications
+ * cjpeg and djpeg.  It is NOT used by the core JPEG library.
+ */
+
+#define JPEG_CJPEG_DJPEG        /* define proper options in jconfig.h */
+#define JPEG_INTERNAL_OPTIONS   /* cjpeg.c,djpeg.c need to see xxx_SUPPORTED */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"             /* get library error codes too */
+#include "cderror.h"            /* get application-specific error codes */
+
+
+/*
+ * Object interface for cjpeg's source file decoding modules
+ */
+
+typedef struct cjpeg_source_struct *cjpeg_source_ptr;
+
+struct cjpeg_source_struct {
+  void (*start_input) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+  JDIMENSION (*get_pixel_rows) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+  void (*finish_input) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+
+  FILE *input_file;
+
+  JSAMPARRAY buffer;
+  JDIMENSION buffer_height;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  JDIMENSION max_pixels;
+#endif
+};
+
+
+/*
+ * Object interface for djpeg's output file encoding modules
+ */
+
+typedef struct djpeg_dest_struct *djpeg_dest_ptr;
+
+struct djpeg_dest_struct {
+  /* start_output is called after jpeg_start_decompress finishes.
+   * The color map will be ready at this time, if one is needed.
+   */
+  void (*start_output) (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo);
+  /* Emit the specified number of pixel rows from the buffer. */
+  void (*put_pixel_rows) (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                          JDIMENSION rows_supplied);
+  /* Finish up at the end of the image. */
+  void (*finish_output) (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo);
+  /* Re-calculate buffer dimensions based on output dimensions (for use with
+     partial image decompression.)  If this is NULL, then the output format
+     does not support partial image decompression (BMP, in particular, cannot
+     support partial decompression because it uses an inversion buffer to write
+     the image in bottom-up order.) */
+  void (*calc_buffer_dimensions) (j_decompress_ptr cinfo,
+                                  djpeg_dest_ptr dinfo);
+
+
+  /* Target file spec; filled in by djpeg.c after object is created. */
+  FILE *output_file;
+
+  /* Output pixel-row buffer.  Created by module init or start_output.
+   * Width is cinfo->output_width * cinfo->output_components;
+   * height is buffer_height.
+   */
+  JSAMPARRAY buffer;
+  JDIMENSION buffer_height;
+};
+
+
+/*
+ * cjpeg/djpeg may need to perform extra passes to convert to or from
+ * the source/destination file format.  The JPEG library does not know
+ * about these passes, but we'd like them to be counted by the progress
+ * monitor.  We use an expanded progress monitor object to hold the
+ * additional pass count.
+ */
+
+struct cdjpeg_progress_mgr {
+  struct jpeg_progress_mgr pub; /* fields known to JPEG library */
+  int completed_extra_passes;   /* extra passes completed */
+  int total_extra_passes;       /* total extra */
+  JDIMENSION max_scans;         /* abort if the number of scans exceeds this
+                                   value and the value is non-zero */
+  boolean report;               /* whether or not to report progress */
+  /* last printed percentage stored here to avoid multiple printouts */
+  int percent_done;
+};
+
+typedef struct cdjpeg_progress_mgr *cd_progress_ptr;
+
+
+/* Module selection routines for I/O modules. */
+
+EXTERN(cjpeg_source_ptr) jinit_read_bmp(j_compress_ptr cinfo,
+                                        boolean use_inversion_array);
+EXTERN(djpeg_dest_ptr) jinit_write_bmp(j_decompress_ptr cinfo, boolean is_os2,
+                                       boolean use_inversion_array);
+EXTERN(cjpeg_source_ptr) jinit_read_gif(j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_gif(j_decompress_ptr cinfo, boolean is_lzw);
+EXTERN(cjpeg_source_ptr) jinit_read_ppm(j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_ppm(j_decompress_ptr cinfo);
+EXTERN(cjpeg_source_ptr) jinit_read_targa(j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_targa(j_decompress_ptr cinfo);
+
+/* cjpeg support routines (in rdswitch.c) */
+
+EXTERN(boolean) read_quant_tables(j_compress_ptr cinfo, char *filename,
+                                  boolean force_baseline);
+EXTERN(boolean) read_scan_script(j_compress_ptr cinfo, char *filename);
+EXTERN(boolean) set_quality_ratings(j_compress_ptr cinfo, char *arg,
+                                    boolean force_baseline);
+EXTERN(boolean) set_quant_slots(j_compress_ptr cinfo, char *arg);
+EXTERN(boolean) set_sample_factors(j_compress_ptr cinfo, char *arg);
+
+/* djpeg support routines (in rdcolmap.c) */
+
+EXTERN(void) read_color_map(j_decompress_ptr cinfo, FILE *infile);
+
+/* common support routines (in cdjpeg.c) */
+
+EXTERN(void) start_progress_monitor(j_common_ptr cinfo,
+                                    cd_progress_ptr progress);
+EXTERN(void) end_progress_monitor(j_common_ptr cinfo);
+EXTERN(boolean) keymatch(char *arg, const char *keyword, int minchars);
+EXTERN(FILE *) read_stdin(void);
+EXTERN(FILE *) write_stdout(void);
+
+/* miscellaneous useful macros */
+
+#ifdef DONT_USE_B_MODE          /* define mode parameters for fopen() */
+#define READ_BINARY     "r"
+#define WRITE_BINARY    "w"
+#else
+#define READ_BINARY     "rb"
+#define WRITE_BINARY    "wb"
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#define EXIT_SUCCESS  0
+#endif
+#ifndef EXIT_WARNING
+#define EXIT_WARNING  2
+#endif
+
+#define IsExtRGB(cs) \
+  (cs == JCS_RGB || (cs >= JCS_EXT_RGB && cs <= JCS_EXT_ARGB))

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/change.log Added

@@ -0,0 +1,336 @@
+libjpeg-turbo note:  This file has been modified by The libjpeg-turbo Project
+to include only information relevant to libjpeg-turbo.  It is included only for
+reference.  Please see ChangeLog.md for information specific to libjpeg-turbo.
+
+
+CHANGE LOG for Independent JPEG Group's JPEG software
+
+
+Version 9d  12-Jan-2020
+-----------------------
+
+Restore GIF read and write support from libjpeg version 6a.
+Thank to Wolfgang Werner (W.W.) Heinz for suggestion.
+
+Add jpegtran -drop option; add options to the crop extension and wipe
+to fill the extra area with content from the source image region,
+instead of gray out.
+
+
+Version 9c  14-Jan-2018
+-----------------------
+
+jpegtran: add an option to the -wipe switch to fill the region
+with the average of adjacent blocks, instead of gray out.
+Thank to Caitlyn Feddock and Maddie Ziegler for inspiration.
+
+
+Version 9b  17-Jan-2016
+-----------------------
+
+Document 'f' specifier for jpegtran -crop specification.
+Thank to Michele Martone for suggestion.
+
+
+Version 9a  19-Jan-2014
+-----------------------
+
+Add jpegtran -wipe option and extension for -crop.
+Thank to Andrew Senior, David Clunie, and Josef Schmid for suggestion.
+
+
+Version 9  13-Jan-2013
+----------------------
+
+Add remark for jpeg_mem_dest() in jdatadst.c.
+Thank to Elie-Gregoire Khoury for the hint.
+
+Correct argument type in format string, avoid compiler warnings.
+Thank to Vincent Torri for hint.
+
+
+Version 8d  15-Jan-2012
+-----------------------
+
+Add cjpeg -rgb option to create RGB JPEG files.
+Using this switch suppresses the conversion from RGB
+colorspace input to the default YCbCr JPEG colorspace.
+Thank to Michael Koch for the initial suggestion.
+
+Add option to disable the region adjustment in the transupp crop code.
+Thank to Jeffrey Friedl for the suggestion.
+
+
+Version 8b  16-May-2010
+-----------------------
+
+Repair problem in new memory source manager with corrupt JPEG data.
+Thank to Ted Campbell and Samuel Chun for the report.
+
+
+Version 8a  28-Feb-2010
+-----------------------
+
+Writing tables-only datastreams via jpeg_write_tables works again.
+
+Support 32-bit BMPs (RGB image with Alpha channel) for read in cjpeg.
+Thank to Brett Blackham for the suggestion.
+
+
+Version 8  10-Jan-2010
+----------------------
+
+Add sanity check in BMP reader module to avoid cjpeg crash for empty input
+image (thank to Isaev Ildar of ISP RAS, Moscow, RU for reporting this error).
+
+Add data source and destination managers for read from and write to
+memory buffers.  New API functions jpeg_mem_src and jpeg_mem_dest.
+Thank to Roberto Boni from Italy for the suggestion.
+
+
+Version 7  27-Jun-2009
+----------------------
+
+New scaled DCTs implemented.
+djpeg now supports scalings N/8 with all N from 1 to 16.
+
+cjpeg -quality option has been extended for support of separate quality
+settings for luminance and chrominance (or in general, for every provided
+quantization table slot).
+New API function jpeg_default_qtables() and q_scale_factor array in library.
+
+Support arithmetic entropy encoding and decoding.
+Added files jaricom.c, jcarith.c, jdarith.c.
+
+jpegtran has a new "lossless" cropping feature.
+
+Implement -perfect option in jpegtran, new API function
+jtransform_perfect_transform() in transupp. (DP 204_perfect.dpatch)
+
+Better error messages for jpegtran fopen failure.
+(DP 203_jpegtran_errmsg.dpatch)
+
+Fix byte order issue with 16bit PPM/PGM files in rdppm.c/wrppm.c:
+according to Netpbm, the de facto standard implementation of the PNM formats,
+the most significant byte is first. (DP 203_rdppm.dpatch)
+
+Add -raw option to rdjpgcom not to mangle the output.
+(DP 205_rdjpgcom_raw.dpatch)
+
+Make rdjpgcom locale aware. (DP 201_rdjpgcom_locale.dpatch)
+
+Add extern "C" to jpeglib.h.
+This avoids the need to put extern "C" { ... } around #include "jpeglib.h"
+in your C++ application.  Defining the symbol DONT_USE_EXTERN_C in the
+configuration prevents this. (DP 202_jpeglib.h_c++.dpatch)
+
+
+Version 6b  27-Mar-1998
+-----------------------
+
+jpegtran has new features for lossless image transformations (rotation
+and flipping) as well as "lossless" reduction to grayscale.
+
+jpegtran now copies comments by default; it has a -copy switch to enable
+copying all APPn blocks as well, or to suppress comments.  (Formerly it
+always suppressed comments and APPn blocks.)  jpegtran now also preserves
+JFIF version and resolution information.
+
+New decompressor library feature: COM and APPn markers found in the input
+file can be saved in memory for later use by the application.  (Before,
+you had to code this up yourself with a custom marker processor.)
+
+There is an unused field "void * client_data" now in compress and decompress
+parameter structs; this may be useful in some applications.
+
+JFIF version number information is now saved by the decoder and accepted by
+the encoder.  jpegtran uses this to copy the source file's version number,
+to ensure "jpegtran -copy all" won't create bogus files that contain JFXX
+extensions but claim to be version 1.01.  Applications that generate their
+own JFXX extension markers also (finally) have a supported way to cause the
+encoder to emit JFIF version number 1.02.
+
+djpeg's trace mode reports JFIF 1.02 thumbnail images as such, rather
+than as unknown APP0 markers.
+
+In -verbose mode, djpeg and rdjpgcom will try to print the contents of
+APP12 markers as text.  Some digital cameras store useful text information
+in APP12 markers.
+
+Handling of truncated data streams is more robust: blocks beyond the one in
+which the error occurs will be output as uniform gray, or left unchanged
+if decoding a progressive JPEG.  The appearance no longer depends on the
+Huffman tables being used.
+
+Huffman tables are checked for validity much more carefully than before.
+
+The configure script now supports building libjpeg as a shared library
+on many flavors of Unix (all the ones that GNU libtool knows how to
+build shared libraries for).  Use "./configure --enable-shared" to
+try this out.
+
+New jconfig file and makefiles for Microsoft Visual C++ and Developer Studio.
+Also, a jconfig file and a build script for Metrowerks CodeWarrior
+on Apple Macintosh.  makefile.dj has been updated for DJGPP v2, and there
+are miscellaneous other minor improvements in the makefiles.
+
+jmemmac.c now knows how to create temporary files following Mac System 7
+conventions.
+
+djpeg's -map switch is now able to read raw-format PPM files reliably.
+
+cjpeg -progressive -restart no longer generates any unnecessary DRI markers.
+
+Multiple calls to jpeg_simple_progression for a single JPEG object
+no longer leak memory.
+
+
+Version 6a  7-Feb-96
+--------------------
+
+Library initialization sequence modified to detect version mismatches
+and struct field packing mismatches between library and calling application.
+This change requires applications to be recompiled, but does not require
+any application source code change.
+
+All routine declarations changed to the style "GLOBAL(type) name ...",
+that is, GLOBAL, LOCAL, METHODDEF, EXTERN are now macros taking the
+routine's return type as an argument.  This makes it possible to add
+Microsoft-style linkage keywords to all the routines by changing just
+these macros.  Note that any application code that was using these macros
+will have to be changed.
+
+DCT coefficient quantization tables are now stored in normal array order
+rather than zigzag order.  Application code that calls jpeg_add_quant_table,
+or otherwise manipulates quantization tables directly, will need to be
+changed.  If you need to make such code work with either older or newer
+versions of the library, a test like "#if JPEG_LIB_VERSION >= 61" is
+recommended.
+
+djpeg's trace capability now dumps DQT tables in natural order, not zigzag
+order.  This allows the trace output to be made into a "-qtables" file
+more easily.
+
+New system-dependent memory manager module for use on Apple Macintosh.
+
+Fix bug in cjpeg's -smooth option: last one or two scanlines would be
+duplicates of the prior line unless the image height mod 16 was 1 or 2.
+
+Repair minor problems in VMS, BCC, MC6 makefiles.
+
+New configure script based on latest GNU Autoconf.
+
+Correct the list of include files needed by MetroWerks C for ccommand().
+
+Numerous small documentation updates.
+
+
+Version 6  2-Aug-95
+-------------------
+
+Progressive JPEG support: library can read and write full progressive JPEG
+files.  A "buffered image" mode supports incremental decoding for on-the-fly
+display of progressive images.  Simply recompiling an existing IJG-v5-based
+decoder with v6 should allow it to read progressive files, though of course
+without any special progressive display.
+
+New "jpegtran" application performs lossless transcoding between different
+JPEG formats; primarily, it can be used to convert baseline to progressive
+JPEG and vice versa.  In support of jpegtran, the library now allows lossless
+reading and writing of JPEG files as DCT coefficient arrays.  This ability
+may be of use in other applications.
+
+Notes for programmers:
+* We changed jpeg_start_decompress() to be able to suspend; this makes all
+decoding modes available to suspending-input applications.  However,
+existing applications that use suspending input will need to be changed
+to check the return value from jpeg_start_decompress().  You don't need to
+do anything if you don't use a suspending data source.
+* We changed the interface to the virtual array routines: access_virt_array
+routines now take a count of the number of rows to access this time.  The
+last parameter to request_virt_array routines is now interpreted as the
+maximum number of rows that may be accessed at once, but not necessarily
+the height of every access.
+
+
+Version 5b  15-Mar-95
+---------------------
+
+Correct bugs with grayscale images having v_samp_factor > 1.
+
+jpeg_write_raw_data() now supports output suspension.
+
+Correct bugs in "configure" script for case of compiling in
+a directory other than the one containing the source files.
+
+Repair bug in jquant1.c: sometimes didn't use as many colors as it could.
+
+Borland C makefile and jconfig file work under either MS-DOS or OS/2.
+
+Miscellaneous improvements to documentation.
+
+
+Version 5a  7-Dec-94
+--------------------
+
+Changed color conversion roundoff behavior so that grayscale values are
+represented exactly.  (This causes test image files to change.)
+
+Make ordered dither use 16x16 instead of 4x4 pattern for a small quality
+improvement.
+
+New configure script based on latest GNU Autoconf.
+Fix configure script to handle CFLAGS correctly.
+Rename *.auto files to *.cfg, so that configure script still works if
+file names have been truncated for DOS.
+
+Fix bug in rdbmp.c: didn't allow for extra data between header and image.
+
+Modify rdppm.c/wrppm.c to handle 2-byte raw PPM/PGM formats for 12-bit data.
+
+Fix several bugs in rdrle.c.
+
+NEED_SHORT_EXTERNAL_NAMES option was broken.
+
+Revise jerror.h/jerror.c for more flexibility in message table.
+
+Repair oversight in jmemname.c NO_MKTEMP case: file could be there
+but unreadable.
+
+
+Version 5  24-Sep-94
+--------------------
+
+Version 5 represents a nearly complete redesign and rewrite of the IJG
+software.  Major user-visible changes include:
+  * Automatic configuration simplifies installation for most Unix systems.
+  * A range of speed vs. image quality tradeoffs are supported.
+    This includes resizing of an image during decompression: scaling down
+    by a factor of 1/2, 1/4, or 1/8 is handled very efficiently.
+  * New programs rdjpgcom and wrjpgcom allow insertion and extraction
+    of text comments in a JPEG file.
+
+The application programmer's interface to the library has changed completely.
+Notable improvements include:
+  * We have eliminated the use of callback routines for handling the
+    uncompressed image data.  The application now sees the library as a
+    set of routines that it calls to read or write image data on a
+    scanline-by-scanline basis.
+  * The application image data is represented in a conventional interleaved-
+    pixel format, rather than as a separate array for each color channel.
+    This can save a copying step in many programs.
+  * The handling of compressed data has been cleaned up: the application can
+    supply routines to source or sink the compressed data.  It is possible to
+    suspend processing on source/sink buffer overrun, although this is not
+    supported in all operating modes.
+  * All static state has been eliminated from the library, so that multiple
+    instances of compression or decompression can be active concurrently.
+  * JPEG abbreviated datastream formats are supported, ie, quantization and
+    Huffman tables can be stored separately from the image data.
+  * And not only that, but the documentation of the library has improved
+    considerably!
+
+
+The last widely used release before the version 5 rewrite was version 4A of
+18-Feb-93.  Change logs before that point have been discarded, since they
+are not of much interest after the rewrite.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cjpeg.1 Added

@@ -0,0 +1,360 @@
+.TH CJPEG 1 "4 November 2020"
+.SH NAME
+cjpeg \- compress an image file to a JPEG file
+.SH SYNOPSIS
+.B cjpeg
+
+.I options
+
+
+.I filename
+
+.LP
+.SH DESCRIPTION
+.LP
+.B cjpeg
+compresses the named image file, or the standard input if no file is
+named, and produces a JPEG/JFIF file on the standard output.
+The currently supported input file formats are: PPM (PBMPLUS color
+format), PGM (PBMPLUS grayscale format), BMP, GIF, and Targa.
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-grayscale
+may be written
+.B \-gray
+or
+.BR \-gr .
+Most of the "basic" switches can be abbreviated to as little as one letter.
+Upper and lower case are equivalent (thus
+.B \-BMP
+is the same as
+.BR \-bmp ).
+British spellings are also accepted (e.g.,
+.BR \-greyscale ),
+though for brevity these are not mentioned below.
+.PP
+The basic switches are:
+.TP
+.BI \-quality " N,..."
+Scale quantization tables to adjust image quality.  Quality is 0 (worst) to
+100 (best); default is 75.  (See below for more info.)
+.TP
+.B \-grayscale
+Create monochrome JPEG file from color input.  Be sure to use this switch when
+compressing a grayscale BMP or GIF file, because
+.B cjpeg
+isn't bright enough to notice whether a BMP or GIF file uses only shades of
+gray.  By saying
+.BR \-grayscale,
+you'll get a smaller JPEG file that takes less time to process.
+.TP
+.B \-rgb
+Create RGB JPEG file.
+Using this switch suppresses the conversion from RGB
+colorspace input to the default YCbCr JPEG colorspace.
+.TP
+.B \-optimize
+Perform optimization of entropy encoding parameters.  Without this, default
+encoding parameters are used.
+.B \-optimize
+usually makes the JPEG file a little smaller, but
+.B cjpeg
+runs somewhat slower and needs much more memory.  Image quality and speed of
+decompression are unaffected by
+.BR \-optimize .
+.TP
+.B \-progressive
+Create progressive JPEG file (see below).
+.TP
+.B \-targa
+Input file is Targa format.  Targa files that contain an "identification"
+field will not be automatically recognized by
+.BR cjpeg ;
+for such files you must specify
+.B \-targa
+to make
+.B cjpeg
+treat the input as Targa format.
+For most Targa files, you won't need this switch.
+.PP
+The
+.B \-quality
+switch lets you trade off compressed file size against quality of the
+reconstructed image: the higher the quality setting, the larger the JPEG file,
+and the closer the output image will be to the original input.  Normally you
+want to use the lowest quality setting (smallest file) that decompresses into
+something visually indistinguishable from the original image.  For this
+purpose the quality setting should generally be between 50 and 95 (the default
+is 75) for photographic images.  If you see defects at
+.B \-quality
+75, then go up 5 or 10 counts at a time until you are happy with the output
+image.  (The optimal setting will vary from one image to another.)
+.PP
+.B \-quality
+100 will generate a quantization table of all 1's, minimizing loss in the
+quantization step (but there is still information loss in subsampling, as well
+as roundoff error.)  For most images, specifying a quality value above
+about 95 will increase the size of the compressed file dramatically, and while
+the quality gain from these higher quality values is measurable (using metrics
+such as PSNR or SSIM), it is rarely perceivable by human vision.
+.PP
+In the other direction, quality values below 50 will produce very small files
+of low image quality.  Settings around 5 to 10 might be useful in preparing an
+index of a large image library, for example.  Try
+.B \-quality
+2 (or so) for some amusing Cubist effects.  (Note: quality
+values below about 25 generate 2-byte quantization tables, which are
+considered optional in the JPEG standard.
+.B cjpeg
+emits a warning message when you give such a quality value, because some
+other JPEG programs may be unable to decode the resulting file.  Use
+.B \-baseline
+if you need to ensure compatibility at low quality values.)
+.PP
+The \fB-quality\fR option has been extended in this version of \fBcjpeg\fR to
+support separate quality settings for luminance and chrominance (or, in
+general, separate settings for every quantization table slot.)  The principle
+is the same as chrominance subsampling:  since the human eye is more sensitive
+to spatial changes in brightness than spatial changes in color, the chrominance
+components can be quantized more than the luminance components without
+incurring any visible image quality loss.  However, unlike subsampling, this
+feature reduces data in the frequency domain instead of the spatial domain,
+which allows for more fine-grained control.  This option is useful in
+quality-sensitive applications, for which the artifacts generated by
+subsampling may be unacceptable.
+.PP
+The \fB-quality\fR option accepts a comma-separated list of parameters, which
+respectively refer to the quality levels that should be assigned to the
+quantization table slots.  If there are more q-table slots than parameters,
+then the last parameter is replicated.  Thus, if only one quality parameter is
+given, this is used for both luminance and chrominance (slots 0 and 1,
+respectively), preserving the legacy behavior of cjpeg v6b and prior.
+More (or customized) quantization tables can be set with the \fB-qtables\fR
+option and assigned to components with the \fB-qslots\fR option (see the
+"wizard" switches below.)
+.PP
+JPEG files generated with separate luminance and chrominance quality are fully
+compliant with standard JPEG decoders.
+.PP
+.BR CAUTION:
+For this setting to be useful, be sure to pass an argument of \fB-sample 1x1\fR
+to \fBcjpeg\fR to disable chrominance subsampling.  Otherwise, the default
+subsampling level (2x2, AKA "4:2:0") will be used.
+.PP
+The
+.B \-progressive
+switch creates a "progressive JPEG" file.  In this type of JPEG file, the data
+is stored in multiple scans of increasing quality.  If the file is being
+transmitted over a slow communications link, the decoder can use the first
+scan to display a low-quality image very quickly, and can then improve the
+display with each subsequent scan.  The final image is exactly equivalent to a
+standard JPEG file of the same quality setting, and the total file size is
+about the same --- often a little smaller.
+.PP
+Switches for advanced users:
+.TP
+.B \-arithmetic
+Use arithmetic coding.
+.B Caution:
+arithmetic coded JPEG is not yet widely implemented, so many decoders will be
+unable to view an arithmetic coded JPEG file at all.
+.TP
+.B \-dct int
+Use accurate integer DCT method (default).
+.TP
+.B \-dct fast
+Use less accurate integer DCT method legacy feature.
+When the Independent JPEG Group's software was first released in 1991, the
+compression time for a 1-megapixel JPEG image on a mainstream PC was measured
+in minutes.  Thus, the \fBfast\fR integer DCT algorithm provided noticeable
+performance benefits.  On modern CPUs running libjpeg-turbo, however, the
+compression time for a 1-megapixel JPEG image is measured in milliseconds, and
+thus the performance benefits of the \fBfast\fR algorithm are much less
+noticeable.  On modern x86/x86-64 CPUs that support AVX2 instructions, the
+\fBfast\fR and \fBint\fR methods have similar performance.  On other types of
+CPUs, the \fBfast\fR method is generally about 5-15% faster than the \fBint\fR
+method.
+
+For quality levels of 90 and below, there should be little or no perceptible
+quality difference between the two algorithms.  For quality levels above 90,
+however, the difference between the \fBfast\fR and \fBint\fR methods becomes
+more pronounced.  With quality=97, for instance, the \fBfast\fR method incurs
+generally about a 1-3 dB loss in PSNR relative to the \fBint\fR method, but
+this can be larger for some images.  Do not use the \fBfast\fR method with
+quality levels above 97.  The algorithm often degenerates at quality=98 and
+above and can actually produce a more lossy image than if lower quality levels
+had been used.  Also, in libjpeg-turbo, the \fBfast\fR method is not fully
+accelerated for quality levels above 97, so it will be slower than the
+\fBint\fR method.
+.TP
+.B \-dct float
+Use floating-point DCT method legacy feature.
+The \fBfloat\fR method does not produce significantly more accurate results
+than the \fBint\fR method, and it is much slower.  The \fBfloat\fR method may
+also give different results on different machines due to varying roundoff
+behavior, whereas the integer methods should give the same results on all
+machines.
+.TP
+.BI \-icc " file"
+Embed ICC color management profile contained in the specified file.
+.TP
+.BI \-restart " N"
+Emit a JPEG restart marker every N MCU rows, or every N MCU blocks if "B" is
+attached to the number.
+.B \-restart 0
+(the default) means no restart markers.
+.TP
+.BI \-smooth " N"
+Smooth the input image to eliminate dithering noise.  N, ranging from 1 to
+100, indicates the strength of smoothing.  0 (the default) means no smoothing.
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, an error will occur.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.BI \-memdst
+Compress to memory instead of a file.  This feature was implemented mainly as a
+way of testing the in-memory destination manager (jpeg_mem_dest()), but it is
+also useful for benchmarking, since it reduces the I/O overhead.
+.TP
+.BI \-report
+Report compression progress.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.TP
+.B \-version
+Print version information and exit.
+.PP
+The
+.B \-restart
+option inserts extra markers that allow a JPEG decoder to resynchronize after
+a transmission error.  Without restart markers, any damage to a compressed
+file will usually ruin the image from the point of the error to the end of the
+image; with restart markers, the damage is usually confined to the portion of
+the image up to the next restart marker.  Of course, the restart markers
+occupy extra space.  We recommend
+.B \-restart 1
+for images that will be transmitted across unreliable networks such as Usenet.
+.PP
+The
+.B \-smooth
+option filters the input to eliminate fine-scale noise.  This is often useful
+when converting dithered images to JPEG: a moderate smoothing factor of 10 to
+50 gets rid of dithering patterns in the input file, resulting in a smaller
+JPEG file and a better-looking image.  Too large a smoothing factor will
+visibly blur the image, however.
+.PP
+Switches for wizards:
+.TP
+.B \-baseline
+Force baseline-compatible quantization tables to be generated.  This clamps
+quantization values to 8 bits even at low quality settings.  (This switch is
+poorly named, since it does not ensure that the output is actually baseline
+JPEG.  For example, you can use
+.B \-baseline
+and
+.B \-progressive
+together.)
+.TP
+.BI \-qtables " file"
+Use the quantization tables given in the specified text file.
+.TP
+.BI \-qslots " N,..."
+Select which quantization table to use for each color component.
+.TP
+.BI \-sample " HxV,..."
+Set JPEG sampling factors for each color component.
+.TP
+.BI \-scans " file"
+Use the scan script given in the specified text file.
+.PP
+The "wizard" switches are intended for experimentation with JPEG.  If you
+don't know what you are doing, \fBdon't use them\fR.  These switches are
+documented further in the file wizard.txt.
+.SH EXAMPLES
+.LP
+This example compresses the PPM file foo.ppm with a quality factor of
+60 and saves the output as foo.jpg:
+.IP
+.B cjpeg \-quality
+.I 60 foo.ppm
+.B >
+.I foo.jpg
+.SH HINTS
+Color GIF files are not the ideal input for JPEG; JPEG is really intended for
+compressing full-color (24-bit) images.  In particular, don't try to convert
+cartoons, line drawings, and other images that have only a few distinct
+colors.  GIF works great on these, JPEG does not.  If you want to convert a
+GIF to JPEG, you should experiment with
+.BR cjpeg 's
+.B \-quality
+and
+.B \-smooth
+options to get a satisfactory conversion.
+.B \-smooth 10
+or so is often helpful.
+.PP
+Avoid running an image through a series of JPEG compression/decompression
+cycles.  Image quality loss will accumulate; after ten or so cycles the image
+may be noticeably worse than it was after one cycle.  It's best to use a
+lossless format while manipulating an image, then convert to JPEG format when
+you are ready to file the image away.
+.PP
+The
+.B \-optimize
+option to
+.B cjpeg
+is worth using when you are making a "final" version for posting or archiving.
+It's also a win when you are using low quality settings to make very small
+JPEG files; the percentage improvement is often a lot more than it is on
+larger files.  (At present,
+.B \-optimize
+mode is always selected when generating progressive JPEG files.)
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+.BR ppm (5),
+.BR pgm (5)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.PP
+This file was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo, to wordsmith certain sections, and to describe
+features not present in libjpeg.
+.SH ISSUES
+Not all variants of BMP and Targa file formats are supported.
+.PP
+The
+.B \-targa
+switch is not a bug, it's a feature.  (It would be a bug if the Targa format
+designers had not been clueless.)

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cjpeg.c Added

@@ -0,0 +1,756 @@
+/*
+ * cjpeg.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * Modified 2003-2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2013-2014, 2017, 2019-2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a command-line user interface for the JPEG compressor.
+ * It should work on any system with Unix- or MS-DOS-style command lines.
+ *
+ * Two different command line styles are permitted, depending on the
+ * compile-time switch TWO_FILE_COMMANDLINE:
+ *      cjpeg options  inputfile outputfile
+ *      cjpeg options  inputfile
+ * In the second style, output is always to standard output, which you'd
+ * normally redirect to a file or pipe to some other program.  Input is
+ * either from a named file or from standard input (typically redirected).
+ * The second style is convenient on Unix but is unhelpful on systems that
+ * don't support pipes.  Also, you MUST use the first style if your system
+ * doesn't do binary I/O to stdin/stdout.
+ * To simplify script writing, the "-outfile" switch is provided.  The syntax
+ *      cjpeg options  -outfile outputfile  inputfile
+ * works regardless of which command line style is used.
+ */
+
+#ifdef CJPEG_FUZZER
+#define JPEG_INTERNALS
+#endif
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "jversion.h"           /* for version message */
+#include "jconfigint.h"
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc(size_t size);
+extern void free(void *ptr);
+#endif
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+
+/* Create the add-on message string table. */
+
+#define JMESSAGE(code, string)  string,
+
+static const char * const cdjpeg_message_table = {
+#include "cderror.h"
+  NULL
+};
+
+
+/*
+ * This routine determines what format the input file is,
+ * and selects the appropriate input-reading module.
+ *
+ * To determine which family of input formats the file belongs to,
+ * we may look only at the first byte of the file, since C does not
+ * guarantee that more than one character can be pushed back with ungetc.
+ * Looking at additional bytes would require one of these approaches:
+ *     1) assume we can fseek() the input file (fails for piped input);
+ *     2) assume we can push back more than one character (works in
+ *        some C implementations, but unportable);
+ *     3) provide our own buffering (breaks input readers that want to use
+ *        stdio directly);
+ * or  4) don't put back the data, and modify the input_init methods to assume
+ *        they start reading after the start of file.
+ * #1 is attractive for MS-DOS but is untenable on Unix.
+ *
+ * The most portable solution for file types that can't be identified by their
+ * first byte is to make the user tell us what they are.  This is also the
+ * only approach for "raw" file types that contain only arbitrary values.
+ * We presently apply this method for Targa files.  Most of the time Targa
+ * files start with 0x00, so we recognize that case.  Potentially, however,
+ * a Targa file could start with any byte value (byte 0 is the length of the
+ * seldom-used ID field), so we provide a switch to force Targa input mode.
+ */
+
+static boolean is_targa;        /* records user -targa switch */
+
+
+LOCAL(cjpeg_source_ptr)
+select_file_type(j_compress_ptr cinfo, FILE *infile)
+{
+  int c;
+
+  if (is_targa) {
+#ifdef TARGA_SUPPORTED
+    return jinit_read_targa(cinfo);
+#else
+    ERREXIT(cinfo, JERR_TGA_NOTCOMP);
+#endif
+  }
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(cinfo, JERR_INPUT_EMPTY);
+  if (ungetc(c, infile) == EOF)
+    ERREXIT(cinfo, JERR_UNGETC_FAILED);
+
+  switch (c) {
+#ifdef BMP_SUPPORTED
+  case 'B':
+    return jinit_read_bmp(cinfo, TRUE);
+#endif
+#ifdef GIF_SUPPORTED
+  case 'G':
+    return jinit_read_gif(cinfo);
+#endif
+#ifdef PPM_SUPPORTED
+  case 'P':
+    return jinit_read_ppm(cinfo);
+#endif
+#ifdef TARGA_SUPPORTED
+  case 0x00:
+    return jinit_read_targa(cinfo);
+#endif
+  default:
+    ERREXIT(cinfo, JERR_UNKNOWN_FORMAT);
+    break;
+  }
+
+  return NULL;                  /* suppress compiler warnings */
+}
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char *progname;    /* program name for error messages */
+static char *icc_filename;      /* for -icc switch */
+static char *outfilename;       /* for -outfile switch */
+boolean memdst;                 /* for -memdst switch */
+boolean report;                 /* for -report switch */
+
+
+#ifdef CJPEG_FUZZER
+
+#include <setjmp.h>
+
+struct my_error_mgr {
+  struct jpeg_error_mgr pub;
+  jmp_buf setjmp_buffer;
+};
+
+void my_error_exit(j_common_ptr cinfo)
+{
+  struct my_error_mgr *myerr = (struct my_error_mgr *)cinfo->err;
+
+  longjmp(myerr->setjmp_buffer, 1);
+}
+
+static void my_emit_message(j_common_ptr cinfo, int msg_level)
+{
+  if (msg_level < 0)
+    cinfo->err->num_warnings++;
+}
+
+#define HANDLE_ERROR() { \
+  if (cinfo.global_state > CSTATE_START) { \
+    if (memdst && outbuffer) \
+      (*cinfo.dest->term_destination) (&cinfo); \
+    jpeg_abort_compress(&cinfo); \
+  } \
+  jpeg_destroy_compress(&cinfo); \
+  if (input_file != stdin && input_file != NULL) \
+    fclose(input_file); \
+  if (memdst) \
+    free(outbuffer); \
+  return EXIT_FAILURE; \
+}
+
+#endif
+
+
+LOCAL(void)
+usage(void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s switches ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "inputfile\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -quality N,...   Compression quality (0..100; 5-95 is most useful range,\n");
+  fprintf(stderr, "                     default is 75)\n");
+  fprintf(stderr, "  -grayscale     Create monochrome JPEG file\n");
+  fprintf(stderr, "  -rgb           Create RGB JPEG file\n");
+#ifdef ENTROPY_OPT_SUPPORTED
+  fprintf(stderr, "  -optimize      Optimize Huffman table (smaller file, but slow compression)\n");
+#endif
+#ifdef C_PROGRESSIVE_SUPPORTED
+  fprintf(stderr, "  -progressive   Create progressive JPEG file\n");
+#endif
+#ifdef TARGA_SUPPORTED
+  fprintf(stderr, "  -targa         Input file is Targa format (usually not needed)\n");
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef C_ARITH_CODING_SUPPORTED
+  fprintf(stderr, "  -arithmetic    Use arithmetic coding\n");
+#endif
+#ifdef DCT_ISLOW_SUPPORTED
+  fprintf(stderr, "  -dct int       Use accurate integer DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_ISLOW ? " (default)" : ""));
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  fprintf(stderr, "  -dct fast      Use less accurate integer DCT method legacy feature%s\n",
+          (JDCT_DEFAULT == JDCT_IFAST ? " (default)" : ""));
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  fprintf(stderr, "  -dct float     Use floating-point DCT method legacy feature%s\n",
+          (JDCT_DEFAULT == JDCT_FLOAT ? " (default)" : ""));
+#endif
+  fprintf(stderr, "  -icc FILE      Embed ICC profile contained in FILE\n");
+  fprintf(stderr, "  -restart N     Set restart interval in rows, or in blocks with B\n");
+#ifdef INPUT_SMOOTHING_SUPPORTED
+  fprintf(stderr, "  -smooth N      Smooth dithered input (N=1..100 is strength)\n");
+#endif
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  fprintf(stderr, "  -memdst        Compress to memory instead of file (useful for benchmarking)\n");
+#endif
+  fprintf(stderr, "  -report        Report compression progress\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "  -version       Print version information and exit\n");
+  fprintf(stderr, "Switches for wizards:\n");
+  fprintf(stderr, "  -baseline      Force baseline quantization tables\n");
+  fprintf(stderr, "  -qtables FILE  Use quantization tables given in FILE\n");
+  fprintf(stderr, "  -qslots N,...    Set component quantization tables\n");
+  fprintf(stderr, "  -sample HxV,...  Set component sampling factors\n");
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  fprintf(stderr, "  -scans FILE    Create multi-scan JPEG per script FILE\n");
+#endif
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(int)
+parse_switches(j_compress_ptr cinfo, int argc, char **argv,
+               int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char *arg;
+  boolean force_baseline;
+  boolean simple_progressive;
+  char *qualityarg = NULL;      /* saves -quality parm if any */
+  char *qtablefile = NULL;      /* saves -qtables filename if any */
+  char *qslotsarg = NULL;       /* saves -qslots parm if any */
+  char *samplearg = NULL;       /* saves -sample parm if any */
+  char *scansarg = NULL;        /* saves -scans parm if any */
+
+  /* Set up default JPEG parameters. */
+
+  force_baseline = FALSE;       /* by default, allow 16-bit quantizers */
+  simple_progressive = FALSE;
+  is_targa = FALSE;
+  icc_filename = NULL;
+  outfilename = NULL;
+  memdst = FALSE;
+  report = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argvargn;
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+        outfilename = NULL;     /* -outfile applies to just one input file */
+        continue;               /* ignore this name if previously processed */
+      }
+      break;                    /* else done parsing switches */
+    }
+    arg++;                      /* advance past switch marker character */
+
+    if (keymatch(arg, "arithmetic", 1)) {
+      /* Use arithmetic coding. */
+#ifdef C_ARITH_CODING_SUPPORTED
+      cinfo->arith_code = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, arithmetic coding not supported\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "baseline", 1)) {
+      /* Force baseline-compatible output (8-bit quantizer values). */
+      force_baseline = TRUE;
+
+    } else if (keymatch(arg, "dct", 2)) {
+      /* Select DCT algorithm. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argvargn, "int", 1)) {
+        cinfo->dct_method = JDCT_ISLOW;
+      } else if (keymatch(argvargn, "fast", 2)) {
+        cinfo->dct_method = JDCT_IFAST;
+      } else if (keymatch(argvargn, "float", 2)) {
+        cinfo->dct_method = JDCT_FLOAT;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (!printed_version) {
+        fprintf(stderr, "%s version %s (build %s)\n",
+                PACKAGE_NAME, VERSION, BUILD);
+        fprintf(stderr, "%s\n\n", JCOPYRIGHT);
+        fprintf(stderr, "Emulating The Independent JPEG Group's software, version %s\n\n",
+                JVERSION);
+        printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "version", 4)) {
+      fprintf(stderr, "%s version %s (build %s)\n",
+              PACKAGE_NAME, VERSION, BUILD);
+      exit(EXIT_SUCCESS);
+
+    } else if (keymatch(arg, "grayscale", 2) ||
+               keymatch(arg, "greyscale", 2)) {
+      /* Force a monochrome JPEG file to be generated. */
+      jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+
+    } else if (keymatch(arg, "rgb", 3)) {
+      /* Force an RGB JPEG file to be generated. */
+      jpeg_set_colorspace(cinfo, JCS_RGB);
+
+    } else if (keymatch(arg, "icc", 1)) {
+      /* Set ICC filename. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      icc_filename = argvargn;
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (ch == 'm' || ch == 'M')
+        lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "optimize", 1) || keymatch(arg, "optimise", 1)) {
+      /* Enable entropy parm optimization. */
+#ifdef ENTROPY_OPT_SUPPORTED
+      cinfo->optimize_coding = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, entropy optimization was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      outfilename = argvargn; /* save it away for later use */
+
+    } else if (keymatch(arg, "progressive", 1)) {
+      /* Select simple progressive mode. */
+#ifdef C_PROGRESSIVE_SUPPORTED
+      simple_progressive = TRUE;
+      /* We must postpone execution until num_components is known. */
+#else
+      fprintf(stderr, "%s: sorry, progressive output was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "memdst", 2)) {
+      /* Use in-memory destination manager */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+      memdst = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, in-memory destination manager was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "quality", 1)) {
+      /* Quality ratings (quantization table scaling factors). */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      qualityarg = argvargn;
+
+    } else if (keymatch(arg, "qslots", 2)) {
+      /* Quantization table slot numbers. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      qslotsarg = argvargn;
+      /* Must delay setting qslots until after we have processed any
+       * colorspace-determining switches, since jpeg_set_colorspace sets
+       * default quant table numbers.
+       */
+
+    } else if (keymatch(arg, "qtables", 2)) {
+      /* Quantization tables fetched from file. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      qtablefile = argvargn;
+      /* We postpone actually reading the file in case -quality comes later. */
+
+    } else if (keymatch(arg, "report", 3)) {
+      report = TRUE;
+
+    } else if (keymatch(arg, "restart", 1)) {
+      /* Restart interval in MCU rows (or in MCUs with 'b'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (lval < 0 || lval > 65535L)
+        usage();
+      if (ch == 'b' || ch == 'B') {
+        cinfo->restart_interval = (unsigned int)lval;
+        cinfo->restart_in_rows = 0; /* else prior '-restart n' overrides me */
+      } else {
+        cinfo->restart_in_rows = (int)lval;
+        /* restart_interval will be computed during startup */
+      }
+
+    } else if (keymatch(arg, "sample", 2)) {
+      /* Set sampling factors. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      samplearg = argvargn;
+      /* Must delay setting sample factors until after we have processed any
+       * colorspace-determining switches, since jpeg_set_colorspace sets
+       * default sampling factors.
+       */
+
+    } else if (keymatch(arg, "scans", 4)) {
+      /* Set scan script. */
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      scansarg = argvargn;
+      /* We must postpone reading the file in case -progressive appears. */
+#else
+      fprintf(stderr, "%s: sorry, multi-scan output was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "smooth", 2)) {
+      /* Set input smoothing factor. */
+      int val;
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%d", &val) != 1)
+        usage();
+      if (val < 0 || val > 100)
+        usage();
+      cinfo->smoothing_factor = val;
+
+    } else if (keymatch(arg, "targa", 1)) {
+      /* Input file is Targa format. */
+      is_targa = TRUE;
+
+    } else {
+      usage();                  /* bogus switch */
+    }
+  }
+
+  /* Post-switch-scanning cleanup */
+
+  if (for_real) {
+
+    /* Set quantization tables for selected quality. */
+    /* Some or all may be overridden if -qtables is present. */
+    if (qualityarg != NULL)     /* process -quality if it was present */
+      if (!set_quality_ratings(cinfo, qualityarg, force_baseline))
+        usage();
+
+    if (qtablefile != NULL)     /* process -qtables if it was present */
+      if (!read_quant_tables(cinfo, qtablefile, force_baseline))
+        usage();
+
+    if (qslotsarg != NULL)      /* process -qslots if it was present */
+      if (!set_quant_slots(cinfo, qslotsarg))
+        usage();
+
+    if (samplearg != NULL)      /* process -sample if it was present */
+      if (!set_sample_factors(cinfo, samplearg))
+        usage();
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+    if (simple_progressive)     /* process -progressive; -scans can override */
+      jpeg_simple_progression(cinfo);
+#endif
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    if (scansarg != NULL)       /* process -scans if it was present */
+      if (!read_scan_script(cinfo, scansarg))
+        usage();
+#endif
+  }
+
+  return argn;                  /* return index of next arg (file name) */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main(int argc, char **argv)
+{
+  struct jpeg_compress_struct cinfo;
+#ifdef CJPEG_FUZZER
+  struct my_error_mgr myerr;
+  struct jpeg_error_mgr &jerr = myerr.pub;
+#else
+  struct jpeg_error_mgr jerr;
+#endif
+  struct cdjpeg_progress_mgr progress;
+  int file_index;
+  cjpeg_source_ptr src_mgr;
+  FILE *input_file = NULL;
+  FILE *icc_file;
+  JOCTET *icc_profile = NULL;
+  long icc_len = 0;
+  FILE *output_file = NULL;
+  unsigned char *outbuffer = NULL;
+  unsigned long outsize = 0;
+  JDIMENSION num_scanlines;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv0;
+  if (progname == NULL || progname0 == 0)
+    progname = "cjpeg";         /* in case C library doesn't provide it */
+
+  /* Initialize the JPEG compression object with default error handling. */
+  cinfo.err = jpeg_std_error(&jerr);
+  jpeg_create_compress(&cinfo);
+  /* Add some application-specific error messages (from cderror.h) */
+  jerr.addon_message_table = cdjpeg_message_table;
+  jerr.first_addon_message = JMSG_FIRSTADDONCODE;
+  jerr.last_addon_message = JMSG_LASTADDONCODE;
+
+  /* Initialize JPEG parameters.
+   * Much of this may be overridden later.
+   * In particular, we don't yet know the input file's color space,
+   * but we need to provide some value for jpeg_set_defaults() to work.
+   */
+
+  cinfo.in_color_space = JCS_RGB; /* arbitrary guess */
+  jpeg_set_defaults(&cinfo);
+
+  /* Scan command line to find file names.
+   * It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are ignored; we will rescan the switches after opening
+   * the input file.
+   */
+
+  file_index = parse_switches(&cinfo, argc, argv, 0, FALSE);
+
+#ifdef TWO_FILE_COMMANDLINE
+  if (!memdst) {
+    /* Must have either -outfile switch or explicit output file name */
+    if (outfilename == NULL) {
+      if (file_index != argc - 2) {
+        fprintf(stderr, "%s: must name one input and one output file\n",
+                progname);
+        usage();
+      }
+      outfilename = argvfile_index + 1;
+    } else {
+      if (file_index != argc - 1) {
+        fprintf(stderr, "%s: must name one input and one output file\n",
+                progname);
+        usage();
+      }
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc - 1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((input_file = fopen(argvfile_index, READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argvfile_index);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    input_file = read_stdin();
+  }
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((output_file = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else if (!memdst) {
+    /* default output file is stdout */
+    output_file = write_stdout();
+  }
+
+  if (icc_filename != NULL) {
+    if ((icc_file = fopen(icc_filename, READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, icc_filename);
+      exit(EXIT_FAILURE);
+    }
+    if (fseek(icc_file, 0, SEEK_END) < 0 ||
+        (icc_len = ftell(icc_file)) < 1 ||
+        fseek(icc_file, 0, SEEK_SET) < 0) {
+      fprintf(stderr, "%s: can't determine size of %s\n", progname,
+              icc_filename);
+      exit(EXIT_FAILURE);
+    }
+    if ((icc_profile = (JOCTET *)malloc(icc_len)) == NULL) {
+      fprintf(stderr, "%s: can't allocate memory for ICC profile\n", progname);
+      fclose(icc_file);
+      exit(EXIT_FAILURE);
+    }
+    if (fread(icc_profile, icc_len, 1, icc_file) < 1) {
+      fprintf(stderr, "%s: can't read ICC profile from %s\n", progname,
+              icc_filename);
+      free(icc_profile);
+      fclose(icc_file);
+      exit(EXIT_FAILURE);
+    }
+    fclose(icc_file);
+  }
+
+#ifdef CJPEG_FUZZER
+  jerr.error_exit = my_error_exit;
+  jerr.emit_message = my_emit_message;
+  if (setjmp(myerr.setjmp_buffer))
+    HANDLE_ERROR()
+#endif
+
+  if (report) {
+    start_progress_monitor((j_common_ptr)&cinfo, &progress);
+    progress.report = report;
+  }
+
+  /* Figure out the input file format, and set up to read it. */
+  src_mgr = select_file_type(&cinfo, input_file);
+  src_mgr->input_file = input_file;
+#ifdef CJPEG_FUZZER
+  src_mgr->max_pixels = 1048576;
+#endif
+
+  /* Read the input file header to obtain file size & colorspace. */
+  (*src_mgr->start_input) (&cinfo, src_mgr);
+
+  /* Now that we know input colorspace, fix colorspace-dependent defaults */
+  jpeg_default_colorspace(&cinfo);
+
+  /* Adjust default compression parameters by re-parsing the options */
+  file_index = parse_switches(&cinfo, argc, argv, 0, TRUE);
+
+  /* Specify data destination for compression */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  if (memdst)
+    jpeg_mem_dest(&cinfo, &outbuffer, &outsize);
+  else
+#endif
+    jpeg_stdio_dest(&cinfo, output_file);
+
+#ifdef CJPEG_FUZZER
+  if (setjmp(myerr.setjmp_buffer))
+    HANDLE_ERROR()
+#endif
+
+  /* Start compressor */
+  jpeg_start_compress(&cinfo, TRUE);
+
+  if (icc_profile != NULL)
+    jpeg_write_icc_profile(&cinfo, icc_profile, (unsigned int)icc_len);
+
+  /* Process data */
+  while (cinfo.next_scanline < cinfo.image_height) {
+    num_scanlines = (*src_mgr->get_pixel_rows) (&cinfo, src_mgr);
+    (void)jpeg_write_scanlines(&cinfo, src_mgr->buffer, num_scanlines);
+  }
+
+  /* Finish compression and release memory */
+  (*src_mgr->finish_input) (&cinfo, src_mgr);
+  jpeg_finish_compress(&cinfo);
+  jpeg_destroy_compress(&cinfo);
+
+  /* Close files, if we opened them */
+  if (input_file != stdin)
+    fclose(input_file);
+  if (output_file != stdout && output_file != NULL)
+    fclose(output_file);
+
+  if (report)
+    end_progress_monitor((j_common_ptr)&cinfo);
+
+  if (memdst) {
+#ifndef CJPEG_FUZZER
+    fprintf(stderr, "Compressed size:  %lu bytes\n", outsize);
+#endif
+    free(outbuffer);
+  }
+
+  free(icc_profile);
+
+  /* All done. */
+  return (jerr.num_warnings ? EXIT_WARNING : EXIT_SUCCESS);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cmakescripts Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cmakescripts/BuildPackages.cmake Added

@@ -0,0 +1,183 @@
+# This file is included from the top-level CMakeLists.txt.  We just store it
+# here to avoid cluttering up that file.
+
+set(PKGNAME ${CMAKE_PROJECT_NAME} CACHE STRING
+  "Distribution package name (default: ${CMAKE_PROJECT_NAME})")
+set(PKGVENDOR "The ${CMAKE_PROJECT_NAME} Project" CACHE STRING
+  "Vendor name to be included in distribution package descriptions (default: The ${CMAKE_PROJECT_NAME} Project)")
+set(PKGURL "http://www.${CMAKE_PROJECT_NAME}.org" CACHE STRING
+  "URL of project web site to be included in distribution package descriptions (default: http://www.${CMAKE_PROJECT_NAME}.org)")
+set(PKGEMAIL "information@${CMAKE_PROJECT_NAME}.org" CACHE STRING
+  "E-mail of project maintainer to be included in distribution package descriptions (default: information@${CMAKE_PROJECT_NAME}.org")
+set(PKGID "com.${CMAKE_PROJECT_NAME}.${PKGNAME}" CACHE STRING
+  "Globally unique package identifier (reverse DNS notation) (default: com.${CMAKE_PROJECT_NAME}.${PKGNAME})")
+
+
+###############################################################################
+# Linux RPM and DEB
+###############################################################################
+
+if(CMAKE_SYSTEM_NAME STREQUAL "Linux")
+
+set(RPMARCH ${CMAKE_SYSTEM_PROCESSOR})
+if(CPU_TYPE STREQUAL "x86_64")
+  set(DEBARCH amd64)
+elseif(CPU_TYPE STREQUAL "arm64")
+  set(DEBARCH ${CPU_TYPE})
+elseif(CPU_TYPE STREQUAL "arm")
+  check_c_source_compiles("
+    #if __ARM_PCS_VFP != 1
+    #error \"float ABI != hard\"
+    #endif
+    int main(void) { return 0; }" HAVE_HARD_FLOAT)
+  if(HAVE_HARD_FLOAT)
+    set(RPMARCH armv7hl)
+    set(DEBARCH armhf)
+  else()
+    set(RPMARCH armel)
+    set(DEBARCH armel)
+  endif()
+elseif(CMAKE_SYSTEM_PROCESSOR_LC STREQUAL "ppc64le")
+  set(DEBARCH ppc64el)
+elseif(CPU_TYPE STREQUAL "powerpc" AND BITS EQUAL 32)
+  set(RPMARCH ppc)
+  set(DEBARCH ppc)
+else()
+  set(DEBARCH ${CMAKE_SYSTEM_PROCESSOR})
+endif()
+message(STATUS "RPM architecture = ${RPMARCH}, DEB architecture = ${DEBARCH}")
+
+# Re-set CMAKE_POSITION_INDEPENDENT_CODE so that the RPM spec file works
+# properly
+boolean_number(CMAKE_POSITION_INDEPENDENT_CODE)
+
+configure_file(release/makerpm.in pkgscripts/makerpm)
+configure_file(release/rpm.spec.in pkgscripts/rpm.spec @ONLY)
+
+add_custom_target(rpm pkgscripts/makerpm
+  SOURCES pkgscripts/makerpm)
+
+configure_file(release/makesrpm.in pkgscripts/makesrpm)
+
+add_custom_target(srpm pkgscripts/makesrpm
+  SOURCES pkgscripts/makesrpm
+  DEPENDS dist)
+
+configure_file(release/makedpkg.in pkgscripts/makedpkg)
+configure_file(release/deb-control.in pkgscripts/deb-control)
+
+add_custom_target(deb pkgscripts/makedpkg
+  SOURCES pkgscripts/makedpkg)
+
+endif() # Linux
+
+
+###############################################################################
+# Windows installer (NullSoft Installer)
+###############################################################################
+
+if(WIN32)
+
+if(MSVC)
+  set(INST_PLATFORM "Visual C++")
+  set(INST_ID vc)
+  set(INST_NAME ${CMAKE_PROJECT_NAME}-${VERSION}-${INST_ID})
+  set(INST_REG_NAME ${CMAKE_PROJECT_NAME})
+elseif(MINGW)
+  set(INST_PLATFORM GCC)
+  set(INST_ID gcc)
+  set(INST_NAME ${CMAKE_PROJECT_NAME}-${VERSION}-${INST_ID})
+  set(INST_REG_NAME ${CMAKE_PROJECT_NAME}-${INST_ID})
+  set(INST_DEFS -DGCC)
+endif()
+
+if(BITS EQUAL 64)
+  set(INST_PLATFORM "${INST_PLATFORM} 64-bit")
+  set(INST_NAME ${INST_NAME}64)
+  set(INST_REG_NAME ${INST_REG_NAME}64)
+  set(INST_DEFS ${INST_DEFS} -DWIN64)
+endif()
+
+if(WITH_JAVA)
+  set(INST_DEFS ${INST_DEFS} -DJAVA)
+endif()
+
+if(MSVC_IDE)
+  set(INST_DEFS ${INST_DEFS} "-DBUILDDIR=${CMAKE_CFG_INTDIR}\\")
+else()
+  set(INST_DEFS ${INST_DEFS} "-DBUILDDIR=")
+endif()
+
+string(REGEX REPLACE "/" "\\\\" INST_DIR ${CMAKE_INSTALL_PREFIX})
+
+configure_file(release/installer.nsi.in installer.nsi @ONLY)
+# TODO: It would be nice to eventually switch to CPack and eliminate this mess,
+# but not today.
+configure_file(win/projectTargets.cmake.in
+  win/${CMAKE_PROJECT_NAME}Targets.cmake @ONLY)
+configure_file(win/${INST_ID}/projectTargets-release.cmake.in
+  win/${CMAKE_PROJECT_NAME}Targets-release.cmake @ONLY)
+
+if(WITH_JAVA)
+  set(JAVA_DEPEND turbojpeg-java)
+endif()
+add_custom_target(installer
+  makensis -nocd ${INST_DEFS} installer.nsi
+  DEPENDS jpeg jpeg-static turbojpeg turbojpeg-static rdjpgcom wrjpgcom
+    cjpeg djpeg jpegtran tjbench ${JAVA_DEPEND}
+  SOURCES installer.nsi)
+
+endif() # WIN32
+
+
+###############################################################################
+# Mac DMG
+###############################################################################
+
+if(APPLE)
+
+set(ARMV8_BUILD "" CACHE PATH
+  "Directory containing Armv8 iOS or macOS build to include in universal binaries")
+
+set(MACOS_APP_CERT_NAME "" CACHE STRING
+  "Name of the Developer ID Application certificate (in the macOS keychain) that should be used to sign the libjpeg-turbo DMG.  Leave this blank to generate an unsigned DMG.")
+set(MACOS_INST_CERT_NAME "" CACHE STRING
+  "Name of the Developer ID Installer certificate (in the macOS keychain) that should be used to sign the libjpeg-turbo installer package.  Leave this blank to generate an unsigned package.")
+
+configure_file(release/makemacpkg.in pkgscripts/makemacpkg)
+configure_file(release/Distribution.xml.in pkgscripts/Distribution.xml)
+configure_file(release/Welcome.rtf.in pkgscripts/Welcome.rtf)
+configure_file(release/uninstall.in pkgscripts/uninstall)
+
+add_custom_target(dmg pkgscripts/makemacpkg
+  SOURCES pkgscripts/makemacpkg)
+
+endif() # APPLE
+
+
+###############################################################################
+# Generic
+###############################################################################
+
+add_custom_target(dist
+  COMMAND git archive --prefix=${CMAKE_PROJECT_NAME}-${VERSION}/ HEAD |
+    gzip > ${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_PROJECT_NAME}-${VERSION}.tar.gz
+    WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR})
+
+configure_file(release/maketarball.in pkgscripts/maketarball)
+
+add_custom_target(tarball pkgscripts/maketarball
+  SOURCES pkgscripts/maketarball)
+
+configure_file(release/libjpeg.pc.in pkgscripts/libjpeg.pc @ONLY)
+
+configure_file(release/libturbojpeg.pc.in pkgscripts/libturbojpeg.pc @ONLY)
+
+include(CMakePackageConfigHelpers)
+write_basic_package_version_file(
+  pkgscripts/${CMAKE_PROJECT_NAME}ConfigVersion.cmake
+  VERSION ${VERSION} COMPATIBILITY AnyNewerVersion)
+
+configure_package_config_file(release/Config.cmake.in
+  pkgscripts/${CMAKE_PROJECT_NAME}Config.cmake
+  INSTALL_DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${CMAKE_PROJECT_NAME})

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cmakescripts/GNUInstallDirs.cmake Added

@@ -0,0 +1,419 @@
+#.rst:
+# GNUInstallDirs
+# --------------
+#
+# Define GNU standard installation directories
+#
+# Provides install directory variables as defined by the
+# `GNU Coding Standards`_.
+#
+# .. _`GNU Coding Standards`: https://www.gnu.org/prep/standards/html_node/Directory-Variables.html
+#
+# Result Variables
+# ^^^^^^^^^^^^^^^^
+#
+# Inclusion of this module defines the following variables:
+#
+# ``CMAKE_INSTALL_<dir>``
+#
+#   Destination for files of a given type.  This value may be passed to
+#   the ``DESTINATION`` options of :command:`install` commands for the
+#   corresponding file type.
+#
+# ``CMAKE_INSTALL_FULL_<dir>``
+#
+#   The absolute path generated from the corresponding ``CMAKE_INSTALL_<dir>``
+#   value.  If the value is not already an absolute path, an absolute path
+#   is constructed typically by prepending the value of the
+#   :variable:`CMAKE_INSTALL_PREFIX` variable.  However, there are some
+#   `special cases`_ as documented below.
+#
+# where ``<dir>`` is one of:
+#
+# ``BINDIR``
+#   user executables (``bin``)
+# ``SBINDIR``
+#   system admin executables (``sbin``)
+# ``LIBEXECDIR``
+#   program executables (``libexec``)
+# ``SYSCONFDIR``
+#   read-only single-machine data (``etc``)
+# ``SHAREDSTATEDIR``
+#   modifiable architecture-independent data (``com``)
+# ``LOCALSTATEDIR``
+#   modifiable single-machine data (``var``)
+# ``LIBDIR``
+#   object code libraries (``lib`` or ``lib64``
+#   or ``lib/<multiarch-tuple>`` on Debian)
+# ``INCLUDEDIR``
+#   C header files (``include``)
+# ``OLDINCLUDEDIR``
+#   C header files for non-gcc (``/usr/include``)
+# ``DATAROOTDIR``
+#   read-only architecture-independent data root (``share``)
+# ``DATADIR``
+#   read-only architecture-independent data (``DATAROOTDIR``)
+# ``INFODIR``
+#   info documentation (``DATAROOTDIR/info``)
+# ``LOCALEDIR``
+#   locale-dependent data (``DATAROOTDIR/locale``)
+# ``MANDIR``
+#   man documentation (``DATAROOTDIR/man``)
+# ``DOCDIR``
+#   documentation root (``DATAROOTDIR/doc/PROJECT_NAME``)
+#
+# If the includer does not define a value the above-shown default will be
+# used and the value will appear in the cache for editing by the user.
+#
+# Special Cases
+# ^^^^^^^^^^^^^
+#
+# The following values of :variable:`CMAKE_INSTALL_PREFIX` are special:
+#
+# ``/``
+#
+#   For ``<dir>`` other than the ``SYSCONFDIR`` and ``LOCALSTATEDIR``,
+#   the value of ``CMAKE_INSTALL_<dir>`` is prefixed with ``usr/`` if
+#   it is not user-specified as an absolute path.  For example, the
+#   ``INCLUDEDIR`` value ``include`` becomes ``usr/include``.
+#   This is required by the `GNU Coding Standards`_, which state:
+#
+#     When building the complete GNU system, the prefix will be empty
+#     and ``/usr`` will be a symbolic link to ``/``.
+#
+# ``/usr``
+#
+#   For ``<dir>`` equal to ``SYSCONFDIR`` or ``LOCALSTATEDIR``, the
+#   ``CMAKE_INSTALL_FULL_<dir>`` is computed by prepending just ``/``
+#   to the value of ``CMAKE_INSTALL_<dir>`` if it is not user-specified
+#   as an absolute path.  For example, the ``SYSCONFDIR`` value ``etc``
+#   becomes ``/etc``.  This is required by the `GNU Coding Standards`_.
+#
+# ``/opt/...``
+#
+#   For ``<dir>`` equal to ``SYSCONFDIR`` or ``LOCALSTATEDIR``, the
+#   ``CMAKE_INSTALL_FULL_<dir>`` is computed by *appending* the prefix
+#   to the value of ``CMAKE_INSTALL_<dir>`` if it is not user-specified
+#   as an absolute path.  For example, the ``SYSCONFDIR`` value ``etc``
+#   becomes ``/etc/opt/...``.  This is defined by the
+#   `Filesystem Hierarchy Standard`_.
+#
+# .. _`Filesystem Hierarchy Standard`: https://refspecs.linuxfoundation.org/FHS_3.0/fhs/index.html
+#
+# Macros
+# ^^^^^^
+#
+# .. command:: GNUInstallDirs_get_absolute_install_dir
+#
+#   ::
+#
+#     GNUInstallDirs_get_absolute_install_dir(absvar var)
+#
+#   Set the given variable ``absvar`` to the absolute path contained
+#   within the variable ``var``.  This is to allow the computation of an
+#   absolute path, accounting for all the special cases documented
+#   above.  While this macro is used to compute the various
+#   ``CMAKE_INSTALL_FULL_<dir>`` variables, it is exposed publicly to
+#   allow users who create additional path variables to also compute
+#   absolute paths where necessary, using the same logic.
+
+#=============================================================================
+# Copyright 2018 Matthias Räncker
+# Copyright 2016, 2019 D. R. Commander
+# Copyright 2016 Dmitry Marakasov
+# Copyright 2016 Roger Leigh
+# Copyright 2015 Alex Turbov
+# Copyright 2014 Rolf Eike Beer
+# Copyright 2014 Daniele E. Domenichelli
+# Copyright 2013 Dimitri John Ledkov
+# Copyright 2011 Alex Neundorf
+# Copyright 2011 Eric NOULARD
+# Copyright 2011, 2013-2015 Kitware, Inc.
+# Copyright 2011 Nikita Krupen'ko
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#
+# * Redistributions of source code must retain the above copyright
+#   notice, this list of conditions and the following disclaimer.
+#
+# * Redistributions in binary form must reproduce the above copyright
+#   notice, this list of conditions and the following disclaimer in the
+#   documentation and/or other materials provided with the distribution.
+#
+# * Neither the names of Kitware, Inc., the Insight Software Consortium,
+#   nor the names of their contributors may be used to endorse or promote
+#   products derived from this software without specific prior written
+#   permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#=============================================================================
+
+# Installation directories
+#
+
+macro(GNUInstallDirs_set_install_dir var docstring)
+  # If CMAKE_INSTALL_PREFIX changes and CMAKE_INSTALL_*DIR is still set to the
+  # default value, then modify it accordingly.  This presumes that the default
+  # value may change based on the prefix.
+
+  set(_GNUInstallDirs_CMAKE_INSTALL_FORCE_${var} "")
+  if(NOT DEFINED CMAKE_INSTALL_${var})
+    set(_GNUInstallDirs_CMAKE_INSTALL_DEFAULT_${var} 1 CACHE INTERNAL
+      "CMAKE_INSTALL_${var} has default value")
+  elseif(DEFINED _GNUInstallDirs_CMAKE_INSTALL_LAST_DEFAULT_${var} AND
+    NOT "${_GNUInstallDirs_CMAKE_INSTALL_LAST_DEFAULT_${var}}" STREQUAL
+      "${CMAKE_INSTALL_DEFAULT_${var}}" AND
+    _GNUInstallDirs_CMAKE_INSTALL_DEFAULT_${var} AND
+    "${_GNUInstallDirs_CMAKE_INSTALL_LAST_${var}}" STREQUAL
+      "${CMAKE_INSTALL_${var}}")
+    set(_GNUInstallDirs_CMAKE_INSTALL_FORCE_${var} "FORCE")
+  endif()
+
+  set(CMAKE_INSTALL_${var} "${CMAKE_INSTALL_DEFAULT_${var}}" CACHE PATH
+    "${docstring} (Default: ${CMAKE_INSTALL_DEFAULT_${var}})"
+    ${_GNUInstallDirs_CMAKE_INSTALL_FORCE_${var}})
+
+  if(NOT CMAKE_INSTALL_${var} STREQUAL CMAKE_INSTALL_DEFAULT_${var})
+    unset(_GNUInstallDirs_CMAKE_INSTALL_DEFAULT_${var} CACHE)
+  endif()
+
+  # Save for next run
+  set(_GNUInstallDirs_CMAKE_INSTALL_LAST_${var} "${CMAKE_INSTALL_${var}}"
+    CACHE INTERNAL "CMAKE_INSTALL_${var} during last run")
+  set(_GNUInstallDirs_CMAKE_INSTALL_LAST_DEFAULT_${var}
+    "${CMAKE_INSTALL_DEFAULT_${var}}" CACHE INTERNAL
+    "CMAKE_INSTALL_DEFAULT_${var} during last run")
+endmacro()
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_BINDIR)
+  set(CMAKE_INSTALL_DEFAULT_BINDIR "bin")
+endif()
+GNUInstallDirs_set_install_dir(BINDIR
+  "Directory into which user executables should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_SBINDIR)
+  set(CMAKE_INSTALL_DEFAULT_SBINDIR "sbin")
+endif()
+GNUInstallDirs_set_install_dir(SBINDIR
+  "Directory into which system admin executables should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_LIBEXECDIR)
+  set(CMAKE_INSTALL_DEFAULT_LIBEXECDIR "libexec")
+endif()
+GNUInstallDirs_set_install_dir(LIBEXECDIR
+  "Directory under which executables run by other programs should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_SYSCONFDIR)
+  set(CMAKE_INSTALL_DEFAULT_SYSCONFDIR "etc")
+endif()
+GNUInstallDirs_set_install_dir(SYSCONFDIR
+  "Directory into which machine-specific read-only ASCII data and configuration files should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_SHAREDSTATEDIR)
+  set(CMAKE_INSTALL_DEFAULT_SHAREDSTATEDIR "com")
+endif()
+GNUInstallDirs_set_install_dir(SHAREDSTATEDIR
+  "Directory into which architecture-independent run-time-modifiable data files should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_LOCALSTATEDIR)
+  set(CMAKE_INSTALL_DEFAULT_LOCALSTATEDIR "var")
+endif()
+GNUInstallDirs_set_install_dir(LOCALSTATEDIR
+  "Directory into which machine-specific run-time-modifiable data files should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_LIBDIR)
+  set(CMAKE_INSTALL_DEFAULT_LIBDIR "lib")
+  # Override this default 'lib' with 'lib64' iff:
+  #  - we are on Linux system but NOT cross-compiling
+  #  - we are NOT on debian
+  #  - we are on a 64 bits system
+  # reason is: amd64 ABI: http://www.x86-64.org/documentation/abi.pdf
+  # For Debian with multiarch, use 'lib/${CMAKE_LIBRARY_ARCHITECTURE}' if
+  # CMAKE_LIBRARY_ARCHITECTURE is set (which contains e.g. "i386-linux-gnu"
+  # and CMAKE_INSTALL_PREFIX is "/usr"
+  # See http://wiki.debian.org/Multiarch
+  if(CMAKE_SYSTEM_NAME MATCHES "^(Linux|kFreeBSD|GNU)$"
+      AND NOT CMAKE_CROSSCOMPILING)
+    if (EXISTS "/etc/debian_version") # is this a debian system ?
+      if(CMAKE_LIBRARY_ARCHITECTURE)
+        if("${CMAKE_INSTALL_PREFIX}" MATCHES "^/usr/?$")
+          set(CMAKE_INSTALL_DEFAULT_LIBDIR "lib/${CMAKE_LIBRARY_ARCHITECTURE}")
+        endif()
+      endif()
+    else() # not debian, rely on CMAKE_SIZEOF_VOID_P:
+      if(NOT DEFINED CMAKE_SIZEOF_VOID_P)
+        message(AUTHOR_WARNING
+          "Unable to determine default CMAKE_INSTALL_LIBDIR directory because no target architecture is known. "
+          "Please enable at least one language before including GNUInstallDirs.")
+      else()
+        if("${CMAKE_SIZEOF_VOID_P}" EQUAL "8")
+          set(CMAKE_INSTALL_DEFAULT_LIBDIR "lib64")
+        elseif(CMAKE_C_COMPILER_ABI MATCHES "ELF X32")
+          set(CMAKE_INSTALL_DEFAULT_LIBDIR "libx32")
+        endif()
+      endif()
+    endif()
+  endif()
+endif()
+GNUInstallDirs_set_install_dir(LIBDIR
+  "Directory into which object files and object code libraries should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_INCLUDEDIR)
+  set(CMAKE_INSTALL_DEFAULT_INCLUDEDIR "include")
+endif()
+GNUInstallDirs_set_install_dir(INCLUDEDIR
+  "Directory into which C header files should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_OLDINCLUDEDIR)
+  set(CMAKE_INSTALL_DEFAULT_OLDINCLUDEDIR "/usr/include")
+endif()
+GNUInstallDirs_set_install_dir(OLDINCLUDEDIR
+  PATH "Directory into which C header files for non-GCC compilers should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_DATAROOTDIR)
+  set(CMAKE_INSTALL_DEFAULT_DATAROOTDIR "share")
+endif()
+GNUInstallDirs_set_install_dir(DATAROOTDIR
+  "The root of the directory tree for read-only architecture-independent data files")
+
+#-----------------------------------------------------------------------------
+# Values whose defaults are relative to DATAROOTDIR.  Store empty values in
+# the cache and store the defaults in local variables if the cache values are
+# not set explicitly.  This auto-updates the defaults as DATAROOTDIR changes.
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_DATADIR)
+  set(CMAKE_INSTALL_DEFAULT_DATADIR "<CMAKE_INSTALL_DATAROOTDIR>")
+endif()
+GNUInstallDirs_set_install_dir(DATADIR
+  "The directory under which read-only architecture-independent data files should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_INFODIR)
+  if(CMAKE_SYSTEM_NAME MATCHES "^(.*BSD|DragonFly)$")
+    set(CMAKE_INSTALL_DEFAULT_INFODIR "info")
+  else()
+    set(CMAKE_INSTALL_DEFAULT_INFODIR "<CMAKE_INSTALL_DATAROOTDIR>/info")
+  endif()
+endif()
+GNUInstallDirs_set_install_dir(INFODIR
+  "The directory into which info documentation files should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_MANDIR)
+  if(CMAKE_SYSTEM_NAME MATCHES "^(.*BSD|DragonFly)$")
+    set(CMAKE_INSTALL_DEFAULT_MANDIR "man")
+  else()
+    set(CMAKE_INSTALL_DEFAULT_MANDIR "<CMAKE_INSTALL_DATAROOTDIR>/man")
+  endif()
+endif()
+GNUInstallDirs_set_install_dir(MANDIR
+  "The directory under which man pages should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_LOCALEDIR)
+  set(CMAKE_INSTALL_DEFAULT_LOCALEDIR "<CMAKE_INSTALL_DATAROOTDIR>/locale")
+endif()
+GNUInstallDirs_set_install_dir(LOCALEDIR
+  "The directory under which locale-specific message catalogs should be installed")
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_DOCDIR)
+  set(CMAKE_INSTALL_DEFAULT_DOCDIR "<CMAKE_INSTALL_DATAROOTDIR>/doc/${PROJECT_NAME}")
+endif()
+GNUInstallDirs_set_install_dir(DOCDIR
+  "The directory into which documentation files (other than info files) should be installed")
+
+#-----------------------------------------------------------------------------
+
+mark_as_advanced(
+  CMAKE_INSTALL_BINDIR
+  CMAKE_INSTALL_SBINDIR
+  CMAKE_INSTALL_LIBEXECDIR
+  CMAKE_INSTALL_SYSCONFDIR
+  CMAKE_INSTALL_SHAREDSTATEDIR
+  CMAKE_INSTALL_LOCALSTATEDIR
+  CMAKE_INSTALL_LIBDIR
+  CMAKE_INSTALL_INCLUDEDIR
+  CMAKE_INSTALL_OLDINCLUDEDIR
+  CMAKE_INSTALL_DATAROOTDIR
+  CMAKE_INSTALL_DATADIR
+  CMAKE_INSTALL_INFODIR
+  CMAKE_INSTALL_LOCALEDIR
+  CMAKE_INSTALL_MANDIR
+  CMAKE_INSTALL_DOCDIR
+  )
+
+macro(GNUInstallDirs_get_absolute_install_dir absvar var)
+  string(REGEX REPLACE "<>" "@" ${var} "${${var}}")
+  # Handle the specific case of an empty CMAKE_INSTALL_DATAROOTDIR
+  if(NOT CMAKE_INSTALL_DATAROOTDIR AND
+    ${var} MATCHES "\@CMAKE_INSTALL_DATAROOTDIR\@/")
+    string(CONFIGURE "${${var}}" ${var} @ONLY)
+    string(REGEX REPLACE "^/" "" ${var} "${${var}}")
+  else()
+    string(CONFIGURE "${${var}}" ${var} @ONLY)
+  endif()
+  if(NOT IS_ABSOLUTE "${${var}}")
+    # Handle special cases:
+    # - CMAKE_INSTALL_PREFIX == /
+    # - CMAKE_INSTALL_PREFIX == /usr
+    # - CMAKE_INSTALL_PREFIX == /opt/...
+    if("${CMAKE_INSTALL_PREFIX}" STREQUAL "/")
+      if("${dir}" STREQUAL "SYSCONFDIR" OR "${dir}" STREQUAL "LOCALSTATEDIR")
+        set(${absvar} "/${${var}}")
+      else()
+        if (NOT "${${var}}" MATCHES "^usr/")
+          set(${var} "usr/${${var}}")
+        endif()
+        set(${absvar} "/${${var}}")
+      endif()
+    elseif("${CMAKE_INSTALL_PREFIX}" MATCHES "^/usr/?$")
+      if("${dir}" STREQUAL "SYSCONFDIR" OR "${dir}" STREQUAL "LOCALSTATEDIR")
+        set(${absvar} "/${${var}}")
+      else()
+        set(${absvar} "${CMAKE_INSTALL_PREFIX}/${${var}}")
+      endif()
+    elseif("${CMAKE_INSTALL_PREFIX}" MATCHES "^/opt/.*")
+      if("${dir}" STREQUAL "SYSCONFDIR" OR "${dir}" STREQUAL "LOCALSTATEDIR")
+        set(${absvar} "/${${var}}${CMAKE_INSTALL_PREFIX}")
+      else()
+        set(${absvar} "${CMAKE_INSTALL_PREFIX}/${${var}}")
+      endif()
+    else()
+      set(${absvar} "${CMAKE_INSTALL_PREFIX}/${${var}}")
+    endif()
+  else()
+    set(${absvar} "${${var}}")
+  endif()
+  string(REGEX REPLACE "/$" "" ${absvar} "${${absvar}}")
+endmacro()
+
+# Result directories
+#
+foreach(dir
+    BINDIR
+    SBINDIR
+    LIBEXECDIR
+    SYSCONFDIR
+    SHAREDSTATEDIR
+    LOCALSTATEDIR
+    LIBDIR
+    INCLUDEDIR
+    OLDINCLUDEDIR
+    DATAROOTDIR
+    DATADIR
+    INFODIR
+    LOCALEDIR
+    MANDIR
+    DOCDIR
+    )
+  GNUInstallDirs_get_absolute_install_dir(CMAKE_INSTALL_FULL_${dir} CMAKE_INSTALL_${dir})
+endforeach()

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cmakescripts/cmake_uninstall.cmake.in Added

@@ -0,0 +1,24 @@
+# This code is from the CMake FAQ
+
+if (NOT EXISTS "@CMAKE_BINARY_DIR@/install_manifest.txt")
+  message(FATAL_ERROR "Cannot find install manifest: \"@CMAKE_BINARY_DIR@/install_manifest.txt\"")
+endif(NOT EXISTS "@CMAKE_BINARY_DIR@/install_manifest.txt")
+
+file(READ "@CMAKE_BINARY_DIR@/install_manifest.txt" files)
+string(REGEX REPLACE "\n" ";" files "${files}")
+list(REVERSE files)
+foreach (file ${files})
+  message(STATUS "Uninstalling \"$ENV{DESTDIR}${file}\"")
+    if (EXISTS "$ENV{DESTDIR}${file}")
+      execute_process(
+        COMMAND "@CMAKE_COMMAND@" -E remove "$ENV{DESTDIR}${file}"
+        OUTPUT_VARIABLE rm_out
+        RESULT_VARIABLE rm_retval
+      )
+    if(NOT ${rm_retval} EQUAL 0)
+      message(FATAL_ERROR "Problem when removing \"$ENV{DESTDIR}${file}\"")
+    endif (NOT ${rm_retval} EQUAL 0)
+  else (EXISTS "$ENV{DESTDIR}${file}")
+    message(STATUS "File \"$ENV{DESTDIR}${file}\" does not exist.")
+  endif (EXISTS "$ENV{DESTDIR}${file}")
+endforeach(file)

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cmakescripts/testclean.cmake Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/cmyk.h Added

@@ -0,0 +1,61 @@
+/*
+ * cmyk.h
+ *
+ * Copyright (C) 2017-2018, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains convenience functions for performing quick & dirty
+ * CMYK<->RGB conversion.  This algorithm is suitable for testing purposes
+ * only.  Properly converting between CMYK and RGB requires a color management
+ * system.
+ */
+
+#ifndef CMYK_H
+#define CMYK_H
+
+#include <jinclude.h>
+#define JPEG_INTERNALS
+#include <jpeglib.h>
+#include "jconfigint.h"
+
+
+/* Fully reversible */
+
+INLINE
+LOCAL(void)
+rgb_to_cmyk(JSAMPLE r, JSAMPLE g, JSAMPLE b, JSAMPLE *c, JSAMPLE *m,
+            JSAMPLE *y, JSAMPLE *k)
+{
+  double ctmp = 1.0 - ((double)r / 255.0);
+  double mtmp = 1.0 - ((double)g / 255.0);
+  double ytmp = 1.0 - ((double)b / 255.0);
+  double ktmp = MIN(MIN(ctmp, mtmp), ytmp);
+
+  if (ktmp == 1.0) ctmp = mtmp = ytmp = 0.0;
+  else {
+    ctmp = (ctmp - ktmp) / (1.0 - ktmp);
+    mtmp = (mtmp - ktmp) / (1.0 - ktmp);
+    ytmp = (ytmp - ktmp) / (1.0 - ktmp);
+  }
+  *c = (JSAMPLE)(255.0 - ctmp * 255.0 + 0.5);
+  *m = (JSAMPLE)(255.0 - mtmp * 255.0 + 0.5);
+  *y = (JSAMPLE)(255.0 - ytmp * 255.0 + 0.5);
+  *k = (JSAMPLE)(255.0 - ktmp * 255.0 + 0.5);
+}
+
+
+/* Fully reversible only for C/M/Y/K values generated with rgb_to_cmyk() */
+
+INLINE
+LOCAL(void)
+cmyk_to_rgb(JSAMPLE c, JSAMPLE m, JSAMPLE y, JSAMPLE k, JSAMPLE *r, JSAMPLE *g,
+            JSAMPLE *b)
+{
+  *r = (JSAMPLE)((double)c * (double)k / 255.0 + 0.5);
+  *g = (JSAMPLE)((double)m * (double)k / 255.0 + 0.5);
+  *b = (JSAMPLE)((double)y * (double)k / 255.0 + 0.5);
+}
+
+
+#endif /* CMYK_H */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/coderules.txt Added

@@ -0,0 +1,78 @@
+IJG JPEG LIBRARY:  CODING RULES
+
+This file was part of the Independent JPEG Group's software:
+Copyright (C) 1991-1996, Thomas G. Lane.
+It was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo.
+For conditions of distribution and use, see the accompanying README.ijg file.
+
+
+Since numerous people will be contributing code and bug fixes, it's important
+to establish a common coding style.  The goal of using similar coding styles
+is much more important than the details of just what that style is.
+
+In general we follow the recommendations of "Recommended C Style and Coding
+Standards" revision 6.1 (Cannon et al. as modified by Spencer, Keppel and
+Brader).  This document is available in the IJG FTP archive (see
+jpeg/doc/cstyle.ms.tbl.Z, or cstyle.txt.Z for those without nroff/tbl).
+
+Block comments should be laid out thusly:
+
+/*
+ *  Block comments in this style.
+ */
+
+We indent statements in K&R style, e.g.,
+        if (test) {
+          then-part;
+        } else {
+          else-part;
+        }
+with two spaces per indentation level.  (This indentation convention is
+handled automatically by GNU Emacs and many other text editors.)
+
+Multi-word names should be written in lower case with underscores, e.g.,
+multi_word_name (not multiWordName).  Preprocessor symbols and enum constants
+are similar but upper case (MULTI_WORD_NAME).  Names should be unique within
+the first fifteen characters.
+
+Note that each function definition must begin with GLOBAL(type), LOCAL(type),
+or METHODDEF(type).  These macros expand to "static type" or just "type" as
+appropriate.  They provide a readable indication of the routine's usage and
+can readily be changed for special needs.  (For instance, special linkage
+keywords can be inserted for use in Windows DLLs.)
+
+A similar solution is used for external function declarations (see the EXTERN
+macro.)
+
+
+The JPEG library is intended to be used within larger programs.  Furthermore,
+we want it to be reentrant so that it can be used by applications that process
+multiple images concurrently.  The following rules support these requirements:
+
+1. Avoid direct use of file I/O, "malloc", error report printouts, etc;
+pass these through the common routines provided.
+
+2. Minimize global namespace pollution.  Functions should be declared static
+wherever possible.  (Note that our method-based calling conventions help this
+a lot: in many modules only the initialization function will ever need to be
+called directly, so only that function need be externally visible.)  All
+global function names should begin with "jpeg_".
+
+3. Don't use global variables; anything that must be used in another module
+should be in the common data structures.
+
+4. Don't use static variables except for read-only constant tables.  Variables
+that should be private to a module can be placed into private structures (see
+the system architecture document, structure.txt).
+
+5. Source file names should begin with "j" for files that are part of the
+library proper; source files that are not part of the library, such as cjpeg.c
+and djpeg.c, do not begin with "j".  Keep compression and decompression code in
+separate source files --- some applications may want only one half of the
+library.
+
+Note: these rules (particularly #4) are not followed religiously in the
+modules that are used in cjpeg/djpeg but are not part of the JPEG library
+proper.  Those modules are not really intended to be used in other
+applications.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/croptest.in Added

@@ -0,0 +1,95 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if  -d $OUTDIR ; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	$*
+}
+
+IMAGE=vgl_6548_0026a.bmp
+WIDTH=128
+HEIGHT=95
+IMGDIR=@CMAKE_CURRENT_SOURCE_DIR@/testimages
+OUTDIR=`mktemp -d /tmp/__croptest_output.XXXXXX`
+EXEDIR=@CMAKE_CURRENT_BINARY_DIR@
+
+if  -d $OUTDIR ; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+exec >$EXEDIR/croptest.log
+
+echo "============================================================"
+echo "$IMAGE ($WIDTH x $HEIGHT)"
+echo "============================================================"
+echo
+
+for PROGARG in "" -progressive; do
+
+	cp $IMGDIR/$IMAGE $OUTDIR
+	basename=`basename $IMAGE .bmp`
+	echo "------------------------------------------------------------"
+	echo "Generating test images"
+	echo "------------------------------------------------------------"
+	echo
+	runme $EXEDIR/cjpeg $PROGARG -grayscale -outfile $OUTDIR/${basename}_GRAY.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg $PROGARG -sample 2x2 -outfile $OUTDIR/${basename}_420.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg $PROGARG -sample 2x1 -outfile $OUTDIR/${basename}_422.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg $PROGARG -sample 1x2 -outfile $OUTDIR/${basename}_440.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg $PROGARG -sample 1x1 -outfile $OUTDIR/${basename}_444.jpg $IMGDIR/${basename}.bmp
+	echo
+
+	for NSARG in "" -nosmooth; do
+
+		for COLORSARG in "" "-colors 256 -dither none -onepass"; do
+
+			for Y in {0..16}; do
+
+				for H in {1..16}; do
+
+					X=$(( (Y*16)%128 ))
+					W=$(( WIDTH-X-7 ))
+					if  $Y -le 15 ; then
+						CROPSPEC="${W}x${H}+${X}+${Y}"
+					else
+						Y2=$(( HEIGHT-H ));
+						CROPSPEC="${W}x${H}+${X}+${Y2}"
+					fi
+
+					echo "------------------------------------------------------------"
+					echo $PROGARG $NSARG $COLORSARG -crop $CROPSPEC
+					echo "------------------------------------------------------------"
+					echo
+					for samp in GRAY 420 422 440 444; do
+						$EXEDIR/djpeg $NSARG $COLORSARG -rgb -outfile $OUTDIR/${basename}_${samp}_full.ppm $OUTDIR/${basename}_${samp}.jpg
+						convert -crop $CROPSPEC $OUTDIR/${basename}_${samp}_full.ppm $OUTDIR/${basename}_${samp}_ref.ppm
+						runme $EXEDIR/djpeg $NSARG $COLORSARG -crop $CROPSPEC -rgb -outfile $OUTDIR/${basename}_${samp}.ppm $OUTDIR/${basename}_${samp}.jpg
+						runme cmp $OUTDIR/${basename}_${samp}.ppm $OUTDIR/${basename}_${samp}_ref.ppm
+					done
+					echo
+
+				done
+
+			done
+
+		done
+
+	done
+
+done
+
+echo SUCCESS!

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/djpeg.1 Added

@@ -0,0 +1,320 @@
+.TH DJPEG 1 "4 November 2020"
+.SH NAME
+djpeg \- decompress a JPEG file to an image file
+.SH SYNOPSIS
+.B djpeg
+
+.I options
+
+
+.I filename
+
+.LP
+.SH DESCRIPTION
+.LP
+.B djpeg
+decompresses the named JPEG file, or the standard input if no file is named,
+and produces an image file on the standard output.  PBMPLUS (PPM/PGM), BMP,
+GIF, or Targa output format can be selected.
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-grayscale
+may be written
+.B \-gray
+or
+.BR \-gr .
+Most of the "basic" switches can be abbreviated to as little as one letter.
+Upper and lower case are equivalent (thus
+.B \-BMP
+is the same as
+.BR \-bmp ).
+British spellings are also accepted (e.g.,
+.BR \-greyscale ),
+though for brevity these are not mentioned below.
+.PP
+The basic switches are:
+.TP
+.BI \-colors " N"
+Reduce image to at most N colors.  This reduces the number of colors used in
+the output image, so that it can be displayed on a colormapped display or
+stored in a colormapped file format.  For example, if you have an 8-bit
+display, you'd need to reduce to 256 or fewer colors.
+.TP
+.BI \-quantize " N"
+Same as
+.BR \-colors .
+.B \-colors
+is the recommended name,
+.B \-quantize
+is provided only for backwards compatibility.
+.TP
+.B \-fast
+Select recommended processing options for fast, low quality output.  (The
+default options are chosen for highest quality output.)  Currently, this is
+equivalent to \fB\-dct fast \-nosmooth \-onepass \-dither ordered\fR.
+.TP
+.B \-grayscale
+Force grayscale output even if JPEG file is color.  Useful for viewing on
+monochrome displays; also,
+.B djpeg
+runs noticeably faster in this mode.
+.TP
+.B \-rgb
+Force RGB output even if JPEG file is grayscale.
+.TP
+.BI \-scale " M/N"
+Scale the output image by a factor M/N.  Currently the scale factor must be
+M/8, where M is an integer between 1 and 16 inclusive, or any reduced fraction
+thereof (such as 1/2, 3/4, etc.)  Scaling is handy if the image is larger than
+your screen; also,
+.B djpeg
+runs much faster when scaling down the output.
+.TP
+.B \-bmp
+Select BMP output format (Windows flavor).  8-bit colormapped format is
+emitted if
+.B \-colors
+or
+.B \-grayscale
+is specified, or if the JPEG file is grayscale; otherwise, 24-bit full-color
+format is emitted.
+.TP
+.B \-gif
+Select GIF output format (LZW-compressed).  Since GIF does not support more
+than 256 colors,
+.B \-colors 256
+is assumed (unless you specify a smaller number of colors).  If you specify
+.BR \-fast,
+the default number of colors is 216.
+.TP
+.B \-gif0
+Select GIF output format (uncompressed).  Since GIF does not support more than
+256 colors,
+.B \-colors 256
+is assumed (unless you specify a smaller number of colors).  If you specify
+.BR \-fast,
+the default number of colors is 216.
+.TP
+.B \-os2
+Select BMP output format (OS/2 1.x flavor).  8-bit colormapped format is
+emitted if
+.B \-colors
+or
+.B \-grayscale
+is specified, or if the JPEG file is grayscale; otherwise, 24-bit full-color
+format is emitted.
+.TP
+.B \-pnm
+Select PBMPLUS (PPM/PGM) output format (this is the default format).
+PGM is emitted if the JPEG file is grayscale or if
+.B \-grayscale
+is specified; otherwise PPM is emitted.
+.TP
+.B \-targa
+Select Targa output format.  Grayscale format is emitted if the JPEG file is
+grayscale or if
+.B \-grayscale
+is specified; otherwise, colormapped format is emitted if
+.B \-colors
+is specified; otherwise, 24-bit full-color format is emitted.
+.PP
+Switches for advanced users:
+.TP
+.B \-dct int
+Use accurate integer DCT method (default).
+.TP
+.B \-dct fast
+Use less accurate integer DCT method legacy feature.
+When the Independent JPEG Group's software was first released in 1991, the
+decompression time for a 1-megapixel JPEG image on a mainstream PC was measured
+in minutes.  Thus, the \fBfast\fR integer DCT algorithm provided noticeable
+performance benefits.  On modern CPUs running libjpeg-turbo, however, the
+decompression time for a 1-megapixel JPEG image is measured in milliseconds,
+and thus the performance benefits of the \fBfast\fR algorithm are much less
+noticeable.  On modern x86/x86-64 CPUs that support AVX2 instructions, the
+\fBfast\fR and \fBint\fR methods have similar performance.  On other types of
+CPUs, the \fBfast\fR method is generally about 5-15% faster than the \fBint\fR
+method.
+
+If the JPEG image was compressed using a quality level of 85 or below, then
+there should be little or no perceptible quality difference between the two
+algorithms.  When decompressing images that were compressed using quality
+levels above 85, however, the difference between the \fBfast\fR and \fBint\fR
+methods becomes more pronounced.  With images compressed using quality=97, for
+instance, the \fBfast\fR method incurs generally about a 4-6 dB loss in PSNR
+relative to the \fBint\fR method, but this can be larger for some images.  If
+you can avoid it, do not use the \fBfast\fR method when decompressing images
+that were compressed using quality levels above 97.  The algorithm often
+degenerates for such images and can actually produce a more lossy output image
+than if the JPEG image had been compressed using lower quality levels.
+.TP
+.B \-dct float
+Use floating-point DCT method legacy feature.
+The \fBfloat\fR method does not produce significantly more accurate results
+than the \fBint\fR method, and it is much slower.  The \fBfloat\fR method may
+also give different results on different machines due to varying roundoff
+behavior, whereas the integer methods should give the same results on all
+machines.
+.TP
+.B \-dither fs
+Use Floyd-Steinberg dithering in color quantization.
+.TP
+.B \-dither ordered
+Use ordered dithering in color quantization.
+.TP
+.B \-dither none
+Do not use dithering in color quantization.
+By default, Floyd-Steinberg dithering is applied when quantizing colors; this
+is slow but usually produces the best results.  Ordered dither is a compromise
+between speed and quality; no dithering is fast but usually looks awful.  Note
+that these switches have no effect unless color quantization is being done.
+Ordered dither is only available in
+.B \-onepass
+mode.
+.TP
+.BI \-icc " file"
+Extract ICC color management profile to the specified file.
+.TP
+.BI \-map " file"
+Quantize to the colors used in the specified image file.  This is useful for
+producing multiple files with identical color maps, or for forcing a
+predefined set of colors to be used.  The
+.I file
+must be a GIF or PPM file. This option overrides
+.B \-colors
+and
+.BR \-onepass .
+.TP
+.B \-nosmooth
+Use a faster, lower-quality upsampling routine.
+.TP
+.B \-onepass
+Use one-pass instead of two-pass color quantization.  The one-pass method is
+faster and needs less memory, but it produces a lower-quality image.
+.B \-onepass
+is ignored unless you also say
+.B \-colors
+.IR N .
+Also, the one-pass method is always used for grayscale output (the two-pass
+method is no improvement then).
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, an error will occur.
+.TP
+.BI \-maxscans " N"
+Abort if the JPEG image contains more than
+.I N
+scans.  This feature demonstrates a method by which applications can guard
+against denial-of-service attacks instigated by specially-crafted malformed
+JPEG images containing numerous scans with missing image data or image data
+consisting only of "EOB runs" (a feature of progressive JPEG images that allows
+potentially hundreds of thousands of adjoining zero-value pixels to be
+represented using only a few bytes.)  Attempting to decompress such malformed
+JPEG images can cause excessive CPU activity, since the decompressor must fully
+process each scan (even if the scan is corrupt) before it can proceed to the
+next scan.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.BI \-memsrc
+Load input file into memory before decompressing.  This feature was implemented
+mainly as a way of testing the in-memory source manager (jpeg_mem_src().)
+.TP
+.BI \-report
+Report decompression progress.
+.TP
+.BI \-skip " Y0,Y1"
+Decompress all rows of the JPEG image except those between Y0 and Y1
+(inclusive.)  Note that if decompression scaling is being used, then Y0 and Y1
+are relative to the scaled image dimensions.
+.TP
+.BI \-crop " WxH+X+Y"
+Decompress only a rectangular subregion of the image, starting at point X,Y
+with width W and height H.  If necessary, X will be shifted left to the nearest
+iMCU boundary, and the width will be increased accordingly.  Note that if
+decompression scaling is being used, then X, Y, W, and H are relative to the
+scaled image dimensions.  Currently this option only works with the
+PBMPLUS (PPM/PGM), GIF, and Targa output formats.
+.TP
+.BI \-strict
+Treat all warnings as fatal.  This feature also demonstrates a method by which
+applications can guard against attacks instigated by specially-crafted
+malformed JPEG images.  Enabling this option will cause the decompressor to
+abort if the JPEG image contains incomplete or corrupt image data.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.TP
+.B \-version
+Print version information and exit.
+.SH EXAMPLES
+.LP
+This example decompresses the JPEG file foo.jpg, quantizes it to
+256 colors, and saves the output in 8-bit BMP format in foo.bmp:
+.IP
+.B djpeg \-colors 256 \-bmp
+.I foo.jpg
+.B >
+.I foo.bmp
+.SH HINTS
+To get a quick preview of an image, use the
+.B \-grayscale
+and/or
+.B \-scale
+switches.
+.B \-grayscale \-scale 1/8
+is the fastest case.
+.PP
+Several options are available that trade off image quality to gain speed.
+.B \-fast
+turns on the recommended settings.
+.PP
+.B \-dct fast
+and/or
+.B \-nosmooth
+gain speed at a small sacrifice in quality.
+When producing a color-quantized image,
+.B \-onepass \-dither ordered
+is fast but much lower quality than the default behavior.
+.B \-dither none
+may give acceptable results in two-pass mode, but is seldom tolerable in
+one-pass mode.
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+.BR ppm (5),
+.BR pgm (5)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.PP
+This file was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo, to wordsmith certain sections, and to describe
+features not present in libjpeg.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/djpeg.c Added

@@ -0,0 +1,868 @@
+/*
+ * djpeg.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2013-2019 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010-2011, 2013-2017, 2019-2020, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a command-line user interface for the JPEG decompressor.
+ * It should work on any system with Unix- or MS-DOS-style command lines.
+ *
+ * Two different command line styles are permitted, depending on the
+ * compile-time switch TWO_FILE_COMMANDLINE:
+ *      djpeg options  inputfile outputfile
+ *      djpeg options  inputfile
+ * In the second style, output is always to standard output, which you'd
+ * normally redirect to a file or pipe to some other program.  Input is
+ * either from a named file or from standard input (typically redirected).
+ * The second style is convenient on Unix but is unhelpful on systems that
+ * don't support pipes.  Also, you MUST use the first style if your system
+ * doesn't do binary I/O to stdin/stdout.
+ * To simplify script writing, the "-outfile" switch is provided.  The syntax
+ *      djpeg options  -outfile outputfile  inputfile
+ * works regardless of which command line style is used.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "jversion.h"           /* for version message */
+#include "jconfigint.h"
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare free() */
+extern void free(void *ptr);
+#endif
+
+#include <ctype.h>              /* to declare isprint() */
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+
+/* Create the add-on message string table. */
+
+#define JMESSAGE(code, string)  string,
+
+static const char * const cdjpeg_message_table = {
+#include "cderror.h"
+  NULL
+};
+
+
+/*
+ * This list defines the known output image formats
+ * (not all of which need be supported by a given version).
+ * You can change the default output format by defining DEFAULT_FMT;
+ * indeed, you had better do so if you undefine PPM_SUPPORTED.
+ */
+
+typedef enum {
+  FMT_BMP,                      /* BMP format (Windows flavor) */
+  FMT_GIF,                      /* GIF format (LZW-compressed) */
+  FMT_GIF0,                     /* GIF format (uncompressed) */
+  FMT_OS2,                      /* BMP format (OS/2 flavor) */
+  FMT_PPM,                      /* PPM/PGM (PBMPLUS formats) */
+  FMT_TARGA,                    /* Targa format */
+  FMT_TIFF                      /* TIFF format */
+} IMAGE_FORMATS;
+
+#ifndef DEFAULT_FMT             /* so can override from CFLAGS in Makefile */
+#define DEFAULT_FMT     FMT_PPM
+#endif
+
+static IMAGE_FORMATS requested_fmt;
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char *progname;    /* program name for error messages */
+static char *icc_filename;      /* for -icc switch */
+JDIMENSION max_scans;           /* for -maxscans switch */
+static char *outfilename;       /* for -outfile switch */
+boolean memsrc;                 /* for -memsrc switch */
+boolean report;                 /* for -report switch */
+boolean skip, crop;
+JDIMENSION skip_start, skip_end;
+JDIMENSION crop_x, crop_y, crop_width, crop_height;
+boolean strict;                 /* for -strict switch */
+#define INPUT_BUF_SIZE  4096
+
+
+LOCAL(void)
+usage(void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s switches ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "inputfile\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -colors N      Reduce image to no more than N colors\n");
+  fprintf(stderr, "  -fast          Fast, low-quality processing\n");
+  fprintf(stderr, "  -grayscale     Force grayscale output\n");
+  fprintf(stderr, "  -rgb           Force RGB output\n");
+  fprintf(stderr, "  -rgb565        Force RGB565 output\n");
+#ifdef IDCT_SCALING_SUPPORTED
+  fprintf(stderr, "  -scale M/N     Scale output image by fraction M/N, eg, 1/8\n");
+#endif
+#ifdef BMP_SUPPORTED
+  fprintf(stderr, "  -bmp           Select BMP output format (Windows style)%s\n",
+          (DEFAULT_FMT == FMT_BMP ? " (default)" : ""));
+#endif
+#ifdef GIF_SUPPORTED
+  fprintf(stderr, "  -gif           Select GIF output format (LZW-compressed)%s\n",
+          (DEFAULT_FMT == FMT_GIF ? " (default)" : ""));
+  fprintf(stderr, "  -gif0          Select GIF output format (uncompressed)%s\n",
+          (DEFAULT_FMT == FMT_GIF0 ? " (default)" : ""));
+#endif
+#ifdef BMP_SUPPORTED
+  fprintf(stderr, "  -os2           Select BMP output format (OS/2 style)%s\n",
+          (DEFAULT_FMT == FMT_OS2 ? " (default)" : ""));
+#endif
+#ifdef PPM_SUPPORTED
+  fprintf(stderr, "  -pnm           Select PBMPLUS (PPM/PGM) output format%s\n",
+          (DEFAULT_FMT == FMT_PPM ? " (default)" : ""));
+#endif
+#ifdef TARGA_SUPPORTED
+  fprintf(stderr, "  -targa         Select Targa output format%s\n",
+          (DEFAULT_FMT == FMT_TARGA ? " (default)" : ""));
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef DCT_ISLOW_SUPPORTED
+  fprintf(stderr, "  -dct int       Use accurate integer DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_ISLOW ? " (default)" : ""));
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  fprintf(stderr, "  -dct fast      Use less accurate integer DCT method legacy feature%s\n",
+          (JDCT_DEFAULT == JDCT_IFAST ? " (default)" : ""));
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  fprintf(stderr, "  -dct float     Use floating-point DCT method legacy feature%s\n",
+          (JDCT_DEFAULT == JDCT_FLOAT ? " (default)" : ""));
+#endif
+  fprintf(stderr, "  -dither fs     Use F-S dithering (default)\n");
+  fprintf(stderr, "  -dither none   Don't use dithering in quantization\n");
+  fprintf(stderr, "  -dither ordered  Use ordered dither (medium speed, quality)\n");
+  fprintf(stderr, "  -icc FILE      Extract ICC profile to FILE\n");
+#ifdef QUANT_2PASS_SUPPORTED
+  fprintf(stderr, "  -map FILE      Map to colors used in named image file\n");
+#endif
+  fprintf(stderr, "  -nosmooth      Don't use high-quality upsampling\n");
+#ifdef QUANT_1PASS_SUPPORTED
+  fprintf(stderr, "  -onepass       Use 1-pass quantization (fast, low quality)\n");
+#endif
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -maxscans N    Maximum number of scans to allow in input file\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  fprintf(stderr, "  -memsrc        Load input file into memory before decompressing\n");
+#endif
+  fprintf(stderr, "  -report        Report decompression progress\n");
+  fprintf(stderr, "  -skip Y0,Y1    Decompress all rows except those between Y0 and Y1 (inclusive)\n");
+  fprintf(stderr, "  -crop WxH+X+Y  Decompress only a rectangular subregion of the image\n");
+  fprintf(stderr, "                 requires PBMPLUS (PPM/PGM), GIF, or Targa output format\n");
+  fprintf(stderr, "  -strict        Treat all warnings as fatal\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "  -version       Print version information and exit\n");
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(int)
+parse_switches(j_decompress_ptr cinfo, int argc, char **argv,
+               int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char *arg;
+
+  /* Set up default JPEG parameters. */
+  requested_fmt = DEFAULT_FMT;  /* set default output file format */
+  icc_filename = NULL;
+  max_scans = 0;
+  outfilename = NULL;
+  memsrc = FALSE;
+  report = FALSE;
+  skip = FALSE;
+  crop = FALSE;
+  strict = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argvargn;
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+        outfilename = NULL;     /* -outfile applies to just one input file */
+        continue;               /* ignore this name if previously processed */
+      }
+      break;                    /* else done parsing switches */
+    }
+    arg++;                      /* advance past switch marker character */
+
+    if (keymatch(arg, "bmp", 1)) {
+      /* BMP output format (Windows flavor). */
+      requested_fmt = FMT_BMP;
+
+    } else if (keymatch(arg, "colors", 1) || keymatch(arg, "colours", 1) ||
+               keymatch(arg, "quantize", 1) || keymatch(arg, "quantise", 1)) {
+      /* Do color quantization. */
+      int val;
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%d", &val) != 1)
+        usage();
+      cinfo->desired_number_of_colors = val;
+      cinfo->quantize_colors = TRUE;
+
+    } else if (keymatch(arg, "dct", 2)) {
+      /* Select IDCT algorithm. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argvargn, "int", 1)) {
+        cinfo->dct_method = JDCT_ISLOW;
+      } else if (keymatch(argvargn, "fast", 2)) {
+        cinfo->dct_method = JDCT_IFAST;
+      } else if (keymatch(argvargn, "float", 2)) {
+        cinfo->dct_method = JDCT_FLOAT;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "dither", 2)) {
+      /* Select dithering algorithm. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argvargn, "fs", 2)) {
+        cinfo->dither_mode = JDITHER_FS;
+      } else if (keymatch(argvargn, "none", 2)) {
+        cinfo->dither_mode = JDITHER_NONE;
+      } else if (keymatch(argvargn, "ordered", 2)) {
+        cinfo->dither_mode = JDITHER_ORDERED;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (!printed_version) {
+        fprintf(stderr, "%s version %s (build %s)\n",
+                PACKAGE_NAME, VERSION, BUILD);
+        fprintf(stderr, "%s\n\n", JCOPYRIGHT);
+        fprintf(stderr, "Emulating The Independent JPEG Group's software, version %s\n\n",
+                JVERSION);
+        printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "version", 4)) {
+      fprintf(stderr, "%s version %s (build %s)\n",
+              PACKAGE_NAME, VERSION, BUILD);
+      exit(EXIT_SUCCESS);
+
+    } else if (keymatch(arg, "fast", 1)) {
+      /* Select recommended processing options for quick-and-dirty output. */
+      cinfo->two_pass_quantize = FALSE;
+      cinfo->dither_mode = JDITHER_ORDERED;
+      if (!cinfo->quantize_colors) /* don't override an earlier -colors */
+        cinfo->desired_number_of_colors = 216;
+      cinfo->dct_method = JDCT_FASTEST;
+      cinfo->do_fancy_upsampling = FALSE;
+
+    } else if (keymatch(arg, "gif", 1)) {
+      /* GIF output format (LZW-compressed). */
+      requested_fmt = FMT_GIF;
+
+    } else if (keymatch(arg, "gif0", 4)) {
+      /* GIF output format (uncompressed). */
+      requested_fmt = FMT_GIF0;
+
+    } else if (keymatch(arg, "grayscale", 2) ||
+               keymatch(arg, "greyscale", 2)) {
+      /* Force monochrome output. */
+      cinfo->out_color_space = JCS_GRAYSCALE;
+
+    } else if (keymatch(arg, "rgb", 2)) {
+      /* Force RGB output. */
+      cinfo->out_color_space = JCS_RGB;
+
+    } else if (keymatch(arg, "rgb565", 2)) {
+      /* Force RGB565 output. */
+      cinfo->out_color_space = JCS_RGB565;
+
+    } else if (keymatch(arg, "icc", 1)) {
+      /* Set ICC filename. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      icc_filename = argvargn;
+      jpeg_save_markers(cinfo, JPEG_APP0 + 2, 0xFFFF);
+
+    } else if (keymatch(arg, "map", 3)) {
+      /* Quantize to a color map taken from an input file. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (for_real) {           /* too expensive to do twice! */
+#ifdef QUANT_2PASS_SUPPORTED    /* otherwise can't quantize to supplied map */
+        FILE *mapfile;
+
+        if ((mapfile = fopen(argvargn, READ_BINARY)) == NULL) {
+          fprintf(stderr, "%s: can't open %s\n", progname, argvargn);
+          exit(EXIT_FAILURE);
+        }
+        read_color_map(cinfo, mapfile);
+        fclose(mapfile);
+        cinfo->quantize_colors = TRUE;
+#else
+        ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+      }
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (ch == 'm' || ch == 'M')
+        lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "maxscans", 4)) {
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%u", &max_scans) != 1)
+        usage();
+
+    } else if (keymatch(arg, "nosmooth", 3)) {
+      /* Suppress fancy upsampling */
+      cinfo->do_fancy_upsampling = FALSE;
+
+    } else if (keymatch(arg, "onepass", 3)) {
+      /* Use fast one-pass quantization. */
+      cinfo->two_pass_quantize = FALSE;
+
+    } else if (keymatch(arg, "os2", 3)) {
+      /* BMP output format (OS/2 flavor). */
+      requested_fmt = FMT_OS2;
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      outfilename = argvargn; /* save it away for later use */
+
+    } else if (keymatch(arg, "memsrc", 2)) {
+      /* Use in-memory source manager */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+      memsrc = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, in-memory source manager was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "pnm", 1) || keymatch(arg, "ppm", 1)) {
+      /* PPM/PGM output format. */
+      requested_fmt = FMT_PPM;
+
+    } else if (keymatch(arg, "report", 2)) {
+      report = TRUE;
+
+    } else if (keymatch(arg, "scale", 2)) {
+      /* Scale the output image by a fraction M/N. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%u/%u",
+                 &cinfo->scale_num, &cinfo->scale_denom) != 2)
+        usage();
+
+    } else if (keymatch(arg, "skip", 2)) {
+      if (++argn >= argc)
+        usage();
+      if (sscanf(argvargn, "%u,%u", &skip_start, &skip_end) != 2 ||
+          skip_start > skip_end)
+        usage();
+      skip = TRUE;
+
+    } else if (keymatch(arg, "crop", 2)) {
+      char c;
+      if (++argn >= argc)
+        usage();
+      if (sscanf(argvargn, "%u%c%u+%u+%u", &crop_width, &c, &crop_height,
+                 &crop_x, &crop_y) != 5 ||
+          (c != 'X' && c != 'x') || crop_width < 1 || crop_height < 1)
+        usage();
+      crop = TRUE;
+
+    } else if (keymatch(arg, "strict", 2)) {
+      strict = TRUE;
+
+    } else if (keymatch(arg, "targa", 1)) {
+      /* Targa output format. */
+      requested_fmt = FMT_TARGA;
+
+    } else {
+      usage();                  /* bogus switch */
+    }
+  }
+
+  return argn;                  /* return index of next arg (file name) */
+}
+
+
+/*
+ * Marker processor for COM and interesting APPn markers.
+ * This replaces the library's built-in processor, which just skips the marker.
+ * We want to print out the marker as text, to the extent possible.
+ * Note this code relies on a non-suspending data source.
+ */
+
+LOCAL(unsigned int)
+jpeg_getc(j_decompress_ptr cinfo)
+/* Read next byte */
+{
+  struct jpeg_source_mgr *datasrc = cinfo->src;
+
+  if (datasrc->bytes_in_buffer == 0) {
+    if (!(*datasrc->fill_input_buffer) (cinfo))
+      ERREXIT(cinfo, JERR_CANT_SUSPEND);
+  }
+  datasrc->bytes_in_buffer--;
+  return *datasrc->next_input_byte++;
+}
+
+
+METHODDEF(boolean)
+print_text_marker(j_decompress_ptr cinfo)
+{
+  boolean traceit = (cinfo->err->trace_level >= 1);
+  long length;
+  unsigned int ch;
+  unsigned int lastch = 0;
+
+  length = jpeg_getc(cinfo) << 8;
+  length += jpeg_getc(cinfo);
+  length -= 2;                  /* discount the length word itself */
+
+  if (traceit) {
+    if (cinfo->unread_marker == JPEG_COM)
+      fprintf(stderr, "Comment, length %ld:\n", (long)length);
+    else                        /* assume it is an APPn otherwise */
+      fprintf(stderr, "APP%d, length %ld:\n",
+              cinfo->unread_marker - JPEG_APP0, (long)length);
+  }
+
+  while (--length >= 0) {
+    ch = jpeg_getc(cinfo);
+    if (traceit) {
+      /* Emit the character in a readable form.
+       * Nonprintables are converted to \nnn form,
+       * while \ is converted to \\.
+       * Newlines in CR, CR/LF, or LF form will be printed as one newline.
+       */
+      if (ch == '\r') {
+        fprintf(stderr, "\n");
+      } else if (ch == '\n') {
+        if (lastch != '\r')
+          fprintf(stderr, "\n");
+      } else if (ch == '\\') {
+        fprintf(stderr, "\\\\");
+      } else if (isprint(ch)) {
+        putc(ch, stderr);
+      } else {
+        fprintf(stderr, "\\%03o", ch);
+      }
+      lastch = ch;
+    }
+  }
+
+  if (traceit)
+    fprintf(stderr, "\n");
+
+  return TRUE;
+}
+
+
+METHODDEF(void)
+my_emit_message(j_common_ptr cinfo, int msg_level)
+{
+  if (msg_level < 0) {
+    /* Treat warning as fatal */
+    cinfo->err->error_exit(cinfo);
+  } else {
+    if (cinfo->err->trace_level >= msg_level)
+      cinfo->err->output_message(cinfo);
+  }
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main(int argc, char **argv)
+{
+  struct jpeg_decompress_struct cinfo;
+  struct jpeg_error_mgr jerr;
+  struct cdjpeg_progress_mgr progress;
+  int file_index;
+  djpeg_dest_ptr dest_mgr = NULL;
+  FILE *input_file;
+  FILE *output_file;
+  unsigned char *inbuffer = NULL;
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  unsigned long insize = 0;
+#endif
+  JDIMENSION num_scanlines;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv0;
+  if (progname == NULL || progname0 == 0)
+    progname = "djpeg";         /* in case C library doesn't provide it */
+
+  /* Initialize the JPEG decompression object with default error handling. */
+  cinfo.err = jpeg_std_error(&jerr);
+  jpeg_create_decompress(&cinfo);
+  /* Add some application-specific error messages (from cderror.h) */
+  jerr.addon_message_table = cdjpeg_message_table;
+  jerr.first_addon_message = JMSG_FIRSTADDONCODE;
+  jerr.last_addon_message = JMSG_LASTADDONCODE;
+
+  /* Insert custom marker processor for COM and APP12.
+   * APP12 is used by some digital camera makers for textual info,
+   * so we provide the ability to display it as text.
+   * If you like, additional APPn marker types can be selected for display,
+   * but don't try to override APP0 or APP14 this way (see libjpeg.txt).
+   */
+  jpeg_set_marker_processor(&cinfo, JPEG_COM, print_text_marker);
+  jpeg_set_marker_processor(&cinfo, JPEG_APP0 + 12, print_text_marker);
+
+  /* Scan command line to find file names. */
+  /* It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are ignored; we will rescan the switches after opening
+   * the input file.
+   * (Exception: tracing level set here controls verbosity for COM markers
+   * found during jpeg_read_header...)
+   */
+
+  file_index = parse_switches(&cinfo, argc, argv, 0, FALSE);
+
+  if (strict)
+    jerr.emit_message = my_emit_message;
+
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have either -outfile switch or explicit output file name */
+  if (outfilename == NULL) {
+    if (file_index != argc - 2) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+    outfilename = argvfile_index + 1;
+  } else {
+    if (file_index != argc - 1) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc - 1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((input_file = fopen(argvfile_index, READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argvfile_index);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    input_file = read_stdin();
+  }
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((output_file = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default output file is stdout */
+    output_file = write_stdout();
+  }
+
+  if (report || max_scans != 0) {
+    start_progress_monitor((j_common_ptr)&cinfo, &progress);
+    progress.report = report;
+    progress.max_scans = max_scans;
+  }
+
+  /* Specify data source for decompression */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  if (memsrc) {
+    size_t nbytes;
+    do {
+      inbuffer = (unsigned char *)realloc(inbuffer, insize + INPUT_BUF_SIZE);
+      if (inbuffer == NULL) {
+        fprintf(stderr, "%s: memory allocation failure\n", progname);
+        exit(EXIT_FAILURE);
+      }
+      nbytes = JFREAD(input_file, &inbufferinsize, INPUT_BUF_SIZE);
+      if (nbytes < INPUT_BUF_SIZE && ferror(input_file)) {
+        if (file_index < argc)
+          fprintf(stderr, "%s: can't read from %s\n", progname,
+                  argvfile_index);
+        else
+          fprintf(stderr, "%s: can't read from stdin\n", progname);
+      }
+      insize += (unsigned long)nbytes;
+    } while (nbytes == INPUT_BUF_SIZE);
+    fprintf(stderr, "Compressed size:  %lu bytes\n", insize);
+    jpeg_mem_src(&cinfo, inbuffer, insize);
+  } else
+#endif
+    jpeg_stdio_src(&cinfo, input_file);
+
+  /* Read file header, set default decompression parameters */
+  (void)jpeg_read_header(&cinfo, TRUE);
+
+  /* Adjust default decompression parameters by re-parsing the options */
+  file_index = parse_switches(&cinfo, argc, argv, 0, TRUE);
+
+  /* Initialize the output module now to let it override any crucial
+   * option settings (for instance, GIF wants to force color quantization).
+   */
+  switch (requested_fmt) {
+#ifdef BMP_SUPPORTED
+  case FMT_BMP:
+    dest_mgr = jinit_write_bmp(&cinfo, FALSE, TRUE);
+    break;
+  case FMT_OS2:
+    dest_mgr = jinit_write_bmp(&cinfo, TRUE, TRUE);
+    break;
+#endif
+#ifdef GIF_SUPPORTED
+  case FMT_GIF:
+    dest_mgr = jinit_write_gif(&cinfo, TRUE);
+    break;
+  case FMT_GIF0:
+    dest_mgr = jinit_write_gif(&cinfo, FALSE);
+    break;
+#endif
+#ifdef PPM_SUPPORTED
+  case FMT_PPM:
+    dest_mgr = jinit_write_ppm(&cinfo);
+    break;
+#endif
+#ifdef TARGA_SUPPORTED
+  case FMT_TARGA:
+    dest_mgr = jinit_write_targa(&cinfo);
+    break;
+#endif
+  default:
+    ERREXIT(&cinfo, JERR_UNSUPPORTED_FORMAT);
+    break;
+  }
+  dest_mgr->output_file = output_file;
+
+  /* Start decompressor */
+  (void)jpeg_start_decompress(&cinfo);
+
+  /* Skip rows */
+  if (skip) {
+    JDIMENSION tmp;
+
+    /* Check for valid skip_end.  We cannot check this value until after
+     * jpeg_start_decompress() is called.  Note that we have already verified
+     * that skip_start <= skip_end.
+     */
+    if (skip_end > cinfo.output_height - 1) {
+      fprintf(stderr, "%s: skip region exceeds image height %d\n", progname,
+              cinfo.output_height);
+      exit(EXIT_FAILURE);
+    }
+
+    /* Write output file header.  This is a hack to ensure that the destination
+     * manager creates an output image of the proper size.
+     */
+    tmp = cinfo.output_height;
+    cinfo.output_height -= (skip_end - skip_start + 1);
+    (*dest_mgr->start_output) (&cinfo, dest_mgr);
+    cinfo.output_height = tmp;
+
+    /* Process data */
+    while (cinfo.output_scanline < skip_start) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+    if ((tmp = jpeg_skip_scanlines(&cinfo, skip_end - skip_start + 1)) !=
+        skip_end - skip_start + 1) {
+      fprintf(stderr, "%s: jpeg_skip_scanlines() returned %d rather than %d\n",
+              progname, tmp, skip_end - skip_start + 1);
+      exit(EXIT_FAILURE);
+    }
+    while (cinfo.output_scanline < cinfo.output_height) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+
+  /* Decompress a subregion */
+  } else if (crop) {
+    JDIMENSION tmp;
+
+    /* Check for valid crop dimensions.  We cannot check these values until
+     * after jpeg_start_decompress() is called.
+     */
+    if (crop_x + crop_width > cinfo.output_width ||
+        crop_y + crop_height > cinfo.output_height) {
+      fprintf(stderr, "%s: crop dimensions exceed image dimensions %d x %d\n",
+              progname, cinfo.output_width, cinfo.output_height);
+      exit(EXIT_FAILURE);
+    }
+
+    jpeg_crop_scanline(&cinfo, &crop_x, &crop_width);
+    if (dest_mgr->calc_buffer_dimensions)
+      (*dest_mgr->calc_buffer_dimensions) (&cinfo, dest_mgr);
+    else
+      ERREXIT(&cinfo, JERR_UNSUPPORTED_FORMAT);
+
+    /* Write output file header.  This is a hack to ensure that the destination
+     * manager creates an output image of the proper size.
+     */
+    tmp = cinfo.output_height;
+    cinfo.output_height = crop_height;
+    (*dest_mgr->start_output) (&cinfo, dest_mgr);
+    cinfo.output_height = tmp;
+
+    /* Process data */
+    if ((tmp = jpeg_skip_scanlines(&cinfo, crop_y)) != crop_y) {
+      fprintf(stderr, "%s: jpeg_skip_scanlines() returned %d rather than %d\n",
+              progname, tmp, crop_y);
+      exit(EXIT_FAILURE);
+    }
+    while (cinfo.output_scanline < crop_y + crop_height) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+    if ((tmp =
+         jpeg_skip_scanlines(&cinfo,
+                             cinfo.output_height - crop_y - crop_height)) !=
+        cinfo.output_height - crop_y - crop_height) {
+      fprintf(stderr, "%s: jpeg_skip_scanlines() returned %d rather than %d\n",
+              progname, tmp, cinfo.output_height - crop_y - crop_height);
+      exit(EXIT_FAILURE);
+    }
+
+  /* Normal full-image decompress */
+  } else {
+    /* Write output file header */
+    (*dest_mgr->start_output) (&cinfo, dest_mgr);
+
+    /* Process data */
+    while (cinfo.output_scanline < cinfo.output_height) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+  }
+
+  /* Hack: count final pass as done in case finish_output does an extra pass.
+   * The library won't have updated completed_passes.
+   */
+  if (report || max_scans != 0)
+    progress.pub.completed_passes = progress.pub.total_passes;
+
+  if (icc_filename != NULL) {
+    FILE *icc_file;
+    JOCTET *icc_profile;
+    unsigned int icc_len;
+
+    if ((icc_file = fopen(icc_filename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, icc_filename);
+      exit(EXIT_FAILURE);
+    }
+    if (jpeg_read_icc_profile(&cinfo, &icc_profile, &icc_len)) {
+      if (fwrite(icc_profile, icc_len, 1, icc_file) < 1) {
+        fprintf(stderr, "%s: can't read ICC profile from %s\n", progname,
+                icc_filename);
+        free(icc_profile);
+        fclose(icc_file);
+        exit(EXIT_FAILURE);
+      }
+      free(icc_profile);
+      fclose(icc_file);
+    } else if (cinfo.err->msg_code != JWRN_BOGUS_ICC)
+      fprintf(stderr, "%s: no ICC profile data in JPEG file\n", progname);
+  }
+
+  /* Finish decompression and release memory.
+   * I must do it in this order because output module has allocated memory
+   * of lifespan JPOOL_IMAGE; it needs to finish before releasing memory.
+   */
+  (*dest_mgr->finish_output) (&cinfo, dest_mgr);
+  (void)jpeg_finish_decompress(&cinfo);
+  jpeg_destroy_decompress(&cinfo);
+
+  /* Close files, if we opened them */
+  if (input_file != stdin)
+    fclose(input_file);
+  if (output_file != stdout)
+    fclose(output_file);
+
+  if (report || max_scans != 0)
+    end_progress_monitor((j_common_ptr)&cinfo);
+
+  if (memsrc)
+    free(inbuffer);
+
+  /* All done. */
+  exit(jerr.num_warnings ? EXIT_WARNING : EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/annotated.html Added

@@ -0,0 +1,83 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: Data Structures</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+</div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+<div class="header">
+  <div class="headertitle">
+<div class="title">Data Structures</div>  </div>
+</div>
+<div class="contents">
+<div class="textblock">Here are the data structures with brief descriptions:</div><div class="directory">
+<table class="directory">
+<tr id="row_0_" class="even"><td class="entry"><span style="width:16px;display:inline-block;">&#160;</span><span class="icona"><span class="icon">C</span></span><a class="el" href="structtjregion.html" target="_self">tjregion</a></td><td class="desc">Cropping region </td></tr>
+<tr id="row_1_"><td class="entry"><span style="width:16px;display:inline-block;">&#160;</span><span class="icona"><span class="icon">C</span></span><a class="el" href="structtjscalingfactor.html" target="_self">tjscalingfactor</a></td><td class="desc">Scaling factor </td></tr>
+<tr id="row_2_" class="even"><td class="entry"><span style="width:16px;display:inline-block;">&#160;</span><span class="icona"><span class="icon">C</span></span><a class="el" href="structtjtransform.html" target="_self">tjtransform</a></td><td class="desc">Lossless transform </td></tr>
+</table>
+</div>
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/bc_s.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/bdwn.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/classes.html Added

@@ -0,0 +1,89 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: Data Structure Index</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+</div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+<div class="header">
+  <div class="headertitle">
+<div class="title">Data Structure Index</div>  </div>
+</div>
+<div class="contents">
+<div class="qindex"><a class="qindex" href="#letter_t">t</a></div>
+<table class="classindex">
+<tr><td rowspan="2" valign="bottom"><a name="letter_t"></a><table border="0" cellspacing="0" cellpadding="0"><tr><td><div class="ah">&#160;&#160;t&#160;&#160;</div></td></tr></table>
+</td>
+<td valign="top"><a class="el" href="structtjscalingfactor.html">tjscalingfactor</a>&#160;&#160;&#160;</td>
+<td valign="top"><a class="el" href="structtjtransform.html">tjtransform</a>&#160;&#160;&#160;</td>
+<td></td></tr>
+<tr><td></td><td></td><td></td></tr>
+<tr><td valign="top"><a class="el" href="structtjregion.html">tjregion</a>&#160;&#160;&#160;</td>
+<td></td><td></td><td></td></tr>
+<tr><td></td><td></td><td></td><td></td></tr>
+</table>
+<div class="qindex"><a class="qindex" href="#letter_t">t</a></div>
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/closed.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/doc.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/doxygen-extra.css Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/doxygen.css Added

@@ -0,0 +1,1732 @@
+/* The standard CSS for doxygen 1.8.20 */
+
+body, table, div, p, dl {
+	font: 400 14px/22px Roboto,sans-serif;
+}
+
+p.reference, p.definition {
+	font: 400 14px/22px Roboto,sans-serif;
+}
+
+/* @group Heading Levels */
+
+h1.groupheader {
+	font-size: 150%;
+}
+
+.title {
+	font: 400 14px/28px Roboto,sans-serif;
+	font-size: 150%;
+	font-weight: bold;
+	margin: 10px 2px;
+}
+
+h2.groupheader {
+	border-bottom: 1px solid #879ECB;
+	color: #354C7B;
+	font-size: 150%;
+	font-weight: normal;
+	margin-top: 1.75em;
+	padding-top: 8px;
+	padding-bottom: 4px;
+	width: 100%;
+}
+
+h3.groupheader {
+	font-size: 100%;
+}
+
+h1, h2, h3, h4, h5, h6 {
+	-webkit-transition: text-shadow 0.5s linear;
+	-moz-transition: text-shadow 0.5s linear;
+	-ms-transition: text-shadow 0.5s linear;
+	-o-transition: text-shadow 0.5s linear;
+	transition: text-shadow 0.5s linear;
+	margin-right: 15px;
+}
+
+h1.glow, h2.glow, h3.glow, h4.glow, h5.glow, h6.glow {
+	text-shadow: 0 0 15px cyan;
+}
+
+dt {
+	font-weight: bold;
+}
+
+ul.multicol {
+	-moz-column-gap: 1em;
+	-webkit-column-gap: 1em;
+	column-gap: 1em;
+	-moz-column-count: 3;
+	-webkit-column-count: 3;
+	column-count: 3;
+}
+
+p.startli, p.startdd {
+	margin-top: 2px;
+}
+
+th p.starttd, th p.intertd, th p.endtd {
+        font-size: 100%;
+        font-weight: 700;
+}
+
+p.starttd {
+	margin-top: 0px;
+}
+
+p.endli {
+	margin-bottom: 0px;
+}
+
+p.enddd {
+	margin-bottom: 4px;
+}
+
+p.endtd {
+	margin-bottom: 2px;
+}
+
+p.interli {
+}
+
+p.interdd {
+}
+
+p.intertd {
+}
+
+/* @end */
+
+caption {
+	font-weight: bold;
+}
+
+span.legend {
+        font-size: 70%;
+        text-align: center;
+}
+
+h3.version {
+        font-size: 90%;
+        text-align: center;
+}
+
+div.qindex, div.navtab{
+	background-color: #EBEFF6;
+	border: 1px solid #A3B4D7;
+	text-align: center;
+}
+
+div.qindex, div.navpath {
+	width: 100%;
+	line-height: 140%;
+}
+
+div.navtab {
+	margin-right: 15px;
+}
+
+/* @group Link Styling */
+
+a {
+	color: #3D578C;
+	font-weight: normal;
+	text-decoration: none;
+}
+
+.contents a:visited {
+	color: #4665A2;
+}
+
+a:hover {
+	text-decoration: underline;
+}
+
+a.qindex {
+	font-weight: bold;
+}
+
+a.qindexHL {
+	font-weight: bold;
+	background-color: #9CAFD4;
+	color: #FFFFFF;
+	border: 1px double #869DCA;
+}
+
+.contents a.qindexHL:visited {
+        color: #FFFFFF;
+}
+
+a.el {
+	font-weight: bold;
+}
+
+a.elRef {
+}
+
+a.code, a.code:visited, a.line, a.line:visited {
+	color: #4665A2; 
+}
+
+a.codeRef, a.codeRef:visited, a.lineRef, a.lineRef:visited {
+	color: #4665A2; 
+}
+
+/* @end */
+
+dl.el {
+	margin-left: -1cm;
+}
+
+ul {
+  overflow: hidden; /*Fixed: list item bullets overlap floating elements*/
+}
+
+#side-nav ul {
+  overflow: visible; /* reset ul rule for scroll bar in GENERATE_TREEVIEW window */
+}
+
+#main-nav ul {
+  overflow: visible; /* reset ul rule for the navigation bar drop down lists */
+}
+
+.fragment {
+  text-align: left;
+  direction: ltr;
+  overflow-x: auto; /*Fixed: fragment lines overlap floating elements*/
+  overflow-y: hidden;
+}
+
+pre.fragment {
+        border: 1px solid #C4CFE5;
+        background-color: #FBFCFD;
+        padding: 4px 6px;
+        margin: 4px 8px 4px 2px;
+        overflow: auto;
+        word-wrap: break-word;
+        font-size:  9pt;
+        line-height: 125%;
+        font-family: monospace, fixed;
+        font-size: 105%;
+}
+
+div.fragment {
+  padding: 0 0 1px 0; /*Fixed: last line underline overlap border*/
+  margin: 4px 8px 4px 2px;
+	background-color: #FBFCFD;
+	border: 1px solid #C4CFE5;
+}
+
+div.line {
+	font-family: monospace, fixed;
+        font-size: 13px;
+	min-height: 13px;
+	line-height: 1.0;
+	text-wrap: unrestricted;
+	white-space: -moz-pre-wrap; /* Moz */
+	white-space: -pre-wrap;     /* Opera 4-6 */
+	white-space: -o-pre-wrap;   /* Opera 7 */
+	white-space: pre-wrap;      /* CSS3  */
+	word-wrap: break-word;      /* IE 5.5+ */
+	text-indent: -53px;
+	padding-left: 53px;
+	padding-bottom: 0px;
+	margin: 0px;
+	-webkit-transition-property: background-color, box-shadow;
+	-webkit-transition-duration: 0.5s;
+	-moz-transition-property: background-color, box-shadow;
+	-moz-transition-duration: 0.5s;
+	-ms-transition-property: background-color, box-shadow;
+	-ms-transition-duration: 0.5s;
+	-o-transition-property: background-color, box-shadow;
+	-o-transition-duration: 0.5s;
+	transition-property: background-color, box-shadow;
+	transition-duration: 0.5s;
+}
+
+div.line:after {
+    content:"\000A";
+    white-space: pre;
+}
+
+div.line.glow {
+	background-color: cyan;
+	box-shadow: 0 0 10px cyan;
+}
+
+
+span.lineno {
+	padding-right: 4px;
+	text-align: right;
+	border-right: 2px solid #0F0;
+	background-color: #E8E8E8;
+        white-space: pre;
+}
+span.lineno a {
+	background-color: #D8D8D8;
+}
+
+span.lineno a:hover {
+	background-color: #C8C8C8;
+}
+
+.lineno {
+	-webkit-touch-callout: none;
+	-webkit-user-select: none;
+	-khtml-user-select: none;
+	-moz-user-select: none;
+	-ms-user-select: none;
+	user-select: none;
+}
+
+div.ah, span.ah {
+	background-color: black;
+	font-weight: bold;
+	color: #FFFFFF;
+	margin-bottom: 3px;
+	margin-top: 3px;
+	padding: 0.2em;
+	border: solid thin #333;
+	border-radius: 0.5em;
+	-webkit-border-radius: .5em;
+	-moz-border-radius: .5em;
+	box-shadow: 2px 2px 3px #999;
+	-webkit-box-shadow: 2px 2px 3px #999;
+	-moz-box-shadow: rgba(0, 0, 0, 0.15) 2px 2px 2px;
+	background-image: -webkit-gradient(linear, left top, left bottom, from(#eee), to(#000),color-stop(0.3, #444));
+	background-image: -moz-linear-gradient(center top, #eee 0%, #444 40%, #000 110%);
+}
+
+div.classindex ul {
+        list-style: none;
+        padding-left: 0;
+}
+
+div.classindex span.ai {
+        display: inline-block;
+}
+
+div.groupHeader {
+	margin-left: 16px;
+	margin-top: 12px;
+	font-weight: bold;
+}
+
+div.groupText {
+	margin-left: 16px;
+	font-style: italic;
+}
+
+body {
+	background-color: white;
+	color: black;
+        margin: 0;
+}
+
+div.contents {
+	margin-top: 10px;
+	margin-left: 12px;
+	margin-right: 8px;
+}
+
+td.indexkey {
+	background-color: #EBEFF6;
+	font-weight: bold;
+	border: 1px solid #C4CFE5;
+	margin: 2px 0px 2px 0;
+	padding: 2px 10px;
+        white-space: nowrap;
+        vertical-align: top;
+}
+
+td.indexvalue {
+	background-color: #EBEFF6;
+	border: 1px solid #C4CFE5;
+	padding: 2px 10px;
+	margin: 2px 0px;
+}
+
+tr.memlist {
+	background-color: #EEF1F7;
+}
+
+p.formulaDsp {
+	text-align: center;
+}
+
+img.formulaDsp {
+	
+}
+
+img.formulaInl, img.inline {
+	vertical-align: middle;
+}
+
+div.center {
+	text-align: center;
+        margin-top: 0px;
+        margin-bottom: 0px;
+        padding: 0px;
+}
+
+div.center img {
+	border: 0px;
+}
+
+address.footer {
+	text-align: right;
+	padding-right: 12px;
+}
+
+img.footer {
+	border: 0px;
+	vertical-align: middle;
+}
+
+/* @group Code Colorization */
+
+span.keyword {
+	color: #008000
+}
+
+span.keywordtype {
+	color: #604020
+}
+
+span.keywordflow {
+	color: #e08000
+}
+
+span.comment {
+	color: #800000
+}
+
+span.preprocessor {
+	color: #806020
+}
+
+span.stringliteral {
+	color: #002080
+}
+
+span.charliteral {
+	color: #008080
+}
+
+span.vhdldigit { 
+	color: #ff00ff 
+}
+
+span.vhdlchar { 
+	color: #000000 
+}
+
+span.vhdlkeyword { 
+	color: #700070 
+}
+
+span.vhdllogic { 
+	color: #ff0000 
+}
+
+blockquote {
+        background-color: #F7F8FB;
+        border-left: 2px solid #9CAFD4;
+        margin: 0 24px 0 4px;
+        padding: 0 12px 0 16px;
+}
+
+blockquote.DocNodeRTL {
+   border-left: 0;
+   border-right: 2px solid #9CAFD4;
+   margin: 0 4px 0 24px;
+   padding: 0 16px 0 12px;
+}
+
+/* @end */
+
+/*
+.search {
+	color: #003399;
+	font-weight: bold;
+}
+
+form.search {
+	margin-bottom: 0px;
+	margin-top: 0px;
+}
+
+input.search {
+	font-size: 75%;
+	color: #000080;
+	font-weight: normal;
+	background-color: #e8eef2;
+}
+*/
+
+td.tiny {
+	font-size: 75%;
+}
+
+.dirtab {
+	padding: 4px;
+	border-collapse: collapse;
+	border: 1px solid #A3B4D7;
+}
+
+th.dirtab {
+	background: #EBEFF6;
+	font-weight: bold;
+}
+
+hr {
+	height: 0px;
+	border: none;
+	border-top: 1px solid #4A6AAA;
+}
+
+hr.footer {
+	height: 1px;
+}
+
+/* @group Member Descriptions */
+
+table.memberdecls {
+	border-spacing: 0px;
+	padding: 0px;
+}
+
+.memberdecls td, .fieldtable tr {
+	-webkit-transition-property: background-color, box-shadow;
+	-webkit-transition-duration: 0.5s;
+	-moz-transition-property: background-color, box-shadow;
+	-moz-transition-duration: 0.5s;
+	-ms-transition-property: background-color, box-shadow;
+	-ms-transition-duration: 0.5s;
+	-o-transition-property: background-color, box-shadow;
+	-o-transition-duration: 0.5s;
+	transition-property: background-color, box-shadow;
+	transition-duration: 0.5s;
+}
+
+.memberdecls td.glow, .fieldtable tr.glow {
+	background-color: cyan;
+	box-shadow: 0 0 15px cyan;
+}
+
+.mdescLeft, .mdescRight,
+.memItemLeft, .memItemRight,
+.memTemplItemLeft, .memTemplItemRight, .memTemplParams {
+	background-color: #F9FAFC;
+	border: none;
+	margin: 4px;
+	padding: 1px 0 0 8px;
+}
+
+.mdescLeft, .mdescRight {
+	padding: 0px 8px 4px 8px;
+	color: #555;
+}
+
+.memSeparator {
+        border-bottom: 1px solid #DEE4F0;
+        line-height: 1px;
+        margin: 0px;
+        padding: 0px;
+}
+
+.memItemLeft, .memTemplItemLeft {
+        white-space: nowrap;
+}
+
+.memItemRight, .memTemplItemRight {
+	width: 100%;
+}
+
+.memTemplParams {
+	color: #4665A2;
+        white-space: nowrap;
+	font-size: 80%;
+}
+
+/* @end */
+
+/* @group Member Details */
+
+/* Styles for detailed member documentation */
+
+.memtitle {
+	padding: 8px;
+	border-top: 1px solid #A8B8D9;
+	border-left: 1px solid #A8B8D9;
+	border-right: 1px solid #A8B8D9;
+	border-top-right-radius: 4px;
+	border-top-left-radius: 4px;
+	margin-bottom: -1px;
+	background-image: url('nav_f.png');
+	background-repeat: repeat-x;
+	background-color: #E2E8F2;
+	line-height: 1.25;
+	font-weight: 300;
+	float:left;
+}
+
+.permalink
+{
+        font-size: 65%;
+        display: inline-block;
+        vertical-align: middle;
+}
+
+.memtemplate {
+	font-size: 80%;
+	color: #4665A2;
+	font-weight: normal;
+	margin-left: 9px;
+}
+
+.memnav {
+	background-color: #EBEFF6;
+	border: 1px solid #A3B4D7;
+	text-align: center;
+	margin: 2px;
+	margin-right: 15px;
+	padding: 2px;
+}
+
+.mempage {
+	width: 100%;
+}
+
+.memitem {
+	padding: 0;
+	margin-bottom: 10px;
+	margin-right: 5px;
+        -webkit-transition: box-shadow 0.5s linear;
+        -moz-transition: box-shadow 0.5s linear;
+        -ms-transition: box-shadow 0.5s linear;
+        -o-transition: box-shadow 0.5s linear;
+        transition: box-shadow 0.5s linear;
+        display: table !important;
+        width: 100%;
+}
+
+.memitem.glow {
+         box-shadow: 0 0 15px cyan;
+}
+
+.memname {
+        font-weight: 400;
+        margin-left: 6px;
+}
+
+.memname td {
+	vertical-align: bottom;
+}
+
+.memproto, dl.reflist dt {
+        border-top: 1px solid #A8B8D9;
+        border-left: 1px solid #A8B8D9;
+        border-right: 1px solid #A8B8D9;
+        padding: 6px 0px 6px 0px;
+        color: #253555;
+        font-weight: bold;
+        text-shadow: 0px 1px 1px rgba(255, 255, 255, 0.9);
+        background-color: #DFE5F1;
+        /* opera specific markup */
+        box-shadow: 5px 5px 5px rgba(0, 0, 0, 0.15);
+        border-top-right-radius: 4px;
+        /* firefox specific markup */
+        -moz-box-shadow: rgba(0, 0, 0, 0.15) 5px 5px 5px;
+        -moz-border-radius-topright: 4px;
+        /* webkit specific markup */
+        -webkit-box-shadow: 5px 5px 5px rgba(0, 0, 0, 0.15);
+        -webkit-border-top-right-radius: 4px;
+
+}
+
+.overload {
+        font-family: "courier new",courier,monospace;
+	font-size: 65%;
+}
+
+.memdoc, dl.reflist dd {
+        border-bottom: 1px solid #A8B8D9;      
+        border-left: 1px solid #A8B8D9;      
+        border-right: 1px solid #A8B8D9; 
+        padding: 6px 10px 2px 10px;
+        background-color: #FBFCFD;
+        border-top-width: 0;
+        background-image:url('nav_g.png');
+        background-repeat:repeat-x;
+        background-color: #FFFFFF;
+        /* opera specific markup */
+        border-bottom-left-radius: 4px;
+        border-bottom-right-radius: 4px;
+        box-shadow: 5px 5px 5px rgba(0, 0, 0, 0.15);
+        /* firefox specific markup */
+        -moz-border-radius-bottomleft: 4px;
+        -moz-border-radius-bottomright: 4px;
+        -moz-box-shadow: rgba(0, 0, 0, 0.15) 5px 5px 5px;
+        /* webkit specific markup */
+        -webkit-border-bottom-left-radius: 4px;
+        -webkit-border-bottom-right-radius: 4px;
+        -webkit-box-shadow: 5px 5px 5px rgba(0, 0, 0, 0.15);
+}
+
+dl.reflist dt {
+        padding: 5px;
+}
+
+dl.reflist dd {
+        margin: 0px 0px 10px 0px;
+        padding: 5px;
+}
+
+.paramkey {
+	text-align: right;
+}
+
+.paramtype {
+	white-space: nowrap;
+}
+
+.paramname {
+	color: #602020;
+	white-space: nowrap;
+}
+.paramname em {
+	font-style: normal;
+}
+.paramname code {
+        line-height: 14px;
+}
+
+.params, .retval, .exception, .tparams {
+        margin-left: 0px;
+        padding-left: 0px;
+}       
+
+.params .paramname, .retval .paramname, .tparams .paramname, .exception .paramname {
+        font-weight: bold;
+        vertical-align: top;
+}
+        
+.params .paramtype, .tparams .paramtype {
+        font-style: italic;
+        vertical-align: top;
+}       
+        
+.params .paramdir, .tparams .paramdir {
+        font-family: "courier new",courier,monospace;
+        vertical-align: top;
+}
+
+table.mlabels {
+	border-spacing: 0px;
+}
+
+td.mlabels-left {
+	width: 100%;
+	padding: 0px;
+}
+
+td.mlabels-right {
+	vertical-align: bottom;
+	padding: 0px;
+	white-space: nowrap;
+}
+
+span.mlabels {
+        margin-left: 8px;
+}
+
+span.mlabel {
+        background-color: #728DC1;
+        border-top:1px solid #5373B4;
+        border-left:1px solid #5373B4;
+        border-right:1px solid #C4CFE5;
+        border-bottom:1px solid #C4CFE5;
+	text-shadow: none;
+	color: white;
+	margin-right: 4px;
+	padding: 2px 3px;
+	border-radius: 3px;
+	font-size: 7pt;
+	white-space: nowrap;
+	vertical-align: middle;
+}
+
+
+
+/* @end */
+
+/* these are for tree view inside a (index) page */
+
+div.directory {
+        margin: 10px 0px;
+        border-top: 1px solid #9CAFD4;
+        border-bottom: 1px solid #9CAFD4;
+        width: 100%;
+}
+
+.directory table {
+        border-collapse:collapse;
+}
+
+.directory td {
+        margin: 0px;
+        padding: 0px;
+	vertical-align: top;
+}
+
+.directory td.entry {
+        white-space: nowrap;
+        padding-right: 6px;
+	padding-top: 3px;
+}
+
+.directory td.entry a {
+        outline:none;
+}
+
+.directory td.entry a img {
+        border: none;
+}
+
+.directory td.desc {
+        width: 100%;
+        padding-left: 6px;
+	padding-right: 6px;
+	padding-top: 3px;
+	border-left: 1px solid rgba(0,0,0,0.05);
+}
+
+.directory tr.even {
+	padding-left: 6px;
+	background-color: #F7F8FB;
+}
+
+.directory img {
+	vertical-align: -30%;
+}
+
+.directory .levels {
+        white-space: nowrap;
+        width: 100%;
+        text-align: right;
+        font-size: 9pt;
+}
+
+.directory .levels span {
+        cursor: pointer;
+        padding-left: 2px;
+        padding-right: 2px;
+	color: #3D578C;
+}
+
+.arrow {
+    color: #9CAFD4;
+    -webkit-user-select: none;
+    -khtml-user-select: none;
+    -moz-user-select: none;
+    -ms-user-select: none;
+    user-select: none;
+    cursor: pointer;
+    font-size: 80%;
+    display: inline-block;
+    width: 16px;
+    height: 22px;
+}
+
+.icon {
+    font-family: Arial, Helvetica;
+    font-weight: bold;
+    font-size: 12px;
+    height: 14px;
+    width: 16px;
+    display: inline-block;
+    background-color: #728DC1;
+    color: white;
+    text-align: center;
+    border-radius: 4px;
+    margin-left: 2px;
+    margin-right: 2px;
+}
+
+.icona {
+    width: 24px;
+    height: 22px;
+    display: inline-block;
+}
+
+.iconfopen {
+    width: 24px;
+    height: 18px;
+    margin-bottom: 4px;
+    background-image:url('folderopen.png');
+    background-position: 0px -4px;
+    background-repeat: repeat-y;
+    vertical-align:top;
+    display: inline-block;
+}
+
+.iconfclosed {
+    width: 24px;
+    height: 18px;
+    margin-bottom: 4px;
+    background-image:url('folderclosed.png');
+    background-position: 0px -4px;
+    background-repeat: repeat-y;
+    vertical-align:top;
+    display: inline-block;
+}
+
+.icondoc {
+    width: 24px;
+    height: 18px;
+    margin-bottom: 4px;
+    background-image:url('doc.png');
+    background-position: 0px -4px;
+    background-repeat: repeat-y;
+    vertical-align:top;
+    display: inline-block;
+}
+
+table.directory {
+    font: 400 14px Roboto,sans-serif;
+}
+
+/* @end */
+
+div.dynheader {
+        margin-top: 8px;
+	-webkit-touch-callout: none;
+	-webkit-user-select: none;
+	-khtml-user-select: none;
+	-moz-user-select: none;
+	-ms-user-select: none;
+	user-select: none;
+}
+
+address {
+	font-style: normal;
+	color: #2A3D61;
+}
+
+table.doxtable caption {
+	caption-side: top;
+}
+
+table.doxtable {
+	border-collapse:collapse;
+        margin-top: 4px;
+        margin-bottom: 4px;
+}
+
+table.doxtable td, table.doxtable th {
+	border: 1px solid #2D4068;
+	padding: 3px 7px 2px;
+}
+
+table.doxtable th {
+	background-color: #374F7F;
+	color: #FFFFFF;
+	font-size: 110%;
+	padding-bottom: 4px;
+	padding-top: 5px;
+}
+
+table.fieldtable {
+        /*width: 100%;*/
+        margin-bottom: 10px;
+        border: 1px solid #A8B8D9;
+        border-spacing: 0px;
+        -moz-border-radius: 4px;
+        -webkit-border-radius: 4px;
+        border-radius: 4px;
+        -moz-box-shadow: rgba(0, 0, 0, 0.15) 2px 2px 2px;
+        -webkit-box-shadow: 2px 2px 2px rgba(0, 0, 0, 0.15);
+        box-shadow: 2px 2px 2px rgba(0, 0, 0, 0.15);
+}
+
+.fieldtable td, .fieldtable th {
+        padding: 3px 7px 2px;
+}
+
+.fieldtable td.fieldtype, .fieldtable td.fieldname {
+        white-space: nowrap;
+        border-right: 1px solid #A8B8D9;
+        border-bottom: 1px solid #A8B8D9;
+        vertical-align: top;
+}
+
+.fieldtable td.fieldname {
+        padding-top: 3px;
+}
+
+.fieldtable td.fielddoc {
+        border-bottom: 1px solid #A8B8D9;
+        /*width: 100%;*/
+}
+
+.fieldtable td.fielddoc p:first-child {
+        margin-top: 0px;
+}       
+        
+.fieldtable td.fielddoc p:last-child {
+        margin-bottom: 2px;
+}
+
+.fieldtable tr:last-child td {
+        border-bottom: none;
+}
+
+.fieldtable th {
+        background-image:url('nav_f.png');
+        background-repeat:repeat-x;
+        background-color: #E2E8F2;
+        font-size: 90%;
+        color: #253555;
+        padding-bottom: 4px;
+        padding-top: 5px;
+        text-align:left;
+        font-weight: 400;
+        -moz-border-radius-topleft: 4px;
+        -moz-border-radius-topright: 4px;
+        -webkit-border-top-left-radius: 4px;
+        -webkit-border-top-right-radius: 4px;
+        border-top-left-radius: 4px;
+        border-top-right-radius: 4px;
+        border-bottom: 1px solid #A8B8D9;
+}
+
+
+.tabsearch {
+	top: 0px;
+	left: 10px;
+	height: 36px;
+	background-image: url('tab_b.png');
+	z-index: 101;
+	overflow: hidden;
+	font-size: 13px;
+}
+
+.navpath ul
+{
+	font-size: 11px;
+	background-image:url('tab_b.png');
+	background-repeat:repeat-x;
+	background-position: 0 -5px;
+	height:30px;
+	line-height:30px;
+	color:#8AA0CC;
+	border:solid 1px #C2CDE4;
+	overflow:hidden;
+	margin:0px;
+	padding:0px;
+}
+
+.navpath li
+{
+	list-style-type:none;
+	float:left;
+	padding-left:10px;
+	padding-right:15px;
+	background-image:url('bc_s.png');
+	background-repeat:no-repeat;
+	background-position:right;
+	color:#364D7C;
+}
+
+.navpath li.navelem a
+{
+	height:32px;
+	display:block;
+	text-decoration: none;
+	outline: none;
+	color: #283A5D;
+	font-family: 'Lucida Grande',Geneva,Helvetica,Arial,sans-serif;
+	text-shadow: 0px 1px 1px rgba(255, 255, 255, 0.9);
+	text-decoration: none;        
+}
+
+.navpath li.navelem a:hover
+{
+	color:#6884BD;
+}
+
+.navpath li.footer
+{
+        list-style-type:none;
+        float:right;
+        padding-left:10px;
+        padding-right:15px;
+        background-image:none;
+        background-repeat:no-repeat;
+        background-position:right;
+        color:#364D7C;
+        font-size: 8pt;
+}
+
+
+div.summary
+{
+	float: right;
+	font-size: 8pt;
+	padding-right: 5px;
+	width: 50%;
+	text-align: right;
+}       
+
+div.summary a
+{
+	white-space: nowrap;
+}
+
+table.classindex
+{
+        margin: 10px;
+        white-space: nowrap;
+        margin-left: 3%;
+        margin-right: 3%;
+        width: 94%;
+        border: 0;
+        border-spacing: 0; 
+        padding: 0;
+}
+
+div.ingroups
+{
+	font-size: 8pt;
+	width: 50%;
+	text-align: left;
+}
+
+div.ingroups a
+{
+	white-space: nowrap;
+}
+
+div.header
+{
+        background-image:url('nav_h.png');
+        background-repeat:repeat-x;
+	background-color: #F9FAFC;
+	margin:  0px;
+	border-bottom: 1px solid #C4CFE5;
+}
+
+div.headertitle
+{
+	padding: 5px 5px 5px 10px;
+}
+
+.PageDocRTL-title div.headertitle {
+  text-align: right;
+  direction: rtl;
+}
+
+dl {
+        padding: 0 0 0 0;
+}
+
+/* dl.note, dl.warning, dl.attention, dl.pre, dl.post, dl.invariant, dl.deprecated, dl.todo, dl.test, dl.bug, dl.examples */
+dl.section {
+	margin-left: 0px;
+	padding-left: 0px;
+}
+
+dl.section.DocNodeRTL {
+  margin-right: 0px;
+  padding-right: 0px;
+}
+
+dl.note {
+  margin-left: -7px;
+  padding-left: 3px;
+  border-left: 4px solid;
+  border-color: #D0C000;
+}
+
+dl.note.DocNodeRTL {
+  margin-left: 0;
+  padding-left: 0;
+  border-left: 0;
+  margin-right: -7px;
+  padding-right: 3px;
+  border-right: 4px solid;
+  border-color: #D0C000;
+}
+
+dl.warning, dl.attention {
+  margin-left: -7px;
+  padding-left: 3px;
+  border-left: 4px solid;
+  border-color: #FF0000;
+}
+
+dl.warning.DocNodeRTL, dl.attention.DocNodeRTL {
+  margin-left: 0;
+  padding-left: 0;
+  border-left: 0;
+  margin-right: -7px;
+  padding-right: 3px;
+  border-right: 4px solid;
+  border-color: #FF0000;
+}
+
+dl.pre, dl.post, dl.invariant {
+  margin-left: -7px;
+  padding-left: 3px;
+  border-left: 4px solid;
+  border-color: #00D000;
+}
+
+dl.pre.DocNodeRTL, dl.post.DocNodeRTL, dl.invariant.DocNodeRTL {
+  margin-left: 0;
+  padding-left: 0;
+  border-left: 0;
+  margin-right: -7px;
+  padding-right: 3px;
+  border-right: 4px solid;
+  border-color: #00D000;
+}
+
+dl.deprecated {
+  margin-left: -7px;
+  padding-left: 3px;
+  border-left: 4px solid;
+  border-color: #505050;
+}
+
+dl.deprecated.DocNodeRTL {
+  margin-left: 0;
+  padding-left: 0;
+  border-left: 0;
+  margin-right: -7px;
+  padding-right: 3px;
+  border-right: 4px solid;
+  border-color: #505050;
+}
+
+dl.todo {
+  margin-left: -7px;
+  padding-left: 3px;
+  border-left: 4px solid;
+  border-color: #00C0E0;
+}
+
+dl.todo.DocNodeRTL {
+  margin-left: 0;
+  padding-left: 0;
+  border-left: 0;
+  margin-right: -7px;
+  padding-right: 3px;
+  border-right: 4px solid;
+  border-color: #00C0E0;
+}
+
+dl.test {
+  margin-left: -7px;
+  padding-left: 3px;
+  border-left: 4px solid;
+  border-color: #3030E0;
+}
+
+dl.test.DocNodeRTL {
+  margin-left: 0;
+  padding-left: 0;
+  border-left: 0;
+  margin-right: -7px;
+  padding-right: 3px;
+  border-right: 4px solid;
+  border-color: #3030E0;
+}
+
+dl.bug {
+  margin-left: -7px;
+  padding-left: 3px;
+  border-left: 4px solid;
+  border-color: #C08050;
+}
+
+dl.bug.DocNodeRTL {
+  margin-left: 0;
+  padding-left: 0;
+  border-left: 0;
+  margin-right: -7px;
+  padding-right: 3px;
+  border-right: 4px solid;
+  border-color: #C08050;
+}
+
+dl.section dd {
+	margin-bottom: 6px;
+}
+
+
+#projectlogo
+{
+	text-align: center;
+	vertical-align: bottom;
+	border-collapse: separate;
+}
+ 
+#projectlogo img
+{ 
+	border: 0px none;
+}
+ 
+#projectalign
+{
+        vertical-align: middle;
+}
+
+#projectname
+{
+	font: 300% Tahoma, Arial,sans-serif;
+	margin: 0px;
+	padding: 2px 0px;
+}
+    
+#projectbrief
+{
+	font: 120% Tahoma, Arial,sans-serif;
+	margin: 0px;
+	padding: 0px;
+}
+
+#projectnumber
+{
+	font: 50% Tahoma, Arial,sans-serif;
+	margin: 0px;
+	padding: 0px;
+}
+
+#titlearea
+{
+	padding: 0px;
+	margin: 0px;
+	width: 100%;
+	border-bottom: 1px solid #5373B4;
+}
+
+.image
+{
+        text-align: center;
+}
+
+.dotgraph
+{
+        text-align: center;
+}
+
+.mscgraph
+{
+        text-align: center;
+}
+
+.plantumlgraph
+{
+        text-align: center;
+}
+
+.diagraph
+{
+        text-align: center;
+}
+
+.caption
+{
+	font-weight: bold;
+}
+
+div.zoom
+{
+	border: 1px solid #90A5CE;
+}
+
+dl.citelist {
+        margin-bottom:50px;
+}
+
+dl.citelist dt {
+        color:#334975;
+        float:left;
+        font-weight:bold;
+        margin-right:10px;
+        padding:5px;
+        text-align:right;
+        width:52px;
+}
+
+dl.citelist dd {
+        margin:2px 0 2px 72px;
+        padding:5px 0;
+}
+
+div.toc {
+        padding: 14px 25px;
+        background-color: #F4F6FA;
+        border: 1px solid #D8DFEE;
+        border-radius: 7px 7px 7px 7px;
+        float: right;
+        height: auto;
+        margin: 0 8px 10px 10px;
+        width: 200px;
+}
+
+.PageDocRTL-title div.toc {
+  float: left !important;
+  text-align: right;
+}
+
+div.toc li {
+        background: url("bdwn.png") no-repeat scroll 0 5px transparent;
+        font: 10px/1.2 Verdana,DejaVu Sans,Geneva,sans-serif;
+        margin-top: 5px;
+        padding-left: 10px;
+        padding-top: 2px;
+}
+
+.PageDocRTL-title div.toc li {
+  background-position-x: right !important;
+  padding-left: 0 !important;
+  padding-right: 10px;
+}
+
+div.toc h3 {
+        font: bold 12px/1.2 Arial,FreeSans,sans-serif;
+	color: #4665A2;
+        border-bottom: 0 none;
+        margin: 0;
+}
+
+div.toc ul {
+        list-style: none outside none;
+        border: medium none;
+        padding: 0px;
+}       
+
+div.toc li.level1 {
+        margin-left: 0px;
+}
+
+div.toc li.level2 {
+        margin-left: 15px;
+}
+
+div.toc li.level3 {
+        margin-left: 30px;
+}
+
+div.toc li.level4 {
+        margin-left: 45px;
+}
+
+.PageDocRTL-title div.toc li.level1 {
+  margin-left: 0 !important;
+  margin-right: 0;
+}
+
+.PageDocRTL-title div.toc li.level2 {
+  margin-left: 0 !important;
+  margin-right: 15px;
+}
+
+.PageDocRTL-title div.toc li.level3 {
+  margin-left: 0 !important;
+  margin-right: 30px;
+}
+
+.PageDocRTL-title div.toc li.level4 {
+  margin-left: 0 !important;
+  margin-right: 45px;
+}
+
+.inherit_header {
+        font-weight: bold;
+        color: gray;
+        cursor: pointer;
+	-webkit-touch-callout: none;
+	-webkit-user-select: none;
+	-khtml-user-select: none;
+	-moz-user-select: none;
+	-ms-user-select: none;
+	user-select: none;
+}
+
+.inherit_header td {
+        padding: 6px 0px 2px 5px;
+}
+
+.inherit {
+        display: none;
+}
+
+tr.heading h2 {
+        margin-top: 12px;
+        margin-bottom: 4px;
+}
+
+/* tooltip related style info */
+
+.ttc {
+        position: absolute;
+        display: none;
+}
+
+#powerTip {
+	cursor: default;
+	white-space: nowrap;
+	background-color: white;
+	border: 1px solid gray;
+	border-radius: 4px 4px 4px 4px;
+	box-shadow: 1px 1px 7px gray;
+	display: none;
+	font-size: smaller;
+	max-width: 80%;
+	opacity: 0.9;
+	padding: 1ex 1em 1em;
+	position: absolute;
+	z-index: 2147483647;
+}
+
+#powerTip div.ttdoc {
+        color: grey;
+	font-style: italic;
+}
+
+#powerTip div.ttname a {
+        font-weight: bold;
+}
+
+#powerTip div.ttname {
+        font-weight: bold;
+}
+
+#powerTip div.ttdeci {
+        color: #006318;
+}
+
+#powerTip div {
+        margin: 0px;
+        padding: 0px;
+        font: 12px/16px Roboto,sans-serif;
+}
+
+#powerTip:before, #powerTip:after {
+	content: "";
+	position: absolute;
+	margin: 0px;
+}
+
+#powerTip.n:after,  #powerTip.n:before,
+#powerTip.s:after,  #powerTip.s:before,
+#powerTip.w:after,  #powerTip.w:before,
+#powerTip.e:after,  #powerTip.e:before,
+#powerTip.ne:after, #powerTip.ne:before,
+#powerTip.se:after, #powerTip.se:before,
+#powerTip.nw:after, #powerTip.nw:before,
+#powerTip.sw:after, #powerTip.sw:before {
+	border: solid transparent;
+	content: " ";
+	height: 0;
+	width: 0;
+	position: absolute;
+}
+
+#powerTip.n:after,  #powerTip.s:after,
+#powerTip.w:after,  #powerTip.e:after,
+#powerTip.nw:after, #powerTip.ne:after,
+#powerTip.sw:after, #powerTip.se:after {
+	border-color: rgba(255, 255, 255, 0);
+}
+
+#powerTip.n:before,  #powerTip.s:before,
+#powerTip.w:before,  #powerTip.e:before,
+#powerTip.nw:before, #powerTip.ne:before,
+#powerTip.sw:before, #powerTip.se:before {
+	border-color: rgba(128, 128, 128, 0);
+}
+
+#powerTip.n:after,  #powerTip.n:before,
+#powerTip.ne:after, #powerTip.ne:before,
+#powerTip.nw:after, #powerTip.nw:before {
+	top: 100%;
+}
+
+#powerTip.n:after, #powerTip.ne:after, #powerTip.nw:after {
+	border-top-color: #FFFFFF;
+	border-width: 10px;
+	margin: 0px -10px;
+}
+#powerTip.n:before {
+	border-top-color: #808080;
+	border-width: 11px;
+	margin: 0px -11px;
+}
+#powerTip.n:after, #powerTip.n:before {
+	left: 50%;
+}
+
+#powerTip.nw:after, #powerTip.nw:before {
+	right: 14px;
+}
+
+#powerTip.ne:after, #powerTip.ne:before {
+	left: 14px;
+}
+
+#powerTip.s:after,  #powerTip.s:before,
+#powerTip.se:after, #powerTip.se:before,
+#powerTip.sw:after, #powerTip.sw:before {
+	bottom: 100%;
+}
+
+#powerTip.s:after, #powerTip.se:after, #powerTip.sw:after {
+	border-bottom-color: #FFFFFF;
+	border-width: 10px;
+	margin: 0px -10px;
+}
+
+#powerTip.s:before, #powerTip.se:before, #powerTip.sw:before {
+	border-bottom-color: #808080;
+	border-width: 11px;
+	margin: 0px -11px;
+}
+
+#powerTip.s:after, #powerTip.s:before {
+	left: 50%;
+}
+
+#powerTip.sw:after, #powerTip.sw:before {
+	right: 14px;
+}
+
+#powerTip.se:after, #powerTip.se:before {
+	left: 14px;
+}
+
+#powerTip.e:after, #powerTip.e:before {
+	left: 100%;
+}
+#powerTip.e:after {
+	border-left-color: #FFFFFF;
+	border-width: 10px;
+	top: 50%;
+	margin-top: -10px;
+}
+#powerTip.e:before {
+	border-left-color: #808080;
+	border-width: 11px;
+	top: 50%;
+	margin-top: -11px;
+}
+
+#powerTip.w:after, #powerTip.w:before {
+	right: 100%;
+}
+#powerTip.w:after {
+	border-right-color: #FFFFFF;
+	border-width: 10px;
+	top: 50%;
+	margin-top: -10px;
+}
+#powerTip.w:before {
+	border-right-color: #808080;
+	border-width: 11px;
+	top: 50%;
+	margin-top: -11px;
+}
+
+@media print
+{
+  #top { display: none; }
+  #side-nav { display: none; }
+  #nav-path { display: none; }
+  body { overflow:visible; }
+  h1, h2, h3, h4, h5, h6 { page-break-after: avoid; }
+  .summary { display: none; }
+  .memitem { page-break-inside: avoid; }
+  #doc-content
+  {
+    margin-left:0 !important;
+    height:auto !important;
+    width:auto !important;
+    overflow:inherit;
+    display:inline;
+  }
+}
+
+/* @group Markdown */
+
+table.markdownTable {
+	border-collapse:collapse;
+        margin-top: 4px;
+        margin-bottom: 4px;
+}
+
+table.markdownTable td, table.markdownTable th {
+	border: 1px solid #2D4068;
+	padding: 3px 7px 2px;
+}
+
+table.markdownTable tr {
+}
+
+th.markdownTableHeadLeft, th.markdownTableHeadRight, th.markdownTableHeadCenter, th.markdownTableHeadNone {
+	background-color: #374F7F;
+	color: #FFFFFF;
+	font-size: 110%;
+	padding-bottom: 4px;
+	padding-top: 5px;
+}
+
+th.markdownTableHeadLeft, td.markdownTableBodyLeft {
+	text-align: left
+}
+
+th.markdownTableHeadRight, td.markdownTableBodyRight {
+	text-align: right
+}
+
+th.markdownTableHeadCenter, td.markdownTableBodyCenter {
+	text-align: center
+}
+
+.DocNodeRTL {
+  text-align: right;
+  direction: rtl;
+}
+
+.DocNodeLTR {
+  text-align: left;
+  direction: ltr;
+}
+
+table.DocNodeRTL {
+   width: auto;
+   margin-right: 0;
+   margin-left: auto;
+}
+
+table.DocNodeLTR {
+   width: auto;
+   margin-right: auto;
+   margin-left: 0;
+}
+
+tt, code, kbd, samp
+{
+  display: inline-block;
+  direction:ltr; 
+}
+/* @end */
+
+u {
+	text-decoration: underline;
+}
+

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/doxygen.svg Added

@@ -0,0 +1,26 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg version="1.1" viewBox="0 0 104 31" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+ <defs>
+  <linearGradient id="a">
+   <stop stop-color="#5373B4" offset="0"/>
+   <stop stop-color="#7C95C6" offset="1"/>
+  </linearGradient>
+  <linearGradient id="d" x1="31.474" x2="31.474" y1="24.821" y2="26.773" gradientUnits="userSpaceOnUse" xlink:href="#a"/>
+  <linearGradient id="c" x1="31.474" x2="31.474" y1="24.821" y2="26.773" gradientTransform="matrix(.6816 0 0 1.0248 72.391 -.91809)" gradientUnits="userSpaceOnUse" xlink:href="#a"/>
+  <linearGradient id="b" x1="56.295" x2="56.295" y1="24.622" y2="26.574" gradientUnits="userSpaceOnUse" xlink:href="#a"/>
+  <linearGradient id="e" x1="49.067" x2="48.956" y1="19.719" y2="9.5227" gradientTransform="matrix(.97968 0 0 1.0207 -.25579 -.25579)" gradientUnits="userSpaceOnUse">
+   <stop stop-color="#C0CCE3" offset="0"/>
+   <stop stop-color="#EEF1F7" offset="1"/>
+  </linearGradient>
+  <filter id="f" x="-.010676" y="-.045304" width="1.0214" height="1.0906" color-interpolation-filters="sRGB">
+   <feGaussianBlur stdDeviation="0.45293203"/>
+  </filter>
+ </defs>
+ <g>
+  <path transform="translate(-2.5759 -27.848)" d="m13.609 32.203v6.8633h-0.05078c-0.40533-0.66867-0.96254-1.1715-1.6719-1.5059-0.69244-0.35193-1.4282-0.52734-2.2051-0.52734-0.96267 0-1.807 0.2027-2.5332 0.60742-0.72622 0.38713-1.3344 0.90556-1.8242 1.5566-0.47289 0.65108-0.83456 1.4092-1.0879 2.2715-0.23644 0.84464-0.35547 1.7236-0.35547 2.6387 0 0.95022 0.11902 1.8643 0.35547 2.7441 0.25333 0.87983 0.615 1.6633 1.0879 2.3496 0.48978 0.66867 1.1065 1.2066 1.8496 1.6113 0.74311 0.38713 1.6044 0.58008 2.584 0.58008 0.86133 0 1.6311-0.15787 2.3066-0.47461 0.69244-0.33434 1.2497-0.87227 1.6719-1.6113h0.05078v1.7422h3.4199v-18.846zm12.875 4.8301c-1.0302 0-1.9596 0.17541-2.7871 0.52734-0.82756 0.33434-1.5358 0.81965-2.127 1.4531-0.59111 0.61588-1.0483 1.3721-1.3691 2.2695-0.32089 0.87983-0.48047 1.866-0.48047 2.957s0.15958 2.0752 0.48047 2.9551c0.32089 0.87983 0.77803 1.6361 1.3691 2.2695 0.59111 0.61588 1.2994 1.0914 2.127 1.4258 0.82756 0.33434 1.7569 0.50195 2.7871 0.50195 1.0302 0 1.9596-0.16762 2.7871-0.50195 0.84444-0.33434 1.5612-0.8099 2.1523-1.4258 0.59111-0.63348 1.0483-1.3897 1.3691-2.2695 0.32089-0.87983 0.48047-1.8641 0.48047-2.9551s-0.15958-2.0772-0.48047-2.957c-0.32089-0.89743-0.77803-1.6536-1.3691-2.2695-0.59111-0.63348-1.3079-1.1188-2.1523-1.4531-0.82756-0.35193-1.7569-0.52734-2.7871-0.52734zm41.715 0c-0.912 0-1.7223 0.18516-2.4316 0.55469-0.69244 0.36953-1.2752 0.87043-1.748 1.5039-0.47289 0.61588-0.83651 1.337-1.0898 2.1641-0.23645 0.80944-0.35352 1.6553-0.35352 2.5352 0 0.93262 0.10007 1.8214 0.30273 2.666 0.21956 0.82704 0.55767 1.556 1.0137 2.1895 0.456 0.61588 1.0387 1.109 1.748 1.4785 0.70933 0.35193 1.5536 0.5293 2.5332 0.5293 0.79378 0 1.5446-0.16762 2.2539-0.50195 0.72622-0.35193 1.2834-0.88986 1.6719-1.6113h0.05078v1.7949c0.01689 0.96782-0.21071 1.7689-0.68359 2.4023-0.456 0.63348-1.1898 0.95117-2.2031 0.95117-0.64178 0-1.2075-0.14228-1.6973-0.42383-0.48978-0.26395-0.81939-0.74731-0.98828-1.4512h-3.5723c0.05067 0.77425 0.25276 1.435 0.60742 1.9805 0.37156 0.56309 0.8287 1.0192 1.3691 1.3711 0.55733 0.35193 1.1656 0.60726 1.8242 0.76562 0.67556 0.17597 1.3328 0.26562 1.9746 0.26562 1.5031 0 2.7025-0.21245 3.5977-0.63477 0.89511-0.42232 1.5798-0.94076 2.0527-1.5566 0.47289-0.59829 0.777-1.2493 0.91211-1.9531 0.152-0.70386 0.22656-1.3295 0.22656-1.875v-12.775h-3.4199v1.8223h-0.05078c-0.43911-0.79185-0.98782-1.3551-1.6465-1.6895-0.64178-0.33434-1.3926-0.50195-2.2539-0.50195zm16.523 0c-0.99644 0-1.9088 0.18516-2.7363 0.55469-0.81067 0.36953-1.5124 0.88018-2.1035 1.5312-0.59111 0.63348-1.0463 1.3897-1.3672 2.2695s-0.48047 1.831-0.48047 2.8516c0 1.0558 0.15108 2.0225 0.45508 2.9023 0.32089 0.87983 0.76758 1.6361 1.3418 2.2695 0.57422 0.63348 1.276 1.1266 2.1035 1.4785 0.82756 0.33434 1.7569 0.50195 2.7871 0.50195 1.4862 0 2.7517-0.35277 3.7988-1.0566 1.0471-0.70387 1.8254-1.8733 2.332-3.5098h-3.168c-0.11822 0.42232-0.43934 0.82772-0.96289 1.2148-0.52355 0.36953-1.1468 0.55274-1.873 0.55273-1.0133 0-1.7916-0.27286-2.332-0.81836-0.54044-0.5455-0.83605-1.4245-0.88672-2.6387h9.4492c0.06756-1.0558-0.01551-2.0673-0.25195-3.0352-0.23644-0.96782-0.62557-1.8293-1.166-2.5859-0.52356-0.75666-1.1998-1.355-2.0273-1.7949-0.82756-0.45751-1.7974-0.6875-2.9121-0.6875zm16.189 0c-0.76 0-1.5023 0.18516-2.2285 0.55469-0.72622 0.35193-1.3174 0.92299-1.7734 1.7148h-0.07617v-1.9004h-3.4199v13.646h3.5977v-7.1523c0-1.3901 0.21909-2.3841 0.6582-2.9824 0.43911-0.61588 1.1494-0.92383 2.1289-0.92383 0.86133 0 1.4611 0.28066 1.7988 0.84375 0.33777 0.5455 0.50586 1.3816 0.50586 2.5078v7.707h3.5976v-8.3926c0-0.84464-0.0765-1.6106-0.22851-2.2969-0.13511-0.70387-0.37971-1.2925-0.73438-1.7676-0.35466-0.49271-0.84386-0.87277-1.4688-1.1367-0.608-0.28155-1.3948-0.42188-2.3574-0.42188zm-66.063 0.36914 4.3066 6.4668-4.7129 7.1797h4.0293l2.7363-4.3027 2.7344 4.3027h4.1055l-4.8398-7.2578 4.3066-6.3887h-3.9766l-2.2793 3.5645-2.3066-3.5645zm13.275 0 4.584 12.803c0.10133 0.26395 0.15234 0.54461 0.15234 0.84375 0 0.40472-0.11707 0.77504-0.35352 1.1094-0.21956 0.33434-0.56617 0.52729-1.0391 0.58008-0.35467 0.0176-0.70979 0.0098-1.0645-0.02539-0.35467-0.03519-0.70128-0.07028-1.0391-0.10547v3.0879c0.37156 0.03519 0.73518 0.06051 1.0898 0.07813 0.37156 0.03519 0.74368 0.05273 1.1152 0.05273 1.2329 0 2.1943-0.23778 2.8867-0.71289 0.69244-0.47511 1.2326-1.2664 1.6211-2.375l5.4727-15.336h-3.7246l-2.8613 9.3438h-0.05078l-2.9648-9.3438zm-37.48 2.4551c0.59111 0 1.0823 0.12279 1.4707 0.36914 0.38844 0.24635 0.6991 0.57184 0.93555 0.97656 0.25333 0.38713 0.43187 0.84515 0.5332 1.373 0.10133 0.5103 0.15234 1.0482 0.15234 1.6113 0 0.56309-0.05101 1.1069-0.15234 1.6348-0.10133 0.5279-0.27137 1.0035-0.50781 1.4258-0.23644 0.40472-0.5556 0.73021-0.96094 0.97656-0.38844 0.24635-0.87959 0.36914-1.4707 0.36914-0.55733 0-1.038-0.12279-1.4434-0.36914-0.38844-0.26395-0.71806-0.59723-0.98828-1.002-0.25333-0.42232-0.43842-0.89788-0.55664-1.4258s-0.17773-1.0561-0.17773-1.584c-1e-7 -0.56309 0.05101-1.0991 0.15234-1.6094 0.11822-0.5279 0.29481-0.99567 0.53125-1.4004 0.25333-0.40472 0.58295-0.73021 0.98828-0.97656 0.40533-0.24635 0.90303-0.36914 1.4941-0.36914zm15.84 0c0.608 0 1.1142 0.13253 1.5195 0.39648 0.42222 0.24635 0.75184 0.57184 0.98828 0.97656 0.25333 0.40472 0.42992 0.87054 0.53125 1.3984 0.10133 0.5279 0.15234 1.0658 0.15234 1.6113 0 0.5455-0.05101 1.0815-0.15234 1.6094-0.10134 0.5103-0.27792 0.97612-0.53125 1.3984-0.23644 0.40472-0.56606 0.73021-0.98828 0.97656-0.40533 0.24635-0.91153 0.36914-1.5195 0.36914-0.608 0-1.1142-0.12279-1.5195-0.36914s-0.73495-0.57184-0.98828-0.97656c-0.23644-0.42232-0.40648-0.88814-0.50781-1.3984-0.10133-0.5279-0.15234-1.0639-0.15234-1.6094 0-0.5455 0.05101-1.0834 0.15234-1.6113 0.10133-0.5279 0.27137-0.99371 0.50781-1.3984 0.25333-0.40472 0.58295-0.73021 0.98828-0.97656 0.40533-0.26395 0.91153-0.39648 1.5195-0.39648zm42.602 0c0.59111 0 1.0803 0.11499 1.4688 0.34375 0.38844 0.22876 0.70105 0.5367 0.9375 0.92383 0.23644 0.38713 0.40648 0.8354 0.50781 1.3457 0.10133 0.49271 0.15039 1.0209 0.15039 1.584 0 0.4927-0.06606 0.96827-0.20117 1.4258-0.11822 0.43992-0.30526 0.83557-0.55859 1.1875-0.25333 0.35193-0.57445 0.63259-0.96289 0.84375-0.38844 0.21116-0.83513 0.31836-1.3418 0.31836-0.55733 0-1.021-0.12474-1.3926-0.37109-0.37156-0.24635-0.67566-0.56209-0.91211-0.94922-0.21956-0.38713-0.38109-0.81786-0.48242-1.293-0.08444-0.49271-0.12695-0.98581-0.12695-1.4785 0-0.5103 0.05101-0.99366 0.15234-1.4512 0.11822-0.47511 0.29676-0.89025 0.5332-1.2422 0.25333-0.36953 0.55744-0.65993 0.91211-0.87109 0.37156-0.21116 0.80974-0.31641 1.3164-0.31641zm15.535 0c0.87822 0 1.529 0.24753 1.9512 0.74023 0.43911 0.49271 0.74322 1.2138 0.91211 2.1641h-5.8535c0.01689-0.26395 0.0679-0.5641 0.15234-0.89844 0.10133-0.33434 0.26287-0.65008 0.48242-0.94922 0.23644-0.29914 0.54055-0.54667 0.91211-0.74023 0.38845-0.21116 0.86914-0.31641 1.4434-0.31641z" filter="url(#f)" opacity=".3" stroke="#969696" xlink:href="#path141"/>
+  <path d="m0.97202 24.161 43.605-0.0019 0.0508 3.3061-43.6 0.04174z" fill="url(#d)" stroke="#000" stroke-width=".5"/>
+  <path d="m10.283 3.5547v6.8633h-0.05078c-0.40533-0.66867-0.96254-1.1715-1.6719-1.5059-0.69244-0.35193-1.4282-0.52734-2.2051-0.52734-0.96267 0-1.807 0.2027-2.5332 0.60742-0.72622 0.38713-1.3344 0.90556-1.8242 1.5566-0.47289 0.65108-0.83456 1.4092-1.0879 2.2715-0.23644 0.84464-0.35547 1.7236-0.35547 2.6387 0 0.95022 0.11902 1.8643 0.35547 2.7441 0.25333 0.87983 0.615 1.6633 1.0879 2.3496 0.48978 0.66867 1.1065 1.2066 1.8496 1.6113 0.74311 0.38713 1.6044 0.58008 2.584 0.58008 0.86133 0 1.6311-0.15787 2.3066-0.47461 0.69244-0.33434 1.2497-0.87227 1.6719-1.6113h0.05078v1.7422h3.4199v-18.846zm12.875 4.8301c-1.0302 0-1.9596 0.17541-2.7871 0.52734-0.82756 0.33434-1.5358 0.81965-2.127 1.4531-0.59111 0.61588-1.0483 1.3721-1.3691 2.2695-0.32089 0.87983-0.48047 1.866-0.48047 2.957s0.15958 2.0752 0.48047 2.9551c0.32089 0.87983 0.77803 1.6361 1.3691 2.2695 0.59111 0.61588 1.2994 1.0914 2.127 1.4258 0.82756 0.33434 1.7569 0.50195 2.7871 0.50195 1.0302 0 1.9596-0.16762 2.7871-0.50195 0.84444-0.33434 1.5612-0.8099 2.1523-1.4258 0.59111-0.63348 1.0483-1.3897 1.3691-2.2695 0.32089-0.87983 0.48047-1.8641 0.48047-2.9551s-0.15958-2.0772-0.48047-2.957c-0.32089-0.89743-0.77803-1.6536-1.3691-2.2695-0.59111-0.63348-1.3079-1.1188-2.1523-1.4531-0.82756-0.35193-1.7569-0.52734-2.7871-0.52734zm41.715 0c-0.912 0-1.7223 0.18516-2.4316 0.55469-0.69244 0.36953-1.2752 0.87043-1.748 1.5039-0.47289 0.61588-0.83651 1.337-1.0898 2.1641-0.23644 0.80944-0.35352 1.6553-0.35352 2.5352 0 0.93262 0.10007 1.8214 0.30273 2.666 0.21956 0.82704 0.55767 1.556 1.0137 2.1895 0.456 0.61588 1.0387 1.109 1.748 1.4785 0.70933 0.35193 1.5536 0.5293 2.5332 0.5293 0.79378 0 1.5446-0.16762 2.2539-0.50195 0.72622-0.35193 1.2834-0.88986 1.6719-1.6113h0.05078v1.7949c0.01689 0.96782-0.21071 1.7689-0.68359 2.4023-0.456 0.63348-1.1898 0.95117-2.2031 0.95117-0.64178 0-1.2075-0.14228-1.6973-0.42383-0.48978-0.26395-0.81939-0.74731-0.98828-1.4512h-3.5723c0.05067 0.77425 0.25276 1.435 0.60742 1.9805 0.37156 0.56309 0.8287 1.0192 1.3691 1.3711 0.55733 0.35193 1.1656 0.60726 1.8242 0.76562 0.67556 0.17597 1.3328 0.26562 1.9746 0.26562 1.5031 0 2.7025-0.21245 3.5977-0.63477 0.89511-0.42232 1.5798-0.94076 2.0527-1.5566 0.47289-0.59829 0.777-1.2493 0.91211-1.9531 0.152-0.70386 0.22656-1.3295 0.22656-1.875v-12.775h-3.4199v1.8223h-0.05078c-0.43911-0.79185-0.98782-1.3551-1.6465-1.6895-0.64178-0.33434-1.3926-0.50195-2.2539-0.50195zm16.523 0c-0.99644 0-1.9088 0.18516-2.7363 0.55469-0.81067 0.36953-1.5124 0.88017-2.1035 1.5312-0.59111 0.63348-1.0463 1.3897-1.3672 2.2695s-0.48047 1.831-0.48047 2.8516c0 1.0558 0.15108 2.0225 0.45508 2.9023 0.32089 0.87983 0.76758 1.6361 1.3418 2.2695 0.57422 0.63348 1.276 1.1266 2.1035 1.4785 0.82756 0.33434 1.7569 0.50195 2.7871 0.50195 1.4862 0 2.7517-0.35278 3.7988-1.0566 1.0471-0.70386 1.8254-1.8733 2.332-3.5098h-3.168c-0.11822 0.42232-0.43934 0.82772-0.96289 1.2148-0.52355 0.36953-1.1468 0.55274-1.873 0.55273-1.0133 0-1.7916-0.27286-2.332-0.81836-0.54044-0.5455-0.83605-1.4245-0.88672-2.6387h9.4492c0.06756-1.0558-0.01551-2.0673-0.25195-3.0352-0.23644-0.96782-0.62557-1.8293-1.166-2.5859-0.52356-0.75666-1.1998-1.355-2.0273-1.7949-0.82756-0.45751-1.7974-0.6875-2.9121-0.6875zm16.189 0c-0.76 0-1.5023 0.18516-2.2285 0.55469-0.72622 0.35193-1.3174 0.923-1.7734 1.7148h-0.07617v-1.9004h-3.4199v13.646h3.5977v-7.1523c0-1.3901 0.21909-2.3841 0.6582-2.9824 0.43911-0.61588 1.1494-0.92383 2.1289-0.92383 0.86133 0 1.461 0.28066 1.7988 0.84375 0.33778 0.5455 0.50586 1.3816 0.50586 2.5078v7.707h3.5977v-8.3926c0-0.84464-0.0765-1.6106-0.22852-2.2969-0.13511-0.70387-0.3797-1.2925-0.73437-1.7676-0.35466-0.49271-0.84386-0.87277-1.4688-1.1367-0.608-0.28155-1.3948-0.42188-2.3574-0.42188zm-66.062 0.36914 4.3066 6.4668-4.7129 7.1797h4.0293l2.7363-4.3027 2.7344 4.3027h4.1055l-4.8398-7.2578 4.3066-6.3887h-3.9766l-2.2793 3.5645-2.3066-3.5645zm13.275 0 4.584 12.803c0.10133 0.26395 0.15234 0.54461 0.15234 0.84375 0 0.40472-0.11707 0.77504-0.35352 1.1094-0.21956 0.33434-0.56617 0.52729-1.0391 0.58008-0.35467 0.0176-0.70979 0.0098-1.0645-0.02539-0.35467-0.03519-0.70128-0.07027-1.0391-0.10547v3.0879c0.37156 0.03519 0.73518 0.06052 1.0898 0.07813 0.37156 0.03519 0.74368 0.05273 1.1152 0.05273 1.2329 0 2.1943-0.23778 2.8867-0.71289 0.69244-0.47511 1.2326-1.2664 1.6211-2.375l5.4727-15.336h-3.7246l-2.8613 9.3437h-0.05078l-2.9648-9.3437zm-37.48 2.4551c0.59111 0 1.0823 0.12279 1.4707 0.36914s0.6991 0.57184 0.93555 0.97656c0.25333 0.38713 0.43187 0.84515 0.5332 1.373 0.10133 0.5103 0.15234 1.0482 0.15234 1.6113 0 0.56309-0.05101 1.1069-0.15234 1.6348-0.10133 0.5279-0.27137 1.0035-0.50781 1.4258-0.23644 0.40472-0.5556 0.73021-0.96094 0.97656-0.38844 0.24635-0.87959 0.36914-1.4707 0.36914-0.55733 0-1.038-0.12279-1.4434-0.36914-0.38844-0.26395-0.71806-0.59723-0.98828-1.002-0.25333-0.42232-0.43842-0.89788-0.55664-1.4258s-0.17773-1.0561-0.17773-1.584c-1e-7 -0.56309 0.05101-1.0991 0.15234-1.6094 0.11822-0.5279 0.29481-0.99567 0.53125-1.4004 0.25333-0.40472 0.58295-0.73021 0.98828-0.97656 0.40533-0.24635 0.90303-0.36914 1.4941-0.36914zm15.84 0c0.608 0 1.1142 0.13254 1.5195 0.39648 0.42222 0.24635 0.75184 0.57184 0.98828 0.97656 0.25333 0.40472 0.42992 0.87054 0.53125 1.3984 0.10133 0.5279 0.15234 1.0658 0.15234 1.6113 0 0.5455-0.05101 1.0815-0.15234 1.6094-0.10133 0.5103-0.27792 0.97612-0.53125 1.3984-0.23644 0.40472-0.56606 0.73021-0.98828 0.97656-0.40533 0.24635-0.91153 0.36914-1.5195 0.36914-0.608 0-1.1142-0.12279-1.5195-0.36914s-0.73495-0.57184-0.98828-0.97656c-0.23644-0.42232-0.40648-0.88813-0.50781-1.3984-0.10133-0.5279-0.15234-1.0639-0.15234-1.6094 0-0.5455 0.05101-1.0834 0.15234-1.6113 0.10133-0.5279 0.27137-0.99371 0.50781-1.3984 0.25333-0.40472 0.58295-0.73021 0.98828-0.97656 0.40533-0.26395 0.91153-0.39648 1.5195-0.39648zm42.602 0c0.59111 0 1.0803 0.11499 1.4688 0.34375 0.38844 0.22876 0.70106 0.5367 0.9375 0.92383 0.23644 0.38713 0.40648 0.8354 0.50781 1.3457 0.10133 0.49271 0.15039 1.0209 0.15039 1.584 0 0.49271-0.06606 0.96827-0.20117 1.4258-0.11822 0.43992-0.30526 0.83557-0.55859 1.1875-0.25333 0.35193-0.57445 0.63259-0.96289 0.84375-0.38844 0.21116-0.83513 0.31836-1.3418 0.31836-0.55733 0-1.021-0.12474-1.3926-0.37109-0.37156-0.24635-0.67566-0.56209-0.91211-0.94922-0.21956-0.38713-0.38109-0.81786-0.48242-1.293-0.08444-0.49271-0.12695-0.98581-0.12695-1.4785 0-0.5103 0.05101-0.99366 0.15234-1.4512 0.11822-0.47511 0.29676-0.89026 0.5332-1.2422 0.25333-0.36953 0.55744-0.65993 0.91211-0.87109 0.37156-0.21116 0.80974-0.31641 1.3164-0.31641zm15.535 0c0.87822 0 1.529 0.24753 1.9512 0.74024 0.43911 0.49271 0.74322 1.2138 0.91211 2.1641h-5.8535c0.01689-0.26395 0.0679-0.5641 0.15234-0.89844 0.10133-0.33434 0.26287-0.65008 0.48242-0.94922 0.23644-0.29914 0.54055-0.54667 0.91211-0.74023 0.38845-0.21116 0.86914-0.31641 1.4434-0.31641z" fill="url(#e)" stroke="#4665A2" stroke-width=".7"/>
+  <path d="m52.988 27.291c0.99602-1.0359 1.3944-1.8725 1.7928-3.1076l3.8247-0.03984c0.3113 1.6096 0.82413 2.5137 1.6335 3.1474z" fill="url(#b)" stroke="#000" stroke-width=".5"/>
+  <path d="m73.89 24.04 28.885-0.2011-0.12476 3.3879-31.033 0.16229c1.2621-1.0234 1.9665-2.2859 2.2724-3.3491z" fill="url(#c)" stroke="#000" stroke-width=".41788"/>
+ </g>
+</svg>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/dynsections.js Added

@@ -0,0 +1,121 @@
+/*
+ @licstart  The following is the entire license notice for the JavaScript code in this file.
+
+ The MIT License (MIT)
+
+ Copyright (C) 1997-2020 by Dimitri van Heesch
+
+ Permission is hereby granted, free of charge, to any person obtaining a copy of this software
+ and associated documentation files (the "Software"), to deal in the Software without restriction,
+ including without limitation the rights to use, copy, modify, merge, publish, distribute,
+ sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
+ furnished to do so, subject to the following conditions:
+
+ The above copyright notice and this permission notice shall be included in all copies or
+ substantial portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING
+ BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+ DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ @licend  The above is the entire license notice for the JavaScript code in this file
+ */
+function toggleVisibility(linkObj)
+{
+ var base = $(linkObj).attr('id');
+ var summary = $('#'+base+'-summary');
+ var content = $('#'+base+'-content');
+ var trigger = $('#'+base+'-trigger');
+ var src=$(trigger).attr('src');
+ if (content.is(':visible')===true) {
+   content.hide();
+   summary.show();
+   $(linkObj).addClass('closed').removeClass('opened');
+   $(trigger).attr('src',src.substring(0,src.length-8)+'closed.png');
+ } else {
+   content.show();
+   summary.hide();
+   $(linkObj).removeClass('closed').addClass('opened');
+   $(trigger).attr('src',src.substring(0,src.length-10)+'open.png');
+ }
+ return false;
+}
+
+function updateStripes()
+{
+  $('table.directory tr').
+       removeClass('even').filter(':visible:even').addClass('even');
+}
+
+function toggleLevel(level)
+{
+  $('table.directory tr').each(function() {
+    var l = this.id.split('_').length-1;
+    var i = $('#img'+this.id.substring(3));
+    var a = $('#arr'+this.id.substring(3));
+    if (l<level+1) {
+      i.removeClass('iconfopen iconfclosed').addClass('iconfopen');
+      a.html('&#9660;');
+      $(this).show();
+    } else if (l==level+1) {
+      i.removeClass('iconfclosed iconfopen').addClass('iconfclosed');
+      a.html('&#9658;');
+      $(this).show();
+    } else {
+      $(this).hide();
+    }
+  });
+  updateStripes();
+}
+
+function toggleFolder(id)
+{
+  // the clicked row
+  var currentRow = $('#row_'+id);
+
+  // all rows after the clicked row
+  var rows = currentRow.nextAll("tr");
+
+  var re = new RegExp('^row_'+id+'\\d+_$', "i"); //only one sub
+
+  // only match elements AFTER this one (can't hide elements before)
+  var childRows = rows.filter(function() { return this.id.match(re); });
+
+  // first row is visible we are HIDING
+  if (childRows.filter(':first').is(':visible')===true) {
+    // replace down arrow by right arrow for current row
+    var currentRowSpans = currentRow.find("span");
+    currentRowSpans.filter(".iconfopen").removeClass("iconfopen").addClass("iconfclosed");
+    currentRowSpans.filter(".arrow").html('&#9658;');
+    rows.filter("id^=row_"+id+"").hide(); // hide all children
+  } else { // we are SHOWING
+    // replace right arrow by down arrow for current row
+    var currentRowSpans = currentRow.find("span");
+    currentRowSpans.filter(".iconfclosed").removeClass("iconfclosed").addClass("iconfopen");
+    currentRowSpans.filter(".arrow").html('&#9660;');
+    // replace down arrows by right arrows for child rows
+    var childRowsSpans = childRows.find("span");
+    childRowsSpans.filter(".iconfopen").removeClass("iconfopen").addClass("iconfclosed");
+    childRowsSpans.filter(".arrow").html('&#9658;');
+    childRows.show(); //show all children
+  }
+  updateStripes();
+}
+
+
+function toggleInherit(id)
+{
+  var rows = $('tr.inherit.'+id);
+  var img = $('tr.inherit_header.'+id+' img');
+  var src = $(img).attr('src');
+  if (rows.filter(':first').is(':visible')===true) {
+    rows.css('display','none');
+    $(img).attr('src',src.substring(0,src.length-8)+'closed.png');
+  } else {
+    rows.css('display','table-row'); // using show() causes jump in firefox
+    $(img).attr('src',src.substring(0,src.length-10)+'open.png');
+  }
+}
+/* @license-end */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/folderclosed.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/folderopen.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/functions.html Added

@@ -0,0 +1,107 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: Data Fields</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+</div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+<div class="contents">
+<div class="textblock">Here is a list of all documented struct and union fields with links to the struct/union documentation for each field:</div><ul>
+<li>customFilter
+: <a class="el" href="structtjtransform.html#afd7fc262df33f741e120ef4183202ef5">tjtransform</a>
+</li>
+<li>data
+: <a class="el" href="structtjtransform.html#a688fe8f1a8ecc12a538d9e561cf338e3">tjtransform</a>
+</li>
+<li>denom
+: <a class="el" href="structtjscalingfactor.html#aefbcdf3e9e62274b2d312c695f133ce3">tjscalingfactor</a>
+</li>
+<li>h
+: <a class="el" href="structtjregion.html#aecefc45a26f4d8b60dd4d825c1710115">tjregion</a>
+</li>
+<li>num
+: <a class="el" href="structtjscalingfactor.html#a9b011e57f981ee23083e2c1aa5e640ec">tjscalingfactor</a>
+</li>
+<li>op
+: <a class="el" href="structtjtransform.html#a2525aab4ba6978a1c273f74fef50e498">tjtransform</a>
+</li>
+<li>options
+: <a class="el" href="structtjtransform.html#ac0e74655baa4402209a21e1ae481c8f6">tjtransform</a>
+</li>
+<li>r
+: <a class="el" href="structtjtransform.html#ac324e5e442abec8a961e5bf219db12cf">tjtransform</a>
+</li>
+<li>w
+: <a class="el" href="structtjregion.html#ab6eb73ceef584fc23c8c8097926dce42">tjregion</a>
+</li>
+<li>x
+: <a class="el" href="structtjregion.html#a4b6a37a93997091b26a75831fa291ad9">tjregion</a>
+</li>
+<li>y
+: <a class="el" href="structtjregion.html#a7b3e0c24cfe87acc80e334cafdcf22c2">tjregion</a>
+</li>
+</ul>
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/functions_vars.html Added

@@ -0,0 +1,107 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: Data Fields - Variables</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+</div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+<div class="contents">
+&#160;<ul>
+<li>customFilter
+: <a class="el" href="structtjtransform.html#afd7fc262df33f741e120ef4183202ef5">tjtransform</a>
+</li>
+<li>data
+: <a class="el" href="structtjtransform.html#a688fe8f1a8ecc12a538d9e561cf338e3">tjtransform</a>
+</li>
+<li>denom
+: <a class="el" href="structtjscalingfactor.html#aefbcdf3e9e62274b2d312c695f133ce3">tjscalingfactor</a>
+</li>
+<li>h
+: <a class="el" href="structtjregion.html#aecefc45a26f4d8b60dd4d825c1710115">tjregion</a>
+</li>
+<li>num
+: <a class="el" href="structtjscalingfactor.html#a9b011e57f981ee23083e2c1aa5e640ec">tjscalingfactor</a>
+</li>
+<li>op
+: <a class="el" href="structtjtransform.html#a2525aab4ba6978a1c273f74fef50e498">tjtransform</a>
+</li>
+<li>options
+: <a class="el" href="structtjtransform.html#ac0e74655baa4402209a21e1ae481c8f6">tjtransform</a>
+</li>
+<li>r
+: <a class="el" href="structtjtransform.html#ac324e5e442abec8a961e5bf219db12cf">tjtransform</a>
+</li>
+<li>w
+: <a class="el" href="structtjregion.html#ab6eb73ceef584fc23c8c8097926dce42">tjregion</a>
+</li>
+<li>x
+: <a class="el" href="structtjregion.html#a4b6a37a93997091b26a75831fa291ad9">tjregion</a>
+</li>
+<li>y
+: <a class="el" href="structtjregion.html#a7b3e0c24cfe87acc80e334cafdcf22c2">tjregion</a>
+</li>
+</ul>
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/group___turbo_j_p_e_g.html Added

@@ -0,0 +1,2872 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: TurboJPEG</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+</div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+<div class="header">
+  <div class="summary">
+<a href="#nested-classes">Data Structures</a> &#124;
+<a href="#define-members">Macros</a> &#124;
+<a href="#typedef-members">Typedefs</a> &#124;
+<a href="#enum-members">Enumerations</a> &#124;
+<a href="#func-members">Functions</a> &#124;
+<a href="#var-members">Variables</a>  </div>
+  <div class="headertitle">
+<div class="title">TurboJPEG</div>  </div>
+</div>
+<div class="contents">
+
+<p>TurboJPEG API.  
+<a href="#details">More...</a></p>
+<table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="nested-classes"></a>
+Data Structures</h2></td></tr>
+<tr class="memitem:"><td class="memItemLeft" align="right" valign="top">struct &#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjscalingfactor.html">tjscalingfactor</a></td></tr>
+<tr class="memdesc:"><td class="mdescLeft">&#160;</td><td class="mdescRight">Scaling factor.  <a href="structtjscalingfactor.html#details">More...</a><br /></td></tr>
+<tr class="separator:"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:"><td class="memItemLeft" align="right" valign="top">struct &#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjregion.html">tjregion</a></td></tr>
+<tr class="memdesc:"><td class="mdescLeft">&#160;</td><td class="mdescRight">Cropping region.  <a href="structtjregion.html#details">More...</a><br /></td></tr>
+<tr class="separator:"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:"><td class="memItemLeft" align="right" valign="top">struct &#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjtransform.html">tjtransform</a></td></tr>
+<tr class="memdesc:"><td class="mdescLeft">&#160;</td><td class="mdescRight">Lossless transform.  <a href="structtjtransform.html#details">More...</a><br /></td></tr>
+<tr class="separator:"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table><table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="define-members"></a>
+Macros</h2></td></tr>
+<tr class="memitem:ga5ef3d169162ce77ce348e292a0b7477c"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga5ef3d169162ce77ce348e292a0b7477c">TJ_NUMSAMP</a></td></tr>
+<tr class="memdesc:ga5ef3d169162ce77ce348e292a0b7477c"><td class="mdescLeft">&#160;</td><td class="mdescRight">The number of chrominance subsampling options.  <a href="group___turbo_j_p_e_g.html#ga5ef3d169162ce77ce348e292a0b7477c">More...</a><br /></td></tr>
+<tr class="separator:ga5ef3d169162ce77ce348e292a0b7477c"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga7010a4402f54a45ba822ad8675a4655e"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td></tr>
+<tr class="memdesc:ga7010a4402f54a45ba822ad8675a4655e"><td class="mdescLeft">&#160;</td><td class="mdescRight">The number of pixel formats.  <a href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">More...</a><br /></td></tr>
+<tr class="separator:ga7010a4402f54a45ba822ad8675a4655e"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga39f57a6fb02d9cf32e7b6890099b5a71"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga39f57a6fb02d9cf32e7b6890099b5a71">TJ_NUMCS</a></td></tr>
+<tr class="memdesc:ga39f57a6fb02d9cf32e7b6890099b5a71"><td class="mdescLeft">&#160;</td><td class="mdescRight">The number of JPEG colorspaces.  <a href="group___turbo_j_p_e_g.html#ga39f57a6fb02d9cf32e7b6890099b5a71">More...</a><br /></td></tr>
+<tr class="separator:ga39f57a6fb02d9cf32e7b6890099b5a71"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga72ecf4ebe6eb702d3c6f5ca27455e1ec"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga72ecf4ebe6eb702d3c6f5ca27455e1ec">TJFLAG_BOTTOMUP</a></td></tr>
+<tr class="memdesc:ga72ecf4ebe6eb702d3c6f5ca27455e1ec"><td class="mdescLeft">&#160;</td><td class="mdescRight">The uncompressed source/destination image is stored in bottom-up (Windows, OpenGL) order, not top-down (X11) order.  <a href="group___turbo_j_p_e_g.html#ga72ecf4ebe6eb702d3c6f5ca27455e1ec">More...</a><br /></td></tr>
+<tr class="separator:ga72ecf4ebe6eb702d3c6f5ca27455e1ec"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga4ee4506c81177a06f77e2504a22efd2d"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga4ee4506c81177a06f77e2504a22efd2d">TJFLAG_FASTUPSAMPLE</a></td></tr>
+<tr class="memdesc:ga4ee4506c81177a06f77e2504a22efd2d"><td class="mdescLeft">&#160;</td><td class="mdescRight">When decompressing an image that was compressed using chrominance subsampling, use the fastest chrominance upsampling algorithm available in the underlying codec.  <a href="group___turbo_j_p_e_g.html#ga4ee4506c81177a06f77e2504a22efd2d">More...</a><br /></td></tr>
+<tr class="separator:ga4ee4506c81177a06f77e2504a22efd2d"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga8808d403c68b62aaa58a4c1e58e98963"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963">TJFLAG_NOREALLOC</a></td></tr>
+<tr class="memdesc:ga8808d403c68b62aaa58a4c1e58e98963"><td class="mdescLeft">&#160;</td><td class="mdescRight">Disable buffer (re)allocation.  <a href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963">More...</a><br /></td></tr>
+<tr class="separator:ga8808d403c68b62aaa58a4c1e58e98963"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gaabce235db80d3f698b27f36cbd453da2"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gaabce235db80d3f698b27f36cbd453da2">TJFLAG_FASTDCT</a></td></tr>
+<tr class="memdesc:gaabce235db80d3f698b27f36cbd453da2"><td class="mdescLeft">&#160;</td><td class="mdescRight">Use the fastest DCT/IDCT algorithm available in the underlying codec.  <a href="group___turbo_j_p_e_g.html#gaabce235db80d3f698b27f36cbd453da2">More...</a><br /></td></tr>
+<tr class="separator:gaabce235db80d3f698b27f36cbd453da2"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gacb233cfd722d66d1ccbf48a7de81f0e0"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">TJFLAG_ACCURATEDCT</a></td></tr>
+<tr class="memdesc:gacb233cfd722d66d1ccbf48a7de81f0e0"><td class="mdescLeft">&#160;</td><td class="mdescRight">Use the most accurate DCT/IDCT algorithm available in the underlying codec.  <a href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">More...</a><br /></td></tr>
+<tr class="separator:gacb233cfd722d66d1ccbf48a7de81f0e0"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga519cfa4ef6c18d9e5b455fdf59306a3a"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga519cfa4ef6c18d9e5b455fdf59306a3a">TJFLAG_STOPONWARNING</a></td></tr>
+<tr class="memdesc:ga519cfa4ef6c18d9e5b455fdf59306a3a"><td class="mdescLeft">&#160;</td><td class="mdescRight">Immediately discontinue the current compression/decompression/transform operation if the underlying codec throws a warning (non-fatal error).  <a href="group___turbo_j_p_e_g.html#ga519cfa4ef6c18d9e5b455fdf59306a3a">More...</a><br /></td></tr>
+<tr class="separator:ga519cfa4ef6c18d9e5b455fdf59306a3a"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga43b426750b46190a25d34a67ef76df1b"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga43b426750b46190a25d34a67ef76df1b">TJFLAG_PROGRESSIVE</a></td></tr>
+<tr class="memdesc:ga43b426750b46190a25d34a67ef76df1b"><td class="mdescLeft">&#160;</td><td class="mdescRight">Use progressive entropy coding in JPEG images generated by the compression and transform functions.  <a href="group___turbo_j_p_e_g.html#ga43b426750b46190a25d34a67ef76df1b">More...</a><br /></td></tr>
+<tr class="separator:ga43b426750b46190a25d34a67ef76df1b"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga163e6482dc5096831feef9c79ff3f805"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga163e6482dc5096831feef9c79ff3f805">TJFLAG_LIMITSCANS</a></td></tr>
+<tr class="memdesc:ga163e6482dc5096831feef9c79ff3f805"><td class="mdescLeft">&#160;</td><td class="mdescRight">Limit the number of progressive JPEG scans that the decompression and transform functions will process.  <a href="group___turbo_j_p_e_g.html#ga163e6482dc5096831feef9c79ff3f805">More...</a><br /></td></tr>
+<tr class="separator:ga163e6482dc5096831feef9c79ff3f805"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga79bde1b4a3e2351e00887e47781b966e"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga79bde1b4a3e2351e00887e47781b966e">TJ_NUMERR</a></td></tr>
+<tr class="memdesc:ga79bde1b4a3e2351e00887e47781b966e"><td class="mdescLeft">&#160;</td><td class="mdescRight">The number of error codes.  <a href="group___turbo_j_p_e_g.html#ga79bde1b4a3e2351e00887e47781b966e">More...</a><br /></td></tr>
+<tr class="separator:ga79bde1b4a3e2351e00887e47781b966e"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga0f6dbd18adf38b7d46ac547f0f4d562c"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga0f6dbd18adf38b7d46ac547f0f4d562c">TJ_NUMXOP</a></td></tr>
+<tr class="memdesc:ga0f6dbd18adf38b7d46ac547f0f4d562c"><td class="mdescLeft">&#160;</td><td class="mdescRight">The number of transform operations.  <a href="group___turbo_j_p_e_g.html#ga0f6dbd18adf38b7d46ac547f0f4d562c">More...</a><br /></td></tr>
+<tr class="separator:ga0f6dbd18adf38b7d46ac547f0f4d562c"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga50e03cb5ed115330e212417429600b00"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga50e03cb5ed115330e212417429600b00">TJXOPT_PERFECT</a></td></tr>
+<tr class="memdesc:ga50e03cb5ed115330e212417429600b00"><td class="mdescLeft">&#160;</td><td class="mdescRight">This option will cause <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> to return an error if the transform is not perfect.  <a href="group___turbo_j_p_e_g.html#ga50e03cb5ed115330e212417429600b00">More...</a><br /></td></tr>
+<tr class="separator:ga50e03cb5ed115330e212417429600b00"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga319826b7eb1583c0595bbe7b95428709"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga319826b7eb1583c0595bbe7b95428709">TJXOPT_TRIM</a></td></tr>
+<tr class="memdesc:ga319826b7eb1583c0595bbe7b95428709"><td class="mdescLeft">&#160;</td><td class="mdescRight">This option will cause <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> to discard any partial MCU blocks that cannot be transformed.  <a href="group___turbo_j_p_e_g.html#ga319826b7eb1583c0595bbe7b95428709">More...</a><br /></td></tr>
+<tr class="separator:ga319826b7eb1583c0595bbe7b95428709"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga9c771a757fc1294add611906b89ab2d2"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga9c771a757fc1294add611906b89ab2d2">TJXOPT_CROP</a></td></tr>
+<tr class="memdesc:ga9c771a757fc1294add611906b89ab2d2"><td class="mdescLeft">&#160;</td><td class="mdescRight">This option will enable lossless cropping.  <a href="group___turbo_j_p_e_g.html#ga9c771a757fc1294add611906b89ab2d2">More...</a><br /></td></tr>
+<tr class="separator:ga9c771a757fc1294add611906b89ab2d2"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga3acee7b48ade1b99e5588736007c2589"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga3acee7b48ade1b99e5588736007c2589">TJXOPT_GRAY</a></td></tr>
+<tr class="memdesc:ga3acee7b48ade1b99e5588736007c2589"><td class="mdescLeft">&#160;</td><td class="mdescRight">This option will discard the color data in the input image and produce a grayscale output image.  <a href="group___turbo_j_p_e_g.html#ga3acee7b48ade1b99e5588736007c2589">More...</a><br /></td></tr>
+<tr class="separator:ga3acee7b48ade1b99e5588736007c2589"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gafbf992bbf6e006705886333703ffab31"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gafbf992bbf6e006705886333703ffab31">TJXOPT_NOOUTPUT</a></td></tr>
+<tr class="memdesc:gafbf992bbf6e006705886333703ffab31"><td class="mdescLeft">&#160;</td><td class="mdescRight">This option will prevent <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> from outputting a JPEG image for this particular transform (this can be used in conjunction with a custom filter to capture the transformed DCT coefficients without transcoding them.)  <a href="group___turbo_j_p_e_g.html#gafbf992bbf6e006705886333703ffab31">More...</a><br /></td></tr>
+<tr class="separator:gafbf992bbf6e006705886333703ffab31"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gad2371c80674584ecc1a7d75e564cf026"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gad2371c80674584ecc1a7d75e564cf026">TJXOPT_PROGRESSIVE</a></td></tr>
+<tr class="memdesc:gad2371c80674584ecc1a7d75e564cf026"><td class="mdescLeft">&#160;</td><td class="mdescRight">This option will enable progressive entropy coding in the output image generated by this particular transform.  <a href="group___turbo_j_p_e_g.html#gad2371c80674584ecc1a7d75e564cf026">More...</a><br /></td></tr>
+<tr class="separator:gad2371c80674584ecc1a7d75e564cf026"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga153b468cfb905d0de61706c838986fe8"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga153b468cfb905d0de61706c838986fe8">TJXOPT_COPYNONE</a></td></tr>
+<tr class="memdesc:ga153b468cfb905d0de61706c838986fe8"><td class="mdescLeft">&#160;</td><td class="mdescRight">This option will prevent <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> from copying any extra markers (including EXIF and ICC profile data) from the source image to the output image.  <a href="group___turbo_j_p_e_g.html#ga153b468cfb905d0de61706c838986fe8">More...</a><br /></td></tr>
+<tr class="separator:ga153b468cfb905d0de61706c838986fe8"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga0aba955473315e405295d978f0c16511"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga0aba955473315e405295d978f0c16511">TJPAD</a>(width)</td></tr>
+<tr class="memdesc:ga0aba955473315e405295d978f0c16511"><td class="mdescLeft">&#160;</td><td class="mdescRight">Pad the given width to the nearest 32-bit boundary.  <a href="group___turbo_j_p_e_g.html#ga0aba955473315e405295d978f0c16511">More...</a><br /></td></tr>
+<tr class="separator:ga0aba955473315e405295d978f0c16511"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga84878bb65404204743aa18cac02781df"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga84878bb65404204743aa18cac02781df">TJSCALED</a>(dimension,  scalingFactor)</td></tr>
+<tr class="memdesc:ga84878bb65404204743aa18cac02781df"><td class="mdescLeft">&#160;</td><td class="mdescRight">Compute the scaled value of <code>dimension</code> using the given scaling factor.  <a href="group___turbo_j_p_e_g.html#ga84878bb65404204743aa18cac02781df">More...</a><br /></td></tr>
+<tr class="separator:ga84878bb65404204743aa18cac02781df"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table><table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="typedef-members"></a>
+Typedefs</h2></td></tr>
+<tr class="memitem:ga504805ec0161f1b505397ca0118bf8fd"><td class="memItemLeft" align="right" valign="top">typedef struct <a class="el" href="structtjtransform.html">tjtransform</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga504805ec0161f1b505397ca0118bf8fd">tjtransform</a></td></tr>
+<tr class="memdesc:ga504805ec0161f1b505397ca0118bf8fd"><td class="mdescLeft">&#160;</td><td class="mdescRight">Lossless transform.  <a href="group___turbo_j_p_e_g.html#ga504805ec0161f1b505397ca0118bf8fd">More...</a><br /></td></tr>
+<tr class="separator:ga504805ec0161f1b505397ca0118bf8fd"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga758d2634ecb4949de7815cba621f5763"><td class="memItemLeft" align="right" valign="top">typedef void *&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a></td></tr>
+<tr class="memdesc:ga758d2634ecb4949de7815cba621f5763"><td class="mdescLeft">&#160;</td><td class="mdescRight">TurboJPEG instance handle.  <a href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">More...</a><br /></td></tr>
+<tr class="separator:ga758d2634ecb4949de7815cba621f5763"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table><table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="enum-members"></a>
+Enumerations</h2></td></tr>
+<tr class="memitem:ga1d047060ea80bb9820d540bb928e9074"><td class="memItemLeft" align="right" valign="top">enum &#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">TJSAMP</a> { <br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074afb8da4f44197837bdec0a4f593dacae3">TJSAMP_444</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a136130902cc578f11f32429b59368404">TJSAMP_422</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a63085dbf683cfe39e513cdb6343e3737">TJSAMP_420</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a3f1c9504842ddc7a48d0f690754b6248">TJSAMP_GRAY</a>, 
+<br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074accf740e6f3aa6ba20ba922cad13cb974">TJSAMP_440</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a28ec62575e5ea295c3fde3001dc628e2">TJSAMP_411</a>
+<br />
+ }</td></tr>
+<tr class="memdesc:ga1d047060ea80bb9820d540bb928e9074"><td class="mdescLeft">&#160;</td><td class="mdescRight">Chrominance subsampling options.  <a href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">More...</a><br /></td></tr>
+<tr class="separator:ga1d047060ea80bb9820d540bb928e9074"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gac916144e26c3817ac514e64ae5d12e2a"><td class="memItemLeft" align="right" valign="top">enum &#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">TJPF</a> { <br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa7ce93230bff449518ce387c17e6ed37c">TJPF_RGB</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aab10624437fb8ef495a0b153e65749839">TJPF_BGR</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa83973bebb7e2dc6fa8bae89ff3f42e01">TJPF_RGBX</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa2a1fbf569ca79897eae886e3376ca4c8">TJPF_BGRX</a>, 
+<br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aaf6603b27147de47e212e75dac027b2af">TJPF_XBGR</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aadae996905efcfa3b42a0bb3bea7f9d84">TJPF_XRGB</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa5431b54b015337705f13118073711a1a">TJPF_GRAY</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa88d2e88fab67f6503cf972e14851cc12">TJPF_RGBA</a>, 
+<br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aac037ff1845cf9b74bb81a3659c2b9fb4">TJPF_BGRA</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa1ba1a7f1631dbeaa49a0a85fc4a40081">TJPF_ABGR</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aae8f846ed9d9de99b6e1dfe448848765c">TJPF_ARGB</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa7f5100ec44c91994e243f1cf55553f8b">TJPF_CMYK</a>, 
+<br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa84c1a6cead7952998e2fb895844a21ed">TJPF_UNKNOWN</a>
+<br />
+ }</td></tr>
+<tr class="memdesc:gac916144e26c3817ac514e64ae5d12e2a"><td class="mdescLeft">&#160;</td><td class="mdescRight">Pixel formats.  <a href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">More...</a><br /></td></tr>
+<tr class="separator:gac916144e26c3817ac514e64ae5d12e2a"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga4f83ad3368e0e29d1957be0efa7c3720"><td class="memItemLeft" align="right" valign="top">enum &#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga4f83ad3368e0e29d1957be0efa7c3720">TJCS</a> { <br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a677cb7ccb85c4038ac41964a2e09e555">TJCS_RGB</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a7389b8f65bb387ffedce3efd0d78ec75">TJCS_YCbCr</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720ab3e7d6a87f695e45b81c1b5262b5a50a">TJCS_GRAY</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a6c8b636152ac8195b869587db315ee53">TJCS_CMYK</a>, 
+<br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a53839e0fe867b76b58d16b0a1a7c598e">TJCS_YCCK</a>
+<br />
+ }</td></tr>
+<tr class="memdesc:ga4f83ad3368e0e29d1957be0efa7c3720"><td class="mdescLeft">&#160;</td><td class="mdescRight">JPEG colorspaces.  <a href="group___turbo_j_p_e_g.html#ga4f83ad3368e0e29d1957be0efa7c3720">More...</a><br /></td></tr>
+<tr class="separator:ga4f83ad3368e0e29d1957be0efa7c3720"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gafbc17cfa57d0d5d11fea35ac025950fe"><td class="memItemLeft" align="right" valign="top">enum &#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gafbc17cfa57d0d5d11fea35ac025950fe">TJERR</a> { <a class="el" href="group___turbo_j_p_e_g.html#ggafbc17cfa57d0d5d11fea35ac025950fea342dd6e2aedb47bb257b4e7568329b59">TJERR_WARNING</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#ggafbc17cfa57d0d5d11fea35ac025950feafc9cceeada13122b09e4851e3788039a">TJERR_FATAL</a>
+ }</td></tr>
+<tr class="memdesc:gafbc17cfa57d0d5d11fea35ac025950fe"><td class="mdescLeft">&#160;</td><td class="mdescRight">Error codes.  <a href="group___turbo_j_p_e_g.html#gafbc17cfa57d0d5d11fea35ac025950fe">More...</a><br /></td></tr>
+<tr class="separator:gafbc17cfa57d0d5d11fea35ac025950fe"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga2de531af4e7e6c4f124908376b354866"><td class="memItemLeft" align="right" valign="top">enum &#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga2de531af4e7e6c4f124908376b354866">TJXOP</a> { <br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866aad88c0366cd3f7d0eac9d7a3fa1c2c27">TJXOP_NONE</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866aa0df69776caa30f0fa28e26332d311ce">TJXOP_HFLIP</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a324eddfbec53b7e691f61e56929d0d5d">TJXOP_VFLIP</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a31060aed199f886afdd417f80499c32d">TJXOP_TRANSPOSE</a>, 
+<br />
+&#160;&#160;<a class="el" href="group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866af3b14d488aea6ece9e5b3df73a74d6a4">TJXOP_TRANSVERSE</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a43b2bbb23bc4bd548422d43fbe9af128">TJXOP_ROT90</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a140952eb8dd0300accfcc22726d69692">TJXOP_ROT180</a>, 
+<a class="el" href="group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a3064ee5dfb7f032df332818587567a08">TJXOP_ROT270</a>
+<br />
+ }</td></tr>
+<tr class="memdesc:ga2de531af4e7e6c4f124908376b354866"><td class="mdescLeft">&#160;</td><td class="mdescRight">Transform operations for <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a>  <a href="group___turbo_j_p_e_g.html#ga2de531af4e7e6c4f124908376b354866">More...</a><br /></td></tr>
+<tr class="separator:ga2de531af4e7e6c4f124908376b354866"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table><table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="func-members"></a>
+Functions</h2></td></tr>
+<tr class="memitem:ga9d63a05fc6d813f4aae06107041a37e8"><td class="memItemLeft" align="right" valign="top">DLLEXPORT <a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga9d63a05fc6d813f4aae06107041a37e8">tjInitCompress</a> (void)</td></tr>
+<tr class="memdesc:ga9d63a05fc6d813f4aae06107041a37e8"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create a TurboJPEG compressor instance.  <a href="group___turbo_j_p_e_g.html#ga9d63a05fc6d813f4aae06107041a37e8">More...</a><br /></td></tr>
+<tr class="separator:ga9d63a05fc6d813f4aae06107041a37e8"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gafbdce0112fd78fd38efae841443a9bcf"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gafbdce0112fd78fd38efae841443a9bcf">tjCompress2</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *srcBuf, int width, int pitch, int height, int pixelFormat, unsigned char **jpegBuf, unsigned long *jpegSize, int jpegSubsamp, int jpegQual, int flags)</td></tr>
+<tr class="memdesc:gafbdce0112fd78fd38efae841443a9bcf"><td class="mdescLeft">&#160;</td><td class="mdescRight">Compress an RGB, grayscale, or CMYK image into a JPEG image.  <a href="group___turbo_j_p_e_g.html#gafbdce0112fd78fd38efae841443a9bcf">More...</a><br /></td></tr>
+<tr class="separator:gafbdce0112fd78fd38efae841443a9bcf"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga7622a459b79aa1007e005b58783f875b"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga7622a459b79aa1007e005b58783f875b">tjCompressFromYUV</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *srcBuf, int width, int pad, int height, int subsamp, unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual, int flags)</td></tr>
+<tr class="memdesc:ga7622a459b79aa1007e005b58783f875b"><td class="mdescLeft">&#160;</td><td class="mdescRight">Compress a YUV planar image into a JPEG image.  <a href="group___turbo_j_p_e_g.html#ga7622a459b79aa1007e005b58783f875b">More...</a><br /></td></tr>
+<tr class="separator:ga7622a459b79aa1007e005b58783f875b"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga29ec5dfbd2d84b8724e951d6fa0d5d9e"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga29ec5dfbd2d84b8724e951d6fa0d5d9e">tjCompressFromYUVPlanes</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char **srcPlanes, int width, const int *strides, int height, int subsamp, unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual, int flags)</td></tr>
+<tr class="memdesc:ga29ec5dfbd2d84b8724e951d6fa0d5d9e"><td class="mdescLeft">&#160;</td><td class="mdescRight">Compress a set of Y, U (Cb), and V (Cr) image planes into a JPEG image.  <a href="group___turbo_j_p_e_g.html#ga29ec5dfbd2d84b8724e951d6fa0d5d9e">More...</a><br /></td></tr>
+<tr class="separator:ga29ec5dfbd2d84b8724e951d6fa0d5d9e"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga67ac12fee79073242cb216e07c9f1f90"><td class="memItemLeft" align="right" valign="top">DLLEXPORT unsigned long&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga67ac12fee79073242cb216e07c9f1f90">tjBufSize</a> (int width, int height, int jpegSubsamp)</td></tr>
+<tr class="memdesc:ga67ac12fee79073242cb216e07c9f1f90"><td class="mdescLeft">&#160;</td><td class="mdescRight">The maximum size of the buffer (in bytes) required to hold a JPEG image with the given parameters.  <a href="group___turbo_j_p_e_g.html#ga67ac12fee79073242cb216e07c9f1f90">More...</a><br /></td></tr>
+<tr class="separator:ga67ac12fee79073242cb216e07c9f1f90"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga2be2b9969d4df9ecce9b05deed273194"><td class="memItemLeft" align="right" valign="top">DLLEXPORT unsigned long&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga2be2b9969d4df9ecce9b05deed273194">tjBufSizeYUV2</a> (int width, int pad, int height, int subsamp)</td></tr>
+<tr class="memdesc:ga2be2b9969d4df9ecce9b05deed273194"><td class="mdescLeft">&#160;</td><td class="mdescRight">The size of the buffer (in bytes) required to hold a YUV planar image with the given parameters.  <a href="group___turbo_j_p_e_g.html#ga2be2b9969d4df9ecce9b05deed273194">More...</a><br /></td></tr>
+<tr class="separator:ga2be2b9969d4df9ecce9b05deed273194"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gab4ab7b24f6e797d79abaaa670373961d"><td class="memItemLeft" align="right" valign="top">DLLEXPORT unsigned long&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gab4ab7b24f6e797d79abaaa670373961d">tjPlaneSizeYUV</a> (int componentID, int width, int stride, int height, int subsamp)</td></tr>
+<tr class="memdesc:gab4ab7b24f6e797d79abaaa670373961d"><td class="mdescLeft">&#160;</td><td class="mdescRight">The size of the buffer (in bytes) required to hold a YUV image plane with the given parameters.  <a href="group___turbo_j_p_e_g.html#gab4ab7b24f6e797d79abaaa670373961d">More...</a><br /></td></tr>
+<tr class="separator:gab4ab7b24f6e797d79abaaa670373961d"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga63fb66bb1e36c74008c4634360becbb1"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga63fb66bb1e36c74008c4634360becbb1">tjPlaneWidth</a> (int componentID, int width, int subsamp)</td></tr>
+<tr class="memdesc:ga63fb66bb1e36c74008c4634360becbb1"><td class="mdescLeft">&#160;</td><td class="mdescRight">The plane width of a YUV image plane with the given parameters.  <a href="group___turbo_j_p_e_g.html#ga63fb66bb1e36c74008c4634360becbb1">More...</a><br /></td></tr>
+<tr class="separator:ga63fb66bb1e36c74008c4634360becbb1"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga1a209696c6a80748f20e134b3c64789f"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga1a209696c6a80748f20e134b3c64789f">tjPlaneHeight</a> (int componentID, int height, int subsamp)</td></tr>
+<tr class="memdesc:ga1a209696c6a80748f20e134b3c64789f"><td class="mdescLeft">&#160;</td><td class="mdescRight">The plane height of a YUV image plane with the given parameters.  <a href="group___turbo_j_p_e_g.html#ga1a209696c6a80748f20e134b3c64789f">More...</a><br /></td></tr>
+<tr class="separator:ga1a209696c6a80748f20e134b3c64789f"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gac519b922cdf446e97d0cdcba513636bf"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gac519b922cdf446e97d0cdcba513636bf">tjEncodeYUV3</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *srcBuf, int width, int pitch, int height, int pixelFormat, unsigned char *dstBuf, int pad, int subsamp, int flags)</td></tr>
+<tr class="memdesc:gac519b922cdf446e97d0cdcba513636bf"><td class="mdescLeft">&#160;</td><td class="mdescRight">Encode an RGB or grayscale image into a YUV planar image.  <a href="group___turbo_j_p_e_g.html#gac519b922cdf446e97d0cdcba513636bf">More...</a><br /></td></tr>
+<tr class="separator:gac519b922cdf446e97d0cdcba513636bf"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gae2d04c72457fe7f4d60cf78ab1b1feb1"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gae2d04c72457fe7f4d60cf78ab1b1feb1">tjEncodeYUVPlanes</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *srcBuf, int width, int pitch, int height, int pixelFormat, unsigned char **dstPlanes, int *strides, int subsamp, int flags)</td></tr>
+<tr class="memdesc:gae2d04c72457fe7f4d60cf78ab1b1feb1"><td class="mdescLeft">&#160;</td><td class="mdescRight">Encode an RGB or grayscale image into separate Y, U (Cb), and V (Cr) image planes.  <a href="group___turbo_j_p_e_g.html#gae2d04c72457fe7f4d60cf78ab1b1feb1">More...</a><br /></td></tr>
+<tr class="separator:gae2d04c72457fe7f4d60cf78ab1b1feb1"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga52300eac3f3d9ef4bab303bc244f62d3"><td class="memItemLeft" align="right" valign="top">DLLEXPORT <a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga52300eac3f3d9ef4bab303bc244f62d3">tjInitDecompress</a> (void)</td></tr>
+<tr class="memdesc:ga52300eac3f3d9ef4bab303bc244f62d3"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create a TurboJPEG decompressor instance.  <a href="group___turbo_j_p_e_g.html#ga52300eac3f3d9ef4bab303bc244f62d3">More...</a><br /></td></tr>
+<tr class="separator:ga52300eac3f3d9ef4bab303bc244f62d3"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga0595681096bba7199cc6f3533cb25f77"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga0595681096bba7199cc6f3533cb25f77">tjDecompressHeader3</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *jpegBuf, unsigned long jpegSize, int *width, int *height, int *jpegSubsamp, int *jpegColorspace)</td></tr>
+<tr class="memdesc:ga0595681096bba7199cc6f3533cb25f77"><td class="mdescLeft">&#160;</td><td class="mdescRight">Retrieve information about a JPEG image without decompressing it.  <a href="group___turbo_j_p_e_g.html#ga0595681096bba7199cc6f3533cb25f77">More...</a><br /></td></tr>
+<tr class="separator:ga0595681096bba7199cc6f3533cb25f77"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gac3854476006b10787bd128f7ede48057"><td class="memItemLeft" align="right" valign="top">DLLEXPORT <a class="el" href="structtjscalingfactor.html">tjscalingfactor</a> *&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gac3854476006b10787bd128f7ede48057">tjGetScalingFactors</a> (int *numscalingfactors)</td></tr>
+<tr class="memdesc:gac3854476006b10787bd128f7ede48057"><td class="mdescLeft">&#160;</td><td class="mdescRight">Returns a list of fractional scaling factors that the JPEG decompressor in this implementation of TurboJPEG supports.  <a href="group___turbo_j_p_e_g.html#gac3854476006b10787bd128f7ede48057">More...</a><br /></td></tr>
+<tr class="separator:gac3854476006b10787bd128f7ede48057"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gae9eccef8b682a48f43a9117c231ed013"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gae9eccef8b682a48f43a9117c231ed013">tjDecompress2</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf, int width, int pitch, int height, int pixelFormat, int flags)</td></tr>
+<tr class="memdesc:gae9eccef8b682a48f43a9117c231ed013"><td class="mdescLeft">&#160;</td><td class="mdescRight">Decompress a JPEG image to an RGB, grayscale, or CMYK image.  <a href="group___turbo_j_p_e_g.html#gae9eccef8b682a48f43a9117c231ed013">More...</a><br /></td></tr>
+<tr class="separator:gae9eccef8b682a48f43a9117c231ed013"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga04d1e839ff9a0860dd1475cff78d3364"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga04d1e839ff9a0860dd1475cff78d3364">tjDecompressToYUV2</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf, int width, int pad, int height, int flags)</td></tr>
+<tr class="memdesc:ga04d1e839ff9a0860dd1475cff78d3364"><td class="mdescLeft">&#160;</td><td class="mdescRight">Decompress a JPEG image to a YUV planar image.  <a href="group___turbo_j_p_e_g.html#ga04d1e839ff9a0860dd1475cff78d3364">More...</a><br /></td></tr>
+<tr class="separator:ga04d1e839ff9a0860dd1475cff78d3364"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gaa59f901a5258ada5bd0185ad59368540"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gaa59f901a5258ada5bd0185ad59368540">tjDecompressToYUVPlanes</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char **dstPlanes, int width, int *strides, int height, int flags)</td></tr>
+<tr class="memdesc:gaa59f901a5258ada5bd0185ad59368540"><td class="mdescLeft">&#160;</td><td class="mdescRight">Decompress a JPEG image into separate Y, U (Cb), and V (Cr) image planes.  <a href="group___turbo_j_p_e_g.html#gaa59f901a5258ada5bd0185ad59368540">More...</a><br /></td></tr>
+<tr class="separator:gaa59f901a5258ada5bd0185ad59368540"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga70abbf38f77a26fd6da8813bef96f695"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga70abbf38f77a26fd6da8813bef96f695">tjDecodeYUV</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *srcBuf, int pad, int subsamp, unsigned char *dstBuf, int width, int pitch, int height, int pixelFormat, int flags)</td></tr>
+<tr class="memdesc:ga70abbf38f77a26fd6da8813bef96f695"><td class="mdescLeft">&#160;</td><td class="mdescRight">Decode a YUV planar image into an RGB or grayscale image.  <a href="group___turbo_j_p_e_g.html#ga70abbf38f77a26fd6da8813bef96f695">More...</a><br /></td></tr>
+<tr class="separator:ga70abbf38f77a26fd6da8813bef96f695"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga10e837c07fa9d25770565b237d3898d9"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga10e837c07fa9d25770565b237d3898d9">tjDecodeYUVPlanes</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char **srcPlanes, const int *strides, int subsamp, unsigned char *dstBuf, int width, int pitch, int height, int pixelFormat, int flags)</td></tr>
+<tr class="memdesc:ga10e837c07fa9d25770565b237d3898d9"><td class="mdescLeft">&#160;</td><td class="mdescRight">Decode a set of Y, U (Cb), and V (Cr) image planes into an RGB or grayscale image.  <a href="group___turbo_j_p_e_g.html#ga10e837c07fa9d25770565b237d3898d9">More...</a><br /></td></tr>
+<tr class="separator:ga10e837c07fa9d25770565b237d3898d9"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga928beff6ac248ceadf01089fc6b41957"><td class="memItemLeft" align="right" valign="top">DLLEXPORT <a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga928beff6ac248ceadf01089fc6b41957">tjInitTransform</a> (void)</td></tr>
+<tr class="memdesc:ga928beff6ac248ceadf01089fc6b41957"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create a new TurboJPEG transformer instance.  <a href="group___turbo_j_p_e_g.html#ga928beff6ac248ceadf01089fc6b41957">More...</a><br /></td></tr>
+<tr class="separator:ga928beff6ac248ceadf01089fc6b41957"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga9cb8abf4cc91881e04a0329b2270be25"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25">tjTransform</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle, const unsigned char *jpegBuf, unsigned long jpegSize, int n, unsigned char **dstBufs, unsigned long *dstSizes, <a class="el" href="structtjtransform.html">tjtransform</a> *transforms, int flags)</td></tr>
+<tr class="memdesc:ga9cb8abf4cc91881e04a0329b2270be25"><td class="mdescLeft">&#160;</td><td class="mdescRight">Losslessly transform a JPEG image into another JPEG image.  <a href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25">More...</a><br /></td></tr>
+<tr class="separator:ga9cb8abf4cc91881e04a0329b2270be25"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga75f355fa27225ba1a4ee392c852394d2"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga75f355fa27225ba1a4ee392c852394d2">tjDestroy</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle)</td></tr>
+<tr class="memdesc:ga75f355fa27225ba1a4ee392c852394d2"><td class="mdescLeft">&#160;</td><td class="mdescRight">Destroy a TurboJPEG compressor, decompressor, or transformer instance.  <a href="group___turbo_j_p_e_g.html#ga75f355fa27225ba1a4ee392c852394d2">More...</a><br /></td></tr>
+<tr class="separator:ga75f355fa27225ba1a4ee392c852394d2"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gaec627dd4c5f30b7a775a7aea3bec5d83"><td class="memItemLeft" align="right" valign="top">DLLEXPORT unsigned char *&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83">tjAlloc</a> (int bytes)</td></tr>
+<tr class="memdesc:gaec627dd4c5f30b7a775a7aea3bec5d83"><td class="mdescLeft">&#160;</td><td class="mdescRight">Allocate an image buffer for use with TurboJPEG.  <a href="group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83">More...</a><br /></td></tr>
+<tr class="separator:gaec627dd4c5f30b7a775a7aea3bec5d83"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gaffbd83c375e79f5db4b5c5d8ad4466e7"><td class="memItemLeft" align="right" valign="top">DLLEXPORT unsigned char *&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gaffbd83c375e79f5db4b5c5d8ad4466e7">tjLoadImage</a> (const char *filename, int *width, int align, int *height, int *pixelFormat, int flags)</td></tr>
+<tr class="memdesc:gaffbd83c375e79f5db4b5c5d8ad4466e7"><td class="mdescLeft">&#160;</td><td class="mdescRight">Load an uncompressed image from disk into memory.  <a href="group___turbo_j_p_e_g.html#gaffbd83c375e79f5db4b5c5d8ad4466e7">More...</a><br /></td></tr>
+<tr class="separator:gaffbd83c375e79f5db4b5c5d8ad4466e7"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga6f445b22d8933ae4815b3370a538d879"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga6f445b22d8933ae4815b3370a538d879">tjSaveImage</a> (const char *filename, unsigned char *buffer, int width, int pitch, int height, int pixelFormat, int flags)</td></tr>
+<tr class="memdesc:ga6f445b22d8933ae4815b3370a538d879"><td class="mdescLeft">&#160;</td><td class="mdescRight">Save an uncompressed image from memory to disk.  <a href="group___turbo_j_p_e_g.html#ga6f445b22d8933ae4815b3370a538d879">More...</a><br /></td></tr>
+<tr class="separator:ga6f445b22d8933ae4815b3370a538d879"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gaea863d2da0cdb609563aabdf9196514b"><td class="memItemLeft" align="right" valign="top">DLLEXPORT void&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gaea863d2da0cdb609563aabdf9196514b">tjFree</a> (unsigned char *buffer)</td></tr>
+<tr class="memdesc:gaea863d2da0cdb609563aabdf9196514b"><td class="mdescLeft">&#160;</td><td class="mdescRight">Free an image buffer previously allocated by TurboJPEG.  <a href="group___turbo_j_p_e_g.html#gaea863d2da0cdb609563aabdf9196514b">More...</a><br /></td></tr>
+<tr class="separator:gaea863d2da0cdb609563aabdf9196514b"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga1ead8574f9f39fbafc6b497124e7aafa"><td class="memItemLeft" align="right" valign="top">DLLEXPORT char *&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa">tjGetErrorStr2</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle)</td></tr>
+<tr class="memdesc:ga1ead8574f9f39fbafc6b497124e7aafa"><td class="mdescLeft">&#160;</td><td class="mdescRight">Returns a descriptive error message explaining why the last command failed.  <a href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa">More...</a><br /></td></tr>
+<tr class="separator:ga1ead8574f9f39fbafc6b497124e7aafa"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga414feeffbf860ebd31c745df203de410"><td class="memItemLeft" align="right" valign="top">DLLEXPORT int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410">tjGetErrorCode</a> (<a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> handle)</td></tr>
+<tr class="memdesc:ga414feeffbf860ebd31c745df203de410"><td class="mdescLeft">&#160;</td><td class="mdescRight">Returns a code indicating the severity of the last error.  <a href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410">More...</a><br /></td></tr>
+<tr class="separator:ga414feeffbf860ebd31c745df203de410"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table><table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="var-members"></a>
+Variables</h2></td></tr>
+<tr class="memitem:ga9e61e7cd47a15a173283ba94e781308c"><td class="memItemLeft" align="right" valign="top">static const int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga9e61e7cd47a15a173283ba94e781308c">tjMCUWidth</a> <a class="el" href="group___turbo_j_p_e_g.html#ga5ef3d169162ce77ce348e292a0b7477c">TJ_NUMSAMP</a></td></tr>
+<tr class="memdesc:ga9e61e7cd47a15a173283ba94e781308c"><td class="mdescLeft">&#160;</td><td class="mdescRight">MCU block width (in pixels) for a given level of chrominance subsampling.  <a href="group___turbo_j_p_e_g.html#ga9e61e7cd47a15a173283ba94e781308c">More...</a><br /></td></tr>
+<tr class="separator:ga9e61e7cd47a15a173283ba94e781308c"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gabd247bb9fecb393eca57366feb8327bf"><td class="memItemLeft" align="right" valign="top">static const int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gabd247bb9fecb393eca57366feb8327bf">tjMCUHeight</a> <a class="el" href="group___turbo_j_p_e_g.html#ga5ef3d169162ce77ce348e292a0b7477c">TJ_NUMSAMP</a></td></tr>
+<tr class="memdesc:gabd247bb9fecb393eca57366feb8327bf"><td class="mdescLeft">&#160;</td><td class="mdescRight">MCU block height (in pixels) for a given level of chrominance subsampling.  <a href="group___turbo_j_p_e_g.html#gabd247bb9fecb393eca57366feb8327bf">More...</a><br /></td></tr>
+<tr class="separator:gabd247bb9fecb393eca57366feb8327bf"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gadd9b446742ac8a3923f7992c7988fea8"><td class="memItemLeft" align="right" valign="top">static const int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gadd9b446742ac8a3923f7992c7988fea8">tjRedOffset</a> <a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td></tr>
+<tr class="memdesc:gadd9b446742ac8a3923f7992c7988fea8"><td class="mdescLeft">&#160;</td><td class="mdescRight">Red offset (in bytes) for a given pixel format.  <a href="group___turbo_j_p_e_g.html#gadd9b446742ac8a3923f7992c7988fea8">More...</a><br /></td></tr>
+<tr class="separator:gadd9b446742ac8a3923f7992c7988fea8"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga82d6e35da441112a411da41923c0ba2f"><td class="memItemLeft" align="right" valign="top">static const int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga82d6e35da441112a411da41923c0ba2f">tjGreenOffset</a> <a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td></tr>
+<tr class="memdesc:ga82d6e35da441112a411da41923c0ba2f"><td class="mdescLeft">&#160;</td><td class="mdescRight">Green offset (in bytes) for a given pixel format.  <a href="group___turbo_j_p_e_g.html#ga82d6e35da441112a411da41923c0ba2f">More...</a><br /></td></tr>
+<tr class="separator:ga82d6e35da441112a411da41923c0ba2f"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga84e2e35d3f08025f976ec1ec53693dea"><td class="memItemLeft" align="right" valign="top">static const int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga84e2e35d3f08025f976ec1ec53693dea">tjBlueOffset</a> <a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td></tr>
+<tr class="memdesc:ga84e2e35d3f08025f976ec1ec53693dea"><td class="mdescLeft">&#160;</td><td class="mdescRight">Blue offset (in bytes) for a given pixel format.  <a href="group___turbo_j_p_e_g.html#ga84e2e35d3f08025f976ec1ec53693dea">More...</a><br /></td></tr>
+<tr class="separator:ga84e2e35d3f08025f976ec1ec53693dea"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ga5af0ab065feefd526debf1e20c43e837"><td class="memItemLeft" align="right" valign="top">static const int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#ga5af0ab065feefd526debf1e20c43e837">tjAlphaOffset</a> <a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td></tr>
+<tr class="memdesc:ga5af0ab065feefd526debf1e20c43e837"><td class="mdescLeft">&#160;</td><td class="mdescRight">Alpha offset (in bytes) for a given pixel format.  <a href="group___turbo_j_p_e_g.html#ga5af0ab065feefd526debf1e20c43e837">More...</a><br /></td></tr>
+<tr class="separator:ga5af0ab065feefd526debf1e20c43e837"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:gad77cf8fe5b2bfd3cb3f53098146abb4c"><td class="memItemLeft" align="right" valign="top">static const int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c">tjPixelSize</a> <a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td></tr>
+<tr class="memdesc:gad77cf8fe5b2bfd3cb3f53098146abb4c"><td class="mdescLeft">&#160;</td><td class="mdescRight">Pixel size (in bytes) for a given pixel format.  <a href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c">More...</a><br /></td></tr>
+<tr class="separator:gad77cf8fe5b2bfd3cb3f53098146abb4c"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table>
+<a name="details" id="details"></a><h2 class="groupheader">Detailed Description</h2>
+<p>TurboJPEG API. </p>
+<p>This API provides an interface for generating, decoding, and transforming planar YUV and JPEG images in memory.</p>
+<p><a class="anchor" id="YUVnotes"></a></p>
+<h2><a class="anchor" id="autotoc_md0"></a>
+YUV Image Format Notes</h2>
+<p>Technically, the JPEG format uses the YCbCr colorspace (which is technically not a colorspace but a color transform), but per the convention of the digital video community, the TurboJPEG API uses "YUV" to refer to an image format consisting of Y, Cb, and Cr image planes.</p>
+<p>Each plane is simply a 2D array of bytes, each byte representing the value of one of the components (Y, Cb, or Cr) at a particular location in the image. The width and height of each plane are determined by the image width, height, and level of chrominance subsampling. The luminance plane width is the image width padded to the nearest multiple of the horizontal subsampling factor (2 in the case of 4:2:0 and 4:2:2, 4 in the case of 4:1:1, 1 in the case of 4:4:4 or grayscale.) Similarly, the luminance plane height is the image height padded to the nearest multiple of the vertical subsampling factor (2 in the case of 4:2:0 or 4:4:0, 1 in the case of 4:4:4 or grayscale.) This is irrespective of any additional padding that may be specified as an argument to the various YUV functions. The chrominance plane width is equal to the luminance plane width divided by the horizontal subsampling factor, and the chrominance plane height is equal to the luminance plane height divided by the vertical subsampling factor.</p>
+<p>For example, if the source image is 35 x 35 pixels and 4:2:2 subsampling is used, then the luminance plane would be 36 x 35 bytes, and each of the chrominance planes would be 18 x 35 bytes. If you specify a line padding of 4 bytes on top of this, then the luminance plane would be 36 x 35 bytes, and each of the chrominance planes would be 20 x 35 bytes. </p>
+<h2 class="groupheader">Macro Definition Documentation</h2>
+<a id="ga39f57a6fb02d9cf32e7b6890099b5a71"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga39f57a6fb02d9cf32e7b6890099b5a71">&#9670;&nbsp;</a></span>TJ_NUMCS</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJ_NUMCS</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The number of JPEG colorspaces. </p>
+
+</div>
+</div>
+<a id="ga79bde1b4a3e2351e00887e47781b966e"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga79bde1b4a3e2351e00887e47781b966e">&#9670;&nbsp;</a></span>TJ_NUMERR</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJ_NUMERR</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The number of error codes. </p>
+
+</div>
+</div>
+<a id="ga7010a4402f54a45ba822ad8675a4655e"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga7010a4402f54a45ba822ad8675a4655e">&#9670;&nbsp;</a></span>TJ_NUMPF</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJ_NUMPF</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The number of pixel formats. </p>
+
+</div>
+</div>
+<a id="ga5ef3d169162ce77ce348e292a0b7477c"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga5ef3d169162ce77ce348e292a0b7477c">&#9670;&nbsp;</a></span>TJ_NUMSAMP</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJ_NUMSAMP</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The number of chrominance subsampling options. </p>
+
+</div>
+</div>
+<a id="ga0f6dbd18adf38b7d46ac547f0f4d562c"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga0f6dbd18adf38b7d46ac547f0f4d562c">&#9670;&nbsp;</a></span>TJ_NUMXOP</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJ_NUMXOP</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The number of transform operations. </p>
+
+</div>
+</div>
+<a id="gacb233cfd722d66d1ccbf48a7de81f0e0"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gacb233cfd722d66d1ccbf48a7de81f0e0">&#9670;&nbsp;</a></span>TJFLAG_ACCURATEDCT</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJFLAG_ACCURATEDCT</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Use the most accurate DCT/IDCT algorithm available in the underlying codec. </p>
+<p>The default if this flag is not specified is implementation-specific. For example, the implementation of TurboJPEG for libjpeg-turbo uses the fast algorithm by default when compressing, because this has been shown to have only a very slight effect on accuracy, but it uses the accurate algorithm when decompressing, because this has been shown to have a larger effect. </p>
+
+</div>
+</div>
+<a id="ga72ecf4ebe6eb702d3c6f5ca27455e1ec"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga72ecf4ebe6eb702d3c6f5ca27455e1ec">&#9670;&nbsp;</a></span>TJFLAG_BOTTOMUP</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJFLAG_BOTTOMUP</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The uncompressed source/destination image is stored in bottom-up (Windows, OpenGL) order, not top-down (X11) order. </p>
+
+</div>
+</div>
+<a id="gaabce235db80d3f698b27f36cbd453da2"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gaabce235db80d3f698b27f36cbd453da2">&#9670;&nbsp;</a></span>TJFLAG_FASTDCT</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJFLAG_FASTDCT</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Use the fastest DCT/IDCT algorithm available in the underlying codec. </p>
+<p>The default if this flag is not specified is implementation-specific. For example, the implementation of TurboJPEG for libjpeg-turbo uses the fast algorithm by default when compressing, because this has been shown to have only a very slight effect on accuracy, but it uses the accurate algorithm when decompressing, because this has been shown to have a larger effect. </p>
+
+</div>
+</div>
+<a id="ga4ee4506c81177a06f77e2504a22efd2d"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga4ee4506c81177a06f77e2504a22efd2d">&#9670;&nbsp;</a></span>TJFLAG_FASTUPSAMPLE</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJFLAG_FASTUPSAMPLE</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>When decompressing an image that was compressed using chrominance subsampling, use the fastest chrominance upsampling algorithm available in the underlying codec. </p>
+<p>The default is to use smooth upsampling, which creates a smooth transition between neighboring chrominance components in order to reduce upsampling artifacts in the decompressed image. </p>
+
+</div>
+</div>
+<a id="ga163e6482dc5096831feef9c79ff3f805"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga163e6482dc5096831feef9c79ff3f805">&#9670;&nbsp;</a></span>TJFLAG_LIMITSCANS</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJFLAG_LIMITSCANS</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Limit the number of progressive JPEG scans that the decompression and transform functions will process. </p>
+<p>If a progressive JPEG image contains an unreasonably large number of scans, then this flag will cause the decompression and transform functions to return an error. The primary purpose of this is to allow security-critical applications to guard against an exploit of the progressive JPEG format described in <a href="https://libjpeg-turbo.org/pmwiki/uploads/About/TwoIssueswiththeJPEGStandard.pdf" target="_blank">this report</a>. </p>
+
+</div>
+</div>
+<a id="ga8808d403c68b62aaa58a4c1e58e98963"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga8808d403c68b62aaa58a4c1e58e98963">&#9670;&nbsp;</a></span>TJFLAG_NOREALLOC</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJFLAG_NOREALLOC</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Disable buffer (re)allocation. </p>
+<p>If passed to one of the JPEG compression or transform functions, this flag will cause those functions to generate an error if the JPEG image buffer is invalid or too small rather than attempting to allocate or reallocate that buffer. This reproduces the behavior of earlier versions of TurboJPEG. </p>
+
+</div>
+</div>
+<a id="ga43b426750b46190a25d34a67ef76df1b"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga43b426750b46190a25d34a67ef76df1b">&#9670;&nbsp;</a></span>TJFLAG_PROGRESSIVE</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJFLAG_PROGRESSIVE</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Use progressive entropy coding in JPEG images generated by the compression and transform functions. </p>
+<p>Progressive entropy coding will generally improve compression relative to baseline entropy coding (the default), but it will reduce compression and decompression performance considerably. </p>
+
+</div>
+</div>
+<a id="ga519cfa4ef6c18d9e5b455fdf59306a3a"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga519cfa4ef6c18d9e5b455fdf59306a3a">&#9670;&nbsp;</a></span>TJFLAG_STOPONWARNING</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJFLAG_STOPONWARNING</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Immediately discontinue the current compression/decompression/transform operation if the underlying codec throws a warning (non-fatal error). </p>
+<p>The default behavior is to allow the operation to complete unless a fatal error is encountered. </p>
+
+</div>
+</div>
+<a id="ga0aba955473315e405295d978f0c16511"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga0aba955473315e405295d978f0c16511">&#9670;&nbsp;</a></span>TJPAD</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJPAD</td>
+          <td>(</td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">width</td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Pad the given width to the nearest 32-bit boundary. </p>
+
+</div>
+</div>
+<a id="ga84878bb65404204743aa18cac02781df"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga84878bb65404204743aa18cac02781df">&#9670;&nbsp;</a></span>TJSCALED</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJSCALED</td>
+          <td>(</td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">dimension, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">scalingFactor&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Compute the scaled value of <code>dimension</code> using the given scaling factor. </p>
+<p>This macro performs the integer equivalent of <code>ceil(dimension * scalingFactor)</code>. </p>
+
+</div>
+</div>
+<a id="ga153b468cfb905d0de61706c838986fe8"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga153b468cfb905d0de61706c838986fe8">&#9670;&nbsp;</a></span>TJXOPT_COPYNONE</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJXOPT_COPYNONE</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>This option will prevent <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> from copying any extra markers (including EXIF and ICC profile data) from the source image to the output image. </p>
+
+</div>
+</div>
+<a id="ga9c771a757fc1294add611906b89ab2d2"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga9c771a757fc1294add611906b89ab2d2">&#9670;&nbsp;</a></span>TJXOPT_CROP</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJXOPT_CROP</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>This option will enable lossless cropping. </p>
+<p>See <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> for more information. </p>
+
+</div>
+</div>
+<a id="ga3acee7b48ade1b99e5588736007c2589"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga3acee7b48ade1b99e5588736007c2589">&#9670;&nbsp;</a></span>TJXOPT_GRAY</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJXOPT_GRAY</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>This option will discard the color data in the input image and produce a grayscale output image. </p>
+
+</div>
+</div>
+<a id="gafbf992bbf6e006705886333703ffab31"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gafbf992bbf6e006705886333703ffab31">&#9670;&nbsp;</a></span>TJXOPT_NOOUTPUT</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJXOPT_NOOUTPUT</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>This option will prevent <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> from outputting a JPEG image for this particular transform (this can be used in conjunction with a custom filter to capture the transformed DCT coefficients without transcoding them.) </p>
+
+</div>
+</div>
+<a id="ga50e03cb5ed115330e212417429600b00"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga50e03cb5ed115330e212417429600b00">&#9670;&nbsp;</a></span>TJXOPT_PERFECT</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJXOPT_PERFECT</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>This option will cause <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> to return an error if the transform is not perfect. </p>
+<p>Lossless transforms operate on MCU blocks, whose size depends on the level of chrominance subsampling used (see <a class="el" href="group___turbo_j_p_e_g.html#ga9e61e7cd47a15a173283ba94e781308c" title="MCU block width (in pixels) for a given level of chrominance subsampling.">tjMCUWidth</a> and <a class="el" href="group___turbo_j_p_e_g.html#gabd247bb9fecb393eca57366feb8327bf" title="MCU block height (in pixels) for a given level of chrominance subsampling.">tjMCUHeight</a>.) If the image's width or height is not evenly divisible by the MCU block size, then there will be partial MCU blocks on the right and/or bottom edges. It is not possible to move these partial MCU blocks to the top or left of the image, so any transform that would require that is "imperfect." If this option is not specified, then any partial MCU blocks that cannot be transformed will be left in place, which will create odd-looking strips on the right or bottom edge of the image. </p>
+
+</div>
+</div>
+<a id="gad2371c80674584ecc1a7d75e564cf026"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gad2371c80674584ecc1a7d75e564cf026">&#9670;&nbsp;</a></span>TJXOPT_PROGRESSIVE</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJXOPT_PROGRESSIVE</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>This option will enable progressive entropy coding in the output image generated by this particular transform. </p>
+<p>Progressive entropy coding will generally improve compression relative to baseline entropy coding (the default), but it will reduce compression and decompression performance considerably. </p>
+
+</div>
+</div>
+<a id="ga319826b7eb1583c0595bbe7b95428709"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga319826b7eb1583c0595bbe7b95428709">&#9670;&nbsp;</a></span>TJXOPT_TRIM</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define TJXOPT_TRIM</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>This option will cause <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> to discard any partial MCU blocks that cannot be transformed. </p>
+
+</div>
+</div>
+<h2 class="groupheader">Typedef Documentation</h2>
+<a id="ga758d2634ecb4949de7815cba621f5763"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga758d2634ecb4949de7815cba621f5763">&#9670;&nbsp;</a></span>tjhandle</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">typedef void* <a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>TurboJPEG instance handle. </p>
+
+</div>
+</div>
+<a id="ga504805ec0161f1b505397ca0118bf8fd"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga504805ec0161f1b505397ca0118bf8fd">&#9670;&nbsp;</a></span>tjtransform</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">typedef struct <a class="el" href="structtjtransform.html">tjtransform</a> <a class="el" href="structtjtransform.html">tjtransform</a></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Lossless transform. </p>
+
+</div>
+</div>
+<h2 class="groupheader">Enumeration Type Documentation</h2>
+<a id="ga4f83ad3368e0e29d1957be0efa7c3720"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga4f83ad3368e0e29d1957be0efa7c3720">&#9670;&nbsp;</a></span>TJCS</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">enum <a class="el" href="group___turbo_j_p_e_g.html#ga4f83ad3368e0e29d1957be0efa7c3720">TJCS</a></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>JPEG colorspaces. </p>
+<table class="fieldtable">
+<tr><th colspan="2">Enumerator</th></tr><tr><td class="fieldname"><a id="gga4f83ad3368e0e29d1957be0efa7c3720a677cb7ccb85c4038ac41964a2e09e555"></a>TJCS_RGB&#160;</td><td class="fielddoc"><p>RGB colorspace. </p>
+<p>When compressing the JPEG image, the R, G, and B components in the source image are reordered into image planes, but no colorspace conversion or subsampling is performed. RGB JPEG images can be decompressed to any of the extended RGB pixel formats or grayscale, but they cannot be decompressed to YUV images. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga4f83ad3368e0e29d1957be0efa7c3720a7389b8f65bb387ffedce3efd0d78ec75"></a>TJCS_YCbCr&#160;</td><td class="fielddoc"><p>YCbCr colorspace. </p>
+<p>YCbCr is not an absolute colorspace but rather a mathematical transformation of RGB designed solely for storage and transmission. YCbCr images must be converted to RGB before they can actually be displayed. In the YCbCr colorspace, the Y (luminance) component represents the black &amp; white portion of the original image, and the Cb and Cr (chrominance) components represent the color portion of the original image. Originally, the analog equivalent of this transformation allowed the same signal to drive both black &amp; white and color televisions, but JPEG images use YCbCr primarily because it allows the color data to be optionally subsampled for the purposes of reducing bandwidth or disk space. YCbCr is the most common JPEG colorspace, and YCbCr JPEG images can be compressed from and decompressed to any of the extended RGB pixel formats or grayscale, or they can be decompressed to YUV planar images. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga4f83ad3368e0e29d1957be0efa7c3720ab3e7d6a87f695e45b81c1b5262b5a50a"></a>TJCS_GRAY&#160;</td><td class="fielddoc"><p>Grayscale colorspace. </p>
+<p>The JPEG image retains only the luminance data (Y component), and any color data from the source image is discarded. Grayscale JPEG images can be compressed from and decompressed to any of the extended RGB pixel formats or grayscale, or they can be decompressed to YUV planar images. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga4f83ad3368e0e29d1957be0efa7c3720a6c8b636152ac8195b869587db315ee53"></a>TJCS_CMYK&#160;</td><td class="fielddoc"><p>CMYK colorspace. </p>
+<p>When compressing the JPEG image, the C, M, Y, and K components in the source image are reordered into image planes, but no colorspace conversion or subsampling is performed. CMYK JPEG images can only be decompressed to CMYK pixels. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga4f83ad3368e0e29d1957be0efa7c3720a53839e0fe867b76b58d16b0a1a7c598e"></a>TJCS_YCCK&#160;</td><td class="fielddoc"><p>YCCK colorspace. </p>
+<p>YCCK (AKA "YCbCrK") is not an absolute colorspace but rather a mathematical transformation of CMYK designed solely for storage and transmission. It is to CMYK as YCbCr is to RGB. CMYK pixels can be reversibly transformed into YCCK, and as with YCbCr, the chrominance components in the YCCK pixels can be subsampled without incurring major perceptual loss. YCCK JPEG images can only be compressed from and decompressed to CMYK pixels. </p>
+</td></tr>
+</table>
+
+</div>
+</div>
+<a id="gafbc17cfa57d0d5d11fea35ac025950fe"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gafbc17cfa57d0d5d11fea35ac025950fe">&#9670;&nbsp;</a></span>TJERR</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">enum <a class="el" href="group___turbo_j_p_e_g.html#gafbc17cfa57d0d5d11fea35ac025950fe">TJERR</a></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Error codes. </p>
+<table class="fieldtable">
+<tr><th colspan="2">Enumerator</th></tr><tr><td class="fieldname"><a id="ggafbc17cfa57d0d5d11fea35ac025950fea342dd6e2aedb47bb257b4e7568329b59"></a>TJERR_WARNING&#160;</td><td class="fielddoc"><p>The error was non-fatal and recoverable, but the image may still be corrupt. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggafbc17cfa57d0d5d11fea35ac025950feafc9cceeada13122b09e4851e3788039a"></a>TJERR_FATAL&#160;</td><td class="fielddoc"><p>The error was fatal and non-recoverable. </p>
+</td></tr>
+</table>
+
+</div>
+</div>
+<a id="gac916144e26c3817ac514e64ae5d12e2a"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gac916144e26c3817ac514e64ae5d12e2a">&#9670;&nbsp;</a></span>TJPF</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">enum <a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">TJPF</a></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Pixel formats. </p>
+<table class="fieldtable">
+<tr><th colspan="2">Enumerator</th></tr><tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aa7ce93230bff449518ce387c17e6ed37c"></a>TJPF_RGB&#160;</td><td class="fielddoc"><p>RGB pixel format. </p>
+<p>The red, green, and blue components in the image are stored in 3-byte pixels in the order R, G, B from lowest to highest byte address within each pixel. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aab10624437fb8ef495a0b153e65749839"></a>TJPF_BGR&#160;</td><td class="fielddoc"><p>BGR pixel format. </p>
+<p>The red, green, and blue components in the image are stored in 3-byte pixels in the order B, G, R from lowest to highest byte address within each pixel. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aa83973bebb7e2dc6fa8bae89ff3f42e01"></a>TJPF_RGBX&#160;</td><td class="fielddoc"><p>RGBX pixel format. </p>
+<p>The red, green, and blue components in the image are stored in 4-byte pixels in the order R, G, B from lowest to highest byte address within each pixel. The X component is ignored when compressing and undefined when decompressing. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aa2a1fbf569ca79897eae886e3376ca4c8"></a>TJPF_BGRX&#160;</td><td class="fielddoc"><p>BGRX pixel format. </p>
+<p>The red, green, and blue components in the image are stored in 4-byte pixels in the order B, G, R from lowest to highest byte address within each pixel. The X component is ignored when compressing and undefined when decompressing. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aaf6603b27147de47e212e75dac027b2af"></a>TJPF_XBGR&#160;</td><td class="fielddoc"><p>XBGR pixel format. </p>
+<p>The red, green, and blue components in the image are stored in 4-byte pixels in the order R, G, B from highest to lowest byte address within each pixel. The X component is ignored when compressing and undefined when decompressing. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aadae996905efcfa3b42a0bb3bea7f9d84"></a>TJPF_XRGB&#160;</td><td class="fielddoc"><p>XRGB pixel format. </p>
+<p>The red, green, and blue components in the image are stored in 4-byte pixels in the order B, G, R from highest to lowest byte address within each pixel. The X component is ignored when compressing and undefined when decompressing. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aa5431b54b015337705f13118073711a1a"></a>TJPF_GRAY&#160;</td><td class="fielddoc"><p>Grayscale pixel format. </p>
+<p>Each 1-byte pixel represents a luminance (brightness) level from 0 to 255. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aa88d2e88fab67f6503cf972e14851cc12"></a>TJPF_RGBA&#160;</td><td class="fielddoc"><p>RGBA pixel format. </p>
+<p>This is the same as <a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa83973bebb7e2dc6fa8bae89ff3f42e01">TJPF_RGBX</a>, except that when decompressing, the X component is guaranteed to be 0xFF, which can be interpreted as an opaque alpha channel. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aac037ff1845cf9b74bb81a3659c2b9fb4"></a>TJPF_BGRA&#160;</td><td class="fielddoc"><p>BGRA pixel format. </p>
+<p>This is the same as <a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa2a1fbf569ca79897eae886e3376ca4c8">TJPF_BGRX</a>, except that when decompressing, the X component is guaranteed to be 0xFF, which can be interpreted as an opaque alpha channel. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aa1ba1a7f1631dbeaa49a0a85fc4a40081"></a>TJPF_ABGR&#160;</td><td class="fielddoc"><p>ABGR pixel format. </p>
+<p>This is the same as <a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aaf6603b27147de47e212e75dac027b2af">TJPF_XBGR</a>, except that when decompressing, the X component is guaranteed to be 0xFF, which can be interpreted as an opaque alpha channel. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aae8f846ed9d9de99b6e1dfe448848765c"></a>TJPF_ARGB&#160;</td><td class="fielddoc"><p>ARGB pixel format. </p>
+<p>This is the same as <a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aadae996905efcfa3b42a0bb3bea7f9d84">TJPF_XRGB</a>, except that when decompressing, the X component is guaranteed to be 0xFF, which can be interpreted as an opaque alpha channel. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aa7f5100ec44c91994e243f1cf55553f8b"></a>TJPF_CMYK&#160;</td><td class="fielddoc"><p>CMYK pixel format. </p>
+<p>Unlike RGB, which is an additive color model used primarily for display, CMYK (Cyan/Magenta/Yellow/Key) is a subtractive color model used primarily for printing. In the CMYK color model, the value of each color component typically corresponds to an amount of cyan, magenta, yellow, or black ink that is applied to a white background. In order to convert between CMYK and RGB, it is necessary to use a color management system (CMS.) A CMS will attempt to map colors within the printer's gamut to perceptually similar colors in the display's gamut and vice versa, but the mapping is typically not 1:1 or reversible, nor can it be defined with a simple formula. Thus, such a conversion is out of scope for a codec library. However, the TurboJPEG API allows for compressing CMYK pixels into a YCCK JPEG image (see <a class="el" href="group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a53839e0fe867b76b58d16b0a1a7c598e" title="YCCK colorspace.">TJCS_YCCK</a>) and decompressing YCCK JPEG images into CMYK pixels. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="ggac916144e26c3817ac514e64ae5d12e2aa84c1a6cead7952998e2fb895844a21ed"></a>TJPF_UNKNOWN&#160;</td><td class="fielddoc"><p>Unknown pixel format. </p>
+<p>Currently this is only used by <a class="el" href="group___turbo_j_p_e_g.html#gaffbd83c375e79f5db4b5c5d8ad4466e7" title="Load an uncompressed image from disk into memory.">tjLoadImage()</a>. </p>
+</td></tr>
+</table>
+
+</div>
+</div>
+<a id="ga1d047060ea80bb9820d540bb928e9074"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga1d047060ea80bb9820d540bb928e9074">&#9670;&nbsp;</a></span>TJSAMP</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">enum <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">TJSAMP</a></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Chrominance subsampling options. </p>
+<p>When pixels are converted from RGB to YCbCr (see <a class="el" href="group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a7389b8f65bb387ffedce3efd0d78ec75" title="YCbCr colorspace.">TJCS_YCbCr</a>) or from CMYK to YCCK (see <a class="el" href="group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a53839e0fe867b76b58d16b0a1a7c598e" title="YCCK colorspace.">TJCS_YCCK</a>) as part of the JPEG compression process, some of the Cb and Cr (chrominance) components can be discarded or averaged together to produce a smaller image with little perceptible loss of image clarity (the human eye is more sensitive to small changes in brightness than to small changes in color.) This is called "chrominance subsampling". </p>
+<table class="fieldtable">
+<tr><th colspan="2">Enumerator</th></tr><tr><td class="fieldname"><a id="gga1d047060ea80bb9820d540bb928e9074afb8da4f44197837bdec0a4f593dacae3"></a>TJSAMP_444&#160;</td><td class="fielddoc"><p>4:4:4 chrominance subsampling (no chrominance subsampling). </p>
+<p>The JPEG or YUV image will contain one chrominance component for every pixel in the source image. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga1d047060ea80bb9820d540bb928e9074a136130902cc578f11f32429b59368404"></a>TJSAMP_422&#160;</td><td class="fielddoc"><p>4:2:2 chrominance subsampling. </p>
+<p>The JPEG or YUV image will contain one chrominance component for every 2x1 block of pixels in the source image. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga1d047060ea80bb9820d540bb928e9074a63085dbf683cfe39e513cdb6343e3737"></a>TJSAMP_420&#160;</td><td class="fielddoc"><p>4:2:0 chrominance subsampling. </p>
+<p>The JPEG or YUV image will contain one chrominance component for every 2x2 block of pixels in the source image. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga1d047060ea80bb9820d540bb928e9074a3f1c9504842ddc7a48d0f690754b6248"></a>TJSAMP_GRAY&#160;</td><td class="fielddoc"><p>Grayscale. </p>
+<p>The JPEG or YUV image will contain no chrominance components. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga1d047060ea80bb9820d540bb928e9074accf740e6f3aa6ba20ba922cad13cb974"></a>TJSAMP_440&#160;</td><td class="fielddoc"><p>4:4:0 chrominance subsampling. </p>
+<p>The JPEG or YUV image will contain one chrominance component for every 1x2 block of pixels in the source image.</p>
+<dl class="section note"><dt>Note</dt><dd>4:4:0 subsampling is not fully accelerated in libjpeg-turbo. </dd></dl>
+</td></tr>
+<tr><td class="fieldname"><a id="gga1d047060ea80bb9820d540bb928e9074a28ec62575e5ea295c3fde3001dc628e2"></a>TJSAMP_411&#160;</td><td class="fielddoc"><p>4:1:1 chrominance subsampling. </p>
+<p>The JPEG or YUV image will contain one chrominance component for every 4x1 block of pixels in the source image. JPEG images compressed with 4:1:1 subsampling will be almost exactly the same size as those compressed with 4:2:0 subsampling, and in the aggregate, both subsampling methods produce approximately the same perceptual quality. However, 4:1:1 is better able to reproduce sharp horizontal features.</p>
+<dl class="section note"><dt>Note</dt><dd>4:1:1 subsampling is not fully accelerated in libjpeg-turbo. </dd></dl>
+</td></tr>
+</table>
+
+</div>
+</div>
+<a id="ga2de531af4e7e6c4f124908376b354866"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga2de531af4e7e6c4f124908376b354866">&#9670;&nbsp;</a></span>TJXOP</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">enum <a class="el" href="group___turbo_j_p_e_g.html#ga2de531af4e7e6c4f124908376b354866">TJXOP</a></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Transform operations for <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a> </p>
+<table class="fieldtable">
+<tr><th colspan="2">Enumerator</th></tr><tr><td class="fieldname"><a id="gga2de531af4e7e6c4f124908376b354866aad88c0366cd3f7d0eac9d7a3fa1c2c27"></a>TJXOP_NONE&#160;</td><td class="fielddoc"><p>Do not transform the position of the image pixels. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga2de531af4e7e6c4f124908376b354866aa0df69776caa30f0fa28e26332d311ce"></a>TJXOP_HFLIP&#160;</td><td class="fielddoc"><p>Flip (mirror) image horizontally. </p>
+<p>This transform is imperfect if there are any partial MCU blocks on the right edge (see <a class="el" href="group___turbo_j_p_e_g.html#ga50e03cb5ed115330e212417429600b00" title="This option will cause tjTransform() to return an error if the transform is not perfect.">TJXOPT_PERFECT</a>.) </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga2de531af4e7e6c4f124908376b354866a324eddfbec53b7e691f61e56929d0d5d"></a>TJXOP_VFLIP&#160;</td><td class="fielddoc"><p>Flip (mirror) image vertically. </p>
+<p>This transform is imperfect if there are any partial MCU blocks on the bottom edge (see <a class="el" href="group___turbo_j_p_e_g.html#ga50e03cb5ed115330e212417429600b00" title="This option will cause tjTransform() to return an error if the transform is not perfect.">TJXOPT_PERFECT</a>.) </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga2de531af4e7e6c4f124908376b354866a31060aed199f886afdd417f80499c32d"></a>TJXOP_TRANSPOSE&#160;</td><td class="fielddoc"><p>Transpose image (flip/mirror along upper left to lower right axis.) This transform is always perfect. </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga2de531af4e7e6c4f124908376b354866af3b14d488aea6ece9e5b3df73a74d6a4"></a>TJXOP_TRANSVERSE&#160;</td><td class="fielddoc"><p>Transverse transpose image (flip/mirror along upper right to lower left axis.) This transform is imperfect if there are any partial MCU blocks in the image (see <a class="el" href="group___turbo_j_p_e_g.html#ga50e03cb5ed115330e212417429600b00" title="This option will cause tjTransform() to return an error if the transform is not perfect.">TJXOPT_PERFECT</a>.) </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga2de531af4e7e6c4f124908376b354866a43b2bbb23bc4bd548422d43fbe9af128"></a>TJXOP_ROT90&#160;</td><td class="fielddoc"><p>Rotate image clockwise by 90 degrees. </p>
+<p>This transform is imperfect if there are any partial MCU blocks on the bottom edge (see <a class="el" href="group___turbo_j_p_e_g.html#ga50e03cb5ed115330e212417429600b00" title="This option will cause tjTransform() to return an error if the transform is not perfect.">TJXOPT_PERFECT</a>.) </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga2de531af4e7e6c4f124908376b354866a140952eb8dd0300accfcc22726d69692"></a>TJXOP_ROT180&#160;</td><td class="fielddoc"><p>Rotate image 180 degrees. </p>
+<p>This transform is imperfect if there are any partial MCU blocks in the image (see <a class="el" href="group___turbo_j_p_e_g.html#ga50e03cb5ed115330e212417429600b00" title="This option will cause tjTransform() to return an error if the transform is not perfect.">TJXOPT_PERFECT</a>.) </p>
+</td></tr>
+<tr><td class="fieldname"><a id="gga2de531af4e7e6c4f124908376b354866a3064ee5dfb7f032df332818587567a08"></a>TJXOP_ROT270&#160;</td><td class="fielddoc"><p>Rotate image counter-clockwise by 90 degrees. </p>
+<p>This transform is imperfect if there are any partial MCU blocks on the right edge (see <a class="el" href="group___turbo_j_p_e_g.html#ga50e03cb5ed115330e212417429600b00" title="This option will cause tjTransform() to return an error if the transform is not perfect.">TJXOPT_PERFECT</a>.) </p>
+</td></tr>
+</table>
+
+</div>
+</div>
+<h2 class="groupheader">Function Documentation</h2>
+<a id="gaec627dd4c5f30b7a775a7aea3bec5d83"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gaec627dd4c5f30b7a775a7aea3bec5d83">&#9670;&nbsp;</a></span>tjAlloc()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT unsigned char* tjAlloc </td>
+          <td>(</td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>bytes</em></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Allocate an image buffer for use with TurboJPEG. </p>
+<p>You should always use this function to allocate the JPEG destination buffer(s) for the compression and transform functions unless you are disabling automatic buffer (re)allocation (by setting <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a>.)</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">bytes</td><td>the number of bytes to allocate</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>a pointer to a newly-allocated buffer with the specified number of bytes.</dd></dl>
+<dl class="section see"><dt>See also</dt><dd><a class="el" href="group___turbo_j_p_e_g.html#gaea863d2da0cdb609563aabdf9196514b" title="Free an image buffer previously allocated by TurboJPEG.">tjFree()</a> </dd></dl>
+
+</div>
+</div>
+<a id="ga67ac12fee79073242cb216e07c9f1f90"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga67ac12fee79073242cb216e07c9f1f90">&#9670;&nbsp;</a></span>tjBufSize()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT unsigned long tjBufSize </td>
+          <td>(</td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>jpegSubsamp</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The maximum size of the buffer (in bytes) required to hold a JPEG image with the given parameters. </p>
+<p>The number of bytes returned by this function is larger than the size of the uncompressed source image. The reason for this is that the JPEG format uses 16-bit coefficients, and it is thus possible for a very high-quality JPEG image with very high-frequency content to expand rather than compress when converted to the JPEG format. Such images represent a very rare corner case, but since there is no way to predict the size of a JPEG image prior to compression, the corner case has to be handled.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">width</td><td>width (in pixels) of the image</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the image</td></tr>
+    <tr><td class="paramname">jpegSubsamp</td><td>the level of chrominance subsampling to be used when generating the JPEG image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>the maximum size of the buffer (in bytes) required to hold the image, or -1 if the arguments are out of bounds. </dd></dl>
+
+</div>
+</div>
+<a id="ga2be2b9969d4df9ecce9b05deed273194"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga2be2b9969d4df9ecce9b05deed273194">&#9670;&nbsp;</a></span>tjBufSizeYUV2()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT unsigned long tjBufSizeYUV2 </td>
+          <td>(</td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pad</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The size of the buffer (in bytes) required to hold a YUV planar image with the given parameters. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">width</td><td>width (in pixels) of the image</td></tr>
+    <tr><td class="paramname">pad</td><td>the width of each line in each plane of the image is padded to the nearest multiple of this number of bytes (must be a power of 2.)</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the image</td></tr>
+    <tr><td class="paramname">subsamp</td><td>level of chrominance subsampling in the image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>the size of the buffer (in bytes) required to hold the image, or -1 if the arguments are out of bounds. </dd></dl>
+
+</div>
+</div>
+<a id="gafbdce0112fd78fd38efae841443a9bcf"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gafbdce0112fd78fd38efae841443a9bcf">&#9670;&nbsp;</a></span>tjCompress2()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjCompress2 </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>srcBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pitch</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pixelFormat</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char **&#160;</td>
+          <td class="paramname"><em>jpegBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned long *&#160;</td>
+          <td class="paramname"><em>jpegSize</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>jpegSubsamp</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>jpegQual</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Compress an RGB, grayscale, or CMYK image into a JPEG image. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG compressor or transformer instance</td></tr>
+    <tr><td class="paramname">srcBuf</td><td>pointer to an image buffer containing RGB, grayscale, or CMYK pixels to be compressed</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the source image</td></tr>
+    <tr><td class="paramname">pitch</td><td>bytes per line in the source image. Normally, this should be <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code> if the image is unpadded, or <code><a class="el" href="group___turbo_j_p_e_g.html#ga0aba955473315e405295d978f0c16511" title="Pad the given width to the nearest 32-bit boundary.">TJPAD</a>(width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat)</code> if each line of the image is padded to the nearest 32-bit boundary, as is the case for Windows bitmaps. You can also be clever and use this parameter to skip lines, etc. Setting this parameter to 0 is the equivalent of setting it to <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code>.</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the source image</td></tr>
+    <tr><td class="paramname">pixelFormat</td><td>pixel format of the source image (see <a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">Pixel formats</a>.)</td></tr>
+    <tr><td class="paramname">jpegBuf</td><td>address of a pointer to an image buffer that will receive the JPEG image. TurboJPEG has the ability to reallocate the JPEG buffer to accommodate the size of the JPEG image. Thus, you can choose to:<ol type="1">
+<li>pre-allocate the JPEG buffer with an arbitrary size using <a class="el" href="group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83" title="Allocate an image buffer for use with TurboJPEG.">tjAlloc()</a> and let TurboJPEG grow the buffer as needed,</li>
+<li>set <code>*jpegBuf</code> to NULL to tell TurboJPEG to allocate the buffer for you, or</li>
+<li>pre-allocate the buffer to a "worst case" size determined by calling <a class="el" href="group___turbo_j_p_e_g.html#ga67ac12fee79073242cb216e07c9f1f90" title="The maximum size of the buffer (in bytes) required to hold a JPEG image with the given parameters.">tjBufSize()</a>. This should ensure that the buffer never has to be re-allocated (setting <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a> guarantees that it won't be.)</li>
+</ol>
+If you choose option 1, <code>*jpegSize</code> should be set to the size of your pre-allocated buffer. In any case, unless you have set <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a>, you should always check <code>*jpegBuf</code> upon return from this function, as it may have changed.</td></tr>
+    <tr><td class="paramname">jpegSize</td><td>pointer to an unsigned long variable that holds the size of the JPEG image buffer. If <code>*jpegBuf</code> points to a pre-allocated buffer, then <code>*jpegSize</code> should be set to the size of the buffer. Upon return, <code>*jpegSize</code> will contain the size of the JPEG image (in bytes.) If <code>*jpegBuf</code> points to a JPEG image buffer that is being reused from a previous call to one of the JPEG compression functions, then <code>*jpegSize</code> is ignored.</td></tr>
+    <tr><td class="paramname">jpegSubsamp</td><td>the level of chrominance subsampling to be used when generating the JPEG image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+    <tr><td class="paramname">jpegQual</td><td>the image quality of the generated JPEG image (1 = worst, 100 = best)</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga7622a459b79aa1007e005b58783f875b"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga7622a459b79aa1007e005b58783f875b">&#9670;&nbsp;</a></span>tjCompressFromYUV()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjCompressFromYUV </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>srcBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pad</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char **&#160;</td>
+          <td class="paramname"><em>jpegBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned long *&#160;</td>
+          <td class="paramname"><em>jpegSize</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>jpegQual</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Compress a YUV planar image into a JPEG image. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG compressor or transformer instance</td></tr>
+    <tr><td class="paramname">srcBuf</td><td>pointer to an image buffer containing a YUV planar image to be compressed. The size of this buffer should match the value returned by <a class="el" href="group___turbo_j_p_e_g.html#ga2be2b9969d4df9ecce9b05deed273194" title="The size of the buffer (in bytes) required to hold a YUV planar image with the given parameters.">tjBufSizeYUV2()</a> for the given image width, height, padding, and level of chrominance subsampling. The Y, U (Cb), and V (Cr) image planes should be stored sequentially in the source buffer (refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a>.)</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the source image. If the width is not an even multiple of the MCU block width (see <a class="el" href="group___turbo_j_p_e_g.html#ga9e61e7cd47a15a173283ba94e781308c" title="MCU block width (in pixels) for a given level of chrominance subsampling.">tjMCUWidth</a>), then an intermediate buffer copy will be performed within TurboJPEG.</td></tr>
+    <tr><td class="paramname">pad</td><td>the line padding used in the source image. For instance, if each line in each plane of the YUV image is padded to the nearest multiple of 4 bytes, then <code>pad</code> should be set to 4.</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the source image. If the height is not an even multiple of the MCU block height (see <a class="el" href="group___turbo_j_p_e_g.html#gabd247bb9fecb393eca57366feb8327bf" title="MCU block height (in pixels) for a given level of chrominance subsampling.">tjMCUHeight</a>), then an intermediate buffer copy will be performed within TurboJPEG.</td></tr>
+    <tr><td class="paramname">subsamp</td><td>the level of chrominance subsampling used in the source image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+    <tr><td class="paramname">jpegBuf</td><td>address of a pointer to an image buffer that will receive the JPEG image. TurboJPEG has the ability to reallocate the JPEG buffer to accommodate the size of the JPEG image. Thus, you can choose to:<ol type="1">
+<li>pre-allocate the JPEG buffer with an arbitrary size using <a class="el" href="group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83" title="Allocate an image buffer for use with TurboJPEG.">tjAlloc()</a> and let TurboJPEG grow the buffer as needed,</li>
+<li>set <code>*jpegBuf</code> to NULL to tell TurboJPEG to allocate the buffer for you, or</li>
+<li>pre-allocate the buffer to a "worst case" size determined by calling <a class="el" href="group___turbo_j_p_e_g.html#ga67ac12fee79073242cb216e07c9f1f90" title="The maximum size of the buffer (in bytes) required to hold a JPEG image with the given parameters.">tjBufSize()</a>. This should ensure that the buffer never has to be re-allocated (setting <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a> guarantees that it won't be.)</li>
+</ol>
+If you choose option 1, <code>*jpegSize</code> should be set to the size of your pre-allocated buffer. In any case, unless you have set <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a>, you should always check <code>*jpegBuf</code> upon return from this function, as it may have changed.</td></tr>
+    <tr><td class="paramname">jpegSize</td><td>pointer to an unsigned long variable that holds the size of the JPEG image buffer. If <code>*jpegBuf</code> points to a pre-allocated buffer, then <code>*jpegSize</code> should be set to the size of the buffer. Upon return, <code>*jpegSize</code> will contain the size of the JPEG image (in bytes.) If <code>*jpegBuf</code> points to a JPEG image buffer that is being reused from a previous call to one of the JPEG compression functions, then <code>*jpegSize</code> is ignored.</td></tr>
+    <tr><td class="paramname">jpegQual</td><td>the image quality of the generated JPEG image (1 = worst, 100 = best)</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga29ec5dfbd2d84b8724e951d6fa0d5d9e"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga29ec5dfbd2d84b8724e951d6fa0d5d9e">&#9670;&nbsp;</a></span>tjCompressFromYUVPlanes()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjCompressFromYUVPlanes </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char **&#160;</td>
+          <td class="paramname"><em>srcPlanes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const int *&#160;</td>
+          <td class="paramname"><em>strides</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char **&#160;</td>
+          <td class="paramname"><em>jpegBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned long *&#160;</td>
+          <td class="paramname"><em>jpegSize</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>jpegQual</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Compress a set of Y, U (Cb), and V (Cr) image planes into a JPEG image. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG compressor or transformer instance</td></tr>
+    <tr><td class="paramname">srcPlanes</td><td>an array of pointers to Y, U (Cb), and V (Cr) image planes (or just a Y plane, if compressing a grayscale image) that contain a YUV image to be compressed. These planes can be contiguous or non-contiguous in memory. The size of each plane should match the value returned by <a class="el" href="group___turbo_j_p_e_g.html#gab4ab7b24f6e797d79abaaa670373961d" title="The size of the buffer (in bytes) required to hold a YUV image plane with the given parameters.">tjPlaneSizeYUV()</a> for the given image width, height, strides, and level of chrominance subsampling. Refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a> for more details.</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the source image. If the width is not an even multiple of the MCU block width (see <a class="el" href="group___turbo_j_p_e_g.html#ga9e61e7cd47a15a173283ba94e781308c" title="MCU block width (in pixels) for a given level of chrominance subsampling.">tjMCUWidth</a>), then an intermediate buffer copy will be performed within TurboJPEG.</td></tr>
+    <tr><td class="paramname">strides</td><td>an array of integers, each specifying the number of bytes per line in the corresponding plane of the YUV source image. Setting the stride for any plane to 0 is the same as setting it to the plane width (see <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a>.) If <code>strides</code> is NULL, then the strides for all planes will be set to their respective plane widths. You can adjust the strides in order to specify an arbitrary amount of line padding in each plane or to create a JPEG image from a subregion of a larger YUV planar image.</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the source image. If the height is not an even multiple of the MCU block height (see <a class="el" href="group___turbo_j_p_e_g.html#gabd247bb9fecb393eca57366feb8327bf" title="MCU block height (in pixels) for a given level of chrominance subsampling.">tjMCUHeight</a>), then an intermediate buffer copy will be performed within TurboJPEG.</td></tr>
+    <tr><td class="paramname">subsamp</td><td>the level of chrominance subsampling used in the source image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+    <tr><td class="paramname">jpegBuf</td><td>address of a pointer to an image buffer that will receive the JPEG image. TurboJPEG has the ability to reallocate the JPEG buffer to accommodate the size of the JPEG image. Thus, you can choose to:<ol type="1">
+<li>pre-allocate the JPEG buffer with an arbitrary size using <a class="el" href="group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83" title="Allocate an image buffer for use with TurboJPEG.">tjAlloc()</a> and let TurboJPEG grow the buffer as needed,</li>
+<li>set <code>*jpegBuf</code> to NULL to tell TurboJPEG to allocate the buffer for you, or</li>
+<li>pre-allocate the buffer to a "worst case" size determined by calling <a class="el" href="group___turbo_j_p_e_g.html#ga67ac12fee79073242cb216e07c9f1f90" title="The maximum size of the buffer (in bytes) required to hold a JPEG image with the given parameters.">tjBufSize()</a>. This should ensure that the buffer never has to be re-allocated (setting <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a> guarantees that it won't be.)</li>
+</ol>
+If you choose option 1, <code>*jpegSize</code> should be set to the size of your pre-allocated buffer. In any case, unless you have set <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a>, you should always check <code>*jpegBuf</code> upon return from this function, as it may have changed.</td></tr>
+    <tr><td class="paramname">jpegSize</td><td>pointer to an unsigned long variable that holds the size of the JPEG image buffer. If <code>*jpegBuf</code> points to a pre-allocated buffer, then <code>*jpegSize</code> should be set to the size of the buffer. Upon return, <code>*jpegSize</code> will contain the size of the JPEG image (in bytes.) If <code>*jpegBuf</code> points to a JPEG image buffer that is being reused from a previous call to one of the JPEG compression functions, then <code>*jpegSize</code> is ignored.</td></tr>
+    <tr><td class="paramname">jpegQual</td><td>the image quality of the generated JPEG image (1 = worst, 100 = best)</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga70abbf38f77a26fd6da8813bef96f695"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga70abbf38f77a26fd6da8813bef96f695">&#9670;&nbsp;</a></span>tjDecodeYUV()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjDecodeYUV </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>srcBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pad</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char *&#160;</td>
+          <td class="paramname"><em>dstBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pitch</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pixelFormat</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Decode a YUV planar image into an RGB or grayscale image. </p>
+<p>This function uses the accelerated color conversion routines in the underlying codec but does not execute any of the other steps in the JPEG decompression process.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG decompressor or transformer instance</td></tr>
+    <tr><td class="paramname">srcBuf</td><td>pointer to an image buffer containing a YUV planar image to be decoded. The size of this buffer should match the value returned by <a class="el" href="group___turbo_j_p_e_g.html#ga2be2b9969d4df9ecce9b05deed273194" title="The size of the buffer (in bytes) required to hold a YUV planar image with the given parameters.">tjBufSizeYUV2()</a> for the given image width, height, padding, and level of chrominance subsampling. The Y, U (Cb), and V (Cr) image planes should be stored sequentially in the source buffer (refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a>.)</td></tr>
+    <tr><td class="paramname">pad</td><td>Use this parameter to specify that the width of each line in each plane of the YUV source image is padded to the nearest multiple of this number of bytes (must be a power of 2.)</td></tr>
+    <tr><td class="paramname">subsamp</td><td>the level of chrominance subsampling used in the YUV source image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+    <tr><td class="paramname">dstBuf</td><td>pointer to an image buffer that will receive the decoded image. This buffer should normally be <code>pitch * height</code> bytes in size, but the <code>dstBuf</code> pointer can also be used to decode into a specific region of a larger buffer.</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the source and destination images</td></tr>
+    <tr><td class="paramname">pitch</td><td>bytes per line in the destination image. Normally, this should be <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code> if the destination image is unpadded, or <code><a class="el" href="group___turbo_j_p_e_g.html#ga0aba955473315e405295d978f0c16511" title="Pad the given width to the nearest 32-bit boundary.">TJPAD</a>(width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat)</code> if each line of the destination image should be padded to the nearest 32-bit boundary, as is the case for Windows bitmaps. You can also be clever and use the pitch parameter to skip lines, etc. Setting this parameter to 0 is the equivalent of setting it to <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code>.</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the source and destination images</td></tr>
+    <tr><td class="paramname">pixelFormat</td><td>pixel format of the destination image (see <a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">Pixel formats</a>.)</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga10e837c07fa9d25770565b237d3898d9"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga10e837c07fa9d25770565b237d3898d9">&#9670;&nbsp;</a></span>tjDecodeYUVPlanes()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjDecodeYUVPlanes </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char **&#160;</td>
+          <td class="paramname"><em>srcPlanes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const int *&#160;</td>
+          <td class="paramname"><em>strides</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char *&#160;</td>
+          <td class="paramname"><em>dstBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pitch</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pixelFormat</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Decode a set of Y, U (Cb), and V (Cr) image planes into an RGB or grayscale image. </p>
+<p>This function uses the accelerated color conversion routines in the underlying codec but does not execute any of the other steps in the JPEG decompression process.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG decompressor or transformer instance</td></tr>
+    <tr><td class="paramname">srcPlanes</td><td>an array of pointers to Y, U (Cb), and V (Cr) image planes (or just a Y plane, if decoding a grayscale image) that contain a YUV image to be decoded. These planes can be contiguous or non-contiguous in memory. The size of each plane should match the value returned by <a class="el" href="group___turbo_j_p_e_g.html#gab4ab7b24f6e797d79abaaa670373961d" title="The size of the buffer (in bytes) required to hold a YUV image plane with the given parameters.">tjPlaneSizeYUV()</a> for the given image width, height, strides, and level of chrominance subsampling. Refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a> for more details.</td></tr>
+    <tr><td class="paramname">strides</td><td>an array of integers, each specifying the number of bytes per line in the corresponding plane of the YUV source image. Setting the stride for any plane to 0 is the same as setting it to the plane width (see <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a>.) If <code>strides</code> is NULL, then the strides for all planes will be set to their respective plane widths. You can adjust the strides in order to specify an arbitrary amount of line padding in each plane or to decode a subregion of a larger YUV planar image.</td></tr>
+    <tr><td class="paramname">subsamp</td><td>the level of chrominance subsampling used in the YUV source image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+    <tr><td class="paramname">dstBuf</td><td>pointer to an image buffer that will receive the decoded image. This buffer should normally be <code>pitch * height</code> bytes in size, but the <code>dstBuf</code> pointer can also be used to decode into a specific region of a larger buffer.</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the source and destination images</td></tr>
+    <tr><td class="paramname">pitch</td><td>bytes per line in the destination image. Normally, this should be <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code> if the destination image is unpadded, or <code><a class="el" href="group___turbo_j_p_e_g.html#ga0aba955473315e405295d978f0c16511" title="Pad the given width to the nearest 32-bit boundary.">TJPAD</a>(width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat)</code> if each line of the destination image should be padded to the nearest 32-bit boundary, as is the case for Windows bitmaps. You can also be clever and use the pitch parameter to skip lines, etc. Setting this parameter to 0 is the equivalent of setting it to <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code>.</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the source and destination images</td></tr>
+    <tr><td class="paramname">pixelFormat</td><td>pixel format of the destination image (see <a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">Pixel formats</a>.)</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="gae9eccef8b682a48f43a9117c231ed013"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gae9eccef8b682a48f43a9117c231ed013">&#9670;&nbsp;</a></span>tjDecompress2()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjDecompress2 </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>jpegBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned long&#160;</td>
+          <td class="paramname"><em>jpegSize</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char *&#160;</td>
+          <td class="paramname"><em>dstBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pitch</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pixelFormat</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Decompress a JPEG image to an RGB, grayscale, or CMYK image. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG decompressor or transformer instance</td></tr>
+    <tr><td class="paramname">jpegBuf</td><td>pointer to a buffer containing the JPEG image to decompress</td></tr>
+    <tr><td class="paramname">jpegSize</td><td>size of the JPEG image (in bytes)</td></tr>
+    <tr><td class="paramname">dstBuf</td><td>pointer to an image buffer that will receive the decompressed image. This buffer should normally be <code>pitch * scaledHeight</code> bytes in size, where <code>scaledHeight</code> can be determined by calling <a class="el" href="group___turbo_j_p_e_g.html#ga84878bb65404204743aa18cac02781df" title="Compute the scaled value of dimension using the given scaling factor.">TJSCALED()</a> with the JPEG image height and one of the scaling factors returned by <a class="el" href="group___turbo_j_p_e_g.html#gac3854476006b10787bd128f7ede48057" title="Returns a list of fractional scaling factors that the JPEG decompressor in this implementation of Tur...">tjGetScalingFactors()</a>. The <code>dstBuf</code> pointer may also be used to decompress into a specific region of a larger buffer.</td></tr>
+    <tr><td class="paramname">width</td><td>desired width (in pixels) of the destination image. If this is different than the width of the JPEG image being decompressed, then TurboJPEG will use scaling in the JPEG decompressor to generate the largest possible image that will fit within the desired width. If <code>width</code> is set to 0, then only the height will be considered when determining the scaled image size.</td></tr>
+    <tr><td class="paramname">pitch</td><td>bytes per line in the destination image. Normally, this is <code>scaledWidth * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code> if the decompressed image is unpadded, else <code><a class="el" href="group___turbo_j_p_e_g.html#ga0aba955473315e405295d978f0c16511" title="Pad the given width to the nearest 32-bit boundary.">TJPAD</a>(scaledWidth * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat)</code> if each line of the decompressed image is padded to the nearest 32-bit boundary, as is the case for Windows bitmaps. (NOTE: <code>scaledWidth</code> can be determined by calling <a class="el" href="group___turbo_j_p_e_g.html#ga84878bb65404204743aa18cac02781df" title="Compute the scaled value of dimension using the given scaling factor.">TJSCALED()</a> with the JPEG image width and one of the scaling factors returned by <a class="el" href="group___turbo_j_p_e_g.html#gac3854476006b10787bd128f7ede48057" title="Returns a list of fractional scaling factors that the JPEG decompressor in this implementation of Tur...">tjGetScalingFactors()</a>.) You can also be clever and use the pitch parameter to skip lines, etc. Setting this parameter to 0 is the equivalent of setting it to <code>scaledWidth * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code>.</td></tr>
+    <tr><td class="paramname">height</td><td>desired height (in pixels) of the destination image. If this is different than the height of the JPEG image being decompressed, then TurboJPEG will use scaling in the JPEG decompressor to generate the largest possible image that will fit within the desired height. If <code>height</code> is set to 0, then only the width will be considered when determining the scaled image size.</td></tr>
+    <tr><td class="paramname">pixelFormat</td><td>pixel format of the destination image (see <a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">Pixel formats</a>.)</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga0595681096bba7199cc6f3533cb25f77"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga0595681096bba7199cc6f3533cb25f77">&#9670;&nbsp;</a></span>tjDecompressHeader3()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjDecompressHeader3 </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>jpegBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned long&#160;</td>
+          <td class="paramname"><em>jpegSize</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>jpegSubsamp</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>jpegColorspace</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Retrieve information about a JPEG image without decompressing it. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG decompressor or transformer instance</td></tr>
+    <tr><td class="paramname">jpegBuf</td><td>pointer to a buffer containing a JPEG image</td></tr>
+    <tr><td class="paramname">jpegSize</td><td>size of the JPEG image (in bytes)</td></tr>
+    <tr><td class="paramname">width</td><td>pointer to an integer variable that will receive the width (in pixels) of the JPEG image</td></tr>
+    <tr><td class="paramname">height</td><td>pointer to an integer variable that will receive the height (in pixels) of the JPEG image</td></tr>
+    <tr><td class="paramname">jpegSubsamp</td><td>pointer to an integer variable that will receive the level of chrominance subsampling used when the JPEG image was compressed (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+    <tr><td class="paramname">jpegColorspace</td><td>pointer to an integer variable that will receive one of the JPEG colorspace constants, indicating the colorspace of the JPEG image (see <a class="el" href="group___turbo_j_p_e_g.html#ga4f83ad3368e0e29d1957be0efa7c3720">JPEG colorspaces</a>.)</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga04d1e839ff9a0860dd1475cff78d3364"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga04d1e839ff9a0860dd1475cff78d3364">&#9670;&nbsp;</a></span>tjDecompressToYUV2()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjDecompressToYUV2 </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>jpegBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned long&#160;</td>
+          <td class="paramname"><em>jpegSize</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char *&#160;</td>
+          <td class="paramname"><em>dstBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pad</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Decompress a JPEG image to a YUV planar image. </p>
+<p>This function performs JPEG decompression but leaves out the color conversion step, so a planar YUV image is generated instead of an RGB image.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG decompressor or transformer instance</td></tr>
+    <tr><td class="paramname">jpegBuf</td><td>pointer to a buffer containing the JPEG image to decompress</td></tr>
+    <tr><td class="paramname">jpegSize</td><td>size of the JPEG image (in bytes)</td></tr>
+    <tr><td class="paramname">dstBuf</td><td>pointer to an image buffer that will receive the YUV image. Use <a class="el" href="group___turbo_j_p_e_g.html#ga2be2b9969d4df9ecce9b05deed273194" title="The size of the buffer (in bytes) required to hold a YUV planar image with the given parameters.">tjBufSizeYUV2()</a> to determine the appropriate size for this buffer based on the image width, height, padding, and level of subsampling. The Y, U (Cb), and V (Cr) image planes will be stored sequentially in the buffer (refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a>.)</td></tr>
+    <tr><td class="paramname">width</td><td>desired width (in pixels) of the YUV image. If this is different than the width of the JPEG image being decompressed, then TurboJPEG will use scaling in the JPEG decompressor to generate the largest possible image that will fit within the desired width. If <code>width</code> is set to 0, then only the height will be considered when determining the scaled image size. If the scaled width is not an even multiple of the MCU block width (see <a class="el" href="group___turbo_j_p_e_g.html#ga9e61e7cd47a15a173283ba94e781308c" title="MCU block width (in pixels) for a given level of chrominance subsampling.">tjMCUWidth</a>), then an intermediate buffer copy will be performed within TurboJPEG.</td></tr>
+    <tr><td class="paramname">pad</td><td>the width of each line in each plane of the YUV image will be padded to the nearest multiple of this number of bytes (must be a power of 2.) To generate images suitable for X Video, <code>pad</code> should be set to 4.</td></tr>
+    <tr><td class="paramname">height</td><td>desired height (in pixels) of the YUV image. If this is different than the height of the JPEG image being decompressed, then TurboJPEG will use scaling in the JPEG decompressor to generate the largest possible image that will fit within the desired height. If <code>height</code> is set to 0, then only the width will be considered when determining the scaled image size. If the scaled height is not an even multiple of the MCU block height (see <a class="el" href="group___turbo_j_p_e_g.html#gabd247bb9fecb393eca57366feb8327bf" title="MCU block height (in pixels) for a given level of chrominance subsampling.">tjMCUHeight</a>), then an intermediate buffer copy will be performed within TurboJPEG.</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="gaa59f901a5258ada5bd0185ad59368540"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gaa59f901a5258ada5bd0185ad59368540">&#9670;&nbsp;</a></span>tjDecompressToYUVPlanes()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjDecompressToYUVPlanes </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>jpegBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned long&#160;</td>
+          <td class="paramname"><em>jpegSize</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char **&#160;</td>
+          <td class="paramname"><em>dstPlanes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>strides</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Decompress a JPEG image into separate Y, U (Cb), and V (Cr) image planes. </p>
+<p>This function performs JPEG decompression but leaves out the color conversion step, so a planar YUV image is generated instead of an RGB image.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG decompressor or transformer instance</td></tr>
+    <tr><td class="paramname">jpegBuf</td><td>pointer to a buffer containing the JPEG image to decompress</td></tr>
+    <tr><td class="paramname">jpegSize</td><td>size of the JPEG image (in bytes)</td></tr>
+    <tr><td class="paramname">dstPlanes</td><td>an array of pointers to Y, U (Cb), and V (Cr) image planes (or just a Y plane, if decompressing a grayscale image) that will receive the YUV image. These planes can be contiguous or non-contiguous in memory. Use <a class="el" href="group___turbo_j_p_e_g.html#gab4ab7b24f6e797d79abaaa670373961d" title="The size of the buffer (in bytes) required to hold a YUV image plane with the given parameters.">tjPlaneSizeYUV()</a> to determine the appropriate size for each plane based on the scaled image width, scaled image height, strides, and level of chrominance subsampling. Refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a> for more details.</td></tr>
+    <tr><td class="paramname">width</td><td>desired width (in pixels) of the YUV image. If this is different than the width of the JPEG image being decompressed, then TurboJPEG will use scaling in the JPEG decompressor to generate the largest possible image that will fit within the desired width. If <code>width</code> is set to 0, then only the height will be considered when determining the scaled image size. If the scaled width is not an even multiple of the MCU block width (see <a class="el" href="group___turbo_j_p_e_g.html#ga9e61e7cd47a15a173283ba94e781308c" title="MCU block width (in pixels) for a given level of chrominance subsampling.">tjMCUWidth</a>), then an intermediate buffer copy will be performed within TurboJPEG.</td></tr>
+    <tr><td class="paramname">strides</td><td>an array of integers, each specifying the number of bytes per line in the corresponding plane of the output image. Setting the stride for any plane to 0 is the same as setting it to the scaled plane width (see <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a>.) If <code>strides</code> is NULL, then the strides for all planes will be set to their respective scaled plane widths. You can adjust the strides in order to add an arbitrary amount of line padding to each plane or to decompress the JPEG image into a subregion of a larger YUV planar image.</td></tr>
+    <tr><td class="paramname">height</td><td>desired height (in pixels) of the YUV image. If this is different than the height of the JPEG image being decompressed, then TurboJPEG will use scaling in the JPEG decompressor to generate the largest possible image that will fit within the desired height. If <code>height</code> is set to 0, then only the width will be considered when determining the scaled image size. If the scaled height is not an even multiple of the MCU block height (see <a class="el" href="group___turbo_j_p_e_g.html#gabd247bb9fecb393eca57366feb8327bf" title="MCU block height (in pixels) for a given level of chrominance subsampling.">tjMCUHeight</a>), then an intermediate buffer copy will be performed within TurboJPEG.</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga75f355fa27225ba1a4ee392c852394d2"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga75f355fa27225ba1a4ee392c852394d2">&#9670;&nbsp;</a></span>tjDestroy()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjDestroy </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Destroy a TurboJPEG compressor, decompressor, or transformer instance. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG compressor, decompressor or transformer instance</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="gac519b922cdf446e97d0cdcba513636bf"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gac519b922cdf446e97d0cdcba513636bf">&#9670;&nbsp;</a></span>tjEncodeYUV3()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjEncodeYUV3 </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>srcBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pitch</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pixelFormat</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char *&#160;</td>
+          <td class="paramname"><em>dstBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pad</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Encode an RGB or grayscale image into a YUV planar image. </p>
+<p>This function uses the accelerated color conversion routines in the underlying codec but does not execute any of the other steps in the JPEG compression process.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG compressor or transformer instance</td></tr>
+    <tr><td class="paramname">srcBuf</td><td>pointer to an image buffer containing RGB or grayscale pixels to be encoded</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the source image</td></tr>
+    <tr><td class="paramname">pitch</td><td>bytes per line in the source image. Normally, this should be <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code> if the image is unpadded, or <code><a class="el" href="group___turbo_j_p_e_g.html#ga0aba955473315e405295d978f0c16511" title="Pad the given width to the nearest 32-bit boundary.">TJPAD</a>(width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat)</code> if each line of the image is padded to the nearest 32-bit boundary, as is the case for Windows bitmaps. You can also be clever and use this parameter to skip lines, etc. Setting this parameter to 0 is the equivalent of setting it to <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code>.</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the source image</td></tr>
+    <tr><td class="paramname">pixelFormat</td><td>pixel format of the source image (see <a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">Pixel formats</a>.)</td></tr>
+    <tr><td class="paramname">dstBuf</td><td>pointer to an image buffer that will receive the YUV image. Use <a class="el" href="group___turbo_j_p_e_g.html#ga2be2b9969d4df9ecce9b05deed273194" title="The size of the buffer (in bytes) required to hold a YUV planar image with the given parameters.">tjBufSizeYUV2()</a> to determine the appropriate size for this buffer based on the image width, height, padding, and level of chrominance subsampling. The Y, U (Cb), and V (Cr) image planes will be stored sequentially in the buffer (refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a>.)</td></tr>
+    <tr><td class="paramname">pad</td><td>the width of each line in each plane of the YUV image will be padded to the nearest multiple of this number of bytes (must be a power of 2.) To generate images suitable for X Video, <code>pad</code> should be set to 4.</td></tr>
+    <tr><td class="paramname">subsamp</td><td>the level of chrominance subsampling to be used when generating the YUV image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.) To generate images suitable for X Video, <code>subsamp</code> should be set to <a class="el" href="group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a63085dbf683cfe39e513cdb6343e3737">TJSAMP_420</a>. This produces an image compatible with the I420 (AKA "YUV420P") format.</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="gae2d04c72457fe7f4d60cf78ab1b1feb1"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gae2d04c72457fe7f4d60cf78ab1b1feb1">&#9670;&nbsp;</a></span>tjEncodeYUVPlanes()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjEncodeYUVPlanes </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>srcBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pitch</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pixelFormat</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char **&#160;</td>
+          <td class="paramname"><em>dstPlanes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>strides</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Encode an RGB or grayscale image into separate Y, U (Cb), and V (Cr) image planes. </p>
+<p>This function uses the accelerated color conversion routines in the underlying codec but does not execute any of the other steps in the JPEG compression process.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG compressor or transformer instance</td></tr>
+    <tr><td class="paramname">srcBuf</td><td>pointer to an image buffer containing RGB or grayscale pixels to be encoded</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the source image</td></tr>
+    <tr><td class="paramname">pitch</td><td>bytes per line in the source image. Normally, this should be <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code> if the image is unpadded, or <code><a class="el" href="group___turbo_j_p_e_g.html#ga0aba955473315e405295d978f0c16511" title="Pad the given width to the nearest 32-bit boundary.">TJPAD</a>(width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat)</code> if each line of the image is padded to the nearest 32-bit boundary, as is the case for Windows bitmaps. You can also be clever and use this parameter to skip lines, etc. Setting this parameter to 0 is the equivalent of setting it to <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code>.</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the source image</td></tr>
+    <tr><td class="paramname">pixelFormat</td><td>pixel format of the source image (see <a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">Pixel formats</a>.)</td></tr>
+    <tr><td class="paramname">dstPlanes</td><td>an array of pointers to Y, U (Cb), and V (Cr) image planes (or just a Y plane, if generating a grayscale image) that will receive the encoded image. These planes can be contiguous or non-contiguous in memory. Use <a class="el" href="group___turbo_j_p_e_g.html#gab4ab7b24f6e797d79abaaa670373961d" title="The size of the buffer (in bytes) required to hold a YUV image plane with the given parameters.">tjPlaneSizeYUV()</a> to determine the appropriate size for each plane based on the image width, height, strides, and level of chrominance subsampling. Refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a> for more details.</td></tr>
+    <tr><td class="paramname">strides</td><td>an array of integers, each specifying the number of bytes per line in the corresponding plane of the output image. Setting the stride for any plane to 0 is the same as setting it to the plane width (see <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a>.) If <code>strides</code> is NULL, then the strides for all planes will be set to their respective plane widths. You can adjust the strides in order to add an arbitrary amount of line padding to each plane or to encode an RGB or grayscale image into a subregion of a larger YUV planar image.</td></tr>
+    <tr><td class="paramname">subsamp</td><td>the level of chrominance subsampling to be used when generating the YUV image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.) To generate images suitable for X Video, <code>subsamp</code> should be set to <a class="el" href="group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a63085dbf683cfe39e513cdb6343e3737">TJSAMP_420</a>. This produces an image compatible with the I420 (AKA "YUV420P") format.</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="gaea863d2da0cdb609563aabdf9196514b"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gaea863d2da0cdb609563aabdf9196514b">&#9670;&nbsp;</a></span>tjFree()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT void tjFree </td>
+          <td>(</td>
+          <td class="paramtype">unsigned char *&#160;</td>
+          <td class="paramname"><em>buffer</em></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Free an image buffer previously allocated by TurboJPEG. </p>
+<p>You should always use this function to free JPEG destination buffer(s) that were automatically (re)allocated by the compression and transform functions or that were manually allocated using <a class="el" href="group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83" title="Allocate an image buffer for use with TurboJPEG.">tjAlloc()</a>.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">buffer</td><td>address of the buffer to free. If the address is NULL, then this function has no effect.</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section see"><dt>See also</dt><dd><a class="el" href="group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83" title="Allocate an image buffer for use with TurboJPEG.">tjAlloc()</a> </dd></dl>
+
+</div>
+</div>
+<a id="ga414feeffbf860ebd31c745df203de410"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga414feeffbf860ebd31c745df203de410">&#9670;&nbsp;</a></span>tjGetErrorCode()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjGetErrorCode </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Returns a code indicating the severity of the last error. </p>
+<p>See <a class="el" href="group___turbo_j_p_e_g.html#gafbc17cfa57d0d5d11fea35ac025950fe">Error codes</a>.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG compressor, decompressor or transformer instance</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>a code indicating the severity of the last error. See <a class="el" href="group___turbo_j_p_e_g.html#gafbc17cfa57d0d5d11fea35ac025950fe">Error codes</a>. </dd></dl>
+
+</div>
+</div>
+<a id="ga1ead8574f9f39fbafc6b497124e7aafa"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga1ead8574f9f39fbafc6b497124e7aafa">&#9670;&nbsp;</a></span>tjGetErrorStr2()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT char* tjGetErrorStr2 </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Returns a descriptive error message explaining why the last command failed. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG compressor, decompressor, or transformer instance, or NULL if the error was generated by a global function (but note that retrieving the error message for a global function is thread-safe only on platforms that support thread-local storage.)</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>a descriptive error message explaining why the last command failed. </dd></dl>
+
+</div>
+</div>
+<a id="gac3854476006b10787bd128f7ede48057"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gac3854476006b10787bd128f7ede48057">&#9670;&nbsp;</a></span>tjGetScalingFactors()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT <a class="el" href="structtjscalingfactor.html">tjscalingfactor</a>* tjGetScalingFactors </td>
+          <td>(</td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>numscalingfactors</em></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Returns a list of fractional scaling factors that the JPEG decompressor in this implementation of TurboJPEG supports. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">numscalingfactors</td><td>pointer to an integer variable that will receive the number of elements in the list</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>a pointer to a list of fractional scaling factors, or NULL if an error is encountered (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga9d63a05fc6d813f4aae06107041a37e8"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga9d63a05fc6d813f4aae06107041a37e8">&#9670;&nbsp;</a></span>tjInitCompress()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT <a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> tjInitCompress </td>
+          <td>(</td>
+          <td class="paramtype">void&#160;</td>
+          <td class="paramname"></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Create a TurboJPEG compressor instance. </p>
+<dl class="section return"><dt>Returns</dt><dd>a handle to the newly-created instance, or NULL if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga52300eac3f3d9ef4bab303bc244f62d3"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga52300eac3f3d9ef4bab303bc244f62d3">&#9670;&nbsp;</a></span>tjInitDecompress()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT <a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> tjInitDecompress </td>
+          <td>(</td>
+          <td class="paramtype">void&#160;</td>
+          <td class="paramname"></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Create a TurboJPEG decompressor instance. </p>
+<dl class="section return"><dt>Returns</dt><dd>a handle to the newly-created instance, or NULL if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga928beff6ac248ceadf01089fc6b41957"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga928beff6ac248ceadf01089fc6b41957">&#9670;&nbsp;</a></span>tjInitTransform()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT <a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a> tjInitTransform </td>
+          <td>(</td>
+          <td class="paramtype">void&#160;</td>
+          <td class="paramname"></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Create a new TurboJPEG transformer instance. </p>
+<dl class="section return"><dt>Returns</dt><dd>a handle to the newly-created instance, or NULL if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="gaffbd83c375e79f5db4b5c5d8ad4466e7"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gaffbd83c375e79f5db4b5c5d8ad4466e7">&#9670;&nbsp;</a></span>tjLoadImage()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT unsigned char* tjLoadImage </td>
+          <td>(</td>
+          <td class="paramtype">const char *&#160;</td>
+          <td class="paramname"><em>filename</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>align</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int *&#160;</td>
+          <td class="paramname"><em>pixelFormat</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Load an uncompressed image from disk into memory. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">filename</td><td>name of a file containing an uncompressed image in Windows BMP or PBMPLUS (PPM/PGM) format</td></tr>
+    <tr><td class="paramname">width</td><td>pointer to an integer variable that will receive the width (in pixels) of the uncompressed image</td></tr>
+    <tr><td class="paramname">align</td><td>row alignment of the image buffer to be returned (must be a power of 2.) For instance, setting this parameter to 4 will cause all rows in the image buffer to be padded to the nearest 32-bit boundary, and setting this parameter to 1 will cause all rows in the image buffer to be unpadded.</td></tr>
+    <tr><td class="paramname">height</td><td>pointer to an integer variable that will receive the height (in pixels) of the uncompressed image</td></tr>
+    <tr><td class="paramname">pixelFormat</td><td>pointer to an integer variable that specifies or will receive the pixel format of the uncompressed image buffer. The behavior of <a class="el" href="group___turbo_j_p_e_g.html#gaffbd83c375e79f5db4b5c5d8ad4466e7" title="Load an uncompressed image from disk into memory.">tjLoadImage()</a> will vary depending on the value of <code>*pixelFormat</code> passed to the function:<ul>
+<li><a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa84c1a6cead7952998e2fb895844a21ed">TJPF_UNKNOWN</a> : The uncompressed image buffer returned by the function will use the most optimal pixel format for the file type, and <code>*pixelFormat</code> will contain the ID of this pixel format upon successful return from the function.</li>
+<li><a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa5431b54b015337705f13118073711a1a">TJPF_GRAY</a> : Only PGM files and 8-bit BMP files with a grayscale colormap can be loaded.</li>
+<li><a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa7f5100ec44c91994e243f1cf55553f8b">TJPF_CMYK</a> : The RGB or grayscale pixels stored in the file will be converted using a quick &amp; dirty algorithm that is suitable only for testing purposes (proper conversion between CMYK and other formats requires a color management system.)</li>
+<li>Other <a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">pixel formats</a> : The uncompressed image buffer will use the specified pixel format, and pixel format conversion will be performed if necessary.</li>
+</ul>
+</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#ga72ecf4ebe6eb702d3c6f5ca27455e1ec">flags</a>.</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>a pointer to a newly-allocated buffer containing the uncompressed image, converted to the chosen pixel format and with the chosen row alignment, or NULL if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a>.) This buffer should be freed using <a class="el" href="group___turbo_j_p_e_g.html#gaea863d2da0cdb609563aabdf9196514b" title="Free an image buffer previously allocated by TurboJPEG.">tjFree()</a>. </dd></dl>
+
+</div>
+</div>
+<a id="ga1a209696c6a80748f20e134b3c64789f"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga1a209696c6a80748f20e134b3c64789f">&#9670;&nbsp;</a></span>tjPlaneHeight()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjPlaneHeight </td>
+          <td>(</td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>componentID</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The plane height of a YUV image plane with the given parameters. </p>
+<p>Refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a> for a description of plane height.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">componentID</td><td>ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the YUV image</td></tr>
+    <tr><td class="paramname">subsamp</td><td>level of chrominance subsampling in the image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>the plane height of a YUV image plane with the given parameters, or -1 if the arguments are out of bounds. </dd></dl>
+
+</div>
+</div>
+<a id="gab4ab7b24f6e797d79abaaa670373961d"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gab4ab7b24f6e797d79abaaa670373961d">&#9670;&nbsp;</a></span>tjPlaneSizeYUV()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT unsigned long tjPlaneSizeYUV </td>
+          <td>(</td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>componentID</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>stride</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The size of the buffer (in bytes) required to hold a YUV image plane with the given parameters. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">componentID</td><td>ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the YUV image. NOTE: this is the width of the whole image, not the plane width.</td></tr>
+    <tr><td class="paramname">stride</td><td>bytes per line in the image plane. Setting this to 0 is the equivalent of setting it to the plane width.</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the YUV image. NOTE: this is the height of the whole image, not the plane height.</td></tr>
+    <tr><td class="paramname">subsamp</td><td>level of chrominance subsampling in the image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>the size of the buffer (in bytes) required to hold the YUV image plane, or -1 if the arguments are out of bounds. </dd></dl>
+
+</div>
+</div>
+<a id="ga63fb66bb1e36c74008c4634360becbb1"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga63fb66bb1e36c74008c4634360becbb1">&#9670;&nbsp;</a></span>tjPlaneWidth()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjPlaneWidth </td>
+          <td>(</td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>componentID</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>subsamp</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The plane width of a YUV image plane with the given parameters. </p>
+<p>Refer to <a class="el" href="group___turbo_j_p_e_g.html#YUVnotes">YUV Image Format Notes</a> for a description of plane width.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">componentID</td><td>ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the YUV image</td></tr>
+    <tr><td class="paramname">subsamp</td><td>level of chrominance subsampling in the image (see <a class="el" href="group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074">Chrominance subsampling options</a>.)</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>the plane width of a YUV image plane with the given parameters, or -1 if the arguments are out of bounds. </dd></dl>
+
+</div>
+</div>
+<a id="ga6f445b22d8933ae4815b3370a538d879"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga6f445b22d8933ae4815b3370a538d879">&#9670;&nbsp;</a></span>tjSaveImage()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjSaveImage </td>
+          <td>(</td>
+          <td class="paramtype">const char *&#160;</td>
+          <td class="paramname"><em>filename</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char *&#160;</td>
+          <td class="paramname"><em>buffer</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>width</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pitch</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>height</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>pixelFormat</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Save an uncompressed image from memory to disk. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">filename</td><td>name of a file to which to save the uncompressed image. The image will be stored in Windows BMP or PBMPLUS (PPM/PGM) format, depending on the file extension.</td></tr>
+    <tr><td class="paramname">buffer</td><td>pointer to an image buffer containing RGB, grayscale, or CMYK pixels to be saved</td></tr>
+    <tr><td class="paramname">width</td><td>width (in pixels) of the uncompressed image</td></tr>
+    <tr><td class="paramname">pitch</td><td>bytes per line in the image buffer. Setting this parameter to 0 is the equivalent of setting it to <code>width * <a class="el" href="group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c" title="Pixel size (in bytes) for a given pixel format.">tjPixelSize</a>pixelFormat</code>.</td></tr>
+    <tr><td class="paramname">height</td><td>height (in pixels) of the uncompressed image</td></tr>
+    <tr><td class="paramname">pixelFormat</td><td>pixel format of the image buffer (see <a class="el" href="group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a">Pixel formats</a>.) If this parameter is set to <a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa5431b54b015337705f13118073711a1a">TJPF_GRAY</a>, then the image will be stored in PGM or 8-bit (indexed color) BMP format. Otherwise, the image will be stored in PPM or 24-bit BMP format. If this parameter is set to <a class="el" href="group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa7f5100ec44c91994e243f1cf55553f8b">TJPF_CMYK</a>, then the CMYK pixels will be converted to RGB using a quick &amp; dirty algorithm that is suitable only for testing (proper conversion between CMYK and other formats requires a color management system.)</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#ga72ecf4ebe6eb702d3c6f5ca27455e1ec">flags</a>.</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a>.) </dd></dl>
+
+</div>
+</div>
+<a id="ga9cb8abf4cc91881e04a0329b2270be25"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga9cb8abf4cc91881e04a0329b2270be25">&#9670;&nbsp;</a></span>tjTransform()</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">DLLEXPORT int tjTransform </td>
+          <td>(</td>
+          <td class="paramtype"><a class="el" href="group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763">tjhandle</a>&#160;</td>
+          <td class="paramname"><em>handle</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const unsigned char *&#160;</td>
+          <td class="paramname"><em>jpegBuf</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned long&#160;</td>
+          <td class="paramname"><em>jpegSize</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>n</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned char **&#160;</td>
+          <td class="paramname"><em>dstBufs</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">unsigned long *&#160;</td>
+          <td class="paramname"><em>dstSizes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype"><a class="el" href="structtjtransform.html">tjtransform</a> *&#160;</td>
+          <td class="paramname"><em>transforms</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>flags</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Losslessly transform a JPEG image into another JPEG image. </p>
+<p>Lossless transforms work by moving the raw DCT coefficients from one JPEG image structure to another without altering the values of the coefficients. While this is typically faster than decompressing the image, transforming it, and re-compressing it, lossless transforms are not free. Each lossless transform requires reading and performing Huffman decoding on all of the coefficients in the source image, regardless of the size of the destination image. Thus, this function provides a means of generating multiple transformed images from the same source or applying multiple transformations simultaneously, in order to eliminate the need to read the source coefficients multiple times.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">handle</td><td>a handle to a TurboJPEG transformer instance</td></tr>
+    <tr><td class="paramname">jpegBuf</td><td>pointer to a buffer containing the JPEG source image to transform</td></tr>
+    <tr><td class="paramname">jpegSize</td><td>size of the JPEG source image (in bytes)</td></tr>
+    <tr><td class="paramname">n</td><td>the number of transformed JPEG images to generate</td></tr>
+    <tr><td class="paramname">dstBufs</td><td>pointer to an array of n image buffers. <code>dstBufsi</code> will receive a JPEG image that has been transformed using the parameters in <code>transformsi</code>. TurboJPEG has the ability to reallocate the JPEG buffer to accommodate the size of the JPEG image. Thus, you can choose to:<ol type="1">
+<li>pre-allocate the JPEG buffer with an arbitrary size using <a class="el" href="group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83" title="Allocate an image buffer for use with TurboJPEG.">tjAlloc()</a> and let TurboJPEG grow the buffer as needed,</li>
+<li>set <code>dstBufsi</code> to NULL to tell TurboJPEG to allocate the buffer for you, or</li>
+<li>pre-allocate the buffer to a "worst case" size determined by calling <a class="el" href="group___turbo_j_p_e_g.html#ga67ac12fee79073242cb216e07c9f1f90" title="The maximum size of the buffer (in bytes) required to hold a JPEG image with the given parameters.">tjBufSize()</a> with the transformed or cropped width and height. Under normal circumstances, this should ensure that the buffer never has to be re-allocated (setting <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a> guarantees that it won't be.) Note, however, that there are some rare cases (such as transforming images with a large amount of embedded EXIF or ICC profile data) in which the output image will be larger than the worst-case size, and <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a> cannot be used in those cases.</li>
+</ol>
+If you choose option 1, <code>dstSizesi</code> should be set to the size of your pre-allocated buffer. In any case, unless you have set <a class="el" href="group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963" title="Disable buffer (re)allocation.">TJFLAG_NOREALLOC</a>, you should always check <code>dstBufsi</code> upon return from this function, as it may have changed.</td></tr>
+    <tr><td class="paramname">dstSizes</td><td>pointer to an array of n unsigned long variables that will receive the actual sizes (in bytes) of each transformed JPEG image. If <code>dstBufsi</code> points to a pre-allocated buffer, then <code>dstSizesi</code> should be set to the size of the buffer. Upon return, <code>dstSizesi</code> will contain the size of the JPEG image (in bytes.)</td></tr>
+    <tr><td class="paramname">transforms</td><td>pointer to an array of n <a class="el" href="structtjtransform.html" title="Lossless transform.">tjtransform</a> structures, each of which specifies the transform parameters and/or cropping region for the corresponding transformed output image.</td></tr>
+    <tr><td class="paramname">flags</td><td>the bitwise OR of one or more of the <a class="el" href="group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0">flags</a></td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if successful, or -1 if an error occurred (see <a class="el" href="group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa" title="Returns a descriptive error message explaining why the last command failed.">tjGetErrorStr2()</a> and <a class="el" href="group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410" title="Returns a code indicating the severity of the last error.">tjGetErrorCode()</a>.) </dd></dl>
+
+</div>
+</div>
+<h2 class="groupheader">Variable Documentation</h2>
+<a id="ga5af0ab065feefd526debf1e20c43e837"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga5af0ab065feefd526debf1e20c43e837">&#9670;&nbsp;</a></span>tjAlphaOffset</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">const int tjAlphaOffset<a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">static</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Alpha offset (in bytes) for a given pixel format. </p>
+<p>This specifies the number of bytes that the Alpha component is offset from the start of the pixel. For instance, if a pixel of format TJ_BGRA is stored in <code>char pixel</code>, then the alpha component will be <code>pixeltjAlphaOffsetTJ_BGRA</code>. This will be -1 if the pixel format does not have an alpha component. </p>
+
+</div>
+</div>
+<a id="ga84e2e35d3f08025f976ec1ec53693dea"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga84e2e35d3f08025f976ec1ec53693dea">&#9670;&nbsp;</a></span>tjBlueOffset</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">const int tjBlueOffset<a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">static</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Blue offset (in bytes) for a given pixel format. </p>
+<p>This specifies the number of bytes that the Blue component is offset from the start of the pixel. For instance, if a pixel of format TJ_BGRX is stored in <code>char pixel</code>, then the blue component will be <code>pixeltjBlueOffsetTJ_BGRX</code>. This will be -1 if the pixel format does not have a blue component. </p>
+
+</div>
+</div>
+<a id="ga82d6e35da441112a411da41923c0ba2f"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga82d6e35da441112a411da41923c0ba2f">&#9670;&nbsp;</a></span>tjGreenOffset</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">const int tjGreenOffset<a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">static</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Green offset (in bytes) for a given pixel format. </p>
+<p>This specifies the number of bytes that the green component is offset from the start of the pixel. For instance, if a pixel of format TJ_BGRX is stored in <code>char pixel</code>, then the green component will be <code>pixeltjGreenOffsetTJ_BGRX</code>. This will be -1 if the pixel format does not have a green component. </p>
+
+</div>
+</div>
+<a id="gabd247bb9fecb393eca57366feb8327bf"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gabd247bb9fecb393eca57366feb8327bf">&#9670;&nbsp;</a></span>tjMCUHeight</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">const int tjMCUHeight<a class="el" href="group___turbo_j_p_e_g.html#ga5ef3d169162ce77ce348e292a0b7477c">TJ_NUMSAMP</a></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">static</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>MCU block height (in pixels) for a given level of chrominance subsampling. </p>
+<p>MCU block sizes:</p><ul>
+<li>8x8 for no subsampling or grayscale</li>
+<li>16x8 for 4:2:2</li>
+<li>8x16 for 4:4:0</li>
+<li>16x16 for 4:2:0</li>
+<li>32x8 for 4:1:1 </li>
+</ul>
+
+</div>
+</div>
+<a id="ga9e61e7cd47a15a173283ba94e781308c"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ga9e61e7cd47a15a173283ba94e781308c">&#9670;&nbsp;</a></span>tjMCUWidth</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">const int tjMCUWidth<a class="el" href="group___turbo_j_p_e_g.html#ga5ef3d169162ce77ce348e292a0b7477c">TJ_NUMSAMP</a></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">static</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>MCU block width (in pixels) for a given level of chrominance subsampling. </p>
+<p>MCU block sizes:</p><ul>
+<li>8x8 for no subsampling or grayscale</li>
+<li>16x8 for 4:2:2</li>
+<li>8x16 for 4:4:0</li>
+<li>16x16 for 4:2:0</li>
+<li>32x8 for 4:1:1 </li>
+</ul>
+
+</div>
+</div>
+<a id="gad77cf8fe5b2bfd3cb3f53098146abb4c"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gad77cf8fe5b2bfd3cb3f53098146abb4c">&#9670;&nbsp;</a></span>tjPixelSize</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">const int tjPixelSize<a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">static</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Pixel size (in bytes) for a given pixel format. </p>
+
+</div>
+</div>
+<a id="gadd9b446742ac8a3923f7992c7988fea8"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#gadd9b446742ac8a3923f7992c7988fea8">&#9670;&nbsp;</a></span>tjRedOffset</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">const int tjRedOffset<a class="el" href="group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e">TJ_NUMPF</a></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">static</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Red offset (in bytes) for a given pixel format. </p>
+<p>This specifies the number of bytes that the red component is offset from the start of the pixel. For instance, if a pixel of format TJ_BGRX is stored in <code>char pixel</code>, then the red component will be <code>pixeltjRedOffsetTJ_BGRX</code>. This will be -1 if the pixel format does not have a red component. </p>
+
+</div>
+</div>
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/index.html Added

@@ -0,0 +1,76 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: Main Page</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+</div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+<div class="header">
+  <div class="headertitle">
+<div class="title">TurboJPEG Documentation</div>  </div>
+</div>
+<div class="contents">
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/jquery.js Added

@@ -0,0 +1,35 @@
+/*! jQuery v3.4.1 | (c) JS Foundation and other contributors | jquery.org/license */
+!function(e,t){"use strict";"object"==typeof module&&"object"==typeof module.exports?module.exports=e.document?t(e,!0):function(e){if(!e.document)throw new Error("jQuery requires a window with a document");return t(e)}:t(e)}("undefined"!=typeof window?window:this,function(C,e){"use strict";var t=,E=C.document,r=Object.getPrototypeOf,s=t.slice,g=t.concat,u=t.push,i=t.indexOf,n={},o=n.toString,v=n.hasOwnProperty,a=v.toString,l=a.call(Object),y={},m=function(e){return"function"==typeof e&&"number"!=typeof e.nodeType},x=function(e){return null!=e&&e===e.window},c={type:!0,src:!0,nonce:!0,noModule:!0};function b(e,t,n){var r,i,o=(n=n||E).createElement("script");if(o.text=e,t)for(r in c)(i=tr||t.getAttribute&&t.getAttribute(r))&&o.setAttribute(r,i);n.head.appendChild(o).parentNode.removeChild(o)}function w(e){return null==e?e+"":"object"==typeof e||"function"==typeof e?no.call(e)||"object":typeof e}var f="3.4.1",k=function(e,t){return new k.fn.init(e,t)},p=/^\s\uFEFF\xA0+|\s\uFEFF\xA0+$/g;function d(e){var t=!!e&&"length"in e&&e.length,n=w(e);return!m(e)&&!x(e)&&("array"===n||0===t||"number"==typeof t&&0<t&&t-1 in e)}k.fn=k.prototype={jquery:f,constructor:k,length:0,toArray:function(){return s.call(this)},get:function(e){return null==e?s.call(this):e<0?thise+this.length:thise},pushStack:function(e){var t=k.merge(this.constructor(),e);return t.prevObject=this,t},each:function(e){return k.each(this,e)},map:function(n){return this.pushStack(k.map(this,function(e,t){return n.call(e,t,e)}))},slice:function(){return this.pushStack(s.apply(this,arguments))},first:function(){return this.eq(0)},last:function(){return this.eq(-1)},eq:function(e){var t=this.length,n=+e+(e<0?t:0);return this.pushStack(0<=n&&n<t?thisn:)},end:function(){return this.prevObject||this.constructor()},push:u,sort:t.sort,splice:t.splice},k.extend=k.fn.extend=function(){var e,t,n,r,i,o,a=arguments0||{},s=1,u=arguments.length,l=!1;for("boolean"==typeof a&&(l=a,a=argumentss||{},s++),"object"==typeof a||m(a)||(a={}),s===u&&(a=this,s--);s<u;s++)if(null!=(e=argumentss))for(t in e)r=et,"__proto__"!==t&&a!==r&&(l&&r&&(k.isPlainObject(r)||(i=Array.isArray(r)))?(n=at,o=i&&!Array.isArray(n)?:i||k.isPlainObject(n)?n:{},i=!1,at=k.extend(l,o,r)):void 0!==r&&(at=r));return a},k.extend({expando:"jQuery"+(f+Math.random()).replace(/\D/g,""),isReady:!0,error:function(e){throw new Error(e)},noop:function(){},isPlainObject:function(e){var t,n;return!(!e||"object Object"!==o.call(e))&&(!(t=r(e))||"function"==typeof(n=v.call(t,"constructor")&&t.constructor)&&a.call(n)===l)},isEmptyObject:function(e){var t;for(t in e)return!1;return!0},globalEval:function(e,t){b(e,{nonce:t&&t.nonce})},each:function(e,t){var n,r=0;if(d(e)){for(n=e.length;r<n;r++)if(!1===t.call(er,r,er))break}else for(r in e)if(!1===t.call(er,r,er))break;return e},trim:function(e){return null==e?"":(e+"").replace(p,"")},makeArray:function(e,t){var n=t||;return null!=e&&(d(Object(e))?k.merge(n,"string"==typeof e?e:e):u.call(n,e)),n},inArray:function(e,t,n){return null==t?-1:i.call(t,e,n)},merge:function(e,t){for(var n=+t.length,r=0,i=e.length;r<n;r++)ei++=tr;return e.length=i,e},grep:function(e,t,n){for(var r=,i=0,o=e.length,a=!n;i<o;i++)!t(ei,i)!==a&&r.push(ei);return r},map:function(e,t,n){var r,i,o=0,a=;if(d(e))for(r=e.length;o<r;o++)null!=(i=t(eo,o,n))&&a.push(i);else for(o in e)null!=(i=t(eo,o,n))&&a.push(i);return g.apply(,a)},guid:1,support:y}),"function"==typeof Symbol&&(k.fnSymbol.iterator=tSymbol.iterator),k.each("Boolean Number String Function Array Date RegExp Object Error Symbol".split(" "),function(e,t){n"object "+t+""=t.toLowerCase()});var h=function(n){var e,d,b,o,i,h,f,g,w,u,l,T,C,a,E,v,s,c,y,k="sizzle"+1*new Date,m=n.document,S=0,r=0,p=ue(),x=ue(),N=ue(),A=ue(),D=function(e,t){return e===t&&(l=!0),0},j={}.hasOwnProperty,t=,q=t.pop,L=t.push,H=t.push,O=t.slice,P=function(e,t){for(var n=0,r=e.length;n<r;n++)if(en===t)return n;return-1},R="checked|selected|async|autofocus|autoplay|controls|defer|disabled|hidden|ismap|loop|multiple|open|readonly|required|scoped",M="\\x20\\t\\r\\n\\f",I="(?:\\\\.|\\w-|^\0-\\xa0)+",W="\\"+M+"*("+I+")(?:"+M+"*(*^$|!~?=)"+M+"*(?:'((?:\\\\.|^\\\\')*)'|\"((?:\\\\.|^\\\\\")*)\"|("+I+"))|)"+M+"*\\",$=":("+I+")(?:\$(('((?:\\\\.|^\\\\')*)'|\"((?:\\\\.|^\\\\\")*)\")|((?:\\\\.|^\\\\()\\|"+W+")*)|.*)\$|)",F=new RegExp(M+"+","g"),B=new RegExp("^"+M+"+|((?:^|^\\\\)(?:\\\\.)*)"+M+"+$","g"),_=new RegExp("^"+M+"*,"+M+"*"),z=new RegExp("^"+M+"*(>+~|"+M+")"+M+"*"),U=new RegExp(M+"|>"),X=new RegExp($),V=new RegExp("^"+I+"$"),G={ID:new RegExp("^#("+I+")"),CLASS:new RegExp("^\\.("+I+")"),TAG:new RegExp("^("+I+"|*)"),ATTR:new RegExp("^"+W),PSEUDO:new RegExp("^"+$),CHILD:new RegExp("^:(only|first|last|nth|nth-last)-(child|of-type)(?:\$"+M+"*(even|odd|((+-|)(\\d*)n|)"+M+"*(?:(+-|)"+M+"*(\\d+)|))"+M+"*\$|)","i"),bool:new RegExp("^(?:"+R+")$","i"),needsContext:new RegExp("^"+M+"*>+~|:(even|odd|eq|gt|lt|nth|first|last)(?:\$"+M+"*((?:-\\d)?\\d*)"+M+"*\$|)(?=^-|$)","i")},Y=/HTML$/i,Q=/^(?:input|select|textarea|button)$/i,J=/^h\d$/i,K=/^^{+\{\s*\native \w/,Z=/^(?:#(\w-+)|(\w+)|\.(\w-+))$/,ee=/+~/,te=new RegExp("\\\\(\\da-f{1,6}"+M+"?|("+M+")|.)","ig"),ne=function(e,t,n){var r="0x"+t-65536;return r!=r||n?t:r<0?String.fromCharCode(r+65536):String.fromCharCode(r>>10|55296,1023&r|56320)},re=/(\0-\x1f\x7f|^-?\d)|^-$|^\0-\x1f\x7f-\uFFFF\w-/g,ie=function(e,t){return t?"\0"===e?"\ufffd":e.slice(0,-1)+"\\"+e.charCodeAt(e.length-1).toString(16)+" ":"\\"+e},oe=function(){T()},ae=be(function(e){return!0===e.disabled&&"fieldset"===e.nodeName.toLowerCase()},{dir:"parentNode",next:"legend"});try{H.apply(t=O.call(m.childNodes),m.childNodes),tm.childNodes.length.nodeType}catch(e){H={apply:t.length?function(e,t){L.apply(e,O.call(t))}:function(e,t){var n=e.length,r=0;while(en++=tr++);e.length=n-1}}}function se(t,e,n,r){var i,o,a,s,u,l,c,f=e&&e.ownerDocument,p=e?e.nodeType:9;if(n=n||,"string"!=typeof t||!t||1!==p&&9!==p&&11!==p)return n;if(!r&&((e?e.ownerDocument||e:m)!==C&&T(e),e=e||C,E)){if(11!==p&&(u=Z.exec(t)))if(i=u1){if(9===p){if(!(a=e.getElementById(i)))return n;if(a.id===i)return n.push(a),n}else if(f&&(a=f.getElementById(i))&&y(e,a)&&a.id===i)return n.push(a),n}else{if(u2)return H.apply(n,e.getElementsByTagName(t)),n;if((i=u3)&&d.getElementsByClassName&&e.getElementsByClassName)return H.apply(n,e.getElementsByClassName(i)),n}if(d.qsa&&!At+" "&&(!v||!v.test(t))&&(1!==p||"object"!==e.nodeName.toLowerCase())){if(c=t,f=e,1===p&&U.test(t)){(s=e.getAttribute("id"))?s=s.replace(re,ie):e.setAttribute("id",s=k),o=(l=h(t)).length;while(o--)lo="#"+s+" "+xe(lo);c=l.join(","),f=ee.test(t)&&ye(e.parentNode)||e}try{return H.apply(n,f.querySelectorAll(c)),n}catch(e){A(t,!0)}finally{s===k&&e.removeAttribute("id")}}}return g(t.replace(B,"$1"),e,n,r)}function ue(){var r=;return function e(t,n){return r.push(t+" ")>b.cacheLength&&delete er.shift(),et+" "=n}}function le(e){return ek=!0,e}function ce(e){var t=C.createElement("fieldset");try{return!!e(t)}catch(e){return!1}finally{t.parentNode&&t.parentNode.removeChild(t),t=null}}function fe(e,t){var n=e.split("|"),r=n.length;while(r--)b.attrHandlenr=t}function pe(e,t){var n=t&&e,r=n&&1===e.nodeType&&1===t.nodeType&&e.sourceIndex-t.sourceIndex;if(r)return r;if(n)while(n=n.nextSibling)if(n===t)return-1;return e?1:-1}function de(t){return function(e){return"input"===e.nodeName.toLowerCase()&&e.type===t}}function he(n){return function(e){var t=e.nodeName.toLowerCase();return("input"===t||"button"===t)&&e.type===n}}function ge(t){return function(e){return"form"in e?e.parentNode&&!1===e.disabled?"label"in e?"label"in e.parentNode?e.parentNode.disabled===t:e.disabled===t:e.isDisabled===t||e.isDisabled!==!t&&ae(e)===t:e.disabled===t:"label"in e&&e.disabled===t}}function ve(a){return le(function(o){return o=+o,le(function(e,t){var n,r=a(,e.length,o),i=r.length;while(i--)en=ri&&(en=!(tn=en))})})}function ye(e){return e&&"undefined"!=typeof e.getElementsByTagName&&e}for(e in d=se.support={},i=se.isXML=function(e){var t=e.namespaceURI,n=(e.ownerDocument||e).documentElement;return!Y.test(t||n&&n.nodeName||"HTML")},T=se.setDocument=function(e){var t,n,r=e?e.ownerDocument||e:m;return r!==C&&9===r.nodeType&&r.documentElement&&(a=(C=r).documentElement,E=!i(C),m!==C&&(n=C.defaultView)&&n.top!==n&&(n.addEventListener?n.addEventListener("unload",oe,!1):n.attachEvent&&n.attachEvent("onunload",oe)),d.attributes=ce(function(e){return e.className="i",!e.getAttribute("className")}),d.getElementsByTagName=ce(function(e){return e.appendChild(C.createComment("")),!e.getElementsByTagName("*").length}),d.getElementsByClassName=K.test(C.getElementsByClassName),d.getById=ce(function(e){return a.appendChild(e).id=k,!C.getElementsByName||!C.getElementsByName(k).length}),d.getById?(b.filter.ID=function(e){var t=e.replace(te,ne);return function(e){return e.getAttribute("id")===t}},b.find.ID=function(e,t){if("undefined"!=typeof t.getElementById&&E){var n=t.getElementById(e);return n?n:}}):(b.filter.ID=function(e){var n=e.replace(te,ne);return function(e){var t="undefined"!=typeof e.getAttributeNode&&e.getAttributeNode("id");return t&&t.value===n}},b.find.ID=function(e,t){if("undefined"!=typeof t.getElementById&&E){var n,r,i,o=t.getElementById(e);if(o){if((n=o.getAttributeNode("id"))&&n.value===e)returno;i=t.getElementsByName(e),r=0;while(o=ir++)if((n=o.getAttributeNode("id"))&&n.value===e)returno}return}}),b.find.TAG=d.getElementsByTagName?function(e,t){return"undefined"!=typeof t.getElementsByTagName?t.getElementsByTagName(e):d.qsa?t.querySelectorAll(e):void 0}:function(e,t){var n,r=,i=0,o=t.getElementsByTagName(e);if("*"===e){while(n=oi++)1===n.nodeType&&r.push(n);return r}return o},b.find.CLASS=d.getElementsByClassName&&function(e,t){if("undefined"!=typeof t.getElementsByClassName&&E)return t.getElementsByClassName(e)},s=,v=,(d.qsa=K.test(C.querySelectorAll))&&(ce(function(e){a.appendChild(e).innerHTML="<a id='"+k+"'></a><select id='"+k+"-\r\\' msallowcapture=''><option selected=''></option></select>",e.querySelectorAll("msallowcapture^=''").length&&v.push("*^$="+M+"*(?:''|\"\")"),e.querySelectorAll("selected").length||v.push("\\"+M+"*(?:value|"+R+")"),e.querySelectorAll("id~="+k+"-").length||v.push("~="),e.querySelectorAll(":checked").length||v.push(":checked"),e.querySelectorAll("a#"+k+"+*").length||v.push(".#.++~")}),ce(function(e){e.innerHTML="<a href='' disabled='disabled'></a><select disabled='disabled'><option/></select>";var t=C.createElement("input");t.setAttribute("type","hidden"),e.appendChild(t).setAttribute("name","D"),e.querySelectorAll("name=d").length&&v.push("name"+M+"**^$|!~?="),2!==e.querySelectorAll(":enabled").length&&v.push(":enabled",":disabled"),a.appendChild(e).disabled=!0,2!==e.querySelectorAll(":disabled").length&&v.push(":enabled",":disabled"),e.querySelectorAll("*,:x"),v.push(",.*:")})),(d.matchesSelector=K.test(c=a.matches||a.webkitMatchesSelector||a.mozMatchesSelector||a.oMatchesSelector||a.msMatchesSelector))&&ce(function(e){d.disconnectedMatch=c.call(e,"*"),c.call(e,"s!='':x"),s.push("!=",$)}),v=v.length&&new RegExp(v.join("|")),s=s.length&&new RegExp(s.join("|")),t=K.test(a.compareDocumentPosition),y=t||K.test(a.contains)?function(e,t){var n=9===e.nodeType?e.documentElement:e,r=t&&t.parentNode;return e===r||!(!r||1!==r.nodeType||!(n.contains?n.contains(r):e.compareDocumentPosition&&16&e.compareDocumentPosition(r)))}:function(e,t){if(t)while(t=t.parentNode)if(t===e)return!0;return!1},D=t?function(e,t){if(e===t)return l=!0,0;var n=!e.compareDocumentPosition-!t.compareDocumentPosition;return n||(1&(n=(e.ownerDocument||e)===(t.ownerDocument||t)?e.compareDocumentPosition(t):1)||!d.sortDetached&&t.compareDocumentPosition(e)===n?e===C||e.ownerDocument===m&&y(m,e)?-1:t===C||t.ownerDocument===m&&y(m,t)?1:u?P(u,e)-P(u,t):0:4&n?-1:1)}:function(e,t){if(e===t)return l=!0,0;var n,r=0,i=e.parentNode,o=t.parentNode,a=e,s=t;if(!i||!o)return e===C?-1:t===C?1:i?-1:o?1:u?P(u,e)-P(u,t):0;if(i===o)return pe(e,t);n=e;while(n=n.parentNode)a.unshift(n);n=t;while(n=n.parentNode)s.unshift(n);while(ar===sr)r++;return r?pe(ar,sr):ar===m?-1:sr===m?1:0}),C},se.matches=function(e,t){return se(e,null,null,t)},se.matchesSelector=function(e,t){if((e.ownerDocument||e)!==C&&T(e),d.matchesSelector&&E&&!At+" "&&(!s||!s.test(t))&&(!v||!v.test(t)))try{var n=c.call(e,t);if(n||d.disconnectedMatch||e.document&&11!==e.document.nodeType)return n}catch(e){A(t,!0)}return 0<se(t,C,null,e).length},se.contains=function(e,t){return(e.ownerDocument||e)!==C&&T(e),y(e,t)},se.attr=function(e,t){(e.ownerDocument||e)!==C&&T(e);var n=b.attrHandlet.toLowerCase(),r=n&&j.call(b.attrHandle,t.toLowerCase())?n(e,t,!E):void 0;return void 0!==r?r:d.attributes||!E?e.getAttribute(t):(r=e.getAttributeNode(t))&&r.specified?r.value:null},se.escape=function(e){return(e+"").replace(re,ie)},se.error=function(e){throw new Error("Syntax error, unrecognized expression: "+e)},se.uniqueSort=function(e){var t,n=,r=0,i=0;if(l=!d.detectDuplicates,u=!d.sortStable&&e.slice(0),e.sort(D),l){while(t=ei++)t===ei&&(r=n.push(i));while(r--)e.splice(nr,1)}return u=null,e},o=se.getText=function(e){var t,n="",r=0,i=e.nodeType;if(i){if(1===i||9===i||11===i){if("string"==typeof e.textContent)return e.textContent;for(e=e.firstChild;e;e=e.nextSibling)n+=o(e)}else if(3===i||4===i)return e.nodeValue}else while(t=er++)n+=o(t);return n},(b=se.selectors={cacheLength:50,createPseudo:le,match:G,attrHandle:{},find:{},relative:{">":{dir:"parentNode",first:!0}," ":{dir:"parentNode"},"+":{dir:"previousSibling",first:!0},"~":{dir:"previousSibling"}},preFilter:{ATTR:function(e){return e1=e1.replace(te,ne),e3=(e3||e4||e5||"").replace(te,ne),"~="===e2&&(e3=" "+e3+" "),e.slice(0,4)},CHILD:function(e){return e1=e1.toLowerCase(),"nth"===e1.slice(0,3)?(e3||se.error(e0),e4=+(e4?e5+(e6||1):2*("even"===e3||"odd"===e3)),e5=+(e7+e8||"odd"===e3)):e3&&se.error(e0),e},PSEUDO:function(e){var t,n=!e6&&e2;return G.CHILD.test(e0)?null:(e3?e2=e4||e5||"":n&&X.test(n)&&(t=h(n,!0))&&(t=n.indexOf(")",n.length-t)-n.length)&&(e0=e0.slice(0,t),e2=n.slice(0,t)),e.slice(0,3))}},filter:{TAG:function(e){var t=e.replace(te,ne).toLowerCase();return"*"===e?function(){return!0}:function(e){return e.nodeName&&e.nodeName.toLowerCase()===t}},CLASS:function(e){var t=pe+" ";return t||(t=new RegExp("(^|"+M+")"+e+"("+M+"|$)"))&&p(e,function(e){return t.test("string"==typeof e.className&&e.className||"undefined"!=typeof e.getAttribute&&e.getAttribute("class")||"")})},ATTR:function(n,r,i){return function(e){var t=se.attr(e,n);return null==t?"!="===r:!r||(t+="","="===r?t===i:"!="===r?t!==i:"^="===r?i&&0===t.indexOf(i):"*="===r?i&&-1<t.indexOf(i):"$="===r?i&&t.slice(-i.length)===i:"~="===r?-1<(" "+t.replace(F," ")+" ").indexOf(i):"|="===r&&(t===i||t.slice(0,i.length+1)===i+"-"))}},CHILD:function(h,e,t,g,v){var y="nth"!==h.slice(0,3),m="last"!==h.slice(-4),x="of-type"===e;return 1===g&&0===v?function(e){return!!e.parentNode}:function(e,t,n){var r,i,o,a,s,u,l=y!==m?"nextSibling":"previousSibling",c=e.parentNode,f=x&&e.nodeName.toLowerCase(),p=!n&&!x,d=!1;if(c){if(y){while(l){a=e;while(a=al)if(x?a.nodeName.toLowerCase()===f:1===a.nodeType)return!1;u=l="only"===h&&!u&&"nextSibling"}return!0}if(u=m?c.firstChild:c.lastChild,m&&p){d=(s=(r=(i=(o=(a=c)k||(ak={}))a.uniqueID||(oa.uniqueID={}))h||)0===S&&r1)&&r2,a=s&&c.childNodess;while(a=++s&&a&&al||(d=s=0)||u.pop())if(1===a.nodeType&&++d&&a===e){ih=S,s,d;break}}else if(p&&(d=s=(r=(i=(o=(a=e)k||(ak={}))a.uniqueID||(oa.uniqueID={}))h||)0===S&&r1),!1===d)while(a=++s&&a&&al||(d=s=0)||u.pop())if((x?a.nodeName.toLowerCase()===f:1===a.nodeType)&&++d&&(p&&((i=(o=ak||(ak={}))a.uniqueID||(oa.uniqueID={}))h=S,d),a===e))break;return(d-=v)===g||d%g==0&&0<=d/g}}},PSEUDO:function(e,o){var t,a=b.pseudose||b.setFilterse.toLowerCase()||se.error("unsupported pseudo: "+e);return ak?a(o):1<a.length?(t=e,e,"",o,b.setFilters.hasOwnProperty(e.toLowerCase())?le(function(e,t){var n,r=a(e,o),i=r.length;while(i--)en=P(e,ri)=!(tn=ri)}):function(e){return a(e,0,t)}):a}},pseudos:{not:le(function(e){var r=,i=,s=f(e.replace(B,"$1"));return sk?le(function(e,t,n,r){var i,o=s(e,null,r,),a=e.length;while(a--)(i=oa)&&(ea=!(ta=i))}):function(e,t,n){return r0=e,s(r,null,n,i),r0=null,!i.pop()}}),has:le(function(t){return function(e){return 0<se(t,e).length}}),contains:le(function(t){return t=t.replace(te,ne),function(e){return-1<(e.textContent||o(e)).indexOf(t)}}),lang:le(function(n){return V.test(n||"")||se.error("unsupported lang: "+n),n=n.replace(te,ne).toLowerCase(),function(e){var t;do{if(t=E?e.lang:e.getAttribute("xml:lang")||e.getAttribute("lang"))return(t=t.toLowerCase())===n||0===t.indexOf(n+"-")}while((e=e.parentNode)&&1===e.nodeType);return!1}}),target:function(e){var t=n.location&&n.location.hash;return t&&t.slice(1)===e.id},root:function(e){return e===a},focus:function(e){return e===C.activeElement&&(!C.hasFocus||C.hasFocus())&&!!(e.type||e.href||~e.tabIndex)},enabled:ge(!1),disabled:ge(!0),checked:function(e){var t=e.nodeName.toLowerCase();return"input"===t&&!!e.checked||"option"===t&&!!e.selected},selected:function(e){return e.parentNode&&e.parentNode.selectedIndex,!0===e.selected},empty:function(e){for(e=e.firstChild;e;e=e.nextSibling)if(e.nodeType<6)return!1;return!0},parent:function(e){return!b.pseudos.empty(e)},header:function(e){return J.test(e.nodeName)},input:function(e){return Q.test(e.nodeName)},button:function(e){var t=e.nodeName.toLowerCase();return"input"===t&&"button"===e.type||"button"===t},text:function(e){var t;return"input"===e.nodeName.toLowerCase()&&"text"===e.type&&(null==(t=e.getAttribute("type"))||"text"===t.toLowerCase())},first:ve(function(){return0}),last:ve(function(e,t){returnt-1}),eq:ve(function(e,t,n){returnn<0?n+t:n}),even:ve(function(e,t){for(var n=0;n<t;n+=2)e.push(n);return e}),odd:ve(function(e,t){for(var n=1;n<t;n+=2)e.push(n);return e}),lt:ve(function(e,t,n){for(var r=n<0?n+t:t<n?t:n;0<=--r;)e.push(r);return e}),gt:ve(function(e,t,n){for(var r=n<0?n+t:n;++r<t;)e.push(r);return e})}}).pseudos.nth=b.pseudos.eq,{radio:!0,checkbox:!0,file:!0,password:!0,image:!0})b.pseudose=de(e);for(e in{submit:!0,reset:!0})b.pseudose=he(e);function me(){}function xe(e){for(var t=0,n=e.length,r="";t<n;t++)r+=et.value;return r}function be(s,e,t){var u=e.dir,l=e.next,c=l||u,f=t&&"parentNode"===c,p=r++;return e.first?function(e,t,n){while(e=eu)if(1===e.nodeType||f)return s(e,t,n);return!1}:function(e,t,n){var r,i,o,a=S,p;if(n){while(e=eu)if((1===e.nodeType||f)&&s(e,t,n))return!0}else while(e=eu)if(1===e.nodeType||f)if(i=(o=ek||(ek={}))e.uniqueID||(oe.uniqueID={}),l&&l===e.nodeName.toLowerCase())e=eu||e;else{if((r=ic)&&r0===S&&r1===p)return a2=r2;if((ic=a)2=s(e,t,n))return!0}return!1}}function we(i){return 1<i.length?function(e,t,n){var r=i.length;while(r--)if(!ir(e,t,n))return!1;return!0}:i0}function Te(e,t,n,r,i){for(var o,a=,s=0,u=e.length,l=null!=t;s<u;s++)(o=es)&&(n&&!n(o,r,i)||(a.push(o),l&&t.push(s)));return a}function Ce(d,h,g,v,y,e){return v&&!vk&&(v=Ce(v)),y&&!yk&&(y=Ce(y,e)),le(function(e,t,n,r){var i,o,a,s=,u=,l=t.length,c=e||function(e,t,n){for(var r=0,i=t.length;r<i;r++)se(e,tr,n);return n}(h||"*",n.nodeType?n:n,),f=!d||!e&&h?c:Te(c,s,d,n,r),p=g?y||(e?d:l||v)?:t:f;if(g&&g(f,p,n,r),v){i=Te(p,u),v(i,,n,r),o=i.length;while(o--)(a=io)&&(puo=!(fuo=a))}if(e){if(y||d){if(y){i=,o=p.length;while(o--)(a=po)&&i.push(fo=a);y(null,p=,i,r)}o=p.length;while(o--)(a=po)&&-1<(i=y?P(e,a):so)&&(ei=!(ti=a))}}else p=Te(p===t?p.splice(l,p.length):p),y?y(null,t,p,r):H.apply(t,p)})}function Ee(e){for(var i,t,n,r=e.length,o=b.relativee0.type,a=o||b.relative" ",s=o?1:0,u=be(function(e){return e===i},a,!0),l=be(function(e){return-1<P(i,e)},a,!0),c=function(e,t,n){var r=!o&&(n||t!==w)||((i=t).nodeType?u(e,t,n):l(e,t,n));return i=null,r};s<r;s++)if(t=b.relativees.type)c=be(we(c),t);else{if((t=b.filteres.type.apply(null,es.matches))k){for(n=++s;n<r;n++)if(b.relativeen.type)break;return Ce(1<s&&we(c),1<s&&xe(e.slice(0,s-1).concat({value:" "===es-2.type?"*":""})).replace(B,"$1"),t,s<n&&Ee(e.slice(s,n)),n<r&&Ee(e=e.slice(n)),n<r&&xe(e))}c.push(t)}return we(c)}return me.prototype=b.filters=b.pseudos,b.setFilters=new me,h=se.tokenize=function(e,t){var n,r,i,o,a,s,u,l=xe+" ";if(l)return t?0:l.slice(0);a=e,s=,u=b.preFilter;while(a){for(o in n&&!(r=_.exec(a))||(r&&(a=a.slice(r0.length)||a),s.push(i=)),n=!1,(r=z.exec(a))&&(n=r.shift(),i.push({value:n,type:r0.replace(B," ")}),a=a.slice(n.length)),b.filter)!(r=Go.exec(a))||uo&&!(r=uo(r))||(n=r.shift(),i.push({value:n,type:o,matches:r}),a=a.slice(n.length));if(!n)break}return t?a.length:a?se.error(e):x(e,s).slice(0)},f=se.compile=function(e,t){var n,v,y,m,x,r,i=,o=,a=Ne+" ";if(!a){t||(t=h(e)),n=t.length;while(n--)(a=Ee(tn))k?i.push(a):o.push(a);(a=N(e,(v=o,m=0<(y=i).length,x=0<v.length,r=function(e,t,n,r,i){var o,a,s,u=0,l="0",c=e&&,f=,p=w,d=e||x&&b.find.TAG("*",i),h=S+=null==p?1:Math.random()||.1,g=d.length;for(i&&(w=t===C||t||i);l!==g&&null!=(o=dl);l++){if(x&&o){a=0,t||o.ownerDocument===C||(T(o),n=!E);while(s=va++)if(s(o,t||C,n)){r.push(o);break}i&&(S=h)}m&&((o=!s&&o)&&u--,e&&c.push(o))}if(u+=l,m&&l!==u){a=0;while(s=ya++)s(c,f,t,n);if(e){if(0<u)while(l--)cl||fl||(fl=q.call(r));f=Te(f)}H.apply(r,f),i&&!e&&0<f.length&&1<u+y.length&&se.uniqueSort(r)}return i&&(S=h,w=p),c},m?le(r):r))).selector=e}return a},g=se.select=function(e,t,n,r){var i,o,a,s,u,l="function"==typeof e&&e,c=!r&&h(e=l.selector||e);if(n=n||,1===c.length){if(2<(o=c0=c0.slice(0)).length&&"ID"===(a=o0).type&&9===t.nodeType&&E&&b.relativeo1.type){if(!(t=(b.find.ID(a.matches0.replace(te,ne),t)||)0))return n;l&&(t=t.parentNode),e=e.slice(o.shift().value.length)}i=G.needsContext.test(e)?0:o.length;while(i--){if(a=oi,b.relatives=a.type)break;if((u=b.finds)&&(r=u(a.matches0.replace(te,ne),ee.test(o0.type)&&ye(t.parentNode)||t))){if(o.splice(i,1),!(e=r.length&&xe(o)))return H.apply(n,r),n;break}}}return(l||f(e,c))(r,t,!E,n,!t||ee.test(e)&&ye(t.parentNode)||t),n},d.sortStable=k.split("").sort(D).join("")===k,d.detectDuplicates=!!l,T(),d.sortDetached=ce(function(e){return 1&e.compareDocumentPosition(C.createElement("fieldset"))}),ce(function(e){return e.innerHTML="<a href='#'></a>","#"===e.firstChild.getAttribute("href")})||fe("type|href|height|width",function(e,t,n){if(!n)return e.getAttribute(t,"type"===t.toLowerCase()?1:2)}),d.attributes&&ce(function(e){return e.innerHTML="<input/>",e.firstChild.setAttribute("value",""),""===e.firstChild.getAttribute("value")})||fe("value",function(e,t,n){if(!n&&"input"===e.nodeName.toLowerCase())return e.defaultValue}),ce(function(e){return null==e.getAttribute("disabled")})||fe(R,function(e,t,n){var r;if(!n)return!0===et?t.toLowerCase():(r=e.getAttributeNode(t))&&r.specified?r.value:null}),se}(C);k.find=h,k.expr=h.selectors,k.expr":"=k.expr.pseudos,k.uniqueSort=k.unique=h.uniqueSort,k.text=h.getText,k.isXMLDoc=h.isXML,k.contains=h.contains,k.escapeSelector=h.escape;var T=function(e,t,n){var r=,i=void 0!==n;while((e=et)&&9!==e.nodeType)if(1===e.nodeType){if(i&&k(e).is(n))break;r.push(e)}return r},S=function(e,t){for(var n=;e;e=e.nextSibling)1===e.nodeType&&e!==t&&n.push(e);return n},N=k.expr.match.needsContext;function A(e,t){return e.nodeName&&e.nodeName.toLowerCase()===t.toLowerCase()}var D=/^<(a-z^\/\0>:\x20\t\r\n\f*)\x20\t\r\n\f*\/?>(?:<\/\1>|)$/i;function j(e,n,r){return m(n)?k.grep(e,function(e,t){return!!n.call(e,t,e)!==r}):n.nodeType?k.grep(e,function(e){return e===n!==r}):"string"!=typeof n?k.grep(e,function(e){return-1<i.call(n,e)!==r}):k.filter(n,e,r)}k.filter=function(e,t,n){var r=t0;return n&&(e=":not("+e+")"),1===t.length&&1===r.nodeType?k.find.matchesSelector(r,e)?r::k.find.matches(e,k.grep(t,function(e){return 1===e.nodeType}))},k.fn.extend({find:function(e){var t,n,r=this.length,i=this;if("string"!=typeof e)return this.pushStack(k(e).filter(function(){for(t=0;t<r;t++)if(k.contains(it,this))return!0}));for(n=this.pushStack(),t=0;t<r;t++)k.find(e,it,n);return 1<r?k.uniqueSort(n):n},filter:function(e){return this.pushStack(j(this,e||,!1))},not:function(e){return this.pushStack(j(this,e||,!0))},is:function(e){return!!j(this,"string"==typeof e&&N.test(e)?k(e):e||,!1).length}});var q,L=/^(?:\s*(<\w\W+>)^>*|#(\w-+))$/;(k.fn.init=function(e,t,n){var r,i;if(!e)return this;if(n=n||q,"string"==typeof e){if(!(r="<"===e0&&">"===ee.length-1&&3<=e.length?null,e,null:L.exec(e))||!r1&&t)return!t||t.jquery?(t||n).find(e):this.constructor(t).find(e);if(r1){if(t=t instanceof k?t0:t,k.merge(this,k.parseHTML(r1,t&&t.nodeType?t.ownerDocument||t:E,!0)),D.test(r1)&&k.isPlainObject(t))for(r in t)m(thisr)?thisr(tr):this.attr(r,tr);return this}return(i=E.getElementById(r2))&&(this0=i,this.length=1),this}return e.nodeType?(this0=e,this.length=1,this):m(e)?void 0!==n.ready?n.ready(e):e(k):k.makeArray(e,this)}).prototype=k.fn,q=k(E);var H=/^(?:parents|prev(?:Until|All))/,O={children:!0,contents:!0,next:!0,prev:!0};function P(e,t){while((e=et)&&1!==e.nodeType);return e}k.fn.extend({has:function(e){var t=k(e,this),n=t.length;return this.filter(function(){for(var e=0;e<n;e++)if(k.contains(this,te))return!0})},closest:function(e,t){var n,r=0,i=this.length,o=,a="string"!=typeof e&&k(e);if(!N.test(e))for(;r<i;r++)for(n=thisr;n&&n!==t;n=n.parentNode)if(n.nodeType<11&&(a?-1<a.index(n):1===n.nodeType&&k.find.matchesSelector(n,e))){o.push(n);break}return this.pushStack(1<o.length?k.uniqueSort(o):o)},index:function(e){return e?"string"==typeof e?i.call(k(e),this0):i.call(this,e.jquery?e0:e):this0&&this0.parentNode?this.first().prevAll().length:-1},add:function(e,t){return this.pushStack(k.uniqueSort(k.merge(this.get(),k(e,t))))},addBack:function(e){return this.add(null==e?this.prevObject:this.prevObject.filter(e))}}),k.each({parent:function(e){var t=e.parentNode;return t&&11!==t.nodeType?t:null},parents:function(e){return T(e,"parentNode")},parentsUntil:function(e,t,n){return T(e,"parentNode",n)},next:function(e){return P(e,"nextSibling")},prev:function(e){return P(e,"previousSibling")},nextAll:function(e){return T(e,"nextSibling")},prevAll:function(e){return T(e,"previousSibling")},nextUntil:function(e,t,n){return T(e,"nextSibling",n)},prevUntil:function(e,t,n){return T(e,"previousSibling",n)},siblings:function(e){return S((e.parentNode||{}).firstChild,e)},children:function(e){return S(e.firstChild)},contents:function(e){return"undefined"!=typeof e.contentDocument?e.contentDocument:(A(e,"template")&&(e=e.content||e),k.merge(,e.childNodes))}},function(r,i){k.fnr=function(e,t){var n=k.map(this,i,e);return"Until"!==r.slice(-5)&&(t=e),t&&"string"==typeof t&&(n=k.filter(t,n)),1<this.length&&(Or||k.uniqueSort(n),H.test(r)&&n.reverse()),this.pushStack(n)}});var R=/^\x20\t\r\n\f+/g;function M(e){return e}function I(e){throw e}function W(e,t,n,r){var i;try{e&&m(i=e.promise)?i.call(e).done(t).fail(n):e&&m(i=e.then)?i.call(e,t,n):t.apply(void 0,e.slice(r))}catch(e){n.apply(void 0,e)}}k.Callbacks=function(r){var e,n;r="string"==typeof r?(e=r,n={},k.each(e.match(R)||,function(e,t){nt=!0}),n):k.extend({},r);var i,t,o,a,s=,u=,l=-1,c=function(){for(a=a||r.once,o=i=!0;u.length;l=-1){t=u.shift();while(++l<s.length)!1===sl.apply(t0,t1)&&r.stopOnFalse&&(l=s.length,t=!1)}r.memory||(t=!1),i=!1,a&&(s=t?:"")},f={add:function(){return s&&(t&&!i&&(l=s.length-1,u.push(t)),function n(e){k.each(e,function(e,t){m(t)?r.unique&&f.has(t)||s.push(t):t&&t.length&&"string"!==w(t)&&n(t)})}(arguments),t&&!i&&c()),this},remove:function(){return k.each(arguments,function(e,t){var n;while(-1<(n=k.inArray(t,s,n)))s.splice(n,1),n<=l&&l--}),this},has:function(e){return e?-1<k.inArray(e,s):0<s.length},empty:function(){return s&&(s=),this},disable:function(){return a=u=,s=t="",this},disabled:function(){return!s},lock:function(){return a=u=,t||i||(s=t=""),this},locked:function(){return!!a},fireWith:function(e,t){return a||(t=e,(t=t||).slice?t.slice():t,u.push(t),i||c()),this},fire:function(){return f.fireWith(this,arguments),this},fired:function(){return!!o}};return f},k.extend({Deferred:function(e){var o="notify","progress",k.Callbacks("memory"),k.Callbacks("memory"),2,"resolve","done",k.Callbacks("once memory"),k.Callbacks("once memory"),0,"resolved","reject","fail",k.Callbacks("once memory"),k.Callbacks("once memory"),1,"rejected",i="pending",a={state:function(){return i},always:function(){return s.done(arguments).fail(arguments),this},"catch":function(e){return a.then(null,e)},pipe:function(){var i=arguments;return k.Deferred(function(r){k.each(o,function(e,t){var n=m(it4)&&it4;st1(function(){var e=n&&n.apply(this,arguments);e&&m(e.promise)?e.promise().progress(r.notify).done(r.resolve).fail(r.reject):rt0+"With"(this,n?e:arguments)})}),i=null}).promise()},then:function(t,n,r){var u=0;function l(i,o,a,s){return function(){var n=this,r=arguments,e=function(){var e,t;if(!(i<u)){if((e=a.apply(n,r))===o.promise())throw new TypeError("Thenable self-resolution");t=e&&("object"==typeof e||"function"==typeof e)&&e.then,m(t)?s?t.call(e,l(u,o,M,s),l(u,o,I,s)):(u++,t.call(e,l(u,o,M,s),l(u,o,I,s),l(u,o,M,o.notifyWith))):(a!==M&&(n=void 0,r=e),(s||o.resolveWith)(n,r))}},t=s?e:function(){try{e()}catch(e){k.Deferred.exceptionHook&&k.Deferred.exceptionHook(e,t.stackTrace),u<=i+1&&(a!==I&&(n=void 0,r=e),o.rejectWith(n,r))}};i?t():(k.Deferred.getStackHook&&(t.stackTrace=k.Deferred.getStackHook()),C.setTimeout(t))}}return k.Deferred(function(e){o03.add(l(0,e,m(r)?r:M,e.notifyWith)),o13.add(l(0,e,m(t)?t:M)),o23.add(l(0,e,m(n)?n:I))}).promise()},promise:function(e){return null!=e?k.extend(e,a):a}},s={};return k.each(o,function(e,t){var n=t2,r=t5;at1=n.add,r&&n.add(function(){i=r},o3-e2.disable,o3-e3.disable,o02.lock,o03.lock),n.add(t3.fire),st0=function(){return st0+"With"(this===s?void 0:this,arguments),this},st0+"With"=n.fireWith}),a.promise(s),e&&e.call(s,s),s},when:function(e){var n=arguments.length,t=n,r=Array(t),i=s.call(arguments),o=k.Deferred(),a=function(t){return function(e){rt=this,it=1<arguments.length?s.call(arguments):e,--n||o.resolveWith(r,i)}};if(n<=1&&(W(e,o.done(a(t)).resolve,o.reject,!n),"pending"===o.state()||m(it&&it.then)))return o.then();while(t--)W(it,a(t),o.reject);return o.promise()}});var $=/^(Eval|Internal|Range|Reference|Syntax|Type|URI)Error$/;k.Deferred.exceptionHook=function(e,t){C.console&&C.console.warn&&e&&$.test(e.name)&&C.console.warn("jQuery.Deferred exception: "+e.message,e.stack,t)},k.readyException=function(e){C.setTimeout(function(){throw e})};var F=k.Deferred();function B(){E.removeEventListener("DOMContentLoaded",B),C.removeEventListener("load",B),k.ready()}k.fn.ready=function(e){return F.then(e)"catch"(function(e){k.readyException(e)}),this},k.extend({isReady:!1,readyWait:1,ready:function(e){(!0===e?--k.readyWait:k.isReady)||(k.isReady=!0)!==e&&0<--k.readyWait||F.resolveWith(E,k)}}),k.ready.then=F.then,"complete"===E.readyState||"loading"!==E.readyState&&!E.documentElement.doScroll?C.setTimeout(k.ready):(E.addEventListener("DOMContentLoaded",B),C.addEventListener("load",B));var _=function(e,t,n,r,i,o,a){var s=0,u=e.length,l=null==n;if("object"===w(n))for(s in i=!0,n)_(e,t,s,ns,!0,o,a);else if(void 0!==r&&(i=!0,m(r)||(a=!0),l&&(a?(t.call(e,r),t=null):(l=t,t=function(e,t,n){return l.call(k(e),n)})),t))for(;s<u;s++)t(es,n,a?r:r.call(es,s,t(es,n)));return i?e:l?t.call(e):u?t(e0,n):o},z=/^-ms-/,U=/-(a-z)/g;function X(e,t){return t.toUpperCase()}function V(e){return e.replace(z,"ms-").replace(U,X)}var G=function(e){return 1===e.nodeType||9===e.nodeType||!+e.nodeType};function Y(){this.expando=k.expando+Y.uid++}Y.uid=1,Y.prototype={cache:function(e){var t=ethis.expando;return t||(t={},G(e)&&(e.nodeType?ethis.expando=t:Object.defineProperty(e,this.expando,{value:t,configurable:!0}))),t},set:function(e,t,n){var r,i=this.cache(e);if("string"==typeof t)iV(t)=n;else for(r in t)iV(r)=tr;return i},get:function(e,t){return void 0===t?this.cache(e):ethis.expando&&ethis.expandoV(t)},access:function(e,t,n){return void 0===t||t&&"string"==typeof t&&void 0===n?this.get(e,t):(this.set(e,t,n),void 0!==n?n:t)},remove:function(e,t){var n,r=ethis.expando;if(void 0!==r){if(void 0!==t){n=(t=Array.isArray(t)?t.map(V):(t=V(t))in r?t:t.match(R)||).length;while(n--)delete rtn}(void 0===t||k.isEmptyObject(r))&&(e.nodeType?ethis.expando=void 0:delete ethis.expando)}},hasData:function(e){var t=ethis.expando;return void 0!==t&&!k.isEmptyObject(t)}};var Q=new Y,J=new Y,K=/^(?:\{\w\W*\}|\\w\W*\)$/,Z=/A-Z/g;function ee(e,t,n){var r,i;if(void 0===n&&1===e.nodeType)if(r="data-"+t.replace(Z,"-$&").toLowerCase(),"string"==typeof(n=e.getAttribute(r))){try{n="true"===(i=n)||"false"!==i&&("null"===i?null:i===+i+""?+i:K.test(i)?JSON.parse(i):i)}catch(e){}J.set(e,t,n)}else n=void 0;return n}k.extend({hasData:function(e){return J.hasData(e)||Q.hasData(e)},data:function(e,t,n){return J.access(e,t,n)},removeData:function(e,t){J.remove(e,t)},_data:function(e,t,n){return Q.access(e,t,n)},_removeData:function(e,t){Q.remove(e,t)}}),k.fn.extend({data:function(n,e){var t,r,i,o=this0,a=o&&o.attributes;if(void 0===n){if(this.length&&(i=J.get(o),1===o.nodeType&&!Q.get(o,"hasDataAttrs"))){t=a.length;while(t--)at&&0===(r=at.name).indexOf("data-")&&(r=V(r.slice(5)),ee(o,r,ir));Q.set(o,"hasDataAttrs",!0)}return i}return"object"==typeof n?this.each(function(){J.set(this,n)}):_(this,function(e){var t;if(o&&void 0===e)return void 0!==(t=J.get(o,n))?t:void 0!==(t=ee(o,n))?t:void 0;this.each(function(){J.set(this,n,e)})},null,e,1<arguments.length,null,!0)},removeData:function(e){return this.each(function(){J.remove(this,e)})}}),k.extend({queue:function(e,t,n){var r;if(e)return t=(t||"fx")+"queue",r=Q.get(e,t),n&&(!r||Array.isArray(n)?r=Q.access(e,t,k.makeArray(n)):r.push(n)),r||},dequeue:function(e,t){t=t||"fx";var n=k.queue(e,t),r=n.length,i=n.shift(),o=k._queueHooks(e,t);"inprogress"===i&&(i=n.shift(),r--),i&&("fx"===t&&n.unshift("inprogress"),delete o.stop,i.call(e,function(){k.dequeue(e,t)},o)),!r&&o&&o.empty.fire()},_queueHooks:function(e,t){var n=t+"queueHooks";return Q.get(e,n)||Q.access(e,n,{empty:k.Callbacks("once memory").add(function(){Q.remove(e,t+"queue",n)})})}}),k.fn.extend({queue:function(t,n){var e=2;return"string"!=typeof t&&(n=t,t="fx",e--),arguments.length<e?k.queue(this0,t):void 0===n?this:this.each(function(){var e=k.queue(this,t,n);k._queueHooks(this,t),"fx"===t&&"inprogress"!==e0&&k.dequeue(this,t)})},dequeue:function(e){return this.each(function(){k.dequeue(this,e)})},clearQueue:function(e){return this.queue(e||"fx",)},promise:function(e,t){var n,r=1,i=k.Deferred(),o=this,a=this.length,s=function(){--r||i.resolveWith(o,o)};"string"!=typeof e&&(t=e,e=void 0),e=e||"fx";while(a--)(n=Q.get(oa,e+"queueHooks"))&&n.empty&&(r++,n.empty.add(s));return s(),i.promise(t)}});var te=/+-?(?:\d*\.|)\d+(?:eE+-?\d+|)/.source,ne=new RegExp("^(?:(+-)=|)("+te+")(a-z%*)$","i"),re="Top","Right","Bottom","Left",ie=E.documentElement,oe=function(e){return k.contains(e.ownerDocument,e)},ae={composed:!0};ie.getRootNode&&(oe=function(e){return k.contains(e.ownerDocument,e)||e.getRootNode(ae)===e.ownerDocument});var se=function(e,t){return"none"===(e=t||e).style.display||""===e.style.display&&oe(e)&&"none"===k.css(e,"display")},ue=function(e,t,n,r){var i,o,a={};for(o in t)ao=e.styleo,e.styleo=to;for(o in i=n.apply(e,r||),t)e.styleo=ao;return i};function le(e,t,n,r){var i,o,a=20,s=r?function(){return r.cur()}:function(){return k.css(e,t,"")},u=s(),l=n&&n3||(k.cssNumbert?"":"px"),c=e.nodeType&&(k.cssNumbert||"px"!==l&&+u)&&ne.exec(k.css(e,t));if(c&&c3!==l){u/=2,l=l||c3,c=+u||1;while(a--)k.style(e,t,c+l),(1-o)*(1-(o=s()/u||.5))<=0&&(a=0),c/=o;c*=2,k.style(e,t,c+l),n=n||}return n&&(c=+c||+u||0,i=n1?c+(n1+1)*n2:+n2,r&&(r.unit=l,r.start=c,r.end=i)),i}var ce={};function fe(e,t){for(var n,r,i,o,a,s,u,l=,c=0,f=e.length;c<f;c++)(r=ec).style&&(n=r.style.display,t?("none"===n&&(lc=Q.get(r,"display")||null,lc||(r.style.display="")),""===r.style.display&&se(r)&&(lc=(u=a=o=void 0,a=(i=r).ownerDocument,s=i.nodeName,(u=ces)||(o=a.body.appendChild(a.createElement(s)),u=k.css(o,"display"),o.parentNode.removeChild(o),"none"===u&&(u="block"),ces=u)))):"none"!==n&&(lc="none",Q.set(r,"display",n)));for(c=0;c<f;c++)null!=lc&&(ec.style.display=lc);return e}k.fn.extend({show:function(){return fe(this,!0)},hide:function(){return fe(this)},toggle:function(e){return"boolean"==typeof e?e?this.show():this.hide():this.each(function(){se(this)?k(this).show():k(this).hide()})}});var pe=/^(?:checkbox|radio)$/i,de=/<(a-z^\/\0>\x20\t\r\n\f*)/i,he=/^$|^module$|\/(?:java|ecma)script/i,ge={option:1,"<select multiple='multiple'>","</select>",thead:1,"<table>","</table>",col:2,"<table><colgroup>","</colgroup></table>",tr:2,"<table><tbody>","</tbody></table>",td:3,"<table><tbody><tr>","</tr></tbody></table>",_default:0,"",""};function ve(e,t){var n;return n="undefined"!=typeof e.getElementsByTagName?e.getElementsByTagName(t||"*"):"undefined"!=typeof e.querySelectorAll?e.querySelectorAll(t||"*"):,void 0===t||t&&A(e,t)?k.merge(e,n):n}function ye(e,t){for(var n=0,r=e.length;n<r;n++)Q.set(en,"globalEval",!t||Q.get(tn,"globalEval"))}ge.optgroup=ge.option,ge.tbody=ge.tfoot=ge.colgroup=ge.caption=ge.thead,ge.th=ge.td;var me,xe,be=/<|&#?\w+;/;function we(e,t,n,r,i){for(var o,a,s,u,l,c,f=t.createDocumentFragment(),p=,d=0,h=e.length;d<h;d++)if((o=ed)||0===o)if("object"===w(o))k.merge(p,o.nodeType?o:o);else if(be.test(o)){a=a||f.appendChild(t.createElement("div")),s=(de.exec(o)||"","")1.toLowerCase(),u=ges||ge._default,a.innerHTML=u1+k.htmlPrefilter(o)+u2,c=u0;while(c--)a=a.lastChild;k.merge(p,a.childNodes),(a=f.firstChild).textContent=""}else p.push(t.createTextNode(o));f.textContent="",d=0;while(o=pd++)if(r&&-1<k.inArray(o,r))i&&i.push(o);else if(l=oe(o),a=ve(f.appendChild(o),"script"),l&&ye(a),n){c=0;while(o=ac++)he.test(o.type||"")&&n.push(o)}return f}me=E.createDocumentFragment().appendChild(E.createElement("div")),(xe=E.createElement("input")).setAttribute("type","radio"),xe.setAttribute("checked","checked"),xe.setAttribute("name","t"),me.appendChild(xe),y.checkClone=me.cloneNode(!0).cloneNode(!0).lastChild.checked,me.innerHTML="<textarea>x</textarea>",y.noCloneChecked=!!me.cloneNode(!0).lastChild.defaultValue;var Te=/^key/,Ce=/^(?:mouse|pointer|contextmenu|drag|drop)|click/,Ee=/^(^.*)(?:\.(.+)|)/;function ke(){return!0}function Se(){return!1}function Ne(e,t){return e===function(){try{return E.activeElement}catch(e){}}()==("focus"===t)}function Ae(e,t,n,r,i,o){var a,s;if("object"==typeof t){for(s in"string"!=typeof n&&(r=r||n,n=void 0),t)Ae(e,s,n,r,ts,o);return e}if(null==r&&null==i?(i=n,r=n=void 0):null==i&&("string"==typeof n?(i=r,r=void 0):(i=r,r=n,n=void 0)),!1===i)i=Se;else if(!i)return e;return 1===o&&(a=i,(i=function(e){return k().off(e),a.apply(this,arguments)}).guid=a.guid||(a.guid=k.guid++)),e.each(function(){k.event.add(this,t,i,r,n)})}function De(e,i,o){o?(Q.set(e,i,!1),k.event.add(e,i,{namespace:!1,handler:function(e){var t,n,r=Q.get(this,i);if(1&e.isTrigger&&thisi){if(r.length)(k.event.speciali||{}).delegateType&&e.stopPropagation();else if(r=s.call(arguments),Q.set(this,i,r),t=o(this,i),thisi(),r!==(n=Q.get(this,i))||t?Q.set(this,i,!1):n={},r!==n)return e.stopImmediatePropagation(),e.preventDefault(),n.value}else r.length&&(Q.set(this,i,{value:k.event.trigger(k.extend(r0,k.Event.prototype),r.slice(1),this)}),e.stopImmediatePropagation())}})):void 0===Q.get(e,i)&&k.event.add(e,i,ke)}k.event={global:{},add:function(t,e,n,r,i){var o,a,s,u,l,c,f,p,d,h,g,v=Q.get(t);if(v){n.handler&&(n=(o=n).handler,i=o.selector),i&&k.find.matchesSelector(ie,i),n.guid||(n.guid=k.guid++),(u=v.events)||(u=v.events={}),(a=v.handle)||(a=v.handle=function(e){return"undefined"!=typeof k&&k.event.triggered!==e.type?k.event.dispatch.apply(t,arguments):void 0}),l=(e=(e||"").match(R)||"").length;while(l--)d=g=(s=Ee.exec(el)||)1,h=(s2||"").split(".").sort(),d&&(f=k.event.speciald||{},d=(i?f.delegateType:f.bindType)||d,f=k.event.speciald||{},c=k.extend({type:d,origType:g,data:r,handler:n,guid:n.guid,selector:i,needsContext:i&&k.expr.match.needsContext.test(i),namespace:h.join(".")},o),(p=ud)||((p=ud=).delegateCount=0,f.setup&&!1!==f.setup.call(t,r,h,a)||t.addEventListener&&t.addEventListener(d,a)),f.add&&(f.add.call(t,c),c.handler.guid||(c.handler.guid=n.guid)),i?p.splice(p.delegateCount++,0,c):p.push(c),k.event.globald=!0)}},remove:function(e,t,n,r,i){var o,a,s,u,l,c,f,p,d,h,g,v=Q.hasData(e)&&Q.get(e);if(v&&(u=v.events)){l=(t=(t||"").match(R)||"").length;while(l--)if(d=g=(s=Ee.exec(tl)||)1,h=(s2||"").split(".").sort(),d){f=k.event.speciald||{},p=ud=(r?f.delegateType:f.bindType)||d||,s=s2&&new RegExp("(^|\\.)"+h.join("\\.(?:.*\\.|)")+"(\\.|$)"),a=o=p.length;while(o--)c=po,!i&&g!==c.origType||n&&n.guid!==c.guid||s&&!s.test(c.namespace)||r&&r!==c.selector&&("**"!==r||!c.selector)||(p.splice(o,1),c.selector&&p.delegateCount--,f.remove&&f.remove.call(e,c));a&&!p.length&&(f.teardown&&!1!==f.teardown.call(e,h,v.handle)||k.removeEvent(e,d,v.handle),delete ud)}else for(d in u)k.event.remove(e,d+tl,n,r,!0);k.isEmptyObject(u)&&Q.remove(e,"handle events")}},dispatch:function(e){var t,n,r,i,o,a,s=k.event.fix(e),u=new Array(arguments.length),l=(Q.get(this,"events")||{})s.type||,c=k.event.specials.type||{};for(u0=s,t=1;t<arguments.length;t++)ut=argumentst;if(s.delegateTarget=this,!c.preDispatch||!1!==c.preDispatch.call(this,s)){a=k.event.handlers.call(this,s,l),t=0;while((i=at++)&&!s.isPropagationStopped()){s.currentTarget=i.elem,n=0;while((o=i.handlersn++)&&!s.isImmediatePropagationStopped())s.rnamespace&&!1!==o.namespace&&!s.rnamespace.test(o.namespace)||(s.handleObj=o,s.data=o.data,void 0!==(r=((k.event.specialo.origType||{}).handle||o.handler).apply(i.elem,u))&&!1===(s.result=r)&&(s.preventDefault(),s.stopPropagation()))}return c.postDispatch&&c.postDispatch.call(this,s),s.result}},handlers:function(e,t){var n,r,i,o,a,s=,u=t.delegateCount,l=e.target;if(u&&l.nodeType&&!("click"===e.type&&1<=e.button))for(;l!==this;l=l.parentNode||this)if(1===l.nodeType&&("click"!==e.type||!0!==l.disabled)){for(o=,a={},n=0;n<u;n++)void 0===ai=(r=tn).selector+" "&&(ai=r.needsContext?-1<k(i,this).index(l):k.find(i,this,null,l).length),ai&&o.push(r);o.length&&s.push({elem:l,handlers:o})}return l=this,u<t.length&&s.push({elem:l,handlers:t.slice(u)}),s},addProp:function(t,e){Object.defineProperty(k.Event.prototype,t,{enumerable:!0,configurable:!0,get:m(e)?function(){if(this.originalEvent)return e(this.originalEvent)}:function(){if(this.originalEvent)return this.originalEventt},set:function(e){Object.defineProperty(this,t,{enumerable:!0,configurable:!0,writable:!0,value:e})}})},fix:function(e){return ek.expando?e:new k.Event(e)},special:{load:{noBubble:!0},click:{setup:function(e){var t=this||e;return pe.test(t.type)&&t.click&&A(t,"input")&&De(t,"click",ke),!1},trigger:function(e){var t=this||e;return pe.test(t.type)&&t.click&&A(t,"input")&&De(t,"click"),!0},_default:function(e){var t=e.target;return pe.test(t.type)&&t.click&&A(t,"input")&&Q.get(t,"click")||A(t,"a")}},beforeunload:{postDispatch:function(e){void 0!==e.result&&e.originalEvent&&(e.originalEvent.returnValue=e.result)}}}},k.removeEvent=function(e,t,n){e.removeEventListener&&e.removeEventListener(t,n)},k.Event=function(e,t){if(!(this instanceof k.Event))return new k.Event(e,t);e&&e.type?(this.originalEvent=e,this.type=e.type,this.isDefaultPrevented=e.defaultPrevented||void 0===e.defaultPrevented&&!1===e.returnValue?ke:Se,this.target=e.target&&3===e.target.nodeType?e.target.parentNode:e.target,this.currentTarget=e.currentTarget,this.relatedTarget=e.relatedTarget):this.type=e,t&&k.extend(this,t),this.timeStamp=e&&e.timeStamp||Date.now(),thisk.expando=!0},k.Event.prototype={constructor:k.Event,isDefaultPrevented:Se,isPropagationStopped:Se,isImmediatePropagationStopped:Se,isSimulated:!1,preventDefault:function(){var e=this.originalEvent;this.isDefaultPrevented=ke,e&&!this.isSimulated&&e.preventDefault()},stopPropagation:function(){var e=this.originalEvent;this.isPropagationStopped=ke,e&&!this.isSimulated&&e.stopPropagation()},stopImmediatePropagation:function(){var e=this.originalEvent;this.isImmediatePropagationStopped=ke,e&&!this.isSimulated&&e.stopImmediatePropagation(),this.stopPropagation()}},k.each({altKey:!0,bubbles:!0,cancelable:!0,changedTouches:!0,ctrlKey:!0,detail:!0,eventPhase:!0,metaKey:!0,pageX:!0,pageY:!0,shiftKey:!0,view:!0,"char":!0,code:!0,charCode:!0,key:!0,keyCode:!0,button:!0,buttons:!0,clientX:!0,clientY:!0,offsetX:!0,offsetY:!0,pointerId:!0,pointerType:!0,screenX:!0,screenY:!0,targetTouches:!0,toElement:!0,touches:!0,which:function(e){var t=e.button;return null==e.which&&Te.test(e.type)?null!=e.charCode?e.charCode:e.keyCode:!e.which&&void 0!==t&&Ce.test(e.type)?1&t?1:2&t?3:4&t?2:0:e.which}},k.event.addProp),k.each({focus:"focusin",blur:"focusout"},function(e,t){k.event.speciale={setup:function(){return De(this,e,Ne),!1},trigger:function(){return De(this,e),!0},delegateType:t}}),k.each({mouseenter:"mouseover",mouseleave:"mouseout",pointerenter:"pointerover",pointerleave:"pointerout"},function(e,i){k.event.speciale={delegateType:i,bindType:i,handle:function(e){var t,n=e.relatedTarget,r=e.handleObj;return n&&(n===this||k.contains(this,n))||(e.type=r.origType,t=r.handler.apply(this,arguments),e.type=i),t}}}),k.fn.extend({on:function(e,t,n,r){return Ae(this,e,t,n,r)},one:function(e,t,n,r){return Ae(this,e,t,n,r,1)},off:function(e,t,n){var r,i;if(e&&e.preventDefault&&e.handleObj)return r=e.handleObj,k(e.delegateTarget).off(r.namespace?r.origType+"."+r.namespace:r.origType,r.selector,r.handler),this;if("object"==typeof e){for(i in e)this.off(i,t,ei);return this}return!1!==t&&"function"!=typeof t||(n=t,t=void 0),!1===n&&(n=Se),this.each(function(){k.event.remove(this,e,n,t)})}});var je=/<(?!area|br|col|embed|hr|img|input|link|meta|param)((a-z^\/\0>\x20\t\r\n\f*)^>*)\/>/gi,qe=/<script|<style|<link/i,Le=/checked\s*(?:^=|=\s*.checked.)/i,He=/^\s*<!(?:\CDATA\|--)|(?:\\|--)>\s*$/g;function Oe(e,t){return A(e,"table")&&A(11!==t.nodeType?t:t.firstChild,"tr")&&k(e).children("tbody")0||e}function Pe(e){return e.type=(null!==e.getAttribute("type"))+"/"+e.type,e}function Re(e){return"true/"===(e.type||"").slice(0,5)?e.type=e.type.slice(5):e.removeAttribute("type"),e}function Me(e,t){var n,r,i,o,a,s,u,l;if(1===t.nodeType){if(Q.hasData(e)&&(o=Q.access(e),a=Q.set(t,o),l=o.events))for(i in delete a.handle,a.events={},l)for(n=0,r=li.length;n<r;n++)k.event.add(t,i,lin);J.hasData(e)&&(s=J.access(e),u=k.extend({},s),J.set(t,u))}}function Ie(n,r,i,o){r=g.apply(,r);var e,t,a,s,u,l,c=0,f=n.length,p=f-1,d=r0,h=m(d);if(h||1<f&&"string"==typeof d&&!y.checkClone&&Le.test(d))return n.each(function(e){var t=n.eq(e);h&&(r0=d.call(this,e,t.html())),Ie(t,r,i,o)});if(f&&(t=(e=we(r,n0.ownerDocument,!1,n,o)).firstChild,1===e.childNodes.length&&(e=t),t||o)){for(s=(a=k.map(ve(e,"script"),Pe)).length;c<f;c++)u=e,c!==p&&(u=k.clone(u,!0,!0),s&&k.merge(a,ve(u,"script"))),i.call(nc,u,c);if(s)for(l=aa.length-1.ownerDocument,k.map(a,Re),c=0;c<s;c++)u=ac,he.test(u.type||"")&&!Q.access(u,"globalEval")&&k.contains(l,u)&&(u.src&&"module"!==(u.type||"").toLowerCase()?k._evalUrl&&!u.noModule&&k._evalUrl(u.src,{nonce:u.nonce||u.getAttribute("nonce")}):b(u.textContent.replace(He,""),u,l))}return n}function We(e,t,n){for(var r,i=t?k.filter(t,e):e,o=0;null!=(r=io);o++)n||1!==r.nodeType||k.cleanData(ve(r)),r.parentNode&&(n&&oe(r)&&ye(ve(r,"script")),r.parentNode.removeChild(r));return e}k.extend({htmlPrefilter:function(e){return e.replace(je,"<$1></$2>")},clone:function(e,t,n){var r,i,o,a,s,u,l,c=e.cloneNode(!0),f=oe(e);if(!(y.noCloneChecked||1!==e.nodeType&&11!==e.nodeType||k.isXMLDoc(e)))for(a=ve(c),r=0,i=(o=ve(e)).length;r<i;r++)s=or,u=ar,void 0,"input"===(l=u.nodeName.toLowerCase())&&pe.test(s.type)?u.checked=s.checked:"input"!==l&&"textarea"!==l||(u.defaultValue=s.defaultValue);if(t)if(n)for(o=o||ve(e),a=a||ve(c),r=0,i=o.length;r<i;r++)Me(or,ar);else Me(e,c);return 0<(a=ve(c,"script")).length&&ye(a,!f&&ve(e,"script")),c},cleanData:function(e){for(var t,n,r,i=k.event.special,o=0;void 0!==(n=eo);o++)if(G(n)){if(t=nQ.expando){if(t.events)for(r in t.events)ir?k.event.remove(n,r):k.removeEvent(n,r,t.handle);nQ.expando=void 0}nJ.expando&&(nJ.expando=void 0)}}}),k.fn.extend({detach:function(e){return We(this,e,!0)},remove:function(e){return We(this,e)},text:function(e){return _(this,function(e){return void 0===e?k.text(this):this.empty().each(function(){1!==this.nodeType&&11!==this.nodeType&&9!==this.nodeType||(this.textContent=e)})},null,e,arguments.length)},append:function(){return Ie(this,arguments,function(e){1!==this.nodeType&&11!==this.nodeType&&9!==this.nodeType||Oe(this,e).appendChild(e)})},prepend:function(){return Ie(this,arguments,function(e){if(1===this.nodeType||11===this.nodeType||9===this.nodeType){var t=Oe(this,e);t.insertBefore(e,t.firstChild)}})},before:function(){return Ie(this,arguments,function(e){this.parentNode&&this.parentNode.insertBefore(e,this)})},after:function(){return Ie(this,arguments,function(e){this.parentNode&&this.parentNode.insertBefore(e,this.nextSibling)})},empty:function(){for(var e,t=0;null!=(e=thist);t++)1===e.nodeType&&(k.cleanData(ve(e,!1)),e.textContent="");return this},clone:function(e,t){return e=null!=e&&e,t=null==t?e:t,this.map(function(){return k.clone(this,e,t)})},html:function(e){return _(this,function(e){var t=this0||{},n=0,r=this.length;if(void 0===e&&1===t.nodeType)return t.innerHTML;if("string"==typeof e&&!qe.test(e)&&!ge(de.exec(e)||"","")1.toLowerCase()){e=k.htmlPrefilter(e);try{for(;n<r;n++)1===(t=thisn||{}).nodeType&&(k.cleanData(ve(t,!1)),t.innerHTML=e);t=0}catch(e){}}t&&this.empty().append(e)},null,e,arguments.length)},replaceWith:function(){var n=;return Ie(this,arguments,function(e){var t=this.parentNode;k.inArray(this,n)<0&&(k.cleanData(ve(this)),t&&t.replaceChild(e,this))},n)}}),k.each({appendTo:"append",prependTo:"prepend",insertBefore:"before",insertAfter:"after",replaceAll:"replaceWith"},function(e,a){k.fne=function(e){for(var t,n=,r=k(e),i=r.length-1,o=0;o<=i;o++)t=o===i?this:this.clone(!0),k(ro)a(t),u.apply(n,t.get());return this.pushStack(n)}});var $e=new RegExp("^("+te+")(?!px)a-z%+$","i"),Fe=function(e){var t=e.ownerDocument.defaultView;return t&&t.opener||(t=C),t.getComputedStyle(e)},Be=new RegExp(re.join("|"),"i");function _e(e,t,n){var r,i,o,a,s=e.style;return(n=n||Fe(e))&&(""!==(a=n.getPropertyValue(t)||nt)||oe(e)||(a=k.style(e,t)),!y.pixelBoxStyles()&&$e.test(a)&&Be.test(t)&&(r=s.width,i=s.minWidth,o=s.maxWidth,s.minWidth=s.maxWidth=s.width=a,a=n.width,s.width=r,s.minWidth=i,s.maxWidth=o)),void 0!==a?a+"":a}function ze(e,t){return{get:function(){if(!e())return(this.get=t).apply(this,arguments);delete this.get}}}!function(){function e(){if(u){s.style.cssText="position:absolute;left:-11111px;width:60px;margin-top:1px;padding:0;border:0",u.style.cssText="position:relative;display:block;box-sizing:border-box;overflow:scroll;margin:auto;border:1px;padding:1px;width:60%;top:1%",ie.appendChild(s).appendChild(u);var e=C.getComputedStyle(u);n="1%"!==e.top,a=12===t(e.marginLeft),u.style.right="60%",o=36===t(e.right),r=36===t(e.width),u.style.position="absolute",i=12===t(u.offsetWidth/3),ie.removeChild(s),u=null}}function t(e){return Math.round(parseFloat(e))}var n,r,i,o,a,s=E.createElement("div"),u=E.createElement("div");u.style&&(u.style.backgroundClip="content-box",u.cloneNode(!0).style.backgroundClip="",y.clearCloneStyle="content-box"===u.style.backgroundClip,k.extend(y,{boxSizingReliable:function(){return e(),r},pixelBoxStyles:function(){return e(),o},pixelPosition:function(){return e(),n},reliableMarginLeft:function(){return e(),a},scrollboxSize:function(){return e(),i}}))}();var Ue="Webkit","Moz","ms",Xe=E.createElement("div").style,Ve={};function Ge(e){var t=k.cssPropse||Vee;return t||(e in Xe?e:Vee=function(e){var t=e0.toUpperCase()+e.slice(1),n=Ue.length;while(n--)if((e=Uen+t)in Xe)return e}(e)||e)}var Ye=/^(none|table(?!-cea).+)/,Qe=/^--/,Je={position:"absolute",visibility:"hidden",display:"block"},Ke={letterSpacing:"0",fontWeight:"400"};function Ze(e,t,n){var r=ne.exec(t);return r?Math.max(0,r2-(n||0))+(r3||"px"):t}function et(e,t,n,r,i,o){var a="width"===t?1:0,s=0,u=0;if(n===(r?"border":"content"))return 0;for(;a<4;a+=2)"margin"===n&&(u+=k.css(e,n+rea,!0,i)),r?("content"===n&&(u-=k.css(e,"padding"+rea,!0,i)),"margin"!==n&&(u-=k.css(e,"border"+rea+"Width",!0,i))):(u+=k.css(e,"padding"+rea,!0,i),"padding"!==n?u+=k.css(e,"border"+rea+"Width",!0,i):s+=k.css(e,"border"+rea+"Width",!0,i));return!r&&0<=o&&(u+=Math.max(0,Math.ceil(e"offset"+t0.toUpperCase()+t.slice(1)-o-u-s-.5))||0),u}function tt(e,t,n){var r=Fe(e),i=(!y.boxSizingReliable()||n)&&"border-box"===k.css(e,"boxSizing",!1,r),o=i,a=_e(e,t,r),s="offset"+t0.toUpperCase()+t.slice(1);if($e.test(a)){if(!n)return a;a="auto"}return(!y.boxSizingReliable()&&i||"auto"===a||!parseFloat(a)&&"inline"===k.css(e,"display",!1,r))&&e.getClientRects().length&&(i="border-box"===k.css(e,"boxSizing",!1,r),(o=s in e)&&(a=es)),(a=parseFloat(a)||0)+et(e,t,n||(i?"border":"content"),o,r,a)+"px"}function nt(e,t,n,r,i){return new nt.prototype.init(e,t,n,r,i)}k.extend({cssHooks:{opacity:{get:function(e,t){if(t){var n=_e(e,"opacity");return""===n?"1":n}}}},cssNumber:{animationIterationCount:!0,columnCount:!0,fillOpacity:!0,flexGrow:!0,flexShrink:!0,fontWeight:!0,gridArea:!0,gridColumn:!0,gridColumnEnd:!0,gridColumnStart:!0,gridRow:!0,gridRowEnd:!0,gridRowStart:!0,lineHeight:!0,opacity:!0,order:!0,orphans:!0,widows:!0,zIndex:!0,zoom:!0},cssProps:{},style:function(e,t,n,r){if(e&&3!==e.nodeType&&8!==e.nodeType&&e.style){var i,o,a,s=V(t),u=Qe.test(t),l=e.style;if(u||(t=Ge(s)),a=k.cssHookst||k.cssHookss,void 0===n)return a&&"get"in a&&void 0!==(i=a.get(e,!1,r))?i:lt;"string"===(o=typeof n)&&(i=ne.exec(n))&&i1&&(n=le(e,t,i),o="number"),null!=n&&n==n&&("number"!==o||u||(n+=i&&i3||(k.cssNumbers?"":"px")),y.clearCloneStyle||""!==n||0!==t.indexOf("background")||(lt="inherit"),a&&"set"in a&&void 0===(n=a.set(e,n,r))||(u?l.setProperty(t,n):lt=n))}},css:function(e,t,n,r){var i,o,a,s=V(t);return Qe.test(t)||(t=Ge(s)),(a=k.cssHookst||k.cssHookss)&&"get"in a&&(i=a.get(e,!0,n)),void 0===i&&(i=_e(e,t,r)),"normal"===i&&t in Ke&&(i=Ket),""===n||n?(o=parseFloat(i),!0===n||isFinite(o)?o||0:i):i}}),k.each("height","width",function(e,u){k.cssHooksu={get:function(e,t,n){if(t)return!Ye.test(k.css(e,"display"))||e.getClientRects().length&&e.getBoundingClientRect().width?tt(e,u,n):ue(e,Je,function(){return tt(e,u,n)})},set:function(e,t,n){var r,i=Fe(e),o=!y.scrollboxSize()&&"absolute"===i.position,a=(o||n)&&"border-box"===k.css(e,"boxSizing",!1,i),s=n?et(e,u,n,a,i):0;return a&&o&&(s-=Math.ceil(e"offset"+u0.toUpperCase()+u.slice(1)-parseFloat(iu)-et(e,u,"border",!1,i)-.5)),s&&(r=ne.exec(t))&&"px"!==(r3||"px")&&(e.styleu=t,t=k.css(e,u)),Ze(0,t,s)}}}),k.cssHooks.marginLeft=ze(y.reliableMarginLeft,function(e,t){if(t)return(parseFloat(_e(e,"marginLeft"))||e.getBoundingClientRect().left-ue(e,{marginLeft:0},function(){return e.getBoundingClientRect().left}))+"px"}),k.each({margin:"",padding:"",border:"Width"},function(i,o){k.cssHooksi+o={expand:function(e){for(var t=0,n={},r="string"==typeof e?e.split(" "):e;t<4;t++)ni+ret+o=rt||rt-2||r0;return n}},"margin"!==i&&(k.cssHooksi+o.set=Ze)}),k.fn.extend({css:function(e,t){return _(this,function(e,t,n){var r,i,o={},a=0;if(Array.isArray(t)){for(r=Fe(e),i=t.length;a<i;a++)ota=k.css(e,ta,!1,r);return o}return void 0!==n?k.style(e,t,n):k.css(e,t)},e,t,1<arguments.length)}}),((k.Tween=nt).prototype={constructor:nt,init:function(e,t,n,r,i,o){this.elem=e,this.prop=n,this.easing=i||k.easing._default,this.options=t,this.start=this.now=this.cur(),this.end=r,this.unit=o||(k.cssNumbern?"":"px")},cur:function(){var e=nt.propHooksthis.prop;return e&&e.get?e.get(this):nt.propHooks._default.get(this)},run:function(e){var t,n=nt.propHooksthis.prop;return this.options.duration?this.pos=t=k.easingthis.easing(e,this.options.duration*e,0,1,this.options.duration):this.pos=t=e,this.now=(this.end-this.start)*t+this.start,this.options.step&&this.options.step.call(this.elem,this.now,this),n&&n.set?n.set(this):nt.propHooks._default.set(this),this}}).init.prototype=nt.prototype,(nt.propHooks={_default:{get:function(e){var t;return 1!==e.elem.nodeType||null!=e.eleme.prop&&null==e.elem.stylee.prop?e.eleme.prop:(t=k.css(e.elem,e.prop,""))&&"auto"!==t?t:0},set:function(e){k.fx.stepe.prop?k.fx.stepe.prop(e):1!==e.elem.nodeType||!k.cssHookse.prop&&null==e.elem.styleGe(e.prop)?e.eleme.prop=e.now:k.style(e.elem,e.prop,e.now+e.unit)}}}).scrollTop=nt.propHooks.scrollLeft={set:function(e){e.elem.nodeType&&e.elem.parentNode&&(e.eleme.prop=e.now)}},k.easing={linear:function(e){return e},swing:function(e){return.5-Math.cos(e*Math.PI)/2},_default:"swing"},k.fx=nt.prototype.init,k.fx.step={};var rt,it,ot,at,st=/^(?:toggle|show|hide)$/,ut=/queueHooks$/;function lt(){it&&(!1===E.hidden&&C.requestAnimationFrame?C.requestAnimationFrame(lt):C.setTimeout(lt,k.fx.interval),k.fx.tick())}function ct(){return C.setTimeout(function(){rt=void 0}),rt=Date.now()}function ft(e,t){var n,r=0,i={height:e};for(t=t?1:0;r<4;r+=2-t)i"margin"+(n=rer)=i"padding"+n=e;return t&&(i.opacity=i.width=e),i}function pt(e,t,n){for(var r,i=(dt.tweenerst||).concat(dt.tweeners"*"),o=0,a=i.length;o<a;o++)if(r=io.call(n,t,e))return r}function dt(o,e,t){var n,a,r=0,i=dt.prefilters.length,s=k.Deferred().always(function(){delete u.elem}),u=function(){if(a)return!1;for(var e=rt||ct(),t=Math.max(0,l.startTime+l.duration-e),n=1-(t/l.duration||0),r=0,i=l.tweens.length;r<i;r++)l.tweensr.run(n);return s.notifyWith(o,l,n,t),n<1&&i?t:(i||s.notifyWith(o,l,1,0),s.resolveWith(o,l),!1)},l=s.promise({elem:o,props:k.extend({},e),opts:k.extend(!0,{specialEasing:{},easing:k.easing._default},t),originalProperties:e,originalOptions:t,startTime:rt||ct(),duration:t.duration,tweens:,createTween:function(e,t){var n=k.Tween(o,l.opts,e,t,l.opts.specialEasinge||l.opts.easing);return l.tweens.push(n),n},stop:function(e){var t=0,n=e?l.tweens.length:0;if(a)return this;for(a=!0;t<n;t++)l.tweenst.run(1);return e?(s.notifyWith(o,l,1,0),s.resolveWith(o,l,e)):s.rejectWith(o,l,e),this}}),c=l.props;for(!function(e,t){var n,r,i,o,a;for(n in e)if(i=tr=V(n),o=en,Array.isArray(o)&&(i=o1,o=en=o0),n!==r&&(er=o,delete en),(a=k.cssHooksr)&&"expand"in a)for(n in o=a.expand(o),delete er,o)n in e||(en=on,tn=i);else tr=i}(c,l.opts.specialEasing);r<i;r++)if(n=dt.prefiltersr.call(l,o,c,l.opts))return m(n.stop)&&(k._queueHooks(l.elem,l.opts.queue).stop=n.stop.bind(n)),n;return k.map(c,pt,l),m(l.opts.start)&&l.opts.start.call(o,l),l.progress(l.opts.progress).done(l.opts.done,l.opts.complete).fail(l.opts.fail).always(l.opts.always),k.fx.timer(k.extend(u,{elem:o,anim:l,queue:l.opts.queue})),l}k.Animation=k.extend(dt,{tweeners:{"*":function(e,t){var n=this.createTween(e,t);return le(n.elem,e,ne.exec(t),n),n}},tweener:function(e,t){m(e)?(t=e,e="*"):e=e.match(R);for(var n,r=0,i=e.length;r<i;r++)n=er,dt.tweenersn=dt.tweenersn||,dt.tweenersn.unshift(t)},prefilters:function(e,t,n){var r,i,o,a,s,u,l,c,f="width"in t||"height"in t,p=this,d={},h=e.style,g=e.nodeType&&se(e),v=Q.get(e,"fxshow");for(r in n.queue||(null==(a=k._queueHooks(e,"fx")).unqueued&&(a.unqueued=0,s=a.empty.fire,a.empty.fire=function(){a.unqueued||s()}),a.unqueued++,p.always(function(){p.always(function(){a.unqueued--,k.queue(e,"fx").length||a.empty.fire()})})),t)if(i=tr,st.test(i)){if(delete tr,o=o||"toggle"===i,i===(g?"hide":"show")){if("show"!==i||!v||void 0===vr)continue;g=!0}dr=v&&vr||k.style(e,r)}if((u=!k.isEmptyObject(t))||!k.isEmptyObject(d))for(r in f&&1===e.nodeType&&(n.overflow=h.overflow,h.overflowX,h.overflowY,null==(l=v&&v.display)&&(l=Q.get(e,"display")),"none"===(c=k.css(e,"display"))&&(l?c=l:(fe(e,!0),l=e.style.display||l,c=k.css(e,"display"),fe(e))),("inline"===c||"inline-block"===c&&null!=l)&&"none"===k.css(e,"float")&&(u||(p.done(function(){h.display=l}),null==l&&(c=h.display,l="none"===c?"":c)),h.display="inline-block")),n.overflow&&(h.overflow="hidden",p.always(function(){h.overflow=n.overflow0,h.overflowX=n.overflow1,h.overflowY=n.overflow2})),u=!1,d)u||(v?"hidden"in v&&(g=v.hidden):v=Q.access(e,"fxshow",{display:l}),o&&(v.hidden=!g),g&&fe(e,!0),p.done(function(){for(r in g||fe(e),Q.remove(e,"fxshow"),d)k.style(e,r,dr)})),u=pt(g?vr:0,r,p),r in v||(vr=u.start,g&&(u.end=u.start,u.start=0))},prefilter:function(e,t){t?dt.prefilters.unshift(e):dt.prefilters.push(e)}}),k.speed=function(e,t,n){var r=e&&"object"==typeof e?k.extend({},e):{complete:n||!n&&t||m(e)&&e,duration:e,easing:n&&t||t&&!m(t)&&t};return k.fx.off?r.duration=0:"number"!=typeof r.duration&&(r.duration in k.fx.speeds?r.duration=k.fx.speedsr.duration:r.duration=k.fx.speeds._default),null!=r.queue&&!0!==r.queue||(r.queue="fx"),r.old=r.complete,r.complete=function(){m(r.old)&&r.old.call(this),r.queue&&k.dequeue(this,r.queue)},r},k.fn.extend({fadeTo:function(e,t,n,r){return this.filter(se).css("opacity",0).show().end().animate({opacity:t},e,n,r)},animate:function(t,e,n,r){var i=k.isEmptyObject(t),o=k.speed(e,n,r),a=function(){var e=dt(this,k.extend({},t),o);(i||Q.get(this,"finish"))&&e.stop(!0)};return a.finish=a,i||!1===o.queue?this.each(a):this.queue(o.queue,a)},stop:function(i,e,o){var a=function(e){var t=e.stop;delete e.stop,t(o)};return"string"!=typeof i&&(o=e,e=i,i=void 0),e&&!1!==i&&this.queue(i||"fx",),this.each(function(){var e=!0,t=null!=i&&i+"queueHooks",n=k.timers,r=Q.get(this);if(t)rt&&rt.stop&&a(rt);else for(t in r)rt&&rt.stop&&ut.test(t)&&a(rt);for(t=n.length;t--;)nt.elem!==this||null!=i&&nt.queue!==i||(nt.anim.stop(o),e=!1,n.splice(t,1));!e&&o||k.dequeue(this,i)})},finish:function(a){return!1!==a&&(a=a||"fx"),this.each(function(){var e,t=Q.get(this),n=ta+"queue",r=ta+"queueHooks",i=k.timers,o=n?n.length:0;for(t.finish=!0,k.queue(this,a,),r&&r.stop&&r.stop.call(this,!0),e=i.length;e--;)ie.elem===this&&ie.queue===a&&(ie.anim.stop(!0),i.splice(e,1));for(e=0;e<o;e++)ne&&ne.finish&&ne.finish.call(this);delete t.finish})}}),k.each("toggle","show","hide",function(e,r){var i=k.fnr;k.fnr=function(e,t,n){return null==e||"boolean"==typeof e?i.apply(this,arguments):this.animate(ft(r,!0),e,t,n)}}),k.each({slideDown:ft("show"),slideUp:ft("hide"),slideToggle:ft("toggle"),fadeIn:{opacity:"show"},fadeOut:{opacity:"hide"},fadeToggle:{opacity:"toggle"}},function(e,r){k.fne=function(e,t,n){return this.animate(r,e,t,n)}}),k.timers=,k.fx.tick=function(){var e,t=0,n=k.timers;for(rt=Date.now();t<n.length;t++)(e=nt)()||nt!==e||n.splice(t--,1);n.length||k.fx.stop(),rt=void 0},k.fx.timer=function(e){k.timers.push(e),k.fx.start()},k.fx.interval=13,k.fx.start=function(){it||(it=!0,lt())},k.fx.stop=function(){it=null},k.fx.speeds={slow:600,fast:200,_default:400},k.fn.delay=function(r,e){return r=k.fx&&k.fx.speedsr||r,e=e||"fx",this.queue(e,function(e,t){var n=C.setTimeout(e,r);t.stop=function(){C.clearTimeout(n)}})},ot=E.createElement("input"),at=E.createElement("select").appendChild(E.createElement("option")),ot.type="checkbox",y.checkOn=""!==ot.value,y.optSelected=at.selected,(ot=E.createElement("input")).value="t",ot.type="radio",y.radioValue="t"===ot.value;var ht,gt=k.expr.attrHandle;k.fn.extend({attr:function(e,t){return _(this,k.attr,e,t,1<arguments.length)},removeAttr:function(e){return this.each(function(){k.removeAttr(this,e)})}}),k.extend({attr:function(e,t,n){var r,i,o=e.nodeType;if(3!==o&&8!==o&&2!==o)return"undefined"==typeof e.getAttribute?k.prop(e,t,n):(1===o&&k.isXMLDoc(e)||(i=k.attrHookst.toLowerCase()||(k.expr.match.bool.test(t)?ht:void 0)),void 0!==n?null===n?void k.removeAttr(e,t):i&&"set"in i&&void 0!==(r=i.set(e,n,t))?r:(e.setAttribute(t,n+""),n):i&&"get"in i&&null!==(r=i.get(e,t))?r:null==(r=k.find.attr(e,t))?void 0:r)},attrHooks:{type:{set:function(e,t){if(!y.radioValue&&"radio"===t&&A(e,"input")){var n=e.value;return e.setAttribute("type",t),n&&(e.value=n),t}}}},removeAttr:function(e,t){var n,r=0,i=t&&t.match(R);if(i&&1===e.nodeType)while(n=ir++)e.removeAttribute(n)}}),ht={set:function(e,t,n){return!1===t?k.removeAttr(e,n):e.setAttribute(n,n),n}},k.each(k.expr.match.bool.source.match(/\w+/g),function(e,t){var a=gtt||k.find.attr;gtt=function(e,t,n){var r,i,o=t.toLowerCase();return n||(i=gto,gto=r,r=null!=a(e,t,n)?o:null,gto=i),r}});var vt=/^(?:input|select|textarea|button)$/i,yt=/^(?:a|area)$/i;function mt(e){return(e.match(R)||).join(" ")}function xt(e){return e.getAttribute&&e.getAttribute("class")||""}function bt(e){return Array.isArray(e)?e:"string"==typeof e&&e.match(R)||}k.fn.extend({prop:function(e,t){return _(this,k.prop,e,t,1<arguments.length)},removeProp:function(e){return this.each(function(){delete thisk.propFixe||e})}}),k.extend({prop:function(e,t,n){var r,i,o=e.nodeType;if(3!==o&&8!==o&&2!==o)return 1===o&&k.isXMLDoc(e)||(t=k.propFixt||t,i=k.propHookst),void 0!==n?i&&"set"in i&&void 0!==(r=i.set(e,n,t))?r:et=n:i&&"get"in i&&null!==(r=i.get(e,t))?r:et},propHooks:{tabIndex:{get:function(e){var t=k.find.attr(e,"tabindex");return t?parseInt(t,10):vt.test(e.nodeName)||yt.test(e.nodeName)&&e.href?0:-1}}},propFix:{"for":"htmlFor","class":"className"}}),y.optSelected||(k.propHooks.selected={get:function(e){var t=e.parentNode;return t&&t.parentNode&&t.parentNode.selectedIndex,null},set:function(e){var t=e.parentNode;t&&(t.selectedIndex,t.parentNode&&t.parentNode.selectedIndex)}}),k.each("tabIndex","readOnly","maxLength","cellSpacing","cellPadding","rowSpan","colSpan","useMap","frameBorder","contentEditable",function(){k.propFixthis.toLowerCase()=this}),k.fn.extend({addClass:function(t){var e,n,r,i,o,a,s,u=0;if(m(t))return this.each(function(e){k(this).addClass(t.call(this,e,xt(this)))});if((e=bt(t)).length)while(n=thisu++)if(i=xt(n),r=1===n.nodeType&&" "+mt(i)+" "){a=0;while(o=ea++)r.indexOf(" "+o+" ")<0&&(r+=o+" ");i!==(s=mt(r))&&n.setAttribute("class",s)}return this},removeClass:function(t){var e,n,r,i,o,a,s,u=0;if(m(t))return this.each(function(e){k(this).removeClass(t.call(this,e,xt(this)))});if(!arguments.length)return this.attr("class","");if((e=bt(t)).length)while(n=thisu++)if(i=xt(n),r=1===n.nodeType&&" "+mt(i)+" "){a=0;while(o=ea++)while(-1<r.indexOf(" "+o+" "))r=r.replace(" "+o+" "," ");i!==(s=mt(r))&&n.setAttribute("class",s)}return this},toggleClass:function(i,t){var o=typeof i,a="string"===o||Array.isArray(i);return"boolean"==typeof t&&a?t?this.addClass(i):this.removeClass(i):m(i)?this.each(function(e){k(this).toggleClass(i.call(this,e,xt(this),t),t)}):this.each(function(){var e,t,n,r;if(a){t=0,n=k(this),r=bt(i);while(e=rt++)n.hasClass(e)?n.removeClass(e):n.addClass(e)}else void 0!==i&&"boolean"!==o||((e=xt(this))&&Q.set(this,"__className__",e),this.setAttribute&&this.setAttribute("class",e||!1===i?"":Q.get(this,"__className__")||""))})},hasClass:function(e){var t,n,r=0;t=" "+e+" ";while(n=thisr++)if(1===n.nodeType&&-1<(" "+mt(xt(n))+" ").indexOf(t))return!0;return!1}});var wt=/\r/g;k.fn.extend({val:function(n){var r,e,i,t=this0;return arguments.length?(i=m(n),this.each(function(e){var t;1===this.nodeType&&(null==(t=i?n.call(this,e,k(this).val()):n)?t="":"number"==typeof t?t+="":Array.isArray(t)&&(t=k.map(t,function(e){return null==e?"":e+""})),(r=k.valHooksthis.type||k.valHooksthis.nodeName.toLowerCase())&&"set"in r&&void 0!==r.set(this,t,"value")||(this.value=t))})):t?(r=k.valHookst.type||k.valHookst.nodeName.toLowerCase())&&"get"in r&&void 0!==(e=r.get(t,"value"))?e:"string"==typeof(e=t.value)?e.replace(wt,""):null==e?"":e:void 0}}),k.extend({valHooks:{option:{get:function(e){var t=k.find.attr(e,"value");return null!=t?t:mt(k.text(e))}},select:{get:function(e){var t,n,r,i=e.options,o=e.selectedIndex,a="select-one"===e.type,s=a?null:,u=a?o+1:i.length;for(r=o<0?u:a?o:0;r<u;r++)if(((n=ir).selected||r===o)&&!n.disabled&&(!n.parentNode.disabled||!A(n.parentNode,"optgroup"))){if(t=k(n).val(),a)return t;s.push(t)}return s},set:function(e,t){var n,r,i=e.options,o=k.makeArray(t),a=i.length;while(a--)((r=ia).selected=-1<k.inArray(k.valHooks.option.get(r),o))&&(n=!0);return n||(e.selectedIndex=-1),o}}}}),k.each("radio","checkbox",function(){k.valHooksthis={set:function(e,t){if(Array.isArray(t))return e.checked=-1<k.inArray(k(e).val(),t)}},y.checkOn||(k.valHooksthis.get=function(e){return null===e.getAttribute("value")?"on":e.value})}),y.focusin="onfocusin"in C;var Tt=/^(?:focusinfocus|focusoutblur)$/,Ct=function(e){e.stopPropagation()};k.extend(k.event,{trigger:function(e,t,n,r){var i,o,a,s,u,l,c,f,p=n||E,d=v.call(e,"type")?e.type:e,h=v.call(e,"namespace")?e.namespace.split("."):;if(o=f=a=n=n||E,3!==n.nodeType&&8!==n.nodeType&&!Tt.test(d+k.event.triggered)&&(-1<d.indexOf(".")&&(d=(h=d.split(".")).shift(),h.sort()),u=d.indexOf(":")<0&&"on"+d,(e=ek.expando?e:new k.Event(d,"object"==typeof e&&e)).isTrigger=r?2:3,e.namespace=h.join("."),e.rnamespace=e.namespace?new RegExp("(^|\\.)"+h.join("\\.(?:.*\\.|)")+"(\\.|$)"):null,e.result=void 0,e.target||(e.target=n),t=null==t?e:k.makeArray(t,e),c=k.event.speciald||{},r||!c.trigger||!1!==c.trigger.apply(n,t))){if(!r&&!c.noBubble&&!x(n)){for(s=c.delegateType||d,Tt.test(s+d)||(o=o.parentNode);o;o=o.parentNode)p.push(o),a=o;a===(n.ownerDocument||E)&&p.push(a.defaultView||a.parentWindow||C)}i=0;while((o=pi++)&&!e.isPropagationStopped())f=o,e.type=1<i?s:c.bindType||d,(l=(Q.get(o,"events")||{})e.type&&Q.get(o,"handle"))&&l.apply(o,t),(l=u&&ou)&&l.apply&&G(o)&&(e.result=l.apply(o,t),!1===e.result&&e.preventDefault());return e.type=d,r||e.isDefaultPrevented()||c._default&&!1!==c._default.apply(p.pop(),t)||!G(n)||u&&m(nd)&&!x(n)&&((a=nu)&&(nu=null),k.event.triggered=d,e.isPropagationStopped()&&f.addEventListener(d,Ct),nd(),e.isPropagationStopped()&&f.removeEventListener(d,Ct),k.event.triggered=void 0,a&&(nu=a)),e.result}},simulate:function(e,t,n){var r=k.extend(new k.Event,n,{type:e,isSimulated:!0});k.event.trigger(r,null,t)}}),k.fn.extend({trigger:function(e,t){return this.each(function(){k.event.trigger(e,t,this)})},triggerHandler:function(e,t){var n=this0;if(n)return k.event.trigger(e,t,n,!0)}}),y.focusin||k.each({focus:"focusin",blur:"focusout"},function(n,r){var i=function(e){k.event.simulate(r,e.target,k.event.fix(e))};k.event.specialr={setup:function(){var e=this.ownerDocument||this,t=Q.access(e,r);t||e.addEventListener(n,i,!0),Q.access(e,r,(t||0)+1)},teardown:function(){var e=this.ownerDocument||this,t=Q.access(e,r)-1;t?Q.access(e,r,t):(e.removeEventListener(n,i,!0),Q.remove(e,r))}}});var Et=C.location,kt=Date.now(),St=/\?/;k.parseXML=function(e){var t;if(!e||"string"!=typeof e)return null;try{t=(new C.DOMParser).parseFromString(e,"text/xml")}catch(e){t=void 0}return t&&!t.getElementsByTagName("parsererror").length||k.error("Invalid XML: "+e),t};var Nt=/\\$/,At=/\r?\n/g,Dt=/^(?:submit|button|image|reset|file)$/i,jt=/^(?:input|select|textarea|keygen)/i;function qt(n,e,r,i){var t;if(Array.isArray(e))k.each(e,function(e,t){r||Nt.test(n)?i(n,t):qt(n+""+("object"==typeof t&&null!=t?e:"")+"",t,r,i)});else if(r||"object"!==w(e))i(n,e);else for(t in e)qt(n+""+t+"",et,r,i)}k.param=function(e,t){var n,r=,i=function(e,t){var n=m(t)?t():t;rr.length=encodeURIComponent(e)+"="+encodeURIComponent(null==n?"":n)};if(null==e)return"";if(Array.isArray(e)||e.jquery&&!k.isPlainObject(e))k.each(e,function(){i(this.name,this.value)});else for(n in e)qt(n,en,t,i);return r.join("&")},k.fn.extend({serialize:function(){return k.param(this.serializeArray())},serializeArray:function(){return this.map(function(){var e=k.prop(this,"elements");return e?k.makeArray(e):this}).filter(function(){var e=this.type;return this.name&&!k(this).is(":disabled")&&jt.test(this.nodeName)&&!Dt.test(e)&&(this.checked||!pe.test(e))}).map(function(e,t){var n=k(this).val();return null==n?null:Array.isArray(n)?k.map(n,function(e){return{name:t.name,value:e.replace(At,"\r\n")}}):{name:t.name,value:n.replace(At,"\r\n")}}).get()}});var Lt=/%20/g,Ht=/#.*$/,Ot=/(?&)_=^&*/,Pt=/^(.*?): \t*(^\r\n*)$/gm,Rt=/^(?:GET|HEAD)$/,Mt=/^\/\//,It={},Wt={},$t="*/".concat("*"),Ft=E.createElement("a");function Bt(o){return function(e,t){"string"!=typeof e&&(t=e,e="*");var n,r=0,i=e.toLowerCase().match(R)||;if(m(t))while(n=ir++)"+"===n0?(n=n.slice(1)||"*",(on=on||).unshift(t)):(on=on||).push(t)}}function _t(t,i,o,a){var s={},u=t===Wt;function l(e){var r;return se=!0,k.each(te||,function(e,t){var n=t(i,o,a);return"string"!=typeof n||u||sn?u?!(r=n):void 0:(i.dataTypes.unshift(n),l(n),!1)}),r}return l(i.dataTypes0)||!s"*"&&l("*")}function zt(e,t){var n,r,i=k.ajaxSettings.flatOptions||{};for(n in t)void 0!==tn&&((in?e:r||(r={}))n=tn);return r&&k.extend(!0,e,r),e}Ft.href=Et.href,k.extend({active:0,lastModified:{},etag:{},ajaxSettings:{url:Et.href,type:"GET",isLocal:/^(?:about|app|app-storage|.+-extension|file|res|widget):$/.test(Et.protocol),global:!0,processData:!0,async:!0,contentType:"application/x-www-form-urlencoded; charset=UTF-8",accepts:{"*":$t,text:"text/plain",html:"text/html",xml:"application/xml, text/xml",json:"application/json, text/javascript"},contents:{xml:/\bxml\b/,html:/\bhtml/,json:/\bjson\b/},responseFields:{xml:"responseXML",text:"responseText",json:"responseJSON"},converters:{"* text":String,"text html":!0,"text json":JSON.parse,"text xml":k.parseXML},flatOptions:{url:!0,context:!0}},ajaxSetup:function(e,t){return t?zt(zt(e,k.ajaxSettings),t):zt(k.ajaxSettings,e)},ajaxPrefilter:Bt(It),ajaxTransport:Bt(Wt),ajax:function(e,t){"object"==typeof e&&(t=e,e=void 0),t=t||{};var c,f,p,n,d,r,h,g,i,o,v=k.ajaxSetup({},t),y=v.context||v,m=v.context&&(y.nodeType||y.jquery)?k(y):k.event,x=k.Deferred(),b=k.Callbacks("once memory"),w=v.statusCode||{},a={},s={},u="canceled",T={readyState:0,getResponseHeader:function(e){var t;if(h){if(!n){n={};while(t=Pt.exec(p))nt1.toLowerCase()+" "=(nt1.toLowerCase()+" "||).concat(t2)}t=ne.toLowerCase()+" "}return null==t?null:t.join(", ")},getAllResponseHeaders:function(){return h?p:null},setRequestHeader:function(e,t){return null==h&&(e=se.toLowerCase()=se.toLowerCase()||e,ae=t),this},overrideMimeType:function(e){return null==h&&(v.mimeType=e),this},statusCode:function(e){var t;if(e)if(h)T.always(eT.status);else for(t in e)wt=wt,et;return this},abort:function(e){var t=e||u;return c&&c.abort(t),l(0,t),this}};if(x.promise(T),v.url=((e||v.url||Et.href)+"").replace(Mt,Et.protocol+"//"),v.type=t.method||t.type||v.method||v.type,v.dataTypes=(v.dataType||"*").toLowerCase().match(R)||"",null==v.crossDomain){r=E.createElement("a");try{r.href=v.url,r.href=r.href,v.crossDomain=Ft.protocol+"//"+Ft.host!=r.protocol+"//"+r.host}catch(e){v.crossDomain=!0}}if(v.data&&v.processData&&"string"!=typeof v.data&&(v.data=k.param(v.data,v.traditional)),_t(It,v,t,T),h)return T;for(i in(g=k.event&&v.global)&&0==k.active++&&k.event.trigger("ajaxStart"),v.type=v.type.toUpperCase(),v.hasContent=!Rt.test(v.type),f=v.url.replace(Ht,""),v.hasContent?v.data&&v.processData&&0===(v.contentType||"").indexOf("application/x-www-form-urlencoded")&&(v.data=v.data.replace(Lt,"+")):(o=v.url.slice(f.length),v.data&&(v.processData||"string"==typeof v.data)&&(f+=(St.test(f)?"&":"?")+v.data,delete v.data),!1===v.cache&&(f=f.replace(Ot,"$1"),o=(St.test(f)?"&":"?")+"_="+kt+++o),v.url=f+o),v.ifModified&&(k.lastModifiedf&&T.setRequestHeader("If-Modified-Since",k.lastModifiedf),k.etagf&&T.setRequestHeader("If-None-Match",k.etagf)),(v.data&&v.hasContent&&!1!==v.contentType||t.contentType)&&T.setRequestHeader("Content-Type",v.contentType),T.setRequestHeader("Accept",v.dataTypes0&&v.acceptsv.dataTypes0?v.acceptsv.dataTypes0+("*"!==v.dataTypes0?", "+$t+"; q=0.01":""):v.accepts"*"),v.headers)T.setRequestHeader(i,v.headersi);if(v.beforeSend&&(!1===v.beforeSend.call(y,T,v)||h))return T.abort();if(u="abort",b.add(v.complete),T.done(v.success),T.fail(v.error),c=_t(Wt,v,t,T)){if(T.readyState=1,g&&m.trigger("ajaxSend",T,v),h)return T;v.async&&0<v.timeout&&(d=C.setTimeout(function(){T.abort("timeout")},v.timeout));try{h=!1,c.send(a,l)}catch(e){if(h)throw e;l(-1,e)}}else l(-1,"No Transport");function l(e,t,n,r){var i,o,a,s,u,l=t;h||(h=!0,d&&C.clearTimeout(d),c=void 0,p=r||"",T.readyState=0<e?4:0,i=200<=e&&e<300||304===e,n&&(s=function(e,t,n){var r,i,o,a,s=e.contents,u=e.dataTypes;while("*"===u0)u.shift(),void 0===r&&(r=e.mimeType||t.getResponseHeader("Content-Type"));if(r)for(i in s)if(si&&si.test(r)){u.unshift(i);break}if(u0in n)o=u0;else{for(i in n){if(!u0||e.convertersi+" "+u0){o=i;break}a||(a=i)}o=o||a}if(o)return o!==u0&&u.unshift(o),no}(v,T,n)),s=function(e,t,n,r){var i,o,a,s,u,l={},c=e.dataTypes.slice();if(c1)for(a in e.converters)la.toLowerCase()=e.convertersa;o=c.shift();while(o)if(e.responseFieldso&&(ne.responseFieldso=t),!u&&r&&e.dataFilter&&(t=e.dataFilter(t,e.dataType)),u=o,o=c.shift())if("*"===o)o=u;else if("*"!==u&&u!==o){if(!(a=lu+" "+o||l"* "+o))for(i in l)if((s=i.split(" "))1===o&&(a=lu+" "+s0||l"* "+s0)){!0===a?a=li:!0!==li&&(o=s0,c.unshift(s1));break}if(!0!==a)if(a&&e"throws")t=a(t);else try{t=a(t)}catch(e){return{state:"parsererror",error:a?e:"No conversion from "+u+" to "+o}}}return{state:"success",data:t}}(v,s,T,i),i?(v.ifModified&&((u=T.getResponseHeader("Last-Modified"))&&(k.lastModifiedf=u),(u=T.getResponseHeader("etag"))&&(k.etagf=u)),204===e||"HEAD"===v.type?l="nocontent":304===e?l="notmodified":(l=s.state,o=s.data,i=!(a=s.error))):(a=l,!e&&l||(l="error",e<0&&(e=0))),T.status=e,T.statusText=(t||l)+"",i?x.resolveWith(y,o,l,T):x.rejectWith(y,T,l,a),T.statusCode(w),w=void 0,g&&m.trigger(i?"ajaxSuccess":"ajaxError",T,v,i?o:a),b.fireWith(y,T,l),g&&(m.trigger("ajaxComplete",T,v),--k.active||k.event.trigger("ajaxStop")))}return T},getJSON:function(e,t,n){return k.get(e,t,n,"json")},getScript:function(e,t){return k.get(e,void 0,t,"script")}}),k.each("get","post",function(e,i){ki=function(e,t,n,r){return m(t)&&(r=r||n,n=t,t=void 0),k.ajax(k.extend({url:e,type:i,dataType:r,data:t,success:n},k.isPlainObject(e)&&e))}}),k._evalUrl=function(e,t){return k.ajax({url:e,type:"GET",dataType:"script",cache:!0,async:!1,global:!1,converters:{"text script":function(){}},dataFilter:function(e){k.globalEval(e,t)}})},k.fn.extend({wrapAll:function(e){var t;return this0&&(m(e)&&(e=e.call(this0)),t=k(e,this0.ownerDocument).eq(0).clone(!0),this0.parentNode&&t.insertBefore(this0),t.map(function(){var e=this;while(e.firstElementChild)e=e.firstElementChild;return e}).append(this)),this},wrapInner:function(n){return m(n)?this.each(function(e){k(this).wrapInner(n.call(this,e))}):this.each(function(){var e=k(this),t=e.contents();t.length?t.wrapAll(n):e.append(n)})},wrap:function(t){var n=m(t);return this.each(function(e){k(this).wrapAll(n?t.call(this,e):t)})},unwrap:function(e){return this.parent(e).not("body").each(function(){k(this).replaceWith(this.childNodes)}),this}}),k.expr.pseudos.hidden=function(e){return!k.expr.pseudos.visible(e)},k.expr.pseudos.visible=function(e){return!!(e.offsetWidth||e.offsetHeight||e.getClientRects().length)},k.ajaxSettings.xhr=function(){try{return new C.XMLHttpRequest}catch(e){}};var Ut={0:200,1223:204},Xt=k.ajaxSettings.xhr();y.cors=!!Xt&&"withCredentials"in Xt,y.ajax=Xt=!!Xt,k.ajaxTransport(function(i){var o,a;if(y.cors||Xt&&!i.crossDomain)return{send:function(e,t){var n,r=i.xhr();if(r.open(i.type,i.url,i.async,i.username,i.password),i.xhrFields)for(n in i.xhrFields)rn=i.xhrFieldsn;for(n in i.mimeType&&r.overrideMimeType&&r.overrideMimeType(i.mimeType),i.crossDomain||e"X-Requested-With"||(e"X-Requested-With"="XMLHttpRequest"),e)r.setRequestHeader(n,en);o=function(e){return function(){o&&(o=a=r.onload=r.onerror=r.onabort=r.ontimeout=r.onreadystatechange=null,"abort"===e?r.abort():"error"===e?"number"!=typeof r.status?t(0,"error"):t(r.status,r.statusText):t(Utr.status||r.status,r.statusText,"text"!==(r.responseType||"text")||"string"!=typeof r.responseText?{binary:r.response}:{text:r.responseText},r.getAllResponseHeaders()))}},r.onload=o(),a=r.onerror=r.ontimeout=o("error"),void 0!==r.onabort?r.onabort=a:r.onreadystatechange=function(){4===r.readyState&&C.setTimeout(function(){o&&a()})},o=o("abort");try{r.send(i.hasContent&&i.data||null)}catch(e){if(o)throw e}},abort:function(){o&&o()}}}),k.ajaxPrefilter(function(e){e.crossDomain&&(e.contents.script=!1)}),k.ajaxSetup({accepts:{script:"text/javascript, application/javascript, application/ecmascript, application/x-ecmascript"},contents:{script:/\b(?:java|ecma)script\b/},converters:{"text script":function(e){return k.globalEval(e),e}}}),k.ajaxPrefilter("script",function(e){void 0===e.cache&&(e.cache=!1),e.crossDomain&&(e.type="GET")}),k.ajaxTransport("script",function(n){var r,i;if(n.crossDomain||n.scriptAttrs)return{send:function(e,t){r=k("<script>").attr(n.scriptAttrs||{}).prop({charset:n.scriptCharset,src:n.url}).on("load error",i=function(e){r.remove(),i=null,e&&t("error"===e.type?404:200,e.type)}),E.head.appendChild(r0)},abort:function(){i&&i()}}});var Vt,Gt=,Yt=/(=)\?(?=&|$)|\?\?/;k.ajaxSetup({jsonp:"callback",jsonpCallback:function(){var e=Gt.pop()||k.expando+"_"+kt++;return thise=!0,e}}),k.ajaxPrefilter("json jsonp",function(e,t,n){var r,i,o,a=!1!==e.jsonp&&(Yt.test(e.url)?"url":"string"==typeof e.data&&0===(e.contentType||"").indexOf("application/x-www-form-urlencoded")&&Yt.test(e.data)&&"data");if(a||"jsonp"===e.dataTypes0)return r=e.jsonpCallback=m(e.jsonpCallback)?e.jsonpCallback():e.jsonpCallback,a?ea=ea.replace(Yt,"$1"+r):!1!==e.jsonp&&(e.url+=(St.test(e.url)?"&":"?")+e.jsonp+"="+r),e.converters"script json"=function(){return o||k.error(r+" was not called"),o0},e.dataTypes0="json",i=Cr,Cr=function(){o=arguments},n.always(function(){void 0===i?k(C).removeProp(r):Cr=i,er&&(e.jsonpCallback=t.jsonpCallback,Gt.push(r)),o&&m(i)&&i(o0),o=i=void 0}),"script"}),y.createHTMLDocument=((Vt=E.implementation.createHTMLDocument("").body).innerHTML="<form></form><form></form>",2===Vt.childNodes.length),k.parseHTML=function(e,t,n){return"string"!=typeof e?:("boolean"==typeof t&&(n=t,t=!1),t||(y.createHTMLDocument?((r=(t=E.implementation.createHTMLDocument("")).createElement("base")).href=E.location.href,t.head.appendChild(r)):t=E),o=!n&&,(i=D.exec(e))?t.createElement(i1):(i=we(e,t,o),o&&o.length&&k(o).remove(),k.merge(,i.childNodes)));var r,i,o},k.fn.load=function(e,t,n){var r,i,o,a=this,s=e.indexOf(" ");return-1<s&&(r=mt(e.slice(s)),e=e.slice(0,s)),m(t)?(n=t,t=void 0):t&&"object"==typeof t&&(i="POST"),0<a.length&&k.ajax({url:e,type:i||"GET",dataType:"html",data:t}).done(function(e){o=arguments,a.html(r?k("<div>").append(k.parseHTML(e)).find(r):e)}).always(n&&function(e,t){a.each(function(){n.apply(this,o||e.responseText,t,e)})}),this},k.each("ajaxStart","ajaxStop","ajaxComplete","ajaxError","ajaxSuccess","ajaxSend",function(e,t){k.fnt=function(e){return this.on(t,e)}}),k.expr.pseudos.animated=function(t){return k.grep(k.timers,function(e){return t===e.elem}).length},k.offset={setOffset:function(e,t,n){var r,i,o,a,s,u,l=k.css(e,"position"),c=k(e),f={};"static"===l&&(e.style.position="relative"),s=c.offset(),o=k.css(e,"top"),u=k.css(e,"left"),("absolute"===l||"fixed"===l)&&-1<(o+u).indexOf("auto")?(a=(r=c.position()).top,i=r.left):(a=parseFloat(o)||0,i=parseFloat(u)||0),m(t)&&(t=t.call(e,n,k.extend({},s))),null!=t.top&&(f.top=t.top-s.top+a),null!=t.left&&(f.left=t.left-s.left+i),"using"in t?t.using.call(e,f):c.css(f)}},k.fn.extend({offset:function(t){if(arguments.length)return void 0===t?this:this.each(function(e){k.offset.setOffset(this,t,e)});var e,n,r=this0;return r?r.getClientRects().length?(e=r.getBoundingClientRect(),n=r.ownerDocument.defaultView,{top:e.top+n.pageYOffset,left:e.left+n.pageXOffset}):{top:0,left:0}:void 0},position:function(){if(this0){var e,t,n,r=this0,i={top:0,left:0};if("fixed"===k.css(r,"position"))t=r.getBoundingClientRect();else{t=this.offset(),n=r.ownerDocument,e=r.offsetParent||n.documentElement;while(e&&(e===n.body||e===n.documentElement)&&"static"===k.css(e,"position"))e=e.parentNode;e&&e!==r&&1===e.nodeType&&((i=k(e).offset()).top+=k.css(e,"borderTopWidth",!0),i.left+=k.css(e,"borderLeftWidth",!0))}return{top:t.top-i.top-k.css(r,"marginTop",!0),left:t.left-i.left-k.css(r,"marginLeft",!0)}}},offsetParent:function(){return this.map(function(){var e=this.offsetParent;while(e&&"static"===k.css(e,"position"))e=e.offsetParent;return e||ie})}}),k.each({scrollLeft:"pageXOffset",scrollTop:"pageYOffset"},function(t,i){var o="pageYOffset"===i;k.fnt=function(e){return _(this,function(e,t,n){var r;if(x(e)?r=e:9===e.nodeType&&(r=e.defaultView),void 0===n)return r?ri:et;r?r.scrollTo(o?r.pageXOffset:n,o?n:r.pageYOffset):et=n},t,e,arguments.length)}}),k.each("top","left",function(e,n){k.cssHooksn=ze(y.pixelPosition,function(e,t){if(t)return t=_e(e,n),$e.test(t)?k(e).position()n+"px":t})}),k.each({Height:"height",Width:"width"},function(a,s){k.each({padding:"inner"+a,content:s,"":"outer"+a},function(r,o){k.fno=function(e,t){var n=arguments.length&&(r||"boolean"!=typeof e),i=r||(!0===e||!0===t?"margin":"border");return _(this,function(e,t,n){var r;return x(e)?0===o.indexOf("outer")?e"inner"+a:e.document.documentElement"client"+a:9===e.nodeType?(r=e.documentElement,Math.max(e.body"scroll"+a,r"scroll"+a,e.body"offset"+a,r"offset"+a,r"client"+a)):void 0===n?k.css(e,t,i):k.style(e,t,n,i)},s,n?e:void 0,n)}})}),k.each("blur focus focusin focusout resize scroll click dblclick mousedown mouseup mousemove mouseover mouseout mouseenter mouseleave change select submit keydown keypress keyup contextmenu".split(" "),function(e,n){k.fnn=function(e,t){return 0<arguments.length?this.on(n,null,e,t):this.trigger(n)}}),k.fn.extend({hover:function(e,t){return this.mouseenter(e).mouseleave(t||e)}}),k.fn.extend({bind:function(e,t,n){return this.on(e,null,t,n)},unbind:function(e,t){return this.off(e,null,t)},delegate:function(e,t,n,r){return this.on(t,e,n,r)},undelegate:function(e,t,n){return 1===arguments.length?this.off(e,"**"):this.off(t,e||"**",n)}}),k.proxy=function(e,t){var n,r,i;if("string"==typeof t&&(n=et,t=e,e=n),m(e))return r=s.call(arguments,2),(i=function(){return e.apply(t||this,r.concat(s.call(arguments)))}).guid=e.guid=e.guid||k.guid++,i},k.holdReady=function(e){e?k.readyWait++:k.ready(!0)},k.isArray=Array.isArray,k.parseJSON=JSON.parse,k.nodeName=A,k.isFunction=m,k.isWindow=x,k.camelCase=V,k.type=w,k.now=Date.now,k.isNumeric=function(e){var t=k.type(e);return("number"===t||"string"===t)&&!isNaN(e-parseFloat(e))},"function"==typeof define&&define.amd&&define("jquery",,function(){return k});var Qt=C.jQuery,Jt=C.$;return k.noConflict=function(e){return C.$===k&&(C.$=Jt),e&&C.jQuery===k&&(C.jQuery=Qt),k},e||(C.jQuery=C.$=k),k});
+/*! jQuery UI - v1.12.1 - 2019-01-27
+* http://jqueryui.com
+* Includes: widget.js, position.js, data.js, disable-selection.js, focusable.js, form-reset-mixin.js, jquery-1-7.js, keycode.js, labels.js, scroll-parent.js, tabbable.js, unique-id.js, widgets/resizable.js, widgets/mouse.js
+* Copyright jQuery Foundation and other contributors; Licensed MIT */
+
+(function(t){"function"==typeof define&&define.amd?define("jquery",t):t(jQuery)})(function(t){function e(t){for(var e=t.css("visibility");"inherit"===e;)t=t.parent(),e=t.css("visibility");return"hidden"!==e}t.ui=t.ui||{},t.ui.version="1.12.1";var i=0,s=Array.prototype.slice;t.cleanData=function(e){return function(i){var s,n,o;for(o=0;null!=(n=io);o++)try{s=t._data(n,"events"),s&&s.remove&&t(n).triggerHandler("remove")}catch(a){}e(i)}}(t.cleanData),t.widget=function(e,i,s){var n,o,a,r={},h=e.split(".")0;e=e.split(".")1;var l=h+"-"+e;return s||(s=i,i=t.Widget),t.isArray(s)&&(s=t.extend.apply(null,{}.concat(s))),t.expr":"l.toLowerCase()=function(e){return!!t.data(e,l)},th=th||{},n=the,o=the=function(t,e){return this._createWidget?(arguments.length&&this._createWidget(t,e),void 0):new o(t,e)},t.extend(o,n,{version:s.version,_proto:t.extend({},s),_childConstructors:}),a=new i,a.options=t.widget.extend({},a.options),t.each(s,function(e,s){return t.isFunction(s)?(re=function(){function t(){return i.prototypee.apply(this,arguments)}function n(t){return i.prototypee.apply(this,t)}return function(){var e,i=this._super,o=this._superApply;return this._super=t,this._superApply=n,e=s.apply(this,arguments),this._super=i,this._superApply=o,e}}(),void 0):(re=s,void 0)}),o.prototype=t.widget.extend(a,{widgetEventPrefix:n?a.widgetEventPrefix||e:e},r,{constructor:o,namespace:h,widgetName:e,widgetFullName:l}),n?(t.each(n._childConstructors,function(e,i){var s=i.prototype;t.widget(s.namespace+"."+s.widgetName,o,i._proto)}),delete n._childConstructors):i._childConstructors.push(o),t.widget.bridge(e,o),o},t.widget.extend=function(e){for(var i,n,o=s.call(arguments,1),a=0,r=o.length;r>a;a++)for(i in oa)n=oai,oa.hasOwnProperty(i)&&void 0!==n&&(ei=t.isPlainObject(n)?t.isPlainObject(ei)?t.widget.extend({},ei,n):t.widget.extend({},n):n);return e},t.widget.bridge=function(e,i){var n=i.prototype.widgetFullName||e;t.fne=function(o){var a="string"==typeof o,r=s.call(arguments,1),h=this;return a?this.length||"instance"!==o?this.each(function(){var i,s=t.data(this,n);return"instance"===o?(h=s,!1):s?t.isFunction(so)&&"_"!==o.charAt(0)?(i=so.apply(s,r),i!==s&&void 0!==i?(h=i&&i.jquery?h.pushStack(i.get()):i,!1):void 0):t.error("no such method '"+o+"' for "+e+" widget instance"):t.error("cannot call methods on "+e+" prior to initialization; "+"attempted to call method '"+o+"'")}):h=void 0:(r.length&&(o=t.widget.extend.apply(null,o.concat(r))),this.each(function(){var e=t.data(this,n);e?(e.option(o||{}),e._init&&e._init()):t.data(this,n,new i(o,this))})),h}},t.Widget=function(){},t.Widget._childConstructors=,t.Widget.prototype={widgetName:"widget",widgetEventPrefix:"",defaultElement:"<div>",options:{classes:{},disabled:!1,create:null},_createWidget:function(e,s){s=t(s||this.defaultElement||this)0,this.element=t(s),this.uuid=i++,this.eventNamespace="."+this.widgetName+this.uuid,this.bindings=t(),this.hoverable=t(),this.focusable=t(),this.classesElementLookup={},s!==this&&(t.data(s,this.widgetFullName,this),this._on(!0,this.element,{remove:function(t){t.target===s&&this.destroy()}}),this.document=t(s.style?s.ownerDocument:s.document||s),this.window=t(this.document0.defaultView||this.document0.parentWindow)),this.options=t.widget.extend({},this.options,this._getCreateOptions(),e),this._create(),this.options.disabled&&this._setOptionDisabled(this.options.disabled),this._trigger("create",null,this._getCreateEventData()),this._init()},_getCreateOptions:function(){return{}},_getCreateEventData:t.noop,_create:t.noop,_init:t.noop,destroy:function(){var e=this;this._destroy(),t.each(this.classesElementLookup,function(t,i){e._removeClass(i,t)}),this.element.off(this.eventNamespace).removeData(this.widgetFullName),this.widget().off(this.eventNamespace).removeAttr("aria-disabled"),this.bindings.off(this.eventNamespace)},_destroy:t.noop,widget:function(){return this.element},option:function(e,i){var s,n,o,a=e;if(0===arguments.length)return t.widget.extend({},this.options);if("string"==typeof e)if(a={},s=e.split("."),e=s.shift(),s.length){for(n=ae=t.widget.extend({},this.optionse),o=0;s.length-1>o;o++)nso=nso||{},n=nso;if(e=s.pop(),1===arguments.length)return void 0===ne?null:ne;ne=i}else{if(1===arguments.length)return void 0===this.optionse?null:this.optionse;ae=i}return this._setOptions(a),this},_setOptions:function(t){var e;for(e in t)this._setOption(e,te);return this},_setOption:function(t,e){return"classes"===t&&this._setOptionClasses(e),this.optionst=e,"disabled"===t&&this._setOptionDisabled(e),this},_setOptionClasses:function(e){var i,s,n;for(i in e)n=this.classesElementLookupi,ei!==this.options.classesi&&n&&n.length&&(s=t(n.get()),this._removeClass(n,i),s.addClass(this._classes({element:s,keys:i,classes:e,add:!0})))},_setOptionDisabled:function(t){this._toggleClass(this.widget(),this.widgetFullName+"-disabled",null,!!t),t&&(this._removeClass(this.hoverable,null,"ui-state-hover"),this._removeClass(this.focusable,null,"ui-state-focus"))},enable:function(){return this._setOptions({disabled:!1})},disable:function(){return this._setOptions({disabled:!0})},_classes:function(e){function i(i,o){var a,r;for(r=0;i.length>r;r++)a=n.classesElementLookupir||t(),a=e.add?t(t.unique(a.get().concat(e.element.get()))):t(a.not(e.element).get()),n.classesElementLookupir=a,s.push(ir),o&&e.classesir&&s.push(e.classesir)}var s=,n=this;return e=t.extend({element:this.element,classes:this.options.classes||{}},e),this._on(e.element,{remove:"_untrackClassesElement"}),e.keys&&i(e.keys.match(/\S+/g)||,!0),e.extra&&i(e.extra.match(/\S+/g)||),s.join(" ")},_untrackClassesElement:function(e){var i=this;t.each(i.classesElementLookup,function(s,n){-1!==t.inArray(e.target,n)&&(i.classesElementLookups=t(n.not(e.target).get()))})},_removeClass:function(t,e,i){return this._toggleClass(t,e,i,!1)},_addClass:function(t,e,i){return this._toggleClass(t,e,i,!0)},_toggleClass:function(t,e,i,s){s="boolean"==typeof s?s:i;var n="string"==typeof t||null===t,o={extra:n?e:i,keys:n?t:e,element:n?this.element:t,add:s};return o.element.toggleClass(this._classes(o),s),this},_on:function(e,i,s){var n,o=this;"boolean"!=typeof e&&(s=i,i=e,e=!1),s?(i=n=t(i),this.bindings=this.bindings.add(i)):(s=i,i=this.element,n=this.widget()),t.each(s,function(s,a){function r(){return e||o.options.disabled!==!0&&!t(this).hasClass("ui-state-disabled")?("string"==typeof a?oa:a).apply(o,arguments):void 0}"string"!=typeof a&&(r.guid=a.guid=a.guid||r.guid||t.guid++);var h=s.match(/^(\w:-*)\s*(.*)$/),l=h1+o.eventNamespace,c=h2;c?n.on(l,c,r):i.on(l,r)})},_off:function(e,i){i=(i||"").split(" ").join(this.eventNamespace+" ")+this.eventNamespace,e.off(i).off(i),this.bindings=t(this.bindings.not(e).get()),this.focusable=t(this.focusable.not(e).get()),this.hoverable=t(this.hoverable.not(e).get())},_delay:function(t,e){function i(){return("string"==typeof t?st:t).apply(s,arguments)}var s=this;return setTimeout(i,e||0)},_hoverable:function(e){this.hoverable=this.hoverable.add(e),this._on(e,{mouseenter:function(e){this._addClass(t(e.currentTarget),null,"ui-state-hover")},mouseleave:function(e){this._removeClass(t(e.currentTarget),null,"ui-state-hover")}})},_focusable:function(e){this.focusable=this.focusable.add(e),this._on(e,{focusin:function(e){this._addClass(t(e.currentTarget),null,"ui-state-focus")},focusout:function(e){this._removeClass(t(e.currentTarget),null,"ui-state-focus")}})},_trigger:function(e,i,s){var n,o,a=this.optionse;if(s=s||{},i=t.Event(i),i.type=(e===this.widgetEventPrefix?e:this.widgetEventPrefix+e).toLowerCase(),i.target=this.element0,o=i.originalEvent)for(n in o)n in i||(in=on);return this.element.trigger(i,s),!(t.isFunction(a)&&a.apply(this.element0,i.concat(s))===!1||i.isDefaultPrevented())}},t.each({show:"fadeIn",hide:"fadeOut"},function(e,i){t.Widget.prototype"_"+e=function(s,n,o){"string"==typeof n&&(n={effect:n});var a,r=n?n===!0||"number"==typeof n?i:n.effect||i:e;n=n||{},"number"==typeof n&&(n={duration:n}),a=!t.isEmptyObject(n),n.complete=o,n.delay&&s.delay(n.delay),a&&t.effects&&t.effects.effectr?se(n):r!==e&&sr?sr(n.duration,n.easing,o):s.queue(function(i){t(this)e(),o&&o.call(s0),i()})}}),t.widget,function(){function e(t,e,i){returnparseFloat(t0)*(u.test(t0)?e/100:1),parseFloat(t1)*(u.test(t1)?i/100:1)}function i(e,i){return parseInt(t.css(e,i),10)||0}function s(e){var i=e0;return 9===i.nodeType?{width:e.width(),height:e.height(),offset:{top:0,left:0}}:t.isWindow(i)?{width:e.width(),height:e.height(),offset:{top:e.scrollTop(),left:e.scrollLeft()}}:i.preventDefault?{width:0,height:0,offset:{top:i.pageY,left:i.pageX}}:{width:e.outerWidth(),height:e.outerHeight(),offset:e.offset()}}var n,o=Math.max,a=Math.abs,r=/left|center|right/,h=/top|center|bottom/,l=/\+\-\d+(\.\d+)?%?/,c=/^\w+/,u=/%$/,d=t.fn.position;t.position={scrollbarWidth:function(){if(void 0!==n)return n;var e,i,s=t("<div style='display:block;position:absolute;width:50px;height:50px;overflow:hidden;'><div style='height:100px;width:auto;'></div></div>"),o=s.children()0;return t("body").append(s),e=o.offsetWidth,s.css("overflow","scroll"),i=o.offsetWidth,e===i&&(i=s0.clientWidth),s.remove(),n=e-i},getScrollInfo:function(e){var i=e.isWindow||e.isDocument?"":e.element.css("overflow-x"),s=e.isWindow||e.isDocument?"":e.element.css("overflow-y"),n="scroll"===i||"auto"===i&&e.width<e.element0.scrollWidth,o="scroll"===s||"auto"===s&&e.height<e.element0.scrollHeight;return{width:o?t.position.scrollbarWidth():0,height:n?t.position.scrollbarWidth():0}},getWithinInfo:function(e){var i=t(e||window),s=t.isWindow(i0),n=!!i0&&9===i0.nodeType,o=!s&&!n;return{element:i,isWindow:s,isDocument:n,offset:o?t(e).offset():{left:0,top:0},scrollLeft:i.scrollLeft(),scrollTop:i.scrollTop(),width:i.outerWidth(),height:i.outerHeight()}}},t.fn.position=function(n){if(!n||!n.of)return d.apply(this,arguments);n=t.extend({},n);var u,p,f,m,g,_,v=t(n.of),b=t.position.getWithinInfo(n.within),y=t.position.getScrollInfo(b),w=(n.collision||"flip").split(" "),x={};return _=s(v),v0.preventDefault&&(n.at="left top"),p=_.width,f=_.height,m=_.offset,g=t.extend({},m),t.each("my","at",function(){var t,e,i=(nthis||"").split(" ");1===i.length&&(i=r.test(i0)?i.concat("center"):h.test(i0)?"center".concat(i):"center","center"),i0=r.test(i0)?i0:"center",i1=h.test(i1)?i1:"center",t=l.exec(i0),e=l.exec(i1),xthis=t?t0:0,e?e0:0,nthis=c.exec(i0)0,c.exec(i1)0}),1===w.length&&(w1=w0),"right"===n.at0?g.left+=p:"center"===n.at0&&(g.left+=p/2),"bottom"===n.at1?g.top+=f:"center"===n.at1&&(g.top+=f/2),u=e(x.at,p,f),g.left+=u0,g.top+=u1,this.each(function(){var s,r,h=t(this),l=h.outerWidth(),c=h.outerHeight(),d=i(this,"marginLeft"),_=i(this,"marginTop"),k=l+d+i(this,"marginRight")+y.width,C=c+_+i(this,"marginBottom")+y.height,D=t.extend({},g),T=e(x.my,h.outerWidth(),h.outerHeight());"right"===n.my0?D.left-=l:"center"===n.my0&&(D.left-=l/2),"bottom"===n.my1?D.top-=c:"center"===n.my1&&(D.top-=c/2),D.left+=T0,D.top+=T1,s={marginLeft:d,marginTop:_},t.each("left","top",function(e,i){t.ui.positionwe&&t.ui.positionwei(D,{targetWidth:p,targetHeight:f,elemWidth:l,elemHeight:c,collisionPosition:s,collisionWidth:k,collisionHeight:C,offset:u0+T0,u1+T1,my:n.my,at:n.at,within:b,elem:h})}),n.using&&(r=function(t){var e=m.left-D.left,i=e+p-l,s=m.top-D.top,r=s+f-c,u={target:{element:v,left:m.left,top:m.top,width:p,height:f},element:{element:h,left:D.left,top:D.top,width:l,height:c},horizontal:0>i?"left":e>0?"right":"center",vertical:0>r?"top":s>0?"bottom":"middle"};l>p&&p>a(e+i)&&(u.horizontal="center"),c>f&&f>a(s+r)&&(u.vertical="middle"),u.important=o(a(e),a(i))>o(a(s),a(r))?"horizontal":"vertical",n.using.call(this,t,u)}),h.offset(t.extend(D,{using:r}))})},t.ui.position={fit:{left:function(t,e){var i,s=e.within,n=s.isWindow?s.scrollLeft:s.offset.left,a=s.width,r=t.left-e.collisionPosition.marginLeft,h=n-r,l=r+e.collisionWidth-a-n;e.collisionWidth>a?h>0&&0>=l?(i=t.left+h+e.collisionWidth-a-n,t.left+=h-i):t.left=l>0&&0>=h?n:h>l?n+a-e.collisionWidth:n:h>0?t.left+=h:l>0?t.left-=l:t.left=o(t.left-r,t.left)},top:function(t,e){var i,s=e.within,n=s.isWindow?s.scrollTop:s.offset.top,a=e.within.height,r=t.top-e.collisionPosition.marginTop,h=n-r,l=r+e.collisionHeight-a-n;e.collisionHeight>a?h>0&&0>=l?(i=t.top+h+e.collisionHeight-a-n,t.top+=h-i):t.top=l>0&&0>=h?n:h>l?n+a-e.collisionHeight:n:h>0?t.top+=h:l>0?t.top-=l:t.top=o(t.top-r,t.top)}},flip:{left:function(t,e){var i,s,n=e.within,o=n.offset.left+n.scrollLeft,r=n.width,h=n.isWindow?n.scrollLeft:n.offset.left,l=t.left-e.collisionPosition.marginLeft,c=l-h,u=l+e.collisionWidth-r-h,d="left"===e.my0?-e.elemWidth:"right"===e.my0?e.elemWidth:0,p="left"===e.at0?e.targetWidth:"right"===e.at0?-e.targetWidth:0,f=-2*e.offset0;0>c?(i=t.left+d+p+f+e.collisionWidth-r-o,(0>i||a(c)>i)&&(t.left+=d+p+f)):u>0&&(s=t.left-e.collisionPosition.marginLeft+d+p+f-h,(s>0||u>a(s))&&(t.left+=d+p+f))},top:function(t,e){var i,s,n=e.within,o=n.offset.top+n.scrollTop,r=n.height,h=n.isWindow?n.scrollTop:n.offset.top,l=t.top-e.collisionPosition.marginTop,c=l-h,u=l+e.collisionHeight-r-h,d="top"===e.my1,p=d?-e.elemHeight:"bottom"===e.my1?e.elemHeight:0,f="top"===e.at1?e.targetHeight:"bottom"===e.at1?-e.targetHeight:0,m=-2*e.offset1;0>c?(s=t.top+p+f+m+e.collisionHeight-r-o,(0>s||a(c)>s)&&(t.top+=p+f+m)):u>0&&(i=t.top-e.collisionPosition.marginTop+p+f+m-h,(i>0||u>a(i))&&(t.top+=p+f+m))}},flipfit:{left:function(){t.ui.position.flip.left.apply(this,arguments),t.ui.position.fit.left.apply(this,arguments)},top:function(){t.ui.position.flip.top.apply(this,arguments),t.ui.position.fit.top.apply(this,arguments)}}}}(),t.ui.position,t.extend(t.expr":",{data:t.expr.createPseudo?t.expr.createPseudo(function(e){return function(i){return!!t.data(i,e)}}):function(e,i,s){return!!t.data(e,s3)}}),t.fn.extend({disableSelection:function(){var t="onselectstart"in document.createElement("div")?"selectstart":"mousedown";return function(){return this.on(t+".ui-disableSelection",function(t){t.preventDefault()})}}(),enableSelection:function(){return this.off(".ui-disableSelection")}}),t.ui.focusable=function(i,s){var n,o,a,r,h,l=i.nodeName.toLowerCase();return"area"===l?(n=i.parentNode,o=n.name,i.href&&o&&"map"===n.nodeName.toLowerCase()?(a=t("imgusemap='#"+o+"'"),a.length>0&&a.is(":visible")):!1):(/^(input|select|textarea|button|object)$/.test(l)?(r=!i.disabled,r&&(h=t(i).closest("fieldset")0,h&&(r=!h.disabled))):r="a"===l?i.href||s:s,r&&t(i).is(":visible")&&e(t(i)))},t.extend(t.expr":",{focusable:function(e){return t.ui.focusable(e,null!=t.attr(e,"tabindex"))}}),t.ui.focusable,t.fn.form=function(){return"string"==typeof this0.form?this.closest("form"):t(this0.form)},t.ui.formResetMixin={_formResetHandler:function(){var e=t(this);setTimeout(function(){var i=e.data("ui-form-reset-instances");t.each(i,function(){this.refresh()})})},_bindFormResetHandler:function(){if(this.form=this.element.form(),this.form.length){var t=this.form.data("ui-form-reset-instances")||;t.length||this.form.on("reset.ui-form-reset",this._formResetHandler),t.push(this),this.form.data("ui-form-reset-instances",t)}},_unbindFormResetHandler:function(){if(this.form.length){var e=this.form.data("ui-form-reset-instances");e.splice(t.inArray(this,e),1),e.length?this.form.data("ui-form-reset-instances",e):this.form.removeData("ui-form-reset-instances").off("reset.ui-form-reset")}}},"1.7"===t.fn.jquery.substring(0,3)&&(t.each("Width","Height",function(e,i){function s(e,i,s,o){return t.each(n,function(){i-=parseFloat(t.css(e,"padding"+this))||0,s&&(i-=parseFloat(t.css(e,"border"+this+"Width"))||0),o&&(i-=parseFloat(t.css(e,"margin"+this))||0)}),i}var n="Width"===i?"Left","Right":"Top","Bottom",o=i.toLowerCase(),a={innerWidth:t.fn.innerWidth,innerHeight:t.fn.innerHeight,outerWidth:t.fn.outerWidth,outerHeight:t.fn.outerHeight};t.fn"inner"+i=function(e){return void 0===e?a"inner"+i.call(this):this.each(function(){t(this).css(o,s(this,e)+"px")})},t.fn"outer"+i=function(e,n){return"number"!=typeof e?a"outer"+i.call(this,e):this.each(function(){t(this).css(o,s(this,e,!0,n)+"px")})}}),t.fn.addBack=function(t){return this.add(null==t?this.prevObject:this.prevObject.filter(t))}),t.ui.keyCode={BACKSPACE:8,COMMA:188,DELETE:46,DOWN:40,END:35,ENTER:13,ESCAPE:27,HOME:36,LEFT:37,PAGE_DOWN:34,PAGE_UP:33,PERIOD:190,RIGHT:39,SPACE:32,TAB:9,UP:38},t.ui.escapeSelector=function(){var t=/(!"#$%&'()*+,./:;<=>?@\^`{|}~)/g;return function(e){return e.replace(t,"\\$1")}}(),t.fn.labels=function(){var e,i,s,n,o;return this0.labels&&this0.labels.length?this.pushStack(this0.labels):(n=this.eq(0).parents("label"),s=this.attr("id"),s&&(e=this.eq(0).parents().last(),o=e.add(e.length?e.siblings():this.siblings()),i="labelfor='"+t.ui.escapeSelector(s)+"'",n=n.add(o.find(i).addBack(i))),this.pushStack(n))},t.fn.scrollParent=function(e){var i=this.css("position"),s="absolute"===i,n=e?/(auto|scroll|hidden)/:/(auto|scroll)/,o=this.parents().filter(function(){var e=t(this);return s&&"static"===e.css("position")?!1:n.test(e.css("overflow")+e.css("overflow-y")+e.css("overflow-x"))}).eq(0);return"fixed"!==i&&o.length?o:t(this0.ownerDocument||document)},t.extend(t.expr":",{tabbable:function(e){var i=t.attr(e,"tabindex"),s=null!=i;return(!s||i>=0)&&t.ui.focusable(e,s)}}),t.fn.extend({uniqueId:function(){var t=0;return function(){return this.each(function(){this.id||(this.id="ui-id-"+ ++t)})}}(),removeUniqueId:function(){return this.each(function(){/^ui-id-\d+$/.test(this.id)&&t(this).removeAttr("id")})}}),t.ui.ie=!!/msie \w.+/.exec(navigator.userAgent.toLowerCase());var n=!1;t(document).on("mouseup",function(){n=!1}),t.widget("ui.mouse",{version:"1.12.1",options:{cancel:"input, textarea, button, select, option",distance:1,delay:0},_mouseInit:function(){var e=this;this.element.on("mousedown."+this.widgetName,function(t){return e._mouseDown(t)}).on("click."+this.widgetName,function(i){return!0===t.data(i.target,e.widgetName+".preventClickEvent")?(t.removeData(i.target,e.widgetName+".preventClickEvent"),i.stopImmediatePropagation(),!1):void 0}),this.started=!1},_mouseDestroy:function(){this.element.off("."+this.widgetName),this._mouseMoveDelegate&&this.document.off("mousemove."+this.widgetName,this._mouseMoveDelegate).off("mouseup."+this.widgetName,this._mouseUpDelegate)},_mouseDown:function(e){if(!n){this._mouseMoved=!1,this._mouseStarted&&this._mouseUp(e),this._mouseDownEvent=e;var i=this,s=1===e.which,o="string"==typeof this.options.cancel&&e.target.nodeName?t(e.target).closest(this.options.cancel).length:!1;return s&&!o&&this._mouseCapture(e)?(this.mouseDelayMet=!this.options.delay,this.mouseDelayMet||(this._mouseDelayTimer=setTimeout(function(){i.mouseDelayMet=!0},this.options.delay)),this._mouseDistanceMet(e)&&this._mouseDelayMet(e)&&(this._mouseStarted=this._mouseStart(e)!==!1,!this._mouseStarted)?(e.preventDefault(),!0):(!0===t.data(e.target,this.widgetName+".preventClickEvent")&&t.removeData(e.target,this.widgetName+".preventClickEvent"),this._mouseMoveDelegate=function(t){return i._mouseMove(t)},this._mouseUpDelegate=function(t){return i._mouseUp(t)},this.document.on("mousemove."+this.widgetName,this._mouseMoveDelegate).on("mouseup."+this.widgetName,this._mouseUpDelegate),e.preventDefault(),n=!0,!0)):!0}},_mouseMove:function(e){if(this._mouseMoved){if(t.ui.ie&&(!document.documentMode||9>document.documentMode)&&!e.button)return this._mouseUp(e);if(!e.which)if(e.originalEvent.altKey||e.originalEvent.ctrlKey||e.originalEvent.metaKey||e.originalEvent.shiftKey)this.ignoreMissingWhich=!0;else if(!this.ignoreMissingWhich)return this._mouseUp(e)}return(e.which||e.button)&&(this._mouseMoved=!0),this._mouseStarted?(this._mouseDrag(e),e.preventDefault()):(this._mouseDistanceMet(e)&&this._mouseDelayMet(e)&&(this._mouseStarted=this._mouseStart(this._mouseDownEvent,e)!==!1,this._mouseStarted?this._mouseDrag(e):this._mouseUp(e)),!this._mouseStarted)},_mouseUp:function(e){this.document.off("mousemove."+this.widgetName,this._mouseMoveDelegate).off("mouseup."+this.widgetName,this._mouseUpDelegate),this._mouseStarted&&(this._mouseStarted=!1,e.target===this._mouseDownEvent.target&&t.data(e.target,this.widgetName+".preventClickEvent",!0),this._mouseStop(e)),this._mouseDelayTimer&&(clearTimeout(this._mouseDelayTimer),delete this._mouseDelayTimer),this.ignoreMissingWhich=!1,n=!1,e.preventDefault()},_mouseDistanceMet:function(t){return Math.max(Math.abs(this._mouseDownEvent.pageX-t.pageX),Math.abs(this._mouseDownEvent.pageY-t.pageY))>=this.options.distance},_mouseDelayMet:function(){return this.mouseDelayMet},_mouseStart:function(){},_mouseDrag:function(){},_mouseStop:function(){},_mouseCapture:function(){return!0}}),t.ui.plugin={add:function(e,i,s){var n,o=t.uie.prototype;for(n in s)o.pluginsn=o.pluginsn||,o.pluginsn.push(i,sn)},call:function(t,e,i,s){var n,o=t.pluginse;if(o&&(s||t.element0.parentNode&&11!==t.element0.parentNode.nodeType))for(n=0;o.length>n;n++)t.optionson0&&on1.apply(t.element,i)}},t.widget("ui.resizable",t.ui.mouse,{version:"1.12.1",widgetEventPrefix:"resize",options:{alsoResize:!1,animate:!1,animateDuration:"slow",animateEasing:"swing",aspectRatio:!1,autoHide:!1,classes:{"ui-resizable-se":"ui-icon ui-icon-gripsmall-diagonal-se"},containment:!1,ghost:!1,grid:!1,handles:"e,s,se",helper:!1,maxHeight:null,maxWidth:null,minHeight:10,minWidth:10,zIndex:90,resize:null,start:null,stop:null},_num:function(t){return parseFloat(t)||0},_isNumber:function(t){return!isNaN(parseFloat(t))},_hasScroll:function(e,i){if("hidden"===t(e).css("overflow"))return!1;var s=i&&"left"===i?"scrollLeft":"scrollTop",n=!1;return es>0?!0:(es=1,n=es>0,es=0,n)},_create:function(){var e,i=this.options,s=this;this._addClass("ui-resizable"),t.extend(this,{_aspectRatio:!!i.aspectRatio,aspectRatio:i.aspectRatio,originalElement:this.element,_proportionallyResizeElements:,_helper:i.helper||i.ghost||i.animate?i.helper||"ui-resizable-helper":null}),this.element0.nodeName.match(/^(canvas|textarea|input|select|button|img)$/i)&&(this.element.wrap(t("<div class='ui-wrapper' style='overflow: hidden;'></div>").css({position:this.element.css("position"),width:this.element.outerWidth(),height:this.element.outerHeight(),top:this.element.css("top"),left:this.element.css("left")})),this.element=this.element.parent().data("ui-resizable",this.element.resizable("instance")),this.elementIsWrapper=!0,e={marginTop:this.originalElement.css("marginTop"),marginRight:this.originalElement.css("marginRight"),marginBottom:this.originalElement.css("marginBottom"),marginLeft:this.originalElement.css("marginLeft")},this.element.css(e),this.originalElement.css("margin",0),this.originalResizeStyle=this.originalElement.css("resize"),this.originalElement.css("resize","none"),this._proportionallyResizeElements.push(this.originalElement.css({position:"static",zoom:1,display:"block"})),this.originalElement.css(e),this._proportionallyResize()),this._setupHandles(),i.autoHide&&t(this.element).on("mouseenter",function(){i.disabled||(s._removeClass("ui-resizable-autohide"),s._handles.show())}).on("mouseleave",function(){i.disabled||s.resizing||(s._addClass("ui-resizable-autohide"),s._handles.hide())}),this._mouseInit()},_destroy:function(){this._mouseDestroy();var e,i=function(e){t(e).removeData("resizable").removeData("ui-resizable").off(".resizable").find(".ui-resizable-handle").remove()};return this.elementIsWrapper&&(i(this.element),e=this.element,this.originalElement.css({position:e.css("position"),width:e.outerWidth(),height:e.outerHeight(),top:e.css("top"),left:e.css("left")}).insertAfter(e),e.remove()),this.originalElement.css("resize",this.originalResizeStyle),i(this.originalElement),this},_setOption:function(t,e){switch(this._super(t,e),t){case"handles":this._removeHandles(),this._setupHandles();break;default:}},_setupHandles:function(){var e,i,s,n,o,a=this.options,r=this;if(this.handles=a.handles||(t(".ui-resizable-handle",this.element).length?{n:".ui-resizable-n",e:".ui-resizable-e",s:".ui-resizable-s",w:".ui-resizable-w",se:".ui-resizable-se",sw:".ui-resizable-sw",ne:".ui-resizable-ne",nw:".ui-resizable-nw"}:"e,s,se"),this._handles=t(),this.handles.constructor===String)for("all"===this.handles&&(this.handles="n,e,s,w,se,sw,ne,nw"),s=this.handles.split(","),this.handles={},i=0;s.length>i;i++)e=t.trim(si),n="ui-resizable-"+e,o=t("<div>"),this._addClass(o,"ui-resizable-handle "+n),o.css({zIndex:a.zIndex}),this.handlese=".ui-resizable-"+e,this.element.append(o);this._renderAxis=function(e){var i,s,n,o;e=e||this.element;for(i in this.handles)this.handlesi.constructor===String?this.handlesi=this.element.children(this.handlesi).first().show():(this.handlesi.jquery||this.handlesi.nodeType)&&(this.handlesi=t(this.handlesi),this._on(this.handlesi,{mousedown:r._mouseDown})),this.elementIsWrapper&&this.originalElement0.nodeName.match(/^(textarea|input|select|button)$/i)&&(s=t(this.handlesi,this.element),o=/sw|ne|nw|se|n|s/.test(i)?s.outerHeight():s.outerWidth(),n="padding",/ne|nw|n/.test(i)?"Top":/se|sw|s/.test(i)?"Bottom":/^e$/.test(i)?"Right":"Left".join(""),e.css(n,o),this._proportionallyResize()),this._handles=this._handles.add(this.handlesi)},this._renderAxis(this.element),this._handles=this._handles.add(this.element.find(".ui-resizable-handle")),this._handles.disableSelection(),this._handles.on("mouseover",function(){r.resizing||(this.className&&(o=this.className.match(/ui-resizable-(se|sw|ne|nw|n|e|s|w)/i)),r.axis=o&&o1?o1:"se")}),a.autoHide&&(this._handles.hide(),this._addClass("ui-resizable-autohide"))},_removeHandles:function(){this._handles.remove()},_mouseCapture:function(e){var i,s,n=!1;for(i in this.handles)s=t(this.handlesi)0,(s===e.target||t.contains(s,e.target))&&(n=!0);return!this.options.disabled&&n},_mouseStart:function(e){var i,s,n,o=this.options,a=this.element;return this.resizing=!0,this._renderProxy(),i=this._num(this.helper.css("left")),s=this._num(this.helper.css("top")),o.containment&&(i+=t(o.containment).scrollLeft()||0,s+=t(o.containment).scrollTop()||0),this.offset=this.helper.offset(),this.position={left:i,top:s},this.size=this._helper?{width:this.helper.width(),height:this.helper.height()}:{width:a.width(),height:a.height()},this.originalSize=this._helper?{width:a.outerWidth(),height:a.outerHeight()}:{width:a.width(),height:a.height()},this.sizeDiff={width:a.outerWidth()-a.width(),height:a.outerHeight()-a.height()},this.originalPosition={left:i,top:s},this.originalMousePosition={left:e.pageX,top:e.pageY},this.aspectRatio="number"==typeof o.aspectRatio?o.aspectRatio:this.originalSize.width/this.originalSize.height||1,n=t(".ui-resizable-"+this.axis).css("cursor"),t("body").css("cursor","auto"===n?this.axis+"-resize":n),this._addClass("ui-resizable-resizing"),this._propagate("start",e),!0},_mouseDrag:function(e){var i,s,n=this.originalMousePosition,o=this.axis,a=e.pageX-n.left||0,r=e.pageY-n.top||0,h=this._changeo;return this._updatePrevProperties(),h?(i=h.apply(this,e,a,r),this._updateVirtualBoundaries(e.shiftKey),(this._aspectRatio||e.shiftKey)&&(i=this._updateRatio(i,e)),i=this._respectSize(i,e),this._updateCache(i),this._propagate("resize",e),s=this._applyChanges(),!this._helper&&this._proportionallyResizeElements.length&&this._proportionallyResize(),t.isEmptyObject(s)||(this._updatePrevProperties(),this._trigger("resize",e,this.ui()),this._applyChanges()),!1):!1},_mouseStop:function(e){this.resizing=!1;var i,s,n,o,a,r,h,l=this.options,c=this;return this._helper&&(i=this._proportionallyResizeElements,s=i.length&&/textarea/i.test(i0.nodeName),n=s&&this._hasScroll(i0,"left")?0:c.sizeDiff.height,o=s?0:c.sizeDiff.width,a={width:c.helper.width()-o,height:c.helper.height()-n},r=parseFloat(c.element.css("left"))+(c.position.left-c.originalPosition.left)||null,h=parseFloat(c.element.css("top"))+(c.position.top-c.originalPosition.top)||null,l.animate||this.element.css(t.extend(a,{top:h,left:r})),c.helper.height(c.size.height),c.helper.width(c.size.width),this._helper&&!l.animate&&this._proportionallyResize()),t("body").css("cursor","auto"),this._removeClass("ui-resizable-resizing"),this._propagate("stop",e),this._helper&&this.helper.remove(),!1},_updatePrevProperties:function(){this.prevPosition={top:this.position.top,left:this.position.left},this.prevSize={width:this.size.width,height:this.size.height}},_applyChanges:function(){var t={};return this.position.top!==this.prevPosition.top&&(t.top=this.position.top+"px"),this.position.left!==this.prevPosition.left&&(t.left=this.position.left+"px"),this.size.width!==this.prevSize.width&&(t.width=this.size.width+"px"),this.size.height!==this.prevSize.height&&(t.height=this.size.height+"px"),this.helper.css(t),t},_updateVirtualBoundaries:function(t){var e,i,s,n,o,a=this.options;o={minWidth:this._isNumber(a.minWidth)?a.minWidth:0,maxWidth:this._isNumber(a.maxWidth)?a.maxWidth:1/0,minHeight:this._isNumber(a.minHeight)?a.minHeight:0,maxHeight:this._isNumber(a.maxHeight)?a.maxHeight:1/0},(this._aspectRatio||t)&&(e=o.minHeight*this.aspectRatio,s=o.minWidth/this.aspectRatio,i=o.maxHeight*this.aspectRatio,n=o.maxWidth/this.aspectRatio,e>o.minWidth&&(o.minWidth=e),s>o.minHeight&&(o.minHeight=s),o.maxWidth>i&&(o.maxWidth=i),o.maxHeight>n&&(o.maxHeight=n)),this._vBoundaries=o},_updateCache:function(t){this.offset=this.helper.offset(),this._isNumber(t.left)&&(this.position.left=t.left),this._isNumber(t.top)&&(this.position.top=t.top),this._isNumber(t.height)&&(this.size.height=t.height),this._isNumber(t.width)&&(this.size.width=t.width)},_updateRatio:function(t){var e=this.position,i=this.size,s=this.axis;return this._isNumber(t.height)?t.width=t.height*this.aspectRatio:this._isNumber(t.width)&&(t.height=t.width/this.aspectRatio),"sw"===s&&(t.left=e.left+(i.width-t.width),t.top=null),"nw"===s&&(t.top=e.top+(i.height-t.height),t.left=e.left+(i.width-t.width)),t},_respectSize:function(t){var e=this._vBoundaries,i=this.axis,s=this._isNumber(t.width)&&e.maxWidth&&e.maxWidth<t.width,n=this._isNumber(t.height)&&e.maxHeight&&e.maxHeight<t.height,o=this._isNumber(t.width)&&e.minWidth&&e.minWidth>t.width,a=this._isNumber(t.height)&&e.minHeight&&e.minHeight>t.height,r=this.originalPosition.left+this.originalSize.width,h=this.originalPosition.top+this.originalSize.height,l=/sw|nw|w/.test(i),c=/nw|ne|n/.test(i);return o&&(t.width=e.minWidth),a&&(t.height=e.minHeight),s&&(t.width=e.maxWidth),n&&(t.height=e.maxHeight),o&&l&&(t.left=r-e.minWidth),s&&l&&(t.left=r-e.maxWidth),a&&c&&(t.top=h-e.minHeight),n&&c&&(t.top=h-e.maxHeight),t.width||t.height||t.left||!t.top?t.width||t.height||t.top||!t.left||(t.left=null):t.top=null,t},_getPaddingPlusBorderDimensions:function(t){for(var e=0,i=,s=t.css("borderTopWidth"),t.css("borderRightWidth"),t.css("borderBottomWidth"),t.css("borderLeftWidth"),n=t.css("paddingTop"),t.css("paddingRight"),t.css("paddingBottom"),t.css("paddingLeft");4>e;e++)ie=parseFloat(se)||0,ie+=parseFloat(ne)||0;return{height:i0+i2,width:i1+i3}},_proportionallyResize:function(){if(this._proportionallyResizeElements.length)for(var t,e=0,i=this.helper||this.element;this._proportionallyResizeElements.length>e;e++)t=this._proportionallyResizeElementse,this.outerDimensions||(this.outerDimensions=this._getPaddingPlusBorderDimensions(t)),t.css({height:i.height()-this.outerDimensions.height||0,width:i.width()-this.outerDimensions.width||0})},_renderProxy:function(){var e=this.element,i=this.options;this.elementOffset=e.offset(),this._helper?(this.helper=this.helper||t("<div style='overflow:hidden;'></div>"),this._addClass(this.helper,this._helper),this.helper.css({width:this.element.outerWidth(),height:this.element.outerHeight(),position:"absolute",left:this.elementOffset.left+"px",top:this.elementOffset.top+"px",zIndex:++i.zIndex}),this.helper.appendTo("body").disableSelection()):this.helper=this.element
+},_change:{e:function(t,e){return{width:this.originalSize.width+e}},w:function(t,e){var i=this.originalSize,s=this.originalPosition;return{left:s.left+e,width:i.width-e}},n:function(t,e,i){var s=this.originalSize,n=this.originalPosition;return{top:n.top+i,height:s.height-i}},s:function(t,e,i){return{height:this.originalSize.height+i}},se:function(e,i,s){return t.extend(this._change.s.apply(this,arguments),this._change.e.apply(this,e,i,s))},sw:function(e,i,s){return t.extend(this._change.s.apply(this,arguments),this._change.w.apply(this,e,i,s))},ne:function(e,i,s){return t.extend(this._change.n.apply(this,arguments),this._change.e.apply(this,e,i,s))},nw:function(e,i,s){return t.extend(this._change.n.apply(this,arguments),this._change.w.apply(this,e,i,s))}},_propagate:function(e,i){t.ui.plugin.call(this,e,i,this.ui()),"resize"!==e&&this._trigger(e,i,this.ui())},plugins:{},ui:function(){return{originalElement:this.originalElement,element:this.element,helper:this.helper,position:this.position,size:this.size,originalSize:this.originalSize,originalPosition:this.originalPosition}}}),t.ui.plugin.add("resizable","animate",{stop:function(e){var i=t(this).resizable("instance"),s=i.options,n=i._proportionallyResizeElements,o=n.length&&/textarea/i.test(n0.nodeName),a=o&&i._hasScroll(n0,"left")?0:i.sizeDiff.height,r=o?0:i.sizeDiff.width,h={width:i.size.width-r,height:i.size.height-a},l=parseFloat(i.element.css("left"))+(i.position.left-i.originalPosition.left)||null,c=parseFloat(i.element.css("top"))+(i.position.top-i.originalPosition.top)||null;i.element.animate(t.extend(h,c&&l?{top:c,left:l}:{}),{duration:s.animateDuration,easing:s.animateEasing,step:function(){var s={width:parseFloat(i.element.css("width")),height:parseFloat(i.element.css("height")),top:parseFloat(i.element.css("top")),left:parseFloat(i.element.css("left"))};n&&n.length&&t(n0).css({width:s.width,height:s.height}),i._updateCache(s),i._propagate("resize",e)}})}}),t.ui.plugin.add("resizable","containment",{start:function(){var e,i,s,n,o,a,r,h=t(this).resizable("instance"),l=h.options,c=h.element,u=l.containment,d=u instanceof t?u.get(0):/parent/.test(u)?c.parent().get(0):u;d&&(h.containerElement=t(d),/document/.test(u)||u===document?(h.containerOffset={left:0,top:0},h.containerPosition={left:0,top:0},h.parentData={element:t(document),left:0,top:0,width:t(document).width(),height:t(document).height()||document.body.parentNode.scrollHeight}):(e=t(d),i=,t("Top","Right","Left","Bottom").each(function(t,s){it=h._num(e.css("padding"+s))}),h.containerOffset=e.offset(),h.containerPosition=e.position(),h.containerSize={height:e.innerHeight()-i3,width:e.innerWidth()-i1},s=h.containerOffset,n=h.containerSize.height,o=h.containerSize.width,a=h._hasScroll(d,"left")?d.scrollWidth:o,r=h._hasScroll(d)?d.scrollHeight:n,h.parentData={element:d,left:s.left,top:s.top,width:a,height:r}))},resize:function(e){var i,s,n,o,a=t(this).resizable("instance"),r=a.options,h=a.containerOffset,l=a.position,c=a._aspectRatio||e.shiftKey,u={top:0,left:0},d=a.containerElement,p=!0;d0!==document&&/static/.test(d.css("position"))&&(u=h),l.left<(a._helper?h.left:0)&&(a.size.width=a.size.width+(a._helper?a.position.left-h.left:a.position.left-u.left),c&&(a.size.height=a.size.width/a.aspectRatio,p=!1),a.position.left=r.helper?h.left:0),l.top<(a._helper?h.top:0)&&(a.size.height=a.size.height+(a._helper?a.position.top-h.top:a.position.top),c&&(a.size.width=a.size.height*a.aspectRatio,p=!1),a.position.top=a._helper?h.top:0),n=a.containerElement.get(0)===a.element.parent().get(0),o=/relative|absolute/.test(a.containerElement.css("position")),n&&o?(a.offset.left=a.parentData.left+a.position.left,a.offset.top=a.parentData.top+a.position.top):(a.offset.left=a.element.offset().left,a.offset.top=a.element.offset().top),i=Math.abs(a.sizeDiff.width+(a._helper?a.offset.left-u.left:a.offset.left-h.left)),s=Math.abs(a.sizeDiff.height+(a._helper?a.offset.top-u.top:a.offset.top-h.top)),i+a.size.width>=a.parentData.width&&(a.size.width=a.parentData.width-i,c&&(a.size.height=a.size.width/a.aspectRatio,p=!1)),s+a.size.height>=a.parentData.height&&(a.size.height=a.parentData.height-s,c&&(a.size.width=a.size.height*a.aspectRatio,p=!1)),p||(a.position.left=a.prevPosition.left,a.position.top=a.prevPosition.top,a.size.width=a.prevSize.width,a.size.height=a.prevSize.height)},stop:function(){var e=t(this).resizable("instance"),i=e.options,s=e.containerOffset,n=e.containerPosition,o=e.containerElement,a=t(e.helper),r=a.offset(),h=a.outerWidth()-e.sizeDiff.width,l=a.outerHeight()-e.sizeDiff.height;e._helper&&!i.animate&&/relative/.test(o.css("position"))&&t(this).css({left:r.left-n.left-s.left,width:h,height:l}),e._helper&&!i.animate&&/static/.test(o.css("position"))&&t(this).css({left:r.left-n.left-s.left,width:h,height:l})}}),t.ui.plugin.add("resizable","alsoResize",{start:function(){var e=t(this).resizable("instance"),i=e.options;t(i.alsoResize).each(function(){var e=t(this);e.data("ui-resizable-alsoresize",{width:parseFloat(e.width()),height:parseFloat(e.height()),left:parseFloat(e.css("left")),top:parseFloat(e.css("top"))})})},resize:function(e,i){var s=t(this).resizable("instance"),n=s.options,o=s.originalSize,a=s.originalPosition,r={height:s.size.height-o.height||0,width:s.size.width-o.width||0,top:s.position.top-a.top||0,left:s.position.left-a.left||0};t(n.alsoResize).each(function(){var e=t(this),s=t(this).data("ui-resizable-alsoresize"),n={},o=e.parents(i.originalElement0).length?"width","height":"width","height","top","left";t.each(o,function(t,e){var i=(se||0)+(re||0);i&&i>=0&&(ne=i||null)}),e.css(n)})},stop:function(){t(this).removeData("ui-resizable-alsoresize")}}),t.ui.plugin.add("resizable","ghost",{start:function(){var e=t(this).resizable("instance"),i=e.size;e.ghost=e.originalElement.clone(),e.ghost.css({opacity:.25,display:"block",position:"relative",height:i.height,width:i.width,margin:0,left:0,top:0}),e._addClass(e.ghost,"ui-resizable-ghost"),t.uiBackCompat!==!1&&"string"==typeof e.options.ghost&&e.ghost.addClass(this.options.ghost),e.ghost.appendTo(e.helper)},resize:function(){var e=t(this).resizable("instance");e.ghost&&e.ghost.css({position:"relative",height:e.size.height,width:e.size.width})},stop:function(){var e=t(this).resizable("instance");e.ghost&&e.helper&&e.helper.get(0).removeChild(e.ghost.get(0))}}),t.ui.plugin.add("resizable","grid",{resize:function(){var e,i=t(this).resizable("instance"),s=i.options,n=i.size,o=i.originalSize,a=i.originalPosition,r=i.axis,h="number"==typeof s.grid?s.grid,s.grid:s.grid,l=h0||1,c=h1||1,u=Math.round((n.width-o.width)/l)*l,d=Math.round((n.height-o.height)/c)*c,p=o.width+u,f=o.height+d,m=s.maxWidth&&p>s.maxWidth,g=s.maxHeight&&f>s.maxHeight,_=s.minWidth&&s.minWidth>p,v=s.minHeight&&s.minHeight>f;s.grid=h,_&&(p+=l),v&&(f+=c),m&&(p-=l),g&&(f-=c),/^(se|s|e)$/.test(r)?(i.size.width=p,i.size.height=f):/^(ne)$/.test(r)?(i.size.width=p,i.size.height=f,i.position.top=a.top-d):/^(sw)$/.test(r)?(i.size.width=p,i.size.height=f,i.position.left=a.left-u):((0>=f-c||0>=p-l)&&(e=i._getPaddingPlusBorderDimensions(this)),f-c>0?(i.size.height=f,i.position.top=a.top-d):(f=c-e.height,i.size.height=f,i.position.top=a.top+o.height-f),p-l>0?(i.size.width=p,i.position.left=a.left-u):(p=l-e.width,i.size.width=p,i.position.left=a.left+o.width-p))}}),t.ui.resizable});/**
+ * Copyright (c) 2007 Ariel Flesler - aflesler ○ gmail • com | https://github.com/flesler
+ * Licensed under MIT
+ * @author Ariel Flesler
+ * @version 2.1.2
+ */
+;(function(f){"use strict";"function"===typeof define&&define.amd?define("jquery",f):"undefined"!==typeof module&&module.exports?module.exports=f(require("jquery")):f(jQuery)})(function($){"use strict";function n(a){return!a.nodeName||-1!==$.inArray(a.nodeName.toLowerCase(),"iframe","#document","html","body")}function h(a){return $.isFunction(a)||$.isPlainObject(a)?a:{top:a,left:a}}var p=$.scrollTo=function(a,d,b){return $(window).scrollTo(a,d,b)};p.defaults={axis:"xy",duration:0,limit:!0};$.fn.scrollTo=function(a,d,b){"object"=== typeof d&&(b=d,d=0);"function"===typeof b&&(b={onAfter:b});"max"===a&&(a=9E9);b=$.extend({},p.defaults,b);d=d||b.duration;var u=b.queue&&1<b.axis.length;u&&(d/=2);b.offset=h(b.offset);b.over=h(b.over);return this.each(function(){function k(a){var k=$.extend({},b,{queue:!0,duration:d,complete:a&&function(){a.call(q,e,b)}});r.animate(f,k)}if(null!==a){var l=n(this),q=l?this.contentWindow||window:this,r=$(q),e=a,f={},t;switch(typeof e){case "number":case "string":if(/^(+-=?)?\d+(\.\d+)?(px|%)?$/.test(e)){e= h(e);break}e=l?$(e):$(e,q);case "object":if(e.length===0)return;if(e.is||e.style)t=(e=$(e)).offset()}var v=$.isFunction(b.offset)&&b.offset(q,e)||b.offset;$.each(b.axis.split(""),function(a,c){var d="x"===c?"Left":"Top",m=d.toLowerCase(),g="scroll"+d,h=rg(),n=p.max(q,c);t?(fg=tm+(l?0:h-r.offset()m),b.margin&&(fg-=parseInt(e.css("margin"+d),10)||0,fg-=parseInt(e.css("border"+d+"Width"),10)||0),fg+=vm||0,b.overm&&(fg+=e"x"===c?"width":"height"()*b.overm)):(d=em,fg=d.slice&& "%"===d.slice(-1)?parseFloat(d)/100*n:d);b.limit&&/^\d+$/.test(fg)&&(fg=0>=fg?0:Math.min(fg,n));!a&&1<b.axis.length&&(h===fg?f={}:u&&(k(b.onAfterFirst),f={}))});k(b.onAfter)}})};p.max=function(a,d){var b="x"===d?"Width":"Height",h="scroll"+b;if(!n(a))return ah-$(a)b.toLowerCase()();var b="client"+b,k=a.ownerDocument||a.document,l=k.documentElement,k=k.body;return Math.max(lh,kh)-Math.min(lb,kb)};$.Tween.propHooks.scrollLeft=$.Tween.propHooks.scrollTop={get:function(a){return $(a.elem)a.prop()}, set:function(a){var d=this.get(a);if(a.options.interrupt&&a._last&&a._last!==d)return $(a.elem).stop();var b=Math.round(a.now);d!==b&&($(a.elem)a.prop(b),a._last=this.get(a))}};return p});
+/*!
+ PowerTip v1.3.1 (2018-04-15)
+ https://stevenbenner.github.io/jquery-powertip/
+ Copyright (c) 2018 Steven Benner (http://stevenbenner.com/).
+ Released under MIT license.
+ https://raw.github.com/stevenbenner/jquery-powertip/master/LICENSE.txt
+*/
+(function(root,factory){if(typeof define==="function"&&define.amd){define("jquery",factory)}else if(typeof module==="object"&&module.exports){module.exports=factory(require("jquery"))}else{factory(root.jQuery)}})(this,function($){var $document=$(document),$window=$(window),$body=$("body");var DATA_DISPLAYCONTROLLER="displayController",DATA_HASACTIVEHOVER="hasActiveHover",DATA_FORCEDOPEN="forcedOpen",DATA_HASMOUSEMOVE="hasMouseMove",DATA_MOUSEONTOTIP="mouseOnToPopup",DATA_ORIGINALTITLE="originalTitle",DATA_POWERTIP="powertip",DATA_POWERTIPJQ="powertipjq",DATA_POWERTIPTARGET="powertiptarget",EVENT_NAMESPACE=".powertip",RAD2DEG=180/Math.PI,MOUSE_EVENTS="click","dblclick","mousedown","mouseup","mousemove","mouseover","mouseout","mouseenter","mouseleave","contextmenu";var session={tooltips:null,isTipOpen:false,isFixedTipOpen:false,isClosing:false,tipOpenImminent:false,activeHover:null,currentX:0,currentY:0,previousX:0,previousY:0,desyncTimeout:null,closeDelayTimeout:null,mouseTrackingActive:false,delayInProgress:false,windowWidth:0,windowHeight:0,scrollTop:0,scrollLeft:0};var Collision={none:0,top:1,bottom:2,left:4,right:8};$.fn.powerTip=function(opts,arg){var targetElements=this,options,tipController;if(!targetElements.length){return targetElements}if($.type(opts)==="string"&&$.powerTipopts){return $.powerTipopts.call(targetElements,targetElements,arg)}options=$.extend({},$.fn.powerTip.defaults,opts);tipController=new TooltipController(options);initTracking();targetElements.each(function elementSetup(){var $this=$(this),dataPowertip=$this.data(DATA_POWERTIP),dataElem=$this.data(DATA_POWERTIPJQ),dataTarget=$this.data(DATA_POWERTIPTARGET),title=$this.attr("title");if(!dataPowertip&&!dataTarget&&!dataElem&&title){$this.data(DATA_POWERTIP,title);$this.data(DATA_ORIGINALTITLE,title);$this.removeAttr("title")}$this.data(DATA_DISPLAYCONTROLLER,new DisplayController($this,options,tipController))});if(!options.manual){$.each(options.openEvents,function(idx,evt){if($.inArray(evt,options.closeEvents)>-1){targetElements.on(evt+EVENT_NAMESPACE,function elementToggle(event){$.powerTip.toggle(this,event)})}else{targetElements.on(evt+EVENT_NAMESPACE,function elementOpen(event){$.powerTip.show(this,event)})}});$.each(options.closeEvents,function(idx,evt){if($.inArray(evt,options.openEvents)<0){targetElements.on(evt+EVENT_NAMESPACE,function elementClose(event){$.powerTip.hide(this,!isMouseEvent(event))})}});targetElements.on("keydown"+EVENT_NAMESPACE,function elementKeyDown(event){if(event.keyCode===27){$.powerTip.hide(this,true)}})}return targetElements};$.fn.powerTip.defaults={fadeInTime:200,fadeOutTime:100,followMouse:false,popupId:"powerTip",popupClass:null,intentSensitivity:7,intentPollInterval:100,closeDelay:100,placement:"n",smartPlacement:false,offset:10,mouseOnToPopup:false,manual:false,openEvents:"mouseenter","focus",closeEvents:"mouseleave","blur"};$.fn.powerTip.smartPlacementLists={n:"n","ne","nw","s",e:"e","ne","se","w","nw","sw","n","s","e",s:"s","se","sw","n",w:"w","nw","sw","e","ne","se","n","s","w",nw:"nw","w","sw","n","s","se","nw",ne:"ne","e","se","n","s","sw","ne",sw:"sw","w","nw","s","n","ne","sw",se:"se","e","ne","s","n","nw","se","nw-alt":"nw-alt","n","ne-alt","sw-alt","s","se-alt","w","e","ne-alt":"ne-alt","n","nw-alt","se-alt","s","sw-alt","e","w","sw-alt":"sw-alt","s","se-alt","nw-alt","n","ne-alt","w","e","se-alt":"se-alt","s","sw-alt","ne-alt","n","nw-alt","e","w"};$.powerTip={show:function apiShowTip(element,event){if(isMouseEvent(event)){trackMouse(event);session.previousX=event.pageX;session.previousY=event.pageY;$(element).data(DATA_DISPLAYCONTROLLER).show()}else{$(element).first().data(DATA_DISPLAYCONTROLLER).show(true,true)}return element},reposition:function apiResetPosition(element){$(element).first().data(DATA_DISPLAYCONTROLLER).resetPosition();return element},hide:function apiCloseTip(element,immediate){var displayController;immediate=element?immediate:true;if(element){displayController=$(element).first().data(DATA_DISPLAYCONTROLLER)}else if(session.activeHover){displayController=session.activeHover.data(DATA_DISPLAYCONTROLLER)}if(displayController){displayController.hide(immediate)}return element},toggle:function apiToggle(element,event){if(session.activeHover&&session.activeHover.is(element)){$.powerTip.hide(element,!isMouseEvent(event))}else{$.powerTip.show(element,event)}return element}};$.powerTip.showTip=$.powerTip.show;$.powerTip.closeTip=$.powerTip.hide;function CSSCoordinates(){var me=this;me.top="auto";me.left="auto";me.right="auto";me.bottom="auto";me.set=function(property,value){if($.isNumeric(value)){meproperty=Math.round(value)}}}function DisplayController(element,options,tipController){var hoverTimer=null,myCloseDelay=null;function openTooltip(immediate,forceOpen){cancelTimer();if(!element.data(DATA_HASACTIVEHOVER)){if(!immediate){session.tipOpenImminent=true;hoverTimer=setTimeout(function intentDelay(){hoverTimer=null;checkForIntent()},options.intentPollInterval)}else{if(forceOpen){element.data(DATA_FORCEDOPEN,true)}closeAnyDelayed();tipController.showTip(element)}}else{cancelClose()}}function closeTooltip(disableDelay){if(myCloseDelay){myCloseDelay=session.closeDelayTimeout=clearTimeout(myCloseDelay);session.delayInProgress=false}cancelTimer();session.tipOpenImminent=false;if(element.data(DATA_HASACTIVEHOVER)){element.data(DATA_FORCEDOPEN,false);if(!disableDelay){session.delayInProgress=true;session.closeDelayTimeout=setTimeout(function closeDelay(){session.closeDelayTimeout=null;tipController.hideTip(element);session.delayInProgress=false;myCloseDelay=null},options.closeDelay);myCloseDelay=session.closeDelayTimeout}else{tipController.hideTip(element)}}}function checkForIntent(){var xDifference=Math.abs(session.previousX-session.currentX),yDifference=Math.abs(session.previousY-session.currentY),totalDifference=xDifference+yDifference;if(totalDifference<options.intentSensitivity){cancelClose();closeAnyDelayed();tipController.showTip(element)}else{session.previousX=session.currentX;session.previousY=session.currentY;openTooltip()}}function cancelTimer(stopClose){hoverTimer=clearTimeout(hoverTimer);if(session.closeDelayTimeout&&myCloseDelay===session.closeDelayTimeout||stopClose){cancelClose()}}function cancelClose(){session.closeDelayTimeout=clearTimeout(session.closeDelayTimeout);session.delayInProgress=false}function closeAnyDelayed(){if(session.delayInProgress&&session.activeHover&&!session.activeHover.is(element)){session.activeHover.data(DATA_DISPLAYCONTROLLER).hide(true)}}function repositionTooltip(){tipController.resetPosition(element)}this.show=openTooltip;this.hide=closeTooltip;this.cancel=cancelTimer;this.resetPosition=repositionTooltip}function PlacementCalculator(){function computePlacementCoords(element,placement,tipWidth,tipHeight,offset){var placementBase=placement.split("-")0,coords=new CSSCoordinates,position;if(isSvgElement(element)){position=getSvgPlacement(element,placementBase)}else{position=getHtmlPlacement(element,placementBase)}switch(placement){case"n":coords.set("left",position.left-tipWidth/2);coords.set("bottom",session.windowHeight-position.top+offset);break;case"e":coords.set("left",position.left+offset);coords.set("top",position.top-tipHeight/2);break;case"s":coords.set("left",position.left-tipWidth/2);coords.set("top",position.top+offset);break;case"w":coords.set("top",position.top-tipHeight/2);coords.set("right",session.windowWidth-position.left+offset);break;case"nw":coords.set("bottom",session.windowHeight-position.top+offset);coords.set("right",session.windowWidth-position.left-20);break;case"nw-alt":coords.set("left",position.left);coords.set("bottom",session.windowHeight-position.top+offset);break;case"ne":coords.set("left",position.left-20);coords.set("bottom",session.windowHeight-position.top+offset);break;case"ne-alt":coords.set("bottom",session.windowHeight-position.top+offset);coords.set("right",session.windowWidth-position.left);break;case"sw":coords.set("top",position.top+offset);coords.set("right",session.windowWidth-position.left-20);break;case"sw-alt":coords.set("left",position.left);coords.set("top",position.top+offset);break;case"se":coords.set("left",position.left-20);coords.set("top",position.top+offset);break;case"se-alt":coords.set("top",position.top+offset);coords.set("right",session.windowWidth-position.left);break}return coords}function getHtmlPlacement(element,placement){var objectOffset=element.offset(),objectWidth=element.outerWidth(),objectHeight=element.outerHeight(),left,top;switch(placement){case"n":left=objectOffset.left+objectWidth/2;top=objectOffset.top;break;case"e":left=objectOffset.left+objectWidth;top=objectOffset.top+objectHeight/2;break;case"s":left=objectOffset.left+objectWidth/2;top=objectOffset.top+objectHeight;break;case"w":left=objectOffset.left;top=objectOffset.top+objectHeight/2;break;case"nw":left=objectOffset.left;top=objectOffset.top;break;case"ne":left=objectOffset.left+objectWidth;top=objectOffset.top;break;case"sw":left=objectOffset.left;top=objectOffset.top+objectHeight;break;case"se":left=objectOffset.left+objectWidth;top=objectOffset.top+objectHeight;break}return{top:top,left:left}}function getSvgPlacement(element,placement){var svgElement=element.closest("svg")0,domElement=element0,point=svgElement.createSVGPoint(),boundingBox=domElement.getBBox(),matrix=domElement.getScreenCTM(),halfWidth=boundingBox.width/2,halfHeight=boundingBox.height/2,placements=,placementKeys="nw","n","ne","e","se","s","sw","w",coords,rotation,steps,x;function pushPlacement(){placements.push(point.matrixTransform(matrix))}point.x=boundingBox.x;point.y=boundingBox.y;pushPlacement();point.x+=halfWidth;pushPlacement();point.x+=halfWidth;pushPlacement();point.y+=halfHeight;pushPlacement();point.y+=halfHeight;pushPlacement();point.x-=halfWidth;pushPlacement();point.x-=halfWidth;pushPlacement();point.y-=halfHeight;pushPlacement();if(placements0.y!==placements1.y||placements0.x!==placements7.x){rotation=Math.atan2(matrix.b,matrix.a)*RAD2DEG;steps=Math.ceil((rotation%360-22.5)/45);if(steps<1){steps+=8}while(steps--){placementKeys.push(placementKeys.shift())}}for(x=0;x<placements.length;x++){if(placementKeysx===placement){coords=placementsx;break}}return{top:coords.y+session.scrollTop,left:coords.x+session.scrollLeft}}this.compute=computePlacementCoords}function TooltipController(options){var placementCalculator=new PlacementCalculator,tipElement=$("#"+options.popupId);if(tipElement.length===0){tipElement=$("<div/>",{id:options.popupId});if($body.length===0){$body=$("body")}$body.append(tipElement);session.tooltips=session.tooltips?session.tooltips.add(tipElement):tipElement}if(options.followMouse){if(!tipElement.data(DATA_HASMOUSEMOVE)){$document.on("mousemove"+EVENT_NAMESPACE,positionTipOnCursor);$window.on("scroll"+EVENT_NAMESPACE,positionTipOnCursor);tipElement.data(DATA_HASMOUSEMOVE,true)}}function beginShowTip(element){element.data(DATA_HASACTIVEHOVER,true);tipElement.queue(function queueTipInit(next){showTip(element);next()})}function showTip(element){var tipContent;if(!element.data(DATA_HASACTIVEHOVER)){return}if(session.isTipOpen){if(!session.isClosing){hideTip(session.activeHover)}tipElement.delay(100).queue(function queueTipAgain(next){showTip(element);next()});return}element.trigger("powerTipPreRender");tipContent=getTooltipContent(element);if(tipContent){tipElement.empty().append(tipContent)}else{return}element.trigger("powerTipRender");session.activeHover=element;session.isTipOpen=true;tipElement.data(DATA_MOUSEONTOTIP,options.mouseOnToPopup);tipElement.addClass(options.popupClass);if(!options.followMouse||element.data(DATA_FORCEDOPEN)){positionTipOnElement(element);session.isFixedTipOpen=true}else{positionTipOnCursor()}if(!element.data(DATA_FORCEDOPEN)&&!options.followMouse){$document.on("click"+EVENT_NAMESPACE,function documentClick(event){var target=event.target;if(target!==element0){if(options.mouseOnToPopup){if(target!==tipElement0&&!$.contains(tipElement0,target)){$.powerTip.hide()}}else{$.powerTip.hide()}}})}if(options.mouseOnToPopup&&!options.manual){tipElement.on("mouseenter"+EVENT_NAMESPACE,function tipMouseEnter(){if(session.activeHover){session.activeHover.data(DATA_DISPLAYCONTROLLER).cancel()}});tipElement.on("mouseleave"+EVENT_NAMESPACE,function tipMouseLeave(){if(session.activeHover){session.activeHover.data(DATA_DISPLAYCONTROLLER).hide()}})}tipElement.fadeIn(options.fadeInTime,function fadeInCallback(){if(!session.desyncTimeout){session.desyncTimeout=setInterval(closeDesyncedTip,500)}element.trigger("powerTipOpen")})}function hideTip(element){session.isClosing=true;session.isTipOpen=false;session.desyncTimeout=clearInterval(session.desyncTimeout);element.data(DATA_HASACTIVEHOVER,false);element.data(DATA_FORCEDOPEN,false);$document.off("click"+EVENT_NAMESPACE);tipElement.off(EVENT_NAMESPACE);tipElement.fadeOut(options.fadeOutTime,function fadeOutCallback(){var coords=new CSSCoordinates;session.activeHover=null;session.isClosing=false;session.isFixedTipOpen=false;tipElement.removeClass();coords.set("top",session.currentY+options.offset);coords.set("left",session.currentX+options.offset);tipElement.css(coords);element.trigger("powerTipClose")})}function positionTipOnCursor(){var tipWidth,tipHeight,coords,collisions,collisionCount;if(!session.isFixedTipOpen&&(session.isTipOpen||session.tipOpenImminent&&tipElement.data(DATA_HASMOUSEMOVE))){tipWidth=tipElement.outerWidth();tipHeight=tipElement.outerHeight();coords=new CSSCoordinates;coords.set("top",session.currentY+options.offset);coords.set("left",session.currentX+options.offset);collisions=getViewportCollisions(coords,tipWidth,tipHeight);if(collisions!==Collision.none){collisionCount=countFlags(collisions);if(collisionCount===1){if(collisions===Collision.right){coords.set("left",session.scrollLeft+session.windowWidth-tipWidth)}else if(collisions===Collision.bottom){coords.set("top",session.scrollTop+session.windowHeight-tipHeight)}}else{coords.set("left",session.currentX-tipWidth-options.offset);coords.set("top",session.currentY-tipHeight-options.offset)}}tipElement.css(coords)}}function positionTipOnElement(element){var priorityList,finalPlacement;if(options.smartPlacement||options.followMouse&&element.data(DATA_FORCEDOPEN)){priorityList=$.fn.powerTip.smartPlacementListsoptions.placement;$.each(priorityList,function(idx,pos){var collisions=getViewportCollisions(placeTooltip(element,pos),tipElement.outerWidth(),tipElement.outerHeight());finalPlacement=pos;return collisions!==Collision.none})}else{placeTooltip(element,options.placement);finalPlacement=options.placement}tipElement.removeClass("w nw sw e ne se n s w se-alt sw-alt ne-alt nw-alt");tipElement.addClass(finalPlacement)}function placeTooltip(element,placement){var iterationCount=0,tipWidth,tipHeight,coords=new CSSCoordinates;coords.set("top",0);coords.set("left",0);tipElement.css(coords);do{tipWidth=tipElement.outerWidth();tipHeight=tipElement.outerHeight();coords=placementCalculator.compute(element,placement,tipWidth,tipHeight,options.offset);tipElement.css(coords)}while(++iterationCount<=5&&(tipWidth!==tipElement.outerWidth()||tipHeight!==tipElement.outerHeight()));return coords}function closeDesyncedTip(){var isDesynced=false,hasDesyncableCloseEvent=$.grep("mouseleave","mouseout","blur","focusout",function(eventType){return $.inArray(eventType,options.closeEvents)!==-1}).length>0;if(session.isTipOpen&&!session.isClosing&&!session.delayInProgress&&hasDesyncableCloseEvent){if(session.activeHover.data(DATA_HASACTIVEHOVER)===false||session.activeHover.is(":disabled")){isDesynced=true}else if(!isMouseOver(session.activeHover)&&!session.activeHover.is(":focus")&&!session.activeHover.data(DATA_FORCEDOPEN)){if(tipElement.data(DATA_MOUSEONTOTIP)){if(!isMouseOver(tipElement)){isDesynced=true}}else{isDesynced=true}}if(isDesynced){hideTip(session.activeHover)}}}this.showTip=beginShowTip;this.hideTip=hideTip;this.resetPosition=positionTipOnElement}function isSvgElement(element){return Boolean(window.SVGElement&&element0instanceof SVGElement)}function isMouseEvent(event){return Boolean(event&&$.inArray(event.type,MOUSE_EVENTS)>-1&&typeof event.pageX==="number")}function initTracking(){if(!session.mouseTrackingActive){session.mouseTrackingActive=true;getViewportDimensions();$(getViewportDimensions);$document.on("mousemove"+EVENT_NAMESPACE,trackMouse);$window.on("resize"+EVENT_NAMESPACE,trackResize);$window.on("scroll"+EVENT_NAMESPACE,trackScroll)}}function getViewportDimensions(){session.scrollLeft=$window.scrollLeft();session.scrollTop=$window.scrollTop();session.windowWidth=$window.width();session.windowHeight=$window.height()}function trackResize(){session.windowWidth=$window.width();session.windowHeight=$window.height()}function trackScroll(){var x=$window.scrollLeft(),y=$window.scrollTop();if(x!==session.scrollLeft){session.currentX+=x-session.scrollLeft;session.scrollLeft=x}if(y!==session.scrollTop){session.currentY+=y-session.scrollTop;session.scrollTop=y}}function trackMouse(event){session.currentX=event.pageX;session.currentY=event.pageY}function isMouseOver(element){var elementPosition=element.offset(),elementBox=element0.getBoundingClientRect(),elementWidth=elementBox.right-elementBox.left,elementHeight=elementBox.bottom-elementBox.top;return session.currentX>=elementPosition.left&&session.currentX<=elementPosition.left+elementWidth&&session.currentY>=elementPosition.top&&session.currentY<=elementPosition.top+elementHeight}function getTooltipContent(element){var tipText=element.data(DATA_POWERTIP),tipObject=element.data(DATA_POWERTIPJQ),tipTarget=element.data(DATA_POWERTIPTARGET),targetElement,content;if(tipText){if($.isFunction(tipText)){tipText=tipText.call(element0)}content=tipText}else if(tipObject){if($.isFunction(tipObject)){tipObject=tipObject.call(element0)}if(tipObject.length>0){content=tipObject.clone(true,true)}}else if(tipTarget){targetElement=$("#"+tipTarget);if(targetElement.length>0){content=targetElement.html()}}return content}function getViewportCollisions(coords,elementWidth,elementHeight){var viewportTop=session.scrollTop,viewportLeft=session.scrollLeft,viewportBottom=viewportTop+session.windowHeight,viewportRight=viewportLeft+session.windowWidth,collisions=Collision.none;if(coords.top<viewportTop||Math.abs(coords.bottom-session.windowHeight)-elementHeight<viewportTop){collisions|=Collision.top}if(coords.top+elementHeight>viewportBottom||Math.abs(coords.bottom-session.windowHeight)>viewportBottom){collisions|=Collision.bottom}if(coords.left<viewportLeft||coords.right+elementWidth>viewportRight){collisions|=Collision.left}if(coords.left+elementWidth>viewportRight||coords.right<viewportLeft){collisions|=Collision.right}return collisions}function countFlags(value){var count=0;while(value){value&=value-1;count++}return count}return $.powerTip});/*!
+ * jQuery UI Touch Punch 0.2.3
+ *
+ * Copyright 2011–2014, Dave Furfero
+ * Dual licensed under the MIT or GPL Version 2 licenses.
+ *
+ * Depends:
+ *  jquery.ui.widget.js
+ *  jquery.ui.mouse.js
+ */
+!function(a){function f(a,b){if(!(a.originalEvent.touches.length>1)){a.preventDefault();var c=a.originalEvent.changedTouches0,d=document.createEvent("MouseEvents");d.initMouseEvent(b,!0,!0,window,1,c.screenX,c.screenY,c.clientX,c.clientY,!1,!1,!1,!1,0,null),a.target.dispatchEvent(d)}}if(a.support.touch="ontouchend"in document,a.support.touch){var e,b=a.ui.mouse.prototype,c=b._mouseInit,d=b._mouseDestroy;b._touchStart=function(a){var b=this;!e&&b._mouseCapture(a.originalEvent.changedTouches0)&&(e=!0,b._touchMoved=!1,f(a,"mouseover"),f(a,"mousemove"),f(a,"mousedown"))},b._touchMove=function(a){e&&(this._touchMoved=!0,f(a,"mousemove"))},b._touchEnd=function(a){e&&(f(a,"mouseup"),f(a,"mouseout"),this._touchMoved||f(a,"click"),e=!1)},b._mouseInit=function(){var b=this;b.element.bind({touchstart:a.proxy(b,"_touchStart"),touchmove:a.proxy(b,"_touchMove"),touchend:a.proxy(b,"_touchEnd")}),c.call(b)},b._mouseDestroy=function(){var b=this;b.element.unbind({touchstart:a.proxy(b,"_touchStart"),touchmove:a.proxy(b,"_touchMove"),touchend:a.proxy(b,"_touchEnd")}),d.call(b)}}}(jQuery);/*! SmartMenus jQuery Plugin - v1.1.0 - September 17, 2017
+ * http://www.smartmenus.org/
+ * Copyright Vasil Dinkov, Vadikom Web Ltd. http://vadikom.com; Licensed MIT */(function(t){"function"==typeof define&&define.amd?define("jquery",t):"object"==typeof module&&"object"==typeof module.exports?module.exports=t(require("jquery")):t(jQuery)})(function($){function initMouseDetection(t){var e=".smartmenus_mouse";if(mouseDetectionEnabled||t)mouseDetectionEnabled&&t&&($(document).off(e),mouseDetectionEnabled=!1);else{var i=!0,s=null,o={mousemove:function(t){var e={x:t.pageX,y:t.pageY,timeStamp:(new Date).getTime()};if(s){var o=Math.abs(s.x-e.x),a=Math.abs(s.y-e.y);if((o>0||a>0)&&2>=o&&2>=a&&300>=e.timeStamp-s.timeStamp&&(mouse=!0,i)){var n=$(t.target).closest("a");n.is("a")&&$.each(menuTrees,function(){return $.contains(this.$root0,n0)?(this.itemEnter({currentTarget:n0}),!1):void 0}),i=!1}}s=e}};otouchEvents?"touchstart":"pointerover pointermove pointerout MSPointerOver MSPointerMove MSPointerOut"=function(t){isTouchEvent(t.originalEvent)&&(mouse=!1)},$(document).on(getEventsNS(o,e)),mouseDetectionEnabled=!0}}function isTouchEvent(t){return!/^(4|mouse)$/.test(t.pointerType)}function getEventsNS(t,e){e||(e="");var i={};for(var s in t)is.split(" ").join(e+" ")+e=ts;return i}var menuTrees=,mouse=!1,touchEvents="ontouchstart"in window,mouseDetectionEnabled=!1,requestAnimationFrame=window.requestAnimationFrame||function(t){return setTimeout(t,1e3/60)},cancelAnimationFrame=window.cancelAnimationFrame||function(t){clearTimeout(t)},canAnimate=!!$.fn.animate;return $.SmartMenus=function(t,e){this.$root=$(t),this.opts=e,this.rootId="",this.accessIdPrefix="",this.$subArrow=null,this.activatedItems=,this.visibleSubMenus=,this.showTimeout=0,this.hideTimeout=0,this.scrollTimeout=0,this.clickActivated=!1,this.focusActivated=!1,this.zIndexInc=0,this.idInc=0,this.$firstLink=null,this.$firstSub=null,this.disabled=!1,this.$disableOverlay=null,this.$touchScrollingSub=null,this.cssTransforms3d="perspective"in t.style||"webkitPerspective"in t.style,this.wasCollapsible=!1,this.init()},$.extend($.SmartMenus,{hideAll:function(){$.each(menuTrees,function(){this.menuHideAll()})},destroy:function(){for(;menuTrees.length;)menuTrees0.destroy();initMouseDetection(!0)},prototype:{init:function(t){var e=this;if(!t){menuTrees.push(this),this.rootId=((new Date).getTime()+Math.random()+"").replace(/\D/g,""),this.accessIdPrefix="sm-"+this.rootId+"-",this.$root.hasClass("sm-rtl")&&(this.opts.rightToLeftSubMenus=!0);var i=".smartmenus";this.$root.data("smartmenus",this).attr("data-smartmenus-id",this.rootId).dataSM("level",1).on(getEventsNS({"mouseover focusin":$.proxy(this.rootOver,this),"mouseout focusout":$.proxy(this.rootOut,this),keydown:$.proxy(this.rootKeyDown,this)},i)).on(getEventsNS({mouseenter:$.proxy(this.itemEnter,this),mouseleave:$.proxy(this.itemLeave,this),mousedown:$.proxy(this.itemDown,this),focus:$.proxy(this.itemFocus,this),blur:$.proxy(this.itemBlur,this),click:$.proxy(this.itemClick,this)},i),"a"),i+=this.rootId,this.opts.hideOnClick&&$(document).on(getEventsNS({touchstart:$.proxy(this.docTouchStart,this),touchmove:$.proxy(this.docTouchMove,this),touchend:$.proxy(this.docTouchEnd,this),click:$.proxy(this.docClick,this)},i)),$(window).on(getEventsNS({"resize orientationchange":$.proxy(this.winResize,this)},i)),this.opts.subIndicators&&(this.$subArrow=$("<span/>").addClass("sub-arrow"),this.opts.subIndicatorsText&&this.$subArrow.html(this.opts.subIndicatorsText)),initMouseDetection()}if(this.$firstSub=this.$root.find("ul").each(function(){e.menuInit($(this))}).eq(0),this.$firstLink=this.$root.find("a").eq(0),this.opts.markCurrentItem){var s=/(index|default)\.^#\?\/*/i,o=/#.*/,a=window.location.href.replace(s,""),n=a.replace(o,"");this.$root.find("a").each(function(){var t=this.href.replace(s,""),i=$(this);(t==a||t==n)&&(i.addClass("current"),e.opts.markCurrentTree&&i.parentsUntil("data-smartmenus-id","ul").each(function(){$(this).dataSM("parent-a").addClass("current")}))})}this.wasCollapsible=this.isCollapsible()},destroy:function(t){if(!t){var e=".smartmenus";this.$root.removeData("smartmenus").removeAttr("data-smartmenus-id").removeDataSM("level").off(e),e+=this.rootId,$(document).off(e),$(window).off(e),this.opts.subIndicators&&(this.$subArrow=null)}this.menuHideAll();var i=this;this.$root.find("ul").each(function(){var t=$(this);t.dataSM("scroll-arrows")&&t.dataSM("scroll-arrows").remove(),t.dataSM("shown-before")&&((i.opts.subMenusMinWidth||i.opts.subMenusMaxWidth)&&t.css({width:"",minWidth:"",maxWidth:""}).removeClass("sm-nowrap"),t.dataSM("scroll-arrows")&&t.dataSM("scroll-arrows").remove(),t.css({zIndex:"",top:"",left:"",marginLeft:"",marginTop:"",display:""})),0==(t.attr("id")||"").indexOf(i.accessIdPrefix)&&t.removeAttr("id")}).removeDataSM("in-mega").removeDataSM("shown-before").removeDataSM("scroll-arrows").removeDataSM("parent-a").removeDataSM("level").removeDataSM("beforefirstshowfired").removeAttr("role").removeAttr("aria-hidden").removeAttr("aria-labelledby").removeAttr("aria-expanded"),this.$root.find("a.has-submenu").each(function(){var t=$(this);0==t.attr("id").indexOf(i.accessIdPrefix)&&t.removeAttr("id")}).removeClass("has-submenu").removeDataSM("sub").removeAttr("aria-haspopup").removeAttr("aria-controls").removeAttr("aria-expanded").closest("li").removeDataSM("sub"),this.opts.subIndicators&&this.$root.find("span.sub-arrow").remove(),this.opts.markCurrentItem&&this.$root.find("a.current").removeClass("current"),t||(this.$root=null,this.$firstLink=null,this.$firstSub=null,this.$disableOverlay&&(this.$disableOverlay.remove(),this.$disableOverlay=null),menuTrees.splice($.inArray(this,menuTrees),1))},disable:function(t){if(!this.disabled){if(this.menuHideAll(),!t&&!this.opts.isPopup&&this.$root.is(":visible")){var e=this.$root.offset();this.$disableOverlay=$('<div class="sm-jquery-disable-overlay"/>').css({position:"absolute",top:e.top,left:e.left,width:this.$root.outerWidth(),height:this.$root.outerHeight(),zIndex:this.getStartZIndex(!0),opacity:0}).appendTo(document.body)}this.disabled=!0}},docClick:function(t){return this.$touchScrollingSub?(this.$touchScrollingSub=null,void 0):((this.visibleSubMenus.length&&!$.contains(this.$root0,t.target)||$(t.target).closest("a").length)&&this.menuHideAll(),void 0)},docTouchEnd:function(){if(this.lastTouch){if(!(!this.visibleSubMenus.length||void 0!==this.lastTouch.x2&&this.lastTouch.x1!=this.lastTouch.x2||void 0!==this.lastTouch.y2&&this.lastTouch.y1!=this.lastTouch.y2||this.lastTouch.target&&$.contains(this.$root0,this.lastTouch.target))){this.hideTimeout&&(clearTimeout(this.hideTimeout),this.hideTimeout=0);var t=this;this.hideTimeout=setTimeout(function(){t.menuHideAll()},350)}this.lastTouch=null}},docTouchMove:function(t){if(this.lastTouch){var e=t.originalEvent.touches0;this.lastTouch.x2=e.pageX,this.lastTouch.y2=e.pageY}},docTouchStart:function(t){var e=t.originalEvent.touches0;this.lastTouch={x1:e.pageX,y1:e.pageY,target:e.target}},enable:function(){this.disabled&&(this.$disableOverlay&&(this.$disableOverlay.remove(),this.$disableOverlay=null),this.disabled=!1)},getClosestMenu:function(t){for(var e=$(t).closest("ul");e.dataSM("in-mega");)e=e.parent().closest("ul");return e0||null},getHeight:function(t){return this.getOffset(t,!0)},getOffset:function(t,e){var i;"none"==t.css("display")&&(i={position:t0.style.position,visibility:t0.style.visibility},t.css({position:"absolute",visibility:"hidden"}).show());var s=t0.getBoundingClientRect&&t0.getBoundingClientRect(),o=s&&(e?s.height||s.bottom-s.top:s.width||s.right-s.left);return o||0===o||(o=e?t0.offsetHeight:t0.offsetWidth),i&&t.hide().css(i),o},getStartZIndex:function(t){var e=parseInt(thist?"$root":"$firstSub".css("z-index"));return!t&&isNaN(e)&&(e=parseInt(this.$root.css("z-index"))),isNaN(e)?1:e},getTouchPoint:function(t){return t.touches&&t.touches0||t.changedTouches&&t.changedTouches0||t},getViewport:function(t){var e=t?"Height":"Width",i=document.documentElement"client"+e,s=window"inner"+e;return s&&(i=Math.min(i,s)),i},getViewportHeight:function(){return this.getViewport(!0)},getViewportWidth:function(){return this.getViewport()},getWidth:function(t){return this.getOffset(t)},handleEvents:function(){return!this.disabled&&this.isCSSOn()},handleItemEvents:function(t){return this.handleEvents()&&!this.isLinkInMegaMenu(t)},isCollapsible:function(){return"static"==this.$firstSub.css("position")},isCSSOn:function(){return"inline"!=this.$firstLink.css("display")},isFixed:function(){var t="fixed"==this.$root.css("position");return t||this.$root.parentsUntil("body").each(function(){return"fixed"==$(this).css("position")?(t=!0,!1):void 0}),t},isLinkInMegaMenu:function(t){return $(this.getClosestMenu(t0)).hasClass("mega-menu")},isTouchMode:function(){return!mouse||this.opts.noMouseOver||this.isCollapsible()},itemActivate:function(t,e){var i=t.closest("ul"),s=i.dataSM("level");if(s>1&&(!this.activatedItemss-2||this.activatedItemss-20!=i.dataSM("parent-a")0)){var o=this;$(i.parentsUntil("data-smartmenus-id","ul").get().reverse()).add(i).each(function(){o.itemActivate($(this).dataSM("parent-a"))})}if((!this.isCollapsible()||e)&&this.menuHideSubMenus(this.activatedItemss-1&&this.activatedItemss-10==t0?s:s-1),this.activatedItemss-1=t,this.$root.triggerHandler("activate.smapi",t0)!==!1){var a=t.dataSM("sub");a&&(this.isTouchMode()||!this.opts.showOnClick||this.clickActivated)&&this.menuShow(a)}},itemBlur:function(t){var e=$(t.currentTarget);this.handleItemEvents(e)&&this.$root.triggerHandler("blur.smapi",e0)},itemClick:function(t){var e=$(t.currentTarget);if(this.handleItemEvents(e)){if(this.$touchScrollingSub&&this.$touchScrollingSub0==e.closest("ul")0)return this.$touchScrollingSub=null,t.stopPropagation(),!1;if(this.$root.triggerHandler("click.smapi",e0)===!1)return!1;var i=$(t.target).is(".sub-arrow"),s=e.dataSM("sub"),o=s?2==s.dataSM("level"):!1,a=this.isCollapsible(),n=/toggle$/.test(this.opts.collapsibleBehavior),r=/link$/.test(this.opts.collapsibleBehavior),h=/^accordion/.test(this.opts.collapsibleBehavior);if(s&&!s.is(":visible")){if((!r||!a||i)&&(this.opts.showOnClick&&o&&(this.clickActivated=!0),this.itemActivate(e,h),s.is(":visible")))return this.focusActivated=!0,!1}else if(a&&(n||i))return this.itemActivate(e,h),this.menuHide(s),n&&(this.focusActivated=!1),!1;return this.opts.showOnClick&&o||e.hasClass("disabled")||this.$root.triggerHandler("select.smapi",e0)===!1?!1:void 0}},itemDown:function(t){var e=$(t.currentTarget);this.handleItemEvents(e)&&e.dataSM("mousedown",!0)},itemEnter:function(t){var e=$(t.currentTarget);if(this.handleItemEvents(e)){if(!this.isTouchMode()){this.showTimeout&&(clearTimeout(this.showTimeout),this.showTimeout=0);var i=this;this.showTimeout=setTimeout(function(){i.itemActivate(e)},this.opts.showOnClick&&1==e.closest("ul").dataSM("level")?1:this.opts.showTimeout)}this.$root.triggerHandler("mouseenter.smapi",e0)}},itemFocus:function(t){var e=$(t.currentTarget);this.handleItemEvents(e)&&(!this.focusActivated||this.isTouchMode()&&e.dataSM("mousedown")||this.activatedItems.length&&this.activatedItemsthis.activatedItems.length-10==e0||this.itemActivate(e,!0),this.$root.triggerHandler("focus.smapi",e0))},itemLeave:function(t){var e=$(t.currentTarget);this.handleItemEvents(e)&&(this.isTouchMode()||(e0.blur(),this.showTimeout&&(clearTimeout(this.showTimeout),this.showTimeout=0)),e.removeDataSM("mousedown"),this.$root.triggerHandler("mouseleave.smapi",e0))},menuHide:function(t){if(this.$root.triggerHandler("beforehide.smapi",t0)!==!1&&(canAnimate&&t.stop(!0,!0),"none"!=t.css("display"))){var e=function(){t.css("z-index","")};this.isCollapsible()?canAnimate&&this.opts.collapsibleHideFunction?this.opts.collapsibleHideFunction.call(this,t,e):t.hide(this.opts.collapsibleHideDuration,e):canAnimate&&this.opts.hideFunction?this.opts.hideFunction.call(this,t,e):t.hide(this.opts.hideDuration,e),t.dataSM("scroll")&&(this.menuScrollStop(t),t.css({"touch-action":"","-ms-touch-action":"","-webkit-transform":"",transform:""}).off(".smartmenus_scroll").removeDataSM("scroll").dataSM("scroll-arrows").hide()),t.dataSM("parent-a").removeClass("highlighted").attr("aria-expanded","false"),t.attr({"aria-expanded":"false","aria-hidden":"true"});var i=t.dataSM("level");this.activatedItems.splice(i-1,1),this.visibleSubMenus.splice($.inArray(t,this.visibleSubMenus),1),this.$root.triggerHandler("hide.smapi",t0)}},menuHideAll:function(){this.showTimeout&&(clearTimeout(this.showTimeout),this.showTimeout=0);for(var t=this.opts.isPopup?1:0,e=this.visibleSubMenus.length-1;e>=t;e--)this.menuHide(this.visibleSubMenuse);this.opts.isPopup&&(canAnimate&&this.$root.stop(!0,!0),this.$root.is(":visible")&&(canAnimate&&this.opts.hideFunction?this.opts.hideFunction.call(this,this.$root):this.$root.hide(this.opts.hideDuration))),this.activatedItems=,this.visibleSubMenus=,this.clickActivated=!1,this.focusActivated=!1,this.zIndexInc=0,this.$root.triggerHandler("hideAll.smapi")},menuHideSubMenus:function(t){for(var e=this.activatedItems.length-1;e>=t;e--){var i=this.activatedItemse.dataSM("sub");i&&this.menuHide(i)}},menuInit:function(t){if(!t.dataSM("in-mega")){t.hasClass("mega-menu")&&t.find("ul").dataSM("in-mega",!0);for(var e=2,i=t0;(i=i.parentNode.parentNode)!=this.$root0;)e++;var s=t.prevAll("a").eq(-1);s.length||(s=t.prevAll().find("a").eq(-1)),s.addClass("has-submenu").dataSM("sub",t),t.dataSM("parent-a",s).dataSM("level",e).parent().dataSM("sub",t);var o=s.attr("id")||this.accessIdPrefix+ ++this.idInc,a=t.attr("id")||this.accessIdPrefix+ ++this.idInc;s.attr({id:o,"aria-haspopup":"true","aria-controls":a,"aria-expanded":"false"}),t.attr({id:a,role:"group","aria-hidden":"true","aria-labelledby":o,"aria-expanded":"false"}),this.opts.subIndicators&&sthis.opts.subIndicatorsPos(this.$subArrow.clone())}},menuPosition:function(t){var e,i,s=t.dataSM("parent-a"),o=s.closest("li"),a=o.parent(),n=t.dataSM("level"),r=this.getWidth(t),h=this.getHeight(t),u=s.offset(),l=u.left,c=u.top,d=this.getWidth(s),m=this.getHeight(s),p=$(window),f=p.scrollLeft(),v=p.scrollTop(),b=this.getViewportWidth(),S=this.getViewportHeight(),g=a.parent().is("data-sm-horizontal-sub")||2==n&&!a.hasClass("sm-vertical"),M=this.opts.rightToLeftSubMenus&&!o.is("data-sm-reverse")||!this.opts.rightToLeftSubMenus&&o.is("data-sm-reverse"),w=2==n?this.opts.mainMenuSubOffsetX:this.opts.subMenusSubOffsetX,T=2==n?this.opts.mainMenuSubOffsetY:this.opts.subMenusSubOffsetY;if(g?(e=M?d-r-w:w,i=this.opts.bottomToTopSubMenus?-h-T:m+T):(e=M?w-r:d-w,i=this.opts.bottomToTopSubMenus?m-T-h:T),this.opts.keepInViewport){var y=l+e,I=c+i;if(M&&f>y?e=g?f-y+e:d-w:!M&&y+r>f+b&&(e=g?f+b-r-y+e:w-r),g||(S>h&&I+h>v+S?i+=v+S-h-I:(h>=S||v>I)&&(i+=v-I)),g&&(I+h>v+S+.49||v>I)||!g&&h>S+.49){var x=this;t.dataSM("scroll-arrows")||t.dataSM("scroll-arrows",$($('<span class="scroll-up"><span class="scroll-up-arrow"></span></span>')0,$('<span class="scroll-down"><span class="scroll-down-arrow"></span></span>')0).on({mouseenter:function(){t.dataSM("scroll").up=$(this).hasClass("scroll-up"),x.menuScroll(t)},mouseleave:function(e){x.menuScrollStop(t),x.menuScrollOut(t,e)},"mousewheel DOMMouseScroll":function(t){t.preventDefault()}}).insertAfter(t));var A=".smartmenus_scroll";if(t.dataSM("scroll",{y:this.cssTransforms3d?0:i-m,step:1,itemH:m,subH:h,arrowDownH:this.getHeight(t.dataSM("scroll-arrows").eq(1))}).on(getEventsNS({mouseover:function(e){x.menuScrollOver(t,e)},mouseout:function(e){x.menuScrollOut(t,e)},"mousewheel DOMMouseScroll":function(e){x.menuScrollMousewheel(t,e)}},A)).dataSM("scroll-arrows").css({top:"auto",left:"0",marginLeft:e+(parseInt(t.css("border-left-width"))||0),width:r-(parseInt(t.css("border-left-width"))||0)-(parseInt(t.css("border-right-width"))||0),zIndex:t.css("z-index")}).eq(g&&this.opts.bottomToTopSubMenus?0:1).show(),this.isFixed()){var C={};CtouchEvents?"touchstart touchmove touchend":"pointerdown pointermove pointerup MSPointerDown MSPointerMove MSPointerUp"=function(e){x.menuScrollTouch(t,e)},t.css({"touch-action":"none","-ms-touch-action":"none"}).on(getEventsNS(C,A))}}}t.css({top:"auto",left:"0",marginLeft:e,marginTop:i-m})},menuScroll:function(t,e,i){var s,o=t.dataSM("scroll"),a=t.dataSM("scroll-arrows"),n=o.up?o.upEnd:o.downEnd;if(!e&&o.momentum){if(o.momentum*=.92,s=o.momentum,.5>s)return this.menuScrollStop(t),void 0}else s=i||(e||!this.opts.scrollAccelerate?this.opts.scrollStep:Math.floor(o.step));var r=t.dataSM("level");if(this.activatedItemsr-1&&this.activatedItemsr-1.dataSM("sub")&&this.activatedItemsr-1.dataSM("sub").is(":visible")&&this.menuHideSubMenus(r-1),o.y=o.up&&o.y>=n||!o.up&&n>=o.y?o.y:Math.abs(n-o.y)>s?o.y+(o.up?s:-s):n,t.css(this.cssTransforms3d?{"-webkit-transform":"translate3d(0, "+o.y+"px, 0)",transform:"translate3d(0, "+o.y+"px, 0)"}:{marginTop:o.y}),mouse&&(o.up&&o.y>o.downEnd||!o.up&&o.y<o.upEnd)&&a.eq(o.up?1:0).show(),o.y==n)mouse&&a.eq(o.up?0:1).hide(),this.menuScrollStop(t);else if(!e){this.opts.scrollAccelerate&&o.step<this.opts.scrollStep&&(o.step+=.2);var h=this;this.scrollTimeout=requestAnimationFrame(function(){h.menuScroll(t)})}},menuScrollMousewheel:function(t,e){if(this.getClosestMenu(e.target)==t0){e=e.originalEvent;var i=(e.wheelDelta||-e.detail)>0;t.dataSM("scroll-arrows").eq(i?0:1).is(":visible")&&(t.dataSM("scroll").up=i,this.menuScroll(t,!0))}e.preventDefault()},menuScrollOut:function(t,e){mouse&&(/^scroll-(up|down)/.test((e.relatedTarget||"").className)||(t0==e.relatedTarget||$.contains(t0,e.relatedTarget))&&this.getClosestMenu(e.relatedTarget)==t0||t.dataSM("scroll-arrows").css("visibility","hidden"))},menuScrollOver:function(t,e){if(mouse&&!/^scroll-(up|down)/.test(e.target.className)&&this.getClosestMenu(e.target)==t0){this.menuScrollRefreshData(t);var i=t.dataSM("scroll"),s=$(window).scrollTop()-t.dataSM("parent-a").offset().top-i.itemH;t.dataSM("scroll-arrows").eq(0).css("margin-top",s).end().eq(1).css("margin-top",s+this.getViewportHeight()-i.arrowDownH).end().css("visibility","visible")}},menuScrollRefreshData:function(t){var e=t.dataSM("scroll"),i=$(window).scrollTop()-t.dataSM("parent-a").offset().top-e.itemH;this.cssTransforms3d&&(i=-(parseFloat(t.css("margin-top"))-i)),$.extend(e,{upEnd:i,downEnd:i+this.getViewportHeight()-e.subH})},menuScrollStop:function(t){return this.scrollTimeout?(cancelAnimationFrame(this.scrollTimeout),this.scrollTimeout=0,t.dataSM("scroll").step=1,!0):void 0},menuScrollTouch:function(t,e){if(e=e.originalEvent,isTouchEvent(e)){var i=this.getTouchPoint(e);if(this.getClosestMenu(i.target)==t0){var s=t.dataSM("scroll");if(/(start|down)$/i.test(e.type))this.menuScrollStop(t)?(e.preventDefault(),this.$touchScrollingSub=t):this.$touchScrollingSub=null,this.menuScrollRefreshData(t),$.extend(s,{touchStartY:i.pageY,touchStartTime:e.timeStamp});else if(/move$/i.test(e.type)){var o=void 0!==s.touchY?s.touchY:s.touchStartY;if(void 0!==o&&o!=i.pageY){this.$touchScrollingSub=t;var a=i.pageY>o;void 0!==s.up&&s.up!=a&&$.extend(s,{touchStartY:i.pageY,touchStartTime:e.timeStamp}),$.extend(s,{up:a,touchY:i.pageY}),this.menuScroll(t,!0,Math.abs(i.pageY-o))}e.preventDefault()}else void 0!==s.touchY&&((s.momentum=15*Math.pow(Math.abs(i.pageY-s.touchStartY)/(e.timeStamp-s.touchStartTime),2))&&(this.menuScrollStop(t),this.menuScroll(t),e.preventDefault()),delete s.touchY)}}},menuShow:function(t){if((t.dataSM("beforefirstshowfired")||(t.dataSM("beforefirstshowfired",!0),this.$root.triggerHandler("beforefirstshow.smapi",t0)!==!1))&&this.$root.triggerHandler("beforeshow.smapi",t0)!==!1&&(t.dataSM("shown-before",!0),canAnimate&&t.stop(!0,!0),!t.is(":visible"))){var e=t.dataSM("parent-a"),i=this.isCollapsible();if((this.opts.keepHighlighted||i)&&e.addClass("highlighted"),i)t.removeClass("sm-nowrap").css({zIndex:"",width:"auto",minWidth:"",maxWidth:"",top:"",left:"",marginLeft:"",marginTop:""});else{if(t.css("z-index",this.zIndexInc=(this.zIndexInc||this.getStartZIndex())+1),(this.opts.subMenusMinWidth||this.opts.subMenusMaxWidth)&&(t.css({width:"auto",minWidth:"",maxWidth:""}).addClass("sm-nowrap"),this.opts.subMenusMinWidth&&t.css("min-width",this.opts.subMenusMinWidth),this.opts.subMenusMaxWidth)){var s=this.getWidth(t);t.css("max-width",this.opts.subMenusMaxWidth),s>this.getWidth(t)&&t.removeClass("sm-nowrap").css("width",this.opts.subMenusMaxWidth)}this.menuPosition(t)}var o=function(){t.css("overflow","")};i?canAnimate&&this.opts.collapsibleShowFunction?this.opts.collapsibleShowFunction.call(this,t,o):t.show(this.opts.collapsibleShowDuration,o):canAnimate&&this.opts.showFunction?this.opts.showFunction.call(this,t,o):t.show(this.opts.showDuration,o),e.attr("aria-expanded","true"),t.attr({"aria-expanded":"true","aria-hidden":"false"}),this.visibleSubMenus.push(t),this.$root.triggerHandler("show.smapi",t0)}},popupHide:function(t){this.hideTimeout&&(clearTimeout(this.hideTimeout),this.hideTimeout=0);var e=this;this.hideTimeout=setTimeout(function(){e.menuHideAll()},t?1:this.opts.hideTimeout)},popupShow:function(t,e){if(!this.opts.isPopup)return alert('SmartMenus jQuery Error:\n\nIf you want to show this menu via the "popupShow" method, set the isPopup:true option.'),void 0;if(this.hideTimeout&&(clearTimeout(this.hideTimeout),this.hideTimeout=0),this.$root.dataSM("shown-before",!0),canAnimate&&this.$root.stop(!0,!0),!this.$root.is(":visible")){this.$root.css({left:t,top:e});var i=this,s=function(){i.$root.css("overflow","")};canAnimate&&this.opts.showFunction?this.opts.showFunction.call(this,this.$root,s):this.$root.show(this.opts.showDuration,s),this.visibleSubMenus0=this.$root}},refresh:function(){this.destroy(!0),this.init(!0)},rootKeyDown:function(t){if(this.handleEvents())switch(t.keyCode){case 27:var e=this.activatedItems0;if(e){this.menuHideAll(),e0.focus();var i=e.dataSM("sub");i&&this.menuHide(i)}break;case 32:var s=$(t.target);if(s.is("a")&&this.handleItemEvents(s)){var i=s.dataSM("sub");i&&!i.is(":visible")&&(this.itemClick({currentTarget:t.target}),t.preventDefault())}}},rootOut:function(t){if(this.handleEvents()&&!this.isTouchMode()&&t.target!=this.$root0&&(this.hideTimeout&&(clearTimeout(this.hideTimeout),this.hideTimeout=0),!this.opts.showOnClick||!this.opts.hideOnClick)){var e=this;this.hideTimeout=setTimeout(function(){e.menuHideAll()},this.opts.hideTimeout)}},rootOver:function(t){this.handleEvents()&&!this.isTouchMode()&&t.target!=this.$root0&&this.hideTimeout&&(clearTimeout(this.hideTimeout),this.hideTimeout=0)},winResize:function(t){if(this.handleEvents()){if(!("onorientationchange"in window)||"orientationchange"==t.type){var e=this.isCollapsible();this.wasCollapsible&&e||(this.activatedItems.length&&this.activatedItemsthis.activatedItems.length-10.blur(),this.menuHideAll()),this.wasCollapsible=e}}else if(this.$disableOverlay){var i=this.$root.offset();this.$disableOverlay.css({top:i.top,left:i.left,width:this.$root.outerWidth(),height:this.$root.outerHeight()})}}}}),$.fn.dataSM=function(t,e){return e?this.data(t+"_smartmenus",e):this.data(t+"_smartmenus")},$.fn.removeDataSM=function(t){return this.removeData(t+"_smartmenus")},$.fn.smartmenus=function(options){if("string"==typeof options){var args=arguments,method=options;return Array.prototype.shift.call(args),this.each(function(){var t=$(this).data("smartmenus");t&&tmethod&&tmethod.apply(t,args)})}return this.each(function(){var dataOpts=$(this).data("sm-options")||null;if(dataOpts)try{dataOpts=eval("("+dataOpts+")")}catch(e){dataOpts=null,alert('ERROR\n\nSmartMenus jQuery init:\nInvalid "data-sm-options" attribute value syntax.')}new $.SmartMenus(this,$.extend({},$.fn.smartmenus.defaults,options,dataOpts))})},$.fn.smartmenus.defaults={isPopup:!1,mainMenuSubOffsetX:0,mainMenuSubOffsetY:0,subMenusSubOffsetX:0,subMenusSubOffsetY:0,subMenusMinWidth:"10em",subMenusMaxWidth:"20em",subIndicators:!0,subIndicatorsPos:"append",subIndicatorsText:"",scrollStep:30,scrollAccelerate:!0,showTimeout:250,hideTimeout:500,showDuration:0,showFunction:null,hideDuration:0,hideFunction:function(t,e){t.fadeOut(200,e)},collapsibleShowDuration:0,collapsibleShowFunction:function(t,e){t.slideDown(200,e)},collapsibleHideDuration:0,collapsibleHideFunction:function(t,e){t.slideUp(200,e)},showOnClick:!1,hideOnClick:!0,noMouseOver:!1,keepInViewport:!0,keepHighlighted:!0,markCurrentItem:!1,markCurrentTree:!0,rightToLeftSubMenus:!1,bottomToTopSubMenus:!1,collapsibleBehavior:"default"},$});
\ No newline at end of file

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/menu.js Added

@@ -0,0 +1,51 @@
+/*
+ @licstart  The following is the entire license notice for the JavaScript code in this file.
+
+ The MIT License (MIT)
+
+ Copyright (C) 1997-2020 by Dimitri van Heesch
+
+ Permission is hereby granted, free of charge, to any person obtaining a copy of this software
+ and associated documentation files (the "Software"), to deal in the Software without restriction,
+ including without limitation the rights to use, copy, modify, merge, publish, distribute,
+ sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
+ furnished to do so, subject to the following conditions:
+
+ The above copyright notice and this permission notice shall be included in all copies or
+ substantial portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING
+ BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+ DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ @licend  The above is the entire license notice for the JavaScript code in this file
+ */
+function initMenu(relPath,searchEnabled,serverSide,searchPage,search) {
+  function makeTree(data,relPath) {
+    var result='';
+    if ('children' in data) {
+      result+='<ul>';
+      for (var i in data.children) {
+        result+='<li><a href="'+relPath+data.childreni.url+'">'+
+                                data.childreni.text+'</a>'+
+                                makeTree(data.childreni,relPath)+'</li>';
+      }
+      result+='</ul>';
+    }
+    return result;
+  }
+
+  $('#main-nav').append(makeTree(menudata,relPath));
+  $('#main-nav').children(':first').addClass('sm sm-dox').attr('id','main-menu');
+  if (searchEnabled) {
+    if (serverSide) {
+      $('#main-menu').append('<li style="float:right"><div id="MSearchBox" class="MSearchBoxInactive"><div class="left"><form id="FSearchBox" action="'+relPath+searchPage+'" method="get"><img id="MSearchSelect" src="'+relPath+'search/mag.svg" alt=""/><input type="text" id="MSearchField" name="query" value="'+search+'" size="20" accesskey="S" onfocus="searchBox.OnSearchFieldFocus(true)" onblur="searchBox.OnSearchFieldFocus(false)"></form></div><div class="right"></div></div></li>');
+    } else {
+      $('#main-menu').append('<li style="float:right"><div id="MSearchBox" class="MSearchBoxInactive"><span class="left"><img id="MSearchSelect" src="'+relPath+'search/mag_sel.svg" onmouseover="return searchBox.OnSearchSelectShow()" onmouseout="return searchBox.OnSearchSelectHide()" alt=""/><input type="text" id="MSearchField" value="'+search+'" accesskey="S" onfocus="searchBox.OnSearchFieldFocus(true)" onblur="searchBox.OnSearchFieldFocus(false)" onkeyup="searchBox.OnSearchFieldChange(event)"/></span><span class="right"><a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="'+relPath+'search/close.svg" alt=""/></a></span></div></li>');
+    }
+  }
+  $('#main-menu').smartmenus();
+}
+/* @license-end */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/menudata.js Added

@@ -0,0 +1,33 @@
+/*
+ @licstart  The following is the entire license notice for the JavaScript code in this file.
+
+ The MIT License (MIT)
+
+ Copyright (C) 1997-2020 by Dimitri van Heesch
+
+ Permission is hereby granted, free of charge, to any person obtaining a copy of this software
+ and associated documentation files (the "Software"), to deal in the Software without restriction,
+ including without limitation the rights to use, copy, modify, merge, publish, distribute,
+ sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
+ furnished to do so, subject to the following conditions:
+
+ The above copyright notice and this permission notice shall be included in all copies or
+ substantial portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING
+ BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+ DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ @licend  The above is the entire license notice for the JavaScript code in this file
+*/
+var menudata={children:
+{text:"Main Page",url:"index.html"},
+{text:"Modules",url:"modules.html"},
+{text:"Data Structures",url:"annotated.html",children:
+{text:"Data Structures",url:"annotated.html"},
+{text:"Data Structure Index",url:"classes.html"},
+{text:"Data Fields",url:"functions.html",children:
+{text:"All",url:"functions.html"},
+{text:"Variables",url:"functions_vars.html"}}}}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/modules.html Added

@@ -0,0 +1,81 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: Modules</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+</div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+<div class="header">
+  <div class="headertitle">
+<div class="title">Modules</div>  </div>
+</div>
+<div class="contents">
+<div class="textblock">Here is a list of all modules:</div><div class="directory">
+<table class="directory">
+<tr id="row_0_" class="even"><td class="entry"><span style="width:16px;display:inline-block;">&#160;</span><a class="el" href="group___turbo_j_p_e_g.html" target="_self">TurboJPEG</a></td><td class="desc">TurboJPEG API </td></tr>
+</table>
+</div>
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/nav_f.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/nav_g.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/nav_h.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/open.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_0.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_0.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_0.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_1.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_1.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_1.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_2.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_2.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_2.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_3.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_3.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_3.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_4.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_4.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_4.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_5.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_5.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_5.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_6.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_6.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_6.js Added

@@ -0,0 +1,103 @@
+var searchData=
+
+  'tj_5fnumcs_8','TJ_NUMCS','../group___turbo_j_p_e_g.html#ga39f57a6fb02d9cf32e7b6890099b5a71',1,'turbojpeg.h',
+  'tj_5fnumerr_9','TJ_NUMERR','../group___turbo_j_p_e_g.html#ga79bde1b4a3e2351e00887e47781b966e',1,'turbojpeg.h',
+  'tj_5fnumpf_10','TJ_NUMPF','../group___turbo_j_p_e_g.html#ga7010a4402f54a45ba822ad8675a4655e',1,'turbojpeg.h',
+  'tj_5fnumsamp_11','TJ_NUMSAMP','../group___turbo_j_p_e_g.html#ga5ef3d169162ce77ce348e292a0b7477c',1,'turbojpeg.h',
+  'tj_5fnumxop_12','TJ_NUMXOP','../group___turbo_j_p_e_g.html#ga0f6dbd18adf38b7d46ac547f0f4d562c',1,'turbojpeg.h',
+  'tjalloc_13','tjAlloc','../group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83',1,'turbojpeg.h',
+  'tjalphaoffset_14','tjAlphaOffset','../group___turbo_j_p_e_g.html#ga5af0ab065feefd526debf1e20c43e837',1,'turbojpeg.h',
+  'tjblueoffset_15','tjBlueOffset','../group___turbo_j_p_e_g.html#ga84e2e35d3f08025f976ec1ec53693dea',1,'turbojpeg.h',
+  'tjbufsize_16','tjBufSize','../group___turbo_j_p_e_g.html#ga67ac12fee79073242cb216e07c9f1f90',1,'turbojpeg.h',
+  'tjbufsizeyuv2_17','tjBufSizeYUV2','../group___turbo_j_p_e_g.html#ga2be2b9969d4df9ecce9b05deed273194',1,'turbojpeg.h',
+  'tjcompress2_18','tjCompress2','../group___turbo_j_p_e_g.html#gafbdce0112fd78fd38efae841443a9bcf',1,'turbojpeg.h',
+  'tjcompressfromyuv_19','tjCompressFromYUV','../group___turbo_j_p_e_g.html#ga7622a459b79aa1007e005b58783f875b',1,'turbojpeg.h',
+  'tjcompressfromyuvplanes_20','tjCompressFromYUVPlanes','../group___turbo_j_p_e_g.html#ga29ec5dfbd2d84b8724e951d6fa0d5d9e',1,'turbojpeg.h',
+  'tjcs_21','TJCS','../group___turbo_j_p_e_g.html#ga4f83ad3368e0e29d1957be0efa7c3720',1,'turbojpeg.h',
+  'tjcs_5fcmyk_22','TJCS_CMYK','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a6c8b636152ac8195b869587db315ee53',1,'turbojpeg.h',
+  'tjcs_5fgray_23','TJCS_GRAY','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720ab3e7d6a87f695e45b81c1b5262b5a50a',1,'turbojpeg.h',
+  'tjcs_5frgb_24','TJCS_RGB','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a677cb7ccb85c4038ac41964a2e09e555',1,'turbojpeg.h',
+  'tjcs_5fycbcr_25','TJCS_YCbCr','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a7389b8f65bb387ffedce3efd0d78ec75',1,'turbojpeg.h',
+  'tjcs_5fycck_26','TJCS_YCCK','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a53839e0fe867b76b58d16b0a1a7c598e',1,'turbojpeg.h',
+  'tjdecodeyuv_27','tjDecodeYUV','../group___turbo_j_p_e_g.html#ga70abbf38f77a26fd6da8813bef96f695',1,'turbojpeg.h',
+  'tjdecodeyuvplanes_28','tjDecodeYUVPlanes','../group___turbo_j_p_e_g.html#ga10e837c07fa9d25770565b237d3898d9',1,'turbojpeg.h',
+  'tjdecompress2_29','tjDecompress2','../group___turbo_j_p_e_g.html#gae9eccef8b682a48f43a9117c231ed013',1,'turbojpeg.h',
+  'tjdecompressheader3_30','tjDecompressHeader3','../group___turbo_j_p_e_g.html#ga0595681096bba7199cc6f3533cb25f77',1,'turbojpeg.h',
+  'tjdecompresstoyuv2_31','tjDecompressToYUV2','../group___turbo_j_p_e_g.html#ga04d1e839ff9a0860dd1475cff78d3364',1,'turbojpeg.h',
+  'tjdecompresstoyuvplanes_32','tjDecompressToYUVPlanes','../group___turbo_j_p_e_g.html#gaa59f901a5258ada5bd0185ad59368540',1,'turbojpeg.h',
+  'tjdestroy_33','tjDestroy','../group___turbo_j_p_e_g.html#ga75f355fa27225ba1a4ee392c852394d2',1,'turbojpeg.h',
+  'tjencodeyuv3_34','tjEncodeYUV3','../group___turbo_j_p_e_g.html#gac519b922cdf446e97d0cdcba513636bf',1,'turbojpeg.h',
+  'tjencodeyuvplanes_35','tjEncodeYUVPlanes','../group___turbo_j_p_e_g.html#gae2d04c72457fe7f4d60cf78ab1b1feb1',1,'turbojpeg.h',
+  'tjerr_36','TJERR','../group___turbo_j_p_e_g.html#gafbc17cfa57d0d5d11fea35ac025950fe',1,'turbojpeg.h',
+  'tjerr_5ffatal_37','TJERR_FATAL','../group___turbo_j_p_e_g.html#ggafbc17cfa57d0d5d11fea35ac025950feafc9cceeada13122b09e4851e3788039a',1,'turbojpeg.h',
+  'tjerr_5fwarning_38','TJERR_WARNING','../group___turbo_j_p_e_g.html#ggafbc17cfa57d0d5d11fea35ac025950fea342dd6e2aedb47bb257b4e7568329b59',1,'turbojpeg.h',
+  'tjflag_5faccuratedct_39','TJFLAG_ACCURATEDCT','../group___turbo_j_p_e_g.html#gacb233cfd722d66d1ccbf48a7de81f0e0',1,'turbojpeg.h',
+  'tjflag_5fbottomup_40','TJFLAG_BOTTOMUP','../group___turbo_j_p_e_g.html#ga72ecf4ebe6eb702d3c6f5ca27455e1ec',1,'turbojpeg.h',
+  'tjflag_5ffastdct_41','TJFLAG_FASTDCT','../group___turbo_j_p_e_g.html#gaabce235db80d3f698b27f36cbd453da2',1,'turbojpeg.h',
+  'tjflag_5ffastupsample_42','TJFLAG_FASTUPSAMPLE','../group___turbo_j_p_e_g.html#ga4ee4506c81177a06f77e2504a22efd2d',1,'turbojpeg.h',
+  'tjflag_5flimitscans_43','TJFLAG_LIMITSCANS','../group___turbo_j_p_e_g.html#ga163e6482dc5096831feef9c79ff3f805',1,'turbojpeg.h',
+  'tjflag_5fnorealloc_44','TJFLAG_NOREALLOC','../group___turbo_j_p_e_g.html#ga8808d403c68b62aaa58a4c1e58e98963',1,'turbojpeg.h',
+  'tjflag_5fprogressive_45','TJFLAG_PROGRESSIVE','../group___turbo_j_p_e_g.html#ga43b426750b46190a25d34a67ef76df1b',1,'turbojpeg.h',
+  'tjflag_5fstoponwarning_46','TJFLAG_STOPONWARNING','../group___turbo_j_p_e_g.html#ga519cfa4ef6c18d9e5b455fdf59306a3a',1,'turbojpeg.h',
+  'tjfree_47','tjFree','../group___turbo_j_p_e_g.html#gaea863d2da0cdb609563aabdf9196514b',1,'turbojpeg.h',
+  'tjgeterrorcode_48','tjGetErrorCode','../group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410',1,'turbojpeg.h',
+  'tjgeterrorstr2_49','tjGetErrorStr2','../group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa',1,'turbojpeg.h',
+  'tjgetscalingfactors_50','tjGetScalingFactors','../group___turbo_j_p_e_g.html#gac3854476006b10787bd128f7ede48057',1,'turbojpeg.h',
+  'tjgreenoffset_51','tjGreenOffset','../group___turbo_j_p_e_g.html#ga82d6e35da441112a411da41923c0ba2f',1,'turbojpeg.h',
+  'tjhandle_52','tjhandle','../group___turbo_j_p_e_g.html#ga758d2634ecb4949de7815cba621f5763',1,'turbojpeg.h',
+  'tjinitcompress_53','tjInitCompress','../group___turbo_j_p_e_g.html#ga9d63a05fc6d813f4aae06107041a37e8',1,'turbojpeg.h',
+  'tjinitdecompress_54','tjInitDecompress','../group___turbo_j_p_e_g.html#ga52300eac3f3d9ef4bab303bc244f62d3',1,'turbojpeg.h',
+  'tjinittransform_55','tjInitTransform','../group___turbo_j_p_e_g.html#ga928beff6ac248ceadf01089fc6b41957',1,'turbojpeg.h',
+  'tjloadimage_56','tjLoadImage','../group___turbo_j_p_e_g.html#gaffbd83c375e79f5db4b5c5d8ad4466e7',1,'turbojpeg.h',
+  'tjmcuheight_57','tjMCUHeight','../group___turbo_j_p_e_g.html#gabd247bb9fecb393eca57366feb8327bf',1,'turbojpeg.h',
+  'tjmcuwidth_58','tjMCUWidth','../group___turbo_j_p_e_g.html#ga9e61e7cd47a15a173283ba94e781308c',1,'turbojpeg.h',
+  'tjpad_59','TJPAD','../group___turbo_j_p_e_g.html#ga0aba955473315e405295d978f0c16511',1,'turbojpeg.h',
+  'tjpf_60','TJPF','../group___turbo_j_p_e_g.html#gac916144e26c3817ac514e64ae5d12e2a',1,'turbojpeg.h',
+  'tjpf_5fabgr_61','TJPF_ABGR','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa1ba1a7f1631dbeaa49a0a85fc4a40081',1,'turbojpeg.h',
+  'tjpf_5fargb_62','TJPF_ARGB','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aae8f846ed9d9de99b6e1dfe448848765c',1,'turbojpeg.h',
+  'tjpf_5fbgr_63','TJPF_BGR','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aab10624437fb8ef495a0b153e65749839',1,'turbojpeg.h',
+  'tjpf_5fbgra_64','TJPF_BGRA','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aac037ff1845cf9b74bb81a3659c2b9fb4',1,'turbojpeg.h',
+  'tjpf_5fbgrx_65','TJPF_BGRX','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa2a1fbf569ca79897eae886e3376ca4c8',1,'turbojpeg.h',
+  'tjpf_5fcmyk_66','TJPF_CMYK','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa7f5100ec44c91994e243f1cf55553f8b',1,'turbojpeg.h',
+  'tjpf_5fgray_67','TJPF_GRAY','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa5431b54b015337705f13118073711a1a',1,'turbojpeg.h',
+  'tjpf_5frgb_68','TJPF_RGB','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa7ce93230bff449518ce387c17e6ed37c',1,'turbojpeg.h',
+  'tjpf_5frgba_69','TJPF_RGBA','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa88d2e88fab67f6503cf972e14851cc12',1,'turbojpeg.h',
+  'tjpf_5frgbx_70','TJPF_RGBX','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa83973bebb7e2dc6fa8bae89ff3f42e01',1,'turbojpeg.h',
+  'tjpf_5funknown_71','TJPF_UNKNOWN','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa84c1a6cead7952998e2fb895844a21ed',1,'turbojpeg.h',
+  'tjpf_5fxbgr_72','TJPF_XBGR','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aaf6603b27147de47e212e75dac027b2af',1,'turbojpeg.h',
+  'tjpf_5fxrgb_73','TJPF_XRGB','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aadae996905efcfa3b42a0bb3bea7f9d84',1,'turbojpeg.h',
+  'tjpixelsize_74','tjPixelSize','../group___turbo_j_p_e_g.html#gad77cf8fe5b2bfd3cb3f53098146abb4c',1,'turbojpeg.h',
+  'tjplaneheight_75','tjPlaneHeight','../group___turbo_j_p_e_g.html#ga1a209696c6a80748f20e134b3c64789f',1,'turbojpeg.h',
+  'tjplanesizeyuv_76','tjPlaneSizeYUV','../group___turbo_j_p_e_g.html#gab4ab7b24f6e797d79abaaa670373961d',1,'turbojpeg.h',
+  'tjplanewidth_77','tjPlaneWidth','../group___turbo_j_p_e_g.html#ga63fb66bb1e36c74008c4634360becbb1',1,'turbojpeg.h',
+  'tjredoffset_78','tjRedOffset','../group___turbo_j_p_e_g.html#gadd9b446742ac8a3923f7992c7988fea8',1,'turbojpeg.h',
+  'tjregion_79','tjregion','../structtjregion.html',1,'',
+  'tjsamp_80','TJSAMP','../group___turbo_j_p_e_g.html#ga1d047060ea80bb9820d540bb928e9074',1,'turbojpeg.h',
+  'tjsamp_5f411_81','TJSAMP_411','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a28ec62575e5ea295c3fde3001dc628e2',1,'turbojpeg.h',
+  'tjsamp_5f420_82','TJSAMP_420','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a63085dbf683cfe39e513cdb6343e3737',1,'turbojpeg.h',
+  'tjsamp_5f422_83','TJSAMP_422','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a136130902cc578f11f32429b59368404',1,'turbojpeg.h',
+  'tjsamp_5f440_84','TJSAMP_440','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074accf740e6f3aa6ba20ba922cad13cb974',1,'turbojpeg.h',
+  'tjsamp_5f444_85','TJSAMP_444','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074afb8da4f44197837bdec0a4f593dacae3',1,'turbojpeg.h',
+  'tjsamp_5fgray_86','TJSAMP_GRAY','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a3f1c9504842ddc7a48d0f690754b6248',1,'turbojpeg.h',
+  'tjsaveimage_87','tjSaveImage','../group___turbo_j_p_e_g.html#ga6f445b22d8933ae4815b3370a538d879',1,'turbojpeg.h',
+  'tjscaled_88','TJSCALED','../group___turbo_j_p_e_g.html#ga84878bb65404204743aa18cac02781df',1,'turbojpeg.h',
+  'tjscalingfactor_89','tjscalingfactor','../structtjscalingfactor.html',1,'',
+  'tjtransform_90','tjtransform','../structtjtransform.html',1,'tjtransform','../group___turbo_j_p_e_g.html#ga504805ec0161f1b505397ca0118bf8fd',1,'tjtransform():&#160;turbojpeg.h','../group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25',1,'tjTransform(tjhandle handle, const unsigned char *jpegBuf, unsigned long jpegSize, int n, unsigned char **dstBufs, unsigned long *dstSizes, tjtransform *transforms, int flags):&#160;turbojpeg.h',
+  'tjxop_91','TJXOP','../group___turbo_j_p_e_g.html#ga2de531af4e7e6c4f124908376b354866',1,'turbojpeg.h',
+  'tjxop_5fhflip_92','TJXOP_HFLIP','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866aa0df69776caa30f0fa28e26332d311ce',1,'turbojpeg.h',
+  'tjxop_5fnone_93','TJXOP_NONE','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866aad88c0366cd3f7d0eac9d7a3fa1c2c27',1,'turbojpeg.h',
+  'tjxop_5frot180_94','TJXOP_ROT180','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a140952eb8dd0300accfcc22726d69692',1,'turbojpeg.h',
+  'tjxop_5frot270_95','TJXOP_ROT270','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a3064ee5dfb7f032df332818587567a08',1,'turbojpeg.h',
+  'tjxop_5frot90_96','TJXOP_ROT90','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a43b2bbb23bc4bd548422d43fbe9af128',1,'turbojpeg.h',
+  'tjxop_5ftranspose_97','TJXOP_TRANSPOSE','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a31060aed199f886afdd417f80499c32d',1,'turbojpeg.h',
+  'tjxop_5ftransverse_98','TJXOP_TRANSVERSE','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866af3b14d488aea6ece9e5b3df73a74d6a4',1,'turbojpeg.h',
+  'tjxop_5fvflip_99','TJXOP_VFLIP','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a324eddfbec53b7e691f61e56929d0d5d',1,'turbojpeg.h',
+  'tjxopt_5fcopynone_100','TJXOPT_COPYNONE','../group___turbo_j_p_e_g.html#ga153b468cfb905d0de61706c838986fe8',1,'turbojpeg.h',
+  'tjxopt_5fcrop_101','TJXOPT_CROP','../group___turbo_j_p_e_g.html#ga9c771a757fc1294add611906b89ab2d2',1,'turbojpeg.h',
+  'tjxopt_5fgray_102','TJXOPT_GRAY','../group___turbo_j_p_e_g.html#ga3acee7b48ade1b99e5588736007c2589',1,'turbojpeg.h',
+  'tjxopt_5fnooutput_103','TJXOPT_NOOUTPUT','../group___turbo_j_p_e_g.html#gafbf992bbf6e006705886333703ffab31',1,'turbojpeg.h',
+  'tjxopt_5fperfect_104','TJXOPT_PERFECT','../group___turbo_j_p_e_g.html#ga50e03cb5ed115330e212417429600b00',1,'turbojpeg.h',
+  'tjxopt_5fprogressive_105','TJXOPT_PROGRESSIVE','../group___turbo_j_p_e_g.html#gad2371c80674584ecc1a7d75e564cf026',1,'turbojpeg.h',
+  'tjxopt_5ftrim_106','TJXOPT_TRIM','../group___turbo_j_p_e_g.html#ga319826b7eb1583c0595bbe7b95428709',1,'turbojpeg.h',
+  'turbojpeg_107','TurboJPEG','../group___turbo_j_p_e_g.html',1,''
+;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_7.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_7.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_7.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_8.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_8.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_8.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_9.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="all_9.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/all_9.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/classes_0.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="classes_0.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/classes_0.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/close.svg Added

@@ -0,0 +1,31 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<svg
+   xmlns:dc="http://purl.org/dc/elements/1.1/"
+   xmlns:cc="http://creativecommons.org/ns#"
+   xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
+   xmlns:svg="http://www.w3.org/2000/svg"
+   xmlns="http://www.w3.org/2000/svg"
+   viewBox="0 0 11 11"
+   height="11"
+   width="11"
+   id="svg2"
+   version="1.1">
+  <metadata
+     id="metadata8">
+    <rdf:RDF>
+      <cc:Work
+         rdf:about="">
+        <dc:format>image/svg+xml</dc:format>
+        <dc:type
+           rdf:resource="http://purl.org/dc/dcmitype/StillImage" />
+        <dc:title></dc:title>
+      </cc:Work>
+    </rdf:RDF>
+  </metadata>
+  <defs
+     id="defs6" />
+  <path
+     id="path12"
+     d="M 5.5 0.5 A 5 5 0 0 0 0.5 5.5 A 5 5 0 0 0 5.5 10.5 A 5 5 0 0 0 10.5 5.5 A 5 5 0 0 0 5.5 0.5 z M 3.5820312 3 A 0.58291923 0.58291923 0 0 1 4 3.1757812 L 5.5 4.6757812 L 7 3.1757812 A 0.58291923 0.58291923 0 0 1 7.4003906 3 A 0.58291923 0.58291923 0 0 1 7.8242188 4 L 6.3242188 5.5 L 7.8242188 7 A 0.58291923 0.58291923 0 1 1 7 7.8242188 L 5.5 6.3242188 L 4 7.8242188 A 0.58291923 0.58291923 0 1 1 3.1757812 7 L 4.6757812 5.5 L 3.1757812 4 A 0.58291923 0.58291923 0 0 1 3.5820312 3 z "
+     style="stroke-width:1.09870648;fill:#bababa;fill-opacity:1" />
+</svg>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/enums_0.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="enums_0.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/enums_0.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/enumvalues_0.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="enumvalues_0.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/enumvalues_0.js Added

@@ -0,0 +1,37 @@
+var searchData=
+
+  'tjcs_5fcmyk_167','TJCS_CMYK','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a6c8b636152ac8195b869587db315ee53',1,'turbojpeg.h',
+  'tjcs_5fgray_168','TJCS_GRAY','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720ab3e7d6a87f695e45b81c1b5262b5a50a',1,'turbojpeg.h',
+  'tjcs_5frgb_169','TJCS_RGB','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a677cb7ccb85c4038ac41964a2e09e555',1,'turbojpeg.h',
+  'tjcs_5fycbcr_170','TJCS_YCbCr','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a7389b8f65bb387ffedce3efd0d78ec75',1,'turbojpeg.h',
+  'tjcs_5fycck_171','TJCS_YCCK','../group___turbo_j_p_e_g.html#gga4f83ad3368e0e29d1957be0efa7c3720a53839e0fe867b76b58d16b0a1a7c598e',1,'turbojpeg.h',
+  'tjerr_5ffatal_172','TJERR_FATAL','../group___turbo_j_p_e_g.html#ggafbc17cfa57d0d5d11fea35ac025950feafc9cceeada13122b09e4851e3788039a',1,'turbojpeg.h',
+  'tjerr_5fwarning_173','TJERR_WARNING','../group___turbo_j_p_e_g.html#ggafbc17cfa57d0d5d11fea35ac025950fea342dd6e2aedb47bb257b4e7568329b59',1,'turbojpeg.h',
+  'tjpf_5fabgr_174','TJPF_ABGR','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa1ba1a7f1631dbeaa49a0a85fc4a40081',1,'turbojpeg.h',
+  'tjpf_5fargb_175','TJPF_ARGB','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aae8f846ed9d9de99b6e1dfe448848765c',1,'turbojpeg.h',
+  'tjpf_5fbgr_176','TJPF_BGR','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aab10624437fb8ef495a0b153e65749839',1,'turbojpeg.h',
+  'tjpf_5fbgra_177','TJPF_BGRA','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aac037ff1845cf9b74bb81a3659c2b9fb4',1,'turbojpeg.h',
+  'tjpf_5fbgrx_178','TJPF_BGRX','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa2a1fbf569ca79897eae886e3376ca4c8',1,'turbojpeg.h',
+  'tjpf_5fcmyk_179','TJPF_CMYK','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa7f5100ec44c91994e243f1cf55553f8b',1,'turbojpeg.h',
+  'tjpf_5fgray_180','TJPF_GRAY','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa5431b54b015337705f13118073711a1a',1,'turbojpeg.h',
+  'tjpf_5frgb_181','TJPF_RGB','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa7ce93230bff449518ce387c17e6ed37c',1,'turbojpeg.h',
+  'tjpf_5frgba_182','TJPF_RGBA','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa88d2e88fab67f6503cf972e14851cc12',1,'turbojpeg.h',
+  'tjpf_5frgbx_183','TJPF_RGBX','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa83973bebb7e2dc6fa8bae89ff3f42e01',1,'turbojpeg.h',
+  'tjpf_5funknown_184','TJPF_UNKNOWN','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aa84c1a6cead7952998e2fb895844a21ed',1,'turbojpeg.h',
+  'tjpf_5fxbgr_185','TJPF_XBGR','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aaf6603b27147de47e212e75dac027b2af',1,'turbojpeg.h',
+  'tjpf_5fxrgb_186','TJPF_XRGB','../group___turbo_j_p_e_g.html#ggac916144e26c3817ac514e64ae5d12e2aadae996905efcfa3b42a0bb3bea7f9d84',1,'turbojpeg.h',
+  'tjsamp_5f411_187','TJSAMP_411','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a28ec62575e5ea295c3fde3001dc628e2',1,'turbojpeg.h',
+  'tjsamp_5f420_188','TJSAMP_420','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a63085dbf683cfe39e513cdb6343e3737',1,'turbojpeg.h',
+  'tjsamp_5f422_189','TJSAMP_422','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a136130902cc578f11f32429b59368404',1,'turbojpeg.h',
+  'tjsamp_5f440_190','TJSAMP_440','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074accf740e6f3aa6ba20ba922cad13cb974',1,'turbojpeg.h',
+  'tjsamp_5f444_191','TJSAMP_444','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074afb8da4f44197837bdec0a4f593dacae3',1,'turbojpeg.h',
+  'tjsamp_5fgray_192','TJSAMP_GRAY','../group___turbo_j_p_e_g.html#gga1d047060ea80bb9820d540bb928e9074a3f1c9504842ddc7a48d0f690754b6248',1,'turbojpeg.h',
+  'tjxop_5fhflip_193','TJXOP_HFLIP','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866aa0df69776caa30f0fa28e26332d311ce',1,'turbojpeg.h',
+  'tjxop_5fnone_194','TJXOP_NONE','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866aad88c0366cd3f7d0eac9d7a3fa1c2c27',1,'turbojpeg.h',
+  'tjxop_5frot180_195','TJXOP_ROT180','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a140952eb8dd0300accfcc22726d69692',1,'turbojpeg.h',
+  'tjxop_5frot270_196','TJXOP_ROT270','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a3064ee5dfb7f032df332818587567a08',1,'turbojpeg.h',
+  'tjxop_5frot90_197','TJXOP_ROT90','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a43b2bbb23bc4bd548422d43fbe9af128',1,'turbojpeg.h',
+  'tjxop_5ftranspose_198','TJXOP_TRANSPOSE','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a31060aed199f886afdd417f80499c32d',1,'turbojpeg.h',
+  'tjxop_5ftransverse_199','TJXOP_TRANSVERSE','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866af3b14d488aea6ece9e5b3df73a74d6a4',1,'turbojpeg.h',
+  'tjxop_5fvflip_200','TJXOP_VFLIP','../group___turbo_j_p_e_g.html#gga2de531af4e7e6c4f124908376b354866a324eddfbec53b7e691f61e56929d0d5d',1,'turbojpeg.h'
+;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/functions_0.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="functions_0.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/functions_0.js Added

@@ -0,0 +1,31 @@
+var searchData=
+
+  'tjalloc_114','tjAlloc','../group___turbo_j_p_e_g.html#gaec627dd4c5f30b7a775a7aea3bec5d83',1,'turbojpeg.h',
+  'tjbufsize_115','tjBufSize','../group___turbo_j_p_e_g.html#ga67ac12fee79073242cb216e07c9f1f90',1,'turbojpeg.h',
+  'tjbufsizeyuv2_116','tjBufSizeYUV2','../group___turbo_j_p_e_g.html#ga2be2b9969d4df9ecce9b05deed273194',1,'turbojpeg.h',
+  'tjcompress2_117','tjCompress2','../group___turbo_j_p_e_g.html#gafbdce0112fd78fd38efae841443a9bcf',1,'turbojpeg.h',
+  'tjcompressfromyuv_118','tjCompressFromYUV','../group___turbo_j_p_e_g.html#ga7622a459b79aa1007e005b58783f875b',1,'turbojpeg.h',
+  'tjcompressfromyuvplanes_119','tjCompressFromYUVPlanes','../group___turbo_j_p_e_g.html#ga29ec5dfbd2d84b8724e951d6fa0d5d9e',1,'turbojpeg.h',
+  'tjdecodeyuv_120','tjDecodeYUV','../group___turbo_j_p_e_g.html#ga70abbf38f77a26fd6da8813bef96f695',1,'turbojpeg.h',
+  'tjdecodeyuvplanes_121','tjDecodeYUVPlanes','../group___turbo_j_p_e_g.html#ga10e837c07fa9d25770565b237d3898d9',1,'turbojpeg.h',
+  'tjdecompress2_122','tjDecompress2','../group___turbo_j_p_e_g.html#gae9eccef8b682a48f43a9117c231ed013',1,'turbojpeg.h',
+  'tjdecompressheader3_123','tjDecompressHeader3','../group___turbo_j_p_e_g.html#ga0595681096bba7199cc6f3533cb25f77',1,'turbojpeg.h',
+  'tjdecompresstoyuv2_124','tjDecompressToYUV2','../group___turbo_j_p_e_g.html#ga04d1e839ff9a0860dd1475cff78d3364',1,'turbojpeg.h',
+  'tjdecompresstoyuvplanes_125','tjDecompressToYUVPlanes','../group___turbo_j_p_e_g.html#gaa59f901a5258ada5bd0185ad59368540',1,'turbojpeg.h',
+  'tjdestroy_126','tjDestroy','../group___turbo_j_p_e_g.html#ga75f355fa27225ba1a4ee392c852394d2',1,'turbojpeg.h',
+  'tjencodeyuv3_127','tjEncodeYUV3','../group___turbo_j_p_e_g.html#gac519b922cdf446e97d0cdcba513636bf',1,'turbojpeg.h',
+  'tjencodeyuvplanes_128','tjEncodeYUVPlanes','../group___turbo_j_p_e_g.html#gae2d04c72457fe7f4d60cf78ab1b1feb1',1,'turbojpeg.h',
+  'tjfree_129','tjFree','../group___turbo_j_p_e_g.html#gaea863d2da0cdb609563aabdf9196514b',1,'turbojpeg.h',
+  'tjgeterrorcode_130','tjGetErrorCode','../group___turbo_j_p_e_g.html#ga414feeffbf860ebd31c745df203de410',1,'turbojpeg.h',
+  'tjgeterrorstr2_131','tjGetErrorStr2','../group___turbo_j_p_e_g.html#ga1ead8574f9f39fbafc6b497124e7aafa',1,'turbojpeg.h',
+  'tjgetscalingfactors_132','tjGetScalingFactors','../group___turbo_j_p_e_g.html#gac3854476006b10787bd128f7ede48057',1,'turbojpeg.h',
+  'tjinitcompress_133','tjInitCompress','../group___turbo_j_p_e_g.html#ga9d63a05fc6d813f4aae06107041a37e8',1,'turbojpeg.h',
+  'tjinitdecompress_134','tjInitDecompress','../group___turbo_j_p_e_g.html#ga52300eac3f3d9ef4bab303bc244f62d3',1,'turbojpeg.h',
+  'tjinittransform_135','tjInitTransform','../group___turbo_j_p_e_g.html#ga928beff6ac248ceadf01089fc6b41957',1,'turbojpeg.h',
+  'tjloadimage_136','tjLoadImage','../group___turbo_j_p_e_g.html#gaffbd83c375e79f5db4b5c5d8ad4466e7',1,'turbojpeg.h',
+  'tjplaneheight_137','tjPlaneHeight','../group___turbo_j_p_e_g.html#ga1a209696c6a80748f20e134b3c64789f',1,'turbojpeg.h',
+  'tjplanesizeyuv_138','tjPlaneSizeYUV','../group___turbo_j_p_e_g.html#gab4ab7b24f6e797d79abaaa670373961d',1,'turbojpeg.h',
+  'tjplanewidth_139','tjPlaneWidth','../group___turbo_j_p_e_g.html#ga63fb66bb1e36c74008c4634360becbb1',1,'turbojpeg.h',
+  'tjsaveimage_140','tjSaveImage','../group___turbo_j_p_e_g.html#ga6f445b22d8933ae4815b3370a538d879',1,'turbojpeg.h',
+  'tjtransform_141','tjTransform','../group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25',1,'turbojpeg.h'
+;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/groups_0.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="groups_0.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/groups_0.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/mag_sel.svg Added

@@ -0,0 +1,74 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+
+
+<svg
+   xmlns:dc="http://purl.org/dc/elements/1.1/"
+   xmlns:cc="http://creativecommons.org/ns#"
+   xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
+   xmlns:svg="http://www.w3.org/2000/svg"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
+   xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
+   version="1.1"
+   id="svg2"
+   width="20"
+   height="19"
+   viewBox="0 0 20 19"
+   sodipodi:docname="mag_sel.svg"
+   inkscape:version="0.92.5 (2060ec1f9f, 2020-04-08)">
+  <metadata
+     id="metadata8">
+    <rdf:RDF>
+      <cc:Work
+         rdf:about="">
+        <dc:format>image/svg+xml</dc:format>
+        <dc:type
+           rdf:resource="http://purl.org/dc/dcmitype/StillImage" />
+        <dc:title></dc:title>
+      </cc:Work>
+    </rdf:RDF>
+  </metadata>
+  <defs
+     id="defs6" />
+  <sodipodi:namedview
+     pagecolor="#ffffff"
+     bordercolor="#666666"
+     borderopacity="1"
+     objecttolerance="10"
+     gridtolerance="10"
+     guidetolerance="10"
+     inkscape:pageopacity="0"
+     inkscape:pageshadow="2"
+     inkscape:window-width="1920"
+     inkscape:window-height="2096"
+     id="namedview4"
+     showgrid="false"
+     fit-margin-top="0"
+     fit-margin-left="0"
+     fit-margin-right="0"
+     fit-margin-bottom="0"
+     inkscape:zoom="32"
+     inkscape:cx="5.9792688"
+     inkscape:cy="1.1436277"
+     inkscape:window-x="1920"
+     inkscape:window-y="27"
+     inkscape:window-maximized="0"
+     inkscape:current-layer="svg2" />
+  <circle
+     style="fill:#000000;fill-opacity:0;stroke:#656565;stroke-width:1.4;stroke-opacity:1;stroke-miterlimit:4;stroke-dasharray:none"
+     id="path4611"
+     cx="5.5"
+     cy="8.5"
+     r="3.5" />
+  <path
+     style="fill:#656565;fill-opacity:1;stroke:none;stroke-width:1px;stroke-linecap:butt;stroke-linejoin:miter;stroke-opacity:1"
+     d="M 11,7 13.5,10 16,7 Z"
+     id="path4609"
+     inkscape:connector-curvature="0"
+     sodipodi:nodetypes="cccc" />
+  <path
+     style="fill:none;stroke:#656565;stroke-width:1.4;stroke-linecap:round;stroke-linejoin:miter;stroke-miterlimit:4;stroke-dasharray:none;stroke-opacity:1"
+     d="m 8.1085854,11.109059 2.7823556,2.782356"
+     id="path4630"
+     inkscape:connector-curvature="0" />
+</svg>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/nomatches.html Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/search.css Added

@@ -0,0 +1,255 @@
+/*---------------- Search Box */
+
+#MSearchBox {
+    white-space : nowrap;
+    background: white;
+    border-radius: 0.65em;
+    box-shadow: inset 0.5px 0.5px 3px 0px #555;
+    z-index: 102;
+}
+
+#MSearchBox .left {
+    display: inline-block;
+    vertical-align: middle;
+    height: 1.4em;
+}
+
+#MSearchSelect {
+    display: inline-block;
+    vertical-align: middle;
+    height: 1.4em;
+    padding: 0 0 0 0.3em;
+    margin: 0;
+}
+
+#MSearchField {
+    display: inline-block;
+    vertical-align: middle;
+    width: 7.5em;
+    height: 1.1em;
+    margin: 0 0.15em;
+    padding: 0;
+    line-height: 1em;
+    border:none;
+    color: #909090;
+    outline: none;
+    font-family: Arial, Verdana, sans-serif;
+    -webkit-border-radius: 0px;
+    border-radius: 0px;
+    background: none;
+}
+
+
+#MSearchBox .right {
+    display: inline-block;
+    vertical-align: middle;
+    width: 1.4em;
+    height: 1.4em;
+}
+
+#MSearchClose {
+    display: none;
+    font-size: inherit;
+    background : none;
+    border: none;
+    margin: 0;
+    padding: 0;
+    outline: none;
+
+}
+
+#MSearchCloseImg {
+    height: 1.4em;
+    padding: 0.3em;
+    margin: 0;
+}
+
+.MSearchBoxActive #MSearchField {
+    color: #000000;
+}
+
+#main-menu > li:last-child {
+    /* This <li> object is the parent of the search bar */
+    display: flex;
+    justify-content: center;
+    align-items: center;
+    height: 36px;
+    margin-right: 1em;
+}
+
+/*---------------- Search filter selection */
+
+#MSearchSelectWindow {
+    display: none;
+    position: absolute;
+    left: 0; top: 0;
+    border: 1px solid #90A5CE;
+    background-color: #F9FAFC;
+    z-index: 10001;
+    padding-top: 4px;
+    padding-bottom: 4px;
+    -moz-border-radius: 4px;
+    -webkit-border-top-left-radius: 4px;
+    -webkit-border-top-right-radius: 4px;
+    -webkit-border-bottom-left-radius: 4px;
+    -webkit-border-bottom-right-radius: 4px;
+    -webkit-box-shadow: 5px 5px 5px rgba(0, 0, 0, 0.15);
+}
+
+.SelectItem {
+    font: 8pt Arial, Verdana, sans-serif;
+    padding-left:  2px;
+    padding-right: 12px;
+    border: 0px;
+}
+
+span.SelectionMark {
+    margin-right: 4px;
+    font-family: monospace;
+    outline-style: none;
+    text-decoration: none;
+}
+
+a.SelectItem {
+    display: block;
+    outline-style: none;
+    color: #000000; 
+    text-decoration: none;
+    padding-left:   6px;
+    padding-right: 12px;
+}
+
+a.SelectItem:focus,
+a.SelectItem:active {
+    color: #000000; 
+    outline-style: none;
+    text-decoration: none;
+}
+
+a.SelectItem:hover {
+    color: #FFFFFF;
+    background-color: #3D578C;
+    outline-style: none;
+    text-decoration: none;
+    cursor: pointer;
+    display: block;
+}
+
+/*---------------- Search results window */
+
+iframe#MSearchResults {
+    width: 60ex;
+    height: 15em;
+}
+
+#MSearchResultsWindow {
+    display: none;
+    position: absolute;
+    left: 0; top: 0;
+    border: 1px solid #000;
+    background-color: #EEF1F7;
+    z-index:10000;
+}
+
+/* ----------------------------------- */
+
+
+#SRIndex {
+    clear:both; 
+    padding-bottom: 15px;
+}
+
+.SREntry {
+    font-size: 10pt;
+    padding-left: 1ex;
+}
+
+.SRPage .SREntry {
+    font-size: 8pt;
+    padding: 1px 5px;
+}
+
+body.SRPage {
+    margin: 5px 2px;
+}
+
+.SRChildren {
+    padding-left: 3ex; padding-bottom: .5em 
+}
+
+.SRPage .SRChildren {
+    display: none;
+}
+
+.SRSymbol {
+    font-weight: bold; 
+    color: #425E97;
+    font-family: Arial, Verdana, sans-serif;
+    text-decoration: none;
+    outline: none;
+}
+
+a.SRScope {
+    display: block;
+    color: #425E97; 
+    font-family: Arial, Verdana, sans-serif;
+    text-decoration: none;
+    outline: none;
+}
+
+a.SRSymbol:focus, a.SRSymbol:active,
+a.SRScope:focus, a.SRScope:active {
+    text-decoration: underline;
+}
+
+span.SRScope {
+    padding-left: 4px;
+}
+
+.SRPage .SRStatus {
+    padding: 2px 5px;
+    font-size: 8pt;
+    font-style: italic;
+}
+
+.SRResult {
+    display: none;
+}
+
+DIV.searchresults {
+    margin-left: 10px;
+    margin-right: 10px;
+}
+
+/*---------------- External search page results */
+
+.searchresult {
+    background-color: #F0F3F8;
+}
+
+.pages b {
+   color: white;
+   padding: 5px 5px 3px 5px;
+   background-image: url("../tab_a.png");
+   background-repeat: repeat-x;
+   text-shadow: 0 1px 1px #000000;
+}
+
+.pages {
+    line-height: 17px;
+    margin-left: 4px;
+    text-decoration: none;
+}
+
+.hl {
+    font-weight: bold;
+}
+
+#searchresults {
+    margin-bottom: 20px;
+}
+
+.searchpages {
+    margin-top: 10px;
+}
+

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/search.js Added

@@ -0,0 +1,814 @@
+/*
+ @licstart  The following is the entire license notice for the JavaScript code in this file.
+
+ The MIT License (MIT)
+
+ Copyright (C) 1997-2020 by Dimitri van Heesch
+
+ Permission is hereby granted, free of charge, to any person obtaining a copy of this software
+ and associated documentation files (the "Software"), to deal in the Software without restriction,
+ including without limitation the rights to use, copy, modify, merge, publish, distribute,
+ sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
+ furnished to do so, subject to the following conditions:
+
+ The above copyright notice and this permission notice shall be included in all copies or
+ substantial portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING
+ BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+ DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ @licend  The above is the entire license notice for the JavaScript code in this file
+ */
+function convertToId(search)
+{
+  var result = '';
+  for (i=0;i<search.length;i++)
+  {
+    var c = search.charAt(i);
+    var cn = c.charCodeAt(0);
+    if (c.match(/a-z0-9\u0080-\uFFFF/))
+    {
+      result+=c;
+    }
+    else if (cn<16)
+    {
+      result+="_0"+cn.toString(16);
+    }
+    else
+    {
+      result+="_"+cn.toString(16);
+    }
+  }
+  return result;
+}
+
+function getXPos(item)
+{
+  var x = 0;
+  if (item.offsetWidth)
+  {
+    while (item && item!=document.body)
+    {
+      x   += item.offsetLeft;
+      item = item.offsetParent;
+    }
+  }
+  return x;
+}
+
+function getYPos(item)
+{
+  var y = 0;
+  if (item.offsetWidth)
+  {
+     while (item && item!=document.body)
+     {
+       y   += item.offsetTop;
+       item = item.offsetParent;
+     }
+  }
+  return y;
+}
+
+/* A class handling everything associated with the search panel.
+
+   Parameters:
+   name - The name of the global variable that will be
+          storing this instance.  Is needed to be able to set timeouts.
+   resultPath - path to use for external files
+*/
+function SearchBox(name, resultsPath, inFrame, label)
+{
+  if (!name || !resultsPath) {  alert("Missing parameters to SearchBox."); }
+
+  // ---------- Instance variables
+  this.name                  = name;
+  this.resultsPath           = resultsPath;
+  this.keyTimeout            = 0;
+  this.keyTimeoutLength      = 500;
+  this.closeSelectionTimeout = 300;
+  this.lastSearchValue       = "";
+  this.lastResultsPage       = "";
+  this.hideTimeout           = 0;
+  this.searchIndex           = 0;
+  this.searchActive          = false;
+  this.insideFrame           = inFrame;
+  this.searchLabel           = label;
+
+  // ----------- DOM Elements
+
+  this.DOMSearchField = function()
+  {  return document.getElementById("MSearchField");  }
+
+  this.DOMSearchSelect = function()
+  {  return document.getElementById("MSearchSelect");  }
+
+  this.DOMSearchSelectWindow = function()
+  {  return document.getElementById("MSearchSelectWindow");  }
+
+  this.DOMPopupSearchResults = function()
+  {  return document.getElementById("MSearchResults");  }
+
+  this.DOMPopupSearchResultsWindow = function()
+  {  return document.getElementById("MSearchResultsWindow");  }
+
+  this.DOMSearchClose = function()
+  {  return document.getElementById("MSearchClose"); }
+
+  this.DOMSearchBox = function()
+  {  return document.getElementById("MSearchBox");  }
+
+  // ------------ Event Handlers
+
+  // Called when focus is added or removed from the search field.
+  this.OnSearchFieldFocus = function(isActive)
+  {
+    this.Activate(isActive);
+  }
+
+  this.OnSearchSelectShow = function()
+  {
+    var searchSelectWindow = this.DOMSearchSelectWindow();
+    var searchField        = this.DOMSearchSelect();
+
+    if (this.insideFrame)
+    {
+      var left = getXPos(searchField);
+      var top  = getYPos(searchField);
+      left += searchField.offsetWidth + 6;
+      top += searchField.offsetHeight;
+
+      // show search selection popup
+      searchSelectWindow.style.display='block';
+      left -= searchSelectWindow.offsetWidth;
+      searchSelectWindow.style.left =  left + 'px';
+      searchSelectWindow.style.top  =  top  + 'px';
+    }
+    else
+    {
+      var left = getXPos(searchField);
+      var top  = getYPos(searchField);
+      top += searchField.offsetHeight;
+
+      // show search selection popup
+      searchSelectWindow.style.display='block';
+      searchSelectWindow.style.left =  left + 'px';
+      searchSelectWindow.style.top  =  top  + 'px';
+    }
+
+    // stop selection hide timer
+    if (this.hideTimeout)
+    {
+      clearTimeout(this.hideTimeout);
+      this.hideTimeout=0;
+    }
+    return false; // to avoid "image drag" default event
+  }
+
+  this.OnSearchSelectHide = function()
+  {
+    this.hideTimeout = setTimeout(this.name +".CloseSelectionWindow()",
+                                  this.closeSelectionTimeout);
+  }
+
+  // Called when the content of the search field is changed.
+  this.OnSearchFieldChange = function(evt)
+  {
+    if (this.keyTimeout) // kill running timer
+    {
+      clearTimeout(this.keyTimeout);
+      this.keyTimeout = 0;
+    }
+
+    var e  = (evt) ? evt : window.event; // for IE
+    if (e.keyCode==40 || e.keyCode==13)
+    {
+      if (e.shiftKey==1)
+      {
+        this.OnSearchSelectShow();
+        var win=this.DOMSearchSelectWindow();
+        for (i=0;i<win.childNodes.length;i++)
+        {
+          var child = win.childNodesi; // get span within a
+          if (child.className=='SelectItem')
+          {
+            child.focus();
+            return;
+          }
+        }
+        return;
+      }
+      else
+      {
+        window.frames.MSearchResults.postMessage("take_focus", "*");
+      }
+    }
+    else if (e.keyCode==27) // Escape out of the search field
+    {
+      this.DOMSearchField().blur();
+      this.DOMPopupSearchResultsWindow().style.display = 'none';
+      this.DOMSearchClose().style.display = 'none';
+      this.lastSearchValue = '';
+      this.Activate(false);
+      return;
+    }
+
+    // strip whitespaces
+    var searchValue = this.DOMSearchField().value.replace(/ +/g, "");
+
+    if (searchValue != this.lastSearchValue) // search value has changed
+    {
+      if (searchValue != "") // non-empty search
+      {
+        // set timer for search update
+        this.keyTimeout = setTimeout(this.name + '.Search()',
+                                     this.keyTimeoutLength);
+      }
+      else // empty search field
+      {
+        this.DOMPopupSearchResultsWindow().style.display = 'none';
+        this.DOMSearchClose().style.display = 'none';
+        this.lastSearchValue = '';
+      }
+    }
+  }
+
+  this.SelectItemCount = function(id)
+  {
+    var count=0;
+    var win=this.DOMSearchSelectWindow();
+    for (i=0;i<win.childNodes.length;i++)
+    {
+      var child = win.childNodesi; // get span within a
+      if (child.className=='SelectItem')
+      {
+        count++;
+      }
+    }
+    return count;
+  }
+
+  this.SelectItemSet = function(id)
+  {
+    var i,j=0;
+    var win=this.DOMSearchSelectWindow();
+    for (i=0;i<win.childNodes.length;i++)
+    {
+      var child = win.childNodesi; // get span within a
+      if (child.className=='SelectItem')
+      {
+        var node = child.firstChild;
+        if (j==id)
+        {
+          node.innerHTML='&#8226;';
+        }
+        else
+        {
+          node.innerHTML='&#160;';
+        }
+        j++;
+      }
+    }
+  }
+
+  // Called when an search filter selection is made.
+  // set item with index id as the active item
+  this.OnSelectItem = function(id)
+  {
+    this.searchIndex = id;
+    this.SelectItemSet(id);
+    var searchValue = this.DOMSearchField().value.replace(/ +/g, "");
+    if (searchValue!="" && this.searchActive) // something was found -> do a search
+    {
+      this.Search();
+    }
+  }
+
+  this.OnSearchSelectKey = function(evt)
+  {
+    var e = (evt) ? evt : window.event; // for IE
+    if (e.keyCode==40 && this.searchIndex<this.SelectItemCount()) // Down
+    {
+      this.searchIndex++;
+      this.OnSelectItem(this.searchIndex);
+    }
+    else if (e.keyCode==38 && this.searchIndex>0) // Up
+    {
+      this.searchIndex--;
+      this.OnSelectItem(this.searchIndex);
+    }
+    else if (e.keyCode==13 || e.keyCode==27)
+    {
+      this.OnSelectItem(this.searchIndex);
+      this.CloseSelectionWindow();
+      this.DOMSearchField().focus();
+    }
+    return false;
+  }
+
+  // --------- Actions
+
+  // Closes the results window.
+  this.CloseResultsWindow = function()
+  {
+    this.DOMPopupSearchResultsWindow().style.display = 'none';
+    this.DOMSearchClose().style.display = 'none';
+    this.Activate(false);
+  }
+
+  this.CloseSelectionWindow = function()
+  {
+    this.DOMSearchSelectWindow().style.display = 'none';
+  }
+
+  // Performs a search.
+  this.Search = function()
+  {
+    this.keyTimeout = 0;
+
+    // strip leading whitespace
+    var searchValue = this.DOMSearchField().value.replace(/^ +/, "");
+
+    var code = searchValue.toLowerCase().charCodeAt(0);
+    var idxChar = searchValue.substr(0, 1).toLowerCase();
+    if ( 0xD800 <= code && code <= 0xDBFF && searchValue > 1) // surrogate pair
+    {
+      idxChar = searchValue.substr(0, 2);
+    }
+
+    var resultsPage;
+    var resultsPageWithSearch;
+    var hasResultsPage;
+
+    var idx = indexSectionsWithContentthis.searchIndex.indexOf(idxChar);
+    if (idx!=-1)
+    {
+       var hexCode=idx.toString(16);
+       resultsPage = this.resultsPath + '/' + indexSectionNamesthis.searchIndex + '_' + hexCode + '.html';
+       resultsPageWithSearch = resultsPage+'?'+escape(searchValue);
+       hasResultsPage = true;
+    }
+    else // nothing available for this search term
+    {
+       resultsPage = this.resultsPath + '/nomatches.html';
+       resultsPageWithSearch = resultsPage;
+       hasResultsPage = false;
+    }
+
+    window.frames.MSearchResults.location = resultsPageWithSearch;
+    var domPopupSearchResultsWindow = this.DOMPopupSearchResultsWindow();
+
+    if (domPopupSearchResultsWindow.style.display!='block')
+    {
+       var domSearchBox = this.DOMSearchBox();
+       this.DOMSearchClose().style.display = 'inline-block';
+       if (this.insideFrame)
+       {
+         var domPopupSearchResults = this.DOMPopupSearchResults();
+         domPopupSearchResultsWindow.style.position = 'relative';
+         domPopupSearchResultsWindow.style.display  = 'block';
+         var width = document.body.clientWidth - 8; // the -8 is for IE :-(
+         domPopupSearchResultsWindow.style.width    = width + 'px';
+         domPopupSearchResults.style.width          = width + 'px';
+       }
+       else
+       {
+         var domPopupSearchResults = this.DOMPopupSearchResults();
+         var left = getXPos(domSearchBox) + 150; // domSearchBox.offsetWidth;
+         var top  = getYPos(domSearchBox) + 20;  // domSearchBox.offsetHeight + 1;
+         domPopupSearchResultsWindow.style.display = 'block';
+         left -= domPopupSearchResults.offsetWidth;
+         domPopupSearchResultsWindow.style.top     = top  + 'px';
+         domPopupSearchResultsWindow.style.left    = left + 'px';
+       }
+    }
+
+    this.lastSearchValue = searchValue;
+    this.lastResultsPage = resultsPage;
+  }
+
+  // -------- Activation Functions
+
+  // Activates or deactivates the search panel, resetting things to
+  // their default values if necessary.
+  this.Activate = function(isActive)
+  {
+    if (isActive || // open it
+        this.DOMPopupSearchResultsWindow().style.display == 'block'
+       )
+    {
+      this.DOMSearchBox().className = 'MSearchBoxActive';
+
+      var searchField = this.DOMSearchField();
+
+      if (searchField.value == this.searchLabel) // clear "Search" term upon entry
+      {
+        searchField.value = '';
+        this.searchActive = true;
+      }
+    }
+    else if (!isActive) // directly remove the panel
+    {
+      this.DOMSearchBox().className = 'MSearchBoxInactive';
+      this.DOMSearchField().value   = this.searchLabel;
+      this.searchActive             = false;
+      this.lastSearchValue          = ''
+      this.lastResultsPage          = '';
+    }
+  }
+}
+
+// -----------------------------------------------------------------------
+
+// The class that handles everything on the search results page.
+function SearchResults(name)
+{
+    // The number of matches from the last run of <Search()>.
+    this.lastMatchCount = 0;
+    this.lastKey = 0;
+    this.repeatOn = false;
+
+    // Toggles the visibility of the passed element ID.
+    this.FindChildElement = function(id)
+    {
+      var parentElement = document.getElementById(id);
+      var element = parentElement.firstChild;
+
+      while (element && element!=parentElement)
+      {
+        if (element.nodeName == 'DIV' && element.className == 'SRChildren')
+        {
+          return element;
+        }
+
+        if (element.nodeName == 'DIV' && element.hasChildNodes())
+        {
+           element = element.firstChild;
+        }
+        else if (element.nextSibling)
+        {
+           element = element.nextSibling;
+        }
+        else
+        {
+          do
+          {
+            element = element.parentNode;
+          }
+          while (element && element!=parentElement && !element.nextSibling);
+
+          if (element && element!=parentElement)
+          {
+            element = element.nextSibling;
+          }
+        }
+      }
+    }
+
+    this.Toggle = function(id)
+    {
+      var element = this.FindChildElement(id);
+      if (element)
+      {
+        if (element.style.display == 'block')
+        {
+          element.style.display = 'none';
+        }
+        else
+        {
+          element.style.display = 'block';
+        }
+      }
+    }
+
+    // Searches for the passed string.  If there is no parameter,
+    // it takes it from the URL query.
+    //
+    // Always returns true, since other documents may try to call it
+    // and that may or may not be possible.
+    this.Search = function(search)
+    {
+      if (!search) // get search word from URL
+      {
+        search = window.location.search;
+        search = search.substring(1);  // Remove the leading '?'
+        search = unescape(search);
+      }
+
+      search = search.replace(/^ +/, ""); // strip leading spaces
+      search = search.replace(/ +$/, ""); // strip trailing spaces
+      search = search.toLowerCase();
+      search = convertToId(search);
+
+      var resultRows = document.getElementsByTagName("div");
+      var matches = 0;
+
+      var i = 0;
+      while (i < resultRows.length)
+      {
+        var row = resultRows.item(i);
+        if (row.className == "SRResult")
+        {
+          var rowMatchName = row.id.toLowerCase();
+          rowMatchName = rowMatchName.replace(/^sr\d*_/, ''); // strip 'sr123_'
+
+          if (search.length<=rowMatchName.length &&
+             rowMatchName.substr(0, search.length)==search)
+          {
+            row.style.display = 'block';
+            matches++;
+          }
+          else
+          {
+            row.style.display = 'none';
+          }
+        }
+        i++;
+      }
+      document.getElementById("Searching").style.display='none';
+      if (matches == 0) // no results
+      {
+        document.getElementById("NoMatches").style.display='block';
+      }
+      else // at least one result
+      {
+        document.getElementById("NoMatches").style.display='none';
+      }
+      this.lastMatchCount = matches;
+      return true;
+    }
+
+    // return the first item with index index or higher that is visible
+    this.NavNext = function(index)
+    {
+      var focusItem;
+      while (1)
+      {
+        var focusName = 'Item'+index;
+        focusItem = document.getElementById(focusName);
+        if (focusItem && focusItem.parentNode.parentNode.style.display=='block')
+        {
+          break;
+        }
+        else if (!focusItem) // last element
+        {
+          break;
+        }
+        focusItem=null;
+        index++;
+      }
+      return focusItem;
+    }
+
+    this.NavPrev = function(index)
+    {
+      var focusItem;
+      while (1)
+      {
+        var focusName = 'Item'+index;
+        focusItem = document.getElementById(focusName);
+        if (focusItem && focusItem.parentNode.parentNode.style.display=='block')
+        {
+          break;
+        }
+        else if (!focusItem) // last element
+        {
+          break;
+        }
+        focusItem=null;
+        index--;
+      }
+      return focusItem;
+    }
+
+    this.ProcessKeys = function(e)
+    {
+      if (e.type == "keydown")
+      {
+        this.repeatOn = false;
+        this.lastKey = e.keyCode;
+      }
+      else if (e.type == "keypress")
+      {
+        if (!this.repeatOn)
+        {
+          if (this.lastKey) this.repeatOn = true;
+          return false; // ignore first keypress after keydown
+        }
+      }
+      else if (e.type == "keyup")
+      {
+        this.lastKey = 0;
+        this.repeatOn = false;
+      }
+      return this.lastKey!=0;
+    }
+
+    this.Nav = function(evt,itemIndex)
+    {
+      var e  = (evt) ? evt : window.event; // for IE
+      if (e.keyCode==13) return true;
+      if (!this.ProcessKeys(e)) return false;
+
+      if (this.lastKey==38) // Up
+      {
+        var newIndex = itemIndex-1;
+        var focusItem = this.NavPrev(newIndex);
+        if (focusItem)
+        {
+          var child = this.FindChildElement(focusItem.parentNode.parentNode.id);
+          if (child && child.style.display == 'block') // children visible
+          {
+            var n=0;
+            var tmpElem;
+            while (1) // search for last child
+            {
+              tmpElem = document.getElementById('Item'+newIndex+'_c'+n);
+              if (tmpElem)
+              {
+                focusItem = tmpElem;
+              }
+              else // found it!
+              {
+                break;
+              }
+              n++;
+            }
+          }
+        }
+        if (focusItem)
+        {
+          focusItem.focus();
+        }
+        else // return focus to search field
+        {
+           parent.document.getElementById("MSearchField").focus();
+        }
+      }
+      else if (this.lastKey==40) // Down
+      {
+        var newIndex = itemIndex+1;
+        var focusItem;
+        var item = document.getElementById('Item'+itemIndex);
+        var elem = this.FindChildElement(item.parentNode.parentNode.id);
+        if (elem && elem.style.display == 'block') // children visible
+        {
+          focusItem = document.getElementById('Item'+itemIndex+'_c0');
+        }
+        if (!focusItem) focusItem = this.NavNext(newIndex);
+        if (focusItem)  focusItem.focus();
+      }
+      else if (this.lastKey==39) // Right
+      {
+        var item = document.getElementById('Item'+itemIndex);
+        var elem = this.FindChildElement(item.parentNode.parentNode.id);
+        if (elem) elem.style.display = 'block';
+      }
+      else if (this.lastKey==37) // Left
+      {
+        var item = document.getElementById('Item'+itemIndex);
+        var elem = this.FindChildElement(item.parentNode.parentNode.id);
+        if (elem) elem.style.display = 'none';
+      }
+      else if (this.lastKey==27) // Escape
+      {
+        parent.searchBox.CloseResultsWindow();
+        parent.document.getElementById("MSearchField").focus();
+      }
+      else if (this.lastKey==13) // Enter
+      {
+        return true;
+      }
+      return false;
+    }
+
+    this.NavChild = function(evt,itemIndex,childIndex)
+    {
+      var e  = (evt) ? evt : window.event; // for IE
+      if (e.keyCode==13) return true;
+      if (!this.ProcessKeys(e)) return false;
+
+      if (this.lastKey==38) // Up
+      {
+        if (childIndex>0)
+        {
+          var newIndex = childIndex-1;
+          document.getElementById('Item'+itemIndex+'_c'+newIndex).focus();
+        }
+        else // already at first child, jump to parent
+        {
+          document.getElementById('Item'+itemIndex).focus();
+        }
+      }
+      else if (this.lastKey==40) // Down
+      {
+        var newIndex = childIndex+1;
+        var elem = document.getElementById('Item'+itemIndex+'_c'+newIndex);
+        if (!elem) // last child, jump to parent next parent
+        {
+          elem = this.NavNext(itemIndex+1);
+        }
+        if (elem)
+        {
+          elem.focus();
+        }
+      }
+      else if (this.lastKey==27) // Escape
+      {
+        parent.searchBox.CloseResultsWindow();
+        parent.document.getElementById("MSearchField").focus();
+      }
+      else if (this.lastKey==13) // Enter
+      {
+        return true;
+      }
+      return false;
+    }
+}
+
+function setKeyActions(elem,action)
+{
+  elem.setAttribute('onkeydown',action);
+  elem.setAttribute('onkeypress',action);
+  elem.setAttribute('onkeyup',action);
+}
+
+function setClassAttr(elem,attr)
+{
+  elem.setAttribute('class',attr);
+  elem.setAttribute('className',attr);
+}
+
+function createResults()
+{
+  var results = document.getElementById("SRResults");
+  for (var e=0; e<searchData.length; e++)
+  {
+    var id = searchDatae0;
+    var srResult = document.createElement('div');
+    srResult.setAttribute('id','SR_'+id);
+    setClassAttr(srResult,'SRResult');
+    var srEntry = document.createElement('div');
+    setClassAttr(srEntry,'SREntry');
+    var srLink = document.createElement('a');
+    srLink.setAttribute('id','Item'+e);
+    setKeyActions(srLink,'return searchResults.Nav(event,'+e+')');
+    setClassAttr(srLink,'SRSymbol');
+    srLink.innerHTML = searchDatae10;
+    srEntry.appendChild(srLink);
+    if (searchDatae1.length==2) // single result
+    {
+      srLink.setAttribute('href',searchDatae110);
+      if (searchDatae111)
+      {
+       srLink.setAttribute('target','_parent');
+      }
+      var srScope = document.createElement('span');
+      setClassAttr(srScope,'SRScope');
+      srScope.innerHTML = searchDatae112;
+      srEntry.appendChild(srScope);
+    }
+    else // multiple results
+    {
+      srLink.setAttribute('href','javascript:searchResults.Toggle("SR_'+id+'")');
+      var srChildren = document.createElement('div');
+      setClassAttr(srChildren,'SRChildren');
+      for (var c=0; c<searchDatae1.length-1; c++)
+      {
+        var srChild = document.createElement('a');
+        srChild.setAttribute('id','Item'+e+'_c'+c);
+        setKeyActions(srChild,'return searchResults.NavChild(event,'+e+','+c+')');
+        setClassAttr(srChild,'SRScope');
+        srChild.setAttribute('href',searchDatae1c+10);
+        if (searchDatae1c+11)
+        {
+         srChild.setAttribute('target','_parent');
+        }
+        srChild.innerHTML = searchDatae1c+12;
+        srChildren.appendChild(srChild);
+      }
+      srEntry.appendChild(srChildren);
+    }
+    srResult.appendChild(srEntry);
+    results.appendChild(srResult);
+  }
+}
+
+function init_search()
+{
+  var results = document.getElementById("MSearchSelectWindow");
+  for (var key in indexSectionLabels)
+  {
+    var link = document.createElement('a');
+    link.setAttribute('class','SelectItem');
+    link.setAttribute('onclick','searchBox.OnSelectItem('+key+')');
+    link.href='javascript:void(0)';
+    link.innerHTML='<span class="SelectionMark">&#160;</span>'+indexSectionLabelskey;
+    results.appendChild(link);
+  }
+  searchBox.OnSelectItem(0);
+}
+/* @license-end */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/search_l.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/search_m.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/search_r.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/searchdata.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/typedefs_0.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="typedefs_0.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/typedefs_0.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_0.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_0.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_0.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_1.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_1.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_1.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_2.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_2.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_2.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_3.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_3.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_3.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_4.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_4.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_4.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_5.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_5.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_5.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_6.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_6.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_6.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_7.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_7.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_7.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_8.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_8.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_8.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_9.html Added

@@ -0,0 +1,36 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html><head><title></title>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<link rel="stylesheet" type="text/css" href="search.css"/>
+<script type="text/javascript" src="variables_9.js"></script>
+<script type="text/javascript" src="search.js"></script>
+</head>
+<body class="SRPage">
+<div id="SRIndex">
+<div class="SRStatus" id="Loading">Loading...</div>
+<div id="SRResults"></div>
+<script type="text/javascript"></script>
+<div class="SRStatus" id="Searching">Searching...</div>
+<div class="SRStatus" id="NoMatches">No Matches</div>
+<script type="text/javascript"></script>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/search/variables_9.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/splitbar.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/structtjregion.html Added

@@ -0,0 +1,173 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: tjregion Struct Reference</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+</div>
+<div class="header">
+  <div class="summary">
+<a href="#pub-attribs">Data Fields</a>  </div>
+  <div class="headertitle">
+<div class="title">tjregion Struct Reference<div class="ingroups"><a class="el" href="group___turbo_j_p_e_g.html">TurboJPEG</a></div></div>  </div>
+</div>
+<div class="contents">
+
+<p>Cropping region.  
+ <a href="structtjregion.html#details">More...</a></p>
+
+<p><code>#include &lt;turbojpeg.h&gt;</code></p>
+<table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="pub-attribs"></a>
+Data Fields</h2></td></tr>
+<tr class="memitem:a4b6a37a93997091b26a75831fa291ad9"><td class="memItemLeft" align="right" valign="top">int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjregion.html#a4b6a37a93997091b26a75831fa291ad9">x</a></td></tr>
+<tr class="memdesc:a4b6a37a93997091b26a75831fa291ad9"><td class="mdescLeft">&#160;</td><td class="mdescRight">The left boundary of the cropping region.  <a href="structtjregion.html#a4b6a37a93997091b26a75831fa291ad9">More...</a><br /></td></tr>
+<tr class="separator:a4b6a37a93997091b26a75831fa291ad9"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:a7b3e0c24cfe87acc80e334cafdcf22c2"><td class="memItemLeft" align="right" valign="top">int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjregion.html#a7b3e0c24cfe87acc80e334cafdcf22c2">y</a></td></tr>
+<tr class="memdesc:a7b3e0c24cfe87acc80e334cafdcf22c2"><td class="mdescLeft">&#160;</td><td class="mdescRight">The upper boundary of the cropping region.  <a href="structtjregion.html#a7b3e0c24cfe87acc80e334cafdcf22c2">More...</a><br /></td></tr>
+<tr class="separator:a7b3e0c24cfe87acc80e334cafdcf22c2"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ab6eb73ceef584fc23c8c8097926dce42"><td class="memItemLeft" align="right" valign="top">int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjregion.html#ab6eb73ceef584fc23c8c8097926dce42">w</a></td></tr>
+<tr class="memdesc:ab6eb73ceef584fc23c8c8097926dce42"><td class="mdescLeft">&#160;</td><td class="mdescRight">The width of the cropping region.  <a href="structtjregion.html#ab6eb73ceef584fc23c8c8097926dce42">More...</a><br /></td></tr>
+<tr class="separator:ab6eb73ceef584fc23c8c8097926dce42"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:aecefc45a26f4d8b60dd4d825c1710115"><td class="memItemLeft" align="right" valign="top">int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjregion.html#aecefc45a26f4d8b60dd4d825c1710115">h</a></td></tr>
+<tr class="memdesc:aecefc45a26f4d8b60dd4d825c1710115"><td class="mdescLeft">&#160;</td><td class="mdescRight">The height of the cropping region.  <a href="structtjregion.html#aecefc45a26f4d8b60dd4d825c1710115">More...</a><br /></td></tr>
+<tr class="separator:aecefc45a26f4d8b60dd4d825c1710115"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table>
+<a name="details" id="details"></a><h2 class="groupheader">Detailed Description</h2>
+<div class="textblock"><p>Cropping region. </p>
+</div><h2 class="groupheader">Field Documentation</h2>
+<a id="aecefc45a26f4d8b60dd4d825c1710115"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#aecefc45a26f4d8b60dd4d825c1710115">&#9670;&nbsp;</a></span>h</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">int tjregion::h</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The height of the cropping region. </p>
+<p>Setting this to 0 is the equivalent of setting it to the height of the source JPEG image - y. </p>
+
+</div>
+</div>
+<a id="ab6eb73ceef584fc23c8c8097926dce42"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ab6eb73ceef584fc23c8c8097926dce42">&#9670;&nbsp;</a></span>w</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">int tjregion::w</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The width of the cropping region. </p>
+<p>Setting this to 0 is the equivalent of setting it to the width of the source JPEG image - x. </p>
+
+</div>
+</div>
+<a id="a4b6a37a93997091b26a75831fa291ad9"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a4b6a37a93997091b26a75831fa291ad9">&#9670;&nbsp;</a></span>x</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">int tjregion::x</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The left boundary of the cropping region. </p>
+<p>This must be evenly divisible by the MCU block width (see <a class="el" href="group___turbo_j_p_e_g.html#ga9e61e7cd47a15a173283ba94e781308c" title="MCU block width (in pixels) for a given level of chrominance subsampling.">tjMCUWidth</a>.) </p>
+
+</div>
+</div>
+<a id="a7b3e0c24cfe87acc80e334cafdcf22c2"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a7b3e0c24cfe87acc80e334cafdcf22c2">&#9670;&nbsp;</a></span>y</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">int tjregion::y</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The upper boundary of the cropping region. </p>
+<p>This must be evenly divisible by the MCU block height (see <a class="el" href="group___turbo_j_p_e_g.html#gabd247bb9fecb393eca57366feb8327bf" title="MCU block height (in pixels) for a given level of chrominance subsampling.">tjMCUHeight</a>.) </p>
+
+</div>
+</div>
+<hr/>The documentation for this struct was generated from the following file:<ul>
+<li>turbojpeg.h</li>
+</ul>
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/structtjscalingfactor.html Added

@@ -0,0 +1,131 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: tjscalingfactor Struct Reference</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+</div>
+<div class="header">
+  <div class="summary">
+<a href="#pub-attribs">Data Fields</a>  </div>
+  <div class="headertitle">
+<div class="title">tjscalingfactor Struct Reference<div class="ingroups"><a class="el" href="group___turbo_j_p_e_g.html">TurboJPEG</a></div></div>  </div>
+</div>
+<div class="contents">
+
+<p>Scaling factor.  
+ <a href="structtjscalingfactor.html#details">More...</a></p>
+
+<p><code>#include &lt;turbojpeg.h&gt;</code></p>
+<table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="pub-attribs"></a>
+Data Fields</h2></td></tr>
+<tr class="memitem:a9b011e57f981ee23083e2c1aa5e640ec"><td class="memItemLeft" align="right" valign="top">int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjscalingfactor.html#a9b011e57f981ee23083e2c1aa5e640ec">num</a></td></tr>
+<tr class="memdesc:a9b011e57f981ee23083e2c1aa5e640ec"><td class="mdescLeft">&#160;</td><td class="mdescRight">Numerator.  <a href="structtjscalingfactor.html#a9b011e57f981ee23083e2c1aa5e640ec">More...</a><br /></td></tr>
+<tr class="separator:a9b011e57f981ee23083e2c1aa5e640ec"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:aefbcdf3e9e62274b2d312c695f133ce3"><td class="memItemLeft" align="right" valign="top">int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjscalingfactor.html#aefbcdf3e9e62274b2d312c695f133ce3">denom</a></td></tr>
+<tr class="memdesc:aefbcdf3e9e62274b2d312c695f133ce3"><td class="mdescLeft">&#160;</td><td class="mdescRight">Denominator.  <a href="structtjscalingfactor.html#aefbcdf3e9e62274b2d312c695f133ce3">More...</a><br /></td></tr>
+<tr class="separator:aefbcdf3e9e62274b2d312c695f133ce3"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table>
+<a name="details" id="details"></a><h2 class="groupheader">Detailed Description</h2>
+<div class="textblock"><p>Scaling factor. </p>
+</div><h2 class="groupheader">Field Documentation</h2>
+<a id="aefbcdf3e9e62274b2d312c695f133ce3"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#aefbcdf3e9e62274b2d312c695f133ce3">&#9670;&nbsp;</a></span>denom</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">int tjscalingfactor::denom</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Denominator. </p>
+
+</div>
+</div>
+<a id="a9b011e57f981ee23083e2c1aa5e640ec"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a9b011e57f981ee23083e2c1aa5e640ec">&#9670;&nbsp;</a></span>num</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">int tjscalingfactor::num</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Numerator. </p>
+
+</div>
+</div>
+<hr/>The documentation for this struct was generated from the following file:<ul>
+<li>turbojpeg.h</li>
+</ul>
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/structtjtransform.html Added

@@ -0,0 +1,201 @@
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml">
+<head>
+<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
+<meta http-equiv="X-UA-Compatible" content="IE=9"/>
+<meta name="generator" content="Doxygen 1.8.20"/>
+<meta name="viewport" content="width=device-width, initial-scale=1"/>
+<title>TurboJPEG: tjtransform Struct Reference</title>
+<link href="tabs.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="jquery.js"></script>
+<script type="text/javascript" src="dynsections.js"></script>
+<link href="search/search.css" rel="stylesheet" type="text/css"/>
+<script type="text/javascript" src="search/searchdata.js"></script>
+<script type="text/javascript" src="search/search.js"></script>
+<link href="doxygen.css" rel="stylesheet" type="text/css" />
+<link href="doxygen-extra.css" rel="stylesheet" type="text/css"/>
+</head>
+<body>
+<div id="top">
+<div id="titlearea">
+<table cellspacing="0" cellpadding="0">
+ <tbody>
+ <tr style="height: 56px;">
+  <td id="projectalign" style="padding-left: 0.5em;">
+   <div id="projectname">TurboJPEG
+   &#160;<span id="projectnumber">2.1</span>
+   </div>
+  </td>
+ </tr>
+ </tbody>
+</table>
+</div>
+
+
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+var searchBox = new SearchBox("searchBox", "search",false,'Search');
+/* @license-end */
+</script>
+<script type="text/javascript" src="menudata.js"></script>
+<script type="text/javascript" src="menu.js"></script>
+<script type="text/javascript">
+/* @license magnet:?xt=urn:btih:cf05388f2679ee054f2beb29a391d25f4e673ac3&amp;dn=gpl-2.0.txt GPL-v2 */
+$(function() {
+  initMenu('',true,false,'search.php','Search');
+  $(document).ready(function() { init_search(); });
+});
+/* @license-end */</script>
+<div id="main-nav"></div>
+
+<div id="MSearchSelectWindow"
+     onmouseover="return searchBox.OnSearchSelectShow()"
+     onmouseout="return searchBox.OnSearchSelectHide()"
+     onkeydown="return searchBox.OnSearchSelectKey(event)">
+</div>
+
+
+<div id="MSearchResultsWindow">
+<iframe src="javascript:void(0)" frameborder="0" 
+        name="MSearchResults" id="MSearchResults">
+</iframe>
+</div>
+
+</div>
+<div class="header">
+  <div class="summary">
+<a href="#pub-attribs">Data Fields</a>  </div>
+  <div class="headertitle">
+<div class="title">tjtransform Struct Reference<div class="ingroups"><a class="el" href="group___turbo_j_p_e_g.html">TurboJPEG</a></div></div>  </div>
+</div>
+<div class="contents">
+
+<p>Lossless transform.  
+ <a href="structtjtransform.html#details">More...</a></p>
+
+<p><code>#include &lt;turbojpeg.h&gt;</code></p>
+<table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="pub-attribs"></a>
+Data Fields</h2></td></tr>
+<tr class="memitem:ac324e5e442abec8a961e5bf219db12cf"><td class="memItemLeft" align="right" valign="top"><a class="el" href="structtjregion.html">tjregion</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjtransform.html#ac324e5e442abec8a961e5bf219db12cf">r</a></td></tr>
+<tr class="memdesc:ac324e5e442abec8a961e5bf219db12cf"><td class="mdescLeft">&#160;</td><td class="mdescRight">Cropping region.  <a href="structtjtransform.html#ac324e5e442abec8a961e5bf219db12cf">More...</a><br /></td></tr>
+<tr class="separator:ac324e5e442abec8a961e5bf219db12cf"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:a2525aab4ba6978a1c273f74fef50e498"><td class="memItemLeft" align="right" valign="top">int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjtransform.html#a2525aab4ba6978a1c273f74fef50e498">op</a></td></tr>
+<tr class="memdesc:a2525aab4ba6978a1c273f74fef50e498"><td class="mdescLeft">&#160;</td><td class="mdescRight">One of the <a class="el" href="group___turbo_j_p_e_g.html#ga2de531af4e7e6c4f124908376b354866">transform operations</a>.  <a href="structtjtransform.html#a2525aab4ba6978a1c273f74fef50e498">More...</a><br /></td></tr>
+<tr class="separator:a2525aab4ba6978a1c273f74fef50e498"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:ac0e74655baa4402209a21e1ae481c8f6"><td class="memItemLeft" align="right" valign="top">int&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjtransform.html#ac0e74655baa4402209a21e1ae481c8f6">options</a></td></tr>
+<tr class="memdesc:ac0e74655baa4402209a21e1ae481c8f6"><td class="mdescLeft">&#160;</td><td class="mdescRight">The bitwise OR of one of more of the <a class="el" href="group___turbo_j_p_e_g.html#ga9c771a757fc1294add611906b89ab2d2">transform options</a>.  <a href="structtjtransform.html#ac0e74655baa4402209a21e1ae481c8f6">More...</a><br /></td></tr>
+<tr class="separator:ac0e74655baa4402209a21e1ae481c8f6"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:a688fe8f1a8ecc12a538d9e561cf338e3"><td class="memItemLeft" align="right" valign="top">void *&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjtransform.html#a688fe8f1a8ecc12a538d9e561cf338e3">data</a></td></tr>
+<tr class="memdesc:a688fe8f1a8ecc12a538d9e561cf338e3"><td class="mdescLeft">&#160;</td><td class="mdescRight">Arbitrary data that can be accessed within the body of the callback function.  <a href="structtjtransform.html#a688fe8f1a8ecc12a538d9e561cf338e3">More...</a><br /></td></tr>
+<tr class="separator:a688fe8f1a8ecc12a538d9e561cf338e3"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:afd7fc262df33f741e120ef4183202ef5"><td class="memItemLeft" align="right" valign="top">int(*&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="structtjtransform.html#afd7fc262df33f741e120ef4183202ef5">customFilter</a> )(short *coeffs, <a class="el" href="structtjregion.html">tjregion</a> arrayRegion, <a class="el" href="structtjregion.html">tjregion</a> planeRegion, int componentIndex, int transformIndex, struct <a class="el" href="structtjtransform.html">tjtransform</a> *transform)</td></tr>
+<tr class="memdesc:afd7fc262df33f741e120ef4183202ef5"><td class="mdescLeft">&#160;</td><td class="mdescRight">A callback function that can be used to modify the DCT coefficients after they are losslessly transformed but before they are transcoded to a new JPEG image.  <a href="structtjtransform.html#afd7fc262df33f741e120ef4183202ef5">More...</a><br /></td></tr>
+<tr class="separator:afd7fc262df33f741e120ef4183202ef5"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table>
+<a name="details" id="details"></a><h2 class="groupheader">Detailed Description</h2>
+<div class="textblock"><p>Lossless transform. </p>
+</div><h2 class="groupheader">Field Documentation</h2>
+<a id="afd7fc262df33f741e120ef4183202ef5"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#afd7fc262df33f741e120ef4183202ef5">&#9670;&nbsp;</a></span>customFilter</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">int(* tjtransform::customFilter) (short *coeffs, <a class="el" href="structtjregion.html">tjregion</a> arrayRegion, <a class="el" href="structtjregion.html">tjregion</a> planeRegion, int componentIndex, int transformIndex, struct <a class="el" href="structtjtransform.html">tjtransform</a> *transform)</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>A callback function that can be used to modify the DCT coefficients after they are losslessly transformed but before they are transcoded to a new JPEG image. </p>
+<p>This allows for custom filters or other transformations to be applied in the frequency domain.</p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">coeffs</td><td>pointer to an array of transformed DCT coefficients. (NOTE: this pointer is not guaranteed to be valid once the callback returns, so applications wishing to hand off the DCT coefficients to another function or library should make a copy of them within the body of the callback.)</td></tr>
+    <tr><td class="paramname">arrayRegion</td><td><a class="el" href="structtjregion.html" title="Cropping region.">tjregion</a> structure containing the width and height of the array pointed to by <code>coeffs</code> as well as its offset relative to the component plane. TurboJPEG implementations may choose to split each component plane into multiple DCT coefficient arrays and call the callback function once for each array.</td></tr>
+    <tr><td class="paramname">planeRegion</td><td><a class="el" href="structtjregion.html" title="Cropping region.">tjregion</a> structure containing the width and height of the component plane to which <code>coeffs</code> belongs</td></tr>
+    <tr><td class="paramname">componentID</td><td>ID number of the component plane to which <code>coeffs</code> belongs (Y, Cb, and Cr have, respectively, ID's of 0, 1, and 2 in typical JPEG images.)</td></tr>
+    <tr><td class="paramname">transformID</td><td>ID number of the transformed image to which <code>coeffs</code> belongs. This is the same as the index of the transform in the <code>transforms</code> array that was passed to <a class="el" href="group___turbo_j_p_e_g.html#ga9cb8abf4cc91881e04a0329b2270be25" title="Losslessly transform a JPEG image into another JPEG image.">tjTransform()</a>.</td></tr>
+    <tr><td class="paramname">transform</td><td>a pointer to a <a class="el" href="structtjtransform.html" title="Lossless transform.">tjtransform</a> structure that specifies the parameters and/or cropping region for this transform</td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>0 if the callback was successful, or -1 if an error occurred. </dd></dl>
+
+</div>
+</div>
+<a id="a688fe8f1a8ecc12a538d9e561cf338e3"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a688fe8f1a8ecc12a538d9e561cf338e3">&#9670;&nbsp;</a></span>data</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">void* tjtransform::data</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Arbitrary data that can be accessed within the body of the callback function. </p>
+
+</div>
+</div>
+<a id="a2525aab4ba6978a1c273f74fef50e498"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a2525aab4ba6978a1c273f74fef50e498">&#9670;&nbsp;</a></span>op</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">int tjtransform::op</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>One of the <a class="el" href="group___turbo_j_p_e_g.html#ga2de531af4e7e6c4f124908376b354866">transform operations</a>. </p>
+
+</div>
+</div>
+<a id="ac0e74655baa4402209a21e1ae481c8f6"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ac0e74655baa4402209a21e1ae481c8f6">&#9670;&nbsp;</a></span>options</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">int tjtransform::options</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>The bitwise OR of one of more of the <a class="el" href="group___turbo_j_p_e_g.html#ga9c771a757fc1294add611906b89ab2d2">transform options</a>. </p>
+
+</div>
+</div>
+<a id="ac324e5e442abec8a961e5bf219db12cf"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ac324e5e442abec8a961e5bf219db12cf">&#9670;&nbsp;</a></span>r</h2>
+
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname"><a class="el" href="structtjregion.html">tjregion</a> tjtransform::r</td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>Cropping region. </p>
+
+</div>
+</div>
+<hr/>The documentation for this struct was generated from the following file:<ul>
+<li>turbojpeg.h</li>
+</ul>
+</div>
+
+<hr class="footer"/><address class="footer"><small>
+Generated by&#160;<a href="http://www.doxygen.org/index.html"><img class="footer" src="doxygen.svg" width="104" height="31" alt="doxygen"/></a> 1.8.20
+</small></address>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/sync_off.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/sync_on.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/tab_a.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/tab_b.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/tab_h.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/tab_s.png Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doc/html/tabs.css Added

@@ -0,0 +1,1 @@
+.sm{position:relative;z-index:9999}.sm,.sm ul,.sm li{display:block;list-style:none;margin:0;padding:0;line-height:normal;direction:ltr;text-align:left;-webkit-tap-highlight-color:rgba(0,0,0,0)}.sm-rtl,.sm-rtl ul,.sm-rtl li{direction:rtl;text-align:right}.sm>li>h1,.sm>li>h2,.sm>li>h3,.sm>li>h4,.sm>li>h5,.sm>li>h6{margin:0;padding:0}.sm ul{display:none}.sm li,.sm a{position:relative}.sm a{display:block}.sm a.disabled{cursor:not-allowed}.sm:after{content:"\00a0";display:block;height:0;font:0/0 serif;clear:both;visibility:hidden;overflow:hidden}.sm,.sm *,.sm *:before,.sm *:after{-moz-box-sizing:border-box;-webkit-box-sizing:border-box;box-sizing:border-box}.sm-dox{background-image:url("tab_b.png")}.sm-dox a,.sm-dox a:focus,.sm-dox a:hover,.sm-dox a:active{padding:0 12px;padding-right:43px;font-family:"Lucida Grande","Geneva","Helvetica",Arial,sans-serif;font-size:13px;font-weight:bold;line-height:36px;text-decoration:none;text-shadow:0 1px 1px rgba(255,255,255,0.9);color:#283a5d;outline:0}.sm-dox a:hover{background-image:url("tab_a.png");background-repeat:repeat-x;color:white;text-shadow:0 1px 1px black}.sm-dox a.current{color:#d23600}.sm-dox a.disabled{color:#bbb}.sm-dox a span.sub-arrow{position:absolute;top:50%;margin-top:-14px;left:auto;right:3px;width:28px;height:28px;overflow:hidden;font:bold 12px/28px monospace!important;text-align:center;text-shadow:none;background:rgba(255,255,255,0.5);-moz-border-radius:5px;-webkit-border-radius:5px;border-radius:5px}.sm-dox a.highlighted span.sub-arrow:before{display:block;content:'-'}.sm-dox>li:first-child>a,.sm-dox>li:first-child>:not(ul) a{-moz-border-radius:5px 5px 0 0;-webkit-border-radius:5px;border-radius:5px 5px 0 0}.sm-dox>li:last-child>a,.sm-dox>li:last-child>*:not(ul) a,.sm-dox>li:last-child>ul,.sm-dox>li:last-child>ul>li:last-child>a,.sm-dox>li:last-child>ul>li:last-child>*:not(ul) a,.sm-dox>li:last-child>ul>li:last-child>ul,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>a,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>*:not(ul) a,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>a,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>*:not(ul) a,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>ul,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>a,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>*:not(ul) a,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>ul{-moz-border-radius:0 0 5px 5px;-webkit-border-radius:0;border-radius:0 0 5px 5px}.sm-dox>li:last-child>a.highlighted,.sm-dox>li:last-child>*:not(ul) a.highlighted,.sm-dox>li:last-child>ul>li:last-child>a.highlighted,.sm-dox>li:last-child>ul>li:last-child>*:not(ul) a.highlighted,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>a.highlighted,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>*:not(ul) a.highlighted,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>a.highlighted,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>*:not(ul) a.highlighted,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>a.highlighted,.sm-dox>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>ul>li:last-child>*:not(ul) a.highlighted{-moz-border-radius:0;-webkit-border-radius:0;border-radius:0}.sm-dox ul{background:rgba(162,162,162,0.1)}.sm-dox ul a,.sm-dox ul a:focus,.sm-dox ul a:hover,.sm-dox ul a:active{font-size:12px;border-left:8px solid transparent;line-height:36px;text-shadow:none;background-color:white;background-image:none}.sm-dox ul a:hover{background-image:url("tab_a.png");background-repeat:repeat-x;color:white;text-shadow:0 1px 1px black}.sm-dox ul ul a,.sm-dox ul ul a:hover,.sm-dox ul ul a:focus,.sm-dox ul ul a:active{border-left:16px solid transparent}.sm-dox ul ul ul a,.sm-dox ul ul ul a:hover,.sm-dox ul ul ul a:focus,.sm-dox ul ul ul a:active{border-left:24px solid transparent}.sm-dox ul ul ul ul a,.sm-dox ul ul ul ul a:hover,.sm-dox ul ul ul ul a:focus,.sm-dox ul ul ul ul a:active{border-left:32px solid transparent}.sm-dox ul ul ul ul ul a,.sm-dox ul ul ul ul ul a:hover,.sm-dox ul ul ul ul ul a:focus,.sm-dox ul ul ul ul ul a:active{border-left:40px solid transparent}@media(min-width:768px){.sm-dox ul{position:absolute;width:12em}.sm-dox li{float:left}.sm-dox.sm-rtl li{float:right}.sm-dox ul li,.sm-dox.sm-rtl ul li,.sm-dox.sm-vertical li{float:none}.sm-dox a{white-space:nowrap}.sm-dox ul a,.sm-dox.sm-vertical a{white-space:normal}.sm-dox .sm-nowrap>li>a,.sm-dox .sm-nowrap>li>:not(ul) a{white-space:nowrap}.sm-dox{padding:0 10px;background-image:url("tab_b.png");line-height:36px}.sm-dox a span.sub-arrow{top:50%;margin-top:-2px;right:12px;width:0;height:0;border-width:4px;border-style:solid dashed dashed dashed;border-color:#283a5d transparent transparent transparent;background:transparent;-moz-border-radius:0;-webkit-border-radius:0;border-radius:0}.sm-dox a,.sm-dox a:focus,.sm-dox a:active,.sm-dox a:hover,.sm-dox a.highlighted{padding:0 12px;background-image:url("tab_s.png");background-repeat:no-repeat;background-position:right;-moz-border-radius:0!important;-webkit-border-radius:0;border-radius:0!important}.sm-dox a:hover{background-image:url("tab_a.png");background-repeat:repeat-x;color:white;text-shadow:0 1px 1px black}.sm-dox a:hover span.sub-arrow{border-color:white transparent transparent transparent}.sm-dox a.has-submenu{padding-right:24px}.sm-dox li{border-top:0}.sm-dox>li>ul:before,.sm-dox>li>ul:after{content:'';position:absolute;top:-18px;left:30px;width:0;height:0;overflow:hidden;border-width:9px;border-style:dashed dashed solid dashed;border-color:transparent transparent #bbb transparent}.sm-dox>li>ul:after{top:-16px;left:31px;border-width:8px;border-color:transparent transparent #fff transparent}.sm-dox ul{border:1px solid #bbb;padding:5px 0;background:#fff;-moz-border-radius:5px!important;-webkit-border-radius:5px;border-radius:5px!important;-moz-box-shadow:0 5px 9px rgba(0,0,0,0.2);-webkit-box-shadow:0 5px 9px rgba(0,0,0,0.2);box-shadow:0 5px 9px rgba(0,0,0,0.2)}.sm-dox ul a span.sub-arrow{right:8px;top:50%;margin-top:-5px;border-width:5px;border-color:transparent transparent transparent #555;border-style:dashed dashed dashed solid}.sm-dox ul a,.sm-dox ul a:hover,.sm-dox ul a:focus,.sm-dox ul a:active,.sm-dox ul a.highlighted{color:#555;background-image:none;border:0!important;color:#555;background-image:none}.sm-dox ul a:hover{background-image:url("tab_a.png");background-repeat:repeat-x;color:white;text-shadow:0 1px 1px black}.sm-dox ul a:hover span.sub-arrow{border-color:transparent transparent transparent white}.sm-dox span.scroll-up,.sm-dox span.scroll-down{position:absolute;display:none;visibility:hidden;overflow:hidden;background:#fff;height:36px}.sm-dox span.scroll-up:hover,.sm-dox span.scroll-down:hover{background:#eee}.sm-dox span.scroll-up:hover span.scroll-up-arrow,.sm-dox span.scroll-up:hover span.scroll-down-arrow{border-color:transparent transparent #d23600 transparent}.sm-dox span.scroll-down:hover span.scroll-down-arrow{border-color:#d23600 transparent transparent transparent}.sm-dox span.scroll-up-arrow,.sm-dox span.scroll-down-arrow{position:absolute;top:0;left:50%;margin-left:-6px;width:0;height:0;overflow:hidden;border-width:6px;border-style:dashed dashed solid dashed;border-color:transparent transparent #555 transparent}.sm-dox span.scroll-down-arrow{top:8px;border-style:solid dashed dashed dashed;border-color:#555 transparent transparent transparent}.sm-dox.sm-rtl a.has-submenu{padding-right:12px;padding-left:24px}.sm-dox.sm-rtl a span.sub-arrow{right:auto;left:12px}.sm-dox.sm-rtl.sm-vertical a.has-submenu{padding:10px 20px}.sm-dox.sm-rtl.sm-vertical a span.sub-arrow{right:auto;left:8px;border-style:dashed solid dashed dashed;border-color:transparent #555 transparent transparent}.sm-dox.sm-rtl>li>ul:before{left:auto;right:30px}.sm-dox.sm-rtl>li>ul:after{left:auto;right:31px}.sm-dox.sm-rtl ul a.has-submenu{padding:10px 20px!important}.sm-dox.sm-rtl ul a span.sub-arrow{right:auto;left:8px;border-style:dashed solid dashed dashed;border-color:transparent #555 transparent transparent}.sm-dox.sm-vertical{padding:10px 0;-moz-border-radius:5px;-webkit-border-radius:5px;border-radius:5px}.sm-dox.sm-vertical a{padding:10px 20px}.sm-dox.sm-vertical a:hover,.sm-dox.sm-vertical a:focus,.sm-dox.sm-vertical a:active,.sm-dox.sm-vertical a.highlighted{background:#fff}.sm-dox.sm-vertical a.disabled{background-image:url("tab_b.png")}.sm-dox.sm-vertical a span.sub-arrow{right:8px;top:50%;margin-top:-5px;border-width:5px;border-style:dashed dashed dashed solid;border-color:transparent transparent transparent #555}.sm-dox.sm-vertical>li>ul:before,.sm-dox.sm-vertical>li>ul:after{display:none}.sm-dox.sm-vertical ul a{padding:10px 20px}.sm-dox.sm-vertical ul a:hover,.sm-dox.sm-vertical ul a:focus,.sm-dox.sm-vertical ul a:active,.sm-dox.sm-vertical ul a.highlighted{background:#eee}.sm-dox.sm-vertical ul a.disabled{background:#fff}}
\ No newline at end of file

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doxygen-extra.css Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/doxygen.config Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/example.txt Added

@@ -0,0 +1,464 @@
+/*
+ * example.txt
+ *
+ * This file illustrates how to use the IJG code as a subroutine library
+ * to read or write JPEG image files.  You should look at this code in
+ * conjunction with the documentation file libjpeg.txt.
+ *
+ * This code will not do anything useful as-is, but it may be helpful as a
+ * skeleton for constructing routines that call the JPEG library.
+ *
+ * We present these routines in the same coding style used in the JPEG code
+ * (ANSI function definitions, etc); but you are of course free to code your
+ * routines in a different style if you prefer.
+ */
+
+/* This example was part of the original libjpeg documentation and has been
+ * unchanged since 1994.  It is, as described in libjpeg.txt, "heavily
+ * commented skeleton code for calling the JPEG library."  It is not meant to
+ * be compiled as a standalone program, since it has no main() function and
+ * does not compress from/decompress to a real image buffer (corollary:
+ * put_scanline_someplace() is not a real function.)  First-time users of
+ * libjpeg-turbo would be better served by looking at tjexample.c, which uses
+ * the more straightforward TurboJPEG API, or at cjpeg.c and djpeg.c, which are
+ * examples of libjpeg API usage that can be (and are) compiled into standalone
+ * programs.  Note that this example, as well as the examples in cjpeg.c and
+ * djpeg.c, interleave disk I/O with JPEG compression/decompression, so none of
+ * these examples is suitable for benchmarking purposes.
+ */
+
+#include <stdio.h>
+
+/*
+ * Include file for users of JPEG library.
+ * You will need to have included system headers that define at least
+ * the typedefs FILE and size_t before you can include jpeglib.h.
+ * (stdio.h is sufficient on ANSI-conforming systems.)
+ * You may also wish to include "jerror.h".
+ */
+
+#include "jpeglib.h"
+
+/*
+ * <setjmp.h> is used for the optional error recovery mechanism shown in
+ * the second part of the example.
+ */
+
+#include <setjmp.h>
+
+
+
+/******************** JPEG COMPRESSION SAMPLE INTERFACE *******************/
+
+/* This half of the example shows how to feed data into the JPEG compressor.
+ * We present a minimal version that does not worry about refinements such
+ * as error recovery (the JPEG code will just exit() if it gets an error).
+ */
+
+
+/*
+ * IMAGE DATA FORMATS:
+ *
+ * The standard input image format is a rectangular array of pixels, with
+ * each pixel having the same number of "component" values (color channels).
+ * Each pixel row is an array of JSAMPLEs (which typically are unsigned chars).
+ * If you are working with color data, then the color values for each pixel
+ * must be adjacent in the row; for example, R,G,B,R,G,B,R,G,B,... for 24-bit
+ * RGB color.
+ *
+ * For this example, we'll assume that this data structure matches the way
+ * our application has stored the image in memory, so we can just pass a
+ * pointer to our image buffer.  In particular, let's say that the image is
+ * RGB color and is described by:
+ */
+
+extern JSAMPLE *image_buffer;   /* Points to large array of R,G,B-order data */
+extern int image_height;        /* Number of rows in image */
+extern int image_width;         /* Number of columns in image */
+
+
+/*
+ * Sample routine for JPEG compression.  We assume that the target file name
+ * and a compression quality factor are passed in.
+ */
+
+GLOBAL(void)
+write_JPEG_file(char *filename, int quality)
+{
+  /* This struct contains the JPEG compression parameters and pointers to
+   * working space (which is allocated as needed by the JPEG library).
+   * It is possible to have several such structures, representing multiple
+   * compression/decompression processes, in existence at once.  We refer
+   * to any one struct (and its associated working data) as a "JPEG object".
+   */
+  struct jpeg_compress_struct cinfo;
+  /* This struct represents a JPEG error handler.  It is declared separately
+   * because applications often want to supply a specialized error handler
+   * (see the second half of this file for an example).  But here we just
+   * take the easy way out and use the standard error handler, which will
+   * print a message on stderr and call exit() if compression fails.
+   * Note that this struct must live as long as the main JPEG parameter
+   * struct, to avoid dangling-pointer problems.
+   */
+  struct jpeg_error_mgr jerr;
+  /* More stuff */
+  FILE *outfile;                /* target file */
+  JSAMPROW row_pointer1;      /* pointer to JSAMPLE rows */
+  int row_stride;               /* physical row width in image buffer */
+
+  /* Step 1: allocate and initialize JPEG compression object */
+
+  /* We have to set up the error handler first, in case the initialization
+   * step fails.  (Unlikely, but it could happen if you are out of memory.)
+   * This routine fills in the contents of struct jerr, and returns jerr's
+   * address which we place into the link field in cinfo.
+   */
+  cinfo.err = jpeg_std_error(&jerr);
+  /* Now we can initialize the JPEG compression object. */
+  jpeg_create_compress(&cinfo);
+
+  /* Step 2: specify data destination (eg, a file) */
+  /* Note: steps 2 and 3 can be done in either order. */
+
+  /* Here we use the library-supplied code to send compressed data to a
+   * stdio stream.  You can also write your own code to do something else.
+   * VERY IMPORTANT: use "b" option to fopen() if you are on a machine that
+   * requires it in order to write binary files.
+   */
+  if ((outfile = fopen(filename, "wb")) == NULL) {
+    fprintf(stderr, "can't open %s\n", filename);
+    exit(1);
+  }
+  jpeg_stdio_dest(&cinfo, outfile);
+
+  /* Step 3: set parameters for compression */
+
+  /* First we supply a description of the input image.
+   * Four fields of the cinfo struct must be filled in:
+   */
+  cinfo.image_width = image_width;      /* image width and height, in pixels */
+  cinfo.image_height = image_height;
+  cinfo.input_components = 3;           /* # of color components per pixel */
+  cinfo.in_color_space = JCS_RGB;       /* colorspace of input image */
+  /* Now use the library's routine to set default compression parameters.
+   * (You must set at least cinfo.in_color_space before calling this,
+   * since the defaults depend on the source color space.)
+   */
+  jpeg_set_defaults(&cinfo);
+  /* Now you can set any non-default parameters you wish to.
+   * Here we just illustrate the use of quality (quantization table) scaling:
+   */
+  jpeg_set_quality(&cinfo, quality, TRUE /* limit to baseline-JPEG values */);
+
+  /* Step 4: Start compressor */
+
+  /* TRUE ensures that we will write a complete interchange-JPEG file.
+   * Pass TRUE unless you are very sure of what you're doing.
+   */
+  jpeg_start_compress(&cinfo, TRUE);
+
+  /* Step 5: while (scan lines remain to be written) */
+  /*           jpeg_write_scanlines(...); */
+
+  /* Here we use the library's state variable cinfo.next_scanline as the
+   * loop counter, so that we don't have to keep track ourselves.
+   * To keep things simple, we pass one scanline per call; you can pass
+   * more if you wish, though.
+   */
+  row_stride = image_width * 3; /* JSAMPLEs per row in image_buffer */
+
+  while (cinfo.next_scanline < cinfo.image_height) {
+    /* jpeg_write_scanlines expects an array of pointers to scanlines.
+     * Here the array is only one element long, but you could pass
+     * more than one scanline at a time if that's more convenient.
+     */
+    row_pointer0 = &image_buffercinfo.next_scanline * row_stride;
+    (void)jpeg_write_scanlines(&cinfo, row_pointer, 1);
+  }
+
+  /* Step 6: Finish compression */
+
+  jpeg_finish_compress(&cinfo);
+  /* After finish_compress, we can close the output file. */
+  fclose(outfile);
+
+  /* Step 7: release JPEG compression object */
+
+  /* This is an important step since it will release a good deal of memory. */
+  jpeg_destroy_compress(&cinfo);
+
+  /* And we're done! */
+}
+
+
+/*
+ * SOME FINE POINTS:
+ *
+ * In the above loop, we ignored the return value of jpeg_write_scanlines,
+ * which is the number of scanlines actually written.  We could get away
+ * with this because we were only relying on the value of cinfo.next_scanline,
+ * which will be incremented correctly.  If you maintain additional loop
+ * variables then you should be careful to increment them properly.
+ * Actually, for output to a stdio stream you needn't worry, because
+ * then jpeg_write_scanlines will write all the lines passed (or else exit
+ * with a fatal error).  Partial writes can only occur if you use a data
+ * destination module that can demand suspension of the compressor.
+ * (If you don't know what that's for, you don't need it.)
+ *
+ * If the compressor requires full-image buffers (for entropy-coding
+ * optimization or a multi-scan JPEG file), it will create temporary
+ * files for anything that doesn't fit within the maximum-memory setting.
+ * (Note that temp files are NOT needed if you use the default parameters.)
+ * On some systems you may need to set up a signal handler to ensure that
+ * temporary files are deleted if the program is interrupted.  See libjpeg.txt.
+ *
+ * Scanlines MUST be supplied in top-to-bottom order if you want your JPEG
+ * files to be compatible with everyone else's.  If you cannot readily read
+ * your data in that order, you'll need an intermediate array to hold the
+ * image.  See rdtarga.c or rdbmp.c for examples of handling bottom-to-top
+ * source data using the JPEG code's internal virtual-array mechanisms.
+ */
+
+
+
+/******************** JPEG DECOMPRESSION SAMPLE INTERFACE *******************/
+
+/* This half of the example shows how to read data from the JPEG decompressor.
+ * It's a bit more refined than the above, in that we show:
+ *   (a) how to modify the JPEG library's standard error-reporting behavior;
+ *   (b) how to allocate workspace using the library's memory manager.
+ *
+ * Just to make this example a little different from the first one, we'll
+ * assume that we do not intend to put the whole image into an in-memory
+ * buffer, but to send it line-by-line someplace else.  We need a one-
+ * scanline-high JSAMPLE array as a work buffer, and we will let the JPEG
+ * memory manager allocate it for us.  This approach is actually quite useful
+ * because we don't need to remember to deallocate the buffer separately: it
+ * will go away automatically when the JPEG object is cleaned up.
+ */
+
+
+/*
+ * ERROR HANDLING:
+ *
+ * The JPEG library's standard error handler (jerror.c) is divided into
+ * several "methods" which you can override individually.  This lets you
+ * adjust the behavior without duplicating a lot of code, which you might
+ * have to update with each future release.
+ *
+ * Our example here shows how to override the "error_exit" method so that
+ * control is returned to the library's caller when a fatal error occurs,
+ * rather than calling exit() as the standard error_exit method does.
+ *
+ * We use C's setjmp/longjmp facility to return control.  This means that the
+ * routine which calls the JPEG library must first execute a setjmp() call to
+ * establish the return point.  We want the replacement error_exit to do a
+ * longjmp().  But we need to make the setjmp buffer accessible to the
+ * error_exit routine.  To do this, we make a private extension of the
+ * standard JPEG error handler object.  (If we were using C++, we'd say we
+ * were making a subclass of the regular error handler.)
+ *
+ * Here's the extended error handler struct:
+ */
+
+struct my_error_mgr {
+  struct jpeg_error_mgr pub;    /* "public" fields */
+
+  jmp_buf setjmp_buffer;        /* for return to caller */
+};
+
+typedef struct my_error_mgr *my_error_ptr;
+
+/*
+ * Here's the routine that will replace the standard error_exit method:
+ */
+
+METHODDEF(void)
+my_error_exit(j_common_ptr cinfo)
+{
+  /* cinfo->err really points to a my_error_mgr struct, so coerce pointer */
+  my_error_ptr myerr = (my_error_ptr)cinfo->err;
+
+  /* Always display the message. */
+  /* We could postpone this until after returning, if we chose. */
+  (*cinfo->err->output_message) (cinfo);
+
+  /* Return control to the setjmp point */
+  longjmp(myerr->setjmp_buffer, 1);
+}
+
+
+METHODDEF(int) do_read_JPEG_file(struct jpeg_decompress_struct *cinfo,
+                                 char *filename);
+
+/*
+ * Sample routine for JPEG decompression.  We assume that the source file name
+ * is passed in.  We want to return 1 on success, 0 on error.
+ */
+
+GLOBAL(int)
+read_JPEG_file(char *filename)
+{
+  /* This struct contains the JPEG decompression parameters and pointers to
+   * working space (which is allocated as needed by the JPEG library).
+   */
+  struct jpeg_decompress_struct cinfo;
+
+  return do_read_JPEG_file(&cinfo, filename);
+}
+
+/*
+ * We call the libjpeg API from within a separate function, because modifying
+ * the local non-volatile jpeg_decompress_struct instance below the setjmp()
+ * return point and then accessing the instance after setjmp() returns would
+ * return in undefined behavior that may potentially overwrite all or part of
+ * the structure.
+ */
+
+METHODDEF(int)
+do_read_JPEG_file(struct jpeg_decompress_struct *cinfo, char *filename)
+{
+  /* We use our private extension JPEG error handler.
+   * Note that this struct must live as long as the main JPEG parameter
+   * struct, to avoid dangling-pointer problems.
+   */
+  struct my_error_mgr jerr;
+  /* More stuff */
+  FILE *infile;                 /* source file */
+  JSAMPARRAY buffer;            /* Output row buffer */
+  int row_stride;               /* physical row width in output buffer */
+
+  /* In this example we want to open the input file before doing anything else,
+   * so that the setjmp() error recovery below can assume the file is open.
+   * VERY IMPORTANT: use "b" option to fopen() if you are on a machine that
+   * requires it in order to read binary files.
+   */
+
+  if ((infile = fopen(filename, "rb")) == NULL) {
+    fprintf(stderr, "can't open %s\n", filename);
+    return 0;
+  }
+
+  /* Step 1: allocate and initialize JPEG decompression object */
+
+  /* We set up the normal JPEG error routines, then override error_exit. */
+  cinfo->err = jpeg_std_error(&jerr.pub);
+  jerr.pub.error_exit = my_error_exit;
+  /* Establish the setjmp return context for my_error_exit to use. */
+  if (setjmp(jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error.
+     * We need to clean up the JPEG object, close the input file, and return.
+     */
+    jpeg_destroy_decompress(cinfo);
+    fclose(infile);
+    return 0;
+  }
+  /* Now we can initialize the JPEG decompression object. */
+  jpeg_create_decompress(cinfo);
+
+  /* Step 2: specify data source (eg, a file) */
+
+  jpeg_stdio_src(cinfo, infile);
+
+  /* Step 3: read file parameters with jpeg_read_header() */
+
+  (void)jpeg_read_header(cinfo, TRUE);
+  /* We can ignore the return value from jpeg_read_header since
+   *   (a) suspension is not possible with the stdio data source, and
+   *   (b) we passed TRUE to reject a tables-only JPEG file as an error.
+   * See libjpeg.txt for more info.
+   */
+
+  /* Step 4: set parameters for decompression */
+
+  /* In this example, we don't need to change any of the defaults set by
+   * jpeg_read_header(), so we do nothing here.
+   */
+
+  /* Step 5: Start decompressor */
+
+  (void)jpeg_start_decompress(cinfo);
+  /* We can ignore the return value since suspension is not possible
+   * with the stdio data source.
+   */
+
+  /* We may need to do some setup of our own at this point before reading
+   * the data.  After jpeg_start_decompress() we have the correct scaled
+   * output image dimensions available, as well as the output colormap
+   * if we asked for color quantization.
+   * In this example, we need to make an output work buffer of the right size.
+   */
+  /* JSAMPLEs per row in output buffer */
+  row_stride = cinfo->output_width * cinfo->output_components;
+  /* Make a one-row-high sample array that will go away when done with image */
+  buffer = (*cinfo->mem->alloc_sarray)
+                ((j_common_ptr)cinfo, JPOOL_IMAGE, row_stride, 1);
+
+  /* Step 6: while (scan lines remain to be read) */
+  /*           jpeg_read_scanlines(...); */
+
+  /* Here we use the library's state variable cinfo->output_scanline as the
+   * loop counter, so that we don't have to keep track ourselves.
+   */
+  while (cinfo->output_scanline < cinfo->output_height) {
+    /* jpeg_read_scanlines expects an array of pointers to scanlines.
+     * Here the array is only one element long, but you could ask for
+     * more than one scanline at a time if that's more convenient.
+     */
+    (void)jpeg_read_scanlines(cinfo, buffer, 1);
+    /* Assume put_scanline_someplace wants a pointer and sample count. */
+    put_scanline_someplace(buffer0, row_stride);
+  }
+
+  /* Step 7: Finish decompression */
+
+  (void)jpeg_finish_decompress(cinfo);
+  /* We can ignore the return value since suspension is not possible
+   * with the stdio data source.
+   */
+
+  /* Step 8: Release JPEG decompression object */
+
+  /* This is an important step since it will release a good deal of memory. */
+  jpeg_destroy_decompress(cinfo);
+
+  /* After finish_decompress, we can close the input file.
+   * Here we postpone it until after no more JPEG errors are possible,
+   * so as to simplify the setjmp error logic above.  (Actually, I don't
+   * think that jpeg_destroy can do an error exit, but why assume anything...)
+   */
+  fclose(infile);
+
+  /* At this point you may want to check to see whether any corrupt-data
+   * warnings occurred (test whether jerr.pub.num_warnings is nonzero).
+   */
+
+  /* And we're done! */
+  return 1;
+}
+
+
+/*
+ * SOME FINE POINTS:
+ *
+ * In the above code, we ignored the return value of jpeg_read_scanlines,
+ * which is the number of scanlines actually read.  We could get away with
+ * this because we asked for only one line at a time and we weren't using
+ * a suspending data source.  See libjpeg.txt for more info.
+ *
+ * We cheated a bit by calling alloc_sarray() after jpeg_start_decompress();
+ * we should have done it beforehand to ensure that the space would be
+ * counted against the JPEG max_memory setting.  In some systems the above
+ * code would risk an out-of-memory error.  However, in general we don't
+ * know the output image dimensions before jpeg_start_decompress(), unless we
+ * call jpeg_calc_output_dimensions().  See libjpeg.txt for more about this.
+ *
+ * Scanlines are returned in the same order as they appear in the JPEG file,
+ * which is standardly top-to-bottom.  If you must emit data bottom-to-top,
+ * you can use one of the virtual arrays provided by the JPEG memory manager
+ * to invert the data.  See wrbmp.c for an example.
+ *
+ * As with compression, some operating modes may require temporary files.
+ * On some systems you may need to set up a signal handler to ensure that
+ * temporary files are deleted if the program is interrupted.  See libjpeg.txt.
+ */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/fuzz Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/fuzz/CMakeLists.txt Added

@@ -0,0 +1,51 @@
+if(NOT ENABLE_STATIC)
+  message(FATAL_ERROR "Fuzz targets require static libraries.")
+endif()
+if(NOT WITH_TURBOJPEG)
+  message(FATAL_ERROR "Fuzz targets require the TurboJPEG API library.")
+endif()
+
+set(FUZZ_BINDIR "" CACHE PATH
+  "Directory into which fuzz targets should be installed")
+if(NOT FUZZ_BINDIR)
+  message(FATAL_ERROR "FUZZ_BINDIR must be specified.")
+endif()
+message(STATUS "FUZZ_BINDIR = ${FUZZ_BINDIR}")
+
+set(FUZZ_LIBRARY "" CACHE STRING
+  "Path to fuzzer library or flags necessary to link with it")
+if(NOT FUZZ_LIBRARY)
+  message(FATAL_ERROR "FUZZ_LIBRARY must be specified.")
+endif()
+message(STATUS "FUZZ_LIBRARY = ${FUZZ_LIBRARY}")
+
+enable_language(CXX)
+
+set(EFFECTIVE_CXX_FLAGS
+  "${CMAKE_CXX_FLAGS} ${CMAKE_CXX_FLAGS_${CMAKE_BUILD_TYPE_UC}}")
+message(STATUS "C++ Compiler flags = ${EFFECTIVE_CXX_FLAGS}")
+
+add_executable(cjpeg_fuzzer cjpeg.cc ../cdjpeg.c ../rdbmp.c ../rdgif.c
+  ../rdppm.c ../rdswitch.c ../rdtarga.c)
+set_property(TARGET cjpeg_fuzzer PROPERTY COMPILE_FLAGS ${COMPILE_FLAGS})
+target_link_libraries(cjpeg_fuzzer ${FUZZ_LIBRARY} jpeg-static)
+install(TARGETS cjpeg_fuzzer RUNTIME DESTINATION ${FUZZ_BINDIR})
+
+macro(add_fuzz_target target source_file)
+  add_executable(${target}_fuzzer ${source_file})
+  target_link_libraries(${target}_fuzzer ${FUZZ_LIBRARY} turbojpeg-static)
+  install(TARGETS ${target}_fuzzer RUNTIME DESTINATION ${FUZZ_BINDIR})
+endmacro()
+
+add_fuzz_target(compress compress.cc)
+
+add_fuzz_target(compress_yuv compress_yuv.cc)
+
+# NOTE: This target is named libjpeg_turbo_fuzzer instead of decompress_fuzzer
+# in order to preserve the corpora from Google's OSS-Fuzz target for
+# libjpeg-turbo, which this target replaces.
+add_fuzz_target(libjpeg_turbo decompress.cc)
+
+add_fuzz_target(decompress_yuv decompress_yuv.cc)
+
+add_fuzz_target(transform transform.cc)

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/fuzz/build.sh Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/fuzz/cjpeg.cc Added

@@ -0,0 +1,89 @@
+/*
+ * Copyright (C)2021 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* This fuzz target wraps cjpeg in order to test esoteric compression options
+   as well as the GIF and Targa readers. */
+
+#define main  cjpeg_main
+#define CJPEG_FUZZER
+extern "C" {
+#include "../cjpeg.c"
+}
+#undef main
+#undef CJPEG_FUZZER
+
+#include <stdint.h>
+#include <unistd.h>
+
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
+{
+  char filenameFILENAME_MAX = { 0 };
+  char *argv1 = {
+    (char *)"cjpeg", (char *)"-dct", (char *)"float", (char *)"-memdst",
+    (char *)"-optimize", (char *)"-quality", (char *)"100,99,98",
+    (char *)"-restart", (char *)"2", (char *)"-sample", (char *)"4x1,2x2,1x2",
+    (char *)"-targa", NULL
+  };
+  char *argv2 = {
+    (char *)"cjpeg", (char *)"-arithmetic", (char *)"-dct", (char *)"float",
+    (char *)"-memdst", (char *)"-quality", (char *)"90,80,70", (char *)"-rgb",
+    (char *)"-sample", (char *)"2x2", (char *)"-smooth", (char *)"50",
+    (char *)"-targa", NULL
+  };
+  int fd = -1;
+#if defined(__has_feature) && __has_feature(memory_sanitizer)
+  char env18 = "JSIMD_FORCENONE=1";
+
+  /* The libjpeg-turbo SIMD extensions produce false positives with
+     MemorySanitizer. */
+  putenv(env);
+#endif
+
+  snprintf(filename, FILENAME_MAX, "/tmp/libjpeg-turbo_cjpeg_fuzz.XXXXXX");
+  if ((fd = mkstemp(filename)) < 0 || write(fd, data, size) < 0)
+    goto bailout;
+
+  argv112 = argv213 = filename;
+
+  cjpeg_main(13, argv1);
+  cjpeg_main(14, argv2);
+
+  argv112 = argv213 = NULL;
+  argv111 = argv212 = filename;
+
+  cjpeg_main(12, argv1);
+  cjpeg_main(13, argv2);
+
+bailout:
+  if (fd >= 0) {
+    close(fd);
+    if (strlen(filename) > 0) unlink(filename);
+  }
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/fuzz/compress.cc Added

@@ -0,0 +1,133 @@
+/*
+ * Copyright (C)2021 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <turbojpeg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+#include <unistd.h>
+
+
+#define NUMTESTS  7
+/* Private flag that triggers different TurboJPEG API behavior when fuzzing */
+#define TJFLAG_FUZZING  (1 << 30)
+
+
+struct test {
+  enum TJPF pf;
+  enum TJSAMP subsamp;
+  int quality;
+};
+
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
+{
+  tjhandle handle = NULL;
+  unsigned char *srcBuf = NULL, *dstBuf = NULL;
+  int width = 0, height = 0, fd = -1, i, ti;
+  char filenameFILENAME_MAX = { 0 };
+  struct test testsNUMTESTS = {
+    { TJPF_RGB, TJSAMP_444, 100 },
+    { TJPF_BGR, TJSAMP_422, 90 },
+    { TJPF_RGBX, TJSAMP_420, 80 },
+    { TJPF_BGRA, TJSAMP_411, 70 },
+    { TJPF_XRGB, TJSAMP_GRAY, 60 },
+    { TJPF_GRAY, TJSAMP_GRAY, 50 },
+    { TJPF_CMYK, TJSAMP_440, 40 }
+  };
+#if defined(__has_feature) && __has_feature(memory_sanitizer)
+  char env18 = "JSIMD_FORCENONE=1";
+
+  /* The libjpeg-turbo SIMD extensions produce false positives with
+     MemorySanitizer. */
+  putenv(env);
+#endif
+
+  snprintf(filename, FILENAME_MAX, "/tmp/libjpeg-turbo_compress_fuzz.XXXXXX");
+  if ((fd = mkstemp(filename)) < 0 || write(fd, data, size) < 0)
+    goto bailout;
+
+  if ((handle = tjInitCompress()) == NULL)
+    goto bailout;
+
+  for (ti = 0; ti < NUMTESTS; ti++) {
+    int flags = TJFLAG_FUZZING, sum = 0, pf = teststi.pf;
+    unsigned long dstSize = 0, maxBufSize;
+
+    /* Test non-default compression options on specific iterations. */
+    if (ti == 0)
+      flags |= TJFLAG_BOTTOMUP | TJFLAG_ACCURATEDCT;
+    else if (ti == 1)
+      flags |= TJFLAG_PROGRESSIVE;
+    if (ti != 2)
+      flags |= TJFLAG_NOREALLOC;
+
+    /* tjLoadImage() refuses to load images larger than 1 Megapixel when
+       FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION is defined (yes, that's a dirty
+       hack), so we don't need to check the width and height here. */
+    if ((srcBuf = tjLoadImage(filename, &width, 1, &height, &pf,
+                              flags)) == NULL)
+      continue;
+
+    maxBufSize = tjBufSize(width, height, teststi.subsamp);
+    if (flags & TJFLAG_NOREALLOC) {
+      if ((dstBuf = (unsigned char *)malloc(maxBufSize)) == NULL)
+        goto bailout;
+    } else
+      dstBuf = NULL;
+
+    if (tjCompress2(handle, srcBuf, width, 0, height, pf, &dstBuf, &dstSize,
+                    teststi.subsamp, teststi.quality, flags) == 0) {
+      /* Touch all of the output pixels in order to catch uninitialized reads
+         when using MemorySanitizer. */
+      for (i = 0; i < dstSize; i++)
+        sum += dstBufi;
+    }
+
+    free(dstBuf);
+    dstBuf = NULL;
+    tjFree(srcBuf);
+    srcBuf = NULL;
+
+    /* Prevent the code above from being optimized out.  This test should never
+       be true, but the compiler doesn't know that. */
+    if (sum > 255 * maxBufSize)
+      goto bailout;
+  }
+
+bailout:
+  free(dstBuf);
+  tjFree(srcBuf);
+  if (fd >= 0) {
+    close(fd);
+    if (strlen(filename) > 0) unlink(filename);
+  }
+  if (handle) tjDestroy(handle);
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/fuzz/compress_yuv.cc Added

@@ -0,0 +1,148 @@
+/*
+ * Copyright (C)2021 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <turbojpeg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+#include <unistd.h>
+
+
+#define NUMTESTS  6
+/* Private flag that triggers different TurboJPEG API behavior when fuzzing */
+#define TJFLAG_FUZZING  (1 << 30)
+
+
+struct test {
+  enum TJPF pf;
+  enum TJSAMP subsamp;
+  int quality;
+};
+
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
+{
+  tjhandle handle = NULL;
+  unsigned char *srcBuf = NULL, *dstBuf = NULL, *yuvBuf = NULL;
+  int width = 0, height = 0, fd = -1, i, ti;
+  char filenameFILENAME_MAX = { 0 };
+  struct test testsNUMTESTS = {
+    { TJPF_XBGR, TJSAMP_444, 100 },
+    { TJPF_XRGB, TJSAMP_422, 90 },
+    { TJPF_BGR, TJSAMP_420, 80 },
+    { TJPF_RGB, TJSAMP_411, 70 },
+    { TJPF_BGR, TJSAMP_GRAY, 60 },
+    { TJPF_GRAY, TJSAMP_GRAY, 50 }
+  };
+  char arithEnv16 = "TJ_ARITHMETIC=0";
+  char restartEnv13 = "TJ_RESTART=0";
+#if defined(__has_feature) && __has_feature(memory_sanitizer)
+  char simdEnv18 = "JSIMD_FORCENONE=1";
+
+  /* The libjpeg-turbo SIMD extensions produce false positives with
+     MemorySanitizer. */
+  putenv(simdEnv);
+#endif
+  putenv(arithEnv);
+  putenv(restartEnv);
+
+  snprintf(filename, FILENAME_MAX, "/tmp/libjpeg-turbo_compress_yuv_fuzz.XXXXXX");
+  if ((fd = mkstemp(filename)) < 0 || write(fd, data, size) < 0)
+    goto bailout;
+
+  if ((handle = tjInitCompress()) == NULL)
+    goto bailout;
+
+  for (ti = 0; ti < NUMTESTS; ti++) {
+    int flags = TJFLAG_FUZZING | TJFLAG_NOREALLOC, sum = 0, pf = teststi.pf;
+    unsigned long dstSize = 0, maxBufSize;
+
+    /* Test non-default compression options on specific iterations. */
+    if (ti == 0)
+      flags |= TJFLAG_BOTTOMUP | TJFLAG_ACCURATEDCT;
+    else if (ti == 1 || ti == 3)
+      flags |= TJFLAG_PROGRESSIVE;
+    if (ti == 2 || ti == 3)
+      arithEnv14 = '1';
+    else
+      arithEnv14 = '0';
+    if (ti == 1 || ti == 2)
+      restartEnv11 = '2';
+    else
+      restartEnv11 = '0';
+
+    /* tjLoadImage() refuses to load images larger than 1 Megapixel when
+       FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION is defined (yes, that's a dirty
+       hack), so we don't need to check the width and height here. */
+    if ((srcBuf = tjLoadImage(filename, &width, 1, &height, &pf,
+                              flags)) == NULL)
+      continue;
+
+    maxBufSize = tjBufSize(width, height, teststi.subsamp);
+    if ((dstBuf = (unsigned char *)malloc(maxBufSize)) == NULL)
+      goto bailout;
+    if ((yuvBuf =
+         (unsigned char *)malloc(tjBufSizeYUV2(width, 1, height,
+                                               teststi.subsamp))) == NULL)
+      goto bailout;
+
+    if (tjEncodeYUV3(handle, srcBuf, width, 0, height, pf, yuvBuf, 1,
+                     teststi.subsamp, flags) == 0 &&
+        tjCompressFromYUV(handle, yuvBuf, width, 1, height, teststi.subsamp,
+                          &dstBuf, &dstSize, teststi.quality, flags) == 0) {
+      /* Touch all of the output pixels in order to catch uninitialized reads
+         when using MemorySanitizer. */
+      for (i = 0; i < dstSize; i++)
+        sum += dstBufi;
+    }
+
+    free(dstBuf);
+    dstBuf = NULL;
+    free(yuvBuf);
+    yuvBuf = NULL;
+    tjFree(srcBuf);
+    srcBuf = NULL;
+
+    /* Prevent the code above from being optimized out.  This test should never
+       be true, but the compiler doesn't know that. */
+    if (sum > 255 * maxBufSize)
+      goto bailout;
+  }
+
+bailout:
+  free(dstBuf);
+  free(yuvBuf);
+  tjFree(srcBuf);
+  if (fd >= 0) {
+    close(fd);
+    if (strlen(filename) > 0) unlink(filename);
+  }
+  if (handle) tjDestroy(handle);
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/fuzz/decompress.cc Added

@@ -0,0 +1,106 @@
+/*
+ * Copyright (C)2021 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <turbojpeg.h>
+#include <stdlib.h>
+#include <stdint.h>
+
+
+#define NUMPF  4
+
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
+{
+  tjhandle handle = NULL;
+  unsigned char *dstBuf = NULL;
+  int width = 0, height = 0, jpegSubsamp, jpegColorspace, pfi;
+  /* TJPF_RGB-TJPF_BGR share the same code paths, as do TJPF_RGBX-TJPF_XRGB and
+     TJPF_RGBA-TJPF_ARGB.  Thus, the pixel formats below should be the minimum
+     necessary to achieve full coverage. */
+  enum TJPF pixelFormatsNUMPF =
+    { TJPF_RGB, TJPF_BGRX, TJPF_GRAY, TJPF_CMYK };
+#if defined(__has_feature) && __has_feature(memory_sanitizer)
+  char env18 = "JSIMD_FORCENONE=1";
+
+  /* The libjpeg-turbo SIMD extensions produce false positives with
+     MemorySanitizer. */
+  putenv(env);
+#endif
+
+  if ((handle = tjInitDecompress()) == NULL)
+    goto bailout;
+
+  /* We ignore the return value of tjDecompressHeader3(), because some JPEG
+     images may have unusual subsampling configurations that the TurboJPEG API
+     cannot identify but can still decompress. */
+  tjDecompressHeader3(handle, data, size, &width, &height, &jpegSubsamp,
+                      &jpegColorspace);
+
+  /* Ignore 0-pixel images and images larger than 1 Megapixel, as Google's
+     OSS-Fuzz target for libjpeg-turbo did.  Casting width to (uint64_t)
+     prevents integer overflow if width * height > INT_MAX. */
+  if (width < 1 || height < 1 || (uint64_t)width * height > 1048576)
+    goto bailout;
+
+  for (pfi = 0; pfi < NUMPF; pfi++) {
+    int pf = pixelFormatspfi, flags = TJFLAG_LIMITSCANS, i, sum = 0;
+    int w = width, h = height;
+
+    /* Test non-default decompression options on the first iteration. */
+    if (pfi == 0)
+      flags |= TJFLAG_BOTTOMUP | TJFLAG_FASTUPSAMPLE | TJFLAG_FASTDCT;
+    /* Test IDCT scaling on the second iteration. */
+    else if (pfi == 1) {
+      w = (width + 1) / 2;
+      h = (height + 1) / 2;
+    }
+
+    if ((dstBuf = (unsigned char *)malloc(w * h * tjPixelSizepf)) == NULL)
+      goto bailout;
+
+    if (tjDecompress2(handle, data, size, dstBuf, w, 0, h, pf, flags) == 0) {
+      /* Touch all of the output pixels in order to catch uninitialized reads
+         when using MemorySanitizer. */
+      for (i = 0; i < w * h * tjPixelSizepf; i++)
+        sum += dstBufi;
+    }
+
+    free(dstBuf);
+    dstBuf = NULL;
+
+    /* Prevent the code above from being optimized out.  This test should never
+       be true, but the compiler doesn't know that. */
+    if (sum > 255 * 1048576 * tjPixelSizepf)
+      goto bailout;
+  }
+
+bailout:
+  free(dstBuf);
+  if (handle) tjDestroy(handle);
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/fuzz/decompress_yuv.cc Added

@@ -0,0 +1,111 @@
+/*
+ * Copyright (C)2021 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <turbojpeg.h>
+#include <stdlib.h>
+#include <stdint.h>
+
+
+#define NUMPF  3
+
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
+{
+  tjhandle handle = NULL;
+  unsigned char *dstBuf = NULL, *yuvBuf = NULL;
+  int width = 0, height = 0, jpegSubsamp, jpegColorspace, pfi;
+  /* TJPF_RGB-TJPF_BGR share the same code paths, as do TJPF_RGBX-TJPF_XRGB and
+     TJPF_RGBA-TJPF_ARGB.  Thus, the pixel formats below should be the minimum
+     necessary to achieve full coverage. */
+  enum TJPF pixelFormatsNUMPF =
+    { TJPF_BGR, TJPF_XRGB, TJPF_GRAY };
+#if defined(__has_feature) && __has_feature(memory_sanitizer)
+  char env18 = "JSIMD_FORCENONE=1";
+
+  /* The libjpeg-turbo SIMD extensions produce false positives with
+     MemorySanitizer. */
+  putenv(env);
+#endif
+
+  if ((handle = tjInitDecompress()) == NULL)
+    goto bailout;
+
+  if (tjDecompressHeader3(handle, data, size, &width, &height, &jpegSubsamp,
+                          &jpegColorspace) < 0)
+    goto bailout;
+
+  /* Ignore 0-pixel images and images larger than 1 Megapixel.  Casting width
+     to (uint64_t) prevents integer overflow if width * height > INT_MAX. */
+  if (width < 1 || height < 1 || (uint64_t)width * height > 1048576)
+    goto bailout;
+
+  for (pfi = 0; pfi < NUMPF; pfi++) {
+    int pf = pixelFormatspfi, flags = TJFLAG_LIMITSCANS, i, sum = 0;
+    int w = width, h = height;
+
+    /* Test non-default decompression options on the first iteration. */
+    if (pfi == 0)
+      flags |= TJFLAG_BOTTOMUP | TJFLAG_FASTUPSAMPLE | TJFLAG_FASTDCT;
+    /* Test IDCT scaling on the second iteration. */
+    else if (pfi == 1) {
+      w = (width + 3) / 4;
+      h = (height + 3) / 4;
+    }
+
+    if ((dstBuf = (unsigned char *)malloc(w * h * tjPixelSizepf)) == NULL)
+      goto bailout;
+    if ((yuvBuf =
+         (unsigned char *)malloc(tjBufSizeYUV2(w, 1, h, jpegSubsamp))) == NULL)
+      goto bailout;
+
+    if (tjDecompressToYUV2(handle, data, size, yuvBuf, w, 1, h, flags) == 0 &&
+        tjDecodeYUV(handle, yuvBuf, 1, jpegSubsamp, dstBuf, w, 0, h, pf,
+                    flags) == 0) {
+      /* Touch all of the output pixels in order to catch uninitialized reads
+         when using MemorySanitizer. */
+      for (i = 0; i < w * h * tjPixelSizepf; i++)
+        sum += dstBufi;
+    }
+
+    free(dstBuf);
+    dstBuf = NULL;
+    free(yuvBuf);
+    yuvBuf = NULL;
+
+    /* Prevent the code above from being optimized out.  This test should never
+       be true, but the compiler doesn't know that. */
+    if (sum > 255 * 1048576 * tjPixelSizepf)
+      goto bailout;
+  }
+
+bailout:
+  free(dstBuf);
+  free(yuvBuf);
+  if (handle) tjDestroy(handle);
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/fuzz/transform.cc Added

@@ -0,0 +1,135 @@
+/*
+ * Copyright (C)2021 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <turbojpeg.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+
+
+#define NUMXFORMS  3
+
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
+{
+  tjhandle handle = NULL;
+  unsigned char *dstBufsNUMXFORMS = { NULL, NULL, NULL };
+  unsigned long dstSizesNUMXFORMS = { 0, 0, 0 }, maxBufSize;
+  int width = 0, height = 0, jpegSubsamp, jpegColorspace, i, t;
+  tjtransform transformsNUMXFORMS;
+#if defined(__has_feature) && __has_feature(memory_sanitizer)
+  char env18 = "JSIMD_FORCENONE=1";
+
+  /* The libjpeg-turbo SIMD extensions produce false positives with
+     MemorySanitizer. */
+  putenv(env);
+#endif
+
+  if ((handle = tjInitTransform()) == NULL)
+    goto bailout;
+
+  /* We ignore the return value of tjDecompressHeader3(), because some JPEG
+     images may have unusual subsampling configurations that the TurboJPEG API
+     cannot identify but can still transform. */
+  tjDecompressHeader3(handle, data, size, &width, &height, &jpegSubsamp,
+                      &jpegColorspace);
+
+  /* Ignore 0-pixel images and images larger than 1 Megapixel.  Casting width
+     to (uint64_t) prevents integer overflow if width * height > INT_MAX. */
+  if (width < 1 || height < 1 || (uint64_t)width * height > 1048576)
+    goto bailout;
+
+  if (jpegSubsamp < 0 || jpegSubsamp >= TJ_NUMSAMP)
+    jpegSubsamp = TJSAMP_444;
+
+  for (t = 0; t < NUMXFORMS; t++)
+    memset(&transformst, 0, sizeof(tjtransform));
+
+  transforms0.op = TJXOP_NONE;
+  transforms0.options = TJXOPT_PROGRESSIVE | TJXOPT_COPYNONE;
+  dstBufs0 = (unsigned char *)malloc(tjBufSize(width, height, jpegSubsamp));
+  if (!dstBufs0)
+    goto bailout;
+
+  transforms1.r.w = (width + 1) / 2;
+  transforms1.r.h = (height + 1) / 2;
+  transforms1.op = TJXOP_TRANSPOSE;
+  transforms1.options = TJXOPT_GRAY | TJXOPT_CROP | TJXOPT_COPYNONE;
+  dstBufs1 =
+    (unsigned char *)malloc(tjBufSize((width + 1) / 2, (height + 1) / 2,
+                                      TJSAMP_GRAY));
+  if (!dstBufs1)
+    goto bailout;
+
+  transforms2.op = TJXOP_ROT90;
+  transforms2.options = TJXOPT_TRIM | TJXOPT_COPYNONE;
+  dstBufs2 = (unsigned char *)malloc(tjBufSize(height, width, jpegSubsamp));
+  if (!dstBufs2)
+    goto bailout;
+
+  maxBufSize = tjBufSize(width, height, jpegSubsamp);
+
+  if (tjTransform(handle, data, size, NUMXFORMS, dstBufs, dstSizes, transforms,
+                  TJFLAG_LIMITSCANS | TJFLAG_NOREALLOC) == 0) {
+    /* Touch all of the output pixels in order to catch uninitialized reads
+       when using MemorySanitizer. */
+    for (t = 0; t < NUMXFORMS; t++) {
+      int sum = 0;
+
+      for (i = 0; i < dstSizest; i++)
+        sum += dstBufsti;
+
+      /* Prevent the code above from being optimized out.  This test should
+         never be true, but the compiler doesn't know that. */
+      if (sum > 255 * maxBufSize)
+        goto bailout;
+    }
+  }
+
+  transforms0.options &= ~TJXOPT_COPYNONE;
+  free(dstBufs0);
+  dstBufs0 = NULL;
+  dstSizes0 = 0;
+
+  if (tjTransform(handle, data, size, 1, dstBufs, dstSizes, transforms,
+                  TJFLAG_LIMITSCANS) == 0) {
+    int sum = 0;
+
+    for (i = 0; i < dstSizes0; i++)
+      sum += dstBufs0i;
+
+    if (sum > 255 * maxBufSize)
+      goto bailout;
+  }
+
+bailout:
+  for (t = 0; t < NUMXFORMS; t++)
+    free(dstBufst);
+  if (handle) tjDestroy(handle);
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jaricom.c Added

@@ -0,0 +1,157 @@
+/*
+ * jaricom.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Developed 1997-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, 2018, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains probability estimation tables for common use in
+ * arithmetic entropy encoding and decoding routines.
+ *
+ * This data represents Table D.2 in
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994 and Table 24 in
+ * Recommendation ITU-T T.82 (1993) | ISO/IEC 11544:1993.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+/* The following #define specifies the packing of the four components
+ * into the compact JLONG representation.
+ * Note that this formula must match the actual arithmetic encoder
+ * and decoder implementation.  The implementation has to be changed
+ * if this formula is changed.
+ * The current organization is leaned on Markus Kuhn's JBIG
+ * implementation (jbig_tab.c).
+ */
+
+#define V(i, a, b, c, d) \
+  (((JLONG)a << 16) | ((JLONG)c << 8) | ((JLONG)d << 7) | b)
+
+const JLONG jpeg_aritab113 + 1 = {
+/*
+ * Index, Qe_Value, Next_Index_LPS, Next_Index_MPS, Switch_MPS
+ */
+  V(   0, 0x5a1d,   1,   1, 1 ),
+  V(   1, 0x2586,  14,   2, 0 ),
+  V(   2, 0x1114,  16,   3, 0 ),
+  V(   3, 0x080b,  18,   4, 0 ),
+  V(   4, 0x03d8,  20,   5, 0 ),
+  V(   5, 0x01da,  23,   6, 0 ),
+  V(   6, 0x00e5,  25,   7, 0 ),
+  V(   7, 0x006f,  28,   8, 0 ),
+  V(   8, 0x0036,  30,   9, 0 ),
+  V(   9, 0x001a,  33,  10, 0 ),
+  V(  10, 0x000d,  35,  11, 0 ),
+  V(  11, 0x0006,   9,  12, 0 ),
+  V(  12, 0x0003,  10,  13, 0 ),
+  V(  13, 0x0001,  12,  13, 0 ),
+  V(  14, 0x5a7f,  15,  15, 1 ),
+  V(  15, 0x3f25,  36,  16, 0 ),
+  V(  16, 0x2cf2,  38,  17, 0 ),
+  V(  17, 0x207c,  39,  18, 0 ),
+  V(  18, 0x17b9,  40,  19, 0 ),
+  V(  19, 0x1182,  42,  20, 0 ),
+  V(  20, 0x0cef,  43,  21, 0 ),
+  V(  21, 0x09a1,  45,  22, 0 ),
+  V(  22, 0x072f,  46,  23, 0 ),
+  V(  23, 0x055c,  48,  24, 0 ),
+  V(  24, 0x0406,  49,  25, 0 ),
+  V(  25, 0x0303,  51,  26, 0 ),
+  V(  26, 0x0240,  52,  27, 0 ),
+  V(  27, 0x01b1,  54,  28, 0 ),
+  V(  28, 0x0144,  56,  29, 0 ),
+  V(  29, 0x00f5,  57,  30, 0 ),
+  V(  30, 0x00b7,  59,  31, 0 ),
+  V(  31, 0x008a,  60,  32, 0 ),
+  V(  32, 0x0068,  62,  33, 0 ),
+  V(  33, 0x004e,  63,  34, 0 ),
+  V(  34, 0x003b,  32,  35, 0 ),
+  V(  35, 0x002c,  33,   9, 0 ),
+  V(  36, 0x5ae1,  37,  37, 1 ),
+  V(  37, 0x484c,  64,  38, 0 ),
+  V(  38, 0x3a0d,  65,  39, 0 ),
+  V(  39, 0x2ef1,  67,  40, 0 ),
+  V(  40, 0x261f,  68,  41, 0 ),
+  V(  41, 0x1f33,  69,  42, 0 ),
+  V(  42, 0x19a8,  70,  43, 0 ),
+  V(  43, 0x1518,  72,  44, 0 ),
+  V(  44, 0x1177,  73,  45, 0 ),
+  V(  45, 0x0e74,  74,  46, 0 ),
+  V(  46, 0x0bfb,  75,  47, 0 ),
+  V(  47, 0x09f8,  77,  48, 0 ),
+  V(  48, 0x0861,  78,  49, 0 ),
+  V(  49, 0x0706,  79,  50, 0 ),
+  V(  50, 0x05cd,  48,  51, 0 ),
+  V(  51, 0x04de,  50,  52, 0 ),
+  V(  52, 0x040f,  50,  53, 0 ),
+  V(  53, 0x0363,  51,  54, 0 ),
+  V(  54, 0x02d4,  52,  55, 0 ),
+  V(  55, 0x025c,  53,  56, 0 ),
+  V(  56, 0x01f8,  54,  57, 0 ),
+  V(  57, 0x01a4,  55,  58, 0 ),
+  V(  58, 0x0160,  56,  59, 0 ),
+  V(  59, 0x0125,  57,  60, 0 ),
+  V(  60, 0x00f6,  58,  61, 0 ),
+  V(  61, 0x00cb,  59,  62, 0 ),
+  V(  62, 0x00ab,  61,  63, 0 ),
+  V(  63, 0x008f,  61,  32, 0 ),
+  V(  64, 0x5b12,  65,  65, 1 ),
+  V(  65, 0x4d04,  80,  66, 0 ),
+  V(  66, 0x412c,  81,  67, 0 ),
+  V(  67, 0x37d8,  82,  68, 0 ),
+  V(  68, 0x2fe8,  83,  69, 0 ),
+  V(  69, 0x293c,  84,  70, 0 ),
+  V(  70, 0x2379,  86,  71, 0 ),
+  V(  71, 0x1edf,  87,  72, 0 ),
+  V(  72, 0x1aa9,  87,  73, 0 ),
+  V(  73, 0x174e,  72,  74, 0 ),
+  V(  74, 0x1424,  72,  75, 0 ),
+  V(  75, 0x119c,  74,  76, 0 ),
+  V(  76, 0x0f6b,  74,  77, 0 ),
+  V(  77, 0x0d51,  75,  78, 0 ),
+  V(  78, 0x0bb6,  77,  79, 0 ),
+  V(  79, 0x0a40,  77,  48, 0 ),
+  V(  80, 0x5832,  80,  81, 1 ),
+  V(  81, 0x4d1c,  88,  82, 0 ),
+  V(  82, 0x438e,  89,  83, 0 ),
+  V(  83, 0x3bdd,  90,  84, 0 ),
+  V(  84, 0x34ee,  91,  85, 0 ),
+  V(  85, 0x2eae,  92,  86, 0 ),
+  V(  86, 0x299a,  93,  87, 0 ),
+  V(  87, 0x2516,  86,  71, 0 ),
+  V(  88, 0x5570,  88,  89, 1 ),
+  V(  89, 0x4ca9,  95,  90, 0 ),
+  V(  90, 0x44d9,  96,  91, 0 ),
+  V(  91, 0x3e22,  97,  92, 0 ),
+  V(  92, 0x3824,  99,  93, 0 ),
+  V(  93, 0x32b4,  99,  94, 0 ),
+  V(  94, 0x2e17,  93,  86, 0 ),
+  V(  95, 0x56a8,  95,  96, 1 ),
+  V(  96, 0x4f46, 101,  97, 0 ),
+  V(  97, 0x47e5, 102,  98, 0 ),
+  V(  98, 0x41cf, 103,  99, 0 ),
+  V(  99, 0x3c3d, 104, 100, 0 ),
+  V( 100, 0x375e,  99,  93, 0 ),
+  V( 101, 0x5231, 105, 102, 0 ),
+  V( 102, 0x4c0f, 106, 103, 0 ),
+  V( 103, 0x4639, 107, 104, 0 ),
+  V( 104, 0x415e, 103,  99, 0 ),
+  V( 105, 0x5627, 105, 106, 1 ),
+  V( 106, 0x50e7, 108, 107, 0 ),
+  V( 107, 0x4b85, 109, 103, 0 ),
+  V( 108, 0x5597, 110, 109, 0 ),
+  V( 109, 0x504f, 111, 107, 0 ),
+  V( 110, 0x5a10, 110, 111, 1 ),
+  V( 111, 0x5522, 112, 109, 0 ),
+  V( 112, 0x59eb, 112, 111, 1 ),
+/*
+ * This last entry is used for fixed probability estimate of 0.5
+ * as recommended in Section 10.3 Table 5 of ITU-T Rec. T.851.
+ */
+  V( 113, 0x5a1d, 113, 113, 0 )
+};

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/CMakeLists.txt Added

@@ -0,0 +1,88 @@
+find_package(Java REQUIRED)
+find_package(JNI REQUIRED)
+
+# Allow the Java compiler flags to be set using an environment variable
+if(NOT DEFINED CMAKE_JAVA_COMPILE_FLAGS AND DEFINED ENV{JAVAFLAGS})
+  set(CMAKE_JAVA_COMPILE_FLAGS $ENV{JAVAFLAGS})
+endif()
+
+include(UseJava)
+
+set(CMAKE_JAVA_COMPILE_FLAGS "${CMAKE_JAVA_COMPILE_FLAGS} -J-Dfile.encoding=UTF8")
+message(STATUS "CMAKE_JAVA_COMPILE_FLAGS = ${CMAKE_JAVA_COMPILE_FLAGS}")
+string(REGEX REPLACE " " ";" CMAKE_JAVA_COMPILE_FLAGS "${CMAKE_JAVA_COMPILE_FLAGS}")
+
+set(JAVAARGS "" CACHE STRING "Additional arguments to pass to java when running unit tests (example: -d32)")
+message(STATUS "JAVAARGS = ${JAVAARGS}")
+
+set(JAVA_SOURCES org/libjpegturbo/turbojpeg/TJ.java
+  org/libjpegturbo/turbojpeg/TJCompressor.java
+  org/libjpegturbo/turbojpeg/TJCustomFilter.java
+  org/libjpegturbo/turbojpeg/TJDecompressor.java
+  org/libjpegturbo/turbojpeg/TJException.java
+  org/libjpegturbo/turbojpeg/TJScalingFactor.java
+  org/libjpegturbo/turbojpeg/TJTransform.java
+  org/libjpegturbo/turbojpeg/TJTransformer.java
+  org/libjpegturbo/turbojpeg/YUVImage.java
+  TJUnitTest.java
+  TJExample.java
+  TJBench.java)
+
+set(TURBOJPEG_DLL_NAME "turbojpeg")
+if(MINGW)
+  set(TURBOJPEG_DLL_NAME "libturbojpeg")
+endif()
+if(WIN32)
+  configure_file(org/libjpegturbo/turbojpeg/TJLoader-win.java.in
+    ${CMAKE_CURRENT_BINARY_DIR}/org/libjpegturbo/turbojpeg/TJLoader.java)
+else()
+  configure_file(org/libjpegturbo/turbojpeg/TJLoader-unix.java.in
+    ${CMAKE_CURRENT_BINARY_DIR}/org/libjpegturbo/turbojpeg/TJLoader.java)
+endif()
+set(JAVA_SOURCES ${JAVA_SOURCES}
+  ${CMAKE_CURRENT_BINARY_DIR}/org/libjpegturbo/turbojpeg/TJLoader.java)
+
+if(MSYS)
+  # UGLY HACK ALERT: If we don't do this, then UseJava.cmake will separate
+  # class path members with a semicolon, which is interpreted as a command
+  # separator by the MSYS shell.
+  set(CMAKE_HOST_SYSTEM_NAME_BAK ${CMAKE_HOST_SYSTEM_NAME})
+  set(CMAKE_HOST_SYSTEM_NAME "MSYS")
+endif()
+add_jar(turbojpeg-java ${JAVA_SOURCES} OUTPUT_NAME turbojpeg
+  ENTRY_POINT TJExample)
+if(MSYS)
+  set(CMAKE_HOST_SYSTEM_NAME ${CMAKE_HOST_SYSTEM_NAME_BAK})
+endif()
+
+add_custom_target(javadoc COMMAND
+  javadoc -notimestamp -d ${CMAKE_CURRENT_SOURCE_DIR}/doc -sourcepath ${CMAKE_CURRENT_SOURCE_DIR} org.libjpegturbo.turbojpeg)
+set(JAVACLASSPATH ${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_FILES_DIRECTORY}/turbojpeg-java.dir)
+if(Java_VERSION_MAJOR GREATER 9)
+  add_custom_target(javah
+    COMMAND javac -h ${CMAKE_CURRENT_SOURCE_DIR} -classpath ${JAVACLASSPATH}
+      -d ${CMAKE_CURRENT_BINARY_DIR}/__unused
+      ${CMAKE_CURRENT_SOURCE_DIR}/org/libjpegturbo/turbojpeg/TJ.java
+      ${CMAKE_CURRENT_SOURCE_DIR}/org/libjpegturbo/turbojpeg/TJCompressor.java
+      ${CMAKE_CURRENT_SOURCE_DIR}/org/libjpegturbo/turbojpeg/TJDecompressor.java
+      ${CMAKE_CURRENT_SOURCE_DIR}/org/libjpegturbo/turbojpeg/TJTransformer.java)
+else()
+  add_custom_target(javah
+    COMMAND javah -d ${CMAKE_CURRENT_SOURCE_DIR} -classpath ${JAVACLASSPATH} org.libjpegturbo.turbojpeg.TJ
+    COMMAND javah -d ${CMAKE_CURRENT_SOURCE_DIR} -classpath ${JAVACLASSPATH} org.libjpegturbo.turbojpeg.TJCompressor
+    COMMAND javah -d ${CMAKE_CURRENT_SOURCE_DIR} -classpath ${JAVACLASSPATH} org.libjpegturbo.turbojpeg.TJDecompressor
+    COMMAND javah -d ${CMAKE_CURRENT_SOURCE_DIR} -classpath ${JAVACLASSPATH} org.libjpegturbo.turbojpeg.TJTransformer)
+endif()
+
+if(NOT DEFINED CMAKE_INSTALL_DEFAULT_JAVADIR)
+  set(CMAKE_INSTALL_DEFAULT_JAVADIR "<CMAKE_INSTALL_DATAROOTDIR>/java")
+endif()
+GNUInstallDirs_set_install_dir(JAVADIR
+  "The directory into which Java classes should be installed")
+GNUInstallDirs_get_absolute_install_dir(CMAKE_INSTALL_FULL_JAVADIR
+  CMAKE_INSTALL_JAVADIR)
+set(CMAKE_INSTALL_JAVADIR ${CMAKE_INSTALL_JAVADIR} PARENT_SCOPE)
+set(CMAKE_INSTALL_FULL_JAVADIR ${CMAKE_INSTALL_FULL_JAVADIR} PARENT_SCOPE)
+report_directory(JAVADIR)
+install_jar(turbojpeg-java ${CMAKE_INSTALL_JAVADIR})
+mark_as_advanced(CLEAR CMAKE_INSTALL_JAVADIR)

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/MANIFEST.MF Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/README Added

@@ -0,0 +1,52 @@
+TurboJPEG Java Wrapper
+======================
+
+The TurboJPEG shared library can optionally be built with a Java Native
+Interface wrapper, which allows the library to be loaded and used directly from
+Java applications.  The Java front end for this is defined in several classes
+located under org/libjpegturbo/turbojpeg.  The source code for these Java
+classes is licensed under a BSD-style license, so the files can be incorporated
+directly into both open source and proprietary projects without restriction.  A
+Java archive (JAR) file containing these classes is also shipped with the
+"official" distribution packages of libjpeg-turbo.
+
+TJExample.java, which should also be located in the same directory as this
+README file, demonstrates how to use the TurboJPEG Java API to compress and
+decompress JPEG images in memory.
+
+
+Performance Pitfalls
+--------------------
+
+The TurboJPEG Java API defines several convenience methods that can allocate
+image buffers or instantiate classes to hold the result of compress,
+decompress, or transform operations.  However, if you use these methods, then
+be mindful of the amount of new data you are creating on the heap.  It may be
+necessary to manually invoke the garbage collector to prevent heap exhaustion
+or to prevent performance degradation.  Background garbage collection can kill
+performance, particularly in a multi-threaded environment (Java pauses all
+threads when the GC runs.)
+
+The TurboJPEG Java API always gives you the option of pre-allocating your own
+source and destination buffers, which allows you to re-use those buffers for
+compressing/decompressing multiple images.  If the image sequence you are
+compressing or decompressing consists of images of the same size, then
+pre-allocating the buffers is recommended.
+
+
+Installation Directory
+----------------------
+
+The TurboJPEG Java Wrapper will look for the TurboJPEG JNI library
+(libturbojpeg.so, libturbojpeg.dylib, or turbojpeg.dll) in the system library
+paths or in any paths specified in LD_LIBRARY_PATH (Un*x), DYLD_LIBRARY_PATH
+(Mac), or PATH (Windows.)  Failing this, on Un*x and Mac systems, the wrapper
+will look for the JNI library under the library directory configured when
+libjpeg-turbo was built.  If that library directory is
+/opt/libjpeg-turbo/lib32, then /opt/libjpeg-turbo/lib64 is also searched, and
+vice versa.
+
+If you installed the JNI library into another directory, then you will need
+to pass an argument of -Djava.library.path={path_to_JNI_library} to java, or
+manipulate LD_LIBRARY_PATH, DYLD_LIBRARY_PATH, or PATH to include the directory
+containing the JNI library.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/TJBench.java Added

@@ -0,0 +1,1026 @@
+/*
+ * Copyright (C)2009-2014, 2016-2019, 2021 D. R. Commander.
+ *                                         All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+import java.io.*;
+import java.awt.image.*;
+import javax.imageio.*;
+import java.util.*;
+import org.libjpegturbo.turbojpeg.*;
+
+final class TJBench {
+
+  private TJBench() {}
+
+  private static int flags = 0, quiet = 0, pf = TJ.PF_BGR, yuvPad = 1;
+  private static boolean compOnly, decompOnly, doTile, doYUV, write = true;
+
+  static final String PIXFORMATSTR = {
+    "RGB", "BGR", "RGBX", "BGRX", "XBGR", "XRGB", "GRAY"
+  };
+
+  static final String SUBNAME_LONG = {
+    "4:4:4", "4:2:2", "4:2:0", "GRAY", "4:4:0", "4:1:1"
+  };
+
+  static final String SUBNAME = {
+    "444", "422", "420", "GRAY", "440", "411"
+  };
+
+  static final String CSNAME = {
+    "RGB", "YCbCr", "GRAY", "CMYK", "YCCK"
+  };
+
+  private static TJScalingFactor sf;
+  private static int xformOp = TJTransform.OP_NONE, xformOpt = 0;
+  private static double benchTime = 5.0, warmup = 1.0;
+
+
+  static double getTime() {
+    return (double)System.nanoTime() / 1.0e9;
+  }
+
+
+  private static String tjErrorMsg;
+  private static int tjErrorCode = -1;
+
+  static void handleTJException(TJException e) throws TJException {
+    String errorMsg = e.getMessage();
+    int errorCode = e.getErrorCode();
+
+    if ((flags & TJ.FLAG_STOPONWARNING) == 0 &&
+        errorCode == TJ.ERR_WARNING) {
+      if (tjErrorMsg == null || !tjErrorMsg.equals(errorMsg) ||
+          tjErrorCode != errorCode) {
+        tjErrorMsg = errorMsg;
+        tjErrorCode = errorCode;
+        System.out.println("WARNING: " + errorMsg);
+      }
+    } else
+      throw e;
+  }
+
+
+  static String formatName(int subsamp, int cs) {
+    if (cs == TJ.CS_YCbCr)
+      return SUBNAME_LONGsubsamp;
+    else if (cs == TJ.CS_YCCK)
+      return CSNAMEcs + " " + SUBNAME_LONGsubsamp;
+    else
+      return CSNAMEcs;
+  }
+
+
+  static String sigFig(double val, int figs) {
+    String format;
+    int digitsAfterDecimal = figs - (int)Math.ceil(Math.log10(Math.abs(val)));
+
+    if (digitsAfterDecimal < 1)
+      format = new String("%.0f");
+    else
+      format = new String("%." + digitsAfterDecimal + "f");
+    return String.format(format, val);
+  }
+
+
+  static byte loadImage(String fileName, int w, int h, int pixelFormat)
+                          throws Exception {
+    BufferedImage img = ImageIO.read(new File(fileName));
+
+    if (img == null)
+      throw new Exception("Could not read " + fileName);
+    w0 = img.getWidth();
+    h0 = img.getHeight();
+
+    int rgb = img.getRGB(0, 0, w0, h0, null, 0, w0);
+    int ps = TJ.getPixelSize(pixelFormat);
+    int rindex = TJ.getRedOffset(pixelFormat);
+    int gindex = TJ.getGreenOffset(pixelFormat);
+    int bindex = TJ.getBlueOffset(pixelFormat);
+    if ((long)w0 * (long)h0 * (long)ps > (long)Integer.MAX_VALUE)
+      throw new Exception("Image is too large");
+    byte dstBuf = new bytew0 * h0 * ps;
+    int pixels = w0 * h0, dstPtr = 0, rgbPtr = 0;
+
+    while (pixels-- > 0) {
+      dstBufdstPtr + rindex = (byte)((rgbrgbPtr >> 16) & 0xff);
+      dstBufdstPtr + gindex = (byte)((rgbrgbPtr >> 8) & 0xff);
+      dstBufdstPtr + bindex = (byte)(rgbrgbPtr & 0xff);
+      dstPtr += ps;
+      rgbPtr++;
+    }
+    return dstBuf;
+  }
+
+
+  static void saveImage(String fileName, byte srcBuf, int w, int h,
+                        int pixelFormat) throws Exception {
+    BufferedImage img = new BufferedImage(w, h, BufferedImage.TYPE_INT_RGB);
+    int pixels = w * h, srcPtr = 0;
+    int ps = TJ.getPixelSize(pixelFormat);
+    int rindex = TJ.getRedOffset(pixelFormat);
+    int gindex = TJ.getGreenOffset(pixelFormat);
+    int bindex = TJ.getBlueOffset(pixelFormat);
+
+    for (int y = 0; y < h; y++) {
+      for (int x = 0; x < w; x++, srcPtr += ps) {
+        int pixel = (srcBufsrcPtr + rindex & 0xff) << 16 |
+                    (srcBufsrcPtr + gindex & 0xff) << 8 |
+                    (srcBufsrcPtr + bindex & 0xff);
+
+        img.setRGB(x, y, pixel);
+      }
+    }
+    ImageIO.write(img, "bmp", new File(fileName));
+  }
+
+
+  /* Decompression test */
+  static void decomp(byte srcBuf, byte jpegBuf, int jpegSize,
+                     byte dstBuf, int w, int h, int subsamp, int jpegQual,
+                     String fileName, int tilew, int tileh) throws Exception {
+    String qualStr = new String(""), sizeStr, tempStr;
+    TJDecompressor tjd;
+    double elapsed, elapsedDecode;
+    int ps = TJ.getPixelSize(pf), i, iter = 0;
+    int scaledw = sf.getScaled(w);
+    int scaledh = sf.getScaled(h);
+    int pitch = scaledw * ps;
+    YUVImage yuvImage = null;
+
+    if (jpegQual > 0)
+      qualStr = new String("_Q" + jpegQual);
+
+    tjd = new TJDecompressor();
+
+    if (dstBuf == null) {
+      if ((long)pitch * (long)scaledh > (long)Integer.MAX_VALUE)
+        throw new Exception("Image is too large");
+      dstBuf = new bytepitch * scaledh;
+    }
+
+    /* Set the destination buffer to gray so we know whether the decompressor
+       attempted to write to it */
+    Arrays.fill(dstBuf, (byte)127);
+
+    if (doYUV) {
+      int width = doTile ? tilew : scaledw;
+      int height = doTile ? tileh : scaledh;
+
+      yuvImage = new YUVImage(width, yuvPad, height, subsamp);
+      Arrays.fill(yuvImage.getBuf(), (byte)127);
+    }
+
+    /* Benchmark */
+    iter = -1;
+    elapsed = elapsedDecode = 0.0;
+    while (true) {
+      int tile = 0;
+      double start = getTime();
+
+      for (int y = 0; y < h; y += tileh) {
+        for (int x = 0; x < w; x += tilew, tile++) {
+          int width = doTile ? Math.min(tilew, w - x) : scaledw;
+          int height = doTile ? Math.min(tileh, h - y) : scaledh;
+
+          try {
+            tjd.setSourceImage(jpegBuftile, jpegSizetile);
+          } catch (TJException e) { handleTJException(e); }
+          if (doYUV) {
+            yuvImage.setBuf(yuvImage.getBuf(), width, yuvPad, height, subsamp);
+            try {
+              tjd.decompressToYUV(yuvImage, flags);
+            } catch (TJException e) { handleTJException(e); }
+            double startDecode = getTime();
+            tjd.setSourceImage(yuvImage);
+            try {
+              tjd.decompress(dstBuf, x, y, width, pitch, height, pf, flags);
+            } catch (TJException e) { handleTJException(e); }
+            if (iter >= 0)
+              elapsedDecode += getTime() - startDecode;
+          } else {
+            try {
+              tjd.decompress(dstBuf, x, y, width, pitch, height, pf, flags);
+            } catch (TJException e) { handleTJException(e); }
+          }
+        }
+      }
+      elapsed += getTime() - start;
+      if (iter >= 0) {
+        iter++;
+        if (elapsed >= benchTime)
+          break;
+      } else if (elapsed >= warmup) {
+        iter = 0;
+        elapsed = elapsedDecode = 0.0;
+      }
+    }
+    if (doYUV)
+      elapsed -= elapsedDecode;
+
+    tjd = null;
+    for (i = 0; i < jpegBuf.length; i++)
+      jpegBufi = null;
+    jpegBuf = null;  jpegSize = null;
+    System.gc();
+
+    if (quiet != 0) {
+      System.out.format("%-6s%s",
+                        sigFig((double)(w * h) / 1000000. *
+                               (double)iter / elapsed, 4),
+                        quiet == 2 ? "\n" : "  ");
+      if (doYUV)
+        System.out.format("%s\n",
+                          sigFig((double)(w * h) / 1000000. *
+                                 (double)iter / elapsedDecode, 4));
+      else if (quiet != 2)
+        System.out.print("\n");
+    } else {
+      System.out.format("%s --> Frame rate:         %f fps\n",
+                        (doYUV ? "Decomp to YUV" : "Decompress   "),
+                        (double)iter / elapsed);
+      System.out.format("                  Throughput:         %f Megapixels/sec\n",
+                        (double)(w * h) / 1000000. * (double)iter / elapsed);
+      if (doYUV) {
+        System.out.format("YUV Decode    --> Frame rate:         %f fps\n",
+                          (double)iter / elapsedDecode);
+        System.out.format("                  Throughput:         %f Megapixels/sec\n",
+                          (double)(w * h) / 1000000. *
+                          (double)iter / elapsedDecode);
+      }
+    }
+
+    if (!write) return;
+
+    if (sf.getNum() != 1 || sf.getDenom() != 1)
+      sizeStr = new String(sf.getNum() + "_" + sf.getDenom());
+    else if (tilew != w || tileh != h)
+      sizeStr = new String(tilew + "x" + tileh);
+    else
+      sizeStr = new String("full");
+    if (decompOnly)
+      tempStr = new String(fileName + "_" + sizeStr + ".bmp");
+    else
+      tempStr = new String(fileName + "_" + SUBNAMEsubsamp + qualStr +
+                           "_" + sizeStr + ".bmp");
+
+    saveImage(tempStr, dstBuf, scaledw, scaledh, pf);
+    int ndx = tempStr.lastIndexOf('.');
+    tempStr = new String(tempStr.substring(0, ndx) + "-err.bmp");
+    if (srcBuf != null && sf.getNum() == 1 && sf.getDenom() == 1) {
+      if (quiet == 0)
+        System.out.println("Compression error written to " + tempStr + ".");
+      if (subsamp == TJ.SAMP_GRAY) {
+        for (int y = 0, index = 0; y < h; y++, index += pitch) {
+          for (int x = 0, index2 = index; x < w; x++, index2 += ps) {
+            int rindex = index2 + TJ.getRedOffset(pf);
+            int gindex = index2 + TJ.getGreenOffset(pf);
+            int bindex = index2 + TJ.getBlueOffset(pf);
+            int lum = (int)((double)(srcBufrindex & 0xff) * 0.299 +
+                            (double)(srcBufgindex & 0xff) * 0.587 +
+                            (double)(srcBufbindex & 0xff) * 0.114 + 0.5);
+
+            if (lum > 255) lum = 255;
+            if (lum < 0) lum = 0;
+            dstBufrindex = (byte)Math.abs((dstBufrindex & 0xff) - lum);
+            dstBufgindex = (byte)Math.abs((dstBufgindex & 0xff) - lum);
+            dstBufbindex = (byte)Math.abs((dstBufbindex & 0xff) - lum);
+          }
+        }
+      } else {
+        for (int y = 0; y < h; y++)
+          for (int x = 0; x < w * ps; x++)
+            dstBufpitch * y + x =
+              (byte)Math.abs((dstBufpitch * y + x & 0xff) -
+                             (srcBufpitch * y + x & 0xff));
+      }
+      saveImage(tempStr, dstBuf, w, h, pf);
+    }
+  }
+
+
+  static void fullTest(byte srcBuf, int w, int h, int subsamp, int jpegQual,
+                       String fileName) throws Exception {
+    TJCompressor tjc;
+    byte tmpBuf;
+    byte jpegBuf;
+    int jpegSize;
+    double start, elapsed, elapsedEncode;
+    int totalJpegSize = 0, tilew, tileh, i, iter;
+    int ps = TJ.getPixelSize(pf);
+    int ntilesw = 1, ntilesh = 1, pitch = w * ps;
+    String pfStr = PIXFORMATSTRpf;
+    YUVImage yuvImage = null;
+
+    if ((long)pitch * (long)h > (long)Integer.MAX_VALUE)
+      throw new Exception("Image is too large");
+    tmpBuf = new bytepitch * h;
+
+    if (quiet == 0)
+      System.out.format(">>>>>  %s (%s) <--> JPEG %s Q%d  <<<<<\n", pfStr,
+                        (flags & TJ.FLAG_BOTTOMUP) != 0 ?
+                        "Bottom-up" : "Top-down",
+                        SUBNAME_LONGsubsamp, jpegQual);
+
+    tjc = new TJCompressor();
+
+    for (tilew = doTile ? 8 : w, tileh = doTile ? 8 : h; ;
+         tilew *= 2, tileh *= 2) {
+      if (tilew > w)
+        tilew = w;
+      if (tileh > h)
+        tileh = h;
+      ntilesw = (w + tilew - 1) / tilew;
+      ntilesh = (h + tileh - 1) / tileh;
+
+      jpegBuf = new bytentilesw * ntileshTJ.bufSize(tilew, tileh, subsamp);
+      jpegSize = new intntilesw * ntilesh;
+
+      /* Compression test */
+      if (quiet == 1)
+        System.out.format("%-4s (%s)  %-5s    %-3d   ", pfStr,
+                          (flags & TJ.FLAG_BOTTOMUP) != 0 ? "BU" : "TD",
+                          SUBNAME_LONGsubsamp, jpegQual);
+      for (i = 0; i < h; i++)
+        System.arraycopy(srcBuf, w * ps * i, tmpBuf, pitch * i, w * ps);
+      tjc.setJPEGQuality(jpegQual);
+      tjc.setSubsamp(subsamp);
+
+      if (doYUV) {
+        yuvImage = new YUVImage(tilew, yuvPad, tileh, subsamp);
+        Arrays.fill(yuvImage.getBuf(), (byte)127);
+      }
+
+      /* Benchmark */
+      iter = -1;
+      elapsed = elapsedEncode = 0.0;
+      while (true) {
+        int tile = 0;
+
+        totalJpegSize = 0;
+        start = getTime();
+        for (int y = 0; y < h; y += tileh) {
+          for (int x = 0; x < w; x += tilew, tile++) {
+            int width = Math.min(tilew, w - x);
+            int height = Math.min(tileh, h - y);
+
+            tjc.setSourceImage(srcBuf, x, y, width, pitch, height, pf);
+            if (doYUV) {
+              double startEncode = getTime();
+
+              yuvImage.setBuf(yuvImage.getBuf(), width, yuvPad, height,
+                              subsamp);
+              tjc.encodeYUV(yuvImage, flags);
+              if (iter >= 0)
+                elapsedEncode += getTime() - startEncode;
+              tjc.setSourceImage(yuvImage);
+            }
+            tjc.compress(jpegBuftile, flags);
+            jpegSizetile = tjc.getCompressedSize();
+            totalJpegSize += jpegSizetile;
+          }
+        }
+        elapsed += getTime() - start;
+        if (iter >= 0) {
+          iter++;
+          if (elapsed >= benchTime)
+            break;
+        } else if (elapsed >= warmup) {
+          iter = 0;
+          elapsed = elapsedEncode = 0.0;
+        }
+      }
+      if (doYUV)
+        elapsed -= elapsedEncode;
+
+      if (quiet == 1)
+        System.out.format("%-5d  %-5d   ", tilew, tileh);
+      if (quiet != 0) {
+        if (doYUV)
+          System.out.format("%-6s%s",
+                            sigFig((double)(w * h) / 1000000. *
+                                   (double)iter / elapsedEncode, 4),
+                            quiet == 2 ? "\n" : "  ");
+        System.out.format("%-6s%s",
+                          sigFig((double)(w * h) / 1000000. *
+                                 (double)iter / elapsed, 4),
+                          quiet == 2 ? "\n" : "  ");
+        System.out.format("%-6s%s",
+                          sigFig((double)(w * h * ps) / (double)totalJpegSize,
+                                 4),
+                          quiet == 2 ? "\n" : "  ");
+      } else {
+        System.out.format("\n%s size: %d x %d\n", doTile ? "Tile" : "Image",
+                          tilew, tileh);
+        if (doYUV) {
+          System.out.format("Encode YUV    --> Frame rate:         %f fps\n",
+                            (double)iter / elapsedEncode);
+          System.out.format("                  Output image size:  %d bytes\n",
+                            yuvImage.getSize());
+          System.out.format("                  Compression ratio:  %f:1\n",
+                            (double)(w * h * ps) / (double)yuvImage.getSize());
+          System.out.format("                  Throughput:         %f Megapixels/sec\n",
+                            (double)(w * h) / 1000000. *
+                            (double)iter / elapsedEncode);
+          System.out.format("                  Output bit stream:  %f Megabits/sec\n",
+                            (double)yuvImage.getSize() * 8. / 1000000. *
+                            (double)iter / elapsedEncode);
+        }
+        System.out.format("%s --> Frame rate:         %f fps\n",
+                          doYUV ? "Comp from YUV" : "Compress     ",
+                          (double)iter / elapsed);
+        System.out.format("                  Output image size:  %d bytes\n",
+                          totalJpegSize);
+        System.out.format("                  Compression ratio:  %f:1\n",
+                          (double)(w * h * ps) / (double)totalJpegSize);
+        System.out.format("                  Throughput:         %f Megapixels/sec\n",
+                          (double)(w * h) / 1000000. * (double)iter / elapsed);
+        System.out.format("                  Output bit stream:  %f Megabits/sec\n",
+                          (double)totalJpegSize * 8. / 1000000. *
+                          (double)iter / elapsed);
+      }
+      if (tilew == w && tileh == h && write) {
+        String tempStr = fileName + "_" + SUBNAMEsubsamp + "_" + "Q" +
+                         jpegQual + ".jpg";
+        FileOutputStream fos = new FileOutputStream(tempStr);
+
+        fos.write(jpegBuf0, 0, jpegSize0);
+        fos.close();
+        if (quiet == 0)
+          System.out.println("Reference image written to " + tempStr);
+      }
+
+      /* Decompression test */
+      if (!compOnly)
+        decomp(srcBuf, jpegBuf, jpegSize, tmpBuf, w, h, subsamp, jpegQual,
+               fileName, tilew, tileh);
+      else if (quiet == 1)
+        System.out.println("N/A");
+
+      if (tilew == w && tileh == h) break;
+    }
+  }
+
+
+  static void decompTest(String fileName) throws Exception {
+    TJTransformer tjt;
+    byte jpegBuf = null;
+    byte srcBuf;
+    int jpegSize = null;
+    int totalJpegSize;
+    double start, elapsed;
+    int ps = TJ.getPixelSize(pf), tile, x, y, iter;
+    // Original image
+    int w = 0, h = 0, ntilesw = 1, ntilesh = 1, subsamp = -1, cs = -1;
+    // Transformed image
+    int tw, th, ttilew, ttileh, tntilesw, tntilesh, tsubsamp;
+
+    FileInputStream fis = new FileInputStream(fileName);
+    if (fis.getChannel().size() > (long)Integer.MAX_VALUE)
+      throw new Exception("Image is too large");
+    int srcSize = (int)fis.getChannel().size();
+    srcBuf = new bytesrcSize;
+    fis.read(srcBuf, 0, srcSize);
+    fis.close();
+
+    int index = fileName.lastIndexOf('.');
+    if (index >= 0)
+      fileName = new String(fileName.substring(0, index));
+
+    tjt = new TJTransformer();
+
+    try {
+      tjt.setSourceImage(srcBuf, srcSize);
+    } catch (TJException e) { handleTJException(e); }
+    w = tjt.getWidth();
+    h = tjt.getHeight();
+    subsamp = tjt.getSubsamp();
+    cs = tjt.getColorspace();
+
+    if (quiet == 1) {
+      System.out.println("All performance values in Mpixels/sec\n");
+      System.out.format("Bitmap     JPEG   JPEG     %s  %s   Xform   Comp    Decomp  ",
+                        (doTile ? "Tile " : "Image"),
+                        (doTile ? "Tile " : "Image"));
+      if (doYUV)
+        System.out.print("Decode");
+      System.out.print("\n");
+      System.out.print("Format     CS     Subsamp  Width  Height  Perf    Ratio   Perf    ");
+      if (doYUV)
+        System.out.print("Perf");
+      System.out.println("\n");
+    } else if (quiet == 0)
+      System.out.format(">>>>>  JPEG %s --> %s (%s)  <<<<<\n",
+                        formatName(subsamp, cs), PIXFORMATSTRpf,
+                        (flags & TJ.FLAG_BOTTOMUP) != 0 ?
+                        "Bottom-up" : "Top-down");
+
+    for (int tilew = doTile ? 16 : w, tileh = doTile ? 16 : h; ;
+         tilew *= 2, tileh *= 2) {
+      if (tilew > w)
+        tilew = w;
+      if (tileh > h)
+        tileh = h;
+      ntilesw = (w + tilew - 1) / tilew;
+      ntilesh = (h + tileh - 1) / tileh;
+
+      tw = w;  th = h;  ttilew = tilew;  ttileh = tileh;
+      if (quiet == 0) {
+        System.out.format("\n%s size: %d x %d", (doTile ? "Tile" : "Image"),
+                          ttilew, ttileh);
+        if (sf.getNum() != 1 || sf.getDenom() != 1)
+          System.out.format(" --> %d x %d", sf.getScaled(tw),
+                            sf.getScaled(th));
+        System.out.println("");
+      } else if (quiet == 1) {
+        System.out.format("%-4s (%s)  %-5s  %-5s    ", PIXFORMATSTRpf,
+                          (flags & TJ.FLAG_BOTTOMUP) != 0 ? "BU" : "TD",
+                          CSNAMEcs, SUBNAME_LONGsubsamp);
+        System.out.format("%-5d  %-5d   ", tilew, tileh);
+      }
+
+      tsubsamp = subsamp;
+      if (doTile || xformOp != TJTransform.OP_NONE || xformOpt != 0) {
+        if (xformOp == TJTransform.OP_TRANSPOSE ||
+            xformOp == TJTransform.OP_TRANSVERSE ||
+            xformOp == TJTransform.OP_ROT90 ||
+            xformOp == TJTransform.OP_ROT270) {
+          tw = h;  th = w;  ttilew = tileh;  ttileh = tilew;
+        }
+
+        if ((xformOpt & TJTransform.OPT_GRAY) != 0)
+          tsubsamp = TJ.SAMP_GRAY;
+        if (xformOp == TJTransform.OP_HFLIP ||
+            xformOp == TJTransform.OP_ROT180)
+          tw = tw - (tw % TJ.getMCUWidth(tsubsamp));
+        if (xformOp == TJTransform.OP_VFLIP ||
+            xformOp == TJTransform.OP_ROT180)
+          th = th - (th % TJ.getMCUHeight(tsubsamp));
+        if (xformOp == TJTransform.OP_TRANSVERSE ||
+            xformOp == TJTransform.OP_ROT90)
+          tw = tw - (tw % TJ.getMCUHeight(tsubsamp));
+        if (xformOp == TJTransform.OP_TRANSVERSE ||
+            xformOp == TJTransform.OP_ROT270)
+          th = th - (th % TJ.getMCUWidth(tsubsamp));
+        tntilesw = (tw + ttilew - 1) / ttilew;
+        tntilesh = (th + ttileh - 1) / ttileh;
+
+        if (xformOp == TJTransform.OP_TRANSPOSE ||
+            xformOp == TJTransform.OP_TRANSVERSE ||
+            xformOp == TJTransform.OP_ROT90 ||
+            xformOp == TJTransform.OP_ROT270) {
+          if (tsubsamp == TJ.SAMP_422)
+            tsubsamp = TJ.SAMP_440;
+          else if (tsubsamp == TJ.SAMP_440)
+            tsubsamp = TJ.SAMP_422;
+        }
+
+        TJTransform t = new TJTransformtntilesw * tntilesh;
+        jpegBuf =
+          new bytetntilesw * tntileshTJ.bufSize(ttilew, ttileh, subsamp);
+
+        for (y = 0, tile = 0; y < th; y += ttileh) {
+          for (x = 0; x < tw; x += ttilew, tile++) {
+            ttile = new TJTransform();
+            ttile.width = Math.min(ttilew, tw - x);
+            ttile.height = Math.min(ttileh, th - y);
+            ttile.x = x;
+            ttile.y = y;
+            ttile.op = xformOp;
+            ttile.options = xformOpt | TJTransform.OPT_TRIM;
+            if ((ttile.options & TJTransform.OPT_NOOUTPUT) != 0 &&
+                jpegBuftile != null)
+              jpegBuftile = null;
+          }
+        }
+
+        iter = -1;
+        elapsed = 0.;
+        while (true) {
+          start = getTime();
+          try {
+            tjt.transform(jpegBuf, t, flags);
+          } catch (TJException e) { handleTJException(e); }
+          jpegSize = tjt.getTransformedSizes();
+          elapsed += getTime() - start;
+          if (iter >= 0) {
+            iter++;
+            if (elapsed >= benchTime)
+              break;
+          } else if (elapsed >= warmup) {
+            iter = 0;
+            elapsed = 0.0;
+          }
+        }
+        t = null;
+
+        for (tile = 0, totalJpegSize = 0; tile < tntilesw * tntilesh; tile++)
+          totalJpegSize += jpegSizetile;
+
+        if (quiet != 0) {
+          System.out.format("%-6s%s%-6s%s",
+                            sigFig((double)(w * h) / 1000000. / elapsed, 4),
+                            quiet == 2 ? "\n" : "  ",
+                            sigFig((double)(w * h * ps) /
+                                   (double)totalJpegSize, 4),
+                            quiet == 2 ? "\n" : "  ");
+        } else if (quiet == 0) {
+          System.out.format("Transform     --> Frame rate:         %f fps\n",
+                            1.0 / elapsed);
+          System.out.format("                  Output image size:  %d bytes\n",
+                            totalJpegSize);
+          System.out.format("                  Compression ratio:  %f:1\n",
+                            (double)(w * h * ps) / (double)totalJpegSize);
+          System.out.format("                  Throughput:         %f Megapixels/sec\n",
+                            (double)(w * h) / 1000000. / elapsed);
+          System.out.format("                  Output bit stream:  %f Megabits/sec\n",
+                            (double)totalJpegSize * 8. / 1000000. / elapsed);
+        }
+      } else {
+        if (quiet == 1)
+          System.out.print("N/A     N/A     ");
+        jpegBuf = new byte1TJ.bufSize(ttilew, ttileh, subsamp);
+        jpegSize = new int1;
+        jpegBuf0 = srcBuf;
+        jpegSize0 = srcSize;
+      }
+
+      if (w == tilew)
+        ttilew = tw;
+      if (h == tileh)
+        ttileh = th;
+      if ((xformOpt & TJTransform.OPT_NOOUTPUT) == 0)
+        decomp(null, jpegBuf, jpegSize, null, tw, th, tsubsamp, 0,
+               fileName, ttilew, ttileh);
+      else if (quiet == 1)
+        System.out.println("N/A");
+
+      jpegBuf = null;
+      jpegSize = null;
+
+      if (tilew == w && tileh == h) break;
+    }
+  }
+
+
+  static void usage() throws Exception {
+    int i;
+    TJScalingFactor scalingFactors = TJ.getScalingFactors();
+    int nsf = scalingFactors.length;
+    String className = new TJBench().getClass().getName();
+
+    System.out.println("\nUSAGE: java " + className);
+    System.out.println("       <Inputfile (BMP)> <Quality> options\n");
+    System.out.println("       java " + className);
+    System.out.println("       <Inputfile (JPG)> options\n");
+    System.out.println("Options:\n");
+    System.out.println("-alloc = Dynamically allocate JPEG image buffers");
+    System.out.println("-bottomup = Test bottom-up compression/decompression");
+    System.out.println("-tile = Test performance of the codec when the image is encoded as separate");
+    System.out.println("     tiles of varying sizes.");
+    System.out.println("-rgb, -bgr, -rgbx, -bgrx, -xbgr, -xrgb =");
+    System.out.println("     Test the specified color conversion path in the codec (default = BGR)");
+    System.out.println("-fastupsample = Use the fastest chrominance upsampling algorithm available in");
+    System.out.println("     the underlying codec");
+    System.out.println("-fastdct = Use the fastest DCT/IDCT algorithms available in the underlying");
+    System.out.println("     codec");
+    System.out.println("-accuratedct = Use the most accurate DCT/IDCT algorithms available in the");
+    System.out.println("     underlying codec");
+    System.out.println("-progressive = Use progressive entropy coding in JPEG images generated by");
+    System.out.println("     compression and transform operations.");
+    System.out.println("-subsamp <s> = When testing JPEG compression, this option specifies the level");
+    System.out.println("     of chrominance subsampling to use (<s> = 444, 422, 440, 420, 411, or");
+    System.out.println("     GRAY).  The default is to test Grayscale, 4:2:0, 4:2:2, and 4:4:4 in");
+    System.out.println("     sequence.");
+    System.out.println("-quiet = Output results in tabular rather than verbose format");
+    System.out.println("-yuv = Test YUV encoding/decoding functions");
+    System.out.println("-yuvpad <p> = If testing YUV encoding/decoding, this specifies the number of");
+    System.out.println("     bytes to which each row of each plane in the intermediate YUV image is");
+    System.out.println("     padded (default = 1)");
+    System.out.println("-scale M/N = Scale down the width/height of the decompressed JPEG image by a");
+    System.out.print("     factor of M/N (M/N = ");
+    for (i = 0; i < nsf; i++) {
+      System.out.format("%d/%d", scalingFactorsi.getNum(),
+                        scalingFactorsi.getDenom());
+      if (nsf == 2 && i != nsf - 1)
+        System.out.print(" or ");
+      else if (nsf > 2) {
+        if (i != nsf - 1)
+          System.out.print(", ");
+        if (i == nsf - 2)
+          System.out.print("or ");
+      }
+      if (i % 8 == 0 && i != 0)
+        System.out.print("\n     ");
+    }
+    System.out.println(")");
+    System.out.println("-hflip, -vflip, -transpose, -transverse, -rot90, -rot180, -rot270 =");
+    System.out.println("     Perform the corresponding lossless transform prior to");
+    System.out.println("     decompression (these options are mutually exclusive)");
+    System.out.println("-grayscale = Perform lossless grayscale conversion prior to decompression");
+    System.out.println("     test (can be combined with the other transforms above)");
+    System.out.println("-copynone = Do not copy any extra markers (including EXIF and ICC profile data)");
+    System.out.println("     when transforming the image.");
+    System.out.println("-benchtime <t> = Run each benchmark for at least <t> seconds (default = 5.0)");
+    System.out.println("-warmup <t> = Run each benchmark for <t> seconds (default = 1.0) prior to");
+    System.out.println("     starting the timer, in order to prime the caches and thus improve the");
+    System.out.println("     consistency of the results.");
+    System.out.println("-componly = Stop after running compression tests.  Do not test decompression.");
+    System.out.println("-nowrite = Do not write reference or output images (improves consistency");
+    System.out.println("     of performance measurements.)");
+    System.out.println("-limitscans = Refuse to decompress or transform progressive JPEG images that");
+    System.out.println("     have an unreasonably large number of scans");
+    System.out.println("-stoponwarning = Immediately discontinue the current");
+    System.out.println("     compression/decompression/transform operation if the underlying codec");
+    System.out.println("     throws a warning (non-fatal error)\n");
+    System.out.println("NOTE:  If the quality is specified as a range (e.g. 90-100), a separate");
+    System.out.println("test will be performed for all quality values in the range.\n");
+    System.exit(1);
+  }
+
+
+  public static void main(String argv) {
+    byte srcBuf = null;
+    int w = 0, h = 0, minQual = -1, maxQual = -1;
+    int minArg = 1, retval = 0;
+    int subsamp = -1;
+
+    try {
+
+      if (argv.length < minArg)
+        usage();
+
+      String tempStr = argv0.toLowerCase();
+      if (tempStr.endsWith(".jpg") || tempStr.endsWith(".jpeg"))
+        decompOnly = true;
+
+      System.out.println("");
+
+      if (!decompOnly) {
+        minArg = 2;
+        if (argv.length < minArg)
+          usage();
+        try {
+          minQual = Integer.parseInt(argv1);
+        } catch (NumberFormatException e) {}
+        if (minQual < 1 || minQual > 100)
+          throw new Exception("Quality must be between 1 and 100.");
+        int dashIndex = argv1.indexOf('-');
+        if (dashIndex > 0 && argv1.length() > dashIndex + 1) {
+          try {
+            maxQual = Integer.parseInt(argv1.substring(dashIndex + 1));
+          } catch (NumberFormatException e) {}
+        }
+        if (maxQual < 1 || maxQual > 100)
+          maxQual = minQual;
+      }
+
+      if (argv.length > minArg) {
+        for (int i = minArg; i < argv.length; i++) {
+          if (argvi.equalsIgnoreCase("-tile")) {
+            doTile = true;  xformOpt |= TJTransform.OPT_CROP;
+          } else if (argvi.equalsIgnoreCase("-fastupsample")) {
+            System.out.println("Using fast upsampling code\n");
+            flags |= TJ.FLAG_FASTUPSAMPLE;
+          } else if (argvi.equalsIgnoreCase("-fastdct")) {
+            System.out.println("Using fastest DCT/IDCT algorithm\n");
+            flags |= TJ.FLAG_FASTDCT;
+          } else if (argvi.equalsIgnoreCase("-accuratedct")) {
+            System.out.println("Using most accurate DCT/IDCT algorithm\n");
+            flags |= TJ.FLAG_ACCURATEDCT;
+          } else if (argvi.equalsIgnoreCase("-progressive")) {
+            System.out.println("Using progressive entropy coding\n");
+            flags |= TJ.FLAG_PROGRESSIVE;
+          } else if (argvi.equalsIgnoreCase("-rgb"))
+            pf = TJ.PF_RGB;
+          else if (argvi.equalsIgnoreCase("-rgbx"))
+            pf = TJ.PF_RGBX;
+          else if (argvi.equalsIgnoreCase("-bgr"))
+            pf = TJ.PF_BGR;
+          else if (argvi.equalsIgnoreCase("-bgrx"))
+            pf = TJ.PF_BGRX;
+          else if (argvi.equalsIgnoreCase("-xbgr"))
+            pf = TJ.PF_XBGR;
+          else if (argvi.equalsIgnoreCase("-xrgb"))
+            pf = TJ.PF_XRGB;
+          else if (argvi.equalsIgnoreCase("-bottomup"))
+            flags |= TJ.FLAG_BOTTOMUP;
+          else if (argvi.equalsIgnoreCase("-quiet"))
+            quiet = 1;
+          else if (argvi.equalsIgnoreCase("-qq"))
+            quiet = 2;
+          else if (argvi.equalsIgnoreCase("-scale") && i < argv.length - 1) {
+            int temp1 = 0, temp2 = 0;
+            boolean match = false, scanned = true;
+            Scanner scanner = new Scanner(argv++i).useDelimiter("/");
+
+            try {
+              temp1 = scanner.nextInt();
+              temp2 = scanner.nextInt();
+            } catch (Exception e) {}
+            if (temp2 <= 0) temp2 = 1;
+            if (temp1 > 0) {
+              TJScalingFactor scalingFactors = TJ.getScalingFactors();
+
+              for (int j = 0; j < scalingFactors.length; j++) {
+                if ((double)temp1 / (double)temp2 ==
+                    (double)scalingFactorsj.getNum() /
+                    (double)scalingFactorsj.getDenom()) {
+                  sf = scalingFactorsj;
+                  match = true;  break;
+                }
+              }
+              if (!match) usage();
+            } else
+              usage();
+          } else if (argvi.equalsIgnoreCase("-hflip"))
+            xformOp = TJTransform.OP_HFLIP;
+          else if (argvi.equalsIgnoreCase("-vflip"))
+            xformOp = TJTransform.OP_VFLIP;
+          else if (argvi.equalsIgnoreCase("-transpose"))
+            xformOp = TJTransform.OP_TRANSPOSE;
+          else if (argvi.equalsIgnoreCase("-transverse"))
+            xformOp = TJTransform.OP_TRANSVERSE;
+          else if (argvi.equalsIgnoreCase("-rot90"))
+            xformOp = TJTransform.OP_ROT90;
+          else if (argvi.equalsIgnoreCase("-rot180"))
+            xformOp = TJTransform.OP_ROT180;
+          else if (argvi.equalsIgnoreCase("-rot270"))
+            xformOp = TJTransform.OP_ROT270;
+          else if (argvi.equalsIgnoreCase("-grayscale"))
+            xformOpt |= TJTransform.OPT_GRAY;
+          else if (argvi.equalsIgnoreCase("-nooutput"))
+            xformOpt |= TJTransform.OPT_NOOUTPUT;
+          else if (argvi.equalsIgnoreCase("-copynone"))
+            xformOpt |= TJTransform.OPT_COPYNONE;
+          else if (argvi.equalsIgnoreCase("-benchtime") &&
+                   i < argv.length - 1) {
+            double temp = -1;
+
+            try {
+              temp = Double.parseDouble(argv++i);
+            } catch (NumberFormatException e) {}
+            if (temp > 0.0)
+              benchTime = temp;
+            else
+              usage();
+          } else if (argvi.equalsIgnoreCase("-warmup") &&
+                     i < argv.length - 1) {
+            double temp = -1;
+
+            try {
+              temp = Double.parseDouble(argv++i);
+            } catch (NumberFormatException e) {}
+            if (temp >= 0.0) {
+              warmup = temp;
+              System.out.format("Warmup time = %.1f seconds\n\n", warmup);
+            } else
+              usage();
+          } else if (argvi.equalsIgnoreCase("-yuv")) {
+            System.out.println("Testing YUV planar encoding/decoding\n");
+            doYUV = true;
+          } else if (argvi.equalsIgnoreCase("-yuvpad") &&
+                     i < argv.length - 1) {
+            int temp = 0;
+
+            try {
+              temp = Integer.parseInt(argv++i);
+            } catch (NumberFormatException e) {}
+            if (temp >= 1)
+              yuvPad = temp;
+          } else if (argvi.equalsIgnoreCase("-subsamp") &&
+                     i < argv.length - 1) {
+            i++;
+            if (argvi.toUpperCase().startsWith("G"))
+              subsamp = TJ.SAMP_GRAY;
+            else if (argvi.equals("444"))
+              subsamp = TJ.SAMP_444;
+            else if (argvi.equals("422"))
+              subsamp = TJ.SAMP_422;
+            else if (argvi.equals("440"))
+              subsamp = TJ.SAMP_440;
+            else if (argvi.equals("420"))
+              subsamp = TJ.SAMP_420;
+            else if (argvi.equals("411"))
+              subsamp = TJ.SAMP_411;
+          } else if (argvi.equalsIgnoreCase("-componly"))
+            compOnly = true;
+          else if (argvi.equalsIgnoreCase("-nowrite"))
+            write = false;
+          else if (argvi.equalsIgnoreCase("-limitscans"))
+            flags |= TJ.FLAG_LIMITSCANS;
+          else if (argvi.equalsIgnoreCase("-stoponwarning"))
+            flags |= TJ.FLAG_STOPONWARNING;
+          else usage();
+        }
+      }
+
+      if (sf == null)
+        sf = new TJScalingFactor(1, 1);
+
+      if ((sf.getNum() != 1 || sf.getDenom() != 1) && doTile) {
+        System.out.println("Disabling tiled compression/decompression tests, because those tests do not");
+        System.out.println("work when scaled decompression is enabled.");
+        doTile = false;
+      }
+
+      if (!decompOnly) {
+        int width = new int1, height = new int1;
+
+        srcBuf = loadImage(argv0, width, height, pf);
+        w = width0;  h = height0;
+        int index = -1;
+        if ((index = argv0.lastIndexOf('.')) >= 0)
+          argv0 = argv0.substring(0, index);
+      }
+
+      if (quiet == 1 && !decompOnly) {
+        System.out.println("All performance values in Mpixels/sec\n");
+        System.out.format("Bitmap     JPEG     JPEG  %s  %s   ",
+                          (doTile ? "Tile " : "Image"),
+                          (doTile ? "Tile " : "Image"));
+        if (doYUV)
+          System.out.print("Encode  ");
+        System.out.print("Comp    Comp    Decomp  ");
+        if (doYUV)
+          System.out.print("Decode");
+        System.out.print("\n");
+        System.out.print("Format     Subsamp  Qual  Width  Height  ");
+        if (doYUV)
+          System.out.print("Perf    ");
+        System.out.print("Perf    Ratio   Perf    ");
+        if (doYUV)
+          System.out.print("Perf");
+        System.out.println("\n");
+      }
+
+      if (decompOnly) {
+        decompTest(argv0);
+        System.out.println("");
+        System.exit(retval);
+      }
+
+      System.gc();
+      if (subsamp >= 0 && subsamp < TJ.NUMSAMP) {
+        for (int i = maxQual; i >= minQual; i--)
+          fullTest(srcBuf, w, h, subsamp, i, argv0);
+        System.out.println("");
+      } else {
+        for (int i = maxQual; i >= minQual; i--)
+          fullTest(srcBuf, w, h, TJ.SAMP_GRAY, i, argv0);
+        System.out.println("");
+        System.gc();
+        for (int i = maxQual; i >= minQual; i--)
+          fullTest(srcBuf, w, h, TJ.SAMP_420, i, argv0);
+        System.out.println("");
+        System.gc();
+        for (int i = maxQual; i >= minQual; i--)
+          fullTest(srcBuf, w, h, TJ.SAMP_422, i, argv0);
+        System.out.println("");
+        System.gc();
+        for (int i = maxQual; i >= minQual; i--)
+          fullTest(srcBuf, w, h, TJ.SAMP_444, i, argv0);
+        System.out.println("");
+      }
+
+    } catch (Exception e) {
+      if (e instanceof TJException) {
+        TJException tje = (TJException)e;
+
+        System.out.println((tje.getErrorCode() == TJ.ERR_WARNING ?
+                            "WARNING: " : "ERROR: ") + tje.getMessage());
+      } else
+        System.out.println("ERROR: " + e.getMessage());
+      e.printStackTrace();
+      retval = -1;
+    }
+
+    System.exit(retval);
+  }
+
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/TJExample.java Added

@@ -0,0 +1,405 @@
+/*
+ * Copyright (C)2011-2012, 2014-2015, 2017-2018 D. R. Commander.
+ *                                              All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * This program demonstrates how to compress, decompress, and transform JPEG
+ * images using the TurboJPEG Java API
+ */
+
+import java.io.*;
+import java.awt.*;
+import java.awt.image.*;
+import java.nio.*;
+import javax.imageio.*;
+import javax.swing.*;
+import org.libjpegturbo.turbojpeg.*;
+
+
+@SuppressWarnings("checkstyle:JavadocType")
+class TJExample implements TJCustomFilter {
+
+  static final String CLASS_NAME =
+    new TJExample().getClass().getName();
+
+  static final int DEFAULT_SUBSAMP = TJ.SAMP_444;
+  static final int DEFAULT_QUALITY = 95;
+
+
+  static final String SUBSAMP_NAME = {
+    "4:4:4", "4:2:2", "4:2:0", "Grayscale", "4:4:0", "4:1:1"
+  };
+
+  static final String COLORSPACE_NAME = {
+    "RGB", "YCbCr", "GRAY", "CMYK", "YCCK"
+  };
+
+
+  /* DCT filter example.  This produces a negative of the image. */
+
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  public void customFilter(ShortBuffer coeffBuffer, Rectangle bufferRegion,
+                           Rectangle planeRegion, int componentIndex,
+                           int transformIndex, TJTransform transform)
+                           throws TJException {
+    for (int i = 0; i < bufferRegion.width * bufferRegion.height; i++) {
+      coeffBuffer.put(i, (short)(-coeffBuffer.get(i)));
+    }
+  }
+
+
+  static void usage() throws Exception {
+    System.out.println("\nUSAGE: java Java options " + CLASS_NAME +
+                       " <Input image> <Output image> options\n");
+
+    System.out.println("Input and output images can be in any image format that the Java Image I/O");
+    System.out.println("extensions understand.  If either filename ends in a .jpg extension, then");
+    System.out.println("the TurboJPEG API will be used to compress or decompress the image.\n");
+
+    System.out.println("Compression Options (used if the output image is a JPEG image)");
+    System.out.println("--------------------------------------------------------------\n");
+
+    System.out.println("-subsamp <444|422|420|gray> = Apply this level of chrominance subsampling when");
+    System.out.println("     compressing the output image.  The default is to use the same level of");
+    System.out.println("     subsampling as in the input image, if the input image is also a JPEG");
+    System.out.println("     image, or to use grayscale if the input image is a grayscale non-JPEG");
+    System.out.println("     image, or to use " +
+                       SUBSAMP_NAMEDEFAULT_SUBSAMP +
+                       " subsampling otherwise.\n");
+
+    System.out.println("-q <1-100> = Compress the output image with this JPEG quality level");
+    System.out.println("     (default = " + DEFAULT_QUALITY + ").\n");
+
+    System.out.println("Decompression Options (used if the input image is a JPEG image)");
+    System.out.println("---------------------------------------------------------------\n");
+
+    System.out.println("-scale M/N = Scale the input image by a factor of M/N when decompressing it.");
+    System.out.print("(M/N = ");
+    for (int i = 0; i < SCALING_FACTORS.length; i++) {
+      System.out.print(SCALING_FACTORSi.getNum() + "/" +
+                       SCALING_FACTORSi.getDenom());
+      if (SCALING_FACTORS.length == 2 && i != SCALING_FACTORS.length - 1)
+        System.out.print(" or ");
+      else if (SCALING_FACTORS.length > 2) {
+        if (i != SCALING_FACTORS.length - 1)
+          System.out.print(", ");
+        if (i == SCALING_FACTORS.length - 2)
+          System.out.print("or ");
+      }
+    }
+    System.out.println(")\n");
+
+    System.out.println("-hflip, -vflip, -transpose, -transverse, -rot90, -rot180, -rot270 =");
+    System.out.println("     Perform one of these lossless transform operations on the input image");
+    System.out.println("     prior to decompressing it (these options are mutually exclusive.)\n");
+
+    System.out.println("-grayscale = Perform lossless grayscale conversion on the input image prior");
+    System.out.println("     to decompressing it (can be combined with the other transform operations");
+    System.out.println("     above.)\n");
+
+    System.out.println("-crop WxH+X+Y = Perform lossless cropping on the input image prior to");
+    System.out.println("     decompressing it.  X and Y specify the upper left corner of the cropping");
+    System.out.println("     region, and W and H specify the width and height of the cropping region.");
+    System.out.println("     X and Y must be evenly divible by the MCU block size (8x8 if the input");
+    System.out.println("     image was compressed using no subsampling or grayscale, 16x8 if it was");
+    System.out.println("     compressed using 4:2:2 subsampling, or 16x16 if it was compressed using");
+    System.out.println("     4:2:0 subsampling.)\n");
+
+    System.out.println("General Options");
+    System.out.println("---------------\n");
+
+    System.out.println("-display = Display output image (Output filename need not be specified in this");
+    System.out.println("     case.)\n");
+
+    System.out.println("-fastupsample = Use the fastest chrominance upsampling algorithm available in");
+    System.out.println("     the underlying codec.\n");
+
+    System.out.println("-fastdct = Use the fastest DCT/IDCT algorithms available in the underlying");
+    System.out.println("     codec.\n");
+
+    System.out.println("-accuratedct = Use the most accurate DCT/IDCT algorithms available in the");
+    System.out.println("     underlying codec.\n");
+
+    System.exit(1);
+  }
+
+
+  public static void main(String argv) {
+
+    try {
+
+      TJScalingFactor scalingFactor = new TJScalingFactor(1, 1);
+      int outSubsamp = -1, outQual = -1;
+      TJTransform xform = new TJTransform();
+      boolean display = false;
+      int flags = 0;
+      int width, height;
+      String inFormat = "jpg", outFormat = "jpg";
+      BufferedImage img = null;
+      byte imgBuf = null;
+
+      if (argv.length < 2)
+        usage();
+
+      if (argv1.substring(0, 2).equalsIgnoreCase("-d"))
+        display = true;
+
+      /* Parse arguments. */
+      for (int i = 2; i < argv.length; i++) {
+        if (argvi.length() < 2)
+          continue;
+        else if (argvi.length() > 2 &&
+                 argvi.substring(0, 3).equalsIgnoreCase("-sc") &&
+                 i < argv.length - 1) {
+          int match = 0;
+          String scaleArg = argv++i.split("/");
+          if (scaleArg.length == 2) {
+            TJScalingFactor tempsf =
+              new TJScalingFactor(Integer.parseInt(scaleArg0),
+                                  Integer.parseInt(scaleArg1));
+            for (int j = 0; j < SCALING_FACTORS.length; j++) {
+              if (tempsf.equals(SCALING_FACTORSj)) {
+                scalingFactor = SCALING_FACTORSj;
+                match = 1;
+                break;
+              }
+            }
+          }
+          if (match != 1)
+            usage();
+        } else if (argvi.length() > 2 &&
+                   argvi.substring(0, 3).equalsIgnoreCase("-su") &&
+                   i < argv.length - 1) {
+          i++;
+          if (argvi.substring(0, 1).equalsIgnoreCase("g"))
+            outSubsamp = TJ.SAMP_GRAY;
+          else if (argvi.equals("444"))
+            outSubsamp = TJ.SAMP_444;
+          else if (argvi.equals("422"))
+            outSubsamp = TJ.SAMP_422;
+          else if (argvi.equals("420"))
+            outSubsamp = TJ.SAMP_420;
+          else
+            usage();
+        } else if (argvi.substring(0, 2).equalsIgnoreCase("-q") &&
+                   i < argv.length - 1) {
+          outQual = Integer.parseInt(argv++i);
+          if (outQual < 1 || outQual > 100)
+            usage();
+        } else if (argvi.substring(0, 2).equalsIgnoreCase("-g"))
+          xform.options |= TJTransform.OPT_GRAY;
+        else if (argvi.equalsIgnoreCase("-hflip"))
+          xform.op = TJTransform.OP_HFLIP;
+        else if (argvi.equalsIgnoreCase("-vflip"))
+          xform.op = TJTransform.OP_VFLIP;
+        else if (argvi.equalsIgnoreCase("-transpose"))
+          xform.op = TJTransform.OP_TRANSPOSE;
+        else if (argvi.equalsIgnoreCase("-transverse"))
+          xform.op = TJTransform.OP_TRANSVERSE;
+        else if (argvi.equalsIgnoreCase("-rot90"))
+          xform.op = TJTransform.OP_ROT90;
+        else if (argvi.equalsIgnoreCase("-rot180"))
+          xform.op = TJTransform.OP_ROT180;
+        else if (argvi.equalsIgnoreCase("-rot270"))
+          xform.op = TJTransform.OP_ROT270;
+        else if (argvi.equalsIgnoreCase("-custom"))
+          xform.cf = new TJExample();
+        else if (argvi.length() > 2 &&
+                 argvi.substring(0, 2).equalsIgnoreCase("-c") &&
+                 i < argv.length - 1) {
+          String cropArg = argv++i.split("x\\+");
+          if (cropArg.length != 4)
+            usage();
+          xform.width = Integer.parseInt(cropArg0);
+          xform.height = Integer.parseInt(cropArg1);
+          xform.x = Integer.parseInt(cropArg2);
+          xform.y = Integer.parseInt(cropArg3);
+          if (xform.x < 0 || xform.y < 0 || xform.width < 1 ||
+              xform.height < 1)
+            usage();
+          xform.options |= TJTransform.OPT_CROP;
+        } else if (argvi.substring(0, 2).equalsIgnoreCase("-d"))
+          display = true;
+        else if (argvi.equalsIgnoreCase("-fastupsample")) {
+          System.out.println("Using fast upsampling code");
+          flags |= TJ.FLAG_FASTUPSAMPLE;
+        } else if (argvi.equalsIgnoreCase("-fastdct")) {
+          System.out.println("Using fastest DCT/IDCT algorithm");
+          flags |= TJ.FLAG_FASTDCT;
+        } else if (argvi.equalsIgnoreCase("-accuratedct")) {
+          System.out.println("Using most accurate DCT/IDCT algorithm");
+          flags |= TJ.FLAG_ACCURATEDCT;
+        } else usage();
+      }
+
+      /* Determine input and output image formats based on file extensions. */
+      String inFileTokens = argv0.split("\\.");
+      if (inFileTokens.length > 1)
+        inFormat = inFileTokensinFileTokens.length - 1;
+      String outFileTokens;
+      if (display)
+        outFormat = "bmp";
+      else {
+        outFileTokens = argv1.split("\\.");
+        if (outFileTokens.length > 1)
+          outFormat = outFileTokensoutFileTokens.length - 1;
+      }
+
+      if (inFormat.equalsIgnoreCase("jpg")) {
+        /* Input image is a JPEG image.  Decompress and/or transform it. */
+        boolean doTransform = (xform.op != TJTransform.OP_NONE ||
+                               xform.options != 0 || xform.cf != null);
+
+        /* Read the JPEG file into memory. */
+        File jpegFile = new File(argv0);
+        FileInputStream fis = new FileInputStream(jpegFile);
+        int jpegSize = fis.available();
+        if (jpegSize < 1) {
+          System.out.println("Input file contains no data");
+          System.exit(1);
+        }
+        byte jpegBuf = new bytejpegSize;
+        fis.read(jpegBuf);
+        fis.close();
+
+        TJDecompressor tjd;
+        if (doTransform) {
+          /* Transform it. */
+          TJTransformer tjt = new TJTransformer(jpegBuf);
+          TJTransform xforms = new TJTransform1;
+          xforms0 = xform;
+          xforms0.options |= TJTransform.OPT_TRIM;
+          TJDecompressor tjds = tjt.transform(xforms, 0);
+          tjd = tjds0;
+          tjt.close();
+        } else
+          tjd = new TJDecompressor(jpegBuf);
+
+        width = tjd.getWidth();
+        height = tjd.getHeight();
+        int inSubsamp = tjd.getSubsamp();
+        int inColorspace = tjd.getColorspace();
+
+        System.out.println((doTransform ? "Transformed" : "Input") +
+                           " Image (jpg):  " + width + " x " + height +
+                           " pixels, " + SUBSAMP_NAMEinSubsamp +
+                           " subsampling, " + COLORSPACE_NAMEinColorspace);
+
+        if (outFormat.equalsIgnoreCase("jpg") && doTransform &&
+            scalingFactor.isOne() && outSubsamp < 0 && outQual < 0) {
+          /* Input image has been transformed, and no re-compression options
+             have been selected.  Write the transformed image to disk and
+             exit. */
+          File outFile = new File(argv1);
+          FileOutputStream fos = new FileOutputStream(outFile);
+          fos.write(tjd.getJPEGBuf(), 0, tjd.getJPEGSize());
+          fos.close();
+          System.exit(0);
+        }
+
+        /* Scaling and/or a non-JPEG output image format and/or compression
+           options have been selected, so we need to decompress the
+           input/transformed image. */
+        width = scalingFactor.getScaled(width);
+        height = scalingFactor.getScaled(height);
+        if (outSubsamp < 0)
+          outSubsamp = inSubsamp;
+
+        if (!outFormat.equalsIgnoreCase("jpg"))
+          img = tjd.decompress(width, height, BufferedImage.TYPE_INT_RGB,
+                               flags);
+        else
+          imgBuf = tjd.decompress(width, 0, height, TJ.PF_BGRX, flags);
+        tjd.close();
+      } else {
+        /* Input image is not a JPEG image.  Load it into memory. */
+        img = ImageIO.read(new File(argv0));
+        if (img == null)
+          throw new Exception("Input image type not supported.");
+        width = img.getWidth();
+        height = img.getHeight();
+        if (outSubsamp < 0) {
+          if (img.getType() == BufferedImage.TYPE_BYTE_GRAY)
+            outSubsamp = TJ.SAMP_GRAY;
+          else
+            outSubsamp = DEFAULT_SUBSAMP;
+        }
+        System.out.println("Input Image:  " + width + " x " + height +
+                           " pixels");
+      }
+      System.gc();
+      if (!display)
+        System.out.print("Output Image (" + outFormat + "):  " + width +
+                         " x " + height + " pixels");
+
+      if (display) {
+        /* Display the uncompressed image */
+        ImageIcon icon = new ImageIcon(img);
+        JLabel label = new JLabel(icon, JLabel.CENTER);
+        JOptionPane.showMessageDialog(null, label, "Output Image",
+                                      JOptionPane.PLAIN_MESSAGE);
+      } else if (outFormat.equalsIgnoreCase("jpg")) {
+        /* Output image format is JPEG.  Compress the uncompressed image. */
+        if (outQual < 0)
+          outQual = DEFAULT_QUALITY;
+        System.out.println(", " + SUBSAMP_NAMEoutSubsamp +
+                           " subsampling, quality = " + outQual);
+
+        TJCompressor tjc = new TJCompressor();
+        tjc.setSubsamp(outSubsamp);
+        tjc.setJPEGQuality(outQual);
+        if (img != null)
+          tjc.setSourceImage(img, 0, 0, 0, 0);
+        else
+          tjc.setSourceImage(imgBuf, 0, 0, width, 0, height, TJ.PF_BGRX);
+        byte jpegBuf = tjc.compress(flags);
+        int jpegSize = tjc.getCompressedSize();
+        tjc.close();
+
+        /* Write the JPEG image to disk. */
+        File outFile = new File(argv1);
+        FileOutputStream fos = new FileOutputStream(outFile);
+        fos.write(jpegBuf, 0, jpegSize);
+        fos.close();
+      } else {
+        /* Output image format is not JPEG.  Save the uncompressed image
+           directly to disk. */
+        System.out.print("\n");
+        File outFile = new File(argv1);
+        ImageIO.write(img, outFormat, outFile);
+      }
+
+    } catch (Exception e) {
+      e.printStackTrace();
+      System.exit(-1);
+    }
+  }
+
+  static final TJScalingFactor SCALING_FACTORS =
+    TJ.getScalingFactors();
+};

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/TJUnitTest.java Added

@@ -0,0 +1,960 @@
+/*
+ * Copyright (C)2011-2018 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * This program tests the various code paths in the TurboJPEG JNI Wrapper
+ */
+
+import java.io.*;
+import java.util.*;
+import java.awt.image.*;
+import javax.imageio.*;
+import java.nio.*;
+import org.libjpegturbo.turbojpeg.*;
+
+@SuppressWarnings("checkstyle:JavadocType")
+final class TJUnitTest {
+
+  private TJUnitTest() {}
+
+  static final String CLASS_NAME =
+    new TJUnitTest().getClass().getName();
+
+  static void usage() {
+    System.out.println("\nUSAGE: java " + CLASS_NAME + " options\n");
+    System.out.println("Options:");
+    System.out.println("-yuv = test YUV encoding/decoding support");
+    System.out.println("-noyuvpad = do not pad each line of each Y, U, and V plane to the nearest");
+    System.out.println("            4-byte boundary");
+    System.out.println("-bi = test BufferedImage support\n");
+    System.exit(1);
+  }
+
+  static final String SUBNAME_LONG = {
+    "4:4:4", "4:2:2", "4:2:0", "GRAY", "4:4:0", "4:1:1"
+  };
+  static final String SUBNAME = {
+    "444", "422", "420", "GRAY", "440", "411"
+  };
+
+  static final String PIXFORMATSTR = {
+    "RGB", "BGR", "RGBX", "BGRX", "XBGR", "XRGB", "Grayscale",
+    "RGBA", "BGRA", "ABGR", "ARGB", "CMYK"
+  };
+
+  static final int FORMATS_3BYTE = {
+    TJ.PF_RGB, TJ.PF_BGR
+  };
+  static final int FORMATS_3BYTEBI = {
+    BufferedImage.TYPE_3BYTE_BGR
+  };
+  static final int FORMATS_4BYTE = {
+    TJ.PF_RGBX, TJ.PF_BGRX, TJ.PF_XBGR, TJ.PF_XRGB, TJ.PF_CMYK
+  };
+  static final int FORMATS_4BYTEBI = {
+    BufferedImage.TYPE_INT_BGR, BufferedImage.TYPE_INT_RGB,
+    BufferedImage.TYPE_4BYTE_ABGR, BufferedImage.TYPE_4BYTE_ABGR_PRE,
+    BufferedImage.TYPE_INT_ARGB, BufferedImage.TYPE_INT_ARGB_PRE
+  };
+  static final int FORMATS_GRAY = {
+    TJ.PF_GRAY
+  };
+  static final int FORMATS_GRAYBI = {
+    BufferedImage.TYPE_BYTE_GRAY
+  };
+  static final int FORMATS_RGB = {
+    TJ.PF_RGB
+  };
+
+  private static boolean doYUV = false;
+  private static int pad = 4;
+  private static boolean bi = false;
+
+  private static int exitStatus = 0;
+
+  static int biTypePF(int biType) {
+    ByteOrder byteOrder = ByteOrder.nativeOrder();
+    switch (biType) {
+    case BufferedImage.TYPE_3BYTE_BGR:
+      return TJ.PF_BGR;
+    case BufferedImage.TYPE_4BYTE_ABGR:
+    case BufferedImage.TYPE_4BYTE_ABGR_PRE:
+      return TJ.PF_ABGR;
+    case BufferedImage.TYPE_BYTE_GRAY:
+      return TJ.PF_GRAY;
+    case BufferedImage.TYPE_INT_BGR:
+      return TJ.PF_RGBX;
+    case BufferedImage.TYPE_INT_RGB:
+      return TJ.PF_BGRX;
+    case BufferedImage.TYPE_INT_ARGB:
+    case BufferedImage.TYPE_INT_ARGB_PRE:
+      return TJ.PF_BGRA;
+    default:
+      return 0;
+    }
+  }
+
+  static String biTypeStr(int biType) {
+    switch (biType) {
+    case BufferedImage.TYPE_3BYTE_BGR:
+      return "3BYTE_BGR";
+    case BufferedImage.TYPE_4BYTE_ABGR:
+      return "4BYTE_ABGR";
+    case BufferedImage.TYPE_4BYTE_ABGR_PRE:
+      return "4BYTE_ABGR_PRE";
+    case BufferedImage.TYPE_BYTE_GRAY:
+      return "BYTE_GRAY";
+    case BufferedImage.TYPE_INT_BGR:
+      return "INT_BGR";
+    case BufferedImage.TYPE_INT_RGB:
+      return "INT_RGB";
+    case BufferedImage.TYPE_INT_ARGB:
+      return "INT_ARGB";
+    case BufferedImage.TYPE_INT_ARGB_PRE:
+      return "INT_ARGB_PRE";
+    default:
+      return "Unknown";
+    }
+  }
+
+  static void initBuf(byte buf, int w, int pitch, int h, int pf, int flags)
+                      throws Exception {
+    int roffset = TJ.getRedOffset(pf);
+    int goffset = TJ.getGreenOffset(pf);
+    int boffset = TJ.getBlueOffset(pf);
+    int aoffset = TJ.getAlphaOffset(pf);
+    int ps = TJ.getPixelSize(pf);
+    int index, row, col, halfway = 16;
+
+    if (pf == TJ.PF_GRAY) {
+      Arrays.fill(buf, (byte)0);
+      for (row = 0; row < h; row++) {
+        for (col = 0; col < w; col++) {
+          if ((flags & TJ.FLAG_BOTTOMUP) != 0)
+            index = pitch * (h - row - 1) + col;
+          else
+            index = pitch * row + col;
+          if (((row / 8) + (col / 8)) % 2 == 0)
+            bufindex = (row < halfway) ? (byte)255 : 0;
+          else
+            bufindex = (row < halfway) ? 76 : (byte)226;
+        }
+      }
+      return;
+    }
+    if (pf == TJ.PF_CMYK) {
+      Arrays.fill(buf, (byte)255);
+      for (row = 0; row < h; row++) {
+        for (col = 0; col < w; col++) {
+          if ((flags & TJ.FLAG_BOTTOMUP) != 0)
+            index = (h - row - 1) * w + col;
+          else
+            index = row * w + col;
+          if (((row / 8) + (col / 8)) % 2 == 0) {
+            if (row >= halfway) bufindex * ps + 3 = 0;
+          } else {
+            bufindex * ps + 2 = 0;
+            if (row < halfway)
+              bufindex * ps + 1 = 0;
+          }
+        }
+      }
+      return;
+    }
+
+    Arrays.fill(buf, (byte)0);
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        if ((flags & TJ.FLAG_BOTTOMUP) != 0)
+          index = pitch * (h - row - 1) + col * ps;
+        else
+          index = pitch * row + col * ps;
+        if (((row / 8) + (col / 8)) % 2 == 0) {
+          if (row < halfway) {
+            bufindex + roffset = (byte)255;
+            bufindex + goffset = (byte)255;
+            bufindex + boffset = (byte)255;
+          }
+        } else {
+          bufindex + roffset = (byte)255;
+          if (row >= halfway)
+            bufindex + goffset = (byte)255;
+        }
+        if (aoffset >= 0)
+          bufindex + aoffset = (byte)255;
+      }
+    }
+  }
+
+  static void initIntBuf(int buf, int w, int pitch, int h, int pf, int flags)
+                         throws Exception {
+    int rshift = TJ.getRedOffset(pf) * 8;
+    int gshift = TJ.getGreenOffset(pf) * 8;
+    int bshift = TJ.getBlueOffset(pf) * 8;
+    int ashift = TJ.getAlphaOffset(pf) * 8;
+    int index, row, col, halfway = 16;
+
+    Arrays.fill(buf, 0);
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        if ((flags & TJ.FLAG_BOTTOMUP) != 0)
+          index = pitch * (h - row - 1) + col;
+        else
+          index = pitch * row + col;
+        if (((row / 8) + (col / 8)) % 2 == 0) {
+          if (row < halfway) {
+            bufindex |= (255 << rshift);
+            bufindex |= (255 << gshift);
+            bufindex |= (255 << bshift);
+          }
+        } else {
+          bufindex |= (255 << rshift);
+          if (row >= halfway)
+            bufindex |= (255 << gshift);
+        }
+        if (ashift >= 0)
+          bufindex |= (255 << ashift);
+      }
+    }
+  }
+
+  static void initImg(BufferedImage img, int pf, int flags) throws Exception {
+    WritableRaster wr = img.getRaster();
+    int imgType = img.getType();
+
+    if (imgType == BufferedImage.TYPE_INT_RGB ||
+        imgType == BufferedImage.TYPE_INT_BGR ||
+        imgType == BufferedImage.TYPE_INT_ARGB ||
+        imgType == BufferedImage.TYPE_INT_ARGB_PRE) {
+      SinglePixelPackedSampleModel sm =
+        (SinglePixelPackedSampleModel)img.getSampleModel();
+      int pitch = sm.getScanlineStride();
+      DataBufferInt db = (DataBufferInt)wr.getDataBuffer();
+      int buf = db.getData();
+      initIntBuf(buf, img.getWidth(), pitch, img.getHeight(), pf, flags);
+    } else {
+      ComponentSampleModel sm = (ComponentSampleModel)img.getSampleModel();
+      int pitch = sm.getScanlineStride();
+      DataBufferByte db = (DataBufferByte)wr.getDataBuffer();
+      byte buf = db.getData();
+      initBuf(buf, img.getWidth(), pitch, img.getHeight(), pf, flags);
+    }
+  }
+
+  static void checkVal(int row, int col, int v, String vname, int cv)
+                       throws Exception {
+    v = (v < 0) ? v + 256 : v;
+    if (v < cv - 1 || v > cv + 1) {
+      throw new Exception("Comp. " + vname + " at " + row + "," + col +
+                          " should be " + cv + ", not " + v);
+    }
+  }
+
+  static void checkVal0(int row, int col, int v, String vname)
+                        throws Exception {
+    v = (v < 0) ? v + 256 : v;
+    if (v > 1) {
+      throw new Exception("Comp. " + vname + " at " + row + "," + col +
+                          " should be 0, not " + v);
+    }
+  }
+
+  static void checkVal255(int row, int col, int v, String vname)
+                          throws Exception {
+    v = (v < 0) ? v + 256 : v;
+    if (v < 254) {
+      throw new Exception("Comp. " + vname + " at " + row + "," + col +
+                          " should be 255, not " + v);
+    }
+  }
+
+  static int checkBuf(byte buf, int w, int pitch, int h, int pf, int subsamp,
+                      TJScalingFactor sf, int flags) throws Exception {
+    int roffset = TJ.getRedOffset(pf);
+    int goffset = TJ.getGreenOffset(pf);
+    int boffset = TJ.getBlueOffset(pf);
+    int aoffset = TJ.getAlphaOffset(pf);
+    int ps = TJ.getPixelSize(pf);
+    int index, row, col, retval = 1;
+    int halfway = 16 * sf.getNum() / sf.getDenom();
+    int blockSize = 8 * sf.getNum() / sf.getDenom();
+
+    try {
+
+      if (pf == TJ.PF_GRAY)
+        roffset = goffset = boffset = 0;
+
+      if (pf == TJ.PF_CMYK) {
+        for (row = 0; row < h; row++) {
+          for (col = 0; col < w; col++) {
+            if ((flags & TJ.FLAG_BOTTOMUP) != 0)
+              index = (h - row - 1) * w + col;
+            else
+              index = row * w + col;
+            byte c = bufindex * ps;
+            byte m = bufindex * ps + 1;
+            byte y = bufindex * ps + 2;
+            byte k = bufindex * ps + 3;
+            checkVal255(row, col, c, "C");
+            if (((row / blockSize) + (col / blockSize)) % 2 == 0) {
+              checkVal255(row, col, m, "M");
+              checkVal255(row, col, y, "Y");
+              if (row < halfway)
+                checkVal255(row, col, k, "K");
+              else
+                checkVal0(row, col, k, "K");
+            } else {
+              checkVal0(row, col, y, "Y");
+              checkVal255(row, col, k, "K");
+              if (row < halfway)
+                checkVal0(row, col, m, "M");
+              else
+                checkVal255(row, col, m, "M");
+            }
+          }
+        }
+        return 1;
+      }
+
+      for (row = 0; row < halfway; row++) {
+        for (col = 0; col < w; col++) {
+          if ((flags & TJ.FLAG_BOTTOMUP) != 0)
+            index = pitch * (h - row - 1) + col * ps;
+          else
+            index = pitch * row + col * ps;
+          byte r = bufindex + roffset;
+          byte g = bufindex + goffset;
+          byte b = bufindex + boffset;
+          byte a = aoffset >= 0 ? bufindex + aoffset : (byte)255;
+          if (((row / blockSize) + (col / blockSize)) % 2 == 0) {
+            if (row < halfway) {
+              checkVal255(row, col, r, "R");
+              checkVal255(row, col, g, "G");
+              checkVal255(row, col, b, "B");
+            } else {
+              checkVal0(row, col, r, "R");
+              checkVal0(row, col, g, "G");
+              checkVal0(row, col, b, "B");
+            }
+          } else {
+            if (subsamp == TJ.SAMP_GRAY) {
+              if (row < halfway) {
+                checkVal(row, col, r, "R", 76);
+                checkVal(row, col, g, "G", 76);
+                checkVal(row, col, b, "B", 76);
+              } else {
+                checkVal(row, col, r, "R", 226);
+                checkVal(row, col, g, "G", 226);
+                checkVal(row, col, b, "B", 226);
+              }
+            } else {
+              checkVal255(row, col, r, "R");
+              if (row < halfway) {
+                checkVal0(row, col, g, "G");
+              } else {
+                checkVal255(row, col, g, "G");
+              }
+              checkVal0(row, col, b, "B");
+            }
+          }
+          checkVal255(row, col, a, "A");
+        }
+      }
+    } catch (Exception e) {
+      System.out.println("\n" + e.getMessage());
+      retval = 0;
+    }
+
+    if (retval == 0) {
+      for (row = 0; row < h; row++) {
+        for (col = 0; col < w; col++) {
+          if (pf == TJ.PF_CMYK) {
+            int c = bufpitch * row + col * ps;
+            int m = bufpitch * row + col * ps + 1;
+            int y = bufpitch * row + col * ps + 2;
+            int k = bufpitch * row + col * ps + 3;
+            if (c < 0) c += 256;
+            if (m < 0) m += 256;
+            if (y < 0) y += 256;
+            if (k < 0) k += 256;
+            System.out.format("%3d/%3d/%3d/%3d ", c, m, y, k);
+          } else {
+            int r = bufpitch * row + col * ps + roffset;
+            int g = bufpitch * row + col * ps + goffset;
+            int b = bufpitch * row + col * ps + boffset;
+            if (r < 0) r += 256;
+            if (g < 0) g += 256;
+            if (b < 0) b += 256;
+            System.out.format("%3d/%3d/%3d ", r, g, b);
+          }
+        }
+        System.out.print("\n");
+      }
+    }
+    return retval;
+  }
+
+  static int checkIntBuf(int buf, int w, int pitch, int h, int pf,
+                         int subsamp, TJScalingFactor sf, int flags)
+                         throws Exception {
+    int rshift = TJ.getRedOffset(pf) * 8;
+    int gshift = TJ.getGreenOffset(pf) * 8;
+    int bshift = TJ.getBlueOffset(pf) * 8;
+    int ashift = TJ.getAlphaOffset(pf) * 8;
+    int index, row, col, retval = 1;
+    int halfway = 16 * sf.getNum() / sf.getDenom();
+    int blockSize = 8 * sf.getNum() / sf.getDenom();
+
+    try {
+      for (row = 0; row < halfway; row++) {
+        for (col = 0; col < w; col++) {
+          if ((flags & TJ.FLAG_BOTTOMUP) != 0)
+            index = pitch * (h - row - 1) + col;
+          else
+            index = pitch * row + col;
+          int r = (bufindex >> rshift) & 0xFF;
+          int g = (bufindex >> gshift) & 0xFF;
+          int b = (bufindex >> bshift) & 0xFF;
+          int a = ashift >= 0 ? (bufindex >> ashift) & 0xFF : 255;
+          if (((row / blockSize) + (col / blockSize)) % 2 == 0) {
+            if (row < halfway) {
+              checkVal255(row, col, r, "R");
+              checkVal255(row, col, g, "G");
+              checkVal255(row, col, b, "B");
+            } else {
+              checkVal0(row, col, r, "R");
+              checkVal0(row, col, g, "G");
+              checkVal0(row, col, b, "B");
+            }
+          } else {
+            if (subsamp == TJ.SAMP_GRAY) {
+              if (row < halfway) {
+                checkVal(row, col, r, "R", 76);
+                checkVal(row, col, g, "G", 76);
+                checkVal(row, col, b, "B", 76);
+              } else {
+                checkVal(row, col, r, "R", 226);
+                checkVal(row, col, g, "G", 226);
+                checkVal(row, col, b, "B", 226);
+              }
+            } else {
+              checkVal255(row, col, r, "R");
+              if (row < halfway) {
+                checkVal0(row, col, g, "G");
+              } else {
+                checkVal255(row, col, g, "G");
+              }
+              checkVal0(row, col, b, "B");
+            }
+          }
+          checkVal255(row, col, a, "A");
+        }
+      }
+    } catch (Exception e) {
+      System.out.println("\n" + e.getMessage());
+      retval = 0;
+    }
+
+    if (retval == 0) {
+      for (row = 0; row < h; row++) {
+        for (col = 0; col < w; col++) {
+          int r = (bufpitch * row + col >> rshift) & 0xFF;
+          int g = (bufpitch * row + col >> gshift) & 0xFF;
+          int b = (bufpitch * row + col >> bshift) & 0xFF;
+          if (r < 0) r += 256;
+          if (g < 0) g += 256;
+          if (b < 0) b += 256;
+          System.out.format("%3d/%3d/%3d ", r, g, b);
+        }
+        System.out.print("\n");
+      }
+    }
+    return retval;
+  }
+
+  static int checkImg(BufferedImage img, int pf, int subsamp,
+                      TJScalingFactor sf, int flags) throws Exception {
+    WritableRaster wr = img.getRaster();
+    int imgType = img.getType();
+    if (imgType == BufferedImage.TYPE_INT_RGB ||
+        imgType == BufferedImage.TYPE_INT_BGR ||
+        imgType == BufferedImage.TYPE_INT_ARGB ||
+        imgType == BufferedImage.TYPE_INT_ARGB_PRE) {
+      SinglePixelPackedSampleModel sm =
+        (SinglePixelPackedSampleModel)img.getSampleModel();
+      int pitch = sm.getScanlineStride();
+      DataBufferInt db = (DataBufferInt)wr.getDataBuffer();
+      int buf = db.getData();
+      return checkIntBuf(buf, img.getWidth(), pitch, img.getHeight(), pf,
+                         subsamp, sf, flags);
+    } else {
+      ComponentSampleModel sm = (ComponentSampleModel)img.getSampleModel();
+      int pitch = sm.getScanlineStride();
+      DataBufferByte db = (DataBufferByte)wr.getDataBuffer();
+      byte buf = db.getData();
+      return checkBuf(buf, img.getWidth(), pitch, img.getHeight(), pf, subsamp,
+                      sf, flags);
+    }
+  }
+
+  static int pad(int v, int p) {
+    return ((v + (p) - 1) & (~((p) - 1)));
+  }
+
+  static int checkBufYUV(byte buf, int size, int w, int h, int subsamp,
+                         TJScalingFactor sf) throws Exception {
+    int row, col;
+    int hsf = TJ.getMCUWidth(subsamp) / 8, vsf = TJ.getMCUHeight(subsamp) / 8;
+    int pw = pad(w, hsf), ph = pad(h, vsf);
+    int cw = pw / hsf, ch = ph / vsf;
+    int ypitch = pad(pw, pad), uvpitch = pad(cw, pad);
+    int retval = 1;
+    int correctsize = ypitch * ph +
+                      (subsamp == TJ.SAMP_GRAY ? 0 : uvpitch * ch * 2);
+    int halfway = 16 * sf.getNum() / sf.getDenom();
+    int blockSize = 8 * sf.getNum() / sf.getDenom();
+
+    try {
+      if (size != correctsize)
+        throw new Exception("Incorrect size " + size + ".  Should be " +
+                            correctsize);
+
+      for (row = 0; row < ph; row++) {
+        for (col = 0; col < pw; col++) {
+          byte y = bufypitch * row + col;
+          if (((row / blockSize) + (col / blockSize)) % 2 == 0) {
+            if (row < halfway)
+              checkVal255(row, col, y, "Y");
+            else
+              checkVal0(row, col, y, "Y");
+          } else {
+            if (row < halfway)
+              checkVal(row, col, y, "Y", 76);
+            else
+              checkVal(row, col, y, "Y", 226);
+          }
+        }
+      }
+      if (subsamp != TJ.SAMP_GRAY) {
+        halfway = 16 / vsf * sf.getNum() / sf.getDenom();
+        for (row = 0; row < ch; row++) {
+          for (col = 0; col < cw; col++) {
+            byte u = bufypitch * ph + (uvpitch * row + col),
+                 v = bufypitch * ph + uvpitch * ch + (uvpitch * row + col);
+            if (((row * vsf / blockSize) + (col * hsf / blockSize)) % 2 == 0) {
+              checkVal(row, col, u, "U", 128);
+              checkVal(row, col, v, "V", 128);
+            } else {
+              if (row < halfway) {
+                checkVal(row, col, u, "U", 85);
+                checkVal255(row, col, v, "V");
+              } else {
+                checkVal0(row, col, u, "U");
+                checkVal(row, col, v, "V", 149);
+              }
+            }
+          }
+        }
+      }
+    } catch (Exception e) {
+      System.out.println("\n" + e.getMessage());
+      retval = 0;
+    }
+
+    if (retval == 0) {
+      for (row = 0; row < ph; row++) {
+        for (col = 0; col < pw; col++) {
+          int y = bufypitch * row + col;
+          if (y < 0) y += 256;
+          System.out.format("%3d ", y);
+        }
+        System.out.print("\n");
+      }
+      System.out.print("\n");
+      for (row = 0; row < ch; row++) {
+        for (col = 0; col < cw; col++) {
+          int u = bufypitch * ph + (uvpitch * row + col);
+          if (u < 0) u += 256;
+          System.out.format("%3d ", u);
+        }
+        System.out.print("\n");
+      }
+      System.out.print("\n");
+      for (row = 0; row < ch; row++) {
+        for (col = 0; col < cw; col++) {
+          int v = bufypitch * ph + uvpitch * ch + (uvpitch * row + col);
+          if (v < 0) v += 256;
+          System.out.format("%3d ", v);
+        }
+        System.out.print("\n");
+      }
+    }
+
+    return retval;
+  }
+
+  static void writeJPEG(byte jpegBuf, int jpegBufSize, String filename)
+                        throws Exception {
+    File file = new File(filename);
+    FileOutputStream fos = new FileOutputStream(file);
+    fos.write(jpegBuf, 0, jpegBufSize);
+    fos.close();
+  }
+
+  static int compTest(TJCompressor tjc, byte dstBuf, int w, int h, int pf,
+                      String baseName, int subsamp, int jpegQual, int flags)
+                      throws Exception {
+    String tempStr;
+    byte srcBuf = null;
+    BufferedImage img = null;
+    String pfStr, pfStrLong;
+    String buStr = (flags & TJ.FLAG_BOTTOMUP) != 0 ? "BU" : "TD";
+    String buStrLong = (flags & TJ.FLAG_BOTTOMUP) != 0 ?
+                       "Bottom-Up" : "Top-Down ";
+    int size = 0, ps, imgType = pf;
+
+    if (bi) {
+      pf = biTypePF(imgType);
+      pfStr = biTypeStr(imgType);
+      pfStrLong = pfStr + " (" + PIXFORMATSTRpf + ")";
+    } else {
+      pfStr = PIXFORMATSTRpf;
+      pfStrLong = pfStr;
+    }
+    ps =  TJ.getPixelSize(pf);
+
+    if (bi) {
+      img = new BufferedImage(w, h, imgType);
+      initImg(img, pf, flags);
+      tempStr = baseName + "_enc_" + pfStr + "_" + buStr + "_" +
+                SUBNAMEsubsamp + "_Q" + jpegQual + ".png";
+      File file = new File(tempStr);
+      ImageIO.write(img, "png", file);
+      tjc.setSourceImage(img, 0, 0, 0, 0);
+    } else {
+      srcBuf = new bytew * h * ps + 1;
+      initBuf(srcBuf, w, w * ps, h, pf, flags);
+      tjc.setSourceImage(srcBuf, 0, 0, w, 0, h, pf);
+    }
+    Arrays.fill(dstBuf, (byte)0);
+
+    tjc.setSubsamp(subsamp);
+    tjc.setJPEGQuality(jpegQual);
+    if (doYUV) {
+      System.out.format("%s %s -> YUV %s ... ", pfStrLong, buStrLong,
+                        SUBNAME_LONGsubsamp);
+      YUVImage yuvImage = tjc.encodeYUV(pad, flags);
+      if (checkBufYUV(yuvImage.getBuf(), yuvImage.getSize(), w, h, subsamp,
+                      new TJScalingFactor(1, 1)) == 1)
+        System.out.print("Passed.\n");
+      else {
+        System.out.print("FAILED!\n");
+        exitStatus = -1;
+      }
+
+      System.out.format("YUV %s %s -> JPEG Q%d ... ", SUBNAME_LONGsubsamp,
+                        buStrLong, jpegQual);
+      tjc.setSourceImage(yuvImage);
+    } else {
+      System.out.format("%s %s -> %s Q%d ... ", pfStrLong, buStrLong,
+                        SUBNAME_LONGsubsamp, jpegQual);
+    }
+    tjc.compress(dstBuf, flags);
+    size = tjc.getCompressedSize();
+
+    tempStr = baseName + "_enc_" + pfStr + "_" + buStr + "_" +
+              SUBNAMEsubsamp + "_Q" + jpegQual + ".jpg";
+    writeJPEG(dstBuf, size, tempStr);
+    System.out.println("Done.\n  Result in " + tempStr);
+
+    return size;
+  }
+
+  static void decompTest(TJDecompressor tjd, byte jpegBuf, int jpegSize,
+                         int w, int h, int pf, String baseName, int subsamp,
+                         int flags, TJScalingFactor sf) throws Exception {
+    String pfStr, pfStrLong, tempStr;
+    String buStrLong = (flags & TJ.FLAG_BOTTOMUP) != 0 ?
+                       "Bottom-Up" : "Top-Down ";
+    int scaledWidth = sf.getScaled(w);
+    int scaledHeight = sf.getScaled(h);
+    int temp1, temp2, imgType = pf;
+    BufferedImage img = null;
+    byte dstBuf = null;
+
+    if (bi) {
+      pf = biTypePF(imgType);
+      pfStr = biTypeStr(imgType);
+      pfStrLong = pfStr + " (" + PIXFORMATSTRpf + ")";
+    } else {
+      pfStr = PIXFORMATSTRpf;
+      pfStrLong = pfStr;
+    }
+
+    tjd.setSourceImage(jpegBuf, jpegSize);
+    if (tjd.getWidth() != w || tjd.getHeight() != h ||
+        tjd.getSubsamp() != subsamp)
+      throw new Exception("Incorrect JPEG header");
+
+    temp1 = scaledWidth;
+    temp2 = scaledHeight;
+    temp1 = tjd.getScaledWidth(temp1, temp2);
+    temp2 = tjd.getScaledHeight(temp1, temp2);
+    if (temp1 != scaledWidth || temp2 != scaledHeight)
+      throw new Exception("Scaled size mismatch");
+
+    if (doYUV) {
+      System.out.format("JPEG -> YUV %s ", SUBNAME_LONGsubsamp);
+      if (!sf.isOne())
+        System.out.format("%d/%d ... ", sf.getNum(), sf.getDenom());
+      else System.out.print("... ");
+      YUVImage yuvImage = tjd.decompressToYUV(scaledWidth, pad, scaledHeight,
+                                              flags);
+      if (checkBufYUV(yuvImage.getBuf(), yuvImage.getSize(), scaledWidth,
+                      scaledHeight, subsamp, sf) == 1)
+        System.out.print("Passed.\n");
+      else {
+        System.out.print("FAILED!\n");  exitStatus = -1;
+      }
+
+      System.out.format("YUV %s -> %s %s ... ", SUBNAME_LONGsubsamp,
+                        pfStrLong, buStrLong);
+      tjd.setSourceImage(yuvImage);
+    } else {
+      System.out.format("JPEG -> %s %s ", pfStrLong, buStrLong);
+      if (!sf.isOne())
+        System.out.format("%d/%d ... ", sf.getNum(), sf.getDenom());
+      else System.out.print("... ");
+    }
+    if (bi)
+      img = tjd.decompress(scaledWidth, scaledHeight, imgType, flags);
+    else
+      dstBuf = tjd.decompress(scaledWidth, 0, scaledHeight, pf, flags);
+
+    if (bi) {
+      tempStr = baseName + "_dec_" + pfStr + "_" +
+                (((flags & TJ.FLAG_BOTTOMUP) != 0) ? "BU" : "TD") + "_" +
+                SUBNAMEsubsamp + "_" +
+                (double)sf.getNum() / (double)sf.getDenom() + "x" + ".png";
+      File file = new File(tempStr);
+      ImageIO.write(img, "png", file);
+    }
+
+    if ((bi && checkImg(img, pf, subsamp, sf, flags) == 1) ||
+        (!bi && checkBuf(dstBuf, scaledWidth,
+                         scaledWidth * TJ.getPixelSize(pf), scaledHeight, pf,
+                         subsamp, sf, flags) == 1))
+      System.out.print("Passed.\n");
+    else {
+      System.out.print("FAILED!\n");
+      exitStatus = -1;
+    }
+  }
+
+  static void decompTest(TJDecompressor tjd, byte jpegBuf, int jpegSize,
+                         int w, int h, int pf, String baseName, int subsamp,
+                         int flags) throws Exception {
+    int i;
+    TJScalingFactor sf = TJ.getScalingFactors();
+    for (i = 0; i < sf.length; i++) {
+      int num = sfi.getNum();
+      int denom = sfi.getDenom();
+      if (subsamp == TJ.SAMP_444 || subsamp == TJ.SAMP_GRAY ||
+          (subsamp == TJ.SAMP_411 && num == 1 &&
+           (denom == 2 || denom == 1)) ||
+          (subsamp != TJ.SAMP_411 && num == 1 &&
+           (denom == 4 || denom == 2 || denom == 1)))
+        decompTest(tjd, jpegBuf, jpegSize, w, h, pf, baseName, subsamp,
+                   flags, sfi);
+    }
+  }
+
+  static void doTest(int w, int h, int formats, int subsamp, String baseName)
+                     throws Exception {
+    TJCompressor tjc = null;
+    TJDecompressor tjd = null;
+    int size;
+    byte dstBuf;
+
+    dstBuf = new byteTJ.bufSize(w, h, subsamp);
+
+    try {
+      tjc = new TJCompressor();
+      tjd = new TJDecompressor();
+
+      for (int pf : formats) {
+        if (pf < 0) continue;
+        for (int i = 0; i < 2; i++) {
+          int flags = 0;
+          if (subsamp == TJ.SAMP_422 || subsamp == TJ.SAMP_420 ||
+              subsamp == TJ.SAMP_440 || subsamp == TJ.SAMP_411)
+            flags |= TJ.FLAG_FASTUPSAMPLE;
+          if (i == 1)
+            flags |= TJ.FLAG_BOTTOMUP;
+          size = compTest(tjc, dstBuf, w, h, pf, baseName, subsamp, 100,
+                          flags);
+          decompTest(tjd, dstBuf, size, w, h, pf, baseName, subsamp, flags);
+          if (pf >= TJ.PF_RGBX && pf <= TJ.PF_XRGB && !bi) {
+            System.out.print("\n");
+            decompTest(tjd, dstBuf, size, w, h, pf + (TJ.PF_RGBA - TJ.PF_RGBX),
+                       baseName, subsamp, flags);
+          }
+          System.out.print("\n");
+        }
+      }
+      System.out.print("--------------------\n\n");
+    } catch (Exception e) {
+      if (tjc != null) tjc.close();
+      if (tjd != null) tjd.close();
+      throw e;
+    }
+    if (tjc != null) tjc.close();
+    if (tjd != null) tjd.close();
+  }
+
+  static void bufSizeTest() throws Exception {
+    int w, h, i, subsamp;
+    byte srcBuf, dstBuf = null;
+    YUVImage dstImage = null;
+    TJCompressor tjc = null;
+    Random r = new Random();
+
+    try {
+      tjc = new TJCompressor();
+      System.out.println("Buffer size regression test");
+      for (subsamp = 0; subsamp < TJ.NUMSAMP; subsamp++) {
+        for (w = 1; w < 48; w++) {
+          int maxh = (w == 1) ? 2048 : 48;
+          for (h = 1; h < maxh; h++) {
+            if (h % 100 == 0)
+              System.out.format("%04d x %04d\b\b\b\b\b\b\b\b\b\b\b", w, h);
+            srcBuf = new bytew * h * 4;
+            if (doYUV)
+              dstImage = new YUVImage(w, pad, h, subsamp);
+            else
+              dstBuf = new byteTJ.bufSize(w, h, subsamp);
+            for (i = 0; i < w * h * 4; i++) {
+              srcBufi = (byte)(r.nextInt(2) * 255);
+            }
+            tjc.setSourceImage(srcBuf, 0, 0, w, 0, h, TJ.PF_BGRX);
+            tjc.setSubsamp(subsamp);
+            tjc.setJPEGQuality(100);
+            if (doYUV)
+              tjc.encodeYUV(dstImage, 0);
+            else
+              tjc.compress(dstBuf, 0);
+
+            srcBuf = new byteh * w * 4;
+            if (doYUV)
+              dstImage = new YUVImage(h, pad, w, subsamp);
+            else
+              dstBuf = new byteTJ.bufSize(h, w, subsamp);
+            for (i = 0; i < h * w * 4; i++) {
+              srcBufi = (byte)(r.nextInt(2) * 255);
+            }
+            tjc.setSourceImage(srcBuf, 0, 0, h, 0, w, TJ.PF_BGRX);
+            if (doYUV)
+              tjc.encodeYUV(dstImage, 0);
+            else
+              tjc.compress(dstBuf, 0);
+          }
+          dstImage = null;
+          dstBuf = null;
+          System.gc();
+        }
+      }
+      System.out.println("Done.      ");
+    } catch (Exception e) {
+      if (tjc != null) tjc.close();
+      throw e;
+    }
+    if (tjc != null) tjc.close();
+  }
+
+  public static void main(String argv) {
+    try {
+      String testName = "javatest";
+      for (int i = 0; i < argv.length; i++) {
+        if (argvi.equalsIgnoreCase("-yuv"))
+          doYUV = true;
+        else if (argvi.equalsIgnoreCase("-noyuvpad"))
+          pad = 1;
+        else if (argvi.equalsIgnoreCase("-bi")) {
+          bi = true;
+          testName = "javabitest";
+        } else
+          usage();
+      }
+      if (doYUV)
+        FORMATS_4BYTE4 = -1;
+      doTest(35, 39, bi ? FORMATS_3BYTEBI : FORMATS_3BYTE, TJ.SAMP_444,
+             testName);
+      doTest(39, 41, bi ? FORMATS_4BYTEBI : FORMATS_4BYTE, TJ.SAMP_444,
+             testName);
+      doTest(41, 35, bi ? FORMATS_3BYTEBI : FORMATS_3BYTE, TJ.SAMP_422,
+             testName);
+      doTest(35, 39, bi ? FORMATS_4BYTEBI : FORMATS_4BYTE, TJ.SAMP_422,
+             testName);
+      doTest(39, 41, bi ? FORMATS_3BYTEBI : FORMATS_3BYTE, TJ.SAMP_420,
+             testName);
+      doTest(41, 35, bi ? FORMATS_4BYTEBI : FORMATS_4BYTE, TJ.SAMP_420,
+             testName);
+      doTest(35, 39, bi ? FORMATS_3BYTEBI : FORMATS_3BYTE, TJ.SAMP_440,
+             testName);
+      doTest(39, 41, bi ? FORMATS_4BYTEBI : FORMATS_4BYTE, TJ.SAMP_440,
+             testName);
+      doTest(41, 35, bi ? FORMATS_3BYTEBI : FORMATS_3BYTE, TJ.SAMP_411,
+             testName);
+      doTest(35, 39, bi ? FORMATS_4BYTEBI : FORMATS_4BYTE, TJ.SAMP_411,
+             testName);
+      doTest(39, 41, bi ? FORMATS_GRAYBI : FORMATS_GRAY, TJ.SAMP_GRAY,
+             testName);
+      doTest(41, 35, bi ? FORMATS_3BYTEBI : FORMATS_3BYTE, TJ.SAMP_GRAY,
+             testName);
+      FORMATS_4BYTE4 = -1;
+      doTest(35, 39, bi ? FORMATS_4BYTEBI : FORMATS_4BYTE, TJ.SAMP_GRAY,
+             testName);
+      if (!bi)
+        bufSizeTest();
+      if (doYUV && !bi) {
+        System.out.print("\n--------------------\n\n");
+        doTest(48, 48, FORMATS_RGB, TJ.SAMP_444, "javatest_yuv0");
+        doTest(48, 48, FORMATS_RGB, TJ.SAMP_422, "javatest_yuv0");
+        doTest(48, 48, FORMATS_RGB, TJ.SAMP_420, "javatest_yuv0");
+        doTest(48, 48, FORMATS_RGB, TJ.SAMP_440, "javatest_yuv0");
+        doTest(48, 48, FORMATS_RGB, TJ.SAMP_411, "javatest_yuv0");
+        doTest(48, 48, FORMATS_RGB, TJ.SAMP_GRAY, "javatest_yuv0");
+        doTest(48, 48, FORMATS_GRAY, TJ.SAMP_GRAY, "javatest_yuv0");
+      }
+    } catch (Exception e) {
+      e.printStackTrace();
+      exitStatus = -1;
+    }
+    System.exit(exitStatus);
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/allclasses-frame.html Added

@@ -0,0 +1,24 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>All Classes</title>
+<link rel="stylesheet" type="text/css" href="stylesheet.css" title="Style">
+</head>
+<body>
+<h1 class="bar">All Classes</h1>
+<div class="indexContainer">
+<ul>
+<li><a href="org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJ</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJCompressor</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg" target="classFrame"><i>TJCustomFilter</i></a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJDecompressor</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJException</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJScalingFactor</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJTransform</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJTransformer</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">YUVImage</a></li>
+</ul>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/allclasses-noframe.html Added

@@ -0,0 +1,24 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>All Classes</title>
+<link rel="stylesheet" type="text/css" href="stylesheet.css" title="Style">
+</head>
+<body>
+<h1 class="bar">All Classes</h1>
+<div class="indexContainer">
+<ul>
+<li><a href="org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><i>TJCustomFilter</i></a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg">TJTransformer</a></li>
+<li><a href="org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></li>
+</ul>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/constant-values.html Added

@@ -0,0 +1,539 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>Constant Field Values</title>
+<link rel="stylesheet" type="text/css" href="stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li><a href="deprecated-list.html">Deprecated</a></li>
+<li><a href="index-all.html">Index</a></li>
+<li><a href="help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?constant-values.html" target="_top">Frames</a></li>
+<li><a href="constant-values.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+<div class="header">
+<h1 title="Constant Field Values" class="title">Constant Field Values</h1>
+<h2 title="Contents">Contents</h2>
+<ul>
+<li><a href="#org.libjpegturbo">org.libjpegturbo.*</a></li>
+</ul>
+</div>
+<div class="constantValuesContainer"><a name="org.libjpegturbo">
+
+</a>
+<h2 title="org.libjpegturbo">org.libjpegturbo.*</h2>
+<ul class="blockList">
+<li class="blockList">
+<table border="0" cellpadding="3" cellspacing="0" summary="Constant Field Values table, listing constant fields, and values">
+<caption><span>org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th scope="col">Constant Field</th>
+<th class="colLast" scope="col">Value</th>
+</tr>
+<tbody>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.CS_CMYK">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#CS_CMYK">CS_CMYK</a></code></td>
+<td class="colLast"><code>3</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.CS_GRAY">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#CS_GRAY">CS_GRAY</a></code></td>
+<td class="colLast"><code>2</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.CS_RGB">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#CS_RGB">CS_RGB</a></code></td>
+<td class="colLast"><code>0</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.CS_YCbCr">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#CS_YCbCr">CS_YCbCr</a></code></td>
+<td class="colLast"><code>1</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.CS_YCCK">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#CS_YCCK">CS_YCCK</a></code></td>
+<td class="colLast"><code>4</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.ERR_FATAL">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#ERR_FATAL">ERR_FATAL</a></code></td>
+<td class="colLast"><code>1</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.ERR_WARNING">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#ERR_WARNING">ERR_WARNING</a></code></td>
+<td class="colLast"><code>0</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_ACCURATEDCT">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_ACCURATEDCT">FLAG_ACCURATEDCT</a></code></td>
+<td class="colLast"><code>4096</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_BOTTOMUP">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP">FLAG_BOTTOMUP</a></code></td>
+<td class="colLast"><code>2</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_FASTDCT">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FASTDCT">FLAG_FASTDCT</a></code></td>
+<td class="colLast"><code>2048</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_FASTUPSAMPLE">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FASTUPSAMPLE">FLAG_FASTUPSAMPLE</a></code></td>
+<td class="colLast"><code>256</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_FORCEMMX">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCEMMX">FLAG_FORCEMMX</a></code></td>
+<td class="colLast"><code>8</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_FORCESSE">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE">FLAG_FORCESSE</a></code></td>
+<td class="colLast"><code>16</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_FORCESSE2">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE2">FLAG_FORCESSE2</a></code></td>
+<td class="colLast"><code>32</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_FORCESSE3">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE3">FLAG_FORCESSE3</a></code></td>
+<td class="colLast"><code>128</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_LIMITSCANS">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_LIMITSCANS">FLAG_LIMITSCANS</a></code></td>
+<td class="colLast"><code>32768</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_PROGRESSIVE">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_PROGRESSIVE">FLAG_PROGRESSIVE</a></code></td>
+<td class="colLast"><code>16384</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.FLAG_STOPONWARNING">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_STOPONWARNING">FLAG_STOPONWARNING</a></code></td>
+<td class="colLast"><code>8192</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.NUMCS">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#NUMCS">NUMCS</a></code></td>
+<td class="colLast"><code>5</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.NUMERR">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#NUMERR">NUMERR</a></code></td>
+<td class="colLast"><code>2</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.NUMPF">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#NUMPF">NUMPF</a></code></td>
+<td class="colLast"><code>12</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.NUMSAMP">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#NUMSAMP">NUMSAMP</a></code></td>
+<td class="colLast"><code>6</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_ABGR">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_ABGR">PF_ABGR</a></code></td>
+<td class="colLast"><code>9</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_ARGB">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_ARGB">PF_ARGB</a></code></td>
+<td class="colLast"><code>10</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_BGR">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_BGR">PF_BGR</a></code></td>
+<td class="colLast"><code>1</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_BGRA">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_BGRA">PF_BGRA</a></code></td>
+<td class="colLast"><code>8</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_BGRX">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_BGRX">PF_BGRX</a></code></td>
+<td class="colLast"><code>3</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_CMYK">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_CMYK">PF_CMYK</a></code></td>
+<td class="colLast"><code>11</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_GRAY">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_GRAY">PF_GRAY</a></code></td>
+<td class="colLast"><code>6</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_RGB">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_RGB">PF_RGB</a></code></td>
+<td class="colLast"><code>0</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_RGBA">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_RGBA">PF_RGBA</a></code></td>
+<td class="colLast"><code>7</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_RGBX">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_RGBX">PF_RGBX</a></code></td>
+<td class="colLast"><code>2</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_XBGR">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_XBGR">PF_XBGR</a></code></td>
+<td class="colLast"><code>4</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.PF_XRGB">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#PF_XRGB">PF_XRGB</a></code></td>
+<td class="colLast"><code>5</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.SAMP_411">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#SAMP_411">SAMP_411</a></code></td>
+<td class="colLast"><code>5</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.SAMP_420">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#SAMP_420">SAMP_420</a></code></td>
+<td class="colLast"><code>2</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.SAMP_422">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#SAMP_422">SAMP_422</a></code></td>
+<td class="colLast"><code>1</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.SAMP_440">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#SAMP_440">SAMP_440</a></code></td>
+<td class="colLast"><code>4</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.SAMP_444">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#SAMP_444">SAMP_444</a></code></td>
+<td class="colLast"><code>0</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJ.SAMP_GRAY">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJ.html#SAMP_GRAY">SAMP_GRAY</a></code></td>
+<td class="colLast"><code>3</code></td>
+</tr>
+</tbody>
+</table>
+</li>
+<li class="blockList">
+<table border="0" cellpadding="3" cellspacing="0" summary="Constant Field Values table, listing constant fields, and values">
+<caption><span>org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th scope="col">Constant Field</th>
+<th class="colLast" scope="col">Value</th>
+</tr>
+<tbody>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.NUMOP">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#NUMOP">NUMOP</a></code></td>
+<td class="colLast"><code>8</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OP_HFLIP">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OP_HFLIP">OP_HFLIP</a></code></td>
+<td class="colLast"><code>1</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OP_NONE">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OP_NONE">OP_NONE</a></code></td>
+<td class="colLast"><code>0</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OP_ROT180">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OP_ROT180">OP_ROT180</a></code></td>
+<td class="colLast"><code>6</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OP_ROT270">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OP_ROT270">OP_ROT270</a></code></td>
+<td class="colLast"><code>7</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OP_ROT90">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OP_ROT90">OP_ROT90</a></code></td>
+<td class="colLast"><code>5</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OP_TRANSPOSE">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OP_TRANSPOSE">OP_TRANSPOSE</a></code></td>
+<td class="colLast"><code>3</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OP_TRANSVERSE">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OP_TRANSVERSE">OP_TRANSVERSE</a></code></td>
+<td class="colLast"><code>4</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OP_VFLIP">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OP_VFLIP">OP_VFLIP</a></code></td>
+<td class="colLast"><code>2</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OPT_COPYNONE">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OPT_COPYNONE">OPT_COPYNONE</a></code></td>
+<td class="colLast"><code>64</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OPT_CROP">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OPT_CROP">OPT_CROP</a></code></td>
+<td class="colLast"><code>4</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OPT_GRAY">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OPT_GRAY">OPT_GRAY</a></code></td>
+<td class="colLast"><code>8</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OPT_NOOUTPUT">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OPT_NOOUTPUT">OPT_NOOUTPUT</a></code></td>
+<td class="colLast"><code>16</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OPT_PERFECT">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT">OPT_PERFECT</a></code></td>
+<td class="colLast"><code>1</code></td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OPT_PROGRESSIVE">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PROGRESSIVE">OPT_PROGRESSIVE</a></code></td>
+<td class="colLast"><code>32</code></td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a name="org.libjpegturbo.turbojpeg.TJTransform.OPT_TRIM">
+
+</a><code>public&nbsp;static&nbsp;final&nbsp;int</code></td>
+<td><code><a href="org/libjpegturbo/turbojpeg/TJTransform.html#OPT_TRIM">OPT_TRIM</a></code></td>
+<td class="colLast"><code>2</code></td>
+</tr>
+</tbody>
+</table>
+</li>
+</ul>
+</div>
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li><a href="deprecated-list.html">Deprecated</a></li>
+<li><a href="index-all.html">Index</a></li>
+<li><a href="help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?constant-values.html" target="_top">Frames</a></li>
+<li><a href="constant-values.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/deprecated-list.html Added

@@ -0,0 +1,252 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>Deprecated List</title>
+<link rel="stylesheet" type="text/css" href="stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li class="navBarCell1Rev">Deprecated</li>
+<li><a href="index-all.html">Index</a></li>
+<li><a href="help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?deprecated-list.html" target="_top">Frames</a></li>
+<li><a href="deprecated-list.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+<div class="header">
+<h1 title="Deprecated API" class="title">Deprecated API</h1>
+<h2 title="Contents">Contents</h2>
+<ul>
+<li><a href="#field">Deprecated Fields</a></li>
+<li><a href="#method">Deprecated Methods</a></li>
+<li><a href="#constructor">Deprecated Constructors</a></li>
+</ul>
+</div>
+<div class="contentContainer"><a name="field">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<table border="0" cellpadding="3" cellspacing="0" summary="Deprecated Fields table, listing deprecated fields, and an explanation">
+<caption><span>Deprecated Fields</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Field and Description</th>
+</tr>
+<tbody>
+<tr class="altColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCEMMX">org.libjpegturbo.turbojpeg.TJ.FLAG_FORCEMMX</a></td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE">org.libjpegturbo.turbojpeg.TJ.FLAG_FORCESSE</a></td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE2">org.libjpegturbo.turbojpeg.TJ.FLAG_FORCESSE2</a></td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE3">org.libjpegturbo.turbojpeg.TJ.FLAG_FORCESSE3</a></td>
+</tr>
+</tbody>
+</table>
+</li>
+</ul>
+<a name="method">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<table border="0" cellpadding="3" cellspacing="0" summary="Deprecated Methods table, listing deprecated methods, and an explanation">
+<caption><span>Deprecated Methods</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Method and Description</th>
+</tr>
+<tbody>
+<tr class="altColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int)">org.libjpegturbo.turbojpeg.TJ.bufSizeYUV(int, int, int)</a>
+<div class="block"><i>Use <a href="org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int,%20int)"><code>TJ.bufSizeYUV(int, int, int, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJCompressor.html#compress(java.awt.image.BufferedImage,%20byte,%20int)">org.libjpegturbo.turbojpeg.TJCompressor.compress(BufferedImage, byte, int)</a>
+<div class="block"><i>Use
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#compress(byte,%20int)"><code>TJCompressor.compress(byte, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJCompressor.html#compress(java.awt.image.BufferedImage,%20int)">org.libjpegturbo.turbojpeg.TJCompressor.compress(BufferedImage, int)</a>
+<div class="block"><i>Use
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#compress(int)"><code>TJCompressor.compress(int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int)">org.libjpegturbo.turbojpeg.TJDecompressor.decompress(byte, int, int, int, int, int)</a>
+<div class="block"><i>Use
+ <a href="org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)"><code>TJDecompressor.decompress(byte, int, int, int, int, int, int, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(byte,%20int)">org.libjpegturbo.turbojpeg.TJDecompressor.decompressToYUV(byte, int)</a>
+<div class="block"><i>Use <a href="org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)"><code>TJDecompressor.decompressToYUV(YUVImage, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int)">org.libjpegturbo.turbojpeg.TJDecompressor.decompressToYUV(int)</a>
+<div class="block"><i>Use <a href="org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)"><code>TJDecompressor.decompressToYUV(int, int, int, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(java.awt.image.BufferedImage,%20byte,%20int)">org.libjpegturbo.turbojpeg.TJCompressor.encodeYUV(BufferedImage, byte, int)</a>
+<div class="block"><i>Use
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(byte,%20int)"><code>TJCompressor.encodeYUV(byte, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(java.awt.image.BufferedImage,%20int)">org.libjpegturbo.turbojpeg.TJCompressor.encodeYUV(BufferedImage, int)</a>
+<div class="block"><i>Use
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)"><code>TJCompressor.encodeYUV(int, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(byte,%20int)">org.libjpegturbo.turbojpeg.TJCompressor.encodeYUV(byte, int)</a>
+<div class="block"><i>Use <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)"><code>TJCompressor.encodeYUV(YUVImage, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int)">org.libjpegturbo.turbojpeg.TJCompressor.encodeYUV(int)</a>
+<div class="block"><i>Use <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)"><code>TJCompressor.encodeYUV(int, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJDecompressor.html#setJPEGImage(byte,%20int)">org.libjpegturbo.turbojpeg.TJDecompressor.setJPEGImage(byte, int)</a>
+<div class="block"><i>Use <a href="org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(byte,%20int)"><code>TJDecompressor.setSourceImage(byte, int)</code></a> instead.</i></div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int)">org.libjpegturbo.turbojpeg.TJCompressor.setSourceImage(byte, int, int, int, int)</a>
+<div class="block"><i>Use
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(byte, int, int, int, int, int, int)</code></a> instead.</i></div>
+</td>
+</tr>
+</tbody>
+</table>
+</li>
+</ul>
+<a name="constructor">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<table border="0" cellpadding="3" cellspacing="0" summary="Deprecated Constructors table, listing deprecated constructors, and an explanation">
+<caption><span>Deprecated Constructors</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Constructor and Description</th>
+</tr>
+<tbody>
+<tr class="altColor">
+<td class="colOne"><a href="org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(byte,%20int,%20int,%20int,%20int)">org.libjpegturbo.turbojpeg.TJCompressor(byte, int, int, int, int)</a>
+<div class="block"><i>Use
+ <a href="org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>TJCompressor.TJCompressor(byte, int, int, int, int, int, int)</code></a> instead.</i></div>
+</td>
+</tr>
+</tbody>
+</table>
+</li>
+</ul>
+</div>
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li class="navBarCell1Rev">Deprecated</li>
+<li><a href="index-all.html">Index</a></li>
+<li><a href="help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?deprecated-list.html" target="_top">Frames</a></li>
+<li><a href="deprecated-list.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/help-doc.html Added

@@ -0,0 +1,210 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>API Help</title>
+<link rel="stylesheet" type="text/css" href="stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li><a href="deprecated-list.html">Deprecated</a></li>
+<li><a href="index-all.html">Index</a></li>
+<li class="navBarCell1Rev">Help</li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?help-doc.html" target="_top">Frames</a></li>
+<li><a href="help-doc.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+<div class="header">
+<h1 class="title">How This API Document Is Organized</h1>
+<div class="subTitle">This API (Application Programming Interface) document has pages corresponding to the items in the navigation bar, described as follows.</div>
+</div>
+<div class="contentContainer">
+<ul class="blockList">
+<li class="blockList">
+<h2>Package</h2>
+<p>Each package has a page that contains a list of its classes and interfaces, with a summary for each. This page can contain six categories:</p>
+<ul>
+<li>Interfaces (italic)</li>
+<li>Classes</li>
+<li>Enums</li>
+<li>Exceptions</li>
+<li>Errors</li>
+<li>Annotation Types</li>
+</ul>
+</li>
+<li class="blockList">
+<h2>Class/Interface</h2>
+<p>Each class, interface, nested class and nested interface has its own separate page. Each of these pages has three sections consisting of a class/interface description, summary tables, and detailed member descriptions:</p>
+<ul>
+<li>Class inheritance diagram</li>
+<li>Direct Subclasses</li>
+<li>All Known Subinterfaces</li>
+<li>All Known Implementing Classes</li>
+<li>Class/interface declaration</li>
+<li>Class/interface description</li>
+</ul>
+<ul>
+<li>Nested Class Summary</li>
+<li>Field Summary</li>
+<li>Constructor Summary</li>
+<li>Method Summary</li>
+</ul>
+<ul>
+<li>Field Detail</li>
+<li>Constructor Detail</li>
+<li>Method Detail</li>
+</ul>
+<p>Each summary entry contains the first sentence from the detailed description for that item. The summary entries are alphabetical, while the detailed descriptions are in the order they appear in the source code. This preserves the logical groupings established by the programmer.</p>
+</li>
+<li class="blockList">
+<h2>Annotation Type</h2>
+<p>Each annotation type has its own separate page with the following sections:</p>
+<ul>
+<li>Annotation Type declaration</li>
+<li>Annotation Type description</li>
+<li>Required Element Summary</li>
+<li>Optional Element Summary</li>
+<li>Element Detail</li>
+</ul>
+</li>
+<li class="blockList">
+<h2>Enum</h2>
+<p>Each enum has its own separate page with the following sections:</p>
+<ul>
+<li>Enum declaration</li>
+<li>Enum description</li>
+<li>Enum Constant Summary</li>
+<li>Enum Constant Detail</li>
+</ul>
+</li>
+<li class="blockList">
+<h2>Tree (Class Hierarchy)</h2>
+<p>There is a <a href="overview-tree.html">Class Hierarchy</a> page for all packages, plus a hierarchy for each package. Each hierarchy page contains a list of classes and a list of interfaces. The classes are organized by inheritance structure starting with <code>java.lang.Object</code>. The interfaces do not inherit from <code>java.lang.Object</code>.</p>
+<ul>
+<li>When viewing the Overview page, clicking on "Tree" displays the hierarchy for all packages.</li>
+<li>When viewing a particular package, class or interface page, clicking "Tree" displays the hierarchy for only that package.</li>
+</ul>
+</li>
+<li class="blockList">
+<h2>Deprecated API</h2>
+<p>The <a href="deprecated-list.html">Deprecated API</a> page lists all of the API that have been deprecated. A deprecated API is not recommended for use, generally due to improvements, and a replacement API is usually given. Deprecated APIs may be removed in future implementations.</p>
+</li>
+<li class="blockList">
+<h2>Index</h2>
+<p>The <a href="index-all.html">Index</a> contains an alphabetic list of all classes, interfaces, constructors, methods, and fields.</p>
+</li>
+<li class="blockList">
+<h2>Prev/Next</h2>
+<p>These links take you to the next or previous class, interface, package, or related page.</p>
+</li>
+<li class="blockList">
+<h2>Frames/No Frames</h2>
+<p>These links show and hide the HTML frames.  All pages are available with or without frames.</p>
+</li>
+<li class="blockList">
+<h2>All Classes</h2>
+<p>The <a href="allclasses-noframe.html">All Classes</a> link shows all classes and interfaces except non-static nested types.</p>
+</li>
+<li class="blockList">
+<h2>Serialized Form</h2>
+<p>Each serializable or externalizable class has a description of its serialization fields and methods. This information is of interest to re-implementors, not to developers using the API. While there is no link in the navigation bar, you can get to this information by going to any serialized class and clicking "Serialized Form" in the "See also" section of the class description.</p>
+</li>
+<li class="blockList">
+<h2>Constant Field Values</h2>
+<p>The <a href="constant-values.html">Constant Field Values</a> page lists the static final fields and their values.</p>
+</li>
+</ul>
+<em>This help file applies to API documentation generated using the standard doclet.</em></div>
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li><a href="deprecated-list.html">Deprecated</a></li>
+<li><a href="index-all.html">Index</a></li>
+<li class="navBarCell1Rev">Help</li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?help-doc.html" target="_top">Frames</a></li>
+<li><a href="help-doc.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/index-all.html Added

@@ -0,0 +1,1034 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>Index</title>
+<link rel="stylesheet" type="text/css" href="./stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="./org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="./org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li><a href="./deprecated-list.html">Deprecated</a></li>
+<li class="navBarCell1Rev">Index</li>
+<li><a href="./help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="./index.html?index-all.html" target="_top">Frames</a></li>
+<li><a href="index-all.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="./allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+<div class="contentContainer"><a href="#_B_">B</a>&nbsp;<a href="#_C_">C</a>&nbsp;<a href="#_D_">D</a>&nbsp;<a href="#_E_">E</a>&nbsp;<a href="#_F_">F</a>&nbsp;<a href="#_G_">G</a>&nbsp;<a href="#_H_">H</a>&nbsp;<a href="#_I_">I</a>&nbsp;<a href="#_J_">J</a>&nbsp;<a href="#_N_">N</a>&nbsp;<a href="#_O_">O</a>&nbsp;<a href="#_P_">P</a>&nbsp;<a href="#_S_">S</a>&nbsp;<a href="#_T_">T</a>&nbsp;<a href="#_Y_">Y</a>&nbsp;<a name="_B_">
+
+</a>
+<h2 class="title">B</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#bufSize(int,%20int,%20int)">bufSize(int, int, int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Returns the maximum size of the buffer (in bytes) required to hold a JPEG
+ image with the given width, height, and level of chrominance subsampling.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int,%20int)">bufSizeYUV(int, int, int, int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Returns the size of the buffer (in bytes) required to hold a YUV planar
+ image with the given width, height, and level of chrominance subsampling.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int)">bufSizeYUV(int, int, int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use <a href="./org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int,%20int)"><code>TJ.bufSizeYUV(int, int, int, int)</code></a> instead.</i></div>
+</div>
+</dd>
+</dl>
+<a name="_C_">
+
+</a>
+<h2 class="title">C</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#cf">cf</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Custom filter instance</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#close()">close()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Free the native structures associated with this compressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#close()">close()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Free the native structures associated with this decompressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#compress(byte,%20int)">compress(byte, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Compress the uncompressed source image associated with this compressor
+ instance and output a JPEG image to the given destination buffer.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#compress(int)">compress(int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Compress the uncompressed source image associated with this compressor
+ instance and return a buffer containing a JPEG image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#compress(java.awt.image.BufferedImage,%20byte,%20int)">compress(BufferedImage, byte, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#compress(byte,%20int)"><code>TJCompressor.compress(byte, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#compress(java.awt.image.BufferedImage,%20int)">compress(BufferedImage, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#compress(int)"><code>TJCompressor.compress(int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#CS_CMYK">CS_CMYK</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">CMYK colorspace.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#CS_GRAY">CS_GRAY</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Grayscale colorspace.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#CS_RGB">CS_RGB</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">RGB colorspace.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#CS_YCbCr">CS_YCbCr</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">YCbCr colorspace.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#CS_YCCK">CS_YCCK</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">YCCK colorspace.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCustomFilter.html#customFilter(java.nio.ShortBuffer,%20java.awt.Rectangle,%20java.awt.Rectangle,%20int,%20int,%20org.libjpegturbo.turbojpeg.TJTransform)">customFilter(ShortBuffer, Rectangle, Rectangle, int, int, TJTransform)</a></span> - Method in interface org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg">TJCustomFilter</a></dt>
+<dd>
+<div class="block">A callback function that can be used to modify the DCT coefficients after
+ they are losslessly transformed but before they are transcoded to a new
+ JPEG image.</div>
+</dd>
+</dl>
+<a name="_D_">
+
+</a>
+<h2 class="title">D</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)">decompress(byte, int, int, int, int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and output a grayscale, RGB, or CMYK image
+ to the given destination buffer.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int)">decompress(byte, int, int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use
+ <a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)"><code>TJDecompressor.decompress(byte, int, int, int, int, int, int, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(int,%20int,%20int,%20int,%20int)">decompress(int, int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance and return a buffer containing the decompressed image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(int,%20int,%20int,%20int,%20int,%20int,%20int,%20int)">decompress(int, int, int, int, int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and output a grayscale, RGB, or CMYK image
+ to the given destination buffer.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(java.awt.image.BufferedImage,%20int)">decompress(BufferedImage, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and output a decompressed/decoded image to
+ the given <code>BufferedImage</code> instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(int,%20int,%20int,%20int)">decompress(int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and return a <code>BufferedImage</code>
+ instance containing the decompressed/decoded image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)">decompressToYUV(YUVImage, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance into a YUV planar image and store it in the given
+ <code>YUVImage</code> instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(byte,%20int)">decompressToYUV(byte, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use <a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)"><code>TJDecompressor.decompressToYUV(YUVImage, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)">decompressToYUV(int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance into a set of Y, U (Cb), and V (Cr) image planes and return a
+ <code>YUVImage</code> instance containing the decompressed image planes.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)">decompressToYUV(int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance into a unified YUV planar image buffer and return a
+ <code>YUVImage</code> instance containing the decompressed image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int)">decompressToYUV(int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use <a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)"><code>TJDecompressor.decompressToYUV(int, int, int, int)</code></a> instead.</i></div>
+</div>
+</dd>
+</dl>
+<a name="_E_">
+
+</a>
+<h2 class="title">E</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)">encodeYUV(YUVImage, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Encode the uncompressed source image associated with this compressor
+ instance into a YUV planar image and store it in the given
+ <code>YUVImage</code> instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(byte,%20int)">encodeYUV(byte, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)"><code>TJCompressor.encodeYUV(YUVImage, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)">encodeYUV(int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Encode the uncompressed source image associated with this compressor
+ instance into a unified YUV planar image buffer and return a
+ <code>YUVImage</code> instance containing the encoded image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)">encodeYUV(int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Encode the uncompressed source image associated with this compressor
+ instance into separate Y, U (Cb), and V (Cr) image planes and return a
+ <code>YUVImage</code> instance containing the encoded image planes.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int)">encodeYUV(int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)"><code>TJCompressor.encodeYUV(int, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(java.awt.image.BufferedImage,%20byte,%20int)">encodeYUV(BufferedImage, byte, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(byte,%20int)"><code>TJCompressor.encodeYUV(byte, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(java.awt.image.BufferedImage,%20int)">encodeYUV(BufferedImage, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)"><code>TJCompressor.encodeYUV(int, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html#equals(org.libjpegturbo.turbojpeg.TJScalingFactor)">equals(TJScalingFactor)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a></dt>
+<dd>
+<div class="block">Returns true or false, depending on whether this instance and
+ <code>other</code> have the same numerator and denominator.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#ERR_FATAL">ERR_FATAL</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">The error was fatal and non-recoverable.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#ERR_WARNING">ERR_WARNING</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">The error was non-fatal and recoverable, but the image may still be
+ corrupt.</div>
+</dd>
+</dl>
+<a name="_F_">
+
+</a>
+<h2 class="title">F</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#finalize()">finalize()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#finalize()">finalize()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_ACCURATEDCT">FLAG_ACCURATEDCT</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Use the most accurate DCT/IDCT algorithm available in the underlying
+ codec.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP">FLAG_BOTTOMUP</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">The uncompressed source/destination image is stored in bottom-up (Windows,
+ OpenGL) order, not top-down (X11) order.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_FASTDCT">FLAG_FASTDCT</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Use the fastest DCT/IDCT algorithm available in the underlying codec.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_FASTUPSAMPLE">FLAG_FASTUPSAMPLE</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">When decompressing an image that was compressed using chrominance
+ subsampling, use the fastest chrominance upsampling algorithm available in
+ the underlying codec.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCEMMX">FLAG_FORCEMMX</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span></div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE">FLAG_FORCESSE</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span></div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE2">FLAG_FORCESSE2</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span></div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE3">FLAG_FORCESSE3</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span></div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_LIMITSCANS">FLAG_LIMITSCANS</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Limit the number of progressive JPEG scans that the decompression and
+ transform operations will process.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_PROGRESSIVE">FLAG_PROGRESSIVE</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Use progressive entropy coding in JPEG images generated by compression and
+ transform operations.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#FLAG_STOPONWARNING">FLAG_STOPONWARNING</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Immediately discontinue the current compression/decompression/transform
+ operation if the underlying codec throws a warning (non-fatal error).</div>
+</dd>
+</dl>
+<a name="_G_">
+
+</a>
+<h2 class="title">G</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#getAlphaOffset(int)">getAlphaOffset(int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">For the given pixel format, returns the number of bytes that the alpha
+ component is offset from the start of the pixel.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#getBlueOffset(int)">getBlueOffset(int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">For the given pixel format, returns the number of bytes that the blue
+ component is offset from the start of the pixel.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#getBuf()">getBuf()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Returns the YUV image buffer (if this image is stored in a unified
+ buffer rather than separate image planes.)</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#getColorspace()">getColorspace()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Returns the colorspace used in the source image (JPEG or YUV) associated
+ with this decompressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#getCompressedSize()">getCompressedSize()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Returns the size of the image (in bytes) generated by the most recent
+ compress operation.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html#getDenom()">getDenom()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a></dt>
+<dd>
+<div class="block">Returns denominator</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJException.html#getErrorCode()">getErrorCode()</a></span> - Method in exception org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></dt>
+<dd>
+<div class="block">Returns a code (one of <a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><code>TJ.ERR_*</code></a>) indicating the severity of the
+ last error.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#getGreenOffset(int)">getGreenOffset(int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">For the given pixel format, returns the number of bytes that the green
+ component is offset from the start of the pixel.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#getHeight()">getHeight()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Returns the height of the source image (JPEG or YUV) associated with this
+ decompressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#getHeight()">getHeight()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Returns the height of the YUV image (or subregion.)</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#getJPEGBuf()">getJPEGBuf()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Returns the JPEG image buffer associated with this decompressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#getJPEGSize()">getJPEGSize()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Returns the size of the JPEG image (in bytes) associated with this
+ decompressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#getMCUHeight(int)">getMCUHeight(int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Returns the MCU block height for the given level of chrominance
+ subsampling.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#getMCUWidth(int)">getMCUWidth(int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Returns the MCU block width for the given level of chrominance
+ subsampling.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html#getNum()">getNum()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a></dt>
+<dd>
+<div class="block">Returns numerator</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#getOffsets()">getOffsets()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Returns the offsets (in bytes) of each plane within the planes of a larger
+ YUV image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#getPad()">getPad()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Returns the line padding used in the YUV image buffer (if this image is
+ stored in a unified buffer rather than separate image planes.)</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#getPixelSize(int)">getPixelSize(int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Returns the pixel size (in bytes) for the given pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#getPlanes()">getPlanes()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Returns the YUV image planes.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#getRedOffset(int)">getRedOffset(int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">For the given pixel format, returns the number of bytes that the red
+ component is offset from the start of the pixel.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html#getScaled(int)">getScaled(int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a></dt>
+<dd>
+<div class="block">Returns the scaled value of <code>dimension</code>.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledHeight(int,%20int)">getScaledHeight(int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Returns the height of the largest scaled-down image that the TurboJPEG
+ decompressor can generate without exceeding the desired image width and
+ height.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledWidth(int,%20int)">getScaledWidth(int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Returns the width of the largest scaled-down image that the TurboJPEG
+ decompressor can generate without exceeding the desired image width and
+ height.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#getScalingFactors()">getScalingFactors()</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Returns a list of fractional scaling factors that the JPEG decompressor in
+ this implementation of TurboJPEG supports.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#getSize()">getSize()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Returns the size (in bytes) of the YUV image buffer (if this image is
+ stored in a unified buffer rather than separate image planes.)</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#getStrides()">getStrides()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Returns the number of bytes per line of each plane in the YUV image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#getSubsamp()">getSubsamp()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Returns the level of chrominance subsampling used in the source image
+ (JPEG or YUV) associated with this decompressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#getSubsamp()">getSubsamp()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Returns the level of chrominance subsampling used in the YUV image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransformer.html#getTransformedSizes()">getTransformedSizes()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg">TJTransformer</a></dt>
+<dd>
+<div class="block">Returns an array containing the sizes of the transformed JPEG images
+ generated by the most recent transform operation.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#getWidth()">getWidth()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Returns the width of the source image (JPEG or YUV) associated with this
+ decompressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#getWidth()">getWidth()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Returns the width of the YUV image (or subregion.)</div>
+</dd>
+</dl>
+<a name="_H_">
+
+</a>
+<h2 class="title">H</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#handle">handle</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#handle">handle</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>&nbsp;</dd>
+</dl>
+<a name="_I_">
+
+</a>
+<h2 class="title">I</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html#isOne()">isOne()</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a></dt>
+<dd>
+<div class="block">Returns true or false, depending on whether this instance is equal to
+ 1/1.</div>
+</dd>
+</dl>
+<a name="_J_">
+
+</a>
+<h2 class="title">J</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegBuf">jpegBuf</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegBufSize">jpegBufSize</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegColorspace">jpegColorspace</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegHeight">jpegHeight</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegSubsamp">jpegSubsamp</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegWidth">jpegWidth</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>&nbsp;</dd>
+</dl>
+<a name="_N_">
+
+</a>
+<h2 class="title">N</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#NUMCS">NUMCS</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">The number of JPEG colorspaces</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#NUMERR">NUMERR</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">The number of error codes</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#NUMOP">NUMOP</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">The number of lossless transform operations</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#NUMPF">NUMPF</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">The number of pixel formats</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#NUMSAMP">NUMSAMP</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">The number of chrominance subsampling options</div>
+</dd>
+</dl>
+<a name="_O_">
+
+</a>
+<h2 class="title">O</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#op">op</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Transform operation (one of <code>OP_*</code>)</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OP_HFLIP">OP_HFLIP</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Flip (mirror) image horizontally.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OP_NONE">OP_NONE</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Do not transform the position of the image pixels.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OP_ROT180">OP_ROT180</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Rotate image 180 degrees.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OP_ROT270">OP_ROT270</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Rotate image counter-clockwise by 90 degrees.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OP_ROT90">OP_ROT90</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Rotate image clockwise by 90 degrees.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OP_TRANSPOSE">OP_TRANSPOSE</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Transpose image (flip/mirror along upper left to lower right axis).</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OP_TRANSVERSE">OP_TRANSVERSE</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Transverse transpose image (flip/mirror along upper right to lower left
+ axis).</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OP_VFLIP">OP_VFLIP</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Flip (mirror) image vertically.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OPT_COPYNONE">OPT_COPYNONE</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">This option will prevent <a href="./org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a> from copying any extra markers (including EXIF
+ and ICC profile data) from the source image to the output image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OPT_CROP">OPT_CROP</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">This option will enable lossless cropping.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OPT_GRAY">OPT_GRAY</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">This option will discard the color data in the input image and produce
+ a grayscale output image.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OPT_NOOUTPUT">OPT_NOOUTPUT</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">This option will prevent <a href="./org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a> from outputting a JPEG image for this
+ particular transform.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT">OPT_PERFECT</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">This option will cause <a href="./org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a> to throw an exception if the transform is not
+ perfect.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PROGRESSIVE">OPT_PROGRESSIVE</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">This option will enable progressive entropy coding in the output image
+ generated by this particular transform.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#OPT_TRIM">OPT_TRIM</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">This option will discard any partial MCU blocks that cannot be
+ transformed.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#options">options</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Transform options (bitwise OR of one or more of <code>OPT_*</code>)</div>
+</dd>
+<dt><a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a> - package org.libjpegturbo.turbojpeg</dt>
+<dd>&nbsp;</dd>
+</dl>
+<a name="_P_">
+
+</a>
+<h2 class="title">P</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_ABGR">PF_ABGR</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">ABGR pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_ARGB">PF_ARGB</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">ARGB pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_BGR">PF_BGR</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">BGR pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_BGRA">PF_BGRA</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">BGRA pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_BGRX">PF_BGRX</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">BGRX pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_CMYK">PF_CMYK</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">CMYK pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_GRAY">PF_GRAY</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Grayscale pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_RGB">PF_RGB</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">RGB pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_RGBA">PF_RGBA</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">RGBA pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_RGBX">PF_RGBX</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">RGBX pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_XBGR">PF_XBGR</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">XBGR pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#PF_XRGB">PF_XRGB</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">XRGB pixel format.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#planeHeight(int,%20int,%20int)">planeHeight(int, int, int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Returns the plane height of a YUV image plane with the given parameters.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#planeSizeYUV(int,%20int,%20int,%20int,%20int)">planeSizeYUV(int, int, int, int, int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Returns the size of the buffer (in bytes) required to hold a YUV image
+ plane with the given parameters.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#planeWidth(int,%20int,%20int)">planeWidth(int, int, int)</a></span> - Static method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Returns the plane width of a YUV image plane with the given parameters.</div>
+</dd>
+</dl>
+<a name="_S_">
+
+</a>
+<h2 class="title">S</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#SAMP_411">SAMP_411</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">4:1:1 chrominance subsampling.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#SAMP_420">SAMP_420</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">4:2:0 chrominance subsampling.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#SAMP_422">SAMP_422</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">4:2:2 chrominance subsampling.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#SAMP_440">SAMP_440</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">4:4:0 chrominance subsampling.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#SAMP_444">SAMP_444</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">4:4:4 chrominance subsampling (no chrominance subsampling).</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJ.html#SAMP_GRAY">SAMP_GRAY</a></span> - Static variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></dt>
+<dd>
+<div class="block">Grayscale.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#setBuf(byte,%20int,%20int,%20int,%20int,%20int)">setBuf(byte, int, int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Assign a set of image planes to this <code>YUVImage</code> instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#setBuf(byte,%20int,%20int,%20int,%20int)">setBuf(byte, int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Assign a unified image buffer to this <code>YUVImage</code> instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#setJPEGImage(byte,%20int)">setJPEGImage(byte, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use <a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(byte,%20int)"><code>TJDecompressor.setSourceImage(byte, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setJPEGQuality(int)">setJPEGQuality(int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Set the JPEG image quality level for subsequent compress operations.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)">setSourceImage(byte, int, int, int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Associate an uncompressed RGB, grayscale, or CMYK source image with this
+ compressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int)">setSourceImage(byte, int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>TJCompressor.setSourceImage(byte, int, int, int, int, int, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)">setSourceImage(BufferedImage, int, int, int, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Associate an uncompressed RGB or grayscale source image with this
+ compressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(org.libjpegturbo.turbojpeg.YUVImage)">setSourceImage(YUVImage)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Associate an uncompressed YUV planar source image with this compressor
+ instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(byte,%20int)">setSourceImage(byte, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Associate the JPEG image of length <code>imageSize</code> bytes stored in
+ <code>jpegImage</code> with this decompressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(org.libjpegturbo.turbojpeg.YUVImage)">setSourceImage(YUVImage)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Associate the specified YUV planar source image with this decompressor
+ instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#setSubsamp(int)">setSubsamp(int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Set the level of chrominance subsampling for subsequent compress/encode
+ operations.</div>
+</dd>
+</dl>
+<a name="_T_">
+
+</a>
+<h2 class="title">T</h2>
+<dl>
+<dt><a href="./org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJ</span></a> - Class in <a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a></dt>
+<dd>
+<div class="block">TurboJPEG utility class (cannot be instantiated)</div>
+</dd>
+<dt><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJCompressor</span></a> - Class in <a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a></dt>
+<dd>
+<div class="block">TurboJPEG compressor</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor()">TJCompressor()</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG compressor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(byte,%20int,%20int,%20int,%20int,%20int,%20int)">TJCompressor(byte, int, int, int, int, int, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG compressor instance and associate the uncompressed
+ source image stored in <code>srcImage</code> with the newly created
+ instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(byte,%20int,%20int,%20int,%20int)">TJCompressor(byte, int, int, int, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block"><span class="strong">Deprecated.</span>
+<div class="block"><i>Use
+ <a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>TJCompressor.TJCompressor(byte, int, int, int, int, int, int)</code></a> instead.</i></div>
+</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)">TJCompressor(BufferedImage, int, int, int, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG compressor instance and associate the uncompressed
+ source image stored in <code>srcImage</code> with the newly created
+ instance.</div>
+</dd>
+<dt><a href="./org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><span class="strong">TJCustomFilter</span></a> - Interface in <a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a></dt>
+<dd>
+<div class="block">Custom filter callback interface</div>
+</dd>
+<dt><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJDecompressor</span></a> - Class in <a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a></dt>
+<dd>
+<div class="block">TurboJPEG decompressor</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#TJDecompressor()">TJDecompressor()</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG decompresssor instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#TJDecompressor(byte)">TJDecompressor(byte)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG decompressor instance and associate the JPEG source
+ image stored in <code>jpegImage</code> with the newly created instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#TJDecompressor(byte,%20int)">TJDecompressor(byte, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG decompressor instance and associate the JPEG source
+ image of length <code>imageSize</code> bytes stored in
+ <code>jpegImage</code> with the newly created instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#TJDecompressor(org.libjpegturbo.turbojpeg.YUVImage)">TJDecompressor(YUVImage)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG decompressor instance and associate the YUV planar
+ source image stored in <code>yuvImage</code> with the newly created
+ instance.</div>
+</dd>
+<dt><a href="./org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJException</span></a> - Exception in <a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJException.html#TJException()">TJException()</a></span> - Constructor for exception org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJException.html#TJException(java.lang.String,%20java.lang.Throwable)">TJException(String, Throwable)</a></span> - Constructor for exception org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJException.html#TJException(java.lang.String)">TJException(String)</a></span> - Constructor for exception org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJException.html#TJException(java.lang.String,%20int)">TJException(String, int)</a></span> - Constructor for exception org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJException.html#TJException(java.lang.Throwable)">TJException(Throwable)</a></span> - Constructor for exception org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></dt>
+<dd>&nbsp;</dd>
+<dt><a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJScalingFactor</span></a> - Class in <a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a></dt>
+<dd>
+<div class="block">Fractional scaling factor</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html#TJScalingFactor(int,%20int)">TJScalingFactor(int, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG scaling factor instance.</div>
+</dd>
+<dt><a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJTransform</span></a> - Class in <a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a></dt>
+<dd>
+<div class="block">Lossless transform parameters</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#TJTransform()">TJTransform()</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Create a new lossless transform instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#TJTransform(int,%20int,%20int,%20int,%20int,%20int,%20org.libjpegturbo.turbojpeg.TJCustomFilter)">TJTransform(int, int, int, int, int, int, TJCustomFilter)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Create a new lossless transform instance with the given parameters.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransform.html#TJTransform(java.awt.Rectangle,%20int,%20int,%20org.libjpegturbo.turbojpeg.TJCustomFilter)">TJTransform(Rectangle, int, int, TJCustomFilter)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></dt>
+<dd>
+<div class="block">Create a new lossless transform instance with the given parameters.</div>
+</dd>
+<dt><a href="./org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJTransformer</span></a> - Class in <a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a></dt>
+<dd>
+<div class="block">TurboJPEG lossless transformer</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransformer.html#TJTransformer()">TJTransformer()</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg">TJTransformer</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG lossless transformer instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransformer.html#TJTransformer(byte)">TJTransformer(byte)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg">TJTransformer</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG lossless transformer instance and associate the JPEG
+ image stored in <code>jpegImage</code> with the newly created instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransformer.html#TJTransformer(byte,%20int)">TJTransformer(byte, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg">TJTransformer</a></dt>
+<dd>
+<div class="block">Create a TurboJPEG lossless transformer instance and associate the JPEG
+ image of length <code>imageSize</code> bytes stored in
+ <code>jpegImage</code> with the newly created instance.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)">transform(byte, TJTransform, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg">TJTransformer</a></dt>
+<dd>
+<div class="block">Losslessly transform the JPEG image associated with this transformer
+ instance into one or more JPEG images stored in the given destination
+ buffers.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJTransformer.html#transform(org.libjpegturbo.turbojpeg.TJTransform,%20int)">transform(TJTransform, int)</a></span> - Method in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg">TJTransformer</a></dt>
+<dd>
+<div class="block">Losslessly transform the JPEG image associated with this transformer
+ instance and return an array of <a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><code>TJDecompressor</code></a> instances, each of
+ which has a transformed JPEG image associated with it.</div>
+</dd>
+</dl>
+<a name="_Y_">
+
+</a>
+<h2 class="title">Y</h2>
+<dl>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#yuvHeight">yuvHeight</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html#yuvImage">yuvImage</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></dt>
+<dd>&nbsp;</dd>
+<dt><a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">YUVImage</span></a> - Class in <a href="./org/libjpegturbo/turbojpeg/package-summary.html">org.libjpegturbo.turbojpeg</a></dt>
+<dd>
+<div class="block">This class encapsulates a YUV planar image and the metadata
+ associated with it.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#YUVImage(int,%20int,%20int,%20int)">YUVImage(int, int, int, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Create a new <code>YUVImage</code> instance backed by separate image
+ planes, and allocate memory for the image planes.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#YUVImage(int,%20int,%20int,%20int)">YUVImage(int, int, int, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Create a new <code>YUVImage</code> instance backed by a unified image
+ buffer, and allocate memory for the image buffer.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#YUVImage(byte,%20int,%20int,%20int,%20int,%20int)">YUVImage(byte, int, int, int, int, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Create a new <code>YUVImage</code> instance from a set of existing image
+ planes.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#YUVImage(byte,%20int,%20int,%20int,%20int)">YUVImage(byte, int, int, int, int)</a></span> - Constructor for class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>
+<div class="block">Create a new <code>YUVImage</code> instance from an existing unified image
+ buffer.</div>
+</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#yuvOffsets">yuvOffsets</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#yuvPad">yuvPad</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#yuvPlanes">yuvPlanes</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#yuvStrides">yuvStrides</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#yuvSubsamp">yuvSubsamp</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>&nbsp;</dd>
+<dt><span class="strong"><a href="./org/libjpegturbo/turbojpeg/YUVImage.html#yuvWidth">yuvWidth</a></span> - Variable in class org.libjpegturbo.turbojpeg.<a href="./org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></dt>
+<dd>&nbsp;</dd>
+</dl>
+<a href="#_B_">B</a>&nbsp;<a href="#_C_">C</a>&nbsp;<a href="#_D_">D</a>&nbsp;<a href="#_E_">E</a>&nbsp;<a href="#_F_">F</a>&nbsp;<a href="#_G_">G</a>&nbsp;<a href="#_H_">H</a>&nbsp;<a href="#_I_">I</a>&nbsp;<a href="#_J_">J</a>&nbsp;<a href="#_N_">N</a>&nbsp;<a href="#_O_">O</a>&nbsp;<a href="#_P_">P</a>&nbsp;<a href="#_S_">S</a>&nbsp;<a href="#_T_">T</a>&nbsp;<a href="#_Y_">Y</a>&nbsp;</div>
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="./org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="./org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li><a href="./deprecated-list.html">Deprecated</a></li>
+<li class="navBarCell1Rev">Index</li>
+<li><a href="./help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="./index.html?index-all.html" target="_top">Frames</a></li>
+<li><a href="index-all.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="./allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/index.html Added

@@ -0,0 +1,71 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Frameset//EN" "http://www.w3.org/TR/html4/frameset.dtd">
+
+<html lang="en">
+<head>
+<title>Generated Documentation (Untitled)</title>
+<script type="text/javascript">
+    tmpTargetPage = "" + window.location.search;
+    if (tmpTargetPage != "" && tmpTargetPage != "undefined")
+        tmpTargetPage = tmpTargetPage.substring(1);
+    if (tmpTargetPage.indexOf(":") != -1 || (tmpTargetPage != "" && !validURL(tmpTargetPage)))
+        tmpTargetPage = "undefined";
+    targetPage = tmpTargetPage;
+    function validURL(url) {
+        try {
+            url = decodeURIComponent(url);
+        }
+        catch (error) {
+            return false;
+        }
+        var pos = url.indexOf(".html");
+        if (pos == -1 || pos != url.length - 5)
+            return false;
+        var allowNumber = false;
+        var allowSep = false;
+        var seenDot = false;
+        for (var i = 0; i < url.length - 5; i++) {
+            var ch = url.charAt(i);
+            if ('a' <= ch && ch <= 'z' ||
+                    'A' <= ch && ch <= 'Z' ||
+                    ch == '$' ||
+                    ch == '_' ||
+                    ch.charCodeAt(0) > 127) {
+                allowNumber = true;
+                allowSep = true;
+            } else if ('0' <= ch && ch <= '9'
+                    || ch == '-') {
+                if (!allowNumber)
+                     return false;
+            } else if (ch == '/' || ch == '.') {
+                if (!allowSep)
+                    return false;
+                allowNumber = false;
+                allowSep = false;
+                if (ch == '.')
+                     seenDot = true;
+                if (ch == '/' && seenDot)
+                     return false;
+            } else {
+                return false;
+            }
+        }
+        return true;
+    }
+    function loadFrames() {
+        if (targetPage != "" && targetPage != "undefined")
+             top.classFrame.location = top.targetPage;
+    }
+</script>
+</head>
+<frameset cols="20%,80%" title="Documentation frame" onload="top.loadFrames()">
+<frame src="allclasses-frame.html" name="packageFrame" title="All classes and interfaces (except non-static nested types)">
+<frame src="org/libjpegturbo/turbojpeg/package-summary.html" name="classFrame" title="Package, class and interface descriptions" scrolling="yes">
+<noframes>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+<h2>Frame Alert</h2>
+<p>This document is designed to be viewed using the frames feature. If you see this message, you are using a non-frame-capable web client. Link to <a href="org/libjpegturbo/turbojpeg/package-summary.html">Non-frame version</a>.</p>
+</noframes>
+</frameset>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/TJ.html Added

@@ -0,0 +1,1381 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>TJ</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev Class</li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJ.html" target="_top">Frames</a></li>
+<li><a href="TJ.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li><a href="#field_summary">Field</a>&nbsp;|&nbsp;</li>
+<li>Constr&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li><a href="#field_detail">Field</a>&nbsp;|&nbsp;</li>
+<li>Constr&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+
+<div class="header">
+<div class="subTitle">org.libjpegturbo.turbojpeg</div>
+<h2 title="Class TJ" class="title">Class TJ</h2>
+</div>
+<div class="contentContainer">
+<ul class="inheritance">
+<li>java.lang.Object</li>
+<li>
+<ul class="inheritance">
+<li>org.libjpegturbo.turbojpeg.TJ</li>
+</ul>
+</li>
+</ul>
+<div class="description">
+<ul class="blockList">
+<li class="blockList">
+<hr>
+<br>
+<pre>public final class <span class="strong">TJ</span>
+extends java.lang.Object</pre>
+<div class="block">TurboJPEG utility class (cannot be instantiated)</div>
+</li>
+</ul>
+</div>
+<div class="summary">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="field_summary">
+
+</a>
+<h3>Field Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Field Summary table, listing fields, and an explanation">
+<caption><span>Fields</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Field and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_CMYK">CS_CMYK</a></strong></code>
+<div class="block">CMYK colorspace.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_GRAY">CS_GRAY</a></strong></code>
+<div class="block">Grayscale colorspace.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_RGB">CS_RGB</a></strong></code>
+<div class="block">RGB colorspace.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_YCbCr">CS_YCbCr</a></strong></code>
+<div class="block">YCbCr colorspace.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_YCCK">CS_YCCK</a></strong></code>
+<div class="block">YCCK colorspace.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#ERR_FATAL">ERR_FATAL</a></strong></code>
+<div class="block">The error was fatal and non-recoverable.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#ERR_WARNING">ERR_WARNING</a></strong></code>
+<div class="block">The error was non-fatal and recoverable, but the image may still be
+ corrupt.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_ACCURATEDCT">FLAG_ACCURATEDCT</a></strong></code>
+<div class="block">Use the most accurate DCT/IDCT algorithm available in the underlying
+ codec.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP">FLAG_BOTTOMUP</a></strong></code>
+<div class="block">The uncompressed source/destination image is stored in bottom-up (Windows,
+ OpenGL) order, not top-down (X11) order.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_FASTDCT">FLAG_FASTDCT</a></strong></code>
+<div class="block">Use the fastest DCT/IDCT algorithm available in the underlying codec.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_FASTUPSAMPLE">FLAG_FASTUPSAMPLE</a></strong></code>
+<div class="block">When decompressing an image that was compressed using chrominance
+ subsampling, use the fastest chrominance upsampling algorithm available in
+ the underlying codec.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCEMMX">FLAG_FORCEMMX</a></strong></code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE">FLAG_FORCESSE</a></strong></code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE2">FLAG_FORCESSE2</a></strong></code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_FORCESSE3">FLAG_FORCESSE3</a></strong></code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_LIMITSCANS">FLAG_LIMITSCANS</a></strong></code>
+<div class="block">Limit the number of progressive JPEG scans that the decompression and
+ transform operations will process.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_PROGRESSIVE">FLAG_PROGRESSIVE</a></strong></code>
+<div class="block">Use progressive entropy coding in JPEG images generated by compression and
+ transform operations.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_STOPONWARNING">FLAG_STOPONWARNING</a></strong></code>
+<div class="block">Immediately discontinue the current compression/decompression/transform
+ operation if the underlying codec throws a warning (non-fatal error).</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#NUMCS">NUMCS</a></strong></code>
+<div class="block">The number of JPEG colorspaces</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#NUMERR">NUMERR</a></strong></code>
+<div class="block">The number of error codes</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#NUMPF">NUMPF</a></strong></code>
+<div class="block">The number of pixel formats</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#NUMSAMP">NUMSAMP</a></strong></code>
+<div class="block">The number of chrominance subsampling options</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_ABGR">PF_ABGR</a></strong></code>
+<div class="block">ABGR pixel format.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_ARGB">PF_ARGB</a></strong></code>
+<div class="block">ARGB pixel format.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_BGR">PF_BGR</a></strong></code>
+<div class="block">BGR pixel format.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_BGRA">PF_BGRA</a></strong></code>
+<div class="block">BGRA pixel format.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_BGRX">PF_BGRX</a></strong></code>
+<div class="block">BGRX pixel format.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_CMYK">PF_CMYK</a></strong></code>
+<div class="block">CMYK pixel format.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_GRAY">PF_GRAY</a></strong></code>
+<div class="block">Grayscale pixel format.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_RGB">PF_RGB</a></strong></code>
+<div class="block">RGB pixel format.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_RGBA">PF_RGBA</a></strong></code>
+<div class="block">RGBA pixel format.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_RGBX">PF_RGBX</a></strong></code>
+<div class="block">RGBX pixel format.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_XBGR">PF_XBGR</a></strong></code>
+<div class="block">XBGR pixel format.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_XRGB">PF_XRGB</a></strong></code>
+<div class="block">XRGB pixel format.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_411">SAMP_411</a></strong></code>
+<div class="block">4:1:1 chrominance subsampling.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_420">SAMP_420</a></strong></code>
+<div class="block">4:2:0 chrominance subsampling.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_422">SAMP_422</a></strong></code>
+<div class="block">4:2:2 chrominance subsampling.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_440">SAMP_440</a></strong></code>
+<div class="block">4:4:0 chrominance subsampling.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444">SAMP_444</a></strong></code>
+<div class="block">4:4:4 chrominance subsampling (no chrominance subsampling).</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_GRAY">SAMP_GRAY</a></strong></code>
+<div class="block">Grayscale.</div>
+</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_summary">
+
+</a>
+<h3>Method Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Method Summary table, listing methods, and an explanation">
+<caption><span>Methods</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Method and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#bufSize(int,%20int,%20int)">bufSize</a></strong>(int&nbsp;width,
+       int&nbsp;height,
+       int&nbsp;jpegSubsamp)</code>
+<div class="block">Returns the maximum size of the buffer (in bytes) required to hold a JPEG
+ image with the given width, height, and level of chrominance subsampling.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int)">bufSizeYUV</a></strong>(int&nbsp;width,
+          int&nbsp;height,
+          int&nbsp;subsamp)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int,%20int)"><code>bufSizeYUV(int, int, int, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int,%20int)">bufSizeYUV</a></strong>(int&nbsp;width,
+          int&nbsp;pad,
+          int&nbsp;height,
+          int&nbsp;subsamp)</code>
+<div class="block">Returns the size of the buffer (in bytes) required to hold a YUV planar
+ image with the given width, height, and level of chrominance subsampling.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getAlphaOffset(int)">getAlphaOffset</a></strong>(int&nbsp;pixelFormat)</code>
+<div class="block">For the given pixel format, returns the number of bytes that the alpha
+ component is offset from the start of the pixel.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getBlueOffset(int)">getBlueOffset</a></strong>(int&nbsp;pixelFormat)</code>
+<div class="block">For the given pixel format, returns the number of bytes that the blue
+ component is offset from the start of the pixel.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getGreenOffset(int)">getGreenOffset</a></strong>(int&nbsp;pixelFormat)</code>
+<div class="block">For the given pixel format, returns the number of bytes that the green
+ component is offset from the start of the pixel.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getMCUHeight(int)">getMCUHeight</a></strong>(int&nbsp;subsamp)</code>
+<div class="block">Returns the MCU block height for the given level of chrominance
+ subsampling.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getMCUWidth(int)">getMCUWidth</a></strong>(int&nbsp;subsamp)</code>
+<div class="block">Returns the MCU block width for the given level of chrominance
+ subsampling.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getPixelSize(int)">getPixelSize</a></strong>(int&nbsp;pixelFormat)</code>
+<div class="block">Returns the pixel size (in bytes) for the given pixel format.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getRedOffset(int)">getRedOffset</a></strong>(int&nbsp;pixelFormat)</code>
+<div class="block">For the given pixel format, returns the number of bytes that the red
+ component is offset from the start of the pixel.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static <a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a></code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getScalingFactors()">getScalingFactors</a></strong>()</code>
+<div class="block">Returns a list of fractional scaling factors that the JPEG decompressor in
+ this implementation of TurboJPEG supports.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#planeHeight(int,%20int,%20int)">planeHeight</a></strong>(int&nbsp;componentID,
+           int&nbsp;height,
+           int&nbsp;subsamp)</code>
+<div class="block">Returns the plane height of a YUV image plane with the given parameters.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#planeSizeYUV(int,%20int,%20int,%20int,%20int)">planeSizeYUV</a></strong>(int&nbsp;componentID,
+            int&nbsp;width,
+            int&nbsp;stride,
+            int&nbsp;height,
+            int&nbsp;subsamp)</code>
+<div class="block">Returns the size of the buffer (in bytes) required to hold a YUV image
+ plane with the given parameters.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJ.html#planeWidth(int,%20int,%20int)">planeWidth</a></strong>(int&nbsp;componentID,
+          int&nbsp;width,
+          int&nbsp;subsamp)</code>
+<div class="block">Returns the plane width of a YUV image plane with the given parameters.</div>
+</td>
+</tr>
+</table>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.lang.Object">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.lang.Object</h3>
+<code>clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait</code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="details">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="field_detail">
+
+</a>
+<h3>Field Detail</h3>
+<a name="NUMSAMP">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>NUMSAMP</h4>
+<pre>public static final&nbsp;int NUMSAMP</pre>
+<div class="block">The number of chrominance subsampling options</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.NUMSAMP">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="SAMP_444">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>SAMP_444</h4>
+<pre>public static final&nbsp;int SAMP_444</pre>
+<div class="block">4:4:4 chrominance subsampling (no chrominance subsampling).  The JPEG
+ or YUV image will contain one chrominance component for every pixel in the
+ source image.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.SAMP_444">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="SAMP_422">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>SAMP_422</h4>
+<pre>public static final&nbsp;int SAMP_422</pre>
+<div class="block">4:2:2 chrominance subsampling.  The JPEG or YUV image will contain one
+ chrominance component for every 2x1 block of pixels in the source image.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.SAMP_422">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="SAMP_420">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>SAMP_420</h4>
+<pre>public static final&nbsp;int SAMP_420</pre>
+<div class="block">4:2:0 chrominance subsampling.  The JPEG or YUV image will contain one
+ chrominance component for every 2x2 block of pixels in the source image.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.SAMP_420">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="SAMP_GRAY">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>SAMP_GRAY</h4>
+<pre>public static final&nbsp;int SAMP_GRAY</pre>
+<div class="block">Grayscale.  The JPEG or YUV image will contain no chrominance components.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.SAMP_GRAY">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="SAMP_440">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>SAMP_440</h4>
+<pre>public static final&nbsp;int SAMP_440</pre>
+<div class="block">4:4:0 chrominance subsampling.  The JPEG or YUV image will contain one
+ chrominance component for every 1x2 block of pixels in the source image.
+ Note that 4:4:0 subsampling is not fully accelerated in libjpeg-turbo.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.SAMP_440">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="SAMP_411">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>SAMP_411</h4>
+<pre>public static final&nbsp;int SAMP_411</pre>
+<div class="block">4:1:1 chrominance subsampling.  The JPEG or YUV image will contain one
+ chrominance component for every 4x1 block of pixels in the source image.
+ JPEG images compressed with 4:1:1 subsampling will be almost exactly the
+ same size as those compressed with 4:2:0 subsampling, and in the
+ aggregate, both subsampling methods produce approximately the same
+ perceptual quality.  However, 4:1:1 is better able to reproduce sharp
+ horizontal features.  Note that 4:1:1 subsampling is not fully accelerated
+ in libjpeg-turbo.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.SAMP_411">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="NUMPF">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>NUMPF</h4>
+<pre>public static final&nbsp;int NUMPF</pre>
+<div class="block">The number of pixel formats</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.NUMPF">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_RGB">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_RGB</h4>
+<pre>public static final&nbsp;int PF_RGB</pre>
+<div class="block">RGB pixel format.  The red, green, and blue components in the image are
+ stored in 3-byte pixels in the order R, G, B from lowest to highest byte
+ address within each pixel.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_RGB">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_BGR">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_BGR</h4>
+<pre>public static final&nbsp;int PF_BGR</pre>
+<div class="block">BGR pixel format.  The red, green, and blue components in the image are
+ stored in 3-byte pixels in the order B, G, R from lowest to highest byte
+ address within each pixel.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_BGR">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_RGBX">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_RGBX</h4>
+<pre>public static final&nbsp;int PF_RGBX</pre>
+<div class="block">RGBX pixel format.  The red, green, and blue components in the image are
+ stored in 4-byte pixels in the order R, G, B from lowest to highest byte
+ address within each pixel.  The X component is ignored when compressing
+ and undefined when decompressing.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_RGBX">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_BGRX">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_BGRX</h4>
+<pre>public static final&nbsp;int PF_BGRX</pre>
+<div class="block">BGRX pixel format.  The red, green, and blue components in the image are
+ stored in 4-byte pixels in the order B, G, R from lowest to highest byte
+ address within each pixel.  The X component is ignored when compressing
+ and undefined when decompressing.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_BGRX">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_XBGR">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_XBGR</h4>
+<pre>public static final&nbsp;int PF_XBGR</pre>
+<div class="block">XBGR pixel format.  The red, green, and blue components in the image are
+ stored in 4-byte pixels in the order R, G, B from highest to lowest byte
+ address within each pixel.  The X component is ignored when compressing
+ and undefined when decompressing.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_XBGR">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_XRGB">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_XRGB</h4>
+<pre>public static final&nbsp;int PF_XRGB</pre>
+<div class="block">XRGB pixel format.  The red, green, and blue components in the image are
+ stored in 4-byte pixels in the order B, G, R from highest to lowest byte
+ address within each pixel.  The X component is ignored when compressing
+ and undefined when decompressing.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_XRGB">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_GRAY">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_GRAY</h4>
+<pre>public static final&nbsp;int PF_GRAY</pre>
+<div class="block">Grayscale pixel format.  Each 1-byte pixel represents a luminance
+ (brightness) level from 0 to 255.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_GRAY">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_RGBA">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_RGBA</h4>
+<pre>public static final&nbsp;int PF_RGBA</pre>
+<div class="block">RGBA pixel format.  This is the same as <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_RGBX"><code>PF_RGBX</code></a>, except that when
+ decompressing, the X byte is guaranteed to be 0xFF, which can be
+ interpreted as an opaque alpha channel.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_RGBA">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_BGRA">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_BGRA</h4>
+<pre>public static final&nbsp;int PF_BGRA</pre>
+<div class="block">BGRA pixel format.  This is the same as <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_BGRX"><code>PF_BGRX</code></a>, except that when
+ decompressing, the X byte is guaranteed to be 0xFF, which can be
+ interpreted as an opaque alpha channel.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_BGRA">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_ABGR">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_ABGR</h4>
+<pre>public static final&nbsp;int PF_ABGR</pre>
+<div class="block">ABGR pixel format.  This is the same as <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_XBGR"><code>PF_XBGR</code></a>, except that when
+ decompressing, the X byte is guaranteed to be 0xFF, which can be
+ interpreted as an opaque alpha channel.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_ABGR">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_ARGB">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_ARGB</h4>
+<pre>public static final&nbsp;int PF_ARGB</pre>
+<div class="block">ARGB pixel format.  This is the same as <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_XRGB"><code>PF_XRGB</code></a>, except that when
+ decompressing, the X byte is guaranteed to be 0xFF, which can be
+ interpreted as an opaque alpha channel.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_ARGB">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="PF_CMYK">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>PF_CMYK</h4>
+<pre>public static final&nbsp;int PF_CMYK</pre>
+<div class="block">CMYK pixel format.  Unlike RGB, which is an additive color model used
+ primarily for display, CMYK (Cyan/Magenta/Yellow/Key) is a subtractive
+ color model used primarily for printing.  In the CMYK color model, the
+ value of each color component typically corresponds to an amount of cyan,
+ magenta, yellow, or black ink that is applied to a white background.  In
+ order to convert between CMYK and RGB, it is necessary to use a color
+ management system (CMS.)  A CMS will attempt to map colors within the
+ printer's gamut to perceptually similar colors in the display's gamut and
+ vice versa, but the mapping is typically not 1:1 or reversible, nor can it
+ be defined with a simple formula.  Thus, such a conversion is out of scope
+ for a codec library.  However, the TurboJPEG API allows for compressing
+ CMYK pixels into a YCCK JPEG image (see <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_YCCK"><code>CS_YCCK</code></a>) and
+ decompressing YCCK JPEG images into CMYK pixels.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.PF_CMYK">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="NUMCS">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>NUMCS</h4>
+<pre>public static final&nbsp;int NUMCS</pre>
+<div class="block">The number of JPEG colorspaces</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.NUMCS">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="CS_RGB">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>CS_RGB</h4>
+<pre>public static final&nbsp;int CS_RGB</pre>
+<div class="block">RGB colorspace.  When compressing the JPEG image, the R, G, and B
+ components in the source image are reordered into image planes, but no
+ colorspace conversion or subsampling is performed.  RGB JPEG images can be
+ decompressed to any of the extended RGB pixel formats or grayscale, but
+ they cannot be decompressed to YUV images.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.CS_RGB">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="CS_YCbCr">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>CS_YCbCr</h4>
+<pre>public static final&nbsp;int CS_YCbCr</pre>
+<div class="block">YCbCr colorspace.  YCbCr is not an absolute colorspace but rather a
+ mathematical transformation of RGB designed solely for storage and
+ transmission.  YCbCr images must be converted to RGB before they can
+ actually be displayed.  In the YCbCr colorspace, the Y (luminance)
+ component represents the black &amp; white portion of the original image,
+ and the Cb and Cr (chrominance) components represent the color portion of
+ the original image.  Originally, the analog equivalent of this
+ transformation allowed the same signal to drive both black &amp; white and
+ color televisions, but JPEG images use YCbCr primarily because it allows
+ the color data to be optionally subsampled for the purposes of reducing
+ bandwidth or disk space.  YCbCr is the most common JPEG colorspace, and
+ YCbCr JPEG images can be compressed from and decompressed to any of the
+ extended RGB pixel formats or grayscale, or they can be decompressed to
+ YUV planar images.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.CS_YCbCr">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="CS_GRAY">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>CS_GRAY</h4>
+<pre>public static final&nbsp;int CS_GRAY</pre>
+<div class="block">Grayscale colorspace.  The JPEG image retains only the luminance data (Y
+ component), and any color data from the source image is discarded.
+ Grayscale JPEG images can be compressed from and decompressed to any of
+ the extended RGB pixel formats or grayscale, or they can be decompressed
+ to YUV planar images.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.CS_GRAY">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="CS_CMYK">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>CS_CMYK</h4>
+<pre>public static final&nbsp;int CS_CMYK</pre>
+<div class="block">CMYK colorspace.  When compressing the JPEG image, the C, M, Y, and K
+ components in the source image are reordered into image planes, but no
+ colorspace conversion or subsampling is performed.  CMYK JPEG images can
+ only be decompressed to CMYK pixels.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.CS_CMYK">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="CS_YCCK">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>CS_YCCK</h4>
+<pre>public static final&nbsp;int CS_YCCK</pre>
+<div class="block">YCCK colorspace.  YCCK (AKA "YCbCrK") is not an absolute colorspace but
+ rather a mathematical transformation of CMYK designed solely for storage
+ and transmission.  It is to CMYK as YCbCr is to RGB.  CMYK pixels can be
+ reversibly transformed into YCCK, and as with YCbCr, the chrominance
+ components in the YCCK pixels can be subsampled without incurring major
+ perceptual loss.  YCCK JPEG images can only be compressed from and
+ decompressed to CMYK pixels.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.CS_YCCK">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_BOTTOMUP">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_BOTTOMUP</h4>
+<pre>public static final&nbsp;int FLAG_BOTTOMUP</pre>
+<div class="block">The uncompressed source/destination image is stored in bottom-up (Windows,
+ OpenGL) order, not top-down (X11) order.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_BOTTOMUP">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_FORCEMMX">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_FORCEMMX</h4>
+<pre>@Deprecated
+public static final&nbsp;int FLAG_FORCEMMX</pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_FORCEMMX">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_FORCESSE">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_FORCESSE</h4>
+<pre>@Deprecated
+public static final&nbsp;int FLAG_FORCESSE</pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_FORCESSE">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_FORCESSE2">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_FORCESSE2</h4>
+<pre>@Deprecated
+public static final&nbsp;int FLAG_FORCESSE2</pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_FORCESSE2">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_FORCESSE3">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_FORCESSE3</h4>
+<pre>@Deprecated
+public static final&nbsp;int FLAG_FORCESSE3</pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_FORCESSE3">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_FASTUPSAMPLE">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_FASTUPSAMPLE</h4>
+<pre>public static final&nbsp;int FLAG_FASTUPSAMPLE</pre>
+<div class="block">When decompressing an image that was compressed using chrominance
+ subsampling, use the fastest chrominance upsampling algorithm available in
+ the underlying codec.  The default is to use smooth upsampling, which
+ creates a smooth transition between neighboring chrominance components in
+ order to reduce upsampling artifacts in the decompressed image.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_FASTUPSAMPLE">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_FASTDCT">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_FASTDCT</h4>
+<pre>public static final&nbsp;int FLAG_FASTDCT</pre>
+<div class="block">Use the fastest DCT/IDCT algorithm available in the underlying codec.  The
+ default if this flag is not specified is implementation-specific.  For
+ example, the implementation of TurboJPEG for libjpeg-turbo uses the fast
+ algorithm by default when compressing, because this has been shown to have
+ only a very slight effect on accuracy, but it uses the accurate algorithm
+ when decompressing, because this has been shown to have a larger effect.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_FASTDCT">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_ACCURATEDCT">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_ACCURATEDCT</h4>
+<pre>public static final&nbsp;int FLAG_ACCURATEDCT</pre>
+<div class="block">Use the most accurate DCT/IDCT algorithm available in the underlying
+ codec.  The default if this flag is not specified is
+ implementation-specific.  For example, the implementation of TurboJPEG for
+ libjpeg-turbo uses the fast algorithm by default when compressing,
+ because this has been shown to have only a very slight effect on accuracy,
+ but it uses the accurate algorithm when decompressing, because this has
+ been shown to have a larger effect.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_ACCURATEDCT">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_STOPONWARNING">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_STOPONWARNING</h4>
+<pre>public static final&nbsp;int FLAG_STOPONWARNING</pre>
+<div class="block">Immediately discontinue the current compression/decompression/transform
+ operation if the underlying codec throws a warning (non-fatal error).  The
+ default behavior is to allow the operation to complete unless a fatal
+ error is encountered.
+ <p>
+ NOTE: due to the design of the TurboJPEG Java API, only certain methods
+ (specifically, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><code>TJDecompressor.decompress*()</code></a> methods
+ with a void return type) will complete and leave the output image in a
+ fully recoverable state after a non-fatal error occurs.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_STOPONWARNING">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_PROGRESSIVE">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_PROGRESSIVE</h4>
+<pre>public static final&nbsp;int FLAG_PROGRESSIVE</pre>
+<div class="block">Use progressive entropy coding in JPEG images generated by compression and
+ transform operations.  Progressive entropy coding will generally improve
+ compression relative to baseline entropy coding (the default), but it will
+ reduce compression and decompression performance considerably.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_PROGRESSIVE">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="FLAG_LIMITSCANS">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>FLAG_LIMITSCANS</h4>
+<pre>public static final&nbsp;int FLAG_LIMITSCANS</pre>
+<div class="block">Limit the number of progressive JPEG scans that the decompression and
+ transform operations will process.  If a progressive JPEG image contains
+ an unreasonably large number of scans, then this flag will cause the
+ decompression and transform operations to throw an error.  The primary
+ purpose of this is to allow security-critical applications to guard
+ against an exploit of the progressive JPEG format described in
+ <a href="https://libjpeg-turbo.org/pmwiki/uploads/About/TwoIssueswiththeJPEGStandard.pdf" target="_blank">this report</a>.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.FLAG_LIMITSCANS">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="NUMERR">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>NUMERR</h4>
+<pre>public static final&nbsp;int NUMERR</pre>
+<div class="block">The number of error codes</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.NUMERR">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="ERR_WARNING">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>ERR_WARNING</h4>
+<pre>public static final&nbsp;int ERR_WARNING</pre>
+<div class="block">The error was non-fatal and recoverable, but the image may still be
+ corrupt.
+ <p>
+ NOTE: due to the design of the TurboJPEG Java API, only certain methods
+ (specifically, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><code>TJDecompressor.decompress*()</code></a> methods
+ with a void return type) will complete and leave the output image in a
+ fully recoverable state after a non-fatal error occurs.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.ERR_WARNING">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="ERR_FATAL">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>ERR_FATAL</h4>
+<pre>public static final&nbsp;int ERR_FATAL</pre>
+<div class="block">The error was fatal and non-recoverable.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJ.ERR_FATAL">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_detail">
+
+</a>
+<h3>Method Detail</h3>
+<a name="getMCUWidth(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getMCUWidth</h4>
+<pre>public static&nbsp;int&nbsp;getMCUWidth(int&nbsp;subsamp)</pre>
+<div class="block">Returns the MCU block width for the given level of chrominance
+ subsampling.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>subsamp</code> - the level of chrominance subsampling (one of
+ <code>SAMP_*</code>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the MCU block width for the given level of chrominance
+ subsampling.</dd></dl>
+</li>
+</ul>
+<a name="getMCUHeight(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getMCUHeight</h4>
+<pre>public static&nbsp;int&nbsp;getMCUHeight(int&nbsp;subsamp)</pre>
+<div class="block">Returns the MCU block height for the given level of chrominance
+ subsampling.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>subsamp</code> - the level of chrominance subsampling (one of
+ <code>SAMP_*</code>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the MCU block height for the given level of chrominance
+ subsampling.</dd></dl>
+</li>
+</ul>
+<a name="getPixelSize(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getPixelSize</h4>
+<pre>public static&nbsp;int&nbsp;getPixelSize(int&nbsp;pixelFormat)</pre>
+<div class="block">Returns the pixel size (in bytes) for the given pixel format.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>pixelFormat</code> - the pixel format (one of <code>PF_*</code>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the pixel size (in bytes) for the given pixel format.</dd></dl>
+</li>
+</ul>
+<a name="getRedOffset(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getRedOffset</h4>
+<pre>public static&nbsp;int&nbsp;getRedOffset(int&nbsp;pixelFormat)</pre>
+<div class="block">For the given pixel format, returns the number of bytes that the red
+ component is offset from the start of the pixel.  For instance, if a pixel
+ of format <code>TJ.PF_BGRX</code> is stored in <code>char pixel</code>,
+ then the red component will be
+ <code>pixelTJ.getRedOffset(TJ.PF_BGRX)</code>.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>pixelFormat</code> - the pixel format (one of <code>PF_*</code>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the red offset for the given pixel format, or -1 if the pixel
+ format does not have a red component.</dd></dl>
+</li>
+</ul>
+<a name="getGreenOffset(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getGreenOffset</h4>
+<pre>public static&nbsp;int&nbsp;getGreenOffset(int&nbsp;pixelFormat)</pre>
+<div class="block">For the given pixel format, returns the number of bytes that the green
+ component is offset from the start of the pixel.  For instance, if a pixel
+ of format <code>TJ.PF_BGRX</code> is stored in <code>char pixel</code>,
+ then the green component will be
+ <code>pixelTJ.getGreenOffset(TJ.PF_BGRX)</code>.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>pixelFormat</code> - the pixel format (one of <code>PF_*</code>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the green offset for the given pixel format, or -1 if the pixel
+ format does not have a green component.</dd></dl>
+</li>
+</ul>
+<a name="getBlueOffset(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getBlueOffset</h4>
+<pre>public static&nbsp;int&nbsp;getBlueOffset(int&nbsp;pixelFormat)</pre>
+<div class="block">For the given pixel format, returns the number of bytes that the blue
+ component is offset from the start of the pixel.  For instance, if a pixel
+ of format <code>TJ.PF_BGRX</code> is stored in <code>char pixel</code>,
+ then the blue component will be
+ <code>pixelTJ.getBlueOffset(TJ.PF_BGRX)</code>.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>pixelFormat</code> - the pixel format (one of <code>PF_*</code>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the blue offset for the given pixel format, or -1 if the pixel
+ format does not have a blue component.</dd></dl>
+</li>
+</ul>
+<a name="getAlphaOffset(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getAlphaOffset</h4>
+<pre>public static&nbsp;int&nbsp;getAlphaOffset(int&nbsp;pixelFormat)</pre>
+<div class="block">For the given pixel format, returns the number of bytes that the alpha
+ component is offset from the start of the pixel.  For instance, if a pixel
+ of format <code>TJ.PF_BGRA</code> is stored in <code>char pixel</code>,
+ then the alpha component will be
+ <code>pixelTJ.getAlphaOffset(TJ.PF_BGRA)</code>.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>pixelFormat</code> - the pixel format (one of <code>PF_*</code>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the alpha offset for the given pixel format, or -1 if the pixel
+ format does not have a alpha component.</dd></dl>
+</li>
+</ul>
+<a name="bufSize(int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>bufSize</h4>
+<pre>public static&nbsp;int&nbsp;bufSize(int&nbsp;width,
+          int&nbsp;height,
+          int&nbsp;jpegSubsamp)</pre>
+<div class="block">Returns the maximum size of the buffer (in bytes) required to hold a JPEG
+ image with the given width, height, and level of chrominance subsampling.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>width</code> - the width (in pixels) of the JPEG image</dd><dd><code>height</code> - the height (in pixels) of the JPEG image</dd><dd><code>jpegSubsamp</code> - the level of chrominance subsampling to be used when
+ generating the JPEG image (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><code>TJ.SAMP_*</code></a>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the maximum size of the buffer (in bytes) required to hold a JPEG
+ image with the given width, height, and level of chrominance subsampling.</dd></dl>
+</li>
+</ul>
+<a name="bufSizeYUV(int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>bufSizeYUV</h4>
+<pre>public static&nbsp;int&nbsp;bufSizeYUV(int&nbsp;width,
+             int&nbsp;pad,
+             int&nbsp;height,
+             int&nbsp;subsamp)</pre>
+<div class="block">Returns the size of the buffer (in bytes) required to hold a YUV planar
+ image with the given width, height, and level of chrominance subsampling.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>width</code> - the width (in pixels) of the YUV image</dd><dd><code>pad</code> - the width of each line in each plane of the image is padded to
+ the nearest multiple of this number of bytes (must be a power of 2.)</dd><dd><code>height</code> - the height (in pixels) of the YUV image</dd><dd><code>subsamp</code> - the level of chrominance subsampling used in the YUV
+ image (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><code>TJ.SAMP_*</code></a>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the size of the buffer (in bytes) required to hold a YUV planar
+ image with the given width, height, and level of chrominance subsampling.</dd></dl>
+</li>
+</ul>
+<a name="bufSizeYUV(int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>bufSizeYUV</h4>
+<pre>@Deprecated
+public static&nbsp;int&nbsp;bufSizeYUV(int&nbsp;width,
+                        int&nbsp;height,
+                        int&nbsp;subsamp)</pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int,%20int)"><code>bufSizeYUV(int, int, int, int)</code></a> instead.</i></div>
+</li>
+</ul>
+<a name="planeSizeYUV(int, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>planeSizeYUV</h4>
+<pre>public static&nbsp;int&nbsp;planeSizeYUV(int&nbsp;componentID,
+               int&nbsp;width,
+               int&nbsp;stride,
+               int&nbsp;height,
+               int&nbsp;subsamp)</pre>
+<div class="block">Returns the size of the buffer (in bytes) required to hold a YUV image
+ plane with the given parameters.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>componentID</code> - ID number of the image plane (0 = Y, 1 = U/Cb,
+ 2 = V/Cr)</dd><dd><code>width</code> - width (in pixels) of the YUV image.  NOTE: this is the width
+ of the whole image, not the plane width.</dd><dd><code>stride</code> - bytes per line in the image plane.</dd><dd><code>height</code> - height (in pixels) of the YUV image.  NOTE: this is the
+ height of the whole image, not the plane height.</dd><dd><code>subsamp</code> - the level of chrominance subsampling used in the YUV
+ image (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><code>TJ.SAMP_*</code></a>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the size of the buffer (in bytes) required to hold a YUV planar
+ image with the given parameters.</dd></dl>
+</li>
+</ul>
+<a name="planeWidth(int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>planeWidth</h4>
+<pre>public static&nbsp;int&nbsp;planeWidth(int&nbsp;componentID,
+             int&nbsp;width,
+             int&nbsp;subsamp)</pre>
+<div class="block">Returns the plane width of a YUV image plane with the given parameters.
+ Refer to <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>YUVImage</code></a> for a description of plane width.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>componentID</code> - ID number of the image plane (0 = Y, 1 = U/Cb,
+ 2 = V/Cr)</dd><dd><code>width</code> - width (in pixels) of the YUV image</dd><dd><code>subsamp</code> - the level of chrominance subsampling used in the YUV image
+ (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><code>TJ.SAMP_*</code></a>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the plane width of a YUV image plane with the given parameters.</dd></dl>
+</li>
+</ul>
+<a name="planeHeight(int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>planeHeight</h4>
+<pre>public static&nbsp;int&nbsp;planeHeight(int&nbsp;componentID,
+              int&nbsp;height,
+              int&nbsp;subsamp)</pre>
+<div class="block">Returns the plane height of a YUV image plane with the given parameters.
+ Refer to <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>YUVImage</code></a> for a description of plane height.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>componentID</code> - ID number of the image plane (0 = Y, 1 = U/Cb,
+ 2 = V/Cr)</dd><dd><code>height</code> - height (in pixels) of the YUV image</dd><dd><code>subsamp</code> - the level of chrominance subsampling used in the YUV image
+ (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><code>TJ.SAMP_*</code></a>)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the plane height of a YUV image plane with the given parameters.</dd></dl>
+</li>
+</ul>
+<a name="getScalingFactors()">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>getScalingFactors</h4>
+<pre>public static&nbsp;<a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a>&nbsp;getScalingFactors()</pre>
+<div class="block">Returns a list of fractional scaling factors that the JPEG decompressor in
+ this implementation of TurboJPEG supports.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>a list of fractional scaling factors that the JPEG decompressor in
+ this implementation of TurboJPEG supports.</dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev Class</li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJ.html" target="_top">Frames</a></li>
+<li><a href="TJ.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li><a href="#field_summary">Field</a>&nbsp;|&nbsp;</li>
+<li>Constr&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li><a href="#field_detail">Field</a>&nbsp;|&nbsp;</li>
+<li>Constr&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/TJCompressor.html Added

@@ -0,0 +1,926 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>TJCompressor</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJCompressor.html" target="_top">Frames</a></li>
+<li><a href="TJCompressor.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+
+<div class="header">
+<div class="subTitle">org.libjpegturbo.turbojpeg</div>
+<h2 title="Class TJCompressor" class="title">Class TJCompressor</h2>
+</div>
+<div class="contentContainer">
+<ul class="inheritance">
+<li>java.lang.Object</li>
+<li>
+<ul class="inheritance">
+<li>org.libjpegturbo.turbojpeg.TJCompressor</li>
+</ul>
+</li>
+</ul>
+<div class="description">
+<ul class="blockList">
+<li class="blockList">
+<dl>
+<dt>All Implemented Interfaces:</dt>
+<dd>java.io.Closeable, java.lang.AutoCloseable</dd>
+</dl>
+<hr>
+<br>
+<pre>public class <span class="strong">TJCompressor</span>
+extends java.lang.Object
+implements java.io.Closeable</pre>
+<div class="block">TurboJPEG compressor</div>
+</li>
+</ul>
+</div>
+<div class="summary">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_summary">
+
+</a>
+<h3>Constructor Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Constructor Summary table, listing constructors, and an explanation">
+<caption><span>Constructors</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Constructor and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor()">TJCompressor</a></strong>()</code>
+<div class="block">Create a TurboJPEG compressor instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)">TJCompressor</a></strong>(java.awt.image.BufferedImage&nbsp;srcImage,
+            int&nbsp;x,
+            int&nbsp;y,
+            int&nbsp;width,
+            int&nbsp;height)</code>
+<div class="block">Create a TurboJPEG compressor instance and associate the uncompressed
+ source image stored in <code>srcImage</code> with the newly created
+ instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(byte,%20int,%20int,%20int,%20int)">TJCompressor</a></strong>(byte&nbsp;srcImage,
+            int&nbsp;width,
+            int&nbsp;pitch,
+            int&nbsp;height,
+            int&nbsp;pixelFormat)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>TJCompressor(byte, int, int, int, int, int, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(byte,%20int,%20int,%20int,%20int,%20int,%20int)">TJCompressor</a></strong>(byte&nbsp;srcImage,
+            int&nbsp;x,
+            int&nbsp;y,
+            int&nbsp;width,
+            int&nbsp;pitch,
+            int&nbsp;height,
+            int&nbsp;pixelFormat)</code>
+<div class="block">Create a TurboJPEG compressor instance and associate the uncompressed
+ source image stored in <code>srcImage</code> with the newly created
+ instance.</div>
+</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_summary">
+
+</a>
+<h3>Method Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Method Summary table, listing methods, and an explanation">
+<caption><span>Methods</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Method and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#close()">close</a></strong>()</code>
+<div class="block">Free the native structures associated with this compressor instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#compress(java.awt.image.BufferedImage,%20byte,%20int)">compress</a></strong>(java.awt.image.BufferedImage&nbsp;srcImage,
+        byte&nbsp;dstBuf,
+        int&nbsp;flags)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#compress(byte,%20int)"><code>compress(byte, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#compress(java.awt.image.BufferedImage,%20int)">compress</a></strong>(java.awt.image.BufferedImage&nbsp;srcImage,
+        int&nbsp;flags)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#compress(int)"><code>compress(int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#compress(byte,%20int)">compress</a></strong>(byte&nbsp;dstBuf,
+        int&nbsp;flags)</code>
+<div class="block">Compress the uncompressed source image associated with this compressor
+ instance and output a JPEG image to the given destination buffer.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#compress(int)">compress</a></strong>(int&nbsp;flags)</code>
+<div class="block">Compress the uncompressed source image associated with this compressor
+ instance and return a buffer containing a JPEG image.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(java.awt.image.BufferedImage,%20byte,%20int)">encodeYUV</a></strong>(java.awt.image.BufferedImage&nbsp;srcImage,
+         byte&nbsp;dstBuf,
+         int&nbsp;flags)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(byte,%20int)"><code>encodeYUV(byte, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(java.awt.image.BufferedImage,%20int)">encodeYUV</a></strong>(java.awt.image.BufferedImage&nbsp;srcImage,
+         int&nbsp;flags)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)"><code>encodeYUV(int, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(byte,%20int)">encodeYUV</a></strong>(byte&nbsp;dstBuf,
+         int&nbsp;flags)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)"><code>encodeYUV(YUVImage, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int)">encodeYUV</a></strong>(int&nbsp;flags)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)"><code>encodeYUV(int, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)">encodeYUV</a></strong>(int&nbsp;strides,
+         int&nbsp;flags)</code>
+<div class="block">Encode the uncompressed source image associated with this compressor
+ instance into separate Y, U (Cb), and V (Cr) image planes and return a
+ <code>YUVImage</code> instance containing the encoded image planes.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)">encodeYUV</a></strong>(int&nbsp;pad,
+         int&nbsp;flags)</code>
+<div class="block">Encode the uncompressed source image associated with this compressor
+ instance into a unified YUV planar image buffer and return a
+ <code>YUVImage</code> instance containing the encoded image.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)">encodeYUV</a></strong>(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;dstImage,
+         int&nbsp;flags)</code>
+<div class="block">Encode the uncompressed source image associated with this compressor
+ instance into a YUV planar image and store it in the given
+ <code>YUVImage</code> instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#finalize()">finalize</a></strong>()</code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#getCompressedSize()">getCompressedSize</a></strong>()</code>
+<div class="block">Returns the size of the image (in bytes) generated by the most recent
+ compress operation.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setJPEGQuality(int)">setJPEGQuality</a></strong>(int&nbsp;quality)</code>
+<div class="block">Set the JPEG image quality level for subsequent compress operations.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)">setSourceImage</a></strong>(java.awt.image.BufferedImage&nbsp;srcImage,
+              int&nbsp;x,
+              int&nbsp;y,
+              int&nbsp;width,
+              int&nbsp;height)</code>
+<div class="block">Associate an uncompressed RGB or grayscale source image with this
+ compressor instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int)">setSourceImage</a></strong>(byte&nbsp;srcImage,
+              int&nbsp;width,
+              int&nbsp;pitch,
+              int&nbsp;height,
+              int&nbsp;pixelFormat)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>setSourceImage(byte, int, int, int, int, int, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)">setSourceImage</a></strong>(byte&nbsp;srcImage,
+              int&nbsp;x,
+              int&nbsp;y,
+              int&nbsp;width,
+              int&nbsp;pitch,
+              int&nbsp;height,
+              int&nbsp;pixelFormat)</code>
+<div class="block">Associate an uncompressed RGB, grayscale, or CMYK source image with this
+ compressor instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(org.libjpegturbo.turbojpeg.YUVImage)">setSourceImage</a></strong>(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;srcImage)</code>
+<div class="block">Associate an uncompressed YUV planar source image with this compressor
+ instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSubsamp(int)">setSubsamp</a></strong>(int&nbsp;newSubsamp)</code>
+<div class="block">Set the level of chrominance subsampling for subsequent compress/encode
+ operations.</div>
+</td>
+</tr>
+</table>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.lang.Object">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.lang.Object</h3>
+<code>clone, equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait</code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="details">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_detail">
+
+</a>
+<h3>Constructor Detail</h3>
+<a name="TJCompressor()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJCompressor</h4>
+<pre>public&nbsp;TJCompressor()
+             throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG compressor instance.</div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="TJCompressor(byte, int, int, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJCompressor</h4>
+<pre>public&nbsp;TJCompressor(byte&nbsp;srcImage,
+            int&nbsp;x,
+            int&nbsp;y,
+            int&nbsp;width,
+            int&nbsp;pitch,
+            int&nbsp;height,
+            int&nbsp;pixelFormat)
+             throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG compressor instance and associate the uncompressed
+ source image stored in <code>srcImage</code> with the newly created
+ instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>srcImage</code> - see <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>setSourceImage(byte, int, int, int, int, int, int)</code></a> for description</dd><dd><code>x</code> - see <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>setSourceImage(byte, int, int, int, int, int, int)</code></a> for description</dd><dd><code>y</code> - see <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>setSourceImage(byte, int, int, int, int, int, int)</code></a> for description</dd><dd><code>width</code> - see <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>setSourceImage(byte, int, int, int, int, int, int)</code></a> for description</dd><dd><code>pitch</code> - see <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>setSourceImage(byte, int, int, int, int, int, int)</code></a> for description</dd><dd><code>height</code> - see <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>setSourceImage(byte, int, int, int, int, int, int)</code></a> for description</dd><dd><code>pixelFormat</code> - pixel format of the source image (one of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_RGB"><code>TJ.PF_*</code></a>)</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="TJCompressor(byte, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJCompressor</h4>
+<pre>@Deprecated
+public&nbsp;TJCompressor(byte&nbsp;srcImage,
+                       int&nbsp;width,
+                       int&nbsp;pitch,
+                       int&nbsp;height,
+                       int&nbsp;pixelFormat)
+             throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#TJCompressor(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>TJCompressor(byte, int, int, int, int, int, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="TJCompressor(java.awt.image.BufferedImage, int, int, int, int)">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>TJCompressor</h4>
+<pre>public&nbsp;TJCompressor(java.awt.image.BufferedImage&nbsp;srcImage,
+            int&nbsp;x,
+            int&nbsp;y,
+            int&nbsp;width,
+            int&nbsp;height)
+             throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG compressor instance and associate the uncompressed
+ source image stored in <code>srcImage</code> with the newly created
+ instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>srcImage</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> for description</dd><dd><code>x</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> for description</dd><dd><code>y</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> for description</dd><dd><code>width</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> for description</dd><dd><code>height</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> for description</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_detail">
+
+</a>
+<h3>Method Detail</h3>
+<a name="setSourceImage(byte, int, int, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setSourceImage</h4>
+<pre>public&nbsp;void&nbsp;setSourceImage(byte&nbsp;srcImage,
+                  int&nbsp;x,
+                  int&nbsp;y,
+                  int&nbsp;width,
+                  int&nbsp;pitch,
+                  int&nbsp;height,
+                  int&nbsp;pixelFormat)
+                    throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Associate an uncompressed RGB, grayscale, or CMYK source image with this
+ compressor instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>srcImage</code> - image buffer containing RGB, grayscale, or CMYK pixels to
+ be compressed or encoded.  This buffer is not modified.</dd><dd><code>x</code> - x offset (in pixels) of the region in the source image from which
+ the JPEG or YUV image should be compressed/encoded</dd><dd><code>y</code> - y offset (in pixels) of the region in the source image from which
+ the JPEG or YUV image should be compressed/encoded</dd><dd><code>width</code> - width (in pixels) of the region in the source image from
+ which the JPEG or YUV image should be compressed/encoded</dd><dd><code>pitch</code> - bytes per line of the source image.  Normally, this should be
+ <code>width * TJ.pixelSize(pixelFormat)</code> if the source image is
+ unpadded, but you can use this parameter to, for instance, specify that
+ the scanlines in the source image are padded to a 4-byte boundary or to
+ compress/encode a JPEG or YUV image from a region of a larger source
+ image.  You can also be clever and use this parameter to skip lines, etc.
+ Setting this parameter to 0 is the equivalent of setting it to
+ <code>width * TJ.pixelSize(pixelFormat)</code>.</dd><dd><code>height</code> - height (in pixels) of the region in the source image from
+ which the JPEG or YUV image should be compressed/encoded</dd><dd><code>pixelFormat</code> - pixel format of the source image (one of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_RGB"><code>TJ.PF_*</code></a>)</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="setSourceImage(byte, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setSourceImage</h4>
+<pre>@Deprecated
+public&nbsp;void&nbsp;setSourceImage(byte&nbsp;srcImage,
+                             int&nbsp;width,
+                             int&nbsp;pitch,
+                             int&nbsp;height,
+                             int&nbsp;pixelFormat)
+                    throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(byte,%20int,%20int,%20int,%20int,%20int,%20int)"><code>setSourceImage(byte, int, int, int, int, int, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="setSourceImage(java.awt.image.BufferedImage, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setSourceImage</h4>
+<pre>public&nbsp;void&nbsp;setSourceImage(java.awt.image.BufferedImage&nbsp;srcImage,
+                  int&nbsp;x,
+                  int&nbsp;y,
+                  int&nbsp;width,
+                  int&nbsp;height)
+                    throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Associate an uncompressed RGB or grayscale source image with this
+ compressor instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>srcImage</code> - a <code>BufferedImage</code> instance containing RGB or
+ grayscale pixels to be compressed or encoded.  This image is not modified.</dd><dd><code>x</code> - x offset (in pixels) of the region in the source image from which
+ the JPEG or YUV image should be compressed/encoded</dd><dd><code>y</code> - y offset (in pixels) of the region in the source image from which
+ the JPEG or YUV image should be compressed/encoded</dd><dd><code>width</code> - width (in pixels) of the region in the source image from
+ which the JPEG or YUV image should be compressed/encoded (0 = use the
+ width of the source image)</dd><dd><code>height</code> - height (in pixels) of the region in the source image from
+ which the JPEG or YUV image should be compressed/encoded (0 = use the
+ height of the source image)</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="setSourceImage(org.libjpegturbo.turbojpeg.YUVImage)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setSourceImage</h4>
+<pre>public&nbsp;void&nbsp;setSourceImage(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;srcImage)
+                    throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Associate an uncompressed YUV planar source image with this compressor
+ instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>srcImage</code> - YUV planar image to be compressed.  This image is not
+ modified.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="setSubsamp(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setSubsamp</h4>
+<pre>public&nbsp;void&nbsp;setSubsamp(int&nbsp;newSubsamp)</pre>
+<div class="block">Set the level of chrominance subsampling for subsequent compress/encode
+ operations.  When pixels are converted from RGB to YCbCr (see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_YCbCr"><code>TJ.CS_YCbCr</code></a>) or from CMYK to YCCK (see <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_YCCK"><code>TJ.CS_YCCK</code></a>) as part
+ of the JPEG compression process, some of the Cb and Cr (chrominance)
+ components can be discarded or averaged together to produce a smaller
+ image with little perceptible loss of image clarity (the human eye is more
+ sensitive to small changes in brightness than to small changes in color.)
+ This is called "chrominance subsampling".
+ <p>
+ NOTE: This method has no effect when compressing a JPEG image from a YUV
+ planar source.  In that case, the level of chrominance subsampling in
+ the JPEG image is determined by the source.  Furthermore, this method has
+ no effect when encoding to a pre-allocated <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>YUVImage</code></a> instance.  In
+ that case, the level of chrominance subsampling is determined by the
+ destination.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>newSubsamp</code> - the level of chrominance subsampling to use in
+ subsequent compress/encode oeprations (one of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444"><code>TJ.SAMP_*</code></a>)</dd></dl>
+</li>
+</ul>
+<a name="setJPEGQuality(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setJPEGQuality</h4>
+<pre>public&nbsp;void&nbsp;setJPEGQuality(int&nbsp;quality)</pre>
+<div class="block">Set the JPEG image quality level for subsequent compress operations.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>quality</code> - the new JPEG image quality level (1 to 100, 1 = worst,
+ 100 = best)</dd></dl>
+</li>
+</ul>
+<a name="compress(byte, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>compress</h4>
+<pre>public&nbsp;void&nbsp;compress(byte&nbsp;dstBuf,
+            int&nbsp;flags)
+              throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Compress the uncompressed source image associated with this compressor
+ instance and output a JPEG image to the given destination buffer.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>dstBuf</code> - buffer that will receive the JPEG image.  Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#bufSize(int,%20int,%20int)"><code>TJ.bufSize(int, int, int)</code></a> to determine the maximum size for this buffer based on
+ the source image's width and height and the desired level of chrominance
+ subsampling.</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="compress(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>compress</h4>
+<pre>public&nbsp;byte&nbsp;compress(int&nbsp;flags)
+                throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Compress the uncompressed source image associated with this compressor
+ instance and return a buffer containing a JPEG image.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Returns:</span></dt><dd>a buffer containing a JPEG image.  The length of this buffer will
+ not be equal to the size of the JPEG image.  Use <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#getCompressedSize()"><code>getCompressedSize()</code></a> to obtain the size of the JPEG image.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="compress(java.awt.image.BufferedImage, byte, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>compress</h4>
+<pre>@Deprecated
+public&nbsp;void&nbsp;compress(java.awt.image.BufferedImage&nbsp;srcImage,
+                       byte&nbsp;dstBuf,
+                       int&nbsp;flags)
+              throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#compress(byte,%20int)"><code>compress(byte, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="compress(java.awt.image.BufferedImage, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>compress</h4>
+<pre>@Deprecated
+public&nbsp;byte&nbsp;compress(java.awt.image.BufferedImage&nbsp;srcImage,
+                         int&nbsp;flags)
+                throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#compress(int)"><code>compress(int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="encodeYUV(org.libjpegturbo.turbojpeg.YUVImage, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>encodeYUV</h4>
+<pre>public&nbsp;void&nbsp;encodeYUV(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;dstImage,
+             int&nbsp;flags)
+               throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Encode the uncompressed source image associated with this compressor
+ instance into a YUV planar image and store it in the given
+ <code>YUVImage</code> instance.   This method uses the accelerated color
+ conversion routines in TurboJPEG's underlying codec but does not execute
+ any of the other steps in the JPEG compression process.  Encoding
+ CMYK source images to YUV is not supported.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>dstImage</code> - <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>YUVImage</code></a> instance that will receive the YUV planar
+ image</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="encodeYUV(byte, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>encodeYUV</h4>
+<pre>@Deprecated
+public&nbsp;void&nbsp;encodeYUV(byte&nbsp;dstBuf,
+                        int&nbsp;flags)
+               throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)"><code>encodeYUV(YUVImage, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="encodeYUV(int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>encodeYUV</h4>
+<pre>public&nbsp;<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;encodeYUV(int&nbsp;pad,
+                 int&nbsp;flags)
+                   throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Encode the uncompressed source image associated with this compressor
+ instance into a unified YUV planar image buffer and return a
+ <code>YUVImage</code> instance containing the encoded image.  This method
+ uses the accelerated color conversion routines in TurboJPEG's underlying
+ codec but does not execute any of the other steps in the JPEG compression
+ process.  Encoding CMYK source images to YUV is not supported.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>pad</code> - the width of each line in each plane of the YUV image will be
+ padded to the nearest multiple of this number of bytes (must be a power of
+ 2.)</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Returns:</span></dt><dd>a YUV planar image.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="encodeYUV(int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>encodeYUV</h4>
+<pre>public&nbsp;<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;encodeYUV(int&nbsp;strides,
+                 int&nbsp;flags)
+                   throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Encode the uncompressed source image associated with this compressor
+ instance into separate Y, U (Cb), and V (Cr) image planes and return a
+ <code>YUVImage</code> instance containing the encoded image planes.  This
+ method uses the accelerated color conversion routines in TurboJPEG's
+ underlying codec but does not execute any of the other steps in the JPEG
+ compression process.  Encoding CMYK source images to YUV is not supported.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>strides</code> - an array of integers, each specifying the number of bytes
+ per line in the corresponding plane of the output image.  Setting the
+ stride for any plane to 0 is the same as setting it to the component width
+ of the plane.  If <code>strides</code> is null, then the strides for all
+ planes will be set to their respective component widths.  You can adjust
+ the strides in order to add an arbitrary amount of line padding to each
+ plane.</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Returns:</span></dt><dd>a YUV planar image.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="encodeYUV(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>encodeYUV</h4>
+<pre>@Deprecated
+public&nbsp;byte&nbsp;encodeYUV(int&nbsp;flags)
+                 throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)"><code>encodeYUV(int, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="encodeYUV(java.awt.image.BufferedImage, byte, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>encodeYUV</h4>
+<pre>@Deprecated
+public&nbsp;void&nbsp;encodeYUV(java.awt.image.BufferedImage&nbsp;srcImage,
+                        byte&nbsp;dstBuf,
+                        int&nbsp;flags)
+               throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(byte,%20int)"><code>encodeYUV(byte, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="encodeYUV(java.awt.image.BufferedImage, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>encodeYUV</h4>
+<pre>@Deprecated
+public&nbsp;byte&nbsp;encodeYUV(java.awt.image.BufferedImage&nbsp;srcImage,
+                          int&nbsp;flags)
+                 throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#setSourceImage(java.awt.image.BufferedImage,%20int,%20int,%20int,%20int)"><code>setSourceImage(BufferedImage, int, int, int, int)</code></a> and
+ <a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html#encodeYUV(int,%20int)"><code>encodeYUV(int, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="getCompressedSize()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getCompressedSize</h4>
+<pre>public&nbsp;int&nbsp;getCompressedSize()</pre>
+<div class="block">Returns the size of the image (in bytes) generated by the most recent
+ compress operation.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the size of the image (in bytes) generated by the most recent
+ compress operation.</dd></dl>
+</li>
+</ul>
+<a name="close()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>close</h4>
+<pre>public&nbsp;void&nbsp;close()
+           throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Free the native structures associated with this compressor instance.</div>
+<dl>
+<dt><strong>Specified by:</strong></dt>
+<dd><code>close</code>&nbsp;in interface&nbsp;<code>java.io.Closeable</code></dd>
+<dt><strong>Specified by:</strong></dt>
+<dd><code>close</code>&nbsp;in interface&nbsp;<code>java.lang.AutoCloseable</code></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="finalize()">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>finalize</h4>
+<pre>protected&nbsp;void&nbsp;finalize()
+                 throws java.lang.Throwable</pre>
+<dl>
+<dt><strong>Overrides:</strong></dt>
+<dd><code>finalize</code>&nbsp;in class&nbsp;<code>java.lang.Object</code></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code>java.lang.Throwable</code></dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJCompressor.html" target="_top">Frames</a></li>
+<li><a href="TJCompressor.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/TJCustomFilter.html Added

@@ -0,0 +1,241 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>TJCustomFilter</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJCustomFilter.html" target="_top">Frames</a></li>
+<li><a href="TJCustomFilter.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li>Constr&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li>Constr&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+
+<div class="header">
+<div class="subTitle">org.libjpegturbo.turbojpeg</div>
+<h2 title="Interface TJCustomFilter" class="title">Interface TJCustomFilter</h2>
+</div>
+<div class="contentContainer">
+<div class="description">
+<ul class="blockList">
+<li class="blockList">
+<hr>
+<br>
+<pre>public interface <span class="strong">TJCustomFilter</span></pre>
+<div class="block">Custom filter callback interface</div>
+</li>
+</ul>
+</div>
+<div class="summary">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="method_summary">
+
+</a>
+<h3>Method Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Method Summary table, listing methods, and an explanation">
+<caption><span>Methods</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Method and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html#customFilter(java.nio.ShortBuffer,%20java.awt.Rectangle,%20java.awt.Rectangle,%20int,%20int,%20org.libjpegturbo.turbojpeg.TJTransform)">customFilter</a></strong>(java.nio.ShortBuffer&nbsp;coeffBuffer,
+            java.awt.Rectangle&nbsp;bufferRegion,
+            java.awt.Rectangle&nbsp;planeRegion,
+            int&nbsp;componentID,
+            int&nbsp;transformID,
+            <a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a>&nbsp;transform)</code>
+<div class="block">A callback function that can be used to modify the DCT coefficients after
+ they are losslessly transformed but before they are transcoded to a new
+ JPEG image.</div>
+</td>
+</tr>
+</table>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="details">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="method_detail">
+
+</a>
+<h3>Method Detail</h3>
+<a name="customFilter(java.nio.ShortBuffer, java.awt.Rectangle, java.awt.Rectangle, int, int, org.libjpegturbo.turbojpeg.TJTransform)">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>customFilter</h4>
+<pre>void&nbsp;customFilter(java.nio.ShortBuffer&nbsp;coeffBuffer,
+                java.awt.Rectangle&nbsp;bufferRegion,
+                java.awt.Rectangle&nbsp;planeRegion,
+                int&nbsp;componentID,
+                int&nbsp;transformID,
+                <a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a>&nbsp;transform)
+                  throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">A callback function that can be used to modify the DCT coefficients after
+ they are losslessly transformed but before they are transcoded to a new
+ JPEG image.  This allows for custom filters or other transformations to be
+ applied in the frequency domain.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>coeffBuffer</code> - a buffer containing transformed DCT coefficients.
+ (NOTE: this buffer is not guaranteed to be valid once the callback
+ returns, so applications wishing to hand off the DCT coefficients to
+ another function or library should make a copy of them within the body of
+ the callback.)</dd><dd><code>bufferRegion</code> - rectangle containing the width and height of
+ <code>coeffBuffer</code> as well as its offset relative to the component
+ plane.  TurboJPEG implementations may choose to split each component plane
+ into multiple DCT coefficient buffers and call the callback function once
+ for each buffer.</dd><dd><code>planeRegion</code> - rectangle containing the width and height of the
+ component plane to which <code>coeffBuffer</code> belongs</dd><dd><code>componentID</code> - ID number of the component plane to which
+ <code>coeffBuffer</code> belongs (Y, Cb, and Cr have, respectively, ID's
+ of 0, 1, and 2 in typical JPEG images.)</dd><dd><code>transformID</code> - ID number of the transformed image to which
+ <code>coeffBuffer</code> belongs.  This is the same as the index of the
+ transform in the <code>transforms</code> array that was passed to <a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a>.</dd><dd><code>transform</code> - a <a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><code>TJTransform</code></a> instance that specifies the
+ parameters and/or cropping region for this transform</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJCustomFilter.html" target="_top">Frames</a></li>
+<li><a href="TJCustomFilter.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li>Constr&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li>Constr&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/TJDecompressor.html Added

@@ -0,0 +1,1255 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>TJDecompressor</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJDecompressor.html" target="_top">Frames</a></li>
+<li><a href="TJDecompressor.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li><a href="#field_summary">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li><a href="#field_detail">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+
+<div class="header">
+<div class="subTitle">org.libjpegturbo.turbojpeg</div>
+<h2 title="Class TJDecompressor" class="title">Class TJDecompressor</h2>
+</div>
+<div class="contentContainer">
+<ul class="inheritance">
+<li>java.lang.Object</li>
+<li>
+<ul class="inheritance">
+<li>org.libjpegturbo.turbojpeg.TJDecompressor</li>
+</ul>
+</li>
+</ul>
+<div class="description">
+<ul class="blockList">
+<li class="blockList">
+<dl>
+<dt>All Implemented Interfaces:</dt>
+<dd>java.io.Closeable, java.lang.AutoCloseable</dd>
+</dl>
+<dl>
+<dt>Direct Known Subclasses:</dt>
+<dd><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg">TJTransformer</a></dd>
+</dl>
+<hr>
+<br>
+<pre>public class <span class="strong">TJDecompressor</span>
+extends java.lang.Object
+implements java.io.Closeable</pre>
+<div class="block">TurboJPEG decompressor</div>
+</li>
+</ul>
+</div>
+<div class="summary">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="field_summary">
+
+</a>
+<h3>Field Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Field Summary table, listing fields, and an explanation">
+<caption><span>Fields</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Field and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected long</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#handle">handle</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>protected byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegBuf">jpegBuf</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegBufSize">jpegBufSize</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegColorspace">jpegColorspace</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegHeight">jpegHeight</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegSubsamp">jpegSubsamp</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegWidth">jpegWidth</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>protected <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#yuvImage">yuvImage</a></strong></code>&nbsp;</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_summary">
+
+</a>
+<h3>Constructor Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Constructor Summary table, listing constructors, and an explanation">
+<caption><span>Constructors</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Constructor and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#TJDecompressor()">TJDecompressor</a></strong>()</code>
+<div class="block">Create a TurboJPEG decompresssor instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#TJDecompressor(byte)">TJDecompressor</a></strong>(byte&nbsp;jpegImage)</code>
+<div class="block">Create a TurboJPEG decompressor instance and associate the JPEG source
+ image stored in <code>jpegImage</code> with the newly created instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#TJDecompressor(byte,%20int)">TJDecompressor</a></strong>(byte&nbsp;jpegImage,
+              int&nbsp;imageSize)</code>
+<div class="block">Create a TurboJPEG decompressor instance and associate the JPEG source
+ image of length <code>imageSize</code> bytes stored in
+ <code>jpegImage</code> with the newly created instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#TJDecompressor(org.libjpegturbo.turbojpeg.YUVImage)">TJDecompressor</a></strong>(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;yuvImage)</code>
+<div class="block">Create a TurboJPEG decompressor instance and associate the YUV planar
+ source image stored in <code>yuvImage</code> with the newly created
+ instance.</div>
+</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_summary">
+
+</a>
+<h3>Method Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Method Summary table, listing methods, and an explanation">
+<caption><span>Methods</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Method and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#close()">close</a></strong>()</code>
+<div class="block">Free the native structures associated with this decompressor instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(java.awt.image.BufferedImage,%20int)">decompress</a></strong>(java.awt.image.BufferedImage&nbsp;dstImage,
+          int&nbsp;flags)</code>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and output a decompressed/decoded image to
+ the given <code>BufferedImage</code> instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int)">decompress</a></strong>(byte&nbsp;dstBuf,
+          int&nbsp;desiredWidth,
+          int&nbsp;pitch,
+          int&nbsp;desiredHeight,
+          int&nbsp;pixelFormat,
+          int&nbsp;flags)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)"><code>decompress(byte, int, int, int, int, int, int, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)">decompress</a></strong>(byte&nbsp;dstBuf,
+          int&nbsp;x,
+          int&nbsp;y,
+          int&nbsp;desiredWidth,
+          int&nbsp;pitch,
+          int&nbsp;desiredHeight,
+          int&nbsp;pixelFormat,
+          int&nbsp;flags)</code>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and output a grayscale, RGB, or CMYK image
+ to the given destination buffer.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(int,%20int,%20int,%20int,%20int,%20int,%20int,%20int)">decompress</a></strong>(int&nbsp;dstBuf,
+          int&nbsp;x,
+          int&nbsp;y,
+          int&nbsp;desiredWidth,
+          int&nbsp;stride,
+          int&nbsp;desiredHeight,
+          int&nbsp;pixelFormat,
+          int&nbsp;flags)</code>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and output a grayscale, RGB, or CMYK image
+ to the given destination buffer.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>java.awt.image.BufferedImage</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(int,%20int,%20int,%20int)">decompress</a></strong>(int&nbsp;desiredWidth,
+          int&nbsp;desiredHeight,
+          int&nbsp;bufferedImageType,
+          int&nbsp;flags)</code>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and return a <code>BufferedImage</code>
+ instance containing the decompressed/decoded image.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(int,%20int,%20int,%20int,%20int)">decompress</a></strong>(int&nbsp;desiredWidth,
+          int&nbsp;pitch,
+          int&nbsp;desiredHeight,
+          int&nbsp;pixelFormat,
+          int&nbsp;flags)</code>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance and return a buffer containing the decompressed image.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(byte,%20int)">decompressToYUV</a></strong>(byte&nbsp;dstBuf,
+               int&nbsp;flags)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)"><code>decompressToYUV(YUVImage, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int)">decompressToYUV</a></strong>(int&nbsp;flags)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)"><code>decompressToYUV(int, int, int, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)">decompressToYUV</a></strong>(int&nbsp;desiredWidth,
+               int&nbsp;strides,
+               int&nbsp;desiredHeight,
+               int&nbsp;flags)</code>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance into a set of Y, U (Cb), and V (Cr) image planes and return a
+ <code>YUVImage</code> instance containing the decompressed image planes.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)">decompressToYUV</a></strong>(int&nbsp;desiredWidth,
+               int&nbsp;pad,
+               int&nbsp;desiredHeight,
+               int&nbsp;flags)</code>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance into a unified YUV planar image buffer and return a
+ <code>YUVImage</code> instance containing the decompressed image.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)">decompressToYUV</a></strong>(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;dstImage,
+               int&nbsp;flags)</code>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance into a YUV planar image and store it in the given
+ <code>YUVImage</code> instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#finalize()">finalize</a></strong>()</code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getColorspace()">getColorspace</a></strong>()</code>
+<div class="block">Returns the colorspace used in the source image (JPEG or YUV) associated
+ with this decompressor instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getHeight()">getHeight</a></strong>()</code>
+<div class="block">Returns the height of the source image (JPEG or YUV) associated with this
+ decompressor instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getJPEGBuf()">getJPEGBuf</a></strong>()</code>
+<div class="block">Returns the JPEG image buffer associated with this decompressor instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getJPEGSize()">getJPEGSize</a></strong>()</code>
+<div class="block">Returns the size of the JPEG image (in bytes) associated with this
+ decompressor instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledHeight(int,%20int)">getScaledHeight</a></strong>(int&nbsp;desiredWidth,
+               int&nbsp;desiredHeight)</code>
+<div class="block">Returns the height of the largest scaled-down image that the TurboJPEG
+ decompressor can generate without exceeding the desired image width and
+ height.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledWidth(int,%20int)">getScaledWidth</a></strong>(int&nbsp;desiredWidth,
+              int&nbsp;desiredHeight)</code>
+<div class="block">Returns the width of the largest scaled-down image that the TurboJPEG
+ decompressor can generate without exceeding the desired image width and
+ height.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getSubsamp()">getSubsamp</a></strong>()</code>
+<div class="block">Returns the level of chrominance subsampling used in the source image
+ (JPEG or YUV) associated with this decompressor instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getWidth()">getWidth</a></strong>()</code>
+<div class="block">Returns the width of the source image (JPEG or YUV) associated with this
+ decompressor instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#setJPEGImage(byte,%20int)">setJPEGImage</a></strong>(byte&nbsp;jpegImage,
+            int&nbsp;imageSize)</code>
+<div class="block"><strong>Deprecated.</strong>&nbsp;
+<div class="block"><i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(byte,%20int)"><code>setSourceImage(byte, int)</code></a> instead.</i></div>
+</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(byte,%20int)">setSourceImage</a></strong>(byte&nbsp;jpegImage,
+              int&nbsp;imageSize)</code>
+<div class="block">Associate the JPEG image of length <code>imageSize</code> bytes stored in
+ <code>jpegImage</code> with this decompressor instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(org.libjpegturbo.turbojpeg.YUVImage)">setSourceImage</a></strong>(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;srcImage)</code>
+<div class="block">Associate the specified YUV planar source image with this decompressor
+ instance.</div>
+</td>
+</tr>
+</table>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.lang.Object">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.lang.Object</h3>
+<code>clone, equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait</code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="details">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="field_detail">
+
+</a>
+<h3>Field Detail</h3>
+<a name="handle">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>handle</h4>
+<pre>protected&nbsp;long handle</pre>
+</li>
+</ul>
+<a name="jpegBuf">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>jpegBuf</h4>
+<pre>protected&nbsp;byte jpegBuf</pre>
+</li>
+</ul>
+<a name="jpegBufSize">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>jpegBufSize</h4>
+<pre>protected&nbsp;int jpegBufSize</pre>
+</li>
+</ul>
+<a name="yuvImage">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>yuvImage</h4>
+<pre>protected&nbsp;<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a> yuvImage</pre>
+</li>
+</ul>
+<a name="jpegWidth">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>jpegWidth</h4>
+<pre>protected&nbsp;int jpegWidth</pre>
+</li>
+</ul>
+<a name="jpegHeight">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>jpegHeight</h4>
+<pre>protected&nbsp;int jpegHeight</pre>
+</li>
+</ul>
+<a name="jpegSubsamp">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>jpegSubsamp</h4>
+<pre>protected&nbsp;int jpegSubsamp</pre>
+</li>
+</ul>
+<a name="jpegColorspace">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>jpegColorspace</h4>
+<pre>protected&nbsp;int jpegColorspace</pre>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_detail">
+
+</a>
+<h3>Constructor Detail</h3>
+<a name="TJDecompressor()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJDecompressor</h4>
+<pre>public&nbsp;TJDecompressor()
+               throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG decompresssor instance.</div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="TJDecompressor(byte)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJDecompressor</h4>
+<pre>public&nbsp;TJDecompressor(byte&nbsp;jpegImage)
+               throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG decompressor instance and associate the JPEG source
+ image stored in <code>jpegImage</code> with the newly created instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>jpegImage</code> - JPEG image buffer (size of the JPEG image is assumed to
+ be the length of the array.)  This buffer is not modified.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="TJDecompressor(byte, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJDecompressor</h4>
+<pre>public&nbsp;TJDecompressor(byte&nbsp;jpegImage,
+              int&nbsp;imageSize)
+               throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG decompressor instance and associate the JPEG source
+ image of length <code>imageSize</code> bytes stored in
+ <code>jpegImage</code> with the newly created instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>jpegImage</code> - JPEG image buffer.  This buffer is not modified.</dd><dd><code>imageSize</code> - size of the JPEG image (in bytes)</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="TJDecompressor(org.libjpegturbo.turbojpeg.YUVImage)">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>TJDecompressor</h4>
+<pre>public&nbsp;TJDecompressor(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;yuvImage)
+               throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG decompressor instance and associate the YUV planar
+ source image stored in <code>yuvImage</code> with the newly created
+ instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>yuvImage</code> - <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>YUVImage</code></a> instance containing a YUV planar
+ image to be decoded.  This image is not modified.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_detail">
+
+</a>
+<h3>Method Detail</h3>
+<a name="setSourceImage(byte, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setSourceImage</h4>
+<pre>public&nbsp;void&nbsp;setSourceImage(byte&nbsp;jpegImage,
+                  int&nbsp;imageSize)
+                    throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Associate the JPEG image of length <code>imageSize</code> bytes stored in
+ <code>jpegImage</code> with this decompressor instance.  This image will
+ be used as the source image for subsequent decompress operations.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>jpegImage</code> - JPEG image buffer.  This buffer is not modified.</dd><dd><code>imageSize</code> - size of the JPEG image (in bytes)</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="setJPEGImage(byte, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setJPEGImage</h4>
+<pre>@Deprecated
+public&nbsp;void&nbsp;setJPEGImage(byte&nbsp;jpegImage,
+                           int&nbsp;imageSize)
+                  throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(byte,%20int)"><code>setSourceImage(byte, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="setSourceImage(org.libjpegturbo.turbojpeg.YUVImage)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setSourceImage</h4>
+<pre>public&nbsp;void&nbsp;setSourceImage(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;srcImage)</pre>
+<div class="block">Associate the specified YUV planar source image with this decompressor
+ instance.  Subsequent decompress operations will decode this image into an
+ RGB or grayscale destination image.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>srcImage</code> - <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>YUVImage</code></a> instance containing a YUV planar image to
+ be decoded.  This image is not modified.</dd></dl>
+</li>
+</ul>
+<a name="getWidth()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getWidth</h4>
+<pre>public&nbsp;int&nbsp;getWidth()</pre>
+<div class="block">Returns the width of the source image (JPEG or YUV) associated with this
+ decompressor instance.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the width of the source image (JPEG or YUV) associated with this
+ decompressor instance.</dd></dl>
+</li>
+</ul>
+<a name="getHeight()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getHeight</h4>
+<pre>public&nbsp;int&nbsp;getHeight()</pre>
+<div class="block">Returns the height of the source image (JPEG or YUV) associated with this
+ decompressor instance.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the height of the source image (JPEG or YUV) associated with this
+ decompressor instance.</dd></dl>
+</li>
+</ul>
+<a name="getSubsamp()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getSubsamp</h4>
+<pre>public&nbsp;int&nbsp;getSubsamp()</pre>
+<div class="block">Returns the level of chrominance subsampling used in the source image
+ (JPEG or YUV) associated with this decompressor instance.  See
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444"><code>TJ.SAMP_*</code></a>.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the level of chrominance subsampling used in the source image
+ (JPEG or YUV) associated with this decompressor instance.</dd></dl>
+</li>
+</ul>
+<a name="getColorspace()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getColorspace</h4>
+<pre>public&nbsp;int&nbsp;getColorspace()</pre>
+<div class="block">Returns the colorspace used in the source image (JPEG or YUV) associated
+ with this decompressor instance.  See <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_RGB"><code>TJ.CS_*</code></a>.  If the
+ source image is YUV, then this always returns <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#CS_YCbCr"><code>TJ.CS_YCbCr</code></a>.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the colorspace used in the source image (JPEG or YUV) associated
+ with this decompressor instance.</dd></dl>
+</li>
+</ul>
+<a name="getJPEGBuf()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getJPEGBuf</h4>
+<pre>public&nbsp;byte&nbsp;getJPEGBuf()</pre>
+<div class="block">Returns the JPEG image buffer associated with this decompressor instance.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the JPEG image buffer associated with this decompressor instance.</dd></dl>
+</li>
+</ul>
+<a name="getJPEGSize()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getJPEGSize</h4>
+<pre>public&nbsp;int&nbsp;getJPEGSize()</pre>
+<div class="block">Returns the size of the JPEG image (in bytes) associated with this
+ decompressor instance.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the size of the JPEG image (in bytes) associated with this
+ decompressor instance.</dd></dl>
+</li>
+</ul>
+<a name="getScaledWidth(int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getScaledWidth</h4>
+<pre>public&nbsp;int&nbsp;getScaledWidth(int&nbsp;desiredWidth,
+                 int&nbsp;desiredHeight)</pre>
+<div class="block">Returns the width of the largest scaled-down image that the TurboJPEG
+ decompressor can generate without exceeding the desired image width and
+ height.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>desiredWidth</code> - desired width (in pixels) of the decompressed image.
+ Setting this to 0 is the same as setting it to the width of the JPEG image
+ (in other words, the width will not be considered when determining the
+ scaled image size.)</dd><dd><code>desiredHeight</code> - desired height (in pixels) of the decompressed image.
+ Setting this to 0 is the same as setting it to the height of the JPEG
+ image (in other words, the height will not be considered when determining
+ the scaled image size.)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the width of the largest scaled-down image that the TurboJPEG
+ decompressor can generate without exceeding the desired image width and
+ height.</dd></dl>
+</li>
+</ul>
+<a name="getScaledHeight(int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getScaledHeight</h4>
+<pre>public&nbsp;int&nbsp;getScaledHeight(int&nbsp;desiredWidth,
+                  int&nbsp;desiredHeight)</pre>
+<div class="block">Returns the height of the largest scaled-down image that the TurboJPEG
+ decompressor can generate without exceeding the desired image width and
+ height.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>desiredWidth</code> - desired width (in pixels) of the decompressed image.
+ Setting this to 0 is the same as setting it to the width of the JPEG image
+ (in other words, the width will not be considered when determining the
+ scaled image size.)</dd><dd><code>desiredHeight</code> - desired height (in pixels) of the decompressed image.
+ Setting this to 0 is the same as setting it to the height of the JPEG
+ image (in other words, the height will not be considered when determining
+ the scaled image size.)</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the height of the largest scaled-down image that the TurboJPEG
+ decompressor can generate without exceeding the desired image width and
+ height.</dd></dl>
+</li>
+</ul>
+<a name="decompress(byte, int, int, int, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompress</h4>
+<pre>public&nbsp;void&nbsp;decompress(byte&nbsp;dstBuf,
+              int&nbsp;x,
+              int&nbsp;y,
+              int&nbsp;desiredWidth,
+              int&nbsp;pitch,
+              int&nbsp;desiredHeight,
+              int&nbsp;pixelFormat,
+              int&nbsp;flags)
+                throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and output a grayscale, RGB, or CMYK image
+ to the given destination buffer.
+ <p>
+ NOTE: The output image is fully recoverable if this method throws a
+ non-fatal <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><code>TJException</code></a> (unless
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_STOPONWARNING"><code>TJ.FLAG_STOPONWARNING</code></a> is specified.)</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>dstBuf</code> - buffer that will receive the decompressed/decoded image.
+ If the source image is a JPEG image, then this buffer should normally be
+ <code>pitch * scaledHeight</code> bytes in size, where
+ <code>scaledHeight</code> can be determined by calling <code>
+ scalingFactor.<a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#getScaled(int)"><code>getScaled</code></a>(jpegHeight)
+ </code> with one of the scaling factors returned from <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getScalingFactors()"><code>TJ.getScalingFactors()</code></a> or by calling <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledHeight(int,%20int)"><code>getScaledHeight(int, int)</code></a>.  If the
+ source image is a YUV image, then this buffer should normally be
+ <code>pitch * height</code> bytes in size, where <code>height</code> is
+ the height of the YUV image.  However, the buffer may also be larger than
+ the dimensions of the source image, in which case the <code>x</code>,
+ <code>y</code>, and <code>pitch</code> parameters can be used to specify
+ the region into which the source image should be decompressed/decoded.</dd><dd><code>x</code> - x offset (in pixels) of the region in the destination image into
+ which the source image should be decompressed/decoded</dd><dd><code>y</code> - y offset (in pixels) of the region in the destination image into
+ which the source image should be decompressed/decoded</dd><dd><code>desiredWidth</code> - If the source image is a JPEG image, then this
+ specifies the desired width (in pixels) of the decompressed image (or
+ image region.)  If the desired destination image dimensions are different
+ than the source image dimensions, then TurboJPEG will use scaling in the
+ JPEG decompressor to generate the largest possible image that will fit
+ within the desired dimensions.  Setting this to 0 is the same as setting
+ it to the width of the JPEG image (in other words, the width will not be
+ considered when determining the scaled image size.)  This parameter is
+ ignored if the source image is a YUV image.</dd><dd><code>pitch</code> - bytes per line of the destination image.  Normally, this
+ should be set to <code>scaledWidth * TJ.pixelSize(pixelFormat)</code> if
+ the destination image is unpadded, but you can use this to, for instance,
+ pad each line of the destination image to a 4-byte boundary or to
+ decompress/decode the source image into a region of a larger image.  NOTE:
+ if the source image is a JPEG image, then <code>scaledWidth</code> can be
+ determined by calling <code>
+ scalingFactor.<a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#getScaled(int)"><code>getScaled</code></a>(jpegWidth)
+ </code> or by calling <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledWidth(int,%20int)"><code>getScaledWidth(int, int)</code></a>.  If the source image is a
+ YUV image, then <code>scaledWidth</code> is the width of the YUV image.
+ Setting this parameter to 0 is the equivalent of setting it to
+ <code>scaledWidth * TJ.pixelSize(pixelFormat)</code>.</dd><dd><code>desiredHeight</code> - If the source image is a JPEG image, then this
+ specifies the desired height (in pixels) of the decompressed image (or
+ image region.)  If the desired destination image dimensions are different
+ than the source image dimensions, then TurboJPEG will use scaling in the
+ JPEG decompressor to generate the largest possible image that will fit
+ within the desired dimensions.  Setting this to 0 is the same as setting
+ it to the height of the JPEG image (in other words, the height will not be
+ considered when determining the scaled image size.)  This parameter is
+ ignored if the source image is a YUV image.</dd><dd><code>pixelFormat</code> - pixel format of the decompressed/decoded image (one of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_RGB"><code>TJ.PF_*</code></a>)</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompress(byte, int, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompress</h4>
+<pre>@Deprecated
+public&nbsp;void&nbsp;decompress(byte&nbsp;dstBuf,
+                         int&nbsp;desiredWidth,
+                         int&nbsp;pitch,
+                         int&nbsp;desiredHeight,
+                         int&nbsp;pixelFormat,
+                         int&nbsp;flags)
+                throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use
+ <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)"><code>decompress(byte, int, int, int, int, int, int, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompress(int, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompress</h4>
+<pre>public&nbsp;byte&nbsp;decompress(int&nbsp;desiredWidth,
+                int&nbsp;pitch,
+                int&nbsp;desiredHeight,
+                int&nbsp;pixelFormat,
+                int&nbsp;flags)
+                  throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance and return a buffer containing the decompressed image.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>desiredWidth</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)"><code>decompress(byte, int, int, int, int, int, int, int)</code></a>
+ for description</dd><dd><code>pitch</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)"><code>decompress(byte, int, int, int, int, int, int, int)</code></a>
+ for description</dd><dd><code>desiredHeight</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)"><code>decompress(byte, int, int, int, int, int, int, int)</code></a>
+ for description</dd><dd><code>pixelFormat</code> - pixel format of the decompressed image (one of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_RGB"><code>TJ.PF_*</code></a>)</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Returns:</span></dt><dd>a buffer containing the decompressed image.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompressToYUV(org.libjpegturbo.turbojpeg.YUVImage, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompressToYUV</h4>
+<pre>public&nbsp;void&nbsp;decompressToYUV(<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;dstImage,
+                   int&nbsp;flags)
+                     throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance into a YUV planar image and store it in the given
+ <code>YUVImage</code> instance.  This method performs JPEG decompression
+ but leaves out the color conversion step, so a planar YUV image is
+ generated instead of an RGB or grayscale image.  This method cannot be
+ used to decompress JPEG source images with the CMYK or YCCK colorspace.
+ <p>
+ NOTE: The YUV planar output image is fully recoverable if this method
+ throws a non-fatal <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><code>TJException</code></a> (unless
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_STOPONWARNING"><code>TJ.FLAG_STOPONWARNING</code></a> is specified.)</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>dstImage</code> - <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>YUVImage</code></a> instance that will receive the YUV planar
+ image.  The level of subsampling specified in this <code>YUVImage</code>
+ instance must match that of the JPEG image, and the width and height
+ specified in the <code>YUVImage</code> instance must match one of the
+ scaled image sizes that TurboJPEG is capable of generating from the JPEG
+ source image.</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompressToYUV(byte, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompressToYUV</h4>
+<pre>@Deprecated
+public&nbsp;void&nbsp;decompressToYUV(byte&nbsp;dstBuf,
+                              int&nbsp;flags)
+                     throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)"><code>decompressToYUV(YUVImage, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompressToYUV(int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompressToYUV</h4>
+<pre>public&nbsp;<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;decompressToYUV(int&nbsp;desiredWidth,
+                       int&nbsp;strides,
+                       int&nbsp;desiredHeight,
+                       int&nbsp;flags)
+                         throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance into a set of Y, U (Cb), and V (Cr) image planes and return a
+ <code>YUVImage</code> instance containing the decompressed image planes.
+ This method performs JPEG decompression but leaves out the color
+ conversion step, so a planar YUV image is generated instead of an RGB or
+ grayscale image.  This method cannot be used to decompress JPEG source
+ images with the CMYK or YCCK colorspace.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>desiredWidth</code> - desired width (in pixels) of the YUV image.  If the
+ desired image dimensions are different than the dimensions of the JPEG
+ image being decompressed, then TurboJPEG will use scaling in the JPEG
+ decompressor to generate the largest possible image that will fit within
+ the desired dimensions.  Setting this to 0 is the same as setting it to
+ the width of the JPEG image (in other words, the width will not be
+ considered when determining the scaled image size.)</dd><dd><code>strides</code> - an array of integers, each specifying the number of bytes
+ per line in the corresponding plane of the output image.  Setting the
+ stride for any plane to 0 is the same as setting it to the scaled
+ component width of the plane.  If <tt>strides</tt> is NULL, then the
+ strides for all planes will be set to their respective scaled component
+ widths.  You can adjust the strides in order to add an arbitrary amount of
+ line padding to each plane.</dd><dd><code>desiredHeight</code> - desired height (in pixels) of the YUV image.  If the
+ desired image dimensions are different than the dimensions of the JPEG
+ image being decompressed, then TurboJPEG will use scaling in the JPEG
+ decompressor to generate the largest possible image that will fit within
+ the desired dimensions.  Setting this to 0 is the same as setting it to
+ the height of the JPEG image (in other words, the height will not be
+ considered when determining the scaled image size.)</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Returns:</span></dt><dd>a YUV planar image.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompressToYUV(int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompressToYUV</h4>
+<pre>public&nbsp;<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a>&nbsp;decompressToYUV(int&nbsp;desiredWidth,
+                       int&nbsp;pad,
+                       int&nbsp;desiredHeight,
+                       int&nbsp;flags)
+                         throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Decompress the JPEG source image associated with this decompressor
+ instance into a unified YUV planar image buffer and return a
+ <code>YUVImage</code> instance containing the decompressed image.  This
+ method performs JPEG decompression but leaves out the color conversion
+ step, so a planar YUV image is generated instead of an RGB or grayscale
+ image.  This method cannot be used to decompress JPEG source images with
+ the CMYK or YCCK colorspace.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>desiredWidth</code> - desired width (in pixels) of the YUV image.  If the
+ desired image dimensions are different than the dimensions of the JPEG
+ image being decompressed, then TurboJPEG will use scaling in the JPEG
+ decompressor to generate the largest possible image that will fit within
+ the desired dimensions.  Setting this to 0 is the same as setting it to
+ the width of the JPEG image (in other words, the width will not be
+ considered when determining the scaled image size.)</dd><dd><code>pad</code> - the width of each line in each plane of the YUV image will be
+ padded to the nearest multiple of this number of bytes (must be a power of
+ 2.)</dd><dd><code>desiredHeight</code> - desired height (in pixels) of the YUV image.  If the
+ desired image dimensions are different than the dimensions of the JPEG
+ image being decompressed, then TurboJPEG will use scaling in the JPEG
+ decompressor to generate the largest possible image that will fit within
+ the desired dimensions.  Setting this to 0 is the same as setting it to
+ the height of the JPEG image (in other words, the height will not be
+ considered when determining the scaled image size.)</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Returns:</span></dt><dd>a YUV planar image.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompressToYUV(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompressToYUV</h4>
+<pre>@Deprecated
+public&nbsp;byte&nbsp;decompressToYUV(int&nbsp;flags)
+                       throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block"><span class="strong">Deprecated.</span>&nbsp;<i>Use <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)"><code>decompressToYUV(int, int, int, int)</code></a> instead.</i></div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompress(int, int, int, int, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompress</h4>
+<pre>public&nbsp;void&nbsp;decompress(int&nbsp;dstBuf,
+              int&nbsp;x,
+              int&nbsp;y,
+              int&nbsp;desiredWidth,
+              int&nbsp;stride,
+              int&nbsp;desiredHeight,
+              int&nbsp;pixelFormat,
+              int&nbsp;flags)
+                throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and output a grayscale, RGB, or CMYK image
+ to the given destination buffer.
+ <p>
+ NOTE: The output image is fully recoverable if this method throws a
+ non-fatal <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><code>TJException</code></a> (unless
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_STOPONWARNING"><code>TJ.FLAG_STOPONWARNING</code></a> is specified.)</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>dstBuf</code> - buffer that will receive the decompressed/decoded image.
+ If the source image is a JPEG image, then this buffer should normally be
+ <code>stride * scaledHeight</code> pixels in size, where
+ <code>scaledHeight</code> can be determined by calling <code>
+ scalingFactor.<a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#getScaled(int)"><code>getScaled</code></a>(jpegHeight)
+ </code> with one of the scaling factors returned from <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getScalingFactors()"><code>TJ.getScalingFactors()</code></a> or by calling <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledHeight(int,%20int)"><code>getScaledHeight(int, int)</code></a>.  If the
+ source image is a YUV image, then this buffer should normally be
+ <code>stride * height</code> pixels in size, where <code>height</code> is
+ the height of the YUV image.  However, the buffer may also be larger than
+ the dimensions of the JPEG image, in which case the <code>x</code>,
+ <code>y</code>, and <code>stride</code> parameters can be used to specify
+ the region into which the source image should be decompressed.</dd><dd><code>x</code> - x offset (in pixels) of the region in the destination image into
+ which the source image should be decompressed/decoded</dd><dd><code>y</code> - y offset (in pixels) of the region in the destination image into
+ which the source image should be decompressed/decoded</dd><dd><code>desiredWidth</code> - If the source image is a JPEG image, then this
+ specifies the desired width (in pixels) of the decompressed image (or
+ image region.)  If the desired destination image dimensions are different
+ than the source image dimensions, then TurboJPEG will use scaling in the
+ JPEG decompressor to generate the largest possible image that will fit
+ within the desired dimensions.  Setting this to 0 is the same as setting
+ it to the width of the JPEG image (in other words, the width will not be
+ considered when determining the scaled image size.)  This parameter is
+ ignored if the source image is a YUV image.</dd><dd><code>stride</code> - pixels per line of the destination image.  Normally, this
+ should be set to <code>scaledWidth</code>, but you can use this to, for
+ instance, decompress the JPEG image into a region of a larger image.
+ NOTE: if the source image is a JPEG image, then <code>scaledWidth</code>
+ can be determined by calling <code>
+ scalingFactor.<a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#getScaled(int)"><code>getScaled</code></a>(jpegWidth)
+ </code> or by calling <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledWidth(int,%20int)"><code>getScaledWidth(int, int)</code></a>.  If the source image is a
+ YUV image, then <code>scaledWidth</code> is the width of the YUV image.
+ Setting this parameter to 0 is the equivalent of setting it to
+ <code>scaledWidth</code>.</dd><dd><code>desiredHeight</code> - If the source image is a JPEG image, then this
+ specifies the desired height (in pixels) of the decompressed image (or
+ image region.)  If the desired destination image dimensions are different
+ than the source image dimensions, then TurboJPEG will use scaling in the
+ JPEG decompressor to generate the largest possible image that will fit
+ within the desired dimensions.  Setting this to 0 is the same as setting
+ it to the height of the JPEG image (in other words, the height will not be
+ considered when determining the scaled image size.)  This parameter is
+ ignored if the source image is a YUV image.</dd><dd><code>pixelFormat</code> - pixel format of the decompressed image (one of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#PF_RGB"><code>TJ.PF_*</code></a>)</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompress(java.awt.image.BufferedImage, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompress</h4>
+<pre>public&nbsp;void&nbsp;decompress(java.awt.image.BufferedImage&nbsp;dstImage,
+              int&nbsp;flags)
+                throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and output a decompressed/decoded image to
+ the given <code>BufferedImage</code> instance.
+ <p>
+ NOTE: The output image is fully recoverable if this method throws a
+ non-fatal <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><code>TJException</code></a> (unless
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_STOPONWARNING"><code>TJ.FLAG_STOPONWARNING</code></a> is specified.)</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>dstImage</code> - a <code>BufferedImage</code> instance that will receive
+ the decompressed/decoded image.  If the source image is a JPEG image, then
+ the width and height of the <code>BufferedImage</code> instance must match
+ one of the scaled image sizes that TurboJPEG is capable of generating from
+ the JPEG image.  If the source image is a YUV image, then the width and
+ height of the <code>BufferedImage</code> instance must match the width and
+ height of the YUV image.</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="decompress(int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>decompress</h4>
+<pre>public&nbsp;java.awt.image.BufferedImage&nbsp;decompress(int&nbsp;desiredWidth,
+                                      int&nbsp;desiredHeight,
+                                      int&nbsp;bufferedImageType,
+                                      int&nbsp;flags)
+                                        throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Decompress the JPEG source image or decode the YUV source image associated
+ with this decompressor instance and return a <code>BufferedImage</code>
+ instance containing the decompressed/decoded image.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>desiredWidth</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)"><code>decompress(byte, int, int, int, int, int, int, int)</code></a> for
+ description</dd><dd><code>desiredHeight</code> - see
+ <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)"><code>decompress(byte, int, int, int, int, int, int, int)</code></a> for
+ description</dd><dd><code>bufferedImageType</code> - the image type of the <code>BufferedImage</code>
+ instance that will be created (for instance,
+ <code>BufferedImage.TYPE_INT_RGB</code>)</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Returns:</span></dt><dd>a <code>BufferedImage</code> instance containing the
+ decompressed/decoded image.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="close()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>close</h4>
+<pre>public&nbsp;void&nbsp;close()
+           throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Free the native structures associated with this decompressor instance.</div>
+<dl>
+<dt><strong>Specified by:</strong></dt>
+<dd><code>close</code>&nbsp;in interface&nbsp;<code>java.io.Closeable</code></dd>
+<dt><strong>Specified by:</strong></dt>
+<dd><code>close</code>&nbsp;in interface&nbsp;<code>java.lang.AutoCloseable</code></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="finalize()">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>finalize</h4>
+<pre>protected&nbsp;void&nbsp;finalize()
+                 throws java.lang.Throwable</pre>
+<dl>
+<dt><strong>Overrides:</strong></dt>
+<dd><code>finalize</code>&nbsp;in class&nbsp;<code>java.lang.Object</code></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code>java.lang.Throwable</code></dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJDecompressor.html" target="_top">Frames</a></li>
+<li><a href="TJDecompressor.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li><a href="#field_summary">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li><a href="#field_detail">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/TJException.html Added

@@ -0,0 +1,340 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>TJException</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJException.html" target="_top">Frames</a></li>
+<li><a href="TJException.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+
+<div class="header">
+<div class="subTitle">org.libjpegturbo.turbojpeg</div>
+<h2 title="Class TJException" class="title">Class TJException</h2>
+</div>
+<div class="contentContainer">
+<ul class="inheritance">
+<li>java.lang.Object</li>
+<li>
+<ul class="inheritance">
+<li>java.lang.Throwable</li>
+<li>
+<ul class="inheritance">
+<li>java.lang.Exception</li>
+<li>
+<ul class="inheritance">
+<li>java.io.IOException</li>
+<li>
+<ul class="inheritance">
+<li>org.libjpegturbo.turbojpeg.TJException</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+<div class="description">
+<ul class="blockList">
+<li class="blockList">
+<dl>
+<dt>All Implemented Interfaces:</dt>
+<dd>java.io.Serializable</dd>
+</dl>
+<hr>
+<br>
+<pre>public class <span class="strong">TJException</span>
+extends java.io.IOException</pre>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../serialized-form.html#org.libjpegturbo.turbojpeg.TJException">Serialized Form</a></dd></dl>
+</li>
+</ul>
+</div>
+<div class="summary">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_summary">
+
+</a>
+<h3>Constructor Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Constructor Summary table, listing constructors, and an explanation">
+<caption><span>Constructors</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Constructor and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJException.html#TJException()">TJException</a></strong>()</code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJException.html#TJException(java.lang.String)">TJException</a></strong>(java.lang.String&nbsp;message)</code>&nbsp;</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJException.html#TJException(java.lang.String,%20int)">TJException</a></strong>(java.lang.String&nbsp;message,
+           int&nbsp;code)</code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJException.html#TJException(java.lang.String,%20java.lang.Throwable)">TJException</a></strong>(java.lang.String&nbsp;message,
+           java.lang.Throwable&nbsp;cause)</code>&nbsp;</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJException.html#TJException(java.lang.Throwable)">TJException</a></strong>(java.lang.Throwable&nbsp;cause)</code>&nbsp;</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_summary">
+
+</a>
+<h3>Method Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Method Summary table, listing methods, and an explanation">
+<caption><span>Methods</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Method and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJException.html#getErrorCode()">getErrorCode</a></strong>()</code>
+<div class="block">Returns a code (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><code>TJ.ERR_*</code></a>) indicating the severity of the
+ last error.</div>
+</td>
+</tr>
+</table>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.lang.Throwable">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.lang.Throwable</h3>
+<code>addSuppressed, fillInStackTrace, getCause, getLocalizedMessage, getMessage, getStackTrace, getSuppressed, initCause, printStackTrace, printStackTrace, printStackTrace, setStackTrace, toString</code></li>
+</ul>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.lang.Object">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.lang.Object</h3>
+<code>clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait</code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="details">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_detail">
+
+</a>
+<h3>Constructor Detail</h3>
+<a name="TJException()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJException</h4>
+<pre>public&nbsp;TJException()</pre>
+</li>
+</ul>
+<a name="TJException(java.lang.String, java.lang.Throwable)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJException</h4>
+<pre>public&nbsp;TJException(java.lang.String&nbsp;message,
+           java.lang.Throwable&nbsp;cause)</pre>
+</li>
+</ul>
+<a name="TJException(java.lang.String)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJException</h4>
+<pre>public&nbsp;TJException(java.lang.String&nbsp;message)</pre>
+</li>
+</ul>
+<a name="TJException(java.lang.String, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJException</h4>
+<pre>public&nbsp;TJException(java.lang.String&nbsp;message,
+           int&nbsp;code)</pre>
+</li>
+</ul>
+<a name="TJException(java.lang.Throwable)">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>TJException</h4>
+<pre>public&nbsp;TJException(java.lang.Throwable&nbsp;cause)</pre>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_detail">
+
+</a>
+<h3>Method Detail</h3>
+<a name="getErrorCode()">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>getErrorCode</h4>
+<pre>public&nbsp;int&nbsp;getErrorCode()</pre>
+<div class="block">Returns a code (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><code>TJ.ERR_*</code></a>) indicating the severity of the
+ last error.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>a code (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><code>TJ.ERR_*</code></a>) indicating the severity of the
+ last error.</dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJException.html" target="_top">Frames</a></li>
+<li><a href="TJException.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/TJScalingFactor.html Added

@@ -0,0 +1,343 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>TJScalingFactor</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJScalingFactor.html" target="_top">Frames</a></li>
+<li><a href="TJScalingFactor.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+
+<div class="header">
+<div class="subTitle">org.libjpegturbo.turbojpeg</div>
+<h2 title="Class TJScalingFactor" class="title">Class TJScalingFactor</h2>
+</div>
+<div class="contentContainer">
+<ul class="inheritance">
+<li>java.lang.Object</li>
+<li>
+<ul class="inheritance">
+<li>org.libjpegturbo.turbojpeg.TJScalingFactor</li>
+</ul>
+</li>
+</ul>
+<div class="description">
+<ul class="blockList">
+<li class="blockList">
+<hr>
+<br>
+<pre>public class <span class="strong">TJScalingFactor</span>
+extends java.lang.Object</pre>
+<div class="block">Fractional scaling factor</div>
+</li>
+</ul>
+</div>
+<div class="summary">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_summary">
+
+</a>
+<h3>Constructor Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Constructor Summary table, listing constructors, and an explanation">
+<caption><span>Constructors</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Constructor and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#TJScalingFactor(int,%20int)">TJScalingFactor</a></strong>(int&nbsp;num,
+               int&nbsp;denom)</code>
+<div class="block">Create a TurboJPEG scaling factor instance.</div>
+</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_summary">
+
+</a>
+<h3>Method Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Method Summary table, listing methods, and an explanation">
+<caption><span>Methods</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Method and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>boolean</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#equals(org.libjpegturbo.turbojpeg.TJScalingFactor)">equals</a></strong>(<a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a>&nbsp;other)</code>
+<div class="block">Returns true or false, depending on whether this instance and
+ <code>other</code> have the same numerator and denominator.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#getDenom()">getDenom</a></strong>()</code>
+<div class="block">Returns denominator</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#getNum()">getNum</a></strong>()</code>
+<div class="block">Returns numerator</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#getScaled(int)">getScaled</a></strong>(int&nbsp;dimension)</code>
+<div class="block">Returns the scaled value of <code>dimension</code>.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>boolean</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html#isOne()">isOne</a></strong>()</code>
+<div class="block">Returns true or false, depending on whether this instance is equal to
+ 1/1.</div>
+</td>
+</tr>
+</table>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.lang.Object">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.lang.Object</h3>
+<code>clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait</code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="details">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_detail">
+
+</a>
+<h3>Constructor Detail</h3>
+<a name="TJScalingFactor(int, int)">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>TJScalingFactor</h4>
+<pre>public&nbsp;TJScalingFactor(int&nbsp;num,
+               int&nbsp;denom)</pre>
+<div class="block">Create a TurboJPEG scaling factor instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>num</code> - numerator</dd><dd><code>denom</code> - denominator</dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_detail">
+
+</a>
+<h3>Method Detail</h3>
+<a name="getNum()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getNum</h4>
+<pre>public&nbsp;int&nbsp;getNum()</pre>
+<div class="block">Returns numerator</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>numerator</dd></dl>
+</li>
+</ul>
+<a name="getDenom()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getDenom</h4>
+<pre>public&nbsp;int&nbsp;getDenom()</pre>
+<div class="block">Returns denominator</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>denominator</dd></dl>
+</li>
+</ul>
+<a name="getScaled(int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getScaled</h4>
+<pre>public&nbsp;int&nbsp;getScaled(int&nbsp;dimension)</pre>
+<div class="block">Returns the scaled value of <code>dimension</code>.  This function
+ performs the integer equivalent of
+ <code>ceil(dimension * scalingFactor)</code>.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>dimension</code> - width or height to multiply by this scaling factor</dd>
+<dt><span class="strong">Returns:</span></dt><dd>the scaled value of <code>dimension</code>.</dd></dl>
+</li>
+</ul>
+<a name="equals(org.libjpegturbo.turbojpeg.TJScalingFactor)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>equals</h4>
+<pre>public&nbsp;boolean&nbsp;equals(<a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a>&nbsp;other)</pre>
+<div class="block">Returns true or false, depending on whether this instance and
+ <code>other</code> have the same numerator and denominator.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>other</code> - the scaling factor against which to compare this one</dd>
+<dt><span class="strong">Returns:</span></dt><dd>true or false, depending on whether this instance and
+ <code>other</code> have the same numerator and denominator.</dd></dl>
+</li>
+</ul>
+<a name="isOne()">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>isOne</h4>
+<pre>public&nbsp;boolean&nbsp;isOne()</pre>
+<div class="block">Returns true or false, depending on whether this instance is equal to
+ 1/1.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>true or false, depending on whether this instance is equal to
+ 1/1.</dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJScalingFactor.html" target="_top">Frames</a></li>
+<li><a href="TJScalingFactor.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/TJTransform.html Added

@@ -0,0 +1,751 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>TJTransform</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJTransform.html" target="_top">Frames</a></li>
+<li><a href="TJTransform.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li><a href="#nested_classes_inherited_from_class_java.awt.geom.Rectangle2D">Nested</a>&nbsp;|&nbsp;</li>
+<li><a href="#field_summary">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#methods_inherited_from_class_java.awt.Rectangle">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li><a href="#field_detail">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li>Method</li>
+</ul>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+
+<div class="header">
+<div class="subTitle">org.libjpegturbo.turbojpeg</div>
+<h2 title="Class TJTransform" class="title">Class TJTransform</h2>
+</div>
+<div class="contentContainer">
+<ul class="inheritance">
+<li>java.lang.Object</li>
+<li>
+<ul class="inheritance">
+<li>java.awt.geom.RectangularShape</li>
+<li>
+<ul class="inheritance">
+<li>java.awt.geom.Rectangle2D</li>
+<li>
+<ul class="inheritance">
+<li>java.awt.Rectangle</li>
+<li>
+<ul class="inheritance">
+<li>org.libjpegturbo.turbojpeg.TJTransform</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+<div class="description">
+<ul class="blockList">
+<li class="blockList">
+<dl>
+<dt>All Implemented Interfaces:</dt>
+<dd>java.awt.Shape, java.io.Serializable, java.lang.Cloneable</dd>
+</dl>
+<hr>
+<br>
+<pre>public class <span class="strong">TJTransform</span>
+extends java.awt.Rectangle</pre>
+<div class="block">Lossless transform parameters</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../serialized-form.html#org.libjpegturbo.turbojpeg.TJTransform">Serialized Form</a></dd></dl>
+</li>
+</ul>
+</div>
+<div class="summary">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="nested_class_summary">
+
+</a>
+<h3>Nested Class Summary</h3>
+<ul class="blockList">
+<li class="blockList"><a name="nested_classes_inherited_from_class_java.awt.geom.Rectangle2D">
+
+</a>
+<h3>Nested classes/interfaces inherited from class&nbsp;java.awt.geom.Rectangle2D</h3>
+<code>java.awt.geom.Rectangle2D.Double, java.awt.geom.Rectangle2D.Float</code></li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="field_summary">
+
+</a>
+<h3>Field Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Field Summary table, listing fields, and an explanation">
+<caption><span>Fields</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Field and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code><a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg">TJCustomFilter</a></code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#cf">cf</a></strong></code>
+<div class="block">Custom filter instance</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#NUMOP">NUMOP</a></strong></code>
+<div class="block">The number of lossless transform operations</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#op">op</a></strong></code>
+<div class="block">Transform operation (one of <code>OP_*</code>)</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OP_HFLIP">OP_HFLIP</a></strong></code>
+<div class="block">Flip (mirror) image horizontally.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OP_NONE">OP_NONE</a></strong></code>
+<div class="block">Do not transform the position of the image pixels.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OP_ROT180">OP_ROT180</a></strong></code>
+<div class="block">Rotate image 180 degrees.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OP_ROT270">OP_ROT270</a></strong></code>
+<div class="block">Rotate image counter-clockwise by 90 degrees.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OP_ROT90">OP_ROT90</a></strong></code>
+<div class="block">Rotate image clockwise by 90 degrees.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OP_TRANSPOSE">OP_TRANSPOSE</a></strong></code>
+<div class="block">Transpose image (flip/mirror along upper left to lower right axis).</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OP_TRANSVERSE">OP_TRANSVERSE</a></strong></code>
+<div class="block">Transverse transpose image (flip/mirror along upper right to lower left
+ axis).</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OP_VFLIP">OP_VFLIP</a></strong></code>
+<div class="block">Flip (mirror) image vertically.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_COPYNONE">OPT_COPYNONE</a></strong></code>
+<div class="block">This option will prevent <a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a> from copying any extra markers (including EXIF
+ and ICC profile data) from the source image to the output image.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_CROP">OPT_CROP</a></strong></code>
+<div class="block">This option will enable lossless cropping.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_GRAY">OPT_GRAY</a></strong></code>
+<div class="block">This option will discard the color data in the input image and produce
+ a grayscale output image.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_NOOUTPUT">OPT_NOOUTPUT</a></strong></code>
+<div class="block">This option will prevent <a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a> from outputting a JPEG image for this
+ particular transform.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT">OPT_PERFECT</a></strong></code>
+<div class="block">This option will cause <a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a> to throw an exception if the transform is not
+ perfect.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PROGRESSIVE">OPT_PROGRESSIVE</a></strong></code>
+<div class="block">This option will enable progressive entropy coding in the output image
+ generated by this particular transform.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>static int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_TRIM">OPT_TRIM</a></strong></code>
+<div class="block">This option will discard any partial MCU blocks that cannot be
+ transformed.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#options">options</a></strong></code>
+<div class="block">Transform options (bitwise OR of one or more of <code>OPT_*</code>)</div>
+</td>
+</tr>
+</table>
+<ul class="blockList">
+<li class="blockList"><a name="fields_inherited_from_class_java.awt.Rectangle">
+
+</a>
+<h3>Fields inherited from class&nbsp;java.awt.Rectangle</h3>
+<code>height, width, x, y</code></li>
+</ul>
+<ul class="blockList">
+<li class="blockList"><a name="fields_inherited_from_class_java.awt.geom.Rectangle2D">
+
+</a>
+<h3>Fields inherited from class&nbsp;java.awt.geom.Rectangle2D</h3>
+<code>OUT_BOTTOM, OUT_LEFT, OUT_RIGHT, OUT_TOP</code></li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_summary">
+
+</a>
+<h3>Constructor Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Constructor Summary table, listing constructors, and an explanation">
+<caption><span>Constructors</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Constructor and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#TJTransform()">TJTransform</a></strong>()</code>
+<div class="block">Create a new lossless transform instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#TJTransform(int,%20int,%20int,%20int,%20int,%20int,%20org.libjpegturbo.turbojpeg.TJCustomFilter)">TJTransform</a></strong>(int&nbsp;x,
+           int&nbsp;y,
+           int&nbsp;w,
+           int&nbsp;h,
+           int&nbsp;op,
+           int&nbsp;options,
+           <a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg">TJCustomFilter</a>&nbsp;cf)</code>
+<div class="block">Create a new lossless transform instance with the given parameters.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#TJTransform(java.awt.Rectangle,%20int,%20int,%20org.libjpegturbo.turbojpeg.TJCustomFilter)">TJTransform</a></strong>(java.awt.Rectangle&nbsp;r,
+           int&nbsp;op,
+           int&nbsp;options,
+           <a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg">TJCustomFilter</a>&nbsp;cf)</code>
+<div class="block">Create a new lossless transform instance with the given parameters.</div>
+</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_summary">
+
+</a>
+<h3>Method Summary</h3>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.awt.Rectangle">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.awt.Rectangle</h3>
+<code>add, add, add, contains, contains, contains, contains, createIntersection, createUnion, equals, getBounds, getBounds2D, getHeight, getLocation, getSize, getWidth, getX, getY, grow, inside, intersection, intersects, isEmpty, move, outcode, reshape, resize, setBounds, setBounds, setLocation, setLocation, setRect, setSize, setSize, toString, translate, union</code></li>
+</ul>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.awt.geom.Rectangle2D">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.awt.geom.Rectangle2D</h3>
+<code>add, add, add, contains, contains, getPathIterator, getPathIterator, hashCode, intersect, intersects, intersectsLine, intersectsLine, outcode, setFrame, setRect, union</code></li>
+</ul>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.awt.geom.RectangularShape">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.awt.geom.RectangularShape</h3>
+<code>clone, contains, contains, getCenterX, getCenterY, getFrame, getMaxX, getMaxY, getMinX, getMinY, intersects, setFrame, setFrame, setFrameFromCenter, setFrameFromCenter, setFrameFromDiagonal, setFrameFromDiagonal</code></li>
+</ul>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.lang.Object">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.lang.Object</h3>
+<code>finalize, getClass, notify, notifyAll, wait, wait, wait</code></li>
+</ul>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.awt.Shape">
+
+</a>
+<h3>Methods inherited from interface&nbsp;java.awt.Shape</h3>
+<code>contains, contains, contains, contains, getPathIterator, getPathIterator, intersects, intersects</code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="details">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="field_detail">
+
+</a>
+<h3>Field Detail</h3>
+<a name="NUMOP">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>NUMOP</h4>
+<pre>public static final&nbsp;int NUMOP</pre>
+<div class="block">The number of lossless transform operations</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.NUMOP">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OP_NONE">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OP_NONE</h4>
+<pre>public static final&nbsp;int OP_NONE</pre>
+<div class="block">Do not transform the position of the image pixels.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OP_NONE">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OP_HFLIP">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OP_HFLIP</h4>
+<pre>public static final&nbsp;int OP_HFLIP</pre>
+<div class="block">Flip (mirror) image horizontally.  This transform is imperfect if there
+ are any partial MCU blocks on the right edge.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT"><code>OPT_PERFECT</code></a>, 
+<a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OP_HFLIP">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OP_VFLIP">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OP_VFLIP</h4>
+<pre>public static final&nbsp;int OP_VFLIP</pre>
+<div class="block">Flip (mirror) image vertically.  This transform is imperfect if there are
+ any partial MCU blocks on the bottom edge.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT"><code>OPT_PERFECT</code></a>, 
+<a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OP_VFLIP">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OP_TRANSPOSE">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OP_TRANSPOSE</h4>
+<pre>public static final&nbsp;int OP_TRANSPOSE</pre>
+<div class="block">Transpose image (flip/mirror along upper left to lower right axis).  This
+ transform is always perfect.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT"><code>OPT_PERFECT</code></a>, 
+<a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OP_TRANSPOSE">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OP_TRANSVERSE">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OP_TRANSVERSE</h4>
+<pre>public static final&nbsp;int OP_TRANSVERSE</pre>
+<div class="block">Transverse transpose image (flip/mirror along upper right to lower left
+ axis).  This transform is imperfect if there are any partial MCU blocks in
+ the image.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT"><code>OPT_PERFECT</code></a>, 
+<a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OP_TRANSVERSE">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OP_ROT90">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OP_ROT90</h4>
+<pre>public static final&nbsp;int OP_ROT90</pre>
+<div class="block">Rotate image clockwise by 90 degrees.  This transform is imperfect if
+ there are any partial MCU blocks on the bottom edge.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT"><code>OPT_PERFECT</code></a>, 
+<a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OP_ROT90">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OP_ROT180">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OP_ROT180</h4>
+<pre>public static final&nbsp;int OP_ROT180</pre>
+<div class="block">Rotate image 180 degrees.  This transform is imperfect if there are any
+ partial MCU blocks in the image.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT"><code>OPT_PERFECT</code></a>, 
+<a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OP_ROT180">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OP_ROT270">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OP_ROT270</h4>
+<pre>public static final&nbsp;int OP_ROT270</pre>
+<div class="block">Rotate image counter-clockwise by 90 degrees.  This transform is imperfect
+ if there are any partial MCU blocks on the right edge.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#OPT_PERFECT"><code>OPT_PERFECT</code></a>, 
+<a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OP_ROT270">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OPT_PERFECT">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OPT_PERFECT</h4>
+<pre>public static final&nbsp;int OPT_PERFECT</pre>
+<div class="block">This option will cause <a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a> to throw an exception if the transform is not
+ perfect.  Lossless transforms operate on MCU blocks, whose size depends on
+ the level of chrominance subsampling used.  If the image's width or height
+ is not evenly divisible by the MCU block size (see <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getMCUWidth(int)"><code>TJ.getMCUWidth(int)</code></a>
+ and <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getMCUHeight(int)"><code>TJ.getMCUHeight(int)</code></a>), then there will be partial MCU blocks on the
+ right and/or bottom edges.   It is not possible to move these partial MCU
+ blocks to the top or left of the image, so any transform that would
+ require that is "imperfect."  If this option is not specified, then any
+ partial MCU blocks that cannot be transformed will be left in place, which
+ will create odd-looking strips on the right or bottom edge of the image.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OPT_PERFECT">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OPT_TRIM">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OPT_TRIM</h4>
+<pre>public static final&nbsp;int OPT_TRIM</pre>
+<div class="block">This option will discard any partial MCU blocks that cannot be
+ transformed.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OPT_TRIM">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OPT_CROP">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OPT_CROP</h4>
+<pre>public static final&nbsp;int OPT_CROP</pre>
+<div class="block">This option will enable lossless cropping.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OPT_CROP">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OPT_GRAY">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OPT_GRAY</h4>
+<pre>public static final&nbsp;int OPT_GRAY</pre>
+<div class="block">This option will discard the color data in the input image and produce
+ a grayscale output image.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OPT_GRAY">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OPT_NOOUTPUT">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OPT_NOOUTPUT</h4>
+<pre>public static final&nbsp;int OPT_NOOUTPUT</pre>
+<div class="block">This option will prevent <a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a> from outputting a JPEG image for this
+ particular transform.  This can be used in conjunction with a custom
+ filter to capture the transformed DCT coefficients without transcoding
+ them.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OPT_NOOUTPUT">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OPT_PROGRESSIVE">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OPT_PROGRESSIVE</h4>
+<pre>public static final&nbsp;int OPT_PROGRESSIVE</pre>
+<div class="block">This option will enable progressive entropy coding in the output image
+ generated by this particular transform.  Progressive entropy coding will
+ generally improve compression relative to baseline entropy coding (the
+ default), but it will reduce compression and decompression performance
+ considerably.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OPT_PROGRESSIVE">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="OPT_COPYNONE">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>OPT_COPYNONE</h4>
+<pre>public static final&nbsp;int OPT_COPYNONE</pre>
+<div class="block">This option will prevent <a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)"><code>TJTransformer.transform()</code></a> from copying any extra markers (including EXIF
+ and ICC profile data) from the source image to the output image.</div>
+<dl><dt><span class="strong">See Also:</span></dt><dd><a href="../../../constant-values.html#org.libjpegturbo.turbojpeg.TJTransform.OPT_COPYNONE">Constant Field Values</a></dd></dl>
+</li>
+</ul>
+<a name="op">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>op</h4>
+<pre>public&nbsp;int op</pre>
+<div class="block">Transform operation (one of <code>OP_*</code>)</div>
+</li>
+</ul>
+<a name="options">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>options</h4>
+<pre>public&nbsp;int options</pre>
+<div class="block">Transform options (bitwise OR of one or more of <code>OPT_*</code>)</div>
+</li>
+</ul>
+<a name="cf">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>cf</h4>
+<pre>public&nbsp;<a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg">TJCustomFilter</a> cf</pre>
+<div class="block">Custom filter instance</div>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_detail">
+
+</a>
+<h3>Constructor Detail</h3>
+<a name="TJTransform()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJTransform</h4>
+<pre>public&nbsp;TJTransform()</pre>
+<div class="block">Create a new lossless transform instance.</div>
+</li>
+</ul>
+<a name="TJTransform(int, int, int, int, int, int, org.libjpegturbo.turbojpeg.TJCustomFilter)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJTransform</h4>
+<pre>public&nbsp;TJTransform(int&nbsp;x,
+           int&nbsp;y,
+           int&nbsp;w,
+           int&nbsp;h,
+           int&nbsp;op,
+           int&nbsp;options,
+           <a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg">TJCustomFilter</a>&nbsp;cf)</pre>
+<div class="block">Create a new lossless transform instance with the given parameters.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>x</code> - the left boundary of the cropping region.  This must be evenly
+ divisible by the MCU block width (see <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getMCUWidth(int)"><code>TJ.getMCUWidth(int)</code></a>)</dd><dd><code>y</code> - the upper boundary of the cropping region.  This must be evenly
+ divisible by the MCU block height (see <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#getMCUHeight(int)"><code>TJ.getMCUHeight(int)</code></a>)</dd><dd><code>w</code> - the width of the cropping region.  Setting this to 0 is the
+ equivalent of setting it to (width of the source JPEG image -
+ <code>x</code>).</dd><dd><code>h</code> - the height of the cropping region.  Setting this to 0 is the
+ equivalent of setting it to (height of the source JPEG image -
+ <code>y</code>).</dd><dd><code>op</code> - one of the transform operations (<code>OP_*</code>)</dd><dd><code>options</code> - the bitwise OR of one or more of the transform options
+ (<code>OPT_*</code>)</dd><dd><code>cf</code> - an instance of an object that implements the <a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><code>TJCustomFilter</code></a> interface, or null if no custom filter is needed</dd></dl>
+</li>
+</ul>
+<a name="TJTransform(java.awt.Rectangle, int, int, org.libjpegturbo.turbojpeg.TJCustomFilter)">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>TJTransform</h4>
+<pre>public&nbsp;TJTransform(java.awt.Rectangle&nbsp;r,
+           int&nbsp;op,
+           int&nbsp;options,
+           <a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg">TJCustomFilter</a>&nbsp;cf)</pre>
+<div class="block">Create a new lossless transform instance with the given parameters.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>r</code> - a <code>Rectangle</code> instance that specifies the cropping
+ region.  See <a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html#TJTransform(int,%20int,%20int,%20int,%20int,%20int,%20org.libjpegturbo.turbojpeg.TJCustomFilter)"><code>TJTransform(int, int, int, int, int, int, TJCustomFilter)</code></a> for more
+ detail.</dd><dd><code>op</code> - one of the transform operations (<code>OP_*</code>)</dd><dd><code>options</code> - the bitwise OR of one or more of the transform options
+ (<code>OPT_*</code>)</dd><dd><code>cf</code> - an instance of an object that implements the <a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><code>TJCustomFilter</code></a> interface, or null if no custom filter is needed</dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJTransform.html" target="_top">Frames</a></li>
+<li><a href="TJTransform.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li><a href="#nested_classes_inherited_from_class_java.awt.geom.Rectangle2D">Nested</a>&nbsp;|&nbsp;</li>
+<li><a href="#field_summary">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#methods_inherited_from_class_java.awt.Rectangle">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li><a href="#field_detail">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li>Method</li>
+</ul>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/TJTransformer.html Added

@@ -0,0 +1,421 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>TJTransformer</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJTransformer.html" target="_top">Frames</a></li>
+<li><a href="TJTransformer.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li><a href="#fields_inherited_from_class_org.libjpegturbo.turbojpeg.TJDecompressor">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+
+<div class="header">
+<div class="subTitle">org.libjpegturbo.turbojpeg</div>
+<h2 title="Class TJTransformer" class="title">Class TJTransformer</h2>
+</div>
+<div class="contentContainer">
+<ul class="inheritance">
+<li>java.lang.Object</li>
+<li>
+<ul class="inheritance">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">org.libjpegturbo.turbojpeg.TJDecompressor</a></li>
+<li>
+<ul class="inheritance">
+<li>org.libjpegturbo.turbojpeg.TJTransformer</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+<div class="description">
+<ul class="blockList">
+<li class="blockList">
+<dl>
+<dt>All Implemented Interfaces:</dt>
+<dd>java.io.Closeable, java.lang.AutoCloseable</dd>
+</dl>
+<hr>
+<br>
+<pre>public class <span class="strong">TJTransformer</span>
+extends <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></pre>
+<div class="block">TurboJPEG lossless transformer</div>
+</li>
+</ul>
+</div>
+<div class="summary">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="field_summary">
+
+</a>
+<h3>Field Summary</h3>
+<ul class="blockList">
+<li class="blockList"><a name="fields_inherited_from_class_org.libjpegturbo.turbojpeg.TJDecompressor">
+
+</a>
+<h3>Fields inherited from class&nbsp;org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></h3>
+<code><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#handle">handle</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegBuf">jpegBuf</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegBufSize">jpegBufSize</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegColorspace">jpegColorspace</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegHeight">jpegHeight</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegSubsamp">jpegSubsamp</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#jpegWidth">jpegWidth</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#yuvImage">yuvImage</a></code></li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_summary">
+
+</a>
+<h3>Constructor Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Constructor Summary table, listing constructors, and an explanation">
+<caption><span>Constructors</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Constructor and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#TJTransformer()">TJTransformer</a></strong>()</code>
+<div class="block">Create a TurboJPEG lossless transformer instance.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#TJTransformer(byte)">TJTransformer</a></strong>(byte&nbsp;jpegImage)</code>
+<div class="block">Create a TurboJPEG lossless transformer instance and associate the JPEG
+ image stored in <code>jpegImage</code> with the newly created instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#TJTransformer(byte,%20int)">TJTransformer</a></strong>(byte&nbsp;jpegImage,
+             int&nbsp;imageSize)</code>
+<div class="block">Create a TurboJPEG lossless transformer instance and associate the JPEG
+ image of length <code>imageSize</code> bytes stored in
+ <code>jpegImage</code> with the newly created instance.</div>
+</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_summary">
+
+</a>
+<h3>Method Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Method Summary table, listing methods, and an explanation">
+<caption><span>Methods</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Method and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#getTransformedSizes()">getTransformedSizes</a></strong>()</code>
+<div class="block">Returns an array containing the sizes of the transformed JPEG images
+ generated by the most recent transform operation.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#transform(byte,%20org.libjpegturbo.turbojpeg.TJTransform,%20int)">transform</a></strong>(byte&nbsp;dstBufs,
+         <a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a>&nbsp;transforms,
+         int&nbsp;flags)</code>
+<div class="block">Losslessly transform the JPEG image associated with this transformer
+ instance into one or more JPEG images stored in the given destination
+ buffers.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html#transform(org.libjpegturbo.turbojpeg.TJTransform,%20int)">transform</a></strong>(<a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a>&nbsp;transforms,
+         int&nbsp;flags)</code>
+<div class="block">Losslessly transform the JPEG image associated with this transformer
+ instance and return an array of <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><code>TJDecompressor</code></a> instances, each of
+ which has a transformed JPEG image associated with it.</div>
+</td>
+</tr>
+</table>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_org.libjpegturbo.turbojpeg.TJDecompressor">
+
+</a>
+<h3>Methods inherited from class&nbsp;org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></h3>
+<code><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#close()">close</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(java.awt.image.BufferedImage,%20int)">decompress</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int)">decompress</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(byte,%20int,%20int,%20int,%20int,%20int,%20int,%20int)">decompress</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(int,%20int,%20int,%20int,%20int,%20int,%20int,%20int)">decompress</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(int,%20int,%20int,%20int)">decompress</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompress(int,%20int,%20int,%20int,%20int)">decompress</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(byte,%20int)">decompressToYUV</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int)">decompressToYUV</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)">decompressToYUV</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(int,%20int,%20int,%20int)">decompressToYUV</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#decompressToYUV(org.libjpegturbo.turbojpeg.YUVImage,%20int)">decompressToYUV</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#finalize()">finalize</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getColorspace()">getColorspace</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getHeight()">getHeight</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getJPEGBuf()">getJPEGBuf</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getJPEGSize()">getJPEGSize</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledHeight(int,%20int)">getScaledHeight</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getScaledWidth(int,%20int)">getScaledWidth</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getSubsamp()">getSubsamp</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#getWidth()">getWidth</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#setJPEGImage(byte,%20int)">setJPEGImage</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(byte,%20int)">setSourceImage</a>, <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html#setSourceImage(org.libjpegturbo.turbojpeg.YUVImage)">setSourceImage</a></code></li>
+</ul>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.lang.Object">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.lang.Object</h3>
+<code>clone, equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait</code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="details">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_detail">
+
+</a>
+<h3>Constructor Detail</h3>
+<a name="TJTransformer()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJTransformer</h4>
+<pre>public&nbsp;TJTransformer()
+              throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG lossless transformer instance.</div>
+<dl><dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="TJTransformer(byte)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>TJTransformer</h4>
+<pre>public&nbsp;TJTransformer(byte&nbsp;jpegImage)
+              throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG lossless transformer instance and associate the JPEG
+ image stored in <code>jpegImage</code> with the newly created instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>jpegImage</code> - JPEG image buffer (size of the JPEG image is assumed to
+ be the length of the array.)  This buffer is not modified.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="TJTransformer(byte, int)">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>TJTransformer</h4>
+<pre>public&nbsp;TJTransformer(byte&nbsp;jpegImage,
+             int&nbsp;imageSize)
+              throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Create a TurboJPEG lossless transformer instance and associate the JPEG
+ image of length <code>imageSize</code> bytes stored in
+ <code>jpegImage</code> with the newly created instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>jpegImage</code> - JPEG image buffer.  This buffer is not modified.</dd><dd><code>imageSize</code> - size of the JPEG image (in bytes)</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_detail">
+
+</a>
+<h3>Method Detail</h3>
+<a name="transform(byte, org.libjpegturbo.turbojpeg.TJTransform, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>transform</h4>
+<pre>public&nbsp;void&nbsp;transform(byte&nbsp;dstBufs,
+             <a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a>&nbsp;transforms,
+             int&nbsp;flags)
+               throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Losslessly transform the JPEG image associated with this transformer
+ instance into one or more JPEG images stored in the given destination
+ buffers.  Lossless transforms work by moving the raw coefficients from one
+ JPEG image structure to another without altering the values of the
+ coefficients.  While this is typically faster than decompressing the
+ image, transforming it, and re-compressing it, lossless transforms are not
+ free.  Each lossless transform requires reading and performing Huffman
+ decoding on all of the coefficients in the source image, regardless of the
+ size of the destination image.  Thus, this method provides a means of
+ generating multiple transformed images from the same source or of applying
+ multiple transformations simultaneously, in order to eliminate the need to
+ read the source coefficients multiple times.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>dstBufs</code> - an array of image buffers.  <code>dstbufsi</code> will
+ receive a JPEG image that has been transformed using the parameters in
+ <code>transformsi</code>.  Use <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#bufSize(int,%20int,%20int)"><code>TJ.bufSize(int, int, int)</code></a> to determine the
+ maximum size for each buffer based on the transformed or cropped width and
+ height and the level of subsampling used in the source image.</dd><dd><code>transforms</code> - an array of <a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><code>TJTransform</code></a> instances, each of
+ which specifies the transform parameters and/or cropping region for the
+ corresponding transformed output image</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="transform(org.libjpegturbo.turbojpeg.TJTransform, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>transform</h4>
+<pre>public&nbsp;<a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a>&nbsp;transform(<a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a>&nbsp;transforms,
+                         int&nbsp;flags)
+                           throws <a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></pre>
+<div class="block">Losslessly transform the JPEG image associated with this transformer
+ instance and return an array of <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><code>TJDecompressor</code></a> instances, each of
+ which has a transformed JPEG image associated with it.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>transforms</code> - an array of <a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><code>TJTransform</code></a> instances, each of
+ which specifies the transform parameters and/or cropping region for the
+ corresponding transformed output image</dd><dd><code>flags</code> - the bitwise OR of one or more of
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#FLAG_BOTTOMUP"><code>TJ.FLAG_*</code></a></dd>
+<dt><span class="strong">Returns:</span></dt><dd>an array of <a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><code>TJDecompressor</code></a> instances, each of
+ which has a transformed JPEG image associated with it.</dd>
+<dt><span class="strong">Throws:</span></dt>
+<dd><code><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></code></dd></dl>
+</li>
+</ul>
+<a name="getTransformedSizes()">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>getTransformedSizes</h4>
+<pre>public&nbsp;int&nbsp;getTransformedSizes()</pre>
+<div class="block">Returns an array containing the sizes of the transformed JPEG images
+ generated by the most recent transform operation.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>an array containing the sizes of the transformed JPEG images
+ generated by the most recent transform operation.</dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Next Class</span></a></li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/TJTransformer.html" target="_top">Frames</a></li>
+<li><a href="TJTransformer.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li><a href="#fields_inherited_from_class_org.libjpegturbo.turbojpeg.TJDecompressor">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li>Field&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/YUVImage.html Added

@@ -0,0 +1,765 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>YUVImage</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li>Next Class</li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/YUVImage.html" target="_top">Frames</a></li>
+<li><a href="YUVImage.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li><a href="#field_summary">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li><a href="#field_detail">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+
+<div class="header">
+<div class="subTitle">org.libjpegturbo.turbojpeg</div>
+<h2 title="Class YUVImage" class="title">Class YUVImage</h2>
+</div>
+<div class="contentContainer">
+<ul class="inheritance">
+<li>java.lang.Object</li>
+<li>
+<ul class="inheritance">
+<li>org.libjpegturbo.turbojpeg.YUVImage</li>
+</ul>
+</li>
+</ul>
+<div class="description">
+<ul class="blockList">
+<li class="blockList">
+<hr>
+<br>
+<pre>public class <span class="strong">YUVImage</span>
+extends java.lang.Object</pre>
+<div class="block">This class encapsulates a YUV planar image and the metadata
+ associated with it.  The TurboJPEG API allows both the JPEG compression and
+ decompression pipelines to be split into stages:  YUV encode, compress from
+ YUV, decompress to YUV, and YUV decode.  A <code>YUVImage</code> instance
+ serves as the destination image for YUV encode and decompress-to-YUV
+ operations and as the source image for compress-from-YUV and YUV decode
+ operations.
+ <p>
+ Technically, the JPEG format uses the YCbCr colorspace (which technically is
+ not a "colorspace" but rather a "color transform"), but per the convention
+ of the digital video community, the TurboJPEG API uses "YUV" to refer to an
+ image format consisting of Y, Cb, and Cr image planes.
+ <p>
+ Each plane is simply a 2D array of bytes, each byte representing the value
+ of one of the components (Y, Cb, or Cr) at a particular location in the
+ image.  The width and height of each plane are determined by the image
+ width, height, and level of chrominance subsampling.  The luminance plane
+ width is the image width padded to the nearest multiple of the horizontal
+ subsampling factor (2 in the case of 4:2:0 and 4:2:2, 4 in the case of
+ 4:1:1, 1 in the case of 4:4:4 or grayscale.)  Similarly, the luminance plane
+ height is the image height padded to the nearest multiple of the vertical
+ subsampling factor (2 in the case of 4:2:0 or 4:4:0, 1 in the case of 4:4:4
+ or grayscale.)  The chrominance plane width is equal to the luminance plane
+ width divided by the horizontal subsampling factor, and the chrominance
+ plane height is equal to the luminance plane height divided by the vertical
+ subsampling factor.
+ <p>
+ For example, if the source image is 35 x 35 pixels and 4:2:2 subsampling is
+ used, then the luminance plane would be 36 x 35 bytes, and each of the
+ chrominance planes would be 18 x 35 bytes.  If you specify a line padding of
+ 4 bytes on top of this, then the luminance plane would be 36 x 35 bytes, and
+ each of the chrominance planes would be 20 x 35 bytes.</div>
+</li>
+</ul>
+</div>
+<div class="summary">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="field_summary">
+
+</a>
+<h3>Field Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Field Summary table, listing fields, and an explanation">
+<caption><span>Fields</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Field and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected long</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#handle">handle</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#yuvHeight">yuvHeight</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#yuvOffsets">yuvOffsets</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#yuvPad">yuvPad</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#yuvPlanes">yuvPlanes</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#yuvStrides">yuvStrides</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#yuvSubsamp">yuvSubsamp</a></strong></code>&nbsp;</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>protected int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#yuvWidth">yuvWidth</a></strong></code>&nbsp;</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_summary">
+
+</a>
+<h3>Constructor Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Constructor Summary table, listing constructors, and an explanation">
+<caption><span>Constructors</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colOne" scope="col">Constructor and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#YUVImage(byte,%20int,%20int,%20int,%20int,%20int)">YUVImage</a></strong>(byte&nbsp;planes,
+        int&nbsp;offsets,
+        int&nbsp;width,
+        int&nbsp;strides,
+        int&nbsp;height,
+        int&nbsp;subsamp)</code>
+<div class="block">Create a new <code>YUVImage</code> instance from a set of existing image
+ planes.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#YUVImage(byte,%20int,%20int,%20int,%20int)">YUVImage</a></strong>(byte&nbsp;yuvImage,
+        int&nbsp;width,
+        int&nbsp;pad,
+        int&nbsp;height,
+        int&nbsp;subsamp)</code>
+<div class="block">Create a new <code>YUVImage</code> instance from an existing unified image
+ buffer.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#YUVImage(int,%20int,%20int,%20int)">YUVImage</a></strong>(int&nbsp;width,
+        int&nbsp;strides,
+        int&nbsp;height,
+        int&nbsp;subsamp)</code>
+<div class="block">Create a new <code>YUVImage</code> instance backed by separate image
+ planes, and allocate memory for the image planes.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colOne"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#YUVImage(int,%20int,%20int,%20int)">YUVImage</a></strong>(int&nbsp;width,
+        int&nbsp;pad,
+        int&nbsp;height,
+        int&nbsp;subsamp)</code>
+<div class="block">Create a new <code>YUVImage</code> instance backed by a unified image
+ buffer, and allocate memory for the image buffer.</div>
+</td>
+</tr>
+</table>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_summary">
+
+</a>
+<h3>Method Summary</h3>
+<table class="overviewSummary" border="0" cellpadding="3" cellspacing="0" summary="Method Summary table, listing methods, and an explanation">
+<caption><span>Methods</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Modifier and Type</th>
+<th class="colLast" scope="col">Method and Description</th>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#getBuf()">getBuf</a></strong>()</code>
+<div class="block">Returns the YUV image buffer (if this image is stored in a unified
+ buffer rather than separate image planes.)</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#getHeight()">getHeight</a></strong>()</code>
+<div class="block">Returns the height of the YUV image (or subregion.)</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#getOffsets()">getOffsets</a></strong>()</code>
+<div class="block">Returns the offsets (in bytes) of each plane within the planes of a larger
+ YUV image.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#getPad()">getPad</a></strong>()</code>
+<div class="block">Returns the line padding used in the YUV image buffer (if this image is
+ stored in a unified buffer rather than separate image planes.)</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>byte</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#getPlanes()">getPlanes</a></strong>()</code>
+<div class="block">Returns the YUV image planes.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#getSize()">getSize</a></strong>()</code>
+<div class="block">Returns the size (in bytes) of the YUV image buffer (if this image is
+ stored in a unified buffer rather than separate image planes.)</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#getStrides()">getStrides</a></strong>()</code>
+<div class="block">Returns the number of bytes per line of each plane in the YUV image.</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#getSubsamp()">getSubsamp</a></strong>()</code>
+<div class="block">Returns the level of chrominance subsampling used in the YUV image.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>int</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#getWidth()">getWidth</a></strong>()</code>
+<div class="block">Returns the width of the YUV image (or subregion.)</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#setBuf(byte,%20int,%20int,%20int,%20int,%20int)">setBuf</a></strong>(byte&nbsp;planes,
+      int&nbsp;offsets,
+      int&nbsp;width,
+      int&nbsp;strides,
+      int&nbsp;height,
+      int&nbsp;subsamp)</code>
+<div class="block">Assign a set of image planes to this <code>YUVImage</code> instance.</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><code>void</code></td>
+<td class="colLast"><code><strong><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html#setBuf(byte,%20int,%20int,%20int,%20int)">setBuf</a></strong>(byte&nbsp;yuvImage,
+      int&nbsp;width,
+      int&nbsp;pad,
+      int&nbsp;height,
+      int&nbsp;subsamp)</code>
+<div class="block">Assign a unified image buffer to this <code>YUVImage</code> instance.</div>
+</td>
+</tr>
+</table>
+<ul class="blockList">
+<li class="blockList"><a name="methods_inherited_from_class_java.lang.Object">
+
+</a>
+<h3>Methods inherited from class&nbsp;java.lang.Object</h3>
+<code>clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait</code></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="details">
+<ul class="blockList">
+<li class="blockList">
+
+<ul class="blockList">
+<li class="blockList"><a name="field_detail">
+
+</a>
+<h3>Field Detail</h3>
+<a name="handle">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>handle</h4>
+<pre>protected&nbsp;long handle</pre>
+</li>
+</ul>
+<a name="yuvPlanes">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>yuvPlanes</h4>
+<pre>protected&nbsp;byte yuvPlanes</pre>
+</li>
+</ul>
+<a name="yuvOffsets">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>yuvOffsets</h4>
+<pre>protected&nbsp;int yuvOffsets</pre>
+</li>
+</ul>
+<a name="yuvStrides">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>yuvStrides</h4>
+<pre>protected&nbsp;int yuvStrides</pre>
+</li>
+</ul>
+<a name="yuvPad">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>yuvPad</h4>
+<pre>protected&nbsp;int yuvPad</pre>
+</li>
+</ul>
+<a name="yuvWidth">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>yuvWidth</h4>
+<pre>protected&nbsp;int yuvWidth</pre>
+</li>
+</ul>
+<a name="yuvHeight">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>yuvHeight</h4>
+<pre>protected&nbsp;int yuvHeight</pre>
+</li>
+</ul>
+<a name="yuvSubsamp">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>yuvSubsamp</h4>
+<pre>protected&nbsp;int yuvSubsamp</pre>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="constructor_detail">
+
+</a>
+<h3>Constructor Detail</h3>
+<a name="YUVImage(int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>YUVImage</h4>
+<pre>public&nbsp;YUVImage(int&nbsp;width,
+        int&nbsp;strides,
+        int&nbsp;height,
+        int&nbsp;subsamp)</pre>
+<div class="block">Create a new <code>YUVImage</code> instance backed by separate image
+ planes, and allocate memory for the image planes.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>width</code> - width (in pixels) of the YUV image</dd><dd><code>strides</code> - an array of integers, each specifying the number of bytes
+ per line in the corresponding plane of the YUV image.  Setting the stride
+ for any plane to 0 is the same as setting it to the plane width (see
+ <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>above</code></a>.)  If <code>strides</code> is null, then the
+ strides for all planes will be set to their respective plane widths.  When
+ using this constructor, the stride for each plane must be equal to or
+ greater than the plane width.</dd><dd><code>height</code> - height (in pixels) of the YUV image</dd><dd><code>subsamp</code> - the level of chrominance subsampling to be used in the YUV
+ image (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444"><code>TJ.SAMP_*</code></a>)</dd></dl>
+</li>
+</ul>
+<a name="YUVImage(int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>YUVImage</h4>
+<pre>public&nbsp;YUVImage(int&nbsp;width,
+        int&nbsp;pad,
+        int&nbsp;height,
+        int&nbsp;subsamp)</pre>
+<div class="block">Create a new <code>YUVImage</code> instance backed by a unified image
+ buffer, and allocate memory for the image buffer.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>width</code> - width (in pixels) of the YUV image</dd><dd><code>pad</code> - Each line of each plane in the YUV image buffer will be padded
+ to this number of bytes (must be a power of 2.)</dd><dd><code>height</code> - height (in pixels) of the YUV image</dd><dd><code>subsamp</code> - the level of chrominance subsampling to be used in the YUV
+ image (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444"><code>TJ.SAMP_*</code></a>)</dd></dl>
+</li>
+</ul>
+<a name="YUVImage(byte, int, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>YUVImage</h4>
+<pre>public&nbsp;YUVImage(byte&nbsp;planes,
+        int&nbsp;offsets,
+        int&nbsp;width,
+        int&nbsp;strides,
+        int&nbsp;height,
+        int&nbsp;subsamp)</pre>
+<div class="block">Create a new <code>YUVImage</code> instance from a set of existing image
+ planes.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>planes</code> - an array of buffers representing the Y, U (Cb), and V (Cr)
+ image planes (or just the Y plane, if the image is grayscale.)   These
+ planes can be contiguous or non-contiguous in memory.  Plane
+ <code>i</code> should be at least <code>offsetsi +
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#planeSizeYUV(int,%20int,%20int,%20int,%20int)"><code>TJ.planeSizeYUV</code></a>(i, width, stridesi, height, subsamp)</code>
+ bytes in size.</dd><dd><code>offsets</code> - If this <code>YUVImage</code> instance represents a
+ subregion of a larger image, then <code>offsetsi</code> specifies the
+ offset (in bytes) of the subregion within plane <code>i</code> of the
+ larger image.  Setting this to null is the same as setting the offsets for
+ all planes to 0.</dd><dd><code>width</code> - width (in pixels) of the new YUV image (or subregion)</dd><dd><code>strides</code> - an array of integers, each specifying the number of bytes
+ per line in the corresponding plane of the YUV image.  Setting the stride
+ for any plane to 0 is the same as setting it to the plane width (see
+ <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>above</code></a>.)  If <code>strides</code> is null, then the
+ strides for all planes will be set to their respective plane widths.  You
+ can adjust the strides in order to add an arbitrary amount of line padding
+ to each plane or to specify that this <code>YUVImage</code> instance is a
+ subregion of a larger image (in which case, <code>stridesi</code> should
+ be set to the plane width of plane <code>i</code> in the larger image.)</dd><dd><code>height</code> - height (in pixels) of the new YUV image (or subregion)</dd><dd><code>subsamp</code> - the level of chrominance subsampling used in the YUV
+ image (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444"><code>TJ.SAMP_*</code></a>)</dd></dl>
+</li>
+</ul>
+<a name="YUVImage(byte, int, int, int, int)">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>YUVImage</h4>
+<pre>public&nbsp;YUVImage(byte&nbsp;yuvImage,
+        int&nbsp;width,
+        int&nbsp;pad,
+        int&nbsp;height,
+        int&nbsp;subsamp)</pre>
+<div class="block">Create a new <code>YUVImage</code> instance from an existing unified image
+ buffer.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>yuvImage</code> - image buffer that contains or will contain YUV planar
+ image data.  Use <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int,%20int)"><code>TJ.bufSizeYUV(int, int, int, int)</code></a> to determine the minimum size for
+ this buffer.  The Y, U (Cb), and V (Cr) image planes are stored
+ sequentially in the buffer (see <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>above</code></a> for a description
+ of the image format.)</dd><dd><code>width</code> - width (in pixels) of the YUV image</dd><dd><code>pad</code> - the line padding used in the YUV image buffer.  For
+ instance, if each line in each plane of the buffer is padded to the
+ nearest multiple of 4 bytes, then <code>pad</code> should be set to 4.</dd><dd><code>height</code> - height (in pixels) of the YUV image</dd><dd><code>subsamp</code> - the level of chrominance subsampling used in the YUV
+ image (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444"><code>TJ.SAMP_*</code></a>)</dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+
+<ul class="blockList">
+<li class="blockList"><a name="method_detail">
+
+</a>
+<h3>Method Detail</h3>
+<a name="setBuf(byte, int, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setBuf</h4>
+<pre>public&nbsp;void&nbsp;setBuf(byte&nbsp;planes,
+          int&nbsp;offsets,
+          int&nbsp;width,
+          int&nbsp;strides,
+          int&nbsp;height,
+          int&nbsp;subsamp)</pre>
+<div class="block">Assign a set of image planes to this <code>YUVImage</code> instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>planes</code> - an array of buffers representing the Y, U (Cb), and V (Cr)
+ image planes (or just the Y plane, if the image is grayscale.)  These
+ planes can be contiguous or non-contiguous in memory.  Plane
+ <code>i</code> should be at least <code>offsetsi +
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#planeSizeYUV(int,%20int,%20int,%20int,%20int)"><code>TJ.planeSizeYUV</code></a>(i, width, stridesi, height, subsamp)</code>
+ bytes in size.</dd><dd><code>offsets</code> - If this <code>YUVImage</code> instance represents a
+ subregion of a larger image, then <code>offsetsi</code> specifies the
+ offset (in bytes) of the subregion within plane <code>i</code> of the
+ larger image.  Setting this to null is the same as setting the offsets for
+ all planes to 0.</dd><dd><code>width</code> - width (in pixels) of the YUV image (or subregion)</dd><dd><code>strides</code> - an array of integers, each specifying the number of bytes
+ per line in the corresponding plane of the YUV image.  Setting the stride
+ for any plane to 0 is the same as setting it to the plane width (see
+ <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>above</code></a>.)  If <code>strides</code> is null, then the
+ strides for all planes will be set to their respective plane widths.  You
+ can adjust the strides in order to add an arbitrary amount of line padding
+ to each plane or to specify that this <code>YUVImage</code> image is a
+ subregion of a larger image (in which case, <code>stridesi</code> should
+ be set to the plane width of plane <code>i</code> in the larger image.)</dd><dd><code>height</code> - height (in pixels) of the YUV image (or subregion)</dd><dd><code>subsamp</code> - the level of chrominance subsampling used in the YUV
+ image (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444"><code>TJ.SAMP_*</code></a>)</dd></dl>
+</li>
+</ul>
+<a name="setBuf(byte, int, int, int, int)">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>setBuf</h4>
+<pre>public&nbsp;void&nbsp;setBuf(byte&nbsp;yuvImage,
+          int&nbsp;width,
+          int&nbsp;pad,
+          int&nbsp;height,
+          int&nbsp;subsamp)</pre>
+<div class="block">Assign a unified image buffer to this <code>YUVImage</code> instance.</div>
+<dl><dt><span class="strong">Parameters:</span></dt><dd><code>yuvImage</code> - image buffer that contains or will contain YUV planar
+ image data.  Use <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#bufSizeYUV(int,%20int,%20int,%20int)"><code>TJ.bufSizeYUV(int, int, int, int)</code></a> to determine the minimum size for
+ this buffer.  The Y, U (Cb), and V (Cr) image planes are stored
+ sequentially in the buffer (see <a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><code>above</code></a> for a description
+ of the image format.)</dd><dd><code>width</code> - width (in pixels) of the YUV image</dd><dd><code>pad</code> - the line padding used in the YUV image buffer.  For
+ instance, if each line in each plane of the buffer is padded to the
+ nearest multiple of 4 bytes, then <code>pad</code> should be set to 4.</dd><dd><code>height</code> - height (in pixels) of the YUV image</dd><dd><code>subsamp</code> - the level of chrominance subsampling used in the YUV
+ image (one of <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444"><code>TJ.SAMP_*</code></a>)</dd></dl>
+</li>
+</ul>
+<a name="getWidth()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getWidth</h4>
+<pre>public&nbsp;int&nbsp;getWidth()</pre>
+<div class="block">Returns the width of the YUV image (or subregion.)</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the width of the YUV image (or subregion)</dd></dl>
+</li>
+</ul>
+<a name="getHeight()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getHeight</h4>
+<pre>public&nbsp;int&nbsp;getHeight()</pre>
+<div class="block">Returns the height of the YUV image (or subregion.)</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the height of the YUV image (or subregion)</dd></dl>
+</li>
+</ul>
+<a name="getPad()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getPad</h4>
+<pre>public&nbsp;int&nbsp;getPad()</pre>
+<div class="block">Returns the line padding used in the YUV image buffer (if this image is
+ stored in a unified buffer rather than separate image planes.)</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the line padding used in the YUV image buffer</dd></dl>
+</li>
+</ul>
+<a name="getStrides()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getStrides</h4>
+<pre>public&nbsp;int&nbsp;getStrides()</pre>
+<div class="block">Returns the number of bytes per line of each plane in the YUV image.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the number of bytes per line of each plane in the YUV image</dd></dl>
+</li>
+</ul>
+<a name="getOffsets()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getOffsets</h4>
+<pre>public&nbsp;int&nbsp;getOffsets()</pre>
+<div class="block">Returns the offsets (in bytes) of each plane within the planes of a larger
+ YUV image.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the offsets (in bytes) of each plane within the planes of a larger
+ YUV image</dd></dl>
+</li>
+</ul>
+<a name="getSubsamp()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getSubsamp</h4>
+<pre>public&nbsp;int&nbsp;getSubsamp()</pre>
+<div class="block">Returns the level of chrominance subsampling used in the YUV image.  See
+ <a href="../../../org/libjpegturbo/turbojpeg/TJ.html#SAMP_444"><code>TJ.SAMP_*</code></a>.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the level of chrominance subsampling used in the YUV image</dd></dl>
+</li>
+</ul>
+<a name="getPlanes()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getPlanes</h4>
+<pre>public&nbsp;byte&nbsp;getPlanes()</pre>
+<div class="block">Returns the YUV image planes.  If the image is stored in a unified buffer,
+ then all image planes will point to that buffer.</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the YUV image planes</dd></dl>
+</li>
+</ul>
+<a name="getBuf()">
+
+</a>
+<ul class="blockList">
+<li class="blockList">
+<h4>getBuf</h4>
+<pre>public&nbsp;byte&nbsp;getBuf()</pre>
+<div class="block">Returns the YUV image buffer (if this image is stored in a unified
+ buffer rather than separate image planes.)</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the YUV image buffer</dd></dl>
+</li>
+</ul>
+<a name="getSize()">
+
+</a>
+<ul class="blockListLast">
+<li class="blockList">
+<h4>getSize</h4>
+<pre>public&nbsp;int&nbsp;getSize()</pre>
+<div class="block">Returns the size (in bytes) of the YUV image buffer (if this image is
+ stored in a unified buffer rather than separate image planes.)</div>
+<dl><dt><span class="strong">Returns:</span></dt><dd>the size (in bytes) of the YUV image buffer</dd></dl>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li class="navBarCell1Rev">Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">Prev Class</span></a></li>
+<li>Next Class</li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/YUVImage.html" target="_top">Frames</a></li>
+<li><a href="YUVImage.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<div>
+<ul class="subNavList">
+<li>Summary:&nbsp;</li>
+<li>Nested&nbsp;|&nbsp;</li>
+<li><a href="#field_summary">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_summary">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_summary">Method</a></li>
+</ul>
+<ul class="subNavList">
+<li>Detail:&nbsp;</li>
+<li><a href="#field_detail">Field</a>&nbsp;|&nbsp;</li>
+<li><a href="#constructor_detail">Constr</a>&nbsp;|&nbsp;</li>
+<li><a href="#method_detail">Method</a></li>
+</ul>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/package-frame.html Added

@@ -0,0 +1,31 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>org.libjpegturbo.turbojpeg</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<h1 class="bar"><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html" target="classFrame">org.libjpegturbo.turbojpeg</a></h1>
+<div class="indexContainer">
+<h2 title="Interfaces">Interfaces</h2>
+<ul title="Interfaces">
+<li><a href="TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg" target="classFrame"><i>TJCustomFilter</i></a></li>
+</ul>
+<h2 title="Classes">Classes</h2>
+<ul title="Classes">
+<li><a href="TJ.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJ</a></li>
+<li><a href="TJCompressor.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJCompressor</a></li>
+<li><a href="TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJDecompressor</a></li>
+<li><a href="TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJScalingFactor</a></li>
+<li><a href="TJTransform.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJTransform</a></li>
+<li><a href="TJTransformer.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJTransformer</a></li>
+<li><a href="YUVImage.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">YUVImage</a></li>
+</ul>
+<h2 title="Exceptions">Exceptions</h2>
+<ul title="Exceptions">
+<li><a href="TJException.html" title="class in org.libjpegturbo.turbojpeg" target="classFrame">TJException</a></li>
+</ul>
+</div>
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/package-summary.html Added

@@ -0,0 +1,202 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>org.libjpegturbo.turbojpeg</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev Package</li>
+<li>Next Package</li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/package-summary.html" target="_top">Frames</a></li>
+<li><a href="package-summary.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+<div class="header">
+<h1 title="Package" class="title">Package&nbsp;org.libjpegturbo.turbojpeg</h1>
+</div>
+<div class="contentContainer">
+<ul class="blockList">
+<li class="blockList">
+<table class="packageSummary" border="0" cellpadding="3" cellspacing="0" summary="Interface Summary table, listing interfaces, and an explanation">
+<caption><span>Interface Summary</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Interface</th>
+<th class="colLast" scope="col">Description</th>
+</tr>
+<tbody>
+<tr class="altColor">
+<td class="colFirst"><a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg">TJCustomFilter</a></td>
+<td class="colLast">
+<div class="block">Custom filter callback interface</div>
+</td>
+</tr>
+</tbody>
+</table>
+</li>
+<li class="blockList">
+<table class="packageSummary" border="0" cellpadding="3" cellspacing="0" summary="Class Summary table, listing classes, and an explanation">
+<caption><span>Class Summary</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Class</th>
+<th class="colLast" scope="col">Description</th>
+</tr>
+<tbody>
+<tr class="altColor">
+<td class="colFirst"><a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg">TJ</a></td>
+<td class="colLast">
+<div class="block">TurboJPEG utility class (cannot be instantiated)</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg">TJCompressor</a></td>
+<td class="colLast">
+<div class="block">TurboJPEG compressor</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg">TJDecompressor</a></td>
+<td class="colLast">
+<div class="block">TurboJPEG decompressor</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg">TJScalingFactor</a></td>
+<td class="colLast">
+<div class="block">Fractional scaling factor</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">TJTransform</a></td>
+<td class="colLast">
+<div class="block">Lossless transform parameters</div>
+</td>
+</tr>
+<tr class="rowColor">
+<td class="colFirst"><a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg">TJTransformer</a></td>
+<td class="colLast">
+<div class="block">TurboJPEG lossless transformer</div>
+</td>
+</tr>
+<tr class="altColor">
+<td class="colFirst"><a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg">YUVImage</a></td>
+<td class="colLast">
+<div class="block">This class encapsulates a YUV planar image and the metadata
+ associated with it.</div>
+</td>
+</tr>
+</tbody>
+</table>
+</li>
+<li class="blockList">
+<table class="packageSummary" border="0" cellpadding="3" cellspacing="0" summary="Exception Summary table, listing exceptions, and an explanation">
+<caption><span>Exception Summary</span><span class="tabEnd">&nbsp;</span></caption>
+<tr>
+<th class="colFirst" scope="col">Exception</th>
+<th class="colLast" scope="col">Description</th>
+</tr>
+<tbody>
+<tr class="altColor">
+<td class="colFirst"><a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">TJException</a></td>
+<td class="colLast">&nbsp;</td>
+</tr>
+</tbody>
+</table>
+</li>
+</ul>
+</div>
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="package-tree.html">Tree</a></li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev Package</li>
+<li>Next Package</li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/package-summary.html" target="_top">Frames</a></li>
+<li><a href="package-summary.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/org/libjpegturbo/turbojpeg/package-tree.html Added

@@ -0,0 +1,160 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>org.libjpegturbo.turbojpeg Class Hierarchy</title>
+<link rel="stylesheet" type="text/css" href="../../../stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li class="navBarCell1Rev">Tree</li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/package-tree.html" target="_top">Frames</a></li>
+<li><a href="package-tree.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+<div class="header">
+<h1 class="title">Hierarchy For Package org.libjpegturbo.turbojpeg</h1>
+</div>
+<div class="contentContainer">
+<h2 title="Class Hierarchy">Class Hierarchy</h2>
+<ul>
+<li type="circle">java.lang.Object
+<ul>
+<li type="circle">java.awt.geom.RectangularShape (implements java.lang.Cloneable, java.awt.Shape)
+<ul>
+<li type="circle">java.awt.geom.Rectangle2D
+<ul>
+<li type="circle">java.awt.Rectangle (implements java.io.Serializable, java.awt.Shape)
+<ul>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJTransform</span></a></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+<li type="circle">java.lang.Throwable (implements java.io.Serializable)
+<ul>
+<li type="circle">java.lang.Exception
+<ul>
+<li type="circle">java.io.IOException
+<ul>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJException</span></a></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJ</span></a></li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJCompressor</span></a> (implements java.io.Closeable)</li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJDecompressor</span></a> (implements java.io.Closeable)
+<ul>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJTransformer</span></a></li>
+</ul>
+</li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJScalingFactor</span></a></li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">YUVImage</span></a></li>
+</ul>
+</li>
+</ul>
+<h2 title="Interface Hierarchy">Interface Hierarchy</h2>
+<ul>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="../../../org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><span class="strong">TJCustomFilter</span></a></li>
+</ul>
+</div>
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="../../../org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li class="navBarCell1Rev">Tree</li>
+<li><a href="../../../deprecated-list.html">Deprecated</a></li>
+<li><a href="../../../index-all.html">Index</a></li>
+<li><a href="../../../help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="../../../index.html?org/libjpegturbo/turbojpeg/package-tree.html" target="_top">Frames</a></li>
+<li><a href="package-tree.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="../../../allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/overview-tree.html Added

@@ -0,0 +1,164 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>Class Hierarchy</title>
+<link rel="stylesheet" type="text/css" href="stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li class="navBarCell1Rev">Tree</li>
+<li><a href="deprecated-list.html">Deprecated</a></li>
+<li><a href="index-all.html">Index</a></li>
+<li><a href="help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?overview-tree.html" target="_top">Frames</a></li>
+<li><a href="overview-tree.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+<div class="header">
+<h1 class="title">Hierarchy For All Packages</h1>
+<span class="strong">Package Hierarchies:</span>
+<ul class="horizontal">
+<li><a href="org/libjpegturbo/turbojpeg/package-tree.html">org.libjpegturbo.turbojpeg</a></li>
+</ul>
+</div>
+<div class="contentContainer">
+<h2 title="Class Hierarchy">Class Hierarchy</h2>
+<ul>
+<li type="circle">java.lang.Object
+<ul>
+<li type="circle">java.awt.geom.RectangularShape (implements java.lang.Cloneable, java.awt.Shape)
+<ul>
+<li type="circle">java.awt.geom.Rectangle2D
+<ul>
+<li type="circle">java.awt.Rectangle (implements java.io.Serializable, java.awt.Shape)
+<ul>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJTransform</span></a></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+<li type="circle">java.lang.Throwable (implements java.io.Serializable)
+<ul>
+<li type="circle">java.lang.Exception
+<ul>
+<li type="circle">java.io.IOException
+<ul>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJException</span></a></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJ.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJ</span></a></li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJCompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJCompressor</span></a> (implements java.io.Closeable)</li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJDecompressor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJDecompressor</span></a> (implements java.io.Closeable)
+<ul>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJTransformer.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJTransformer</span></a></li>
+</ul>
+</li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJScalingFactor.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">TJScalingFactor</span></a></li>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/YUVImage.html" title="class in org.libjpegturbo.turbojpeg"><span class="strong">YUVImage</span></a></li>
+</ul>
+</li>
+</ul>
+<h2 title="Interface Hierarchy">Interface Hierarchy</h2>
+<ul>
+<li type="circle">org.libjpegturbo.turbojpeg.<a href="org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg"><span class="strong">TJCustomFilter</span></a></li>
+</ul>
+</div>
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li class="navBarCell1Rev">Tree</li>
+<li><a href="deprecated-list.html">Deprecated</a></li>
+<li><a href="index-all.html">Index</a></li>
+<li><a href="help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?overview-tree.html" target="_top">Frames</a></li>
+<li><a href="overview-tree.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/package-list Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/resources Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/resources/background.gif Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/resources/tab.gif Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/resources/titlebar.gif Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/resources/titlebar_end.gif Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/script.js Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/serialized-form.html Added

@@ -0,0 +1,176 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
+
+<html lang="en">
+<head>
+<title>Serialized Form</title>
+<link rel="stylesheet" type="text/css" href="stylesheet.css" title="Style">
+</head>
+<body>
+<script type="text/javascript">
+</script>
+<noscript>
+<div>JavaScript is disabled on your browser.</div>
+</noscript>
+
+<div class="topNav"><a name="navbar_top">
+
+</a><a href="#skip-navbar_top" title="Skip navigation links"></a><a name="navbar_top_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li><a href="deprecated-list.html">Deprecated</a></li>
+<li><a href="index-all.html">Index</a></li>
+<li><a href="help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?serialized-form.html" target="_top">Frames</a></li>
+<li><a href="serialized-form.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_top">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_top">
+
+</a></div>
+
+<div class="header">
+<h1 title="Serialized Form" class="title">Serialized Form</h1>
+</div>
+<div class="serializedFormContainer">
+<ul class="blockList">
+<li class="blockList">
+<h2 title="Package">Package&nbsp;org.libjpegturbo.turbojpeg</h2>
+<ul class="blockList">
+<li class="blockList"><a name="org.libjpegturbo.turbojpeg.TJException">
+
+</a>
+<h3>Class <a href="org/libjpegturbo/turbojpeg/TJException.html" title="class in org.libjpegturbo.turbojpeg">org.libjpegturbo.turbojpeg.TJException</a> extends java.io.IOException implements Serializable</h3>
+<dl class="nameValue">
+<dt>serialVersionUID:</dt>
+<dd>1L</dd>
+</dl>
+<ul class="blockList">
+<li class="blockList"><a name="serializedForm">
+
+</a>
+<h3>Serialized Fields</h3>
+<ul class="blockList">
+<li class="blockListLast">
+<h4>errorCode</h4>
+<pre>int errorCode</pre>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+<li class="blockList"><a name="org.libjpegturbo.turbojpeg.TJTransform">
+
+</a>
+<h3>Class <a href="org/libjpegturbo/turbojpeg/TJTransform.html" title="class in org.libjpegturbo.turbojpeg">org.libjpegturbo.turbojpeg.TJTransform</a> extends java.awt.Rectangle implements Serializable</h3>
+<dl class="nameValue">
+<dt>serialVersionUID:</dt>
+<dd>-127367705761430371L</dd>
+</dl>
+<ul class="blockList">
+<li class="blockList"><a name="serializedForm">
+
+</a>
+<h3>Serialized Fields</h3>
+<ul class="blockList">
+<li class="blockList">
+<h4>op</h4>
+<pre>int op</pre>
+<div class="block">Transform operation (one of <code>OP_*</code>)</div>
+</li>
+<li class="blockList">
+<h4>options</h4>
+<pre>int options</pre>
+<div class="block">Transform options (bitwise OR of one or more of <code>OPT_*</code>)</div>
+</li>
+<li class="blockListLast">
+<h4>cf</h4>
+<pre><a href="org/libjpegturbo/turbojpeg/TJCustomFilter.html" title="interface in org.libjpegturbo.turbojpeg">TJCustomFilter</a> cf</pre>
+<div class="block">Custom filter instance</div>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</div>
+
+<div class="bottomNav"><a name="navbar_bottom">
+
+</a><a href="#skip-navbar_bottom" title="Skip navigation links"></a><a name="navbar_bottom_firstrow">
+
+</a>
+<ul class="navList" title="Navigation">
+<li><a href="org/libjpegturbo/turbojpeg/package-summary.html">Package</a></li>
+<li>Class</li>
+<li><a href="org/libjpegturbo/turbojpeg/package-tree.html">Tree</a></li>
+<li><a href="deprecated-list.html">Deprecated</a></li>
+<li><a href="index-all.html">Index</a></li>
+<li><a href="help-doc.html">Help</a></li>
+</ul>
+</div>
+<div class="subNav">
+<ul class="navList">
+<li>Prev</li>
+<li>Next</li>
+</ul>
+<ul class="navList">
+<li><a href="index.html?serialized-form.html" target="_top">Frames</a></li>
+<li><a href="serialized-form.html" target="_top">No Frames</a></li>
+</ul>
+<ul class="navList" id="allclasses_navbar_bottom">
+<li><a href="allclasses-noframe.html">All Classes</a></li>
+</ul>
+<div>
+<script type="text/javascript">
+</script>
+</div>
+<a name="skip-navbar_bottom">
+
+</a></div>
+
+</body>
+</html>

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/doc/stylesheet.css Added

@@ -0,0 +1,474 @@
+/* Javadoc style sheet */
+/*
+Overall document style
+*/
+body {
+    background-color:#ffffff;
+    color:#353833;
+    font-family:Arial, Helvetica, sans-serif;
+    font-size:76%;
+    margin:0;
+}
+a:link, a:visited {
+    text-decoration:none;
+    color:#4c6b87;
+}
+a:hover, a:focus {
+    text-decoration:none;
+    color:#bb7a2a;
+}
+a:active {
+    text-decoration:none;
+    color:#4c6b87;
+}
+aname {
+    color:#353833;
+}
+aname:hover {
+    text-decoration:none;
+    color:#353833;
+}
+pre {
+    font-size:1.3em;
+}
+h1 {
+    font-size:1.8em;
+}
+h2 {
+    font-size:1.5em;
+}
+h3 {
+    font-size:1.4em;
+}
+h4 {
+    font-size:1.3em;
+}
+h5 {
+    font-size:1.2em;
+}
+h6 {
+    font-size:1.1em;
+}
+ul {
+    list-style-type:disc;
+}
+code, tt {
+    font-size:1.2em;
+}
+dt code {
+    font-size:1.2em;
+}
+table tr td dt code {
+    font-size:1.2em;
+    vertical-align:top;
+}
+sup {
+    font-size:.6em;
+}
+/*
+Document title and Copyright styles
+*/
+.clear {
+    clear:both;
+    height:0px;
+    overflow:hidden;
+}
+.aboutLanguage {
+    float:right;
+    padding:0px 21px;
+    font-size:.8em;
+    z-index:200;
+    margin-top:-7px;
+}
+.legalCopy {
+    margin-left:.5em;
+}
+.bar a, .bar a:link, .bar a:visited, .bar a:active {
+    color:#FFFFFF;
+    text-decoration:none;
+}
+.bar a:hover, .bar a:focus {
+    color:#bb7a2a;
+}
+.tab {
+    background-color:#0066FF;
+    background-image:url(resources/titlebar.gif);
+    background-position:left top;
+    background-repeat:no-repeat;
+    color:#ffffff;
+    padding:8px;
+    width:5em;
+    font-weight:bold;
+}
+/*
+Navigation bar styles
+*/
+.bar {
+    background-image:url(resources/background.gif);
+    background-repeat:repeat-x;
+    color:#FFFFFF;
+    padding:.8em .5em .4em .8em;
+    height:auto;/*height:1.8em;*/
+    font-size:1em;
+    margin:0;
+}
+.topNav {
+    background-image:url(resources/background.gif);
+    background-repeat:repeat-x;
+    color:#FFFFFF;
+    float:left;
+    padding:0;
+    width:100%;
+    clear:right;
+    height:2.8em;
+    padding-top:10px;
+    overflow:hidden;
+}
+.bottomNav {
+    margin-top:10px;
+    background-image:url(resources/background.gif);
+    background-repeat:repeat-x;
+    color:#FFFFFF;
+    float:left;
+    padding:0;
+    width:100%;
+    clear:right;
+    height:2.8em;
+    padding-top:10px;
+    overflow:hidden;
+}
+.subNav {
+    background-color:#dee3e9;
+    border-bottom:1px solid #9eadc0;
+    float:left;
+    width:100%;
+    overflow:hidden;
+}
+.subNav div {
+    clear:left;
+    float:left;
+    padding:0 0 5px 6px;
+}
+ul.navList, ul.subNavList {
+    float:left;
+    margin:0 25px 0 0;
+    padding:0;
+}
+ul.navList li{
+    list-style:none;
+    float:left;
+    padding:3px 6px;
+}
+ul.subNavList li{
+    list-style:none;
+    float:left;
+    font-size:90%;
+}
+.topNav a:link, .topNav a:active, .topNav a:visited, .bottomNav a:link, .bottomNav a:active, .bottomNav a:visited {
+    color:#FFFFFF;
+    text-decoration:none;
+}
+.topNav a:hover, .bottomNav a:hover {
+    text-decoration:none;
+    color:#bb7a2a;
+}
+.navBarCell1Rev {
+    background-image:url(resources/tab.gif);
+    background-color:#a88834;
+    color:#FFFFFF;
+    margin: auto 5px;
+    border:1px solid #c9aa44;
+}
+/*
+Page header and footer styles
+*/
+.header, .footer {
+    clear:both;
+    margin:0 20px;
+    padding:5px 0 0 0;
+}
+.indexHeader {
+    margin:10px;
+    position:relative;
+}
+.indexHeader h1 {
+    font-size:1.3em;
+}
+.title {
+    color:#2c4557;
+    margin:10px 0;
+}
+.subTitle {
+    margin:5px 0 0 0;
+}
+.header ul {
+    margin:0 0 25px 0;
+    padding:0;
+}
+.footer ul {
+    margin:20px 0 5px 0;
+}
+.header ul li, .footer ul li {
+    list-style:none;
+    font-size:1.2em;
+}
+/*
+Heading styles
+*/
+div.details ul.blockList ul.blockList ul.blockList li.blockList h4, div.details ul.blockList ul.blockList ul.blockListLast li.blockList h4 {
+    background-color:#dee3e9;
+    border-top:1px solid #9eadc0;
+    border-bottom:1px solid #9eadc0;
+    margin:0 0 6px -8px;
+    padding:2px 5px;
+}
+ul.blockList ul.blockList ul.blockList li.blockList h3 {
+    background-color:#dee3e9;
+    border-top:1px solid #9eadc0;
+    border-bottom:1px solid #9eadc0;
+    margin:0 0 6px -8px;
+    padding:2px 5px;
+}
+ul.blockList ul.blockList li.blockList h3 {
+    padding:0;
+    margin:15px 0;
+}
+ul.blockList li.blockList h2 {
+    padding:0px 0 20px 0;
+}
+/*
+Page layout container styles
+*/
+.contentContainer, .sourceContainer, .classUseContainer, .serializedFormContainer, .constantValuesContainer {
+    clear:both;
+    padding:10px 20px;
+    position:relative;
+}
+.indexContainer {
+    margin:10px;
+    position:relative;
+    font-size:1.0em;
+}
+.indexContainer h2 {
+    font-size:1.1em;
+    padding:0 0 3px 0;
+}
+.indexContainer ul {
+    margin:0;
+    padding:0;
+}
+.indexContainer ul li {
+    list-style:none;
+}
+.contentContainer .description dl dt, .contentContainer .details dl dt, .serializedFormContainer dl dt {
+    font-size:1.1em;
+    font-weight:bold;
+    margin:10px 0 0 0;
+    color:#4E4E4E;
+}
+.contentContainer .description dl dd, .contentContainer .details dl dd, .serializedFormContainer dl dd {
+    margin:10px 0 10px 20px;
+}
+.serializedFormContainer dl.nameValue dt {
+    margin-left:1px;
+    font-size:1.1em;
+    display:inline;
+    font-weight:bold;
+}
+.serializedFormContainer dl.nameValue dd {
+    margin:0 0 0 1px;
+    font-size:1.1em;
+    display:inline;
+}
+/*
+List styles
+*/
+ul.horizontal li {
+    display:inline;
+    font-size:0.9em;
+}
+ul.inheritance {
+    margin:0;
+    padding:0;
+}
+ul.inheritance li {
+    display:inline;
+    list-style:none;
+}
+ul.inheritance li ul.inheritance {
+    margin-left:15px;
+    padding-left:15px;
+    padding-top:1px;
+}
+ul.blockList, ul.blockListLast {
+    margin:10px 0 10px 0;
+    padding:0;
+}
+ul.blockList li.blockList, ul.blockListLast li.blockList {
+    list-style:none;
+    margin-bottom:25px;
+}
+ul.blockList ul.blockList li.blockList, ul.blockList ul.blockListLast li.blockList {
+    padding:0px 20px 5px 10px;
+    border:1px solid #9eadc0;
+    background-color:#f9f9f9;
+}
+ul.blockList ul.blockList ul.blockList li.blockList, ul.blockList ul.blockList ul.blockListLast li.blockList {
+    padding:0 0 5px 8px;
+    background-color:#ffffff;
+    border:1px solid #9eadc0;
+    border-top:none;
+}
+ul.blockList ul.blockList ul.blockList ul.blockList li.blockList {
+    margin-left:0;
+    padding-left:0;
+    padding-bottom:15px;
+    border:none;
+    border-bottom:1px solid #9eadc0;
+}
+ul.blockList ul.blockList ul.blockList ul.blockList li.blockListLast {
+    list-style:none;
+    border-bottom:none;
+    padding-bottom:0;
+}
+table tr td dl, table tr td dl dt, table tr td dl dd {
+    margin-top:0;
+    margin-bottom:1px;
+}
+/*
+Table styles
+*/
+.contentContainer table, .classUseContainer table, .constantValuesContainer table {
+    border-bottom:1px solid #9eadc0;
+    width:100%;
+}
+.contentContainer ul li table, .classUseContainer ul li table, .constantValuesContainer ul li table {
+    width:100%;
+}
+.contentContainer .description table, .contentContainer .details table {
+    border-bottom:none;
+}
+.contentContainer ul li table th.colOne, .contentContainer ul li table th.colFirst, .contentContainer ul li table th.colLast, .classUseContainer ul li table th, .constantValuesContainer ul li table th, .contentContainer ul li table td.colOne, .contentContainer ul li table td.colFirst, .contentContainer ul li table td.colLast, .classUseContainer ul li table td, .constantValuesContainer ul li table td{
+    vertical-align:top;
+    padding-right:20px;
+}
+.contentContainer ul li table th.colLast, .classUseContainer ul li table th.colLast,.constantValuesContainer ul li table th.colLast,
+.contentContainer ul li table td.colLast, .classUseContainer ul li table td.colLast,.constantValuesContainer ul li table td.colLast,
+.contentContainer ul li table th.colOne, .classUseContainer ul li table th.colOne,
+.contentContainer ul li table td.colOne, .classUseContainer ul li table td.colOne {
+    padding-right:3px;
+}
+.overviewSummary caption, .packageSummary caption, .contentContainer ul.blockList li.blockList caption, .summary caption, .classUseContainer caption, .constantValuesContainer caption {
+    position:relative;
+    text-align:left;
+    background-repeat:no-repeat;
+    color:#FFFFFF;
+    font-weight:bold;
+    clear:none;
+    overflow:hidden;
+    padding:0px;
+    margin:0px;
+}
+caption a:link, caption a:hover, caption a:active, caption a:visited {
+    color:#FFFFFF;
+}
+.overviewSummary caption span, .packageSummary caption span, .contentContainer ul.blockList li.blockList caption span, .summary caption span, .classUseContainer caption span, .constantValuesContainer caption span {
+    white-space:nowrap;
+    padding-top:8px;
+    padding-left:8px;
+    display:block;
+    float:left;
+    background-image:url(resources/titlebar.gif);
+    height:18px;
+}
+.overviewSummary .tabEnd, .packageSummary .tabEnd, .contentContainer ul.blockList li.blockList .tabEnd, .summary .tabEnd, .classUseContainer .tabEnd, .constantValuesContainer .tabEnd {
+    width:10px;
+    background-image:url(resources/titlebar_end.gif);
+    background-repeat:no-repeat;
+    background-position:top right;
+    position:relative;
+    float:left;
+}
+ul.blockList ul.blockList li.blockList table {
+    margin:0 0 12px 0px;
+    width:100%;
+}
+.tableSubHeadingColor {
+    background-color: #EEEEFF;
+}
+.altColor {
+    background-color:#eeeeef;
+}
+.rowColor {
+    background-color:#ffffff;
+}
+.overviewSummary td, .packageSummary td, .contentContainer ul.blockList li.blockList td, .summary td, .classUseContainer td, .constantValuesContainer td {
+    text-align:left;
+    padding:3px 3px 3px 7px;
+}
+th.colFirst, th.colLast, th.colOne, .constantValuesContainer th {
+    background:#dee3e9;
+    border-top:1px solid #9eadc0;
+    border-bottom:1px solid #9eadc0;
+    text-align:left;
+    padding:3px 3px 3px 7px;
+}
+td.colOne a:link, td.colOne a:active, td.colOne a:visited, td.colOne a:hover, td.colFirst a:link, td.colFirst a:active, td.colFirst a:visited, td.colFirst a:hover, td.colLast a:link, td.colLast a:active, td.colLast a:visited, td.colLast a:hover, .constantValuesContainer td a:link, .constantValuesContainer td a:active, .constantValuesContainer td a:visited, .constantValuesContainer td a:hover {
+    font-weight:bold;
+}
+td.colFirst, th.colFirst {
+    border-left:1px solid #9eadc0;
+    white-space:nowrap;
+}
+td.colLast, th.colLast {
+    border-right:1px solid #9eadc0;
+}
+td.colOne, th.colOne {
+    border-right:1px solid #9eadc0;
+    border-left:1px solid #9eadc0;
+}
+table.overviewSummary  {
+    padding:0px;
+    margin-left:0px;
+}
+table.overviewSummary td.colFirst, table.overviewSummary th.colFirst,
+table.overviewSummary td.colOne, table.overviewSummary th.colOne {
+    width:25%;
+    vertical-align:middle;
+}
+table.packageSummary td.colFirst, table.overviewSummary th.colFirst {
+    width:25%;
+    vertical-align:middle;
+}
+/*
+Content styles
+*/
+.description pre {
+    margin-top:0;
+}
+.deprecatedContent {
+    margin:0;
+    padding:10px 0;
+}
+.docSummary {
+    padding:0;
+}
+/*
+Formatting effect styles
+*/
+.sourceLineNo {
+    color:green;
+    padding:0 30px 0 0;
+}
+h1.hidden {
+    visibility:hidden;
+    overflow:hidden;
+    font-size:.9em;
+}
+.block {
+    display:block;
+    margin:3px 0 0 0;
+}
+.strong {
+    font-weight:bold;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJ.java Added

@@ -0,0 +1,596 @@
+/*
+ * Copyright (C)2011-2013, 2017-2018, 2020-2021 D. R. Commander.
+ *                                              All Rights Reserved.
+ * Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+/**
+ * TurboJPEG utility class (cannot be instantiated)
+ */
+public final class TJ {
+
+  private TJ() {}
+
+  /**
+   * The number of chrominance subsampling options
+   */
+  public static final int NUMSAMP   = 6;
+  /**
+   * 4:4:4 chrominance subsampling (no chrominance subsampling).  The JPEG
+   * or YUV image will contain one chrominance component for every pixel in the
+   * source image.
+   */
+  public static final int SAMP_444  = 0;
+  /**
+   * 4:2:2 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 2x1 block of pixels in the source image.
+   */
+  public static final int SAMP_422  = 1;
+  /**
+   * 4:2:0 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 2x2 block of pixels in the source image.
+   */
+  public static final int SAMP_420  = 2;
+  /**
+   * Grayscale.  The JPEG or YUV image will contain no chrominance components.
+   */
+  public static final int SAMP_GRAY = 3;
+  /**
+   * 4:4:0 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 1x2 block of pixels in the source image.
+   * Note that 4:4:0 subsampling is not fully accelerated in libjpeg-turbo.
+   */
+  public static final int SAMP_440  = 4;
+  /**
+   * 4:1:1 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 4x1 block of pixels in the source image.
+   * JPEG images compressed with 4:1:1 subsampling will be almost exactly the
+   * same size as those compressed with 4:2:0 subsampling, and in the
+   * aggregate, both subsampling methods produce approximately the same
+   * perceptual quality.  However, 4:1:1 is better able to reproduce sharp
+   * horizontal features.  Note that 4:1:1 subsampling is not fully accelerated
+   * in libjpeg-turbo.
+   */
+  public static final int SAMP_411  = 5;
+
+
+  /**
+   * Returns the MCU block width for the given level of chrominance
+   * subsampling.
+   *
+   * @param subsamp the level of chrominance subsampling (one of
+   * <code>SAMP_*</code>)
+   *
+   * @return the MCU block width for the given level of chrominance
+   * subsampling.
+   */
+  public static int getMCUWidth(int subsamp) {
+    checkSubsampling(subsamp);
+    return MCU_WIDTHsubsamp;
+  }
+
+  private static final int MCU_WIDTH = {
+    8, 16, 16, 8, 8, 32
+  };
+
+
+  /**
+   * Returns the MCU block height for the given level of chrominance
+   * subsampling.
+   *
+   * @param subsamp the level of chrominance subsampling (one of
+   * <code>SAMP_*</code>)
+   *
+   * @return the MCU block height for the given level of chrominance
+   * subsampling.
+   */
+  public static int getMCUHeight(int subsamp) {
+    checkSubsampling(subsamp);
+    return MCU_HEIGHTsubsamp;
+  }
+
+  private static final int MCU_HEIGHT = {
+    8, 8, 16, 8, 16, 8
+  };
+
+
+  /**
+   * The number of pixel formats
+   */
+  public static final int NUMPF   = 12;
+  /**
+   * RGB pixel format.  The red, green, and blue components in the image are
+   * stored in 3-byte pixels in the order R, G, B from lowest to highest byte
+   * address within each pixel.
+   */
+  public static final int PF_RGB  = 0;
+  /**
+   * BGR pixel format.  The red, green, and blue components in the image are
+   * stored in 3-byte pixels in the order B, G, R from lowest to highest byte
+   * address within each pixel.
+   */
+  public static final int PF_BGR  = 1;
+  /**
+   * RGBX pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order R, G, B from lowest to highest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  public static final int PF_RGBX = 2;
+  /**
+   * BGRX pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order B, G, R from lowest to highest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  public static final int PF_BGRX = 3;
+  /**
+   * XBGR pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order R, G, B from highest to lowest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  public static final int PF_XBGR = 4;
+  /**
+   * XRGB pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order B, G, R from highest to lowest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  public static final int PF_XRGB = 5;
+  /**
+   * Grayscale pixel format.  Each 1-byte pixel represents a luminance
+   * (brightness) level from 0 to 255.
+   */
+  public static final int PF_GRAY = 6;
+  /**
+   * RGBA pixel format.  This is the same as {@link #PF_RGBX}, except that when
+   * decompressing, the X byte is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  public static final int PF_RGBA = 7;
+  /**
+   * BGRA pixel format.  This is the same as {@link #PF_BGRX}, except that when
+   * decompressing, the X byte is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  public static final int PF_BGRA = 8;
+  /**
+   * ABGR pixel format.  This is the same as {@link #PF_XBGR}, except that when
+   * decompressing, the X byte is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  public static final int PF_ABGR = 9;
+  /**
+   * ARGB pixel format.  This is the same as {@link #PF_XRGB}, except that when
+   * decompressing, the X byte is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  public static final int PF_ARGB = 10;
+  /**
+   * CMYK pixel format.  Unlike RGB, which is an additive color model used
+   * primarily for display, CMYK (Cyan/Magenta/Yellow/Key) is a subtractive
+   * color model used primarily for printing.  In the CMYK color model, the
+   * value of each color component typically corresponds to an amount of cyan,
+   * magenta, yellow, or black ink that is applied to a white background.  In
+   * order to convert between CMYK and RGB, it is necessary to use a color
+   * management system (CMS.)  A CMS will attempt to map colors within the
+   * printer's gamut to perceptually similar colors in the display's gamut and
+   * vice versa, but the mapping is typically not 1:1 or reversible, nor can it
+   * be defined with a simple formula.  Thus, such a conversion is out of scope
+   * for a codec library.  However, the TurboJPEG API allows for compressing
+   * CMYK pixels into a YCCK JPEG image (see {@link #CS_YCCK}) and
+   * decompressing YCCK JPEG images into CMYK pixels.
+   */
+  public static final int PF_CMYK = 11;
+
+
+  /**
+   * Returns the pixel size (in bytes) for the given pixel format.
+   *
+   * @param pixelFormat the pixel format (one of <code>PF_*</code>)
+   *
+   * @return the pixel size (in bytes) for the given pixel format.
+   */
+  public static int getPixelSize(int pixelFormat) {
+    checkPixelFormat(pixelFormat);
+    return PIXEL_SIZEpixelFormat;
+  }
+
+  private static final int PIXEL_SIZE = {
+    3, 3, 4, 4, 4, 4, 1, 4, 4, 4, 4, 4
+  };
+
+
+  /**
+   * For the given pixel format, returns the number of bytes that the red
+   * component is offset from the start of the pixel.  For instance, if a pixel
+   * of format <code>TJ.PF_BGRX</code> is stored in <code>char pixel</code>,
+   * then the red component will be
+   * <code>pixelTJ.getRedOffset(TJ.PF_BGRX)</code>.
+   *
+   * @param pixelFormat the pixel format (one of <code>PF_*</code>)
+   *
+   * @return the red offset for the given pixel format, or -1 if the pixel
+   * format does not have a red component.
+   */
+  public static int getRedOffset(int pixelFormat) {
+    checkPixelFormat(pixelFormat);
+    return RED_OFFSETpixelFormat;
+  }
+
+  private static final int RED_OFFSET = {
+    0, 2, 0, 2, 3, 1, -1, 0, 2, 3, 1, -1
+  };
+
+
+  /**
+   * For the given pixel format, returns the number of bytes that the green
+   * component is offset from the start of the pixel.  For instance, if a pixel
+   * of format <code>TJ.PF_BGRX</code> is stored in <code>char pixel</code>,
+   * then the green component will be
+   * <code>pixelTJ.getGreenOffset(TJ.PF_BGRX)</code>.
+   *
+   * @param pixelFormat the pixel format (one of <code>PF_*</code>)
+   *
+   * @return the green offset for the given pixel format, or -1 if the pixel
+   * format does not have a green component.
+   */
+  public static int getGreenOffset(int pixelFormat) {
+    checkPixelFormat(pixelFormat);
+    return GREEN_OFFSETpixelFormat;
+  }
+
+  private static final int GREEN_OFFSET = {
+    1, 1, 1, 1, 2, 2, -1, 1, 1, 2, 2, -1
+  };
+
+
+  /**
+   * For the given pixel format, returns the number of bytes that the blue
+   * component is offset from the start of the pixel.  For instance, if a pixel
+   * of format <code>TJ.PF_BGRX</code> is stored in <code>char pixel</code>,
+   * then the blue component will be
+   * <code>pixelTJ.getBlueOffset(TJ.PF_BGRX)</code>.
+   *
+   * @param pixelFormat the pixel format (one of <code>PF_*</code>)
+   *
+   * @return the blue offset for the given pixel format, or -1 if the pixel
+   * format does not have a blue component.
+   */
+  public static int getBlueOffset(int pixelFormat) {
+    checkPixelFormat(pixelFormat);
+    return BLUE_OFFSETpixelFormat;
+  }
+
+  private static final int BLUE_OFFSET = {
+    2, 0, 2, 0, 1, 3, -1, 2, 0, 1, 3, -1
+  };
+
+
+  /**
+   * For the given pixel format, returns the number of bytes that the alpha
+   * component is offset from the start of the pixel.  For instance, if a pixel
+   * of format <code>TJ.PF_BGRA</code> is stored in <code>char pixel</code>,
+   * then the alpha component will be
+   * <code>pixelTJ.getAlphaOffset(TJ.PF_BGRA)</code>.
+   *
+   * @param pixelFormat the pixel format (one of <code>PF_*</code>)
+   *
+   * @return the alpha offset for the given pixel format, or -1 if the pixel
+   * format does not have a alpha component.
+   */
+  public static int getAlphaOffset(int pixelFormat) {
+    checkPixelFormat(pixelFormat);
+    return ALPHA_OFFSETpixelFormat;
+  }
+
+  private static final int ALPHA_OFFSET = {
+    -1, -1, -1, -1, -1, -1, -1, 3, 3, 0, 0, -1
+  };
+
+
+  /**
+   * The number of JPEG colorspaces
+   */
+  public static final int NUMCS = 5;
+  /**
+   * RGB colorspace.  When compressing the JPEG image, the R, G, and B
+   * components in the source image are reordered into image planes, but no
+   * colorspace conversion or subsampling is performed.  RGB JPEG images can be
+   * decompressed to any of the extended RGB pixel formats or grayscale, but
+   * they cannot be decompressed to YUV images.
+   */
+  public static final int CS_RGB = 0;
+  /**
+   * YCbCr colorspace.  YCbCr is not an absolute colorspace but rather a
+   * mathematical transformation of RGB designed solely for storage and
+   * transmission.  YCbCr images must be converted to RGB before they can
+   * actually be displayed.  In the YCbCr colorspace, the Y (luminance)
+   * component represents the black &amp; white portion of the original image,
+   * and the Cb and Cr (chrominance) components represent the color portion of
+   * the original image.  Originally, the analog equivalent of this
+   * transformation allowed the same signal to drive both black &amp; white and
+   * color televisions, but JPEG images use YCbCr primarily because it allows
+   * the color data to be optionally subsampled for the purposes of reducing
+   * bandwidth or disk space.  YCbCr is the most common JPEG colorspace, and
+   * YCbCr JPEG images can be compressed from and decompressed to any of the
+   * extended RGB pixel formats or grayscale, or they can be decompressed to
+   * YUV planar images.
+   */
+  @SuppressWarnings("checkstyle:ConstantName")
+  public static final int CS_YCbCr = 1;
+  /**
+   * Grayscale colorspace.  The JPEG image retains only the luminance data (Y
+   * component), and any color data from the source image is discarded.
+   * Grayscale JPEG images can be compressed from and decompressed to any of
+   * the extended RGB pixel formats or grayscale, or they can be decompressed
+   * to YUV planar images.
+   */
+  public static final int CS_GRAY = 2;
+  /**
+   * CMYK colorspace.  When compressing the JPEG image, the C, M, Y, and K
+   * components in the source image are reordered into image planes, but no
+   * colorspace conversion or subsampling is performed.  CMYK JPEG images can
+   * only be decompressed to CMYK pixels.
+   */
+  public static final int CS_CMYK = 3;
+  /**
+   * YCCK colorspace.  YCCK (AKA "YCbCrK") is not an absolute colorspace but
+   * rather a mathematical transformation of CMYK designed solely for storage
+   * and transmission.  It is to CMYK as YCbCr is to RGB.  CMYK pixels can be
+   * reversibly transformed into YCCK, and as with YCbCr, the chrominance
+   * components in the YCCK pixels can be subsampled without incurring major
+   * perceptual loss.  YCCK JPEG images can only be compressed from and
+   * decompressed to CMYK pixels.
+   */
+  public static final int CS_YCCK = 4;
+
+
+  /**
+   * The uncompressed source/destination image is stored in bottom-up (Windows,
+   * OpenGL) order, not top-down (X11) order.
+   */
+  public static final int FLAG_BOTTOMUP      = 2;
+
+  @SuppressWarnings("checkstyle:JavadocVariable")
+  @Deprecated
+  public static final int FLAG_FORCEMMX      = 8;
+  @SuppressWarnings("checkstyle:JavadocVariable")
+  @Deprecated
+  public static final int FLAG_FORCESSE      = 16;
+  @SuppressWarnings("checkstyle:JavadocVariable")
+  @Deprecated
+  public static final int FLAG_FORCESSE2     = 32;
+  @SuppressWarnings("checkstyle:JavadocVariable")
+  @Deprecated
+  public static final int FLAG_FORCESSE3     = 128;
+
+  /**
+   * When decompressing an image that was compressed using chrominance
+   * subsampling, use the fastest chrominance upsampling algorithm available in
+   * the underlying codec.  The default is to use smooth upsampling, which
+   * creates a smooth transition between neighboring chrominance components in
+   * order to reduce upsampling artifacts in the decompressed image.
+   */
+  public static final int FLAG_FASTUPSAMPLE  = 256;
+  /**
+   * Use the fastest DCT/IDCT algorithm available in the underlying codec.  The
+   * default if this flag is not specified is implementation-specific.  For
+   * example, the implementation of TurboJPEG for libjpeg-turbo uses the fast
+   * algorithm by default when compressing, because this has been shown to have
+   * only a very slight effect on accuracy, but it uses the accurate algorithm
+   * when decompressing, because this has been shown to have a larger effect.
+   */
+  public static final int FLAG_FASTDCT       = 2048;
+  /**
+   * Use the most accurate DCT/IDCT algorithm available in the underlying
+   * codec.  The default if this flag is not specified is
+   * implementation-specific.  For example, the implementation of TurboJPEG for
+   * libjpeg-turbo uses the fast algorithm by default when compressing,
+   * because this has been shown to have only a very slight effect on accuracy,
+   * but it uses the accurate algorithm when decompressing, because this has
+   * been shown to have a larger effect.
+   */
+  public static final int FLAG_ACCURATEDCT   = 4096;
+  /**
+   * Immediately discontinue the current compression/decompression/transform
+   * operation if the underlying codec throws a warning (non-fatal error).  The
+   * default behavior is to allow the operation to complete unless a fatal
+   * error is encountered.
+   * <p>
+   * NOTE: due to the design of the TurboJPEG Java API, only certain methods
+   * (specifically, {@link TJDecompressor TJDecompressor.decompress*()} methods
+   * with a void return type) will complete and leave the output image in a
+   * fully recoverable state after a non-fatal error occurs.
+   */
+  public static final int FLAG_STOPONWARNING = 8192;
+  /**
+   * Use progressive entropy coding in JPEG images generated by compression and
+   * transform operations.  Progressive entropy coding will generally improve
+   * compression relative to baseline entropy coding (the default), but it will
+   * reduce compression and decompression performance considerably.
+   */
+  public static final int FLAG_PROGRESSIVE   = 16384;
+  /**
+   * Limit the number of progressive JPEG scans that the decompression and
+   * transform operations will process.  If a progressive JPEG image contains
+   * an unreasonably large number of scans, then this flag will cause the
+   * decompression and transform operations to throw an error.  The primary
+   * purpose of this is to allow security-critical applications to guard
+   * against an exploit of the progressive JPEG format described in
+   * <a href="https://libjpeg-turbo.org/pmwiki/uploads/About/TwoIssueswiththeJPEGStandard.pdf" target="_blank">this report</a>.
+   */
+  public static final int FLAG_LIMITSCANS    = 32768;
+
+
+  /**
+   * The number of error codes
+   */
+  public static final int NUMERR = 2;
+  /**
+   * The error was non-fatal and recoverable, but the image may still be
+   * corrupt.
+   * <p>
+   * NOTE: due to the design of the TurboJPEG Java API, only certain methods
+   * (specifically, {@link TJDecompressor TJDecompressor.decompress*()} methods
+   * with a void return type) will complete and leave the output image in a
+   * fully recoverable state after a non-fatal error occurs.
+   */
+  public static final int ERR_WARNING = 0;
+  /**
+   * The error was fatal and non-recoverable.
+   */
+  public static final int ERR_FATAL = 1;
+
+
+  /**
+   * Returns the maximum size of the buffer (in bytes) required to hold a JPEG
+   * image with the given width, height, and level of chrominance subsampling.
+   *
+   * @param width the width (in pixels) of the JPEG image
+   *
+   * @param height the height (in pixels) of the JPEG image
+   *
+   * @param jpegSubsamp the level of chrominance subsampling to be used when
+   * generating the JPEG image (one of {@link TJ TJ.SAMP_*})
+   *
+   * @return the maximum size of the buffer (in bytes) required to hold a JPEG
+   * image with the given width, height, and level of chrominance subsampling.
+   */
+  public static native int bufSize(int width, int height, int jpegSubsamp);
+
+  /**
+   * Returns the size of the buffer (in bytes) required to hold a YUV planar
+   * image with the given width, height, and level of chrominance subsampling.
+   *
+   * @param width the width (in pixels) of the YUV image
+   *
+   * @param pad the width of each line in each plane of the image is padded to
+   * the nearest multiple of this number of bytes (must be a power of 2.)
+   *
+   * @param height the height (in pixels) of the YUV image
+   *
+   * @param subsamp the level of chrominance subsampling used in the YUV
+   * image (one of {@link TJ TJ.SAMP_*})
+   *
+   * @return the size of the buffer (in bytes) required to hold a YUV planar
+   * image with the given width, height, and level of chrominance subsampling.
+   */
+  public static native int bufSizeYUV(int width, int pad, int height,
+                                      int subsamp);
+
+  /**
+   * @deprecated Use {@link #bufSizeYUV(int, int, int, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public static native int bufSizeYUV(int width, int height, int subsamp);
+
+  /**
+   * Returns the size of the buffer (in bytes) required to hold a YUV image
+   * plane with the given parameters.
+   *
+   * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb,
+   * 2 = V/Cr)
+   *
+   * @param width width (in pixels) of the YUV image.  NOTE: this is the width
+   * of the whole image, not the plane width.
+   *
+   * @param stride bytes per line in the image plane.
+   *
+   * @param height height (in pixels) of the YUV image.  NOTE: this is the
+   * height of the whole image, not the plane height.
+   *
+   * @param subsamp the level of chrominance subsampling used in the YUV
+   * image (one of {@link TJ TJ.SAMP_*})
+   *
+   * @return the size of the buffer (in bytes) required to hold a YUV planar
+   * image with the given parameters.
+   */
+  public static native int planeSizeYUV(int componentID, int width, int stride,
+                                        int height, int subsamp);
+
+  /**
+   * Returns the plane width of a YUV image plane with the given parameters.
+   * Refer to {@link YUVImage YUVImage} for a description of plane width.
+   *
+   * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb,
+   * 2 = V/Cr)
+   *
+   * @param width width (in pixels) of the YUV image
+   *
+   * @param subsamp the level of chrominance subsampling used in the YUV image
+   * (one of {@link TJ TJ.SAMP_*})
+   *
+   * @return the plane width of a YUV image plane with the given parameters.
+   */
+  public static native int planeWidth(int componentID, int width, int subsamp);
+
+  /**
+   * Returns the plane height of a YUV image plane with the given parameters.
+   * Refer to {@link YUVImage YUVImage} for a description of plane height.
+   *
+   * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb,
+   * 2 = V/Cr)
+   *
+   * @param height height (in pixels) of the YUV image
+   *
+   * @param subsamp the level of chrominance subsampling used in the YUV image
+   * (one of {@link TJ TJ.SAMP_*})
+   *
+   * @return the plane height of a YUV image plane with the given parameters.
+   */
+  public static native int planeHeight(int componentID, int height,
+                                       int subsamp);
+
+  /**
+   * Returns a list of fractional scaling factors that the JPEG decompressor in
+   * this implementation of TurboJPEG supports.
+   *
+   * @return a list of fractional scaling factors that the JPEG decompressor in
+   * this implementation of TurboJPEG supports.
+   */
+  public static native TJScalingFactor getScalingFactors();
+
+  static {
+    TJLoader.load();
+  }
+
+  private static void checkPixelFormat(int pixelFormat) {
+    if (pixelFormat < 0 || pixelFormat >= NUMPF)
+      throw new IllegalArgumentException("Invalid pixel format");
+  }
+
+  private static void checkSubsampling(int subsamp) {
+    if (subsamp < 0 || subsamp >= NUMSAMP)
+      throw new IllegalArgumentException("Invalid subsampling type");
+  }
+
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJCompressor.java Added

@@ -0,0 +1,684 @@
+/*
+ * Copyright (C)2011-2015, 2018, 2020 D. R. Commander.  All Rights Reserved.
+ * Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+import java.awt.image.*;
+import java.nio.*;
+import java.io.*;
+
+/**
+ * TurboJPEG compressor
+ */
+public class TJCompressor implements Closeable {
+
+  private static final String NO_ASSOC_ERROR =
+    "No source image is associated with this instance";
+
+  /**
+   * Create a TurboJPEG compressor instance.
+   */
+  public TJCompressor() throws TJException {
+    init();
+  }
+
+  /**
+   * Create a TurboJPEG compressor instance and associate the uncompressed
+   * source image stored in <code>srcImage</code> with the newly created
+   * instance.
+   *
+   * @param srcImage see {@link #setSourceImage} for description
+   *
+   * @param x see {@link #setSourceImage} for description
+   *
+   * @param y see {@link #setSourceImage} for description
+   *
+   * @param width see {@link #setSourceImage} for description
+   *
+   * @param pitch see {@link #setSourceImage} for description
+   *
+   * @param height see {@link #setSourceImage} for description
+   *
+   * @param pixelFormat pixel format of the source image (one of
+   * {@link TJ#PF_RGB TJ.PF_*})
+   */
+  public TJCompressor(byte srcImage, int x, int y, int width, int pitch,
+                      int height, int pixelFormat) throws TJException {
+    setSourceImage(srcImage, x, y, width, pitch, height, pixelFormat);
+  }
+
+  /**
+   * @deprecated Use
+   * {@link #TJCompressor(byte, int, int, int, int, int, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public TJCompressor(byte srcImage, int width, int pitch, int height,
+                      int pixelFormat) throws TJException {
+    setSourceImage(srcImage, width, pitch, height, pixelFormat);
+  }
+
+  /**
+   * Create a TurboJPEG compressor instance and associate the uncompressed
+   * source image stored in <code>srcImage</code> with the newly created
+   * instance.
+   *
+   * @param srcImage see
+   * {@link #setSourceImage(BufferedImage, int, int, int, int)} for description
+   *
+   * @param x see
+   * {@link #setSourceImage(BufferedImage, int, int, int, int)} for description
+   *
+   * @param y see
+   * {@link #setSourceImage(BufferedImage, int, int, int, int)} for description
+   *
+   * @param width see
+   * {@link #setSourceImage(BufferedImage, int, int, int, int)} for description
+   *
+   * @param height see
+   * {@link #setSourceImage(BufferedImage, int, int, int, int)} for description
+   */
+  public TJCompressor(BufferedImage srcImage, int x, int y, int width,
+                      int height) throws TJException {
+    setSourceImage(srcImage, x, y, width, height);
+  }
+
+  /**
+   * Associate an uncompressed RGB, grayscale, or CMYK source image with this
+   * compressor instance.
+   *
+   * @param srcImage image buffer containing RGB, grayscale, or CMYK pixels to
+   * be compressed or encoded.  This buffer is not modified.
+   *
+   * @param x x offset (in pixels) of the region in the source image from which
+   * the JPEG or YUV image should be compressed/encoded
+   *
+   * @param y y offset (in pixels) of the region in the source image from which
+   * the JPEG or YUV image should be compressed/encoded
+   *
+   * @param width width (in pixels) of the region in the source image from
+   * which the JPEG or YUV image should be compressed/encoded
+   *
+   * @param pitch bytes per line of the source image.  Normally, this should be
+   * <code>width * TJ.pixelSize(pixelFormat)</code> if the source image is
+   * unpadded, but you can use this parameter to, for instance, specify that
+   * the scanlines in the source image are padded to a 4-byte boundary or to
+   * compress/encode a JPEG or YUV image from a region of a larger source
+   * image.  You can also be clever and use this parameter to skip lines, etc.
+   * Setting this parameter to 0 is the equivalent of setting it to
+   * <code>width * TJ.pixelSize(pixelFormat)</code>.
+   *
+   * @param height height (in pixels) of the region in the source image from
+   * which the JPEG or YUV image should be compressed/encoded
+   *
+   * @param pixelFormat pixel format of the source image (one of
+   * {@link TJ#PF_RGB TJ.PF_*})
+   */
+  public void setSourceImage(byte srcImage, int x, int y, int width,
+                             int pitch, int height, int pixelFormat)
+                             throws TJException {
+    if (handle == 0) init();
+    if (srcImage == null || x < 0 || y < 0 || width < 1 || height < 1 ||
+        pitch < 0 || pixelFormat < 0 || pixelFormat >= TJ.NUMPF)
+      throw new IllegalArgumentException("Invalid argument in setSourceImage()");
+    srcBuf = srcImage;
+    srcWidth = width;
+    if (pitch == 0)
+      srcPitch = width * TJ.getPixelSize(pixelFormat);
+    else
+      srcPitch = pitch;
+    srcHeight = height;
+    srcPixelFormat = pixelFormat;
+    srcX = x;
+    srcY = y;
+    srcBufInt = null;
+    srcYUVImage = null;
+  }
+
+  /**
+   * @deprecated Use
+   * {@link #setSourceImage(byte, int, int, int, int, int, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public void setSourceImage(byte srcImage, int width, int pitch,
+                             int height, int pixelFormat) throws TJException {
+    setSourceImage(srcImage, 0, 0, width, pitch, height, pixelFormat);
+    srcX = srcY = -1;
+  }
+
+  /**
+   * Associate an uncompressed RGB or grayscale source image with this
+   * compressor instance.
+   *
+   * @param srcImage a <code>BufferedImage</code> instance containing RGB or
+   * grayscale pixels to be compressed or encoded.  This image is not modified.
+   *
+   * @param x x offset (in pixels) of the region in the source image from which
+   * the JPEG or YUV image should be compressed/encoded
+   *
+   * @param y y offset (in pixels) of the region in the source image from which
+   * the JPEG or YUV image should be compressed/encoded
+   *
+   * @param width width (in pixels) of the region in the source image from
+   * which the JPEG or YUV image should be compressed/encoded (0 = use the
+   * width of the source image)
+   *
+   * @param height height (in pixels) of the region in the source image from
+   * which the JPEG or YUV image should be compressed/encoded (0 = use the
+   * height of the source image)
+   */
+  public void setSourceImage(BufferedImage srcImage, int x, int y, int width,
+                             int height) throws TJException {
+    if (handle == 0) init();
+    if (srcImage == null || x < 0 || y < 0 || width < 0 || height < 0)
+      throw new IllegalArgumentException("Invalid argument in setSourceImage()");
+    srcX = x;
+    srcY = y;
+    srcWidth = (width == 0) ? srcImage.getWidth() : width;
+    srcHeight = (height == 0) ? srcImage.getHeight() : height;
+    if (x + width > srcImage.getWidth() || y + height > srcImage.getHeight())
+      throw new IllegalArgumentException("Compression region exceeds the bounds of the source image");
+
+    int pixelFormat;
+    boolean intPixels = false;
+    if (byteOrder == null)
+      byteOrder = ByteOrder.nativeOrder();
+    switch (srcImage.getType()) {
+    case BufferedImage.TYPE_3BYTE_BGR:
+      pixelFormat = TJ.PF_BGR;  break;
+    case BufferedImage.TYPE_4BYTE_ABGR:
+    case BufferedImage.TYPE_4BYTE_ABGR_PRE:
+      pixelFormat = TJ.PF_XBGR;  break;
+    case BufferedImage.TYPE_BYTE_GRAY:
+      pixelFormat = TJ.PF_GRAY;  break;
+    case BufferedImage.TYPE_INT_BGR:
+      if (byteOrder == ByteOrder.BIG_ENDIAN)
+        pixelFormat = TJ.PF_XBGR;
+      else
+        pixelFormat = TJ.PF_RGBX;
+      intPixels = true;  break;
+    case BufferedImage.TYPE_INT_RGB:
+    case BufferedImage.TYPE_INT_ARGB:
+    case BufferedImage.TYPE_INT_ARGB_PRE:
+      if (byteOrder == ByteOrder.BIG_ENDIAN)
+        pixelFormat = TJ.PF_XRGB;
+      else
+        pixelFormat = TJ.PF_BGRX;
+      intPixels = true;  break;
+    default:
+      throw new IllegalArgumentException("Unsupported BufferedImage format");
+    }
+    srcPixelFormat = pixelFormat;
+
+    WritableRaster wr = srcImage.getRaster();
+    if (intPixels) {
+      SinglePixelPackedSampleModel sm =
+        (SinglePixelPackedSampleModel)srcImage.getSampleModel();
+      srcStride = sm.getScanlineStride();
+      DataBufferInt db = (DataBufferInt)wr.getDataBuffer();
+      srcBufInt = db.getData();
+      srcBuf = null;
+    } else {
+      ComponentSampleModel sm =
+        (ComponentSampleModel)srcImage.getSampleModel();
+      int pixelSize = sm.getPixelStride();
+      if (pixelSize != TJ.getPixelSize(pixelFormat))
+        throw new IllegalArgumentException("Inconsistency between pixel format and pixel size in BufferedImage");
+      srcPitch = sm.getScanlineStride();
+      DataBufferByte db = (DataBufferByte)wr.getDataBuffer();
+      srcBuf = db.getData();
+      srcBufInt = null;
+    }
+    srcYUVImage = null;
+  }
+
+  /**
+   * Associate an uncompressed YUV planar source image with this compressor
+   * instance.
+   *
+   * @param srcImage YUV planar image to be compressed.  This image is not
+   * modified.
+   */
+  public void setSourceImage(YUVImage srcImage) throws TJException {
+    if (handle == 0) init();
+    if (srcImage == null)
+      throw new IllegalArgumentException("Invalid argument in setSourceImage()");
+    srcYUVImage = srcImage;
+    srcBuf = null;
+    srcBufInt = null;
+  }
+
+  /**
+   * Set the level of chrominance subsampling for subsequent compress/encode
+   * operations.  When pixels are converted from RGB to YCbCr (see
+   * {@link TJ#CS_YCbCr}) or from CMYK to YCCK (see {@link TJ#CS_YCCK}) as part
+   * of the JPEG compression process, some of the Cb and Cr (chrominance)
+   * components can be discarded or averaged together to produce a smaller
+   * image with little perceptible loss of image clarity (the human eye is more
+   * sensitive to small changes in brightness than to small changes in color.)
+   * This is called "chrominance subsampling".
+   * <p>
+   * NOTE: This method has no effect when compressing a JPEG image from a YUV
+   * planar source.  In that case, the level of chrominance subsampling in
+   * the JPEG image is determined by the source.  Furthermore, this method has
+   * no effect when encoding to a pre-allocated {@link YUVImage} instance.  In
+   * that case, the level of chrominance subsampling is determined by the
+   * destination.
+   *
+   * @param newSubsamp the level of chrominance subsampling to use in
+   * subsequent compress/encode oeprations (one of
+   * {@link TJ#SAMP_444 TJ.SAMP_*})
+   */
+  public void setSubsamp(int newSubsamp) {
+    if (newSubsamp < 0 || newSubsamp >= TJ.NUMSAMP)
+      throw new IllegalArgumentException("Invalid argument in setSubsamp()");
+    subsamp = newSubsamp;
+  }
+
+  /**
+   * Set the JPEG image quality level for subsequent compress operations.
+   *
+   * @param quality the new JPEG image quality level (1 to 100, 1 = worst,
+   * 100 = best)
+   */
+  public void setJPEGQuality(int quality) {
+    if (quality < 1 || quality > 100)
+      throw new IllegalArgumentException("Invalid argument in setJPEGQuality()");
+    jpegQuality = quality;
+  }
+
+  /**
+   * Compress the uncompressed source image associated with this compressor
+   * instance and output a JPEG image to the given destination buffer.
+   *
+   * @param dstBuf buffer that will receive the JPEG image.  Use
+   * {@link TJ#bufSize} to determine the maximum size for this buffer based on
+   * the source image's width and height and the desired level of chrominance
+   * subsampling.
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   */
+  public void compress(byte dstBuf, int flags) throws TJException {
+    if (dstBuf == null || flags < 0)
+      throw new IllegalArgumentException("Invalid argument in compress()");
+    if (srcBuf == null && srcBufInt == null && srcYUVImage == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (jpegQuality < 0)
+      throw new IllegalStateException("JPEG Quality not set");
+    if (subsamp < 0 && srcYUVImage == null)
+      throw new IllegalStateException("Subsampling level not set");
+
+    if (srcYUVImage != null)
+      compressedSize = compressFromYUV(srcYUVImage.getPlanes(),
+                                       srcYUVImage.getOffsets(),
+                                       srcYUVImage.getWidth(),
+                                       srcYUVImage.getStrides(),
+                                       srcYUVImage.getHeight(),
+                                       srcYUVImage.getSubsamp(),
+                                       dstBuf, jpegQuality, flags);
+    else if (srcBuf != null) {
+      if (srcX >= 0 && srcY >= 0)
+        compressedSize = compress(srcBuf, srcX, srcY, srcWidth, srcPitch,
+                                  srcHeight, srcPixelFormat, dstBuf, subsamp,
+                                  jpegQuality, flags);
+      else
+        compressedSize = compress(srcBuf, srcWidth, srcPitch, srcHeight,
+                                  srcPixelFormat, dstBuf, subsamp, jpegQuality,
+                                  flags);
+    } else if (srcBufInt != null) {
+      if (srcX >= 0 && srcY >= 0)
+        compressedSize = compress(srcBufInt, srcX, srcY, srcWidth, srcStride,
+                                  srcHeight, srcPixelFormat, dstBuf, subsamp,
+                                  jpegQuality, flags);
+      else
+        compressedSize = compress(srcBufInt, srcWidth, srcStride, srcHeight,
+                                  srcPixelFormat, dstBuf, subsamp, jpegQuality,
+                                  flags);
+    }
+  }
+
+  /**
+   * Compress the uncompressed source image associated with this compressor
+   * instance and return a buffer containing a JPEG image.
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   *
+   * @return a buffer containing a JPEG image.  The length of this buffer will
+   * not be equal to the size of the JPEG image.  Use {@link
+   * #getCompressedSize} to obtain the size of the JPEG image.
+   */
+  public byte compress(int flags) throws TJException {
+    byte buf;
+    if (srcYUVImage != null) {
+      buf = new byteTJ.bufSize(srcYUVImage.getWidth(),
+                                srcYUVImage.getHeight(),
+                                srcYUVImage.getSubsamp());
+    } else {
+      checkSourceImage();
+      buf = new byteTJ.bufSize(srcWidth, srcHeight, subsamp);
+    }
+    compress(buf, flags);
+    return buf;
+  }
+
+  /**
+   * @deprecated Use
+   * {@link #setSourceImage(BufferedImage, int, int, int, int)} and
+   * {@link #compress(byte, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public void compress(BufferedImage srcImage, byte dstBuf, int flags)
+                       throws TJException {
+    setSourceImage(srcImage, 0, 0, 0, 0);
+    compress(dstBuf, flags);
+  }
+
+  /**
+   * @deprecated Use
+   * {@link #setSourceImage(BufferedImage, int, int, int, int)} and
+   * {@link #compress(int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public byte compress(BufferedImage srcImage, int flags)
+                         throws TJException {
+    setSourceImage(srcImage, 0, 0, 0, 0);
+    return compress(flags);
+  }
+
+  /**
+   * Encode the uncompressed source image associated with this compressor
+   * instance into a YUV planar image and store it in the given
+   * <code>YUVImage</code> instance.   This method uses the accelerated color
+   * conversion routines in TurboJPEG's underlying codec but does not execute
+   * any of the other steps in the JPEG compression process.  Encoding
+   * CMYK source images to YUV is not supported.
+   *
+   * @param dstImage {@link YUVImage} instance that will receive the YUV planar
+   * image
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   */
+  public void encodeYUV(YUVImage dstImage, int flags) throws TJException {
+    if (dstImage == null || flags < 0)
+      throw new IllegalArgumentException("Invalid argument in encodeYUV()");
+    if (srcBuf == null && srcBufInt == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (srcYUVImage != null)
+      throw new IllegalStateException("Source image is not correct type");
+    checkSubsampling();
+    if (srcWidth != dstImage.getWidth() || srcHeight != dstImage.getHeight())
+      throw new IllegalStateException("Destination image is the wrong size");
+
+    if (srcBufInt != null) {
+      encodeYUV(srcBufInt, srcX, srcY, srcWidth, srcStride, srcHeight,
+                srcPixelFormat, dstImage.getPlanes(), dstImage.getOffsets(),
+                dstImage.getStrides(), dstImage.getSubsamp(), flags);
+    } else {
+      encodeYUV(srcBuf, srcX, srcY, srcWidth, srcPitch, srcHeight,
+                srcPixelFormat, dstImage.getPlanes(), dstImage.getOffsets(),
+                dstImage.getStrides(), dstImage.getSubsamp(), flags);
+    }
+    compressedSize = 0;
+  }
+
+  /**
+   * @deprecated Use {@link #encodeYUV(YUVImage, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public void encodeYUV(byte dstBuf, int flags) throws TJException {
+    if (dstBuf == null)
+      throw new IllegalArgumentException("Invalid argument in encodeYUV()");
+    checkSourceImage();
+    checkSubsampling();
+    YUVImage dstYUVImage = new YUVImage(dstBuf, srcWidth, 4, srcHeight,
+                                        subsamp);
+    encodeYUV(dstYUVImage, flags);
+  }
+
+  /**
+   * Encode the uncompressed source image associated with this compressor
+   * instance into a unified YUV planar image buffer and return a
+   * <code>YUVImage</code> instance containing the encoded image.  This method
+   * uses the accelerated color conversion routines in TurboJPEG's underlying
+   * codec but does not execute any of the other steps in the JPEG compression
+   * process.  Encoding CMYK source images to YUV is not supported.
+   *
+   * @param pad the width of each line in each plane of the YUV image will be
+   * padded to the nearest multiple of this number of bytes (must be a power of
+   * 2.)
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   *
+   * @return a YUV planar image.
+   */
+  public YUVImage encodeYUV(int pad, int flags) throws TJException {
+    checkSourceImage();
+    checkSubsampling();
+    if (pad < 1 || ((pad & (pad - 1)) != 0))
+      throw new IllegalStateException("Invalid argument in encodeYUV()");
+    YUVImage dstYUVImage = new YUVImage(srcWidth, pad, srcHeight, subsamp);
+    encodeYUV(dstYUVImage, flags);
+    return dstYUVImage;
+  }
+
+  /**
+   * Encode the uncompressed source image associated with this compressor
+   * instance into separate Y, U (Cb), and V (Cr) image planes and return a
+   * <code>YUVImage</code> instance containing the encoded image planes.  This
+   * method uses the accelerated color conversion routines in TurboJPEG's
+   * underlying codec but does not execute any of the other steps in the JPEG
+   * compression process.  Encoding CMYK source images to YUV is not supported.
+   *
+   * @param strides an array of integers, each specifying the number of bytes
+   * per line in the corresponding plane of the output image.  Setting the
+   * stride for any plane to 0 is the same as setting it to the component width
+   * of the plane.  If <code>strides</code> is null, then the strides for all
+   * planes will be set to their respective component widths.  You can adjust
+   * the strides in order to add an arbitrary amount of line padding to each
+   * plane.
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   *
+   * @return a YUV planar image.
+   */
+  public YUVImage encodeYUV(int strides, int flags) throws TJException {
+    checkSourceImage();
+    checkSubsampling();
+    YUVImage dstYUVImage = new YUVImage(srcWidth, strides, srcHeight, subsamp);
+    encodeYUV(dstYUVImage, flags);
+    return dstYUVImage;
+  }
+
+  /**
+   * @deprecated Use {@link #encodeYUV(int, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public byte encodeYUV(int flags) throws TJException {
+    checkSourceImage();
+    checkSubsampling();
+    YUVImage dstYUVImage = new YUVImage(srcWidth, 4, srcHeight, subsamp);
+    encodeYUV(dstYUVImage, flags);
+    return dstYUVImage.getBuf();
+  }
+
+  /**
+   * @deprecated Use
+   * {@link #setSourceImage(BufferedImage, int, int, int, int)} and
+   * {@link #encodeYUV(byte, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public void encodeYUV(BufferedImage srcImage, byte dstBuf, int flags)
+                        throws TJException {
+    setSourceImage(srcImage, 0, 0, 0, 0);
+    encodeYUV(dstBuf, flags);
+  }
+
+  /**
+   * @deprecated Use
+   * {@link #setSourceImage(BufferedImage, int, int, int, int)} and
+   * {@link #encodeYUV(int, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public byte encodeYUV(BufferedImage srcImage, int flags)
+                          throws TJException {
+    setSourceImage(srcImage, 0, 0, 0, 0);
+    return encodeYUV(flags);
+  }
+
+  /**
+   * Returns the size of the image (in bytes) generated by the most recent
+   * compress operation.
+   *
+   * @return the size of the image (in bytes) generated by the most recent
+   * compress operation.
+   */
+  public int getCompressedSize() {
+    return compressedSize;
+  }
+
+  /**
+   * Free the native structures associated with this compressor instance.
+   */
+  @Override
+  public void close() throws TJException {
+    if (handle != 0)
+      destroy();
+  }
+
+  @SuppressWarnings("checkstyle:DesignForExtension")
+  @Override
+  protected void finalize() throws Throwable {
+    try {
+      close();
+    } catch (TJException e) {
+    } finally {
+      super.finalize();
+    }
+  };
+
+  private native void init() throws TJException;
+
+  private native void destroy() throws TJException;
+
+  // JPEG size in bytes is returned
+  @SuppressWarnings("checkstyle:HiddenField")
+  @Deprecated
+  private native int compress(byte srcBuf, int width, int pitch,
+    int height, int pixelFormat, byte jpegBuf, int jpegSubsamp, int jpegQual,
+    int flags) throws TJException;
+
+  @SuppressWarnings("checkstyle:HiddenField")
+  private native int compress(byte srcBuf, int x, int y, int width,
+    int pitch, int height, int pixelFormat, byte jpegBuf, int jpegSubsamp,
+    int jpegQual, int flags) throws TJException;
+
+  @SuppressWarnings("checkstyle:HiddenField")
+  @Deprecated
+  private native int compress(int srcBuf, int width, int stride,
+    int height, int pixelFormat, byte jpegBuf, int jpegSubsamp, int jpegQual,
+    int flags) throws TJException;
+
+  @SuppressWarnings("checkstyle:HiddenField")
+  private native int compress(int srcBuf, int x, int y, int width,
+    int stride, int height, int pixelFormat, byte jpegBuf, int jpegSubsamp,
+    int jpegQual, int flags) throws TJException;
+
+  @SuppressWarnings("checkstyle:HiddenField")
+  private native int compressFromYUV(byte srcPlanes, int srcOffsets,
+    int width, int srcStrides, int height, int subsamp, byte jpegBuf,
+    int jpegQual, int flags)
+    throws TJException;
+
+  @SuppressWarnings("checkstyle:HiddenField")
+  @Deprecated
+  private native void encodeYUV(byte srcBuf, int width, int pitch,
+    int height, int pixelFormat, byte dstBuf, int subsamp, int flags)
+    throws TJException;
+
+  @SuppressWarnings("checkstyle:HiddenField")
+  private native void encodeYUV(byte srcBuf, int x, int y, int width,
+    int pitch, int height, int pixelFormat, byte dstPlanes,
+    int dstOffsets, int dstStrides, int subsamp, int flags)
+    throws TJException;
+
+  @SuppressWarnings("checkstyle:HiddenField")
+  @Deprecated
+  private native void encodeYUV(int srcBuf, int width, int stride,
+    int height, int pixelFormat, byte dstBuf, int subsamp, int flags)
+    throws TJException;
+
+  @SuppressWarnings("checkstyle:HiddenField")
+  private native void encodeYUV(int srcBuf, int x, int y, int width,
+    int srcStride, int height, int pixelFormat, byte dstPlanes,
+    int dstOffsets, int dstStrides, int subsamp, int flags)
+    throws TJException;
+
+  static {
+    TJLoader.load();
+  }
+
+  private void checkSourceImage() {
+    if (srcWidth < 1 || srcHeight < 1)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+  }
+
+  private void checkSubsampling() {
+    if (subsamp < 0)
+      throw new IllegalStateException("Subsampling level not set");
+  }
+
+  private long handle = 0;
+  private byte srcBuf = null;
+  private int srcBufInt = null;
+  private int srcWidth = 0;
+  private int srcHeight = 0;
+  private int srcX = -1;
+  private int srcY = -1;
+  private int srcPitch = 0;
+  private int srcStride = 0;
+  private int srcPixelFormat = -1;
+  private YUVImage srcYUVImage = null;
+  private int subsamp = -1;
+  private int jpegQuality = -1;
+  private int compressedSize = 0;
+  private int yuvPad = 4;
+  private ByteOrder byteOrder = null;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJCustomFilter.java Added

@@ -0,0 +1,76 @@
+/*
+ * Copyright (C)2011, 2013 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+import java.awt.*;
+import java.nio.*;
+
+/**
+ * Custom filter callback interface
+ */
+public interface TJCustomFilter {
+
+  /**
+   * A callback function that can be used to modify the DCT coefficients after
+   * they are losslessly transformed but before they are transcoded to a new
+   * JPEG image.  This allows for custom filters or other transformations to be
+   * applied in the frequency domain.
+   *
+   * @param coeffBuffer a buffer containing transformed DCT coefficients.
+   * (NOTE: this buffer is not guaranteed to be valid once the callback
+   * returns, so applications wishing to hand off the DCT coefficients to
+   * another function or library should make a copy of them within the body of
+   * the callback.)
+   *
+   * @param bufferRegion rectangle containing the width and height of
+   * <code>coeffBuffer</code> as well as its offset relative to the component
+   * plane.  TurboJPEG implementations may choose to split each component plane
+   * into multiple DCT coefficient buffers and call the callback function once
+   * for each buffer.
+   *
+   * @param planeRegion rectangle containing the width and height of the
+   * component plane to which <code>coeffBuffer</code> belongs
+   *
+   * @param componentID ID number of the component plane to which
+   * <code>coeffBuffer</code> belongs (Y, Cb, and Cr have, respectively, ID's
+   * of 0, 1, and 2 in typical JPEG images.)
+   *
+   * @param transformID ID number of the transformed image to which
+   * <code>coeffBuffer</code> belongs.  This is the same as the index of the
+   * transform in the <code>transforms</code> array that was passed to {@link
+   * TJTransformer#transform TJTransformer.transform()}.
+   *
+   * @param transform a {@link TJTransform} instance that specifies the
+   * parameters and/or cropping region for this transform
+   */
+  void customFilter(ShortBuffer coeffBuffer, Rectangle bufferRegion,
+                    Rectangle planeRegion, int componentID, int transformID,
+                    TJTransform transform)
+    throws TJException;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJDecompressor.java Added

@@ -0,0 +1,931 @@
+/*
+ * Copyright (C)2011-2015, 2018 D. R. Commander.  All Rights Reserved.
+ * Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+import java.awt.image.*;
+import java.nio.*;
+import java.io.*;
+
+/**
+ * TurboJPEG decompressor
+ */
+public class TJDecompressor implements Closeable {
+
+  private static final String NO_ASSOC_ERROR =
+    "No JPEG image is associated with this instance";
+
+  /**
+   * Create a TurboJPEG decompresssor instance.
+   */
+  public TJDecompressor() throws TJException {
+    init();
+  }
+
+  /**
+   * Create a TurboJPEG decompressor instance and associate the JPEG source
+   * image stored in <code>jpegImage</code> with the newly created instance.
+   *
+   * @param jpegImage JPEG image buffer (size of the JPEG image is assumed to
+   * be the length of the array.)  This buffer is not modified.
+   */
+  public TJDecompressor(byte jpegImage) throws TJException {
+    init();
+    setSourceImage(jpegImage, jpegImage.length);
+  }
+
+  /**
+   * Create a TurboJPEG decompressor instance and associate the JPEG source
+   * image of length <code>imageSize</code> bytes stored in
+   * <code>jpegImage</code> with the newly created instance.
+   *
+   * @param jpegImage JPEG image buffer.  This buffer is not modified.
+   *
+   * @param imageSize size of the JPEG image (in bytes)
+   */
+  public TJDecompressor(byte jpegImage, int imageSize) throws TJException {
+    init();
+    setSourceImage(jpegImage, imageSize);
+  }
+
+  /**
+   * Create a TurboJPEG decompressor instance and associate the YUV planar
+   * source image stored in <code>yuvImage</code> with the newly created
+   * instance.
+   *
+   * @param yuvImage {@link YUVImage} instance containing a YUV planar
+   * image to be decoded.  This image is not modified.
+   */
+  @SuppressWarnings("checkstyle:HiddenField")
+  public TJDecompressor(YUVImage yuvImage) throws TJException {
+    init();
+    setSourceImage(yuvImage);
+  }
+
+  /**
+   * Associate the JPEG image of length <code>imageSize</code> bytes stored in
+   * <code>jpegImage</code> with this decompressor instance.  This image will
+   * be used as the source image for subsequent decompress operations.
+   *
+   * @param jpegImage JPEG image buffer.  This buffer is not modified.
+   *
+   * @param imageSize size of the JPEG image (in bytes)
+   */
+  public void setSourceImage(byte jpegImage, int imageSize)
+                             throws TJException {
+    if (jpegImage == null || imageSize < 1)
+      throw new IllegalArgumentException("Invalid argument in setSourceImage()");
+    jpegBuf = jpegImage;
+    jpegBufSize = imageSize;
+    decompressHeader(jpegBuf, jpegBufSize);
+    yuvImage = null;
+  }
+
+  /**
+   * @deprecated Use {@link #setSourceImage(byte, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public void setJPEGImage(byte jpegImage, int imageSize)
+                           throws TJException {
+    setSourceImage(jpegImage, imageSize);
+  }
+
+  /**
+   * Associate the specified YUV planar source image with this decompressor
+   * instance.  Subsequent decompress operations will decode this image into an
+   * RGB or grayscale destination image.
+   *
+   * @param srcImage {@link YUVImage} instance containing a YUV planar image to
+   * be decoded.  This image is not modified.
+   */
+  public void setSourceImage(YUVImage srcImage) {
+    if (srcImage == null)
+      throw new IllegalArgumentException("Invalid argument in setSourceImage()");
+    yuvImage = srcImage;
+    jpegBuf = null;
+    jpegBufSize = 0;
+  }
+
+
+  /**
+   * Returns the width of the source image (JPEG or YUV) associated with this
+   * decompressor instance.
+   *
+   * @return the width of the source image (JPEG or YUV) associated with this
+   * decompressor instance.
+   */
+  public int getWidth() {
+    if (yuvImage != null)
+      return yuvImage.getWidth();
+    if (jpegWidth < 1)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return jpegWidth;
+  }
+
+  /**
+   * Returns the height of the source image (JPEG or YUV) associated with this
+   * decompressor instance.
+   *
+   * @return the height of the source image (JPEG or YUV) associated with this
+   * decompressor instance.
+   */
+  public int getHeight() {
+    if (yuvImage != null)
+      return yuvImage.getHeight();
+    if (jpegHeight < 1)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return jpegHeight;
+  }
+
+  /**
+   * Returns the level of chrominance subsampling used in the source image
+   * (JPEG or YUV) associated with this decompressor instance.  See
+   * {@link TJ#SAMP_444 TJ.SAMP_*}.
+   *
+   * @return the level of chrominance subsampling used in the source image
+   * (JPEG or YUV) associated with this decompressor instance.
+   */
+  public int getSubsamp() {
+    if (yuvImage != null)
+      return yuvImage.getSubsamp();
+    if (jpegSubsamp < 0)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (jpegSubsamp >= TJ.NUMSAMP)
+      throw new IllegalStateException("JPEG header information is invalid");
+    return jpegSubsamp;
+  }
+
+  /**
+   * Returns the colorspace used in the source image (JPEG or YUV) associated
+   * with this decompressor instance.  See {@link TJ#CS_RGB TJ.CS_*}.  If the
+   * source image is YUV, then this always returns {@link TJ#CS_YCbCr}.
+   *
+   * @return the colorspace used in the source image (JPEG or YUV) associated
+   * with this decompressor instance.
+   */
+  public int getColorspace() {
+    if (yuvImage != null)
+      return TJ.CS_YCbCr;
+    if (jpegColorspace < 0)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (jpegColorspace >= TJ.NUMCS)
+      throw new IllegalStateException("JPEG header information is invalid");
+    return jpegColorspace;
+  }
+
+  /**
+   * Returns the JPEG image buffer associated with this decompressor instance.
+   *
+   * @return the JPEG image buffer associated with this decompressor instance.
+   */
+  public byte getJPEGBuf() {
+    if (jpegBuf == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return jpegBuf;
+  }
+
+  /**
+   * Returns the size of the JPEG image (in bytes) associated with this
+   * decompressor instance.
+   *
+   * @return the size of the JPEG image (in bytes) associated with this
+   * decompressor instance.
+   */
+  public int getJPEGSize() {
+    if (jpegBufSize < 1)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return jpegBufSize;
+  }
+
+  /**
+   * Returns the width of the largest scaled-down image that the TurboJPEG
+   * decompressor can generate without exceeding the desired image width and
+   * height.
+   *
+   * @param desiredWidth desired width (in pixels) of the decompressed image.
+   * Setting this to 0 is the same as setting it to the width of the JPEG image
+   * (in other words, the width will not be considered when determining the
+   * scaled image size.)
+   *
+   * @param desiredHeight desired height (in pixels) of the decompressed image.
+   * Setting this to 0 is the same as setting it to the height of the JPEG
+   * image (in other words, the height will not be considered when determining
+   * the scaled image size.)
+   *
+   * @return the width of the largest scaled-down image that the TurboJPEG
+   * decompressor can generate without exceeding the desired image width and
+   * height.
+   */
+  public int getScaledWidth(int desiredWidth, int desiredHeight) {
+    if (jpegWidth < 1 || jpegHeight < 1)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (desiredWidth < 0 || desiredHeight < 0)
+      throw new IllegalArgumentException("Invalid argument in getScaledWidth()");
+    TJScalingFactor sf = TJ.getScalingFactors();
+    if (desiredWidth == 0)
+      desiredWidth = jpegWidth;
+    if (desiredHeight == 0)
+      desiredHeight = jpegHeight;
+    int scaledWidth = jpegWidth, scaledHeight = jpegHeight;
+    for (int i = 0; i < sf.length; i++) {
+      scaledWidth = sfi.getScaled(jpegWidth);
+      scaledHeight = sfi.getScaled(jpegHeight);
+      if (scaledWidth <= desiredWidth && scaledHeight <= desiredHeight)
+        break;
+    }
+    if (scaledWidth > desiredWidth || scaledHeight > desiredHeight)
+      throw new IllegalArgumentException("Could not scale down to desired image dimensions");
+    return scaledWidth;
+  }
+
+  /**
+   * Returns the height of the largest scaled-down image that the TurboJPEG
+   * decompressor can generate without exceeding the desired image width and
+   * height.
+   *
+   * @param desiredWidth desired width (in pixels) of the decompressed image.
+   * Setting this to 0 is the same as setting it to the width of the JPEG image
+   * (in other words, the width will not be considered when determining the
+   * scaled image size.)
+   *
+   * @param desiredHeight desired height (in pixels) of the decompressed image.
+   * Setting this to 0 is the same as setting it to the height of the JPEG
+   * image (in other words, the height will not be considered when determining
+   * the scaled image size.)
+   *
+   * @return the height of the largest scaled-down image that the TurboJPEG
+   * decompressor can generate without exceeding the desired image width and
+   * height.
+   */
+  public int getScaledHeight(int desiredWidth, int desiredHeight) {
+    if (jpegWidth < 1 || jpegHeight < 1)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (desiredWidth < 0 || desiredHeight < 0)
+      throw new IllegalArgumentException("Invalid argument in getScaledHeight()");
+    TJScalingFactor sf = TJ.getScalingFactors();
+    if (desiredWidth == 0)
+      desiredWidth = jpegWidth;
+    if (desiredHeight == 0)
+      desiredHeight = jpegHeight;
+    int scaledWidth = jpegWidth, scaledHeight = jpegHeight;
+    for (int i = 0; i < sf.length; i++) {
+      scaledWidth = sfi.getScaled(jpegWidth);
+      scaledHeight = sfi.getScaled(jpegHeight);
+      if (scaledWidth <= desiredWidth && scaledHeight <= desiredHeight)
+        break;
+    }
+    if (scaledWidth > desiredWidth || scaledHeight > desiredHeight)
+      throw new IllegalArgumentException("Could not scale down to desired image dimensions");
+    return scaledHeight;
+  }
+
+  /**
+   * Decompress the JPEG source image or decode the YUV source image associated
+   * with this decompressor instance and output a grayscale, RGB, or CMYK image
+   * to the given destination buffer.
+   * <p>
+   * NOTE: The output image is fully recoverable if this method throws a
+   * non-fatal {@link TJException} (unless
+   * {@link TJ#FLAG_STOPONWARNING TJ.FLAG_STOPONWARNING} is specified.)
+   *
+   * @param dstBuf buffer that will receive the decompressed/decoded image.
+   * If the source image is a JPEG image, then this buffer should normally be
+   * <code>pitch * scaledHeight</code> bytes in size, where
+   * <code>scaledHeight</code> can be determined by calling <code>
+   * scalingFactor.{@link TJScalingFactor#getScaled getScaled}(jpegHeight)
+   * </code> with one of the scaling factors returned from {@link
+   * TJ#getScalingFactors} or by calling {@link #getScaledHeight}.  If the
+   * source image is a YUV image, then this buffer should normally be
+   * <code>pitch * height</code> bytes in size, where <code>height</code> is
+   * the height of the YUV image.  However, the buffer may also be larger than
+   * the dimensions of the source image, in which case the <code>x</code>,
+   * <code>y</code>, and <code>pitch</code> parameters can be used to specify
+   * the region into which the source image should be decompressed/decoded.
+   *
+   * @param x x offset (in pixels) of the region in the destination image into
+   * which the source image should be decompressed/decoded
+   *
+   * @param y y offset (in pixels) of the region in the destination image into
+   * which the source image should be decompressed/decoded
+   *
+   * @param desiredWidth If the source image is a JPEG image, then this
+   * specifies the desired width (in pixels) of the decompressed image (or
+   * image region.)  If the desired destination image dimensions are different
+   * than the source image dimensions, then TurboJPEG will use scaling in the
+   * JPEG decompressor to generate the largest possible image that will fit
+   * within the desired dimensions.  Setting this to 0 is the same as setting
+   * it to the width of the JPEG image (in other words, the width will not be
+   * considered when determining the scaled image size.)  This parameter is
+   * ignored if the source image is a YUV image.
+   *
+   * @param pitch bytes per line of the destination image.  Normally, this
+   * should be set to <code>scaledWidth * TJ.pixelSize(pixelFormat)</code> if
+   * the destination image is unpadded, but you can use this to, for instance,
+   * pad each line of the destination image to a 4-byte boundary or to
+   * decompress/decode the source image into a region of a larger image.  NOTE:
+   * if the source image is a JPEG image, then <code>scaledWidth</code> can be
+   * determined by calling <code>
+   * scalingFactor.{@link TJScalingFactor#getScaled getScaled}(jpegWidth)
+   * </code> or by calling {@link #getScaledWidth}.  If the source image is a
+   * YUV image, then <code>scaledWidth</code> is the width of the YUV image.
+   * Setting this parameter to 0 is the equivalent of setting it to
+   * <code>scaledWidth * TJ.pixelSize(pixelFormat)</code>.
+   *
+   * @param desiredHeight If the source image is a JPEG image, then this
+   * specifies the desired height (in pixels) of the decompressed image (or
+   * image region.)  If the desired destination image dimensions are different
+   * than the source image dimensions, then TurboJPEG will use scaling in the
+   * JPEG decompressor to generate the largest possible image that will fit
+   * within the desired dimensions.  Setting this to 0 is the same as setting
+   * it to the height of the JPEG image (in other words, the height will not be
+   * considered when determining the scaled image size.)  This parameter is
+   * ignored if the source image is a YUV image.
+   *
+   * @param pixelFormat pixel format of the decompressed/decoded image (one of
+   * {@link TJ#PF_RGB TJ.PF_*})
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   */
+  public void decompress(byte dstBuf, int x, int y, int desiredWidth,
+                         int pitch, int desiredHeight, int pixelFormat,
+                         int flags) throws TJException {
+    if (jpegBuf == null && yuvImage == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (dstBuf == null || x < 0 || y < 0 || pitch < 0 ||
+        (yuvImage != null && (desiredWidth < 0 || desiredHeight < 0)) ||
+        pixelFormat < 0 || pixelFormat >= TJ.NUMPF || flags < 0)
+      throw new IllegalArgumentException("Invalid argument in decompress()");
+    if (yuvImage != null)
+      decodeYUV(yuvImage.getPlanes(), yuvImage.getOffsets(),
+                yuvImage.getStrides(), yuvImage.getSubsamp(), dstBuf, x, y,
+                yuvImage.getWidth(), pitch, yuvImage.getHeight(), pixelFormat,
+                flags);
+    else {
+      if (x > 0 || y > 0)
+        decompress(jpegBuf, jpegBufSize, dstBuf, x, y, desiredWidth, pitch,
+                   desiredHeight, pixelFormat, flags);
+      else
+        decompress(jpegBuf, jpegBufSize, dstBuf, desiredWidth, pitch,
+                   desiredHeight, pixelFormat, flags);
+    }
+  }
+
+  /**
+   * @deprecated Use
+   * {@link #decompress(byte, int, int, int, int, int, int, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public void decompress(byte dstBuf, int desiredWidth, int pitch,
+                         int desiredHeight, int pixelFormat, int flags)
+                         throws TJException {
+    decompress(dstBuf, 0, 0, desiredWidth, pitch, desiredHeight, pixelFormat,
+               flags);
+  }
+
+  /**
+   * Decompress the JPEG source image associated with this decompressor
+   * instance and return a buffer containing the decompressed image.
+   *
+   * @param desiredWidth see
+   * {@link #decompress(byte, int, int, int, int, int, int, int)}
+   * for description
+   *
+   * @param pitch see
+   * {@link #decompress(byte, int, int, int, int, int, int, int)}
+   * for description
+   *
+   * @param desiredHeight see
+   * {@link #decompress(byte, int, int, int, int, int, int, int)}
+   * for description
+   *
+   * @param pixelFormat pixel format of the decompressed image (one of
+   * {@link TJ#PF_RGB TJ.PF_*})
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   *
+   * @return a buffer containing the decompressed image.
+   */
+  public byte decompress(int desiredWidth, int pitch, int desiredHeight,
+                           int pixelFormat, int flags) throws TJException {
+    if (pitch < 0 ||
+        (yuvImage == null && (desiredWidth < 0 || desiredHeight < 0)) ||
+        pixelFormat < 0 || pixelFormat >= TJ.NUMPF || flags < 0)
+      throw new IllegalArgumentException("Invalid argument in decompress()");
+    int pixelSize = TJ.getPixelSize(pixelFormat);
+    int scaledWidth = getScaledWidth(desiredWidth, desiredHeight);
+    int scaledHeight = getScaledHeight(desiredWidth, desiredHeight);
+    if (pitch == 0)
+      pitch = scaledWidth * pixelSize;
+    byte buf = new bytepitch * scaledHeight;
+    decompress(buf, desiredWidth, pitch, desiredHeight, pixelFormat, flags);
+    return buf;
+  }
+
+  /**
+   * Decompress the JPEG source image associated with this decompressor
+   * instance into a YUV planar image and store it in the given
+   * <code>YUVImage</code> instance.  This method performs JPEG decompression
+   * but leaves out the color conversion step, so a planar YUV image is
+   * generated instead of an RGB or grayscale image.  This method cannot be
+   * used to decompress JPEG source images with the CMYK or YCCK colorspace.
+   * <p>
+   * NOTE: The YUV planar output image is fully recoverable if this method
+   * throws a non-fatal {@link TJException} (unless
+   * {@link TJ#FLAG_STOPONWARNING TJ.FLAG_STOPONWARNING} is specified.)
+   *
+   * @param dstImage {@link YUVImage} instance that will receive the YUV planar
+   * image.  The level of subsampling specified in this <code>YUVImage</code>
+   * instance must match that of the JPEG image, and the width and height
+   * specified in the <code>YUVImage</code> instance must match one of the
+   * scaled image sizes that TurboJPEG is capable of generating from the JPEG
+   * source image.
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   */
+  public void decompressToYUV(YUVImage dstImage, int flags)
+                              throws TJException {
+    if (jpegBuf == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (dstImage == null || flags < 0)
+      throw new IllegalArgumentException("Invalid argument in decompressToYUV()");
+    int scaledWidth = getScaledWidth(dstImage.getWidth(),
+                                     dstImage.getHeight());
+    int scaledHeight = getScaledHeight(dstImage.getWidth(),
+                                       dstImage.getHeight());
+    if (scaledWidth != dstImage.getWidth() ||
+        scaledHeight != dstImage.getHeight())
+      throw new IllegalArgumentException("YUVImage dimensions do not match one of the scaled image sizes that TurboJPEG is capable of generating.");
+    if (jpegSubsamp != dstImage.getSubsamp())
+      throw new IllegalArgumentException("YUVImage subsampling level does not match that of the JPEG image");
+
+    decompressToYUV(jpegBuf, jpegBufSize, dstImage.getPlanes(),
+                    dstImage.getOffsets(), dstImage.getWidth(),
+                    dstImage.getStrides(), dstImage.getHeight(), flags);
+  }
+
+  /**
+   * @deprecated Use {@link #decompressToYUV(YUVImage, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public void decompressToYUV(byte dstBuf, int flags) throws TJException {
+    YUVImage dstYUVImage = new YUVImage(dstBuf, jpegWidth, 4, jpegHeight,
+                                        jpegSubsamp);
+    decompressToYUV(dstYUVImage, flags);
+  }
+
+  /**
+   * Decompress the JPEG source image associated with this decompressor
+   * instance into a set of Y, U (Cb), and V (Cr) image planes and return a
+   * <code>YUVImage</code> instance containing the decompressed image planes.
+   * This method performs JPEG decompression but leaves out the color
+   * conversion step, so a planar YUV image is generated instead of an RGB or
+   * grayscale image.  This method cannot be used to decompress JPEG source
+   * images with the CMYK or YCCK colorspace.
+   *
+   * @param desiredWidth desired width (in pixels) of the YUV image.  If the
+   * desired image dimensions are different than the dimensions of the JPEG
+   * image being decompressed, then TurboJPEG will use scaling in the JPEG
+   * decompressor to generate the largest possible image that will fit within
+   * the desired dimensions.  Setting this to 0 is the same as setting it to
+   * the width of the JPEG image (in other words, the width will not be
+   * considered when determining the scaled image size.)
+   *
+   * @param strides an array of integers, each specifying the number of bytes
+   * per line in the corresponding plane of the output image.  Setting the
+   * stride for any plane to 0 is the same as setting it to the scaled
+   * component width of the plane.  If <tt>strides</tt> is NULL, then the
+   * strides for all planes will be set to their respective scaled component
+   * widths.  You can adjust the strides in order to add an arbitrary amount of
+   * line padding to each plane.
+   *
+   * @param desiredHeight desired height (in pixels) of the YUV image.  If the
+   * desired image dimensions are different than the dimensions of the JPEG
+   * image being decompressed, then TurboJPEG will use scaling in the JPEG
+   * decompressor to generate the largest possible image that will fit within
+   * the desired dimensions.  Setting this to 0 is the same as setting it to
+   * the height of the JPEG image (in other words, the height will not be
+   * considered when determining the scaled image size.)
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   *
+   * @return a YUV planar image.
+   */
+  public YUVImage decompressToYUV(int desiredWidth, int strides,
+                                  int desiredHeight,
+                                  int flags) throws TJException {
+    if (flags < 0)
+      throw new IllegalArgumentException("Invalid argument in decompressToYUV()");
+    if (jpegWidth < 1 || jpegHeight < 1 || jpegSubsamp < 0)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (jpegSubsamp >= TJ.NUMSAMP)
+      throw new IllegalStateException("JPEG header information is invalid");
+    if (yuvImage != null)
+      throw new IllegalStateException("Source image is the wrong type");
+
+    int scaledWidth = getScaledWidth(desiredWidth, desiredHeight);
+    int scaledHeight = getScaledHeight(desiredWidth, desiredHeight);
+    YUVImage dstYUVImage = new YUVImage(scaledWidth, null, scaledHeight,
+                                        jpegSubsamp);
+    decompressToYUV(dstYUVImage, flags);
+    return dstYUVImage;
+  }
+
+  /**
+   * Decompress the JPEG source image associated with this decompressor
+   * instance into a unified YUV planar image buffer and return a
+   * <code>YUVImage</code> instance containing the decompressed image.  This
+   * method performs JPEG decompression but leaves out the color conversion
+   * step, so a planar YUV image is generated instead of an RGB or grayscale
+   * image.  This method cannot be used to decompress JPEG source images with
+   * the CMYK or YCCK colorspace.
+   *
+   * @param desiredWidth desired width (in pixels) of the YUV image.  If the
+   * desired image dimensions are different than the dimensions of the JPEG
+   * image being decompressed, then TurboJPEG will use scaling in the JPEG
+   * decompressor to generate the largest possible image that will fit within
+   * the desired dimensions.  Setting this to 0 is the same as setting it to
+   * the width of the JPEG image (in other words, the width will not be
+   * considered when determining the scaled image size.)
+   *
+   * @param pad the width of each line in each plane of the YUV image will be
+   * padded to the nearest multiple of this number of bytes (must be a power of
+   * 2.)
+   *
+   * @param desiredHeight desired height (in pixels) of the YUV image.  If the
+   * desired image dimensions are different than the dimensions of the JPEG
+   * image being decompressed, then TurboJPEG will use scaling in the JPEG
+   * decompressor to generate the largest possible image that will fit within
+   * the desired dimensions.  Setting this to 0 is the same as setting it to
+   * the height of the JPEG image (in other words, the height will not be
+   * considered when determining the scaled image size.)
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   *
+   * @return a YUV planar image.
+   */
+  public YUVImage decompressToYUV(int desiredWidth, int pad, int desiredHeight,
+                                  int flags) throws TJException {
+    if (flags < 0)
+      throw new IllegalArgumentException("Invalid argument in decompressToYUV()");
+    if (jpegWidth < 1 || jpegHeight < 1 || jpegSubsamp < 0)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (jpegSubsamp >= TJ.NUMSAMP)
+      throw new IllegalStateException("JPEG header information is invalid");
+    if (yuvImage != null)
+      throw new IllegalStateException("Source image is the wrong type");
+
+    int scaledWidth = getScaledWidth(desiredWidth, desiredHeight);
+    int scaledHeight = getScaledHeight(desiredWidth, desiredHeight);
+    YUVImage dstYUVImage = new YUVImage(scaledWidth, pad, scaledHeight,
+                                        jpegSubsamp);
+    decompressToYUV(dstYUVImage, flags);
+    return dstYUVImage;
+  }
+
+  /**
+   * @deprecated Use {@link #decompressToYUV(int, int, int, int)} instead.
+   */
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  @Deprecated
+  public byte decompressToYUV(int flags) throws TJException {
+    YUVImage dstYUVImage = new YUVImage(jpegWidth, 4, jpegHeight, jpegSubsamp);
+    decompressToYUV(dstYUVImage, flags);
+    return dstYUVImage.getBuf();
+  }
+
+  /**
+   * Decompress the JPEG source image or decode the YUV source image associated
+   * with this decompressor instance and output a grayscale, RGB, or CMYK image
+   * to the given destination buffer.
+   * <p>
+   * NOTE: The output image is fully recoverable if this method throws a
+   * non-fatal {@link TJException} (unless
+   * {@link TJ#FLAG_STOPONWARNING TJ.FLAG_STOPONWARNING} is specified.)
+   *
+   * @param dstBuf buffer that will receive the decompressed/decoded image.
+   * If the source image is a JPEG image, then this buffer should normally be
+   * <code>stride * scaledHeight</code> pixels in size, where
+   * <code>scaledHeight</code> can be determined by calling <code>
+   * scalingFactor.{@link TJScalingFactor#getScaled getScaled}(jpegHeight)
+   * </code> with one of the scaling factors returned from {@link
+   * TJ#getScalingFactors} or by calling {@link #getScaledHeight}.  If the
+   * source image is a YUV image, then this buffer should normally be
+   * <code>stride * height</code> pixels in size, where <code>height</code> is
+   * the height of the YUV image.  However, the buffer may also be larger than
+   * the dimensions of the JPEG image, in which case the <code>x</code>,
+   * <code>y</code>, and <code>stride</code> parameters can be used to specify
+   * the region into which the source image should be decompressed.
+   *
+   * @param x x offset (in pixels) of the region in the destination image into
+   * which the source image should be decompressed/decoded
+   *
+   * @param y y offset (in pixels) of the region in the destination image into
+   * which the source image should be decompressed/decoded
+   *
+   * @param desiredWidth If the source image is a JPEG image, then this
+   * specifies the desired width (in pixels) of the decompressed image (or
+   * image region.)  If the desired destination image dimensions are different
+   * than the source image dimensions, then TurboJPEG will use scaling in the
+   * JPEG decompressor to generate the largest possible image that will fit
+   * within the desired dimensions.  Setting this to 0 is the same as setting
+   * it to the width of the JPEG image (in other words, the width will not be
+   * considered when determining the scaled image size.)  This parameter is
+   * ignored if the source image is a YUV image.
+   *
+   * @param stride pixels per line of the destination image.  Normally, this
+   * should be set to <code>scaledWidth</code>, but you can use this to, for
+   * instance, decompress the JPEG image into a region of a larger image.
+   * NOTE: if the source image is a JPEG image, then <code>scaledWidth</code>
+   * can be determined by calling <code>
+   * scalingFactor.{@link TJScalingFactor#getScaled getScaled}(jpegWidth)
+   * </code> or by calling {@link #getScaledWidth}.  If the source image is a
+   * YUV image, then <code>scaledWidth</code> is the width of the YUV image.
+   * Setting this parameter to 0 is the equivalent of setting it to
+   * <code>scaledWidth</code>.
+   *
+   * @param desiredHeight If the source image is a JPEG image, then this
+   * specifies the desired height (in pixels) of the decompressed image (or
+   * image region.)  If the desired destination image dimensions are different
+   * than the source image dimensions, then TurboJPEG will use scaling in the
+   * JPEG decompressor to generate the largest possible image that will fit
+   * within the desired dimensions.  Setting this to 0 is the same as setting
+   * it to the height of the JPEG image (in other words, the height will not be
+   * considered when determining the scaled image size.)  This parameter is
+   * ignored if the source image is a YUV image.
+   *
+   * @param pixelFormat pixel format of the decompressed image (one of
+   * {@link TJ#PF_RGB TJ.PF_*})
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   */
+  public void decompress(int dstBuf, int x, int y, int desiredWidth,
+                         int stride, int desiredHeight, int pixelFormat,
+                         int flags) throws TJException {
+    if (jpegBuf == null && yuvImage == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (dstBuf == null || x < 0 || y < 0 || stride < 0 ||
+        (yuvImage != null && (desiredWidth < 0 || desiredHeight < 0)) ||
+        pixelFormat < 0 || pixelFormat >= TJ.NUMPF || flags < 0)
+      throw new IllegalArgumentException("Invalid argument in decompress()");
+    if (yuvImage != null)
+      decodeYUV(yuvImage.getPlanes(), yuvImage.getOffsets(),
+                yuvImage.getStrides(), yuvImage.getSubsamp(), dstBuf, x, y,
+                yuvImage.getWidth(), stride, yuvImage.getHeight(), pixelFormat,
+                flags);
+    else
+      decompress(jpegBuf, jpegBufSize, dstBuf, x, y, desiredWidth, stride,
+                 desiredHeight, pixelFormat, flags);
+  }
+
+  /**
+   * Decompress the JPEG source image or decode the YUV source image associated
+   * with this decompressor instance and output a decompressed/decoded image to
+   * the given <code>BufferedImage</code> instance.
+   * <p>
+   * NOTE: The output image is fully recoverable if this method throws a
+   * non-fatal {@link TJException} (unless
+   * {@link TJ#FLAG_STOPONWARNING TJ.FLAG_STOPONWARNING} is specified.)
+   *
+   * @param dstImage a <code>BufferedImage</code> instance that will receive
+   * the decompressed/decoded image.  If the source image is a JPEG image, then
+   * the width and height of the <code>BufferedImage</code> instance must match
+   * one of the scaled image sizes that TurboJPEG is capable of generating from
+   * the JPEG image.  If the source image is a YUV image, then the width and
+   * height of the <code>BufferedImage</code> instance must match the width and
+   * height of the YUV image.
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   */
+  public void decompress(BufferedImage dstImage, int flags)
+                         throws TJException {
+    if (dstImage == null || flags < 0)
+      throw new IllegalArgumentException("Invalid argument in decompress()");
+    int desiredWidth = dstImage.getWidth();
+    int desiredHeight = dstImage.getHeight();
+    int scaledWidth, scaledHeight;
+
+    if (yuvImage != null) {
+      if (desiredWidth != yuvImage.getWidth() ||
+          desiredHeight != yuvImage.getHeight())
+        throw new IllegalArgumentException("BufferedImage dimensions do not match the dimensions of the source image.");
+      scaledWidth = yuvImage.getWidth();
+      scaledHeight = yuvImage.getHeight();
+    } else {
+      scaledWidth = getScaledWidth(desiredWidth, desiredHeight);
+      scaledHeight = getScaledHeight(desiredWidth, desiredHeight);
+      if (scaledWidth != desiredWidth || scaledHeight != desiredHeight)
+        throw new IllegalArgumentException("BufferedImage dimensions do not match one of the scaled image sizes that TurboJPEG is capable of generating.");
+    }
+    int pixelFormat;  boolean intPixels = false;
+    if (byteOrder == null)
+      byteOrder = ByteOrder.nativeOrder();
+    switch (dstImage.getType()) {
+    case BufferedImage.TYPE_3BYTE_BGR:
+      pixelFormat = TJ.PF_BGR;  break;
+    case BufferedImage.TYPE_4BYTE_ABGR:
+    case BufferedImage.TYPE_4BYTE_ABGR_PRE:
+      pixelFormat = TJ.PF_XBGR;  break;
+    case BufferedImage.TYPE_BYTE_GRAY:
+      pixelFormat = TJ.PF_GRAY;  break;
+    case BufferedImage.TYPE_INT_BGR:
+      if (byteOrder == ByteOrder.BIG_ENDIAN)
+        pixelFormat = TJ.PF_XBGR;
+      else
+        pixelFormat = TJ.PF_RGBX;
+      intPixels = true;  break;
+    case BufferedImage.TYPE_INT_RGB:
+      if (byteOrder == ByteOrder.BIG_ENDIAN)
+        pixelFormat = TJ.PF_XRGB;
+      else
+        pixelFormat = TJ.PF_BGRX;
+      intPixels = true;  break;
+    case BufferedImage.TYPE_INT_ARGB:
+    case BufferedImage.TYPE_INT_ARGB_PRE:
+      if (byteOrder == ByteOrder.BIG_ENDIAN)
+        pixelFormat = TJ.PF_ARGB;
+      else
+        pixelFormat = TJ.PF_BGRA;
+      intPixels = true;  break;
+    default:
+      throw new IllegalArgumentException("Unsupported BufferedImage format");
+    }
+    WritableRaster wr = dstImage.getRaster();
+    if (intPixels) {
+      SinglePixelPackedSampleModel sm =
+        (SinglePixelPackedSampleModel)dstImage.getSampleModel();
+      int stride = sm.getScanlineStride();
+      DataBufferInt db = (DataBufferInt)wr.getDataBuffer();
+      int buf = db.getData();
+      if (yuvImage != null)
+        decodeYUV(yuvImage.getPlanes(), yuvImage.getOffsets(),
+                  yuvImage.getStrides(), yuvImage.getSubsamp(), buf, 0, 0,
+                  yuvImage.getWidth(), stride, yuvImage.getHeight(),
+                  pixelFormat, flags);
+      else {
+        if (jpegBuf == null)
+          throw new IllegalStateException(NO_ASSOC_ERROR);
+        decompress(jpegBuf, jpegBufSize, buf, 0, 0, scaledWidth, stride,
+                   scaledHeight, pixelFormat, flags);
+      }
+    } else {
+      ComponentSampleModel sm =
+        (ComponentSampleModel)dstImage.getSampleModel();
+      int pixelSize = sm.getPixelStride();
+      if (pixelSize != TJ.getPixelSize(pixelFormat))
+        throw new IllegalArgumentException("Inconsistency between pixel format and pixel size in BufferedImage");
+      int pitch = sm.getScanlineStride();
+      DataBufferByte db = (DataBufferByte)wr.getDataBuffer();
+      byte buf = db.getData();
+      decompress(buf, 0, 0, scaledWidth, pitch, scaledHeight, pixelFormat,
+                 flags);
+    }
+  }
+
+  /**
+   * Decompress the JPEG source image or decode the YUV source image associated
+   * with this decompressor instance and return a <code>BufferedImage</code>
+   * instance containing the decompressed/decoded image.
+   *
+   * @param desiredWidth see
+   * {@link #decompress(byte, int, int, int, int, int, int, int)} for
+   * description
+   *
+   * @param desiredHeight see
+   * {@link #decompress(byte, int, int, int, int, int, int, int)} for
+   * description
+   *
+   * @param bufferedImageType the image type of the <code>BufferedImage</code>
+   * instance that will be created (for instance,
+   * <code>BufferedImage.TYPE_INT_RGB</code>)
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   *
+   * @return a <code>BufferedImage</code> instance containing the
+   * decompressed/decoded image.
+   */
+  public BufferedImage decompress(int desiredWidth, int desiredHeight,
+                                  int bufferedImageType, int flags)
+                                  throws TJException {
+    if ((yuvImage == null && (desiredWidth < 0 || desiredHeight < 0)) ||
+        flags < 0)
+      throw new IllegalArgumentException("Invalid argument in decompress()");
+    int scaledWidth = getScaledWidth(desiredWidth, desiredHeight);
+    int scaledHeight = getScaledHeight(desiredWidth, desiredHeight);
+    BufferedImage img = new BufferedImage(scaledWidth, scaledHeight,
+                                          bufferedImageType);
+    decompress(img, flags);
+    return img;
+  }
+
+  /**
+   * Free the native structures associated with this decompressor instance.
+   */
+  @Override
+  public void close() throws TJException {
+    if (handle != 0)
+      destroy();
+  }
+
+  @SuppressWarnings("checkstyle:DesignForExtension")
+  @Override
+  protected void finalize() throws Throwable {
+    try {
+      close();
+    } catch (TJException e) {
+    } finally {
+      super.finalize();
+    }
+  };
+
+  private native void init() throws TJException;
+
+  private native void destroy() throws TJException;
+
+  private native void decompressHeader(byte srcBuf, int size)
+    throws TJException;
+
+  @Deprecated
+  private native void decompress(byte srcBuf, int size, byte dstBuf,
+    int desiredWidth, int pitch, int desiredHeight, int pixelFormat, int flags)
+    throws TJException;
+
+  private native void decompress(byte srcBuf, int size, byte dstBuf, int x,
+    int y, int desiredWidth, int pitch, int desiredHeight, int pixelFormat,
+    int flags) throws TJException;
+
+  @Deprecated
+  private native void decompress(byte srcBuf, int size, int dstBuf,
+    int desiredWidth, int stride, int desiredHeight, int pixelFormat,
+    int flags) throws TJException;
+
+  private native void decompress(byte srcBuf, int size, int dstBuf, int x,
+    int y, int desiredWidth, int stride, int desiredHeight, int pixelFormat,
+    int flags) throws TJException;
+
+  @Deprecated
+  private native void decompressToYUV(byte srcBuf, int size, byte dstBuf,
+    int flags) throws TJException;
+
+  private native void decompressToYUV(byte srcBuf, int size,
+    byte dstPlanes, int dstOffsets, int desiredWidth, int dstStrides,
+    int desiredheight, int flags) throws TJException;
+
+  private native void decodeYUV(byte srcPlanes, int srcOffsets,
+    int srcStrides, int subsamp, byte dstBuf, int x, int y, int width,
+    int pitch, int height, int pixelFormat, int flags) throws TJException;
+
+  private native void decodeYUV(byte srcPlanes, int srcOffsets,
+    int srcStrides, int subsamp, int dstBuf, int x, int y, int width,
+    int stride, int height, int pixelFormat, int flags) throws TJException;
+
+  static {
+    TJLoader.load();
+  }
+
+  protected long handle = 0;
+  protected byte jpegBuf = null;
+  protected int jpegBufSize = 0;
+  protected YUVImage yuvImage = null;
+  protected int jpegWidth = 0;
+  protected int jpegHeight = 0;
+  protected int jpegSubsamp = -1;
+  protected int jpegColorspace = -1;
+  private ByteOrder byteOrder = null;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJException.java Added

@@ -0,0 +1,78 @@
+/*
+ * Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.
+ * Copyright (C)2017-2018 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+import java.io.IOException;
+
+@SuppressWarnings("checkstyle:JavadocType")
+public class TJException extends IOException {
+
+  private static final long serialVersionUID = 1L;
+
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  public TJException() {
+    super();
+  }
+
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  public TJException(String message, Throwable cause) {
+    super(message, cause);
+  }
+
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  public TJException(String message) {
+    super(message);
+  }
+
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  public TJException(String message, int code) {
+    super(message);
+    if (errorCode >= 0 && errorCode < TJ.NUMERR)
+      errorCode = code;
+  }
+
+  @SuppressWarnings("checkstyle:JavadocMethod")
+  public TJException(Throwable cause) {
+    super(cause);
+  }
+
+  /**
+   * Returns a code (one of {@link TJ TJ.ERR_*}) indicating the severity of the
+   * last error.
+   *
+   * @return a code (one of {@link TJ TJ.ERR_*}) indicating the severity of the
+   * last error.
+   */
+  public int getErrorCode() {
+    return errorCode;
+  }
+
+  private int errorCode = TJ.ERR_FATAL;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJLoader-unix.java.in Added

@@ -0,0 +1,59 @@
+/*
+ * Copyright (C)2011-2013, 2016, 2020 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+final class TJLoader {
+  static void load() {
+    try {
+      System.loadLibrary("turbojpeg");
+    } catch (java.lang.UnsatisfiedLinkError e) {
+      String os = System.getProperty("os.name").toLowerCase();
+      if (os.indexOf("mac") >= 0) {
+        try {
+          System.load("@CMAKE_INSTALL_FULL_LIBDIR@/libturbojpeg.dylib");
+        } catch (java.lang.UnsatisfiedLinkError e2) {
+          System.load("/usr/lib/libturbojpeg.dylib");
+        }
+      } else {
+        try {
+          System.load("@CMAKE_INSTALL_FULL_LIBDIR@/libturbojpeg.so");
+        } catch (java.lang.UnsatisfiedLinkError e3) {
+          String libdir = "@CMAKE_INSTALL_FULL_LIBDIR@";
+          if (libdir.equals("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib64")) {
+            System.load("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib32/libturbojpeg.so");
+          } else if (libdir.equals("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib32")) {
+            System.load("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib64/libturbojpeg.so");
+          } else {
+            throw e3;
+          }
+        }
+      }
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJLoader-win.java.in Added

@@ -0,0 +1,35 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+final class TJLoader {
+  static void load() {
+    System.loadLibrary("@TURBOJPEG_DLL_NAME@");
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJScalingFactor.java Added

@@ -0,0 +1,115 @@
+/*
+ * Copyright (C)2011, 2018 D. R. Commander.  All Rights Reserved.
+ * Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+/**
+ * Fractional scaling factor
+ */
+public class TJScalingFactor {
+
+  /**
+   * Create a TurboJPEG scaling factor instance.
+   *
+   * @param num numerator
+   * @param denom denominator
+   */
+  @SuppressWarnings("checkstyle:HiddenField")
+  public TJScalingFactor(int num, int denom) {
+    if (num < 1 || denom < 1)
+      throw new IllegalArgumentException("Numerator and denominator must be >= 1");
+    this.num = num;
+    this.denom = denom;
+  }
+
+  /**
+   * Returns numerator
+   *
+   * @return numerator
+   */
+  public int getNum() {
+    return num;
+  }
+
+  /**
+   * Returns denominator
+   *
+   * @return denominator
+   */
+  public int getDenom() {
+    return denom;
+  }
+
+  /**
+   * Returns the scaled value of <code>dimension</code>.  This function
+   * performs the integer equivalent of
+   * <code>ceil(dimension * scalingFactor)</code>.
+   *
+   * @param dimension width or height to multiply by this scaling factor
+   *
+   * @return the scaled value of <code>dimension</code>.
+   */
+  public int getScaled(int dimension) {
+    return (dimension * num + denom - 1) / denom;
+  }
+
+  /**
+   * Returns true or false, depending on whether this instance and
+   * <code>other</code> have the same numerator and denominator.
+   *
+   * @param other the scaling factor against which to compare this one
+   *
+   * @return true or false, depending on whether this instance and
+   * <code>other</code> have the same numerator and denominator.
+   */
+  public boolean equals(TJScalingFactor other) {
+    return this.num == other.num && this.denom == other.denom;
+  }
+
+  /**
+   * Returns true or false, depending on whether this instance is equal to
+   * 1/1.
+   *
+   * @return true or false, depending on whether this instance is equal to
+   * 1/1.
+   */
+  public boolean isOne() {
+    return num == 1 && denom == 1;
+  }
+
+  /**
+   * Numerator
+   */
+  private int num = 1;
+
+  /**
+   * Denominator
+   */
+  private int denom = 1;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJTransform.java Added

@@ -0,0 +1,227 @@
+/*
+ * Copyright (C)2011, 2013, 2018 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+import java.awt.*;
+
+/**
+ * Lossless transform parameters
+ */
+public class TJTransform extends Rectangle {
+
+  private static final long serialVersionUID = -127367705761430371L;
+
+  /**
+   * The number of lossless transform operations
+   */
+  public static final int NUMOP         = 8;
+  /**
+   * Do not transform the position of the image pixels.
+   */
+  public static final int OP_NONE       = 0;
+  /**
+   * Flip (mirror) image horizontally.  This transform is imperfect if there
+   * are any partial MCU blocks on the right edge.
+   * @see #OPT_PERFECT
+   */
+  public static final int OP_HFLIP      = 1;
+  /**
+   * Flip (mirror) image vertically.  This transform is imperfect if there are
+   * any partial MCU blocks on the bottom edge.
+   * @see #OPT_PERFECT
+   */
+  public static final int OP_VFLIP      = 2;
+  /**
+   * Transpose image (flip/mirror along upper left to lower right axis).  This
+   * transform is always perfect.
+   * @see #OPT_PERFECT
+   */
+  public static final int OP_TRANSPOSE  = 3;
+  /**
+   * Transverse transpose image (flip/mirror along upper right to lower left
+   * axis).  This transform is imperfect if there are any partial MCU blocks in
+   * the image.
+   * @see #OPT_PERFECT
+   */
+  public static final int OP_TRANSVERSE = 4;
+  /**
+   * Rotate image clockwise by 90 degrees.  This transform is imperfect if
+   * there are any partial MCU blocks on the bottom edge.
+   * @see #OPT_PERFECT
+   */
+  public static final int OP_ROT90      = 5;
+  /**
+   * Rotate image 180 degrees.  This transform is imperfect if there are any
+   * partial MCU blocks in the image.
+   * @see #OPT_PERFECT
+   */
+  public static final int OP_ROT180     = 6;
+  /**
+   * Rotate image counter-clockwise by 90 degrees.  This transform is imperfect
+   * if there are any partial MCU blocks on the right edge.
+   * @see #OPT_PERFECT
+   */
+  public static final int OP_ROT270     = 7;
+
+
+  /**
+   * This option will cause {@link TJTransformer#transform
+   * TJTransformer.transform()} to throw an exception if the transform is not
+   * perfect.  Lossless transforms operate on MCU blocks, whose size depends on
+   * the level of chrominance subsampling used.  If the image's width or height
+   * is not evenly divisible by the MCU block size (see {@link TJ#getMCUWidth}
+   * and {@link TJ#getMCUHeight}), then there will be partial MCU blocks on the
+   * right and/or bottom edges.   It is not possible to move these partial MCU
+   * blocks to the top or left of the image, so any transform that would
+   * require that is "imperfect."  If this option is not specified, then any
+   * partial MCU blocks that cannot be transformed will be left in place, which
+   * will create odd-looking strips on the right or bottom edge of the image.
+   */
+  public static final int OPT_PERFECT     = 1;
+  /**
+   * This option will discard any partial MCU blocks that cannot be
+   * transformed.
+   */
+  public static final int OPT_TRIM        = 2;
+  /**
+   * This option will enable lossless cropping.
+   */
+  public static final int OPT_CROP        = 4;
+  /**
+   * This option will discard the color data in the input image and produce
+   * a grayscale output image.
+   */
+  public static final int OPT_GRAY        = 8;
+  /**
+   * This option will prevent {@link TJTransformer#transform
+   * TJTransformer.transform()} from outputting a JPEG image for this
+   * particular transform.  This can be used in conjunction with a custom
+   * filter to capture the transformed DCT coefficients without transcoding
+   * them.
+   */
+  public static final int OPT_NOOUTPUT    = 16;
+  /**
+   * This option will enable progressive entropy coding in the output image
+   * generated by this particular transform.  Progressive entropy coding will
+   * generally improve compression relative to baseline entropy coding (the
+   * default), but it will reduce compression and decompression performance
+   * considerably.
+   */
+  public static final int OPT_PROGRESSIVE = 32;
+  /**
+   * This option will prevent {@link TJTransformer#transform
+   * TJTransformer.transform()} from copying any extra markers (including EXIF
+   * and ICC profile data) from the source image to the output image.
+   */
+  public static final int OPT_COPYNONE    = 64;
+
+
+  /**
+   * Create a new lossless transform instance.
+   */
+  public TJTransform() {
+  }
+
+  /**
+   * Create a new lossless transform instance with the given parameters.
+   *
+   * @param x the left boundary of the cropping region.  This must be evenly
+   * divisible by the MCU block width (see {@link TJ#getMCUWidth})
+   *
+   * @param y the upper boundary of the cropping region.  This must be evenly
+   * divisible by the MCU block height (see {@link TJ#getMCUHeight})
+   *
+   * @param w the width of the cropping region.  Setting this to 0 is the
+   * equivalent of setting it to (width of the source JPEG image -
+   * <code>x</code>).
+   *
+   * @param h the height of the cropping region.  Setting this to 0 is the
+   * equivalent of setting it to (height of the source JPEG image -
+   * <code>y</code>).
+   *
+   * @param op one of the transform operations (<code>OP_*</code>)
+   *
+   * @param options the bitwise OR of one or more of the transform options
+   * (<code>OPT_*</code>)
+   *
+   * @param cf an instance of an object that implements the {@link
+   * TJCustomFilter} interface, or null if no custom filter is needed
+   */
+  @SuppressWarnings("checkstyle:HiddenField")
+  public TJTransform(int x, int y, int w, int h, int op, int options,
+                     TJCustomFilter cf) {
+    super(x, y, w, h);
+    this.op = op;
+    this.options = options;
+    this.cf = cf;
+  }
+
+  /**
+   * Create a new lossless transform instance with the given parameters.
+   *
+   * @param r a <code>Rectangle</code> instance that specifies the cropping
+   * region.  See {@link
+   * #TJTransform(int, int, int, int, int, int, TJCustomFilter)} for more
+   * detail.
+   *
+   * @param op one of the transform operations (<code>OP_*</code>)
+   *
+   * @param options the bitwise OR of one or more of the transform options
+   * (<code>OPT_*</code>)
+   *
+   * @param cf an instance of an object that implements the {@link
+   * TJCustomFilter} interface, or null if no custom filter is needed
+   */
+  @SuppressWarnings("checkstyle:HiddenField")
+  public TJTransform(Rectangle r, int op, int options,
+                     TJCustomFilter cf) {
+    super(r);
+    this.op = op;
+    this.options = options;
+    this.cf = cf;
+  }
+
+  /**
+   * Transform operation (one of <code>OP_*</code>)
+   */
+  @SuppressWarnings("checkstyle:VisibilityModifier")
+  public int op = 0;
+
+  /**
+   * Transform options (bitwise OR of one or more of <code>OPT_*</code>)
+   */
+  @SuppressWarnings("checkstyle:VisibilityModifier")
+  public int options = 0;
+
+  /**
+   * Custom filter instance
+   */
+  @SuppressWarnings("checkstyle:VisibilityModifier")
+  public TJCustomFilter cf = null;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/TJTransformer.java Added

@@ -0,0 +1,163 @@
+/*
+ * Copyright (C)2011, 2013-2015 D. R. Commander.  All Rights Reserved.
+ * Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+/**
+ * TurboJPEG lossless transformer
+ */
+public class TJTransformer extends TJDecompressor {
+
+  /**
+   * Create a TurboJPEG lossless transformer instance.
+   */
+  public TJTransformer() throws TJException {
+    init();
+  }
+
+  /**
+   * Create a TurboJPEG lossless transformer instance and associate the JPEG
+   * image stored in <code>jpegImage</code> with the newly created instance.
+   *
+   * @param jpegImage JPEG image buffer (size of the JPEG image is assumed to
+   * be the length of the array.)  This buffer is not modified.
+   */
+  public TJTransformer(byte jpegImage) throws TJException {
+    init();
+    setSourceImage(jpegImage, jpegImage.length);
+  }
+
+  /**
+   * Create a TurboJPEG lossless transformer instance and associate the JPEG
+   * image of length <code>imageSize</code> bytes stored in
+   * <code>jpegImage</code> with the newly created instance.
+   *
+   * @param jpegImage JPEG image buffer.  This buffer is not modified.
+   *
+   * @param imageSize size of the JPEG image (in bytes)
+   */
+  public TJTransformer(byte jpegImage, int imageSize) throws TJException {
+    init();
+    setSourceImage(jpegImage, imageSize);
+  }
+
+  /**
+   * Losslessly transform the JPEG image associated with this transformer
+   * instance into one or more JPEG images stored in the given destination
+   * buffers.  Lossless transforms work by moving the raw coefficients from one
+   * JPEG image structure to another without altering the values of the
+   * coefficients.  While this is typically faster than decompressing the
+   * image, transforming it, and re-compressing it, lossless transforms are not
+   * free.  Each lossless transform requires reading and performing Huffman
+   * decoding on all of the coefficients in the source image, regardless of the
+   * size of the destination image.  Thus, this method provides a means of
+   * generating multiple transformed images from the same source or of applying
+   * multiple transformations simultaneously, in order to eliminate the need to
+   * read the source coefficients multiple times.
+   *
+   * @param dstBufs an array of image buffers.  <code>dstbufsi</code> will
+   * receive a JPEG image that has been transformed using the parameters in
+   * <code>transformsi</code>.  Use {@link TJ#bufSize} to determine the
+   * maximum size for each buffer based on the transformed or cropped width and
+   * height and the level of subsampling used in the source image.
+   *
+   * @param transforms an array of {@link TJTransform} instances, each of
+   * which specifies the transform parameters and/or cropping region for the
+   * corresponding transformed output image
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   */
+  public void transform(byte dstBufs, TJTransform transforms,
+                        int flags) throws TJException {
+    if (jpegBuf == null)
+      throw new IllegalStateException("JPEG buffer not initialized");
+    transformedSizes = transform(jpegBuf, jpegBufSize, dstBufs, transforms,
+                                 flags);
+  }
+
+  /**
+   * Losslessly transform the JPEG image associated with this transformer
+   * instance and return an array of {@link TJDecompressor} instances, each of
+   * which has a transformed JPEG image associated with it.
+   *
+   * @param transforms an array of {@link TJTransform} instances, each of
+   * which specifies the transform parameters and/or cropping region for the
+   * corresponding transformed output image
+   *
+   * @param flags the bitwise OR of one or more of
+   * {@link TJ#FLAG_BOTTOMUP TJ.FLAG_*}
+   *
+   * @return an array of {@link TJDecompressor} instances, each of
+   * which has a transformed JPEG image associated with it.
+   */
+  public TJDecompressor transform(TJTransform transforms, int flags)
+                                    throws TJException {
+    byte dstBufs = new bytetransforms.length;
+    if (jpegWidth < 1 || jpegHeight < 1)
+      throw new IllegalStateException("JPEG buffer not initialized");
+    for (int i = 0; i < transforms.length; i++) {
+      int w = jpegWidth, h = jpegHeight;
+      if ((transformsi.options & TJTransform.OPT_CROP) != 0) {
+        if (transformsi.width != 0) w = transformsi.width;
+        if (transformsi.height != 0) h = transformsi.height;
+      }
+      dstBufsi = new byteTJ.bufSize(w, h, jpegSubsamp);
+    }
+    TJDecompressor tjd = new TJDecompressortransforms.length;
+    transform(dstBufs, transforms, flags);
+    for (int i = 0; i < transforms.length; i++)
+      tjdi = new TJDecompressor(dstBufsi, transformedSizesi);
+    return tjd;
+  }
+
+  /**
+   * Returns an array containing the sizes of the transformed JPEG images
+   * generated by the most recent transform operation.
+   *
+   * @return an array containing the sizes of the transformed JPEG images
+   * generated by the most recent transform operation.
+   */
+  public int getTransformedSizes() {
+    if (transformedSizes == null)
+      throw new IllegalStateException("No image has been transformed yet");
+    return transformedSizes;
+  }
+
+  private native void init() throws TJException;
+
+  private native int transform(byte srcBuf, int srcSize, byte dstBufs,
+    TJTransform transforms, int flags) throws TJException;
+
+  static {
+    TJLoader.load();
+  }
+
+  private int transformedSizes = null;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org/libjpegturbo/turbojpeg/YUVImage.java Added

@@ -0,0 +1,445 @@
+/*
+ * Copyright (C)2014, 2017 D. R. Commander.  All Rights Reserved.
+ * Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+package org.libjpegturbo.turbojpeg;
+
+/**
+ * This class encapsulates a YUV planar image and the metadata
+ * associated with it.  The TurboJPEG API allows both the JPEG compression and
+ * decompression pipelines to be split into stages:  YUV encode, compress from
+ * YUV, decompress to YUV, and YUV decode.  A <code>YUVImage</code> instance
+ * serves as the destination image for YUV encode and decompress-to-YUV
+ * operations and as the source image for compress-from-YUV and YUV decode
+ * operations.
+ * <p>
+ * Technically, the JPEG format uses the YCbCr colorspace (which technically is
+ * not a "colorspace" but rather a "color transform"), but per the convention
+ * of the digital video community, the TurboJPEG API uses "YUV" to refer to an
+ * image format consisting of Y, Cb, and Cr image planes.
+ * <p>
+ * Each plane is simply a 2D array of bytes, each byte representing the value
+ * of one of the components (Y, Cb, or Cr) at a particular location in the
+ * image.  The width and height of each plane are determined by the image
+ * width, height, and level of chrominance subsampling.  The luminance plane
+ * width is the image width padded to the nearest multiple of the horizontal
+ * subsampling factor (2 in the case of 4:2:0 and 4:2:2, 4 in the case of
+ * 4:1:1, 1 in the case of 4:4:4 or grayscale.)  Similarly, the luminance plane
+ * height is the image height padded to the nearest multiple of the vertical
+ * subsampling factor (2 in the case of 4:2:0 or 4:4:0, 1 in the case of 4:4:4
+ * or grayscale.)  The chrominance plane width is equal to the luminance plane
+ * width divided by the horizontal subsampling factor, and the chrominance
+ * plane height is equal to the luminance plane height divided by the vertical
+ * subsampling factor.
+ * <p>
+ * For example, if the source image is 35 x 35 pixels and 4:2:2 subsampling is
+ * used, then the luminance plane would be 36 x 35 bytes, and each of the
+ * chrominance planes would be 18 x 35 bytes.  If you specify a line padding of
+ * 4 bytes on top of this, then the luminance plane would be 36 x 35 bytes, and
+ * each of the chrominance planes would be 20 x 35 bytes.
+ */
+public class YUVImage {
+
+  private static final String NO_ASSOC_ERROR =
+    "No image data is associated with this instance";
+
+  /**
+   * Create a new <code>YUVImage</code> instance backed by separate image
+   * planes, and allocate memory for the image planes.
+   *
+   * @param width width (in pixels) of the YUV image
+   *
+   * @param strides an array of integers, each specifying the number of bytes
+   * per line in the corresponding plane of the YUV image.  Setting the stride
+   * for any plane to 0 is the same as setting it to the plane width (see
+   * {@link YUVImage above}.)  If <code>strides</code> is null, then the
+   * strides for all planes will be set to their respective plane widths.  When
+   * using this constructor, the stride for each plane must be equal to or
+   * greater than the plane width.
+   *
+   * @param height height (in pixels) of the YUV image
+   *
+   * @param subsamp the level of chrominance subsampling to be used in the YUV
+   * image (one of {@link TJ#SAMP_444 TJ.SAMP_*})
+   */
+  public YUVImage(int width, int strides, int height, int subsamp) {
+    setBuf(null, null, width, strides, height, subsamp, true);
+  }
+
+  /**
+   * Create a new <code>YUVImage</code> instance backed by a unified image
+   * buffer, and allocate memory for the image buffer.
+   *
+   * @param width width (in pixels) of the YUV image
+   *
+   * @param pad Each line of each plane in the YUV image buffer will be padded
+   * to this number of bytes (must be a power of 2.)
+   *
+   * @param height height (in pixels) of the YUV image
+   *
+   * @param subsamp the level of chrominance subsampling to be used in the YUV
+   * image (one of {@link TJ#SAMP_444 TJ.SAMP_*})
+   */
+  public YUVImage(int width, int pad, int height, int subsamp) {
+    setBuf(new byteTJ.bufSizeYUV(width, pad, height, subsamp), width, pad,
+           height, subsamp);
+  }
+
+  /**
+   * Create a new <code>YUVImage</code> instance from a set of existing image
+   * planes.
+   *
+   * @param planes an array of buffers representing the Y, U (Cb), and V (Cr)
+   * image planes (or just the Y plane, if the image is grayscale.)   These
+   * planes can be contiguous or non-contiguous in memory.  Plane
+   * <code>i</code> should be at least <code>offsetsi +
+   * {@link TJ#planeSizeYUV TJ.planeSizeYUV}(i, width, stridesi, height, subsamp)</code>
+   * bytes in size.
+   *
+   * @param offsets If this <code>YUVImage</code> instance represents a
+   * subregion of a larger image, then <code>offsetsi</code> specifies the
+   * offset (in bytes) of the subregion within plane <code>i</code> of the
+   * larger image.  Setting this to null is the same as setting the offsets for
+   * all planes to 0.
+   *
+   * @param width width (in pixels) of the new YUV image (or subregion)
+   *
+   * @param strides an array of integers, each specifying the number of bytes
+   * per line in the corresponding plane of the YUV image.  Setting the stride
+   * for any plane to 0 is the same as setting it to the plane width (see
+   * {@link YUVImage above}.)  If <code>strides</code> is null, then the
+   * strides for all planes will be set to their respective plane widths.  You
+   * can adjust the strides in order to add an arbitrary amount of line padding
+   * to each plane or to specify that this <code>YUVImage</code> instance is a
+   * subregion of a larger image (in which case, <code>stridesi</code> should
+   * be set to the plane width of plane <code>i</code> in the larger image.)
+   *
+   * @param height height (in pixels) of the new YUV image (or subregion)
+   *
+   * @param subsamp the level of chrominance subsampling used in the YUV
+   * image (one of {@link TJ#SAMP_444 TJ.SAMP_*})
+   */
+  public YUVImage(byte planes, int offsets, int width, int strides,
+                  int height, int subsamp) {
+    setBuf(planes, offsets, width, strides, height, subsamp, false);
+  }
+
+  /**
+   * Create a new <code>YUVImage</code> instance from an existing unified image
+   * buffer.
+   *
+   * @param yuvImage image buffer that contains or will contain YUV planar
+   * image data.  Use {@link TJ#bufSizeYUV} to determine the minimum size for
+   * this buffer.  The Y, U (Cb), and V (Cr) image planes are stored
+   * sequentially in the buffer (see {@link YUVImage above} for a description
+   * of the image format.)
+   *
+   * @param width width (in pixels) of the YUV image
+   *
+   * @param pad the line padding used in the YUV image buffer.  For
+   * instance, if each line in each plane of the buffer is padded to the
+   * nearest multiple of 4 bytes, then <code>pad</code> should be set to 4.
+   *
+   * @param height height (in pixels) of the YUV image
+   *
+   * @param subsamp the level of chrominance subsampling used in the YUV
+   * image (one of {@link TJ#SAMP_444 TJ.SAMP_*})
+   */
+  public YUVImage(byte yuvImage, int width, int pad, int height,
+                  int subsamp) {
+    setBuf(yuvImage, width, pad, height, subsamp);
+  }
+
+  /**
+   * Assign a set of image planes to this <code>YUVImage</code> instance.
+   *
+   * @param planes an array of buffers representing the Y, U (Cb), and V (Cr)
+   * image planes (or just the Y plane, if the image is grayscale.)  These
+   * planes can be contiguous or non-contiguous in memory.  Plane
+   * <code>i</code> should be at least <code>offsetsi +
+   * {@link TJ#planeSizeYUV TJ.planeSizeYUV}(i, width, stridesi, height, subsamp)</code>
+   * bytes in size.
+   *
+   * @param offsets If this <code>YUVImage</code> instance represents a
+   * subregion of a larger image, then <code>offsetsi</code> specifies the
+   * offset (in bytes) of the subregion within plane <code>i</code> of the
+   * larger image.  Setting this to null is the same as setting the offsets for
+   * all planes to 0.
+   *
+   * @param width width (in pixels) of the YUV image (or subregion)
+   *
+   * @param strides an array of integers, each specifying the number of bytes
+   * per line in the corresponding plane of the YUV image.  Setting the stride
+   * for any plane to 0 is the same as setting it to the plane width (see
+   * {@link YUVImage above}.)  If <code>strides</code> is null, then the
+   * strides for all planes will be set to their respective plane widths.  You
+   * can adjust the strides in order to add an arbitrary amount of line padding
+   * to each plane or to specify that this <code>YUVImage</code> image is a
+   * subregion of a larger image (in which case, <code>stridesi</code> should
+   * be set to the plane width of plane <code>i</code> in the larger image.)
+   *
+   * @param height height (in pixels) of the YUV image (or subregion)
+   *
+   * @param subsamp the level of chrominance subsampling used in the YUV
+   * image (one of {@link TJ#SAMP_444 TJ.SAMP_*})
+   */
+  public void setBuf(byte planes, int offsets, int width, int strides,
+                     int height, int subsamp) {
+    setBuf(planes, offsets, width, strides, height, subsamp, false);
+  }
+
+  private void setBuf(byte planes, int offsets, int width, int strides,
+                     int height, int subsamp, boolean alloc) {
+    if ((planes == null && !alloc) || width < 1 || height < 1 || subsamp < 0 ||
+        subsamp >= TJ.NUMSAMP)
+      throw new IllegalArgumentException("Invalid argument in YUVImage::setBuf()");
+
+    int nc = (subsamp == TJ.SAMP_GRAY ? 1 : 3);
+    if ((planes != null && planes.length != nc) ||
+        (offsets != null && offsets.length != nc) ||
+        (strides != null && strides.length != nc))
+      throw new IllegalArgumentException("YUVImage::setBuf(): planes, offsets, or strides array is the wrong size");
+
+    if (planes == null)
+      planes = new bytenc;
+    if (offsets == null)
+      offsets = new intnc;
+    if (strides == null)
+      strides = new intnc;
+
+    for (int i = 0; i < nc; i++) {
+      int pw = TJ.planeWidth(i, width, subsamp);
+      int ph = TJ.planeHeight(i, height, subsamp);
+      int planeSize = TJ.planeSizeYUV(i, width, stridesi, height, subsamp);
+
+      if (stridesi == 0)
+        stridesi = pw;
+      if (alloc) {
+        if (stridesi < pw)
+          throw new IllegalArgumentException("Stride must be >= plane width when allocating a new YUV image");
+        planesi = new bytestridesi * ph;
+      }
+      if (planesi == null || offsetsi < 0)
+        throw new IllegalArgumentException("Invalid argument in YUVImage::setBuf()");
+      if (stridesi < 0 && offsetsi - planeSize + pw < 0)
+        throw new IllegalArgumentException("Stride for plane " + i +
+                                           " would cause memory to be accessed below plane boundary");
+      if (planesi.length < offsetsi + planeSize)
+        throw new IllegalArgumentException("Image plane " + i +
+                                           " is not large enough");
+    }
+
+    yuvPlanes = planes;
+    yuvOffsets = offsets;
+    yuvWidth = width;
+    yuvStrides = strides;
+    yuvHeight = height;
+    yuvSubsamp = subsamp;
+  }
+
+  /**
+   * Assign a unified image buffer to this <code>YUVImage</code> instance.
+   *
+   * @param yuvImage image buffer that contains or will contain YUV planar
+   * image data.  Use {@link TJ#bufSizeYUV} to determine the minimum size for
+   * this buffer.  The Y, U (Cb), and V (Cr) image planes are stored
+   * sequentially in the buffer (see {@link YUVImage above} for a description
+   * of the image format.)
+   *
+   * @param width width (in pixels) of the YUV image
+   *
+   * @param pad the line padding used in the YUV image buffer.  For
+   * instance, if each line in each plane of the buffer is padded to the
+   * nearest multiple of 4 bytes, then <code>pad</code> should be set to 4.
+   *
+   * @param height height (in pixels) of the YUV image
+   *
+   * @param subsamp the level of chrominance subsampling used in the YUV
+   * image (one of {@link TJ#SAMP_444 TJ.SAMP_*})
+   */
+  public void setBuf(byte yuvImage, int width, int pad, int height,
+                     int subsamp) {
+    if (yuvImage == null || width < 1 || pad < 1 || ((pad & (pad - 1)) != 0) ||
+        height < 1 || subsamp < 0 || subsamp >= TJ.NUMSAMP)
+      throw new IllegalArgumentException("Invalid argument in YUVImage::setBuf()");
+    if (yuvImage.length < TJ.bufSizeYUV(width, pad, height, subsamp))
+      throw new IllegalArgumentException("YUV image buffer is not large enough");
+
+    int nc = (subsamp == TJ.SAMP_GRAY ? 1 : 3);
+    byte planes = new bytenc;
+    int strides = new intnc;
+    int offsets = new intnc;
+
+    planes0 = yuvImage;
+    strides0 = pad(TJ.planeWidth(0, width, subsamp), pad);
+    if (subsamp != TJ.SAMP_GRAY) {
+      strides1 = strides2 = pad(TJ.planeWidth(1, width, subsamp), pad);
+      planes1 = planes2 = yuvImage;
+      offsets1 = offsets0 +
+        strides0 * TJ.planeHeight(0, height, subsamp);
+      offsets2 = offsets1 +
+        strides1 * TJ.planeHeight(1, height, subsamp);
+    }
+
+    yuvPad = pad;
+    setBuf(planes, offsets, width, strides, height, subsamp);
+  }
+
+  /**
+   * Returns the width of the YUV image (or subregion.)
+   *
+   * @return the width of the YUV image (or subregion)
+   */
+  public int getWidth() {
+    if (yuvWidth < 1)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return yuvWidth;
+  }
+
+  /**
+   * Returns the height of the YUV image (or subregion.)
+   *
+   * @return the height of the YUV image (or subregion)
+   */
+  public int getHeight() {
+    if (yuvHeight < 1)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return yuvHeight;
+  }
+
+  /**
+   * Returns the line padding used in the YUV image buffer (if this image is
+   * stored in a unified buffer rather than separate image planes.)
+   *
+   * @return the line padding used in the YUV image buffer
+   */
+  public int getPad() {
+    if (yuvPlanes == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    if (yuvPad < 1 || ((yuvPad & (yuvPad - 1)) != 0))
+      throw new IllegalStateException("Image is not stored in a unified buffer");
+    return yuvPad;
+  }
+
+  /**
+   * Returns the number of bytes per line of each plane in the YUV image.
+   *
+   * @return the number of bytes per line of each plane in the YUV image
+   */
+  public int getStrides() {
+    if (yuvStrides == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return yuvStrides;
+  }
+
+  /**
+   * Returns the offsets (in bytes) of each plane within the planes of a larger
+   * YUV image.
+   *
+   * @return the offsets (in bytes) of each plane within the planes of a larger
+   * YUV image
+   */
+  public int getOffsets() {
+    if (yuvOffsets == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return yuvOffsets;
+  }
+
+  /**
+   * Returns the level of chrominance subsampling used in the YUV image.  See
+   * {@link TJ#SAMP_444 TJ.SAMP_*}.
+   *
+   * @return the level of chrominance subsampling used in the YUV image
+   */
+  public int getSubsamp() {
+    if (yuvSubsamp < 0 || yuvSubsamp >= TJ.NUMSAMP)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return yuvSubsamp;
+  }
+
+  /**
+   * Returns the YUV image planes.  If the image is stored in a unified buffer,
+   * then all image planes will point to that buffer.
+   *
+   * @return the YUV image planes
+   */
+  public byte getPlanes() {
+    if (yuvPlanes == null)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    return yuvPlanes;
+  }
+
+  /**
+   * Returns the YUV image buffer (if this image is stored in a unified
+   * buffer rather than separate image planes.)
+   *
+   * @return the YUV image buffer
+   */
+  public byte getBuf() {
+    if (yuvPlanes == null || yuvSubsamp < 0 || yuvSubsamp >= TJ.NUMSAMP)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    int nc = (yuvSubsamp == TJ.SAMP_GRAY ? 1 : 3);
+    for (int i = 1; i < nc; i++) {
+      if (yuvPlanesi != yuvPlanes0)
+        throw new IllegalStateException("Image is not stored in a unified buffer");
+    }
+    return yuvPlanes0;
+  }
+
+  /**
+   * Returns the size (in bytes) of the YUV image buffer (if this image is
+   * stored in a unified buffer rather than separate image planes.)
+   *
+   * @return the size (in bytes) of the YUV image buffer
+   */
+  public int getSize() {
+    if (yuvPlanes == null || yuvSubsamp < 0 || yuvSubsamp >= TJ.NUMSAMP)
+      throw new IllegalStateException(NO_ASSOC_ERROR);
+    int nc = (yuvSubsamp == TJ.SAMP_GRAY ? 1 : 3);
+    if (yuvPad < 1)
+      throw new IllegalStateException("Image is not stored in a unified buffer");
+    for (int i = 1; i < nc; i++) {
+      if (yuvPlanesi != yuvPlanes0)
+        throw new IllegalStateException("Image is not stored in a unified buffer");
+    }
+    return TJ.bufSizeYUV(yuvWidth, yuvPad, yuvHeight, yuvSubsamp);
+  }
+
+  private static int pad(int v, int p) {
+    return (v + p - 1) & (~(p - 1));
+  }
+
+  protected long handle = 0;
+  protected byte yuvPlanes = null;
+  protected int yuvOffsets = null;
+  protected int yuvStrides = null;
+  protected int yuvPad = 0;
+  protected int yuvWidth = 0;
+  protected int yuvHeight = 0;
+  protected int yuvSubsamp = -1;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org_libjpegturbo_turbojpeg_TJ.h Added

@@ -0,0 +1,149 @@
+/* DO NOT EDIT THIS FILE - it is machine generated */
+#include <jni.h>
+/* Header for class org_libjpegturbo_turbojpeg_TJ */
+
+#ifndef _Included_org_libjpegturbo_turbojpeg_TJ
+#define _Included_org_libjpegturbo_turbojpeg_TJ
+#ifdef __cplusplus
+extern "C" {
+#endif
+#undef org_libjpegturbo_turbojpeg_TJ_NUMSAMP
+#define org_libjpegturbo_turbojpeg_TJ_NUMSAMP 6L
+#undef org_libjpegturbo_turbojpeg_TJ_SAMP_444
+#define org_libjpegturbo_turbojpeg_TJ_SAMP_444 0L
+#undef org_libjpegturbo_turbojpeg_TJ_SAMP_422
+#define org_libjpegturbo_turbojpeg_TJ_SAMP_422 1L
+#undef org_libjpegturbo_turbojpeg_TJ_SAMP_420
+#define org_libjpegturbo_turbojpeg_TJ_SAMP_420 2L
+#undef org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY
+#define org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY 3L
+#undef org_libjpegturbo_turbojpeg_TJ_SAMP_440
+#define org_libjpegturbo_turbojpeg_TJ_SAMP_440 4L
+#undef org_libjpegturbo_turbojpeg_TJ_SAMP_411
+#define org_libjpegturbo_turbojpeg_TJ_SAMP_411 5L
+#undef org_libjpegturbo_turbojpeg_TJ_NUMPF
+#define org_libjpegturbo_turbojpeg_TJ_NUMPF 12L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_RGB
+#define org_libjpegturbo_turbojpeg_TJ_PF_RGB 0L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_BGR
+#define org_libjpegturbo_turbojpeg_TJ_PF_BGR 1L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_RGBX
+#define org_libjpegturbo_turbojpeg_TJ_PF_RGBX 2L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_BGRX
+#define org_libjpegturbo_turbojpeg_TJ_PF_BGRX 3L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_XBGR
+#define org_libjpegturbo_turbojpeg_TJ_PF_XBGR 4L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_XRGB
+#define org_libjpegturbo_turbojpeg_TJ_PF_XRGB 5L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_GRAY
+#define org_libjpegturbo_turbojpeg_TJ_PF_GRAY 6L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_RGBA
+#define org_libjpegturbo_turbojpeg_TJ_PF_RGBA 7L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_BGRA
+#define org_libjpegturbo_turbojpeg_TJ_PF_BGRA 8L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_ABGR
+#define org_libjpegturbo_turbojpeg_TJ_PF_ABGR 9L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_ARGB
+#define org_libjpegturbo_turbojpeg_TJ_PF_ARGB 10L
+#undef org_libjpegturbo_turbojpeg_TJ_PF_CMYK
+#define org_libjpegturbo_turbojpeg_TJ_PF_CMYK 11L
+#undef org_libjpegturbo_turbojpeg_TJ_NUMCS
+#define org_libjpegturbo_turbojpeg_TJ_NUMCS 5L
+#undef org_libjpegturbo_turbojpeg_TJ_CS_RGB
+#define org_libjpegturbo_turbojpeg_TJ_CS_RGB 0L
+#undef org_libjpegturbo_turbojpeg_TJ_CS_YCbCr
+#define org_libjpegturbo_turbojpeg_TJ_CS_YCbCr 1L
+#undef org_libjpegturbo_turbojpeg_TJ_CS_GRAY
+#define org_libjpegturbo_turbojpeg_TJ_CS_GRAY 2L
+#undef org_libjpegturbo_turbojpeg_TJ_CS_CMYK
+#define org_libjpegturbo_turbojpeg_TJ_CS_CMYK 3L
+#undef org_libjpegturbo_turbojpeg_TJ_CS_YCCK
+#define org_libjpegturbo_turbojpeg_TJ_CS_YCCK 4L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_BOTTOMUP
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_BOTTOMUP 2L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_FORCEMMX
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_FORCEMMX 8L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_FORCESSE
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_FORCESSE 16L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_FORCESSE2
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_FORCESSE2 32L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_FORCESSE3
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_FORCESSE3 128L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_FASTUPSAMPLE
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_FASTUPSAMPLE 256L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_FASTDCT
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_FASTDCT 2048L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_ACCURATEDCT
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_ACCURATEDCT 4096L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_STOPONWARNING
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_STOPONWARNING 8192L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_PROGRESSIVE
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_PROGRESSIVE 16384L
+#undef org_libjpegturbo_turbojpeg_TJ_FLAG_LIMITSCANS
+#define org_libjpegturbo_turbojpeg_TJ_FLAG_LIMITSCANS 32768L
+#undef org_libjpegturbo_turbojpeg_TJ_NUMERR
+#define org_libjpegturbo_turbojpeg_TJ_NUMERR 2L
+#undef org_libjpegturbo_turbojpeg_TJ_ERR_WARNING
+#define org_libjpegturbo_turbojpeg_TJ_ERR_WARNING 0L
+#undef org_libjpegturbo_turbojpeg_TJ_ERR_FATAL
+#define org_libjpegturbo_turbojpeg_TJ_ERR_FATAL 1L
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJ
+ * Method:    bufSize
+ * Signature: (III)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSize
+  (JNIEnv *, jclass, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJ
+ * Method:    bufSizeYUV
+ * Signature: (IIII)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII
+  (JNIEnv *, jclass, jint, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJ
+ * Method:    bufSizeYUV
+ * Signature: (III)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__III
+  (JNIEnv *, jclass, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJ
+ * Method:    planeSizeYUV
+ * Signature: (IIIII)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeSizeYUV__IIIII
+  (JNIEnv *, jclass, jint, jint, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJ
+ * Method:    planeWidth
+ * Signature: (III)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeWidth__III
+  (JNIEnv *, jclass, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJ
+ * Method:    planeHeight
+ * Signature: (III)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeHeight__III
+  (JNIEnv *, jclass, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJ
+ * Method:    getScalingFactors
+ * Signature: ()Lorg/libjpegturbo/turbojpeg/TJScalingFactor;
+ */
+JNIEXPORT jobjectArray JNICALL Java_org_libjpegturbo_turbojpeg_TJ_getScalingFactors
+  (JNIEnv *, jclass);
+
+#ifdef __cplusplus
+}
+#endif
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org_libjpegturbo_turbojpeg_TJCompressor.h Added

@@ -0,0 +1,101 @@
+/* DO NOT EDIT THIS FILE - it is machine generated */
+#include <jni.h>
+/* Header for class org_libjpegturbo_turbojpeg_TJCompressor */
+
+#ifndef _Included_org_libjpegturbo_turbojpeg_TJCompressor
+#define _Included_org_libjpegturbo_turbojpeg_TJCompressor
+#ifdef __cplusplus
+extern "C" {
+#endif
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    init
+ * Signature: ()V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_init
+  (JNIEnv *, jobject);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    destroy
+ * Signature: ()V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy
+  (JNIEnv *, jobject);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    compress
+ * Signature: (BIIIIBIII)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIII_3BIII
+  (JNIEnv *, jobject, jbyteArray, jint, jint, jint, jint, jbyteArray, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    compress
+ * Signature: (BIIIIIIBIII)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIIIII_3BIII
+  (JNIEnv *, jobject, jbyteArray, jint, jint, jint, jint, jint, jint, jbyteArray, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    compress
+ * Signature: (IIIIIBIII)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIII_3BIII
+  (JNIEnv *, jobject, jintArray, jint, jint, jint, jint, jbyteArray, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    compress
+ * Signature: (IIIIIIIBIII)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIIIII_3BIII
+  (JNIEnv *, jobject, jintArray, jint, jint, jint, jint, jint, jint, jbyteArray, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    compressFromYUV
+ * Signature: (BIIIIIBII)I
+ */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compressFromYUV___3_3B_3II_3III_3BII
+  (JNIEnv *, jobject, jobjectArray, jintArray, jint, jintArray, jint, jint, jbyteArray, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    encodeYUV
+ * Signature: (BIIIIBII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIII_3BII
+  (JNIEnv *, jobject, jbyteArray, jint, jint, jint, jint, jbyteArray, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    encodeYUV
+ * Signature: (BIIIIIIBIIII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIIIII_3_3B_3I_3III
+  (JNIEnv *, jobject, jbyteArray, jint, jint, jint, jint, jint, jint, jobjectArray, jintArray, jintArray, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    encodeYUV
+ * Signature: (IIIIIBII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIII_3BII
+  (JNIEnv *, jobject, jintArray, jint, jint, jint, jint, jbyteArray, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJCompressor
+ * Method:    encodeYUV
+ * Signature: (IIIIIIIBIIII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIIIII_3_3B_3I_3III
+  (JNIEnv *, jobject, jintArray, jint, jint, jint, jint, jint, jint, jobjectArray, jintArray, jintArray, jint, jint);
+
+#ifdef __cplusplus
+}
+#endif
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org_libjpegturbo_turbojpeg_TJDecompressor.h Added

@@ -0,0 +1,101 @@
+/* DO NOT EDIT THIS FILE - it is machine generated */
+#include <jni.h>
+/* Header for class org_libjpegturbo_turbojpeg_TJDecompressor */
+
+#ifndef _Included_org_libjpegturbo_turbojpeg_TJDecompressor
+#define _Included_org_libjpegturbo_turbojpeg_TJDecompressor
+#ifdef __cplusplus
+extern "C" {
+#endif
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    init
+ * Signature: ()V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_init
+  (JNIEnv *, jobject);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    destroy
+ * Signature: ()V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_destroy
+  (JNIEnv *, jobject);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    decompressHeader
+ * Signature: (BI)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressHeader
+  (JNIEnv *, jobject, jbyteArray, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    decompress
+ * Signature: (BIBIIIII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIII
+  (JNIEnv *, jobject, jbyteArray, jint, jbyteArray, jint, jint, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    decompress
+ * Signature: (BIBIIIIIII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIIIII
+  (JNIEnv *, jobject, jbyteArray, jint, jbyteArray, jint, jint, jint, jint, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    decompress
+ * Signature: (BIIIIIII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIII
+  (JNIEnv *, jobject, jbyteArray, jint, jintArray, jint, jint, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    decompress
+ * Signature: (BIIIIIIIII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIIIII
+  (JNIEnv *, jobject, jbyteArray, jint, jintArray, jint, jint, jint, jint, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    decompressToYUV
+ * Signature: (BIBI)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3BI
+  (JNIEnv *, jobject, jbyteArray, jint, jbyteArray, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    decompressToYUV
+ * Signature: (BIBIIIII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3_3B_3II_3III
+  (JNIEnv *, jobject, jbyteArray, jint, jobjectArray, jintArray, jint, jintArray, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    decodeYUV
+ * Signature: (BIIIBIIIIIII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3BIIIIIII
+  (JNIEnv *, jobject, jobjectArray, jintArray, jintArray, jint, jbyteArray, jint, jint, jint, jint, jint, jint, jint);
+
+/*
+ * Class:     org_libjpegturbo_turbojpeg_TJDecompressor
+ * Method:    decodeYUV
+ * Signature: (BIIIIIIIIIII)V
+ */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3IIIIIIII
+  (JNIEnv *, jobject, jobjectArray, jintArray, jintArray, jint, jintArray, jint, jint, jint, jint, jint, jint, jint);
+
+#ifdef __cplusplus
+}
+#endif
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/java/org_libjpegturbo_turbojpeg_TJTransformer.h Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcapimin.c Added

@@ -0,0 +1,295 @@
+/*
+ * jcapimin.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * Modified 2003-2010 by Guido Vollbeding.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains application interface code for the compression half
+ * of the JPEG library.  These are the "minimum" API routines that may be
+ * needed in either the normal full-compression case or the transcoding-only
+ * case.
+ *
+ * Most of the routines intended to be called directly by an application
+ * are in this file or in jcapistd.c.  But also see jcparam.c for
+ * parameter-setup helper routines, jcomapi.c for routines shared by
+ * compression and decompression, and jctrans.c for the transcoding case.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Initialization of a JPEG compression object.
+ * The error manager must already be set up (in case memory manager fails).
+ */
+
+GLOBAL(void)
+jpeg_CreateCompress(j_compress_ptr cinfo, int version, size_t structsize)
+{
+  int i;
+
+  /* Guard against version mismatches between library and caller. */
+  cinfo->mem = NULL;            /* so jpeg_destroy knows mem mgr not called */
+  if (version != JPEG_LIB_VERSION)
+    ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
+  if (structsize != sizeof(struct jpeg_compress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE,
+             (int)sizeof(struct jpeg_compress_struct), (int)structsize);
+
+  /* For debugging purposes, we zero the whole master structure.
+   * But the application has already set the err pointer, and may have set
+   * client_data, so we have to save and restore those fields.
+   * Note: if application hasn't set client_data, tools like Purify may
+   * complain here.
+   */
+  {
+    struct jpeg_error_mgr *err = cinfo->err;
+    void *client_data = cinfo->client_data; /* ignore Purify complaint here */
+    MEMZERO(cinfo, sizeof(struct jpeg_compress_struct));
+    cinfo->err = err;
+    cinfo->client_data = client_data;
+  }
+  cinfo->is_decompressor = FALSE;
+
+  /* Initialize a memory manager instance for this object */
+  jinit_memory_mgr((j_common_ptr)cinfo);
+
+  /* Zero out pointers to permanent structures. */
+  cinfo->progress = NULL;
+  cinfo->dest = NULL;
+
+  cinfo->comp_info = NULL;
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    cinfo->quant_tbl_ptrsi = NULL;
+#if JPEG_LIB_VERSION >= 70
+    cinfo->q_scale_factori = 100;
+#endif
+  }
+
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    cinfo->dc_huff_tbl_ptrsi = NULL;
+    cinfo->ac_huff_tbl_ptrsi = NULL;
+  }
+
+#if JPEG_LIB_VERSION >= 80
+  /* Must do it here for emit_dqt in case jpeg_write_tables is used */
+  cinfo->block_size = DCTSIZE;
+  cinfo->natural_order = jpeg_natural_order;
+  cinfo->lim_Se = DCTSIZE2 - 1;
+#endif
+
+  cinfo->script_space = NULL;
+
+  cinfo->input_gamma = 1.0;     /* in case application forgets */
+
+  /* OK, I'm ready */
+  cinfo->global_state = CSTATE_START;
+}
+
+
+/*
+ * Destruction of a JPEG compression object
+ */
+
+GLOBAL(void)
+jpeg_destroy_compress(j_compress_ptr cinfo)
+{
+  jpeg_destroy((j_common_ptr)cinfo); /* use common routine */
+}
+
+
+/*
+ * Abort processing of a JPEG compression operation,
+ * but don't destroy the object itself.
+ */
+
+GLOBAL(void)
+jpeg_abort_compress(j_compress_ptr cinfo)
+{
+  jpeg_abort((j_common_ptr)cinfo); /* use common routine */
+}
+
+
+/*
+ * Forcibly suppress or un-suppress all quantization and Huffman tables.
+ * Marks all currently defined tables as already written (if suppress)
+ * or not written (if !suppress).  This will control whether they get emitted
+ * by a subsequent jpeg_start_compress call.
+ *
+ * This routine is exported for use by applications that want to produce
+ * abbreviated JPEG datastreams.  It logically belongs in jcparam.c, but
+ * since it is called by jpeg_start_compress, we put it here --- otherwise
+ * jcparam.o would be linked whether the application used it or not.
+ */
+
+GLOBAL(void)
+jpeg_suppress_tables(j_compress_ptr cinfo, boolean suppress)
+{
+  int i;
+  JQUANT_TBL *qtbl;
+  JHUFF_TBL *htbl;
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    if ((qtbl = cinfo->quant_tbl_ptrsi) != NULL)
+      qtbl->sent_table = suppress;
+  }
+
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    if ((htbl = cinfo->dc_huff_tbl_ptrsi) != NULL)
+      htbl->sent_table = suppress;
+    if ((htbl = cinfo->ac_huff_tbl_ptrsi) != NULL)
+      htbl->sent_table = suppress;
+  }
+}
+
+
+/*
+ * Finish JPEG compression.
+ *
+ * If a multipass operating mode was selected, this may do a great deal of
+ * work including most of the actual output.
+ */
+
+GLOBAL(void)
+jpeg_finish_compress(j_compress_ptr cinfo)
+{
+  JDIMENSION iMCU_row;
+
+  if (cinfo->global_state == CSTATE_SCANNING ||
+      cinfo->global_state == CSTATE_RAW_OK) {
+    /* Terminate first pass */
+    if (cinfo->next_scanline < cinfo->image_height)
+      ERREXIT(cinfo, JERR_TOO_LITTLE_DATA);
+    (*cinfo->master->finish_pass) (cinfo);
+  } else if (cinfo->global_state != CSTATE_WRCOEFS)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Perform any remaining passes */
+  while (!cinfo->master->is_last_pass) {
+    (*cinfo->master->prepare_for_pass) (cinfo);
+    for (iMCU_row = 0; iMCU_row < cinfo->total_iMCU_rows; iMCU_row++) {
+      if (cinfo->progress != NULL) {
+        cinfo->progress->pass_counter = (long)iMCU_row;
+        cinfo->progress->pass_limit = (long)cinfo->total_iMCU_rows;
+        (*cinfo->progress->progress_monitor) ((j_common_ptr)cinfo);
+      }
+      /* We bypass the main controller and invoke coef controller directly;
+       * all work is being done from the coefficient buffer.
+       */
+      if (!(*cinfo->coef->compress_data) (cinfo, (JSAMPIMAGE)NULL))
+        ERREXIT(cinfo, JERR_CANT_SUSPEND);
+    }
+    (*cinfo->master->finish_pass) (cinfo);
+  }
+  /* Write EOI, do final cleanup */
+  (*cinfo->marker->write_file_trailer) (cinfo);
+  (*cinfo->dest->term_destination) (cinfo);
+  /* We can use jpeg_abort to release memory and reset global_state */
+  jpeg_abort((j_common_ptr)cinfo);
+}
+
+
+/*
+ * Write a special marker.
+ * This is only recommended for writing COM or APPn markers.
+ * Must be called after jpeg_start_compress() and before
+ * first call to jpeg_write_scanlines() or jpeg_write_raw_data().
+ */
+
+GLOBAL(void)
+jpeg_write_marker(j_compress_ptr cinfo, int marker, const JOCTET *dataptr,
+                  unsigned int datalen)
+{
+  void (*write_marker_byte) (j_compress_ptr info, int val);
+
+  if (cinfo->next_scanline != 0 ||
+      (cinfo->global_state != CSTATE_SCANNING &&
+       cinfo->global_state != CSTATE_RAW_OK &&
+       cinfo->global_state != CSTATE_WRCOEFS))
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  (*cinfo->marker->write_marker_header) (cinfo, marker, datalen);
+  write_marker_byte = cinfo->marker->write_marker_byte; /* copy for speed */
+  while (datalen--) {
+    (*write_marker_byte) (cinfo, *dataptr);
+    dataptr++;
+  }
+}
+
+/* Same, but piecemeal. */
+
+GLOBAL(void)
+jpeg_write_m_header(j_compress_ptr cinfo, int marker, unsigned int datalen)
+{
+  if (cinfo->next_scanline != 0 ||
+      (cinfo->global_state != CSTATE_SCANNING &&
+       cinfo->global_state != CSTATE_RAW_OK &&
+       cinfo->global_state != CSTATE_WRCOEFS))
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  (*cinfo->marker->write_marker_header) (cinfo, marker, datalen);
+}
+
+GLOBAL(void)
+jpeg_write_m_byte(j_compress_ptr cinfo, int val)
+{
+  (*cinfo->marker->write_marker_byte) (cinfo, val);
+}
+
+
+/*
+ * Alternate compression function: just write an abbreviated table file.
+ * Before calling this, all parameters and a data destination must be set up.
+ *
+ * To produce a pair of files containing abbreviated tables and abbreviated
+ * image data, one would proceed as follows:
+ *
+ *              initialize JPEG object
+ *              set JPEG parameters
+ *              set destination to table file
+ *              jpeg_write_tables(cinfo);
+ *              set destination to image file
+ *              jpeg_start_compress(cinfo, FALSE);
+ *              write data...
+ *              jpeg_finish_compress(cinfo);
+ *
+ * jpeg_write_tables has the side effect of marking all tables written
+ * (same as jpeg_suppress_tables(..., TRUE)).  Thus a subsequent start_compress
+ * will not re-emit the tables unless it is passed write_all_tables=TRUE.
+ */
+
+GLOBAL(void)
+jpeg_write_tables(j_compress_ptr cinfo)
+{
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* (Re)initialize error mgr and destination modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr)cinfo);
+  (*cinfo->dest->init_destination) (cinfo);
+  /* Initialize the marker writer ... bit of a crock to do it here. */
+  jinit_marker_writer(cinfo);
+  /* Write them tables! */
+  (*cinfo->marker->write_tables_only) (cinfo);
+  /* And clean up. */
+  (*cinfo->dest->term_destination) (cinfo);
+  /*
+   * In library releases up through v6a, we called jpeg_abort() here to free
+   * any working memory allocated by the destination manager and marker
+   * writer.  Some applications had a problem with that: they allocated space
+   * of their own from the library memory manager, and didn't want it to go
+   * away during write_tables.  So now we do nothing.  This will cause a
+   * memory leak if an app calls write_tables repeatedly without doing a full
+   * compression cycle or otherwise resetting the JPEG object.  However, that
+   * seems less bad than unexpectedly freeing memory in the normal case.
+   * An app that prefers the old behavior can call jpeg_abort for itself after
+   * each call to jpeg_write_tables().
+   */
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcapistd.c Added

@@ -0,0 +1,162 @@
+/*
+ * jcapistd.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains application interface code for the compression half
+ * of the JPEG library.  These are the "standard" API routines that are
+ * used in the normal full-compression case.  They are not used by a
+ * transcoding-only application.  Note that if an application links in
+ * jpeg_start_compress, it will end up linking in the entire compressor.
+ * We thus must separate this file from jcapimin.c to avoid linking the
+ * whole compression library into a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Compression initialization.
+ * Before calling this, all parameters and a data destination must be set up.
+ *
+ * We require a write_all_tables parameter as a failsafe check when writing
+ * multiple datastreams from the same compression object.  Since prior runs
+ * will have left all the tables marked sent_table=TRUE, a subsequent run
+ * would emit an abbreviated stream (no tables) by default.  This may be what
+ * is wanted, but for safety's sake it should not be the default behavior:
+ * programmers should have to make a deliberate choice to emit abbreviated
+ * images.  Therefore the documentation and examples should encourage people
+ * to pass write_all_tables=TRUE; then it will take active thought to do the
+ * wrong thing.
+ */
+
+GLOBAL(void)
+jpeg_start_compress(j_compress_ptr cinfo, boolean write_all_tables)
+{
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (write_all_tables)
+    jpeg_suppress_tables(cinfo, FALSE); /* mark all tables to be written */
+
+  /* (Re)initialize error mgr and destination modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr)cinfo);
+  (*cinfo->dest->init_destination) (cinfo);
+  /* Perform master selection of active modules */
+  jinit_compress_master(cinfo);
+  /* Set up for the first pass */
+  (*cinfo->master->prepare_for_pass) (cinfo);
+  /* Ready for application to drive first pass through jpeg_write_scanlines
+   * or jpeg_write_raw_data.
+   */
+  cinfo->next_scanline = 0;
+  cinfo->global_state = (cinfo->raw_data_in ? CSTATE_RAW_OK : CSTATE_SCANNING);
+}
+
+
+/*
+ * Write some scanlines of data to the JPEG compressor.
+ *
+ * The return value will be the number of lines actually written.
+ * This should be less than the supplied num_lines only in case that
+ * the data destination module has requested suspension of the compressor,
+ * or if more than image_height scanlines are passed in.
+ *
+ * Note: we warn about excess calls to jpeg_write_scanlines() since
+ * this likely signals an application programmer error.  However,
+ * excess scanlines passed in the last valid call are *silently* ignored,
+ * so that the application need not adjust num_lines for end-of-image
+ * when using a multiple-scanline buffer.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_write_scanlines(j_compress_ptr cinfo, JSAMPARRAY scanlines,
+                     JDIMENSION num_lines)
+{
+  JDIMENSION row_ctr, rows_left;
+
+  if (cinfo->global_state != CSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->next_scanline >= cinfo->image_height)
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long)cinfo->next_scanline;
+    cinfo->progress->pass_limit = (long)cinfo->image_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr)cinfo);
+  }
+
+  /* Give master control module another chance if this is first call to
+   * jpeg_write_scanlines.  This lets output of the frame/scan headers be
+   * delayed so that application can write COM, etc, markers between
+   * jpeg_start_compress and jpeg_write_scanlines.
+   */
+  if (cinfo->master->call_pass_startup)
+    (*cinfo->master->pass_startup) (cinfo);
+
+  /* Ignore any extra scanlines at bottom of image. */
+  rows_left = cinfo->image_height - cinfo->next_scanline;
+  if (num_lines > rows_left)
+    num_lines = rows_left;
+
+  row_ctr = 0;
+  (*cinfo->main->process_data) (cinfo, scanlines, &row_ctr, num_lines);
+  cinfo->next_scanline += row_ctr;
+  return row_ctr;
+}
+
+
+/*
+ * Alternate entry point to write raw data.
+ * Processes exactly one iMCU row per call, unless suspended.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_write_raw_data(j_compress_ptr cinfo, JSAMPIMAGE data,
+                    JDIMENSION num_lines)
+{
+  JDIMENSION lines_per_iMCU_row;
+
+  if (cinfo->global_state != CSTATE_RAW_OK)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->next_scanline >= cinfo->image_height) {
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+    return 0;
+  }
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long)cinfo->next_scanline;
+    cinfo->progress->pass_limit = (long)cinfo->image_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr)cinfo);
+  }
+
+  /* Give master control module another chance if this is first call to
+   * jpeg_write_raw_data.  This lets output of the frame/scan headers be
+   * delayed so that application can write COM, etc, markers between
+   * jpeg_start_compress and jpeg_write_raw_data.
+   */
+  if (cinfo->master->call_pass_startup)
+    (*cinfo->master->pass_startup) (cinfo);
+
+  /* Verify that at least one iMCU row has been passed. */
+  lines_per_iMCU_row = cinfo->max_v_samp_factor * DCTSIZE;
+  if (num_lines < lines_per_iMCU_row)
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* Directly compress the row. */
+  if (!(*cinfo->coef->compress_data) (cinfo, data)) {
+    /* If compressor did not consume the whole row, suspend processing. */
+    return 0;
+  }
+
+  /* OK, we processed one iMCU row. */
+  cinfo->next_scanline += lines_per_iMCU_row;
+  return lines_per_iMCU_row;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcarith.c Added

@@ -0,0 +1,932 @@
+/*
+ * jcarith.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Developed 1997-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, 2018, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains portable arithmetic entropy encoding routines for JPEG
+ * (implementing Recommendation ITU-T T.81 | ISO/IEC 10918-1).
+ *
+ * Both sequential and progressive modes are supported in this single module.
+ *
+ * Suspension is not currently supported in this module.
+ *
+ * NOTE: All referenced figures are from
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Expanded entropy encoder object for arithmetic encoding. */
+
+typedef struct {
+  struct jpeg_entropy_encoder pub; /* public fields */
+
+  JLONG c; /* C register, base of coding interval, layout as in sec. D.1.3 */
+  JLONG a;               /* A register, normalized size of coding interval */
+  JLONG sc;        /* counter for stacked 0xFF values which might overflow */
+  JLONG zc;          /* counter for pending 0x00 output values which might *
+                          * be discarded at the end ("Pacman" termination) */
+  int ct;  /* bit shift counter, determines when next byte will be written */
+  int buffer;                /* buffer for most recent output byte != 0xFF */
+
+  int last_dc_valMAX_COMPS_IN_SCAN; /* last DC coef for each component */
+  int dc_contextMAX_COMPS_IN_SCAN; /* context index for DC conditioning */
+
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+  int next_restart_num;         /* next restart number to write (0-7) */
+
+  /* Pointers to statistics areas (these workspaces have image lifespan) */
+  unsigned char *dc_statsNUM_ARITH_TBLS;
+  unsigned char *ac_statsNUM_ARITH_TBLS;
+
+  /* Statistics bin for coding with fixed probability 0.5 */
+  unsigned char fixed_bin4;
+} arith_entropy_encoder;
+
+typedef arith_entropy_encoder *arith_entropy_ptr;
+
+/* The following two definitions specify the allocation chunk size
+ * for the statistics area.
+ * According to sections F.1.4.4.1.3 and F.1.4.4.2, we need at least
+ * 49 statistics bins for DC, and 245 statistics bins for AC coding.
+ *
+ * We use a compact representation with 1 byte per statistics bin,
+ * thus the numbers directly represent byte sizes.
+ * This 1 byte per statistics bin contains the meaning of the MPS
+ * (more probable symbol) in the highest bit (mask 0x80), and the
+ * index into the probability estimation state machine table
+ * in the lower bits (mask 0x7F).
+ */
+
+#define DC_STAT_BINS  64
+#define AC_STAT_BINS  256
+
+/* NOTE: Uncomment the following #define if you want to use the
+ * given formula for calculating the AC conditioning parameter Kx
+ * for spectral selection progressive coding in section G.1.3.2
+ * of the spec (Kx = Kmin + SRL (8 + Se - Kmin) 4).
+ * Although the spec and P&M authors claim that this "has proven
+ * to give good results for 8 bit precision samples", I'm not
+ * convinced yet that this is really beneficial.
+ * Early tests gave only very marginal compression enhancements
+ * (a few - around 5 or so - bytes even for very large files),
+ * which would turn out rather negative if we'd suppress the
+ * DAC (Define Arithmetic Conditioning) marker segments for
+ * the default parameters in the future.
+ * Note that currently the marker writing module emits 12-byte
+ * DAC segments for a full-component scan in a color image.
+ * This is not worth worrying about IMHO. However, since the
+ * spec defines the default values to be used if the tables
+ * are omitted (unlike Huffman tables, which are required
+ * anyway), one might optimize this behaviour in the future,
+ * and then it would be disadvantageous to use custom tables if
+ * they don't provide sufficient gain to exceed the DAC size.
+ *
+ * On the other hand, I'd consider it as a reasonable result
+ * that the conditioning has no significant influence on the
+ * compression performance. This means that the basic
+ * statistical model is already rather stable.
+ *
+ * Thus, at the moment, we use the default conditioning values
+ * anyway, and do not use the custom formula.
+ *
+#define CALCULATE_SPECTRAL_CONDITIONING
+ */
+
+/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than JLONG.
+ * We assume that int right shift is unsigned if JLONG right shift is,
+ * which should be safe.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define ISHIFT_TEMPS    int ishift_temp;
+#define IRIGHT_SHIFT(x, shft) \
+  ((ishift_temp = (x)) < 0 ? \
+   (ishift_temp >> (shft)) | ((~0) << (16 - (shft))) : \
+   (ishift_temp >> (shft)))
+#else
+#define ISHIFT_TEMPS
+#define IRIGHT_SHIFT(x, shft)   ((x) >> (shft))
+#endif
+
+
+LOCAL(void)
+emit_byte(int val, j_compress_ptr cinfo)
+/* Write next output byte; we do not support suspension in this module. */
+{
+  struct jpeg_destination_mgr *dest = cinfo->dest;
+
+  *dest->next_output_byte++ = (JOCTET)val;
+  if (--dest->free_in_buffer == 0)
+    if (!(*dest->empty_output_buffer) (cinfo))
+      ERREXIT(cinfo, JERR_CANT_SUSPEND);
+}
+
+
+/*
+ * Finish up at the end of an arithmetic-compressed scan.
+ */
+
+METHODDEF(void)
+finish_pass(j_compress_ptr cinfo)
+{
+  arith_entropy_ptr e = (arith_entropy_ptr)cinfo->entropy;
+  JLONG temp;
+
+  /* Section D.1.8: Termination of encoding */
+
+  /* Find the e->c in the coding interval with the largest
+   * number of trailing zero bits */
+  if ((temp = (e->a - 1 + e->c) & 0xFFFF0000UL) < e->c)
+    e->c = temp + 0x8000L;
+  else
+    e->c = temp;
+  /* Send remaining bytes to output */
+  e->c <<= e->ct;
+  if (e->c & 0xF8000000UL) {
+    /* One final overflow has to be handled */
+    if (e->buffer >= 0) {
+      if (e->zc)
+        do emit_byte(0x00, cinfo);
+        while (--e->zc);
+      emit_byte(e->buffer + 1, cinfo);
+      if (e->buffer + 1 == 0xFF)
+        emit_byte(0x00, cinfo);
+    }
+    e->zc += e->sc;  /* carry-over converts stacked 0xFF bytes to 0x00 */
+    e->sc = 0;
+  } else {
+    if (e->buffer == 0)
+      ++e->zc;
+    else if (e->buffer >= 0) {
+      if (e->zc)
+        do emit_byte(0x00, cinfo);
+        while (--e->zc);
+      emit_byte(e->buffer, cinfo);
+    }
+    if (e->sc) {
+      if (e->zc)
+        do emit_byte(0x00, cinfo);
+        while (--e->zc);
+      do {
+        emit_byte(0xFF, cinfo);
+        emit_byte(0x00, cinfo);
+      } while (--e->sc);
+    }
+  }
+  /* Output final bytes only if they are not 0x00 */
+  if (e->c & 0x7FFF800L) {
+    if (e->zc)  /* output final pending zero bytes */
+      do emit_byte(0x00, cinfo);
+      while (--e->zc);
+    emit_byte((e->c >> 19) & 0xFF, cinfo);
+    if (((e->c >> 19) & 0xFF) == 0xFF)
+      emit_byte(0x00, cinfo);
+    if (e->c & 0x7F800L) {
+      emit_byte((e->c >> 11) & 0xFF, cinfo);
+      if (((e->c >> 11) & 0xFF) == 0xFF)
+        emit_byte(0x00, cinfo);
+    }
+  }
+}
+
+
+/*
+ * The core arithmetic encoding routine (common in JPEG and JBIG).
+ * This needs to go as fast as possible.
+ * Machine-dependent optimization facilities
+ * are not utilized in this portable implementation.
+ * However, this code should be fairly efficient and
+ * may be a good base for further optimizations anyway.
+ *
+ * Parameter 'val' to be encoded may be 0 or 1 (binary decision).
+ *
+ * Note: I've added full "Pacman" termination support to the
+ * byte output routines, which is equivalent to the optional
+ * Discard_final_zeros procedure (Figure D.15) in the spec.
+ * Thus, we always produce the shortest possible output
+ * stream compliant to the spec (no trailing zero bytes,
+ * except for FF stuffing).
+ *
+ * I've also introduced a new scheme for accessing
+ * the probability estimation state machine table,
+ * derived from Markus Kuhn's JBIG implementation.
+ */
+
+LOCAL(void)
+arith_encode(j_compress_ptr cinfo, unsigned char *st, int val)
+{
+  register arith_entropy_ptr e = (arith_entropy_ptr)cinfo->entropy;
+  register unsigned char nl, nm;
+  register JLONG qe, temp;
+  register int sv;
+
+  /* Fetch values from our compact representation of Table D.2:
+   * Qe values and probability estimation state machine
+   */
+  sv = *st;
+  qe = jpeg_aritabsv & 0x7F;  /* => Qe_Value */
+  nl = qe & 0xFF;  qe >>= 8;    /* Next_Index_LPS + Switch_MPS */
+  nm = qe & 0xFF;  qe >>= 8;    /* Next_Index_MPS */
+
+  /* Encode & estimation procedures per sections D.1.4 & D.1.5 */
+  e->a -= qe;
+  if (val != (sv >> 7)) {
+    /* Encode the less probable symbol */
+    if (e->a >= qe) {
+      /* If the interval size (qe) for the less probable symbol (LPS)
+       * is larger than the interval size for the MPS, then exchange
+       * the two symbols for coding efficiency, otherwise code the LPS
+       * as usual: */
+      e->c += e->a;
+      e->a = qe;
+    }
+    *st = (sv & 0x80) ^ nl;     /* Estimate_after_LPS */
+  } else {
+    /* Encode the more probable symbol */
+    if (e->a >= 0x8000L)
+      return;  /* A >= 0x8000 -> ready, no renormalization required */
+    if (e->a < qe) {
+      /* If the interval size (qe) for the less probable symbol (LPS)
+       * is larger than the interval size for the MPS, then exchange
+       * the two symbols for coding efficiency: */
+      e->c += e->a;
+      e->a = qe;
+    }
+    *st = (sv & 0x80) ^ nm;     /* Estimate_after_MPS */
+  }
+
+  /* Renormalization & data output per section D.1.6 */
+  do {
+    e->a <<= 1;
+    e->c <<= 1;
+    if (--e->ct == 0) {
+      /* Another byte is ready for output */
+      temp = e->c >> 19;
+      if (temp > 0xFF) {
+        /* Handle overflow over all stacked 0xFF bytes */
+        if (e->buffer >= 0) {
+          if (e->zc)
+            do emit_byte(0x00, cinfo);
+            while (--e->zc);
+          emit_byte(e->buffer + 1, cinfo);
+          if (e->buffer + 1 == 0xFF)
+            emit_byte(0x00, cinfo);
+        }
+        e->zc += e->sc;  /* carry-over converts stacked 0xFF bytes to 0x00 */
+        e->sc = 0;
+        /* Note: The 3 spacer bits in the C register guarantee
+         * that the new buffer byte can't be 0xFF here
+         * (see page 160 in the P&M JPEG book). */
+        e->buffer = temp & 0xFF;  /* new output byte, might overflow later */
+      } else if (temp == 0xFF) {
+        ++e->sc;  /* stack 0xFF byte (which might overflow later) */
+      } else {
+        /* Output all stacked 0xFF bytes, they will not overflow any more */
+        if (e->buffer == 0)
+          ++e->zc;
+        else if (e->buffer >= 0) {
+          if (e->zc)
+            do emit_byte(0x00, cinfo);
+            while (--e->zc);
+          emit_byte(e->buffer, cinfo);
+        }
+        if (e->sc) {
+          if (e->zc)
+            do emit_byte(0x00, cinfo);
+            while (--e->zc);
+          do {
+            emit_byte(0xFF, cinfo);
+            emit_byte(0x00, cinfo);
+          } while (--e->sc);
+        }
+        e->buffer = temp & 0xFF;  /* new output byte (can still overflow) */
+      }
+      e->c &= 0x7FFFFL;
+      e->ct += 8;
+    }
+  } while (e->a < 0x8000L);
+}
+
+
+/*
+ * Emit a restart marker & resynchronize predictions.
+ */
+
+LOCAL(void)
+emit_restart(j_compress_ptr cinfo, int restart_num)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  int ci;
+  jpeg_component_info *compptr;
+
+  finish_pass(cinfo);
+
+  emit_byte(0xFF, cinfo);
+  emit_byte(JPEG_RST0 + restart_num, cinfo);
+
+  /* Re-initialize statistics areas */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    /* DC needs no table for refinement scan */
+    if (cinfo->progressive_mode == 0 || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+      MEMZERO(entropy->dc_statscompptr->dc_tbl_no, DC_STAT_BINS);
+      /* Reset DC predictions to 0 */
+      entropy->last_dc_valci = 0;
+      entropy->dc_contextci = 0;
+    }
+    /* AC needs no table when not present */
+    if (cinfo->progressive_mode == 0 || cinfo->Se) {
+      MEMZERO(entropy->ac_statscompptr->ac_tbl_no, AC_STAT_BINS);
+    }
+  }
+
+  /* Reset arithmetic encoding variables */
+  entropy->c = 0;
+  entropy->a = 0x10000L;
+  entropy->sc = 0;
+  entropy->zc = 0;
+  entropy->ct = 11;
+  entropy->buffer = -1;  /* empty */
+}
+
+
+/*
+ * MCU encoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_first(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  JBLOCKROW block;
+  unsigned char *st;
+  int blkn, ci, tbl;
+  int v, v2, m;
+  ISHIFT_TEMPS
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      emit_restart(cinfo, entropy->next_restart_num);
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_datablkn;
+    ci = cinfo->MCU_membershipblkn;
+    tbl = cinfo->cur_comp_infoci->dc_tbl_no;
+
+    /* Compute the DC value after the required point transform by Al.
+     * This is simply an arithmetic right shift.
+     */
+    m = IRIGHT_SHIFT((int)((*block)0), cinfo->Al);
+
+    /* Sections F.1.4.1 & F.1.4.4.1: Encoding of DC coefficients */
+
+    /* Table F.4: Point to statistics bin S0 for DC coefficient coding */
+    st = entropy->dc_statstbl + entropy->dc_contextci;
+
+    /* Figure F.4: Encode_DC_DIFF */
+    if ((v = m - entropy->last_dc_valci) == 0) {
+      arith_encode(cinfo, st, 0);
+      entropy->dc_contextci = 0;      /* zero diff category */
+    } else {
+      entropy->last_dc_valci = m;
+      arith_encode(cinfo, st, 1);
+      /* Figure F.6: Encoding nonzero value v */
+      /* Figure F.7: Encoding the sign of v */
+      if (v > 0) {
+        arith_encode(cinfo, st + 1, 0); /* Table F.4: SS = S0 + 1 */
+        st += 2;                        /* Table F.4: SP = S0 + 2 */
+        entropy->dc_contextci = 4;    /* small positive diff category */
+      } else {
+        v = -v;
+        arith_encode(cinfo, st + 1, 1); /* Table F.4: SS = S0 + 1 */
+        st += 3;                        /* Table F.4: SN = S0 + 3 */
+        entropy->dc_contextci = 8;    /* small negative diff category */
+      }
+      /* Figure F.8: Encoding the magnitude category of v */
+      m = 0;
+      if (v -= 1) {
+        arith_encode(cinfo, st, 1);
+        m = 1;
+        v2 = v;
+        st = entropy->dc_statstbl + 20; /* Table F.4: X1 = 20 */
+        while (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st += 1;
+        }
+      }
+      arith_encode(cinfo, st, 0);
+      /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
+      if (m < (int)((1L << cinfo->arith_dc_Ltbl) >> 1))
+        entropy->dc_contextci = 0;    /* zero diff category */
+      else if (m > (int)((1L << cinfo->arith_dc_Utbl) >> 1))
+        entropy->dc_contextci += 8;   /* large diff category */
+      /* Figure F.9: Encoding the magnitude bit pattern of v */
+      st += 14;
+      while (m >>= 1)
+        arith_encode(cinfo, st, (m & v) ? 1 : 0);
+    }
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+encode_mcu_AC_first(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  JBLOCKROW block;
+  unsigned char *st;
+  int tbl, k, ke;
+  int v, v2, m;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      emit_restart(cinfo, entropy->next_restart_num);
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  /* Encode the MCU data block */
+  block = MCU_data0;
+  tbl = cinfo->cur_comp_info0->ac_tbl_no;
+
+  /* Sections F.1.4.2 & F.1.4.4.2: Encoding of AC coefficients */
+
+  /* Establish EOB (end-of-block) index */
+  for (ke = cinfo->Se; ke > 0; ke--)
+    /* We must apply the point transform by Al.  For AC coefficients this
+     * is an integer division with rounding towards 0.  To do this portably
+     * in C, we shift after obtaining the absolute value.
+     */
+    if ((v = (*block)jpeg_natural_orderke) >= 0) {
+      if (v >>= cinfo->Al) break;
+    } else {
+      v = -v;
+      if (v >>= cinfo->Al) break;
+    }
+
+  /* Figure F.5: Encode_AC_Coefficients */
+  for (k = cinfo->Ss; k <= ke; k++) {
+    st = entropy->ac_statstbl + 3 * (k - 1);
+    arith_encode(cinfo, st, 0);         /* EOB decision */
+    for (;;) {
+      if ((v = (*block)jpeg_natural_orderk) >= 0) {
+        if (v >>= cinfo->Al) {
+          arith_encode(cinfo, st + 1, 1);
+          arith_encode(cinfo, entropy->fixed_bin, 0);
+          break;
+        }
+      } else {
+        v = -v;
+        if (v >>= cinfo->Al) {
+          arith_encode(cinfo, st + 1, 1);
+          arith_encode(cinfo, entropy->fixed_bin, 1);
+          break;
+        }
+      }
+      arith_encode(cinfo, st + 1, 0);  st += 3;  k++;
+    }
+    st += 2;
+    /* Figure F.8: Encoding the magnitude category of v */
+    m = 0;
+    if (v -= 1) {
+      arith_encode(cinfo, st, 1);
+      m = 1;
+      v2 = v;
+      if (v2 >>= 1) {
+        arith_encode(cinfo, st, 1);
+        m <<= 1;
+        st = entropy->ac_statstbl +
+             (k <= cinfo->arith_ac_Ktbl ? 189 : 217);
+        while (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st += 1;
+        }
+      }
+    }
+    arith_encode(cinfo, st, 0);
+    /* Figure F.9: Encoding the magnitude bit pattern of v */
+    st += 14;
+    while (m >>= 1)
+      arith_encode(cinfo, st, (m & v) ? 1 : 0);
+  }
+  /* Encode EOB decision only if k <= cinfo->Se */
+  if (k <= cinfo->Se) {
+    st = entropy->ac_statstbl + 3 * (k - 1);
+    arith_encode(cinfo, st, 1);
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for DC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  unsigned char *st;
+  int Al, blkn;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      emit_restart(cinfo, entropy->next_restart_num);
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  st = entropy->fixed_bin;      /* use fixed probability estimation */
+  Al = cinfo->Al;
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    /* We simply emit the Al'th bit of the DC coefficient value. */
+    arith_encode(cinfo, st, (MCU_datablkn00 >> Al) & 1);
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for AC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  JBLOCKROW block;
+  unsigned char *st;
+  int tbl, k, ke, kex;
+  int v;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      emit_restart(cinfo, entropy->next_restart_num);
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  /* Encode the MCU data block */
+  block = MCU_data0;
+  tbl = cinfo->cur_comp_info0->ac_tbl_no;
+
+  /* Section G.1.3.3: Encoding of AC coefficients */
+
+  /* Establish EOB (end-of-block) index */
+  for (ke = cinfo->Se; ke > 0; ke--)
+    /* We must apply the point transform by Al.  For AC coefficients this
+     * is an integer division with rounding towards 0.  To do this portably
+     * in C, we shift after obtaining the absolute value.
+     */
+    if ((v = (*block)jpeg_natural_orderke) >= 0) {
+      if (v >>= cinfo->Al) break;
+    } else {
+      v = -v;
+      if (v >>= cinfo->Al) break;
+    }
+
+  /* Establish EOBx (previous stage end-of-block) index */
+  for (kex = ke; kex > 0; kex--)
+    if ((v = (*block)jpeg_natural_orderkex) >= 0) {
+      if (v >>= cinfo->Ah) break;
+    } else {
+      v = -v;
+      if (v >>= cinfo->Ah) break;
+    }
+
+  /* Figure G.10: Encode_AC_Coefficients_SA */
+  for (k = cinfo->Ss; k <= ke; k++) {
+    st = entropy->ac_statstbl + 3 * (k - 1);
+    if (k > kex)
+      arith_encode(cinfo, st, 0);       /* EOB decision */
+    for (;;) {
+      if ((v = (*block)jpeg_natural_orderk) >= 0) {
+        if (v >>= cinfo->Al) {
+          if (v >> 1)                   /* previously nonzero coef */
+            arith_encode(cinfo, st + 2, (v & 1));
+          else {                        /* newly nonzero coef */
+            arith_encode(cinfo, st + 1, 1);
+            arith_encode(cinfo, entropy->fixed_bin, 0);
+          }
+          break;
+        }
+      } else {
+        v = -v;
+        if (v >>= cinfo->Al) {
+          if (v >> 1)                   /* previously nonzero coef */
+            arith_encode(cinfo, st + 2, (v & 1));
+          else {                        /* newly nonzero coef */
+            arith_encode(cinfo, st + 1, 1);
+            arith_encode(cinfo, entropy->fixed_bin, 1);
+          }
+          break;
+        }
+      }
+      arith_encode(cinfo, st + 1, 0);  st += 3;  k++;
+    }
+  }
+  /* Encode EOB decision only if k <= cinfo->Se */
+  if (k <= cinfo->Se) {
+    st = entropy->ac_statstbl + 3 * (k - 1);
+    arith_encode(cinfo, st, 1);
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Encode and output one MCU's worth of arithmetic-compressed coefficients.
+ */
+
+METHODDEF(boolean)
+encode_mcu(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  jpeg_component_info *compptr;
+  JBLOCKROW block;
+  unsigned char *st;
+  int blkn, ci, tbl, k, ke;
+  int v, v2, m;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      emit_restart(cinfo, entropy->next_restart_num);
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_datablkn;
+    ci = cinfo->MCU_membershipblkn;
+    compptr = cinfo->cur_comp_infoci;
+
+    /* Sections F.1.4.1 & F.1.4.4.1: Encoding of DC coefficients */
+
+    tbl = compptr->dc_tbl_no;
+
+    /* Table F.4: Point to statistics bin S0 for DC coefficient coding */
+    st = entropy->dc_statstbl + entropy->dc_contextci;
+
+    /* Figure F.4: Encode_DC_DIFF */
+    if ((v = (*block)0 - entropy->last_dc_valci) == 0) {
+      arith_encode(cinfo, st, 0);
+      entropy->dc_contextci = 0;      /* zero diff category */
+    } else {
+      entropy->last_dc_valci = (*block)0;
+      arith_encode(cinfo, st, 1);
+      /* Figure F.6: Encoding nonzero value v */
+      /* Figure F.7: Encoding the sign of v */
+      if (v > 0) {
+        arith_encode(cinfo, st + 1, 0); /* Table F.4: SS = S0 + 1 */
+        st += 2;                        /* Table F.4: SP = S0 + 2 */
+        entropy->dc_contextci = 4;    /* small positive diff category */
+      } else {
+        v = -v;
+        arith_encode(cinfo, st + 1, 1); /* Table F.4: SS = S0 + 1 */
+        st += 3;                        /* Table F.4: SN = S0 + 3 */
+        entropy->dc_contextci = 8;    /* small negative diff category */
+      }
+      /* Figure F.8: Encoding the magnitude category of v */
+      m = 0;
+      if (v -= 1) {
+        arith_encode(cinfo, st, 1);
+        m = 1;
+        v2 = v;
+        st = entropy->dc_statstbl + 20; /* Table F.4: X1 = 20 */
+        while (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st += 1;
+        }
+      }
+      arith_encode(cinfo, st, 0);
+      /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
+      if (m < (int)((1L << cinfo->arith_dc_Ltbl) >> 1))
+        entropy->dc_contextci = 0;    /* zero diff category */
+      else if (m > (int)((1L << cinfo->arith_dc_Utbl) >> 1))
+        entropy->dc_contextci += 8;   /* large diff category */
+      /* Figure F.9: Encoding the magnitude bit pattern of v */
+      st += 14;
+      while (m >>= 1)
+        arith_encode(cinfo, st, (m & v) ? 1 : 0);
+    }
+
+    /* Sections F.1.4.2 & F.1.4.4.2: Encoding of AC coefficients */
+
+    tbl = compptr->ac_tbl_no;
+
+    /* Establish EOB (end-of-block) index */
+    for (ke = DCTSIZE2 - 1; ke > 0; ke--)
+      if ((*block)jpeg_natural_orderke) break;
+
+    /* Figure F.5: Encode_AC_Coefficients */
+    for (k = 1; k <= ke; k++) {
+      st = entropy->ac_statstbl + 3 * (k - 1);
+      arith_encode(cinfo, st, 0);       /* EOB decision */
+      while ((v = (*block)jpeg_natural_orderk) == 0) {
+        arith_encode(cinfo, st + 1, 0);  st += 3;  k++;
+      }
+      arith_encode(cinfo, st + 1, 1);
+      /* Figure F.6: Encoding nonzero value v */
+      /* Figure F.7: Encoding the sign of v */
+      if (v > 0) {
+        arith_encode(cinfo, entropy->fixed_bin, 0);
+      } else {
+        v = -v;
+        arith_encode(cinfo, entropy->fixed_bin, 1);
+      }
+      st += 2;
+      /* Figure F.8: Encoding the magnitude category of v */
+      m = 0;
+      if (v -= 1) {
+        arith_encode(cinfo, st, 1);
+        m = 1;
+        v2 = v;
+        if (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st = entropy->ac_statstbl +
+               (k <= cinfo->arith_ac_Ktbl ? 189 : 217);
+          while (v2 >>= 1) {
+            arith_encode(cinfo, st, 1);
+            m <<= 1;
+            st += 1;
+          }
+        }
+      }
+      arith_encode(cinfo, st, 0);
+      /* Figure F.9: Encoding the magnitude bit pattern of v */
+      st += 14;
+      while (m >>= 1)
+        arith_encode(cinfo, st, (m & v) ? 1 : 0);
+    }
+    /* Encode EOB decision only if k <= DCTSIZE2 - 1 */
+    if (k <= DCTSIZE2 - 1) {
+      st = entropy->ac_statstbl + 3 * (k - 1);
+      arith_encode(cinfo, st, 1);
+    }
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Initialize for an arithmetic-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass(j_compress_ptr cinfo, boolean gather_statistics)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  int ci, tbl;
+  jpeg_component_info *compptr;
+
+  if (gather_statistics)
+    /* Make sure to avoid that in the master control logic!
+     * We are fully adaptive here and need no extra
+     * statistics gathering pass!
+     */
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+
+  /* We assume jcmaster.c already validated the progressive scan parameters. */
+
+  /* Select execution routines */
+  if (cinfo->progressive_mode) {
+    if (cinfo->Ah == 0) {
+      if (cinfo->Ss == 0)
+        entropy->pub.encode_mcu = encode_mcu_DC_first;
+      else
+        entropy->pub.encode_mcu = encode_mcu_AC_first;
+    } else {
+      if (cinfo->Ss == 0)
+        entropy->pub.encode_mcu = encode_mcu_DC_refine;
+      else
+        entropy->pub.encode_mcu = encode_mcu_AC_refine;
+    }
+  } else
+    entropy->pub.encode_mcu = encode_mcu;
+
+  /* Allocate & initialize requested statistics areas */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    /* DC needs no table for refinement scan */
+    if (cinfo->progressive_mode == 0 || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+      tbl = compptr->dc_tbl_no;
+      if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+      if (entropy->dc_statstbl == NULL)
+        entropy->dc_statstbl = (unsigned char *)(*cinfo->mem->alloc_small)
+          ((j_common_ptr)cinfo, JPOOL_IMAGE, DC_STAT_BINS);
+      MEMZERO(entropy->dc_statstbl, DC_STAT_BINS);
+      /* Initialize DC predictions to 0 */
+      entropy->last_dc_valci = 0;
+      entropy->dc_contextci = 0;
+    }
+    /* AC needs no table when not present */
+    if (cinfo->progressive_mode == 0 || cinfo->Se) {
+      tbl = compptr->ac_tbl_no;
+      if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+      if (entropy->ac_statstbl == NULL)
+        entropy->ac_statstbl = (unsigned char *)(*cinfo->mem->alloc_small)
+          ((j_common_ptr)cinfo, JPOOL_IMAGE, AC_STAT_BINS);
+      MEMZERO(entropy->ac_statstbl, AC_STAT_BINS);
+#ifdef CALCULATE_SPECTRAL_CONDITIONING
+      if (cinfo->progressive_mode)
+        /* Section G.1.3.2: Set appropriate arithmetic conditioning value Kx */
+        cinfo->arith_ac_Ktbl = cinfo->Ss +
+                                 ((8 + cinfo->Se - cinfo->Ss) >> 4);
+#endif
+    }
+  }
+
+  /* Initialize arithmetic encoding variables */
+  entropy->c = 0;
+  entropy->a = 0x10000L;
+  entropy->sc = 0;
+  entropy->zc = 0;
+  entropy->ct = 11;
+  entropy->buffer = -1;  /* empty */
+
+  /* Initialize restart stuff */
+  entropy->restarts_to_go = cinfo->restart_interval;
+  entropy->next_restart_num = 0;
+}
+
+
+/*
+ * Module initialization routine for arithmetic entropy encoding.
+ */
+
+GLOBAL(void)
+jinit_arith_encoder(j_compress_ptr cinfo)
+{
+  arith_entropy_ptr entropy;
+  int i;
+
+  entropy = (arith_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(arith_entropy_encoder));
+  cinfo->entropy = (struct jpeg_entropy_encoder *)entropy;
+  entropy->pub.start_pass = start_pass;
+  entropy->pub.finish_pass = finish_pass;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    entropy->dc_statsi = NULL;
+    entropy->ac_statsi = NULL;
+  }
+
+  /* Initialize index for fixed probability estimation */
+  entropy->fixed_bin0 = 113;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jccoefct.c Added

@@ -0,0 +1,449 @@
+/*
+ * jccoefct.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the coefficient buffer controller for compression.
+ * This controller is the top level of the JPEG compressor proper.
+ * The coefficient buffer lies between forward-DCT and entropy encoding steps.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* We use a full-image coefficient buffer when doing Huffman optimization,
+ * and also for writing multiple-scan JPEG files.  In all cases, the DCT
+ * step is run during the first pass, and subsequent passes need only read
+ * the buffered coefficients.
+ */
+#ifdef ENTROPY_OPT_SUPPORTED
+#define FULL_COEF_BUFFER_SUPPORTED
+#else
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+#define FULL_COEF_BUFFER_SUPPORTED
+#endif
+#endif
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_coef_controller pub; /* public fields */
+
+  JDIMENSION iMCU_row_num;      /* iMCU row # within image */
+  JDIMENSION mcu_ctr;           /* counts MCUs processed in current row */
+  int MCU_vert_offset;          /* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;    /* number of such rows needed */
+
+  /* For single-pass compression, it's sufficient to buffer just one MCU
+   * (although this may prove a bit slow in practice).  We allocate a
+   * workspace of C_MAX_BLOCKS_IN_MCU coefficient blocks, and reuse it for each
+   * MCU constructed and sent.  In multi-pass modes, this array points to the
+   * current MCU's blocks within the virtual arrays.
+   */
+  JBLOCKROW MCU_bufferC_MAX_BLOCKS_IN_MCU;
+
+  /* In multi-pass modes, we need a virtual block array for each component. */
+  jvirt_barray_ptr whole_imageMAX_COMPONENTS;
+} my_coef_controller;
+
+typedef my_coef_controller *my_coef_ptr;
+
+
+/* Forward declarations */
+METHODDEF(boolean) compress_data(j_compress_ptr cinfo, JSAMPIMAGE input_buf);
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+METHODDEF(boolean) compress_first_pass(j_compress_ptr cinfo,
+                                       JSAMPIMAGE input_buf);
+METHODDEF(boolean) compress_output(j_compress_ptr cinfo, JSAMPIMAGE input_buf);
+#endif
+
+
+LOCAL(void)
+start_iMCU_row(j_compress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row */
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (coef->iMCU_row_num < (cinfo->total_iMCU_rows - 1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info0->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info0->last_row_height;
+  }
+
+  coef->mcu_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_coef(j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+
+  coef->iMCU_row_num = 0;
+  start_iMCU_row(cinfo);
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+    if (coef->whole_image0 != NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    coef->pub.compress_data = compress_data;
+    break;
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+  case JBUF_SAVE_AND_PASS:
+    if (coef->whole_image0 == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    coef->pub.compress_data = compress_first_pass;
+    break;
+  case JBUF_CRANK_DEST:
+    if (coef->whole_image0 == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    coef->pub.compress_data = compress_output;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+}
+
+
+/*
+ * Process some data in the single-pass case.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the image.
+ * Returns TRUE if the iMCU row is completed, FALSE if suspended.
+ *
+ * NB: input_buf contains a plane for each component in image,
+ * which we index according to the component's SOF position.
+ */
+
+METHODDEF(boolean)
+compress_data(j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
+  JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  int blkn, bi, ci, yindex, yoffset, blockcnt;
+  JDIMENSION ypos, xpos;
+  jpeg_component_info *compptr;
+
+  /* Loop to write as much as one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->mcu_ctr; MCU_col_num <= last_MCU_col;
+         MCU_col_num++) {
+      /* Determine where data comes from in input_buf and do the DCT thing.
+       * Each call on forward_DCT processes a horizontal row of DCT blocks
+       * as wide as an MCU; we rely on having allocated the MCU_buffer blocks
+       * sequentially.  Dummy blocks at the right or bottom edge are filled in
+       * specially.  The data in them does not matter for image reconstruction,
+       * so we fill them with values that will encode to the smallest amount of
+       * data, viz: all zeroes in the AC entries, DC entries equal to previous
+       * block's DC value.  (Thanks to Thomas Kinsman for this idea.)
+       */
+      blkn = 0;
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+        compptr = cinfo->cur_comp_infoci;
+        blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width :
+                                                  compptr->last_col_width;
+        xpos = MCU_col_num * compptr->MCU_sample_width;
+        ypos = yoffset * DCTSIZE; /* ypos == (yoffset+yindex) * DCTSIZE */
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          if (coef->iMCU_row_num < last_iMCU_row ||
+              yoffset + yindex < compptr->last_row_height) {
+            (*cinfo->fdct->forward_DCT) (cinfo, compptr,
+                                         input_bufcompptr->component_index,
+                                         coef->MCU_bufferblkn,
+                                         ypos, xpos, (JDIMENSION)blockcnt);
+            if (blockcnt < compptr->MCU_width) {
+              /* Create some dummy blocks at the right edge of the image. */
+              jzero_far((void *)coef->MCU_bufferblkn + blockcnt,
+                        (compptr->MCU_width - blockcnt) * sizeof(JBLOCK));
+              for (bi = blockcnt; bi < compptr->MCU_width; bi++) {
+                coef->MCU_bufferblkn + bi00 =
+                  coef->MCU_bufferblkn + bi - 100;
+              }
+            }
+          } else {
+            /* Create a row of dummy blocks at the bottom of the image. */
+            jzero_far((void *)coef->MCU_bufferblkn,
+                      compptr->MCU_width * sizeof(JBLOCK));
+            for (bi = 0; bi < compptr->MCU_width; bi++) {
+              coef->MCU_bufferblkn + bi00 =
+                coef->MCU_bufferblkn - 100;
+            }
+          }
+          blkn += compptr->MCU_width;
+          ypos += DCTSIZE;
+        }
+      }
+      /* Try to write the MCU.  In event of a suspension failure, we will
+       * re-DCT the MCU on restart (a bit inefficient, could be fixed...)
+       */
+      if (!(*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->mcu_ctr = MCU_col_num;
+        return FALSE;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->mcu_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  coef->iMCU_row_num++;
+  start_iMCU_row(cinfo);
+  return TRUE;
+}
+
+
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+
+/*
+ * Process some data in the first pass of a multi-pass case.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the image.
+ * This amount of data is read from the source buffer, DCT'd and quantized,
+ * and saved into the virtual arrays.  We also generate suitable dummy blocks
+ * as needed at the right and lower edges.  (The dummy blocks are constructed
+ * in the virtual arrays, which have been padded appropriately.)  This makes
+ * it possible for subsequent passes not to worry about real vs. dummy blocks.
+ *
+ * We must also emit the data to the entropy encoder.  This is conveniently
+ * done by calling compress_output() after we've loaded the current strip
+ * of the virtual arrays.
+ *
+ * NB: input_buf contains a plane for each component in image.  All
+ * components are DCT'd and loaded into the virtual arrays in this pass.
+ * However, it may be that only a subset of the components are emitted to
+ * the entropy encoder during this first pass; be careful about looking
+ * at the scan-dependent variables (MCU dimensions, etc).
+ */
+
+METHODDEF(boolean)
+compress_first_pass(j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  JDIMENSION blocks_across, MCUs_across, MCUindex;
+  int bi, ci, h_samp_factor, block_row, block_rows, ndummy;
+  JCOEF lastDC;
+  jpeg_component_info *compptr;
+  JBLOCKARRAY buffer;
+  JBLOCKROW thisblockrow, lastblockrow;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Align the virtual buffer for this component. */
+    buffer = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr)cinfo, coef->whole_imageci,
+       coef->iMCU_row_num * compptr->v_samp_factor,
+       (JDIMENSION)compptr->v_samp_factor, TRUE);
+    /* Count non-dummy DCT block rows in this iMCU row. */
+    if (coef->iMCU_row_num < last_iMCU_row)
+      block_rows = compptr->v_samp_factor;
+    else {
+      /* NB: can't use last_row_height here, since may not be set! */
+      block_rows = (int)(compptr->height_in_blocks % compptr->v_samp_factor);
+      if (block_rows == 0) block_rows = compptr->v_samp_factor;
+    }
+    blocks_across = compptr->width_in_blocks;
+    h_samp_factor = compptr->h_samp_factor;
+    /* Count number of dummy blocks to be added at the right margin. */
+    ndummy = (int)(blocks_across % h_samp_factor);
+    if (ndummy > 0)
+      ndummy = h_samp_factor - ndummy;
+    /* Perform DCT for all non-dummy blocks in this iMCU row.  Each call
+     * on forward_DCT processes a complete horizontal row of DCT blocks.
+     */
+    for (block_row = 0; block_row < block_rows; block_row++) {
+      thisblockrow = bufferblock_row;
+      (*cinfo->fdct->forward_DCT) (cinfo, compptr,
+                                   input_bufci, thisblockrow,
+                                   (JDIMENSION)(block_row * DCTSIZE),
+                                   (JDIMENSION)0, blocks_across);
+      if (ndummy > 0) {
+        /* Create dummy blocks at the right edge of the image. */
+        thisblockrow += blocks_across; /* => first dummy block */
+        jzero_far((void *)thisblockrow, ndummy * sizeof(JBLOCK));
+        lastDC = thisblockrow-10;
+        for (bi = 0; bi < ndummy; bi++) {
+          thisblockrowbi0 = lastDC;
+        }
+      }
+    }
+    /* If at end of image, create dummy block rows as needed.
+     * The tricky part here is that within each MCU, we want the DC values
+     * of the dummy blocks to match the last real block's DC value.
+     * This squeezes a few more bytes out of the resulting file...
+     */
+    if (coef->iMCU_row_num == last_iMCU_row) {
+      blocks_across += ndummy;  /* include lower right corner */
+      MCUs_across = blocks_across / h_samp_factor;
+      for (block_row = block_rows; block_row < compptr->v_samp_factor;
+           block_row++) {
+        thisblockrow = bufferblock_row;
+        lastblockrow = bufferblock_row - 1;
+        jzero_far((void *)thisblockrow,
+                  (size_t)(blocks_across * sizeof(JBLOCK)));
+        for (MCUindex = 0; MCUindex < MCUs_across; MCUindex++) {
+          lastDC = lastblockrowh_samp_factor - 10;
+          for (bi = 0; bi < h_samp_factor; bi++) {
+            thisblockrowbi0 = lastDC;
+          }
+          thisblockrow += h_samp_factor; /* advance to next MCU in row */
+          lastblockrow += h_samp_factor;
+        }
+      }
+    }
+  }
+  /* NB: compress_output will increment iMCU_row_num if successful.
+   * A suspension return will result in redoing all the work above next time.
+   */
+
+  /* Emit data to the entropy encoder, sharing code with subsequent passes */
+  return compress_output(cinfo, input_buf);
+}
+
+
+/*
+ * Process some data in subsequent passes of a multi-pass case.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the scan.
+ * The data is obtained from the virtual arrays and fed to the entropy coder.
+ * Returns TRUE if the iMCU row is completed, FALSE if suspended.
+ *
+ * NB: input_buf is ignored; it is likely to be a NULL pointer.
+ */
+
+METHODDEF(boolean)
+compress_output(j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
+  int blkn, ci, xindex, yindex, yoffset;
+  JDIMENSION start_col;
+  JBLOCKARRAY bufferMAX_COMPS_IN_SCAN;
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  /* Align the virtual buffers for the components used in this scan.
+   * NB: during first pass, this is safe only because the buffers will
+   * already be aligned properly, so jmemmgr.c won't need to do any I/O.
+   */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    bufferci = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr)cinfo, coef->whole_imagecompptr->component_index,
+       coef->iMCU_row_num * compptr->v_samp_factor,
+       (JDIMENSION)compptr->v_samp_factor, FALSE);
+  }
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
+         MCU_col_num++) {
+      /* Construct list of pointers to DCT blocks belonging to this MCU */
+      blkn = 0;                 /* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+        compptr = cinfo->cur_comp_infoci;
+        start_col = MCU_col_num * compptr->MCU_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          buffer_ptr = bufferciyindex + yoffset + start_col;
+          for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+            coef->MCU_bufferblkn++ = buffer_ptr++;
+          }
+        }
+      }
+      /* Try to write the MCU. */
+      if (!(*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->mcu_ctr = MCU_col_num;
+        return FALSE;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->mcu_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  coef->iMCU_row_num++;
+  start_iMCU_row(cinfo);
+  return TRUE;
+}
+
+#endif /* FULL_COEF_BUFFER_SUPPORTED */
+
+
+/*
+ * Initialize coefficient buffer controller.
+ */
+
+GLOBAL(void)
+jinit_c_coef_controller(j_compress_ptr cinfo, boolean need_full_buffer)
+{
+  my_coef_ptr coef;
+
+  coef = (my_coef_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_coef_controller));
+  cinfo->coef = (struct jpeg_c_coef_controller *)coef;
+  coef->pub.start_pass = start_pass_coef;
+
+  /* Create the coefficient buffer. */
+  if (need_full_buffer) {
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+    /* Allocate a full-image virtual array for each component, */
+    /* padded to a multiple of samp_factor DCT blocks in each direction. */
+    int ci;
+    jpeg_component_info *compptr;
+
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+         ci++, compptr++) {
+      coef->whole_imageci = (*cinfo->mem->request_virt_barray)
+        ((j_common_ptr)cinfo, JPOOL_IMAGE, FALSE,
+         (JDIMENSION)jround_up((long)compptr->width_in_blocks,
+                               (long)compptr->h_samp_factor),
+         (JDIMENSION)jround_up((long)compptr->height_in_blocks,
+                               (long)compptr->v_samp_factor),
+         (JDIMENSION)compptr->v_samp_factor);
+    }
+#else
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif
+  } else {
+    /* We only need a single-MCU buffer. */
+    JBLOCKROW buffer;
+    int i;
+
+    buffer = (JBLOCKROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  C_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
+    for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
+      coef->MCU_bufferi = buffer + i;
+    }
+    coef->whole_image0 = NULL; /* flag for no virtual arrays */
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jccolext.c Added

@@ -0,0 +1,144 @@
+/*
+ * jccolext.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2012, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains input colorspace conversion routines.
+ */
+
+
+/* This file is included by jccolor.c */
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ *
+ * Note that we change from the application's interleaved-pixel format
+ * to our internal noninterleaved, one-plane-per-component format.
+ * The input buffer is therefore three times as wide as the output buffer.
+ *
+ * A starting row offset is provided only for the output buffer.  The caller
+ * can easily adjust the passed input_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+INLINE
+LOCAL(void)
+rgb_ycc_convert_internal(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                         JSAMPIMAGE output_buf, JDIMENSION output_row,
+                         int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  register int r, g, b;
+  register JLONG *ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf0output_row;
+    outptr1 = output_buf1output_row;
+    outptr2 = output_buf2output_row;
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = inptrRGB_RED;
+      g = inptrRGB_GREEN;
+      b = inptrRGB_BLUE;
+      inptr += RGB_PIXELSIZE;
+      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
+       * must be too; we do not need an explicit range-limiting operation.
+       * Hence the value being shifted is never negative, and we don't
+       * need the general RIGHT_SHIFT macro.
+       */
+      /* Y */
+      outptr0col = (JSAMPLE)((ctabr + R_Y_OFF + ctabg + G_Y_OFF +
+                                ctabb + B_Y_OFF) >> SCALEBITS);
+      /* Cb */
+      outptr1col = (JSAMPLE)((ctabr + R_CB_OFF + ctabg + G_CB_OFF +
+                                ctabb + B_CB_OFF) >> SCALEBITS);
+      /* Cr */
+      outptr2col = (JSAMPLE)((ctabr + R_CR_OFF + ctabg + G_CR_OFF +
+                                ctabb + B_CR_OFF) >> SCALEBITS);
+    }
+  }
+}
+
+
+/**************** Cases other than RGB -> YCbCr **************/
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles RGB->grayscale conversion, which is the same
+ * as the RGB->Y portion of RGB->YCbCr.
+ * We assume rgb_ycc_start has been called (we only use the Y tables).
+ */
+
+INLINE
+LOCAL(void)
+rgb_gray_convert_internal(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                          JSAMPIMAGE output_buf, JDIMENSION output_row,
+                          int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  register int r, g, b;
+  register JLONG *ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf0output_row;
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = inptrRGB_RED;
+      g = inptrRGB_GREEN;
+      b = inptrRGB_BLUE;
+      inptr += RGB_PIXELSIZE;
+      /* Y */
+      outptrcol = (JSAMPLE)((ctabr + R_Y_OFF + ctabg + G_Y_OFF +
+                               ctabb + B_Y_OFF) >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles extended RGB->plain RGB conversion
+ */
+
+INLINE
+LOCAL(void)
+rgb_rgb_convert_internal(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                         JSAMPIMAGE output_buf, JDIMENSION output_row,
+                         int num_rows)
+{
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf0output_row;
+    outptr1 = output_buf1output_row;
+    outptr2 = output_buf2output_row;
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      outptr0col = inptrRGB_RED;
+      outptr1col = inptrRGB_GREEN;
+      outptr2col = inptrRGB_BLUE;
+      inptr += RGB_PIXELSIZE;
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jccolor.c Added

@@ -0,0 +1,710 @@
+/*
+ * jccolor.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2009-2012, 2015, D. R. Commander.
+ * Copyright (C) 2014, MIPS Technologies, Inc., California.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains input colorspace conversion routines.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jsimd.h"
+#include "jconfigint.h"
+
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_color_converter pub; /* public fields */
+
+  /* Private state for RGB->YCC conversion */
+  JLONG *rgb_ycc_tab;           /* => table for RGB to YCbCr conversion */
+} my_color_converter;
+
+typedef my_color_converter *my_cconvert_ptr;
+
+
+/**************** RGB -> YCbCr conversion: most common case **************/
+
+/*
+ * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
+ * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
+ * The conversion equations to be implemented are therefore
+ *      Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+ *      Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + CENTERJSAMPLE
+ *      Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B  + CENTERJSAMPLE
+ * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
+ * Note: older versions of the IJG code used a zero offset of MAXJSAMPLE/2,
+ * rather than CENTERJSAMPLE, for Cb and Cr.  This gave equal positive and
+ * negative swings for Cb/Cr, but meant that grayscale values (Cb=Cr=0)
+ * were not represented exactly.  Now we sacrifice exact representation of
+ * maximum red and maximum blue in order to get exact grayscales.
+ *
+ * To avoid floating-point arithmetic, we represent the fractional constants
+ * as integers scaled up by 2^16 (about 4 digits precision); we have to divide
+ * the products by 2^16, with appropriate rounding, to get the correct answer.
+ *
+ * For even more speed, we avoid doing any multiplications in the inner loop
+ * by precalculating the constants times R,G,B for all possible values.
+ * For 8-bit JSAMPLEs this is very reasonable (only 256 entries per table);
+ * for 12-bit samples it is still acceptable.  It's not very reasonable for
+ * 16-bit samples, but if you want lossless storage you shouldn't be changing
+ * colorspace anyway.
+ * The CENTERJSAMPLE offsets and the rounding fudge-factor of 0.5 are included
+ * in the tables to save adding them separately in the inner loop.
+ */
+
+#define SCALEBITS       16      /* speediest right-shift on some machines */
+#define CBCR_OFFSET     ((JLONG)CENTERJSAMPLE << SCALEBITS)
+#define ONE_HALF        ((JLONG)1 << (SCALEBITS - 1))
+#define FIX(x)          ((JLONG)((x) * (1L << SCALEBITS) + 0.5))
+
+/* We allocate one big table and divide it up into eight parts, instead of
+ * doing eight alloc_small requests.  This lets us use a single table base
+ * address, which can be held in a register in the inner loops on many
+ * machines (more than can hold all eight addresses, anyway).
+ */
+
+#define R_Y_OFF         0                       /* offset to R => Y section */
+#define G_Y_OFF         (1 * (MAXJSAMPLE + 1))  /* offset to G => Y section */
+#define B_Y_OFF         (2 * (MAXJSAMPLE + 1))  /* etc. */
+#define R_CB_OFF        (3 * (MAXJSAMPLE + 1))
+#define G_CB_OFF        (4 * (MAXJSAMPLE + 1))
+#define B_CB_OFF        (5 * (MAXJSAMPLE + 1))
+#define R_CR_OFF        B_CB_OFF                /* B=>Cb, R=>Cr are the same */
+#define G_CR_OFF        (6 * (MAXJSAMPLE + 1))
+#define B_CR_OFF        (7 * (MAXJSAMPLE + 1))
+#define TABLE_SIZE      (8 * (MAXJSAMPLE + 1))
+
+
+/* Include inline routines for colorspace extensions */
+
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define rgb_ycc_convert_internal  extrgb_ycc_convert_internal
+#define rgb_gray_convert_internal  extrgb_gray_convert_internal
+#define rgb_rgb_convert_internal  extrgb_rgb_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define rgb_ycc_convert_internal  extrgbx_ycc_convert_internal
+#define rgb_gray_convert_internal  extrgbx_gray_convert_internal
+#define rgb_rgb_convert_internal  extrgbx_rgb_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define rgb_ycc_convert_internal  extbgr_ycc_convert_internal
+#define rgb_gray_convert_internal  extbgr_gray_convert_internal
+#define rgb_rgb_convert_internal  extbgr_rgb_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define rgb_ycc_convert_internal  extbgrx_ycc_convert_internal
+#define rgb_gray_convert_internal  extbgrx_gray_convert_internal
+#define rgb_rgb_convert_internal  extbgrx_rgb_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define rgb_ycc_convert_internal  extxbgr_ycc_convert_internal
+#define rgb_gray_convert_internal  extxbgr_gray_convert_internal
+#define rgb_rgb_convert_internal  extxbgr_rgb_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define rgb_ycc_convert_internal  extxrgb_ycc_convert_internal
+#define rgb_gray_convert_internal  extxrgb_gray_convert_internal
+#define rgb_rgb_convert_internal  extxrgb_rgb_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
+
+
+/*
+ * Initialize for RGB->YCC colorspace conversion.
+ */
+
+METHODDEF(void)
+rgb_ycc_start(j_compress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  JLONG *rgb_ycc_tab;
+  JLONG i;
+
+  /* Allocate and fill in the conversion tables. */
+  cconvert->rgb_ycc_tab = rgb_ycc_tab = (JLONG *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (TABLE_SIZE * sizeof(JLONG)));
+
+  for (i = 0; i <= MAXJSAMPLE; i++) {
+    rgb_ycc_tabi + R_Y_OFF = FIX(0.29900) * i;
+    rgb_ycc_tabi + G_Y_OFF = FIX(0.58700) * i;
+    rgb_ycc_tabi + B_Y_OFF = FIX(0.11400) * i   + ONE_HALF;
+    rgb_ycc_tabi + R_CB_OFF = (-FIX(0.16874)) * i;
+    rgb_ycc_tabi + G_CB_OFF = (-FIX(0.33126)) * i;
+    /* We use a rounding fudge-factor of 0.5-epsilon for Cb and Cr.
+     * This ensures that the maximum output will round to MAXJSAMPLE
+     * not MAXJSAMPLE+1, and thus that we don't have to range-limit.
+     */
+    rgb_ycc_tabi + B_CB_OFF = FIX(0.50000) * i  + CBCR_OFFSET + ONE_HALF - 1;
+/*  B=>Cb and R=>Cr tables are the same
+    rgb_ycc_tabi + R_CR_OFF = FIX(0.50000) * i  + CBCR_OFFSET + ONE_HALF - 1;
+*/
+    rgb_ycc_tabi + G_CR_OFF = (-FIX(0.41869)) * i;
+    rgb_ycc_tabi + B_CR_OFF = (-FIX(0.08131)) * i;
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ */
+
+METHODDEF(void)
+rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                JSAMPIMAGE output_buf, JDIMENSION output_row, int num_rows)
+{
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    extrgb_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                num_rows);
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    extrgbx_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  case JCS_EXT_BGR:
+    extbgr_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                num_rows);
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    extbgrx_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    extxbgr_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    extxrgb_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  default:
+    rgb_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                             num_rows);
+    break;
+  }
+}
+
+
+/**************** Cases other than RGB -> YCbCr **************/
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ */
+
+METHODDEF(void)
+rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                 JSAMPIMAGE output_buf, JDIMENSION output_row, int num_rows)
+{
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    extrgb_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    extrgbx_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+    break;
+  case JCS_EXT_BGR:
+    extbgr_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    extbgrx_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    extxbgr_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    extxrgb_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+    break;
+  default:
+    rgb_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                              num_rows);
+    break;
+  }
+}
+
+
+/*
+ * Extended RGB to plain RGB conversion
+ */
+
+METHODDEF(void)
+rgb_rgb_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                JSAMPIMAGE output_buf, JDIMENSION output_row, int num_rows)
+{
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    extrgb_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                num_rows);
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    extrgbx_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  case JCS_EXT_BGR:
+    extbgr_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                num_rows);
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    extbgrx_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    extxbgr_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    extxrgb_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                 num_rows);
+    break;
+  default:
+    rgb_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                             num_rows);
+    break;
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles Adobe-style CMYK->YCCK conversion,
+ * where we convert R=1-C, G=1-M, and B=1-Y to YCbCr using the same
+ * conversion as above, while passing K (black) unchanged.
+ * We assume rgb_ycc_start has been called.
+ */
+
+METHODDEF(void)
+cmyk_ycck_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                  JSAMPIMAGE output_buf, JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  register int r, g, b;
+  register JLONG *ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2, outptr3;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf0output_row;
+    outptr1 = output_buf1output_row;
+    outptr2 = output_buf2output_row;
+    outptr3 = output_buf3output_row;
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = MAXJSAMPLE - inptr0;
+      g = MAXJSAMPLE - inptr1;
+      b = MAXJSAMPLE - inptr2;
+      /* K passes through as-is */
+      outptr3col = inptr3;
+      inptr += 4;
+      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
+       * must be too; we do not need an explicit range-limiting operation.
+       * Hence the value being shifted is never negative, and we don't
+       * need the general RIGHT_SHIFT macro.
+       */
+      /* Y */
+      outptr0col = (JSAMPLE)((ctabr + R_Y_OFF + ctabg + G_Y_OFF +
+                                ctabb + B_Y_OFF) >> SCALEBITS);
+      /* Cb */
+      outptr1col = (JSAMPLE)((ctabr + R_CB_OFF + ctabg + G_CB_OFF +
+                                ctabb + B_CB_OFF) >> SCALEBITS);
+      /* Cr */
+      outptr2col = (JSAMPLE)((ctabr + R_CR_OFF + ctabg + G_CR_OFF +
+                                ctabb + B_CR_OFF) >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles grayscale output with no conversion.
+ * The source can be either plain grayscale or YCbCr (since Y == gray).
+ */
+
+METHODDEF(void)
+grayscale_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                  JSAMPIMAGE output_buf, JDIMENSION output_row, int num_rows)
+{
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+  int instride = cinfo->input_components;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf0output_row;
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      outptrcol = inptr0;
+      inptr += instride;
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles multi-component colorspaces without conversion.
+ * We assume input_components == num_components.
+ */
+
+METHODDEF(void)
+null_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows)
+{
+  register JSAMPROW inptr;
+  register JSAMPROW outptr, outptr0, outptr1, outptr2, outptr3;
+  register JDIMENSION col;
+  register int ci;
+  int nc = cinfo->num_components;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  if (nc == 3) {
+    while (--num_rows >= 0) {
+      inptr = *input_buf++;
+      outptr0 = output_buf0output_row;
+      outptr1 = output_buf1output_row;
+      outptr2 = output_buf2output_row;
+      output_row++;
+      for (col = 0; col < num_cols; col++) {
+        outptr0col = *inptr++;
+        outptr1col = *inptr++;
+        outptr2col = *inptr++;
+      }
+    }
+  } else if (nc == 4) {
+    while (--num_rows >= 0) {
+      inptr = *input_buf++;
+      outptr0 = output_buf0output_row;
+      outptr1 = output_buf1output_row;
+      outptr2 = output_buf2output_row;
+      outptr3 = output_buf3output_row;
+      output_row++;
+      for (col = 0; col < num_cols; col++) {
+        outptr0col = *inptr++;
+        outptr1col = *inptr++;
+        outptr2col = *inptr++;
+        outptr3col = *inptr++;
+      }
+    }
+  } else {
+    while (--num_rows >= 0) {
+      /* It seems fastest to make a separate pass for each component. */
+      for (ci = 0; ci < nc; ci++) {
+        inptr = *input_buf;
+        outptr = output_bufcioutput_row;
+        for (col = 0; col < num_cols; col++) {
+          outptrcol = inptrci;
+          inptr += nc;
+        }
+      }
+      input_buf++;
+      output_row++;
+    }
+  }
+}
+
+
+/*
+ * Empty method for start_pass.
+ */
+
+METHODDEF(void)
+null_method(j_compress_ptr cinfo)
+{
+  /* no work needed */
+}
+
+
+/*
+ * Module initialization routine for input colorspace conversion.
+ */
+
+GLOBAL(void)
+jinit_color_converter(j_compress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert;
+
+  cconvert = (my_cconvert_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_color_converter));
+  cinfo->cconvert = (struct jpeg_color_converter *)cconvert;
+  /* set start_pass to null method until we find out differently */
+  cconvert->pub.start_pass = null_method;
+
+  /* Make sure input_components agrees with in_color_space */
+  switch (cinfo->in_color_space) {
+  case JCS_GRAYSCALE:
+    if (cinfo->input_components != 1)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  case JCS_RGB:
+  case JCS_EXT_RGB:
+  case JCS_EXT_RGBX:
+  case JCS_EXT_BGR:
+  case JCS_EXT_BGRX:
+  case JCS_EXT_XBGR:
+  case JCS_EXT_XRGB:
+  case JCS_EXT_RGBA:
+  case JCS_EXT_BGRA:
+  case JCS_EXT_ABGR:
+  case JCS_EXT_ARGB:
+    if (cinfo->input_components != rgb_pixelsizecinfo->in_color_space)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  case JCS_YCbCr:
+    if (cinfo->input_components != 3)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  case JCS_CMYK:
+  case JCS_YCCK:
+    if (cinfo->input_components != 4)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  default:                      /* JCS_UNKNOWN can be anything */
+    if (cinfo->input_components < 1)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+  }
+
+  /* Check num_components, set conversion method based on requested space */
+  switch (cinfo->jpeg_color_space) {
+  case JCS_GRAYSCALE:
+    if (cinfo->num_components != 1)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_GRAYSCALE)
+      cconvert->pub.color_convert = grayscale_convert;
+    else if (cinfo->in_color_space == JCS_RGB ||
+             cinfo->in_color_space == JCS_EXT_RGB ||
+             cinfo->in_color_space == JCS_EXT_RGBX ||
+             cinfo->in_color_space == JCS_EXT_BGR ||
+             cinfo->in_color_space == JCS_EXT_BGRX ||
+             cinfo->in_color_space == JCS_EXT_XBGR ||
+             cinfo->in_color_space == JCS_EXT_XRGB ||
+             cinfo->in_color_space == JCS_EXT_RGBA ||
+             cinfo->in_color_space == JCS_EXT_BGRA ||
+             cinfo->in_color_space == JCS_EXT_ABGR ||
+             cinfo->in_color_space == JCS_EXT_ARGB) {
+      if (jsimd_can_rgb_gray())
+        cconvert->pub.color_convert = jsimd_rgb_gray_convert;
+      else {
+        cconvert->pub.start_pass = rgb_ycc_start;
+        cconvert->pub.color_convert = rgb_gray_convert;
+      }
+    } else if (cinfo->in_color_space == JCS_YCbCr)
+      cconvert->pub.color_convert = grayscale_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_RGB:
+    if (cinfo->num_components != 3)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (rgb_redcinfo->in_color_space == 0 &&
+        rgb_greencinfo->in_color_space == 1 &&
+        rgb_bluecinfo->in_color_space == 2 &&
+        rgb_pixelsizecinfo->in_color_space == 3) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else if (cinfo->in_color_space == JCS_RGB ||
+               cinfo->in_color_space == JCS_EXT_RGB ||
+               cinfo->in_color_space == JCS_EXT_RGBX ||
+               cinfo->in_color_space == JCS_EXT_BGR ||
+               cinfo->in_color_space == JCS_EXT_BGRX ||
+               cinfo->in_color_space == JCS_EXT_XBGR ||
+               cinfo->in_color_space == JCS_EXT_XRGB ||
+               cinfo->in_color_space == JCS_EXT_RGBA ||
+               cinfo->in_color_space == JCS_EXT_BGRA ||
+               cinfo->in_color_space == JCS_EXT_ABGR ||
+               cinfo->in_color_space == JCS_EXT_ARGB)
+      cconvert->pub.color_convert = rgb_rgb_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_YCbCr:
+    if (cinfo->num_components != 3)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_RGB ||
+        cinfo->in_color_space == JCS_EXT_RGB ||
+        cinfo->in_color_space == JCS_EXT_RGBX ||
+        cinfo->in_color_space == JCS_EXT_BGR ||
+        cinfo->in_color_space == JCS_EXT_BGRX ||
+        cinfo->in_color_space == JCS_EXT_XBGR ||
+        cinfo->in_color_space == JCS_EXT_XRGB ||
+        cinfo->in_color_space == JCS_EXT_RGBA ||
+        cinfo->in_color_space == JCS_EXT_BGRA ||
+        cinfo->in_color_space == JCS_EXT_ABGR ||
+        cinfo->in_color_space == JCS_EXT_ARGB) {
+      if (jsimd_can_rgb_ycc())
+        cconvert->pub.color_convert = jsimd_rgb_ycc_convert;
+      else {
+        cconvert->pub.start_pass = rgb_ycc_start;
+        cconvert->pub.color_convert = rgb_ycc_convert;
+      }
+    } else if (cinfo->in_color_space == JCS_YCbCr) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_CMYK:
+    if (cinfo->num_components != 4)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_CMYK) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_YCCK:
+    if (cinfo->num_components != 4)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_CMYK) {
+      cconvert->pub.start_pass = rgb_ycc_start;
+      cconvert->pub.color_convert = cmyk_ycck_convert;
+    } else if (cinfo->in_color_space == JCS_YCCK) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  default:                      /* allow null conversion of JCS_UNKNOWN */
+    if (cinfo->jpeg_color_space != cinfo->in_color_space ||
+        cinfo->num_components != cinfo->input_components)
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+#if defined(__mips__)
+    if (jsimd_c_can_null_convert())
+      cconvert->pub.color_convert = jsimd_c_null_convert;
+    else
+#endif
+      cconvert->pub.color_convert = null_convert;
+    break;
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcdctmgr.c Added

@@ -0,0 +1,720 @@
+/*
+ * jcdctmgr.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2011, 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the forward-DCT management logic.
+ * This code selects a particular DCT implementation to be used,
+ * and it performs related housekeeping chores including coefficient
+ * quantization.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"               /* Private declarations for DCT subsystem */
+#include "jsimddct.h"
+
+
+/* Private subobject for this module */
+
+typedef void (*forward_DCT_method_ptr) (DCTELEM *data);
+typedef void (*float_DCT_method_ptr) (FAST_FLOAT *data);
+
+typedef void (*convsamp_method_ptr) (JSAMPARRAY sample_data,
+                                     JDIMENSION start_col,
+                                     DCTELEM *workspace);
+typedef void (*float_convsamp_method_ptr) (JSAMPARRAY sample_data,
+                                           JDIMENSION start_col,
+                                           FAST_FLOAT *workspace);
+
+typedef void (*quantize_method_ptr) (JCOEFPTR coef_block, DCTELEM *divisors,
+                                     DCTELEM *workspace);
+typedef void (*float_quantize_method_ptr) (JCOEFPTR coef_block,
+                                           FAST_FLOAT *divisors,
+                                           FAST_FLOAT *workspace);
+
+METHODDEF(void) quantize(JCOEFPTR, DCTELEM *, DCTELEM *);
+
+typedef struct {
+  struct jpeg_forward_dct pub;  /* public fields */
+
+  /* Pointer to the DCT routine actually in use */
+  forward_DCT_method_ptr dct;
+  convsamp_method_ptr convsamp;
+  quantize_method_ptr quantize;
+
+  /* The actual post-DCT divisors --- not identical to the quant table
+   * entries, because of scaling (especially for an unnormalized DCT).
+   * Each table is given in normal array order.
+   */
+  DCTELEM *divisorsNUM_QUANT_TBLS;
+
+  /* work area for FDCT subroutine */
+  DCTELEM *workspace;
+
+#ifdef DCT_FLOAT_SUPPORTED
+  /* Same as above for the floating-point case. */
+  float_DCT_method_ptr float_dct;
+  float_convsamp_method_ptr float_convsamp;
+  float_quantize_method_ptr float_quantize;
+  FAST_FLOAT *float_divisorsNUM_QUANT_TBLS;
+  FAST_FLOAT *float_workspace;
+#endif
+} my_fdct_controller;
+
+typedef my_fdct_controller *my_fdct_ptr;
+
+
+#if BITS_IN_JSAMPLE == 8
+
+/*
+ * Find the highest bit in an integer through binary search.
+ */
+
+LOCAL(int)
+flss(UINT16 val)
+{
+  int bit;
+
+  bit = 16;
+
+  if (!val)
+    return 0;
+
+  if (!(val & 0xff00)) {
+    bit -= 8;
+    val <<= 8;
+  }
+  if (!(val & 0xf000)) {
+    bit -= 4;
+    val <<= 4;
+  }
+  if (!(val & 0xc000)) {
+    bit -= 2;
+    val <<= 2;
+  }
+  if (!(val & 0x8000)) {
+    bit -= 1;
+    val <<= 1;
+  }
+
+  return bit;
+}
+
+
+/*
+ * Compute values to do a division using reciprocal.
+ *
+ * This implementation is based on an algorithm described in
+ *   "How to optimize for the Pentium family of microprocessors"
+ *   (http://www.agner.org/assem/).
+ * More information about the basic algorithm can be found in
+ * the paper "Integer Division Using Reciprocals" by Robert Alverson.
+ *
+ * The basic idea is to replace x/d by x * d^-1. In order to store
+ * d^-1 with enough precision we shift it left a few places. It turns
+ * out that this algoright gives just enough precision, and also fits
+ * into DCTELEM:
+ *
+ *   b = (the number of significant bits in divisor) - 1
+ *   r = (word size) + b
+ *   f = 2^r / divisor
+ *
+ * f will not be an integer for most cases, so we need to compensate
+ * for the rounding error introduced:
+ *
+ *   no fractional part:
+ *
+ *       result = input >> r
+ *
+ *   fractional part of f < 0.5:
+ *
+ *       round f down to nearest integer
+ *       result = ((input + 1) * f) >> r
+ *
+ *   fractional part of f > 0.5:
+ *
+ *       round f up to nearest integer
+ *       result = (input * f) >> r
+ *
+ * This is the original algorithm that gives truncated results. But we
+ * want properly rounded results, so we replace "input" with
+ * "input + divisor/2".
+ *
+ * In order to allow SIMD implementations we also tweak the values to
+ * allow the same calculation to be made at all times:
+ *
+ *   dctbl0 = f rounded to nearest integer
+ *   dctbl1 = divisor / 2 (+ 1 if fractional part of f < 0.5)
+ *   dctbl2 = 1 << ((word size) * 2 - r)
+ *   dctbl3 = r - (word size)
+ *
+ * dctbl2 is for stupid instruction sets where the shift operation
+ * isn't member wise (e.g. MMX).
+ *
+ * The reason dctbl2 and dctbl3 reduce the shift with (word size)
+ * is that most SIMD implementations have a "multiply and store top
+ * half" operation.
+ *
+ * Lastly, we store each of the values in their own table instead
+ * of in a consecutive manner, yet again in order to allow SIMD
+ * routines.
+ */
+
+LOCAL(int)
+compute_reciprocal(UINT16 divisor, DCTELEM *dtbl)
+{
+  UDCTELEM2 fq, fr;
+  UDCTELEM c;
+  int b, r;
+
+  if (divisor == 1) {
+    /* divisor == 1 means unquantized, so these reciprocal/correction/shift
+     * values will cause the C quantization algorithm to act like the
+     * identity function.  Since only the C quantization algorithm is used in
+     * these cases, the scale value is irrelevant.
+     */
+    dtblDCTSIZE2 * 0 = (DCTELEM)1;                        /* reciprocal */
+    dtblDCTSIZE2 * 1 = (DCTELEM)0;                        /* correction */
+    dtblDCTSIZE2 * 2 = (DCTELEM)1;                        /* scale */
+    dtblDCTSIZE2 * 3 = -(DCTELEM)(sizeof(DCTELEM) * 8);   /* shift */
+    return 0;
+  }
+
+  b = flss(divisor) - 1;
+  r  = sizeof(DCTELEM) * 8 + b;
+
+  fq = ((UDCTELEM2)1 << r) / divisor;
+  fr = ((UDCTELEM2)1 << r) % divisor;
+
+  c = divisor / 2;                      /* for rounding */
+
+  if (fr == 0) {                        /* divisor is power of two */
+    /* fq will be one bit too large to fit in DCTELEM, so adjust */
+    fq >>= 1;
+    r--;
+  } else if (fr <= (divisor / 2U)) {    /* fractional part is < 0.5 */
+    c++;
+  } else {                              /* fractional part is > 0.5 */
+    fq++;
+  }
+
+  dtblDCTSIZE2 * 0 = (DCTELEM)fq;     /* reciprocal */
+  dtblDCTSIZE2 * 1 = (DCTELEM)c;      /* correction + roundfactor */
+#ifdef WITH_SIMD
+  dtblDCTSIZE2 * 2 = (DCTELEM)(1 << (sizeof(DCTELEM) * 8 * 2 - r)); /* scale */
+#else
+  dtblDCTSIZE2 * 2 = 1;
+#endif
+  dtblDCTSIZE2 * 3 = (DCTELEM)r - sizeof(DCTELEM) * 8; /* shift */
+
+  if (r <= 16) return 0;
+  else return 1;
+}
+
+#endif
+
+
+/*
+ * Initialize for a processing pass.
+ * Verify that all referenced Q-tables are present, and set up
+ * the divisor table for each one.
+ * In the current implementation, DCT of all components is done during
+ * the first pass, even if only some components will be output in the
+ * first scan.  Hence all components should be examined here.
+ */
+
+METHODDEF(void)
+start_pass_fdctmgr(j_compress_ptr cinfo)
+{
+  my_fdct_ptr fdct = (my_fdct_ptr)cinfo->fdct;
+  int ci, qtblno, i;
+  jpeg_component_info *compptr;
+  JQUANT_TBL *qtbl;
+  DCTELEM *dtbl;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    qtblno = compptr->quant_tbl_no;
+    /* Make sure specified quantization table is present */
+    if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
+        cinfo->quant_tbl_ptrsqtblno == NULL)
+      ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
+    qtbl = cinfo->quant_tbl_ptrsqtblno;
+    /* Compute divisors for this quant table */
+    /* We may do this more than once for same table, but it's not a big deal */
+    switch (cinfo->dct_method) {
+#ifdef DCT_ISLOW_SUPPORTED
+    case JDCT_ISLOW:
+      /* For LL&M IDCT method, divisors are equal to raw quantization
+       * coefficients multiplied by 8 (to counteract scaling).
+       */
+      if (fdct->divisorsqtblno == NULL) {
+        fdct->divisorsqtblno = (DCTELEM *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                      (DCTSIZE2 * 4) * sizeof(DCTELEM));
+      }
+      dtbl = fdct->divisorsqtblno;
+      for (i = 0; i < DCTSIZE2; i++) {
+#if BITS_IN_JSAMPLE == 8
+        if (!compute_reciprocal(qtbl->quantvali << 3, &dtbli) &&
+            fdct->quantize == jsimd_quantize)
+          fdct->quantize = quantize;
+#else
+        dtbli = ((DCTELEM)qtbl->quantvali) << 3;
+#endif
+      }
+      break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+    case JDCT_IFAST:
+      {
+        /* For AA&N IDCT method, divisors are equal to quantization
+         * coefficients scaled by scalefactorrow*scalefactorcol, where
+         *   scalefactor0 = 1
+         *   scalefactork = cos(k*PI/16) * sqrt(2)    for k=1..7
+         * We apply a further scale factor of 8.
+         */
+#define CONST_BITS  14
+        static const INT16 aanscalesDCTSIZE2 = {
+          /* precomputed values scaled up by 14 bits */
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+          21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+          19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+           8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+           4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+        };
+        SHIFT_TEMPS
+
+        if (fdct->divisorsqtblno == NULL) {
+          fdct->divisorsqtblno = (DCTELEM *)
+            (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                        (DCTSIZE2 * 4) * sizeof(DCTELEM));
+        }
+        dtbl = fdct->divisorsqtblno;
+        for (i = 0; i < DCTSIZE2; i++) {
+#if BITS_IN_JSAMPLE == 8
+          if (!compute_reciprocal(
+                DESCALE(MULTIPLY16V16((JLONG)qtbl->quantvali,
+                                      (JLONG)aanscalesi),
+                        CONST_BITS - 3), &dtbli) &&
+              fdct->quantize == jsimd_quantize)
+            fdct->quantize = quantize;
+#else
+          dtbli = (DCTELEM)
+            DESCALE(MULTIPLY16V16((JLONG)qtbl->quantvali,
+                                  (JLONG)aanscalesi),
+                    CONST_BITS - 3);
+#endif
+        }
+      }
+      break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+    case JDCT_FLOAT:
+      {
+        /* For float AA&N IDCT method, divisors are equal to quantization
+         * coefficients scaled by scalefactorrow*scalefactorcol, where
+         *   scalefactor0 = 1
+         *   scalefactork = cos(k*PI/16) * sqrt(2)    for k=1..7
+         * We apply a further scale factor of 8.
+         * What's actually stored is 1/divisor so that the inner loop can
+         * use a multiplication rather than a division.
+         */
+        FAST_FLOAT *fdtbl;
+        int row, col;
+        static const double aanscalefactorDCTSIZE = {
+          1.0, 1.387039845, 1.306562965, 1.175875602,
+          1.0, 0.785694958, 0.541196100, 0.275899379
+        };
+
+        if (fdct->float_divisorsqtblno == NULL) {
+          fdct->float_divisorsqtblno = (FAST_FLOAT *)
+            (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                        DCTSIZE2 * sizeof(FAST_FLOAT));
+        }
+        fdtbl = fdct->float_divisorsqtblno;
+        i = 0;
+        for (row = 0; row < DCTSIZE; row++) {
+          for (col = 0; col < DCTSIZE; col++) {
+            fdtbli = (FAST_FLOAT)
+              (1.0 / (((double)qtbl->quantvali *
+                       aanscalefactorrow * aanscalefactorcol * 8.0)));
+            i++;
+          }
+        }
+      }
+      break;
+#endif
+    default:
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+      break;
+    }
+  }
+}
+
+
+/*
+ * Load data into workspace, applying unsigned->signed conversion.
+ */
+
+METHODDEF(void)
+convsamp(JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace)
+{
+  register DCTELEM *workspaceptr;
+  register JSAMPROW elemptr;
+  register int elemr;
+
+  workspaceptr = workspace;
+  for (elemr = 0; elemr < DCTSIZE; elemr++) {
+    elemptr = sample_dataelemr + start_col;
+
+#if DCTSIZE == 8                /* unroll the inner loop */
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+#else
+    {
+      register int elemc;
+      for (elemc = DCTSIZE; elemc > 0; elemc--)
+        *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    }
+#endif
+  }
+}
+
+
+/*
+ * Quantize/descale the coefficients, and store into coef_blocks.
+ */
+
+METHODDEF(void)
+quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
+{
+  int i;
+  DCTELEM temp;
+  JCOEFPTR output_ptr = coef_block;
+
+#if BITS_IN_JSAMPLE == 8
+
+  UDCTELEM recip, corr;
+  int shift;
+  UDCTELEM2 product;
+
+  for (i = 0; i < DCTSIZE2; i++) {
+    temp = workspacei;
+    recip = divisorsi + DCTSIZE2 * 0;
+    corr =  divisorsi + DCTSIZE2 * 1;
+    shift = divisorsi + DCTSIZE2 * 3;
+
+    if (temp < 0) {
+      temp = -temp;
+      product = (UDCTELEM2)(temp + corr) * recip;
+      product >>= shift + sizeof(DCTELEM) * 8;
+      temp = (DCTELEM)product;
+      temp = -temp;
+    } else {
+      product = (UDCTELEM2)(temp + corr) * recip;
+      product >>= shift + sizeof(DCTELEM) * 8;
+      temp = (DCTELEM)product;
+    }
+    output_ptri = (JCOEF)temp;
+  }
+
+#else
+
+  register DCTELEM qval;
+
+  for (i = 0; i < DCTSIZE2; i++) {
+    qval = divisorsi;
+    temp = workspacei;
+    /* Divide the coefficient value by qval, ensuring proper rounding.
+     * Since C does not specify the direction of rounding for negative
+     * quotients, we have to force the dividend positive for portability.
+     *
+     * In most files, at least half of the output values will be zero
+     * (at default quantization settings, more like three-quarters...)
+     * so we should ensure that this case is fast.  On many machines,
+     * a comparison is enough cheaper than a divide to make a special test
+     * a win.  Since both inputs will be nonnegative, we need only test
+     * for a < b to discover whether a/b is 0.
+     * If your machine's division is fast enough, define FAST_DIVIDE.
+     */
+#ifdef FAST_DIVIDE
+#define DIVIDE_BY(a, b)  a /= b
+#else
+#define DIVIDE_BY(a, b)  if (a >= b) a /= b;  else a = 0
+#endif
+    if (temp < 0) {
+      temp = -temp;
+      temp += qval >> 1;        /* for rounding */
+      DIVIDE_BY(temp, qval);
+      temp = -temp;
+    } else {
+      temp += qval >> 1;        /* for rounding */
+      DIVIDE_BY(temp, qval);
+    }
+    output_ptri = (JCOEF)temp;
+  }
+
+#endif
+
+}
+
+
+/*
+ * Perform forward DCT on one or more blocks of a component.
+ *
+ * The input samples are taken from the sample_data array starting at
+ * position start_row/start_col, and moving to the right for any additional
+ * blocks. The quantized coefficients are returned in coef_blocks.
+ */
+
+METHODDEF(void)
+forward_DCT(j_compress_ptr cinfo, jpeg_component_info *compptr,
+            JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+            JDIMENSION start_row, JDIMENSION start_col, JDIMENSION num_blocks)
+/* This version is used for integer DCT implementations. */
+{
+  /* This routine is heavily used, so it's worth coding it tightly. */
+  my_fdct_ptr fdct = (my_fdct_ptr)cinfo->fdct;
+  DCTELEM *divisors = fdct->divisorscompptr->quant_tbl_no;
+  DCTELEM *workspace;
+  JDIMENSION bi;
+
+  /* Make sure the compiler doesn't look up these every pass */
+  forward_DCT_method_ptr do_dct = fdct->dct;
+  convsamp_method_ptr do_convsamp = fdct->convsamp;
+  quantize_method_ptr do_quantize = fdct->quantize;
+  workspace = fdct->workspace;
+
+  sample_data += start_row;     /* fold in the vertical offset once */
+
+  for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
+    /* Load data into workspace, applying unsigned->signed conversion */
+    (*do_convsamp) (sample_data, start_col, workspace);
+
+    /* Perform the DCT */
+    (*do_dct) (workspace);
+
+    /* Quantize/descale the coefficients, and store into coef_blocks */
+    (*do_quantize) (coef_blocksbi, divisors, workspace);
+  }
+}
+
+
+#ifdef DCT_FLOAT_SUPPORTED
+
+METHODDEF(void)
+convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+               FAST_FLOAT *workspace)
+{
+  register FAST_FLOAT *workspaceptr;
+  register JSAMPROW elemptr;
+  register int elemr;
+
+  workspaceptr = workspace;
+  for (elemr = 0; elemr < DCTSIZE; elemr++) {
+    elemptr = sample_dataelemr + start_col;
+#if DCTSIZE == 8                /* unroll the inner loop */
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+#else
+    {
+      register int elemc;
+      for (elemc = DCTSIZE; elemc > 0; elemc--)
+        *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    }
+#endif
+  }
+}
+
+
+METHODDEF(void)
+quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+               FAST_FLOAT *workspace)
+{
+  register FAST_FLOAT temp;
+  register int i;
+  register JCOEFPTR output_ptr = coef_block;
+
+  for (i = 0; i < DCTSIZE2; i++) {
+    /* Apply the quantization and scaling factor */
+    temp = workspacei * divisorsi;
+
+    /* Round to nearest integer.
+     * Since C does not specify the direction of rounding for negative
+     * quotients, we have to force the dividend positive for portability.
+     * The maximum coefficient size is +-16K (for 12-bit data), so this
+     * code should work for either 16-bit or 32-bit ints.
+     */
+    output_ptri = (JCOEF)((int)(temp + (FAST_FLOAT)16384.5) - 16384);
+  }
+}
+
+
+METHODDEF(void)
+forward_DCT_float(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                  JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+                  JDIMENSION start_row, JDIMENSION start_col,
+                  JDIMENSION num_blocks)
+/* This version is used for floating-point DCT implementations. */
+{
+  /* This routine is heavily used, so it's worth coding it tightly. */
+  my_fdct_ptr fdct = (my_fdct_ptr)cinfo->fdct;
+  FAST_FLOAT *divisors = fdct->float_divisorscompptr->quant_tbl_no;
+  FAST_FLOAT *workspace;
+  JDIMENSION bi;
+
+
+  /* Make sure the compiler doesn't look up these every pass */
+  float_DCT_method_ptr do_dct = fdct->float_dct;
+  float_convsamp_method_ptr do_convsamp = fdct->float_convsamp;
+  float_quantize_method_ptr do_quantize = fdct->float_quantize;
+  workspace = fdct->float_workspace;
+
+  sample_data += start_row;     /* fold in the vertical offset once */
+
+  for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
+    /* Load data into workspace, applying unsigned->signed conversion */
+    (*do_convsamp) (sample_data, start_col, workspace);
+
+    /* Perform the DCT */
+    (*do_dct) (workspace);
+
+    /* Quantize/descale the coefficients, and store into coef_blocks */
+    (*do_quantize) (coef_blocksbi, divisors, workspace);
+  }
+}
+
+#endif /* DCT_FLOAT_SUPPORTED */
+
+
+/*
+ * Initialize FDCT manager.
+ */
+
+GLOBAL(void)
+jinit_forward_dct(j_compress_ptr cinfo)
+{
+  my_fdct_ptr fdct;
+  int i;
+
+  fdct = (my_fdct_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_fdct_controller));
+  cinfo->fdct = (struct jpeg_forward_dct *)fdct;
+  fdct->pub.start_pass = start_pass_fdctmgr;
+
+  /* First determine the DCT... */
+  switch (cinfo->dct_method) {
+#ifdef DCT_ISLOW_SUPPORTED
+  case JDCT_ISLOW:
+    fdct->pub.forward_DCT = forward_DCT;
+    if (jsimd_can_fdct_islow())
+      fdct->dct = jsimd_fdct_islow;
+    else
+      fdct->dct = jpeg_fdct_islow;
+    break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  case JDCT_IFAST:
+    fdct->pub.forward_DCT = forward_DCT;
+    if (jsimd_can_fdct_ifast())
+      fdct->dct = jsimd_fdct_ifast;
+    else
+      fdct->dct = jpeg_fdct_ifast;
+    break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  case JDCT_FLOAT:
+    fdct->pub.forward_DCT = forward_DCT_float;
+    if (jsimd_can_fdct_float())
+      fdct->float_dct = jsimd_fdct_float;
+    else
+      fdct->float_dct = jpeg_fdct_float;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+    break;
+  }
+
+  /* ...then the supporting stages. */
+  switch (cinfo->dct_method) {
+#ifdef DCT_ISLOW_SUPPORTED
+  case JDCT_ISLOW:
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  case JDCT_IFAST:
+#endif
+#if defined(DCT_ISLOW_SUPPORTED) || defined(DCT_IFAST_SUPPORTED)
+    if (jsimd_can_convsamp())
+      fdct->convsamp = jsimd_convsamp;
+    else
+      fdct->convsamp = convsamp;
+    if (jsimd_can_quantize())
+      fdct->quantize = jsimd_quantize;
+    else
+      fdct->quantize = quantize;
+    break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  case JDCT_FLOAT:
+    if (jsimd_can_convsamp_float())
+      fdct->float_convsamp = jsimd_convsamp_float;
+    else
+      fdct->float_convsamp = convsamp_float;
+    if (jsimd_can_quantize_float())
+      fdct->float_quantize = jsimd_quantize_float;
+    else
+      fdct->float_quantize = quantize_float;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+    break;
+  }
+
+  /* Allocate workspace memory */
+#ifdef DCT_FLOAT_SUPPORTED
+  if (cinfo->dct_method == JDCT_FLOAT)
+    fdct->float_workspace = (FAST_FLOAT *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  sizeof(FAST_FLOAT) * DCTSIZE2);
+  else
+#endif
+    fdct->workspace = (DCTELEM *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  sizeof(DCTELEM) * DCTSIZE2);
+
+  /* Mark divisor tables unallocated */
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    fdct->divisorsi = NULL;
+#ifdef DCT_FLOAT_SUPPORTED
+    fdct->float_divisorsi = NULL;
+#endif
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jchuff.c Added

@@ -0,0 +1,1137 @@
+/*
+ * jchuff.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2014-2016, 2018-2021, D. R. Commander.
+ * Copyright (C) 2015, Matthieu Darbois.
+ * Copyright (C) 2018, Matthias Räncker.
+ * Copyright (C) 2020, Arm Limited.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains Huffman entropy encoding routines.
+ *
+ * Much of the complexity here has to do with supporting output suspension.
+ * If the data destination module demands suspension, we want to be able to
+ * back up to the start of the current MCU.  To do this, we copy state
+ * variables into local working storage, and update them back to the
+ * permanent JPEG objects only upon successful completion of an MCU.
+ *
+ * NOTE: All referenced figures are from
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jsimd.h"
+#include "jconfigint.h"
+#include <limits.h>
+
+/*
+ * NOTE: If USE_CLZ_INTRINSIC is defined, then clz/bsr instructions will be
+ * used for bit counting rather than the lookup table.  This will reduce the
+ * memory footprint by 64k, which is important for some mobile applications
+ * that create many isolated instances of libjpeg-turbo (web browsers, for
+ * instance.)  This may improve performance on some mobile platforms as well.
+ * This feature is enabled by default only on Arm processors, because some x86
+ * chips have a slow implementation of bsr, and the use of clz/bsr cannot be
+ * shown to have a significant performance impact even on the x86 chips that
+ * have a fast implementation of it.  When building for Armv6, you can
+ * explicitly disable the use of clz/bsr by adding -mthumb to the compiler
+ * flags (this defines __thumb__).
+ */
+
+/* NOTE: Both GCC and Clang define __GNUC__ */
+#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
+    defined(_M_ARM) || defined(_M_ARM64)
+#if !defined(__thumb__) || defined(__thumb2__)
+#define USE_CLZ_INTRINSIC
+#endif
+#endif
+
+#ifdef USE_CLZ_INTRINSIC
+#if defined(_MSC_VER) && !defined(__clang__)
+#define JPEG_NBITS_NONZERO(x)  (32 - _CountLeadingZeros(x))
+#else
+#define JPEG_NBITS_NONZERO(x)  (32 - __builtin_clz(x))
+#endif
+#define JPEG_NBITS(x)          (x ? JPEG_NBITS_NONZERO(x) : 0)
+#else
+#include "jpeg_nbits_table.h"
+#define JPEG_NBITS(x)          (jpeg_nbits_tablex)
+#define JPEG_NBITS_NONZERO(x)  JPEG_NBITS(x)
+#endif
+
+
+/* Expanded entropy encoder object for Huffman encoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+#if defined(__x86_64__) && defined(__ILP32__)
+typedef unsigned long long bit_buf_type;
+#else
+typedef size_t bit_buf_type;
+#endif
+
+/* NOTE: The more optimal Huffman encoding algorithm is only used by the
+ * intrinsics implementation of the Arm Neon SIMD extensions, which is why we
+ * retain the old Huffman encoder behavior when using the GAS implementation.
+ */
+#if defined(WITH_SIMD) && !(defined(__arm__) || defined(__aarch64__) || \
+                            defined(_M_ARM) || defined(_M_ARM64))
+typedef unsigned long long simd_bit_buf_type;
+#else
+typedef bit_buf_type simd_bit_buf_type;
+#endif
+
+#if (defined(SIZEOF_SIZE_T) && SIZEOF_SIZE_T == 8) || defined(_WIN64) || \
+    (defined(__x86_64__) && defined(__ILP32__))
+#define BIT_BUF_SIZE  64
+#elif (defined(SIZEOF_SIZE_T) && SIZEOF_SIZE_T == 4) || defined(_WIN32)
+#define BIT_BUF_SIZE  32
+#else
+#error Cannot determine word size
+#endif
+#define SIMD_BIT_BUF_SIZE  (sizeof(simd_bit_buf_type) * 8)
+
+typedef struct {
+  union {
+    bit_buf_type c;
+    simd_bit_buf_type simd;
+  } put_buffer;                         /* current bit accumulation buffer */
+  int free_bits;                        /* # of bits available in it */
+                                        /* (Neon GAS: # of bits now in it) */
+  int last_dc_valMAX_COMPS_IN_SCAN;   /* last DC coef for each component */
+} savable_state;
+
+typedef struct {
+  struct jpeg_entropy_encoder pub; /* public fields */
+
+  savable_state saved;          /* Bit buffer & DC state at start of MCU */
+
+  /* These fields are NOT loaded into local working state. */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+  int next_restart_num;         /* next restart number to write (0-7) */
+
+  /* Pointers to derived tables (these workspaces have image lifespan) */
+  c_derived_tbl *dc_derived_tblsNUM_HUFF_TBLS;
+  c_derived_tbl *ac_derived_tblsNUM_HUFF_TBLS;
+
+#ifdef ENTROPY_OPT_SUPPORTED    /* Statistics tables for optimization */
+  long *dc_count_ptrsNUM_HUFF_TBLS;
+  long *ac_count_ptrsNUM_HUFF_TBLS;
+#endif
+
+  int simd;
+} huff_entropy_encoder;
+
+typedef huff_entropy_encoder *huff_entropy_ptr;
+
+/* Working state while writing an MCU.
+ * This struct contains all the fields that are needed by subroutines.
+ */
+
+typedef struct {
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
+  savable_state cur;            /* Current bit buffer & DC state */
+  j_compress_ptr cinfo;         /* dump_buffer needs access to this */
+  int simd;
+} working_state;
+
+
+/* Forward declarations */
+METHODDEF(boolean) encode_mcu_huff(j_compress_ptr cinfo, JBLOCKROW *MCU_data);
+METHODDEF(void) finish_pass_huff(j_compress_ptr cinfo);
+#ifdef ENTROPY_OPT_SUPPORTED
+METHODDEF(boolean) encode_mcu_gather(j_compress_ptr cinfo,
+                                     JBLOCKROW *MCU_data);
+METHODDEF(void) finish_pass_gather(j_compress_ptr cinfo);
+#endif
+
+
+/*
+ * Initialize for a Huffman-compressed scan.
+ * If gather_statistics is TRUE, we do not output anything during the scan,
+ * just count the Huffman symbols used and generate Huffman code tables.
+ */
+
+METHODDEF(void)
+start_pass_huff(j_compress_ptr cinfo, boolean gather_statistics)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  int ci, dctbl, actbl;
+  jpeg_component_info *compptr;
+
+  if (gather_statistics) {
+#ifdef ENTROPY_OPT_SUPPORTED
+    entropy->pub.encode_mcu = encode_mcu_gather;
+    entropy->pub.finish_pass = finish_pass_gather;
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    entropy->pub.encode_mcu = encode_mcu_huff;
+    entropy->pub.finish_pass = finish_pass_huff;
+  }
+
+  entropy->simd = jsimd_can_huff_encode_one_block();
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    dctbl = compptr->dc_tbl_no;
+    actbl = compptr->ac_tbl_no;
+    if (gather_statistics) {
+#ifdef ENTROPY_OPT_SUPPORTED
+      /* Check for invalid table indexes */
+      /* (make_c_derived_tbl does this in the other path) */
+      if (dctbl < 0 || dctbl >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, dctbl);
+      if (actbl < 0 || actbl >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, actbl);
+      /* Allocate and zero the statistics tables */
+      /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
+      if (entropy->dc_count_ptrsdctbl == NULL)
+        entropy->dc_count_ptrsdctbl = (long *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                      257 * sizeof(long));
+      MEMZERO(entropy->dc_count_ptrsdctbl, 257 * sizeof(long));
+      if (entropy->ac_count_ptrsactbl == NULL)
+        entropy->ac_count_ptrsactbl = (long *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                      257 * sizeof(long));
+      MEMZERO(entropy->ac_count_ptrsactbl, 257 * sizeof(long));
+#endif
+    } else {
+      /* Compute derived values for Huffman tables */
+      /* We may do this more than once for a table, but it's not expensive */
+      jpeg_make_c_derived_tbl(cinfo, TRUE, dctbl,
+                              &entropy->dc_derived_tblsdctbl);
+      jpeg_make_c_derived_tbl(cinfo, FALSE, actbl,
+                              &entropy->ac_derived_tblsactbl);
+    }
+    /* Initialize DC predictions to 0 */
+    entropy->saved.last_dc_valci = 0;
+  }
+
+  /* Initialize bit buffer to empty */
+  if (entropy->simd) {
+    entropy->saved.put_buffer.simd = 0;
+#if defined(__aarch64__) && !defined(NEON_INTRINSICS)
+    entropy->saved.free_bits = 0;
+#else
+    entropy->saved.free_bits = SIMD_BIT_BUF_SIZE;
+#endif
+  } else {
+    entropy->saved.put_buffer.c = 0;
+    entropy->saved.free_bits = BIT_BUF_SIZE;
+  }
+
+  /* Initialize restart stuff */
+  entropy->restarts_to_go = cinfo->restart_interval;
+  entropy->next_restart_num = 0;
+}
+
+
+/*
+ * Compute the derived values for a Huffman table.
+ * This routine also performs some validation checks on the table.
+ *
+ * Note this is also used by jcphuff.c.
+ */
+
+GLOBAL(void)
+jpeg_make_c_derived_tbl(j_compress_ptr cinfo, boolean isDC, int tblno,
+                        c_derived_tbl **pdtbl)
+{
+  JHUFF_TBL *htbl;
+  c_derived_tbl *dtbl;
+  int p, i, l, lastp, si, maxsymbol;
+  char huffsize257;
+  unsigned int huffcode257;
+  unsigned int code;
+
+  /* Note that huffsize and huffcode are filled in code-length order,
+   * paralleling the order of the symbols themselves in htbl->huffval.
+   */
+
+  /* Find the input Huffman table */
+  if (tblno < 0 || tblno >= NUM_HUFF_TBLS)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+  htbl =
+    isDC ? cinfo->dc_huff_tbl_ptrstblno : cinfo->ac_huff_tbl_ptrstblno;
+  if (htbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+
+  /* Allocate a workspace if we haven't already done so. */
+  if (*pdtbl == NULL)
+    *pdtbl = (c_derived_tbl *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  sizeof(c_derived_tbl));
+  dtbl = *pdtbl;
+
+  /* Figure C.1: make table of Huffman code length for each symbol */
+
+  p = 0;
+  for (l = 1; l <= 16; l++) {
+    i = (int)htbl->bitsl;
+    if (i < 0 || p + i > 256)   /* protect against table overrun */
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    while (i--)
+      huffsizep++ = (char)l;
+  }
+  huffsizep = 0;
+  lastp = p;
+
+  /* Figure C.2: generate the codes themselves */
+  /* We also validate that the counts represent a legal Huffman code tree. */
+
+  code = 0;
+  si = huffsize0;
+  p = 0;
+  while (huffsizep) {
+    while (((int)huffsizep) == si) {
+      huffcodep++ = code;
+      code++;
+    }
+    /* code is now 1 more than the last code used for codelength si; but
+     * it must still fit in si bits, since no code is allowed to be all ones.
+     */
+    if (((JLONG)code) >= (((JLONG)1) << si))
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    code <<= 1;
+    si++;
+  }
+
+  /* Figure C.3: generate encoding tables */
+  /* These are code and size indexed by symbol value */
+
+  /* Set all codeless symbols to have code length 0;
+   * this lets us detect duplicate VAL entries here, and later
+   * allows emit_bits to detect any attempt to emit such symbols.
+   */
+  MEMZERO(dtbl->ehufco, sizeof(dtbl->ehufco));
+  MEMZERO(dtbl->ehufsi, sizeof(dtbl->ehufsi));
+
+  /* This is also a convenient place to check for out-of-range
+   * and duplicated VAL entries.  We allow 0..255 for AC symbols
+   * but only 0..15 for DC.  (We could constrain them further
+   * based on data depth and mode, but this seems enough.)
+   */
+  maxsymbol = isDC ? 15 : 255;
+
+  for (p = 0; p < lastp; p++) {
+    i = htbl->huffvalp;
+    if (i < 0 || i > maxsymbol || dtbl->ehufsii)
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    dtbl->ehufcoi = huffcodep;
+    dtbl->ehufsii = huffsizep;
+  }
+}
+
+
+/* Outputting bytes to the file */
+
+/* Emit a byte, taking 'action' if must suspend. */
+#define emit_byte(state, val, action) { \
+  *(state)->next_output_byte++ = (JOCTET)(val); \
+  if (--(state)->free_in_buffer == 0) \
+    if (!dump_buffer(state)) \
+      { action; } \
+}
+
+
+LOCAL(boolean)
+dump_buffer(working_state *state)
+/* Empty the output buffer; return TRUE if successful, FALSE if must suspend */
+{
+  struct jpeg_destination_mgr *dest = state->cinfo->dest;
+
+  if (!(*dest->empty_output_buffer) (state->cinfo))
+    return FALSE;
+  /* After a successful buffer dump, must reset buffer pointers */
+  state->next_output_byte = dest->next_output_byte;
+  state->free_in_buffer = dest->free_in_buffer;
+  return TRUE;
+}
+
+
+/* Outputting bits to the file */
+
+/* Output byte b and, speculatively, an additional 0 byte.  0xFF must be
+ * encoded as 0xFF 0x00, so the output buffer pointer is advanced by 2 if the
+ * byte is 0xFF.  Otherwise, the output buffer pointer is advanced by 1, and
+ * the speculative 0 byte will be overwritten by the next byte.
+ */
+#define EMIT_BYTE(b) { \
+  buffer0 = (JOCTET)(b); \
+  buffer1 = 0; \
+  buffer -= -2 + ((JOCTET)(b) < 0xFF); \
+}
+
+/* Output the entire bit buffer.  If there are no 0xFF bytes in it, then write
+ * directly to the output buffer.  Otherwise, use the EMIT_BYTE() macro to
+ * encode 0xFF as 0xFF 0x00.
+ */
+#if BIT_BUF_SIZE == 64
+
+#define FLUSH() { \
+  if (put_buffer & 0x8080808080808080 & ~(put_buffer + 0x0101010101010101)) { \
+    EMIT_BYTE(put_buffer >> 56) \
+    EMIT_BYTE(put_buffer >> 48) \
+    EMIT_BYTE(put_buffer >> 40) \
+    EMIT_BYTE(put_buffer >> 32) \
+    EMIT_BYTE(put_buffer >> 24) \
+    EMIT_BYTE(put_buffer >> 16) \
+    EMIT_BYTE(put_buffer >>  8) \
+    EMIT_BYTE(put_buffer      ) \
+  } else { \
+    buffer0 = (JOCTET)(put_buffer >> 56); \
+    buffer1 = (JOCTET)(put_buffer >> 48); \
+    buffer2 = (JOCTET)(put_buffer >> 40); \
+    buffer3 = (JOCTET)(put_buffer >> 32); \
+    buffer4 = (JOCTET)(put_buffer >> 24); \
+    buffer5 = (JOCTET)(put_buffer >> 16); \
+    buffer6 = (JOCTET)(put_buffer >> 8); \
+    buffer7 = (JOCTET)(put_buffer); \
+    buffer += 8; \
+  } \
+}
+
+#else
+
+#define FLUSH() { \
+  if (put_buffer & 0x80808080 & ~(put_buffer + 0x01010101)) { \
+    EMIT_BYTE(put_buffer >> 24) \
+    EMIT_BYTE(put_buffer >> 16) \
+    EMIT_BYTE(put_buffer >>  8) \
+    EMIT_BYTE(put_buffer      ) \
+  } else { \
+    buffer0 = (JOCTET)(put_buffer >> 24); \
+    buffer1 = (JOCTET)(put_buffer >> 16); \
+    buffer2 = (JOCTET)(put_buffer >> 8); \
+    buffer3 = (JOCTET)(put_buffer); \
+    buffer += 4; \
+  } \
+}
+
+#endif
+
+/* Fill the bit buffer to capacity with the leading bits from code, then output
+ * the bit buffer and put the remaining bits from code into the bit buffer.
+ */
+#define PUT_AND_FLUSH(code, size) { \
+  put_buffer = (put_buffer << (size + free_bits)) | (code >> -free_bits); \
+  FLUSH() \
+  free_bits += BIT_BUF_SIZE; \
+  put_buffer = code; \
+}
+
+/* Insert code into the bit buffer and output the bit buffer if needed.
+ * NOTE: We can't flush with free_bits == 0, since the left shift in
+ * PUT_AND_FLUSH() would have undefined behavior.
+ */
+#define PUT_BITS(code, size) { \
+  free_bits -= size; \
+  if (free_bits < 0) \
+    PUT_AND_FLUSH(code, size) \
+  else \
+    put_buffer = (put_buffer << size) | code; \
+}
+
+#define PUT_CODE(code, size) { \
+  temp &= (((JLONG)1) << nbits) - 1; \
+  temp |= code << nbits; \
+  nbits += size; \
+  PUT_BITS(temp, nbits) \
+}
+
+
+/* Although it is exceedingly rare, it is possible for a Huffman-encoded
+ * coefficient block to be larger than the 128-byte unencoded block.  For each
+ * of the 64 coefficients, PUT_BITS is invoked twice, and each invocation can
+ * theoretically store 16 bits (for a maximum of 2048 bits or 256 bytes per
+ * encoded block.)  If, for instance, one artificially sets the AC
+ * coefficients to alternating values of 32767 and -32768 (using the JPEG
+ * scanning order-- 1, 8, 16, etc.), then this will produce an encoded block
+ * larger than 200 bytes.
+ */
+#define BUFSIZE  (DCTSIZE2 * 8)
+
+#define LOAD_BUFFER() { \
+  if (state->free_in_buffer < BUFSIZE) { \
+    localbuf = 1; \
+    buffer = _buffer; \
+  } else \
+    buffer = state->next_output_byte; \
+}
+
+#define STORE_BUFFER() { \
+  if (localbuf) { \
+    size_t bytes, bytestocopy; \
+    bytes = buffer - _buffer; \
+    buffer = _buffer; \
+    while (bytes > 0) { \
+      bytestocopy = MIN(bytes, state->free_in_buffer); \
+      MEMCOPY(state->next_output_byte, buffer, bytestocopy); \
+      state->next_output_byte += bytestocopy; \
+      buffer += bytestocopy; \
+      state->free_in_buffer -= bytestocopy; \
+      if (state->free_in_buffer == 0) \
+        if (!dump_buffer(state)) return FALSE; \
+      bytes -= bytestocopy; \
+    } \
+  } else { \
+    state->free_in_buffer -= (buffer - state->next_output_byte); \
+    state->next_output_byte = buffer; \
+  } \
+}
+
+
+LOCAL(boolean)
+flush_bits(working_state *state)
+{
+  JOCTET _bufferBUFSIZE, *buffer, temp;
+  simd_bit_buf_type put_buffer;  int put_bits;
+  int localbuf = 0;
+
+  if (state->simd) {
+#if defined(__aarch64__) && !defined(NEON_INTRINSICS)
+    put_bits = state->cur.free_bits;
+#else
+    put_bits = SIMD_BIT_BUF_SIZE - state->cur.free_bits;
+#endif
+    put_buffer = state->cur.put_buffer.simd;
+  } else {
+    put_bits = BIT_BUF_SIZE - state->cur.free_bits;
+    put_buffer = state->cur.put_buffer.c;
+  }
+
+  LOAD_BUFFER()
+
+  while (put_bits >= 8) {
+    put_bits -= 8;
+    temp = (JOCTET)(put_buffer >> put_bits);
+    EMIT_BYTE(temp)
+  }
+  if (put_bits) {
+    /* fill partial byte with ones */
+    temp = (JOCTET)((put_buffer << (8 - put_bits)) | (0xFF >> put_bits));
+    EMIT_BYTE(temp)
+  }
+
+  if (state->simd) {                    /* and reset bit buffer to empty */
+    state->cur.put_buffer.simd = 0;
+#if defined(__aarch64__) && !defined(NEON_INTRINSICS)
+    state->cur.free_bits = 0;
+#else
+    state->cur.free_bits = SIMD_BIT_BUF_SIZE;
+#endif
+  } else {
+    state->cur.put_buffer.c = 0;
+    state->cur.free_bits = BIT_BUF_SIZE;
+  }
+  STORE_BUFFER()
+
+  return TRUE;
+}
+
+
+/* Encode a single block's worth of coefficients */
+
+LOCAL(boolean)
+encode_one_block_simd(working_state *state, JCOEFPTR block, int last_dc_val,
+                      c_derived_tbl *dctbl, c_derived_tbl *actbl)
+{
+  JOCTET _bufferBUFSIZE, *buffer;
+  int localbuf = 0;
+
+  LOAD_BUFFER()
+
+  buffer = jsimd_huff_encode_one_block(state, buffer, block, last_dc_val,
+                                       dctbl, actbl);
+
+  STORE_BUFFER()
+
+  return TRUE;
+}
+
+LOCAL(boolean)
+encode_one_block(working_state *state, JCOEFPTR block, int last_dc_val,
+                 c_derived_tbl *dctbl, c_derived_tbl *actbl)
+{
+  int temp, nbits, free_bits;
+  bit_buf_type put_buffer;
+  JOCTET _bufferBUFSIZE, *buffer;
+  int localbuf = 0;
+
+  free_bits = state->cur.free_bits;
+  put_buffer = state->cur.put_buffer.c;
+  LOAD_BUFFER()
+
+  /* Encode the DC coefficient difference per section F.1.2.1 */
+
+  temp = block0 - last_dc_val;
+
+  /* This is a well-known technique for obtaining the absolute value without a
+   * branch.  It is derived from an assembly language technique presented in
+   * "How to Optimize for the Pentium Processors", Copyright (c) 1996, 1997 by
+   * Agner Fog.  This code assumes we are on a two's complement machine.
+   */
+  nbits = temp >> (CHAR_BIT * sizeof(int) - 1);
+  temp += nbits;
+  nbits ^= temp;
+
+  /* Find the number of bits needed for the magnitude of the coefficient */
+  nbits = JPEG_NBITS(nbits);
+
+  /* Emit the Huffman-coded symbol for the number of bits.
+   * Emit that number of bits of the value, if positive,
+   * or the complement of its magnitude, if negative.
+   */
+  PUT_CODE(dctbl->ehufconbits, dctbl->ehufsinbits)
+
+  /* Encode the AC coefficients per section F.1.2.2 */
+
+  {
+    int r = 0;                  /* r = run length of zeros */
+
+/* Manually unroll the k loop to eliminate the counter variable.  This
+ * improves performance greatly on systems with a limited number of
+ * registers (such as x86.)
+ */
+#define kloop(jpeg_natural_order_of_k) { \
+  if ((temp = blockjpeg_natural_order_of_k) == 0) { \
+    r += 16; \
+  } else { \
+    /* Branch-less absolute value, bitwise complement, etc., same as above */ \
+    nbits = temp >> (CHAR_BIT * sizeof(int) - 1); \
+    temp += nbits; \
+    nbits ^= temp; \
+    nbits = JPEG_NBITS_NONZERO(nbits); \
+    /* if run length > 15, must emit special run-length-16 codes (0xF0) */ \
+    while (r >= 16 * 16) { \
+      r -= 16 * 16; \
+      PUT_BITS(actbl->ehufco0xf0, actbl->ehufsi0xf0) \
+    } \
+    /* Emit Huffman symbol for run length / number of bits */ \
+    r += nbits; \
+    PUT_CODE(actbl->ehufcor, actbl->ehufsir) \
+    r = 0; \
+  } \
+}
+
+    /* One iteration for each value in jpeg_natural_order */
+    kloop(1);   kloop(8);   kloop(16);  kloop(9);   kloop(2);   kloop(3);
+    kloop(10);  kloop(17);  kloop(24);  kloop(32);  kloop(25);  kloop(18);
+    kloop(11);  kloop(4);   kloop(5);   kloop(12);  kloop(19);  kloop(26);
+    kloop(33);  kloop(40);  kloop(48);  kloop(41);  kloop(34);  kloop(27);
+    kloop(20);  kloop(13);  kloop(6);   kloop(7);   kloop(14);  kloop(21);
+    kloop(28);  kloop(35);  kloop(42);  kloop(49);  kloop(56);  kloop(57);
+    kloop(50);  kloop(43);  kloop(36);  kloop(29);  kloop(22);  kloop(15);
+    kloop(23);  kloop(30);  kloop(37);  kloop(44);  kloop(51);  kloop(58);
+    kloop(59);  kloop(52);  kloop(45);  kloop(38);  kloop(31);  kloop(39);
+    kloop(46);  kloop(53);  kloop(60);  kloop(61);  kloop(54);  kloop(47);
+    kloop(55);  kloop(62);  kloop(63);
+
+    /* If the last coef(s) were zero, emit an end-of-block code */
+    if (r > 0) {
+      PUT_BITS(actbl->ehufco0, actbl->ehufsi0)
+    }
+  }
+
+  state->cur.put_buffer.c = put_buffer;
+  state->cur.free_bits = free_bits;
+  STORE_BUFFER()
+
+  return TRUE;
+}
+
+
+/*
+ * Emit a restart marker & resynchronize predictions.
+ */
+
+LOCAL(boolean)
+emit_restart(working_state *state, int restart_num)
+{
+  int ci;
+
+  if (!flush_bits(state))
+    return FALSE;
+
+  emit_byte(state, 0xFF, return FALSE);
+  emit_byte(state, JPEG_RST0 + restart_num, return FALSE);
+
+  /* Re-initialize DC predictions to 0 */
+  for (ci = 0; ci < state->cinfo->comps_in_scan; ci++)
+    state->cur.last_dc_valci = 0;
+
+  /* The restart counter is not updated until we successfully write the MCU. */
+
+  return TRUE;
+}
+
+
+/*
+ * Encode and output one MCU's worth of Huffman-compressed coefficients.
+ */
+
+METHODDEF(boolean)
+encode_mcu_huff(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  working_state state;
+  int blkn, ci;
+  jpeg_component_info *compptr;
+
+  /* Load up working state */
+  state.next_output_byte = cinfo->dest->next_output_byte;
+  state.free_in_buffer = cinfo->dest->free_in_buffer;
+  state.cur = entropy->saved;
+  state.cinfo = cinfo;
+  state.simd = entropy->simd;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (!emit_restart(&state, entropy->next_restart_num))
+        return FALSE;
+  }
+
+  /* Encode the MCU data blocks */
+  if (entropy->simd) {
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      ci = cinfo->MCU_membershipblkn;
+      compptr = cinfo->cur_comp_infoci;
+      if (!encode_one_block_simd(&state,
+                                 MCU_datablkn0, state.cur.last_dc_valci,
+                                 entropy->dc_derived_tblscompptr->dc_tbl_no,
+                                 entropy->ac_derived_tblscompptr->ac_tbl_no))
+        return FALSE;
+      /* Update last_dc_val */
+      state.cur.last_dc_valci = MCU_datablkn00;
+    }
+  } else {
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      ci = cinfo->MCU_membershipblkn;
+      compptr = cinfo->cur_comp_infoci;
+      if (!encode_one_block(&state,
+                            MCU_datablkn0, state.cur.last_dc_valci,
+                            entropy->dc_derived_tblscompptr->dc_tbl_no,
+                            entropy->ac_derived_tblscompptr->ac_tbl_no))
+        return FALSE;
+      /* Update last_dc_val */
+      state.cur.last_dc_valci = MCU_datablkn00;
+    }
+  }
+
+  /* Completed MCU, so update state */
+  cinfo->dest->next_output_byte = state.next_output_byte;
+  cinfo->dest->free_in_buffer = state.free_in_buffer;
+  entropy->saved = state.cur;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Finish up at the end of a Huffman-compressed scan.
+ */
+
+METHODDEF(void)
+finish_pass_huff(j_compress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  working_state state;
+
+  /* Load up working state ... flush_bits needs it */
+  state.next_output_byte = cinfo->dest->next_output_byte;
+  state.free_in_buffer = cinfo->dest->free_in_buffer;
+  state.cur = entropy->saved;
+  state.cinfo = cinfo;
+  state.simd = entropy->simd;
+
+  /* Flush out the last data */
+  if (!flush_bits(&state))
+    ERREXIT(cinfo, JERR_CANT_SUSPEND);
+
+  /* Update state */
+  cinfo->dest->next_output_byte = state.next_output_byte;
+  cinfo->dest->free_in_buffer = state.free_in_buffer;
+  entropy->saved = state.cur;
+}
+
+
+/*
+ * Huffman coding optimization.
+ *
+ * We first scan the supplied data and count the number of uses of each symbol
+ * that is to be Huffman-coded. (This process MUST agree with the code above.)
+ * Then we build a Huffman coding tree for the observed counts.
+ * Symbols which are not needed at all for the particular image are not
+ * assigned any code, which saves space in the DHT marker as well as in
+ * the compressed data.
+ */
+
+#ifdef ENTROPY_OPT_SUPPORTED
+
+
+/* Process a single block's worth of coefficients */
+
+LOCAL(void)
+htest_one_block(j_compress_ptr cinfo, JCOEFPTR block, int last_dc_val,
+                long dc_counts, long ac_counts)
+{
+  register int temp;
+  register int nbits;
+  register int k, r;
+
+  /* Encode the DC coefficient difference per section F.1.2.1 */
+
+  temp = block0 - last_dc_val;
+  if (temp < 0)
+    temp = -temp;
+
+  /* Find the number of bits needed for the magnitude of the coefficient */
+  nbits = 0;
+  while (temp) {
+    nbits++;
+    temp >>= 1;
+  }
+  /* Check for out-of-range coefficient values.
+   * Since we're encoding a difference, the range limit is twice as much.
+   */
+  if (nbits > MAX_COEF_BITS + 1)
+    ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+
+  /* Count the Huffman symbol for the number of bits */
+  dc_countsnbits++;
+
+  /* Encode the AC coefficients per section F.1.2.2 */
+
+  r = 0;                        /* r = run length of zeros */
+
+  for (k = 1; k < DCTSIZE2; k++) {
+    if ((temp = blockjpeg_natural_orderk) == 0) {
+      r++;
+    } else {
+      /* if run length > 15, must emit special run-length-16 codes (0xF0) */
+      while (r > 15) {
+        ac_counts0xF0++;
+        r -= 16;
+      }
+
+      /* Find the number of bits needed for the magnitude of the coefficient */
+      if (temp < 0)
+        temp = -temp;
+
+      /* Find the number of bits needed for the magnitude of the coefficient */
+      nbits = 1;                /* there must be at least one 1 bit */
+      while ((temp >>= 1))
+        nbits++;
+      /* Check for out-of-range coefficient values */
+      if (nbits > MAX_COEF_BITS)
+        ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+
+      /* Count Huffman symbol for run length / number of bits */
+      ac_counts(r << 4) + nbits++;
+
+      r = 0;
+    }
+  }
+
+  /* If the last coef(s) were zero, emit an end-of-block code */
+  if (r > 0)
+    ac_counts0++;
+}
+
+
+/*
+ * Trial-encode one MCU's worth of Huffman-compressed coefficients.
+ * No data is actually output, so no suspension return is possible.
+ */
+
+METHODDEF(boolean)
+encode_mcu_gather(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  int blkn, ci;
+  jpeg_component_info *compptr;
+
+  /* Take care of restart intervals if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      /* Re-initialize DC predictions to 0 */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+        entropy->saved.last_dc_valci = 0;
+      /* Update restart state */
+      entropy->restarts_to_go = cinfo->restart_interval;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    ci = cinfo->MCU_membershipblkn;
+    compptr = cinfo->cur_comp_infoci;
+    htest_one_block(cinfo, MCU_datablkn0, entropy->saved.last_dc_valci,
+                    entropy->dc_count_ptrscompptr->dc_tbl_no,
+                    entropy->ac_count_ptrscompptr->ac_tbl_no);
+    entropy->saved.last_dc_valci = MCU_datablkn00;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Generate the best Huffman code table for the given counts, fill htbl.
+ * Note this is also used by jcphuff.c.
+ *
+ * The JPEG standard requires that no symbol be assigned a codeword of all
+ * one bits (so that padding bits added at the end of a compressed segment
+ * can't look like a valid code).  Because of the canonical ordering of
+ * codewords, this just means that there must be an unused slot in the
+ * longest codeword length category.  Annex K (Clause K.2) of
+ * Rec. ITU-T T.81 (1992) | ISO/IEC 10918-1:1994 suggests reserving such a slot
+ * by pretending that symbol 256 is a valid symbol with count 1.  In theory
+ * that's not optimal; giving it count zero but including it in the symbol set
+ * anyway should give a better Huffman code.  But the theoretically better code
+ * actually seems to come out worse in practice, because it produces more
+ * all-ones bytes (which incur stuffed zero bytes in the final file).  In any
+ * case the difference is tiny.
+ *
+ * The JPEG standard requires Huffman codes to be no more than 16 bits long.
+ * If some symbols have a very small but nonzero probability, the Huffman tree
+ * must be adjusted to meet the code length restriction.  We currently use
+ * the adjustment method suggested in JPEG section K.2.  This method is *not*
+ * optimal; it may not choose the best possible limited-length code.  But
+ * typically only very-low-frequency symbols will be given less-than-optimal
+ * lengths, so the code is almost optimal.  Experimental comparisons against
+ * an optimal limited-length-code algorithm indicate that the difference is
+ * microscopic --- usually less than a hundredth of a percent of total size.
+ * So the extra complexity of an optimal algorithm doesn't seem worthwhile.
+ */
+
+GLOBAL(void)
+jpeg_gen_optimal_table(j_compress_ptr cinfo, JHUFF_TBL *htbl, long freq)
+{
+#define MAX_CLEN  32            /* assumed maximum initial code length */
+  UINT8 bitsMAX_CLEN + 1;     /* bitsk = # of symbols with code length k */
+  int codesize257;            /* codesizek = code length of symbol k */
+  int others257;              /* next symbol in current branch of tree */
+  int c1, c2;
+  int p, i, j;
+  long v;
+
+  /* This algorithm is explained in section K.2 of the JPEG standard */
+
+  MEMZERO(bits, sizeof(bits));
+  MEMZERO(codesize, sizeof(codesize));
+  for (i = 0; i < 257; i++)
+    othersi = -1;             /* init links to empty */
+
+  freq256 = 1;                /* make sure 256 has a nonzero count */
+  /* Including the pseudo-symbol 256 in the Huffman procedure guarantees
+   * that no real symbol is given code-value of all ones, because 256
+   * will be placed last in the largest codeword category.
+   */
+
+  /* Huffman's basic algorithm to assign optimal code lengths to symbols */
+
+  for (;;) {
+    /* Find the smallest nonzero frequency, set c1 = its symbol */
+    /* In case of ties, take the larger symbol number */
+    c1 = -1;
+    v = 1000000000L;
+    for (i = 0; i <= 256; i++) {
+      if (freqi && freqi <= v) {
+        v = freqi;
+        c1 = i;
+      }
+    }
+
+    /* Find the next smallest nonzero frequency, set c2 = its symbol */
+    /* In case of ties, take the larger symbol number */
+    c2 = -1;
+    v = 1000000000L;
+    for (i = 0; i <= 256; i++) {
+      if (freqi && freqi <= v && i != c1) {
+        v = freqi;
+        c2 = i;
+      }
+    }
+
+    /* Done if we've merged everything into one frequency */
+    if (c2 < 0)
+      break;
+
+    /* Else merge the two counts/trees */
+    freqc1 += freqc2;
+    freqc2 = 0;
+
+    /* Increment the codesize of everything in c1's tree branch */
+    codesizec1++;
+    while (othersc1 >= 0) {
+      c1 = othersc1;
+      codesizec1++;
+    }
+
+    othersc1 = c2;            /* chain c2 onto c1's tree branch */
+
+    /* Increment the codesize of everything in c2's tree branch */
+    codesizec2++;
+    while (othersc2 >= 0) {
+      c2 = othersc2;
+      codesizec2++;
+    }
+  }
+
+  /* Now count the number of symbols of each code length */
+  for (i = 0; i <= 256; i++) {
+    if (codesizei) {
+      /* The JPEG standard seems to think that this can't happen, */
+      /* but I'm paranoid... */
+      if (codesizei > MAX_CLEN)
+        ERREXIT(cinfo, JERR_HUFF_CLEN_OVERFLOW);
+
+      bitscodesizei++;
+    }
+  }
+
+  /* JPEG doesn't allow symbols with code lengths over 16 bits, so if the pure
+   * Huffman procedure assigned any such lengths, we must adjust the coding.
+   * Here is what Rec. ITU-T T.81 | ISO/IEC 10918-1 says about how this next
+   * bit works: Since symbols are paired for the longest Huffman code, the
+   * symbols are removed from this length category two at a time.  The prefix
+   * for the pair (which is one bit shorter) is allocated to one of the pair;
+   * then, skipping the BITS entry for that prefix length, a code word from the
+   * next shortest nonzero BITS entry is converted into a prefix for two code
+   * words one bit longer.
+   */
+
+  for (i = MAX_CLEN; i > 16; i--) {
+    while (bitsi > 0) {
+      j = i - 2;                /* find length of new prefix to be used */
+      while (bitsj == 0)
+        j--;
+
+      bitsi -= 2;             /* remove two symbols */
+      bitsi - 1++;            /* one goes in this length */
+      bitsj + 1 += 2;         /* two new symbols in this length */
+      bitsj--;                /* symbol of this length is now a prefix */
+    }
+  }
+
+  /* Remove the count for the pseudo-symbol 256 from the largest codelength */
+  while (bitsi == 0)          /* find largest codelength still in use */
+    i--;
+  bitsi--;
+
+  /* Return final symbol counts (only for lengths 0..16) */
+  MEMCOPY(htbl->bits, bits, sizeof(htbl->bits));
+
+  /* Return a list of the symbols sorted by code length */
+  /* It's not real clear to me why we don't need to consider the codelength
+   * changes made above, but Rec. ITU-T T.81 | ISO/IEC 10918-1 seems to think
+   * this works.
+   */
+  p = 0;
+  for (i = 1; i <= MAX_CLEN; i++) {
+    for (j = 0; j <= 255; j++) {
+      if (codesizej == i) {
+        htbl->huffvalp = (UINT8)j;
+        p++;
+      }
+    }
+  }
+
+  /* Set sent_table FALSE so updated table will be written to JPEG file. */
+  htbl->sent_table = FALSE;
+}
+
+
+/*
+ * Finish up a statistics-gathering pass and create the new Huffman tables.
+ */
+
+METHODDEF(void)
+finish_pass_gather(j_compress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  int ci, dctbl, actbl;
+  jpeg_component_info *compptr;
+  JHUFF_TBL **htblptr;
+  boolean did_dcNUM_HUFF_TBLS;
+  boolean did_acNUM_HUFF_TBLS;
+
+  /* It's important not to apply jpeg_gen_optimal_table more than once
+   * per table, because it clobbers the input frequency counts!
+   */
+  MEMZERO(did_dc, sizeof(did_dc));
+  MEMZERO(did_ac, sizeof(did_ac));
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    dctbl = compptr->dc_tbl_no;
+    actbl = compptr->ac_tbl_no;
+    if (!did_dcdctbl) {
+      htblptr = &cinfo->dc_huff_tbl_ptrsdctbl;
+      if (*htblptr == NULL)
+        *htblptr = jpeg_alloc_huff_table((j_common_ptr)cinfo);
+      jpeg_gen_optimal_table(cinfo, *htblptr, entropy->dc_count_ptrsdctbl);
+      did_dcdctbl = TRUE;
+    }
+    if (!did_acactbl) {
+      htblptr = &cinfo->ac_huff_tbl_ptrsactbl;
+      if (*htblptr == NULL)
+        *htblptr = jpeg_alloc_huff_table((j_common_ptr)cinfo);
+      jpeg_gen_optimal_table(cinfo, *htblptr, entropy->ac_count_ptrsactbl);
+      did_acactbl = TRUE;
+    }
+  }
+}
+
+
+#endif /* ENTROPY_OPT_SUPPORTED */
+
+
+/*
+ * Module initialization routine for Huffman entropy encoding.
+ */
+
+GLOBAL(void)
+jinit_huff_encoder(j_compress_ptr cinfo)
+{
+  huff_entropy_ptr entropy;
+  int i;
+
+  entropy = (huff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(huff_entropy_encoder));
+  cinfo->entropy = (struct jpeg_entropy_encoder *)entropy;
+  entropy->pub.start_pass = start_pass_huff;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->dc_derived_tblsi = entropy->ac_derived_tblsi = NULL;
+#ifdef ENTROPY_OPT_SUPPORTED
+    entropy->dc_count_ptrsi = entropy->ac_count_ptrsi = NULL;
+#endif
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jchuff.h Added

@@ -0,0 +1,42 @@
+/*
+ * jchuff.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains declarations for Huffman entropy encoding routines
+ * that are shared between the sequential encoder (jchuff.c) and the
+ * progressive encoder (jcphuff.c).  No other modules need to see these.
+ */
+
+/* The legal range of a DCT coefficient is
+ *  -1024 .. +1023  for 8-bit data;
+ * -16384 .. +16383 for 12-bit data.
+ * Hence the magnitude should always fit in 10 or 14 bits respectively.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MAX_COEF_BITS  10
+#else
+#define MAX_COEF_BITS  14
+#endif
+
+/* Derived data constructed for each Huffman table */
+
+typedef struct {
+  unsigned int ehufco256;     /* code for each symbol */
+  char ehufsi256;             /* length of code for each symbol */
+  /* If no code has been allocated for a symbol S, ehufsiS contains 0 */
+} c_derived_tbl;
+
+/* Expand a Huffman table definition into the derived format */
+EXTERN(void) jpeg_make_c_derived_tbl(j_compress_ptr cinfo, boolean isDC,
+                                     int tblno, c_derived_tbl **pdtbl);
+
+/* Generate an optimal table definition given the specified counts */
+EXTERN(void) jpeg_gen_optimal_table(j_compress_ptr cinfo, JHUFF_TBL *htbl,
+                                    long freq);

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcicc.c Added

@@ -0,0 +1,105 @@
+/*
+ * jcicc.c
+ *
+ * Copyright (C) 1997-1998, Thomas G. Lane, Todd Newman.
+ * Copyright (C) 2017, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file provides code to write International Color Consortium (ICC) device
+ * profiles embedded in JFIF JPEG image files.  The ICC has defined a standard
+ * for including such data in JPEG "APP2" markers.  The code given here does
+ * not know anything about the internal structure of the ICC profile data; it
+ * just knows how to embed the profile data in a JPEG file while writing it.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+
+/*
+ * Since an ICC profile can be larger than the maximum size of a JPEG marker
+ * (64K), we need provisions to split it into multiple markers.  The format
+ * defined by the ICC specifies one or more APP2 markers containing the
+ * following data:
+ *      Identifying string      ASCII "ICC_PROFILE\0"  (12 bytes)
+ *      Marker sequence number  1 for first APP2, 2 for next, etc (1 byte)
+ *      Number of markers       Total number of APP2's used (1 byte)
+ *      Profile data            (remainder of APP2 data)
+ * Decoders should use the marker sequence numbers to reassemble the profile,
+ * rather than assuming that the APP2 markers appear in the correct sequence.
+ */
+
+#define ICC_MARKER  (JPEG_APP0 + 2)     /* JPEG marker code for ICC */
+#define ICC_OVERHEAD_LEN  14            /* size of non-profile data in APP2 */
+#define MAX_BYTES_IN_MARKER  65533      /* maximum data len of a JPEG marker */
+#define MAX_DATA_BYTES_IN_MARKER  (MAX_BYTES_IN_MARKER - ICC_OVERHEAD_LEN)
+
+
+/*
+ * This routine writes the given ICC profile data into a JPEG file.  It *must*
+ * be called AFTER calling jpeg_start_compress() and BEFORE the first call to
+ * jpeg_write_scanlines().  (This ordering ensures that the APP2 marker(s) will
+ * appear after the SOI and JFIF or Adobe markers, but before all else.)
+ */
+
+GLOBAL(void)
+jpeg_write_icc_profile(j_compress_ptr cinfo, const JOCTET *icc_data_ptr,
+                       unsigned int icc_data_len)
+{
+  unsigned int num_markers;     /* total number of markers we'll write */
+  int cur_marker = 1;           /* per spec, counting starts at 1 */
+  unsigned int length;          /* number of bytes to write in this marker */
+
+  if (icc_data_ptr == NULL || icc_data_len == 0)
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+  if (cinfo->global_state < CSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Calculate the number of markers we'll need, rounding up of course */
+  num_markers = icc_data_len / MAX_DATA_BYTES_IN_MARKER;
+  if (num_markers * MAX_DATA_BYTES_IN_MARKER != icc_data_len)
+    num_markers++;
+
+  while (icc_data_len > 0) {
+    /* length of profile to put in this marker */
+    length = icc_data_len;
+    if (length > MAX_DATA_BYTES_IN_MARKER)
+      length = MAX_DATA_BYTES_IN_MARKER;
+    icc_data_len -= length;
+
+    /* Write the JPEG marker header (APP2 code and marker length) */
+    jpeg_write_m_header(cinfo, ICC_MARKER,
+                        (unsigned int)(length + ICC_OVERHEAD_LEN));
+
+    /* Write the marker identifying string "ICC_PROFILE" (null-terminated).  We
+     * code it in this less-than-transparent way so that the code works even if
+     * the local character set is not ASCII.
+     */
+    jpeg_write_m_byte(cinfo, 0x49);
+    jpeg_write_m_byte(cinfo, 0x43);
+    jpeg_write_m_byte(cinfo, 0x43);
+    jpeg_write_m_byte(cinfo, 0x5F);
+    jpeg_write_m_byte(cinfo, 0x50);
+    jpeg_write_m_byte(cinfo, 0x52);
+    jpeg_write_m_byte(cinfo, 0x4F);
+    jpeg_write_m_byte(cinfo, 0x46);
+    jpeg_write_m_byte(cinfo, 0x49);
+    jpeg_write_m_byte(cinfo, 0x4C);
+    jpeg_write_m_byte(cinfo, 0x45);
+    jpeg_write_m_byte(cinfo, 0x0);
+
+    /* Add the sequencing info */
+    jpeg_write_m_byte(cinfo, cur_marker);
+    jpeg_write_m_byte(cinfo, (int)num_markers);
+
+    /* Add the profile data */
+    while (length--) {
+      jpeg_write_m_byte(cinfo, *icc_data_ptr);
+      icc_data_ptr++;
+    }
+    cur_marker++;
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcinit.c Added

@@ -0,0 +1,80 @@
+/*
+ * jcinit.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains initialization logic for the JPEG compressor.
+ * This routine is in charge of selecting the modules to be executed and
+ * making an initialization call to each one.
+ *
+ * Logically, this code belongs in jcmaster.c.  It's split out because
+ * linking this routine implies linking the entire compression library.
+ * For a transcoding-only application, we want to be able to use jcmaster.c
+ * without linking in the whole library.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jpegcomp.h"
+
+
+/*
+ * Master selection of compression modules.
+ * This is done once at the start of processing an image.  We determine
+ * which modules will be used and give them appropriate initialization calls.
+ */
+
+GLOBAL(void)
+jinit_compress_master(j_compress_ptr cinfo)
+{
+  /* Initialize master control (includes parameter checking/processing) */
+  jinit_c_master_control(cinfo, FALSE /* full compression */);
+
+  /* Preprocessing */
+  if (!cinfo->raw_data_in) {
+    jinit_color_converter(cinfo);
+    jinit_downsampler(cinfo);
+    jinit_c_prep_controller(cinfo, FALSE /* never need full buffer here */);
+  }
+  /* Forward DCT */
+  jinit_forward_dct(cinfo);
+  /* Entropy encoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+#ifdef C_ARITH_CODING_SUPPORTED
+    jinit_arith_encoder(cinfo);
+#else
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+#endif
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+      jinit_phuff_encoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_encoder(cinfo);
+  }
+
+  /* Need a full-image coefficient buffer in any multi-pass mode. */
+  jinit_c_coef_controller(cinfo, (boolean)(cinfo->num_scans > 1 ||
+                                           cinfo->optimize_coding));
+  jinit_c_main_controller(cinfo, FALSE /* never need full buffer here */);
+
+  jinit_marker_writer(cinfo);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr)cinfo);
+
+  /* Write the datastream header (SOI) immediately.
+   * Frame and scan headers are postponed till later.
+   * This lets application insert special markers after the SOI.
+   */
+  (*cinfo->marker->write_file_header) (cinfo);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcmainct.c Added

@@ -0,0 +1,162 @@
+/*
+ * jcmainct.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the main buffer controller for compression.
+ * The main buffer lies between the pre-processor and the JPEG
+ * compressor proper; it holds downsampled data in the JPEG colorspace.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_main_controller pub; /* public fields */
+
+  JDIMENSION cur_iMCU_row;      /* number of current iMCU row */
+  JDIMENSION rowgroup_ctr;      /* counts row groups received in iMCU row */
+  boolean suspended;            /* remember if we suspended output */
+  J_BUF_MODE pass_mode;         /* current operating mode */
+
+  /* If using just a strip buffer, this points to the entire set of buffers
+   * (we allocate one for each component).  In the full-image case, this
+   * points to the currently accessible strips of the virtual arrays.
+   */
+  JSAMPARRAY bufferMAX_COMPONENTS;
+} my_main_controller;
+
+typedef my_main_controller *my_main_ptr;
+
+
+/* Forward declarations */
+METHODDEF(void) process_data_simple_main(j_compress_ptr cinfo,
+                                         JSAMPARRAY input_buf,
+                                         JDIMENSION *in_row_ctr,
+                                         JDIMENSION in_rows_avail);
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_main(j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+
+  /* Do nothing in raw-data mode. */
+  if (cinfo->raw_data_in)
+    return;
+
+  if (pass_mode != JBUF_PASS_THRU)
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  main_ptr->cur_iMCU_row = 0;   /* initialize counters */
+  main_ptr->rowgroup_ctr = 0;
+  main_ptr->suspended = FALSE;
+  main_ptr->pass_mode = pass_mode;      /* save mode for use by process_data */
+  main_ptr->pub.process_data = process_data_simple_main;
+}
+
+
+/*
+ * Process some data.
+ * This routine handles the simple pass-through mode,
+ * where we have only a strip buffer.
+ */
+
+METHODDEF(void)
+process_data_simple_main(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                         JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail)
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+
+  while (main_ptr->cur_iMCU_row < cinfo->total_iMCU_rows) {
+    /* Read input data if we haven't filled the main buffer yet */
+    if (main_ptr->rowgroup_ctr < DCTSIZE)
+      (*cinfo->prep->pre_process_data) (cinfo, input_buf, in_row_ctr,
+                                        in_rows_avail, main_ptr->buffer,
+                                        &main_ptr->rowgroup_ctr,
+                                        (JDIMENSION)DCTSIZE);
+
+    /* If we don't have a full iMCU row buffered, return to application for
+     * more data.  Note that preprocessor will always pad to fill the iMCU row
+     * at the bottom of the image.
+     */
+    if (main_ptr->rowgroup_ctr != DCTSIZE)
+      return;
+
+    /* Send the completed row to the compressor */
+    if (!(*cinfo->coef->compress_data) (cinfo, main_ptr->buffer)) {
+      /* If compressor did not consume the whole row, then we must need to
+       * suspend processing and return to the application.  In this situation
+       * we pretend we didn't yet consume the last input row; otherwise, if
+       * it happened to be the last row of the image, the application would
+       * think we were done.
+       */
+      if (!main_ptr->suspended) {
+        (*in_row_ctr)--;
+        main_ptr->suspended = TRUE;
+      }
+      return;
+    }
+    /* We did finish the row.  Undo our little suspension hack if a previous
+     * call suspended; then mark the main buffer empty.
+     */
+    if (main_ptr->suspended) {
+      (*in_row_ctr)++;
+      main_ptr->suspended = FALSE;
+    }
+    main_ptr->rowgroup_ctr = 0;
+    main_ptr->cur_iMCU_row++;
+  }
+}
+
+
+/*
+ * Initialize main buffer controller.
+ */
+
+GLOBAL(void)
+jinit_c_main_controller(j_compress_ptr cinfo, boolean need_full_buffer)
+{
+  my_main_ptr main_ptr;
+  int ci;
+  jpeg_component_info *compptr;
+
+  main_ptr = (my_main_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_main_controller));
+  cinfo->main = (struct jpeg_c_main_controller *)main_ptr;
+  main_ptr->pub.start_pass = start_pass_main;
+
+  /* We don't need to create a buffer in raw-data mode. */
+  if (cinfo->raw_data_in)
+    return;
+
+  /* Create the buffer.  It holds downsampled data, so each component
+   * may be of a different size.
+   */
+  if (need_full_buffer) {
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+  } else {
+    /* Allocate a strip buffer for each component */
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+         ci++, compptr++) {
+      main_ptr->bufferci = (*cinfo->mem->alloc_sarray)
+        ((j_common_ptr)cinfo, JPOOL_IMAGE,
+         compptr->width_in_blocks * DCTSIZE,
+         (JDIMENSION)(compptr->v_samp_factor * DCTSIZE));
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcmarker.c Added

@@ -0,0 +1,664 @@
+/*
+ * jcmarker.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * Modified 2003-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write JPEG datastream markers.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jpegcomp.h"
+
+
+typedef enum {                  /* JPEG marker codes */
+  M_SOF0  = 0xc0,
+  M_SOF1  = 0xc1,
+  M_SOF2  = 0xc2,
+  M_SOF3  = 0xc3,
+
+  M_SOF5  = 0xc5,
+  M_SOF6  = 0xc6,
+  M_SOF7  = 0xc7,
+
+  M_JPG   = 0xc8,
+  M_SOF9  = 0xc9,
+  M_SOF10 = 0xca,
+  M_SOF11 = 0xcb,
+
+  M_SOF13 = 0xcd,
+  M_SOF14 = 0xce,
+  M_SOF15 = 0xcf,
+
+  M_DHT   = 0xc4,
+
+  M_DAC   = 0xcc,
+
+  M_RST0  = 0xd0,
+  M_RST1  = 0xd1,
+  M_RST2  = 0xd2,
+  M_RST3  = 0xd3,
+  M_RST4  = 0xd4,
+  M_RST5  = 0xd5,
+  M_RST6  = 0xd6,
+  M_RST7  = 0xd7,
+
+  M_SOI   = 0xd8,
+  M_EOI   = 0xd9,
+  M_SOS   = 0xda,
+  M_DQT   = 0xdb,
+  M_DNL   = 0xdc,
+  M_DRI   = 0xdd,
+  M_DHP   = 0xde,
+  M_EXP   = 0xdf,
+
+  M_APP0  = 0xe0,
+  M_APP1  = 0xe1,
+  M_APP2  = 0xe2,
+  M_APP3  = 0xe3,
+  M_APP4  = 0xe4,
+  M_APP5  = 0xe5,
+  M_APP6  = 0xe6,
+  M_APP7  = 0xe7,
+  M_APP8  = 0xe8,
+  M_APP9  = 0xe9,
+  M_APP10 = 0xea,
+  M_APP11 = 0xeb,
+  M_APP12 = 0xec,
+  M_APP13 = 0xed,
+  M_APP14 = 0xee,
+  M_APP15 = 0xef,
+
+  M_JPG0  = 0xf0,
+  M_JPG13 = 0xfd,
+  M_COM   = 0xfe,
+
+  M_TEM   = 0x01,
+
+  M_ERROR = 0x100
+} JPEG_MARKER;
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_marker_writer pub; /* public fields */
+
+  unsigned int last_restart_interval; /* last DRI value emitted; 0 after SOI */
+} my_marker_writer;
+
+typedef my_marker_writer *my_marker_ptr;
+
+
+/*
+ * Basic output routines.
+ *
+ * Note that we do not support suspension while writing a marker.
+ * Therefore, an application using suspension must ensure that there is
+ * enough buffer space for the initial markers (typ. 600-700 bytes) before
+ * calling jpeg_start_compress, and enough space to write the trailing EOI
+ * (a few bytes) before calling jpeg_finish_compress.  Multipass compression
+ * modes are not supported at all with suspension, so those two are the only
+ * points where markers will be written.
+ */
+
+LOCAL(void)
+emit_byte(j_compress_ptr cinfo, int val)
+/* Emit a byte */
+{
+  struct jpeg_destination_mgr *dest = cinfo->dest;
+
+  *(dest->next_output_byte)++ = (JOCTET)val;
+  if (--dest->free_in_buffer == 0) {
+    if (!(*dest->empty_output_buffer) (cinfo))
+      ERREXIT(cinfo, JERR_CANT_SUSPEND);
+  }
+}
+
+
+LOCAL(void)
+emit_marker(j_compress_ptr cinfo, JPEG_MARKER mark)
+/* Emit a marker code */
+{
+  emit_byte(cinfo, 0xFF);
+  emit_byte(cinfo, (int)mark);
+}
+
+
+LOCAL(void)
+emit_2bytes(j_compress_ptr cinfo, int value)
+/* Emit a 2-byte integer; these are always MSB first in JPEG files */
+{
+  emit_byte(cinfo, (value >> 8) & 0xFF);
+  emit_byte(cinfo, value & 0xFF);
+}
+
+
+/*
+ * Routines to write specific marker types.
+ */
+
+LOCAL(int)
+emit_dqt(j_compress_ptr cinfo, int index)
+/* Emit a DQT marker */
+/* Returns the precision used (0 = 8bits, 1 = 16bits) for baseline checking */
+{
+  JQUANT_TBL *qtbl = cinfo->quant_tbl_ptrsindex;
+  int prec;
+  int i;
+
+  if (qtbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, index);
+
+  prec = 0;
+  for (i = 0; i < DCTSIZE2; i++) {
+    if (qtbl->quantvali > 255)
+      prec = 1;
+  }
+
+  if (!qtbl->sent_table) {
+    emit_marker(cinfo, M_DQT);
+
+    emit_2bytes(cinfo, prec ? DCTSIZE2 * 2 + 1 + 2 : DCTSIZE2 + 1 + 2);
+
+    emit_byte(cinfo, index + (prec << 4));
+
+    for (i = 0; i < DCTSIZE2; i++) {
+      /* The table entries must be emitted in zigzag order. */
+      unsigned int qval = qtbl->quantvaljpeg_natural_orderi;
+      if (prec)
+        emit_byte(cinfo, (int)(qval >> 8));
+      emit_byte(cinfo, (int)(qval & 0xFF));
+    }
+
+    qtbl->sent_table = TRUE;
+  }
+
+  return prec;
+}
+
+
+LOCAL(void)
+emit_dht(j_compress_ptr cinfo, int index, boolean is_ac)
+/* Emit a DHT marker */
+{
+  JHUFF_TBL *htbl;
+  int length, i;
+
+  if (is_ac) {
+    htbl = cinfo->ac_huff_tbl_ptrsindex;
+    index += 0x10;              /* output index has AC bit set */
+  } else {
+    htbl = cinfo->dc_huff_tbl_ptrsindex;
+  }
+
+  if (htbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, index);
+
+  if (!htbl->sent_table) {
+    emit_marker(cinfo, M_DHT);
+
+    length = 0;
+    for (i = 1; i <= 16; i++)
+      length += htbl->bitsi;
+
+    emit_2bytes(cinfo, length + 2 + 1 + 16);
+    emit_byte(cinfo, index);
+
+    for (i = 1; i <= 16; i++)
+      emit_byte(cinfo, htbl->bitsi);
+
+    for (i = 0; i < length; i++)
+      emit_byte(cinfo, htbl->huffvali);
+
+    htbl->sent_table = TRUE;
+  }
+}
+
+
+LOCAL(void)
+emit_dac(j_compress_ptr cinfo)
+/* Emit a DAC marker */
+/* Since the useful info is so small, we want to emit all the tables in */
+/* one DAC marker.  Therefore this routine does its own scan of the table. */
+{
+#ifdef C_ARITH_CODING_SUPPORTED
+  char dc_in_useNUM_ARITH_TBLS;
+  char ac_in_useNUM_ARITH_TBLS;
+  int length, i;
+  jpeg_component_info *compptr;
+
+  for (i = 0; i < NUM_ARITH_TBLS; i++)
+    dc_in_usei = ac_in_usei = 0;
+
+  for (i = 0; i < cinfo->comps_in_scan; i++) {
+    compptr = cinfo->cur_comp_infoi;
+    /* DC needs no table for refinement scan */
+    if (cinfo->Ss == 0 && cinfo->Ah == 0)
+      dc_in_usecompptr->dc_tbl_no = 1;
+    /* AC needs no table when not present */
+    if (cinfo->Se)
+      ac_in_usecompptr->ac_tbl_no = 1;
+  }
+
+  length = 0;
+  for (i = 0; i < NUM_ARITH_TBLS; i++)
+    length += dc_in_usei + ac_in_usei;
+
+  if (length) {
+    emit_marker(cinfo, M_DAC);
+
+    emit_2bytes(cinfo, length * 2 + 2);
+
+    for (i = 0; i < NUM_ARITH_TBLS; i++) {
+      if (dc_in_usei) {
+        emit_byte(cinfo, i);
+        emit_byte(cinfo, cinfo->arith_dc_Li + (cinfo->arith_dc_Ui << 4));
+      }
+      if (ac_in_usei) {
+        emit_byte(cinfo, i + 0x10);
+        emit_byte(cinfo, cinfo->arith_ac_Ki);
+      }
+    }
+  }
+#endif /* C_ARITH_CODING_SUPPORTED */
+}
+
+
+LOCAL(void)
+emit_dri(j_compress_ptr cinfo)
+/* Emit a DRI marker */
+{
+  emit_marker(cinfo, M_DRI);
+
+  emit_2bytes(cinfo, 4);        /* fixed length */
+
+  emit_2bytes(cinfo, (int)cinfo->restart_interval);
+}
+
+
+LOCAL(void)
+emit_sof(j_compress_ptr cinfo, JPEG_MARKER code)
+/* Emit a SOF marker */
+{
+  int ci;
+  jpeg_component_info *compptr;
+
+  emit_marker(cinfo, code);
+
+  emit_2bytes(cinfo, 3 * cinfo->num_components + 2 + 5 + 1); /* length */
+
+  /* Make sure image isn't bigger than SOF field can handle */
+  if ((long)cinfo->_jpeg_height > 65535L || (long)cinfo->_jpeg_width > 65535L)
+    ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int)65535);
+
+  emit_byte(cinfo, cinfo->data_precision);
+  emit_2bytes(cinfo, (int)cinfo->_jpeg_height);
+  emit_2bytes(cinfo, (int)cinfo->_jpeg_width);
+
+  emit_byte(cinfo, cinfo->num_components);
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    emit_byte(cinfo, compptr->component_id);
+    emit_byte(cinfo, (compptr->h_samp_factor << 4) + compptr->v_samp_factor);
+    emit_byte(cinfo, compptr->quant_tbl_no);
+  }
+}
+
+
+LOCAL(void)
+emit_sos(j_compress_ptr cinfo)
+/* Emit a SOS marker */
+{
+  int i, td, ta;
+  jpeg_component_info *compptr;
+
+  emit_marker(cinfo, M_SOS);
+
+  emit_2bytes(cinfo, 2 * cinfo->comps_in_scan + 2 + 1 + 3); /* length */
+
+  emit_byte(cinfo, cinfo->comps_in_scan);
+
+  for (i = 0; i < cinfo->comps_in_scan; i++) {
+    compptr = cinfo->cur_comp_infoi;
+    emit_byte(cinfo, compptr->component_id);
+
+    /* We emit 0 for unused field(s); this is recommended by the P&M text
+     * but does not seem to be specified in the standard.
+     */
+
+    /* DC needs no table for refinement scan */
+    td = cinfo->Ss == 0 && cinfo->Ah == 0 ? compptr->dc_tbl_no : 0;
+    /* AC needs no table when not present */
+    ta = cinfo->Se ? compptr->ac_tbl_no : 0;
+
+    emit_byte(cinfo, (td << 4) + ta);
+  }
+
+  emit_byte(cinfo, cinfo->Ss);
+  emit_byte(cinfo, cinfo->Se);
+  emit_byte(cinfo, (cinfo->Ah << 4) + cinfo->Al);
+}
+
+
+LOCAL(void)
+emit_jfif_app0(j_compress_ptr cinfo)
+/* Emit a JFIF-compliant APP0 marker */
+{
+  /*
+   * Length of APP0 block       (2 bytes)
+   * Block ID                   (4 bytes - ASCII "JFIF")
+   * Zero byte                  (1 byte to terminate the ID string)
+   * Version Major, Minor       (2 bytes - major first)
+   * Units                      (1 byte - 0x00 = none, 0x01 = inch, 0x02 = cm)
+   * Xdpu                       (2 bytes - dots per unit horizontal)
+   * Ydpu                       (2 bytes - dots per unit vertical)
+   * Thumbnail X size           (1 byte)
+   * Thumbnail Y size           (1 byte)
+   */
+
+  emit_marker(cinfo, M_APP0);
+
+  emit_2bytes(cinfo, 2 + 4 + 1 + 2 + 1 + 2 + 2 + 1 + 1); /* length */
+
+  emit_byte(cinfo, 0x4A);       /* Identifier: ASCII "JFIF" */
+  emit_byte(cinfo, 0x46);
+  emit_byte(cinfo, 0x49);
+  emit_byte(cinfo, 0x46);
+  emit_byte(cinfo, 0);
+  emit_byte(cinfo, cinfo->JFIF_major_version); /* Version fields */
+  emit_byte(cinfo, cinfo->JFIF_minor_version);
+  emit_byte(cinfo, cinfo->density_unit); /* Pixel size information */
+  emit_2bytes(cinfo, (int)cinfo->X_density);
+  emit_2bytes(cinfo, (int)cinfo->Y_density);
+  emit_byte(cinfo, 0);          /* No thumbnail image */
+  emit_byte(cinfo, 0);
+}
+
+
+LOCAL(void)
+emit_adobe_app14(j_compress_ptr cinfo)
+/* Emit an Adobe APP14 marker */
+{
+  /*
+   * Length of APP14 block      (2 bytes)
+   * Block ID                   (5 bytes - ASCII "Adobe")
+   * Version Number             (2 bytes - currently 100)
+   * Flags0                     (2 bytes - currently 0)
+   * Flags1                     (2 bytes - currently 0)
+   * Color transform            (1 byte)
+   *
+   * Although Adobe TN 5116 mentions Version = 101, all the Adobe files
+   * now in circulation seem to use Version = 100, so that's what we write.
+   *
+   * We write the color transform byte as 1 if the JPEG color space is
+   * YCbCr, 2 if it's YCCK, 0 otherwise.  Adobe's definition has to do with
+   * whether the encoder performed a transformation, which is pretty useless.
+   */
+
+  emit_marker(cinfo, M_APP14);
+
+  emit_2bytes(cinfo, 2 + 5 + 2 + 2 + 2 + 1); /* length */
+
+  emit_byte(cinfo, 0x41);       /* Identifier: ASCII "Adobe" */
+  emit_byte(cinfo, 0x64);
+  emit_byte(cinfo, 0x6F);
+  emit_byte(cinfo, 0x62);
+  emit_byte(cinfo, 0x65);
+  emit_2bytes(cinfo, 100);      /* Version */
+  emit_2bytes(cinfo, 0);        /* Flags0 */
+  emit_2bytes(cinfo, 0);        /* Flags1 */
+  switch (cinfo->jpeg_color_space) {
+  case JCS_YCbCr:
+    emit_byte(cinfo, 1);        /* Color transform = 1 */
+    break;
+  case JCS_YCCK:
+    emit_byte(cinfo, 2);        /* Color transform = 2 */
+    break;
+  default:
+    emit_byte(cinfo, 0);        /* Color transform = 0 */
+    break;
+  }
+}
+
+
+/*
+ * These routines allow writing an arbitrary marker with parameters.
+ * The only intended use is to emit COM or APPn markers after calling
+ * write_file_header and before calling write_frame_header.
+ * Other uses are not guaranteed to produce desirable results.
+ * Counting the parameter bytes properly is the caller's responsibility.
+ */
+
+METHODDEF(void)
+write_marker_header(j_compress_ptr cinfo, int marker, unsigned int datalen)
+/* Emit an arbitrary marker header */
+{
+  if (datalen > (unsigned int)65533)            /* safety check */
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  emit_marker(cinfo, (JPEG_MARKER)marker);
+
+  emit_2bytes(cinfo, (int)(datalen + 2));       /* total length */
+}
+
+METHODDEF(void)
+write_marker_byte(j_compress_ptr cinfo, int val)
+/* Emit one byte of marker parameters following write_marker_header */
+{
+  emit_byte(cinfo, val);
+}
+
+
+/*
+ * Write datastream header.
+ * This consists of an SOI and optional APPn markers.
+ * We recommend use of the JFIF marker, but not the Adobe marker,
+ * when using YCbCr or grayscale data.  The JFIF marker should NOT
+ * be used for any other JPEG colorspace.  The Adobe marker is helpful
+ * to distinguish RGB, CMYK, and YCCK colorspaces.
+ * Note that an application can write additional header markers after
+ * jpeg_start_compress returns.
+ */
+
+METHODDEF(void)
+write_file_header(j_compress_ptr cinfo)
+{
+  my_marker_ptr marker = (my_marker_ptr)cinfo->marker;
+
+  emit_marker(cinfo, M_SOI);    /* first the SOI */
+
+  /* SOI is defined to reset restart interval to 0 */
+  marker->last_restart_interval = 0;
+
+  if (cinfo->write_JFIF_header) /* next an optional JFIF APP0 */
+    emit_jfif_app0(cinfo);
+  if (cinfo->write_Adobe_marker) /* next an optional Adobe APP14 */
+    emit_adobe_app14(cinfo);
+}
+
+
+/*
+ * Write frame header.
+ * This consists of DQT and SOFn markers.
+ * Note that we do not emit the SOF until we have emitted the DQT(s).
+ * This avoids compatibility problems with incorrect implementations that
+ * try to error-check the quant table numbers as soon as they see the SOF.
+ */
+
+METHODDEF(void)
+write_frame_header(j_compress_ptr cinfo)
+{
+  int ci, prec;
+  boolean is_baseline;
+  jpeg_component_info *compptr;
+
+  /* Emit DQT for each quantization table.
+   * Note that emit_dqt() suppresses any duplicate tables.
+   */
+  prec = 0;
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    prec += emit_dqt(cinfo, compptr->quant_tbl_no);
+  }
+  /* now prec is nonzero iff there are any 16-bit quant tables. */
+
+  /* Check for a non-baseline specification.
+   * Note we assume that Huffman table numbers won't be changed later.
+   */
+  if (cinfo->arith_code || cinfo->progressive_mode ||
+      cinfo->data_precision != 8) {
+    is_baseline = FALSE;
+  } else {
+    is_baseline = TRUE;
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+         ci++, compptr++) {
+      if (compptr->dc_tbl_no > 1 || compptr->ac_tbl_no > 1)
+        is_baseline = FALSE;
+    }
+    if (prec && is_baseline) {
+      is_baseline = FALSE;
+      /* If it's baseline except for quantizer size, warn the user */
+      TRACEMS(cinfo, 0, JTRC_16BIT_TABLES);
+    }
+  }
+
+  /* Emit the proper SOF marker */
+  if (cinfo->arith_code) {
+    if (cinfo->progressive_mode)
+      emit_sof(cinfo, M_SOF10); /* SOF code for progressive arithmetic */
+    else
+      emit_sof(cinfo, M_SOF9);  /* SOF code for sequential arithmetic */
+  } else {
+    if (cinfo->progressive_mode)
+      emit_sof(cinfo, M_SOF2);  /* SOF code for progressive Huffman */
+    else if (is_baseline)
+      emit_sof(cinfo, M_SOF0);  /* SOF code for baseline implementation */
+    else
+      emit_sof(cinfo, M_SOF1);  /* SOF code for non-baseline Huffman file */
+  }
+}
+
+
+/*
+ * Write scan header.
+ * This consists of DHT or DAC markers, optional DRI, and SOS.
+ * Compressed data will be written following the SOS.
+ */
+
+METHODDEF(void)
+write_scan_header(j_compress_ptr cinfo)
+{
+  my_marker_ptr marker = (my_marker_ptr)cinfo->marker;
+  int i;
+  jpeg_component_info *compptr;
+
+  if (cinfo->arith_code) {
+    /* Emit arith conditioning info.  We may have some duplication
+     * if the file has multiple scans, but it's so small it's hardly
+     * worth worrying about.
+     */
+    emit_dac(cinfo);
+  } else {
+    /* Emit Huffman tables.
+     * Note that emit_dht() suppresses any duplicate tables.
+     */
+    for (i = 0; i < cinfo->comps_in_scan; i++) {
+      compptr = cinfo->cur_comp_infoi;
+      /* DC needs no table for refinement scan */
+      if (cinfo->Ss == 0 && cinfo->Ah == 0)
+        emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
+      /* AC needs no table when not present */
+      if (cinfo->Se)
+        emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
+    }
+  }
+
+  /* Emit DRI if required --- note that DRI value could change for each scan.
+   * We avoid wasting space with unnecessary DRIs, however.
+   */
+  if (cinfo->restart_interval != marker->last_restart_interval) {
+    emit_dri(cinfo);
+    marker->last_restart_interval = cinfo->restart_interval;
+  }
+
+  emit_sos(cinfo);
+}
+
+
+/*
+ * Write datastream trailer.
+ */
+
+METHODDEF(void)
+write_file_trailer(j_compress_ptr cinfo)
+{
+  emit_marker(cinfo, M_EOI);
+}
+
+
+/*
+ * Write an abbreviated table-specification datastream.
+ * This consists of SOI, DQT and DHT tables, and EOI.
+ * Any table that is defined and not marked sent_table = TRUE will be
+ * emitted.  Note that all tables will be marked sent_table = TRUE at exit.
+ */
+
+METHODDEF(void)
+write_tables_only(j_compress_ptr cinfo)
+{
+  int i;
+
+  emit_marker(cinfo, M_SOI);
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    if (cinfo->quant_tbl_ptrsi != NULL)
+      (void)emit_dqt(cinfo, i);
+  }
+
+  if (!cinfo->arith_code) {
+    for (i = 0; i < NUM_HUFF_TBLS; i++) {
+      if (cinfo->dc_huff_tbl_ptrsi != NULL)
+        emit_dht(cinfo, i, FALSE);
+      if (cinfo->ac_huff_tbl_ptrsi != NULL)
+        emit_dht(cinfo, i, TRUE);
+    }
+  }
+
+  emit_marker(cinfo, M_EOI);
+}
+
+
+/*
+ * Initialize the marker writer module.
+ */
+
+GLOBAL(void)
+jinit_marker_writer(j_compress_ptr cinfo)
+{
+  my_marker_ptr marker;
+
+  /* Create the subobject */
+  marker = (my_marker_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_marker_writer));
+  cinfo->marker = (struct jpeg_marker_writer *)marker;
+  /* Initialize method pointers */
+  marker->pub.write_file_header = write_file_header;
+  marker->pub.write_frame_header = write_frame_header;
+  marker->pub.write_scan_header = write_scan_header;
+  marker->pub.write_file_trailer = write_file_trailer;
+  marker->pub.write_tables_only = write_tables_only;
+  marker->pub.write_marker_header = write_marker_header;
+  marker->pub.write_marker_byte = write_marker_byte;
+  /* Initialize private state */
+  marker->last_restart_interval = 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcmaster.c Added

@@ -0,0 +1,640 @@
+/*
+ * jcmaster.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2003-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2016, 2018, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains master control logic for the JPEG compressor.
+ * These routines are concerned with parameter validation, initial setup,
+ * and inter-pass control (determining the number of passes and the work
+ * to be done in each pass).
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jpegcomp.h"
+#include "jconfigint.h"
+
+
+/* Private state */
+
+typedef enum {
+  main_pass,                    /* input data, also do first output step */
+  huff_opt_pass,                /* Huffman code optimization pass */
+  output_pass                   /* data output pass */
+} c_pass_type;
+
+typedef struct {
+  struct jpeg_comp_master pub;  /* public fields */
+
+  c_pass_type pass_type;        /* the type of the current pass */
+
+  int pass_number;              /* # of passes completed */
+  int total_passes;             /* total # of passes needed */
+
+  int scan_number;              /* current index in scan_info */
+
+  /*
+   * This is here so we can add libjpeg-turbo version/build information to the
+   * global string table without introducing a new global symbol.  Adding this
+   * information to the global string table allows one to examine a binary
+   * object and determine which version of libjpeg-turbo it was built from or
+   * linked against.
+   */
+  const char *jpeg_version;
+
+} my_comp_master;
+
+typedef my_comp_master *my_master_ptr;
+
+
+/*
+ * Support routines that do various essential calculations.
+ */
+
+#if JPEG_LIB_VERSION >= 70
+/*
+ * Compute JPEG image dimensions and related values.
+ * NOTE: this is exported for possible use by application.
+ * Hence it mustn't do anything that can't be done twice.
+ */
+
+GLOBAL(void)
+jpeg_calc_jpeg_dimensions(j_compress_ptr cinfo)
+/* Do computations that are needed before master selection phase */
+{
+  /* Hardwire it to "no scaling" */
+  cinfo->jpeg_width = cinfo->image_width;
+  cinfo->jpeg_height = cinfo->image_height;
+  cinfo->min_DCT_h_scaled_size = DCTSIZE;
+  cinfo->min_DCT_v_scaled_size = DCTSIZE;
+}
+#endif
+
+
+LOCAL(void)
+initial_setup(j_compress_ptr cinfo, boolean transcode_only)
+/* Do computations that are needed before master selection phase */
+{
+  int ci;
+  jpeg_component_info *compptr;
+  long samplesperrow;
+  JDIMENSION jd_samplesperrow;
+
+#if JPEG_LIB_VERSION >= 70
+#if JPEG_LIB_VERSION >= 80
+  if (!transcode_only)
+#endif
+    jpeg_calc_jpeg_dimensions(cinfo);
+#endif
+
+  /* Sanity check on image dimensions */
+  if (cinfo->_jpeg_height <= 0 || cinfo->_jpeg_width <= 0 ||
+      cinfo->num_components <= 0 || cinfo->input_components <= 0)
+    ERREXIT(cinfo, JERR_EMPTY_IMAGE);
+
+  /* Make sure image isn't bigger than I can handle */
+  if ((long)cinfo->_jpeg_height > (long)JPEG_MAX_DIMENSION ||
+      (long)cinfo->_jpeg_width > (long)JPEG_MAX_DIMENSION)
+    ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int)JPEG_MAX_DIMENSION);
+
+  /* Width of an input scanline must be representable as JDIMENSION. */
+  samplesperrow = (long)cinfo->image_width * (long)cinfo->input_components;
+  jd_samplesperrow = (JDIMENSION)samplesperrow;
+  if ((long)jd_samplesperrow != samplesperrow)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+
+  /* For now, precision must match compiled-in value... */
+  if (cinfo->data_precision != BITS_IN_JSAMPLE)
+    ERREXIT1(cinfo, JERR_BAD_PRECISION, cinfo->data_precision);
+
+  /* Check that number of components won't exceed internal array sizes */
+  if (cinfo->num_components > MAX_COMPONENTS)
+    ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+             MAX_COMPONENTS);
+
+  /* Compute maximum sampling factors; check factor validity */
+  cinfo->max_h_samp_factor = 1;
+  cinfo->max_v_samp_factor = 1;
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    if (compptr->h_samp_factor <= 0 ||
+        compptr->h_samp_factor > MAX_SAMP_FACTOR ||
+        compptr->v_samp_factor <= 0 ||
+        compptr->v_samp_factor > MAX_SAMP_FACTOR)
+      ERREXIT(cinfo, JERR_BAD_SAMPLING);
+    cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
+                                   compptr->h_samp_factor);
+    cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
+                                   compptr->v_samp_factor);
+  }
+
+  /* Compute dimensions of components */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Fill in the correct component_index value; don't rely on application */
+    compptr->component_index = ci;
+    /* For compression, we never do DCT scaling. */
+#if JPEG_LIB_VERSION >= 70
+    compptr->DCT_h_scaled_size = compptr->DCT_v_scaled_size = DCTSIZE;
+#else
+    compptr->DCT_scaled_size = DCTSIZE;
+#endif
+    /* Size in DCT blocks */
+    compptr->width_in_blocks = (JDIMENSION)
+      jdiv_round_up((long)cinfo->_jpeg_width * (long)compptr->h_samp_factor,
+                    (long)(cinfo->max_h_samp_factor * DCTSIZE));
+    compptr->height_in_blocks = (JDIMENSION)
+      jdiv_round_up((long)cinfo->_jpeg_height * (long)compptr->v_samp_factor,
+                    (long)(cinfo->max_v_samp_factor * DCTSIZE));
+    /* Size in samples */
+    compptr->downsampled_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->_jpeg_width * (long)compptr->h_samp_factor,
+                    (long)cinfo->max_h_samp_factor);
+    compptr->downsampled_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->_jpeg_height * (long)compptr->v_samp_factor,
+                    (long)cinfo->max_v_samp_factor);
+    /* Mark component needed (this flag isn't actually used for compression) */
+    compptr->component_needed = TRUE;
+  }
+
+  /* Compute number of fully interleaved MCU rows (number of times that
+   * main controller will call coefficient controller).
+   */
+  cinfo->total_iMCU_rows = (JDIMENSION)
+    jdiv_round_up((long)cinfo->_jpeg_height,
+                  (long)(cinfo->max_v_samp_factor * DCTSIZE));
+}
+
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+
+LOCAL(void)
+validate_script(j_compress_ptr cinfo)
+/* Verify that the scan script in cinfo->scan_info is valid; also
+ * determine whether it uses progressive JPEG, and set cinfo->progressive_mode.
+ */
+{
+  const jpeg_scan_info *scanptr;
+  int scanno, ncomps, ci, coefi, thisi;
+  int Ss, Se, Ah, Al;
+  boolean component_sentMAX_COMPONENTS;
+#ifdef C_PROGRESSIVE_SUPPORTED
+  int *last_bitpos_ptr;
+  int last_bitposMAX_COMPONENTSDCTSIZE2;
+  /* -1 until that coefficient has been seen; then last Al for it */
+#endif
+
+  if (cinfo->num_scans <= 0)
+    ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, 0);
+
+  /* For sequential JPEG, all scans must have Ss=0, Se=DCTSIZE2-1;
+   * for progressive JPEG, no scan can have this.
+   */
+  scanptr = cinfo->scan_info;
+  if (scanptr->Ss != 0 || scanptr->Se != DCTSIZE2 - 1) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+    cinfo->progressive_mode = TRUE;
+    last_bitpos_ptr = &last_bitpos00;
+    for (ci = 0; ci < cinfo->num_components; ci++)
+      for (coefi = 0; coefi < DCTSIZE2; coefi++)
+        *last_bitpos_ptr++ = -1;
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    cinfo->progressive_mode = FALSE;
+    for (ci = 0; ci < cinfo->num_components; ci++)
+      component_sentci = FALSE;
+  }
+
+  for (scanno = 1; scanno <= cinfo->num_scans; scanptr++, scanno++) {
+    /* Validate component indexes */
+    ncomps = scanptr->comps_in_scan;
+    if (ncomps <= 0 || ncomps > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, ncomps, MAX_COMPS_IN_SCAN);
+    for (ci = 0; ci < ncomps; ci++) {
+      thisi = scanptr->component_indexci;
+      if (thisi < 0 || thisi >= cinfo->num_components)
+        ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+      /* Components must appear in SOF order within each scan */
+      if (ci > 0 && thisi <= scanptr->component_indexci - 1)
+        ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+    }
+    /* Validate progression parameters */
+    Ss = scanptr->Ss;
+    Se = scanptr->Se;
+    Ah = scanptr->Ah;
+    Al = scanptr->Al;
+    if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+      /* Rec. ITU-T T.81 | ISO/IEC 10918-1 simply gives the ranges 0..13 for Ah
+       * and Al, but that seems wrong: the upper bound ought to depend on data
+       * precision.  Perhaps they really meant 0..N+1 for N-bit precision.
+       * Here we allow 0..10 for 8-bit data; Al larger than 10 results in
+       * out-of-range reconstructed DC values during the first DC scan,
+       * which might cause problems for some decoders.
+       */
+#if BITS_IN_JSAMPLE == 8
+#define MAX_AH_AL  10
+#else
+#define MAX_AH_AL  13
+#endif
+      if (Ss < 0 || Ss >= DCTSIZE2 || Se < Ss || Se >= DCTSIZE2 ||
+          Ah < 0 || Ah > MAX_AH_AL || Al < 0 || Al > MAX_AH_AL)
+        ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      if (Ss == 0) {
+        if (Se != 0)            /* DC and AC together not OK */
+          ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      } else {
+        if (ncomps != 1)        /* AC scans must be for only one component */
+          ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      }
+      for (ci = 0; ci < ncomps; ci++) {
+        last_bitpos_ptr = &last_bitposscanptr->component_indexci0;
+        if (Ss != 0 && last_bitpos_ptr0 < 0) /* AC without prior DC scan */
+          ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+        for (coefi = Ss; coefi <= Se; coefi++) {
+          if (last_bitpos_ptrcoefi < 0) {
+            /* first scan of this coefficient */
+            if (Ah != 0)
+              ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+          } else {
+            /* not first scan */
+            if (Ah != last_bitpos_ptrcoefi || Al != Ah - 1)
+              ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+          }
+          last_bitpos_ptrcoefi = Al;
+        }
+      }
+#endif
+    } else {
+      /* For sequential JPEG, all progression parameters must be these: */
+      if (Ss != 0 || Se != DCTSIZE2 - 1 || Ah != 0 || Al != 0)
+        ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      /* Make sure components are not sent twice */
+      for (ci = 0; ci < ncomps; ci++) {
+        thisi = scanptr->component_indexci;
+        if (component_sentthisi)
+          ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+        component_sentthisi = TRUE;
+      }
+    }
+  }
+
+  /* Now verify that everything got sent. */
+  if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+    /* For progressive mode, we only check that at least some DC data
+     * got sent for each component; the spec does not require that all bits
+     * of all coefficients be transmitted.  Would it be wiser to enforce
+     * transmission of all coefficient bits??
+     */
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      if (last_bitposci0 < 0)
+        ERREXIT(cinfo, JERR_MISSING_DATA);
+    }
+#endif
+  } else {
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      if (!component_sentci)
+        ERREXIT(cinfo, JERR_MISSING_DATA);
+    }
+  }
+}
+
+#endif /* C_MULTISCAN_FILES_SUPPORTED */
+
+
+LOCAL(void)
+select_scan_parameters(j_compress_ptr cinfo)
+/* Set up the scan parameters for the current scan */
+{
+  int ci;
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  if (cinfo->scan_info != NULL) {
+    /* Prepare for current scan --- the script is already validated */
+    my_master_ptr master = (my_master_ptr)cinfo->master;
+    const jpeg_scan_info *scanptr = cinfo->scan_info + master->scan_number;
+
+    cinfo->comps_in_scan = scanptr->comps_in_scan;
+    for (ci = 0; ci < scanptr->comps_in_scan; ci++) {
+      cinfo->cur_comp_infoci =
+        &cinfo->comp_infoscanptr->component_indexci;
+    }
+    cinfo->Ss = scanptr->Ss;
+    cinfo->Se = scanptr->Se;
+    cinfo->Ah = scanptr->Ah;
+    cinfo->Al = scanptr->Al;
+  } else
+#endif
+  {
+    /* Prepare for single sequential-JPEG scan containing all components */
+    if (cinfo->num_components > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+               MAX_COMPS_IN_SCAN);
+    cinfo->comps_in_scan = cinfo->num_components;
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      cinfo->cur_comp_infoci = &cinfo->comp_infoci;
+    }
+    cinfo->Ss = 0;
+    cinfo->Se = DCTSIZE2 - 1;
+    cinfo->Ah = 0;
+    cinfo->Al = 0;
+  }
+}
+
+
+LOCAL(void)
+per_scan_setup(j_compress_ptr cinfo)
+/* Do computations that are needed before processing a JPEG scan */
+/* cinfo->comps_in_scan and cinfo->cur_comp_info are already set */
+{
+  int ci, mcublks, tmp;
+  jpeg_component_info *compptr;
+
+  if (cinfo->comps_in_scan == 1) {
+
+    /* Noninterleaved (single-component) scan */
+    compptr = cinfo->cur_comp_info0;
+
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = compptr->width_in_blocks;
+    cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
+
+    /* For noninterleaved scan, always one block per MCU */
+    compptr->MCU_width = 1;
+    compptr->MCU_height = 1;
+    compptr->MCU_blocks = 1;
+    compptr->MCU_sample_width = DCTSIZE;
+    compptr->last_col_width = 1;
+    /* For noninterleaved scans, it is convenient to define last_row_height
+     * as the number of block rows present in the last iMCU row.
+     */
+    tmp = (int)(compptr->height_in_blocks % compptr->v_samp_factor);
+    if (tmp == 0) tmp = compptr->v_samp_factor;
+    compptr->last_row_height = tmp;
+
+    /* Prepare array describing MCU composition */
+    cinfo->blocks_in_MCU = 1;
+    cinfo->MCU_membership0 = 0;
+
+  } else {
+
+    /* Interleaved (multi-component) scan */
+    if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
+               MAX_COMPS_IN_SCAN);
+
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = (JDIMENSION)
+      jdiv_round_up((long)cinfo->_jpeg_width,
+                    (long)(cinfo->max_h_samp_factor * DCTSIZE));
+    cinfo->MCU_rows_in_scan = (JDIMENSION)
+      jdiv_round_up((long)cinfo->_jpeg_height,
+                    (long)(cinfo->max_v_samp_factor * DCTSIZE));
+
+    cinfo->blocks_in_MCU = 0;
+
+    for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+      compptr = cinfo->cur_comp_infoci;
+      /* Sampling factors give # of blocks of component in each MCU */
+      compptr->MCU_width = compptr->h_samp_factor;
+      compptr->MCU_height = compptr->v_samp_factor;
+      compptr->MCU_blocks = compptr->MCU_width * compptr->MCU_height;
+      compptr->MCU_sample_width = compptr->MCU_width * DCTSIZE;
+      /* Figure number of non-dummy blocks in last MCU column & row */
+      tmp = (int)(compptr->width_in_blocks % compptr->MCU_width);
+      if (tmp == 0) tmp = compptr->MCU_width;
+      compptr->last_col_width = tmp;
+      tmp = (int)(compptr->height_in_blocks % compptr->MCU_height);
+      if (tmp == 0) tmp = compptr->MCU_height;
+      compptr->last_row_height = tmp;
+      /* Prepare array describing MCU composition */
+      mcublks = compptr->MCU_blocks;
+      if (cinfo->blocks_in_MCU + mcublks > C_MAX_BLOCKS_IN_MCU)
+        ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+      while (mcublks-- > 0) {
+        cinfo->MCU_membershipcinfo->blocks_in_MCU++ = ci;
+      }
+    }
+
+  }
+
+  /* Convert restart specified in rows to actual MCU count. */
+  /* Note that count must fit in 16 bits, so we provide limiting. */
+  if (cinfo->restart_in_rows > 0) {
+    long nominal = (long)cinfo->restart_in_rows * (long)cinfo->MCUs_per_row;
+    cinfo->restart_interval = (unsigned int)MIN(nominal, 65535L);
+  }
+}
+
+
+/*
+ * Per-pass setup.
+ * This is called at the beginning of each pass.  We determine which modules
+ * will be active during this pass and give them appropriate start_pass calls.
+ * We also set is_last_pass to indicate whether any more passes will be
+ * required.
+ */
+
+METHODDEF(void)
+prepare_for_pass(j_compress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+
+  switch (master->pass_type) {
+  case main_pass:
+    /* Initial pass: will collect input data, and do either Huffman
+     * optimization or data output for the first scan.
+     */
+    select_scan_parameters(cinfo);
+    per_scan_setup(cinfo);
+    if (!cinfo->raw_data_in) {
+      (*cinfo->cconvert->start_pass) (cinfo);
+      (*cinfo->downsample->start_pass) (cinfo);
+      (*cinfo->prep->start_pass) (cinfo, JBUF_PASS_THRU);
+    }
+    (*cinfo->fdct->start_pass) (cinfo);
+    (*cinfo->entropy->start_pass) (cinfo, cinfo->optimize_coding);
+    (*cinfo->coef->start_pass) (cinfo,
+                                (master->total_passes > 1 ?
+                                 JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+    (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
+    if (cinfo->optimize_coding) {
+      /* No immediate data output; postpone writing frame/scan headers */
+      master->pub.call_pass_startup = FALSE;
+    } else {
+      /* Will write frame/scan headers at first jpeg_write_scanlines call */
+      master->pub.call_pass_startup = TRUE;
+    }
+    break;
+#ifdef ENTROPY_OPT_SUPPORTED
+  case huff_opt_pass:
+    /* Do Huffman optimization for a scan after the first one. */
+    select_scan_parameters(cinfo);
+    per_scan_setup(cinfo);
+    if (cinfo->Ss != 0 || cinfo->Ah == 0 || cinfo->arith_code) {
+      (*cinfo->entropy->start_pass) (cinfo, TRUE);
+      (*cinfo->coef->start_pass) (cinfo, JBUF_CRANK_DEST);
+      master->pub.call_pass_startup = FALSE;
+      break;
+    }
+    /* Special case: Huffman DC refinement scans need no Huffman table
+     * and therefore we can skip the optimization pass for them.
+     */
+    master->pass_type = output_pass;
+    master->pass_number++;
+#endif
+    FALLTHROUGH                 /*FALLTHROUGH*/
+  case output_pass:
+    /* Do a data-output pass. */
+    /* We need not repeat per-scan setup if prior optimization pass did it. */
+    if (!cinfo->optimize_coding) {
+      select_scan_parameters(cinfo);
+      per_scan_setup(cinfo);
+    }
+    (*cinfo->entropy->start_pass) (cinfo, FALSE);
+    (*cinfo->coef->start_pass) (cinfo, JBUF_CRANK_DEST);
+    /* We emit frame/scan headers now */
+    if (master->scan_number == 0)
+      (*cinfo->marker->write_frame_header) (cinfo);
+    (*cinfo->marker->write_scan_header) (cinfo);
+    master->pub.call_pass_startup = FALSE;
+    break;
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+  }
+
+  master->pub.is_last_pass = (master->pass_number == master->total_passes - 1);
+
+  /* Set up progress monitor's pass info if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->completed_passes = master->pass_number;
+    cinfo->progress->total_passes = master->total_passes;
+  }
+}
+
+
+/*
+ * Special start-of-pass hook.
+ * This is called by jpeg_write_scanlines if call_pass_startup is TRUE.
+ * In single-pass processing, we need this hook because we don't want to
+ * write frame/scan headers during jpeg_start_compress; we want to let the
+ * application write COM markers etc. between jpeg_start_compress and the
+ * jpeg_write_scanlines loop.
+ * In multi-pass processing, this routine is not used.
+ */
+
+METHODDEF(void)
+pass_startup(j_compress_ptr cinfo)
+{
+  cinfo->master->call_pass_startup = FALSE; /* reset flag so call only once */
+
+  (*cinfo->marker->write_frame_header) (cinfo);
+  (*cinfo->marker->write_scan_header) (cinfo);
+}
+
+
+/*
+ * Finish up at end of pass.
+ */
+
+METHODDEF(void)
+finish_pass_master(j_compress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+
+  /* The entropy coder always needs an end-of-pass call,
+   * either to analyze statistics or to flush its output buffer.
+   */
+  (*cinfo->entropy->finish_pass) (cinfo);
+
+  /* Update state for next pass */
+  switch (master->pass_type) {
+  case main_pass:
+    /* next pass is either output of scan 0 (after optimization)
+     * or output of scan 1 (if no optimization).
+     */
+    master->pass_type = output_pass;
+    if (!cinfo->optimize_coding)
+      master->scan_number++;
+    break;
+  case huff_opt_pass:
+    /* next pass is always output of current scan */
+    master->pass_type = output_pass;
+    break;
+  case output_pass:
+    /* next pass is either optimization or output of next scan */
+    if (cinfo->optimize_coding)
+      master->pass_type = huff_opt_pass;
+    master->scan_number++;
+    break;
+  }
+
+  master->pass_number++;
+}
+
+
+/*
+ * Initialize master compression control.
+ */
+
+GLOBAL(void)
+jinit_c_master_control(j_compress_ptr cinfo, boolean transcode_only)
+{
+  my_master_ptr master;
+
+  master = (my_master_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_comp_master));
+  cinfo->master = (struct jpeg_comp_master *)master;
+  master->pub.prepare_for_pass = prepare_for_pass;
+  master->pub.pass_startup = pass_startup;
+  master->pub.finish_pass = finish_pass_master;
+  master->pub.is_last_pass = FALSE;
+
+  /* Validate parameters, determine derived values */
+  initial_setup(cinfo, transcode_only);
+
+  if (cinfo->scan_info != NULL) {
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    validate_script(cinfo);
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    cinfo->progressive_mode = FALSE;
+    cinfo->num_scans = 1;
+  }
+
+  if (cinfo->progressive_mode && !cinfo->arith_code)  /*  TEMPORARY HACK ??? */
+    cinfo->optimize_coding = TRUE; /* assume default tables no good for progressive mode */
+
+  /* Initialize my private state */
+  if (transcode_only) {
+    /* no main pass in transcoding */
+    if (cinfo->optimize_coding)
+      master->pass_type = huff_opt_pass;
+    else
+      master->pass_type = output_pass;
+  } else {
+    /* for normal compression, first pass is always this type: */
+    master->pass_type = main_pass;
+  }
+  master->scan_number = 0;
+  master->pass_number = 0;
+  if (cinfo->optimize_coding)
+    master->total_passes = cinfo->num_scans * 2;
+  else
+    master->total_passes = cinfo->num_scans;
+
+  master->jpeg_version = PACKAGE_NAME " version " VERSION " (build " BUILD ")";
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcomapi.c Added

@@ -0,0 +1,109 @@
+/*
+ * jcomapi.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains application interface routines that are used for both
+ * compression and decompression.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Abort processing of a JPEG compression or decompression operation,
+ * but don't destroy the object itself.
+ *
+ * For this, we merely clean up all the nonpermanent memory pools.
+ * Note that temp files (virtual arrays) are not allowed to belong to
+ * the permanent pool, so we will be able to close all temp files here.
+ * Closing a data source or destination, if necessary, is the application's
+ * responsibility.
+ */
+
+GLOBAL(void)
+jpeg_abort(j_common_ptr cinfo)
+{
+  int pool;
+
+  /* Do nothing if called on a not-initialized or destroyed JPEG object. */
+  if (cinfo->mem == NULL)
+    return;
+
+  /* Releasing pools in reverse order might help avoid fragmentation
+   * with some (brain-damaged) malloc libraries.
+   */
+  for (pool = JPOOL_NUMPOOLS - 1; pool > JPOOL_PERMANENT; pool--) {
+    (*cinfo->mem->free_pool) (cinfo, pool);
+  }
+
+  /* Reset overall state for possible reuse of object */
+  if (cinfo->is_decompressor) {
+    cinfo->global_state = DSTATE_START;
+    /* Try to keep application from accessing now-deleted marker list.
+     * A bit kludgy to do it here, but this is the most central place.
+     */
+    ((j_decompress_ptr)cinfo)->marker_list = NULL;
+  } else {
+    cinfo->global_state = CSTATE_START;
+  }
+}
+
+
+/*
+ * Destruction of a JPEG object.
+ *
+ * Everything gets deallocated except the master jpeg_compress_struct itself
+ * and the error manager struct.  Both of these are supplied by the application
+ * and must be freed, if necessary, by the application.  (Often they are on
+ * the stack and so don't need to be freed anyway.)
+ * Closing a data source or destination, if necessary, is the application's
+ * responsibility.
+ */
+
+GLOBAL(void)
+jpeg_destroy(j_common_ptr cinfo)
+{
+  /* We need only tell the memory manager to release everything. */
+  /* NB: mem pointer is NULL if memory mgr failed to initialize. */
+  if (cinfo->mem != NULL)
+    (*cinfo->mem->self_destruct) (cinfo);
+  cinfo->mem = NULL;            /* be safe if jpeg_destroy is called twice */
+  cinfo->global_state = 0;      /* mark it destroyed */
+}
+
+
+/*
+ * Convenience routines for allocating quantization and Huffman tables.
+ * (Would jutils.c be a more reasonable place to put these?)
+ */
+
+GLOBAL(JQUANT_TBL *)
+jpeg_alloc_quant_table(j_common_ptr cinfo)
+{
+  JQUANT_TBL *tbl;
+
+  tbl = (JQUANT_TBL *)
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, sizeof(JQUANT_TBL));
+  tbl->sent_table = FALSE;      /* make sure this is false in any new table */
+  return tbl;
+}
+
+
+GLOBAL(JHUFF_TBL *)
+jpeg_alloc_huff_table(j_common_ptr cinfo)
+{
+  JHUFF_TBL *tbl;
+
+  tbl = (JHUFF_TBL *)
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, sizeof(JHUFF_TBL));
+  tbl->sent_table = FALSE;      /* make sure this is false in any new table */
+  return tbl;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jconfig.h.in Added

@@ -0,0 +1,68 @@
+/* Version ID for the JPEG library.
+ * Might be useful for tests like "#if JPEG_LIB_VERSION >= 60".
+ */
+#define JPEG_LIB_VERSION  @JPEG_LIB_VERSION@
+
+/* libjpeg-turbo version */
+#define LIBJPEG_TURBO_VERSION  @VERSION@
+
+/* libjpeg-turbo version in integer form */
+#define LIBJPEG_TURBO_VERSION_NUMBER  @LIBJPEG_TURBO_VERSION_NUMBER@
+
+/* Support arithmetic encoding */
+#cmakedefine C_ARITH_CODING_SUPPORTED 1
+
+/* Support arithmetic decoding */
+#cmakedefine D_ARITH_CODING_SUPPORTED 1
+
+/* Support in-memory source/destination managers */
+#cmakedefine MEM_SRCDST_SUPPORTED 1
+
+/* Use accelerated SIMD routines. */
+#cmakedefine WITH_SIMD 1
+
+/*
+ * Define BITS_IN_JSAMPLE as either
+ *   8   for 8-bit sample values (the usual setting)
+ *   12  for 12-bit sample values
+ * Only 8 and 12 are legal data precisions for lossy JPEG according to the
+ * JPEG standard, and the IJG code does not support anything else!
+ * We do not support run-time selection of data precision, sorry.
+ */
+
+#define BITS_IN_JSAMPLE  @BITS_IN_JSAMPLE@      /* use 8 or 12 */
+
+/* Define to 1 if you have the <locale.h> header file. */
+#cmakedefine HAVE_LOCALE_H 1
+
+/* Define to 1 if you have the <stddef.h> header file. */
+#cmakedefine HAVE_STDDEF_H 1
+
+/* Define to 1 if you have the <stdlib.h> header file. */
+#cmakedefine HAVE_STDLIB_H 1
+
+/* Define if you need to include <sys/types.h> to get size_t. */
+#cmakedefine NEED_SYS_TYPES_H 1
+
+/* Define if you have BSD-like bzero and bcopy in <strings.h> rather than
+   memset/memcpy in <string.h>. */
+#cmakedefine NEED_BSD_STRINGS 1
+
+/* Define to 1 if the system has the type `unsigned char'. */
+#cmakedefine HAVE_UNSIGNED_CHAR 1
+
+/* Define to 1 if the system has the type `unsigned short'. */
+#cmakedefine HAVE_UNSIGNED_SHORT 1
+
+/* Compiler does not support pointers to undefined structures. */
+#cmakedefine INCOMPLETE_TYPES_BROKEN 1
+
+/* Define if your (broken) compiler shifts signed values as if they were
+   unsigned. */
+#cmakedefine RIGHT_SHIFT_IS_UNSIGNED 1
+
+/* Define to empty if `const' does not conform to ANSI C. */
+/* #undef const */
+
+/* Define to `unsigned int' if <sys/types.h> does not define. */
+/* #undef size_t */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jconfig.txt Added

@@ -0,0 +1,136 @@
+/*
+ * jconfig.txt
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file documents the configuration options that are required to
+ * customize the JPEG software for a particular system.
+ *
+ * The actual configuration options for a particular installation are stored
+ * in jconfig.h.  On many machines, jconfig.h can be generated automatically
+ * or copied from one of the "canned" jconfig files that we supply.  But if
+ * you need to generate a jconfig.h file by hand, this file tells you how.
+ *
+ * DO NOT EDIT THIS FILE --- IT WON'T ACCOMPLISH ANYTHING.
+ * EDIT A COPY NAMED JCONFIG.H.
+ */
+
+
+/*
+ * These symbols indicate the properties of your machine or compiler.
+ * #define the symbol if yes, #undef it if no.
+ */
+
+/* Does your compiler support the declaration "unsigned char" ?
+ * How about "unsigned short" ?
+ */
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+
+/* Define "void" as "char" if your compiler doesn't know about type void.
+ * NOTE: be sure to define void such that "void *" represents the most general
+ * pointer type, e.g., that returned by malloc().
+ */
+/* #define void char */
+
+/* Define "const" as empty if your compiler doesn't know the "const" keyword.
+ */
+/* #define const */
+
+/* Define this if your system has an ANSI-conforming <stddef.h> file.
+ */
+#define HAVE_STDDEF_H
+
+/* Define this if your system has an ANSI-conforming <stdlib.h> file.
+ */
+#define HAVE_STDLIB_H
+
+/* Define this if your system does not have an ANSI/SysV <string.h>,
+ * but does have a BSD-style <strings.h>.
+ */
+#undef NEED_BSD_STRINGS
+
+/* Define this if your system does not provide typedef size_t in any of the
+ * ANSI-standard places (stddef.h, stdlib.h, or stdio.h), but places it in
+ * <sys/types.h> instead.
+ */
+#undef NEED_SYS_TYPES_H
+
+/* Although a real ANSI C compiler can deal perfectly well with pointers to
+ * unspecified structures (see "incomplete types" in the spec), a few pre-ANSI
+ * and pseudo-ANSI compilers get confused.  To keep one of these bozos happy,
+ * define INCOMPLETE_TYPES_BROKEN.  This is not recommended unless you
+ * actually get "missing structure definition" warnings or errors while
+ * compiling the JPEG code.
+ */
+#undef INCOMPLETE_TYPES_BROKEN
+
+/* Define "boolean" as unsigned char, not int, on Windows systems.
+ */
+#ifdef _WIN32
+#ifndef __RPCNDR_H__            /* don't conflict if rpcndr.h already read */
+typedef unsigned char boolean;
+#endif
+#define HAVE_BOOLEAN            /* prevent jmorecfg.h from redefining it */
+#endif
+
+
+/*
+ * The following options affect code selection within the JPEG library,
+ * but they don't need to be visible to applications using the library.
+ * To minimize application namespace pollution, the symbols won't be
+ * defined unless JPEG_INTERNALS has been defined.
+ */
+
+#ifdef JPEG_INTERNALS
+
+/* Define this if your compiler implements ">>" on signed values as a logical
+ * (unsigned) shift; leave it undefined if ">>" is a signed (arithmetic) shift,
+ * which is the normal and rational definition.
+ */
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+
+#endif /* JPEG_INTERNALS */
+
+
+/*
+ * The remaining options do not affect the JPEG library proper,
+ * but only the sample applications cjpeg/djpeg (see cjpeg.c, djpeg.c).
+ * Other applications can ignore these.
+ */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+/* These defines indicate which image (non-JPEG) file formats are allowed. */
+
+#define BMP_SUPPORTED           /* BMP image file format */
+#define GIF_SUPPORTED           /* GIF image file format */
+#define PPM_SUPPORTED           /* PBMPLUS PPM/PGM image file format */
+#define TARGA_SUPPORTED         /* Targa image file format */
+
+/* Define this if you want to name both input and output files on the command
+ * line, rather than using stdout and optionally stdin.  You MUST do this if
+ * your system can't cope with binary I/O to stdin/stdout.  See comments at
+ * head of cjpeg.c or djpeg.c.
+ */
+#undef TWO_FILE_COMMANDLINE
+
+/* By default, we open image files with fopen(..., "rb") or fopen(..., "wb").
+ * This is necessary on systems that distinguish text files from binary files,
+ * and is harmless on most systems that don't.  If you have one of the rare
+ * systems that complains about the "b" spec, define this symbol.
+ */
+#undef DONT_USE_B_MODE
+
+/* Define this if you want percent-done progress reports from cjpeg/djpeg.
+ */
+#undef PROGRESS_REPORT
+
+
+#endif /* JPEG_CJPEG_DJPEG */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jconfigint.h.in Added

@@ -0,0 +1,44 @@
+/* libjpeg-turbo build number */
+#define BUILD  "@BUILD@"
+
+/* Compiler's inline keyword */
+#undef inline
+
+/* How to obtain function inlining. */
+#define INLINE  @INLINE@
+
+/* How to obtain thread-local storage */
+#define THREAD_LOCAL  @THREAD_LOCAL@
+
+/* Define to the full name of this package. */
+#define PACKAGE_NAME  "@CMAKE_PROJECT_NAME@"
+
+/* Version number of package */
+#define VERSION  "@VERSION@"
+
+/* The size of `size_t', as computed by sizeof. */
+#define SIZEOF_SIZE_T  @SIZE_T@
+
+/* Define if your compiler has __builtin_ctzl() and sizeof(unsigned long) == sizeof(size_t). */
+#cmakedefine HAVE_BUILTIN_CTZL
+
+/* Define to 1 if you have the <intrin.h> header file. */
+#cmakedefine HAVE_INTRIN_H
+
+#if defined(_MSC_VER) && defined(HAVE_INTRIN_H)
+#if (SIZEOF_SIZE_T == 8)
+#define HAVE_BITSCANFORWARD64
+#elif (SIZEOF_SIZE_T == 4)
+#define HAVE_BITSCANFORWARD
+#endif
+#endif
+
+#if defined(__has_attribute)
+#if __has_attribute(fallthrough)
+#define FALLTHROUGH  __attribute__((fallthrough));
+#else
+#define FALLTHROUGH
+#endif
+#else
+#define FALLTHROUGH
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcparam.c Added

@@ -0,0 +1,541 @@
+/*
+ * jcparam.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * Modified 2003-2008 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2018, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains optional default-setting code for the JPEG compressor.
+ * Applications do not have to use this file, but those that don't use it
+ * must know a lot more about the innards of the JPEG code.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jstdhuff.c"
+
+
+/*
+ * Quantization table setup routines
+ */
+
+GLOBAL(void)
+jpeg_add_quant_table(j_compress_ptr cinfo, int which_tbl,
+                     const unsigned int *basic_table, int scale_factor,
+                     boolean force_baseline)
+/* Define a quantization table equal to the basic_table times
+ * a scale factor (given as a percentage).
+ * If force_baseline is TRUE, the computed quantization table entries
+ * are limited to 1..255 for JPEG baseline compatibility.
+ */
+{
+  JQUANT_TBL **qtblptr;
+  int i;
+  long temp;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (which_tbl < 0 || which_tbl >= NUM_QUANT_TBLS)
+    ERREXIT1(cinfo, JERR_DQT_INDEX, which_tbl);
+
+  qtblptr = &cinfo->quant_tbl_ptrswhich_tbl;
+
+  if (*qtblptr == NULL)
+    *qtblptr = jpeg_alloc_quant_table((j_common_ptr)cinfo);
+
+  for (i = 0; i < DCTSIZE2; i++) {
+    temp = ((long)basic_tablei * scale_factor + 50L) / 100L;
+    /* limit the values to the valid range */
+    if (temp <= 0L) temp = 1L;
+    if (temp > 32767L) temp = 32767L; /* max quantizer needed for 12 bits */
+    if (force_baseline && temp > 255L)
+      temp = 255L;              /* limit to baseline range if requested */
+    (*qtblptr)->quantvali = (UINT16)temp;
+  }
+
+  /* Initialize sent_table FALSE so table will be written to JPEG file. */
+  (*qtblptr)->sent_table = FALSE;
+}
+
+
+/* These are the sample quantization tables given in Annex K (Clause K.1) of
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+ * The spec says that the values given produce "good" quality, and
+ * when divided by 2, "very good" quality.
+ */
+static const unsigned int std_luminance_quant_tblDCTSIZE2 = {
+  16,  11,  10,  16,  24,  40,  51,  61,
+  12,  12,  14,  19,  26,  58,  60,  55,
+  14,  13,  16,  24,  40,  57,  69,  56,
+  14,  17,  22,  29,  51,  87,  80,  62,
+  18,  22,  37,  56,  68, 109, 103,  77,
+  24,  35,  55,  64,  81, 104, 113,  92,
+  49,  64,  78,  87, 103, 121, 120, 101,
+  72,  92,  95,  98, 112, 100, 103,  99
+};
+static const unsigned int std_chrominance_quant_tblDCTSIZE2 = {
+  17,  18,  24,  47,  99,  99,  99,  99,
+  18,  21,  26,  66,  99,  99,  99,  99,
+  24,  26,  56,  99,  99,  99,  99,  99,
+  47,  66,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99
+};
+
+
+#if JPEG_LIB_VERSION >= 70
+GLOBAL(void)
+jpeg_default_qtables(j_compress_ptr cinfo, boolean force_baseline)
+/* Set or change the 'quality' (quantization) setting, using default tables
+ * and straight percentage-scaling quality scales.
+ * This entry point allows different scalings for luminance and chrominance.
+ */
+{
+  /* Set up two quantization tables using the specified scaling */
+  jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
+                       cinfo->q_scale_factor0, force_baseline);
+  jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
+                       cinfo->q_scale_factor1, force_baseline);
+}
+#endif
+
+
+GLOBAL(void)
+jpeg_set_linear_quality(j_compress_ptr cinfo, int scale_factor,
+                        boolean force_baseline)
+/* Set or change the 'quality' (quantization) setting, using default tables
+ * and a straight percentage-scaling quality scale.  In most cases it's better
+ * to use jpeg_set_quality (below); this entry point is provided for
+ * applications that insist on a linear percentage scaling.
+ */
+{
+  /* Set up two quantization tables using the specified scaling */
+  jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
+                       scale_factor, force_baseline);
+  jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
+                       scale_factor, force_baseline);
+}
+
+
+GLOBAL(int)
+jpeg_quality_scaling(int quality)
+/* Convert a user-specified quality rating to a percentage scaling factor
+ * for an underlying quantization table, using our recommended scaling curve.
+ * The input 'quality' factor should be 0 (terrible) to 100 (very good).
+ */
+{
+  /* Safety limit on quality factor.  Convert 0 to 1 to avoid zero divide. */
+  if (quality <= 0) quality = 1;
+  if (quality > 100) quality = 100;
+
+  /* The basic table is used as-is (scaling 100) for a quality of 50.
+   * Qualities 50..100 are converted to scaling percentage 200 - 2*Q;
+   * note that at Q=100 the scaling is 0, which will cause jpeg_add_quant_table
+   * to make all the table entries 1 (hence, minimum quantization loss).
+   * Qualities 1..50 are converted to scaling percentage 5000/Q.
+   */
+  if (quality < 50)
+    quality = 5000 / quality;
+  else
+    quality = 200 - quality * 2;
+
+  return quality;
+}
+
+
+GLOBAL(void)
+jpeg_set_quality(j_compress_ptr cinfo, int quality, boolean force_baseline)
+/* Set or change the 'quality' (quantization) setting, using default tables.
+ * This is the standard quality-adjusting entry point for typical user
+ * interfaces; only those who want detailed control over quantization tables
+ * would use the preceding three routines directly.
+ */
+{
+  /* Convert user 0-100 rating to percentage scaling */
+  quality = jpeg_quality_scaling(quality);
+
+  /* Set up standard quality tables */
+  jpeg_set_linear_quality(cinfo, quality, force_baseline);
+}
+
+
+/*
+ * Default parameter setup for compression.
+ *
+ * Applications that don't choose to use this routine must do their
+ * own setup of all these parameters.  Alternately, you can call this
+ * to establish defaults and then alter parameters selectively.  This
+ * is the recommended approach since, if we add any new parameters,
+ * your code will still work (they'll be set to reasonable defaults).
+ */
+
+GLOBAL(void)
+jpeg_set_defaults(j_compress_ptr cinfo)
+{
+  int i;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Allocate comp_info array large enough for maximum component count.
+   * Array is made permanent in case application wants to compress
+   * multiple images at same param settings.
+   */
+  if (cinfo->comp_info == NULL)
+    cinfo->comp_info = (jpeg_component_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                  MAX_COMPONENTS * sizeof(jpeg_component_info));
+
+  /* Initialize everything not dependent on the color space */
+
+#if JPEG_LIB_VERSION >= 70
+  cinfo->scale_num = 1;         /* 1:1 scaling */
+  cinfo->scale_denom = 1;
+#endif
+  cinfo->data_precision = BITS_IN_JSAMPLE;
+  /* Set up two quantization tables using default quality of 75 */
+  jpeg_set_quality(cinfo, 75, TRUE);
+  /* Set up two Huffman tables */
+  std_huff_tables((j_common_ptr)cinfo);
+
+  /* Initialize default arithmetic coding conditioning */
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    cinfo->arith_dc_Li = 0;
+    cinfo->arith_dc_Ui = 1;
+    cinfo->arith_ac_Ki = 5;
+  }
+
+  /* Default is no multiple-scan output */
+  cinfo->scan_info = NULL;
+  cinfo->num_scans = 0;
+
+  /* Expect normal source image, not raw downsampled data */
+  cinfo->raw_data_in = FALSE;
+
+  /* Use Huffman coding, not arithmetic coding, by default */
+  cinfo->arith_code = FALSE;
+
+  /* By default, don't do extra passes to optimize entropy coding */
+  cinfo->optimize_coding = FALSE;
+  /* The standard Huffman tables are only valid for 8-bit data precision.
+   * If the precision is higher, force optimization on so that usable
+   * tables will be computed.  This test can be removed if default tables
+   * are supplied that are valid for the desired precision.
+   */
+  if (cinfo->data_precision > 8)
+    cinfo->optimize_coding = TRUE;
+
+  /* By default, use the simpler non-cosited sampling alignment */
+  cinfo->CCIR601_sampling = FALSE;
+
+#if JPEG_LIB_VERSION >= 70
+  /* By default, apply fancy downsampling */
+  cinfo->do_fancy_downsampling = TRUE;
+#endif
+
+  /* No input smoothing */
+  cinfo->smoothing_factor = 0;
+
+  /* DCT algorithm preference */
+  cinfo->dct_method = JDCT_DEFAULT;
+
+  /* No restart markers */
+  cinfo->restart_interval = 0;
+  cinfo->restart_in_rows = 0;
+
+  /* Fill in default JFIF marker parameters.  Note that whether the marker
+   * will actually be written is determined by jpeg_set_colorspace.
+   *
+   * By default, the library emits JFIF version code 1.01.
+   * An application that wants to emit JFIF 1.02 extension markers should set
+   * JFIF_minor_version to 2.  We could probably get away with just defaulting
+   * to 1.02, but there may still be some decoders in use that will complain
+   * about that; saying 1.01 should minimize compatibility problems.
+   */
+  cinfo->JFIF_major_version = 1; /* Default JFIF version = 1.01 */
+  cinfo->JFIF_minor_version = 1;
+  cinfo->density_unit = 0;      /* Pixel size is unknown by default */
+  cinfo->X_density = 1;         /* Pixel aspect ratio is square by default */
+  cinfo->Y_density = 1;
+
+  /* Choose JPEG colorspace based on input space, set defaults accordingly */
+
+  jpeg_default_colorspace(cinfo);
+}
+
+
+/*
+ * Select an appropriate JPEG colorspace for in_color_space.
+ */
+
+GLOBAL(void)
+jpeg_default_colorspace(j_compress_ptr cinfo)
+{
+  switch (cinfo->in_color_space) {
+  case JCS_GRAYSCALE:
+    jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+    break;
+  case JCS_RGB:
+  case JCS_EXT_RGB:
+  case JCS_EXT_RGBX:
+  case JCS_EXT_BGR:
+  case JCS_EXT_BGRX:
+  case JCS_EXT_XBGR:
+  case JCS_EXT_XRGB:
+  case JCS_EXT_RGBA:
+  case JCS_EXT_BGRA:
+  case JCS_EXT_ABGR:
+  case JCS_EXT_ARGB:
+    jpeg_set_colorspace(cinfo, JCS_YCbCr);
+    break;
+  case JCS_YCbCr:
+    jpeg_set_colorspace(cinfo, JCS_YCbCr);
+    break;
+  case JCS_CMYK:
+    jpeg_set_colorspace(cinfo, JCS_CMYK); /* By default, no translation */
+    break;
+  case JCS_YCCK:
+    jpeg_set_colorspace(cinfo, JCS_YCCK);
+    break;
+  case JCS_UNKNOWN:
+    jpeg_set_colorspace(cinfo, JCS_UNKNOWN);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+  }
+}
+
+
+/*
+ * Set the JPEG colorspace, and choose colorspace-dependent default values.
+ */
+
+GLOBAL(void)
+jpeg_set_colorspace(j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
+{
+  jpeg_component_info *compptr;
+  int ci;
+
+#define SET_COMP(index, id, hsamp, vsamp, quant, dctbl, actbl) \
+  (compptr = &cinfo->comp_infoindex, \
+   compptr->component_id = (id), \
+   compptr->h_samp_factor = (hsamp), \
+   compptr->v_samp_factor = (vsamp), \
+   compptr->quant_tbl_no = (quant), \
+   compptr->dc_tbl_no = (dctbl), \
+   compptr->ac_tbl_no = (actbl) )
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* For all colorspaces, we use Q and Huff tables 0 for luminance components,
+   * tables 1 for chrominance components.
+   */
+
+  cinfo->jpeg_color_space = colorspace;
+
+  cinfo->write_JFIF_header = FALSE; /* No marker for non-JFIF colorspaces */
+  cinfo->write_Adobe_marker = FALSE; /* write no Adobe marker by default */
+
+  switch (colorspace) {
+  case JCS_GRAYSCALE:
+    cinfo->write_JFIF_header = TRUE; /* Write a JFIF marker */
+    cinfo->num_components = 1;
+    /* JFIF specifies component ID 1 */
+    SET_COMP(0, 1, 1, 1, 0, 0, 0);
+    break;
+  case JCS_RGB:
+    cinfo->write_Adobe_marker = TRUE; /* write Adobe marker to flag RGB */
+    cinfo->num_components = 3;
+    SET_COMP(0, 0x52 /* 'R' */, 1, 1, 0, 0, 0);
+    SET_COMP(1, 0x47 /* 'G' */, 1, 1, 0, 0, 0);
+    SET_COMP(2, 0x42 /* 'B' */, 1, 1, 0, 0, 0);
+    break;
+  case JCS_YCbCr:
+    cinfo->write_JFIF_header = TRUE; /* Write a JFIF marker */
+    cinfo->num_components = 3;
+    /* JFIF specifies component IDs 1,2,3 */
+    /* We default to 2x2 subsamples of chrominance */
+    SET_COMP(0, 1, 2, 2, 0, 0, 0);
+    SET_COMP(1, 2, 1, 1, 1, 1, 1);
+    SET_COMP(2, 3, 1, 1, 1, 1, 1);
+    break;
+  case JCS_CMYK:
+    cinfo->write_Adobe_marker = TRUE; /* write Adobe marker to flag CMYK */
+    cinfo->num_components = 4;
+    SET_COMP(0, 0x43 /* 'C' */, 1, 1, 0, 0, 0);
+    SET_COMP(1, 0x4D /* 'M' */, 1, 1, 0, 0, 0);
+    SET_COMP(2, 0x59 /* 'Y' */, 1, 1, 0, 0, 0);
+    SET_COMP(3, 0x4B /* 'K' */, 1, 1, 0, 0, 0);
+    break;
+  case JCS_YCCK:
+    cinfo->write_Adobe_marker = TRUE; /* write Adobe marker to flag YCCK */
+    cinfo->num_components = 4;
+    SET_COMP(0, 1, 2, 2, 0, 0, 0);
+    SET_COMP(1, 2, 1, 1, 1, 1, 1);
+    SET_COMP(2, 3, 1, 1, 1, 1, 1);
+    SET_COMP(3, 4, 2, 2, 0, 0, 0);
+    break;
+  case JCS_UNKNOWN:
+    cinfo->num_components = cinfo->input_components;
+    if (cinfo->num_components < 1 || cinfo->num_components > MAX_COMPONENTS)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+               MAX_COMPONENTS);
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      SET_COMP(ci, ci, 1, 1, 0, 0, 0);
+    }
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+  }
+}
+
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+
+LOCAL(jpeg_scan_info *)
+fill_a_scan(jpeg_scan_info *scanptr, int ci, int Ss, int Se, int Ah, int Al)
+/* Support routine: generate one scan for specified component */
+{
+  scanptr->comps_in_scan = 1;
+  scanptr->component_index0 = ci;
+  scanptr->Ss = Ss;
+  scanptr->Se = Se;
+  scanptr->Ah = Ah;
+  scanptr->Al = Al;
+  scanptr++;
+  return scanptr;
+}
+
+LOCAL(jpeg_scan_info *)
+fill_scans(jpeg_scan_info *scanptr, int ncomps, int Ss, int Se, int Ah, int Al)
+/* Support routine: generate one scan for each component */
+{
+  int ci;
+
+  for (ci = 0; ci < ncomps; ci++) {
+    scanptr->comps_in_scan = 1;
+    scanptr->component_index0 = ci;
+    scanptr->Ss = Ss;
+    scanptr->Se = Se;
+    scanptr->Ah = Ah;
+    scanptr->Al = Al;
+    scanptr++;
+  }
+  return scanptr;
+}
+
+LOCAL(jpeg_scan_info *)
+fill_dc_scans(jpeg_scan_info *scanptr, int ncomps, int Ah, int Al)
+/* Support routine: generate interleaved DC scan if possible, else N scans */
+{
+  int ci;
+
+  if (ncomps <= MAX_COMPS_IN_SCAN) {
+    /* Single interleaved DC scan */
+    scanptr->comps_in_scan = ncomps;
+    for (ci = 0; ci < ncomps; ci++)
+      scanptr->component_indexci = ci;
+    scanptr->Ss = scanptr->Se = 0;
+    scanptr->Ah = Ah;
+    scanptr->Al = Al;
+    scanptr++;
+  } else {
+    /* Noninterleaved DC scan for each component */
+    scanptr = fill_scans(scanptr, ncomps, 0, 0, Ah, Al);
+  }
+  return scanptr;
+}
+
+
+/*
+ * Create a recommended progressive-JPEG script.
+ * cinfo->num_components and cinfo->jpeg_color_space must be correct.
+ */
+
+GLOBAL(void)
+jpeg_simple_progression(j_compress_ptr cinfo)
+{
+  int ncomps = cinfo->num_components;
+  int nscans;
+  jpeg_scan_info *scanptr;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Figure space needed for script.  Calculation must match code below! */
+  if (ncomps == 3 && cinfo->jpeg_color_space == JCS_YCbCr) {
+    /* Custom script for YCbCr color images. */
+    nscans = 10;
+  } else {
+    /* All-purpose script for other color spaces. */
+    if (ncomps > MAX_COMPS_IN_SCAN)
+      nscans = 6 * ncomps;      /* 2 DC + 4 AC scans per component */
+    else
+      nscans = 2 + 4 * ncomps;  /* 2 DC scans; 4 AC scans per component */
+  }
+
+  /* Allocate space for script.
+   * We need to put it in the permanent pool in case the application performs
+   * multiple compressions without changing the settings.  To avoid a memory
+   * leak if jpeg_simple_progression is called repeatedly for the same JPEG
+   * object, we try to re-use previously allocated space, and we allocate
+   * enough space to handle YCbCr even if initially asked for grayscale.
+   */
+  if (cinfo->script_space == NULL || cinfo->script_space_size < nscans) {
+    cinfo->script_space_size = MAX(nscans, 10);
+    cinfo->script_space = (jpeg_scan_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                        cinfo->script_space_size * sizeof(jpeg_scan_info));
+  }
+  scanptr = cinfo->script_space;
+  cinfo->scan_info = scanptr;
+  cinfo->num_scans = nscans;
+
+  if (ncomps == 3 && cinfo->jpeg_color_space == JCS_YCbCr) {
+    /* Custom script for YCbCr color images. */
+    /* Initial DC scan */
+    scanptr = fill_dc_scans(scanptr, ncomps, 0, 1);
+    /* Initial AC scan: get some luma data out in a hurry */
+    scanptr = fill_a_scan(scanptr, 0, 1, 5, 0, 2);
+    /* Chroma data is too small to be worth expending many scans on */
+    scanptr = fill_a_scan(scanptr, 2, 1, 63, 0, 1);
+    scanptr = fill_a_scan(scanptr, 1, 1, 63, 0, 1);
+    /* Complete spectral selection for luma AC */
+    scanptr = fill_a_scan(scanptr, 0, 6, 63, 0, 2);
+    /* Refine next bit of luma AC */
+    scanptr = fill_a_scan(scanptr, 0, 1, 63, 2, 1);
+    /* Finish DC successive approximation */
+    scanptr = fill_dc_scans(scanptr, ncomps, 1, 0);
+    /* Finish AC successive approximation */
+    scanptr = fill_a_scan(scanptr, 2, 1, 63, 1, 0);
+    scanptr = fill_a_scan(scanptr, 1, 1, 63, 1, 0);
+    /* Luma bottom bit comes last since it's usually largest scan */
+    scanptr = fill_a_scan(scanptr, 0, 1, 63, 1, 0);
+  } else {
+    /* All-purpose script for other color spaces. */
+    /* Successive approximation first pass */
+    scanptr = fill_dc_scans(scanptr, ncomps, 0, 1);
+    scanptr = fill_scans(scanptr, ncomps, 1, 5, 0, 2);
+    scanptr = fill_scans(scanptr, ncomps, 6, 63, 0, 2);
+    /* Successive approximation second pass */
+    scanptr = fill_scans(scanptr, ncomps, 1, 63, 2, 1);
+    /* Successive approximation final pass */
+    scanptr = fill_dc_scans(scanptr, ncomps, 1, 0);
+    scanptr = fill_scans(scanptr, ncomps, 1, 63, 1, 0);
+  }
+}
+
+#endif /* C_PROGRESSIVE_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcphuff.c Added

@@ -0,0 +1,1114 @@
+/*
+ * jcphuff.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, 2015, 2018, 2021, D. R. Commander.
+ * Copyright (C) 2016, 2018, Matthieu Darbois.
+ * Copyright (C) 2020, Arm Limited.
+ * Copyright (C) 2021, Alex Richardson.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains Huffman entropy encoding routines for progressive JPEG.
+ *
+ * We do not support output suspension in this module, since the library
+ * currently does not allow multiple-scan files to be written with output
+ * suspension.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jsimd.h"
+#include "jconfigint.h"
+#include <limits.h>
+
+#ifdef HAVE_INTRIN_H
+#include <intrin.h>
+#ifdef _MSC_VER
+#ifdef HAVE_BITSCANFORWARD64
+#pragma intrinsic(_BitScanForward64)
+#endif
+#ifdef HAVE_BITSCANFORWARD
+#pragma intrinsic(_BitScanForward)
+#endif
+#endif
+#endif
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+
+/*
+ * NOTE: If USE_CLZ_INTRINSIC is defined, then clz/bsr instructions will be
+ * used for bit counting rather than the lookup table.  This will reduce the
+ * memory footprint by 64k, which is important for some mobile applications
+ * that create many isolated instances of libjpeg-turbo (web browsers, for
+ * instance.)  This may improve performance on some mobile platforms as well.
+ * This feature is enabled by default only on Arm processors, because some x86
+ * chips have a slow implementation of bsr, and the use of clz/bsr cannot be
+ * shown to have a significant performance impact even on the x86 chips that
+ * have a fast implementation of it.  When building for Armv6, you can
+ * explicitly disable the use of clz/bsr by adding -mthumb to the compiler
+ * flags (this defines __thumb__).
+ */
+
+/* NOTE: Both GCC and Clang define __GNUC__ */
+#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
+    defined(_M_ARM) || defined(_M_ARM64)
+#if !defined(__thumb__) || defined(__thumb2__)
+#define USE_CLZ_INTRINSIC
+#endif
+#endif
+
+#ifdef USE_CLZ_INTRINSIC
+#if defined(_MSC_VER) && !defined(__clang__)
+#define JPEG_NBITS_NONZERO(x)  (32 - _CountLeadingZeros(x))
+#else
+#define JPEG_NBITS_NONZERO(x)  (32 - __builtin_clz(x))
+#endif
+#define JPEG_NBITS(x)          (x ? JPEG_NBITS_NONZERO(x) : 0)
+#else
+#include "jpeg_nbits_table.h"
+#define JPEG_NBITS(x)          (jpeg_nbits_tablex)
+#define JPEG_NBITS_NONZERO(x)  JPEG_NBITS(x)
+#endif
+
+
+/* Expanded entropy encoder object for progressive Huffman encoding. */
+
+typedef struct {
+  struct jpeg_entropy_encoder pub; /* public fields */
+
+  /* Pointer to routine to prepare data for encode_mcu_AC_first() */
+  void (*AC_first_prepare) (const JCOEF *block,
+                            const int *jpeg_natural_order_start, int Sl,
+                            int Al, JCOEF *values, size_t *zerobits);
+  /* Pointer to routine to prepare data for encode_mcu_AC_refine() */
+  int (*AC_refine_prepare) (const JCOEF *block,
+                            const int *jpeg_natural_order_start, int Sl,
+                            int Al, JCOEF *absvalues, size_t *bits);
+
+  /* Mode flag: TRUE for optimization, FALSE for actual data output */
+  boolean gather_statistics;
+
+  /* Bit-level coding status.
+   * next_output_byte/free_in_buffer are local copies of cinfo->dest fields.
+   */
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
+  size_t put_buffer;            /* current bit-accumulation buffer */
+  int put_bits;                 /* # of bits now in it */
+  j_compress_ptr cinfo;         /* link to cinfo (needed for dump_buffer) */
+
+  /* Coding status for DC components */
+  int last_dc_valMAX_COMPS_IN_SCAN; /* last DC coef for each component */
+
+  /* Coding status for AC components */
+  int ac_tbl_no;                /* the table number of the single component */
+  unsigned int EOBRUN;          /* run length of EOBs */
+  unsigned int BE;              /* # of buffered correction bits before MCU */
+  char *bit_buffer;             /* buffer for correction bits (1 per char) */
+  /* packing correction bits tightly would save some space but cost time... */
+
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+  int next_restart_num;         /* next restart number to write (0-7) */
+
+  /* Pointers to derived tables (these workspaces have image lifespan).
+   * Since any one scan codes only DC or only AC, we only need one set
+   * of tables, not one for DC and one for AC.
+   */
+  c_derived_tbl *derived_tblsNUM_HUFF_TBLS;
+
+  /* Statistics tables for optimization; again, one set is enough */
+  long *count_ptrsNUM_HUFF_TBLS;
+} phuff_entropy_encoder;
+
+typedef phuff_entropy_encoder *phuff_entropy_ptr;
+
+/* MAX_CORR_BITS is the number of bits the AC refinement correction-bit
+ * buffer can hold.  Larger sizes may slightly improve compression, but
+ * 1000 is already well into the realm of overkill.
+ * The minimum safe size is 64 bits.
+ */
+
+#define MAX_CORR_BITS  1000     /* Max # of correction bits I can buffer */
+
+/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than JLONG.
+ * We assume that int right shift is unsigned if JLONG right shift is,
+ * which should be safe.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define ISHIFT_TEMPS    int ishift_temp;
+#define IRIGHT_SHIFT(x, shft) \
+  ((ishift_temp = (x)) < 0 ? \
+   (ishift_temp >> (shft)) | ((~0) << (16 - (shft))) : \
+   (ishift_temp >> (shft)))
+#else
+#define ISHIFT_TEMPS
+#define IRIGHT_SHIFT(x, shft)   ((x) >> (shft))
+#endif
+
+#define PAD(v, p)  ((v + (p) - 1) & (~((p) - 1)))
+
+/* Forward declarations */
+METHODDEF(boolean) encode_mcu_DC_first(j_compress_ptr cinfo,
+                                       JBLOCKROW *MCU_data);
+METHODDEF(void) encode_mcu_AC_first_prepare
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *values, size_t *zerobits);
+METHODDEF(boolean) encode_mcu_AC_first(j_compress_ptr cinfo,
+                                       JBLOCKROW *MCU_data);
+METHODDEF(boolean) encode_mcu_DC_refine(j_compress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(int) encode_mcu_AC_refine_prepare
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *absvalues, size_t *bits);
+METHODDEF(boolean) encode_mcu_AC_refine(j_compress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(void) finish_pass_phuff(j_compress_ptr cinfo);
+METHODDEF(void) finish_pass_gather_phuff(j_compress_ptr cinfo);
+
+
+/* Count bit loop zeroes */
+INLINE
+METHODDEF(int)
+count_zeroes(size_t *x)
+{
+#if defined(HAVE_BUILTIN_CTZL)
+  int result;
+  result = __builtin_ctzl(*x);
+  *x >>= result;
+#elif defined(HAVE_BITSCANFORWARD64)
+  unsigned long result;
+  _BitScanForward64(&result, *x);
+  *x >>= result;
+#elif defined(HAVE_BITSCANFORWARD)
+  unsigned long result;
+  _BitScanForward(&result, *x);
+  *x >>= result;
+#else
+  int result = 0;
+  while ((*x & 1) == 0) {
+    ++result;
+    *x >>= 1;
+  }
+#endif
+  return (int)result;
+}
+
+
+/*
+ * Initialize for a Huffman-compressed scan using progressive JPEG.
+ */
+
+METHODDEF(void)
+start_pass_phuff(j_compress_ptr cinfo, boolean gather_statistics)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  boolean is_DC_band;
+  int ci, tbl;
+  jpeg_component_info *compptr;
+
+  entropy->cinfo = cinfo;
+  entropy->gather_statistics = gather_statistics;
+
+  is_DC_band = (cinfo->Ss == 0);
+
+  /* We assume jcmaster.c already validated the scan parameters. */
+
+  /* Select execution routines */
+  if (cinfo->Ah == 0) {
+    if (is_DC_band)
+      entropy->pub.encode_mcu = encode_mcu_DC_first;
+    else
+      entropy->pub.encode_mcu = encode_mcu_AC_first;
+    if (jsimd_can_encode_mcu_AC_first_prepare())
+      entropy->AC_first_prepare = jsimd_encode_mcu_AC_first_prepare;
+    else
+      entropy->AC_first_prepare = encode_mcu_AC_first_prepare;
+  } else {
+    if (is_DC_band)
+      entropy->pub.encode_mcu = encode_mcu_DC_refine;
+    else {
+      entropy->pub.encode_mcu = encode_mcu_AC_refine;
+      if (jsimd_can_encode_mcu_AC_refine_prepare())
+        entropy->AC_refine_prepare = jsimd_encode_mcu_AC_refine_prepare;
+      else
+        entropy->AC_refine_prepare = encode_mcu_AC_refine_prepare;
+      /* AC refinement needs a correction bit buffer */
+      if (entropy->bit_buffer == NULL)
+        entropy->bit_buffer = (char *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                      MAX_CORR_BITS * sizeof(char));
+    }
+  }
+  if (gather_statistics)
+    entropy->pub.finish_pass = finish_pass_gather_phuff;
+  else
+    entropy->pub.finish_pass = finish_pass_phuff;
+
+  /* Only DC coefficients may be interleaved, so cinfo->comps_in_scan = 1
+   * for AC coefficients.
+   */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    /* Initialize DC predictions to 0 */
+    entropy->last_dc_valci = 0;
+    /* Get table index */
+    if (is_DC_band) {
+      if (cinfo->Ah != 0)       /* DC refinement needs no table */
+        continue;
+      tbl = compptr->dc_tbl_no;
+    } else {
+      entropy->ac_tbl_no = tbl = compptr->ac_tbl_no;
+    }
+    if (gather_statistics) {
+      /* Check for invalid table index */
+      /* (make_c_derived_tbl does this in the other path) */
+      if (tbl < 0 || tbl >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tbl);
+      /* Allocate and zero the statistics tables */
+      /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
+      if (entropy->count_ptrstbl == NULL)
+        entropy->count_ptrstbl = (long *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                      257 * sizeof(long));
+      MEMZERO(entropy->count_ptrstbl, 257 * sizeof(long));
+    } else {
+      /* Compute derived values for Huffman table */
+      /* We may do this more than once for a table, but it's not expensive */
+      jpeg_make_c_derived_tbl(cinfo, is_DC_band, tbl,
+                              &entropy->derived_tblstbl);
+    }
+  }
+
+  /* Initialize AC stuff */
+  entropy->EOBRUN = 0;
+  entropy->BE = 0;
+
+  /* Initialize bit buffer to empty */
+  entropy->put_buffer = 0;
+  entropy->put_bits = 0;
+
+  /* Initialize restart stuff */
+  entropy->restarts_to_go = cinfo->restart_interval;
+  entropy->next_restart_num = 0;
+}
+
+
+/* Outputting bytes to the file.
+ * NB: these must be called only when actually outputting,
+ * that is, entropy->gather_statistics == FALSE.
+ */
+
+/* Emit a byte */
+#define emit_byte(entropy, val) { \
+  *(entropy)->next_output_byte++ = (JOCTET)(val); \
+  if (--(entropy)->free_in_buffer == 0) \
+    dump_buffer(entropy); \
+}
+
+
+LOCAL(void)
+dump_buffer(phuff_entropy_ptr entropy)
+/* Empty the output buffer; we do not support suspension in this module. */
+{
+  struct jpeg_destination_mgr *dest = entropy->cinfo->dest;
+
+  if (!(*dest->empty_output_buffer) (entropy->cinfo))
+    ERREXIT(entropy->cinfo, JERR_CANT_SUSPEND);
+  /* After a successful buffer dump, must reset buffer pointers */
+  entropy->next_output_byte = dest->next_output_byte;
+  entropy->free_in_buffer = dest->free_in_buffer;
+}
+
+
+/* Outputting bits to the file */
+
+/* Only the right 24 bits of put_buffer are used; the valid bits are
+ * left-justified in this part.  At most 16 bits can be passed to emit_bits
+ * in one call, and we never retain more than 7 bits in put_buffer
+ * between calls, so 24 bits are sufficient.
+ */
+
+LOCAL(void)
+emit_bits(phuff_entropy_ptr entropy, unsigned int code, int size)
+/* Emit some bits, unless we are in gather mode */
+{
+  /* This routine is heavily used, so it's worth coding tightly. */
+  register size_t put_buffer = (size_t)code;
+  register int put_bits = entropy->put_bits;
+
+  /* if size is 0, caller used an invalid Huffman table entry */
+  if (size == 0)
+    ERREXIT(entropy->cinfo, JERR_HUFF_MISSING_CODE);
+
+  if (entropy->gather_statistics)
+    return;                     /* do nothing if we're only getting stats */
+
+  put_buffer &= (((size_t)1) << size) - 1; /* mask off any extra bits in code */
+
+  put_bits += size;             /* new number of bits in buffer */
+
+  put_buffer <<= 24 - put_bits; /* align incoming bits */
+
+  put_buffer |= entropy->put_buffer; /* and merge with old buffer contents */
+
+  while (put_bits >= 8) {
+    int c = (int)((put_buffer >> 16) & 0xFF);
+
+    emit_byte(entropy, c);
+    if (c == 0xFF) {            /* need to stuff a zero byte? */
+      emit_byte(entropy, 0);
+    }
+    put_buffer <<= 8;
+    put_bits -= 8;
+  }
+
+  entropy->put_buffer = put_buffer; /* update variables */
+  entropy->put_bits = put_bits;
+}
+
+
+LOCAL(void)
+flush_bits(phuff_entropy_ptr entropy)
+{
+  emit_bits(entropy, 0x7F, 7); /* fill any partial byte with ones */
+  entropy->put_buffer = 0;     /* and reset bit-buffer to empty */
+  entropy->put_bits = 0;
+}
+
+
+/*
+ * Emit (or just count) a Huffman symbol.
+ */
+
+LOCAL(void)
+emit_symbol(phuff_entropy_ptr entropy, int tbl_no, int symbol)
+{
+  if (entropy->gather_statistics)
+    entropy->count_ptrstbl_nosymbol++;
+  else {
+    c_derived_tbl *tbl = entropy->derived_tblstbl_no;
+    emit_bits(entropy, tbl->ehufcosymbol, tbl->ehufsisymbol);
+  }
+}
+
+
+/*
+ * Emit bits from a correction bit buffer.
+ */
+
+LOCAL(void)
+emit_buffered_bits(phuff_entropy_ptr entropy, char *bufstart,
+                   unsigned int nbits)
+{
+  if (entropy->gather_statistics)
+    return;                     /* no real work */
+
+  while (nbits > 0) {
+    emit_bits(entropy, (unsigned int)(*bufstart), 1);
+    bufstart++;
+    nbits--;
+  }
+}
+
+
+/*
+ * Emit any pending EOBRUN symbol.
+ */
+
+LOCAL(void)
+emit_eobrun(phuff_entropy_ptr entropy)
+{
+  register int temp, nbits;
+
+  if (entropy->EOBRUN > 0) {    /* if there is any pending EOBRUN */
+    temp = entropy->EOBRUN;
+    nbits = JPEG_NBITS_NONZERO(temp) - 1;
+    /* safety check: shouldn't happen given limited correction-bit buffer */
+    if (nbits > 14)
+      ERREXIT(entropy->cinfo, JERR_HUFF_MISSING_CODE);
+
+    emit_symbol(entropy, entropy->ac_tbl_no, nbits << 4);
+    if (nbits)
+      emit_bits(entropy, entropy->EOBRUN, nbits);
+
+    entropy->EOBRUN = 0;
+
+    /* Emit any buffered correction bits */
+    emit_buffered_bits(entropy, entropy->bit_buffer, entropy->BE);
+    entropy->BE = 0;
+  }
+}
+
+
+/*
+ * Emit a restart marker & resynchronize predictions.
+ */
+
+LOCAL(void)
+emit_restart(phuff_entropy_ptr entropy, int restart_num)
+{
+  int ci;
+
+  emit_eobrun(entropy);
+
+  if (!entropy->gather_statistics) {
+    flush_bits(entropy);
+    emit_byte(entropy, 0xFF);
+    emit_byte(entropy, JPEG_RST0 + restart_num);
+  }
+
+  if (entropy->cinfo->Ss == 0) {
+    /* Re-initialize DC predictions to 0 */
+    for (ci = 0; ci < entropy->cinfo->comps_in_scan; ci++)
+      entropy->last_dc_valci = 0;
+  } else {
+    /* Re-initialize all AC-related fields to 0 */
+    entropy->EOBRUN = 0;
+    entropy->BE = 0;
+  }
+}
+
+
+/*
+ * MCU encoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_first(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  register int temp, temp2, temp3;
+  register int nbits;
+  int blkn, ci;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+  jpeg_component_info *compptr;
+  ISHIFT_TEMPS
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_datablkn;
+    ci = cinfo->MCU_membershipblkn;
+    compptr = cinfo->cur_comp_infoci;
+
+    /* Compute the DC value after the required point transform by Al.
+     * This is simply an arithmetic right shift.
+     */
+    temp2 = IRIGHT_SHIFT((int)((*block)0), Al);
+
+    /* DC differences are figured on the point-transformed values. */
+    temp = temp2 - entropy->last_dc_valci;
+    entropy->last_dc_valci = temp2;
+
+    /* Encode the DC coefficient difference per section G.1.2.1 */
+
+    /* This is a well-known technique for obtaining the absolute value without
+     * a branch.  It is derived from an assembly language technique presented
+     * in "How to Optimize for the Pentium Processors", Copyright (c) 1996,
+     * 1997 by Agner Fog.
+     */
+    temp3 = temp >> (CHAR_BIT * sizeof(int) - 1);
+    temp ^= temp3;
+    temp -= temp3;              /* temp is abs value of input */
+    /* For a negative input, want temp2 = bitwise complement of abs(input) */
+    temp2 = temp ^ temp3;
+
+    /* Find the number of bits needed for the magnitude of the coefficient */
+    nbits = JPEG_NBITS(temp);
+    /* Check for out-of-range coefficient values.
+     * Since we're encoding a difference, the range limit is twice as much.
+     */
+    if (nbits > MAX_COEF_BITS + 1)
+      ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+
+    /* Count/emit the Huffman-coded symbol for the number of bits */
+    emit_symbol(entropy, compptr->dc_tbl_no, nbits);
+
+    /* Emit that number of bits of the value, if positive, */
+    /* or the complement of its magnitude, if negative. */
+    if (nbits)                  /* emit_bits rejects calls with size 0 */
+      emit_bits(entropy, (unsigned int)temp2, nbits);
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Data preparation for encode_mcu_AC_first().
+ */
+
+#define COMPUTE_ABSVALUES_AC_FIRST(Sl) { \
+  for (k = 0; k < Sl; k++) { \
+    temp = blockjpeg_natural_order_startk; \
+    if (temp == 0) \
+      continue; \
+    /* We must apply the point transform by Al.  For AC coefficients this \
+     * is an integer division with rounding towards 0.  To do this portably \
+     * in C, we shift after obtaining the absolute value; so the code is \
+     * interwoven with finding the abs value (temp) and output bits (temp2). \
+     */ \
+    temp2 = temp >> (CHAR_BIT * sizeof(int) - 1); \
+    temp ^= temp2; \
+    temp -= temp2;              /* temp is abs value of input */ \
+    temp >>= Al;                /* apply the point transform */ \
+    /* Watch out for case that nonzero coef is zero after point transform */ \
+    if (temp == 0) \
+      continue; \
+    /* For a negative coef, want temp2 = bitwise complement of abs(coef) */ \
+    temp2 ^= temp; \
+    valuesk = temp; \
+    valuesk + DCTSIZE2 = temp2; \
+    zerobits |= ((size_t)1U) << k; \
+  } \
+}
+
+METHODDEF(void)
+encode_mcu_AC_first_prepare(const JCOEF *block,
+                            const int *jpeg_natural_order_start, int Sl,
+                            int Al, JCOEF *values, size_t *bits)
+{
+  register int k, temp, temp2;
+  size_t zerobits = 0U;
+  int Sl0 = Sl;
+
+#if SIZEOF_SIZE_T == 4
+  if (Sl0 > 32)
+    Sl0 = 32;
+#endif
+
+  COMPUTE_ABSVALUES_AC_FIRST(Sl0);
+
+  bits0 = zerobits;
+#if SIZEOF_SIZE_T == 4
+  zerobits = 0U;
+
+  if (Sl > 32) {
+    Sl -= 32;
+    jpeg_natural_order_start += 32;
+    values += 32;
+
+    COMPUTE_ABSVALUES_AC_FIRST(Sl);
+  }
+  bits1 = zerobits;
+#endif
+}
+
+/*
+ * MCU encoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+#define ENCODE_COEFS_AC_FIRST(label) { \
+  while (zerobits) { \
+    r = count_zeroes(&zerobits); \
+    cvalue += r; \
+label \
+    temp  = cvalue0; \
+    temp2 = cvalueDCTSIZE2; \
+    \
+    /* if run length > 15, must emit special run-length-16 codes (0xF0) */ \
+    while (r > 15) { \
+      emit_symbol(entropy, entropy->ac_tbl_no, 0xF0); \
+      r -= 16; \
+    } \
+    \
+    /* Find the number of bits needed for the magnitude of the coefficient */ \
+    nbits = JPEG_NBITS_NONZERO(temp);  /* there must be at least one 1 bit */ \
+    /* Check for out-of-range coefficient values */ \
+    if (nbits > MAX_COEF_BITS) \
+      ERREXIT(cinfo, JERR_BAD_DCT_COEF); \
+    \
+    /* Count/emit Huffman symbol for run length / number of bits */ \
+    emit_symbol(entropy, entropy->ac_tbl_no, (r << 4) + nbits); \
+    \
+    /* Emit that number of bits of the value, if positive, */ \
+    /* or the complement of its magnitude, if negative. */ \
+    emit_bits(entropy, (unsigned int)temp2, nbits); \
+    \
+    cvalue++; \
+    zerobits >>= 1; \
+  } \
+}
+
+METHODDEF(boolean)
+encode_mcu_AC_first(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  register int temp, temp2;
+  register int nbits, r;
+  int Sl = cinfo->Se - cinfo->Ss + 1;
+  int Al = cinfo->Al;
+  JCOEF values_unaligned2 * DCTSIZE2 + 15;
+  JCOEF *values;
+  const JCOEF *cvalue;
+  size_t zerobits;
+  size_t bits8 / SIZEOF_SIZE_T;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+#ifdef WITH_SIMD
+  cvalue = values = (JCOEF *)PAD((JUINTPTR)values_unaligned, 16);
+#else
+  /* Not using SIMD, so alignment is not needed */
+  cvalue = values = values_unaligned;
+#endif
+
+  /* Prepare data */
+  entropy->AC_first_prepare(MCU_data00, jpeg_natural_order + cinfo->Ss,
+                            Sl, Al, values, bits);
+
+  zerobits = bits0;
+#if SIZEOF_SIZE_T == 4
+  zerobits |= bits1;
+#endif
+
+  /* Emit any pending EOBRUN */
+  if (zerobits && (entropy->EOBRUN > 0))
+    emit_eobrun(entropy);
+
+#if SIZEOF_SIZE_T == 4
+  zerobits = bits0;
+#endif
+
+  /* Encode the AC coefficients per section G.1.2.2, fig. G.3 */
+
+  ENCODE_COEFS_AC_FIRST((void)0;);
+
+#if SIZEOF_SIZE_T == 4
+  zerobits = bits1;
+  if (zerobits) {
+    int diff = ((values + DCTSIZE2 / 2) - cvalue);
+    r = count_zeroes(&zerobits);
+    r += diff;
+    cvalue += r;
+    goto first_iter_ac_first;
+  }
+
+  ENCODE_COEFS_AC_FIRST(first_iter_ac_first:);
+#endif
+
+  if (cvalue < (values + Sl)) { /* If there are trailing zeroes, */
+    entropy->EOBRUN++;          /* count an EOB */
+    if (entropy->EOBRUN == 0x7FFF)
+      emit_eobrun(entropy);     /* force it out to avoid overflow */
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for DC successive approximation refinement scan.
+ * Note: we assume such scans can be multi-component, although the spec
+ * is not very clear on the point.
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  register int temp;
+  int blkn;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_datablkn;
+
+    /* We simply emit the Al'th bit of the DC coefficient value. */
+    temp = (*block)0;
+    emit_bits(entropy, (unsigned int)(temp >> Al), 1);
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Data preparation for encode_mcu_AC_refine().
+ */
+
+#define COMPUTE_ABSVALUES_AC_REFINE(Sl, koffset) { \
+  /* It is convenient to make a pre-pass to determine the transformed \
+   * coefficients' absolute values and the EOB position. \
+   */ \
+  for (k = 0; k < Sl; k++) { \
+    temp = blockjpeg_natural_order_startk; \
+    /* We must apply the point transform by Al.  For AC coefficients this \
+     * is an integer division with rounding towards 0.  To do this portably \
+     * in C, we shift after obtaining the absolute value. \
+     */ \
+    temp2 = temp >> (CHAR_BIT * sizeof(int) - 1); \
+    temp ^= temp2; \
+    temp -= temp2;              /* temp is abs value of input */ \
+    temp >>= Al;                /* apply the point transform */ \
+    if (temp != 0) { \
+      zerobits |= ((size_t)1U) << k; \
+      signbits |= ((size_t)(temp2 + 1)) << k; \
+    } \
+    absvaluesk = (JCOEF)temp; /* save abs value for main pass */ \
+    if (temp == 1) \
+      EOB = k + koffset;        /* EOB = index of last newly-nonzero coef */ \
+  } \
+}
+
+METHODDEF(int)
+encode_mcu_AC_refine_prepare(const JCOEF *block,
+                             const int *jpeg_natural_order_start, int Sl,
+                             int Al, JCOEF *absvalues, size_t *bits)
+{
+  register int k, temp, temp2;
+  int EOB = 0;
+  size_t zerobits = 0U, signbits = 0U;
+  int Sl0 = Sl;
+
+#if SIZEOF_SIZE_T == 4
+  if (Sl0 > 32)
+    Sl0 = 32;
+#endif
+
+  COMPUTE_ABSVALUES_AC_REFINE(Sl0, 0);
+
+  bits0 = zerobits;
+#if SIZEOF_SIZE_T == 8
+  bits1 = signbits;
+#else
+  bits2 = signbits;
+
+  zerobits = 0U;
+  signbits = 0U;
+
+  if (Sl > 32) {
+    Sl -= 32;
+    jpeg_natural_order_start += 32;
+    absvalues += 32;
+
+    COMPUTE_ABSVALUES_AC_REFINE(Sl, 32);
+  }
+
+  bits1 = zerobits;
+  bits3 = signbits;
+#endif
+
+  return EOB;
+}
+
+
+/*
+ * MCU encoding for AC successive approximation refinement scan.
+ */
+
+#define ENCODE_COEFS_AC_REFINE(label) { \
+  while (zerobits) { \
+    idx = count_zeroes(&zerobits); \
+    r += idx; \
+    cabsvalue += idx; \
+    signbits >>= idx; \
+label \
+    /* Emit any required ZRLs, but not if they can be folded into EOB */ \
+    while (r > 15 && (cabsvalue <= EOBPTR)) { \
+      /* emit any pending EOBRUN and the BE correction bits */ \
+      emit_eobrun(entropy); \
+      /* Emit ZRL */ \
+      emit_symbol(entropy, entropy->ac_tbl_no, 0xF0); \
+      r -= 16; \
+      /* Emit buffered correction bits that must be associated with ZRL */ \
+      emit_buffered_bits(entropy, BR_buffer, BR); \
+      BR_buffer = entropy->bit_buffer; /* BE bits are gone now */ \
+      BR = 0; \
+    } \
+    \
+    temp = *cabsvalue++; \
+    \
+    /* If the coef was previously nonzero, it only needs a correction bit. \
+     * NOTE: a straight translation of the spec's figure G.7 would suggest \
+     * that we also need to test r > 15.  But if r > 15, we can only get here \
+     * if k > EOB, which implies that this coefficient is not 1. \
+     */ \
+    if (temp > 1) { \
+      /* The correction bit is the next bit of the absolute value. */ \
+      BR_bufferBR++ = (char)(temp & 1); \
+      signbits >>= 1; \
+      zerobits >>= 1; \
+      continue; \
+    } \
+    \
+    /* Emit any pending EOBRUN and the BE correction bits */ \
+    emit_eobrun(entropy); \
+    \
+    /* Count/emit Huffman symbol for run length / number of bits */ \
+    emit_symbol(entropy, entropy->ac_tbl_no, (r << 4) + 1); \
+    \
+    /* Emit output bit for newly-nonzero coef */ \
+    temp = signbits & 1; /* ((*block)jpeg_natural_order_startk < 0) ? 0 : 1 */ \
+    emit_bits(entropy, (unsigned int)temp, 1); \
+    \
+    /* Emit buffered correction bits that must be associated with this code */ \
+    emit_buffered_bits(entropy, BR_buffer, BR); \
+    BR_buffer = entropy->bit_buffer; /* BE bits are gone now */ \
+    BR = 0; \
+    r = 0;                      /* reset zero run length */ \
+    signbits >>= 1; \
+    zerobits >>= 1; \
+  } \
+}
+
+METHODDEF(boolean)
+encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  register int temp, r, idx;
+  char *BR_buffer;
+  unsigned int BR;
+  int Sl = cinfo->Se - cinfo->Ss + 1;
+  int Al = cinfo->Al;
+  JCOEF absvalues_unalignedDCTSIZE2 + 15;
+  JCOEF *absvalues;
+  const JCOEF *cabsvalue, *EOBPTR;
+  size_t zerobits, signbits;
+  size_t bits16 / SIZEOF_SIZE_T;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+#ifdef WITH_SIMD
+  cabsvalue = absvalues = (JCOEF *)PAD((JUINTPTR)absvalues_unaligned, 16);
+#else
+  /* Not using SIMD, so alignment is not needed */
+  cabsvalue = absvalues = absvalues_unaligned;
+#endif
+
+  /* Prepare data */
+  EOBPTR = absvalues +
+    entropy->AC_refine_prepare(MCU_data00, jpeg_natural_order + cinfo->Ss,
+                               Sl, Al, absvalues, bits);
+
+  /* Encode the AC coefficients per section G.1.2.3, fig. G.7 */
+
+  r = 0;                        /* r = run length of zeros */
+  BR = 0;                       /* BR = count of buffered bits added now */
+  BR_buffer = entropy->bit_buffer + entropy->BE; /* Append bits to buffer */
+
+  zerobits = bits0;
+#if SIZEOF_SIZE_T == 8
+  signbits = bits1;
+#else
+  signbits = bits2;
+#endif
+  ENCODE_COEFS_AC_REFINE((void)0;);
+
+#if SIZEOF_SIZE_T == 4
+  zerobits = bits1;
+  signbits = bits3;
+
+  if (zerobits) {
+    int diff = ((absvalues + DCTSIZE2 / 2) - cabsvalue);
+    idx = count_zeroes(&zerobits);
+    signbits >>= idx;
+    idx += diff;
+    r += idx;
+    cabsvalue += idx;
+    goto first_iter_ac_refine;
+  }
+
+  ENCODE_COEFS_AC_REFINE(first_iter_ac_refine:);
+#endif
+
+  r |= (int)((absvalues + Sl) - cabsvalue);
+
+  if (r > 0 || BR > 0) {        /* If there are trailing zeroes, */
+    entropy->EOBRUN++;          /* count an EOB */
+    entropy->BE += BR;          /* concat my correction bits to older ones */
+    /* We force out the EOB if we risk either:
+     * 1. overflow of the EOB counter;
+     * 2. overflow of the correction bit buffer during the next MCU.
+     */
+    if (entropy->EOBRUN == 0x7FFF ||
+        entropy->BE > (MAX_CORR_BITS - DCTSIZE2 + 1))
+      emit_eobrun(entropy);
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Finish up at the end of a Huffman-compressed progressive scan.
+ */
+
+METHODDEF(void)
+finish_pass_phuff(j_compress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Flush out any buffered data */
+  emit_eobrun(entropy);
+  flush_bits(entropy);
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+}
+
+
+/*
+ * Finish up a statistics-gathering pass and create the new Huffman tables.
+ */
+
+METHODDEF(void)
+finish_pass_gather_phuff(j_compress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  boolean is_DC_band;
+  int ci, tbl;
+  jpeg_component_info *compptr;
+  JHUFF_TBL **htblptr;
+  boolean didNUM_HUFF_TBLS;
+
+  /* Flush out buffered data (all we care about is counting the EOB symbol) */
+  emit_eobrun(entropy);
+
+  is_DC_band = (cinfo->Ss == 0);
+
+  /* It's important not to apply jpeg_gen_optimal_table more than once
+   * per table, because it clobbers the input frequency counts!
+   */
+  MEMZERO(did, sizeof(did));
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    if (is_DC_band) {
+      if (cinfo->Ah != 0)       /* DC refinement needs no table */
+        continue;
+      tbl = compptr->dc_tbl_no;
+    } else {
+      tbl = compptr->ac_tbl_no;
+    }
+    if (!didtbl) {
+      if (is_DC_band)
+        htblptr = &cinfo->dc_huff_tbl_ptrstbl;
+      else
+        htblptr = &cinfo->ac_huff_tbl_ptrstbl;
+      if (*htblptr == NULL)
+        *htblptr = jpeg_alloc_huff_table((j_common_ptr)cinfo);
+      jpeg_gen_optimal_table(cinfo, *htblptr, entropy->count_ptrstbl);
+      didtbl = TRUE;
+    }
+  }
+}
+
+
+/*
+ * Module initialization routine for progressive Huffman entropy encoding.
+ */
+
+GLOBAL(void)
+jinit_phuff_encoder(j_compress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy;
+  int i;
+
+  entropy = (phuff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(phuff_entropy_encoder));
+  cinfo->entropy = (struct jpeg_entropy_encoder *)entropy;
+  entropy->pub.start_pass = start_pass_phuff;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->derived_tblsi = NULL;
+    entropy->count_ptrsi = NULL;
+  }
+  entropy->bit_buffer = NULL;   /* needed only in AC refinement scan */
+}
+
+#endif /* C_PROGRESSIVE_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcprepct.c Added

@@ -0,0 +1,351 @@
+/*
+ * jcprepct.c
+ *
+ * This file is part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the compression preprocessing controller.
+ * This controller manages the color conversion, downsampling,
+ * and edge expansion steps.
+ *
+ * Most of the complexity here is associated with buffering input rows
+ * as required by the downsampler.  See the comments at the head of
+ * jcsample.c for the downsampler's needs.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* At present, jcsample.c can request context rows only for smoothing.
+ * In the future, we might also need context rows for CCIR601 sampling
+ * or other more-complex downsampling procedures.  The code to support
+ * context rows should be compiled only if needed.
+ */
+#ifdef INPUT_SMOOTHING_SUPPORTED
+#define CONTEXT_ROWS_SUPPORTED
+#endif
+
+
+/*
+ * For the simple (no-context-row) case, we just need to buffer one
+ * row group's worth of pixels for the downsampling step.  At the bottom of
+ * the image, we pad to a full row group by replicating the last pixel row.
+ * The downsampler's last output row is then replicated if needed to pad
+ * out to a full iMCU row.
+ *
+ * When providing context rows, we must buffer three row groups' worth of
+ * pixels.  Three row groups are physically allocated, but the row pointer
+ * arrays are made five row groups high, with the extra pointers above and
+ * below "wrapping around" to point to the last and first real row groups.
+ * This allows the downsampler to access the proper context rows.
+ * At the top and bottom of the image, we create dummy context rows by
+ * copying the first or last real pixel row.  This copying could be avoided
+ * by pointer hacking as is done in jdmainct.c, but it doesn't seem worth the
+ * trouble on the compression side.
+ */
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_prep_controller pub; /* public fields */
+
+  /* Downsampling input buffer.  This buffer holds color-converted data
+   * until we have enough to do a downsample step.
+   */
+  JSAMPARRAY color_bufMAX_COMPONENTS;
+
+  JDIMENSION rows_to_go;        /* counts rows remaining in source image */
+  int next_buf_row;             /* index of next row to store in color_buf */
+
+#ifdef CONTEXT_ROWS_SUPPORTED   /* only needed for context case */
+  int this_row_group;           /* starting row index of group to process */
+  int next_buf_stop;            /* downsample when we reach this index */
+#endif
+} my_prep_controller;
+
+typedef my_prep_controller *my_prep_ptr;
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_prep(j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_prep_ptr prep = (my_prep_ptr)cinfo->prep;
+
+  if (pass_mode != JBUF_PASS_THRU)
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  /* Initialize total-height counter for detecting bottom of image */
+  prep->rows_to_go = cinfo->image_height;
+  /* Mark the conversion buffer empty */
+  prep->next_buf_row = 0;
+#ifdef CONTEXT_ROWS_SUPPORTED
+  /* Preset additional state variables for context mode.
+   * These aren't used in non-context mode, so we needn't test which mode.
+   */
+  prep->this_row_group = 0;
+  /* Set next_buf_stop to stop after two row groups have been read in. */
+  prep->next_buf_stop = 2 * cinfo->max_v_samp_factor;
+#endif
+}
+
+
+/*
+ * Expand an image vertically from height input_rows to height output_rows,
+ * by duplicating the bottom row.
+ */
+
+LOCAL(void)
+expand_bottom_edge(JSAMPARRAY image_data, JDIMENSION num_cols, int input_rows,
+                   int output_rows)
+{
+  register int row;
+
+  for (row = input_rows; row < output_rows; row++) {
+    jcopy_sample_rows(image_data, input_rows - 1, image_data, row, 1,
+                      num_cols);
+  }
+}
+
+
+/*
+ * Process some data in the simple no-context case.
+ *
+ * Preprocessor output data is counted in "row groups".  A row group
+ * is defined to be v_samp_factor sample rows of each component.
+ * Downsampling will produce this much data from each max_v_samp_factor
+ * input rows.
+ */
+
+METHODDEF(void)
+pre_process_data(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                 JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail,
+                 JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+                 JDIMENSION out_row_groups_avail)
+{
+  my_prep_ptr prep = (my_prep_ptr)cinfo->prep;
+  int numrows, ci;
+  JDIMENSION inrows;
+  jpeg_component_info *compptr;
+
+  while (*in_row_ctr < in_rows_avail &&
+         *out_row_group_ctr < out_row_groups_avail) {
+    /* Do color conversion to fill the conversion buffer. */
+    inrows = in_rows_avail - *in_row_ctr;
+    numrows = cinfo->max_v_samp_factor - prep->next_buf_row;
+    numrows = (int)MIN((JDIMENSION)numrows, inrows);
+    (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
+                                       prep->color_buf,
+                                       (JDIMENSION)prep->next_buf_row,
+                                       numrows);
+    *in_row_ctr += numrows;
+    prep->next_buf_row += numrows;
+    prep->rows_to_go -= numrows;
+    /* If at bottom of image, pad to fill the conversion buffer. */
+    if (prep->rows_to_go == 0 &&
+        prep->next_buf_row < cinfo->max_v_samp_factor) {
+      for (ci = 0; ci < cinfo->num_components; ci++) {
+        expand_bottom_edge(prep->color_bufci, cinfo->image_width,
+                           prep->next_buf_row, cinfo->max_v_samp_factor);
+      }
+      prep->next_buf_row = cinfo->max_v_samp_factor;
+    }
+    /* If we've filled the conversion buffer, empty it. */
+    if (prep->next_buf_row == cinfo->max_v_samp_factor) {
+      (*cinfo->downsample->downsample) (cinfo,
+                                        prep->color_buf, (JDIMENSION)0,
+                                        output_buf, *out_row_group_ctr);
+      prep->next_buf_row = 0;
+      (*out_row_group_ctr)++;
+    }
+    /* If at bottom of image, pad the output to a full iMCU height.
+     * Note we assume the caller is providing a one-iMCU-height output buffer!
+     */
+    if (prep->rows_to_go == 0 && *out_row_group_ctr < out_row_groups_avail) {
+      for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+           ci++, compptr++) {
+        expand_bottom_edge(output_bufci, compptr->width_in_blocks * DCTSIZE,
+                           (int)(*out_row_group_ctr * compptr->v_samp_factor),
+                           (int)(out_row_groups_avail * compptr->v_samp_factor));
+      }
+      *out_row_group_ctr = out_row_groups_avail;
+      break;                    /* can exit outer loop without test */
+    }
+  }
+}
+
+
+#ifdef CONTEXT_ROWS_SUPPORTED
+
+/*
+ * Process some data in the context case.
+ */
+
+METHODDEF(void)
+pre_process_context(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                    JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail,
+                    JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+                    JDIMENSION out_row_groups_avail)
+{
+  my_prep_ptr prep = (my_prep_ptr)cinfo->prep;
+  int numrows, ci;
+  int buf_height = cinfo->max_v_samp_factor * 3;
+  JDIMENSION inrows;
+
+  while (*out_row_group_ctr < out_row_groups_avail) {
+    if (*in_row_ctr < in_rows_avail) {
+      /* Do color conversion to fill the conversion buffer. */
+      inrows = in_rows_avail - *in_row_ctr;
+      numrows = prep->next_buf_stop - prep->next_buf_row;
+      numrows = (int)MIN((JDIMENSION)numrows, inrows);
+      (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
+                                         prep->color_buf,
+                                         (JDIMENSION)prep->next_buf_row,
+                                         numrows);
+      /* Pad at top of image, if first time through */
+      if (prep->rows_to_go == cinfo->image_height) {
+        for (ci = 0; ci < cinfo->num_components; ci++) {
+          int row;
+          for (row = 1; row <= cinfo->max_v_samp_factor; row++) {
+            jcopy_sample_rows(prep->color_bufci, 0, prep->color_bufci,
+                              -row, 1, cinfo->image_width);
+          }
+        }
+      }
+      *in_row_ctr += numrows;
+      prep->next_buf_row += numrows;
+      prep->rows_to_go -= numrows;
+    } else {
+      /* Return for more data, unless we are at the bottom of the image. */
+      if (prep->rows_to_go != 0)
+        break;
+      /* When at bottom of image, pad to fill the conversion buffer. */
+      if (prep->next_buf_row < prep->next_buf_stop) {
+        for (ci = 0; ci < cinfo->num_components; ci++) {
+          expand_bottom_edge(prep->color_bufci, cinfo->image_width,
+                             prep->next_buf_row, prep->next_buf_stop);
+        }
+        prep->next_buf_row = prep->next_buf_stop;
+      }
+    }
+    /* If we've gotten enough data, downsample a row group. */
+    if (prep->next_buf_row == prep->next_buf_stop) {
+      (*cinfo->downsample->downsample) (cinfo, prep->color_buf,
+                                        (JDIMENSION)prep->this_row_group,
+                                        output_buf, *out_row_group_ctr);
+      (*out_row_group_ctr)++;
+      /* Advance pointers with wraparound as necessary. */
+      prep->this_row_group += cinfo->max_v_samp_factor;
+      if (prep->this_row_group >= buf_height)
+        prep->this_row_group = 0;
+      if (prep->next_buf_row >= buf_height)
+        prep->next_buf_row = 0;
+      prep->next_buf_stop = prep->next_buf_row + cinfo->max_v_samp_factor;
+    }
+  }
+}
+
+
+/*
+ * Create the wrapped-around downsampling input buffer needed for context mode.
+ */
+
+LOCAL(void)
+create_context_buffer(j_compress_ptr cinfo)
+{
+  my_prep_ptr prep = (my_prep_ptr)cinfo->prep;
+  int rgroup_height = cinfo->max_v_samp_factor;
+  int ci, i;
+  jpeg_component_info *compptr;
+  JSAMPARRAY true_buffer, fake_buffer;
+
+  /* Grab enough space for fake row pointers for all the components;
+   * we need five row groups' worth of pointers for each component.
+   */
+  fake_buffer = (JSAMPARRAY)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (cinfo->num_components * 5 * rgroup_height) *
+                                sizeof(JSAMPROW));
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Allocate the actual buffer space (3 row groups) for this component.
+     * We make the buffer wide enough to allow the downsampler to edge-expand
+     * horizontally within the buffer, if it so chooses.
+     */
+    true_buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE,
+       (JDIMENSION)(((long)compptr->width_in_blocks * DCTSIZE *
+                     cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+       (JDIMENSION)(3 * rgroup_height));
+    /* Copy true buffer row pointers into the middle of the fake row array */
+    MEMCOPY(fake_buffer + rgroup_height, true_buffer,
+            3 * rgroup_height * sizeof(JSAMPROW));
+    /* Fill in the above and below wraparound pointers */
+    for (i = 0; i < rgroup_height; i++) {
+      fake_bufferi = true_buffer2 * rgroup_height + i;
+      fake_buffer4 * rgroup_height + i = true_bufferi;
+    }
+    prep->color_bufci = fake_buffer + rgroup_height;
+    fake_buffer += 5 * rgroup_height; /* point to space for next component */
+  }
+}
+
+#endif /* CONTEXT_ROWS_SUPPORTED */
+
+
+/*
+ * Initialize preprocessing controller.
+ */
+
+GLOBAL(void)
+jinit_c_prep_controller(j_compress_ptr cinfo, boolean need_full_buffer)
+{
+  my_prep_ptr prep;
+  int ci;
+  jpeg_component_info *compptr;
+
+  if (need_full_buffer)         /* safety check */
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  prep = (my_prep_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_prep_controller));
+  cinfo->prep = (struct jpeg_c_prep_controller *)prep;
+  prep->pub.start_pass = start_pass_prep;
+
+  /* Allocate the color conversion buffer.
+   * We make the buffer wide enough to allow the downsampler to edge-expand
+   * horizontally within the buffer, if it so chooses.
+   */
+  if (cinfo->downsample->need_context_rows) {
+    /* Set up to provide context rows */
+#ifdef CONTEXT_ROWS_SUPPORTED
+    prep->pub.pre_process_data = pre_process_context;
+    create_context_buffer(cinfo);
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    /* No context, just make it tall enough for one row group */
+    prep->pub.pre_process_data = pre_process_data;
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+         ci++, compptr++) {
+      prep->color_bufci = (*cinfo->mem->alloc_sarray)
+        ((j_common_ptr)cinfo, JPOOL_IMAGE,
+         (JDIMENSION)(((long)compptr->width_in_blocks * DCTSIZE *
+                       cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+         (JDIMENSION)cinfo->max_v_samp_factor);
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcsample.c Added

@@ -0,0 +1,522 @@
+/*
+ * jcsample.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2014, MIPS Technologies, Inc., California.
+ * Copyright (C) 2015, 2019, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains downsampling routines.
+ *
+ * Downsampling input data is counted in "row groups".  A row group
+ * is defined to be max_v_samp_factor pixel rows of each component,
+ * from which the downsampler produces v_samp_factor sample rows.
+ * A single row group is processed in each call to the downsampler module.
+ *
+ * The downsampler is responsible for edge-expansion of its output data
+ * to fill an integral number of DCT blocks horizontally.  The source buffer
+ * may be modified if it is helpful for this purpose (the source buffer is
+ * allocated wide enough to correspond to the desired output width).
+ * The caller (the prep controller) is responsible for vertical padding.
+ *
+ * The downsampler may request "context rows" by setting need_context_rows
+ * during startup.  In this case, the input arrays will contain at least
+ * one row group's worth of pixels above and below the passed-in data;
+ * the caller will create dummy rows at image top and bottom by replicating
+ * the first or last real pixel row.
+ *
+ * An excellent reference for image resampling is
+ *   Digital Image Warping, George Wolberg, 1990.
+ *   Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7.
+ *
+ * The downsampling algorithm used here is a simple average of the source
+ * pixels covered by the output pixel.  The hi-falutin sampling literature
+ * refers to this as a "box filter".  In general the characteristics of a box
+ * filter are not very good, but for the specific cases we normally use (1:1
+ * and 2:1 ratios) the box is equivalent to a "triangle filter" which is not
+ * nearly so bad.  If you intend to use other sampling ratios, you'd be well
+ * advised to improve this code.
+ *
+ * A simple input-smoothing capability is provided.  This is mainly intended
+ * for cleaning up color-dithered GIF input files (if you find it inadequate,
+ * we suggest using an external filtering program such as pnmconvol).  When
+ * enabled, each input pixel P is replaced by a weighted sum of itself and its
+ * eight neighbors.  P's weight is 1-8*SF and each neighbor's weight is SF,
+ * where SF = (smoothing_factor / 1024).
+ * Currently, smoothing is only supported for 2h2v sampling factors.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jsimd.h"
+
+
+/* Pointer to routine to downsample a single component */
+typedef void (*downsample1_ptr) (j_compress_ptr cinfo,
+                                 jpeg_component_info *compptr,
+                                 JSAMPARRAY input_data,
+                                 JSAMPARRAY output_data);
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_downsampler pub;  /* public fields */
+
+  /* Downsampling method pointers, one per component */
+  downsample1_ptr methodsMAX_COMPONENTS;
+} my_downsampler;
+
+typedef my_downsampler *my_downsample_ptr;
+
+
+/*
+ * Initialize for a downsampling pass.
+ */
+
+METHODDEF(void)
+start_pass_downsample(j_compress_ptr cinfo)
+{
+  /* no work for now */
+}
+
+
+/*
+ * Expand a component horizontally from width input_cols to width output_cols,
+ * by duplicating the rightmost samples.
+ */
+
+LOCAL(void)
+expand_right_edge(JSAMPARRAY image_data, int num_rows, JDIMENSION input_cols,
+                  JDIMENSION output_cols)
+{
+  register JSAMPROW ptr;
+  register JSAMPLE pixval;
+  register int count;
+  int row;
+  int numcols = (int)(output_cols - input_cols);
+
+  if (numcols > 0) {
+    for (row = 0; row < num_rows; row++) {
+      ptr = image_datarow + input_cols;
+      pixval = ptr-1;
+      for (count = numcols; count > 0; count--)
+        *ptr++ = pixval;
+    }
+  }
+}
+
+
+/*
+ * Do downsampling for a whole row group (all components).
+ *
+ * In this version we simply downsample each component independently.
+ */
+
+METHODDEF(void)
+sep_downsample(j_compress_ptr cinfo, JSAMPIMAGE input_buf,
+               JDIMENSION in_row_index, JSAMPIMAGE output_buf,
+               JDIMENSION out_row_group_index)
+{
+  my_downsample_ptr downsample = (my_downsample_ptr)cinfo->downsample;
+  int ci;
+  jpeg_component_info *compptr;
+  JSAMPARRAY in_ptr, out_ptr;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    in_ptr = input_bufci + in_row_index;
+    out_ptr = output_bufci + (out_row_group_index * compptr->v_samp_factor);
+    (*downsample->methodsci) (cinfo, compptr, in_ptr, out_ptr);
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * One row group is processed per call.
+ * This version handles arbitrary integral sampling ratios, without smoothing.
+ * Note that this version is not actually used for customary sampling ratios.
+ */
+
+METHODDEF(void)
+int_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+               JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow, h_expand, v_expand, numpix, numpix2, h, v;
+  JDIMENSION outcol, outcol_h;  /* outcol_h == outcol*h_expand */
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  JSAMPROW inptr, outptr;
+  JLONG outvalue;
+
+  h_expand = cinfo->max_h_samp_factor / compptr->h_samp_factor;
+  v_expand = cinfo->max_v_samp_factor / compptr->v_samp_factor;
+  numpix = h_expand * v_expand;
+  numpix2 = numpix / 2;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data, cinfo->max_v_samp_factor, cinfo->image_width,
+                    output_cols * h_expand);
+
+  inrow = 0;
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_dataoutrow;
+    for (outcol = 0, outcol_h = 0; outcol < output_cols;
+         outcol++, outcol_h += h_expand) {
+      outvalue = 0;
+      for (v = 0; v < v_expand; v++) {
+        inptr = input_datainrow + v + outcol_h;
+        for (h = 0; h < h_expand; h++) {
+          outvalue += (JLONG)(*inptr++);
+        }
+      }
+      *outptr++ = (JSAMPLE)((outvalue + numpix2) / numpix);
+    }
+    inrow += v_expand;
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the special case of a full-size component,
+ * without smoothing.
+ */
+
+METHODDEF(void)
+fullsize_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  /* Copy the data */
+  jcopy_sample_rows(input_data, 0, output_data, 0, cinfo->max_v_samp_factor,
+                    cinfo->image_width);
+  /* Edge-expand */
+  expand_right_edge(output_data, cinfo->max_v_samp_factor, cinfo->image_width,
+                    compptr->width_in_blocks * DCTSIZE);
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the common case of 2:1 horizontal and 1:1 vertical,
+ * without smoothing.
+ *
+ * A note about the "bias" calculations: when rounding fractional values to
+ * integer, we do not want to always round 0.5 up to the next integer.
+ * If we did that, we'd introduce a noticeable bias towards larger values.
+ * Instead, this code is arranged so that 0.5 will be rounded up or down at
+ * alternate pixel locations (a simple ordered dither pattern).
+ */
+
+METHODDEF(void)
+h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int outrow;
+  JDIMENSION outcol;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr, outptr;
+  register int bias;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data, cinfo->max_v_samp_factor, cinfo->image_width,
+                    output_cols * 2);
+
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_dataoutrow;
+    inptr = input_dataoutrow;
+    bias = 0;                   /* bias = 0,1,0,1,... for successive samples */
+    for (outcol = 0; outcol < output_cols; outcol++) {
+      *outptr++ = (JSAMPLE)((inptr0 + inptr1 + bias) >> 1);
+      bias ^= 1;                /* 0=>1, 1=>0 */
+      inptr += 2;
+    }
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+ * without smoothing.
+ */
+
+METHODDEF(void)
+h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow;
+  JDIMENSION outcol;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr0, inptr1, outptr;
+  register int bias;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data, cinfo->max_v_samp_factor, cinfo->image_width,
+                    output_cols * 2);
+
+  inrow = 0;
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_dataoutrow;
+    inptr0 = input_datainrow;
+    inptr1 = input_datainrow + 1;
+    bias = 1;                   /* bias = 1,2,1,2,... for successive samples */
+    for (outcol = 0; outcol < output_cols; outcol++) {
+      *outptr++ =
+        (JSAMPLE)((inptr00 + inptr01 + inptr10 + inptr11 + bias) >> 2);
+      bias ^= 3;                /* 1=>2, 2=>1 */
+      inptr0 += 2;  inptr1 += 2;
+    }
+    inrow += 2;
+  }
+}
+
+
+#ifdef INPUT_SMOOTHING_SUPPORTED
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+ * with smoothing.  One row of context is required.
+ */
+
+METHODDEF(void)
+h2v2_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow;
+  JDIMENSION colctr;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr0, inptr1, above_ptr, below_ptr, outptr;
+  JLONG membersum, neighsum, memberscale, neighscale;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
+                    cinfo->image_width, output_cols * 2);
+
+  /* We don't bother to form the individual "smoothed" input pixel values;
+   * we can directly compute the output which is the average of the four
+   * smoothed values.  Each of the four member pixels contributes a fraction
+   * (1-8*SF) to its own smoothed image and a fraction SF to each of the three
+   * other smoothed pixels, therefore a total fraction (1-5*SF)/4 to the final
+   * output.  The four corner-adjacent neighbor pixels contribute a fraction
+   * SF to just one smoothed pixel, or SF/4 to the final output; while the
+   * eight edge-adjacent neighbors contribute SF to each of two smoothed
+   * pixels, or SF/2 overall.  In order to use integer arithmetic, these
+   * factors are scaled by 2^16 = 65536.
+   * Also recall that SF = smoothing_factor / 1024.
+   */
+
+  memberscale = 16384 - cinfo->smoothing_factor * 80; /* scaled (1-5*SF)/4 */
+  neighscale = cinfo->smoothing_factor * 16; /* scaled SF/4 */
+
+  inrow = 0;
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_dataoutrow;
+    inptr0 = input_datainrow;
+    inptr1 = input_datainrow + 1;
+    above_ptr = input_datainrow - 1;
+    below_ptr = input_datainrow + 2;
+
+    /* Special case for first column: pretend column -1 is same as column 0 */
+    membersum = inptr00 + inptr01 + inptr10 + inptr11;
+    neighsum = above_ptr0 + above_ptr1 + below_ptr0 + below_ptr1 +
+               inptr00 + inptr02 + inptr10 + inptr12;
+    neighsum += neighsum;
+    neighsum += above_ptr0 + above_ptr2 + below_ptr0 + below_ptr2;
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
+    inptr0 += 2;  inptr1 += 2;  above_ptr += 2;  below_ptr += 2;
+
+    for (colctr = output_cols - 2; colctr > 0; colctr--) {
+      /* sum of pixels directly mapped to this output element */
+      membersum = inptr00 + inptr01 + inptr10 + inptr11;
+      /* sum of edge-neighbor pixels */
+      neighsum = above_ptr0 + above_ptr1 + below_ptr0 + below_ptr1 +
+                 inptr0-1 + inptr02 + inptr1-1 + inptr12;
+      /* The edge-neighbors count twice as much as corner-neighbors */
+      neighsum += neighsum;
+      /* Add in the corner-neighbors */
+      neighsum += above_ptr-1 + above_ptr2 + below_ptr-1 + below_ptr2;
+      /* form final output scaled up by 2^16 */
+      membersum = membersum * memberscale + neighsum * neighscale;
+      /* round, descale and output it */
+      *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
+      inptr0 += 2;  inptr1 += 2;  above_ptr += 2;  below_ptr += 2;
+    }
+
+    /* Special case for last column */
+    membersum = inptr00 + inptr01 + inptr10 + inptr11;
+    neighsum = above_ptr0 + above_ptr1 + below_ptr0 + below_ptr1 +
+               inptr0-1 + inptr01 + inptr1-1 + inptr11;
+    neighsum += neighsum;
+    neighsum += above_ptr-1 + above_ptr1 + below_ptr-1 + below_ptr1;
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr = (JSAMPLE)((membersum + 32768) >> 16);
+
+    inrow += 2;
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the special case of a full-size component,
+ * with smoothing.  One row of context is required.
+ */
+
+METHODDEF(void)
+fullsize_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                           JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int outrow;
+  JDIMENSION colctr;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr, above_ptr, below_ptr, outptr;
+  JLONG membersum, neighsum, memberscale, neighscale;
+  int colsum, lastcolsum, nextcolsum;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
+                    cinfo->image_width, output_cols);
+
+  /* Each of the eight neighbor pixels contributes a fraction SF to the
+   * smoothed pixel, while the main pixel contributes (1-8*SF).  In order
+   * to use integer arithmetic, these factors are multiplied by 2^16 = 65536.
+   * Also recall that SF = smoothing_factor / 1024.
+   */
+
+  memberscale = 65536L - cinfo->smoothing_factor * 512L; /* scaled 1-8*SF */
+  neighscale = cinfo->smoothing_factor * 64; /* scaled SF */
+
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_dataoutrow;
+    inptr = input_dataoutrow;
+    above_ptr = input_dataoutrow - 1;
+    below_ptr = input_dataoutrow + 1;
+
+    /* Special case for first column */
+    colsum = (*above_ptr++) + (*below_ptr++) + inptr0;
+    membersum = *inptr++;
+    nextcolsum = above_ptr0 + below_ptr0 + inptr0;
+    neighsum = colsum + (colsum - membersum) + nextcolsum;
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
+    lastcolsum = colsum;  colsum = nextcolsum;
+
+    for (colctr = output_cols - 2; colctr > 0; colctr--) {
+      membersum = *inptr++;
+      above_ptr++;  below_ptr++;
+      nextcolsum = above_ptr0 + below_ptr0 + inptr0;
+      neighsum = lastcolsum + (colsum - membersum) + nextcolsum;
+      membersum = membersum * memberscale + neighsum * neighscale;
+      *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
+      lastcolsum = colsum;  colsum = nextcolsum;
+    }
+
+    /* Special case for last column */
+    membersum = *inptr;
+    neighsum = lastcolsum + (colsum - membersum) + colsum;
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr = (JSAMPLE)((membersum + 32768) >> 16);
+
+  }
+}
+
+#endif /* INPUT_SMOOTHING_SUPPORTED */
+
+
+/*
+ * Module initialization routine for downsampling.
+ * Note that we must select a routine for each component.
+ */
+
+GLOBAL(void)
+jinit_downsampler(j_compress_ptr cinfo)
+{
+  my_downsample_ptr downsample;
+  int ci;
+  jpeg_component_info *compptr;
+  boolean smoothok = TRUE;
+
+  downsample = (my_downsample_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_downsampler));
+  cinfo->downsample = (struct jpeg_downsampler *)downsample;
+  downsample->pub.start_pass = start_pass_downsample;
+  downsample->pub.downsample = sep_downsample;
+  downsample->pub.need_context_rows = FALSE;
+
+  if (cinfo->CCIR601_sampling)
+    ERREXIT(cinfo, JERR_CCIR601_NOTIMPL);
+
+  /* Verify we can handle the sampling factors, and set up method pointers */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    if (compptr->h_samp_factor == cinfo->max_h_samp_factor &&
+        compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+#ifdef INPUT_SMOOTHING_SUPPORTED
+      if (cinfo->smoothing_factor) {
+        downsample->methodsci = fullsize_smooth_downsample;
+        downsample->pub.need_context_rows = TRUE;
+      } else
+#endif
+        downsample->methodsci = fullsize_downsample;
+    } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
+               compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+      smoothok = FALSE;
+      if (jsimd_can_h2v1_downsample())
+        downsample->methodsci = jsimd_h2v1_downsample;
+      else
+        downsample->methodsci = h2v1_downsample;
+    } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
+               compptr->v_samp_factor * 2 == cinfo->max_v_samp_factor) {
+#ifdef INPUT_SMOOTHING_SUPPORTED
+      if (cinfo->smoothing_factor) {
+#if defined(__mips__)
+        if (jsimd_can_h2v2_smooth_downsample())
+          downsample->methodsci = jsimd_h2v2_smooth_downsample;
+        else
+#endif
+          downsample->methodsci = h2v2_smooth_downsample;
+        downsample->pub.need_context_rows = TRUE;
+      } else
+#endif
+      {
+        if (jsimd_can_h2v2_downsample())
+          downsample->methodsci = jsimd_h2v2_downsample;
+        else
+          downsample->methodsci = h2v2_downsample;
+      }
+    } else if ((cinfo->max_h_samp_factor % compptr->h_samp_factor) == 0 &&
+               (cinfo->max_v_samp_factor % compptr->v_samp_factor) == 0) {
+      smoothok = FALSE;
+      downsample->methodsci = int_downsample;
+    } else
+      ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL);
+  }
+
+#ifdef INPUT_SMOOTHING_SUPPORTED
+  if (cinfo->smoothing_factor && !smoothok)
+    TRACEMS(cinfo, 0, JTRC_SMOOTH_NOTIMPL);
+#endif
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jcstest.c Added

@@ -0,0 +1,126 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* This program demonstrates how to check for the colorspace extension
+   capabilities of libjpeg-turbo at both compile time and run time. */
+
+#include <stdio.h>
+#include <jpeglib.h>
+#include <jerror.h>
+#include <setjmp.h>
+
+#ifndef JCS_EXTENSIONS
+#define JCS_EXT_RGB  6
+#endif
+#if !defined(JCS_EXTENSIONS) || !defined(JCS_ALPHA_EXTENSIONS)
+#define JCS_EXT_RGBA  12
+#endif
+
+static char lasterrorJMSG_LENGTH_MAX = "No error";
+
+typedef struct _error_mgr {
+  struct jpeg_error_mgr pub;
+  jmp_buf jb;
+} error_mgr;
+
+static void my_error_exit(j_common_ptr cinfo)
+{
+  error_mgr *myerr = (error_mgr *)cinfo->err;
+  (*cinfo->err->output_message) (cinfo);
+  longjmp(myerr->jb, 1);
+}
+
+static void my_output_message(j_common_ptr cinfo)
+{
+  (*cinfo->err->format_message) (cinfo, lasterror);
+}
+
+int main(void)
+{
+  int jcs_valid = -1, jcs_alpha_valid = -1;
+  struct jpeg_compress_struct cinfo;
+  error_mgr jerr;
+
+  printf("libjpeg-turbo colorspace extensions:\n");
+#if JCS_EXTENSIONS
+  printf("  Present at compile time\n");
+#else
+  printf("  Not present at compile time\n");
+#endif
+
+  cinfo.err = jpeg_std_error(&jerr.pub);
+  jerr.pub.error_exit = my_error_exit;
+  jerr.pub.output_message = my_output_message;
+
+  if (setjmp(jerr.jb)) {
+    /* this will execute if libjpeg has an error */
+    jcs_valid = 0;
+    goto done;
+  }
+
+  jpeg_create_compress(&cinfo);
+  cinfo.input_components = 3;
+  jpeg_set_defaults(&cinfo);
+  cinfo.in_color_space = JCS_EXT_RGB;
+  jpeg_default_colorspace(&cinfo);
+  jcs_valid = 1;
+
+done:
+  if (jcs_valid)
+    printf("  Working properly\n");
+  else
+    printf("  Not working properly.  Error returned was:\n    %s\n",
+           lasterror);
+
+  printf("libjpeg-turbo alpha colorspace extensions:\n");
+#if JCS_ALPHA_EXTENSIONS
+  printf("  Present at compile time\n");
+#else
+  printf("  Not present at compile time\n");
+#endif
+
+  if (setjmp(jerr.jb)) {
+    /* this will execute if libjpeg has an error */
+    jcs_alpha_valid = 0;
+    goto done2;
+  }
+
+  cinfo.in_color_space = JCS_EXT_RGBA;
+  jpeg_default_colorspace(&cinfo);
+  jcs_alpha_valid = 1;
+
+done2:
+  if (jcs_alpha_valid)
+    printf("  Working properly\n");
+  else
+    printf("  Not working properly.  Error returned was:\n    %s\n",
+           lasterror);
+
+  jpeg_destroy_compress(&cinfo);
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jctrans.c Added

@@ -0,0 +1,401 @@
+/*
+ * jctrans.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1995-1998, Thomas G. Lane.
+ * Modified 2000-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains library routines for transcoding compression,
+ * that is, writing raw DCT coefficient arrays to an output JPEG file.
+ * The routines in jcapimin.c will also be needed by a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jpegcomp.h"
+
+
+/* Forward declarations */
+LOCAL(void) transencode_master_selection(j_compress_ptr cinfo,
+                                         jvirt_barray_ptr *coef_arrays);
+LOCAL(void) transencode_coef_controller(j_compress_ptr cinfo,
+                                        jvirt_barray_ptr *coef_arrays);
+
+
+/*
+ * Compression initialization for writing raw-coefficient data.
+ * Before calling this, all parameters and a data destination must be set up.
+ * Call jpeg_finish_compress() to actually write the data.
+ *
+ * The number of passed virtual arrays must match cinfo->num_components.
+ * Note that the virtual arrays need not be filled or even realized at
+ * the time write_coefficients is called; indeed, if the virtual arrays
+ * were requested from this compression object's memory manager, they
+ * typically will be realized during this routine and filled afterwards.
+ */
+
+GLOBAL(void)
+jpeg_write_coefficients(j_compress_ptr cinfo, jvirt_barray_ptr *coef_arrays)
+{
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Mark all tables to be written */
+  jpeg_suppress_tables(cinfo, FALSE);
+  /* (Re)initialize error mgr and destination modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr)cinfo);
+  (*cinfo->dest->init_destination) (cinfo);
+  /* Perform master selection of active modules */
+  transencode_master_selection(cinfo, coef_arrays);
+  /* Wait for jpeg_finish_compress() call */
+  cinfo->next_scanline = 0;     /* so jpeg_write_marker works */
+  cinfo->global_state = CSTATE_WRCOEFS;
+}
+
+
+/*
+ * Initialize the compression object with default parameters,
+ * then copy from the source object all parameters needed for lossless
+ * transcoding.  Parameters that can be varied without loss (such as
+ * scan script and Huffman optimization) are left in their default states.
+ */
+
+GLOBAL(void)
+jpeg_copy_critical_parameters(j_decompress_ptr srcinfo, j_compress_ptr dstinfo)
+{
+  JQUANT_TBL **qtblptr;
+  jpeg_component_info *incomp, *outcomp;
+  JQUANT_TBL *c_quant, *slot_quant;
+  int tblno, ci, coefi;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (dstinfo->global_state != CSTATE_START)
+    ERREXIT1(dstinfo, JERR_BAD_STATE, dstinfo->global_state);
+  /* Copy fundamental image dimensions */
+  dstinfo->image_width = srcinfo->image_width;
+  dstinfo->image_height = srcinfo->image_height;
+  dstinfo->input_components = srcinfo->num_components;
+  dstinfo->in_color_space = srcinfo->jpeg_color_space;
+#if JPEG_LIB_VERSION >= 70
+  dstinfo->jpeg_width = srcinfo->output_width;
+  dstinfo->jpeg_height = srcinfo->output_height;
+  dstinfo->min_DCT_h_scaled_size = srcinfo->min_DCT_h_scaled_size;
+  dstinfo->min_DCT_v_scaled_size = srcinfo->min_DCT_v_scaled_size;
+#endif
+  /* Initialize all parameters to default values */
+  jpeg_set_defaults(dstinfo);
+  /* jpeg_set_defaults may choose wrong colorspace, eg YCbCr if input is RGB.
+   * Fix it to get the right header markers for the image colorspace.
+   */
+  jpeg_set_colorspace(dstinfo, srcinfo->jpeg_color_space);
+  dstinfo->data_precision = srcinfo->data_precision;
+  dstinfo->CCIR601_sampling = srcinfo->CCIR601_sampling;
+  /* Copy the source's quantization tables. */
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    if (srcinfo->quant_tbl_ptrstblno != NULL) {
+      qtblptr = &dstinfo->quant_tbl_ptrstblno;
+      if (*qtblptr == NULL)
+        *qtblptr = jpeg_alloc_quant_table((j_common_ptr)dstinfo);
+      MEMCOPY((*qtblptr)->quantval, srcinfo->quant_tbl_ptrstblno->quantval,
+              sizeof((*qtblptr)->quantval));
+      (*qtblptr)->sent_table = FALSE;
+    }
+  }
+  /* Copy the source's per-component info.
+   * Note we assume jpeg_set_defaults has allocated the dest comp_info array.
+   */
+  dstinfo->num_components = srcinfo->num_components;
+  if (dstinfo->num_components < 1 || dstinfo->num_components > MAX_COMPONENTS)
+    ERREXIT2(dstinfo, JERR_COMPONENT_COUNT, dstinfo->num_components,
+             MAX_COMPONENTS);
+  for (ci = 0, incomp = srcinfo->comp_info, outcomp = dstinfo->comp_info;
+       ci < dstinfo->num_components; ci++, incomp++, outcomp++) {
+    outcomp->component_id = incomp->component_id;
+    outcomp->h_samp_factor = incomp->h_samp_factor;
+    outcomp->v_samp_factor = incomp->v_samp_factor;
+    outcomp->quant_tbl_no = incomp->quant_tbl_no;
+    /* Make sure saved quantization table for component matches the qtable
+     * slot.  If not, the input file re-used this qtable slot.
+     * IJG encoder currently cannot duplicate this.
+     */
+    tblno = outcomp->quant_tbl_no;
+    if (tblno < 0 || tblno >= NUM_QUANT_TBLS ||
+        srcinfo->quant_tbl_ptrstblno == NULL)
+      ERREXIT1(dstinfo, JERR_NO_QUANT_TABLE, tblno);
+    slot_quant = srcinfo->quant_tbl_ptrstblno;
+    c_quant = incomp->quant_table;
+    if (c_quant != NULL) {
+      for (coefi = 0; coefi < DCTSIZE2; coefi++) {
+        if (c_quant->quantvalcoefi != slot_quant->quantvalcoefi)
+          ERREXIT1(dstinfo, JERR_MISMATCHED_QUANT_TABLE, tblno);
+      }
+    }
+    /* Note: we do not copy the source's Huffman table assignments;
+     * instead we rely on jpeg_set_colorspace to have made a suitable choice.
+     */
+  }
+  /* Also copy JFIF version and resolution information, if available.
+   * Strictly speaking this isn't "critical" info, but it's nearly
+   * always appropriate to copy it if available.  In particular,
+   * if the application chooses to copy JFIF 1.02 extension markers from
+   * the source file, we need to copy the version to make sure we don't
+   * emit a file that has 1.02 extensions but a claimed version of 1.01.
+   * We will *not*, however, copy version info from mislabeled "2.01" files.
+   */
+  if (srcinfo->saw_JFIF_marker) {
+    if (srcinfo->JFIF_major_version == 1) {
+      dstinfo->JFIF_major_version = srcinfo->JFIF_major_version;
+      dstinfo->JFIF_minor_version = srcinfo->JFIF_minor_version;
+    }
+    dstinfo->density_unit = srcinfo->density_unit;
+    dstinfo->X_density = srcinfo->X_density;
+    dstinfo->Y_density = srcinfo->Y_density;
+  }
+}
+
+
+/*
+ * Master selection of compression modules for transcoding.
+ * This substitutes for jcinit.c's initialization of the full compressor.
+ */
+
+LOCAL(void)
+transencode_master_selection(j_compress_ptr cinfo,
+                             jvirt_barray_ptr *coef_arrays)
+{
+  /* Although we don't actually use input_components for transcoding,
+   * jcmaster.c's initial_setup will complain if input_components is 0.
+   */
+  cinfo->input_components = 1;
+  /* Initialize master control (includes parameter checking/processing) */
+  jinit_c_master_control(cinfo, TRUE /* transcode only */);
+
+  /* Entropy encoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+#ifdef C_ARITH_CODING_SUPPORTED
+    jinit_arith_encoder(cinfo);
+#else
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+#endif
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+      jinit_phuff_encoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_encoder(cinfo);
+  }
+
+  /* We need a special coefficient buffer controller. */
+  transencode_coef_controller(cinfo, coef_arrays);
+
+  jinit_marker_writer(cinfo);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr)cinfo);
+
+  /* Write the datastream header (SOI, JFIF) immediately.
+   * Frame and scan headers are postponed till later.
+   * This lets application insert special markers after the SOI.
+   */
+  (*cinfo->marker->write_file_header) (cinfo);
+}
+
+
+/*
+ * The rest of this file is a special implementation of the coefficient
+ * buffer controller.  This is similar to jccoefct.c, but it handles only
+ * output from presupplied virtual arrays.  Furthermore, we generate any
+ * dummy padding blocks on-the-fly rather than expecting them to be present
+ * in the arrays.
+ */
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_coef_controller pub; /* public fields */
+
+  JDIMENSION iMCU_row_num;      /* iMCU row # within image */
+  JDIMENSION mcu_ctr;           /* counts MCUs processed in current row */
+  int MCU_vert_offset;          /* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;    /* number of such rows needed */
+
+  /* Virtual block array for each component. */
+  jvirt_barray_ptr *whole_image;
+
+  /* Workspace for constructing dummy blocks at right/bottom edges. */
+  JBLOCKROW dummy_bufferC_MAX_BLOCKS_IN_MCU;
+} my_coef_controller;
+
+typedef my_coef_controller *my_coef_ptr;
+
+
+LOCAL(void)
+start_iMCU_row(j_compress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row */
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (coef->iMCU_row_num < (cinfo->total_iMCU_rows - 1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info0->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info0->last_row_height;
+  }
+
+  coef->mcu_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_coef(j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+
+  if (pass_mode != JBUF_CRANK_DEST)
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  coef->iMCU_row_num = 0;
+  start_iMCU_row(cinfo);
+}
+
+
+/*
+ * Process some data.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the scan.
+ * The data is obtained from the virtual arrays and fed to the entropy coder.
+ * Returns TRUE if the iMCU row is completed, FALSE if suspended.
+ *
+ * NB: input_buf is ignored; it is likely to be a NULL pointer.
+ */
+
+METHODDEF(boolean)
+compress_output(j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
+  JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  int blkn, ci, xindex, yindex, yoffset, blockcnt;
+  JDIMENSION start_col;
+  JBLOCKARRAY bufferMAX_COMPS_IN_SCAN;
+  JBLOCKROW MCU_bufferC_MAX_BLOCKS_IN_MCU;
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  /* Align the virtual buffers for the components used in this scan. */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    bufferci = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr)cinfo, coef->whole_imagecompptr->component_index,
+       coef->iMCU_row_num * compptr->v_samp_factor,
+       (JDIMENSION)compptr->v_samp_factor, FALSE);
+  }
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
+         MCU_col_num++) {
+      /* Construct list of pointers to DCT blocks belonging to this MCU */
+      blkn = 0;                 /* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+        compptr = cinfo->cur_comp_infoci;
+        start_col = MCU_col_num * compptr->MCU_width;
+        blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width :
+                                                  compptr->last_col_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          if (coef->iMCU_row_num < last_iMCU_row ||
+              yindex + yoffset < compptr->last_row_height) {
+            /* Fill in pointers to real blocks in this row */
+            buffer_ptr = bufferciyindex + yoffset + start_col;
+            for (xindex = 0; xindex < blockcnt; xindex++)
+              MCU_bufferblkn++ = buffer_ptr++;
+          } else {
+            /* At bottom of image, need a whole row of dummy blocks */
+            xindex = 0;
+          }
+          /* Fill in any dummy blocks needed in this row.
+           * Dummy blocks are filled in the same way as in jccoefct.c:
+           * all zeroes in the AC entries, DC entries equal to previous
+           * block's DC value.  The init routine has already zeroed the
+           * AC entries, so we need only set the DC entries correctly.
+           */
+          for (; xindex < compptr->MCU_width; xindex++) {
+            MCU_bufferblkn = coef->dummy_bufferblkn;
+            MCU_bufferblkn00 = MCU_bufferblkn - 100;
+            blkn++;
+          }
+        }
+      }
+      /* Try to write the MCU. */
+      if (!(*cinfo->entropy->encode_mcu) (cinfo, MCU_buffer)) {
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->mcu_ctr = MCU_col_num;
+        return FALSE;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->mcu_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  coef->iMCU_row_num++;
+  start_iMCU_row(cinfo);
+  return TRUE;
+}
+
+
+/*
+ * Initialize coefficient buffer controller.
+ *
+ * Each passed coefficient array must be the right size for that
+ * coefficient: width_in_blocks wide and height_in_blocks high,
+ * with unitheight at least v_samp_factor.
+ */
+
+LOCAL(void)
+transencode_coef_controller(j_compress_ptr cinfo,
+                            jvirt_barray_ptr *coef_arrays)
+{
+  my_coef_ptr coef;
+  JBLOCKROW buffer;
+  int i;
+
+  coef = (my_coef_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_coef_controller));
+  cinfo->coef = (struct jpeg_c_coef_controller *)coef;
+  coef->pub.start_pass = start_pass_coef;
+  coef->pub.compress_data = compress_output;
+
+  /* Save pointer to virtual arrays */
+  coef->whole_image = coef_arrays;
+
+  /* Allocate and pre-zero space for dummy DCT blocks. */
+  buffer = (JBLOCKROW)
+    (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                C_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
+  jzero_far((void *)buffer, C_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
+  for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
+    coef->dummy_bufferi = buffer + i;
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdapimin.c Added

@@ -0,0 +1,408 @@
+/*
+ * jdapimin.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains application interface code for the decompression half
+ * of the JPEG library.  These are the "minimum" API routines that may be
+ * needed in either the normal full-decompression case or the
+ * transcoding-only case.
+ *
+ * Most of the routines intended to be called directly by an application
+ * are in this file or in jdapistd.c.  But also see jcomapi.c for routines
+ * shared by compression and decompression, and jdtrans.c for the transcoding
+ * case.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdmaster.h"
+#include "jconfigint.h"
+
+
+/*
+ * Initialization of a JPEG decompression object.
+ * The error manager must already be set up (in case memory manager fails).
+ */
+
+GLOBAL(void)
+jpeg_CreateDecompress(j_decompress_ptr cinfo, int version, size_t structsize)
+{
+  int i;
+
+  /* Guard against version mismatches between library and caller. */
+  cinfo->mem = NULL;            /* so jpeg_destroy knows mem mgr not called */
+  if (version != JPEG_LIB_VERSION)
+    ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
+  if (structsize != sizeof(struct jpeg_decompress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE,
+             (int)sizeof(struct jpeg_decompress_struct), (int)structsize);
+
+  /* For debugging purposes, we zero the whole master structure.
+   * But the application has already set the err pointer, and may have set
+   * client_data, so we have to save and restore those fields.
+   * Note: if application hasn't set client_data, tools like Purify may
+   * complain here.
+   */
+  {
+    struct jpeg_error_mgr *err = cinfo->err;
+    void *client_data = cinfo->client_data; /* ignore Purify complaint here */
+    MEMZERO(cinfo, sizeof(struct jpeg_decompress_struct));
+    cinfo->err = err;
+    cinfo->client_data = client_data;
+  }
+  cinfo->is_decompressor = TRUE;
+
+  /* Initialize a memory manager instance for this object */
+  jinit_memory_mgr((j_common_ptr)cinfo);
+
+  /* Zero out pointers to permanent structures. */
+  cinfo->progress = NULL;
+  cinfo->src = NULL;
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++)
+    cinfo->quant_tbl_ptrsi = NULL;
+
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    cinfo->dc_huff_tbl_ptrsi = NULL;
+    cinfo->ac_huff_tbl_ptrsi = NULL;
+  }
+
+  /* Initialize marker processor so application can override methods
+   * for COM, APPn markers before calling jpeg_read_header.
+   */
+  cinfo->marker_list = NULL;
+  jinit_marker_reader(cinfo);
+
+  /* And initialize the overall input controller. */
+  jinit_input_controller(cinfo);
+
+  /* OK, I'm ready */
+  cinfo->global_state = DSTATE_START;
+
+  /* The master struct is used to store extension parameters, so we allocate it
+   * here.
+   */
+  cinfo->master = (struct jpeg_decomp_master *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                sizeof(my_decomp_master));
+  MEMZERO(cinfo->master, sizeof(my_decomp_master));
+}
+
+
+/*
+ * Destruction of a JPEG decompression object
+ */
+
+GLOBAL(void)
+jpeg_destroy_decompress(j_decompress_ptr cinfo)
+{
+  jpeg_destroy((j_common_ptr)cinfo); /* use common routine */
+}
+
+
+/*
+ * Abort processing of a JPEG decompression operation,
+ * but don't destroy the object itself.
+ */
+
+GLOBAL(void)
+jpeg_abort_decompress(j_decompress_ptr cinfo)
+{
+  jpeg_abort((j_common_ptr)cinfo); /* use common routine */
+}
+
+
+/*
+ * Set default decompression parameters.
+ */
+
+LOCAL(void)
+default_decompress_parms(j_decompress_ptr cinfo)
+{
+  /* Guess the input colorspace, and set output colorspace accordingly. */
+  /* (Wish JPEG committee had provided a real way to specify this...) */
+  /* Note application may override our guesses. */
+  switch (cinfo->num_components) {
+  case 1:
+    cinfo->jpeg_color_space = JCS_GRAYSCALE;
+    cinfo->out_color_space = JCS_GRAYSCALE;
+    break;
+
+  case 3:
+    if (cinfo->saw_JFIF_marker) {
+      cinfo->jpeg_color_space = JCS_YCbCr; /* JFIF implies YCbCr */
+    } else if (cinfo->saw_Adobe_marker) {
+      switch (cinfo->Adobe_transform) {
+      case 0:
+        cinfo->jpeg_color_space = JCS_RGB;
+        break;
+      case 1:
+        cinfo->jpeg_color_space = JCS_YCbCr;
+        break;
+      default:
+        WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+        cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+        break;
+      }
+    } else {
+      /* Saw no special markers, try to guess from the component IDs */
+      int cid0 = cinfo->comp_info0.component_id;
+      int cid1 = cinfo->comp_info1.component_id;
+      int cid2 = cinfo->comp_info2.component_id;
+
+      if (cid0 == 1 && cid1 == 2 && cid2 == 3)
+        cinfo->jpeg_color_space = JCS_YCbCr; /* assume JFIF w/out marker */
+      else if (cid0 == 82 && cid1 == 71 && cid2 == 66)
+        cinfo->jpeg_color_space = JCS_RGB; /* ASCII 'R', 'G', 'B' */
+      else {
+        TRACEMS3(cinfo, 1, JTRC_UNKNOWN_IDS, cid0, cid1, cid2);
+        cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+      }
+    }
+    /* Always guess RGB is proper output colorspace. */
+    cinfo->out_color_space = JCS_RGB;
+    break;
+
+  case 4:
+    if (cinfo->saw_Adobe_marker) {
+      switch (cinfo->Adobe_transform) {
+      case 0:
+        cinfo->jpeg_color_space = JCS_CMYK;
+        break;
+      case 2:
+        cinfo->jpeg_color_space = JCS_YCCK;
+        break;
+      default:
+        WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+        cinfo->jpeg_color_space = JCS_YCCK; /* assume it's YCCK */
+        break;
+      }
+    } else {
+      /* No special markers, assume straight CMYK. */
+      cinfo->jpeg_color_space = JCS_CMYK;
+    }
+    cinfo->out_color_space = JCS_CMYK;
+    break;
+
+  default:
+    cinfo->jpeg_color_space = JCS_UNKNOWN;
+    cinfo->out_color_space = JCS_UNKNOWN;
+    break;
+  }
+
+  /* Set defaults for other decompression parameters. */
+  cinfo->scale_num = 1;         /* 1:1 scaling */
+  cinfo->scale_denom = 1;
+  cinfo->output_gamma = 1.0;
+  cinfo->buffered_image = FALSE;
+  cinfo->raw_data_out = FALSE;
+  cinfo->dct_method = JDCT_DEFAULT;
+  cinfo->do_fancy_upsampling = TRUE;
+  cinfo->do_block_smoothing = TRUE;
+  cinfo->quantize_colors = FALSE;
+  /* We set these in case application only sets quantize_colors. */
+  cinfo->dither_mode = JDITHER_FS;
+#ifdef QUANT_2PASS_SUPPORTED
+  cinfo->two_pass_quantize = TRUE;
+#else
+  cinfo->two_pass_quantize = FALSE;
+#endif
+  cinfo->desired_number_of_colors = 256;
+  cinfo->colormap = NULL;
+  /* Initialize for no mode change in buffered-image mode. */
+  cinfo->enable_1pass_quant = FALSE;
+  cinfo->enable_external_quant = FALSE;
+  cinfo->enable_2pass_quant = FALSE;
+}
+
+
+/*
+ * Decompression startup: read start of JPEG datastream to see what's there.
+ * Need only initialize JPEG object and supply a data source before calling.
+ *
+ * This routine will read as far as the first SOS marker (ie, actual start of
+ * compressed data), and will save all tables and parameters in the JPEG
+ * object.  It will also initialize the decompression parameters to default
+ * values, and finally return JPEG_HEADER_OK.  On return, the application may
+ * adjust the decompression parameters and then call jpeg_start_decompress.
+ * (Or, if the application only wanted to determine the image parameters,
+ * the data need not be decompressed.  In that case, call jpeg_abort or
+ * jpeg_destroy to release any temporary space.)
+ * If an abbreviated (tables only) datastream is presented, the routine will
+ * return JPEG_HEADER_TABLES_ONLY upon reaching EOI.  The application may then
+ * re-use the JPEG object to read the abbreviated image datastream(s).
+ * It is unnecessary (but OK) to call jpeg_abort in this case.
+ * The JPEG_SUSPENDED return code only occurs if the data source module
+ * requests suspension of the decompressor.  In this case the application
+ * should load more source data and then re-call jpeg_read_header to resume
+ * processing.
+ * If a non-suspending data source is used and require_image is TRUE, then the
+ * return code need not be inspected since only JPEG_HEADER_OK is possible.
+ *
+ * This routine is now just a front end to jpeg_consume_input, with some
+ * extra error checking.
+ */
+
+GLOBAL(int)
+jpeg_read_header(j_decompress_ptr cinfo, boolean require_image)
+{
+  int retcode;
+
+  if (cinfo->global_state != DSTATE_START &&
+      cinfo->global_state != DSTATE_INHEADER)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  retcode = jpeg_consume_input(cinfo);
+
+  switch (retcode) {
+  case JPEG_REACHED_SOS:
+    retcode = JPEG_HEADER_OK;
+    break;
+  case JPEG_REACHED_EOI:
+    if (require_image)          /* Complain if application wanted an image */
+      ERREXIT(cinfo, JERR_NO_IMAGE);
+    /* Reset to start state; it would be safer to require the application to
+     * call jpeg_abort, but we can't change it now for compatibility reasons.
+     * A side effect is to free any temporary memory (there shouldn't be any).
+     */
+    jpeg_abort((j_common_ptr)cinfo); /* sets state = DSTATE_START */
+    retcode = JPEG_HEADER_TABLES_ONLY;
+    break;
+  case JPEG_SUSPENDED:
+    /* no work */
+    break;
+  }
+
+  return retcode;
+}
+
+
+/*
+ * Consume data in advance of what the decompressor requires.
+ * This can be called at any time once the decompressor object has
+ * been created and a data source has been set up.
+ *
+ * This routine is essentially a state machine that handles a couple
+ * of critical state-transition actions, namely initial setup and
+ * transition from header scanning to ready-for-start_decompress.
+ * All the actual input is done via the input controller's consume_input
+ * method.
+ */
+
+GLOBAL(int)
+jpeg_consume_input(j_decompress_ptr cinfo)
+{
+  int retcode = JPEG_SUSPENDED;
+
+  /* NB: every possible DSTATE value should be listed in this switch */
+  switch (cinfo->global_state) {
+  case DSTATE_START:
+    /* Start-of-datastream actions: reset appropriate modules */
+    (*cinfo->inputctl->reset_input_controller) (cinfo);
+    /* Initialize application's data source module */
+    (*cinfo->src->init_source) (cinfo);
+    cinfo->global_state = DSTATE_INHEADER;
+    FALLTHROUGH                 /*FALLTHROUGH*/
+  case DSTATE_INHEADER:
+    retcode = (*cinfo->inputctl->consume_input) (cinfo);
+    if (retcode == JPEG_REACHED_SOS) { /* Found SOS, prepare to decompress */
+      /* Set up default parameters based on header data */
+      default_decompress_parms(cinfo);
+      /* Set global state: ready for start_decompress */
+      cinfo->global_state = DSTATE_READY;
+    }
+    break;
+  case DSTATE_READY:
+    /* Can't advance past first SOS until start_decompress is called */
+    retcode = JPEG_REACHED_SOS;
+    break;
+  case DSTATE_PRELOAD:
+  case DSTATE_PRESCAN:
+  case DSTATE_SCANNING:
+  case DSTATE_RAW_OK:
+  case DSTATE_BUFIMAGE:
+  case DSTATE_BUFPOST:
+  case DSTATE_STOPPING:
+    retcode = (*cinfo->inputctl->consume_input) (cinfo);
+    break;
+  default:
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  }
+  return retcode;
+}
+
+
+/*
+ * Have we finished reading the input file?
+ */
+
+GLOBAL(boolean)
+jpeg_input_complete(j_decompress_ptr cinfo)
+{
+  /* Check for valid jpeg object */
+  if (cinfo->global_state < DSTATE_START ||
+      cinfo->global_state > DSTATE_STOPPING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return cinfo->inputctl->eoi_reached;
+}
+
+
+/*
+ * Is there more than one scan?
+ */
+
+GLOBAL(boolean)
+jpeg_has_multiple_scans(j_decompress_ptr cinfo)
+{
+  /* Only valid after jpeg_read_header completes */
+  if (cinfo->global_state < DSTATE_READY ||
+      cinfo->global_state > DSTATE_STOPPING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return cinfo->inputctl->has_multiple_scans;
+}
+
+
+/*
+ * Finish JPEG decompression.
+ *
+ * This will normally just verify the file trailer and release temp storage.
+ *
+ * Returns FALSE if suspended.  The return value need be inspected only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(boolean)
+jpeg_finish_decompress(j_decompress_ptr cinfo)
+{
+  if ((cinfo->global_state == DSTATE_SCANNING ||
+       cinfo->global_state == DSTATE_RAW_OK) && !cinfo->buffered_image) {
+    /* Terminate final pass of non-buffered mode */
+    if (cinfo->output_scanline < cinfo->output_height)
+      ERREXIT(cinfo, JERR_TOO_LITTLE_DATA);
+    (*cinfo->master->finish_output_pass) (cinfo);
+    cinfo->global_state = DSTATE_STOPPING;
+  } else if (cinfo->global_state == DSTATE_BUFIMAGE) {
+    /* Finishing after a buffered-image operation */
+    cinfo->global_state = DSTATE_STOPPING;
+  } else if (cinfo->global_state != DSTATE_STOPPING) {
+    /* STOPPING = repeat call after a suspension, anything else is error */
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  }
+  /* Read until EOI */
+  while (!cinfo->inputctl->eoi_reached) {
+    if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
+      return FALSE;             /* Suspend, come back later */
+  }
+  /* Do final cleanup */
+  (*cinfo->src->term_source) (cinfo);
+  /* We can use jpeg_abort to release memory and reset global_state */
+  jpeg_abort((j_common_ptr)cinfo);
+  return TRUE;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdapistd.c Added

@@ -0,0 +1,674 @@
+/*
+ * jdapistd.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2015-2020, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains application interface code for the decompression half
+ * of the JPEG library.  These are the "standard" API routines that are
+ * used in the normal full-decompression case.  They are not used by a
+ * transcoding-only application.  Note that if an application links in
+ * jpeg_start_decompress, it will end up linking in the entire decompressor.
+ * We thus must separate this file from jdapimin.c to avoid linking the
+ * whole decompression library into a transcoder.
+ */
+
+#include "jinclude.h"
+#include "jdmainct.h"
+#include "jdcoefct.h"
+#include "jdmaster.h"
+#include "jdmerge.h"
+#include "jdsample.h"
+#include "jmemsys.h"
+
+/* Forward declarations */
+LOCAL(boolean) output_pass_setup(j_decompress_ptr cinfo);
+
+
+/*
+ * Decompression initialization.
+ * jpeg_read_header must be completed before calling this.
+ *
+ * If a multipass operating mode was selected, this will do all but the
+ * last pass, and thus may take a great deal of time.
+ *
+ * Returns FALSE if suspended.  The return value need be inspected only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(boolean)
+jpeg_start_decompress(j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state == DSTATE_READY) {
+    /* First call: initialize master control, select active modules */
+    jinit_master_decompress(cinfo);
+    if (cinfo->buffered_image) {
+      /* No more work here; expecting jpeg_start_output next */
+      cinfo->global_state = DSTATE_BUFIMAGE;
+      return TRUE;
+    }
+    cinfo->global_state = DSTATE_PRELOAD;
+  }
+  if (cinfo->global_state == DSTATE_PRELOAD) {
+    /* If file has multiple scans, absorb them all into the coef buffer */
+    if (cinfo->inputctl->has_multiple_scans) {
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+      for (;;) {
+        int retcode;
+        /* Call progress monitor hook if present */
+        if (cinfo->progress != NULL)
+          (*cinfo->progress->progress_monitor) ((j_common_ptr)cinfo);
+        /* Absorb some more input */
+        retcode = (*cinfo->inputctl->consume_input) (cinfo);
+        if (retcode == JPEG_SUSPENDED)
+          return FALSE;
+        if (retcode == JPEG_REACHED_EOI)
+          break;
+        /* Advance progress counter if appropriate */
+        if (cinfo->progress != NULL &&
+            (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+          if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+            /* jdmaster underestimated number of scans; ratchet up one scan */
+            cinfo->progress->pass_limit += (long)cinfo->total_iMCU_rows;
+          }
+        }
+      }
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+    }
+    cinfo->output_scan_number = cinfo->input_scan_number;
+  } else if (cinfo->global_state != DSTATE_PRESCAN)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Perform any dummy output passes, and set up for the final pass */
+  return output_pass_setup(cinfo);
+}
+
+
+/*
+ * Set up for an output pass, and perform any dummy pass(es) needed.
+ * Common subroutine for jpeg_start_decompress and jpeg_start_output.
+ * Entry: global_state = DSTATE_PRESCAN only if previously suspended.
+ * Exit: If done, returns TRUE and sets global_state for proper output mode.
+ *       If suspended, returns FALSE and sets global_state = DSTATE_PRESCAN.
+ */
+
+LOCAL(boolean)
+output_pass_setup(j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state != DSTATE_PRESCAN) {
+    /* First call: do pass setup */
+    (*cinfo->master->prepare_for_output_pass) (cinfo);
+    cinfo->output_scanline = 0;
+    cinfo->global_state = DSTATE_PRESCAN;
+  }
+  /* Loop over any required dummy passes */
+  while (cinfo->master->is_dummy_pass) {
+#ifdef QUANT_2PASS_SUPPORTED
+    /* Crank through the dummy pass */
+    while (cinfo->output_scanline < cinfo->output_height) {
+      JDIMENSION last_scanline;
+      /* Call progress monitor hook if present */
+      if (cinfo->progress != NULL) {
+        cinfo->progress->pass_counter = (long)cinfo->output_scanline;
+        cinfo->progress->pass_limit = (long)cinfo->output_height;
+        (*cinfo->progress->progress_monitor) ((j_common_ptr)cinfo);
+      }
+      /* Process some data */
+      last_scanline = cinfo->output_scanline;
+      (*cinfo->main->process_data) (cinfo, (JSAMPARRAY)NULL,
+                                    &cinfo->output_scanline, (JDIMENSION)0);
+      if (cinfo->output_scanline == last_scanline)
+        return FALSE;           /* No progress made, must suspend */
+    }
+    /* Finish up dummy pass, and set up for another one */
+    (*cinfo->master->finish_output_pass) (cinfo);
+    (*cinfo->master->prepare_for_output_pass) (cinfo);
+    cinfo->output_scanline = 0;
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif /* QUANT_2PASS_SUPPORTED */
+  }
+  /* Ready for application to drive output pass through
+   * jpeg_read_scanlines or jpeg_read_raw_data.
+   */
+  cinfo->global_state = cinfo->raw_data_out ? DSTATE_RAW_OK : DSTATE_SCANNING;
+  return TRUE;
+}
+
+
+/*
+ * Enable partial scanline decompression
+ *
+ * Must be called after jpeg_start_decompress() and before any calls to
+ * jpeg_read_scanlines() or jpeg_skip_scanlines().
+ *
+ * Refer to libjpeg.txt for more information.
+ */
+
+GLOBAL(void)
+jpeg_crop_scanline(j_decompress_ptr cinfo, JDIMENSION *xoffset,
+                   JDIMENSION *width)
+{
+  int ci, align, orig_downsampled_width;
+  JDIMENSION input_xoffset;
+  boolean reinit_upsampler = FALSE;
+  jpeg_component_info *compptr;
+
+  if (cinfo->global_state != DSTATE_SCANNING || cinfo->output_scanline != 0)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (!xoffset || !width)
+    ERREXIT(cinfo, JERR_BAD_CROP_SPEC);
+
+  /* xoffset and width must fall within the output image dimensions. */
+  if (*width == 0 || *xoffset + *width > cinfo->output_width)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+
+  /* No need to do anything if the caller wants the entire width. */
+  if (*width == cinfo->output_width)
+    return;
+
+  /* Ensuring the proper alignment of xoffset is tricky.  At minimum, it
+   * must align with an MCU boundary, because:
+   *
+   *   (1) The IDCT is performed in blocks, and it is not feasible to modify
+   *       the algorithm so that it can transform partial blocks.
+   *   (2) Because of the SIMD extensions, any input buffer passed to the
+   *       upsampling and color conversion routines must be aligned to the
+   *       SIMD word size (for instance, 128-bit in the case of SSE2.)  The
+   *       easiest way to accomplish this without copying data is to ensure
+   *       that upsampling and color conversion begin at the start of the
+   *       first MCU column that will be inverse transformed.
+   *
+   * In practice, we actually impose a stricter alignment requirement.  We
+   * require that xoffset be a multiple of the maximum MCU column width of all
+   * of the components (the "iMCU column width.")  This is to simplify the
+   * single-pass decompression case, allowing us to use the same MCU column
+   * width for all of the components.
+   */
+  if (cinfo->comps_in_scan == 1 && cinfo->num_components == 1)
+    align = cinfo->_min_DCT_scaled_size;
+  else
+    align = cinfo->_min_DCT_scaled_size * cinfo->max_h_samp_factor;
+
+  /* Adjust xoffset to the nearest iMCU boundary <= the requested value */
+  input_xoffset = *xoffset;
+  *xoffset = (input_xoffset / align) * align;
+
+  /* Adjust the width so that the right edge of the output image is as
+   * requested (only the left edge is altered.)  It is important that calling
+   * programs check this value after this function returns, so that they can
+   * allocate an output buffer with the appropriate size.
+   */
+  *width = *width + input_xoffset - *xoffset;
+  cinfo->output_width = *width;
+
+  /* Set the first and last iMCU columns that we must decompress.  These values
+   * will be used in single-scan decompressions.
+   */
+  cinfo->master->first_iMCU_col = (JDIMENSION)(long)(*xoffset) / (long)align;
+  cinfo->master->last_iMCU_col =
+    (JDIMENSION)jdiv_round_up((long)(*xoffset + cinfo->output_width),
+                              (long)align) - 1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    int hsf = (cinfo->comps_in_scan == 1 && cinfo->num_components == 1) ?
+              1 : compptr->h_samp_factor;
+
+    /* Set downsampled_width to the new output width. */
+    orig_downsampled_width = compptr->downsampled_width;
+    compptr->downsampled_width =
+      (JDIMENSION)jdiv_round_up((long)(cinfo->output_width *
+                                       compptr->h_samp_factor),
+                                (long)cinfo->max_h_samp_factor);
+    if (compptr->downsampled_width < 2 && orig_downsampled_width >= 2)
+      reinit_upsampler = TRUE;
+
+    /* Set the first and last iMCU columns that we must decompress.  These
+     * values will be used in multi-scan decompressions.
+     */
+    cinfo->master->first_MCU_colci =
+      (JDIMENSION)(long)(*xoffset * hsf) / (long)align;
+    cinfo->master->last_MCU_colci =
+      (JDIMENSION)jdiv_round_up((long)((*xoffset + cinfo->output_width) * hsf),
+                                (long)align) - 1;
+  }
+
+  if (reinit_upsampler) {
+    cinfo->master->jinit_upsampler_no_alloc = TRUE;
+    jinit_upsampler(cinfo);
+    cinfo->master->jinit_upsampler_no_alloc = FALSE;
+  }
+}
+
+
+/*
+ * Read some scanlines of data from the JPEG decompressor.
+ *
+ * The return value will be the number of lines actually read.
+ * This may be less than the number requested in several cases,
+ * including bottom of image, data source suspension, and operating
+ * modes that emit multiple scanlines at a time.
+ *
+ * Note: we warn about excess calls to jpeg_read_scanlines() since
+ * this likely signals an application programmer error.  However,
+ * an oversize buffer (max_lines > scanlines remaining) is not an error.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_read_scanlines(j_decompress_ptr cinfo, JSAMPARRAY scanlines,
+                    JDIMENSION max_lines)
+{
+  JDIMENSION row_ctr;
+
+  if (cinfo->global_state != DSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->output_scanline >= cinfo->output_height) {
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+    return 0;
+  }
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long)cinfo->output_scanline;
+    cinfo->progress->pass_limit = (long)cinfo->output_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr)cinfo);
+  }
+
+  /* Process some data */
+  row_ctr = 0;
+  (*cinfo->main->process_data) (cinfo, scanlines, &row_ctr, max_lines);
+  cinfo->output_scanline += row_ctr;
+  return row_ctr;
+}
+
+
+/* Dummy color convert function used by jpeg_skip_scanlines() */
+LOCAL(void)
+noop_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+             JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+
+/* Dummy quantize function used by jpeg_skip_scanlines() */
+LOCAL(void)
+noop_quantize(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+              JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+
+/*
+ * In some cases, it is best to call jpeg_read_scanlines() and discard the
+ * output, rather than skipping the scanlines, because this allows us to
+ * maintain the internal state of the context-based upsampler.  In these cases,
+ * we set up and tear down a dummy color converter in order to avoid valgrind
+ * errors and to achieve the best possible performance.
+ */
+
+LOCAL(void)
+read_and_discard_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
+{
+  JDIMENSION n;
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+  JSAMPLE dummy_sample1 = { 0 };
+  JSAMPROW dummy_row = dummy_sample;
+  JSAMPARRAY scanlines = NULL;
+  void (*color_convert) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows) = NULL;
+  void (*color_quantize) (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                          JSAMPARRAY output_buf, int num_rows) = NULL;
+
+  if (cinfo->cconvert && cinfo->cconvert->color_convert) {
+    color_convert = cinfo->cconvert->color_convert;
+    cinfo->cconvert->color_convert = noop_convert;
+    /* This just prevents UBSan from complaining about adding 0 to a NULL
+     * pointer.  The pointer isn't actually used.
+     */
+    scanlines = &dummy_row;
+  }
+
+  if (cinfo->cquantize && cinfo->cquantize->color_quantize) {
+    color_quantize = cinfo->cquantize->color_quantize;
+    cinfo->cquantize->color_quantize = noop_quantize;
+  }
+
+  if (master->using_merged_upsample && cinfo->max_v_samp_factor == 2) {
+    my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+    scanlines = &upsample->spare_row;
+  }
+
+  for (n = 0; n < num_lines; n++)
+    jpeg_read_scanlines(cinfo, scanlines, 1);
+
+  if (color_convert)
+    cinfo->cconvert->color_convert = color_convert;
+
+  if (color_quantize)
+    cinfo->cquantize->color_quantize = color_quantize;
+}
+
+
+/*
+ * Called by jpeg_skip_scanlines().  This partially skips a decompress block by
+ * incrementing the rowgroup counter.
+ */
+
+LOCAL(void)
+increment_simple_rowgroup_ctr(j_decompress_ptr cinfo, JDIMENSION rows)
+{
+  JDIMENSION rows_left;
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+
+  if (master->using_merged_upsample && cinfo->max_v_samp_factor == 2) {
+    read_and_discard_scanlines(cinfo, rows);
+    return;
+  }
+
+  /* Increment the counter to the next row group after the skipped rows. */
+  main_ptr->rowgroup_ctr += rows / cinfo->max_v_samp_factor;
+
+  /* Partially skipping a row group would involve modifying the internal state
+   * of the upsampler, so read the remaining rows into a dummy buffer instead.
+   */
+  rows_left = rows % cinfo->max_v_samp_factor;
+  cinfo->output_scanline += rows - rows_left;
+
+  read_and_discard_scanlines(cinfo, rows_left);
+}
+
+/*
+ * Skips some scanlines of data from the JPEG decompressor.
+ *
+ * The return value will be the number of lines actually skipped.  If skipping
+ * num_lines would move beyond the end of the image, then the actual number of
+ * lines remaining in the image is returned.  Otherwise, the return value will
+ * be equal to num_lines.
+ *
+ * Refer to libjpeg.txt for more information.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
+  JDIMENSION i, x;
+  int y;
+  JDIMENSION lines_per_iMCU_row, lines_left_in_iMCU_row, lines_after_iMCU_row;
+  JDIMENSION lines_to_skip, lines_to_read;
+
+  /* Two-pass color quantization is not supported. */
+  if (cinfo->quantize_colors && cinfo->two_pass_quantize)
+    ERREXIT(cinfo, JERR_NOTIMPL);
+
+  if (cinfo->global_state != DSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Do not skip past the bottom of the image. */
+  if (cinfo->output_scanline + num_lines >= cinfo->output_height) {
+    num_lines = cinfo->output_height - cinfo->output_scanline;
+    cinfo->output_scanline = cinfo->output_height;
+    (*cinfo->inputctl->finish_input_pass) (cinfo);
+    cinfo->inputctl->eoi_reached = TRUE;
+    return num_lines;
+  }
+
+  if (num_lines == 0)
+    return 0;
+
+  lines_per_iMCU_row = cinfo->_min_DCT_scaled_size * cinfo->max_v_samp_factor;
+  lines_left_in_iMCU_row =
+    (lines_per_iMCU_row - (cinfo->output_scanline % lines_per_iMCU_row)) %
+    lines_per_iMCU_row;
+  lines_after_iMCU_row = num_lines - lines_left_in_iMCU_row;
+
+  /* Skip the lines remaining in the current iMCU row.  When upsampling
+   * requires context rows, we need the previous and next rows in order to read
+   * the current row.  This adds some complexity.
+   */
+  if (cinfo->upsample->need_context_rows) {
+    /* If the skipped lines would not move us past the current iMCU row, we
+     * read the lines and ignore them.  There might be a faster way of doing
+     * this, but we are facing increasing complexity for diminishing returns.
+     * The increasing complexity would be a by-product of meddling with the
+     * state machine used to skip context rows.  Near the end of an iMCU row,
+     * the next iMCU row may have already been entropy-decoded.  In this unique
+     * case, we will read the next iMCU row if we cannot skip past it as well.
+     */
+    if ((num_lines < lines_left_in_iMCU_row + 1) ||
+        (lines_left_in_iMCU_row <= 1 && main_ptr->buffer_full &&
+         lines_after_iMCU_row < lines_per_iMCU_row + 1)) {
+      read_and_discard_scanlines(cinfo, num_lines);
+      return num_lines;
+    }
+
+    /* If the next iMCU row has already been entropy-decoded, make sure that
+     * we do not skip too far.
+     */
+    if (lines_left_in_iMCU_row <= 1 && main_ptr->buffer_full) {
+      cinfo->output_scanline += lines_left_in_iMCU_row + lines_per_iMCU_row;
+      lines_after_iMCU_row -= lines_per_iMCU_row;
+    } else {
+      cinfo->output_scanline += lines_left_in_iMCU_row;
+    }
+
+    /* If we have just completed the first block, adjust the buffer pointers */
+    if (main_ptr->iMCU_row_ctr == 0 ||
+        (main_ptr->iMCU_row_ctr == 1 && lines_left_in_iMCU_row > 2))
+      set_wraparound_pointers(cinfo);
+    main_ptr->buffer_full = FALSE;
+    main_ptr->rowgroup_ctr = 0;
+    main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
+    if (!master->using_merged_upsample) {
+      upsample->next_row_out = cinfo->max_v_samp_factor;
+      upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+    }
+  }
+
+  /* Skipping is much simpler when context rows are not required. */
+  else {
+    if (num_lines < lines_left_in_iMCU_row) {
+      increment_simple_rowgroup_ctr(cinfo, num_lines);
+      return num_lines;
+    } else {
+      cinfo->output_scanline += lines_left_in_iMCU_row;
+      main_ptr->buffer_full = FALSE;
+      main_ptr->rowgroup_ctr = 0;
+      if (!master->using_merged_upsample) {
+        upsample->next_row_out = cinfo->max_v_samp_factor;
+        upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+      }
+    }
+  }
+
+  /* Calculate how many full iMCU rows we can skip. */
+  if (cinfo->upsample->need_context_rows)
+    lines_to_skip = ((lines_after_iMCU_row - 1) / lines_per_iMCU_row) *
+                    lines_per_iMCU_row;
+  else
+    lines_to_skip = (lines_after_iMCU_row / lines_per_iMCU_row) *
+                    lines_per_iMCU_row;
+  /* Calculate the number of lines that remain to be skipped after skipping all
+   * of the full iMCU rows that we can.  We will not read these lines unless we
+   * have to.
+   */
+  lines_to_read = lines_after_iMCU_row - lines_to_skip;
+
+  /* For images requiring multiple scans (progressive, non-interleaved, etc.),
+   * all of the entropy decoding occurs in jpeg_start_decompress(), assuming
+   * that the input data source is non-suspending.  This makes skipping easy.
+   */
+  if (cinfo->inputctl->has_multiple_scans) {
+    if (cinfo->upsample->need_context_rows) {
+      cinfo->output_scanline += lines_to_skip;
+      cinfo->output_iMCU_row += lines_to_skip / lines_per_iMCU_row;
+      main_ptr->iMCU_row_ctr += lines_to_skip / lines_per_iMCU_row;
+      /* It is complex to properly move to the middle of a context block, so
+       * read the remaining lines instead of skipping them.
+       */
+      read_and_discard_scanlines(cinfo, lines_to_read);
+    } else {
+      cinfo->output_scanline += lines_to_skip;
+      cinfo->output_iMCU_row += lines_to_skip / lines_per_iMCU_row;
+      increment_simple_rowgroup_ctr(cinfo, lines_to_read);
+    }
+    if (!master->using_merged_upsample)
+      upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+    return num_lines;
+  }
+
+  /* Skip the iMCU rows that we can safely skip. */
+  for (i = 0; i < lines_to_skip; i += lines_per_iMCU_row) {
+    for (y = 0; y < coef->MCU_rows_per_iMCU_row; y++) {
+      for (x = 0; x < cinfo->MCUs_per_row; x++) {
+        /* Calling decode_mcu() with a NULL pointer causes it to discard the
+         * decoded coefficients.  This is ~5% faster for large subsets, but
+         * it's tough to tell a difference for smaller images.
+         */
+        if (!cinfo->entropy->insufficient_data)
+          cinfo->master->last_good_iMCU_row = cinfo->input_iMCU_row;
+        (*cinfo->entropy->decode_mcu) (cinfo, NULL);
+      }
+    }
+    cinfo->input_iMCU_row++;
+    cinfo->output_iMCU_row++;
+    if (cinfo->input_iMCU_row < cinfo->total_iMCU_rows)
+      start_iMCU_row(cinfo);
+    else
+      (*cinfo->inputctl->finish_input_pass) (cinfo);
+  }
+  cinfo->output_scanline += lines_to_skip;
+
+  if (cinfo->upsample->need_context_rows) {
+    /* Context-based upsampling keeps track of iMCU rows. */
+    main_ptr->iMCU_row_ctr += lines_to_skip / lines_per_iMCU_row;
+
+    /* It is complex to properly move to the middle of a context block, so
+     * read the remaining lines instead of skipping them.
+     */
+    read_and_discard_scanlines(cinfo, lines_to_read);
+  } else {
+    increment_simple_rowgroup_ctr(cinfo, lines_to_read);
+  }
+
+  /* Since skipping lines involves skipping the upsampling step, the value of
+   * "rows_to_go" will become invalid unless we set it here.  NOTE: This is a
+   * bit odd, since "rows_to_go" seems to be redundantly keeping track of
+   * output_scanline.
+   */
+  if (!master->using_merged_upsample)
+    upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+
+  /* Always skip the requested number of lines. */
+  return num_lines;
+}
+
+/*
+ * Alternate entry point to read raw data.
+ * Processes exactly one iMCU row per call, unless suspended.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_read_raw_data(j_decompress_ptr cinfo, JSAMPIMAGE data,
+                   JDIMENSION max_lines)
+{
+  JDIMENSION lines_per_iMCU_row;
+
+  if (cinfo->global_state != DSTATE_RAW_OK)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->output_scanline >= cinfo->output_height) {
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+    return 0;
+  }
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long)cinfo->output_scanline;
+    cinfo->progress->pass_limit = (long)cinfo->output_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr)cinfo);
+  }
+
+  /* Verify that at least one iMCU row can be returned. */
+  lines_per_iMCU_row = cinfo->max_v_samp_factor * cinfo->_min_DCT_scaled_size;
+  if (max_lines < lines_per_iMCU_row)
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* Decompress directly into user's buffer. */
+  if (!(*cinfo->coef->decompress_data) (cinfo, data))
+    return 0;                   /* suspension forced, can do nothing more */
+
+  /* OK, we processed one iMCU row. */
+  cinfo->output_scanline += lines_per_iMCU_row;
+  return lines_per_iMCU_row;
+}
+
+
+/* Additional entry points for buffered-image mode. */
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+
+/*
+ * Initialize for an output pass in buffered-image mode.
+ */
+
+GLOBAL(boolean)
+jpeg_start_output(j_decompress_ptr cinfo, int scan_number)
+{
+  if (cinfo->global_state != DSTATE_BUFIMAGE &&
+      cinfo->global_state != DSTATE_PRESCAN)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Limit scan number to valid range */
+  if (scan_number <= 0)
+    scan_number = 1;
+  if (cinfo->inputctl->eoi_reached && scan_number > cinfo->input_scan_number)
+    scan_number = cinfo->input_scan_number;
+  cinfo->output_scan_number = scan_number;
+  /* Perform any dummy output passes, and set up for the real pass */
+  return output_pass_setup(cinfo);
+}
+
+
+/*
+ * Finish up after an output pass in buffered-image mode.
+ *
+ * Returns FALSE if suspended.  The return value need be inspected only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(boolean)
+jpeg_finish_output(j_decompress_ptr cinfo)
+{
+  if ((cinfo->global_state == DSTATE_SCANNING ||
+       cinfo->global_state == DSTATE_RAW_OK) && cinfo->buffered_image) {
+    /* Terminate this pass. */
+    /* We do not require the whole pass to have been completed. */
+    (*cinfo->master->finish_output_pass) (cinfo);
+    cinfo->global_state = DSTATE_BUFPOST;
+  } else if (cinfo->global_state != DSTATE_BUFPOST) {
+    /* BUFPOST = repeat call after a suspension, anything else is error */
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  }
+  /* Read markers looking for SOS or EOI */
+  while (cinfo->input_scan_number <= cinfo->output_scan_number &&
+         !cinfo->inputctl->eoi_reached) {
+    if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
+      return FALSE;             /* Suspend, come back later */
+  }
+  cinfo->global_state = DSTATE_BUFIMAGE;
+  return TRUE;
+}
+
+#endif /* D_MULTISCAN_FILES_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdarith.c Added

@@ -0,0 +1,782 @@
+/*
+ * jdarith.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Developed 1997-2015 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015-2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains portable arithmetic entropy encoding routines for JPEG
+ * (implementing Recommendation ITU-T T.81 | ISO/IEC 10918-1).
+ *
+ * Both sequential and progressive modes are supported in this single module.
+ *
+ * Suspension is not currently supported in this module.
+ *
+ * NOTE: All referenced figures are from
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+#define NEG_1  ((unsigned int)-1)
+
+
+/* Expanded entropy decoder object for arithmetic decoding. */
+
+typedef struct {
+  struct jpeg_entropy_decoder pub; /* public fields */
+
+  JLONG c;       /* C register, base of coding interval + input bit buffer */
+  JLONG a;               /* A register, normalized size of coding interval */
+  int ct;     /* bit shift counter, # of bits left in bit buffer part of C */
+                                                         /* init: ct = -16 */
+                                                         /* run: ct = 0..7 */
+                                                         /* error: ct = -1 */
+  int last_dc_valMAX_COMPS_IN_SCAN; /* last DC coef for each component */
+  int dc_contextMAX_COMPS_IN_SCAN; /* context index for DC conditioning */
+
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+
+  /* Pointers to statistics areas (these workspaces have image lifespan) */
+  unsigned char *dc_statsNUM_ARITH_TBLS;
+  unsigned char *ac_statsNUM_ARITH_TBLS;
+
+  /* Statistics bin for coding with fixed probability 0.5 */
+  unsigned char fixed_bin4;
+} arith_entropy_decoder;
+
+typedef arith_entropy_decoder *arith_entropy_ptr;
+
+/* The following two definitions specify the allocation chunk size
+ * for the statistics area.
+ * According to sections F.1.4.4.1.3 and F.1.4.4.2, we need at least
+ * 49 statistics bins for DC, and 245 statistics bins for AC coding.
+ *
+ * We use a compact representation with 1 byte per statistics bin,
+ * thus the numbers directly represent byte sizes.
+ * This 1 byte per statistics bin contains the meaning of the MPS
+ * (more probable symbol) in the highest bit (mask 0x80), and the
+ * index into the probability estimation state machine table
+ * in the lower bits (mask 0x7F).
+ */
+
+#define DC_STAT_BINS  64
+#define AC_STAT_BINS  256
+
+
+LOCAL(int)
+get_byte(j_decompress_ptr cinfo)
+/* Read next input byte; we do not support suspension in this module. */
+{
+  struct jpeg_source_mgr *src = cinfo->src;
+
+  if (src->bytes_in_buffer == 0)
+    if (!(*src->fill_input_buffer) (cinfo))
+      ERREXIT(cinfo, JERR_CANT_SUSPEND);
+  src->bytes_in_buffer--;
+  return *src->next_input_byte++;
+}
+
+
+/*
+ * The core arithmetic decoding routine (common in JPEG and JBIG).
+ * This needs to go as fast as possible.
+ * Machine-dependent optimization facilities
+ * are not utilized in this portable implementation.
+ * However, this code should be fairly efficient and
+ * may be a good base for further optimizations anyway.
+ *
+ * Return value is 0 or 1 (binary decision).
+ *
+ * Note: I've changed the handling of the code base & bit
+ * buffer register C compared to other implementations
+ * based on the standards layout & procedures.
+ * While it also contains both the actual base of the
+ * coding interval (16 bits) and the next-bits buffer,
+ * the cut-point between these two parts is floating
+ * (instead of fixed) with the bit shift counter CT.
+ * Thus, we also need only one (variable instead of
+ * fixed size) shift for the LPS/MPS decision, and
+ * we can do away with any renormalization update
+ * of C (except for new data insertion, of course).
+ *
+ * I've also introduced a new scheme for accessing
+ * the probability estimation state machine table,
+ * derived from Markus Kuhn's JBIG implementation.
+ */
+
+LOCAL(int)
+arith_decode(j_decompress_ptr cinfo, unsigned char *st)
+{
+  register arith_entropy_ptr e = (arith_entropy_ptr)cinfo->entropy;
+  register unsigned char nl, nm;
+  register JLONG qe, temp;
+  register int sv, data;
+
+  /* Renormalization & data input per section D.2.6 */
+  while (e->a < 0x8000L) {
+    if (--e->ct < 0) {
+      /* Need to fetch next data byte */
+      if (cinfo->unread_marker)
+        data = 0;               /* stuff zero data */
+      else {
+        data = get_byte(cinfo); /* read next input byte */
+        if (data == 0xFF) {     /* zero stuff or marker code */
+          do data = get_byte(cinfo);
+          while (data == 0xFF); /* swallow extra 0xFF bytes */
+          if (data == 0)
+            data = 0xFF;        /* discard stuffed zero byte */
+          else {
+            /* Note: Different from the Huffman decoder, hitting
+             * a marker while processing the compressed data
+             * segment is legal in arithmetic coding.
+             * The convention is to supply zero data
+             * then until decoding is complete.
+             */
+            cinfo->unread_marker = data;
+            data = 0;
+          }
+        }
+      }
+      e->c = (e->c << 8) | data; /* insert data into C register */
+      if ((e->ct += 8) < 0)      /* update bit shift counter */
+        /* Need more initial bytes */
+        if (++e->ct == 0)
+          /* Got 2 initial bytes -> re-init A and exit loop */
+          e->a = 0x8000L; /* => e->a = 0x10000L after loop exit */
+    }
+    e->a <<= 1;
+  }
+
+  /* Fetch values from our compact representation of Table D.2:
+   * Qe values and probability estimation state machine
+   */
+  sv = *st;
+  qe = jpeg_aritabsv & 0x7F;  /* => Qe_Value */
+  nl = qe & 0xFF;  qe >>= 8;    /* Next_Index_LPS + Switch_MPS */
+  nm = qe & 0xFF;  qe >>= 8;    /* Next_Index_MPS */
+
+  /* Decode & estimation procedures per sections D.2.4 & D.2.5 */
+  temp = e->a - qe;
+  e->a = temp;
+  temp <<= e->ct;
+  if (e->c >= temp) {
+    e->c -= temp;
+    /* Conditional LPS (less probable symbol) exchange */
+    if (e->a < qe) {
+      e->a = qe;
+      *st = (sv & 0x80) ^ nm;   /* Estimate_after_MPS */
+    } else {
+      e->a = qe;
+      *st = (sv & 0x80) ^ nl;   /* Estimate_after_LPS */
+      sv ^= 0x80;               /* Exchange LPS/MPS */
+    }
+  } else if (e->a < 0x8000L) {
+    /* Conditional MPS (more probable symbol) exchange */
+    if (e->a < qe) {
+      *st = (sv & 0x80) ^ nl;   /* Estimate_after_LPS */
+      sv ^= 0x80;               /* Exchange LPS/MPS */
+    } else {
+      *st = (sv & 0x80) ^ nm;   /* Estimate_after_MPS */
+    }
+  }
+
+  return sv >> 7;
+}
+
+
+/*
+ * Check for a restart marker & resynchronize decoder.
+ */
+
+LOCAL(void)
+process_restart(j_decompress_ptr cinfo)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  int ci;
+  jpeg_component_info *compptr;
+
+  /* Advance past the RSTn marker */
+  if (!(*cinfo->marker->read_restart_marker) (cinfo))
+    ERREXIT(cinfo, JERR_CANT_SUSPEND);
+
+  /* Re-initialize statistics areas */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    if (!cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+      MEMZERO(entropy->dc_statscompptr->dc_tbl_no, DC_STAT_BINS);
+      /* Reset DC predictions to 0 */
+      entropy->last_dc_valci = 0;
+      entropy->dc_contextci = 0;
+    }
+    if (!cinfo->progressive_mode || cinfo->Ss) {
+      MEMZERO(entropy->ac_statscompptr->ac_tbl_no, AC_STAT_BINS);
+    }
+  }
+
+  /* Reset arithmetic decoding variables */
+  entropy->c = 0;
+  entropy->a = 0;
+  entropy->ct = -16;    /* force reading 2 initial bytes to fill C */
+
+  /* Reset restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Arithmetic MCU decoding.
+ * Each of these routines decodes and returns one MCU's worth of
+ * arithmetic-compressed coefficients.
+ * The coefficients are reordered from zigzag order into natural array order,
+ * but are not dequantized.
+ *
+ * The i'th block of the MCU is stored into the block pointed to by
+ * MCU_datai.  WE ASSUME THIS AREA IS INITIALLY ZEROED BY THE CALLER.
+ */
+
+/*
+ * MCU decoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  JBLOCKROW block;
+  unsigned char *st;
+  int blkn, ci, tbl, sign;
+  int v, m;
+
+  /* Process restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      process_restart(cinfo);
+    entropy->restarts_to_go--;
+  }
+
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
+
+  /* Outer loop handles each block in the MCU */
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_datablkn;
+    ci = cinfo->MCU_membershipblkn;
+    tbl = cinfo->cur_comp_infoci->dc_tbl_no;
+
+    /* Sections F.2.4.1 & F.1.4.4.1: Decoding of DC coefficients */
+
+    /* Table F.4: Point to statistics bin S0 for DC coefficient coding */
+    st = entropy->dc_statstbl + entropy->dc_contextci;
+
+    /* Figure F.19: Decode_DC_DIFF */
+    if (arith_decode(cinfo, st) == 0)
+      entropy->dc_contextci = 0;
+    else {
+      /* Figure F.21: Decoding nonzero value v */
+      /* Figure F.22: Decoding the sign of v */
+      sign = arith_decode(cinfo, st + 1);
+      st += 2;  st += sign;
+      /* Figure F.23: Decoding the magnitude category of v */
+      if ((m = arith_decode(cinfo, st)) != 0) {
+        st = entropy->dc_statstbl + 20;       /* Table F.4: X1 = 20 */
+        while (arith_decode(cinfo, st)) {
+          if ((m <<= 1) == 0x8000) {
+            WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+            entropy->ct = -1;                   /* magnitude overflow */
+            return TRUE;
+          }
+          st += 1;
+        }
+      }
+      /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
+      if (m < (int)((1L << cinfo->arith_dc_Ltbl) >> 1))
+        entropy->dc_contextci = 0;               /* zero diff category */
+      else if (m > (int)((1L << cinfo->arith_dc_Utbl) >> 1))
+        entropy->dc_contextci = 12 + (sign * 4); /* large diff category */
+      else
+        entropy->dc_contextci = 4 + (sign * 4);  /* small diff category */
+      v = m;
+      /* Figure F.24: Decoding the magnitude bit pattern of v */
+      st += 14;
+      while (m >>= 1)
+        if (arith_decode(cinfo, st)) v |= m;
+      v += 1;  if (sign) v = -v;
+      entropy->last_dc_valci = (entropy->last_dc_valci + v) & 0xffff;
+    }
+
+    /* Scale and output the DC coefficient (assumes jpeg_natural_order0=0) */
+    (*block)0 = (JCOEF)LEFT_SHIFT(entropy->last_dc_valci, cinfo->Al);
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  JBLOCKROW block;
+  unsigned char *st;
+  int tbl, sign, k;
+  int v, m;
+
+  /* Process restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      process_restart(cinfo);
+    entropy->restarts_to_go--;
+  }
+
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
+
+  /* There is always only one block per MCU */
+  block = MCU_data0;
+  tbl = cinfo->cur_comp_info0->ac_tbl_no;
+
+  /* Sections F.2.4.2 & F.1.4.4.2: Decoding of AC coefficients */
+
+  /* Figure F.20: Decode_AC_coefficients */
+  for (k = cinfo->Ss; k <= cinfo->Se; k++) {
+    st = entropy->ac_statstbl + 3 * (k - 1);
+    if (arith_decode(cinfo, st)) break;         /* EOB flag */
+    while (arith_decode(cinfo, st + 1) == 0) {
+      st += 3;  k++;
+      if (k > cinfo->Se) {
+        WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+        entropy->ct = -1;                       /* spectral overflow */
+        return TRUE;
+      }
+    }
+    /* Figure F.21: Decoding nonzero value v */
+    /* Figure F.22: Decoding the sign of v */
+    sign = arith_decode(cinfo, entropy->fixed_bin);
+    st += 2;
+    /* Figure F.23: Decoding the magnitude category of v */
+    if ((m = arith_decode(cinfo, st)) != 0) {
+      if (arith_decode(cinfo, st)) {
+        m <<= 1;
+        st = entropy->ac_statstbl +
+             (k <= cinfo->arith_ac_Ktbl ? 189 : 217);
+        while (arith_decode(cinfo, st)) {
+          if ((m <<= 1) == 0x8000) {
+            WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+            entropy->ct = -1;                   /* magnitude overflow */
+            return TRUE;
+          }
+          st += 1;
+        }
+      }
+    }
+    v = m;
+    /* Figure F.24: Decoding the magnitude bit pattern of v */
+    st += 14;
+    while (m >>= 1)
+      if (arith_decode(cinfo, st)) v |= m;
+    v += 1;  if (sign) v = -v;
+    /* Scale and output coefficient in natural (dezigzagged) order */
+    (*block)jpeg_natural_orderk = (JCOEF)((unsigned)v << cinfo->Al);
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for DC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_refine(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  unsigned char *st;
+  int p1, blkn;
+
+  /* Process restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      process_restart(cinfo);
+    entropy->restarts_to_go--;
+  }
+
+  st = entropy->fixed_bin;      /* use fixed probability estimation */
+  p1 = 1 << cinfo->Al;          /* 1 in the bit position being coded */
+
+  /* Outer loop handles each block in the MCU */
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    /* Encoded data is simply the next bit of the two's-complement DC value */
+    if (arith_decode(cinfo, st))
+      MCU_datablkn00 |= p1;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_refine(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  JBLOCKROW block;
+  JCOEFPTR thiscoef;
+  unsigned char *st;
+  int tbl, k, kex;
+  int p1, m1;
+
+  /* Process restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      process_restart(cinfo);
+    entropy->restarts_to_go--;
+  }
+
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
+
+  /* There is always only one block per MCU */
+  block = MCU_data0;
+  tbl = cinfo->cur_comp_info0->ac_tbl_no;
+
+  p1 = 1 << cinfo->Al;          /* 1 in the bit position being coded */
+  m1 = (NEG_1) << cinfo->Al;    /* -1 in the bit position being coded */
+
+  /* Establish EOBx (previous stage end-of-block) index */
+  for (kex = cinfo->Se; kex > 0; kex--)
+    if ((*block)jpeg_natural_orderkex) break;
+
+  for (k = cinfo->Ss; k <= cinfo->Se; k++) {
+    st = entropy->ac_statstbl + 3 * (k - 1);
+    if (k > kex)
+      if (arith_decode(cinfo, st)) break;       /* EOB flag */
+    for (;;) {
+      thiscoef = *block + jpeg_natural_orderk;
+      if (*thiscoef) {                          /* previously nonzero coef */
+        if (arith_decode(cinfo, st + 2)) {
+          if (*thiscoef < 0)
+            *thiscoef += m1;
+          else
+            *thiscoef += p1;
+        }
+        break;
+      }
+      if (arith_decode(cinfo, st + 1)) {        /* newly nonzero coef */
+        if (arith_decode(cinfo, entropy->fixed_bin))
+          *thiscoef = m1;
+        else
+          *thiscoef = p1;
+        break;
+      }
+      st += 3;  k++;
+      if (k > cinfo->Se) {
+        WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+        entropy->ct = -1;                       /* spectral overflow */
+        return TRUE;
+      }
+    }
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Decode one MCU's worth of arithmetic-compressed coefficients.
+ */
+
+METHODDEF(boolean)
+decode_mcu(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  jpeg_component_info *compptr;
+  JBLOCKROW block;
+  unsigned char *st;
+  int blkn, ci, tbl, sign, k;
+  int v, m;
+
+  /* Process restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      process_restart(cinfo);
+    entropy->restarts_to_go--;
+  }
+
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
+
+  /* Outer loop handles each block in the MCU */
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_data ? MCU_datablkn : NULL;
+    ci = cinfo->MCU_membershipblkn;
+    compptr = cinfo->cur_comp_infoci;
+
+    /* Sections F.2.4.1 & F.1.4.4.1: Decoding of DC coefficients */
+
+    tbl = compptr->dc_tbl_no;
+
+    /* Table F.4: Point to statistics bin S0 for DC coefficient coding */
+    st = entropy->dc_statstbl + entropy->dc_contextci;
+
+    /* Figure F.19: Decode_DC_DIFF */
+    if (arith_decode(cinfo, st) == 0)
+      entropy->dc_contextci = 0;
+    else {
+      /* Figure F.21: Decoding nonzero value v */
+      /* Figure F.22: Decoding the sign of v */
+      sign = arith_decode(cinfo, st + 1);
+      st += 2;  st += sign;
+      /* Figure F.23: Decoding the magnitude category of v */
+      if ((m = arith_decode(cinfo, st)) != 0) {
+        st = entropy->dc_statstbl + 20;       /* Table F.4: X1 = 20 */
+        while (arith_decode(cinfo, st)) {
+          if ((m <<= 1) == 0x8000) {
+            WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+            entropy->ct = -1;                   /* magnitude overflow */
+            return TRUE;
+          }
+          st += 1;
+        }
+      }
+      /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
+      if (m < (int)((1L << cinfo->arith_dc_Ltbl) >> 1))
+        entropy->dc_contextci = 0;               /* zero diff category */
+      else if (m > (int)((1L << cinfo->arith_dc_Utbl) >> 1))
+        entropy->dc_contextci = 12 + (sign * 4); /* large diff category */
+      else
+        entropy->dc_contextci = 4 + (sign * 4);  /* small diff category */
+      v = m;
+      /* Figure F.24: Decoding the magnitude bit pattern of v */
+      st += 14;
+      while (m >>= 1)
+        if (arith_decode(cinfo, st)) v |= m;
+      v += 1;  if (sign) v = -v;
+      entropy->last_dc_valci = (entropy->last_dc_valci + v) & 0xffff;
+    }
+
+    if (block)
+      (*block)0 = (JCOEF)entropy->last_dc_valci;
+
+    /* Sections F.2.4.2 & F.1.4.4.2: Decoding of AC coefficients */
+
+    tbl = compptr->ac_tbl_no;
+
+    /* Figure F.20: Decode_AC_coefficients */
+    for (k = 1; k <= DCTSIZE2 - 1; k++) {
+      st = entropy->ac_statstbl + 3 * (k - 1);
+      if (arith_decode(cinfo, st)) break;       /* EOB flag */
+      while (arith_decode(cinfo, st + 1) == 0) {
+        st += 3;  k++;
+        if (k > DCTSIZE2 - 1) {
+          WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+          entropy->ct = -1;                     /* spectral overflow */
+          return TRUE;
+        }
+      }
+      /* Figure F.21: Decoding nonzero value v */
+      /* Figure F.22: Decoding the sign of v */
+      sign = arith_decode(cinfo, entropy->fixed_bin);
+      st += 2;
+      /* Figure F.23: Decoding the magnitude category of v */
+      if ((m = arith_decode(cinfo, st)) != 0) {
+        if (arith_decode(cinfo, st)) {
+          m <<= 1;
+          st = entropy->ac_statstbl +
+               (k <= cinfo->arith_ac_Ktbl ? 189 : 217);
+          while (arith_decode(cinfo, st)) {
+            if ((m <<= 1) == 0x8000) {
+              WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+              entropy->ct = -1;                 /* magnitude overflow */
+              return TRUE;
+            }
+            st += 1;
+          }
+        }
+      }
+      v = m;
+      /* Figure F.24: Decoding the magnitude bit pattern of v */
+      st += 14;
+      while (m >>= 1)
+        if (arith_decode(cinfo, st)) v |= m;
+      v += 1;  if (sign) v = -v;
+      if (block)
+        (*block)jpeg_natural_orderk = (JCOEF)v;
+    }
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Initialize for an arithmetic-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass(j_decompress_ptr cinfo)
+{
+  arith_entropy_ptr entropy = (arith_entropy_ptr)cinfo->entropy;
+  int ci, tbl;
+  jpeg_component_info *compptr;
+
+  if (cinfo->progressive_mode) {
+    /* Validate progressive scan parameters */
+    if (cinfo->Ss == 0) {
+      if (cinfo->Se != 0)
+        goto bad;
+    } else {
+      /* need not check Ss/Se < 0 since they came from unsigned bytes */
+      if (cinfo->Se < cinfo->Ss || cinfo->Se > DCTSIZE2 - 1)
+        goto bad;
+      /* AC scans may have only one component */
+      if (cinfo->comps_in_scan != 1)
+        goto bad;
+    }
+    if (cinfo->Ah != 0) {
+      /* Successive approximation refinement scan: must have Al = Ah-1. */
+      if (cinfo->Ah - 1 != cinfo->Al)
+        goto bad;
+    }
+    if (cinfo->Al > 13) {       /* need not check for < 0 */
+bad:
+      ERREXIT4(cinfo, JERR_BAD_PROGRESSION,
+               cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
+    }
+    /* Update progression status, and verify that scan order is legal.
+     * Note that inter-scan inconsistencies are treated as warnings
+     * not fatal errors ... not clear if this is right way to behave.
+     */
+    for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+      int coefi, cindex = cinfo->cur_comp_infoci->component_index;
+      int *coef_bit_ptr = &cinfo->coef_bitscindex0;
+      int *prev_coef_bit_ptr =
+        &cinfo->coef_bitscindex + cinfo->num_components0;
+      if (cinfo->Ss && coef_bit_ptr0 < 0) /* AC without prior DC scan */
+        WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
+      for (coefi = MIN(cinfo->Ss, 1); coefi <= MAX(cinfo->Se, 9); coefi++) {
+        if (cinfo->input_scan_number > 1)
+          prev_coef_bit_ptrcoefi = coef_bit_ptrcoefi;
+        else
+          prev_coef_bit_ptrcoefi = 0;
+      }
+      for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
+        int expected = (coef_bit_ptrcoefi < 0) ? 0 : coef_bit_ptrcoefi;
+        if (cinfo->Ah != expected)
+          WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
+        coef_bit_ptrcoefi = cinfo->Al;
+      }
+    }
+    /* Select MCU decoding routine */
+    if (cinfo->Ah == 0) {
+      if (cinfo->Ss == 0)
+        entropy->pub.decode_mcu = decode_mcu_DC_first;
+      else
+        entropy->pub.decode_mcu = decode_mcu_AC_first;
+    } else {
+      if (cinfo->Ss == 0)
+        entropy->pub.decode_mcu = decode_mcu_DC_refine;
+      else
+        entropy->pub.decode_mcu = decode_mcu_AC_refine;
+    }
+  } else {
+    /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
+     * This ought to be an error condition, but we make it a warning.
+     */
+    if (cinfo->Ss != 0 || cinfo->Ah != 0 || cinfo->Al != 0 ||
+        (cinfo->Se < DCTSIZE2 && cinfo->Se != DCTSIZE2 - 1))
+      WARNMS(cinfo, JWRN_NOT_SEQUENTIAL);
+    /* Select MCU decoding routine */
+    entropy->pub.decode_mcu = decode_mcu;
+  }
+
+  /* Allocate & initialize requested statistics areas */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    if (!cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+      tbl = compptr->dc_tbl_no;
+      if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+      if (entropy->dc_statstbl == NULL)
+        entropy->dc_statstbl = (unsigned char *)(*cinfo->mem->alloc_small)
+          ((j_common_ptr)cinfo, JPOOL_IMAGE, DC_STAT_BINS);
+      MEMZERO(entropy->dc_statstbl, DC_STAT_BINS);
+      /* Initialize DC predictions to 0 */
+      entropy->last_dc_valci = 0;
+      entropy->dc_contextci = 0;
+    }
+    if (!cinfo->progressive_mode || cinfo->Ss) {
+      tbl = compptr->ac_tbl_no;
+      if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+      if (entropy->ac_statstbl == NULL)
+        entropy->ac_statstbl = (unsigned char *)(*cinfo->mem->alloc_small)
+          ((j_common_ptr)cinfo, JPOOL_IMAGE, AC_STAT_BINS);
+      MEMZERO(entropy->ac_statstbl, AC_STAT_BINS);
+    }
+  }
+
+  /* Initialize arithmetic decoding variables */
+  entropy->c = 0;
+  entropy->a = 0;
+  entropy->ct = -16;    /* force reading 2 initial bytes to fill C */
+  entropy->pub.insufficient_data = FALSE;
+
+  /* Initialize restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Module initialization routine for arithmetic entropy decoding.
+ */
+
+GLOBAL(void)
+jinit_arith_decoder(j_decompress_ptr cinfo)
+{
+  arith_entropy_ptr entropy;
+  int i;
+
+  entropy = (arith_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(arith_entropy_decoder));
+  cinfo->entropy = (struct jpeg_entropy_decoder *)entropy;
+  entropy->pub.start_pass = start_pass;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    entropy->dc_statsi = NULL;
+    entropy->ac_statsi = NULL;
+  }
+
+  /* Initialize index for fixed probability estimation */
+  entropy->fixed_bin0 = 113;
+
+  if (cinfo->progressive_mode) {
+    /* Create progression status table */
+    int *coef_bit_ptr, ci;
+    cinfo->coef_bits = (int (*)DCTSIZE2)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  cinfo->num_components * 2 * DCTSIZE2 *
+                                  sizeof(int));
+    coef_bit_ptr = &cinfo->coef_bits00;
+    for (ci = 0; ci < cinfo->num_components; ci++)
+      for (i = 0; i < DCTSIZE2; i++)
+        *coef_bit_ptr++ = -1;
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdatadst-tj.c Added

@@ -0,0 +1,202 @@
+/*
+ * jdatadst-tj.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2012 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, 2014, 2016, 2019, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains compression data destination routines for the case of
+ * emitting JPEG data to memory or to a file (or any stdio stream).
+ * While these routines are sufficient for most applications,
+ * some will want to use a different destination manager.
+ * IMPORTANT: we assume that fwrite() will correctly transcribe an array of
+ * JOCTETs into 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc(size_t size);
+extern void free(void *ptr);
+#endif
+void jpeg_mem_dest_tj(j_compress_ptr cinfo, unsigned char **outbuffer,
+                      unsigned long *outsize, boolean alloc);
+
+
+#define OUTPUT_BUF_SIZE  4096   /* choose an efficiently fwrite'able size */
+
+
+/* Expanded data destination object for memory output */
+
+typedef struct {
+  struct jpeg_destination_mgr pub; /* public fields */
+
+  unsigned char **outbuffer;    /* target buffer */
+  unsigned long *outsize;
+  unsigned char *newbuffer;     /* newly allocated buffer */
+  JOCTET *buffer;               /* start of buffer */
+  size_t bufsize;
+  boolean alloc;
+} my_mem_destination_mgr;
+
+typedef my_mem_destination_mgr *my_mem_dest_ptr;
+
+
+/*
+ * Initialize destination --- called by jpeg_start_compress
+ * before any data is actually written.
+ */
+
+METHODDEF(void)
+init_mem_destination(j_compress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+
+
+/*
+ * Empty the output buffer --- called whenever buffer fills up.
+ *
+ * In typical applications, this should write the entire output buffer
+ * (ignoring the current state of next_output_byte & free_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been dumped.
+ *
+ * In applications that need to be able to suspend compression due to output
+ * overrun, a FALSE return indicates that the buffer cannot be emptied now.
+ * In this situation, the compressor will return to its caller (possibly with
+ * an indication that it has not accepted all the supplied scanlines).  The
+ * application should resume compression after it has made more room in the
+ * output buffer.  Note that there are substantial restrictions on the use of
+ * suspension --- see the documentation.
+ *
+ * When suspending, the compressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_output_byte & free_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point will be regenerated after resumption, so do not
+ * write it out when emptying the buffer externally.
+ */
+
+METHODDEF(boolean)
+empty_mem_output_buffer(j_compress_ptr cinfo)
+{
+  size_t nextsize;
+  JOCTET *nextbuffer;
+  my_mem_dest_ptr dest = (my_mem_dest_ptr)cinfo->dest;
+
+  if (!dest->alloc) ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* Try to allocate new buffer with double size */
+  nextsize = dest->bufsize * 2;
+  nextbuffer = (JOCTET *)malloc(nextsize);
+
+  if (nextbuffer == NULL)
+    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
+
+  MEMCOPY(nextbuffer, dest->buffer, dest->bufsize);
+
+  free(dest->newbuffer);
+
+  dest->newbuffer = nextbuffer;
+
+  dest->pub.next_output_byte = nextbuffer + dest->bufsize;
+  dest->pub.free_in_buffer = dest->bufsize;
+
+  dest->buffer = nextbuffer;
+  dest->bufsize = nextsize;
+
+  return TRUE;
+}
+
+
+/*
+ * Terminate destination --- called by jpeg_finish_compress
+ * after all data has been written.  Usually needs to flush buffer.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_mem_destination(j_compress_ptr cinfo)
+{
+  my_mem_dest_ptr dest = (my_mem_dest_ptr)cinfo->dest;
+
+  if (dest->alloc) *dest->outbuffer = dest->buffer;
+  *dest->outsize = (unsigned long)(dest->bufsize - dest->pub.free_in_buffer);
+}
+
+
+/*
+ * Prepare for output to a memory buffer.
+ * The caller may supply an own initial buffer with appropriate size.
+ * Otherwise, or when the actual data output exceeds the given size,
+ * the library adapts the buffer size as necessary.
+ * The standard library functions malloc/free are used for allocating
+ * larger memory, so the buffer is available to the application after
+ * finishing compression, and then the application is responsible for
+ * freeing the requested memory.
+ */
+
+GLOBAL(void)
+jpeg_mem_dest_tj(j_compress_ptr cinfo, unsigned char **outbuffer,
+                 unsigned long *outsize, boolean alloc)
+{
+  boolean reused = FALSE;
+  my_mem_dest_ptr dest;
+
+  if (outbuffer == NULL || outsize == NULL)     /* sanity check */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* The destination object is made permanent so that multiple JPEG images
+   * can be written to the same buffer without re-executing jpeg_mem_dest.
+   */
+  if (cinfo->dest == NULL) {    /* first time for this JPEG object? */
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_mem_destination_mgr));
+    dest = (my_mem_dest_ptr)cinfo->dest;
+    dest->newbuffer = NULL;
+    dest->buffer = NULL;
+  } else if (cinfo->dest->init_destination != init_mem_destination) {
+    /* It is unsafe to reuse the existing destination manager unless it was
+     * created by this function.
+     */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+  }
+
+  dest = (my_mem_dest_ptr)cinfo->dest;
+  dest->pub.init_destination = init_mem_destination;
+  dest->pub.empty_output_buffer = empty_mem_output_buffer;
+  dest->pub.term_destination = term_mem_destination;
+  if (dest->buffer == *outbuffer && *outbuffer != NULL && alloc)
+    reused = TRUE;
+  dest->outbuffer = outbuffer;
+  dest->outsize = outsize;
+  dest->alloc = alloc;
+
+  if (*outbuffer == NULL || *outsize == 0) {
+    if (alloc) {
+      /* Allocate initial buffer */
+      dest->newbuffer = *outbuffer = (unsigned char *)malloc(OUTPUT_BUF_SIZE);
+      if (dest->newbuffer == NULL)
+        ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
+      *outsize = OUTPUT_BUF_SIZE;
+    } else
+      ERREXIT(cinfo, JERR_BUFFER_SIZE);
+  }
+
+  dest->pub.next_output_byte = dest->buffer = *outbuffer;
+  if (!reused)
+    dest->bufsize = *outsize;
+  dest->pub.free_in_buffer = dest->bufsize;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdatadst.c Added

@@ -0,0 +1,292 @@
+/*
+ * jdatadst.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2012 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains compression data destination routines for the case of
+ * emitting JPEG data to memory or to a file (or any stdio stream).
+ * While these routines are sufficient for most applications,
+ * some will want to use a different destination manager.
+ * IMPORTANT: we assume that fwrite() will correctly transcribe an array of
+ * JOCTETs into 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc(size_t size);
+extern void free(void *ptr);
+#endif
+
+
+/* Expanded data destination object for stdio output */
+
+typedef struct {
+  struct jpeg_destination_mgr pub; /* public fields */
+
+  FILE *outfile;                /* target stream */
+  JOCTET *buffer;               /* start of buffer */
+} my_destination_mgr;
+
+typedef my_destination_mgr *my_dest_ptr;
+
+#define OUTPUT_BUF_SIZE  4096   /* choose an efficiently fwrite'able size */
+
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/* Expanded data destination object for memory output */
+
+typedef struct {
+  struct jpeg_destination_mgr pub; /* public fields */
+
+  unsigned char **outbuffer;    /* target buffer */
+  unsigned long *outsize;
+  unsigned char *newbuffer;     /* newly allocated buffer */
+  JOCTET *buffer;               /* start of buffer */
+  size_t bufsize;
+} my_mem_destination_mgr;
+
+typedef my_mem_destination_mgr *my_mem_dest_ptr;
+#endif
+
+
+/*
+ * Initialize destination --- called by jpeg_start_compress
+ * before any data is actually written.
+ */
+
+METHODDEF(void)
+init_destination(j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr)cinfo->dest;
+
+  /* Allocate the output buffer --- it will be released when done with image */
+  dest->buffer = (JOCTET *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                OUTPUT_BUF_SIZE * sizeof(JOCTET));
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(void)
+init_mem_destination(j_compress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+#endif
+
+
+/*
+ * Empty the output buffer --- called whenever buffer fills up.
+ *
+ * In typical applications, this should write the entire output buffer
+ * (ignoring the current state of next_output_byte & free_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been dumped.
+ *
+ * In applications that need to be able to suspend compression due to output
+ * overrun, a FALSE return indicates that the buffer cannot be emptied now.
+ * In this situation, the compressor will return to its caller (possibly with
+ * an indication that it has not accepted all the supplied scanlines).  The
+ * application should resume compression after it has made more room in the
+ * output buffer.  Note that there are substantial restrictions on the use of
+ * suspension --- see the documentation.
+ *
+ * When suspending, the compressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_output_byte & free_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point will be regenerated after resumption, so do not
+ * write it out when emptying the buffer externally.
+ */
+
+METHODDEF(boolean)
+empty_output_buffer(j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr)cinfo->dest;
+
+  if (JFWRITE(dest->outfile, dest->buffer, OUTPUT_BUF_SIZE) !=
+      (size_t)OUTPUT_BUF_SIZE)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+
+  return TRUE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(boolean)
+empty_mem_output_buffer(j_compress_ptr cinfo)
+{
+  size_t nextsize;
+  JOCTET *nextbuffer;
+  my_mem_dest_ptr dest = (my_mem_dest_ptr)cinfo->dest;
+
+  /* Try to allocate new buffer with double size */
+  nextsize = dest->bufsize * 2;
+  nextbuffer = (JOCTET *)malloc(nextsize);
+
+  if (nextbuffer == NULL)
+    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
+
+  MEMCOPY(nextbuffer, dest->buffer, dest->bufsize);
+
+  free(dest->newbuffer);
+
+  dest->newbuffer = nextbuffer;
+
+  dest->pub.next_output_byte = nextbuffer + dest->bufsize;
+  dest->pub.free_in_buffer = dest->bufsize;
+
+  dest->buffer = nextbuffer;
+  dest->bufsize = nextsize;
+
+  return TRUE;
+}
+#endif
+
+
+/*
+ * Terminate destination --- called by jpeg_finish_compress
+ * after all data has been written.  Usually needs to flush buffer.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_destination(j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr)cinfo->dest;
+  size_t datacount = OUTPUT_BUF_SIZE - dest->pub.free_in_buffer;
+
+  /* Write any data remaining in the buffer */
+  if (datacount > 0) {
+    if (JFWRITE(dest->outfile, dest->buffer, datacount) != datacount)
+      ERREXIT(cinfo, JERR_FILE_WRITE);
+  }
+  fflush(dest->outfile);
+  /* Make sure we wrote the output file OK */
+  if (ferror(dest->outfile))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(void)
+term_mem_destination(j_compress_ptr cinfo)
+{
+  my_mem_dest_ptr dest = (my_mem_dest_ptr)cinfo->dest;
+
+  *dest->outbuffer = dest->buffer;
+  *dest->outsize = (unsigned long)(dest->bufsize - dest->pub.free_in_buffer);
+}
+#endif
+
+
+/*
+ * Prepare for output to a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing compression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_dest(j_compress_ptr cinfo, FILE *outfile)
+{
+  my_dest_ptr dest;
+
+  /* The destination object is made permanent so that multiple JPEG images
+   * can be written to the same file without re-executing jpeg_stdio_dest.
+   */
+  if (cinfo->dest == NULL) {    /* first time for this JPEG object? */
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_destination_mgr));
+  } else if (cinfo->dest->init_destination != init_destination) {
+    /* It is unsafe to reuse the existing destination manager unless it was
+     * created by this function.  Otherwise, there is no guarantee that the
+     * opaque structure is the right size.  Note that we could just create a
+     * new structure, but the old structure would not be freed until
+     * jpeg_destroy_compress() was called.
+     */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+  }
+
+  dest = (my_dest_ptr)cinfo->dest;
+  dest->pub.init_destination = init_destination;
+  dest->pub.empty_output_buffer = empty_output_buffer;
+  dest->pub.term_destination = term_destination;
+  dest->outfile = outfile;
+}
+
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/*
+ * Prepare for output to a memory buffer.
+ * The caller may supply an own initial buffer with appropriate size.
+ * Otherwise, or when the actual data output exceeds the given size,
+ * the library adapts the buffer size as necessary.
+ * The standard library functions malloc/free are used for allocating
+ * larger memory, so the buffer is available to the application after
+ * finishing compression, and then the application is responsible for
+ * freeing the requested memory.
+ * Note:  An initial buffer supplied by the caller is expected to be
+ * managed by the application.  The library does not free such buffer
+ * when allocating a larger buffer.
+ */
+
+GLOBAL(void)
+jpeg_mem_dest(j_compress_ptr cinfo, unsigned char **outbuffer,
+              unsigned long *outsize)
+{
+  my_mem_dest_ptr dest;
+
+  if (outbuffer == NULL || outsize == NULL)     /* sanity check */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* The destination object is made permanent so that multiple JPEG images
+   * can be written to the same buffer without re-executing jpeg_mem_dest.
+   */
+  if (cinfo->dest == NULL) {    /* first time for this JPEG object? */
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_mem_destination_mgr));
+  } else if (cinfo->dest->init_destination != init_mem_destination) {
+    /* It is unsafe to reuse the existing destination manager unless it was
+     * created by this function.
+     */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+  }
+
+  dest = (my_mem_dest_ptr)cinfo->dest;
+  dest->pub.init_destination = init_mem_destination;
+  dest->pub.empty_output_buffer = empty_mem_output_buffer;
+  dest->pub.term_destination = term_mem_destination;
+  dest->outbuffer = outbuffer;
+  dest->outsize = outsize;
+  dest->newbuffer = NULL;
+
+  if (*outbuffer == NULL || *outsize == 0) {
+    /* Allocate initial buffer */
+    dest->newbuffer = *outbuffer = (unsigned char *)malloc(OUTPUT_BUF_SIZE);
+    if (dest->newbuffer == NULL)
+      ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
+    *outsize = OUTPUT_BUF_SIZE;
+  }
+
+  dest->pub.next_output_byte = dest->buffer = *outbuffer;
+  dest->pub.free_in_buffer = dest->bufsize = *outsize;
+}
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdatasrc-tj.c Added

@@ -0,0 +1,194 @@
+/*
+ * jdatasrc-tj.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, 2016, 2019, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains decompression data source routines for the case of
+ * reading JPEG data from memory or from a file (or any stdio stream).
+ * While these routines are sufficient for most applications,
+ * some will want to use a different source manager.
+ * IMPORTANT: we assume that fread() will correctly transcribe an array of
+ * JOCTETs from 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+void jpeg_mem_src_tj(j_decompress_ptr cinfo, const unsigned char *inbuffer,
+                     unsigned long insize);
+
+
+/*
+ * Initialize source --- called by jpeg_read_header
+ * before any data is actually read.
+ */
+
+METHODDEF(void)
+init_mem_source(j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+
+
+/*
+ * Fill the input buffer --- called whenever buffer is emptied.
+ *
+ * In typical applications, this should read fresh data into the buffer
+ * (ignoring the current state of next_input_byte & bytes_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been reloaded.  It is not necessary to
+ * fill the buffer entirely, only to obtain at least one more byte.
+ *
+ * There is no such thing as an EOF return.  If the end of the file has been
+ * reached, the routine has a choice of ERREXIT() or inserting fake data into
+ * the buffer.  In most cases, generating a warning message and inserting a
+ * fake EOI marker is the best course of action --- this will allow the
+ * decompressor to output however much of the image is there.  However,
+ * the resulting error message is misleading if the real problem is an empty
+ * input file, so we handle that case specially.
+ *
+ * In applications that need to be able to suspend compression due to input
+ * not being available yet, a FALSE return indicates that no more data can be
+ * obtained right now, but more may be forthcoming later.  In this situation,
+ * the decompressor will return to its caller (with an indication of the
+ * number of scanlines it has read, if any).  The application should resume
+ * decompression after it has loaded more data into the input buffer.  Note
+ * that there are substantial restrictions on the use of suspension --- see
+ * the documentation.
+ *
+ * When suspending, the decompressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_input_byte & bytes_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point must be rescanned after resumption, so move it to
+ * the front of the buffer rather than discarding it.
+ */
+
+METHODDEF(boolean)
+fill_mem_input_buffer(j_decompress_ptr cinfo)
+{
+  static const JOCTET mybuffer4 = {
+    (JOCTET)0xFF, (JOCTET)JPEG_EOI, 0, 0
+  };
+
+  /* The whole JPEG data is expected to reside in the supplied memory
+   * buffer, so any request for more data beyond the given buffer size
+   * is treated as an error.
+   */
+  WARNMS(cinfo, JWRN_JPEG_EOF);
+
+  /* Insert a fake EOI marker */
+
+  cinfo->src->next_input_byte = mybuffer;
+  cinfo->src->bytes_in_buffer = 2;
+
+  return TRUE;
+}
+
+
+/*
+ * Skip data --- used to skip over a potentially large amount of
+ * uninteresting data (such as an APPn marker).
+ *
+ * Writers of suspendable-input applications must note that skip_input_data
+ * is not granted the right to give a suspension return.  If the skip extends
+ * beyond the data currently in the buffer, the buffer can be marked empty so
+ * that the next read will cause a fill_input_buffer call that can suspend.
+ * Arranging for additional bytes to be discarded before reloading the input
+ * buffer is the application writer's problem.
+ */
+
+METHODDEF(void)
+skip_input_data(j_decompress_ptr cinfo, long num_bytes)
+{
+  struct jpeg_source_mgr *src = cinfo->src;
+
+  /* Just a dumb implementation for now.  Could use fseek() except
+   * it doesn't work on pipes.  Not clear that being smart is worth
+   * any trouble anyway --- large skips are infrequent.
+   */
+  if (num_bytes > 0) {
+    while (num_bytes > (long)src->bytes_in_buffer) {
+      num_bytes -= (long)src->bytes_in_buffer;
+      (void)(*src->fill_input_buffer) (cinfo);
+      /* note we assume that fill_input_buffer will never return FALSE,
+       * so suspension need not be handled.
+       */
+    }
+    src->next_input_byte += (size_t)num_bytes;
+    src->bytes_in_buffer -= (size_t)num_bytes;
+  }
+}
+
+
+/*
+ * An additional method that can be provided by data source modules is the
+ * resync_to_restart method for error recovery in the presence of RST markers.
+ * For the moment, this source module just uses the default resync method
+ * provided by the JPEG library.  That method assumes that no backtracking
+ * is possible.
+ */
+
+
+/*
+ * Terminate source --- called by jpeg_finish_decompress
+ * after all data has been read.  Often a no-op.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_source(j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+
+
+/*
+ * Prepare for input from a supplied memory buffer.
+ * The buffer must contain the whole JPEG data.
+ */
+
+GLOBAL(void)
+jpeg_mem_src_tj(j_decompress_ptr cinfo, const unsigned char *inbuffer,
+                unsigned long insize)
+{
+  struct jpeg_source_mgr *src;
+
+  if (inbuffer == NULL || insize == 0)  /* Treat empty input as fatal error */
+    ERREXIT(cinfo, JERR_INPUT_EMPTY);
+
+  /* The source object is made permanent so that a series of JPEG images
+   * can be read from the same buffer by calling jpeg_mem_src only before
+   * the first one.
+   */
+  if (cinfo->src == NULL) {     /* first time for this JPEG object? */
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                  sizeof(struct jpeg_source_mgr));
+  } else if (cinfo->src->init_source != init_mem_source) {
+    /* It is unsafe to reuse the existing source manager unless it was created
+     * by this function.
+     */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+  }
+
+  src = cinfo->src;
+  src->init_source = init_mem_source;
+  src->fill_input_buffer = fill_mem_input_buffer;
+  src->skip_input_data = skip_input_data;
+  src->resync_to_restart = jpeg_resync_to_restart; /* use default method */
+  src->term_source = term_source;
+  src->bytes_in_buffer = (size_t)insize;
+  src->next_input_byte = (const JOCTET *)inbuffer;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdatasrc.c Added

@@ -0,0 +1,295 @@
+/*
+ * jdatasrc.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains decompression data source routines for the case of
+ * reading JPEG data from memory or from a file (or any stdio stream).
+ * While these routines are sufficient for most applications,
+ * some will want to use a different source manager.
+ * IMPORTANT: we assume that fread() will correctly transcribe an array of
+ * JOCTETs from 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+
+/* Expanded data source object for stdio input */
+
+typedef struct {
+  struct jpeg_source_mgr pub;   /* public fields */
+
+  FILE *infile;                 /* source stream */
+  JOCTET *buffer;               /* start of buffer */
+  boolean start_of_file;        /* have we gotten any data yet? */
+} my_source_mgr;
+
+typedef my_source_mgr *my_src_ptr;
+
+#define INPUT_BUF_SIZE  4096    /* choose an efficiently fread'able size */
+
+
+/*
+ * Initialize source --- called by jpeg_read_header
+ * before any data is actually read.
+ */
+
+METHODDEF(void)
+init_source(j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr)cinfo->src;
+
+  /* We reset the empty-input-file flag for each image,
+   * but we don't clear the input buffer.
+   * This is correct behavior for reading a series of images from one source.
+   */
+  src->start_of_file = TRUE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(void)
+init_mem_source(j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+#endif
+
+
+/*
+ * Fill the input buffer --- called whenever buffer is emptied.
+ *
+ * In typical applications, this should read fresh data into the buffer
+ * (ignoring the current state of next_input_byte & bytes_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been reloaded.  It is not necessary to
+ * fill the buffer entirely, only to obtain at least one more byte.
+ *
+ * There is no such thing as an EOF return.  If the end of the file has been
+ * reached, the routine has a choice of ERREXIT() or inserting fake data into
+ * the buffer.  In most cases, generating a warning message and inserting a
+ * fake EOI marker is the best course of action --- this will allow the
+ * decompressor to output however much of the image is there.  However,
+ * the resulting error message is misleading if the real problem is an empty
+ * input file, so we handle that case specially.
+ *
+ * In applications that need to be able to suspend compression due to input
+ * not being available yet, a FALSE return indicates that no more data can be
+ * obtained right now, but more may be forthcoming later.  In this situation,
+ * the decompressor will return to its caller (with an indication of the
+ * number of scanlines it has read, if any).  The application should resume
+ * decompression after it has loaded more data into the input buffer.  Note
+ * that there are substantial restrictions on the use of suspension --- see
+ * the documentation.
+ *
+ * When suspending, the decompressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_input_byte & bytes_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point must be rescanned after resumption, so move it to
+ * the front of the buffer rather than discarding it.
+ */
+
+METHODDEF(boolean)
+fill_input_buffer(j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr)cinfo->src;
+  size_t nbytes;
+
+  nbytes = JFREAD(src->infile, src->buffer, INPUT_BUF_SIZE);
+
+  if (nbytes <= 0) {
+    if (src->start_of_file)     /* Treat empty input file as fatal error */
+      ERREXIT(cinfo, JERR_INPUT_EMPTY);
+    WARNMS(cinfo, JWRN_JPEG_EOF);
+    /* Insert a fake EOI marker */
+    src->buffer0 = (JOCTET)0xFF;
+    src->buffer1 = (JOCTET)JPEG_EOI;
+    nbytes = 2;
+  }
+
+  src->pub.next_input_byte = src->buffer;
+  src->pub.bytes_in_buffer = nbytes;
+  src->start_of_file = FALSE;
+
+  return TRUE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(boolean)
+fill_mem_input_buffer(j_decompress_ptr cinfo)
+{
+  static const JOCTET mybuffer4 = {
+    (JOCTET)0xFF, (JOCTET)JPEG_EOI, 0, 0
+  };
+
+  /* The whole JPEG data is expected to reside in the supplied memory
+   * buffer, so any request for more data beyond the given buffer size
+   * is treated as an error.
+   */
+  WARNMS(cinfo, JWRN_JPEG_EOF);
+
+  /* Insert a fake EOI marker */
+
+  cinfo->src->next_input_byte = mybuffer;
+  cinfo->src->bytes_in_buffer = 2;
+
+  return TRUE;
+}
+#endif
+
+
+/*
+ * Skip data --- used to skip over a potentially large amount of
+ * uninteresting data (such as an APPn marker).
+ *
+ * Writers of suspendable-input applications must note that skip_input_data
+ * is not granted the right to give a suspension return.  If the skip extends
+ * beyond the data currently in the buffer, the buffer can be marked empty so
+ * that the next read will cause a fill_input_buffer call that can suspend.
+ * Arranging for additional bytes to be discarded before reloading the input
+ * buffer is the application writer's problem.
+ */
+
+METHODDEF(void)
+skip_input_data(j_decompress_ptr cinfo, long num_bytes)
+{
+  struct jpeg_source_mgr *src = cinfo->src;
+
+  /* Just a dumb implementation for now.  Could use fseek() except
+   * it doesn't work on pipes.  Not clear that being smart is worth
+   * any trouble anyway --- large skips are infrequent.
+   */
+  if (num_bytes > 0) {
+    while (num_bytes > (long)src->bytes_in_buffer) {
+      num_bytes -= (long)src->bytes_in_buffer;
+      (void)(*src->fill_input_buffer) (cinfo);
+      /* note we assume that fill_input_buffer will never return FALSE,
+       * so suspension need not be handled.
+       */
+    }
+    src->next_input_byte += (size_t)num_bytes;
+    src->bytes_in_buffer -= (size_t)num_bytes;
+  }
+}
+
+
+/*
+ * An additional method that can be provided by data source modules is the
+ * resync_to_restart method for error recovery in the presence of RST markers.
+ * For the moment, this source module just uses the default resync method
+ * provided by the JPEG library.  That method assumes that no backtracking
+ * is possible.
+ */
+
+
+/*
+ * Terminate source --- called by jpeg_finish_decompress
+ * after all data has been read.  Often a no-op.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_source(j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+
+
+/*
+ * Prepare for input from a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing decompression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_src(j_decompress_ptr cinfo, FILE *infile)
+{
+  my_src_ptr src;
+
+  /* The source object and input buffer are made permanent so that a series
+   * of JPEG images can be read from the same file by calling jpeg_stdio_src
+   * only before the first one.  (If we discarded the buffer at the end of
+   * one image, we'd likely lose the start of the next one.)
+   */
+  if (cinfo->src == NULL) {     /* first time for this JPEG object? */
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_source_mgr));
+    src = (my_src_ptr)cinfo->src;
+    src->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                  INPUT_BUF_SIZE * sizeof(JOCTET));
+  } else if (cinfo->src->init_source != init_source) {
+    /* It is unsafe to reuse the existing source manager unless it was created
+     * by this function.  Otherwise, there is no guarantee that the opaque
+     * structure is the right size.  Note that we could just create a new
+     * structure, but the old structure would not be freed until
+     * jpeg_destroy_decompress() was called.
+     */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+  }
+
+  src = (my_src_ptr)cinfo->src;
+  src->pub.init_source = init_source;
+  src->pub.fill_input_buffer = fill_input_buffer;
+  src->pub.skip_input_data = skip_input_data;
+  src->pub.resync_to_restart = jpeg_resync_to_restart; /* use default method */
+  src->pub.term_source = term_source;
+  src->infile = infile;
+  src->pub.bytes_in_buffer = 0; /* forces fill_input_buffer on first read */
+  src->pub.next_input_byte = NULL; /* until buffer loaded */
+}
+
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/*
+ * Prepare for input from a supplied memory buffer.
+ * The buffer must contain the whole JPEG data.
+ */
+
+GLOBAL(void)
+jpeg_mem_src(j_decompress_ptr cinfo, const unsigned char *inbuffer,
+             unsigned long insize)
+{
+  struct jpeg_source_mgr *src;
+
+  if (inbuffer == NULL || insize == 0)  /* Treat empty input as fatal error */
+    ERREXIT(cinfo, JERR_INPUT_EMPTY);
+
+  /* The source object is made permanent so that a series of JPEG images
+   * can be read from the same buffer by calling jpeg_mem_src only before
+   * the first one.
+   */
+  if (cinfo->src == NULL) {     /* first time for this JPEG object? */
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                  sizeof(struct jpeg_source_mgr));
+  } else if (cinfo->src->init_source != init_mem_source) {
+    /* It is unsafe to reuse the existing source manager unless it was created
+     * by this function.
+     */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+  }
+
+  src = cinfo->src;
+  src->init_source = init_mem_source;
+  src->fill_input_buffer = fill_mem_input_buffer;
+  src->skip_input_data = skip_input_data;
+  src->resync_to_restart = jpeg_resync_to_restart; /* use default method */
+  src->term_source = term_source;
+  src->bytes_in_buffer = (size_t)insize;
+  src->next_input_byte = (const JOCTET *)inbuffer;
+}
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdcoefct.c Added

@@ -0,0 +1,878 @@
+/*
+ * jdcoefct.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2010, 2015-2016, 2019-2020, D. R. Commander.
+ * Copyright (C) 2015, 2020, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the coefficient buffer controller for decompression.
+ * This controller is the top level of the JPEG decompressor proper.
+ * The coefficient buffer lies between entropy decoding and inverse-DCT steps.
+ *
+ * In buffered-image mode, this controller is the interface between
+ * input-oriented processing and output-oriented processing.
+ * Also, the input side (only) is used when reading a file for transcoding.
+ */
+
+#include "jinclude.h"
+#include "jdcoefct.h"
+#include "jpegcomp.h"
+
+
+/* Forward declarations */
+METHODDEF(int) decompress_onepass(j_decompress_ptr cinfo,
+                                  JSAMPIMAGE output_buf);
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+METHODDEF(int) decompress_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
+#endif
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+LOCAL(boolean) smoothing_ok(j_decompress_ptr cinfo);
+METHODDEF(int) decompress_smooth_data(j_decompress_ptr cinfo,
+                                      JSAMPIMAGE output_buf);
+#endif
+
+
+/*
+ * Initialize for an input processing pass.
+ */
+
+METHODDEF(void)
+start_input_pass(j_decompress_ptr cinfo)
+{
+  cinfo->input_iMCU_row = 0;
+  start_iMCU_row(cinfo);
+}
+
+
+/*
+ * Initialize for an output processing pass.
+ */
+
+METHODDEF(void)
+start_output_pass(j_decompress_ptr cinfo)
+{
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+
+  /* If multipass, check to see whether to use block smoothing on this pass */
+  if (coef->pub.coef_arrays != NULL) {
+    if (cinfo->do_block_smoothing && smoothing_ok(cinfo))
+      coef->pub.decompress_data = decompress_smooth_data;
+    else
+      coef->pub.decompress_data = decompress_data;
+  }
+#endif
+  cinfo->output_iMCU_row = 0;
+}
+
+
+/*
+ * Decompress and return some data in the single-pass case.
+ * Always attempts to emit one fully interleaved MCU row ("iMCU" row).
+ * Input and output must run in lockstep since we have only a one-MCU buffer.
+ * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED.
+ *
+ * NB: output_buf contains a plane for each component in image,
+ * which we index according to the component's SOF position.
+ */
+
+METHODDEF(int)
+decompress_onepass(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
+  JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  int blkn, ci, xindex, yindex, yoffset, useful_width;
+  JSAMPARRAY output_ptr;
+  JDIMENSION start_col, output_col;
+  jpeg_component_info *compptr;
+  inverse_DCT_method_ptr inverse_DCT;
+
+  /* Loop to process as much as one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->MCU_ctr; MCU_col_num <= last_MCU_col;
+         MCU_col_num++) {
+      /* Try to fetch an MCU.  Entropy decoder expects buffer to be zeroed. */
+      jzero_far((void *)coef->MCU_buffer0,
+                (size_t)(cinfo->blocks_in_MCU * sizeof(JBLOCK)));
+      if (!cinfo->entropy->insufficient_data)
+        cinfo->master->last_good_iMCU_row = cinfo->input_iMCU_row;
+      if (!(*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->MCU_ctr = MCU_col_num;
+        return JPEG_SUSPENDED;
+      }
+
+      /* Only perform the IDCT on blocks that are contained within the desired
+       * cropping region.
+       */
+      if (MCU_col_num >= cinfo->master->first_iMCU_col &&
+          MCU_col_num <= cinfo->master->last_iMCU_col) {
+        /* Determine where data should go in output_buf and do the IDCT thing.
+         * We skip dummy blocks at the right and bottom edges (but blkn gets
+         * incremented past them!).  Note the inner loop relies on having
+         * allocated the MCU_buffer blocks sequentially.
+         */
+        blkn = 0;               /* index of current DCT block within MCU */
+        for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+          compptr = cinfo->cur_comp_infoci;
+          /* Don't bother to IDCT an uninteresting component. */
+          if (!compptr->component_needed) {
+            blkn += compptr->MCU_blocks;
+            continue;
+          }
+          inverse_DCT = cinfo->idct->inverse_DCTcompptr->component_index;
+          useful_width = (MCU_col_num < last_MCU_col) ?
+                         compptr->MCU_width : compptr->last_col_width;
+          output_ptr = output_bufcompptr->component_index +
+                       yoffset * compptr->_DCT_scaled_size;
+          start_col = (MCU_col_num - cinfo->master->first_iMCU_col) *
+                      compptr->MCU_sample_width;
+          for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+            if (cinfo->input_iMCU_row < last_iMCU_row ||
+                yoffset + yindex < compptr->last_row_height) {
+              output_col = start_col;
+              for (xindex = 0; xindex < useful_width; xindex++) {
+                (*inverse_DCT) (cinfo, compptr,
+                                (JCOEFPTR)coef->MCU_bufferblkn + xindex,
+                                output_ptr, output_col);
+                output_col += compptr->_DCT_scaled_size;
+              }
+            }
+            blkn += compptr->MCU_width;
+            output_ptr += compptr->_DCT_scaled_size;
+          }
+        }
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->MCU_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  cinfo->output_iMCU_row++;
+  if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) {
+    start_iMCU_row(cinfo);
+    return JPEG_ROW_COMPLETED;
+  }
+  /* Completed the scan */
+  (*cinfo->inputctl->finish_input_pass) (cinfo);
+  return JPEG_SCAN_COMPLETED;
+}
+
+
+/*
+ * Dummy consume-input routine for single-pass operation.
+ */
+
+METHODDEF(int)
+dummy_consume_data(j_decompress_ptr cinfo)
+{
+  return JPEG_SUSPENDED;        /* Always indicate nothing was done */
+}
+
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+
+/*
+ * Consume input data and store it in the full-image coefficient buffer.
+ * We read as much as one fully interleaved MCU row ("iMCU" row) per call,
+ * ie, v_samp_factor block rows for each component in the scan.
+ * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED.
+ */
+
+METHODDEF(int)
+consume_data(j_decompress_ptr cinfo)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
+  int blkn, ci, xindex, yindex, yoffset;
+  JDIMENSION start_col;
+  JBLOCKARRAY bufferMAX_COMPS_IN_SCAN;
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  /* Align the virtual buffers for the components used in this scan. */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    bufferci = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr)cinfo, coef->whole_imagecompptr->component_index,
+       cinfo->input_iMCU_row * compptr->v_samp_factor,
+       (JDIMENSION)compptr->v_samp_factor, TRUE);
+    /* Note: entropy decoder expects buffer to be zeroed,
+     * but this is handled automatically by the memory manager
+     * because we requested a pre-zeroed array.
+     */
+  }
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->MCU_ctr; MCU_col_num < cinfo->MCUs_per_row;
+         MCU_col_num++) {
+      /* Construct list of pointers to DCT blocks belonging to this MCU */
+      blkn = 0;                 /* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+        compptr = cinfo->cur_comp_infoci;
+        start_col = MCU_col_num * compptr->MCU_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          buffer_ptr = bufferciyindex + yoffset + start_col;
+          for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+            coef->MCU_bufferblkn++ = buffer_ptr++;
+          }
+        }
+      }
+      if (!cinfo->entropy->insufficient_data)
+        cinfo->master->last_good_iMCU_row = cinfo->input_iMCU_row;
+      /* Try to fetch the MCU. */
+      if (!(*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->MCU_ctr = MCU_col_num;
+        return JPEG_SUSPENDED;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->MCU_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) {
+    start_iMCU_row(cinfo);
+    return JPEG_ROW_COMPLETED;
+  }
+  /* Completed the scan */
+  (*cinfo->inputctl->finish_input_pass) (cinfo);
+  return JPEG_SCAN_COMPLETED;
+}
+
+
+/*
+ * Decompress and return some data in the multi-pass case.
+ * Always attempts to emit one fully interleaved MCU row ("iMCU" row).
+ * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED.
+ *
+ * NB: output_buf contains a plane for each component in image.
+ */
+
+METHODDEF(int)
+decompress_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  JDIMENSION block_num;
+  int ci, block_row, block_rows;
+  JBLOCKARRAY buffer;
+  JBLOCKROW buffer_ptr;
+  JSAMPARRAY output_ptr;
+  JDIMENSION output_col;
+  jpeg_component_info *compptr;
+  inverse_DCT_method_ptr inverse_DCT;
+
+  /* Force some input to be done if we are getting ahead of the input. */
+  while (cinfo->input_scan_number < cinfo->output_scan_number ||
+         (cinfo->input_scan_number == cinfo->output_scan_number &&
+          cinfo->input_iMCU_row <= cinfo->output_iMCU_row)) {
+    if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
+      return JPEG_SUSPENDED;
+  }
+
+  /* OK, output from the virtual arrays. */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Don't bother to IDCT an uninteresting component. */
+    if (!compptr->component_needed)
+      continue;
+    /* Align the virtual buffer for this component. */
+    buffer = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr)cinfo, coef->whole_imageci,
+       cinfo->output_iMCU_row * compptr->v_samp_factor,
+       (JDIMENSION)compptr->v_samp_factor, FALSE);
+    /* Count non-dummy DCT block rows in this iMCU row. */
+    if (cinfo->output_iMCU_row < last_iMCU_row)
+      block_rows = compptr->v_samp_factor;
+    else {
+      /* NB: can't use last_row_height here; it is input-side-dependent! */
+      block_rows = (int)(compptr->height_in_blocks % compptr->v_samp_factor);
+      if (block_rows == 0) block_rows = compptr->v_samp_factor;
+    }
+    inverse_DCT = cinfo->idct->inverse_DCTci;
+    output_ptr = output_bufci;
+    /* Loop over all DCT blocks to be processed. */
+    for (block_row = 0; block_row < block_rows; block_row++) {
+      buffer_ptr = bufferblock_row + cinfo->master->first_MCU_colci;
+      output_col = 0;
+      for (block_num = cinfo->master->first_MCU_colci;
+           block_num <= cinfo->master->last_MCU_colci; block_num++) {
+        (*inverse_DCT) (cinfo, compptr, (JCOEFPTR)buffer_ptr, output_ptr,
+                        output_col);
+        buffer_ptr++;
+        output_col += compptr->_DCT_scaled_size;
+      }
+      output_ptr += compptr->_DCT_scaled_size;
+    }
+  }
+
+  if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows)
+    return JPEG_ROW_COMPLETED;
+  return JPEG_SCAN_COMPLETED;
+}
+
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+
+
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+
+/*
+ * This code applies interblock smoothing; the first 9 AC coefficients are
+ * estimated from the DC values of a DCT block and its 24 neighboring blocks.
+ * We apply smoothing only for progressive JPEG decoding, and only if
+ * the coefficients it can estimate are not yet known to full precision.
+ */
+
+/* Natural-order array positions of the first 9 zigzag-order coefficients */
+#define Q01_POS  1
+#define Q10_POS  8
+#define Q20_POS  16
+#define Q11_POS  9
+#define Q02_POS  2
+#define Q03_POS  3
+#define Q12_POS  10
+#define Q21_POS  17
+#define Q30_POS  24
+
+/*
+ * Determine whether block smoothing is applicable and safe.
+ * We also latch the current states of the coef_bits entries for the
+ * AC coefficients; otherwise, if the input side of the decompressor
+ * advances into a new scan, we might think the coefficients are known
+ * more accurately than they really are.
+ */
+
+LOCAL(boolean)
+smoothing_ok(j_decompress_ptr cinfo)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  boolean smoothing_useful = FALSE;
+  int ci, coefi;
+  jpeg_component_info *compptr;
+  JQUANT_TBL *qtable;
+  int *coef_bits, *prev_coef_bits;
+  int *coef_bits_latch, *prev_coef_bits_latch;
+
+  if (!cinfo->progressive_mode || cinfo->coef_bits == NULL)
+    return FALSE;
+
+  /* Allocate latch area if not already done */
+  if (coef->coef_bits_latch == NULL)
+    coef->coef_bits_latch = (int *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  cinfo->num_components * 2 *
+                                  (SAVED_COEFS * sizeof(int)));
+  coef_bits_latch = coef->coef_bits_latch;
+  prev_coef_bits_latch =
+    &coef->coef_bits_latchcinfo->num_components * SAVED_COEFS;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* All components' quantization values must already be latched. */
+    if ((qtable = compptr->quant_table) == NULL)
+      return FALSE;
+    /* Verify DC & first 9 AC quantizers are nonzero to avoid zero-divide. */
+    if (qtable->quantval0 == 0 ||
+        qtable->quantvalQ01_POS == 0 ||
+        qtable->quantvalQ10_POS == 0 ||
+        qtable->quantvalQ20_POS == 0 ||
+        qtable->quantvalQ11_POS == 0 ||
+        qtable->quantvalQ02_POS == 0 ||
+        qtable->quantvalQ03_POS == 0 ||
+        qtable->quantvalQ12_POS == 0 ||
+        qtable->quantvalQ21_POS == 0 ||
+        qtable->quantvalQ30_POS == 0)
+      return FALSE;
+    /* DC values must be at least partly known for all components. */
+    coef_bits = cinfo->coef_bitsci;
+    prev_coef_bits = cinfo->coef_bitsci + cinfo->num_components;
+    if (coef_bits0 < 0)
+      return FALSE;
+    coef_bits_latch0 = coef_bits0;
+    /* Block smoothing is helpful if some AC coefficients remain inaccurate. */
+    for (coefi = 1; coefi < SAVED_COEFS; coefi++) {
+      if (cinfo->input_scan_number > 1)
+        prev_coef_bits_latchcoefi = prev_coef_bitscoefi;
+      else
+        prev_coef_bits_latchcoefi = -1;
+      coef_bits_latchcoefi = coef_bitscoefi;
+      if (coef_bitscoefi != 0)
+        smoothing_useful = TRUE;
+    }
+    coef_bits_latch += SAVED_COEFS;
+    prev_coef_bits_latch += SAVED_COEFS;
+  }
+
+  return smoothing_useful;
+}
+
+
+/*
+ * Variant of decompress_data for use when doing block smoothing.
+ */
+
+METHODDEF(int)
+decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  JDIMENSION block_num, last_block_column;
+  int ci, block_row, block_rows, access_rows;
+  JBLOCKARRAY buffer;
+  JBLOCKROW buffer_ptr, prev_prev_block_row, prev_block_row;
+  JBLOCKROW next_block_row, next_next_block_row;
+  JSAMPARRAY output_ptr;
+  JDIMENSION output_col;
+  jpeg_component_info *compptr;
+  inverse_DCT_method_ptr inverse_DCT;
+  boolean change_dc;
+  JCOEF *workspace;
+  int *coef_bits;
+  JQUANT_TBL *quanttbl;
+  JLONG Q00, Q01, Q02, Q03 = 0, Q10, Q11, Q12 = 0, Q20, Q21 = 0, Q30 = 0, num;
+  int DC01, DC02, DC03, DC04, DC05, DC06, DC07, DC08, DC09, DC10, DC11, DC12,
+      DC13, DC14, DC15, DC16, DC17, DC18, DC19, DC20, DC21, DC22, DC23, DC24,
+      DC25;
+  int Al, pred;
+
+  /* Keep a local variable to avoid looking it up more than once */
+  workspace = coef->workspace;
+
+  /* Force some input to be done if we are getting ahead of the input. */
+  while (cinfo->input_scan_number <= cinfo->output_scan_number &&
+         !cinfo->inputctl->eoi_reached) {
+    if (cinfo->input_scan_number == cinfo->output_scan_number) {
+      /* If input is working on current scan, we ordinarily want it to
+       * have completed the current row.  But if input scan is DC,
+       * we want it to keep two rows ahead so that next two block rows' DC
+       * values are up to date.
+       */
+      JDIMENSION delta = (cinfo->Ss == 0) ? 2 : 0;
+      if (cinfo->input_iMCU_row > cinfo->output_iMCU_row + delta)
+        break;
+    }
+    if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
+      return JPEG_SUSPENDED;
+  }
+
+  /* OK, output from the virtual arrays. */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Don't bother to IDCT an uninteresting component. */
+    if (!compptr->component_needed)
+      continue;
+    /* Count non-dummy DCT block rows in this iMCU row. */
+    if (cinfo->output_iMCU_row < last_iMCU_row - 1) {
+      block_rows = compptr->v_samp_factor;
+      access_rows = block_rows * 3; /* this and next two iMCU rows */
+    } else if (cinfo->output_iMCU_row < last_iMCU_row) {
+      block_rows = compptr->v_samp_factor;
+      access_rows = block_rows * 2; /* this and next iMCU row */
+    } else {
+      /* NB: can't use last_row_height here; it is input-side-dependent! */
+      block_rows = (int)(compptr->height_in_blocks % compptr->v_samp_factor);
+      if (block_rows == 0) block_rows = compptr->v_samp_factor;
+      access_rows = block_rows; /* this iMCU row only */
+    }
+    /* Align the virtual buffer for this component. */
+    if (cinfo->output_iMCU_row > 1) {
+      access_rows += 2 * compptr->v_samp_factor; /* prior two iMCU rows too */
+      buffer = (*cinfo->mem->access_virt_barray)
+        ((j_common_ptr)cinfo, coef->whole_imageci,
+         (cinfo->output_iMCU_row - 2) * compptr->v_samp_factor,
+         (JDIMENSION)access_rows, FALSE);
+      buffer += 2 * compptr->v_samp_factor; /* point to current iMCU row */
+    } else if (cinfo->output_iMCU_row > 0) {
+      buffer = (*cinfo->mem->access_virt_barray)
+        ((j_common_ptr)cinfo, coef->whole_imageci,
+         (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor,
+         (JDIMENSION)access_rows, FALSE);
+      buffer += compptr->v_samp_factor; /* point to current iMCU row */
+    } else {
+      buffer = (*cinfo->mem->access_virt_barray)
+        ((j_common_ptr)cinfo, coef->whole_imageci,
+         (JDIMENSION)0, (JDIMENSION)access_rows, FALSE);
+    }
+    /* Fetch component-dependent info.
+     * If the current scan is incomplete, then we use the component-dependent
+     * info from the previous scan.
+     */
+    if (cinfo->output_iMCU_row > cinfo->master->last_good_iMCU_row)
+      coef_bits =
+        coef->coef_bits_latch + ((ci + cinfo->num_components) * SAVED_COEFS);
+    else
+      coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS);
+
+    /* We only do DC interpolation if no AC coefficient data is available. */
+    change_dc =
+      coef_bits1 == -1 && coef_bits2 == -1 && coef_bits3 == -1 &&
+      coef_bits4 == -1 && coef_bits5 == -1 && coef_bits6 == -1 &&
+      coef_bits7 == -1 && coef_bits8 == -1 && coef_bits9 == -1;
+
+    quanttbl = compptr->quant_table;
+    Q00 = quanttbl->quantval0;
+    Q01 = quanttbl->quantvalQ01_POS;
+    Q10 = quanttbl->quantvalQ10_POS;
+    Q20 = quanttbl->quantvalQ20_POS;
+    Q11 = quanttbl->quantvalQ11_POS;
+    Q02 = quanttbl->quantvalQ02_POS;
+    if (change_dc) {
+      Q03 = quanttbl->quantvalQ03_POS;
+      Q12 = quanttbl->quantvalQ12_POS;
+      Q21 = quanttbl->quantvalQ21_POS;
+      Q30 = quanttbl->quantvalQ30_POS;
+    }
+    inverse_DCT = cinfo->idct->inverse_DCTci;
+    output_ptr = output_bufci;
+    /* Loop over all DCT blocks to be processed. */
+    for (block_row = 0; block_row < block_rows; block_row++) {
+      buffer_ptr = bufferblock_row + cinfo->master->first_MCU_colci;
+
+      if (block_row > 0 || cinfo->output_iMCU_row > 0)
+        prev_block_row =
+          bufferblock_row - 1 + cinfo->master->first_MCU_colci;
+      else
+        prev_block_row = buffer_ptr;
+
+      if (block_row > 1 || cinfo->output_iMCU_row > 1)
+        prev_prev_block_row =
+          bufferblock_row - 2 + cinfo->master->first_MCU_colci;
+      else
+        prev_prev_block_row = prev_block_row;
+
+      if (block_row < block_rows - 1 || cinfo->output_iMCU_row < last_iMCU_row)
+        next_block_row =
+          bufferblock_row + 1 + cinfo->master->first_MCU_colci;
+      else
+        next_block_row = buffer_ptr;
+
+      if (block_row < block_rows - 2 ||
+          cinfo->output_iMCU_row < last_iMCU_row - 1)
+        next_next_block_row =
+          bufferblock_row + 2 + cinfo->master->first_MCU_colci;
+      else
+        next_next_block_row = next_block_row;
+
+      /* We fetch the surrounding DC values using a sliding-register approach.
+       * Initialize all 25 here so as to do the right thing on narrow pics.
+       */
+      DC01 = DC02 = DC03 = DC04 = DC05 = (int)prev_prev_block_row00;
+      DC06 = DC07 = DC08 = DC09 = DC10 = (int)prev_block_row00;
+      DC11 = DC12 = DC13 = DC14 = DC15 = (int)buffer_ptr00;
+      DC16 = DC17 = DC18 = DC19 = DC20 = (int)next_block_row00;
+      DC21 = DC22 = DC23 = DC24 = DC25 = (int)next_next_block_row00;
+      output_col = 0;
+      last_block_column = compptr->width_in_blocks - 1;
+      for (block_num = cinfo->master->first_MCU_colci;
+           block_num <= cinfo->master->last_MCU_colci; block_num++) {
+        /* Fetch current DCT block into workspace so we can modify it. */
+        jcopy_block_row(buffer_ptr, (JBLOCKROW)workspace, (JDIMENSION)1);
+        /* Update DC values */
+        if (block_num == cinfo->master->first_MCU_colci &&
+            block_num < last_block_column) {
+          DC04 = (int)prev_prev_block_row10;
+          DC09 = (int)prev_block_row10;
+          DC14 = (int)buffer_ptr10;
+          DC19 = (int)next_block_row10;
+          DC24 = (int)next_next_block_row10;
+        }
+        if (block_num + 1 < last_block_column) {
+          DC05 = (int)prev_prev_block_row20;
+          DC10 = (int)prev_block_row20;
+          DC15 = (int)buffer_ptr20;
+          DC20 = (int)next_block_row20;
+          DC25 = (int)next_next_block_row20;
+        }
+        /* If DC interpolation is enabled, compute coefficient estimates using
+         * a Gaussian-like kernel, keeping the averages of the DC values.
+         *
+         * If DC interpolation is disabled, compute coefficient estimates using
+         * an algorithm similar to the one described in Section K.8 of the JPEG
+         * standard, except applied to a 5x5 window rather than a 3x3 window.
+         *
+         * An estimate is applied only if the coefficient is still zero and is
+         * not known to be fully accurate.
+         */
+        /* AC01 */
+        if ((Al = coef_bits1) != 0 && workspace1 == 0) {
+          num = Q00 * (change_dc ?
+                (-DC01 - DC02 + DC04 + DC05 - 3 * DC06 + 13 * DC07 -
+                 13 * DC09 + 3 * DC10 - 3 * DC11 + 38 * DC12 - 38 * DC14 +
+                 3 * DC15 - 3 * DC16 + 13 * DC17 - 13 * DC19 + 3 * DC20 -
+                 DC21 - DC22 + DC24 + DC25) :
+                (-7 * DC11 + 50 * DC12 - 50 * DC14 + 7 * DC15));
+          if (num >= 0) {
+            pred = (int)(((Q01 << 7) + num) / (Q01 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+          } else {
+            pred = (int)(((Q01 << 7) - num) / (Q01 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+            pred = -pred;
+          }
+          workspace1 = (JCOEF)pred;
+        }
+        /* AC10 */
+        if ((Al = coef_bits2) != 0 && workspace8 == 0) {
+          num = Q00 * (change_dc ?
+                (-DC01 - 3 * DC02 - 3 * DC03 - 3 * DC04 - DC05 - DC06 +
+                 13 * DC07 + 38 * DC08 + 13 * DC09 - DC10 + DC16 -
+                 13 * DC17 - 38 * DC18 - 13 * DC19 + DC20 + DC21 +
+                 3 * DC22 + 3 * DC23 + 3 * DC24 + DC25) :
+                (-7 * DC03 + 50 * DC08 - 50 * DC18 + 7 * DC23));
+          if (num >= 0) {
+            pred = (int)(((Q10 << 7) + num) / (Q10 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+          } else {
+            pred = (int)(((Q10 << 7) - num) / (Q10 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+            pred = -pred;
+          }
+          workspace8 = (JCOEF)pred;
+        }
+        /* AC20 */
+        if ((Al = coef_bits3) != 0 && workspace16 == 0) {
+          num = Q00 * (change_dc ?
+                (DC03 + 2 * DC07 + 7 * DC08 + 2 * DC09 - 5 * DC12 - 14 * DC13 -
+                 5 * DC14 + 2 * DC17 + 7 * DC18 + 2 * DC19 + DC23) :
+                (-DC03 + 13 * DC08 - 24 * DC13 + 13 * DC18 - DC23));
+          if (num >= 0) {
+            pred = (int)(((Q20 << 7) + num) / (Q20 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+          } else {
+            pred = (int)(((Q20 << 7) - num) / (Q20 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+            pred = -pred;
+          }
+          workspace16 = (JCOEF)pred;
+        }
+        /* AC11 */
+        if ((Al = coef_bits4) != 0 && workspace9 == 0) {
+          num = Q00 * (change_dc ?
+                (-DC01 + DC05 + 9 * DC07 - 9 * DC09 - 9 * DC17 +
+                 9 * DC19 + DC21 - DC25) :
+                (DC10 + DC16 - 10 * DC17 + 10 * DC19 - DC02 - DC20 + DC22 -
+                 DC24 + DC04 - DC06 + 10 * DC07 - 10 * DC09));
+          if (num >= 0) {
+            pred = (int)(((Q11 << 7) + num) / (Q11 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+          } else {
+            pred = (int)(((Q11 << 7) - num) / (Q11 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+            pred = -pred;
+          }
+          workspace9 = (JCOEF)pred;
+        }
+        /* AC02 */
+        if ((Al = coef_bits5) != 0 && workspace2 == 0) {
+          num = Q00 * (change_dc ?
+                (2 * DC07 - 5 * DC08 + 2 * DC09 + DC11 + 7 * DC12 - 14 * DC13 +
+                 7 * DC14 + DC15 + 2 * DC17 - 5 * DC18 + 2 * DC19) :
+                (-DC11 + 13 * DC12 - 24 * DC13 + 13 * DC14 - DC15));
+          if (num >= 0) {
+            pred = (int)(((Q02 << 7) + num) / (Q02 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+          } else {
+            pred = (int)(((Q02 << 7) - num) / (Q02 << 8));
+            if (Al > 0 && pred >= (1 << Al))
+              pred = (1 << Al) - 1;
+            pred = -pred;
+          }
+          workspace2 = (JCOEF)pred;
+        }
+        if (change_dc) {
+          /* AC03 */
+          if ((Al = coef_bits6) != 0 && workspace3 == 0) {
+            num = Q00 * (DC07 - DC09 + 2 * DC12 - 2 * DC14 + DC17 - DC19);
+            if (num >= 0) {
+              pred = (int)(((Q03 << 7) + num) / (Q03 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+            } else {
+              pred = (int)(((Q03 << 7) - num) / (Q03 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+              pred = -pred;
+            }
+            workspace3 = (JCOEF)pred;
+          }
+          /* AC12 */
+          if ((Al = coef_bits7) != 0 && workspace10 == 0) {
+            num = Q00 * (DC07 - 3 * DC08 + DC09 - DC17 + 3 * DC18 - DC19);
+            if (num >= 0) {
+              pred = (int)(((Q12 << 7) + num) / (Q12 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+            } else {
+              pred = (int)(((Q12 << 7) - num) / (Q12 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+              pred = -pred;
+            }
+            workspace10 = (JCOEF)pred;
+          }
+          /* AC21 */
+          if ((Al = coef_bits8) != 0 && workspace17 == 0) {
+            num = Q00 * (DC07 - DC09 - 3 * DC12 + 3 * DC14 + DC17 - DC19);
+            if (num >= 0) {
+              pred = (int)(((Q21 << 7) + num) / (Q21 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+            } else {
+              pred = (int)(((Q21 << 7) - num) / (Q21 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+              pred = -pred;
+            }
+            workspace17 = (JCOEF)pred;
+          }
+          /* AC30 */
+          if ((Al = coef_bits9) != 0 && workspace24 == 0) {
+            num = Q00 * (DC07 + 2 * DC08 + DC09 - DC17 - 2 * DC18 - DC19);
+            if (num >= 0) {
+              pred = (int)(((Q30 << 7) + num) / (Q30 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+            } else {
+              pred = (int)(((Q30 << 7) - num) / (Q30 << 8));
+              if (Al > 0 && pred >= (1 << Al))
+                pred = (1 << Al) - 1;
+              pred = -pred;
+            }
+            workspace24 = (JCOEF)pred;
+          }
+          /* coef_bits0 is non-negative.  Otherwise this function would not
+           * be called.
+           */
+          num = Q00 *
+                (-2 * DC01 - 6 * DC02 - 8 * DC03 - 6 * DC04 - 2 * DC05 -
+                 6 * DC06 + 6 * DC07 + 42 * DC08 + 6 * DC09 - 6 * DC10 -
+                 8 * DC11 + 42 * DC12 + 152 * DC13 + 42 * DC14 - 8 * DC15 -
+                 6 * DC16 + 6 * DC17 + 42 * DC18 + 6 * DC19 - 6 * DC20 -
+                 2 * DC21 - 6 * DC22 - 8 * DC23 - 6 * DC24 - 2 * DC25);
+          if (num >= 0) {
+            pred = (int)(((Q00 << 7) + num) / (Q00 << 8));
+          } else {
+            pred = (int)(((Q00 << 7) - num) / (Q00 << 8));
+            pred = -pred;
+          }
+          workspace0 = (JCOEF)pred;
+        }  /* change_dc */
+
+        /* OK, do the IDCT */
+        (*inverse_DCT) (cinfo, compptr, (JCOEFPTR)workspace, output_ptr,
+                        output_col);
+        /* Advance for next column */
+        DC01 = DC02;  DC02 = DC03;  DC03 = DC04;  DC04 = DC05;
+        DC06 = DC07;  DC07 = DC08;  DC08 = DC09;  DC09 = DC10;
+        DC11 = DC12;  DC12 = DC13;  DC13 = DC14;  DC14 = DC15;
+        DC16 = DC17;  DC17 = DC18;  DC18 = DC19;  DC19 = DC20;
+        DC21 = DC22;  DC22 = DC23;  DC23 = DC24;  DC24 = DC25;
+        buffer_ptr++, prev_block_row++, next_block_row++,
+          prev_prev_block_row++, next_next_block_row++;
+        output_col += compptr->_DCT_scaled_size;
+      }
+      output_ptr += compptr->_DCT_scaled_size;
+    }
+  }
+
+  if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows)
+    return JPEG_ROW_COMPLETED;
+  return JPEG_SCAN_COMPLETED;
+}
+
+#endif /* BLOCK_SMOOTHING_SUPPORTED */
+
+
+/*
+ * Initialize coefficient buffer controller.
+ */
+
+GLOBAL(void)
+jinit_d_coef_controller(j_decompress_ptr cinfo, boolean need_full_buffer)
+{
+  my_coef_ptr coef;
+
+  coef = (my_coef_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_coef_controller));
+  cinfo->coef = (struct jpeg_d_coef_controller *)coef;
+  coef->pub.start_input_pass = start_input_pass;
+  coef->pub.start_output_pass = start_output_pass;
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  coef->coef_bits_latch = NULL;
+#endif
+
+  /* Create the coefficient buffer. */
+  if (need_full_buffer) {
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+    /* Allocate a full-image virtual array for each component, */
+    /* padded to a multiple of samp_factor DCT blocks in each direction. */
+    /* Note we ask for a pre-zeroed array. */
+    int ci, access_rows;
+    jpeg_component_info *compptr;
+
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+         ci++, compptr++) {
+      access_rows = compptr->v_samp_factor;
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+      /* If block smoothing could be used, need a bigger window */
+      if (cinfo->progressive_mode)
+        access_rows *= 5;
+#endif
+      coef->whole_imageci = (*cinfo->mem->request_virt_barray)
+        ((j_common_ptr)cinfo, JPOOL_IMAGE, TRUE,
+         (JDIMENSION)jround_up((long)compptr->width_in_blocks,
+                               (long)compptr->h_samp_factor),
+         (JDIMENSION)jround_up((long)compptr->height_in_blocks,
+                               (long)compptr->v_samp_factor),
+         (JDIMENSION)access_rows);
+    }
+    coef->pub.consume_data = consume_data;
+    coef->pub.decompress_data = decompress_data;
+    coef->pub.coef_arrays = coef->whole_image; /* link to virtual arrays */
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    /* We only need a single-MCU buffer. */
+    JBLOCKROW buffer;
+    int i;
+
+    buffer = (JBLOCKROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  D_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
+    for (i = 0; i < D_MAX_BLOCKS_IN_MCU; i++) {
+      coef->MCU_bufferi = buffer + i;
+    }
+    coef->pub.consume_data = dummy_consume_data;
+    coef->pub.decompress_data = decompress_onepass;
+    coef->pub.coef_arrays = NULL; /* flag for no virtual arrays */
+  }
+
+  /* Allocate the workspace buffer */
+  coef->workspace = (JCOEF *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(JCOEF) * DCTSIZE2);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdcoefct.h Added

@@ -0,0 +1,83 @@
+/*
+ * jdcoefct.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2020, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+
+
+/* Block smoothing is only applicable for progressive JPEG, so: */
+#ifndef D_PROGRESSIVE_SUPPORTED
+#undef BLOCK_SMOOTHING_SUPPORTED
+#endif
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_coef_controller pub; /* public fields */
+
+  /* These variables keep track of the current location of the input side. */
+  /* cinfo->input_iMCU_row is also used for this. */
+  JDIMENSION MCU_ctr;           /* counts MCUs processed in current row */
+  int MCU_vert_offset;          /* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;    /* number of such rows needed */
+
+  /* The output side's location is represented by cinfo->output_iMCU_row. */
+
+  /* In single-pass modes, it's sufficient to buffer just one MCU.
+   * We allocate a workspace of D_MAX_BLOCKS_IN_MCU coefficient blocks,
+   * and let the entropy decoder write into that workspace each time.
+   * In multi-pass modes, this array points to the current MCU's blocks
+   * within the virtual arrays; it is used only by the input side.
+   */
+  JBLOCKROW MCU_bufferD_MAX_BLOCKS_IN_MCU;
+
+  /* Temporary workspace for one MCU */
+  JCOEF *workspace;
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+  /* In multi-pass modes, we need a virtual block array for each component. */
+  jvirt_barray_ptr whole_imageMAX_COMPONENTS;
+#endif
+
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  /* When doing block smoothing, we latch coefficient Al values here */
+  int *coef_bits_latch;
+#define SAVED_COEFS  10         /* we save coef_bits0..9 */
+#endif
+} my_coef_controller;
+
+typedef my_coef_controller *my_coef_ptr;
+
+
+LOCAL(void)
+start_iMCU_row(j_decompress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row (input side) */
+{
+  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (cinfo->input_iMCU_row < (cinfo->total_iMCU_rows - 1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info0->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info0->last_row_height;
+  }
+
+  coef->MCU_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdcol565.c Added

@@ -0,0 +1,384 @@
+/*
+ * jdcol565.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modifications:
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+/* This file is included by jdcolor.c */
+
+
+INLINE
+LOCAL(void)
+ycc_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                            JDIMENSION input_row, JSAMPARRAY output_buf,
+                            int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  register int *Crrtab = cconvert->Cr_r_tab;
+  register int *Cbbtab = cconvert->Cb_b_tab;
+  register JLONG *Crgtab = cconvert->Cr_g_tab;
+  register JLONG *Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
+      r = range_limity + Crrtabcr;
+      g = range_limity + ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                            SCALEBITS));
+      b = range_limity + Cbbtabcb;
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16 *)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
+      r = range_limity + Crrtabcr;
+      g = range_limity + ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                            SCALEBITS));
+      b = range_limity + Cbbtabcb;
+      rgb = PACK_SHORT_565(r, g, b);
+
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
+      r = range_limity + Crrtabcr;
+      g = range_limity + ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                            SCALEBITS));
+      b = range_limity + Cbbtabcb;
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      y  = *inptr0;
+      cb = *inptr1;
+      cr = *inptr2;
+      r = range_limity + Crrtabcr;
+      g = range_limity + ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                            SCALEBITS));
+      b = range_limity + Cbbtabcb;
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16 *)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+ycc_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                             JDIMENSION input_row, JSAMPARRAY output_buf,
+                             int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  register int *Crrtab = cconvert->Cr_r_tab;
+  register int *Cbbtab = cconvert->Cb_b_tab;
+  register JLONG *Crgtab = cconvert->Cr_g_tab;
+  register JLONG *Cbgtab = cconvert->Cb_g_tab;
+  JLONG d0 = dither_matrixcinfo->output_scanline & DITHER_MASK;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
+      r = range_limitDITHER_565_R(y + Crrtabcr, d0);
+      g = range_limitDITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                                     SCALEBITS)), d0);
+      b = range_limitDITHER_565_B(y + Cbbtabcb, d0);
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16 *)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
+      r = range_limitDITHER_565_R(y + Crrtabcr, d0);
+      g = range_limitDITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                                     SCALEBITS)), d0);
+      b = range_limitDITHER_565_B(y + Cbbtabcb, d0);
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_SHORT_565(r, g, b);
+
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *inptr2++;
+      r = range_limitDITHER_565_R(y + Crrtabcr, d0);
+      g = range_limitDITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                                     SCALEBITS)), d0);
+      b = range_limitDITHER_565_B(y + Cbbtabcb, d0);
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      y  = *inptr0;
+      cb = *inptr1;
+      cr = *inptr2;
+      r = range_limitDITHER_565_R(y + Crrtabcr, d0);
+      g = range_limitDITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                                     SCALEBITS)), d0);
+      b = range_limitDITHER_565_B(y + Cbbtabcb, d0);
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16 *)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+rgb_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                            JDIMENSION input_row, JSAMPARRAY output_buf,
+                            int num_rows)
+{
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      r = *inptr0++;
+      g = *inptr1++;
+      b = *inptr2++;
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16 *)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      r = *inptr0++;
+      g = *inptr1++;
+      b = *inptr2++;
+      rgb = PACK_SHORT_565(r, g, b);
+
+      r = *inptr0++;
+      g = *inptr1++;
+      b = *inptr2++;
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      r = *inptr0;
+      g = *inptr1;
+      b = *inptr2;
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16 *)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+rgb_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                             JDIMENSION input_row, JSAMPARRAY output_buf,
+                             int num_rows)
+{
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  JDIMENSION num_cols = cinfo->output_width;
+  JLONG d0 = dither_matrixcinfo->output_scanline & DITHER_MASK;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      r = range_limitDITHER_565_R(*inptr0++, d0);
+      g = range_limitDITHER_565_G(*inptr1++, d0);
+      b = range_limitDITHER_565_B(*inptr2++, d0);
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16 *)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      r = range_limitDITHER_565_R(*inptr0++, d0);
+      g = range_limitDITHER_565_G(*inptr1++, d0);
+      b = range_limitDITHER_565_B(*inptr2++, d0);
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_SHORT_565(r, g, b);
+
+      r = range_limitDITHER_565_R(*inptr0++, d0);
+      g = range_limitDITHER_565_G(*inptr1++, d0);
+      b = range_limitDITHER_565_B(*inptr2++, d0);
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      r = range_limitDITHER_565_R(*inptr0, d0);
+      g = range_limitDITHER_565_G(*inptr1, d0);
+      b = range_limitDITHER_565_B(*inptr2, d0);
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16 *)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+gray_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                             JDIMENSION input_row, JSAMPARRAY output_buf,
+                             int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int g;
+
+    inptr = input_buf0input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      g = *inptr++;
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16 *)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      g = *inptr++;
+      rgb = PACK_SHORT_565(g, g, g);
+      g = *inptr++;
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g));
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      g = *inptr;
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16 *)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+gray_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                              JDIMENSION input_row, JSAMPARRAY output_buf,
+                              int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  JDIMENSION num_cols = cinfo->output_width;
+  JLONG d0 = dither_matrixcinfo->output_scanline & DITHER_MASK;
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int g;
+
+    inptr = input_buf0input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      g = *inptr++;
+      g = range_limitDITHER_565_R(g, d0);
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16 *)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      g = *inptr++;
+      g = range_limitDITHER_565_R(g, d0);
+      rgb = PACK_SHORT_565(g, g, g);
+      d0 = DITHER_ROTATE(d0);
+
+      g = *inptr++;
+      g = range_limitDITHER_565_R(g, d0);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g));
+      d0 = DITHER_ROTATE(d0);
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      g = *inptr;
+      g = range_limitDITHER_565_R(g, d0);
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16 *)outptr = (INT16)rgb;
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdcolext.c Added

@@ -0,0 +1,141 @@
+/*
+ * jdcolext.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2011, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+
+/* This file is included by jdcolor.c */
+
+
+/*
+ * Convert some rows of samples to the output colorspace.
+ *
+ * Note that we change from noninterleaved, one-plane-per-component format
+ * to interleaved-pixel format.  The output buffer is therefore three times
+ * as wide as the input buffer.
+ * A starting row offset is provided only for the input buffer.  The caller
+ * can easily adjust the passed output_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+INLINE
+LOCAL(void)
+ycc_rgb_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  register int *Crrtab = cconvert->Cr_r_tab;
+  register int *Cbbtab = cconvert->Cb_b_tab;
+  register JLONG *Crgtab = cconvert->Cr_g_tab;
+  register JLONG *Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = inptr0col;
+      cb = inptr1col;
+      cr = inptr2col;
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptrRGB_RED =   range_limity + Crrtabcr;
+      outptrRGB_GREEN = range_limity +
+                              ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                                SCALEBITS));
+      outptrRGB_BLUE =  range_limity + Cbbtabcb;
+      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
+      /* alpha channel value */
+#ifdef RGB_ALPHA
+      outptrRGB_ALPHA = 0xFF;
+#endif
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+
+/*
+ * Convert grayscale to RGB: just duplicate the graylevel three times.
+ * This is provided to support applications that don't want to cope
+ * with grayscale as a separate case.
+ */
+
+INLINE
+LOCAL(void)
+gray_rgb_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                          JDIMENSION input_row, JSAMPARRAY output_buf,
+                          int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr = input_buf0input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      outptrRGB_RED = outptrRGB_GREEN = outptrRGB_BLUE = inptrcol;
+      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
+      /* alpha channel value */
+#ifdef RGB_ALPHA
+      outptrRGB_ALPHA = 0xFF;
+#endif
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+
+/*
+ * Convert RGB to extended RGB: just swap the order of source pixels
+ */
+
+INLINE
+LOCAL(void)
+rgb_rgb_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      outptrRGB_RED = inptr0col;
+      outptrRGB_GREEN = inptr1col;
+      outptrRGB_BLUE = inptr2col;
+      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
+      /* alpha channel value */
+#ifdef RGB_ALPHA
+      outptrRGB_ALPHA = 0xFF;
+#endif
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdcolor.c Added

@@ -0,0 +1,882 @@
+/*
+ * jdcolor.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2009, 2011-2012, 2014-2015, D. R. Commander.
+ * Copyright (C) 2013, Linaro Limited.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jsimd.h"
+#include "jconfigint.h"
+
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_color_deconverter pub; /* public fields */
+
+  /* Private state for YCC->RGB conversion */
+  int *Cr_r_tab;                /* => table for Cr to R conversion */
+  int *Cb_b_tab;                /* => table for Cb to B conversion */
+  JLONG *Cr_g_tab;              /* => table for Cr to G conversion */
+  JLONG *Cb_g_tab;              /* => table for Cb to G conversion */
+
+  /* Private state for RGB->Y conversion */
+  JLONG *rgb_y_tab;             /* => table for RGB to Y conversion */
+} my_color_deconverter;
+
+typedef my_color_deconverter *my_cconvert_ptr;
+
+
+/**************** YCbCr -> RGB conversion: most common case **************/
+/****************   RGB -> Y   conversion: less common case **************/
+
+/*
+ * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
+ * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
+ * The conversion equations to be implemented are therefore
+ *
+ *      R = Y                + 1.40200 * Cr
+ *      G = Y - 0.34414 * Cb - 0.71414 * Cr
+ *      B = Y + 1.77200 * Cb
+ *
+ *      Y = 0.29900 * R + 0.58700 * G + 0.11400 * B
+ *
+ * where Cb and Cr represent the incoming values less CENTERJSAMPLE.
+ * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
+ *
+ * To avoid floating-point arithmetic, we represent the fractional constants
+ * as integers scaled up by 2^16 (about 4 digits precision); we have to divide
+ * the products by 2^16, with appropriate rounding, to get the correct answer.
+ * Notice that Y, being an integral input, does not contribute any fraction
+ * so it need not participate in the rounding.
+ *
+ * For even more speed, we avoid doing any multiplications in the inner loop
+ * by precalculating the constants times Cb and Cr for all possible values.
+ * For 8-bit JSAMPLEs this is very reasonable (only 256 entries per table);
+ * for 12-bit samples it is still acceptable.  It's not very reasonable for
+ * 16-bit samples, but if you want lossless storage you shouldn't be changing
+ * colorspace anyway.
+ * The Cr=>R and Cb=>B values can be rounded to integers in advance; the
+ * values for the G calculation are left scaled up, since we must add them
+ * together before rounding.
+ */
+
+#define SCALEBITS       16      /* speediest right-shift on some machines */
+#define ONE_HALF        ((JLONG)1 << (SCALEBITS - 1))
+#define FIX(x)          ((JLONG)((x) * (1L << SCALEBITS) + 0.5))
+
+/* We allocate one big table for RGB->Y conversion and divide it up into
+ * three parts, instead of doing three alloc_small requests.  This lets us
+ * use a single table base address, which can be held in a register in the
+ * inner loops on many machines (more than can hold all three addresses,
+ * anyway).
+ */
+
+#define R_Y_OFF         0                       /* offset to R => Y section */
+#define G_Y_OFF         (1 * (MAXJSAMPLE + 1))  /* offset to G => Y section */
+#define B_Y_OFF         (2 * (MAXJSAMPLE + 1))  /* etc. */
+#define TABLE_SIZE      (3 * (MAXJSAMPLE + 1))
+
+
+/* Include inline routines for colorspace extensions */
+
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define ycc_rgb_convert_internal  ycc_extrgb_convert_internal
+#define gray_rgb_convert_internal  gray_extrgb_convert_internal
+#define rgb_rgb_convert_internal  rgb_extrgb_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_ALPHA  3
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define ycc_rgb_convert_internal  ycc_extrgbx_convert_internal
+#define gray_rgb_convert_internal  gray_extrgbx_convert_internal
+#define rgb_rgb_convert_internal  rgb_extrgbx_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define ycc_rgb_convert_internal  ycc_extbgr_convert_internal
+#define gray_rgb_convert_internal  gray_extbgr_convert_internal
+#define rgb_rgb_convert_internal  rgb_extbgr_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_ALPHA  3
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define ycc_rgb_convert_internal  ycc_extbgrx_convert_internal
+#define gray_rgb_convert_internal  gray_extbgrx_convert_internal
+#define rgb_rgb_convert_internal  rgb_extbgrx_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_ALPHA  0
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define ycc_rgb_convert_internal  ycc_extxbgr_convert_internal
+#define gray_rgb_convert_internal  gray_extxbgr_convert_internal
+#define rgb_rgb_convert_internal  rgb_extxbgr_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_ALPHA  0
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define ycc_rgb_convert_internal  ycc_extxrgb_convert_internal
+#define gray_rgb_convert_internal  gray_extxrgb_convert_internal
+#define rgb_rgb_convert_internal  rgb_extxrgb_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
+
+
+/*
+ * Initialize tables for YCC->RGB colorspace conversion.
+ */
+
+LOCAL(void)
+build_ycc_rgb_table(j_decompress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  int i;
+  JLONG x;
+  SHIFT_TEMPS
+
+  cconvert->Cr_r_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE + 1) * sizeof(int));
+  cconvert->Cb_b_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE + 1) * sizeof(int));
+  cconvert->Cr_g_tab = (JLONG *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE + 1) * sizeof(JLONG));
+  cconvert->Cb_g_tab = (JLONG *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE + 1) * sizeof(JLONG));
+
+  for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
+    /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
+    /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
+    /* Cr=>R value is nearest int to 1.40200 * x */
+    cconvert->Cr_r_tabi = (int)
+                    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+    /* Cb=>B value is nearest int to 1.77200 * x */
+    cconvert->Cb_b_tabi = (int)
+                    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+    /* Cr=>G value is scaled-up -0.71414 * x */
+    cconvert->Cr_g_tabi = (-FIX(0.71414)) * x;
+    /* Cb=>G value is scaled-up -0.34414 * x */
+    /* We also add in ONE_HALF so that need not do it in inner loop */
+    cconvert->Cb_g_tabi = (-FIX(0.34414)) * x + ONE_HALF;
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the output colorspace.
+ */
+
+METHODDEF(void)
+ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    ycc_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                num_rows);
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    ycc_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  case JCS_EXT_BGR:
+    ycc_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                num_rows);
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    ycc_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    ycc_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    ycc_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  default:
+    ycc_rgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                             num_rows);
+    break;
+  }
+}
+
+
+/**************** Cases other than YCbCr -> RGB **************/
+
+
+/*
+ * Initialize for RGB->grayscale colorspace conversion.
+ */
+
+LOCAL(void)
+build_rgb_y_table(j_decompress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  JLONG *rgb_y_tab;
+  JLONG i;
+
+  /* Allocate and fill in the conversion tables. */
+  cconvert->rgb_y_tab = rgb_y_tab = (JLONG *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (TABLE_SIZE * sizeof(JLONG)));
+
+  for (i = 0; i <= MAXJSAMPLE; i++) {
+    rgb_y_tabi + R_Y_OFF = FIX(0.29900) * i;
+    rgb_y_tabi + G_Y_OFF = FIX(0.58700) * i;
+    rgb_y_tabi + B_Y_OFF = FIX(0.11400) * i + ONE_HALF;
+  }
+}
+
+
+/*
+ * Convert RGB to grayscale.
+ */
+
+METHODDEF(void)
+rgb_gray_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                 JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  register int r, g, b;
+  register JLONG *ctab = cconvert->rgb_y_tab;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      r = inptr0col;
+      g = inptr1col;
+      b = inptr2col;
+      /* Y */
+      outptrcol = (JSAMPLE)((ctabr + R_Y_OFF + ctabg + G_Y_OFF +
+                               ctabb + B_Y_OFF) >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
+ * Color conversion for no colorspace change: just copy the data,
+ * converting from separate-planes to interleaved representation.
+ */
+
+METHODDEF(void)
+null_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+             JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, inptr0, inptr1, inptr2, inptr3, outptr;
+  register JDIMENSION col;
+  register int num_components = cinfo->num_components;
+  JDIMENSION num_cols = cinfo->output_width;
+  int ci;
+
+  if (num_components == 3) {
+    while (--num_rows >= 0) {
+      inptr0 = input_buf0input_row;
+      inptr1 = input_buf1input_row;
+      inptr2 = input_buf2input_row;
+      input_row++;
+      outptr = *output_buf++;
+      for (col = 0; col < num_cols; col++) {
+        *outptr++ = inptr0col;
+        *outptr++ = inptr1col;
+        *outptr++ = inptr2col;
+      }
+    }
+  } else if (num_components == 4) {
+    while (--num_rows >= 0) {
+      inptr0 = input_buf0input_row;
+      inptr1 = input_buf1input_row;
+      inptr2 = input_buf2input_row;
+      inptr3 = input_buf3input_row;
+      input_row++;
+      outptr = *output_buf++;
+      for (col = 0; col < num_cols; col++) {
+        *outptr++ = inptr0col;
+        *outptr++ = inptr1col;
+        *outptr++ = inptr2col;
+        *outptr++ = inptr3col;
+      }
+    }
+  } else {
+    while (--num_rows >= 0) {
+      for (ci = 0; ci < num_components; ci++) {
+        inptr = input_bufciinput_row;
+        outptr = *output_buf;
+        for (col = 0; col < num_cols; col++) {
+          outptrci = inptrcol;
+          outptr += num_components;
+        }
+      }
+      output_buf++;
+      input_row++;
+    }
+  }
+}
+
+
+/*
+ * Color conversion for grayscale: just copy the data.
+ * This also works for YCbCr -> grayscale conversion, in which
+ * we just copy the Y (luminance) component and ignore chrominance.
+ */
+
+METHODDEF(void)
+grayscale_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                  JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  jcopy_sample_rows(input_buf0, (int)input_row, output_buf, 0, num_rows,
+                    cinfo->output_width);
+}
+
+
+/*
+ * Convert grayscale to RGB
+ */
+
+METHODDEF(void)
+gray_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                 JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    gray_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    gray_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+    break;
+  case JCS_EXT_BGR:
+    gray_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    gray_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    gray_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    gray_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+    break;
+  default:
+    gray_rgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                              num_rows);
+    break;
+  }
+}
+
+
+/*
+ * Convert plain RGB to extended RGB
+ */
+
+METHODDEF(void)
+rgb_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    rgb_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                num_rows);
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    rgb_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  case JCS_EXT_BGR:
+    rgb_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                num_rows);
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    rgb_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    rgb_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    rgb_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                 num_rows);
+    break;
+  default:
+    rgb_rgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                             num_rows);
+    break;
+  }
+}
+
+
+/*
+ * Adobe-style YCCK->CMYK conversion.
+ * We convert YCbCr to R=1-C, G=1-M, and B=1-Y using the same
+ * conversion as above, while passing K (black) unchanged.
+ * We assume build_ycc_rgb_table has been called.
+ */
+
+METHODDEF(void)
+ycck_cmyk_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                  JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr)cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2, inptr3;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  register int *Crrtab = cconvert->Cr_r_tab;
+  register int *Cbbtab = cconvert->Cb_b_tab;
+  register JLONG *Crgtab = cconvert->Cr_g_tab;
+  register JLONG *Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    inptr3 = input_buf3input_row;
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = inptr0col;
+      cb = inptr1col;
+      cr = inptr2col;
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptr0 = range_limitMAXJSAMPLE - (y + Crrtabcr);   /* red */
+      outptr1 = range_limitMAXJSAMPLE - (y +                 /* green */
+                              ((int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr,
+                                                 SCALEBITS)));
+      outptr2 = range_limitMAXJSAMPLE - (y + Cbbtabcb);   /* blue */
+      /* K passes through unchanged */
+      outptr3 = inptr3col;
+      outptr += 4;
+    }
+  }
+}
+
+
+/*
+ * RGB565 conversion
+ */
+
+#define PACK_SHORT_565_LE(r, g, b) \
+  ((((r) << 8) & 0xF800) | (((g) << 3) & 0x7E0) | ((b) >> 3))
+#define PACK_SHORT_565_BE(r, g, b) \
+  (((r) & 0xF8) | ((g) >> 5) | (((g) << 11) & 0xE000) | (((b) << 5) & 0x1F00))
+
+#define PACK_TWO_PIXELS_LE(l, r)    ((r << 16) | l)
+#define PACK_TWO_PIXELS_BE(l, r)    ((l << 16) | r)
+
+#define PACK_NEED_ALIGNMENT(ptr)    (((size_t)(ptr)) & 3)
+
+#define WRITE_TWO_ALIGNED_PIXELS(addr, pixels)  ((*(int *)(addr)) = pixels)
+
+#define DITHER_565_R(r, dither)  ((r) + ((dither) & 0xFF))
+#define DITHER_565_G(g, dither)  ((g) + (((dither) & 0xFF) >> 1))
+#define DITHER_565_B(b, dither)  ((b) + ((dither) & 0xFF))
+
+
+/* Declarations for ordered dithering
+ *
+ * We use a 4x4 ordered dither array packed into 32 bits.  This array is
+ * sufficient for dithering RGB888 to RGB565.
+ */
+
+#define DITHER_MASK       0x3
+#define DITHER_ROTATE(x)  ((((x) & 0xFF) << 24) | (((x) >> 8) & 0x00FFFFFF))
+static const JLONG dither_matrix4 = {
+  0x0008020A,
+  0x0C040E06,
+  0x030B0109,
+  0x0F070D05
+};
+
+
+static INLINE boolean is_big_endian(void)
+{
+  int test_value = 1;
+  if (*(char *)&test_value != 1)
+    return TRUE;
+  return FALSE;
+}
+
+
+/* Include inline routines for RGB565 conversion */
+
+#define PACK_SHORT_565  PACK_SHORT_565_LE
+#define PACK_TWO_PIXELS  PACK_TWO_PIXELS_LE
+#define ycc_rgb565_convert_internal  ycc_rgb565_convert_le
+#define ycc_rgb565D_convert_internal  ycc_rgb565D_convert_le
+#define rgb_rgb565_convert_internal  rgb_rgb565_convert_le
+#define rgb_rgb565D_convert_internal  rgb_rgb565D_convert_le
+#define gray_rgb565_convert_internal  gray_rgb565_convert_le
+#define gray_rgb565D_convert_internal  gray_rgb565D_convert_le
+#include "jdcol565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef ycc_rgb565_convert_internal
+#undef ycc_rgb565D_convert_internal
+#undef rgb_rgb565_convert_internal
+#undef rgb_rgb565D_convert_internal
+#undef gray_rgb565_convert_internal
+#undef gray_rgb565D_convert_internal
+
+#define PACK_SHORT_565  PACK_SHORT_565_BE
+#define PACK_TWO_PIXELS  PACK_TWO_PIXELS_BE
+#define ycc_rgb565_convert_internal  ycc_rgb565_convert_be
+#define ycc_rgb565D_convert_internal  ycc_rgb565D_convert_be
+#define rgb_rgb565_convert_internal  rgb_rgb565_convert_be
+#define rgb_rgb565D_convert_internal  rgb_rgb565D_convert_be
+#define gray_rgb565_convert_internal  gray_rgb565_convert_be
+#define gray_rgb565D_convert_internal  gray_rgb565D_convert_be
+#include "jdcol565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef ycc_rgb565_convert_internal
+#undef ycc_rgb565D_convert_internal
+#undef rgb_rgb565_convert_internal
+#undef rgb_rgb565D_convert_internal
+#undef gray_rgb565_convert_internal
+#undef gray_rgb565D_convert_internal
+
+
+METHODDEF(void)
+ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                   JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    ycc_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    ycc_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+ycc_rgb565D_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    ycc_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    ycc_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+rgb_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                   JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    rgb_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    rgb_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+rgb_rgb565D_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    rgb_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    rgb_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+gray_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    gray_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    gray_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+gray_rgb565D_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                     JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    gray_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    gray_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+/*
+ * Empty method for start_pass.
+ */
+
+METHODDEF(void)
+start_pass_dcolor(j_decompress_ptr cinfo)
+{
+  /* no work needed */
+}
+
+
+/*
+ * Module initialization routine for output colorspace conversion.
+ */
+
+GLOBAL(void)
+jinit_color_deconverter(j_decompress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert;
+  int ci;
+
+  cconvert = (my_cconvert_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_color_deconverter));
+  cinfo->cconvert = (struct jpeg_color_deconverter *)cconvert;
+  cconvert->pub.start_pass = start_pass_dcolor;
+
+  /* Make sure num_components agrees with jpeg_color_space */
+  switch (cinfo->jpeg_color_space) {
+  case JCS_GRAYSCALE:
+    if (cinfo->num_components != 1)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+
+  case JCS_RGB:
+  case JCS_YCbCr:
+    if (cinfo->num_components != 3)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+
+  case JCS_CMYK:
+  case JCS_YCCK:
+    if (cinfo->num_components != 4)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+
+  default:                      /* JCS_UNKNOWN can be anything */
+    if (cinfo->num_components < 1)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+  }
+
+  /* Set out_color_components and conversion method based on requested space.
+   * Also clear the component_needed flags for any unused components,
+   * so that earlier pipeline stages can avoid useless computation.
+   */
+
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    cinfo->out_color_components = 1;
+    if (cinfo->jpeg_color_space == JCS_GRAYSCALE ||
+        cinfo->jpeg_color_space == JCS_YCbCr) {
+      cconvert->pub.color_convert = grayscale_convert;
+      /* For color->grayscale conversion, only the Y (0) component is needed */
+      for (ci = 1; ci < cinfo->num_components; ci++)
+        cinfo->comp_infoci.component_needed = FALSE;
+    } else if (cinfo->jpeg_color_space == JCS_RGB) {
+      cconvert->pub.color_convert = rgb_gray_convert;
+      build_rgb_y_table(cinfo);
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_RGB:
+  case JCS_EXT_RGB:
+  case JCS_EXT_RGBX:
+  case JCS_EXT_BGR:
+  case JCS_EXT_BGRX:
+  case JCS_EXT_XBGR:
+  case JCS_EXT_XRGB:
+  case JCS_EXT_RGBA:
+  case JCS_EXT_BGRA:
+  case JCS_EXT_ABGR:
+  case JCS_EXT_ARGB:
+    cinfo->out_color_components = rgb_pixelsizecinfo->out_color_space;
+    if (cinfo->jpeg_color_space == JCS_YCbCr) {
+      if (jsimd_can_ycc_rgb())
+        cconvert->pub.color_convert = jsimd_ycc_rgb_convert;
+      else {
+        cconvert->pub.color_convert = ycc_rgb_convert;
+        build_ycc_rgb_table(cinfo);
+      }
+    } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+      cconvert->pub.color_convert = gray_rgb_convert;
+    } else if (cinfo->jpeg_color_space == JCS_RGB) {
+      if (rgb_redcinfo->out_color_space == 0 &&
+          rgb_greencinfo->out_color_space == 1 &&
+          rgb_bluecinfo->out_color_space == 2 &&
+          rgb_pixelsizecinfo->out_color_space == 3)
+        cconvert->pub.color_convert = null_convert;
+      else
+        cconvert->pub.color_convert = rgb_rgb_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_RGB565:
+    cinfo->out_color_components = 3;
+    if (cinfo->dither_mode == JDITHER_NONE) {
+      if (cinfo->jpeg_color_space == JCS_YCbCr) {
+        if (jsimd_can_ycc_rgb565())
+          cconvert->pub.color_convert = jsimd_ycc_rgb565_convert;
+        else {
+          cconvert->pub.color_convert = ycc_rgb565_convert;
+          build_ycc_rgb_table(cinfo);
+        }
+      } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+        cconvert->pub.color_convert = gray_rgb565_convert;
+      } else if (cinfo->jpeg_color_space == JCS_RGB) {
+        cconvert->pub.color_convert = rgb_rgb565_convert;
+      } else
+        ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    } else {
+      /* only ordered dithering is supported */
+      if (cinfo->jpeg_color_space == JCS_YCbCr) {
+        cconvert->pub.color_convert = ycc_rgb565D_convert;
+        build_ycc_rgb_table(cinfo);
+      } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+        cconvert->pub.color_convert = gray_rgb565D_convert;
+      } else if (cinfo->jpeg_color_space == JCS_RGB) {
+        cconvert->pub.color_convert = rgb_rgb565D_convert;
+      } else
+        ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    }
+    break;
+
+  case JCS_CMYK:
+    cinfo->out_color_components = 4;
+    if (cinfo->jpeg_color_space == JCS_YCCK) {
+      cconvert->pub.color_convert = ycck_cmyk_convert;
+      build_ycc_rgb_table(cinfo);
+    } else if (cinfo->jpeg_color_space == JCS_CMYK) {
+      cconvert->pub.color_convert = null_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  default:
+    /* Permit null conversion to same output space */
+    if (cinfo->out_color_space == cinfo->jpeg_color_space) {
+      cinfo->out_color_components = cinfo->num_components;
+      cconvert->pub.color_convert = null_convert;
+    } else                      /* unsupported non-null conversion */
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+  }
+
+  if (cinfo->quantize_colors)
+    cinfo->output_components = 1; /* single colormapped output component */
+  else
+    cinfo->output_components = cinfo->out_color_components;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdct.h Added

@@ -0,0 +1,208 @@
+/*
+ * jdct.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This include file contains common declarations for the forward and
+ * inverse DCT modules.  These declarations are private to the DCT managers
+ * (jcdctmgr.c, jddctmgr.c) and the individual DCT algorithms.
+ * The individual DCT algorithms are kept in separate files to ease
+ * machine-dependent tuning (e.g., assembly coding).
+ */
+
+
+/*
+ * A forward DCT routine is given a pointer to a work area of type DCTELEM;
+ * the DCT is to be performed in-place in that buffer.  Type DCTELEM is int
+ * for 8-bit samples, JLONG for 12-bit samples.  (NOTE: Floating-point DCT
+ * implementations use an array of type FAST_FLOAT, instead.)
+ * The DCT inputs are expected to be signed (range +-CENTERJSAMPLE).
+ * The DCT outputs are returned scaled up by a factor of 8; they therefore
+ * have a range of +-8K for 8-bit data, +-128K for 12-bit data.  This
+ * convention improves accuracy in integer implementations and saves some
+ * work in floating-point ones.
+ * Quantization of the output coefficients is done by jcdctmgr.c. This
+ * step requires an unsigned type and also one with twice the bits.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#ifndef WITH_SIMD
+typedef int DCTELEM;            /* 16 or 32 bits is fine */
+typedef unsigned int UDCTELEM;
+typedef unsigned long long UDCTELEM2;
+#else
+typedef short DCTELEM;          /* prefer 16 bit with SIMD for parellelism */
+typedef unsigned short UDCTELEM;
+typedef unsigned int UDCTELEM2;
+#endif
+#else
+typedef JLONG DCTELEM;          /* must have 32 bits */
+typedef unsigned long long UDCTELEM2;
+#endif
+
+
+/*
+ * An inverse DCT routine is given a pointer to the input JBLOCK and a pointer
+ * to an output sample array.  The routine must dequantize the input data as
+ * well as perform the IDCT; for dequantization, it uses the multiplier table
+ * pointed to by compptr->dct_table.  The output data is to be placed into the
+ * sample array starting at a specified column.  (Any row offset needed will
+ * be applied to the array pointer before it is passed to the IDCT code.)
+ * Note that the number of samples emitted by the IDCT routine is
+ * DCT_scaled_size * DCT_scaled_size.
+ */
+
+/* typedef inverse_DCT_method_ptr is declared in jpegint.h */
+
+/*
+ * Each IDCT routine has its own ideas about the best dct_table element type.
+ */
+
+typedef MULTIPLIER ISLOW_MULT_TYPE;  /* short or int, whichever is faster */
+#if BITS_IN_JSAMPLE == 8
+typedef MULTIPLIER IFAST_MULT_TYPE;  /* 16 bits is OK, use short if faster */
+#define IFAST_SCALE_BITS  2          /* fractional bits in scale factors */
+#else
+typedef JLONG IFAST_MULT_TYPE;       /* need 32 bits for scaled quantizers */
+#define IFAST_SCALE_BITS  13         /* fractional bits in scale factors */
+#endif
+typedef FAST_FLOAT FLOAT_MULT_TYPE;  /* preferred floating type */
+
+
+/*
+ * Each IDCT routine is responsible for range-limiting its results and
+ * converting them to unsigned form (0..MAXJSAMPLE).  The raw outputs could
+ * be quite far out of range if the input data is corrupt, so a bulletproof
+ * range-limiting step is required.  We use a mask-and-table-lookup method
+ * to do the combined operations quickly.  See the comments with
+ * prepare_range_limit_table (in jdmaster.c) for more info.
+ */
+
+#define IDCT_range_limit(cinfo)  ((cinfo)->sample_range_limit + CENTERJSAMPLE)
+
+#define RANGE_MASK  (MAXJSAMPLE * 4 + 3) /* 2 bits wider than legal samples */
+
+
+/* Extern declarations for the forward and inverse DCT routines. */
+
+EXTERN(void) jpeg_fdct_islow(DCTELEM *data);
+EXTERN(void) jpeg_fdct_ifast(DCTELEM *data);
+EXTERN(void) jpeg_fdct_float(FAST_FLOAT *data);
+
+EXTERN(void) jpeg_idct_islow(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_ifast(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_float(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_7x7(j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_6x6(j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_5x5(j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_4x4(j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_3x3(j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_2x2(j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_1x1(j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_9x9(j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_10x10(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_11x11(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_12x12(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_13x13(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_14x14(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_15x15(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_16x16(j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr, JCOEFPTR coef_block,
+                             JSAMPARRAY output_buf, JDIMENSION output_col);
+
+
+/*
+ * Macros for handling fixed-point arithmetic; these are used by many
+ * but not all of the DCT/IDCT modules.
+ *
+ * All values are expected to be of type JLONG.
+ * Fractional constants are scaled left by CONST_BITS bits.
+ * CONST_BITS is defined within each module using these macros,
+ * and may differ from one module to the next.
+ */
+
+#define ONE          ((JLONG)1)
+#define CONST_SCALE  (ONE << CONST_BITS)
+
+/* Convert a positive real constant to an integer scaled by CONST_SCALE.
+ * Caution: some C compilers fail to reduce "FIX(constant)" at compile time,
+ * thus causing a lot of useless floating-point operations at run time.
+ */
+
+#define FIX(x)  ((JLONG)((x) * CONST_SCALE + 0.5))
+
+/* Descale and correctly round a JLONG value that's scaled by N bits.
+ * We assume RIGHT_SHIFT rounds towards minus infinity, so adding
+ * the fudge factor is correct for either sign of X.
+ */
+
+#define DESCALE(x, n)  RIGHT_SHIFT((x) + (ONE << ((n) - 1)), n)
+
+/* Multiply a JLONG variable by a JLONG constant to yield a JLONG result.
+ * This macro is used only when the two inputs will actually be no more than
+ * 16 bits wide, so that a 16x16->32 bit multiply can be used instead of a
+ * full 32x32 multiply.  This provides a useful speedup on many machines.
+ * Unfortunately there is no way to specify a 16x16->32 multiply portably
+ * in C, but some C compilers will do the right thing if you provide the
+ * correct combination of casts.
+ */
+
+#ifdef SHORTxSHORT_32           /* may work if 'int' is 32 bits */
+#define MULTIPLY16C16(var, const)  (((INT16)(var)) * ((INT16)(const)))
+#endif
+#ifdef SHORTxLCONST_32          /* known to work with Microsoft C 6.0 */
+#define MULTIPLY16C16(var, const)  (((INT16)(var)) * ((JLONG)(const)))
+#endif
+
+#ifndef MULTIPLY16C16           /* default definition */
+#define MULTIPLY16C16(var, const)  ((var) * (const))
+#endif
+
+/* Same except both inputs are variables. */
+
+#ifdef SHORTxSHORT_32           /* may work if 'int' is 32 bits */
+#define MULTIPLY16V16(var1, var2)  (((INT16)(var1)) * ((INT16)(var2)))
+#endif
+
+#ifndef MULTIPLY16V16           /* default definition */
+#define MULTIPLY16V16(var1, var2)  ((var1) * (var2))
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jddctmgr.c Added

@@ -0,0 +1,352 @@
+/*
+ * jddctmgr.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2002-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2010, 2015, D. R. Commander.
+ * Copyright (C) 2013, MIPS Technologies, Inc., California.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the inverse-DCT management logic.
+ * This code selects a particular IDCT implementation to be used,
+ * and it performs related housekeeping chores.  No code in this file
+ * is executed per IDCT step, only during output pass setup.
+ *
+ * Note that the IDCT routines are responsible for performing coefficient
+ * dequantization as well as the IDCT proper.  This module sets up the
+ * dequantization multiplier table needed by the IDCT routine.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"               /* Private declarations for DCT subsystem */
+#include "jsimddct.h"
+#include "jpegcomp.h"
+
+
+/*
+ * The decompressor input side (jdinput.c) saves away the appropriate
+ * quantization table for each component at the start of the first scan
+ * involving that component.  (This is necessary in order to correctly
+ * decode files that reuse Q-table slots.)
+ * When we are ready to make an output pass, the saved Q-table is converted
+ * to a multiplier table that will actually be used by the IDCT routine.
+ * The multiplier table contents are IDCT-method-dependent.  To support
+ * application changes in IDCT method between scans, we can remake the
+ * multiplier tables if necessary.
+ * In buffered-image mode, the first output pass may occur before any data
+ * has been seen for some components, and thus before their Q-tables have
+ * been saved away.  To handle this case, multiplier tables are preset
+ * to zeroes; the result of the IDCT will be a neutral gray level.
+ */
+
+
+/* Private subobject for this module */
+
+typedef struct {
+  struct jpeg_inverse_dct pub;  /* public fields */
+
+  /* This array contains the IDCT method code that each multiplier table
+   * is currently set up for, or -1 if it's not yet set up.
+   * The actual multiplier tables are pointed to by dct_table in the
+   * per-component comp_info structures.
+   */
+  int cur_methodMAX_COMPONENTS;
+} my_idct_controller;
+
+typedef my_idct_controller *my_idct_ptr;
+
+
+/* Allocated multiplier tables: big enough for any supported variant */
+
+typedef union {
+  ISLOW_MULT_TYPE islow_arrayDCTSIZE2;
+#ifdef DCT_IFAST_SUPPORTED
+  IFAST_MULT_TYPE ifast_arrayDCTSIZE2;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  FLOAT_MULT_TYPE float_arrayDCTSIZE2;
+#endif
+} multiplier_table;
+
+
+/* The current scaled-IDCT routines require ISLOW-style multiplier tables,
+ * so be sure to compile that code if either ISLOW or SCALING is requested.
+ */
+#ifdef DCT_ISLOW_SUPPORTED
+#define PROVIDE_ISLOW_TABLES
+#else
+#ifdef IDCT_SCALING_SUPPORTED
+#define PROVIDE_ISLOW_TABLES
+#endif
+#endif
+
+
+/*
+ * Prepare for an output pass.
+ * Here we select the proper IDCT routine for each component and build
+ * a matching multiplier table.
+ */
+
+METHODDEF(void)
+start_pass(j_decompress_ptr cinfo)
+{
+  my_idct_ptr idct = (my_idct_ptr)cinfo->idct;
+  int ci, i;
+  jpeg_component_info *compptr;
+  int method = 0;
+  inverse_DCT_method_ptr method_ptr = NULL;
+  JQUANT_TBL *qtbl;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Select the proper IDCT routine for this component's scaling */
+    switch (compptr->_DCT_scaled_size) {
+#ifdef IDCT_SCALING_SUPPORTED
+    case 1:
+      method_ptr = jpeg_idct_1x1;
+      method = JDCT_ISLOW;      /* jidctred uses islow-style table */
+      break;
+    case 2:
+      if (jsimd_can_idct_2x2())
+        method_ptr = jsimd_idct_2x2;
+      else
+        method_ptr = jpeg_idct_2x2;
+      method = JDCT_ISLOW;      /* jidctred uses islow-style table */
+      break;
+    case 3:
+      method_ptr = jpeg_idct_3x3;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 4:
+      if (jsimd_can_idct_4x4())
+        method_ptr = jsimd_idct_4x4;
+      else
+        method_ptr = jpeg_idct_4x4;
+      method = JDCT_ISLOW;      /* jidctred uses islow-style table */
+      break;
+    case 5:
+      method_ptr = jpeg_idct_5x5;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 6:
+#if defined(__mips__)
+      if (jsimd_can_idct_6x6())
+        method_ptr = jsimd_idct_6x6;
+      else
+#endif
+      method_ptr = jpeg_idct_6x6;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 7:
+      method_ptr = jpeg_idct_7x7;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+#endif
+    case DCTSIZE:
+      switch (cinfo->dct_method) {
+#ifdef DCT_ISLOW_SUPPORTED
+      case JDCT_ISLOW:
+        if (jsimd_can_idct_islow())
+          method_ptr = jsimd_idct_islow;
+        else
+          method_ptr = jpeg_idct_islow;
+        method = JDCT_ISLOW;
+        break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+      case JDCT_IFAST:
+        if (jsimd_can_idct_ifast())
+          method_ptr = jsimd_idct_ifast;
+        else
+          method_ptr = jpeg_idct_ifast;
+        method = JDCT_IFAST;
+        break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+      case JDCT_FLOAT:
+        if (jsimd_can_idct_float())
+          method_ptr = jsimd_idct_float;
+        else
+          method_ptr = jpeg_idct_float;
+        method = JDCT_FLOAT;
+        break;
+#endif
+      default:
+        ERREXIT(cinfo, JERR_NOT_COMPILED);
+        break;
+      }
+      break;
+#ifdef IDCT_SCALING_SUPPORTED
+    case 9:
+      method_ptr = jpeg_idct_9x9;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 10:
+      method_ptr = jpeg_idct_10x10;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 11:
+      method_ptr = jpeg_idct_11x11;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 12:
+#if defined(__mips__)
+      if (jsimd_can_idct_12x12())
+        method_ptr = jsimd_idct_12x12;
+      else
+#endif
+      method_ptr = jpeg_idct_12x12;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 13:
+      method_ptr = jpeg_idct_13x13;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 14:
+      method_ptr = jpeg_idct_14x14;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 15:
+      method_ptr = jpeg_idct_15x15;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 16:
+      method_ptr = jpeg_idct_16x16;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+#endif
+    default:
+      ERREXIT1(cinfo, JERR_BAD_DCTSIZE, compptr->_DCT_scaled_size);
+      break;
+    }
+    idct->pub.inverse_DCTci = method_ptr;
+    /* Create multiplier table from quant table.
+     * However, we can skip this if the component is uninteresting
+     * or if we already built the table.  Also, if no quant table
+     * has yet been saved for the component, we leave the
+     * multiplier table all-zero; we'll be reading zeroes from the
+     * coefficient controller's buffer anyway.
+     */
+    if (!compptr->component_needed || idct->cur_methodci == method)
+      continue;
+    qtbl = compptr->quant_table;
+    if (qtbl == NULL)           /* happens if no data yet for component */
+      continue;
+    idct->cur_methodci = method;
+    switch (method) {
+#ifdef PROVIDE_ISLOW_TABLES
+    case JDCT_ISLOW:
+      {
+        /* For LL&M IDCT method, multipliers are equal to raw quantization
+         * coefficients, but are stored as ints to ensure access efficiency.
+         */
+        ISLOW_MULT_TYPE *ismtbl = (ISLOW_MULT_TYPE *)compptr->dct_table;
+        for (i = 0; i < DCTSIZE2; i++) {
+          ismtbli = (ISLOW_MULT_TYPE)qtbl->quantvali;
+        }
+      }
+      break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+    case JDCT_IFAST:
+      {
+        /* For AA&N IDCT method, multipliers are equal to quantization
+         * coefficients scaled by scalefactorrow*scalefactorcol, where
+         *   scalefactor0 = 1
+         *   scalefactork = cos(k*PI/16) * sqrt(2)    for k=1..7
+         * For integer operation, the multiplier table is to be scaled by
+         * IFAST_SCALE_BITS.
+         */
+        IFAST_MULT_TYPE *ifmtbl = (IFAST_MULT_TYPE *)compptr->dct_table;
+#define CONST_BITS  14
+        static const INT16 aanscalesDCTSIZE2 = {
+          /* precomputed values scaled up by 14 bits */
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+          21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+          19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+           8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+           4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+        };
+        SHIFT_TEMPS
+
+        for (i = 0; i < DCTSIZE2; i++) {
+          ifmtbli = (IFAST_MULT_TYPE)
+            DESCALE(MULTIPLY16V16((JLONG)qtbl->quantvali,
+                                  (JLONG)aanscalesi),
+                    CONST_BITS - IFAST_SCALE_BITS);
+        }
+      }
+      break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+    case JDCT_FLOAT:
+      {
+        /* For float AA&N IDCT method, multipliers are equal to quantization
+         * coefficients scaled by scalefactorrow*scalefactorcol, where
+         *   scalefactor0 = 1
+         *   scalefactork = cos(k*PI/16) * sqrt(2)    for k=1..7
+         */
+        FLOAT_MULT_TYPE *fmtbl = (FLOAT_MULT_TYPE *)compptr->dct_table;
+        int row, col;
+        static const double aanscalefactorDCTSIZE = {
+          1.0, 1.387039845, 1.306562965, 1.175875602,
+          1.0, 0.785694958, 0.541196100, 0.275899379
+        };
+
+        i = 0;
+        for (row = 0; row < DCTSIZE; row++) {
+          for (col = 0; col < DCTSIZE; col++) {
+            fmtbli = (FLOAT_MULT_TYPE)
+              ((double)qtbl->quantvali *
+               aanscalefactorrow * aanscalefactorcol);
+            i++;
+          }
+        }
+      }
+      break;
+#endif
+    default:
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+      break;
+    }
+  }
+}
+
+
+/*
+ * Initialize IDCT manager.
+ */
+
+GLOBAL(void)
+jinit_inverse_dct(j_decompress_ptr cinfo)
+{
+  my_idct_ptr idct;
+  int ci;
+  jpeg_component_info *compptr;
+
+  idct = (my_idct_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_idct_controller));
+  cinfo->idct = (struct jpeg_inverse_dct *)idct;
+  idct->pub.start_pass = start_pass;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Allocate and pre-zero a multiplier table for each component */
+    compptr->dct_table =
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  sizeof(multiplier_table));
+    MEMZERO(compptr->dct_table, sizeof(multiplier_table));
+    /* Mark multiplier table not yet set up for any method */
+    idct->cur_methodci = -1;
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdhuff.c Added

@@ -0,0 +1,834 @@
+/*
+ * jdhuff.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2016, 2018-2019, D. R. Commander.
+ * Copyright (C) 2018, Matthias Räncker.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains Huffman entropy decoding routines.
+ *
+ * Much of the complexity here has to do with supporting input suspension.
+ * If the data source module demands suspension, we want to be able to back
+ * up to the start of the current MCU.  To do this, we copy state variables
+ * into local working storage, and update them back to the permanent
+ * storage only upon successful completion of an MCU.
+ *
+ * NOTE: All referenced figures are from
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdhuff.h"             /* Declarations shared with jdphuff.c */
+#include "jpegcomp.h"
+#include "jstdhuff.c"
+
+
+/*
+ * Expanded entropy decoder object for Huffman decoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+typedef struct {
+  int last_dc_valMAX_COMPS_IN_SCAN; /* last DC coef for each component */
+} savable_state;
+
+typedef struct {
+  struct jpeg_entropy_decoder pub; /* public fields */
+
+  /* These fields are loaded into local variables at start of each MCU.
+   * In case of suspension, we exit WITHOUT updating them.
+   */
+  bitread_perm_state bitstate;  /* Bit buffer at start of MCU */
+  savable_state saved;          /* Other state at start of MCU */
+
+  /* These fields are NOT loaded into local working state. */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+
+  /* Pointers to derived tables (these workspaces have image lifespan) */
+  d_derived_tbl *dc_derived_tblsNUM_HUFF_TBLS;
+  d_derived_tbl *ac_derived_tblsNUM_HUFF_TBLS;
+
+  /* Precalculated info set up by start_pass for use in decode_mcu: */
+
+  /* Pointers to derived tables to be used for each block within an MCU */
+  d_derived_tbl *dc_cur_tblsD_MAX_BLOCKS_IN_MCU;
+  d_derived_tbl *ac_cur_tblsD_MAX_BLOCKS_IN_MCU;
+  /* Whether we care about the DC and AC coefficient values for each block */
+  boolean dc_neededD_MAX_BLOCKS_IN_MCU;
+  boolean ac_neededD_MAX_BLOCKS_IN_MCU;
+} huff_entropy_decoder;
+
+typedef huff_entropy_decoder *huff_entropy_ptr;
+
+
+/*
+ * Initialize for a Huffman-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass_huff_decoder(j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  int ci, blkn, dctbl, actbl;
+  d_derived_tbl **pdtbl;
+  jpeg_component_info *compptr;
+
+  /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
+   * This ought to be an error condition, but we make it a warning because
+   * there are some baseline files out there with all zeroes in these bytes.
+   */
+  if (cinfo->Ss != 0 || cinfo->Se != DCTSIZE2 - 1 ||
+      cinfo->Ah != 0 || cinfo->Al != 0)
+    WARNMS(cinfo, JWRN_NOT_SEQUENTIAL);
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    dctbl = compptr->dc_tbl_no;
+    actbl = compptr->ac_tbl_no;
+    /* Compute derived values for Huffman tables */
+    /* We may do this more than once for a table, but it's not expensive */
+    pdtbl = (d_derived_tbl **)(entropy->dc_derived_tbls) + dctbl;
+    jpeg_make_d_derived_tbl(cinfo, TRUE, dctbl, pdtbl);
+    pdtbl = (d_derived_tbl **)(entropy->ac_derived_tbls) + actbl;
+    jpeg_make_d_derived_tbl(cinfo, FALSE, actbl, pdtbl);
+    /* Initialize DC predictions to 0 */
+    entropy->saved.last_dc_valci = 0;
+  }
+
+  /* Precalculate decoding info for each block in an MCU of this scan */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    ci = cinfo->MCU_membershipblkn;
+    compptr = cinfo->cur_comp_infoci;
+    /* Precalculate which table to use for each block */
+    entropy->dc_cur_tblsblkn = entropy->dc_derived_tblscompptr->dc_tbl_no;
+    entropy->ac_cur_tblsblkn = entropy->ac_derived_tblscompptr->ac_tbl_no;
+    /* Decide whether we really care about the coefficient values */
+    if (compptr->component_needed) {
+      entropy->dc_neededblkn = TRUE;
+      /* we don't need the ACs if producing a 1/8th-size image */
+      entropy->ac_neededblkn = (compptr->_DCT_scaled_size > 1);
+    } else {
+      entropy->dc_neededblkn = entropy->ac_neededblkn = FALSE;
+    }
+  }
+
+  /* Initialize bitread state variables */
+  entropy->bitstate.bits_left = 0;
+  entropy->bitstate.get_buffer = 0; /* unnecessary, but keeps Purify quiet */
+  entropy->pub.insufficient_data = FALSE;
+
+  /* Initialize restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Compute the derived values for a Huffman table.
+ * This routine also performs some validation checks on the table.
+ *
+ * Note this is also used by jdphuff.c.
+ */
+
+GLOBAL(void)
+jpeg_make_d_derived_tbl(j_decompress_ptr cinfo, boolean isDC, int tblno,
+                        d_derived_tbl **pdtbl)
+{
+  JHUFF_TBL *htbl;
+  d_derived_tbl *dtbl;
+  int p, i, l, si, numsymbols;
+  int lookbits, ctr;
+  char huffsize257;
+  unsigned int huffcode257;
+  unsigned int code;
+
+  /* Note that huffsize and huffcode are filled in code-length order,
+   * paralleling the order of the symbols themselves in htbl->huffval.
+   */
+
+  /* Find the input Huffman table */
+  if (tblno < 0 || tblno >= NUM_HUFF_TBLS)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+  htbl =
+    isDC ? cinfo->dc_huff_tbl_ptrstblno : cinfo->ac_huff_tbl_ptrstblno;
+  if (htbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+
+  /* Allocate a workspace if we haven't already done so. */
+  if (*pdtbl == NULL)
+    *pdtbl = (d_derived_tbl *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  sizeof(d_derived_tbl));
+  dtbl = *pdtbl;
+  dtbl->pub = htbl;             /* fill in back link */
+
+  /* Figure C.1: make table of Huffman code length for each symbol */
+
+  p = 0;
+  for (l = 1; l <= 16; l++) {
+    i = (int)htbl->bitsl;
+    if (i < 0 || p + i > 256)   /* protect against table overrun */
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    while (i--)
+      huffsizep++ = (char)l;
+  }
+  huffsizep = 0;
+  numsymbols = p;
+
+  /* Figure C.2: generate the codes themselves */
+  /* We also validate that the counts represent a legal Huffman code tree. */
+
+  code = 0;
+  si = huffsize0;
+  p = 0;
+  while (huffsizep) {
+    while (((int)huffsizep) == si) {
+      huffcodep++ = code;
+      code++;
+    }
+    /* code is now 1 more than the last code used for codelength si; but
+     * it must still fit in si bits, since no code is allowed to be all ones.
+     */
+    if (((JLONG)code) >= (((JLONG)1) << si))
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    code <<= 1;
+    si++;
+  }
+
+  /* Figure F.15: generate decoding tables for bit-sequential decoding */
+
+  p = 0;
+  for (l = 1; l <= 16; l++) {
+    if (htbl->bitsl) {
+      /* valoffsetl = huffval index of 1st symbol of code length l,
+       * minus the minimum code of length l
+       */
+      dtbl->valoffsetl = (JLONG)p - (JLONG)huffcodep;
+      p += htbl->bitsl;
+      dtbl->maxcodel = huffcodep - 1; /* maximum code of length l */
+    } else {
+      dtbl->maxcodel = -1;    /* -1 if no codes of this length */
+    }
+  }
+  dtbl->valoffset17 = 0;
+  dtbl->maxcode17 = 0xFFFFFL; /* ensures jpeg_huff_decode terminates */
+
+  /* Compute lookahead tables to speed up decoding.
+   * First we set all the table entries to 0, indicating "too long";
+   * then we iterate through the Huffman codes that are short enough and
+   * fill in all the entries that correspond to bit sequences starting
+   * with that code.
+   */
+
+  for (i = 0; i < (1 << HUFF_LOOKAHEAD); i++)
+    dtbl->lookupi = (HUFF_LOOKAHEAD + 1) << HUFF_LOOKAHEAD;
+
+  p = 0;
+  for (l = 1; l <= HUFF_LOOKAHEAD; l++) {
+    for (i = 1; i <= (int)htbl->bitsl; i++, p++) {
+      /* l = current code's length, p = its index in huffcode & huffval. */
+      /* Generate left-justified code followed by all possible bit sequences */
+      lookbits = huffcodep << (HUFF_LOOKAHEAD - l);
+      for (ctr = 1 << (HUFF_LOOKAHEAD - l); ctr > 0; ctr--) {
+        dtbl->lookuplookbits = (l << HUFF_LOOKAHEAD) | htbl->huffvalp;
+        lookbits++;
+      }
+    }
+  }
+
+  /* Validate symbols as being reasonable.
+   * For AC tables, we make no check, but accept all byte values 0..255.
+   * For DC tables, we require the symbols to be in range 0..15.
+   * (Tighter bounds could be applied depending on the data depth and mode,
+   * but this is sufficient to ensure safe decoding.)
+   */
+  if (isDC) {
+    for (i = 0; i < numsymbols; i++) {
+      int sym = htbl->huffvali;
+      if (sym < 0 || sym > 15)
+        ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    }
+  }
+}
+
+
+/*
+ * Out-of-line code for bit fetching (shared with jdphuff.c).
+ * See jdhuff.h for info about usage.
+ * Note: current values of get_buffer and bits_left are passed as parameters,
+ * but are returned in the corresponding fields of the state struct.
+ *
+ * On most machines MIN_GET_BITS should be 25 to allow the full 32-bit width
+ * of get_buffer to be used.  (On machines with wider words, an even larger
+ * buffer could be used.)  However, on some machines 32-bit shifts are
+ * quite slow and take time proportional to the number of places shifted.
+ * (This is true with most PC compilers, for instance.)  In this case it may
+ * be a win to set MIN_GET_BITS to the minimum value of 15.  This reduces the
+ * average shift distance at the cost of more calls to jpeg_fill_bit_buffer.
+ */
+
+#ifdef SLOW_SHIFT_32
+#define MIN_GET_BITS  15        /* minimum allowable value */
+#else
+#define MIN_GET_BITS  (BIT_BUF_SIZE - 7)
+#endif
+
+
+GLOBAL(boolean)
+jpeg_fill_bit_buffer(bitread_working_state *state,
+                     register bit_buf_type get_buffer, register int bits_left,
+                     int nbits)
+/* Load up the bit buffer to a depth of at least nbits */
+{
+  /* Copy heavily used state fields into locals (hopefully registers) */
+  register const JOCTET *next_input_byte = state->next_input_byte;
+  register size_t bytes_in_buffer = state->bytes_in_buffer;
+  j_decompress_ptr cinfo = state->cinfo;
+
+  /* Attempt to load at least MIN_GET_BITS bits into get_buffer. */
+  /* (It is assumed that no request will be for more than that many bits.) */
+  /* We fail to do so only if we hit a marker or are forced to suspend. */
+
+  if (cinfo->unread_marker == 0) {      /* cannot advance past a marker */
+    while (bits_left < MIN_GET_BITS) {
+      register int c;
+
+      /* Attempt to read a byte */
+      if (bytes_in_buffer == 0) {
+        if (!(*cinfo->src->fill_input_buffer) (cinfo))
+          return FALSE;
+        next_input_byte = cinfo->src->next_input_byte;
+        bytes_in_buffer = cinfo->src->bytes_in_buffer;
+      }
+      bytes_in_buffer--;
+      c = *next_input_byte++;
+
+      /* If it's 0xFF, check and discard stuffed zero byte */
+      if (c == 0xFF) {
+        /* Loop here to discard any padding FF's on terminating marker,
+         * so that we can save a valid unread_marker value.  NOTE: we will
+         * accept multiple FF's followed by a 0 as meaning a single FF data
+         * byte.  This data pattern is not valid according to the standard.
+         */
+        do {
+          if (bytes_in_buffer == 0) {
+            if (!(*cinfo->src->fill_input_buffer) (cinfo))
+              return FALSE;
+            next_input_byte = cinfo->src->next_input_byte;
+            bytes_in_buffer = cinfo->src->bytes_in_buffer;
+          }
+          bytes_in_buffer--;
+          c = *next_input_byte++;
+        } while (c == 0xFF);
+
+        if (c == 0) {
+          /* Found FF/00, which represents an FF data byte */
+          c = 0xFF;
+        } else {
+          /* Oops, it's actually a marker indicating end of compressed data.
+           * Save the marker code for later use.
+           * Fine point: it might appear that we should save the marker into
+           * bitread working state, not straight into permanent state.  But
+           * once we have hit a marker, we cannot need to suspend within the
+           * current MCU, because we will read no more bytes from the data
+           * source.  So it is OK to update permanent state right away.
+           */
+          cinfo->unread_marker = c;
+          /* See if we need to insert some fake zero bits. */
+          goto no_more_bytes;
+        }
+      }
+
+      /* OK, load c into get_buffer */
+      get_buffer = (get_buffer << 8) | c;
+      bits_left += 8;
+    } /* end while */
+  } else {
+no_more_bytes:
+    /* We get here if we've read the marker that terminates the compressed
+     * data segment.  There should be enough bits in the buffer register
+     * to satisfy the request; if so, no problem.
+     */
+    if (nbits > bits_left) {
+      /* Uh-oh.  Report corrupted data to user and stuff zeroes into
+       * the data stream, so that we can produce some kind of image.
+       * We use a nonvolatile flag to ensure that only one warning message
+       * appears per data segment.
+       */
+      if (!cinfo->entropy->insufficient_data) {
+        WARNMS(cinfo, JWRN_HIT_MARKER);
+        cinfo->entropy->insufficient_data = TRUE;
+      }
+      /* Fill the buffer with zero bits */
+      get_buffer <<= MIN_GET_BITS - bits_left;
+      bits_left = MIN_GET_BITS;
+    }
+  }
+
+  /* Unload the local registers */
+  state->next_input_byte = next_input_byte;
+  state->bytes_in_buffer = bytes_in_buffer;
+  state->get_buffer = get_buffer;
+  state->bits_left = bits_left;
+
+  return TRUE;
+}
+
+
+/* Macro version of the above, which performs much better but does not
+   handle markers.  We have to hand off any blocks with markers to the
+   slower routines. */
+
+#define GET_BYTE { \
+  register int c0, c1; \
+  c0 = *buffer++; \
+  c1 = *buffer; \
+  /* Pre-execute most common case */ \
+  get_buffer = (get_buffer << 8) | c0; \
+  bits_left += 8; \
+  if (c0 == 0xFF) { \
+    /* Pre-execute case of FF/00, which represents an FF data byte */ \
+    buffer++; \
+    if (c1 != 0) { \
+      /* Oops, it's actually a marker indicating end of compressed data. */ \
+      cinfo->unread_marker = c1; \
+      /* Back out pre-execution and fill the buffer with zero bits */ \
+      buffer -= 2; \
+      get_buffer &= ~0xFF; \
+    } \
+  } \
+}
+
+#if SIZEOF_SIZE_T == 8 || defined(_WIN64) || (defined(__x86_64__) && defined(__ILP32__))
+
+/* Pre-fetch 48 bytes, because the holding register is 64-bit */
+#define FILL_BIT_BUFFER_FAST \
+  if (bits_left <= 16) { \
+    GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE \
+  }
+
+#else
+
+/* Pre-fetch 16 bytes, because the holding register is 32-bit */
+#define FILL_BIT_BUFFER_FAST \
+  if (bits_left <= 16) { \
+    GET_BYTE GET_BYTE \
+  }
+
+#endif
+
+
+/*
+ * Out-of-line code for Huffman code decoding.
+ * See jdhuff.h for info about usage.
+ */
+
+GLOBAL(int)
+jpeg_huff_decode(bitread_working_state *state,
+                 register bit_buf_type get_buffer, register int bits_left,
+                 d_derived_tbl *htbl, int min_bits)
+{
+  register int l = min_bits;
+  register JLONG code;
+
+  /* HUFF_DECODE has determined that the code is at least min_bits */
+  /* bits long, so fetch that many bits in one swoop. */
+
+  CHECK_BIT_BUFFER(*state, l, return -1);
+  code = GET_BITS(l);
+
+  /* Collect the rest of the Huffman code one bit at a time. */
+  /* This is per Figure F.16. */
+
+  while (code > htbl->maxcodel) {
+    code <<= 1;
+    CHECK_BIT_BUFFER(*state, 1, return -1);
+    code |= GET_BITS(1);
+    l++;
+  }
+
+  /* Unload the local registers */
+  state->get_buffer = get_buffer;
+  state->bits_left = bits_left;
+
+  /* With garbage input we may reach the sentinel value l = 17. */
+
+  if (l > 16) {
+    WARNMS(state->cinfo, JWRN_HUFF_BAD_CODE);
+    return 0;                   /* fake a zero as the safest result */
+  }
+
+  return htbl->pub->huffval(int)(code + htbl->valoffsetl);
+}
+
+
+/*
+ * Figure F.12: extend sign bit.
+ * On some machines, a shift and add will be faster than a table lookup.
+ */
+
+#define AVOID_TABLES
+#ifdef AVOID_TABLES
+
+#define NEG_1  ((unsigned int)-1)
+#define HUFF_EXTEND(x, s) \
+  ((x) + ((((x) - (1 << ((s) - 1))) >> 31) & (((NEG_1) << (s)) + 1)))
+
+#else
+
+#define HUFF_EXTEND(x, s) \
+  ((x) < extend_tests ? (x) + extend_offsets : (x))
+
+static const int extend_test16 = {   /* entry n is 2**(n-1) */
+  0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080,
+  0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000
+};
+
+static const int extend_offset16 = { /* entry n is (-1 << n) + 1 */
+  0, ((-1) << 1) + 1, ((-1) << 2) + 1, ((-1) << 3) + 1, ((-1) << 4) + 1,
+  ((-1) << 5) + 1, ((-1) << 6) + 1, ((-1) << 7) + 1, ((-1) << 8) + 1,
+  ((-1) << 9) + 1, ((-1) << 10) + 1, ((-1) << 11) + 1, ((-1) << 12) + 1,
+  ((-1) << 13) + 1, ((-1) << 14) + 1, ((-1) << 15) + 1
+};
+
+#endif /* AVOID_TABLES */
+
+
+/*
+ * Check for a restart marker & resynchronize decoder.
+ * Returns FALSE if must suspend.
+ */
+
+LOCAL(boolean)
+process_restart(j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  int ci;
+
+  /* Throw away any unused bits remaining in bit buffer; */
+  /* include any full bytes in next_marker's count of discarded bytes */
+  cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8;
+  entropy->bitstate.bits_left = 0;
+
+  /* Advance past the RSTn marker */
+  if (!(*cinfo->marker->read_restart_marker) (cinfo))
+    return FALSE;
+
+  /* Re-initialize DC predictions to 0 */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+    entropy->saved.last_dc_valci = 0;
+
+  /* Reset restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+
+  /* Reset out-of-data flag, unless read_restart_marker left us smack up
+   * against a marker.  In that case we will end up treating the next data
+   * segment as empty, and we can avoid producing bogus output pixels by
+   * leaving the flag set.
+   */
+  if (cinfo->unread_marker == 0)
+    entropy->pub.insufficient_data = FALSE;
+
+  return TRUE;
+}
+
+
+#if defined(__has_feature)
+#if __has_feature(undefined_behavior_sanitizer)
+__attribute__((no_sanitize("signed-integer-overflow"),
+               no_sanitize("unsigned-integer-overflow")))
+#endif
+#endif
+LOCAL(boolean)
+decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  BITREAD_STATE_VARS;
+  int blkn;
+  savable_state state;
+  /* Outer loop handles each block in the MCU */
+
+  /* Load up working state */
+  BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
+  state = entropy->saved;
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    JBLOCKROW block = MCU_data ? MCU_datablkn : NULL;
+    d_derived_tbl *dctbl = entropy->dc_cur_tblsblkn;
+    d_derived_tbl *actbl = entropy->ac_cur_tblsblkn;
+    register int s, k, r;
+
+    /* Decode a single block's worth of coefficients */
+
+    /* Section F.2.2.1: decode the DC coefficient difference */
+    HUFF_DECODE(s, br_state, dctbl, return FALSE, label1);
+    if (s) {
+      CHECK_BIT_BUFFER(br_state, s, return FALSE);
+      r = GET_BITS(s);
+      s = HUFF_EXTEND(r, s);
+    }
+
+    if (entropy->dc_neededblkn) {
+      /* Convert DC difference to actual value, update last_dc_val */
+      int ci = cinfo->MCU_membershipblkn;
+      /* Certain malformed JPEG images produce repeated DC coefficient
+       * differences of 2047 or -2047, which causes state.last_dc_valci to
+       * grow until it overflows or underflows a 32-bit signed integer.  This
+       * behavior is, to the best of our understanding, innocuous, and it is
+       * unclear how to work around it without potentially affecting
+       * performance.  Thus, we (hopefully temporarily) suppress UBSan integer
+       * overflow errors for this function and decode_mcu_fast().
+       */
+      s += state.last_dc_valci;
+      state.last_dc_valci = s;
+      if (block) {
+        /* Output the DC coefficient (assumes jpeg_natural_order0 = 0) */
+        (*block)0 = (JCOEF)s;
+      }
+    }
+
+    if (entropy->ac_neededblkn && block) {
+
+      /* Section F.2.2.2: decode the AC coefficients */
+      /* Since zeroes are skipped, output area must be cleared beforehand */
+      for (k = 1; k < DCTSIZE2; k++) {
+        HUFF_DECODE(s, br_state, actbl, return FALSE, label2);
+
+        r = s >> 4;
+        s &= 15;
+
+        if (s) {
+          k += r;
+          CHECK_BIT_BUFFER(br_state, s, return FALSE);
+          r = GET_BITS(s);
+          s = HUFF_EXTEND(r, s);
+          /* Output coefficient in natural (dezigzagged) order.
+           * Note: the extra entries in jpeg_natural_order will save us
+           * if k >= DCTSIZE2, which could happen if the data is corrupted.
+           */
+          (*block)jpeg_natural_orderk = (JCOEF)s;
+        } else {
+          if (r != 15)
+            break;
+          k += 15;
+        }
+      }
+
+    } else {
+
+      /* Section F.2.2.2: decode the AC coefficients */
+      /* In this path we just discard the values */
+      for (k = 1; k < DCTSIZE2; k++) {
+        HUFF_DECODE(s, br_state, actbl, return FALSE, label3);
+
+        r = s >> 4;
+        s &= 15;
+
+        if (s) {
+          k += r;
+          CHECK_BIT_BUFFER(br_state, s, return FALSE);
+          DROP_BITS(s);
+        } else {
+          if (r != 15)
+            break;
+          k += 15;
+        }
+      }
+    }
+  }
+
+  /* Completed MCU, so update state */
+  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
+  entropy->saved = state;
+  return TRUE;
+}
+
+
+#if defined(__has_feature)
+#if __has_feature(undefined_behavior_sanitizer)
+__attribute__((no_sanitize("signed-integer-overflow"),
+               no_sanitize("unsigned-integer-overflow")))
+#endif
+#endif
+LOCAL(boolean)
+decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  BITREAD_STATE_VARS;
+  JOCTET *buffer;
+  int blkn;
+  savable_state state;
+  /* Outer loop handles each block in the MCU */
+
+  /* Load up working state */
+  BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
+  buffer = (JOCTET *)br_state.next_input_byte;
+  state = entropy->saved;
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    JBLOCKROW block = MCU_data ? MCU_datablkn : NULL;
+    d_derived_tbl *dctbl = entropy->dc_cur_tblsblkn;
+    d_derived_tbl *actbl = entropy->ac_cur_tblsblkn;
+    register int s, k, r, l;
+
+    HUFF_DECODE_FAST(s, l, dctbl);
+    if (s) {
+      FILL_BIT_BUFFER_FAST
+      r = GET_BITS(s);
+      s = HUFF_EXTEND(r, s);
+    }
+
+    if (entropy->dc_neededblkn) {
+      int ci = cinfo->MCU_membershipblkn;
+      /* Refer to the comment in decode_mcu_slow() regarding the supression of
+       * a UBSan integer overflow error in this line of code.
+       */
+      s += state.last_dc_valci;
+      state.last_dc_valci = s;
+      if (block)
+        (*block)0 = (JCOEF)s;
+    }
+
+    if (entropy->ac_neededblkn && block) {
+
+      for (k = 1; k < DCTSIZE2; k++) {
+        HUFF_DECODE_FAST(s, l, actbl);
+        r = s >> 4;
+        s &= 15;
+
+        if (s) {
+          k += r;
+          FILL_BIT_BUFFER_FAST
+          r = GET_BITS(s);
+          s = HUFF_EXTEND(r, s);
+          (*block)jpeg_natural_orderk = (JCOEF)s;
+        } else {
+          if (r != 15) break;
+          k += 15;
+        }
+      }
+
+    } else {
+
+      for (k = 1; k < DCTSIZE2; k++) {
+        HUFF_DECODE_FAST(s, l, actbl);
+        r = s >> 4;
+        s &= 15;
+
+        if (s) {
+          k += r;
+          FILL_BIT_BUFFER_FAST
+          DROP_BITS(s);
+        } else {
+          if (r != 15) break;
+          k += 15;
+        }
+      }
+    }
+  }
+
+  if (cinfo->unread_marker != 0) {
+    cinfo->unread_marker = 0;
+    return FALSE;
+  }
+
+  br_state.bytes_in_buffer -= (buffer - br_state.next_input_byte);
+  br_state.next_input_byte = buffer;
+  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
+  entropy->saved = state;
+  return TRUE;
+}
+
+
+/*
+ * Decode and return one MCU's worth of Huffman-compressed coefficients.
+ * The coefficients are reordered from zigzag order into natural array order,
+ * but are not dequantized.
+ *
+ * The i'th block of the MCU is stored into the block pointed to by
+ * MCU_datai.  WE ASSUME THIS AREA HAS BEEN ZEROED BY THE CALLER.
+ * (Wholesale zeroing is usually a little faster than retail...)
+ *
+ * Returns FALSE if data source requested suspension.  In that case no
+ * changes have been made to permanent state.  (Exception: some output
+ * coefficients may already have been assigned.  This is harmless for
+ * this module, since we'll just re-assign them on the next call.)
+ */
+
+#define BUFSIZE  (DCTSIZE2 * 8)
+
+METHODDEF(boolean)
+decode_mcu(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
+  int usefast = 1;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (!process_restart(cinfo))
+        return FALSE;
+    usefast = 0;
+  }
+
+  if (cinfo->src->bytes_in_buffer < BUFSIZE * (size_t)cinfo->blocks_in_MCU ||
+      cinfo->unread_marker != 0)
+    usefast = 0;
+
+  /* If we've run out of data, just leave the MCU set to zeroes.
+   * This way, we return uniform gray for the remainder of the segment.
+   */
+  if (!entropy->pub.insufficient_data) {
+
+    if (usefast) {
+      if (!decode_mcu_fast(cinfo, MCU_data)) goto use_slow;
+    } else {
+use_slow:
+      if (!decode_mcu_slow(cinfo, MCU_data)) return FALSE;
+    }
+
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * Module initialization routine for Huffman entropy decoding.
+ */
+
+GLOBAL(void)
+jinit_huff_decoder(j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy;
+  int i;
+
+  /* Motion JPEG frames typically do not include the Huffman tables if they
+     are the default tables.  Thus, if the tables are not set by the time
+     the Huffman decoder is initialized (usually within the body of
+     jpeg_start_decompress()), we set them to default values. */
+  std_huff_tables((j_common_ptr)cinfo);
+
+  entropy = (huff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(huff_entropy_decoder));
+  cinfo->entropy = (struct jpeg_entropy_decoder *)entropy;
+  entropy->pub.start_pass = start_pass_huff_decoder;
+  entropy->pub.decode_mcu = decode_mcu;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->dc_derived_tblsi = entropy->ac_derived_tblsi = NULL;
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdhuff.h Added

@@ -0,0 +1,247 @@
+/*
+ * jdhuff.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010-2011, 2015-2016, 2021, D. R. Commander.
+ * Copyright (C) 2018, Matthias Räncker.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains declarations for Huffman entropy decoding routines
+ * that are shared between the sequential decoder (jdhuff.c) and the
+ * progressive decoder (jdphuff.c).  No other modules need to see these.
+ */
+
+#include "jconfigint.h"
+
+
+/* Derived data constructed for each Huffman table */
+
+#define HUFF_LOOKAHEAD  8       /* # of bits of lookahead */
+
+typedef struct {
+  /* Basic tables: (element 0 of each array is unused) */
+  JLONG maxcode18;            /* largest code of length k (-1 if none) */
+  /* (maxcode17 is a sentinel to ensure jpeg_huff_decode terminates) */
+  JLONG valoffset18;          /* huffval offset for codes of length k */
+  /* valoffsetk = huffval index of 1st symbol of code length k, less
+   * the smallest code of length k; so given a code of length k, the
+   * corresponding symbol is huffvalcode + valoffsetk
+   */
+
+  /* Link to public Huffman table (needed only in jpeg_huff_decode) */
+  JHUFF_TBL *pub;
+
+  /* Lookahead table: indexed by the next HUFF_LOOKAHEAD bits of
+   * the input data stream.  If the next Huffman code is no more
+   * than HUFF_LOOKAHEAD bits long, we can obtain its length and
+   * the corresponding symbol directly from this tables.
+   *
+   * The lower 8 bits of each table entry contain the number of
+   * bits in the corresponding Huffman code, or HUFF_LOOKAHEAD + 1
+   * if too long.  The next 8 bits of each entry contain the
+   * symbol.
+   */
+  int lookup1 << HUFF_LOOKAHEAD;
+} d_derived_tbl;
+
+/* Expand a Huffman table definition into the derived format */
+EXTERN(void) jpeg_make_d_derived_tbl(j_decompress_ptr cinfo, boolean isDC,
+                                     int tblno, d_derived_tbl **pdtbl);
+
+
+/*
+ * Fetching the next N bits from the input stream is a time-critical operation
+ * for the Huffman decoders.  We implement it with a combination of inline
+ * macros and out-of-line subroutines.  Note that N (the number of bits
+ * demanded at one time) never exceeds 15 for JPEG use.
+ *
+ * We read source bytes into get_buffer and dole out bits as needed.
+ * If get_buffer already contains enough bits, they are fetched in-line
+ * by the macros CHECK_BIT_BUFFER and GET_BITS.  When there aren't enough
+ * bits, jpeg_fill_bit_buffer is called; it will attempt to fill get_buffer
+ * as full as possible (not just to the number of bits needed; this
+ * prefetching reduces the overhead cost of calling jpeg_fill_bit_buffer).
+ * Note that jpeg_fill_bit_buffer may return FALSE to indicate suspension.
+ * On TRUE return, jpeg_fill_bit_buffer guarantees that get_buffer contains
+ * at least the requested number of bits --- dummy zeroes are inserted if
+ * necessary.
+ */
+
+#if !defined(_WIN32) && !defined(SIZEOF_SIZE_T)
+#error Cannot determine word size
+#endif
+
+#if SIZEOF_SIZE_T == 8 || defined(_WIN64)
+
+typedef size_t bit_buf_type;            /* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  64                /* size of buffer in bits */
+
+#elif defined(__x86_64__) && defined(__ILP32__)
+
+typedef unsigned long long bit_buf_type; /* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  64                 /* size of buffer in bits */
+
+#else
+
+typedef unsigned long bit_buf_type;     /* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  32                /* size of buffer in bits */
+
+#endif
+
+/* If long is > 32 bits on your machine, and shifting/masking longs is
+ * reasonably fast, making bit_buf_type be long and setting BIT_BUF_SIZE
+ * appropriately should be a win.  Unfortunately we can't define the size
+ * with something like  #define BIT_BUF_SIZE (sizeof(bit_buf_type)*8)
+ * because not all machines measure sizeof in 8-bit bytes.
+ */
+
+typedef struct {                /* Bitreading state saved across MCUs */
+  bit_buf_type get_buffer;      /* current bit-extraction buffer */
+  int bits_left;                /* # of unused bits in it */
+} bitread_perm_state;
+
+typedef struct {                /* Bitreading working state within an MCU */
+  /* Current data source location */
+  /* We need a copy, rather than munging the original, in case of suspension */
+  const JOCTET *next_input_byte; /* => next byte to read from source */
+  size_t bytes_in_buffer;       /* # of bytes remaining in source buffer */
+  /* Bit input buffer --- note these values are kept in register variables,
+   * not in this struct, inside the inner loops.
+   */
+  bit_buf_type get_buffer;      /* current bit-extraction buffer */
+  int bits_left;                /* # of unused bits in it */
+  /* Pointer needed by jpeg_fill_bit_buffer. */
+  j_decompress_ptr cinfo;       /* back link to decompress master record */
+} bitread_working_state;
+
+/* Macros to declare and load/save bitread local variables. */
+#define BITREAD_STATE_VARS \
+  register bit_buf_type get_buffer; \
+  register int bits_left; \
+  bitread_working_state br_state
+
+#define BITREAD_LOAD_STATE(cinfop, permstate) \
+  br_state.cinfo = cinfop; \
+  br_state.next_input_byte = cinfop->src->next_input_byte; \
+  br_state.bytes_in_buffer = cinfop->src->bytes_in_buffer; \
+  get_buffer = permstate.get_buffer; \
+  bits_left = permstate.bits_left;
+
+#define BITREAD_SAVE_STATE(cinfop, permstate) \
+  cinfop->src->next_input_byte = br_state.next_input_byte; \
+  cinfop->src->bytes_in_buffer = br_state.bytes_in_buffer; \
+  permstate.get_buffer = get_buffer; \
+  permstate.bits_left = bits_left
+
+/*
+ * These macros provide the in-line portion of bit fetching.
+ * Use CHECK_BIT_BUFFER to ensure there are N bits in get_buffer
+ * before using GET_BITS, PEEK_BITS, or DROP_BITS.
+ * The variables get_buffer and bits_left are assumed to be locals,
+ * but the state struct might not be (jpeg_huff_decode needs this).
+ *      CHECK_BIT_BUFFER(state, n, action);
+ *              Ensure there are N bits in get_buffer; if suspend, take action.
+ *      val = GET_BITS(n);
+ *              Fetch next N bits.
+ *      val = PEEK_BITS(n);
+ *              Fetch next N bits without removing them from the buffer.
+ *      DROP_BITS(n);
+ *              Discard next N bits.
+ * The value N should be a simple variable, not an expression, because it
+ * is evaluated multiple times.
+ */
+
+#define CHECK_BIT_BUFFER(state, nbits, action) { \
+  if (bits_left < (nbits)) { \
+    if (!jpeg_fill_bit_buffer(&(state), get_buffer, bits_left, nbits)) \
+      { action; } \
+    get_buffer = (state).get_buffer;  bits_left = (state).bits_left; \
+  } \
+}
+
+#define GET_BITS(nbits) \
+  (((int)(get_buffer >> (bits_left -= (nbits)))) & ((1 << (nbits)) - 1))
+
+#define PEEK_BITS(nbits) \
+  (((int)(get_buffer >> (bits_left -  (nbits)))) & ((1 << (nbits)) - 1))
+
+#define DROP_BITS(nbits) \
+  (bits_left -= (nbits))
+
+/* Load up the bit buffer to a depth of at least nbits */
+EXTERN(boolean) jpeg_fill_bit_buffer(bitread_working_state *state,
+                                     register bit_buf_type get_buffer,
+                                     register int bits_left, int nbits);
+
+
+/*
+ * Code for extracting next Huffman-coded symbol from input bit stream.
+ * Again, this is time-critical and we make the main paths be macros.
+ *
+ * We use a lookahead table to process codes of up to HUFF_LOOKAHEAD bits
+ * without looping.  Usually, more than 95% of the Huffman codes will be 8
+ * or fewer bits long.  The few overlength codes are handled with a loop,
+ * which need not be inline code.
+ *
+ * Notes about the HUFF_DECODE macro:
+ * 1. Near the end of the data segment, we may fail to get enough bits
+ *    for a lookahead.  In that case, we do it the hard way.
+ * 2. If the lookahead table contains no entry, the next code must be
+ *    more than HUFF_LOOKAHEAD bits long.
+ * 3. jpeg_huff_decode returns -1 if forced to suspend.
+ */
+
+#define HUFF_DECODE(result, state, htbl, failaction, slowlabel) { \
+  register int nb, look; \
+  if (bits_left < HUFF_LOOKAHEAD) { \
+    if (!jpeg_fill_bit_buffer(&state, get_buffer, bits_left, 0)) \
+      { failaction; } \
+    get_buffer = state.get_buffer;  bits_left = state.bits_left; \
+    if (bits_left < HUFF_LOOKAHEAD) { \
+      nb = 1;  goto slowlabel; \
+    } \
+  } \
+  look = PEEK_BITS(HUFF_LOOKAHEAD); \
+  if ((nb = (htbl->lookuplook >> HUFF_LOOKAHEAD)) <= HUFF_LOOKAHEAD) { \
+    DROP_BITS(nb); \
+    result = htbl->lookuplook & ((1 << HUFF_LOOKAHEAD) - 1); \
+  } else { \
+slowlabel: \
+    if ((result = \
+         jpeg_huff_decode(&state, get_buffer, bits_left, htbl, nb)) < 0) \
+      { failaction; } \
+    get_buffer = state.get_buffer;  bits_left = state.bits_left; \
+  } \
+}
+
+#define HUFF_DECODE_FAST(s, nb, htbl) \
+  FILL_BIT_BUFFER_FAST; \
+  s = PEEK_BITS(HUFF_LOOKAHEAD); \
+  s = htbl->lookups; \
+  nb = s >> HUFF_LOOKAHEAD; \
+  /* Pre-execute the common case of nb <= HUFF_LOOKAHEAD */ \
+  DROP_BITS(nb); \
+  s = s & ((1 << HUFF_LOOKAHEAD) - 1); \
+  if (nb > HUFF_LOOKAHEAD) { \
+    /* Equivalent of jpeg_huff_decode() */ \
+    /* Don't use GET_BITS() here because we don't want to modify bits_left */ \
+    s = (get_buffer >> bits_left) & ((1 << (nb)) - 1); \
+    while (s > htbl->maxcodenb) { \
+      s <<= 1; \
+      s |= GET_BITS(1); \
+      nb++; \
+    } \
+    if (nb > 16) \
+      s = 0; \
+    else \
+      s = htbl->pub->huffval(int)(s + htbl->valoffsetnb) & 0xFF; \
+  }
+
+/* Out-of-line case for Huffman code fetching */
+EXTERN(int) jpeg_huff_decode(bitread_working_state *state,
+                             register bit_buf_type get_buffer,
+                             register int bits_left, d_derived_tbl *htbl,
+                             int min_bits);

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdicc.c Added

@@ -0,0 +1,171 @@
+/*
+ * jdicc.c
+ *
+ * Copyright (C) 1997-1998, Thomas G. Lane, Todd Newman.
+ * Copyright (C) 2017, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file provides code to read International Color Consortium (ICC) device
+ * profiles embedded in JFIF JPEG image files.  The ICC has defined a standard
+ * for including such data in JPEG "APP2" markers.  The code given here does
+ * not know anything about the internal structure of the ICC profile data; it
+ * just knows how to get the profile data from a JPEG file while reading it.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc() */
+extern void *malloc(size_t size);
+#endif
+
+
+#define ICC_MARKER  (JPEG_APP0 + 2)     /* JPEG marker code for ICC */
+#define ICC_OVERHEAD_LEN  14            /* size of non-profile data in APP2 */
+
+
+/*
+ * Handy subroutine to test whether a saved marker is an ICC profile marker.
+ */
+
+LOCAL(boolean)
+marker_is_icc(jpeg_saved_marker_ptr marker)
+{
+  return
+    marker->marker == ICC_MARKER &&
+    marker->data_length >= ICC_OVERHEAD_LEN &&
+    /* verify the identifying string */
+    marker->data0 == 0x49 &&
+    marker->data1 == 0x43 &&
+    marker->data2 == 0x43 &&
+    marker->data3 == 0x5F &&
+    marker->data4 == 0x50 &&
+    marker->data5 == 0x52 &&
+    marker->data6 == 0x4F &&
+    marker->data7 == 0x46 &&
+    marker->data8 == 0x49 &&
+    marker->data9 == 0x4C &&
+    marker->data10 == 0x45 &&
+    marker->data11 == 0x0;
+}
+
+
+/*
+ * See if there was an ICC profile in the JPEG file being read; if so,
+ * reassemble and return the profile data.
+ *
+ * TRUE is returned if an ICC profile was found, FALSE if not.  If TRUE is
+ * returned, *icc_data_ptr is set to point to the returned data, and
+ * *icc_data_len is set to its length.
+ *
+ * IMPORTANT: the data at *icc_data_ptr is allocated with malloc() and must be
+ * freed by the caller with free() when the caller no longer needs it.
+ * (Alternatively, we could write this routine to use the IJG library's memory
+ * allocator, so that the data would be freed implicitly when
+ * jpeg_finish_decompress() is called.  But it seems likely that many
+ * applications will prefer to have the data stick around after decompression
+ * finishes.)
+ */
+
+GLOBAL(boolean)
+jpeg_read_icc_profile(j_decompress_ptr cinfo, JOCTET **icc_data_ptr,
+                      unsigned int *icc_data_len)
+{
+  jpeg_saved_marker_ptr marker;
+  int num_markers = 0;
+  int seq_no;
+  JOCTET *icc_data;
+  unsigned int total_length;
+#define MAX_SEQ_NO  255         /* sufficient since marker numbers are bytes */
+  char marker_presentMAX_SEQ_NO + 1;      /* 1 if marker found */
+  unsigned int data_lengthMAX_SEQ_NO + 1; /* size of profile data in marker */
+  unsigned int data_offsetMAX_SEQ_NO + 1; /* offset for data in marker */
+
+  if (icc_data_ptr == NULL || icc_data_len == NULL)
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+  if (cinfo->global_state < DSTATE_READY)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  *icc_data_ptr = NULL;         /* avoid confusion if FALSE return */
+  *icc_data_len = 0;
+
+  /* This first pass over the saved markers discovers whether there are
+   * any ICC markers and verifies the consistency of the marker numbering.
+   */
+
+  for (seq_no = 1; seq_no <= MAX_SEQ_NO; seq_no++)
+    marker_presentseq_no = 0;
+
+  for (marker = cinfo->marker_list; marker != NULL; marker = marker->next) {
+    if (marker_is_icc(marker)) {
+      if (num_markers == 0)
+        num_markers = marker->data13;
+      else if (num_markers != marker->data13) {
+        WARNMS(cinfo, JWRN_BOGUS_ICC);  /* inconsistent num_markers fields */
+        return FALSE;
+      }
+      seq_no = marker->data12;
+      if (seq_no <= 0 || seq_no > num_markers) {
+        WARNMS(cinfo, JWRN_BOGUS_ICC);  /* bogus sequence number */
+        return FALSE;
+      }
+      if (marker_presentseq_no) {
+        WARNMS(cinfo, JWRN_BOGUS_ICC);  /* duplicate sequence numbers */
+        return FALSE;
+      }
+      marker_presentseq_no = 1;
+      data_lengthseq_no = marker->data_length - ICC_OVERHEAD_LEN;
+    }
+  }
+
+  if (num_markers == 0)
+    return FALSE;
+
+  /* Check for missing markers, count total space needed,
+   * compute offset of each marker's part of the data.
+   */
+
+  total_length = 0;
+  for (seq_no = 1; seq_no <= num_markers; seq_no++) {
+    if (marker_presentseq_no == 0) {
+      WARNMS(cinfo, JWRN_BOGUS_ICC);  /* missing sequence number */
+      return FALSE;
+    }
+    data_offsetseq_no = total_length;
+    total_length += data_lengthseq_no;
+  }
+
+  if (total_length == 0) {
+    WARNMS(cinfo, JWRN_BOGUS_ICC);  /* found only empty markers? */
+    return FALSE;
+  }
+
+  /* Allocate space for assembled data */
+  icc_data = (JOCTET *)malloc(total_length * sizeof(JOCTET));
+  if (icc_data == NULL)
+    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 11);  /* oops, out of memory */
+
+  /* and fill it in */
+  for (marker = cinfo->marker_list; marker != NULL; marker = marker->next) {
+    if (marker_is_icc(marker)) {
+      JOCTET FAR *src_ptr;
+      JOCTET *dst_ptr;
+      unsigned int length;
+      seq_no = marker->data12;
+      dst_ptr = icc_data + data_offsetseq_no;
+      src_ptr = marker->data + ICC_OVERHEAD_LEN;
+      length = data_lengthseq_no;
+      while (length--) {
+        *dst_ptr++ = *src_ptr++;
+      }
+    }
+  }
+
+  *icc_data_ptr = icc_data;
+  *icc_data_len = total_length;
+
+  return TRUE;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdinput.c Added

@@ -0,0 +1,408 @@
+/*
+ * jdinput.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2016, 2018, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains input control logic for the JPEG decompressor.
+ * These routines are concerned with controlling the decompressor's input
+ * processing (marker reading and coefficient decoding).  The actual input
+ * reading is done in jdmarker.c, jdhuff.c, and jdphuff.c.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jpegcomp.h"
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_input_controller pub; /* public fields */
+
+  boolean inheaders;            /* TRUE until first SOS is reached */
+} my_input_controller;
+
+typedef my_input_controller *my_inputctl_ptr;
+
+
+/* Forward declarations */
+METHODDEF(int) consume_markers(j_decompress_ptr cinfo);
+
+
+/*
+ * Routines to calculate various quantities related to the size of the image.
+ */
+
+LOCAL(void)
+initial_setup(j_decompress_ptr cinfo)
+/* Called once, when first SOS marker is reached */
+{
+  int ci;
+  jpeg_component_info *compptr;
+
+  /* Make sure image isn't bigger than I can handle */
+  if ((long)cinfo->image_height > (long)JPEG_MAX_DIMENSION ||
+      (long)cinfo->image_width > (long)JPEG_MAX_DIMENSION)
+    ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int)JPEG_MAX_DIMENSION);
+
+  /* For now, precision must match compiled-in value... */
+  if (cinfo->data_precision != BITS_IN_JSAMPLE)
+    ERREXIT1(cinfo, JERR_BAD_PRECISION, cinfo->data_precision);
+
+  /* Check that number of components won't exceed internal array sizes */
+  if (cinfo->num_components > MAX_COMPONENTS)
+    ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+             MAX_COMPONENTS);
+
+  /* Compute maximum sampling factors; check factor validity */
+  cinfo->max_h_samp_factor = 1;
+  cinfo->max_v_samp_factor = 1;
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    if (compptr->h_samp_factor <= 0 ||
+        compptr->h_samp_factor > MAX_SAMP_FACTOR ||
+        compptr->v_samp_factor <= 0 ||
+        compptr->v_samp_factor > MAX_SAMP_FACTOR)
+      ERREXIT(cinfo, JERR_BAD_SAMPLING);
+    cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
+                                   compptr->h_samp_factor);
+    cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
+                                   compptr->v_samp_factor);
+  }
+
+#if JPEG_LIB_VERSION >= 80
+  cinfo->block_size = DCTSIZE;
+  cinfo->natural_order = jpeg_natural_order;
+  cinfo->lim_Se = DCTSIZE2 - 1;
+#endif
+
+  /* We initialize DCT_scaled_size and min_DCT_scaled_size to DCTSIZE.
+   * In the full decompressor, this will be overridden by jdmaster.c;
+   * but in the transcoder, jdmaster.c is not used, so we must do it here.
+   */
+#if JPEG_LIB_VERSION >= 70
+  cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = DCTSIZE;
+#else
+  cinfo->min_DCT_scaled_size = DCTSIZE;
+#endif
+
+  /* Compute dimensions of components */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+#if JPEG_LIB_VERSION >= 70
+    compptr->DCT_h_scaled_size = compptr->DCT_v_scaled_size = DCTSIZE;
+#else
+    compptr->DCT_scaled_size = DCTSIZE;
+#endif
+    /* Size in DCT blocks */
+    compptr->width_in_blocks = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * (long)compptr->h_samp_factor,
+                    (long)(cinfo->max_h_samp_factor * DCTSIZE));
+    compptr->height_in_blocks = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * (long)compptr->v_samp_factor,
+                    (long)(cinfo->max_v_samp_factor * DCTSIZE));
+    /* Set the first and last MCU columns to decompress from multi-scan images.
+     * By default, decompress all of the MCU columns.
+     */
+    cinfo->master->first_MCU_colci = 0;
+    cinfo->master->last_MCU_colci = compptr->width_in_blocks - 1;
+    /* downsampled_width and downsampled_height will also be overridden by
+     * jdmaster.c if we are doing full decompression.  The transcoder library
+     * doesn't use these values, but the calling application might.
+     */
+    /* Size in samples */
+    compptr->downsampled_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * (long)compptr->h_samp_factor,
+                    (long)cinfo->max_h_samp_factor);
+    compptr->downsampled_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * (long)compptr->v_samp_factor,
+                    (long)cinfo->max_v_samp_factor);
+    /* Mark component needed, until color conversion says otherwise */
+    compptr->component_needed = TRUE;
+    /* Mark no quantization table yet saved for component */
+    compptr->quant_table = NULL;
+  }
+
+  /* Compute number of fully interleaved MCU rows. */
+  cinfo->total_iMCU_rows = (JDIMENSION)
+    jdiv_round_up((long)cinfo->image_height,
+                  (long)(cinfo->max_v_samp_factor * DCTSIZE));
+
+  /* Decide whether file contains multiple scans */
+  if (cinfo->comps_in_scan < cinfo->num_components || cinfo->progressive_mode)
+    cinfo->inputctl->has_multiple_scans = TRUE;
+  else
+    cinfo->inputctl->has_multiple_scans = FALSE;
+}
+
+
+LOCAL(void)
+per_scan_setup(j_decompress_ptr cinfo)
+/* Do computations that are needed before processing a JPEG scan */
+/* cinfo->comps_in_scan and cinfo->cur_comp_info were set from SOS marker */
+{
+  int ci, mcublks, tmp;
+  jpeg_component_info *compptr;
+
+  if (cinfo->comps_in_scan == 1) {
+
+    /* Noninterleaved (single-component) scan */
+    compptr = cinfo->cur_comp_info0;
+
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = compptr->width_in_blocks;
+    cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
+
+    /* For noninterleaved scan, always one block per MCU */
+    compptr->MCU_width = 1;
+    compptr->MCU_height = 1;
+    compptr->MCU_blocks = 1;
+    compptr->MCU_sample_width = compptr->_DCT_scaled_size;
+    compptr->last_col_width = 1;
+    /* For noninterleaved scans, it is convenient to define last_row_height
+     * as the number of block rows present in the last iMCU row.
+     */
+    tmp = (int)(compptr->height_in_blocks % compptr->v_samp_factor);
+    if (tmp == 0) tmp = compptr->v_samp_factor;
+    compptr->last_row_height = tmp;
+
+    /* Prepare array describing MCU composition */
+    cinfo->blocks_in_MCU = 1;
+    cinfo->MCU_membership0 = 0;
+
+  } else {
+
+    /* Interleaved (multi-component) scan */
+    if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
+               MAX_COMPS_IN_SCAN);
+
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width,
+                    (long)(cinfo->max_h_samp_factor * DCTSIZE));
+    cinfo->MCU_rows_in_scan = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height,
+                    (long)(cinfo->max_v_samp_factor * DCTSIZE));
+
+    cinfo->blocks_in_MCU = 0;
+
+    for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+      compptr = cinfo->cur_comp_infoci;
+      /* Sampling factors give # of blocks of component in each MCU */
+      compptr->MCU_width = compptr->h_samp_factor;
+      compptr->MCU_height = compptr->v_samp_factor;
+      compptr->MCU_blocks = compptr->MCU_width * compptr->MCU_height;
+      compptr->MCU_sample_width = compptr->MCU_width *
+                                  compptr->_DCT_scaled_size;
+      /* Figure number of non-dummy blocks in last MCU column & row */
+      tmp = (int)(compptr->width_in_blocks % compptr->MCU_width);
+      if (tmp == 0) tmp = compptr->MCU_width;
+      compptr->last_col_width = tmp;
+      tmp = (int)(compptr->height_in_blocks % compptr->MCU_height);
+      if (tmp == 0) tmp = compptr->MCU_height;
+      compptr->last_row_height = tmp;
+      /* Prepare array describing MCU composition */
+      mcublks = compptr->MCU_blocks;
+      if (cinfo->blocks_in_MCU + mcublks > D_MAX_BLOCKS_IN_MCU)
+        ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+      while (mcublks-- > 0) {
+        cinfo->MCU_membershipcinfo->blocks_in_MCU++ = ci;
+      }
+    }
+
+  }
+}
+
+
+/*
+ * Save away a copy of the Q-table referenced by each component present
+ * in the current scan, unless already saved during a prior scan.
+ *
+ * In a multiple-scan JPEG file, the encoder could assign different components
+ * the same Q-table slot number, but change table definitions between scans
+ * so that each component uses a different Q-table.  (The IJG encoder is not
+ * currently capable of doing this, but other encoders might.)  Since we want
+ * to be able to dequantize all the components at the end of the file, this
+ * means that we have to save away the table actually used for each component.
+ * We do this by copying the table at the start of the first scan containing
+ * the component.
+ * Rec. ITU-T T.81 | ISO/IEC 10918-1 prohibits the encoder from changing the
+ * contents of a Q-table slot between scans of a component using that slot.  If
+ * the encoder does so anyway, this decoder will simply use the Q-table values
+ * that were current at the start of the first scan for the component.
+ *
+ * The decompressor output side looks only at the saved quant tables,
+ * not at the current Q-table slots.
+ */
+
+LOCAL(void)
+latch_quant_tables(j_decompress_ptr cinfo)
+{
+  int ci, qtblno;
+  jpeg_component_info *compptr;
+  JQUANT_TBL *qtbl;
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    /* No work if we already saved Q-table for this component */
+    if (compptr->quant_table != NULL)
+      continue;
+    /* Make sure specified quantization table is present */
+    qtblno = compptr->quant_tbl_no;
+    if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
+        cinfo->quant_tbl_ptrsqtblno == NULL)
+      ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
+    /* OK, save away the quantization table */
+    qtbl = (JQUANT_TBL *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  sizeof(JQUANT_TBL));
+    MEMCOPY(qtbl, cinfo->quant_tbl_ptrsqtblno, sizeof(JQUANT_TBL));
+    compptr->quant_table = qtbl;
+  }
+}
+
+
+/*
+ * Initialize the input modules to read a scan of compressed data.
+ * The first call to this is done by jdmaster.c after initializing
+ * the entire decompressor (during jpeg_start_decompress).
+ * Subsequent calls come from consume_markers, below.
+ */
+
+METHODDEF(void)
+start_input_pass(j_decompress_ptr cinfo)
+{
+  per_scan_setup(cinfo);
+  latch_quant_tables(cinfo);
+  (*cinfo->entropy->start_pass) (cinfo);
+  (*cinfo->coef->start_input_pass) (cinfo);
+  cinfo->inputctl->consume_input = cinfo->coef->consume_data;
+}
+
+
+/*
+ * Finish up after inputting a compressed-data scan.
+ * This is called by the coefficient controller after it's read all
+ * the expected data of the scan.
+ */
+
+METHODDEF(void)
+finish_input_pass(j_decompress_ptr cinfo)
+{
+  cinfo->inputctl->consume_input = consume_markers;
+}
+
+
+/*
+ * Read JPEG markers before, between, or after compressed-data scans.
+ * Change state as necessary when a new scan is reached.
+ * Return value is JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
+ *
+ * The consume_input method pointer points either here or to the
+ * coefficient controller's consume_data routine, depending on whether
+ * we are reading a compressed data segment or inter-segment markers.
+ */
+
+METHODDEF(int)
+consume_markers(j_decompress_ptr cinfo)
+{
+  my_inputctl_ptr inputctl = (my_inputctl_ptr)cinfo->inputctl;
+  int val;
+
+  if (inputctl->pub.eoi_reached) /* After hitting EOI, read no further */
+    return JPEG_REACHED_EOI;
+
+  val = (*cinfo->marker->read_markers) (cinfo);
+
+  switch (val) {
+  case JPEG_REACHED_SOS:        /* Found SOS */
+    if (inputctl->inheaders) {  /* 1st SOS */
+      initial_setup(cinfo);
+      inputctl->inheaders = FALSE;
+      /* Note: start_input_pass must be called by jdmaster.c
+       * before any more input can be consumed.  jdapimin.c is
+       * responsible for enforcing this sequencing.
+       */
+    } else {                    /* 2nd or later SOS marker */
+      if (!inputctl->pub.has_multiple_scans)
+        ERREXIT(cinfo, JERR_EOI_EXPECTED); /* Oops, I wasn't expecting this! */
+      start_input_pass(cinfo);
+    }
+    break;
+  case JPEG_REACHED_EOI:        /* Found EOI */
+    inputctl->pub.eoi_reached = TRUE;
+    if (inputctl->inheaders) {  /* Tables-only datastream, apparently */
+      if (cinfo->marker->saw_SOF)
+        ERREXIT(cinfo, JERR_SOF_NO_SOS);
+    } else {
+      /* Prevent infinite loop in coef ctlr's decompress_data routine
+       * if user set output_scan_number larger than number of scans.
+       */
+      if (cinfo->output_scan_number > cinfo->input_scan_number)
+        cinfo->output_scan_number = cinfo->input_scan_number;
+    }
+    break;
+  case JPEG_SUSPENDED:
+    break;
+  }
+
+  return val;
+}
+
+
+/*
+ * Reset state to begin a fresh datastream.
+ */
+
+METHODDEF(void)
+reset_input_controller(j_decompress_ptr cinfo)
+{
+  my_inputctl_ptr inputctl = (my_inputctl_ptr)cinfo->inputctl;
+
+  inputctl->pub.consume_input = consume_markers;
+  inputctl->pub.has_multiple_scans = FALSE; /* "unknown" would be better */
+  inputctl->pub.eoi_reached = FALSE;
+  inputctl->inheaders = TRUE;
+  /* Reset other modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr)cinfo);
+  (*cinfo->marker->reset_marker_reader) (cinfo);
+  /* Reset progression state -- would be cleaner if entropy decoder did this */
+  cinfo->coef_bits = NULL;
+}
+
+
+/*
+ * Initialize the input controller module.
+ * This is called only once, when the decompression object is created.
+ */
+
+GLOBAL(void)
+jinit_input_controller(j_decompress_ptr cinfo)
+{
+  my_inputctl_ptr inputctl;
+
+  /* Create subobject in permanent pool */
+  inputctl = (my_inputctl_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                sizeof(my_input_controller));
+  cinfo->inputctl = (struct jpeg_input_controller *)inputctl;
+  /* Initialize method pointers */
+  inputctl->pub.consume_input = consume_markers;
+  inputctl->pub.reset_input_controller = reset_input_controller;
+  inputctl->pub.start_input_pass = start_input_pass;
+  inputctl->pub.finish_input_pass = finish_input_pass;
+  /* Initialize state: can't use reset_input_controller since we don't
+   * want to try to reset other modules yet.
+   */
+  inputctl->pub.has_multiple_scans = FALSE; /* "unknown" would be better */
+  inputctl->pub.eoi_reached = FALSE;
+  inputctl->inheaders = TRUE;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdmainct.c Added

@@ -0,0 +1,461 @@
+/*
+ * jdmainct.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the main buffer controller for decompression.
+ * The main buffer lies between the JPEG decompressor proper and the
+ * post-processor; it holds downsampled data in the JPEG colorspace.
+ *
+ * Note that this code is bypassed in raw-data mode, since the application
+ * supplies the equivalent of the main buffer in that case.
+ */
+
+#include "jinclude.h"
+#include "jdmainct.h"
+#include "jconfigint.h"
+
+
+/*
+ * In the current system design, the main buffer need never be a full-image
+ * buffer; any full-height buffers will be found inside the coefficient or
+ * postprocessing controllers.  Nonetheless, the main controller is not
+ * trivial.  Its responsibility is to provide context rows for upsampling/
+ * rescaling, and doing this in an efficient fashion is a bit tricky.
+ *
+ * Postprocessor input data is counted in "row groups".  A row group
+ * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size)
+ * sample rows of each component.  (We require DCT_scaled_size values to be
+ * chosen such that these numbers are integers.  In practice DCT_scaled_size
+ * values will likely be powers of two, so we actually have the stronger
+ * condition that DCT_scaled_size / min_DCT_scaled_size is an integer.)
+ * Upsampling will typically produce max_v_samp_factor pixel rows from each
+ * row group (times any additional scale factor that the upsampler is
+ * applying).
+ *
+ * The coefficient controller will deliver data to us one iMCU row at a time;
+ * each iMCU row contains v_samp_factor * DCT_scaled_size sample rows, or
+ * exactly min_DCT_scaled_size row groups.  (This amount of data corresponds
+ * to one row of MCUs when the image is fully interleaved.)  Note that the
+ * number of sample rows varies across components, but the number of row
+ * groups does not.  Some garbage sample rows may be included in the last iMCU
+ * row at the bottom of the image.
+ *
+ * Depending on the vertical scaling algorithm used, the upsampler may need
+ * access to the sample row(s) above and below its current input row group.
+ * The upsampler is required to set need_context_rows TRUE at global selection
+ * time if so.  When need_context_rows is FALSE, this controller can simply
+ * obtain one iMCU row at a time from the coefficient controller and dole it
+ * out as row groups to the postprocessor.
+ *
+ * When need_context_rows is TRUE, this controller guarantees that the buffer
+ * passed to postprocessing contains at least one row group's worth of samples
+ * above and below the row group(s) being processed.  Note that the context
+ * rows "above" the first passed row group appear at negative row offsets in
+ * the passed buffer.  At the top and bottom of the image, the required
+ * context rows are manufactured by duplicating the first or last real sample
+ * row; this avoids having special cases in the upsampling inner loops.
+ *
+ * The amount of context is fixed at one row group just because that's a
+ * convenient number for this controller to work with.  The existing
+ * upsamplers really only need one sample row of context.  An upsampler
+ * supporting arbitrary output rescaling might wish for more than one row
+ * group of context when shrinking the image; tough, we don't handle that.
+ * (This is justified by the assumption that downsizing will be handled mostly
+ * by adjusting the DCT_scaled_size values, so that the actual scale factor at
+ * the upsample step needn't be much less than one.)
+ *
+ * To provide the desired context, we have to retain the last two row groups
+ * of one iMCU row while reading in the next iMCU row.  (The last row group
+ * can't be processed until we have another row group for its below-context,
+ * and so we have to save the next-to-last group too for its above-context.)
+ * We could do this most simply by copying data around in our buffer, but
+ * that'd be very slow.  We can avoid copying any data by creating a rather
+ * strange pointer structure.  Here's how it works.  We allocate a workspace
+ * consisting of M+2 row groups (where M = min_DCT_scaled_size is the number
+ * of row groups per iMCU row).  We create two sets of redundant pointers to
+ * the workspace.  Labeling the physical row groups 0 to M+1, the synthesized
+ * pointer lists look like this:
+ *                   M+1                          M-1
+ * master pointer --> 0         master pointer --> 0
+ *                    1                            1
+ *                   ...                          ...
+ *                   M-3                          M-3
+ *                   M-2                           M
+ *                   M-1                          M+1
+ *                    M                           M-2
+ *                   M+1                          M-1
+ *                    0                            0
+ * We read alternate iMCU rows using each master pointer; thus the last two
+ * row groups of the previous iMCU row remain un-overwritten in the workspace.
+ * The pointer lists are set up so that the required context rows appear to
+ * be adjacent to the proper places when we pass the pointer lists to the
+ * upsampler.
+ *
+ * The above pictures describe the normal state of the pointer lists.
+ * At top and bottom of the image, we diddle the pointer lists to duplicate
+ * the first or last sample row as necessary (this is cheaper than copying
+ * sample rows around).
+ *
+ * This scheme breaks down if M < 2, ie, min_DCT_scaled_size is 1.  In that
+ * situation each iMCU row provides only one row group so the buffering logic
+ * must be different (eg, we must read two iMCU rows before we can emit the
+ * first row group).  For now, we simply do not support providing context
+ * rows when min_DCT_scaled_size is 1.  That combination seems unlikely to
+ * be worth providing --- if someone wants a 1/8th-size preview, they probably
+ * want it quick and dirty, so a context-free upsampler is sufficient.
+ */
+
+
+/* Forward declarations */
+METHODDEF(void) process_data_simple_main(j_decompress_ptr cinfo,
+                                         JSAMPARRAY output_buf,
+                                         JDIMENSION *out_row_ctr,
+                                         JDIMENSION out_rows_avail);
+METHODDEF(void) process_data_context_main(j_decompress_ptr cinfo,
+                                          JSAMPARRAY output_buf,
+                                          JDIMENSION *out_row_ctr,
+                                          JDIMENSION out_rows_avail);
+#ifdef QUANT_2PASS_SUPPORTED
+METHODDEF(void) process_data_crank_post(j_decompress_ptr cinfo,
+                                        JSAMPARRAY output_buf,
+                                        JDIMENSION *out_row_ctr,
+                                        JDIMENSION out_rows_avail);
+#endif
+
+
+LOCAL(void)
+alloc_funny_pointers(j_decompress_ptr cinfo)
+/* Allocate space for the funny pointer lists.
+ * This is done only once, not once per pass.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+  int ci, rgroup;
+  int M = cinfo->_min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf;
+
+  /* Get top-level space for component array pointers.
+   * We alloc both arrays with one call to save a few cycles.
+   */
+  main_ptr->xbuffer0 = (JSAMPIMAGE)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                cinfo->num_components * 2 * sizeof(JSAMPARRAY));
+  main_ptr->xbuffer1 = main_ptr->xbuffer0 + cinfo->num_components;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
+      cinfo->_min_DCT_scaled_size; /* height of a row group of component */
+    /* Get space for pointer lists --- M+4 row groups in each list.
+     * We alloc both pointer lists with one call to save a few cycles.
+     */
+    xbuf = (JSAMPARRAY)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  2 * (rgroup * (M + 4)) * sizeof(JSAMPROW));
+    xbuf += rgroup;             /* want one row group at negative offsets */
+    main_ptr->xbuffer0ci = xbuf;
+    xbuf += rgroup * (M + 4);
+    main_ptr->xbuffer1ci = xbuf;
+  }
+}
+
+
+LOCAL(void)
+make_funny_pointers(j_decompress_ptr cinfo)
+/* Create the funny pointer lists discussed in the comments above.
+ * The actual workspace is already allocated (in main_ptr->buffer),
+ * and the space for the pointer lists is allocated too.
+ * This routine just fills in the curiously ordered lists.
+ * This will be repeated at the beginning of each pass.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+  int ci, i, rgroup;
+  int M = cinfo->_min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY buf, xbuf0, xbuf1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
+      cinfo->_min_DCT_scaled_size; /* height of a row group of component */
+    xbuf0 = main_ptr->xbuffer0ci;
+    xbuf1 = main_ptr->xbuffer1ci;
+    /* First copy the workspace pointers as-is */
+    buf = main_ptr->bufferci;
+    for (i = 0; i < rgroup * (M + 2); i++) {
+      xbuf0i = xbuf1i = bufi;
+    }
+    /* In the second list, put the last four row groups in swapped order */
+    for (i = 0; i < rgroup * 2; i++) {
+      xbuf1rgroup * (M - 2) + i = bufrgroup * M + i;
+      xbuf1rgroup * M + i = bufrgroup * (M - 2) + i;
+    }
+    /* The wraparound pointers at top and bottom will be filled later
+     * (see set_wraparound_pointers, below).  Initially we want the "above"
+     * pointers to duplicate the first actual data line.  This only needs
+     * to happen in xbuffer0.
+     */
+    for (i = 0; i < rgroup; i++) {
+      xbuf0i - rgroup = xbuf00;
+    }
+  }
+}
+
+
+LOCAL(void)
+set_bottom_pointers(j_decompress_ptr cinfo)
+/* Change the pointer lists to duplicate the last sample row at the bottom
+ * of the image.  whichptr indicates which xbuffer holds the final iMCU row.
+ * Also sets rowgroups_avail to indicate number of nondummy row groups in row.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+  int ci, i, rgroup, iMCUheight, rows_left;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Count sample rows in one iMCU row and in one row group */
+    iMCUheight = compptr->v_samp_factor * compptr->_DCT_scaled_size;
+    rgroup = iMCUheight / cinfo->_min_DCT_scaled_size;
+    /* Count nondummy sample rows remaining for this component */
+    rows_left = (int)(compptr->downsampled_height % (JDIMENSION)iMCUheight);
+    if (rows_left == 0) rows_left = iMCUheight;
+    /* Count nondummy row groups.  Should get same answer for each component,
+     * so we need only do it once.
+     */
+    if (ci == 0) {
+      main_ptr->rowgroups_avail = (JDIMENSION)((rows_left - 1) / rgroup + 1);
+    }
+    /* Duplicate the last real sample row rgroup*2 times; this pads out the
+     * last partial rowgroup and ensures at least one full rowgroup of context.
+     */
+    xbuf = main_ptr->xbuffermain_ptr->whichptrci;
+    for (i = 0; i < rgroup * 2; i++) {
+      xbufrows_left + i = xbufrows_left - 1;
+    }
+  }
+}
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_main(j_decompress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+    if (cinfo->upsample->need_context_rows) {
+      main_ptr->pub.process_data = process_data_context_main;
+      make_funny_pointers(cinfo); /* Create the xbuffer lists */
+      main_ptr->whichptr = 0;   /* Read first iMCU row into xbuffer0 */
+      main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
+      main_ptr->iMCU_row_ctr = 0;
+    } else {
+      /* Simple case with no context needed */
+      main_ptr->pub.process_data = process_data_simple_main;
+    }
+    main_ptr->buffer_full = FALSE;      /* Mark buffer empty */
+    main_ptr->rowgroup_ctr = 0;
+    break;
+#ifdef QUANT_2PASS_SUPPORTED
+  case JBUF_CRANK_DEST:
+    /* For last pass of 2-pass quantization, just crank the postprocessor */
+    main_ptr->pub.process_data = process_data_crank_post;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+}
+
+
+/*
+ * Process some data.
+ * This handles the simple case where no context is required.
+ */
+
+METHODDEF(void)
+process_data_simple_main(j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+                         JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+  JDIMENSION rowgroups_avail;
+
+  /* Read input data if we haven't filled the main buffer yet */
+  if (!main_ptr->buffer_full) {
+    if (!(*cinfo->coef->decompress_data) (cinfo, main_ptr->buffer))
+      return;                   /* suspension forced, can do nothing more */
+    main_ptr->buffer_full = TRUE;       /* OK, we have an iMCU row to work with */
+  }
+
+  /* There are always min_DCT_scaled_size row groups in an iMCU row. */
+  rowgroups_avail = (JDIMENSION)cinfo->_min_DCT_scaled_size;
+  /* Note: at the bottom of the image, we may pass extra garbage row groups
+   * to the postprocessor.  The postprocessor has to check for bottom
+   * of image anyway (at row resolution), so no point in us doing it too.
+   */
+
+  /* Feed the postprocessor */
+  (*cinfo->post->post_process_data) (cinfo, main_ptr->buffer,
+                                     &main_ptr->rowgroup_ctr, rowgroups_avail,
+                                     output_buf, out_row_ctr, out_rows_avail);
+
+  /* Has postprocessor consumed all the data yet? If so, mark buffer empty */
+  if (main_ptr->rowgroup_ctr >= rowgroups_avail) {
+    main_ptr->buffer_full = FALSE;
+    main_ptr->rowgroup_ctr = 0;
+  }
+}
+
+
+/*
+ * Process some data.
+ * This handles the case where context rows must be provided.
+ */
+
+METHODDEF(void)
+process_data_context_main(j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+                          JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+
+  /* Read input data if we haven't filled the main buffer yet */
+  if (!main_ptr->buffer_full) {
+    if (!(*cinfo->coef->decompress_data) (cinfo,
+                                          main_ptr->xbuffermain_ptr->whichptr))
+      return;                   /* suspension forced, can do nothing more */
+    main_ptr->buffer_full = TRUE;       /* OK, we have an iMCU row to work with */
+    main_ptr->iMCU_row_ctr++;   /* count rows received */
+  }
+
+  /* Postprocessor typically will not swallow all the input data it is handed
+   * in one call (due to filling the output buffer first).  Must be prepared
+   * to exit and restart.  This switch lets us keep track of how far we got.
+   * Note that each case falls through to the next on successful completion.
+   */
+  switch (main_ptr->context_state) {
+  case CTX_POSTPONED_ROW:
+    /* Call postprocessor using previously set pointers for postponed row */
+    (*cinfo->post->post_process_data) (cinfo,
+                                       main_ptr->xbuffermain_ptr->whichptr,
+                                       &main_ptr->rowgroup_ctr,
+                                       main_ptr->rowgroups_avail, output_buf,
+                                       out_row_ctr, out_rows_avail);
+    if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail)
+      return;                   /* Need to suspend */
+    main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
+    if (*out_row_ctr >= out_rows_avail)
+      return;                   /* Postprocessor exactly filled output buf */
+    FALLTHROUGH                 /*FALLTHROUGH*/
+  case CTX_PREPARE_FOR_IMCU:
+    /* Prepare to process first M-1 row groups of this iMCU row */
+    main_ptr->rowgroup_ctr = 0;
+    main_ptr->rowgroups_avail = (JDIMENSION)(cinfo->_min_DCT_scaled_size - 1);
+    /* Check for bottom of image: if so, tweak pointers to "duplicate"
+     * the last sample row, and adjust rowgroups_avail to ignore padding rows.
+     */
+    if (main_ptr->iMCU_row_ctr == cinfo->total_iMCU_rows)
+      set_bottom_pointers(cinfo);
+    main_ptr->context_state = CTX_PROCESS_IMCU;
+    FALLTHROUGH                 /*FALLTHROUGH*/
+  case CTX_PROCESS_IMCU:
+    /* Call postprocessor using previously set pointers */
+    (*cinfo->post->post_process_data) (cinfo,
+                                       main_ptr->xbuffermain_ptr->whichptr,
+                                       &main_ptr->rowgroup_ctr,
+                                       main_ptr->rowgroups_avail, output_buf,
+                                       out_row_ctr, out_rows_avail);
+    if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail)
+      return;                   /* Need to suspend */
+    /* After the first iMCU, change wraparound pointers to normal state */
+    if (main_ptr->iMCU_row_ctr == 1)
+      set_wraparound_pointers(cinfo);
+    /* Prepare to load new iMCU row using other xbuffer list */
+    main_ptr->whichptr ^= 1;    /* 0=>1 or 1=>0 */
+    main_ptr->buffer_full = FALSE;
+    /* Still need to process last row group of this iMCU row, */
+    /* which is saved at index M+1 of the other xbuffer */
+    main_ptr->rowgroup_ctr = (JDIMENSION)(cinfo->_min_DCT_scaled_size + 1);
+    main_ptr->rowgroups_avail = (JDIMENSION)(cinfo->_min_DCT_scaled_size + 2);
+    main_ptr->context_state = CTX_POSTPONED_ROW;
+  }
+}
+
+
+/*
+ * Process some data.
+ * Final pass of two-pass quantization: just call the postprocessor.
+ * Source data will be the postprocessor controller's internal buffer.
+ */
+
+#ifdef QUANT_2PASS_SUPPORTED
+
+METHODDEF(void)
+process_data_crank_post(j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+                        JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
+{
+  (*cinfo->post->post_process_data) (cinfo, (JSAMPIMAGE)NULL,
+                                     (JDIMENSION *)NULL, (JDIMENSION)0,
+                                     output_buf, out_row_ctr, out_rows_avail);
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
+
+
+/*
+ * Initialize main buffer controller.
+ */
+
+GLOBAL(void)
+jinit_d_main_controller(j_decompress_ptr cinfo, boolean need_full_buffer)
+{
+  my_main_ptr main_ptr;
+  int ci, rgroup, ngroups;
+  jpeg_component_info *compptr;
+
+  main_ptr = (my_main_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_main_controller));
+  cinfo->main = (struct jpeg_d_main_controller *)main_ptr;
+  main_ptr->pub.start_pass = start_pass_main;
+
+  if (need_full_buffer)         /* shouldn't happen */
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  /* Allocate the workspace.
+   * ngroups is the number of row groups we need.
+   */
+  if (cinfo->upsample->need_context_rows) {
+    if (cinfo->_min_DCT_scaled_size < 2) /* unsupported, see comments above */
+      ERREXIT(cinfo, JERR_NOTIMPL);
+    alloc_funny_pointers(cinfo); /* Alloc space for xbuffer lists */
+    ngroups = cinfo->_min_DCT_scaled_size + 2;
+  } else {
+    ngroups = cinfo->_min_DCT_scaled_size;
+  }
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
+      cinfo->_min_DCT_scaled_size; /* height of a row group of component */
+    main_ptr->bufferci = (*cinfo->mem->alloc_sarray)
+                        ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                         compptr->width_in_blocks * compptr->_DCT_scaled_size,
+                         (JDIMENSION)(rgroup * ngroups));
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdmainct.h Added

@@ -0,0 +1,71 @@
+/*
+ * jdmainct.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+#include "jpegcomp.h"
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_main_controller pub; /* public fields */
+
+  /* Pointer to allocated workspace (M or M+2 row groups). */
+  JSAMPARRAY bufferMAX_COMPONENTS;
+
+  boolean buffer_full;          /* Have we gotten an iMCU row from decoder? */
+  JDIMENSION rowgroup_ctr;      /* counts row groups output to postprocessor */
+
+  /* Remaining fields are only used in the context case. */
+
+  /* These are the master pointers to the funny-order pointer lists. */
+  JSAMPIMAGE xbuffer2;        /* pointers to weird pointer lists */
+
+  int whichptr;                 /* indicates which pointer set is now in use */
+  int context_state;            /* process_data state machine status */
+  JDIMENSION rowgroups_avail;   /* row groups available to postprocessor */
+  JDIMENSION iMCU_row_ctr;      /* counts iMCU rows to detect image top/bot */
+} my_main_controller;
+
+typedef my_main_controller *my_main_ptr;
+
+
+/* context_state values: */
+#define CTX_PREPARE_FOR_IMCU    0       /* need to prepare for MCU row */
+#define CTX_PROCESS_IMCU        1       /* feeding iMCU to postprocessor */
+#define CTX_POSTPONED_ROW       2       /* feeding postponed row group */
+
+
+LOCAL(void)
+set_wraparound_pointers(j_decompress_ptr cinfo)
+/* Set up the "wraparound" pointers at top and bottom of the pointer lists.
+ * This changes the pointer list state from top-of-image to the normal state.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
+  int ci, i, rgroup;
+  int M = cinfo->_min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf0, xbuf1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
+      cinfo->_min_DCT_scaled_size; /* height of a row group of component */
+    xbuf0 = main_ptr->xbuffer0ci;
+    xbuf1 = main_ptr->xbuffer1ci;
+    for (i = 0; i < rgroup; i++) {
+      xbuf0i - rgroup = xbuf0rgroup * (M + 1) + i;
+      xbuf1i - rgroup = xbuf1rgroup * (M + 1) + i;
+      xbuf0rgroup * (M + 2) + i = xbuf0i;
+      xbuf1rgroup * (M + 2) + i = xbuf1i;
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdmarker.c Added

@@ -0,0 +1,1374 @@
+/*
+ * jdmarker.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2012, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to decode JPEG datastream markers.
+ * Most of the complexity arises from our desire to support input
+ * suspension: if not all of the data for a marker is available,
+ * we must exit back to the application.  On resumption, we reprocess
+ * the marker.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+typedef enum {                  /* JPEG marker codes */
+  M_SOF0  = 0xc0,
+  M_SOF1  = 0xc1,
+  M_SOF2  = 0xc2,
+  M_SOF3  = 0xc3,
+
+  M_SOF5  = 0xc5,
+  M_SOF6  = 0xc6,
+  M_SOF7  = 0xc7,
+
+  M_JPG   = 0xc8,
+  M_SOF9  = 0xc9,
+  M_SOF10 = 0xca,
+  M_SOF11 = 0xcb,
+
+  M_SOF13 = 0xcd,
+  M_SOF14 = 0xce,
+  M_SOF15 = 0xcf,
+
+  M_DHT   = 0xc4,
+
+  M_DAC   = 0xcc,
+
+  M_RST0  = 0xd0,
+  M_RST1  = 0xd1,
+  M_RST2  = 0xd2,
+  M_RST3  = 0xd3,
+  M_RST4  = 0xd4,
+  M_RST5  = 0xd5,
+  M_RST6  = 0xd6,
+  M_RST7  = 0xd7,
+
+  M_SOI   = 0xd8,
+  M_EOI   = 0xd9,
+  M_SOS   = 0xda,
+  M_DQT   = 0xdb,
+  M_DNL   = 0xdc,
+  M_DRI   = 0xdd,
+  M_DHP   = 0xde,
+  M_EXP   = 0xdf,
+
+  M_APP0  = 0xe0,
+  M_APP1  = 0xe1,
+  M_APP2  = 0xe2,
+  M_APP3  = 0xe3,
+  M_APP4  = 0xe4,
+  M_APP5  = 0xe5,
+  M_APP6  = 0xe6,
+  M_APP7  = 0xe7,
+  M_APP8  = 0xe8,
+  M_APP9  = 0xe9,
+  M_APP10 = 0xea,
+  M_APP11 = 0xeb,
+  M_APP12 = 0xec,
+  M_APP13 = 0xed,
+  M_APP14 = 0xee,
+  M_APP15 = 0xef,
+
+  M_JPG0  = 0xf0,
+  M_JPG13 = 0xfd,
+  M_COM   = 0xfe,
+
+  M_TEM   = 0x01,
+
+  M_ERROR = 0x100
+} JPEG_MARKER;
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_marker_reader pub; /* public fields */
+
+  /* Application-overridable marker processing methods */
+  jpeg_marker_parser_method process_COM;
+  jpeg_marker_parser_method process_APPn16;
+
+  /* Limit on marker data length to save for each marker type */
+  unsigned int length_limit_COM;
+  unsigned int length_limit_APPn16;
+
+  /* Status of COM/APPn marker saving */
+  jpeg_saved_marker_ptr cur_marker;     /* NULL if not processing a marker */
+  unsigned int bytes_read;              /* data bytes read so far in marker */
+  /* Note: cur_marker is not linked into marker_list until it's all read. */
+} my_marker_reader;
+
+typedef my_marker_reader *my_marker_ptr;
+
+
+/*
+ * Macros for fetching data from the data source module.
+ *
+ * At all times, cinfo->src->next_input_byte and ->bytes_in_buffer reflect
+ * the current restart point; we update them only when we have reached a
+ * suitable place to restart if a suspension occurs.
+ */
+
+/* Declare and initialize local copies of input pointer/count */
+#define INPUT_VARS(cinfo) \
+  struct jpeg_source_mgr *datasrc = (cinfo)->src; \
+  const JOCTET *next_input_byte = datasrc->next_input_byte; \
+  size_t bytes_in_buffer = datasrc->bytes_in_buffer
+
+/* Unload the local copies --- do this only at a restart boundary */
+#define INPUT_SYNC(cinfo) \
+  ( datasrc->next_input_byte = next_input_byte, \
+    datasrc->bytes_in_buffer = bytes_in_buffer )
+
+/* Reload the local copies --- used only in MAKE_BYTE_AVAIL */
+#define INPUT_RELOAD(cinfo) \
+  ( next_input_byte = datasrc->next_input_byte, \
+    bytes_in_buffer = datasrc->bytes_in_buffer )
+
+/* Internal macro for INPUT_BYTE and INPUT_2BYTES: make a byte available.
+ * Note we do *not* do INPUT_SYNC before calling fill_input_buffer,
+ * but we must reload the local copies after a successful fill.
+ */
+#define MAKE_BYTE_AVAIL(cinfo, action) \
+  if (bytes_in_buffer == 0) { \
+    if (!(*datasrc->fill_input_buffer) (cinfo)) \
+      { action; } \
+    INPUT_RELOAD(cinfo); \
+  }
+
+/* Read a byte into variable V.
+ * If must suspend, take the specified action (typically "return FALSE").
+ */
+#define INPUT_BYTE(cinfo, V, action) \
+  MAKESTMT( MAKE_BYTE_AVAIL(cinfo, action); \
+            bytes_in_buffer--; \
+            V = *next_input_byte++; )
+
+/* As above, but read two bytes interpreted as an unsigned 16-bit integer.
+ * V should be declared unsigned int or perhaps JLONG.
+ */
+#define INPUT_2BYTES(cinfo, V, action) \
+  MAKESTMT( MAKE_BYTE_AVAIL(cinfo, action); \
+            bytes_in_buffer--; \
+            V = ((unsigned int)(*next_input_byte++)) << 8; \
+            MAKE_BYTE_AVAIL(cinfo, action); \
+            bytes_in_buffer--; \
+            V += *next_input_byte++; )
+
+
+/*
+ * Routines to process JPEG markers.
+ *
+ * Entry condition: JPEG marker itself has been read and its code saved
+ *   in cinfo->unread_marker; input restart point is just after the marker.
+ *
+ * Exit: if return TRUE, have read and processed any parameters, and have
+ *   updated the restart point to point after the parameters.
+ *   If return FALSE, was forced to suspend before reaching end of
+ *   marker parameters; restart point has not been moved.  Same routine
+ *   will be called again after application supplies more input data.
+ *
+ * This approach to suspension assumes that all of a marker's parameters
+ * can fit into a single input bufferload.  This should hold for "normal"
+ * markers.  Some COM/APPn markers might have large parameter segments
+ * that might not fit.  If we are simply dropping such a marker, we use
+ * skip_input_data to get past it, and thereby put the problem on the
+ * source manager's shoulders.  If we are saving the marker's contents
+ * into memory, we use a slightly different convention: when forced to
+ * suspend, the marker processor updates the restart point to the end of
+ * what it's consumed (ie, the end of the buffer) before returning FALSE.
+ * On resumption, cinfo->unread_marker still contains the marker code,
+ * but the data source will point to the next chunk of marker data.
+ * The marker processor must retain internal state to deal with this.
+ *
+ * Note that we don't bother to avoid duplicate trace messages if a
+ * suspension occurs within marker parameters.  Other side effects
+ * require more care.
+ */
+
+
+LOCAL(boolean)
+get_soi(j_decompress_ptr cinfo)
+/* Process an SOI marker */
+{
+  int i;
+
+  TRACEMS(cinfo, 1, JTRC_SOI);
+
+  if (cinfo->marker->saw_SOI)
+    ERREXIT(cinfo, JERR_SOI_DUPLICATE);
+
+  /* Reset all parameters that are defined to be reset by SOI */
+
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    cinfo->arith_dc_Li = 0;
+    cinfo->arith_dc_Ui = 1;
+    cinfo->arith_ac_Ki = 5;
+  }
+  cinfo->restart_interval = 0;
+
+  /* Set initial assumptions for colorspace etc */
+
+  cinfo->jpeg_color_space = JCS_UNKNOWN;
+  cinfo->CCIR601_sampling = FALSE; /* Assume non-CCIR sampling??? */
+
+  cinfo->saw_JFIF_marker = FALSE;
+  cinfo->JFIF_major_version = 1; /* set default JFIF APP0 values */
+  cinfo->JFIF_minor_version = 1;
+  cinfo->density_unit = 0;
+  cinfo->X_density = 1;
+  cinfo->Y_density = 1;
+  cinfo->saw_Adobe_marker = FALSE;
+  cinfo->Adobe_transform = 0;
+
+  cinfo->marker->saw_SOI = TRUE;
+
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_sof(j_decompress_ptr cinfo, boolean is_prog, boolean is_arith)
+/* Process a SOFn marker */
+{
+  JLONG length;
+  int c, ci;
+  jpeg_component_info *compptr;
+  INPUT_VARS(cinfo);
+
+  cinfo->progressive_mode = is_prog;
+  cinfo->arith_code = is_arith;
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+
+  INPUT_BYTE(cinfo, cinfo->data_precision, return FALSE);
+  INPUT_2BYTES(cinfo, cinfo->image_height, return FALSE);
+  INPUT_2BYTES(cinfo, cinfo->image_width, return FALSE);
+  INPUT_BYTE(cinfo, cinfo->num_components, return FALSE);
+
+  length -= 8;
+
+  TRACEMS4(cinfo, 1, JTRC_SOF, cinfo->unread_marker,
+           (int)cinfo->image_width, (int)cinfo->image_height,
+           cinfo->num_components);
+
+  if (cinfo->marker->saw_SOF)
+    ERREXIT(cinfo, JERR_SOF_DUPLICATE);
+
+  /* We don't support files in which the image height is initially specified */
+  /* as 0 and is later redefined by DNL.  As long as we have to check that,  */
+  /* might as well have a general sanity check. */
+  if (cinfo->image_height <= 0 || cinfo->image_width <= 0 ||
+      cinfo->num_components <= 0)
+    ERREXIT(cinfo, JERR_EMPTY_IMAGE);
+
+  if (length != (cinfo->num_components * 3))
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  if (cinfo->comp_info == NULL) /* do only once, even if suspend */
+    cinfo->comp_info = (jpeg_component_info *)(*cinfo->mem->alloc_small)
+                        ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                         cinfo->num_components * sizeof(jpeg_component_info));
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    compptr->component_index = ci;
+    INPUT_BYTE(cinfo, compptr->component_id, return FALSE);
+    INPUT_BYTE(cinfo, c, return FALSE);
+    compptr->h_samp_factor = (c >> 4) & 15;
+    compptr->v_samp_factor = (c     ) & 15;
+    INPUT_BYTE(cinfo, compptr->quant_tbl_no, return FALSE);
+
+    TRACEMS4(cinfo, 1, JTRC_SOF_COMPONENT,
+             compptr->component_id, compptr->h_samp_factor,
+             compptr->v_samp_factor, compptr->quant_tbl_no);
+  }
+
+  cinfo->marker->saw_SOF = TRUE;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_sos(j_decompress_ptr cinfo)
+/* Process a SOS marker */
+{
+  JLONG length;
+  int i, ci, n, c, cc, pi;
+  jpeg_component_info *compptr;
+  INPUT_VARS(cinfo);
+
+  if (!cinfo->marker->saw_SOF)
+    ERREXIT(cinfo, JERR_SOS_NO_SOF);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+
+  INPUT_BYTE(cinfo, n, return FALSE); /* Number of components */
+
+  TRACEMS1(cinfo, 1, JTRC_SOS, n);
+
+  if (length != (n * 2 + 6) || n < 1 || n > MAX_COMPS_IN_SCAN)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  cinfo->comps_in_scan = n;
+
+  /* Collect the component-spec parameters */
+
+  for (i = 0; i < MAX_COMPS_IN_SCAN; i++)
+    cinfo->cur_comp_infoi = NULL;
+
+  for (i = 0; i < n; i++) {
+    INPUT_BYTE(cinfo, cc, return FALSE);
+    INPUT_BYTE(cinfo, c, return FALSE);
+
+    for (ci = 0, compptr = cinfo->comp_info;
+         ci < cinfo->num_components && ci < MAX_COMPS_IN_SCAN;
+         ci++, compptr++) {
+      if (cc == compptr->component_id && !cinfo->cur_comp_infoci)
+        goto id_found;
+    }
+
+    ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc);
+
+id_found:
+
+    cinfo->cur_comp_infoi = compptr;
+    compptr->dc_tbl_no = (c >> 4) & 15;
+    compptr->ac_tbl_no = (c     ) & 15;
+
+    TRACEMS3(cinfo, 1, JTRC_SOS_COMPONENT, cc,
+             compptr->dc_tbl_no, compptr->ac_tbl_no);
+
+    /* This CSi (cc) should differ from the previous CSi */
+    for (pi = 0; pi < i; pi++) {
+      if (cinfo->cur_comp_infopi == compptr) {
+        ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc);
+      }
+    }
+  }
+
+  /* Collect the additional scan parameters Ss, Se, Ah/Al. */
+  INPUT_BYTE(cinfo, c, return FALSE);
+  cinfo->Ss = c;
+  INPUT_BYTE(cinfo, c, return FALSE);
+  cinfo->Se = c;
+  INPUT_BYTE(cinfo, c, return FALSE);
+  cinfo->Ah = (c >> 4) & 15;
+  cinfo->Al = (c     ) & 15;
+
+  TRACEMS4(cinfo, 1, JTRC_SOS_PARAMS, cinfo->Ss, cinfo->Se,
+           cinfo->Ah, cinfo->Al);
+
+  /* Prepare to scan data & restart markers */
+  cinfo->marker->next_restart_num = 0;
+
+  /* Count another SOS marker */
+  cinfo->input_scan_number++;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+#ifdef D_ARITH_CODING_SUPPORTED
+
+LOCAL(boolean)
+get_dac(j_decompress_ptr cinfo)
+/* Process a DAC marker */
+{
+  JLONG length;
+  int index, val;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+
+  while (length > 0) {
+    INPUT_BYTE(cinfo, index, return FALSE);
+    INPUT_BYTE(cinfo, val, return FALSE);
+
+    length -= 2;
+
+    TRACEMS2(cinfo, 1, JTRC_DAC, index, val);
+
+    if (index < 0 || index >= (2 * NUM_ARITH_TBLS))
+      ERREXIT1(cinfo, JERR_DAC_INDEX, index);
+
+    if (index >= NUM_ARITH_TBLS) { /* define AC table */
+      cinfo->arith_ac_Kindex - NUM_ARITH_TBLS = (UINT8)val;
+    } else {                    /* define DC table */
+      cinfo->arith_dc_Lindex = (UINT8)(val & 0x0F);
+      cinfo->arith_dc_Uindex = (UINT8)(val >> 4);
+      if (cinfo->arith_dc_Lindex > cinfo->arith_dc_Uindex)
+        ERREXIT1(cinfo, JERR_DAC_VALUE, val);
+    }
+  }
+
+  if (length != 0)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+#else /* !D_ARITH_CODING_SUPPORTED */
+
+#define get_dac(cinfo)  skip_variable(cinfo)
+
+#endif /* D_ARITH_CODING_SUPPORTED */
+
+
+LOCAL(boolean)
+get_dht(j_decompress_ptr cinfo)
+/* Process a DHT marker */
+{
+  JLONG length;
+  UINT8 bits17;
+  UINT8 huffval256;
+  int i, index, count;
+  JHUFF_TBL **htblptr;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+
+  while (length > 16) {
+    INPUT_BYTE(cinfo, index, return FALSE);
+
+    TRACEMS1(cinfo, 1, JTRC_DHT, index);
+
+    bits0 = 0;
+    count = 0;
+    for (i = 1; i <= 16; i++) {
+      INPUT_BYTE(cinfo, bitsi, return FALSE);
+      count += bitsi;
+    }
+
+    length -= 1 + 16;
+
+    TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
+             bits1, bits2, bits3, bits4,
+             bits5, bits6, bits7, bits8);
+    TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
+             bits9, bits10, bits11, bits12,
+             bits13, bits14, bits15, bits16);
+
+    /* Here we just do minimal validation of the counts to avoid walking
+     * off the end of our table space.  jdhuff.c will check more carefully.
+     */
+    if (count > 256 || ((JLONG)count) > length)
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+
+    for (i = 0; i < count; i++)
+      INPUT_BYTE(cinfo, huffvali, return FALSE);
+
+    MEMZERO(&huffvalcount, (256 - count) * sizeof(UINT8));
+
+    length -= count;
+
+    if (index & 0x10) {         /* AC table definition */
+      index -= 0x10;
+      if (index < 0 || index >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_DHT_INDEX, index);
+      htblptr = &cinfo->ac_huff_tbl_ptrsindex;
+    } else {                    /* DC table definition */
+      if (index < 0 || index >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_DHT_INDEX, index);
+      htblptr = &cinfo->dc_huff_tbl_ptrsindex;
+    }
+
+    if (*htblptr == NULL)
+      *htblptr = jpeg_alloc_huff_table((j_common_ptr)cinfo);
+
+    MEMCOPY((*htblptr)->bits, bits, sizeof((*htblptr)->bits));
+    MEMCOPY((*htblptr)->huffval, huffval, sizeof((*htblptr)->huffval));
+  }
+
+  if (length != 0)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_dqt(j_decompress_ptr cinfo)
+/* Process a DQT marker */
+{
+  JLONG length;
+  int n, i, prec;
+  unsigned int tmp;
+  JQUANT_TBL *quant_ptr;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+
+  while (length > 0) {
+    INPUT_BYTE(cinfo, n, return FALSE);
+    prec = n >> 4;
+    n &= 0x0F;
+
+    TRACEMS2(cinfo, 1, JTRC_DQT, n, prec);
+
+    if (n >= NUM_QUANT_TBLS)
+      ERREXIT1(cinfo, JERR_DQT_INDEX, n);
+
+    if (cinfo->quant_tbl_ptrsn == NULL)
+      cinfo->quant_tbl_ptrsn = jpeg_alloc_quant_table((j_common_ptr)cinfo);
+    quant_ptr = cinfo->quant_tbl_ptrsn;
+
+    for (i = 0; i < DCTSIZE2; i++) {
+      if (prec)
+        INPUT_2BYTES(cinfo, tmp, return FALSE);
+      else
+        INPUT_BYTE(cinfo, tmp, return FALSE);
+      /* We convert the zigzag-order table to natural array order. */
+      quant_ptr->quantvaljpeg_natural_orderi = (UINT16)tmp;
+    }
+
+    if (cinfo->err->trace_level >= 2) {
+      for (i = 0; i < DCTSIZE2; i += 8) {
+        TRACEMS8(cinfo, 2, JTRC_QUANTVALS,
+                 quant_ptr->quantvali,     quant_ptr->quantvali + 1,
+                 quant_ptr->quantvali + 2, quant_ptr->quantvali + 3,
+                 quant_ptr->quantvali + 4, quant_ptr->quantvali + 5,
+                 quant_ptr->quantvali + 6, quant_ptr->quantvali + 7);
+      }
+    }
+
+    length -= DCTSIZE2 + 1;
+    if (prec) length -= DCTSIZE2;
+  }
+
+  if (length != 0)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_dri(j_decompress_ptr cinfo)
+/* Process a DRI marker */
+{
+  JLONG length;
+  unsigned int tmp;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+
+  if (length != 4)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_2BYTES(cinfo, tmp, return FALSE);
+
+  TRACEMS1(cinfo, 1, JTRC_DRI, tmp);
+
+  cinfo->restart_interval = tmp;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+/*
+ * Routines for processing APPn and COM markers.
+ * These are either saved in memory or discarded, per application request.
+ * APP0 and APP14 are specially checked to see if they are
+ * JFIF and Adobe markers, respectively.
+ */
+
+#define APP0_DATA_LEN   14      /* Length of interesting data in APP0 */
+#define APP14_DATA_LEN  12      /* Length of interesting data in APP14 */
+#define APPN_DATA_LEN   14      /* Must be the largest of the above!! */
+
+
+LOCAL(void)
+examine_app0(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
+             JLONG remaining)
+/* Examine first few bytes from an APP0.
+ * Take appropriate action if it is a JFIF marker.
+ * datalen is # of bytes at data, remaining is length of rest of marker data.
+ */
+{
+  JLONG totallen = (JLONG)datalen + remaining;
+
+  if (datalen >= APP0_DATA_LEN &&
+      data0 == 0x4A &&
+      data1 == 0x46 &&
+      data2 == 0x49 &&
+      data3 == 0x46 &&
+      data4 == 0) {
+    /* Found JFIF APP0 marker: save info */
+    cinfo->saw_JFIF_marker = TRUE;
+    cinfo->JFIF_major_version = data5;
+    cinfo->JFIF_minor_version = data6;
+    cinfo->density_unit = data7;
+    cinfo->X_density = (data8 << 8) + data9;
+    cinfo->Y_density = (data10 << 8) + data11;
+    /* Check version.
+     * Major version must be 1, anything else signals an incompatible change.
+     * (We used to treat this as an error, but now it's a nonfatal warning,
+     * because some bozo at Hijaak couldn't read the spec.)
+     * Minor version should be 0..2, but process anyway if newer.
+     */
+    if (cinfo->JFIF_major_version != 1)
+      WARNMS2(cinfo, JWRN_JFIF_MAJOR,
+              cinfo->JFIF_major_version, cinfo->JFIF_minor_version);
+    /* Generate trace messages */
+    TRACEMS5(cinfo, 1, JTRC_JFIF,
+             cinfo->JFIF_major_version, cinfo->JFIF_minor_version,
+             cinfo->X_density, cinfo->Y_density, cinfo->density_unit);
+    /* Validate thumbnail dimensions and issue appropriate messages */
+    if (data12 | data13)
+      TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL, data12, data13);
+    totallen -= APP0_DATA_LEN;
+    if (totallen != ((JLONG)data12 * (JLONG)data13 * (JLONG)3))
+      TRACEMS1(cinfo, 1, JTRC_JFIF_BADTHUMBNAILSIZE, (int)totallen);
+  } else if (datalen >= 6 &&
+             data0 == 0x4A &&
+             data1 == 0x46 &&
+             data2 == 0x58 &&
+             data3 == 0x58 &&
+             data4 == 0) {
+    /* Found JFIF "JFXX" extension APP0 marker */
+    /* The library doesn't actually do anything with these,
+     * but we try to produce a helpful trace message.
+     */
+    switch (data5) {
+    case 0x10:
+      TRACEMS1(cinfo, 1, JTRC_THUMB_JPEG, (int)totallen);
+      break;
+    case 0x11:
+      TRACEMS1(cinfo, 1, JTRC_THUMB_PALETTE, (int)totallen);
+      break;
+    case 0x13:
+      TRACEMS1(cinfo, 1, JTRC_THUMB_RGB, (int)totallen);
+      break;
+    default:
+      TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION, data5, (int)totallen);
+      break;
+    }
+  } else {
+    /* Start of APP0 does not match "JFIF" or "JFXX", or too short */
+    TRACEMS1(cinfo, 1, JTRC_APP0, (int)totallen);
+  }
+}
+
+
+LOCAL(void)
+examine_app14(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
+              JLONG remaining)
+/* Examine first few bytes from an APP14.
+ * Take appropriate action if it is an Adobe marker.
+ * datalen is # of bytes at data, remaining is length of rest of marker data.
+ */
+{
+  unsigned int version, flags0, flags1, transform;
+
+  if (datalen >= APP14_DATA_LEN &&
+      data0 == 0x41 &&
+      data1 == 0x64 &&
+      data2 == 0x6F &&
+      data3 == 0x62 &&
+      data4 == 0x65) {
+    /* Found Adobe APP14 marker */
+    version = (data5 << 8) + data6;
+    flags0 = (data7 << 8) + data8;
+    flags1 = (data9 << 8) + data10;
+    transform = data11;
+    TRACEMS4(cinfo, 1, JTRC_ADOBE, version, flags0, flags1, transform);
+    cinfo->saw_Adobe_marker = TRUE;
+    cinfo->Adobe_transform = (UINT8)transform;
+  } else {
+    /* Start of APP14 does not match "Adobe", or too short */
+    TRACEMS1(cinfo, 1, JTRC_APP14, (int)(datalen + remaining));
+  }
+}
+
+
+METHODDEF(boolean)
+get_interesting_appn(j_decompress_ptr cinfo)
+/* Process an APP0 or APP14 marker without saving it */
+{
+  JLONG length;
+  JOCTET bAPPN_DATA_LEN;
+  unsigned int i, numtoread;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+
+  /* get the interesting part of the marker data */
+  if (length >= APPN_DATA_LEN)
+    numtoread = APPN_DATA_LEN;
+  else if (length > 0)
+    numtoread = (unsigned int)length;
+  else
+    numtoread = 0;
+  for (i = 0; i < numtoread; i++)
+    INPUT_BYTE(cinfo, bi, return FALSE);
+  length -= numtoread;
+
+  /* process it */
+  switch (cinfo->unread_marker) {
+  case M_APP0:
+    examine_app0(cinfo, (JOCTET *)b, numtoread, length);
+    break;
+  case M_APP14:
+    examine_app14(cinfo, (JOCTET *)b, numtoread, length);
+    break;
+  default:
+    /* can't get here unless jpeg_save_markers chooses wrong processor */
+    ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, cinfo->unread_marker);
+    break;
+  }
+
+  /* skip any remaining data -- could be lots */
+  INPUT_SYNC(cinfo);
+  if (length > 0)
+    (*cinfo->src->skip_input_data) (cinfo, (long)length);
+
+  return TRUE;
+}
+
+
+#ifdef SAVE_MARKERS_SUPPORTED
+
+METHODDEF(boolean)
+save_marker(j_decompress_ptr cinfo)
+/* Save an APPn or COM marker into the marker list */
+{
+  my_marker_ptr marker = (my_marker_ptr)cinfo->marker;
+  jpeg_saved_marker_ptr cur_marker = marker->cur_marker;
+  unsigned int bytes_read, data_length;
+  JOCTET *data;
+  JLONG length = 0;
+  INPUT_VARS(cinfo);
+
+  if (cur_marker == NULL) {
+    /* begin reading a marker */
+    INPUT_2BYTES(cinfo, length, return FALSE);
+    length -= 2;
+    if (length >= 0) {          /* watch out for bogus length word */
+      /* figure out how much we want to save */
+      unsigned int limit;
+      if (cinfo->unread_marker == (int)M_COM)
+        limit = marker->length_limit_COM;
+      else
+        limit = marker->length_limit_APPncinfo->unread_marker - (int)M_APP0;
+      if ((unsigned int)length < limit)
+        limit = (unsigned int)length;
+      /* allocate and initialize the marker item */
+      cur_marker = (jpeg_saved_marker_ptr)
+        (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                    sizeof(struct jpeg_marker_struct) + limit);
+      cur_marker->next = NULL;
+      cur_marker->marker = (UINT8)cinfo->unread_marker;
+      cur_marker->original_length = (unsigned int)length;
+      cur_marker->data_length = limit;
+      /* data area is just beyond the jpeg_marker_struct */
+      data = cur_marker->data = (JOCTET *)(cur_marker + 1);
+      marker->cur_marker = cur_marker;
+      marker->bytes_read = 0;
+      bytes_read = 0;
+      data_length = limit;
+    } else {
+      /* deal with bogus length word */
+      bytes_read = data_length = 0;
+      data = NULL;
+    }
+  } else {
+    /* resume reading a marker */
+    bytes_read = marker->bytes_read;
+    data_length = cur_marker->data_length;
+    data = cur_marker->data + bytes_read;
+  }
+
+  while (bytes_read < data_length) {
+    INPUT_SYNC(cinfo);          /* move the restart point to here */
+    marker->bytes_read = bytes_read;
+    /* If there's not at least one byte in buffer, suspend */
+    MAKE_BYTE_AVAIL(cinfo, return FALSE);
+    /* Copy bytes with reasonable rapidity */
+    while (bytes_read < data_length && bytes_in_buffer > 0) {
+      *data++ = *next_input_byte++;
+      bytes_in_buffer--;
+      bytes_read++;
+    }
+  }
+
+  /* Done reading what we want to read */
+  if (cur_marker != NULL) {     /* will be NULL if bogus length word */
+    /* Add new marker to end of list */
+    if (cinfo->marker_list == NULL) {
+      cinfo->marker_list = cur_marker;
+    } else {
+      jpeg_saved_marker_ptr prev = cinfo->marker_list;
+      while (prev->next != NULL)
+        prev = prev->next;
+      prev->next = cur_marker;
+    }
+    /* Reset pointer & calc remaining data length */
+    data = cur_marker->data;
+    length = cur_marker->original_length - data_length;
+  }
+  /* Reset to initial state for next marker */
+  marker->cur_marker = NULL;
+
+  /* Process the marker if interesting; else just make a generic trace msg */
+  switch (cinfo->unread_marker) {
+  case M_APP0:
+    examine_app0(cinfo, data, data_length, length);
+    break;
+  case M_APP14:
+    examine_app14(cinfo, data, data_length, length);
+    break;
+  default:
+    TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker,
+             (int)(data_length + length));
+    break;
+  }
+
+  /* skip any remaining data -- could be lots */
+  INPUT_SYNC(cinfo);            /* do before skip_input_data */
+  if (length > 0)
+    (*cinfo->src->skip_input_data) (cinfo, (long)length);
+
+  return TRUE;
+}
+
+#endif /* SAVE_MARKERS_SUPPORTED */
+
+
+METHODDEF(boolean)
+skip_variable(j_decompress_ptr cinfo)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  JLONG length;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+
+  TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker, (int)length);
+
+  INPUT_SYNC(cinfo);            /* do before skip_input_data */
+  if (length > 0)
+    (*cinfo->src->skip_input_data) (cinfo, (long)length);
+
+  return TRUE;
+}
+
+
+/*
+ * Find the next JPEG marker, save it in cinfo->unread_marker.
+ * Returns FALSE if had to suspend before reaching a marker;
+ * in that case cinfo->unread_marker is unchanged.
+ *
+ * Note that the result might not be a valid marker code,
+ * but it will never be 0 or FF.
+ */
+
+LOCAL(boolean)
+next_marker(j_decompress_ptr cinfo)
+{
+  int c;
+  INPUT_VARS(cinfo);
+
+  for (;;) {
+    INPUT_BYTE(cinfo, c, return FALSE);
+    /* Skip any non-FF bytes.
+     * This may look a bit inefficient, but it will not occur in a valid file.
+     * We sync after each discarded byte so that a suspending data source
+     * can discard the byte from its buffer.
+     */
+    while (c != 0xFF) {
+      cinfo->marker->discarded_bytes++;
+      INPUT_SYNC(cinfo);
+      INPUT_BYTE(cinfo, c, return FALSE);
+    }
+    /* This loop swallows any duplicate FF bytes.  Extra FFs are legal as
+     * pad bytes, so don't count them in discarded_bytes.  We assume there
+     * will not be so many consecutive FF bytes as to overflow a suspending
+     * data source's input buffer.
+     */
+    do {
+      INPUT_BYTE(cinfo, c, return FALSE);
+    } while (c == 0xFF);
+    if (c != 0)
+      break;                    /* found a valid marker, exit loop */
+    /* Reach here if we found a stuffed-zero data sequence (FF/00).
+     * Discard it and loop back to try again.
+     */
+    cinfo->marker->discarded_bytes += 2;
+    INPUT_SYNC(cinfo);
+  }
+
+  if (cinfo->marker->discarded_bytes != 0) {
+    WARNMS2(cinfo, JWRN_EXTRANEOUS_DATA, cinfo->marker->discarded_bytes, c);
+    cinfo->marker->discarded_bytes = 0;
+  }
+
+  cinfo->unread_marker = c;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+first_marker(j_decompress_ptr cinfo)
+/* Like next_marker, but used to obtain the initial SOI marker. */
+/* For this marker, we do not allow preceding garbage or fill; otherwise,
+ * we might well scan an entire input file before realizing it ain't JPEG.
+ * If an application wants to process non-JFIF files, it must seek to the
+ * SOI before calling the JPEG library.
+ */
+{
+  int c, c2;
+  INPUT_VARS(cinfo);
+
+  INPUT_BYTE(cinfo, c, return FALSE);
+  INPUT_BYTE(cinfo, c2, return FALSE);
+  if (c != 0xFF || c2 != (int)M_SOI)
+    ERREXIT2(cinfo, JERR_NO_SOI, c, c2);
+
+  cinfo->unread_marker = c2;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+/*
+ * Read markers until SOS or EOI.
+ *
+ * Returns same codes as are defined for jpeg_consume_input:
+ * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
+ */
+
+METHODDEF(int)
+read_markers(j_decompress_ptr cinfo)
+{
+  /* Outer loop repeats once for each marker. */
+  for (;;) {
+    /* Collect the marker proper, unless we already did. */
+    /* NB: first_marker() enforces the requirement that SOI appear first. */
+    if (cinfo->unread_marker == 0) {
+      if (!cinfo->marker->saw_SOI) {
+        if (!first_marker(cinfo))
+          return JPEG_SUSPENDED;
+      } else {
+        if (!next_marker(cinfo))
+          return JPEG_SUSPENDED;
+      }
+    }
+    /* At this point cinfo->unread_marker contains the marker code and the
+     * input point is just past the marker proper, but before any parameters.
+     * A suspension will cause us to return with this state still true.
+     */
+    switch (cinfo->unread_marker) {
+    case M_SOI:
+      if (!get_soi(cinfo))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF0:                /* Baseline */
+    case M_SOF1:                /* Extended sequential, Huffman */
+      if (!get_sof(cinfo, FALSE, FALSE))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF2:                /* Progressive, Huffman */
+      if (!get_sof(cinfo, TRUE, FALSE))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF9:                /* Extended sequential, arithmetic */
+      if (!get_sof(cinfo, FALSE, TRUE))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF10:               /* Progressive, arithmetic */
+      if (!get_sof(cinfo, TRUE, TRUE))
+        return JPEG_SUSPENDED;
+      break;
+
+    /* Currently unsupported SOFn types */
+    case M_SOF3:                /* Lossless, Huffman */
+    case M_SOF5:                /* Differential sequential, Huffman */
+    case M_SOF6:                /* Differential progressive, Huffman */
+    case M_SOF7:                /* Differential lossless, Huffman */
+    case M_JPG:                 /* Reserved for JPEG extensions */
+    case M_SOF11:               /* Lossless, arithmetic */
+    case M_SOF13:               /* Differential sequential, arithmetic */
+    case M_SOF14:               /* Differential progressive, arithmetic */
+    case M_SOF15:               /* Differential lossless, arithmetic */
+      ERREXIT1(cinfo, JERR_SOF_UNSUPPORTED, cinfo->unread_marker);
+      break;
+
+    case M_SOS:
+      if (!get_sos(cinfo))
+        return JPEG_SUSPENDED;
+      cinfo->unread_marker = 0; /* processed the marker */
+      return JPEG_REACHED_SOS;
+
+    case M_EOI:
+      TRACEMS(cinfo, 1, JTRC_EOI);
+      cinfo->unread_marker = 0; /* processed the marker */
+      return JPEG_REACHED_EOI;
+
+    case M_DAC:
+      if (!get_dac(cinfo))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_DHT:
+      if (!get_dht(cinfo))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_DQT:
+      if (!get_dqt(cinfo))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_DRI:
+      if (!get_dri(cinfo))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_APP0:
+    case M_APP1:
+    case M_APP2:
+    case M_APP3:
+    case M_APP4:
+    case M_APP5:
+    case M_APP6:
+    case M_APP7:
+    case M_APP8:
+    case M_APP9:
+    case M_APP10:
+    case M_APP11:
+    case M_APP12:
+    case M_APP13:
+    case M_APP14:
+    case M_APP15:
+      if (!(*((my_marker_ptr)cinfo->marker)->process_APPn
+               cinfo->unread_marker - (int)M_APP0) (cinfo))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_COM:
+      if (!(*((my_marker_ptr)cinfo->marker)->process_COM) (cinfo))
+        return JPEG_SUSPENDED;
+      break;
+
+    case M_RST0:                /* these are all parameterless */
+    case M_RST1:
+    case M_RST2:
+    case M_RST3:
+    case M_RST4:
+    case M_RST5:
+    case M_RST6:
+    case M_RST7:
+    case M_TEM:
+      TRACEMS1(cinfo, 1, JTRC_PARMLESS_MARKER, cinfo->unread_marker);
+      break;
+
+    case M_DNL:                 /* Ignore DNL ... perhaps the wrong thing */
+      if (!skip_variable(cinfo))
+        return JPEG_SUSPENDED;
+      break;
+
+    default:                    /* must be DHP, EXP, JPGn, or RESn */
+      /* For now, we treat the reserved markers as fatal errors since they are
+       * likely to be used to signal incompatible JPEG Part 3 extensions.
+       * Once the JPEG 3 version-number marker is well defined, this code
+       * ought to change!
+       */
+      ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, cinfo->unread_marker);
+      break;
+    }
+    /* Successfully processed marker, so reset state variable */
+    cinfo->unread_marker = 0;
+  } /* end loop */
+}
+
+
+/*
+ * Read a restart marker, which is expected to appear next in the datastream;
+ * if the marker is not there, take appropriate recovery action.
+ * Returns FALSE if suspension is required.
+ *
+ * This is called by the entropy decoder after it has read an appropriate
+ * number of MCUs.  cinfo->unread_marker may be nonzero if the entropy decoder
+ * has already read a marker from the data source.  Under normal conditions
+ * cinfo->unread_marker will be reset to 0 before returning; if not reset,
+ * it holds a marker which the decoder will be unable to read past.
+ */
+
+METHODDEF(boolean)
+read_restart_marker(j_decompress_ptr cinfo)
+{
+  /* Obtain a marker unless we already did. */
+  /* Note that next_marker will complain if it skips any data. */
+  if (cinfo->unread_marker == 0) {
+    if (!next_marker(cinfo))
+      return FALSE;
+  }
+
+  if (cinfo->unread_marker ==
+      ((int)M_RST0 + cinfo->marker->next_restart_num)) {
+    /* Normal case --- swallow the marker and let entropy decoder continue */
+    TRACEMS1(cinfo, 3, JTRC_RST, cinfo->marker->next_restart_num);
+    cinfo->unread_marker = 0;
+  } else {
+    /* Uh-oh, the restart markers have been messed up. */
+    /* Let the data source manager determine how to resync. */
+    if (!(*cinfo->src->resync_to_restart) (cinfo,
+                                           cinfo->marker->next_restart_num))
+      return FALSE;
+  }
+
+  /* Update next-restart state */
+  cinfo->marker->next_restart_num = (cinfo->marker->next_restart_num + 1) & 7;
+
+  return TRUE;
+}
+
+
+/*
+ * This is the default resync_to_restart method for data source managers
+ * to use if they don't have any better approach.  Some data source managers
+ * may be able to back up, or may have additional knowledge about the data
+ * which permits a more intelligent recovery strategy; such managers would
+ * presumably supply their own resync method.
+ *
+ * read_restart_marker calls resync_to_restart if it finds a marker other than
+ * the restart marker it was expecting.  (This code is *not* used unless
+ * a nonzero restart interval has been declared.)  cinfo->unread_marker is
+ * the marker code actually found (might be anything, except 0 or FF).
+ * The desired restart marker number (0..7) is passed as a parameter.
+ * This routine is supposed to apply whatever error recovery strategy seems
+ * appropriate in order to position the input stream to the next data segment.
+ * Note that cinfo->unread_marker is treated as a marker appearing before
+ * the current data-source input point; usually it should be reset to zero
+ * before returning.
+ * Returns FALSE if suspension is required.
+ *
+ * This implementation is substantially constrained by wanting to treat the
+ * input as a data stream; this means we can't back up.  Therefore, we have
+ * only the following actions to work with:
+ *   1. Simply discard the marker and let the entropy decoder resume at next
+ *      byte of file.
+ *   2. Read forward until we find another marker, discarding intervening
+ *      data.  (In theory we could look ahead within the current bufferload,
+ *      without having to discard data if we don't find the desired marker.
+ *      This idea is not implemented here, in part because it makes behavior
+ *      dependent on buffer size and chance buffer-boundary positions.)
+ *   3. Leave the marker unread (by failing to zero cinfo->unread_marker).
+ *      This will cause the entropy decoder to process an empty data segment,
+ *      inserting dummy zeroes, and then we will reprocess the marker.
+ *
+ * #2 is appropriate if we think the desired marker lies ahead, while #3 is
+ * appropriate if the found marker is a future restart marker (indicating
+ * that we have missed the desired restart marker, probably because it got
+ * corrupted).
+ * We apply #2 or #3 if the found marker is a restart marker no more than
+ * two counts behind or ahead of the expected one.  We also apply #2 if the
+ * found marker is not a legal JPEG marker code (it's certainly bogus data).
+ * If the found marker is a restart marker more than 2 counts away, we do #1
+ * (too much risk that the marker is erroneous; with luck we will be able to
+ * resync at some future point).
+ * For any valid non-restart JPEG marker, we apply #3.  This keeps us from
+ * overrunning the end of a scan.  An implementation limited to single-scan
+ * files might find it better to apply #2 for markers other than EOI, since
+ * any other marker would have to be bogus data in that case.
+ */
+
+GLOBAL(boolean)
+jpeg_resync_to_restart(j_decompress_ptr cinfo, int desired)
+{
+  int marker = cinfo->unread_marker;
+  int action = 1;
+
+  /* Always put up a warning. */
+  WARNMS2(cinfo, JWRN_MUST_RESYNC, marker, desired);
+
+  /* Outer loop handles repeated decision after scanning forward. */
+  for (;;) {
+    if (marker < (int)M_SOF0)
+      action = 2;               /* invalid marker */
+    else if (marker < (int)M_RST0 || marker > (int)M_RST7)
+      action = 3;               /* valid non-restart marker */
+    else {
+      if (marker == ((int)M_RST0 + ((desired + 1) & 7)) ||
+          marker == ((int)M_RST0 + ((desired + 2) & 7)))
+        action = 3;             /* one of the next two expected restarts */
+      else if (marker == ((int)M_RST0 + ((desired - 1) & 7)) ||
+               marker == ((int)M_RST0 + ((desired - 2) & 7)))
+        action = 2;             /* a prior restart, so advance */
+      else
+        action = 1;             /* desired restart or too far away */
+    }
+    TRACEMS2(cinfo, 4, JTRC_RECOVERY_ACTION, marker, action);
+    switch (action) {
+    case 1:
+      /* Discard marker and let entropy decoder resume processing. */
+      cinfo->unread_marker = 0;
+      return TRUE;
+    case 2:
+      /* Scan to the next marker, and repeat the decision loop. */
+      if (!next_marker(cinfo))
+        return FALSE;
+      marker = cinfo->unread_marker;
+      break;
+    case 3:
+      /* Return without advancing past this marker. */
+      /* Entropy decoder will be forced to process an empty segment. */
+      return TRUE;
+    }
+  } /* end loop */
+}
+
+
+/*
+ * Reset marker processing state to begin a fresh datastream.
+ */
+
+METHODDEF(void)
+reset_marker_reader(j_decompress_ptr cinfo)
+{
+  my_marker_ptr marker = (my_marker_ptr)cinfo->marker;
+
+  cinfo->comp_info = NULL;              /* until allocated by get_sof */
+  cinfo->input_scan_number = 0;         /* no SOS seen yet */
+  cinfo->unread_marker = 0;             /* no pending marker */
+  marker->pub.saw_SOI = FALSE;          /* set internal state too */
+  marker->pub.saw_SOF = FALSE;
+  marker->pub.discarded_bytes = 0;
+  marker->cur_marker = NULL;
+}
+
+
+/*
+ * Initialize the marker reader module.
+ * This is called only once, when the decompression object is created.
+ */
+
+GLOBAL(void)
+jinit_marker_reader(j_decompress_ptr cinfo)
+{
+  my_marker_ptr marker;
+  int i;
+
+  /* Create subobject in permanent pool */
+  marker = (my_marker_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_PERMANENT,
+                                sizeof(my_marker_reader));
+  cinfo->marker = (struct jpeg_marker_reader *)marker;
+  /* Initialize public method pointers */
+  marker->pub.reset_marker_reader = reset_marker_reader;
+  marker->pub.read_markers = read_markers;
+  marker->pub.read_restart_marker = read_restart_marker;
+  /* Initialize COM/APPn processing.
+   * By default, we examine and then discard APP0 and APP14,
+   * but simply discard COM and all other APPn.
+   */
+  marker->process_COM = skip_variable;
+  marker->length_limit_COM = 0;
+  for (i = 0; i < 16; i++) {
+    marker->process_APPni = skip_variable;
+    marker->length_limit_APPni = 0;
+  }
+  marker->process_APPn0 = get_interesting_appn;
+  marker->process_APPn14 = get_interesting_appn;
+  /* Reset marker processing state */
+  reset_marker_reader(cinfo);
+}
+
+
+/*
+ * Control saving of COM and APPn markers into marker_list.
+ */
+
+#ifdef SAVE_MARKERS_SUPPORTED
+
+GLOBAL(void)
+jpeg_save_markers(j_decompress_ptr cinfo, int marker_code,
+                  unsigned int length_limit)
+{
+  my_marker_ptr marker = (my_marker_ptr)cinfo->marker;
+  long maxlength;
+  jpeg_marker_parser_method processor;
+
+  /* Length limit mustn't be larger than what we can allocate
+   * (should only be a concern in a 16-bit environment).
+   */
+  maxlength = cinfo->mem->max_alloc_chunk - sizeof(struct jpeg_marker_struct);
+  if (((long)length_limit) > maxlength)
+    length_limit = (unsigned int)maxlength;
+
+  /* Choose processor routine to use.
+   * APP0/APP14 have special requirements.
+   */
+  if (length_limit) {
+    processor = save_marker;
+    /* If saving APP0/APP14, save at least enough for our internal use. */
+    if (marker_code == (int)M_APP0 && length_limit < APP0_DATA_LEN)
+      length_limit = APP0_DATA_LEN;
+    else if (marker_code == (int)M_APP14 && length_limit < APP14_DATA_LEN)
+      length_limit = APP14_DATA_LEN;
+  } else {
+    processor = skip_variable;
+    /* If discarding APP0/APP14, use our regular on-the-fly processor. */
+    if (marker_code == (int)M_APP0 || marker_code == (int)M_APP14)
+      processor = get_interesting_appn;
+  }
+
+  if (marker_code == (int)M_COM) {
+    marker->process_COM = processor;
+    marker->length_limit_COM = length_limit;
+  } else if (marker_code >= (int)M_APP0 && marker_code <= (int)M_APP15) {
+    marker->process_APPnmarker_code - (int)M_APP0 = processor;
+    marker->length_limit_APPnmarker_code - (int)M_APP0 = length_limit;
+  } else
+    ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, marker_code);
+}
+
+#endif /* SAVE_MARKERS_SUPPORTED */
+
+
+/*
+ * Install a special processing method for COM or APPn markers.
+ */
+
+GLOBAL(void)
+jpeg_set_marker_processor(j_decompress_ptr cinfo, int marker_code,
+                          jpeg_marker_parser_method routine)
+{
+  my_marker_ptr marker = (my_marker_ptr)cinfo->marker;
+
+  if (marker_code == (int)M_COM)
+    marker->process_COM = routine;
+  else if (marker_code >= (int)M_APP0 && marker_code <= (int)M_APP15)
+    marker->process_APPnmarker_code - (int)M_APP0 = routine;
+  else
+    ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, marker_code);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdmaster.c Added

@@ -0,0 +1,726 @@
+/*
+ * jdmaster.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2002-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2016, 2019, D. R. Commander.
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains master control logic for the JPEG decompressor.
+ * These routines are concerned with selecting the modules to be executed
+ * and with determining the number of passes and the work to be done in each
+ * pass.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jpegcomp.h"
+#include "jdmaster.h"
+
+
+/*
+ * Determine whether merged upsample/color conversion should be used.
+ * CRUCIAL: this must match the actual capabilities of jdmerge.c!
+ */
+
+LOCAL(boolean)
+use_merged_upsample(j_decompress_ptr cinfo)
+{
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+  /* Merging is the equivalent of plain box-filter upsampling */
+  if (cinfo->do_fancy_upsampling || cinfo->CCIR601_sampling)
+    return FALSE;
+  /* jdmerge.c only supports YCC=>RGB and YCC=>RGB565 color conversion */
+  if (cinfo->jpeg_color_space != JCS_YCbCr || cinfo->num_components != 3 ||
+      (cinfo->out_color_space != JCS_RGB &&
+       cinfo->out_color_space != JCS_RGB565 &&
+       cinfo->out_color_space != JCS_EXT_RGB &&
+       cinfo->out_color_space != JCS_EXT_RGBX &&
+       cinfo->out_color_space != JCS_EXT_BGR &&
+       cinfo->out_color_space != JCS_EXT_BGRX &&
+       cinfo->out_color_space != JCS_EXT_XBGR &&
+       cinfo->out_color_space != JCS_EXT_XRGB &&
+       cinfo->out_color_space != JCS_EXT_RGBA &&
+       cinfo->out_color_space != JCS_EXT_BGRA &&
+       cinfo->out_color_space != JCS_EXT_ABGR &&
+       cinfo->out_color_space != JCS_EXT_ARGB))
+    return FALSE;
+  if ((cinfo->out_color_space == JCS_RGB565 &&
+       cinfo->out_color_components != 3) ||
+      (cinfo->out_color_space != JCS_RGB565 &&
+       cinfo->out_color_components != rgb_pixelsizecinfo->out_color_space))
+    return FALSE;
+  /* and it only handles 2h1v or 2h2v sampling ratios */
+  if (cinfo->comp_info0.h_samp_factor != 2 ||
+      cinfo->comp_info1.h_samp_factor != 1 ||
+      cinfo->comp_info2.h_samp_factor != 1 ||
+      cinfo->comp_info0.v_samp_factor >  2 ||
+      cinfo->comp_info1.v_samp_factor != 1 ||
+      cinfo->comp_info2.v_samp_factor != 1)
+    return FALSE;
+  /* furthermore, it doesn't work if we've scaled the IDCTs differently */
+  if (cinfo->comp_info0._DCT_scaled_size != cinfo->_min_DCT_scaled_size ||
+      cinfo->comp_info1._DCT_scaled_size != cinfo->_min_DCT_scaled_size ||
+      cinfo->comp_info2._DCT_scaled_size != cinfo->_min_DCT_scaled_size)
+    return FALSE;
+  /* ??? also need to test for upsample-time rescaling, when & if supported */
+  return TRUE;                  /* by golly, it'll work... */
+#else
+  return FALSE;
+#endif
+}
+
+
+/*
+ * Compute output image dimensions and related values.
+ * NOTE: this is exported for possible use by application.
+ * Hence it mustn't do anything that can't be done twice.
+ */
+
+#if JPEG_LIB_VERSION >= 80
+GLOBAL(void)
+#else
+LOCAL(void)
+#endif
+jpeg_core_output_dimensions(j_decompress_ptr cinfo)
+/* Do computations that are needed before master selection phase.
+ * This function is used for transcoding and full decompression.
+ */
+{
+#ifdef IDCT_SCALING_SUPPORTED
+  int ci;
+  jpeg_component_info *compptr;
+
+  /* Compute actual output image dimensions and DCT scaling choices. */
+  if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom) {
+    /* Provide 1/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 1;
+    cinfo->_min_DCT_v_scaled_size = 1;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 2) {
+    /* Provide 2/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 2L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 2L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 2;
+    cinfo->_min_DCT_v_scaled_size = 2;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 3) {
+    /* Provide 3/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 3L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 3L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 3;
+    cinfo->_min_DCT_v_scaled_size = 3;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 4) {
+    /* Provide 4/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 4L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 4L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 4;
+    cinfo->_min_DCT_v_scaled_size = 4;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 5) {
+    /* Provide 5/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 5L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 5L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 5;
+    cinfo->_min_DCT_v_scaled_size = 5;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 6) {
+    /* Provide 6/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 6L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 6L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 6;
+    cinfo->_min_DCT_v_scaled_size = 6;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 7) {
+    /* Provide 7/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 7L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 7L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 7;
+    cinfo->_min_DCT_v_scaled_size = 7;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 8) {
+    /* Provide 8/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 8L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 8L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 8;
+    cinfo->_min_DCT_v_scaled_size = 8;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 9) {
+    /* Provide 9/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 9L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 9L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 9;
+    cinfo->_min_DCT_v_scaled_size = 9;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 10) {
+    /* Provide 10/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 10L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 10L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 10;
+    cinfo->_min_DCT_v_scaled_size = 10;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 11) {
+    /* Provide 11/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 11L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 11L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 11;
+    cinfo->_min_DCT_v_scaled_size = 11;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 12) {
+    /* Provide 12/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 12L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 12L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 12;
+    cinfo->_min_DCT_v_scaled_size = 12;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 13) {
+    /* Provide 13/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 13L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 13L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 13;
+    cinfo->_min_DCT_v_scaled_size = 13;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 14) {
+    /* Provide 14/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 14L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 14L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 14;
+    cinfo->_min_DCT_v_scaled_size = 14;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 15) {
+    /* Provide 15/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 15L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 15L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 15;
+    cinfo->_min_DCT_v_scaled_size = 15;
+  } else {
+    /* Provide 16/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width * 16L, (long)DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height * 16L, (long)DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 16;
+    cinfo->_min_DCT_v_scaled_size = 16;
+  }
+
+  /* Recompute dimensions of components */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    compptr->_DCT_h_scaled_size = cinfo->_min_DCT_h_scaled_size;
+    compptr->_DCT_v_scaled_size = cinfo->_min_DCT_v_scaled_size;
+  }
+
+#else /* !IDCT_SCALING_SUPPORTED */
+
+  /* Hardwire it to "no scaling" */
+  cinfo->output_width = cinfo->image_width;
+  cinfo->output_height = cinfo->image_height;
+  /* jdinput.c has already initialized DCT_scaled_size,
+   * and has computed unscaled downsampled_width and downsampled_height.
+   */
+
+#endif /* IDCT_SCALING_SUPPORTED */
+}
+
+
+/*
+ * Compute output image dimensions and related values.
+ * NOTE: this is exported for possible use by application.
+ * Hence it mustn't do anything that can't be done twice.
+ * Also note that it may be called before the master module is initialized!
+ */
+
+GLOBAL(void)
+jpeg_calc_output_dimensions(j_decompress_ptr cinfo)
+/* Do computations that are needed before master selection phase */
+{
+#ifdef IDCT_SCALING_SUPPORTED
+  int ci;
+  jpeg_component_info *compptr;
+#endif
+
+  /* Prevent application from calling me at wrong times */
+  if (cinfo->global_state != DSTATE_READY)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Compute core output image dimensions and DCT scaling choices. */
+  jpeg_core_output_dimensions(cinfo);
+
+#ifdef IDCT_SCALING_SUPPORTED
+
+  /* In selecting the actual DCT scaling for each component, we try to
+   * scale up the chroma components via IDCT scaling rather than upsampling.
+   * This saves time if the upsampler gets to use 1:1 scaling.
+   * Note this code adapts subsampling ratios which are powers of 2.
+   */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    int ssize = cinfo->_min_DCT_scaled_size;
+    while (ssize < DCTSIZE &&
+           ((cinfo->max_h_samp_factor * cinfo->_min_DCT_scaled_size) %
+            (compptr->h_samp_factor * ssize * 2) == 0) &&
+           ((cinfo->max_v_samp_factor * cinfo->_min_DCT_scaled_size) %
+            (compptr->v_samp_factor * ssize * 2) == 0)) {
+      ssize = ssize * 2;
+    }
+#if JPEG_LIB_VERSION >= 70
+    compptr->DCT_h_scaled_size = compptr->DCT_v_scaled_size = ssize;
+#else
+    compptr->DCT_scaled_size = ssize;
+#endif
+  }
+
+  /* Recompute downsampled dimensions of components;
+   * application needs to know these if using raw downsampled data.
+   */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Size in samples, after IDCT scaling */
+    compptr->downsampled_width = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_width *
+                    (long)(compptr->h_samp_factor * compptr->_DCT_scaled_size),
+                    (long)(cinfo->max_h_samp_factor * DCTSIZE));
+    compptr->downsampled_height = (JDIMENSION)
+      jdiv_round_up((long)cinfo->image_height *
+                    (long)(compptr->v_samp_factor * compptr->_DCT_scaled_size),
+                    (long)(cinfo->max_v_samp_factor * DCTSIZE));
+  }
+
+#else /* !IDCT_SCALING_SUPPORTED */
+
+  /* Hardwire it to "no scaling" */
+  cinfo->output_width = cinfo->image_width;
+  cinfo->output_height = cinfo->image_height;
+  /* jdinput.c has already initialized DCT_scaled_size to DCTSIZE,
+   * and has computed unscaled downsampled_width and downsampled_height.
+   */
+
+#endif /* IDCT_SCALING_SUPPORTED */
+
+  /* Report number of components in selected colorspace. */
+  /* Probably this should be in the color conversion module... */
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    cinfo->out_color_components = 1;
+    break;
+  case JCS_RGB:
+  case JCS_EXT_RGB:
+  case JCS_EXT_RGBX:
+  case JCS_EXT_BGR:
+  case JCS_EXT_BGRX:
+  case JCS_EXT_XBGR:
+  case JCS_EXT_XRGB:
+  case JCS_EXT_RGBA:
+  case JCS_EXT_BGRA:
+  case JCS_EXT_ABGR:
+  case JCS_EXT_ARGB:
+    cinfo->out_color_components = rgb_pixelsizecinfo->out_color_space;
+    break;
+  case JCS_YCbCr:
+  case JCS_RGB565:
+    cinfo->out_color_components = 3;
+    break;
+  case JCS_CMYK:
+  case JCS_YCCK:
+    cinfo->out_color_components = 4;
+    break;
+  default:                      /* else must be same colorspace as in file */
+    cinfo->out_color_components = cinfo->num_components;
+    break;
+  }
+  cinfo->output_components = (cinfo->quantize_colors ? 1 :
+                              cinfo->out_color_components);
+
+  /* See if upsampler will want to emit more than one row at a time */
+  if (use_merged_upsample(cinfo))
+    cinfo->rec_outbuf_height = cinfo->max_v_samp_factor;
+  else
+    cinfo->rec_outbuf_height = 1;
+}
+
+
+/*
+ * Several decompression processes need to range-limit values to the range
+ * 0..MAXJSAMPLE; the input value may fall somewhat outside this range
+ * due to noise introduced by quantization, roundoff error, etc.  These
+ * processes are inner loops and need to be as fast as possible.  On most
+ * machines, particularly CPUs with pipelines or instruction prefetch,
+ * a (subscript-check-less) C table lookup
+ *              x = sample_range_limitx;
+ * is faster than explicit tests
+ *              if (x < 0)  x = 0;
+ *              else if (x > MAXJSAMPLE)  x = MAXJSAMPLE;
+ * These processes all use a common table prepared by the routine below.
+ *
+ * For most steps we can mathematically guarantee that the initial value
+ * of x is within MAXJSAMPLE+1 of the legal range, so a table running from
+ * -(MAXJSAMPLE+1) to 2*MAXJSAMPLE+1 is sufficient.  But for the initial
+ * limiting step (just after the IDCT), a wildly out-of-range value is
+ * possible if the input data is corrupt.  To avoid any chance of indexing
+ * off the end of memory and getting a bad-pointer trap, we perform the
+ * post-IDCT limiting thus:
+ *              x = range_limitx & MASK;
+ * where MASK is 2 bits wider than legal sample data, ie 10 bits for 8-bit
+ * samples.  Under normal circumstances this is more than enough range and
+ * a correct output will be generated; with bogus input data the mask will
+ * cause wraparound, and we will safely generate a bogus-but-in-range output.
+ * For the post-IDCT step, we want to convert the data from signed to unsigned
+ * representation by adding CENTERJSAMPLE at the same time that we limit it.
+ * So the post-IDCT limiting table ends up looking like this:
+ *   CENTERJSAMPLE,CENTERJSAMPLE+1,...,MAXJSAMPLE,
+ *   MAXJSAMPLE (repeat 2*(MAXJSAMPLE+1)-CENTERJSAMPLE times),
+ *   0          (repeat 2*(MAXJSAMPLE+1)-CENTERJSAMPLE times),
+ *   0,1,...,CENTERJSAMPLE-1
+ * Negative inputs select values from the upper half of the table after
+ * masking.
+ *
+ * We can save some space by overlapping the start of the post-IDCT table
+ * with the simpler range limiting table.  The post-IDCT table begins at
+ * sample_range_limit + CENTERJSAMPLE.
+ */
+
+LOCAL(void)
+prepare_range_limit_table(j_decompress_ptr cinfo)
+/* Allocate and fill in the sample_range_limit table */
+{
+  JSAMPLE *table;
+  int i;
+
+  table = (JSAMPLE *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                (5 * (MAXJSAMPLE + 1) + CENTERJSAMPLE) * sizeof(JSAMPLE));
+  table += (MAXJSAMPLE + 1);    /* allow negative subscripts of simple table */
+  cinfo->sample_range_limit = table;
+  /* First segment of "simple" table: limitx = 0 for x < 0 */
+  MEMZERO(table - (MAXJSAMPLE + 1), (MAXJSAMPLE + 1) * sizeof(JSAMPLE));
+  /* Main part of "simple" table: limitx = x */
+  for (i = 0; i <= MAXJSAMPLE; i++)
+    tablei = (JSAMPLE)i;
+  table += CENTERJSAMPLE;       /* Point to where post-IDCT table starts */
+  /* End of simple table, rest of first half of post-IDCT table */
+  for (i = CENTERJSAMPLE; i < 2 * (MAXJSAMPLE + 1); i++)
+    tablei = MAXJSAMPLE;
+  /* Second half of post-IDCT table */
+  MEMZERO(table + (2 * (MAXJSAMPLE + 1)),
+          (2 * (MAXJSAMPLE + 1) - CENTERJSAMPLE) * sizeof(JSAMPLE));
+  MEMCOPY(table + (4 * (MAXJSAMPLE + 1) - CENTERJSAMPLE),
+          cinfo->sample_range_limit, CENTERJSAMPLE * sizeof(JSAMPLE));
+}
+
+
+/*
+ * Master selection of decompression modules.
+ * This is done once at jpeg_start_decompress time.  We determine
+ * which modules will be used and give them appropriate initialization calls.
+ * We also initialize the decompressor input side to begin consuming data.
+ *
+ * Since jpeg_read_header has finished, we know what is in the SOF
+ * and (first) SOS markers.  We also have all the application parameter
+ * settings.
+ */
+
+LOCAL(void)
+master_selection(j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+  boolean use_c_buffer;
+  long samplesperrow;
+  JDIMENSION jd_samplesperrow;
+
+  /* Initialize dimensions and other stuff */
+  jpeg_calc_output_dimensions(cinfo);
+  prepare_range_limit_table(cinfo);
+
+  /* Width of an output scanline must be representable as JDIMENSION. */
+  samplesperrow = (long)cinfo->output_width *
+                  (long)cinfo->out_color_components;
+  jd_samplesperrow = (JDIMENSION)samplesperrow;
+  if ((long)jd_samplesperrow != samplesperrow)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+
+  /* Initialize my private state */
+  master->pass_number = 0;
+  master->using_merged_upsample = use_merged_upsample(cinfo);
+
+  /* Color quantizer selection */
+  master->quantizer_1pass = NULL;
+  master->quantizer_2pass = NULL;
+  /* No mode changes if not using buffered-image mode. */
+  if (!cinfo->quantize_colors || !cinfo->buffered_image) {
+    cinfo->enable_1pass_quant = FALSE;
+    cinfo->enable_external_quant = FALSE;
+    cinfo->enable_2pass_quant = FALSE;
+  }
+  if (cinfo->quantize_colors) {
+    if (cinfo->raw_data_out)
+      ERREXIT(cinfo, JERR_NOTIMPL);
+    /* 2-pass quantizer only works in 3-component color space. */
+    if (cinfo->out_color_components != 3) {
+      cinfo->enable_1pass_quant = TRUE;
+      cinfo->enable_external_quant = FALSE;
+      cinfo->enable_2pass_quant = FALSE;
+      cinfo->colormap = NULL;
+    } else if (cinfo->colormap != NULL) {
+      cinfo->enable_external_quant = TRUE;
+    } else if (cinfo->two_pass_quantize) {
+      cinfo->enable_2pass_quant = TRUE;
+    } else {
+      cinfo->enable_1pass_quant = TRUE;
+    }
+
+    if (cinfo->enable_1pass_quant) {
+#ifdef QUANT_1PASS_SUPPORTED
+      jinit_1pass_quantizer(cinfo);
+      master->quantizer_1pass = cinfo->cquantize;
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    }
+
+    /* We use the 2-pass code to map to external colormaps. */
+    if (cinfo->enable_2pass_quant || cinfo->enable_external_quant) {
+#ifdef QUANT_2PASS_SUPPORTED
+      jinit_2pass_quantizer(cinfo);
+      master->quantizer_2pass = cinfo->cquantize;
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    }
+    /* If both quantizers are initialized, the 2-pass one is left active;
+     * this is necessary for starting with quantization to an external map.
+     */
+  }
+
+  /* Post-processing: in particular, color conversion first */
+  if (!cinfo->raw_data_out) {
+    if (master->using_merged_upsample) {
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+      jinit_merged_upsampler(cinfo); /* does color conversion too */
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else {
+      jinit_color_deconverter(cinfo);
+      jinit_upsampler(cinfo);
+    }
+    jinit_d_post_controller(cinfo, cinfo->enable_2pass_quant);
+  }
+  /* Inverse DCT */
+  jinit_inverse_dct(cinfo);
+  /* Entropy decoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+#ifdef D_ARITH_CODING_SUPPORTED
+    jinit_arith_decoder(cinfo);
+#else
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+#endif
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef D_PROGRESSIVE_SUPPORTED
+      jinit_phuff_decoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_decoder(cinfo);
+  }
+
+  /* Initialize principal buffer controllers. */
+  use_c_buffer = cinfo->inputctl->has_multiple_scans || cinfo->buffered_image;
+  jinit_d_coef_controller(cinfo, use_c_buffer);
+
+  if (!cinfo->raw_data_out)
+    jinit_d_main_controller(cinfo, FALSE /* never need full buffer here */);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr)cinfo);
+
+  /* Initialize input side of decompressor to consume first scan. */
+  (*cinfo->inputctl->start_input_pass) (cinfo);
+
+  /* Set the first and last iMCU columns to decompress from single-scan images.
+   * By default, decompress all of the iMCU columns.
+   */
+  cinfo->master->first_iMCU_col = 0;
+  cinfo->master->last_iMCU_col = cinfo->MCUs_per_row - 1;
+  cinfo->master->last_good_iMCU_row = 0;
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+  /* If jpeg_start_decompress will read the whole file, initialize
+   * progress monitoring appropriately.  The input step is counted
+   * as one pass.
+   */
+  if (cinfo->progress != NULL && !cinfo->buffered_image &&
+      cinfo->inputctl->has_multiple_scans) {
+    int nscans;
+    /* Estimate number of scans to set pass_limit. */
+    if (cinfo->progressive_mode) {
+      /* Arbitrarily estimate 2 interleaved DC scans + 3 AC scans/component. */
+      nscans = 2 + 3 * cinfo->num_components;
+    } else {
+      /* For a nonprogressive multiscan file, estimate 1 scan per component. */
+      nscans = cinfo->num_components;
+    }
+    cinfo->progress->pass_counter = 0L;
+    cinfo->progress->pass_limit = (long)cinfo->total_iMCU_rows * nscans;
+    cinfo->progress->completed_passes = 0;
+    cinfo->progress->total_passes = (cinfo->enable_2pass_quant ? 3 : 2);
+    /* Count the input pass as done */
+    master->pass_number++;
+  }
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+}
+
+
+/*
+ * Per-pass setup.
+ * This is called at the beginning of each output pass.  We determine which
+ * modules will be active during this pass and give them appropriate
+ * start_pass calls.  We also set is_dummy_pass to indicate whether this
+ * is a "real" output pass or a dummy pass for color quantization.
+ * (In the latter case, jdapistd.c will crank the pass to completion.)
+ */
+
+METHODDEF(void)
+prepare_for_output_pass(j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+
+  if (master->pub.is_dummy_pass) {
+#ifdef QUANT_2PASS_SUPPORTED
+    /* Final pass of 2-pass quantization */
+    master->pub.is_dummy_pass = FALSE;
+    (*cinfo->cquantize->start_pass) (cinfo, FALSE);
+    (*cinfo->post->start_pass) (cinfo, JBUF_CRANK_DEST);
+    (*cinfo->main->start_pass) (cinfo, JBUF_CRANK_DEST);
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif /* QUANT_2PASS_SUPPORTED */
+  } else {
+    if (cinfo->quantize_colors && cinfo->colormap == NULL) {
+      /* Select new quantization method */
+      if (cinfo->two_pass_quantize && cinfo->enable_2pass_quant) {
+        cinfo->cquantize = master->quantizer_2pass;
+        master->pub.is_dummy_pass = TRUE;
+      } else if (cinfo->enable_1pass_quant) {
+        cinfo->cquantize = master->quantizer_1pass;
+      } else {
+        ERREXIT(cinfo, JERR_MODE_CHANGE);
+      }
+    }
+    (*cinfo->idct->start_pass) (cinfo);
+    (*cinfo->coef->start_output_pass) (cinfo);
+    if (!cinfo->raw_data_out) {
+      if (!master->using_merged_upsample)
+        (*cinfo->cconvert->start_pass) (cinfo);
+      (*cinfo->upsample->start_pass) (cinfo);
+      if (cinfo->quantize_colors)
+        (*cinfo->cquantize->start_pass) (cinfo, master->pub.is_dummy_pass);
+      (*cinfo->post->start_pass) (cinfo,
+            (master->pub.is_dummy_pass ? JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+      (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
+    }
+  }
+
+  /* Set up progress monitor's pass info if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->completed_passes = master->pass_number;
+    cinfo->progress->total_passes = master->pass_number +
+                                    (master->pub.is_dummy_pass ? 2 : 1);
+    /* In buffered-image mode, we assume one more output pass if EOI not
+     * yet reached, but no more passes if EOI has been reached.
+     */
+    if (cinfo->buffered_image && !cinfo->inputctl->eoi_reached) {
+      cinfo->progress->total_passes += (cinfo->enable_2pass_quant ? 2 : 1);
+    }
+  }
+}
+
+
+/*
+ * Finish up at end of an output pass.
+ */
+
+METHODDEF(void)
+finish_output_pass(j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+
+  if (cinfo->quantize_colors)
+    (*cinfo->cquantize->finish_pass) (cinfo);
+  master->pass_number++;
+}
+
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+
+/*
+ * Switch to a new external colormap between output passes.
+ */
+
+GLOBAL(void)
+jpeg_new_colormap(j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+
+  /* Prevent application from calling me at wrong times */
+  if (cinfo->global_state != DSTATE_BUFIMAGE)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (cinfo->quantize_colors && cinfo->enable_external_quant &&
+      cinfo->colormap != NULL) {
+    /* Select 2-pass quantizer for external colormap use */
+    cinfo->cquantize = master->quantizer_2pass;
+    /* Notify quantizer of colormap change */
+    (*cinfo->cquantize->new_color_map) (cinfo);
+    master->pub.is_dummy_pass = FALSE; /* just in case */
+  } else
+    ERREXIT(cinfo, JERR_MODE_CHANGE);
+}
+
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+
+
+/*
+ * Initialize master decompression control and select active modules.
+ * This is performed at the start of jpeg_start_decompress.
+ */
+
+GLOBAL(void)
+jinit_master_decompress(j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr)cinfo->master;
+
+  master->pub.prepare_for_output_pass = prepare_for_output_pass;
+  master->pub.finish_output_pass = finish_output_pass;
+
+  master->pub.is_dummy_pass = FALSE;
+  master->pub.jinit_upsampler_no_alloc = FALSE;
+
+  master_selection(cinfo);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdmaster.h Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdmerge.c Added

@@ -0,0 +1,588 @@
+/*
+ * jdmerge.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2009, 2011, 2014-2015, 2020, D. R. Commander.
+ * Copyright (C) 2013, Linaro Limited.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ *
+ * This file combines functions from jdsample.c and jdcolor.c;
+ * read those files first to understand what's going on.
+ *
+ * When the chroma components are to be upsampled by simple replication
+ * (ie, box filtering), we can save some work in color conversion by
+ * calculating all the output pixels corresponding to a pair of chroma
+ * samples at one time.  In the conversion equations
+ *      R = Y           + K1 * Cr
+ *      G = Y + K2 * Cb + K3 * Cr
+ *      B = Y + K4 * Cb
+ * only the Y term varies among the group of pixels corresponding to a pair
+ * of chroma samples, so the rest of the terms can be calculated just once.
+ * At typical sampling ratios, this eliminates half or three-quarters of the
+ * multiplications needed for color conversion.
+ *
+ * This file currently provides implementations for the following cases:
+ *      YCbCr => RGB color conversion only.
+ *      Sampling ratios of 2h1v or 2h2v.
+ *      No scaling needed at upsample time.
+ *      Corner-aligned (non-CCIR601) sampling alignment.
+ * Other special cases could be added, but in most applications these are
+ * the only common cases.  (For uncommon cases we fall back on the more
+ * general code in jdsample.c and jdcolor.c.)
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdmerge.h"
+#include "jsimd.h"
+#include "jconfigint.h"
+
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+
+
+#define SCALEBITS       16      /* speediest right-shift on some machines */
+#define ONE_HALF        ((JLONG)1 << (SCALEBITS - 1))
+#define FIX(x)          ((JLONG)((x) * (1L << SCALEBITS) + 0.5))
+
+
+/* Include inline routines for colorspace extensions */
+
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define h2v1_merged_upsample_internal  extrgb_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal  extrgb_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_ALPHA  3
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define h2v1_merged_upsample_internal  extrgbx_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal  extrgbx_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define h2v1_merged_upsample_internal  extbgr_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal  extbgr_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_ALPHA  3
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define h2v1_merged_upsample_internal  extbgrx_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal  extbgrx_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_ALPHA  0
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define h2v1_merged_upsample_internal  extxbgr_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal  extxbgr_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_ALPHA  0
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define h2v1_merged_upsample_internal  extxrgb_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal  extxrgb_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+
+/*
+ * Initialize tables for YCC->RGB colorspace conversion.
+ * This is taken directly from jdcolor.c; see that file for more info.
+ */
+
+LOCAL(void)
+build_ycc_rgb_table(j_decompress_ptr cinfo)
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  int i;
+  JLONG x;
+  SHIFT_TEMPS
+
+  upsample->Cr_r_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE + 1) * sizeof(int));
+  upsample->Cb_b_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE + 1) * sizeof(int));
+  upsample->Cr_g_tab = (JLONG *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE + 1) * sizeof(JLONG));
+  upsample->Cb_g_tab = (JLONG *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (MAXJSAMPLE + 1) * sizeof(JLONG));
+
+  for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
+    /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
+    /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
+    /* Cr=>R value is nearest int to 1.40200 * x */
+    upsample->Cr_r_tabi = (int)
+                    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+    /* Cb=>B value is nearest int to 1.77200 * x */
+    upsample->Cb_b_tabi = (int)
+                    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+    /* Cr=>G value is scaled-up -0.71414 * x */
+    upsample->Cr_g_tabi = (-FIX(0.71414)) * x;
+    /* Cb=>G value is scaled-up -0.34414 * x */
+    /* We also add in ONE_HALF so that need not do it in inner loop */
+    upsample->Cb_g_tabi = (-FIX(0.34414)) * x + ONE_HALF;
+  }
+}
+
+
+/*
+ * Initialize for an upsampling pass.
+ */
+
+METHODDEF(void)
+start_pass_merged_upsample(j_decompress_ptr cinfo)
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+
+  /* Mark the spare buffer empty */
+  upsample->spare_full = FALSE;
+  /* Initialize total-height counter for detecting bottom of image */
+  upsample->rows_to_go = cinfo->output_height;
+}
+
+
+/*
+ * Control routine to do upsampling (and color conversion).
+ *
+ * The control routine just handles the row buffering considerations.
+ */
+
+METHODDEF(void)
+merged_2v_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                   JDIMENSION *in_row_group_ctr,
+                   JDIMENSION in_row_groups_avail, JSAMPARRAY output_buf,
+                   JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
+/* 2:1 vertical sampling case: may need a spare row. */
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  JSAMPROW work_ptrs2;
+  JDIMENSION num_rows;          /* number of rows returned to caller */
+
+  if (upsample->spare_full) {
+    /* If we have a spare row saved from a previous cycle, just return it. */
+    JDIMENSION size = upsample->out_row_width;
+    if (cinfo->out_color_space == JCS_RGB565)
+      size = cinfo->output_width * 2;
+    jcopy_sample_rows(&upsample->spare_row, 0, output_buf + *out_row_ctr, 0, 1,
+                      size);
+    num_rows = 1;
+    upsample->spare_full = FALSE;
+  } else {
+    /* Figure number of rows to return to caller. */
+    num_rows = 2;
+    /* Not more than the distance to the end of the image. */
+    if (num_rows > upsample->rows_to_go)
+      num_rows = upsample->rows_to_go;
+    /* And not more than what the client can accept: */
+    out_rows_avail -= *out_row_ctr;
+    if (num_rows > out_rows_avail)
+      num_rows = out_rows_avail;
+    /* Create output pointer array for upsampler. */
+    work_ptrs0 = output_buf*out_row_ctr;
+    if (num_rows > 1) {
+      work_ptrs1 = output_buf*out_row_ctr + 1;
+    } else {
+      work_ptrs1 = upsample->spare_row;
+      upsample->spare_full = TRUE;
+    }
+    /* Now do the upsampling. */
+    (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr, work_ptrs);
+  }
+
+  /* Adjust counts */
+  *out_row_ctr += num_rows;
+  upsample->rows_to_go -= num_rows;
+  /* When the buffer is emptied, declare this input row group consumed */
+  if (!upsample->spare_full)
+    (*in_row_group_ctr)++;
+}
+
+
+METHODDEF(void)
+merged_1v_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                   JDIMENSION *in_row_group_ctr,
+                   JDIMENSION in_row_groups_avail, JSAMPARRAY output_buf,
+                   JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
+/* 1:1 vertical sampling case: much easier, never need a spare row. */
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+
+  /* Just do the upsampling. */
+  (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr,
+                         output_buf + *out_row_ctr);
+  /* Adjust counts */
+  (*out_row_ctr)++;
+  (*in_row_group_ctr)++;
+}
+
+
+/*
+ * These are the routines invoked by the control routines to do
+ * the actual upsampling/conversion.  One row group is processed per call.
+ *
+ * Note: since we may be writing directly into application-supplied buffers,
+ * we have to be honest about the output width; we can't assume the buffer
+ * has been rounded up to an even width.
+ */
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ */
+
+METHODDEF(void)
+h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                     JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    extrgb_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                         output_buf);
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    extrgbx_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                          output_buf);
+    break;
+  case JCS_EXT_BGR:
+    extbgr_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                         output_buf);
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    extbgrx_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                          output_buf);
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    extxbgr_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                          output_buf);
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    extxrgb_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                          output_buf);
+    break;
+  default:
+    h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                  output_buf);
+    break;
+  }
+}
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+ */
+
+METHODDEF(void)
+h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                     JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    extrgb_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                         output_buf);
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    extrgbx_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                          output_buf);
+    break;
+  case JCS_EXT_BGR:
+    extbgr_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                         output_buf);
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    extbgrx_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                          output_buf);
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    extxbgr_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                          output_buf);
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    extxrgb_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                          output_buf);
+    break;
+  default:
+    h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                  output_buf);
+    break;
+  }
+}
+
+
+/*
+ * RGB565 conversion
+ */
+
+#define PACK_SHORT_565_LE(r, g, b) \
+  ((((r) << 8) & 0xF800) | (((g) << 3) & 0x7E0) | ((b) >> 3))
+#define PACK_SHORT_565_BE(r, g, b) \
+  (((r) & 0xF8) | ((g) >> 5) | (((g) << 11) & 0xE000) | (((b) << 5) & 0x1F00))
+
+#define PACK_TWO_PIXELS_LE(l, r)    ((r << 16) | l)
+#define PACK_TWO_PIXELS_BE(l, r)    ((l << 16) | r)
+
+#define WRITE_TWO_PIXELS_LE(addr, pixels) { \
+  ((INT16 *)(addr))0 = (INT16)(pixels); \
+  ((INT16 *)(addr))1 = (INT16)((pixels) >> 16); \
+}
+#define WRITE_TWO_PIXELS_BE(addr, pixels) { \
+  ((INT16 *)(addr))1 = (INT16)(pixels); \
+  ((INT16 *)(addr))0 = (INT16)((pixels) >> 16); \
+}
+
+#define DITHER_565_R(r, dither)  ((r) + ((dither) & 0xFF))
+#define DITHER_565_G(g, dither)  ((g) + (((dither) & 0xFF) >> 1))
+#define DITHER_565_B(b, dither)  ((b) + ((dither) & 0xFF))
+
+
+/* Declarations for ordered dithering
+ *
+ * We use a 4x4 ordered dither array packed into 32 bits.  This array is
+ * sufficient for dithering RGB888 to RGB565.
+ */
+
+#define DITHER_MASK       0x3
+#define DITHER_ROTATE(x)  ((((x) & 0xFF) << 24) | (((x) >> 8) & 0x00FFFFFF))
+static const JLONG dither_matrix4 = {
+  0x0008020A,
+  0x0C040E06,
+  0x030B0109,
+  0x0F070D05
+};
+
+
+/* Include inline routines for RGB565 conversion */
+
+#define PACK_SHORT_565  PACK_SHORT_565_LE
+#define PACK_TWO_PIXELS  PACK_TWO_PIXELS_LE
+#define WRITE_TWO_PIXELS  WRITE_TWO_PIXELS_LE
+#define h2v1_merged_upsample_565_internal  h2v1_merged_upsample_565_le
+#define h2v1_merged_upsample_565D_internal  h2v1_merged_upsample_565D_le
+#define h2v2_merged_upsample_565_internal  h2v2_merged_upsample_565_le
+#define h2v2_merged_upsample_565D_internal  h2v2_merged_upsample_565D_le
+#include "jdmrg565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef WRITE_TWO_PIXELS
+#undef h2v1_merged_upsample_565_internal
+#undef h2v1_merged_upsample_565D_internal
+#undef h2v2_merged_upsample_565_internal
+#undef h2v2_merged_upsample_565D_internal
+
+#define PACK_SHORT_565  PACK_SHORT_565_BE
+#define PACK_TWO_PIXELS  PACK_TWO_PIXELS_BE
+#define WRITE_TWO_PIXELS  WRITE_TWO_PIXELS_BE
+#define h2v1_merged_upsample_565_internal  h2v1_merged_upsample_565_be
+#define h2v1_merged_upsample_565D_internal  h2v1_merged_upsample_565D_be
+#define h2v2_merged_upsample_565_internal  h2v2_merged_upsample_565_be
+#define h2v2_merged_upsample_565D_internal  h2v2_merged_upsample_565D_be
+#include "jdmrg565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef WRITE_TWO_PIXELS
+#undef h2v1_merged_upsample_565_internal
+#undef h2v1_merged_upsample_565D_internal
+#undef h2v2_merged_upsample_565_internal
+#undef h2v2_merged_upsample_565D_internal
+
+
+static INLINE boolean is_big_endian(void)
+{
+  int test_value = 1;
+  if (*(char *)&test_value != 1)
+    return TRUE;
+  return FALSE;
+}
+
+
+METHODDEF(void)
+h2v1_merged_upsample_565(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v1_merged_upsample_565_be(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+  else
+    h2v1_merged_upsample_565_le(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+}
+
+
+METHODDEF(void)
+h2v1_merged_upsample_565D(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                          JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v1_merged_upsample_565D_be(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+  else
+    h2v1_merged_upsample_565D_le(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+}
+
+
+METHODDEF(void)
+h2v2_merged_upsample_565(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v2_merged_upsample_565_be(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+  else
+    h2v2_merged_upsample_565_le(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+}
+
+
+METHODDEF(void)
+h2v2_merged_upsample_565D(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                          JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v2_merged_upsample_565D_be(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+  else
+    h2v2_merged_upsample_565D_le(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+}
+
+
+/*
+ * Module initialization routine for merged upsampling/color conversion.
+ *
+ * NB: this is called under the conditions determined by use_merged_upsample()
+ * in jdmaster.c.  That routine MUST correspond to the actual capabilities
+ * of this module; no safety checks are made here.
+ */
+
+GLOBAL(void)
+jinit_merged_upsampler(j_decompress_ptr cinfo)
+{
+  my_merged_upsample_ptr upsample;
+
+  upsample = (my_merged_upsample_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_merged_upsampler));
+  cinfo->upsample = (struct jpeg_upsampler *)upsample;
+  upsample->pub.start_pass = start_pass_merged_upsample;
+  upsample->pub.need_context_rows = FALSE;
+
+  upsample->out_row_width = cinfo->output_width * cinfo->out_color_components;
+
+  if (cinfo->max_v_samp_factor == 2) {
+    upsample->pub.upsample = merged_2v_upsample;
+    if (jsimd_can_h2v2_merged_upsample())
+      upsample->upmethod = jsimd_h2v2_merged_upsample;
+    else
+      upsample->upmethod = h2v2_merged_upsample;
+    if (cinfo->out_color_space == JCS_RGB565) {
+      if (cinfo->dither_mode != JDITHER_NONE) {
+        upsample->upmethod = h2v2_merged_upsample_565D;
+      } else {
+        upsample->upmethod = h2v2_merged_upsample_565;
+      }
+    }
+    /* Allocate a spare row buffer */
+    upsample->spare_row = (JSAMPROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                (size_t)(upsample->out_row_width * sizeof(JSAMPLE)));
+  } else {
+    upsample->pub.upsample = merged_1v_upsample;
+    if (jsimd_can_h2v1_merged_upsample())
+      upsample->upmethod = jsimd_h2v1_merged_upsample;
+    else
+      upsample->upmethod = h2v1_merged_upsample;
+    if (cinfo->out_color_space == JCS_RGB565) {
+      if (cinfo->dither_mode != JDITHER_NONE) {
+        upsample->upmethod = h2v1_merged_upsample_565D;
+      } else {
+        upsample->upmethod = h2v1_merged_upsample_565;
+      }
+    }
+    /* No spare row needed */
+    upsample->spare_row = NULL;
+  }
+
+  build_ycc_rgb_table(cinfo);
+}
+
+#endif /* UPSAMPLE_MERGING_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdmerge.h Added

@@ -0,0 +1,47 @@
+/*
+ * jdmerge.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_upsampler pub;    /* public fields */
+
+  /* Pointer to routine to do actual upsampling/conversion of one row group */
+  void (*upmethod) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+  /* Private state for YCC->RGB conversion */
+  int *Cr_r_tab;                /* => table for Cr to R conversion */
+  int *Cb_b_tab;                /* => table for Cb to B conversion */
+  JLONG *Cr_g_tab;              /* => table for Cr to G conversion */
+  JLONG *Cb_g_tab;              /* => table for Cb to G conversion */
+
+  /* For 2:1 vertical sampling, we produce two output rows at a time.
+   * We need a "spare" row buffer to hold the second output row if the
+   * application provides just a one-row buffer; we also use the spare
+   * to discard the dummy last row if the image height is odd.
+   */
+  JSAMPROW spare_row;
+  boolean spare_full;           /* T if spare buffer is occupied */
+
+  JDIMENSION out_row_width;     /* samples per output row */
+  JDIMENSION rows_to_go;        /* counts rows remaining in image */
+} my_merged_upsampler;
+
+typedef my_merged_upsampler *my_merged_upsample_ptr;
+
+#endif /* UPSAMPLE_MERGING_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdmrg565.c Added

@@ -0,0 +1,354 @@
+/*
+ * jdmrg565.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2014-2015, 2018, 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ */
+
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                                  JDIMENSION in_row_group_ctr,
+                                  JSAMPARRAY output_buf)
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int *Crrtab = upsample->Cr_r_tab;
+  int *Cbbtab = upsample->Cb_b_tab;
+  JLONG *Crgtab = upsample->Cr_g_tab;
+  JLONG *Cbgtab = upsample->Cb_g_tab;
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf0in_row_group_ctr;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr = output_buf0;
+
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = *inptr1++;
+    cr = *inptr2++;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = *inptr0++;
+    r = range_limity + cred;
+    g = range_limity + cgreen;
+    b = range_limity + cblue;
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = *inptr0++;
+    r = range_limity + cred;
+    g = range_limity + cgreen;
+    b = range_limity + cblue;
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr, rgb);
+    outptr += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = *inptr1;
+    cr = *inptr2;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+    y  = *inptr0;
+    r = range_limity + cred;
+    g = range_limity + cgreen;
+    b = range_limity + cblue;
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16 *)outptr = (INT16)rgb;
+  }
+}
+
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_565D_internal(j_decompress_ptr cinfo,
+                                   JSAMPIMAGE input_buf,
+                                   JDIMENSION in_row_group_ctr,
+                                   JSAMPARRAY output_buf)
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int *Crrtab = upsample->Cr_r_tab;
+  int *Cbbtab = upsample->Cb_b_tab;
+  JLONG *Crgtab = upsample->Cr_g_tab;
+  JLONG *Cbgtab = upsample->Cb_g_tab;
+  JLONG d0 = dither_matrixcinfo->output_scanline & DITHER_MASK;
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf0in_row_group_ctr;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr = output_buf0;
+
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = *inptr1++;
+    cr = *inptr2++;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = *inptr0++;
+    r = range_limitDITHER_565_R(y + cred, d0);
+    g = range_limitDITHER_565_G(y + cgreen, d0);
+    b = range_limitDITHER_565_B(y + cblue, d0);
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = *inptr0++;
+    r = range_limitDITHER_565_R(y + cred, d0);
+    g = range_limitDITHER_565_G(y + cgreen, d0);
+    b = range_limitDITHER_565_B(y + cblue, d0);
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr, rgb);
+    outptr += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = *inptr1;
+    cr = *inptr2;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+    y  = *inptr0;
+    r = range_limitDITHER_565_R(y + cred, d0);
+    g = range_limitDITHER_565_G(y + cgreen, d0);
+    b = range_limitDITHER_565_B(y + cblue, d0);
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16 *)outptr = (INT16)rgb;
+  }
+}
+
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                                  JDIMENSION in_row_group_ctr,
+                                  JSAMPARRAY output_buf)
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int *Crrtab = upsample->Cr_r_tab;
+  int *Cbbtab = upsample->Cb_b_tab;
+  JLONG *Crgtab = upsample->Cr_g_tab;
+  JLONG *Cbgtab = upsample->Cb_g_tab;
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf0in_row_group_ctr * 2;
+  inptr01 = input_buf0in_row_group_ctr * 2 + 1;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr0 = output_buf0;
+  outptr1 = output_buf1;
+
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = *inptr1++;
+    cr = *inptr2++;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = *inptr00++;
+    r = range_limity + cred;
+    g = range_limity + cgreen;
+    b = range_limity + cblue;
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = *inptr00++;
+    r = range_limity + cred;
+    g = range_limity + cgreen;
+    b = range_limity + cblue;
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr0, rgb);
+    outptr0 += 4;
+
+    y  = *inptr01++;
+    r = range_limity + cred;
+    g = range_limity + cgreen;
+    b = range_limity + cblue;
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = *inptr01++;
+    r = range_limity + cred;
+    g = range_limity + cgreen;
+    b = range_limity + cblue;
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr1, rgb);
+    outptr1 += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = *inptr1;
+    cr = *inptr2;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+
+    y  = *inptr00;
+    r = range_limity + cred;
+    g = range_limity + cgreen;
+    b = range_limity + cblue;
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16 *)outptr0 = (INT16)rgb;
+
+    y  = *inptr01;
+    r = range_limity + cred;
+    g = range_limity + cgreen;
+    b = range_limity + cblue;
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16 *)outptr1 = (INT16)rgb;
+  }
+}
+
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_565D_internal(j_decompress_ptr cinfo,
+                                   JSAMPIMAGE input_buf,
+                                   JDIMENSION in_row_group_ctr,
+                                   JSAMPARRAY output_buf)
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int *Crrtab = upsample->Cr_r_tab;
+  int *Cbbtab = upsample->Cb_b_tab;
+  JLONG *Crgtab = upsample->Cr_g_tab;
+  JLONG *Cbgtab = upsample->Cb_g_tab;
+  JLONG d0 = dither_matrixcinfo->output_scanline & DITHER_MASK;
+  JLONG d1 = dither_matrix(cinfo->output_scanline + 1) & DITHER_MASK;
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf0in_row_group_ctr * 2;
+  inptr01 = input_buf0in_row_group_ctr * 2 + 1;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr0 = output_buf0;
+  outptr1 = output_buf1;
+
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = *inptr1++;
+    cr = *inptr2++;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = *inptr00++;
+    r = range_limitDITHER_565_R(y + cred, d0);
+    g = range_limitDITHER_565_G(y + cgreen, d0);
+    b = range_limitDITHER_565_B(y + cblue, d0);
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = *inptr00++;
+    r = range_limitDITHER_565_R(y + cred, d0);
+    g = range_limitDITHER_565_G(y + cgreen, d0);
+    b = range_limitDITHER_565_B(y + cblue, d0);
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr0, rgb);
+    outptr0 += 4;
+
+    y  = *inptr01++;
+    r = range_limitDITHER_565_R(y + cred, d1);
+    g = range_limitDITHER_565_G(y + cgreen, d1);
+    b = range_limitDITHER_565_B(y + cblue, d1);
+    d1 = DITHER_ROTATE(d1);
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = *inptr01++;
+    r = range_limitDITHER_565_R(y + cred, d1);
+    g = range_limitDITHER_565_G(y + cgreen, d1);
+    b = range_limitDITHER_565_B(y + cblue, d1);
+    d1 = DITHER_ROTATE(d1);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr1, rgb);
+    outptr1 += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = *inptr1;
+    cr = *inptr2;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+
+    y  = *inptr00;
+    r = range_limitDITHER_565_R(y + cred, d0);
+    g = range_limitDITHER_565_G(y + cgreen, d0);
+    b = range_limitDITHER_565_B(y + cblue, d0);
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16 *)outptr0 = (INT16)rgb;
+
+    y  = *inptr01;
+    r = range_limitDITHER_565_R(y + cred, d1);
+    g = range_limitDITHER_565_G(y + cgreen, d1);
+    b = range_limitDITHER_565_B(y + cblue, d1);
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16 *)outptr1 = (INT16)rgb;
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdmrgext.c Added

@@ -0,0 +1,184 @@
+/*
+ * jdmrgext.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, 2015, 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ */
+
+
+/* This file is included by jdmerge.c */
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ */
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                              JDIMENSION in_row_group_ctr,
+                              JSAMPARRAY output_buf)
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int *Crrtab = upsample->Cr_r_tab;
+  int *Cbbtab = upsample->Cb_b_tab;
+  JLONG *Crgtab = upsample->Cr_g_tab;
+  JLONG *Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf0in_row_group_ctr;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr = output_buf0;
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = *inptr1++;
+    cr = *inptr2++;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = *inptr0++;
+    outptrRGB_RED =   range_limity + cred;
+    outptrRGB_GREEN = range_limity + cgreen;
+    outptrRGB_BLUE =  range_limity + cblue;
+#ifdef RGB_ALPHA
+    outptrRGB_ALPHA = 0xFF;
+#endif
+    outptr += RGB_PIXELSIZE;
+    y  = *inptr0++;
+    outptrRGB_RED =   range_limity + cred;
+    outptrRGB_GREEN = range_limity + cgreen;
+    outptrRGB_BLUE =  range_limity + cblue;
+#ifdef RGB_ALPHA
+    outptrRGB_ALPHA = 0xFF;
+#endif
+    outptr += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = *inptr1;
+    cr = *inptr2;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+    y  = *inptr0;
+    outptrRGB_RED =   range_limity + cred;
+    outptrRGB_GREEN = range_limity + cgreen;
+    outptrRGB_BLUE =  range_limity + cblue;
+#ifdef RGB_ALPHA
+    outptrRGB_ALPHA = 0xFF;
+#endif
+  }
+}
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+ */
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                              JDIMENSION in_row_group_ctr,
+                              JSAMPARRAY output_buf)
+{
+  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int *Crrtab = upsample->Cr_r_tab;
+  int *Cbbtab = upsample->Cb_b_tab;
+  JLONG *Crgtab = upsample->Cr_g_tab;
+  JLONG *Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf0in_row_group_ctr * 2;
+  inptr01 = input_buf0in_row_group_ctr * 2 + 1;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr0 = output_buf0;
+  outptr1 = output_buf1;
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = *inptr1++;
+    cr = *inptr2++;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = *inptr00++;
+    outptr0RGB_RED =   range_limity + cred;
+    outptr0RGB_GREEN = range_limity + cgreen;
+    outptr0RGB_BLUE =  range_limity + cblue;
+#ifdef RGB_ALPHA
+    outptr0RGB_ALPHA = 0xFF;
+#endif
+    outptr0 += RGB_PIXELSIZE;
+    y  = *inptr00++;
+    outptr0RGB_RED =   range_limity + cred;
+    outptr0RGB_GREEN = range_limity + cgreen;
+    outptr0RGB_BLUE =  range_limity + cblue;
+#ifdef RGB_ALPHA
+    outptr0RGB_ALPHA = 0xFF;
+#endif
+    outptr0 += RGB_PIXELSIZE;
+    y  = *inptr01++;
+    outptr1RGB_RED =   range_limity + cred;
+    outptr1RGB_GREEN = range_limity + cgreen;
+    outptr1RGB_BLUE =  range_limity + cblue;
+#ifdef RGB_ALPHA
+    outptr1RGB_ALPHA = 0xFF;
+#endif
+    outptr1 += RGB_PIXELSIZE;
+    y  = *inptr01++;
+    outptr1RGB_RED =   range_limity + cred;
+    outptr1RGB_GREEN = range_limity + cgreen;
+    outptr1RGB_BLUE =  range_limity + cblue;
+#ifdef RGB_ALPHA
+    outptr1RGB_ALPHA = 0xFF;
+#endif
+    outptr1 += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = *inptr1;
+    cr = *inptr2;
+    cred = Crrtabcr;
+    cgreen = (int)RIGHT_SHIFT(Cbgtabcb + Crgtabcr, SCALEBITS);
+    cblue = Cbbtabcb;
+    y  = *inptr00;
+    outptr0RGB_RED =   range_limity + cred;
+    outptr0RGB_GREEN = range_limity + cgreen;
+    outptr0RGB_BLUE =  range_limity + cblue;
+#ifdef RGB_ALPHA
+    outptr0RGB_ALPHA = 0xFF;
+#endif
+    y  = *inptr01;
+    outptr1RGB_RED =   range_limity + cred;
+    outptr1RGB_GREEN = range_limity + cgreen;
+    outptr1RGB_BLUE =  range_limity + cblue;
+#ifdef RGB_ALPHA
+    outptr1RGB_ALPHA = 0xFF;
+#endif
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdphuff.c Added

@@ -0,0 +1,679 @@
+/*
+ * jdphuff.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015-2016, 2018-2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains Huffman entropy decoding routines for progressive JPEG.
+ *
+ * Much of the complexity here has to do with supporting input suspension.
+ * If the data source module demands suspension, we want to be able to back
+ * up to the start of the current MCU.  To do this, we copy state variables
+ * into local working storage, and update them back to the permanent
+ * storage only upon successful completion of an MCU.
+ *
+ * NOTE: All referenced figures are from
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdhuff.h"             /* Declarations shared with jdhuff.c */
+#include <limits.h>
+
+
+#ifdef D_PROGRESSIVE_SUPPORTED
+
+/*
+ * Expanded entropy decoder object for progressive Huffman decoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+typedef struct {
+  unsigned int EOBRUN;                  /* remaining EOBs in EOBRUN */
+  int last_dc_valMAX_COMPS_IN_SCAN;   /* last DC coef for each component */
+} savable_state;
+
+typedef struct {
+  struct jpeg_entropy_decoder pub; /* public fields */
+
+  /* These fields are loaded into local variables at start of each MCU.
+   * In case of suspension, we exit WITHOUT updating them.
+   */
+  bitread_perm_state bitstate;  /* Bit buffer at start of MCU */
+  savable_state saved;          /* Other state at start of MCU */
+
+  /* These fields are NOT loaded into local working state. */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+
+  /* Pointers to derived tables (these workspaces have image lifespan) */
+  d_derived_tbl *derived_tblsNUM_HUFF_TBLS;
+
+  d_derived_tbl *ac_derived_tbl; /* active table during an AC scan */
+} phuff_entropy_decoder;
+
+typedef phuff_entropy_decoder *phuff_entropy_ptr;
+
+/* Forward declarations */
+METHODDEF(boolean) decode_mcu_DC_first(j_decompress_ptr cinfo,
+                                       JBLOCKROW *MCU_data);
+METHODDEF(boolean) decode_mcu_AC_first(j_decompress_ptr cinfo,
+                                       JBLOCKROW *MCU_data);
+METHODDEF(boolean) decode_mcu_DC_refine(j_decompress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(boolean) decode_mcu_AC_refine(j_decompress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+
+
+/*
+ * Initialize for a Huffman-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass_phuff_decoder(j_decompress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  boolean is_DC_band, bad;
+  int ci, coefi, tbl;
+  d_derived_tbl **pdtbl;
+  int *coef_bit_ptr, *prev_coef_bit_ptr;
+  jpeg_component_info *compptr;
+
+  is_DC_band = (cinfo->Ss == 0);
+
+  /* Validate scan parameters */
+  bad = FALSE;
+  if (is_DC_band) {
+    if (cinfo->Se != 0)
+      bad = TRUE;
+  } else {
+    /* need not check Ss/Se < 0 since they came from unsigned bytes */
+    if (cinfo->Ss > cinfo->Se || cinfo->Se >= DCTSIZE2)
+      bad = TRUE;
+    /* AC scans may have only one component */
+    if (cinfo->comps_in_scan != 1)
+      bad = TRUE;
+  }
+  if (cinfo->Ah != 0) {
+    /* Successive approximation refinement scan: must have Al = Ah-1. */
+    if (cinfo->Al != cinfo->Ah - 1)
+      bad = TRUE;
+  }
+  if (cinfo->Al > 13)           /* need not check for < 0 */
+    bad = TRUE;
+  /* Arguably the maximum Al value should be less than 13 for 8-bit precision,
+   * but the spec doesn't say so, and we try to be liberal about what we
+   * accept.  Note: large Al values could result in out-of-range DC
+   * coefficients during early scans, leading to bizarre displays due to
+   * overflows in the IDCT math.  But we won't crash.
+   */
+  if (bad)
+    ERREXIT4(cinfo, JERR_BAD_PROGRESSION,
+             cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
+  /* Update progression status, and verify that scan order is legal.
+   * Note that inter-scan inconsistencies are treated as warnings
+   * not fatal errors ... not clear if this is right way to behave.
+   */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    int cindex = cinfo->cur_comp_infoci->component_index;
+    coef_bit_ptr = &cinfo->coef_bitscindex0;
+    prev_coef_bit_ptr = &cinfo->coef_bitscindex + cinfo->num_components0;
+    if (!is_DC_band && coef_bit_ptr0 < 0) /* AC without prior DC scan */
+      WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
+    for (coefi = MIN(cinfo->Ss, 1); coefi <= MAX(cinfo->Se, 9); coefi++) {
+      if (cinfo->input_scan_number > 1)
+        prev_coef_bit_ptrcoefi = coef_bit_ptrcoefi;
+      else
+        prev_coef_bit_ptrcoefi = 0;
+    }
+    for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
+      int expected = (coef_bit_ptrcoefi < 0) ? 0 : coef_bit_ptrcoefi;
+      if (cinfo->Ah != expected)
+        WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
+      coef_bit_ptrcoefi = cinfo->Al;
+    }
+  }
+
+  /* Select MCU decoding routine */
+  if (cinfo->Ah == 0) {
+    if (is_DC_band)
+      entropy->pub.decode_mcu = decode_mcu_DC_first;
+    else
+      entropy->pub.decode_mcu = decode_mcu_AC_first;
+  } else {
+    if (is_DC_band)
+      entropy->pub.decode_mcu = decode_mcu_DC_refine;
+    else
+      entropy->pub.decode_mcu = decode_mcu_AC_refine;
+  }
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_infoci;
+    /* Make sure requested tables are present, and compute derived tables.
+     * We may build same derived table more than once, but it's not expensive.
+     */
+    if (is_DC_band) {
+      if (cinfo->Ah == 0) {     /* DC refinement needs no table */
+        tbl = compptr->dc_tbl_no;
+        pdtbl = (d_derived_tbl **)(entropy->derived_tbls) + tbl;
+        jpeg_make_d_derived_tbl(cinfo, TRUE, tbl, pdtbl);
+      }
+    } else {
+      tbl = compptr->ac_tbl_no;
+      pdtbl = (d_derived_tbl **)(entropy->derived_tbls) + tbl;
+      jpeg_make_d_derived_tbl(cinfo, FALSE, tbl, pdtbl);
+      /* remember the single active table */
+      entropy->ac_derived_tbl = entropy->derived_tblstbl;
+    }
+    /* Initialize DC predictions to 0 */
+    entropy->saved.last_dc_valci = 0;
+  }
+
+  /* Initialize bitread state variables */
+  entropy->bitstate.bits_left = 0;
+  entropy->bitstate.get_buffer = 0; /* unnecessary, but keeps Purify quiet */
+  entropy->pub.insufficient_data = FALSE;
+
+  /* Initialize private state variables */
+  entropy->saved.EOBRUN = 0;
+
+  /* Initialize restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Figure F.12: extend sign bit.
+ * On some machines, a shift and add will be faster than a table lookup.
+ */
+
+#define AVOID_TABLES
+#ifdef AVOID_TABLES
+
+#define NEG_1  ((unsigned)-1)
+#define HUFF_EXTEND(x, s) \
+  ((x) < (1 << ((s) - 1)) ? (x) + (((NEG_1) << (s)) + 1) : (x))
+
+#else
+
+#define HUFF_EXTEND(x, s) \
+  ((x) < extend_tests ? (x) + extend_offsets : (x))
+
+static const int extend_test16 = {   /* entry n is 2**(n-1) */
+  0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080,
+  0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000
+};
+
+static const int extend_offset16 = { /* entry n is (-1 << n) + 1 */
+  0, ((-1) << 1) + 1, ((-1) << 2) + 1, ((-1) << 3) + 1, ((-1) << 4) + 1,
+  ((-1) << 5) + 1, ((-1) << 6) + 1, ((-1) << 7) + 1, ((-1) << 8) + 1,
+  ((-1) << 9) + 1, ((-1) << 10) + 1, ((-1) << 11) + 1, ((-1) << 12) + 1,
+  ((-1) << 13) + 1, ((-1) << 14) + 1, ((-1) << 15) + 1
+};
+
+#endif /* AVOID_TABLES */
+
+
+/*
+ * Check for a restart marker & resynchronize decoder.
+ * Returns FALSE if must suspend.
+ */
+
+LOCAL(boolean)
+process_restart(j_decompress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  int ci;
+
+  /* Throw away any unused bits remaining in bit buffer; */
+  /* include any full bytes in next_marker's count of discarded bytes */
+  cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8;
+  entropy->bitstate.bits_left = 0;
+
+  /* Advance past the RSTn marker */
+  if (!(*cinfo->marker->read_restart_marker) (cinfo))
+    return FALSE;
+
+  /* Re-initialize DC predictions to 0 */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+    entropy->saved.last_dc_valci = 0;
+  /* Re-init EOB run count, too */
+  entropy->saved.EOBRUN = 0;
+
+  /* Reset restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+
+  /* Reset out-of-data flag, unless read_restart_marker left us smack up
+   * against a marker.  In that case we will end up treating the next data
+   * segment as empty, and we can avoid producing bogus output pixels by
+   * leaving the flag set.
+   */
+  if (cinfo->unread_marker == 0)
+    entropy->pub.insufficient_data = FALSE;
+
+  return TRUE;
+}
+
+
+/*
+ * Huffman MCU decoding.
+ * Each of these routines decodes and returns one MCU's worth of
+ * Huffman-compressed coefficients.
+ * The coefficients are reordered from zigzag order into natural array order,
+ * but are not dequantized.
+ *
+ * The i'th block of the MCU is stored into the block pointed to by
+ * MCU_datai.  WE ASSUME THIS AREA IS INITIALLY ZEROED BY THE CALLER.
+ *
+ * We return FALSE if data source requested suspension.  In that case no
+ * changes have been made to permanent state.  (Exception: some output
+ * coefficients may already have been assigned.  This is harmless for
+ * spectral selection, since we'll just re-assign them on the next call.
+ * Successive approximation AC refinement has to be more careful, however.)
+ */
+
+/*
+ * MCU decoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  int Al = cinfo->Al;
+  register int s, r;
+  int blkn, ci;
+  JBLOCKROW block;
+  BITREAD_STATE_VARS;
+  savable_state state;
+  d_derived_tbl *tbl;
+  jpeg_component_info *compptr;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (!process_restart(cinfo))
+        return FALSE;
+  }
+
+  /* If we've run out of data, just leave the MCU set to zeroes.
+   * This way, we return uniform gray for the remainder of the segment.
+   */
+  if (!entropy->pub.insufficient_data) {
+
+    /* Load up working state */
+    BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
+    state = entropy->saved;
+
+    /* Outer loop handles each block in the MCU */
+
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      block = MCU_datablkn;
+      ci = cinfo->MCU_membershipblkn;
+      compptr = cinfo->cur_comp_infoci;
+      tbl = entropy->derived_tblscompptr->dc_tbl_no;
+
+      /* Decode a single block's worth of coefficients */
+
+      /* Section F.2.2.1: decode the DC coefficient difference */
+      HUFF_DECODE(s, br_state, tbl, return FALSE, label1);
+      if (s) {
+        CHECK_BIT_BUFFER(br_state, s, return FALSE);
+        r = GET_BITS(s);
+        s = HUFF_EXTEND(r, s);
+      }
+
+      /* Convert DC difference to actual value, update last_dc_val */
+      if ((state.last_dc_valci >= 0 &&
+           s > INT_MAX - state.last_dc_valci) ||
+          (state.last_dc_valci < 0 && s < INT_MIN - state.last_dc_valci))
+        ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+      s += state.last_dc_valci;
+      state.last_dc_valci = s;
+      /* Scale and output the coefficient (assumes jpeg_natural_order0=0) */
+      (*block)0 = (JCOEF)LEFT_SHIFT(s, Al);
+    }
+
+    /* Completed MCU, so update state */
+    BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
+    entropy->saved = state;
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  int Se = cinfo->Se;
+  int Al = cinfo->Al;
+  register int s, k, r;
+  unsigned int EOBRUN;
+  JBLOCKROW block;
+  BITREAD_STATE_VARS;
+  d_derived_tbl *tbl;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (!process_restart(cinfo))
+        return FALSE;
+  }
+
+  /* If we've run out of data, just leave the MCU set to zeroes.
+   * This way, we return uniform gray for the remainder of the segment.
+   */
+  if (!entropy->pub.insufficient_data) {
+
+    /* Load up working state.
+     * We can avoid loading/saving bitread state if in an EOB run.
+     */
+    EOBRUN = entropy->saved.EOBRUN;     /* only part of saved state we need */
+
+    /* There is always only one block per MCU */
+
+    if (EOBRUN > 0)             /* if it's a band of zeroes... */
+      EOBRUN--;                 /* ...process it now (we do nothing) */
+    else {
+      BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
+      block = MCU_data0;
+      tbl = entropy->ac_derived_tbl;
+
+      for (k = cinfo->Ss; k <= Se; k++) {
+        HUFF_DECODE(s, br_state, tbl, return FALSE, label2);
+        r = s >> 4;
+        s &= 15;
+        if (s) {
+          k += r;
+          CHECK_BIT_BUFFER(br_state, s, return FALSE);
+          r = GET_BITS(s);
+          s = HUFF_EXTEND(r, s);
+          /* Scale and output coefficient in natural (dezigzagged) order */
+          (*block)jpeg_natural_orderk = (JCOEF)LEFT_SHIFT(s, Al);
+        } else {
+          if (r == 15) {        /* ZRL */
+            k += 15;            /* skip 15 zeroes in band */
+          } else {              /* EOBr, run length is 2^r + appended bits */
+            EOBRUN = 1 << r;
+            if (r) {            /* EOBr, r > 0 */
+              CHECK_BIT_BUFFER(br_state, r, return FALSE);
+              r = GET_BITS(r);
+              EOBRUN += r;
+            }
+            EOBRUN--;           /* this band is processed at this moment */
+            break;              /* force end-of-band */
+          }
+        }
+      }
+
+      BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
+    }
+
+    /* Completed MCU, so update state */
+    entropy->saved.EOBRUN = EOBRUN;     /* only part of saved state we need */
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for DC successive approximation refinement scan.
+ * Note: we assume such scans can be multi-component, although the spec
+ * is not very clear on the point.
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_refine(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  int p1 = 1 << cinfo->Al;      /* 1 in the bit position being coded */
+  int blkn;
+  JBLOCKROW block;
+  BITREAD_STATE_VARS;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (!process_restart(cinfo))
+        return FALSE;
+  }
+
+  /* Not worth the cycles to check insufficient_data here,
+   * since we will not change the data anyway if we read zeroes.
+   */
+
+  /* Load up working state */
+  BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
+
+  /* Outer loop handles each block in the MCU */
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_datablkn;
+
+    /* Encoded data is simply the next bit of the two's-complement DC value */
+    CHECK_BIT_BUFFER(br_state, 1, return FALSE);
+    if (GET_BITS(1))
+      (*block)0 |= p1;
+    /* Note: since we use |=, repeating the assignment later is safe */
+  }
+
+  /* Completed MCU, so update state */
+  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
+
+  /* Account for restart interval (no-op if not using restarts) */
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_refine(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
+  int Se = cinfo->Se;
+  int p1 = 1 << cinfo->Al;        /* 1 in the bit position being coded */
+  int m1 = (NEG_1) << cinfo->Al;  /* -1 in the bit position being coded */
+  register int s, k, r;
+  unsigned int EOBRUN;
+  JBLOCKROW block;
+  JCOEFPTR thiscoef;
+  BITREAD_STATE_VARS;
+  d_derived_tbl *tbl;
+  int num_newnz;
+  int newnz_posDCTSIZE2;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (!process_restart(cinfo))
+        return FALSE;
+  }
+
+  /* If we've run out of data, don't modify the MCU.
+   */
+  if (!entropy->pub.insufficient_data) {
+
+    /* Load up working state */
+    BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
+    EOBRUN = entropy->saved.EOBRUN; /* only part of saved state we need */
+
+    /* There is always only one block per MCU */
+    block = MCU_data0;
+    tbl = entropy->ac_derived_tbl;
+
+    /* If we are forced to suspend, we must undo the assignments to any newly
+     * nonzero coefficients in the block, because otherwise we'd get confused
+     * next time about which coefficients were already nonzero.
+     * But we need not undo addition of bits to already-nonzero coefficients;
+     * instead, we can test the current bit to see if we already did it.
+     */
+    num_newnz = 0;
+
+    /* initialize coefficient loop counter to start of band */
+    k = cinfo->Ss;
+
+    if (EOBRUN == 0) {
+      for (; k <= Se; k++) {
+        HUFF_DECODE(s, br_state, tbl, goto undoit, label3);
+        r = s >> 4;
+        s &= 15;
+        if (s) {
+          if (s != 1)           /* size of new coef should always be 1 */
+            WARNMS(cinfo, JWRN_HUFF_BAD_CODE);
+          CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+          if (GET_BITS(1))
+            s = p1;             /* newly nonzero coef is positive */
+          else
+            s = m1;             /* newly nonzero coef is negative */
+        } else {
+          if (r != 15) {
+            EOBRUN = 1 << r;    /* EOBr, run length is 2^r + appended bits */
+            if (r) {
+              CHECK_BIT_BUFFER(br_state, r, goto undoit);
+              r = GET_BITS(r);
+              EOBRUN += r;
+            }
+            break;              /* rest of block is handled by EOB logic */
+          }
+          /* note s = 0 for processing ZRL */
+        }
+        /* Advance over already-nonzero coefs and r still-zero coefs,
+         * appending correction bits to the nonzeroes.  A correction bit is 1
+         * if the absolute value of the coefficient must be increased.
+         */
+        do {
+          thiscoef = *block + jpeg_natural_orderk;
+          if (*thiscoef != 0) {
+            CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+            if (GET_BITS(1)) {
+              if ((*thiscoef & p1) == 0) { /* do nothing if already set it */
+                if (*thiscoef >= 0)
+                  *thiscoef += p1;
+                else
+                  *thiscoef += m1;
+              }
+            }
+          } else {
+            if (--r < 0)
+              break;            /* reached target zero coefficient */
+          }
+          k++;
+        } while (k <= Se);
+        if (s) {
+          int pos = jpeg_natural_orderk;
+          /* Output newly nonzero coefficient */
+          (*block)pos = (JCOEF)s;
+          /* Remember its position in case we have to suspend */
+          newnz_posnum_newnz++ = pos;
+        }
+      }
+    }
+
+    if (EOBRUN > 0) {
+      /* Scan any remaining coefficient positions after the end-of-band
+       * (the last newly nonzero coefficient, if any).  Append a correction
+       * bit to each already-nonzero coefficient.  A correction bit is 1
+       * if the absolute value of the coefficient must be increased.
+       */
+      for (; k <= Se; k++) {
+        thiscoef = *block + jpeg_natural_orderk;
+        if (*thiscoef != 0) {
+          CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+          if (GET_BITS(1)) {
+            if ((*thiscoef & p1) == 0) { /* do nothing if already changed it */
+              if (*thiscoef >= 0)
+                *thiscoef += p1;
+              else
+                *thiscoef += m1;
+            }
+          }
+        }
+      }
+      /* Count one block completed in EOB run */
+      EOBRUN--;
+    }
+
+    /* Completed MCU, so update state */
+    BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
+    entropy->saved.EOBRUN = EOBRUN; /* only part of saved state we need */
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;
+
+  return TRUE;
+
+undoit:
+  /* Re-zero any output coefficients that we made newly nonzero */
+  while (num_newnz > 0)
+    (*block)newnz_pos--num_newnz = 0;
+
+  return FALSE;
+}
+
+
+/*
+ * Module initialization routine for progressive Huffman entropy decoding.
+ */
+
+GLOBAL(void)
+jinit_phuff_decoder(j_decompress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy;
+  int *coef_bit_ptr;
+  int ci, i;
+
+  entropy = (phuff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(phuff_entropy_decoder));
+  cinfo->entropy = (struct jpeg_entropy_decoder *)entropy;
+  entropy->pub.start_pass = start_pass_phuff_decoder;
+
+  /* Mark derived tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->derived_tblsi = NULL;
+  }
+
+  /* Create progression status table */
+  cinfo->coef_bits = (int (*)DCTSIZE2)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                cinfo->num_components * 2 * DCTSIZE2 *
+                                sizeof(int));
+  coef_bit_ptr = &cinfo->coef_bits00;
+  for (ci = 0; ci < cinfo->num_components; ci++)
+    for (i = 0; i < DCTSIZE2; i++)
+      *coef_bit_ptr++ = -1;
+}
+
+#endif /* D_PROGRESSIVE_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdpostct.c Added

@@ -0,0 +1,294 @@
+/*
+ * jdpostct.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the decompression postprocessing controller.
+ * This controller manages the upsampling, color conversion, and color
+ * quantization/reduction steps; specifically, it controls the buffering
+ * between upsample/color conversion and color quantization/reduction.
+ *
+ * If no color quantization/reduction is required, then this module has no
+ * work to do, and it just hands off to the upsample/color conversion code.
+ * An integrated upsample/convert/quantize process would replace this module
+ * entirely.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_post_controller pub; /* public fields */
+
+  /* Color quantization source buffer: this holds output data from
+   * the upsample/color conversion step to be passed to the quantizer.
+   * For two-pass color quantization, we need a full-image buffer;
+   * for one-pass operation, a strip buffer is sufficient.
+   */
+  jvirt_sarray_ptr whole_image; /* virtual array, or NULL if one-pass */
+  JSAMPARRAY buffer;            /* strip buffer, or current strip of virtual */
+  JDIMENSION strip_height;      /* buffer size in rows */
+  /* for two-pass mode only: */
+  JDIMENSION starting_row;      /* row # of first row in current strip */
+  JDIMENSION next_row;          /* index of next row to fill/empty in strip */
+} my_post_controller;
+
+typedef my_post_controller *my_post_ptr;
+
+
+/* Forward declarations */
+METHODDEF(void) post_process_1pass(j_decompress_ptr cinfo,
+                                   JSAMPIMAGE input_buf,
+                                   JDIMENSION *in_row_group_ctr,
+                                   JDIMENSION in_row_groups_avail,
+                                   JSAMPARRAY output_buf,
+                                   JDIMENSION *out_row_ctr,
+                                   JDIMENSION out_rows_avail);
+#ifdef QUANT_2PASS_SUPPORTED
+METHODDEF(void) post_process_prepass(j_decompress_ptr cinfo,
+                                     JSAMPIMAGE input_buf,
+                                     JDIMENSION *in_row_group_ctr,
+                                     JDIMENSION in_row_groups_avail,
+                                     JSAMPARRAY output_buf,
+                                     JDIMENSION *out_row_ctr,
+                                     JDIMENSION out_rows_avail);
+METHODDEF(void) post_process_2pass(j_decompress_ptr cinfo,
+                                   JSAMPIMAGE input_buf,
+                                   JDIMENSION *in_row_group_ctr,
+                                   JDIMENSION in_row_groups_avail,
+                                   JSAMPARRAY output_buf,
+                                   JDIMENSION *out_row_ctr,
+                                   JDIMENSION out_rows_avail);
+#endif
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_dpost(j_decompress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_post_ptr post = (my_post_ptr)cinfo->post;
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+    if (cinfo->quantize_colors) {
+      /* Single-pass processing with color quantization. */
+      post->pub.post_process_data = post_process_1pass;
+      /* We could be doing buffered-image output before starting a 2-pass
+       * color quantization; in that case, jinit_d_post_controller did not
+       * allocate a strip buffer.  Use the virtual-array buffer as workspace.
+       */
+      if (post->buffer == NULL) {
+        post->buffer = (*cinfo->mem->access_virt_sarray)
+          ((j_common_ptr)cinfo, post->whole_image,
+           (JDIMENSION)0, post->strip_height, TRUE);
+      }
+    } else {
+      /* For single-pass processing without color quantization,
+       * I have no work to do; just call the upsampler directly.
+       */
+      post->pub.post_process_data = cinfo->upsample->upsample;
+    }
+    break;
+#ifdef QUANT_2PASS_SUPPORTED
+  case JBUF_SAVE_AND_PASS:
+    /* First pass of 2-pass quantization */
+    if (post->whole_image == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    post->pub.post_process_data = post_process_prepass;
+    break;
+  case JBUF_CRANK_DEST:
+    /* Second pass of 2-pass quantization */
+    if (post->whole_image == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    post->pub.post_process_data = post_process_2pass;
+    break;
+#endif /* QUANT_2PASS_SUPPORTED */
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+  post->starting_row = post->next_row = 0;
+}
+
+
+/*
+ * Process some data in the one-pass (strip buffer) case.
+ * This is used for color precision reduction as well as one-pass quantization.
+ */
+
+METHODDEF(void)
+post_process_1pass(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                   JDIMENSION *in_row_group_ctr,
+                   JDIMENSION in_row_groups_avail, JSAMPARRAY output_buf,
+                   JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
+{
+  my_post_ptr post = (my_post_ptr)cinfo->post;
+  JDIMENSION num_rows, max_rows;
+
+  /* Fill the buffer, but not more than what we can dump out in one go. */
+  /* Note we rely on the upsampler to detect bottom of image. */
+  max_rows = out_rows_avail - *out_row_ctr;
+  if (max_rows > post->strip_height)
+    max_rows = post->strip_height;
+  num_rows = 0;
+  (*cinfo->upsample->upsample) (cinfo, input_buf, in_row_group_ctr,
+                                in_row_groups_avail, post->buffer, &num_rows,
+                                max_rows);
+  /* Quantize and emit data. */
+  (*cinfo->cquantize->color_quantize) (cinfo, post->buffer,
+                                       output_buf + *out_row_ctr,
+                                       (int)num_rows);
+  *out_row_ctr += num_rows;
+}
+
+
+#ifdef QUANT_2PASS_SUPPORTED
+
+/*
+ * Process some data in the first pass of 2-pass quantization.
+ */
+
+METHODDEF(void)
+post_process_prepass(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                     JDIMENSION *in_row_group_ctr,
+                     JDIMENSION in_row_groups_avail, JSAMPARRAY output_buf,
+                     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
+{
+  my_post_ptr post = (my_post_ptr)cinfo->post;
+  JDIMENSION old_next_row, num_rows;
+
+  /* Reposition virtual buffer if at start of strip. */
+  if (post->next_row == 0) {
+    post->buffer = (*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr)cinfo, post->whole_image,
+         post->starting_row, post->strip_height, TRUE);
+  }
+
+  /* Upsample some data (up to a strip height's worth). */
+  old_next_row = post->next_row;
+  (*cinfo->upsample->upsample) (cinfo, input_buf, in_row_group_ctr,
+                                in_row_groups_avail, post->buffer,
+                                &post->next_row, post->strip_height);
+
+  /* Allow quantizer to scan new data.  No data is emitted, */
+  /* but we advance out_row_ctr so outer loop can tell when we're done. */
+  if (post->next_row > old_next_row) {
+    num_rows = post->next_row - old_next_row;
+    (*cinfo->cquantize->color_quantize) (cinfo, post->buffer + old_next_row,
+                                         (JSAMPARRAY)NULL, (int)num_rows);
+    *out_row_ctr += num_rows;
+  }
+
+  /* Advance if we filled the strip. */
+  if (post->next_row >= post->strip_height) {
+    post->starting_row += post->strip_height;
+    post->next_row = 0;
+  }
+}
+
+
+/*
+ * Process some data in the second pass of 2-pass quantization.
+ */
+
+METHODDEF(void)
+post_process_2pass(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                   JDIMENSION *in_row_group_ctr,
+                   JDIMENSION in_row_groups_avail, JSAMPARRAY output_buf,
+                   JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
+{
+  my_post_ptr post = (my_post_ptr)cinfo->post;
+  JDIMENSION num_rows, max_rows;
+
+  /* Reposition virtual buffer if at start of strip. */
+  if (post->next_row == 0) {
+    post->buffer = (*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr)cinfo, post->whole_image,
+         post->starting_row, post->strip_height, FALSE);
+  }
+
+  /* Determine number of rows to emit. */
+  num_rows = post->strip_height - post->next_row; /* available in strip */
+  max_rows = out_rows_avail - *out_row_ctr; /* available in output area */
+  if (num_rows > max_rows)
+    num_rows = max_rows;
+  /* We have to check bottom of image here, can't depend on upsampler. */
+  max_rows = cinfo->output_height - post->starting_row;
+  if (num_rows > max_rows)
+    num_rows = max_rows;
+
+  /* Quantize and emit data. */
+  (*cinfo->cquantize->color_quantize) (cinfo, post->buffer + post->next_row,
+                                       output_buf + *out_row_ctr,
+                                       (int)num_rows);
+  *out_row_ctr += num_rows;
+
+  /* Advance if we filled the strip. */
+  post->next_row += num_rows;
+  if (post->next_row >= post->strip_height) {
+    post->starting_row += post->strip_height;
+    post->next_row = 0;
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
+
+
+/*
+ * Initialize postprocessing controller.
+ */
+
+GLOBAL(void)
+jinit_d_post_controller(j_decompress_ptr cinfo, boolean need_full_buffer)
+{
+  my_post_ptr post;
+
+  post = (my_post_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_post_controller));
+  cinfo->post = (struct jpeg_d_post_controller *)post;
+  post->pub.start_pass = start_pass_dpost;
+  post->whole_image = NULL;     /* flag for no virtual arrays */
+  post->buffer = NULL;          /* flag for no strip buffer */
+
+  /* Create the quantization buffer, if needed */
+  if (cinfo->quantize_colors) {
+    /* The buffer strip height is max_v_samp_factor, which is typically
+     * an efficient number of rows for upsampling to return.
+     * (In the presence of output rescaling, we might want to be smarter?)
+     */
+    post->strip_height = (JDIMENSION)cinfo->max_v_samp_factor;
+    if (need_full_buffer) {
+      /* Two-pass color quantization: need full-image storage. */
+      /* We round up the number of rows to a multiple of the strip height. */
+#ifdef QUANT_2PASS_SUPPORTED
+      post->whole_image = (*cinfo->mem->request_virt_sarray)
+        ((j_common_ptr)cinfo, JPOOL_IMAGE, FALSE,
+         cinfo->output_width * cinfo->out_color_components,
+         (JDIMENSION)jround_up((long)cinfo->output_height,
+                               (long)post->strip_height),
+         post->strip_height);
+#else
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif /* QUANT_2PASS_SUPPORTED */
+    } else {
+      /* One-pass color quantization: just make a strip buffer. */
+      post->buffer = (*cinfo->mem->alloc_sarray)
+        ((j_common_ptr)cinfo, JPOOL_IMAGE,
+         cinfo->output_width * cinfo->out_color_components,
+         post->strip_height);
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdsample.c Added

@@ -0,0 +1,524 @@
+/*
+ * jdsample.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2010, 2015-2016, D. R. Commander.
+ * Copyright (C) 2014, MIPS Technologies, Inc., California.
+ * Copyright (C) 2015, Google, Inc.
+ * Copyright (C) 2019-2020, Arm Limited.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains upsampling routines.
+ *
+ * Upsampling input data is counted in "row groups".  A row group
+ * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size)
+ * sample rows of each component.  Upsampling will normally produce
+ * max_v_samp_factor pixel rows from each row group (but this could vary
+ * if the upsampler is applying a scale factor of its own).
+ *
+ * An excellent reference for image resampling is
+ *   Digital Image Warping, George Wolberg, 1990.
+ *   Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7.
+ */
+
+#include "jinclude.h"
+#include "jdsample.h"
+#include "jsimd.h"
+#include "jpegcomp.h"
+
+
+
+/*
+ * Initialize for an upsampling pass.
+ */
+
+METHODDEF(void)
+start_pass_upsample(j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
+
+  /* Mark the conversion buffer empty */
+  upsample->next_row_out = cinfo->max_v_samp_factor;
+  /* Initialize total-height counter for detecting bottom of image */
+  upsample->rows_to_go = cinfo->output_height;
+}
+
+
+/*
+ * Control routine to do upsampling (and color conversion).
+ *
+ * In this version we upsample each component independently.
+ * We upsample one row group into the conversion buffer, then apply
+ * color conversion a row at a time.
+ */
+
+METHODDEF(void)
+sep_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+             JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail,
+             JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+             JDIMENSION out_rows_avail)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
+  int ci;
+  jpeg_component_info *compptr;
+  JDIMENSION num_rows;
+
+  /* Fill the conversion buffer, if it's empty */
+  if (upsample->next_row_out >= cinfo->max_v_samp_factor) {
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+         ci++, compptr++) {
+      /* Invoke per-component upsample method.  Notice we pass a POINTER
+       * to color_bufci, so that fullsize_upsample can change it.
+       */
+      (*upsample->methodsci) (cinfo, compptr,
+        input_bufci + (*in_row_group_ctr * upsample->rowgroup_heightci),
+        upsample->color_buf + ci);
+    }
+    upsample->next_row_out = 0;
+  }
+
+  /* Color-convert and emit rows */
+
+  /* How many we have in the buffer: */
+  num_rows = (JDIMENSION)(cinfo->max_v_samp_factor - upsample->next_row_out);
+  /* Not more than the distance to the end of the image.  Need this test
+   * in case the image height is not a multiple of max_v_samp_factor:
+   */
+  if (num_rows > upsample->rows_to_go)
+    num_rows = upsample->rows_to_go;
+  /* And not more than what the client can accept: */
+  out_rows_avail -= *out_row_ctr;
+  if (num_rows > out_rows_avail)
+    num_rows = out_rows_avail;
+
+  (*cinfo->cconvert->color_convert) (cinfo, upsample->color_buf,
+                                     (JDIMENSION)upsample->next_row_out,
+                                     output_buf + *out_row_ctr, (int)num_rows);
+
+  /* Adjust counts */
+  *out_row_ctr += num_rows;
+  upsample->rows_to_go -= num_rows;
+  upsample->next_row_out += num_rows;
+  /* When the buffer is emptied, declare this input row group consumed */
+  if (upsample->next_row_out >= cinfo->max_v_samp_factor)
+    (*in_row_group_ctr)++;
+}
+
+
+/*
+ * These are the routines invoked by sep_upsample to upsample pixel values
+ * of a single component.  One row group is processed per call.
+ */
+
+
+/*
+ * For full-size components, we just make color_bufci point at the
+ * input buffer, and thus avoid copying any data.  Note that this is
+ * safe only because sep_upsample doesn't declare the input row group
+ * "consumed" until we are done color converting and emitting it.
+ */
+
+METHODDEF(void)
+fullsize_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  *output_data_ptr = input_data;
+}
+
+
+/*
+ * This is a no-op version used for "uninteresting" components.
+ * These components will not be referenced by color conversion.
+ */
+
+METHODDEF(void)
+noop_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  *output_data_ptr = NULL;      /* safety check */
+}
+
+
+/*
+ * This version handles any integral sampling ratios.
+ * This is not used for typical JPEG files, so it need not be fast.
+ * Nor, for that matter, is it particularly accurate: the algorithm is
+ * simple replication of the input pixel onto the corresponding output
+ * pixels.  The hi-falutin sampling literature refers to this as a
+ * "box filter".  A box filter tends to introduce visible artifacts,
+ * so if you are actually going to use 3:1 or 4:1 sampling ratios
+ * you would be well advised to improve this code.
+ */
+
+METHODDEF(void)
+int_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+             JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register JSAMPLE invalue;
+  register int h;
+  JSAMPROW outend;
+  int h_expand, v_expand;
+  int inrow, outrow;
+
+  h_expand = upsample->h_expandcompptr->component_index;
+  v_expand = upsample->v_expandcompptr->component_index;
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    /* Generate one output row with proper horizontal expansion */
+    inptr = input_datainrow;
+    outptr = output_dataoutrow;
+    outend = outptr + cinfo->output_width;
+    while (outptr < outend) {
+      invalue = *inptr++;
+      for (h = h_expand; h > 0; h--) {
+        *outptr++ = invalue;
+      }
+    }
+    /* Generate any additional output rows by duplicating the first one */
+    if (v_expand > 1) {
+      jcopy_sample_rows(output_data, outrow, output_data, outrow + 1,
+                        v_expand - 1, cinfo->output_width);
+    }
+    inrow++;
+    outrow += v_expand;
+  }
+}
+
+
+/*
+ * Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+ * It's still a box filter.
+ */
+
+METHODDEF(void)
+h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register JSAMPLE invalue;
+  JSAMPROW outend;
+  int inrow;
+
+  for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) {
+    inptr = input_datainrow;
+    outptr = output_datainrow;
+    outend = outptr + cinfo->output_width;
+    while (outptr < outend) {
+      invalue = *inptr++;
+      *outptr++ = invalue;
+      *outptr++ = invalue;
+    }
+  }
+}
+
+
+/*
+ * Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+ * It's still a box filter.
+ */
+
+METHODDEF(void)
+h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register JSAMPLE invalue;
+  JSAMPROW outend;
+  int inrow, outrow;
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    inptr = input_datainrow;
+    outptr = output_dataoutrow;
+    outend = outptr + cinfo->output_width;
+    while (outptr < outend) {
+      invalue = *inptr++;
+      *outptr++ = invalue;
+      *outptr++ = invalue;
+    }
+    jcopy_sample_rows(output_data, outrow, output_data, outrow + 1, 1,
+                      cinfo->output_width);
+    inrow++;
+    outrow += 2;
+  }
+}
+
+
+/*
+ * Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+ *
+ * The upsampling algorithm is linear interpolation between pixel centers,
+ * also known as a "triangle filter".  This is a good compromise between
+ * speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+ * of the way between input pixel centers.
+ *
+ * A note about the "bias" calculations: when rounding fractional values to
+ * integer, we do not want to always round 0.5 up to the next integer.
+ * If we did that, we'd introduce a noticeable bias towards larger values.
+ * Instead, this code is arranged so that 0.5 will be rounded up or down at
+ * alternate pixel locations (a simple ordered dither pattern).
+ */
+
+METHODDEF(void)
+h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register int invalue;
+  register JDIMENSION colctr;
+  int inrow;
+
+  for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) {
+    inptr = input_datainrow;
+    outptr = output_datainrow;
+    /* Special case for first column */
+    invalue = *inptr++;
+    *outptr++ = (JSAMPLE)invalue;
+    *outptr++ = (JSAMPLE)((invalue * 3 + inptr0 + 2) >> 2);
+
+    for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
+      /* General case: 3/4 * nearer pixel + 1/4 * further pixel */
+      invalue = (*inptr++) * 3;
+      *outptr++ = (JSAMPLE)((invalue + inptr-2 + 1) >> 2);
+      *outptr++ = (JSAMPLE)((invalue + inptr0 + 2) >> 2);
+    }
+
+    /* Special case for last column */
+    invalue = *inptr;
+    *outptr++ = (JSAMPLE)((invalue * 3 + inptr-1 + 1) >> 2);
+    *outptr++ = (JSAMPLE)invalue;
+  }
+}
+
+
+/*
+ * Fancy processing for 1:1 horizontal and 2:1 vertical (4:4:0 subsampling).
+ *
+ * This is a less common case, but it can be encountered when losslessly
+ * rotating/transposing a JPEG file that uses 4:2:2 chroma subsampling.
+ */
+
+METHODDEF(void)
+h1v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr0, inptr1, outptr;
+#if BITS_IN_JSAMPLE == 8
+  int thiscolsum, bias;
+#else
+  JLONG thiscolsum, bias;
+#endif
+  JDIMENSION colctr;
+  int inrow, outrow, v;
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    for (v = 0; v < 2; v++) {
+      /* inptr0 points to nearest input row, inptr1 points to next nearest */
+      inptr0 = input_datainrow;
+      if (v == 0) {             /* next nearest is row above */
+        inptr1 = input_datainrow - 1;
+        bias = 1;
+      } else {                  /* next nearest is row below */
+        inptr1 = input_datainrow + 1;
+        bias = 2;
+      }
+      outptr = output_dataoutrow++;
+
+      for (colctr = 0; colctr < compptr->downsampled_width; colctr++) {
+        thiscolsum = (*inptr0++) * 3 + (*inptr1++);
+        *outptr++ = (JSAMPLE)((thiscolsum + bias) >> 2);
+      }
+    }
+    inrow++;
+  }
+}
+
+
+/*
+ * Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+ * Again a triangle filter; see comments for h2v1 case, above.
+ *
+ * It is OK for us to reference the adjacent input rows because we demanded
+ * context from the main buffer controller (see initialization code).
+ */
+
+METHODDEF(void)
+h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr0, inptr1, outptr;
+#if BITS_IN_JSAMPLE == 8
+  register int thiscolsum, lastcolsum, nextcolsum;
+#else
+  register JLONG thiscolsum, lastcolsum, nextcolsum;
+#endif
+  register JDIMENSION colctr;
+  int inrow, outrow, v;
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    for (v = 0; v < 2; v++) {
+      /* inptr0 points to nearest input row, inptr1 points to next nearest */
+      inptr0 = input_datainrow;
+      if (v == 0)               /* next nearest is row above */
+        inptr1 = input_datainrow - 1;
+      else                      /* next nearest is row below */
+        inptr1 = input_datainrow + 1;
+      outptr = output_dataoutrow++;
+
+      /* Special case for first column */
+      thiscolsum = (*inptr0++) * 3 + (*inptr1++);
+      nextcolsum = (*inptr0++) * 3 + (*inptr1++);
+      *outptr++ = (JSAMPLE)((thiscolsum * 4 + 8) >> 4);
+      *outptr++ = (JSAMPLE)((thiscolsum * 3 + nextcolsum + 7) >> 4);
+      lastcolsum = thiscolsum;  thiscolsum = nextcolsum;
+
+      for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
+        /* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
+        /* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
+        nextcolsum = (*inptr0++) * 3 + (*inptr1++);
+        *outptr++ = (JSAMPLE)((thiscolsum * 3 + lastcolsum + 8) >> 4);
+        *outptr++ = (JSAMPLE)((thiscolsum * 3 + nextcolsum + 7) >> 4);
+        lastcolsum = thiscolsum;  thiscolsum = nextcolsum;
+      }
+
+      /* Special case for last column */
+      *outptr++ = (JSAMPLE)((thiscolsum * 3 + lastcolsum + 8) >> 4);
+      *outptr++ = (JSAMPLE)((thiscolsum * 4 + 7) >> 4);
+    }
+    inrow++;
+  }
+}
+
+
+/*
+ * Module initialization routine for upsampling.
+ */
+
+GLOBAL(void)
+jinit_upsampler(j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample;
+  int ci;
+  jpeg_component_info *compptr;
+  boolean need_buffer, do_fancy;
+  int h_in_group, v_in_group, h_out_group, v_out_group;
+
+  if (!cinfo->master->jinit_upsampler_no_alloc) {
+    upsample = (my_upsample_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  sizeof(my_upsampler));
+    cinfo->upsample = (struct jpeg_upsampler *)upsample;
+    upsample->pub.start_pass = start_pass_upsample;
+    upsample->pub.upsample = sep_upsample;
+    upsample->pub.need_context_rows = FALSE; /* until we find out differently */
+  } else
+    upsample = (my_upsample_ptr)cinfo->upsample;
+
+  if (cinfo->CCIR601_sampling)  /* this isn't supported */
+    ERREXIT(cinfo, JERR_CCIR601_NOTIMPL);
+
+  /* jdmainct.c doesn't support context rows when min_DCT_scaled_size = 1,
+   * so don't ask for it.
+   */
+  do_fancy = cinfo->do_fancy_upsampling && cinfo->_min_DCT_scaled_size > 1;
+
+  /* Verify we can handle the sampling factors, select per-component methods,
+   * and create storage as needed.
+   */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Compute size of an "input group" after IDCT scaling.  This many samples
+     * are to be converted to max_h_samp_factor * max_v_samp_factor pixels.
+     */
+    h_in_group = (compptr->h_samp_factor * compptr->_DCT_scaled_size) /
+                 cinfo->_min_DCT_scaled_size;
+    v_in_group = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
+                 cinfo->_min_DCT_scaled_size;
+    h_out_group = cinfo->max_h_samp_factor;
+    v_out_group = cinfo->max_v_samp_factor;
+    upsample->rowgroup_heightci = v_in_group; /* save for use later */
+    need_buffer = TRUE;
+    if (!compptr->component_needed) {
+      /* Don't bother to upsample an uninteresting component. */
+      upsample->methodsci = noop_upsample;
+      need_buffer = FALSE;
+    } else if (h_in_group == h_out_group && v_in_group == v_out_group) {
+      /* Fullsize components can be processed without any work. */
+      upsample->methodsci = fullsize_upsample;
+      need_buffer = FALSE;
+    } else if (h_in_group * 2 == h_out_group && v_in_group == v_out_group) {
+      /* Special cases for 2h1v upsampling */
+      if (do_fancy && compptr->downsampled_width > 2) {
+        if (jsimd_can_h2v1_fancy_upsample())
+          upsample->methodsci = jsimd_h2v1_fancy_upsample;
+        else
+          upsample->methodsci = h2v1_fancy_upsample;
+      } else {
+        if (jsimd_can_h2v1_upsample())
+          upsample->methodsci = jsimd_h2v1_upsample;
+        else
+          upsample->methodsci = h2v1_upsample;
+      }
+    } else if (h_in_group == h_out_group &&
+               v_in_group * 2 == v_out_group && do_fancy) {
+      /* Non-fancy upsampling is handled by the generic method */
+#if defined(__arm__) || defined(__aarch64__) || \
+    defined(_M_ARM) || defined(_M_ARM64)
+      if (jsimd_can_h1v2_fancy_upsample())
+        upsample->methodsci = jsimd_h1v2_fancy_upsample;
+      else
+#endif
+        upsample->methodsci = h1v2_fancy_upsample;
+      upsample->pub.need_context_rows = TRUE;
+    } else if (h_in_group * 2 == h_out_group &&
+               v_in_group * 2 == v_out_group) {
+      /* Special cases for 2h2v upsampling */
+      if (do_fancy && compptr->downsampled_width > 2) {
+        if (jsimd_can_h2v2_fancy_upsample())
+          upsample->methodsci = jsimd_h2v2_fancy_upsample;
+        else
+          upsample->methodsci = h2v2_fancy_upsample;
+        upsample->pub.need_context_rows = TRUE;
+      } else {
+        if (jsimd_can_h2v2_upsample())
+          upsample->methodsci = jsimd_h2v2_upsample;
+        else
+          upsample->methodsci = h2v2_upsample;
+      }
+    } else if ((h_out_group % h_in_group) == 0 &&
+               (v_out_group % v_in_group) == 0) {
+      /* Generic integral-factors upsampling method */
+#if defined(__mips__)
+      if (jsimd_can_int_upsample())
+        upsample->methodsci = jsimd_int_upsample;
+      else
+#endif
+        upsample->methodsci = int_upsample;
+      upsample->h_expandci = (UINT8)(h_out_group / h_in_group);
+      upsample->v_expandci = (UINT8)(v_out_group / v_in_group);
+    } else
+      ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL);
+    if (need_buffer && !cinfo->master->jinit_upsampler_no_alloc) {
+      upsample->color_bufci = (*cinfo->mem->alloc_sarray)
+        ((j_common_ptr)cinfo, JPOOL_IMAGE,
+         (JDIMENSION)jround_up((long)cinfo->output_width,
+                               (long)cinfo->max_h_samp_factor),
+         (JDIMENSION)cinfo->max_v_samp_factor);
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdsample.h Added

@@ -0,0 +1,50 @@
+/*
+ * jdsample.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+
+
+/* Pointer to routine to upsample a single component */
+typedef void (*upsample1_ptr) (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JSAMPARRAY input_data,
+                               JSAMPARRAY *output_data_ptr);
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_upsampler pub;    /* public fields */
+
+  /* Color conversion buffer.  When using separate upsampling and color
+   * conversion steps, this buffer holds one upsampled row group until it
+   * has been color converted and output.
+   * Note: we do not allocate any storage for component(s) which are full-size,
+   * ie do not need rescaling.  The corresponding entry of color_buf is
+   * simply set to point to the input data array, thereby avoiding copying.
+   */
+  JSAMPARRAY color_bufMAX_COMPONENTS;
+
+  /* Per-component upsampling method pointers */
+  upsample1_ptr methodsMAX_COMPONENTS;
+
+  int next_row_out;             /* counts rows emitted from color_buf */
+  JDIMENSION rows_to_go;        /* counts rows remaining in image */
+
+  /* Height of an input row group for each component. */
+  int rowgroup_heightMAX_COMPONENTS;
+
+  /* These arrays save pixel expansion factors so that int_expand need not
+   * recompute them each time.  They are unused for other upsampling methods.
+   */
+  UINT8 h_expandMAX_COMPONENTS;
+  UINT8 v_expandMAX_COMPONENTS;
+} my_upsampler;
+
+typedef my_upsampler *my_upsample_ptr;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jdtrans.c Added

@@ -0,0 +1,156 @@
+/*
+ * jdtrans.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains library routines for transcoding decompression,
+ * that is, reading raw DCT coefficient arrays from an input JPEG file.
+ * The routines in jdapimin.c will also be needed by a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jpegcomp.h"
+
+
+/* Forward declarations */
+LOCAL(void) transdecode_master_selection(j_decompress_ptr cinfo);
+
+
+/*
+ * Read the coefficient arrays from a JPEG file.
+ * jpeg_read_header must be completed before calling this.
+ *
+ * The entire image is read into a set of virtual coefficient-block arrays,
+ * one per component.  The return value is a pointer to the array of
+ * virtual-array descriptors.  These can be manipulated directly via the
+ * JPEG memory manager, or handed off to jpeg_write_coefficients().
+ * To release the memory occupied by the virtual arrays, call
+ * jpeg_finish_decompress() when done with the data.
+ *
+ * An alternative usage is to simply obtain access to the coefficient arrays
+ * during a buffered-image-mode decompression operation.  This is allowed
+ * after any jpeg_finish_output() call.  The arrays can be accessed until
+ * jpeg_finish_decompress() is called.  (Note that any call to the library
+ * may reposition the arrays, so don't rely on access_virt_barray() results
+ * to stay valid across library calls.)
+ *
+ * Returns NULL if suspended.  This case need be checked only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(jvirt_barray_ptr *)
+jpeg_read_coefficients(j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state == DSTATE_READY) {
+    /* First call: initialize active modules */
+    transdecode_master_selection(cinfo);
+    cinfo->global_state = DSTATE_RDCOEFS;
+  }
+  if (cinfo->global_state == DSTATE_RDCOEFS) {
+    /* Absorb whole file into the coef buffer */
+    for (;;) {
+      int retcode;
+      /* Call progress monitor hook if present */
+      if (cinfo->progress != NULL)
+        (*cinfo->progress->progress_monitor) ((j_common_ptr)cinfo);
+      /* Absorb some more input */
+      retcode = (*cinfo->inputctl->consume_input) (cinfo);
+      if (retcode == JPEG_SUSPENDED)
+        return NULL;
+      if (retcode == JPEG_REACHED_EOI)
+        break;
+      /* Advance progress counter if appropriate */
+      if (cinfo->progress != NULL &&
+          (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+        if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+          /* startup underestimated number of scans; ratchet up one scan */
+          cinfo->progress->pass_limit += (long)cinfo->total_iMCU_rows;
+        }
+      }
+    }
+    /* Set state so that jpeg_finish_decompress does the right thing */
+    cinfo->global_state = DSTATE_STOPPING;
+  }
+  /* At this point we should be in state DSTATE_STOPPING if being used
+   * standalone, or in state DSTATE_BUFIMAGE if being invoked to get access
+   * to the coefficients during a full buffered-image-mode decompression.
+   */
+  if ((cinfo->global_state == DSTATE_STOPPING ||
+       cinfo->global_state == DSTATE_BUFIMAGE) && cinfo->buffered_image) {
+    return cinfo->coef->coef_arrays;
+  }
+  /* Oops, improper usage */
+  ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return NULL;                  /* keep compiler happy */
+}
+
+
+/*
+ * Master selection of decompression modules for transcoding.
+ * This substitutes for jdmaster.c's initialization of the full decompressor.
+ */
+
+LOCAL(void)
+transdecode_master_selection(j_decompress_ptr cinfo)
+{
+  /* This is effectively a buffered-image operation. */
+  cinfo->buffered_image = TRUE;
+
+#if JPEG_LIB_VERSION >= 80
+  /* Compute output image dimensions and related values. */
+  jpeg_core_output_dimensions(cinfo);
+#endif
+
+  /* Entropy decoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+#ifdef D_ARITH_CODING_SUPPORTED
+    jinit_arith_decoder(cinfo);
+#else
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+#endif
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef D_PROGRESSIVE_SUPPORTED
+      jinit_phuff_decoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_decoder(cinfo);
+  }
+
+  /* Always get a full-image coefficient buffer. */
+  jinit_d_coef_controller(cinfo, TRUE);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr)cinfo);
+
+  /* Initialize input side of decompressor to consume first scan. */
+  (*cinfo->inputctl->start_input_pass) (cinfo);
+
+  /* Initialize progress monitoring. */
+  if (cinfo->progress != NULL) {
+    int nscans;
+    /* Estimate number of scans to set pass_limit. */
+    if (cinfo->progressive_mode) {
+      /* Arbitrarily estimate 2 interleaved DC scans + 3 AC scans/component. */
+      nscans = 2 + 3 * cinfo->num_components;
+    } else if (cinfo->inputctl->has_multiple_scans) {
+      /* For a nonprogressive multiscan file, estimate 1 scan per component. */
+      nscans = cinfo->num_components;
+    } else {
+      nscans = 1;
+    }
+    cinfo->progress->pass_counter = 0L;
+    cinfo->progress->pass_limit = (long)cinfo->total_iMCU_rows * nscans;
+    cinfo->progress->completed_passes = 0;
+    cinfo->progress->total_passes = 1;
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jerror.c Added

@@ -0,0 +1,251 @@
+/*
+ * jerror.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains simple error-reporting and trace-message routines.
+ * These are suitable for Unix-like systems and others where writing to
+ * stderr is the right thing to do.  Many applications will want to replace
+ * some or all of these routines.
+ *
+ * If you define USE_WINDOWS_MESSAGEBOX in jconfig.h or in the makefile,
+ * you get a Windows-specific hack to display error messages in a dialog box.
+ * It ain't much, but it beats dropping error messages into the bit bucket,
+ * which is what happens to output to stderr under most Windows C compilers.
+ *
+ * These routines are used by both the compression and decompression code.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jversion.h"
+#include "jerror.h"
+
+#ifdef USE_WINDOWS_MESSAGEBOX
+#include <windows.h>
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+
+
+/*
+ * Create the message string table.
+ * We do this from the master message list in jerror.h by re-reading
+ * jerror.h with a suitable definition for macro JMESSAGE.
+ * The message table is made an external symbol just in case any applications
+ * want to refer to it directly.
+ */
+
+#define JMESSAGE(code, string)  string,
+
+const char * const jpeg_std_message_table = {
+#include "jerror.h"
+  NULL
+};
+
+
+/*
+ * Error exit handler: must not return to caller.
+ *
+ * Applications may override this if they want to get control back after
+ * an error.  Typically one would longjmp somewhere instead of exiting.
+ * The setjmp buffer can be made a private field within an expanded error
+ * handler object.  Note that the info needed to generate an error message
+ * is stored in the error object, so you can generate the message now or
+ * later, at your convenience.
+ * You should make sure that the JPEG object is cleaned up (with jpeg_abort
+ * or jpeg_destroy) at some point.
+ */
+
+METHODDEF(void)
+error_exit(j_common_ptr cinfo)
+{
+  /* Always display the message */
+  (*cinfo->err->output_message) (cinfo);
+
+  /* Let the memory manager delete any temp files before we die */
+  jpeg_destroy(cinfo);
+
+  exit(EXIT_FAILURE);
+}
+
+
+/*
+ * Actual output of an error or trace message.
+ * Applications may override this method to send JPEG messages somewhere
+ * other than stderr.
+ *
+ * On Windows, printing to stderr is generally completely useless,
+ * so we provide optional code to produce an error-dialog popup.
+ * Most Windows applications will still prefer to override this routine,
+ * but if they don't, it'll do something at least marginally useful.
+ *
+ * NOTE: to use the library in an environment that doesn't support the
+ * C stdio library, you may have to delete the call to fprintf() entirely,
+ * not just not use this routine.
+ */
+
+METHODDEF(void)
+output_message(j_common_ptr cinfo)
+{
+  char bufferJMSG_LENGTH_MAX;
+
+  /* Create the message */
+  (*cinfo->err->format_message) (cinfo, buffer);
+
+#ifdef USE_WINDOWS_MESSAGEBOX
+  /* Display it in a message dialog box */
+  MessageBox(GetActiveWindow(), buffer, "JPEG Library Error",
+             MB_OK | MB_ICONERROR);
+#else
+  /* Send it to stderr, adding a newline */
+  fprintf(stderr, "%s\n", buffer);
+#endif
+}
+
+
+/*
+ * Decide whether to emit a trace or warning message.
+ * msg_level is one of:
+ *   -1: recoverable corrupt-data warning, may want to abort.
+ *    0: important advisory messages (always display to user).
+ *    1: first level of tracing detail.
+ *    2,3,...: successively more detailed tracing messages.
+ * An application might override this method if it wanted to abort on warnings
+ * or change the policy about which messages to display.
+ */
+
+METHODDEF(void)
+emit_message(j_common_ptr cinfo, int msg_level)
+{
+  struct jpeg_error_mgr *err = cinfo->err;
+
+  if (msg_level < 0) {
+    /* It's a warning message.  Since corrupt files may generate many warnings,
+     * the policy implemented here is to show only the first warning,
+     * unless trace_level >= 3.
+     */
+    if (err->num_warnings == 0 || err->trace_level >= 3)
+      (*err->output_message) (cinfo);
+    /* Always count warnings in num_warnings. */
+    err->num_warnings++;
+  } else {
+    /* It's a trace message.  Show it if trace_level >= msg_level. */
+    if (err->trace_level >= msg_level)
+      (*err->output_message) (cinfo);
+  }
+}
+
+
+/*
+ * Format a message string for the most recent JPEG error or message.
+ * The message is stored into buffer, which should be at least JMSG_LENGTH_MAX
+ * characters.  Note that no '\n' character is added to the string.
+ * Few applications should need to override this method.
+ */
+
+METHODDEF(void)
+format_message(j_common_ptr cinfo, char *buffer)
+{
+  struct jpeg_error_mgr *err = cinfo->err;
+  int msg_code = err->msg_code;
+  const char *msgtext = NULL;
+  const char *msgptr;
+  char ch;
+  boolean isstring;
+
+  /* Look up message string in proper table */
+  if (msg_code > 0 && msg_code <= err->last_jpeg_message) {
+    msgtext = err->jpeg_message_tablemsg_code;
+  } else if (err->addon_message_table != NULL &&
+             msg_code >= err->first_addon_message &&
+             msg_code <= err->last_addon_message) {
+    msgtext = err->addon_message_tablemsg_code - err->first_addon_message;
+  }
+
+  /* Defend against bogus message number */
+  if (msgtext == NULL) {
+    err->msg_parm.i0 = msg_code;
+    msgtext = err->jpeg_message_table0;
+  }
+
+  /* Check for string parameter, as indicated by %s in the message text */
+  isstring = FALSE;
+  msgptr = msgtext;
+  while ((ch = *msgptr++) != '\0') {
+    if (ch == '%') {
+      if (*msgptr == 's') isstring = TRUE;
+      break;
+    }
+  }
+
+  /* Format the message into the passed buffer */
+  if (isstring)
+    sprintf(buffer, msgtext, err->msg_parm.s);
+  else
+    sprintf(buffer, msgtext,
+            err->msg_parm.i0, err->msg_parm.i1,
+            err->msg_parm.i2, err->msg_parm.i3,
+            err->msg_parm.i4, err->msg_parm.i5,
+            err->msg_parm.i6, err->msg_parm.i7);
+}
+
+
+/*
+ * Reset error state variables at start of a new image.
+ * This is called during compression startup to reset trace/error
+ * processing to default state, without losing any application-specific
+ * method pointers.  An application might possibly want to override
+ * this method if it has additional error processing state.
+ */
+
+METHODDEF(void)
+reset_error_mgr(j_common_ptr cinfo)
+{
+  cinfo->err->num_warnings = 0;
+  /* trace_level is not reset since it is an application-supplied parameter */
+  cinfo->err->msg_code = 0;     /* may be useful as a flag for "no error" */
+}
+
+
+/*
+ * Fill in the standard error-handling methods in a jpeg_error_mgr object.
+ * Typical call is:
+ *      struct jpeg_compress_struct cinfo;
+ *      struct jpeg_error_mgr err;
+ *
+ *      cinfo.err = jpeg_std_error(&err);
+ * after which the application may override some of the methods.
+ */
+
+GLOBAL(struct jpeg_error_mgr *)
+jpeg_std_error(struct jpeg_error_mgr *err)
+{
+  err->error_exit = error_exit;
+  err->emit_message = emit_message;
+  err->output_message = output_message;
+  err->format_message = format_message;
+  err->reset_error_mgr = reset_error_mgr;
+
+  err->trace_level = 0;         /* default = no tracing */
+  err->num_warnings = 0;        /* no warnings emitted yet */
+  err->msg_code = 0;            /* may be useful as a flag for "no error" */
+
+  /* Initialize message table pointers */
+  err->jpeg_message_table = jpeg_std_message_table;
+  err->last_jpeg_message = (int)JMSG_LASTMSGCODE - 1;
+
+  err->addon_message_table = NULL;
+  err->first_addon_message = 0; /* for safety */
+  err->last_addon_message = 0;
+
+  return err;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jerror.h Added

@@ -0,0 +1,329 @@
+/*
+ * jerror.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * Modified 1997-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2014, 2017, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file defines the error and message codes for the JPEG library.
+ * Edit this file to add new codes, or to translate the message strings to
+ * some other language.
+ * A set of error-reporting macros are defined too.  Some applications using
+ * the JPEG library may wish to include this file to get the error codes
+ * and/or the macros.
+ */
+
+/*
+ * To define the enum list of message codes, include this file without
+ * defining macro JMESSAGE.  To create a message string table, include it
+ * again with a suitable JMESSAGE definition (see jerror.c for an example).
+ */
+#ifndef JMESSAGE
+#ifndef JERROR_H
+/* First time through, define the enum list */
+#define JMAKE_ENUM_LIST
+#else
+/* Repeated inclusions of this file are no-ops unless JMESSAGE is defined */
+#define JMESSAGE(code, string)
+#endif /* JERROR_H */
+#endif /* JMESSAGE */
+
+#ifdef JMAKE_ENUM_LIST
+
+typedef enum {
+
+#define JMESSAGE(code, string)  code,
+
+#endif /* JMAKE_ENUM_LIST */
+
+JMESSAGE(JMSG_NOMESSAGE, "Bogus message code %d") /* Must be first entry! */
+
+/* For maintenance convenience, list is alphabetical by message code name */
+#if JPEG_LIB_VERSION < 70
+JMESSAGE(JERR_ARITH_NOTIMPL, "Sorry, arithmetic coding is not implemented")
+#endif
+JMESSAGE(JERR_BAD_ALIGN_TYPE, "ALIGN_TYPE is wrong, please fix")
+JMESSAGE(JERR_BAD_ALLOC_CHUNK, "MAX_ALLOC_CHUNK is wrong, please fix")
+JMESSAGE(JERR_BAD_BUFFER_MODE, "Bogus buffer control mode")
+JMESSAGE(JERR_BAD_COMPONENT_ID, "Invalid component ID %d in SOS")
+#if JPEG_LIB_VERSION >= 70
+JMESSAGE(JERR_BAD_CROP_SPEC, "Invalid crop request")
+#endif
+JMESSAGE(JERR_BAD_DCT_COEF, "DCT coefficient out of range")
+JMESSAGE(JERR_BAD_DCTSIZE, "IDCT output block size %d not supported")
+#if JPEG_LIB_VERSION >= 70
+JMESSAGE(JERR_BAD_DROP_SAMPLING,
+         "Component index %d: mismatching sampling ratio %d:%d, %d:%d, %c")
+#endif
+JMESSAGE(JERR_BAD_HUFF_TABLE, "Bogus Huffman table definition")
+JMESSAGE(JERR_BAD_IN_COLORSPACE, "Bogus input colorspace")
+JMESSAGE(JERR_BAD_J_COLORSPACE, "Bogus JPEG colorspace")
+JMESSAGE(JERR_BAD_LENGTH, "Bogus marker length")
+JMESSAGE(JERR_BAD_LIB_VERSION,
+         "Wrong JPEG library version: library is %d, caller expects %d")
+JMESSAGE(JERR_BAD_MCU_SIZE, "Sampling factors too large for interleaved scan")
+JMESSAGE(JERR_BAD_POOL_ID, "Invalid memory pool code %d")
+JMESSAGE(JERR_BAD_PRECISION, "Unsupported JPEG data precision %d")
+JMESSAGE(JERR_BAD_PROGRESSION,
+         "Invalid progressive parameters Ss=%d Se=%d Ah=%d Al=%d")
+JMESSAGE(JERR_BAD_PROG_SCRIPT,
+         "Invalid progressive parameters at scan script entry %d")
+JMESSAGE(JERR_BAD_SAMPLING, "Bogus sampling factors")
+JMESSAGE(JERR_BAD_SCAN_SCRIPT, "Invalid scan script at entry %d")
+JMESSAGE(JERR_BAD_STATE, "Improper call to JPEG library in state %d")
+JMESSAGE(JERR_BAD_STRUCT_SIZE,
+         "JPEG parameter struct mismatch: library thinks size is %u, caller expects %u")
+JMESSAGE(JERR_BAD_VIRTUAL_ACCESS, "Bogus virtual array access")
+JMESSAGE(JERR_BUFFER_SIZE, "Buffer passed to JPEG library is too small")
+JMESSAGE(JERR_CANT_SUSPEND, "Suspension not allowed here")
+JMESSAGE(JERR_CCIR601_NOTIMPL, "CCIR601 sampling not implemented yet")
+JMESSAGE(JERR_COMPONENT_COUNT, "Too many color components: %d, max %d")
+JMESSAGE(JERR_CONVERSION_NOTIMPL, "Unsupported color conversion request")
+JMESSAGE(JERR_DAC_INDEX, "Bogus DAC index %d")
+JMESSAGE(JERR_DAC_VALUE, "Bogus DAC value 0x%x")
+JMESSAGE(JERR_DHT_INDEX, "Bogus DHT index %d")
+JMESSAGE(JERR_DQT_INDEX, "Bogus DQT index %d")
+JMESSAGE(JERR_EMPTY_IMAGE, "Empty JPEG image (DNL not supported)")
+JMESSAGE(JERR_EMS_READ, "Read from EMS failed")
+JMESSAGE(JERR_EMS_WRITE, "Write to EMS failed")
+JMESSAGE(JERR_EOI_EXPECTED, "Didn't expect more than one scan")
+JMESSAGE(JERR_FILE_READ, "Input file read error")
+JMESSAGE(JERR_FILE_WRITE, "Output file write error --- out of disk space?")
+JMESSAGE(JERR_FRACT_SAMPLE_NOTIMPL, "Fractional sampling not implemented yet")
+JMESSAGE(JERR_HUFF_CLEN_OVERFLOW, "Huffman code size table overflow")
+JMESSAGE(JERR_HUFF_MISSING_CODE, "Missing Huffman code table entry")
+JMESSAGE(JERR_IMAGE_TOO_BIG, "Maximum supported image dimension is %u pixels")
+JMESSAGE(JERR_INPUT_EMPTY, "Empty input file")
+JMESSAGE(JERR_INPUT_EOF, "Premature end of input file")
+JMESSAGE(JERR_MISMATCHED_QUANT_TABLE,
+         "Cannot transcode due to multiple use of quantization table %d")
+JMESSAGE(JERR_MISSING_DATA, "Scan script does not transmit all data")
+JMESSAGE(JERR_MODE_CHANGE, "Invalid color quantization mode change")
+JMESSAGE(JERR_NOTIMPL, "Not implemented yet")
+JMESSAGE(JERR_NOT_COMPILED, "Requested feature was omitted at compile time")
+#if JPEG_LIB_VERSION >= 70
+JMESSAGE(JERR_NO_ARITH_TABLE, "Arithmetic table 0x%02x was not defined")
+#endif
+JMESSAGE(JERR_NO_BACKING_STORE, "Backing store not supported")
+JMESSAGE(JERR_NO_HUFF_TABLE, "Huffman table 0x%02x was not defined")
+JMESSAGE(JERR_NO_IMAGE, "JPEG datastream contains no image")
+JMESSAGE(JERR_NO_QUANT_TABLE, "Quantization table 0x%02x was not defined")
+JMESSAGE(JERR_NO_SOI, "Not a JPEG file: starts with 0x%02x 0x%02x")
+JMESSAGE(JERR_OUT_OF_MEMORY, "Insufficient memory (case %d)")
+JMESSAGE(JERR_QUANT_COMPONENTS,
+         "Cannot quantize more than %d color components")
+JMESSAGE(JERR_QUANT_FEW_COLORS, "Cannot quantize to fewer than %d colors")
+JMESSAGE(JERR_QUANT_MANY_COLORS, "Cannot quantize to more than %d colors")
+JMESSAGE(JERR_SOF_DUPLICATE, "Invalid JPEG file structure: two SOF markers")
+JMESSAGE(JERR_SOF_NO_SOS, "Invalid JPEG file structure: missing SOS marker")
+JMESSAGE(JERR_SOF_UNSUPPORTED, "Unsupported JPEG process: SOF type 0x%02x")
+JMESSAGE(JERR_SOI_DUPLICATE, "Invalid JPEG file structure: two SOI markers")
+JMESSAGE(JERR_SOS_NO_SOF, "Invalid JPEG file structure: SOS before SOF")
+JMESSAGE(JERR_TFILE_CREATE, "Failed to create temporary file %s")
+JMESSAGE(JERR_TFILE_READ, "Read failed on temporary file")
+JMESSAGE(JERR_TFILE_SEEK, "Seek failed on temporary file")
+JMESSAGE(JERR_TFILE_WRITE,
+         "Write failed on temporary file --- out of disk space?")
+JMESSAGE(JERR_TOO_LITTLE_DATA, "Application transferred too few scanlines")
+JMESSAGE(JERR_UNKNOWN_MARKER, "Unsupported marker type 0x%02x")
+JMESSAGE(JERR_VIRTUAL_BUG, "Virtual array controller messed up")
+JMESSAGE(JERR_WIDTH_OVERFLOW, "Image too wide for this implementation")
+JMESSAGE(JERR_XMS_READ, "Read from XMS failed")
+JMESSAGE(JERR_XMS_WRITE, "Write to XMS failed")
+JMESSAGE(JMSG_COPYRIGHT, JCOPYRIGHT_SHORT)
+JMESSAGE(JMSG_VERSION, JVERSION)
+JMESSAGE(JTRC_16BIT_TABLES,
+         "Caution: quantization tables are too coarse for baseline JPEG")
+JMESSAGE(JTRC_ADOBE,
+         "Adobe APP14 marker: version %d, flags 0x%04x 0x%04x, transform %d")
+JMESSAGE(JTRC_APP0, "Unknown APP0 marker (not JFIF), length %u")
+JMESSAGE(JTRC_APP14, "Unknown APP14 marker (not Adobe), length %u")
+JMESSAGE(JTRC_DAC, "Define Arithmetic Table 0x%02x: 0x%02x")
+JMESSAGE(JTRC_DHT, "Define Huffman Table 0x%02x")
+JMESSAGE(JTRC_DQT, "Define Quantization Table %d  precision %d")
+JMESSAGE(JTRC_DRI, "Define Restart Interval %u")
+JMESSAGE(JTRC_EMS_CLOSE, "Freed EMS handle %u")
+JMESSAGE(JTRC_EMS_OPEN, "Obtained EMS handle %u")
+JMESSAGE(JTRC_EOI, "End Of Image")
+JMESSAGE(JTRC_HUFFBITS, "        %3d %3d %3d %3d %3d %3d %3d %3d")
+JMESSAGE(JTRC_JFIF, "JFIF APP0 marker: version %d.%02d, density %dx%d  %d")
+JMESSAGE(JTRC_JFIF_BADTHUMBNAILSIZE,
+         "Warning: thumbnail image size does not match data length %u")
+JMESSAGE(JTRC_JFIF_EXTENSION, "JFIF extension marker: type 0x%02x, length %u")
+JMESSAGE(JTRC_JFIF_THUMBNAIL, "    with %d x %d thumbnail image")
+JMESSAGE(JTRC_MISC_MARKER, "Miscellaneous marker 0x%02x, length %u")
+JMESSAGE(JTRC_PARMLESS_MARKER, "Unexpected marker 0x%02x")
+JMESSAGE(JTRC_QUANTVALS, "        %4u %4u %4u %4u %4u %4u %4u %4u")
+JMESSAGE(JTRC_QUANT_3_NCOLORS, "Quantizing to %d = %d*%d*%d colors")
+JMESSAGE(JTRC_QUANT_NCOLORS, "Quantizing to %d colors")
+JMESSAGE(JTRC_QUANT_SELECTED, "Selected %d colors for quantization")
+JMESSAGE(JTRC_RECOVERY_ACTION, "At marker 0x%02x, recovery action %d")
+JMESSAGE(JTRC_RST, "RST%d")
+JMESSAGE(JTRC_SMOOTH_NOTIMPL,
+         "Smoothing not supported with nonstandard sampling ratios")
+JMESSAGE(JTRC_SOF, "Start Of Frame 0x%02x: width=%u, height=%u, components=%d")
+JMESSAGE(JTRC_SOF_COMPONENT, "    Component %d: %dhx%dv q=%d")
+JMESSAGE(JTRC_SOI, "Start of Image")
+JMESSAGE(JTRC_SOS, "Start Of Scan: %d components")
+JMESSAGE(JTRC_SOS_COMPONENT, "    Component %d: dc=%d ac=%d")
+JMESSAGE(JTRC_SOS_PARAMS, "  Ss=%d, Se=%d, Ah=%d, Al=%d")
+JMESSAGE(JTRC_TFILE_CLOSE, "Closed temporary file %s")
+JMESSAGE(JTRC_TFILE_OPEN, "Opened temporary file %s")
+JMESSAGE(JTRC_THUMB_JPEG,
+         "JFIF extension marker: JPEG-compressed thumbnail image, length %u")
+JMESSAGE(JTRC_THUMB_PALETTE,
+         "JFIF extension marker: palette thumbnail image, length %u")
+JMESSAGE(JTRC_THUMB_RGB,
+         "JFIF extension marker: RGB thumbnail image, length %u")
+JMESSAGE(JTRC_UNKNOWN_IDS,
+         "Unrecognized component IDs %d %d %d, assuming YCbCr")
+JMESSAGE(JTRC_XMS_CLOSE, "Freed XMS handle %u")
+JMESSAGE(JTRC_XMS_OPEN, "Obtained XMS handle %u")
+JMESSAGE(JWRN_ADOBE_XFORM, "Unknown Adobe color transform code %d")
+#if JPEG_LIB_VERSION >= 70
+JMESSAGE(JWRN_ARITH_BAD_CODE, "Corrupt JPEG data: bad arithmetic code")
+#endif
+JMESSAGE(JWRN_BOGUS_PROGRESSION,
+         "Inconsistent progression sequence for component %d coefficient %d")
+JMESSAGE(JWRN_EXTRANEOUS_DATA,
+         "Corrupt JPEG data: %u extraneous bytes before marker 0x%02x")
+JMESSAGE(JWRN_HIT_MARKER, "Corrupt JPEG data: premature end of data segment")
+JMESSAGE(JWRN_HUFF_BAD_CODE, "Corrupt JPEG data: bad Huffman code")
+JMESSAGE(JWRN_JFIF_MAJOR, "Warning: unknown JFIF revision number %d.%02d")
+JMESSAGE(JWRN_JPEG_EOF, "Premature end of JPEG file")
+JMESSAGE(JWRN_MUST_RESYNC,
+         "Corrupt JPEG data: found marker 0x%02x instead of RST%d")
+JMESSAGE(JWRN_NOT_SEQUENTIAL, "Invalid SOS parameters for sequential JPEG")
+JMESSAGE(JWRN_TOO_MUCH_DATA, "Application transferred too many scanlines")
+#if JPEG_LIB_VERSION < 70
+JMESSAGE(JERR_BAD_CROP_SPEC, "Invalid crop request")
+#if defined(C_ARITH_CODING_SUPPORTED) || defined(D_ARITH_CODING_SUPPORTED)
+JMESSAGE(JERR_NO_ARITH_TABLE, "Arithmetic table 0x%02x was not defined")
+JMESSAGE(JWRN_ARITH_BAD_CODE, "Corrupt JPEG data: bad arithmetic code")
+#endif
+#endif
+JMESSAGE(JWRN_BOGUS_ICC, "Corrupt JPEG data: bad ICC marker")
+#if JPEG_LIB_VERSION < 70
+JMESSAGE(JERR_BAD_DROP_SAMPLING,
+         "Component index %d: mismatching sampling ratio %d:%d, %d:%d, %c")
+#endif
+
+#ifdef JMAKE_ENUM_LIST
+
+  JMSG_LASTMSGCODE
+} J_MESSAGE_CODE;
+
+#undef JMAKE_ENUM_LIST
+#endif /* JMAKE_ENUM_LIST */
+
+/* Zap JMESSAGE macro so that future re-inclusions do nothing by default */
+#undef JMESSAGE
+
+
+#ifndef JERROR_H
+#define JERROR_H
+
+/* Macros to simplify using the error and trace message stuff */
+/* The first parameter is either type of cinfo pointer */
+
+/* Fatal errors (print message and exit) */
+#define ERREXIT(cinfo, code) \
+  ((cinfo)->err->msg_code = (code), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
+#define ERREXIT1(cinfo, code, p1) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i0 = (p1), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
+#define ERREXIT2(cinfo, code, p1, p2) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i0 = (p1), \
+   (cinfo)->err->msg_parm.i1 = (p2), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
+#define ERREXIT3(cinfo, code, p1, p2, p3) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i0 = (p1), \
+   (cinfo)->err->msg_parm.i1 = (p2), \
+   (cinfo)->err->msg_parm.i2 = (p3), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
+#define ERREXIT4(cinfo, code, p1, p2, p3, p4) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i0 = (p1), \
+   (cinfo)->err->msg_parm.i1 = (p2), \
+   (cinfo)->err->msg_parm.i2 = (p3), \
+   (cinfo)->err->msg_parm.i3 = (p4), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
+#define ERREXIT6(cinfo, code, p1, p2, p3, p4, p5, p6) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i0 = (p1), \
+   (cinfo)->err->msg_parm.i1 = (p2), \
+   (cinfo)->err->msg_parm.i2 = (p3), \
+   (cinfo)->err->msg_parm.i3 = (p4), \
+   (cinfo)->err->msg_parm.i4 = (p5), \
+   (cinfo)->err->msg_parm.i5 = (p6), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
+#define ERREXITS(cinfo, code, str) \
+  ((cinfo)->err->msg_code = (code), \
+   strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
+
+#define MAKESTMT(stuff)         do { stuff } while (0)
+
+/* Nonfatal errors (we can keep going, but the data is probably corrupt) */
+#define WARNMS(cinfo, code) \
+  ((cinfo)->err->msg_code = (code), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), -1))
+#define WARNMS1(cinfo, code, p1) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i0 = (p1), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), -1))
+#define WARNMS2(cinfo, code, p1, p2) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i0 = (p1), \
+   (cinfo)->err->msg_parm.i1 = (p2), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), -1))
+
+/* Informational/debugging messages */
+#define TRACEMS(cinfo, lvl, code) \
+  ((cinfo)->err->msg_code = (code), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), (lvl)))
+#define TRACEMS1(cinfo, lvl, code, p1) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i0 = (p1), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), (lvl)))
+#define TRACEMS2(cinfo, lvl, code, p1, p2) \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i0 = (p1), \
+   (cinfo)->err->msg_parm.i1 = (p2), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), (lvl)))
+#define TRACEMS3(cinfo, lvl, code, p1, p2, p3) \
+  MAKESTMT(int *_mp = (cinfo)->err->msg_parm.i; \
+           _mp0 = (p1);  _mp1 = (p2);  _mp2 = (p3); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), (lvl)); )
+#define TRACEMS4(cinfo, lvl, code, p1, p2, p3, p4) \
+  MAKESTMT(int *_mp = (cinfo)->err->msg_parm.i; \
+           _mp0 = (p1);  _mp1 = (p2);  _mp2 = (p3);  _mp3 = (p4); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), (lvl)); )
+#define TRACEMS5(cinfo, lvl, code, p1, p2, p3, p4, p5) \
+  MAKESTMT(int *_mp = (cinfo)->err->msg_parm.i; \
+           _mp0 = (p1);  _mp1 = (p2);  _mp2 = (p3);  _mp3 = (p4); \
+           _mp4 = (p5); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), (lvl)); )
+#define TRACEMS8(cinfo, lvl, code, p1, p2, p3, p4, p5, p6, p7, p8) \
+  MAKESTMT(int *_mp = (cinfo)->err->msg_parm.i; \
+           _mp0 = (p1);  _mp1 = (p2);  _mp2 = (p3);  _mp3 = (p4); \
+           _mp4 = (p5);  _mp5 = (p6);  _mp6 = (p7);  _mp7 = (p8); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), (lvl)); )
+#define TRACEMSS(cinfo, lvl, code, str) \
+  ((cinfo)->err->msg_code = (code), \
+   strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr)(cinfo), (lvl)))
+
+#endif /* JERROR_H */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jfdctflt.c Added

@@ -0,0 +1,169 @@
+/*
+ * jfdctflt.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a floating-point implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * This implementation should be more accurate than either of the integer
+ * DCT implementations.  However, it may not give the same results on all
+ * machines because of differences in roundoff behavior.  Speed will depend
+ * on the hardware's floating point capacity.
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with a fixed-point
+ * implementation, accuracy is lost due to imprecise representation of the
+ * scaled quantization values.  However, that problem does not arise if
+ * we use floating point arithmetic.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"               /* Private declarations for DCT subsystem */
+
+#ifdef DCT_FLOAT_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_float(FAST_FLOAT *data)
+{
+  FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
+  FAST_FLOAT z1, z2, z3, z4, z5, z11, z13;
+  FAST_FLOAT *dataptr;
+  int ctr;
+
+  /* Pass 1: process rows. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE - 1; ctr >= 0; ctr--) {
+    tmp0 = dataptr0 + dataptr7;
+    tmp7 = dataptr0 - dataptr7;
+    tmp1 = dataptr1 + dataptr6;
+    tmp6 = dataptr1 - dataptr6;
+    tmp2 = dataptr2 + dataptr5;
+    tmp5 = dataptr2 - dataptr5;
+    tmp3 = dataptr3 + dataptr4;
+    tmp4 = dataptr3 - dataptr4;
+
+    /* Even part */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+
+    dataptr0 = tmp10 + tmp11; /* phase 3 */
+    dataptr4 = tmp10 - tmp11;
+
+    z1 = (tmp12 + tmp13) * ((FAST_FLOAT)0.707106781); /* c4 */
+    dataptr2 = tmp13 + z1;    /* phase 5 */
+    dataptr6 = tmp13 - z1;
+
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = (tmp10 - tmp12) * ((FAST_FLOAT)0.382683433); /* c6 */
+    z2 = ((FAST_FLOAT)0.541196100) * tmp10 + z5; /* c2-c6 */
+    z4 = ((FAST_FLOAT)1.306562965) * tmp12 + z5; /* c2+c6 */
+    z3 = tmp11 * ((FAST_FLOAT)0.707106781); /* c4 */
+
+    z11 = tmp7 + z3;            /* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr5 = z13 + z2;      /* phase 6 */
+    dataptr3 = z13 - z2;
+    dataptr1 = z11 + z4;
+    dataptr7 = z11 - z4;
+
+    dataptr += DCTSIZE;         /* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE - 1; ctr >= 0; ctr--) {
+    tmp0 = dataptrDCTSIZE * 0 + dataptrDCTSIZE * 7;
+    tmp7 = dataptrDCTSIZE * 0 - dataptrDCTSIZE * 7;
+    tmp1 = dataptrDCTSIZE * 1 + dataptrDCTSIZE * 6;
+    tmp6 = dataptrDCTSIZE * 1 - dataptrDCTSIZE * 6;
+    tmp2 = dataptrDCTSIZE * 2 + dataptrDCTSIZE * 5;
+    tmp5 = dataptrDCTSIZE * 2 - dataptrDCTSIZE * 5;
+    tmp3 = dataptrDCTSIZE * 3 + dataptrDCTSIZE * 4;
+    tmp4 = dataptrDCTSIZE * 3 - dataptrDCTSIZE * 4;
+
+    /* Even part */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+
+    dataptrDCTSIZE * 0 = tmp10 + tmp11; /* phase 3 */
+    dataptrDCTSIZE * 4 = tmp10 - tmp11;
+
+    z1 = (tmp12 + tmp13) * ((FAST_FLOAT)0.707106781); /* c4 */
+    dataptrDCTSIZE * 2 = tmp13 + z1; /* phase 5 */
+    dataptrDCTSIZE * 6 = tmp13 - z1;
+
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = (tmp10 - tmp12) * ((FAST_FLOAT)0.382683433); /* c6 */
+    z2 = ((FAST_FLOAT)0.541196100) * tmp10 + z5; /* c2-c6 */
+    z4 = ((FAST_FLOAT)1.306562965) * tmp12 + z5; /* c2+c6 */
+    z3 = tmp11 * ((FAST_FLOAT)0.707106781); /* c4 */
+
+    z11 = tmp7 + z3;            /* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptrDCTSIZE * 5 = z13 + z2; /* phase 6 */
+    dataptrDCTSIZE * 3 = z13 - z2;
+    dataptrDCTSIZE * 1 = z11 + z4;
+    dataptrDCTSIZE * 7 = z11 - z4;
+
+    dataptr++;                  /* advance pointer to next column */
+  }
+}
+
+#endif /* DCT_FLOAT_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jfdctfst.c Added

@@ -0,0 +1,227 @@
+/*
+ * jfdctfst.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a fast, not so accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with fixed-point math,
+ * accuracy is lost due to imprecise representation of the scaled
+ * quantization values.  The smaller the quantization table entry, the less
+ * precise the scaled value, so this implementation does worse with high-
+ * quality-setting files than with low-quality ones.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"               /* Private declarations for DCT subsystem */
+
+#ifdef DCT_IFAST_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling decisions are generally the same as in the LL&M algorithm;
+ * see jfdctint.c for more details.  However, we choose to descale
+ * (right shift) multiplication products as soon as they are formed,
+ * rather than carrying additional fractional bits into subsequent additions.
+ * This compromises accuracy slightly, but it lets us save a few shifts.
+ * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples)
+ * everywhere except in the multiplications proper; this saves a good deal
+ * of work on 16-bit-int machines.
+ *
+ * Again to save a few shifts, the intermediate results between pass 1 and
+ * pass 2 are not upscaled, but are represented only to integral precision.
+ *
+ * A final compromise is to represent the multiplicative constants to only
+ * 8 fractional bits, rather than 13.  This saves some shifting work on some
+ * machines, and may also reduce the cost of multiplication (since there
+ * are fewer one-bits in the constants).
+ */
+
+#define CONST_BITS  8
+
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 8
+#define FIX_0_382683433  ((JLONG)98)            /* FIX(0.382683433) */
+#define FIX_0_541196100  ((JLONG)139)           /* FIX(0.541196100) */
+#define FIX_0_707106781  ((JLONG)181)           /* FIX(0.707106781) */
+#define FIX_1_306562965  ((JLONG)334)           /* FIX(1.306562965) */
+#else
+#define FIX_0_382683433  FIX(0.382683433)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_707106781  FIX(0.707106781)
+#define FIX_1_306562965  FIX(1.306562965)
+#endif
+
+
+/* We can gain a little more speed, with a further compromise in accuracy,
+ * by omitting the addition in a descaling shift.  This yields an incorrectly
+ * rounded result half the time...
+ */
+
+#ifndef USE_ACCURATE_ROUNDING
+#undef DESCALE
+#define DESCALE(x, n)  RIGHT_SHIFT(x, n)
+#endif
+
+
+/* Multiply a DCTELEM variable by an JLONG constant, and immediately
+ * descale to yield a DCTELEM result.
+ */
+
+#define MULTIPLY(var, const)  ((DCTELEM)DESCALE((var) * (const), CONST_BITS))
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_ifast(DCTELEM *data)
+{
+  DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  DCTELEM tmp10, tmp11, tmp12, tmp13;
+  DCTELEM z1, z2, z3, z4, z5, z11, z13;
+  DCTELEM *dataptr;
+  int ctr;
+  SHIFT_TEMPS
+
+  /* Pass 1: process rows. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE - 1; ctr >= 0; ctr--) {
+    tmp0 = dataptr0 + dataptr7;
+    tmp7 = dataptr0 - dataptr7;
+    tmp1 = dataptr1 + dataptr6;
+    tmp6 = dataptr1 - dataptr6;
+    tmp2 = dataptr2 + dataptr5;
+    tmp5 = dataptr2 - dataptr5;
+    tmp3 = dataptr3 + dataptr4;
+    tmp4 = dataptr3 - dataptr4;
+
+    /* Even part */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+
+    dataptr0 = tmp10 + tmp11; /* phase 3 */
+    dataptr4 = tmp10 - tmp11;
+
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
+    dataptr2 = tmp13 + z1;    /* phase 5 */
+    dataptr6 = tmp13 - z1;
+
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
+    z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
+    z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
+    z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
+
+    z11 = tmp7 + z3;            /* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr5 = z13 + z2;      /* phase 6 */
+    dataptr3 = z13 - z2;
+    dataptr1 = z11 + z4;
+    dataptr7 = z11 - z4;
+
+    dataptr += DCTSIZE;         /* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE - 1; ctr >= 0; ctr--) {
+    tmp0 = dataptrDCTSIZE * 0 + dataptrDCTSIZE * 7;
+    tmp7 = dataptrDCTSIZE * 0 - dataptrDCTSIZE * 7;
+    tmp1 = dataptrDCTSIZE * 1 + dataptrDCTSIZE * 6;
+    tmp6 = dataptrDCTSIZE * 1 - dataptrDCTSIZE * 6;
+    tmp2 = dataptrDCTSIZE * 2 + dataptrDCTSIZE * 5;
+    tmp5 = dataptrDCTSIZE * 2 - dataptrDCTSIZE * 5;
+    tmp3 = dataptrDCTSIZE * 3 + dataptrDCTSIZE * 4;
+    tmp4 = dataptrDCTSIZE * 3 - dataptrDCTSIZE * 4;
+
+    /* Even part */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+
+    dataptrDCTSIZE * 0 = tmp10 + tmp11; /* phase 3 */
+    dataptrDCTSIZE * 4 = tmp10 - tmp11;
+
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
+    dataptrDCTSIZE * 2 = tmp13 + z1; /* phase 5 */
+    dataptrDCTSIZE * 6 = tmp13 - z1;
+
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
+    z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
+    z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
+    z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
+
+    z11 = tmp7 + z3;            /* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptrDCTSIZE * 5 = z13 + z2; /* phase 6 */
+    dataptrDCTSIZE * 3 = z13 - z2;
+    dataptrDCTSIZE * 1 = z11 + z4;
+    dataptrDCTSIZE * 7 = z11 - z4;
+
+    dataptr++;                  /* advance pointer to next column */
+  }
+}
+
+#endif /* DCT_IFAST_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jfdctint.c Added

@@ -0,0 +1,288 @@
+/*
+ * jfdctint.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a slower but more accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on an algorithm described in
+ *   C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT
+ *   Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics,
+ *   Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991.
+ * The primary algorithm described there uses 11 multiplies and 29 adds.
+ * We use their alternate method with 12 multiplies and 32 adds.
+ * The advantage of this method is that no data path contains more than one
+ * multiplication; this allows a very simple and accurate implementation in
+ * scaled fixed-point arithmetic, with a minimal number of shifts.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"               /* Private declarations for DCT subsystem */
+
+#ifdef DCT_ISLOW_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * The poop on this scaling stuff is as follows:
+ *
+ * Each 1-D DCT step produces outputs which are a factor of sqrt(N)
+ * larger than the true DCT outputs.  The final outputs are therefore
+ * a factor of N larger than desired; since N=8 this can be cured by
+ * a simple right shift at the end of the algorithm.  The advantage of
+ * this arrangement is that we save two multiplications per 1-D DCT,
+ * because the y0 and y4 outputs need not be divided by sqrt(N).
+ * In the IJG code, this factor of 8 is removed by the quantization step
+ * (in jcdctmgr.c), NOT in this module.
+ *
+ * We have to do addition and subtraction of the integer inputs, which
+ * is no problem, and multiplication by fractional constants, which is
+ * a problem to do in integer arithmetic.  We multiply all the constants
+ * by CONST_SCALE and convert them to integer constants (thus retaining
+ * CONST_BITS bits of precision in the constants).  After doing a
+ * multiplication we have to divide the product by CONST_SCALE, with proper
+ * rounding, to produce the correct output.  This division can be done
+ * cheaply as a right shift of CONST_BITS bits.  We postpone shifting
+ * as long as possible so that partial sums can be added together with
+ * full fractional precision.
+ *
+ * The outputs of the first pass are scaled up by PASS1_BITS bits so that
+ * they are represented to better-than-integral precision.  These outputs
+ * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
+ * with the recommended scaling.  (For 12-bit sample data, the intermediate
+ * array is JLONG anyway.)
+ *
+ * To avoid overflow of the 32-bit intermediate results in pass 2, we must
+ * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
+ * shows that the values given below are the most effective.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  13
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 13
+#define FIX_0_298631336  ((JLONG)2446)          /* FIX(0.298631336) */
+#define FIX_0_390180644  ((JLONG)3196)          /* FIX(0.390180644) */
+#define FIX_0_541196100  ((JLONG)4433)          /* FIX(0.541196100) */
+#define FIX_0_765366865  ((JLONG)6270)          /* FIX(0.765366865) */
+#define FIX_0_899976223  ((JLONG)7373)          /* FIX(0.899976223) */
+#define FIX_1_175875602  ((JLONG)9633)          /* FIX(1.175875602) */
+#define FIX_1_501321110  ((JLONG)12299)         /* FIX(1.501321110) */
+#define FIX_1_847759065  ((JLONG)15137)         /* FIX(1.847759065) */
+#define FIX_1_961570560  ((JLONG)16069)         /* FIX(1.961570560) */
+#define FIX_2_053119869  ((JLONG)16819)         /* FIX(2.053119869) */
+#define FIX_2_562915447  ((JLONG)20995)         /* FIX(2.562915447) */
+#define FIX_3_072711026  ((JLONG)25172)         /* FIX(3.072711026) */
+#else
+#define FIX_0_298631336  FIX(0.298631336)
+#define FIX_0_390180644  FIX(0.390180644)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_765366865  FIX(0.765366865)
+#define FIX_0_899976223  FIX(0.899976223)
+#define FIX_1_175875602  FIX(1.175875602)
+#define FIX_1_501321110  FIX(1.501321110)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_1_961570560  FIX(1.961570560)
+#define FIX_2_053119869  FIX(2.053119869)
+#define FIX_2_562915447  FIX(2.562915447)
+#define FIX_3_072711026  FIX(3.072711026)
+#endif
+
+
+/* Multiply an JLONG variable by an JLONG constant to yield an JLONG result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
+ * For 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MULTIPLY(var, const)  MULTIPLY16C16(var, const)
+#else
+#define MULTIPLY(var, const)  ((var) * (const))
+#endif
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_islow(DCTELEM *data)
+{
+  JLONG tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  JLONG tmp10, tmp11, tmp12, tmp13;
+  JLONG z1, z2, z3, z4, z5;
+  DCTELEM *dataptr;
+  int ctr;
+  SHIFT_TEMPS
+
+  /* Pass 1: process rows. */
+  /* Note results are scaled up by sqrt(8) compared to a true DCT; */
+  /* furthermore, we scale the results by 2**PASS1_BITS. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE - 1; ctr >= 0; ctr--) {
+    tmp0 = dataptr0 + dataptr7;
+    tmp7 = dataptr0 - dataptr7;
+    tmp1 = dataptr1 + dataptr6;
+    tmp6 = dataptr1 - dataptr6;
+    tmp2 = dataptr2 + dataptr5;
+    tmp5 = dataptr2 - dataptr5;
+    tmp3 = dataptr3 + dataptr4;
+    tmp4 = dataptr3 - dataptr4;
+
+    /* Even part per LL&M figure 1 --- note that published figure is faulty;
+     * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
+     */
+
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+
+    dataptr0 = (DCTELEM)LEFT_SHIFT(tmp10 + tmp11, PASS1_BITS);
+    dataptr4 = (DCTELEM)LEFT_SHIFT(tmp10 - tmp11, PASS1_BITS);
+
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
+    dataptr2 = (DCTELEM)DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
+                                  CONST_BITS - PASS1_BITS);
+    dataptr6 = (DCTELEM)DESCALE(z1 + MULTIPLY(tmp12, -FIX_1_847759065),
+                                  CONST_BITS - PASS1_BITS);
+
+    /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
+     * cK represents cos(K*pi/16).
+     * i0..i3 in the paper are tmp4..tmp7 here.
+     */
+
+    z1 = tmp4 + tmp7;
+    z2 = tmp5 + tmp6;
+    z3 = tmp4 + tmp6;
+    z4 = tmp5 + tmp7;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+
+    tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, -FIX_0_899976223); /* sqrt(2) * ( c7-c3) */
+    z2 = MULTIPLY(z2, -FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, -FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, -FIX_0_390180644); /* sqrt(2) * ( c5-c3) */
+
+    z3 += z5;
+    z4 += z5;
+
+    dataptr7 = (DCTELEM)DESCALE(tmp4 + z1 + z3, CONST_BITS - PASS1_BITS);
+    dataptr5 = (DCTELEM)DESCALE(tmp5 + z2 + z4, CONST_BITS - PASS1_BITS);
+    dataptr3 = (DCTELEM)DESCALE(tmp6 + z2 + z3, CONST_BITS - PASS1_BITS);
+    dataptr1 = (DCTELEM)DESCALE(tmp7 + z1 + z4, CONST_BITS - PASS1_BITS);
+
+    dataptr += DCTSIZE;         /* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns.
+   * We remove the PASS1_BITS scaling, but leave the results scaled up
+   * by an overall factor of 8.
+   */
+
+  dataptr = data;
+  for (ctr = DCTSIZE - 1; ctr >= 0; ctr--) {
+    tmp0 = dataptrDCTSIZE * 0 + dataptrDCTSIZE * 7;
+    tmp7 = dataptrDCTSIZE * 0 - dataptrDCTSIZE * 7;
+    tmp1 = dataptrDCTSIZE * 1 + dataptrDCTSIZE * 6;
+    tmp6 = dataptrDCTSIZE * 1 - dataptrDCTSIZE * 6;
+    tmp2 = dataptrDCTSIZE * 2 + dataptrDCTSIZE * 5;
+    tmp5 = dataptrDCTSIZE * 2 - dataptrDCTSIZE * 5;
+    tmp3 = dataptrDCTSIZE * 3 + dataptrDCTSIZE * 4;
+    tmp4 = dataptrDCTSIZE * 3 - dataptrDCTSIZE * 4;
+
+    /* Even part per LL&M figure 1 --- note that published figure is faulty;
+     * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
+     */
+
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+
+    dataptrDCTSIZE * 0 = (DCTELEM)DESCALE(tmp10 + tmp11, PASS1_BITS);
+    dataptrDCTSIZE * 4 = (DCTELEM)DESCALE(tmp10 - tmp11, PASS1_BITS);
+
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
+    dataptrDCTSIZE * 2 =
+      (DCTELEM)DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
+                       CONST_BITS + PASS1_BITS);
+    dataptrDCTSIZE * 6 =
+      (DCTELEM)DESCALE(z1 + MULTIPLY(tmp12, -FIX_1_847759065),
+                       CONST_BITS + PASS1_BITS);
+
+    /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
+     * cK represents cos(K*pi/16).
+     * i0..i3 in the paper are tmp4..tmp7 here.
+     */
+
+    z1 = tmp4 + tmp7;
+    z2 = tmp5 + tmp6;
+    z3 = tmp4 + tmp6;
+    z4 = tmp5 + tmp7;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+
+    tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, -FIX_0_899976223); /* sqrt(2) * ( c7-c3) */
+    z2 = MULTIPLY(z2, -FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, -FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, -FIX_0_390180644); /* sqrt(2) * ( c5-c3) */
+
+    z3 += z5;
+    z4 += z5;
+
+    dataptrDCTSIZE * 7 = (DCTELEM)DESCALE(tmp4 + z1 + z3,
+                                            CONST_BITS + PASS1_BITS);
+    dataptrDCTSIZE * 5 = (DCTELEM)DESCALE(tmp5 + z2 + z4,
+                                            CONST_BITS + PASS1_BITS);
+    dataptrDCTSIZE * 3 = (DCTELEM)DESCALE(tmp6 + z2 + z3,
+                                            CONST_BITS + PASS1_BITS);
+    dataptrDCTSIZE * 1 = (DCTELEM)DESCALE(tmp7 + z1 + z4,
+                                            CONST_BITS + PASS1_BITS);
+
+    dataptr++;                  /* advance pointer to next column */
+  }
+}
+
+#endif /* DCT_ISLOW_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jidctflt.c Added

@@ -0,0 +1,240 @@
+/*
+ * jidctflt.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * Modified 2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a floating-point implementation of the
+ * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
+ * must also perform dequantization of the input coefficients.
+ *
+ * This implementation should be more accurate than either of the integer
+ * IDCT implementations.  However, it may not give the same results on all
+ * machines because of differences in roundoff behavior.  Speed will depend
+ * on the hardware's floating point capacity.
+ *
+ * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
+ * on each row (or vice versa, but it's more convenient to emit a row at
+ * a time).  Direct algorithms are also available, but they are much more
+ * complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with a fixed-point
+ * implementation, accuracy is lost due to imprecise representation of the
+ * scaled quantization values.  However, that problem does not arise if
+ * we use floating point arithmetic.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"               /* Private declarations for DCT subsystem */
+
+#ifdef DCT_FLOAT_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce a float result.
+ */
+
+#define DEQUANTIZE(coef, quantval)  (((FAST_FLOAT)(coef)) * (quantval))
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ */
+
+GLOBAL(void)
+jpeg_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
+  FAST_FLOAT z5, z10, z11, z12, z13;
+  JCOEFPTR inptr;
+  FLOAT_MULT_TYPE *quantptr;
+  FAST_FLOAT *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int ctr;
+  FAST_FLOAT workspaceDCTSIZE2; /* buffers data between passes */
+#define _0_125  ((FLOAT_MULT_TYPE)0.125)
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (FLOAT_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any column in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * column DCT calculations can be simplified this way.
+     */
+
+    if (inptrDCTSIZE * 1 == 0 && inptrDCTSIZE * 2 == 0 &&
+        inptrDCTSIZE * 3 == 0 && inptrDCTSIZE * 4 == 0 &&
+        inptrDCTSIZE * 5 == 0 && inptrDCTSIZE * 6 == 0 &&
+        inptrDCTSIZE * 7 == 0) {
+      /* AC terms all zero */
+      FAST_FLOAT dcval = DEQUANTIZE(inptrDCTSIZE * 0,
+                                    quantptrDCTSIZE * 0 * _0_125);
+
+      wsptrDCTSIZE * 0 = dcval;
+      wsptrDCTSIZE * 1 = dcval;
+      wsptrDCTSIZE * 2 = dcval;
+      wsptrDCTSIZE * 3 = dcval;
+      wsptrDCTSIZE * 4 = dcval;
+      wsptrDCTSIZE * 5 = dcval;
+      wsptrDCTSIZE * 6 = dcval;
+      wsptrDCTSIZE * 7 = dcval;
+
+      inptr++;                  /* advance pointers to next column */
+      quantptr++;
+      wsptr++;
+      continue;
+    }
+
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0 * _0_125);
+    tmp1 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2 * _0_125);
+    tmp2 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4 * _0_125);
+    tmp3 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6 * _0_125);
+
+    tmp10 = tmp0 + tmp2;        /* phase 3 */
+    tmp11 = tmp0 - tmp2;
+
+    tmp13 = tmp1 + tmp3;        /* phases 5-3 */
+    tmp12 = (tmp1 - tmp3) * ((FAST_FLOAT)1.414213562) - tmp13; /* 2*c4 */
+
+    tmp0 = tmp10 + tmp13;       /* phase 2 */
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    tmp4 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1 * _0_125);
+    tmp5 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3 * _0_125);
+    tmp6 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5 * _0_125);
+    tmp7 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7 * _0_125);
+
+    z13 = tmp6 + tmp5;          /* phase 6 */
+    z10 = tmp6 - tmp5;
+    z11 = tmp4 + tmp7;
+    z12 = tmp4 - tmp7;
+
+    tmp7 = z11 + z13;           /* phase 5 */
+    tmp11 = (z11 - z13) * ((FAST_FLOAT)1.414213562); /* 2*c4 */
+
+    z5 = (z10 + z12) * ((FAST_FLOAT)1.847759065); /* 2*c2 */
+    tmp10 = z5 - z12 * ((FAST_FLOAT)1.082392200); /* 2*(c2-c6) */
+    tmp12 = z5 - z10 * ((FAST_FLOAT)2.613125930); /* 2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;        /* phase 2 */
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 - tmp5;
+
+    wsptrDCTSIZE * 0 = tmp0 + tmp7;
+    wsptrDCTSIZE * 7 = tmp0 - tmp7;
+    wsptrDCTSIZE * 1 = tmp1 + tmp6;
+    wsptrDCTSIZE * 6 = tmp1 - tmp6;
+    wsptrDCTSIZE * 2 = tmp2 + tmp5;
+    wsptrDCTSIZE * 5 = tmp2 - tmp5;
+    wsptrDCTSIZE * 3 = tmp3 + tmp4;
+    wsptrDCTSIZE * 4 = tmp3 - tmp4;
+
+    inptr++;                    /* advance pointers to next column */
+    quantptr++;
+    wsptr++;
+  }
+
+  /* Pass 2: process rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_bufctr + output_col;
+    /* Rows of zeroes can be exploited in the same way as we did with columns.
+     * However, the column calculation has created many nonzero AC terms, so
+     * the simplification applies less often (typically 5% to 10% of the time).
+     * And testing floats for zero is relatively expensive, so we don't bother.
+     */
+
+    /* Even part */
+
+    /* Apply signed->unsigned and prepare float->int conversion */
+    z5 = wsptr0 + ((FAST_FLOAT)CENTERJSAMPLE + (FAST_FLOAT)0.5);
+    tmp10 = z5 + wsptr4;
+    tmp11 = z5 - wsptr4;
+
+    tmp13 = wsptr2 + wsptr6;
+    tmp12 = (wsptr2 - wsptr6) * ((FAST_FLOAT)1.414213562) - tmp13;
+
+    tmp0 = tmp10 + tmp13;
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z13 = wsptr5 + wsptr3;
+    z10 = wsptr5 - wsptr3;
+    z11 = wsptr1 + wsptr7;
+    z12 = wsptr1 - wsptr7;
+
+    tmp7 = z11 + z13;
+    tmp11 = (z11 - z13) * ((FAST_FLOAT)1.414213562);
+
+    z5 = (z10 + z12) * ((FAST_FLOAT)1.847759065); /* 2*c2 */
+    tmp10 = z5 - z12 * ((FAST_FLOAT)1.082392200); /* 2*(c2-c6) */
+    tmp12 = z5 - z10 * ((FAST_FLOAT)2.613125930); /* 2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 - tmp5;
+
+    /* Final output stage: float->int conversion and range-limit */
+
+    outptr0 = range_limit((int)(tmp0 + tmp7)) & RANGE_MASK;
+    outptr7 = range_limit((int)(tmp0 - tmp7)) & RANGE_MASK;
+    outptr1 = range_limit((int)(tmp1 + tmp6)) & RANGE_MASK;
+    outptr6 = range_limit((int)(tmp1 - tmp6)) & RANGE_MASK;
+    outptr2 = range_limit((int)(tmp2 + tmp5)) & RANGE_MASK;
+    outptr5 = range_limit((int)(tmp2 - tmp5)) & RANGE_MASK;
+    outptr3 = range_limit((int)(tmp3 + tmp4)) & RANGE_MASK;
+    outptr4 = range_limit((int)(tmp3 - tmp4)) & RANGE_MASK;
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
+  }
+}
+
+#endif /* DCT_FLOAT_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jidctfst.c Added

@@ -0,0 +1,371 @@
+/*
+ * jidctfst.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a fast, not so accurate integer implementation of the
+ * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
+ * must also perform dequantization of the input coefficients.
+ *
+ * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
+ * on each row (or vice versa, but it's more convenient to emit a row at
+ * a time).  Direct algorithms are also available, but they are much more
+ * complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with fixed-point math,
+ * accuracy is lost due to imprecise representation of the scaled
+ * quantization values.  The smaller the quantization table entry, the less
+ * precise the scaled value, so this implementation does worse with high-
+ * quality-setting files than with low-quality ones.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"               /* Private declarations for DCT subsystem */
+
+#ifdef DCT_IFAST_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling decisions are generally the same as in the LL&M algorithm;
+ * see jidctint.c for more details.  However, we choose to descale
+ * (right shift) multiplication products as soon as they are formed,
+ * rather than carrying additional fractional bits into subsequent additions.
+ * This compromises accuracy slightly, but it lets us save a few shifts.
+ * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples)
+ * everywhere except in the multiplications proper; this saves a good deal
+ * of work on 16-bit-int machines.
+ *
+ * The dequantized coefficients are not integers because the AA&N scaling
+ * factors have been incorporated.  We represent them scaled up by PASS1_BITS,
+ * so that the first and second IDCT rounds have the same input scaling.
+ * For 8-bit JSAMPLEs, we choose IFAST_SCALE_BITS = PASS1_BITS so as to
+ * avoid a descaling shift; this compromises accuracy rather drastically
+ * for small quantization table entries, but it saves a lot of shifts.
+ * For 12-bit JSAMPLEs, there's no hope of using 16x16 multiplies anyway,
+ * so we use a much larger scaling factor to preserve accuracy.
+ *
+ * A final compromise is to represent the multiplicative constants to only
+ * 8 fractional bits, rather than 13.  This saves some shifting work on some
+ * machines, and may also reduce the cost of multiplication (since there
+ * are fewer one-bits in the constants).
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  8
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  8
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 8
+#define FIX_1_082392200  ((JLONG)277)           /* FIX(1.082392200) */
+#define FIX_1_414213562  ((JLONG)362)           /* FIX(1.414213562) */
+#define FIX_1_847759065  ((JLONG)473)           /* FIX(1.847759065) */
+#define FIX_2_613125930  ((JLONG)669)           /* FIX(2.613125930) */
+#else
+#define FIX_1_082392200  FIX(1.082392200)
+#define FIX_1_414213562  FIX(1.414213562)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_2_613125930  FIX(2.613125930)
+#endif
+
+
+/* We can gain a little more speed, with a further compromise in accuracy,
+ * by omitting the addition in a descaling shift.  This yields an incorrectly
+ * rounded result half the time...
+ */
+
+#ifndef USE_ACCURATE_ROUNDING
+#undef DESCALE
+#define DESCALE(x, n)  RIGHT_SHIFT(x, n)
+#endif
+
+
+/* Multiply a DCTELEM variable by an JLONG constant, and immediately
+ * descale to yield a DCTELEM result.
+ */
+
+#define MULTIPLY(var, const)  ((DCTELEM)DESCALE((var) * (const), CONST_BITS))
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce a DCTELEM result.  For 8-bit data a 16x16->16
+ * multiplication will do.  For 12-bit data, the multiplier table is
+ * declared JLONG, so a 32-bit multiply will be used.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define DEQUANTIZE(coef, quantval)  (((IFAST_MULT_TYPE)(coef)) * (quantval))
+#else
+#define DEQUANTIZE(coef, quantval) \
+  DESCALE((coef) * (quantval), IFAST_SCALE_BITS - PASS1_BITS)
+#endif
+
+
+/* Like DESCALE, but applies to a DCTELEM and produces an int.
+ * We assume that int right shift is unsigned if JLONG right shift is.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define ISHIFT_TEMPS    DCTELEM ishift_temp;
+#if BITS_IN_JSAMPLE == 8
+#define DCTELEMBITS  16         /* DCTELEM may be 16 or 32 bits */
+#else
+#define DCTELEMBITS  32         /* DCTELEM must be 32 bits */
+#endif
+#define IRIGHT_SHIFT(x, shft) \
+  ((ishift_temp = (x)) < 0 ? \
+   (ishift_temp >> (shft)) | ((~((DCTELEM)0)) << (DCTELEMBITS - (shft))) : \
+   (ishift_temp >> (shft)))
+#else
+#define ISHIFT_TEMPS
+#define IRIGHT_SHIFT(x, shft)   ((x) >> (shft))
+#endif
+
+#ifdef USE_ACCURATE_ROUNDING
+#define IDESCALE(x, n)  ((int)IRIGHT_SHIFT((x) + (1 << ((n) - 1)), n))
+#else
+#define IDESCALE(x, n)  ((int)IRIGHT_SHIFT(x, n))
+#endif
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ */
+
+GLOBAL(void)
+jpeg_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  DCTELEM tmp10, tmp11, tmp12, tmp13;
+  DCTELEM z5, z10, z11, z12, z13;
+  JCOEFPTR inptr;
+  IFAST_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspaceDCTSIZE2;      /* buffers data between passes */
+  SHIFT_TEMPS                   /* for DESCALE */
+  ISHIFT_TEMPS                  /* for IDESCALE */
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (IFAST_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any column in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * column DCT calculations can be simplified this way.
+     */
+
+    if (inptrDCTSIZE * 1 == 0 && inptrDCTSIZE * 2 == 0 &&
+        inptrDCTSIZE * 3 == 0 && inptrDCTSIZE * 4 == 0 &&
+        inptrDCTSIZE * 5 == 0 && inptrDCTSIZE * 6 == 0 &&
+        inptrDCTSIZE * 7 == 0) {
+      /* AC terms all zero */
+      int dcval = (int)DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+
+      wsptrDCTSIZE * 0 = dcval;
+      wsptrDCTSIZE * 1 = dcval;
+      wsptrDCTSIZE * 2 = dcval;
+      wsptrDCTSIZE * 3 = dcval;
+      wsptrDCTSIZE * 4 = dcval;
+      wsptrDCTSIZE * 5 = dcval;
+      wsptrDCTSIZE * 6 = dcval;
+      wsptrDCTSIZE * 7 = dcval;
+
+      inptr++;                  /* advance pointers to next column */
+      quantptr++;
+      wsptr++;
+      continue;
+    }
+
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp1 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    tmp2 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    tmp3 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    tmp10 = tmp0 + tmp2;        /* phase 3 */
+    tmp11 = tmp0 - tmp2;
+
+    tmp13 = tmp1 + tmp3;        /* phases 5-3 */
+    tmp12 = MULTIPLY(tmp1 - tmp3, FIX_1_414213562) - tmp13; /* 2*c4 */
+
+    tmp0 = tmp10 + tmp13;       /* phase 2 */
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    tmp4 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    tmp5 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    tmp6 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    tmp7 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+
+    z13 = tmp6 + tmp5;          /* phase 6 */
+    z10 = tmp6 - tmp5;
+    z11 = tmp4 + tmp7;
+    z12 = tmp4 - tmp7;
+
+    tmp7 = z11 + z13;           /* phase 5 */
+    tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
+
+    z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
+    tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
+    tmp12 = MULTIPLY(z10, -FIX_2_613125930) + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;        /* phase 2 */
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    wsptrDCTSIZE * 0 = (int)(tmp0 + tmp7);
+    wsptrDCTSIZE * 7 = (int)(tmp0 - tmp7);
+    wsptrDCTSIZE * 1 = (int)(tmp1 + tmp6);
+    wsptrDCTSIZE * 6 = (int)(tmp1 - tmp6);
+    wsptrDCTSIZE * 2 = (int)(tmp2 + tmp5);
+    wsptrDCTSIZE * 5 = (int)(tmp2 - tmp5);
+    wsptrDCTSIZE * 4 = (int)(tmp3 + tmp4);
+    wsptrDCTSIZE * 3 = (int)(tmp3 - tmp4);
+
+    inptr++;                    /* advance pointers to next column */
+    quantptr++;
+    wsptr++;
+  }
+
+  /* Pass 2: process rows from work array, store into output array. */
+  /* Note that we must descale the results by a factor of 8 == 2**3, */
+  /* and also undo the PASS1_BITS scaling. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_bufctr + output_col;
+    /* Rows of zeroes can be exploited in the same way as we did with columns.
+     * However, the column calculation has created many nonzero AC terms, so
+     * the simplification applies less often (typically 5% to 10% of the time).
+     * On machines with very fast multiplication, it's possible that the
+     * test takes more time than it's worth.  In that case this section
+     * may be commented out.
+     */
+
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr1 == 0 && wsptr2 == 0 && wsptr3 == 0 && wsptr4 == 0 &&
+        wsptr5 == 0 && wsptr6 == 0 && wsptr7 == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval =
+        range_limitIDESCALE(wsptr0, PASS1_BITS + 3) & RANGE_MASK;
+
+      outptr0 = dcval;
+      outptr1 = dcval;
+      outptr2 = dcval;
+      outptr3 = dcval;
+      outptr4 = dcval;
+      outptr5 = dcval;
+      outptr6 = dcval;
+      outptr7 = dcval;
+
+      wsptr += DCTSIZE;         /* advance pointer to next row */
+      continue;
+    }
+#endif
+
+    /* Even part */
+
+    tmp10 = ((DCTELEM)wsptr0 + (DCTELEM)wsptr4);
+    tmp11 = ((DCTELEM)wsptr0 - (DCTELEM)wsptr4);
+
+    tmp13 = ((DCTELEM)wsptr2 + (DCTELEM)wsptr6);
+    tmp12 =
+      MULTIPLY((DCTELEM)wsptr2 - (DCTELEM)wsptr6, FIX_1_414213562) - tmp13;
+
+    tmp0 = tmp10 + tmp13;
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z13 = (DCTELEM)wsptr5 + (DCTELEM)wsptr3;
+    z10 = (DCTELEM)wsptr5 - (DCTELEM)wsptr3;
+    z11 = (DCTELEM)wsptr1 + (DCTELEM)wsptr7;
+    z12 = (DCTELEM)wsptr1 - (DCTELEM)wsptr7;
+
+    tmp7 = z11 + z13;           /* phase 5 */
+    tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
+
+    z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
+    tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
+    tmp12 = MULTIPLY(z10, -FIX_2_613125930) + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;        /* phase 2 */
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    /* Final output stage: scale down by a factor of 8 and range-limit */
+
+    outptr0 =
+      range_limitIDESCALE(tmp0 + tmp7, PASS1_BITS + 3) & RANGE_MASK;
+    outptr7 =
+      range_limitIDESCALE(tmp0 - tmp7, PASS1_BITS + 3) & RANGE_MASK;
+    outptr1 =
+      range_limitIDESCALE(tmp1 + tmp6, PASS1_BITS + 3) & RANGE_MASK;
+    outptr6 =
+      range_limitIDESCALE(tmp1 - tmp6, PASS1_BITS + 3) & RANGE_MASK;
+    outptr2 =
+      range_limitIDESCALE(tmp2 + tmp5, PASS1_BITS + 3) & RANGE_MASK;
+    outptr5 =
+      range_limitIDESCALE(tmp2 - tmp5, PASS1_BITS + 3) & RANGE_MASK;
+    outptr4 =
+      range_limitIDESCALE(tmp3 + tmp4, PASS1_BITS + 3) & RANGE_MASK;
+    outptr3 =
+      range_limitIDESCALE(tmp3 - tmp4, PASS1_BITS + 3) & RANGE_MASK;
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
+  }
+}
+
+#endif /* DCT_IFAST_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jidctint.c Added

@@ -0,0 +1,2627 @@
+/*
+ * jidctint.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * Modification developed 2002-2018 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a slower but more accurate integer implementation of the
+ * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
+ * must also perform dequantization of the input coefficients.
+ *
+ * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
+ * on each row (or vice versa, but it's more convenient to emit a row at
+ * a time).  Direct algorithms are also available, but they are much more
+ * complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on an algorithm described in
+ *   C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT
+ *   Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics,
+ *   Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991.
+ * The primary algorithm described there uses 11 multiplies and 29 adds.
+ * We use their alternate method with 12 multiplies and 32 adds.
+ * The advantage of this method is that no data path contains more than one
+ * multiplication; this allows a very simple and accurate implementation in
+ * scaled fixed-point arithmetic, with a minimal number of shifts.
+ *
+ * We also provide IDCT routines with various output sample block sizes for
+ * direct resolution reduction or enlargement without additional resampling:
+ * NxN (N=1...16) pixels for one 8x8 input DCT block.
+ *
+ * For N<8 we simply take the corresponding low-frequency coefficients of
+ * the 8x8 input DCT block and apply an NxN point IDCT on the sub-block
+ * to yield the downscaled outputs.
+ * This can be seen as direct low-pass downsampling from the DCT domain
+ * point of view rather than the usual spatial domain point of view,
+ * yielding significant computational savings and results at least
+ * as good as common bilinear (averaging) spatial downsampling.
+ *
+ * For N>8 we apply a partial NxN IDCT on the 8 input coefficients as
+ * lower frequencies and higher frequencies assumed to be zero.
+ * It turns out that the computational effort is similar to the 8x8 IDCT
+ * regarding the output size.
+ * Furthermore, the scaling and descaling is the same for all IDCT sizes.
+ *
+ * CAUTION: We rely on the FIX() macro except for the N=1,2,4,8 cases
+ * since there would be too many additional constants to pre-calculate.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"               /* Private declarations for DCT subsystem */
+
+#ifdef DCT_ISLOW_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCT blocks. /* deliberate syntax err */
+#endif
+
+
+/*
+ * The poop on this scaling stuff is as follows:
+ *
+ * Each 1-D IDCT step produces outputs which are a factor of sqrt(N)
+ * larger than the true IDCT outputs.  The final outputs are therefore
+ * a factor of N larger than desired; since N=8 this can be cured by
+ * a simple right shift at the end of the algorithm.  The advantage of
+ * this arrangement is that we save two multiplications per 1-D IDCT,
+ * because the y0 and y4 inputs need not be divided by sqrt(N).
+ *
+ * We have to do addition and subtraction of the integer inputs, which
+ * is no problem, and multiplication by fractional constants, which is
+ * a problem to do in integer arithmetic.  We multiply all the constants
+ * by CONST_SCALE and convert them to integer constants (thus retaining
+ * CONST_BITS bits of precision in the constants).  After doing a
+ * multiplication we have to divide the product by CONST_SCALE, with proper
+ * rounding, to produce the correct output.  This division can be done
+ * cheaply as a right shift of CONST_BITS bits.  We postpone shifting
+ * as long as possible so that partial sums can be added together with
+ * full fractional precision.
+ *
+ * The outputs of the first pass are scaled up by PASS1_BITS bits so that
+ * they are represented to better-than-integral precision.  These outputs
+ * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
+ * with the recommended scaling.  (To scale up 12-bit sample data further, an
+ * intermediate JLONG array would be needed.)
+ *
+ * To avoid overflow of the 32-bit intermediate results in pass 2, we must
+ * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
+ * shows that the values given below are the most effective.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  13
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 13
+#define FIX_0_298631336  ((JLONG)2446)          /* FIX(0.298631336) */
+#define FIX_0_390180644  ((JLONG)3196)          /* FIX(0.390180644) */
+#define FIX_0_541196100  ((JLONG)4433)          /* FIX(0.541196100) */
+#define FIX_0_765366865  ((JLONG)6270)          /* FIX(0.765366865) */
+#define FIX_0_899976223  ((JLONG)7373)          /* FIX(0.899976223) */
+#define FIX_1_175875602  ((JLONG)9633)          /* FIX(1.175875602) */
+#define FIX_1_501321110  ((JLONG)12299)         /* FIX(1.501321110) */
+#define FIX_1_847759065  ((JLONG)15137)         /* FIX(1.847759065) */
+#define FIX_1_961570560  ((JLONG)16069)         /* FIX(1.961570560) */
+#define FIX_2_053119869  ((JLONG)16819)         /* FIX(2.053119869) */
+#define FIX_2_562915447  ((JLONG)20995)         /* FIX(2.562915447) */
+#define FIX_3_072711026  ((JLONG)25172)         /* FIX(3.072711026) */
+#else
+#define FIX_0_298631336  FIX(0.298631336)
+#define FIX_0_390180644  FIX(0.390180644)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_765366865  FIX(0.765366865)
+#define FIX_0_899976223  FIX(0.899976223)
+#define FIX_1_175875602  FIX(1.175875602)
+#define FIX_1_501321110  FIX(1.501321110)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_1_961570560  FIX(1.961570560)
+#define FIX_2_053119869  FIX(2.053119869)
+#define FIX_2_562915447  FIX(2.562915447)
+#define FIX_3_072711026  FIX(3.072711026)
+#endif
+
+
+/* Multiply an JLONG variable by an JLONG constant to yield an JLONG result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
+ * For 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MULTIPLY(var, const)  MULTIPLY16C16(var, const)
+#else
+#define MULTIPLY(var, const)  ((var) * (const))
+#endif
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce an int result.  In this module, both inputs and result
+ * are 16 bits or less, so either int or short multiply will work.
+ */
+
+#define DEQUANTIZE(coef, quantval)  (((ISLOW_MULT_TYPE)(coef)) * (quantval))
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ */
+
+GLOBAL(void)
+jpeg_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp3;
+  JLONG tmp10, tmp11, tmp12, tmp13;
+  JLONG z1, z2, z3, z4, z5;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspaceDCTSIZE2;      /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+  /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
+  /* furthermore, we scale the results by 2**PASS1_BITS. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any column in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * column DCT calculations can be simplified this way.
+     */
+
+    if (inptrDCTSIZE * 1 == 0 && inptrDCTSIZE * 2 == 0 &&
+        inptrDCTSIZE * 3 == 0 && inptrDCTSIZE * 4 == 0 &&
+        inptrDCTSIZE * 5 == 0 && inptrDCTSIZE * 6 == 0 &&
+        inptrDCTSIZE * 7 == 0) {
+      /* AC terms all zero */
+      int dcval = LEFT_SHIFT(DEQUANTIZE(inptrDCTSIZE * 0,
+                             quantptrDCTSIZE * 0), PASS1_BITS);
+
+      wsptrDCTSIZE * 0 = dcval;
+      wsptrDCTSIZE * 1 = dcval;
+      wsptrDCTSIZE * 2 = dcval;
+      wsptrDCTSIZE * 3 = dcval;
+      wsptrDCTSIZE * 4 = dcval;
+      wsptrDCTSIZE * 5 = dcval;
+      wsptrDCTSIZE * 6 = dcval;
+      wsptrDCTSIZE * 7 = dcval;
+
+      inptr++;                  /* advance pointers to next column */
+      quantptr++;
+      wsptr++;
+      continue;
+    }
+
+    /* Even part: reverse the even part of the forward DCT. */
+    /* The rotator is sqrt(2)*c(-6). */
+
+    z2 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
+    tmp2 = z1 + MULTIPLY(z3, -FIX_1_847759065);
+    tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
+
+    z2 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+
+    tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS);
+    tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS);
+
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    tmp0 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+    tmp1 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    tmp2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    tmp3 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+
+    z1 = tmp0 + tmp3;
+    z2 = tmp1 + tmp2;
+    z3 = tmp0 + tmp2;
+    z4 = tmp1 + tmp3;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+
+    tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, -FIX_0_899976223); /* sqrt(2) * ( c7-c3) */
+    z2 = MULTIPLY(z2, -FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, -FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, -FIX_0_390180644); /* sqrt(2) * ( c5-c3) */
+
+    z3 += z5;
+    z4 += z5;
+
+    tmp0 += z1 + z3;
+    tmp1 += z2 + z4;
+    tmp2 += z2 + z3;
+    tmp3 += z1 + z4;
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    wsptrDCTSIZE * 0 = (int)DESCALE(tmp10 + tmp3, CONST_BITS - PASS1_BITS);
+    wsptrDCTSIZE * 7 = (int)DESCALE(tmp10 - tmp3, CONST_BITS - PASS1_BITS);
+    wsptrDCTSIZE * 1 = (int)DESCALE(tmp11 + tmp2, CONST_BITS - PASS1_BITS);
+    wsptrDCTSIZE * 6 = (int)DESCALE(tmp11 - tmp2, CONST_BITS - PASS1_BITS);
+    wsptrDCTSIZE * 2 = (int)DESCALE(tmp12 + tmp1, CONST_BITS - PASS1_BITS);
+    wsptrDCTSIZE * 5 = (int)DESCALE(tmp12 - tmp1, CONST_BITS - PASS1_BITS);
+    wsptrDCTSIZE * 3 = (int)DESCALE(tmp13 + tmp0, CONST_BITS - PASS1_BITS);
+    wsptrDCTSIZE * 4 = (int)DESCALE(tmp13 - tmp0, CONST_BITS - PASS1_BITS);
+
+    inptr++;                    /* advance pointers to next column */
+    quantptr++;
+    wsptr++;
+  }
+
+  /* Pass 2: process rows from work array, store into output array. */
+  /* Note that we must descale the results by a factor of 8 == 2**3, */
+  /* and also undo the PASS1_BITS scaling. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_bufctr + output_col;
+    /* Rows of zeroes can be exploited in the same way as we did with columns.
+     * However, the column calculation has created many nonzero AC terms, so
+     * the simplification applies less often (typically 5% to 10% of the time).
+     * On machines with very fast multiplication, it's possible that the
+     * test takes more time than it's worth.  In that case this section
+     * may be commented out.
+     */
+
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr1 == 0 && wsptr2 == 0 && wsptr3 == 0 && wsptr4 == 0 &&
+        wsptr5 == 0 && wsptr6 == 0 && wsptr7 == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit(int)DESCALE((JLONG)wsptr0,
+                                               PASS1_BITS + 3) & RANGE_MASK;
+
+      outptr0 = dcval;
+      outptr1 = dcval;
+      outptr2 = dcval;
+      outptr3 = dcval;
+      outptr4 = dcval;
+      outptr5 = dcval;
+      outptr6 = dcval;
+      outptr7 = dcval;
+
+      wsptr += DCTSIZE;         /* advance pointer to next row */
+      continue;
+    }
+#endif
+
+    /* Even part: reverse the even part of the forward DCT. */
+    /* The rotator is sqrt(2)*c(-6). */
+
+    z2 = (JLONG)wsptr2;
+    z3 = (JLONG)wsptr6;
+
+    z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
+    tmp2 = z1 + MULTIPLY(z3, -FIX_1_847759065);
+    tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
+
+    tmp0 = LEFT_SHIFT((JLONG)wsptr0 + (JLONG)wsptr4, CONST_BITS);
+    tmp1 = LEFT_SHIFT((JLONG)wsptr0 - (JLONG)wsptr4, CONST_BITS);
+
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    tmp0 = (JLONG)wsptr7;
+    tmp1 = (JLONG)wsptr5;
+    tmp2 = (JLONG)wsptr3;
+    tmp3 = (JLONG)wsptr1;
+
+    z1 = tmp0 + tmp3;
+    z2 = tmp1 + tmp2;
+    z3 = tmp0 + tmp2;
+    z4 = tmp1 + tmp3;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+
+    tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, -FIX_0_899976223); /* sqrt(2) * ( c7-c3) */
+    z2 = MULTIPLY(z2, -FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, -FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, -FIX_0_390180644); /* sqrt(2) * ( c5-c3) */
+
+    z3 += z5;
+    z4 += z5;
+
+    tmp0 += z1 + z3;
+    tmp1 += z2 + z4;
+    tmp2 += z2 + z3;
+    tmp3 += z1 + z4;
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    outptr0 = range_limit(int)DESCALE(tmp10 + tmp3,
+                                         CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr7 = range_limit(int)DESCALE(tmp10 - tmp3,
+                                         CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr1 = range_limit(int)DESCALE(tmp11 + tmp2,
+                                         CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr6 = range_limit(int)DESCALE(tmp11 - tmp2,
+                                         CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr2 = range_limit(int)DESCALE(tmp12 + tmp1,
+                                         CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr5 = range_limit(int)DESCALE(tmp12 - tmp1,
+                                         CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr3 = range_limit(int)DESCALE(tmp13 + tmp0,
+                                         CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr4 = range_limit(int)DESCALE(tmp13 - tmp0,
+                                         CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
+  }
+}
+
+#ifdef IDCT_SCALING_SUPPORTED
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 7x7 output block.
+ *
+ * Optimized algorithm with 12 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/14).
+ */
+
+GLOBAL(void)
+jpeg_idct_7x7(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+              JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp10, tmp11, tmp12, tmp13;
+  JLONG z1, z2, z3;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace7 * 7;         /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 7; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp13 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp13 = LEFT_SHIFT(tmp13, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp13 += ONE << (CONST_BITS - PASS1_BITS - 1);
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    tmp10 = MULTIPLY(z2 - z3, FIX(0.881747734));     /* c4 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.314692123));     /* c6 */
+    tmp11 = tmp10 + tmp12 + tmp13 - MULTIPLY(z2, FIX(1.841218003)); /* c2+c4-c6 */
+    tmp0 = z1 + z3;
+    z2 -= tmp0;
+    tmp0 = MULTIPLY(tmp0, FIX(1.274162392)) + tmp13; /* c2 */
+    tmp10 += tmp0 - MULTIPLY(z3, FIX(0.077722536));  /* c2-c4-c6 */
+    tmp12 += tmp0 - MULTIPLY(z1, FIX(2.470602249));  /* c2+c4+c6 */
+    tmp13 += MULTIPLY(z2, FIX(1.414213562));         /* c0 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+
+    tmp1 = MULTIPLY(z1 + z2, FIX(0.935414347));      /* (c3+c1-c5)/2 */
+    tmp2 = MULTIPLY(z1 - z2, FIX(0.170262339));      /* (c3+c5-c1)/2 */
+    tmp0 = tmp1 - tmp2;
+    tmp1 += tmp2;
+    tmp2 = MULTIPLY(z2 + z3, -FIX(1.378756276));     /* -c1 */
+    tmp1 += tmp2;
+    z2 = MULTIPLY(z1 + z3, FIX(0.613604268));        /* c5 */
+    tmp0 += z2;
+    tmp2 += z2 + MULTIPLY(z3, FIX(1.870828693));     /* c3+c1-c5 */
+
+    /* Final output stage */
+
+    wsptr7 * 0 = (int)RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS - PASS1_BITS);
+    wsptr7 * 6 = (int)RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS - PASS1_BITS);
+    wsptr7 * 1 = (int)RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS - PASS1_BITS);
+    wsptr7 * 5 = (int)RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS - PASS1_BITS);
+    wsptr7 * 2 = (int)RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS - PASS1_BITS);
+    wsptr7 * 4 = (int)RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS - PASS1_BITS);
+    wsptr7 * 3 = (int)RIGHT_SHIFT(tmp13, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 7 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 7; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp13 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    tmp13 = LEFT_SHIFT(tmp13, CONST_BITS);
+
+    z1 = (JLONG)wsptr2;
+    z2 = (JLONG)wsptr4;
+    z3 = (JLONG)wsptr6;
+
+    tmp10 = MULTIPLY(z2 - z3, FIX(0.881747734));     /* c4 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.314692123));     /* c6 */
+    tmp11 = tmp10 + tmp12 + tmp13 - MULTIPLY(z2, FIX(1.841218003)); /* c2+c4-c6 */
+    tmp0 = z1 + z3;
+    z2 -= tmp0;
+    tmp0 = MULTIPLY(tmp0, FIX(1.274162392)) + tmp13; /* c2 */
+    tmp10 += tmp0 - MULTIPLY(z3, FIX(0.077722536));  /* c2-c4-c6 */
+    tmp12 += tmp0 - MULTIPLY(z1, FIX(2.470602249));  /* c2+c4+c6 */
+    tmp13 += MULTIPLY(z2, FIX(1.414213562));         /* c0 */
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z3 = (JLONG)wsptr5;
+
+    tmp1 = MULTIPLY(z1 + z2, FIX(0.935414347));      /* (c3+c1-c5)/2 */
+    tmp2 = MULTIPLY(z1 - z2, FIX(0.170262339));      /* (c3+c5-c1)/2 */
+    tmp0 = tmp1 - tmp2;
+    tmp1 += tmp2;
+    tmp2 = MULTIPLY(z2 + z3, -FIX(1.378756276));     /* -c1 */
+    tmp1 += tmp2;
+    z2 = MULTIPLY(z1 + z3, FIX(0.613604268));        /* c5 */
+    tmp0 += z2;
+    tmp2 += z2 + MULTIPLY(z3, FIX(1.870828693));     /* c3+c1-c5 */
+
+    /* Final output stage */
+
+    outptr0 = range_limit(int)RIGHT_SHIFT(tmp10 + tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr6 = range_limit(int)RIGHT_SHIFT(tmp10 - tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr1 = range_limit(int)RIGHT_SHIFT(tmp11 + tmp1,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr5 = range_limit(int)RIGHT_SHIFT(tmp11 - tmp1,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr2 = range_limit(int)RIGHT_SHIFT(tmp12 + tmp2,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr4 = range_limit(int)RIGHT_SHIFT(tmp12 - tmp2,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr3 = range_limit(int)RIGHT_SHIFT(tmp13,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+
+    wsptr += 7;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 6x6 output block.
+ *
+ * Optimized algorithm with 3 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/12).
+ */
+
+GLOBAL(void)
+jpeg_idct_6x6(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+              JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp10, tmp11, tmp12;
+  JLONG z1, z2, z3;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace6 * 6;         /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 6; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS - PASS1_BITS - 1);
+    tmp2 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    tmp10 = MULTIPLY(tmp2, FIX(0.707106781));   /* c4 */
+    tmp1 = tmp0 + tmp10;
+    tmp11 = RIGHT_SHIFT(tmp0 - tmp10 - tmp10, CONST_BITS - PASS1_BITS);
+    tmp10 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    tmp0 = MULTIPLY(tmp10, FIX(1.224744871));   /* c2 */
+    tmp10 = tmp1 + tmp0;
+    tmp12 = tmp1 - tmp0;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    tmp1 = MULTIPLY(z1 + z3, FIX(0.366025404)); /* c5 */
+    tmp0 = tmp1 + LEFT_SHIFT(z1 + z2, CONST_BITS);
+    tmp2 = tmp1 + LEFT_SHIFT(z3 - z2, CONST_BITS);
+    tmp1 = LEFT_SHIFT(z1 - z2 - z3, PASS1_BITS);
+
+    /* Final output stage */
+
+    wsptr6 * 0 = (int)RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS - PASS1_BITS);
+    wsptr6 * 5 = (int)RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS - PASS1_BITS);
+    wsptr6 * 1 = (int)(tmp11 + tmp1);
+    wsptr6 * 4 = (int)(tmp11 - tmp1);
+    wsptr6 * 2 = (int)RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS - PASS1_BITS);
+    wsptr6 * 3 = (int)RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 6 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 6; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    tmp2 = (JLONG)wsptr4;
+    tmp10 = MULTIPLY(tmp2, FIX(0.707106781));   /* c4 */
+    tmp1 = tmp0 + tmp10;
+    tmp11 = tmp0 - tmp10 - tmp10;
+    tmp10 = (JLONG)wsptr2;
+    tmp0 = MULTIPLY(tmp10, FIX(1.224744871));   /* c2 */
+    tmp10 = tmp1 + tmp0;
+    tmp12 = tmp1 - tmp0;
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z3 = (JLONG)wsptr5;
+    tmp1 = MULTIPLY(z1 + z3, FIX(0.366025404)); /* c5 */
+    tmp0 = tmp1 + LEFT_SHIFT(z1 + z2, CONST_BITS);
+    tmp2 = tmp1 + LEFT_SHIFT(z3 - z2, CONST_BITS);
+    tmp1 = LEFT_SHIFT(z1 - z2 - z3, CONST_BITS);
+
+    /* Final output stage */
+
+    outptr0 = range_limit(int)RIGHT_SHIFT(tmp10 + tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr5 = range_limit(int)RIGHT_SHIFT(tmp10 - tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr1 = range_limit(int)RIGHT_SHIFT(tmp11 + tmp1,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr4 = range_limit(int)RIGHT_SHIFT(tmp11 - tmp1,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr2 = range_limit(int)RIGHT_SHIFT(tmp12 + tmp2,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr3 = range_limit(int)RIGHT_SHIFT(tmp12 - tmp2,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+
+    wsptr += 6;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 5x5 output block.
+ *
+ * Optimized algorithm with 5 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/10).
+ */
+
+GLOBAL(void)
+jpeg_idct_5x5(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+              JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp10, tmp11, tmp12;
+  JLONG z1, z2, z3;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace5 * 5;         /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 5; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp12 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp12 = LEFT_SHIFT(tmp12, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp12 += ONE << (CONST_BITS - PASS1_BITS - 1);
+    tmp0 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    tmp1 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    z1 = MULTIPLY(tmp0 + tmp1, FIX(0.790569415)); /* (c2+c4)/2 */
+    z2 = MULTIPLY(tmp0 - tmp1, FIX(0.353553391)); /* (c2-c4)/2 */
+    z3 = tmp12 + z2;
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z1;
+    tmp12 -= LEFT_SHIFT(z2, 2);
+
+    /* Odd part */
+
+    z2 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));     /* c3 */
+    tmp0 = z1 + MULTIPLY(z2, FIX(0.513743148));   /* c1-c3 */
+    tmp1 = z1 - MULTIPLY(z3, FIX(2.176250899));   /* c1+c3 */
+
+    /* Final output stage */
+
+    wsptr5 * 0 = (int)RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS - PASS1_BITS);
+    wsptr5 * 4 = (int)RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS - PASS1_BITS);
+    wsptr5 * 1 = (int)RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS - PASS1_BITS);
+    wsptr5 * 3 = (int)RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS - PASS1_BITS);
+    wsptr5 * 2 = (int)RIGHT_SHIFT(tmp12, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 5 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 5; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp12 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    tmp12 = LEFT_SHIFT(tmp12, CONST_BITS);
+    tmp0 = (JLONG)wsptr2;
+    tmp1 = (JLONG)wsptr4;
+    z1 = MULTIPLY(tmp0 + tmp1, FIX(0.790569415)); /* (c2+c4)/2 */
+    z2 = MULTIPLY(tmp0 - tmp1, FIX(0.353553391)); /* (c2-c4)/2 */
+    z3 = tmp12 + z2;
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z1;
+    tmp12 -= LEFT_SHIFT(z2, 2);
+
+    /* Odd part */
+
+    z2 = (JLONG)wsptr1;
+    z3 = (JLONG)wsptr3;
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));     /* c3 */
+    tmp0 = z1 + MULTIPLY(z2, FIX(0.513743148));   /* c1-c3 */
+    tmp1 = z1 - MULTIPLY(z3, FIX(2.176250899));   /* c1+c3 */
+
+    /* Final output stage */
+
+    outptr0 = range_limit(int)RIGHT_SHIFT(tmp10 + tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr4 = range_limit(int)RIGHT_SHIFT(tmp10 - tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr1 = range_limit(int)RIGHT_SHIFT(tmp11 + tmp1,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr3 = range_limit(int)RIGHT_SHIFT(tmp11 - tmp1,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr2 = range_limit(int)RIGHT_SHIFT(tmp12,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+
+    wsptr += 5;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 3x3 output block.
+ *
+ * Optimized algorithm with 2 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/6).
+ */
+
+GLOBAL(void)
+jpeg_idct_3x3(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+              JDIMENSION output_col)
+{
+  JLONG tmp0, tmp2, tmp10, tmp12;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace3 * 3;         /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 3; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS - PASS1_BITS - 1);
+    tmp2 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    tmp12 = MULTIPLY(tmp2, FIX(0.707106781)); /* c2 */
+    tmp10 = tmp0 + tmp12;
+    tmp2 = tmp0 - tmp12 - tmp12;
+
+    /* Odd part */
+
+    tmp12 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    tmp0 = MULTIPLY(tmp12, FIX(1.224744871)); /* c1 */
+
+    /* Final output stage */
+
+    wsptr3 * 0 = (int)RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS - PASS1_BITS);
+    wsptr3 * 2 = (int)RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS - PASS1_BITS);
+    wsptr3 * 1 = (int)RIGHT_SHIFT(tmp2, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 3 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 3; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    tmp2 = (JLONG)wsptr2;
+    tmp12 = MULTIPLY(tmp2, FIX(0.707106781)); /* c2 */
+    tmp10 = tmp0 + tmp12;
+    tmp2 = tmp0 - tmp12 - tmp12;
+
+    /* Odd part */
+
+    tmp12 = (JLONG)wsptr1;
+    tmp0 = MULTIPLY(tmp12, FIX(1.224744871)); /* c1 */
+
+    /* Final output stage */
+
+    outptr0 = range_limit(int)RIGHT_SHIFT(tmp10 + tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr2 = range_limit(int)RIGHT_SHIFT(tmp10 - tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr1 = range_limit(int)RIGHT_SHIFT(tmp2,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+
+    wsptr += 3;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 9x9 output block.
+ *
+ * Optimized algorithm with 10 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/18).
+ */
+
+GLOBAL(void)
+jpeg_idct_9x9(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+              JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp3, tmp10, tmp11, tmp12, tmp13, tmp14;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace8 * 9;         /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS - PASS1_BITS - 1);
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    tmp3 = MULTIPLY(z3, FIX(0.707106781));      /* c6 */
+    tmp1 = tmp0 + tmp3;
+    tmp2 = tmp0 - tmp3 - tmp3;
+
+    tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */
+    tmp11 = tmp2 + tmp0;
+    tmp14 = tmp2 - tmp0 - tmp0;
+
+    tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */
+    tmp2 = MULTIPLY(z1, FIX(1.083350441));      /* c4 */
+    tmp3 = MULTIPLY(z2, FIX(0.245575608));      /* c8 */
+
+    tmp10 = tmp1 + tmp0 - tmp3;
+    tmp12 = tmp1 - tmp0 + tmp2;
+    tmp13 = tmp1 - tmp2 + tmp3;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+
+    z2 = MULTIPLY(z2, -FIX(1.224744871));            /* -c3 */
+
+    tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955));      /* c5 */
+    tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525));      /* c7 */
+    tmp0 = tmp2 + tmp3 - z2;
+    tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481));      /* c1 */
+    tmp2 += z2 - tmp1;
+    tmp3 += z2 + tmp1;
+    tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */
+
+    /* Final output stage */
+
+    wsptr8 * 0 = (int)RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS - PASS1_BITS);
+    wsptr8 * 8 = (int)RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS - PASS1_BITS);
+    wsptr8 * 1 = (int)RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS - PASS1_BITS);
+    wsptr8 * 7 = (int)RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS - PASS1_BITS);
+    wsptr8 * 2 = (int)RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS - PASS1_BITS);
+    wsptr8 * 6 = (int)RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS - PASS1_BITS);
+    wsptr8 * 3 = (int)RIGHT_SHIFT(tmp13 + tmp3, CONST_BITS - PASS1_BITS);
+    wsptr8 * 5 = (int)RIGHT_SHIFT(tmp13 - tmp3, CONST_BITS - PASS1_BITS);
+    wsptr8 * 4 = (int)RIGHT_SHIFT(tmp14, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 9 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 9; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+
+    z1 = (JLONG)wsptr2;
+    z2 = (JLONG)wsptr4;
+    z3 = (JLONG)wsptr6;
+
+    tmp3 = MULTIPLY(z3, FIX(0.707106781));      /* c6 */
+    tmp1 = tmp0 + tmp3;
+    tmp2 = tmp0 - tmp3 - tmp3;
+
+    tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */
+    tmp11 = tmp2 + tmp0;
+    tmp14 = tmp2 - tmp0 - tmp0;
+
+    tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */
+    tmp2 = MULTIPLY(z1, FIX(1.083350441));      /* c4 */
+    tmp3 = MULTIPLY(z2, FIX(0.245575608));      /* c8 */
+
+    tmp10 = tmp1 + tmp0 - tmp3;
+    tmp12 = tmp1 - tmp0 + tmp2;
+    tmp13 = tmp1 - tmp2 + tmp3;
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z3 = (JLONG)wsptr5;
+    z4 = (JLONG)wsptr7;
+
+    z2 = MULTIPLY(z2, -FIX(1.224744871));            /* -c3 */
+
+    tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955));      /* c5 */
+    tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525));      /* c7 */
+    tmp0 = tmp2 + tmp3 - z2;
+    tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481));      /* c1 */
+    tmp2 += z2 - tmp1;
+    tmp3 += z2 + tmp1;
+    tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */
+
+    /* Final output stage */
+
+    outptr0 = range_limit(int)RIGHT_SHIFT(tmp10 + tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr8 = range_limit(int)RIGHT_SHIFT(tmp10 - tmp0,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr1 = range_limit(int)RIGHT_SHIFT(tmp11 + tmp1,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr7 = range_limit(int)RIGHT_SHIFT(tmp11 - tmp1,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr2 = range_limit(int)RIGHT_SHIFT(tmp12 + tmp2,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr6 = range_limit(int)RIGHT_SHIFT(tmp12 - tmp2,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr3 = range_limit(int)RIGHT_SHIFT(tmp13 + tmp3,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr5 = range_limit(int)RIGHT_SHIFT(tmp13 - tmp3,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr4 = range_limit(int)RIGHT_SHIFT(tmp14,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 10x10 output block.
+ *
+ * Optimized algorithm with 12 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/20).
+ */
+
+GLOBAL(void)
+jpeg_idct_10x10(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24;
+  JLONG z1, z2, z3, z4, z5;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace8 * 10;        /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z3 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z3 += ONE << (CONST_BITS - PASS1_BITS - 1);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    z1 = MULTIPLY(z4, FIX(1.144122806));         /* c4 */
+    z2 = MULTIPLY(z4, FIX(0.437016024));         /* c8 */
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z2;
+
+    tmp22 = RIGHT_SHIFT(z3 - LEFT_SHIFT(z1 - z2, 1),
+                        CONST_BITS - PASS1_BITS); /* c0 = (c4-c8)*2 */
+
+    z2 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));    /* c6 */
+    tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */
+    tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */
+
+    tmp20 = tmp10 + tmp12;
+    tmp24 = tmp10 - tmp12;
+    tmp21 = tmp11 + tmp13;
+    tmp23 = tmp11 - tmp13;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+
+    tmp11 = z2 + z4;
+    tmp13 = z2 - z4;
+
+    tmp12 = MULTIPLY(tmp13, FIX(0.309016994));        /* (c3-c7)/2 */
+    z5 = LEFT_SHIFT(z3, CONST_BITS);
+
+    z2 = MULTIPLY(tmp11, FIX(0.951056516));           /* (c3+c7)/2 */
+    z4 = z5 + tmp12;
+
+    tmp10 = MULTIPLY(z1, FIX(1.396802247)) + z2 + z4; /* c1 */
+    tmp14 = MULTIPLY(z1, FIX(0.221231742)) - z2 + z4; /* c9 */
+
+    z2 = MULTIPLY(tmp11, FIX(0.587785252));           /* (c1-c9)/2 */
+    z4 = z5 - tmp12 - LEFT_SHIFT(tmp13, CONST_BITS - 1);
+
+    tmp12 = LEFT_SHIFT(z1 - tmp13 - z3, PASS1_BITS);
+
+    tmp11 = MULTIPLY(z1, FIX(1.260073511)) - z2 - z4; /* c3 */
+    tmp13 = MULTIPLY(z1, FIX(0.642039522)) - z2 + z4; /* c7 */
+
+    /* Final output stage */
+
+    wsptr8 * 0 = (int)RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 9 = (int)RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 1 = (int)RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 8 = (int)RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 2 = (int)(tmp22 + tmp12);
+    wsptr8 * 7 = (int)(tmp22 - tmp12);
+    wsptr8 * 3 = (int)RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 6 = (int)RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 4 = (int)RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 5 = (int)RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 10 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 10; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z3 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    z4 = (JLONG)wsptr4;
+    z1 = MULTIPLY(z4, FIX(1.144122806));         /* c4 */
+    z2 = MULTIPLY(z4, FIX(0.437016024));         /* c8 */
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z2;
+
+    tmp22 = z3 - LEFT_SHIFT(z1 - z2, 1);         /* c0 = (c4-c8)*2 */
+
+    z2 = (JLONG)wsptr2;
+    z3 = (JLONG)wsptr6;
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));    /* c6 */
+    tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */
+    tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */
+
+    tmp20 = tmp10 + tmp12;
+    tmp24 = tmp10 - tmp12;
+    tmp21 = tmp11 + tmp13;
+    tmp23 = tmp11 - tmp13;
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z3 = (JLONG)wsptr5;
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    z4 = (JLONG)wsptr7;
+
+    tmp11 = z2 + z4;
+    tmp13 = z2 - z4;
+
+    tmp12 = MULTIPLY(tmp13, FIX(0.309016994));        /* (c3-c7)/2 */
+
+    z2 = MULTIPLY(tmp11, FIX(0.951056516));           /* (c3+c7)/2 */
+    z4 = z3 + tmp12;
+
+    tmp10 = MULTIPLY(z1, FIX(1.396802247)) + z2 + z4; /* c1 */
+    tmp14 = MULTIPLY(z1, FIX(0.221231742)) - z2 + z4; /* c9 */
+
+    z2 = MULTIPLY(tmp11, FIX(0.587785252));           /* (c1-c9)/2 */
+    z4 = z3 - tmp12 - LEFT_SHIFT(tmp13, CONST_BITS - 1);
+
+    tmp12 = LEFT_SHIFT(z1 - tmp13, CONST_BITS) - z3;
+
+    tmp11 = MULTIPLY(z1, FIX(1.260073511)) - z2 - z4; /* c3 */
+    tmp13 = MULTIPLY(z1, FIX(0.642039522)) - z2 + z4; /* c7 */
+
+    /* Final output stage */
+
+    outptr0 = range_limit(int)RIGHT_SHIFT(tmp20 + tmp10,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr9 = range_limit(int)RIGHT_SHIFT(tmp20 - tmp10,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr1 = range_limit(int)RIGHT_SHIFT(tmp21 + tmp11,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr8 = range_limit(int)RIGHT_SHIFT(tmp21 - tmp11,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr2 = range_limit(int)RIGHT_SHIFT(tmp22 + tmp12,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr7 = range_limit(int)RIGHT_SHIFT(tmp22 - tmp12,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr3 = range_limit(int)RIGHT_SHIFT(tmp23 + tmp13,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr6 = range_limit(int)RIGHT_SHIFT(tmp23 - tmp13,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr4 = range_limit(int)RIGHT_SHIFT(tmp24 + tmp14,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+    outptr5 = range_limit(int)RIGHT_SHIFT(tmp24 - tmp14,
+                                             CONST_BITS + PASS1_BITS + 3) &
+                            RANGE_MASK;
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing an 11x11 output block.
+ *
+ * Optimized algorithm with 24 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/22).
+ */
+
+GLOBAL(void)
+jpeg_idct_11x11(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace8 * 11;        /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp10 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp10 = LEFT_SHIFT(tmp10, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp10 += ONE << (CONST_BITS - PASS1_BITS - 1);
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    tmp20 = MULTIPLY(z2 - z3, FIX(2.546640132));     /* c2+c4 */
+    tmp23 = MULTIPLY(z2 - z1, FIX(0.430815045));     /* c2-c6 */
+    z4 = z1 + z3;
+    tmp24 = MULTIPLY(z4, -FIX(1.155664402));         /* -(c2-c10) */
+    z4 -= z2;
+    tmp25 = tmp10 + MULTIPLY(z4, FIX(1.356927976));  /* c2 */
+    tmp21 = tmp20 + tmp23 + tmp25 -
+            MULTIPLY(z2, FIX(1.821790775));          /* c2+c4+c10-c6 */
+    tmp20 += tmp25 + MULTIPLY(z3, FIX(2.115825087)); /* c4+c6 */
+    tmp23 += tmp25 - MULTIPLY(z1, FIX(1.513598477)); /* c6+c8 */
+    tmp24 += tmp25;
+    tmp22 = tmp24 - MULTIPLY(z3, FIX(0.788749120));  /* c8+c10 */
+    tmp24 += MULTIPLY(z2, FIX(1.944413522)) -        /* c2+c8 */
+             MULTIPLY(z1, FIX(1.390975730));         /* c4+c10 */
+    tmp25 = tmp10 - MULTIPLY(z4, FIX(1.414213562));  /* c0 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+
+    tmp11 = z1 + z2;
+    tmp14 = MULTIPLY(tmp11 + z3 + z4, FIX(0.398430003)); /* c9 */
+    tmp11 = MULTIPLY(tmp11, FIX(0.887983902));           /* c3-c9 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(0.670361295));         /* c5-c9 */
+    tmp13 = tmp14 + MULTIPLY(z1 + z4, FIX(0.366151574)); /* c7-c9 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(0.923107866));              /* c7+c5+c3-c1-2*c9 */
+    z1    = tmp14 - MULTIPLY(z2 + z3, FIX(1.163011579)); /* c7+c9 */
+    tmp11 += z1 + MULTIPLY(z2, FIX(2.073276588));        /* c1+c7+3*c9-c3 */
+    tmp12 += z1 - MULTIPLY(z3, FIX(1.192193623));        /* c3+c5-c7-c9 */
+    z1    = MULTIPLY(z2 + z4, -FIX(1.798248910));        /* -(c1+c9) */
+    tmp11 += z1;
+    tmp13 += z1 + MULTIPLY(z4, FIX(2.102458632));        /* c1+c5+c9-c7 */
+    tmp14 += MULTIPLY(z2, -FIX(1.467221301)) +           /* -(c5+c9) */
+             MULTIPLY(z3, FIX(1.001388905)) -            /* c1-c9 */
+             MULTIPLY(z4, FIX(1.684843907));             /* c3+c9 */
+
+    /* Final output stage */
+
+    wsptr8 * 0  = (int)RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 10 = (int)RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 1  = (int)RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 9  = (int)RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 2  = (int)RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 8  = (int)RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 3  = (int)RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 7  = (int)RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 4  = (int)RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 6  = (int)RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 5  = (int)RIGHT_SHIFT(tmp25, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 11 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 11; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp10 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    tmp10 = LEFT_SHIFT(tmp10, CONST_BITS);
+
+    z1 = (JLONG)wsptr2;
+    z2 = (JLONG)wsptr4;
+    z3 = (JLONG)wsptr6;
+
+    tmp20 = MULTIPLY(z2 - z3, FIX(2.546640132));     /* c2+c4 */
+    tmp23 = MULTIPLY(z2 - z1, FIX(0.430815045));     /* c2-c6 */
+    z4 = z1 + z3;
+    tmp24 = MULTIPLY(z4, -FIX(1.155664402));         /* -(c2-c10) */
+    z4 -= z2;
+    tmp25 = tmp10 + MULTIPLY(z4, FIX(1.356927976));  /* c2 */
+    tmp21 = tmp20 + tmp23 + tmp25 -
+            MULTIPLY(z2, FIX(1.821790775));          /* c2+c4+c10-c6 */
+    tmp20 += tmp25 + MULTIPLY(z3, FIX(2.115825087)); /* c4+c6 */
+    tmp23 += tmp25 - MULTIPLY(z1, FIX(1.513598477)); /* c6+c8 */
+    tmp24 += tmp25;
+    tmp22 = tmp24 - MULTIPLY(z3, FIX(0.788749120));  /* c8+c10 */
+    tmp24 += MULTIPLY(z2, FIX(1.944413522)) -        /* c2+c8 */
+             MULTIPLY(z1, FIX(1.390975730));         /* c4+c10 */
+    tmp25 = tmp10 - MULTIPLY(z4, FIX(1.414213562));  /* c0 */
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z3 = (JLONG)wsptr5;
+    z4 = (JLONG)wsptr7;
+
+    tmp11 = z1 + z2;
+    tmp14 = MULTIPLY(tmp11 + z3 + z4, FIX(0.398430003)); /* c9 */
+    tmp11 = MULTIPLY(tmp11, FIX(0.887983902));           /* c3-c9 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(0.670361295));         /* c5-c9 */
+    tmp13 = tmp14 + MULTIPLY(z1 + z4, FIX(0.366151574)); /* c7-c9 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(0.923107866));              /* c7+c5+c3-c1-2*c9 */
+    z1    = tmp14 - MULTIPLY(z2 + z3, FIX(1.163011579)); /* c7+c9 */
+    tmp11 += z1 + MULTIPLY(z2, FIX(2.073276588));        /* c1+c7+3*c9-c3 */
+    tmp12 += z1 - MULTIPLY(z3, FIX(1.192193623));        /* c3+c5-c7-c9 */
+    z1    = MULTIPLY(z2 + z4, -FIX(1.798248910));        /* -(c1+c9) */
+    tmp11 += z1;
+    tmp13 += z1 + MULTIPLY(z4, FIX(2.102458632));        /* c1+c5+c9-c7 */
+    tmp14 += MULTIPLY(z2, -FIX(1.467221301)) +           /* -(c5+c9) */
+             MULTIPLY(z3, FIX(1.001388905)) -            /* c1-c9 */
+             MULTIPLY(z4, FIX(1.684843907));             /* c3+c9 */
+
+    /* Final output stage */
+
+    outptr0  = range_limit(int)RIGHT_SHIFT(tmp20 + tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr10 = range_limit(int)RIGHT_SHIFT(tmp20 - tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr1  = range_limit(int)RIGHT_SHIFT(tmp21 + tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr9  = range_limit(int)RIGHT_SHIFT(tmp21 - tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr2  = range_limit(int)RIGHT_SHIFT(tmp22 + tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr8  = range_limit(int)RIGHT_SHIFT(tmp22 - tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr3  = range_limit(int)RIGHT_SHIFT(tmp23 + tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr7  = range_limit(int)RIGHT_SHIFT(tmp23 - tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr4  = range_limit(int)RIGHT_SHIFT(tmp24 + tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr6  = range_limit(int)RIGHT_SHIFT(tmp24 - tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr5  = range_limit(int)RIGHT_SHIFT(tmp25,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 12x12 output block.
+ *
+ * Optimized algorithm with 15 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/24).
+ */
+
+GLOBAL(void)
+jpeg_idct_12x12(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace8 * 12;        /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z3 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z3 += ONE << (CONST_BITS - PASS1_BITS - 1);
+
+    z4 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    z4 = MULTIPLY(z4, FIX(1.224744871)); /* c4 */
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z4 = MULTIPLY(z1, FIX(1.366025404)); /* c2 */
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+    z2 = LEFT_SHIFT(z2, CONST_BITS);
+
+    tmp12 = z1 - z2;
+
+    tmp21 = z3 + tmp12;
+    tmp24 = z3 - tmp12;
+
+    tmp12 = z4 + z2;
+
+    tmp20 = tmp10 + tmp12;
+    tmp25 = tmp10 - tmp12;
+
+    tmp12 = z4 - z1 - z2;
+
+    tmp22 = tmp11 + tmp12;
+    tmp23 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+
+    tmp11 = MULTIPLY(z2, FIX(1.306562965));                  /* c3 */
+    tmp14 = MULTIPLY(z2, -FIX_0_541196100);                  /* -c9 */
+
+    tmp10 = z1 + z3;
+    tmp15 = MULTIPLY(tmp10 + z4, FIX(0.860918669));          /* c7 */
+    tmp12 = tmp15 + MULTIPLY(tmp10, FIX(0.261052384));       /* c5-c7 */
+    tmp10 = tmp12 + tmp11 + MULTIPLY(z1, FIX(0.280143716));  /* c1-c5 */
+    tmp13 = MULTIPLY(z3 + z4, -FIX(1.045510580));            /* -(c7+c11) */
+    tmp12 += tmp13 + tmp14 - MULTIPLY(z3, FIX(1.478575242)); /* c1+c5-c7-c11 */
+    tmp13 += tmp15 - tmp11 + MULTIPLY(z4, FIX(1.586706681)); /* c1+c11 */
+    tmp15 += tmp14 - MULTIPLY(z1, FIX(0.676326758)) -        /* c7-c11 */
+             MULTIPLY(z4, FIX(1.982889723));                 /* c5+c7 */
+
+    z1 -= z4;
+    z2 -= z3;
+    z3 = MULTIPLY(z1 + z2, FIX_0_541196100);                 /* c9 */
+    tmp11 = z3 + MULTIPLY(z1, FIX_0_765366865);              /* c3-c9 */
+    tmp14 = z3 - MULTIPLY(z2, FIX_1_847759065);              /* c3+c9 */
+
+    /* Final output stage */
+
+    wsptr8 * 0  = (int)RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 11 = (int)RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 1  = (int)RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 10 = (int)RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 2  = (int)RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 9  = (int)RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 3  = (int)RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 8  = (int)RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 4  = (int)RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 7  = (int)RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 5  = (int)RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS - PASS1_BITS);
+    wsptr8 * 6  = (int)RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 12 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 12; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z3 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+
+    z4 = (JLONG)wsptr4;
+    z4 = MULTIPLY(z4, FIX(1.224744871)); /* c4 */
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    z1 = (JLONG)wsptr2;
+    z4 = MULTIPLY(z1, FIX(1.366025404)); /* c2 */
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    z2 = (JLONG)wsptr6;
+    z2 = LEFT_SHIFT(z2, CONST_BITS);
+
+    tmp12 = z1 - z2;
+
+    tmp21 = z3 + tmp12;
+    tmp24 = z3 - tmp12;
+
+    tmp12 = z4 + z2;
+
+    tmp20 = tmp10 + tmp12;
+    tmp25 = tmp10 - tmp12;
+
+    tmp12 = z4 - z1 - z2;
+
+    tmp22 = tmp11 + tmp12;
+    tmp23 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z3 = (JLONG)wsptr5;
+    z4 = (JLONG)wsptr7;
+
+    tmp11 = MULTIPLY(z2, FIX(1.306562965));                  /* c3 */
+    tmp14 = MULTIPLY(z2, -FIX_0_541196100);                  /* -c9 */
+
+    tmp10 = z1 + z3;
+    tmp15 = MULTIPLY(tmp10 + z4, FIX(0.860918669));          /* c7 */
+    tmp12 = tmp15 + MULTIPLY(tmp10, FIX(0.261052384));       /* c5-c7 */
+    tmp10 = tmp12 + tmp11 + MULTIPLY(z1, FIX(0.280143716));  /* c1-c5 */
+    tmp13 = MULTIPLY(z3 + z4, -FIX(1.045510580));            /* -(c7+c11) */
+    tmp12 += tmp13 + tmp14 - MULTIPLY(z3, FIX(1.478575242)); /* c1+c5-c7-c11 */
+    tmp13 += tmp15 - tmp11 + MULTIPLY(z4, FIX(1.586706681)); /* c1+c11 */
+    tmp15 += tmp14 - MULTIPLY(z1, FIX(0.676326758)) -        /* c7-c11 */
+             MULTIPLY(z4, FIX(1.982889723));                 /* c5+c7 */
+
+    z1 -= z4;
+    z2 -= z3;
+    z3 = MULTIPLY(z1 + z2, FIX_0_541196100);                 /* c9 */
+    tmp11 = z3 + MULTIPLY(z1, FIX_0_765366865);              /* c3-c9 */
+    tmp14 = z3 - MULTIPLY(z2, FIX_1_847759065);              /* c3+c9 */
+
+    /* Final output stage */
+
+    outptr0  = range_limit(int)RIGHT_SHIFT(tmp20 + tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr11 = range_limit(int)RIGHT_SHIFT(tmp20 - tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr1  = range_limit(int)RIGHT_SHIFT(tmp21 + tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr10 = range_limit(int)RIGHT_SHIFT(tmp21 - tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr2  = range_limit(int)RIGHT_SHIFT(tmp22 + tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr9  = range_limit(int)RIGHT_SHIFT(tmp22 - tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr3  = range_limit(int)RIGHT_SHIFT(tmp23 + tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr8  = range_limit(int)RIGHT_SHIFT(tmp23 - tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr4  = range_limit(int)RIGHT_SHIFT(tmp24 + tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr7  = range_limit(int)RIGHT_SHIFT(tmp24 - tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr5  = range_limit(int)RIGHT_SHIFT(tmp25 + tmp15,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr6  = range_limit(int)RIGHT_SHIFT(tmp25 - tmp15,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 13x13 output block.
+ *
+ * Optimized algorithm with 29 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/26).
+ */
+
+GLOBAL(void)
+jpeg_idct_13x13(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace8 * 13;        /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z1 += ONE << (CONST_BITS - PASS1_BITS - 1);
+
+    z2 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    tmp12 = MULTIPLY(tmp10, FIX(1.155388986));                /* (c4+c6)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.096834934)) + z1;           /* (c4-c6)/2 */
+
+    tmp20 = MULTIPLY(z2, FIX(1.373119086)) + tmp12 + tmp13;   /* c2 */
+    tmp22 = MULTIPLY(z2, FIX(0.501487041)) - tmp12 + tmp13;   /* c10 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.316450131));                /* (c8-c12)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.486914739)) + z1;           /* (c8+c12)/2 */
+
+    tmp21 = MULTIPLY(z2, FIX(1.058554052)) - tmp12 + tmp13;   /* c6 */
+    tmp25 = MULTIPLY(z2, -FIX(1.252223920)) + tmp12 + tmp13;  /* c4 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.435816023));                /* (c2-c10)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.937303064)) - z1;           /* (c2+c10)/2 */
+
+    tmp23 = MULTIPLY(z2, -FIX(0.170464608)) - tmp12 - tmp13;  /* c12 */
+    tmp24 = MULTIPLY(z2, -FIX(0.803364869)) + tmp12 - tmp13;  /* c8 */
+
+    tmp26 = MULTIPLY(tmp11 - z2, FIX(1.414213562)) + z1;      /* c0 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.322312651));     /* c3 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(1.163874945));     /* c5 */
+    tmp15 = z1 + z4;
+    tmp13 = MULTIPLY(tmp15, FIX(0.937797057));       /* c7 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(2.020082300));          /* c7+c5+c3-c1 */
+    tmp14 = MULTIPLY(z2 + z3, -FIX(0.338443458));    /* -c11 */
+    tmp11 += tmp14 + MULTIPLY(z2, FIX(0.837223564)); /* c5+c9+c11-c3 */
+    tmp12 += tmp14 - MULTIPLY(z3, FIX(1.572116027)); /* c1+c5-c9-c11 */
+    tmp14 = MULTIPLY(z2 + z4, -FIX(1.163874945));    /* -c5 */
+    tmp11 += tmp14;
+    tmp13 += tmp14 + MULTIPLY(z4, FIX(2.205608352)); /* c3+c5+c9-c7 */
+    tmp14 = MULTIPLY(z3 + z4, -FIX(0.657217813));    /* -c9 */
+    tmp12 += tmp14;
+    tmp13 += tmp14;
+    tmp15 = MULTIPLY(tmp15, FIX(0.338443458));       /* c11 */
+    tmp14 = tmp15 + MULTIPLY(z1, FIX(0.318774355)) - /* c9-c11 */
+            MULTIPLY(z2, FIX(0.466105296));          /* c1-c7 */
+    z1    = MULTIPLY(z3 - z2, FIX(0.937797057));     /* c7 */
+    tmp14 += z1;
+    tmp15 += z1 + MULTIPLY(z3, FIX(0.384515595)) -   /* c3-c7 */
+             MULTIPLY(z4, FIX(1.742345811));         /* c1+c11 */
+
+    /* Final output stage */
+
+    wsptr8 * 0  = (int)RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 12 = (int)RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 1  = (int)RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 11 = (int)RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 2  = (int)RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 10 = (int)RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 3  = (int)RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 9  = (int)RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 4  = (int)RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 8  = (int)RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 5  = (int)RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS - PASS1_BITS);
+    wsptr8 * 7  = (int)RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS - PASS1_BITS);
+    wsptr8 * 6  = (int)RIGHT_SHIFT(tmp26, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 13 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 13; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z1 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+
+    z2 = (JLONG)wsptr2;
+    z3 = (JLONG)wsptr4;
+    z4 = (JLONG)wsptr6;
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    tmp12 = MULTIPLY(tmp10, FIX(1.155388986));                /* (c4+c6)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.096834934)) + z1;           /* (c4-c6)/2 */
+
+    tmp20 = MULTIPLY(z2, FIX(1.373119086)) + tmp12 + tmp13;   /* c2 */
+    tmp22 = MULTIPLY(z2, FIX(0.501487041)) - tmp12 + tmp13;   /* c10 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.316450131));                /* (c8-c12)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.486914739)) + z1;           /* (c8+c12)/2 */
+
+    tmp21 = MULTIPLY(z2, FIX(1.058554052)) - tmp12 + tmp13;   /* c6 */
+    tmp25 = MULTIPLY(z2, -FIX(1.252223920)) + tmp12 + tmp13;  /* c4 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.435816023));                /* (c2-c10)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.937303064)) - z1;           /* (c2+c10)/2 */
+
+    tmp23 = MULTIPLY(z2, -FIX(0.170464608)) - tmp12 - tmp13;  /* c12 */
+    tmp24 = MULTIPLY(z2, -FIX(0.803364869)) + tmp12 - tmp13;  /* c8 */
+
+    tmp26 = MULTIPLY(tmp11 - z2, FIX(1.414213562)) + z1;      /* c0 */
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z3 = (JLONG)wsptr5;
+    z4 = (JLONG)wsptr7;
+
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.322312651));     /* c3 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(1.163874945));     /* c5 */
+    tmp15 = z1 + z4;
+    tmp13 = MULTIPLY(tmp15, FIX(0.937797057));       /* c7 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(2.020082300));          /* c7+c5+c3-c1 */
+    tmp14 = MULTIPLY(z2 + z3, -FIX(0.338443458));    /* -c11 */
+    tmp11 += tmp14 + MULTIPLY(z2, FIX(0.837223564)); /* c5+c9+c11-c3 */
+    tmp12 += tmp14 - MULTIPLY(z3, FIX(1.572116027)); /* c1+c5-c9-c11 */
+    tmp14 = MULTIPLY(z2 + z4, -FIX(1.163874945));    /* -c5 */
+    tmp11 += tmp14;
+    tmp13 += tmp14 + MULTIPLY(z4, FIX(2.205608352)); /* c3+c5+c9-c7 */
+    tmp14 = MULTIPLY(z3 + z4, -FIX(0.657217813));    /* -c9 */
+    tmp12 += tmp14;
+    tmp13 += tmp14;
+    tmp15 = MULTIPLY(tmp15, FIX(0.338443458));       /* c11 */
+    tmp14 = tmp15 + MULTIPLY(z1, FIX(0.318774355)) - /* c9-c11 */
+            MULTIPLY(z2, FIX(0.466105296));          /* c1-c7 */
+    z1    = MULTIPLY(z3 - z2, FIX(0.937797057));     /* c7 */
+    tmp14 += z1;
+    tmp15 += z1 + MULTIPLY(z3, FIX(0.384515595)) -   /* c3-c7 */
+             MULTIPLY(z4, FIX(1.742345811));         /* c1+c11 */
+
+    /* Final output stage */
+
+    outptr0  = range_limit(int)RIGHT_SHIFT(tmp20 + tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr12 = range_limit(int)RIGHT_SHIFT(tmp20 - tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr1  = range_limit(int)RIGHT_SHIFT(tmp21 + tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr11 = range_limit(int)RIGHT_SHIFT(tmp21 - tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr2  = range_limit(int)RIGHT_SHIFT(tmp22 + tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr10 = range_limit(int)RIGHT_SHIFT(tmp22 - tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr3  = range_limit(int)RIGHT_SHIFT(tmp23 + tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr9  = range_limit(int)RIGHT_SHIFT(tmp23 - tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr4  = range_limit(int)RIGHT_SHIFT(tmp24 + tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr8  = range_limit(int)RIGHT_SHIFT(tmp24 - tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr5  = range_limit(int)RIGHT_SHIFT(tmp25 + tmp15,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr7  = range_limit(int)RIGHT_SHIFT(tmp25 - tmp15,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr6  = range_limit(int)RIGHT_SHIFT(tmp26,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 14x14 output block.
+ *
+ * Optimized algorithm with 20 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/28).
+ */
+
+GLOBAL(void)
+jpeg_idct_14x14(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace8 * 14;        /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z1 += ONE << (CONST_BITS - PASS1_BITS - 1);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    z2 = MULTIPLY(z4, FIX(1.274162392));         /* c4 */
+    z3 = MULTIPLY(z4, FIX(0.314692123));         /* c12 */
+    z4 = MULTIPLY(z4, FIX(0.881747734));         /* c8 */
+
+    tmp10 = z1 + z2;
+    tmp11 = z1 + z3;
+    tmp12 = z1 - z4;
+
+    tmp23 = RIGHT_SHIFT(z1 - LEFT_SHIFT(z2 + z3 - z4, 1),
+                        CONST_BITS - PASS1_BITS); /* c0 = (c4+c12-c8)*2 */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    z3 = MULTIPLY(z1 + z2, FIX(1.105676686));    /* c6 */
+
+    tmp13 = z3 + MULTIPLY(z1, FIX(0.273079590)); /* c2-c6 */
+    tmp14 = z3 - MULTIPLY(z2, FIX(1.719280954)); /* c6+c10 */
+    tmp15 = MULTIPLY(z1, FIX(0.613604268)) -     /* c10 */
+            MULTIPLY(z2, FIX(1.378756276));      /* c2 */
+
+    tmp20 = tmp10 + tmp13;
+    tmp26 = tmp10 - tmp13;
+    tmp21 = tmp11 + tmp14;
+    tmp25 = tmp11 - tmp14;
+    tmp22 = tmp12 + tmp15;
+    tmp24 = tmp12 - tmp15;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+    tmp13 = LEFT_SHIFT(z4, CONST_BITS);
+
+    tmp14 = z1 + z3;
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.334852607));           /* c3 */
+    tmp12 = MULTIPLY(tmp14, FIX(1.197448846));             /* c5 */
+    tmp10 = tmp11 + tmp12 + tmp13 - MULTIPLY(z1, FIX(1.126980169)); /* c3+c5-c1 */
+    tmp14 = MULTIPLY(tmp14, FIX(0.752406978));             /* c9 */
+    tmp16 = tmp14 - MULTIPLY(z1, FIX(1.061150426));        /* c9+c11-c13 */
+    z1    -= z2;
+    tmp15 = MULTIPLY(z1, FIX(0.467085129)) - tmp13;        /* c11 */
+    tmp16 += tmp15;
+    z1    += z4;
+    z4    = MULTIPLY(z2 + z3, -FIX(0.158341681)) - tmp13;  /* -c13 */
+    tmp11 += z4 - MULTIPLY(z2, FIX(0.424103948));          /* c3-c9-c13 */
+    tmp12 += z4 - MULTIPLY(z3, FIX(2.373959773));          /* c3+c5-c13 */
+    z4    = MULTIPLY(z3 - z2, FIX(1.405321284));           /* c1 */
+    tmp14 += z4 + tmp13 - MULTIPLY(z3, FIX(1.6906431334)); /* c1+c9-c11 */
+    tmp15 += z4 + MULTIPLY(z2, FIX(0.674957567));          /* c1+c11-c5 */
+
+    tmp13 = LEFT_SHIFT(z1 - z3, PASS1_BITS);
+
+    /* Final output stage */
+
+    wsptr8 * 0  = (int)RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 13 = (int)RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 1  = (int)RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 12 = (int)RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 2  = (int)RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 11 = (int)RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 3  = (int)(tmp23 + tmp13);
+    wsptr8 * 10 = (int)(tmp23 - tmp13);
+    wsptr8 * 4  = (int)RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 9  = (int)RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 5  = (int)RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS - PASS1_BITS);
+    wsptr8 * 8  = (int)RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS - PASS1_BITS);
+    wsptr8 * 6  = (int)RIGHT_SHIFT(tmp26 + tmp16, CONST_BITS - PASS1_BITS);
+    wsptr8 * 7  = (int)RIGHT_SHIFT(tmp26 - tmp16, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 14 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 14; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z1 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    z4 = (JLONG)wsptr4;
+    z2 = MULTIPLY(z4, FIX(1.274162392));         /* c4 */
+    z3 = MULTIPLY(z4, FIX(0.314692123));         /* c12 */
+    z4 = MULTIPLY(z4, FIX(0.881747734));         /* c8 */
+
+    tmp10 = z1 + z2;
+    tmp11 = z1 + z3;
+    tmp12 = z1 - z4;
+
+    tmp23 = z1 - LEFT_SHIFT(z2 + z3 - z4, 1);    /* c0 = (c4+c12-c8)*2 */
+
+    z1 = (JLONG)wsptr2;
+    z2 = (JLONG)wsptr6;
+
+    z3 = MULTIPLY(z1 + z2, FIX(1.105676686));    /* c6 */
+
+    tmp13 = z3 + MULTIPLY(z1, FIX(0.273079590)); /* c2-c6 */
+    tmp14 = z3 - MULTIPLY(z2, FIX(1.719280954)); /* c6+c10 */
+    tmp15 = MULTIPLY(z1, FIX(0.613604268)) -     /* c10 */
+            MULTIPLY(z2, FIX(1.378756276));      /* c2 */
+
+    tmp20 = tmp10 + tmp13;
+    tmp26 = tmp10 - tmp13;
+    tmp21 = tmp11 + tmp14;
+    tmp25 = tmp11 - tmp14;
+    tmp22 = tmp12 + tmp15;
+    tmp24 = tmp12 - tmp15;
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z3 = (JLONG)wsptr5;
+    z4 = (JLONG)wsptr7;
+    z4 = LEFT_SHIFT(z4, CONST_BITS);
+
+    tmp14 = z1 + z3;
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.334852607));           /* c3 */
+    tmp12 = MULTIPLY(tmp14, FIX(1.197448846));             /* c5 */
+    tmp10 = tmp11 + tmp12 + z4 - MULTIPLY(z1, FIX(1.126980169)); /* c3+c5-c1 */
+    tmp14 = MULTIPLY(tmp14, FIX(0.752406978));             /* c9 */
+    tmp16 = tmp14 - MULTIPLY(z1, FIX(1.061150426));        /* c9+c11-c13 */
+    z1    -= z2;
+    tmp15 = MULTIPLY(z1, FIX(0.467085129)) - z4;           /* c11 */
+    tmp16 += tmp15;
+    tmp13 = MULTIPLY(z2 + z3, -FIX(0.158341681)) - z4;     /* -c13 */
+    tmp11 += tmp13 - MULTIPLY(z2, FIX(0.424103948));       /* c3-c9-c13 */
+    tmp12 += tmp13 - MULTIPLY(z3, FIX(2.373959773));       /* c3+c5-c13 */
+    tmp13 = MULTIPLY(z3 - z2, FIX(1.405321284));           /* c1 */
+    tmp14 += tmp13 + z4 - MULTIPLY(z3, FIX(1.6906431334)); /* c1+c9-c11 */
+    tmp15 += tmp13 + MULTIPLY(z2, FIX(0.674957567));       /* c1+c11-c5 */
+
+    tmp13 = LEFT_SHIFT(z1 - z3, CONST_BITS) + z4;
+
+    /* Final output stage */
+
+    outptr0  = range_limit(int)RIGHT_SHIFT(tmp20 + tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr13 = range_limit(int)RIGHT_SHIFT(tmp20 - tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr1  = range_limit(int)RIGHT_SHIFT(tmp21 + tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr12 = range_limit(int)RIGHT_SHIFT(tmp21 - tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr2  = range_limit(int)RIGHT_SHIFT(tmp22 + tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr11 = range_limit(int)RIGHT_SHIFT(tmp22 - tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr3  = range_limit(int)RIGHT_SHIFT(tmp23 + tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr10 = range_limit(int)RIGHT_SHIFT(tmp23 - tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr4  = range_limit(int)RIGHT_SHIFT(tmp24 + tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr9  = range_limit(int)RIGHT_SHIFT(tmp24 - tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr5  = range_limit(int)RIGHT_SHIFT(tmp25 + tmp15,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr8  = range_limit(int)RIGHT_SHIFT(tmp25 - tmp15,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr6  = range_limit(int)RIGHT_SHIFT(tmp26 + tmp16,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr7  = range_limit(int)RIGHT_SHIFT(tmp26 - tmp16,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 15x15 output block.
+ *
+ * Optimized algorithm with 22 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/30).
+ */
+
+GLOBAL(void)
+jpeg_idct_15x15(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26, tmp27;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace8 * 15;        /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z1 += ONE << (CONST_BITS - PASS1_BITS - 1);
+
+    z2 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    tmp10 = MULTIPLY(z4, FIX(0.437016024)); /* c12 */
+    tmp11 = MULTIPLY(z4, FIX(1.144122806)); /* c6 */
+
+    tmp12 = z1 - tmp10;
+    tmp13 = z1 + tmp11;
+    z1 -= LEFT_SHIFT(tmp11 - tmp10, 1);     /* c0 = (c6-c12)*2 */
+
+    z4 = z2 - z3;
+    z3 += z2;
+    tmp10 = MULTIPLY(z3, FIX(1.337628990)); /* (c2+c4)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.045680613)); /* (c2-c4)/2 */
+    z2 = MULTIPLY(z2, FIX(1.439773946));    /* c4+c14 */
+
+    tmp20 = tmp13 + tmp10 + tmp11;
+    tmp23 = tmp12 - tmp10 + tmp11 + z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.547059574)); /* (c8+c14)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.399234004)); /* (c8-c14)/2 */
+
+    tmp25 = tmp13 - tmp10 - tmp11;
+    tmp26 = tmp12 + tmp10 - tmp11 - z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.790569415)); /* (c6+c12)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.353553391)); /* (c6-c12)/2 */
+
+    tmp21 = tmp12 + tmp10 + tmp11;
+    tmp24 = tmp13 - tmp10 + tmp11;
+    tmp11 += tmp11;
+    tmp22 = z1 + tmp11;                     /* c10 = c6-c12 */
+    tmp27 = z1 - tmp11 - tmp11;             /* c0 = (c6-c12)*2 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    z3 = MULTIPLY(z4, FIX(1.224744871));                    /* c5 */
+    z4 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+
+    tmp13 = z2 - z4;
+    tmp15 = MULTIPLY(z1 + tmp13, FIX(0.831253876));         /* c9 */
+    tmp11 = tmp15 + MULTIPLY(z1, FIX(0.513743148));         /* c3-c9 */
+    tmp14 = tmp15 - MULTIPLY(tmp13, FIX(2.176250899));      /* c3+c9 */
+
+    tmp13 = MULTIPLY(z2, -FIX(0.831253876));                /* -c9 */
+    tmp15 = MULTIPLY(z2, -FIX(1.344997024));                /* -c3 */
+    z2 = z1 - z4;
+    tmp12 = z3 + MULTIPLY(z2, FIX(1.406466353));            /* c1 */
+
+    tmp10 = tmp12 + MULTIPLY(z4, FIX(2.457431844)) - tmp15; /* c1+c7 */
+    tmp16 = tmp12 - MULTIPLY(z1, FIX(1.112434820)) + tmp13; /* c1-c13 */
+    tmp12 = MULTIPLY(z2, FIX(1.224744871)) - z3;            /* c5 */
+    z2 = MULTIPLY(z1 + z4, FIX(0.575212477));               /* c11 */
+    tmp13 += z2 + MULTIPLY(z1, FIX(0.475753014)) - z3;      /* c7-c11 */
+    tmp15 += z2 - MULTIPLY(z4, FIX(0.869244010)) + z3;      /* c11+c13 */
+
+    /* Final output stage */
+
+    wsptr8 * 0  = (int)RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 14 = (int)RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 1  = (int)RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 13 = (int)RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 2  = (int)RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 12 = (int)RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 3  = (int)RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 11 = (int)RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 4  = (int)RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 10 = (int)RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS - PASS1_BITS);
+    wsptr8 * 5  = (int)RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS - PASS1_BITS);
+    wsptr8 * 9  = (int)RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS - PASS1_BITS);
+    wsptr8 * 6  = (int)RIGHT_SHIFT(tmp26 + tmp16, CONST_BITS - PASS1_BITS);
+    wsptr8 * 8  = (int)RIGHT_SHIFT(tmp26 - tmp16, CONST_BITS - PASS1_BITS);
+    wsptr8 * 7  = (int)RIGHT_SHIFT(tmp27, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 15 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 15; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z1 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+
+    z2 = (JLONG)wsptr2;
+    z3 = (JLONG)wsptr4;
+    z4 = (JLONG)wsptr6;
+
+    tmp10 = MULTIPLY(z4, FIX(0.437016024)); /* c12 */
+    tmp11 = MULTIPLY(z4, FIX(1.144122806)); /* c6 */
+
+    tmp12 = z1 - tmp10;
+    tmp13 = z1 + tmp11;
+    z1 -= LEFT_SHIFT(tmp11 - tmp10, 1);     /* c0 = (c6-c12)*2 */
+
+    z4 = z2 - z3;
+    z3 += z2;
+    tmp10 = MULTIPLY(z3, FIX(1.337628990)); /* (c2+c4)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.045680613)); /* (c2-c4)/2 */
+    z2 = MULTIPLY(z2, FIX(1.439773946));    /* c4+c14 */
+
+    tmp20 = tmp13 + tmp10 + tmp11;
+    tmp23 = tmp12 - tmp10 + tmp11 + z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.547059574)); /* (c8+c14)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.399234004)); /* (c8-c14)/2 */
+
+    tmp25 = tmp13 - tmp10 - tmp11;
+    tmp26 = tmp12 + tmp10 - tmp11 - z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.790569415)); /* (c6+c12)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.353553391)); /* (c6-c12)/2 */
+
+    tmp21 = tmp12 + tmp10 + tmp11;
+    tmp24 = tmp13 - tmp10 + tmp11;
+    tmp11 += tmp11;
+    tmp22 = z1 + tmp11;                     /* c10 = c6-c12 */
+    tmp27 = z1 - tmp11 - tmp11;             /* c0 = (c6-c12)*2 */
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z4 = (JLONG)wsptr5;
+    z3 = MULTIPLY(z4, FIX(1.224744871));                    /* c5 */
+    z4 = (JLONG)wsptr7;
+
+    tmp13 = z2 - z4;
+    tmp15 = MULTIPLY(z1 + tmp13, FIX(0.831253876));         /* c9 */
+    tmp11 = tmp15 + MULTIPLY(z1, FIX(0.513743148));         /* c3-c9 */
+    tmp14 = tmp15 - MULTIPLY(tmp13, FIX(2.176250899));      /* c3+c9 */
+
+    tmp13 = MULTIPLY(z2, -FIX(0.831253876));                /* -c9 */
+    tmp15 = MULTIPLY(z2, -FIX(1.344997024));                /* -c3 */
+    z2 = z1 - z4;
+    tmp12 = z3 + MULTIPLY(z2, FIX(1.406466353));            /* c1 */
+
+    tmp10 = tmp12 + MULTIPLY(z4, FIX(2.457431844)) - tmp15; /* c1+c7 */
+    tmp16 = tmp12 - MULTIPLY(z1, FIX(1.112434820)) + tmp13; /* c1-c13 */
+    tmp12 = MULTIPLY(z2, FIX(1.224744871)) - z3;            /* c5 */
+    z2 = MULTIPLY(z1 + z4, FIX(0.575212477));               /* c11 */
+    tmp13 += z2 + MULTIPLY(z1, FIX(0.475753014)) - z3;      /* c7-c11 */
+    tmp15 += z2 - MULTIPLY(z4, FIX(0.869244010)) + z3;      /* c11+c13 */
+
+    /* Final output stage */
+
+    outptr0  = range_limit(int)RIGHT_SHIFT(tmp20 + tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr14 = range_limit(int)RIGHT_SHIFT(tmp20 - tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr1  = range_limit(int)RIGHT_SHIFT(tmp21 + tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr13 = range_limit(int)RIGHT_SHIFT(tmp21 - tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr2  = range_limit(int)RIGHT_SHIFT(tmp22 + tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr12 = range_limit(int)RIGHT_SHIFT(tmp22 - tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr3  = range_limit(int)RIGHT_SHIFT(tmp23 + tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr11 = range_limit(int)RIGHT_SHIFT(tmp23 - tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr4  = range_limit(int)RIGHT_SHIFT(tmp24 + tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr10 = range_limit(int)RIGHT_SHIFT(tmp24 - tmp14,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr5  = range_limit(int)RIGHT_SHIFT(tmp25 + tmp15,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr9  = range_limit(int)RIGHT_SHIFT(tmp25 - tmp15,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr6  = range_limit(int)RIGHT_SHIFT(tmp26 + tmp16,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr8  = range_limit(int)RIGHT_SHIFT(tmp26 - tmp16,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr7  = range_limit(int)RIGHT_SHIFT(tmp27,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 16x16 output block.
+ *
+ * Optimized algorithm with 28 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/32).
+ */
+
+GLOBAL(void)
+jpeg_idct_16x16(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp3, tmp10, tmp11, tmp12, tmp13;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26, tmp27;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace8 * 16;        /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS - PASS1_BITS - 1);
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 4, quantptrDCTSIZE * 4);
+    tmp1 = MULTIPLY(z1, FIX(1.306562965));      /* c416 = c28 */
+    tmp2 = MULTIPLY(z1, FIX_0_541196100);       /* c1216 = c68 */
+
+    tmp10 = tmp0 + tmp1;
+    tmp11 = tmp0 - tmp1;
+    tmp12 = tmp0 + tmp2;
+    tmp13 = tmp0 - tmp2;
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+    z3 = z1 - z2;
+    z4 = MULTIPLY(z3, FIX(0.275899379));        /* c1416 = c78 */
+    z3 = MULTIPLY(z3, FIX(1.387039845));        /* c216 = c18 */
+
+    tmp0 = z3 + MULTIPLY(z2, FIX_2_562915447);  /* (c6+c2)16 = (c3+c1)8 */
+    tmp1 = z4 + MULTIPLY(z1, FIX_0_899976223);  /* (c6-c14)16 = (c3-c7)8 */
+    tmp2 = z3 - MULTIPLY(z1, FIX(0.601344887)); /* (c2-c10)16 = (c1-c5)8 */
+    tmp3 = z4 - MULTIPLY(z2, FIX(0.509795579)); /* (c10-c14)16 = (c5-c7)8 */
+
+    tmp20 = tmp10 + tmp0;
+    tmp27 = tmp10 - tmp0;
+    tmp21 = tmp12 + tmp1;
+    tmp26 = tmp12 - tmp1;
+    tmp22 = tmp13 + tmp2;
+    tmp25 = tmp13 - tmp2;
+    tmp23 = tmp11 + tmp3;
+    tmp24 = tmp11 - tmp3;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+
+    tmp11 = z1 + z3;
+
+    tmp1  = MULTIPLY(z1 + z2, FIX(1.353318001));   /* c3 */
+    tmp2  = MULTIPLY(tmp11,   FIX(1.247225013));   /* c5 */
+    tmp3  = MULTIPLY(z1 + z4, FIX(1.093201867));   /* c7 */
+    tmp10 = MULTIPLY(z1 - z4, FIX(0.897167586));   /* c9 */
+    tmp11 = MULTIPLY(tmp11,   FIX(0.666655658));   /* c11 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.410524528));   /* c13 */
+    tmp0  = tmp1 + tmp2 + tmp3 -
+            MULTIPLY(z1, FIX(2.286341144));        /* c7+c5+c3-c1 */
+    tmp13 = tmp10 + tmp11 + tmp12 -
+            MULTIPLY(z1, FIX(1.835730603));        /* c9+c11+c13-c15 */
+    z1    = MULTIPLY(z2 + z3, FIX(0.138617169));   /* c15 */
+    tmp1  += z1 + MULTIPLY(z2, FIX(0.071888074));  /* c9+c11-c3-c15 */
+    tmp2  += z1 - MULTIPLY(z3, FIX(1.125726048));  /* c5+c7+c15-c3 */
+    z1    = MULTIPLY(z3 - z2, FIX(1.407403738));   /* c1 */
+    tmp11 += z1 - MULTIPLY(z3, FIX(0.766367282));  /* c1+c11-c9-c13 */
+    tmp12 += z1 + MULTIPLY(z2, FIX(1.971951411));  /* c1+c5+c13-c7 */
+    z2    += z4;
+    z1    = MULTIPLY(z2, -FIX(0.666655658));       /* -c11 */
+    tmp1  += z1;
+    tmp3  += z1 + MULTIPLY(z4, FIX(1.065388962));  /* c3+c11+c15-c7 */
+    z2    = MULTIPLY(z2, -FIX(1.247225013));       /* -c5 */
+    tmp10 += z2 + MULTIPLY(z4, FIX(3.141271809));  /* c1+c5+c9-c13 */
+    tmp12 += z2;
+    z2    = MULTIPLY(z3 + z4, -FIX(1.353318001));  /* -c3 */
+    tmp2  += z2;
+    tmp3  += z2;
+    z2    = MULTIPLY(z4 - z3, FIX(0.410524528));   /* c13 */
+    tmp10 += z2;
+    tmp11 += z2;
+
+    /* Final output stage */
+
+    wsptr8 * 0  = (int)RIGHT_SHIFT(tmp20 + tmp0,  CONST_BITS - PASS1_BITS);
+    wsptr8 * 15 = (int)RIGHT_SHIFT(tmp20 - tmp0,  CONST_BITS - PASS1_BITS);
+    wsptr8 * 1  = (int)RIGHT_SHIFT(tmp21 + tmp1,  CONST_BITS - PASS1_BITS);
+    wsptr8 * 14 = (int)RIGHT_SHIFT(tmp21 - tmp1,  CONST_BITS - PASS1_BITS);
+    wsptr8 * 2  = (int)RIGHT_SHIFT(tmp22 + tmp2,  CONST_BITS - PASS1_BITS);
+    wsptr8 * 13 = (int)RIGHT_SHIFT(tmp22 - tmp2,  CONST_BITS - PASS1_BITS);
+    wsptr8 * 3  = (int)RIGHT_SHIFT(tmp23 + tmp3,  CONST_BITS - PASS1_BITS);
+    wsptr8 * 12 = (int)RIGHT_SHIFT(tmp23 - tmp3,  CONST_BITS - PASS1_BITS);
+    wsptr8 * 4  = (int)RIGHT_SHIFT(tmp24 + tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 11 = (int)RIGHT_SHIFT(tmp24 - tmp10, CONST_BITS - PASS1_BITS);
+    wsptr8 * 5  = (int)RIGHT_SHIFT(tmp25 + tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 10 = (int)RIGHT_SHIFT(tmp25 - tmp11, CONST_BITS - PASS1_BITS);
+    wsptr8 * 6  = (int)RIGHT_SHIFT(tmp26 + tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 9  = (int)RIGHT_SHIFT(tmp26 - tmp12, CONST_BITS - PASS1_BITS);
+    wsptr8 * 7  = (int)RIGHT_SHIFT(tmp27 + tmp13, CONST_BITS - PASS1_BITS);
+    wsptr8 * 8  = (int)RIGHT_SHIFT(tmp27 - tmp13, CONST_BITS - PASS1_BITS);
+  }
+
+  /* Pass 2: process 16 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 16; ctr++) {
+    outptr = output_bufctr + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG)wsptr0 + (ONE << (PASS1_BITS + 2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+
+    z1 = (JLONG)wsptr4;
+    tmp1 = MULTIPLY(z1, FIX(1.306562965));      /* c416 = c28 */
+    tmp2 = MULTIPLY(z1, FIX_0_541196100);       /* c1216 = c68 */
+
+    tmp10 = tmp0 + tmp1;
+    tmp11 = tmp0 - tmp1;
+    tmp12 = tmp0 + tmp2;
+    tmp13 = tmp0 - tmp2;
+
+    z1 = (JLONG)wsptr2;
+    z2 = (JLONG)wsptr6;
+    z3 = z1 - z2;
+    z4 = MULTIPLY(z3, FIX(0.275899379));        /* c1416 = c78 */
+    z3 = MULTIPLY(z3, FIX(1.387039845));        /* c216 = c18 */
+
+    tmp0 = z3 + MULTIPLY(z2, FIX_2_562915447);  /* (c6+c2)16 = (c3+c1)8 */
+    tmp1 = z4 + MULTIPLY(z1, FIX_0_899976223);  /* (c6-c14)16 = (c3-c7)8 */
+    tmp2 = z3 - MULTIPLY(z1, FIX(0.601344887)); /* (c2-c10)16 = (c1-c5)8 */
+    tmp3 = z4 - MULTIPLY(z2, FIX(0.509795579)); /* (c10-c14)16 = (c5-c7)8 */
+
+    tmp20 = tmp10 + tmp0;
+    tmp27 = tmp10 - tmp0;
+    tmp21 = tmp12 + tmp1;
+    tmp26 = tmp12 - tmp1;
+    tmp22 = tmp13 + tmp2;
+    tmp25 = tmp13 - tmp2;
+    tmp23 = tmp11 + tmp3;
+    tmp24 = tmp11 - tmp3;
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr1;
+    z2 = (JLONG)wsptr3;
+    z3 = (JLONG)wsptr5;
+    z4 = (JLONG)wsptr7;
+
+    tmp11 = z1 + z3;
+
+    tmp1  = MULTIPLY(z1 + z2, FIX(1.353318001));   /* c3 */
+    tmp2  = MULTIPLY(tmp11,   FIX(1.247225013));   /* c5 */
+    tmp3  = MULTIPLY(z1 + z4, FIX(1.093201867));   /* c7 */
+    tmp10 = MULTIPLY(z1 - z4, FIX(0.897167586));   /* c9 */
+    tmp11 = MULTIPLY(tmp11,   FIX(0.666655658));   /* c11 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.410524528));   /* c13 */
+    tmp0  = tmp1 + tmp2 + tmp3 -
+            MULTIPLY(z1, FIX(2.286341144));        /* c7+c5+c3-c1 */
+    tmp13 = tmp10 + tmp11 + tmp12 -
+            MULTIPLY(z1, FIX(1.835730603));        /* c9+c11+c13-c15 */
+    z1    = MULTIPLY(z2 + z3, FIX(0.138617169));   /* c15 */
+    tmp1  += z1 + MULTIPLY(z2, FIX(0.071888074));  /* c9+c11-c3-c15 */
+    tmp2  += z1 - MULTIPLY(z3, FIX(1.125726048));  /* c5+c7+c15-c3 */
+    z1    = MULTIPLY(z3 - z2, FIX(1.407403738));   /* c1 */
+    tmp11 += z1 - MULTIPLY(z3, FIX(0.766367282));  /* c1+c11-c9-c13 */
+    tmp12 += z1 + MULTIPLY(z2, FIX(1.971951411));  /* c1+c5+c13-c7 */
+    z2    += z4;
+    z1    = MULTIPLY(z2, -FIX(0.666655658));       /* -c11 */
+    tmp1  += z1;
+    tmp3  += z1 + MULTIPLY(z4, FIX(1.065388962));  /* c3+c11+c15-c7 */
+    z2    = MULTIPLY(z2, -FIX(1.247225013));       /* -c5 */
+    tmp10 += z2 + MULTIPLY(z4, FIX(3.141271809));  /* c1+c5+c9-c13 */
+    tmp12 += z2;
+    z2    = MULTIPLY(z3 + z4, -FIX(1.353318001));  /* -c3 */
+    tmp2  += z2;
+    tmp3  += z2;
+    z2    = MULTIPLY(z4 - z3, FIX(0.410524528));   /* c13 */
+    tmp10 += z2;
+    tmp11 += z2;
+
+    /* Final output stage */
+
+    outptr0  = range_limit(int)RIGHT_SHIFT(tmp20 + tmp0,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr15 = range_limit(int)RIGHT_SHIFT(tmp20 - tmp0,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr1  = range_limit(int)RIGHT_SHIFT(tmp21 + tmp1,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr14 = range_limit(int)RIGHT_SHIFT(tmp21 - tmp1,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr2  = range_limit(int)RIGHT_SHIFT(tmp22 + tmp2,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr13 = range_limit(int)RIGHT_SHIFT(tmp22 - tmp2,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr3  = range_limit(int)RIGHT_SHIFT(tmp23 + tmp3,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr12 = range_limit(int)RIGHT_SHIFT(tmp23 - tmp3,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr4  = range_limit(int)RIGHT_SHIFT(tmp24 + tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr11 = range_limit(int)RIGHT_SHIFT(tmp24 - tmp10,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr5  = range_limit(int)RIGHT_SHIFT(tmp25 + tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr10 = range_limit(int)RIGHT_SHIFT(tmp25 - tmp11,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr6  = range_limit(int)RIGHT_SHIFT(tmp26 + tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr9  = range_limit(int)RIGHT_SHIFT(tmp26 - tmp12,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr7  = range_limit(int)RIGHT_SHIFT(tmp27 + tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+    outptr8  = range_limit(int)RIGHT_SHIFT(tmp27 - tmp13,
+                                              CONST_BITS + PASS1_BITS + 3) &
+                             RANGE_MASK;
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+#endif /* IDCT_SCALING_SUPPORTED */
+#endif /* DCT_ISLOW_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jidctred.c Added

@@ -0,0 +1,409 @@
+/*
+ * jidctred.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains inverse-DCT routines that produce reduced-size output:
+ * either 4x4, 2x2, or 1x1 pixels from an 8x8 DCT block.
+ *
+ * The implementation is based on the Loeffler, Ligtenberg and Moschytz (LL&M)
+ * algorithm used in jidctint.c.  We simply replace each 8-to-8 1-D IDCT step
+ * with an 8-to-4 step that produces the four averages of two adjacent outputs
+ * (or an 8-to-2 step producing two averages of four outputs, for 2x2 output).
+ * These steps were derived by computing the corresponding values at the end
+ * of the normal LL&M code, then simplifying as much as possible.
+ *
+ * 1x1 is trivial: just take the DC coefficient divided by 8.
+ *
+ * See jidctint.c for additional comments.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"               /* Private declarations for DCT subsystem */
+
+#ifdef IDCT_SCALING_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling is the same as in jidctint.c. */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  13
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 13
+#define FIX_0_211164243  ((JLONG)1730)          /* FIX(0.211164243) */
+#define FIX_0_509795579  ((JLONG)4176)          /* FIX(0.509795579) */
+#define FIX_0_601344887  ((JLONG)4926)          /* FIX(0.601344887) */
+#define FIX_0_720959822  ((JLONG)5906)          /* FIX(0.720959822) */
+#define FIX_0_765366865  ((JLONG)6270)          /* FIX(0.765366865) */
+#define FIX_0_850430095  ((JLONG)6967)          /* FIX(0.850430095) */
+#define FIX_0_899976223  ((JLONG)7373)          /* FIX(0.899976223) */
+#define FIX_1_061594337  ((JLONG)8697)          /* FIX(1.061594337) */
+#define FIX_1_272758580  ((JLONG)10426)         /* FIX(1.272758580) */
+#define FIX_1_451774981  ((JLONG)11893)         /* FIX(1.451774981) */
+#define FIX_1_847759065  ((JLONG)15137)         /* FIX(1.847759065) */
+#define FIX_2_172734803  ((JLONG)17799)         /* FIX(2.172734803) */
+#define FIX_2_562915447  ((JLONG)20995)         /* FIX(2.562915447) */
+#define FIX_3_624509785  ((JLONG)29692)         /* FIX(3.624509785) */
+#else
+#define FIX_0_211164243  FIX(0.211164243)
+#define FIX_0_509795579  FIX(0.509795579)
+#define FIX_0_601344887  FIX(0.601344887)
+#define FIX_0_720959822  FIX(0.720959822)
+#define FIX_0_765366865  FIX(0.765366865)
+#define FIX_0_850430095  FIX(0.850430095)
+#define FIX_0_899976223  FIX(0.899976223)
+#define FIX_1_061594337  FIX(1.061594337)
+#define FIX_1_272758580  FIX(1.272758580)
+#define FIX_1_451774981  FIX(1.451774981)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_2_172734803  FIX(2.172734803)
+#define FIX_2_562915447  FIX(2.562915447)
+#define FIX_3_624509785  FIX(3.624509785)
+#endif
+
+
+/* Multiply a JLONG variable by a JLONG constant to yield a JLONG result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
+ * For 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MULTIPLY(var, const)  MULTIPLY16C16(var, const)
+#else
+#define MULTIPLY(var, const)  ((var) * (const))
+#endif
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce an int result.  In this module, both inputs and result
+ * are 16 bits or less, so either int or short multiply will work.
+ */
+
+#define DEQUANTIZE(coef, quantval)  (((ISLOW_MULT_TYPE)(coef)) * (quantval))
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 4x4 output block.
+ */
+
+GLOBAL(void)
+jpeg_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+              JDIMENSION output_col)
+{
+  JLONG tmp0, tmp2, tmp10, tmp12;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspaceDCTSIZE * 4;   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; inptr++, quantptr++, wsptr++, ctr--) {
+    /* Don't bother to process column 4, because second pass won't use it */
+    if (ctr == DCTSIZE - 4)
+      continue;
+    if (inptrDCTSIZE * 1 == 0 && inptrDCTSIZE * 2 == 0 &&
+        inptrDCTSIZE * 3 == 0 && inptrDCTSIZE * 5 == 0 &&
+        inptrDCTSIZE * 6 == 0 && inptrDCTSIZE * 7 == 0) {
+      /* AC terms all zero; we need not examine term 4 for 4x4 output */
+      int dcval = LEFT_SHIFT(DEQUANTIZE(inptrDCTSIZE * 0,
+                                        quantptrDCTSIZE * 0), PASS1_BITS);
+
+      wsptrDCTSIZE * 0 = dcval;
+      wsptrDCTSIZE * 1 = dcval;
+      wsptrDCTSIZE * 2 = dcval;
+      wsptrDCTSIZE * 3 = dcval;
+
+      continue;
+    }
+
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS + 1);
+
+    z2 = DEQUANTIZE(inptrDCTSIZE * 2, quantptrDCTSIZE * 2);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 6, quantptrDCTSIZE * 6);
+
+    tmp2 = MULTIPLY(z2, FIX_1_847759065) + MULTIPLY(z3, -FIX_0_765366865);
+
+    tmp10 = tmp0 + tmp2;
+    tmp12 = tmp0 - tmp2;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+    z2 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    z3 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    z4 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+
+    tmp0 = MULTIPLY(z1, -FIX_0_211164243) + /* sqrt(2) * ( c3-c1) */
+           MULTIPLY(z2,  FIX_1_451774981) + /* sqrt(2) * ( c3+c7) */
+           MULTIPLY(z3, -FIX_2_172734803) + /* sqrt(2) * (-c1-c5) */
+           MULTIPLY(z4,  FIX_1_061594337);  /* sqrt(2) * ( c5+c7) */
+
+    tmp2 = MULTIPLY(z1, -FIX_0_509795579) + /* sqrt(2) * (c7-c5) */
+           MULTIPLY(z2, -FIX_0_601344887) + /* sqrt(2) * (c5-c1) */
+           MULTIPLY(z3,  FIX_0_899976223) + /* sqrt(2) * (c3-c7) */
+           MULTIPLY(z4,  FIX_2_562915447);  /* sqrt(2) * (c1+c3) */
+
+    /* Final output stage */
+
+    wsptrDCTSIZE * 0 =
+      (int)DESCALE(tmp10 + tmp2, CONST_BITS - PASS1_BITS + 1);
+    wsptrDCTSIZE * 3 =
+      (int)DESCALE(tmp10 - tmp2, CONST_BITS - PASS1_BITS + 1);
+    wsptrDCTSIZE * 1 =
+      (int)DESCALE(tmp12 + tmp0, CONST_BITS - PASS1_BITS + 1);
+    wsptrDCTSIZE * 2 =
+      (int)DESCALE(tmp12 - tmp0, CONST_BITS - PASS1_BITS + 1);
+  }
+
+  /* Pass 2: process 4 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 4; ctr++) {
+    outptr = output_bufctr + output_col;
+    /* It's not clear whether a zero row test is worthwhile here ... */
+
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr1 == 0 && wsptr2 == 0 && wsptr3 == 0 &&
+        wsptr5 == 0 && wsptr6 == 0 && wsptr7 == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit(int)DESCALE((JLONG)wsptr0,
+                                               PASS1_BITS + 3) & RANGE_MASK;
+
+      outptr0 = dcval;
+      outptr1 = dcval;
+      outptr2 = dcval;
+      outptr3 = dcval;
+
+      wsptr += DCTSIZE;         /* advance pointer to next row */
+      continue;
+    }
+#endif
+
+    /* Even part */
+
+    tmp0 = LEFT_SHIFT((JLONG)wsptr0, CONST_BITS + 1);
+
+    tmp2 = MULTIPLY((JLONG)wsptr2,  FIX_1_847759065) +
+           MULTIPLY((JLONG)wsptr6, -FIX_0_765366865);
+
+    tmp10 = tmp0 + tmp2;
+    tmp12 = tmp0 - tmp2;
+
+    /* Odd part */
+
+    z1 = (JLONG)wsptr7;
+    z2 = (JLONG)wsptr5;
+    z3 = (JLONG)wsptr3;
+    z4 = (JLONG)wsptr1;
+
+    tmp0 = MULTIPLY(z1, -FIX_0_211164243) + /* sqrt(2) * ( c3-c1) */
+           MULTIPLY(z2,  FIX_1_451774981) + /* sqrt(2) * ( c3+c7) */
+           MULTIPLY(z3, -FIX_2_172734803) + /* sqrt(2) * (-c1-c5) */
+           MULTIPLY(z4,  FIX_1_061594337);  /* sqrt(2) * ( c5+c7) */
+
+    tmp2 = MULTIPLY(z1, -FIX_0_509795579) + /* sqrt(2) * (c7-c5) */
+           MULTIPLY(z2, -FIX_0_601344887) + /* sqrt(2) * (c5-c1) */
+           MULTIPLY(z3, FIX_0_899976223) +  /* sqrt(2) * (c3-c7) */
+           MULTIPLY(z4, FIX_2_562915447);   /* sqrt(2) * (c1+c3) */
+
+    /* Final output stage */
+
+    outptr0 = range_limit(int)DESCALE(tmp10 + tmp2,
+                                         CONST_BITS + PASS1_BITS + 3 + 1) &
+                            RANGE_MASK;
+    outptr3 = range_limit(int)DESCALE(tmp10 - tmp2,
+                                         CONST_BITS + PASS1_BITS + 3 + 1) &
+                            RANGE_MASK;
+    outptr1 = range_limit(int)DESCALE(tmp12 + tmp0,
+                                         CONST_BITS + PASS1_BITS + 3 + 1) &
+                            RANGE_MASK;
+    outptr2 = range_limit(int)DESCALE(tmp12 - tmp0,
+                                         CONST_BITS + PASS1_BITS + 3 + 1) &
+                            RANGE_MASK;
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 2x2 output block.
+ */
+
+GLOBAL(void)
+jpeg_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+              JDIMENSION output_col)
+{
+  JLONG tmp0, tmp10, z1;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspaceDCTSIZE * 2;   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; inptr++, quantptr++, wsptr++, ctr--) {
+    /* Don't bother to process columns 2,4,6 */
+    if (ctr == DCTSIZE - 2 || ctr == DCTSIZE - 4 || ctr == DCTSIZE - 6)
+      continue;
+    if (inptrDCTSIZE * 1 == 0 && inptrDCTSIZE * 3 == 0 &&
+        inptrDCTSIZE * 5 == 0 && inptrDCTSIZE * 7 == 0) {
+      /* AC terms all zero; we need not examine terms 2,4,6 for 2x2 output */
+      int dcval = LEFT_SHIFT(DEQUANTIZE(inptrDCTSIZE * 0,
+                             quantptrDCTSIZE * 0), PASS1_BITS);
+
+      wsptrDCTSIZE * 0 = dcval;
+      wsptrDCTSIZE * 1 = dcval;
+
+      continue;
+    }
+
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 0, quantptrDCTSIZE * 0);
+    tmp10 = LEFT_SHIFT(z1, CONST_BITS + 2);
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptrDCTSIZE * 7, quantptrDCTSIZE * 7);
+    tmp0 = MULTIPLY(z1, -FIX_0_720959822);  /* sqrt(2) * ( c7-c5+c3-c1) */
+    z1 = DEQUANTIZE(inptrDCTSIZE * 5, quantptrDCTSIZE * 5);
+    tmp0 += MULTIPLY(z1, FIX_0_850430095);  /* sqrt(2) * (-c1+c3+c5+c7) */
+    z1 = DEQUANTIZE(inptrDCTSIZE * 3, quantptrDCTSIZE * 3);
+    tmp0 += MULTIPLY(z1, -FIX_1_272758580); /* sqrt(2) * (-c1+c3-c5-c7) */
+    z1 = DEQUANTIZE(inptrDCTSIZE * 1, quantptrDCTSIZE * 1);
+    tmp0 += MULTIPLY(z1, FIX_3_624509785);  /* sqrt(2) * ( c1+c3+c5+c7) */
+
+    /* Final output stage */
+
+    wsptrDCTSIZE * 0 =
+      (int)DESCALE(tmp10 + tmp0, CONST_BITS - PASS1_BITS + 2);
+    wsptrDCTSIZE * 1 =
+      (int)DESCALE(tmp10 - tmp0, CONST_BITS - PASS1_BITS + 2);
+  }
+
+  /* Pass 2: process 2 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 2; ctr++) {
+    outptr = output_bufctr + output_col;
+    /* It's not clear whether a zero row test is worthwhile here ... */
+
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr1 == 0 && wsptr3 == 0 && wsptr5 == 0 && wsptr7 == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit(int)DESCALE((JLONG)wsptr0,
+                                               PASS1_BITS + 3) & RANGE_MASK;
+
+      outptr0 = dcval;
+      outptr1 = dcval;
+
+      wsptr += DCTSIZE;         /* advance pointer to next row */
+      continue;
+    }
+#endif
+
+    /* Even part */
+
+    tmp10 = LEFT_SHIFT((JLONG)wsptr0, CONST_BITS + 2);
+
+    /* Odd part */
+
+    tmp0 = MULTIPLY((JLONG)wsptr7, -FIX_0_720959822) + /* sqrt(2) * ( c7-c5+c3-c1) */
+           MULTIPLY((JLONG)wsptr5,  FIX_0_850430095) + /* sqrt(2) * (-c1+c3+c5+c7) */
+           MULTIPLY((JLONG)wsptr3, -FIX_1_272758580) + /* sqrt(2) * (-c1+c3-c5-c7) */
+           MULTIPLY((JLONG)wsptr1,  FIX_3_624509785);  /* sqrt(2) * ( c1+c3+c5+c7) */
+
+    /* Final output stage */
+
+    outptr0 = range_limit(int)DESCALE(tmp10 + tmp0,
+                                         CONST_BITS + PASS1_BITS + 3 + 2) &
+                            RANGE_MASK;
+    outptr1 = range_limit(int)DESCALE(tmp10 - tmp0,
+                                         CONST_BITS + PASS1_BITS + 3 + 2) &
+                            RANGE_MASK;
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 1x1 output block.
+ */
+
+GLOBAL(void)
+jpeg_idct_1x1(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+              JDIMENSION output_col)
+{
+  int dcval;
+  ISLOW_MULT_TYPE *quantptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  SHIFT_TEMPS
+
+  /* We hardly need an inverse DCT routine for this: just take the
+   * average pixel value, which is one-eighth of the DC coefficient.
+   */
+  quantptr = (ISLOW_MULT_TYPE *)compptr->dct_table;
+  dcval = DEQUANTIZE(coef_block0, quantptr0);
+  dcval = (int)DESCALE((JLONG)dcval, 3);
+
+  output_buf0output_col = range_limitdcval & RANGE_MASK;
+}
+
+#endif /* IDCT_SCALING_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jinclude.h Added

@@ -0,0 +1,88 @@
+/*
+ * jinclude.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file exists to provide a single place to fix any problems with
+ * including the wrong system include files.  (Common problems are taken
+ * care of by the standard jconfig symbols, but on really weird systems
+ * you may have to edit this file.)
+ *
+ * NOTE: this file is NOT intended to be included by applications using the
+ * JPEG library.  Most applications need only include jpeglib.h.
+ */
+
+
+/* Include auto-config file to find out which system include files we need. */
+
+#include "jconfig.h"            /* auto configuration options */
+#define JCONFIG_INCLUDED        /* so that jpeglib.h doesn't do it again */
+
+/*
+ * We need the NULL macro and size_t typedef.
+ * On an ANSI-conforming system it is sufficient to include <stddef.h>.
+ * Otherwise, we get them from <stdlib.h> or <stdio.h>; we may have to
+ * pull in <sys/types.h> as well.
+ * Note that the core JPEG library does not require <stdio.h>;
+ * only the default error handler and data source/destination modules do.
+ * But we must pull it in because of the references to FILE in jpeglib.h.
+ * You can remove those references if you want to compile without <stdio.h>.
+ */
+
+#ifdef HAVE_STDDEF_H
+#include <stddef.h>
+#endif
+
+#ifdef HAVE_STDLIB_H
+#include <stdlib.h>
+#endif
+
+#ifdef NEED_SYS_TYPES_H
+#include <sys/types.h>
+#endif
+
+#include <stdio.h>
+
+/*
+ * We need memory copying and zeroing functions, plus strncpy().
+ * ANSI and System V implementations declare these in <string.h>.
+ * BSD doesn't have the mem() functions, but it does have bcopy()/bzero().
+ * Some systems may declare memset and memcpy in <memory.h>.
+ *
+ * NOTE: we assume the size parameters to these functions are of type size_t.
+ * Change the casts in these macros if not!
+ */
+
+#ifdef NEED_BSD_STRINGS
+
+#include <strings.h>
+#define MEMZERO(target, size) \
+  bzero((void *)(target), (size_t)(size))
+#define MEMCOPY(dest, src, size) \
+  bcopy((const void *)(src), (void *)(dest), (size_t)(size))
+
+#else /* not BSD, assume ANSI/SysV string lib */
+
+#include <string.h>
+#define MEMZERO(target, size) \
+  memset((void *)(target), 0, (size_t)(size))
+#define MEMCOPY(dest, src, size) \
+  memcpy((void *)(dest), (const void *)(src), (size_t)(size))
+
+#endif
+
+/*
+ * The modules that use fread() and fwrite() always invoke them through
+ * these macros.  On some systems you may need to twiddle the argument casts.
+ * CAUTION: argument order is different from underlying functions!
+ */
+
+#define JFREAD(file, buf, sizeofbuf) \
+  ((size_t)fread((void *)(buf), (size_t)1, (size_t)(sizeofbuf), (file)))
+#define JFWRITE(file, buf, sizeofbuf) \
+  ((size_t)fwrite((const void *)(buf), (size_t)1, (size_t)(sizeofbuf), (file)))

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jmemmgr.c Added

@@ -0,0 +1,1179 @@
+/*
+ * jmemmgr.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2016, 2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the JPEG system-independent memory management
+ * routines.  This code is usable across a wide variety of machines; most
+ * of the system dependencies have been isolated in a separate file.
+ * The major functions provided here are:
+ *   * pool-based allocation and freeing of memory;
+ *   * policy decisions about how to divide available memory among the
+ *     virtual arrays;
+ *   * control logic for swapping virtual arrays between main memory and
+ *     backing storage.
+ * The separate system-dependent file provides the actual backing-storage
+ * access code, and it contains the policy decision about how much total
+ * main memory to use.
+ * This file is system-dependent in the sense that some of its functions
+ * are unnecessary in some systems.  For example, if there is enough virtual
+ * memory so that backing storage will never be used, much of the virtual
+ * array control logic could be removed.  (Of course, if you have that much
+ * memory then you shouldn't care about a little bit of unused code...)
+ */
+
+#define JPEG_INTERNALS
+#define AM_MEMORY_MANAGER       /* we define jvirt_Xarray_control structs */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"            /* import the system-dependent declarations */
+#if !defined(_MSC_VER) || _MSC_VER > 1600
+#include <stdint.h>
+#endif
+#include <limits.h>
+
+#ifndef NO_GETENV
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare getenv() */
+extern char *getenv(const char *name);
+#endif
+#endif
+
+
+LOCAL(size_t)
+round_up_pow2(size_t a, size_t b)
+/* a rounded up to the next multiple of b, i.e. ceil(a/b)*b */
+/* Assumes a >= 0, b > 0, and b is a power of 2 */
+{
+  return ((a + b - 1) & (~(b - 1)));
+}
+
+
+/*
+ * Some important notes:
+ *   The allocation routines provided here must never return NULL.
+ *   They should exit to error_exit if unsuccessful.
+ *
+ *   It's not a good idea to try to merge the sarray and barray routines,
+ *   even though they are textually almost the same, because samples are
+ *   usually stored as bytes while coefficients are shorts or ints.  Thus,
+ *   in machines where byte pointers have a different representation from
+ *   word pointers, the resulting machine code could not be the same.
+ */
+
+
+/*
+ * Many machines require storage alignment: longs must start on 4-byte
+ * boundaries, doubles on 8-byte boundaries, etc.  On such machines, malloc()
+ * always returns pointers that are multiples of the worst-case alignment
+ * requirement, and we had better do so too.
+ * There isn't any really portable way to determine the worst-case alignment
+ * requirement.  This module assumes that the alignment requirement is
+ * multiples of ALIGN_SIZE.
+ * By default, we define ALIGN_SIZE as sizeof(double).  This is necessary on
+ * some workstations (where doubles really do need 8-byte alignment) and will
+ * work fine on nearly everything.  If your machine has lesser alignment needs,
+ * you can save a few bytes by making ALIGN_SIZE smaller.
+ * The only place I know of where this will NOT work is certain Macintosh
+ * 680x0 compilers that define double as a 10-byte IEEE extended float.
+ * Doing 10-byte alignment is counterproductive because longwords won't be
+ * aligned well.  Put "#define ALIGN_SIZE 4" in jconfig.h if you have
+ * such a compiler.
+ */
+
+#ifndef ALIGN_SIZE              /* so can override from jconfig.h */
+#ifndef WITH_SIMD
+#define ALIGN_SIZE  sizeof(double)
+#else
+#define ALIGN_SIZE  32 /* Most of the SIMD instructions we support require
+                          16-byte (128-bit) alignment, but AVX2 requires
+                          32-byte alignment. */
+#endif
+#endif
+
+/*
+ * We allocate objects from "pools", where each pool is gotten with a single
+ * request to jpeg_get_small() or jpeg_get_large().  There is no per-object
+ * overhead within a pool, except for alignment padding.  Each pool has a
+ * header with a link to the next pool of the same class.
+ * Small and large pool headers are identical.
+ */
+
+typedef struct small_pool_struct *small_pool_ptr;
+
+typedef struct small_pool_struct {
+  small_pool_ptr next;          /* next in list of pools */
+  size_t bytes_used;            /* how many bytes already used within pool */
+  size_t bytes_left;            /* bytes still available in this pool */
+} small_pool_hdr;
+
+typedef struct large_pool_struct *large_pool_ptr;
+
+typedef struct large_pool_struct {
+  large_pool_ptr next;          /* next in list of pools */
+  size_t bytes_used;            /* how many bytes already used within pool */
+  size_t bytes_left;            /* bytes still available in this pool */
+} large_pool_hdr;
+
+/*
+ * Here is the full definition of a memory manager object.
+ */
+
+typedef struct {
+  struct jpeg_memory_mgr pub;   /* public fields */
+
+  /* Each pool identifier (lifetime class) names a linked list of pools. */
+  small_pool_ptr small_listJPOOL_NUMPOOLS;
+  large_pool_ptr large_listJPOOL_NUMPOOLS;
+
+  /* Since we only have one lifetime class of virtual arrays, only one
+   * linked list is necessary (for each datatype).  Note that the virtual
+   * array control blocks being linked together are actually stored somewhere
+   * in the small-pool list.
+   */
+  jvirt_sarray_ptr virt_sarray_list;
+  jvirt_barray_ptr virt_barray_list;
+
+  /* This counts total space obtained from jpeg_get_small/large */
+  size_t total_space_allocated;
+
+  /* alloc_sarray and alloc_barray set this value for use by virtual
+   * array routines.
+   */
+  JDIMENSION last_rowsperchunk; /* from most recent alloc_sarray/barray */
+} my_memory_mgr;
+
+typedef my_memory_mgr *my_mem_ptr;
+
+
+/*
+ * The control blocks for virtual arrays.
+ * Note that these blocks are allocated in the "small" pool area.
+ * System-dependent info for the associated backing store (if any) is hidden
+ * inside the backing_store_info struct.
+ */
+
+struct jvirt_sarray_control {
+  JSAMPARRAY mem_buffer;        /* => the in-memory buffer */
+  JDIMENSION rows_in_array;     /* total virtual array height */
+  JDIMENSION samplesperrow;     /* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;         /* max rows accessed by access_virt_sarray */
+  JDIMENSION rows_in_mem;       /* height of memory buffer */
+  JDIMENSION rowsperchunk;      /* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;     /* first logical row # in the buffer */
+  JDIMENSION first_undef_row;   /* row # of first uninitialized row */
+  boolean pre_zero;             /* pre-zero mode requested? */
+  boolean dirty;                /* do current buffer contents need written? */
+  boolean b_s_open;             /* is backing-store data valid? */
+  jvirt_sarray_ptr next;        /* link to next virtual sarray control block */
+  backing_store_info b_s_info;  /* System-dependent control info */
+};
+
+struct jvirt_barray_control {
+  JBLOCKARRAY mem_buffer;       /* => the in-memory buffer */
+  JDIMENSION rows_in_array;     /* total virtual array height */
+  JDIMENSION blocksperrow;      /* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;         /* max rows accessed by access_virt_barray */
+  JDIMENSION rows_in_mem;       /* height of memory buffer */
+  JDIMENSION rowsperchunk;      /* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;     /* first logical row # in the buffer */
+  JDIMENSION first_undef_row;   /* row # of first uninitialized row */
+  boolean pre_zero;             /* pre-zero mode requested? */
+  boolean dirty;                /* do current buffer contents need written? */
+  boolean b_s_open;             /* is backing-store data valid? */
+  jvirt_barray_ptr next;        /* link to next virtual barray control block */
+  backing_store_info b_s_info;  /* System-dependent control info */
+};
+
+
+#ifdef MEM_STATS                /* optional extra stuff for statistics */
+
+LOCAL(void)
+print_mem_stats(j_common_ptr cinfo, int pool_id)
+{
+  my_mem_ptr mem = (my_mem_ptr)cinfo->mem;
+  small_pool_ptr shdr_ptr;
+  large_pool_ptr lhdr_ptr;
+
+  /* Since this is only a debugging stub, we can cheat a little by using
+   * fprintf directly rather than going through the trace message code.
+   * This is helpful because message parm array can't handle longs.
+   */
+  fprintf(stderr, "Freeing pool %d, total space = %ld\n",
+          pool_id, mem->total_space_allocated);
+
+  for (lhdr_ptr = mem->large_listpool_id; lhdr_ptr != NULL;
+       lhdr_ptr = lhdr_ptr->next) {
+    fprintf(stderr, "  Large chunk used %ld\n", (long)lhdr_ptr->bytes_used);
+  }
+
+  for (shdr_ptr = mem->small_listpool_id; shdr_ptr != NULL;
+       shdr_ptr = shdr_ptr->next) {
+    fprintf(stderr, "  Small chunk used %ld free %ld\n",
+            (long)shdr_ptr->bytes_used, (long)shdr_ptr->bytes_left);
+  }
+}
+
+#endif /* MEM_STATS */
+
+
+LOCAL(void)
+out_of_memory(j_common_ptr cinfo, int which)
+/* Report an out-of-memory error and stop execution */
+/* If we compiled MEM_STATS support, report alloc requests before dying */
+{
+#ifdef MEM_STATS
+  cinfo->err->trace_level = 2;  /* force self_destruct to report stats */
+#endif
+  ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, which);
+}
+
+
+/*
+ * Allocation of "small" objects.
+ *
+ * For these, we use pooled storage.  When a new pool must be created,
+ * we try to get enough space for the current request plus a "slop" factor,
+ * where the slop will be the amount of leftover space in the new pool.
+ * The speed vs. space tradeoff is largely determined by the slop values.
+ * A different slop value is provided for each pool class (lifetime),
+ * and we also distinguish the first pool of a class from later ones.
+ * NOTE: the values given work fairly well on both 16- and 32-bit-int
+ * machines, but may be too small if longs are 64 bits or more.
+ *
+ * Since we do not know what alignment malloc() gives us, we have to
+ * allocate ALIGN_SIZE-1 extra space per pool to have room for alignment
+ * adjustment.
+ */
+
+static const size_t first_pool_slopJPOOL_NUMPOOLS = {
+  1600,                         /* first PERMANENT pool */
+  16000                         /* first IMAGE pool */
+};
+
+static const size_t extra_pool_slopJPOOL_NUMPOOLS = {
+  0,                            /* additional PERMANENT pools */
+  5000                          /* additional IMAGE pools */
+};
+
+#define MIN_SLOP  50            /* greater than 0 to avoid futile looping */
+
+
+METHODDEF(void *)
+alloc_small(j_common_ptr cinfo, int pool_id, size_t sizeofobject)
+/* Allocate a "small" object */
+{
+  my_mem_ptr mem = (my_mem_ptr)cinfo->mem;
+  small_pool_ptr hdr_ptr, prev_hdr_ptr;
+  char *data_ptr;
+  size_t min_request, slop;
+
+  /*
+   * Round up the requested size to a multiple of ALIGN_SIZE in order
+   * to assure alignment for the next object allocated in the same pool
+   * and so that algorithms can straddle outside the proper area up
+   * to the next alignment.
+   */
+  if (sizeofobject > MAX_ALLOC_CHUNK) {
+    /* This prevents overflow/wrap-around in round_up_pow2() if sizeofobject
+       is close to SIZE_MAX. */
+    out_of_memory(cinfo, 7);
+  }
+  sizeofobject = round_up_pow2(sizeofobject, ALIGN_SIZE);
+
+  /* Check for unsatisfiable request (do now to ensure no overflow below) */
+  if ((sizeof(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) >
+      MAX_ALLOC_CHUNK)
+    out_of_memory(cinfo, 1);    /* request exceeds malloc's ability */
+
+  /* See if space is available in any existing pool */
+  if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
+  prev_hdr_ptr = NULL;
+  hdr_ptr = mem->small_listpool_id;
+  while (hdr_ptr != NULL) {
+    if (hdr_ptr->bytes_left >= sizeofobject)
+      break;                    /* found pool with enough space */
+    prev_hdr_ptr = hdr_ptr;
+    hdr_ptr = hdr_ptr->next;
+  }
+
+  /* Time to make a new pool? */
+  if (hdr_ptr == NULL) {
+    /* min_request is what we need now, slop is what will be leftover */
+    min_request = sizeof(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1;
+    if (prev_hdr_ptr == NULL)   /* first pool in class? */
+      slop = first_pool_sloppool_id;
+    else
+      slop = extra_pool_sloppool_id;
+    /* Don't ask for more than MAX_ALLOC_CHUNK */
+    if (slop > (size_t)(MAX_ALLOC_CHUNK - min_request))
+      slop = (size_t)(MAX_ALLOC_CHUNK - min_request);
+    /* Try to get space, if fail reduce slop and try again */
+    for (;;) {
+      hdr_ptr = (small_pool_ptr)jpeg_get_small(cinfo, min_request + slop);
+      if (hdr_ptr != NULL)
+        break;
+      slop /= 2;
+      if (slop < MIN_SLOP)      /* give up when it gets real small */
+        out_of_memory(cinfo, 2); /* jpeg_get_small failed */
+    }
+    mem->total_space_allocated += min_request + slop;
+    /* Success, initialize the new pool header and add to end of list */
+    hdr_ptr->next = NULL;
+    hdr_ptr->bytes_used = 0;
+    hdr_ptr->bytes_left = sizeofobject + slop;
+    if (prev_hdr_ptr == NULL)   /* first pool in class? */
+      mem->small_listpool_id = hdr_ptr;
+    else
+      prev_hdr_ptr->next = hdr_ptr;
+  }
+
+  /* OK, allocate the object from the current pool */
+  data_ptr = (char *)hdr_ptr; /* point to first data byte in pool... */
+  data_ptr += sizeof(small_pool_hdr); /* ...by skipping the header... */
+  if ((size_t)data_ptr % ALIGN_SIZE) /* ...and adjust for alignment */
+    data_ptr += ALIGN_SIZE - (size_t)data_ptr % ALIGN_SIZE;
+  data_ptr += hdr_ptr->bytes_used; /* point to place for object */
+  hdr_ptr->bytes_used += sizeofobject;
+  hdr_ptr->bytes_left -= sizeofobject;
+
+  return (void *)data_ptr;
+}
+
+
+/*
+ * Allocation of "large" objects.
+ *
+ * The external semantics of these are the same as "small" objects.  However,
+ * the pool management heuristics are quite different.  We assume that each
+ * request is large enough that it may as well be passed directly to
+ * jpeg_get_large; the pool management just links everything together
+ * so that we can free it all on demand.
+ * Note: the major use of "large" objects is in JSAMPARRAY and JBLOCKARRAY
+ * structures.  The routines that create these structures (see below)
+ * deliberately bunch rows together to ensure a large request size.
+ */
+
+METHODDEF(void *)
+alloc_large(j_common_ptr cinfo, int pool_id, size_t sizeofobject)
+/* Allocate a "large" object */
+{
+  my_mem_ptr mem = (my_mem_ptr)cinfo->mem;
+  large_pool_ptr hdr_ptr;
+  char *data_ptr;
+
+  /*
+   * Round up the requested size to a multiple of ALIGN_SIZE so that
+   * algorithms can straddle outside the proper area up to the next
+   * alignment.
+   */
+  if (sizeofobject > MAX_ALLOC_CHUNK) {
+    /* This prevents overflow/wrap-around in round_up_pow2() if sizeofobject
+       is close to SIZE_MAX. */
+    out_of_memory(cinfo, 8);
+  }
+  sizeofobject = round_up_pow2(sizeofobject, ALIGN_SIZE);
+
+  /* Check for unsatisfiable request (do now to ensure no overflow below) */
+  if ((sizeof(large_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) >
+      MAX_ALLOC_CHUNK)
+    out_of_memory(cinfo, 3);    /* request exceeds malloc's ability */
+
+  /* Always make a new pool */
+  if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
+
+  hdr_ptr = (large_pool_ptr)jpeg_get_large(cinfo, sizeofobject +
+                                           sizeof(large_pool_hdr) +
+                                           ALIGN_SIZE - 1);
+  if (hdr_ptr == NULL)
+    out_of_memory(cinfo, 4);    /* jpeg_get_large failed */
+  mem->total_space_allocated += sizeofobject + sizeof(large_pool_hdr) +
+                                ALIGN_SIZE - 1;
+
+  /* Success, initialize the new pool header and add to list */
+  hdr_ptr->next = mem->large_listpool_id;
+  /* We maintain space counts in each pool header for statistical purposes,
+   * even though they are not needed for allocation.
+   */
+  hdr_ptr->bytes_used = sizeofobject;
+  hdr_ptr->bytes_left = 0;
+  mem->large_listpool_id = hdr_ptr;
+
+  data_ptr = (char *)hdr_ptr; /* point to first data byte in pool... */
+  data_ptr += sizeof(small_pool_hdr); /* ...by skipping the header... */
+  if ((size_t)data_ptr % ALIGN_SIZE) /* ...and adjust for alignment */
+    data_ptr += ALIGN_SIZE - (size_t)data_ptr % ALIGN_SIZE;
+
+  return (void *)data_ptr;
+}
+
+
+/*
+ * Creation of 2-D sample arrays.
+ *
+ * To minimize allocation overhead and to allow I/O of large contiguous
+ * blocks, we allocate the sample rows in groups of as many rows as possible
+ * without exceeding MAX_ALLOC_CHUNK total bytes per allocation request.
+ * NB: the virtual array control routines, later in this file, know about
+ * this chunking of rows.  The rowsperchunk value is left in the mem manager
+ * object so that it can be saved away if this sarray is the workspace for
+ * a virtual array.
+ *
+ * Since we are often upsampling with a factor 2, we align the size (not
+ * the start) to 2 * ALIGN_SIZE so that the upsampling routines don't have
+ * to be as careful about size.
+ */
+
+METHODDEF(JSAMPARRAY)
+alloc_sarray(j_common_ptr cinfo, int pool_id, JDIMENSION samplesperrow,
+             JDIMENSION numrows)
+/* Allocate a 2-D sample array */
+{
+  my_mem_ptr mem = (my_mem_ptr)cinfo->mem;
+  JSAMPARRAY result;
+  JSAMPROW workspace;
+  JDIMENSION rowsperchunk, currow, i;
+  long ltemp;
+
+  /* Make sure each row is properly aligned */
+  if ((ALIGN_SIZE % sizeof(JSAMPLE)) != 0)
+    out_of_memory(cinfo, 5);    /* safety check */
+
+  if (samplesperrow > MAX_ALLOC_CHUNK) {
+    /* This prevents overflow/wrap-around in round_up_pow2() if sizeofobject
+       is close to SIZE_MAX. */
+    out_of_memory(cinfo, 9);
+  }
+  samplesperrow = (JDIMENSION)round_up_pow2(samplesperrow, (2 * ALIGN_SIZE) /
+                                                           sizeof(JSAMPLE));
+
+  /* Calculate max # of rows allowed in one allocation chunk */
+  ltemp = (MAX_ALLOC_CHUNK - sizeof(large_pool_hdr)) /
+          ((long)samplesperrow * sizeof(JSAMPLE));
+  if (ltemp <= 0)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+  if (ltemp < (long)numrows)
+    rowsperchunk = (JDIMENSION)ltemp;
+  else
+    rowsperchunk = numrows;
+  mem->last_rowsperchunk = rowsperchunk;
+
+  /* Get space for row pointers (small object) */
+  result = (JSAMPARRAY)alloc_small(cinfo, pool_id,
+                                   (size_t)(numrows * sizeof(JSAMPROW)));
+
+  /* Get the rows themselves (large objects) */
+  currow = 0;
+  while (currow < numrows) {
+    rowsperchunk = MIN(rowsperchunk, numrows - currow);
+    workspace = (JSAMPROW)alloc_large(cinfo, pool_id,
+      (size_t)((size_t)rowsperchunk * (size_t)samplesperrow *
+               sizeof(JSAMPLE)));
+    for (i = rowsperchunk; i > 0; i--) {
+      resultcurrow++ = workspace;
+      workspace += samplesperrow;
+    }
+  }
+
+  return result;
+}
+
+
+/*
+ * Creation of 2-D coefficient-block arrays.
+ * This is essentially the same as the code for sample arrays, above.
+ */
+
+METHODDEF(JBLOCKARRAY)
+alloc_barray(j_common_ptr cinfo, int pool_id, JDIMENSION blocksperrow,
+             JDIMENSION numrows)
+/* Allocate a 2-D coefficient-block array */
+{
+  my_mem_ptr mem = (my_mem_ptr)cinfo->mem;
+  JBLOCKARRAY result;
+  JBLOCKROW workspace;
+  JDIMENSION rowsperchunk, currow, i;
+  long ltemp;
+
+  /* Make sure each row is properly aligned */
+  if ((sizeof(JBLOCK) % ALIGN_SIZE) != 0)
+    out_of_memory(cinfo, 6);    /* safety check */
+
+  /* Calculate max # of rows allowed in one allocation chunk */
+  ltemp = (MAX_ALLOC_CHUNK - sizeof(large_pool_hdr)) /
+          ((long)blocksperrow * sizeof(JBLOCK));
+  if (ltemp <= 0)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+  if (ltemp < (long)numrows)
+    rowsperchunk = (JDIMENSION)ltemp;
+  else
+    rowsperchunk = numrows;
+  mem->last_rowsperchunk = rowsperchunk;
+
+  /* Get space for row pointers (small object) */
+  result = (JBLOCKARRAY)alloc_small(cinfo, pool_id,
+                                    (size_t)(numrows * sizeof(JBLOCKROW)));
+
+  /* Get the rows themselves (large objects) */
+  currow = 0;
+  while (currow < numrows) {
+    rowsperchunk = MIN(rowsperchunk, numrows - currow);
+    workspace = (JBLOCKROW)alloc_large(cinfo, pool_id,
+        (size_t)((size_t)rowsperchunk * (size_t)blocksperrow *
+                  sizeof(JBLOCK)));
+    for (i = rowsperchunk; i > 0; i--) {
+      resultcurrow++ = workspace;
+      workspace += blocksperrow;
+    }
+  }
+
+  return result;
+}
+
+
+/*
+ * About virtual array management:
+ *
+ * The above "normal" array routines are only used to allocate strip buffers
+ * (as wide as the image, but just a few rows high).  Full-image-sized buffers
+ * are handled as "virtual" arrays.  The array is still accessed a strip at a
+ * time, but the memory manager must save the whole array for repeated
+ * accesses.  The intended implementation is that there is a strip buffer in
+ * memory (as high as is possible given the desired memory limit), plus a
+ * backing file that holds the rest of the array.
+ *
+ * The request_virt_array routines are told the total size of the image and
+ * the maximum number of rows that will be accessed at once.  The in-memory
+ * buffer must be at least as large as the maxaccess value.
+ *
+ * The request routines create control blocks but not the in-memory buffers.
+ * That is postponed until realize_virt_arrays is called.  At that time the
+ * total amount of space needed is known (approximately, anyway), so free
+ * memory can be divided up fairly.
+ *
+ * The access_virt_array routines are responsible for making a specific strip
+ * area accessible (after reading or writing the backing file, if necessary).
+ * Note that the access routines are told whether the caller intends to modify
+ * the accessed strip; during a read-only pass this saves having to rewrite
+ * data to disk.  The access routines are also responsible for pre-zeroing
+ * any newly accessed rows, if pre-zeroing was requested.
+ *
+ * In current usage, the access requests are usually for nonoverlapping
+ * strips; that is, successive access start_row numbers differ by exactly
+ * num_rows = maxaccess.  This means we can get good performance with simple
+ * buffer dump/reload logic, by making the in-memory buffer be a multiple
+ * of the access height; then there will never be accesses across bufferload
+ * boundaries.  The code will still work with overlapping access requests,
+ * but it doesn't handle bufferload overlaps very efficiently.
+ */
+
+
+METHODDEF(jvirt_sarray_ptr)
+request_virt_sarray(j_common_ptr cinfo, int pool_id, boolean pre_zero,
+                    JDIMENSION samplesperrow, JDIMENSION numrows,
+                    JDIMENSION maxaccess)
+/* Request a virtual 2-D sample array */
+{
+  my_mem_ptr mem = (my_mem_ptr)cinfo->mem;
+  jvirt_sarray_ptr result;
+
+  /* Only IMAGE-lifetime virtual arrays are currently supported */
+  if (pool_id != JPOOL_IMAGE)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
+
+  /* get control block */
+  result = (jvirt_sarray_ptr)alloc_small(cinfo, pool_id,
+                                         sizeof(struct jvirt_sarray_control));
+
+  result->mem_buffer = NULL;    /* marks array not yet realized */
+  result->rows_in_array = numrows;
+  result->samplesperrow = samplesperrow;
+  result->maxaccess = maxaccess;
+  result->pre_zero = pre_zero;
+  result->b_s_open = FALSE;     /* no associated backing-store object */
+  result->next = mem->virt_sarray_list; /* add to list of virtual arrays */
+  mem->virt_sarray_list = result;
+
+  return result;
+}
+
+
+METHODDEF(jvirt_barray_ptr)
+request_virt_barray(j_common_ptr cinfo, int pool_id, boolean pre_zero,
+                    JDIMENSION blocksperrow, JDIMENSION numrows,
+                    JDIMENSION maxaccess)
+/* Request a virtual 2-D coefficient-block array */
+{
+  my_mem_ptr mem = (my_mem_ptr)cinfo->mem;
+  jvirt_barray_ptr result;
+
+  /* Only IMAGE-lifetime virtual arrays are currently supported */
+  if (pool_id != JPOOL_IMAGE)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
+
+  /* get control block */
+  result = (jvirt_barray_ptr)alloc_small(cinfo, pool_id,
+                                         sizeof(struct jvirt_barray_control));
+
+  result->mem_buffer = NULL;    /* marks array not yet realized */
+  result->rows_in_array = numrows;
+  result->blocksperrow = blocksperrow;
+  result->maxaccess = maxaccess;
+  result->pre_zero = pre_zero;
+  result->b_s_open = FALSE;     /* no associated backing-store object */
+  result->next = mem->virt_barray_list; /* add to list of virtual arrays */
+  mem->virt_barray_list = result;
+
+  return result;
+}
+
+
+METHODDEF(void)
+realize_virt_arrays(j_common_ptr cinfo)
+/* Allocate the in-memory buffers for any unrealized virtual arrays */
+{
+  my_mem_ptr mem = (my_mem_ptr)cinfo->mem;
+  size_t space_per_minheight, maximum_space, avail_mem;
+  size_t minheights, max_minheights;
+  jvirt_sarray_ptr sptr;
+  jvirt_barray_ptr bptr;
+
+  /* Compute the minimum space needed (maxaccess rows in each buffer)
+   * and the maximum space needed (full image height in each buffer).
+   * These may be of use to the system-dependent jpeg_mem_available routine.
+   */
+  space_per_minheight = 0;
+  maximum_space = 0;
+  for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
+    if (sptr->mem_buffer == NULL) { /* if not realized yet */
+      size_t new_space = (long)sptr->rows_in_array *
+                         (long)sptr->samplesperrow * sizeof(JSAMPLE);
+
+      space_per_minheight += (long)sptr->maxaccess *
+                             (long)sptr->samplesperrow * sizeof(JSAMPLE);
+      if (SIZE_MAX - maximum_space < new_space)
+        out_of_memory(cinfo, 10);
+      maximum_space += new_space;
+    }
+  }
+  for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
+    if (bptr->mem_buffer == NULL) { /* if not realized yet */
+      size_t new_space = (long)bptr->rows_in_array *
+                         (long)bptr->blocksperrow * sizeof(JBLOCK);
+
+      space_per_minheight += (long)bptr->maxaccess *
+                             (long)bptr->blocksperrow * sizeof(JBLOCK);
+      if (SIZE_MAX - maximum_space < new_space)
+        out_of_memory(cinfo, 11);
+      maximum_space += new_space;
+    }
+  }
+
+  if (space_per_minheight <= 0)
+    return;                     /* no unrealized arrays, no work */
+
+  /* Determine amount of memory to actually use; this is system-dependent. */
+  avail_mem = jpeg_mem_available(cinfo, space_per_minheight, maximum_space,
+                                 mem->total_space_allocated);
+
+  /* If the maximum space needed is available, make all the buffers full
+   * height; otherwise parcel it out with the same number of minheights
+   * in each buffer.
+   */
+  if (avail_mem >= maximum_space)
+    max_minheights = 1000000000L;
+  else {
+    max_minheights = avail_mem / space_per_minheight;
+    /* If there doesn't seem to be enough space, try to get the minimum
+     * anyway.  This allows a "stub" implementation of jpeg_mem_available().
+     */
+    if (max_minheights <= 0)
+      max_minheights = 1;
+  }
+
+  /* Allocate the in-memory buffers and initialize backing store as needed. */
+
+  for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
+    if (sptr->mem_buffer == NULL) { /* if not realized yet */
+      minheights = ((long)sptr->rows_in_array - 1L) / sptr->maxaccess + 1L;
+      if (minheights <= max_minheights) {
+        /* This buffer fits in memory */
+        sptr->rows_in_mem = sptr->rows_in_array;
+      } else {
+        /* It doesn't fit in memory, create backing store. */
+        sptr->rows_in_mem = (JDIMENSION)(max_minheights * sptr->maxaccess);
+        jpeg_open_backing_store(cinfo, &sptr->b_s_info,
+                                (long)sptr->rows_in_array *
+                                (long)sptr->samplesperrow *
+                                (long)sizeof(JSAMPLE));
+        sptr->b_s_open = TRUE;
+      }
+      sptr->mem_buffer = alloc_sarray(cinfo, JPOOL_IMAGE,
+                                      sptr->samplesperrow, sptr->rows_in_mem);
+      sptr->rowsperchunk = mem->last_rowsperchunk;
+      sptr->cur_start_row = 0;
+      sptr->first_undef_row = 0;
+      sptr->dirty = FALSE;
+    }
+  }
+
+  for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
+    if (bptr->mem_buffer == NULL) { /* if not realized yet */
+      minheights = ((long)bptr->rows_in_array - 1L) / bptr->maxaccess + 1L;
+      if (minheights <= max_minheights) {
+        /* This buffer fits in memory */
+        bptr->rows_in_mem = bptr->rows_in_array;
+      } else {
+        /* It doesn't fit in memory, create backing store. */
+        bptr->rows_in_mem = (JDIMENSION)(max_minheights * bptr->maxaccess);
+        jpeg_open_backing_store(cinfo, &bptr->b_s_info,
+                                (long)bptr->rows_in_array *
+                                (long)bptr->blocksperrow *
+                                (long)sizeof(JBLOCK));
+        bptr->b_s_open = TRUE;
+      }
+      bptr->mem_buffer = alloc_barray(cinfo, JPOOL_IMAGE,
+                                      bptr->blocksperrow, bptr->rows_in_mem);
+      bptr->rowsperchunk = mem->last_rowsperchunk;
+      bptr->cur_start_row = 0;
+      bptr->first_undef_row = 0;
+      bptr->dirty = FALSE;
+    }
+  }
+}
+
+
+LOCAL(void)
+do_sarray_io(j_common_ptr cinfo, jvirt_sarray_ptr ptr, boolean writing)
+/* Do backing store read or write of a virtual sample array */
+{
+  long bytesperrow, file_offset, byte_count, rows, thisrow, i;
+
+  bytesperrow = (long)ptr->samplesperrow * sizeof(JSAMPLE);
+  file_offset = ptr->cur_start_row * bytesperrow;
+  /* Loop to read or write each allocation chunk in mem_buffer */
+  for (i = 0; i < (long)ptr->rows_in_mem; i += ptr->rowsperchunk) {
+    /* One chunk, but check for short chunk at end of buffer */
+    rows = MIN((long)ptr->rowsperchunk, (long)ptr->rows_in_mem - i);
+    /* Transfer no more than is currently defined */
+    thisrow = (long)ptr->cur_start_row + i;
+    rows = MIN(rows, (long)ptr->first_undef_row - thisrow);
+    /* Transfer no more than fits in file */
+    rows = MIN(rows, (long)ptr->rows_in_array - thisrow);
+    if (rows <= 0)              /* this chunk might be past end of file! */
+      break;
+    byte_count = rows * bytesperrow;
+    if (writing)
+      (*ptr->b_s_info.write_backing_store) (cinfo, &ptr->b_s_info,
+                                            (void *)ptr->mem_bufferi,
+                                            file_offset, byte_count);
+    else
+      (*ptr->b_s_info.read_backing_store) (cinfo, &ptr->b_s_info,
+                                           (void *)ptr->mem_bufferi,
+                                           file_offset, byte_count);
+    file_offset += byte_count;
+  }
+}
+
+
+LOCAL(void)
+do_barray_io(j_common_ptr cinfo, jvirt_barray_ptr ptr, boolean writing)
+/* Do backing store read or write of a virtual coefficient-block array */
+{
+  long bytesperrow, file_offset, byte_count, rows, thisrow, i;
+
+  bytesperrow = (long)ptr->blocksperrow * sizeof(JBLOCK);
+  file_offset = ptr->cur_start_row * bytesperrow;
+  /* Loop to read or write each allocation chunk in mem_buffer */
+  for (i = 0; i < (long)ptr->rows_in_mem; i += ptr->rowsperchunk) {
+    /* One chunk, but check for short chunk at end of buffer */
+    rows = MIN((long)ptr->rowsperchunk, (long)ptr->rows_in_mem - i);
+    /* Transfer no more than is currently defined */
+    thisrow = (long)ptr->cur_start_row + i;
+    rows = MIN(rows, (long)ptr->first_undef_row - thisrow);
+    /* Transfer no more than fits in file */
+    rows = MIN(rows, (long)ptr->rows_in_array - thisrow);
+    if (rows <= 0)              /* this chunk might be past end of file! */
+      break;
+    byte_count = rows * bytesperrow;
+    if (writing)
+      (*ptr->b_s_info.write_backing_store) (cinfo, &ptr->b_s_info,
+                                            (void *)ptr->mem_bufferi,
+                                            file_offset, byte_count);
+    else
+      (*ptr->b_s_info.read_backing_store) (cinfo, &ptr->b_s_info,
+                                           (void *)ptr->mem_bufferi,
+                                           file_offset, byte_count);
+    file_offset += byte_count;
+  }
+}
+
+
+METHODDEF(JSAMPARRAY)
+access_virt_sarray(j_common_ptr cinfo, jvirt_sarray_ptr ptr,
+                   JDIMENSION start_row, JDIMENSION num_rows, boolean writable)
+/* Access the part of a virtual sample array starting at start_row */
+/* and extending for num_rows rows.  writable is true if  */
+/* caller intends to modify the accessed area. */
+{
+  JDIMENSION end_row = start_row + num_rows;
+  JDIMENSION undef_row;
+
+  /* debugging check */
+  if (end_row > ptr->rows_in_array || num_rows > ptr->maxaccess ||
+      ptr->mem_buffer == NULL)
+    ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+
+  /* Make the desired part of the virtual array accessible */
+  if (start_row < ptr->cur_start_row ||
+      end_row > ptr->cur_start_row + ptr->rows_in_mem) {
+    if (!ptr->b_s_open)
+      ERREXIT(cinfo, JERR_VIRTUAL_BUG);
+    /* Flush old buffer contents if necessary */
+    if (ptr->dirty) {
+      do_sarray_io(cinfo, ptr, TRUE);
+      ptr->dirty = FALSE;
+    }
+    /* Decide what part of virtual array to access.
+     * Algorithm: if target address > current window, assume forward scan,
+     * load starting at target address.  If target address < current window,
+     * assume backward scan, load so that target area is top of window.
+     * Note that when switching from forward write to forward read, will have
+     * start_row = 0, so the limiting case applies and we load from 0 anyway.
+     */
+    if (start_row > ptr->cur_start_row) {
+      ptr->cur_start_row = start_row;
+    } else {
+      /* use long arithmetic here to avoid overflow & unsigned problems */
+      long ltemp;
+
+      ltemp = (long)end_row - (long)ptr->rows_in_mem;
+      if (ltemp < 0)
+        ltemp = 0;              /* don't fall off front end of file */
+      ptr->cur_start_row = (JDIMENSION)ltemp;
+    }
+    /* Read in the selected part of the array.
+     * During the initial write pass, we will do no actual read
+     * because the selected part is all undefined.
+     */
+    do_sarray_io(cinfo, ptr, FALSE);
+  }
+  /* Ensure the accessed part of the array is defined; prezero if needed.
+   * To improve locality of access, we only prezero the part of the array
+   * that the caller is about to access, not the entire in-memory array.
+   */
+  if (ptr->first_undef_row < end_row) {
+    if (ptr->first_undef_row < start_row) {
+      if (writable)             /* writer skipped over a section of array */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;    /* but reader is allowed to read ahead */
+    } else {
+      undef_row = ptr->first_undef_row;
+    }
+    if (writable)
+      ptr->first_undef_row = end_row;
+    if (ptr->pre_zero) {
+      size_t bytesperrow = (size_t)ptr->samplesperrow * sizeof(JSAMPLE);
+      undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
+      end_row -= ptr->cur_start_row;
+      while (undef_row < end_row) {
+        jzero_far((void *)ptr->mem_bufferundef_row, bytesperrow);
+        undef_row++;
+      }
+    } else {
+      if (!writable)            /* reader looking at undefined data */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+    }
+  }
+  /* Flag the buffer dirty if caller will write in it */
+  if (writable)
+    ptr->dirty = TRUE;
+  /* Return address of proper part of the buffer */
+  return ptr->mem_buffer + (start_row - ptr->cur_start_row);
+}
+
+
+METHODDEF(JBLOCKARRAY)
+access_virt_barray(j_common_ptr cinfo, jvirt_barray_ptr ptr,
+                   JDIMENSION start_row, JDIMENSION num_rows, boolean writable)
+/* Access the part of a virtual block array starting at start_row */
+/* and extending for num_rows rows.  writable is true if  */
+/* caller intends to modify the accessed area. */
+{
+  JDIMENSION end_row = start_row + num_rows;
+  JDIMENSION undef_row;
+
+  /* debugging check */
+  if (end_row > ptr->rows_in_array || num_rows > ptr->maxaccess ||
+      ptr->mem_buffer == NULL)
+    ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+
+  /* Make the desired part of the virtual array accessible */
+  if (start_row < ptr->cur_start_row ||
+      end_row > ptr->cur_start_row + ptr->rows_in_mem) {
+    if (!ptr->b_s_open)
+      ERREXIT(cinfo, JERR_VIRTUAL_BUG);
+    /* Flush old buffer contents if necessary */
+    if (ptr->dirty) {
+      do_barray_io(cinfo, ptr, TRUE);
+      ptr->dirty = FALSE;
+    }
+    /* Decide what part of virtual array to access.
+     * Algorithm: if target address > current window, assume forward scan,
+     * load starting at target address.  If target address < current window,
+     * assume backward scan, load so that target area is top of window.
+     * Note that when switching from forward write to forward read, will have
+     * start_row = 0, so the limiting case applies and we load from 0 anyway.
+     */
+    if (start_row > ptr->cur_start_row) {
+      ptr->cur_start_row = start_row;
+    } else {
+      /* use long arithmetic here to avoid overflow & unsigned problems */
+      long ltemp;
+
+      ltemp = (long)end_row - (long)ptr->rows_in_mem;
+      if (ltemp < 0)
+        ltemp = 0;              /* don't fall off front end of file */
+      ptr->cur_start_row = (JDIMENSION)ltemp;
+    }
+    /* Read in the selected part of the array.
+     * During the initial write pass, we will do no actual read
+     * because the selected part is all undefined.
+     */
+    do_barray_io(cinfo, ptr, FALSE);
+  }
+  /* Ensure the accessed part of the array is defined; prezero if needed.
+   * To improve locality of access, we only prezero the part of the array
+   * that the caller is about to access, not the entire in-memory array.
+   */
+  if (ptr->first_undef_row < end_row) {
+    if (ptr->first_undef_row < start_row) {
+      if (writable)             /* writer skipped over a section of array */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;    /* but reader is allowed to read ahead */
+    } else {
+      undef_row = ptr->first_undef_row;
+    }
+    if (writable)
+      ptr->first_undef_row = end_row;
+    if (ptr->pre_zero) {
+      size_t bytesperrow = (size_t)ptr->blocksperrow * sizeof(JBLOCK);
+      undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
+      end_row -= ptr->cur_start_row;
+      while (undef_row < end_row) {
+        jzero_far((void *)ptr->mem_bufferundef_row, bytesperrow);
+        undef_row++;
+      }
+    } else {
+      if (!writable)            /* reader looking at undefined data */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+    }
+  }
+  /* Flag the buffer dirty if caller will write in it */
+  if (writable)
+    ptr->dirty = TRUE;
+  /* Return address of proper part of the buffer */
+  return ptr->mem_buffer + (start_row - ptr->cur_start_row);
+}
+
+
+/*
+ * Release all objects belonging to a specified pool.
+ */
+
+METHODDEF(void)
+free_pool(j_common_ptr cinfo, int pool_id)
+{
+  my_mem_ptr mem = (my_mem_ptr)cinfo->mem;
+  small_pool_ptr shdr_ptr;
+  large_pool_ptr lhdr_ptr;
+  size_t space_freed;
+
+  if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
+
+#ifdef MEM_STATS
+  if (cinfo->err->trace_level > 1)
+    print_mem_stats(cinfo, pool_id); /* print pool's memory usage statistics */
+#endif
+
+  /* If freeing IMAGE pool, close any virtual arrays first */
+  if (pool_id == JPOOL_IMAGE) {
+    jvirt_sarray_ptr sptr;
+    jvirt_barray_ptr bptr;
+
+    for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
+      if (sptr->b_s_open) {     /* there may be no backing store */
+        sptr->b_s_open = FALSE; /* prevent recursive close if error */
+        (*sptr->b_s_info.close_backing_store) (cinfo, &sptr->b_s_info);
+      }
+    }
+    mem->virt_sarray_list = NULL;
+    for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
+      if (bptr->b_s_open) {     /* there may be no backing store */
+        bptr->b_s_open = FALSE; /* prevent recursive close if error */
+        (*bptr->b_s_info.close_backing_store) (cinfo, &bptr->b_s_info);
+      }
+    }
+    mem->virt_barray_list = NULL;
+  }
+
+  /* Release large objects */
+  lhdr_ptr = mem->large_listpool_id;
+  mem->large_listpool_id = NULL;
+
+  while (lhdr_ptr != NULL) {
+    large_pool_ptr next_lhdr_ptr = lhdr_ptr->next;
+    space_freed = lhdr_ptr->bytes_used +
+                  lhdr_ptr->bytes_left +
+                  sizeof(large_pool_hdr) + ALIGN_SIZE - 1;
+    jpeg_free_large(cinfo, (void *)lhdr_ptr, space_freed);
+    mem->total_space_allocated -= space_freed;
+    lhdr_ptr = next_lhdr_ptr;
+  }
+
+  /* Release small objects */
+  shdr_ptr = mem->small_listpool_id;
+  mem->small_listpool_id = NULL;
+
+  while (shdr_ptr != NULL) {
+    small_pool_ptr next_shdr_ptr = shdr_ptr->next;
+    space_freed = shdr_ptr->bytes_used + shdr_ptr->bytes_left +
+                  sizeof(small_pool_hdr) + ALIGN_SIZE - 1;
+    jpeg_free_small(cinfo, (void *)shdr_ptr, space_freed);
+    mem->total_space_allocated -= space_freed;
+    shdr_ptr = next_shdr_ptr;
+  }
+}
+
+
+/*
+ * Close up shop entirely.
+ * Note that this cannot be called unless cinfo->mem is non-NULL.
+ */
+
+METHODDEF(void)
+self_destruct(j_common_ptr cinfo)
+{
+  int pool;
+
+  /* Close all backing store, release all memory.
+   * Releasing pools in reverse order might help avoid fragmentation
+   * with some (brain-damaged) malloc libraries.
+   */
+  for (pool = JPOOL_NUMPOOLS - 1; pool >= JPOOL_PERMANENT; pool--) {
+    free_pool(cinfo, pool);
+  }
+
+  /* Release the memory manager control block too. */
+  jpeg_free_small(cinfo, (void *)cinfo->mem, sizeof(my_memory_mgr));
+  cinfo->mem = NULL;            /* ensures I will be called only once */
+
+  jpeg_mem_term(cinfo);         /* system-dependent cleanup */
+}
+
+
+/*
+ * Memory manager initialization.
+ * When this is called, only the error manager pointer is valid in cinfo!
+ */
+
+GLOBAL(void)
+jinit_memory_mgr(j_common_ptr cinfo)
+{
+  my_mem_ptr mem;
+  long max_to_use;
+  int pool;
+  size_t test_mac;
+
+  cinfo->mem = NULL;            /* for safety if init fails */
+
+  /* Check for configuration errors.
+   * sizeof(ALIGN_TYPE) should be a power of 2; otherwise, it probably
+   * doesn't reflect any real hardware alignment requirement.
+   * The test is a little tricky: for X>0, X and X-1 have no one-bits
+   * in common if and only if X is a power of 2, ie has only one one-bit.
+   * Some compilers may give an "unreachable code" warning here; ignore it.
+   */
+  if ((ALIGN_SIZE & (ALIGN_SIZE - 1)) != 0)
+    ERREXIT(cinfo, JERR_BAD_ALIGN_TYPE);
+  /* MAX_ALLOC_CHUNK must be representable as type size_t, and must be
+   * a multiple of ALIGN_SIZE.
+   * Again, an "unreachable code" warning may be ignored here.
+   * But a "constant too large" warning means you need to fix MAX_ALLOC_CHUNK.
+   */
+  test_mac = (size_t)MAX_ALLOC_CHUNK;
+  if ((long)test_mac != MAX_ALLOC_CHUNK ||
+      (MAX_ALLOC_CHUNK % ALIGN_SIZE) != 0)
+    ERREXIT(cinfo, JERR_BAD_ALLOC_CHUNK);
+
+  max_to_use = jpeg_mem_init(cinfo); /* system-dependent initialization */
+
+  /* Attempt to allocate memory manager's control block */
+  mem = (my_mem_ptr)jpeg_get_small(cinfo, sizeof(my_memory_mgr));
+
+  if (mem == NULL) {
+    jpeg_mem_term(cinfo);       /* system-dependent cleanup */
+    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 0);
+  }
+
+  /* OK, fill in the method pointers */
+  mem->pub.alloc_small = alloc_small;
+  mem->pub.alloc_large = alloc_large;
+  mem->pub.alloc_sarray = alloc_sarray;
+  mem->pub.alloc_barray = alloc_barray;
+  mem->pub.request_virt_sarray = request_virt_sarray;
+  mem->pub.request_virt_barray = request_virt_barray;
+  mem->pub.realize_virt_arrays = realize_virt_arrays;
+  mem->pub.access_virt_sarray = access_virt_sarray;
+  mem->pub.access_virt_barray = access_virt_barray;
+  mem->pub.free_pool = free_pool;
+  mem->pub.self_destruct = self_destruct;
+
+  /* Make MAX_ALLOC_CHUNK accessible to other modules */
+  mem->pub.max_alloc_chunk = MAX_ALLOC_CHUNK;
+
+  /* Initialize working state */
+  mem->pub.max_memory_to_use = max_to_use;
+
+  for (pool = JPOOL_NUMPOOLS - 1; pool >= JPOOL_PERMANENT; pool--) {
+    mem->small_listpool = NULL;
+    mem->large_listpool = NULL;
+  }
+  mem->virt_sarray_list = NULL;
+  mem->virt_barray_list = NULL;
+
+  mem->total_space_allocated = sizeof(my_memory_mgr);
+
+  /* Declare ourselves open for business */
+  cinfo->mem = &mem->pub;
+
+  /* Check for an environment variable JPEGMEM; if found, override the
+   * default max_memory setting from jpeg_mem_init.  Note that the
+   * surrounding application may again override this value.
+   * If your system doesn't support getenv(), define NO_GETENV to disable
+   * this feature.
+   */
+#ifndef NO_GETENV
+  {
+    char *memenv;
+
+    if ((memenv = getenv("JPEGMEM")) != NULL) {
+      char ch = 'x';
+
+      if (sscanf(memenv, "%ld%c", &max_to_use, &ch) > 0) {
+        if (ch == 'm' || ch == 'M')
+          max_to_use *= 1000L;
+        mem->pub.max_memory_to_use = max_to_use * 1000L;
+      }
+    }
+  }
+#endif
+
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jmemnobs.c Added

@@ -0,0 +1,115 @@
+/*
+ * jmemnobs.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1992-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2017-2018, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file provides a really simple implementation of the system-
+ * dependent portion of the JPEG memory manager.  This implementation
+ * assumes that no backing-store files are needed: all required space
+ * can be obtained from malloc().
+ * This is very portable in the sense that it'll compile on almost anything,
+ * but you'd better have lots of main memory (or virtual memory) if you want
+ * to process big images.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"            /* import the system-dependent declarations */
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc(size_t size);
+extern void free(void *ptr);
+#endif
+
+
+/*
+ * Memory allocation and freeing are controlled by the regular library
+ * routines malloc() and free().
+ */
+
+GLOBAL(void *)
+jpeg_get_small(j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *)malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_small(j_common_ptr cinfo, void *object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * "Large" objects are treated the same as "small" ones.
+ */
+
+GLOBAL(void *)
+jpeg_get_large(j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *)malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_large(j_common_ptr cinfo, void *object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * This routine computes the total memory space available for allocation.
+ */
+
+GLOBAL(size_t)
+jpeg_mem_available(j_common_ptr cinfo, size_t min_bytes_needed,
+                   size_t max_bytes_needed, size_t already_allocated)
+{
+  if (cinfo->mem->max_memory_to_use) {
+    if ((size_t)cinfo->mem->max_memory_to_use > already_allocated)
+      return cinfo->mem->max_memory_to_use - already_allocated;
+    else
+      return 0;
+  } else {
+    /* Here we always say, "we got all you want bud!" */
+    return max_bytes_needed;
+  }
+}
+
+
+/*
+ * Backing store (temporary file) management.
+ * Since jpeg_mem_available always promised the moon,
+ * this should never be called and we can just error out.
+ */
+
+GLOBAL(void)
+jpeg_open_backing_store(j_common_ptr cinfo, backing_store_ptr info,
+                        long total_bytes_needed)
+{
+  ERREXIT(cinfo, JERR_NO_BACKING_STORE);
+}
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.  Here, there isn't any.
+ */
+
+GLOBAL(long)
+jpeg_mem_init(j_common_ptr cinfo)
+{
+  return 0;                     /* just set max_memory_to_use to 0 */
+}
+
+GLOBAL(void)
+jpeg_mem_term(j_common_ptr cinfo)
+{
+  /* no work */
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jmemsys.h Added

@@ -0,0 +1,178 @@
+/*
+ * jmemsys.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1992-1997, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This include file defines the interface between the system-independent
+ * and system-dependent portions of the JPEG memory manager.  No other
+ * modules need include it.  (The system-independent portion is jmemmgr.c;
+ * there are several different versions of the system-dependent portion.)
+ *
+ * This file works as-is for the system-dependent memory managers supplied
+ * in the IJG distribution.  You may need to modify it if you write a
+ * custom memory manager.  If system-dependent changes are needed in
+ * this file, the best method is to #ifdef them based on a configuration
+ * symbol supplied in jconfig.h.
+ */
+
+
+/*
+ * These two functions are used to allocate and release small chunks of
+ * memory.  (Typically the total amount requested through jpeg_get_small is
+ * no more than 20K or so; this will be requested in chunks of a few K each.)
+ * Behavior should be the same as for the standard library functions malloc
+ * and free; in particular, jpeg_get_small must return NULL on failure.
+ * On most systems, these ARE malloc and free.  jpeg_free_small is passed the
+ * size of the object being freed, just in case it's needed.
+ */
+
+EXTERN(void *) jpeg_get_small(j_common_ptr cinfo, size_t sizeofobject);
+EXTERN(void) jpeg_free_small(j_common_ptr cinfo, void *object,
+                             size_t sizeofobject);
+
+/*
+ * These two functions are used to allocate and release large chunks of
+ * memory (up to the total free space designated by jpeg_mem_available).
+ * These are identical to the jpeg_get/free_small routines; but we keep them
+ * separate anyway, in case a different allocation strategy is desirable for
+ * large chunks.
+ */
+
+EXTERN(void *) jpeg_get_large(j_common_ptr cinfo, size_t sizeofobject);
+EXTERN(void) jpeg_free_large(j_common_ptr cinfo, void *object,
+                             size_t sizeofobject);
+
+/*
+ * The macro MAX_ALLOC_CHUNK designates the maximum number of bytes that may
+ * be requested in a single call to jpeg_get_large (and jpeg_get_small for that
+ * matter, but that case should never come into play).  This macro was needed
+ * to model the 64Kb-segment-size limit of far addressing on 80x86 machines.
+ * On machines with flat address spaces, any large constant may be used.
+ *
+ * NB: jmemmgr.c expects that MAX_ALLOC_CHUNK will be representable as type
+ * size_t and will be a multiple of sizeof(align_type).
+ */
+
+#ifndef MAX_ALLOC_CHUNK         /* may be overridden in jconfig.h */
+#define MAX_ALLOC_CHUNK  1000000000L
+#endif
+
+/*
+ * This routine computes the total space still available for allocation by
+ * jpeg_get_large.  If more space than this is needed, backing store will be
+ * used.  NOTE: any memory already allocated must not be counted.
+ *
+ * There is a minimum space requirement, corresponding to the minimum
+ * feasible buffer sizes; jmemmgr.c will request that much space even if
+ * jpeg_mem_available returns zero.  The maximum space needed, enough to hold
+ * all working storage in memory, is also passed in case it is useful.
+ * Finally, the total space already allocated is passed.  If no better
+ * method is available, cinfo->mem->max_memory_to_use - already_allocated
+ * is often a suitable calculation.
+ *
+ * It is OK for jpeg_mem_available to underestimate the space available
+ * (that'll just lead to more backing-store access than is really necessary).
+ * However, an overestimate will lead to failure.  Hence it's wise to subtract
+ * a slop factor from the true available space.  5% should be enough.
+ *
+ * On machines with lots of virtual memory, any large constant may be returned.
+ * Conversely, zero may be returned to always use the minimum amount of memory.
+ */
+
+EXTERN(size_t) jpeg_mem_available(j_common_ptr cinfo, size_t min_bytes_needed,
+                                  size_t max_bytes_needed,
+                                  size_t already_allocated);
+
+
+/*
+ * This structure holds whatever state is needed to access a single
+ * backing-store object.  The read/write/close method pointers are called
+ * by jmemmgr.c to manipulate the backing-store object; all other fields
+ * are private to the system-dependent backing store routines.
+ */
+
+#define TEMP_NAME_LENGTH   64   /* max length of a temporary file's name */
+
+
+#ifdef USE_MSDOS_MEMMGR         /* DOS-specific junk */
+
+typedef unsigned short XMSH;    /* type of extended-memory handles */
+typedef unsigned short EMSH;    /* type of expanded-memory handles */
+
+typedef union {
+  short file_handle;            /* DOS file handle if it's a temp file */
+  XMSH xms_handle;              /* handle if it's a chunk of XMS */
+  EMSH ems_handle;              /* handle if it's a chunk of EMS */
+} handle_union;
+
+#endif /* USE_MSDOS_MEMMGR */
+
+#ifdef USE_MAC_MEMMGR           /* Mac-specific junk */
+#include <Files.h>
+#endif /* USE_MAC_MEMMGR */
+
+
+typedef struct backing_store_struct *backing_store_ptr;
+
+typedef struct backing_store_struct {
+  /* Methods for reading/writing/closing this backing-store object */
+  void (*read_backing_store) (j_common_ptr cinfo, backing_store_ptr info,
+                              void *buffer_address, long file_offset,
+                              long byte_count);
+  void (*write_backing_store) (j_common_ptr cinfo, backing_store_ptr info,
+                               void *buffer_address, long file_offset,
+                               long byte_count);
+  void (*close_backing_store) (j_common_ptr cinfo, backing_store_ptr info);
+
+  /* Private fields for system-dependent backing-store management */
+#ifdef USE_MSDOS_MEMMGR
+  /* For the MS-DOS manager (jmemdos.c), we need: */
+  handle_union handle;          /* reference to backing-store storage object */
+  char temp_nameTEMP_NAME_LENGTH; /* name if it's a file */
+#else
+#ifdef USE_MAC_MEMMGR
+  /* For the Mac manager (jmemmac.c), we need: */
+  short temp_file;              /* file reference number to temp file */
+  FSSpec tempSpec;              /* the FSSpec for the temp file */
+  char temp_nameTEMP_NAME_LENGTH; /* name if it's a file */
+#else
+  /* For a typical implementation with temp files, we need: */
+  FILE *temp_file;              /* stdio reference to temp file */
+  char temp_nameTEMP_NAME_LENGTH; /* name of temp file */
+#endif
+#endif
+} backing_store_info;
+
+
+/*
+ * Initial opening of a backing-store object.  This must fill in the
+ * read/write/close pointers in the object.  The read/write routines
+ * may take an error exit if the specified maximum file size is exceeded.
+ * (If jpeg_mem_available always returns a large value, this routine can
+ * just take an error exit.)
+ */
+
+EXTERN(void) jpeg_open_backing_store(j_common_ptr cinfo,
+                                     backing_store_ptr info,
+                                     long total_bytes_needed);
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.  jpeg_mem_init will be called before anything is
+ * allocated (and, therefore, nothing in cinfo is of use except the error
+ * manager pointer).  It should return a suitable default value for
+ * max_memory_to_use; this may subsequently be overridden by the surrounding
+ * application.  (Note that max_memory_to_use is only important if
+ * jpeg_mem_available chooses to consult it ... no one else will.)
+ * jpeg_mem_term may assume that all requested memory has been freed and that
+ * all opened backing-store objects have been closed.
+ */
+
+EXTERN(long) jpeg_mem_init(j_common_ptr cinfo);
+EXTERN(void) jpeg_mem_term(j_common_ptr cinfo);

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jmorecfg.h Added

@@ -0,0 +1,386 @@
+/*
+ * jmorecfg.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 1997-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2011, 2014-2015, 2018, 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains additional configuration options that customize the
+ * JPEG software for special applications or support machine-dependent
+ * optimizations.  Most users will not need to touch this file.
+ */
+
+
+/*
+ * Maximum number of components (color channels) allowed in JPEG image.
+ * To meet the letter of Rec. ITU-T T.81 | ISO/IEC 10918-1, set this to 255.
+ * However, darn few applications need more than 4 channels (maybe 5 for CMYK +
+ * alpha mask).  We recommend 10 as a reasonable compromise; use 4 if you are
+ * really short on memory.  (Each allowed component costs a hundred or so
+ * bytes of storage, whether actually used in an image or not.)
+ */
+
+#define MAX_COMPONENTS  10      /* maximum number of image components */
+
+
+/*
+ * Basic data types.
+ * You may need to change these if you have a machine with unusual data
+ * type sizes; for example, "char" not 8 bits, "short" not 16 bits,
+ * or "long" not 32 bits.  We don't care whether "int" is 16 or 32 bits,
+ * but it had better be at least 16.
+ */
+
+/* Representation of a single sample (pixel element value).
+ * We frequently allocate large arrays of these, so it's important to keep
+ * them small.  But if you have memory to burn and access to char or short
+ * arrays is very slow on your hardware, you might want to change these.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+/* JSAMPLE should be the smallest type that will hold the values 0..255.
+ */
+
+typedef unsigned char JSAMPLE;
+#define GETJSAMPLE(value)  ((int)(value))
+
+#define MAXJSAMPLE      255
+#define CENTERJSAMPLE   128
+
+#endif /* BITS_IN_JSAMPLE == 8 */
+
+
+#if BITS_IN_JSAMPLE == 12
+/* JSAMPLE should be the smallest type that will hold the values 0..4095.
+ * On nearly all machines "short" will do nicely.
+ */
+
+typedef short JSAMPLE;
+#define GETJSAMPLE(value)  ((int)(value))
+
+#define MAXJSAMPLE      4095
+#define CENTERJSAMPLE   2048
+
+#endif /* BITS_IN_JSAMPLE == 12 */
+
+
+/* Representation of a DCT frequency coefficient.
+ * This should be a signed value of at least 16 bits; "short" is usually OK.
+ * Again, we allocate large arrays of these, but you can change to int
+ * if you have memory to burn and "short" is really slow.
+ */
+
+typedef short JCOEF;
+
+
+/* Compressed datastreams are represented as arrays of JOCTET.
+ * These must be EXACTLY 8 bits wide, at least once they are written to
+ * external storage.  Note that when using the stdio data source/destination
+ * managers, this is also the data type passed to fread/fwrite.
+ */
+
+typedef unsigned char JOCTET;
+#define GETJOCTET(value)  (value)
+
+
+/* These typedefs are used for various table entries and so forth.
+ * They must be at least as wide as specified; but making them too big
+ * won't cost a huge amount of memory, so we don't provide special
+ * extraction code like we did for JSAMPLE.  (In other words, these
+ * typedefs live at a different point on the speed/space tradeoff curve.)
+ */
+
+/* UINT8 must hold at least the values 0..255. */
+
+typedef unsigned char UINT8;
+
+/* UINT16 must hold at least the values 0..65535. */
+
+#ifdef HAVE_UNSIGNED_SHORT
+typedef unsigned short UINT16;
+#else /* not HAVE_UNSIGNED_SHORT */
+typedef unsigned int UINT16;
+#endif /* HAVE_UNSIGNED_SHORT */
+
+/* INT16 must hold at least the values -32768..32767. */
+
+#ifndef XMD_H                   /* X11/xmd.h correctly defines INT16 */
+typedef short INT16;
+#endif
+
+/* INT32 must hold at least signed 32-bit values.
+ *
+ * NOTE: The INT32 typedef dates back to libjpeg v5 (1994.)  Integers were
+ * sometimes 16-bit back then (MS-DOS), which is why INT32 is typedef'd to
+ * long.  It also wasn't common (or at least as common) in 1994 for INT32 to be
+ * defined by platform headers.  Since then, however, INT32 is defined in
+ * several other common places:
+ *
+ * Xmd.h (X11 header) typedefs INT32 to int on 64-bit platforms and long on
+ * 32-bit platforms (i.e always a 32-bit signed type.)
+ *
+ * basetsd.h (Win32 header) typedefs INT32 to int (always a 32-bit signed type
+ * on modern platforms.)
+ *
+ * qglobal.h (Qt header) typedefs INT32 to int (always a 32-bit signed type on
+ * modern platforms.)
+ *
+ * This is a recipe for conflict, since "long" and "int" aren't always
+ * compatible types.  Since the definition of INT32 has technically been part
+ * of the libjpeg API for more than 20 years, we can't remove it, but we do not
+ * use it internally any longer.  We instead define a separate type (JLONG)
+ * for internal use, which ensures that internal behavior will always be the
+ * same regardless of any external headers that may be included.
+ */
+
+#ifndef XMD_H                   /* X11/xmd.h correctly defines INT32 */
+#ifndef _BASETSD_H_             /* Microsoft defines it in basetsd.h */
+#ifndef _BASETSD_H              /* MinGW is slightly different */
+#ifndef QGLOBAL_H               /* Qt defines it in qglobal.h */
+typedef long INT32;
+#endif
+#endif
+#endif
+#endif
+
+/* Datatype used for image dimensions.  The JPEG standard only supports
+ * images up to 64K*64K due to 16-bit fields in SOF markers.  Therefore
+ * "unsigned int" is sufficient on all machines.  However, if you need to
+ * handle larger images and you don't mind deviating from the spec, you
+ * can change this datatype.  (Note that changing this datatype will
+ * potentially require modifying the SIMD code.  The x86-64 SIMD extensions,
+ * in particular, assume a 32-bit JDIMENSION.)
+ */
+
+typedef unsigned int JDIMENSION;
+
+#define JPEG_MAX_DIMENSION  65500L  /* a tad under 64K to prevent overflows */
+
+
+/* These macros are used in all function definitions and extern declarations.
+ * You could modify them if you need to change function linkage conventions;
+ * in particular, you'll need to do that to make the library a Windows DLL.
+ * Another application is to make all functions global for use with debuggers
+ * or code profilers that require it.
+ */
+
+/* a function called through method pointers: */
+#define METHODDEF(type)         static type
+/* a function used only in its module: */
+#define LOCAL(type)             static type
+/* a function referenced thru EXTERNs: */
+#define GLOBAL(type)            type
+/* a reference to a GLOBAL function: */
+#define EXTERN(type)            extern type
+
+
+/* Originally, this macro was used as a way of defining function prototypes
+ * for both modern compilers as well as older compilers that did not support
+ * prototype parameters.  libjpeg-turbo has never supported these older,
+ * non-ANSI compilers, but the macro is still included because there is some
+ * software out there that uses it.
+ */
+
+#define JMETHOD(type, methodname, arglist)  type (*methodname) arglist
+
+
+/* libjpeg-turbo no longer supports platforms that have far symbols (MS-DOS),
+ * but again, some software relies on this macro.
+ */
+
+#undef FAR
+#define FAR
+
+
+/*
+ * On a few systems, type boolean and/or its values FALSE, TRUE may appear
+ * in standard header files.  Or you may have conflicts with application-
+ * specific header files that you want to include together with these files.
+ * Defining HAVE_BOOLEAN before including jpeglib.h should make it work.
+ */
+
+#ifndef HAVE_BOOLEAN
+typedef int boolean;
+#endif
+#ifndef FALSE                   /* in case these macros already exist */
+#define FALSE   0               /* values of boolean */
+#endif
+#ifndef TRUE
+#define TRUE    1
+#endif
+
+
+/*
+ * The remaining options affect code selection within the JPEG library,
+ * but they don't need to be visible to most applications using the library.
+ * To minimize application namespace pollution, the symbols won't be
+ * defined unless JPEG_INTERNALS or JPEG_INTERNAL_OPTIONS has been defined.
+ */
+
+#ifdef JPEG_INTERNALS
+#define JPEG_INTERNAL_OPTIONS
+#endif
+
+#ifdef JPEG_INTERNAL_OPTIONS
+
+
+/*
+ * These defines indicate whether to include various optional functions.
+ * Undefining some of these symbols will produce a smaller but less capable
+ * library.  Note that you can leave certain source files out of the
+ * compilation/linking process if you've #undef'd the corresponding symbols.
+ * (You may HAVE to do that if your compiler doesn't like null source files.)
+ */
+
+/* Capability options common to encoder and decoder: */
+
+#define DCT_ISLOW_SUPPORTED     /* accurate integer method */
+#define DCT_IFAST_SUPPORTED     /* less accurate int method legacy feature */
+#define DCT_FLOAT_SUPPORTED     /* floating-point method legacy feature */
+
+/* Encoder capability options: */
+
+#define C_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
+#define C_PROGRESSIVE_SUPPORTED     /* Progressive JPEG? (Requires MULTISCAN)*/
+#define ENTROPY_OPT_SUPPORTED       /* Optimization of entropy coding parms? */
+/* Note: if you selected 12-bit data precision, it is dangerous to turn off
+ * ENTROPY_OPT_SUPPORTED.  The standard Huffman tables are only good for 8-bit
+ * precision, so jchuff.c normally uses entropy optimization to compute
+ * usable tables for higher precision.  If you don't want to do optimization,
+ * you'll have to supply different default Huffman tables.
+ * The exact same statements apply for progressive JPEG: the default tables
+ * don't work for progressive mode.  (This may get fixed, however.)
+ */
+#define INPUT_SMOOTHING_SUPPORTED   /* Input image smoothing option? */
+
+/* Decoder capability options: */
+
+#define D_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
+#define D_PROGRESSIVE_SUPPORTED     /* Progressive JPEG? (Requires MULTISCAN)*/
+#define SAVE_MARKERS_SUPPORTED      /* jpeg_save_markers() needed? */
+#define BLOCK_SMOOTHING_SUPPORTED   /* Block smoothing? (Progressive only) */
+#define IDCT_SCALING_SUPPORTED      /* Output rescaling via IDCT? */
+#undef  UPSAMPLE_SCALING_SUPPORTED  /* Output rescaling at upsample stage? */
+#define UPSAMPLE_MERGING_SUPPORTED  /* Fast path for sloppy upsampling? */
+#define QUANT_1PASS_SUPPORTED       /* 1-pass color quantization? */
+#define QUANT_2PASS_SUPPORTED       /* 2-pass color quantization? */
+
+/* more capability options later, no doubt */
+
+
+/*
+ * The RGB_RED, RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE macros are a vestigial
+ * feature of libjpeg.  The idea was that, if an application developer needed
+ * to compress from/decompress to a BGR/BGRX/RGBX/XBGR/XRGB buffer, they could
+ * change these macros, rebuild libjpeg, and link their application statically
+ * with it.  In reality, few people ever did this, because there were some
+ * severe restrictions involved (cjpeg and djpeg no longer worked properly,
+ * compressing/decompressing RGB JPEGs no longer worked properly, and the color
+ * quantizer wouldn't work with pixel sizes other than 3.)  Furthermore, since
+ * all of the O/S-supplied versions of libjpeg were built with the default
+ * values of RGB_RED, RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE, many applications
+ * have come to regard these values as immutable.
+ *
+ * The libjpeg-turbo colorspace extensions provide a much cleaner way of
+ * compressing from/decompressing to buffers with arbitrary component orders
+ * and pixel sizes.  Thus, we do not support changing the values of RGB_RED,
+ * RGB_GREEN, RGB_BLUE, or RGB_PIXELSIZE.  In addition to the restrictions
+ * listed above, changing these values will also break the SIMD extensions and
+ * the regression tests.
+ */
+
+#define RGB_RED         0       /* Offset of Red in an RGB scanline element */
+#define RGB_GREEN       1       /* Offset of Green */
+#define RGB_BLUE        2       /* Offset of Blue */
+#define RGB_PIXELSIZE   3       /* JSAMPLEs per RGB scanline element */
+
+#define JPEG_NUMCS  17
+
+#define EXT_RGB_RED         0
+#define EXT_RGB_GREEN       1
+#define EXT_RGB_BLUE        2
+#define EXT_RGB_PIXELSIZE   3
+
+#define EXT_RGBX_RED        0
+#define EXT_RGBX_GREEN      1
+#define EXT_RGBX_BLUE       2
+#define EXT_RGBX_PIXELSIZE  4
+
+#define EXT_BGR_RED         2
+#define EXT_BGR_GREEN       1
+#define EXT_BGR_BLUE        0
+#define EXT_BGR_PIXELSIZE   3
+
+#define EXT_BGRX_RED        2
+#define EXT_BGRX_GREEN      1
+#define EXT_BGRX_BLUE       0
+#define EXT_BGRX_PIXELSIZE  4
+
+#define EXT_XBGR_RED        3
+#define EXT_XBGR_GREEN      2
+#define EXT_XBGR_BLUE       1
+#define EXT_XBGR_PIXELSIZE  4
+
+#define EXT_XRGB_RED        1
+#define EXT_XRGB_GREEN      2
+#define EXT_XRGB_BLUE       3
+#define EXT_XRGB_PIXELSIZE  4
+
+static const int rgb_redJPEG_NUMCS = {
+  -1, -1, RGB_RED, -1, -1, -1, EXT_RGB_RED, EXT_RGBX_RED,
+  EXT_BGR_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED,
+  EXT_RGBX_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED,
+  -1
+};
+
+static const int rgb_greenJPEG_NUMCS = {
+  -1, -1, RGB_GREEN, -1, -1, -1, EXT_RGB_GREEN, EXT_RGBX_GREEN,
+  EXT_BGR_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN,
+  EXT_RGBX_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN,
+  -1
+};
+
+static const int rgb_blueJPEG_NUMCS = {
+  -1, -1, RGB_BLUE, -1, -1, -1, EXT_RGB_BLUE, EXT_RGBX_BLUE,
+  EXT_BGR_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE,
+  EXT_RGBX_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE,
+  -1
+};
+
+static const int rgb_pixelsizeJPEG_NUMCS = {
+  -1, -1, RGB_PIXELSIZE, -1, -1, -1, EXT_RGB_PIXELSIZE, EXT_RGBX_PIXELSIZE,
+  EXT_BGR_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE,
+  EXT_RGBX_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE,
+  -1
+};
+
+/* Definitions for speed-related optimizations. */
+
+/* On some machines (notably 68000 series) "int" is 32 bits, but multiplying
+ * two 16-bit shorts is faster than multiplying two ints.  Define MULTIPLIER
+ * as short on such a machine.  MULTIPLIER must be at least 16 bits wide.
+ */
+
+#ifndef MULTIPLIER
+#ifndef WITH_SIMD
+#define MULTIPLIER  int         /* type for fastest integer multiply */
+#else
+#define MULTIPLIER  short       /* prefer 16-bit with SIMD for parellelism */
+#endif
+#endif
+
+
+/* FAST_FLOAT should be either float or double, whichever is done faster
+ * by your compiler.  (Note that this type is only used in the floating point
+ * DCT routines, so it only matters if you've defined DCT_FLOAT_SUPPORTED.)
+ */
+
+#ifndef FAST_FLOAT
+#define FAST_FLOAT  float
+#endif
+
+#endif /* JPEG_INTERNAL_OPTIONS */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jpeg_nbits_table.h Added

@@ -0,0 +1,4098 @@
+static const unsigned char jpeg_nbits_table65536 = {
+   0,  1,  2,  2,  3,  3,  3,  3,  4,  4,  4,  4,  4,  4,  4,  4,
+   5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+};

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jpegcomp.h Added

@@ -0,0 +1,32 @@
+/*
+ * jpegcomp.h
+ *
+ * Copyright (C) 2010, 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * JPEG compatibility macros
+ * These declarations are considered internal to the JPEG library; most
+ * applications using the library shouldn't need to include this file.
+ */
+
+#if JPEG_LIB_VERSION >= 70
+#define _DCT_scaled_size  DCT_h_scaled_size
+#define _DCT_h_scaled_size  DCT_h_scaled_size
+#define _DCT_v_scaled_size  DCT_v_scaled_size
+#define _min_DCT_scaled_size  min_DCT_h_scaled_size
+#define _min_DCT_h_scaled_size  min_DCT_h_scaled_size
+#define _min_DCT_v_scaled_size  min_DCT_v_scaled_size
+#define _jpeg_width  jpeg_width
+#define _jpeg_height  jpeg_height
+#define JERR_ARITH_NOTIMPL  JERR_NOT_COMPILED
+#else
+#define _DCT_scaled_size  DCT_scaled_size
+#define _DCT_h_scaled_size  DCT_scaled_size
+#define _DCT_v_scaled_size  DCT_scaled_size
+#define _min_DCT_scaled_size  min_DCT_scaled_size
+#define _min_DCT_h_scaled_size  min_DCT_scaled_size
+#define _min_DCT_v_scaled_size  min_DCT_scaled_size
+#define _jpeg_width  image_width
+#define _jpeg_height  image_height
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jpegint.h Added

@@ -0,0 +1,384 @@
+/*
+ * jpegint.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 1997-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015-2016, 2019, 2021, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * Copyright (C) 2021, Alex Richardson.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file provides common declarations for the various JPEG modules.
+ * These declarations are considered internal to the JPEG library; most
+ * applications using the library shouldn't need to include this file.
+ */
+
+
+/* Declarations for both compression & decompression */
+
+typedef enum {            /* Operating modes for buffer controllers */
+  JBUF_PASS_THRU,         /* Plain stripwise operation */
+  /* Remaining modes require a full-image buffer to have been created */
+  JBUF_SAVE_SOURCE,       /* Run source subobject only, save output */
+  JBUF_CRANK_DEST,        /* Run dest subobject only, using saved data */
+  JBUF_SAVE_AND_PASS      /* Run both subobjects, save output */
+} J_BUF_MODE;
+
+/* Values of global_state field (jdapi.c has some dependencies on ordering!) */
+#define CSTATE_START     100    /* after create_compress */
+#define CSTATE_SCANNING  101    /* start_compress done, write_scanlines OK */
+#define CSTATE_RAW_OK    102    /* start_compress done, write_raw_data OK */
+#define CSTATE_WRCOEFS   103    /* jpeg_write_coefficients done */
+#define DSTATE_START     200    /* after create_decompress */
+#define DSTATE_INHEADER  201    /* reading header markers, no SOS yet */
+#define DSTATE_READY     202    /* found SOS, ready for start_decompress */
+#define DSTATE_PRELOAD   203    /* reading multiscan file in start_decompress*/
+#define DSTATE_PRESCAN   204    /* performing dummy pass for 2-pass quant */
+#define DSTATE_SCANNING  205    /* start_decompress done, read_scanlines OK */
+#define DSTATE_RAW_OK    206    /* start_decompress done, read_raw_data OK */
+#define DSTATE_BUFIMAGE  207    /* expecting jpeg_start_output */
+#define DSTATE_BUFPOST   208    /* looking for SOS/EOI in jpeg_finish_output */
+#define DSTATE_RDCOEFS   209    /* reading file in jpeg_read_coefficients */
+#define DSTATE_STOPPING  210    /* looking for EOI in jpeg_finish_decompress */
+
+
+/* JLONG must hold at least signed 32-bit values. */
+typedef long JLONG;
+
+/* JUINTPTR must hold pointer values. */
+#ifdef __UINTPTR_TYPE__
+/*
+ * __UINTPTR_TYPE__ is GNU-specific and available in GCC 4.6+ and Clang 3.0+.
+ * Fortunately, that is sufficient to support the few architectures for which
+ * sizeof(void *) != sizeof(size_t).  The only other options would require C99
+ * or Clang-specific builtins.
+ */
+typedef __UINTPTR_TYPE__ JUINTPTR;
+#else
+typedef size_t JUINTPTR;
+#endif
+
+/*
+ * Left shift macro that handles a negative operand without causing any
+ * sanitizer warnings
+ */
+
+#define LEFT_SHIFT(a, b)  ((JLONG)((unsigned long)(a) << (b)))
+
+
+/* Declarations for compression modules */
+
+/* Master control module */
+struct jpeg_comp_master {
+  void (*prepare_for_pass) (j_compress_ptr cinfo);
+  void (*pass_startup) (j_compress_ptr cinfo);
+  void (*finish_pass) (j_compress_ptr cinfo);
+
+  /* State variables made visible to other modules */
+  boolean call_pass_startup;    /* True if pass_startup must be called */
+  boolean is_last_pass;         /* True during last pass */
+};
+
+/* Main buffer control (downsampled-data buffer) */
+struct jpeg_c_main_controller {
+  void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*process_data) (j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                        JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail);
+};
+
+/* Compression preprocessing (downsampling input buffer control) */
+struct jpeg_c_prep_controller {
+  void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*pre_process_data) (j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                            JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail,
+                            JSAMPIMAGE output_buf,
+                            JDIMENSION *out_row_group_ctr,
+                            JDIMENSION out_row_groups_avail);
+};
+
+/* Coefficient buffer control */
+struct jpeg_c_coef_controller {
+  void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode);
+  boolean (*compress_data) (j_compress_ptr cinfo, JSAMPIMAGE input_buf);
+};
+
+/* Colorspace conversion */
+struct jpeg_color_converter {
+  void (*start_pass) (j_compress_ptr cinfo);
+  void (*color_convert) (j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                         JSAMPIMAGE output_buf, JDIMENSION output_row,
+                         int num_rows);
+};
+
+/* Downsampling */
+struct jpeg_downsampler {
+  void (*start_pass) (j_compress_ptr cinfo);
+  void (*downsample) (j_compress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION in_row_index, JSAMPIMAGE output_buf,
+                      JDIMENSION out_row_group_index);
+
+  boolean need_context_rows;    /* TRUE if need rows above & below */
+};
+
+/* Forward DCT (also controls coefficient quantization) */
+struct jpeg_forward_dct {
+  void (*start_pass) (j_compress_ptr cinfo);
+  /* perhaps this should be an array??? */
+  void (*forward_DCT) (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+                       JDIMENSION start_row, JDIMENSION start_col,
+                       JDIMENSION num_blocks);
+};
+
+/* Entropy encoding */
+struct jpeg_entropy_encoder {
+  void (*start_pass) (j_compress_ptr cinfo, boolean gather_statistics);
+  boolean (*encode_mcu) (j_compress_ptr cinfo, JBLOCKROW *MCU_data);
+  void (*finish_pass) (j_compress_ptr cinfo);
+};
+
+/* Marker writing */
+struct jpeg_marker_writer {
+  void (*write_file_header) (j_compress_ptr cinfo);
+  void (*write_frame_header) (j_compress_ptr cinfo);
+  void (*write_scan_header) (j_compress_ptr cinfo);
+  void (*write_file_trailer) (j_compress_ptr cinfo);
+  void (*write_tables_only) (j_compress_ptr cinfo);
+  /* These routines are exported to allow insertion of extra markers */
+  /* Probably only COM and APPn markers should be written this way */
+  void (*write_marker_header) (j_compress_ptr cinfo, int marker,
+                               unsigned int datalen);
+  void (*write_marker_byte) (j_compress_ptr cinfo, int val);
+};
+
+
+/* Declarations for decompression modules */
+
+/* Master control module */
+struct jpeg_decomp_master {
+  void (*prepare_for_output_pass) (j_decompress_ptr cinfo);
+  void (*finish_output_pass) (j_decompress_ptr cinfo);
+
+  /* State variables made visible to other modules */
+  boolean is_dummy_pass;        /* True during 1st pass for 2-pass quant */
+
+  /* Partial decompression variables */
+  JDIMENSION first_iMCU_col;
+  JDIMENSION last_iMCU_col;
+  JDIMENSION first_MCU_colMAX_COMPONENTS;
+  JDIMENSION last_MCU_colMAX_COMPONENTS;
+  boolean jinit_upsampler_no_alloc;
+
+  /* Last iMCU row that was successfully decoded */
+  JDIMENSION last_good_iMCU_row;
+};
+
+/* Input control module */
+struct jpeg_input_controller {
+  int (*consume_input) (j_decompress_ptr cinfo);
+  void (*reset_input_controller) (j_decompress_ptr cinfo);
+  void (*start_input_pass) (j_decompress_ptr cinfo);
+  void (*finish_input_pass) (j_decompress_ptr cinfo);
+
+  /* State variables made visible to other modules */
+  boolean has_multiple_scans;   /* True if file has multiple scans */
+  boolean eoi_reached;          /* True when EOI has been consumed */
+};
+
+/* Main buffer control (downsampled-data buffer) */
+struct jpeg_d_main_controller {
+  void (*start_pass) (j_decompress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*process_data) (j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+                        JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
+};
+
+/* Coefficient buffer control */
+struct jpeg_d_coef_controller {
+  void (*start_input_pass) (j_decompress_ptr cinfo);
+  int (*consume_data) (j_decompress_ptr cinfo);
+  void (*start_output_pass) (j_decompress_ptr cinfo);
+  int (*decompress_data) (j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
+  /* Pointer to array of coefficient virtual arrays, or NULL if none */
+  jvirt_barray_ptr *coef_arrays;
+};
+
+/* Decompression postprocessing (color quantization buffer control) */
+struct jpeg_d_post_controller {
+  void (*start_pass) (j_decompress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*post_process_data) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                             JDIMENSION *in_row_group_ctr,
+                             JDIMENSION in_row_groups_avail,
+                             JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                             JDIMENSION out_rows_avail);
+};
+
+/* Marker reading & parsing */
+struct jpeg_marker_reader {
+  void (*reset_marker_reader) (j_decompress_ptr cinfo);
+  /* Read markers until SOS or EOI.
+   * Returns same codes as are defined for jpeg_consume_input:
+   * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
+   */
+  int (*read_markers) (j_decompress_ptr cinfo);
+  /* Read a restart marker --- exported for use by entropy decoder only */
+  jpeg_marker_parser_method read_restart_marker;
+
+  /* State of marker reader --- nominally internal, but applications
+   * supplying COM or APPn handlers might like to know the state.
+   */
+  boolean saw_SOI;              /* found SOI? */
+  boolean saw_SOF;              /* found SOF? */
+  int next_restart_num;         /* next restart number expected (0-7) */
+  unsigned int discarded_bytes; /* # of bytes skipped looking for a marker */
+};
+
+/* Entropy decoding */
+struct jpeg_entropy_decoder {
+  void (*start_pass) (j_decompress_ptr cinfo);
+  boolean (*decode_mcu) (j_decompress_ptr cinfo, JBLOCKROW *MCU_data);
+
+  /* This is here to share code between baseline and progressive decoders; */
+  /* other modules probably should not use it */
+  boolean insufficient_data;    /* set TRUE after emitting warning */
+};
+
+/* Inverse DCT (also performs dequantization) */
+typedef void (*inverse_DCT_method_ptr) (j_decompress_ptr cinfo,
+                                        jpeg_component_info *compptr,
+                                        JCOEFPTR coef_block,
+                                        JSAMPARRAY output_buf,
+                                        JDIMENSION output_col);
+
+struct jpeg_inverse_dct {
+  void (*start_pass) (j_decompress_ptr cinfo);
+  /* It is useful to allow each component to have a separate IDCT method. */
+  inverse_DCT_method_ptr inverse_DCTMAX_COMPONENTS;
+};
+
+/* Upsampling (note that upsampler must also call color converter) */
+struct jpeg_upsampler {
+  void (*start_pass) (j_decompress_ptr cinfo);
+  void (*upsample) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail, JSAMPARRAY output_buf,
+                    JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
+
+  boolean need_context_rows;    /* TRUE if need rows above & below */
+};
+
+/* Colorspace conversion */
+struct jpeg_color_deconverter {
+  void (*start_pass) (j_decompress_ptr cinfo);
+  void (*color_convert) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows);
+};
+
+/* Color quantization or color precision reduction */
+struct jpeg_color_quantizer {
+  void (*start_pass) (j_decompress_ptr cinfo, boolean is_pre_scan);
+  void (*color_quantize) (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                          JSAMPARRAY output_buf, int num_rows);
+  void (*finish_pass) (j_decompress_ptr cinfo);
+  void (*new_color_map) (j_decompress_ptr cinfo);
+};
+
+
+/* Miscellaneous useful macros */
+
+#undef MAX
+#define MAX(a, b)       ((a) > (b) ? (a) : (b))
+#undef MIN
+#define MIN(a, b)       ((a) < (b) ? (a) : (b))
+
+
+/* We assume that right shift corresponds to signed division by 2 with
+ * rounding towards minus infinity.  This is correct for typical "arithmetic
+ * shift" instructions that shift in copies of the sign bit.  But some
+ * C compilers implement >> with an unsigned shift.  For these machines you
+ * must define RIGHT_SHIFT_IS_UNSIGNED.
+ * RIGHT_SHIFT provides a proper signed right shift of a JLONG quantity.
+ * It is only applied with constant shift counts.  SHIFT_TEMPS must be
+ * included in the variables of any routine using RIGHT_SHIFT.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define SHIFT_TEMPS     JLONG shift_temp;
+#define RIGHT_SHIFT(x, shft) \
+  ((shift_temp = (x)) < 0 ? \
+   (shift_temp >> (shft)) | ((~((JLONG)0)) << (32 - (shft))) : \
+   (shift_temp >> (shft)))
+#else
+#define SHIFT_TEMPS
+#define RIGHT_SHIFT(x, shft)    ((x) >> (shft))
+#endif
+
+
+/* Compression module initialization routines */
+EXTERN(void) jinit_compress_master(j_compress_ptr cinfo);
+EXTERN(void) jinit_c_master_control(j_compress_ptr cinfo,
+                                    boolean transcode_only);
+EXTERN(void) jinit_c_main_controller(j_compress_ptr cinfo,
+                                     boolean need_full_buffer);
+EXTERN(void) jinit_c_prep_controller(j_compress_ptr cinfo,
+                                     boolean need_full_buffer);
+EXTERN(void) jinit_c_coef_controller(j_compress_ptr cinfo,
+                                     boolean need_full_buffer);
+EXTERN(void) jinit_color_converter(j_compress_ptr cinfo);
+EXTERN(void) jinit_downsampler(j_compress_ptr cinfo);
+EXTERN(void) jinit_forward_dct(j_compress_ptr cinfo);
+EXTERN(void) jinit_huff_encoder(j_compress_ptr cinfo);
+EXTERN(void) jinit_phuff_encoder(j_compress_ptr cinfo);
+EXTERN(void) jinit_arith_encoder(j_compress_ptr cinfo);
+EXTERN(void) jinit_marker_writer(j_compress_ptr cinfo);
+/* Decompression module initialization routines */
+EXTERN(void) jinit_master_decompress(j_decompress_ptr cinfo);
+EXTERN(void) jinit_d_main_controller(j_decompress_ptr cinfo,
+                                     boolean need_full_buffer);
+EXTERN(void) jinit_d_coef_controller(j_decompress_ptr cinfo,
+                                     boolean need_full_buffer);
+EXTERN(void) jinit_d_post_controller(j_decompress_ptr cinfo,
+                                     boolean need_full_buffer);
+EXTERN(void) jinit_input_controller(j_decompress_ptr cinfo);
+EXTERN(void) jinit_marker_reader(j_decompress_ptr cinfo);
+EXTERN(void) jinit_huff_decoder(j_decompress_ptr cinfo);
+EXTERN(void) jinit_phuff_decoder(j_decompress_ptr cinfo);
+EXTERN(void) jinit_arith_decoder(j_decompress_ptr cinfo);
+EXTERN(void) jinit_inverse_dct(j_decompress_ptr cinfo);
+EXTERN(void) jinit_upsampler(j_decompress_ptr cinfo);
+EXTERN(void) jinit_color_deconverter(j_decompress_ptr cinfo);
+EXTERN(void) jinit_1pass_quantizer(j_decompress_ptr cinfo);
+EXTERN(void) jinit_2pass_quantizer(j_decompress_ptr cinfo);
+EXTERN(void) jinit_merged_upsampler(j_decompress_ptr cinfo);
+/* Memory manager initialization */
+EXTERN(void) jinit_memory_mgr(j_common_ptr cinfo);
+
+/* Utility routines in jutils.c */
+EXTERN(long) jdiv_round_up(long a, long b);
+EXTERN(long) jround_up(long a, long b);
+EXTERN(void) jcopy_sample_rows(JSAMPARRAY input_array, int source_row,
+                               JSAMPARRAY output_array, int dest_row,
+                               int num_rows, JDIMENSION num_cols);
+EXTERN(void) jcopy_block_row(JBLOCKROW input_row, JBLOCKROW output_row,
+                             JDIMENSION num_blocks);
+EXTERN(void) jzero_far(void *target, size_t bytestozero);
+/* Constant tables in jutils.c */
+#if 0                           /* This table is not actually needed in v6a */
+extern const int jpeg_zigzag_order; /* natural coef order to zigzag order */
+#endif
+extern const int jpeg_natural_order; /* zigzag coef order to natural order */
+
+/* Arithmetic coding probability estimation tables in jaricom.c */
+extern const JLONG jpeg_aritab;
+
+/* Suppress undefined-structure complaints if necessary. */
+
+#ifdef INCOMPLETE_TYPES_BROKEN
+#ifndef AM_MEMORY_MANAGER       /* only jmemmgr.c defines these */
+struct jvirt_sarray_control { long dummy; };
+struct jvirt_barray_control { long dummy; };
+#endif
+#endif /* INCOMPLETE_TYPES_BROKEN */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jpeglib.h Added

@@ -0,0 +1,1132 @@
+/*
+ * jpeglib.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * Modified 2002-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2013-2014, 2016-2017, 2020, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file defines the application interface for the JPEG library.
+ * Most applications using the library need only include this file,
+ * and perhaps jerror.h if they want to know the exact error codes.
+ */
+
+#ifndef JPEGLIB_H
+#define JPEGLIB_H
+
+/*
+ * First we include the configuration files that record how this
+ * installation of the JPEG library is set up.  jconfig.h can be
+ * generated automatically for many systems.  jmorecfg.h contains
+ * manual configuration options that most people need not worry about.
+ */
+
+#ifndef JCONFIG_INCLUDED        /* in case jinclude.h already did */
+#include "jconfig.h"            /* widely used configuration options */
+#endif
+#include "jmorecfg.h"           /* seldom changed options */
+
+
+#ifdef __cplusplus
+#ifndef DONT_USE_EXTERN_C
+extern "C" {
+#endif
+#endif
+
+
+/* Various constants determining the sizes of things.
+ * All of these are specified by the JPEG standard, so don't change them
+ * if you want to be compatible.
+ */
+
+#define DCTSIZE             8   /* The basic DCT block is 8x8 samples */
+#define DCTSIZE2            64  /* DCTSIZE squared; # of elements in a block */
+#define NUM_QUANT_TBLS      4   /* Quantization tables are numbered 0..3 */
+#define NUM_HUFF_TBLS       4   /* Huffman tables are numbered 0..3 */
+#define NUM_ARITH_TBLS      16  /* Arith-coding tables are numbered 0..15 */
+#define MAX_COMPS_IN_SCAN   4   /* JPEG limit on # of components in one scan */
+#define MAX_SAMP_FACTOR     4   /* JPEG limit on sampling factors */
+/* Unfortunately, some bozo at Adobe saw no reason to be bound by the standard;
+ * the PostScript DCT filter can emit files with many more than 10 blocks/MCU.
+ * If you happen to run across such a file, you can up D_MAX_BLOCKS_IN_MCU
+ * to handle it.  We even let you do this from the jconfig.h file.  However,
+ * we strongly discourage changing C_MAX_BLOCKS_IN_MCU; just because Adobe
+ * sometimes emits noncompliant files doesn't mean you should too.
+ */
+#define C_MAX_BLOCKS_IN_MCU   10 /* compressor's limit on blocks per MCU */
+#ifndef D_MAX_BLOCKS_IN_MCU
+#define D_MAX_BLOCKS_IN_MCU   10 /* decompressor's limit on blocks per MCU */
+#endif
+
+
+/* Data structures for images (arrays of samples and of DCT coefficients).
+ */
+
+typedef JSAMPLE *JSAMPROW;      /* ptr to one image row of pixel samples. */
+typedef JSAMPROW *JSAMPARRAY;   /* ptr to some rows (a 2-D sample array) */
+typedef JSAMPARRAY *JSAMPIMAGE; /* a 3-D sample array: top index is color */
+
+typedef JCOEF JBLOCKDCTSIZE2; /* one block of coefficients */
+typedef JBLOCK *JBLOCKROW;      /* pointer to one row of coefficient blocks */
+typedef JBLOCKROW *JBLOCKARRAY;         /* a 2-D array of coefficient blocks */
+typedef JBLOCKARRAY *JBLOCKIMAGE;       /* a 3-D array of coefficient blocks */
+
+typedef JCOEF *JCOEFPTR;        /* useful in a couple of places */
+
+
+/* Types for JPEG compression parameters and working tables. */
+
+
+/* DCT coefficient quantization tables. */
+
+typedef struct {
+  /* This array gives the coefficient quantizers in natural array order
+   * (not the zigzag order in which they are stored in a JPEG DQT marker).
+   * CAUTION: IJG versions prior to v6a kept this array in zigzag order.
+   */
+  UINT16 quantvalDCTSIZE2;    /* quantization step for each coefficient */
+  /* This field is used only during compression.  It's initialized FALSE when
+   * the table is created, and set TRUE when it's been output to the file.
+   * You could suppress output of a table by setting this to TRUE.
+   * (See jpeg_suppress_tables for an example.)
+   */
+  boolean sent_table;           /* TRUE when table has been output */
+} JQUANT_TBL;
+
+
+/* Huffman coding tables. */
+
+typedef struct {
+  /* These two fields directly represent the contents of a JPEG DHT marker */
+  UINT8 bits17;               /* bitsk = # of symbols with codes of */
+                                /* length k bits; bits0 is unused */
+  UINT8 huffval256;           /* The symbols, in order of incr code length */
+  /* This field is used only during compression.  It's initialized FALSE when
+   * the table is created, and set TRUE when it's been output to the file.
+   * You could suppress output of a table by setting this to TRUE.
+   * (See jpeg_suppress_tables for an example.)
+   */
+  boolean sent_table;           /* TRUE when table has been output */
+} JHUFF_TBL;
+
+
+/* Basic info about one component (color channel). */
+
+typedef struct {
+  /* These values are fixed over the whole image. */
+  /* For compression, they must be supplied by parameter setup; */
+  /* for decompression, they are read from the SOF marker. */
+  int component_id;             /* identifier for this component (0..255) */
+  int component_index;          /* its index in SOF or cinfo->comp_info */
+  int h_samp_factor;            /* horizontal sampling factor (1..4) */
+  int v_samp_factor;            /* vertical sampling factor (1..4) */
+  int quant_tbl_no;             /* quantization table selector (0..3) */
+  /* These values may vary between scans. */
+  /* For compression, they must be supplied by parameter setup; */
+  /* for decompression, they are read from the SOS marker. */
+  /* The decompressor output side may not use these variables. */
+  int dc_tbl_no;                /* DC entropy table selector (0..3) */
+  int ac_tbl_no;                /* AC entropy table selector (0..3) */
+
+  /* Remaining fields should be treated as private by applications. */
+
+  /* These values are computed during compression or decompression startup: */
+  /* Component's size in DCT blocks.
+   * Any dummy blocks added to complete an MCU are not counted; therefore
+   * these values do not depend on whether a scan is interleaved or not.
+   */
+  JDIMENSION width_in_blocks;
+  JDIMENSION height_in_blocks;
+  /* Size of a DCT block in samples.  Always DCTSIZE for compression.
+   * For decompression this is the size of the output from one DCT block,
+   * reflecting any scaling we choose to apply during the IDCT step.
+   * Values from 1 to 16 are supported.
+   * Note that different components may receive different IDCT scalings.
+   */
+#if JPEG_LIB_VERSION >= 70
+  int DCT_h_scaled_size;
+  int DCT_v_scaled_size;
+#else
+  int DCT_scaled_size;
+#endif
+  /* The downsampled dimensions are the component's actual, unpadded number
+   * of samples at the main buffer (preprocessing/compression interface), thus
+   * downsampled_width = ceil(image_width * Hi/Hmax)
+   * and similarly for height.  For decompression, IDCT scaling is included, so
+   * downsampled_width = ceil(image_width * Hi/Hmax * DCT_h_scaled_size/DCTSIZE)
+   */
+  JDIMENSION downsampled_width;  /* actual width in samples */
+  JDIMENSION downsampled_height; /* actual height in samples */
+  /* This flag is used only for decompression.  In cases where some of the
+   * components will be ignored (eg grayscale output from YCbCr image),
+   * we can skip most computations for the unused components.
+   */
+  boolean component_needed;     /* do we need the value of this component? */
+
+  /* These values are computed before starting a scan of the component. */
+  /* The decompressor output side may not use these variables. */
+  int MCU_width;                /* number of blocks per MCU, horizontally */
+  int MCU_height;               /* number of blocks per MCU, vertically */
+  int MCU_blocks;               /* MCU_width * MCU_height */
+  int MCU_sample_width;         /* MCU width in samples, MCU_width*DCT_h_scaled_size */
+  int last_col_width;           /* # of non-dummy blocks across in last MCU */
+  int last_row_height;          /* # of non-dummy blocks down in last MCU */
+
+  /* Saved quantization table for component; NULL if none yet saved.
+   * See jdinput.c comments about the need for this information.
+   * This field is currently used only for decompression.
+   */
+  JQUANT_TBL *quant_table;
+
+  /* Private per-component storage for DCT or IDCT subsystem. */
+  void *dct_table;
+} jpeg_component_info;
+
+
+/* The script for encoding a multiple-scan file is an array of these: */
+
+typedef struct {
+  int comps_in_scan;            /* number of components encoded in this scan */
+  int component_indexMAX_COMPS_IN_SCAN; /* their SOF/comp_info indexes */
+  int Ss, Se;                   /* progressive JPEG spectral selection parms */
+  int Ah, Al;                   /* progressive JPEG successive approx. parms */
+} jpeg_scan_info;
+
+/* The decompressor can save APPn and COM markers in a list of these: */
+
+typedef struct jpeg_marker_struct *jpeg_saved_marker_ptr;
+
+struct jpeg_marker_struct {
+  jpeg_saved_marker_ptr next;   /* next in list, or NULL */
+  UINT8 marker;                 /* marker code: JPEG_COM, or JPEG_APP0+n */
+  unsigned int original_length; /* # bytes of data in the file */
+  unsigned int data_length;     /* # bytes of data saved at data */
+  JOCTET *data;                 /* the data contained in the marker */
+  /* the marker length word is not counted in data_length or original_length */
+};
+
+/* Known color spaces. */
+
+#define JCS_EXTENSIONS  1
+#define JCS_ALPHA_EXTENSIONS  1
+
+typedef enum {
+  JCS_UNKNOWN,            /* error/unspecified */
+  JCS_GRAYSCALE,          /* monochrome */
+  JCS_RGB,                /* red/green/blue as specified by the RGB_RED,
+                             RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE macros */
+  JCS_YCbCr,              /* Y/Cb/Cr (also known as YUV) */
+  JCS_CMYK,               /* C/M/Y/K */
+  JCS_YCCK,               /* Y/Cb/Cr/K */
+  JCS_EXT_RGB,            /* red/green/blue */
+  JCS_EXT_RGBX,           /* red/green/blue/x */
+  JCS_EXT_BGR,            /* blue/green/red */
+  JCS_EXT_BGRX,           /* blue/green/red/x */
+  JCS_EXT_XBGR,           /* x/blue/green/red */
+  JCS_EXT_XRGB,           /* x/red/green/blue */
+  /* When out_color_space it set to JCS_EXT_RGBX, JCS_EXT_BGRX, JCS_EXT_XBGR,
+     or JCS_EXT_XRGB during decompression, the X byte is undefined, and in
+     order to ensure the best performance, libjpeg-turbo can set that byte to
+     whatever value it wishes.  Use the following colorspace constants to
+     ensure that the X byte is set to 0xFF, so that it can be interpreted as an
+     opaque alpha channel. */
+  JCS_EXT_RGBA,           /* red/green/blue/alpha */
+  JCS_EXT_BGRA,           /* blue/green/red/alpha */
+  JCS_EXT_ABGR,           /* alpha/blue/green/red */
+  JCS_EXT_ARGB,           /* alpha/red/green/blue */
+  JCS_RGB565              /* 5-bit red/6-bit green/5-bit blue */
+} J_COLOR_SPACE;
+
+/* DCT/IDCT algorithm options. */
+
+typedef enum {
+  JDCT_ISLOW,             /* accurate integer method */
+  JDCT_IFAST,             /* less accurate integer method legacy feature */
+  JDCT_FLOAT              /* floating-point method legacy feature */
+} J_DCT_METHOD;
+
+#ifndef JDCT_DEFAULT            /* may be overridden in jconfig.h */
+#define JDCT_DEFAULT  JDCT_ISLOW
+#endif
+#ifndef JDCT_FASTEST            /* may be overridden in jconfig.h */
+#define JDCT_FASTEST  JDCT_IFAST
+#endif
+
+/* Dithering options for decompression. */
+
+typedef enum {
+  JDITHER_NONE,           /* no dithering */
+  JDITHER_ORDERED,        /* simple ordered dither */
+  JDITHER_FS              /* Floyd-Steinberg error diffusion dither */
+} J_DITHER_MODE;
+
+
+/* Common fields between JPEG compression and decompression master structs. */
+
+#define jpeg_common_fields \
+  struct jpeg_error_mgr *err;   /* Error handler module */ \
+  struct jpeg_memory_mgr *mem;  /* Memory manager module */ \
+  struct jpeg_progress_mgr *progress; /* Progress monitor, or NULL if none */ \
+  void *client_data;            /* Available for use by application */ \
+  boolean is_decompressor;      /* So common code can tell which is which */ \
+  int global_state              /* For checking call sequence validity */
+
+/* Routines that are to be used by both halves of the library are declared
+ * to receive a pointer to this structure.  There are no actual instances of
+ * jpeg_common_struct, only of jpeg_compress_struct and jpeg_decompress_struct.
+ */
+struct jpeg_common_struct {
+  jpeg_common_fields;           /* Fields common to both master struct types */
+  /* Additional fields follow in an actual jpeg_compress_struct or
+   * jpeg_decompress_struct.  All three structs must agree on these
+   * initial fields!  (This would be a lot cleaner in C++.)
+   */
+};
+
+typedef struct jpeg_common_struct *j_common_ptr;
+typedef struct jpeg_compress_struct *j_compress_ptr;
+typedef struct jpeg_decompress_struct *j_decompress_ptr;
+
+
+/* Master record for a compression instance */
+
+struct jpeg_compress_struct {
+  jpeg_common_fields;           /* Fields shared with jpeg_decompress_struct */
+
+  /* Destination for compressed data */
+  struct jpeg_destination_mgr *dest;
+
+  /* Description of source image --- these fields must be filled in by
+   * outer application before starting compression.  in_color_space must
+   * be correct before you can even call jpeg_set_defaults().
+   */
+
+  JDIMENSION image_width;       /* input image width */
+  JDIMENSION image_height;      /* input image height */
+  int input_components;         /* # of color components in input image */
+  J_COLOR_SPACE in_color_space; /* colorspace of input image */
+
+  double input_gamma;           /* image gamma of input image */
+
+  /* Compression parameters --- these fields must be set before calling
+   * jpeg_start_compress().  We recommend calling jpeg_set_defaults() to
+   * initialize everything to reasonable defaults, then changing anything
+   * the application specifically wants to change.  That way you won't get
+   * burnt when new parameters are added.  Also note that there are several
+   * helper routines to simplify changing parameters.
+   */
+
+#if JPEG_LIB_VERSION >= 70
+  unsigned int scale_num, scale_denom; /* fraction by which to scale image */
+
+  JDIMENSION jpeg_width;        /* scaled JPEG image width */
+  JDIMENSION jpeg_height;       /* scaled JPEG image height */
+  /* Dimensions of actual JPEG image that will be written to file,
+   * derived from input dimensions by scaling factors above.
+   * These fields are computed by jpeg_start_compress().
+   * You can also use jpeg_calc_jpeg_dimensions() to determine these values
+   * in advance of calling jpeg_start_compress().
+   */
+#endif
+
+  int data_precision;           /* bits of precision in image data */
+
+  int num_components;           /* # of color components in JPEG image */
+  J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
+
+  jpeg_component_info *comp_info;
+  /* comp_infoi describes component that appears i'th in SOF */
+
+  JQUANT_TBL *quant_tbl_ptrsNUM_QUANT_TBLS;
+#if JPEG_LIB_VERSION >= 70
+  int q_scale_factorNUM_QUANT_TBLS;
+#endif
+  /* ptrs to coefficient quantization tables, or NULL if not defined,
+   * and corresponding scale factors (percentage, initialized 100).
+   */
+
+  JHUFF_TBL *dc_huff_tbl_ptrsNUM_HUFF_TBLS;
+  JHUFF_TBL *ac_huff_tbl_ptrsNUM_HUFF_TBLS;
+  /* ptrs to Huffman coding tables, or NULL if not defined */
+
+  UINT8 arith_dc_LNUM_ARITH_TBLS; /* L values for DC arith-coding tables */
+  UINT8 arith_dc_UNUM_ARITH_TBLS; /* U values for DC arith-coding tables */
+  UINT8 arith_ac_KNUM_ARITH_TBLS; /* Kx values for AC arith-coding tables */
+
+  int num_scans;                /* # of entries in scan_info array */
+  const jpeg_scan_info *scan_info; /* script for multi-scan file, or NULL */
+  /* The default value of scan_info is NULL, which causes a single-scan
+   * sequential JPEG file to be emitted.  To create a multi-scan file,
+   * set num_scans and scan_info to point to an array of scan definitions.
+   */
+
+  boolean raw_data_in;          /* TRUE=caller supplies downsampled data */
+  boolean arith_code;           /* TRUE=arithmetic coding, FALSE=Huffman */
+  boolean optimize_coding;      /* TRUE=optimize entropy encoding parms */
+  boolean CCIR601_sampling;     /* TRUE=first samples are cosited */
+#if JPEG_LIB_VERSION >= 70
+  boolean do_fancy_downsampling; /* TRUE=apply fancy downsampling */
+#endif
+  int smoothing_factor;         /* 1..100, or 0 for no input smoothing */
+  J_DCT_METHOD dct_method;      /* DCT algorithm selector */
+
+  /* The restart interval can be specified in absolute MCUs by setting
+   * restart_interval, or in MCU rows by setting restart_in_rows
+   * (in which case the correct restart_interval will be figured
+   * for each scan).
+   */
+  unsigned int restart_interval; /* MCUs per restart, or 0 for no restart */
+  int restart_in_rows;          /* if > 0, MCU rows per restart interval */
+
+  /* Parameters controlling emission of special markers. */
+
+  boolean write_JFIF_header;    /* should a JFIF marker be written? */
+  UINT8 JFIF_major_version;     /* What to write for the JFIF version number */
+  UINT8 JFIF_minor_version;
+  /* These three values are not used by the JPEG code, merely copied */
+  /* into the JFIF APP0 marker.  density_unit can be 0 for unknown, */
+  /* 1 for dots/inch, or 2 for dots/cm.  Note that the pixel aspect */
+  /* ratio is defined by X_density/Y_density even when density_unit=0. */
+  UINT8 density_unit;           /* JFIF code for pixel size units */
+  UINT16 X_density;             /* Horizontal pixel density */
+  UINT16 Y_density;             /* Vertical pixel density */
+  boolean write_Adobe_marker;   /* should an Adobe marker be written? */
+
+  /* State variable: index of next scanline to be written to
+   * jpeg_write_scanlines().  Application may use this to control its
+   * processing loop, e.g., "while (next_scanline < image_height)".
+   */
+
+  JDIMENSION next_scanline;     /* 0 .. image_height-1  */
+
+  /* Remaining fields are known throughout compressor, but generally
+   * should not be touched by a surrounding application.
+   */
+
+  /*
+   * These fields are computed during compression startup
+   */
+  boolean progressive_mode;     /* TRUE if scan script uses progressive mode */
+  int max_h_samp_factor;        /* largest h_samp_factor */
+  int max_v_samp_factor;        /* largest v_samp_factor */
+
+#if JPEG_LIB_VERSION >= 70
+  int min_DCT_h_scaled_size;    /* smallest DCT_h_scaled_size of any component */
+  int min_DCT_v_scaled_size;    /* smallest DCT_v_scaled_size of any component */
+#endif
+
+  JDIMENSION total_iMCU_rows;   /* # of iMCU rows to be input to coef ctlr */
+  /* The coefficient controller receives data in units of MCU rows as defined
+   * for fully interleaved scans (whether the JPEG file is interleaved or not).
+   * There are v_samp_factor * DCTSIZE sample rows of each component in an
+   * "iMCU" (interleaved MCU) row.
+   */
+
+  /*
+   * These fields are valid during any one scan.
+   * They describe the components and MCUs actually appearing in the scan.
+   */
+  int comps_in_scan;            /* # of JPEG components in this scan */
+  jpeg_component_info *cur_comp_infoMAX_COMPS_IN_SCAN;
+  /* *cur_comp_infoi describes component that appears i'th in SOS */
+
+  JDIMENSION MCUs_per_row;      /* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;  /* # of MCU rows in the image */
+
+  int blocks_in_MCU;            /* # of DCT blocks per MCU */
+  int MCU_membershipC_MAX_BLOCKS_IN_MCU;
+  /* MCU_membershipi is index in cur_comp_info of component owning */
+  /* i'th block in an MCU */
+
+  int Ss, Se, Ah, Al;           /* progressive JPEG parameters for scan */
+
+#if JPEG_LIB_VERSION >= 80
+  int block_size;               /* the basic DCT block size: 1..16 */
+  const int *natural_order;     /* natural-order position array */
+  int lim_Se;                   /* min( Se, DCTSIZE2-1 ) */
+#endif
+
+  /*
+   * Links to compression subobjects (methods and private variables of modules)
+   */
+  struct jpeg_comp_master *master;
+  struct jpeg_c_main_controller *main;
+  struct jpeg_c_prep_controller *prep;
+  struct jpeg_c_coef_controller *coef;
+  struct jpeg_marker_writer *marker;
+  struct jpeg_color_converter *cconvert;
+  struct jpeg_downsampler *downsample;
+  struct jpeg_forward_dct *fdct;
+  struct jpeg_entropy_encoder *entropy;
+  jpeg_scan_info *script_space; /* workspace for jpeg_simple_progression */
+  int script_space_size;
+};
+
+
+/* Master record for a decompression instance */
+
+struct jpeg_decompress_struct {
+  jpeg_common_fields;           /* Fields shared with jpeg_compress_struct */
+
+  /* Source of compressed data */
+  struct jpeg_source_mgr *src;
+
+  /* Basic description of image --- filled in by jpeg_read_header(). */
+  /* Application may inspect these values to decide how to process image. */
+
+  JDIMENSION image_width;       /* nominal image width (from SOF marker) */
+  JDIMENSION image_height;      /* nominal image height */
+  int num_components;           /* # of color components in JPEG image */
+  J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
+
+  /* Decompression processing parameters --- these fields must be set before
+   * calling jpeg_start_decompress().  Note that jpeg_read_header() initializes
+   * them to default values.
+   */
+
+  J_COLOR_SPACE out_color_space; /* colorspace for output */
+
+  unsigned int scale_num, scale_denom; /* fraction by which to scale image */
+
+  double output_gamma;          /* image gamma wanted in output */
+
+  boolean buffered_image;       /* TRUE=multiple output passes */
+  boolean raw_data_out;         /* TRUE=downsampled data wanted */
+
+  J_DCT_METHOD dct_method;      /* IDCT algorithm selector */
+  boolean do_fancy_upsampling;  /* TRUE=apply fancy upsampling */
+  boolean do_block_smoothing;   /* TRUE=apply interblock smoothing */
+
+  boolean quantize_colors;      /* TRUE=colormapped output wanted */
+  /* the following are ignored if not quantize_colors: */
+  J_DITHER_MODE dither_mode;    /* type of color dithering to use */
+  boolean two_pass_quantize;    /* TRUE=use two-pass color quantization */
+  int desired_number_of_colors; /* max # colors to use in created colormap */
+  /* these are significant only in buffered-image mode: */
+  boolean enable_1pass_quant;   /* enable future use of 1-pass quantizer */
+  boolean enable_external_quant;/* enable future use of external colormap */
+  boolean enable_2pass_quant;   /* enable future use of 2-pass quantizer */
+
+  /* Description of actual output image that will be returned to application.
+   * These fields are computed by jpeg_start_decompress().
+   * You can also use jpeg_calc_output_dimensions() to determine these values
+   * in advance of calling jpeg_start_decompress().
+   */
+
+  JDIMENSION output_width;      /* scaled image width */
+  JDIMENSION output_height;     /* scaled image height */
+  int out_color_components;     /* # of color components in out_color_space */
+  int output_components;        /* # of color components returned */
+  /* output_components is 1 (a colormap index) when quantizing colors;
+   * otherwise it equals out_color_components.
+   */
+  int rec_outbuf_height;        /* min recommended height of scanline buffer */
+  /* If the buffer passed to jpeg_read_scanlines() is less than this many rows
+   * high, space and time will be wasted due to unnecessary data copying.
+   * Usually rec_outbuf_height will be 1 or 2, at most 4.
+   */
+
+  /* When quantizing colors, the output colormap is described by these fields.
+   * The application can supply a colormap by setting colormap non-NULL before
+   * calling jpeg_start_decompress; otherwise a colormap is created during
+   * jpeg_start_decompress or jpeg_start_output.
+   * The map has out_color_components rows and actual_number_of_colors columns.
+   */
+  int actual_number_of_colors;  /* number of entries in use */
+  JSAMPARRAY colormap;          /* The color map as a 2-D pixel array */
+
+  /* State variables: these variables indicate the progress of decompression.
+   * The application may examine these but must not modify them.
+   */
+
+  /* Row index of next scanline to be read from jpeg_read_scanlines().
+   * Application may use this to control its processing loop, e.g.,
+   * "while (output_scanline < output_height)".
+   */
+  JDIMENSION output_scanline;   /* 0 .. output_height-1  */
+
+  /* Current input scan number and number of iMCU rows completed in scan.
+   * These indicate the progress of the decompressor input side.
+   */
+  int input_scan_number;        /* Number of SOS markers seen so far */
+  JDIMENSION input_iMCU_row;    /* Number of iMCU rows completed */
+
+  /* The "output scan number" is the notional scan being displayed by the
+   * output side.  The decompressor will not allow output scan/row number
+   * to get ahead of input scan/row, but it can fall arbitrarily far behind.
+   */
+  int output_scan_number;       /* Nominal scan number being displayed */
+  JDIMENSION output_iMCU_row;   /* Number of iMCU rows read */
+
+  /* Current progression status.  coef_bitsci indicates the precision
+   * with which component c's DCT coefficient i (in zigzag order) is known.
+   * It is -1 when no data has yet been received, otherwise it is the point
+   * transform (shift) value for the most recent scan of the coefficient
+   * (thus, 0 at completion of the progression).
+   * This pointer is NULL when reading a non-progressive file.
+   */
+  int (*coef_bits)DCTSIZE2;   /* -1 or current Al value for each coef */
+
+  /* Internal JPEG parameters --- the application usually need not look at
+   * these fields.  Note that the decompressor output side may not use
+   * any parameters that can change between scans.
+   */
+
+  /* Quantization and Huffman tables are carried forward across input
+   * datastreams when processing abbreviated JPEG datastreams.
+   */
+
+  JQUANT_TBL *quant_tbl_ptrsNUM_QUANT_TBLS;
+  /* ptrs to coefficient quantization tables, or NULL if not defined */
+
+  JHUFF_TBL *dc_huff_tbl_ptrsNUM_HUFF_TBLS;
+  JHUFF_TBL *ac_huff_tbl_ptrsNUM_HUFF_TBLS;
+  /* ptrs to Huffman coding tables, or NULL if not defined */
+
+  /* These parameters are never carried across datastreams, since they
+   * are given in SOF/SOS markers or defined to be reset by SOI.
+   */
+
+  int data_precision;           /* bits of precision in image data */
+
+  jpeg_component_info *comp_info;
+  /* comp_infoi describes component that appears i'th in SOF */
+
+#if JPEG_LIB_VERSION >= 80
+  boolean is_baseline;          /* TRUE if Baseline SOF0 encountered */
+#endif
+  boolean progressive_mode;     /* TRUE if SOFn specifies progressive mode */
+  boolean arith_code;           /* TRUE=arithmetic coding, FALSE=Huffman */
+
+  UINT8 arith_dc_LNUM_ARITH_TBLS; /* L values for DC arith-coding tables */
+  UINT8 arith_dc_UNUM_ARITH_TBLS; /* U values for DC arith-coding tables */
+  UINT8 arith_ac_KNUM_ARITH_TBLS; /* Kx values for AC arith-coding tables */
+
+  unsigned int restart_interval; /* MCUs per restart interval, or 0 for no restart */
+
+  /* These fields record data obtained from optional markers recognized by
+   * the JPEG library.
+   */
+  boolean saw_JFIF_marker;      /* TRUE iff a JFIF APP0 marker was found */
+  /* Data copied from JFIF marker; only valid if saw_JFIF_marker is TRUE: */
+  UINT8 JFIF_major_version;     /* JFIF version number */
+  UINT8 JFIF_minor_version;
+  UINT8 density_unit;           /* JFIF code for pixel size units */
+  UINT16 X_density;             /* Horizontal pixel density */
+  UINT16 Y_density;             /* Vertical pixel density */
+  boolean saw_Adobe_marker;     /* TRUE iff an Adobe APP14 marker was found */
+  UINT8 Adobe_transform;        /* Color transform code from Adobe marker */
+
+  boolean CCIR601_sampling;     /* TRUE=first samples are cosited */
+
+  /* Aside from the specific data retained from APPn markers known to the
+   * library, the uninterpreted contents of any or all APPn and COM markers
+   * can be saved in a list for examination by the application.
+   */
+  jpeg_saved_marker_ptr marker_list; /* Head of list of saved markers */
+
+  /* Remaining fields are known throughout decompressor, but generally
+   * should not be touched by a surrounding application.
+   */
+
+  /*
+   * These fields are computed during decompression startup
+   */
+  int max_h_samp_factor;        /* largest h_samp_factor */
+  int max_v_samp_factor;        /* largest v_samp_factor */
+
+#if JPEG_LIB_VERSION >= 70
+  int min_DCT_h_scaled_size;    /* smallest DCT_h_scaled_size of any component */
+  int min_DCT_v_scaled_size;    /* smallest DCT_v_scaled_size of any component */
+#else
+  int min_DCT_scaled_size;      /* smallest DCT_scaled_size of any component */
+#endif
+
+  JDIMENSION total_iMCU_rows;   /* # of iMCU rows in image */
+  /* The coefficient controller's input and output progress is measured in
+   * units of "iMCU" (interleaved MCU) rows.  These are the same as MCU rows
+   * in fully interleaved JPEG scans, but are used whether the scan is
+   * interleaved or not.  We define an iMCU row as v_samp_factor DCT block
+   * rows of each component.  Therefore, the IDCT output contains
+   * v_samp_factor*DCT_v_scaled_size sample rows of a component per iMCU row.
+   */
+
+  JSAMPLE *sample_range_limit;  /* table for fast range-limiting */
+
+  /*
+   * These fields are valid during any one scan.
+   * They describe the components and MCUs actually appearing in the scan.
+   * Note that the decompressor output side must not use these fields.
+   */
+  int comps_in_scan;            /* # of JPEG components in this scan */
+  jpeg_component_info *cur_comp_infoMAX_COMPS_IN_SCAN;
+  /* *cur_comp_infoi describes component that appears i'th in SOS */
+
+  JDIMENSION MCUs_per_row;      /* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;  /* # of MCU rows in the image */
+
+  int blocks_in_MCU;            /* # of DCT blocks per MCU */
+  int MCU_membershipD_MAX_BLOCKS_IN_MCU;
+  /* MCU_membershipi is index in cur_comp_info of component owning */
+  /* i'th block in an MCU */
+
+  int Ss, Se, Ah, Al;           /* progressive JPEG parameters for scan */
+
+#if JPEG_LIB_VERSION >= 80
+  /* These fields are derived from Se of first SOS marker.
+   */
+  int block_size;               /* the basic DCT block size: 1..16 */
+  const int *natural_order; /* natural-order position array for entropy decode */
+  int lim_Se;                   /* min( Se, DCTSIZE2-1 ) for entropy decode */
+#endif
+
+  /* This field is shared between entropy decoder and marker parser.
+   * It is either zero or the code of a JPEG marker that has been
+   * read from the data source, but has not yet been processed.
+   */
+  int unread_marker;
+
+  /*
+   * Links to decompression subobjects (methods, private variables of modules)
+   */
+  struct jpeg_decomp_master *master;
+  struct jpeg_d_main_controller *main;
+  struct jpeg_d_coef_controller *coef;
+  struct jpeg_d_post_controller *post;
+  struct jpeg_input_controller *inputctl;
+  struct jpeg_marker_reader *marker;
+  struct jpeg_entropy_decoder *entropy;
+  struct jpeg_inverse_dct *idct;
+  struct jpeg_upsampler *upsample;
+  struct jpeg_color_deconverter *cconvert;
+  struct jpeg_color_quantizer *cquantize;
+};
+
+
+/* "Object" declarations for JPEG modules that may be supplied or called
+ * directly by the surrounding application.
+ * As with all objects in the JPEG library, these structs only define the
+ * publicly visible methods and state variables of a module.  Additional
+ * private fields may exist after the public ones.
+ */
+
+
+/* Error handler object */
+
+struct jpeg_error_mgr {
+  /* Error exit handler: does not return to caller */
+  void (*error_exit) (j_common_ptr cinfo);
+  /* Conditionally emit a trace or warning message */
+  void (*emit_message) (j_common_ptr cinfo, int msg_level);
+  /* Routine that actually outputs a trace or error message */
+  void (*output_message) (j_common_ptr cinfo);
+  /* Format a message string for the most recent JPEG error or message */
+  void (*format_message) (j_common_ptr cinfo, char *buffer);
+#define JMSG_LENGTH_MAX  200    /* recommended size of format_message buffer */
+  /* Reset error state variables at start of a new image */
+  void (*reset_error_mgr) (j_common_ptr cinfo);
+
+  /* The message ID code and any parameters are saved here.
+   * A message can have one string parameter or up to 8 int parameters.
+   */
+  int msg_code;
+#define JMSG_STR_PARM_MAX  80
+  union {
+    int i8;
+    char sJMSG_STR_PARM_MAX;
+  } msg_parm;
+
+  /* Standard state variables for error facility */
+
+  int trace_level;              /* max msg_level that will be displayed */
+
+  /* For recoverable corrupt-data errors, we emit a warning message,
+   * but keep going unless emit_message chooses to abort.  emit_message
+   * should count warnings in num_warnings.  The surrounding application
+   * can check for bad data by seeing if num_warnings is nonzero at the
+   * end of processing.
+   */
+  long num_warnings;            /* number of corrupt-data warnings */
+
+  /* These fields point to the table(s) of error message strings.
+   * An application can change the table pointer to switch to a different
+   * message list (typically, to change the language in which errors are
+   * reported).  Some applications may wish to add additional error codes
+   * that will be handled by the JPEG library error mechanism; the second
+   * table pointer is used for this purpose.
+   *
+   * First table includes all errors generated by JPEG library itself.
+   * Error code 0 is reserved for a "no such error string" message.
+   */
+  const char * const *jpeg_message_table; /* Library errors */
+  int last_jpeg_message;    /* Table contains strings 0..last_jpeg_message */
+  /* Second table can be added by application (see cjpeg/djpeg for example).
+   * It contains strings numbered first_addon_message..last_addon_message.
+   */
+  const char * const *addon_message_table; /* Non-library errors */
+  int first_addon_message;      /* code for first string in addon table */
+  int last_addon_message;       /* code for last string in addon table */
+};
+
+
+/* Progress monitor object */
+
+struct jpeg_progress_mgr {
+  void (*progress_monitor) (j_common_ptr cinfo);
+
+  long pass_counter;            /* work units completed in this pass */
+  long pass_limit;              /* total number of work units in this pass */
+  int completed_passes;         /* passes completed so far */
+  int total_passes;             /* total number of passes expected */
+};
+
+
+/* Data destination object for compression */
+
+struct jpeg_destination_mgr {
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
+
+  void (*init_destination) (j_compress_ptr cinfo);
+  boolean (*empty_output_buffer) (j_compress_ptr cinfo);
+  void (*term_destination) (j_compress_ptr cinfo);
+};
+
+
+/* Data source object for decompression */
+
+struct jpeg_source_mgr {
+  const JOCTET *next_input_byte; /* => next byte to read from buffer */
+  size_t bytes_in_buffer;       /* # of bytes remaining in buffer */
+
+  void (*init_source) (j_decompress_ptr cinfo);
+  boolean (*fill_input_buffer) (j_decompress_ptr cinfo);
+  void (*skip_input_data) (j_decompress_ptr cinfo, long num_bytes);
+  boolean (*resync_to_restart) (j_decompress_ptr cinfo, int desired);
+  void (*term_source) (j_decompress_ptr cinfo);
+};
+
+
+/* Memory manager object.
+ * Allocates "small" objects (a few K total), "large" objects (tens of K),
+ * and "really big" objects (virtual arrays with backing store if needed).
+ * The memory manager does not allow individual objects to be freed; rather,
+ * each created object is assigned to a pool, and whole pools can be freed
+ * at once.  This is faster and more convenient than remembering exactly what
+ * to free, especially where malloc()/free() are not too speedy.
+ * NB: alloc routines never return NULL.  They exit to error_exit if not
+ * successful.
+ */
+
+#define JPOOL_PERMANENT  0      /* lasts until master record is destroyed */
+#define JPOOL_IMAGE      1      /* lasts until done with image/datastream */
+#define JPOOL_NUMPOOLS   2
+
+typedef struct jvirt_sarray_control *jvirt_sarray_ptr;
+typedef struct jvirt_barray_control *jvirt_barray_ptr;
+
+
+struct jpeg_memory_mgr {
+  /* Method pointers */
+  void *(*alloc_small) (j_common_ptr cinfo, int pool_id, size_t sizeofobject);
+  void *(*alloc_large) (j_common_ptr cinfo, int pool_id,
+                        size_t sizeofobject);
+  JSAMPARRAY (*alloc_sarray) (j_common_ptr cinfo, int pool_id,
+                              JDIMENSION samplesperrow, JDIMENSION numrows);
+  JBLOCKARRAY (*alloc_barray) (j_common_ptr cinfo, int pool_id,
+                               JDIMENSION blocksperrow, JDIMENSION numrows);
+  jvirt_sarray_ptr (*request_virt_sarray) (j_common_ptr cinfo, int pool_id,
+                                           boolean pre_zero,
+                                           JDIMENSION samplesperrow,
+                                           JDIMENSION numrows,
+                                           JDIMENSION maxaccess);
+  jvirt_barray_ptr (*request_virt_barray) (j_common_ptr cinfo, int pool_id,
+                                           boolean pre_zero,
+                                           JDIMENSION blocksperrow,
+                                           JDIMENSION numrows,
+                                           JDIMENSION maxaccess);
+  void (*realize_virt_arrays) (j_common_ptr cinfo);
+  JSAMPARRAY (*access_virt_sarray) (j_common_ptr cinfo, jvirt_sarray_ptr ptr,
+                                    JDIMENSION start_row, JDIMENSION num_rows,
+                                    boolean writable);
+  JBLOCKARRAY (*access_virt_barray) (j_common_ptr cinfo, jvirt_barray_ptr ptr,
+                                     JDIMENSION start_row, JDIMENSION num_rows,
+                                     boolean writable);
+  void (*free_pool) (j_common_ptr cinfo, int pool_id);
+  void (*self_destruct) (j_common_ptr cinfo);
+
+  /* Limit on memory allocation for this JPEG object.  (Note that this is
+   * merely advisory, not a guaranteed maximum; it only affects the space
+   * used for virtual-array buffers.)  May be changed by outer application
+   * after creating the JPEG object.
+   */
+  long max_memory_to_use;
+
+  /* Maximum allocation request accepted by alloc_large. */
+  long max_alloc_chunk;
+};
+
+
+/* Routine signature for application-supplied marker processing methods.
+ * Need not pass marker code since it is stored in cinfo->unread_marker.
+ */
+typedef boolean (*jpeg_marker_parser_method) (j_decompress_ptr cinfo);
+
+
+/* Originally, this macro was used as a way of defining function prototypes
+ * for both modern compilers as well as older compilers that did not support
+ * prototype parameters.  libjpeg-turbo has never supported these older,
+ * non-ANSI compilers, but the macro is still included because there is some
+ * software out there that uses it.
+ */
+
+#define JPP(arglist)    arglist
+
+
+/* Default error-management setup */
+EXTERN(struct jpeg_error_mgr *) jpeg_std_error(struct jpeg_error_mgr *err);
+
+/* Initialization of JPEG compression objects.
+ * jpeg_create_compress() and jpeg_create_decompress() are the exported
+ * names that applications should call.  These expand to calls on
+ * jpeg_CreateCompress and jpeg_CreateDecompress with additional information
+ * passed for version mismatch checking.
+ * NB: you must set up the error-manager BEFORE calling jpeg_create_xxx.
+ */
+#define jpeg_create_compress(cinfo) \
+  jpeg_CreateCompress((cinfo), JPEG_LIB_VERSION, \
+                      (size_t)sizeof(struct jpeg_compress_struct))
+#define jpeg_create_decompress(cinfo) \
+  jpeg_CreateDecompress((cinfo), JPEG_LIB_VERSION, \
+                        (size_t)sizeof(struct jpeg_decompress_struct))
+EXTERN(void) jpeg_CreateCompress(j_compress_ptr cinfo, int version,
+                                 size_t structsize);
+EXTERN(void) jpeg_CreateDecompress(j_decompress_ptr cinfo, int version,
+                                   size_t structsize);
+/* Destruction of JPEG compression objects */
+EXTERN(void) jpeg_destroy_compress(j_compress_ptr cinfo);
+EXTERN(void) jpeg_destroy_decompress(j_decompress_ptr cinfo);
+
+/* Standard data source and destination managers: stdio streams. */
+/* Caller is responsible for opening the file before and closing after. */
+EXTERN(void) jpeg_stdio_dest(j_compress_ptr cinfo, FILE *outfile);
+EXTERN(void) jpeg_stdio_src(j_decompress_ptr cinfo, FILE *infile);
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/* Data source and destination managers: memory buffers. */
+EXTERN(void) jpeg_mem_dest(j_compress_ptr cinfo, unsigned char **outbuffer,
+                           unsigned long *outsize);
+EXTERN(void) jpeg_mem_src(j_decompress_ptr cinfo,
+                          const unsigned char *inbuffer, unsigned long insize);
+#endif
+
+/* Default parameter setup for compression */
+EXTERN(void) jpeg_set_defaults(j_compress_ptr cinfo);
+/* Compression parameter setup aids */
+EXTERN(void) jpeg_set_colorspace(j_compress_ptr cinfo,
+                                 J_COLOR_SPACE colorspace);
+EXTERN(void) jpeg_default_colorspace(j_compress_ptr cinfo);
+EXTERN(void) jpeg_set_quality(j_compress_ptr cinfo, int quality,
+                              boolean force_baseline);
+EXTERN(void) jpeg_set_linear_quality(j_compress_ptr cinfo, int scale_factor,
+                                     boolean force_baseline);
+#if JPEG_LIB_VERSION >= 70
+EXTERN(void) jpeg_default_qtables(j_compress_ptr cinfo,
+                                  boolean force_baseline);
+#endif
+EXTERN(void) jpeg_add_quant_table(j_compress_ptr cinfo, int which_tbl,
+                                  const unsigned int *basic_table,
+                                  int scale_factor, boolean force_baseline);
+EXTERN(int) jpeg_quality_scaling(int quality);
+EXTERN(void) jpeg_simple_progression(j_compress_ptr cinfo);
+EXTERN(void) jpeg_suppress_tables(j_compress_ptr cinfo, boolean suppress);
+EXTERN(JQUANT_TBL *) jpeg_alloc_quant_table(j_common_ptr cinfo);
+EXTERN(JHUFF_TBL *) jpeg_alloc_huff_table(j_common_ptr cinfo);
+
+/* Main entry points for compression */
+EXTERN(void) jpeg_start_compress(j_compress_ptr cinfo,
+                                 boolean write_all_tables);
+EXTERN(JDIMENSION) jpeg_write_scanlines(j_compress_ptr cinfo,
+                                        JSAMPARRAY scanlines,
+                                        JDIMENSION num_lines);
+EXTERN(void) jpeg_finish_compress(j_compress_ptr cinfo);
+
+#if JPEG_LIB_VERSION >= 70
+/* Precalculate JPEG dimensions for current compression parameters. */
+EXTERN(void) jpeg_calc_jpeg_dimensions(j_compress_ptr cinfo);
+#endif
+
+/* Replaces jpeg_write_scanlines when writing raw downsampled data. */
+EXTERN(JDIMENSION) jpeg_write_raw_data(j_compress_ptr cinfo, JSAMPIMAGE data,
+                                       JDIMENSION num_lines);
+
+/* Write a special marker.  See libjpeg.txt concerning safe usage. */
+EXTERN(void) jpeg_write_marker(j_compress_ptr cinfo, int marker,
+                               const JOCTET *dataptr, unsigned int datalen);
+/* Same, but piecemeal. */
+EXTERN(void) jpeg_write_m_header(j_compress_ptr cinfo, int marker,
+                                 unsigned int datalen);
+EXTERN(void) jpeg_write_m_byte(j_compress_ptr cinfo, int val);
+
+/* Alternate compression function: just write an abbreviated table file */
+EXTERN(void) jpeg_write_tables(j_compress_ptr cinfo);
+
+/* Write ICC profile.  See libjpeg.txt for usage information. */
+EXTERN(void) jpeg_write_icc_profile(j_compress_ptr cinfo,
+                                    const JOCTET *icc_data_ptr,
+                                    unsigned int icc_data_len);
+
+
+/* Decompression startup: read start of JPEG datastream to see what's there */
+EXTERN(int) jpeg_read_header(j_decompress_ptr cinfo, boolean require_image);
+/* Return value is one of: */
+#define JPEG_SUSPENDED           0 /* Suspended due to lack of input data */
+#define JPEG_HEADER_OK           1 /* Found valid image datastream */
+#define JPEG_HEADER_TABLES_ONLY  2 /* Found valid table-specs-only datastream */
+/* If you pass require_image = TRUE (normal case), you need not check for
+ * a TABLES_ONLY return code; an abbreviated file will cause an error exit.
+ * JPEG_SUSPENDED is only possible if you use a data source module that can
+ * give a suspension return (the stdio source module doesn't).
+ */
+
+/* Main entry points for decompression */
+EXTERN(boolean) jpeg_start_decompress(j_decompress_ptr cinfo);
+EXTERN(JDIMENSION) jpeg_read_scanlines(j_decompress_ptr cinfo,
+                                       JSAMPARRAY scanlines,
+                                       JDIMENSION max_lines);
+EXTERN(JDIMENSION) jpeg_skip_scanlines(j_decompress_ptr cinfo,
+                                       JDIMENSION num_lines);
+EXTERN(void) jpeg_crop_scanline(j_decompress_ptr cinfo, JDIMENSION *xoffset,
+                                JDIMENSION *width);
+EXTERN(boolean) jpeg_finish_decompress(j_decompress_ptr cinfo);
+
+/* Replaces jpeg_read_scanlines when reading raw downsampled data. */
+EXTERN(JDIMENSION) jpeg_read_raw_data(j_decompress_ptr cinfo, JSAMPIMAGE data,
+                                      JDIMENSION max_lines);
+
+/* Additional entry points for buffered-image mode. */
+EXTERN(boolean) jpeg_has_multiple_scans(j_decompress_ptr cinfo);
+EXTERN(boolean) jpeg_start_output(j_decompress_ptr cinfo, int scan_number);
+EXTERN(boolean) jpeg_finish_output(j_decompress_ptr cinfo);
+EXTERN(boolean) jpeg_input_complete(j_decompress_ptr cinfo);
+EXTERN(void) jpeg_new_colormap(j_decompress_ptr cinfo);
+EXTERN(int) jpeg_consume_input(j_decompress_ptr cinfo);
+/* Return value is one of: */
+/* #define JPEG_SUSPENDED       0    Suspended due to lack of input data */
+#define JPEG_REACHED_SOS        1 /* Reached start of new scan */
+#define JPEG_REACHED_EOI        2 /* Reached end of image */
+#define JPEG_ROW_COMPLETED      3 /* Completed one iMCU row */
+#define JPEG_SCAN_COMPLETED     4 /* Completed last iMCU row of a scan */
+
+/* Precalculate output dimensions for current decompression parameters. */
+#if JPEG_LIB_VERSION >= 80
+EXTERN(void) jpeg_core_output_dimensions(j_decompress_ptr cinfo);
+#endif
+EXTERN(void) jpeg_calc_output_dimensions(j_decompress_ptr cinfo);
+
+/* Control saving of COM and APPn markers into marker_list. */
+EXTERN(void) jpeg_save_markers(j_decompress_ptr cinfo, int marker_code,
+                               unsigned int length_limit);
+
+/* Install a special processing method for COM or APPn markers. */
+EXTERN(void) jpeg_set_marker_processor(j_decompress_ptr cinfo,
+                                       int marker_code,
+                                       jpeg_marker_parser_method routine);
+
+/* Read or write raw DCT coefficients --- useful for lossless transcoding. */
+EXTERN(jvirt_barray_ptr *) jpeg_read_coefficients(j_decompress_ptr cinfo);
+EXTERN(void) jpeg_write_coefficients(j_compress_ptr cinfo,
+                                     jvirt_barray_ptr *coef_arrays);
+EXTERN(void) jpeg_copy_critical_parameters(j_decompress_ptr srcinfo,
+                                           j_compress_ptr dstinfo);
+
+/* If you choose to abort compression or decompression before completing
+ * jpeg_finish_(de)compress, then you need to clean up to release memory,
+ * temporary files, etc.  You can just call jpeg_destroy_(de)compress
+ * if you're done with the JPEG object, but if you want to clean it up and
+ * reuse it, call this:
+ */
+EXTERN(void) jpeg_abort_compress(j_compress_ptr cinfo);
+EXTERN(void) jpeg_abort_decompress(j_decompress_ptr cinfo);
+
+/* Generic versions of jpeg_abort and jpeg_destroy that work on either
+ * flavor of JPEG object.  These may be more convenient in some places.
+ */
+EXTERN(void) jpeg_abort(j_common_ptr cinfo);
+EXTERN(void) jpeg_destroy(j_common_ptr cinfo);
+
+/* Default restart-marker-resync procedure for use by data source modules */
+EXTERN(boolean) jpeg_resync_to_restart(j_decompress_ptr cinfo, int desired);
+
+/* Read ICC profile.  See libjpeg.txt for usage information. */
+EXTERN(boolean) jpeg_read_icc_profile(j_decompress_ptr cinfo,
+                                      JOCTET **icc_data_ptr,
+                                      unsigned int *icc_data_len);
+
+
+/* These marker codes are exported since applications and data source modules
+ * are likely to want to use them.
+ */
+
+#define JPEG_RST0       0xD0    /* RST0 marker code */
+#define JPEG_EOI        0xD9    /* EOI marker code */
+#define JPEG_APP0       0xE0    /* APP0 marker code */
+#define JPEG_COM        0xFE    /* COM marker code */
+
+
+/* If we have a brain-damaged compiler that emits warnings (or worse, errors)
+ * for structure definitions that are never filled in, keep it quiet by
+ * supplying dummy definitions for the various substructures.
+ */
+
+#ifdef INCOMPLETE_TYPES_BROKEN
+#ifndef JPEG_INTERNALS          /* will be defined in jpegint.h */
+struct jvirt_sarray_control { long dummy; };
+struct jvirt_barray_control { long dummy; };
+struct jpeg_comp_master { long dummy; };
+struct jpeg_c_main_controller { long dummy; };
+struct jpeg_c_prep_controller { long dummy; };
+struct jpeg_c_coef_controller { long dummy; };
+struct jpeg_marker_writer { long dummy; };
+struct jpeg_color_converter { long dummy; };
+struct jpeg_downsampler { long dummy; };
+struct jpeg_forward_dct { long dummy; };
+struct jpeg_entropy_encoder { long dummy; };
+struct jpeg_decomp_master { long dummy; };
+struct jpeg_d_main_controller { long dummy; };
+struct jpeg_d_coef_controller { long dummy; };
+struct jpeg_d_post_controller { long dummy; };
+struct jpeg_input_controller { long dummy; };
+struct jpeg_marker_reader { long dummy; };
+struct jpeg_entropy_decoder { long dummy; };
+struct jpeg_inverse_dct { long dummy; };
+struct jpeg_upsampler { long dummy; };
+struct jpeg_color_deconverter { long dummy; };
+struct jpeg_color_quantizer { long dummy; };
+#endif /* JPEG_INTERNALS */
+#endif /* INCOMPLETE_TYPES_BROKEN */
+
+
+/*
+ * The JPEG library modules define JPEG_INTERNALS before including this file.
+ * The internal structure declarations are read only when that is true.
+ * Applications using the library should not include jpegint.h, but may wish
+ * to include jerror.h.
+ */
+
+#ifdef JPEG_INTERNALS
+#include "jpegint.h"            /* fetch private declarations */
+#include "jerror.h"             /* fetch error codes too */
+#endif
+
+#ifdef __cplusplus
+#ifndef DONT_USE_EXTERN_C
+}
+#endif
+#endif
+
+#endif /* JPEGLIB_H */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jpegtran.1 Added

@@ -0,0 +1,362 @@
+.TH JPEGTRAN 1 "13 July 2021"
+.SH NAME
+jpegtran \- lossless transformation of JPEG files
+.SH SYNOPSIS
+.B jpegtran
+
+.I options
+
+
+.I filename
+
+.LP
+.SH DESCRIPTION
+.LP
+.B jpegtran
+performs various useful transformations of JPEG files.
+It can translate the coded representation from one variant of JPEG to another,
+for example from baseline JPEG to progressive JPEG or vice versa.  It can also
+perform some rearrangements of the image data, for example turning an image
+from landscape to portrait format by rotation.
+.PP
+For EXIF files and JPEG files containing Exif data, you may prefer to use
+.B exiftran
+instead.
+.PP
+.B jpegtran
+works by rearranging the compressed data (DCT coefficients), without
+ever fully decoding the image.  Therefore, its transformations are lossless:
+there is no image degradation at all, which would not be true if you used
+.B djpeg
+followed by
+.B cjpeg
+to accomplish the same conversion.  But by the same token,
+.B jpegtran
+cannot perform lossy operations such as changing the image quality.  However,
+while the image data is losslessly transformed, metadata can be removed.  See
+the
+.B \-copy
+option for specifics.
+.PP
+.B jpegtran
+reads the named JPEG/JFIF file, or the standard input if no file is
+named, and produces a JPEG/JFIF file on the standard output.
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-optimize
+may be written
+.B \-opt
+or
+.BR \-o .
+Upper and lower case are equivalent.
+British spellings are also accepted (e.g.,
+.BR \-optimise ),
+though for brevity these are not mentioned below.
+.PP
+To specify the coded JPEG representation used in the output file,
+.B jpegtran
+accepts a subset of the switches recognized by
+.BR cjpeg :
+.TP
+.B \-optimize
+Perform optimization of entropy encoding parameters.
+.TP
+.B \-progressive
+Create progressive JPEG file.
+.TP
+.BI \-restart " N"
+Emit a JPEG restart marker every N MCU rows, or every N MCU blocks if "B" is
+attached to the number.
+.TP
+.B \-arithmetic
+Use arithmetic coding.
+.TP
+.BI \-scans " file"
+Use the scan script given in the specified text file.
+.PP
+See
+.BR cjpeg (1)
+for more details about these switches.
+If you specify none of these switches, you get a plain baseline-JPEG output
+file.  The quality setting and so forth are determined by the input file.
+.PP
+The image can be losslessly transformed by giving one of these switches:
+.TP
+.B \-flip horizontal
+Mirror image horizontally (left-right).
+.TP
+.B \-flip vertical
+Mirror image vertically (top-bottom).
+.TP
+.B \-rotate 90
+Rotate image 90 degrees clockwise.
+.TP
+.B \-rotate 180
+Rotate image 180 degrees.
+.TP
+.B \-rotate 270
+Rotate image 270 degrees clockwise (or 90 ccw).
+.TP
+.B \-transpose
+Transpose image (across UL-to-LR axis).
+.TP
+.B \-transverse
+Transverse transpose (across UR-to-LL axis).
+.PP
+The transpose transformation has no restrictions regarding image dimensions.
+The other transformations operate rather oddly if the image dimensions are not
+a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
+transform complete blocks of DCT coefficient data in the desired way.
+.PP
+.BR jpegtran 's
+default behavior when transforming an odd-size image is designed
+to preserve exact reversibility and mathematical consistency of the
+transformation set.  As stated, transpose is able to flip the entire image
+area.  Horizontal mirroring leaves any partial iMCU column at the right edge
+untouched, but is able to flip all rows of the image.  Similarly, vertical
+mirroring leaves any partial iMCU row at the bottom edge untouched, but is
+able to flip all columns.  The other transforms can be built up as sequences
+of transpose and flip operations; for consistency, their actions on edge
+pixels are defined to be the same as the end result of the corresponding
+transpose-and-flip sequence.
+.PP
+For practical use, you may prefer to discard any untransformable edge pixels
+rather than having a strange-looking strip along the right and/or bottom edges
+of a transformed image.  To do this, add the
+.B \-trim
+switch:
+.TP
+.B \-trim
+Drop non-transformable edge blocks.
+.IP
+Obviously, a transformation with
+.B \-trim
+is not reversible, so strictly speaking
+.B jpegtran
+with this switch is not lossless.  Also, the expected mathematical
+equivalences between the transformations no longer hold.  For example,
+.B \-rot 270 -trim
+trims only the bottom edge, but
+.B \-rot 90 -trim
+followed by
+.B \-rot 180 -trim
+trims both edges.
+.TP
+.B \-perfect
+If you are only interested in perfect transformations, add the
+.B \-perfect
+switch.  This causes
+.B jpegtran
+to fail with an error if the transformation is not perfect.
+.IP
+For example, you may want to do
+.IP
+.B (jpegtran \-rot 90 -perfect
+.I foo.jpg
+.B || djpeg
+.I foo.jpg
+.B | pnmflip \-r90 | cjpeg)
+.IP
+to do a perfect rotation, if available, or an approximated one if not.
+.PP
+This version of \fBjpegtran\fR also offers a lossless crop option, which
+discards data outside of a given image region but losslessly preserves what is
+inside.  Like the rotate and flip transforms, lossless crop is restricted by
+the current JPEG format; the upper left corner of the selected region must fall
+on an iMCU boundary.  If it doesn't, then it is silently moved up and/or left
+to the nearest iMCU boundary (the lower right corner is unchanged.)  Thus, the
+output image covers at least the requested region, but it may cover more.  The
+adjustment of the region dimensions may be optionally disabled by attaching an
+'f' character ("force") to the width or height number.
+
+The image can be losslessly cropped by giving the switch:
+.TP
+.B \-crop WxH+X+Y
+Crop the image to a rectangular region of width W and height H, starting at
+point X,Y.  The lossless crop feature discards data outside of a given image
+region but losslessly preserves what is inside.  Like the rotate and flip
+transforms, lossless crop is restricted by the current JPEG format; the upper
+left corner of the selected region must fall on an iMCU boundary.  If it
+doesn't, then it is silently moved up and/or left to the nearest iMCU boundary
+(the lower right corner is unchanged.)
+.PP
+If W or H is larger than the width/height of the input image, then the output
+image is expanded in size, and the expanded region is filled in with zeros
+(neutral gray).  Attaching an 'f' character ("flatten") to the width number
+will cause each block in the expanded region to be filled in with the DC
+coefficient of the nearest block in the input image rather than grayed out.
+Attaching an 'r' character ("reflect") to the width number will cause the
+expanded region to be filled in with repeated reflections of the input image
+rather than grayed out.
+.PP
+A complementary lossless wipe option is provided to discard (gray out) data
+inside a given image region while losslessly preserving what is outside:
+.TP
+.B \-wipe WxH+X+Y
+Wipe (gray out) a rectangular region of width W and height H from the input
+image, starting at point X,Y.
+.PP
+Attaching an 'f' character ("flatten") to the width number will cause the
+region to be filled with the average of adjacent blocks rather than grayed out.
+If the wipe region and the region outside the wipe region, when adjusted to the
+nearest iMCU boundary, form two horizontally adjacent rectangles, then
+attaching an 'r' character ("reflect") to the width number will cause the wipe
+region to be filled with repeated reflections of the outside region rather than
+grayed out.
+.PP
+A lossless drop option is also provided, which allows another JPEG image to be
+inserted ("dropped") into the input image data at a given position, replacing
+the existing image data at that position:
+.TP
+.B \-drop +X+Y filename
+Drop (insert) another image at point X,Y
+.PP
+Both the input image and the drop image must have the same subsampling level.
+It is best if they also have the same quantization (quality.)  Otherwise, the
+quantization of the output image will be adapted to accommodate the higher of
+the input image quality and the drop image quality.  The trim option can be
+used with the drop option to requantize the drop image to match the input
+image.  Note that a grayscale image can be dropped into a full-color image or
+vice versa, as long as the full-color image has no vertical subsampling.  If
+the input image is grayscale and the drop image is full-color, then the
+chrominance channels from the drop image will be discarded.
+.PP
+Other not-strictly-lossless transformation switches are:
+.TP
+.B \-grayscale
+Force grayscale output.
+.IP
+This option discards the chrominance channels if the input image is YCbCr
+(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
+luminance channel is preserved exactly, so this is a better method of reducing
+to grayscale than decompression, conversion, and recompression.  This switch
+is particularly handy for fixing a monochrome picture that was mistakenly
+encoded as a color JPEG.  (In such a case, the space savings from getting rid
+of the near-empty chroma channels won't be large; but the decoding time for
+a grayscale JPEG is substantially less than that for a color JPEG.)
+.PP
+.B jpegtran
+also recognizes these switches that control what to do with "extra" markers,
+such as comment blocks:
+.TP
+.B \-copy none
+Copy no extra markers from source file.  This setting suppresses all
+comments and other metadata in the source file.
+.TP
+.B \-copy comments
+Copy only comment markers.  This setting copies comments from the source file
+but discards any other metadata.
+.TP
+.B \-copy icc
+Copy only ICC profile markers.  This setting copies the ICC profile from the
+source file but discards any other metadata.
+.TP
+.B \-copy all
+Copy all extra markers.  This setting preserves miscellaneous markers
+found in the source file, such as JFIF thumbnails, Exif data, and Photoshop
+settings.  In some files, these extra markers can be sizable.  Note that this
+option will copy thumbnails as-is; they will not be transformed.
+.PP
+The default behavior is \fB-copy comments\fR.  (Note: in IJG releases v6 and
+v6a, \fBjpegtran\fR always did the equivalent of \fB-copy none\fR.)
+.PP
+Additional switches recognized by jpegtran are:
+.TP
+.BI \-icc " file"
+Embed ICC color management profile contained in the specified file.  Note that
+this will cause \fBjpegtran\fR to ignore any APP2 markers in the input file,
+even if \fB-copy all\fR or \fB-copy icc\fR is specified.
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, an error will occur.
+.TP
+.BI \-maxscans " N"
+Abort if the input image contains more than
+.I N
+scans.  This feature demonstrates a method by which applications can guard
+against denial-of-service attacks instigated by specially-crafted malformed
+JPEG images containing numerous scans with missing image data or image data
+consisting only of "EOB runs" (a feature of progressive JPEG images that allows
+potentially hundreds of thousands of adjoining zero-value pixels to be
+represented using only a few bytes.)  Attempting to transform such malformed
+JPEG images can cause excessive CPU activity, since the decompressor must fully
+process each scan (even if the scan is corrupt) before it can proceed to the
+next scan.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.BI \-report
+Report transformation progress.
+.TP
+.BI \-strict
+Treat all warnings as fatal.  This feature also demonstrates a method by which
+applications can guard against attacks instigated by specially-crafted
+malformed JPEG images.  Enabling this option will cause the decompressor to
+abort if the input image contains incomplete or corrupt image data.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.TP
+.B \-version
+Print version information and exit.
+.SH EXAMPLES
+.LP
+This example converts a baseline JPEG file to progressive form:
+.IP
+.B jpegtran \-progressive
+.I foo.jpg
+.B >
+.I fooprog.jpg
+.PP
+This example rotates an image 90 degrees clockwise, discarding any
+unrotatable edge pixels:
+.IP
+.B jpegtran \-rot 90 -trim
+.I foo.jpg
+.B >
+.I foo90.jpg
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.PP
+This file was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo and to wordsmith certain sections.
+.SH BUGS
+The transform options can't transform odd-size images perfectly.  Use
+.B \-trim
+or
+.B \-perfect
+if you don't like the results.
+.PP
+The entire image is read into memory and then written out again, even in
+cases where this isn't really necessary.  Expect swapping on large images,
+especially when using the more complex transform options.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jpegtran.c Added

@@ -0,0 +1,739 @@
+/*
+ * jpegtran.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1995-2019, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2014, 2017, 2019-2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a command-line user interface for JPEG transcoding.
+ * It is very similar to cjpeg.c, and partly to djpeg.c, but provides
+ * lossless transcoding between different JPEG file formats.  It also
+ * provides some lossless and sort-of-lossless transformations of JPEG data.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "transupp.h"           /* Support routines for jpegtran */
+#include "jversion.h"           /* for version message */
+#include "jconfigint.h"
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char *progname;    /* program name for error messages */
+static char *icc_filename;      /* for -icc switch */
+JDIMENSION max_scans;           /* for -maxscans switch */
+static char *outfilename;       /* for -outfile switch */
+static char *dropfilename;      /* for -drop switch */
+boolean report;                 /* for -report switch */
+boolean strict;                 /* for -strict switch */
+static JCOPY_OPTION copyoption; /* -copy switch */
+static jpeg_transform_info transformoption; /* image transformation options */
+
+
+LOCAL(void)
+usage(void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s switches ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "inputfile\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -copy none     Copy no extra markers from source file\n");
+  fprintf(stderr, "  -copy comments Copy only comment markers (default)\n");
+  fprintf(stderr, "  -copy icc      Copy only ICC profile markers\n");
+  fprintf(stderr, "  -copy all      Copy all extra markers\n");
+#ifdef ENTROPY_OPT_SUPPORTED
+  fprintf(stderr, "  -optimize      Optimize Huffman table (smaller file, but slow compression)\n");
+#endif
+#ifdef C_PROGRESSIVE_SUPPORTED
+  fprintf(stderr, "  -progressive   Create progressive JPEG file\n");
+#endif
+  fprintf(stderr, "Switches for modifying the image:\n");
+#if TRANSFORMS_SUPPORTED
+  fprintf(stderr, "  -crop WxH+X+Y  Crop to a rectangular region\n");
+  fprintf(stderr, "  -drop +X+Y filename          Drop (insert) another image\n");
+  fprintf(stderr, "  -flip horizontal|vertical  Mirror image (left-right or top-bottom)\n");
+  fprintf(stderr, "  -grayscale     Reduce to grayscale (omit color data)\n");
+  fprintf(stderr, "  -perfect       Fail if there is non-transformable edge blocks\n");
+  fprintf(stderr, "  -rotate 90|180|270         Rotate image (degrees clockwise)\n");
+#endif
+#if TRANSFORMS_SUPPORTED
+  fprintf(stderr, "  -transpose     Transpose image\n");
+  fprintf(stderr, "  -transverse    Transverse transpose image\n");
+  fprintf(stderr, "  -trim          Drop non-transformable edge blocks\n");
+  fprintf(stderr, "                 with -drop: Requantize drop file to match source file\n");
+  fprintf(stderr, "  -wipe WxH+X+Y  Wipe (gray out) a rectangular region\n");
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef C_ARITH_CODING_SUPPORTED
+  fprintf(stderr, "  -arithmetic    Use arithmetic coding\n");
+#endif
+  fprintf(stderr, "  -icc FILE      Embed ICC profile contained in FILE\n");
+  fprintf(stderr, "  -restart N     Set restart interval in rows, or in blocks with B\n");
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -maxscans N    Maximum number of scans to allow in input file\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+  fprintf(stderr, "  -report        Report transformation progress\n");
+  fprintf(stderr, "  -strict        Treat all warnings as fatal\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "  -version       Print version information and exit\n");
+  fprintf(stderr, "Switches for wizards:\n");
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  fprintf(stderr, "  -scans FILE    Create multi-scan JPEG per script FILE\n");
+#endif
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(void)
+select_transform(JXFORM_CODE transform)
+/* Silly little routine to detect multiple transform options,
+ * which we can't handle.
+ */
+{
+#if TRANSFORMS_SUPPORTED
+  if (transformoption.transform == JXFORM_NONE ||
+      transformoption.transform == transform) {
+    transformoption.transform = transform;
+  } else {
+    fprintf(stderr, "%s: can only do one image transformation at a time\n",
+            progname);
+    usage();
+  }
+#else
+  fprintf(stderr, "%s: sorry, image transformation was not compiled\n",
+          progname);
+  exit(EXIT_FAILURE);
+#endif
+}
+
+
+LOCAL(int)
+parse_switches(j_compress_ptr cinfo, int argc, char **argv,
+               int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char *arg;
+  boolean simple_progressive;
+  char *scansarg = NULL;        /* saves -scans parm if any */
+
+  /* Set up default JPEG parameters. */
+  simple_progressive = FALSE;
+  icc_filename = NULL;
+  max_scans = 0;
+  outfilename = NULL;
+  report = FALSE;
+  strict = FALSE;
+  copyoption = JCOPYOPT_DEFAULT;
+  transformoption.transform = JXFORM_NONE;
+  transformoption.perfect = FALSE;
+  transformoption.trim = FALSE;
+  transformoption.force_grayscale = FALSE;
+  transformoption.crop = FALSE;
+  transformoption.slow_hflip = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argvargn;
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+        outfilename = NULL;     /* -outfile applies to just one input file */
+        continue;               /* ignore this name if previously processed */
+      }
+      break;                    /* else done parsing switches */
+    }
+    arg++;                      /* advance past switch marker character */
+
+    if (keymatch(arg, "arithmetic", 1)) {
+      /* Use arithmetic coding. */
+#ifdef C_ARITH_CODING_SUPPORTED
+      cinfo->arith_code = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, arithmetic coding not supported\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "copy", 2)) {
+      /* Select which extra markers to copy. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argvargn, "none", 1)) {
+        copyoption = JCOPYOPT_NONE;
+      } else if (keymatch(argvargn, "comments", 1)) {
+        copyoption = JCOPYOPT_COMMENTS;
+      } else if (keymatch(argvargn, "icc", 1)) {
+        copyoption = JCOPYOPT_ICC;
+      } else if (keymatch(argvargn, "all", 1)) {
+        copyoption = JCOPYOPT_ALL;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "crop", 2)) {
+      /* Perform lossless cropping. */
+#if TRANSFORMS_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (transformoption.crop /* reject multiple crop/drop/wipe requests */ ||
+          !jtransform_parse_crop_spec(&transformoption, argvargn)) {
+        fprintf(stderr, "%s: bogus -crop argument '%s'\n",
+                progname, argvargn);
+        exit(EXIT_FAILURE);
+      }
+#else
+      select_transform(JXFORM_NONE);    /* force an error */
+#endif
+
+    } else if (keymatch(arg, "drop", 2)) {
+#if TRANSFORMS_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (transformoption.crop /* reject multiple crop/drop/wipe requests */ ||
+          !jtransform_parse_crop_spec(&transformoption, argvargn) ||
+          transformoption.crop_width_set != JCROP_UNSET ||
+          transformoption.crop_height_set != JCROP_UNSET) {
+        fprintf(stderr, "%s: bogus -drop argument '%s'\n",
+                progname, argvargn);
+        exit(EXIT_FAILURE);
+      }
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      dropfilename = argvargn;
+      select_transform(JXFORM_DROP);
+#else
+      select_transform(JXFORM_NONE);    /* force an error */
+#endif
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (!printed_version) {
+        fprintf(stderr, "%s version %s (build %s)\n",
+                PACKAGE_NAME, VERSION, BUILD);
+        fprintf(stderr, "%s\n\n", JCOPYRIGHT);
+        fprintf(stderr, "Emulating The Independent JPEG Group's software, version %s\n\n",
+                JVERSION);
+        printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "version", 4)) {
+      fprintf(stderr, "%s version %s (build %s)\n",
+              PACKAGE_NAME, VERSION, BUILD);
+      exit(EXIT_SUCCESS);
+
+    } else if (keymatch(arg, "flip", 1)) {
+      /* Mirror left-right or top-bottom. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argvargn, "horizontal", 1))
+        select_transform(JXFORM_FLIP_H);
+      else if (keymatch(argvargn, "vertical", 1))
+        select_transform(JXFORM_FLIP_V);
+      else
+        usage();
+
+    } else if (keymatch(arg, "grayscale", 1) ||
+               keymatch(arg, "greyscale", 1)) {
+      /* Force to grayscale. */
+#if TRANSFORMS_SUPPORTED
+      transformoption.force_grayscale = TRUE;
+#else
+      select_transform(JXFORM_NONE);    /* force an error */
+#endif
+
+    } else if (keymatch(arg, "icc", 1)) {
+      /* Set ICC filename. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      icc_filename = argvargn;
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (ch == 'm' || ch == 'M')
+        lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "maxscans", 4)) {
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%u", &max_scans) != 1)
+        usage();
+
+    } else if (keymatch(arg, "optimize", 1) || keymatch(arg, "optimise", 1)) {
+      /* Enable entropy parm optimization. */
+#ifdef ENTROPY_OPT_SUPPORTED
+      cinfo->optimize_coding = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, entropy optimization was not compiled\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      outfilename = argvargn; /* save it away for later use */
+
+    } else if (keymatch(arg, "perfect", 2)) {
+      /* Fail if there is any partial edge MCUs that the transform can't
+       * handle. */
+      transformoption.perfect = TRUE;
+
+    } else if (keymatch(arg, "progressive", 2)) {
+      /* Select simple progressive mode. */
+#ifdef C_PROGRESSIVE_SUPPORTED
+      simple_progressive = TRUE;
+      /* We must postpone execution until num_components is known. */
+#else
+      fprintf(stderr, "%s: sorry, progressive output was not compiled\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "report", 3)) {
+      report = TRUE;
+
+    } else if (keymatch(arg, "restart", 1)) {
+      /* Restart interval in MCU rows (or in MCUs with 'b'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argvargn, "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (lval < 0 || lval > 65535L)
+        usage();
+      if (ch == 'b' || ch == 'B') {
+        cinfo->restart_interval = (unsigned int)lval;
+        cinfo->restart_in_rows = 0; /* else prior '-restart n' overrides me */
+      } else {
+        cinfo->restart_in_rows = (int)lval;
+        /* restart_interval will be computed during startup */
+      }
+
+    } else if (keymatch(arg, "rotate", 2)) {
+      /* Rotate 90, 180, or 270 degrees (measured clockwise). */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argvargn, "90", 2))
+        select_transform(JXFORM_ROT_90);
+      else if (keymatch(argvargn, "180", 3))
+        select_transform(JXFORM_ROT_180);
+      else if (keymatch(argvargn, "270", 3))
+        select_transform(JXFORM_ROT_270);
+      else
+        usage();
+
+    } else if (keymatch(arg, "scans", 1)) {
+      /* Set scan script. */
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      scansarg = argvargn;
+      /* We must postpone reading the file in case -progressive appears. */
+#else
+      fprintf(stderr, "%s: sorry, multi-scan output was not compiled\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "strict", 2)) {
+      strict = TRUE;
+
+    } else if (keymatch(arg, "transpose", 1)) {
+      /* Transpose (across UL-to-LR axis). */
+      select_transform(JXFORM_TRANSPOSE);
+
+    } else if (keymatch(arg, "transverse", 6)) {
+      /* Transverse transpose (across UR-to-LL axis). */
+      select_transform(JXFORM_TRANSVERSE);
+
+    } else if (keymatch(arg, "trim", 3)) {
+      /* Trim off any partial edge MCUs that the transform can't handle. */
+      transformoption.trim = TRUE;
+
+    } else if (keymatch(arg, "wipe", 1)) {
+#if TRANSFORMS_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (transformoption.crop /* reject multiple crop/drop/wipe requests */ ||
+          !jtransform_parse_crop_spec(&transformoption, argvargn)) {
+        fprintf(stderr, "%s: bogus -wipe argument '%s'\n",
+                progname, argvargn);
+        exit(EXIT_FAILURE);
+      }
+      select_transform(JXFORM_WIPE);
+#else
+      select_transform(JXFORM_NONE);    /* force an error */
+#endif
+
+    } else {
+      usage();                  /* bogus switch */
+    }
+  }
+
+  /* Post-switch-scanning cleanup */
+
+  if (for_real) {
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+    if (simple_progressive)     /* process -progressive; -scans can override */
+      jpeg_simple_progression(cinfo);
+#endif
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    if (scansarg != NULL)       /* process -scans if it was present */
+      if (!read_scan_script(cinfo, scansarg))
+        usage();
+#endif
+  }
+
+  return argn;                  /* return index of next arg (file name) */
+}
+
+
+METHODDEF(void)
+my_emit_message(j_common_ptr cinfo, int msg_level)
+{
+  if (msg_level < 0) {
+    /* Treat warning as fatal */
+    cinfo->err->error_exit(cinfo);
+  } else {
+    if (cinfo->err->trace_level >= msg_level)
+      cinfo->err->output_message(cinfo);
+  }
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main(int argc, char **argv)
+{
+  struct jpeg_decompress_struct srcinfo;
+#if TRANSFORMS_SUPPORTED
+  struct jpeg_decompress_struct dropinfo;
+  struct jpeg_error_mgr jdroperr;
+  FILE *drop_file;
+#endif
+  struct jpeg_compress_struct dstinfo;
+  struct jpeg_error_mgr jsrcerr, jdsterr;
+  struct cdjpeg_progress_mgr src_progress, dst_progress;
+  jvirt_barray_ptr *src_coef_arrays;
+  jvirt_barray_ptr *dst_coef_arrays;
+  int file_index;
+  /* We assume all-in-memory processing and can therefore use only a
+   * single file pointer for sequential input and output operation.
+   */
+  FILE *fp;
+  FILE *icc_file;
+  JOCTET *icc_profile = NULL;
+  long icc_len = 0;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv0;
+  if (progname == NULL || progname0 == 0)
+    progname = "jpegtran";      /* in case C library doesn't provide it */
+
+  /* Initialize the JPEG decompression object with default error handling. */
+  srcinfo.err = jpeg_std_error(&jsrcerr);
+  jpeg_create_decompress(&srcinfo);
+  /* Initialize the JPEG compression object with default error handling. */
+  dstinfo.err = jpeg_std_error(&jdsterr);
+  jpeg_create_compress(&dstinfo);
+
+  /* Scan command line to find file names.
+   * It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are mostly ignored; we will rescan the switches after
+   * opening the input file.  Also note that most of the switches affect the
+   * destination JPEG object, so we parse into that and then copy over what
+   * needs to affect the source too.
+   */
+
+  file_index = parse_switches(&dstinfo, argc, argv, 0, FALSE);
+  jsrcerr.trace_level = jdsterr.trace_level;
+  srcinfo.mem->max_memory_to_use = dstinfo.mem->max_memory_to_use;
+
+  if (strict)
+    jsrcerr.emit_message = my_emit_message;
+
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have either -outfile switch or explicit output file name */
+  if (outfilename == NULL) {
+    if (file_index != argc - 2) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+    outfilename = argvfile_index + 1;
+  } else {
+    if (file_index != argc - 1) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc - 1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((fp = fopen(argvfile_index, READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s for reading\n", progname,
+              argvfile_index);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    fp = read_stdin();
+  }
+
+  if (icc_filename != NULL) {
+    if ((icc_file = fopen(icc_filename, READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, icc_filename);
+      exit(EXIT_FAILURE);
+    }
+    if (fseek(icc_file, 0, SEEK_END) < 0 ||
+        (icc_len = ftell(icc_file)) < 1 ||
+        fseek(icc_file, 0, SEEK_SET) < 0) {
+      fprintf(stderr, "%s: can't determine size of %s\n", progname,
+              icc_filename);
+      exit(EXIT_FAILURE);
+    }
+    if ((icc_profile = (JOCTET *)malloc(icc_len)) == NULL) {
+      fprintf(stderr, "%s: can't allocate memory for ICC profile\n", progname);
+      fclose(icc_file);
+      exit(EXIT_FAILURE);
+    }
+    if (fread(icc_profile, icc_len, 1, icc_file) < 1) {
+      fprintf(stderr, "%s: can't read ICC profile from %s\n", progname,
+              icc_filename);
+      free(icc_profile);
+      fclose(icc_file);
+      exit(EXIT_FAILURE);
+    }
+    fclose(icc_file);
+    if (copyoption == JCOPYOPT_ALL)
+      copyoption = JCOPYOPT_ALL_EXCEPT_ICC;
+    if (copyoption == JCOPYOPT_ICC)
+      copyoption = JCOPYOPT_NONE;
+  }
+
+  if (report) {
+    start_progress_monitor((j_common_ptr)&dstinfo, &dst_progress);
+    dst_progress.report = report;
+  }
+  if (report || max_scans != 0) {
+    start_progress_monitor((j_common_ptr)&srcinfo, &src_progress);
+    src_progress.report = report;
+    src_progress.max_scans = max_scans;
+  }
+#if TRANSFORMS_SUPPORTED
+  /* Open the drop file. */
+  if (dropfilename != NULL) {
+    if ((drop_file = fopen(dropfilename, READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s for reading\n", progname,
+              dropfilename);
+      exit(EXIT_FAILURE);
+    }
+    dropinfo.err = jpeg_std_error(&jdroperr);
+    jpeg_create_decompress(&dropinfo);
+    jpeg_stdio_src(&dropinfo, drop_file);
+  } else {
+    drop_file = NULL;
+  }
+#endif
+
+  /* Specify data source for decompression */
+  jpeg_stdio_src(&srcinfo, fp);
+
+  /* Enable saving of extra markers that we want to copy */
+  jcopy_markers_setup(&srcinfo, copyoption);
+
+  /* Read file header */
+  (void)jpeg_read_header(&srcinfo, TRUE);
+
+#if TRANSFORMS_SUPPORTED
+  if (dropfilename != NULL) {
+    (void)jpeg_read_header(&dropinfo, TRUE);
+    transformoption.crop_width = dropinfo.image_width;
+    transformoption.crop_width_set = JCROP_POS;
+    transformoption.crop_height = dropinfo.image_height;
+    transformoption.crop_height_set = JCROP_POS;
+    transformoption.drop_ptr = &dropinfo;
+  }
+#endif
+
+  /* Any space needed by a transform option must be requested before
+   * jpeg_read_coefficients so that memory allocation will be done right.
+   */
+#if TRANSFORMS_SUPPORTED
+  /* Fail right away if -perfect is given and transformation is not perfect.
+   */
+  if (!jtransform_request_workspace(&srcinfo, &transformoption)) {
+    fprintf(stderr, "%s: transformation is not perfect\n", progname);
+    exit(EXIT_FAILURE);
+  }
+#endif
+
+  /* Read source file as DCT coefficients */
+  src_coef_arrays = jpeg_read_coefficients(&srcinfo);
+
+#if TRANSFORMS_SUPPORTED
+  if (dropfilename != NULL) {
+    transformoption.drop_coef_arrays = jpeg_read_coefficients(&dropinfo);
+  }
+#endif
+
+  /* Initialize destination compression parameters from source values */
+  jpeg_copy_critical_parameters(&srcinfo, &dstinfo);
+
+  /* Adjust destination parameters if required by transform options;
+   * also find out which set of coefficient arrays will hold the output.
+   */
+#if TRANSFORMS_SUPPORTED
+  dst_coef_arrays = jtransform_adjust_parameters(&srcinfo, &dstinfo,
+                                                 src_coef_arrays,
+                                                 &transformoption);
+#else
+  dst_coef_arrays = src_coef_arrays;
+#endif
+
+  /* Close input file, if we opened it.
+   * Note: we assume that jpeg_read_coefficients consumed all input
+   * until JPEG_REACHED_EOI, and that jpeg_finish_decompress will
+   * only consume more while (!cinfo->inputctl->eoi_reached).
+   * We cannot call jpeg_finish_decompress here since we still need the
+   * virtual arrays allocated from the source object for processing.
+   */
+  if (fp != stdin)
+    fclose(fp);
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((fp = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s for writing\n", progname,
+              outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default output file is stdout */
+    fp = write_stdout();
+  }
+
+  /* Adjust default compression parameters by re-parsing the options */
+  file_index = parse_switches(&dstinfo, argc, argv, 0, TRUE);
+
+  /* Specify data destination for compression */
+  jpeg_stdio_dest(&dstinfo, fp);
+
+  /* Start compressor (note no image data is actually written here) */
+  jpeg_write_coefficients(&dstinfo, dst_coef_arrays);
+
+  /* Copy to the output file any extra markers that we want to preserve */
+  jcopy_markers_execute(&srcinfo, &dstinfo, copyoption);
+
+  if (icc_profile != NULL)
+    jpeg_write_icc_profile(&dstinfo, icc_profile, (unsigned int)icc_len);
+
+  /* Execute image transformation, if any */
+#if TRANSFORMS_SUPPORTED
+  jtransform_execute_transformation(&srcinfo, &dstinfo, src_coef_arrays,
+                                    &transformoption);
+#endif
+
+  /* Finish compression and release memory */
+  jpeg_finish_compress(&dstinfo);
+  jpeg_destroy_compress(&dstinfo);
+#if TRANSFORMS_SUPPORTED
+  if (dropfilename != NULL) {
+    (void)jpeg_finish_decompress(&dropinfo);
+    jpeg_destroy_decompress(&dropinfo);
+  }
+#endif
+  (void)jpeg_finish_decompress(&srcinfo);
+  jpeg_destroy_decompress(&srcinfo);
+
+  /* Close output file, if we opened it */
+  if (fp != stdout)
+    fclose(fp);
+#if TRANSFORMS_SUPPORTED
+  if (drop_file != NULL)
+    fclose(drop_file);
+#endif
+
+  if (report)
+    end_progress_monitor((j_common_ptr)&dstinfo);
+  if (report || max_scans != 0)
+    end_progress_monitor((j_common_ptr)&srcinfo);
+
+  free(icc_profile);
+
+  /* All done. */
+#if TRANSFORMS_SUPPORTED
+  if (dropfilename != NULL)
+    exit(jsrcerr.num_warnings + jdroperr.num_warnings +
+         jdsterr.num_warnings ? EXIT_WARNING : EXIT_SUCCESS);
+#endif
+  exit(jsrcerr.num_warnings + jdsterr.num_warnings ?
+       EXIT_WARNING : EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jquant1.c Added

@@ -0,0 +1,856 @@
+/*
+ * jquant1.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains 1-pass color quantization (color mapping) routines.
+ * These routines provide mapping to a fixed color map using equally spaced
+ * color values.  Optional Floyd-Steinberg or ordered dithering is available.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+#ifdef QUANT_1PASS_SUPPORTED
+
+
+/*
+ * The main purpose of 1-pass quantization is to provide a fast, if not very
+ * high quality, colormapped output capability.  A 2-pass quantizer usually
+ * gives better visual quality; however, for quantized grayscale output this
+ * quantizer is perfectly adequate.  Dithering is highly recommended with this
+ * quantizer, though you can turn it off if you really want to.
+ *
+ * In 1-pass quantization the colormap must be chosen in advance of seeing the
+ * image.  We use a map consisting of all combinations of Ncolorsi color
+ * values for the i'th component.  The Ncolors values are chosen so that
+ * their product, the total number of colors, is no more than that requested.
+ * (In most cases, the product will be somewhat less.)
+ *
+ * Since the colormap is orthogonal, the representative value for each color
+ * component can be determined without considering the other components;
+ * then these indexes can be combined into a colormap index by a standard
+ * N-dimensional-array-subscript calculation.  Most of the arithmetic involved
+ * can be precalculated and stored in the lookup table colorindex.
+ * colorindexij maps pixel value j in component i to the nearest
+ * representative value (grid plane) for that component; this index is
+ * multiplied by the array stride for component i, so that the
+ * index of the colormap entry closest to a given pixel value is just
+ *    sum( colorindexcomponent-numberpixel-component-value )
+ * Aside from being fast, this scheme allows for variable spacing between
+ * representative values with no additional lookup cost.
+ *
+ * If gamma correction has been applied in color conversion, it might be wise
+ * to adjust the color grid spacing so that the representative colors are
+ * equidistant in linear space.  At this writing, gamma correction is not
+ * implemented by jdcolor, so nothing is done here.
+ */
+
+
+/* Declarations for ordered dithering.
+ *
+ * We use a standard 16x16 ordered dither array.  The basic concept of ordered
+ * dithering is described in many references, for instance Dale Schumacher's
+ * chapter II.2 of Graphics Gems II (James Arvo, ed. Academic Press, 1991).
+ * In place of Schumacher's comparisons against a "threshold" value, we add a
+ * "dither" value to the input pixel and then round the result to the nearest
+ * output value.  The dither value is equivalent to (0.5 - threshold) times
+ * the distance between output values.  For ordered dithering, we assume that
+ * the output colors are equally spaced; if not, results will probably be
+ * worse, since the dither may be too much or too little at a given point.
+ *
+ * The normal calculation would be to form pixel value + dither, range-limit
+ * this to 0..MAXJSAMPLE, and then index into the colorindex table as usual.
+ * We can skip the separate range-limiting step by extending the colorindex
+ * table in both directions.
+ */
+
+#define ODITHER_SIZE  16        /* dimension of dither matrix */
+/* NB: if ODITHER_SIZE is not a power of 2, ODITHER_MASK uses will break */
+#define ODITHER_CELLS  (ODITHER_SIZE * ODITHER_SIZE) /* # cells in matrix */
+#define ODITHER_MASK  (ODITHER_SIZE - 1) /* mask for wrapping around
+                                            counters */
+
+typedef int ODITHER_MATRIXODITHER_SIZEODITHER_SIZE;
+typedef int (*ODITHER_MATRIX_PTR)ODITHER_SIZE;
+
+static const UINT8 base_dither_matrixODITHER_SIZEODITHER_SIZE = {
+  /* Bayer's order-4 dither array.  Generated by the code given in
+   * Stephen Hawley's article "Ordered Dithering" in Graphics Gems I.
+   * The values in this array must range from 0 to ODITHER_CELLS-1.
+   */
+  {   0,192, 48,240, 12,204, 60,252,  3,195, 51,243, 15,207, 63,255 },
+  { 128, 64,176,112,140, 76,188,124,131, 67,179,115,143, 79,191,127 },
+  {  32,224, 16,208, 44,236, 28,220, 35,227, 19,211, 47,239, 31,223 },
+  { 160, 96,144, 80,172,108,156, 92,163, 99,147, 83,175,111,159, 95 },
+  {   8,200, 56,248,  4,196, 52,244, 11,203, 59,251,  7,199, 55,247 },
+  { 136, 72,184,120,132, 68,180,116,139, 75,187,123,135, 71,183,119 },
+  {  40,232, 24,216, 36,228, 20,212, 43,235, 27,219, 39,231, 23,215 },
+  { 168,104,152, 88,164,100,148, 84,171,107,155, 91,167,103,151, 87 },
+  {   2,194, 50,242, 14,206, 62,254,  1,193, 49,241, 13,205, 61,253 },
+  { 130, 66,178,114,142, 78,190,126,129, 65,177,113,141, 77,189,125 },
+  {  34,226, 18,210, 46,238, 30,222, 33,225, 17,209, 45,237, 29,221 },
+  { 162, 98,146, 82,174,110,158, 94,161, 97,145, 81,173,109,157, 93 },
+  {  10,202, 58,250,  6,198, 54,246,  9,201, 57,249,  5,197, 53,245 },
+  { 138, 74,186,122,134, 70,182,118,137, 73,185,121,133, 69,181,117 },
+  {  42,234, 26,218, 38,230, 22,214, 41,233, 25,217, 37,229, 21,213 },
+  { 170,106,154, 90,166,102,150, 86,169,105,153, 89,165,101,149, 85 }
+};
+
+
+/* Declarations for Floyd-Steinberg dithering.
+ *
+ * Errors are accumulated into the array fserrors, at a resolution of
+ * 1/16th of a pixel count.  The error at a given pixel is propagated
+ * to its not-yet-processed neighbors using the standard F-S fractions,
+ *              ...     (here)  7/16
+ *              3/16    5/16    1/16
+ * We work left-to-right on even rows, right-to-left on odd rows.
+ *
+ * We can get away with a single array (holding one row's worth of errors)
+ * by using it to store the current row's errors at pixel columns not yet
+ * processed, but the next row's errors at columns already processed.  We
+ * need only a few extra variables to hold the errors immediately around the
+ * current column.  (If we are lucky, those variables are in registers, but
+ * even if not, they're probably cheaper to access than array elements are.)
+ *
+ * The fserrors array is indexed component#position.
+ * We provide (#columns + 2) entries per component; the extra entry at each
+ * end saves us from special-casing the first and last pixels.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+typedef INT16 FSERROR;          /* 16 bits should be enough */
+typedef int LOCFSERROR;         /* use 'int' for calculation temps */
+#else
+typedef JLONG FSERROR;          /* may need more than 16 bits */
+typedef JLONG LOCFSERROR;       /* be sure calculation temps are big enough */
+#endif
+
+typedef FSERROR *FSERRPTR;      /* pointer to error array */
+
+
+/* Private subobject */
+
+#define MAX_Q_COMPS  4          /* max components I can handle */
+
+typedef struct {
+  struct jpeg_color_quantizer pub; /* public fields */
+
+  /* Initially allocated colormap is saved here */
+  JSAMPARRAY sv_colormap;       /* The color map as a 2-D pixel array */
+  int sv_actual;                /* number of entries in use */
+
+  JSAMPARRAY colorindex;        /* Precomputed mapping for speed */
+  /* colorindexij = index of color closest to pixel value j in component i,
+   * premultiplied as described above.  Since colormap indexes must fit into
+   * JSAMPLEs, the entries of this array will too.
+   */
+  boolean is_padded;            /* is the colorindex padded for odither? */
+
+  int NcolorsMAX_Q_COMPS;     /* # of values allocated to each component */
+
+  /* Variables for ordered dithering */
+  int row_index;                /* cur row's vertical index in dither matrix */
+  ODITHER_MATRIX_PTR oditherMAX_Q_COMPS; /* one dither array per component */
+
+  /* Variables for Floyd-Steinberg dithering */
+  FSERRPTR fserrorsMAX_Q_COMPS; /* accumulated errors */
+  boolean on_odd_row;           /* flag to remember which row we are on */
+} my_cquantizer;
+
+typedef my_cquantizer *my_cquantize_ptr;
+
+
+/*
+ * Policy-making subroutines for create_colormap and create_colorindex.
+ * These routines determine the colormap to be used.  The rest of the module
+ * only assumes that the colormap is orthogonal.
+ *
+ *  * select_ncolors decides how to divvy up the available colors
+ *    among the components.
+ *  * output_value defines the set of representative values for a component.
+ *  * largest_input_value defines the mapping from input values to
+ *    representative values for a component.
+ * Note that the latter two routines may impose different policies for
+ * different components, though this is not currently done.
+ */
+
+
+LOCAL(int)
+select_ncolors(j_decompress_ptr cinfo, int Ncolors)
+/* Determine allocation of desired colors to components, */
+/* and fill in Ncolors array to indicate choice. */
+/* Return value is total number of colors (product of Ncolors values). */
+{
+  int nc = cinfo->out_color_components; /* number of color components */
+  int max_colors = cinfo->desired_number_of_colors;
+  int total_colors, iroot, i, j;
+  boolean changed;
+  long temp;
+  int RGB_order3 = { RGB_GREEN, RGB_RED, RGB_BLUE };
+  RGB_order0 = rgb_greencinfo->out_color_space;
+  RGB_order1 = rgb_redcinfo->out_color_space;
+  RGB_order2 = rgb_bluecinfo->out_color_space;
+
+  /* We can allocate at least the nc'th root of max_colors per component. */
+  /* Compute floor(nc'th root of max_colors). */
+  iroot = 1;
+  do {
+    iroot++;
+    temp = iroot;               /* set temp = iroot ** nc */
+    for (i = 1; i < nc; i++)
+      temp *= iroot;
+  } while (temp <= (long)max_colors); /* repeat till iroot exceeds root */
+  iroot--;                      /* now iroot = floor(root) */
+
+  /* Must have at least 2 color values per component */
+  if (iroot < 2)
+    ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, (int)temp);
+
+  /* Initialize to iroot color values for each component */
+  total_colors = 1;
+  for (i = 0; i < nc; i++) {
+    Ncolorsi = iroot;
+    total_colors *= iroot;
+  }
+  /* We may be able to increment the count for one or more components without
+   * exceeding max_colors, though we know not all can be incremented.
+   * Sometimes, the first component can be incremented more than once!
+   * (Example: for 16 colors, we start at 2*2*2, go to 3*2*2, then 4*2*2.)
+   * In RGB colorspace, try to increment G first, then R, then B.
+   */
+  do {
+    changed = FALSE;
+    for (i = 0; i < nc; i++) {
+      j = (cinfo->out_color_space == JCS_RGB ? RGB_orderi : i);
+      /* calculate new total_colors if Ncolorsj is incremented */
+      temp = total_colors / Ncolorsj;
+      temp *= Ncolorsj + 1;   /* done in long arith to avoid oflo */
+      if (temp > (long)max_colors)
+        break;                  /* won't fit, done with this pass */
+      Ncolorsj++;             /* OK, apply the increment */
+      total_colors = (int)temp;
+      changed = TRUE;
+    }
+  } while (changed);
+
+  return total_colors;
+}
+
+
+LOCAL(int)
+output_value(j_decompress_ptr cinfo, int ci, int j, int maxj)
+/* Return j'th output value, where j will range from 0 to maxj */
+/* The output values must fall in 0..MAXJSAMPLE in increasing order */
+{
+  /* We always provide values 0 and MAXJSAMPLE for each component;
+   * any additional values are equally spaced between these limits.
+   * (Forcing the upper and lower values to the limits ensures that
+   * dithering can't produce a color outside the selected gamut.)
+   */
+  return (int)(((JLONG)j * MAXJSAMPLE + maxj / 2) / maxj);
+}
+
+
+LOCAL(int)
+largest_input_value(j_decompress_ptr cinfo, int ci, int j, int maxj)
+/* Return largest input value that should map to j'th output value */
+/* Must have largest(j=0) >= 0, and largest(j=maxj) >= MAXJSAMPLE */
+{
+  /* Breakpoints are halfway between values returned by output_value */
+  return (int)(((JLONG)(2 * j + 1) * MAXJSAMPLE + maxj) / (2 * maxj));
+}
+
+
+/*
+ * Create the colormap.
+ */
+
+LOCAL(void)
+create_colormap(j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  JSAMPARRAY colormap;          /* Created colormap */
+  int total_colors;             /* Number of distinct output colors */
+  int i, j, k, nci, blksize, blkdist, ptr, val;
+
+  /* Select number of colors for each component */
+  total_colors = select_ncolors(cinfo, cquantize->Ncolors);
+
+  /* Report selected color counts */
+  if (cinfo->out_color_components == 3)
+    TRACEMS4(cinfo, 1, JTRC_QUANT_3_NCOLORS, total_colors,
+             cquantize->Ncolors0, cquantize->Ncolors1,
+             cquantize->Ncolors2);
+  else
+    TRACEMS1(cinfo, 1, JTRC_QUANT_NCOLORS, total_colors);
+
+  /* Allocate and fill in the colormap. */
+  /* The colors are ordered in the map in standard row-major order, */
+  /* i.e. rightmost (highest-indexed) color changes most rapidly. */
+
+  colormap = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE,
+     (JDIMENSION)total_colors, (JDIMENSION)cinfo->out_color_components);
+
+  /* blksize is number of adjacent repeated entries for a component */
+  /* blkdist is distance between groups of identical entries for a component */
+  blkdist = total_colors;
+
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    /* fill in colormap entries for i'th color component */
+    nci = cquantize->Ncolorsi; /* # of distinct values for this color */
+    blksize = blkdist / nci;
+    for (j = 0; j < nci; j++) {
+      /* Compute j'th output value (out of nci) for component */
+      val = output_value(cinfo, i, j, nci - 1);
+      /* Fill in all colormap entries that have this value of this component */
+      for (ptr = j * blksize; ptr < total_colors; ptr += blkdist) {
+        /* fill in blksize entries beginning at ptr */
+        for (k = 0; k < blksize; k++)
+          colormapiptr + k = (JSAMPLE)val;
+      }
+    }
+    blkdist = blksize;          /* blksize of this color is blkdist of next */
+  }
+
+  /* Save the colormap in private storage,
+   * where it will survive color quantization mode changes.
+   */
+  cquantize->sv_colormap = colormap;
+  cquantize->sv_actual = total_colors;
+}
+
+
+/*
+ * Create the color index table.
+ */
+
+LOCAL(void)
+create_colorindex(j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  JSAMPROW indexptr;
+  int i, j, k, nci, blksize, val, pad;
+
+  /* For ordered dither, we pad the color index tables by MAXJSAMPLE in
+   * each direction (input index values can be -MAXJSAMPLE .. 2*MAXJSAMPLE).
+   * This is not necessary in the other dithering modes.  However, we
+   * flag whether it was done in case user changes dithering mode.
+   */
+  if (cinfo->dither_mode == JDITHER_ORDERED) {
+    pad = MAXJSAMPLE * 2;
+    cquantize->is_padded = TRUE;
+  } else {
+    pad = 0;
+    cquantize->is_padded = FALSE;
+  }
+
+  cquantize->colorindex = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE,
+     (JDIMENSION)(MAXJSAMPLE + 1 + pad),
+     (JDIMENSION)cinfo->out_color_components);
+
+  /* blksize is number of adjacent repeated entries for a component */
+  blksize = cquantize->sv_actual;
+
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    /* fill in colorindex entries for i'th color component */
+    nci = cquantize->Ncolorsi; /* # of distinct values for this color */
+    blksize = blksize / nci;
+
+    /* adjust colorindex pointers to provide padding at negative indexes. */
+    if (pad)
+      cquantize->colorindexi += MAXJSAMPLE;
+
+    /* in loop, val = index of current output value, */
+    /* and k = largest j that maps to current val */
+    indexptr = cquantize->colorindexi;
+    val = 0;
+    k = largest_input_value(cinfo, i, 0, nci - 1);
+    for (j = 0; j <= MAXJSAMPLE; j++) {
+      while (j > k)             /* advance val if past boundary */
+        k = largest_input_value(cinfo, i, ++val, nci - 1);
+      /* premultiply so that no multiplication needed in main processing */
+      indexptrj = (JSAMPLE)(val * blksize);
+    }
+    /* Pad at both ends if necessary */
+    if (pad)
+      for (j = 1; j <= MAXJSAMPLE; j++) {
+        indexptr-j = indexptr0;
+        indexptrMAXJSAMPLE + j = indexptrMAXJSAMPLE;
+      }
+  }
+}
+
+
+/*
+ * Create an ordered-dither array for a component having ncolors
+ * distinct output values.
+ */
+
+LOCAL(ODITHER_MATRIX_PTR)
+make_odither_array(j_decompress_ptr cinfo, int ncolors)
+{
+  ODITHER_MATRIX_PTR odither;
+  int j, k;
+  JLONG num, den;
+
+  odither = (ODITHER_MATRIX_PTR)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(ODITHER_MATRIX));
+  /* The inter-value distance for this color is MAXJSAMPLE/(ncolors-1).
+   * Hence the dither value for the matrix cell with fill order f
+   * (f=0..N-1) should be (N-1-2*f)/(2*N) * MAXJSAMPLE/(ncolors-1).
+   * On 16-bit-int machine, be careful to avoid overflow.
+   */
+  den = 2 * ODITHER_CELLS * ((JLONG)(ncolors - 1));
+  for (j = 0; j < ODITHER_SIZE; j++) {
+    for (k = 0; k < ODITHER_SIZE; k++) {
+      num = ((JLONG)(ODITHER_CELLS - 1 -
+                     2 * ((int)base_dither_matrixjk))) * MAXJSAMPLE;
+      /* Ensure round towards zero despite C's lack of consistency
+       * about rounding negative values in integer division...
+       */
+      oditherjk = (int)(num < 0 ? -((-num) / den) : num / den);
+    }
+  }
+  return odither;
+}
+
+
+/*
+ * Create the ordered-dither tables.
+ * Components having the same number of representative colors may
+ * share a dither table.
+ */
+
+LOCAL(void)
+create_odither_tables(j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  ODITHER_MATRIX_PTR odither;
+  int i, j, nci;
+
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    nci = cquantize->Ncolorsi; /* # of distinct values for this color */
+    odither = NULL;             /* search for matching prior component */
+    for (j = 0; j < i; j++) {
+      if (nci == cquantize->Ncolorsj) {
+        odither = cquantize->oditherj;
+        break;
+      }
+    }
+    if (odither == NULL)        /* need a new table? */
+      odither = make_odither_array(cinfo, nci);
+    cquantize->oditheri = odither;
+  }
+}
+
+
+/*
+ * Map some rows of pixels to the output colormapped representation.
+ */
+
+METHODDEF(void)
+color_quantize(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+               JSAMPARRAY output_buf, int num_rows)
+/* General case, no dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  JSAMPARRAY colorindex = cquantize->colorindex;
+  register int pixcode, ci;
+  register JSAMPROW ptrin, ptrout;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+  register int nc = cinfo->out_color_components;
+
+  for (row = 0; row < num_rows; row++) {
+    ptrin = input_bufrow;
+    ptrout = output_bufrow;
+    for (col = width; col > 0; col--) {
+      pixcode = 0;
+      for (ci = 0; ci < nc; ci++) {
+        pixcode += colorindexci*ptrin++;
+      }
+      *ptrout++ = (JSAMPLE)pixcode;
+    }
+  }
+}
+
+
+METHODDEF(void)
+color_quantize3(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                JSAMPARRAY output_buf, int num_rows)
+/* Fast path for out_color_components==3, no dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  register int pixcode;
+  register JSAMPROW ptrin, ptrout;
+  JSAMPROW colorindex0 = cquantize->colorindex0;
+  JSAMPROW colorindex1 = cquantize->colorindex1;
+  JSAMPROW colorindex2 = cquantize->colorindex2;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    ptrin = input_bufrow;
+    ptrout = output_bufrow;
+    for (col = width; col > 0; col--) {
+      pixcode  = colorindex0*ptrin++;
+      pixcode += colorindex1*ptrin++;
+      pixcode += colorindex2*ptrin++;
+      *ptrout++ = (JSAMPLE)pixcode;
+    }
+  }
+}
+
+
+METHODDEF(void)
+quantize_ord_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                    JSAMPARRAY output_buf, int num_rows)
+/* General case, with ordered dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  register JSAMPROW input_ptr;
+  register JSAMPROW output_ptr;
+  JSAMPROW colorindex_ci;
+  int *dither;                  /* points to active row of dither matrix */
+  int row_index, col_index;     /* current indexes into dither matrix */
+  int nc = cinfo->out_color_components;
+  int ci;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    /* Initialize output values to 0 so can process components separately */
+    jzero_far((void *)output_bufrow, (size_t)(width * sizeof(JSAMPLE)));
+    row_index = cquantize->row_index;
+    for (ci = 0; ci < nc; ci++) {
+      input_ptr = input_bufrow + ci;
+      output_ptr = output_bufrow;
+      colorindex_ci = cquantize->colorindexci;
+      dither = cquantize->odithercirow_index;
+      col_index = 0;
+
+      for (col = width; col > 0; col--) {
+        /* Form pixel value + dither, range-limit to 0..MAXJSAMPLE,
+         * select output value, accumulate into output code for this pixel.
+         * Range-limiting need not be done explicitly, as we have extended
+         * the colorindex table to produce the right answers for out-of-range
+         * inputs.  The maximum dither is +- MAXJSAMPLE; this sets the
+         * required amount of padding.
+         */
+        *output_ptr +=
+          colorindex_ci*input_ptr + dithercol_index;
+        input_ptr += nc;
+        output_ptr++;
+        col_index = (col_index + 1) & ODITHER_MASK;
+      }
+    }
+    /* Advance row index for next row */
+    row_index = (row_index + 1) & ODITHER_MASK;
+    cquantize->row_index = row_index;
+  }
+}
+
+
+METHODDEF(void)
+quantize3_ord_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                     JSAMPARRAY output_buf, int num_rows)
+/* Fast path for out_color_components==3, with ordered dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  register int pixcode;
+  register JSAMPROW input_ptr;
+  register JSAMPROW output_ptr;
+  JSAMPROW colorindex0 = cquantize->colorindex0;
+  JSAMPROW colorindex1 = cquantize->colorindex1;
+  JSAMPROW colorindex2 = cquantize->colorindex2;
+  int *dither0;                 /* points to active row of dither matrix */
+  int *dither1;
+  int *dither2;
+  int row_index, col_index;     /* current indexes into dither matrix */
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    row_index = cquantize->row_index;
+    input_ptr = input_bufrow;
+    output_ptr = output_bufrow;
+    dither0 = cquantize->odither0row_index;
+    dither1 = cquantize->odither1row_index;
+    dither2 = cquantize->odither2row_index;
+    col_index = 0;
+
+    for (col = width; col > 0; col--) {
+      pixcode  = colorindex0(*input_ptr++) + dither0col_index;
+      pixcode += colorindex1(*input_ptr++) + dither1col_index;
+      pixcode += colorindex2(*input_ptr++) + dither2col_index;
+      *output_ptr++ = (JSAMPLE)pixcode;
+      col_index = (col_index + 1) & ODITHER_MASK;
+    }
+    row_index = (row_index + 1) & ODITHER_MASK;
+    cquantize->row_index = row_index;
+  }
+}
+
+
+METHODDEF(void)
+quantize_fs_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                   JSAMPARRAY output_buf, int num_rows)
+/* General case, with Floyd-Steinberg dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  register LOCFSERROR cur;      /* current error or pixel value */
+  LOCFSERROR belowerr;          /* error for pixel below cur */
+  LOCFSERROR bpreverr;          /* error for below/prev col */
+  LOCFSERROR bnexterr;          /* error for below/next col */
+  LOCFSERROR delta;
+  register FSERRPTR errorptr;   /* => fserrors at column before current */
+  register JSAMPROW input_ptr;
+  register JSAMPROW output_ptr;
+  JSAMPROW colorindex_ci;
+  JSAMPROW colormap_ci;
+  int pixcode;
+  int nc = cinfo->out_color_components;
+  int dir;                      /* 1 for left-to-right, -1 for right-to-left */
+  int dirnc;                    /* dir * nc */
+  int ci;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+  JSAMPLE *range_limit = cinfo->sample_range_limit;
+  SHIFT_TEMPS
+
+  for (row = 0; row < num_rows; row++) {
+    /* Initialize output values to 0 so can process components separately */
+    jzero_far((void *)output_bufrow, (size_t)(width * sizeof(JSAMPLE)));
+    for (ci = 0; ci < nc; ci++) {
+      input_ptr = input_bufrow + ci;
+      output_ptr = output_bufrow;
+      if (cquantize->on_odd_row) {
+        /* work right to left in this row */
+        input_ptr += (width - 1) * nc; /* so point to rightmost pixel */
+        output_ptr += width - 1;
+        dir = -1;
+        dirnc = -nc;
+        errorptr = cquantize->fserrorsci + (width + 1); /* => entry after last column */
+      } else {
+        /* work left to right in this row */
+        dir = 1;
+        dirnc = nc;
+        errorptr = cquantize->fserrorsci; /* => entry before first column */
+      }
+      colorindex_ci = cquantize->colorindexci;
+      colormap_ci = cquantize->sv_colormapci;
+      /* Preset error values: no error propagated to first pixel from left */
+      cur = 0;
+      /* and no error propagated to row below yet */
+      belowerr = bpreverr = 0;
+
+      for (col = width; col > 0; col--) {
+        /* cur holds the error propagated from the previous pixel on the
+         * current line.  Add the error propagated from the previous line
+         * to form the complete error correction term for this pixel, and
+         * round the error term (which is expressed * 16) to an integer.
+         * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
+         * for either sign of the error value.
+         * Note: errorptr points to *previous* column's array entry.
+         */
+        cur = RIGHT_SHIFT(cur + errorptrdir + 8, 4);
+        /* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
+         * The maximum error is +- MAXJSAMPLE; this sets the required size
+         * of the range_limit array.
+         */
+        cur += *input_ptr;
+        cur = range_limitcur;
+        /* Select output value, accumulate into output code for this pixel */
+        pixcode = colorindex_cicur;
+        *output_ptr += (JSAMPLE)pixcode;
+        /* Compute actual representation error at this pixel */
+        /* Note: we can do this even though we don't have the final */
+        /* pixel code, because the colormap is orthogonal. */
+        cur -= colormap_cipixcode;
+        /* Compute error fractions to be propagated to adjacent pixels.
+         * Add these into the running sums, and simultaneously shift the
+         * next-line error sums left by 1 column.
+         */
+        bnexterr = cur;
+        delta = cur * 2;
+        cur += delta;           /* form error * 3 */
+        errorptr0 = (FSERROR)(bpreverr + cur);
+        cur += delta;           /* form error * 5 */
+        bpreverr = belowerr + cur;
+        belowerr = bnexterr;
+        cur += delta;           /* form error * 7 */
+        /* At this point cur contains the 7/16 error value to be propagated
+         * to the next pixel on the current line, and all the errors for the
+         * next line have been shifted over. We are therefore ready to move on.
+         */
+        input_ptr += dirnc;     /* advance input ptr to next column */
+        output_ptr += dir;      /* advance output ptr to next column */
+        errorptr += dir;        /* advance errorptr to current column */
+      }
+      /* Post-loop cleanup: we must unload the final error value into the
+       * final fserrors entry.  Note we need not unload belowerr because
+       * it is for the dummy column before or after the actual array.
+       */
+      errorptr0 = (FSERROR)bpreverr; /* unload prev err into array */
+    }
+    cquantize->on_odd_row = (cquantize->on_odd_row ? FALSE : TRUE);
+  }
+}
+
+
+/*
+ * Allocate workspace for Floyd-Steinberg errors.
+ */
+
+LOCAL(void)
+alloc_fs_workspace(j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  size_t arraysize;
+  int i;
+
+  arraysize = (size_t)((cinfo->output_width + 2) * sizeof(FSERROR));
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    cquantize->fserrorsi = (FSERRPTR)
+      (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE, arraysize);
+  }
+}
+
+
+/*
+ * Initialize for one-pass color quantization.
+ */
+
+METHODDEF(void)
+start_pass_1_quant(j_decompress_ptr cinfo, boolean is_pre_scan)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  size_t arraysize;
+  int i;
+
+  /* Install my colormap. */
+  cinfo->colormap = cquantize->sv_colormap;
+  cinfo->actual_number_of_colors = cquantize->sv_actual;
+
+  /* Initialize for desired dithering mode. */
+  switch (cinfo->dither_mode) {
+  case JDITHER_NONE:
+    if (cinfo->out_color_components == 3)
+      cquantize->pub.color_quantize = color_quantize3;
+    else
+      cquantize->pub.color_quantize = color_quantize;
+    break;
+  case JDITHER_ORDERED:
+    if (cinfo->out_color_components == 3)
+      cquantize->pub.color_quantize = quantize3_ord_dither;
+    else
+      cquantize->pub.color_quantize = quantize_ord_dither;
+    cquantize->row_index = 0;   /* initialize state for ordered dither */
+    /* If user changed to ordered dither from another mode,
+     * we must recreate the color index table with padding.
+     * This will cost extra space, but probably isn't very likely.
+     */
+    if (!cquantize->is_padded)
+      create_colorindex(cinfo);
+    /* Create ordered-dither tables if we didn't already. */
+    if (cquantize->odither0 == NULL)
+      create_odither_tables(cinfo);
+    break;
+  case JDITHER_FS:
+    cquantize->pub.color_quantize = quantize_fs_dither;
+    cquantize->on_odd_row = FALSE; /* initialize state for F-S dither */
+    /* Allocate Floyd-Steinberg workspace if didn't already. */
+    if (cquantize->fserrors0 == NULL)
+      alloc_fs_workspace(cinfo);
+    /* Initialize the propagated errors to zero. */
+    arraysize = (size_t)((cinfo->output_width + 2) * sizeof(FSERROR));
+    for (i = 0; i < cinfo->out_color_components; i++)
+      jzero_far((void *)cquantize->fserrorsi, arraysize);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+    break;
+  }
+}
+
+
+/*
+ * Finish up at the end of the pass.
+ */
+
+METHODDEF(void)
+finish_pass_1_quant(j_decompress_ptr cinfo)
+{
+  /* no work in 1-pass case */
+}
+
+
+/*
+ * Switch to a new external colormap between output passes.
+ * Shouldn't get to this module!
+ */
+
+METHODDEF(void)
+new_color_map_1_quant(j_decompress_ptr cinfo)
+{
+  ERREXIT(cinfo, JERR_MODE_CHANGE);
+}
+
+
+/*
+ * Module initialization routine for 1-pass color quantization.
+ */
+
+GLOBAL(void)
+jinit_1pass_quantizer(j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize;
+
+  cquantize = (my_cquantize_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_cquantizer));
+  cinfo->cquantize = (struct jpeg_color_quantizer *)cquantize;
+  cquantize->pub.start_pass = start_pass_1_quant;
+  cquantize->pub.finish_pass = finish_pass_1_quant;
+  cquantize->pub.new_color_map = new_color_map_1_quant;
+  cquantize->fserrors0 = NULL; /* Flag FS workspace not allocated */
+  cquantize->odither0 = NULL; /* Also flag odither arrays not allocated */
+
+  /* Make sure my internal arrays won't overflow */
+  if (cinfo->out_color_components > MAX_Q_COMPS)
+    ERREXIT1(cinfo, JERR_QUANT_COMPONENTS, MAX_Q_COMPS);
+  /* Make sure colormap indexes can be represented by JSAMPLEs */
+  if (cinfo->desired_number_of_colors > (MAXJSAMPLE + 1))
+    ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXJSAMPLE + 1);
+
+  /* Create the colormap and color index table. */
+  create_colormap(cinfo);
+  create_colorindex(cinfo);
+
+  /* Allocate Floyd-Steinberg workspace now if requested.
+   * We do this now since it may affect the memory manager's space
+   * calculations.  If the user changes to FS dither mode in a later pass, we
+   * will allocate the space then, and will possibly overrun the
+   * max_memory_to_use setting.
+   */
+  if (cinfo->dither_mode == JDITHER_FS)
+    alloc_fs_workspace(cinfo);
+}
+
+#endif /* QUANT_1PASS_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jquant2.c Added

@@ -0,0 +1,1285 @@
+/*
+ * jquant2.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2014-2015, 2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains 2-pass color quantization (color mapping) routines.
+ * These routines provide selection of a custom color map for an image,
+ * followed by mapping of the image to that color map, with optional
+ * Floyd-Steinberg dithering.
+ * It is also possible to use just the second pass to map to an arbitrary
+ * externally-given color map.
+ *
+ * Note: ordered dithering is not supported, since there isn't any fast
+ * way to compute intercolor distances; it's unclear that ordered dither's
+ * fundamental assumptions even hold with an irregularly spaced color map.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+#ifdef QUANT_2PASS_SUPPORTED
+
+
+/*
+ * This module implements the well-known Heckbert paradigm for color
+ * quantization.  Most of the ideas used here can be traced back to
+ * Heckbert's seminal paper
+ *   Heckbert, Paul.  "Color Image Quantization for Frame Buffer Display",
+ *   Proc. SIGGRAPH '82, Computer Graphics v.16 #3 (July 1982), pp 297-304.
+ *
+ * In the first pass over the image, we accumulate a histogram showing the
+ * usage count of each possible color.  To keep the histogram to a reasonable
+ * size, we reduce the precision of the input; typical practice is to retain
+ * 5 or 6 bits per color, so that 8 or 4 different input values are counted
+ * in the same histogram cell.
+ *
+ * Next, the color-selection step begins with a box representing the whole
+ * color space, and repeatedly splits the "largest" remaining box until we
+ * have as many boxes as desired colors.  Then the mean color in each
+ * remaining box becomes one of the possible output colors.
+ *
+ * The second pass over the image maps each input pixel to the closest output
+ * color (optionally after applying a Floyd-Steinberg dithering correction).
+ * This mapping is logically trivial, but making it go fast enough requires
+ * considerable care.
+ *
+ * Heckbert-style quantizers vary a good deal in their policies for choosing
+ * the "largest" box and deciding where to cut it.  The particular policies
+ * used here have proved out well in experimental comparisons, but better ones
+ * may yet be found.
+ *
+ * In earlier versions of the IJG code, this module quantized in YCbCr color
+ * space, processing the raw upsampled data without a color conversion step.
+ * This allowed the color conversion math to be done only once per colormap
+ * entry, not once per pixel.  However, that optimization precluded other
+ * useful optimizations (such as merging color conversion with upsampling)
+ * and it also interfered with desired capabilities such as quantizing to an
+ * externally-supplied colormap.  We have therefore abandoned that approach.
+ * The present code works in the post-conversion color space, typically RGB.
+ *
+ * To improve the visual quality of the results, we actually work in scaled
+ * RGB space, giving G distances more weight than R, and R in turn more than
+ * B.  To do everything in integer math, we must use integer scale factors.
+ * The 2/3/1 scale factors used here correspond loosely to the relative
+ * weights of the colors in the NTSC grayscale equation.
+ * If you want to use this code to quantize a non-RGB color space, you'll
+ * probably need to change these scale factors.
+ */
+
+#define R_SCALE  2              /* scale R distances by this much */
+#define G_SCALE  3              /* scale G distances by this much */
+#define B_SCALE  1              /* and B by this much */
+
+static const int c_scales3 = { R_SCALE, G_SCALE, B_SCALE };
+#define C0_SCALE  c_scalesrgb_redcinfo->out_color_space
+#define C1_SCALE  c_scalesrgb_greencinfo->out_color_space
+#define C2_SCALE  c_scalesrgb_bluecinfo->out_color_space
+
+/*
+ * First we have the histogram data structure and routines for creating it.
+ *
+ * The number of bits of precision can be adjusted by changing these symbols.
+ * We recommend keeping 6 bits for G and 5 each for R and B.
+ * If you have plenty of memory and cycles, 6 bits all around gives marginally
+ * better results; if you are short of memory, 5 bits all around will save
+ * some space but degrade the results.
+ * To maintain a fully accurate histogram, we'd need to allocate a "long"
+ * (preferably unsigned long) for each cell.  In practice this is overkill;
+ * we can get by with 16 bits per cell.  Few of the cell counts will overflow,
+ * and clamping those that do overflow to the maximum value will give close-
+ * enough results.  This reduces the recommended histogram size from 256Kb
+ * to 128Kb, which is a useful savings on PC-class machines.
+ * (In the second pass the histogram space is re-used for pixel mapping data;
+ * in that capacity, each cell must be able to store zero to the number of
+ * desired colors.  16 bits/cell is plenty for that too.)
+ * Since the JPEG code is intended to run in small memory model on 80x86
+ * machines, we can't just allocate the histogram in one chunk.  Instead
+ * of a true 3-D array, we use a row of pointers to 2-D arrays.  Each
+ * pointer corresponds to a C0 value (typically 2^5 = 32 pointers) and
+ * each 2-D array has 2^6*2^5 = 2048 or 2^6*2^6 = 4096 entries.
+ */
+
+#define MAXNUMCOLORS  (MAXJSAMPLE + 1) /* maximum size of colormap */
+
+/* These will do the right thing for either R,G,B or B,G,R color order,
+ * but you may not like the results for other color orders.
+ */
+#define HIST_C0_BITS  5         /* bits of precision in R/B histogram */
+#define HIST_C1_BITS  6         /* bits of precision in G histogram */
+#define HIST_C2_BITS  5         /* bits of precision in B/R histogram */
+
+/* Number of elements along histogram axes. */
+#define HIST_C0_ELEMS  (1 << HIST_C0_BITS)
+#define HIST_C1_ELEMS  (1 << HIST_C1_BITS)
+#define HIST_C2_ELEMS  (1 << HIST_C2_BITS)
+
+/* These are the amounts to shift an input value to get a histogram index. */
+#define C0_SHIFT  (BITS_IN_JSAMPLE - HIST_C0_BITS)
+#define C1_SHIFT  (BITS_IN_JSAMPLE - HIST_C1_BITS)
+#define C2_SHIFT  (BITS_IN_JSAMPLE - HIST_C2_BITS)
+
+
+typedef UINT16 histcell;        /* histogram cell; prefer an unsigned type */
+
+typedef histcell *histptr;      /* for pointers to histogram cells */
+
+typedef histcell hist1dHIST_C2_ELEMS; /* typedefs for the array */
+typedef hist1d *hist2d;         /* type for the 2nd-level pointers */
+typedef hist2d *hist3d;         /* type for top-level pointer */
+
+
+/* Declarations for Floyd-Steinberg dithering.
+ *
+ * Errors are accumulated into the array fserrors, at a resolution of
+ * 1/16th of a pixel count.  The error at a given pixel is propagated
+ * to its not-yet-processed neighbors using the standard F-S fractions,
+ *              ...     (here)  7/16
+ *              3/16    5/16    1/16
+ * We work left-to-right on even rows, right-to-left on odd rows.
+ *
+ * We can get away with a single array (holding one row's worth of errors)
+ * by using it to store the current row's errors at pixel columns not yet
+ * processed, but the next row's errors at columns already processed.  We
+ * need only a few extra variables to hold the errors immediately around the
+ * current column.  (If we are lucky, those variables are in registers, but
+ * even if not, they're probably cheaper to access than array elements are.)
+ *
+ * The fserrors array has (#columns + 2) entries; the extra entry at
+ * each end saves us from special-casing the first and last pixels.
+ * Each entry is three values long, one value for each color component.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+typedef INT16 FSERROR;          /* 16 bits should be enough */
+typedef int LOCFSERROR;         /* use 'int' for calculation temps */
+#else
+typedef JLONG FSERROR;          /* may need more than 16 bits */
+typedef JLONG LOCFSERROR;       /* be sure calculation temps are big enough */
+#endif
+
+typedef FSERROR *FSERRPTR;      /* pointer to error array */
+
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_color_quantizer pub; /* public fields */
+
+  /* Space for the eventually created colormap is stashed here */
+  JSAMPARRAY sv_colormap;       /* colormap allocated at init time */
+  int desired;                  /* desired # of colors = size of colormap */
+
+  /* Variables for accumulating image statistics */
+  hist3d histogram;             /* pointer to the histogram */
+
+  boolean needs_zeroed;         /* TRUE if next pass must zero histogram */
+
+  /* Variables for Floyd-Steinberg dithering */
+  FSERRPTR fserrors;            /* accumulated errors */
+  boolean on_odd_row;           /* flag to remember which row we are on */
+  int *error_limiter;           /* table for clamping the applied error */
+} my_cquantizer;
+
+typedef my_cquantizer *my_cquantize_ptr;
+
+
+/*
+ * Prescan some rows of pixels.
+ * In this module the prescan simply updates the histogram, which has been
+ * initialized to zeroes by start_pass.
+ * An output_buf parameter is required by the method signature, but no data
+ * is actually output (in fact the buffer controller is probably passing a
+ * NULL pointer).
+ */
+
+METHODDEF(void)
+prescan_quantize(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                 JSAMPARRAY output_buf, int num_rows)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  register JSAMPROW ptr;
+  register histptr histp;
+  register hist3d histogram = cquantize->histogram;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    ptr = input_bufrow;
+    for (col = width; col > 0; col--) {
+      /* get pixel value and index into the histogram */
+      histp = &histogramptr0 >> C0_SHIFT
+                        ptr1 >> C1_SHIFT
+                        ptr2 >> C2_SHIFT;
+      /* increment, check for overflow and undo increment if so. */
+      if (++(*histp) <= 0)
+        (*histp)--;
+      ptr += 3;
+    }
+  }
+}
+
+
+/*
+ * Next we have the really interesting routines: selection of a colormap
+ * given the completed histogram.
+ * These routines work with a list of "boxes", each representing a rectangular
+ * subset of the input color space (to histogram precision).
+ */
+
+typedef struct {
+  /* The bounds of the box (inclusive); expressed as histogram indexes */
+  int c0min, c0max;
+  int c1min, c1max;
+  int c2min, c2max;
+  /* The volume (actually 2-norm) of the box */
+  JLONG volume;
+  /* The number of nonzero histogram cells within this box */
+  long colorcount;
+} box;
+
+typedef box *boxptr;
+
+
+LOCAL(boxptr)
+find_biggest_color_pop(boxptr boxlist, int numboxes)
+/* Find the splittable box with the largest color population */
+/* Returns NULL if no splittable boxes remain */
+{
+  register boxptr boxp;
+  register int i;
+  register long maxc = 0;
+  boxptr which = NULL;
+
+  for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
+    if (boxp->colorcount > maxc && boxp->volume > 0) {
+      which = boxp;
+      maxc = boxp->colorcount;
+    }
+  }
+  return which;
+}
+
+
+LOCAL(boxptr)
+find_biggest_volume(boxptr boxlist, int numboxes)
+/* Find the splittable box with the largest (scaled) volume */
+/* Returns NULL if no splittable boxes remain */
+{
+  register boxptr boxp;
+  register int i;
+  register JLONG maxv = 0;
+  boxptr which = NULL;
+
+  for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
+    if (boxp->volume > maxv) {
+      which = boxp;
+      maxv = boxp->volume;
+    }
+  }
+  return which;
+}
+
+
+LOCAL(void)
+update_box(j_decompress_ptr cinfo, boxptr boxp)
+/* Shrink the min/max bounds of a box to enclose only nonzero elements, */
+/* and recompute its volume and population */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  histptr histp;
+  int c0, c1, c2;
+  int c0min, c0max, c1min, c1max, c2min, c2max;
+  JLONG dist0, dist1, dist2;
+  long ccount;
+
+  c0min = boxp->c0min;  c0max = boxp->c0max;
+  c1min = boxp->c1min;  c1max = boxp->c1max;
+  c2min = boxp->c2min;  c2max = boxp->c2max;
+
+  if (c0max > c0min)
+    for (c0 = c0min; c0 <= c0max; c0++)
+      for (c1 = c1min; c1 <= c1max; c1++) {
+        histp = &histogramc0c1c2min;
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c0min = c0min = c0;
+            goto have_c0min;
+          }
+      }
+have_c0min:
+  if (c0max > c0min)
+    for (c0 = c0max; c0 >= c0min; c0--)
+      for (c1 = c1min; c1 <= c1max; c1++) {
+        histp = &histogramc0c1c2min;
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c0max = c0max = c0;
+            goto have_c0max;
+          }
+      }
+have_c0max:
+  if (c1max > c1min)
+    for (c1 = c1min; c1 <= c1max; c1++)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+        histp = &histogramc0c1c2min;
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c1min = c1min = c1;
+            goto have_c1min;
+          }
+      }
+have_c1min:
+  if (c1max > c1min)
+    for (c1 = c1max; c1 >= c1min; c1--)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+        histp = &histogramc0c1c2min;
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c1max = c1max = c1;
+            goto have_c1max;
+          }
+      }
+have_c1max:
+  if (c2max > c2min)
+    for (c2 = c2min; c2 <= c2max; c2++)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+        histp = &histogramc0c1minc2;
+        for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+          if (*histp != 0) {
+            boxp->c2min = c2min = c2;
+            goto have_c2min;
+          }
+      }
+have_c2min:
+  if (c2max > c2min)
+    for (c2 = c2max; c2 >= c2min; c2--)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+        histp = &histogramc0c1minc2;
+        for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+          if (*histp != 0) {
+            boxp->c2max = c2max = c2;
+            goto have_c2max;
+          }
+      }
+have_c2max:
+
+  /* Update box volume.
+   * We use 2-norm rather than real volume here; this biases the method
+   * against making long narrow boxes, and it has the side benefit that
+   * a box is splittable iff norm > 0.
+   * Since the differences are expressed in histogram-cell units,
+   * we have to shift back to JSAMPLE units to get consistent distances;
+   * after which, we scale according to the selected distance scale factors.
+   */
+  dist0 = ((c0max - c0min) << C0_SHIFT) * C0_SCALE;
+  dist1 = ((c1max - c1min) << C1_SHIFT) * C1_SCALE;
+  dist2 = ((c2max - c2min) << C2_SHIFT) * C2_SCALE;
+  boxp->volume = dist0 * dist0 + dist1 * dist1 + dist2 * dist2;
+
+  /* Now scan remaining volume of box and compute population */
+  ccount = 0;
+  for (c0 = c0min; c0 <= c0max; c0++)
+    for (c1 = c1min; c1 <= c1max; c1++) {
+      histp = &histogramc0c1c2min;
+      for (c2 = c2min; c2 <= c2max; c2++, histp++)
+        if (*histp != 0) {
+          ccount++;
+        }
+    }
+  boxp->colorcount = ccount;
+}
+
+
+LOCAL(int)
+median_cut(j_decompress_ptr cinfo, boxptr boxlist, int numboxes,
+           int desired_colors)
+/* Repeatedly select and split the largest box until we have enough boxes */
+{
+  int n, lb;
+  int c0, c1, c2, cmax;
+  register boxptr b1, b2;
+
+  while (numboxes < desired_colors) {
+    /* Select box to split.
+     * Current algorithm: by population for first half, then by volume.
+     */
+    if (numboxes * 2 <= desired_colors) {
+      b1 = find_biggest_color_pop(boxlist, numboxes);
+    } else {
+      b1 = find_biggest_volume(boxlist, numboxes);
+    }
+    if (b1 == NULL)             /* no splittable boxes left! */
+      break;
+    b2 = &boxlistnumboxes;    /* where new box will go */
+    /* Copy the color bounds to the new box. */
+    b2->c0max = b1->c0max;  b2->c1max = b1->c1max;  b2->c2max = b1->c2max;
+    b2->c0min = b1->c0min;  b2->c1min = b1->c1min;  b2->c2min = b1->c2min;
+    /* Choose which axis to split the box on.
+     * Current algorithm: longest scaled axis.
+     * See notes in update_box about scaling distances.
+     */
+    c0 = ((b1->c0max - b1->c0min) << C0_SHIFT) * C0_SCALE;
+    c1 = ((b1->c1max - b1->c1min) << C1_SHIFT) * C1_SCALE;
+    c2 = ((b1->c2max - b1->c2min) << C2_SHIFT) * C2_SCALE;
+    /* We want to break any ties in favor of green, then red, blue last.
+     * This code does the right thing for R,G,B or B,G,R color orders only.
+     */
+    if (rgb_redcinfo->out_color_space == 0) {
+      cmax = c1;  n = 1;
+      if (c0 > cmax) { cmax = c0;  n = 0; }
+      if (c2 > cmax) { n = 2; }
+    } else {
+      cmax = c1;  n = 1;
+      if (c2 > cmax) { cmax = c2;  n = 2; }
+      if (c0 > cmax) { n = 0; }
+    }
+    /* Choose split point along selected axis, and update box bounds.
+     * Current algorithm: split at halfway point.
+     * (Since the box has been shrunk to minimum volume,
+     * any split will produce two nonempty subboxes.)
+     * Note that lb value is max for lower box, so must be < old max.
+     */
+    switch (n) {
+    case 0:
+      lb = (b1->c0max + b1->c0min) / 2;
+      b1->c0max = lb;
+      b2->c0min = lb + 1;
+      break;
+    case 1:
+      lb = (b1->c1max + b1->c1min) / 2;
+      b1->c1max = lb;
+      b2->c1min = lb + 1;
+      break;
+    case 2:
+      lb = (b1->c2max + b1->c2min) / 2;
+      b1->c2max = lb;
+      b2->c2min = lb + 1;
+      break;
+    }
+    /* Update stats for boxes */
+    update_box(cinfo, b1);
+    update_box(cinfo, b2);
+    numboxes++;
+  }
+  return numboxes;
+}
+
+
+LOCAL(void)
+compute_color(j_decompress_ptr cinfo, boxptr boxp, int icolor)
+/* Compute representative color for a box, put it in colormapicolor */
+{
+  /* Current algorithm: mean weighted by pixels (not colors) */
+  /* Note it is important to get the rounding correct! */
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  histptr histp;
+  int c0, c1, c2;
+  int c0min, c0max, c1min, c1max, c2min, c2max;
+  long count;
+  long total = 0;
+  long c0total = 0;
+  long c1total = 0;
+  long c2total = 0;
+
+  c0min = boxp->c0min;  c0max = boxp->c0max;
+  c1min = boxp->c1min;  c1max = boxp->c1max;
+  c2min = boxp->c2min;  c2max = boxp->c2max;
+
+  for (c0 = c0min; c0 <= c0max; c0++)
+    for (c1 = c1min; c1 <= c1max; c1++) {
+      histp = &histogramc0c1c2min;
+      for (c2 = c2min; c2 <= c2max; c2++) {
+        if ((count = *histp++) != 0) {
+          total += count;
+          c0total += ((c0 << C0_SHIFT) + ((1 << C0_SHIFT) >> 1)) * count;
+          c1total += ((c1 << C1_SHIFT) + ((1 << C1_SHIFT) >> 1)) * count;
+          c2total += ((c2 << C2_SHIFT) + ((1 << C2_SHIFT) >> 1)) * count;
+        }
+      }
+    }
+
+  cinfo->colormap0icolor = (JSAMPLE)((c0total + (total >> 1)) / total);
+  cinfo->colormap1icolor = (JSAMPLE)((c1total + (total >> 1)) / total);
+  cinfo->colormap2icolor = (JSAMPLE)((c2total + (total >> 1)) / total);
+}
+
+
+LOCAL(void)
+select_colors(j_decompress_ptr cinfo, int desired_colors)
+/* Master routine for color selection */
+{
+  boxptr boxlist;
+  int numboxes;
+  int i;
+
+  /* Allocate workspace for box list */
+  boxlist = (boxptr)(*cinfo->mem->alloc_small)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE, desired_colors * sizeof(box));
+  /* Initialize one box containing whole space */
+  numboxes = 1;
+  boxlist0.c0min = 0;
+  boxlist0.c0max = MAXJSAMPLE >> C0_SHIFT;
+  boxlist0.c1min = 0;
+  boxlist0.c1max = MAXJSAMPLE >> C1_SHIFT;
+  boxlist0.c2min = 0;
+  boxlist0.c2max = MAXJSAMPLE >> C2_SHIFT;
+  /* Shrink it to actually-used volume and set its statistics */
+  update_box(cinfo, &boxlist0);
+  /* Perform median-cut to produce final box list */
+  numboxes = median_cut(cinfo, boxlist, numboxes, desired_colors);
+  /* Compute the representative color for each box, fill colormap */
+  for (i = 0; i < numboxes; i++)
+    compute_color(cinfo, &boxlisti, i);
+  cinfo->actual_number_of_colors = numboxes;
+  TRACEMS1(cinfo, 1, JTRC_QUANT_SELECTED, numboxes);
+}
+
+
+/*
+ * These routines are concerned with the time-critical task of mapping input
+ * colors to the nearest color in the selected colormap.
+ *
+ * We re-use the histogram space as an "inverse color map", essentially a
+ * cache for the results of nearest-color searches.  All colors within a
+ * histogram cell will be mapped to the same colormap entry, namely the one
+ * closest to the cell's center.  This may not be quite the closest entry to
+ * the actual input color, but it's almost as good.  A zero in the cache
+ * indicates we haven't found the nearest color for that cell yet; the array
+ * is cleared to zeroes before starting the mapping pass.  When we find the
+ * nearest color for a cell, its colormap index plus one is recorded in the
+ * cache for future use.  The pass2 scanning routines call fill_inverse_cmap
+ * when they need to use an unfilled entry in the cache.
+ *
+ * Our method of efficiently finding nearest colors is based on the "locally
+ * sorted search" idea described by Heckbert and on the incremental distance
+ * calculation described by Spencer W. Thomas in chapter III.1 of Graphics
+ * Gems II (James Arvo, ed.  Academic Press, 1991).  Thomas points out that
+ * the distances from a given colormap entry to each cell of the histogram can
+ * be computed quickly using an incremental method: the differences between
+ * distances to adjacent cells themselves differ by a constant.  This allows a
+ * fairly fast implementation of the "brute force" approach of computing the
+ * distance from every colormap entry to every histogram cell.  Unfortunately,
+ * it needs a work array to hold the best-distance-so-far for each histogram
+ * cell (because the inner loop has to be over cells, not colormap entries).
+ * The work array elements have to be JLONGs, so the work array would need
+ * 256Kb at our recommended precision.  This is not feasible in DOS machines.
+ *
+ * To get around these problems, we apply Thomas' method to compute the
+ * nearest colors for only the cells within a small subbox of the histogram.
+ * The work array need be only as big as the subbox, so the memory usage
+ * problem is solved.  Furthermore, we need not fill subboxes that are never
+ * referenced in pass2; many images use only part of the color gamut, so a
+ * fair amount of work is saved.  An additional advantage of this
+ * approach is that we can apply Heckbert's locality criterion to quickly
+ * eliminate colormap entries that are far away from the subbox; typically
+ * three-fourths of the colormap entries are rejected by Heckbert's criterion,
+ * and we need not compute their distances to individual cells in the subbox.
+ * The speed of this approach is heavily influenced by the subbox size: too
+ * small means too much overhead, too big loses because Heckbert's criterion
+ * can't eliminate as many colormap entries.  Empirically the best subbox
+ * size seems to be about 1/512th of the histogram (1/8th in each direction).
+ *
+ * Thomas' article also describes a refined method which is asymptotically
+ * faster than the brute-force method, but it is also far more complex and
+ * cannot efficiently be applied to small subboxes.  It is therefore not
+ * useful for programs intended to be portable to DOS machines.  On machines
+ * with plenty of memory, filling the whole histogram in one shot with Thomas'
+ * refined method might be faster than the present code --- but then again,
+ * it might not be any faster, and it's certainly more complicated.
+ */
+
+
+/* log2(histogram cells in update box) for each axis; this can be adjusted */
+#define BOX_C0_LOG  (HIST_C0_BITS - 3)
+#define BOX_C1_LOG  (HIST_C1_BITS - 3)
+#define BOX_C2_LOG  (HIST_C2_BITS - 3)
+
+#define BOX_C0_ELEMS  (1 << BOX_C0_LOG) /* # of hist cells in update box */
+#define BOX_C1_ELEMS  (1 << BOX_C1_LOG)
+#define BOX_C2_ELEMS  (1 << BOX_C2_LOG)
+
+#define BOX_C0_SHIFT  (C0_SHIFT + BOX_C0_LOG)
+#define BOX_C1_SHIFT  (C1_SHIFT + BOX_C1_LOG)
+#define BOX_C2_SHIFT  (C2_SHIFT + BOX_C2_LOG)
+
+
+/*
+ * The next three routines implement inverse colormap filling.  They could
+ * all be folded into one big routine, but splitting them up this way saves
+ * some stack space (the mindist and bestdist arrays need not coexist)
+ * and may allow some compilers to produce better code by registerizing more
+ * inner-loop variables.
+ */
+
+LOCAL(int)
+find_nearby_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
+                   JSAMPLE colorlist)
+/* Locate the colormap entries close enough to an update box to be candidates
+ * for the nearest entry to some cell(s) in the update box.  The update box
+ * is specified by the center coordinates of its first cell.  The number of
+ * candidate colormap entries is returned, and their colormap indexes are
+ * placed in colorlist.
+ * This routine uses Heckbert's "locally sorted search" criterion to select
+ * the colors that need further consideration.
+ */
+{
+  int numcolors = cinfo->actual_number_of_colors;
+  int maxc0, maxc1, maxc2;
+  int centerc0, centerc1, centerc2;
+  int i, x, ncolors;
+  JLONG minmaxdist, min_dist, max_dist, tdist;
+  JLONG mindistMAXNUMCOLORS;  /* min distance to colormap entry i */
+
+  /* Compute true coordinates of update box's upper corner and center.
+   * Actually we compute the coordinates of the center of the upper-corner
+   * histogram cell, which are the upper bounds of the volume we care about.
+   * Note that since ">>" rounds down, the "center" values may be closer to
+   * min than to max; hence comparisons to them must be "<=", not "<".
+   */
+  maxc0 = minc0 + ((1 << BOX_C0_SHIFT) - (1 << C0_SHIFT));
+  centerc0 = (minc0 + maxc0) >> 1;
+  maxc1 = minc1 + ((1 << BOX_C1_SHIFT) - (1 << C1_SHIFT));
+  centerc1 = (minc1 + maxc1) >> 1;
+  maxc2 = minc2 + ((1 << BOX_C2_SHIFT) - (1 << C2_SHIFT));
+  centerc2 = (minc2 + maxc2) >> 1;
+
+  /* For each color in colormap, find:
+   *  1. its minimum squared-distance to any point in the update box
+   *     (zero if color is within update box);
+   *  2. its maximum squared-distance to any point in the update box.
+   * Both of these can be found by considering only the corners of the box.
+   * We save the minimum distance for each color in mindist;
+   * only the smallest maximum distance is of interest.
+   */
+  minmaxdist = 0x7FFFFFFFL;
+
+  for (i = 0; i < numcolors; i++) {
+    /* We compute the squared-c0-distance term, then add in the other two. */
+    x = cinfo->colormap0i;
+    if (x < minc0) {
+      tdist = (x - minc0) * C0_SCALE;
+      min_dist = tdist * tdist;
+      tdist = (x - maxc0) * C0_SCALE;
+      max_dist = tdist * tdist;
+    } else if (x > maxc0) {
+      tdist = (x - maxc0) * C0_SCALE;
+      min_dist = tdist * tdist;
+      tdist = (x - minc0) * C0_SCALE;
+      max_dist = tdist * tdist;
+    } else {
+      /* within cell range so no contribution to min_dist */
+      min_dist = 0;
+      if (x <= centerc0) {
+        tdist = (x - maxc0) * C0_SCALE;
+        max_dist = tdist * tdist;
+      } else {
+        tdist = (x - minc0) * C0_SCALE;
+        max_dist = tdist * tdist;
+      }
+    }
+
+    x = cinfo->colormap1i;
+    if (x < minc1) {
+      tdist = (x - minc1) * C1_SCALE;
+      min_dist += tdist * tdist;
+      tdist = (x - maxc1) * C1_SCALE;
+      max_dist += tdist * tdist;
+    } else if (x > maxc1) {
+      tdist = (x - maxc1) * C1_SCALE;
+      min_dist += tdist * tdist;
+      tdist = (x - minc1) * C1_SCALE;
+      max_dist += tdist * tdist;
+    } else {
+      /* within cell range so no contribution to min_dist */
+      if (x <= centerc1) {
+        tdist = (x - maxc1) * C1_SCALE;
+        max_dist += tdist * tdist;
+      } else {
+        tdist = (x - minc1) * C1_SCALE;
+        max_dist += tdist * tdist;
+      }
+    }
+
+    x = cinfo->colormap2i;
+    if (x < minc2) {
+      tdist = (x - minc2) * C2_SCALE;
+      min_dist += tdist * tdist;
+      tdist = (x - maxc2) * C2_SCALE;
+      max_dist += tdist * tdist;
+    } else if (x > maxc2) {
+      tdist = (x - maxc2) * C2_SCALE;
+      min_dist += tdist * tdist;
+      tdist = (x - minc2) * C2_SCALE;
+      max_dist += tdist * tdist;
+    } else {
+      /* within cell range so no contribution to min_dist */
+      if (x <= centerc2) {
+        tdist = (x - maxc2) * C2_SCALE;
+        max_dist += tdist * tdist;
+      } else {
+        tdist = (x - minc2) * C2_SCALE;
+        max_dist += tdist * tdist;
+      }
+    }
+
+    mindisti = min_dist;      /* save away the results */
+    if (max_dist < minmaxdist)
+      minmaxdist = max_dist;
+  }
+
+  /* Now we know that no cell in the update box is more than minmaxdist
+   * away from some colormap entry.  Therefore, only colors that are
+   * within minmaxdist of some part of the box need be considered.
+   */
+  ncolors = 0;
+  for (i = 0; i < numcolors; i++) {
+    if (mindisti <= minmaxdist)
+      colorlistncolors++ = (JSAMPLE)i;
+  }
+  return ncolors;
+}
+
+
+LOCAL(void)
+find_best_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
+                 int numcolors, JSAMPLE colorlist, JSAMPLE bestcolor)
+/* Find the closest colormap entry for each cell in the update box,
+ * given the list of candidate colors prepared by find_nearby_colors.
+ * Return the indexes of the closest entries in the bestcolor array.
+ * This routine uses Thomas' incremental distance calculation method to
+ * find the distance from a colormap entry to successive cells in the box.
+ */
+{
+  int ic0, ic1, ic2;
+  int i, icolor;
+  register JLONG *bptr;         /* pointer into bestdist array */
+  JSAMPLE *cptr;                /* pointer into bestcolor array */
+  JLONG dist0, dist1;           /* initial distance values */
+  register JLONG dist2;         /* current distance in inner loop */
+  JLONG xx0, xx1;               /* distance increments */
+  register JLONG xx2;
+  JLONG inc0, inc1, inc2;       /* initial values for increments */
+  /* This array holds the distance to the nearest-so-far color for each cell */
+  JLONG bestdistBOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS;
+
+  /* Initialize best-distance for each cell of the update box */
+  bptr = bestdist;
+  for (i = BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS - 1; i >= 0; i--)
+    *bptr++ = 0x7FFFFFFFL;
+
+  /* For each color selected by find_nearby_colors,
+   * compute its distance to the center of each cell in the box.
+   * If that's less than best-so-far, update best distance and color number.
+   */
+
+  /* Nominal steps between cell centers ("x" in Thomas article) */
+#define STEP_C0  ((1 << C0_SHIFT) * C0_SCALE)
+#define STEP_C1  ((1 << C1_SHIFT) * C1_SCALE)
+#define STEP_C2  ((1 << C2_SHIFT) * C2_SCALE)
+
+  for (i = 0; i < numcolors; i++) {
+    icolor = colorlisti;
+    /* Compute (square of) distance from minc0/c1/c2 to this color */
+    inc0 = (minc0 - cinfo->colormap0icolor) * C0_SCALE;
+    dist0 = inc0 * inc0;
+    inc1 = (minc1 - cinfo->colormap1icolor) * C1_SCALE;
+    dist0 += inc1 * inc1;
+    inc2 = (minc2 - cinfo->colormap2icolor) * C2_SCALE;
+    dist0 += inc2 * inc2;
+    /* Form the initial difference increments */
+    inc0 = inc0 * (2 * STEP_C0) + STEP_C0 * STEP_C0;
+    inc1 = inc1 * (2 * STEP_C1) + STEP_C1 * STEP_C1;
+    inc2 = inc2 * (2 * STEP_C2) + STEP_C2 * STEP_C2;
+    /* Now loop over all cells in box, updating distance per Thomas method */
+    bptr = bestdist;
+    cptr = bestcolor;
+    xx0 = inc0;
+    for (ic0 = BOX_C0_ELEMS - 1; ic0 >= 0; ic0--) {
+      dist1 = dist0;
+      xx1 = inc1;
+      for (ic1 = BOX_C1_ELEMS - 1; ic1 >= 0; ic1--) {
+        dist2 = dist1;
+        xx2 = inc2;
+        for (ic2 = BOX_C2_ELEMS - 1; ic2 >= 0; ic2--) {
+          if (dist2 < *bptr) {
+            *bptr = dist2;
+            *cptr = (JSAMPLE)icolor;
+          }
+          dist2 += xx2;
+          xx2 += 2 * STEP_C2 * STEP_C2;
+          bptr++;
+          cptr++;
+        }
+        dist1 += xx1;
+        xx1 += 2 * STEP_C1 * STEP_C1;
+      }
+      dist0 += xx0;
+      xx0 += 2 * STEP_C0 * STEP_C0;
+    }
+  }
+}
+
+
+LOCAL(void)
+fill_inverse_cmap(j_decompress_ptr cinfo, int c0, int c1, int c2)
+/* Fill the inverse-colormap entries in the update box that contains */
+/* histogram cell c0/c1/c2.  (Only that one cell MUST be filled, but */
+/* we can fill as many others as we wish.) */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  int minc0, minc1, minc2;      /* lower left corner of update box */
+  int ic0, ic1, ic2;
+  register JSAMPLE *cptr;       /* pointer into bestcolor array */
+  register histptr cachep;      /* pointer into main cache array */
+  /* This array lists the candidate colormap indexes. */
+  JSAMPLE colorlistMAXNUMCOLORS;
+  int numcolors;                /* number of candidate colors */
+  /* This array holds the actually closest colormap index for each cell. */
+  JSAMPLE bestcolorBOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS;
+
+  /* Convert cell coordinates to update box ID */
+  c0 >>= BOX_C0_LOG;
+  c1 >>= BOX_C1_LOG;
+  c2 >>= BOX_C2_LOG;
+
+  /* Compute true coordinates of update box's origin corner.
+   * Actually we compute the coordinates of the center of the corner
+   * histogram cell, which are the lower bounds of the volume we care about.
+   */
+  minc0 = (c0 << BOX_C0_SHIFT) + ((1 << C0_SHIFT) >> 1);
+  minc1 = (c1 << BOX_C1_SHIFT) + ((1 << C1_SHIFT) >> 1);
+  minc2 = (c2 << BOX_C2_SHIFT) + ((1 << C2_SHIFT) >> 1);
+
+  /* Determine which colormap entries are close enough to be candidates
+   * for the nearest entry to some cell in the update box.
+   */
+  numcolors = find_nearby_colors(cinfo, minc0, minc1, minc2, colorlist);
+
+  /* Determine the actually nearest colors. */
+  find_best_colors(cinfo, minc0, minc1, minc2, numcolors, colorlist,
+                   bestcolor);
+
+  /* Save the best color numbers (plus 1) in the main cache array */
+  c0 <<= BOX_C0_LOG;            /* convert ID back to base cell indexes */
+  c1 <<= BOX_C1_LOG;
+  c2 <<= BOX_C2_LOG;
+  cptr = bestcolor;
+  for (ic0 = 0; ic0 < BOX_C0_ELEMS; ic0++) {
+    for (ic1 = 0; ic1 < BOX_C1_ELEMS; ic1++) {
+      cachep = &histogramc0 + ic0c1 + ic1c2;
+      for (ic2 = 0; ic2 < BOX_C2_ELEMS; ic2++) {
+        *cachep++ = (histcell)((*cptr++) + 1);
+      }
+    }
+  }
+}
+
+
+/*
+ * Map some rows of pixels to the output colormapped representation.
+ */
+
+METHODDEF(void)
+pass2_no_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                JSAMPARRAY output_buf, int num_rows)
+/* This version performs no dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  register JSAMPROW inptr, outptr;
+  register histptr cachep;
+  register int c0, c1, c2;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    inptr = input_bufrow;
+    outptr = output_bufrow;
+    for (col = width; col > 0; col--) {
+      /* get pixel value and index into the cache */
+      c0 = (*inptr++) >> C0_SHIFT;
+      c1 = (*inptr++) >> C1_SHIFT;
+      c2 = (*inptr++) >> C2_SHIFT;
+      cachep = &histogramc0c1c2;
+      /* If we have not seen this color before, find nearest colormap entry */
+      /* and update the cache */
+      if (*cachep == 0)
+        fill_inverse_cmap(cinfo, c0, c1, c2);
+      /* Now emit the colormap index for this cell */
+      *outptr++ = (JSAMPLE)(*cachep - 1);
+    }
+  }
+}
+
+
+METHODDEF(void)
+pass2_fs_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                JSAMPARRAY output_buf, int num_rows)
+/* This version performs Floyd-Steinberg dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  register LOCFSERROR cur0, cur1, cur2; /* current error or pixel value */
+  LOCFSERROR belowerr0, belowerr1, belowerr2; /* error for pixel below cur */
+  LOCFSERROR bpreverr0, bpreverr1, bpreverr2; /* error for below/prev col */
+  register FSERRPTR errorptr;   /* => fserrors at column before current */
+  JSAMPROW inptr;               /* => current input pixel */
+  JSAMPROW outptr;              /* => current output pixel */
+  histptr cachep;
+  int dir;                      /* +1 or -1 depending on direction */
+  int dir3;                     /* 3*dir, for advancing inptr & errorptr */
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+  JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int *error_limit = cquantize->error_limiter;
+  JSAMPROW colormap0 = cinfo->colormap0;
+  JSAMPROW colormap1 = cinfo->colormap1;
+  JSAMPROW colormap2 = cinfo->colormap2;
+  SHIFT_TEMPS
+
+  for (row = 0; row < num_rows; row++) {
+    inptr = input_bufrow;
+    outptr = output_bufrow;
+    if (cquantize->on_odd_row) {
+      /* work right to left in this row */
+      inptr += (width - 1) * 3; /* so point to rightmost pixel */
+      outptr += width - 1;
+      dir = -1;
+      dir3 = -3;
+      errorptr = cquantize->fserrors + (width + 1) * 3; /* => entry after last column */
+      cquantize->on_odd_row = FALSE; /* flip for next time */
+    } else {
+      /* work left to right in this row */
+      dir = 1;
+      dir3 = 3;
+      errorptr = cquantize->fserrors; /* => entry before first real column */
+      cquantize->on_odd_row = TRUE; /* flip for next time */
+    }
+    /* Preset error values: no error propagated to first pixel from left */
+    cur0 = cur1 = cur2 = 0;
+    /* and no error propagated to row below yet */
+    belowerr0 = belowerr1 = belowerr2 = 0;
+    bpreverr0 = bpreverr1 = bpreverr2 = 0;
+
+    for (col = width; col > 0; col--) {
+      /* curN holds the error propagated from the previous pixel on the
+       * current line.  Add the error propagated from the previous line
+       * to form the complete error correction term for this pixel, and
+       * round the error term (which is expressed * 16) to an integer.
+       * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
+       * for either sign of the error value.
+       * Note: errorptr points to *previous* column's array entry.
+       */
+      cur0 = RIGHT_SHIFT(cur0 + errorptrdir3 + 0 + 8, 4);
+      cur1 = RIGHT_SHIFT(cur1 + errorptrdir3 + 1 + 8, 4);
+      cur2 = RIGHT_SHIFT(cur2 + errorptrdir3 + 2 + 8, 4);
+      /* Limit the error using transfer function set by init_error_limit.
+       * See comments with init_error_limit for rationale.
+       */
+      cur0 = error_limitcur0;
+      cur1 = error_limitcur1;
+      cur2 = error_limitcur2;
+      /* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
+       * The maximum error is +- MAXJSAMPLE (or less with error limiting);
+       * this sets the required size of the range_limit array.
+       */
+      cur0 += inptr0;
+      cur1 += inptr1;
+      cur2 += inptr2;
+      cur0 = range_limitcur0;
+      cur1 = range_limitcur1;
+      cur2 = range_limitcur2;
+      /* Index into the cache with adjusted pixel value */
+      cachep =
+        &histogramcur0 >> C0_SHIFTcur1 >> C1_SHIFTcur2 >> C2_SHIFT;
+      /* If we have not seen this color before, find nearest colormap */
+      /* entry and update the cache */
+      if (*cachep == 0)
+        fill_inverse_cmap(cinfo, cur0 >> C0_SHIFT, cur1 >> C1_SHIFT,
+                          cur2 >> C2_SHIFT);
+      /* Now emit the colormap index for this cell */
+      {
+        register int pixcode = *cachep - 1;
+        *outptr = (JSAMPLE)pixcode;
+        /* Compute representation error for this pixel */
+        cur0 -= colormap0pixcode;
+        cur1 -= colormap1pixcode;
+        cur2 -= colormap2pixcode;
+      }
+      /* Compute error fractions to be propagated to adjacent pixels.
+       * Add these into the running sums, and simultaneously shift the
+       * next-line error sums left by 1 column.
+       */
+      {
+        register LOCFSERROR bnexterr;
+
+        bnexterr = cur0;        /* Process component 0 */
+        errorptr0 = (FSERROR)(bpreverr0 + cur0 * 3);
+        bpreverr0 = belowerr0 + cur0 * 5;
+        belowerr0 = bnexterr;
+        cur0 *= 7;
+        bnexterr = cur1;        /* Process component 1 */
+        errorptr1 = (FSERROR)(bpreverr1 + cur1 * 3);
+        bpreverr1 = belowerr1 + cur1 * 5;
+        belowerr1 = bnexterr;
+        cur1 *= 7;
+        bnexterr = cur2;        /* Process component 2 */
+        errorptr2 = (FSERROR)(bpreverr2 + cur2 * 3);
+        bpreverr2 = belowerr2 + cur2 * 5;
+        belowerr2 = bnexterr;
+        cur2 *= 7;
+      }
+      /* At this point curN contains the 7/16 error value to be propagated
+       * to the next pixel on the current line, and all the errors for the
+       * next line have been shifted over.  We are therefore ready to move on.
+       */
+      inptr += dir3;            /* Advance pixel pointers to next column */
+      outptr += dir;
+      errorptr += dir3;         /* advance errorptr to current column */
+    }
+    /* Post-loop cleanup: we must unload the final error values into the
+     * final fserrors entry.  Note we need not unload belowerrN because
+     * it is for the dummy column before or after the actual array.
+     */
+    errorptr0 = (FSERROR)bpreverr0; /* unload prev errs into array */
+    errorptr1 = (FSERROR)bpreverr1;
+    errorptr2 = (FSERROR)bpreverr2;
+  }
+}
+
+
+/*
+ * Initialize the error-limiting transfer function (lookup table).
+ * The raw F-S error computation can potentially compute error values of up to
+ * +- MAXJSAMPLE.  But we want the maximum correction applied to a pixel to be
+ * much less, otherwise obviously wrong pixels will be created.  (Typical
+ * effects include weird fringes at color-area boundaries, isolated bright
+ * pixels in a dark area, etc.)  The standard advice for avoiding this problem
+ * is to ensure that the "corners" of the color cube are allocated as output
+ * colors; then repeated errors in the same direction cannot cause cascading
+ * error buildup.  However, that only prevents the error from getting
+ * completely out of hand; Aaron Giles reports that error limiting improves
+ * the results even with corner colors allocated.
+ * A simple clamping of the error values to about +- MAXJSAMPLE/8 works pretty
+ * well, but the smoother transfer function used below is even better.  Thanks
+ * to Aaron Giles for this idea.
+ */
+
+LOCAL(void)
+init_error_limit(j_decompress_ptr cinfo)
+/* Allocate and fill in the error_limiter table */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  int *table;
+  int in, out;
+
+  table = (int *)(*cinfo->mem->alloc_small)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE, (MAXJSAMPLE * 2 + 1) * sizeof(int));
+  table += MAXJSAMPLE;          /* so can index -MAXJSAMPLE .. +MAXJSAMPLE */
+  cquantize->error_limiter = table;
+
+#define STEPSIZE  ((MAXJSAMPLE + 1) / 16)
+  /* Map errors 1:1 up to +- MAXJSAMPLE/16 */
+  out = 0;
+  for (in = 0; in < STEPSIZE; in++, out++) {
+    tablein = out;  table-in = -out;
+  }
+  /* Map errors 1:2 up to +- 3*MAXJSAMPLE/16 */
+  for (; in < STEPSIZE * 3; in++, out += (in & 1) ? 0 : 1) {
+    tablein = out;  table-in = -out;
+  }
+  /* Clamp the rest to final out value (which is (MAXJSAMPLE+1)/8) */
+  for (; in <= MAXJSAMPLE; in++) {
+    tablein = out;  table-in = -out;
+  }
+#undef STEPSIZE
+}
+
+
+/*
+ * Finish up at the end of each pass.
+ */
+
+METHODDEF(void)
+finish_pass1(j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+
+  /* Select the representative colors and fill in cinfo->colormap */
+  cinfo->colormap = cquantize->sv_colormap;
+  select_colors(cinfo, cquantize->desired);
+  /* Force next pass to zero the color index table */
+  cquantize->needs_zeroed = TRUE;
+}
+
+
+METHODDEF(void)
+finish_pass2(j_decompress_ptr cinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * Initialize for each processing pass.
+ */
+
+METHODDEF(void)
+start_pass_2_quant(j_decompress_ptr cinfo, boolean is_pre_scan)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  int i;
+
+  /* Only F-S dithering or no dithering is supported. */
+  /* If user asks for ordered dither, give them F-S. */
+  if (cinfo->dither_mode != JDITHER_NONE)
+    cinfo->dither_mode = JDITHER_FS;
+
+  if (is_pre_scan) {
+    /* Set up method pointers */
+    cquantize->pub.color_quantize = prescan_quantize;
+    cquantize->pub.finish_pass = finish_pass1;
+    cquantize->needs_zeroed = TRUE; /* Always zero histogram */
+  } else {
+    /* Set up method pointers */
+    if (cinfo->dither_mode == JDITHER_FS)
+      cquantize->pub.color_quantize = pass2_fs_dither;
+    else
+      cquantize->pub.color_quantize = pass2_no_dither;
+    cquantize->pub.finish_pass = finish_pass2;
+
+    /* Make sure color count is acceptable */
+    i = cinfo->actual_number_of_colors;
+    if (i < 1)
+      ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, 1);
+    if (i > MAXNUMCOLORS)
+      ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXNUMCOLORS);
+
+    if (cinfo->dither_mode == JDITHER_FS) {
+      size_t arraysize =
+        (size_t)((cinfo->output_width + 2) * (3 * sizeof(FSERROR)));
+      /* Allocate Floyd-Steinberg workspace if we didn't already. */
+      if (cquantize->fserrors == NULL)
+        cquantize->fserrors = (FSERRPTR)(*cinfo->mem->alloc_large)
+          ((j_common_ptr)cinfo, JPOOL_IMAGE, arraysize);
+      /* Initialize the propagated errors to zero. */
+      jzero_far((void *)cquantize->fserrors, arraysize);
+      /* Make the error-limit table if we didn't already. */
+      if (cquantize->error_limiter == NULL)
+        init_error_limit(cinfo);
+      cquantize->on_odd_row = FALSE;
+    }
+
+  }
+  /* Zero the histogram or inverse color map, if necessary */
+  if (cquantize->needs_zeroed) {
+    for (i = 0; i < HIST_C0_ELEMS; i++) {
+      jzero_far((void *)histogrami,
+                HIST_C1_ELEMS * HIST_C2_ELEMS * sizeof(histcell));
+    }
+    cquantize->needs_zeroed = FALSE;
+  }
+}
+
+
+/*
+ * Switch to a new external colormap between output passes.
+ */
+
+METHODDEF(void)
+new_color_map_2_quant(j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr)cinfo->cquantize;
+
+  /* Reset the inverse color map */
+  cquantize->needs_zeroed = TRUE;
+}
+
+
+/*
+ * Module initialization routine for 2-pass color quantization.
+ */
+
+GLOBAL(void)
+jinit_2pass_quantizer(j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize;
+  int i;
+
+  cquantize = (my_cquantize_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(my_cquantizer));
+  cinfo->cquantize = (struct jpeg_color_quantizer *)cquantize;
+  cquantize->pub.start_pass = start_pass_2_quant;
+  cquantize->pub.new_color_map = new_color_map_2_quant;
+  cquantize->fserrors = NULL;   /* flag optional arrays not allocated */
+  cquantize->error_limiter = NULL;
+
+  /* Make sure jdmaster didn't give me a case I can't handle */
+  if (cinfo->out_color_components != 3)
+    ERREXIT(cinfo, JERR_NOTIMPL);
+
+  /* Allocate the histogram/inverse colormap storage */
+  cquantize->histogram = (hist3d)(*cinfo->mem->alloc_small)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE, HIST_C0_ELEMS * sizeof(hist2d));
+  for (i = 0; i < HIST_C0_ELEMS; i++) {
+    cquantize->histogrami = (hist2d)(*cinfo->mem->alloc_large)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE,
+       HIST_C1_ELEMS * HIST_C2_ELEMS * sizeof(histcell));
+  }
+  cquantize->needs_zeroed = TRUE; /* histogram is garbage now */
+
+  /* Allocate storage for the completed colormap, if required.
+   * We do this now since it may affect the memory manager's space
+   * calculations.
+   */
+  if (cinfo->enable_2pass_quant) {
+    /* Make sure color count is acceptable */
+    int desired = cinfo->desired_number_of_colors;
+    /* Lower bound on # of colors ... somewhat arbitrary as long as > 0 */
+    if (desired < 8)
+      ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, 8);
+    /* Make sure colormap indexes can be represented by JSAMPLEs */
+    if (desired > MAXNUMCOLORS)
+      ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXNUMCOLORS);
+    cquantize->sv_colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE, (JDIMENSION)desired, (JDIMENSION)3);
+    cquantize->desired = desired;
+  } else
+    cquantize->sv_colormap = NULL;
+
+  /* Only F-S dithering or no dithering is supported. */
+  /* If user asks for ordered dither, give them F-S. */
+  if (cinfo->dither_mode != JDITHER_NONE)
+    cinfo->dither_mode = JDITHER_FS;
+
+  /* Allocate Floyd-Steinberg workspace if necessary.
+   * This isn't really needed until pass 2, but again it may affect the memory
+   * manager's space calculations.  Although we will cope with a later change
+   * in dither_mode, we do not promise to honor max_memory_to_use if
+   * dither_mode changes.
+   */
+  if (cinfo->dither_mode == JDITHER_FS) {
+    cquantize->fserrors = (FSERRPTR)(*cinfo->mem->alloc_large)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE,
+       (size_t)((cinfo->output_width + 2) * (3 * sizeof(FSERROR))));
+    /* Might as well create the error-limiting table too. */
+    init_error_limit(cinfo);
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jsimd.h Added

@@ -0,0 +1,123 @@
+/*
+ * jsimd.h
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2011, 2014, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ * Copyright (C) 2020, Arm Limited.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ */
+
+#include "jchuff.h"             /* Declarations shared with jcphuff.c */
+
+EXTERN(int) jsimd_can_rgb_ycc(void);
+EXTERN(int) jsimd_can_rgb_gray(void);
+EXTERN(int) jsimd_can_ycc_rgb(void);
+EXTERN(int) jsimd_can_ycc_rgb565(void);
+EXTERN(int) jsimd_c_can_null_convert(void);
+
+EXTERN(void) jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                                   JSAMPIMAGE output_buf,
+                                   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                                    JSAMPIMAGE output_buf,
+                                    JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_ycc_rgb_convert(j_decompress_ptr cinfo,
+                                   JSAMPIMAGE input_buf, JDIMENSION input_row,
+                                   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo,
+                                      JSAMPIMAGE input_buf,
+                                      JDIMENSION input_row,
+                                      JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_c_null_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                                  JSAMPIMAGE output_buf, JDIMENSION output_row,
+                                  int num_rows);
+
+EXTERN(int) jsimd_can_h2v2_downsample(void);
+EXTERN(int) jsimd_can_h2v1_downsample(void);
+
+EXTERN(void) jsimd_h2v2_downsample(j_compress_ptr cinfo,
+                                   jpeg_component_info *compptr,
+                                   JSAMPARRAY input_data,
+                                   JSAMPARRAY output_data);
+
+EXTERN(int) jsimd_can_h2v2_smooth_downsample(void);
+
+EXTERN(void) jsimd_h2v2_smooth_downsample(j_compress_ptr cinfo,
+                                          jpeg_component_info *compptr,
+                                          JSAMPARRAY input_data,
+                                          JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample(j_compress_ptr cinfo,
+                                   jpeg_component_info *compptr,
+                                   JSAMPARRAY input_data,
+                                   JSAMPARRAY output_data);
+
+EXTERN(int) jsimd_can_h2v2_upsample(void);
+EXTERN(int) jsimd_can_h2v1_upsample(void);
+EXTERN(int) jsimd_can_int_upsample(void);
+
+EXTERN(void) jsimd_h2v2_upsample(j_decompress_ptr cinfo,
+                                 jpeg_component_info *compptr,
+                                 JSAMPARRAY input_data,
+                                 JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v1_upsample(j_decompress_ptr cinfo,
+                                 jpeg_component_info *compptr,
+                                 JSAMPARRAY input_data,
+                                 JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_int_upsample(j_decompress_ptr cinfo,
+                                jpeg_component_info *compptr,
+                                JSAMPARRAY input_data,
+                                JSAMPARRAY *output_data_ptr);
+
+EXTERN(int) jsimd_can_h2v2_fancy_upsample(void);
+EXTERN(int) jsimd_can_h2v1_fancy_upsample(void);
+EXTERN(int) jsimd_can_h1v2_fancy_upsample(void);
+
+EXTERN(void) jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo,
+                                       jpeg_component_info *compptr,
+                                       JSAMPARRAY input_data,
+                                       JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo,
+                                       jpeg_component_info *compptr,
+                                       JSAMPARRAY input_data,
+                                       JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h1v2_fancy_upsample(j_decompress_ptr cinfo,
+                                       jpeg_component_info *compptr,
+                                       JSAMPARRAY input_data,
+                                       JSAMPARRAY *output_data_ptr);
+
+EXTERN(int) jsimd_can_h2v2_merged_upsample(void);
+EXTERN(int) jsimd_can_h2v1_merged_upsample(void);
+
+EXTERN(void) jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo,
+                                        JSAMPIMAGE input_buf,
+                                        JDIMENSION in_row_group_ctr,
+                                        JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo,
+                                        JSAMPIMAGE input_buf,
+                                        JDIMENSION in_row_group_ctr,
+                                        JSAMPARRAY output_buf);
+
+EXTERN(int) jsimd_can_huff_encode_one_block(void);
+
+EXTERN(JOCTET *) jsimd_huff_encode_one_block(void *state, JOCTET *buffer,
+                                             JCOEFPTR block, int last_dc_val,
+                                             c_derived_tbl *dctbl,
+                                             c_derived_tbl *actbl);
+
+EXTERN(int) jsimd_can_encode_mcu_AC_first_prepare(void);
+
+EXTERN(void) jsimd_encode_mcu_AC_first_prepare
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *values, size_t *zerobits);
+
+EXTERN(int) jsimd_can_encode_mcu_AC_refine_prepare(void);
+
+EXTERN(int) jsimd_encode_mcu_AC_refine_prepare
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *absvalues, size_t *bits);

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jsimd_none.c Added

@@ -0,0 +1,431 @@
+/*
+ * jsimd_none.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2009-2011, 2014, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ * Copyright (C) 2020, Arm Limited.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains stubs for when there is no SIMD support available.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jsimd.h"
+#include "jdct.h"
+#include "jsimddct.h"
+
+GLOBAL(int)
+jsimd_can_rgb_ycc(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_c_can_null_convert(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                      JSAMPIMAGE output_buf, JDIMENSION output_row,
+                      int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                       JSAMPIMAGE output_buf, JDIMENSION output_row,
+                       int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION input_row, JSAMPARRAY output_buf,
+                      int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_c_null_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                     JSAMPIMAGE output_buf, JDIMENSION output_row,
+                     int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_smooth_downsample(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v2_smooth_downsample(j_compress_ptr cinfo,
+                             jpeg_component_info *compptr,
+                             JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_int_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_int_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                   JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h1v2_fancy_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h1v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(int)
+jsimd_can_convsamp(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
+               DCTELEM *workspace)
+{
+}
+
+GLOBAL(void)
+jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow(DCTELEM *data)
+{
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast(DCTELEM *data)
+{
+}
+
+GLOBAL(void)
+jsimd_fdct_float(FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
+{
+}
+
+GLOBAL(void)
+jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_6x6(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_12x12(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_6x6(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_12x12(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block(void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET *)
+jsimd_huff_encode_one_block(void *state, JOCTET *buffer, JCOEFPTR block,
+                            int last_dc_val, c_derived_tbl *dctbl,
+                            c_derived_tbl *actbl)
+{
+  return NULL;
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_first_prepare(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_encode_mcu_AC_first_prepare(const JCOEF *block,
+                                  const int *jpeg_natural_order_start, int Sl,
+                                  int Al, JCOEF *values, size_t *zerobits)
+{
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_refine_prepare(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block,
+                                   const int *jpeg_natural_order_start, int Sl,
+                                   int Al, JCOEF *absvalues, size_t *bits)
+{
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jsimddct.h Added

@@ -0,0 +1,70 @@
+/*
+ * jsimddct.h
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ */
+
+EXTERN(int) jsimd_can_convsamp(void);
+EXTERN(int) jsimd_can_convsamp_float(void);
+
+EXTERN(void) jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
+                            DCTELEM *workspace);
+EXTERN(void) jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+                                  FAST_FLOAT *workspace);
+
+EXTERN(int) jsimd_can_fdct_islow(void);
+EXTERN(int) jsimd_can_fdct_ifast(void);
+EXTERN(int) jsimd_can_fdct_float(void);
+
+EXTERN(void) jsimd_fdct_islow(DCTELEM *data);
+EXTERN(void) jsimd_fdct_ifast(DCTELEM *data);
+EXTERN(void) jsimd_fdct_float(FAST_FLOAT *data);
+
+EXTERN(int) jsimd_can_quantize(void);
+EXTERN(int) jsimd_can_quantize_float(void);
+
+EXTERN(void) jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors,
+                            DCTELEM *workspace);
+EXTERN(void) jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                                  FAST_FLOAT *workspace);
+
+EXTERN(int) jsimd_can_idct_2x2(void);
+EXTERN(int) jsimd_can_idct_4x4(void);
+EXTERN(int) jsimd_can_idct_6x6(void);
+EXTERN(int) jsimd_can_idct_12x12(void);
+
+EXTERN(void) jsimd_idct_2x2(j_decompress_ptr cinfo,
+                            jpeg_component_info *compptr, JCOEFPTR coef_block,
+                            JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4(j_decompress_ptr cinfo,
+                            jpeg_component_info *compptr, JCOEFPTR coef_block,
+                            JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jsimd_idct_6x6(j_decompress_ptr cinfo,
+                            jpeg_component_info *compptr, JCOEFPTR coef_block,
+                            JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jsimd_idct_12x12(j_decompress_ptr cinfo,
+                              jpeg_component_info *compptr,
+                              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                              JDIMENSION output_col);
+
+EXTERN(int) jsimd_can_idct_islow(void);
+EXTERN(int) jsimd_can_idct_ifast(void);
+EXTERN(int) jsimd_can_idct_float(void);
+
+EXTERN(void) jsimd_idct_islow(j_decompress_ptr cinfo,
+                              jpeg_component_info *compptr,
+                              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                              JDIMENSION output_col);
+EXTERN(void) jsimd_idct_ifast(j_decompress_ptr cinfo,
+                              jpeg_component_info *compptr,
+                              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                              JDIMENSION output_col);
+EXTERN(void) jsimd_idct_float(j_decompress_ptr cinfo,
+                              jpeg_component_info *compptr,
+                              JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                              JDIMENSION output_col);

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jstdhuff.c Added

@@ -0,0 +1,143 @@
+/*
+ * jstdhuff.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to set the default Huffman tables, if they are
+ * not already set.
+ */
+
+/*
+ * Huffman table setup routines
+ */
+
+LOCAL(void)
+add_huff_table(j_common_ptr cinfo, JHUFF_TBL **htblptr, const UINT8 *bits,
+               const UINT8 *val)
+/* Define a Huffman table */
+{
+  int nsymbols, len;
+
+  if (*htblptr == NULL)
+    *htblptr = jpeg_alloc_huff_table(cinfo);
+  else
+    return;
+
+  /* Copy the number-of-symbols-of-each-code-length counts */
+  MEMCOPY((*htblptr)->bits, bits, sizeof((*htblptr)->bits));
+
+  /* Validate the counts.  We do this here mainly so we can copy the right
+   * number of symbols from the val array, without risking marching off
+   * the end of memory.  jchuff.c will do a more thorough test later.
+   */
+  nsymbols = 0;
+  for (len = 1; len <= 16; len++)
+    nsymbols += bitslen;
+  if (nsymbols < 1 || nsymbols > 256)
+    ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+
+  MEMCOPY((*htblptr)->huffval, val, nsymbols * sizeof(UINT8));
+  MEMZERO(&((*htblptr)->huffvalnsymbols), (256 - nsymbols) * sizeof(UINT8));
+
+  /* Initialize sent_table FALSE so table will be written to JPEG file. */
+  (*htblptr)->sent_table = FALSE;
+}
+
+
+LOCAL(void)
+std_huff_tables(j_common_ptr cinfo)
+/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */
+/* IMPORTANT: these are only valid for 8-bit data precision! */
+{
+  JHUFF_TBL **dc_huff_tbl_ptrs, **ac_huff_tbl_ptrs;
+
+  static const UINT8 bits_dc_luminance17 = {
+    /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0
+  };
+  static const UINT8 val_dc_luminance = {
+    0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11
+  };
+
+  static const UINT8 bits_dc_chrominance17 = {
+    /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0
+  };
+  static const UINT8 val_dc_chrominance = {
+    0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11
+  };
+
+  static const UINT8 bits_ac_luminance17 = {
+    /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d
+  };
+  static const UINT8 val_ac_luminance = {
+    0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
+    0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
+    0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
+    0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
+    0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
+    0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
+    0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
+    0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
+    0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
+    0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
+    0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
+    0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
+    0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
+    0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
+    0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
+    0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
+    0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
+    0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
+    0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
+    0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+    0xf9, 0xfa
+  };
+
+  static const UINT8 bits_ac_chrominance17 = {
+    /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77
+  };
+  static const UINT8 val_ac_chrominance = {
+    0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
+    0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
+    0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
+    0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
+    0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
+    0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
+    0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
+    0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
+    0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+    0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
+    0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
+    0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
+    0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
+    0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
+    0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
+    0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
+    0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
+    0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
+    0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
+    0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+    0xf9, 0xfa
+  };
+
+  if (cinfo->is_decompressor) {
+    dc_huff_tbl_ptrs = ((j_decompress_ptr)cinfo)->dc_huff_tbl_ptrs;
+    ac_huff_tbl_ptrs = ((j_decompress_ptr)cinfo)->ac_huff_tbl_ptrs;
+  } else {
+    dc_huff_tbl_ptrs = ((j_compress_ptr)cinfo)->dc_huff_tbl_ptrs;
+    ac_huff_tbl_ptrs = ((j_compress_ptr)cinfo)->ac_huff_tbl_ptrs;
+  }
+
+  add_huff_table(cinfo, &dc_huff_tbl_ptrs0, bits_dc_luminance,
+                 val_dc_luminance);
+  add_huff_table(cinfo, &ac_huff_tbl_ptrs0, bits_ac_luminance,
+                 val_ac_luminance);
+  add_huff_table(cinfo, &dc_huff_tbl_ptrs1, bits_dc_chrominance,
+                 val_dc_chrominance);
+  add_huff_table(cinfo, &ac_huff_tbl_ptrs1, bits_ac_chrominance,
+                 val_ac_chrominance);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jutils.c Added

@@ -0,0 +1,133 @@
+/*
+ * jutils.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code
+ * relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains tables and miscellaneous utility routines needed
+ * for both compression and decompression.
+ * Note we prefix all global names with "j" to minimize conflicts with
+ * a surrounding application.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * jpeg_zigzag_orderi is the zigzag-order position of the i'th element
+ * of a DCT block read in natural order (left to right, top to bottom).
+ */
+
+#if 0                           /* This table is not actually needed in v6a */
+
+const int jpeg_zigzag_orderDCTSIZE2 = {
+   0,  1,  5,  6, 14, 15, 27, 28,
+   2,  4,  7, 13, 16, 26, 29, 42,
+   3,  8, 12, 17, 25, 30, 41, 43,
+   9, 11, 18, 24, 31, 40, 44, 53,
+  10, 19, 23, 32, 39, 45, 52, 54,
+  20, 22, 33, 38, 46, 51, 55, 60,
+  21, 34, 37, 47, 50, 56, 59, 61,
+  35, 36, 48, 49, 57, 58, 62, 63
+};
+
+#endif
+
+/*
+ * jpeg_natural_orderi is the natural-order position of the i'th element
+ * of zigzag order.
+ *
+ * When reading corrupted data, the Huffman decoders could attempt
+ * to reference an entry beyond the end of this array (if the decoded
+ * zero run length reaches past the end of the block).  To prevent
+ * wild stores without adding an inner-loop test, we put some extra
+ * "63"s after the real entries.  This will cause the extra coefficient
+ * to be stored in location 63 of the block, not somewhere random.
+ * The worst case would be a run-length of 15, which means we need 16
+ * fake entries.
+ */
+
+const int jpeg_natural_orderDCTSIZE2 + 16 = {
+  0,  1,  8, 16,  9,  2,  3, 10,
+ 17, 24, 32, 25, 18, 11,  4,  5,
+ 12, 19, 26, 33, 40, 48, 41, 34,
+ 27, 20, 13,  6,  7, 14, 21, 28,
+ 35, 42, 49, 56, 57, 50, 43, 36,
+ 29, 22, 15, 23, 30, 37, 44, 51,
+ 58, 59, 52, 45, 38, 31, 39, 46,
+ 53, 60, 61, 54, 47, 55, 62, 63,
+ 63, 63, 63, 63, 63, 63, 63, 63, /* extra entries for safety in decoder */
+ 63, 63, 63, 63, 63, 63, 63, 63
+};
+
+
+/*
+ * Arithmetic utilities
+ */
+
+GLOBAL(long)
+jdiv_round_up(long a, long b)
+/* Compute a/b rounded up to next integer, ie, ceil(a/b) */
+/* Assumes a >= 0, b > 0 */
+{
+  return (a + b - 1L) / b;
+}
+
+
+GLOBAL(long)
+jround_up(long a, long b)
+/* Compute a rounded up to next multiple of b, ie, ceil(a/b)*b */
+/* Assumes a >= 0, b > 0 */
+{
+  a += b - 1L;
+  return a - (a % b);
+}
+
+
+GLOBAL(void)
+jcopy_sample_rows(JSAMPARRAY input_array, int source_row,
+                  JSAMPARRAY output_array, int dest_row, int num_rows,
+                  JDIMENSION num_cols)
+/* Copy some rows of samples from one place to another.
+ * num_rows rows are copied from input_arraysource_row++
+ * to output_arraydest_row++; these areas may overlap for duplication.
+ * The source and destination arrays must be at least as wide as num_cols.
+ */
+{
+  register JSAMPROW inptr, outptr;
+  register size_t count = (size_t)(num_cols * sizeof(JSAMPLE));
+  register int row;
+
+  input_array += source_row;
+  output_array += dest_row;
+
+  for (row = num_rows; row > 0; row--) {
+    inptr = *input_array++;
+    outptr = *output_array++;
+    MEMCOPY(outptr, inptr, count);
+  }
+}
+
+
+GLOBAL(void)
+jcopy_block_row(JBLOCKROW input_row, JBLOCKROW output_row,
+                JDIMENSION num_blocks)
+/* Copy a row of coefficient blocks from one place to another. */
+{
+  MEMCOPY(output_row, input_row, num_blocks * (DCTSIZE2 * sizeof(JCOEF)));
+}
+
+
+GLOBAL(void)
+jzero_far(void *target, size_t bytestozero)
+/* Zero out a chunk of memory. */
+/* This might be sample-array data, block-array data, or alloc_large data. */
+{
+  MEMZERO(target, bytestozero);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/jversion.h Added

@@ -0,0 +1,54 @@
+/*
+ * jversion.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2012-2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains software version identification.
+ */
+
+
+#if JPEG_LIB_VERSION >= 80
+
+#define JVERSION        "8d  15-Jan-2012"
+
+#elif JPEG_LIB_VERSION >= 70
+
+#define JVERSION        "7  27-Jun-2009"
+
+#else
+
+#define JVERSION        "6b  27-Mar-1998"
+
+#endif
+
+/*
+ * NOTE: It is our convention to place the authors in the following order:
+ * - libjpeg-turbo authors (2009-) in descending order of the date of their
+ *   most recent contribution to the project, then in ascending order of the
+ *   date of their first contribution to the project, then in alphabetical
+ *   order
+ * - Upstream authors in descending order of the date of the first inclusion of
+ *   their code
+ */
+
+#define JCOPYRIGHT \
+  "Copyright (C) 2009-2021 D. R. Commander\n" \
+  "Copyright (C) 2015, 2020 Google, Inc.\n" \
+  "Copyright (C) 2019-2020 Arm Limited\n" \
+  "Copyright (C) 2015-2016, 2018 Matthieu Darbois\n" \
+  "Copyright (C) 2011-2016 Siarhei Siamashka\n" \
+  "Copyright (C) 2015 Intel Corporation\n" \
+  "Copyright (C) 2013-2014 Linaro Limited\n" \
+  "Copyright (C) 2013-2014 MIPS Technologies, Inc.\n" \
+  "Copyright (C) 2009, 2012 Pierre Ossman for Cendio AB\n" \
+  "Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies)\n" \
+  "Copyright (C) 1999-2006 MIYASAKA Masaru\n" \
+  "Copyright (C) 1991-2020 Thomas G. Lane, Guido Vollbeding"
+
+#define JCOPYRIGHT_SHORT \
+  "Copyright (C) 1991-2021 The libjpeg-turbo Project and many others"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/libjpeg.map.in Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/libjpeg.txt Added

@@ -0,0 +1,3162 @@
+USING THE IJG JPEG LIBRARY
+
+This file was part of the Independent JPEG Group's software:
+Copyright (C) 1994-2013, Thomas G. Lane, Guido Vollbeding.
+libjpeg-turbo Modifications:
+Copyright (C) 2010, 2014-2018, 2020, D. R. Commander.
+Copyright (C) 2015, Google, Inc.
+For conditions of distribution and use, see the accompanying README.ijg file.
+
+
+This file describes how to use the IJG JPEG library within an application
+program.  Read it if you want to write a program that uses the library.
+
+The file example.txt provides heavily commented skeleton code for calling the
+JPEG library.  Also see jpeglib.h (the include file to be used by application
+programs) for full details about data structures and function parameter lists.
+The library source code, of course, is the ultimate reference.
+
+Note that there have been *major* changes from the application interface
+presented by IJG version 4 and earlier versions.  The old design had several
+inherent limitations, and it had accumulated a lot of cruft as we added
+features while trying to minimize application-interface changes.  We have
+sacrificed backward compatibility in the version 5 rewrite, but we think the
+improvements justify this.
+
+
+TABLE OF CONTENTS
+-----------------
+
+Overview:
+        Functions provided by the library
+        Outline of typical usage
+Basic library usage:
+        Data formats
+        Compression details
+        Decompression details
+        Partial image decompression
+        Mechanics of usage: include files, linking, etc
+Advanced features:
+        Compression parameter selection
+        Decompression parameter selection
+        Special color spaces
+        Error handling
+        Compressed data handling (source and destination managers)
+        I/O suspension
+        Progressive JPEG support
+        Buffered-image mode
+        Abbreviated datastreams and multiple images
+        Special markers
+        ICC profiles
+        Raw (downsampled) image data
+        Really raw data: DCT coefficients
+        Progress monitoring
+        Memory management
+        Memory usage
+        Library compile-time options
+        Portability considerations
+
+You should read at least the overview and basic usage sections before trying
+to program with the library.  The sections on advanced features can be read
+if and when you need them.
+
+
+OVERVIEW
+========
+
+Functions provided by the library
+---------------------------------
+
+The IJG JPEG library provides C code to read and write JPEG-compressed image
+files.  The surrounding application program receives or supplies image data a
+scanline at a time, using a straightforward uncompressed image format.  All
+details of color conversion and other preprocessing/postprocessing can be
+handled by the library.
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  The application indirectly selects use of this code
+by specifying the format in which it wishes to supply or receive image data.
+For example, if colormapped output is requested, then the decompression
+library automatically invokes color quantization.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+A word about functions *not* provided by the library.  We handle a subset of
+the ISO JPEG standard; most baseline, extended-sequential, and progressive
+JPEG processes are supported.  (Our subset includes all features now in common
+use.)  Unsupported ISO options include:
+        * Hierarchical storage
+        * Lossless JPEG
+        * DNL marker
+        * Nonintegral subsampling ratios
+We support both 8- and 12-bit data precision, but this is a compile-time
+choice rather than a run-time choice; hence it is difficult to use both
+precisions in a single application.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, this library is
+used by the free LIBTIFF library to support JPEG compression in TIFF.)
+
+
+Outline of typical usage
+------------------------
+
+The rough outline of a JPEG compression operation is:
+
+        Allocate and initialize a JPEG compression object
+        Specify the destination for the compressed data (eg, a file)
+        Set parameters for compression, including image size & colorspace
+        jpeg_start_compress(...);
+        while (scan lines remain to be written)
+                jpeg_write_scanlines(...);
+        jpeg_finish_compress(...);
+        Release the JPEG compression object
+
+A JPEG compression object holds parameters and working state for the JPEG
+library.  We make creation/destruction of the object separate from starting
+or finishing compression of an image; the same object can be re-used for a
+series of image compression operations.  This makes it easy to re-use the
+same parameter settings for a sequence of images.  Re-use of a JPEG object
+also has important implications for processing abbreviated JPEG datastreams,
+as discussed later.
+
+The image data to be compressed is supplied to jpeg_write_scanlines() from
+in-memory buffers.  If the application is doing file-to-file compression,
+reading image data from the source file is the application's responsibility.
+The library emits compressed data by calling a "data destination manager",
+which typically will write the data into a file; but the application can
+provide its own destination manager to do something else.
+
+Similarly, the rough outline of a JPEG decompression operation is:
+
+        Allocate and initialize a JPEG decompression object
+        Specify the source of the compressed data (eg, a file)
+        Call jpeg_read_header() to obtain image info
+        Set parameters for decompression
+        jpeg_start_decompress(...);
+        while (scan lines remain to be read)
+                jpeg_read_scanlines(...);
+        jpeg_finish_decompress(...);
+        Release the JPEG decompression object
+
+This is comparable to the compression outline except that reading the
+datastream header is a separate step.  This is helpful because information
+about the image's size, colorspace, etc is available when the application
+selects decompression parameters.  For example, the application can choose an
+output scaling ratio that will fit the image into the available screen size.
+
+The decompression library obtains compressed data by calling a data source
+manager, which typically will read the data from a file; but other behaviors
+can be obtained with a custom source manager.  Decompressed data is delivered
+into in-memory buffers passed to jpeg_read_scanlines().
+
+It is possible to abort an incomplete compression or decompression operation
+by calling jpeg_abort(); or, if you do not need to retain the JPEG object,
+simply release it by calling jpeg_destroy().
+
+JPEG compression and decompression objects are two separate struct types.
+However, they share some common fields, and certain routines such as
+jpeg_destroy() can work on either type of object.
+
+The JPEG library has no static variables: all state is in the compression
+or decompression object.  Therefore it is possible to process multiple
+compression and decompression operations concurrently, using multiple JPEG
+objects.
+
+Both compression and decompression can be done in an incremental memory-to-
+memory fashion, if suitable source/destination managers are used.  See the
+section on "I/O suspension" for more details.
+
+
+BASIC LIBRARY USAGE
+===================
+
+Data formats
+------------
+
+Before diving into procedural details, it is helpful to understand the
+image data format that the JPEG library expects or returns.
+
+The standard input image format is a rectangular array of pixels, with each
+pixel having the same number of "component" or "sample" values (color
+channels).  You must specify how many components there are and the colorspace
+interpretation of the components.  Most applications will use RGB data
+(three components per pixel) or grayscale data (one component per pixel).
+PLEASE NOTE THAT RGB DATA IS THREE SAMPLES PER PIXEL, GRAYSCALE ONLY ONE.
+A remarkable number of people manage to miss this, only to find that their
+programs don't work with grayscale JPEG files.
+
+There is no provision for colormapped input.  JPEG files are always full-color
+or full grayscale (or sometimes another colorspace such as CMYK).  You can
+feed in a colormapped image by expanding it to full-color format.  However
+JPEG often doesn't work very well with source data that has been colormapped,
+because of dithering noise.  This is discussed in more detail in the JPEG FAQ
+and the other references mentioned in the README.ijg file.
+
+Pixels are stored by scanlines, with each scanline running from left to
+right.  The component values for each pixel are adjacent in the row; for
+example, R,G,B,R,G,B,R,G,B,... for 24-bit RGB color.  Each scanline is an
+array of data type JSAMPLE --- which is typically "unsigned char", unless
+you've changed jmorecfg.h.  (You can also change the RGB pixel layout, say
+to B,G,R order, by modifying jmorecfg.h.  But see the restrictions listed in
+that file before doing so.)
+
+A 2-D array of pixels is formed by making a list of pointers to the starts of
+scanlines; so the scanlines need not be physically adjacent in memory.  Even
+if you process just one scanline at a time, you must make a one-element
+pointer array to conform to this structure.  Pointers to JSAMPLE rows are of
+type JSAMPROW, and the pointer to the pointer array is of type JSAMPARRAY.
+
+The library accepts or supplies one or more complete scanlines per call.
+It is not possible to process part of a row at a time.  Scanlines are always
+processed top-to-bottom.  You can process an entire image in one call if you
+have it all in memory, but usually it's simplest to process one scanline at
+a time.
+
+For best results, source data values should have the precision specified by
+BITS_IN_JSAMPLE (normally 8 bits).  For instance, if you choose to compress
+data that's only 6 bits/channel, you should left-justify each value in a
+byte before passing it to the compressor.  If you need to compress data
+that has more than 8 bits/channel, compile with BITS_IN_JSAMPLE = 12.
+(See "Library compile-time options", later.)
+
+
+The data format returned by the decompressor is the same in all details,
+except that colormapped output is supported.  (Again, a JPEG file is never
+colormapped.  But you can ask the decompressor to perform on-the-fly color
+quantization to deliver colormapped output.)  If you request colormapped
+output then the returned data array contains a single JSAMPLE per pixel;
+its value is an index into a color map.  The color map is represented as
+a 2-D JSAMPARRAY in which each row holds the values of one color component,
+that is, colormapij is the value of the i'th color component for pixel
+value (map index) j.  Note that since the colormap indexes are stored in
+JSAMPLEs, the maximum number of colors is limited by the size of JSAMPLE
+(ie, at most 256 colors for an 8-bit JPEG library).
+
+
+Compression details
+-------------------
+
+Here we revisit the JPEG compression outline given in the overview.
+
+1. Allocate and initialize a JPEG compression object.
+
+A JPEG compression object is a "struct jpeg_compress_struct".  (It also has
+a bunch of subsidiary structures which are allocated via malloc(), but the
+application doesn't control those directly.)  This struct can be just a local
+variable in the calling routine, if a single routine is going to execute the
+whole JPEG compression sequence.  Otherwise it can be static or allocated
+from malloc().
+
+You will also need a structure representing a JPEG error handler.  The part
+of this that the library cares about is a "struct jpeg_error_mgr".  If you
+are providing your own error handler, you'll typically want to embed the
+jpeg_error_mgr struct in a larger structure; this is discussed later under
+"Error handling".  For now we'll assume you are just using the default error
+handler.  The default error handler will print JPEG error/warning messages
+on stderr, and it will call exit() if a fatal error occurs.
+
+You must initialize the error handler structure, store a pointer to it into
+the JPEG object's "err" field, and then call jpeg_create_compress() to
+initialize the rest of the JPEG object.
+
+Typical code for this step, if you are using the default error handler, is
+
+        struct jpeg_compress_struct cinfo;
+        struct jpeg_error_mgr jerr;
+        ...
+        cinfo.err = jpeg_std_error(&jerr);
+        jpeg_create_compress(&cinfo);
+
+jpeg_create_compress allocates a small amount of memory, so it could fail
+if you are out of memory.  In that case it will exit via the error handler;
+that's why the error handler must be initialized first.
+
+
+2. Specify the destination for the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library delivers compressed data to a
+"data destination" module.  The library includes one data destination
+module which knows how to write to a stdio stream.  You can use your own
+destination module if you want to do something else, as discussed later.
+
+If you use the standard destination module, you must open the target stdio
+stream beforehand.  Typical code for this step looks like:
+
+        FILE *outfile;
+        ...
+        if ((outfile = fopen(filename, "wb")) == NULL) {
+            fprintf(stderr, "can't open %s\n", filename);
+            exit(1);
+        }
+        jpeg_stdio_dest(&cinfo, outfile);
+
+where the last line invokes the standard destination module.
+
+WARNING: it is critical that the binary compressed data be delivered to the
+output file unchanged.  On non-Unix systems the stdio library may perform
+newline translation or otherwise corrupt binary data.  To suppress this
+behavior, you may need to use a "b" option to fopen (as shown above), or use
+setmode() or another routine to put the stdio stream in binary mode.  See
+cjpeg.c and djpeg.c for code that has been found to work on many systems.
+
+You can select the data destination after setting other parameters (step 3),
+if that's more convenient.  You may not change the destination between
+calling jpeg_start_compress() and jpeg_finish_compress().
+
+
+3. Set parameters for compression, including image size & colorspace.
+
+You must supply information about the source image by setting the following
+fields in the JPEG object (cinfo structure):
+
+        image_width             Width of image, in pixels
+        image_height            Height of image, in pixels
+        input_components        Number of color channels (samples per pixel)
+        in_color_space          Color space of source image
+
+The image dimensions are, hopefully, obvious.  JPEG supports image dimensions
+of 1 to 64K pixels in either direction.  The input color space is typically
+RGB or grayscale, and input_components is 3 or 1 accordingly.  (See "Special
+color spaces", later, for more info.)  The in_color_space field must be
+assigned one of the J_COLOR_SPACE enum constants, typically JCS_RGB or
+JCS_GRAYSCALE.
+
+JPEG has a large number of compression parameters that determine how the
+image is encoded.  Most applications don't need or want to know about all
+these parameters.  You can set all the parameters to reasonable defaults by
+calling jpeg_set_defaults(); then, if there are particular values you want
+to change, you can do so after that.  The "Compression parameter selection"
+section tells about all the parameters.
+
+You must set in_color_space correctly before calling jpeg_set_defaults(),
+because the defaults depend on the source image colorspace.  However the
+other three source image parameters need not be valid until you call
+jpeg_start_compress().  There's no harm in calling jpeg_set_defaults() more
+than once, if that happens to be convenient.
+
+Typical code for a 24-bit RGB source image is
+
+        cinfo.image_width = Width;      /* image width and height, in pixels */
+        cinfo.image_height = Height;
+        cinfo.input_components = 3;     /* # of color components per pixel */
+        cinfo.in_color_space = JCS_RGB; /* colorspace of input image */
+
+        jpeg_set_defaults(&cinfo);
+        /* Make optional parameter settings here */
+
+
+4. jpeg_start_compress(...);
+
+After you have established the data destination and set all the necessary
+source image info and other parameters, call jpeg_start_compress() to begin
+a compression cycle.  This will initialize internal state, allocate working
+storage, and emit the first few bytes of the JPEG datastream header.
+
+Typical code:
+
+        jpeg_start_compress(&cinfo, TRUE);
+
+The "TRUE" parameter ensures that a complete JPEG interchange datastream
+will be written.  This is appropriate in most cases.  If you think you might
+want to use an abbreviated datastream, read the section on abbreviated
+datastreams, below.
+
+Once you have called jpeg_start_compress(), you may not alter any JPEG
+parameters or other fields of the JPEG object until you have completed
+the compression cycle.
+
+
+5. while (scan lines remain to be written)
+        jpeg_write_scanlines(...);
+
+Now write all the required image data by calling jpeg_write_scanlines()
+one or more times.  You can pass one or more scanlines in each call, up
+to the total image height.  In most applications it is convenient to pass
+just one or a few scanlines at a time.  The expected format for the passed
+data is discussed under "Data formats", above.
+
+Image data should be written in top-to-bottom scanline order.
+Rec. ITU-T T.81 | ISO/IEC 10918-1 says, "Applications determine which edges of
+a source image are defined as top, bottom, left, and right."  However, if you
+want your files to be compatible with everyone else's, then top-to-bottom order
+must be used.  If the source data must be read in bottom-to-top order, then you
+can use the JPEG library's virtual array mechanism to invert the data
+efficiently.  Examples of this can be found in the sample application cjpeg.
+
+The library maintains a count of the number of scanlines written so far
+in the next_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.next_scanline < cinfo.image_height)".
+
+Code for this step depends heavily on the way that you store the source data.
+example.txt shows the following code for the case of a full-size 2-D source
+array containing 3-byte RGB pixels:
+
+        JSAMPROW row_pointer1;        /* pointer to a single row */
+        int row_stride;                 /* physical row width in buffer */
+
+        row_stride = image_width * 3;   /* JSAMPLEs per row in image_buffer */
+
+        while (cinfo.next_scanline < cinfo.image_height) {
+            row_pointer0 = &image_buffercinfo.next_scanline * row_stride;
+            jpeg_write_scanlines(&cinfo, row_pointer, 1);
+        }
+
+jpeg_write_scanlines() returns the number of scanlines actually written.
+This will normally be equal to the number passed in, so you can usually
+ignore the return value.  It is different in just two cases:
+  * If you try to write more scanlines than the declared image height,
+    the additional scanlines are ignored.
+  * If you use a suspending data destination manager, output buffer overrun
+    will cause the compressor to return before accepting all the passed lines.
+    This feature is discussed under "I/O suspension", below.  The normal
+    stdio destination manager will NOT cause this to happen.
+In any case, the return value is the same as the change in the value of
+next_scanline.
+
+
+6. jpeg_finish_compress(...);
+
+After all the image data has been written, call jpeg_finish_compress() to
+complete the compression cycle.  This step is ESSENTIAL to ensure that the
+last bufferload of data is written to the data destination.
+jpeg_finish_compress() also releases working memory associated with the JPEG
+object.
+
+Typical code:
+
+        jpeg_finish_compress(&cinfo);
+
+If using the stdio destination manager, don't forget to close the output
+stdio stream (if necessary) afterwards.
+
+If you have requested a multi-pass operating mode, such as Huffman code
+optimization, jpeg_finish_compress() will perform the additional passes using
+data buffered by the first pass.  In this case jpeg_finish_compress() may take
+quite a while to complete.  With the default compression parameters, this will
+not happen.
+
+It is an error to call jpeg_finish_compress() before writing the necessary
+total number of scanlines.  If you wish to abort compression, call
+jpeg_abort() as discussed below.
+
+After completing a compression cycle, you may dispose of the JPEG object
+as discussed next, or you may use it to compress another image.  In that case
+return to step 2, 3, or 4 as appropriate.  If you do not change the
+destination manager, the new datastream will be written to the same target.
+If you do not change any JPEG parameters, the new datastream will be written
+with the same parameters as before.  Note that you can change the input image
+dimensions freely between cycles, but if you change the input colorspace, you
+should call jpeg_set_defaults() to adjust for the new colorspace; and then
+you'll need to repeat all of step 3.
+
+
+7. Release the JPEG compression object.
+
+When you are done with a JPEG compression object, destroy it by calling
+jpeg_destroy_compress().  This will free all subsidiary memory (regardless of
+the previous state of the object).  Or you can call jpeg_destroy(), which
+works for either compression or decompression objects --- this may be more
+convenient if you are sharing code between compression and decompression
+cases.  (Actually, these routines are equivalent except for the declared type
+of the passed pointer.  To avoid gripes from ANSI C compilers, jpeg_destroy()
+should be passed a j_common_ptr.)
+
+If you allocated the jpeg_compress_struct structure from malloc(), freeing
+it is your responsibility --- jpeg_destroy() won't.  Ditto for the error
+handler structure.
+
+Typical code:
+
+        jpeg_destroy_compress(&cinfo);
+
+
+8. Aborting.
+
+If you decide to abort a compression cycle before finishing, you can clean up
+in either of two ways:
+
+* If you don't need the JPEG object any more, just call
+  jpeg_destroy_compress() or jpeg_destroy() to release memory.  This is
+  legitimate at any point after calling jpeg_create_compress() --- in fact,
+  it's safe even if jpeg_create_compress() fails.
+
+* If you want to re-use the JPEG object, call jpeg_abort_compress(), or call
+  jpeg_abort() which works on both compression and decompression objects.
+  This will return the object to an idle state, releasing any working memory.
+  jpeg_abort() is allowed at any time after successful object creation.
+
+Note that cleaning up the data destination, if required, is your
+responsibility; neither of these routines will call term_destination().
+(See "Compressed data handling", below, for more about that.)
+
+jpeg_destroy() and jpeg_abort() are the only safe calls to make on a JPEG
+object that has reported an error by calling error_exit (see "Error handling"
+for more info).  The internal state of such an object is likely to be out of
+whack.  Either of these two routines will return the object to a known state.
+
+
+Decompression details
+---------------------
+
+Here we revisit the JPEG decompression outline given in the overview.
+
+1. Allocate and initialize a JPEG decompression object.
+
+This is just like initialization for compression, as discussed above,
+except that the object is a "struct jpeg_decompress_struct" and you
+call jpeg_create_decompress().  Error handling is exactly the same.
+
+Typical code:
+
+        struct jpeg_decompress_struct cinfo;
+        struct jpeg_error_mgr jerr;
+        ...
+        cinfo.err = jpeg_std_error(&jerr);
+        jpeg_create_decompress(&cinfo);
+
+(Both here and in the IJG code, we usually use variable name "cinfo" for
+both compression and decompression objects.)
+
+
+2. Specify the source of the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library reads compressed data from a "data
+source" module.  The library includes one data source module which knows how
+to read from a stdio stream.  You can use your own source module if you want
+to do something else, as discussed later.
+
+If you use the standard source module, you must open the source stdio stream
+beforehand.  Typical code for this step looks like:
+
+        FILE *infile;
+        ...
+        if ((infile = fopen(filename, "rb")) == NULL) {
+            fprintf(stderr, "can't open %s\n", filename);
+            exit(1);
+        }
+        jpeg_stdio_src(&cinfo, infile);
+
+where the last line invokes the standard source module.
+
+WARNING: it is critical that the binary compressed data be read unchanged.
+On non-Unix systems the stdio library may perform newline translation or
+otherwise corrupt binary data.  To suppress this behavior, you may need to use
+a "b" option to fopen (as shown above), or use setmode() or another routine to
+put the stdio stream in binary mode.  See cjpeg.c and djpeg.c for code that
+has been found to work on many systems.
+
+You may not change the data source between calling jpeg_read_header() and
+jpeg_finish_decompress().  If you wish to read a series of JPEG images from
+a single source file, you should repeat the jpeg_read_header() to
+jpeg_finish_decompress() sequence without reinitializing either the JPEG
+object or the data source module; this prevents buffered input data from
+being discarded.
+
+
+3. Call jpeg_read_header() to obtain image info.
+
+Typical code for this step is just
+
+        jpeg_read_header(&cinfo, TRUE);
+
+This will read the source datastream header markers, up to the beginning
+of the compressed data proper.  On return, the image dimensions and other
+info have been stored in the JPEG object.  The application may wish to
+consult this information before selecting decompression parameters.
+
+More complex code is necessary if
+  * A suspending data source is used --- in that case jpeg_read_header()
+    may return before it has read all the header data.  See "I/O suspension",
+    below.  The normal stdio source manager will NOT cause this to happen.
+  * Abbreviated JPEG files are to be processed --- see the section on
+    abbreviated datastreams.  Standard applications that deal only in
+    interchange JPEG files need not be concerned with this case either.
+
+It is permissible to stop at this point if you just wanted to find out the
+image dimensions and other header info for a JPEG file.  In that case,
+call jpeg_destroy() when you are done with the JPEG object, or call
+jpeg_abort() to return it to an idle state before selecting a new data
+source and reading another header.
+
+
+4. Set parameters for decompression.
+
+jpeg_read_header() sets appropriate default decompression parameters based on
+the properties of the image (in particular, its colorspace).  However, you
+may well want to alter these defaults before beginning the decompression.
+For example, the default is to produce full color output from a color file.
+If you want colormapped output you must ask for it.  Other options allow the
+returned image to be scaled and allow various speed/quality tradeoffs to be
+selected.  "Decompression parameter selection", below, gives details.
+
+If the defaults are appropriate, nothing need be done at this step.
+
+Note that all default values are set by each call to jpeg_read_header().
+If you reuse a decompression object, you cannot expect your parameter
+settings to be preserved across cycles, as you can for compression.
+You must set desired parameter values each time.
+
+
+5. jpeg_start_decompress(...);
+
+Once the parameter values are satisfactory, call jpeg_start_decompress() to
+begin decompression.  This will initialize internal state, allocate working
+memory, and prepare for returning data.
+
+Typical code is just
+
+        jpeg_start_decompress(&cinfo);
+
+If you have requested a multi-pass operating mode, such as 2-pass color
+quantization, jpeg_start_decompress() will do everything needed before data
+output can begin.  In this case jpeg_start_decompress() may take quite a while
+to complete.  With a single-scan (non progressive) JPEG file and default
+decompression parameters, this will not happen; jpeg_start_decompress() will
+return quickly.
+
+After this call, the final output image dimensions, including any requested
+scaling, are available in the JPEG object; so is the selected colormap, if
+colormapped output has been requested.  Useful fields include
+
+        output_width            image width and height, as scaled
+        output_height
+        out_color_components    # of color components in out_color_space
+        output_components       # of color components returned per pixel
+        colormap                the selected colormap, if any
+        actual_number_of_colors         number of entries in colormap
+
+output_components is 1 (a colormap index) when quantizing colors; otherwise it
+equals out_color_components.  It is the number of JSAMPLE values that will be
+emitted per pixel in the output arrays.
+
+Typically you will need to allocate data buffers to hold the incoming image.
+You will need output_width * output_components JSAMPLEs per scanline in your
+output buffer, and a total of output_height scanlines will be returned.
+
+Note: if you are using the JPEG library's internal memory manager to allocate
+data buffers (as djpeg does), then the manager's protocol requires that you
+request large buffers *before* calling jpeg_start_decompress().  This is a
+little tricky since the output_XXX fields are not normally valid then.  You
+can make them valid by calling jpeg_calc_output_dimensions() after setting the
+relevant parameters (scaling, output color space, and quantization flag).
+
+
+6. while (scan lines remain to be read)
+        jpeg_read_scanlines(...);
+
+Now you can read the decompressed image data by calling jpeg_read_scanlines()
+one or more times.  At each call, you pass in the maximum number of scanlines
+to be read (ie, the height of your working buffer); jpeg_read_scanlines()
+will return up to that many lines.  The return value is the number of lines
+actually read.  The format of the returned data is discussed under "Data
+formats", above.  Don't forget that grayscale and color JPEGs will return
+different data formats!
+
+Image data is returned in top-to-bottom scanline order.  If you must write
+out the image in bottom-to-top order, you can use the JPEG library's virtual
+array mechanism to invert the data efficiently.  Examples of this can be
+found in the sample application djpeg.
+
+The library maintains a count of the number of scanlines returned so far
+in the output_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.output_scanline < cinfo.output_height)".  (Note that the test
+should NOT be against image_height, unless you never use scaling.  The
+image_height field is the height of the original unscaled image.)
+The return value always equals the change in the value of output_scanline.
+
+If you don't use a suspending data source, it is safe to assume that
+jpeg_read_scanlines() reads at least one scanline per call, until the
+bottom of the image has been reached.
+
+If you use a buffer larger than one scanline, it is NOT safe to assume that
+jpeg_read_scanlines() fills it.  (The current implementation returns only a
+few scanlines per call, no matter how large a buffer you pass.)  So you must
+always provide a loop that calls jpeg_read_scanlines() repeatedly until the
+whole image has been read.
+
+
+7. jpeg_finish_decompress(...);
+
+After all the image data has been read, call jpeg_finish_decompress() to
+complete the decompression cycle.  This causes working memory associated
+with the JPEG object to be released.
+
+Typical code:
+
+        jpeg_finish_decompress(&cinfo);
+
+If using the stdio source manager, don't forget to close the source stdio
+stream if necessary.
+
+It is an error to call jpeg_finish_decompress() before reading the correct
+total number of scanlines.  If you wish to abort decompression, call
+jpeg_abort() as discussed below.
+
+After completing a decompression cycle, you may dispose of the JPEG object as
+discussed next, or you may use it to decompress another image.  In that case
+return to step 2 or 3 as appropriate.  If you do not change the source
+manager, the next image will be read from the same source.
+
+
+8. Release the JPEG decompression object.
+
+When you are done with a JPEG decompression object, destroy it by calling
+jpeg_destroy_decompress() or jpeg_destroy().  The previous discussion of
+destroying compression objects applies here too.
+
+Typical code:
+
+        jpeg_destroy_decompress(&cinfo);
+
+
+9. Aborting.
+
+You can abort a decompression cycle by calling jpeg_destroy_decompress() or
+jpeg_destroy() if you don't need the JPEG object any more, or
+jpeg_abort_decompress() or jpeg_abort() if you want to reuse the object.
+The previous discussion of aborting compression cycles applies here too.
+
+
+Partial image decompression
+---------------------------
+
+Partial image decompression is convenient for performance-critical applications
+that wish to view only a portion of a large JPEG image without decompressing
+the whole thing.  It it also useful in memory-constrained environments (such as
+on mobile devices.)  This library provides the following functions to support
+partial image decompression:
+
+1. Skipping rows when decompressing
+
+        jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines);
+
+This function provides application programmers with the ability to skip over
+multiple rows in the JPEG image.
+
+Suspending data sources are not supported by this function.  Calling
+jpeg_skip_scanlines() with a suspending data source will result in undefined
+behavior.  Two-pass color quantization is also not supported by this function.
+Calling jpeg_skip_scanlines() with two-pass color quantization enabled will
+result in an error.
+
+jpeg_skip_scanlines() will not allow skipping past the bottom of the image.  If
+the value of num_lines is large enough to skip past the bottom of the image,
+then the function will skip to the end of the image instead.
+
+If the value of num_lines is valid, then jpeg_skip_scanlines() will always
+skip all of the input rows requested.  There is no need to inspect the return
+value of the function in that case.
+
+Best results will be achieved by calling jpeg_skip_scanlines() for large chunks
+of rows.  The function should be viewed as a way to quickly jump to a
+particular vertical offset in the JPEG image in order to decode a subset of the
+image.  Used in this manner, it will provide significant performance
+improvements.
+
+Calling jpeg_skip_scanlines() for small values of num_lines has several
+potential drawbacks:
+    1) JPEG decompression occurs in blocks, so if jpeg_skip_scanlines() is
+       called from the middle of a decompression block, then it is likely that
+       much of the decompression work has already been done for the first
+       couple of rows that need to be skipped.
+    2) When this function returns, it must leave the decompressor in a state
+       such that it is ready to read the next line.  This may involve
+       decompressing a block that must be partially skipped.
+These issues are especially tricky for cases in which upsampling requires
+context rows.  In the worst case, jpeg_skip_scanlines() will perform similarly
+to jpeg_read_scanlines() (since it will actually call jpeg_read_scanlines().)
+
+2. Decompressing partial scanlines
+
+        jpeg_crop_scanline (j_decompress_ptr cinfo, JDIMENSION *xoffset,
+                            JDIMENSION *width)
+
+This function provides application programmers with the ability to decompress
+only a portion of each row in the JPEG image.  It must be called after
+jpeg_start_decompress() and before any calls to jpeg_read_scanlines() or
+jpeg_skip_scanlines().
+
+If xoffset and width do not form a valid subset of the image row, then this
+function will generate an error.  Note that if the output image is scaled, then
+xoffset and width are relative to the scaled image dimensions.
+
+xoffset and width are passed by reference because xoffset must fall on an iMCU
+boundary.  If it doesn't, then it will be moved left to the nearest iMCU
+boundary, and width will be increased accordingly.  If the calling program does
+not like the adjusted values of xoffset and width, then it can call
+jpeg_crop_scanline() again with new values (for instance, if it wants to move
+xoffset to the nearest iMCU boundary to the right instead of to the left.)
+
+After calling this function, cinfo->output_width will be set to the adjusted
+width.  This value should be used when allocating an output buffer to pass to
+jpeg_read_scanlines().
+
+The output image from a partial-width decompression will be identical to the
+corresponding image region from a full decode, with one exception:  The "fancy"
+(smooth) h2v2 (4:2:0) and h2v1 (4:2:2) upsampling algorithms fill in the
+missing chroma components by averaging the chroma components from neighboring
+pixels, except on the right and left edges of the image (where there are no
+neighboring pixels.)  When performing a partial-width decompression, these
+"fancy" upsampling algorithms may treat the left and right edges of the partial
+image region as if they are the left and right edges of the image, meaning that
+the upsampling algorithm may be simplified.  The result is that the pixels on
+the left or right edge of the partial image may not be exactly identical to the
+corresponding pixels in the original image.
+
+
+Mechanics of usage: include files, linking, etc
+-----------------------------------------------
+
+Applications using the JPEG library should include the header file jpeglib.h
+to obtain declarations of data types and routines.  Before including
+jpeglib.h, include system headers that define at least the typedefs FILE and
+size_t.  On ANSI-conforming systems, including <stdio.h> is sufficient; on
+older Unix systems, you may need <sys/types.h> to define size_t.
+
+If the application needs to refer to individual JPEG library error codes, also
+include jerror.h to define those symbols.
+
+jpeglib.h indirectly includes the files jconfig.h and jmorecfg.h.  If you are
+installing the JPEG header files in a system directory, you will want to
+install all four files: jpeglib.h, jerror.h, jconfig.h, jmorecfg.h.
+
+The most convenient way to include the JPEG code into your executable program
+is to prepare a library file ("libjpeg.a", or a corresponding name on non-Unix
+machines) and reference it at your link step.  If you use only half of the
+library (only compression or only decompression), only that much code will be
+included from the library, unless your linker is hopelessly brain-damaged.
+The supplied makefiles build libjpeg.a automatically (see install.txt).
+
+While you can build the JPEG library as a shared library if the whim strikes
+you, we don't really recommend it.  The trouble with shared libraries is that
+at some point you'll probably try to substitute a new version of the library
+without recompiling the calling applications.  That generally doesn't work
+because the parameter struct declarations usually change with each new
+version.  In other words, the library's API is *not* guaranteed binary
+compatible across versions; we only try to ensure source-code compatibility.
+(In hindsight, it might have been smarter to hide the parameter structs from
+applications and introduce a ton of access functions instead.  Too late now,
+however.)
+
+It may be worth pointing out that the core JPEG library does not actually
+require the stdio library: only the default source/destination managers and
+error handler need it.  You can use the library in a stdio-less environment
+if you replace those modules and use jmemnobs.c (or another memory manager of
+your own devising).  More info about the minimum system library requirements
+may be found in jinclude.h.
+
+
+ADVANCED FEATURES
+=================
+
+Compression parameter selection
+-------------------------------
+
+This section describes all the optional parameters you can set for JPEG
+compression, as well as the "helper" routines provided to assist in this
+task.  Proper setting of some parameters requires detailed understanding
+of the JPEG standard; if you don't know what a parameter is for, it's best
+not to mess with it!  See REFERENCES in the README.ijg file for pointers to
+more info about JPEG.
+
+It's a good idea to call jpeg_set_defaults() first, even if you plan to set
+all the parameters; that way your code is more likely to work with future JPEG
+libraries that have additional parameters.  For the same reason, we recommend
+you use a helper routine where one is provided, in preference to twiddling
+cinfo fields directly.
+
+The helper routines are:
+
+jpeg_set_defaults (j_compress_ptr cinfo)
+        This routine sets all JPEG parameters to reasonable defaults, using
+        only the input image's color space (field in_color_space, which must
+        already be set in cinfo).  Many applications will only need to use
+        this routine and perhaps jpeg_set_quality().
+
+jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
+        Sets the JPEG file's colorspace (field jpeg_color_space) as specified,
+        and sets other color-space-dependent parameters appropriately.  See
+        "Special color spaces", below, before using this.  A large number of
+        parameters, including all per-component parameters, are set by this
+        routine; if you want to twiddle individual parameters you should call
+        jpeg_set_colorspace() before rather than after.
+
+jpeg_default_colorspace (j_compress_ptr cinfo)
+        Selects an appropriate JPEG colorspace based on cinfo->in_color_space,
+        and calls jpeg_set_colorspace().  This is actually a subroutine of
+        jpeg_set_defaults().  It's broken out in case you want to change
+        just the colorspace-dependent JPEG parameters.
+
+jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)
+        Constructs JPEG quantization tables appropriate for the indicated
+        quality setting.  The quality value is expressed on the 0..100 scale
+        recommended by IJG (cjpeg's "-quality" switch uses this routine).
+        Note that the exact mapping from quality values to tables may change
+        in future IJG releases as more is learned about DCT quantization.
+        If the force_baseline parameter is TRUE, then the quantization table
+        entries are constrained to the range 1..255 for full JPEG baseline
+        compatibility.  In the current implementation, this only makes a
+        difference for quality settings below 25, and it effectively prevents
+        very small/low quality files from being generated.  The IJG decoder
+        is capable of reading the non-baseline files generated at low quality
+        settings when force_baseline is FALSE, but other decoders may not be.
+
+jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
+                         boolean force_baseline)
+        Same as jpeg_set_quality() except that the generated tables are the
+        sample tables given in Annex K (Clause K.1) of
+        Rec. ITU-T T.81 (1992) | ISO/IEC 10918-1:1994, multiplied by the
+        specified scale factor (which is expressed as a percentage; thus
+        scale_factor = 100 reproduces the spec's tables).  Note that larger
+        scale factors give lower quality.  This entry point is useful for
+        conforming to the Adobe PostScript DCT conventions, but we do not
+        recommend linear scaling as a user-visible quality scale otherwise.
+        force_baseline again constrains the computed table entries to 1..255.
+
+int jpeg_quality_scaling (int quality)
+        Converts a value on the IJG-recommended quality scale to a linear
+        scaling percentage.  Note that this routine may change or go away
+        in future releases --- IJG may choose to adopt a scaling method that
+        can't be expressed as a simple scalar multiplier, in which case the
+        premise of this routine collapses.  Caveat user.
+
+jpeg_default_qtables (j_compress_ptr cinfo, boolean force_baseline)
+        libjpeg v7+ API/ABI emulation only
+        Set default quantization tables with linear q_scale_factor values
+        (see below).
+
+jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
+                      const unsigned int *basic_table,
+                      int scale_factor, boolean force_baseline)
+        Allows an arbitrary quantization table to be created.  which_tbl
+        indicates which table slot to fill.  basic_table points to an array
+        of 64 unsigned ints given in normal array order.  These values are
+        multiplied by scale_factor/100 and then clamped to the range 1..65535
+        (or to 1..255 if force_baseline is TRUE).
+        CAUTION: prior to library version 6a, jpeg_add_quant_table expected
+        the basic table to be given in JPEG zigzag order.  If you need to
+        write code that works with either older or newer versions of this
+        routine, you must check the library version number.  Something like
+        "#if JPEG_LIB_VERSION >= 61" is the right test.
+
+jpeg_simple_progression (j_compress_ptr cinfo)
+        Generates a default scan script for writing a progressive-JPEG file.
+        This is the recommended method of creating a progressive file,
+        unless you want to make a custom scan sequence.  You must ensure that
+        the JPEG color space is set correctly before calling this routine.
+
+
+Compression parameters (cinfo fields) include:
+
+boolean arith_code
+        If TRUE, use arithmetic coding.
+        If FALSE, use Huffman coding.
+
+J_DCT_METHOD dct_method
+        Selects the algorithm used for the DCT step.  Choices are:
+                JDCT_ISLOW: accurate integer method
+                JDCT_IFAST: less accurate integer method legacy feature
+                JDCT_FLOAT: floating-point method legacy feature
+                JDCT_DEFAULT: default method (normally JDCT_ISLOW)
+                JDCT_FASTEST: fastest method (normally JDCT_IFAST)
+        When the Independent JPEG Group's software was first released in 1991,
+        the compression time for a 1-megapixel JPEG image on a mainstream PC
+        was measured in minutes.  Thus, JDCT_IFAST provided noticeable
+        performance benefits.  On modern CPUs running libjpeg-turbo, however,
+        the compression time for a 1-megapixel JPEG image is measured in
+        milliseconds, and thus the performance benefits of JDCT_IFAST are much
+        less noticeable.  On modern x86/x86-64 CPUs that support AVX2
+        instructions, JDCT_IFAST and JDCT_ISLOW have similar performance.  On
+        other types of CPUs, JDCT_IFAST is generally about 5-15% faster than
+        JDCT_ISLOW.
+
+        For quality levels of 90 and below, there should be little or no
+        perceptible quality difference between the two algorithms.  For quality
+        levels above 90, however, the difference between JDCT_IFAST and
+        JDCT_ISLOW becomes more pronounced.  With quality=97, for instance,
+        JDCT_IFAST incurs generally about a 1-3 dB loss in PSNR relative to
+        JDCT_ISLOW, but this can be larger for some images.  Do not use
+        JDCT_IFAST with quality levels above 97.  The algorithm often
+        degenerates at quality=98 and above and can actually produce a more
+        lossy image than if lower quality levels had been used.  Also, in
+        libjpeg-turbo, JDCT_IFAST is not fully accelerated for quality levels
+        above 97, so it will be slower than JDCT_ISLOW.
+
+        JDCT_FLOAT does not produce significantly more accurate results than
+        JDCT_ISLOW, and it is much slower.  JDCT_FLOAT may also give different
+        results on different machines due to varying roundoff behavior, whereas
+        the integer methods should give the same results on all machines.
+
+J_COLOR_SPACE jpeg_color_space
+int num_components
+        The JPEG color space and corresponding number of components; see
+        "Special color spaces", below, for more info.  We recommend using
+        jpeg_set_color_space() if you want to change these.
+
+boolean optimize_coding
+        TRUE causes the compressor to compute optimal Huffman coding tables
+        for the image.  This requires an extra pass over the data and
+        therefore costs a good deal of space and time.  The default is
+        FALSE, which tells the compressor to use the supplied or default
+        Huffman tables.  In most cases optimal tables save only a few percent
+        of file size compared to the default tables.  Note that when this is
+        TRUE, you need not supply Huffman tables at all, and any you do
+        supply will be overwritten.
+
+unsigned int restart_interval
+int restart_in_rows
+        To emit restart markers in the JPEG file, set one of these nonzero.
+        Set restart_interval to specify the exact interval in MCU blocks.
+        Set restart_in_rows to specify the interval in MCU rows.  (If
+        restart_in_rows is not 0, then restart_interval is set after the
+        image width in MCUs is computed.)  Defaults are zero (no restarts).
+        One restart marker per MCU row is often a good choice.
+        NOTE: the overhead of restart markers is higher in grayscale JPEG
+        files than in color files, and MUCH higher in progressive JPEGs.
+        If you use restarts, you may want to use larger intervals in those
+        cases.
+
+const jpeg_scan_info *scan_info
+int num_scans
+        By default, scan_info is NULL; this causes the compressor to write a
+        single-scan sequential JPEG file.  If not NULL, scan_info points to
+        an array of scan definition records of length num_scans.  The
+        compressor will then write a JPEG file having one scan for each scan
+        definition record.  This is used to generate noninterleaved or
+        progressive JPEG files.  The library checks that the scan array
+        defines a valid JPEG scan sequence.  (jpeg_simple_progression creates
+        a suitable scan definition array for progressive JPEG.)  This is
+        discussed further under "Progressive JPEG support".
+
+int smoothing_factor
+        If non-zero, the input image is smoothed; the value should be 1 for
+        minimal smoothing to 100 for maximum smoothing.  Consult jcsample.c
+        for details of the smoothing algorithm.  The default is zero.
+
+boolean write_JFIF_header
+        If TRUE, a JFIF APP0 marker is emitted.  jpeg_set_defaults() and
+        jpeg_set_colorspace() set this TRUE if a JFIF-legal JPEG color space
+        (ie, YCbCr or grayscale) is selected, otherwise FALSE.
+
+UINT8 JFIF_major_version
+UINT8 JFIF_minor_version
+        The version number to be written into the JFIF marker.
+        jpeg_set_defaults() initializes the version to 1.01 (major=minor=1).
+        You should set it to 1.02 (major=1, minor=2) if you plan to write
+        any JFIF 1.02 extension markers.
+
+UINT8 density_unit
+UINT16 X_density
+UINT16 Y_density
+        The resolution information to be written into the JFIF marker;
+        not used otherwise.  density_unit may be 0 for unknown,
+        1 for dots/inch, or 2 for dots/cm.  The default values are 0,1,1
+        indicating square pixels of unknown size.
+
+boolean write_Adobe_marker
+        If TRUE, an Adobe APP14 marker is emitted.  jpeg_set_defaults() and
+        jpeg_set_colorspace() set this TRUE if JPEG color space RGB, CMYK,
+        or YCCK is selected, otherwise FALSE.  It is generally a bad idea
+        to set both write_JFIF_header and write_Adobe_marker.  In fact,
+        you probably shouldn't change the default settings at all --- the
+        default behavior ensures that the JPEG file's color space can be
+        recognized by the decoder.
+
+JQUANT_TBL *quant_tbl_ptrsNUM_QUANT_TBLS
+        Pointers to coefficient quantization tables, one per table slot,
+        or NULL if no table is defined for a slot.  Usually these should
+        be set via one of the above helper routines; jpeg_add_quant_table()
+        is general enough to define any quantization table.  The other
+        routines will set up table slot 0 for luminance quality and table
+        slot 1 for chrominance.
+
+int q_scale_factorNUM_QUANT_TBLS
+        libjpeg v7+ API/ABI emulation only
+        Linear quantization scaling factors (0-100, default 100)
+        for use with jpeg_default_qtables().
+        See rdswitch.c and cjpeg.c for an example of usage.
+        Note that the q_scale_factor values use "linear" scales, so JPEG
+        quality levels chosen by the user must be converted to these scales
+        using jpeg_quality_scaling().  Here is an example that corresponds to
+        cjpeg -quality 90,70:
+
+                jpeg_set_defaults(cinfo);
+
+                /* Set luminance quality 90. */
+                cinfo->q_scale_factor0 = jpeg_quality_scaling(90);
+                /* Set chrominance quality 70. */
+                cinfo->q_scale_factor1 = jpeg_quality_scaling(70);
+
+                jpeg_default_qtables(cinfo, force_baseline);
+
+        CAUTION: Setting separate quality levels for chrominance and luminance
+        is mainly only useful if chrominance subsampling is disabled.  2x2
+        chrominance subsampling (AKA "4:2:0") is the default, but you can
+        explicitly disable subsampling as follows:
+
+                cinfo->comp_info0.v_samp_factor = 1;
+                cinfo->comp_info0.h_samp_factor = 1;
+
+JHUFF_TBL *dc_huff_tbl_ptrsNUM_HUFF_TBLS
+JHUFF_TBL *ac_huff_tbl_ptrsNUM_HUFF_TBLS
+        Pointers to Huffman coding tables, one per table slot, or NULL if
+        no table is defined for a slot.  Slots 0 and 1 are filled with the
+        JPEG sample tables by jpeg_set_defaults().  If you need to allocate
+        more table structures, jpeg_alloc_huff_table() may be used.
+        Note that optimal Huffman tables can be computed for an image
+        by setting optimize_coding, as discussed above; there's seldom
+        any need to mess with providing your own Huffman tables.
+
+
+libjpeg v7+ API/ABI emulation only
+The actual dimensions of the JPEG image that will be written to the file are
+given by the following fields.  These are computed from the input image
+dimensions and the compression parameters by jpeg_start_compress().  You can
+also call jpeg_calc_jpeg_dimensions() to obtain the values that will result
+from the current parameter settings.  This can be useful if you are trying
+to pick a scaling ratio that will get close to a desired target size.
+
+JDIMENSION jpeg_width           Actual dimensions of output image.
+JDIMENSION jpeg_height
+
+
+Per-component parameters are stored in the struct cinfo.comp_infoi for
+component number i.  Note that components here refer to components of the
+JPEG color space, *not* the source image color space.  A suitably large
+comp_info array is allocated by jpeg_set_defaults(); if you choose not
+to use that routine, it's up to you to allocate the array.
+
+int component_id
+        The one-byte identifier code to be recorded in the JPEG file for
+        this component.  For the standard color spaces, we recommend you
+        leave the default values alone.
+
+int h_samp_factor
+int v_samp_factor
+        Horizontal and vertical sampling factors for the component; must
+        be 1..4 according to the JPEG standard.  Note that larger sampling
+        factors indicate a higher-resolution component; many people find
+        this behavior quite unintuitive.  The default values are 2,2 for
+        luminance components and 1,1 for chrominance components, except
+        for grayscale where 1,1 is used.
+
+int quant_tbl_no
+        Quantization table number for component.  The default value is
+        0 for luminance components and 1 for chrominance components.
+
+int dc_tbl_no
+int ac_tbl_no
+        DC and AC entropy coding table numbers.  The default values are
+        0 for luminance components and 1 for chrominance components.
+
+int component_index
+        Must equal the component's index in comp_info.  (Beginning in
+        release v6, the compressor library will fill this in automatically;
+        you don't have to.)
+
+
+Decompression parameter selection
+---------------------------------
+
+Decompression parameter selection is somewhat simpler than compression
+parameter selection, since all of the JPEG internal parameters are
+recorded in the source file and need not be supplied by the application.
+(Unless you are working with abbreviated files, in which case see
+"Abbreviated datastreams", below.)  Decompression parameters control
+the postprocessing done on the image to deliver it in a format suitable
+for the application's use.  Many of the parameters control speed/quality
+tradeoffs, in which faster decompression may be obtained at the price of
+a poorer-quality image.  The defaults select the highest quality (slowest)
+processing.
+
+The following fields in the JPEG object are set by jpeg_read_header() and
+may be useful to the application in choosing decompression parameters:
+
+JDIMENSION image_width                  Width and height of image
+JDIMENSION image_height
+int num_components                      Number of color components
+J_COLOR_SPACE jpeg_color_space          Colorspace of image
+boolean saw_JFIF_marker                 TRUE if a JFIF APP0 marker was seen
+  UINT8 JFIF_major_version              Version information from JFIF marker
+  UINT8 JFIF_minor_version
+  UINT8 density_unit                    Resolution data from JFIF marker
+  UINT16 X_density
+  UINT16 Y_density
+boolean saw_Adobe_marker                TRUE if an Adobe APP14 marker was seen
+  UINT8 Adobe_transform                 Color transform code from Adobe marker
+
+The JPEG color space, unfortunately, is something of a guess since the JPEG
+standard proper does not provide a way to record it.  In practice most files
+adhere to the JFIF or Adobe conventions, and the decoder will recognize these
+correctly.  See "Special color spaces", below, for more info.
+
+
+The decompression parameters that determine the basic properties of the
+returned image are:
+
+J_COLOR_SPACE out_color_space
+        Output color space.  jpeg_read_header() sets an appropriate default
+        based on jpeg_color_space; typically it will be RGB or grayscale.
+        The application can change this field to request output in a different
+        colorspace.  For example, set it to JCS_GRAYSCALE to get grayscale
+        output from a color file.  (This is useful for previewing: grayscale
+        output is faster than full color since the color components need not
+        be processed.)  Note that not all possible color space transforms are
+        currently implemented; you may need to extend jdcolor.c if you want an
+        unusual conversion.
+
+unsigned int scale_num, scale_denom
+        Scale the image by the fraction scale_num/scale_denom.  Default is
+        1/1, or no scaling.  Currently, the only supported scaling ratios
+        are M/8 with all M from 1 to 16, or any reduced fraction thereof (such
+        as 1/2, 3/4, etc.)  (The library design allows for arbitrary
+        scaling ratios but this is not likely to be implemented any time soon.)
+        Smaller scaling ratios permit significantly faster decoding since
+        fewer pixels need be processed and a simpler IDCT method can be used.
+
+boolean quantize_colors
+        If set TRUE, colormapped output will be delivered.  Default is FALSE,
+        meaning that full-color output will be delivered.
+
+The next three parameters are relevant only if quantize_colors is TRUE.
+
+int desired_number_of_colors
+        Maximum number of colors to use in generating a library-supplied color
+        map (the actual number of colors is returned in a different field).
+        Default 256.  Ignored when the application supplies its own color map.
+
+boolean two_pass_quantize
+        If TRUE, an extra pass over the image is made to select a custom color
+        map for the image.  This usually looks a lot better than the one-size-
+        fits-all colormap that is used otherwise.  Default is TRUE.  Ignored
+        when the application supplies its own color map.
+
+J_DITHER_MODE dither_mode
+        Selects color dithering method.  Supported values are:
+                JDITHER_NONE    no dithering: fast, very low quality
+                JDITHER_ORDERED ordered dither: moderate speed and quality
+                JDITHER_FS      Floyd-Steinberg dither: slow, high quality
+        Default is JDITHER_FS.  (At present, ordered dither is implemented
+        only in the single-pass, standard-colormap case.  If you ask for
+        ordered dither when two_pass_quantize is TRUE or when you supply
+        an external color map, you'll get F-S dithering.)
+
+When quantize_colors is TRUE, the target color map is described by the next
+two fields.  colormap is set to NULL by jpeg_read_header().  The application
+can supply a color map by setting colormap non-NULL and setting
+actual_number_of_colors to the map size.  Otherwise, jpeg_start_decompress()
+selects a suitable color map and sets these two fields itself.
+Implementation restriction: at present, an externally supplied colormap is
+only accepted for 3-component output color spaces.
+
+JSAMPARRAY colormap
+        The color map, represented as a 2-D pixel array of out_color_components
+        rows and actual_number_of_colors columns.  Ignored if not quantizing.
+        CAUTION: if the JPEG library creates its own colormap, the storage
+        pointed to by this field is released by jpeg_finish_decompress().
+        Copy the colormap somewhere else first, if you want to save it.
+
+int actual_number_of_colors
+        The number of colors in the color map.
+
+Additional decompression parameters that the application may set include:
+
+J_DCT_METHOD dct_method
+        Selects the algorithm used for the DCT step.  Choices are:
+                JDCT_ISLOW: accurate integer method
+                JDCT_IFAST: less accurate integer method legacy feature
+                JDCT_FLOAT: floating-point method legacy feature
+                JDCT_DEFAULT: default method (normally JDCT_ISLOW)
+                JDCT_FASTEST: fastest method (normally JDCT_IFAST)
+        When the Independent JPEG Group's software was first released in 1991,
+        the decompression time for a 1-megapixel JPEG image on a mainstream PC
+        was measured in minutes.  Thus, JDCT_IFAST provided noticeable
+        performance benefits.  On modern CPUs running libjpeg-turbo, however,
+        the decompression time for a 1-megapixel JPEG image is measured in
+        milliseconds, and thus the performance benefits of JDCT_IFAST are much
+        less noticeable.  On modern x86/x86-64 CPUs that support AVX2
+        instructions, JDCT_IFAST and JDCT_ISLOW have similar performance.  On
+        other types of CPUs, JDCT_IFAST is generally about 5-15% faster than
+        JDCT_ISLOW.
+
+        If the JPEG image was compressed using a quality level of 85 or below,
+        then there should be little or no perceptible quality difference
+        between the two algorithms.  When decompressing images that were
+        compressed using quality levels above 85, however, the difference
+        between JDCT_IFAST and JDCT_ISLOW becomes more pronounced.  With images
+        compressed using quality=97, for instance, JDCT_IFAST incurs generally
+        about a 4-6 dB loss in PSNR relative to JDCT_ISLOW, but this can be
+        larger for some images.  If you can avoid it, do not use JDCT_IFAST
+        when decompressing images that were compressed using quality levels
+        above 97.  The algorithm often degenerates for such images and can
+        actually produce a more lossy output image than if the JPEG image had
+        been compressed using lower quality levels.
+
+        JDCT_FLOAT does not produce significantly more accurate results than
+        JDCT_ISLOW, and it is much slower.  JDCT_FLOAT may also give different
+        results on different machines due to varying roundoff behavior, whereas
+        the integer methods should give the same results on all machines.
+
+boolean do_fancy_upsampling
+        If TRUE, do careful upsampling of chroma components.  If FALSE,
+        a faster but sloppier method is used.  Default is TRUE.  The visual
+        impact of the sloppier method is often very small.
+
+boolean do_block_smoothing
+        If TRUE, interblock smoothing is applied in early stages of decoding
+        progressive JPEG files; if FALSE, not.  Default is TRUE.  Early
+        progression stages look "fuzzy" with smoothing, "blocky" without.
+        In any case, block smoothing ceases to be applied after the first few
+        AC coefficients are known to full accuracy, so it is relevant only
+        when using buffered-image mode for progressive images.
+
+boolean enable_1pass_quant
+boolean enable_external_quant
+boolean enable_2pass_quant
+        These are significant only in buffered-image mode, which is
+        described in its own section below.
+
+
+The output image dimensions are given by the following fields.  These are
+computed from the source image dimensions and the decompression parameters
+by jpeg_start_decompress().  You can also call jpeg_calc_output_dimensions()
+to obtain the values that will result from the current parameter settings.
+This can be useful if you are trying to pick a scaling ratio that will get
+close to a desired target size.  It's also important if you are using the
+JPEG library's memory manager to allocate output buffer space, because you
+are supposed to request such buffers *before* jpeg_start_decompress().
+
+JDIMENSION output_width         Actual dimensions of output image.
+JDIMENSION output_height
+int out_color_components        Number of color components in out_color_space.
+int output_components           Number of color components returned.
+int rec_outbuf_height           Recommended height of scanline buffer.
+
+When quantizing colors, output_components is 1, indicating a single color map
+index per pixel.  Otherwise it equals out_color_components.  The output arrays
+are required to be output_width * output_components JSAMPLEs wide.
+
+rec_outbuf_height is the recommended minimum height (in scanlines) of the
+buffer passed to jpeg_read_scanlines().  If the buffer is smaller, the
+library will still work, but time will be wasted due to unnecessary data
+copying.  In high-quality modes, rec_outbuf_height is always 1, but some
+faster, lower-quality modes set it to larger values (typically 2 to 4).
+If you are going to ask for a high-speed processing mode, you may as well
+go to the trouble of honoring rec_outbuf_height so as to avoid data copying.
+(An output buffer larger than rec_outbuf_height lines is OK, but won't
+provide any material speed improvement over that height.)
+
+
+Special color spaces
+--------------------
+
+The JPEG standard itself is "color blind" and doesn't specify any particular
+color space.  It is customary to convert color data to a luminance/chrominance
+color space before compressing, since this permits greater compression.  The
+existing de-facto JPEG file format standards specify YCbCr or grayscale data
+(JFIF), or grayscale, RGB, YCbCr, CMYK, or YCCK (Adobe).  For special
+applications such as multispectral images, other color spaces can be used,
+but it must be understood that such files will be unportable.
+
+The JPEG library can handle the most common colorspace conversions (namely
+RGB <=> YCbCr and CMYK <=> YCCK).  It can also deal with data of an unknown
+color space, passing it through without conversion.  If you deal extensively
+with an unusual color space, you can easily extend the library to understand
+additional color spaces and perform appropriate conversions.
+
+For compression, the source data's color space is specified by field
+in_color_space.  This is transformed to the JPEG file's color space given
+by jpeg_color_space.  jpeg_set_defaults() chooses a reasonable JPEG color
+space depending on in_color_space, but you can override this by calling
+jpeg_set_colorspace().  Of course you must select a supported transformation.
+jccolor.c currently supports the following transformations:
+        RGB => YCbCr
+        RGB => GRAYSCALE
+        YCbCr => GRAYSCALE
+        CMYK => YCCK
+plus the null transforms: GRAYSCALE => GRAYSCALE, RGB => RGB,
+YCbCr => YCbCr, CMYK => CMYK, YCCK => YCCK, and UNKNOWN => UNKNOWN.
+
+The de-facto file format standards (JFIF and Adobe) specify APPn markers that
+indicate the color space of the JPEG file.  It is important to ensure that
+these are written correctly, or omitted if the JPEG file's color space is not
+one of the ones supported by the de-facto standards.  jpeg_set_colorspace()
+will set the compression parameters to include or omit the APPn markers
+properly, so long as it is told the truth about the JPEG color space.
+For example, if you are writing some random 3-component color space without
+conversion, don't try to fake out the library by setting in_color_space and
+jpeg_color_space to JCS_YCbCr; use JCS_UNKNOWN.  You may want to write an
+APPn marker of your own devising to identify the colorspace --- see "Special
+markers", below.
+
+When told that the color space is UNKNOWN, the library will default to using
+luminance-quality compression parameters for all color components.  You may
+well want to change these parameters.  See the source code for
+jpeg_set_colorspace(), in jcparam.c, for details.
+
+For decompression, the JPEG file's color space is given in jpeg_color_space,
+and this is transformed to the output color space out_color_space.
+jpeg_read_header's setting of jpeg_color_space can be relied on if the file
+conforms to JFIF or Adobe conventions, but otherwise it is no better than a
+guess.  If you know the JPEG file's color space for certain, you can override
+jpeg_read_header's guess by setting jpeg_color_space.  jpeg_read_header also
+selects a default output color space based on (its guess of) jpeg_color_space;
+set out_color_space to override this.  Again, you must select a supported
+transformation.  jdcolor.c currently supports
+        YCbCr => RGB
+        YCbCr => GRAYSCALE
+        RGB => GRAYSCALE
+        GRAYSCALE => RGB
+        YCCK => CMYK
+as well as the null transforms.  (Since GRAYSCALE=>RGB is provided, an
+application can force grayscale JPEGs to look like color JPEGs if it only
+wants to handle one case.)
+
+The two-pass color quantizer, jquant2.c, is specialized to handle RGB data
+(it weights distances appropriately for RGB colors).  You'll need to modify
+the code if you want to use it for non-RGB output color spaces.  Note that
+jquant2.c is used to map to an application-supplied colormap as well as for
+the normal two-pass colormap selection process.
+
+CAUTION: it appears that Adobe Photoshop writes inverted data in CMYK JPEG
+files: 0 represents 100% ink coverage, rather than 0% ink as you'd expect.
+This is arguably a bug in Photoshop, but if you need to work with Photoshop
+CMYK files, you will have to deal with it in your application.  We cannot
+"fix" this in the library by inverting the data during the CMYK<=>YCCK
+transform, because that would break other applications, notably Ghostscript.
+Photoshop versions prior to 3.0 write EPS files containing JPEG-encoded CMYK
+data in the same inverted-YCCK representation used in bare JPEG files, but
+the surrounding PostScript code performs an inversion using the PS image
+operator.  I am told that Photoshop 3.0 will write uninverted YCCK in
+EPS/JPEG files, and will omit the PS-level inversion.  (But the data
+polarity used in bare JPEG files will not change in 3.0.)  In either case,
+the JPEG library must not invert the data itself, or else Ghostscript would
+read these EPS files incorrectly.
+
+
+Error handling
+--------------
+
+When the default error handler is used, any error detected inside the JPEG
+routines will cause a message to be printed on stderr, followed by exit().
+You can supply your own error handling routines to override this behavior
+and to control the treatment of nonfatal warnings and trace/debug messages.
+The file example.txt illustrates the most common case, which is to have the
+application regain control after an error rather than exiting.
+
+The JPEG library never writes any message directly; it always goes through
+the error handling routines.  Three classes of messages are recognized:
+  * Fatal errors: the library cannot continue.
+  * Warnings: the library can continue, but the data is corrupt, and a
+    damaged output image is likely to result.
+  * Trace/informational messages.  These come with a trace level indicating
+    the importance of the message; you can control the verbosity of the
+    program by adjusting the maximum trace level that will be displayed.
+
+You may, if you wish, simply replace the entire JPEG error handling module
+(jerror.c) with your own code.  However, you can avoid code duplication by
+only replacing some of the routines depending on the behavior you need.
+This is accomplished by calling jpeg_std_error() as usual, but then overriding
+some of the method pointers in the jpeg_error_mgr struct, as illustrated by
+example.txt.
+
+All of the error handling routines will receive a pointer to the JPEG object
+(a j_common_ptr which points to either a jpeg_compress_struct or a
+jpeg_decompress_struct; if you need to tell which, test the is_decompressor
+field).  This struct includes a pointer to the error manager struct in its
+"err" field.  Frequently, custom error handler routines will need to access
+additional data which is not known to the JPEG library or the standard error
+handler.  The most convenient way to do this is to embed either the JPEG
+object or the jpeg_error_mgr struct in a larger structure that contains
+additional fields; then casting the passed pointer provides access to the
+additional fields.  Again, see example.txt for one way to do it.  (Beginning
+with IJG version 6b, there is also a void pointer "client_data" in each
+JPEG object, which the application can also use to find related data.
+The library does not touch client_data at all.)
+
+The individual methods that you might wish to override are:
+
+error_exit (j_common_ptr cinfo)
+        Receives control for a fatal error.  Information sufficient to
+        generate the error message has been stored in cinfo->err; call
+        output_message to display it.  Control must NOT return to the caller;
+        generally this routine will exit() or longjmp() somewhere.
+        Typically you would override this routine to get rid of the exit()
+        default behavior.  Note that if you continue processing, you should
+        clean up the JPEG object with jpeg_abort() or jpeg_destroy().
+
+output_message (j_common_ptr cinfo)
+        Actual output of any JPEG message.  Override this to send messages
+        somewhere other than stderr.  Note that this method does not know
+        how to generate a message, only where to send it.
+
+format_message (j_common_ptr cinfo, char *buffer)
+        Constructs a readable error message string based on the error info
+        stored in cinfo->err.  This method is called by output_message.  Few
+        applications should need to override this method.  One possible
+        reason for doing so is to implement dynamic switching of error message
+        language.
+
+emit_message (j_common_ptr cinfo, int msg_level)
+        Decide whether or not to emit a warning or trace message; if so,
+        calls output_message.  The main reason for overriding this method
+        would be to abort on warnings.  msg_level is -1 for warnings,
+        0 and up for trace messages.
+
+Only error_exit() and emit_message() are called from the rest of the JPEG
+library; the other two are internal to the error handler.
+
+The actual message texts are stored in an array of strings which is pointed to
+by the field err->jpeg_message_table.  The messages are numbered from 0 to
+err->last_jpeg_message, and it is these code numbers that are used in the
+JPEG library code.  You could replace the message texts (for instance, with
+messages in French or German) by changing the message table pointer.  See
+jerror.h for the default texts.  CAUTION: this table will almost certainly
+change or grow from one library version to the next.
+
+It may be useful for an application to add its own message texts that are
+handled by the same mechanism.  The error handler supports a second "add-on"
+message table for this purpose.  To define an addon table, set the pointer
+err->addon_message_table and the message numbers err->first_addon_message and
+err->last_addon_message.  If you number the addon messages beginning at 1000
+or so, you won't have to worry about conflicts with the library's built-in
+messages.  See the sample applications cjpeg/djpeg for an example of using
+addon messages (the addon messages are defined in cderror.h).
+
+Actual invocation of the error handler is done via macros defined in jerror.h:
+        ERREXITn(...)   for fatal errors
+        WARNMSn(...)    for corrupt-data warnings
+        TRACEMSn(...)   for trace and informational messages.
+These macros store the message code and any additional parameters into the
+error handler struct, then invoke the error_exit() or emit_message() method.
+The variants of each macro are for varying numbers of additional parameters.
+The additional parameters are inserted into the generated message using
+standard printf() format codes.
+
+See jerror.h and jerror.c for further details.
+
+
+Compressed data handling (source and destination managers)
+----------------------------------------------------------
+
+The JPEG compression library sends its compressed data to a "destination
+manager" module.  The default destination manager just writes the data to a
+memory buffer or to a stdio stream, but you can provide your own manager to
+do something else.  Similarly, the decompression library calls a "source
+manager" to obtain the compressed data; you can provide your own source
+manager if you want the data to come from somewhere other than a memory
+buffer or a stdio stream.
+
+In both cases, compressed data is processed a bufferload at a time: the
+destination or source manager provides a work buffer, and the library invokes
+the manager only when the buffer is filled or emptied.  (You could define a
+one-character buffer to force the manager to be invoked for each byte, but
+that would be rather inefficient.)  The buffer's size and location are
+controlled by the manager, not by the library.  For example, the memory
+source manager just makes the buffer pointer and length point to the original
+data in memory.  In this case the buffer-reload procedure will be invoked
+only if the decompressor ran off the end of the datastream, which would
+indicate an erroneous datastream.
+
+The work buffer is defined as an array of datatype JOCTET, which is generally
+"char" or "unsigned char".  On a machine where char is not exactly 8 bits
+wide, you must define JOCTET as a wider data type and then modify the data
+source and destination modules to transcribe the work arrays into 8-bit units
+on external storage.
+
+A data destination manager struct contains a pointer and count defining the
+next byte to write in the work buffer and the remaining free space:
+
+        JOCTET *next_output_byte;   /* => next byte to write in buffer */
+        size_t free_in_buffer;      /* # of byte spaces remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is filled.  The manager's empty_output_buffer method must reset the pointer
+and count.  The manager is expected to remember the buffer's starting address
+and total size in private fields not visible to the library.
+
+A data destination manager provides three methods:
+
+init_destination (j_compress_ptr cinfo)
+        Initialize destination.  This is called by jpeg_start_compress()
+        before any data is actually written.  It must initialize
+        next_output_byte and free_in_buffer.  free_in_buffer must be
+        initialized to a positive value.
+
+empty_output_buffer (j_compress_ptr cinfo)
+        This is called whenever the buffer has filled (free_in_buffer
+        reaches zero).  In typical applications, it should write out the
+        *entire* buffer (use the saved start address and buffer length;
+        ignore the current state of next_output_byte and free_in_buffer).
+        Then reset the pointer & count to the start of the buffer, and
+        return TRUE indicating that the buffer has been dumped.
+        free_in_buffer must be set to a positive value when TRUE is
+        returned.  A FALSE return should only be used when I/O suspension is
+        desired (this operating mode is discussed in the next section).
+
+term_destination (j_compress_ptr cinfo)
+        Terminate destination --- called by jpeg_finish_compress() after all
+        data has been written.  In most applications, this must flush any
+        data remaining in the buffer.  Use either next_output_byte or
+        free_in_buffer to determine how much data is in the buffer.
+
+term_destination() is NOT called by jpeg_abort() or jpeg_destroy().  If you
+want the destination manager to be cleaned up during an abort, you must do it
+yourself.
+
+You will also need code to create a jpeg_destination_mgr struct, fill in its
+method pointers, and insert a pointer to the struct into the "dest" field of
+the JPEG compression object.  This can be done in-line in your setup code if
+you like, but it's probably cleaner to provide a separate routine similar to
+the jpeg_stdio_dest() or jpeg_mem_dest() routines of the supplied destination
+managers.
+
+Decompression source managers follow a parallel design, but with some
+additional frammishes.  The source manager struct contains a pointer and count
+defining the next byte to read from the work buffer and the number of bytes
+remaining:
+
+        const JOCTET *next_input_byte;  /* => next byte to read from buffer */
+        size_t bytes_in_buffer;         /* # of bytes remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is emptied.  The manager's fill_input_buffer method must reset the pointer and
+count.  In most applications, the manager must remember the buffer's starting
+address and total size in private fields not visible to the library.
+
+A data source manager provides five methods:
+
+init_source (j_decompress_ptr cinfo)
+        Initialize source.  This is called by jpeg_read_header() before any
+        data is actually read.  Unlike init_destination(), it may leave
+        bytes_in_buffer set to 0 (in which case a fill_input_buffer() call
+        will occur immediately).
+
+fill_input_buffer (j_decompress_ptr cinfo)
+        This is called whenever bytes_in_buffer has reached zero and more
+        data is wanted.  In typical applications, it should read fresh data
+        into the buffer (ignoring the current state of next_input_byte and
+        bytes_in_buffer), reset the pointer & count to the start of the
+        buffer, and return TRUE indicating that the buffer has been reloaded.
+        It is not necessary to fill the buffer entirely, only to obtain at
+        least one more byte.  bytes_in_buffer MUST be set to a positive value
+        if TRUE is returned.  A FALSE return should only be used when I/O
+        suspension is desired (this mode is discussed in the next section).
+
+skip_input_data (j_decompress_ptr cinfo, long num_bytes)
+        Skip num_bytes worth of data.  The buffer pointer and count should
+        be advanced over num_bytes input bytes, refilling the buffer as
+        needed.  This is used to skip over a potentially large amount of
+        uninteresting data (such as an APPn marker).  In some applications
+        it may be possible to optimize away the reading of the skipped data,
+        but it's not clear that being smart is worth much trouble; large
+        skips are uncommon.  bytes_in_buffer may be zero on return.
+        A zero or negative skip count should be treated as a no-op.
+
+resync_to_restart (j_decompress_ptr cinfo, int desired)
+        This routine is called only when the decompressor has failed to find
+        a restart (RSTn) marker where one is expected.  Its mission is to
+        find a suitable point for resuming decompression.  For most
+        applications, we recommend that you just use the default resync
+        procedure, jpeg_resync_to_restart().  However, if you are able to back
+        up in the input data stream, or if you have a-priori knowledge about
+        the likely location of restart markers, you may be able to do better.
+        Read the read_restart_marker() and jpeg_resync_to_restart() routines
+        in jdmarker.c if you think you'd like to implement your own resync
+        procedure.
+
+term_source (j_decompress_ptr cinfo)
+        Terminate source --- called by jpeg_finish_decompress() after all
+        data has been read.  Often a no-op.
+
+For both fill_input_buffer() and skip_input_data(), there is no such thing
+as an EOF return.  If the end of the file has been reached, the routine has
+a choice of exiting via ERREXIT() or inserting fake data into the buffer.
+In most cases, generating a warning message and inserting a fake EOI marker
+is the best course of action --- this will allow the decompressor to output
+however much of the image is there.  In pathological cases, the decompressor
+may swallow the EOI and again demand data ... just keep feeding it fake EOIs.
+jdatasrc.c illustrates the recommended error recovery behavior.
+
+term_source() is NOT called by jpeg_abort() or jpeg_destroy().  If you want
+the source manager to be cleaned up during an abort, you must do it yourself.
+
+You will also need code to create a jpeg_source_mgr struct, fill in its method
+pointers, and insert a pointer to the struct into the "src" field of the JPEG
+decompression object.  This can be done in-line in your setup code if you
+like, but it's probably cleaner to provide a separate routine similar to the
+jpeg_stdio_src() or jpeg_mem_src() routines of the supplied source managers.
+
+For more information, consult the memory and stdio source and destination
+managers in jdatasrc.c and jdatadst.c.
+
+
+I/O suspension
+--------------
+
+Some applications need to use the JPEG library as an incremental memory-to-
+memory filter: when the compressed data buffer is filled or emptied, they want
+control to return to the outer loop, rather than expecting that the buffer can
+be emptied or reloaded within the data source/destination manager subroutine.
+The library supports this need by providing an "I/O suspension" mode, which we
+describe in this section.
+
+The I/O suspension mode is not a panacea: nothing is guaranteed about the
+maximum amount of time spent in any one call to the library, so it will not
+eliminate response-time problems in single-threaded applications.  If you
+need guaranteed response time, we suggest you "bite the bullet" and implement
+a real multi-tasking capability.
+
+To use I/O suspension, cooperation is needed between the calling application
+and the data source or destination manager; you will always need a custom
+source/destination manager.  (Please read the previous section if you haven't
+already.)  The basic idea is that the empty_output_buffer() or
+fill_input_buffer() routine is a no-op, merely returning FALSE to indicate
+that it has done nothing.  Upon seeing this, the JPEG library suspends
+operation and returns to its caller.  The surrounding application is
+responsible for emptying or refilling the work buffer before calling the
+JPEG library again.
+
+Compression suspension:
+
+For compression suspension, use an empty_output_buffer() routine that returns
+FALSE; typically it will not do anything else.  This will cause the
+compressor to return to the caller of jpeg_write_scanlines(), with the return
+value indicating that not all the supplied scanlines have been accepted.
+The application must make more room in the output buffer, adjust the output
+buffer pointer/count appropriately, and then call jpeg_write_scanlines()
+again, pointing to the first unconsumed scanline.
+
+When forced to suspend, the compressor will backtrack to a convenient stopping
+point (usually the start of the current MCU); it will regenerate some output
+data when restarted.  Therefore, although empty_output_buffer() is only
+called when the buffer is filled, you should NOT write out the entire buffer
+after a suspension.  Write only the data up to the current position of
+next_output_byte/free_in_buffer.  The data beyond that point will be
+regenerated after resumption.
+
+Because of the backtracking behavior, a good-size output buffer is essential
+for efficiency; you don't want the compressor to suspend often.  (In fact, an
+overly small buffer could lead to infinite looping, if a single MCU required
+more data than would fit in the buffer.)  We recommend a buffer of at least
+several Kbytes.  You may want to insert explicit code to ensure that you don't
+call jpeg_write_scanlines() unless there is a reasonable amount of space in
+the output buffer; in other words, flush the buffer before trying to compress
+more data.
+
+The compressor does not allow suspension while it is trying to write JPEG
+markers at the beginning and end of the file.  This means that:
+  * At the beginning of a compression operation, there must be enough free
+    space in the output buffer to hold the header markers (typically 600 or
+    so bytes).  The recommended buffer size is bigger than this anyway, so
+    this is not a problem as long as you start with an empty buffer.  However,
+    this restriction might catch you if you insert large special markers, such
+    as a JFIF thumbnail image, without flushing the buffer afterwards.
+  * When you call jpeg_finish_compress(), there must be enough space in the
+    output buffer to emit any buffered data and the final EOI marker.  In the
+    current implementation, half a dozen bytes should suffice for this, but
+    for safety's sake we recommend ensuring that at least 100 bytes are free
+    before calling jpeg_finish_compress().
+
+A more significant restriction is that jpeg_finish_compress() cannot suspend.
+This means you cannot use suspension with multi-pass operating modes, namely
+Huffman code optimization and multiple-scan output.  Those modes write the
+whole file during jpeg_finish_compress(), which will certainly result in
+buffer overrun.  (Note that this restriction applies only to compression,
+not decompression.  The decompressor supports input suspension in all of its
+operating modes.)
+
+Decompression suspension:
+
+For decompression suspension, use a fill_input_buffer() routine that simply
+returns FALSE (except perhaps during error recovery, as discussed below).
+This will cause the decompressor to return to its caller with an indication
+that suspension has occurred.  This can happen at four places:
+  * jpeg_read_header(): will return JPEG_SUSPENDED.
+  * jpeg_start_decompress(): will return FALSE, rather than its usual TRUE.
+  * jpeg_read_scanlines(): will return the number of scanlines already
+        completed (possibly 0).
+  * jpeg_finish_decompress(): will return FALSE, rather than its usual TRUE.
+The surrounding application must recognize these cases, load more data into
+the input buffer, and repeat the call.  In the case of jpeg_read_scanlines(),
+increment the passed pointers past any scanlines successfully read.
+
+Just as with compression, the decompressor will typically backtrack to a
+convenient restart point before suspending.  When fill_input_buffer() is
+called, next_input_byte/bytes_in_buffer point to the current restart point,
+which is where the decompressor will backtrack to if FALSE is returned.
+The data beyond that position must NOT be discarded if you suspend; it needs
+to be re-read upon resumption.  In most implementations, you'll need to shift
+this data down to the start of your work buffer and then load more data after
+it.  Again, this behavior means that a several-Kbyte work buffer is essential
+for decent performance; furthermore, you should load a reasonable amount of
+new data before resuming decompression.  (If you loaded, say, only one new
+byte each time around, you could waste a LOT of cycles.)
+
+The skip_input_data() source manager routine requires special care in a
+suspension scenario.  This routine is NOT granted the ability to suspend the
+decompressor; it can decrement bytes_in_buffer to zero, but no more.  If the
+requested skip distance exceeds the amount of data currently in the input
+buffer, then skip_input_data() must set bytes_in_buffer to zero and record the
+additional skip distance somewhere else.  The decompressor will immediately
+call fill_input_buffer(), which should return FALSE, which will cause a
+suspension return.  The surrounding application must then arrange to discard
+the recorded number of bytes before it resumes loading the input buffer.
+(Yes, this design is rather baroque, but it avoids complexity in the far more
+common case where a non-suspending source manager is used.)
+
+If the input data has been exhausted, we recommend that you emit a warning
+and insert dummy EOI markers just as a non-suspending data source manager
+would do.  This can be handled either in the surrounding application logic or
+within fill_input_buffer(); the latter is probably more efficient.  If
+fill_input_buffer() knows that no more data is available, it can set the
+pointer/count to point to a dummy EOI marker and then return TRUE just as
+though it had read more data in a non-suspending situation.
+
+The decompressor does not attempt to suspend within standard JPEG markers;
+instead it will backtrack to the start of the marker and reprocess the whole
+marker next time.  Hence the input buffer must be large enough to hold the
+longest standard marker in the file.  Standard JPEG markers should normally
+not exceed a few hundred bytes each (DHT tables are typically the longest).
+We recommend at least a 2K buffer for performance reasons, which is much
+larger than any correct marker is likely to be.  For robustness against
+damaged marker length counts, you may wish to insert a test in your
+application for the case that the input buffer is completely full and yet
+the decoder has suspended without consuming any data --- otherwise, if this
+situation did occur, it would lead to an endless loop.  (The library can't
+provide this test since it has no idea whether "the buffer is full", or
+even whether there is a fixed-size input buffer.)
+
+The input buffer would need to be 64K to allow for arbitrary COM or APPn
+markers, but these are handled specially: they are either saved into allocated
+memory, or skipped over by calling skip_input_data().  In the former case,
+suspension is handled correctly, and in the latter case, the problem of
+buffer overrun is placed on skip_input_data's shoulders, as explained above.
+Note that if you provide your own marker handling routine for large markers,
+you should consider how to deal with buffer overflow.
+
+Multiple-buffer management:
+
+In some applications it is desirable to store the compressed data in a linked
+list of buffer areas, so as to avoid data copying.  This can be handled by
+having empty_output_buffer() or fill_input_buffer() set the pointer and count
+to reference the next available buffer; FALSE is returned only if no more
+buffers are available.  Although seemingly straightforward, there is a
+pitfall in this approach: the backtrack that occurs when FALSE is returned
+could back up into an earlier buffer.  For example, when fill_input_buffer()
+is called, the current pointer & count indicate the backtrack restart point.
+Since fill_input_buffer() will set the pointer and count to refer to a new
+buffer, the restart position must be saved somewhere else.  Suppose a second
+call to fill_input_buffer() occurs in the same library call, and no
+additional input data is available, so fill_input_buffer must return FALSE.
+If the JPEG library has not moved the pointer/count forward in the current
+buffer, then *the correct restart point is the saved position in the prior
+buffer*.  Prior buffers may be discarded only after the library establishes
+a restart point within a later buffer.  Similar remarks apply for output into
+a chain of buffers.
+
+The library will never attempt to backtrack over a skip_input_data() call,
+so any skipped data can be permanently discarded.  You still have to deal
+with the case of skipping not-yet-received data, however.
+
+It's much simpler to use only a single buffer; when fill_input_buffer() is
+called, move any unconsumed data (beyond the current pointer/count) down to
+the beginning of this buffer and then load new data into the remaining buffer
+space.  This approach requires a little more data copying but is far easier
+to get right.
+
+
+Progressive JPEG support
+------------------------
+
+Progressive JPEG rearranges the stored data into a series of scans of
+increasing quality.  In situations where a JPEG file is transmitted across a
+slow communications link, a decoder can generate a low-quality image very
+quickly from the first scan, then gradually improve the displayed quality as
+more scans are received.  The final image after all scans are complete is
+identical to that of a regular (sequential) JPEG file of the same quality
+setting.  Progressive JPEG files are often slightly smaller than equivalent
+sequential JPEG files, but the possibility of incremental display is the main
+reason for using progressive JPEG.
+
+The IJG encoder library generates progressive JPEG files when given a
+suitable "scan script" defining how to divide the data into scans.
+Creation of progressive JPEG files is otherwise transparent to the encoder.
+Progressive JPEG files can also be read transparently by the decoder library.
+If the decoding application simply uses the library as defined above, it
+will receive a final decoded image without any indication that the file was
+progressive.  Of course, this approach does not allow incremental display.
+To perform incremental display, an application needs to use the decoder
+library's "buffered-image" mode, in which it receives a decoded image
+multiple times.
+
+Each displayed scan requires about as much work to decode as a full JPEG
+image of the same size, so the decoder must be fairly fast in relation to the
+data transmission rate in order to make incremental display useful.  However,
+it is possible to skip displaying the image and simply add the incoming bits
+to the decoder's coefficient buffer.  This is fast because only Huffman
+decoding need be done, not IDCT, upsampling, colorspace conversion, etc.
+The IJG decoder library allows the application to switch dynamically between
+displaying the image and simply absorbing the incoming bits.  A properly
+coded application can automatically adapt the number of display passes to
+suit the time available as the image is received.  Also, a final
+higher-quality display cycle can be performed from the buffered data after
+the end of the file is reached.
+
+Progressive compression:
+
+To create a progressive JPEG file (or a multiple-scan sequential JPEG file),
+set the scan_info cinfo field to point to an array of scan descriptors, and
+perform compression as usual.  Instead of constructing your own scan list,
+you can call the jpeg_simple_progression() helper routine to create a
+recommended progression sequence; this method should be used by all
+applications that don't want to get involved in the nitty-gritty of
+progressive scan sequence design.  (If you want to provide user control of
+scan sequences, you may wish to borrow the scan script reading code found
+in rdswitch.c, so that you can read scan script files just like cjpeg's.)
+When scan_info is not NULL, the compression library will store DCT'd data
+into a buffer array as jpeg_write_scanlines() is called, and will emit all
+the requested scans during jpeg_finish_compress().  This implies that
+multiple-scan output cannot be created with a suspending data destination
+manager, since jpeg_finish_compress() does not support suspension.  We
+should also note that the compressor currently forces Huffman optimization
+mode when creating a progressive JPEG file, because the default Huffman
+tables are unsuitable for progressive files.
+
+Progressive decompression:
+
+When buffered-image mode is not used, the decoder library will read all of
+a multi-scan file during jpeg_start_decompress(), so that it can provide a
+final decoded image.  (Here "multi-scan" means either progressive or
+multi-scan sequential.)  This makes multi-scan files transparent to the
+decoding application.  However, existing applications that used suspending
+input with version 5 of the IJG library will need to be modified to check
+for a suspension return from jpeg_start_decompress().
+
+To perform incremental display, an application must use the library's
+buffered-image mode.  This is described in the next section.
+
+
+Buffered-image mode
+-------------------
+
+In buffered-image mode, the library stores the partially decoded image in a
+coefficient buffer, from which it can be read out as many times as desired.
+This mode is typically used for incremental display of progressive JPEG files,
+but it can be used with any JPEG file.  Each scan of a progressive JPEG file
+adds more data (more detail) to the buffered image.  The application can
+display in lockstep with the source file (one display pass per input scan),
+or it can allow input processing to outrun display processing.  By making
+input and display processing run independently, it is possible for the
+application to adapt progressive display to a wide range of data transmission
+rates.
+
+The basic control flow for buffered-image decoding is
+
+        jpeg_create_decompress()
+        set data source
+        jpeg_read_header()
+        set overall decompression parameters
+        cinfo.buffered_image = TRUE;    /* select buffered-image mode */
+        jpeg_start_decompress()
+        for (each output pass) {
+            adjust output decompression parameters if required
+            jpeg_start_output()         /* start a new output pass */
+            for (all scanlines in image) {
+                jpeg_read_scanlines()
+                display scanlines
+            }
+            jpeg_finish_output()        /* terminate output pass */
+        }
+        jpeg_finish_decompress()
+        jpeg_destroy_decompress()
+
+This differs from ordinary unbuffered decoding in that there is an additional
+level of looping.  The application can choose how many output passes to make
+and how to display each pass.
+
+The simplest approach to displaying progressive images is to do one display
+pass for each scan appearing in the input file.  In this case the outer loop
+condition is typically
+        while (!jpeg_input_complete(&cinfo))
+and the start-output call should read
+        jpeg_start_output(&cinfo, cinfo.input_scan_number);
+The second parameter to jpeg_start_output() indicates which scan of the input
+file is to be displayed; the scans are numbered starting at 1 for this
+purpose.  (You can use a loop counter starting at 1 if you like, but using
+the library's input scan counter is easier.)  The library automatically reads
+data as necessary to complete each requested scan, and jpeg_finish_output()
+advances to the next scan or end-of-image marker (hence input_scan_number
+will be incremented by the time control arrives back at jpeg_start_output()).
+With this technique, data is read from the input file only as needed, and
+input and output processing run in lockstep.
+
+After reading the final scan and reaching the end of the input file, the
+buffered image remains available; it can be read additional times by
+repeating the jpeg_start_output()/jpeg_read_scanlines()/jpeg_finish_output()
+sequence.  For example, a useful technique is to use fast one-pass color
+quantization for display passes made while the image is arriving, followed by
+a final display pass using two-pass quantization for highest quality.  This
+is done by changing the library parameters before the final output pass.
+Changing parameters between passes is discussed in detail below.
+
+In general the last scan of a progressive file cannot be recognized as such
+until after it is read, so a post-input display pass is the best approach if
+you want special processing in the final pass.
+
+When done with the image, be sure to call jpeg_finish_decompress() to release
+the buffered image (or just use jpeg_destroy_decompress()).
+
+If input data arrives faster than it can be displayed, the application can
+cause the library to decode input data in advance of what's needed to produce
+output.  This is done by calling the routine jpeg_consume_input().
+The return value is one of the following:
+        JPEG_REACHED_SOS:    reached an SOS marker (the start of a new scan)
+        JPEG_REACHED_EOI:    reached the EOI marker (end of image)
+        JPEG_ROW_COMPLETED:  completed reading one MCU row of compressed data
+        JPEG_SCAN_COMPLETED: completed reading last MCU row of current scan
+        JPEG_SUSPENDED:      suspended before completing any of the above
+(JPEG_SUSPENDED can occur only if a suspending data source is used.)  This
+routine can be called at any time after initializing the JPEG object.  It
+reads some additional data and returns when one of the indicated significant
+events occurs.  (If called after the EOI marker is reached, it will
+immediately return JPEG_REACHED_EOI without attempting to read more data.)
+
+The library's output processing will automatically call jpeg_consume_input()
+whenever the output processing overtakes the input; thus, simple lockstep
+display requires no direct calls to jpeg_consume_input().  But by adding
+calls to jpeg_consume_input(), you can absorb data in advance of what is
+being displayed.  This has two benefits:
+  * You can limit buildup of unprocessed data in your input buffer.
+  * You can eliminate extra display passes by paying attention to the
+    state of the library's input processing.
+
+The first of these benefits only requires interspersing calls to
+jpeg_consume_input() with your display operations and any other processing
+you may be doing.  To avoid wasting cycles due to backtracking, it's best to
+call jpeg_consume_input() only after a hundred or so new bytes have arrived.
+This is discussed further under "I/O suspension", above.  (Note: the JPEG
+library currently is not thread-safe.  You must not call jpeg_consume_input()
+from one thread of control if a different library routine is working on the
+same JPEG object in another thread.)
+
+When input arrives fast enough that more than one new scan is available
+before you start a new output pass, you may as well skip the output pass
+corresponding to the completed scan.  This occurs for free if you pass
+cinfo.input_scan_number as the target scan number to jpeg_start_output().
+The input_scan_number field is simply the index of the scan currently being
+consumed by the input processor.  You can ensure that this is up-to-date by
+emptying the input buffer just before calling jpeg_start_output(): call
+jpeg_consume_input() repeatedly until it returns JPEG_SUSPENDED or
+JPEG_REACHED_EOI.
+
+The target scan number passed to jpeg_start_output() is saved in the
+cinfo.output_scan_number field.  The library's output processing calls
+jpeg_consume_input() whenever the current input scan number and row within
+that scan is less than or equal to the current output scan number and row.
+Thus, input processing can "get ahead" of the output processing but is not
+allowed to "fall behind".  You can achieve several different effects by
+manipulating this interlock rule.  For example, if you pass a target scan
+number greater than the current input scan number, the output processor will
+wait until that scan starts to arrive before producing any output.  (To avoid
+an infinite loop, the target scan number is automatically reset to the last
+scan number when the end of image is reached.  Thus, if you specify a large
+target scan number, the library will just absorb the entire input file and
+then perform an output pass.  This is effectively the same as what
+jpeg_start_decompress() does when you don't select buffered-image mode.)
+When you pass a target scan number equal to the current input scan number,
+the image is displayed no faster than the current input scan arrives.  The
+final possibility is to pass a target scan number less than the current input
+scan number; this disables the input/output interlock and causes the output
+processor to simply display whatever it finds in the image buffer, without
+waiting for input.  (However, the library will not accept a target scan
+number less than one, so you can't avoid waiting for the first scan.)
+
+When data is arriving faster than the output display processing can advance
+through the image, jpeg_consume_input() will store data into the buffered
+image beyond the point at which the output processing is reading data out
+again.  If the input arrives fast enough, it may "wrap around" the buffer to
+the point where the input is more than one whole scan ahead of the output.
+If the output processing simply proceeds through its display pass without
+paying attention to the input, the effect seen on-screen is that the lower
+part of the image is one or more scans better in quality than the upper part.
+Then, when the next output scan is started, you have a choice of what target
+scan number to use.  The recommended choice is to use the current input scan
+number at that time, which implies that you've skipped the output scans
+corresponding to the input scans that were completed while you processed the
+previous output scan.  In this way, the decoder automatically adapts its
+speed to the arriving data, by skipping output scans as necessary to keep up
+with the arriving data.
+
+When using this strategy, you'll want to be sure that you perform a final
+output pass after receiving all the data; otherwise your last display may not
+be full quality across the whole screen.  So the right outer loop logic is
+something like this:
+        do {
+            absorb any waiting input by calling jpeg_consume_input()
+            final_pass = jpeg_input_complete(&cinfo);
+            adjust output decompression parameters if required
+            jpeg_start_output(&cinfo, cinfo.input_scan_number);
+            ...
+            jpeg_finish_output()
+        } while (!final_pass);
+rather than quitting as soon as jpeg_input_complete() returns TRUE.  This
+arrangement makes it simple to use higher-quality decoding parameters
+for the final pass.  But if you don't want to use special parameters for
+the final pass, the right loop logic is like this:
+        for (;;) {
+            absorb any waiting input by calling jpeg_consume_input()
+            jpeg_start_output(&cinfo, cinfo.input_scan_number);
+            ...
+            jpeg_finish_output()
+            if (jpeg_input_complete(&cinfo) &&
+                cinfo.input_scan_number == cinfo.output_scan_number)
+              break;
+        }
+In this case you don't need to know in advance whether an output pass is to
+be the last one, so it's not necessary to have reached EOF before starting
+the final output pass; rather, what you want to test is whether the output
+pass was performed in sync with the final input scan.  This form of the loop
+will avoid an extra output pass whenever the decoder is able (or nearly able)
+to keep up with the incoming data.
+
+When the data transmission speed is high, you might begin a display pass,
+then find that much or all of the file has arrived before you can complete
+the pass.  (You can detect this by noting the JPEG_REACHED_EOI return code
+from jpeg_consume_input(), or equivalently by testing jpeg_input_complete().)
+In this situation you may wish to abort the current display pass and start a
+new one using the newly arrived information.  To do so, just call
+jpeg_finish_output() and then start a new pass with jpeg_start_output().
+
+A variant strategy is to abort and restart display if more than one complete
+scan arrives during an output pass; this can be detected by noting
+JPEG_REACHED_SOS returns and/or examining cinfo.input_scan_number.  This
+idea should be employed with caution, however, since the display process
+might never get to the bottom of the image before being aborted, resulting
+in the lower part of the screen being several passes worse than the upper.
+In most cases it's probably best to abort an output pass only if the whole
+file has arrived and you want to begin the final output pass immediately.
+
+When receiving data across a communication link, we recommend always using
+the current input scan number for the output target scan number; if a
+higher-quality final pass is to be done, it should be started (aborting any
+incomplete output pass) as soon as the end of file is received.  However,
+many other strategies are possible.  For example, the application can examine
+the parameters of the current input scan and decide whether to display it or
+not.  If the scan contains only chroma data, one might choose not to use it
+as the target scan, expecting that the scan will be small and will arrive
+quickly.  To skip to the next scan, call jpeg_consume_input() until it
+returns JPEG_REACHED_SOS or JPEG_REACHED_EOI.  Or just use the next higher
+number as the target scan for jpeg_start_output(); but that method doesn't
+let you inspect the next scan's parameters before deciding to display it.
+
+
+In buffered-image mode, jpeg_start_decompress() never performs input and
+thus never suspends.  An application that uses input suspension with
+buffered-image mode must be prepared for suspension returns from these
+routines:
+* jpeg_start_output() performs input only if you request 2-pass quantization
+  and the target scan isn't fully read yet.  (This is discussed below.)
+* jpeg_read_scanlines(), as always, returns the number of scanlines that it
+  was able to produce before suspending.
+* jpeg_finish_output() will read any markers following the target scan,
+  up to the end of the file or the SOS marker that begins another scan.
+  (But it reads no input if jpeg_consume_input() has already reached the
+  end of the file or a SOS marker beyond the target output scan.)
+* jpeg_finish_decompress() will read until the end of file, and thus can
+  suspend if the end hasn't already been reached (as can be tested by
+  calling jpeg_input_complete()).
+jpeg_start_output(), jpeg_finish_output(), and jpeg_finish_decompress()
+all return TRUE if they completed their tasks, FALSE if they had to suspend.
+In the event of a FALSE return, the application must load more input data
+and repeat the call.  Applications that use non-suspending data sources need
+not check the return values of these three routines.
+
+
+It is possible to change decoding parameters between output passes in the
+buffered-image mode.  The decoder library currently supports only very
+limited changes of parameters.  ONLY THE FOLLOWING parameter changes are
+allowed after jpeg_start_decompress() is called:
+* dct_method can be changed before each call to jpeg_start_output().
+  For example, one could use a fast DCT method for early scans, changing
+  to a higher quality method for the final scan.
+* dither_mode can be changed before each call to jpeg_start_output();
+  of course this has no impact if not using color quantization.  Typically
+  one would use ordered dither for initial passes, then switch to
+  Floyd-Steinberg dither for the final pass.  Caution: changing dither mode
+  can cause more memory to be allocated by the library.  Although the amount
+  of memory involved is not large (a scanline or so), it may cause the
+  initial max_memory_to_use specification to be exceeded, which in the worst
+  case would result in an out-of-memory failure.
+* do_block_smoothing can be changed before each call to jpeg_start_output().
+  This setting is relevant only when decoding a progressive JPEG image.
+  During the first DC-only scan, block smoothing provides a very "fuzzy" look
+  instead of the very "blocky" look seen without it; which is better seems a
+  matter of personal taste.  But block smoothing is nearly always a win
+  during later stages, especially when decoding a successive-approximation
+  image: smoothing helps to hide the slight blockiness that otherwise shows
+  up on smooth gradients until the lowest coefficient bits are sent.
+* Color quantization mode can be changed under the rules described below.
+  You *cannot* change between full-color and quantized output (because that
+  would alter the required I/O buffer sizes), but you can change which
+  quantization method is used.
+
+When generating color-quantized output, changing quantization method is a
+very useful way of switching between high-speed and high-quality display.
+The library allows you to change among its three quantization methods:
+1. Single-pass quantization to a fixed color cube.
+   Selected by cinfo.two_pass_quantize = FALSE and cinfo.colormap = NULL.
+2. Single-pass quantization to an application-supplied colormap.
+   Selected by setting cinfo.colormap to point to the colormap (the value of
+   two_pass_quantize is ignored); also set cinfo.actual_number_of_colors.
+3. Two-pass quantization to a colormap chosen specifically for the image.
+   Selected by cinfo.two_pass_quantize = TRUE and cinfo.colormap = NULL.
+   (This is the default setting selected by jpeg_read_header, but it is
+   probably NOT what you want for the first pass of progressive display!)
+These methods offer successively better quality and lesser speed.  However,
+only the first method is available for quantizing in non-RGB color spaces.
+
+IMPORTANT: because the different quantizer methods have very different
+working-storage requirements, the library requires you to indicate which
+one(s) you intend to use before you call jpeg_start_decompress().  (If we did
+not require this, the max_memory_to_use setting would be a complete fiction.)
+You do this by setting one or more of these three cinfo fields to TRUE:
+        enable_1pass_quant              Fixed color cube colormap
+        enable_external_quant           Externally-supplied colormap
+        enable_2pass_quant              Two-pass custom colormap
+All three are initialized FALSE by jpeg_read_header().  But
+jpeg_start_decompress() automatically sets TRUE the one selected by the
+current two_pass_quantize and colormap settings, so you only need to set the
+enable flags for any other quantization methods you plan to change to later.
+
+After setting the enable flags correctly at jpeg_start_decompress() time, you
+can change to any enabled quantization method by setting two_pass_quantize
+and colormap properly just before calling jpeg_start_output().  The following
+special rules apply:
+1. You must explicitly set cinfo.colormap to NULL when switching to 1-pass
+   or 2-pass mode from a different mode, or when you want the 2-pass
+   quantizer to be re-run to generate a new colormap.
+2. To switch to an external colormap, or to change to a different external
+   colormap than was used on the prior pass, you must call
+   jpeg_new_colormap() after setting cinfo.colormap.
+NOTE: if you want to use the same colormap as was used in the prior pass,
+you should not do either of these things.  This will save some nontrivial
+switchover costs.
+(These requirements exist because cinfo.colormap will always be non-NULL
+after completing a prior output pass, since both the 1-pass and 2-pass
+quantizers set it to point to their output colormaps.  Thus you have to
+do one of these two things to notify the library that something has changed.
+Yup, it's a bit klugy, but it's necessary to do it this way for backwards
+compatibility.)
+
+Note that in buffered-image mode, the library generates any requested colormap
+during jpeg_start_output(), not during jpeg_start_decompress().
+
+When using two-pass quantization, jpeg_start_output() makes a pass over the
+buffered image to determine the optimum color map; it therefore may take a
+significant amount of time, whereas ordinarily it does little work.  The
+progress monitor hook is called during this pass, if defined.  It is also
+important to realize that if the specified target scan number is greater than
+or equal to the current input scan number, jpeg_start_output() will attempt
+to consume input as it makes this pass.  If you use a suspending data source,
+you need to check for a FALSE return from jpeg_start_output() under these
+conditions.  The combination of 2-pass quantization and a not-yet-fully-read
+target scan is the only case in which jpeg_start_output() will consume input.
+
+
+Application authors who support buffered-image mode may be tempted to use it
+for all JPEG images, even single-scan ones.  This will work, but it is
+inefficient: there is no need to create an image-sized coefficient buffer for
+single-scan images.  Requesting buffered-image mode for such an image wastes
+memory.  Worse, it can cost time on large images, since the buffered data has
+to be swapped out or written to a temporary file.  If you are concerned about
+maximum performance on baseline JPEG files, you should use buffered-image
+mode only when the incoming file actually has multiple scans.  This can be
+tested by calling jpeg_has_multiple_scans(), which will return a correct
+result at any time after jpeg_read_header() completes.
+
+It is also worth noting that when you use jpeg_consume_input() to let input
+processing get ahead of output processing, the resulting pattern of access to
+the coefficient buffer is quite nonsequential.  It's best to use the memory
+manager jmemnobs.c if you can (ie, if you have enough real or virtual main
+memory).  If not, at least make sure that max_memory_to_use is set as high as
+possible.  If the JPEG memory manager has to use a temporary file, you will
+probably see a lot of disk traffic and poor performance.  (This could be
+improved with additional work on the memory manager, but we haven't gotten
+around to it yet.)
+
+In some applications it may be convenient to use jpeg_consume_input() for all
+input processing, including reading the initial markers; that is, you may
+wish to call jpeg_consume_input() instead of jpeg_read_header() during
+startup.  This works, but note that you must check for JPEG_REACHED_SOS and
+JPEG_REACHED_EOI return codes as the equivalent of jpeg_read_header's codes.
+Once the first SOS marker has been reached, you must call
+jpeg_start_decompress() before jpeg_consume_input() will consume more input;
+it'll just keep returning JPEG_REACHED_SOS until you do.  If you read a
+tables-only file this way, jpeg_consume_input() will return JPEG_REACHED_EOI
+without ever returning JPEG_REACHED_SOS; be sure to check for this case.
+If this happens, the decompressor will not read any more input until you call
+jpeg_abort() to reset it.  It is OK to call jpeg_consume_input() even when not
+using buffered-image mode, but in that case it's basically a no-op after the
+initial markers have been read: it will just return JPEG_SUSPENDED.
+
+
+Abbreviated datastreams and multiple images
+-------------------------------------------
+
+A JPEG compression or decompression object can be reused to process multiple
+images.  This saves a small amount of time per image by eliminating the
+"create" and "destroy" operations, but that isn't the real purpose of the
+feature.  Rather, reuse of an object provides support for abbreviated JPEG
+datastreams.  Object reuse can also simplify processing a series of images in
+a single input or output file.  This section explains these features.
+
+A JPEG file normally contains several hundred bytes worth of quantization
+and Huffman tables.  In a situation where many images will be stored or
+transmitted with identical tables, this may represent an annoying overhead.
+The JPEG standard therefore permits tables to be omitted.  The standard
+defines three classes of JPEG datastreams:
+  * "Interchange" datastreams contain an image and all tables needed to decode
+     the image.  These are the usual kind of JPEG file.
+  * "Abbreviated image" datastreams contain an image, but are missing some or
+    all of the tables needed to decode that image.
+  * "Abbreviated table specification" (henceforth "tables-only") datastreams
+    contain only table specifications.
+To decode an abbreviated image, it is necessary to load the missing table(s)
+into the decoder beforehand.  This can be accomplished by reading a separate
+tables-only file.  A variant scheme uses a series of images in which the first
+image is an interchange (complete) datastream, while subsequent ones are
+abbreviated and rely on the tables loaded by the first image.  It is assumed
+that once the decoder has read a table, it will remember that table until a
+new definition for the same table number is encountered.
+
+It is the application designer's responsibility to figure out how to associate
+the correct tables with an abbreviated image.  While abbreviated datastreams
+can be useful in a closed environment, their use is strongly discouraged in
+any situation where data exchange with other applications might be needed.
+Caveat designer.
+
+The JPEG library provides support for reading and writing any combination of
+tables-only datastreams and abbreviated images.  In both compression and
+decompression objects, a quantization or Huffman table will be retained for
+the lifetime of the object, unless it is overwritten by a new table definition.
+
+
+To create abbreviated image datastreams, it is only necessary to tell the
+compressor not to emit some or all of the tables it is using.  Each
+quantization and Huffman table struct contains a boolean field "sent_table",
+which normally is initialized to FALSE.  For each table used by the image, the
+header-writing process emits the table and sets sent_table = TRUE unless it is
+already TRUE.  (In normal usage, this prevents outputting the same table
+definition multiple times, as would otherwise occur because the chroma
+components typically share tables.)  Thus, setting this field to TRUE before
+calling jpeg_start_compress() will prevent the table from being written at
+all.
+
+If you want to create a "pure" abbreviated image file containing no tables,
+just call "jpeg_suppress_tables(&cinfo, TRUE)" after constructing all the
+tables.  If you want to emit some but not all tables, you'll need to set the
+individual sent_table fields directly.
+
+To create an abbreviated image, you must also call jpeg_start_compress()
+with a second parameter of FALSE, not TRUE.  Otherwise jpeg_start_compress()
+will force all the sent_table fields to FALSE.  (This is a safety feature to
+prevent abbreviated images from being created accidentally.)
+
+To create a tables-only file, perform the same parameter setup that you
+normally would, but instead of calling jpeg_start_compress() and so on, call
+jpeg_write_tables(&cinfo).  This will write an abbreviated datastream
+containing only SOI, DQT and/or DHT markers, and EOI.  All the quantization
+and Huffman tables that are currently defined in the compression object will
+be emitted unless their sent_tables flag is already TRUE, and then all the
+sent_tables flags will be set TRUE.
+
+A sure-fire way to create matching tables-only and abbreviated image files
+is to proceed as follows:
+
+        create JPEG compression object
+        set JPEG parameters
+        set destination to tables-only file
+        jpeg_write_tables(&cinfo);
+        set destination to image file
+        jpeg_start_compress(&cinfo, FALSE);
+        write data...
+        jpeg_finish_compress(&cinfo);
+
+Since the JPEG parameters are not altered between writing the table file and
+the abbreviated image file, the same tables are sure to be used.  Of course,
+you can repeat the jpeg_start_compress() ... jpeg_finish_compress() sequence
+many times to produce many abbreviated image files matching the table file.
+
+You cannot suppress output of the computed Huffman tables when Huffman
+optimization is selected.  (If you could, there'd be no way to decode the
+image...)  Generally, you don't want to set optimize_coding = TRUE when
+you are trying to produce abbreviated files.
+
+In some cases you might want to compress an image using tables which are
+not stored in the application, but are defined in an interchange or
+tables-only file readable by the application.  This can be done by setting up
+a JPEG decompression object to read the specification file, then copying the
+tables into your compression object.  See jpeg_copy_critical_parameters()
+for an example of copying quantization tables.
+
+
+To read abbreviated image files, you simply need to load the proper tables
+into the decompression object before trying to read the abbreviated image.
+If the proper tables are stored in the application program, you can just
+allocate the table structs and fill in their contents directly.  For example,
+to load a fixed quantization table into table slot "n":
+
+    if (cinfo.quant_tbl_ptrsn == NULL)
+      cinfo.quant_tbl_ptrsn = jpeg_alloc_quant_table((j_common_ptr) &cinfo);
+    quant_ptr = cinfo.quant_tbl_ptrsn;        /* quant_ptr is JQUANT_TBL* */
+    for (i = 0; i < 64; i++) {
+      /* Qtable is desired quantization table, in natural array order */
+      quant_ptr->quantvali = Qtablei;
+    }
+
+Code to load a fixed Huffman table is typically (for AC table "n"):
+
+    if (cinfo.ac_huff_tbl_ptrsn == NULL)
+      cinfo.ac_huff_tbl_ptrsn = jpeg_alloc_huff_table((j_common_ptr) &cinfo);
+    huff_ptr = cinfo.ac_huff_tbl_ptrsn;       /* huff_ptr is JHUFF_TBL* */
+    for (i = 1; i <= 16; i++) {
+      /* countsi is number of Huffman codes of length i bits, i=1..16 */
+      huff_ptr->bitsi = countsi;
+    }
+    for (i = 0; i < 256; i++) {
+      /* symbols is the list of Huffman symbols, in code-length order */
+      huff_ptr->huffvali = symbolsi;
+    }
+
+(Note that trying to set cinfo.quant_tbl_ptrsn to point directly at a
+constant JQUANT_TBL object is not safe.  If the incoming file happened to
+contain a quantization table definition, your master table would get
+overwritten!  Instead allocate a working table copy and copy the master table
+into it, as illustrated above.  Ditto for Huffman tables, of course.)
+
+You might want to read the tables from a tables-only file, rather than
+hard-wiring them into your application.  The jpeg_read_header() call is
+sufficient to read a tables-only file.  You must pass a second parameter of
+FALSE to indicate that you do not require an image to be present.  Thus, the
+typical scenario is
+
+        create JPEG decompression object
+        set source to tables-only file
+        jpeg_read_header(&cinfo, FALSE);
+        set source to abbreviated image file
+        jpeg_read_header(&cinfo, TRUE);
+        set decompression parameters
+        jpeg_start_decompress(&cinfo);
+        read data...
+        jpeg_finish_decompress(&cinfo);
+
+In some cases, you may want to read a file without knowing whether it contains
+an image or just tables.  In that case, pass FALSE and check the return value
+from jpeg_read_header(): it will be JPEG_HEADER_OK if an image was found,
+JPEG_HEADER_TABLES_ONLY if only tables were found.  (A third return value,
+JPEG_SUSPENDED, is possible when using a suspending data source manager.)
+Note that jpeg_read_header() will not complain if you read an abbreviated
+image for which you haven't loaded the missing tables; the missing-table check
+occurs later, in jpeg_start_decompress().
+
+
+It is possible to read a series of images from a single source file by
+repeating the jpeg_read_header() ... jpeg_finish_decompress() sequence,
+without releasing/recreating the JPEG object or the data source module.
+(If you did reinitialize, any partial bufferload left in the data source
+buffer at the end of one image would be discarded, causing you to lose the
+start of the next image.)  When you use this method, stored tables are
+automatically carried forward, so some of the images can be abbreviated images
+that depend on tables from earlier images.
+
+If you intend to write a series of images into a single destination file,
+you might want to make a specialized data destination module that doesn't
+flush the output buffer at term_destination() time.  This would speed things
+up by some trifling amount.  Of course, you'd need to remember to flush the
+buffer after the last image.  You can make the later images be abbreviated
+ones by passing FALSE to jpeg_start_compress().
+
+
+Special markers
+---------------
+
+Some applications may need to insert or extract special data in the JPEG
+datastream.  The JPEG standard provides marker types "COM" (comment) and
+"APP0" through "APP15" (application) to hold application-specific data.
+Unfortunately, the use of these markers is not specified by the standard.
+COM markers are fairly widely used to hold user-supplied text.  The JFIF file
+format spec uses APP0 markers with specified initial strings to hold certain
+data.  Adobe applications use APP14 markers beginning with the string "Adobe"
+for miscellaneous data.  Other APPn markers are rarely seen, but might
+contain almost anything.
+
+If you wish to store user-supplied text, we recommend you use COM markers
+and place readable 7-bit ASCII text in them.  Newline conventions are not
+standardized --- expect to find LF (Unix style), CR/LF (DOS style), or CR
+(Mac style).  A robust COM reader should be able to cope with random binary
+garbage, including nulls, since some applications generate COM markers
+containing non-ASCII junk.  (But yours should not be one of them.)
+
+For program-supplied data, use an APPn marker, and be sure to begin it with an
+identifying string so that you can tell whether the marker is actually yours.
+It's probably best to avoid using APP0 or APP14 for any private markers.
+(NOTE: the upcoming SPIFF standard will use APP8 markers; we recommend you
+not use APP8 markers for any private purposes, either.)
+
+Keep in mind that at most 65533 bytes can be put into one marker, but you
+can have as many markers as you like.
+
+By default, the IJG compression library will write a JFIF APP0 marker if the
+selected JPEG colorspace is grayscale or YCbCr, or an Adobe APP14 marker if
+the selected colorspace is RGB, CMYK, or YCCK.  You can disable this, but
+we don't recommend it.  The decompression library will recognize JFIF and
+Adobe markers and will set the JPEG colorspace properly when one is found.
+
+
+You can write special markers immediately following the datastream header by
+calling jpeg_write_marker() after jpeg_start_compress() and before the first
+call to jpeg_write_scanlines().  When you do this, the markers appear after
+the SOI and the JFIF APP0 and Adobe APP14 markers (if written), but before
+all else.  Specify the marker type parameter as "JPEG_COM" for COM or
+"JPEG_APP0 + n" for APPn.  (Actually, jpeg_write_marker will let you write
+any marker type, but we don't recommend writing any other kinds of marker.)
+For example, to write a user comment string pointed to by comment_text:
+        jpeg_write_marker(cinfo, JPEG_COM, comment_text, strlen(comment_text));
+
+If it's not convenient to store all the marker data in memory at once,
+you can instead call jpeg_write_m_header() followed by multiple calls to
+jpeg_write_m_byte().  If you do it this way, it's your responsibility to
+call jpeg_write_m_byte() exactly the number of times given in the length
+parameter to jpeg_write_m_header().  (This method lets you empty the
+output buffer partway through a marker, which might be important when
+using a suspending data destination module.  In any case, if you are using
+a suspending destination, you should flush its buffer after inserting
+any special markers.  See "I/O suspension".)
+
+Or, if you prefer to synthesize the marker byte sequence yourself,
+you can just cram it straight into the data destination module.
+
+If you are writing JFIF 1.02 extension markers (thumbnail images), don't
+forget to set cinfo.JFIF_minor_version = 2 so that the encoder will write the
+correct JFIF version number in the JFIF header marker.  The library's default
+is to write version 1.01, but that's wrong if you insert any 1.02 extension
+markers.  (We could probably get away with just defaulting to 1.02, but there
+used to be broken decoders that would complain about unknown minor version
+numbers.  To reduce compatibility risks it's safest not to write 1.02 unless
+you are actually using 1.02 extensions.)
+
+
+When reading, two methods of handling special markers are available:
+1. You can ask the library to save the contents of COM and/or APPn markers
+into memory, and then examine them at your leisure afterwards.
+2. You can supply your own routine to process COM and/or APPn markers
+on-the-fly as they are read.
+The first method is simpler to use, especially if you are using a suspending
+data source; writing a marker processor that copes with input suspension is
+not easy (consider what happens if the marker is longer than your available
+input buffer).  However, the second method conserves memory since the marker
+data need not be kept around after it's been processed.
+
+For either method, you'd normally set up marker handling after creating a
+decompression object and before calling jpeg_read_header(), because the
+markers of interest will typically be near the head of the file and so will
+be scanned by jpeg_read_header.  Once you've established a marker handling
+method, it will be used for the life of that decompression object
+(potentially many datastreams), unless you change it.  Marker handling is
+determined separately for COM markers and for each APPn marker code.
+
+
+To save the contents of special markers in memory, call
+        jpeg_save_markers(cinfo, marker_code, length_limit)
+where marker_code is the marker type to save, JPEG_COM or JPEG_APP0+n.
+(To arrange to save all the special marker types, you need to call this
+routine 17 times, for COM and APP0-APP15.)  If the incoming marker is longer
+than length_limit data bytes, only length_limit bytes will be saved; this
+parameter allows you to avoid chewing up memory when you only need to see the
+first few bytes of a potentially large marker.  If you want to save all the
+data, set length_limit to 0xFFFF; that is enough since marker lengths are only
+16 bits.  As a special case, setting length_limit to 0 prevents that marker
+type from being saved at all.  (That is the default behavior, in fact.)
+
+After jpeg_read_header() completes, you can examine the special markers by
+following the cinfo->marker_list pointer chain.  All the special markers in
+the file appear in this list, in order of their occurrence in the file (but
+omitting any markers of types you didn't ask for).  Both the original data
+length and the saved data length are recorded for each list entry; the latter
+will not exceed length_limit for the particular marker type.  Note that these
+lengths exclude the marker length word, whereas the stored representation
+within the JPEG file includes it.  (Hence the maximum data length is really
+only 65533.)
+
+It is possible that additional special markers appear in the file beyond the
+SOS marker at which jpeg_read_header stops; if so, the marker list will be
+extended during reading of the rest of the file.  This is not expected to be
+common, however.  If you are short on memory you may want to reset the length
+limit to zero for all marker types after finishing jpeg_read_header, to
+ensure that the max_memory_to_use setting cannot be exceeded due to addition
+of later markers.
+
+The marker list remains stored until you call jpeg_finish_decompress or
+jpeg_abort, at which point the memory is freed and the list is set to empty.
+(jpeg_destroy also releases the storage, of course.)
+
+Note that the library is internally interested in APP0 and APP14 markers;
+if you try to set a small nonzero length limit on these types, the library
+will silently force the length up to the minimum it wants.  (But you can set
+a zero length limit to prevent them from being saved at all.)  Also, in a
+16-bit environment, the maximum length limit may be constrained to less than
+65533 by malloc() limitations.  It is therefore best not to assume that the
+effective length limit is exactly what you set it to be.
+
+
+If you want to supply your own marker-reading routine, you do it by calling
+jpeg_set_marker_processor().  A marker processor routine must have the
+signature
+        boolean jpeg_marker_parser_method (j_decompress_ptr cinfo)
+Although the marker code is not explicitly passed, the routine can find it
+in cinfo->unread_marker.  At the time of call, the marker proper has been
+read from the data source module.  The processor routine is responsible for
+reading the marker length word and the remaining parameter bytes, if any.
+Return TRUE to indicate success.  (FALSE should be returned only if you are
+using a suspending data source and it tells you to suspend.  See the standard
+marker processors in jdmarker.c for appropriate coding methods if you need to
+use a suspending data source.)
+
+If you override the default APP0 or APP14 processors, it is up to you to
+recognize JFIF and Adobe markers if you want colorspace recognition to occur
+properly.  We recommend copying and extending the default processors if you
+want to do that.  (A better idea is to save these marker types for later
+examination by calling jpeg_save_markers(); that method doesn't interfere
+with the library's own processing of these markers.)
+
+jpeg_set_marker_processor() and jpeg_save_markers() are mutually exclusive
+--- if you call one it overrides any previous call to the other, for the
+particular marker type specified.
+
+A simple example of an external COM processor can be found in djpeg.c.
+Also, see jpegtran.c for an example of using jpeg_save_markers.
+
+
+ICC profiles
+------------
+
+Two functions are provided for writing and reading International Color
+Consortium (ICC) device profiles embedded in JFIF JPEG image files:
+
+        void jpeg_write_icc_profile (j_compress_ptr cinfo,
+                                     const JOCTET *icc_data_ptr,
+                                     unsigned int icc_data_len);
+        boolean jpeg_read_icc_profile (j_decompress_ptr cinfo,
+                                       JOCTET **icc_data_ptr,
+                                       unsigned int *icc_data_len);
+
+The ICC has defined a standard for including such data in JPEG "APP2" markers.
+The aforementioned functions do not know anything about the internal structure
+of the ICC profile data; they just know how to embed the profile data into a
+JPEG file while writing it, or to extract the profile data from a JPEG file
+while reading it.
+
+jpeg_write_icc_profile() must be called after calling jpeg_start_compress() and
+before the first call to jpeg_write_scanlines() or jpeg_write_raw_data().  This
+ordering ensures that the APP2 marker(s) will appear after the SOI and JFIF or
+Adobe markers, but before all other data.
+
+jpeg_read_icc_profile() returns TRUE if an ICC profile was found and FALSE
+otherwise.  If an ICC profile was found, then the function will allocate a
+memory region containing the profile and will return a pointer to that memory
+region in *icc_data_ptr, as well as the length of the region in *icc_data_len.
+This memory region is allocated by the library using malloc() and must be freed
+by the caller using free() when the memory region is no longer needed.  Callers
+wishing to use jpeg_read_icc_profile() must call
+
+        jpeg_save_markers(cinfo, JPEG_APP0 + 2, 0xFFFF);
+
+prior to calling jpeg_read_header().  jpeg_read_icc_profile() can be called at
+any point between jpeg_read_header() and jpeg_finish_decompress().
+
+
+Raw (downsampled) image data
+----------------------------
+
+Some applications need to supply already-downsampled image data to the JPEG
+compressor, or to receive raw downsampled data from the decompressor.  The
+library supports this requirement by allowing the application to write or
+read raw data, bypassing the normal preprocessing or postprocessing steps.
+The interface is different from the standard one and is somewhat harder to
+use.  If your interest is merely in bypassing color conversion, we recommend
+that you use the standard interface and simply set jpeg_color_space =
+in_color_space (or jpeg_color_space = out_color_space for decompression).
+The mechanism described in this section is necessary only to supply or
+receive downsampled image data, in which not all components have the same
+dimensions.
+
+
+To compress raw data, you must supply the data in the colorspace to be used
+in the JPEG file (please read the earlier section on Special color spaces)
+and downsampled to the sampling factors specified in the JPEG parameters.
+You must supply the data in the format used internally by the JPEG library,
+namely a JSAMPIMAGE array.  This is an array of pointers to two-dimensional
+arrays, each of type JSAMPARRAY.  Each 2-D array holds the values for one
+color component.  This structure is necessary since the components are of
+different sizes.  If the image dimensions are not a multiple of the MCU size,
+you must also pad the data correctly (usually, this is done by replicating
+the last column and/or row).  The data must be padded to a multiple of a DCT
+block in each component: that is, each downsampled row must contain a
+multiple of 8 valid samples, and there must be a multiple of 8 sample rows
+for each component.  (For applications such as conversion of digital TV
+images, the standard image size is usually a multiple of the DCT block size,
+so that no padding need actually be done.)
+
+The procedure for compression of raw data is basically the same as normal
+compression, except that you call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  Before calling jpeg_start_compress(), you must do
+the following:
+  * Set cinfo->raw_data_in to TRUE.  (It is set FALSE by jpeg_set_defaults().)
+    This notifies the library that you will be supplying raw data.
+  * Ensure jpeg_color_space is correct --- an explicit jpeg_set_colorspace()
+    call is a good idea.  Note that since color conversion is bypassed,
+    in_color_space is ignored, except that jpeg_set_defaults() uses it to
+    choose the default jpeg_color_space setting.
+  * Ensure the sampling factors, cinfo->comp_infoi.h_samp_factor and
+    cinfo->comp_infoi.v_samp_factor, are correct.  Since these indicate the
+    dimensions of the data you are supplying, it's wise to set them
+    explicitly, rather than assuming the library's defaults are what you want.
+
+To pass raw data to the library, call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  The two routines work similarly except that
+jpeg_write_raw_data takes a JSAMPIMAGE data array rather than JSAMPARRAY.
+The scanlines count passed to and returned from jpeg_write_raw_data is
+measured in terms of the component with the largest v_samp_factor.
+
+jpeg_write_raw_data() processes one MCU row per call, which is to say
+v_samp_factor*DCTSIZE sample rows of each component.  The passed num_lines
+value must be at least max_v_samp_factor*DCTSIZE, and the return value will
+be exactly that amount (or possibly some multiple of that amount, in future
+library versions).  This is true even on the last call at the bottom of the
+image; don't forget to pad your data as necessary.
+
+The required dimensions of the supplied data can be computed for each
+component as
+        cinfo->comp_infoi.width_in_blocks*DCTSIZE  samples per row
+        cinfo->comp_infoi.height_in_blocks*DCTSIZE rows in image
+after jpeg_start_compress() has initialized those fields.  If the valid data
+is smaller than this, it must be padded appropriately.  For some sampling
+factors and image sizes, additional dummy DCT blocks are inserted to make
+the image a multiple of the MCU dimensions.  The library creates such dummy
+blocks itself; it does not read them from your supplied data.  Therefore you
+need never pad by more than DCTSIZE samples.  An example may help here.
+Assume 2h2v downsampling of YCbCr data, that is
+        cinfo->comp_info0.h_samp_factor = 2           for Y
+        cinfo->comp_info0.v_samp_factor = 2
+        cinfo->comp_info1.h_samp_factor = 1           for Cb
+        cinfo->comp_info1.v_samp_factor = 1
+        cinfo->comp_info2.h_samp_factor = 1           for Cr
+        cinfo->comp_info2.v_samp_factor = 1
+and suppose that the nominal image dimensions (cinfo->image_width and
+cinfo->image_height) are 101x101 pixels.  Then jpeg_start_compress() will
+compute downsampled_width = 101 and width_in_blocks = 13 for Y,
+downsampled_width = 51 and width_in_blocks = 7 for Cb and Cr (and the same
+for the height fields).  You must pad the Y data to at least 13*8 = 104
+columns and rows, the Cb/Cr data to at least 7*8 = 56 columns and rows.  The
+MCU height is max_v_samp_factor = 2 DCT rows so you must pass at least 16
+scanlines on each call to jpeg_write_raw_data(), which is to say 16 actual
+sample rows of Y and 8 each of Cb and Cr.  A total of 7 MCU rows are needed,
+so you must pass a total of 7*16 = 112 "scanlines".  The last DCT block row
+of Y data is dummy, so it doesn't matter what you pass for it in the data
+arrays, but the scanlines count must total up to 112 so that all of the Cb
+and Cr data gets passed.
+
+Output suspension is supported with raw-data compression: if the data
+destination module suspends, jpeg_write_raw_data() will return 0.
+In this case the same data rows must be passed again on the next call.
+
+
+Decompression with raw data output implies bypassing all postprocessing:
+you cannot ask for rescaling or color quantization, for instance.  More
+seriously, you must deal with the color space and sampling factors present in
+the incoming file.  If your application only handles, say, 2h1v YCbCr data,
+you must check for and fail on other color spaces or other sampling factors.
+The library will not convert to a different color space for you.
+
+To obtain raw data output, set cinfo->raw_data_out = TRUE before
+jpeg_start_decompress() (it is set FALSE by jpeg_read_header()).  Be sure to
+verify that the color space and sampling factors are ones you can handle.
+Then call jpeg_read_raw_data() in place of jpeg_read_scanlines().  The
+decompression process is otherwise the same as usual.
+
+jpeg_read_raw_data() returns one MCU row per call, and thus you must pass a
+buffer of at least max_v_samp_factor*DCTSIZE scanlines (scanline counting is
+the same as for raw-data compression).  The buffer you pass must be large
+enough to hold the actual data plus padding to DCT-block boundaries.  As with
+compression, any entirely dummy DCT blocks are not processed so you need not
+allocate space for them, but the total scanline count includes them.  The
+above example of computing buffer dimensions for raw-data compression is
+equally valid for decompression.
+
+Input suspension is supported with raw-data decompression: if the data source
+module suspends, jpeg_read_raw_data() will return 0.  You can also use
+buffered-image mode to read raw data in multiple passes.
+
+
+Really raw data: DCT coefficients
+---------------------------------
+
+It is possible to read or write the contents of a JPEG file as raw DCT
+coefficients.  This facility is mainly intended for use in lossless
+transcoding between different JPEG file formats.  Other possible applications
+include lossless cropping of a JPEG image, lossless reassembly of a
+multi-strip or multi-tile TIFF/JPEG file into a single JPEG datastream, etc.
+
+To read the contents of a JPEG file as DCT coefficients, open the file and do
+jpeg_read_header() as usual.  But instead of calling jpeg_start_decompress()
+and jpeg_read_scanlines(), call jpeg_read_coefficients().  This will read the
+entire image into a set of virtual coefficient-block arrays, one array per
+component.  The return value is a pointer to an array of virtual-array
+descriptors.  Each virtual array can be accessed directly using the JPEG
+memory manager's access_virt_barray method (see Memory management, below,
+and also read structure.txt's discussion of virtual array handling).  Or,
+for simple transcoding to a different JPEG file format, the array list can
+just be handed directly to jpeg_write_coefficients().
+
+Each block in the block arrays contains quantized coefficient values in
+normal array order (not JPEG zigzag order).  The block arrays contain only
+DCT blocks containing real data; any entirely-dummy blocks added to fill out
+interleaved MCUs at the right or bottom edges of the image are discarded
+during reading and are not stored in the block arrays.  (The size of each
+block array can be determined from the width_in_blocks and height_in_blocks
+fields of the component's comp_info entry.)  This is also the data format
+expected by jpeg_write_coefficients().
+
+When you are done using the virtual arrays, call jpeg_finish_decompress()
+to release the array storage and return the decompression object to an idle
+state; or just call jpeg_destroy() if you don't need to reuse the object.
+
+If you use a suspending data source, jpeg_read_coefficients() will return
+NULL if it is forced to suspend; a non-NULL return value indicates successful
+completion.  You need not test for a NULL return value when using a
+non-suspending data source.
+
+It is also possible to call jpeg_read_coefficients() to obtain access to the
+decoder's coefficient arrays during a normal decode cycle in buffered-image
+mode.  This frammish might be useful for progressively displaying an incoming
+image and then re-encoding it without loss.  To do this, decode in buffered-
+image mode as discussed previously, then call jpeg_read_coefficients() after
+the last jpeg_finish_output() call.  The arrays will be available for your use
+until you call jpeg_finish_decompress().
+
+
+To write the contents of a JPEG file as DCT coefficients, you must provide
+the DCT coefficients stored in virtual block arrays.  You can either pass
+block arrays read from an input JPEG file by jpeg_read_coefficients(), or
+allocate virtual arrays from the JPEG compression object and fill them
+yourself.  In either case, jpeg_write_coefficients() is substituted for
+jpeg_start_compress() and jpeg_write_scanlines().  Thus the sequence is
+  * Create compression object
+  * Set all compression parameters as necessary
+  * Request virtual arrays if needed
+  * jpeg_write_coefficients()
+  * jpeg_finish_compress()
+  * Destroy or re-use compression object
+jpeg_write_coefficients() is passed a pointer to an array of virtual block
+array descriptors; the number of arrays is equal to cinfo.num_components.
+
+The virtual arrays need only have been requested, not realized, before
+jpeg_write_coefficients() is called.  A side-effect of
+jpeg_write_coefficients() is to realize any virtual arrays that have been
+requested from the compression object's memory manager.  Thus, when obtaining
+the virtual arrays from the compression object, you should fill the arrays
+after calling jpeg_write_coefficients().  The data is actually written out
+when you call jpeg_finish_compress(); jpeg_write_coefficients() only writes
+the file header.
+
+When writing raw DCT coefficients, it is crucial that the JPEG quantization
+tables and sampling factors match the way the data was encoded, or the
+resulting file will be invalid.  For transcoding from an existing JPEG file,
+we recommend using jpeg_copy_critical_parameters().  This routine initializes
+all the compression parameters to default values (like jpeg_set_defaults()),
+then copies the critical information from a source decompression object.
+The decompression object should have just been used to read the entire
+JPEG input file --- that is, it should be awaiting jpeg_finish_decompress().
+
+jpeg_write_coefficients() marks all tables stored in the compression object
+as needing to be written to the output file (thus, it acts like
+jpeg_start_compress(cinfo, TRUE)).  This is for safety's sake, to avoid
+emitting abbreviated JPEG files by accident.  If you really want to emit an
+abbreviated JPEG file, call jpeg_suppress_tables(), or set the tables'
+individual sent_table flags, between calling jpeg_write_coefficients() and
+jpeg_finish_compress().
+
+
+Progress monitoring
+-------------------
+
+Some applications may need to regain control from the JPEG library every so
+often.  The typical use of this feature is to produce a percent-done bar or
+other progress display.  (For a simple example, see cjpeg.c or djpeg.c.)
+Although you do get control back frequently during the data-transferring pass
+(the jpeg_read_scanlines or jpeg_write_scanlines loop), any additional passes
+will occur inside jpeg_finish_compress or jpeg_start_decompress; those
+routines may take a long time to execute, and you don't get control back
+until they are done.
+
+You can define a progress-monitor routine which will be called periodically
+by the library.  No guarantees are made about how often this call will occur,
+so we don't recommend you use it for mouse tracking or anything like that.
+At present, a call will occur once per MCU row, scanline, or sample row
+group, whichever unit is convenient for the current processing mode; so the
+wider the image, the longer the time between calls.  During the data
+transferring pass, only one call occurs per call of jpeg_read_scanlines or
+jpeg_write_scanlines, so don't pass a large number of scanlines at once if
+you want fine resolution in the progress count.  (If you really need to use
+the callback mechanism for time-critical tasks like mouse tracking, you could
+insert additional calls inside some of the library's inner loops.)
+
+To establish a progress-monitor callback, create a struct jpeg_progress_mgr,
+fill in its progress_monitor field with a pointer to your callback routine,
+and set cinfo->progress to point to the struct.  The callback will be called
+whenever cinfo->progress is non-NULL.  (This pointer is set to NULL by
+jpeg_create_compress or jpeg_create_decompress; the library will not change
+it thereafter.  So if you allocate dynamic storage for the progress struct,
+make sure it will live as long as the JPEG object does.  Allocating from the
+JPEG memory manager with lifetime JPOOL_PERMANENT will work nicely.)  You
+can use the same callback routine for both compression and decompression.
+
+The jpeg_progress_mgr struct contains four fields which are set by the library:
+        long pass_counter;      /* work units completed in this pass */
+        long pass_limit;        /* total number of work units in this pass */
+        int completed_passes;   /* passes completed so far */
+        int total_passes;       /* total number of passes expected */
+During any one pass, pass_counter increases from 0 up to (not including)
+pass_limit; the step size is usually but not necessarily 1.  The pass_limit
+value may change from one pass to another.  The expected total number of
+passes is in total_passes, and the number of passes already completed is in
+completed_passes.  Thus the fraction of work completed may be estimated as
+                completed_passes + (pass_counter/pass_limit)
+                --------------------------------------------
+                                total_passes
+ignoring the fact that the passes may not be equal amounts of work.
+
+When decompressing, pass_limit can even change within a pass, because it
+depends on the number of scans in the JPEG file, which isn't always known in
+advance.  The computed fraction-of-work-done may jump suddenly (if the library
+discovers it has overestimated the number of scans) or even decrease (in the
+opposite case).  It is not wise to put great faith in the work estimate.
+
+When using the decompressor's buffered-image mode, the progress monitor work
+estimate is likely to be completely unhelpful, because the library has no way
+to know how many output passes will be demanded of it.  Currently, the library
+sets total_passes based on the assumption that there will be one more output
+pass if the input file end hasn't yet been read (jpeg_input_complete() isn't
+TRUE), but no more output passes if the file end has been reached when the
+output pass is started.  This means that total_passes will rise as additional
+output passes are requested.  If you have a way of determining the input file
+size, estimating progress based on the fraction of the file that's been read
+will probably be more useful than using the library's value.
+
+
+Memory management
+-----------------
+
+This section covers some key facts about the JPEG library's built-in memory
+manager.  For more info, please read structure.txt's section about the memory
+manager, and consult the source code if necessary.
+
+All memory and temporary file allocation within the library is done via the
+memory manager.  If necessary, you can replace the "back end" of the memory
+manager to control allocation yourself (for example, if you don't want the
+library to use malloc() and free() for some reason).
+
+Some data is allocated "permanently" and will not be freed until the JPEG
+object is destroyed.  Most data is allocated "per image" and is freed by
+jpeg_finish_compress, jpeg_finish_decompress, or jpeg_abort.  You can call the
+memory manager yourself to allocate structures that will automatically be
+freed at these times.  Typical code for this is
+  ptr = (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE, size);
+Use JPOOL_PERMANENT to get storage that lasts as long as the JPEG object.
+Use alloc_large instead of alloc_small for anything bigger than a few Kbytes.
+There are also alloc_sarray and alloc_barray routines that automatically
+build 2-D sample or block arrays.
+
+The library's minimum space requirements to process an image depend on the
+image's width, but not on its height, because the library ordinarily works
+with "strip" buffers that are as wide as the image but just a few rows high.
+Some operating modes (eg, two-pass color quantization) require full-image
+buffers.  Such buffers are treated as "virtual arrays": only the current strip
+need be in memory, and the rest can be swapped out to a temporary file.
+
+When using temporary files, the library will make the in-memory buffers for
+its virtual arrays just big enough to stay within a "maximum memory" setting.
+Your application can set this limit by setting cinfo->mem->max_memory_to_use
+after creating the JPEG object.  (Of course, there is still a minimum size for
+the buffers, so the max-memory setting is effective only if it is bigger than
+the minimum space needed.)  If you allocate any large structures yourself, you
+must allocate them before jpeg_start_compress() or jpeg_start_decompress() in
+order to have them counted against the max memory limit.  Also keep in mind
+that space allocated with alloc_small() is ignored, on the assumption that
+it's too small to be worth worrying about; so a reasonable safety margin
+should be left when setting max_memory_to_use.
+
+NOTE: Unless you develop your own memory manager back end, then temporary files
+will never be used.  The back end provided in libjpeg-turbo (jmemnobs.c) simply
+malloc()s and free()s virtual arrays, and an error occurs if the required
+memory exceeds the limit specified in cinfo->mem->max_memory_to_use.
+
+
+Memory usage
+------------
+
+Working memory requirements while performing compression or decompression
+depend on image dimensions, image characteristics (such as colorspace and
+JPEG process), and operating mode (application-selected options).
+
+As of v6b, the decompressor requires:
+ 1. About 24K in more-or-less-fixed-size data.  This varies a bit depending
+    on operating mode and image characteristics (particularly color vs.
+    grayscale), but it doesn't depend on image dimensions.
+ 2. Strip buffers (of size proportional to the image width) for IDCT and
+    upsampling results.  The worst case for commonly used sampling factors
+    is about 34 bytes * width in pixels for a color image.  A grayscale image
+    only needs about 8 bytes per pixel column.
+ 3. A full-image DCT coefficient buffer is needed to decode a multi-scan JPEG
+    file (including progressive JPEGs), or whenever you select buffered-image
+    mode.  This takes 2 bytes/coefficient.  At typical 2x2 sampling, that's
+    3 bytes per pixel for a color image.  Worst case (1x1 sampling) requires
+    6 bytes/pixel.  For grayscale, figure 2 bytes/pixel.
+ 4. To perform 2-pass color quantization, the decompressor also needs a
+    128K color lookup table and a full-image pixel buffer (3 bytes/pixel).
+This does not count any memory allocated by the application, such as a
+buffer to hold the final output image.
+
+The above figures are valid for 8-bit JPEG data precision and a machine with
+32-bit ints.  For 12-bit JPEG data, double the size of the strip buffers and
+quantization pixel buffer.  The "fixed-size" data will be somewhat smaller
+with 16-bit ints, larger with 64-bit ints.  Also, CMYK or other unusual
+color spaces will require different amounts of space.
+
+The full-image coefficient and pixel buffers, if needed at all, do not
+have to be fully RAM resident; you can have the library use temporary
+files instead when the total memory usage would exceed a limit you set.
+(But if your OS supports virtual memory, it's probably better to just use
+jmemnobs and let the OS do the swapping.)
+
+The compressor's memory requirements are similar, except that it has no need
+for color quantization.  Also, it needs a full-image DCT coefficient buffer
+if Huffman-table optimization is asked for, even if progressive mode is not
+requested.
+
+If you need more detailed information about memory usage in a particular
+situation, you can enable the MEM_STATS code in jmemmgr.c.
+
+
+Library compile-time options
+----------------------------
+
+A number of compile-time options are available by modifying jmorecfg.h.
+
+The JPEG standard provides for both the baseline 8-bit DCT process and
+a 12-bit DCT process.  The IJG code supports 12-bit lossy JPEG if you define
+BITS_IN_JSAMPLE as 12 rather than 8.  Note that this causes JSAMPLE to be
+larger than a char, so it affects the surrounding application's image data.
+The sample applications cjpeg and djpeg can support 12-bit mode only for PPM
+and GIF file formats; you must disable the other file formats to compile a
+12-bit cjpeg or djpeg.  (install.txt has more information about that.)
+At present, a 12-bit library can handle *only* 12-bit images, not both
+precisions.
+
+Note that a 12-bit library always compresses in Huffman optimization mode,
+in order to generate valid Huffman tables.  This is necessary because our
+default Huffman tables only cover 8-bit data.  If you need to output 12-bit
+files in one pass, you'll have to supply suitable default Huffman tables.
+You may also want to supply your own DCT quantization tables; the existing
+quality-scaling code has been developed for 8-bit use, and probably doesn't
+generate especially good tables for 12-bit.
+
+The maximum number of components (color channels) in the image is determined
+by MAX_COMPONENTS.  The JPEG standard allows up to 255 components, but we
+expect that few applications will need more than four or so.
+
+On machines with unusual data type sizes, you may be able to improve
+performance or reduce memory space by tweaking the various typedefs in
+jmorecfg.h.  In particular, on some RISC CPUs, access to arrays of "short"s
+is quite slow; consider trading memory for speed by making JCOEF, INT16, and
+UINT16 be "int" or "unsigned int".  UINT8 is also a candidate to become int.
+You probably don't want to make JSAMPLE be int unless you have lots of memory
+to burn.
+
+You can reduce the size of the library by compiling out various optional
+functions.  To do this, undefine xxx_SUPPORTED symbols as necessary.
+
+You can also save a few K by not having text error messages in the library;
+the standard error message table occupies about 5Kb.  This is particularly
+reasonable for embedded applications where there's no good way to display
+a message anyway.  To do this, remove the creation of the message table
+(jpeg_std_message_table) from jerror.c, and alter format_message to do
+something reasonable without it.  You could output the numeric value of the
+message code number, for example.  If you do this, you can also save a couple
+more K by modifying the TRACEMSn() macros in jerror.h to expand to nothing;
+you don't need trace capability anyway, right?
+
+
+Portability considerations
+--------------------------
+
+The JPEG library has been written to be extremely portable; the sample
+applications cjpeg and djpeg are slightly less so.  This section summarizes
+the design goals in this area.  (If you encounter any bugs that cause the
+library to be less portable than is claimed here, we'd appreciate hearing
+about them.)
+
+The code works fine on ANSI C and C++ compilers, using any of the popular
+system include file setups, and some not-so-popular ones too.
+
+The code is not dependent on the exact sizes of the C data types.  As
+distributed, we make the assumptions that
+        char    is at least 8 bits wide
+        short   is at least 16 bits wide
+        int     is at least 16 bits wide
+        long    is at least 32 bits wide
+(These are the minimum requirements of the ANSI C standard.)  Wider types will
+work fine, although memory may be used inefficiently if char is much larger
+than 8 bits or short is much bigger than 16 bits.  The code should work
+equally well with 16- or 32-bit ints.
+
+In a system where these assumptions are not met, you may be able to make the
+code work by modifying the typedefs in jmorecfg.h.  However, you will probably
+have difficulty if int is less than 16 bits wide, since references to plain
+int abound in the code.
+
+char can be either signed or unsigned, although the code runs faster if an
+unsigned char type is available.  If char is wider than 8 bits, you will need
+to redefine JOCTET and/or provide custom data source/destination managers so
+that JOCTET represents exactly 8 bits of data on external storage.
+
+The JPEG library proper does not assume ASCII representation of characters.
+But some of the image file I/O modules in cjpeg/djpeg do have ASCII
+dependencies in file-header manipulation; so does cjpeg's select_file_type()
+routine.
+
+The JPEG library does not rely heavily on the C library.  In particular, C
+stdio is used only by the data source/destination modules and the error
+handler, all of which are application-replaceable.  (cjpeg/djpeg are more
+heavily dependent on stdio.)  malloc and free are called only from the memory
+manager "back end" module, so you can use a different memory allocator by
+replacing that one file.
+
+More info about porting the code may be gleaned by reading jconfig.txt,
+jmorecfg.h, and jinclude.h.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/md5 Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/md5/CMakeLists.txt Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/md5/md5.c Added

@@ -0,0 +1,275 @@
+/*
+ * This code implements the MD5 message-digest algorithm.
+ * The algorithm is due to Ron Rivest.  This code was
+ * written by Colin Plumb in 1993, no copyright is claimed.
+ * This code is in the public domain; do with it what you wish.
+ *
+ * Equivalent code is available from RSA Data Security, Inc.
+ * This code has been tested against that, and is equivalent,
+ * except that you don't need to include two pages of legalese
+ * with every copy.
+ *
+ * To compute the message digest of a chunk of bytes, declare an
+ * MD5Context structure, pass it to MD5Init, call MD5Update as
+ * needed on buffers full of bytes, and then call MD5Final, which
+ * will fill a supplied 16-byte array with the digest.
+ * ----------------------------------------------------------------------------
+ * libjpeg-turbo Modifications:
+ * Copyright (C)2018, D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ * ----------------------------------------------------------------------------
+ */
+
+#include <string.h>             /* for memcpy() */
+#include "md5.h"
+
+#if (BYTE_ORDER == LITTLE_ENDIAN)
+#define byteReverse(buf, len)   /* Nothing */
+#else
+/*
+ * Note: this code is harmless on little-endian machines.
+ */
+static void byteReverse(unsigned char *buf, unsigned int longs)
+{
+  uint32 t;
+  do {
+    t = (uint32)((unsigned int)buf3 << 8 | buf2) << 16 |
+        ((unsigned int)buf1 << 8 | buf0);
+    *(uint32 *)buf = t;
+    buf += 4;
+  } while (--longs);
+}
+#endif
+
+/*
+ * Start MD5 accumulation.  Set bit count to 0 and buffer to mysterious
+ * initialization constants.
+ */
+void MD5Init(struct MD5Context *ctx)
+{
+  ctx->buf0 = 0x67452301;
+  ctx->buf1 = 0xefcdab89;
+  ctx->buf2 = 0x98badcfe;
+  ctx->buf3 = 0x10325476;
+
+  ctx->bits0 = 0;
+  ctx->bits1 = 0;
+}
+
+/*
+ * Update context to reflect the concatenation of another buffer full
+ * of bytes.
+ */
+void MD5Update(struct MD5Context *ctx, unsigned char *buf, unsigned int len)
+{
+  uint32 t;
+
+  /* Update bitcount */
+
+  t = ctx->bits0;
+  if ((ctx->bits0 = t + ((uint32)len << 3)) < t)
+    ctx->bits1++;             /* Carry from low to high */
+  ctx->bits1 += len >> 29;
+
+  t = (t >> 3) & 0x3f;          /* Bytes already in shsInfo->data */
+
+  /* Handle any leading odd-sized chunks */
+
+  if (t) {
+    unsigned char *p = (unsigned char *)ctx->in + t;
+
+    t = 64 - t;
+    if (len < t) {
+      memcpy(p, buf, len);
+      return;
+    }
+    memcpy(p, buf, t);
+    byteReverse(ctx->in, 16);
+    MD5Transform(ctx->buf, (uint32 *)ctx->in);
+    buf += t;
+    len -= t;
+  }
+  /* Process data in 64-byte chunks */
+
+  while (len >= 64) {
+    memcpy(ctx->in, buf, 64);
+    byteReverse(ctx->in, 16);
+    MD5Transform(ctx->buf, (uint32 *)ctx->in);
+    buf += 64;
+    len -= 64;
+  }
+
+  /* Handle any remaining bytes of data. */
+
+  memcpy(ctx->in, buf, len);
+}
+
+/*
+ * Final wrapup - pad to 64-byte boundary with the bit pattern
+ * 1 0* (64-bit count of bits processed, MSB-first)
+ */
+void MD5Final(unsigned char digest16, struct MD5Context *ctx)
+{
+  unsigned int count;
+  unsigned char *p;
+  uint32 *in32 = (uint32 *)ctx->in;
+
+  /* Compute number of bytes mod 64 */
+  count = (ctx->bits0 >> 3) & 0x3F;
+
+  /* Set the first char of padding to 0x80.  This is safe since there is
+     always at least one byte free */
+  p = ctx->in + count;
+  *p++ = 0x80;
+
+  /* Bytes of padding needed to make 64 bytes */
+  count = 64 - 1 - count;
+
+  /* Pad out to 56 mod 64 */
+  if (count < 8) {
+    /* Two lots of padding:  Pad the first block to 64 bytes */
+    memset(p, 0, count);
+    byteReverse(ctx->in, 16);
+    MD5Transform(ctx->buf, (uint32 *)ctx->in);
+
+    /* Now fill the next block with 56 bytes */
+    memset(ctx->in, 0, 56);
+  } else {
+    /* Pad block to 56 bytes */
+    memset(p, 0, count - 8);
+  }
+  byteReverse(ctx->in, 14);
+
+  /* Append length in bits and transform */
+  in3214 = ctx->bits0;
+  in3215 = ctx->bits1;
+
+  MD5Transform(ctx->buf, (uint32 *)ctx->in);
+  byteReverse((unsigned char *)ctx->buf, 4);
+  memcpy(digest, ctx->buf, 16);
+  memset(ctx, 0, sizeof(struct MD5Context));     /* In case it's sensitive */
+}
+
+
+/* The four core functions - F1 is optimized somewhat */
+
+/* #define F1(x, y, z) (x & y | ~x & z) */
+#define F1(x, y, z)  (z ^ (x & (y ^ z)))
+#define F2(x, y, z)  F1(z, x, y)
+#define F3(x, y, z)  (x ^ y ^ z)
+#define F4(x, y, z)  (y ^ (x | ~z))
+
+/* This is the central step in the MD5 algorithm. */
+#define MD5STEP(f, w, x, y, z, data, s) \
+  ( w += f(x, y, z) + data,  w = w << s | w >> (32 - s),  w += x )
+
+/*
+ * The core of the MD5 algorithm, this alters an existing MD5 hash to
+ * reflect the addition of 16 longwords of new data.  MD5Update blocks
+ * the data and converts bytes into longwords for this routine.
+ */
+void MD5Transform(uint32 buf4, uint32 in16)
+{
+  register uint32 a, b, c, d;
+
+  a = buf0;
+  b = buf1;
+  c = buf2;
+  d = buf3;
+
+  MD5STEP(F1, a, b, c, d, in0 + 0xd76aa478, 7);
+  MD5STEP(F1, d, a, b, c, in1 + 0xe8c7b756, 12);
+  MD5STEP(F1, c, d, a, b, in2 + 0x242070db, 17);
+  MD5STEP(F1, b, c, d, a, in3 + 0xc1bdceee, 22);
+  MD5STEP(F1, a, b, c, d, in4 + 0xf57c0faf, 7);
+  MD5STEP(F1, d, a, b, c, in5 + 0x4787c62a, 12);
+  MD5STEP(F1, c, d, a, b, in6 + 0xa8304613, 17);
+  MD5STEP(F1, b, c, d, a, in7 + 0xfd469501, 22);
+  MD5STEP(F1, a, b, c, d, in8 + 0x698098d8, 7);
+  MD5STEP(F1, d, a, b, c, in9 + 0x8b44f7af, 12);
+  MD5STEP(F1, c, d, a, b, in10 + 0xffff5bb1, 17);
+  MD5STEP(F1, b, c, d, a, in11 + 0x895cd7be, 22);
+  MD5STEP(F1, a, b, c, d, in12 + 0x6b901122, 7);
+  MD5STEP(F1, d, a, b, c, in13 + 0xfd987193, 12);
+  MD5STEP(F1, c, d, a, b, in14 + 0xa679438e, 17);
+  MD5STEP(F1, b, c, d, a, in15 + 0x49b40821, 22);
+
+  MD5STEP(F2, a, b, c, d, in1 + 0xf61e2562, 5);
+  MD5STEP(F2, d, a, b, c, in6 + 0xc040b340, 9);
+  MD5STEP(F2, c, d, a, b, in11 + 0x265e5a51, 14);
+  MD5STEP(F2, b, c, d, a, in0 + 0xe9b6c7aa, 20);
+  MD5STEP(F2, a, b, c, d, in5 + 0xd62f105d, 5);
+  MD5STEP(F2, d, a, b, c, in10 + 0x02441453, 9);
+  MD5STEP(F2, c, d, a, b, in15 + 0xd8a1e681, 14);
+  MD5STEP(F2, b, c, d, a, in4 + 0xe7d3fbc8, 20);
+  MD5STEP(F2, a, b, c, d, in9 + 0x21e1cde6, 5);
+  MD5STEP(F2, d, a, b, c, in14 + 0xc33707d6, 9);
+  MD5STEP(F2, c, d, a, b, in3 + 0xf4d50d87, 14);
+  MD5STEP(F2, b, c, d, a, in8 + 0x455a14ed, 20);
+  MD5STEP(F2, a, b, c, d, in13 + 0xa9e3e905, 5);
+  MD5STEP(F2, d, a, b, c, in2 + 0xfcefa3f8, 9);
+  MD5STEP(F2, c, d, a, b, in7 + 0x676f02d9, 14);
+  MD5STEP(F2, b, c, d, a, in12 + 0x8d2a4c8a, 20);
+
+  MD5STEP(F3, a, b, c, d, in5 + 0xfffa3942, 4);
+  MD5STEP(F3, d, a, b, c, in8 + 0x8771f681, 11);
+  MD5STEP(F3, c, d, a, b, in11 + 0x6d9d6122, 16);
+  MD5STEP(F3, b, c, d, a, in14 + 0xfde5380c, 23);
+  MD5STEP(F3, a, b, c, d, in1 + 0xa4beea44, 4);
+  MD5STEP(F3, d, a, b, c, in4 + 0x4bdecfa9, 11);
+  MD5STEP(F3, c, d, a, b, in7 + 0xf6bb4b60, 16);
+  MD5STEP(F3, b, c, d, a, in10 + 0xbebfbc70, 23);
+  MD5STEP(F3, a, b, c, d, in13 + 0x289b7ec6, 4);
+  MD5STEP(F3, d, a, b, c, in0 + 0xeaa127fa, 11);
+  MD5STEP(F3, c, d, a, b, in3 + 0xd4ef3085, 16);
+  MD5STEP(F3, b, c, d, a, in6 + 0x04881d05, 23);
+  MD5STEP(F3, a, b, c, d, in9 + 0xd9d4d039, 4);
+  MD5STEP(F3, d, a, b, c, in12 + 0xe6db99e5, 11);
+  MD5STEP(F3, c, d, a, b, in15 + 0x1fa27cf8, 16);
+  MD5STEP(F3, b, c, d, a, in2 + 0xc4ac5665, 23);
+
+  MD5STEP(F4, a, b, c, d, in0 + 0xf4292244, 6);
+  MD5STEP(F4, d, a, b, c, in7 + 0x432aff97, 10);
+  MD5STEP(F4, c, d, a, b, in14 + 0xab9423a7, 15);
+  MD5STEP(F4, b, c, d, a, in5 + 0xfc93a039, 21);
+  MD5STEP(F4, a, b, c, d, in12 + 0x655b59c3, 6);
+  MD5STEP(F4, d, a, b, c, in3 + 0x8f0ccc92, 10);
+  MD5STEP(F4, c, d, a, b, in10 + 0xffeff47d, 15);
+  MD5STEP(F4, b, c, d, a, in1 + 0x85845dd1, 21);
+  MD5STEP(F4, a, b, c, d, in8 + 0x6fa87e4f, 6);
+  MD5STEP(F4, d, a, b, c, in15 + 0xfe2ce6e0, 10);
+  MD5STEP(F4, c, d, a, b, in6 + 0xa3014314, 15);
+  MD5STEP(F4, b, c, d, a, in13 + 0x4e0811a1, 21);
+  MD5STEP(F4, a, b, c, d, in4 + 0xf7537e82, 6);
+  MD5STEP(F4, d, a, b, c, in11 + 0xbd3af235, 10);
+  MD5STEP(F4, c, d, a, b, in2 + 0x2ad7d2bb, 15);
+  MD5STEP(F4, b, c, d, a, in9 + 0xeb86d391, 21);
+
+  buf0 += a;
+  buf1 += b;
+  buf2 += c;
+  buf3 += d;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/md5/md5.h Added

@@ -0,0 +1,57 @@
+/*
+ * libjpeg-turbo Modifications:
+ * Copyright (C)2018 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef MD5_H
+#define MD5_H
+
+#include <sys/types.h>
+#ifdef __amigaos4__
+#include <machine/endian.h>
+#endif
+
+/*  On machines where "long" is 64 bits, we need to declare
+    uint32 as something guaranteed to be 32 bits.  */
+
+typedef unsigned int uint32;
+
+typedef struct MD5Context {
+  uint32 buf4;
+  uint32 bits2;
+  unsigned char in64;
+} MD5_CTX;
+
+extern void MD5Init(struct MD5Context *ctx);
+extern void MD5Update(struct MD5Context *ctx, unsigned char *buf,
+                      unsigned int len);
+extern void MD5Final(unsigned char digest16, struct MD5Context *ctx);
+extern void MD5Transform(uint32 buf4, uint32 in16);
+extern char *MD5File(const char *, char *);
+extern char *MD5FileChunk(const char *, char *, off_t, off_t);
+
+#endif /* !MD5_H */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/md5/md5cmp.c Added

@@ -0,0 +1,59 @@
+/*
+ * Copyright (C)2013, 2016 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+#include "./md5.h"
+#include "../tjutil.h"
+
+int main(int argc, char *argv)
+{
+  char *md5sum = NULL, buf65;
+
+  if (argc < 3) {
+    fprintf(stderr, "USAGE: %s <correct MD5 sum> <file>\n", argv0);
+    return -1;
+  }
+
+  if (strlen(argv1) != 32)
+    fprintf(stderr, "WARNING: MD5 hash size is wrong.\n");
+
+  md5sum = MD5File(argv2, buf);
+  if (!md5sum) {
+    perror("Could not obtain MD5 sum");
+    return -1;
+  }
+
+  if (!strcasecmp(md5sum, argv1)) {
+    fprintf(stderr, "%s: OK\n", argv2);
+    return 0;
+  } else {
+    fprintf(stderr, "%s: FAILED.  Checksum is %s\n", argv2, md5sum);
+    return -1;
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/md5/md5hl.c Added

@@ -0,0 +1,125 @@
+/* mdXhl.c
+ * ----------------------------------------------------------------------------
+ * "THE BEER-WARE LICENSE" (Revision 42):
+ * <phk@FreeBSD.org> wrote this file.  As long as you retain this notice you
+ * can do whatever you want with this stuff. If we meet some day, and you think
+ * this stuff is worth it, you can buy me a beer in return.   Poul-Henning Kamp
+ * ----------------------------------------------------------------------------
+ * libjpeg-turbo Modifications:
+ * Copyright (C)2016, 2018-2019 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ * ----------------------------------------------------------------------------
+ */
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#ifdef _WIN32
+#include <io.h>
+#define close  _close
+#define fstat  _fstat
+#define lseek  _lseek
+#define read  _read
+#define stat  _stat
+#else
+#include <unistd.h>
+#endif
+
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#define LENGTH  16
+
+#include "./md5.h"
+
+static char *MD5End(MD5_CTX *ctx, char *buf)
+{
+  int i;
+  unsigned char digestLENGTH;
+  static const char hex = "0123456789abcdef";
+
+  if (!buf)
+    buf = malloc(2 * LENGTH + 1);
+  if (!buf)
+    return 0;
+  MD5Final(digest, ctx);
+  for (i = 0; i < LENGTH; i++) {
+    bufi + i = hexdigesti >> 4;
+    bufi + i + 1 = hexdigesti & 0x0f;
+  }
+  bufi + i = '\0';
+  return buf;
+}
+
+char *MD5File(const char *filename, char *buf)
+{
+  return (MD5FileChunk(filename, buf, 0, 0));
+}
+
+char *MD5FileChunk(const char *filename, char *buf, off_t ofs, off_t len)
+{
+  unsigned char bufferBUFSIZ;
+  MD5_CTX ctx;
+  struct stat stbuf;
+  int f, i, e;
+  off_t n;
+
+  MD5Init(&ctx);
+#ifdef _WIN32
+  f = _open(filename, O_RDONLY | O_BINARY);
+#else
+  f = open(filename, O_RDONLY);
+#endif
+  if (f < 0)
+    return 0;
+  if (fstat(f, &stbuf) < 0)
+    return 0;
+  if (ofs > stbuf.st_size)
+    ofs = stbuf.st_size;
+  if ((len == 0) || (len > stbuf.st_size - ofs))
+    len = stbuf.st_size - ofs;
+  if (lseek(f, ofs, SEEK_SET) < 0)
+    return 0;
+  n = len;
+  i = 0;
+  while (n > 0) {
+    if (n > sizeof(buffer))
+      i = read(f, buffer, sizeof(buffer));
+    else
+      i = read(f, buffer, n);
+    if (i < 0)
+      break;
+    MD5Update(&ctx, buffer, i);
+    n -= i;
+  }
+  e = errno;
+  close(f);
+  errno = e;
+  if (i < 0)
+    return 0;
+  return (MD5End(&ctx, buf));
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/rdbmp.c Added

@@ -0,0 +1,689 @@
+/*
+ * rdbmp.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2017 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Modified 2011 by Siarhei Siamashka.
+ * Copyright (C) 2015, 2017-2018, 2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in Microsoft "BMP"
+ * format (MS Windows 3.x, OS/2 1.x, and OS/2 2.x flavors).
+ * Currently, only 8-, 24-, and 32-bit images are supported, not 1-bit or
+ * 4-bit (feeding such low-depth images into JPEG would be silly anyway).
+ * Also, we don't support RLE-compressed files.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed BMP format).
+ *
+ * This code contributed by James Arthur Boucher.
+ */
+
+#include "cmyk.h"
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef BMP_SUPPORTED
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int)(x))
+
+
+#define ReadOK(file, buffer, len) \
+  (JFREAD(file, buffer, len) == ((size_t)(len)))
+
+static int alpha_indexJPEG_NUMCS = {
+  -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 3, 0, 0, -1
+};
+
+
+/* Private version of data source object */
+
+typedef struct _bmp_source_struct *bmp_source_ptr;
+
+typedef struct _bmp_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  j_compress_ptr cinfo;         /* back link saves passing separate parm */
+
+  JSAMPARRAY colormap;          /* BMP colormap (converted to my format) */
+
+  jvirt_sarray_ptr whole_image; /* Needed to reverse row order */
+  JDIMENSION source_row;        /* Current source row number */
+  JDIMENSION row_width;         /* Physical width of scanlines in file */
+
+  int bits_per_pixel;           /* remembers 8-, 24-, or 32-bit format */
+  int cmap_length;              /* colormap length */
+
+  boolean use_inversion_array;  /* TRUE = preload the whole image, which is
+                                   stored in bottom-up order, and feed it to
+                                   the calling program in top-down order
+
+                                   FALSE = the calling program will maintain
+                                   its own image buffer and read the rows in
+                                   bottom-up order */
+
+  U_CHAR *iobuffer;             /* I/O buffer (used to buffer a single row from
+                                   disk if use_inversion_array == FALSE) */
+} bmp_source_struct;
+
+
+LOCAL(int)
+read_byte(bmp_source_ptr sinfo)
+/* Read next byte from BMP file */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int c;
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  return c;
+}
+
+
+LOCAL(void)
+read_colormap(bmp_source_ptr sinfo, int cmaplen, int mapentrysize)
+/* Read the colormap from a BMP file */
+{
+  int i, gray = 1;
+
+  switch (mapentrysize) {
+  case 3:
+    /* BGR format (occurs in OS/2 files) */
+    for (i = 0; i < cmaplen; i++) {
+      sinfo->colormap2i = (JSAMPLE)read_byte(sinfo);
+      sinfo->colormap1i = (JSAMPLE)read_byte(sinfo);
+      sinfo->colormap0i = (JSAMPLE)read_byte(sinfo);
+      if (sinfo->colormap2i != sinfo->colormap1i ||
+          sinfo->colormap1i != sinfo->colormap0i)
+        gray = 0;
+    }
+    break;
+  case 4:
+    /* BGR0 format (occurs in MS Windows files) */
+    for (i = 0; i < cmaplen; i++) {
+      sinfo->colormap2i = (JSAMPLE)read_byte(sinfo);
+      sinfo->colormap1i = (JSAMPLE)read_byte(sinfo);
+      sinfo->colormap0i = (JSAMPLE)read_byte(sinfo);
+      (void)read_byte(sinfo);
+      if (sinfo->colormap2i != sinfo->colormap1i ||
+          sinfo->colormap1i != sinfo->colormap0i)
+        gray = 0;
+    }
+    break;
+  default:
+    ERREXIT(sinfo->cinfo, JERR_BMP_BADCMAP);
+    break;
+  }
+
+  if (sinfo->cinfo->in_color_space == JCS_UNKNOWN && gray)
+    sinfo->cinfo->in_color_space = JCS_GRAYSCALE;
+
+  if (sinfo->cinfo->in_color_space == JCS_GRAYSCALE && !gray)
+    ERREXIT(sinfo->cinfo, JERR_BAD_IN_COLORSPACE);
+}
+
+
+/*
+ * Read one row of pixels.
+ * The image has been read into the whole_image array, but is otherwise
+ * unprocessed.  We must read it out in top-to-bottom row order, and if
+ * it is an 8-bit image, we must expand colormapped pixels to 24bit format.
+ */
+
+METHODDEF(JDIMENSION)
+get_8bit_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit colormap indexes */
+{
+  bmp_source_ptr source = (bmp_source_ptr)sinfo;
+  register JSAMPARRAY colormap = source->colormap;
+  int cmaplen = source->cmap_length;
+  JSAMPARRAY image_ptr;
+  register int t;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  if (source->use_inversion_array) {
+    /* Fetch next row from virtual array */
+    source->source_row--;
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr)cinfo, source->whole_image,
+       source->source_row, (JDIMENSION)1, FALSE);
+    inptr = image_ptr0;
+  } else {
+    if (!ReadOK(source->pub.input_file, source->iobuffer, source->row_width))
+      ERREXIT(cinfo, JERR_INPUT_EOF);
+    inptr = source->iobuffer;
+  }
+
+  /* Expand the colormap indexes to real data */
+  outptr = source->pub.buffer0;
+  if (cinfo->in_color_space == JCS_GRAYSCALE) {
+    for (col = cinfo->image_width; col > 0; col--) {
+      t = *inptr++;
+      if (t >= cmaplen)
+        ERREXIT(cinfo, JERR_BMP_OUTOFRANGE);
+      *outptr++ = colormap0t;
+    }
+  } else if (cinfo->in_color_space == JCS_CMYK) {
+    for (col = cinfo->image_width; col > 0; col--) {
+      t = *inptr++;
+      if (t >= cmaplen)
+        ERREXIT(cinfo, JERR_BMP_OUTOFRANGE);
+      rgb_to_cmyk(colormap0t, colormap1t, colormap2t, outptr,
+                  outptr + 1, outptr + 2, outptr + 3);
+      outptr += 4;
+    }
+  } else {
+    register int rindex = rgb_redcinfo->in_color_space;
+    register int gindex = rgb_greencinfo->in_color_space;
+    register int bindex = rgb_bluecinfo->in_color_space;
+    register int aindex = alpha_indexcinfo->in_color_space;
+    register int ps = rgb_pixelsizecinfo->in_color_space;
+
+    if (aindex >= 0) {
+      for (col = cinfo->image_width; col > 0; col--) {
+        t = *inptr++;
+        if (t >= cmaplen)
+          ERREXIT(cinfo, JERR_BMP_OUTOFRANGE);
+        outptrrindex = colormap0t;
+        outptrgindex = colormap1t;
+        outptrbindex = colormap2t;
+        outptraindex = 0xFF;
+        outptr += ps;
+      }
+    } else {
+      for (col = cinfo->image_width; col > 0; col--) {
+        t = *inptr++;
+        if (t >= cmaplen)
+          ERREXIT(cinfo, JERR_BMP_OUTOFRANGE);
+        outptrrindex = colormap0t;
+        outptrgindex = colormap1t;
+        outptrbindex = colormap2t;
+        outptr += ps;
+      }
+    }
+  }
+
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_24bit_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 24-bit pixels */
+{
+  bmp_source_ptr source = (bmp_source_ptr)sinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  if (source->use_inversion_array) {
+    /* Fetch next row from virtual array */
+    source->source_row--;
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr)cinfo, source->whole_image,
+       source->source_row, (JDIMENSION)1, FALSE);
+    inptr = image_ptr0;
+  } else {
+    if (!ReadOK(source->pub.input_file, source->iobuffer, source->row_width))
+      ERREXIT(cinfo, JERR_INPUT_EOF);
+    inptr = source->iobuffer;
+  }
+
+  /* Transfer data.  Note source values are in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  outptr = source->pub.buffer0;
+  if (cinfo->in_color_space == JCS_EXT_BGR) {
+    MEMCOPY(outptr, inptr, source->row_width);
+  } else if (cinfo->in_color_space == JCS_CMYK) {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE b = *inptr++, g = *inptr++, r = *inptr++;
+      rgb_to_cmyk(r, g, b, outptr, outptr + 1, outptr + 2, outptr + 3);
+      outptr += 4;
+    }
+  } else {
+    register int rindex = rgb_redcinfo->in_color_space;
+    register int gindex = rgb_greencinfo->in_color_space;
+    register int bindex = rgb_bluecinfo->in_color_space;
+    register int aindex = alpha_indexcinfo->in_color_space;
+    register int ps = rgb_pixelsizecinfo->in_color_space;
+
+    if (aindex >= 0) {
+      for (col = cinfo->image_width; col > 0; col--) {
+        outptrbindex = *inptr++;
+        outptrgindex = *inptr++;
+        outptrrindex = *inptr++;
+        outptraindex = 0xFF;
+        outptr += ps;
+      }
+    } else {
+      for (col = cinfo->image_width; col > 0; col--) {
+        outptrbindex = *inptr++;
+        outptrgindex = *inptr++;
+        outptrrindex = *inptr++;
+        outptr += ps;
+      }
+    }
+  }
+
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_32bit_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 32-bit pixels */
+{
+  bmp_source_ptr source = (bmp_source_ptr)sinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  if (source->use_inversion_array) {
+    /* Fetch next row from virtual array */
+    source->source_row--;
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr)cinfo, source->whole_image,
+       source->source_row, (JDIMENSION)1, FALSE);
+    inptr = image_ptr0;
+  } else {
+    if (!ReadOK(source->pub.input_file, source->iobuffer, source->row_width))
+      ERREXIT(cinfo, JERR_INPUT_EOF);
+    inptr = source->iobuffer;
+  }
+
+  /* Transfer data.  Note source values are in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  outptr = source->pub.buffer0;
+  if (cinfo->in_color_space == JCS_EXT_BGRX ||
+      cinfo->in_color_space == JCS_EXT_BGRA) {
+    MEMCOPY(outptr, inptr, source->row_width);
+  } else if (cinfo->in_color_space == JCS_CMYK) {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE b = *inptr++, g = *inptr++, r = *inptr++;
+      rgb_to_cmyk(r, g, b, outptr, outptr + 1, outptr + 2, outptr + 3);
+      inptr++;                          /* skip the 4th byte (Alpha channel) */
+      outptr += 4;
+    }
+  } else {
+    register int rindex = rgb_redcinfo->in_color_space;
+    register int gindex = rgb_greencinfo->in_color_space;
+    register int bindex = rgb_bluecinfo->in_color_space;
+    register int aindex = alpha_indexcinfo->in_color_space;
+    register int ps = rgb_pixelsizecinfo->in_color_space;
+
+    if (aindex >= 0) {
+      for (col = cinfo->image_width; col > 0; col--) {
+        outptrbindex = *inptr++;
+        outptrgindex = *inptr++;
+        outptrrindex = *inptr++;
+        outptraindex = *inptr++;
+        outptr += ps;
+      }
+    } else {
+      for (col = cinfo->image_width; col > 0; col--) {
+        outptrbindex = *inptr++;
+        outptrgindex = *inptr++;
+        outptrrindex = *inptr++;
+        inptr++;                        /* skip the 4th byte (Alpha channel) */
+        outptr += ps;
+      }
+    }
+  }
+
+  return 1;
+}
+
+
+/*
+ * This method loads the image into whole_image during the first call on
+ * get_pixel_rows.  The get_pixel_rows pointer is then adjusted to call
+ * get_8bit_row, get_24bit_row, or get_32bit_row on subsequent calls.
+ */
+
+METHODDEF(JDIMENSION)
+preload_image(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  bmp_source_ptr source = (bmp_source_ptr)sinfo;
+  register FILE *infile = source->pub.input_file;
+  register JSAMPROW out_ptr;
+  JSAMPARRAY image_ptr;
+  JDIMENSION row;
+  cd_progress_ptr progress = (cd_progress_ptr)cinfo->progress;
+
+  /* Read the data into a virtual array in input-file row order. */
+  for (row = 0; row < cinfo->image_height; row++) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long)row;
+      progress->pub.pass_limit = (long)cinfo->image_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr)cinfo);
+    }
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr)cinfo, source->whole_image, row, (JDIMENSION)1, TRUE);
+    out_ptr = image_ptr0;
+    if (fread(out_ptr, 1, source->row_width, infile) != source->row_width) {
+      if (feof(infile))
+        ERREXIT(cinfo, JERR_INPUT_EOF);
+      else
+        ERREXIT(cinfo, JERR_FILE_READ);
+    }
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Set up to read from the virtual array in top-to-bottom order */
+  switch (source->bits_per_pixel) {
+  case 8:
+    source->pub.get_pixel_rows = get_8bit_row;
+    break;
+  case 24:
+    source->pub.get_pixel_rows = get_24bit_row;
+    break;
+  case 32:
+    source->pub.get_pixel_rows = get_32bit_row;
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+  }
+  source->source_row = cinfo->image_height;
+
+  /* And read the first row */
+  return (*source->pub.get_pixel_rows) (cinfo, sinfo);
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_bmp(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  bmp_source_ptr source = (bmp_source_ptr)sinfo;
+  U_CHAR bmpfileheader14;
+  U_CHAR bmpinfoheader64;
+
+#define GET_2B(array, offset) \
+  ((unsigned short)UCH(arrayoffset) + \
+   (((unsigned short)UCH(arrayoffset + 1)) << 8))
+#define GET_4B(array, offset) \
+  ((unsigned int)UCH(arrayoffset) + \
+   (((unsigned int)UCH(arrayoffset + 1)) << 8) + \
+   (((unsigned int)UCH(arrayoffset + 2)) << 16) + \
+   (((unsigned int)UCH(arrayoffset + 3)) << 24))
+
+  int bfOffBits;
+  int headerSize;
+  int biWidth;
+  int biHeight;
+  unsigned short biPlanes;
+  unsigned int biCompression;
+  int biXPelsPerMeter, biYPelsPerMeter;
+  int biClrUsed = 0;
+  int mapentrysize = 0;         /* 0 indicates no colormap */
+  int bPad;
+  JDIMENSION row_width = 0;
+
+  /* Read and verify the bitmap file header */
+  if (!ReadOK(source->pub.input_file, bmpfileheader, 14))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  if (GET_2B(bmpfileheader, 0) != 0x4D42) /* 'BM' */
+    ERREXIT(cinfo, JERR_BMP_NOT);
+  bfOffBits = GET_4B(bmpfileheader, 10);
+  /* We ignore the remaining fileheader fields */
+
+  /* The infoheader might be 12 bytes (OS/2 1.x), 40 bytes (Windows),
+   * or 64 bytes (OS/2 2.x).  Check the first 4 bytes to find out which.
+   */
+  if (!ReadOK(source->pub.input_file, bmpinfoheader, 4))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  headerSize = GET_4B(bmpinfoheader, 0);
+  if (headerSize < 12 || headerSize > 64 || (headerSize + 14) > bfOffBits)
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+  if (!ReadOK(source->pub.input_file, bmpinfoheader + 4, headerSize - 4))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+
+  switch (headerSize) {
+  case 12:
+    /* Decode OS/2 1.x header (Microsoft calls this a BITMAPCOREHEADER) */
+    biWidth = (int)GET_2B(bmpinfoheader, 4);
+    biHeight = (int)GET_2B(bmpinfoheader, 6);
+    biPlanes = GET_2B(bmpinfoheader, 8);
+    source->bits_per_pixel = (int)GET_2B(bmpinfoheader, 10);
+
+    switch (source->bits_per_pixel) {
+    case 8:                     /* colormapped image */
+      mapentrysize = 3;         /* OS/2 uses RGBTRIPLE colormap */
+      TRACEMS2(cinfo, 1, JTRC_BMP_OS2_MAPPED, biWidth, biHeight);
+      break;
+    case 24:                    /* RGB image */
+    case 32:                    /* RGB image + Alpha channel */
+      TRACEMS3(cinfo, 1, JTRC_BMP_OS2, biWidth, biHeight,
+               source->bits_per_pixel);
+      break;
+    default:
+      ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+      break;
+    }
+    break;
+  case 40:
+  case 64:
+    /* Decode Windows 3.x header (Microsoft calls this a BITMAPINFOHEADER) */
+    /* or OS/2 2.x header, which has additional fields that we ignore */
+    biWidth = (int)GET_4B(bmpinfoheader, 4);
+    biHeight = (int)GET_4B(bmpinfoheader, 8);
+    biPlanes = GET_2B(bmpinfoheader, 12);
+    source->bits_per_pixel = (int)GET_2B(bmpinfoheader, 14);
+    biCompression = GET_4B(bmpinfoheader, 16);
+    biXPelsPerMeter = (int)GET_4B(bmpinfoheader, 24);
+    biYPelsPerMeter = (int)GET_4B(bmpinfoheader, 28);
+    biClrUsed = GET_4B(bmpinfoheader, 32);
+    /* biSizeImage, biClrImportant fields are ignored */
+
+    switch (source->bits_per_pixel) {
+    case 8:                     /* colormapped image */
+      mapentrysize = 4;         /* Windows uses RGBQUAD colormap */
+      TRACEMS2(cinfo, 1, JTRC_BMP_MAPPED, biWidth, biHeight);
+      break;
+    case 24:                    /* RGB image */
+    case 32:                    /* RGB image + Alpha channel */
+      TRACEMS3(cinfo, 1, JTRC_BMP, biWidth, biHeight, source->bits_per_pixel);
+      break;
+    default:
+      ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+      break;
+    }
+    if (biCompression != 0)
+      ERREXIT(cinfo, JERR_BMP_COMPRESSED);
+
+    if (biXPelsPerMeter > 0 && biYPelsPerMeter > 0) {
+      /* Set JFIF density parameters from the BMP data */
+      cinfo->X_density = (UINT16)(biXPelsPerMeter / 100); /* 100 cm per meter */
+      cinfo->Y_density = (UINT16)(biYPelsPerMeter / 100);
+      cinfo->density_unit = 2;  /* dots/cm */
+    }
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+    return;
+  }
+
+  if (biWidth <= 0 || biHeight <= 0)
+    ERREXIT(cinfo, JERR_BMP_EMPTY);
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  if (sinfo->max_pixels &&
+      (unsigned long long)biWidth * biHeight > sinfo->max_pixels)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+#endif
+  if (biPlanes != 1)
+    ERREXIT(cinfo, JERR_BMP_BADPLANES);
+
+  /* Compute distance to bitmap data --- will adjust for colormap below */
+  bPad = bfOffBits - (headerSize + 14);
+
+  /* Read the colormap, if any */
+  if (mapentrysize > 0) {
+    if (biClrUsed <= 0)
+      biClrUsed = 256;          /* assume it's 256 */
+    else if (biClrUsed > 256)
+      ERREXIT(cinfo, JERR_BMP_BADCMAP);
+    /* Allocate space to store the colormap */
+    source->colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE, (JDIMENSION)biClrUsed, (JDIMENSION)3);
+    source->cmap_length = (int)biClrUsed;
+    /* and read it from the file */
+    read_colormap(source, (int)biClrUsed, mapentrysize);
+    /* account for size of colormap */
+    bPad -= biClrUsed * mapentrysize;
+  }
+
+  /* Skip any remaining pad bytes */
+  if (bPad < 0)                 /* incorrect bfOffBits value? */
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+  while (--bPad >= 0) {
+    (void)read_byte(source);
+  }
+
+  /* Compute row width in file, including padding to 4-byte boundary */
+  switch (source->bits_per_pixel) {
+  case 8:
+    if (cinfo->in_color_space == JCS_UNKNOWN)
+      cinfo->in_color_space = JCS_EXT_RGB;
+    if (IsExtRGB(cinfo->in_color_space))
+      cinfo->input_components = rgb_pixelsizecinfo->in_color_space;
+    else if (cinfo->in_color_space == JCS_GRAYSCALE)
+      cinfo->input_components = 1;
+    else if (cinfo->in_color_space == JCS_CMYK)
+      cinfo->input_components = 4;
+    else
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    row_width = (JDIMENSION)biWidth;
+    break;
+  case 24:
+    if (cinfo->in_color_space == JCS_UNKNOWN)
+      cinfo->in_color_space = JCS_EXT_BGR;
+    if (IsExtRGB(cinfo->in_color_space))
+      cinfo->input_components = rgb_pixelsizecinfo->in_color_space;
+    else if (cinfo->in_color_space == JCS_CMYK)
+      cinfo->input_components = 4;
+    else
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    if ((unsigned long long)biWidth * 3ULL > 0xFFFFFFFFULL)
+      ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+    row_width = (JDIMENSION)biWidth * 3;
+    break;
+  case 32:
+    if (cinfo->in_color_space == JCS_UNKNOWN)
+      cinfo->in_color_space = JCS_EXT_BGRA;
+    if (IsExtRGB(cinfo->in_color_space))
+      cinfo->input_components = rgb_pixelsizecinfo->in_color_space;
+    else if (cinfo->in_color_space == JCS_CMYK)
+      cinfo->input_components = 4;
+    else
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    if ((unsigned long long)biWidth * 4ULL > 0xFFFFFFFFULL)
+      ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+    row_width = (JDIMENSION)biWidth * 4;
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+  }
+  while ((row_width & 3) != 0) row_width++;
+  source->row_width = row_width;
+
+  if (source->use_inversion_array) {
+    /* Allocate space for inversion array, prepare for preload pass */
+    source->whole_image = (*cinfo->mem->request_virt_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE, FALSE,
+       row_width, (JDIMENSION)biHeight, (JDIMENSION)1);
+    source->pub.get_pixel_rows = preload_image;
+    if (cinfo->progress != NULL) {
+      cd_progress_ptr progress = (cd_progress_ptr)cinfo->progress;
+      progress->total_extra_passes++; /* count file input as separate pass */
+    }
+  } else {
+    source->iobuffer = (U_CHAR *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE, row_width);
+    switch (source->bits_per_pixel) {
+    case 8:
+      source->pub.get_pixel_rows = get_8bit_row;
+      break;
+    case 24:
+      source->pub.get_pixel_rows = get_24bit_row;
+      break;
+    case 32:
+      source->pub.get_pixel_rows = get_32bit_row;
+      break;
+    default:
+      ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+    }
+  }
+
+  /* Ensure that biWidth * cinfo->input_components doesn't exceed the maximum
+     value of the JDIMENSION type.  This is only a danger with BMP files, since
+     their width and height fields are 32-bit integers. */
+  if ((unsigned long long)biWidth *
+      (unsigned long long)cinfo->input_components > 0xFFFFFFFFULL)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+  /* Allocate one-row buffer for returned data */
+  source->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE,
+     (JDIMENSION)biWidth * (JDIMENSION)cinfo->input_components, (JDIMENSION)1);
+  source->pub.buffer_height = 1;
+
+  cinfo->data_precision = 8;
+  cinfo->image_width = (JDIMENSION)biWidth;
+  cinfo->image_height = (JDIMENSION)biHeight;
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_bmp(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for BMP format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_bmp(j_compress_ptr cinfo, boolean use_inversion_array)
+{
+  bmp_source_ptr source;
+
+  /* Create module interface object */
+  source = (bmp_source_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(bmp_source_struct));
+  source->cinfo = cinfo;        /* make back link for subroutines */
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_bmp;
+  source->pub.finish_input = finish_input_bmp;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  source->pub.max_pixels = 0;
+#endif
+
+  source->use_inversion_array = use_inversion_array;
+
+  return (cjpeg_source_ptr)source;
+}
+
+#endif /* BMP_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/rdcolmap.c Added

@@ -0,0 +1,253 @@
+/*
+ * rdcolmap.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file implements djpeg's "-map file" switch.  It reads a source image
+ * and constructs a colormap to be supplied to the JPEG decompressor.
+ *
+ * Currently, these file formats are supported for the map file:
+ *   GIF: the contents of the GIF's global colormap are used.
+ *   PPM (either text or raw flavor): the entire file is read and
+ *      each unique pixel value is entered in the map.
+ * Note that reading a large PPM file will be horrendously slow.
+ * Typically, a PPM-format map file should contain just one pixel
+ * of each desired color.  Such a file can be extracted from an
+ * ordinary image PPM file with ppmtomap(1).
+ *
+ * Rescaling a PPM that has a maxval unequal to MAXJSAMPLE is not
+ * currently implemented.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef QUANT_2PASS_SUPPORTED    /* otherwise can't quantize to supplied map */
+
+/* Portions of this code are based on the PBMPLUS library, which is:
+**
+** Copyright (C) 1988 by Jef Poskanzer.
+**
+** Permission to use, copy, modify, and distribute this software and its
+** documentation for any purpose and without fee is hereby granted, provided
+** that the above copyright notice appear in all copies and that both that
+** copyright notice and this permission notice appear in supporting
+** documentation.  This software is provided "as is" without express or
+** implied warranty.
+*/
+
+
+/*
+ * Add a (potentially) new color to the color map.
+ */
+
+LOCAL(void)
+add_map_entry(j_decompress_ptr cinfo, int R, int G, int B)
+{
+  JSAMPROW colormap0 = cinfo->colormap0;
+  JSAMPROW colormap1 = cinfo->colormap1;
+  JSAMPROW colormap2 = cinfo->colormap2;
+  int ncolors = cinfo->actual_number_of_colors;
+  int index;
+
+  /* Check for duplicate color. */
+  for (index = 0; index < ncolors; index++) {
+    if (colormap0index == R && colormap1index == G &&
+        colormap2index == B)
+      return;                   /* color is already in map */
+  }
+
+  /* Check for map overflow. */
+  if (ncolors >= (MAXJSAMPLE + 1))
+    ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, (MAXJSAMPLE + 1));
+
+  /* OK, add color to map. */
+  colormap0ncolors = (JSAMPLE)R;
+  colormap1ncolors = (JSAMPLE)G;
+  colormap2ncolors = (JSAMPLE)B;
+  cinfo->actual_number_of_colors++;
+}
+
+
+/*
+ * Extract color map from a GIF file.
+ */
+
+LOCAL(void)
+read_gif_map(j_decompress_ptr cinfo, FILE *infile)
+{
+  int header13;
+  int i, colormaplen;
+  int R, G, B;
+
+  /* Initial 'G' has already been read by read_color_map */
+  /* Read the rest of the GIF header and logical screen descriptor */
+  for (i = 1; i < 13; i++) {
+    if ((headeri = getc(infile)) == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+  }
+
+  /* Verify GIF Header */
+  if (header1 != 'I' || header2 != 'F')
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* There must be a global color map. */
+  if ((header10 & 0x80) == 0)
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* OK, fetch it. */
+  colormaplen = 2 << (header10 & 0x07);
+
+  for (i = 0; i < colormaplen; i++) {
+    R = getc(infile);
+    G = getc(infile);
+    B = getc(infile);
+    if (R == EOF || G == EOF || B == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    add_map_entry(cinfo,
+                  R << (BITS_IN_JSAMPLE - 8),
+                  G << (BITS_IN_JSAMPLE - 8),
+                  B << (BITS_IN_JSAMPLE - 8));
+  }
+}
+
+
+/* Support routines for reading PPM */
+
+
+LOCAL(int)
+pbm_getc(FILE *infile)
+/* Read next char, skipping over any comments */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(infile);
+  if (ch == '#') {
+    do {
+      ch = getc(infile);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(unsigned int)
+read_pbm_integer(j_decompress_ptr cinfo, FILE *infile)
+/* Read an unsigned decimal integer from the PPM file */
+/* Swallows one trailing character after the integer */
+/* Note that on a 16-bit-int machine, only values up to 64k can be read. */
+/* This should not be a problem in practice. */
+{
+  register int ch;
+  register unsigned int val;
+
+  /* Skip any leading whitespace */
+  do {
+    ch = pbm_getc(infile);
+    if (ch == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+  } while (ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r');
+
+  if (ch < '0' || ch > '9')
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  val = ch - '0';
+  while ((ch = pbm_getc(infile)) >= '0' && ch <= '9') {
+    val *= 10;
+    val += ch - '0';
+  }
+  return val;
+}
+
+
+/*
+ * Extract color map from a PPM file.
+ */
+
+LOCAL(void)
+read_ppm_map(j_decompress_ptr cinfo, FILE *infile)
+{
+  int c;
+  unsigned int w, h, maxval, row, col;
+  int R, G, B;
+
+  /* Initial 'P' has already been read by read_color_map */
+  c = getc(infile);             /* save format discriminator for a sec */
+
+  /* while we fetch the remaining header info */
+  w = read_pbm_integer(cinfo, infile);
+  h = read_pbm_integer(cinfo, infile);
+  maxval = read_pbm_integer(cinfo, infile);
+
+  if (w <= 0 || h <= 0 || maxval <= 0) /* error check */
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* For now, we don't support rescaling from an unusual maxval. */
+  if (maxval != (unsigned int)MAXJSAMPLE)
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  switch (c) {
+  case '3':                     /* it's a text-format PPM file */
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        R = read_pbm_integer(cinfo, infile);
+        G = read_pbm_integer(cinfo, infile);
+        B = read_pbm_integer(cinfo, infile);
+        add_map_entry(cinfo, R, G, B);
+      }
+    }
+    break;
+
+  case '6':                     /* it's a raw-format PPM file */
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        R = getc(infile);
+        G = getc(infile);
+        B = getc(infile);
+        if (R == EOF || G == EOF || B == EOF)
+          ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+        add_map_entry(cinfo, R, G, B);
+      }
+    }
+    break;
+
+  default:
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    break;
+  }
+}
+
+
+/*
+ * Main entry point from djpeg.c.
+ *  Input: opened input file (from file name argument on command line).
+ *  Output: colormap and actual_number_of_colors fields are set in cinfo.
+ */
+
+GLOBAL(void)
+read_color_map(j_decompress_ptr cinfo, FILE *infile)
+{
+  /* Allocate space for a color map of maximum supported size. */
+  cinfo->colormap = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE,
+     (JDIMENSION)(MAXJSAMPLE + 1), (JDIMENSION)3);
+  cinfo->actual_number_of_colors = 0; /* initialize map to empty */
+
+  /* Read first byte to determine file format */
+  switch (getc(infile)) {
+  case 'G':
+    read_gif_map(cinfo, infile);
+    break;
+  case 'P':
+    read_ppm_map(cinfo, infile);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    break;
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/rdgif.c Added

@@ -0,0 +1,695 @@
+/*
+ * rdgif.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2019 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in GIF format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed GIF format).
+ */
+
+/*
+ * This code is loosely based on giftoppm from the PBMPLUS distribution
+ * of Feb. 1991.  That file contains the following copyright notice:
+ * +-------------------------------------------------------------------+
+ * | Copyright 1990, David Koblas.                                     |
+ * |   Permission to use, copy, modify, and distribute this software   |
+ * |   and its documentation for any purpose and without fee is hereby |
+ * |   granted, provided that the above copyright notice appear in all |
+ * |   copies and that both that copyright notice and this permission  |
+ * |   notice appear in supporting documentation.  This software is    |
+ * |   provided "as is" without express or implied warranty.           |
+ * +-------------------------------------------------------------------+
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef GIF_SUPPORTED
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int)(x))
+
+
+#define ReadOK(file, buffer, len) \
+  (JFREAD(file, buffer, len) == ((size_t)(len)))
+
+
+#define MAXCOLORMAPSIZE  256    /* max # of colors in a GIF colormap */
+#define NUMCOLORS        3      /* # of colors */
+#define CM_RED           0      /* color component numbers */
+#define CM_GREEN         1
+#define CM_BLUE          2
+
+#define MAX_LZW_BITS     12     /* maximum LZW code size */
+#define LZW_TABLE_SIZE   (1 << MAX_LZW_BITS) /* # of possible LZW symbols */
+
+/* Macros for extracting header data --- note we assume chars may be signed */
+
+#define LM_to_uint(array, offset) \
+  ((unsigned int)UCH(arrayoffset) + \
+   (((unsigned int)UCH(arrayoffset + 1)) << 8))
+
+#define BitSet(byte, bit)       ((byte) & (bit))
+#define INTERLACE       0x40    /* mask for bit signifying interlaced image */
+#define COLORMAPFLAG    0x80    /* mask for bit signifying colormap presence */
+
+
+/*
+ * LZW decompression tables look like this:
+ *   symbol_headK = prefix symbol of any LZW symbol K (0..LZW_TABLE_SIZE-1)
+ *   symbol_tailK = suffix byte   of any LZW symbol K (0..LZW_TABLE_SIZE-1)
+ * Note that entries 0..end_code of the above tables are not used,
+ * since those symbols represent raw bytes or special codes.
+ *
+ * The stack represents the not-yet-used expansion of the last LZW symbol.
+ * In the worst case, a symbol could expand to as many bytes as there are
+ * LZW symbols, so we allocate LZW_TABLE_SIZE bytes for the stack.
+ * (This is conservative since that number includes the raw-byte symbols.)
+ */
+
+
+/* Private version of data source object */
+
+typedef struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  j_compress_ptr cinfo;         /* back link saves passing separate parm */
+
+  JSAMPARRAY colormap;          /* GIF colormap (converted to my format) */
+
+  /* State for GetCode and LZWReadByte */
+  U_CHAR code_buf256 + 4;     /* current input data block */
+  int last_byte;                /* # of bytes in code_buf */
+  int last_bit;                 /* # of bits in code_buf */
+  int cur_bit;                  /* next bit index to read */
+  boolean first_time;           /* flags first call to GetCode */
+  boolean out_of_blocks;        /* TRUE if hit terminator data block */
+
+  int input_code_size;          /* codesize given in GIF file */
+  int clear_code, end_code;     /* values for Clear and End codes */
+
+  int code_size;                /* current actual code size */
+  int limit_code;               /* 2^code_size */
+  int max_code;                 /* first unused code value */
+
+  /* Private state for LZWReadByte */
+  int oldcode;                  /* previous LZW symbol */
+  int firstcode;                /* first byte of oldcode's expansion */
+
+  /* LZW symbol table and expansion stack */
+  UINT16 *symbol_head;          /* => table of prefix symbols */
+  UINT8  *symbol_tail;          /* => table of suffix bytes */
+  UINT8  *symbol_stack;         /* => stack for symbol expansions */
+  UINT8  *sp;                   /* stack pointer */
+
+  /* State for interlaced image processing */
+  boolean is_interlaced;        /* TRUE if have interlaced image */
+  jvirt_sarray_ptr interlaced_image; /* full image in interlaced order */
+  JDIMENSION cur_row_number;    /* need to know actual row number */
+  JDIMENSION pass2_offset;      /* # of pixel rows in pass 1 */
+  JDIMENSION pass3_offset;      /* # of pixel rows in passes 1&2 */
+  JDIMENSION pass4_offset;      /* # of pixel rows in passes 1,2,3 */
+} gif_source_struct;
+
+typedef gif_source_struct *gif_source_ptr;
+
+
+/* Forward declarations */
+METHODDEF(JDIMENSION) get_pixel_rows(j_compress_ptr cinfo,
+                                     cjpeg_source_ptr sinfo);
+METHODDEF(JDIMENSION) load_interlaced_image(j_compress_ptr cinfo,
+                                            cjpeg_source_ptr sinfo);
+METHODDEF(JDIMENSION) get_interlaced_row(j_compress_ptr cinfo,
+                                         cjpeg_source_ptr sinfo);
+
+
+LOCAL(int)
+ReadByte(gif_source_ptr sinfo)
+/* Read next byte from GIF file */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int c;
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  return c;
+}
+
+
+LOCAL(int)
+GetDataBlock(gif_source_ptr sinfo, U_CHAR *buf)
+/* Read a GIF data block, which has a leading count byte */
+/* A zero-length block marks the end of a data block sequence */
+{
+  int count;
+
+  count = ReadByte(sinfo);
+  if (count > 0) {
+    if (!ReadOK(sinfo->pub.input_file, buf, count))
+      ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  }
+  return count;
+}
+
+
+LOCAL(void)
+SkipDataBlocks(gif_source_ptr sinfo)
+/* Skip a series of data blocks, until a block terminator is found */
+{
+  U_CHAR buf256;
+
+  while (GetDataBlock(sinfo, buf) > 0)
+    /* skip */;
+}
+
+
+LOCAL(void)
+ReInitLZW(gif_source_ptr sinfo)
+/* (Re)initialize LZW state; shared code for startup and Clear processing */
+{
+  sinfo->code_size = sinfo->input_code_size + 1;
+  sinfo->limit_code = sinfo->clear_code << 1;   /* 2^code_size */
+  sinfo->max_code = sinfo->clear_code + 2;      /* first unused code value */
+  sinfo->sp = sinfo->symbol_stack;              /* init stack to empty */
+}
+
+
+LOCAL(void)
+InitLZWCode(gif_source_ptr sinfo)
+/* Initialize for a series of LZWReadByte (and hence GetCode) calls */
+{
+  /* GetCode initialization */
+  sinfo->last_byte = 2;         /* make safe to "recopy last two bytes" */
+  sinfo->code_buf0 = 0;
+  sinfo->code_buf1 = 0;
+  sinfo->last_bit = 0;          /* nothing in the buffer */
+  sinfo->cur_bit = 0;           /* force buffer load on first call */
+  sinfo->first_time = TRUE;
+  sinfo->out_of_blocks = FALSE;
+
+  /* LZWReadByte initialization: */
+  /* compute special code values (note that these do not change later) */
+  sinfo->clear_code = 1 << sinfo->input_code_size;
+  sinfo->end_code = sinfo->clear_code + 1;
+  ReInitLZW(sinfo);
+}
+
+
+LOCAL(int)
+GetCode(gif_source_ptr sinfo)
+/* Fetch the next code_size bits from the GIF data */
+/* We assume code_size is less than 16 */
+{
+  register int accum;
+  int offs, count;
+
+  while (sinfo->cur_bit + sinfo->code_size > sinfo->last_bit) {
+    /* Time to reload the buffer */
+    /* First time, share code with Clear case */
+    if (sinfo->first_time) {
+      sinfo->first_time = FALSE;
+      return sinfo->clear_code;
+    }
+    if (sinfo->out_of_blocks) {
+      WARNMS(sinfo->cinfo, JWRN_GIF_NOMOREDATA);
+      return sinfo->end_code;   /* fake something useful */
+    }
+    /* preserve last two bytes of what we have -- assume code_size <= 16 */
+    sinfo->code_buf0 = sinfo->code_bufsinfo->last_byte-2;
+    sinfo->code_buf1 = sinfo->code_bufsinfo->last_byte-1;
+    /* Load more bytes; set flag if we reach the terminator block */
+    if ((count = GetDataBlock(sinfo, &sinfo->code_buf2)) == 0) {
+      sinfo->out_of_blocks = TRUE;
+      WARNMS(sinfo->cinfo, JWRN_GIF_NOMOREDATA);
+      return sinfo->end_code;   /* fake something useful */
+    }
+    /* Reset counters */
+    sinfo->cur_bit = (sinfo->cur_bit - sinfo->last_bit) + 16;
+    sinfo->last_byte = 2 + count;
+    sinfo->last_bit = sinfo->last_byte * 8;
+  }
+
+  /* Form up next 24 bits in accum */
+  offs = sinfo->cur_bit >> 3;   /* byte containing cur_bit */
+  accum = UCH(sinfo->code_bufoffs + 2);
+  accum <<= 8;
+  accum |= UCH(sinfo->code_bufoffs + 1);
+  accum <<= 8;
+  accum |= UCH(sinfo->code_bufoffs);
+
+  /* Right-align cur_bit in accum, then mask off desired number of bits */
+  accum >>= (sinfo->cur_bit & 7);
+  sinfo->cur_bit += sinfo->code_size;
+  return accum & ((1 << sinfo->code_size) - 1);
+}
+
+
+LOCAL(int)
+LZWReadByte(gif_source_ptr sinfo)
+/* Read an LZW-compressed byte */
+{
+  register int code;            /* current working code */
+  int incode;                   /* saves actual input code */
+
+  /* If any codes are stacked from a previously read symbol, return them */
+  if (sinfo->sp > sinfo->symbol_stack)
+    return (int)(*(--sinfo->sp));
+
+  /* Time to read a new symbol */
+  code = GetCode(sinfo);
+
+  if (code == sinfo->clear_code) {
+    /* Reinit state, swallow any extra Clear codes, and */
+    /* return next code, which is expected to be a raw byte. */
+    ReInitLZW(sinfo);
+    do {
+      code = GetCode(sinfo);
+    } while (code == sinfo->clear_code);
+    if (code > sinfo->clear_code) { /* make sure it is a raw byte */
+      WARNMS(sinfo->cinfo, JWRN_GIF_BADDATA);
+      code = 0;                 /* use something valid */
+    }
+    /* make firstcode, oldcode valid! */
+    sinfo->firstcode = sinfo->oldcode = code;
+    return code;
+  }
+
+  if (code == sinfo->end_code) {
+    /* Skip the rest of the image, unless GetCode already read terminator */
+    if (!sinfo->out_of_blocks) {
+      SkipDataBlocks(sinfo);
+      sinfo->out_of_blocks = TRUE;
+    }
+    /* Complain that there's not enough data */
+    WARNMS(sinfo->cinfo, JWRN_GIF_ENDCODE);
+    /* Pad data with 0's */
+    return 0;                   /* fake something usable */
+  }
+
+  /* Got normal raw byte or LZW symbol */
+  incode = code;                /* save for a moment */
+
+  if (code >= sinfo->max_code) { /* special case for not-yet-defined symbol */
+    /* code == max_code is OK; anything bigger is bad data */
+    if (code > sinfo->max_code) {
+      WARNMS(sinfo->cinfo, JWRN_GIF_BADDATA);
+      incode = 0;               /* prevent creation of loops in symbol table */
+    }
+    /* this symbol will be defined as oldcode/firstcode */
+    *(sinfo->sp++) = (UINT8)sinfo->firstcode;
+    code = sinfo->oldcode;
+  }
+
+  /* If it's a symbol, expand it into the stack */
+  while (code >= sinfo->clear_code) {
+    *(sinfo->sp++) = sinfo->symbol_tailcode; /* tail is a byte value */
+    code = sinfo->symbol_headcode; /* head is another LZW symbol */
+  }
+  /* At this point code just represents a raw byte */
+  sinfo->firstcode = code;      /* save for possible future use */
+
+  /* If there's room in table... */
+  if ((code = sinfo->max_code) < LZW_TABLE_SIZE) {
+    /* Define a new symbol = prev sym + head of this sym's expansion */
+    sinfo->symbol_headcode = (UINT16)sinfo->oldcode;
+    sinfo->symbol_tailcode = (UINT8)sinfo->firstcode;
+    sinfo->max_code++;
+    /* Is it time to increase code_size? */
+    if (sinfo->max_code >= sinfo->limit_code &&
+        sinfo->code_size < MAX_LZW_BITS) {
+      sinfo->code_size++;
+      sinfo->limit_code <<= 1;  /* keep equal to 2^code_size */
+    }
+  }
+
+  sinfo->oldcode = incode;      /* save last input symbol for future use */
+  return sinfo->firstcode;      /* return first byte of symbol's expansion */
+}
+
+
+LOCAL(void)
+ReadColorMap(gif_source_ptr sinfo, int cmaplen, JSAMPARRAY cmap)
+/* Read a GIF colormap */
+{
+  int i;
+
+  for (i = 0; i < cmaplen; i++) {
+#if BITS_IN_JSAMPLE == 8
+#define UPSCALE(x)  (x)
+#else
+#define UPSCALE(x)  ((x) << (BITS_IN_JSAMPLE - 8))
+#endif
+    cmapCM_REDi   = (JSAMPLE)UPSCALE(ReadByte(sinfo));
+    cmapCM_GREENi = (JSAMPLE)UPSCALE(ReadByte(sinfo));
+    cmapCM_BLUEi  = (JSAMPLE)UPSCALE(ReadByte(sinfo));
+  }
+}
+
+
+LOCAL(void)
+DoExtension(gif_source_ptr sinfo)
+/* Process an extension block */
+/* Currently we ignore 'em all */
+{
+  int extlabel;
+
+  /* Read extension label byte */
+  extlabel = ReadByte(sinfo);
+  TRACEMS1(sinfo->cinfo, 1, JTRC_GIF_EXTENSION, extlabel);
+  /* Skip the data block(s) associated with the extension */
+  SkipDataBlocks(sinfo);
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_gif(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  gif_source_ptr source = (gif_source_ptr)sinfo;
+  U_CHAR hdrbuf10;            /* workspace for reading control blocks */
+  unsigned int width, height;   /* image dimensions */
+  int colormaplen, aspectRatio;
+  int c;
+
+  /* Read and verify GIF Header */
+  if (!ReadOK(source->pub.input_file, hdrbuf, 6))
+    ERREXIT(cinfo, JERR_GIF_NOT);
+  if (hdrbuf0 != 'G' || hdrbuf1 != 'I' || hdrbuf2 != 'F')
+    ERREXIT(cinfo, JERR_GIF_NOT);
+  /* Check for expected version numbers.
+   * If unknown version, give warning and try to process anyway;
+   * this is per recommendation in GIF89a standard.
+   */
+  if ((hdrbuf3 != '8' || hdrbuf4 != '7' || hdrbuf5 != 'a') &&
+      (hdrbuf3 != '8' || hdrbuf4 != '9' || hdrbuf5 != 'a'))
+    TRACEMS3(cinfo, 1, JTRC_GIF_BADVERSION, hdrbuf3, hdrbuf4, hdrbuf5);
+
+  /* Read and decipher Logical Screen Descriptor */
+  if (!ReadOK(source->pub.input_file, hdrbuf, 7))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  width = LM_to_uint(hdrbuf, 0);
+  height = LM_to_uint(hdrbuf, 2);
+  if (width == 0 || height == 0)
+    ERREXIT(cinfo, JERR_GIF_EMPTY);
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  if (sinfo->max_pixels &&
+      (unsigned long long)width * height > sinfo->max_pixels)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+#endif
+  /* we ignore the color resolution, sort flag, and background color index */
+  aspectRatio = UCH(hdrbuf6);
+  if (aspectRatio != 0 && aspectRatio != 49)
+    TRACEMS(cinfo, 1, JTRC_GIF_NONSQUARE);
+
+  /* Allocate space to store the colormap */
+  source->colormap = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE, (JDIMENSION)MAXCOLORMAPSIZE,
+     (JDIMENSION)NUMCOLORS);
+  colormaplen = 0;              /* indicate initialization */
+
+  /* Read global colormap if header indicates it is present */
+  if (BitSet(hdrbuf4, COLORMAPFLAG)) {
+    colormaplen = 2 << (hdrbuf4 & 0x07);
+    ReadColorMap(source, colormaplen, source->colormap);
+  }
+
+  /* Scan until we reach start of desired image.
+   * We don't currently support skipping images, but could add it easily.
+   */
+  for (;;) {
+    c = ReadByte(source);
+
+    if (c == ';')               /* GIF terminator?? */
+      ERREXIT(cinfo, JERR_GIF_IMAGENOTFOUND);
+
+    if (c == '!') {             /* Extension */
+      DoExtension(source);
+      continue;
+    }
+
+    if (c != ',') {             /* Not an image separator? */
+      WARNMS1(cinfo, JWRN_GIF_CHAR, c);
+      continue;
+    }
+
+    /* Read and decipher Local Image Descriptor */
+    if (!ReadOK(source->pub.input_file, hdrbuf, 9))
+      ERREXIT(cinfo, JERR_INPUT_EOF);
+    /* we ignore top/left position info, also sort flag */
+    width = LM_to_uint(hdrbuf, 4);
+    height = LM_to_uint(hdrbuf, 6);
+    if (width == 0 || height == 0)
+      ERREXIT(cinfo, JERR_GIF_EMPTY);
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+    if (sinfo->max_pixels &&
+        (unsigned long long)width * height > sinfo->max_pixels)
+      ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+#endif
+    source->is_interlaced = (BitSet(hdrbuf8, INTERLACE) != 0);
+
+    /* Read local colormap if header indicates it is present */
+    /* Note: if we wanted to support skipping images, */
+    /* we'd need to skip rather than read colormap for ignored images */
+    if (BitSet(hdrbuf8, COLORMAPFLAG)) {
+      colormaplen = 2 << (hdrbuf8 & 0x07);
+      ReadColorMap(source, colormaplen, source->colormap);
+    }
+
+    source->input_code_size = ReadByte(source); /* get min-code-size byte */
+    if (source->input_code_size < 2 || source->input_code_size > 8)
+      ERREXIT1(cinfo, JERR_GIF_CODESIZE, source->input_code_size);
+
+    /* Reached desired image, so break out of loop */
+    /* If we wanted to skip this image, */
+    /* we'd call SkipDataBlocks and then continue the loop */
+    break;
+  }
+
+  /* Prepare to read selected image: first initialize LZW decompressor */
+  source->symbol_head = (UINT16 *)
+    (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                LZW_TABLE_SIZE * sizeof(UINT16));
+  source->symbol_tail = (UINT8 *)
+    (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                LZW_TABLE_SIZE * sizeof(UINT8));
+  source->symbol_stack = (UINT8 *)
+    (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                LZW_TABLE_SIZE * sizeof(UINT8));
+  InitLZWCode(source);
+
+  /*
+   * If image is interlaced, we read it into a full-size sample array,
+   * decompressing as we go; then get_interlaced_row selects rows from the
+   * sample array in the proper order.
+   */
+  if (source->is_interlaced) {
+    /* We request the virtual array now, but can't access it until virtual
+     * arrays have been allocated.  Hence, the actual work of reading the
+     * image is postponed until the first call to get_pixel_rows.
+     */
+    source->interlaced_image = (*cinfo->mem->request_virt_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE, FALSE,
+       (JDIMENSION)width, (JDIMENSION)height, (JDIMENSION)1);
+    if (cinfo->progress != NULL) {
+      cd_progress_ptr progress = (cd_progress_ptr)cinfo->progress;
+      progress->total_extra_passes++; /* count file input as separate pass */
+    }
+    source->pub.get_pixel_rows = load_interlaced_image;
+  } else {
+    source->pub.get_pixel_rows = get_pixel_rows;
+  }
+
+  /* Create compressor input buffer. */
+  source->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE, (JDIMENSION)width * NUMCOLORS,
+     (JDIMENSION)1);
+  source->pub.buffer_height = 1;
+
+  /* Pad colormap for safety. */
+  for (c = colormaplen; c < source->clear_code; c++) {
+    source->colormapCM_REDc   =
+    source->colormapCM_GREENc =
+    source->colormapCM_BLUEc  = CENTERJSAMPLE;
+  }
+
+  /* Return info about the image. */
+  cinfo->in_color_space = JCS_RGB;
+  cinfo->input_components = NUMCOLORS;
+  cinfo->data_precision = BITS_IN_JSAMPLE; /* we always rescale data to this */
+  cinfo->image_width = width;
+  cinfo->image_height = height;
+
+  TRACEMS3(cinfo, 1, JTRC_GIF, width, height, colormaplen);
+}
+
+
+/*
+ * Read one row of pixels.
+ * This version is used for noninterlaced GIF images:
+ * we read directly from the GIF file.
+ */
+
+METHODDEF(JDIMENSION)
+get_pixel_rows(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  gif_source_ptr source = (gif_source_ptr)sinfo;
+  register int c;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  register JSAMPARRAY colormap = source->colormap;
+
+  ptr = source->pub.buffer0;
+  for (col = cinfo->image_width; col > 0; col--) {
+    c = LZWReadByte(source);
+    *ptr++ = colormapCM_REDc;
+    *ptr++ = colormapCM_GREENc;
+    *ptr++ = colormapCM_BLUEc;
+  }
+  return 1;
+}
+
+
+/*
+ * Read one row of pixels.
+ * This version is used for the first call on get_pixel_rows when
+ * reading an interlaced GIF file: we read the whole image into memory.
+ */
+
+METHODDEF(JDIMENSION)
+load_interlaced_image(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  gif_source_ptr source = (gif_source_ptr)sinfo;
+  register JSAMPROW sptr;
+  register JDIMENSION col;
+  JDIMENSION row;
+  cd_progress_ptr progress = (cd_progress_ptr)cinfo->progress;
+
+  /* Read the interlaced image into the virtual array we've created. */
+  for (row = 0; row < cinfo->image_height; row++) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long)row;
+      progress->pub.pass_limit = (long)cinfo->image_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr)cinfo);
+    }
+    sptr = *(*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr)cinfo, source->interlaced_image, row, (JDIMENSION)1,
+       TRUE);
+    for (col = cinfo->image_width; col > 0; col--) {
+      *sptr++ = (JSAMPLE)LZWReadByte(source);
+    }
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Replace method pointer so subsequent calls don't come here. */
+  source->pub.get_pixel_rows = get_interlaced_row;
+  /* Initialize for get_interlaced_row, and perform first call on it. */
+  source->cur_row_number = 0;
+  source->pass2_offset = (cinfo->image_height + 7) / 8;
+  source->pass3_offset = source->pass2_offset + (cinfo->image_height + 3) / 8;
+  source->pass4_offset = source->pass3_offset + (cinfo->image_height + 1) / 4;
+
+  return get_interlaced_row(cinfo, sinfo);
+}
+
+
+/*
+ * Read one row of pixels.
+ * This version is used for interlaced GIF images:
+ * we read from the virtual array.
+ */
+
+METHODDEF(JDIMENSION)
+get_interlaced_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  gif_source_ptr source = (gif_source_ptr)sinfo;
+  register int c;
+  register JSAMPROW sptr, ptr;
+  register JDIMENSION col;
+  register JSAMPARRAY colormap = source->colormap;
+  JDIMENSION irow;
+
+  /* Figure out which row of interlaced image is needed, and access it. */
+  switch ((int)(source->cur_row_number & 7)) {
+  case 0:                       /* first-pass row */
+    irow = source->cur_row_number >> 3;
+    break;
+  case 4:                       /* second-pass row */
+    irow = (source->cur_row_number >> 3) + source->pass2_offset;
+    break;
+  case 2:                       /* third-pass row */
+  case 6:
+    irow = (source->cur_row_number >> 2) + source->pass3_offset;
+    break;
+  default:                      /* fourth-pass row */
+    irow = (source->cur_row_number >> 1) + source->pass4_offset;
+  }
+  sptr = *(*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr)cinfo, source->interlaced_image, irow, (JDIMENSION)1,
+     FALSE);
+  /* Scan the row, expand colormap, and output */
+  ptr = source->pub.buffer0;
+  for (col = cinfo->image_width; col > 0; col--) {
+    c = *sptr++;
+    *ptr++ = colormapCM_REDc;
+    *ptr++ = colormapCM_GREENc;
+    *ptr++ = colormapCM_BLUEc;
+  }
+  source->cur_row_number++;     /* for next time */
+  return 1;
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_gif(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for GIF format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_gif(j_compress_ptr cinfo)
+{
+  gif_source_ptr source;
+
+  /* Create module interface object */
+  source = (gif_source_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(gif_source_struct));
+  source->cinfo = cinfo;        /* make back link for subroutines */
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_gif;
+  source->pub.finish_input = finish_input_gif;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  source->pub.max_pixels = 0;
+#endif
+
+  return (cjpeg_source_ptr)source;
+}
+
+#endif /* GIF_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/rdjpgcom.1 Added

@@ -0,0 +1,63 @@
+.TH RDJPGCOM 1 "02 April 2009"
+.SH NAME
+rdjpgcom \- display text comments from a JPEG file
+.SH SYNOPSIS
+.B rdjpgcom
+
+.B \-raw
+
+
+.B \-verbose
+
+
+.I filename
+
+.LP
+.SH DESCRIPTION
+.LP
+.B rdjpgcom
+reads the named JPEG/JFIF file, or the standard input if no file is named,
+and prints any text comments found in the file on the standard output.
+.PP
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+.SH OPTIONS
+.TP
+.B \-raw
+Normally
+.B rdjpgcom
+escapes non-printable characters in comments, for security reasons.
+This option avoids that.
+.PP
+.B \-verbose
+Causes
+.B rdjpgcom
+to also display the JPEG image dimensions.
+.PP
+Switch names may be abbreviated, and are not case sensitive.
+.SH HINTS
+.B rdjpgcom
+does not depend on the IJG JPEG library.  Its source code is intended as an
+illustration of the minimum amount of code required to parse a JPEG file
+header correctly.
+.PP
+In
+.B \-verbose
+mode,
+.B rdjpgcom
+will also attempt to print the contents of any "APP12" markers as text.
+Some digital cameras produce APP12 markers containing useful textual
+information.  If you like, you can modify the source code to print
+other APPn marker types as well.
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR wrjpgcom (1)
+.SH AUTHOR
+Independent JPEG Group

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/rdjpgcom.c Added

@@ -0,0 +1,510 @@
+/*
+ * rdjpgcom.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * Modified 2009 by Bill Allombert, Guido Vollbeding.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a very simple stand-alone application that displays
+ * the text in COM (comment) markers in a JFIF file.
+ * This may be useful as an example of the minimum logic needed to parse
+ * JPEG markers.
+ */
+
+#define JPEG_CJPEG_DJPEG        /* to get the command-line config symbols */
+#include "jinclude.h"           /* get auto-config symbols, <stdio.h> */
+
+#ifdef HAVE_LOCALE_H
+#include <locale.h>             /* Bill Allombert: use locale for isprint */
+#endif
+#include <ctype.h>              /* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>              /* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>                 /* to declare setmode() */
+#endif
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+#ifdef DONT_USE_B_MODE          /* define mode parameters for fopen() */
+#define READ_BINARY     "r"
+#else
+#define READ_BINARY     "rb"
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#define EXIT_SUCCESS  0
+#endif
+
+
+/*
+ * These macros are used to read the input file.
+ * To reuse this code in another application, you might need to change these.
+ */
+
+static FILE *infile;            /* input JPEG file */
+
+/* Return next input byte, or EOF if no more */
+#define NEXTBYTE()  getc(infile)
+
+
+/* Error exit handler */
+#define ERREXIT(msg)  (fprintf(stderr, "%s\n", msg), exit(EXIT_FAILURE))
+
+
+/* Read one byte, testing for EOF */
+static int
+read_1_byte(void)
+{
+  int c;
+
+  c = NEXTBYTE();
+  if (c == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return c;
+}
+
+/* Read 2 bytes, convert to unsigned int */
+/* All 2-byte quantities in JPEG markers are MSB first */
+static unsigned int
+read_2_bytes(void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  if (c1 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  c2 = NEXTBYTE();
+  if (c2 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return (((unsigned int)c1) << 8) + ((unsigned int)c2);
+}
+
+
+/*
+ * JPEG markers consist of one or more 0xFF bytes, followed by a marker
+ * code byte (which is not an FF).  Here are the marker codes of interest
+ * in this program.  (See jdmarker.c for a more complete list.)
+ */
+
+#define M_SOF0   0xC0           /* Start Of Frame N */
+#define M_SOF1   0xC1           /* N indicates which compression process */
+#define M_SOF2   0xC2           /* Only SOF0-SOF2 are now in common use */
+#define M_SOF3   0xC3
+#define M_SOF5   0xC5           /* NB: codes C4 and CC are NOT SOF markers */
+#define M_SOF6   0xC6
+#define M_SOF7   0xC7
+#define M_SOF9   0xC9
+#define M_SOF10  0xCA
+#define M_SOF11  0xCB
+#define M_SOF13  0xCD
+#define M_SOF14  0xCE
+#define M_SOF15  0xCF
+#define M_SOI    0xD8           /* Start Of Image (beginning of datastream) */
+#define M_EOI    0xD9           /* End Of Image (end of datastream) */
+#define M_SOS    0xDA           /* Start Of Scan (begins compressed data) */
+#define M_APP12  0xEC           /* (we don't bother to list all 16 APPn's) */
+#define M_COM    0xFE           /* COMment */
+
+
+/*
+ * Find the next JPEG marker and return its marker code.
+ * We expect at least one FF byte, possibly more if the compressor used FFs
+ * to pad the file.
+ * There could also be non-FF garbage between markers.  The treatment of such
+ * garbage is unspecified; we choose to skip over it but emit a warning msg.
+ * NB: this routine must not be used after seeing SOS marker, since it will
+ * not deal correctly with FF/00 sequences in the compressed image data...
+ */
+
+static int
+next_marker(void)
+{
+  int c;
+  int discarded_bytes = 0;
+
+  /* Find 0xFF byte; count and skip any non-FFs. */
+  c = read_1_byte();
+  while (c != 0xFF) {
+    discarded_bytes++;
+    c = read_1_byte();
+  }
+  /* Get marker code byte, swallowing any duplicate FF bytes.  Extra FFs
+   * are legal as pad bytes, so don't count them in discarded_bytes.
+   */
+  do {
+    c = read_1_byte();
+  } while (c == 0xFF);
+
+  if (discarded_bytes != 0) {
+    fprintf(stderr, "Warning: garbage data found in JPEG file\n");
+  }
+
+  return c;
+}
+
+
+/*
+ * Read the initial marker, which should be SOI.
+ * For a JFIF file, the first two bytes of the file should be literally
+ * 0xFF M_SOI.  To be more general, we could use next_marker, but if the
+ * input file weren't actually JPEG at all, next_marker might read the whole
+ * file and then return a misleading error message...
+ */
+
+static int
+first_marker(void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  c2 = NEXTBYTE();
+  if (c1 != 0xFF || c2 != M_SOI)
+    ERREXIT("Not a JPEG file");
+  return c2;
+}
+
+
+/*
+ * Most types of marker are followed by a variable-length parameter segment.
+ * This routine skips over the parameters for any marker we don't otherwise
+ * want to process.
+ * Note that we MUST skip the parameter segment explicitly in order not to
+ * be fooled by 0xFF bytes that might appear within the parameter segment;
+ * such bytes do NOT introduce new markers.
+ */
+
+static void
+skip_variable(void)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    (void)read_1_byte();
+    length--;
+  }
+}
+
+
+/*
+ * Process a COM marker.
+ * We want to print out the marker contents as legible text;
+ * we must guard against non-text junk and varying newline representations.
+ */
+
+static void
+process_COM(int raw)
+{
+  unsigned int length;
+  int ch;
+  int lastch = 0;
+
+  /* Bill Allombert: set locale properly for isprint */
+#ifdef HAVE_LOCALE_H
+  setlocale(LC_CTYPE, "");
+#endif
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+
+  while (length > 0) {
+    ch = read_1_byte();
+    if (raw) {
+      putc(ch, stdout);
+    /* Emit the character in a readable form.
+     * Nonprintables are converted to \nnn form,
+     * while \ is converted to \\.
+     * Newlines in CR, CR/LF, or LF form will be printed as one newline.
+     */
+    } else if (ch == '\r') {
+      printf("\n");
+    } else if (ch == '\n') {
+      if (lastch != '\r')
+        printf("\n");
+    } else if (ch == '\\') {
+      printf("\\\\");
+    } else if (isprint(ch)) {
+      putc(ch, stdout);
+    } else {
+      printf("\\%03o", ch);
+    }
+    lastch = ch;
+    length--;
+  }
+  printf("\n");
+
+  /* Bill Allombert: revert to C locale */
+#ifdef HAVE_LOCALE_H
+  setlocale(LC_CTYPE, "C");
+#endif
+}
+
+
+/*
+ * Process a SOFn marker.
+ * This code is only needed if you want to know the image dimensions...
+ */
+
+static void
+process_SOFn(int marker)
+{
+  unsigned int length;
+  unsigned int image_height, image_width;
+  int data_precision, num_components;
+  const char *process;
+  int ci;
+
+  length = read_2_bytes();      /* usual parameter length count */
+
+  data_precision = read_1_byte();
+  image_height = read_2_bytes();
+  image_width = read_2_bytes();
+  num_components = read_1_byte();
+
+  switch (marker) {
+  case M_SOF0:  process = "Baseline";  break;
+  case M_SOF1:  process = "Extended sequential";  break;
+  case M_SOF2:  process = "Progressive";  break;
+  case M_SOF3:  process = "Lossless";  break;
+  case M_SOF5:  process = "Differential sequential";  break;
+  case M_SOF6:  process = "Differential progressive";  break;
+  case M_SOF7:  process = "Differential lossless";  break;
+  case M_SOF9:  process = "Extended sequential, arithmetic coding";  break;
+  case M_SOF10: process = "Progressive, arithmetic coding";  break;
+  case M_SOF11: process = "Lossless, arithmetic coding";  break;
+  case M_SOF13: process = "Differential sequential, arithmetic coding";  break;
+  case M_SOF14:
+    process = "Differential progressive, arithmetic coding";  break;
+  case M_SOF15: process = "Differential lossless, arithmetic coding";  break;
+  default:      process = "Unknown";  break;
+  }
+
+  printf("JPEG image is %uw * %uh, %d color components, %d bits per sample\n",
+         image_width, image_height, num_components, data_precision);
+  printf("JPEG process: %s\n", process);
+
+  if (length != (unsigned int)(8 + num_components * 3))
+    ERREXIT("Bogus SOF marker length");
+
+  for (ci = 0; ci < num_components; ci++) {
+    (void)read_1_byte();        /* Component ID code */
+    (void)read_1_byte();        /* H, V sampling factors */
+    (void)read_1_byte();        /* Quantization table number */
+  }
+}
+
+
+/*
+ * Parse the marker stream until SOS or EOI is seen;
+ * display any COM markers.
+ * While the companion program wrjpgcom will always insert COM markers before
+ * SOFn, other implementations might not, so we scan to SOS before stopping.
+ * If we were only interested in the image dimensions, we would stop at SOFn.
+ * (Conversely, if we only cared about COM markers, there would be no need
+ * for special code to handle SOFn; we could treat it like other markers.)
+ */
+
+static int
+scan_JPEG_header(int verbose, int raw)
+{
+  int marker;
+
+  /* Expect SOI at start of file */
+  if (first_marker() != M_SOI)
+    ERREXIT("Expected SOI marker first");
+
+  /* Scan miscellaneous markers until we reach SOS. */
+  for (;;) {
+    marker = next_marker();
+    switch (marker) {
+      /* Note that marker codes 0xC4, 0xC8, 0xCC are not, and must not be,
+       * treated as SOFn.  C4 in particular is actually DHT.
+       */
+    case M_SOF0:                /* Baseline */
+    case M_SOF1:                /* Extended sequential, Huffman */
+    case M_SOF2:                /* Progressive, Huffman */
+    case M_SOF3:                /* Lossless, Huffman */
+    case M_SOF5:                /* Differential sequential, Huffman */
+    case M_SOF6:                /* Differential progressive, Huffman */
+    case M_SOF7:                /* Differential lossless, Huffman */
+    case M_SOF9:                /* Extended sequential, arithmetic */
+    case M_SOF10:               /* Progressive, arithmetic */
+    case M_SOF11:               /* Lossless, arithmetic */
+    case M_SOF13:               /* Differential sequential, arithmetic */
+    case M_SOF14:               /* Differential progressive, arithmetic */
+    case M_SOF15:               /* Differential lossless, arithmetic */
+      if (verbose)
+        process_SOFn(marker);
+      else
+        skip_variable();
+      break;
+
+    case M_SOS:                 /* stop before hitting compressed data */
+      return marker;
+
+    case M_EOI:                 /* in case it's a tables-only JPEG stream */
+      return marker;
+
+    case M_COM:
+      process_COM(raw);
+      break;
+
+    case M_APP12:
+      /* Some digital camera makers put useful textual information into
+       * APP12 markers, so we print those out too when in -verbose mode.
+       */
+      if (verbose) {
+        printf("APP12 contains:\n");
+        process_COM(raw);
+      } else
+        skip_variable();
+      break;
+
+    default:                    /* Anything else just gets skipped */
+      skip_variable();          /* we assume it has a parameter count... */
+      break;
+    }
+  } /* end loop */
+}
+
+
+/* Command line parsing code */
+
+static const char *progname;    /* program name for error messages */
+
+
+static void
+usage(void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "rdjpgcom displays any textual comments in a JPEG file.\n");
+
+  fprintf(stderr, "Usage: %s switches inputfile\n", progname);
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -raw        Display non-printable characters in comments (unsafe)\n");
+  fprintf(stderr, "  -verbose    Also display dimensions of JPEG image\n");
+
+  exit(EXIT_FAILURE);
+}
+
+
+static int
+keymatch(char *arg, const char *keyword, int minchars)
+/* Case-insensitive matching of (possibly abbreviated) keyword switches. */
+/* keyword is the constant keyword (must be lower case already), */
+/* minchars is length of minimum legal abbreviation. */
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return 0;                 /* arg longer than keyword, no good */
+    if (isupper(ca))            /* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return 0;                 /* no good */
+    nmatched++;                 /* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return 0;
+  return 1;                     /* A-OK */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main(int argc, char **argv)
+{
+  int argn;
+  char *arg;
+  int verbose = 0, raw = 0;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv0;
+  if (progname == NULL || progname0 == 0)
+    progname = "rdjpgcom";      /* in case C library doesn't provide it */
+
+  /* Parse switches, if any */
+  for (argn = 1; argn < argc; argn++) {
+    arg = argvargn;
+    if (arg0 != '-')
+      break;                    /* not switch, must be file name */
+    arg++;                      /* advance over '-' */
+    if (keymatch(arg, "verbose", 1)) {
+      verbose++;
+    } else if (keymatch(arg, "raw", 1)) {
+      raw = 1;
+    } else
+      usage();
+  }
+
+  /* Open the input file. */
+  /* Unix style: expect zero or one file name */
+  if (argn < argc - 1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+  if (argn < argc) {
+    if ((infile = fopen(argvargn, READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argvargn);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+#ifdef USE_SETMODE              /* need to hack file mode? */
+    setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+    if ((infile = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open stdin\n", progname);
+      exit(EXIT_FAILURE);
+    }
+#else
+    infile = stdin;
+#endif
+  }
+
+  /* Scan the JPEG headers. */
+  (void)scan_JPEG_header(verbose, raw);
+
+  /* All done. */
+  exit(EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/rdppm.c Added

@@ -0,0 +1,779 @@
+/*
+ * rdppm.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2009 by Bill Allombert, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015-2017, 2020-2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in PPM/PGM format.
+ * The extended 2-byte-per-sample raw PPM/PGM formats are supported.
+ * The PBMPLUS library is NOT required to compile this software
+ * (but it is highly useful as a set of PPM image manipulation programs).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed PPM format).
+ */
+
+#include "cmyk.h"
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef PPM_SUPPORTED
+
+
+/* Portions of this code are based on the PBMPLUS library, which is:
+**
+** Copyright (C) 1988 by Jef Poskanzer.
+**
+** Permission to use, copy, modify, and distribute this software and its
+** documentation for any purpose and without fee is hereby granted, provided
+** that the above copyright notice appear in all copies and that both that
+** copyright notice and this permission notice appear in supporting
+** documentation.  This software is provided "as is" without express or
+** implied warranty.
+*/
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int)(x))
+
+
+#define ReadOK(file, buffer, len) \
+  (JFREAD(file, buffer, len) == ((size_t)(len)))
+
+static int alpha_indexJPEG_NUMCS = {
+  -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 3, 0, 0, -1
+};
+
+
+/* Private version of data source object */
+
+typedef struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  /* Usually these two pointers point to the same place: */
+  U_CHAR *iobuffer;             /* fread's I/O buffer */
+  JSAMPROW pixrow;              /* compressor input buffer */
+  size_t buffer_width;          /* width of I/O buffer */
+  JSAMPLE *rescale;             /* => maxval-remapping array, or NULL */
+  unsigned int maxval;
+} ppm_source_struct;
+
+typedef ppm_source_struct *ppm_source_ptr;
+
+
+LOCAL(int)
+pbm_getc(FILE *infile)
+/* Read next char, skipping over any comments */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(infile);
+  if (ch == '#') {
+    do {
+      ch = getc(infile);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(unsigned int)
+read_pbm_integer(j_compress_ptr cinfo, FILE *infile, unsigned int maxval)
+/* Read an unsigned decimal integer from the PPM file */
+/* Swallows one trailing character after the integer */
+/* Note that on a 16-bit-int machine, only values up to 64k can be read. */
+/* This should not be a problem in practice. */
+{
+  register int ch;
+  register unsigned int val;
+
+  /* Skip any leading whitespace */
+  do {
+    ch = pbm_getc(infile);
+    if (ch == EOF)
+      ERREXIT(cinfo, JERR_INPUT_EOF);
+  } while (ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r');
+
+  if (ch < '0' || ch > '9')
+    ERREXIT(cinfo, JERR_PPM_NONNUMERIC);
+
+  val = ch - '0';
+  while ((ch = pbm_getc(infile)) >= '0' && ch <= '9') {
+    val *= 10;
+    val += ch - '0';
+    if (val > maxval)
+      ERREXIT(cinfo, JERR_PPM_OUTOFRANGE);
+  }
+
+  return val;
+}
+
+
+/*
+ * Read one row of pixels.
+ *
+ * We provide several different versions depending on input file format.
+ * In all cases, input is scaled to the size of JSAMPLE.
+ *
+ * A really fast path is provided for reading byte/sample raw files with
+ * maxval = MAXJSAMPLE, which is the normal case for 8-bit data.
+ */
+
+
+METHODDEF(JDIMENSION)
+get_text_gray_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  FILE *infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  ptr = source->pub.buffer0;
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescaleread_pbm_integer(cinfo, infile, maxval);
+  }
+  return 1;
+}
+
+
+#define GRAY_RGB_READ_LOOP(read_op, alpha_set_op) { \
+  for (col = cinfo->image_width; col > 0; col--) { \
+    ptrrindex = ptrgindex = ptrbindex = read_op; \
+    alpha_set_op \
+    ptr += ps; \
+  } \
+}
+
+METHODDEF(JDIMENSION)
+get_text_gray_rgb_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PGM files with any maxval and
+   converting to extended RGB */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  FILE *infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+  register int rindex = rgb_redcinfo->in_color_space;
+  register int gindex = rgb_greencinfo->in_color_space;
+  register int bindex = rgb_bluecinfo->in_color_space;
+  register int aindex = alpha_indexcinfo->in_color_space;
+  register int ps = rgb_pixelsizecinfo->in_color_space;
+
+  ptr = source->pub.buffer0;
+  if (maxval == MAXJSAMPLE) {
+    if (aindex >= 0)
+      GRAY_RGB_READ_LOOP(read_pbm_integer(cinfo, infile, maxval),
+                         ptraindex = 0xFF;)
+    else
+      GRAY_RGB_READ_LOOP(read_pbm_integer(cinfo, infile, maxval),)
+  } else {
+    if (aindex >= 0)
+      GRAY_RGB_READ_LOOP(rescaleread_pbm_integer(cinfo, infile, maxval),
+                         ptraindex = 0xFF;)
+    else
+      GRAY_RGB_READ_LOOP(rescaleread_pbm_integer(cinfo, infile, maxval),)
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_text_gray_cmyk_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PGM files with any maxval and
+   converting to CMYK */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  FILE *infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  ptr = source->pub.buffer0;
+  if (maxval == MAXJSAMPLE) {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE gray = read_pbm_integer(cinfo, infile, maxval);
+      rgb_to_cmyk(gray, gray, gray, ptr, ptr + 1, ptr + 2, ptr + 3);
+      ptr += 4;
+    }
+  } else {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE gray = rescaleread_pbm_integer(cinfo, infile, maxval);
+      rgb_to_cmyk(gray, gray, gray, ptr, ptr + 1, ptr + 2, ptr + 3);
+      ptr += 4;
+    }
+  }
+  return 1;
+}
+
+
+#define RGB_READ_LOOP(read_op, alpha_set_op) { \
+  for (col = cinfo->image_width; col > 0; col--) { \
+    ptrrindex = read_op; \
+    ptrgindex = read_op; \
+    ptrbindex = read_op; \
+    alpha_set_op \
+    ptr += ps; \
+  } \
+}
+
+METHODDEF(JDIMENSION)
+get_text_rgb_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  FILE *infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+  register int rindex = rgb_redcinfo->in_color_space;
+  register int gindex = rgb_greencinfo->in_color_space;
+  register int bindex = rgb_bluecinfo->in_color_space;
+  register int aindex = alpha_indexcinfo->in_color_space;
+  register int ps = rgb_pixelsizecinfo->in_color_space;
+
+  ptr = source->pub.buffer0;
+  if (maxval == MAXJSAMPLE) {
+    if (aindex >= 0)
+      RGB_READ_LOOP(read_pbm_integer(cinfo, infile, maxval),
+                    ptraindex = 0xFF;)
+    else
+      RGB_READ_LOOP(read_pbm_integer(cinfo, infile, maxval),)
+  } else {
+    if (aindex >= 0)
+      RGB_READ_LOOP(rescaleread_pbm_integer(cinfo, infile, maxval),
+                    ptraindex = 0xFF;)
+    else
+      RGB_READ_LOOP(rescaleread_pbm_integer(cinfo, infile, maxval),)
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_text_rgb_cmyk_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PPM files with any maxval and
+   converting to CMYK */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  FILE *infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  ptr = source->pub.buffer0;
+  if (maxval == MAXJSAMPLE) {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE r = read_pbm_integer(cinfo, infile, maxval);
+      JSAMPLE g = read_pbm_integer(cinfo, infile, maxval);
+      JSAMPLE b = read_pbm_integer(cinfo, infile, maxval);
+      rgb_to_cmyk(r, g, b, ptr, ptr + 1, ptr + 2, ptr + 3);
+      ptr += 4;
+    }
+  } else {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE r = rescaleread_pbm_integer(cinfo, infile, maxval);
+      JSAMPLE g = rescaleread_pbm_integer(cinfo, infile, maxval);
+      JSAMPLE b = rescaleread_pbm_integer(cinfo, infile, maxval);
+      rgb_to_cmyk(r, g, b, ptr, ptr + 1, ptr + 2, ptr + 3);
+      ptr += 4;
+    }
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_scaled_gray_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (!ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer0;
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescaleUCH(*bufferptr++);
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_gray_rgb_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PGM files with any maxval
+   and converting to extended RGB */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+  register int rindex = rgb_redcinfo->in_color_space;
+  register int gindex = rgb_greencinfo->in_color_space;
+  register int bindex = rgb_bluecinfo->in_color_space;
+  register int aindex = alpha_indexcinfo->in_color_space;
+  register int ps = rgb_pixelsizecinfo->in_color_space;
+
+  if (!ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer0;
+  bufferptr = source->iobuffer;
+  if (maxval == MAXJSAMPLE) {
+    if (aindex >= 0)
+      GRAY_RGB_READ_LOOP(*bufferptr++, ptraindex = 0xFF;)
+    else
+      GRAY_RGB_READ_LOOP(*bufferptr++,)
+  } else {
+    if (aindex >= 0)
+      GRAY_RGB_READ_LOOP(rescaleUCH(*bufferptr++), ptraindex = 0xFF;)
+    else
+      GRAY_RGB_READ_LOOP(rescaleUCH(*bufferptr++),)
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_gray_cmyk_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PGM files with any maxval
+   and converting to CMYK */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  if (!ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer0;
+  bufferptr = source->iobuffer;
+  if (maxval == MAXJSAMPLE) {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE gray = *bufferptr++;
+      rgb_to_cmyk(gray, gray, gray, ptr, ptr + 1, ptr + 2, ptr + 3);
+      ptr += 4;
+    }
+  } else {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE gray = rescaleUCH(*bufferptr++);
+      rgb_to_cmyk(gray, gray, gray, ptr, ptr + 1, ptr + 2, ptr + 3);
+      ptr += 4;
+    }
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_rgb_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+  register int rindex = rgb_redcinfo->in_color_space;
+  register int gindex = rgb_greencinfo->in_color_space;
+  register int bindex = rgb_bluecinfo->in_color_space;
+  register int aindex = alpha_indexcinfo->in_color_space;
+  register int ps = rgb_pixelsizecinfo->in_color_space;
+
+  if (!ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer0;
+  bufferptr = source->iobuffer;
+  if (maxval == MAXJSAMPLE) {
+    if (aindex >= 0)
+      RGB_READ_LOOP(*bufferptr++, ptraindex = 0xFF;)
+    else
+      RGB_READ_LOOP(*bufferptr++,)
+  } else {
+    if (aindex >= 0)
+      RGB_READ_LOOP(rescaleUCH(*bufferptr++), ptraindex = 0xFF;)
+    else
+      RGB_READ_LOOP(rescaleUCH(*bufferptr++),)
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_rgb_cmyk_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PPM files with any maxval and
+   converting to CMYK */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  if (!ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer0;
+  bufferptr = source->iobuffer;
+  if (maxval == MAXJSAMPLE) {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE r = *bufferptr++;
+      JSAMPLE g = *bufferptr++;
+      JSAMPLE b = *bufferptr++;
+      rgb_to_cmyk(r, g, b, ptr, ptr + 1, ptr + 2, ptr + 3);
+      ptr += 4;
+    }
+  } else {
+    for (col = cinfo->image_width; col > 0; col--) {
+      JSAMPLE r = rescaleUCH(*bufferptr++);
+      JSAMPLE g = rescaleUCH(*bufferptr++);
+      JSAMPLE b = rescaleUCH(*bufferptr++);
+      rgb_to_cmyk(r, g, b, ptr, ptr + 1, ptr + 2, ptr + 3);
+      ptr += 4;
+    }
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_raw_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format files with maxval = MAXJSAMPLE.
+ * In this case we just read right into the JSAMPLE buffer!
+ * Note that same code works for PPM and PGM files.
+ */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+
+  if (!ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_word_gray_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-word-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  if (!ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer0;
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    register unsigned int temp;
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    if (temp > maxval)
+      ERREXIT(cinfo, JERR_PPM_OUTOFRANGE);
+    *ptr++ = rescaletemp;
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_word_rgb_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-word-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+  register int rindex = rgb_redcinfo->in_color_space;
+  register int gindex = rgb_greencinfo->in_color_space;
+  register int bindex = rgb_bluecinfo->in_color_space;
+  register int aindex = alpha_indexcinfo->in_color_space;
+  register int ps = rgb_pixelsizecinfo->in_color_space;
+
+  if (!ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer0;
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    register unsigned int temp;
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    if (temp > maxval)
+      ERREXIT(cinfo, JERR_PPM_OUTOFRANGE);
+    ptrrindex = rescaletemp;
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    if (temp > maxval)
+      ERREXIT(cinfo, JERR_PPM_OUTOFRANGE);
+    ptrgindex = rescaletemp;
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    if (temp > maxval)
+      ERREXIT(cinfo, JERR_PPM_OUTOFRANGE);
+    ptrbindex = rescaletemp;
+    if (aindex >= 0)
+      ptraindex = 0xFF;
+    ptr += ps;
+  }
+  return 1;
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_ppm(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  ppm_source_ptr source = (ppm_source_ptr)sinfo;
+  int c;
+  unsigned int w, h, maxval;
+  boolean need_iobuffer, use_raw_buffer, need_rescale;
+
+  if (getc(source->pub.input_file) != 'P')
+    ERREXIT(cinfo, JERR_PPM_NOT);
+
+  c = getc(source->pub.input_file); /* subformat discriminator character */
+
+  /* detect unsupported variants (ie, PBM) before trying to read header */
+  switch (c) {
+  case '2':                     /* it's a text-format PGM file */
+  case '3':                     /* it's a text-format PPM file */
+  case '5':                     /* it's a raw-format PGM file */
+  case '6':                     /* it's a raw-format PPM file */
+    break;
+  default:
+    ERREXIT(cinfo, JERR_PPM_NOT);
+    break;
+  }
+
+  /* fetch the remaining header info */
+  w = read_pbm_integer(cinfo, source->pub.input_file, 65535);
+  h = read_pbm_integer(cinfo, source->pub.input_file, 65535);
+  maxval = read_pbm_integer(cinfo, source->pub.input_file, 65535);
+
+  if (w <= 0 || h <= 0 || maxval <= 0) /* error check */
+    ERREXIT(cinfo, JERR_PPM_NOT);
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  if (sinfo->max_pixels && (unsigned long long)w * h > sinfo->max_pixels)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+#endif
+
+  cinfo->data_precision = BITS_IN_JSAMPLE; /* we always rescale data to this */
+  cinfo->image_width = (JDIMENSION)w;
+  cinfo->image_height = (JDIMENSION)h;
+  source->maxval = maxval;
+
+  /* initialize flags to most common settings */
+  need_iobuffer = TRUE;         /* do we need an I/O buffer? */
+  use_raw_buffer = FALSE;       /* do we map input buffer onto I/O buffer? */
+  need_rescale = TRUE;          /* do we need a rescale array? */
+
+  switch (c) {
+  case '2':                     /* it's a text-format PGM file */
+    if (cinfo->in_color_space == JCS_UNKNOWN)
+      cinfo->in_color_space = JCS_GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_PGM_TEXT, w, h);
+    if (cinfo->in_color_space == JCS_GRAYSCALE)
+      source->pub.get_pixel_rows = get_text_gray_row;
+    else if (IsExtRGB(cinfo->in_color_space))
+      source->pub.get_pixel_rows = get_text_gray_rgb_row;
+    else if (cinfo->in_color_space == JCS_CMYK)
+      source->pub.get_pixel_rows = get_text_gray_cmyk_row;
+    else
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    need_iobuffer = FALSE;
+    break;
+
+  case '3':                     /* it's a text-format PPM file */
+    if (cinfo->in_color_space == JCS_UNKNOWN)
+      cinfo->in_color_space = JCS_EXT_RGB;
+    TRACEMS2(cinfo, 1, JTRC_PPM_TEXT, w, h);
+    if (IsExtRGB(cinfo->in_color_space))
+      source->pub.get_pixel_rows = get_text_rgb_row;
+    else if (cinfo->in_color_space == JCS_CMYK)
+      source->pub.get_pixel_rows = get_text_rgb_cmyk_row;
+    else
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    need_iobuffer = FALSE;
+    break;
+
+  case '5':                     /* it's a raw-format PGM file */
+    if (cinfo->in_color_space == JCS_UNKNOWN)
+      cinfo->in_color_space = JCS_GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_PGM, w, h);
+    if (maxval > 255) {
+      if (cinfo->in_color_space == JCS_GRAYSCALE)
+        source->pub.get_pixel_rows = get_word_gray_row;
+      else
+        ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    } else if (maxval == MAXJSAMPLE && sizeof(JSAMPLE) == sizeof(U_CHAR) &&
+               cinfo->in_color_space == JCS_GRAYSCALE) {
+      source->pub.get_pixel_rows = get_raw_row;
+      use_raw_buffer = TRUE;
+      need_rescale = FALSE;
+    } else {
+      if (cinfo->in_color_space == JCS_GRAYSCALE)
+        source->pub.get_pixel_rows = get_scaled_gray_row;
+      else if (IsExtRGB(cinfo->in_color_space))
+        source->pub.get_pixel_rows = get_gray_rgb_row;
+      else if (cinfo->in_color_space == JCS_CMYK)
+        source->pub.get_pixel_rows = get_gray_cmyk_row;
+      else
+        ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    }
+    break;
+
+  case '6':                     /* it's a raw-format PPM file */
+    if (cinfo->in_color_space == JCS_UNKNOWN)
+      cinfo->in_color_space = JCS_EXT_RGB;
+    TRACEMS2(cinfo, 1, JTRC_PPM, w, h);
+    if (maxval > 255) {
+      if (IsExtRGB(cinfo->in_color_space))
+        source->pub.get_pixel_rows = get_word_rgb_row;
+      else
+        ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    } else if (maxval == MAXJSAMPLE && sizeof(JSAMPLE) == sizeof(U_CHAR) &&
+#if RGB_RED == 0 && RGB_GREEN == 1 && RGB_BLUE == 2 && RGB_PIXELSIZE == 3
+               (cinfo->in_color_space == JCS_EXT_RGB ||
+                cinfo->in_color_space == JCS_RGB)) {
+#else
+               cinfo->in_color_space == JCS_EXT_RGB) {
+#endif
+      source->pub.get_pixel_rows = get_raw_row;
+      use_raw_buffer = TRUE;
+      need_rescale = FALSE;
+    } else {
+      if (IsExtRGB(cinfo->in_color_space))
+        source->pub.get_pixel_rows = get_rgb_row;
+      else if (cinfo->in_color_space == JCS_CMYK)
+        source->pub.get_pixel_rows = get_rgb_cmyk_row;
+      else
+        ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    }
+    break;
+  }
+
+  if (IsExtRGB(cinfo->in_color_space))
+    cinfo->input_components = rgb_pixelsizecinfo->in_color_space;
+  else if (cinfo->in_color_space == JCS_GRAYSCALE)
+    cinfo->input_components = 1;
+  else if (cinfo->in_color_space == JCS_CMYK)
+    cinfo->input_components = 4;
+
+  /* Allocate space for I/O buffer: 1 or 3 bytes or words/pixel. */
+  if (need_iobuffer) {
+    if (c == '6')
+      source->buffer_width = (size_t)w * 3 *
+        ((maxval <= 255) ? sizeof(U_CHAR) : (2 * sizeof(U_CHAR)));
+    else
+      source->buffer_width = (size_t)w *
+        ((maxval <= 255) ? sizeof(U_CHAR) : (2 * sizeof(U_CHAR)));
+    source->iobuffer = (U_CHAR *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  source->buffer_width);
+  }
+
+  /* Create compressor input buffer. */
+  if (use_raw_buffer) {
+    /* For unscaled raw-input case, we can just map it onto the I/O buffer. */
+    /* Synthesize a JSAMPARRAY pointer structure */
+    source->pixrow = (JSAMPROW)source->iobuffer;
+    source->pub.buffer = &source->pixrow;
+    source->pub.buffer_height = 1;
+  } else {
+    /* Need to translate anyway, so make a separate sample buffer. */
+    source->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE,
+       (JDIMENSION)w * cinfo->input_components, (JDIMENSION)1);
+    source->pub.buffer_height = 1;
+  }
+
+  /* Compute the rescaling array if required. */
+  if (need_rescale) {
+    long val, half_maxval;
+
+    /* On 16-bit-int machines we have to be careful of maxval = 65535 */
+    source->rescale = (JSAMPLE *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  (size_t)(((long)MAX(maxval, 255) + 1L) *
+                                           sizeof(JSAMPLE)));
+    MEMZERO(source->rescale, (size_t)(((long)MAX(maxval, 255) + 1L) *
+                                      sizeof(JSAMPLE)));
+    half_maxval = maxval / 2;
+    for (val = 0; val <= (long)maxval; val++) {
+      /* The multiplication here must be done in 32 bits to avoid overflow */
+      source->rescaleval = (JSAMPLE)((val * MAXJSAMPLE + half_maxval) /
+                                        maxval);
+    }
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_ppm(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for PPM format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_ppm(j_compress_ptr cinfo)
+{
+  ppm_source_ptr source;
+
+  /* Create module interface object */
+  source = (ppm_source_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(ppm_source_struct));
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_ppm;
+  source->pub.finish_input = finish_input_ppm;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  source->pub.max_pixels = 0;
+#endif
+
+  return (cjpeg_source_ptr)source;
+}
+
+#endif /* PPM_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/rdswitch.c Added

@@ -0,0 +1,424 @@
+/*
+ * rdswitch.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2018, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to process some of cjpeg's more complicated
+ * command-line switches.  Switches processed here are:
+ *      -qtables file           Read quantization tables from text file
+ *      -scans file             Read scan script from text file
+ *      -quality N,N,...      Set quality ratings
+ *      -qslots N,N,...       Set component quantization table selectors
+ *      -sample HxV,HxV,...   Set component sampling factors
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include <ctype.h>              /* to declare isdigit(), isspace() */
+
+
+LOCAL(int)
+text_getc(FILE *file)
+/* Read next char, skipping over any comments (# to end of line) */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(file);
+  if (ch == '#') {
+    do {
+      ch = getc(file);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(boolean)
+read_text_integer(FILE *file, long *result, int *termchar)
+/* Read an unsigned decimal integer from a file, store it in result */
+/* Reads one trailing character after the integer; returns it in termchar */
+{
+  register int ch;
+  register long val;
+
+  /* Skip any leading whitespace, detect EOF */
+  do {
+    ch = text_getc(file);
+    if (ch == EOF) {
+      *termchar = ch;
+      return FALSE;
+    }
+  } while (isspace(ch));
+
+  if (!isdigit(ch)) {
+    *termchar = ch;
+    return FALSE;
+  }
+
+  val = ch - '0';
+  while ((ch = text_getc(file)) != EOF) {
+    if (!isdigit(ch))
+      break;
+    val *= 10;
+    val += ch - '0';
+  }
+  *result = val;
+  *termchar = ch;
+  return TRUE;
+}
+
+
+#if JPEG_LIB_VERSION < 70
+static int q_scale_factorNUM_QUANT_TBLS = { 100, 100, 100, 100 };
+#endif
+
+GLOBAL(boolean)
+read_quant_tables(j_compress_ptr cinfo, char *filename, boolean force_baseline)
+/* Read a set of quantization tables from the specified file.
+ * The file is plain ASCII text: decimal numbers with whitespace between.
+ * Comments preceded by '#' may be included in the file.
+ * There may be one to NUM_QUANT_TBLS tables in the file, each of 64 values.
+ * The tables are implicitly numbered 0,1,etc.
+ * NOTE: does not affect the qslots mapping, which will default to selecting
+ * table 0 for luminance (or primary) components, 1 for chrominance components.
+ * You must use -qslots if you want a different component->table mapping.
+ */
+{
+  FILE *fp;
+  int tblno, i, termchar;
+  long val;
+  unsigned int tableDCTSIZE2;
+
+  if ((fp = fopen(filename, "r")) == NULL) {
+    fprintf(stderr, "Can't open table file %s\n", filename);
+    return FALSE;
+  }
+  tblno = 0;
+
+  while (read_text_integer(fp, &val, &termchar)) { /* read 1st element of table */
+    if (tblno >= NUM_QUANT_TBLS) {
+      fprintf(stderr, "Too many tables in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    table0 = (unsigned int)val;
+    for (i = 1; i < DCTSIZE2; i++) {
+      if (!read_text_integer(fp, &val, &termchar)) {
+        fprintf(stderr, "Invalid table data in file %s\n", filename);
+        fclose(fp);
+        return FALSE;
+      }
+      tablei = (unsigned int)val;
+    }
+#if JPEG_LIB_VERSION >= 70
+    jpeg_add_quant_table(cinfo, tblno, table, cinfo->q_scale_factortblno,
+                         force_baseline);
+#else
+    jpeg_add_quant_table(cinfo, tblno, table, q_scale_factortblno,
+                         force_baseline);
+#endif
+    tblno++;
+  }
+
+  if (termchar != EOF) {
+    fprintf(stderr, "Non-numeric data in file %s\n", filename);
+    fclose(fp);
+    return FALSE;
+  }
+
+  fclose(fp);
+  return TRUE;
+}
+
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+
+LOCAL(boolean)
+read_scan_integer(FILE *file, long *result, int *termchar)
+/* Variant of read_text_integer that always looks for a non-space termchar;
+ * this simplifies parsing of punctuation in scan scripts.
+ */
+{
+  register int ch;
+
+  if (!read_text_integer(file, result, termchar))
+    return FALSE;
+  ch = *termchar;
+  while (ch != EOF && isspace(ch))
+    ch = text_getc(file);
+  if (isdigit(ch)) {            /* oops, put it back */
+    if (ungetc(ch, file) == EOF)
+      return FALSE;
+    ch = ' ';
+  } else {
+    /* Any separators other than ';' and ':' are ignored;
+     * this allows user to insert commas, etc, if desired.
+     */
+    if (ch != EOF && ch != ';' && ch != ':')
+      ch = ' ';
+  }
+  *termchar = ch;
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+read_scan_script(j_compress_ptr cinfo, char *filename)
+/* Read a scan script from the specified text file.
+ * Each entry in the file defines one scan to be emitted.
+ * Entries are separated by semicolons ';'.
+ * An entry contains one to four component indexes,
+ * optionally followed by a colon ':' and four progressive-JPEG parameters.
+ * The component indexes denote which component(s) are to be transmitted
+ * in the current scan.  The first component has index 0.
+ * Sequential JPEG is used if the progressive-JPEG parameters are omitted.
+ * The file is free format text: any whitespace may appear between numbers
+ * and the ':' and ';' punctuation marks.  Also, other punctuation (such
+ * as commas or dashes) can be placed between numbers if desired.
+ * Comments preceded by '#' may be included in the file.
+ * Note: we do very little validity checking here;
+ * jcmaster.c will validate the script parameters.
+ */
+{
+  FILE *fp;
+  int scanno, ncomps, termchar;
+  long val;
+  jpeg_scan_info *scanptr;
+#define MAX_SCANS  100          /* quite arbitrary limit */
+  jpeg_scan_info scansMAX_SCANS;
+
+  if ((fp = fopen(filename, "r")) == NULL) {
+    fprintf(stderr, "Can't open scan definition file %s\n", filename);
+    return FALSE;
+  }
+  scanptr = scans;
+  scanno = 0;
+
+  while (read_scan_integer(fp, &val, &termchar)) {
+    if (scanno >= MAX_SCANS) {
+      fprintf(stderr, "Too many scans defined in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    scanptr->component_index0 = (int)val;
+    ncomps = 1;
+    while (termchar == ' ') {
+      if (ncomps >= MAX_COMPS_IN_SCAN) {
+        fprintf(stderr, "Too many components in one scan in file %s\n",
+                filename);
+        fclose(fp);
+        return FALSE;
+      }
+      if (!read_scan_integer(fp, &val, &termchar))
+        goto bogus;
+      scanptr->component_indexncomps = (int)val;
+      ncomps++;
+    }
+    scanptr->comps_in_scan = ncomps;
+    if (termchar == ':') {
+      if (!read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+        goto bogus;
+      scanptr->Ss = (int)val;
+      if (!read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+        goto bogus;
+      scanptr->Se = (int)val;
+      if (!read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+        goto bogus;
+      scanptr->Ah = (int)val;
+      if (!read_scan_integer(fp, &val, &termchar))
+        goto bogus;
+      scanptr->Al = (int)val;
+    } else {
+      /* set non-progressive parameters */
+      scanptr->Ss = 0;
+      scanptr->Se = DCTSIZE2 - 1;
+      scanptr->Ah = 0;
+      scanptr->Al = 0;
+    }
+    if (termchar != ';' && termchar != EOF) {
+bogus:
+      fprintf(stderr, "Invalid scan entry format in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    scanptr++, scanno++;
+  }
+
+  if (termchar != EOF) {
+    fprintf(stderr, "Non-numeric data in file %s\n", filename);
+    fclose(fp);
+    return FALSE;
+  }
+
+  if (scanno > 0) {
+    /* Stash completed scan list in cinfo structure.
+     * NOTE: for cjpeg's use, JPOOL_IMAGE is the right lifetime for this data,
+     * but if you want to compress multiple images you'd want JPOOL_PERMANENT.
+     */
+    scanptr = (jpeg_scan_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  scanno * sizeof(jpeg_scan_info));
+    MEMCOPY(scanptr, scans, scanno * sizeof(jpeg_scan_info));
+    cinfo->scan_info = scanptr;
+    cinfo->num_scans = scanno;
+  }
+
+  fclose(fp);
+  return TRUE;
+}
+
+#endif /* C_MULTISCAN_FILES_SUPPORTED */
+
+
+#if JPEG_LIB_VERSION < 70
+/* These are the sample quantization tables given in Annex K (Clause K.1) of
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+ * The spec says that the values given produce "good" quality, and
+ * when divided by 2, "very good" quality.
+ */
+static const unsigned int std_luminance_quant_tblDCTSIZE2 = {
+  16,  11,  10,  16,  24,  40,  51,  61,
+  12,  12,  14,  19,  26,  58,  60,  55,
+  14,  13,  16,  24,  40,  57,  69,  56,
+  14,  17,  22,  29,  51,  87,  80,  62,
+  18,  22,  37,  56,  68, 109, 103,  77,
+  24,  35,  55,  64,  81, 104, 113,  92,
+  49,  64,  78,  87, 103, 121, 120, 101,
+  72,  92,  95,  98, 112, 100, 103,  99
+};
+static const unsigned int std_chrominance_quant_tblDCTSIZE2 = {
+  17,  18,  24,  47,  99,  99,  99,  99,
+  18,  21,  26,  66,  99,  99,  99,  99,
+  24,  26,  56,  99,  99,  99,  99,  99,
+  47,  66,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99
+};
+
+
+LOCAL(void)
+jpeg_default_qtables(j_compress_ptr cinfo, boolean force_baseline)
+{
+  jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl, q_scale_factor0,
+                       force_baseline);
+  jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl, q_scale_factor1,
+                       force_baseline);
+}
+#endif
+
+
+GLOBAL(boolean)
+set_quality_ratings(j_compress_ptr cinfo, char *arg, boolean force_baseline)
+/* Process a quality-ratings parameter string, of the form
+ *     N,N,...
+ * If there are more q-table slots than parameters, the last value is replicated.
+ */
+{
+  int val = 75;                 /* default value */
+  int tblno;
+  char ch;
+
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    if (*arg) {
+      ch = ',';                 /* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c", &val, &ch) < 1)
+        return FALSE;
+      if (ch != ',')            /* syntax check */
+        return FALSE;
+      /* Convert user 0-100 rating to percentage scaling */
+#if JPEG_LIB_VERSION >= 70
+      cinfo->q_scale_factortblno = jpeg_quality_scaling(val);
+#else
+      q_scale_factortblno = jpeg_quality_scaling(val);
+#endif
+      while (*arg && *arg++ != ','); /* advance to next segment of arg
+                                        string */
+    } else {
+      /* reached end of parameter, set remaining factors to last value */
+#if JPEG_LIB_VERSION >= 70
+      cinfo->q_scale_factortblno = jpeg_quality_scaling(val);
+#else
+      q_scale_factortblno = jpeg_quality_scaling(val);
+#endif
+    }
+  }
+  jpeg_default_qtables(cinfo, force_baseline);
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+set_quant_slots(j_compress_ptr cinfo, char *arg)
+/* Process a quantization-table-selectors parameter string, of the form
+ *     N,N,...
+ * If there are more components than parameters, the last value is replicated.
+ */
+{
+  int val = 0;                  /* default table # */
+  int ci;
+  char ch;
+
+  for (ci = 0; ci < MAX_COMPONENTS; ci++) {
+    if (*arg) {
+      ch = ',';                 /* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c", &val, &ch) < 1)
+        return FALSE;
+      if (ch != ',')            /* syntax check */
+        return FALSE;
+      if (val < 0 || val >= NUM_QUANT_TBLS) {
+        fprintf(stderr, "JPEG quantization tables are numbered 0..%d\n",
+                NUM_QUANT_TBLS - 1);
+        return FALSE;
+      }
+      cinfo->comp_infoci.quant_tbl_no = val;
+      while (*arg && *arg++ != ','); /* advance to next segment of arg
+                                        string */
+    } else {
+      /* reached end of parameter, set remaining components to last table */
+      cinfo->comp_infoci.quant_tbl_no = val;
+    }
+  }
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+set_sample_factors(j_compress_ptr cinfo, char *arg)
+/* Process a sample-factors parameter string, of the form
+ *     HxV,HxV,...
+ * If there are more components than parameters, "1x1" is assumed for the rest.
+ */
+{
+  int ci, val1, val2;
+  char ch1, ch2;
+
+  for (ci = 0; ci < MAX_COMPONENTS; ci++) {
+    if (*arg) {
+      ch2 = ',';                /* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c%d%c", &val1, &ch1, &val2, &ch2) < 3)
+        return FALSE;
+      if ((ch1 != 'x' && ch1 != 'X') || ch2 != ',') /* syntax check */
+        return FALSE;
+      if (val1 <= 0 || val1 > 4 || val2 <= 0 || val2 > 4) {
+        fprintf(stderr, "JPEG sampling factors must be 1..4\n");
+        return FALSE;
+      }
+      cinfo->comp_infoci.h_samp_factor = val1;
+      cinfo->comp_infoci.v_samp_factor = val2;
+      while (*arg && *arg++ != ',');  /* advance to next segment of arg
+                                         string */
+    } else {
+      /* reached end of parameter, set remaining components to 1x1 sampling */
+      cinfo->comp_infoci.h_samp_factor = 1;
+      cinfo->comp_infoci.v_samp_factor = 1;
+    }
+  }
+  return TRUE;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/rdtarga.c Added

@@ -0,0 +1,508 @@
+/*
+ * rdtarga.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * Modified 2017 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2018, 2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in Targa format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed Targa format).
+ *
+ * Based on code contributed by Lee Daniel Crocker.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef TARGA_SUPPORTED
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int)(x))
+
+
+#define ReadOK(file, buffer, len) \
+  (JFREAD(file, buffer, len) == ((size_t)(len)))
+
+
+/* Private version of data source object */
+
+typedef struct _tga_source_struct *tga_source_ptr;
+
+typedef struct _tga_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  j_compress_ptr cinfo;         /* back link saves passing separate parm */
+
+  JSAMPARRAY colormap;          /* Targa colormap (converted to my format) */
+
+  jvirt_sarray_ptr whole_image; /* Needed if funny input row order */
+  JDIMENSION current_row;       /* Current logical row number to read */
+
+  /* Pointer to routine to extract next Targa pixel from input file */
+  void (*read_pixel) (tga_source_ptr sinfo);
+
+  /* Result of read_pixel is delivered here: */
+  U_CHAR tga_pixel4;
+
+  int pixel_size;               /* Bytes per Targa pixel (1 to 4) */
+  int cmap_length;              /* colormap length */
+
+  /* State info for reading RLE-coded pixels; both counts must be init to 0 */
+  int block_count;              /* # of pixels remaining in RLE block */
+  int dup_pixel_count;          /* # of times to duplicate previous pixel */
+
+  /* This saves the correct pixel-row-expansion method for preload_image */
+  JDIMENSION (*get_pixel_rows) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+} tga_source_struct;
+
+
+/* For expanding 5-bit pixel values to 8-bit with best rounding */
+
+static const UINT8 c5to8bits32 = {
+    0,   8,  16,  25,  33,  41,  49,  58,
+   66,  74,  82,  90,  99, 107, 115, 123,
+  132, 140, 148, 156, 165, 173, 181, 189,
+  197, 206, 214, 222, 230, 239, 247, 255
+};
+
+
+
+LOCAL(int)
+read_byte(tga_source_ptr sinfo)
+/* Read next byte from Targa file */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int c;
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  return c;
+}
+
+
+LOCAL(void)
+read_colormap(tga_source_ptr sinfo, int cmaplen, int mapentrysize)
+/* Read the colormap from a Targa file */
+{
+  int i;
+
+  /* Presently only handles 24-bit BGR format */
+  if (mapentrysize != 24)
+    ERREXIT(sinfo->cinfo, JERR_TGA_BADCMAP);
+
+  for (i = 0; i < cmaplen; i++) {
+    sinfo->colormap2i = (JSAMPLE)read_byte(sinfo);
+    sinfo->colormap1i = (JSAMPLE)read_byte(sinfo);
+    sinfo->colormap0i = (JSAMPLE)read_byte(sinfo);
+  }
+}
+
+
+/*
+ * read_pixel methods: get a single pixel from Targa file into tga_pixel
+ */
+
+METHODDEF(void)
+read_non_rle_pixel(tga_source_ptr sinfo)
+/* Read one Targa pixel from the input file; no RLE expansion */
+{
+  register int i;
+
+  for (i = 0; i < sinfo->pixel_size; i++) {
+    sinfo->tga_pixeli = (U_CHAR)read_byte(sinfo);
+  }
+}
+
+
+METHODDEF(void)
+read_rle_pixel(tga_source_ptr sinfo)
+/* Read one Targa pixel from the input file, expanding RLE data as needed */
+{
+  register int i;
+
+  /* Duplicate previously read pixel? */
+  if (sinfo->dup_pixel_count > 0) {
+    sinfo->dup_pixel_count--;
+    return;
+  }
+
+  /* Time to read RLE block header? */
+  if (--sinfo->block_count < 0) { /* decrement pixels remaining in block */
+    i = read_byte(sinfo);
+    if (i & 0x80) {             /* Start of duplicate-pixel block? */
+      sinfo->dup_pixel_count = i & 0x7F; /* number of dups after this one */
+      sinfo->block_count = 0;   /* then read new block header */
+    } else {
+      sinfo->block_count = i & 0x7F; /* number of pixels after this one */
+    }
+  }
+
+  /* Read next pixel */
+  for (i = 0; i < sinfo->pixel_size; i++) {
+    sinfo->tga_pixeli = (U_CHAR)read_byte(sinfo);
+  }
+}
+
+
+/*
+ * Read one row of pixels.
+ *
+ * We provide several different versions depending on input file format.
+ */
+
+
+METHODDEF(JDIMENSION)
+get_8bit_gray_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit grayscale pixels */
+{
+  tga_source_ptr source = (tga_source_ptr)sinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = source->pub.buffer0;
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    *ptr++ = (JSAMPLE)UCH(source->tga_pixel0);
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_8bit_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit colormap indexes */
+{
+  tga_source_ptr source = (tga_source_ptr)sinfo;
+  register int t;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  register JSAMPARRAY colormap = source->colormap;
+  int cmaplen = source->cmap_length;
+
+  ptr = source->pub.buffer0;
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    t = UCH(source->tga_pixel0);
+    if (t >= cmaplen)
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    *ptr++ = colormap0t;
+    *ptr++ = colormap1t;
+    *ptr++ = colormap2t;
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_16bit_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 16-bit pixels */
+{
+  tga_source_ptr source = (tga_source_ptr)sinfo;
+  register int t;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = source->pub.buffer0;
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    t = UCH(source->tga_pixel0);
+    t += UCH(source->tga_pixel1) << 8;
+    /* We expand 5 bit data to 8 bit sample width.
+     * The format of the 16-bit (LSB first) input word is
+     *     xRRRRRGGGGGBBBBB
+     */
+    ptr2 = (JSAMPLE)c5to8bitst & 0x1F;
+    t >>= 5;
+    ptr1 = (JSAMPLE)c5to8bitst & 0x1F;
+    t >>= 5;
+    ptr0 = (JSAMPLE)c5to8bitst & 0x1F;
+    ptr += 3;
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_24bit_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 24-bit pixels */
+{
+  tga_source_ptr source = (tga_source_ptr)sinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = source->pub.buffer0;
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    *ptr++ = (JSAMPLE)UCH(source->tga_pixel2); /* change BGR to RGB order */
+    *ptr++ = (JSAMPLE)UCH(source->tga_pixel1);
+    *ptr++ = (JSAMPLE)UCH(source->tga_pixel0);
+  }
+  return 1;
+}
+
+/*
+ * Targa also defines a 32-bit pixel format with order B,G,R,A.
+ * We presently ignore the attribute byte, so the code for reading
+ * these pixels is identical to the 24-bit routine above.
+ * This works because the actual pixel length is only known to read_pixel.
+ */
+
+#define get_32bit_row  get_24bit_row
+
+
+/*
+ * This method is for re-reading the input data in standard top-down
+ * row order.  The entire image has already been read into whole_image
+ * with proper conversion of pixel format, but it's in a funny row order.
+ */
+
+METHODDEF(JDIMENSION)
+get_memory_row(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr)sinfo;
+  JDIMENSION source_row;
+
+  /* Compute row of source that maps to current_row of normal order */
+  /* For now, assume image is bottom-up and not interlaced. */
+  /* NEEDS WORK to support interlaced images! */
+  source_row = cinfo->image_height - source->current_row - 1;
+
+  /* Fetch that row from virtual array */
+  source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr)cinfo, source->whole_image,
+     source_row, (JDIMENSION)1, FALSE);
+
+  source->current_row++;
+  return 1;
+}
+
+
+/*
+ * This method loads the image into whole_image during the first call on
+ * get_pixel_rows.  The get_pixel_rows pointer is then adjusted to call
+ * get_memory_row on subsequent calls.
+ */
+
+METHODDEF(JDIMENSION)
+preload_image(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr)sinfo;
+  JDIMENSION row;
+  cd_progress_ptr progress = (cd_progress_ptr)cinfo->progress;
+
+  /* Read the data into a virtual array in input-file row order. */
+  for (row = 0; row < cinfo->image_height; row++) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long)row;
+      progress->pub.pass_limit = (long)cinfo->image_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr)cinfo);
+    }
+    source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr)cinfo, source->whole_image, row, (JDIMENSION)1, TRUE);
+    (*source->get_pixel_rows) (cinfo, sinfo);
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Set up to read from the virtual array in unscrambled order */
+  source->pub.get_pixel_rows = get_memory_row;
+  source->current_row = 0;
+  /* And read the first row */
+  return get_memory_row(cinfo, sinfo);
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_tga(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr)sinfo;
+  U_CHAR targaheader18;
+  int idlen, cmaptype, subtype, flags, interlace_type, components;
+  unsigned int width, height, maplen;
+  boolean is_bottom_up;
+
+#define GET_2B(offset) \
+  ((unsigned int)UCH(targaheaderoffset) + \
+   (((unsigned int)UCH(targaheaderoffset + 1)) << 8))
+
+  if (!ReadOK(source->pub.input_file, targaheader, 18))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+
+  /* Pretend "15-bit" pixels are 16-bit --- we ignore attribute bit anyway */
+  if (targaheader16 == 15)
+    targaheader16 = 16;
+
+  idlen = UCH(targaheader0);
+  cmaptype = UCH(targaheader1);
+  subtype = UCH(targaheader2);
+  maplen = GET_2B(5);
+  width = GET_2B(12);
+  height = GET_2B(14);
+  source->pixel_size = UCH(targaheader16) >> 3;
+  flags = UCH(targaheader17); /* Image Descriptor byte */
+
+  is_bottom_up = ((flags & 0x20) == 0); /* bit 5 set => top-down */
+  interlace_type = flags >> 6;  /* bits 6/7 are interlace code */
+
+  if (cmaptype > 1 ||           /* cmaptype must be 0 or 1 */
+      source->pixel_size < 1 || source->pixel_size > 4 ||
+      (UCH(targaheader16) & 7) != 0 || /* bits/pixel must be multiple of 8 */
+      interlace_type != 0 ||      /* currently don't allow interlaced image */
+      width == 0 || height == 0)  /* image width/height must be non-zero */
+    ERREXIT(cinfo, JERR_TGA_BADPARMS);
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  if (sinfo->max_pixels &&
+      (unsigned long long)width * height > sinfo->max_pixels)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+#endif
+
+  if (subtype > 8) {
+    /* It's an RLE-coded file */
+    source->read_pixel = read_rle_pixel;
+    source->block_count = source->dup_pixel_count = 0;
+    subtype -= 8;
+  } else {
+    /* Non-RLE file */
+    source->read_pixel = read_non_rle_pixel;
+  }
+
+  /* Now should have subtype 1, 2, or 3 */
+  components = 3;               /* until proven different */
+  cinfo->in_color_space = JCS_RGB;
+
+  switch (subtype) {
+  case 1:                       /* Colormapped image */
+    if (source->pixel_size == 1 && cmaptype == 1)
+      source->get_pixel_rows = get_8bit_row;
+    else
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    TRACEMS2(cinfo, 1, JTRC_TGA_MAPPED, width, height);
+    break;
+  case 2:                       /* RGB image */
+    switch (source->pixel_size) {
+    case 2:
+      source->get_pixel_rows = get_16bit_row;
+      break;
+    case 3:
+      source->get_pixel_rows = get_24bit_row;
+      break;
+    case 4:
+      source->get_pixel_rows = get_32bit_row;
+      break;
+    default:
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+      break;
+    }
+    TRACEMS2(cinfo, 1, JTRC_TGA, width, height);
+    break;
+  case 3:                       /* Grayscale image */
+    components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    if (source->pixel_size == 1)
+      source->get_pixel_rows = get_8bit_gray_row;
+    else
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    TRACEMS2(cinfo, 1, JTRC_TGA_GRAY, width, height);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    break;
+  }
+
+  if (is_bottom_up) {
+    /* Create a virtual array to buffer the upside-down image. */
+    source->whole_image = (*cinfo->mem->request_virt_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE, FALSE,
+       (JDIMENSION)width * components, (JDIMENSION)height, (JDIMENSION)1);
+    if (cinfo->progress != NULL) {
+      cd_progress_ptr progress = (cd_progress_ptr)cinfo->progress;
+      progress->total_extra_passes++; /* count file input as separate pass */
+    }
+    /* source->pub.buffer will point to the virtual array. */
+    source->pub.buffer_height = 1; /* in case anyone looks at it */
+    source->pub.get_pixel_rows = preload_image;
+  } else {
+    /* Don't need a virtual array, but do need a one-row input buffer. */
+    source->whole_image = NULL;
+    source->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE,
+       (JDIMENSION)width * components, (JDIMENSION)1);
+    source->pub.buffer_height = 1;
+    source->pub.get_pixel_rows = source->get_pixel_rows;
+  }
+
+  while (idlen--)               /* Throw away ID field */
+    (void)read_byte(source);
+
+  if (maplen > 0) {
+    if (maplen > 256 || GET_2B(3) != 0)
+      ERREXIT(cinfo, JERR_TGA_BADCMAP);
+    /* Allocate space to store the colormap */
+    source->colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE, (JDIMENSION)maplen, (JDIMENSION)3);
+    source->cmap_length = (int)maplen;
+    /* and read it from the file */
+    read_colormap(source, (int)maplen, UCH(targaheader7));
+  } else {
+    if (cmaptype)               /* but you promised a cmap! */
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    source->colormap = NULL;
+    source->cmap_length = 0;
+  }
+
+  cinfo->input_components = components;
+  cinfo->data_precision = 8;
+  cinfo->image_width = width;
+  cinfo->image_height = height;
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_tga(j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for Targa format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_targa(j_compress_ptr cinfo)
+{
+  tga_source_ptr source;
+
+  /* Create module interface object */
+  source = (tga_source_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(tga_source_struct));
+  source->cinfo = cinfo;        /* make back link for subroutines */
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_tga;
+  source->pub.finish_input = finish_input_tga;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  source->pub.max_pixels = 0;
+#endif
+
+  return (cjpeg_source_ptr)source;
+}
+
+#endif /* TARGA_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/Config.cmake.in Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/Distribution.xml.in Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/License.rtf Added

@@ -0,0 +1,20 @@
+{\rtf1\ansi\ansicpg1252\cocoartf1038\cocoasubrtf350
+{\fonttbl\f0\fswiss\fcharset0 Helvetica;}
+{\colortbl;\red255\green255\blue255;}
+{\*\listtable{\list\listtemplateid1\listhybrid{\listlevel\levelnfc23\levelnfcn23\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{disc\}}{\leveltext\leveltemplateid1\'01\uc0\u8226 ;}{\levelnumbers;}\fi-360\li720\lin720 }{\listname ;}\listid1}}
+{\*\listoverridetable{\listoverride\listid1\listoverridecount0\ls1}}
+\margl1440\margr1440\vieww9820\viewh8480\viewkind0
+\deftab720
+\pard\pardeftab720
+
+\f0\fs24 \cf0 Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:\
+\
+\pard\tx220\tx720\pardeftab720\li720\fi-720
+\ls1\ilvl0\cf0 {\listtext	\'95	}Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.\
+{\listtext	\'95	}Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.\
+{\listtext	\'95	}Neither the name of the libjpeg-turbo Project nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.\
+\pard\pardeftab720\qc
+\cf0 \
+\pard\pardeftab720
+\cf0 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS", AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;  LOSS OF USE, DATA, OR PROFITS;  OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\
+}
\ No newline at end of file

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/ReadMe.txt Added

@@ -0,0 +1,5 @@
+libjpeg-turbo is a JPEG image codec that uses SIMD instructions to accelerate baseline JPEG compression and decompression on x86, x86-64, Arm, PowerPC, and MIPS systems, as well as progressive JPEG compression on x86, x86-64, and Arm systems.  On such systems, libjpeg-turbo is generally 2-6x as fast as libjpeg, all else being equal.  On other types of systems, libjpeg-turbo can still outperform libjpeg by a significant amount, by virtue of its highly-optimized Huffman coding routines.  In many cases, the performance of libjpeg-turbo rivals that of proprietary high-speed JPEG codecs.
+
+libjpeg-turbo implements both the traditional libjpeg API as well as the less powerful but more straightforward TurboJPEG API.  libjpeg-turbo also features colorspace extensions that allow it to compress from/decompress to 32-bit and big-endian pixel buffers (RGBX, XBGR, etc.), as well as a full-featured Java interface.
+
+libjpeg-turbo was originally based on libjpeg/SIMD, an MMX-accelerated derivative of libjpeg v6b developed by Miyasaka Masaru.  The TigerVNC and VirtualGL projects made numerous enhancements to the codec in 2009, and in early 2010, libjpeg-turbo spun off into an independent project, with the goal of making high-speed JPEG compression/decompression technology available to a broader range of users and developers.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/Welcome.rtf.in Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/deb-control.in Added

@@ -0,0 +1,31 @@
+Package: {__PKGNAME}
+Version: @VERSION@-@BUILD@
+Section: misc
+Priority: optional
+Architecture: {__ARCH}
+Essential: no
+Maintainer: @PKGVENDOR@ <@PKGEMAIL@>
+Homepage: @PKGURL@
+Installed-Size: {__SIZE}
+Description: A SIMD-accelerated JPEG codec that provides both the libjpeg and TurboJPEG APIs
+ libjpeg-turbo is a JPEG image codec that uses SIMD instructions to accelerate
+ baseline JPEG compression and decompression on x86, x86-64, Arm, PowerPC, and
+ MIPS systems, as well as progressive JPEG compression on x86, x86-64, and
+ Arm systems.  On such systems, libjpeg-turbo is generally 2-6x as fast as
+ libjpeg, all else being equal.  On other types of systems, libjpeg-turbo can
+ still outperform libjpeg by a significant amount, by virtue of its
+ highly-optimized Huffman coding routines.  In many cases, the performance of
+ libjpeg-turbo rivals that of proprietary high-speed JPEG codecs.
+ .
+ libjpeg-turbo implements both the traditional libjpeg API as well as the less
+ powerful but more straightforward TurboJPEG API.  libjpeg-turbo also features
+ colorspace extensions that allow it to compress from/decompress to 32-bit and
+ big-endian pixel buffers (RGBX, XBGR, etc.), as well as a full-featured Java
+ interface.
+ .
+ libjpeg-turbo was originally based on libjpeg/SIMD, an MMX-accelerated
+ derivative of libjpeg v6b developed by Miyasaka Masaru.  The TigerVNC and
+ VirtualGL projects made numerous enhancements to the codec in 2009, and in
+ early 2010, libjpeg-turbo spun off into an independent project, with the goal
+ of making high-speed JPEG compression/decompression technology available to a
+ broader range of users and developers.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/installer.nsi.in Added

@@ -0,0 +1,202 @@
+!include x64.nsh
+Name "@CMAKE_PROJECT_NAME@ SDK for @INST_PLATFORM@"
+OutFile "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}@INST_NAME@.exe"
+InstallDir "@INST_DIR@"
+
+SetCompressor bzip2
+
+Page directory
+Page instfiles
+
+UninstPage uninstConfirm
+UninstPage instfiles
+
+Section "@CMAKE_PROJECT_NAME@ SDK for @INST_PLATFORM@ (required)"
+!ifdef WIN64
+	${If} ${RunningX64}
+	${DisableX64FSRedirection}
+	${Endif}
+!endif
+	SectionIn RO
+!ifdef GCC
+	IfFileExists $SYSDIR/libturbojpeg.dll exists 0
+!else
+	IfFileExists $SYSDIR/turbojpeg.dll exists 0
+!endif
+	goto notexists
+	exists:
+!ifdef GCC
+	MessageBox MB_OK "An existing version of the @CMAKE_PROJECT_NAME@ SDK for @INST_PLATFORM@ is already installed.  Please uninstall it first."
+!else
+	MessageBox MB_OK "An existing version of the @CMAKE_PROJECT_NAME@ SDK for @INST_PLATFORM@ or the TurboJPEG SDK is already installed.  Please uninstall it first."
+!endif
+	quit
+
+	notexists:
+	SetOutPath $SYSDIR
+!ifdef GCC
+	File "@CMAKE_CURRENT_BINARY_DIR@\libturbojpeg.dll"
+!else
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}turbojpeg.dll"
+!endif
+	SetOutPath $INSTDIR\bin
+!ifdef GCC
+	File "@CMAKE_CURRENT_BINARY_DIR@\libturbojpeg.dll"
+!else
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}turbojpeg.dll"
+!endif
+!ifdef GCC
+	File "@CMAKE_CURRENT_BINARY_DIR@\libjpeg-@SO_MAJOR_VERSION@.dll"
+!else
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}jpeg@SO_MAJOR_VERSION@.dll"
+!endif
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}cjpeg.exe"
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}djpeg.exe"
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}jpegtran.exe"
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}tjbench.exe"
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}rdjpgcom.exe"
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}wrjpgcom.exe"
+	SetOutPath $INSTDIR\lib
+!ifdef GCC
+	File "@CMAKE_CURRENT_BINARY_DIR@\libturbojpeg.dll.a"
+	File "@CMAKE_CURRENT_BINARY_DIR@\libturbojpeg.a"
+	File "@CMAKE_CURRENT_BINARY_DIR@\libjpeg.dll.a"
+	File "@CMAKE_CURRENT_BINARY_DIR@\libjpeg.a"
+!else
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}turbojpeg.lib"
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}turbojpeg-static.lib"
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}jpeg.lib"
+	File "@CMAKE_CURRENT_BINARY_DIR@\${BUILDDIR}jpeg-static.lib"
+!endif
+	SetOutPath $INSTDIR\lib\pkgconfig
+	File "@CMAKE_CURRENT_BINARY_DIR@\pkgscripts\libjpeg.pc"
+	File "@CMAKE_CURRENT_BINARY_DIR@\pkgscripts\libturbojpeg.pc"
+	SetOutPath $INSTDIR\lib\cmake\@CMAKE_PROJECT_NAME@
+	File "@CMAKE_CURRENT_BINARY_DIR@\pkgscripts\@CMAKE_PROJECT_NAME@Config.cmake"
+	File "@CMAKE_CURRENT_BINARY_DIR@\pkgscripts\@CMAKE_PROJECT_NAME@ConfigVersion.cmake"
+	File "@CMAKE_CURRENT_BINARY_DIR@\win\@CMAKE_PROJECT_NAME@Targets.cmake"
+	File "@CMAKE_CURRENT_BINARY_DIR@\win\@CMAKE_PROJECT_NAME@Targets-release.cmake"
+!ifdef JAVA
+	SetOutPath $INSTDIR\classes
+	File "@CMAKE_CURRENT_BINARY_DIR@\java\turbojpeg.jar"
+!endif
+	SetOutPath $INSTDIR\include
+	File "@CMAKE_CURRENT_BINARY_DIR@\jconfig.h"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\jerror.h"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\jmorecfg.h"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\jpeglib.h"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\turbojpeg.h"
+	SetOutPath $INSTDIR\doc
+	File "@CMAKE_CURRENT_SOURCE_DIR@\README.ijg"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\README.md"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\LICENSE.md"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\example.txt"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\libjpeg.txt"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\structure.txt"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\usage.txt"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\wizard.txt"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\tjexample.c"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\java\TJExample.java"
+!ifdef GCC
+	SetOutPath $INSTDIR\man\man1
+	File "@CMAKE_CURRENT_SOURCE_DIR@\cjpeg.1"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\djpeg.1"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\jpegtran.1"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\rdjpgcom.1"
+	File "@CMAKE_CURRENT_SOURCE_DIR@\wrjpgcom.1"
+!endif
+
+	WriteRegStr HKLM "SOFTWARE\@INST_REG_NAME@ @VERSION@" "Install_Dir" "$INSTDIR"
+
+	WriteRegStr HKLM "Software\Microsoft\Windows\CurrentVersion\Uninstall\@INST_REG_NAME@ @VERSION@" "DisplayName" "@CMAKE_PROJECT_NAME@ SDK v@VERSION@ for @INST_PLATFORM@"
+	WriteRegStr HKLM "Software\Microsoft\Windows\CurrentVersion\Uninstall\@INST_REG_NAME@ @VERSION@" "UninstallString" '"$INSTDIR\uninstall_@VERSION@.exe"'
+	WriteRegDWORD HKLM "Software\Microsoft\Windows\CurrentVersion\Uninstall\@INST_REG_NAME@ @VERSION@" "NoModify" 1
+	WriteRegDWORD HKLM "Software\Microsoft\Windows\CurrentVersion\Uninstall\@INST_REG_NAME@ @VERSION@" "NoRepair" 1
+	WriteUninstaller "uninstall_@VERSION@.exe"
+SectionEnd
+
+Section "Uninstall"
+!ifdef WIN64
+	${If} ${RunningX64}
+	${DisableX64FSRedirection}
+	${Endif}
+!endif
+
+	SetShellVarContext all
+
+	DeleteRegKey HKLM "Software\Microsoft\Windows\CurrentVersion\Uninstall\@INST_REG_NAME@ @VERSION@"
+	DeleteRegKey HKLM "SOFTWARE\@INST_REG_NAME@ @VERSION@"
+
+!ifdef GCC
+	Delete $INSTDIR\bin\libjpeg-@SO_MAJOR_VERSION@.dll
+	Delete $INSTDIR\bin\libturbojpeg.dll
+	Delete $SYSDIR\libturbojpeg.dll
+	Delete $INSTDIR\lib\libturbojpeg.dll.a
+	Delete $INSTDIR\lib\libturbojpeg.a
+	Delete $INSTDIR\lib\libjpeg.dll.a
+	Delete $INSTDIR\lib\libjpeg.a
+!else
+	Delete $INSTDIR\bin\jpeg@SO_MAJOR_VERSION@.dll
+	Delete $INSTDIR\bin\turbojpeg.dll
+	Delete $SYSDIR\turbojpeg.dll
+	Delete $INSTDIR\lib\jpeg.lib
+	Delete $INSTDIR\lib\jpeg-static.lib
+	Delete $INSTDIR\lib\turbojpeg.lib
+	Delete $INSTDIR\lib\turbojpeg-static.lib
+!endif
+	Delete $INSTDIR\lib\pkgconfig\libjpeg.pc
+	Delete $INSTDIR\lib\pkgconfig\libturbojpeg.pc
+	Delete $INSTDIR\lib\cmake\@CMAKE_PROJECT_NAME@\@CMAKE_PROJECT_NAME@Config.cmake
+	Delete $INSTDIR\lib\cmake\@CMAKE_PROJECT_NAME@\@CMAKE_PROJECT_NAME@ConfigVersion.cmake
+	Delete $INSTDIR\lib\cmake\@CMAKE_PROJECT_NAME@\@CMAKE_PROJECT_NAME@Targets.cmake
+	Delete $INSTDIR\lib\cmake\@CMAKE_PROJECT_NAME@\@CMAKE_PROJECT_NAME@Targets-release.cmake
+!ifdef JAVA
+	Delete $INSTDIR\classes\turbojpeg.jar
+!endif
+	Delete $INSTDIR\bin\cjpeg.exe
+	Delete $INSTDIR\bin\djpeg.exe
+	Delete $INSTDIR\bin\jpegtran.exe
+	Delete $INSTDIR\bin\tjbench.exe
+	Delete $INSTDIR\bin\rdjpgcom.exe
+	Delete $INSTDIR\bin\wrjpgcom.exe
+	Delete $INSTDIR\include\jconfig.h
+	Delete $INSTDIR\include\jerror.h
+	Delete $INSTDIR\include\jmorecfg.h
+	Delete $INSTDIR\include\jpeglib.h
+	Delete $INSTDIR\include\turbojpeg.h
+	Delete $INSTDIR\uninstall_@VERSION@.exe
+	Delete $INSTDIR\doc\README.ijg
+	Delete $INSTDIR\doc\README.md
+	Delete $INSTDIR\doc\LICENSE.md
+	Delete $INSTDIR\doc\example.txt
+	Delete $INSTDIR\doc\libjpeg.txt
+	Delete $INSTDIR\doc\structure.txt
+	Delete $INSTDIR\doc\usage.txt
+	Delete $INSTDIR\doc\wizard.txt
+	Delete $INSTDIR\doc\tjexample.c
+	Delete $INSTDIR\doc\TJExample.java
+!ifdef GCC
+	Delete $INSTDIR\man\man1\cjpeg.1
+	Delete $INSTDIR\man\man1\djpeg.1
+	Delete $INSTDIR\man\man1\jpegtran.1
+	Delete $INSTDIR\man\man1\rdjpgcom.1
+	Delete $INSTDIR\man\man1\wrjpgcom.1
+!endif
+
+	RMDir "$INSTDIR\include"
+	RMDir "$INSTDIR\lib\pkgconfig"
+	RMDir "$INSTDIR\lib\cmake\@CMAKE_PROJECT_NAME@"
+	RMDir "$INSTDIR\lib\cmake"
+	RMDir "$INSTDIR\lib"
+	RMDir "$INSTDIR\doc"
+!ifdef GCC
+	RMDir "$INSTDIR\man\man1"
+	RMDir "$INSTDIR\man"
+!endif
+!ifdef JAVA
+	RMDir "$INSTDIR\classes"
+!endif
+	RMDir "$INSTDIR\bin"
+	RMDir "$INSTDIR"
+
+SectionEnd

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/libjpeg.pc.in Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/libturbojpeg.pc.in Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/makedpkg.in Added

@@ -0,0 +1,121 @@
+#!/bin/sh
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+TMPDIR=
+SUDO=
+
+onexit()
+{
+	if  ! "$TMPDIR" = "" ; then
+		$SUDO rm -rf $TMPDIR
+	fi
+}
+
+uid()
+{
+	id | cut -f2 -d = | cut -f1 -d \(;
+}
+
+safedirmove ()
+{
+	if  "$1" = "$2" ; then
+		return 0
+	fi
+	if  "$1" = "" -o ! -d "$1" ; then
+		echo safedirmove: source dir $1 is not valid
+		return 1
+	fi
+	if  "$2" = "" -o -e "$2" ; then
+		echo safedirmove: dest dir $2 is not valid
+		return 1
+	fi
+	if  "$3" = "" -o -e "$3" ; then
+		echo safedirmove: tmp dir $3 is not valid
+		return 1
+	fi
+	mkdir -p $3
+	mv $1/* $3/
+	rmdir $1
+	mkdir -p $2
+	mv $3/* $2/
+	rmdir $3
+	return 0
+}
+
+makedeb()
+{
+	SUPPLEMENT=$1
+	DIRNAME=$PKGNAME
+
+	if  $SUPPLEMENT = 1 ; then
+		PKGNAME=$PKGNAME\32
+		if  "$DEBARCH" = "i386" ; then
+			DEBARCH=amd64
+		else
+			DEBARCH=arm64
+		fi
+	fi
+
+	umask 022
+	rm -f $PKGNAME\_$VERSION\_$DEBARCH.deb
+	TMPDIR=`mktemp -d /tmp/$PKGNAME-build.XXXXXX`
+	mkdir $TMPDIR/DEBIAN
+
+	if  $SUPPLEMENT = 1 ; then
+		DESTDIR=$TMPDIR @CMAKE_MAKE_PROGRAM@ install
+		rm -rf $TMPDIR$BINDIR
+		if  "$DATAROOTDIR" != "$PREFIX" ; then
+			rm -rf $TMPDIR$DATAROOTDIR
+		fi
+		if  "$JAVADIR" != "" ; then
+			rm -rf $TMPDIR$JAVADIR
+		fi
+		rm -rf $TMPDIR$DOCDIR
+		rm -rf $TMPDIR$INCLUDEDIR
+		rm -rf $TMPDIR$MANDIR
+	else
+		DESTDIR=$TMPDIR @CMAKE_MAKE_PROGRAM@ install
+		if  "$PREFIX" = "@CMAKE_INSTALL_DEFAULT_PREFIX@" -a "$DOCDIR" = "@CMAKE_INSTALL_DEFAULT_PREFIX@/doc" ; then
+			safedirmove $TMPDIR/$DOCDIR $TMPDIR/usr/share/doc/$PKGNAME-$VERSION $TMPDIR/__tmpdoc
+			ln -fs /usr/share/doc/$DIRNAME-$VERSION $TMPDIR$DOCDIR
+		fi
+	fi
+
+	SIZE=`du -s $TMPDIR | cut -f1`
+	(cat pkgscripts/deb-control | sed s/{__PKGNAME}/$PKGNAME/g \
+		| sed s/{__ARCH}/$DEBARCH/g | sed s/{__SIZE}/$SIZE/g \
+		> $TMPDIR/DEBIAN/control)
+
+	/sbin/ldconfig -n $TMPDIR$LIBDIR
+
+	$SUDO chown -Rh root:root $TMPDIR/*
+	dpkg -b $TMPDIR $PKGNAME\_$VERSION\_$DEBARCH.deb
+}
+
+PKGNAME=@PKGNAME@
+VERSION=@VERSION@
+DEBARCH=@DEBARCH@
+PREFIX=@CMAKE_INSTALL_PREFIX@
+BINDIR=@CMAKE_INSTALL_FULL_BINDIR@
+DATAROOTDIR=@CMAKE_INSTALL_FULL_DATAROOTDIR@
+DOCDIR=@CMAKE_INSTALL_FULL_DOCDIR@
+INCLUDEDIR=@CMAKE_INSTALL_FULL_INCLUDEDIR@
+JAVADIR=@CMAKE_INSTALL_FULL_JAVADIR@
+LIBDIR=@CMAKE_INSTALL_FULL_LIBDIR@
+MANDIR=@CMAKE_INSTALL_FULL_MANDIR@
+
+if  ! `uid` -eq 0 ; then
+	SUDO=sudo
+fi
+
+makedeb 0
+if  "$DEBARCH" = "i386" -o "$DEBARCH" = "armel" -o "$DEBARCH" = "armhf" ; then
+	makedeb 1
+fi
+
+exit

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/makemacpkg.in Added

@@ -0,0 +1,207 @@
+#!/bin/sh
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+TMPDIR=
+
+onexit()
+{
+	if  ! "$TMPDIR" = "" ; then
+		rm -rf $TMPDIR
+	fi
+}
+
+safedirmove ()
+{
+	if  "$1" = "$2" ; then
+		return 0
+	fi
+	if  "$1" = "" -o ! -d "$1" ; then
+		echo safedirmove: source dir $1 is not valid
+		return 1
+	fi
+	if  "$2" = "" -o -e "$2" ; then
+		echo safedirmove: dest dir $2 is not valid
+		return 1
+	fi
+	if  "$3" = "" -o -e "$3" ; then
+		echo safedirmove: tmp dir $3 is not valid
+		return 1
+	fi
+	mkdir -p $3
+	mv $1/* $3/
+	rmdir $1
+	mkdir -p $2
+	mv $3/* $2/
+	rmdir $3
+	return 0
+}
+
+usage()
+{
+	echo "$0 -lipo path to lipo"
+	exit 1
+}
+
+PKGNAME=@PKGNAME@
+VERSION=@VERSION@
+BUILD=@BUILD@
+SRCDIR=@CMAKE_CURRENT_SOURCE_DIR@
+BUILDDIRARMV8=@ARMV8_BUILD@
+WITH_JAVA=@WITH_JAVA@
+MACOS_APP_CERT_NAME="@MACOS_APP_CERT_NAME@"
+MACOS_INST_CERT_NAME="@MACOS_INST_CERT_NAME@"
+LIPO=lipo
+
+PREFIX=@CMAKE_INSTALL_PREFIX@
+BINDIR=@CMAKE_INSTALL_FULL_BINDIR@
+DOCDIR=@CMAKE_INSTALL_FULL_DOCDIR@
+LIBDIR=@CMAKE_INSTALL_FULL_LIBDIR@
+
+LIBJPEG_DSO_NAME=libjpeg.@SO_MAJOR_VERSION@.@SO_AGE@.@SO_MINOR_VERSION@.dylib
+TURBOJPEG_DSO_NAME=libturbojpeg.@TURBOJPEG_SO_VERSION@.dylib
+
+while  $# -gt 0 ; do
+	case $1 in
+	-h*)
+		usage 0
+		;;
+	-lipo)
+		if  $# -gt 1 ; then
+			if  ! "$2" =~ -.* ; then
+				LIPO=$2;  shift
+			fi
+		fi
+		;;
+	esac
+	shift
+done
+
+if  -f $PKGNAME-$VERSION.dmg ; then
+	rm -f $PKGNAME-$VERSION.dmg
+fi
+
+umask 022
+TMPDIR=`mktemp -d /tmp/$PKGNAME-build.XXXXXX`
+PKGROOT=$TMPDIR/pkg/Package_Root
+mkdir -p $PKGROOT
+
+DESTDIR=$PKGROOT @CMAKE_MAKE_PROGRAM@ install
+
+if  "$PREFIX" = "@CMAKE_INSTALL_DEFAULT_PREFIX@" -a "$DOCDIR" = "@CMAKE_INSTALL_DEFAULT_PREFIX@/doc" ; then
+	mkdir -p $PKGROOT/Library/Documentation
+	safedirmove $PKGROOT$DOCDIR $PKGROOT/Library/Documentation/$PKGNAME $TMPDIR/__tmpdoc
+	ln -fs /Library/Documentation/$PKGNAME $PKGROOT$DOCDIR
+fi
+
+install_subbuild()
+{
+	BUILDDIR=$1
+	ARCHNAME=$2
+	DIRNAME=$3
+	LIPOARCH=$4
+
+	if  ! -d $BUILDDIR ; then
+		echo ERROR: $ARCHNAME build directory $BUILDDIR does not exist
+		exit 1
+	fi
+	if  ! -f $BUILDDIR/Makefile -a ! -f $BUILDDIR/build.ninja ; then
+		echo ERROR: $ARCHNAME build directory $BUILDDIR is not configured
+		exit 1
+	fi
+	mkdir -p $TMPDIR/dist.$DIRNAME
+	pushd $BUILDDIR
+	DESTDIR=$TMPDIR/dist.$DIRNAME @CMAKE_MAKE_PROGRAM@ install
+	popd
+	$LIPO -create \
+		$PKGROOT/$LIBDIR/$LIBJPEG_DSO_NAME \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$LIBDIR/$LIBJPEG_DSO_NAME \
+		-output $PKGROOT/$LIBDIR/$LIBJPEG_DSO_NAME
+	$LIPO -create \
+		$PKGROOT/$LIBDIR/libjpeg.a \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$LIBDIR/libjpeg.a \
+		-output $PKGROOT/$LIBDIR/libjpeg.a
+	$LIPO -create \
+		$PKGROOT/$LIBDIR/$TURBOJPEG_DSO_NAME \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$LIBDIR/$TURBOJPEG_DSO_NAME \
+		-output $PKGROOT/$LIBDIR/$TURBOJPEG_DSO_NAME
+	$LIPO -create \
+		$PKGROOT/$LIBDIR/libturbojpeg.a \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$LIBDIR/libturbojpeg.a \
+		-output $PKGROOT/$LIBDIR/libturbojpeg.a
+	$LIPO -create \
+		$PKGROOT/$BINDIR/cjpeg \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$BINDIR/cjpeg \
+		-output $PKGROOT/$BINDIR/cjpeg
+	$LIPO -create \
+		$PKGROOT/$BINDIR/djpeg \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$BINDIR/djpeg \
+		-output $PKGROOT/$BINDIR/djpeg
+	$LIPO -create \
+		$PKGROOT/$BINDIR/jpegtran \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$BINDIR/jpegtran \
+		-output $PKGROOT/$BINDIR/jpegtran
+	$LIPO -create \
+		$PKGROOT/$BINDIR/tjbench \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$BINDIR/tjbench \
+		-output $PKGROOT/$BINDIR/tjbench
+	$LIPO -create \
+		$PKGROOT/$BINDIR/rdjpgcom \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$BINDIR/rdjpgcom \
+		-output $PKGROOT/$BINDIR/rdjpgcom
+	$LIPO -create \
+		$PKGROOT/$BINDIR/wrjpgcom \
+		-arch $LIPOARCH $TMPDIR/dist.$DIRNAME/$BINDIR/wrjpgcom \
+		-output $PKGROOT/$BINDIR/wrjpgcom
+}
+
+if  "$BUILDDIRARMV8" != "" ; then
+	install_subbuild $BUILDDIRARMV8 Armv8 armv8 arm64
+fi
+
+install_name_tool -id $LIBDIR/$LIBJPEG_DSO_NAME $PKGROOT/$LIBDIR/$LIBJPEG_DSO_NAME
+install_name_tool -id $LIBDIR/$TURBOJPEG_DSO_NAME $PKGROOT/$LIBDIR/$TURBOJPEG_DSO_NAME
+
+if  "$PREFIX" = "@CMAKE_INSTALL_DEFAULT_PREFIX@" -a "$LIBDIR" = "@CMAKE_INSTALL_DEFAULT_PREFIX@/lib" ; then
+	if  ! -h $PKGROOT/$PREFIX/lib64 ; then
+		ln -fs lib $PKGROOT/$PREFIX/lib64
+	fi
+fi
+
+mkdir -p $TMPDIR/pkg
+
+install -m 755 pkgscripts/uninstall $PKGROOT/$BINDIR/
+
+find $PKGROOT -type f | while read file; do xattr -c $file; done
+
+cp $SRCDIR/release/License.rtf pkgscripts/Welcome.rtf $SRCDIR/release/ReadMe.txt $TMPDIR/pkg/
+
+mkdir $TMPDIR/dmg
+pkgbuild --root $PKGROOT --version $VERSION.$BUILD --identifier @PKGID@ \
+	$TMPDIR/pkg/$PKGNAME.pkg
+SUFFIX=
+if  "$MACOS_INST_CERT_NAME" != "" ; then
+	SUFFIX=-unsigned
+fi
+productbuild --distribution pkgscripts/Distribution.xml \
+	--package-path $TMPDIR/pkg/ --resources $TMPDIR/pkg/ \
+	$TMPDIR/dmg/$PKGNAME$SUFFIX.pkg
+if  "$MACOS_INST_CERT_NAME" != "" ; then
+	productsign --sign "$MACOS_INST_CERT_NAME" --timestamp \
+		$TMPDIR/dmg/$PKGNAME$SUFFIX.pkg $TMPDIR/dmg/$PKGNAME.pkg
+	rm -r $TMPDIR/dmg/$PKGNAME$SUFFIX.pkg
+	pkgutil --check-signature $TMPDIR/dmg/$PKGNAME.pkg
+fi
+hdiutil create -fs HFS+ -volname $PKGNAME-$VERSION \
+	-srcfolder "$TMPDIR/dmg" $TMPDIR/$PKGNAME-$VERSION.dmg
+if  "$MACOS_APP_CERT_NAME" != "" ; then
+	codesign -s "$MACOS_APP_CERT_NAME" --timestamp $TMPDIR/$PKGNAME-$VERSION.dmg
+	codesign -vv $TMPDIR/$PKGNAME-$VERSION.dmg
+fi
+cp $TMPDIR/$PKGNAME-$VERSION.dmg .
+
+exit

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/makerpm.in Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/makesrpm.in Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/maketarball.in Added

@@ -0,0 +1,51 @@
+#!/bin/sh
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+TMPDIR=
+SUDO=
+
+onexit()
+{
+	if  ! "$TMPDIR" = "" ; then
+		rm -rf $TMPDIR
+	fi
+}
+
+uid()
+{
+	id | cut -f2 -d = | cut -f1 -d \(;
+}
+
+PKGNAME=@PKGNAME@
+VERSION=@VERSION@
+ARCH=@CPU_TYPE@
+OS=@CMAKE_SYSTEM_NAME@
+PREFIX=@CMAKE_INSTALL_PREFIX@
+
+umask 022
+rm -f $PKGNAME-$VERSION-$OS-$ARCH.tar.bz2
+TMPDIR=`mktemp -d /tmp/$PKGNAME-build.XXXXXX`
+mkdir -p $TMPDIR/install
+
+DESTDIR=$TMPDIR/install @CMAKE_MAKE_PROGRAM@ install
+echo tartest >$TMPDIR/tartest
+GNUTAR=0
+BSDTAR=0
+tar cf $TMPDIR/tartest.tar --owner=root --group=root -C $TMPDIR tartest >/dev/null 2>&1 && GNUTAR=1
+if  "$GNUTAR" = "1" ; then
+	tar cf - --owner=root --group=root -C $TMPDIR/install .$PREFIX | bzip2 -c >$PKGNAME-$VERSION-$OS-$ARCH.tar.bz2
+else
+	tar cf $TMPDIR/tartest.tar --uid 0 --gid 0 -C $TMPDIR tartest >/dev/null 2>&1 && BSDTAR=1
+	if  "$BSDTAR" = "1" ; then
+		tar cf - --uid=0 --gid=0 -C $TMPDIR/install .$PREFIX | bzip2 -c >$PKGNAME-$VERSION-$OS-$ARCH.tar.bz2
+	else
+		tar cf - -C $TMPDIR/install .$PREFIX | bzip2 -c >$PKGNAME-$VERSION-$OS-$ARCH.tar.bz2
+	fi
+fi
+
+exit

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/rpm.spec.in Added

@@ -0,0 +1,232 @@
+%global _docdir %{_defaultdocdir}/%{name}-%{version}
+%define _prefix  @CMAKE_INSTALL_PREFIX@
+%define _bindir  @CMAKE_INSTALL_FULL_BINDIR@
+%define _datarootdir  @CMAKE_INSTALL_FULL_DATAROOTDIR@
+%define _includedir  @CMAKE_INSTALL_FULL_INCLUDEDIR@
+%define _javadir  @CMAKE_INSTALL_FULL_JAVADIR@
+%define _mandir  @CMAKE_INSTALL_FULL_MANDIR@
+%define _enable_static  @ENABLE_STATIC@
+%define _enable_shared  @ENABLE_SHARED@
+%define _with_turbojpeg  @WITH_TURBOJPEG@
+%define _with_java  @WITH_JAVA@
+
+%if "%{?__isa_bits:1}" == "1"
+%define _bits  %{__isa_bits}
+%else
+# RPM < 4.6
+%if "%{_lib}" == "lib64"
+%define _bits  64
+%else
+%define _bits  32
+%endif
+%endif
+
+%if "%{_bits}" == "64"
+%define _syslibdir /usr/lib64
+%else
+%define _syslibdir /usr/lib
+%endif
+
+#-->%if 1
+%if "%{_bits}" == "64"
+%define _libdir  %{_exec_prefix}/lib64
+%else
+%if "%{_prefix}" == "/opt/libjpeg-turbo"
+%define _libdir  %{_exec_prefix}/lib32
+%endif
+%endif
+#-->%else
+%define _libdir  @CMAKE_INSTALL_FULL_LIBDIR@
+#-->%endif
+
+Summary: A SIMD-accelerated JPEG codec that provides both the libjpeg and TurboJPEG APIs
+Name: @PKGNAME@
+Version: @VERSION@
+Vendor: @PKGVENDOR@
+URL: @PKGURL@
+Group: System Environment/Libraries
+#-->Source0: http://prdownloads.sourceforge.net/@CMAKE_PROJECT_NAME@/@CMAKE_PROJECT_NAME@-%{version}.tar.gz
+Release: @BUILD@
+License: BSD-style
+BuildRoot: %{_blddir}/%{name}-buildroot-%{version}-%{release}
+Requires: /sbin/ldconfig
+%if "%{_bits}" == "64"
+Provides: %{name} = %{version}-%{release}, @CMAKE_PROJECT_NAME@ = %{version}-%{release}, libturbojpeg.so()(64bit)
+%else
+Provides: %{name} = %{version}-%{release}, @CMAKE_PROJECT_NAME@ = %{version}-%{release}, libturbojpeg.so
+%endif
+
+%description
+libjpeg-turbo is a JPEG image codec that uses SIMD instructions to accelerate
+baseline JPEG compression and decompression on x86, x86-64, Arm, PowerPC, and
+MIPS systems, as well as progressive JPEG compression on x86, x86-64, and Arm
+systems.  On such systems, libjpeg-turbo is generally 2-6x as fast as libjpeg,
+all else being equal.  On other types of systems, libjpeg-turbo can still
+outperform libjpeg by a significant amount, by virtue of its highly-optimized
+Huffman coding routines.  In many cases, the performance of libjpeg-turbo
+rivals that of proprietary high-speed JPEG codecs.
+
+libjpeg-turbo implements both the traditional libjpeg API as well as the less
+powerful but more straightforward TurboJPEG API.  libjpeg-turbo also features
+colorspace extensions that allow it to compress from/decompress to 32-bit and
+big-endian pixel buffers (RGBX, XBGR, etc.), as well as a full-featured Java
+interface.
+
+libjpeg-turbo was originally based on libjpeg/SIMD, an MMX-accelerated
+derivative of libjpeg v6b developed by Miyasaka Masaru.  The TigerVNC and
+VirtualGL projects made numerous enhancements to the codec in 2009, and in
+early 2010, libjpeg-turbo spun off into an independent project, with the goal
+of making high-speed JPEG compression/decompression technology available to a
+broader range of users and developers.
+
+#-->%prep
+#-->%setup -q -n @CMAKE_PROJECT_NAME@-%{version}
+
+#-->%build
+#-->cmake -G"Unix Makefiles" -DCMAKE_BUILD_TYPE=@CMAKE_BUILD_TYPE@ \
+#-->  -DBUILD=%{release} \
+#-->  -DCMAKE_INSTALL_BINDIR=%{_bindir} \
+#-->  -DCMAKE_INSTALL_DATAROOTDIR=%{_datarootdir} \
+#-->  -DCMAKE_INSTALL_DOCDIR=%{_docdir} \
+#-->  -DCMAKE_INSTALL_INCLUDEDIR=%{_includedir} \
+#-->  -DCMAKE_INSTALL_JAVADIR=%{_javadir} \
+#-->  -DCMAKE_INSTALL_LIBDIR=%{_libdir} \
+#-->  -DCMAKE_INSTALL_MANDIR=%{_mandir} \
+#-->  -DCMAKE_INSTALL_PREFIX=%{_prefix} \
+#-->  -DCMAKE_POSITION_INDEPENDENT_CODE=@CMAKE_POSITION_INDEPENDENT_CODE@ \
+#-->  -DENABLE_SHARED=@ENABLE_SHARED@ -DENABLE_STATIC=@ENABLE_STATIC@ \
+#-->  -DSO_MAJOR_VERSION=@SO_MAJOR_VERSION@ \
+#-->  -DSO_MINOR_VERSION=@SO_MINOR_VERSION@ \
+#-->  -DJPEG_LIB_VERSION=@JPEG_LIB_VERSION@ \
+#-->  -DREQUIRE_SIMD=@REQUIRE_SIMD@ \
+#-->  -DWITH_12BIT=@WITH_12BIT@ -DWITH_ARITH_DEC=@WITH_ARITH_DEC@ \
+#-->  -DWITH_ARITH_ENC=@WITH_ARITH_ENC@ -DWITH_JAVA=@WITH_JAVA@ \
+#-->  -DWITH_JPEG7=@WITH_JPEG7@ -DWITH_JPEG8=@WITH_JPEG8@ \
+#-->  -DWITH_MEM_SRCDST=@WITH_MEM_SRCDST@ -DWITH_SIMD=@WITH_SIMD@ \
+#-->  -DWITH_TURBOJPEG=@WITH_TURBOJPEG@ .
+#-->make DESTDIR=$RPM_BUILD_ROOT
+
+%install
+rm -rf $RPM_BUILD_ROOT
+DESTDIR=$RPM_BUILD_ROOT @CMAKE_MAKE_PROGRAM@ install
+/sbin/ldconfig -n $RPM_BUILD_ROOT%{_libdir}
+
+#-->%if 0
+
+# This is only needed to support in-tree RPM generation via 'make rpm'.  When
+# building from a SRPM, we control where things are installed via CMake
+# variables.
+
+safedirmove ()
+{
+	if  "$1" = "$2" ; then
+		return 0
+	fi
+	if  "$1" = "" -o ! -d "$1" ; then
+		echo safedirmove: source dir $1 is not valid
+		return 1
+	fi
+	if  "$2" = "" -o -e "$2" ; then
+		echo safedirmove: dest dir $2 is not valid
+		return 1
+	fi
+	if  "$3" = "" -o -e "$3" ; then
+		echo safedirmove: tmp dir $3 is not valid
+		return 1
+	fi
+	mkdir -p $3
+	mv $1/* $3/
+	rmdir $1
+	mkdir -p $2
+	mv $3/* $2/
+	rmdir $3
+	return 0
+}
+
+LJT_DOCDIR=@CMAKE_INSTALL_FULL_DOCDIR@
+if  ! "$LJT_DOCDIR" = "%{_docdir}" ; then
+	safedirmove $RPM_BUILD_ROOT/$LJT_DOCDIR $RPM_BUILD_ROOT/%{_docdir} $RPM_BUILD_ROOT/__tmpdoc
+fi
+
+#-->%endif
+
+LJT_DOCDIR=@CMAKE_INSTALL_FULL_DOCDIR@
+if  "%{_prefix}" = "@CMAKE_INSTALL_DEFAULT_PREFIX@" -a "$LJT_DOCDIR" = "@CMAKE_INSTALL_DEFAULT_PREFIX@/doc" ; then
+	ln -fs %{_docdir} $RPM_BUILD_ROOT/$LJT_DOCDIR
+fi
+
+%post -p /sbin/ldconfig
+
+%postun -p /sbin/ldconfig
+
+%clean
+rm -rf $RPM_BUILD_ROOT
+
+%files
+%defattr(-,root,root)
+%dir %{_docdir}
+%doc %{_docdir}/*
+%dir %{_prefix}
+%if "%{_prefix}" == "@CMAKE_INSTALL_DEFAULT_PREFIX@" && "%{_docdir}" != "%{_prefix}/doc"
+	%{_prefix}/doc
+%endif
+%dir %{_bindir}
+%{_bindir}/cjpeg
+%{_bindir}/djpeg
+%{_bindir}/jpegtran
+%if "%{_with_turbojpeg}" == "1"
+	%{_bindir}/tjbench
+%endif
+%{_bindir}/rdjpgcom
+%{_bindir}/wrjpgcom
+%if "%{_libdir}" != "%{_syslibdir}"
+%dir %{_libdir}
+%endif
+%if "%{_enable_shared}" == "1"
+	%{_libdir}/libjpeg.so.@SO_MAJOR_VERSION@.@SO_AGE@.@SO_MINOR_VERSION@
+	%{_libdir}/libjpeg.so.@SO_MAJOR_VERSION@
+	%{_libdir}/libjpeg.so
+%endif
+%if "%{_enable_static}" == "1"
+	%{_libdir}/libjpeg.a
+%endif
+%dir %{_libdir}/pkgconfig
+%{_libdir}/pkgconfig/libjpeg.pc
+%dir %{_libdir}/cmake
+%dir %{_libdir}/cmake/@CMAKE_PROJECT_NAME@
+%{_libdir}/cmake/@CMAKE_PROJECT_NAME@
+%if "%{_with_turbojpeg}" == "1"
+	%if "%{_enable_shared}" == "1" || "%{_with_java}" == "1"
+		%{_libdir}/libturbojpeg.so.@TURBOJPEG_SO_VERSION@
+		%{_libdir}/libturbojpeg.so.@TURBOJPEG_SO_MAJOR_VERSION@
+		%{_libdir}/libturbojpeg.so
+	%endif
+	%if "%{_enable_static}" == "1"
+		%{_libdir}/libturbojpeg.a
+	%endif
+	%{_libdir}/pkgconfig/libturbojpeg.pc
+%endif
+%dir %{_includedir}
+%{_includedir}/jconfig.h
+%{_includedir}/jerror.h
+%{_includedir}/jmorecfg.h
+%{_includedir}/jpeglib.h
+%if "%{_with_turbojpeg}" == "1"
+	%{_includedir}/turbojpeg.h
+%endif
+%dir %{_mandir}
+%dir %{_mandir}/man1
+%{_mandir}/man1/cjpeg.1*
+%{_mandir}/man1/djpeg.1*
+%{_mandir}/man1/jpegtran.1*
+%{_mandir}/man1/rdjpgcom.1*
+%{_mandir}/man1/wrjpgcom.1*
+%if "%{_prefix}" != "%{_datarootdir}"
+	%dir %{_datarootdir}
+%endif
+%if "%{_with_java}" == "1"
+	%dir %{_javadir}
+	%{_javadir}/turbojpeg.jar
+%endif
+
+%changelog

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/release/uninstall.in Added

@@ -0,0 +1,120 @@
+# Copyright (C)2009-2011, 2013, 2016, 2020 D. R. Commander.
+#                                          All Rights Reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+#
+# - Redistributions of source code must retain the above copyright notice,
+#   this list of conditions and the following disclaimer.
+# - Redistributions in binary form must reproduce the above copyright notice,
+#   this list of conditions and the following disclaimer in the documentation
+#   and/or other materials provided with the distribution.
+# - Neither the name of the libjpeg-turbo Project nor the names of its
+#   contributors may be used to endorse or promote products derived from this
+#   software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+# ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+# POSSIBILITY OF SUCH DAMAGE.
+
+#!/bin/sh
+
+if  ! "`id -u`" = "0" ; then
+	echo "ERROR: This script must be executed as root"
+	exit -1
+fi
+
+PKGNAME=@PKGNAME@
+PKGID=@PKGID@
+RECEIPT=/Library/Receipts/$PKGNAME.pkg
+
+LSBOM=
+if  -d $RECEIPT ; then
+	LSBOM='lsbom -s -f -l '$RECEIPT'/Contents/Archive.bom'
+else
+	LSBOM='pkgutil --files '$PKGID
+fi
+
+mylsbom()
+{
+	$LSBOM || (echo "ERROR: Could not list package contents"; exit -1)
+}
+
+echo Removing package files ...
+EXITSTATUS=0
+pushd /
+mylsbom | while read file; do
+	if  ! -d "$file" ; then rm "$file" 2>&1 || EXITSTATUS=-1; fi
+done
+popd
+
+echo Removing package directories ...
+PREFIX=@CMAKE_INSTALL_PREFIX@
+BINDIR=@CMAKE_INSTALL_FULL_BINDIR@
+DATAROOTDIR=@CMAKE_INSTALL_FULL_DATAROOTDIR@
+INCLUDEDIR=@CMAKE_INSTALL_FULL_INCLUDEDIR@
+JAVADIR=@CMAKE_INSTALL_FULL_JAVADIR@
+LIBDIR=@CMAKE_INSTALL_FULL_LIBDIR@
+MANDIR=@CMAKE_INSTALL_FULL_MANDIR@
+
+if  -d $BINDIR ; then
+	rmdir $BINDIR 2>&1 || EXITSTATUS=-1
+fi
+if  -d $LIBDIR/pkgconfig ; then
+	rmdir $LIBDIR/pkgconfig 2>&1 || EXITSTATUS=-1
+fi
+if  -d $LIBDIR/cmake/@CMAKE_PROJECT_NAME@ ; then
+	rmdir $LIBDIR/cmake/@CMAKE_PROJECT_NAME@ || EXITSTATUS=-1
+fi
+if  -d $LIBDIR/cmake ; then
+	rmdir $LIBDIR/cmake || EXITSTATUS=-1
+fi
+if  -d $LIBDIR ; then
+	rmdir $LIBDIR 2>&1 || EXITSTATUS=-1
+fi
+if  -d $INCLUDEDIR ; then
+	rmdir $INCLUDEDIR 2>&1 || EXITSTATUS=-1
+fi
+if  "$PREFIX" = "@CMAKE_INSTALL_DEFAULT_PREFIX@" -a "$LIBDIR" = "@CMAKE_INSTALL_DEFAULT_PREFIX@/lib" ; then
+	if  -h $LIBDIR\32 ; then
+		rm $LIBDIR\32 2>&1 || EXITSTATUS=-1
+	fi
+	if  -h $LIBDIR\64 ; then
+		rm $LIBDIR\64 2>&1 || EXITSTATUS=-1
+	fi
+fi
+if  -d $MANDIR/man1 ; then
+	rmdir $MANDIR/man1 2>&1 || EXITSTATUS=-1
+fi
+if  -d $MANDIR ; then
+	rmdir $MANDIR 2>&1 || EXITSTATUS=-1
+fi
+if  -d "$JAVADIR" ; then
+	rmdir $JAVADIR 2>&1 || EXITSTATUS=-1
+fi
+if  -d $DATAROOTDIR -a "$DATAROOTDIR" != "$PREFIX" ; then
+	rmdir $DATAROOTDIR 2>&1 || EXITSTATUS=-1
+fi
+if  "$PREFIX" = "@CMAKE_INSTALL_DEFAULT_PREFIX@" -a -h "$PREFIX/doc" ; then
+	rm $PREFIX/doc 2>&1 || EXITSTATUS=-1
+fi
+rmdir $PREFIX 2>&1 || EXITSTATUS=-1
+rmdir /Library/Documentation/$PKGNAME 2>&1 || EXITSTATUS=-1
+
+if  -d $RECEIPT ; then
+	echo Removing package receipt ...
+	rm -r $RECEIPT 2>&1 || EXITSTATUS=-1
+else
+	echo Forgetting package $PKGID ...
+	pkgutil --forget $PKGID
+fi
+
+exit $EXITSTATUS

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/sharedlib Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/sharedlib/CMakeLists.txt Added

@@ -0,0 +1,102 @@
+# Anything that must be linked against the shared C library on Windows must
+# be built in this subdirectory, because CMake doesn't allow us to override
+# the compiler flags for each build type except at directory scope.  Note
+# to CMake developers:  Add a COMPILE_FLAGS_<CONFIG> target property, or
+# better yet, provide a friendly way of configuring a Windows target to use the
+# static C library.
+
+set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/..)
+set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/..)
+set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/..)
+
+if(MSVC)
+  # Build all configurations against shared C library
+  foreach(var CMAKE_C_FLAGS CMAKE_C_FLAGS_DEBUG CMAKE_C_FLAGS_RELEASE
+    CMAKE_C_FLAGS_MINSIZEREL CMAKE_C_FLAGS_RELWITHDEBINFO)
+    if(${var} MATCHES "/MT")
+      string(REGEX REPLACE "/MT" "/MD" ${var} "${${var}}")
+    endif()
+  endforeach()
+endif()
+
+foreach(src ${JPEG_SOURCES})
+  set(JPEG_SRCS ${JPEG_SRCS} ../${src})
+endforeach()
+
+if(WITH_SIMD AND (MSVC_IDE OR XCODE))
+  # This tells CMake that the "source" files haven't been generated yet
+  set_source_files_properties(${SIMD_OBJS} PROPERTIES GENERATED 1)
+endif()
+
+if(WIN32)
+  if(WITH_MEM_SRCDST)
+    set(DEFFILE ../win/jpeg${SO_MAJOR_VERSION}-memsrcdst.def)
+  else()
+    set(DEFFILE ../win/jpeg${SO_MAJOR_VERSION}.def)
+  endif()
+endif()
+add_library(jpeg SHARED ${JPEG_SRCS} ${DEFFILE} $<TARGET_OBJECTS:simd>
+  ${SIMD_OBJS})
+
+set_target_properties(jpeg PROPERTIES SOVERSION ${SO_MAJOR_VERSION}
+  VERSION ${SO_MAJOR_VERSION}.${SO_AGE}.${SO_MINOR_VERSION})
+if(APPLE AND (NOT CMAKE_OSX_DEPLOYMENT_TARGET OR
+              CMAKE_OSX_DEPLOYMENT_TARGET VERSION_GREATER 10.4))
+  if(NOT CMAKE_SHARED_LIBRARY_RUNTIME_C_FLAG)
+    set(CMAKE_SHARED_LIBRARY_RUNTIME_C_FLAG "-Wl,-rpath,")
+  endif()
+  set_target_properties(jpeg PROPERTIES MACOSX_RPATH 1)
+endif()
+if(MAPFLAG)
+  set_target_properties(jpeg PROPERTIES
+    LINK_FLAGS "${MAPFLAG}${CMAKE_CURRENT_BINARY_DIR}/../libjpeg.map")
+endif()
+if(MSVC)
+  set_target_properties(jpeg PROPERTIES
+    RUNTIME_OUTPUT_NAME jpeg${SO_MAJOR_VERSION})
+  # The jsimd_*.c file is built using /MT, so this prevents a linker warning.
+  set_target_properties(jpeg PROPERTIES LINK_FLAGS "/NODEFAULTLIB:LIBCMT /NODEFAULTLIB:LIBCMTD")
+elseif(MINGW)
+  set_target_properties(jpeg PROPERTIES SUFFIX -${SO_MAJOR_VERSION}.dll)
+endif()
+
+if(WIN32)
+  set(USE_SETMODE "-DUSE_SETMODE")
+endif()
+if(WITH_12BIT)
+  set(COMPILE_FLAGS "-DGIF_SUPPORTED -DPPM_SUPPORTED ${USE_SETMODE}")
+else()
+  set(COMPILE_FLAGS "-DBMP_SUPPORTED -DGIF_SUPPORTED -DPPM_SUPPORTED -DTARGA_SUPPORTED ${USE_SETMODE}")
+  set(CJPEG_BMP_SOURCES ../rdbmp.c ../rdtarga.c)
+  set(DJPEG_BMP_SOURCES ../wrbmp.c ../wrtarga.c)
+endif()
+
+add_executable(cjpeg ../cjpeg.c ../cdjpeg.c ../rdgif.c ../rdppm.c
+  ../rdswitch.c ${CJPEG_BMP_SOURCES})
+set_property(TARGET cjpeg PROPERTY COMPILE_FLAGS ${COMPILE_FLAGS})
+target_link_libraries(cjpeg jpeg)
+
+add_executable(djpeg ../djpeg.c ../cdjpeg.c ../rdcolmap.c ../rdswitch.c
+  ../wrgif.c ../wrppm.c ${DJPEG_BMP_SOURCES})
+set_property(TARGET djpeg PROPERTY COMPILE_FLAGS ${COMPILE_FLAGS})
+target_link_libraries(djpeg jpeg)
+
+add_executable(jpegtran ../jpegtran.c ../cdjpeg.c ../rdswitch.c ../transupp.c)
+target_link_libraries(jpegtran jpeg)
+set_property(TARGET jpegtran PROPERTY COMPILE_FLAGS "${USE_SETMODE}")
+
+add_executable(jcstest ../jcstest.c)
+target_link_libraries(jcstest jpeg)
+
+install(TARGETS jpeg EXPORT ${CMAKE_PROJECT_NAME}Targets
+  INCLUDES DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
+  ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+install(TARGETS cjpeg djpeg jpegtran
+  RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+if(NOT CMAKE_VERSION VERSION_LESS "3.1" AND MSVC AND
+  CMAKE_C_LINKER_SUPPORTS_PDB)
+  install(FILES "$<TARGET_PDB_FILE:jpeg>"
+    DESTINATION ${CMAKE_INSTALL_BINDIR} OPTIONAL)
+endif()

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/CMakeLists.txt Added

@@ -0,0 +1,453 @@
+macro(simd_fail message)
+  if(REQUIRE_SIMD)
+    message(FATAL_ERROR "${message}.")
+  else()
+    message(WARNING "${message}.  Performance will suffer.")
+    set(WITH_SIMD 0 PARENT_SCOPE)
+  endif()
+endmacro()
+
+
+###############################################################################
+# x86-64 (NASM)
+###############################################################################
+
+if(CPU_TYPE STREQUAL "x86_64" OR CPU_TYPE STREQUAL "i386")
+
+set(CMAKE_ASM_NASM_FLAGS_DEBUG_INIT "-g")
+set(CMAKE_ASM_NASM_FLAGS_RELWITHDEBINFO_INIT "-g")
+
+# Allow the location of the NASM executable to be specified using the ASM_NASM
+# environment variable.  This should happen automatically, but unfortunately
+# enable_language(ASM_NASM) doesn't parse the ASM_NASM environment variable
+# until after CMAKE_ASM_NASM_COMPILER has been populated with the results of
+# searching for NASM or YASM in the PATH.
+if(NOT DEFINED CMAKE_ASM_NASM_COMPILER AND DEFINED ENV{ASM_NASM})
+  set(CMAKE_ASM_NASM_COMPILER $ENV{ASM_NASM})
+endif()
+
+if(CPU_TYPE STREQUAL "x86_64")
+  if(CYGWIN)
+    set(CMAKE_ASM_NASM_OBJECT_FORMAT win64)
+  endif()
+  if(CMAKE_C_COMPILER_ABI MATCHES "ELF X32")
+    set(CMAKE_ASM_NASM_OBJECT_FORMAT elfx32)
+  endif()
+elseif(CPU_TYPE STREQUAL "i386")
+  if(BORLAND)
+    set(CMAKE_ASM_NASM_OBJECT_FORMAT obj)
+  elseif(CYGWIN)
+    set(CMAKE_ASM_NASM_OBJECT_FORMAT win32)
+  endif()
+endif()
+
+if(NOT REQUIRE_SIMD)
+  include(CheckLanguage)
+  check_language(ASM_NASM)
+  if(NOT CMAKE_ASM_NASM_COMPILER)
+    simd_fail("SIMD extensions disabled: could not find NASM compiler")
+    return()
+  endif()
+endif()
+enable_language(ASM_NASM)
+message(STATUS "CMAKE_ASM_NASM_COMPILER = ${CMAKE_ASM_NASM_COMPILER}")
+
+if(CMAKE_ASM_NASM_OBJECT_FORMAT MATCHES "^macho")
+  set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -DMACHO")
+elseif(CMAKE_ASM_NASM_OBJECT_FORMAT MATCHES "^elf")
+  set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -DELF")
+  set(CMAKE_ASM_NASM_DEBUG_FORMAT "dwarf2")
+endif()
+if(CPU_TYPE STREQUAL "x86_64")
+  if(WIN32 OR CYGWIN)
+    set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -DWIN64")
+  endif()
+  set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -D__x86_64__")
+elseif(CPU_TYPE STREQUAL "i386")
+  if(BORLAND)
+    set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -DOBJ32")
+  elseif(WIN32 OR CYGWIN)
+    set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -DWIN32")
+  endif()
+endif()
+
+message(STATUS "CMAKE_ASM_NASM_OBJECT_FORMAT = ${CMAKE_ASM_NASM_OBJECT_FORMAT}")
+
+if(NOT CMAKE_ASM_NASM_OBJECT_FORMAT)
+  simd_fail("SIMD extensions disabled: could not determine NASM object format")
+  return()
+endif()
+
+get_filename_component(CMAKE_ASM_NASM_COMPILER_TYPE
+  "${CMAKE_ASM_NASM_COMPILER}" NAME_WE)
+if(CMAKE_ASM_NASM_COMPILER_TYPE MATCHES "yasm")
+  foreach(var CMAKE_ASM_NASM_FLAGS_DEBUG CMAKE_ASM_NASM_FLAGS_RELWITHDEBINFO)
+    if(${var} STREQUAL "-g")
+      if(CMAKE_ASM_NASM_DEBUG_FORMAT)
+        set_property(CACHE ${var} PROPERTY VALUE "-g ${CMAKE_ASM_NASM_DEBUG_FORMAT}")
+      else()
+        set_property(CACHE ${var} PROPERTY VALUE "")
+      endif()
+    endif()
+  endforeach()
+endif()
+
+if(NOT WIN32 AND (CMAKE_POSITION_INDEPENDENT_CODE OR ENABLE_SHARED))
+  set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -DPIC")
+endif()
+
+string(TOUPPER ${CMAKE_BUILD_TYPE} CMAKE_BUILD_TYPE_UC)
+set(EFFECTIVE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} ${CMAKE_ASM_NASM_FLAGS_${CMAKE_BUILD_TYPE_UC}}")
+message(STATUS "CMAKE_ASM_NASM_FLAGS = ${EFFECTIVE_ASM_NASM_FLAGS}")
+
+set(CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS} -I\"${CMAKE_CURRENT_SOURCE_DIR}/nasm/\" -I\"${CMAKE_CURRENT_SOURCE_DIR}/${CPU_TYPE}/\"")
+
+set(GREP grep)
+if(CMAKE_SYSTEM_NAME STREQUAL "SunOS")
+  set(GREP ggrep)
+endif()
+add_custom_target(jsimdcfg COMMAND
+  ${CMAKE_C_COMPILER} -E -I${CMAKE_BINARY_DIR} -I${CMAKE_CURRENT_BINARY_DIR}
+    -I${CMAKE_CURRENT_SOURCE_DIR}
+    ${CMAKE_CURRENT_SOURCE_DIR}/nasm/jsimdcfg.inc.h |
+  ${GREP} -E '^\;%|^\ %' | sed 's%_cpp_protection_%%' |
+  sed 's@% define@%define@g' >${CMAKE_CURRENT_SOURCE_DIR}/nasm/jsimdcfg.inc)
+
+if(CPU_TYPE STREQUAL "x86_64")
+  set(SIMD_SOURCES x86_64/jsimdcpu.asm x86_64/jfdctflt-sse.asm
+    x86_64/jccolor-sse2.asm x86_64/jcgray-sse2.asm x86_64/jchuff-sse2.asm
+    x86_64/jcphuff-sse2.asm x86_64/jcsample-sse2.asm x86_64/jdcolor-sse2.asm
+    x86_64/jdmerge-sse2.asm x86_64/jdsample-sse2.asm x86_64/jfdctfst-sse2.asm
+    x86_64/jfdctint-sse2.asm x86_64/jidctflt-sse2.asm x86_64/jidctfst-sse2.asm
+    x86_64/jidctint-sse2.asm x86_64/jidctred-sse2.asm x86_64/jquantf-sse2.asm
+    x86_64/jquanti-sse2.asm
+    x86_64/jccolor-avx2.asm x86_64/jcgray-avx2.asm x86_64/jcsample-avx2.asm
+    x86_64/jdcolor-avx2.asm x86_64/jdmerge-avx2.asm x86_64/jdsample-avx2.asm
+    x86_64/jfdctint-avx2.asm x86_64/jidctint-avx2.asm x86_64/jquanti-avx2.asm)
+else()
+  set(SIMD_SOURCES i386/jsimdcpu.asm i386/jfdctflt-3dn.asm
+    i386/jidctflt-3dn.asm i386/jquant-3dn.asm
+    i386/jccolor-mmx.asm i386/jcgray-mmx.asm i386/jcsample-mmx.asm
+    i386/jdcolor-mmx.asm i386/jdmerge-mmx.asm i386/jdsample-mmx.asm
+    i386/jfdctfst-mmx.asm i386/jfdctint-mmx.asm i386/jidctfst-mmx.asm
+    i386/jidctint-mmx.asm i386/jidctred-mmx.asm i386/jquant-mmx.asm
+    i386/jfdctflt-sse.asm i386/jidctflt-sse.asm i386/jquant-sse.asm
+    i386/jccolor-sse2.asm i386/jcgray-sse2.asm i386/jchuff-sse2.asm
+    i386/jcphuff-sse2.asm i386/jcsample-sse2.asm i386/jdcolor-sse2.asm
+    i386/jdmerge-sse2.asm i386/jdsample-sse2.asm i386/jfdctfst-sse2.asm
+    i386/jfdctint-sse2.asm i386/jidctflt-sse2.asm i386/jidctfst-sse2.asm
+    i386/jidctint-sse2.asm i386/jidctred-sse2.asm i386/jquantf-sse2.asm
+    i386/jquanti-sse2.asm
+    i386/jccolor-avx2.asm i386/jcgray-avx2.asm i386/jcsample-avx2.asm
+    i386/jdcolor-avx2.asm i386/jdmerge-avx2.asm i386/jdsample-avx2.asm
+    i386/jfdctint-avx2.asm i386/jidctint-avx2.asm i386/jquanti-avx2.asm)
+endif()
+
+if(MSVC_IDE)
+  set(OBJDIR "${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_CFG_INTDIR}")
+  string(REGEX REPLACE " " ";" CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS}")
+elseif(XCODE)
+  set(OBJDIR "${CMAKE_CURRENT_BINARY_DIR}")
+  string(REGEX REPLACE " " ";" CMAKE_ASM_NASM_FLAGS "${CMAKE_ASM_NASM_FLAGS}")
+endif()
+
+file(GLOB INC_FILES nasm/*.inc)
+
+foreach(file ${SIMD_SOURCES})
+  set(OBJECT_DEPENDS "")
+  if(${file} MATCHES jccolor)
+    string(REGEX REPLACE "jccolor" "jccolext" DEPFILE ${file})
+    set(OBJECT_DEPENDS ${OBJECT_DEPENDS}
+      ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE})
+  endif()
+  if(${file} MATCHES jcgray)
+    string(REGEX REPLACE "jcgray" "jcgryext" DEPFILE ${file})
+    set(OBJECT_DEPENDS ${OBJECT_DEPENDS}
+      ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE})
+  endif()
+  if(${file} MATCHES jdcolor)
+    string(REGEX REPLACE "jdcolor" "jdcolext" DEPFILE ${file})
+    set(OBJECT_DEPENDS ${OBJECT_DEPENDS}
+      ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE})
+  endif()
+  if(${file} MATCHES jdmerge)
+    string(REGEX REPLACE "jdmerge" "jdmrgext" DEPFILE ${file})
+    set(OBJECT_DEPENDS ${OBJECT_DEPENDS}
+      ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE})
+  endif()
+  set(OBJECT_DEPENDS ${OBJECT_DEPENDS} ${INC_FILES})
+  if(MSVC_IDE OR XCODE)
+    # The CMake Visual Studio generators do not work properly with the ASM_NASM
+    # language, so we have to go rogue here and use a custom command like we
+    # did in prior versions of libjpeg-turbo.  (This is why we can't have nice
+    # things.)
+    string(REGEX REPLACE "${CPU_TYPE}/" "" filename ${file})
+    set(SIMD_OBJ ${OBJDIR}/${filename}${CMAKE_C_OUTPUT_EXTENSION})
+    add_custom_command(OUTPUT ${SIMD_OBJ} DEPENDS ${file} ${OBJECT_DEPENDS}
+      COMMAND ${CMAKE_ASM_NASM_COMPILER} -f${CMAKE_ASM_NASM_OBJECT_FORMAT}
+        ${CMAKE_ASM_NASM_FLAGS} ${CMAKE_CURRENT_SOURCE_DIR}/${file}
+        -o${SIMD_OBJ})
+    set(SIMD_OBJS ${SIMD_OBJS} ${SIMD_OBJ})
+  else()
+    set_source_files_properties(${file} PROPERTIES OBJECT_DEPENDS
+      "${OBJECT_DEPENDS}")
+  endif()
+endforeach()
+
+if(MSVC_IDE OR XCODE)
+  set(SIMD_OBJS ${SIMD_OBJS} PARENT_SCOPE)
+  add_library(simd OBJECT ${CPU_TYPE}/jsimd.c)
+  add_custom_target(simd-objs DEPENDS ${SIMD_OBJS})
+  add_dependencies(simd simd-objs)
+else()
+  add_library(simd OBJECT ${SIMD_SOURCES} ${CPU_TYPE}/jsimd.c)
+endif()
+if(NOT WIN32 AND (CMAKE_POSITION_INDEPENDENT_CODE OR ENABLE_SHARED))
+  set_target_properties(simd PROPERTIES POSITION_INDEPENDENT_CODE 1)
+endif()
+
+
+###############################################################################
+# Arm (Intrinsics or GAS)
+###############################################################################
+
+elseif(CPU_TYPE STREQUAL "arm64" OR CPU_TYPE STREQUAL "arm")
+
+# If Neon instructions are not explicitly enabled at compile time (e.g. using
+# -mfpu=neon) with an AArch32 Linux or Android build, then the AArch32 SIMD
+# dispatcher will parse /proc/cpuinfo to determine whether the Neon SIMD
+# extensions can be enabled at run time.  In order to support all AArch32 CPUs
+# using the same code base, i.e. to support run-time FPU and Neon
+# auto-detection, it is necessary to compile the scalar C source code using
+# -mfloat-abi=soft (which is usually the default) but compile the intrinsics
+# implementation of the Neon SIMD extensions using -mfloat-abi=softfp.  The
+# following test determines whether -mfloat-abi=softfp should be explicitly
+# added to the compile flags for the intrinsics implementation of the Neon SIMD
+# extensions.
+if(BITS EQUAL 32)
+  check_c_source_compiles("
+    #if defined(__ARM_NEON__) || (!defined(__linux__) && !defined(ANDROID) && !defined(__ANDROID__))
+    #error \"Neon run-time auto-detection will not be used\"
+    #endif
+    #if __ARM_PCS_VFP == 1
+    #error \"float ABI = hard\"
+    #endif
+    #if __SOFTFP__ != 1
+    #error \"float ABI = softfp\"
+    #endif
+    int main(void) { return 0; }" NEED_SOFTFP_FOR_INTRINSICS)
+  if(NEED_SOFTFP_FOR_INTRINSICS)
+    set(SOFTFP_FLAG -mfloat-abi=softfp)
+  endif()
+endif()
+
+include(CheckSymbolExists)
+if(BITS EQUAL 32)
+  set(CMAKE_REQUIRED_FLAGS "-mfpu=neon ${SOFTFP_FLAG}")
+endif()
+check_symbol_exists(vld1_s16_x3 arm_neon.h HAVE_VLD1_S16_X3)
+check_symbol_exists(vld1_u16_x2 arm_neon.h HAVE_VLD1_U16_X2)
+check_symbol_exists(vld1q_u8_x4 arm_neon.h HAVE_VLD1Q_U8_X4)
+if(BITS EQUAL 32)
+  unset(CMAKE_REQUIRED_FLAGS)
+endif()
+configure_file(arm/neon-compat.h.in arm/neon-compat.h @ONLY)
+include_directories(${CMAKE_CURRENT_BINARY_DIR}/arm)
+
+# GCC (as of this writing) and some older versions of Clang do not have a full
+# or optimal set of Neon intrinsics, so for performance reasons, when using
+# those compilers, we default to using the older GAS implementation of the Neon
+# SIMD extensions for certain algorithms.  The presence or absence of the three
+# intrinsics we tested above is a reasonable proxy for this.  We always default
+# to using the full Neon intrinsics implementation when building for macOS or
+# iOS, to avoid the need for gas-preprocessor.
+if((HAVE_VLD1_S16_X3 AND HAVE_VLD1_U16_X2 AND HAVE_VLD1Q_U8_X4) OR APPLE)
+  set(DEFAULT_NEON_INTRINSICS 1)
+else()
+  set(DEFAULT_NEON_INTRINSICS 0)
+endif()
+option(NEON_INTRINSICS
+  "Because GCC (as of this writing) and some older versions of Clang do not have a full or optimal set of Neon intrinsics, for performance reasons, the default when building libjpeg-turbo with those compilers is to continue using the older GAS implementation of the Neon SIMD extensions for certain algorithms.  Setting this option forces the full Neon intrinsics implementation to be used with all compilers.  Unsetting this option forces the hybrid GAS/intrinsics implementation to be used with all compilers."
+  ${DEFAULT_NEON_INTRINSICS})
+boolean_number(NEON_INTRINSICS PARENT_SCOPE)
+if(NEON_INTRINSICS)
+  add_definitions(-DNEON_INTRINSICS)
+  message(STATUS "Use full Neon SIMD intrinsics implementation (NEON_INTRINSICS = ${NEON_INTRINSICS})")
+else()
+  message(STATUS "Use partial Neon SIMD intrinsics implementation (NEON_INTRINSICS = ${NEON_INTRINSICS})")
+endif()
+
+set(SIMD_SOURCES arm/jcgray-neon.c arm/jcphuff-neon.c arm/jcsample-neon.c
+  arm/jdmerge-neon.c arm/jdsample-neon.c arm/jfdctfst-neon.c
+  arm/jidctred-neon.c arm/jquanti-neon.c)
+if(NEON_INTRINSICS)
+  set(SIMD_SOURCES ${SIMD_SOURCES} arm/jccolor-neon.c arm/jidctint-neon.c)
+endif()
+if(NEON_INTRINSICS OR BITS EQUAL 64)
+  set(SIMD_SOURCES ${SIMD_SOURCES} arm/jidctfst-neon.c)
+endif()
+if(NEON_INTRINSICS OR BITS EQUAL 32)
+  set(SIMD_SOURCES ${SIMD_SOURCES} arm/aarch${BITS}/jchuff-neon.c
+    arm/jdcolor-neon.c arm/jfdctint-neon.c)
+endif()
+if(BITS EQUAL 32)
+  set_source_files_properties(${SIMD_SOURCES} COMPILE_FLAGS "-mfpu=neon ${SOFTFP_FLAG}")
+endif()
+if(NOT NEON_INTRINSICS)
+  enable_language(ASM)
+
+  set(CMAKE_ASM_FLAGS "${CMAKE_C_FLAGS} ${CMAKE_ASM_FLAGS}")
+
+  string(TOUPPER ${CMAKE_BUILD_TYPE} CMAKE_BUILD_TYPE_UC)
+  set(EFFECTIVE_ASM_FLAGS "${CMAKE_ASM_FLAGS} ${CMAKE_ASM_FLAGS_${CMAKE_BUILD_TYPE_UC}}")
+  message(STATUS "CMAKE_ASM_FLAGS = ${EFFECTIVE_ASM_FLAGS}")
+
+  set(SIMD_SOURCES ${SIMD_SOURCES} arm/aarch${BITS}/jsimd_neon.S)
+endif()
+
+add_library(simd OBJECT ${SIMD_SOURCES} arm/aarch${BITS}/jsimd.c)
+
+if(CMAKE_POSITION_INDEPENDENT_CODE OR ENABLE_SHARED)
+  set_target_properties(simd PROPERTIES POSITION_INDEPENDENT_CODE 1)
+endif()
+
+
+###############################################################################
+# MIPS (GAS)
+###############################################################################
+
+elseif(CPU_TYPE STREQUAL "mips" OR CPU_TYPE STREQUAL "mipsel")
+
+enable_language(ASM)
+
+string(TOUPPER ${CMAKE_BUILD_TYPE} CMAKE_BUILD_TYPE_UC)
+set(EFFECTIVE_ASM_FLAGS "${CMAKE_ASM_FLAGS} ${CMAKE_ASM_FLAGS_${CMAKE_BUILD_TYPE_UC}}")
+message(STATUS "CMAKE_ASM_FLAGS = ${EFFECTIVE_ASM_FLAGS}")
+
+set(CMAKE_REQUIRED_FLAGS -mdspr2)
+
+check_c_source_compiles("
+  #if !(defined(__mips__) && __mips_isa_rev >= 2)
+  #error MIPS DSPr2 is currently only available on MIPS32r2 platforms.
+  #endif
+  int main(void) {
+    int c = 0, a = 0, b = 0;
+    __asm__ __volatile__ (
+      \"precr.qb.ph %c, %a, %b\"
+      : c \"=r\" (c)
+      : a \"r\" (a), b \"r\" (b)
+    );
+    return c;
+  }" HAVE_DSPR2)
+
+unset(CMAKE_REQUIRED_FLAGS)
+
+if(NOT HAVE_DSPR2)
+  simd_fail("SIMD extensions not available for this CPU")
+  return()
+endif()
+
+add_library(simd OBJECT mips/jsimd_dspr2.S mips/jsimd.c)
+
+if(CMAKE_POSITION_INDEPENDENT_CODE OR ENABLE_SHARED)
+  set_target_properties(simd PROPERTIES POSITION_INDEPENDENT_CODE 1)
+endif()
+
+###############################################################################
+# MIPS64 (Intrinsics)
+###############################################################################
+
+elseif(CPU_TYPE STREQUAL "loongson" OR CPU_TYPE MATCHES "^mips64")
+
+set(CMAKE_REQUIRED_FLAGS -Wa,-mloongson-mmi,-mloongson-ext)
+
+check_c_source_compiles("
+  int main(void) {
+    int c = 0, a = 0, b = 0;
+    asm (
+      \"paddb %0, %1, %2\"
+      : \"=f\" (c)
+      : \"f\" (a), \"f\" (b)
+    );
+    return c;
+  }" HAVE_MMI)
+
+unset(CMAKE_REQUIRED_FLAGS)
+
+if(NOT HAVE_MMI)
+  simd_fail("SIMD extensions not available for this CPU")
+  return()
+endif()
+
+set(SIMD_SOURCES mips64/jccolor-mmi.c mips64/jcgray-mmi.c mips64/jcsample-mmi.c
+  mips64/jdcolor-mmi.c mips64/jdmerge-mmi.c mips64/jdsample-mmi.c
+  mips64/jfdctfst-mmi.c mips64/jfdctint-mmi.c mips64/jidctfst-mmi.c
+  mips64/jidctint-mmi.c mips64/jquanti-mmi.c)
+
+if(CMAKE_COMPILER_IS_GNUCC)
+  foreach(file ${SIMD_SOURCES})
+    set_property(SOURCE ${file} APPEND_STRING PROPERTY COMPILE_FLAGS
+      " -fno-strict-aliasing")
+  endforeach()
+endif()
+foreach(file ${SIMD_SOURCES})
+  set_property(SOURCE ${file} APPEND_STRING PROPERTY COMPILE_FLAGS
+    " -Wa,-mloongson-mmi,-mloongson-ext")
+endforeach()
+
+add_library(simd OBJECT ${SIMD_SOURCES} mips64/jsimd.c)
+
+if(CMAKE_POSITION_INDEPENDENT_CODE OR ENABLE_SHARED)
+  set_target_properties(simd PROPERTIES POSITION_INDEPENDENT_CODE 1)
+endif()
+
+###############################################################################
+# PowerPC (Intrinsics)
+###############################################################################
+
+elseif(CPU_TYPE STREQUAL "powerpc")
+
+set(CMAKE_REQUIRED_FLAGS -maltivec)
+
+check_c_source_compiles("
+  #include <altivec.h>
+  int main(void) {
+    __vector int vi = { 0, 0, 0, 0 };
+    int i4;
+    vec_st(vi, 0, i);
+    return i0;
+  }" HAVE_ALTIVEC)
+
+unset(CMAKE_REQUIRED_FLAGS)
+
+if(NOT HAVE_ALTIVEC)
+  simd_fail("SIMD extensions not available for this CPU (PowerPC SPE)")
+  return()
+endif()
+
+set(SIMD_SOURCES powerpc/jccolor-altivec.c powerpc/jcgray-altivec.c
+  powerpc/jcsample-altivec.c powerpc/jdcolor-altivec.c
+  powerpc/jdmerge-altivec.c powerpc/jdsample-altivec.c
+  powerpc/jfdctfst-altivec.c powerpc/jfdctint-altivec.c
+  powerpc/jidctfst-altivec.c powerpc/jidctint-altivec.c
+  powerpc/jquanti-altivec.c)
+
+set_source_files_properties(${SIMD_SOURCES} PROPERTIES
+  COMPILE_FLAGS -maltivec)
+
+add_library(simd OBJECT ${SIMD_SOURCES} powerpc/jsimd.c)
+
+if(CMAKE_POSITION_INDEPENDENT_CODE OR ENABLE_SHARED)
+  set_target_properties(simd PROPERTIES POSITION_INDEPENDENT_CODE 1)
+endif()
+
+
+###############################################################################
+# None
+###############################################################################
+
+else()
+
+simd_fail("SIMD extensions not available for this CPU (${CMAKE_SYSTEM_PROCESSOR})")
+
+endif() # CPU_TYPE

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch32 Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch32/jccolext-neon.c Added

@@ -0,0 +1,148 @@
+/*
+ * jccolext-neon.c - colorspace conversion (32-bit Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jccolor-neon.c */
+
+
+/* RGB -> YCbCr conversion is defined by the following equations:
+ *    Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+ *    Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + 128
+ *    Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B  + 128
+ *
+ * Avoid floating point arithmetic by using shifted integer constants:
+ *    0.29899597 = 19595 * 2^-16
+ *    0.58700561 = 38470 * 2^-16
+ *    0.11399841 =  7471 * 2^-16
+ *    0.16874695 = 11059 * 2^-16
+ *    0.33125305 = 21709 * 2^-16
+ *    0.50000000 = 32768 * 2^-16
+ *    0.41868592 = 27439 * 2^-16
+ *    0.08131409 =  5329 * 2^-16
+ * These constants are defined in jccolor-neon.c
+ *
+ * We add the fixed-point equivalent of 0.5 to Cb and Cr, which effectively
+ * rounds up or down the result via integer truncation.
+ */
+
+void jsimd_rgb_ycc_convert_neon(JDIMENSION image_width, JSAMPARRAY input_buf,
+                                JSAMPIMAGE output_buf, JDIMENSION output_row,
+                                int num_rows)
+{
+  /* Pointer to RGB(X/A) input data */
+  JSAMPROW inptr;
+  /* Pointers to Y, Cb, and Cr output data */
+  JSAMPROW outptr0, outptr1, outptr2;
+  /* Allocate temporary buffer for final (image_width % 8) pixels in row. */
+  ALIGN(16) uint8_t tmp_buf8 * RGB_PIXELSIZE;
+
+  /* Set up conversion constants. */
+#ifdef HAVE_VLD1_U16_X2
+  const uint16x4x2_t consts = vld1_u16_x2(jsimd_rgb_ycc_neon_consts);
+#else
+  /* GCC does not currently support the intrinsic vld1_<type>_x2(). */
+  const uint16x4_t consts1 = vld1_u16(jsimd_rgb_ycc_neon_consts);
+  const uint16x4_t consts2 = vld1_u16(jsimd_rgb_ycc_neon_consts + 4);
+  const uint16x4x2_t consts = { { consts1, consts2 } };
+#endif
+  const uint32x4_t scaled_128_5 = vdupq_n_u32((128 << 16) + 32767);
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf0output_row;
+    outptr1 = output_buf1output_row;
+    outptr2 = output_buf2output_row;
+    output_row++;
+
+    int cols_remaining = image_width;
+    for (; cols_remaining > 0; cols_remaining -= 8) {
+
+      /* To prevent buffer overread by the vector load instructions, the last
+       * (image_width % 8) columns of data are first memcopied to a temporary
+       * buffer large enough to accommodate the vector load.
+       */
+      if (cols_remaining < 8) {
+        memcpy(tmp_buf, inptr, cols_remaining * RGB_PIXELSIZE);
+        inptr = tmp_buf;
+      }
+
+#if RGB_PIXELSIZE == 4
+      uint8x8x4_t input_pixels = vld4_u8(inptr);
+#else
+      uint8x8x3_t input_pixels = vld3_u8(inptr);
+#endif
+      uint16x8_t r = vmovl_u8(input_pixels.valRGB_RED);
+      uint16x8_t g = vmovl_u8(input_pixels.valRGB_GREEN);
+      uint16x8_t b = vmovl_u8(input_pixels.valRGB_BLUE);
+
+      /* Compute Y = 0.29900 * R + 0.58700 * G + 0.11400 * B */
+      uint32x4_t y_low = vmull_lane_u16(vget_low_u16(r), consts.val0, 0);
+      y_low = vmlal_lane_u16(y_low, vget_low_u16(g), consts.val0, 1);
+      y_low = vmlal_lane_u16(y_low, vget_low_u16(b), consts.val0, 2);
+      uint32x4_t y_high = vmull_lane_u16(vget_high_u16(r), consts.val0, 0);
+      y_high = vmlal_lane_u16(y_high, vget_high_u16(g), consts.val0, 1);
+      y_high = vmlal_lane_u16(y_high, vget_high_u16(b), consts.val0, 2);
+
+      /* Compute Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + 128 */
+      uint32x4_t cb_low = scaled_128_5;
+      cb_low = vmlsl_lane_u16(cb_low, vget_low_u16(r), consts.val0, 3);
+      cb_low = vmlsl_lane_u16(cb_low, vget_low_u16(g), consts.val1, 0);
+      cb_low = vmlal_lane_u16(cb_low, vget_low_u16(b), consts.val1, 1);
+      uint32x4_t cb_high = scaled_128_5;
+      cb_high = vmlsl_lane_u16(cb_high, vget_high_u16(r), consts.val0, 3);
+      cb_high = vmlsl_lane_u16(cb_high, vget_high_u16(g), consts.val1, 0);
+      cb_high = vmlal_lane_u16(cb_high, vget_high_u16(b), consts.val1, 1);
+
+      /* Compute Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B  + 128 */
+      uint32x4_t cr_low = scaled_128_5;
+      cr_low = vmlal_lane_u16(cr_low, vget_low_u16(r), consts.val1, 1);
+      cr_low = vmlsl_lane_u16(cr_low, vget_low_u16(g), consts.val1, 2);
+      cr_low = vmlsl_lane_u16(cr_low, vget_low_u16(b), consts.val1, 3);
+      uint32x4_t cr_high = scaled_128_5;
+      cr_high = vmlal_lane_u16(cr_high, vget_high_u16(r), consts.val1, 1);
+      cr_high = vmlsl_lane_u16(cr_high, vget_high_u16(g), consts.val1, 2);
+      cr_high = vmlsl_lane_u16(cr_high, vget_high_u16(b), consts.val1, 3);
+
+      /* Descale Y values (rounding right shift) and narrow to 16-bit. */
+      uint16x8_t y_u16 = vcombine_u16(vrshrn_n_u32(y_low, 16),
+                                      vrshrn_n_u32(y_high, 16));
+      /* Descale Cb values (right shift) and narrow to 16-bit. */
+      uint16x8_t cb_u16 = vcombine_u16(vshrn_n_u32(cb_low, 16),
+                                       vshrn_n_u32(cb_high, 16));
+      /* Descale Cr values (right shift) and narrow to 16-bit. */
+      uint16x8_t cr_u16 = vcombine_u16(vshrn_n_u32(cr_low, 16),
+                                       vshrn_n_u32(cr_high, 16));
+      /* Narrow Y, Cb, and Cr values to 8-bit and store to memory.  Buffer
+       * overwrite is permitted up to the next multiple of ALIGN_SIZE bytes.
+       */
+      vst1_u8(outptr0, vmovn_u16(y_u16));
+      vst1_u8(outptr1, vmovn_u16(cb_u16));
+      vst1_u8(outptr2, vmovn_u16(cr_u16));
+
+      /* Increment pointers. */
+      inptr += (8 * RGB_PIXELSIZE);
+      outptr0 += 8;
+      outptr1 += 8;
+      outptr2 += 8;
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch32/jchuff-neon.c Added

@@ -0,0 +1,334 @@
+/*
+ * jchuff-neon.c - Huffman entropy encoding (32-bit Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ *
+ * NOTE: All referenced figures are from
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+ */
+
+#define JPEG_INTERNALS
+#include "../../../jinclude.h"
+#include "../../../jpeglib.h"
+#include "../../../jsimd.h"
+#include "../../../jdct.h"
+#include "../../../jsimddct.h"
+#include "../../jsimd.h"
+#include "../jchuff.h"
+#include "neon-compat.h"
+
+#include <limits.h>
+
+#include <arm_neon.h>
+
+
+JOCTET *jsimd_huff_encode_one_block_neon(void *state, JOCTET *buffer,
+                                         JCOEFPTR block, int last_dc_val,
+                                         c_derived_tbl *dctbl,
+                                         c_derived_tbl *actbl)
+{
+  uint8_t block_nbitsDCTSIZE2;
+  uint16_t block_diffDCTSIZE2;
+
+  /* Load rows of coefficients from DCT block in zig-zag order. */
+
+  /* Compute DC coefficient difference value. (F.1.1.5.1) */
+  int16x8_t row0 = vdupq_n_s16(block0 - last_dc_val);
+  row0 = vld1q_lane_s16(block +  1, row0, 1);
+  row0 = vld1q_lane_s16(block +  8, row0, 2);
+  row0 = vld1q_lane_s16(block + 16, row0, 3);
+  row0 = vld1q_lane_s16(block +  9, row0, 4);
+  row0 = vld1q_lane_s16(block +  2, row0, 5);
+  row0 = vld1q_lane_s16(block +  3, row0, 6);
+  row0 = vld1q_lane_s16(block + 10, row0, 7);
+
+  int16x8_t row1 = vld1q_dup_s16(block + 17);
+  row1 = vld1q_lane_s16(block + 24, row1, 1);
+  row1 = vld1q_lane_s16(block + 32, row1, 2);
+  row1 = vld1q_lane_s16(block + 25, row1, 3);
+  row1 = vld1q_lane_s16(block + 18, row1, 4);
+  row1 = vld1q_lane_s16(block + 11, row1, 5);
+  row1 = vld1q_lane_s16(block +  4, row1, 6);
+  row1 = vld1q_lane_s16(block +  5, row1, 7);
+
+  int16x8_t row2 = vld1q_dup_s16(block + 12);
+  row2 = vld1q_lane_s16(block + 19, row2, 1);
+  row2 = vld1q_lane_s16(block + 26, row2, 2);
+  row2 = vld1q_lane_s16(block + 33, row2, 3);
+  row2 = vld1q_lane_s16(block + 40, row2, 4);
+  row2 = vld1q_lane_s16(block + 48, row2, 5);
+  row2 = vld1q_lane_s16(block + 41, row2, 6);
+  row2 = vld1q_lane_s16(block + 34, row2, 7);
+
+  int16x8_t row3 = vld1q_dup_s16(block + 27);
+  row3 = vld1q_lane_s16(block + 20, row3, 1);
+  row3 = vld1q_lane_s16(block + 13, row3, 2);
+  row3 = vld1q_lane_s16(block +  6, row3, 3);
+  row3 = vld1q_lane_s16(block +  7, row3, 4);
+  row3 = vld1q_lane_s16(block + 14, row3, 5);
+  row3 = vld1q_lane_s16(block + 21, row3, 6);
+  row3 = vld1q_lane_s16(block + 28, row3, 7);
+
+  int16x8_t abs_row0 = vabsq_s16(row0);
+  int16x8_t abs_row1 = vabsq_s16(row1);
+  int16x8_t abs_row2 = vabsq_s16(row2);
+  int16x8_t abs_row3 = vabsq_s16(row3);
+
+  int16x8_t row0_lz = vclzq_s16(abs_row0);
+  int16x8_t row1_lz = vclzq_s16(abs_row1);
+  int16x8_t row2_lz = vclzq_s16(abs_row2);
+  int16x8_t row3_lz = vclzq_s16(abs_row3);
+
+  /* Compute number of bits required to represent each coefficient. */
+  uint8x8_t row0_nbits = vsub_u8(vdup_n_u8(16),
+                                 vmovn_u16(vreinterpretq_u16_s16(row0_lz)));
+  uint8x8_t row1_nbits = vsub_u8(vdup_n_u8(16),
+                                 vmovn_u16(vreinterpretq_u16_s16(row1_lz)));
+  uint8x8_t row2_nbits = vsub_u8(vdup_n_u8(16),
+                                 vmovn_u16(vreinterpretq_u16_s16(row2_lz)));
+  uint8x8_t row3_nbits = vsub_u8(vdup_n_u8(16),
+                                 vmovn_u16(vreinterpretq_u16_s16(row3_lz)));
+
+  vst1_u8(block_nbits + 0 * DCTSIZE, row0_nbits);
+  vst1_u8(block_nbits + 1 * DCTSIZE, row1_nbits);
+  vst1_u8(block_nbits + 2 * DCTSIZE, row2_nbits);
+  vst1_u8(block_nbits + 3 * DCTSIZE, row3_nbits);
+
+  uint16x8_t row0_mask =
+    vshlq_u16(vreinterpretq_u16_s16(vshrq_n_s16(row0, 15)),
+              vnegq_s16(row0_lz));
+  uint16x8_t row1_mask =
+    vshlq_u16(vreinterpretq_u16_s16(vshrq_n_s16(row1, 15)),
+              vnegq_s16(row1_lz));
+  uint16x8_t row2_mask =
+    vshlq_u16(vreinterpretq_u16_s16(vshrq_n_s16(row2, 15)),
+              vnegq_s16(row2_lz));
+  uint16x8_t row3_mask =
+    vshlq_u16(vreinterpretq_u16_s16(vshrq_n_s16(row3, 15)),
+              vnegq_s16(row3_lz));
+
+  uint16x8_t row0_diff = veorq_u16(vreinterpretq_u16_s16(abs_row0), row0_mask);
+  uint16x8_t row1_diff = veorq_u16(vreinterpretq_u16_s16(abs_row1), row1_mask);
+  uint16x8_t row2_diff = veorq_u16(vreinterpretq_u16_s16(abs_row2), row2_mask);
+  uint16x8_t row3_diff = veorq_u16(vreinterpretq_u16_s16(abs_row3), row3_mask);
+
+  /* Store diff values for rows 0, 1, 2, and 3. */
+  vst1q_u16(block_diff + 0 * DCTSIZE, row0_diff);
+  vst1q_u16(block_diff + 1 * DCTSIZE, row1_diff);
+  vst1q_u16(block_diff + 2 * DCTSIZE, row2_diff);
+  vst1q_u16(block_diff + 3 * DCTSIZE, row3_diff);
+
+  /* Load last four rows of coefficients from DCT block in zig-zag order. */
+  int16x8_t row4 = vld1q_dup_s16(block + 35);
+  row4 = vld1q_lane_s16(block + 42, row4, 1);
+  row4 = vld1q_lane_s16(block + 49, row4, 2);
+  row4 = vld1q_lane_s16(block + 56, row4, 3);
+  row4 = vld1q_lane_s16(block + 57, row4, 4);
+  row4 = vld1q_lane_s16(block + 50, row4, 5);
+  row4 = vld1q_lane_s16(block + 43, row4, 6);
+  row4 = vld1q_lane_s16(block + 36, row4, 7);
+
+  int16x8_t row5 = vld1q_dup_s16(block + 29);
+  row5 = vld1q_lane_s16(block + 22, row5, 1);
+  row5 = vld1q_lane_s16(block + 15, row5, 2);
+  row5 = vld1q_lane_s16(block + 23, row5, 3);
+  row5 = vld1q_lane_s16(block + 30, row5, 4);
+  row5 = vld1q_lane_s16(block + 37, row5, 5);
+  row5 = vld1q_lane_s16(block + 44, row5, 6);
+  row5 = vld1q_lane_s16(block + 51, row5, 7);
+
+  int16x8_t row6 = vld1q_dup_s16(block + 58);
+  row6 = vld1q_lane_s16(block + 59, row6, 1);
+  row6 = vld1q_lane_s16(block + 52, row6, 2);
+  row6 = vld1q_lane_s16(block + 45, row6, 3);
+  row6 = vld1q_lane_s16(block + 38, row6, 4);
+  row6 = vld1q_lane_s16(block + 31, row6, 5);
+  row6 = vld1q_lane_s16(block + 39, row6, 6);
+  row6 = vld1q_lane_s16(block + 46, row6, 7);
+
+  int16x8_t row7 = vld1q_dup_s16(block + 53);
+  row7 = vld1q_lane_s16(block + 60, row7, 1);
+  row7 = vld1q_lane_s16(block + 61, row7, 2);
+  row7 = vld1q_lane_s16(block + 54, row7, 3);
+  row7 = vld1q_lane_s16(block + 47, row7, 4);
+  row7 = vld1q_lane_s16(block + 55, row7, 5);
+  row7 = vld1q_lane_s16(block + 62, row7, 6);
+  row7 = vld1q_lane_s16(block + 63, row7, 7);
+
+  int16x8_t abs_row4 = vabsq_s16(row4);
+  int16x8_t abs_row5 = vabsq_s16(row5);
+  int16x8_t abs_row6 = vabsq_s16(row6);
+  int16x8_t abs_row7 = vabsq_s16(row7);
+
+  int16x8_t row4_lz = vclzq_s16(abs_row4);
+  int16x8_t row5_lz = vclzq_s16(abs_row5);
+  int16x8_t row6_lz = vclzq_s16(abs_row6);
+  int16x8_t row7_lz = vclzq_s16(abs_row7);
+
+  /* Compute number of bits required to represent each coefficient. */
+  uint8x8_t row4_nbits = vsub_u8(vdup_n_u8(16),
+                                 vmovn_u16(vreinterpretq_u16_s16(row4_lz)));
+  uint8x8_t row5_nbits = vsub_u8(vdup_n_u8(16),
+                                 vmovn_u16(vreinterpretq_u16_s16(row5_lz)));
+  uint8x8_t row6_nbits = vsub_u8(vdup_n_u8(16),
+                                 vmovn_u16(vreinterpretq_u16_s16(row6_lz)));
+  uint8x8_t row7_nbits = vsub_u8(vdup_n_u8(16),
+                                 vmovn_u16(vreinterpretq_u16_s16(row7_lz)));
+
+  vst1_u8(block_nbits + 4 * DCTSIZE, row4_nbits);
+  vst1_u8(block_nbits + 5 * DCTSIZE, row5_nbits);
+  vst1_u8(block_nbits + 6 * DCTSIZE, row6_nbits);
+  vst1_u8(block_nbits + 7 * DCTSIZE, row7_nbits);
+
+  uint16x8_t row4_mask =
+    vshlq_u16(vreinterpretq_u16_s16(vshrq_n_s16(row4, 15)),
+              vnegq_s16(row4_lz));
+  uint16x8_t row5_mask =
+    vshlq_u16(vreinterpretq_u16_s16(vshrq_n_s16(row5, 15)),
+              vnegq_s16(row5_lz));
+  uint16x8_t row6_mask =
+    vshlq_u16(vreinterpretq_u16_s16(vshrq_n_s16(row6, 15)),
+              vnegq_s16(row6_lz));
+  uint16x8_t row7_mask =
+    vshlq_u16(vreinterpretq_u16_s16(vshrq_n_s16(row7, 15)),
+              vnegq_s16(row7_lz));
+
+  uint16x8_t row4_diff = veorq_u16(vreinterpretq_u16_s16(abs_row4), row4_mask);
+  uint16x8_t row5_diff = veorq_u16(vreinterpretq_u16_s16(abs_row5), row5_mask);
+  uint16x8_t row6_diff = veorq_u16(vreinterpretq_u16_s16(abs_row6), row6_mask);
+  uint16x8_t row7_diff = veorq_u16(vreinterpretq_u16_s16(abs_row7), row7_mask);
+
+  /* Store diff values for rows 4, 5, 6, and 7. */
+  vst1q_u16(block_diff + 4 * DCTSIZE, row4_diff);
+  vst1q_u16(block_diff + 5 * DCTSIZE, row5_diff);
+  vst1q_u16(block_diff + 6 * DCTSIZE, row6_diff);
+  vst1q_u16(block_diff + 7 * DCTSIZE, row7_diff);
+
+  /* Construct bitmap to accelerate encoding of AC coefficients.  A set bit
+   * means that the corresponding coefficient != 0.
+   */
+  uint8x8_t row0_nbits_gt0 = vcgt_u8(row0_nbits, vdup_n_u8(0));
+  uint8x8_t row1_nbits_gt0 = vcgt_u8(row1_nbits, vdup_n_u8(0));
+  uint8x8_t row2_nbits_gt0 = vcgt_u8(row2_nbits, vdup_n_u8(0));
+  uint8x8_t row3_nbits_gt0 = vcgt_u8(row3_nbits, vdup_n_u8(0));
+  uint8x8_t row4_nbits_gt0 = vcgt_u8(row4_nbits, vdup_n_u8(0));
+  uint8x8_t row5_nbits_gt0 = vcgt_u8(row5_nbits, vdup_n_u8(0));
+  uint8x8_t row6_nbits_gt0 = vcgt_u8(row6_nbits, vdup_n_u8(0));
+  uint8x8_t row7_nbits_gt0 = vcgt_u8(row7_nbits, vdup_n_u8(0));
+
+  /* { 0x80, 0x40, 0x20, 0x10, 0x08, 0x04, 0x02, 0x01 } */
+  const uint8x8_t bitmap_mask =
+    vreinterpret_u8_u64(vmov_n_u64(0x0102040810204080));
+
+  row0_nbits_gt0 = vand_u8(row0_nbits_gt0, bitmap_mask);
+  row1_nbits_gt0 = vand_u8(row1_nbits_gt0, bitmap_mask);
+  row2_nbits_gt0 = vand_u8(row2_nbits_gt0, bitmap_mask);
+  row3_nbits_gt0 = vand_u8(row3_nbits_gt0, bitmap_mask);
+  row4_nbits_gt0 = vand_u8(row4_nbits_gt0, bitmap_mask);
+  row5_nbits_gt0 = vand_u8(row5_nbits_gt0, bitmap_mask);
+  row6_nbits_gt0 = vand_u8(row6_nbits_gt0, bitmap_mask);
+  row7_nbits_gt0 = vand_u8(row7_nbits_gt0, bitmap_mask);
+
+  uint8x8_t bitmap_rows_10 = vpadd_u8(row1_nbits_gt0, row0_nbits_gt0);
+  uint8x8_t bitmap_rows_32 = vpadd_u8(row3_nbits_gt0, row2_nbits_gt0);
+  uint8x8_t bitmap_rows_54 = vpadd_u8(row5_nbits_gt0, row4_nbits_gt0);
+  uint8x8_t bitmap_rows_76 = vpadd_u8(row7_nbits_gt0, row6_nbits_gt0);
+  uint8x8_t bitmap_rows_3210 = vpadd_u8(bitmap_rows_32, bitmap_rows_10);
+  uint8x8_t bitmap_rows_7654 = vpadd_u8(bitmap_rows_76, bitmap_rows_54);
+  uint8x8_t bitmap = vpadd_u8(bitmap_rows_7654, bitmap_rows_3210);
+
+  /* Shift left to remove DC bit. */
+  bitmap = vreinterpret_u8_u64(vshl_n_u64(vreinterpret_u64_u8(bitmap), 1));
+  /* Move bitmap to 32-bit scalar registers. */
+  uint32_t bitmap_1_32 = vget_lane_u32(vreinterpret_u32_u8(bitmap), 1);
+  uint32_t bitmap_33_63 = vget_lane_u32(vreinterpret_u32_u8(bitmap), 0);
+
+  /* Set up state and bit buffer for output bitstream. */
+  working_state *state_ptr = (working_state *)state;
+  int free_bits = state_ptr->cur.free_bits;
+  size_t put_buffer = state_ptr->cur.put_buffer;
+
+  /* Encode DC coefficient. */
+
+  unsigned int nbits = block_nbits0;
+  /* Emit Huffman-coded symbol and additional diff bits. */
+  unsigned int diff = block_diff0;
+  PUT_CODE(dctbl->ehufconbits, dctbl->ehufsinbits, diff)
+
+  /* Encode AC coefficients. */
+
+  unsigned int r = 0;  /* r = run length of zeros */
+  unsigned int i = 1;  /* i = number of coefficients encoded */
+  /* Code and size information for a run length of 16 zero coefficients */
+  const unsigned int code_0xf0 = actbl->ehufco0xf0;
+  const unsigned int size_0xf0 = actbl->ehufsi0xf0;
+
+  while (bitmap_1_32 != 0) {
+    r = BUILTIN_CLZ(bitmap_1_32);
+    i += r;
+    bitmap_1_32 <<= r;
+    nbits = block_nbitsi;
+    diff = block_diffi;
+    while (r > 15) {
+      /* If run length > 15, emit special run-length-16 codes. */
+      PUT_BITS(code_0xf0, size_0xf0)
+      r -= 16;
+    }
+    /* Emit Huffman symbol for run length / number of bits. (F.1.2.2.1) */
+    unsigned int rs = (r << 4) + nbits;
+    PUT_CODE(actbl->ehufcors, actbl->ehufsirs, diff)
+    i++;
+    bitmap_1_32 <<= 1;
+  }
+
+  r = 33 - i;
+  i = 33;
+
+  while (bitmap_33_63 != 0) {
+    unsigned int leading_zeros = BUILTIN_CLZ(bitmap_33_63);
+    r += leading_zeros;
+    i += leading_zeros;
+    bitmap_33_63 <<= leading_zeros;
+    nbits = block_nbitsi;
+    diff = block_diffi;
+    while (r > 15) {
+      /* If run length > 15, emit special run-length-16 codes. */
+      PUT_BITS(code_0xf0, size_0xf0)
+      r -= 16;
+    }
+    /* Emit Huffman symbol for run length / number of bits. (F.1.2.2.1) */
+    unsigned int rs = (r << 4) + nbits;
+    PUT_CODE(actbl->ehufcors, actbl->ehufsirs, diff)
+    r = 0;
+    i++;
+    bitmap_33_63 <<= 1;
+  }
+
+  /* If the last coefficient(s) were zero, emit an end-of-block (EOB) code.
+   * The value of RS for the EOB code is 0.
+   */
+  if (i != 64) {
+    PUT_BITS(actbl->ehufco0, actbl->ehufsi0)
+  }
+
+  state_ptr->cur.put_buffer = put_buffer;
+  state_ptr->cur.free_bits = free_bits;
+
+  return buffer;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch32/jsimd.c Added

@@ -0,0 +1,983 @@
+/*
+ * jsimd_arm.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2011, Nokia Corporation and/or its subsidiary(-ies).
+ * Copyright (C) 2009-2011, 2013-2014, 2016, 2018, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ * Copyright (C) 2019, Google LLC.
+ * Copyright (C) 2020, Arm Limited.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 32-bit Arm architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../../../jinclude.h"
+#include "../../../jpeglib.h"
+#include "../../../jsimd.h"
+#include "../../../jdct.h"
+#include "../../../jsimddct.h"
+#include "../../jsimd.h"
+
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+
+static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
+
+#if !defined(__ARM_NEON__) && (defined(__linux__) || defined(ANDROID) || defined(__ANDROID__))
+
+#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT  (1024 * 1024)
+
+LOCAL(int)
+check_feature(char *buffer, char *feature)
+{
+  char *p;
+
+  if (*feature == 0)
+    return 0;
+  if (strncmp(buffer, "Features", 8) != 0)
+    return 0;
+  buffer += 8;
+  while (isspace(*buffer))
+    buffer++;
+
+  /* Check if 'feature' is present in the buffer as a separate word */
+  while ((p = strstr(buffer, feature))) {
+    if (p > buffer && !isspace(*(p - 1))) {
+      buffer++;
+      continue;
+    }
+    p += strlen(feature);
+    if (*p != 0 && !isspace(*p)) {
+      buffer++;
+      continue;
+    }
+    return 1;
+  }
+  return 0;
+}
+
+LOCAL(int)
+parse_proc_cpuinfo(int bufsize)
+{
+  char *buffer = (char *)malloc(bufsize);
+  FILE *fd;
+
+  simd_support = 0;
+
+  if (!buffer)
+    return 0;
+
+  fd = fopen("/proc/cpuinfo", "r");
+  if (fd) {
+    while (fgets(buffer, bufsize, fd)) {
+      if (!strchr(buffer, '\n') && !feof(fd)) {
+        /* "impossible" happened - insufficient size of the buffer! */
+        fclose(fd);
+        free(buffer);
+        return 0;
+      }
+      if (check_feature(buffer, "neon"))
+        simd_support |= JSIMD_NEON;
+    }
+    fclose(fd);
+  }
+  free(buffer);
+  return 1;
+}
+
+#endif
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd(void)
+{
+#ifndef NO_GETENV
+  char *env = NULL;
+#endif
+#if !defined(__ARM_NEON__) && (defined(__linux__) || defined(ANDROID) || defined(__ANDROID__))
+  int bufsize = 1024; /* an initial guess for the line buffer size limit */
+#endif
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = 0;
+
+#if defined(__ARM_NEON__)
+  simd_support |= JSIMD_NEON;
+#elif defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+  /* We still have a chance to use Neon regardless of globally used
+   * -mcpu/-mfpu options passed to gcc by performing runtime detection via
+   * /proc/cpuinfo parsing on linux/android */
+  while (!parse_proc_cpuinfo(bufsize)) {
+    bufsize *= 2;
+    if (bufsize > SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT)
+      break;
+  }
+#endif
+
+#ifndef NO_GETENV
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCENEON");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = JSIMD_NEON;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+#endif
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                      JSAMPIMAGE output_buf, JDIMENSION output_row,
+                      int num_rows)
+{
+  void (*neonfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    neonfct = jsimd_extrgb_ycc_convert_neon;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    neonfct = jsimd_extrgbx_ycc_convert_neon;
+    break;
+  case JCS_EXT_BGR:
+    neonfct = jsimd_extbgr_ycc_convert_neon;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    neonfct = jsimd_extbgrx_ycc_convert_neon;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    neonfct = jsimd_extxbgr_ycc_convert_neon;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    neonfct = jsimd_extxrgb_ycc_convert_neon;
+    break;
+  default:
+    neonfct = jsimd_extrgb_ycc_convert_neon;
+    break;
+  }
+
+  neonfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                       JSAMPIMAGE output_buf, JDIMENSION output_row,
+                       int num_rows)
+{
+  void (*neonfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    neonfct = jsimd_extrgb_gray_convert_neon;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    neonfct = jsimd_extrgbx_gray_convert_neon;
+    break;
+  case JCS_EXT_BGR:
+    neonfct = jsimd_extbgr_gray_convert_neon;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    neonfct = jsimd_extbgrx_gray_convert_neon;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    neonfct = jsimd_extxbgr_gray_convert_neon;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    neonfct = jsimd_extxrgb_gray_convert_neon;
+    break;
+  default:
+    neonfct = jsimd_extrgb_gray_convert_neon;
+    break;
+  }
+
+  neonfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION input_row, JSAMPARRAY output_buf,
+                      int num_rows)
+{
+  void (*neonfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    neonfct = jsimd_ycc_extrgb_convert_neon;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    neonfct = jsimd_ycc_extrgbx_convert_neon;
+    break;
+  case JCS_EXT_BGR:
+    neonfct = jsimd_ycc_extbgr_convert_neon;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    neonfct = jsimd_ycc_extbgrx_convert_neon;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    neonfct = jsimd_ycc_extxbgr_convert_neon;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    neonfct = jsimd_ycc_extxrgb_convert_neon;
+    break;
+  default:
+    neonfct = jsimd_ycc_extrgb_convert_neon;
+    break;
+  }
+
+  neonfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+  jsimd_ycc_rgb565_convert_neon(cinfo->output_width, input_buf, input_row,
+                                output_buf, num_rows);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_neon(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_neon(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_upsample_neon(cinfo->max_v_samp_factor, cinfo->output_width,
+                           input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_upsample_neon(cinfo->max_v_samp_factor, cinfo->output_width,
+                           input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h1v2_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_fancy_upsample_neon(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_fancy_upsample_neon(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h1v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h1v2_fancy_upsample_neon(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*neonfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      neonfct = jsimd_h2v2_extrgb_merged_upsample_neon;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      neonfct = jsimd_h2v2_extrgbx_merged_upsample_neon;
+      break;
+    case JCS_EXT_BGR:
+      neonfct = jsimd_h2v2_extbgr_merged_upsample_neon;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      neonfct = jsimd_h2v2_extbgrx_merged_upsample_neon;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      neonfct = jsimd_h2v2_extxbgr_merged_upsample_neon;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      neonfct = jsimd_h2v2_extxrgb_merged_upsample_neon;
+      break;
+    default:
+      neonfct = jsimd_h2v2_extrgb_merged_upsample_neon;
+      break;
+  }
+
+  neonfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*neonfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      neonfct = jsimd_h2v1_extrgb_merged_upsample_neon;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      neonfct = jsimd_h2v1_extrgbx_merged_upsample_neon;
+      break;
+    case JCS_EXT_BGR:
+      neonfct = jsimd_h2v1_extbgr_merged_upsample_neon;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      neonfct = jsimd_h2v1_extbgrx_merged_upsample_neon;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      neonfct = jsimd_h2v1_extxbgr_merged_upsample_neon;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      neonfct = jsimd_h2v1_extxrgb_merged_upsample_neon;
+      break;
+    default:
+      neonfct = jsimd_h2v1_extrgb_merged_upsample_neon;
+      break;
+  }
+
+  neonfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
+               DCTELEM *workspace)
+{
+  jsimd_convsamp_neon(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow(DCTELEM *data)
+{
+  jsimd_fdct_islow_neon(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast(DCTELEM *data)
+{
+  jsimd_fdct_ifast_neon(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float(FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
+{
+  jsimd_quantize_neon(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON && simd_huffman)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET *)
+jsimd_huff_encode_one_block(void *state, JOCTET *buffer, JCOEFPTR block,
+                            int last_dc_val, c_derived_tbl *dctbl,
+                            c_derived_tbl *actbl)
+{
+  return jsimd_huff_encode_one_block_neon(state, buffer, block, last_dc_val,
+                                          dctbl, actbl);
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_first_prepare(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_encode_mcu_AC_first_prepare(const JCOEF *block,
+                                  const int *jpeg_natural_order_start, int Sl,
+                                  int Al, JCOEF *values, size_t *zerobits)
+{
+  jsimd_encode_mcu_AC_first_prepare_neon(block, jpeg_natural_order_start,
+                                         Sl, Al, values, zerobits);
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_refine_prepare(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block,
+                                   const int *jpeg_natural_order_start, int Sl,
+                                   int Al, JCOEF *absvalues, size_t *bits)
+{
+  return jsimd_encode_mcu_AC_refine_prepare_neon(block,
+                                                 jpeg_natural_order_start, Sl,
+                                                 Al, absvalues, bits);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch32/jsimd_neon.S Added

@@ -0,0 +1,1200 @@
+/*
+ * Armv7 Neon optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2009-2011, Nokia Corporation and/or its subsidiary(-ies).
+ *                          All Rights Reserved.
+ * Author:  Siarhei Siamashka <siarhei.siamashka@nokia.com>
+ * Copyright (C) 2014, Siarhei Siamashka.  All Rights Reserved.
+ * Copyright (C) 2014, Linaro Limited.  All Rights Reserved.
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack, "", %progbits  /* mark stack as non-executable */
+#endif
+
+.text
+.fpu neon
+.arch armv7a
+.object_arch armv4
+.arm
+.syntax unified
+
+
+/*****************************************************************************/
+
+/* Supplementary macro for setting function attributes */
+.macro asm_function fname
+#ifdef __APPLE__
+    .private_extern _\fname
+    .globl _\fname
+_\fname:
+#else
+    .global \fname
+#ifdef __ELF__
+    .hidden \fname
+    .type \fname, %function
+#endif
+\fname:
+#endif
+.endm
+
+
+#define CENTERJSAMPLE  128
+
+/*****************************************************************************/
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ *
+ * GLOBAL(void)
+ * jsimd_idct_islow_neon(void *dct_table, JCOEFPTR coef_block,
+ *                       JSAMPARRAY output_buf, JDIMENSION output_col)
+ */
+
+#define FIX_0_298631336  (2446)
+#define FIX_0_390180644  (3196)
+#define FIX_0_541196100  (4433)
+#define FIX_0_765366865  (6270)
+#define FIX_0_899976223  (7373)
+#define FIX_1_175875602  (9633)
+#define FIX_1_501321110  (12299)
+#define FIX_1_847759065  (15137)
+#define FIX_1_961570560  (16069)
+#define FIX_2_053119869  (16819)
+#define FIX_2_562915447  (20995)
+#define FIX_3_072711026  (25172)
+
+#define FIX_1_175875602_MINUS_1_961570560  (FIX_1_175875602 - FIX_1_961570560)
+#define FIX_1_175875602_MINUS_0_390180644  (FIX_1_175875602 - FIX_0_390180644)
+#define FIX_0_541196100_MINUS_1_847759065  (FIX_0_541196100 - FIX_1_847759065)
+#define FIX_3_072711026_MINUS_2_562915447  (FIX_3_072711026 - FIX_2_562915447)
+#define FIX_0_298631336_MINUS_0_899976223  (FIX_0_298631336 - FIX_0_899976223)
+#define FIX_1_501321110_MINUS_0_899976223  (FIX_1_501321110 - FIX_0_899976223)
+#define FIX_2_053119869_MINUS_2_562915447  (FIX_2_053119869 - FIX_2_562915447)
+#define FIX_0_541196100_PLUS_0_765366865   (FIX_0_541196100 + FIX_0_765366865)
+
+/*
+ * Reference SIMD-friendly 1-D ISLOW iDCT C implementation.
+ * Uses some ideas from the comments in 'simd/jiss2int-64.asm'
+ */
+#define REF_1D_IDCT(xrow0, xrow1, xrow2, xrow3, xrow4, xrow5, xrow6, xrow7) { \
+  DCTELEM row0, row1, row2, row3, row4, row5, row6, row7; \
+  JLONG   q1, q2, q3, q4, q5, q6, q7; \
+  JLONG   tmp11_plus_tmp2, tmp11_minus_tmp2; \
+  \
+  /* 1-D iDCT input data */ \
+  row0 = xrow0; \
+  row1 = xrow1; \
+  row2 = xrow2; \
+  row3 = xrow3; \
+  row4 = xrow4; \
+  row5 = xrow5; \
+  row6 = xrow6; \
+  row7 = xrow7; \
+  \
+  q5 = row7 + row3; \
+  q4 = row5 + row1; \
+  q6 = MULTIPLY(q5, FIX_1_175875602_MINUS_1_961570560) + \
+       MULTIPLY(q4, FIX_1_175875602); \
+  q7 = MULTIPLY(q5, FIX_1_175875602) + \
+       MULTIPLY(q4, FIX_1_175875602_MINUS_0_390180644); \
+  q2 = MULTIPLY(row2, FIX_0_541196100) + \
+       MULTIPLY(row6, FIX_0_541196100_MINUS_1_847759065); \
+  q4 = q6; \
+  q3 = ((JLONG)row0 - (JLONG)row4) << 13; \
+  q6 += MULTIPLY(row5, -FIX_2_562915447) + \
+        MULTIPLY(row3, FIX_3_072711026_MINUS_2_562915447); \
+  /* now we can use q1 (reloadable constants have been used up) */ \
+  q1 = q3 + q2; \
+  q4 += MULTIPLY(row7, FIX_0_298631336_MINUS_0_899976223) + \
+        MULTIPLY(row1, -FIX_0_899976223); \
+  q5 = q7; \
+  q1 = q1 + q6; \
+  q7 += MULTIPLY(row7, -FIX_0_899976223) + \
+        MULTIPLY(row1, FIX_1_501321110_MINUS_0_899976223); \
+  \
+  /* (tmp11 + tmp2) has been calculated (out_row1 before descale) */ \
+  tmp11_plus_tmp2 = q1; \
+  row1 = 0; \
+  \
+  q1 = q1 - q6; \
+  q5 += MULTIPLY(row5, FIX_2_053119869_MINUS_2_562915447) + \
+        MULTIPLY(row3, -FIX_2_562915447); \
+  q1 = q1 - q6; \
+  q6 = MULTIPLY(row2, FIX_0_541196100_PLUS_0_765366865) + \
+       MULTIPLY(row6, FIX_0_541196100); \
+  q3 = q3 - q2; \
+  \
+  /* (tmp11 - tmp2) has been calculated (out_row6 before descale) */ \
+  tmp11_minus_tmp2 = q1; \
+  \
+  q1 = ((JLONG)row0 + (JLONG)row4) << 13; \
+  q2 = q1 + q6; \
+  q1 = q1 - q6; \
+  \
+  /* pick up the results */ \
+  tmp0  = q4; \
+  tmp1  = q5; \
+  tmp2  = (tmp11_plus_tmp2 - tmp11_minus_tmp2) / 2; \
+  tmp3  = q7; \
+  tmp10 = q2; \
+  tmp11 = (tmp11_plus_tmp2 + tmp11_minus_tmp2) / 2; \
+  tmp12 = q3; \
+  tmp13 = q1; \
+}
+
+#define XFIX_0_899976223                    d00
+#define XFIX_0_541196100                    d01
+#define XFIX_2_562915447                    d02
+#define XFIX_0_298631336_MINUS_0_899976223  d03
+#define XFIX_1_501321110_MINUS_0_899976223  d10
+#define XFIX_2_053119869_MINUS_2_562915447  d11
+#define XFIX_0_541196100_PLUS_0_765366865   d12
+#define XFIX_1_175875602                    d13
+#define XFIX_1_175875602_MINUS_0_390180644  d20
+#define XFIX_0_541196100_MINUS_1_847759065  d21
+#define XFIX_3_072711026_MINUS_2_562915447  d22
+#define XFIX_1_175875602_MINUS_1_961570560  d23
+
+.balign 16
+jsimd_idct_islow_neon_consts:
+  .short FIX_0_899976223                    /* d00 */
+  .short FIX_0_541196100                    /* d01 */
+  .short FIX_2_562915447                    /* d02 */
+  .short FIX_0_298631336_MINUS_0_899976223  /* d03 */
+  .short FIX_1_501321110_MINUS_0_899976223  /* d10 */
+  .short FIX_2_053119869_MINUS_2_562915447  /* d11 */
+  .short FIX_0_541196100_PLUS_0_765366865   /* d12 */
+  .short FIX_1_175875602                    /* d13 */
+  /* reloadable constants */
+  .short FIX_1_175875602_MINUS_0_390180644  /* d20 */
+  .short FIX_0_541196100_MINUS_1_847759065  /* d21 */
+  .short FIX_3_072711026_MINUS_2_562915447  /* d22 */
+  .short FIX_1_175875602_MINUS_1_961570560  /* d23 */
+
+asm_function jsimd_idct_islow_neon
+
+    DCT_TABLE       .req r0
+    COEF_BLOCK      .req r1
+    OUTPUT_BUF      .req r2
+    OUTPUT_COL      .req r3
+    TMP1            .req r0
+    TMP2            .req r1
+    TMP3            .req r2
+    TMP4            .req ip
+
+    ROW0L           .req d16
+    ROW0R           .req d17
+    ROW1L           .req d18
+    ROW1R           .req d19
+    ROW2L           .req d20
+    ROW2R           .req d21
+    ROW3L           .req d22
+    ROW3R           .req d23
+    ROW4L           .req d24
+    ROW4R           .req d25
+    ROW5L           .req d26
+    ROW5R           .req d27
+    ROW6L           .req d28
+    ROW6R           .req d29
+    ROW7L           .req d30
+    ROW7R           .req d31
+
+    /* Load and dequantize coefficients into Neon registers
+     * with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d16     | d17     ( q8  )
+     *   1 | d18     | d19     ( q9  )
+     *   2 | d20     | d21     ( q10 )
+     *   3 | d22     | d23     ( q11 )
+     *   4 | d24     | d25     ( q12 )
+     *   5 | d26     | d27     ( q13 )
+     *   6 | d28     | d29     ( q14 )
+     *   7 | d30     | d31     ( q15 )
+     */
+    adr             ip, jsimd_idct_islow_neon_consts
+    vld1.16         {d16, d17, d18, d19}, COEF_BLOCK, :128!
+    vld1.16         {d0, d1, d2, d3}, DCT_TABLE, :128!
+    vld1.16         {d20, d21, d22, d23}, COEF_BLOCK, :128!
+    vmul.s16        q8, q8, q0
+    vld1.16         {d4, d5, d6, d7}, DCT_TABLE, :128!
+    vmul.s16        q9, q9, q1
+    vld1.16         {d24, d25, d26, d27}, COEF_BLOCK, :128!
+    vmul.s16        q10, q10, q2
+    vld1.16         {d0, d1, d2, d3}, DCT_TABLE, :128!
+    vmul.s16        q11, q11, q3
+    vld1.16         {d28, d29, d30, d31}, COEF_BLOCK, :128
+    vmul.s16        q12, q12, q0
+    vld1.16         {d4, d5, d6, d7}, DCT_TABLE, :128!
+    vmul.s16        q14, q14, q2
+    vmul.s16        q13, q13, q1
+    vld1.16         {d0, d1, d2, d3}, ip, :128  /* load constants */
+    add             ip, ip, #16
+    vmul.s16        q15, q15, q3
+    vpush           {d8 - d15}                    /* save Neon registers */
+    /* 1-D IDCT, pass 1, left 4x8 half */
+    vadd.s16        d4, ROW7L, ROW3L
+    vadd.s16        d5, ROW5L, ROW1L
+    vmull.s16       q6, d4, XFIX_1_175875602_MINUS_1_961570560
+    vmlal.s16       q6, d5, XFIX_1_175875602
+    vmull.s16       q7, d4, XFIX_1_175875602
+      /* Check for the zero coefficients in the right 4x8 half */
+      push            {r4, r5}
+    vmlal.s16       q7, d5, XFIX_1_175875602_MINUS_0_390180644
+    vsubl.s16       q3, ROW0L, ROW4L
+      ldrd            r4, COEF_BLOCK, #(-96 + 2 * (4 + 1 * 8))
+    vmull.s16       q2, ROW2L, XFIX_0_541196100
+    vmlal.s16       q2, ROW6L, XFIX_0_541196100_MINUS_1_847759065
+      orr             r0, r4, r5
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW5L, XFIX_2_562915447
+      ldrd            r4, COEF_BLOCK, #(-96 + 2 * (4 + 2 * 8))
+    vmlal.s16       q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447
+    vshl.s32        q3, q3, #13
+      orr             r0, r0, r4
+    vmlsl.s16       q4, ROW1L, XFIX_0_899976223
+      orr             r0, r0, r5
+    vadd.s32        q1, q3, q2
+      ldrd            r4, COEF_BLOCK, #(-96 + 2 * (4 + 3 * 8))
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+      orr             r0, r0, r4
+    vmlsl.s16       q7, ROW7L, XFIX_0_899976223
+      orr             r0, r0, r5
+    vmlal.s16       q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223
+    vrshrn.s32      ROW1L, q1, #11
+      ldrd            r4, COEF_BLOCK, #(-96 + 2 * (4 + 4 * 8))
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW5L, XFIX_2_053119869_MINUS_2_562915447
+      orr             r0, r0, r4
+    vmlsl.s16       q5, ROW3L, XFIX_2_562915447
+      orr             r0, r0, r5
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865
+      ldrd            r4, COEF_BLOCK, #(-96 + 2 * (4 + 5 * 8))
+    vmlal.s16       q6, ROW6L, XFIX_0_541196100
+    vsub.s32        q3, q3, q2
+      orr             r0, r0, r4
+    vrshrn.s32      ROW6L, q1, #11
+      orr             r0, r0, r5
+    vadd.s32        q1, q3, q5
+      ldrd            r4, COEF_BLOCK, #(-96 + 2 * (4 + 6 * 8))
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW0L, ROW4L
+      orr             r0, r0, r4
+    vrshrn.s32      ROW2L, q1, #11
+      orr             r0, r0, r5
+    vrshrn.s32      ROW5L, q3, #11
+      ldrd            r4, COEF_BLOCK, #(-96 + 2 * (4 + 7 * 8))
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW7L, XFIX_0_298631336_MINUS_0_899976223
+      orr             r0, r0, r4
+    vadd.s32        q2, q5, q6
+      orrs            r0, r0, r5
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+      ldrd            r4, COEF_BLOCK, #(-96 + 2 * (4 + 0 * 8))
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+      orr             r0, r4, r5
+    vsub.s32        q3, q1, q4
+      pop             {r4, r5}
+    vrshrn.s32      ROW7L, q2, #11
+    vrshrn.s32      ROW3L, q5, #11
+    vrshrn.s32      ROW0L, q6, #11
+    vrshrn.s32      ROW4L, q3, #11
+
+      beq             3f  /* Go to do some special handling for the sparse
+                             right 4x8 half */
+
+    /* 1-D IDCT, pass 1, right 4x8 half */
+    vld1.s16        {d2}, ip, :64  /* reload constants */
+    vadd.s16        d10, ROW7R, ROW3R
+    vadd.s16        d8, ROW5R, ROW1R
+      /* Transpose left 4x8 half */
+      vtrn.16         ROW6L, ROW7L
+    vmull.s16       q6, d10, XFIX_1_175875602_MINUS_1_961570560
+    vmlal.s16       q6, d8, XFIX_1_175875602
+      vtrn.16         ROW2L, ROW3L
+    vmull.s16       q7, d10, XFIX_1_175875602
+    vmlal.s16       q7, d8, XFIX_1_175875602_MINUS_0_390180644
+      vtrn.16         ROW0L, ROW1L
+    vsubl.s16       q3, ROW0R, ROW4R
+    vmull.s16       q2, ROW2R, XFIX_0_541196100
+    vmlal.s16       q2, ROW6R, XFIX_0_541196100_MINUS_1_847759065
+      vtrn.16         ROW4L, ROW5L
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW5R, XFIX_2_562915447
+    vmlal.s16       q6, ROW3R, XFIX_3_072711026_MINUS_2_562915447
+      vtrn.32         ROW1L, ROW3L
+    vshl.s32        q3, q3, #13
+    vmlsl.s16       q4, ROW1R, XFIX_0_899976223
+      vtrn.32         ROW4L, ROW6L
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+      vtrn.32         ROW0L, ROW2L
+    vmlsl.s16       q7, ROW7R, XFIX_0_899976223
+    vmlal.s16       q7, ROW1R, XFIX_1_501321110_MINUS_0_899976223
+    vrshrn.s32      ROW1R, q1, #11
+      vtrn.32         ROW5L, ROW7L
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW5R, XFIX_2_053119869_MINUS_2_562915447
+    vmlsl.s16       q5, ROW3R, XFIX_2_562915447
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2R, XFIX_0_541196100_PLUS_0_765366865
+    vmlal.s16       q6, ROW6R, XFIX_0_541196100
+    vsub.s32        q3, q3, q2
+    vrshrn.s32      ROW6R, q1, #11
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW0R, ROW4R
+    vrshrn.s32      ROW2R, q1, #11
+    vrshrn.s32      ROW5R, q3, #11
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW7R, XFIX_0_298631336_MINUS_0_899976223
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vrshrn.s32      ROW7R, q2, #11
+    vrshrn.s32      ROW3R, q5, #11
+    vrshrn.s32      ROW0R, q6, #11
+    vrshrn.s32      ROW4R, q3, #11
+    /* Transpose right 4x8 half */
+    vtrn.16         ROW6R, ROW7R
+    vtrn.16         ROW2R, ROW3R
+    vtrn.16         ROW0R, ROW1R
+    vtrn.16         ROW4R, ROW5R
+    vtrn.32         ROW1R, ROW3R
+    vtrn.32         ROW4R, ROW6R
+    vtrn.32         ROW0R, ROW2R
+    vtrn.32         ROW5R, ROW7R
+
+1:  /* 1-D IDCT, pass 2 (normal variant), left 4x8 half */
+    vld1.s16        {d2}, ip, :64               /* reload constants */
+    vmull.s16       q6, ROW1R, XFIX_1_175875602   /* ROW5L <-> ROW1R */
+    vmlal.s16       q6, ROW1L, XFIX_1_175875602
+    vmlal.s16       q6, ROW3R, XFIX_1_175875602_MINUS_1_961570560  /* ROW7L <-> ROW3R */
+    vmlal.s16       q6, ROW3L, XFIX_1_175875602_MINUS_1_961570560
+    vmull.s16       q7, ROW3R, XFIX_1_175875602   /* ROW7L <-> ROW3R */
+    vmlal.s16       q7, ROW3L, XFIX_1_175875602
+    vmlal.s16       q7, ROW1R, XFIX_1_175875602_MINUS_0_390180644  /* ROW5L <-> ROW1R */
+    vmlal.s16       q7, ROW1L, XFIX_1_175875602_MINUS_0_390180644
+    vsubl.s16       q3, ROW0L, ROW0R              /* ROW4L <-> ROW0R */
+    vmull.s16       q2, ROW2L, XFIX_0_541196100
+    vmlal.s16       q2, ROW2R, XFIX_0_541196100_MINUS_1_847759065  /* ROW6L <-> ROW2R */
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW1R, XFIX_2_562915447   /* ROW5L <-> ROW1R */
+    vmlal.s16       q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447
+    vshl.s32        q3, q3, #13
+    vmlsl.s16       q4, ROW1L, XFIX_0_899976223
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+    vmlsl.s16       q7, ROW3R, XFIX_0_899976223   /* ROW7L <-> ROW3R */
+    vmlal.s16       q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223
+    vshrn.s32       ROW1L, q1, #16
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW1R, XFIX_2_053119869_MINUS_2_562915447  /* ROW5L <-> ROW1R */
+    vmlsl.s16       q5, ROW3L, XFIX_2_562915447
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865
+    vmlal.s16       q6, ROW2R, XFIX_0_541196100   /* ROW6L <-> ROW2R */
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW2R, q1, #16                /* ROW6L <-> ROW2R */
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW0L, ROW0R              /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW2L, q1, #16
+    vshrn.s32       ROW1R, q3, #16                /* ROW5L <-> ROW1R */
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW3R, XFIX_0_298631336_MINUS_0_899976223  /* ROW7L <-> ROW3R */
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW3R, q2, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW3L, q5, #16
+    vshrn.s32       ROW0L, q6, #16
+    vshrn.s32       ROW0R, q3, #16                /* ROW4L <-> ROW0R */
+    /* 1-D IDCT, pass 2, right 4x8 half */
+    vld1.s16        {d2}, ip, :64               /* reload constants */
+    vmull.s16       q6, ROW5R, XFIX_1_175875602
+    vmlal.s16       q6, ROW5L, XFIX_1_175875602   /* ROW5L <-> ROW1R */
+    vmlal.s16       q6, ROW7R, XFIX_1_175875602_MINUS_1_961570560
+    vmlal.s16       q6, ROW7L, XFIX_1_175875602_MINUS_1_961570560  /* ROW7L <-> ROW3R */
+    vmull.s16       q7, ROW7R, XFIX_1_175875602
+    vmlal.s16       q7, ROW7L, XFIX_1_175875602   /* ROW7L <-> ROW3R */
+    vmlal.s16       q7, ROW5R, XFIX_1_175875602_MINUS_0_390180644
+    vmlal.s16       q7, ROW5L, XFIX_1_175875602_MINUS_0_390180644  /* ROW5L <-> ROW1R */
+    vsubl.s16       q3, ROW4L, ROW4R              /* ROW4L <-> ROW0R */
+    vmull.s16       q2, ROW6L, XFIX_0_541196100   /* ROW6L <-> ROW2R */
+    vmlal.s16       q2, ROW6R, XFIX_0_541196100_MINUS_1_847759065
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW5R, XFIX_2_562915447
+    vmlal.s16       q6, ROW7L, XFIX_3_072711026_MINUS_2_562915447  /* ROW7L <-> ROW3R */
+    vshl.s32        q3, q3, #13
+    vmlsl.s16       q4, ROW5L, XFIX_0_899976223   /* ROW5L <-> ROW1R */
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+    vmlsl.s16       q7, ROW7R, XFIX_0_899976223
+    vmlal.s16       q7, ROW5L, XFIX_1_501321110_MINUS_0_899976223  /* ROW5L <-> ROW1R */
+    vshrn.s32       ROW5L, q1, #16                /* ROW5L <-> ROW1R */
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW5R, XFIX_2_053119869_MINUS_2_562915447
+    vmlsl.s16       q5, ROW7L, XFIX_2_562915447   /* ROW7L <-> ROW3R */
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW6L, XFIX_0_541196100_PLUS_0_765366865  /* ROW6L <-> ROW2R */
+    vmlal.s16       q6, ROW6R, XFIX_0_541196100
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW6R, q1, #16
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW4L, ROW4R              /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW6L, q1, #16                /* ROW6L <-> ROW2R */
+    vshrn.s32       ROW5R, q3, #16
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW7R, XFIX_0_298631336_MINUS_0_899976223
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW7R, q2, #16
+    vshrn.s32       ROW7L, q5, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW4L, q6, #16                /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW4R, q3, #16
+
+2:  /* Descale to 8-bit and range limit */
+    vqrshrn.s16     d16, q8, #2
+    vqrshrn.s16     d17, q9, #2
+    vqrshrn.s16     d18, q10, #2
+    vqrshrn.s16     d19, q11, #2
+    vpop            {d8 - d15}                    /* restore Neon registers */
+    vqrshrn.s16     d20, q12, #2
+      /* Transpose the final 8-bit samples and do signed->unsigned conversion */
+      vtrn.16         q8, q9
+    vqrshrn.s16     d21, q13, #2
+    vqrshrn.s16     d22, q14, #2
+      vmov.u8         q0, #(CENTERJSAMPLE)
+    vqrshrn.s16     d23, q15, #2
+      vtrn.8          d16, d17
+      vtrn.8          d18, d19
+      vadd.u8         q8, q8, q0
+      vadd.u8         q9, q9, q0
+      vtrn.16         q10, q11
+        /* Store results to the output buffer */
+        ldmia           OUTPUT_BUF!, {TMP1, TMP2}
+        add             TMP1, TMP1, OUTPUT_COL
+        add             TMP2, TMP2, OUTPUT_COL
+        vst1.8          {d16}, TMP1
+      vtrn.8          d20, d21
+        vst1.8          {d17}, TMP2
+        ldmia           OUTPUT_BUF!, {TMP1, TMP2}
+        add             TMP1, TMP1, OUTPUT_COL
+        add             TMP2, TMP2, OUTPUT_COL
+        vst1.8          {d18}, TMP1
+      vadd.u8         q10, q10, q0
+        vst1.8          {d19}, TMP2
+        ldmia           OUTPUT_BUF, {TMP1, TMP2, TMP3, TMP4}
+        add             TMP1, TMP1, OUTPUT_COL
+        add             TMP2, TMP2, OUTPUT_COL
+        add             TMP3, TMP3, OUTPUT_COL
+        add             TMP4, TMP4, OUTPUT_COL
+      vtrn.8          d22, d23
+        vst1.8          {d20}, TMP1
+      vadd.u8         q11, q11, q0
+        vst1.8          {d21}, TMP2
+        vst1.8          {d22}, TMP3
+        vst1.8          {d23}, TMP4
+    bx              lr
+
+3:  /* Left 4x8 half is done, right 4x8 half contains mostly zeros */
+
+    /* Transpose left 4x8 half */
+    vtrn.16         ROW6L, ROW7L
+    vtrn.16         ROW2L, ROW3L
+    vtrn.16         ROW0L, ROW1L
+    vtrn.16         ROW4L, ROW5L
+    vshl.s16        ROW0R, ROW0R, #2  /* PASS1_BITS */
+    vtrn.32         ROW1L, ROW3L
+    vtrn.32         ROW4L, ROW6L
+    vtrn.32         ROW0L, ROW2L
+    vtrn.32         ROW5L, ROW7L
+
+    cmp             r0, #0
+    beq             4f  /* Right 4x8 half has all zeros, go to 'sparse' second
+                           pass */
+
+    /* Only row 0 is non-zero for the right 4x8 half  */
+    vdup.s16        ROW1R, ROW0R1
+    vdup.s16        ROW2R, ROW0R2
+    vdup.s16        ROW3R, ROW0R3
+    vdup.s16        ROW4R, ROW0R0
+    vdup.s16        ROW5R, ROW0R1
+    vdup.s16        ROW6R, ROW0R2
+    vdup.s16        ROW7R, ROW0R3
+    vdup.s16        ROW0R, ROW0R0
+    b               1b  /* Go to 'normal' second pass */
+
+4:  /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), left 4x8 half */
+    vld1.s16        {d2}, ip, :64               /* reload constants */
+    vmull.s16       q6, ROW1L, XFIX_1_175875602
+    vmlal.s16       q6, ROW3L, XFIX_1_175875602_MINUS_1_961570560
+    vmull.s16       q7, ROW3L, XFIX_1_175875602
+    vmlal.s16       q7, ROW1L, XFIX_1_175875602_MINUS_0_390180644
+    vmull.s16       q2, ROW2L, XFIX_0_541196100
+    vshll.s16       q3, ROW0L, #13
+    vmov            q4, q6
+    vmlal.s16       q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447
+    vmlsl.s16       q4, ROW1L, XFIX_0_899976223
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vmlal.s16       q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223
+    vadd.s32        q1, q1, q6
+    vadd.s32        q6, q6, q6
+    vmlsl.s16       q5, ROW3L, XFIX_2_562915447
+    vshrn.s32       ROW1L, q1, #16
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW2R, q1, #16                /* ROW6L <-> ROW2R */
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vshll.s16       q5, ROW0L, #13
+    vshrn.s32       ROW2L, q1, #16
+    vshrn.s32       ROW1R, q3, #16                /* ROW5L <-> ROW1R */
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW3R, q2, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW3L, q5, #16
+    vshrn.s32       ROW0L, q6, #16
+    vshrn.s32       ROW0R, q3, #16                /* ROW4L <-> ROW0R */
+    /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), right 4x8 half */
+    vld1.s16        {d2}, ip, :64               /* reload constants */
+    vmull.s16       q6, ROW5L, XFIX_1_175875602
+    vmlal.s16       q6, ROW7L, XFIX_1_175875602_MINUS_1_961570560
+    vmull.s16       q7, ROW7L, XFIX_1_175875602
+    vmlal.s16       q7, ROW5L, XFIX_1_175875602_MINUS_0_390180644
+    vmull.s16       q2, ROW6L, XFIX_0_541196100
+    vshll.s16       q3, ROW4L, #13
+    vmov            q4, q6
+    vmlal.s16       q6, ROW7L, XFIX_3_072711026_MINUS_2_562915447
+    vmlsl.s16       q4, ROW5L, XFIX_0_899976223
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vmlal.s16       q7, ROW5L, XFIX_1_501321110_MINUS_0_899976223
+    vadd.s32        q1, q1, q6
+    vadd.s32        q6, q6, q6
+    vmlsl.s16       q5, ROW7L, XFIX_2_562915447
+    vshrn.s32       ROW5L, q1, #16                /* ROW5L <-> ROW1R */
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW6L, XFIX_0_541196100_PLUS_0_765366865
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW6R, q1, #16
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vshll.s16       q5, ROW4L, #13
+    vshrn.s32       ROW6L, q1, #16                /* ROW6L <-> ROW2R */
+    vshrn.s32       ROW5R, q3, #16
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW7R, q2, #16
+    vshrn.s32       ROW7L, q5, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW4L, q6, #16                /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW4R, q3, #16
+    b               2b                            /* Go to epilogue */
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+
+    .unreq          ROW0L
+    .unreq          ROW0R
+    .unreq          ROW1L
+    .unreq          ROW1R
+    .unreq          ROW2L
+    .unreq          ROW2R
+    .unreq          ROW3L
+    .unreq          ROW3R
+    .unreq          ROW4L
+    .unreq          ROW4R
+    .unreq          ROW5L
+    .unreq          ROW5R
+    .unreq          ROW6L
+    .unreq          ROW6R
+    .unreq          ROW7L
+    .unreq          ROW7R
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_idct_ifast_neon
+ *
+ * This function contains a fast, not so accurate integer implementation of
+ * the inverse DCT (Discrete Cosine Transform). It uses the same calculations
+ * and produces exactly the same output as IJG's original 'jpeg_idct_ifast'
+ * function from jidctfst.c
+ *
+ * Normally 1-D AAN DCT needs 5 multiplications and 29 additions.
+ * But in Arm Neon case some extra additions are required because VQDMULH
+ * instruction can't handle the constants larger than 1. So the expressions
+ * like "x * 1.082392200" have to be converted to "x * 0.082392200 + x",
+ * which introduces an extra addition. Overall, there are 6 extra additions
+ * per 1-D IDCT pass, totalling to 5 VQDMULH and 35 VADD/VSUB instructions.
+ */
+
+#define XFIX_1_082392200  d00
+#define XFIX_1_414213562  d01
+#define XFIX_1_847759065  d02
+#define XFIX_2_613125930  d03
+
+.balign 16
+jsimd_idct_ifast_neon_consts:
+  .short (277 * 128 - 256 * 128)  /* XFIX_1_082392200 */
+  .short (362 * 128 - 256 * 128)  /* XFIX_1_414213562 */
+  .short (473 * 128 - 256 * 128)  /* XFIX_1_847759065 */
+  .short (669 * 128 - 512 * 128)  /* XFIX_2_613125930 */
+
+asm_function jsimd_idct_ifast_neon
+
+    DCT_TABLE       .req r0
+    COEF_BLOCK      .req r1
+    OUTPUT_BUF      .req r2
+    OUTPUT_COL      .req r3
+    TMP1            .req r0
+    TMP2            .req r1
+    TMP3            .req r2
+    TMP4            .req ip
+
+    /* Load and dequantize coefficients into Neon registers
+     * with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d16     | d17     ( q8  )
+     *   1 | d18     | d19     ( q9  )
+     *   2 | d20     | d21     ( q10 )
+     *   3 | d22     | d23     ( q11 )
+     *   4 | d24     | d25     ( q12 )
+     *   5 | d26     | d27     ( q13 )
+     *   6 | d28     | d29     ( q14 )
+     *   7 | d30     | d31     ( q15 )
+     */
+    adr             ip, jsimd_idct_ifast_neon_consts
+    vld1.16         {d16, d17, d18, d19}, COEF_BLOCK, :128!
+    vld1.16         {d0, d1, d2, d3}, DCT_TABLE, :128!
+    vld1.16         {d20, d21, d22, d23}, COEF_BLOCK, :128!
+    vmul.s16        q8, q8, q0
+    vld1.16         {d4, d5, d6, d7}, DCT_TABLE, :128!
+    vmul.s16        q9, q9, q1
+    vld1.16         {d24, d25, d26, d27}, COEF_BLOCK, :128!
+    vmul.s16        q10, q10, q2
+    vld1.16         {d0, d1, d2, d3}, DCT_TABLE, :128!
+    vmul.s16        q11, q11, q3
+    vld1.16         {d28, d29, d30, d31}, COEF_BLOCK, :128
+    vmul.s16        q12, q12, q0
+    vld1.16         {d4, d5, d6, d7}, DCT_TABLE, :128!
+    vmul.s16        q14, q14, q2
+    vmul.s16        q13, q13, q1
+    vld1.16         {d0}, ip, :64  /* load constants */
+    vmul.s16        q15, q15, q3
+    vpush           {d8 - d13}       /* save Neon registers */
+    /* 1-D IDCT, pass 1 */
+    vsub.s16        q2, q10, q14
+    vadd.s16        q14, q10, q14
+    vsub.s16        q1, q11, q13
+    vadd.s16        q13, q11, q13
+    vsub.s16        q5, q9, q15
+    vadd.s16        q15, q9, q15
+    vqdmulh.s16     q4, q2, XFIX_1_414213562
+    vqdmulh.s16     q6, q1, XFIX_2_613125930
+    vadd.s16        q3, q1, q1
+    vsub.s16        q1, q5, q1
+    vadd.s16        q10, q2, q4
+    vqdmulh.s16     q4, q1, XFIX_1_847759065
+    vsub.s16        q2, q15, q13
+    vadd.s16        q3, q3, q6
+    vqdmulh.s16     q6, q2, XFIX_1_414213562
+    vadd.s16        q1, q1, q4
+    vqdmulh.s16     q4, q5, XFIX_1_082392200
+    vsub.s16        q10, q10, q14
+    vadd.s16        q2, q2, q6
+    vsub.s16        q6, q8, q12
+    vadd.s16        q12, q8, q12
+    vadd.s16        q9, q5, q4
+    vadd.s16        q5, q6, q10
+    vsub.s16        q10, q6, q10
+    vadd.s16        q6, q15, q13
+    vadd.s16        q8, q12, q14
+    vsub.s16        q3, q6, q3
+    vsub.s16        q12, q12, q14
+    vsub.s16        q3, q3, q1
+    vsub.s16        q1, q9, q1
+    vadd.s16        q2, q3, q2
+    vsub.s16        q15, q8, q6
+    vadd.s16        q1, q1, q2
+    vadd.s16        q8, q8, q6
+    vadd.s16        q14, q5, q3
+    vsub.s16        q9, q5, q3
+    vsub.s16        q13, q10, q2
+    vadd.s16        q10, q10, q2
+      /* Transpose */
+      vtrn.16         q8, q9
+    vsub.s16        q11, q12, q1
+      vtrn.16         q14, q15
+    vadd.s16        q12, q12, q1
+      vtrn.16         q10, q11
+      vtrn.16         q12, q13
+      vtrn.32         q9, q11
+      vtrn.32         q12, q14
+      vtrn.32         q8, q10
+      vtrn.32         q13, q15
+      vswp            d28, d21
+      vswp            d26, d19
+    /* 1-D IDCT, pass 2 */
+    vsub.s16        q2, q10, q14
+      vswp            d30, d23
+    vadd.s16        q14, q10, q14
+      vswp            d24, d17
+    vsub.s16        q1, q11, q13
+    vadd.s16        q13, q11, q13
+    vsub.s16        q5, q9, q15
+    vadd.s16        q15, q9, q15
+    vqdmulh.s16     q4, q2, XFIX_1_414213562
+    vqdmulh.s16     q6, q1, XFIX_2_613125930
+    vadd.s16        q3, q1, q1
+    vsub.s16        q1, q5, q1
+    vadd.s16        q10, q2, q4
+    vqdmulh.s16     q4, q1, XFIX_1_847759065
+    vsub.s16        q2, q15, q13
+    vadd.s16        q3, q3, q6
+    vqdmulh.s16     q6, q2, XFIX_1_414213562
+    vadd.s16        q1, q1, q4
+    vqdmulh.s16     q4, q5, XFIX_1_082392200
+    vsub.s16        q10, q10, q14
+    vadd.s16        q2, q2, q6
+    vsub.s16        q6, q8, q12
+    vadd.s16        q12, q8, q12
+    vadd.s16        q9, q5, q4
+    vadd.s16        q5, q6, q10
+    vsub.s16        q10, q6, q10
+    vadd.s16        q6, q15, q13
+    vadd.s16        q8, q12, q14
+    vsub.s16        q3, q6, q3
+    vsub.s16        q12, q12, q14
+    vsub.s16        q3, q3, q1
+    vsub.s16        q1, q9, q1
+    vadd.s16        q2, q3, q2
+    vsub.s16        q15, q8, q6
+    vadd.s16        q1, q1, q2
+    vadd.s16        q8, q8, q6
+    vadd.s16        q14, q5, q3
+    vsub.s16        q9, q5, q3
+    vsub.s16        q13, q10, q2
+    vpop            {d8 - d13}    /* restore Neon registers */
+    vadd.s16        q10, q10, q2
+    vsub.s16        q11, q12, q1
+    vadd.s16        q12, q12, q1
+    /* Descale to 8-bit and range limit */
+    vmov.u8         q0, #0x80
+    vqshrn.s16      d16, q8, #5
+    vqshrn.s16      d17, q9, #5
+    vqshrn.s16      d18, q10, #5
+    vqshrn.s16      d19, q11, #5
+    vqshrn.s16      d20, q12, #5
+    vqshrn.s16      d21, q13, #5
+    vqshrn.s16      d22, q14, #5
+    vqshrn.s16      d23, q15, #5
+    vadd.u8         q8, q8, q0
+    vadd.u8         q9, q9, q0
+    vadd.u8         q10, q10, q0
+    vadd.u8         q11, q11, q0
+    /* Transpose the final 8-bit samples */
+    vtrn.16         q8, q9
+    vtrn.16         q10, q11
+    vtrn.32         q8, q10
+    vtrn.32         q9, q11
+    vtrn.8          d16, d17
+    vtrn.8          d18, d19
+      /* Store results to the output buffer */
+      ldmia           OUTPUT_BUF!, {TMP1, TMP2}
+      add             TMP1, TMP1, OUTPUT_COL
+      add             TMP2, TMP2, OUTPUT_COL
+      vst1.8          {d16}, TMP1
+      vst1.8          {d17}, TMP2
+      ldmia           OUTPUT_BUF!, {TMP1, TMP2}
+      add             TMP1, TMP1, OUTPUT_COL
+      add             TMP2, TMP2, OUTPUT_COL
+      vst1.8          {d18}, TMP1
+    vtrn.8          d20, d21
+      vst1.8          {d19}, TMP2
+      ldmia           OUTPUT_BUF, {TMP1, TMP2, TMP3, TMP4}
+      add             TMP1, TMP1, OUTPUT_COL
+      add             TMP2, TMP2, OUTPUT_COL
+      add             TMP3, TMP3, OUTPUT_COL
+      add             TMP4, TMP4, OUTPUT_COL
+      vst1.8          {d20}, TMP1
+    vtrn.8          d22, d23
+      vst1.8          {d21}, TMP2
+      vst1.8          {d22}, TMP3
+      vst1.8          {d23}, TMP4
+    bx              lr
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_extrgb_ycc_convert_neon
+ * jsimd_extbgr_ycc_convert_neon
+ * jsimd_extrgbx_ycc_convert_neon
+ * jsimd_extbgrx_ycc_convert_neon
+ * jsimd_extxbgr_ycc_convert_neon
+ * jsimd_extxrgb_ycc_convert_neon
+ *
+ * Colorspace conversion RGB -> YCbCr
+ */
+
+.macro do_store size
+  .if \size == 8
+    vst1.8          {d20}, Y!
+    vst1.8          {d21}, U!
+    vst1.8          {d22}, V!
+  .elseif \size == 4
+    vst1.8          {d200}, Y!
+    vst1.8          {d201}, Y!
+    vst1.8          {d202}, Y!
+    vst1.8          {d203}, Y!
+    vst1.8          {d210}, U!
+    vst1.8          {d211}, U!
+    vst1.8          {d212}, U!
+    vst1.8          {d213}, U!
+    vst1.8          {d220}, V!
+    vst1.8          {d221}, V!
+    vst1.8          {d222}, V!
+    vst1.8          {d223}, V!
+  .elseif \size == 2
+    vst1.8          {d204}, Y!
+    vst1.8          {d205}, Y!
+    vst1.8          {d214}, U!
+    vst1.8          {d215}, U!
+    vst1.8          {d224}, V!
+    vst1.8          {d225}, V!
+  .elseif \size == 1
+    vst1.8          {d206}, Y!
+    vst1.8          {d216}, U!
+    vst1.8          {d226}, V!
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_load bpp, size
+  .if \bpp == 24
+    .if \size == 8
+      vld3.8        {d10, d11, d12}, RGB!
+      pld           RGB, #128
+    .elseif \size == 4
+      vld3.8        {d100, d110, d120}, RGB!
+      vld3.8        {d101, d111, d121}, RGB!
+      vld3.8        {d102, d112, d122}, RGB!
+      vld3.8        {d103, d113, d123}, RGB!
+    .elseif \size == 2
+      vld3.8        {d104, d114, d124}, RGB!
+      vld3.8        {d105, d115, d125}, RGB!
+    .elseif \size == 1
+      vld3.8        {d106, d116, d126}, RGB!
+    .else
+      .error unsupported macroblock size
+    .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      vld4.8        {d10, d11, d12, d13}, RGB!
+      pld           RGB, #128
+    .elseif \size == 4
+      vld4.8        {d100, d110, d120, d130}, RGB!
+      vld4.8        {d101, d111, d121, d131}, RGB!
+      vld4.8        {d102, d112, d122, d132}, RGB!
+      vld4.8        {d103, d113, d123, d133}, RGB!
+    .elseif \size == 2
+      vld4.8        {d104, d114, d124, d134}, RGB!
+      vld4.8        {d105, d115, d125, d135}, RGB!
+    .elseif \size == 1
+      vld4.8        {d106, d116, d126, d136}, RGB!
+    .else
+      .error unsupported macroblock size
+    .endif
+  .else
+    .error unsupported bpp
+  .endif
+.endm
+
+.macro generate_jsimd_rgb_ycc_convert_neon colorid, bpp, r_offs, g_offs, b_offs
+
+/*
+ * 2-stage pipelined RGB->YCbCr conversion
+ */
+
+.macro do_rgb_to_yuv_stage1
+    vmovl.u8        q2, d1\r_offs  /* r = { d4, d5 } */
+    vmovl.u8        q3, d1\g_offs  /* g = { d6, d7 } */
+    vmovl.u8        q4, d1\b_offs  /* b = { d8, d9 } */
+    vmull.u16       q7, d4, d00
+    vmlal.u16       q7, d6, d01
+    vmlal.u16       q7, d8, d02
+    vmull.u16       q8, d5, d00
+    vmlal.u16       q8, d7, d01
+    vmlal.u16       q8, d9, d02
+    vrev64.32       q9, q1
+    vrev64.32       q13, q1
+    vmlsl.u16       q9, d4, d03
+    vmlsl.u16       q9, d6, d10
+    vmlal.u16       q9, d8, d11
+    vmlsl.u16       q13, d5, d03
+    vmlsl.u16       q13, d7, d10
+    vmlal.u16       q13, d9, d11
+    vrev64.32       q14, q1
+    vrev64.32       q15, q1
+    vmlal.u16       q14, d4, d11
+    vmlsl.u16       q14, d6, d12
+    vmlsl.u16       q14, d8, d13
+    vmlal.u16       q15, d5, d11
+    vmlsl.u16       q15, d7, d12
+    vmlsl.u16       q15, d9, d13
+.endm
+
+.macro do_rgb_to_yuv_stage2
+    vrshrn.u32      d20, q7, #16
+    vrshrn.u32      d21, q8, #16
+    vshrn.u32       d22, q9, #16
+    vshrn.u32       d23, q13, #16
+    vshrn.u32       d24, q14, #16
+    vshrn.u32       d25, q15, #16
+    vmovn.u16       d20, q10       /* d20 = y */
+    vmovn.u16       d21, q11       /* d21 = u */
+    vmovn.u16       d22, q12       /* d22 = v */
+.endm
+
+.macro do_rgb_to_yuv
+    do_rgb_to_yuv_stage1
+    do_rgb_to_yuv_stage2
+.endm
+
+.macro do_rgb_to_yuv_stage2_store_load_stage1
+      vrshrn.u32      d20, q7, #16
+      vrshrn.u32      d21, q8, #16
+      vshrn.u32       d22, q9, #16
+    vrev64.32       q9, q1
+      vshrn.u32       d23, q13, #16
+    vrev64.32       q13, q1
+      vshrn.u32       d24, q14, #16
+      vshrn.u32       d25, q15, #16
+    do_load         \bpp, 8
+      vmovn.u16       d20, q10     /* d20 = y */
+    vmovl.u8        q2, d1\r_offs  /* r = { d4, d5 } */
+      vmovn.u16       d21, q11     /* d21 = u */
+    vmovl.u8        q3, d1\g_offs  /* g = { d6, d7 } */
+      vmovn.u16       d22, q12     /* d22 = v */
+    vmovl.u8        q4, d1\b_offs  /* b = { d8, d9 } */
+    vmull.u16       q7, d4, d00
+    vmlal.u16       q7, d6, d01
+    vmlal.u16       q7, d8, d02
+      vst1.8          {d20}, Y!
+    vmull.u16       q8, d5, d00
+    vmlal.u16       q8, d7, d01
+    vmlal.u16       q8, d9, d02
+    vmlsl.u16       q9, d4, d03
+    vmlsl.u16       q9, d6, d10
+    vmlal.u16       q9, d8, d11
+      vst1.8          {d21}, U!
+    vmlsl.u16       q13, d5, d03
+    vmlsl.u16       q13, d7, d10
+    vmlal.u16       q13, d9, d11
+    vrev64.32       q14, q1
+    vrev64.32       q15, q1
+    vmlal.u16       q14, d4, d11
+    vmlsl.u16       q14, d6, d12
+    vmlsl.u16       q14, d8, d13
+      vst1.8          {d22}, V!
+    vmlal.u16       q15, d5, d11
+    vmlsl.u16       q15, d7, d12
+    vmlsl.u16       q15, d9, d13
+.endm
+
+.balign 16
+jsimd_\colorid\()_ycc_neon_consts:
+  .short 19595, 38470, 7471,  11059
+  .short 21709, 32768, 27439, 5329
+  .short 32767, 128,   32767, 128
+  .short 32767, 128,   32767, 128
+
+asm_function jsimd_\colorid\()_ycc_convert_neon
+    OUTPUT_WIDTH    .req r0
+    INPUT_BUF       .req r1
+    OUTPUT_BUF      .req r2
+    OUTPUT_ROW      .req r3
+    NUM_ROWS        .req r4
+
+    OUTPUT_BUF0     .req r5
+    OUTPUT_BUF1     .req r6
+    OUTPUT_BUF2     .req OUTPUT_BUF
+
+    RGB             .req r7
+    Y               .req r8
+    U               .req r9
+    V               .req r10
+    N               .req ip
+
+    /* Load constants to d0, d1, d2, d3 */
+    adr             ip, jsimd_\colorid\()_ycc_neon_consts
+    vld1.16         {d0, d1, d2, d3}, ip, :128
+
+    /* Save Arm registers and handle input arguments */
+    push            {r4, r5, r6, r7, r8, r9, r10, lr}
+    ldr             NUM_ROWS, sp, #(4 * 8)
+    ldr             OUTPUT_BUF0, OUTPUT_BUF
+    ldr             OUTPUT_BUF1, OUTPUT_BUF, #4
+    ldr             OUTPUT_BUF2, OUTPUT_BUF, #8
+    .unreq          OUTPUT_BUF
+
+    /* Save Neon registers */
+    vpush           {d8 - d15}
+
+    /* Outer loop over scanlines */
+    cmp             NUM_ROWS, #1
+    blt             9f
+0:
+    ldr             Y, OUTPUT_BUF0, OUTPUT_ROW, lsl #2
+    ldr             U, OUTPUT_BUF1, OUTPUT_ROW, lsl #2
+    mov             N, OUTPUT_WIDTH
+    ldr             V, OUTPUT_BUF2, OUTPUT_ROW, lsl #2
+    add             OUTPUT_ROW, OUTPUT_ROW, #1
+    ldr             RGB, INPUT_BUF, #4
+
+    /* Inner loop over pixels */
+    subs            N, N, #8
+    blt             3f
+    do_load         \bpp, 8
+    do_rgb_to_yuv_stage1
+    subs            N, N, #8
+    blt             2f
+1:
+    do_rgb_to_yuv_stage2_store_load_stage1
+    subs            N, N, #8
+    bge             1b
+2:
+    do_rgb_to_yuv_stage2
+    do_store        8
+    tst             N, #7
+    beq             8f
+3:
+    tst             N, #4
+    beq             3f
+    do_load         \bpp, 4
+3:
+    tst             N, #2
+    beq             4f
+    do_load         \bpp, 2
+4:
+    tst             N, #1
+    beq             5f
+    do_load         \bpp, 1
+5:
+    do_rgb_to_yuv
+    tst             N, #4
+    beq             6f
+    do_store        4
+6:
+    tst             N, #2
+    beq             7f
+    do_store        2
+7:
+    tst             N, #1
+    beq             8f
+    do_store        1
+8:
+    subs            NUM_ROWS, NUM_ROWS, #1
+    bgt             0b
+9:
+    /* Restore all registers and return */
+    vpop            {d8 - d15}
+    pop             {r4, r5, r6, r7, r8, r9, r10, pc}
+
+    .unreq          OUTPUT_WIDTH
+    .unreq          OUTPUT_ROW
+    .unreq          INPUT_BUF
+    .unreq          NUM_ROWS
+    .unreq          OUTPUT_BUF0
+    .unreq          OUTPUT_BUF1
+    .unreq          OUTPUT_BUF2
+    .unreq          RGB
+    .unreq          Y
+    .unreq          U
+    .unreq          V
+    .unreq          N
+
+.purgem do_rgb_to_yuv
+.purgem do_rgb_to_yuv_stage1
+.purgem do_rgb_to_yuv_stage2
+.purgem do_rgb_to_yuv_stage2_store_load_stage1
+
+.endm
+
+/*--------------------------------- id ----- bpp R  G  B */
+generate_jsimd_rgb_ycc_convert_neon extrgb,  24, 0, 1, 2
+generate_jsimd_rgb_ycc_convert_neon extbgr,  24, 2, 1, 0
+generate_jsimd_rgb_ycc_convert_neon extrgbx, 32, 0, 1, 2
+generate_jsimd_rgb_ycc_convert_neon extbgrx, 32, 2, 1, 0
+generate_jsimd_rgb_ycc_convert_neon extxbgr, 32, 3, 2, 1
+generate_jsimd_rgb_ycc_convert_neon extxrgb, 32, 1, 2, 3
+
+.purgem do_load
+.purgem do_store

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch64 Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch64/jccolext-neon.c Added

@@ -0,0 +1,316 @@
+/*
+ * jccolext-neon.c - colorspace conversion (64-bit Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jccolor-neon.c */
+
+
+/* RGB -> YCbCr conversion is defined by the following equations:
+ *    Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+ *    Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + 128
+ *    Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B  + 128
+ *
+ * Avoid floating point arithmetic by using shifted integer constants:
+ *    0.29899597 = 19595 * 2^-16
+ *    0.58700561 = 38470 * 2^-16
+ *    0.11399841 =  7471 * 2^-16
+ *    0.16874695 = 11059 * 2^-16
+ *    0.33125305 = 21709 * 2^-16
+ *    0.50000000 = 32768 * 2^-16
+ *    0.41868592 = 27439 * 2^-16
+ *    0.08131409 =  5329 * 2^-16
+ * These constants are defined in jccolor-neon.c
+ *
+ * We add the fixed-point equivalent of 0.5 to Cb and Cr, which effectively
+ * rounds up or down the result via integer truncation.
+ */
+
+void jsimd_rgb_ycc_convert_neon(JDIMENSION image_width, JSAMPARRAY input_buf,
+                                JSAMPIMAGE output_buf, JDIMENSION output_row,
+                                int num_rows)
+{
+  /* Pointer to RGB(X/A) input data */
+  JSAMPROW inptr;
+  /* Pointers to Y, Cb, and Cr output data */
+  JSAMPROW outptr0, outptr1, outptr2;
+  /* Allocate temporary buffer for final (image_width % 16) pixels in row. */
+  ALIGN(16) uint8_t tmp_buf16 * RGB_PIXELSIZE;
+
+  /* Set up conversion constants. */
+  const uint16x8_t consts = vld1q_u16(jsimd_rgb_ycc_neon_consts);
+  const uint32x4_t scaled_128_5 = vdupq_n_u32((128 << 16) + 32767);
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf0output_row;
+    outptr1 = output_buf1output_row;
+    outptr2 = output_buf2output_row;
+    output_row++;
+
+    int cols_remaining = image_width;
+    for (; cols_remaining >= 16; cols_remaining -= 16) {
+
+#if RGB_PIXELSIZE == 4
+      uint8x16x4_t input_pixels = vld4q_u8(inptr);
+#else
+      uint8x16x3_t input_pixels = vld3q_u8(inptr);
+#endif
+      uint16x8_t r_l = vmovl_u8(vget_low_u8(input_pixels.valRGB_RED));
+      uint16x8_t g_l = vmovl_u8(vget_low_u8(input_pixels.valRGB_GREEN));
+      uint16x8_t b_l = vmovl_u8(vget_low_u8(input_pixels.valRGB_BLUE));
+      uint16x8_t r_h = vmovl_u8(vget_high_u8(input_pixels.valRGB_RED));
+      uint16x8_t g_h = vmovl_u8(vget_high_u8(input_pixels.valRGB_GREEN));
+      uint16x8_t b_h = vmovl_u8(vget_high_u8(input_pixels.valRGB_BLUE));
+
+      /* Compute Y = 0.29900 * R + 0.58700 * G + 0.11400 * B */
+      uint32x4_t y_ll = vmull_laneq_u16(vget_low_u16(r_l), consts, 0);
+      y_ll = vmlal_laneq_u16(y_ll, vget_low_u16(g_l), consts, 1);
+      y_ll = vmlal_laneq_u16(y_ll, vget_low_u16(b_l), consts, 2);
+      uint32x4_t y_lh = vmull_laneq_u16(vget_high_u16(r_l), consts, 0);
+      y_lh = vmlal_laneq_u16(y_lh, vget_high_u16(g_l), consts, 1);
+      y_lh = vmlal_laneq_u16(y_lh, vget_high_u16(b_l), consts, 2);
+      uint32x4_t y_hl = vmull_laneq_u16(vget_low_u16(r_h), consts, 0);
+      y_hl = vmlal_laneq_u16(y_hl, vget_low_u16(g_h), consts, 1);
+      y_hl = vmlal_laneq_u16(y_hl, vget_low_u16(b_h), consts, 2);
+      uint32x4_t y_hh = vmull_laneq_u16(vget_high_u16(r_h), consts, 0);
+      y_hh = vmlal_laneq_u16(y_hh, vget_high_u16(g_h), consts, 1);
+      y_hh = vmlal_laneq_u16(y_hh, vget_high_u16(b_h), consts, 2);
+
+      /* Compute Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + 128 */
+      uint32x4_t cb_ll = scaled_128_5;
+      cb_ll = vmlsl_laneq_u16(cb_ll, vget_low_u16(r_l), consts, 3);
+      cb_ll = vmlsl_laneq_u16(cb_ll, vget_low_u16(g_l), consts, 4);
+      cb_ll = vmlal_laneq_u16(cb_ll, vget_low_u16(b_l), consts, 5);
+      uint32x4_t cb_lh = scaled_128_5;
+      cb_lh = vmlsl_laneq_u16(cb_lh, vget_high_u16(r_l), consts, 3);
+      cb_lh = vmlsl_laneq_u16(cb_lh, vget_high_u16(g_l), consts, 4);
+      cb_lh = vmlal_laneq_u16(cb_lh, vget_high_u16(b_l), consts, 5);
+      uint32x4_t cb_hl = scaled_128_5;
+      cb_hl = vmlsl_laneq_u16(cb_hl, vget_low_u16(r_h), consts, 3);
+      cb_hl = vmlsl_laneq_u16(cb_hl, vget_low_u16(g_h), consts, 4);
+      cb_hl = vmlal_laneq_u16(cb_hl, vget_low_u16(b_h), consts, 5);
+      uint32x4_t cb_hh = scaled_128_5;
+      cb_hh = vmlsl_laneq_u16(cb_hh, vget_high_u16(r_h), consts, 3);
+      cb_hh = vmlsl_laneq_u16(cb_hh, vget_high_u16(g_h), consts, 4);
+      cb_hh = vmlal_laneq_u16(cb_hh, vget_high_u16(b_h), consts, 5);
+
+      /* Compute Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B  + 128 */
+      uint32x4_t cr_ll = scaled_128_5;
+      cr_ll = vmlal_laneq_u16(cr_ll, vget_low_u16(r_l), consts, 5);
+      cr_ll = vmlsl_laneq_u16(cr_ll, vget_low_u16(g_l), consts, 6);
+      cr_ll = vmlsl_laneq_u16(cr_ll, vget_low_u16(b_l), consts, 7);
+      uint32x4_t cr_lh = scaled_128_5;
+      cr_lh = vmlal_laneq_u16(cr_lh, vget_high_u16(r_l), consts, 5);
+      cr_lh = vmlsl_laneq_u16(cr_lh, vget_high_u16(g_l), consts, 6);
+      cr_lh = vmlsl_laneq_u16(cr_lh, vget_high_u16(b_l), consts, 7);
+      uint32x4_t cr_hl = scaled_128_5;
+      cr_hl = vmlal_laneq_u16(cr_hl, vget_low_u16(r_h), consts, 5);
+      cr_hl = vmlsl_laneq_u16(cr_hl, vget_low_u16(g_h), consts, 6);
+      cr_hl = vmlsl_laneq_u16(cr_hl, vget_low_u16(b_h), consts, 7);
+      uint32x4_t cr_hh = scaled_128_5;
+      cr_hh = vmlal_laneq_u16(cr_hh, vget_high_u16(r_h), consts, 5);
+      cr_hh = vmlsl_laneq_u16(cr_hh, vget_high_u16(g_h), consts, 6);
+      cr_hh = vmlsl_laneq_u16(cr_hh, vget_high_u16(b_h), consts, 7);
+
+      /* Descale Y values (rounding right shift) and narrow to 16-bit. */
+      uint16x8_t y_l = vcombine_u16(vrshrn_n_u32(y_ll, 16),
+                                    vrshrn_n_u32(y_lh, 16));
+      uint16x8_t y_h = vcombine_u16(vrshrn_n_u32(y_hl, 16),
+                                    vrshrn_n_u32(y_hh, 16));
+      /* Descale Cb values (right shift) and narrow to 16-bit. */
+      uint16x8_t cb_l = vcombine_u16(vshrn_n_u32(cb_ll, 16),
+                                     vshrn_n_u32(cb_lh, 16));
+      uint16x8_t cb_h = vcombine_u16(vshrn_n_u32(cb_hl, 16),
+                                     vshrn_n_u32(cb_hh, 16));
+      /* Descale Cr values (right shift) and narrow to 16-bit. */
+      uint16x8_t cr_l = vcombine_u16(vshrn_n_u32(cr_ll, 16),
+                                     vshrn_n_u32(cr_lh, 16));
+      uint16x8_t cr_h = vcombine_u16(vshrn_n_u32(cr_hl, 16),
+                                     vshrn_n_u32(cr_hh, 16));
+      /* Narrow Y, Cb, and Cr values to 8-bit and store to memory.  Buffer
+       * overwrite is permitted up to the next multiple of ALIGN_SIZE bytes.
+       */
+      vst1q_u8(outptr0, vcombine_u8(vmovn_u16(y_l), vmovn_u16(y_h)));
+      vst1q_u8(outptr1, vcombine_u8(vmovn_u16(cb_l), vmovn_u16(cb_h)));
+      vst1q_u8(outptr2, vcombine_u8(vmovn_u16(cr_l), vmovn_u16(cr_h)));
+
+      /* Increment pointers. */
+      inptr += (16 * RGB_PIXELSIZE);
+      outptr0 += 16;
+      outptr1 += 16;
+      outptr2 += 16;
+    }
+
+    if (cols_remaining > 8) {
+      /* To prevent buffer overread by the vector load instructions, the last
+       * (image_width % 16) columns of data are first memcopied to a temporary
+       * buffer large enough to accommodate the vector load.
+       */
+      memcpy(tmp_buf, inptr, cols_remaining * RGB_PIXELSIZE);
+      inptr = tmp_buf;
+
+#if RGB_PIXELSIZE == 4
+      uint8x16x4_t input_pixels = vld4q_u8(inptr);
+#else
+      uint8x16x3_t input_pixels = vld3q_u8(inptr);
+#endif
+      uint16x8_t r_l = vmovl_u8(vget_low_u8(input_pixels.valRGB_RED));
+      uint16x8_t g_l = vmovl_u8(vget_low_u8(input_pixels.valRGB_GREEN));
+      uint16x8_t b_l = vmovl_u8(vget_low_u8(input_pixels.valRGB_BLUE));
+      uint16x8_t r_h = vmovl_u8(vget_high_u8(input_pixels.valRGB_RED));
+      uint16x8_t g_h = vmovl_u8(vget_high_u8(input_pixels.valRGB_GREEN));
+      uint16x8_t b_h = vmovl_u8(vget_high_u8(input_pixels.valRGB_BLUE));
+
+      /* Compute Y = 0.29900 * R + 0.58700 * G + 0.11400 * B */
+      uint32x4_t y_ll = vmull_laneq_u16(vget_low_u16(r_l), consts, 0);
+      y_ll = vmlal_laneq_u16(y_ll, vget_low_u16(g_l), consts, 1);
+      y_ll = vmlal_laneq_u16(y_ll, vget_low_u16(b_l), consts, 2);
+      uint32x4_t y_lh = vmull_laneq_u16(vget_high_u16(r_l), consts, 0);
+      y_lh = vmlal_laneq_u16(y_lh, vget_high_u16(g_l), consts, 1);
+      y_lh = vmlal_laneq_u16(y_lh, vget_high_u16(b_l), consts, 2);
+      uint32x4_t y_hl = vmull_laneq_u16(vget_low_u16(r_h), consts, 0);
+      y_hl = vmlal_laneq_u16(y_hl, vget_low_u16(g_h), consts, 1);
+      y_hl = vmlal_laneq_u16(y_hl, vget_low_u16(b_h), consts, 2);
+      uint32x4_t y_hh = vmull_laneq_u16(vget_high_u16(r_h), consts, 0);
+      y_hh = vmlal_laneq_u16(y_hh, vget_high_u16(g_h), consts, 1);
+      y_hh = vmlal_laneq_u16(y_hh, vget_high_u16(b_h), consts, 2);
+
+      /* Compute Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + 128 */
+      uint32x4_t cb_ll = scaled_128_5;
+      cb_ll = vmlsl_laneq_u16(cb_ll, vget_low_u16(r_l), consts, 3);
+      cb_ll = vmlsl_laneq_u16(cb_ll, vget_low_u16(g_l), consts, 4);
+      cb_ll = vmlal_laneq_u16(cb_ll, vget_low_u16(b_l), consts, 5);
+      uint32x4_t cb_lh = scaled_128_5;
+      cb_lh = vmlsl_laneq_u16(cb_lh, vget_high_u16(r_l), consts, 3);
+      cb_lh = vmlsl_laneq_u16(cb_lh, vget_high_u16(g_l), consts, 4);
+      cb_lh = vmlal_laneq_u16(cb_lh, vget_high_u16(b_l), consts, 5);
+      uint32x4_t cb_hl = scaled_128_5;
+      cb_hl = vmlsl_laneq_u16(cb_hl, vget_low_u16(r_h), consts, 3);
+      cb_hl = vmlsl_laneq_u16(cb_hl, vget_low_u16(g_h), consts, 4);
+      cb_hl = vmlal_laneq_u16(cb_hl, vget_low_u16(b_h), consts, 5);
+      uint32x4_t cb_hh = scaled_128_5;
+      cb_hh = vmlsl_laneq_u16(cb_hh, vget_high_u16(r_h), consts, 3);
+      cb_hh = vmlsl_laneq_u16(cb_hh, vget_high_u16(g_h), consts, 4);
+      cb_hh = vmlal_laneq_u16(cb_hh, vget_high_u16(b_h), consts, 5);
+
+      /* Compute Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B  + 128 */
+      uint32x4_t cr_ll = scaled_128_5;
+      cr_ll = vmlal_laneq_u16(cr_ll, vget_low_u16(r_l), consts, 5);
+      cr_ll = vmlsl_laneq_u16(cr_ll, vget_low_u16(g_l), consts, 6);
+      cr_ll = vmlsl_laneq_u16(cr_ll, vget_low_u16(b_l), consts, 7);
+      uint32x4_t cr_lh = scaled_128_5;
+      cr_lh = vmlal_laneq_u16(cr_lh, vget_high_u16(r_l), consts, 5);
+      cr_lh = vmlsl_laneq_u16(cr_lh, vget_high_u16(g_l), consts, 6);
+      cr_lh = vmlsl_laneq_u16(cr_lh, vget_high_u16(b_l), consts, 7);
+      uint32x4_t cr_hl = scaled_128_5;
+      cr_hl = vmlal_laneq_u16(cr_hl, vget_low_u16(r_h), consts, 5);
+      cr_hl = vmlsl_laneq_u16(cr_hl, vget_low_u16(g_h), consts, 6);
+      cr_hl = vmlsl_laneq_u16(cr_hl, vget_low_u16(b_h), consts, 7);
+      uint32x4_t cr_hh = scaled_128_5;
+      cr_hh = vmlal_laneq_u16(cr_hh, vget_high_u16(r_h), consts, 5);
+      cr_hh = vmlsl_laneq_u16(cr_hh, vget_high_u16(g_h), consts, 6);
+      cr_hh = vmlsl_laneq_u16(cr_hh, vget_high_u16(b_h), consts, 7);
+
+      /* Descale Y values (rounding right shift) and narrow to 16-bit. */
+      uint16x8_t y_l = vcombine_u16(vrshrn_n_u32(y_ll, 16),
+                                    vrshrn_n_u32(y_lh, 16));
+      uint16x8_t y_h = vcombine_u16(vrshrn_n_u32(y_hl, 16),
+                                    vrshrn_n_u32(y_hh, 16));
+      /* Descale Cb values (right shift) and narrow to 16-bit. */
+      uint16x8_t cb_l = vcombine_u16(vshrn_n_u32(cb_ll, 16),
+                                     vshrn_n_u32(cb_lh, 16));
+      uint16x8_t cb_h = vcombine_u16(vshrn_n_u32(cb_hl, 16),
+                                     vshrn_n_u32(cb_hh, 16));
+      /* Descale Cr values (right shift) and narrow to 16-bit. */
+      uint16x8_t cr_l = vcombine_u16(vshrn_n_u32(cr_ll, 16),
+                                     vshrn_n_u32(cr_lh, 16));
+      uint16x8_t cr_h = vcombine_u16(vshrn_n_u32(cr_hl, 16),
+                                     vshrn_n_u32(cr_hh, 16));
+      /* Narrow Y, Cb, and Cr values to 8-bit and store to memory.  Buffer
+       * overwrite is permitted up to the next multiple of ALIGN_SIZE bytes.
+       */
+      vst1q_u8(outptr0, vcombine_u8(vmovn_u16(y_l), vmovn_u16(y_h)));
+      vst1q_u8(outptr1, vcombine_u8(vmovn_u16(cb_l), vmovn_u16(cb_h)));
+      vst1q_u8(outptr2, vcombine_u8(vmovn_u16(cr_l), vmovn_u16(cr_h)));
+
+    } else if (cols_remaining > 0) {
+      /* To prevent buffer overread by the vector load instructions, the last
+       * (image_width % 8) columns of data are first memcopied to a temporary
+       * buffer large enough to accommodate the vector load.
+       */
+      memcpy(tmp_buf, inptr, cols_remaining * RGB_PIXELSIZE);
+      inptr = tmp_buf;
+
+#if RGB_PIXELSIZE == 4
+      uint8x8x4_t input_pixels = vld4_u8(inptr);
+#else
+      uint8x8x3_t input_pixels = vld3_u8(inptr);
+#endif
+      uint16x8_t r = vmovl_u8(input_pixels.valRGB_RED);
+      uint16x8_t g = vmovl_u8(input_pixels.valRGB_GREEN);
+      uint16x8_t b = vmovl_u8(input_pixels.valRGB_BLUE);
+
+      /* Compute Y = 0.29900 * R + 0.58700 * G + 0.11400 * B */
+      uint32x4_t y_l = vmull_laneq_u16(vget_low_u16(r), consts, 0);
+      y_l = vmlal_laneq_u16(y_l, vget_low_u16(g), consts, 1);
+      y_l = vmlal_laneq_u16(y_l, vget_low_u16(b), consts, 2);
+      uint32x4_t y_h = vmull_laneq_u16(vget_high_u16(r), consts, 0);
+      y_h = vmlal_laneq_u16(y_h, vget_high_u16(g), consts, 1);
+      y_h = vmlal_laneq_u16(y_h, vget_high_u16(b), consts, 2);
+
+      /* Compute Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + 128 */
+      uint32x4_t cb_l = scaled_128_5;
+      cb_l = vmlsl_laneq_u16(cb_l, vget_low_u16(r), consts, 3);
+      cb_l = vmlsl_laneq_u16(cb_l, vget_low_u16(g), consts, 4);
+      cb_l = vmlal_laneq_u16(cb_l, vget_low_u16(b), consts, 5);
+      uint32x4_t cb_h = scaled_128_5;
+      cb_h = vmlsl_laneq_u16(cb_h, vget_high_u16(r), consts, 3);
+      cb_h = vmlsl_laneq_u16(cb_h, vget_high_u16(g), consts, 4);
+      cb_h = vmlal_laneq_u16(cb_h, vget_high_u16(b), consts, 5);
+
+      /* Compute Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B  + 128 */
+      uint32x4_t cr_l = scaled_128_5;
+      cr_l = vmlal_laneq_u16(cr_l, vget_low_u16(r), consts, 5);
+      cr_l = vmlsl_laneq_u16(cr_l, vget_low_u16(g), consts, 6);
+      cr_l = vmlsl_laneq_u16(cr_l, vget_low_u16(b), consts, 7);
+      uint32x4_t cr_h = scaled_128_5;
+      cr_h = vmlal_laneq_u16(cr_h, vget_high_u16(r), consts, 5);
+      cr_h = vmlsl_laneq_u16(cr_h, vget_high_u16(g), consts, 6);
+      cr_h = vmlsl_laneq_u16(cr_h, vget_high_u16(b), consts, 7);
+
+      /* Descale Y values (rounding right shift) and narrow to 16-bit. */
+      uint16x8_t y_u16 = vcombine_u16(vrshrn_n_u32(y_l, 16),
+                                      vrshrn_n_u32(y_h, 16));
+      /* Descale Cb values (right shift) and narrow to 16-bit. */
+      uint16x8_t cb_u16 = vcombine_u16(vshrn_n_u32(cb_l, 16),
+                                       vshrn_n_u32(cb_h, 16));
+      /* Descale Cr values (right shift) and narrow to 16-bit. */
+      uint16x8_t cr_u16 = vcombine_u16(vshrn_n_u32(cr_l, 16),
+                                       vshrn_n_u32(cr_h, 16));
+      /* Narrow Y, Cb, and Cr values to 8-bit and store to memory.  Buffer
+       * overwrite is permitted up to the next multiple of ALIGN_SIZE bytes.
+       */
+      vst1_u8(outptr0, vmovn_u16(y_u16));
+      vst1_u8(outptr1, vmovn_u16(cb_u16));
+      vst1_u8(outptr2, vmovn_u16(cr_u16));
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch64/jchuff-neon.c Added

@@ -0,0 +1,403 @@
+/*
+ * jchuff-neon.c - Huffman entropy encoding (64-bit Arm Neon)
+ *
+ * Copyright (C) 2020-2021, Arm Limited.  All Rights Reserved.
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ *
+ * NOTE: All referenced figures are from
+ * Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+ */
+
+#define JPEG_INTERNALS
+#include "../../../jinclude.h"
+#include "../../../jpeglib.h"
+#include "../../../jsimd.h"
+#include "../../../jdct.h"
+#include "../../../jsimddct.h"
+#include "../../jsimd.h"
+#include "../align.h"
+#include "../jchuff.h"
+#include "neon-compat.h"
+
+#include <limits.h>
+
+#include <arm_neon.h>
+
+
+ALIGN(16) static const uint8_t jsimd_huff_encode_one_block_consts = {
+    0,   1,   2,   3,  16,  17,  32,  33,
+   18,  19,   4,   5,   6,   7,  20,  21,
+   34,  35,  48,  49, 255, 255,  50,  51,
+   36,  37,  22,  23,   8,   9,  10,  11,
+  255, 255,   6,   7,  20,  21,  34,  35,
+   48,  49, 255, 255,  50,  51,  36,  37,
+   54,  55,  40,  41,  26,  27,  12,  13,
+   14,  15,  28,  29,  42,  43,  56,  57,
+    6,   7,  20,  21,  34,  35,  48,  49,
+   50,  51,  36,  37,  22,  23,   8,   9,
+   26,  27,  12,  13, 255, 255,  14,  15,
+   28,  29,  42,  43,  56,  57, 255, 255,
+   52,  53,  54,  55,  40,  41,  26,  27,
+   12,  13, 255, 255,  14,  15,  28,  29,
+   26,  27,  40,  41,  42,  43,  28,  29,
+   14,  15,  30,  31,  44,  45,  46,  47
+};
+
+JOCTET *jsimd_huff_encode_one_block_neon(void *state, JOCTET *buffer,
+                                         JCOEFPTR block, int last_dc_val,
+                                         c_derived_tbl *dctbl,
+                                         c_derived_tbl *actbl)
+{
+  uint16_t block_diffDCTSIZE2;
+
+  /* Load lookup table indices for rows of zig-zag ordering. */
+#ifdef HAVE_VLD1Q_U8_X4
+  const uint8x16x4_t idx_rows_0123 =
+    vld1q_u8_x4(jsimd_huff_encode_one_block_consts + 0 * DCTSIZE);
+  const uint8x16x4_t idx_rows_4567 =
+    vld1q_u8_x4(jsimd_huff_encode_one_block_consts + 8 * DCTSIZE);
+#else
+  /* GCC does not currently support intrinsics vl1dq_<type>_x4(). */
+  const uint8x16x4_t idx_rows_0123 = { {
+    vld1q_u8(jsimd_huff_encode_one_block_consts + 0 * DCTSIZE),
+    vld1q_u8(jsimd_huff_encode_one_block_consts + 2 * DCTSIZE),
+    vld1q_u8(jsimd_huff_encode_one_block_consts + 4 * DCTSIZE),
+    vld1q_u8(jsimd_huff_encode_one_block_consts + 6 * DCTSIZE)
+  } };
+  const uint8x16x4_t idx_rows_4567 = { {
+    vld1q_u8(jsimd_huff_encode_one_block_consts + 8 * DCTSIZE),
+    vld1q_u8(jsimd_huff_encode_one_block_consts + 10 * DCTSIZE),
+    vld1q_u8(jsimd_huff_encode_one_block_consts + 12 * DCTSIZE),
+    vld1q_u8(jsimd_huff_encode_one_block_consts + 14 * DCTSIZE)
+  } };
+#endif
+
+  /* Load 8x8 block of DCT coefficients. */
+#ifdef HAVE_VLD1Q_U8_X4
+  const int8x16x4_t tbl_rows_0123 =
+    vld1q_s8_x4((int8_t *)(block + 0 * DCTSIZE));
+  const int8x16x4_t tbl_rows_4567 =
+    vld1q_s8_x4((int8_t *)(block + 4 * DCTSIZE));
+#else
+  const int8x16x4_t tbl_rows_0123 = { {
+    vld1q_s8((int8_t *)(block + 0 * DCTSIZE)),
+    vld1q_s8((int8_t *)(block + 1 * DCTSIZE)),
+    vld1q_s8((int8_t *)(block + 2 * DCTSIZE)),
+    vld1q_s8((int8_t *)(block + 3 * DCTSIZE))
+  } };
+  const int8x16x4_t tbl_rows_4567 = { {
+    vld1q_s8((int8_t *)(block + 4 * DCTSIZE)),
+    vld1q_s8((int8_t *)(block + 5 * DCTSIZE)),
+    vld1q_s8((int8_t *)(block + 6 * DCTSIZE)),
+    vld1q_s8((int8_t *)(block + 7 * DCTSIZE))
+  } };
+#endif
+
+  /* Initialise extra lookup tables. */
+  const int8x16x4_t tbl_rows_2345 = { {
+    tbl_rows_0123.val2, tbl_rows_0123.val3,
+    tbl_rows_4567.val0, tbl_rows_4567.val1
+  } };
+  const int8x16x3_t tbl_rows_567 =
+    { { tbl_rows_4567.val1, tbl_rows_4567.val2, tbl_rows_4567.val3 } };
+
+  /* Shuffle coefficients into zig-zag order. */
+  int16x8_t row0 =
+    vreinterpretq_s16_s8(vqtbl4q_s8(tbl_rows_0123, idx_rows_0123.val0));
+  int16x8_t row1 =
+    vreinterpretq_s16_s8(vqtbl4q_s8(tbl_rows_0123, idx_rows_0123.val1));
+  int16x8_t row2 =
+    vreinterpretq_s16_s8(vqtbl4q_s8(tbl_rows_2345, idx_rows_0123.val2));
+  int16x8_t row3 =
+    vreinterpretq_s16_s8(vqtbl4q_s8(tbl_rows_0123, idx_rows_0123.val3));
+  int16x8_t row4 =
+    vreinterpretq_s16_s8(vqtbl4q_s8(tbl_rows_4567, idx_rows_4567.val0));
+  int16x8_t row5 =
+    vreinterpretq_s16_s8(vqtbl4q_s8(tbl_rows_2345, idx_rows_4567.val1));
+  int16x8_t row6 =
+    vreinterpretq_s16_s8(vqtbl4q_s8(tbl_rows_4567, idx_rows_4567.val2));
+  int16x8_t row7 =
+    vreinterpretq_s16_s8(vqtbl3q_s8(tbl_rows_567, idx_rows_4567.val3));
+
+  /* Compute DC coefficient difference value (F.1.1.5.1). */
+  row0 = vsetq_lane_s16(block0 - last_dc_val, row0, 0);
+  /* Initialize AC coefficient lanes not reachable by lookup tables. */
+  row1 =
+    vsetq_lane_s16(vgetq_lane_s16(vreinterpretq_s16_s8(tbl_rows_4567.val0),
+                                  0), row1, 2);
+  row2 =
+    vsetq_lane_s16(vgetq_lane_s16(vreinterpretq_s16_s8(tbl_rows_0123.val1),
+                                  4), row2, 0);
+  row2 =
+    vsetq_lane_s16(vgetq_lane_s16(vreinterpretq_s16_s8(tbl_rows_4567.val2),
+                                  0), row2, 5);
+  row5 =
+    vsetq_lane_s16(vgetq_lane_s16(vreinterpretq_s16_s8(tbl_rows_0123.val1),
+                                  7), row5, 2);
+  row5 =
+    vsetq_lane_s16(vgetq_lane_s16(vreinterpretq_s16_s8(tbl_rows_4567.val2),
+                                  3), row5, 7);
+  row6 =
+    vsetq_lane_s16(vgetq_lane_s16(vreinterpretq_s16_s8(tbl_rows_0123.val3),
+                                  7), row6, 5);
+
+  /* DCT block is now in zig-zag order; start Huffman encoding process. */
+  int16x8_t abs_row0 = vabsq_s16(row0);
+  int16x8_t abs_row1 = vabsq_s16(row1);
+  int16x8_t abs_row2 = vabsq_s16(row2);
+  int16x8_t abs_row3 = vabsq_s16(row3);
+  int16x8_t abs_row4 = vabsq_s16(row4);
+  int16x8_t abs_row5 = vabsq_s16(row5);
+  int16x8_t abs_row6 = vabsq_s16(row6);
+  int16x8_t abs_row7 = vabsq_s16(row7);
+
+  /* For negative coeffs: diff = abs(coeff) -1 = ~abs(coeff) */
+  uint16x8_t row0_diff =
+    vreinterpretq_u16_s16(veorq_s16(abs_row0, vshrq_n_s16(row0, 15)));
+  uint16x8_t row1_diff =
+    vreinterpretq_u16_s16(veorq_s16(abs_row1, vshrq_n_s16(row1, 15)));
+  uint16x8_t row2_diff =
+    vreinterpretq_u16_s16(veorq_s16(abs_row2, vshrq_n_s16(row2, 15)));
+  uint16x8_t row3_diff =
+    vreinterpretq_u16_s16(veorq_s16(abs_row3, vshrq_n_s16(row3, 15)));
+  uint16x8_t row4_diff =
+    vreinterpretq_u16_s16(veorq_s16(abs_row4, vshrq_n_s16(row4, 15)));
+  uint16x8_t row5_diff =
+    vreinterpretq_u16_s16(veorq_s16(abs_row5, vshrq_n_s16(row5, 15)));
+  uint16x8_t row6_diff =
+    vreinterpretq_u16_s16(veorq_s16(abs_row6, vshrq_n_s16(row6, 15)));
+  uint16x8_t row7_diff =
+    vreinterpretq_u16_s16(veorq_s16(abs_row7, vshrq_n_s16(row7, 15)));
+
+  /* Construct bitmap to accelerate encoding of AC coefficients.  A set bit
+   * means that the corresponding coefficient != 0.
+   */
+  uint8x8_t abs_row0_gt0 = vmovn_u16(vcgtq_u16(vreinterpretq_u16_s16(abs_row0),
+                                               vdupq_n_u16(0)));
+  uint8x8_t abs_row1_gt0 = vmovn_u16(vcgtq_u16(vreinterpretq_u16_s16(abs_row1),
+                                               vdupq_n_u16(0)));
+  uint8x8_t abs_row2_gt0 = vmovn_u16(vcgtq_u16(vreinterpretq_u16_s16(abs_row2),
+                                               vdupq_n_u16(0)));
+  uint8x8_t abs_row3_gt0 = vmovn_u16(vcgtq_u16(vreinterpretq_u16_s16(abs_row3),
+                                               vdupq_n_u16(0)));
+  uint8x8_t abs_row4_gt0 = vmovn_u16(vcgtq_u16(vreinterpretq_u16_s16(abs_row4),
+                                               vdupq_n_u16(0)));
+  uint8x8_t abs_row5_gt0 = vmovn_u16(vcgtq_u16(vreinterpretq_u16_s16(abs_row5),
+                                               vdupq_n_u16(0)));
+  uint8x8_t abs_row6_gt0 = vmovn_u16(vcgtq_u16(vreinterpretq_u16_s16(abs_row6),
+                                               vdupq_n_u16(0)));
+  uint8x8_t abs_row7_gt0 = vmovn_u16(vcgtq_u16(vreinterpretq_u16_s16(abs_row7),
+                                               vdupq_n_u16(0)));
+
+  /* { 0x80, 0x40, 0x20, 0x10, 0x08, 0x04, 0x02, 0x01 } */
+  const uint8x8_t bitmap_mask =
+    vreinterpret_u8_u64(vmov_n_u64(0x0102040810204080));
+
+  abs_row0_gt0 = vand_u8(abs_row0_gt0, bitmap_mask);
+  abs_row1_gt0 = vand_u8(abs_row1_gt0, bitmap_mask);
+  abs_row2_gt0 = vand_u8(abs_row2_gt0, bitmap_mask);
+  abs_row3_gt0 = vand_u8(abs_row3_gt0, bitmap_mask);
+  abs_row4_gt0 = vand_u8(abs_row4_gt0, bitmap_mask);
+  abs_row5_gt0 = vand_u8(abs_row5_gt0, bitmap_mask);
+  abs_row6_gt0 = vand_u8(abs_row6_gt0, bitmap_mask);
+  abs_row7_gt0 = vand_u8(abs_row7_gt0, bitmap_mask);
+
+  uint8x8_t bitmap_rows_10 = vpadd_u8(abs_row1_gt0, abs_row0_gt0);
+  uint8x8_t bitmap_rows_32 = vpadd_u8(abs_row3_gt0, abs_row2_gt0);
+  uint8x8_t bitmap_rows_54 = vpadd_u8(abs_row5_gt0, abs_row4_gt0);
+  uint8x8_t bitmap_rows_76 = vpadd_u8(abs_row7_gt0, abs_row6_gt0);
+  uint8x8_t bitmap_rows_3210 = vpadd_u8(bitmap_rows_32, bitmap_rows_10);
+  uint8x8_t bitmap_rows_7654 = vpadd_u8(bitmap_rows_76, bitmap_rows_54);
+  uint8x8_t bitmap_all = vpadd_u8(bitmap_rows_7654, bitmap_rows_3210);
+
+  /* Shift left to remove DC bit. */
+  bitmap_all =
+    vreinterpret_u8_u64(vshl_n_u64(vreinterpret_u64_u8(bitmap_all), 1));
+  /* Count bits set (number of non-zero coefficients) in bitmap. */
+  unsigned int non_zero_coefficients = vaddv_u8(vcnt_u8(bitmap_all));
+  /* Move bitmap to 64-bit scalar register. */
+  uint64_t bitmap = vget_lane_u64(vreinterpret_u64_u8(bitmap_all), 0);
+
+  /* Set up state and bit buffer for output bitstream. */
+  working_state *state_ptr = (working_state *)state;
+  int free_bits = state_ptr->cur.free_bits;
+  size_t put_buffer = state_ptr->cur.put_buffer;
+
+  /* Encode DC coefficient. */
+
+  /* Find nbits required to specify sign and amplitude of coefficient. */
+#if defined(_MSC_VER) && !defined(__clang__)
+  unsigned int lz = BUILTIN_CLZ(vgetq_lane_s16(abs_row0, 0));
+#else
+  unsigned int lz;
+  __asm__("clz %w0, %w1" : "=r"(lz) : "r"(vgetq_lane_s16(abs_row0, 0)));
+#endif
+  unsigned int nbits = 32 - lz;
+  /* Emit Huffman-coded symbol and additional diff bits. */
+  unsigned int diff = (unsigned int)(vgetq_lane_u16(row0_diff, 0) << lz) >> lz;
+  PUT_CODE(dctbl->ehufconbits, dctbl->ehufsinbits, diff)
+
+  /* Encode AC coefficients. */
+
+  unsigned int r = 0;  /* r = run length of zeros */
+  unsigned int i = 1;  /* i = number of coefficients encoded */
+  /* Code and size information for a run length of 16 zero coefficients */
+  const unsigned int code_0xf0 = actbl->ehufco0xf0;
+  const unsigned int size_0xf0 = actbl->ehufsi0xf0;
+
+  /* The most efficient method of computing nbits and diff depends on the
+   * number of non-zero coefficients.  If the bitmap is not too sparse (> 8
+   * non-zero AC coefficients), it is beneficial to use Neon; else we compute
+   * nbits and diff on demand using scalar code.
+   */
+  if (non_zero_coefficients > 8) {
+    uint8_t block_nbitsDCTSIZE2;
+
+    int16x8_t row0_lz = vclzq_s16(abs_row0);
+    int16x8_t row1_lz = vclzq_s16(abs_row1);
+    int16x8_t row2_lz = vclzq_s16(abs_row2);
+    int16x8_t row3_lz = vclzq_s16(abs_row3);
+    int16x8_t row4_lz = vclzq_s16(abs_row4);
+    int16x8_t row5_lz = vclzq_s16(abs_row5);
+    int16x8_t row6_lz = vclzq_s16(abs_row6);
+    int16x8_t row7_lz = vclzq_s16(abs_row7);
+    /* Compute nbits needed to specify magnitude of each coefficient. */
+    uint8x8_t row0_nbits = vsub_u8(vdup_n_u8(16),
+                                   vmovn_u16(vreinterpretq_u16_s16(row0_lz)));
+    uint8x8_t row1_nbits = vsub_u8(vdup_n_u8(16),
+                                   vmovn_u16(vreinterpretq_u16_s16(row1_lz)));
+    uint8x8_t row2_nbits = vsub_u8(vdup_n_u8(16),
+                                   vmovn_u16(vreinterpretq_u16_s16(row2_lz)));
+    uint8x8_t row3_nbits = vsub_u8(vdup_n_u8(16),
+                                   vmovn_u16(vreinterpretq_u16_s16(row3_lz)));
+    uint8x8_t row4_nbits = vsub_u8(vdup_n_u8(16),
+                                   vmovn_u16(vreinterpretq_u16_s16(row4_lz)));
+    uint8x8_t row5_nbits = vsub_u8(vdup_n_u8(16),
+                                   vmovn_u16(vreinterpretq_u16_s16(row5_lz)));
+    uint8x8_t row6_nbits = vsub_u8(vdup_n_u8(16),
+                                   vmovn_u16(vreinterpretq_u16_s16(row6_lz)));
+    uint8x8_t row7_nbits = vsub_u8(vdup_n_u8(16),
+                                   vmovn_u16(vreinterpretq_u16_s16(row7_lz)));
+    /* Store nbits. */
+    vst1_u8(block_nbits + 0 * DCTSIZE, row0_nbits);
+    vst1_u8(block_nbits + 1 * DCTSIZE, row1_nbits);
+    vst1_u8(block_nbits + 2 * DCTSIZE, row2_nbits);
+    vst1_u8(block_nbits + 3 * DCTSIZE, row3_nbits);
+    vst1_u8(block_nbits + 4 * DCTSIZE, row4_nbits);
+    vst1_u8(block_nbits + 5 * DCTSIZE, row5_nbits);
+    vst1_u8(block_nbits + 6 * DCTSIZE, row6_nbits);
+    vst1_u8(block_nbits + 7 * DCTSIZE, row7_nbits);
+    /* Mask bits not required to specify sign and amplitude of diff. */
+    row0_diff = vshlq_u16(row0_diff, row0_lz);
+    row1_diff = vshlq_u16(row1_diff, row1_lz);
+    row2_diff = vshlq_u16(row2_diff, row2_lz);
+    row3_diff = vshlq_u16(row3_diff, row3_lz);
+    row4_diff = vshlq_u16(row4_diff, row4_lz);
+    row5_diff = vshlq_u16(row5_diff, row5_lz);
+    row6_diff = vshlq_u16(row6_diff, row6_lz);
+    row7_diff = vshlq_u16(row7_diff, row7_lz);
+    row0_diff = vshlq_u16(row0_diff, vnegq_s16(row0_lz));
+    row1_diff = vshlq_u16(row1_diff, vnegq_s16(row1_lz));
+    row2_diff = vshlq_u16(row2_diff, vnegq_s16(row2_lz));
+    row3_diff = vshlq_u16(row3_diff, vnegq_s16(row3_lz));
+    row4_diff = vshlq_u16(row4_diff, vnegq_s16(row4_lz));
+    row5_diff = vshlq_u16(row5_diff, vnegq_s16(row5_lz));
+    row6_diff = vshlq_u16(row6_diff, vnegq_s16(row6_lz));
+    row7_diff = vshlq_u16(row7_diff, vnegq_s16(row7_lz));
+    /* Store diff bits. */
+    vst1q_u16(block_diff + 0 * DCTSIZE, row0_diff);
+    vst1q_u16(block_diff + 1 * DCTSIZE, row1_diff);
+    vst1q_u16(block_diff + 2 * DCTSIZE, row2_diff);
+    vst1q_u16(block_diff + 3 * DCTSIZE, row3_diff);
+    vst1q_u16(block_diff + 4 * DCTSIZE, row4_diff);
+    vst1q_u16(block_diff + 5 * DCTSIZE, row5_diff);
+    vst1q_u16(block_diff + 6 * DCTSIZE, row6_diff);
+    vst1q_u16(block_diff + 7 * DCTSIZE, row7_diff);
+
+    while (bitmap != 0) {
+      r = BUILTIN_CLZLL(bitmap);
+      i += r;
+      bitmap <<= r;
+      nbits = block_nbitsi;
+      diff = block_diffi;
+      while (r > 15) {
+        /* If run length > 15, emit special run-length-16 codes. */
+        PUT_BITS(code_0xf0, size_0xf0)
+        r -= 16;
+      }
+      /* Emit Huffman symbol for run length / number of bits. (F.1.2.2.1) */
+      unsigned int rs = (r << 4) + nbits;
+      PUT_CODE(actbl->ehufcors, actbl->ehufsirs, diff)
+      i++;
+      bitmap <<= 1;
+    }
+  } else if (bitmap != 0) {
+    uint16_t block_absDCTSIZE2;
+    /* Store absolute value of coefficients. */
+    vst1q_u16(block_abs + 0 * DCTSIZE, vreinterpretq_u16_s16(abs_row0));
+    vst1q_u16(block_abs + 1 * DCTSIZE, vreinterpretq_u16_s16(abs_row1));
+    vst1q_u16(block_abs + 2 * DCTSIZE, vreinterpretq_u16_s16(abs_row2));
+    vst1q_u16(block_abs + 3 * DCTSIZE, vreinterpretq_u16_s16(abs_row3));
+    vst1q_u16(block_abs + 4 * DCTSIZE, vreinterpretq_u16_s16(abs_row4));
+    vst1q_u16(block_abs + 5 * DCTSIZE, vreinterpretq_u16_s16(abs_row5));
+    vst1q_u16(block_abs + 6 * DCTSIZE, vreinterpretq_u16_s16(abs_row6));
+    vst1q_u16(block_abs + 7 * DCTSIZE, vreinterpretq_u16_s16(abs_row7));
+    /* Store diff bits. */
+    vst1q_u16(block_diff + 0 * DCTSIZE, row0_diff);
+    vst1q_u16(block_diff + 1 * DCTSIZE, row1_diff);
+    vst1q_u16(block_diff + 2 * DCTSIZE, row2_diff);
+    vst1q_u16(block_diff + 3 * DCTSIZE, row3_diff);
+    vst1q_u16(block_diff + 4 * DCTSIZE, row4_diff);
+    vst1q_u16(block_diff + 5 * DCTSIZE, row5_diff);
+    vst1q_u16(block_diff + 6 * DCTSIZE, row6_diff);
+    vst1q_u16(block_diff + 7 * DCTSIZE, row7_diff);
+
+    /* Same as above but must mask diff bits and compute nbits on demand. */
+    while (bitmap != 0) {
+      r = BUILTIN_CLZLL(bitmap);
+      i += r;
+      bitmap <<= r;
+      lz = BUILTIN_CLZ(block_absi);
+      nbits = 32 - lz;
+      diff = (unsigned int)(block_diffi << lz) >> lz;
+      while (r > 15) {
+        /* If run length > 15, emit special run-length-16 codes. */
+        PUT_BITS(code_0xf0, size_0xf0)
+        r -= 16;
+      }
+      /* Emit Huffman symbol for run length / number of bits. (F.1.2.2.1) */
+      unsigned int rs = (r << 4) + nbits;
+      PUT_CODE(actbl->ehufcors, actbl->ehufsirs, diff)
+      i++;
+      bitmap <<= 1;
+    }
+  }
+
+  /* If the last coefficient(s) were zero, emit an end-of-block (EOB) code.
+   * The value of RS for the EOB code is 0.
+   */
+  if (i != 64) {
+    PUT_BITS(actbl->ehufco0, actbl->ehufsi0)
+  }
+
+  state_ptr->cur.put_buffer = put_buffer;
+  state_ptr->cur.free_bits = free_bits;
+
+  return buffer;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch64/jsimd.c Added

@@ -0,0 +1,1063 @@
+/*
+ * jsimd_arm64.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2011, Nokia Corporation and/or its subsidiary(-ies).
+ * Copyright (C) 2009-2011, 2013-2014, 2016, 2018, 2020, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ * Copyright (C) 2020, Arm Limited.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 64-bit Arm architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../../../jinclude.h"
+#include "../../../jpeglib.h"
+#include "../../../jsimd.h"
+#include "../../../jdct.h"
+#include "../../../jsimddct.h"
+#include "../../jsimd.h"
+#include "jconfigint.h"
+
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+
+#define JSIMD_FASTLD3  1
+#define JSIMD_FASTST3  2
+#define JSIMD_FASTTBL  4
+
+static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
+static unsigned int simd_features = JSIMD_FASTLD3 | JSIMD_FASTST3 |
+                                    JSIMD_FASTTBL;
+
+#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+
+#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT  (1024 * 1024)
+
+LOCAL(int)
+check_cpuinfo(char *buffer, const char *field, char *value)
+{
+  char *p;
+
+  if (*value == 0)
+    return 0;
+  if (strncmp(buffer, field, strlen(field)) != 0)
+    return 0;
+  buffer += strlen(field);
+  while (isspace(*buffer))
+    buffer++;
+
+  /* Check if 'value' is present in the buffer as a separate word */
+  while ((p = strstr(buffer, value))) {
+    if (p > buffer && !isspace(*(p - 1))) {
+      buffer++;
+      continue;
+    }
+    p += strlen(value);
+    if (*p != 0 && !isspace(*p)) {
+      buffer++;
+      continue;
+    }
+    return 1;
+  }
+  return 0;
+}
+
+LOCAL(int)
+parse_proc_cpuinfo(int bufsize)
+{
+  char *buffer = (char *)malloc(bufsize);
+  FILE *fd;
+
+  if (!buffer)
+    return 0;
+
+  fd = fopen("/proc/cpuinfo", "r");
+  if (fd) {
+    while (fgets(buffer, bufsize, fd)) {
+      if (!strchr(buffer, '\n') && !feof(fd)) {
+        /* "impossible" happened - insufficient size of the buffer! */
+        fclose(fd);
+        free(buffer);
+        return 0;
+      }
+      if (check_cpuinfo(buffer, "CPU part", "0xd03") ||
+          check_cpuinfo(buffer, "CPU part", "0xd07"))
+        /* The Cortex-A53 has a slow tbl implementation.  We can gain a few
+           percent speedup by disabling the use of that instruction.  The
+           speedup on Cortex-A57 is more subtle but still measurable. */
+        simd_features &= ~JSIMD_FASTTBL;
+      else if (check_cpuinfo(buffer, "CPU part", "0x0a1"))
+        /* The SIMD version of Huffman encoding is slower than the C version on
+           Cavium ThunderX.  Also, ld3 and st3 are abyssmally slow on that
+           CPU. */
+        simd_huffman = simd_features = 0;
+    }
+    fclose(fd);
+  }
+  free(buffer);
+  return 1;
+}
+
+#endif
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+
+/*
+ * Armv8 architectures support Neon extensions by default.
+ * It is no longer optional as it was with Armv7.
+ */
+
+
+LOCAL(void)
+init_simd(void)
+{
+#ifndef NO_GETENV
+  char *env = NULL;
+#endif
+#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+  int bufsize = 1024; /* an initial guess for the line buffer size limit */
+#endif
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = 0;
+
+  simd_support |= JSIMD_NEON;
+#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+  while (!parse_proc_cpuinfo(bufsize)) {
+    bufsize *= 2;
+    if (bufsize > SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT)
+      break;
+  }
+#endif
+
+#ifndef NO_GETENV
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCENEON");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = JSIMD_NEON;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+  env = getenv("JSIMD_FASTLD3");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_features |= JSIMD_FASTLD3;
+  if ((env != NULL) && (strcmp(env, "0") == 0))
+    simd_features &= ~JSIMD_FASTLD3;
+  env = getenv("JSIMD_FASTST3");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_features |= JSIMD_FASTST3;
+  if ((env != NULL) && (strcmp(env, "0") == 0))
+    simd_features &= ~JSIMD_FASTST3;
+#endif
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                      JSAMPIMAGE output_buf, JDIMENSION output_row,
+                      int num_rows)
+{
+  void (*neonfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+#ifndef NEON_INTRINSICS
+    if (simd_features & JSIMD_FASTLD3)
+#endif
+      neonfct = jsimd_extrgb_ycc_convert_neon;
+#ifndef NEON_INTRINSICS
+    else
+      neonfct = jsimd_extrgb_ycc_convert_neon_slowld3;
+#endif
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    neonfct = jsimd_extrgbx_ycc_convert_neon;
+    break;
+  case JCS_EXT_BGR:
+#ifndef NEON_INTRINSICS
+    if (simd_features & JSIMD_FASTLD3)
+#endif
+      neonfct = jsimd_extbgr_ycc_convert_neon;
+#ifndef NEON_INTRINSICS
+    else
+      neonfct = jsimd_extbgr_ycc_convert_neon_slowld3;
+#endif
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    neonfct = jsimd_extbgrx_ycc_convert_neon;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    neonfct = jsimd_extxbgr_ycc_convert_neon;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    neonfct = jsimd_extxrgb_ycc_convert_neon;
+    break;
+  default:
+#ifndef NEON_INTRINSICS
+    if (simd_features & JSIMD_FASTLD3)
+#endif
+      neonfct = jsimd_extrgb_ycc_convert_neon;
+#ifndef NEON_INTRINSICS
+    else
+      neonfct = jsimd_extrgb_ycc_convert_neon_slowld3;
+#endif
+    break;
+  }
+
+  neonfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                       JSAMPIMAGE output_buf, JDIMENSION output_row,
+                       int num_rows)
+{
+  void (*neonfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    neonfct = jsimd_extrgb_gray_convert_neon;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    neonfct = jsimd_extrgbx_gray_convert_neon;
+    break;
+  case JCS_EXT_BGR:
+    neonfct = jsimd_extbgr_gray_convert_neon;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    neonfct = jsimd_extbgrx_gray_convert_neon;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    neonfct = jsimd_extxbgr_gray_convert_neon;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    neonfct = jsimd_extxrgb_gray_convert_neon;
+    break;
+  default:
+    neonfct = jsimd_extrgb_gray_convert_neon;
+    break;
+  }
+
+  neonfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION input_row, JSAMPARRAY output_buf,
+                      int num_rows)
+{
+  void (*neonfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+#ifndef NEON_INTRINSICS
+    if (simd_features & JSIMD_FASTST3)
+#endif
+      neonfct = jsimd_ycc_extrgb_convert_neon;
+#ifndef NEON_INTRINSICS
+    else
+      neonfct = jsimd_ycc_extrgb_convert_neon_slowst3;
+#endif
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    neonfct = jsimd_ycc_extrgbx_convert_neon;
+    break;
+  case JCS_EXT_BGR:
+#ifndef NEON_INTRINSICS
+    if (simd_features & JSIMD_FASTST3)
+#endif
+      neonfct = jsimd_ycc_extbgr_convert_neon;
+#ifndef NEON_INTRINSICS
+    else
+      neonfct = jsimd_ycc_extbgr_convert_neon_slowst3;
+#endif
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    neonfct = jsimd_ycc_extbgrx_convert_neon;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    neonfct = jsimd_ycc_extxbgr_convert_neon;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    neonfct = jsimd_ycc_extxrgb_convert_neon;
+    break;
+  default:
+#ifndef NEON_INTRINSICS
+    if (simd_features & JSIMD_FASTST3)
+#endif
+      neonfct = jsimd_ycc_extrgb_convert_neon;
+#ifndef NEON_INTRINSICS
+    else
+      neonfct = jsimd_ycc_extrgb_convert_neon_slowst3;
+#endif
+    break;
+  }
+
+  neonfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+  jsimd_ycc_rgb565_convert_neon(cinfo->output_width, input_buf, input_row,
+                                output_buf, num_rows);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_neon(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_neon(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_upsample_neon(cinfo->max_v_samp_factor, cinfo->output_width,
+                           input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_upsample_neon(cinfo->max_v_samp_factor, cinfo->output_width,
+                           input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h1v2_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_fancy_upsample_neon(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_fancy_upsample_neon(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h1v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h1v2_fancy_upsample_neon(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*neonfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      neonfct = jsimd_h2v2_extrgb_merged_upsample_neon;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      neonfct = jsimd_h2v2_extrgbx_merged_upsample_neon;
+      break;
+    case JCS_EXT_BGR:
+      neonfct = jsimd_h2v2_extbgr_merged_upsample_neon;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      neonfct = jsimd_h2v2_extbgrx_merged_upsample_neon;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      neonfct = jsimd_h2v2_extxbgr_merged_upsample_neon;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      neonfct = jsimd_h2v2_extxrgb_merged_upsample_neon;
+      break;
+    default:
+      neonfct = jsimd_h2v2_extrgb_merged_upsample_neon;
+      break;
+  }
+
+  neonfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*neonfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      neonfct = jsimd_h2v1_extrgb_merged_upsample_neon;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      neonfct = jsimd_h2v1_extrgbx_merged_upsample_neon;
+      break;
+    case JCS_EXT_BGR:
+      neonfct = jsimd_h2v1_extbgr_merged_upsample_neon;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      neonfct = jsimd_h2v1_extbgrx_merged_upsample_neon;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      neonfct = jsimd_h2v1_extxbgr_merged_upsample_neon;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      neonfct = jsimd_h2v1_extxrgb_merged_upsample_neon;
+      break;
+    default:
+      neonfct = jsimd_h2v1_extrgb_merged_upsample_neon;
+      break;
+  }
+
+  neonfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
+               DCTELEM *workspace)
+{
+  jsimd_convsamp_neon(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow(DCTELEM *data)
+{
+  jsimd_fdct_islow_neon(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast(DCTELEM *data)
+{
+  jsimd_fdct_ifast_neon(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float(FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
+{
+  jsimd_quantize_neon(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_NEON && simd_huffman)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET *)
+jsimd_huff_encode_one_block(void *state, JOCTET *buffer, JCOEFPTR block,
+                            int last_dc_val, c_derived_tbl *dctbl,
+                            c_derived_tbl *actbl)
+{
+#ifndef NEON_INTRINSICS
+  if (simd_features & JSIMD_FASTTBL)
+#endif
+    return jsimd_huff_encode_one_block_neon(state, buffer, block, last_dc_val,
+                                            dctbl, actbl);
+#ifndef NEON_INTRINSICS
+  else
+    return jsimd_huff_encode_one_block_neon_slowtbl(state, buffer, block,
+                                                    last_dc_val, dctbl, actbl);
+#endif
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_first_prepare(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (SIZEOF_SIZE_T != 8)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_encode_mcu_AC_first_prepare(const JCOEF *block,
+                                  const int *jpeg_natural_order_start, int Sl,
+                                  int Al, JCOEF *values, size_t *zerobits)
+{
+  jsimd_encode_mcu_AC_first_prepare_neon(block, jpeg_natural_order_start,
+                                         Sl, Al, values, zerobits);
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_refine_prepare(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (SIZEOF_SIZE_T != 8)
+    return 0;
+
+  if (simd_support & JSIMD_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block,
+                                   const int *jpeg_natural_order_start, int Sl,
+                                   int Al, JCOEF *absvalues, size_t *bits)
+{
+  return jsimd_encode_mcu_AC_refine_prepare_neon(block,
+                                                 jpeg_natural_order_start,
+                                                 Sl, Al, absvalues, bits);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/aarch64/jsimd_neon.S Added

@@ -0,0 +1,2252 @@
+/*
+ * Armv8 Neon optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2009-2011, Nokia Corporation and/or its subsidiary(-ies).
+ *                          All Rights Reserved.
+ * Author:  Siarhei Siamashka <siarhei.siamashka@nokia.com>
+ * Copyright (C) 2013-2014, Linaro Limited.  All Rights Reserved.
+ * Author:  Ragesh Radhakrishnan <ragesh.r@linaro.org>
+ * Copyright (C) 2014-2016, 2020, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.  All Rights Reserved.
+ * Copyright (C) 2016, Siarhei Siamashka.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack, "", %progbits  /* mark stack as non-executable */
+#endif
+
+#if defined(__APPLE__)
+.section __DATA, __const
+#elif defined(_WIN32)
+.section .rdata
+#else
+.section .rodata, "a", %progbits
+#endif
+
+/* Constants for jsimd_idct_islow_neon() */
+
+#define F_0_298   2446  /* FIX(0.298631336) */
+#define F_0_390   3196  /* FIX(0.390180644) */
+#define F_0_541   4433  /* FIX(0.541196100) */
+#define F_0_765   6270  /* FIX(0.765366865) */
+#define F_0_899   7373  /* FIX(0.899976223) */
+#define F_1_175   9633  /* FIX(1.175875602) */
+#define F_1_501  12299  /* FIX(1.501321110) */
+#define F_1_847  15137  /* FIX(1.847759065) */
+#define F_1_961  16069  /* FIX(1.961570560) */
+#define F_2_053  16819  /* FIX(2.053119869) */
+#define F_2_562  20995  /* FIX(2.562915447) */
+#define F_3_072  25172  /* FIX(3.072711026) */
+
+.balign 16
+Ljsimd_idct_islow_neon_consts:
+  .short F_0_298
+  .short -F_0_390
+  .short F_0_541
+  .short F_0_765
+  .short - F_0_899
+  .short F_1_175
+  .short F_1_501
+  .short - F_1_847
+  .short - F_1_961
+  .short F_2_053
+  .short - F_2_562
+  .short F_3_072
+  .short 0          /* padding */
+  .short 0
+  .short 0
+  .short 0
+
+#undef F_0_298
+#undef F_0_390
+#undef F_0_541
+#undef F_0_765
+#undef F_0_899
+#undef F_1_175
+#undef F_1_501
+#undef F_1_847
+#undef F_1_961
+#undef F_2_053
+#undef F_2_562
+#undef F_3_072
+
+/* Constants for jsimd_ycc_*_neon() */
+
+.balign 16
+Ljsimd_ycc_rgb_neon_consts:
+  .short 0,      0,     0,      0
+  .short 22971, -11277, -23401, 29033
+  .short -128,  -128,   -128,   -128
+  .short -128,  -128,   -128,   -128
+
+/* Constants for jsimd_*_ycc_neon() */
+
+.balign 16
+Ljsimd_rgb_ycc_neon_consts:
+  .short 19595, 38470, 7471, 11059
+  .short 21709, 32768, 27439, 5329
+  .short 32767, 128, 32767, 128
+  .short 32767, 128, 32767, 128
+
+/* Constants for jsimd_fdct_islow_neon() */
+
+#define F_0_298   2446  /* FIX(0.298631336) */
+#define F_0_390   3196  /* FIX(0.390180644) */
+#define F_0_541   4433  /* FIX(0.541196100) */
+#define F_0_765   6270  /* FIX(0.765366865) */
+#define F_0_899   7373  /* FIX(0.899976223) */
+#define F_1_175   9633  /* FIX(1.175875602) */
+#define F_1_501  12299  /* FIX(1.501321110) */
+#define F_1_847  15137  /* FIX(1.847759065) */
+#define F_1_961  16069  /* FIX(1.961570560) */
+#define F_2_053  16819  /* FIX(2.053119869) */
+#define F_2_562  20995  /* FIX(2.562915447) */
+#define F_3_072  25172  /* FIX(3.072711026) */
+
+.balign 16
+Ljsimd_fdct_islow_neon_consts:
+  .short F_0_298
+  .short -F_0_390
+  .short F_0_541
+  .short F_0_765
+  .short - F_0_899
+  .short F_1_175
+  .short F_1_501
+  .short - F_1_847
+  .short - F_1_961
+  .short F_2_053
+  .short - F_2_562
+  .short F_3_072
+  .short 0          /* padding */
+  .short 0
+  .short 0
+  .short 0
+
+#undef F_0_298
+#undef F_0_390
+#undef F_0_541
+#undef F_0_765
+#undef F_0_899
+#undef F_1_175
+#undef F_1_501
+#undef F_1_847
+#undef F_1_961
+#undef F_2_053
+#undef F_2_562
+#undef F_3_072
+
+/* Constants for jsimd_huff_encode_one_block_neon() */
+
+.balign 16
+Ljsimd_huff_encode_one_block_neon_consts:
+    .byte 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, \
+          0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80
+    .byte    0,   1,   2,   3,  16,  17,  32,  33, \
+            18,  19,   4,   5,   6,   7,  20,  21  /* L0 => L3 : 4 lines OK */
+    .byte   34,  35,  48,  49, 255, 255,  50,  51, \
+            36,  37,  22,  23,   8,   9,  10,  11  /* L0 => L3 : 4 lines OK */
+    .byte    8,   9,  22,  23,  36,  37,  50,  51, \
+           255, 255, 255, 255, 255, 255,  52,  53  /* L1 => L4 : 4 lines OK */
+    .byte   54,  55,  40,  41,  26,  27,  12,  13, \
+            14,  15,  28,  29,  42,  43,  56,  57  /* L0 => L3 : 4 lines OK */
+    .byte    6,   7,  20,  21,  34,  35,  48,  49, \
+            50,  51,  36,  37,  22,  23,   8,   9  /* L4 => L7 : 4 lines OK */
+    .byte   42,  43,  28,  29,  14,  15,  30,  31, \
+            44,  45,  58,  59, 255, 255, 255, 255  /* L1 => L4 : 4 lines OK */
+    .byte  255, 255, 255, 255,  56,  57,  42,  43, \
+            28,  29,  14,  15,  30,  31,  44,  45  /* L3 => L6 : 4 lines OK */
+    .byte   26,  27,  40,  41,  42,  43,  28,  29, \
+            14,  15,  30,  31,  44,  45,  46,  47  /* L5 => L7 : 3 lines OK */
+    .byte  255, 255, 255, 255,   0,   1, 255, 255, \
+           255, 255, 255, 255, 255, 255, 255, 255  /* L4 : 1 lines OK */
+    .byte  255, 255, 255, 255, 255, 255, 255, 255, \
+             0,   1,  16,  17,   2,   3, 255, 255  /* L5 => L6 : 2 lines OK */
+    .byte  255, 255, 255, 255, 255, 255, 255, 255, \
+           255, 255, 255, 255,   8,   9,  22,  23  /* L5 => L6 : 2 lines OK */
+    .byte    4,   5,   6,   7, 255, 255, 255, 255, \
+           255, 255, 255, 255, 255, 255, 255, 255  /* L7 : 1 line OK */
+
+
+/*****************************************************************************/
+
+/* Supplementary macro for setting function attributes */
+.macro asm_function fname
+#ifdef __APPLE__
+    .private_extern _\fname
+    .globl _\fname
+_\fname:
+#else
+    .global \fname
+#ifdef __ELF__
+    .hidden \fname
+    .type \fname, %function
+#endif
+\fname:
+#endif
+.endm
+
+/* Get symbol location */
+.macro get_symbol_loc reg, symbol
+#ifdef __APPLE__
+    adrp            \reg, \symbol@PAGE
+    add             \reg, \reg, \symbol@PAGEOFF
+#else
+    adrp            \reg, \symbol
+    add             \reg, \reg, :lo12:\symbol
+#endif
+.endm
+
+.macro transpose_8x8 l0, l1, l2, l3, l4, l5, l6, l7, t0, t1, t2, t3
+    trn1            \t0\().8h, \l0\().8h, \l1\().8h
+    trn1            \t1\().8h, \l2\().8h, \l3\().8h
+    trn1            \t2\().8h, \l4\().8h, \l5\().8h
+    trn1            \t3\().8h, \l6\().8h, \l7\().8h
+    trn2            \l1\().8h, \l0\().8h, \l1\().8h
+    trn2            \l3\().8h, \l2\().8h, \l3\().8h
+    trn2            \l5\().8h, \l4\().8h, \l5\().8h
+    trn2            \l7\().8h, \l6\().8h, \l7\().8h
+
+    trn1            \l4\().4s, \t2\().4s, \t3\().4s
+    trn2            \t3\().4s, \t2\().4s, \t3\().4s
+    trn1            \t2\().4s, \t0\().4s, \t1\().4s
+    trn2            \l2\().4s, \t0\().4s, \t1\().4s
+    trn1            \t0\().4s, \l1\().4s, \l3\().4s
+    trn2            \l3\().4s, \l1\().4s, \l3\().4s
+    trn2            \t1\().4s, \l5\().4s, \l7\().4s
+    trn1            \l5\().4s, \l5\().4s, \l7\().4s
+
+    trn2            \l6\().2d, \l2\().2d, \t3\().2d
+    trn1            \l0\().2d, \t2\().2d, \l4\().2d
+    trn1            \l1\().2d, \t0\().2d, \l5\().2d
+    trn2            \l7\().2d, \l3\().2d, \t1\().2d
+    trn1            \l2\().2d, \l2\().2d, \t3\().2d
+    trn2            \l4\().2d, \t2\().2d, \l4\().2d
+    trn1            \l3\().2d, \l3\().2d, \t1\().2d
+    trn2            \l5\().2d, \t0\().2d, \l5\().2d
+.endm
+
+
+#define CENTERJSAMPLE  128
+
+/*****************************************************************************/
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ *
+ * GLOBAL(void)
+ * jsimd_idct_islow_neon(void *dct_table, JCOEFPTR coef_block,
+ *                       JSAMPARRAY output_buf, JDIMENSION output_col)
+ */
+
+#define CONST_BITS  13
+#define PASS1_BITS  2
+
+#define XFIX_P_0_298  v0.h0
+#define XFIX_N_0_390  v0.h1
+#define XFIX_P_0_541  v0.h2
+#define XFIX_P_0_765  v0.h3
+#define XFIX_N_0_899  v0.h4
+#define XFIX_P_1_175  v0.h5
+#define XFIX_P_1_501  v0.h6
+#define XFIX_N_1_847  v0.h7
+#define XFIX_N_1_961  v1.h0
+#define XFIX_P_2_053  v1.h1
+#define XFIX_N_2_562  v1.h2
+#define XFIX_P_3_072  v1.h3
+
+asm_function jsimd_idct_islow_neon
+    DCT_TABLE       .req x0
+    COEF_BLOCK      .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_COL      .req x3
+    TMP1            .req x0
+    TMP2            .req x1
+    TMP3            .req x9
+    TMP4            .req x10
+    TMP5            .req x11
+    TMP6            .req x12
+    TMP7            .req x13
+    TMP8            .req x14
+
+    /* OUTPUT_COL is a JDIMENSION (unsigned int) argument, so the ABI doesn't
+       guarantee that the upper (unused) 32 bits of x3 are valid.  This
+       instruction ensures that those bits are set to zero. */
+    uxtw x3, w3
+
+    sub             sp, sp, #64
+    get_symbol_loc  x15, Ljsimd_idct_islow_neon_consts
+    mov             x10, sp
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, x10, #32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, x10, #32
+    ld1             {v0.8h, v1.8h}, x15
+    ld1             {v2.8h, v3.8h, v4.8h, v5.8h}, COEF_BLOCK, #64
+    ld1             {v18.8h, v19.8h, v20.8h, v21.8h}, DCT_TABLE, #64
+    ld1             {v6.8h, v7.8h, v8.8h, v9.8h}, COEF_BLOCK, #64
+    ld1             {v22.8h, v23.8h, v24.8h, v25.8h}, DCT_TABLE, #64
+
+    cmeq            v16.8h, v3.8h, #0
+    cmeq            v26.8h, v4.8h, #0
+    cmeq            v27.8h, v5.8h, #0
+    cmeq            v28.8h, v6.8h, #0
+    cmeq            v29.8h, v7.8h, #0
+    cmeq            v30.8h, v8.8h, #0
+    cmeq            v31.8h, v9.8h, #0
+
+    and             v10.16b, v16.16b, v26.16b
+    and             v11.16b, v27.16b, v28.16b
+    and             v12.16b, v29.16b, v30.16b
+    and             v13.16b, v31.16b, v10.16b
+    and             v14.16b, v11.16b, v12.16b
+    mul             v2.8h, v2.8h, v18.8h
+    and             v15.16b, v13.16b, v14.16b
+    shl             v10.8h, v2.8h, #(PASS1_BITS)
+    sqxtn           v16.8b, v15.8h
+    mov             TMP1, v16.d0
+    mvn             TMP2, TMP1
+
+    cbnz            TMP2, 2f
+    /* case all AC coeffs are zeros */
+    dup             v2.2d, v10.d0
+    dup             v6.2d, v10.d1
+    mov             v3.16b, v2.16b
+    mov             v7.16b, v6.16b
+    mov             v4.16b, v2.16b
+    mov             v8.16b, v6.16b
+    mov             v5.16b, v2.16b
+    mov             v9.16b, v6.16b
+1:
+    /* for this transpose, we should organise data like this:
+     * 00, 01, 02, 03, 40, 41, 42, 43
+     * 10, 11, 12, 13, 50, 51, 52, 53
+     * 20, 21, 22, 23, 60, 61, 62, 63
+     * 30, 31, 32, 33, 70, 71, 72, 73
+     * 04, 05, 06, 07, 44, 45, 46, 47
+     * 14, 15, 16, 17, 54, 55, 56, 57
+     * 24, 25, 26, 27, 64, 65, 66, 67
+     * 34, 35, 36, 37, 74, 75, 76, 77
+     */
+    trn1            v28.8h, v2.8h, v3.8h
+    trn1            v29.8h, v4.8h, v5.8h
+    trn1            v30.8h, v6.8h, v7.8h
+    trn1            v31.8h, v8.8h, v9.8h
+    trn2            v16.8h, v2.8h, v3.8h
+    trn2            v17.8h, v4.8h, v5.8h
+    trn2            v18.8h, v6.8h, v7.8h
+    trn2            v19.8h, v8.8h, v9.8h
+    trn1            v2.4s, v28.4s, v29.4s
+    trn1            v6.4s, v30.4s, v31.4s
+    trn1            v3.4s, v16.4s, v17.4s
+    trn1            v7.4s, v18.4s, v19.4s
+    trn2            v4.4s, v28.4s, v29.4s
+    trn2            v8.4s, v30.4s, v31.4s
+    trn2            v5.4s, v16.4s, v17.4s
+    trn2            v9.4s, v18.4s, v19.4s
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.8h, v4.8h, v8.8h           /* z2 + z3 = DEQUANTIZE(inptrDCTSIZE*2, quantptrDCTSIZE*2) + DEQUANTIZE(inptrDCTSIZE*6, quantptrDCTSIZE*6) */
+    add             v22.8h, v2.8h, v6.8h           /* z2 + z3 = DEQUANTIZE(inptrDCTSIZE*0, quantptrDCTSIZE*0) + DEQUANTIZE(inptrDCTSIZE*4, quantptrDCTSIZE*4) */
+    smull2          v19.4s, v18.8h, XFIX_P_0_541   /* z1h z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sub             v26.8h, v2.8h, v6.8h           /* z2 - z3 = DEQUANTIZE(inptrDCTSIZE*0, quantptrDCTSIZE*0) - DEQUANTIZE(inptrDCTSIZE*4, quantptrDCTSIZE*4) */
+    smull           v18.4s, v18.4h, XFIX_P_0_541   /* z1l z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sshll2          v23.4s, v22.8h, #(CONST_BITS)  /* tmp0h tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v21.16b, v19.16b               /* tmp3 = z1 */
+    mov             v20.16b, v18.16b               /* tmp3 = z1 */
+    smlal2          v19.4s, v8.8h, XFIX_N_1_847    /* tmp2h tmp2 = z1 + MULTIPLY(z3, -FIX_1_847759065); */
+    smlal           v18.4s, v8.4h, XFIX_N_1_847    /* tmp2l tmp2 = z1 + MULTIPLY(z3, -FIX_1_847759065); */
+    sshll2          v27.4s, v26.8h, #(CONST_BITS)  /* tmp1h tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal2          v21.4s, v4.8h, XFIX_P_0_765    /* tmp3h tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    smlal           v20.4s, v4.4h, XFIX_P_0_765    /* tmp3l tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    sshll           v22.4s, v22.4h, #(CONST_BITS)  /* tmp0l tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    sshll           v26.4s, v26.4h, #(CONST_BITS)  /* tmp1l tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    add             v2.4s, v22.4s, v20.4s          /* tmp10l tmp10 = tmp0 + tmp3; */
+    sub             v6.4s, v22.4s, v20.4s          /* tmp13l tmp13 = tmp0 - tmp3; */
+    add             v8.4s, v26.4s, v18.4s          /* tmp11l tmp11 = tmp1 + tmp2; */
+    sub             v4.4s, v26.4s, v18.4s          /* tmp12l tmp12 = tmp1 - tmp2; */
+    add             v28.4s, v23.4s, v21.4s         /* tmp10h tmp10 = tmp0 + tmp3; */
+    sub             v31.4s, v23.4s, v21.4s         /* tmp13h tmp13 = tmp0 - tmp3; */
+    add             v29.4s, v27.4s, v19.4s         /* tmp11h tmp11 = tmp1 + tmp2; */
+    sub             v30.4s, v27.4s, v19.4s         /* tmp12h tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.8h, v9.8h, v5.8h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptrDCTSIZE*7, quantptrDCTSIZE*7) + DEQUANTIZE(inptrDCTSIZE*3, quantptrDCTSIZE*3) */
+    add             v24.8h, v7.8h, v3.8h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptrDCTSIZE*5, quantptrDCTSIZE*5) + DEQUANTIZE(inptrDCTSIZE*1, quantptrDCTSIZE*1) */
+    add             v18.8h, v9.8h, v3.8h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptrDCTSIZE*7, quantptrDCTSIZE*7) + DEQUANTIZE(inptrDCTSIZE*1, quantptrDCTSIZE*1) */
+    add             v20.8h, v7.8h, v5.8h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptrDCTSIZE*5, quantptrDCTSIZE*5) + DEQUANTIZE(inptrDCTSIZE*3, quantptrDCTSIZE*3) */
+    add             v26.8h, v22.8h, v24.8h  /* z5 = z3 + z4 */
+
+    smull2          v11.4s, v9.8h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull2          v13.4s, v7.8h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull2          v15.4s, v5.8h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull2          v17.4s, v3.8h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull2          v27.4s, v26.8h, XFIX_P_1_175  /* z5h z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull2          v23.4s, v22.8h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, -FIX_1_961570560) */
+    smull2          v25.4s, v24.8h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, -FIX_0_390180644) */
+    smull2          v19.4s, v18.8h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, -FIX_0_899976223) */
+    smull2          v21.4s, v20.8h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, -FIX_2_562915447) */
+
+    smull           v10.4s, v9.4h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull           v12.4s, v7.4h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull           v14.4s, v5.4h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull           v16.4s, v3.4h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull           v26.4s, v26.4h, XFIX_P_1_175  /* z5l z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull           v22.4s, v22.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, -FIX_1_961570560) */
+    smull           v24.4s, v24.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, -FIX_0_390180644) */
+    smull           v18.4s, v18.4h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, -FIX_0_899976223) */
+    smull           v20.4s, v20.4h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, -FIX_2_562915447) */
+
+    add             v23.4s, v23.4s, v27.4s  /* z3 += z5 */
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v25.4s, v25.4s, v27.4s  /* z4 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v11.4s, v11.4s, v19.4s  /* tmp0 += z1 */
+    add             v10.4s, v10.4s, v18.4s  /* tmp0 += z1 */
+    add             v13.4s, v13.4s, v21.4s  /* tmp1 += z2 */
+    add             v12.4s, v12.4s, v20.4s  /* tmp1 += z2 */
+    add             v15.4s, v15.4s, v21.4s  /* tmp2 += z2 */
+    add             v14.4s, v14.4s, v20.4s  /* tmp2 += z2 */
+    add             v17.4s, v17.4s, v19.4s  /* tmp3 += z1 */
+    add             v16.4s, v16.4s, v18.4s  /* tmp3 += z1 */
+
+    add             v11.4s, v11.4s, v23.4s  /* tmp0 += z3 */
+    add             v10.4s, v10.4s, v22.4s  /* tmp0 += z3 */
+    add             v13.4s, v13.4s, v25.4s  /* tmp1 += z4 */
+    add             v12.4s, v12.4s, v24.4s  /* tmp1 += z4 */
+    add             v17.4s, v17.4s, v25.4s  /* tmp3 += z4 */
+    add             v16.4s, v16.4s, v24.4s  /* tmp3 += z4 */
+    add             v15.4s, v15.4s, v23.4s  /* tmp2 += z3 */
+    add             v14.4s, v14.4s, v22.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v18.4s, v2.4s, v16.4s   /* tmp10 + tmp3 */
+    add             v19.4s, v28.4s, v17.4s  /* tmp10 + tmp3 */
+    sub             v20.4s, v2.4s, v16.4s   /* tmp10 - tmp3 */
+    sub             v21.4s, v28.4s, v17.4s  /* tmp10 - tmp3 */
+    add             v22.4s, v8.4s, v14.4s   /* tmp11 + tmp2 */
+    add             v23.4s, v29.4s, v15.4s  /* tmp11 + tmp2 */
+    sub             v24.4s, v8.4s, v14.4s   /* tmp11 - tmp2 */
+    sub             v25.4s, v29.4s, v15.4s  /* tmp11 - tmp2 */
+    add             v26.4s, v4.4s, v12.4s   /* tmp12 + tmp1 */
+    add             v27.4s, v30.4s, v13.4s  /* tmp12 + tmp1 */
+    sub             v28.4s, v4.4s, v12.4s   /* tmp12 - tmp1 */
+    sub             v29.4s, v30.4s, v13.4s  /* tmp12 - tmp1 */
+    add             v14.4s, v6.4s, v10.4s   /* tmp13 + tmp0 */
+    add             v15.4s, v31.4s, v11.4s  /* tmp13 + tmp0 */
+    sub             v16.4s, v6.4s, v10.4s   /* tmp13 - tmp0 */
+    sub             v17.4s, v31.4s, v11.4s  /* tmp13 - tmp0 */
+
+    shrn            v2.4h, v18.4s, #16  /* wsptrDCTSIZE*0 = (int)DESCALE(tmp10 + tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn            v9.4h, v20.4s, #16  /* wsptrDCTSIZE*7 = (int)DESCALE(tmp10 - tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn            v3.4h, v22.4s, #16  /* wsptrDCTSIZE*1 = (int)DESCALE(tmp11 + tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn            v8.4h, v24.4s, #16  /* wsptrDCTSIZE*6 = (int)DESCALE(tmp11 - tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn            v4.4h, v26.4s, #16  /* wsptrDCTSIZE*2 = (int)DESCALE(tmp12 + tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn            v7.4h, v28.4s, #16  /* wsptrDCTSIZE*5 = (int)DESCALE(tmp12 - tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn            v5.4h, v14.4s, #16  /* wsptrDCTSIZE*3 = (int)DESCALE(tmp13 + tmp0, CONST_BITS+PASS1_BITS+3) */
+    shrn            v6.4h, v16.4s, #16  /* wsptrDCTSIZE*4 = (int)DESCALE(tmp13 - tmp0, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v2.8h, v19.4s, #16  /* wsptrDCTSIZE*0 = (int)DESCALE(tmp10 + tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v9.8h, v21.4s, #16  /* wsptrDCTSIZE*7 = (int)DESCALE(tmp10 - tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v3.8h, v23.4s, #16  /* wsptrDCTSIZE*1 = (int)DESCALE(tmp11 + tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v8.8h, v25.4s, #16  /* wsptrDCTSIZE*6 = (int)DESCALE(tmp11 - tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v4.8h, v27.4s, #16  /* wsptrDCTSIZE*2 = (int)DESCALE(tmp12 + tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v7.8h, v29.4s, #16  /* wsptrDCTSIZE*5 = (int)DESCALE(tmp12 - tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v5.8h, v15.4s, #16  /* wsptrDCTSIZE*3 = (int)DESCALE(tmp13 + tmp0, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v6.8h, v17.4s, #16  /* wsptrDCTSIZE*4 = (int)DESCALE(tmp13 - tmp0, CONST_BITS+PASS1_BITS+3) */
+    movi            v0.16b, #(CENTERJSAMPLE)
+    /* Prepare pointers (dual-issue with Neon instructions) */
+      ldp             TMP1, TMP2, OUTPUT_BUF, 16
+    sqrshrn         v28.8b, v2.8h, #(CONST_BITS + PASS1_BITS + 3 - 16)
+      ldp             TMP3, TMP4, OUTPUT_BUF, 16
+    sqrshrn         v29.8b, v3.8h, #(CONST_BITS + PASS1_BITS + 3 - 16)
+      add             TMP1, TMP1, OUTPUT_COL
+    sqrshrn         v30.8b, v4.8h, #(CONST_BITS + PASS1_BITS + 3 - 16)
+      add             TMP2, TMP2, OUTPUT_COL
+    sqrshrn         v31.8b, v5.8h, #(CONST_BITS + PASS1_BITS + 3 - 16)
+      add             TMP3, TMP3, OUTPUT_COL
+    sqrshrn2        v28.16b, v6.8h, #(CONST_BITS + PASS1_BITS + 3 - 16)
+      add             TMP4, TMP4, OUTPUT_COL
+    sqrshrn2        v29.16b, v7.8h, #(CONST_BITS + PASS1_BITS + 3 - 16)
+      ldp             TMP5, TMP6, OUTPUT_BUF, 16
+    sqrshrn2        v30.16b, v8.8h, #(CONST_BITS + PASS1_BITS + 3 - 16)
+      ldp             TMP7, TMP8, OUTPUT_BUF, 16
+    sqrshrn2        v31.16b, v9.8h, #(CONST_BITS + PASS1_BITS + 3 - 16)
+      add             TMP5, TMP5, OUTPUT_COL
+    add             v16.16b, v28.16b, v0.16b
+      add             TMP6, TMP6, OUTPUT_COL
+    add             v18.16b, v29.16b, v0.16b
+      add             TMP7, TMP7, OUTPUT_COL
+    add             v20.16b, v30.16b, v0.16b
+      add             TMP8, TMP8, OUTPUT_COL
+    add             v22.16b, v31.16b, v0.16b
+
+    /* Transpose the final 8-bit samples */
+    trn1            v28.16b, v16.16b, v18.16b
+    trn1            v30.16b, v20.16b, v22.16b
+    trn2            v29.16b, v16.16b, v18.16b
+    trn2            v31.16b, v20.16b, v22.16b
+
+    trn1            v16.8h, v28.8h, v30.8h
+    trn2            v18.8h, v28.8h, v30.8h
+    trn1            v20.8h, v29.8h, v31.8h
+    trn2            v22.8h, v29.8h, v31.8h
+
+    uzp1            v28.4s, v16.4s, v18.4s
+    uzp2            v30.4s, v16.4s, v18.4s
+    uzp1            v29.4s, v20.4s, v22.4s
+    uzp2            v31.4s, v20.4s, v22.4s
+
+    /* Store results to the output buffer */
+    st1             {v28.d}0, TMP1
+    st1             {v29.d}0, TMP2
+    st1             {v28.d}1, TMP3
+    st1             {v29.d}1, TMP4
+    st1             {v30.d}0, TMP5
+    st1             {v31.d}0, TMP6
+    st1             {v30.d}1, TMP7
+    st1             {v31.d}1, TMP8
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, sp, #32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, sp, #32
+    blr             x30
+
+.balign 16
+2:
+    mul             v3.8h, v3.8h, v19.8h
+    mul             v4.8h, v4.8h, v20.8h
+    mul             v5.8h, v5.8h, v21.8h
+    add             TMP4, xzr, TMP2, LSL #32
+    mul             v6.8h, v6.8h, v22.8h
+    mul             v7.8h, v7.8h, v23.8h
+    adds            TMP3, xzr, TMP2, LSR #32
+    mul             v8.8h, v8.8h, v24.8h
+    mul             v9.8h, v9.8h, v25.8h
+    b.ne            3f
+    /* Right AC coef is zero */
+    dup             v15.2d, v10.d1
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.4h, v4.4h, v8.4h           /* z2 + z3 = DEQUANTIZE(inptrDCTSIZE*2, quantptrDCTSIZE*2) + DEQUANTIZE(inptrDCTSIZE*6, quantptrDCTSIZE*6) */
+    add             v22.4h, v2.4h, v6.4h           /* z2 + z3 = DEQUANTIZE(inptrDCTSIZE*0, quantptrDCTSIZE*0) + DEQUANTIZE(inptrDCTSIZE*4, quantptrDCTSIZE*4) */
+    sub             v26.4h, v2.4h, v6.4h           /* z2 - z3 = DEQUANTIZE(inptrDCTSIZE*0, quantptrDCTSIZE*0) - DEQUANTIZE(inptrDCTSIZE*4, quantptrDCTSIZE*4) */
+    smull           v18.4s, v18.4h, XFIX_P_0_541   /* z1l z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sshll           v22.4s, v22.4h, #(CONST_BITS)  /* tmp0l tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v20.16b, v18.16b               /* tmp3 = z1 */
+    sshll           v26.4s, v26.4h, #(CONST_BITS)  /* tmp1l tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal           v18.4s, v8.4h, XFIX_N_1_847    /* tmp2l tmp2 = z1 + MULTIPLY(z3, -FIX_1_847759065); */
+    smlal           v20.4s, v4.4h, XFIX_P_0_765    /* tmp3l tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    add             v2.4s, v22.4s, v20.4s          /* tmp10l tmp10 = tmp0 + tmp3; */
+    sub             v6.4s, v22.4s, v20.4s          /* tmp13l tmp13 = tmp0 - tmp3; */
+    add             v8.4s, v26.4s, v18.4s          /* tmp11l tmp11 = tmp1 + tmp2; */
+    sub             v4.4s, v26.4s, v18.4s          /* tmp12l tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.4h, v9.4h, v5.4h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptrDCTSIZE*7, quantptrDCTSIZE*7) + DEQUANTIZE(inptrDCTSIZE*3, quantptrDCTSIZE*3) */
+    add             v24.4h, v7.4h, v3.4h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptrDCTSIZE*5, quantptrDCTSIZE*5) + DEQUANTIZE(inptrDCTSIZE*1, quantptrDCTSIZE*1) */
+    add             v18.4h, v9.4h, v3.4h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptrDCTSIZE*7, quantptrDCTSIZE*7) + DEQUANTIZE(inptrDCTSIZE*1, quantptrDCTSIZE*1) */
+    add             v20.4h, v7.4h, v5.4h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptrDCTSIZE*5, quantptrDCTSIZE*5) + DEQUANTIZE(inptrDCTSIZE*3, quantptrDCTSIZE*3) */
+    add             v26.4h, v22.4h, v24.4h  /* z5 = z3 + z4 */
+
+    smull           v10.4s, v9.4h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull           v12.4s, v7.4h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull           v14.4s, v5.4h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull           v16.4s, v3.4h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull           v26.4s, v26.4h, XFIX_P_1_175  /* z5l z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull           v22.4s, v22.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, -FIX_1_961570560) */
+    smull           v24.4s, v24.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, -FIX_0_390180644) */
+    smull           v18.4s, v18.4h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, -FIX_0_899976223) */
+    smull           v20.4s, v20.4h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, -FIX_2_562915447) */
+
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v10.4s, v10.4s, v18.4s  /* tmp0 += z1 */
+    add             v12.4s, v12.4s, v20.4s  /* tmp1 += z2 */
+    add             v14.4s, v14.4s, v20.4s  /* tmp2 += z2 */
+    add             v16.4s, v16.4s, v18.4s  /* tmp3 += z1 */
+
+    add             v10.4s, v10.4s, v22.4s  /* tmp0 += z3 */
+    add             v12.4s, v12.4s, v24.4s  /* tmp1 += z4 */
+    add             v16.4s, v16.4s, v24.4s  /* tmp3 += z4 */
+    add             v14.4s, v14.4s, v22.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v18.4s, v2.4s, v16.4s  /* tmp10 + tmp3 */
+    sub             v20.4s, v2.4s, v16.4s  /* tmp10 - tmp3 */
+    add             v22.4s, v8.4s, v14.4s  /* tmp11 + tmp2 */
+    sub             v24.4s, v8.4s, v14.4s  /* tmp11 - tmp2 */
+    add             v26.4s, v4.4s, v12.4s  /* tmp12 + tmp1 */
+    sub             v28.4s, v4.4s, v12.4s  /* tmp12 - tmp1 */
+    add             v14.4s, v6.4s, v10.4s  /* tmp13 + tmp0 */
+    sub             v16.4s, v6.4s, v10.4s  /* tmp13 - tmp0 */
+
+    rshrn           v2.4h, v18.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*0 = (int)DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v3.4h, v22.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*1 = (int)DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v4.4h, v26.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*2 = (int)DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v5.4h, v14.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*3 = (int)DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v2.8h, v16.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*4 = (int)DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v3.8h, v28.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*5 = (int)DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v4.8h, v24.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*6 = (int)DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v5.8h, v20.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*7 = (int)DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    mov             v6.16b, v15.16b
+    mov             v7.16b, v15.16b
+    mov             v8.16b, v15.16b
+    mov             v9.16b, v15.16b
+    b               1b
+
+.balign 16
+3:
+    cbnz            TMP4, 4f
+    /* Left AC coef is zero */
+    dup             v14.2d, v10.d0
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.8h, v4.8h, v8.8h           /* z2 + z3 = DEQUANTIZE(inptrDCTSIZE*2, quantptrDCTSIZE*2) + DEQUANTIZE(inptrDCTSIZE*6, quantptrDCTSIZE*6) */
+    add             v22.8h, v2.8h, v6.8h           /* z2 + z3 = DEQUANTIZE(inptrDCTSIZE*0, quantptrDCTSIZE*0) + DEQUANTIZE(inptrDCTSIZE*4, quantptrDCTSIZE*4) */
+    smull2          v19.4s, v18.8h, XFIX_P_0_541   /* z1h z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sub             v26.8h, v2.8h, v6.8h           /* z2 - z3 = DEQUANTIZE(inptrDCTSIZE*0, quantptrDCTSIZE*0) - DEQUANTIZE(inptrDCTSIZE*4, quantptrDCTSIZE*4) */
+    sshll2          v23.4s, v22.8h, #(CONST_BITS)  /* tmp0h tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v21.16b, v19.16b               /* tmp3 = z1 */
+    smlal2          v19.4s, v8.8h, XFIX_N_1_847    /* tmp2h tmp2 = z1 + MULTIPLY(z3, -FIX_1_847759065); */
+    sshll2          v27.4s, v26.8h, #(CONST_BITS)  /* tmp1h tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal2          v21.4s, v4.8h, XFIX_P_0_765    /* tmp3h tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    add             v28.4s, v23.4s, v21.4s         /* tmp10h tmp10 = tmp0 + tmp3; */
+    sub             v31.4s, v23.4s, v21.4s         /* tmp13h tmp13 = tmp0 - tmp3; */
+    add             v29.4s, v27.4s, v19.4s         /* tmp11h tmp11 = tmp1 + tmp2; */
+    sub             v30.4s, v27.4s, v19.4s         /* tmp12h tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.8h, v9.8h, v5.8h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptrDCTSIZE*7, quantptrDCTSIZE*7) + DEQUANTIZE(inptrDCTSIZE*3, quantptrDCTSIZE*3) */
+    add             v24.8h, v7.8h, v3.8h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptrDCTSIZE*5, quantptrDCTSIZE*5) + DEQUANTIZE(inptrDCTSIZE*1, quantptrDCTSIZE*1) */
+    add             v18.8h, v9.8h, v3.8h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptrDCTSIZE*7, quantptrDCTSIZE*7) + DEQUANTIZE(inptrDCTSIZE*1, quantptrDCTSIZE*1) */
+    add             v20.8h, v7.8h, v5.8h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptrDCTSIZE*5, quantptrDCTSIZE*5) + DEQUANTIZE(inptrDCTSIZE*3, quantptrDCTSIZE*3) */
+    add             v26.8h, v22.8h, v24.8h  /* z5 = z3 + z4 */
+
+    smull2          v11.4s, v9.8h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull2          v13.4s, v7.8h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull2          v15.4s, v5.8h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull2          v17.4s, v3.8h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull2          v27.4s, v26.8h, XFIX_P_1_175  /* z5h z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull2          v23.4s, v22.8h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, -FIX_1_961570560) */
+    smull2          v25.4s, v24.8h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, -FIX_0_390180644) */
+    smull2          v19.4s, v18.8h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, -FIX_0_899976223) */
+    smull2          v21.4s, v20.8h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, -FIX_2_562915447) */
+
+    add             v23.4s, v23.4s, v27.4s  /* z3 += z5 */
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v25.4s, v25.4s, v27.4s  /* z4 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v11.4s, v11.4s, v19.4s  /* tmp0 += z1 */
+    add             v13.4s, v13.4s, v21.4s  /* tmp1 += z2 */
+    add             v15.4s, v15.4s, v21.4s  /* tmp2 += z2 */
+    add             v17.4s, v17.4s, v19.4s  /* tmp3 += z1 */
+
+    add             v11.4s, v11.4s, v23.4s  /* tmp0 += z3 */
+    add             v13.4s, v13.4s, v25.4s  /* tmp1 += z4 */
+    add             v17.4s, v17.4s, v25.4s  /* tmp3 += z4 */
+    add             v15.4s, v15.4s, v23.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v19.4s, v28.4s, v17.4s  /* tmp10 + tmp3 */
+    sub             v21.4s, v28.4s, v17.4s  /* tmp10 - tmp3 */
+    add             v23.4s, v29.4s, v15.4s  /* tmp11 + tmp2 */
+    sub             v25.4s, v29.4s, v15.4s  /* tmp11 - tmp2 */
+    add             v27.4s, v30.4s, v13.4s  /* tmp12 + tmp1 */
+    sub             v29.4s, v30.4s, v13.4s  /* tmp12 - tmp1 */
+    add             v15.4s, v31.4s, v11.4s  /* tmp13 + tmp0 */
+    sub             v17.4s, v31.4s, v11.4s  /* tmp13 - tmp0 */
+
+    mov             v2.16b, v14.16b
+    mov             v3.16b, v14.16b
+    mov             v4.16b, v14.16b
+    mov             v5.16b, v14.16b
+    rshrn           v6.4h, v19.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*0 = (int)DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v7.4h, v23.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*1 = (int)DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v8.4h, v27.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*2 = (int)DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v9.4h, v15.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*3 = (int)DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v6.8h, v17.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*4 = (int)DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v7.8h, v29.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*5 = (int)DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v8.8h, v25.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*6 = (int)DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v9.8h, v21.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*7 = (int)DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    b               1b
+
+.balign 16
+4:
+    /* "No" AC coef is zero */
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.8h, v4.8h, v8.8h           /* z2 + z3 = DEQUANTIZE(inptrDCTSIZE*2, quantptrDCTSIZE*2) + DEQUANTIZE(inptrDCTSIZE*6, quantptrDCTSIZE*6) */
+    add             v22.8h, v2.8h, v6.8h           /* z2 + z3 = DEQUANTIZE(inptrDCTSIZE*0, quantptrDCTSIZE*0) + DEQUANTIZE(inptrDCTSIZE*4, quantptrDCTSIZE*4) */
+    smull2          v19.4s, v18.8h, XFIX_P_0_541   /* z1h z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sub             v26.8h, v2.8h, v6.8h           /* z2 - z3 = DEQUANTIZE(inptrDCTSIZE*0, quantptrDCTSIZE*0) - DEQUANTIZE(inptrDCTSIZE*4, quantptrDCTSIZE*4) */
+    smull           v18.4s, v18.4h, XFIX_P_0_541   /* z1l z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sshll2          v23.4s, v22.8h, #(CONST_BITS)  /* tmp0h tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v21.16b, v19.16b               /* tmp3 = z1 */
+    mov             v20.16b, v18.16b               /* tmp3 = z1 */
+    smlal2          v19.4s, v8.8h, XFIX_N_1_847    /* tmp2h tmp2 = z1 + MULTIPLY(z3, -FIX_1_847759065); */
+    smlal           v18.4s, v8.4h, XFIX_N_1_847    /* tmp2l tmp2 = z1 + MULTIPLY(z3, -FIX_1_847759065); */
+    sshll2          v27.4s, v26.8h, #(CONST_BITS)  /* tmp1h tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal2          v21.4s, v4.8h, XFIX_P_0_765    /* tmp3h tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    smlal           v20.4s, v4.4h, XFIX_P_0_765    /* tmp3l tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    sshll           v22.4s, v22.4h, #(CONST_BITS)  /* tmp0l tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    sshll           v26.4s, v26.4h, #(CONST_BITS)  /* tmp1l tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    add             v2.4s, v22.4s, v20.4s          /* tmp10l tmp10 = tmp0 + tmp3; */
+    sub             v6.4s, v22.4s, v20.4s          /* tmp13l tmp13 = tmp0 - tmp3; */
+    add             v8.4s, v26.4s, v18.4s          /* tmp11l tmp11 = tmp1 + tmp2; */
+    sub             v4.4s, v26.4s, v18.4s          /* tmp12l tmp12 = tmp1 - tmp2; */
+    add             v28.4s, v23.4s, v21.4s         /* tmp10h tmp10 = tmp0 + tmp3; */
+    sub             v31.4s, v23.4s, v21.4s         /* tmp13h tmp13 = tmp0 - tmp3; */
+    add             v29.4s, v27.4s, v19.4s         /* tmp11h tmp11 = tmp1 + tmp2; */
+    sub             v30.4s, v27.4s, v19.4s         /* tmp12h tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.8h, v9.8h, v5.8h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptrDCTSIZE*7, quantptrDCTSIZE*7) + DEQUANTIZE(inptrDCTSIZE*3, quantptrDCTSIZE*3) */
+    add             v24.8h, v7.8h, v3.8h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptrDCTSIZE*5, quantptrDCTSIZE*5) + DEQUANTIZE(inptrDCTSIZE*1, quantptrDCTSIZE*1) */
+    add             v18.8h, v9.8h, v3.8h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptrDCTSIZE*7, quantptrDCTSIZE*7) + DEQUANTIZE(inptrDCTSIZE*1, quantptrDCTSIZE*1) */
+    add             v20.8h, v7.8h, v5.8h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptrDCTSIZE*5, quantptrDCTSIZE*5) + DEQUANTIZE(inptrDCTSIZE*3, quantptrDCTSIZE*3) */
+    add             v26.8h, v22.8h, v24.8h  /* z5 = z3 + z4 */
+
+    smull2          v11.4s, v9.8h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull2          v13.4s, v7.8h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull2          v15.4s, v5.8h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull2          v17.4s, v3.8h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull2          v27.4s, v26.8h, XFIX_P_1_175  /* z5h z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull2          v23.4s, v22.8h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, -FIX_1_961570560) */
+    smull2          v25.4s, v24.8h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, -FIX_0_390180644) */
+    smull2          v19.4s, v18.8h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, -FIX_0_899976223) */
+    smull2          v21.4s, v20.8h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, -FIX_2_562915447) */
+
+    smull           v10.4s, v9.4h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull           v12.4s, v7.4h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull           v14.4s, v5.4h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull           v16.4s, v3.4h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull           v26.4s, v26.4h, XFIX_P_1_175  /* z5l z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull           v22.4s, v22.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, -FIX_1_961570560) */
+    smull           v24.4s, v24.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, -FIX_0_390180644) */
+    smull           v18.4s, v18.4h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, -FIX_0_899976223) */
+    smull           v20.4s, v20.4h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, -FIX_2_562915447) */
+
+    add             v23.4s, v23.4s, v27.4s  /* z3 += z5 */
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v25.4s, v25.4s, v27.4s  /* z4 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v11.4s, v11.4s, v19.4s  /* tmp0 += z1 */
+    add             v10.4s, v10.4s, v18.4s  /* tmp0 += z1 */
+    add             v13.4s, v13.4s, v21.4s  /* tmp1 += z2 */
+    add             v12.4s, v12.4s, v20.4s  /* tmp1 += z2 */
+    add             v15.4s, v15.4s, v21.4s  /* tmp2 += z2 */
+    add             v14.4s, v14.4s, v20.4s  /* tmp2 += z2 */
+    add             v17.4s, v17.4s, v19.4s  /* tmp3 += z1 */
+    add             v16.4s, v16.4s, v18.4s  /* tmp3 += z1 */
+
+    add             v11.4s, v11.4s, v23.4s  /* tmp0 += z3 */
+    add             v10.4s, v10.4s, v22.4s  /* tmp0 += z3 */
+    add             v13.4s, v13.4s, v25.4s  /* tmp1 += z4 */
+    add             v12.4s, v12.4s, v24.4s  /* tmp1 += z4 */
+    add             v17.4s, v17.4s, v25.4s  /* tmp3 += z4 */
+    add             v16.4s, v16.4s, v24.4s  /* tmp3 += z4 */
+    add             v15.4s, v15.4s, v23.4s  /* tmp2 += z3 */
+    add             v14.4s, v14.4s, v22.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v18.4s, v2.4s, v16.4s   /* tmp10 + tmp3 */
+    add             v19.4s, v28.4s, v17.4s  /* tmp10 + tmp3 */
+    sub             v20.4s, v2.4s, v16.4s   /* tmp10 - tmp3 */
+    sub             v21.4s, v28.4s, v17.4s  /* tmp10 - tmp3 */
+    add             v22.4s, v8.4s, v14.4s   /* tmp11 + tmp2 */
+    add             v23.4s, v29.4s, v15.4s  /* tmp11 + tmp2 */
+    sub             v24.4s, v8.4s, v14.4s   /* tmp11 - tmp2 */
+    sub             v25.4s, v29.4s, v15.4s  /* tmp11 - tmp2 */
+    add             v26.4s, v4.4s, v12.4s   /* tmp12 + tmp1 */
+    add             v27.4s, v30.4s, v13.4s  /* tmp12 + tmp1 */
+    sub             v28.4s, v4.4s, v12.4s   /* tmp12 - tmp1 */
+    sub             v29.4s, v30.4s, v13.4s  /* tmp12 - tmp1 */
+    add             v14.4s, v6.4s, v10.4s   /* tmp13 + tmp0 */
+    add             v15.4s, v31.4s, v11.4s  /* tmp13 + tmp0 */
+    sub             v16.4s, v6.4s, v10.4s   /* tmp13 - tmp0 */
+    sub             v17.4s, v31.4s, v11.4s  /* tmp13 - tmp0 */
+
+    rshrn           v2.4h, v18.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*0 = (int)DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v3.4h, v22.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*1 = (int)DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v4.4h, v26.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*2 = (int)DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v5.4h, v14.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*3 = (int)DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn           v6.4h, v19.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*0 = (int)DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v7.4h, v23.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*1 = (int)DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v8.4h, v27.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*2 = (int)DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v9.4h, v15.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*3 = (int)DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v2.8h, v16.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*4 = (int)DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v3.8h, v28.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*5 = (int)DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v4.8h, v24.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*6 = (int)DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v5.8h, v20.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*7 = (int)DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn2          v6.8h, v17.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*4 = (int)DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v7.8h, v29.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*5 = (int)DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v8.8h, v25.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*6 = (int)DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v9.8h, v21.4s, #(CONST_BITS - PASS1_BITS)  /* wsptrDCTSIZE*7 = (int)DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    b               1b
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+    .unreq          TMP5
+    .unreq          TMP6
+    .unreq          TMP7
+    .unreq          TMP8
+
+#undef CENTERJSAMPLE
+#undef CONST_BITS
+#undef PASS1_BITS
+#undef XFIX_P_0_298
+#undef XFIX_N_0_390
+#undef XFIX_P_0_541
+#undef XFIX_P_0_765
+#undef XFIX_N_0_899
+#undef XFIX_P_1_175
+#undef XFIX_P_1_501
+#undef XFIX_N_1_847
+#undef XFIX_N_1_961
+#undef XFIX_P_2_053
+#undef XFIX_N_2_562
+#undef XFIX_P_3_072
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_ycc_extrgb_convert_neon
+ * jsimd_ycc_extbgr_convert_neon
+ * jsimd_ycc_extrgbx_convert_neon
+ * jsimd_ycc_extbgrx_convert_neon
+ * jsimd_ycc_extxbgr_convert_neon
+ * jsimd_ycc_extxrgb_convert_neon
+ *
+ * Colorspace conversion YCbCr -> RGB
+ */
+
+.macro do_load size
+  .if \size == 8
+    ld1             {v4.8b}, U, 8
+    ld1             {v5.8b}, V, 8
+    ld1             {v0.8b}, Y, 8
+    prfm            pldl1keep, U, #64
+    prfm            pldl1keep, V, #64
+    prfm            pldl1keep, Y, #64
+  .elseif \size == 4
+    ld1             {v4.b}0, U, 1
+    ld1             {v4.b}1, U, 1
+    ld1             {v4.b}2, U, 1
+    ld1             {v4.b}3, U, 1
+    ld1             {v5.b}0, V, 1
+    ld1             {v5.b}1, V, 1
+    ld1             {v5.b}2, V, 1
+    ld1             {v5.b}3, V, 1
+    ld1             {v0.b}0, Y, 1
+    ld1             {v0.b}1, Y, 1
+    ld1             {v0.b}2, Y, 1
+    ld1             {v0.b}3, Y, 1
+  .elseif \size == 2
+    ld1             {v4.b}4, U, 1
+    ld1             {v4.b}5, U, 1
+    ld1             {v5.b}4, V, 1
+    ld1             {v5.b}5, V, 1
+    ld1             {v0.b}4, Y, 1
+    ld1             {v0.b}5, Y, 1
+  .elseif \size == 1
+    ld1             {v4.b}6, U, 1
+    ld1             {v5.b}6, V, 1
+    ld1             {v0.b}6, Y, 1
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_store bpp, size, fast_st3
+  .if \bpp == 24
+    .if \size == 8
+      .if \fast_st3 == 1
+        st3         {v10.8b, v11.8b, v12.8b}, RGB, 24
+      .else
+        st1         {v10.b}0, RGB, #1
+        st1         {v11.b}0, RGB, #1
+        st1         {v12.b}0, RGB, #1
+
+        st1         {v10.b}1, RGB, #1
+        st1         {v11.b}1, RGB, #1
+        st1         {v12.b}1, RGB, #1
+
+        st1         {v10.b}2, RGB, #1
+        st1         {v11.b}2, RGB, #1
+        st1         {v12.b}2, RGB, #1
+
+        st1         {v10.b}3, RGB, #1
+        st1         {v11.b}3, RGB, #1
+        st1         {v12.b}3, RGB, #1
+
+        st1         {v10.b}4, RGB, #1
+        st1         {v11.b}4, RGB, #1
+        st1         {v12.b}4, RGB, #1
+
+        st1         {v10.b}5, RGB, #1
+        st1         {v11.b}5, RGB, #1
+        st1         {v12.b}5, RGB, #1
+
+        st1         {v10.b}6, RGB, #1
+        st1         {v11.b}6, RGB, #1
+        st1         {v12.b}6, RGB, #1
+
+        st1         {v10.b}7, RGB, #1
+        st1         {v11.b}7, RGB, #1
+        st1         {v12.b}7, RGB, #1
+      .endif
+    .elseif \size == 4
+      st3           {v10.b, v11.b, v12.b}0, RGB, 3
+      st3           {v10.b, v11.b, v12.b}1, RGB, 3
+      st3           {v10.b, v11.b, v12.b}2, RGB, 3
+      st3           {v10.b, v11.b, v12.b}3, RGB, 3
+    .elseif \size == 2
+      st3           {v10.b, v11.b, v12.b}4, RGB, 3
+      st3           {v10.b, v11.b, v12.b}5, RGB, 3
+    .elseif \size == 1
+      st3           {v10.b, v11.b, v12.b}6, RGB, 3
+    .else
+     .error unsupported macroblock size
+    .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      st4           {v10.8b, v11.8b, v12.8b, v13.8b}, RGB, 32
+    .elseif \size == 4
+      st4           {v10.b, v11.b, v12.b, v13.b}0, RGB, 4
+      st4           {v10.b, v11.b, v12.b, v13.b}1, RGB, 4
+      st4           {v10.b, v11.b, v12.b, v13.b}2, RGB, 4
+      st4           {v10.b, v11.b, v12.b, v13.b}3, RGB, 4
+    .elseif \size == 2
+      st4           {v10.b, v11.b, v12.b, v13.b}4, RGB, 4
+      st4           {v10.b, v11.b, v12.b, v13.b}5, RGB, 4
+    .elseif \size == 1
+      st4           {v10.b, v11.b, v12.b, v13.b}6, RGB, 4
+    .else
+      .error unsupported macroblock size
+    .endif
+  .elseif \bpp == 16
+    .if \size == 8
+      st1           {v25.8h}, RGB, 16
+    .elseif \size == 4
+      st1           {v25.4h}, RGB, 8
+    .elseif \size == 2
+      st1           {v25.h}4, RGB, 2
+      st1           {v25.h}5, RGB, 2
+    .elseif \size == 1
+      st1           {v25.h}6, RGB, 2
+    .else
+      .error unsupported macroblock size
+    .endif
+  .else
+    .error unsupported bpp
+  .endif
+.endm
+
+.macro generate_jsimd_ycc_rgb_convert_neon colorid, bpp, r_offs, rsize, \
+                                           g_offs, gsize, b_offs, bsize, \
+                                           defsize, fast_st3
+
+/*
+ * 2-stage pipelined YCbCr->RGB conversion
+ */
+
+.macro do_yuv_to_rgb_stage1
+    uaddw           v6.8h, v2.8h, v4.8b     /* q3 = u - 128 */
+    uaddw           v8.8h, v2.8h, v5.8b     /* q2 = v - 128 */
+    smull           v20.4s, v6.4h, v1.h1  /* multiply by -11277 */
+    smlal           v20.4s, v8.4h, v1.h2  /* multiply by -23401 */
+    smull2          v22.4s, v6.8h, v1.h1  /* multiply by -11277 */
+    smlal2          v22.4s, v8.8h, v1.h2  /* multiply by -23401 */
+    smull           v24.4s, v8.4h, v1.h0  /* multiply by 22971 */
+    smull2          v26.4s, v8.8h, v1.h0  /* multiply by 22971 */
+    smull           v28.4s, v6.4h, v1.h3  /* multiply by 29033 */
+    smull2          v30.4s, v6.8h, v1.h3  /* multiply by 29033 */
+.endm
+
+.macro do_yuv_to_rgb_stage2
+    rshrn           v20.4h, v20.4s, #15
+    rshrn2          v20.8h, v22.4s, #15
+    rshrn           v24.4h, v24.4s, #14
+    rshrn2          v24.8h, v26.4s, #14
+    rshrn           v28.4h, v28.4s, #14
+    rshrn2          v28.8h, v30.4s, #14
+    uaddw           v20.8h, v20.8h, v0.8b
+    uaddw           v24.8h, v24.8h, v0.8b
+    uaddw           v28.8h, v28.8h, v0.8b
+  .if \bpp != 16
+    sqxtun          v1\g_offs\defsize, v20.8h
+    sqxtun          v1\r_offs\defsize, v24.8h
+    sqxtun          v1\b_offs\defsize, v28.8h
+  .else
+    sqshlu          v21.8h, v20.8h, #8
+    sqshlu          v25.8h, v24.8h, #8
+    sqshlu          v29.8h, v28.8h, #8
+    sri             v25.8h, v21.8h, #5
+    sri             v25.8h, v29.8h, #11
+  .endif
+.endm
+
+.macro do_yuv_to_rgb_stage2_store_load_stage1 fast_st3
+    rshrn           v20.4h, v20.4s, #15
+    rshrn           v24.4h, v24.4s, #14
+    rshrn           v28.4h, v28.4s, #14
+    ld1             {v4.8b}, U, 8
+    rshrn2          v20.8h, v22.4s, #15
+    rshrn2          v24.8h, v26.4s, #14
+    rshrn2          v28.8h, v30.4s, #14
+    ld1             {v5.8b}, V, 8
+    uaddw           v20.8h, v20.8h, v0.8b
+    uaddw           v24.8h, v24.8h, v0.8b
+    uaddw           v28.8h, v28.8h, v0.8b
+  .if \bpp != 16  /**************** rgb24/rgb32 ******************************/
+    sqxtun          v1\g_offs\defsize, v20.8h
+    ld1             {v0.8b}, Y, 8
+    sqxtun          v1\r_offs\defsize, v24.8h
+    prfm            pldl1keep, U, #64
+    prfm            pldl1keep, V, #64
+    prfm            pldl1keep, Y, #64
+    sqxtun          v1\b_offs\defsize, v28.8h
+    uaddw           v6.8h, v2.8h, v4.8b     /* v6.16b = u - 128 */
+    uaddw           v8.8h, v2.8h, v5.8b     /* q2 = v - 128 */
+    smull           v20.4s, v6.4h, v1.h1  /* multiply by -11277 */
+    smlal           v20.4s, v8.4h, v1.h2  /* multiply by -23401 */
+    smull2          v22.4s, v6.8h, v1.h1  /* multiply by -11277 */
+    smlal2          v22.4s, v8.8h, v1.h2  /* multiply by -23401 */
+    smull           v24.4s, v8.4h, v1.h0  /* multiply by 22971 */
+    smull2          v26.4s, v8.8h, v1.h0  /* multiply by 22971 */
+  .else  /**************************** rgb565 ********************************/
+    sqshlu          v21.8h, v20.8h, #8
+    sqshlu          v25.8h, v24.8h, #8
+    sqshlu          v29.8h, v28.8h, #8
+    uaddw           v6.8h, v2.8h, v4.8b     /* v6.16b = u - 128 */
+    uaddw           v8.8h, v2.8h, v5.8b     /* q2 = v - 128 */
+    ld1             {v0.8b}, Y, 8
+    smull           v20.4s, v6.4h, v1.h1  /* multiply by -11277 */
+    smlal           v20.4s, v8.4h, v1.h2  /* multiply by -23401 */
+    smull2          v22.4s, v6.8h, v1.h1  /* multiply by -11277 */
+    smlal2          v22.4s, v8.8h, v1.h2  /* multiply by -23401 */
+    sri             v25.8h, v21.8h, #5
+    smull           v24.4s, v8.4h, v1.h0  /* multiply by 22971 */
+    smull2          v26.4s, v8.8h, v1.h0  /* multiply by 22971 */
+    prfm            pldl1keep, U, #64
+    prfm            pldl1keep, V, #64
+    prfm            pldl1keep, Y, #64
+    sri             v25.8h, v29.8h, #11
+  .endif
+    do_store        \bpp, 8, \fast_st3
+    smull           v28.4s, v6.4h, v1.h3  /* multiply by 29033 */
+    smull2          v30.4s, v6.8h, v1.h3  /* multiply by 29033 */
+.endm
+
+.macro do_yuv_to_rgb
+    do_yuv_to_rgb_stage1
+    do_yuv_to_rgb_stage2
+.endm
+
+.if \fast_st3 == 1
+asm_function jsimd_ycc_\colorid\()_convert_neon
+.else
+asm_function jsimd_ycc_\colorid\()_convert_neon_slowst3
+.endif
+    OUTPUT_WIDTH    .req w0
+    INPUT_BUF       .req x1
+    INPUT_ROW       .req w2
+    OUTPUT_BUF      .req x3
+    NUM_ROWS        .req w4
+
+    INPUT_BUF0      .req x5
+    INPUT_BUF1      .req x6
+    INPUT_BUF2      .req x1
+
+    RGB             .req x7
+    Y               .req x9
+    U               .req x10
+    V               .req x11
+    N               .req w15
+
+    sub             sp, sp, 64
+    mov             x9, sp
+
+    /* Load constants to d1, d2, d3 (v0.4h is just used for padding) */
+    get_symbol_loc  x15, Ljsimd_ycc_rgb_neon_consts
+
+    /* Save Neon registers */
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, x9, 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, x9, 32
+    ld1             {v0.4h, v1.4h}, x15, 16
+    ld1             {v2.8h}, x15
+
+    ldr             INPUT_BUF0, INPUT_BUF
+    ldr             INPUT_BUF1, INPUT_BUF, #8
+    ldr             INPUT_BUF2, INPUT_BUF, #16
+    .unreq          INPUT_BUF
+
+    /* Initially set v10, v11.4h, v12.8b, d13 to 0xFF */
+    movi            v10.16b, #255
+    movi            v13.16b, #255
+
+    /* Outer loop over scanlines */
+    cmp             NUM_ROWS, #1
+    b.lt            9f
+0:
+    ldr             Y, INPUT_BUF0, INPUT_ROW, uxtw #3
+    ldr             U, INPUT_BUF1, INPUT_ROW, uxtw #3
+    mov             N, OUTPUT_WIDTH
+    ldr             V, INPUT_BUF2, INPUT_ROW, uxtw #3
+    add             INPUT_ROW, INPUT_ROW, #1
+    ldr             RGB, OUTPUT_BUF, #8
+
+    /* Inner loop over pixels */
+    subs            N, N, #8
+    b.lt            3f
+    do_load         8
+    do_yuv_to_rgb_stage1
+    subs            N, N, #8
+    b.lt            2f
+1:
+    do_yuv_to_rgb_stage2_store_load_stage1 \fast_st3
+    subs            N, N, #8
+    b.ge            1b
+2:
+    do_yuv_to_rgb_stage2
+    do_store        \bpp, 8, \fast_st3
+    tst             N, #7
+    b.eq            8f
+3:
+    tst             N, #4
+    b.eq            3f
+    do_load         4
+3:
+    tst             N, #2
+    b.eq            4f
+    do_load         2
+4:
+    tst             N, #1
+    b.eq            5f
+    do_load         1
+5:
+    do_yuv_to_rgb
+    tst             N, #4
+    b.eq            6f
+    do_store        \bpp, 4, \fast_st3
+6:
+    tst             N, #2
+    b.eq            7f
+    do_store        \bpp, 2, \fast_st3
+7:
+    tst             N, #1
+    b.eq            8f
+    do_store        \bpp, 1, \fast_st3
+8:
+    subs            NUM_ROWS, NUM_ROWS, #1
+    b.gt            0b
+9:
+    /* Restore all registers and return */
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, sp, 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, sp, 32
+    br              x30
+    .unreq          OUTPUT_WIDTH
+    .unreq          INPUT_ROW
+    .unreq          OUTPUT_BUF
+    .unreq          NUM_ROWS
+    .unreq          INPUT_BUF0
+    .unreq          INPUT_BUF1
+    .unreq          INPUT_BUF2
+    .unreq          RGB
+    .unreq          Y
+    .unreq          U
+    .unreq          V
+    .unreq          N
+
+.purgem do_yuv_to_rgb
+.purgem do_yuv_to_rgb_stage1
+.purgem do_yuv_to_rgb_stage2
+.purgem do_yuv_to_rgb_stage2_store_load_stage1
+
+.endm
+
+/*--------------------------------- id ----- bpp R  rsize G  gsize B  bsize defsize fast_st3*/
+generate_jsimd_ycc_rgb_convert_neon extrgb,  24, 0, .4h,  1, .4h,  2, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extbgr,  24, 2, .4h,  1, .4h,  0, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extrgbx, 32, 0, .4h,  1, .4h,  2, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extbgrx, 32, 2, .4h,  1, .4h,  0, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extxbgr, 32, 3, .4h,  2, .4h,  1, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extxrgb, 32, 1, .4h,  2, .4h,  3, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon rgb565,  16, 0, .4h,  0, .4h,  0, .4h,  .8b,    1
+
+generate_jsimd_ycc_rgb_convert_neon extrgb,  24, 0, .4h,  1, .4h,  2, .4h,  .8b,    0
+generate_jsimd_ycc_rgb_convert_neon extbgr,  24, 2, .4h,  1, .4h,  0, .4h,  .8b,    0
+
+.purgem do_load
+.purgem do_store
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_extrgb_ycc_convert_neon
+ * jsimd_extbgr_ycc_convert_neon
+ * jsimd_extrgbx_ycc_convert_neon
+ * jsimd_extbgrx_ycc_convert_neon
+ * jsimd_extxbgr_ycc_convert_neon
+ * jsimd_extxrgb_ycc_convert_neon
+ *
+ * Colorspace conversion RGB -> YCbCr
+ */
+
+.macro do_store size
+  .if \size == 8
+    st1             {v20.8b}, Y, #8
+    st1             {v21.8b}, U, #8
+    st1             {v22.8b}, V, #8
+  .elseif \size == 4
+    st1             {v20.b}0, Y, #1
+    st1             {v20.b}1, Y, #1
+    st1             {v20.b}2, Y, #1
+    st1             {v20.b}3, Y, #1
+    st1             {v21.b}0, U, #1
+    st1             {v21.b}1, U, #1
+    st1             {v21.b}2, U, #1
+    st1             {v21.b}3, U, #1
+    st1             {v22.b}0, V, #1
+    st1             {v22.b}1, V, #1
+    st1             {v22.b}2, V, #1
+    st1             {v22.b}3, V, #1
+  .elseif \size == 2
+    st1             {v20.b}4, Y, #1
+    st1             {v20.b}5, Y, #1
+    st1             {v21.b}4, U, #1
+    st1             {v21.b}5, U, #1
+    st1             {v22.b}4, V, #1
+    st1             {v22.b}5, V, #1
+  .elseif \size == 1
+    st1             {v20.b}6, Y, #1
+    st1             {v21.b}6, U, #1
+    st1             {v22.b}6, V, #1
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_load bpp, size, fast_ld3
+  .if \bpp == 24
+    .if \size == 8
+      .if \fast_ld3 == 1
+        ld3         {v10.8b, v11.8b, v12.8b}, RGB, #24
+      .else
+        ld1         {v10.b}0, RGB, #1
+        ld1         {v11.b}0, RGB, #1
+        ld1         {v12.b}0, RGB, #1
+
+        ld1         {v10.b}1, RGB, #1
+        ld1         {v11.b}1, RGB, #1
+        ld1         {v12.b}1, RGB, #1
+
+        ld1         {v10.b}2, RGB, #1
+        ld1         {v11.b}2, RGB, #1
+        ld1         {v12.b}2, RGB, #1
+
+        ld1         {v10.b}3, RGB, #1
+        ld1         {v11.b}3, RGB, #1
+        ld1         {v12.b}3, RGB, #1
+
+        ld1         {v10.b}4, RGB, #1
+        ld1         {v11.b}4, RGB, #1
+        ld1         {v12.b}4, RGB, #1
+
+        ld1         {v10.b}5, RGB, #1
+        ld1         {v11.b}5, RGB, #1
+        ld1         {v12.b}5, RGB, #1
+
+        ld1         {v10.b}6, RGB, #1
+        ld1         {v11.b}6, RGB, #1
+        ld1         {v12.b}6, RGB, #1
+
+        ld1         {v10.b}7, RGB, #1
+        ld1         {v11.b}7, RGB, #1
+        ld1         {v12.b}7, RGB, #1
+      .endif
+      prfm          pldl1keep, RGB, #128
+    .elseif \size == 4
+      ld3           {v10.b, v11.b, v12.b}0, RGB, #3
+      ld3           {v10.b, v11.b, v12.b}1, RGB, #3
+      ld3           {v10.b, v11.b, v12.b}2, RGB, #3
+      ld3           {v10.b, v11.b, v12.b}3, RGB, #3
+    .elseif \size == 2
+      ld3           {v10.b, v11.b, v12.b}4, RGB, #3
+      ld3           {v10.b, v11.b, v12.b}5, RGB, #3
+    .elseif \size == 1
+      ld3           {v10.b, v11.b, v12.b}6, RGB, #3
+    .else
+      .error unsupported macroblock size
+    .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      ld4           {v10.8b, v11.8b, v12.8b, v13.8b}, RGB, #32
+      prfm          pldl1keep, RGB, #128
+    .elseif \size == 4
+      ld4           {v10.b, v11.b, v12.b, v13.b}0, RGB, #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}1, RGB, #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}2, RGB, #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}3, RGB, #4
+    .elseif \size == 2
+      ld4           {v10.b, v11.b, v12.b, v13.b}4, RGB, #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}5, RGB, #4
+    .elseif \size == 1
+      ld4           {v10.b, v11.b, v12.b, v13.b}6, RGB, #4
+    .else
+      .error unsupported macroblock size
+    .endif
+  .else
+    .error unsupported bpp
+  .endif
+.endm
+
+.macro generate_jsimd_rgb_ycc_convert_neon colorid, bpp, r_offs, g_offs, \
+                                           b_offs, fast_ld3
+
+/*
+ * 2-stage pipelined RGB->YCbCr conversion
+ */
+
+.macro do_rgb_to_yuv_stage1
+    ushll           v4.8h, v1\r_offs\().8b, #0  /* r = v4 */
+    ushll           v6.8h, v1\g_offs\().8b, #0  /* g = v6 */
+    ushll           v8.8h, v1\b_offs\().8b, #0  /* b = v8 */
+    rev64           v18.4s, v1.4s
+    rev64           v26.4s, v1.4s
+    rev64           v28.4s, v1.4s
+    rev64           v30.4s, v1.4s
+    umull           v14.4s, v4.4h, v0.h0
+    umull2          v16.4s, v4.8h, v0.h0
+    umlsl           v18.4s, v4.4h, v0.h3
+    umlsl2          v26.4s, v4.8h, v0.h3
+    umlal           v28.4s, v4.4h, v0.h5
+    umlal2          v30.4s, v4.8h, v0.h5
+    umlal           v14.4s, v6.4h, v0.h1
+    umlal2          v16.4s, v6.8h, v0.h1
+    umlsl           v18.4s, v6.4h, v0.h4
+    umlsl2          v26.4s, v6.8h, v0.h4
+    umlsl           v28.4s, v6.4h, v0.h6
+    umlsl2          v30.4s, v6.8h, v0.h6
+    umlal           v14.4s, v8.4h, v0.h2
+    umlal2          v16.4s, v8.8h, v0.h2
+    umlal           v18.4s, v8.4h, v0.h5
+    umlal2          v26.4s, v8.8h, v0.h5
+    umlsl           v28.4s, v8.4h, v0.h7
+    umlsl2          v30.4s, v8.8h, v0.h7
+.endm
+
+.macro do_rgb_to_yuv_stage2
+    rshrn           v20.4h, v14.4s, #16
+    shrn            v22.4h, v18.4s, #16
+    shrn            v24.4h, v28.4s, #16
+    rshrn2          v20.8h, v16.4s, #16
+    shrn2           v22.8h, v26.4s, #16
+    shrn2           v24.8h, v30.4s, #16
+    xtn             v20.8b, v20.8h       /* v20 = y */
+    xtn             v21.8b, v22.8h       /* v21 = u */
+    xtn             v22.8b, v24.8h       /* v22 = v */
+.endm
+
+.macro do_rgb_to_yuv
+    do_rgb_to_yuv_stage1
+    do_rgb_to_yuv_stage2
+.endm
+
+/* TODO: expand macros and interleave instructions if some in-order
+ *       AArch64 processor actually can dual-issue LOAD/STORE with ALU */
+.macro do_rgb_to_yuv_stage2_store_load_stage1 fast_ld3
+    do_rgb_to_yuv_stage2
+    do_load         \bpp, 8, \fast_ld3
+    st1             {v20.8b}, Y, #8
+    st1             {v21.8b}, U, #8
+    st1             {v22.8b}, V, #8
+    do_rgb_to_yuv_stage1
+.endm
+
+.if \fast_ld3 == 1
+asm_function jsimd_\colorid\()_ycc_convert_neon
+.else
+asm_function jsimd_\colorid\()_ycc_convert_neon_slowld3
+.endif
+    OUTPUT_WIDTH    .req w0
+    INPUT_BUF       .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_ROW      .req w3
+    NUM_ROWS        .req w4
+
+    OUTPUT_BUF0     .req x5
+    OUTPUT_BUF1     .req x6
+    OUTPUT_BUF2     .req x2  /* OUTPUT_BUF */
+
+    RGB             .req x7
+    Y               .req x9
+    U               .req x10
+    V               .req x11
+    N               .req w12
+
+    /* Load constants to d0, d1, d2, d3 */
+    get_symbol_loc  x13, Ljsimd_rgb_ycc_neon_consts
+    ld1             {v0.8h, v1.8h}, x13
+
+    ldr             OUTPUT_BUF0, OUTPUT_BUF
+    ldr             OUTPUT_BUF1, OUTPUT_BUF, #8
+    ldr             OUTPUT_BUF2, OUTPUT_BUF, #16
+    .unreq          OUTPUT_BUF
+
+    /* Save Neon registers */
+    sub             sp, sp, #64
+    mov             x9, sp
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, x9, 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, x9, 32
+
+    /* Outer loop over scanlines */
+    cmp             NUM_ROWS, #1
+    b.lt            9f
+0:
+    ldr             Y, OUTPUT_BUF0, OUTPUT_ROW, uxtw #3
+    ldr             U, OUTPUT_BUF1, OUTPUT_ROW, uxtw #3
+    mov             N, OUTPUT_WIDTH
+    ldr             V, OUTPUT_BUF2, OUTPUT_ROW, uxtw #3
+    add             OUTPUT_ROW, OUTPUT_ROW, #1
+    ldr             RGB, INPUT_BUF, #8
+
+    /* Inner loop over pixels */
+    subs            N, N, #8
+    b.lt            3f
+    do_load         \bpp, 8, \fast_ld3
+    do_rgb_to_yuv_stage1
+    subs            N, N, #8
+    b.lt            2f
+1:
+    do_rgb_to_yuv_stage2_store_load_stage1 \fast_ld3
+    subs            N, N, #8
+    b.ge            1b
+2:
+    do_rgb_to_yuv_stage2
+    do_store        8
+    tst             N, #7
+    b.eq            8f
+3:
+    tbz             N, #2, 3f
+    do_load         \bpp, 4, \fast_ld3
+3:
+    tbz             N, #1, 4f
+    do_load         \bpp, 2, \fast_ld3
+4:
+    tbz             N, #0, 5f
+    do_load         \bpp, 1, \fast_ld3
+5:
+    do_rgb_to_yuv
+    tbz             N, #2, 6f
+    do_store        4
+6:
+    tbz             N, #1, 7f
+    do_store        2
+7:
+    tbz             N, #0, 8f
+    do_store        1
+8:
+    subs            NUM_ROWS, NUM_ROWS, #1
+    b.gt            0b
+9:
+    /* Restore all registers and return */
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, sp, 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, sp, 32
+    br              x30
+
+    .unreq          OUTPUT_WIDTH
+    .unreq          OUTPUT_ROW
+    .unreq          INPUT_BUF
+    .unreq          NUM_ROWS
+    .unreq          OUTPUT_BUF0
+    .unreq          OUTPUT_BUF1
+    .unreq          OUTPUT_BUF2
+    .unreq          RGB
+    .unreq          Y
+    .unreq          U
+    .unreq          V
+    .unreq          N
+
+.purgem do_rgb_to_yuv
+.purgem do_rgb_to_yuv_stage1
+.purgem do_rgb_to_yuv_stage2
+.purgem do_rgb_to_yuv_stage2_store_load_stage1
+
+.endm
+
+/*--------------------------------- id ----- bpp R  G  B  Fast LD3 */
+generate_jsimd_rgb_ycc_convert_neon extrgb,  24, 0, 1, 2, 1
+generate_jsimd_rgb_ycc_convert_neon extbgr,  24, 2, 1, 0, 1
+generate_jsimd_rgb_ycc_convert_neon extrgbx, 32, 0, 1, 2, 1
+generate_jsimd_rgb_ycc_convert_neon extbgrx, 32, 2, 1, 0, 1
+generate_jsimd_rgb_ycc_convert_neon extxbgr, 32, 3, 2, 1, 1
+generate_jsimd_rgb_ycc_convert_neon extxrgb, 32, 1, 2, 3, 1
+
+generate_jsimd_rgb_ycc_convert_neon extrgb,  24, 0, 1, 2, 0
+generate_jsimd_rgb_ycc_convert_neon extbgr,  24, 2, 1, 0, 0
+
+.purgem do_load
+.purgem do_store
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_fdct_islow_neon
+ *
+ * This file contains a slower but more accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform). The following code is based
+ * directly on the IJG''s original jfdctint.c; see the jfdctint.c for
+ * more details.
+ *
+ * TODO: can be combined with 'jsimd_convsamp_neon' to get
+ *       rid of a bunch of VLD1.16 instructions
+ */
+
+#define CONST_BITS  13
+#define PASS1_BITS  2
+
+#define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2  (CONST_BITS + PASS1_BITS)
+
+#define XFIX_P_0_298  v0.h0
+#define XFIX_N_0_390  v0.h1
+#define XFIX_P_0_541  v0.h2
+#define XFIX_P_0_765  v0.h3
+#define XFIX_N_0_899  v0.h4
+#define XFIX_P_1_175  v0.h5
+#define XFIX_P_1_501  v0.h6
+#define XFIX_N_1_847  v0.h7
+#define XFIX_N_1_961  v1.h0
+#define XFIX_P_2_053  v1.h1
+#define XFIX_N_2_562  v1.h2
+#define XFIX_P_3_072  v1.h3
+
+asm_function jsimd_fdct_islow_neon
+
+    DATA            .req x0
+    TMP             .req x9
+
+    /* Load constants */
+    get_symbol_loc  TMP, Ljsimd_fdct_islow_neon_consts
+    ld1             {v0.8h, v1.8h}, TMP
+
+    /* Save Neon registers */
+    sub             sp, sp, #64
+    mov             x10, sp
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, x10, 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, x10, 32
+
+    /* Load all DATA into Neon registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d16     | d17    | v16.8h
+     *   1 | d18     | d19    | v17.8h
+     *   2 | d20     | d21    | v18.8h
+     *   3 | d22     | d23    | v19.8h
+     *   4 | d24     | d25    | v20.8h
+     *   5 | d26     | d27    | v21.8h
+     *   6 | d28     | d29    | v22.8h
+     *   7 | d30     | d31    | v23.8h
+     */
+
+    ld1             {v16.8h, v17.8h, v18.8h, v19.8h}, DATA, 64
+    ld1             {v20.8h, v21.8h, v22.8h, v23.8h}, DATA
+    sub             DATA, DATA, #64
+
+    /* Transpose */
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v31, v2, v3, v4
+    /* 1-D FDCT */
+    add             v24.8h, v16.8h, v23.8h  /* tmp0 = dataptr0 + dataptr7; */
+    sub             v31.8h, v16.8h, v23.8h  /* tmp7 = dataptr0 - dataptr7; */
+    add             v25.8h, v17.8h, v22.8h  /* tmp1 = dataptr1 + dataptr6; */
+    sub             v30.8h, v17.8h, v22.8h  /* tmp6 = dataptr1 - dataptr6; */
+    add             v26.8h, v18.8h, v21.8h  /* tmp2 = dataptr2 + dataptr5; */
+    sub             v29.8h, v18.8h, v21.8h  /* tmp5 = dataptr2 - dataptr5; */
+    add             v27.8h, v19.8h, v20.8h  /* tmp3 = dataptr3 + dataptr4; */
+    sub             v28.8h, v19.8h, v20.8h  /* tmp4 = dataptr3 - dataptr4; */
+
+    /* even part */
+
+    add             v8.8h, v24.8h, v27.8h   /* tmp10 = tmp0 + tmp3; */
+    sub             v9.8h, v24.8h, v27.8h   /* tmp13 = tmp0 - tmp3; */
+    add             v10.8h, v25.8h, v26.8h  /* tmp11 = tmp1 + tmp2; */
+    sub             v11.8h, v25.8h, v26.8h  /* tmp12 = tmp1 - tmp2; */
+
+    add             v16.8h, v8.8h, v10.8h  /* tmp10 + tmp11 */
+    sub             v20.8h, v8.8h, v10.8h  /* tmp10 - tmp11 */
+
+    add             v18.8h, v11.8h, v9.8h  /* tmp12 + tmp13 */
+
+    shl             v16.8h, v16.8h, #PASS1_BITS  /* dataptr0 = (DCTELEM)LEFT_SHIFT(tmp10 + tmp11, PASS1_BITS); */
+    shl             v20.8h, v20.8h, #PASS1_BITS  /* dataptr4 = (DCTELEM)LEFT_SHIFT(tmp10 - tmp11, PASS1_BITS); */
+
+    smull2          v24.4s, v18.8h, XFIX_P_0_541  /* z1 hi = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    smull           v18.4s, v18.4h, XFIX_P_0_541  /* z1 lo = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    mov             v22.16b, v18.16b
+    mov             v25.16b, v24.16b
+
+    smlal           v18.4s, v9.4h, XFIX_P_0_765   /* lo z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal2          v24.4s, v9.8h, XFIX_P_0_765   /* hi z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal           v22.4s, v11.4h, XFIX_N_1_847  /* lo z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+    smlal2          v25.4s, v11.8h, XFIX_N_1_847  /* hi z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+
+    rshrn           v18.4h, v18.4s, #DESCALE_P1
+    rshrn           v22.4h, v22.4s, #DESCALE_P1
+    rshrn2          v18.8h, v24.4s, #DESCALE_P1  /* dataptr2 = (DCTELEM)DESCALE(z1 + MULTIPLY(tmp13, XFIX_P_0_765), CONST_BITS-PASS1_BITS); */
+    rshrn2          v22.8h, v25.4s, #DESCALE_P1  /* dataptr6 = (DCTELEM)DESCALE(z1 + MULTIPLY(tmp12, XFIX_N_1_847), CONST_BITS-PASS1_BITS); */
+
+    /* Odd part */
+
+    add             v8.8h, v28.8h, v31.8h        /* z1 = tmp4 + tmp7; */
+    add             v9.8h, v29.8h, v30.8h        /* z2 = tmp5 + tmp6; */
+    add             v10.8h, v28.8h, v30.8h       /* z3 = tmp4 + tmp6; */
+    add             v11.8h, v29.8h, v31.8h       /* z4 = tmp5 + tmp7; */
+    smull           v4.4s, v10.4h, XFIX_P_1_175  /* z5 lo = z3 lo * XFIX_P_1_175 */
+    smull2          v5.4s, v10.8h, XFIX_P_1_175
+    smlal           v4.4s, v11.4h, XFIX_P_1_175  /* z5 = MULTIPLY(z3 + z4, FIX_1_175875602); */
+    smlal2          v5.4s, v11.8h, XFIX_P_1_175
+
+    smull2          v24.4s, v28.8h, XFIX_P_0_298
+    smull2          v25.4s, v29.8h, XFIX_P_2_053
+    smull2          v26.4s, v30.8h, XFIX_P_3_072
+    smull2          v27.4s, v31.8h, XFIX_P_1_501
+    smull           v28.4s, v28.4h, XFIX_P_0_298  /* tmp4 = MULTIPLY(tmp4, FIX_0_298631336); */
+    smull           v29.4s, v29.4h, XFIX_P_2_053  /* tmp5 = MULTIPLY(tmp5, FIX_2_053119869); */
+    smull           v30.4s, v30.4h, XFIX_P_3_072  /* tmp6 = MULTIPLY(tmp6, FIX_3_072711026); */
+    smull           v31.4s, v31.4h, XFIX_P_1_501  /* tmp7 = MULTIPLY(tmp7, FIX_1_501321110); */
+
+    smull2          v12.4s, v8.8h, XFIX_N_0_899
+    smull2          v13.4s, v9.8h, XFIX_N_2_562
+    smull2          v14.4s, v10.8h, XFIX_N_1_961
+    smull2          v15.4s, v11.8h, XFIX_N_0_390
+    smull           v8.4s, v8.4h, XFIX_N_0_899    /* z1 = MULTIPLY(z1, -FIX_0_899976223); */
+    smull           v9.4s, v9.4h, XFIX_N_2_562    /* z2 = MULTIPLY(z2, -FIX_2_562915447); */
+    smull           v10.4s, v10.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, -FIX_1_961570560); */
+    smull           v11.4s, v11.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, -FIX_0_390180644); */
+
+    add             v10.4s, v10.4s, v4.4s  /* z3 += z5 */
+    add             v14.4s, v14.4s, v5.4s
+    add             v11.4s, v11.4s, v4.4s  /* z4 += z5 */
+    add             v15.4s, v15.4s, v5.4s
+
+    add             v28.4s, v28.4s, v8.4s   /* tmp4 += z1 */
+    add             v24.4s, v24.4s, v12.4s
+    add             v29.4s, v29.4s, v9.4s   /* tmp5 += z2 */
+    add             v25.4s, v25.4s, v13.4s
+    add             v30.4s, v30.4s, v10.4s  /* tmp6 += z3 */
+    add             v26.4s, v26.4s, v14.4s
+    add             v31.4s, v31.4s, v11.4s  /* tmp7 += z4 */
+    add             v27.4s, v27.4s, v15.4s
+
+    add             v28.4s, v28.4s, v10.4s  /* tmp4 += z3 */
+    add             v24.4s, v24.4s, v14.4s
+    add             v29.4s, v29.4s, v11.4s  /* tmp5 += z4 */
+    add             v25.4s, v25.4s, v15.4s
+    add             v30.4s, v30.4s, v9.4s   /* tmp6 += z2 */
+    add             v26.4s, v26.4s, v13.4s
+    add             v31.4s, v31.4s, v8.4s   /* tmp7 += z1 */
+    add             v27.4s, v27.4s, v12.4s
+
+    rshrn           v23.4h, v28.4s, #DESCALE_P1
+    rshrn           v21.4h, v29.4s, #DESCALE_P1
+    rshrn           v19.4h, v30.4s, #DESCALE_P1
+    rshrn           v17.4h, v31.4s, #DESCALE_P1
+    rshrn2          v23.8h, v24.4s, #DESCALE_P1  /* dataptr7 = (DCTELEM)DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v21.8h, v25.4s, #DESCALE_P1  /* dataptr5 = (DCTELEM)DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS); */
+    rshrn2          v19.8h, v26.4s, #DESCALE_P1  /* dataptr3 = (DCTELEM)DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v17.8h, v27.4s, #DESCALE_P1  /* dataptr1 = (DCTELEM)DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS); */
+
+    /* Transpose */
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v31, v2, v3, v4
+
+    /* 1-D FDCT */
+    add             v24.8h, v16.8h, v23.8h  /* tmp0 = dataptr0 + dataptr7; */
+    sub             v31.8h, v16.8h, v23.8h  /* tmp7 = dataptr0 - dataptr7; */
+    add             v25.8h, v17.8h, v22.8h  /* tmp1 = dataptr1 + dataptr6; */
+    sub             v30.8h, v17.8h, v22.8h  /* tmp6 = dataptr1 - dataptr6; */
+    add             v26.8h, v18.8h, v21.8h  /* tmp2 = dataptr2 + dataptr5; */
+    sub             v29.8h, v18.8h, v21.8h  /* tmp5 = dataptr2 - dataptr5; */
+    add             v27.8h, v19.8h, v20.8h  /* tmp3 = dataptr3 + dataptr4; */
+    sub             v28.8h, v19.8h, v20.8h  /* tmp4 = dataptr3 - dataptr4; */
+
+    /* even part */
+    add             v8.8h, v24.8h, v27.8h   /* tmp10 = tmp0 + tmp3; */
+    sub             v9.8h, v24.8h, v27.8h   /* tmp13 = tmp0 - tmp3; */
+    add             v10.8h, v25.8h, v26.8h  /* tmp11 = tmp1 + tmp2; */
+    sub             v11.8h, v25.8h, v26.8h  /* tmp12 = tmp1 - tmp2; */
+
+    add             v16.8h, v8.8h, v10.8h  /* tmp10 + tmp11 */
+    sub             v20.8h, v8.8h, v10.8h  /* tmp10 - tmp11 */
+
+    add             v18.8h, v11.8h, v9.8h  /* tmp12 + tmp13 */
+
+    srshr           v16.8h, v16.8h, #PASS1_BITS  /* dataptr0 = (DCTELEM)DESCALE(tmp10 + tmp11, PASS1_BITS); */
+    srshr           v20.8h, v20.8h, #PASS1_BITS  /* dataptr4 = (DCTELEM)DESCALE(tmp10 - tmp11, PASS1_BITS); */
+
+    smull2          v24.4s, v18.8h, XFIX_P_0_541  /* z1 hi = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    smull           v18.4s, v18.4h, XFIX_P_0_541  /* z1 lo = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    mov             v22.16b, v18.16b
+    mov             v25.16b, v24.16b
+
+    smlal           v18.4s, v9.4h, XFIX_P_0_765   /* lo z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal2          v24.4s, v9.8h, XFIX_P_0_765   /* hi z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal           v22.4s, v11.4h, XFIX_N_1_847  /* lo z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+    smlal2          v25.4s, v11.8h, XFIX_N_1_847  /* hi z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+
+    rshrn           v18.4h, v18.4s, #DESCALE_P2
+    rshrn           v22.4h, v22.4s, #DESCALE_P2
+    rshrn2          v18.8h, v24.4s, #DESCALE_P2  /* dataptr2 = (DCTELEM)DESCALE(z1 + MULTIPLY(tmp13, XFIX_P_0_765), CONST_BITS-PASS1_BITS); */
+    rshrn2          v22.8h, v25.4s, #DESCALE_P2  /* dataptr6 = (DCTELEM)DESCALE(z1 + MULTIPLY(tmp12, XFIX_N_1_847), CONST_BITS-PASS1_BITS); */
+
+    /* Odd part */
+    add             v8.8h, v28.8h, v31.8h   /* z1 = tmp4 + tmp7; */
+    add             v9.8h, v29.8h, v30.8h   /* z2 = tmp5 + tmp6; */
+    add             v10.8h, v28.8h, v30.8h  /* z3 = tmp4 + tmp6; */
+    add             v11.8h, v29.8h, v31.8h  /* z4 = tmp5 + tmp7; */
+
+    smull           v4.4s, v10.4h, XFIX_P_1_175  /* z5 lo = z3 lo * XFIX_P_1_175 */
+    smull2          v5.4s, v10.8h, XFIX_P_1_175
+    smlal           v4.4s, v11.4h, XFIX_P_1_175  /* z5 = MULTIPLY(z3 + z4, FIX_1_175875602); */
+    smlal2          v5.4s, v11.8h, XFIX_P_1_175
+
+    smull2          v24.4s, v28.8h, XFIX_P_0_298
+    smull2          v25.4s, v29.8h, XFIX_P_2_053
+    smull2          v26.4s, v30.8h, XFIX_P_3_072
+    smull2          v27.4s, v31.8h, XFIX_P_1_501
+    smull           v28.4s, v28.4h, XFIX_P_0_298  /* tmp4 = MULTIPLY(tmp4, FIX_0_298631336); */
+    smull           v29.4s, v29.4h, XFIX_P_2_053  /* tmp5 = MULTIPLY(tmp5, FIX_2_053119869); */
+    smull           v30.4s, v30.4h, XFIX_P_3_072  /* tmp6 = MULTIPLY(tmp6, FIX_3_072711026); */
+    smull           v31.4s, v31.4h, XFIX_P_1_501  /* tmp7 = MULTIPLY(tmp7, FIX_1_501321110); */
+
+    smull2          v12.4s, v8.8h, XFIX_N_0_899
+    smull2          v13.4s, v9.8h, XFIX_N_2_562
+    smull2          v14.4s, v10.8h, XFIX_N_1_961
+    smull2          v15.4s, v11.8h, XFIX_N_0_390
+    smull           v8.4s, v8.4h, XFIX_N_0_899    /* z1 = MULTIPLY(z1, -FIX_0_899976223); */
+    smull           v9.4s, v9.4h, XFIX_N_2_562    /* z2 = MULTIPLY(z2, -FIX_2_562915447); */
+    smull           v10.4s, v10.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, -FIX_1_961570560); */
+    smull           v11.4s, v11.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, -FIX_0_390180644); */
+
+    add             v10.4s, v10.4s, v4.4s
+    add             v14.4s, v14.4s, v5.4s
+    add             v11.4s, v11.4s, v4.4s
+    add             v15.4s, v15.4s, v5.4s
+
+    add             v28.4s, v28.4s, v8.4s   /* tmp4 += z1 */
+    add             v24.4s, v24.4s, v12.4s
+    add             v29.4s, v29.4s, v9.4s   /* tmp5 += z2 */
+    add             v25.4s, v25.4s, v13.4s
+    add             v30.4s, v30.4s, v10.4s  /* tmp6 += z3 */
+    add             v26.4s, v26.4s, v14.4s
+    add             v31.4s, v31.4s, v11.4s  /* tmp7 += z4 */
+    add             v27.4s, v27.4s, v15.4s
+
+    add             v28.4s, v28.4s, v10.4s  /* tmp4 += z3 */
+    add             v24.4s, v24.4s, v14.4s
+    add             v29.4s, v29.4s, v11.4s  /* tmp5 += z4 */
+    add             v25.4s, v25.4s, v15.4s
+    add             v30.4s, v30.4s, v9.4s   /* tmp6 += z2 */
+    add             v26.4s, v26.4s, v13.4s
+    add             v31.4s, v31.4s, v8.4s   /* tmp7 += z1 */
+    add             v27.4s, v27.4s, v12.4s
+
+    rshrn           v23.4h, v28.4s, #DESCALE_P2
+    rshrn           v21.4h, v29.4s, #DESCALE_P2
+    rshrn           v19.4h, v30.4s, #DESCALE_P2
+    rshrn           v17.4h, v31.4s, #DESCALE_P2
+    rshrn2          v23.8h, v24.4s, #DESCALE_P2  /* dataptr7 = (DCTELEM)DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v21.8h, v25.4s, #DESCALE_P2  /* dataptr5 = (DCTELEM)DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS); */
+    rshrn2          v19.8h, v26.4s, #DESCALE_P2  /* dataptr3 = (DCTELEM)DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v17.8h, v27.4s, #DESCALE_P2  /* dataptr1 = (DCTELEM)DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS); */
+
+    /* store results */
+    st1             {v16.8h, v17.8h, v18.8h, v19.8h}, DATA, 64
+    st1             {v20.8h, v21.8h, v22.8h, v23.8h}, DATA
+
+    /* Restore Neon registers */
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, sp, 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, sp, 32
+
+    br              x30
+
+    .unreq          DATA
+    .unreq          TMP
+
+#undef XFIX_P_0_298
+#undef XFIX_N_0_390
+#undef XFIX_P_0_541
+#undef XFIX_P_0_765
+#undef XFIX_N_0_899
+#undef XFIX_P_1_175
+#undef XFIX_P_1_501
+#undef XFIX_N_1_847
+#undef XFIX_N_1_961
+#undef XFIX_P_2_053
+#undef XFIX_N_2_562
+#undef XFIX_P_3_072
+
+
+/*****************************************************************************/
+
+/*
+ * GLOBAL(JOCTET *)
+ * jsimd_huff_encode_one_block(working_state *state, JOCTET *buffer,
+ *                             JCOEFPTR block, int last_dc_val,
+ *                             c_derived_tbl *dctbl, c_derived_tbl *actbl)
+ *
+ */
+
+    BUFFER          .req x1
+    PUT_BUFFER      .req x6
+    PUT_BITS        .req x7
+    PUT_BITSw       .req w7
+
+.macro emit_byte
+    sub             PUT_BITS, PUT_BITS, #0x8
+    lsr             x19, PUT_BUFFER, PUT_BITS
+    uxtb            w19, w19
+    strb            w19, BUFFER, #1!
+    cmp             w19, #0xff
+    b.ne            14f
+    strb            wzr, BUFFER, #1!
+14:
+.endm
+.macro put_bits CODE, SIZE
+    lsl             PUT_BUFFER, PUT_BUFFER, \SIZE
+    add             PUT_BITS, PUT_BITS, \SIZE
+    orr             PUT_BUFFER, PUT_BUFFER, \CODE
+.endm
+.macro checkbuf31
+    cmp             PUT_BITS, #0x20
+    b.lt            31f
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+31:
+.endm
+.macro checkbuf47
+    cmp             PUT_BITS, #0x30
+    b.lt            47f
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+47:
+.endm
+
+.macro generate_jsimd_huff_encode_one_block fast_tbl
+
+.if \fast_tbl == 1
+asm_function jsimd_huff_encode_one_block_neon
+.else
+asm_function jsimd_huff_encode_one_block_neon_slowtbl
+.endif
+    sub             sp, sp, 272
+    sub             BUFFER, BUFFER, #0x1    /* BUFFER=buffer-- */
+    /* Save Arm registers */
+    stp             x19, x20, sp
+    get_symbol_loc  x15, Ljsimd_huff_encode_one_block_neon_consts
+    ldr             PUT_BUFFER, x0, #0x10
+    ldr             PUT_BITSw, x0, #0x18
+    ldrsh           w12, x2               /* load DC coeff in w12 */
+    /* prepare data */
+.if \fast_tbl == 1
+    ld1             {v23.16b}, x15, #16
+    ld1             {v0.16b, v1.16b, v2.16b, v3.16b}, x15, #64
+    ld1             {v4.16b, v5.16b, v6.16b, v7.16b}, x15, #64
+    ld1             {v16.16b, v17.16b, v18.16b, v19.16b}, x15, #64
+    ld1             {v24.16b, v25.16b, v26.16b, v27.16b}, x2, #64
+    ld1             {v28.16b, v29.16b, v30.16b, v31.16b}, x2, #64
+    sub             w12, w12, w3      /* last_dc_val, not used afterwards */
+    /* ZigZag 8x8 */
+    tbl             v0.16b, {v24.16b, v25.16b, v26.16b, v27.16b}, v0.16b
+    tbl             v1.16b, {v24.16b, v25.16b, v26.16b, v27.16b}, v1.16b
+    tbl             v2.16b, {v25.16b, v26.16b, v27.16b, v28.16b}, v2.16b
+    tbl             v3.16b, {v24.16b, v25.16b, v26.16b, v27.16b}, v3.16b
+    tbl             v4.16b, {v28.16b, v29.16b, v30.16b, v31.16b}, v4.16b
+    tbl             v5.16b, {v25.16b, v26.16b, v27.16b, v28.16b}, v5.16b
+    tbl             v6.16b, {v27.16b, v28.16b, v29.16b, v30.16b}, v6.16b
+    tbl             v7.16b, {v29.16b, v30.16b, v31.16b}, v7.16b
+    ins             v0.h0, w12
+    tbx             v1.16b, {v28.16b}, v16.16b
+    tbx             v2.16b, {v29.16b, v30.16b}, v17.16b
+    tbx             v5.16b, {v29.16b, v30.16b}, v18.16b
+    tbx             v6.16b, {v31.16b}, v19.16b
+.else
+      add             x13, x2, #0x22
+      sub             w12, w12, w3    /* last_dc_val, not used afterwards */
+    ld1             {v23.16b}, x15
+      add             x14, x2, #0x18
+      add             x3, x2, #0x36
+    ins             v0.h0, w12
+      add             x9, x2, #0x2
+    ld1             {v1.h}0, x13
+      add             x15, x2, #0x30
+    ld1             {v2.h}0, x14
+      add             x19, x2, #0x26
+    ld1             {v3.h}0, x3
+      add             x20, x2, #0x28
+    ld1             {v0.h}1, x9
+      add             x12, x2, #0x10
+    ld1             {v1.h}1, x15
+      add             x13, x2, #0x40
+    ld1             {v2.h}1, x19
+      add             x14, x2, #0x34
+    ld1             {v3.h}1, x20
+      add             x3, x2, #0x1a
+    ld1             {v0.h}2, x12
+      add             x9, x2, #0x20
+    ld1             {v1.h}2, x13
+      add             x15, x2, #0x32
+    ld1             {v2.h}2, x14
+      add             x19, x2, #0x42
+    ld1             {v3.h}2, x3
+      add             x20, x2, #0xc
+    ld1             {v0.h}3, x9
+      add             x12, x2, #0x12
+    ld1             {v1.h}3, x15
+      add             x13, x2, #0x24
+    ld1             {v2.h}3, x19
+      add             x14, x2, #0x50
+    ld1             {v3.h}3, x20
+      add             x3, x2, #0xe
+    ld1             {v0.h}4, x12
+      add             x9, x2, #0x4
+    ld1             {v1.h}4, x13
+      add             x15, x2, #0x16
+    ld1             {v2.h}4, x14
+      add             x19, x2, #0x60
+    ld1             {v3.h}4, x3
+      add             x20, x2, #0x1c
+    ld1             {v0.h}5, x9
+      add             x12, x2, #0x6
+    ld1             {v1.h}5, x15
+      add             x13, x2, #0x8
+    ld1             {v2.h}5, x19
+      add             x14, x2, #0x52
+    ld1             {v3.h}5, x20
+      add             x3, x2, #0x2a
+    ld1             {v0.h}6, x12
+      add             x9, x2, #0x14
+    ld1             {v1.h}6, x13
+      add             x15, x2, #0xa
+    ld1             {v2.h}6, x14
+      add             x19, x2, #0x44
+    ld1             {v3.h}6, x3
+      add             x20, x2, #0x38
+    ld1             {v0.h}7, x9
+      add             x12, x2, #0x46
+    ld1             {v1.h}7, x15
+      add             x13, x2, #0x3a
+    ld1             {v2.h}7, x19
+      add             x14, x2, #0x74
+    ld1             {v3.h}7, x20
+      add             x3, x2, #0x6a
+    ld1             {v4.h}0, x12
+      add             x9, x2, #0x54
+    ld1             {v5.h}0, x13
+      add             x15, x2, #0x2c
+    ld1             {v6.h}0, x14
+      add             x19, x2, #0x76
+    ld1             {v7.h}0, x3
+      add             x20, x2, #0x78
+    ld1             {v4.h}1, x9
+      add             x12, x2, #0x62
+    ld1             {v5.h}1, x15
+      add             x13, x2, #0x1e
+    ld1             {v6.h}1, x19
+      add             x14, x2, #0x68
+    ld1             {v7.h}1, x20
+      add             x3, x2, #0x7a
+    ld1             {v4.h}2, x12
+      add             x9, x2, #0x70
+    ld1             {v5.h}2, x13
+      add             x15, x2, #0x2e
+    ld1             {v6.h}2, x14
+      add             x19, x2, #0x5a
+    ld1             {v7.h}2, x3
+      add             x20, x2, #0x6c
+    ld1             {v4.h}3, x9
+      add             x12, x2, #0x72
+    ld1             {v5.h}3, x15
+      add             x13, x2, #0x3c
+    ld1             {v6.h}3, x19
+      add             x14, x2, #0x4c
+    ld1             {v7.h}3, x20
+      add             x3, x2, #0x5e
+    ld1             {v4.h}4, x12
+      add             x9, x2, #0x64
+    ld1             {v5.h}4, x13
+      add             x15, x2, #0x4a
+    ld1             {v6.h}4, x14
+      add             x19, x2, #0x3e
+    ld1             {v7.h}4, x3
+      add             x20, x2, #0x6e
+    ld1             {v4.h}5, x9
+      add             x12, x2, #0x56
+    ld1             {v5.h}5, x15
+      add             x13, x2, #0x58
+    ld1             {v6.h}5, x19
+      add             x14, x2, #0x4e
+    ld1             {v7.h}5, x20
+      add             x3, x2, #0x7c
+    ld1             {v4.h}6, x12
+      add             x9, x2, #0x48
+    ld1             {v5.h}6, x13
+      add             x15, x2, #0x66
+    ld1             {v6.h}6, x14
+      add             x19, x2, #0x5c
+    ld1             {v7.h}6, x3
+      add             x20, x2, #0x7e
+    ld1             {v4.h}7, x9
+    ld1             {v5.h}7, x15
+    ld1             {v6.h}7, x19
+    ld1             {v7.h}7, x20
+.endif
+    cmlt            v24.8h, v0.8h, #0
+    cmlt            v25.8h, v1.8h, #0
+    cmlt            v26.8h, v2.8h, #0
+    cmlt            v27.8h, v3.8h, #0
+    cmlt            v28.8h, v4.8h, #0
+    cmlt            v29.8h, v5.8h, #0
+    cmlt            v30.8h, v6.8h, #0
+    cmlt            v31.8h, v7.8h, #0
+    abs             v0.8h, v0.8h
+    abs             v1.8h, v1.8h
+    abs             v2.8h, v2.8h
+    abs             v3.8h, v3.8h
+    abs             v4.8h, v4.8h
+    abs             v5.8h, v5.8h
+    abs             v6.8h, v6.8h
+    abs             v7.8h, v7.8h
+    eor             v24.16b, v24.16b, v0.16b
+    eor             v25.16b, v25.16b, v1.16b
+    eor             v26.16b, v26.16b, v2.16b
+    eor             v27.16b, v27.16b, v3.16b
+    eor             v28.16b, v28.16b, v4.16b
+    eor             v29.16b, v29.16b, v5.16b
+    eor             v30.16b, v30.16b, v6.16b
+    eor             v31.16b, v31.16b, v7.16b
+    cmeq            v16.8h, v0.8h, #0
+    cmeq            v17.8h, v1.8h, #0
+    cmeq            v18.8h, v2.8h, #0
+    cmeq            v19.8h, v3.8h, #0
+    cmeq            v20.8h, v4.8h, #0
+    cmeq            v21.8h, v5.8h, #0
+    cmeq            v22.8h, v6.8h, #0
+    xtn             v16.8b, v16.8h
+    xtn             v18.8b, v18.8h
+    xtn             v20.8b, v20.8h
+    xtn             v22.8b, v22.8h
+      umov            w14, v0.h0
+    xtn2            v16.16b, v17.8h
+      umov            w13, v24.h0
+    xtn2            v18.16b, v19.8h
+      clz             w14, w14
+    xtn2            v20.16b, v21.8h
+      lsl             w13, w13, w14
+    cmeq            v17.8h, v7.8h, #0
+      sub             w12, w14, #32
+    xtn2            v22.16b, v17.8h
+      lsr             w13, w13, w14
+    and             v16.16b, v16.16b, v23.16b
+      neg             w12, w12
+    and             v18.16b, v18.16b, v23.16b
+      add             x3, x4, #0x400           /* r1 = dctbl->ehufsi */
+    and             v20.16b, v20.16b, v23.16b
+      add             x15, sp, #0x90           /* x15 = t2 */
+    and             v22.16b, v22.16b, v23.16b
+      ldr             w10, x4, x12, lsl #2
+    addp            v16.16b, v16.16b, v18.16b
+      ldrb            w11, x3, x12
+    addp            v20.16b, v20.16b, v22.16b
+      checkbuf47
+    addp            v16.16b, v16.16b, v20.16b
+      put_bits        x10, x11
+    addp            v16.16b, v16.16b, v18.16b
+      checkbuf47
+    umov            x9, v16.D0
+      put_bits        x13, x12
+    cnt             v17.8b, v16.8b
+      mvn             x9, x9
+    addv            B18, v17.8b
+      add             x4, x5, #0x400   /* x4 = actbl->ehufsi */
+    umov            w12, v18.b0
+      lsr             x9, x9, #0x1     /* clear AC coeff */
+    ldr             w13, x5, #0x3c0  /* x13 = actbl->ehufco0xf0 */
+    rbit            x9, x9             /* x9 = index0 */
+    ldrb            w14, x4, #0xf0   /* x14 = actbl->ehufsi0xf0 */
+    cmp             w12, #(64-8)
+    add             x11, sp, #16
+    b.lt            4f
+    cbz             x9, 6f
+    st1             {v0.8h, v1.8h, v2.8h, v3.8h}, x11, #64
+    st1             {v4.8h, v5.8h, v6.8h, v7.8h}, x11, #64
+    st1             {v24.8h, v25.8h, v26.8h, v27.8h}, x11, #64
+    st1             {v28.8h, v29.8h, v30.8h, v31.8h}, x11, #64
+1:
+    clz             x2, x9
+    add             x15, x15, x2, lsl #1
+    lsl             x9, x9, x2
+    ldrh            w20, x15, #-126
+2:
+    cmp             x2, #0x10
+    b.lt            3f
+    sub             x2, x2, #0x10
+    checkbuf47
+    put_bits        x13, x14
+    b               2b
+3:
+    clz             w20, w20
+    ldrh            w3, x15, #2!
+    sub             w11, w20, #32
+    lsl             w3, w3, w20
+    neg             w11, w11
+    lsr             w3, w3, w20
+    add             x2, x11, x2, lsl #4
+    lsl             x9, x9, #0x1
+    ldr             w12, x5, x2, lsl #2
+    ldrb            w10, x4, x2
+    checkbuf31
+    put_bits        x12, x10
+    put_bits        x3, x11
+    cbnz            x9, 1b
+    b               6f
+4:
+    movi            v21.8h, #0x0010
+    clz             v0.8h, v0.8h
+    clz             v1.8h, v1.8h
+    clz             v2.8h, v2.8h
+    clz             v3.8h, v3.8h
+    clz             v4.8h, v4.8h
+    clz             v5.8h, v5.8h
+    clz             v6.8h, v6.8h
+    clz             v7.8h, v7.8h
+    ushl            v24.8h, v24.8h, v0.8h
+    ushl            v25.8h, v25.8h, v1.8h
+    ushl            v26.8h, v26.8h, v2.8h
+    ushl            v27.8h, v27.8h, v3.8h
+    ushl            v28.8h, v28.8h, v4.8h
+    ushl            v29.8h, v29.8h, v5.8h
+    ushl            v30.8h, v30.8h, v6.8h
+    ushl            v31.8h, v31.8h, v7.8h
+    neg             v0.8h, v0.8h
+    neg             v1.8h, v1.8h
+    neg             v2.8h, v2.8h
+    neg             v3.8h, v3.8h
+    neg             v4.8h, v4.8h
+    neg             v5.8h, v5.8h
+    neg             v6.8h, v6.8h
+    neg             v7.8h, v7.8h
+    ushl            v24.8h, v24.8h, v0.8h
+    ushl            v25.8h, v25.8h, v1.8h
+    ushl            v26.8h, v26.8h, v2.8h
+    ushl            v27.8h, v27.8h, v3.8h
+    ushl            v28.8h, v28.8h, v4.8h
+    ushl            v29.8h, v29.8h, v5.8h
+    ushl            v30.8h, v30.8h, v6.8h
+    ushl            v31.8h, v31.8h, v7.8h
+    add             v0.8h, v21.8h, v0.8h
+    add             v1.8h, v21.8h, v1.8h
+    add             v2.8h, v21.8h, v2.8h
+    add             v3.8h, v21.8h, v3.8h
+    add             v4.8h, v21.8h, v4.8h
+    add             v5.8h, v21.8h, v5.8h
+    add             v6.8h, v21.8h, v6.8h
+    add             v7.8h, v21.8h, v7.8h
+    st1             {v0.8h, v1.8h, v2.8h, v3.8h}, x11, #64
+    st1             {v4.8h, v5.8h, v6.8h, v7.8h}, x11, #64
+    st1             {v24.8h, v25.8h, v26.8h, v27.8h}, x11, #64
+    st1             {v28.8h, v29.8h, v30.8h, v31.8h}, x11, #64
+1:
+    clz             x2, x9
+    add             x15, x15, x2, lsl #1
+    lsl             x9, x9, x2
+    ldrh            w11, x15, #-126
+2:
+    cmp             x2, #0x10
+    b.lt            3f
+    sub             x2, x2, #0x10
+    checkbuf47
+    put_bits        x13, x14
+    b               2b
+3:
+    ldrh            w3, x15, #2!
+    add             x2, x11, x2, lsl #4
+    lsl             x9, x9, #0x1
+    ldr             w12, x5, x2, lsl #2
+    ldrb            w10, x4, x2
+    checkbuf31
+    put_bits        x12, x10
+    put_bits        x3, x11
+    cbnz            x9, 1b
+6:
+    add             x13, sp, #0x10e
+    cmp             x15, x13
+    b.hs            1f
+    ldr             w12, x5
+    ldrb            w14, x4
+    checkbuf47
+    put_bits        x12, x14
+1:
+    str             PUT_BUFFER, x0, #0x10
+    str             PUT_BITSw, x0, #0x18
+    ldp             x19, x20, sp, 16
+    add             x0, BUFFER, #0x1
+    add             sp, sp, 256
+    br              x30
+
+.endm
+
+generate_jsimd_huff_encode_one_block 1
+generate_jsimd_huff_encode_one_block 0
+
+    .unreq          BUFFER
+    .unreq          PUT_BUFFER
+    .unreq          PUT_BITS
+    .unreq          PUT_BITSw
+
+.purgem emit_byte
+.purgem put_bits
+.purgem checkbuf31
+.purgem checkbuf47

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/align.h Added

@@ -0,0 +1,28 @@
+/*
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* How to obtain memory alignment for structures and variables */
+#if defined(_MSC_VER)
+#define ALIGN(alignment)  __declspec(align(alignment))
+#elif defined(__clang__) || defined(__GNUC__)
+#define ALIGN(alignment)  __attribute__((aligned(alignment)))
+#else
+#error "Unknown compiler"
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jccolor-neon.c Added

@@ -0,0 +1,160 @@
+/*
+ * jccolor-neon.c - colorspace conversion (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+#include "neon-compat.h"
+
+#include <arm_neon.h>
+
+
+/* RGB -> YCbCr conversion constants */
+
+#define F_0_298  19595
+#define F_0_587  38470
+#define F_0_113  7471
+#define F_0_168  11059
+#define F_0_331  21709
+#define F_0_500  32768
+#define F_0_418  27439
+#define F_0_081  5329
+
+ALIGN(16) static const uint16_t jsimd_rgb_ycc_neon_consts = {
+  F_0_298, F_0_587, F_0_113, F_0_168,
+  F_0_331, F_0_500, F_0_418, F_0_081
+};
+
+
+/* Include inline routines for colorspace extensions. */
+
+#if defined(__aarch64__) || defined(_M_ARM64)
+#include "aarch64/jccolext-neon.c"
+#else
+#include "aarch32/jccolext-neon.c"
+#endif
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_rgb_ycc_convert_neon  jsimd_extrgb_ycc_convert_neon
+#if defined(__aarch64__) || defined(_M_ARM64)
+#include "aarch64/jccolext-neon.c"
+#else
+#include "aarch32/jccolext-neon.c"
+#endif
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_neon
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define jsimd_rgb_ycc_convert_neon  jsimd_extrgbx_ycc_convert_neon
+#if defined(__aarch64__) || defined(_M_ARM64)
+#include "aarch64/jccolext-neon.c"
+#else
+#include "aarch32/jccolext-neon.c"
+#endif
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_neon
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define jsimd_rgb_ycc_convert_neon  jsimd_extbgr_ycc_convert_neon
+#if defined(__aarch64__) || defined(_M_ARM64)
+#include "aarch64/jccolext-neon.c"
+#else
+#include "aarch32/jccolext-neon.c"
+#endif
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_neon
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define jsimd_rgb_ycc_convert_neon  jsimd_extbgrx_ycc_convert_neon
+#if defined(__aarch64__) || defined(_M_ARM64)
+#include "aarch64/jccolext-neon.c"
+#else
+#include "aarch32/jccolext-neon.c"
+#endif
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_neon
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define jsimd_rgb_ycc_convert_neon  jsimd_extxbgr_ycc_convert_neon
+#if defined(__aarch64__) || defined(_M_ARM64)
+#include "aarch64/jccolext-neon.c"
+#else
+#include "aarch32/jccolext-neon.c"
+#endif
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_neon
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define jsimd_rgb_ycc_convert_neon  jsimd_extxrgb_ycc_convert_neon
+#if defined(__aarch64__) || defined(_M_ARM64)
+#include "aarch64/jccolext-neon.c"
+#else
+#include "aarch32/jccolext-neon.c"
+#endif
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_neon

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jcgray-neon.c Added

@@ -0,0 +1,120 @@
+/*
+ * jcgray-neon.c - grayscale colorspace conversion (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+
+#include <arm_neon.h>
+
+
+/* RGB -> Grayscale conversion constants */
+
+#define F_0_298  19595
+#define F_0_587  38470
+#define F_0_113  7471
+
+
+/* Include inline routines for colorspace extensions. */
+
+#include "jcgryext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_rgb_gray_convert_neon  jsimd_extrgb_gray_convert_neon
+#include "jcgryext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_neon
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define jsimd_rgb_gray_convert_neon  jsimd_extrgbx_gray_convert_neon
+#include "jcgryext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_neon
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define jsimd_rgb_gray_convert_neon  jsimd_extbgr_gray_convert_neon
+#include "jcgryext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_neon
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define jsimd_rgb_gray_convert_neon  jsimd_extbgrx_gray_convert_neon
+#include "jcgryext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_neon
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define jsimd_rgb_gray_convert_neon  jsimd_extxbgr_gray_convert_neon
+#include "jcgryext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_neon
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define jsimd_rgb_gray_convert_neon  jsimd_extxrgb_gray_convert_neon
+#include "jcgryext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_neon

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jcgryext-neon.c Added

@@ -0,0 +1,106 @@
+/*
+ * jcgryext-neon.c - grayscale colorspace conversion (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jcgray-neon.c */
+
+
+/* RGB -> Grayscale conversion is defined by the following equation:
+ *    Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+ *
+ * Avoid floating point arithmetic by using shifted integer constants:
+ *    0.29899597 = 19595 * 2^-16
+ *    0.58700561 = 38470 * 2^-16
+ *    0.11399841 =  7471 * 2^-16
+ * These constants are defined in jcgray-neon.c
+ *
+ * This is the same computation as the RGB -> Y portion of RGB -> YCbCr.
+ */
+
+void jsimd_rgb_gray_convert_neon(JDIMENSION image_width, JSAMPARRAY input_buf,
+                                 JSAMPIMAGE output_buf, JDIMENSION output_row,
+                                 int num_rows)
+{
+  JSAMPROW inptr;
+  JSAMPROW outptr;
+  /* Allocate temporary buffer for final (image_width % 16) pixels in row. */
+  ALIGN(16) uint8_t tmp_buf16 * RGB_PIXELSIZE;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf0output_row;
+    output_row++;
+
+    int cols_remaining = image_width;
+    for (; cols_remaining > 0; cols_remaining -= 16) {
+
+      /* To prevent buffer overread by the vector load instructions, the last
+       * (image_width % 16) columns of data are first memcopied to a temporary
+       * buffer large enough to accommodate the vector load.
+       */
+      if (cols_remaining < 16) {
+        memcpy(tmp_buf, inptr, cols_remaining * RGB_PIXELSIZE);
+        inptr = tmp_buf;
+      }
+
+#if RGB_PIXELSIZE == 4
+      uint8x16x4_t input_pixels = vld4q_u8(inptr);
+#else
+      uint8x16x3_t input_pixels = vld3q_u8(inptr);
+#endif
+      uint16x8_t r_l = vmovl_u8(vget_low_u8(input_pixels.valRGB_RED));
+      uint16x8_t r_h = vmovl_u8(vget_high_u8(input_pixels.valRGB_RED));
+      uint16x8_t g_l = vmovl_u8(vget_low_u8(input_pixels.valRGB_GREEN));
+      uint16x8_t g_h = vmovl_u8(vget_high_u8(input_pixels.valRGB_GREEN));
+      uint16x8_t b_l = vmovl_u8(vget_low_u8(input_pixels.valRGB_BLUE));
+      uint16x8_t b_h = vmovl_u8(vget_high_u8(input_pixels.valRGB_BLUE));
+
+      /* Compute Y = 0.29900 * R + 0.58700 * G + 0.11400 * B */
+      uint32x4_t y_ll = vmull_n_u16(vget_low_u16(r_l), F_0_298);
+      uint32x4_t y_lh = vmull_n_u16(vget_high_u16(r_l), F_0_298);
+      uint32x4_t y_hl = vmull_n_u16(vget_low_u16(r_h), F_0_298);
+      uint32x4_t y_hh = vmull_n_u16(vget_high_u16(r_h), F_0_298);
+      y_ll = vmlal_n_u16(y_ll, vget_low_u16(g_l), F_0_587);
+      y_lh = vmlal_n_u16(y_lh, vget_high_u16(g_l), F_0_587);
+      y_hl = vmlal_n_u16(y_hl, vget_low_u16(g_h), F_0_587);
+      y_hh = vmlal_n_u16(y_hh, vget_high_u16(g_h), F_0_587);
+      y_ll = vmlal_n_u16(y_ll, vget_low_u16(b_l), F_0_113);
+      y_lh = vmlal_n_u16(y_lh, vget_high_u16(b_l), F_0_113);
+      y_hl = vmlal_n_u16(y_hl, vget_low_u16(b_h), F_0_113);
+      y_hh = vmlal_n_u16(y_hh, vget_high_u16(b_h), F_0_113);
+
+      /* Descale Y values (rounding right shift) and narrow to 16-bit. */
+      uint16x8_t y_l = vcombine_u16(vrshrn_n_u32(y_ll, 16),
+                                    vrshrn_n_u32(y_lh, 16));
+      uint16x8_t y_h = vcombine_u16(vrshrn_n_u32(y_hl, 16),
+                                    vrshrn_n_u32(y_hh, 16));
+
+      /* Narrow Y values to 8-bit and store to memory.  Buffer overwrite is
+       * permitted up to the next multiple of ALIGN_SIZE bytes.
+       */
+      vst1q_u8(outptr, vcombine_u8(vmovn_u16(y_l), vmovn_u16(y_h)));
+
+      /* Increment pointers. */
+      inptr += (16 * RGB_PIXELSIZE);
+      outptr += 16;
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jchuff.h Added

@@ -0,0 +1,131 @@
+/*
+ * jchuff.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2018, 2021, D. R. Commander.
+ * Copyright (C) 2018, Matthias Räncker.
+ * Copyright (C) 2020-2021, Arm Limited.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+/* Expanded entropy encoder object for Huffman encoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+#if defined(__aarch64__) || defined(_M_ARM64)
+#define BIT_BUF_SIZE  64
+#else
+#define BIT_BUF_SIZE  32
+#endif
+
+typedef struct {
+  size_t put_buffer;                    /* current bit accumulation buffer */
+  int free_bits;                        /* # of bits available in it */
+  int last_dc_valMAX_COMPS_IN_SCAN;   /* last DC coef for each component */
+} savable_state;
+
+typedef struct {
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
+  savable_state cur;            /* Current bit buffer & DC state */
+  j_compress_ptr cinfo;         /* dump_buffer needs access to this */
+  int simd;
+} working_state;
+
+/* Outputting bits to the file */
+
+/* Output byte b and, speculatively, an additional 0 byte. 0xFF must be encoded
+ * as 0xFF 0x00, so the output buffer pointer is advanced by 2 if the byte is
+ * 0xFF.  Otherwise, the output buffer pointer is advanced by 1, and the
+ * speculative 0 byte will be overwritten by the next byte.
+ */
+#define EMIT_BYTE(b) { \
+  buffer0 = (JOCTET)(b); \
+  buffer1 = 0; \
+  buffer -= -2 + ((JOCTET)(b) < 0xFF); \
+}
+
+/* Output the entire bit buffer.  If there are no 0xFF bytes in it, then write
+ * directly to the output buffer.  Otherwise, use the EMIT_BYTE() macro to
+ * encode 0xFF as 0xFF 0x00.
+ */
+#if defined(__aarch64__) || defined(_M_ARM64)
+
+#define FLUSH() { \
+  if (put_buffer & 0x8080808080808080 & ~(put_buffer + 0x0101010101010101)) { \
+    EMIT_BYTE(put_buffer >> 56) \
+    EMIT_BYTE(put_buffer >> 48) \
+    EMIT_BYTE(put_buffer >> 40) \
+    EMIT_BYTE(put_buffer >> 32) \
+    EMIT_BYTE(put_buffer >> 24) \
+    EMIT_BYTE(put_buffer >> 16) \
+    EMIT_BYTE(put_buffer >>  8) \
+    EMIT_BYTE(put_buffer      ) \
+  } else { \
+    *((uint64_t *)buffer) = BUILTIN_BSWAP64(put_buffer); \
+    buffer += 8; \
+  } \
+}
+
+#else
+
+#if defined(_MSC_VER) && !defined(__clang__)
+#define SPLAT() { \
+  buffer0 = (JOCTET)(put_buffer >> 24); \
+  buffer1 = (JOCTET)(put_buffer >> 16); \
+  buffer2 = (JOCTET)(put_buffer >>  8); \
+  buffer3 = (JOCTET)(put_buffer      ); \
+  buffer += 4; \
+}
+#else
+#define SPLAT() { \
+  put_buffer = __builtin_bswap32(put_buffer); \
+  __asm__("str %1, %0, #4" : "+r" (buffer) : "r" (put_buffer)); \
+}
+#endif
+
+#define FLUSH() { \
+  if (put_buffer & 0x80808080 & ~(put_buffer + 0x01010101)) { \
+    EMIT_BYTE(put_buffer >> 24) \
+    EMIT_BYTE(put_buffer >> 16) \
+    EMIT_BYTE(put_buffer >>  8) \
+    EMIT_BYTE(put_buffer      ) \
+  } else { \
+    SPLAT(); \
+  } \
+}
+
+#endif
+
+/* Fill the bit buffer to capacity with the leading bits from code, then output
+ * the bit buffer and put the remaining bits from code into the bit buffer.
+ */
+#define PUT_AND_FLUSH(code, size) { \
+  put_buffer = (put_buffer << (size + free_bits)) | (code >> -free_bits); \
+  FLUSH() \
+  free_bits += BIT_BUF_SIZE; \
+  put_buffer = code; \
+}
+
+/* Insert code into the bit buffer and output the bit buffer if needed.
+ * NOTE: We can't flush with free_bits == 0, since the left shift in
+ * PUT_AND_FLUSH() would have undefined behavior.
+ */
+#define PUT_BITS(code, size) { \
+  free_bits -= size; \
+  if (free_bits < 0) \
+    PUT_AND_FLUSH(code, size) \
+  else \
+    put_buffer = (put_buffer << size) | code; \
+}
+
+#define PUT_CODE(code, size, diff) { \
+  diff |= code << nbits; \
+  nbits += size; \
+  PUT_BITS(diff, nbits) \
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jcphuff-neon.c Added

@@ -0,0 +1,622 @@
+/*
+ * jcphuff-neon.c - prepare data for progressive Huffman encoding (Arm Neon)
+ *
+ * Copyright (C) 2020-2021, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "jconfigint.h"
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "neon-compat.h"
+
+#include <arm_neon.h>
+
+
+/* Data preparation for encode_mcu_AC_first().
+ *
+ * The equivalent scalar C function (encode_mcu_AC_first_prepare()) can be
+ * found in jcphuff.c.
+ */
+
+void jsimd_encode_mcu_AC_first_prepare_neon
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *values, size_t *zerobits)
+{
+  JCOEF *values_ptr = values;
+  JCOEF *diff_values_ptr = values + DCTSIZE2;
+
+  /* Rows of coefficients to zero (since they haven't been processed) */
+  int i, rows_to_zero = 8;
+
+  for (i = 0; i < Sl / 16; i++) {
+    int16x8_t coefs1 = vld1q_dup_s16(block + jpeg_natural_order_start0);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start1, coefs1, 1);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start2, coefs1, 2);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start3, coefs1, 3);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start4, coefs1, 4);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start5, coefs1, 5);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start6, coefs1, 6);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start7, coefs1, 7);
+    int16x8_t coefs2 = vld1q_dup_s16(block + jpeg_natural_order_start8);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start9, coefs2, 1);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start10, coefs2, 2);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start11, coefs2, 3);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start12, coefs2, 4);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start13, coefs2, 5);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start14, coefs2, 6);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start15, coefs2, 7);
+
+    /* Isolate sign of coefficients. */
+    int16x8_t sign_coefs1 = vshrq_n_s16(coefs1, 15);
+    int16x8_t sign_coefs2 = vshrq_n_s16(coefs2, 15);
+    /* Compute absolute value of coefficients and apply point transform Al. */
+    int16x8_t abs_coefs1 = vabsq_s16(coefs1);
+    int16x8_t abs_coefs2 = vabsq_s16(coefs2);
+    coefs1 = vshlq_s16(abs_coefs1, vdupq_n_s16(-Al));
+    coefs2 = vshlq_s16(abs_coefs2, vdupq_n_s16(-Al));
+
+    /* Compute diff values. */
+    int16x8_t diff1 = veorq_s16(coefs1, sign_coefs1);
+    int16x8_t diff2 = veorq_s16(coefs2, sign_coefs2);
+
+    /* Store transformed coefficients and diff values. */
+    vst1q_s16(values_ptr, coefs1);
+    vst1q_s16(values_ptr + DCTSIZE, coefs2);
+    vst1q_s16(diff_values_ptr, diff1);
+    vst1q_s16(diff_values_ptr + DCTSIZE, diff2);
+    values_ptr += 16;
+    diff_values_ptr += 16;
+    jpeg_natural_order_start += 16;
+    rows_to_zero -= 2;
+  }
+
+  /* Same operation but for remaining partial vector */
+  int remaining_coefs = Sl % 16;
+  if (remaining_coefs > 8) {
+    int16x8_t coefs1 = vld1q_dup_s16(block + jpeg_natural_order_start0);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start1, coefs1, 1);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start2, coefs1, 2);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start3, coefs1, 3);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start4, coefs1, 4);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start5, coefs1, 5);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start6, coefs1, 6);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start7, coefs1, 7);
+    int16x8_t coefs2 = vdupq_n_s16(0);
+    switch (remaining_coefs) {
+    case 15:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start14, coefs2, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 14:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start13, coefs2, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 13:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start12, coefs2, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 12:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start11, coefs2, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 11:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start10, coefs2, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 10:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start9, coefs2, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 9:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start8, coefs2, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    default:
+      break;
+    }
+
+    /* Isolate sign of coefficients. */
+    int16x8_t sign_coefs1 = vshrq_n_s16(coefs1, 15);
+    int16x8_t sign_coefs2 = vshrq_n_s16(coefs2, 15);
+    /* Compute absolute value of coefficients and apply point transform Al. */
+    int16x8_t abs_coefs1 = vabsq_s16(coefs1);
+    int16x8_t abs_coefs2 = vabsq_s16(coefs2);
+    coefs1 = vshlq_s16(abs_coefs1, vdupq_n_s16(-Al));
+    coefs2 = vshlq_s16(abs_coefs2, vdupq_n_s16(-Al));
+
+    /* Compute diff values. */
+    int16x8_t diff1 = veorq_s16(coefs1, sign_coefs1);
+    int16x8_t diff2 = veorq_s16(coefs2, sign_coefs2);
+
+    /* Store transformed coefficients and diff values. */
+    vst1q_s16(values_ptr, coefs1);
+    vst1q_s16(values_ptr + DCTSIZE, coefs2);
+    vst1q_s16(diff_values_ptr, diff1);
+    vst1q_s16(diff_values_ptr + DCTSIZE, diff2);
+    values_ptr += 16;
+    diff_values_ptr += 16;
+    rows_to_zero -= 2;
+
+  } else if (remaining_coefs > 0) {
+    int16x8_t coefs = vdupq_n_s16(0);
+
+    switch (remaining_coefs) {
+    case 8:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start7, coefs, 7);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 7:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start6, coefs, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 6:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start5, coefs, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 5:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start4, coefs, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 4:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start3, coefs, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 3:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start2, coefs, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 2:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start1, coefs, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 1:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start0, coefs, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    default:
+      break;
+    }
+
+    /* Isolate sign of coefficients. */
+    int16x8_t sign_coefs = vshrq_n_s16(coefs, 15);
+    /* Compute absolute value of coefficients and apply point transform Al. */
+    int16x8_t abs_coefs = vabsq_s16(coefs);
+    coefs = vshlq_s16(abs_coefs, vdupq_n_s16(-Al));
+
+    /* Compute diff values. */
+    int16x8_t diff = veorq_s16(coefs, sign_coefs);
+
+    /* Store transformed coefficients and diff values. */
+    vst1q_s16(values_ptr, coefs);
+    vst1q_s16(diff_values_ptr, diff);
+    values_ptr += 8;
+    diff_values_ptr += 8;
+    rows_to_zero--;
+  }
+
+  /* Zero remaining memory in the values and diff_values blocks. */
+  for (i = 0; i < rows_to_zero; i++) {
+    vst1q_s16(values_ptr, vdupq_n_s16(0));
+    vst1q_s16(diff_values_ptr, vdupq_n_s16(0));
+    values_ptr += 8;
+    diff_values_ptr += 8;
+  }
+
+  /* Construct zerobits bitmap.  A set bit means that the corresponding
+   * coefficient != 0.
+   */
+  int16x8_t row0 = vld1q_s16(values + 0 * DCTSIZE);
+  int16x8_t row1 = vld1q_s16(values + 1 * DCTSIZE);
+  int16x8_t row2 = vld1q_s16(values + 2 * DCTSIZE);
+  int16x8_t row3 = vld1q_s16(values + 3 * DCTSIZE);
+  int16x8_t row4 = vld1q_s16(values + 4 * DCTSIZE);
+  int16x8_t row5 = vld1q_s16(values + 5 * DCTSIZE);
+  int16x8_t row6 = vld1q_s16(values + 6 * DCTSIZE);
+  int16x8_t row7 = vld1q_s16(values + 7 * DCTSIZE);
+
+  uint8x8_t row0_eq0 = vmovn_u16(vceqq_s16(row0, vdupq_n_s16(0)));
+  uint8x8_t row1_eq0 = vmovn_u16(vceqq_s16(row1, vdupq_n_s16(0)));
+  uint8x8_t row2_eq0 = vmovn_u16(vceqq_s16(row2, vdupq_n_s16(0)));
+  uint8x8_t row3_eq0 = vmovn_u16(vceqq_s16(row3, vdupq_n_s16(0)));
+  uint8x8_t row4_eq0 = vmovn_u16(vceqq_s16(row4, vdupq_n_s16(0)));
+  uint8x8_t row5_eq0 = vmovn_u16(vceqq_s16(row5, vdupq_n_s16(0)));
+  uint8x8_t row6_eq0 = vmovn_u16(vceqq_s16(row6, vdupq_n_s16(0)));
+  uint8x8_t row7_eq0 = vmovn_u16(vceqq_s16(row7, vdupq_n_s16(0)));
+
+  /* { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 } */
+  const uint8x8_t bitmap_mask =
+    vreinterpret_u8_u64(vmov_n_u64(0x8040201008040201));
+
+  row0_eq0 = vand_u8(row0_eq0, bitmap_mask);
+  row1_eq0 = vand_u8(row1_eq0, bitmap_mask);
+  row2_eq0 = vand_u8(row2_eq0, bitmap_mask);
+  row3_eq0 = vand_u8(row3_eq0, bitmap_mask);
+  row4_eq0 = vand_u8(row4_eq0, bitmap_mask);
+  row5_eq0 = vand_u8(row5_eq0, bitmap_mask);
+  row6_eq0 = vand_u8(row6_eq0, bitmap_mask);
+  row7_eq0 = vand_u8(row7_eq0, bitmap_mask);
+
+  uint8x8_t bitmap_rows_01 = vpadd_u8(row0_eq0, row1_eq0);
+  uint8x8_t bitmap_rows_23 = vpadd_u8(row2_eq0, row3_eq0);
+  uint8x8_t bitmap_rows_45 = vpadd_u8(row4_eq0, row5_eq0);
+  uint8x8_t bitmap_rows_67 = vpadd_u8(row6_eq0, row7_eq0);
+  uint8x8_t bitmap_rows_0123 = vpadd_u8(bitmap_rows_01, bitmap_rows_23);
+  uint8x8_t bitmap_rows_4567 = vpadd_u8(bitmap_rows_45, bitmap_rows_67);
+  uint8x8_t bitmap_all = vpadd_u8(bitmap_rows_0123, bitmap_rows_4567);
+
+#if defined(__aarch64__) || defined(_M_ARM64)
+  /* Move bitmap to a 64-bit scalar register. */
+  uint64_t bitmap = vget_lane_u64(vreinterpret_u64_u8(bitmap_all), 0);
+  /* Store zerobits bitmap. */
+  *zerobits = ~bitmap;
+#else
+  /* Move bitmap to two 32-bit scalar registers. */
+  uint32_t bitmap0 = vget_lane_u32(vreinterpret_u32_u8(bitmap_all), 0);
+  uint32_t bitmap1 = vget_lane_u32(vreinterpret_u32_u8(bitmap_all), 1);
+  /* Store zerobits bitmap. */
+  zerobits0 = ~bitmap0;
+  zerobits1 = ~bitmap1;
+#endif
+}
+
+
+/* Data preparation for encode_mcu_AC_refine().
+ *
+ * The equivalent scalar C function (encode_mcu_AC_refine_prepare()) can be
+ * found in jcphuff.c.
+ */
+
+int jsimd_encode_mcu_AC_refine_prepare_neon
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *absvalues, size_t *bits)
+{
+  /* Temporary storage buffers for data used to compute the signbits bitmap and
+   * the end-of-block (EOB) position
+   */
+  uint8_t coef_sign_bits64;
+  uint8_t coef_eq1_bits64;
+
+  JCOEF *absvalues_ptr = absvalues;
+  uint8_t *coef_sign_bits_ptr = coef_sign_bits;
+  uint8_t *eq1_bits_ptr = coef_eq1_bits;
+
+  /* Rows of coefficients to zero (since they haven't been processed) */
+  int i, rows_to_zero = 8;
+
+  for (i = 0; i < Sl / 16; i++) {
+    int16x8_t coefs1 = vld1q_dup_s16(block + jpeg_natural_order_start0);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start1, coefs1, 1);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start2, coefs1, 2);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start3, coefs1, 3);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start4, coefs1, 4);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start5, coefs1, 5);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start6, coefs1, 6);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start7, coefs1, 7);
+    int16x8_t coefs2 = vld1q_dup_s16(block + jpeg_natural_order_start8);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start9, coefs2, 1);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start10, coefs2, 2);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start11, coefs2, 3);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start12, coefs2, 4);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start13, coefs2, 5);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start14, coefs2, 6);
+    coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start15, coefs2, 7);
+
+    /* Compute and store data for signbits bitmap. */
+    uint8x8_t sign_coefs1 =
+      vmovn_u16(vreinterpretq_u16_s16(vshrq_n_s16(coefs1, 15)));
+    uint8x8_t sign_coefs2 =
+      vmovn_u16(vreinterpretq_u16_s16(vshrq_n_s16(coefs2, 15)));
+    vst1_u8(coef_sign_bits_ptr, sign_coefs1);
+    vst1_u8(coef_sign_bits_ptr + DCTSIZE, sign_coefs2);
+
+    /* Compute absolute value of coefficients and apply point transform Al. */
+    int16x8_t abs_coefs1 = vabsq_s16(coefs1);
+    int16x8_t abs_coefs2 = vabsq_s16(coefs2);
+    coefs1 = vshlq_s16(abs_coefs1, vdupq_n_s16(-Al));
+    coefs2 = vshlq_s16(abs_coefs2, vdupq_n_s16(-Al));
+    vst1q_s16(absvalues_ptr, coefs1);
+    vst1q_s16(absvalues_ptr + DCTSIZE, coefs2);
+
+    /* Test whether transformed coefficient values == 1 (used to find EOB
+     * position.)
+     */
+    uint8x8_t coefs_eq11 = vmovn_u16(vceqq_s16(coefs1, vdupq_n_s16(1)));
+    uint8x8_t coefs_eq12 = vmovn_u16(vceqq_s16(coefs2, vdupq_n_s16(1)));
+    vst1_u8(eq1_bits_ptr, coefs_eq11);
+    vst1_u8(eq1_bits_ptr + DCTSIZE, coefs_eq12);
+
+    absvalues_ptr += 16;
+    coef_sign_bits_ptr += 16;
+    eq1_bits_ptr += 16;
+    jpeg_natural_order_start += 16;
+    rows_to_zero -= 2;
+  }
+
+  /* Same operation but for remaining partial vector */
+  int remaining_coefs = Sl % 16;
+  if (remaining_coefs > 8) {
+    int16x8_t coefs1 = vld1q_dup_s16(block + jpeg_natural_order_start0);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start1, coefs1, 1);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start2, coefs1, 2);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start3, coefs1, 3);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start4, coefs1, 4);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start5, coefs1, 5);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start6, coefs1, 6);
+    coefs1 = vld1q_lane_s16(block + jpeg_natural_order_start7, coefs1, 7);
+    int16x8_t coefs2 = vdupq_n_s16(0);
+    switch (remaining_coefs) {
+    case 15:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start14, coefs2, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 14:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start13, coefs2, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 13:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start12, coefs2, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 12:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start11, coefs2, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 11:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start10, coefs2, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 10:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start9, coefs2, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 9:
+      coefs2 = vld1q_lane_s16(block + jpeg_natural_order_start8, coefs2, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    default:
+      break;
+    }
+
+    /* Compute and store data for signbits bitmap. */
+    uint8x8_t sign_coefs1 =
+      vmovn_u16(vreinterpretq_u16_s16(vshrq_n_s16(coefs1, 15)));
+    uint8x8_t sign_coefs2 =
+      vmovn_u16(vreinterpretq_u16_s16(vshrq_n_s16(coefs2, 15)));
+    vst1_u8(coef_sign_bits_ptr, sign_coefs1);
+    vst1_u8(coef_sign_bits_ptr + DCTSIZE, sign_coefs2);
+
+    /* Compute absolute value of coefficients and apply point transform Al. */
+    int16x8_t abs_coefs1 = vabsq_s16(coefs1);
+    int16x8_t abs_coefs2 = vabsq_s16(coefs2);
+    coefs1 = vshlq_s16(abs_coefs1, vdupq_n_s16(-Al));
+    coefs2 = vshlq_s16(abs_coefs2, vdupq_n_s16(-Al));
+    vst1q_s16(absvalues_ptr, coefs1);
+    vst1q_s16(absvalues_ptr + DCTSIZE, coefs2);
+
+    /* Test whether transformed coefficient values == 1 (used to find EOB
+     * position.)
+     */
+    uint8x8_t coefs_eq11 = vmovn_u16(vceqq_s16(coefs1, vdupq_n_s16(1)));
+    uint8x8_t coefs_eq12 = vmovn_u16(vceqq_s16(coefs2, vdupq_n_s16(1)));
+    vst1_u8(eq1_bits_ptr, coefs_eq11);
+    vst1_u8(eq1_bits_ptr + DCTSIZE, coefs_eq12);
+
+    absvalues_ptr += 16;
+    coef_sign_bits_ptr += 16;
+    eq1_bits_ptr += 16;
+    jpeg_natural_order_start += 16;
+    rows_to_zero -= 2;
+
+  } else if (remaining_coefs > 0) {
+    int16x8_t coefs = vdupq_n_s16(0);
+
+    switch (remaining_coefs) {
+    case 8:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start7, coefs, 7);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 7:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start6, coefs, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 6:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start5, coefs, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 5:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start4, coefs, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 4:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start3, coefs, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 3:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start2, coefs, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 2:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start1, coefs, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 1:
+      coefs = vld1q_lane_s16(block + jpeg_natural_order_start0, coefs, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    default:
+      break;
+    }
+
+    /* Compute and store data for signbits bitmap. */
+    uint8x8_t sign_coefs =
+      vmovn_u16(vreinterpretq_u16_s16(vshrq_n_s16(coefs, 15)));
+    vst1_u8(coef_sign_bits_ptr, sign_coefs);
+
+    /* Compute absolute value of coefficients and apply point transform Al. */
+    int16x8_t abs_coefs = vabsq_s16(coefs);
+    coefs = vshlq_s16(abs_coefs, vdupq_n_s16(-Al));
+    vst1q_s16(absvalues_ptr, coefs);
+
+    /* Test whether transformed coefficient values == 1 (used to find EOB
+     * position.)
+     */
+    uint8x8_t coefs_eq1 = vmovn_u16(vceqq_s16(coefs, vdupq_n_s16(1)));
+    vst1_u8(eq1_bits_ptr, coefs_eq1);
+
+    absvalues_ptr += 8;
+    coef_sign_bits_ptr += 8;
+    eq1_bits_ptr += 8;
+    rows_to_zero--;
+  }
+
+  /* Zero remaining memory in blocks. */
+  for (i = 0; i < rows_to_zero; i++) {
+    vst1q_s16(absvalues_ptr, vdupq_n_s16(0));
+    vst1_u8(coef_sign_bits_ptr, vdup_n_u8(0));
+    vst1_u8(eq1_bits_ptr, vdup_n_u8(0));
+    absvalues_ptr += 8;
+    coef_sign_bits_ptr += 8;
+    eq1_bits_ptr += 8;
+  }
+
+  /* Construct zerobits bitmap. */
+  int16x8_t abs_row0 = vld1q_s16(absvalues + 0 * DCTSIZE);
+  int16x8_t abs_row1 = vld1q_s16(absvalues + 1 * DCTSIZE);
+  int16x8_t abs_row2 = vld1q_s16(absvalues + 2 * DCTSIZE);
+  int16x8_t abs_row3 = vld1q_s16(absvalues + 3 * DCTSIZE);
+  int16x8_t abs_row4 = vld1q_s16(absvalues + 4 * DCTSIZE);
+  int16x8_t abs_row5 = vld1q_s16(absvalues + 5 * DCTSIZE);
+  int16x8_t abs_row6 = vld1q_s16(absvalues + 6 * DCTSIZE);
+  int16x8_t abs_row7 = vld1q_s16(absvalues + 7 * DCTSIZE);
+
+  uint8x8_t abs_row0_eq0 = vmovn_u16(vceqq_s16(abs_row0, vdupq_n_s16(0)));
+  uint8x8_t abs_row1_eq0 = vmovn_u16(vceqq_s16(abs_row1, vdupq_n_s16(0)));
+  uint8x8_t abs_row2_eq0 = vmovn_u16(vceqq_s16(abs_row2, vdupq_n_s16(0)));
+  uint8x8_t abs_row3_eq0 = vmovn_u16(vceqq_s16(abs_row3, vdupq_n_s16(0)));
+  uint8x8_t abs_row4_eq0 = vmovn_u16(vceqq_s16(abs_row4, vdupq_n_s16(0)));
+  uint8x8_t abs_row5_eq0 = vmovn_u16(vceqq_s16(abs_row5, vdupq_n_s16(0)));
+  uint8x8_t abs_row6_eq0 = vmovn_u16(vceqq_s16(abs_row6, vdupq_n_s16(0)));
+  uint8x8_t abs_row7_eq0 = vmovn_u16(vceqq_s16(abs_row7, vdupq_n_s16(0)));
+
+  /* { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 } */
+  const uint8x8_t bitmap_mask =
+    vreinterpret_u8_u64(vmov_n_u64(0x8040201008040201));
+
+  abs_row0_eq0 = vand_u8(abs_row0_eq0, bitmap_mask);
+  abs_row1_eq0 = vand_u8(abs_row1_eq0, bitmap_mask);
+  abs_row2_eq0 = vand_u8(abs_row2_eq0, bitmap_mask);
+  abs_row3_eq0 = vand_u8(abs_row3_eq0, bitmap_mask);
+  abs_row4_eq0 = vand_u8(abs_row4_eq0, bitmap_mask);
+  abs_row5_eq0 = vand_u8(abs_row5_eq0, bitmap_mask);
+  abs_row6_eq0 = vand_u8(abs_row6_eq0, bitmap_mask);
+  abs_row7_eq0 = vand_u8(abs_row7_eq0, bitmap_mask);
+
+  uint8x8_t bitmap_rows_01 = vpadd_u8(abs_row0_eq0, abs_row1_eq0);
+  uint8x8_t bitmap_rows_23 = vpadd_u8(abs_row2_eq0, abs_row3_eq0);
+  uint8x8_t bitmap_rows_45 = vpadd_u8(abs_row4_eq0, abs_row5_eq0);
+  uint8x8_t bitmap_rows_67 = vpadd_u8(abs_row6_eq0, abs_row7_eq0);
+  uint8x8_t bitmap_rows_0123 = vpadd_u8(bitmap_rows_01, bitmap_rows_23);
+  uint8x8_t bitmap_rows_4567 = vpadd_u8(bitmap_rows_45, bitmap_rows_67);
+  uint8x8_t bitmap_all = vpadd_u8(bitmap_rows_0123, bitmap_rows_4567);
+
+#if defined(__aarch64__) || defined(_M_ARM64)
+  /* Move bitmap to a 64-bit scalar register. */
+  uint64_t bitmap = vget_lane_u64(vreinterpret_u64_u8(bitmap_all), 0);
+  /* Store zerobits bitmap. */
+  bits0 = ~bitmap;
+#else
+  /* Move bitmap to two 32-bit scalar registers. */
+  uint32_t bitmap0 = vget_lane_u32(vreinterpret_u32_u8(bitmap_all), 0);
+  uint32_t bitmap1 = vget_lane_u32(vreinterpret_u32_u8(bitmap_all), 1);
+  /* Store zerobits bitmap. */
+  bits0 = ~bitmap0;
+  bits1 = ~bitmap1;
+#endif
+
+  /* Construct signbits bitmap. */
+  uint8x8_t signbits_row0 = vld1_u8(coef_sign_bits + 0 * DCTSIZE);
+  uint8x8_t signbits_row1 = vld1_u8(coef_sign_bits + 1 * DCTSIZE);
+  uint8x8_t signbits_row2 = vld1_u8(coef_sign_bits + 2 * DCTSIZE);
+  uint8x8_t signbits_row3 = vld1_u8(coef_sign_bits + 3 * DCTSIZE);
+  uint8x8_t signbits_row4 = vld1_u8(coef_sign_bits + 4 * DCTSIZE);
+  uint8x8_t signbits_row5 = vld1_u8(coef_sign_bits + 5 * DCTSIZE);
+  uint8x8_t signbits_row6 = vld1_u8(coef_sign_bits + 6 * DCTSIZE);
+  uint8x8_t signbits_row7 = vld1_u8(coef_sign_bits + 7 * DCTSIZE);
+
+  signbits_row0 = vand_u8(signbits_row0, bitmap_mask);
+  signbits_row1 = vand_u8(signbits_row1, bitmap_mask);
+  signbits_row2 = vand_u8(signbits_row2, bitmap_mask);
+  signbits_row3 = vand_u8(signbits_row3, bitmap_mask);
+  signbits_row4 = vand_u8(signbits_row4, bitmap_mask);
+  signbits_row5 = vand_u8(signbits_row5, bitmap_mask);
+  signbits_row6 = vand_u8(signbits_row6, bitmap_mask);
+  signbits_row7 = vand_u8(signbits_row7, bitmap_mask);
+
+  bitmap_rows_01 = vpadd_u8(signbits_row0, signbits_row1);
+  bitmap_rows_23 = vpadd_u8(signbits_row2, signbits_row3);
+  bitmap_rows_45 = vpadd_u8(signbits_row4, signbits_row5);
+  bitmap_rows_67 = vpadd_u8(signbits_row6, signbits_row7);
+  bitmap_rows_0123 = vpadd_u8(bitmap_rows_01, bitmap_rows_23);
+  bitmap_rows_4567 = vpadd_u8(bitmap_rows_45, bitmap_rows_67);
+  bitmap_all = vpadd_u8(bitmap_rows_0123, bitmap_rows_4567);
+
+#if defined(__aarch64__) || defined(_M_ARM64)
+  /* Move bitmap to a 64-bit scalar register. */
+  bitmap = vget_lane_u64(vreinterpret_u64_u8(bitmap_all), 0);
+  /* Store signbits bitmap. */
+  bits1 = ~bitmap;
+#else
+  /* Move bitmap to two 32-bit scalar registers. */
+  bitmap0 = vget_lane_u32(vreinterpret_u32_u8(bitmap_all), 0);
+  bitmap1 = vget_lane_u32(vreinterpret_u32_u8(bitmap_all), 1);
+  /* Store signbits bitmap. */
+  bits2 = ~bitmap0;
+  bits3 = ~bitmap1;
+#endif
+
+  /* Construct bitmap to find EOB position (the index of the last coefficient
+   * equal to 1.)
+   */
+  uint8x8_t row0_eq1 = vld1_u8(coef_eq1_bits + 0 * DCTSIZE);
+  uint8x8_t row1_eq1 = vld1_u8(coef_eq1_bits + 1 * DCTSIZE);
+  uint8x8_t row2_eq1 = vld1_u8(coef_eq1_bits + 2 * DCTSIZE);
+  uint8x8_t row3_eq1 = vld1_u8(coef_eq1_bits + 3 * DCTSIZE);
+  uint8x8_t row4_eq1 = vld1_u8(coef_eq1_bits + 4 * DCTSIZE);
+  uint8x8_t row5_eq1 = vld1_u8(coef_eq1_bits + 5 * DCTSIZE);
+  uint8x8_t row6_eq1 = vld1_u8(coef_eq1_bits + 6 * DCTSIZE);
+  uint8x8_t row7_eq1 = vld1_u8(coef_eq1_bits + 7 * DCTSIZE);
+
+  row0_eq1 = vand_u8(row0_eq1, bitmap_mask);
+  row1_eq1 = vand_u8(row1_eq1, bitmap_mask);
+  row2_eq1 = vand_u8(row2_eq1, bitmap_mask);
+  row3_eq1 = vand_u8(row3_eq1, bitmap_mask);
+  row4_eq1 = vand_u8(row4_eq1, bitmap_mask);
+  row5_eq1 = vand_u8(row5_eq1, bitmap_mask);
+  row6_eq1 = vand_u8(row6_eq1, bitmap_mask);
+  row7_eq1 = vand_u8(row7_eq1, bitmap_mask);
+
+  bitmap_rows_01 = vpadd_u8(row0_eq1, row1_eq1);
+  bitmap_rows_23 = vpadd_u8(row2_eq1, row3_eq1);
+  bitmap_rows_45 = vpadd_u8(row4_eq1, row5_eq1);
+  bitmap_rows_67 = vpadd_u8(row6_eq1, row7_eq1);
+  bitmap_rows_0123 = vpadd_u8(bitmap_rows_01, bitmap_rows_23);
+  bitmap_rows_4567 = vpadd_u8(bitmap_rows_45, bitmap_rows_67);
+  bitmap_all = vpadd_u8(bitmap_rows_0123, bitmap_rows_4567);
+
+#if defined(__aarch64__) || defined(_M_ARM64)
+  /* Move bitmap to a 64-bit scalar register. */
+  bitmap = vget_lane_u64(vreinterpret_u64_u8(bitmap_all), 0);
+
+  /* Return EOB position. */
+  if (bitmap == 0) {
+    /* EOB position is defined to be 0 if all coefficients != 1. */
+    return 0;
+  } else {
+    return 63 - BUILTIN_CLZLL(bitmap);
+  }
+#else
+  /* Move bitmap to two 32-bit scalar registers. */
+  bitmap0 = vget_lane_u32(vreinterpret_u32_u8(bitmap_all), 0);
+  bitmap1 = vget_lane_u32(vreinterpret_u32_u8(bitmap_all), 1);
+
+  /* Return EOB position. */
+  if (bitmap0 == 0 && bitmap1 == 0) {
+    return 0;
+  } else if (bitmap1 != 0) {
+    return 63 - BUILTIN_CLZ(bitmap1);
+  } else {
+    return 31 - BUILTIN_CLZ(bitmap0);
+  }
+#endif
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jcsample-neon.c Added

@@ -0,0 +1,192 @@
+/*
+ * jcsample-neon.c - downsampling (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+
+#include <arm_neon.h>
+
+
+ALIGN(16) static const uint8_t jsimd_h2_downsample_consts = {
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,   /* Pad 0 */
+  0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,   /* Pad 1 */
+  0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0E,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,   /* Pad 2 */
+  0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0D, 0x0D,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,   /* Pad 3 */
+  0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0C, 0x0C, 0x0C,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,   /* Pad 4 */
+  0x08, 0x09, 0x0A, 0x0B, 0x0B, 0x0B, 0x0B, 0x0B,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,   /* Pad 5 */
+  0x08, 0x09, 0x0A, 0x0A, 0x0A, 0x0A, 0x0A, 0x0A,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,   /* Pad 6 */
+  0x08, 0x09, 0x09, 0x09, 0x09, 0x09, 0x09, 0x09,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,   /* Pad 7 */
+  0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,   /* Pad 8 */
+  0x07, 0x07, 0x07, 0x07, 0x07, 0x07, 0x07, 0x07,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x06,   /* Pad 9 */
+  0x06, 0x06, 0x06, 0x06, 0x06, 0x06, 0x06, 0x06,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x05, 0x05,   /* Pad 10 */
+  0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05,
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x04, 0x04, 0x04,   /* Pad 11 */
+  0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04,
+  0x00, 0x01, 0x02, 0x03, 0x03, 0x03, 0x03, 0x03,   /* Pad 12 */
+  0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03,
+  0x00, 0x01, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02,   /* Pad 13 */
+  0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02,
+  0x00, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01,   /* Pad 14 */
+  0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01,
+  0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,   /* Pad 15 */
+  0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
+};
+
+
+/* Downsample pixel values of a single component.
+ * This version handles the common case of 2:1 horizontal and 1:1 vertical,
+ * without smoothing.
+ */
+
+void jsimd_h2v1_downsample_neon(JDIMENSION image_width, int max_v_samp_factor,
+                                JDIMENSION v_samp_factor,
+                                JDIMENSION width_in_blocks,
+                                JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  JSAMPROW inptr, outptr;
+  /* Load expansion mask to pad remaining elements of last DCT block. */
+  const int mask_offset = 16 * ((width_in_blocks * 2 * DCTSIZE) - image_width);
+  const uint8x16_t expand_mask =
+    vld1q_u8(&jsimd_h2_downsample_constsmask_offset);
+  /* Load bias pattern (alternating every pixel.) */
+  /* { 0, 1, 0, 1, 0, 1, 0, 1 } */
+  const uint16x8_t bias = vreinterpretq_u16_u32(vdupq_n_u32(0x00010000));
+  unsigned i, outrow;
+
+  for (outrow = 0; outrow < v_samp_factor; outrow++) {
+    outptr = output_dataoutrow;
+    inptr = input_dataoutrow;
+
+    /* Downsample all but the last DCT block of pixels. */
+    for (i = 0; i < width_in_blocks - 1; i++) {
+      uint8x16_t pixels = vld1q_u8(inptr + i * 2 * DCTSIZE);
+      /* Add adjacent pixel values, widen to 16-bit, and add bias. */
+      uint16x8_t samples_u16 = vpadalq_u8(bias, pixels);
+      /* Divide total by 2 and narrow to 8-bit. */
+      uint8x8_t samples_u8 = vshrn_n_u16(samples_u16, 1);
+      /* Store samples to memory. */
+      vst1_u8(outptr + i * DCTSIZE, samples_u8);
+    }
+
+    /* Load pixels in last DCT block into a table. */
+    uint8x16_t pixels = vld1q_u8(inptr + (width_in_blocks - 1) * 2 * DCTSIZE);
+#if defined(__aarch64__) || defined(_M_ARM64)
+    /* Pad the empty elements with the value of the last pixel. */
+    pixels = vqtbl1q_u8(pixels, expand_mask);
+#else
+    uint8x8x2_t table = { { vget_low_u8(pixels), vget_high_u8(pixels) } };
+    pixels = vcombine_u8(vtbl2_u8(table, vget_low_u8(expand_mask)),
+                         vtbl2_u8(table, vget_high_u8(expand_mask)));
+#endif
+    /* Add adjacent pixel values, widen to 16-bit, and add bias. */
+    uint16x8_t samples_u16 = vpadalq_u8(bias, pixels);
+    /* Divide total by 2, narrow to 8-bit, and store. */
+    uint8x8_t samples_u8 = vshrn_n_u16(samples_u16, 1);
+    vst1_u8(outptr + (width_in_blocks - 1) * DCTSIZE, samples_u8);
+  }
+}
+
+
+/* Downsample pixel values of a single component.
+ * This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+ * without smoothing.
+ */
+
+void jsimd_h2v2_downsample_neon(JDIMENSION image_width, int max_v_samp_factor,
+                                JDIMENSION v_samp_factor,
+                                JDIMENSION width_in_blocks,
+                                JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  JSAMPROW inptr0, inptr1, outptr;
+  /* Load expansion mask to pad remaining elements of last DCT block. */
+  const int mask_offset = 16 * ((width_in_blocks * 2 * DCTSIZE) - image_width);
+  const uint8x16_t expand_mask =
+    vld1q_u8(&jsimd_h2_downsample_constsmask_offset);
+  /* Load bias pattern (alternating every pixel.) */
+  /* { 1, 2, 1, 2, 1, 2, 1, 2 } */
+  const uint16x8_t bias = vreinterpretq_u16_u32(vdupq_n_u32(0x00020001));
+  unsigned i, outrow;
+
+  for (outrow = 0; outrow < v_samp_factor; outrow++) {
+    outptr = output_dataoutrow;
+    inptr0 = input_dataoutrow;
+    inptr1 = input_dataoutrow + 1;
+
+    /* Downsample all but the last DCT block of pixels. */
+    for (i = 0; i < width_in_blocks - 1; i++) {
+      uint8x16_t pixels_r0 = vld1q_u8(inptr0 + i * 2 * DCTSIZE);
+      uint8x16_t pixels_r1 = vld1q_u8(inptr1 + i * 2 * DCTSIZE);
+      /* Add adjacent pixel values in row 0, widen to 16-bit, and add bias. */
+      uint16x8_t samples_u16 = vpadalq_u8(bias, pixels_r0);
+      /* Add adjacent pixel values in row 1, widen to 16-bit, and accumulate.
+       */
+      samples_u16 = vpadalq_u8(samples_u16, pixels_r1);
+      /* Divide total by 4 and narrow to 8-bit. */
+      uint8x8_t samples_u8 = vshrn_n_u16(samples_u16, 2);
+      /* Store samples to memory and increment pointers. */
+      vst1_u8(outptr + i * DCTSIZE, samples_u8);
+    }
+
+    /* Load pixels in last DCT block into a table. */
+    uint8x16_t pixels_r0 =
+      vld1q_u8(inptr0 + (width_in_blocks - 1) * 2 * DCTSIZE);
+    uint8x16_t pixels_r1 =
+      vld1q_u8(inptr1 + (width_in_blocks - 1) * 2 * DCTSIZE);
+#if defined(__aarch64__) || defined(_M_ARM64)
+    /* Pad the empty elements with the value of the last pixel. */
+    pixels_r0 = vqtbl1q_u8(pixels_r0, expand_mask);
+    pixels_r1 = vqtbl1q_u8(pixels_r1, expand_mask);
+#else
+    uint8x8x2_t table_r0 =
+      { { vget_low_u8(pixels_r0), vget_high_u8(pixels_r0) } };
+    uint8x8x2_t table_r1 =
+      { { vget_low_u8(pixels_r1), vget_high_u8(pixels_r1) } };
+    pixels_r0 = vcombine_u8(vtbl2_u8(table_r0, vget_low_u8(expand_mask)),
+                            vtbl2_u8(table_r0, vget_high_u8(expand_mask)));
+    pixels_r1 = vcombine_u8(vtbl2_u8(table_r1, vget_low_u8(expand_mask)),
+                            vtbl2_u8(table_r1, vget_high_u8(expand_mask)));
+#endif
+    /* Add adjacent pixel values in row 0, widen to 16-bit, and add bias. */
+    uint16x8_t samples_u16 = vpadalq_u8(bias, pixels_r0);
+    /* Add adjacent pixel values in row 1, widen to 16-bit, and accumulate. */
+    samples_u16 = vpadalq_u8(samples_u16, pixels_r1);
+    /* Divide total by 4, narrow to 8-bit, and store. */
+    uint8x8_t samples_u8 = vshrn_n_u16(samples_u16, 2);
+    vst1_u8(outptr + (width_in_blocks - 1) * DCTSIZE, samples_u8);
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jdcolext-neon.c Added

@@ -0,0 +1,374 @@
+/*
+ * jdcolext-neon.c - colorspace conversion (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdcolor-neon.c. */
+
+
+/* YCbCr -> RGB conversion is defined by the following equations:
+ *    R = Y                        + 1.40200 * (Cr - 128)
+ *    G = Y - 0.34414 * (Cb - 128) - 0.71414 * (Cr - 128)
+ *    B = Y + 1.77200 * (Cb - 128)
+ *
+ * Scaled integer constants are used to avoid floating-point arithmetic:
+ *    0.3441467 = 11277 * 2^-15
+ *    0.7141418 = 23401 * 2^-15
+ *    1.4020386 = 22971 * 2^-14
+ *    1.7720337 = 29033 * 2^-14
+ * These constants are defined in jdcolor-neon.c.
+ *
+ * To ensure correct results, rounding is used when descaling.
+ */
+
+/* Notes on safe memory access for YCbCr -> RGB conversion routines:
+ *
+ * Input memory buffers can be safely overread up to the next multiple of
+ * ALIGN_SIZE bytes, since they are always allocated by alloc_sarray() in
+ * jmemmgr.c.
+ *
+ * The output buffer cannot safely be written beyond output_width, since
+ * output_buf points to a possibly unpadded row in the decompressed image
+ * buffer allocated by the calling program.
+ */
+
+void jsimd_ycc_rgb_convert_neon(JDIMENSION output_width, JSAMPIMAGE input_buf,
+                                JDIMENSION input_row, JSAMPARRAY output_buf,
+                                int num_rows)
+{
+  JSAMPROW outptr;
+  /* Pointers to Y, Cb, and Cr data */
+  JSAMPROW inptr0, inptr1, inptr2;
+
+  const int16x4_t consts = vld1_s16(jsimd_ycc_rgb_convert_neon_consts);
+  const int16x8_t neg_128 = vdupq_n_s16(-128);
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+    int cols_remaining = output_width;
+    for (; cols_remaining >= 16; cols_remaining -= 16) {
+      uint8x16_t y  = vld1q_u8(inptr0);
+      uint8x16_t cb = vld1q_u8(inptr1);
+      uint8x16_t cr = vld1q_u8(inptr2);
+      /* Subtract 128 from Cb and Cr. */
+      int16x8_t cr_128_l =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128),
+                                       vget_low_u8(cr)));
+      int16x8_t cr_128_h =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128),
+                                       vget_high_u8(cr)));
+      int16x8_t cb_128_l =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128),
+                                       vget_low_u8(cb)));
+      int16x8_t cb_128_h =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128),
+                                       vget_high_u8(cb)));
+      /* Compute G-Y: - 0.34414 * (Cb - 128) - 0.71414 * (Cr - 128) */
+      int32x4_t g_sub_y_ll = vmull_lane_s16(vget_low_s16(cb_128_l), consts, 0);
+      int32x4_t g_sub_y_lh = vmull_lane_s16(vget_high_s16(cb_128_l),
+                                            consts, 0);
+      int32x4_t g_sub_y_hl = vmull_lane_s16(vget_low_s16(cb_128_h), consts, 0);
+      int32x4_t g_sub_y_hh = vmull_lane_s16(vget_high_s16(cb_128_h),
+                                            consts, 0);
+      g_sub_y_ll = vmlsl_lane_s16(g_sub_y_ll, vget_low_s16(cr_128_l),
+                                  consts, 1);
+      g_sub_y_lh = vmlsl_lane_s16(g_sub_y_lh, vget_high_s16(cr_128_l),
+                                  consts, 1);
+      g_sub_y_hl = vmlsl_lane_s16(g_sub_y_hl, vget_low_s16(cr_128_h),
+                                  consts, 1);
+      g_sub_y_hh = vmlsl_lane_s16(g_sub_y_hh, vget_high_s16(cr_128_h),
+                                  consts, 1);
+      /* Descale G components: shift right 15, round, and narrow to 16-bit. */
+      int16x8_t g_sub_y_l = vcombine_s16(vrshrn_n_s32(g_sub_y_ll, 15),
+                                         vrshrn_n_s32(g_sub_y_lh, 15));
+      int16x8_t g_sub_y_h = vcombine_s16(vrshrn_n_s32(g_sub_y_hl, 15),
+                                         vrshrn_n_s32(g_sub_y_hh, 15));
+      /* Compute R-Y: 1.40200 * (Cr - 128) */
+      int16x8_t r_sub_y_l = vqrdmulhq_lane_s16(vshlq_n_s16(cr_128_l, 1),
+                                               consts, 2);
+      int16x8_t r_sub_y_h = vqrdmulhq_lane_s16(vshlq_n_s16(cr_128_h, 1),
+                                               consts, 2);
+      /* Compute B-Y: 1.77200 * (Cb - 128) */
+      int16x8_t b_sub_y_l = vqrdmulhq_lane_s16(vshlq_n_s16(cb_128_l, 1),
+                                               consts, 3);
+      int16x8_t b_sub_y_h = vqrdmulhq_lane_s16(vshlq_n_s16(cb_128_h, 1),
+                                               consts, 3);
+      /* Add Y. */
+      int16x8_t r_l =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y_l),
+                                       vget_low_u8(y)));
+      int16x8_t r_h =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y_h),
+                                       vget_high_u8(y)));
+      int16x8_t b_l =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y_l),
+                                       vget_low_u8(y)));
+      int16x8_t b_h =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y_h),
+                                       vget_high_u8(y)));
+      int16x8_t g_l =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y_l),
+                                       vget_low_u8(y)));
+      int16x8_t g_h =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y_h),
+                                       vget_high_u8(y)));
+
+#if RGB_PIXELSIZE == 4
+      uint8x16x4_t rgba;
+      /* Convert each component to unsigned and narrow, clamping to 0-255. */
+      rgba.valRGB_RED = vcombine_u8(vqmovun_s16(r_l), vqmovun_s16(r_h));
+      rgba.valRGB_GREEN = vcombine_u8(vqmovun_s16(g_l), vqmovun_s16(g_h));
+      rgba.valRGB_BLUE = vcombine_u8(vqmovun_s16(b_l), vqmovun_s16(b_h));
+      /* Set alpha channel to opaque (0xFF). */
+      rgba.valRGB_ALPHA = vdupq_n_u8(0xFF);
+      /* Store RGBA pixel data to memory. */
+      vst4q_u8(outptr, rgba);
+#elif RGB_PIXELSIZE == 3
+      uint8x16x3_t rgb;
+      /* Convert each component to unsigned and narrow, clamping to 0-255. */
+      rgb.valRGB_RED = vcombine_u8(vqmovun_s16(r_l), vqmovun_s16(r_h));
+      rgb.valRGB_GREEN = vcombine_u8(vqmovun_s16(g_l), vqmovun_s16(g_h));
+      rgb.valRGB_BLUE = vcombine_u8(vqmovun_s16(b_l), vqmovun_s16(b_h));
+      /* Store RGB pixel data to memory. */
+      vst3q_u8(outptr, rgb);
+#else
+      /* Pack R, G, and B values in ratio 5:6:5. */
+      uint16x8_t rgb565_l = vqshluq_n_s16(r_l, 8);
+      rgb565_l = vsriq_n_u16(rgb565_l, vqshluq_n_s16(g_l, 8), 5);
+      rgb565_l = vsriq_n_u16(rgb565_l, vqshluq_n_s16(b_l, 8), 11);
+      uint16x8_t rgb565_h = vqshluq_n_s16(r_h, 8);
+      rgb565_h = vsriq_n_u16(rgb565_h, vqshluq_n_s16(g_h, 8), 5);
+      rgb565_h = vsriq_n_u16(rgb565_h, vqshluq_n_s16(b_h, 8), 11);
+      /* Store RGB pixel data to memory. */
+      vst1q_u16((uint16_t *)outptr, rgb565_l);
+      vst1q_u16(((uint16_t *)outptr) + 8, rgb565_h);
+#endif
+
+      /* Increment pointers. */
+      inptr0 += 16;
+      inptr1 += 16;
+      inptr2 += 16;
+      outptr += (RGB_PIXELSIZE * 16);
+    }
+
+    if (cols_remaining >= 8) {
+      uint8x8_t y  = vld1_u8(inptr0);
+      uint8x8_t cb = vld1_u8(inptr1);
+      uint8x8_t cr = vld1_u8(inptr2);
+      /* Subtract 128 from Cb and Cr. */
+      int16x8_t cr_128 =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cr));
+      int16x8_t cb_128 =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cb));
+      /* Compute G-Y: - 0.34414 * (Cb - 128) - 0.71414 * (Cr - 128) */
+      int32x4_t g_sub_y_l = vmull_lane_s16(vget_low_s16(cb_128), consts, 0);
+      int32x4_t g_sub_y_h = vmull_lane_s16(vget_high_s16(cb_128), consts, 0);
+      g_sub_y_l = vmlsl_lane_s16(g_sub_y_l, vget_low_s16(cr_128), consts, 1);
+      g_sub_y_h = vmlsl_lane_s16(g_sub_y_h, vget_high_s16(cr_128), consts, 1);
+      /* Descale G components: shift right 15, round, and narrow to 16-bit. */
+      int16x8_t g_sub_y = vcombine_s16(vrshrn_n_s32(g_sub_y_l, 15),
+                                       vrshrn_n_s32(g_sub_y_h, 15));
+      /* Compute R-Y: 1.40200 * (Cr - 128) */
+      int16x8_t r_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cr_128, 1),
+                                             consts, 2);
+      /* Compute B-Y: 1.77200 * (Cb - 128) */
+      int16x8_t b_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cb_128, 1),
+                                             consts, 3);
+      /* Add Y. */
+      int16x8_t r =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y), y));
+      int16x8_t b =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y), y));
+      int16x8_t g =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y), y));
+
+#if RGB_PIXELSIZE == 4
+      uint8x8x4_t rgba;
+      /* Convert each component to unsigned and narrow, clamping to 0-255. */
+      rgba.valRGB_RED = vqmovun_s16(r);
+      rgba.valRGB_GREEN = vqmovun_s16(g);
+      rgba.valRGB_BLUE = vqmovun_s16(b);
+      /* Set alpha channel to opaque (0xFF). */
+      rgba.valRGB_ALPHA = vdup_n_u8(0xFF);
+      /* Store RGBA pixel data to memory. */
+      vst4_u8(outptr, rgba);
+#elif RGB_PIXELSIZE == 3
+      uint8x8x3_t rgb;
+      /* Convert each component to unsigned and narrow, clamping to 0-255. */
+      rgb.valRGB_RED = vqmovun_s16(r);
+      rgb.valRGB_GREEN = vqmovun_s16(g);
+      rgb.valRGB_BLUE = vqmovun_s16(b);
+      /* Store RGB pixel data to memory. */
+      vst3_u8(outptr, rgb);
+#else
+      /* Pack R, G, and B values in ratio 5:6:5. */
+      uint16x8_t rgb565 = vqshluq_n_s16(r, 8);
+      rgb565 = vsriq_n_u16(rgb565, vqshluq_n_s16(g, 8), 5);
+      rgb565 = vsriq_n_u16(rgb565, vqshluq_n_s16(b, 8), 11);
+      /* Store RGB pixel data to memory. */
+      vst1q_u16((uint16_t *)outptr, rgb565);
+#endif
+
+      /* Increment pointers. */
+      inptr0 += 8;
+      inptr1 += 8;
+      inptr2 += 8;
+      outptr += (RGB_PIXELSIZE * 8);
+      cols_remaining -= 8;
+    }
+
+    /* Handle the tail elements. */
+    if (cols_remaining > 0) {
+      uint8x8_t y  = vld1_u8(inptr0);
+      uint8x8_t cb = vld1_u8(inptr1);
+      uint8x8_t cr = vld1_u8(inptr2);
+      /* Subtract 128 from Cb and Cr. */
+      int16x8_t cr_128 =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cr));
+      int16x8_t cb_128 =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cb));
+      /* Compute G-Y: - 0.34414 * (Cb - 128) - 0.71414 * (Cr - 128) */
+      int32x4_t g_sub_y_l = vmull_lane_s16(vget_low_s16(cb_128), consts, 0);
+      int32x4_t g_sub_y_h = vmull_lane_s16(vget_high_s16(cb_128), consts, 0);
+      g_sub_y_l = vmlsl_lane_s16(g_sub_y_l, vget_low_s16(cr_128), consts, 1);
+      g_sub_y_h = vmlsl_lane_s16(g_sub_y_h, vget_high_s16(cr_128), consts, 1);
+      /* Descale G components: shift right 15, round, and narrow to 16-bit. */
+      int16x8_t g_sub_y = vcombine_s16(vrshrn_n_s32(g_sub_y_l, 15),
+                                       vrshrn_n_s32(g_sub_y_h, 15));
+      /* Compute R-Y: 1.40200 * (Cr - 128) */
+      int16x8_t r_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cr_128, 1),
+                                             consts, 2);
+      /* Compute B-Y: 1.77200 * (Cb - 128) */
+      int16x8_t b_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cb_128, 1),
+                                             consts, 3);
+      /* Add Y. */
+      int16x8_t r =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y), y));
+      int16x8_t b =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y), y));
+      int16x8_t g =
+        vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y), y));
+
+#if RGB_PIXELSIZE == 4
+      uint8x8x4_t rgba;
+      /* Convert each component to unsigned and narrow, clamping to 0-255. */
+      rgba.valRGB_RED = vqmovun_s16(r);
+      rgba.valRGB_GREEN = vqmovun_s16(g);
+      rgba.valRGB_BLUE = vqmovun_s16(b);
+      /* Set alpha channel to opaque (0xFF). */
+      rgba.valRGB_ALPHA = vdup_n_u8(0xFF);
+      /* Store RGBA pixel data to memory. */
+      switch (cols_remaining) {
+      case 7:
+        vst4_lane_u8(outptr + 6 * RGB_PIXELSIZE, rgba, 6);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 6:
+        vst4_lane_u8(outptr + 5 * RGB_PIXELSIZE, rgba, 5);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 5:
+        vst4_lane_u8(outptr + 4 * RGB_PIXELSIZE, rgba, 4);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 4:
+        vst4_lane_u8(outptr + 3 * RGB_PIXELSIZE, rgba, 3);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 3:
+        vst4_lane_u8(outptr + 2 * RGB_PIXELSIZE, rgba, 2);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 2:
+        vst4_lane_u8(outptr + RGB_PIXELSIZE, rgba, 1);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 1:
+        vst4_lane_u8(outptr, rgba, 0);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      default:
+        break;
+      }
+#elif RGB_PIXELSIZE == 3
+      uint8x8x3_t rgb;
+      /* Convert each component to unsigned and narrow, clamping to 0-255. */
+      rgb.valRGB_RED = vqmovun_s16(r);
+      rgb.valRGB_GREEN = vqmovun_s16(g);
+      rgb.valRGB_BLUE = vqmovun_s16(b);
+      /* Store RGB pixel data to memory. */
+      switch (cols_remaining) {
+      case 7:
+        vst3_lane_u8(outptr + 6 * RGB_PIXELSIZE, rgb, 6);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 6:
+        vst3_lane_u8(outptr + 5 * RGB_PIXELSIZE, rgb, 5);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 5:
+        vst3_lane_u8(outptr + 4 * RGB_PIXELSIZE, rgb, 4);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 4:
+        vst3_lane_u8(outptr + 3 * RGB_PIXELSIZE, rgb, 3);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 3:
+        vst3_lane_u8(outptr + 2 * RGB_PIXELSIZE, rgb, 2);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 2:
+        vst3_lane_u8(outptr + RGB_PIXELSIZE, rgb, 1);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 1:
+        vst3_lane_u8(outptr, rgb, 0);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      default:
+        break;
+      }
+#else
+      /* Pack R, G, and B values in ratio 5:6:5. */
+      uint16x8_t rgb565 = vqshluq_n_s16(r, 8);
+      rgb565 = vsriq_n_u16(rgb565, vqshluq_n_s16(g, 8), 5);
+      rgb565 = vsriq_n_u16(rgb565, vqshluq_n_s16(b, 8), 11);
+      /* Store RGB565 pixel data to memory. */
+      switch (cols_remaining) {
+      case 7:
+        vst1q_lane_u16((uint16_t *)(outptr + 6 * RGB_PIXELSIZE), rgb565, 6);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 6:
+        vst1q_lane_u16((uint16_t *)(outptr + 5 * RGB_PIXELSIZE), rgb565, 5);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 5:
+        vst1q_lane_u16((uint16_t *)(outptr + 4 * RGB_PIXELSIZE), rgb565, 4);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 4:
+        vst1q_lane_u16((uint16_t *)(outptr + 3 * RGB_PIXELSIZE), rgb565, 3);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 3:
+        vst1q_lane_u16((uint16_t *)(outptr + 2 * RGB_PIXELSIZE), rgb565, 2);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 2:
+        vst1q_lane_u16((uint16_t *)(outptr + RGB_PIXELSIZE), rgb565, 1);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      case 1:
+        vst1q_lane_u16((uint16_t *)outptr, rgb565, 0);
+        FALLTHROUGH             /*FALLTHROUGH*/
+      default:
+        break;
+      }
+#endif
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jdcolor-neon.c Added

@@ -0,0 +1,142 @@
+/*
+ * jdcolor-neon.c - colorspace conversion (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "jconfigint.h"
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+
+#include <arm_neon.h>
+
+
+/* YCbCr -> RGB conversion constants */
+
+#define F_0_344  11277  /* 0.3441467 = 11277 * 2^-15 */
+#define F_0_714  23401  /* 0.7141418 = 23401 * 2^-15 */
+#define F_1_402  22971  /* 1.4020386 = 22971 * 2^-14 */
+#define F_1_772  29033  /* 1.7720337 = 29033 * 2^-14 */
+
+ALIGN(16) static const int16_t jsimd_ycc_rgb_convert_neon_consts = {
+  -F_0_344, F_0_714, F_1_402, F_1_772
+};
+
+
+/* Include inline routines for colorspace extensions. */
+
+#include "jdcolext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_ycc_rgb_convert_neon  jsimd_ycc_extrgb_convert_neon
+#include "jdcolext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_neon
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_ALPHA  3
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define jsimd_ycc_rgb_convert_neon  jsimd_ycc_extrgbx_convert_neon
+#include "jdcolext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_neon
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define jsimd_ycc_rgb_convert_neon  jsimd_ycc_extbgr_convert_neon
+#include "jdcolext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_neon
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_ALPHA  3
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define jsimd_ycc_rgb_convert_neon  jsimd_ycc_extbgrx_convert_neon
+#include "jdcolext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_neon
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_ALPHA  0
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define jsimd_ycc_rgb_convert_neon  jsimd_ycc_extxbgr_convert_neon
+#include "jdcolext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_neon
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_ALPHA  0
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define jsimd_ycc_rgb_convert_neon  jsimd_ycc_extxrgb_convert_neon
+#include "jdcolext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_neon
+
+/* YCbCr -> RGB565 Conversion */
+
+#define RGB_PIXELSIZE  2
+#define jsimd_ycc_rgb_convert_neon  jsimd_ycc_rgb565_convert_neon
+#include "jdcolext-neon.c"
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_neon

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jdmerge-neon.c Added

@@ -0,0 +1,145 @@
+/*
+ * jdmerge-neon.c - merged upsampling/color conversion (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "jconfigint.h"
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+
+#include <arm_neon.h>
+
+
+/* YCbCr -> RGB conversion constants */
+
+#define F_0_344  11277  /* 0.3441467 = 11277 * 2^-15 */
+#define F_0_714  23401  /* 0.7141418 = 23401 * 2^-15 */
+#define F_1_402  22971  /* 1.4020386 = 22971 * 2^-14 */
+#define F_1_772  29033  /* 1.7720337 = 29033 * 2^-14 */
+
+ALIGN(16) static const int16_t jsimd_ycc_rgb_convert_neon_consts = {
+  -F_0_344, F_0_714, F_1_402, F_1_772
+};
+
+
+/* Include inline routines for colorspace extensions. */
+
+#include "jdmrgext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_neon  jsimd_h2v1_extrgb_merged_upsample_neon
+#define jsimd_h2v2_merged_upsample_neon  jsimd_h2v2_extrgb_merged_upsample_neon
+#include "jdmrgext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_neon
+#undef jsimd_h2v2_merged_upsample_neon
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_ALPHA  3
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_neon  jsimd_h2v1_extrgbx_merged_upsample_neon
+#define jsimd_h2v2_merged_upsample_neon  jsimd_h2v2_extrgbx_merged_upsample_neon
+#include "jdmrgext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_neon
+#undef jsimd_h2v2_merged_upsample_neon
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_neon  jsimd_h2v1_extbgr_merged_upsample_neon
+#define jsimd_h2v2_merged_upsample_neon  jsimd_h2v2_extbgr_merged_upsample_neon
+#include "jdmrgext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_neon
+#undef jsimd_h2v2_merged_upsample_neon
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_ALPHA  3
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_neon  jsimd_h2v1_extbgrx_merged_upsample_neon
+#define jsimd_h2v2_merged_upsample_neon  jsimd_h2v2_extbgrx_merged_upsample_neon
+#include "jdmrgext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_neon
+#undef jsimd_h2v2_merged_upsample_neon
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_ALPHA  0
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_neon  jsimd_h2v1_extxbgr_merged_upsample_neon
+#define jsimd_h2v2_merged_upsample_neon  jsimd_h2v2_extxbgr_merged_upsample_neon
+#include "jdmrgext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_neon
+#undef jsimd_h2v2_merged_upsample_neon
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_ALPHA  0
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_neon  jsimd_h2v1_extxrgb_merged_upsample_neon
+#define jsimd_h2v2_merged_upsample_neon  jsimd_h2v2_extxrgb_merged_upsample_neon
+#include "jdmrgext-neon.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_ALPHA
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_neon

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jdmrgext-neon.c Added

@@ -0,0 +1,723 @@
+/*
+ * jdmrgext-neon.c - merged upsampling/color conversion (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdmerge-neon.c. */
+
+
+/* These routines combine simple (non-fancy, i.e. non-smooth) h2v1 or h2v2
+ * chroma upsampling and YCbCr -> RGB color conversion into a single function.
+ *
+ * As with the standalone functions, YCbCr -> RGB conversion is defined by the
+ * following equations:
+ *    R = Y                        + 1.40200 * (Cr - 128)
+ *    G = Y - 0.34414 * (Cb - 128) - 0.71414 * (Cr - 128)
+ *    B = Y + 1.77200 * (Cb - 128)
+ *
+ * Scaled integer constants are used to avoid floating-point arithmetic:
+ *    0.3441467 = 11277 * 2^-15
+ *    0.7141418 = 23401 * 2^-15
+ *    1.4020386 = 22971 * 2^-14
+ *    1.7720337 = 29033 * 2^-14
+ * These constants are defined in jdmerge-neon.c.
+ *
+ * To ensure correct results, rounding is used when descaling.
+ */
+
+/* Notes on safe memory access for merged upsampling/YCbCr -> RGB conversion
+ * routines:
+ *
+ * Input memory buffers can be safely overread up to the next multiple of
+ * ALIGN_SIZE bytes, since they are always allocated by alloc_sarray() in
+ * jmemmgr.c.
+ *
+ * The output buffer cannot safely be written beyond output_width, since
+ * output_buf points to a possibly unpadded row in the decompressed image
+ * buffer allocated by the calling program.
+ */
+
+/* Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ */
+
+void jsimd_h2v1_merged_upsample_neon(JDIMENSION output_width,
+                                     JSAMPIMAGE input_buf,
+                                     JDIMENSION in_row_group_ctr,
+                                     JSAMPARRAY output_buf)
+{
+  JSAMPROW outptr;
+  /* Pointers to Y, Cb, and Cr data */
+  JSAMPROW inptr0, inptr1, inptr2;
+
+  const int16x4_t consts = vld1_s16(jsimd_ycc_rgb_convert_neon_consts);
+  const int16x8_t neg_128 = vdupq_n_s16(-128);
+
+  inptr0 = input_buf0in_row_group_ctr;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr = output_buf0;
+
+  int cols_remaining = output_width;
+  for (; cols_remaining >= 16; cols_remaining -= 16) {
+    /* De-interleave Y component values into two separate vectors, one
+     * containing the component values with even-numbered indices and one
+     * containing the component values with odd-numbered indices.
+     */
+    uint8x8x2_t y = vld2_u8(inptr0);
+    uint8x8_t cb = vld1_u8(inptr1);
+    uint8x8_t cr = vld1_u8(inptr2);
+    /* Subtract 128 from Cb and Cr. */
+    int16x8_t cr_128 =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cr));
+    int16x8_t cb_128 =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cb));
+    /* Compute G-Y: - 0.34414 * (Cb - 128) - 0.71414 * (Cr - 128) */
+    int32x4_t g_sub_y_l = vmull_lane_s16(vget_low_s16(cb_128), consts, 0);
+    int32x4_t g_sub_y_h = vmull_lane_s16(vget_high_s16(cb_128), consts, 0);
+    g_sub_y_l = vmlsl_lane_s16(g_sub_y_l, vget_low_s16(cr_128), consts, 1);
+    g_sub_y_h = vmlsl_lane_s16(g_sub_y_h, vget_high_s16(cr_128), consts, 1);
+    /* Descale G components: shift right 15, round, and narrow to 16-bit. */
+    int16x8_t g_sub_y = vcombine_s16(vrshrn_n_s32(g_sub_y_l, 15),
+                                     vrshrn_n_s32(g_sub_y_h, 15));
+    /* Compute R-Y: 1.40200 * (Cr - 128) */
+    int16x8_t r_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cr_128, 1), consts, 2);
+    /* Compute B-Y: 1.77200 * (Cb - 128) */
+    int16x8_t b_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cb_128, 1), consts, 3);
+    /* Add the chroma-derived values (G-Y, R-Y, and B-Y) to both the "even" and
+     * "odd" Y component values.  This effectively upsamples the chroma
+     * components horizontally.
+     */
+    int16x8_t g_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y.val0));
+    int16x8_t r_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y.val0));
+    int16x8_t b_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y.val0));
+    int16x8_t g_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y.val1));
+    int16x8_t r_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y.val1));
+    int16x8_t b_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y.val1));
+    /* Convert each component to unsigned and narrow, clamping to 0-255.
+     * Re-interleave the "even" and "odd" component values.
+     */
+    uint8x8x2_t r = vzip_u8(vqmovun_s16(r_even), vqmovun_s16(r_odd));
+    uint8x8x2_t g = vzip_u8(vqmovun_s16(g_even), vqmovun_s16(g_odd));
+    uint8x8x2_t b = vzip_u8(vqmovun_s16(b_even), vqmovun_s16(b_odd));
+
+#ifdef RGB_ALPHA
+    uint8x16x4_t rgba;
+    rgba.valRGB_RED = vcombine_u8(r.val0, r.val1);
+    rgba.valRGB_GREEN = vcombine_u8(g.val0, g.val1);
+    rgba.valRGB_BLUE = vcombine_u8(b.val0, b.val1);
+    /* Set alpha channel to opaque (0xFF). */
+    rgba.valRGB_ALPHA = vdupq_n_u8(0xFF);
+    /* Store RGBA pixel data to memory. */
+    vst4q_u8(outptr, rgba);
+#else
+    uint8x16x3_t rgb;
+    rgb.valRGB_RED = vcombine_u8(r.val0, r.val1);
+    rgb.valRGB_GREEN = vcombine_u8(g.val0, g.val1);
+    rgb.valRGB_BLUE = vcombine_u8(b.val0, b.val1);
+    /* Store RGB pixel data to memory. */
+    vst3q_u8(outptr, rgb);
+#endif
+
+    /* Increment pointers. */
+    inptr0 += 16;
+    inptr1 += 8;
+    inptr2 += 8;
+    outptr += (RGB_PIXELSIZE * 16);
+  }
+
+  if (cols_remaining > 0) {
+    /* De-interleave Y component values into two separate vectors, one
+     * containing the component values with even-numbered indices and one
+     * containing the component values with odd-numbered indices.
+     */
+    uint8x8x2_t y = vld2_u8(inptr0);
+    uint8x8_t cb = vld1_u8(inptr1);
+    uint8x8_t cr = vld1_u8(inptr2);
+    /* Subtract 128 from Cb and Cr. */
+    int16x8_t cr_128 =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cr));
+    int16x8_t cb_128 =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cb));
+    /* Compute G-Y: - 0.34414 * (Cb - 128) - 0.71414 * (Cr - 128) */
+    int32x4_t g_sub_y_l = vmull_lane_s16(vget_low_s16(cb_128), consts, 0);
+    int32x4_t g_sub_y_h = vmull_lane_s16(vget_high_s16(cb_128), consts, 0);
+    g_sub_y_l = vmlsl_lane_s16(g_sub_y_l, vget_low_s16(cr_128), consts, 1);
+    g_sub_y_h = vmlsl_lane_s16(g_sub_y_h, vget_high_s16(cr_128), consts, 1);
+    /* Descale G components: shift right 15, round, and narrow to 16-bit. */
+    int16x8_t g_sub_y = vcombine_s16(vrshrn_n_s32(g_sub_y_l, 15),
+                                     vrshrn_n_s32(g_sub_y_h, 15));
+    /* Compute R-Y: 1.40200 * (Cr - 128) */
+    int16x8_t r_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cr_128, 1), consts, 2);
+    /* Compute B-Y: 1.77200 * (Cb - 128) */
+    int16x8_t b_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cb_128, 1), consts, 3);
+    /* Add the chroma-derived values (G-Y, R-Y, and B-Y) to both the "even" and
+     * "odd" Y component values.  This effectively upsamples the chroma
+     * components horizontally.
+     */
+    int16x8_t g_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y.val0));
+    int16x8_t r_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y.val0));
+    int16x8_t b_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y.val0));
+    int16x8_t g_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y.val1));
+    int16x8_t r_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y.val1));
+    int16x8_t b_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y.val1));
+    /* Convert each component to unsigned and narrow, clamping to 0-255.
+     * Re-interleave the "even" and "odd" component values.
+     */
+    uint8x8x2_t r = vzip_u8(vqmovun_s16(r_even), vqmovun_s16(r_odd));
+    uint8x8x2_t g = vzip_u8(vqmovun_s16(g_even), vqmovun_s16(g_odd));
+    uint8x8x2_t b = vzip_u8(vqmovun_s16(b_even), vqmovun_s16(b_odd));
+
+#ifdef RGB_ALPHA
+    uint8x8x4_t rgba_h;
+    rgba_h.valRGB_RED = r.val1;
+    rgba_h.valRGB_GREEN = g.val1;
+    rgba_h.valRGB_BLUE = b.val1;
+    /* Set alpha channel to opaque (0xFF). */
+    rgba_h.valRGB_ALPHA = vdup_n_u8(0xFF);
+    uint8x8x4_t rgba_l;
+    rgba_l.valRGB_RED = r.val0;
+    rgba_l.valRGB_GREEN = g.val0;
+    rgba_l.valRGB_BLUE = b.val0;
+    /* Set alpha channel to opaque (0xFF). */
+    rgba_l.valRGB_ALPHA = vdup_n_u8(0xFF);
+    /* Store RGBA pixel data to memory. */
+    switch (cols_remaining) {
+    case 15:
+      vst4_lane_u8(outptr + 14 * RGB_PIXELSIZE, rgba_h, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 14:
+      vst4_lane_u8(outptr + 13 * RGB_PIXELSIZE, rgba_h, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 13:
+      vst4_lane_u8(outptr + 12 * RGB_PIXELSIZE, rgba_h, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 12:
+      vst4_lane_u8(outptr + 11 * RGB_PIXELSIZE, rgba_h, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 11:
+      vst4_lane_u8(outptr + 10 * RGB_PIXELSIZE, rgba_h, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 10:
+      vst4_lane_u8(outptr + 9 * RGB_PIXELSIZE, rgba_h, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 9:
+      vst4_lane_u8(outptr + 8 * RGB_PIXELSIZE, rgba_h, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 8:
+      vst4_u8(outptr, rgba_l);
+      break;
+    case 7:
+      vst4_lane_u8(outptr + 6 * RGB_PIXELSIZE, rgba_l, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 6:
+      vst4_lane_u8(outptr + 5 * RGB_PIXELSIZE, rgba_l, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 5:
+      vst4_lane_u8(outptr + 4 * RGB_PIXELSIZE, rgba_l, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 4:
+      vst4_lane_u8(outptr + 3 * RGB_PIXELSIZE, rgba_l, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 3:
+      vst4_lane_u8(outptr + 2 * RGB_PIXELSIZE, rgba_l, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 2:
+      vst4_lane_u8(outptr + RGB_PIXELSIZE, rgba_l, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 1:
+      vst4_lane_u8(outptr, rgba_l, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    default:
+      break;
+    }
+#else
+    uint8x8x3_t rgb_h;
+    rgb_h.valRGB_RED = r.val1;
+    rgb_h.valRGB_GREEN = g.val1;
+    rgb_h.valRGB_BLUE = b.val1;
+    uint8x8x3_t rgb_l;
+    rgb_l.valRGB_RED = r.val0;
+    rgb_l.valRGB_GREEN = g.val0;
+    rgb_l.valRGB_BLUE = b.val0;
+    /* Store RGB pixel data to memory. */
+    switch (cols_remaining) {
+    case 15:
+      vst3_lane_u8(outptr + 14 * RGB_PIXELSIZE, rgb_h, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 14:
+      vst3_lane_u8(outptr + 13 * RGB_PIXELSIZE, rgb_h, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 13:
+      vst3_lane_u8(outptr + 12 * RGB_PIXELSIZE, rgb_h, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 12:
+      vst3_lane_u8(outptr + 11 * RGB_PIXELSIZE, rgb_h, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 11:
+      vst3_lane_u8(outptr + 10 * RGB_PIXELSIZE, rgb_h, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 10:
+      vst3_lane_u8(outptr + 9 * RGB_PIXELSIZE, rgb_h, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 9:
+      vst3_lane_u8(outptr + 8 * RGB_PIXELSIZE, rgb_h, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 8:
+      vst3_u8(outptr, rgb_l);
+      break;
+    case 7:
+      vst3_lane_u8(outptr + 6 * RGB_PIXELSIZE, rgb_l, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 6:
+      vst3_lane_u8(outptr + 5 * RGB_PIXELSIZE, rgb_l, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 5:
+      vst3_lane_u8(outptr + 4 * RGB_PIXELSIZE, rgb_l, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 4:
+      vst3_lane_u8(outptr + 3 * RGB_PIXELSIZE, rgb_l, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 3:
+      vst3_lane_u8(outptr + 2 * RGB_PIXELSIZE, rgb_l, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 2:
+      vst3_lane_u8(outptr + RGB_PIXELSIZE, rgb_l, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 1:
+      vst3_lane_u8(outptr, rgb_l, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    default:
+      break;
+    }
+#endif
+  }
+}
+
+
+/* Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+ *
+ * See comments above for details regarding color conversion and safe memory
+ * access.
+ */
+
+void jsimd_h2v2_merged_upsample_neon(JDIMENSION output_width,
+                                     JSAMPIMAGE input_buf,
+                                     JDIMENSION in_row_group_ctr,
+                                     JSAMPARRAY output_buf)
+{
+  JSAMPROW outptr0, outptr1;
+  /* Pointers to Y (both rows), Cb, and Cr data */
+  JSAMPROW inptr0_0, inptr0_1, inptr1, inptr2;
+
+  const int16x4_t consts = vld1_s16(jsimd_ycc_rgb_convert_neon_consts);
+  const int16x8_t neg_128 = vdupq_n_s16(-128);
+
+  inptr0_0 = input_buf0in_row_group_ctr * 2;
+  inptr0_1 = input_buf0in_row_group_ctr * 2 + 1;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr0 = output_buf0;
+  outptr1 = output_buf1;
+
+  int cols_remaining = output_width;
+  for (; cols_remaining >= 16; cols_remaining -= 16) {
+    /* For each row, de-interleave Y component values into two separate
+     * vectors, one containing the component values with even-numbered indices
+     * and one containing the component values with odd-numbered indices.
+     */
+    uint8x8x2_t y0 = vld2_u8(inptr0_0);
+    uint8x8x2_t y1 = vld2_u8(inptr0_1);
+    uint8x8_t cb = vld1_u8(inptr1);
+    uint8x8_t cr = vld1_u8(inptr2);
+    /* Subtract 128 from Cb and Cr. */
+    int16x8_t cr_128 =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cr));
+    int16x8_t cb_128 =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cb));
+    /* Compute G-Y: - 0.34414 * (Cb - 128) - 0.71414 * (Cr - 128) */
+    int32x4_t g_sub_y_l = vmull_lane_s16(vget_low_s16(cb_128), consts, 0);
+    int32x4_t g_sub_y_h = vmull_lane_s16(vget_high_s16(cb_128), consts, 0);
+    g_sub_y_l = vmlsl_lane_s16(g_sub_y_l, vget_low_s16(cr_128), consts, 1);
+    g_sub_y_h = vmlsl_lane_s16(g_sub_y_h, vget_high_s16(cr_128), consts, 1);
+    /* Descale G components: shift right 15, round, and narrow to 16-bit. */
+    int16x8_t g_sub_y = vcombine_s16(vrshrn_n_s32(g_sub_y_l, 15),
+                                     vrshrn_n_s32(g_sub_y_h, 15));
+    /* Compute R-Y: 1.40200 * (Cr - 128) */
+    int16x8_t r_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cr_128, 1), consts, 2);
+    /* Compute B-Y: 1.77200 * (Cb - 128) */
+    int16x8_t b_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cb_128, 1), consts, 3);
+    /* For each row, add the chroma-derived values (G-Y, R-Y, and B-Y) to both
+     * the "even" and "odd" Y component values.  This effectively upsamples the
+     * chroma components both horizontally and vertically.
+     */
+    int16x8_t g0_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y0.val0));
+    int16x8_t r0_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y0.val0));
+    int16x8_t b0_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y0.val0));
+    int16x8_t g0_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y0.val1));
+    int16x8_t r0_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y0.val1));
+    int16x8_t b0_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y0.val1));
+    int16x8_t g1_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y1.val0));
+    int16x8_t r1_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y1.val0));
+    int16x8_t b1_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y1.val0));
+    int16x8_t g1_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y1.val1));
+    int16x8_t r1_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y1.val1));
+    int16x8_t b1_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y1.val1));
+    /* Convert each component to unsigned and narrow, clamping to 0-255.
+     * Re-interleave the "even" and "odd" component values.
+     */
+    uint8x8x2_t r0 = vzip_u8(vqmovun_s16(r0_even), vqmovun_s16(r0_odd));
+    uint8x8x2_t r1 = vzip_u8(vqmovun_s16(r1_even), vqmovun_s16(r1_odd));
+    uint8x8x2_t g0 = vzip_u8(vqmovun_s16(g0_even), vqmovun_s16(g0_odd));
+    uint8x8x2_t g1 = vzip_u8(vqmovun_s16(g1_even), vqmovun_s16(g1_odd));
+    uint8x8x2_t b0 = vzip_u8(vqmovun_s16(b0_even), vqmovun_s16(b0_odd));
+    uint8x8x2_t b1 = vzip_u8(vqmovun_s16(b1_even), vqmovun_s16(b1_odd));
+
+#ifdef RGB_ALPHA
+    uint8x16x4_t rgba0, rgba1;
+    rgba0.valRGB_RED = vcombine_u8(r0.val0, r0.val1);
+    rgba1.valRGB_RED = vcombine_u8(r1.val0, r1.val1);
+    rgba0.valRGB_GREEN = vcombine_u8(g0.val0, g0.val1);
+    rgba1.valRGB_GREEN = vcombine_u8(g1.val0, g1.val1);
+    rgba0.valRGB_BLUE = vcombine_u8(b0.val0, b0.val1);
+    rgba1.valRGB_BLUE = vcombine_u8(b1.val0, b1.val1);
+    /* Set alpha channel to opaque (0xFF). */
+    rgba0.valRGB_ALPHA = vdupq_n_u8(0xFF);
+    rgba1.valRGB_ALPHA = vdupq_n_u8(0xFF);
+    /* Store RGBA pixel data to memory. */
+    vst4q_u8(outptr0, rgba0);
+    vst4q_u8(outptr1, rgba1);
+#else
+    uint8x16x3_t rgb0, rgb1;
+    rgb0.valRGB_RED = vcombine_u8(r0.val0, r0.val1);
+    rgb1.valRGB_RED = vcombine_u8(r1.val0, r1.val1);
+    rgb0.valRGB_GREEN = vcombine_u8(g0.val0, g0.val1);
+    rgb1.valRGB_GREEN = vcombine_u8(g1.val0, g1.val1);
+    rgb0.valRGB_BLUE = vcombine_u8(b0.val0, b0.val1);
+    rgb1.valRGB_BLUE = vcombine_u8(b1.val0, b1.val1);
+    /* Store RGB pixel data to memory. */
+    vst3q_u8(outptr0, rgb0);
+    vst3q_u8(outptr1, rgb1);
+#endif
+
+    /* Increment pointers. */
+    inptr0_0 += 16;
+    inptr0_1 += 16;
+    inptr1 += 8;
+    inptr2 += 8;
+    outptr0 += (RGB_PIXELSIZE * 16);
+    outptr1 += (RGB_PIXELSIZE * 16);
+  }
+
+  if (cols_remaining > 0) {
+    /* For each row, de-interleave Y component values into two separate
+     * vectors, one containing the component values with even-numbered indices
+     * and one containing the component values with odd-numbered indices.
+     */
+    uint8x8x2_t y0 = vld2_u8(inptr0_0);
+    uint8x8x2_t y1 = vld2_u8(inptr0_1);
+    uint8x8_t cb = vld1_u8(inptr1);
+    uint8x8_t cr = vld1_u8(inptr2);
+    /* Subtract 128 from Cb and Cr. */
+    int16x8_t cr_128 =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cr));
+    int16x8_t cb_128 =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(neg_128), cb));
+    /* Compute G-Y: - 0.34414 * (Cb - 128) - 0.71414 * (Cr - 128) */
+    int32x4_t g_sub_y_l = vmull_lane_s16(vget_low_s16(cb_128), consts, 0);
+    int32x4_t g_sub_y_h = vmull_lane_s16(vget_high_s16(cb_128), consts, 0);
+    g_sub_y_l = vmlsl_lane_s16(g_sub_y_l, vget_low_s16(cr_128), consts, 1);
+    g_sub_y_h = vmlsl_lane_s16(g_sub_y_h, vget_high_s16(cr_128), consts, 1);
+    /* Descale G components: shift right 15, round, and narrow to 16-bit. */
+    int16x8_t g_sub_y = vcombine_s16(vrshrn_n_s32(g_sub_y_l, 15),
+                                     vrshrn_n_s32(g_sub_y_h, 15));
+    /* Compute R-Y: 1.40200 * (Cr - 128) */
+    int16x8_t r_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cr_128, 1), consts, 2);
+    /* Compute B-Y: 1.77200 * (Cb - 128) */
+    int16x8_t b_sub_y = vqrdmulhq_lane_s16(vshlq_n_s16(cb_128, 1), consts, 3);
+    /* For each row, add the chroma-derived values (G-Y, R-Y, and B-Y) to both
+     * the "even" and "odd" Y component values.  This effectively upsamples the
+     * chroma components both horizontally and vertically.
+     */
+    int16x8_t g0_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y0.val0));
+    int16x8_t r0_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y0.val0));
+    int16x8_t b0_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y0.val0));
+    int16x8_t g0_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y0.val1));
+    int16x8_t r0_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y0.val1));
+    int16x8_t b0_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y0.val1));
+    int16x8_t g1_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y1.val0));
+    int16x8_t r1_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y1.val0));
+    int16x8_t b1_even =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y1.val0));
+    int16x8_t g1_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(g_sub_y),
+                                     y1.val1));
+    int16x8_t r1_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(r_sub_y),
+                                     y1.val1));
+    int16x8_t b1_odd =
+      vreinterpretq_s16_u16(vaddw_u8(vreinterpretq_u16_s16(b_sub_y),
+                                     y1.val1));
+    /* Convert each component to unsigned and narrow, clamping to 0-255.
+     * Re-interleave the "even" and "odd" component values.
+     */
+    uint8x8x2_t r0 = vzip_u8(vqmovun_s16(r0_even), vqmovun_s16(r0_odd));
+    uint8x8x2_t r1 = vzip_u8(vqmovun_s16(r1_even), vqmovun_s16(r1_odd));
+    uint8x8x2_t g0 = vzip_u8(vqmovun_s16(g0_even), vqmovun_s16(g0_odd));
+    uint8x8x2_t g1 = vzip_u8(vqmovun_s16(g1_even), vqmovun_s16(g1_odd));
+    uint8x8x2_t b0 = vzip_u8(vqmovun_s16(b0_even), vqmovun_s16(b0_odd));
+    uint8x8x2_t b1 = vzip_u8(vqmovun_s16(b1_even), vqmovun_s16(b1_odd));
+
+#ifdef RGB_ALPHA
+    uint8x8x4_t rgba0_h, rgba1_h;
+    rgba0_h.valRGB_RED = r0.val1;
+    rgba1_h.valRGB_RED = r1.val1;
+    rgba0_h.valRGB_GREEN = g0.val1;
+    rgba1_h.valRGB_GREEN = g1.val1;
+    rgba0_h.valRGB_BLUE = b0.val1;
+    rgba1_h.valRGB_BLUE = b1.val1;
+    /* Set alpha channel to opaque (0xFF). */
+    rgba0_h.valRGB_ALPHA = vdup_n_u8(0xFF);
+    rgba1_h.valRGB_ALPHA = vdup_n_u8(0xFF);
+
+    uint8x8x4_t rgba0_l, rgba1_l;
+    rgba0_l.valRGB_RED = r0.val0;
+    rgba1_l.valRGB_RED = r1.val0;
+    rgba0_l.valRGB_GREEN = g0.val0;
+    rgba1_l.valRGB_GREEN = g1.val0;
+    rgba0_l.valRGB_BLUE = b0.val0;
+    rgba1_l.valRGB_BLUE = b1.val0;
+    /* Set alpha channel to opaque (0xFF). */
+    rgba0_l.valRGB_ALPHA = vdup_n_u8(0xFF);
+    rgba1_l.valRGB_ALPHA = vdup_n_u8(0xFF);
+    /* Store RGBA pixel data to memory. */
+    switch (cols_remaining) {
+    case 15:
+      vst4_lane_u8(outptr0 + 14 * RGB_PIXELSIZE, rgba0_h, 6);
+      vst4_lane_u8(outptr1 + 14 * RGB_PIXELSIZE, rgba1_h, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 14:
+      vst4_lane_u8(outptr0 + 13 * RGB_PIXELSIZE, rgba0_h, 5);
+      vst4_lane_u8(outptr1 + 13 * RGB_PIXELSIZE, rgba1_h, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 13:
+      vst4_lane_u8(outptr0 + 12 * RGB_PIXELSIZE, rgba0_h, 4);
+      vst4_lane_u8(outptr1 + 12 * RGB_PIXELSIZE, rgba1_h, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 12:
+      vst4_lane_u8(outptr0 + 11 * RGB_PIXELSIZE, rgba0_h, 3);
+      vst4_lane_u8(outptr1 + 11 * RGB_PIXELSIZE, rgba1_h, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 11:
+      vst4_lane_u8(outptr0 + 10 * RGB_PIXELSIZE, rgba0_h, 2);
+      vst4_lane_u8(outptr1 + 10 * RGB_PIXELSIZE, rgba1_h, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 10:
+      vst4_lane_u8(outptr0 + 9 * RGB_PIXELSIZE, rgba0_h, 1);
+      vst4_lane_u8(outptr1 + 9 * RGB_PIXELSIZE, rgba1_h, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 9:
+      vst4_lane_u8(outptr0 + 8 * RGB_PIXELSIZE, rgba0_h, 0);
+      vst4_lane_u8(outptr1 + 8 * RGB_PIXELSIZE, rgba1_h, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 8:
+      vst4_u8(outptr0, rgba0_l);
+      vst4_u8(outptr1, rgba1_l);
+      break;
+    case 7:
+      vst4_lane_u8(outptr0 + 6 * RGB_PIXELSIZE, rgba0_l, 6);
+      vst4_lane_u8(outptr1 + 6 * RGB_PIXELSIZE, rgba1_l, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 6:
+      vst4_lane_u8(outptr0 + 5 * RGB_PIXELSIZE, rgba0_l, 5);
+      vst4_lane_u8(outptr1 + 5 * RGB_PIXELSIZE, rgba1_l, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 5:
+      vst4_lane_u8(outptr0 + 4 * RGB_PIXELSIZE, rgba0_l, 4);
+      vst4_lane_u8(outptr1 + 4 * RGB_PIXELSIZE, rgba1_l, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 4:
+      vst4_lane_u8(outptr0 + 3 * RGB_PIXELSIZE, rgba0_l, 3);
+      vst4_lane_u8(outptr1 + 3 * RGB_PIXELSIZE, rgba1_l, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 3:
+      vst4_lane_u8(outptr0 + 2 * RGB_PIXELSIZE, rgba0_l, 2);
+      vst4_lane_u8(outptr1 + 2 * RGB_PIXELSIZE, rgba1_l, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 2:
+      vst4_lane_u8(outptr0 + 1 * RGB_PIXELSIZE, rgba0_l, 1);
+      vst4_lane_u8(outptr1 + 1 * RGB_PIXELSIZE, rgba1_l, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 1:
+      vst4_lane_u8(outptr0, rgba0_l, 0);
+      vst4_lane_u8(outptr1, rgba1_l, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    default:
+      break;
+    }
+#else
+    uint8x8x3_t rgb0_h, rgb1_h;
+    rgb0_h.valRGB_RED = r0.val1;
+    rgb1_h.valRGB_RED = r1.val1;
+    rgb0_h.valRGB_GREEN = g0.val1;
+    rgb1_h.valRGB_GREEN = g1.val1;
+    rgb0_h.valRGB_BLUE = b0.val1;
+    rgb1_h.valRGB_BLUE = b1.val1;
+
+    uint8x8x3_t rgb0_l, rgb1_l;
+    rgb0_l.valRGB_RED = r0.val0;
+    rgb1_l.valRGB_RED = r1.val0;
+    rgb0_l.valRGB_GREEN = g0.val0;
+    rgb1_l.valRGB_GREEN = g1.val0;
+    rgb0_l.valRGB_BLUE = b0.val0;
+    rgb1_l.valRGB_BLUE = b1.val0;
+    /* Store RGB pixel data to memory. */
+    switch (cols_remaining) {
+    case 15:
+      vst3_lane_u8(outptr0 + 14 * RGB_PIXELSIZE, rgb0_h, 6);
+      vst3_lane_u8(outptr1 + 14 * RGB_PIXELSIZE, rgb1_h, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 14:
+      vst3_lane_u8(outptr0 + 13 * RGB_PIXELSIZE, rgb0_h, 5);
+      vst3_lane_u8(outptr1 + 13 * RGB_PIXELSIZE, rgb1_h, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 13:
+      vst3_lane_u8(outptr0 + 12 * RGB_PIXELSIZE, rgb0_h, 4);
+      vst3_lane_u8(outptr1 + 12 * RGB_PIXELSIZE, rgb1_h, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 12:
+      vst3_lane_u8(outptr0 + 11 * RGB_PIXELSIZE, rgb0_h, 3);
+      vst3_lane_u8(outptr1 + 11 * RGB_PIXELSIZE, rgb1_h, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 11:
+      vst3_lane_u8(outptr0 + 10 * RGB_PIXELSIZE, rgb0_h, 2);
+      vst3_lane_u8(outptr1 + 10 * RGB_PIXELSIZE, rgb1_h, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 10:
+      vst3_lane_u8(outptr0 + 9 * RGB_PIXELSIZE, rgb0_h, 1);
+      vst3_lane_u8(outptr1 + 9 * RGB_PIXELSIZE, rgb1_h, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 9:
+      vst3_lane_u8(outptr0 + 8 * RGB_PIXELSIZE, rgb0_h, 0);
+      vst3_lane_u8(outptr1 + 8 * RGB_PIXELSIZE, rgb1_h, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 8:
+      vst3_u8(outptr0, rgb0_l);
+      vst3_u8(outptr1, rgb1_l);
+      break;
+    case 7:
+      vst3_lane_u8(outptr0 + 6 * RGB_PIXELSIZE, rgb0_l, 6);
+      vst3_lane_u8(outptr1 + 6 * RGB_PIXELSIZE, rgb1_l, 6);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 6:
+      vst3_lane_u8(outptr0 + 5 * RGB_PIXELSIZE, rgb0_l, 5);
+      vst3_lane_u8(outptr1 + 5 * RGB_PIXELSIZE, rgb1_l, 5);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 5:
+      vst3_lane_u8(outptr0 + 4 * RGB_PIXELSIZE, rgb0_l, 4);
+      vst3_lane_u8(outptr1 + 4 * RGB_PIXELSIZE, rgb1_l, 4);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 4:
+      vst3_lane_u8(outptr0 + 3 * RGB_PIXELSIZE, rgb0_l, 3);
+      vst3_lane_u8(outptr1 + 3 * RGB_PIXELSIZE, rgb1_l, 3);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 3:
+      vst3_lane_u8(outptr0 + 2 * RGB_PIXELSIZE, rgb0_l, 2);
+      vst3_lane_u8(outptr1 + 2 * RGB_PIXELSIZE, rgb1_l, 2);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 2:
+      vst3_lane_u8(outptr0 + 1 * RGB_PIXELSIZE, rgb0_l, 1);
+      vst3_lane_u8(outptr1 + 1 * RGB_PIXELSIZE, rgb1_l, 1);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    case 1:
+      vst3_lane_u8(outptr0, rgb0_l, 0);
+      vst3_lane_u8(outptr1, rgb1_l, 0);
+      FALLTHROUGH               /*FALLTHROUGH*/
+    default:
+      break;
+    }
+#endif
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jdsample-neon.c Added

@@ -0,0 +1,569 @@
+/*
+ * jdsample-neon.c - upsampling (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+
+#include <arm_neon.h>
+
+
+/* The diagram below shows a row of samples produced by h2v1 downsampling.
+ *
+ *                s0        s1        s2
+ *            +---------+---------+---------+
+ *            |         |         |         |
+ *            | p0   p1 | p2   p3 | p4   p5 |
+ *            |         |         |         |
+ *            +---------+---------+---------+
+ *
+ * Samples s0-s2 were created by averaging the original pixel component values
+ * centered at positions p0-p5 above.  To approximate those original pixel
+ * component values, we proportionally blend the adjacent samples in each row.
+ *
+ * An upsampled pixel component value is computed by blending the sample
+ * containing the pixel center with the nearest neighboring sample, in the
+ * ratio 3:1.  For example:
+ *     p1(upsampled) = 3/4 * s0 + 1/4 * s1
+ *     p2(upsampled) = 3/4 * s1 + 1/4 * s0
+ * When computing the first and last pixel component values in the row, there
+ * is no adjacent sample to blend, so:
+ *     p0(upsampled) = s0
+ *     p5(upsampled) = s2
+ */
+
+void jsimd_h2v1_fancy_upsample_neon(int max_v_samp_factor,
+                                    JDIMENSION downsampled_width,
+                                    JSAMPARRAY input_data,
+                                    JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr;
+  int inrow;
+  unsigned colctr;
+  /* Set up constants. */
+  const uint16x8_t one_u16 = vdupq_n_u16(1);
+  const uint8x8_t three_u8 = vdup_n_u8(3);
+
+  for (inrow = 0; inrow < max_v_samp_factor; inrow++) {
+    inptr = input_datainrow;
+    outptr = output_datainrow;
+    /* First pixel component value in this row of the original image */
+    *outptr = (JSAMPLE)GETJSAMPLE(*inptr);
+
+    /*    3/4 * containing sample + 1/4 * nearest neighboring sample
+     * For p1: containing sample = s0, nearest neighboring sample = s1
+     * For p2: containing sample = s1, nearest neighboring sample = s0
+     */
+    uint8x16_t s0 = vld1q_u8(inptr);
+    uint8x16_t s1 = vld1q_u8(inptr + 1);
+    /* Multiplication makes vectors twice as wide.  '_l' and '_h' suffixes
+     * denote low half and high half respectively.
+     */
+    uint16x8_t s1_add_3s0_l =
+      vmlal_u8(vmovl_u8(vget_low_u8(s1)), vget_low_u8(s0), three_u8);
+    uint16x8_t s1_add_3s0_h =
+      vmlal_u8(vmovl_u8(vget_high_u8(s1)), vget_high_u8(s0), three_u8);
+    uint16x8_t s0_add_3s1_l =
+      vmlal_u8(vmovl_u8(vget_low_u8(s0)), vget_low_u8(s1), three_u8);
+    uint16x8_t s0_add_3s1_h =
+      vmlal_u8(vmovl_u8(vget_high_u8(s0)), vget_high_u8(s1), three_u8);
+    /* Add ordered dithering bias to odd pixel values. */
+    s0_add_3s1_l = vaddq_u16(s0_add_3s1_l, one_u16);
+    s0_add_3s1_h = vaddq_u16(s0_add_3s1_h, one_u16);
+
+    /* The offset is initially 1, because the first pixel component has already
+     * been stored.  However, in subsequent iterations of the SIMD loop, this
+     * offset is (2 * colctr - 1) to stay within the bounds of the sample
+     * buffers without having to resort to a slow scalar tail case for the last
+     * (downsampled_width % 16) samples.  See "Creation of 2-D sample arrays"
+     * in jmemmgr.c for more details.
+     */
+    unsigned outptr_offset = 1;
+    uint8x16x2_t output_pixels;
+
+    /* We use software pipelining to maximise performance.  The code indented
+     * an extra two spaces begins the next iteration of the loop.
+     */
+    for (colctr = 16; colctr < downsampled_width; colctr += 16) {
+
+        s0 = vld1q_u8(inptr + colctr - 1);
+        s1 = vld1q_u8(inptr + colctr);
+
+      /* Right-shift by 2 (divide by 4), narrow to 8-bit, and combine. */
+      output_pixels.val0 = vcombine_u8(vrshrn_n_u16(s1_add_3s0_l, 2),
+                                         vrshrn_n_u16(s1_add_3s0_h, 2));
+      output_pixels.val1 = vcombine_u8(vshrn_n_u16(s0_add_3s1_l, 2),
+                                         vshrn_n_u16(s0_add_3s1_h, 2));
+
+        /* Multiplication makes vectors twice as wide.  '_l' and '_h' suffixes
+         * denote low half and high half respectively.
+         */
+        s1_add_3s0_l =
+          vmlal_u8(vmovl_u8(vget_low_u8(s1)), vget_low_u8(s0), three_u8);
+        s1_add_3s0_h =
+          vmlal_u8(vmovl_u8(vget_high_u8(s1)), vget_high_u8(s0), three_u8);
+        s0_add_3s1_l =
+          vmlal_u8(vmovl_u8(vget_low_u8(s0)), vget_low_u8(s1), three_u8);
+        s0_add_3s1_h =
+          vmlal_u8(vmovl_u8(vget_high_u8(s0)), vget_high_u8(s1), three_u8);
+        /* Add ordered dithering bias to odd pixel values. */
+        s0_add_3s1_l = vaddq_u16(s0_add_3s1_l, one_u16);
+        s0_add_3s1_h = vaddq_u16(s0_add_3s1_h, one_u16);
+
+      /* Store pixel component values to memory. */
+      vst2q_u8(outptr + outptr_offset, output_pixels);
+      outptr_offset = 2 * colctr - 1;
+    }
+
+    /* Complete the last iteration of the loop. */
+
+    /* Right-shift by 2 (divide by 4), narrow to 8-bit, and combine. */
+    output_pixels.val0 = vcombine_u8(vrshrn_n_u16(s1_add_3s0_l, 2),
+                                       vrshrn_n_u16(s1_add_3s0_h, 2));
+    output_pixels.val1 = vcombine_u8(vshrn_n_u16(s0_add_3s1_l, 2),
+                                       vshrn_n_u16(s0_add_3s1_h, 2));
+    /* Store pixel component values to memory. */
+    vst2q_u8(outptr + outptr_offset, output_pixels);
+
+    /* Last pixel component value in this row of the original image */
+    outptr2 * downsampled_width - 1 =
+      GETJSAMPLE(inptrdownsampled_width - 1);
+  }
+}
+
+
+/* The diagram below shows an array of samples produced by h2v2 downsampling.
+ *
+ *                s0        s1        s2
+ *            +---------+---------+---------+
+ *            | p0   p1 | p2   p3 | p4   p5 |
+ *       sA   |         |         |         |
+ *            | p6   p7 | p8   p9 | p10  p11|
+ *            +---------+---------+---------+
+ *            | p12  p13| p14  p15| p16  p17|
+ *       sB   |         |         |         |
+ *            | p18  p19| p20  p21| p22  p23|
+ *            +---------+---------+---------+
+ *            | p24  p25| p26  p27| p28  p29|
+ *       sC   |         |         |         |
+ *            | p30  p31| p32  p33| p34  p35|
+ *            +---------+---------+---------+
+ *
+ * Samples s0A-s2C were created by averaging the original pixel component
+ * values centered at positions p0-p35 above.  To approximate one of those
+ * original pixel component values, we proportionally blend the sample
+ * containing the pixel center with the nearest neighboring samples in each
+ * row, column, and diagonal.
+ *
+ * An upsampled pixel component value is computed by first blending the sample
+ * containing the pixel center with the nearest neighboring samples in the
+ * same column, in the ratio 3:1, and then blending each column sum with the
+ * nearest neighboring column sum, in the ratio 3:1.  For example:
+ *     p14(upsampled) = 3/4 * (3/4 * s1B + 1/4 * s1A) +
+ *                      1/4 * (3/4 * s0B + 1/4 * s0A)
+ *                    = 9/16 * s1B + 3/16 * s1A + 3/16 * s0B + 1/16 * s0A
+ * When computing the first and last pixel component values in the row, there
+ * is no horizontally adjacent sample to blend, so:
+ *     p12(upsampled) = 3/4 * s0B + 1/4 * s0A
+ *     p23(upsampled) = 3/4 * s2B + 1/4 * s2C
+ * When computing the first and last pixel component values in the column,
+ * there is no vertically adjacent sample to blend, so:
+ *     p2(upsampled) = 3/4 * s1A + 1/4 * s0A
+ *     p33(upsampled) = 3/4 * s1C + 1/4 * s2C
+ * When computing the corner pixel component values, there is no adjacent
+ * sample to blend, so:
+ *     p0(upsampled) = s0A
+ *     p35(upsampled) = s2C
+ */
+
+void jsimd_h2v2_fancy_upsample_neon(int max_v_samp_factor,
+                                    JDIMENSION downsampled_width,
+                                    JSAMPARRAY input_data,
+                                    JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr0, inptr1, inptr2, outptr0, outptr1;
+  int inrow, outrow;
+  unsigned colctr;
+  /* Set up constants. */
+  const uint16x8_t seven_u16 = vdupq_n_u16(7);
+  const uint8x8_t three_u8 = vdup_n_u8(3);
+  const uint16x8_t three_u16 = vdupq_n_u16(3);
+
+  inrow = outrow = 0;
+  while (outrow < max_v_samp_factor) {
+    inptr0 = input_datainrow - 1;
+    inptr1 = input_datainrow;
+    inptr2 = input_datainrow + 1;
+    /* Suffixes 0 and 1 denote the upper and lower rows of output pixels,
+     * respectively.
+     */
+    outptr0 = output_dataoutrow++;
+    outptr1 = output_dataoutrow++;
+
+    /* First pixel component value in this row of the original image */
+    int s0colsum0 = GETJSAMPLE(*inptr1) * 3 + GETJSAMPLE(*inptr0);
+    *outptr0 = (JSAMPLE)((s0colsum0 * 4 + 8) >> 4);
+    int s0colsum1 = GETJSAMPLE(*inptr1) * 3 + GETJSAMPLE(*inptr2);
+    *outptr1 = (JSAMPLE)((s0colsum1 * 4 + 8) >> 4);
+
+    /* Step 1: Blend samples vertically in columns s0 and s1.
+     * Leave the divide by 4 until the end, when it can be done for both
+     * dimensions at once, right-shifting by 4.
+     */
+
+    /* Load and compute s0colsum0 and s0colsum1. */
+    uint8x16_t s0A = vld1q_u8(inptr0);
+    uint8x16_t s0B = vld1q_u8(inptr1);
+    uint8x16_t s0C = vld1q_u8(inptr2);
+    /* Multiplication makes vectors twice as wide.  '_l' and '_h' suffixes
+     * denote low half and high half respectively.
+     */
+    uint16x8_t s0colsum0_l = vmlal_u8(vmovl_u8(vget_low_u8(s0A)),
+                                      vget_low_u8(s0B), three_u8);
+    uint16x8_t s0colsum0_h = vmlal_u8(vmovl_u8(vget_high_u8(s0A)),
+                                      vget_high_u8(s0B), three_u8);
+    uint16x8_t s0colsum1_l = vmlal_u8(vmovl_u8(vget_low_u8(s0C)),
+                                      vget_low_u8(s0B), three_u8);
+    uint16x8_t s0colsum1_h = vmlal_u8(vmovl_u8(vget_high_u8(s0C)),
+                                      vget_high_u8(s0B), three_u8);
+    /* Load and compute s1colsum0 and s1colsum1. */
+    uint8x16_t s1A = vld1q_u8(inptr0 + 1);
+    uint8x16_t s1B = vld1q_u8(inptr1 + 1);
+    uint8x16_t s1C = vld1q_u8(inptr2 + 1);
+    uint16x8_t s1colsum0_l = vmlal_u8(vmovl_u8(vget_low_u8(s1A)),
+                                      vget_low_u8(s1B), three_u8);
+    uint16x8_t s1colsum0_h = vmlal_u8(vmovl_u8(vget_high_u8(s1A)),
+                                      vget_high_u8(s1B), three_u8);
+    uint16x8_t s1colsum1_l = vmlal_u8(vmovl_u8(vget_low_u8(s1C)),
+                                      vget_low_u8(s1B), three_u8);
+    uint16x8_t s1colsum1_h = vmlal_u8(vmovl_u8(vget_high_u8(s1C)),
+                                      vget_high_u8(s1B), three_u8);
+
+    /* Step 2: Blend the already-blended columns. */
+
+    uint16x8_t output0_p1_l = vmlaq_u16(s1colsum0_l, s0colsum0_l, three_u16);
+    uint16x8_t output0_p1_h = vmlaq_u16(s1colsum0_h, s0colsum0_h, three_u16);
+    uint16x8_t output0_p2_l = vmlaq_u16(s0colsum0_l, s1colsum0_l, three_u16);
+    uint16x8_t output0_p2_h = vmlaq_u16(s0colsum0_h, s1colsum0_h, three_u16);
+    uint16x8_t output1_p1_l = vmlaq_u16(s1colsum1_l, s0colsum1_l, three_u16);
+    uint16x8_t output1_p1_h = vmlaq_u16(s1colsum1_h, s0colsum1_h, three_u16);
+    uint16x8_t output1_p2_l = vmlaq_u16(s0colsum1_l, s1colsum1_l, three_u16);
+    uint16x8_t output1_p2_h = vmlaq_u16(s0colsum1_h, s1colsum1_h, three_u16);
+    /* Add ordered dithering bias to odd pixel values. */
+    output0_p1_l = vaddq_u16(output0_p1_l, seven_u16);
+    output0_p1_h = vaddq_u16(output0_p1_h, seven_u16);
+    output1_p1_l = vaddq_u16(output1_p1_l, seven_u16);
+    output1_p1_h = vaddq_u16(output1_p1_h, seven_u16);
+    /* Right-shift by 4 (divide by 16), narrow to 8-bit, and combine. */
+    uint8x16x2_t output_pixels0 = { {
+      vcombine_u8(vshrn_n_u16(output0_p1_l, 4), vshrn_n_u16(output0_p1_h, 4)),
+      vcombine_u8(vrshrn_n_u16(output0_p2_l, 4), vrshrn_n_u16(output0_p2_h, 4))
+    } };
+    uint8x16x2_t output_pixels1 = { {
+      vcombine_u8(vshrn_n_u16(output1_p1_l, 4), vshrn_n_u16(output1_p1_h, 4)),
+      vcombine_u8(vrshrn_n_u16(output1_p2_l, 4), vrshrn_n_u16(output1_p2_h, 4))
+    } };
+
+    /* Store pixel component values to memory.
+     * The minimum size of the output buffer for each row is 64 bytes => no
+     * need to worry about buffer overflow here.  See "Creation of 2-D sample
+     * arrays" in jmemmgr.c for more details.
+     */
+    vst2q_u8(outptr0 + 1, output_pixels0);
+    vst2q_u8(outptr1 + 1, output_pixels1);
+
+    /* The first pixel of the image shifted our loads and stores by one byte.
+     * We have to re-align on a 32-byte boundary at some point before the end
+     * of the row (we do it now on the 32/33 pixel boundary) to stay within the
+     * bounds of the sample buffers without having to resort to a slow scalar
+     * tail case for the last (downsampled_width % 16) samples.  See "Creation
+     * of 2-D sample arrays" in jmemmgr.c for more details.
+     */
+    for (colctr = 16; colctr < downsampled_width; colctr += 16) {
+      /* Step 1: Blend samples vertically in columns s0 and s1. */
+
+      /* Load and compute s0colsum0 and s0colsum1. */
+      s0A = vld1q_u8(inptr0 + colctr - 1);
+      s0B = vld1q_u8(inptr1 + colctr - 1);
+      s0C = vld1q_u8(inptr2 + colctr - 1);
+      s0colsum0_l = vmlal_u8(vmovl_u8(vget_low_u8(s0A)), vget_low_u8(s0B),
+                             three_u8);
+      s0colsum0_h = vmlal_u8(vmovl_u8(vget_high_u8(s0A)), vget_high_u8(s0B),
+                             three_u8);
+      s0colsum1_l = vmlal_u8(vmovl_u8(vget_low_u8(s0C)), vget_low_u8(s0B),
+                             three_u8);
+      s0colsum1_h = vmlal_u8(vmovl_u8(vget_high_u8(s0C)), vget_high_u8(s0B),
+                             three_u8);
+      /* Load and compute s1colsum0 and s1colsum1. */
+      s1A = vld1q_u8(inptr0 + colctr);
+      s1B = vld1q_u8(inptr1 + colctr);
+      s1C = vld1q_u8(inptr2 + colctr);
+      s1colsum0_l = vmlal_u8(vmovl_u8(vget_low_u8(s1A)), vget_low_u8(s1B),
+                             three_u8);
+      s1colsum0_h = vmlal_u8(vmovl_u8(vget_high_u8(s1A)), vget_high_u8(s1B),
+                             three_u8);
+      s1colsum1_l = vmlal_u8(vmovl_u8(vget_low_u8(s1C)), vget_low_u8(s1B),
+                             three_u8);
+      s1colsum1_h = vmlal_u8(vmovl_u8(vget_high_u8(s1C)), vget_high_u8(s1B),
+                             three_u8);
+
+      /* Step 2: Blend the already-blended columns. */
+
+      output0_p1_l = vmlaq_u16(s1colsum0_l, s0colsum0_l, three_u16);
+      output0_p1_h = vmlaq_u16(s1colsum0_h, s0colsum0_h, three_u16);
+      output0_p2_l = vmlaq_u16(s0colsum0_l, s1colsum0_l, three_u16);
+      output0_p2_h = vmlaq_u16(s0colsum0_h, s1colsum0_h, three_u16);
+      output1_p1_l = vmlaq_u16(s1colsum1_l, s0colsum1_l, three_u16);
+      output1_p1_h = vmlaq_u16(s1colsum1_h, s0colsum1_h, three_u16);
+      output1_p2_l = vmlaq_u16(s0colsum1_l, s1colsum1_l, three_u16);
+      output1_p2_h = vmlaq_u16(s0colsum1_h, s1colsum1_h, three_u16);
+      /* Add ordered dithering bias to odd pixel values. */
+      output0_p1_l = vaddq_u16(output0_p1_l, seven_u16);
+      output0_p1_h = vaddq_u16(output0_p1_h, seven_u16);
+      output1_p1_l = vaddq_u16(output1_p1_l, seven_u16);
+      output1_p1_h = vaddq_u16(output1_p1_h, seven_u16);
+      /* Right-shift by 4 (divide by 16), narrow to 8-bit, and combine. */
+      output_pixels0.val0 = vcombine_u8(vshrn_n_u16(output0_p1_l, 4),
+                                          vshrn_n_u16(output0_p1_h, 4));
+      output_pixels0.val1 = vcombine_u8(vrshrn_n_u16(output0_p2_l, 4),
+                                          vrshrn_n_u16(output0_p2_h, 4));
+      output_pixels1.val0 = vcombine_u8(vshrn_n_u16(output1_p1_l, 4),
+                                          vshrn_n_u16(output1_p1_h, 4));
+      output_pixels1.val1 = vcombine_u8(vrshrn_n_u16(output1_p2_l, 4),
+                                          vrshrn_n_u16(output1_p2_h, 4));
+      /* Store pixel component values to memory. */
+      vst2q_u8(outptr0 + 2 * colctr - 1, output_pixels0);
+      vst2q_u8(outptr1 + 2 * colctr - 1, output_pixels1);
+    }
+
+    /* Last pixel component value in this row of the original image */
+    int s1colsum0 = GETJSAMPLE(inptr1downsampled_width - 1) * 3 +
+                    GETJSAMPLE(inptr0downsampled_width - 1);
+    outptr02 * downsampled_width - 1 = (JSAMPLE)((s1colsum0 * 4 + 7) >> 4);
+    int s1colsum1 = GETJSAMPLE(inptr1downsampled_width - 1) * 3 +
+                    GETJSAMPLE(inptr2downsampled_width - 1);
+    outptr12 * downsampled_width - 1 = (JSAMPLE)((s1colsum1 * 4 + 7) >> 4);
+    inrow++;
+  }
+}
+
+
+/* The diagram below shows a column of samples produced by h1v2 downsampling
+ * (or by losslessly rotating or transposing an h2v1-downsampled image.)
+ *
+ *            +---------+
+ *            |   p0    |
+ *     sA     |         |
+ *            |   p1    |
+ *            +---------+
+ *            |   p2    |
+ *     sB     |         |
+ *            |   p3    |
+ *            +---------+
+ *            |   p4    |
+ *     sC     |         |
+ *            |   p5    |
+ *            +---------+
+ *
+ * Samples sA-sC were created by averaging the original pixel component values
+ * centered at positions p0-p5 above.  To approximate those original pixel
+ * component values, we proportionally blend the adjacent samples in each
+ * column.
+ *
+ * An upsampled pixel component value is computed by blending the sample
+ * containing the pixel center with the nearest neighboring sample, in the
+ * ratio 3:1.  For example:
+ *     p1(upsampled) = 3/4 * sA + 1/4 * sB
+ *     p2(upsampled) = 3/4 * sB + 1/4 * sA
+ * When computing the first and last pixel component values in the column,
+ * there is no adjacent sample to blend, so:
+ *     p0(upsampled) = sA
+ *     p5(upsampled) = sC
+ */
+
+void jsimd_h1v2_fancy_upsample_neon(int max_v_samp_factor,
+                                    JDIMENSION downsampled_width,
+                                    JSAMPARRAY input_data,
+                                    JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr0, inptr1, inptr2, outptr0, outptr1;
+  int inrow, outrow;
+  unsigned colctr;
+  /* Set up constants. */
+  const uint16x8_t one_u16 = vdupq_n_u16(1);
+  const uint8x8_t three_u8 = vdup_n_u8(3);
+
+  inrow = outrow = 0;
+  while (outrow < max_v_samp_factor) {
+    inptr0 = input_datainrow - 1;
+    inptr1 = input_datainrow;
+    inptr2 = input_datainrow + 1;
+    /* Suffixes 0 and 1 denote the upper and lower rows of output pixels,
+     * respectively.
+     */
+    outptr0 = output_dataoutrow++;
+    outptr1 = output_dataoutrow++;
+    inrow++;
+
+    /* The size of the input and output buffers is always a multiple of 32
+     * bytes => no need to worry about buffer overflow when reading/writing
+     * memory.  See "Creation of 2-D sample arrays" in jmemmgr.c for more
+     * details.
+     */
+    for (colctr = 0; colctr < downsampled_width; colctr += 16) {
+      /* Load samples. */
+      uint8x16_t sA = vld1q_u8(inptr0 + colctr);
+      uint8x16_t sB = vld1q_u8(inptr1 + colctr);
+      uint8x16_t sC = vld1q_u8(inptr2 + colctr);
+      /* Blend samples vertically. */
+      uint16x8_t colsum0_l = vmlal_u8(vmovl_u8(vget_low_u8(sA)),
+                                      vget_low_u8(sB), three_u8);
+      uint16x8_t colsum0_h = vmlal_u8(vmovl_u8(vget_high_u8(sA)),
+                                      vget_high_u8(sB), three_u8);
+      uint16x8_t colsum1_l = vmlal_u8(vmovl_u8(vget_low_u8(sC)),
+                                      vget_low_u8(sB), three_u8);
+      uint16x8_t colsum1_h = vmlal_u8(vmovl_u8(vget_high_u8(sC)),
+                                      vget_high_u8(sB), three_u8);
+      /* Add ordered dithering bias to pixel values in even output rows. */
+      colsum0_l = vaddq_u16(colsum0_l, one_u16);
+      colsum0_h = vaddq_u16(colsum0_h, one_u16);
+      /* Right-shift by 2 (divide by 4), narrow to 8-bit, and combine. */
+      uint8x16_t output_pixels0 = vcombine_u8(vshrn_n_u16(colsum0_l, 2),
+                                              vshrn_n_u16(colsum0_h, 2));
+      uint8x16_t output_pixels1 = vcombine_u8(vrshrn_n_u16(colsum1_l, 2),
+                                              vrshrn_n_u16(colsum1_h, 2));
+      /* Store pixel component values to memory. */
+      vst1q_u8(outptr0 + colctr, output_pixels0);
+      vst1q_u8(outptr1 + colctr, output_pixels1);
+    }
+  }
+}
+
+
+/* The diagram below shows a row of samples produced by h2v1 downsampling.
+ *
+ *                s0        s1
+ *            +---------+---------+
+ *            |         |         |
+ *            | p0   p1 | p2   p3 |
+ *            |         |         |
+ *            +---------+---------+
+ *
+ * Samples s0 and s1 were created by averaging the original pixel component
+ * values centered at positions p0-p3 above.  To approximate those original
+ * pixel component values, we duplicate the samples horizontally:
+ *     p0(upsampled) = p1(upsampled) = s0
+ *     p2(upsampled) = p3(upsampled) = s1
+ */
+
+void jsimd_h2v1_upsample_neon(int max_v_samp_factor, JDIMENSION output_width,
+                              JSAMPARRAY input_data,
+                              JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr;
+  int inrow;
+  unsigned colctr;
+
+  for (inrow = 0; inrow < max_v_samp_factor; inrow++) {
+    inptr = input_datainrow;
+    outptr = output_datainrow;
+    for (colctr = 0; 2 * colctr < output_width; colctr += 16) {
+      uint8x16_t samples = vld1q_u8(inptr + colctr);
+      /* Duplicate the samples.  The store operation below interleaves them so
+       * that adjacent pixel component values take on the same sample value,
+       * per above.
+       */
+      uint8x16x2_t output_pixels = { { samples, samples } };
+      /* Store pixel component values to memory.
+       * Due to the way sample buffers are allocated, we don't need to worry
+       * about tail cases when output_width is not a multiple of 32.  See
+       * "Creation of 2-D sample arrays" in jmemmgr.c for details.
+       */
+      vst2q_u8(outptr + 2 * colctr, output_pixels);
+    }
+  }
+}
+
+
+/* The diagram below shows an array of samples produced by h2v2 downsampling.
+ *
+ *                s0        s1
+ *            +---------+---------+
+ *            | p0   p1 | p2   p3 |
+ *       sA   |         |         |
+ *            | p4   p5 | p6   p7 |
+ *            +---------+---------+
+ *            | p8   p9 | p10  p11|
+ *       sB   |         |         |
+ *            | p12  p13| p14  p15|
+ *            +---------+---------+
+ *
+ * Samples s0A-s1B were created by averaging the original pixel component
+ * values centered at positions p0-p15 above.  To approximate those original
+ * pixel component values, we duplicate the samples both horizontally and
+ * vertically:
+ *     p0(upsampled) = p1(upsampled) = p4(upsampled) = p5(upsampled) = s0A
+ *     p2(upsampled) = p3(upsampled) = p6(upsampled) = p7(upsampled) = s1A
+ *     p8(upsampled) = p9(upsampled) = p12(upsampled) = p13(upsampled) = s0B
+ *     p10(upsampled) = p11(upsampled) = p14(upsampled) = p15(upsampled) = s1B
+ */
+
+void jsimd_h2v2_upsample_neon(int max_v_samp_factor, JDIMENSION output_width,
+                              JSAMPARRAY input_data,
+                              JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr0, outptr1;
+  int inrow, outrow;
+  unsigned colctr;
+
+  for (inrow = 0, outrow = 0; outrow < max_v_samp_factor; inrow++) {
+    inptr = input_datainrow;
+    outptr0 = output_dataoutrow++;
+    outptr1 = output_dataoutrow++;
+
+    for (colctr = 0; 2 * colctr < output_width; colctr += 16) {
+      uint8x16_t samples = vld1q_u8(inptr + colctr);
+      /* Duplicate the samples.  The store operation below interleaves them so
+       * that adjacent pixel component values take on the same sample value,
+       * per above.
+       */
+      uint8x16x2_t output_pixels = { { samples, samples } };
+      /* Store pixel component values for both output rows to memory.
+       * Due to the way sample buffers are allocated, we don't need to worry
+       * about tail cases when output_width is not a multiple of 32.  See
+       * "Creation of 2-D sample arrays" in jmemmgr.c for details.
+       */
+      vst2q_u8(outptr0 + 2 * colctr, output_pixels);
+      vst2q_u8(outptr1 + 2 * colctr, output_pixels);
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jfdctfst-neon.c Added

@@ -0,0 +1,214 @@
+/*
+ * jfdctfst-neon.c - fast integer FDCT (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+
+#include <arm_neon.h>
+
+
+/* jsimd_fdct_ifast_neon() performs a fast, not so accurate forward DCT
+ * (Discrete Cosine Transform) on one block of samples.  It uses the same
+ * calculations and produces exactly the same output as IJG's original
+ * jpeg_fdct_ifast() function, which can be found in jfdctfst.c.
+ *
+ * Scaled integer constants are used to avoid floating-point arithmetic:
+ *    0.382683433 = 12544 * 2^-15
+ *    0.541196100 = 17795 * 2^-15
+ *    0.707106781 = 23168 * 2^-15
+ *    0.306562965 =  9984 * 2^-15
+ *
+ * See jfdctfst.c for further details of the DCT algorithm.  Where possible,
+ * the variable names and comments here in jsimd_fdct_ifast_neon() match up
+ * with those in jpeg_fdct_ifast().
+ */
+
+#define F_0_382  12544
+#define F_0_541  17792
+#define F_0_707  23168
+#define F_0_306  9984
+
+
+ALIGN(16) static const int16_t jsimd_fdct_ifast_neon_consts = {
+  F_0_382, F_0_541, F_0_707, F_0_306
+};
+
+void jsimd_fdct_ifast_neon(DCTELEM *data)
+{
+  /* Load an 8x8 block of samples into Neon registers.  De-interleaving loads
+   * are used, followed by vuzp to transpose the block such that we have a
+   * column of samples per vector - allowing all rows to be processed at once.
+   */
+  int16x8x4_t data1 = vld4q_s16(data);
+  int16x8x4_t data2 = vld4q_s16(data + 4 * DCTSIZE);
+
+  int16x8x2_t cols_04 = vuzpq_s16(data1.val0, data2.val0);
+  int16x8x2_t cols_15 = vuzpq_s16(data1.val1, data2.val1);
+  int16x8x2_t cols_26 = vuzpq_s16(data1.val2, data2.val2);
+  int16x8x2_t cols_37 = vuzpq_s16(data1.val3, data2.val3);
+
+  int16x8_t col0 = cols_04.val0;
+  int16x8_t col1 = cols_15.val0;
+  int16x8_t col2 = cols_26.val0;
+  int16x8_t col3 = cols_37.val0;
+  int16x8_t col4 = cols_04.val1;
+  int16x8_t col5 = cols_15.val1;
+  int16x8_t col6 = cols_26.val1;
+  int16x8_t col7 = cols_37.val1;
+
+  /* Pass 1: process rows. */
+
+  /* Load DCT conversion constants. */
+  const int16x4_t consts = vld1_s16(jsimd_fdct_ifast_neon_consts);
+
+  int16x8_t tmp0 = vaddq_s16(col0, col7);
+  int16x8_t tmp7 = vsubq_s16(col0, col7);
+  int16x8_t tmp1 = vaddq_s16(col1, col6);
+  int16x8_t tmp6 = vsubq_s16(col1, col6);
+  int16x8_t tmp2 = vaddq_s16(col2, col5);
+  int16x8_t tmp5 = vsubq_s16(col2, col5);
+  int16x8_t tmp3 = vaddq_s16(col3, col4);
+  int16x8_t tmp4 = vsubq_s16(col3, col4);
+
+  /* Even part */
+  int16x8_t tmp10 = vaddq_s16(tmp0, tmp3);    /* phase 2 */
+  int16x8_t tmp13 = vsubq_s16(tmp0, tmp3);
+  int16x8_t tmp11 = vaddq_s16(tmp1, tmp2);
+  int16x8_t tmp12 = vsubq_s16(tmp1, tmp2);
+
+  col0 = vaddq_s16(tmp10, tmp11);             /* phase 3 */
+  col4 = vsubq_s16(tmp10, tmp11);
+
+  int16x8_t z1 = vqdmulhq_lane_s16(vaddq_s16(tmp12, tmp13), consts, 2);
+  col2 = vaddq_s16(tmp13, z1);                /* phase 5 */
+  col6 = vsubq_s16(tmp13, z1);
+
+  /* Odd part */
+  tmp10 = vaddq_s16(tmp4, tmp5);              /* phase 2 */
+  tmp11 = vaddq_s16(tmp5, tmp6);
+  tmp12 = vaddq_s16(tmp6, tmp7);
+
+  int16x8_t z5 = vqdmulhq_lane_s16(vsubq_s16(tmp10, tmp12), consts, 0);
+  int16x8_t z2 = vqdmulhq_lane_s16(tmp10, consts, 1);
+  z2 = vaddq_s16(z2, z5);
+  int16x8_t z4 = vqdmulhq_lane_s16(tmp12, consts, 3);
+  z5 = vaddq_s16(tmp12, z5);
+  z4 = vaddq_s16(z4, z5);
+  int16x8_t z3 = vqdmulhq_lane_s16(tmp11, consts, 2);
+
+  int16x8_t z11 = vaddq_s16(tmp7, z3);        /* phase 5 */
+  int16x8_t z13 = vsubq_s16(tmp7, z3);
+
+  col5 = vaddq_s16(z13, z2);                  /* phase 6 */
+  col3 = vsubq_s16(z13, z2);
+  col1 = vaddq_s16(z11, z4);
+  col7 = vsubq_s16(z11, z4);
+
+  /* Transpose to work on columns in pass 2. */
+  int16x8x2_t cols_01 = vtrnq_s16(col0, col1);
+  int16x8x2_t cols_23 = vtrnq_s16(col2, col3);
+  int16x8x2_t cols_45 = vtrnq_s16(col4, col5);
+  int16x8x2_t cols_67 = vtrnq_s16(col6, col7);
+
+  int32x4x2_t cols_0145_l = vtrnq_s32(vreinterpretq_s32_s16(cols_01.val0),
+                                      vreinterpretq_s32_s16(cols_45.val0));
+  int32x4x2_t cols_0145_h = vtrnq_s32(vreinterpretq_s32_s16(cols_01.val1),
+                                      vreinterpretq_s32_s16(cols_45.val1));
+  int32x4x2_t cols_2367_l = vtrnq_s32(vreinterpretq_s32_s16(cols_23.val0),
+                                      vreinterpretq_s32_s16(cols_67.val0));
+  int32x4x2_t cols_2367_h = vtrnq_s32(vreinterpretq_s32_s16(cols_23.val1),
+                                      vreinterpretq_s32_s16(cols_67.val1));
+
+  int32x4x2_t rows_04 = vzipq_s32(cols_0145_l.val0, cols_2367_l.val0);
+  int32x4x2_t rows_15 = vzipq_s32(cols_0145_h.val0, cols_2367_h.val0);
+  int32x4x2_t rows_26 = vzipq_s32(cols_0145_l.val1, cols_2367_l.val1);
+  int32x4x2_t rows_37 = vzipq_s32(cols_0145_h.val1, cols_2367_h.val1);
+
+  int16x8_t row0 = vreinterpretq_s16_s32(rows_04.val0);
+  int16x8_t row1 = vreinterpretq_s16_s32(rows_15.val0);
+  int16x8_t row2 = vreinterpretq_s16_s32(rows_26.val0);
+  int16x8_t row3 = vreinterpretq_s16_s32(rows_37.val0);
+  int16x8_t row4 = vreinterpretq_s16_s32(rows_04.val1);
+  int16x8_t row5 = vreinterpretq_s16_s32(rows_15.val1);
+  int16x8_t row6 = vreinterpretq_s16_s32(rows_26.val1);
+  int16x8_t row7 = vreinterpretq_s16_s32(rows_37.val1);
+
+  /* Pass 2: process columns. */
+
+  tmp0 = vaddq_s16(row0, row7);
+  tmp7 = vsubq_s16(row0, row7);
+  tmp1 = vaddq_s16(row1, row6);
+  tmp6 = vsubq_s16(row1, row6);
+  tmp2 = vaddq_s16(row2, row5);
+  tmp5 = vsubq_s16(row2, row5);
+  tmp3 = vaddq_s16(row3, row4);
+  tmp4 = vsubq_s16(row3, row4);
+
+  /* Even part */
+  tmp10 = vaddq_s16(tmp0, tmp3);              /* phase 2 */
+  tmp13 = vsubq_s16(tmp0, tmp3);
+  tmp11 = vaddq_s16(tmp1, tmp2);
+  tmp12 = vsubq_s16(tmp1, tmp2);
+
+  row0 = vaddq_s16(tmp10, tmp11);             /* phase 3 */
+  row4 = vsubq_s16(tmp10, tmp11);
+
+  z1 = vqdmulhq_lane_s16(vaddq_s16(tmp12, tmp13), consts, 2);
+  row2 = vaddq_s16(tmp13, z1);                /* phase 5 */
+  row6 = vsubq_s16(tmp13, z1);
+
+  /* Odd part */
+  tmp10 = vaddq_s16(tmp4, tmp5);              /* phase 2 */
+  tmp11 = vaddq_s16(tmp5, tmp6);
+  tmp12 = vaddq_s16(tmp6, tmp7);
+
+  z5 = vqdmulhq_lane_s16(vsubq_s16(tmp10, tmp12), consts, 0);
+  z2 = vqdmulhq_lane_s16(tmp10, consts, 1);
+  z2 = vaddq_s16(z2, z5);
+  z4 = vqdmulhq_lane_s16(tmp12, consts, 3);
+  z5 = vaddq_s16(tmp12, z5);
+  z4 = vaddq_s16(z4, z5);
+  z3 = vqdmulhq_lane_s16(tmp11, consts, 2);
+
+  z11 = vaddq_s16(tmp7, z3);                  /* phase 5 */
+  z13 = vsubq_s16(tmp7, z3);
+
+  row5 = vaddq_s16(z13, z2);                  /* phase 6 */
+  row3 = vsubq_s16(z13, z2);
+  row1 = vaddq_s16(z11, z4);
+  row7 = vsubq_s16(z11, z4);
+
+  vst1q_s16(data + 0 * DCTSIZE, row0);
+  vst1q_s16(data + 1 * DCTSIZE, row1);
+  vst1q_s16(data + 2 * DCTSIZE, row2);
+  vst1q_s16(data + 3 * DCTSIZE, row3);
+  vst1q_s16(data + 4 * DCTSIZE, row4);
+  vst1q_s16(data + 5 * DCTSIZE, row5);
+  vst1q_s16(data + 6 * DCTSIZE, row6);
+  vst1q_s16(data + 7 * DCTSIZE, row7);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jfdctint-neon.c Added

@@ -0,0 +1,376 @@
+/*
+ * jfdctint-neon.c - accurate integer FDCT (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+#include "neon-compat.h"
+
+#include <arm_neon.h>
+
+
+/* jsimd_fdct_islow_neon() performs a slower but more accurate forward DCT
+ * (Discrete Cosine Transform) on one block of samples.  It uses the same
+ * calculations and produces exactly the same output as IJG's original
+ * jpeg_fdct_islow() function, which can be found in jfdctint.c.
+ *
+ * Scaled integer constants are used to avoid floating-point arithmetic:
+ *    0.298631336 =  2446 * 2^-13
+ *    0.390180644 =  3196 * 2^-13
+ *    0.541196100 =  4433 * 2^-13
+ *    0.765366865 =  6270 * 2^-13
+ *    0.899976223 =  7373 * 2^-13
+ *    1.175875602 =  9633 * 2^-13
+ *    1.501321110 = 12299 * 2^-13
+ *    1.847759065 = 15137 * 2^-13
+ *    1.961570560 = 16069 * 2^-13
+ *    2.053119869 = 16819 * 2^-13
+ *    2.562915447 = 20995 * 2^-13
+ *    3.072711026 = 25172 * 2^-13
+ *
+ * See jfdctint.c for further details of the DCT algorithm.  Where possible,
+ * the variable names and comments here in jsimd_fdct_islow_neon() match up
+ * with those in jpeg_fdct_islow().
+ */
+
+#define CONST_BITS  13
+#define PASS1_BITS  2
+
+#define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2  (CONST_BITS + PASS1_BITS)
+
+#define F_0_298  2446
+#define F_0_390  3196
+#define F_0_541  4433
+#define F_0_765  6270
+#define F_0_899  7373
+#define F_1_175  9633
+#define F_1_501  12299
+#define F_1_847  15137
+#define F_1_961  16069
+#define F_2_053  16819
+#define F_2_562  20995
+#define F_3_072  25172
+
+
+ALIGN(16) static const int16_t jsimd_fdct_islow_neon_consts = {
+  F_0_298, -F_0_390,  F_0_541,  F_0_765,
+ -F_0_899,  F_1_175,  F_1_501, -F_1_847,
+ -F_1_961,  F_2_053, -F_2_562,  F_3_072
+};
+
+void jsimd_fdct_islow_neon(DCTELEM *data)
+{
+  /* Load DCT constants. */
+#ifdef HAVE_VLD1_S16_X3
+  const int16x4x3_t consts = vld1_s16_x3(jsimd_fdct_islow_neon_consts);
+#else
+  /* GCC does not currently support the intrinsic vld1_<type>_x3(). */
+  const int16x4_t consts1 = vld1_s16(jsimd_fdct_islow_neon_consts);
+  const int16x4_t consts2 = vld1_s16(jsimd_fdct_islow_neon_consts + 4);
+  const int16x4_t consts3 = vld1_s16(jsimd_fdct_islow_neon_consts + 8);
+  const int16x4x3_t consts = { { consts1, consts2, consts3 } };
+#endif
+
+  /* Load an 8x8 block of samples into Neon registers.  De-interleaving loads
+   * are used, followed by vuzp to transpose the block such that we have a
+   * column of samples per vector - allowing all rows to be processed at once.
+   */
+  int16x8x4_t s_rows_0123 = vld4q_s16(data);
+  int16x8x4_t s_rows_4567 = vld4q_s16(data + 4 * DCTSIZE);
+
+  int16x8x2_t cols_04 = vuzpq_s16(s_rows_0123.val0, s_rows_4567.val0);
+  int16x8x2_t cols_15 = vuzpq_s16(s_rows_0123.val1, s_rows_4567.val1);
+  int16x8x2_t cols_26 = vuzpq_s16(s_rows_0123.val2, s_rows_4567.val2);
+  int16x8x2_t cols_37 = vuzpq_s16(s_rows_0123.val3, s_rows_4567.val3);
+
+  int16x8_t col0 = cols_04.val0;
+  int16x8_t col1 = cols_15.val0;
+  int16x8_t col2 = cols_26.val0;
+  int16x8_t col3 = cols_37.val0;
+  int16x8_t col4 = cols_04.val1;
+  int16x8_t col5 = cols_15.val1;
+  int16x8_t col6 = cols_26.val1;
+  int16x8_t col7 = cols_37.val1;
+
+  /* Pass 1: process rows. */
+
+  int16x8_t tmp0 = vaddq_s16(col0, col7);
+  int16x8_t tmp7 = vsubq_s16(col0, col7);
+  int16x8_t tmp1 = vaddq_s16(col1, col6);
+  int16x8_t tmp6 = vsubq_s16(col1, col6);
+  int16x8_t tmp2 = vaddq_s16(col2, col5);
+  int16x8_t tmp5 = vsubq_s16(col2, col5);
+  int16x8_t tmp3 = vaddq_s16(col3, col4);
+  int16x8_t tmp4 = vsubq_s16(col3, col4);
+
+  /* Even part */
+  int16x8_t tmp10 = vaddq_s16(tmp0, tmp3);
+  int16x8_t tmp13 = vsubq_s16(tmp0, tmp3);
+  int16x8_t tmp11 = vaddq_s16(tmp1, tmp2);
+  int16x8_t tmp12 = vsubq_s16(tmp1, tmp2);
+
+  col0 = vshlq_n_s16(vaddq_s16(tmp10, tmp11), PASS1_BITS);
+  col4 = vshlq_n_s16(vsubq_s16(tmp10, tmp11), PASS1_BITS);
+
+  int16x8_t tmp12_add_tmp13 = vaddq_s16(tmp12, tmp13);
+  int32x4_t z1_l =
+    vmull_lane_s16(vget_low_s16(tmp12_add_tmp13), consts.val0, 2);
+  int32x4_t z1_h =
+    vmull_lane_s16(vget_high_s16(tmp12_add_tmp13), consts.val0, 2);
+
+  int32x4_t col2_scaled_l =
+    vmlal_lane_s16(z1_l, vget_low_s16(tmp13), consts.val0, 3);
+  int32x4_t col2_scaled_h =
+    vmlal_lane_s16(z1_h, vget_high_s16(tmp13), consts.val0, 3);
+  col2 = vcombine_s16(vrshrn_n_s32(col2_scaled_l, DESCALE_P1),
+                      vrshrn_n_s32(col2_scaled_h, DESCALE_P1));
+
+  int32x4_t col6_scaled_l =
+    vmlal_lane_s16(z1_l, vget_low_s16(tmp12), consts.val1, 3);
+  int32x4_t col6_scaled_h =
+    vmlal_lane_s16(z1_h, vget_high_s16(tmp12), consts.val1, 3);
+  col6 = vcombine_s16(vrshrn_n_s32(col6_scaled_l, DESCALE_P1),
+                      vrshrn_n_s32(col6_scaled_h, DESCALE_P1));
+
+  /* Odd part */
+  int16x8_t z1 = vaddq_s16(tmp4, tmp7);
+  int16x8_t z2 = vaddq_s16(tmp5, tmp6);
+  int16x8_t z3 = vaddq_s16(tmp4, tmp6);
+  int16x8_t z4 = vaddq_s16(tmp5, tmp7);
+  /* sqrt(2) * c3 */
+  int32x4_t z5_l = vmull_lane_s16(vget_low_s16(z3), consts.val1, 1);
+  int32x4_t z5_h = vmull_lane_s16(vget_high_s16(z3), consts.val1, 1);
+  z5_l = vmlal_lane_s16(z5_l, vget_low_s16(z4), consts.val1, 1);
+  z5_h = vmlal_lane_s16(z5_h, vget_high_s16(z4), consts.val1, 1);
+
+  /* sqrt(2) * (-c1+c3+c5-c7) */
+  int32x4_t tmp4_l = vmull_lane_s16(vget_low_s16(tmp4), consts.val0, 0);
+  int32x4_t tmp4_h = vmull_lane_s16(vget_high_s16(tmp4), consts.val0, 0);
+  /* sqrt(2) * ( c1+c3-c5+c7) */
+  int32x4_t tmp5_l = vmull_lane_s16(vget_low_s16(tmp5), consts.val2, 1);
+  int32x4_t tmp5_h = vmull_lane_s16(vget_high_s16(tmp5), consts.val2, 1);
+  /* sqrt(2) * ( c1+c3+c5-c7) */
+  int32x4_t tmp6_l = vmull_lane_s16(vget_low_s16(tmp6), consts.val2, 3);
+  int32x4_t tmp6_h = vmull_lane_s16(vget_high_s16(tmp6), consts.val2, 3);
+  /* sqrt(2) * ( c1+c3-c5-c7) */
+  int32x4_t tmp7_l = vmull_lane_s16(vget_low_s16(tmp7), consts.val1, 2);
+  int32x4_t tmp7_h = vmull_lane_s16(vget_high_s16(tmp7), consts.val1, 2);
+
+  /* sqrt(2) * (c7-c3) */
+  z1_l = vmull_lane_s16(vget_low_s16(z1), consts.val1, 0);
+  z1_h = vmull_lane_s16(vget_high_s16(z1), consts.val1, 0);
+  /* sqrt(2) * (-c1-c3) */
+  int32x4_t z2_l = vmull_lane_s16(vget_low_s16(z2), consts.val2, 2);
+  int32x4_t z2_h = vmull_lane_s16(vget_high_s16(z2), consts.val2, 2);
+  /* sqrt(2) * (-c3-c5) */
+  int32x4_t z3_l = vmull_lane_s16(vget_low_s16(z3), consts.val2, 0);
+  int32x4_t z3_h = vmull_lane_s16(vget_high_s16(z3), consts.val2, 0);
+  /* sqrt(2) * (c5-c3) */
+  int32x4_t z4_l = vmull_lane_s16(vget_low_s16(z4), consts.val0, 1);
+  int32x4_t z4_h = vmull_lane_s16(vget_high_s16(z4), consts.val0, 1);
+
+  z3_l = vaddq_s32(z3_l, z5_l);
+  z3_h = vaddq_s32(z3_h, z5_h);
+  z4_l = vaddq_s32(z4_l, z5_l);
+  z4_h = vaddq_s32(z4_h, z5_h);
+
+  tmp4_l = vaddq_s32(tmp4_l, z1_l);
+  tmp4_h = vaddq_s32(tmp4_h, z1_h);
+  tmp4_l = vaddq_s32(tmp4_l, z3_l);
+  tmp4_h = vaddq_s32(tmp4_h, z3_h);
+  col7 = vcombine_s16(vrshrn_n_s32(tmp4_l, DESCALE_P1),
+                      vrshrn_n_s32(tmp4_h, DESCALE_P1));
+
+  tmp5_l = vaddq_s32(tmp5_l, z2_l);
+  tmp5_h = vaddq_s32(tmp5_h, z2_h);
+  tmp5_l = vaddq_s32(tmp5_l, z4_l);
+  tmp5_h = vaddq_s32(tmp5_h, z4_h);
+  col5 = vcombine_s16(vrshrn_n_s32(tmp5_l, DESCALE_P1),
+                      vrshrn_n_s32(tmp5_h, DESCALE_P1));
+
+  tmp6_l = vaddq_s32(tmp6_l, z2_l);
+  tmp6_h = vaddq_s32(tmp6_h, z2_h);
+  tmp6_l = vaddq_s32(tmp6_l, z3_l);
+  tmp6_h = vaddq_s32(tmp6_h, z3_h);
+  col3 = vcombine_s16(vrshrn_n_s32(tmp6_l, DESCALE_P1),
+                      vrshrn_n_s32(tmp6_h, DESCALE_P1));
+
+  tmp7_l = vaddq_s32(tmp7_l, z1_l);
+  tmp7_h = vaddq_s32(tmp7_h, z1_h);
+  tmp7_l = vaddq_s32(tmp7_l, z4_l);
+  tmp7_h = vaddq_s32(tmp7_h, z4_h);
+  col1 = vcombine_s16(vrshrn_n_s32(tmp7_l, DESCALE_P1),
+                      vrshrn_n_s32(tmp7_h, DESCALE_P1));
+
+  /* Transpose to work on columns in pass 2. */
+  int16x8x2_t cols_01 = vtrnq_s16(col0, col1);
+  int16x8x2_t cols_23 = vtrnq_s16(col2, col3);
+  int16x8x2_t cols_45 = vtrnq_s16(col4, col5);
+  int16x8x2_t cols_67 = vtrnq_s16(col6, col7);
+
+  int32x4x2_t cols_0145_l = vtrnq_s32(vreinterpretq_s32_s16(cols_01.val0),
+                                      vreinterpretq_s32_s16(cols_45.val0));
+  int32x4x2_t cols_0145_h = vtrnq_s32(vreinterpretq_s32_s16(cols_01.val1),
+                                      vreinterpretq_s32_s16(cols_45.val1));
+  int32x4x2_t cols_2367_l = vtrnq_s32(vreinterpretq_s32_s16(cols_23.val0),
+                                      vreinterpretq_s32_s16(cols_67.val0));
+  int32x4x2_t cols_2367_h = vtrnq_s32(vreinterpretq_s32_s16(cols_23.val1),
+                                      vreinterpretq_s32_s16(cols_67.val1));
+
+  int32x4x2_t rows_04 = vzipq_s32(cols_0145_l.val0, cols_2367_l.val0);
+  int32x4x2_t rows_15 = vzipq_s32(cols_0145_h.val0, cols_2367_h.val0);
+  int32x4x2_t rows_26 = vzipq_s32(cols_0145_l.val1, cols_2367_l.val1);
+  int32x4x2_t rows_37 = vzipq_s32(cols_0145_h.val1, cols_2367_h.val1);
+
+  int16x8_t row0 = vreinterpretq_s16_s32(rows_04.val0);
+  int16x8_t row1 = vreinterpretq_s16_s32(rows_15.val0);
+  int16x8_t row2 = vreinterpretq_s16_s32(rows_26.val0);
+  int16x8_t row3 = vreinterpretq_s16_s32(rows_37.val0);
+  int16x8_t row4 = vreinterpretq_s16_s32(rows_04.val1);
+  int16x8_t row5 = vreinterpretq_s16_s32(rows_15.val1);
+  int16x8_t row6 = vreinterpretq_s16_s32(rows_26.val1);
+  int16x8_t row7 = vreinterpretq_s16_s32(rows_37.val1);
+
+  /* Pass 2: process columns. */
+
+  tmp0 = vaddq_s16(row0, row7);
+  tmp7 = vsubq_s16(row0, row7);
+  tmp1 = vaddq_s16(row1, row6);
+  tmp6 = vsubq_s16(row1, row6);
+  tmp2 = vaddq_s16(row2, row5);
+  tmp5 = vsubq_s16(row2, row5);
+  tmp3 = vaddq_s16(row3, row4);
+  tmp4 = vsubq_s16(row3, row4);
+
+  /* Even part */
+  tmp10 = vaddq_s16(tmp0, tmp3);
+  tmp13 = vsubq_s16(tmp0, tmp3);
+  tmp11 = vaddq_s16(tmp1, tmp2);
+  tmp12 = vsubq_s16(tmp1, tmp2);
+
+  row0 = vrshrq_n_s16(vaddq_s16(tmp10, tmp11), PASS1_BITS);
+  row4 = vrshrq_n_s16(vsubq_s16(tmp10, tmp11), PASS1_BITS);
+
+  tmp12_add_tmp13 = vaddq_s16(tmp12, tmp13);
+  z1_l = vmull_lane_s16(vget_low_s16(tmp12_add_tmp13), consts.val0, 2);
+  z1_h = vmull_lane_s16(vget_high_s16(tmp12_add_tmp13), consts.val0, 2);
+
+  int32x4_t row2_scaled_l =
+    vmlal_lane_s16(z1_l, vget_low_s16(tmp13), consts.val0, 3);
+  int32x4_t row2_scaled_h =
+    vmlal_lane_s16(z1_h, vget_high_s16(tmp13), consts.val0, 3);
+  row2 = vcombine_s16(vrshrn_n_s32(row2_scaled_l, DESCALE_P2),
+                      vrshrn_n_s32(row2_scaled_h, DESCALE_P2));
+
+  int32x4_t row6_scaled_l =
+    vmlal_lane_s16(z1_l, vget_low_s16(tmp12), consts.val1, 3);
+  int32x4_t row6_scaled_h =
+    vmlal_lane_s16(z1_h, vget_high_s16(tmp12), consts.val1, 3);
+  row6 = vcombine_s16(vrshrn_n_s32(row6_scaled_l, DESCALE_P2),
+                      vrshrn_n_s32(row6_scaled_h, DESCALE_P2));
+
+  /* Odd part */
+  z1 = vaddq_s16(tmp4, tmp7);
+  z2 = vaddq_s16(tmp5, tmp6);
+  z3 = vaddq_s16(tmp4, tmp6);
+  z4 = vaddq_s16(tmp5, tmp7);
+  /* sqrt(2) * c3 */
+  z5_l = vmull_lane_s16(vget_low_s16(z3), consts.val1, 1);
+  z5_h = vmull_lane_s16(vget_high_s16(z3), consts.val1, 1);
+  z5_l = vmlal_lane_s16(z5_l, vget_low_s16(z4), consts.val1, 1);
+  z5_h = vmlal_lane_s16(z5_h, vget_high_s16(z4), consts.val1, 1);
+
+  /* sqrt(2) * (-c1+c3+c5-c7) */
+  tmp4_l = vmull_lane_s16(vget_low_s16(tmp4), consts.val0, 0);
+  tmp4_h = vmull_lane_s16(vget_high_s16(tmp4), consts.val0, 0);
+  /* sqrt(2) * ( c1+c3-c5+c7) */
+  tmp5_l = vmull_lane_s16(vget_low_s16(tmp5), consts.val2, 1);
+  tmp5_h = vmull_lane_s16(vget_high_s16(tmp5), consts.val2, 1);
+  /* sqrt(2) * ( c1+c3+c5-c7) */
+  tmp6_l = vmull_lane_s16(vget_low_s16(tmp6), consts.val2, 3);
+  tmp6_h = vmull_lane_s16(vget_high_s16(tmp6), consts.val2, 3);
+  /* sqrt(2) * ( c1+c3-c5-c7) */
+  tmp7_l = vmull_lane_s16(vget_low_s16(tmp7), consts.val1, 2);
+  tmp7_h = vmull_lane_s16(vget_high_s16(tmp7), consts.val1, 2);
+
+  /* sqrt(2) * (c7-c3) */
+  z1_l = vmull_lane_s16(vget_low_s16(z1), consts.val1, 0);
+  z1_h = vmull_lane_s16(vget_high_s16(z1), consts.val1, 0);
+  /* sqrt(2) * (-c1-c3) */
+  z2_l = vmull_lane_s16(vget_low_s16(z2), consts.val2, 2);
+  z2_h = vmull_lane_s16(vget_high_s16(z2), consts.val2, 2);
+  /* sqrt(2) * (-c3-c5) */
+  z3_l = vmull_lane_s16(vget_low_s16(z3), consts.val2, 0);
+  z3_h = vmull_lane_s16(vget_high_s16(z3), consts.val2, 0);
+  /* sqrt(2) * (c5-c3) */
+  z4_l = vmull_lane_s16(vget_low_s16(z4), consts.val0, 1);
+  z4_h = vmull_lane_s16(vget_high_s16(z4), consts.val0, 1);
+
+  z3_l = vaddq_s32(z3_l, z5_l);
+  z3_h = vaddq_s32(z3_h, z5_h);
+  z4_l = vaddq_s32(z4_l, z5_l);
+  z4_h = vaddq_s32(z4_h, z5_h);
+
+  tmp4_l = vaddq_s32(tmp4_l, z1_l);
+  tmp4_h = vaddq_s32(tmp4_h, z1_h);
+  tmp4_l = vaddq_s32(tmp4_l, z3_l);
+  tmp4_h = vaddq_s32(tmp4_h, z3_h);
+  row7 = vcombine_s16(vrshrn_n_s32(tmp4_l, DESCALE_P2),
+                      vrshrn_n_s32(tmp4_h, DESCALE_P2));
+
+  tmp5_l = vaddq_s32(tmp5_l, z2_l);
+  tmp5_h = vaddq_s32(tmp5_h, z2_h);
+  tmp5_l = vaddq_s32(tmp5_l, z4_l);
+  tmp5_h = vaddq_s32(tmp5_h, z4_h);
+  row5 = vcombine_s16(vrshrn_n_s32(tmp5_l, DESCALE_P2),
+                      vrshrn_n_s32(tmp5_h, DESCALE_P2));
+
+  tmp6_l = vaddq_s32(tmp6_l, z2_l);
+  tmp6_h = vaddq_s32(tmp6_h, z2_h);
+  tmp6_l = vaddq_s32(tmp6_l, z3_l);
+  tmp6_h = vaddq_s32(tmp6_h, z3_h);
+  row3 = vcombine_s16(vrshrn_n_s32(tmp6_l, DESCALE_P2),
+                      vrshrn_n_s32(tmp6_h, DESCALE_P2));
+
+  tmp7_l = vaddq_s32(tmp7_l, z1_l);
+  tmp7_h = vaddq_s32(tmp7_h, z1_h);
+  tmp7_l = vaddq_s32(tmp7_l, z4_l);
+  tmp7_h = vaddq_s32(tmp7_h, z4_h);
+  row1 = vcombine_s16(vrshrn_n_s32(tmp7_l, DESCALE_P2),
+                      vrshrn_n_s32(tmp7_h, DESCALE_P2));
+
+  vst1q_s16(data + 0 * DCTSIZE, row0);
+  vst1q_s16(data + 1 * DCTSIZE, row1);
+  vst1q_s16(data + 2 * DCTSIZE, row2);
+  vst1q_s16(data + 3 * DCTSIZE, row3);
+  vst1q_s16(data + 4 * DCTSIZE, row4);
+  vst1q_s16(data + 5 * DCTSIZE, row5);
+  vst1q_s16(data + 6 * DCTSIZE, row6);
+  vst1q_s16(data + 7 * DCTSIZE, row7);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jidctfst-neon.c Added

@@ -0,0 +1,472 @@
+/*
+ * jidctfst-neon.c - fast integer IDCT (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+
+#include <arm_neon.h>
+
+
+/* jsimd_idct_ifast_neon() performs dequantization and a fast, not so accurate
+ * inverse DCT (Discrete Cosine Transform) on one block of coefficients.  It
+ * uses the same calculations and produces exactly the same output as IJG's
+ * original jpeg_idct_ifast() function, which can be found in jidctfst.c.
+ *
+ * Scaled integer constants are used to avoid floating-point arithmetic:
+ *    0.082392200 =  2688 * 2^-15
+ *    0.414213562 = 13568 * 2^-15
+ *    0.847759065 = 27776 * 2^-15
+ *    0.613125930 = 20096 * 2^-15
+ *
+ * See jidctfst.c for further details of the IDCT algorithm.  Where possible,
+ * the variable names and comments here in jsimd_idct_ifast_neon() match up
+ * with those in jpeg_idct_ifast().
+ */
+
+#define PASS1_BITS  2
+
+#define F_0_082  2688
+#define F_0_414  13568
+#define F_0_847  27776
+#define F_0_613  20096
+
+
+ALIGN(16) static const int16_t jsimd_idct_ifast_neon_consts = {
+  F_0_082, F_0_414, F_0_847, F_0_613
+};
+
+void jsimd_idct_ifast_neon(void *dct_table, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  IFAST_MULT_TYPE *quantptr = dct_table;
+
+  /* Load DCT coefficients. */
+  int16x8_t row0 = vld1q_s16(coef_block + 0 * DCTSIZE);
+  int16x8_t row1 = vld1q_s16(coef_block + 1 * DCTSIZE);
+  int16x8_t row2 = vld1q_s16(coef_block + 2 * DCTSIZE);
+  int16x8_t row3 = vld1q_s16(coef_block + 3 * DCTSIZE);
+  int16x8_t row4 = vld1q_s16(coef_block + 4 * DCTSIZE);
+  int16x8_t row5 = vld1q_s16(coef_block + 5 * DCTSIZE);
+  int16x8_t row6 = vld1q_s16(coef_block + 6 * DCTSIZE);
+  int16x8_t row7 = vld1q_s16(coef_block + 7 * DCTSIZE);
+
+  /* Load quantization table values for DC coefficients. */
+  int16x8_t quant_row0 = vld1q_s16(quantptr + 0 * DCTSIZE);
+  /* Dequantize DC coefficients. */
+  row0 = vmulq_s16(row0, quant_row0);
+
+  /* Construct bitmap to test if all AC coefficients are 0. */
+  int16x8_t bitmap = vorrq_s16(row1, row2);
+  bitmap = vorrq_s16(bitmap, row3);
+  bitmap = vorrq_s16(bitmap, row4);
+  bitmap = vorrq_s16(bitmap, row5);
+  bitmap = vorrq_s16(bitmap, row6);
+  bitmap = vorrq_s16(bitmap, row7);
+
+  int64_t left_ac_bitmap = vgetq_lane_s64(vreinterpretq_s64_s16(bitmap), 0);
+  int64_t right_ac_bitmap = vgetq_lane_s64(vreinterpretq_s64_s16(bitmap), 1);
+
+  /* Load IDCT conversion constants. */
+  const int16x4_t consts = vld1_s16(jsimd_idct_ifast_neon_consts);
+
+  if (left_ac_bitmap == 0 && right_ac_bitmap == 0) {
+    /* All AC coefficients are zero.
+     * Compute DC values and duplicate into vectors.
+     */
+    int16x8_t dcval = row0;
+    row1 = dcval;
+    row2 = dcval;
+    row3 = dcval;
+    row4 = dcval;
+    row5 = dcval;
+    row6 = dcval;
+    row7 = dcval;
+  } else if (left_ac_bitmap == 0) {
+    /* AC coefficients are zero for columns 0, 1, 2, and 3.
+     * Use DC values for these columns.
+     */
+    int16x4_t dcval = vget_low_s16(row0);
+
+    /* Commence regular fast IDCT computation for columns 4, 5, 6, and 7. */
+
+    /* Load quantization table. */
+    int16x4_t quant_row1 = vld1_s16(quantptr + 1 * DCTSIZE + 4);
+    int16x4_t quant_row2 = vld1_s16(quantptr + 2 * DCTSIZE + 4);
+    int16x4_t quant_row3 = vld1_s16(quantptr + 3 * DCTSIZE + 4);
+    int16x4_t quant_row4 = vld1_s16(quantptr + 4 * DCTSIZE + 4);
+    int16x4_t quant_row5 = vld1_s16(quantptr + 5 * DCTSIZE + 4);
+    int16x4_t quant_row6 = vld1_s16(quantptr + 6 * DCTSIZE + 4);
+    int16x4_t quant_row7 = vld1_s16(quantptr + 7 * DCTSIZE + 4);
+
+    /* Even part: dequantize DCT coefficients. */
+    int16x4_t tmp0 = vget_high_s16(row0);
+    int16x4_t tmp1 = vmul_s16(vget_high_s16(row2), quant_row2);
+    int16x4_t tmp2 = vmul_s16(vget_high_s16(row4), quant_row4);
+    int16x4_t tmp3 = vmul_s16(vget_high_s16(row6), quant_row6);
+
+    int16x4_t tmp10 = vadd_s16(tmp0, tmp2);   /* phase 3 */
+    int16x4_t tmp11 = vsub_s16(tmp0, tmp2);
+
+    int16x4_t tmp13 = vadd_s16(tmp1, tmp3);   /* phases 5-3 */
+    int16x4_t tmp1_sub_tmp3 = vsub_s16(tmp1, tmp3);
+    int16x4_t tmp12 = vqdmulh_lane_s16(tmp1_sub_tmp3, consts, 1);
+    tmp12 = vadd_s16(tmp12, tmp1_sub_tmp3);
+    tmp12 = vsub_s16(tmp12, tmp13);
+
+    tmp0 = vadd_s16(tmp10, tmp13);            /* phase 2 */
+    tmp3 = vsub_s16(tmp10, tmp13);
+    tmp1 = vadd_s16(tmp11, tmp12);
+    tmp2 = vsub_s16(tmp11, tmp12);
+
+    /* Odd part: dequantize DCT coefficients. */
+    int16x4_t tmp4 = vmul_s16(vget_high_s16(row1), quant_row1);
+    int16x4_t tmp5 = vmul_s16(vget_high_s16(row3), quant_row3);
+    int16x4_t tmp6 = vmul_s16(vget_high_s16(row5), quant_row5);
+    int16x4_t tmp7 = vmul_s16(vget_high_s16(row7), quant_row7);
+
+    int16x4_t z13 = vadd_s16(tmp6, tmp5);     /* phase 6 */
+    int16x4_t neg_z10 = vsub_s16(tmp5, tmp6);
+    int16x4_t z11 = vadd_s16(tmp4, tmp7);
+    int16x4_t z12 = vsub_s16(tmp4, tmp7);
+
+    tmp7 = vadd_s16(z11, z13);                /* phase 5 */
+    int16x4_t z11_sub_z13 = vsub_s16(z11, z13);
+    tmp11 = vqdmulh_lane_s16(z11_sub_z13, consts, 1);
+    tmp11 = vadd_s16(tmp11, z11_sub_z13);
+
+    int16x4_t z10_add_z12 = vsub_s16(z12, neg_z10);
+    int16x4_t z5 = vqdmulh_lane_s16(z10_add_z12, consts, 2);
+    z5 = vadd_s16(z5, z10_add_z12);
+    tmp10 = vqdmulh_lane_s16(z12, consts, 0);
+    tmp10 = vadd_s16(tmp10, z12);
+    tmp10 = vsub_s16(tmp10, z5);
+    tmp12 = vqdmulh_lane_s16(neg_z10, consts, 3);
+    tmp12 = vadd_s16(tmp12, vadd_s16(neg_z10, neg_z10));
+    tmp12 = vadd_s16(tmp12, z5);
+
+    tmp6 = vsub_s16(tmp12, tmp7);             /* phase 2 */
+    tmp5 = vsub_s16(tmp11, tmp6);
+    tmp4 = vadd_s16(tmp10, tmp5);
+
+    row0 = vcombine_s16(dcval, vadd_s16(tmp0, tmp7));
+    row7 = vcombine_s16(dcval, vsub_s16(tmp0, tmp7));
+    row1 = vcombine_s16(dcval, vadd_s16(tmp1, tmp6));
+    row6 = vcombine_s16(dcval, vsub_s16(tmp1, tmp6));
+    row2 = vcombine_s16(dcval, vadd_s16(tmp2, tmp5));
+    row5 = vcombine_s16(dcval, vsub_s16(tmp2, tmp5));
+    row4 = vcombine_s16(dcval, vadd_s16(tmp3, tmp4));
+    row3 = vcombine_s16(dcval, vsub_s16(tmp3, tmp4));
+  } else if (right_ac_bitmap == 0) {
+    /* AC coefficients are zero for columns 4, 5, 6, and 7.
+     * Use DC values for these columns.
+     */
+    int16x4_t dcval = vget_high_s16(row0);
+
+    /* Commence regular fast IDCT computation for columns 0, 1, 2, and 3. */
+
+    /* Load quantization table. */
+    int16x4_t quant_row1 = vld1_s16(quantptr + 1 * DCTSIZE);
+    int16x4_t quant_row2 = vld1_s16(quantptr + 2 * DCTSIZE);
+    int16x4_t quant_row3 = vld1_s16(quantptr + 3 * DCTSIZE);
+    int16x4_t quant_row4 = vld1_s16(quantptr + 4 * DCTSIZE);
+    int16x4_t quant_row5 = vld1_s16(quantptr + 5 * DCTSIZE);
+    int16x4_t quant_row6 = vld1_s16(quantptr + 6 * DCTSIZE);
+    int16x4_t quant_row7 = vld1_s16(quantptr + 7 * DCTSIZE);
+
+    /* Even part: dequantize DCT coefficients. */
+    int16x4_t tmp0 = vget_low_s16(row0);
+    int16x4_t tmp1 = vmul_s16(vget_low_s16(row2), quant_row2);
+    int16x4_t tmp2 = vmul_s16(vget_low_s16(row4), quant_row4);
+    int16x4_t tmp3 = vmul_s16(vget_low_s16(row6), quant_row6);
+
+    int16x4_t tmp10 = vadd_s16(tmp0, tmp2);   /* phase 3 */
+    int16x4_t tmp11 = vsub_s16(tmp0, tmp2);
+
+    int16x4_t tmp13 = vadd_s16(tmp1, tmp3);   /* phases 5-3 */
+    int16x4_t tmp1_sub_tmp3 = vsub_s16(tmp1, tmp3);
+    int16x4_t tmp12 = vqdmulh_lane_s16(tmp1_sub_tmp3, consts, 1);
+    tmp12 = vadd_s16(tmp12, tmp1_sub_tmp3);
+    tmp12 = vsub_s16(tmp12, tmp13);
+
+    tmp0 = vadd_s16(tmp10, tmp13);            /* phase 2 */
+    tmp3 = vsub_s16(tmp10, tmp13);
+    tmp1 = vadd_s16(tmp11, tmp12);
+    tmp2 = vsub_s16(tmp11, tmp12);
+
+    /* Odd part: dequantize DCT coefficients. */
+    int16x4_t tmp4 = vmul_s16(vget_low_s16(row1), quant_row1);
+    int16x4_t tmp5 = vmul_s16(vget_low_s16(row3), quant_row3);
+    int16x4_t tmp6 = vmul_s16(vget_low_s16(row5), quant_row5);
+    int16x4_t tmp7 = vmul_s16(vget_low_s16(row7), quant_row7);
+
+    int16x4_t z13 = vadd_s16(tmp6, tmp5);     /* phase 6 */
+    int16x4_t neg_z10 = vsub_s16(tmp5, tmp6);
+    int16x4_t z11 = vadd_s16(tmp4, tmp7);
+    int16x4_t z12 = vsub_s16(tmp4, tmp7);
+
+    tmp7 = vadd_s16(z11, z13);                /* phase 5 */
+    int16x4_t z11_sub_z13 = vsub_s16(z11, z13);
+    tmp11 = vqdmulh_lane_s16(z11_sub_z13, consts, 1);
+    tmp11 = vadd_s16(tmp11, z11_sub_z13);
+
+    int16x4_t z10_add_z12 = vsub_s16(z12, neg_z10);
+    int16x4_t z5 = vqdmulh_lane_s16(z10_add_z12, consts, 2);
+    z5 = vadd_s16(z5, z10_add_z12);
+    tmp10 = vqdmulh_lane_s16(z12, consts, 0);
+    tmp10 = vadd_s16(tmp10, z12);
+    tmp10 = vsub_s16(tmp10, z5);
+    tmp12 = vqdmulh_lane_s16(neg_z10, consts, 3);
+    tmp12 = vadd_s16(tmp12, vadd_s16(neg_z10, neg_z10));
+    tmp12 = vadd_s16(tmp12, z5);
+
+    tmp6 = vsub_s16(tmp12, tmp7);             /* phase 2 */
+    tmp5 = vsub_s16(tmp11, tmp6);
+    tmp4 = vadd_s16(tmp10, tmp5);
+
+    row0 = vcombine_s16(vadd_s16(tmp0, tmp7), dcval);
+    row7 = vcombine_s16(vsub_s16(tmp0, tmp7), dcval);
+    row1 = vcombine_s16(vadd_s16(tmp1, tmp6), dcval);
+    row6 = vcombine_s16(vsub_s16(tmp1, tmp6), dcval);
+    row2 = vcombine_s16(vadd_s16(tmp2, tmp5), dcval);
+    row5 = vcombine_s16(vsub_s16(tmp2, tmp5), dcval);
+    row4 = vcombine_s16(vadd_s16(tmp3, tmp4), dcval);
+    row3 = vcombine_s16(vsub_s16(tmp3, tmp4), dcval);
+  } else {
+    /* Some AC coefficients are non-zero; full IDCT calculation required. */
+
+    /* Load quantization table. */
+    int16x8_t quant_row1 = vld1q_s16(quantptr + 1 * DCTSIZE);
+    int16x8_t quant_row2 = vld1q_s16(quantptr + 2 * DCTSIZE);
+    int16x8_t quant_row3 = vld1q_s16(quantptr + 3 * DCTSIZE);
+    int16x8_t quant_row4 = vld1q_s16(quantptr + 4 * DCTSIZE);
+    int16x8_t quant_row5 = vld1q_s16(quantptr + 5 * DCTSIZE);
+    int16x8_t quant_row6 = vld1q_s16(quantptr + 6 * DCTSIZE);
+    int16x8_t quant_row7 = vld1q_s16(quantptr + 7 * DCTSIZE);
+
+    /* Even part: dequantize DCT coefficients. */
+    int16x8_t tmp0 = row0;
+    int16x8_t tmp1 = vmulq_s16(row2, quant_row2);
+    int16x8_t tmp2 = vmulq_s16(row4, quant_row4);
+    int16x8_t tmp3 = vmulq_s16(row6, quant_row6);
+
+    int16x8_t tmp10 = vaddq_s16(tmp0, tmp2);   /* phase 3 */
+    int16x8_t tmp11 = vsubq_s16(tmp0, tmp2);
+
+    int16x8_t tmp13 = vaddq_s16(tmp1, tmp3);   /* phases 5-3 */
+    int16x8_t tmp1_sub_tmp3 = vsubq_s16(tmp1, tmp3);
+    int16x8_t tmp12 = vqdmulhq_lane_s16(tmp1_sub_tmp3, consts, 1);
+    tmp12 = vaddq_s16(tmp12, tmp1_sub_tmp3);
+    tmp12 = vsubq_s16(tmp12, tmp13);
+
+    tmp0 = vaddq_s16(tmp10, tmp13);            /* phase 2 */
+    tmp3 = vsubq_s16(tmp10, tmp13);
+    tmp1 = vaddq_s16(tmp11, tmp12);
+    tmp2 = vsubq_s16(tmp11, tmp12);
+
+    /* Odd part: dequantize DCT coefficients. */
+    int16x8_t tmp4 = vmulq_s16(row1, quant_row1);
+    int16x8_t tmp5 = vmulq_s16(row3, quant_row3);
+    int16x8_t tmp6 = vmulq_s16(row5, quant_row5);
+    int16x8_t tmp7 = vmulq_s16(row7, quant_row7);
+
+    int16x8_t z13 = vaddq_s16(tmp6, tmp5);     /* phase 6 */
+    int16x8_t neg_z10 = vsubq_s16(tmp5, tmp6);
+    int16x8_t z11 = vaddq_s16(tmp4, tmp7);
+    int16x8_t z12 = vsubq_s16(tmp4, tmp7);
+
+    tmp7 = vaddq_s16(z11, z13);                /* phase 5 */
+    int16x8_t z11_sub_z13 = vsubq_s16(z11, z13);
+    tmp11 = vqdmulhq_lane_s16(z11_sub_z13, consts, 1);
+    tmp11 = vaddq_s16(tmp11, z11_sub_z13);
+
+    int16x8_t z10_add_z12 = vsubq_s16(z12, neg_z10);
+    int16x8_t z5 = vqdmulhq_lane_s16(z10_add_z12, consts, 2);
+    z5 = vaddq_s16(z5, z10_add_z12);
+    tmp10 = vqdmulhq_lane_s16(z12, consts, 0);
+    tmp10 = vaddq_s16(tmp10, z12);
+    tmp10 = vsubq_s16(tmp10, z5);
+    tmp12 = vqdmulhq_lane_s16(neg_z10, consts, 3);
+    tmp12 = vaddq_s16(tmp12, vaddq_s16(neg_z10, neg_z10));
+    tmp12 = vaddq_s16(tmp12, z5);
+
+    tmp6 = vsubq_s16(tmp12, tmp7);             /* phase 2 */
+    tmp5 = vsubq_s16(tmp11, tmp6);
+    tmp4 = vaddq_s16(tmp10, tmp5);
+
+    row0 = vaddq_s16(tmp0, tmp7);
+    row7 = vsubq_s16(tmp0, tmp7);
+    row1 = vaddq_s16(tmp1, tmp6);
+    row6 = vsubq_s16(tmp1, tmp6);
+    row2 = vaddq_s16(tmp2, tmp5);
+    row5 = vsubq_s16(tmp2, tmp5);
+    row4 = vaddq_s16(tmp3, tmp4);
+    row3 = vsubq_s16(tmp3, tmp4);
+  }
+
+  /* Transpose rows to work on columns in pass 2. */
+  int16x8x2_t rows_01 = vtrnq_s16(row0, row1);
+  int16x8x2_t rows_23 = vtrnq_s16(row2, row3);
+  int16x8x2_t rows_45 = vtrnq_s16(row4, row5);
+  int16x8x2_t rows_67 = vtrnq_s16(row6, row7);
+
+  int32x4x2_t rows_0145_l = vtrnq_s32(vreinterpretq_s32_s16(rows_01.val0),
+                                      vreinterpretq_s32_s16(rows_45.val0));
+  int32x4x2_t rows_0145_h = vtrnq_s32(vreinterpretq_s32_s16(rows_01.val1),
+                                      vreinterpretq_s32_s16(rows_45.val1));
+  int32x4x2_t rows_2367_l = vtrnq_s32(vreinterpretq_s32_s16(rows_23.val0),
+                                      vreinterpretq_s32_s16(rows_67.val0));
+  int32x4x2_t rows_2367_h = vtrnq_s32(vreinterpretq_s32_s16(rows_23.val1),
+                                      vreinterpretq_s32_s16(rows_67.val1));
+
+  int32x4x2_t cols_04 = vzipq_s32(rows_0145_l.val0, rows_2367_l.val0);
+  int32x4x2_t cols_15 = vzipq_s32(rows_0145_h.val0, rows_2367_h.val0);
+  int32x4x2_t cols_26 = vzipq_s32(rows_0145_l.val1, rows_2367_l.val1);
+  int32x4x2_t cols_37 = vzipq_s32(rows_0145_h.val1, rows_2367_h.val1);
+
+  int16x8_t col0 = vreinterpretq_s16_s32(cols_04.val0);
+  int16x8_t col1 = vreinterpretq_s16_s32(cols_15.val0);
+  int16x8_t col2 = vreinterpretq_s16_s32(cols_26.val0);
+  int16x8_t col3 = vreinterpretq_s16_s32(cols_37.val0);
+  int16x8_t col4 = vreinterpretq_s16_s32(cols_04.val1);
+  int16x8_t col5 = vreinterpretq_s16_s32(cols_15.val1);
+  int16x8_t col6 = vreinterpretq_s16_s32(cols_26.val1);
+  int16x8_t col7 = vreinterpretq_s16_s32(cols_37.val1);
+
+  /* 1-D IDCT, pass 2 */
+
+  /* Even part */
+  int16x8_t tmp10 = vaddq_s16(col0, col4);
+  int16x8_t tmp11 = vsubq_s16(col0, col4);
+
+  int16x8_t tmp13 = vaddq_s16(col2, col6);
+  int16x8_t col2_sub_col6 = vsubq_s16(col2, col6);
+  int16x8_t tmp12 = vqdmulhq_lane_s16(col2_sub_col6, consts, 1);
+  tmp12 = vaddq_s16(tmp12, col2_sub_col6);
+  tmp12 = vsubq_s16(tmp12, tmp13);
+
+  int16x8_t tmp0 = vaddq_s16(tmp10, tmp13);
+  int16x8_t tmp3 = vsubq_s16(tmp10, tmp13);
+  int16x8_t tmp1 = vaddq_s16(tmp11, tmp12);
+  int16x8_t tmp2 = vsubq_s16(tmp11, tmp12);
+
+  /* Odd part */
+  int16x8_t z13 = vaddq_s16(col5, col3);
+  int16x8_t neg_z10 = vsubq_s16(col3, col5);
+  int16x8_t z11 = vaddq_s16(col1, col7);
+  int16x8_t z12 = vsubq_s16(col1, col7);
+
+  int16x8_t tmp7 = vaddq_s16(z11, z13);      /* phase 5 */
+  int16x8_t z11_sub_z13 = vsubq_s16(z11, z13);
+  tmp11 = vqdmulhq_lane_s16(z11_sub_z13, consts, 1);
+  tmp11 = vaddq_s16(tmp11, z11_sub_z13);
+
+  int16x8_t z10_add_z12 = vsubq_s16(z12, neg_z10);
+  int16x8_t z5 = vqdmulhq_lane_s16(z10_add_z12, consts, 2);
+  z5 = vaddq_s16(z5, z10_add_z12);
+  tmp10 = vqdmulhq_lane_s16(z12, consts, 0);
+  tmp10 = vaddq_s16(tmp10, z12);
+  tmp10 = vsubq_s16(tmp10, z5);
+  tmp12 = vqdmulhq_lane_s16(neg_z10, consts, 3);
+  tmp12 = vaddq_s16(tmp12, vaddq_s16(neg_z10, neg_z10));
+  tmp12 = vaddq_s16(tmp12, z5);
+
+  int16x8_t tmp6 = vsubq_s16(tmp12, tmp7);   /* phase 2 */
+  int16x8_t tmp5 = vsubq_s16(tmp11, tmp6);
+  int16x8_t tmp4 = vaddq_s16(tmp10, tmp5);
+
+  col0 = vaddq_s16(tmp0, tmp7);
+  col7 = vsubq_s16(tmp0, tmp7);
+  col1 = vaddq_s16(tmp1, tmp6);
+  col6 = vsubq_s16(tmp1, tmp6);
+  col2 = vaddq_s16(tmp2, tmp5);
+  col5 = vsubq_s16(tmp2, tmp5);
+  col4 = vaddq_s16(tmp3, tmp4);
+  col3 = vsubq_s16(tmp3, tmp4);
+
+  /* Scale down by a factor of 8, narrowing to 8-bit. */
+  int8x16_t cols_01_s8 = vcombine_s8(vqshrn_n_s16(col0, PASS1_BITS + 3),
+                                     vqshrn_n_s16(col1, PASS1_BITS + 3));
+  int8x16_t cols_45_s8 = vcombine_s8(vqshrn_n_s16(col4, PASS1_BITS + 3),
+                                     vqshrn_n_s16(col5, PASS1_BITS + 3));
+  int8x16_t cols_23_s8 = vcombine_s8(vqshrn_n_s16(col2, PASS1_BITS + 3),
+                                     vqshrn_n_s16(col3, PASS1_BITS + 3));
+  int8x16_t cols_67_s8 = vcombine_s8(vqshrn_n_s16(col6, PASS1_BITS + 3),
+                                     vqshrn_n_s16(col7, PASS1_BITS + 3));
+  /* Clamp to range 0-255. */
+  uint8x16_t cols_01 =
+    vreinterpretq_u8_s8
+      (vaddq_s8(cols_01_s8, vreinterpretq_s8_u8(vdupq_n_u8(CENTERJSAMPLE))));
+  uint8x16_t cols_45 =
+    vreinterpretq_u8_s8
+      (vaddq_s8(cols_45_s8, vreinterpretq_s8_u8(vdupq_n_u8(CENTERJSAMPLE))));
+  uint8x16_t cols_23 =
+    vreinterpretq_u8_s8
+      (vaddq_s8(cols_23_s8, vreinterpretq_s8_u8(vdupq_n_u8(CENTERJSAMPLE))));
+  uint8x16_t cols_67 =
+    vreinterpretq_u8_s8
+      (vaddq_s8(cols_67_s8, vreinterpretq_s8_u8(vdupq_n_u8(CENTERJSAMPLE))));
+
+  /* Transpose block to prepare for store. */
+  uint32x4x2_t cols_0415 = vzipq_u32(vreinterpretq_u32_u8(cols_01),
+                                     vreinterpretq_u32_u8(cols_45));
+  uint32x4x2_t cols_2637 = vzipq_u32(vreinterpretq_u32_u8(cols_23),
+                                     vreinterpretq_u32_u8(cols_67));
+
+  uint8x16x2_t cols_0145 = vtrnq_u8(vreinterpretq_u8_u32(cols_0415.val0),
+                                    vreinterpretq_u8_u32(cols_0415.val1));
+  uint8x16x2_t cols_2367 = vtrnq_u8(vreinterpretq_u8_u32(cols_2637.val0),
+                                    vreinterpretq_u8_u32(cols_2637.val1));
+  uint16x8x2_t rows_0426 = vtrnq_u16(vreinterpretq_u16_u8(cols_0145.val0),
+                                     vreinterpretq_u16_u8(cols_2367.val0));
+  uint16x8x2_t rows_1537 = vtrnq_u16(vreinterpretq_u16_u8(cols_0145.val1),
+                                     vreinterpretq_u16_u8(cols_2367.val1));
+
+  uint8x16_t rows_04 = vreinterpretq_u8_u16(rows_0426.val0);
+  uint8x16_t rows_15 = vreinterpretq_u8_u16(rows_1537.val0);
+  uint8x16_t rows_26 = vreinterpretq_u8_u16(rows_0426.val1);
+  uint8x16_t rows_37 = vreinterpretq_u8_u16(rows_1537.val1);
+
+  JSAMPROW outptr0 = output_buf0 + output_col;
+  JSAMPROW outptr1 = output_buf1 + output_col;
+  JSAMPROW outptr2 = output_buf2 + output_col;
+  JSAMPROW outptr3 = output_buf3 + output_col;
+  JSAMPROW outptr4 = output_buf4 + output_col;
+  JSAMPROW outptr5 = output_buf5 + output_col;
+  JSAMPROW outptr6 = output_buf6 + output_col;
+  JSAMPROW outptr7 = output_buf7 + output_col;
+
+  /* Store DCT block to memory. */
+  vst1q_lane_u64((uint64_t *)outptr0, vreinterpretq_u64_u8(rows_04), 0);
+  vst1q_lane_u64((uint64_t *)outptr1, vreinterpretq_u64_u8(rows_15), 0);
+  vst1q_lane_u64((uint64_t *)outptr2, vreinterpretq_u64_u8(rows_26), 0);
+  vst1q_lane_u64((uint64_t *)outptr3, vreinterpretq_u64_u8(rows_37), 0);
+  vst1q_lane_u64((uint64_t *)outptr4, vreinterpretq_u64_u8(rows_04), 1);
+  vst1q_lane_u64((uint64_t *)outptr5, vreinterpretq_u64_u8(rows_15), 1);
+  vst1q_lane_u64((uint64_t *)outptr6, vreinterpretq_u64_u8(rows_26), 1);
+  vst1q_lane_u64((uint64_t *)outptr7, vreinterpretq_u64_u8(rows_37), 1);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jidctint-neon.c Added

@@ -0,0 +1,802 @@
+/*
+ * jidctint-neon.c - accurate integer IDCT (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "jconfigint.h"
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+#include "neon-compat.h"
+
+#include <arm_neon.h>
+
+
+#define CONST_BITS  13
+#define PASS1_BITS  2
+
+#define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2  (CONST_BITS + PASS1_BITS + 3)
+
+/* The computation of the inverse DCT requires the use of constants known at
+ * compile time.  Scaled integer constants are used to avoid floating-point
+ * arithmetic:
+ *    0.298631336 =  2446 * 2^-13
+ *    0.390180644 =  3196 * 2^-13
+ *    0.541196100 =  4433 * 2^-13
+ *    0.765366865 =  6270 * 2^-13
+ *    0.899976223 =  7373 * 2^-13
+ *    1.175875602 =  9633 * 2^-13
+ *    1.501321110 = 12299 * 2^-13
+ *    1.847759065 = 15137 * 2^-13
+ *    1.961570560 = 16069 * 2^-13
+ *    2.053119869 = 16819 * 2^-13
+ *    2.562915447 = 20995 * 2^-13
+ *    3.072711026 = 25172 * 2^-13
+ */
+
+#define F_0_298  2446
+#define F_0_390  3196
+#define F_0_541  4433
+#define F_0_765  6270
+#define F_0_899  7373
+#define F_1_175  9633
+#define F_1_501  12299
+#define F_1_847  15137
+#define F_1_961  16069
+#define F_2_053  16819
+#define F_2_562  20995
+#define F_3_072  25172
+
+#define F_1_175_MINUS_1_961  (F_1_175 - F_1_961)
+#define F_1_175_MINUS_0_390  (F_1_175 - F_0_390)
+#define F_0_541_MINUS_1_847  (F_0_541 - F_1_847)
+#define F_3_072_MINUS_2_562  (F_3_072 - F_2_562)
+#define F_0_298_MINUS_0_899  (F_0_298 - F_0_899)
+#define F_1_501_MINUS_0_899  (F_1_501 - F_0_899)
+#define F_2_053_MINUS_2_562  (F_2_053 - F_2_562)
+#define F_0_541_PLUS_0_765   (F_0_541 + F_0_765)
+
+
+ALIGN(16) static const int16_t jsimd_idct_islow_neon_consts = {
+  F_0_899,             F_0_541,
+  F_2_562,             F_0_298_MINUS_0_899,
+  F_1_501_MINUS_0_899, F_2_053_MINUS_2_562,
+  F_0_541_PLUS_0_765,  F_1_175,
+  F_1_175_MINUS_0_390, F_0_541_MINUS_1_847,
+  F_3_072_MINUS_2_562, F_1_175_MINUS_1_961,
+  0, 0, 0, 0
+};
+
+
+/* Forward declaration of regular and sparse IDCT helper functions */
+
+static INLINE void jsimd_idct_islow_pass1_regular(int16x4_t row0,
+                                                  int16x4_t row1,
+                                                  int16x4_t row2,
+                                                  int16x4_t row3,
+                                                  int16x4_t row4,
+                                                  int16x4_t row5,
+                                                  int16x4_t row6,
+                                                  int16x4_t row7,
+                                                  int16x4_t quant_row0,
+                                                  int16x4_t quant_row1,
+                                                  int16x4_t quant_row2,
+                                                  int16x4_t quant_row3,
+                                                  int16x4_t quant_row4,
+                                                  int16x4_t quant_row5,
+                                                  int16x4_t quant_row6,
+                                                  int16x4_t quant_row7,
+                                                  int16_t *workspace_1,
+                                                  int16_t *workspace_2);
+
+static INLINE void jsimd_idct_islow_pass1_sparse(int16x4_t row0,
+                                                 int16x4_t row1,
+                                                 int16x4_t row2,
+                                                 int16x4_t row3,
+                                                 int16x4_t quant_row0,
+                                                 int16x4_t quant_row1,
+                                                 int16x4_t quant_row2,
+                                                 int16x4_t quant_row3,
+                                                 int16_t *workspace_1,
+                                                 int16_t *workspace_2);
+
+static INLINE void jsimd_idct_islow_pass2_regular(int16_t *workspace,
+                                                  JSAMPARRAY output_buf,
+                                                  JDIMENSION output_col,
+                                                  unsigned buf_offset);
+
+static INLINE void jsimd_idct_islow_pass2_sparse(int16_t *workspace,
+                                                 JSAMPARRAY output_buf,
+                                                 JDIMENSION output_col,
+                                                 unsigned buf_offset);
+
+
+/* Perform dequantization and inverse DCT on one block of coefficients.  For
+ * reference, the C implementation (jpeg_idct_slow()) can be found in
+ * jidctint.c.
+ *
+ * Optimization techniques used for fast data access:
+ *
+ * In each pass, the inverse DCT is computed for the left and right 4x8 halves
+ * of the DCT block.  This avoids spilling due to register pressure, and the
+ * increased granularity allows for an optimized calculation depending on the
+ * values of the DCT coefficients.  Between passes, intermediate data is stored
+ * in 4x8 workspace buffers.
+ *
+ * Transposing the 8x8 DCT block after each pass can be achieved by transposing
+ * each of the four 4x4 quadrants and swapping quadrants 1 and 2 (refer to the
+ * diagram below.)  Swapping quadrants is cheap, since the second pass can just
+ * swap the workspace buffer pointers.
+ *
+ *      +-------+-------+                   +-------+-------+
+ *      |       |       |                   |       |       |
+ *      |   0   |   1   |                   |   0   |   2   |
+ *      |       |       |    transpose      |       |       |
+ *      +-------+-------+     ------>       +-------+-------+
+ *      |       |       |                   |       |       |
+ *      |   2   |   3   |                   |   1   |   3   |
+ *      |       |       |                   |       |       |
+ *      +-------+-------+                   +-------+-------+
+ *
+ * Optimization techniques used to accelerate the inverse DCT calculation:
+ *
+ * In a DCT coefficient block, the coefficients are increasingly likely to be 0
+ * as you move diagonally from top left to bottom right.  If whole rows of
+ * coefficients are 0, then the inverse DCT calculation can be simplified.  On
+ * the first pass of the inverse DCT, we test for three special cases before
+ * defaulting to a full "regular" inverse DCT:
+ *
+ * 1) Coefficients in rows 4-7 are all zero.  In this case, we perform a
+ *    "sparse" simplified inverse DCT on rows 0-3.
+ * 2) AC coefficients (rows 1-7) are all zero.  In this case, the inverse DCT
+ *    result is equal to the dequantized DC coefficients.
+ * 3) AC and DC coefficients are all zero.  In this case, the inverse DCT
+ *    result is all zero.  For the left 4x8 half, this is handled identically
+ *    to Case 2 above.  For the right 4x8 half, we do no work and signal that
+ *    the "sparse" algorithm is required for the second pass.
+ *
+ * In the second pass, only a single special case is tested: whether the AC and
+ * DC coefficients were all zero in the right 4x8 block during the first pass
+ * (refer to Case 3 above.)  If this is the case, then a "sparse" variant of
+ * the second pass is performed for both the left and right halves of the DCT
+ * block.  (The transposition after the first pass means that the right 4x8
+ * block during the first pass becomes rows 4-7 during the second pass.)
+ */
+
+void jsimd_idct_islow_neon(void *dct_table, JCOEFPTR coef_block,
+                           JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  ISLOW_MULT_TYPE *quantptr = dct_table;
+
+  int16_t workspace_l8 * DCTSIZE / 2;
+  int16_t workspace_r8 * DCTSIZE / 2;
+
+  /* Compute IDCT first pass on left 4x8 coefficient block. */
+
+  /* Load DCT coefficients in left 4x8 block. */
+  int16x4_t row0 = vld1_s16(coef_block + 0 * DCTSIZE);
+  int16x4_t row1 = vld1_s16(coef_block + 1 * DCTSIZE);
+  int16x4_t row2 = vld1_s16(coef_block + 2 * DCTSIZE);
+  int16x4_t row3 = vld1_s16(coef_block + 3 * DCTSIZE);
+  int16x4_t row4 = vld1_s16(coef_block + 4 * DCTSIZE);
+  int16x4_t row5 = vld1_s16(coef_block + 5 * DCTSIZE);
+  int16x4_t row6 = vld1_s16(coef_block + 6 * DCTSIZE);
+  int16x4_t row7 = vld1_s16(coef_block + 7 * DCTSIZE);
+
+  /* Load quantization table for left 4x8 block. */
+  int16x4_t quant_row0 = vld1_s16(quantptr + 0 * DCTSIZE);
+  int16x4_t quant_row1 = vld1_s16(quantptr + 1 * DCTSIZE);
+  int16x4_t quant_row2 = vld1_s16(quantptr + 2 * DCTSIZE);
+  int16x4_t quant_row3 = vld1_s16(quantptr + 3 * DCTSIZE);
+  int16x4_t quant_row4 = vld1_s16(quantptr + 4 * DCTSIZE);
+  int16x4_t quant_row5 = vld1_s16(quantptr + 5 * DCTSIZE);
+  int16x4_t quant_row6 = vld1_s16(quantptr + 6 * DCTSIZE);
+  int16x4_t quant_row7 = vld1_s16(quantptr + 7 * DCTSIZE);
+
+  /* Construct bitmap to test if DCT coefficients in left 4x8 block are 0. */
+  int16x4_t bitmap = vorr_s16(row7, row6);
+  bitmap = vorr_s16(bitmap, row5);
+  bitmap = vorr_s16(bitmap, row4);
+  int64_t bitmap_rows_4567 = vget_lane_s64(vreinterpret_s64_s16(bitmap), 0);
+
+  if (bitmap_rows_4567 == 0) {
+    bitmap = vorr_s16(bitmap, row3);
+    bitmap = vorr_s16(bitmap, row2);
+    bitmap = vorr_s16(bitmap, row1);
+    int64_t left_ac_bitmap = vget_lane_s64(vreinterpret_s64_s16(bitmap), 0);
+
+    if (left_ac_bitmap == 0) {
+      int16x4_t dcval = vshl_n_s16(vmul_s16(row0, quant_row0), PASS1_BITS);
+      int16x4x4_t quadrant = { { dcval, dcval, dcval, dcval } };
+      /* Store 4x4 blocks to workspace, transposing in the process. */
+      vst4_s16(workspace_l, quadrant);
+      vst4_s16(workspace_r, quadrant);
+    } else {
+      jsimd_idct_islow_pass1_sparse(row0, row1, row2, row3, quant_row0,
+                                    quant_row1, quant_row2, quant_row3,
+                                    workspace_l, workspace_r);
+    }
+  } else {
+    jsimd_idct_islow_pass1_regular(row0, row1, row2, row3, row4, row5,
+                                   row6, row7, quant_row0, quant_row1,
+                                   quant_row2, quant_row3, quant_row4,
+                                   quant_row5, quant_row6, quant_row7,
+                                   workspace_l, workspace_r);
+  }
+
+  /* Compute IDCT first pass on right 4x8 coefficient block. */
+
+  /* Load DCT coefficients in right 4x8 block. */
+  row0 = vld1_s16(coef_block + 0 * DCTSIZE + 4);
+  row1 = vld1_s16(coef_block + 1 * DCTSIZE + 4);
+  row2 = vld1_s16(coef_block + 2 * DCTSIZE + 4);
+  row3 = vld1_s16(coef_block + 3 * DCTSIZE + 4);
+  row4 = vld1_s16(coef_block + 4 * DCTSIZE + 4);
+  row5 = vld1_s16(coef_block + 5 * DCTSIZE + 4);
+  row6 = vld1_s16(coef_block + 6 * DCTSIZE + 4);
+  row7 = vld1_s16(coef_block + 7 * DCTSIZE + 4);
+
+  /* Load quantization table for right 4x8 block. */
+  quant_row0 = vld1_s16(quantptr + 0 * DCTSIZE + 4);
+  quant_row1 = vld1_s16(quantptr + 1 * DCTSIZE + 4);
+  quant_row2 = vld1_s16(quantptr + 2 * DCTSIZE + 4);
+  quant_row3 = vld1_s16(quantptr + 3 * DCTSIZE + 4);
+  quant_row4 = vld1_s16(quantptr + 4 * DCTSIZE + 4);
+  quant_row5 = vld1_s16(quantptr + 5 * DCTSIZE + 4);
+  quant_row6 = vld1_s16(quantptr + 6 * DCTSIZE + 4);
+  quant_row7 = vld1_s16(quantptr + 7 * DCTSIZE + 4);
+
+  /* Construct bitmap to test if DCT coefficients in right 4x8 block are 0. */
+  bitmap = vorr_s16(row7, row6);
+  bitmap = vorr_s16(bitmap, row5);
+  bitmap = vorr_s16(bitmap, row4);
+  bitmap_rows_4567 = vget_lane_s64(vreinterpret_s64_s16(bitmap), 0);
+  bitmap = vorr_s16(bitmap, row3);
+  bitmap = vorr_s16(bitmap, row2);
+  bitmap = vorr_s16(bitmap, row1);
+  int64_t right_ac_bitmap = vget_lane_s64(vreinterpret_s64_s16(bitmap), 0);
+
+  /* If this remains non-zero, a "regular" second pass will be performed. */
+  int64_t right_ac_dc_bitmap = 1;
+
+  if (right_ac_bitmap == 0) {
+    bitmap = vorr_s16(bitmap, row0);
+    right_ac_dc_bitmap = vget_lane_s64(vreinterpret_s64_s16(bitmap), 0);
+
+    if (right_ac_dc_bitmap != 0) {
+      int16x4_t dcval = vshl_n_s16(vmul_s16(row0, quant_row0), PASS1_BITS);
+      int16x4x4_t quadrant = { { dcval, dcval, dcval, dcval } };
+      /* Store 4x4 blocks to workspace, transposing in the process. */
+      vst4_s16(workspace_l + 4 * DCTSIZE / 2, quadrant);
+      vst4_s16(workspace_r + 4 * DCTSIZE / 2, quadrant);
+    }
+  } else {
+    if (bitmap_rows_4567 == 0) {
+      jsimd_idct_islow_pass1_sparse(row0, row1, row2, row3, quant_row0,
+                                    quant_row1, quant_row2, quant_row3,
+                                    workspace_l + 4 * DCTSIZE / 2,
+                                    workspace_r + 4 * DCTSIZE / 2);
+    } else {
+      jsimd_idct_islow_pass1_regular(row0, row1, row2, row3, row4, row5,
+                                     row6, row7, quant_row0, quant_row1,
+                                     quant_row2, quant_row3, quant_row4,
+                                     quant_row5, quant_row6, quant_row7,
+                                     workspace_l + 4 * DCTSIZE / 2,
+                                     workspace_r + 4 * DCTSIZE / 2);
+    }
+  }
+
+  /* Second pass: compute IDCT on rows in workspace. */
+
+  /* If all coefficients in right 4x8 block are 0, use "sparse" second pass. */
+  if (right_ac_dc_bitmap == 0) {
+    jsimd_idct_islow_pass2_sparse(workspace_l, output_buf, output_col, 0);
+    jsimd_idct_islow_pass2_sparse(workspace_r, output_buf, output_col, 4);
+  } else {
+    jsimd_idct_islow_pass2_regular(workspace_l, output_buf, output_col, 0);
+    jsimd_idct_islow_pass2_regular(workspace_r, output_buf, output_col, 4);
+  }
+}
+
+
+/* Perform dequantization and the first pass of the accurate inverse DCT on a
+ * 4x8 block of coefficients.  (To process the full 8x8 DCT block, this
+ * function-- or some other optimized variant-- needs to be called for both the
+ * left and right 4x8 blocks.)
+ *
+ * This "regular" version assumes that no optimization can be made to the IDCT
+ * calculation, since no useful set of AC coefficients is all 0.
+ *
+ * The original C implementation of the accurate IDCT (jpeg_idct_slow()) can be
+ * found in jidctint.c.  Algorithmic changes made here are documented inline.
+ */
+
+static INLINE void jsimd_idct_islow_pass1_regular(int16x4_t row0,
+                                                  int16x4_t row1,
+                                                  int16x4_t row2,
+                                                  int16x4_t row3,
+                                                  int16x4_t row4,
+                                                  int16x4_t row5,
+                                                  int16x4_t row6,
+                                                  int16x4_t row7,
+                                                  int16x4_t quant_row0,
+                                                  int16x4_t quant_row1,
+                                                  int16x4_t quant_row2,
+                                                  int16x4_t quant_row3,
+                                                  int16x4_t quant_row4,
+                                                  int16x4_t quant_row5,
+                                                  int16x4_t quant_row6,
+                                                  int16x4_t quant_row7,
+                                                  int16_t *workspace_1,
+                                                  int16_t *workspace_2)
+{
+  /* Load constants for IDCT computation. */
+#ifdef HAVE_VLD1_S16_X3
+  const int16x4x3_t consts = vld1_s16_x3(jsimd_idct_islow_neon_consts);
+#else
+  const int16x4_t consts1 = vld1_s16(jsimd_idct_islow_neon_consts);
+  const int16x4_t consts2 = vld1_s16(jsimd_idct_islow_neon_consts + 4);
+  const int16x4_t consts3 = vld1_s16(jsimd_idct_islow_neon_consts + 8);
+  const int16x4x3_t consts = { { consts1, consts2, consts3 } };
+#endif
+
+  /* Even part */
+  int16x4_t z2_s16 = vmul_s16(row2, quant_row2);
+  int16x4_t z3_s16 = vmul_s16(row6, quant_row6);
+
+  int32x4_t tmp2 = vmull_lane_s16(z2_s16, consts.val0, 1);
+  int32x4_t tmp3 = vmull_lane_s16(z2_s16, consts.val1, 2);
+  tmp2 = vmlal_lane_s16(tmp2, z3_s16, consts.val2, 1);
+  tmp3 = vmlal_lane_s16(tmp3, z3_s16, consts.val0, 1);
+
+  z2_s16 = vmul_s16(row0, quant_row0);
+  z3_s16 = vmul_s16(row4, quant_row4);
+
+  int32x4_t tmp0 = vshll_n_s16(vadd_s16(z2_s16, z3_s16), CONST_BITS);
+  int32x4_t tmp1 = vshll_n_s16(vsub_s16(z2_s16, z3_s16), CONST_BITS);
+
+  int32x4_t tmp10 = vaddq_s32(tmp0, tmp3);
+  int32x4_t tmp13 = vsubq_s32(tmp0, tmp3);
+  int32x4_t tmp11 = vaddq_s32(tmp1, tmp2);
+  int32x4_t tmp12 = vsubq_s32(tmp1, tmp2);
+
+  /* Odd part */
+  int16x4_t tmp0_s16 = vmul_s16(row7, quant_row7);
+  int16x4_t tmp1_s16 = vmul_s16(row5, quant_row5);
+  int16x4_t tmp2_s16 = vmul_s16(row3, quant_row3);
+  int16x4_t tmp3_s16 = vmul_s16(row1, quant_row1);
+
+  z3_s16 = vadd_s16(tmp0_s16, tmp2_s16);
+  int16x4_t z4_s16 = vadd_s16(tmp1_s16, tmp3_s16);
+
+  /* Implementation as per jpeg_idct_islow() in jidctint.c:
+   *   z5 = (z3 + z4) * 1.175875602;
+   *   z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+   *   z3 += z5;  z4 += z5;
+   *
+   * This implementation:
+   *   z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+   *   z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+   */
+
+  int32x4_t z3 = vmull_lane_s16(z3_s16, consts.val2, 3);
+  int32x4_t z4 = vmull_lane_s16(z3_s16, consts.val1, 3);
+  z3 = vmlal_lane_s16(z3, z4_s16, consts.val1, 3);
+  z4 = vmlal_lane_s16(z4, z4_s16, consts.val2, 0);
+
+  /* Implementation as per jpeg_idct_islow() in jidctint.c:
+   *   z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+   *   tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+   *   tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+   *   z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+   *   tmp0 += z1 + z3;  tmp1 += z2 + z4;
+   *   tmp2 += z2 + z3;  tmp3 += z1 + z4;
+   *
+   * This implementation:
+   *   tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+   *   tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+   *   tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+   *   tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+   *   tmp0 += z3;  tmp1 += z4;
+   *   tmp2 += z3;  tmp3 += z4;
+   */
+
+  tmp0 = vmull_lane_s16(tmp0_s16, consts.val0, 3);
+  tmp1 = vmull_lane_s16(tmp1_s16, consts.val1, 1);
+  tmp2 = vmull_lane_s16(tmp2_s16, consts.val2, 2);
+  tmp3 = vmull_lane_s16(tmp3_s16, consts.val1, 0);
+
+  tmp0 = vmlsl_lane_s16(tmp0, tmp3_s16, consts.val0, 0);
+  tmp1 = vmlsl_lane_s16(tmp1, tmp2_s16, consts.val0, 2);
+  tmp2 = vmlsl_lane_s16(tmp2, tmp1_s16, consts.val0, 2);
+  tmp3 = vmlsl_lane_s16(tmp3, tmp0_s16, consts.val0, 0);
+
+  tmp0 = vaddq_s32(tmp0, z3);
+  tmp1 = vaddq_s32(tmp1, z4);
+  tmp2 = vaddq_s32(tmp2, z3);
+  tmp3 = vaddq_s32(tmp3, z4);
+
+  /* Final output stage: descale and narrow to 16-bit. */
+  int16x4x4_t rows_0123 = { {
+    vrshrn_n_s32(vaddq_s32(tmp10, tmp3), DESCALE_P1),
+    vrshrn_n_s32(vaddq_s32(tmp11, tmp2), DESCALE_P1),
+    vrshrn_n_s32(vaddq_s32(tmp12, tmp1), DESCALE_P1),
+    vrshrn_n_s32(vaddq_s32(tmp13, tmp0), DESCALE_P1)
+  } };
+  int16x4x4_t rows_4567 = { {
+    vrshrn_n_s32(vsubq_s32(tmp13, tmp0), DESCALE_P1),
+    vrshrn_n_s32(vsubq_s32(tmp12, tmp1), DESCALE_P1),
+    vrshrn_n_s32(vsubq_s32(tmp11, tmp2), DESCALE_P1),
+    vrshrn_n_s32(vsubq_s32(tmp10, tmp3), DESCALE_P1)
+  } };
+
+  /* Store 4x4 blocks to the intermediate workspace, ready for the second pass.
+   * (VST4 transposes the blocks.  We need to operate on rows in the next
+   * pass.)
+   */
+  vst4_s16(workspace_1, rows_0123);
+  vst4_s16(workspace_2, rows_4567);
+}
+
+
+/* Perform dequantization and the first pass of the accurate inverse DCT on a
+ * 4x8 block of coefficients.
+ *
+ * This "sparse" version assumes that the AC coefficients in rows 4-7 are all
+ * 0.  This simplifies the IDCT calculation, accelerating overall performance.
+ */
+
+static INLINE void jsimd_idct_islow_pass1_sparse(int16x4_t row0,
+                                                 int16x4_t row1,
+                                                 int16x4_t row2,
+                                                 int16x4_t row3,
+                                                 int16x4_t quant_row0,
+                                                 int16x4_t quant_row1,
+                                                 int16x4_t quant_row2,
+                                                 int16x4_t quant_row3,
+                                                 int16_t *workspace_1,
+                                                 int16_t *workspace_2)
+{
+  /* Load constants for IDCT computation. */
+#ifdef HAVE_VLD1_S16_X3
+  const int16x4x3_t consts = vld1_s16_x3(jsimd_idct_islow_neon_consts);
+#else
+  const int16x4_t consts1 = vld1_s16(jsimd_idct_islow_neon_consts);
+  const int16x4_t consts2 = vld1_s16(jsimd_idct_islow_neon_consts + 4);
+  const int16x4_t consts3 = vld1_s16(jsimd_idct_islow_neon_consts + 8);
+  const int16x4x3_t consts = { { consts1, consts2, consts3 } };
+#endif
+
+  /* Even part (z3 is all 0) */
+  int16x4_t z2_s16 = vmul_s16(row2, quant_row2);
+
+  int32x4_t tmp2 = vmull_lane_s16(z2_s16, consts.val0, 1);
+  int32x4_t tmp3 = vmull_lane_s16(z2_s16, consts.val1, 2);
+
+  z2_s16 = vmul_s16(row0, quant_row0);
+  int32x4_t tmp0 = vshll_n_s16(z2_s16, CONST_BITS);
+  int32x4_t tmp1 = vshll_n_s16(z2_s16, CONST_BITS);
+
+  int32x4_t tmp10 = vaddq_s32(tmp0, tmp3);
+  int32x4_t tmp13 = vsubq_s32(tmp0, tmp3);
+  int32x4_t tmp11 = vaddq_s32(tmp1, tmp2);
+  int32x4_t tmp12 = vsubq_s32(tmp1, tmp2);
+
+  /* Odd part (tmp0 and tmp1 are both all 0) */
+  int16x4_t tmp2_s16 = vmul_s16(row3, quant_row3);
+  int16x4_t tmp3_s16 = vmul_s16(row1, quant_row1);
+
+  int16x4_t z3_s16 = tmp2_s16;
+  int16x4_t z4_s16 = tmp3_s16;
+
+  int32x4_t z3 = vmull_lane_s16(z3_s16, consts.val2, 3);
+  int32x4_t z4 = vmull_lane_s16(z3_s16, consts.val1, 3);
+  z3 = vmlal_lane_s16(z3, z4_s16, consts.val1, 3);
+  z4 = vmlal_lane_s16(z4, z4_s16, consts.val2, 0);
+
+  tmp0 = vmlsl_lane_s16(z3, tmp3_s16, consts.val0, 0);
+  tmp1 = vmlsl_lane_s16(z4, tmp2_s16, consts.val0, 2);
+  tmp2 = vmlal_lane_s16(z3, tmp2_s16, consts.val2, 2);
+  tmp3 = vmlal_lane_s16(z4, tmp3_s16, consts.val1, 0);
+
+  /* Final output stage: descale and narrow to 16-bit. */
+  int16x4x4_t rows_0123 = { {
+    vrshrn_n_s32(vaddq_s32(tmp10, tmp3), DESCALE_P1),
+    vrshrn_n_s32(vaddq_s32(tmp11, tmp2), DESCALE_P1),
+    vrshrn_n_s32(vaddq_s32(tmp12, tmp1), DESCALE_P1),
+    vrshrn_n_s32(vaddq_s32(tmp13, tmp0), DESCALE_P1)
+  } };
+  int16x4x4_t rows_4567 = { {
+    vrshrn_n_s32(vsubq_s32(tmp13, tmp0), DESCALE_P1),
+    vrshrn_n_s32(vsubq_s32(tmp12, tmp1), DESCALE_P1),
+    vrshrn_n_s32(vsubq_s32(tmp11, tmp2), DESCALE_P1),
+    vrshrn_n_s32(vsubq_s32(tmp10, tmp3), DESCALE_P1)
+  } };
+
+  /* Store 4x4 blocks to the intermediate workspace, ready for the second pass.
+   * (VST4 transposes the blocks.  We need to operate on rows in the next
+   * pass.)
+   */
+  vst4_s16(workspace_1, rows_0123);
+  vst4_s16(workspace_2, rows_4567);
+}
+
+
+/* Perform the second pass of the accurate inverse DCT on a 4x8 block of
+ * coefficients.  (To process the full 8x8 DCT block, this function-- or some
+ * other optimized variant-- needs to be called for both the right and left 4x8
+ * blocks.)
+ *
+ * This "regular" version assumes that no optimization can be made to the IDCT
+ * calculation, since no useful set of coefficient values are all 0 after the
+ * first pass.
+ *
+ * Again, the original C implementation of the accurate IDCT (jpeg_idct_slow())
+ * can be found in jidctint.c.  Algorithmic changes made here are documented
+ * inline.
+ */
+
+static INLINE void jsimd_idct_islow_pass2_regular(int16_t *workspace,
+                                                  JSAMPARRAY output_buf,
+                                                  JDIMENSION output_col,
+                                                  unsigned buf_offset)
+{
+  /* Load constants for IDCT computation. */
+#ifdef HAVE_VLD1_S16_X3
+  const int16x4x3_t consts = vld1_s16_x3(jsimd_idct_islow_neon_consts);
+#else
+  const int16x4_t consts1 = vld1_s16(jsimd_idct_islow_neon_consts);
+  const int16x4_t consts2 = vld1_s16(jsimd_idct_islow_neon_consts + 4);
+  const int16x4_t consts3 = vld1_s16(jsimd_idct_islow_neon_consts + 8);
+  const int16x4x3_t consts = { { consts1, consts2, consts3 } };
+#endif
+
+  /* Even part */
+  int16x4_t z2_s16 = vld1_s16(workspace + 2 * DCTSIZE / 2);
+  int16x4_t z3_s16 = vld1_s16(workspace + 6 * DCTSIZE / 2);
+
+  int32x4_t tmp2 = vmull_lane_s16(z2_s16, consts.val0, 1);
+  int32x4_t tmp3 = vmull_lane_s16(z2_s16, consts.val1, 2);
+  tmp2 = vmlal_lane_s16(tmp2, z3_s16, consts.val2, 1);
+  tmp3 = vmlal_lane_s16(tmp3, z3_s16, consts.val0, 1);
+
+  z2_s16 = vld1_s16(workspace + 0 * DCTSIZE / 2);
+  z3_s16 = vld1_s16(workspace + 4 * DCTSIZE / 2);
+
+  int32x4_t tmp0 = vshll_n_s16(vadd_s16(z2_s16, z3_s16), CONST_BITS);
+  int32x4_t tmp1 = vshll_n_s16(vsub_s16(z2_s16, z3_s16), CONST_BITS);
+
+  int32x4_t tmp10 = vaddq_s32(tmp0, tmp3);
+  int32x4_t tmp13 = vsubq_s32(tmp0, tmp3);
+  int32x4_t tmp11 = vaddq_s32(tmp1, tmp2);
+  int32x4_t tmp12 = vsubq_s32(tmp1, tmp2);
+
+  /* Odd part */
+  int16x4_t tmp0_s16 = vld1_s16(workspace + 7 * DCTSIZE / 2);
+  int16x4_t tmp1_s16 = vld1_s16(workspace + 5 * DCTSIZE / 2);
+  int16x4_t tmp2_s16 = vld1_s16(workspace + 3 * DCTSIZE / 2);
+  int16x4_t tmp3_s16 = vld1_s16(workspace + 1 * DCTSIZE / 2);
+
+  z3_s16 = vadd_s16(tmp0_s16, tmp2_s16);
+  int16x4_t z4_s16 = vadd_s16(tmp1_s16, tmp3_s16);
+
+  /* Implementation as per jpeg_idct_islow() in jidctint.c:
+   *   z5 = (z3 + z4) * 1.175875602;
+   *   z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+   *   z3 += z5;  z4 += z5;
+   *
+   * This implementation:
+   *   z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+   *   z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+   */
+
+  int32x4_t z3 = vmull_lane_s16(z3_s16, consts.val2, 3);
+  int32x4_t z4 = vmull_lane_s16(z3_s16, consts.val1, 3);
+  z3 = vmlal_lane_s16(z3, z4_s16, consts.val1, 3);
+  z4 = vmlal_lane_s16(z4, z4_s16, consts.val2, 0);
+
+  /* Implementation as per jpeg_idct_islow() in jidctint.c:
+   *   z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+   *   tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+   *   tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+   *   z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+   *   tmp0 += z1 + z3;  tmp1 += z2 + z4;
+   *   tmp2 += z2 + z3;  tmp3 += z1 + z4;
+   *
+   * This implementation:
+   *   tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+   *   tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+   *   tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+   *   tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+   *   tmp0 += z3;  tmp1 += z4;
+   *   tmp2 += z3;  tmp3 += z4;
+   */
+
+  tmp0 = vmull_lane_s16(tmp0_s16, consts.val0, 3);
+  tmp1 = vmull_lane_s16(tmp1_s16, consts.val1, 1);
+  tmp2 = vmull_lane_s16(tmp2_s16, consts.val2, 2);
+  tmp3 = vmull_lane_s16(tmp3_s16, consts.val1, 0);
+
+  tmp0 = vmlsl_lane_s16(tmp0, tmp3_s16, consts.val0, 0);
+  tmp1 = vmlsl_lane_s16(tmp1, tmp2_s16, consts.val0, 2);
+  tmp2 = vmlsl_lane_s16(tmp2, tmp1_s16, consts.val0, 2);
+  tmp3 = vmlsl_lane_s16(tmp3, tmp0_s16, consts.val0, 0);
+
+  tmp0 = vaddq_s32(tmp0, z3);
+  tmp1 = vaddq_s32(tmp1, z4);
+  tmp2 = vaddq_s32(tmp2, z3);
+  tmp3 = vaddq_s32(tmp3, z4);
+
+  /* Final output stage: descale and narrow to 16-bit. */
+  int16x8_t cols_02_s16 = vcombine_s16(vaddhn_s32(tmp10, tmp3),
+                                       vaddhn_s32(tmp12, tmp1));
+  int16x8_t cols_13_s16 = vcombine_s16(vaddhn_s32(tmp11, tmp2),
+                                       vaddhn_s32(tmp13, tmp0));
+  int16x8_t cols_46_s16 = vcombine_s16(vsubhn_s32(tmp13, tmp0),
+                                       vsubhn_s32(tmp11, tmp2));
+  int16x8_t cols_57_s16 = vcombine_s16(vsubhn_s32(tmp12, tmp1),
+                                       vsubhn_s32(tmp10, tmp3));
+  /* Descale and narrow to 8-bit. */
+  int8x8_t cols_02_s8 = vqrshrn_n_s16(cols_02_s16, DESCALE_P2 - 16);
+  int8x8_t cols_13_s8 = vqrshrn_n_s16(cols_13_s16, DESCALE_P2 - 16);
+  int8x8_t cols_46_s8 = vqrshrn_n_s16(cols_46_s16, DESCALE_P2 - 16);
+  int8x8_t cols_57_s8 = vqrshrn_n_s16(cols_57_s16, DESCALE_P2 - 16);
+  /* Clamp to range 0-255. */
+  uint8x8_t cols_02_u8 = vadd_u8(vreinterpret_u8_s8(cols_02_s8),
+                                 vdup_n_u8(CENTERJSAMPLE));
+  uint8x8_t cols_13_u8 = vadd_u8(vreinterpret_u8_s8(cols_13_s8),
+                                 vdup_n_u8(CENTERJSAMPLE));
+  uint8x8_t cols_46_u8 = vadd_u8(vreinterpret_u8_s8(cols_46_s8),
+                                 vdup_n_u8(CENTERJSAMPLE));
+  uint8x8_t cols_57_u8 = vadd_u8(vreinterpret_u8_s8(cols_57_s8),
+                                 vdup_n_u8(CENTERJSAMPLE));
+
+  /* Transpose 4x8 block and store to memory.  (Zipping adjacent columns
+   * together allows us to store 16-bit elements.)
+   */
+  uint8x8x2_t cols_01_23 = vzip_u8(cols_02_u8, cols_13_u8);
+  uint8x8x2_t cols_45_67 = vzip_u8(cols_46_u8, cols_57_u8);
+  uint16x4x4_t cols_01_23_45_67 = { {
+    vreinterpret_u16_u8(cols_01_23.val0),
+    vreinterpret_u16_u8(cols_01_23.val1),
+    vreinterpret_u16_u8(cols_45_67.val0),
+    vreinterpret_u16_u8(cols_45_67.val1)
+  } };
+
+  JSAMPROW outptr0 = output_bufbuf_offset + 0 + output_col;
+  JSAMPROW outptr1 = output_bufbuf_offset + 1 + output_col;
+  JSAMPROW outptr2 = output_bufbuf_offset + 2 + output_col;
+  JSAMPROW outptr3 = output_bufbuf_offset + 3 + output_col;
+  /* VST4 of 16-bit elements completes the transpose. */
+  vst4_lane_u16((uint16_t *)outptr0, cols_01_23_45_67, 0);
+  vst4_lane_u16((uint16_t *)outptr1, cols_01_23_45_67, 1);
+  vst4_lane_u16((uint16_t *)outptr2, cols_01_23_45_67, 2);
+  vst4_lane_u16((uint16_t *)outptr3, cols_01_23_45_67, 3);
+}
+
+
+/* Performs the second pass of the accurate inverse DCT on a 4x8 block
+ * of coefficients.
+ *
+ * This "sparse" version assumes that the coefficient values (after the first
+ * pass) in rows 4-7 are all 0.  This simplifies the IDCT calculation,
+ * accelerating overall performance.
+ */
+
+static INLINE void jsimd_idct_islow_pass2_sparse(int16_t *workspace,
+                                                 JSAMPARRAY output_buf,
+                                                 JDIMENSION output_col,
+                                                 unsigned buf_offset)
+{
+  /* Load constants for IDCT computation. */
+#ifdef HAVE_VLD1_S16_X3
+  const int16x4x3_t consts = vld1_s16_x3(jsimd_idct_islow_neon_consts);
+#else
+  const int16x4_t consts1 = vld1_s16(jsimd_idct_islow_neon_consts);
+  const int16x4_t consts2 = vld1_s16(jsimd_idct_islow_neon_consts + 4);
+  const int16x4_t consts3 = vld1_s16(jsimd_idct_islow_neon_consts + 8);
+  const int16x4x3_t consts = { { consts1, consts2, consts3 } };
+#endif
+
+  /* Even part (z3 is all 0) */
+  int16x4_t z2_s16 = vld1_s16(workspace + 2 * DCTSIZE / 2);
+
+  int32x4_t tmp2 = vmull_lane_s16(z2_s16, consts.val0, 1);
+  int32x4_t tmp3 = vmull_lane_s16(z2_s16, consts.val1, 2);
+
+  z2_s16 = vld1_s16(workspace + 0 * DCTSIZE / 2);
+  int32x4_t tmp0 = vshll_n_s16(z2_s16, CONST_BITS);
+  int32x4_t tmp1 = vshll_n_s16(z2_s16, CONST_BITS);
+
+  int32x4_t tmp10 = vaddq_s32(tmp0, tmp3);
+  int32x4_t tmp13 = vsubq_s32(tmp0, tmp3);
+  int32x4_t tmp11 = vaddq_s32(tmp1, tmp2);
+  int32x4_t tmp12 = vsubq_s32(tmp1, tmp2);
+
+  /* Odd part (tmp0 and tmp1 are both all 0) */
+  int16x4_t tmp2_s16 = vld1_s16(workspace + 3 * DCTSIZE / 2);
+  int16x4_t tmp3_s16 = vld1_s16(workspace + 1 * DCTSIZE / 2);
+
+  int16x4_t z3_s16 = tmp2_s16;
+  int16x4_t z4_s16 = tmp3_s16;
+
+  int32x4_t z3 = vmull_lane_s16(z3_s16, consts.val2, 3);
+  z3 = vmlal_lane_s16(z3, z4_s16, consts.val1, 3);
+  int32x4_t z4 = vmull_lane_s16(z3_s16, consts.val1, 3);
+  z4 = vmlal_lane_s16(z4, z4_s16, consts.val2, 0);
+
+  tmp0 = vmlsl_lane_s16(z3, tmp3_s16, consts.val0, 0);
+  tmp1 = vmlsl_lane_s16(z4, tmp2_s16, consts.val0, 2);
+  tmp2 = vmlal_lane_s16(z3, tmp2_s16, consts.val2, 2);
+  tmp3 = vmlal_lane_s16(z4, tmp3_s16, consts.val1, 0);
+
+  /* Final output stage: descale and narrow to 16-bit. */
+  int16x8_t cols_02_s16 = vcombine_s16(vaddhn_s32(tmp10, tmp3),
+                                       vaddhn_s32(tmp12, tmp1));
+  int16x8_t cols_13_s16 = vcombine_s16(vaddhn_s32(tmp11, tmp2),
+                                       vaddhn_s32(tmp13, tmp0));
+  int16x8_t cols_46_s16 = vcombine_s16(vsubhn_s32(tmp13, tmp0),
+                                       vsubhn_s32(tmp11, tmp2));
+  int16x8_t cols_57_s16 = vcombine_s16(vsubhn_s32(tmp12, tmp1),
+                                       vsubhn_s32(tmp10, tmp3));
+  /* Descale and narrow to 8-bit. */
+  int8x8_t cols_02_s8 = vqrshrn_n_s16(cols_02_s16, DESCALE_P2 - 16);
+  int8x8_t cols_13_s8 = vqrshrn_n_s16(cols_13_s16, DESCALE_P2 - 16);
+  int8x8_t cols_46_s8 = vqrshrn_n_s16(cols_46_s16, DESCALE_P2 - 16);
+  int8x8_t cols_57_s8 = vqrshrn_n_s16(cols_57_s16, DESCALE_P2 - 16);
+  /* Clamp to range 0-255. */
+  uint8x8_t cols_02_u8 = vadd_u8(vreinterpret_u8_s8(cols_02_s8),
+                                 vdup_n_u8(CENTERJSAMPLE));
+  uint8x8_t cols_13_u8 = vadd_u8(vreinterpret_u8_s8(cols_13_s8),
+                                 vdup_n_u8(CENTERJSAMPLE));
+  uint8x8_t cols_46_u8 = vadd_u8(vreinterpret_u8_s8(cols_46_s8),
+                                 vdup_n_u8(CENTERJSAMPLE));
+  uint8x8_t cols_57_u8 = vadd_u8(vreinterpret_u8_s8(cols_57_s8),
+                                 vdup_n_u8(CENTERJSAMPLE));
+
+  /* Transpose 4x8 block and store to memory.  (Zipping adjacent columns
+   * together allows us to store 16-bit elements.)
+   */
+  uint8x8x2_t cols_01_23 = vzip_u8(cols_02_u8, cols_13_u8);
+  uint8x8x2_t cols_45_67 = vzip_u8(cols_46_u8, cols_57_u8);
+  uint16x4x4_t cols_01_23_45_67 = { {
+    vreinterpret_u16_u8(cols_01_23.val0),
+    vreinterpret_u16_u8(cols_01_23.val1),
+    vreinterpret_u16_u8(cols_45_67.val0),
+    vreinterpret_u16_u8(cols_45_67.val1)
+  } };
+
+  JSAMPROW outptr0 = output_bufbuf_offset + 0 + output_col;
+  JSAMPROW outptr1 = output_bufbuf_offset + 1 + output_col;
+  JSAMPROW outptr2 = output_bufbuf_offset + 2 + output_col;
+  JSAMPROW outptr3 = output_bufbuf_offset + 3 + output_col;
+  /* VST4 of 16-bit elements completes the transpose. */
+  vst4_lane_u16((uint16_t *)outptr0, cols_01_23_45_67, 0);
+  vst4_lane_u16((uint16_t *)outptr1, cols_01_23_45_67, 1);
+  vst4_lane_u16((uint16_t *)outptr2, cols_01_23_45_67, 2);
+  vst4_lane_u16((uint16_t *)outptr3, cols_01_23_45_67, 3);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jidctred-neon.c Added

@@ -0,0 +1,486 @@
+/*
+ * jidctred-neon.c - reduced-size IDCT (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "align.h"
+#include "neon-compat.h"
+
+#include <arm_neon.h>
+
+
+#define CONST_BITS  13
+#define PASS1_BITS  2
+
+#define F_0_211  1730
+#define F_0_509  4176
+#define F_0_601  4926
+#define F_0_720  5906
+#define F_0_765  6270
+#define F_0_850  6967
+#define F_0_899  7373
+#define F_1_061  8697
+#define F_1_272  10426
+#define F_1_451  11893
+#define F_1_847  15137
+#define F_2_172  17799
+#define F_2_562  20995
+#define F_3_624  29692
+
+
+/* jsimd_idct_2x2_neon() is an inverse DCT function that produces reduced-size
+ * 2x2 output from an 8x8 DCT block.  It uses the same calculations and
+ * produces exactly the same output as IJG's original jpeg_idct_2x2() function
+ * from jpeg-6b, which can be found in jidctred.c.
+ *
+ * Scaled integer constants are used to avoid floating-point arithmetic:
+ *    0.720959822 =  5906 * 2^-13
+ *    0.850430095 =  6967 * 2^-13
+ *    1.272758580 = 10426 * 2^-13
+ *    3.624509785 = 29692 * 2^-13
+ *
+ * See jidctred.c for further details of the 2x2 IDCT algorithm.  Where
+ * possible, the variable names and comments here in jsimd_idct_2x2_neon()
+ * match up with those in jpeg_idct_2x2().
+ */
+
+ALIGN(16) static const int16_t jsimd_idct_2x2_neon_consts = {
+  -F_0_720, F_0_850, -F_1_272, F_3_624
+};
+
+void jsimd_idct_2x2_neon(void *dct_table, JCOEFPTR coef_block,
+                         JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  ISLOW_MULT_TYPE *quantptr = dct_table;
+
+  /* Load DCT coefficients. */
+  int16x8_t row0 = vld1q_s16(coef_block + 0 * DCTSIZE);
+  int16x8_t row1 = vld1q_s16(coef_block + 1 * DCTSIZE);
+  int16x8_t row3 = vld1q_s16(coef_block + 3 * DCTSIZE);
+  int16x8_t row5 = vld1q_s16(coef_block + 5 * DCTSIZE);
+  int16x8_t row7 = vld1q_s16(coef_block + 7 * DCTSIZE);
+
+  /* Load quantization table values. */
+  int16x8_t quant_row0 = vld1q_s16(quantptr + 0 * DCTSIZE);
+  int16x8_t quant_row1 = vld1q_s16(quantptr + 1 * DCTSIZE);
+  int16x8_t quant_row3 = vld1q_s16(quantptr + 3 * DCTSIZE);
+  int16x8_t quant_row5 = vld1q_s16(quantptr + 5 * DCTSIZE);
+  int16x8_t quant_row7 = vld1q_s16(quantptr + 7 * DCTSIZE);
+
+  /* Dequantize DCT coefficients. */
+  row0 = vmulq_s16(row0, quant_row0);
+  row1 = vmulq_s16(row1, quant_row1);
+  row3 = vmulq_s16(row3, quant_row3);
+  row5 = vmulq_s16(row5, quant_row5);
+  row7 = vmulq_s16(row7, quant_row7);
+
+  /* Load IDCT conversion constants. */
+  const int16x4_t consts = vld1_s16(jsimd_idct_2x2_neon_consts);
+
+  /* Pass 1: process columns from input, put results in vectors row0 and
+   * row1.
+   */
+
+  /* Even part */
+  int32x4_t tmp10_l = vshll_n_s16(vget_low_s16(row0), CONST_BITS + 2);
+  int32x4_t tmp10_h = vshll_n_s16(vget_high_s16(row0), CONST_BITS + 2);
+
+  /* Odd part */
+  int32x4_t tmp0_l = vmull_lane_s16(vget_low_s16(row1), consts, 3);
+  tmp0_l = vmlal_lane_s16(tmp0_l, vget_low_s16(row3), consts, 2);
+  tmp0_l = vmlal_lane_s16(tmp0_l, vget_low_s16(row5), consts, 1);
+  tmp0_l = vmlal_lane_s16(tmp0_l, vget_low_s16(row7), consts, 0);
+  int32x4_t tmp0_h = vmull_lane_s16(vget_high_s16(row1), consts, 3);
+  tmp0_h = vmlal_lane_s16(tmp0_h, vget_high_s16(row3), consts, 2);
+  tmp0_h = vmlal_lane_s16(tmp0_h, vget_high_s16(row5), consts, 1);
+  tmp0_h = vmlal_lane_s16(tmp0_h, vget_high_s16(row7), consts, 0);
+
+  /* Final output stage: descale and narrow to 16-bit. */
+  row0 = vcombine_s16(vrshrn_n_s32(vaddq_s32(tmp10_l, tmp0_l), CONST_BITS),
+                      vrshrn_n_s32(vaddq_s32(tmp10_h, tmp0_h), CONST_BITS));
+  row1 = vcombine_s16(vrshrn_n_s32(vsubq_s32(tmp10_l, tmp0_l), CONST_BITS),
+                      vrshrn_n_s32(vsubq_s32(tmp10_h, tmp0_h), CONST_BITS));
+
+  /* Transpose two rows, ready for second pass. */
+  int16x8x2_t cols_0246_1357 = vtrnq_s16(row0, row1);
+  int16x8_t cols_0246 = cols_0246_1357.val0;
+  int16x8_t cols_1357 = cols_0246_1357.val1;
+  /* Duplicate columns such that each is accessible in its own vector. */
+  int32x4x2_t cols_1155_3377 = vtrnq_s32(vreinterpretq_s32_s16(cols_1357),
+                                         vreinterpretq_s32_s16(cols_1357));
+  int16x8_t cols_1155 = vreinterpretq_s16_s32(cols_1155_3377.val0);
+  int16x8_t cols_3377 = vreinterpretq_s16_s32(cols_1155_3377.val1);
+
+  /* Pass 2: process two rows, store to output array. */
+
+  /* Even part: we're only interested in col0; the top half of tmp10 is "don't
+   * care."
+   */
+  int32x4_t tmp10 = vshll_n_s16(vget_low_s16(cols_0246), CONST_BITS + 2);
+
+  /* Odd part: we're only interested in the bottom half of tmp0. */
+  int32x4_t tmp0 = vmull_lane_s16(vget_low_s16(cols_1155), consts, 3);
+  tmp0 = vmlal_lane_s16(tmp0, vget_low_s16(cols_3377), consts, 2);
+  tmp0 = vmlal_lane_s16(tmp0, vget_high_s16(cols_1155), consts, 1);
+  tmp0 = vmlal_lane_s16(tmp0, vget_high_s16(cols_3377), consts, 0);
+
+  /* Final output stage: descale and clamp to range 0-255. */
+  int16x8_t output_s16 = vcombine_s16(vaddhn_s32(tmp10, tmp0),
+                                      vsubhn_s32(tmp10, tmp0));
+  output_s16 = vrsraq_n_s16(vdupq_n_s16(CENTERJSAMPLE), output_s16,
+                            CONST_BITS + PASS1_BITS + 3 + 2 - 16);
+  /* Narrow to 8-bit and convert to unsigned. */
+  uint8x8_t output_u8 = vqmovun_s16(output_s16);
+
+  /* Store 2x2 block to memory. */
+  vst1_lane_u8(output_buf0 + output_col, output_u8, 0);
+  vst1_lane_u8(output_buf1 + output_col, output_u8, 1);
+  vst1_lane_u8(output_buf0 + output_col + 1, output_u8, 4);
+  vst1_lane_u8(output_buf1 + output_col + 1, output_u8, 5);
+}
+
+
+/* jsimd_idct_4x4_neon() is an inverse DCT function that produces reduced-size
+ * 4x4 output from an 8x8 DCT block.  It uses the same calculations and
+ * produces exactly the same output as IJG's original jpeg_idct_4x4() function
+ * from jpeg-6b, which can be found in jidctred.c.
+ *
+ * Scaled integer constants are used to avoid floating-point arithmetic:
+ *    0.211164243 =  1730 * 2^-13
+ *    0.509795579 =  4176 * 2^-13
+ *    0.601344887 =  4926 * 2^-13
+ *    0.765366865 =  6270 * 2^-13
+ *    0.899976223 =  7373 * 2^-13
+ *    1.061594337 =  8697 * 2^-13
+ *    1.451774981 = 11893 * 2^-13
+ *    1.847759065 = 15137 * 2^-13
+ *    2.172734803 = 17799 * 2^-13
+ *    2.562915447 = 20995 * 2^-13
+ *
+ * See jidctred.c for further details of the 4x4 IDCT algorithm.  Where
+ * possible, the variable names and comments here in jsimd_idct_4x4_neon()
+ * match up with those in jpeg_idct_4x4().
+ */
+
+ALIGN(16) static const int16_t jsimd_idct_4x4_neon_consts = {
+  F_1_847, -F_0_765, -F_0_211,  F_1_451,
+ -F_2_172,  F_1_061, -F_0_509, -F_0_601,
+  F_0_899,  F_2_562,        0,        0
+};
+
+void jsimd_idct_4x4_neon(void *dct_table, JCOEFPTR coef_block,
+                         JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  ISLOW_MULT_TYPE *quantptr = dct_table;
+
+  /* Load DCT coefficients. */
+  int16x8_t row0  = vld1q_s16(coef_block + 0 * DCTSIZE);
+  int16x8_t row1  = vld1q_s16(coef_block + 1 * DCTSIZE);
+  int16x8_t row2  = vld1q_s16(coef_block + 2 * DCTSIZE);
+  int16x8_t row3  = vld1q_s16(coef_block + 3 * DCTSIZE);
+  int16x8_t row5  = vld1q_s16(coef_block + 5 * DCTSIZE);
+  int16x8_t row6  = vld1q_s16(coef_block + 6 * DCTSIZE);
+  int16x8_t row7  = vld1q_s16(coef_block + 7 * DCTSIZE);
+
+  /* Load quantization table values for DC coefficients. */
+  int16x8_t quant_row0 = vld1q_s16(quantptr + 0 * DCTSIZE);
+  /* Dequantize DC coefficients. */
+  row0 = vmulq_s16(row0, quant_row0);
+
+  /* Construct bitmap to test if all AC coefficients are 0. */
+  int16x8_t bitmap = vorrq_s16(row1, row2);
+  bitmap = vorrq_s16(bitmap, row3);
+  bitmap = vorrq_s16(bitmap, row5);
+  bitmap = vorrq_s16(bitmap, row6);
+  bitmap = vorrq_s16(bitmap, row7);
+
+  int64_t left_ac_bitmap = vgetq_lane_s64(vreinterpretq_s64_s16(bitmap), 0);
+  int64_t right_ac_bitmap = vgetq_lane_s64(vreinterpretq_s64_s16(bitmap), 1);
+
+  /* Load constants for IDCT computation. */
+#ifdef HAVE_VLD1_S16_X3
+  const int16x4x3_t consts = vld1_s16_x3(jsimd_idct_4x4_neon_consts);
+#else
+  /* GCC does not currently support the intrinsic vld1_<type>_x3(). */
+  const int16x4_t consts1 = vld1_s16(jsimd_idct_4x4_neon_consts);
+  const int16x4_t consts2 = vld1_s16(jsimd_idct_4x4_neon_consts + 4);
+  const int16x4_t consts3 = vld1_s16(jsimd_idct_4x4_neon_consts + 8);
+  const int16x4x3_t consts = { { consts1, consts2, consts3 } };
+#endif
+
+  if (left_ac_bitmap == 0 && right_ac_bitmap == 0) {
+    /* All AC coefficients are zero.
+     * Compute DC values and duplicate into row vectors 0, 1, 2, and 3.
+     */
+    int16x8_t dcval = vshlq_n_s16(row0, PASS1_BITS);
+    row0 = dcval;
+    row1 = dcval;
+    row2 = dcval;
+    row3 = dcval;
+  } else if (left_ac_bitmap == 0) {
+    /* AC coefficients are zero for columns 0, 1, 2, and 3.
+     * Compute DC values for these columns.
+     */
+    int16x4_t dcval = vshl_n_s16(vget_low_s16(row0), PASS1_BITS);
+
+    /* Commence regular IDCT computation for columns 4, 5, 6, and 7. */
+
+    /* Load quantization table. */
+    int16x4_t quant_row1 = vld1_s16(quantptr + 1 * DCTSIZE + 4);
+    int16x4_t quant_row2 = vld1_s16(quantptr + 2 * DCTSIZE + 4);
+    int16x4_t quant_row3 = vld1_s16(quantptr + 3 * DCTSIZE + 4);
+    int16x4_t quant_row5 = vld1_s16(quantptr + 5 * DCTSIZE + 4);
+    int16x4_t quant_row6 = vld1_s16(quantptr + 6 * DCTSIZE + 4);
+    int16x4_t quant_row7 = vld1_s16(quantptr + 7 * DCTSIZE + 4);
+
+    /* Even part */
+    int32x4_t tmp0 = vshll_n_s16(vget_high_s16(row0), CONST_BITS + 1);
+
+    int16x4_t z2 = vmul_s16(vget_high_s16(row2), quant_row2);
+    int16x4_t z3 = vmul_s16(vget_high_s16(row6), quant_row6);
+
+    int32x4_t tmp2 = vmull_lane_s16(z2, consts.val0, 0);
+    tmp2 = vmlal_lane_s16(tmp2, z3, consts.val0, 1);
+
+    int32x4_t tmp10 = vaddq_s32(tmp0, tmp2);
+    int32x4_t tmp12 = vsubq_s32(tmp0, tmp2);
+
+    /* Odd part */
+    int16x4_t z1 = vmul_s16(vget_high_s16(row7), quant_row7);
+    z2 = vmul_s16(vget_high_s16(row5), quant_row5);
+    z3 = vmul_s16(vget_high_s16(row3), quant_row3);
+    int16x4_t z4 = vmul_s16(vget_high_s16(row1), quant_row1);
+
+    tmp0 = vmull_lane_s16(z1, consts.val0, 2);
+    tmp0 = vmlal_lane_s16(tmp0, z2, consts.val0, 3);
+    tmp0 = vmlal_lane_s16(tmp0, z3, consts.val1, 0);
+    tmp0 = vmlal_lane_s16(tmp0, z4, consts.val1, 1);
+
+    tmp2 = vmull_lane_s16(z1, consts.val1, 2);
+    tmp2 = vmlal_lane_s16(tmp2, z2, consts.val1, 3);
+    tmp2 = vmlal_lane_s16(tmp2, z3, consts.val2, 0);
+    tmp2 = vmlal_lane_s16(tmp2, z4, consts.val2, 1);
+
+    /* Final output stage: descale and narrow to 16-bit. */
+    row0 = vcombine_s16(dcval, vrshrn_n_s32(vaddq_s32(tmp10, tmp2),
+                                            CONST_BITS - PASS1_BITS + 1));
+    row3 = vcombine_s16(dcval, vrshrn_n_s32(vsubq_s32(tmp10, tmp2),
+                                            CONST_BITS - PASS1_BITS + 1));
+    row1 = vcombine_s16(dcval, vrshrn_n_s32(vaddq_s32(tmp12, tmp0),
+                                            CONST_BITS - PASS1_BITS + 1));
+    row2 = vcombine_s16(dcval, vrshrn_n_s32(vsubq_s32(tmp12, tmp0),
+                                            CONST_BITS - PASS1_BITS + 1));
+  } else if (right_ac_bitmap == 0) {
+    /* AC coefficients are zero for columns 4, 5, 6, and 7.
+     * Compute DC values for these columns.
+     */
+    int16x4_t dcval = vshl_n_s16(vget_high_s16(row0), PASS1_BITS);
+
+    /* Commence regular IDCT computation for columns 0, 1, 2, and 3. */
+
+    /* Load quantization table. */
+    int16x4_t quant_row1 = vld1_s16(quantptr + 1 * DCTSIZE);
+    int16x4_t quant_row2 = vld1_s16(quantptr + 2 * DCTSIZE);
+    int16x4_t quant_row3 = vld1_s16(quantptr + 3 * DCTSIZE);
+    int16x4_t quant_row5 = vld1_s16(quantptr + 5 * DCTSIZE);
+    int16x4_t quant_row6 = vld1_s16(quantptr + 6 * DCTSIZE);
+    int16x4_t quant_row7 = vld1_s16(quantptr + 7 * DCTSIZE);
+
+    /* Even part */
+    int32x4_t tmp0 = vshll_n_s16(vget_low_s16(row0), CONST_BITS + 1);
+
+    int16x4_t z2 = vmul_s16(vget_low_s16(row2), quant_row2);
+    int16x4_t z3 = vmul_s16(vget_low_s16(row6), quant_row6);
+
+    int32x4_t tmp2 = vmull_lane_s16(z2, consts.val0, 0);
+    tmp2 = vmlal_lane_s16(tmp2, z3, consts.val0, 1);
+
+    int32x4_t tmp10 = vaddq_s32(tmp0, tmp2);
+    int32x4_t tmp12 = vsubq_s32(tmp0, tmp2);
+
+    /* Odd part */
+    int16x4_t z1 = vmul_s16(vget_low_s16(row7), quant_row7);
+    z2 = vmul_s16(vget_low_s16(row5), quant_row5);
+    z3 = vmul_s16(vget_low_s16(row3), quant_row3);
+    int16x4_t z4 = vmul_s16(vget_low_s16(row1), quant_row1);
+
+    tmp0 = vmull_lane_s16(z1, consts.val0, 2);
+    tmp0 = vmlal_lane_s16(tmp0, z2, consts.val0, 3);
+    tmp0 = vmlal_lane_s16(tmp0, z3, consts.val1, 0);
+    tmp0 = vmlal_lane_s16(tmp0, z4, consts.val1, 1);
+
+    tmp2 = vmull_lane_s16(z1, consts.val1, 2);
+    tmp2 = vmlal_lane_s16(tmp2, z2, consts.val1, 3);
+    tmp2 = vmlal_lane_s16(tmp2, z3, consts.val2, 0);
+    tmp2 = vmlal_lane_s16(tmp2, z4, consts.val2, 1);
+
+    /* Final output stage: descale and narrow to 16-bit. */
+    row0 = vcombine_s16(vrshrn_n_s32(vaddq_s32(tmp10, tmp2),
+                                     CONST_BITS - PASS1_BITS + 1), dcval);
+    row3 = vcombine_s16(vrshrn_n_s32(vsubq_s32(tmp10, tmp2),
+                                     CONST_BITS - PASS1_BITS + 1), dcval);
+    row1 = vcombine_s16(vrshrn_n_s32(vaddq_s32(tmp12, tmp0),
+                                     CONST_BITS - PASS1_BITS + 1), dcval);
+    row2 = vcombine_s16(vrshrn_n_s32(vsubq_s32(tmp12, tmp0),
+                                     CONST_BITS - PASS1_BITS + 1), dcval);
+  } else {
+    /* All AC coefficients are non-zero; full IDCT calculation required. */
+    int16x8_t quant_row1 = vld1q_s16(quantptr + 1 * DCTSIZE);
+    int16x8_t quant_row2 = vld1q_s16(quantptr + 2 * DCTSIZE);
+    int16x8_t quant_row3 = vld1q_s16(quantptr + 3 * DCTSIZE);
+    int16x8_t quant_row5 = vld1q_s16(quantptr + 5 * DCTSIZE);
+    int16x8_t quant_row6 = vld1q_s16(quantptr + 6 * DCTSIZE);
+    int16x8_t quant_row7 = vld1q_s16(quantptr + 7 * DCTSIZE);
+
+    /* Even part */
+    int32x4_t tmp0_l = vshll_n_s16(vget_low_s16(row0), CONST_BITS + 1);
+    int32x4_t tmp0_h = vshll_n_s16(vget_high_s16(row0), CONST_BITS + 1);
+
+    int16x8_t z2 = vmulq_s16(row2, quant_row2);
+    int16x8_t z3 = vmulq_s16(row6, quant_row6);
+
+    int32x4_t tmp2_l = vmull_lane_s16(vget_low_s16(z2), consts.val0, 0);
+    int32x4_t tmp2_h = vmull_lane_s16(vget_high_s16(z2), consts.val0, 0);
+    tmp2_l = vmlal_lane_s16(tmp2_l, vget_low_s16(z3), consts.val0, 1);
+    tmp2_h = vmlal_lane_s16(tmp2_h, vget_high_s16(z3), consts.val0, 1);
+
+    int32x4_t tmp10_l = vaddq_s32(tmp0_l, tmp2_l);
+    int32x4_t tmp10_h = vaddq_s32(tmp0_h, tmp2_h);
+    int32x4_t tmp12_l = vsubq_s32(tmp0_l, tmp2_l);
+    int32x4_t tmp12_h = vsubq_s32(tmp0_h, tmp2_h);
+
+    /* Odd part */
+    int16x8_t z1 = vmulq_s16(row7, quant_row7);
+    z2 = vmulq_s16(row5, quant_row5);
+    z3 = vmulq_s16(row3, quant_row3);
+    int16x8_t z4 = vmulq_s16(row1, quant_row1);
+
+    tmp0_l = vmull_lane_s16(vget_low_s16(z1), consts.val0, 2);
+    tmp0_l = vmlal_lane_s16(tmp0_l, vget_low_s16(z2), consts.val0, 3);
+    tmp0_l = vmlal_lane_s16(tmp0_l, vget_low_s16(z3), consts.val1, 0);
+    tmp0_l = vmlal_lane_s16(tmp0_l, vget_low_s16(z4), consts.val1, 1);
+    tmp0_h = vmull_lane_s16(vget_high_s16(z1), consts.val0, 2);
+    tmp0_h = vmlal_lane_s16(tmp0_h, vget_high_s16(z2), consts.val0, 3);
+    tmp0_h = vmlal_lane_s16(tmp0_h, vget_high_s16(z3), consts.val1, 0);
+    tmp0_h = vmlal_lane_s16(tmp0_h, vget_high_s16(z4), consts.val1, 1);
+
+    tmp2_l = vmull_lane_s16(vget_low_s16(z1), consts.val1, 2);
+    tmp2_l = vmlal_lane_s16(tmp2_l, vget_low_s16(z2), consts.val1, 3);
+    tmp2_l = vmlal_lane_s16(tmp2_l, vget_low_s16(z3), consts.val2, 0);
+    tmp2_l = vmlal_lane_s16(tmp2_l, vget_low_s16(z4), consts.val2, 1);
+    tmp2_h = vmull_lane_s16(vget_high_s16(z1), consts.val1, 2);
+    tmp2_h = vmlal_lane_s16(tmp2_h, vget_high_s16(z2), consts.val1, 3);
+    tmp2_h = vmlal_lane_s16(tmp2_h, vget_high_s16(z3), consts.val2, 0);
+    tmp2_h = vmlal_lane_s16(tmp2_h, vget_high_s16(z4), consts.val2, 1);
+
+    /* Final output stage: descale and narrow to 16-bit. */
+    row0 = vcombine_s16(vrshrn_n_s32(vaddq_s32(tmp10_l, tmp2_l),
+                                     CONST_BITS - PASS1_BITS + 1),
+                        vrshrn_n_s32(vaddq_s32(tmp10_h, tmp2_h),
+                                     CONST_BITS - PASS1_BITS + 1));
+    row3 = vcombine_s16(vrshrn_n_s32(vsubq_s32(tmp10_l, tmp2_l),
+                                     CONST_BITS - PASS1_BITS + 1),
+                        vrshrn_n_s32(vsubq_s32(tmp10_h, tmp2_h),
+                                     CONST_BITS - PASS1_BITS + 1));
+    row1 = vcombine_s16(vrshrn_n_s32(vaddq_s32(tmp12_l, tmp0_l),
+                                     CONST_BITS - PASS1_BITS + 1),
+                        vrshrn_n_s32(vaddq_s32(tmp12_h, tmp0_h),
+                                     CONST_BITS - PASS1_BITS + 1));
+    row2 = vcombine_s16(vrshrn_n_s32(vsubq_s32(tmp12_l, tmp0_l),
+                                     CONST_BITS - PASS1_BITS + 1),
+                        vrshrn_n_s32(vsubq_s32(tmp12_h, tmp0_h),
+                                     CONST_BITS - PASS1_BITS + 1));
+  }
+
+  /* Transpose 8x4 block to perform IDCT on rows in second pass. */
+  int16x8x2_t row_01 = vtrnq_s16(row0, row1);
+  int16x8x2_t row_23 = vtrnq_s16(row2, row3);
+
+  int32x4x2_t cols_0426 = vtrnq_s32(vreinterpretq_s32_s16(row_01.val0),
+                                    vreinterpretq_s32_s16(row_23.val0));
+  int32x4x2_t cols_1537 = vtrnq_s32(vreinterpretq_s32_s16(row_01.val1),
+                                    vreinterpretq_s32_s16(row_23.val1));
+
+  int16x4_t col0 = vreinterpret_s16_s32(vget_low_s32(cols_0426.val0));
+  int16x4_t col1 = vreinterpret_s16_s32(vget_low_s32(cols_1537.val0));
+  int16x4_t col2 = vreinterpret_s16_s32(vget_low_s32(cols_0426.val1));
+  int16x4_t col3 = vreinterpret_s16_s32(vget_low_s32(cols_1537.val1));
+  int16x4_t col5 = vreinterpret_s16_s32(vget_high_s32(cols_1537.val0));
+  int16x4_t col6 = vreinterpret_s16_s32(vget_high_s32(cols_0426.val1));
+  int16x4_t col7 = vreinterpret_s16_s32(vget_high_s32(cols_1537.val1));
+
+  /* Commence second pass of IDCT. */
+
+  /* Even part */
+  int32x4_t tmp0 = vshll_n_s16(col0, CONST_BITS + 1);
+  int32x4_t tmp2 = vmull_lane_s16(col2, consts.val0, 0);
+  tmp2 = vmlal_lane_s16(tmp2, col6, consts.val0, 1);
+
+  int32x4_t tmp10 = vaddq_s32(tmp0, tmp2);
+  int32x4_t tmp12 = vsubq_s32(tmp0, tmp2);
+
+  /* Odd part */
+  tmp0 = vmull_lane_s16(col7, consts.val0, 2);
+  tmp0 = vmlal_lane_s16(tmp0, col5, consts.val0, 3);
+  tmp0 = vmlal_lane_s16(tmp0, col3, consts.val1, 0);
+  tmp0 = vmlal_lane_s16(tmp0, col1, consts.val1, 1);
+
+  tmp2 = vmull_lane_s16(col7, consts.val1, 2);
+  tmp2 = vmlal_lane_s16(tmp2, col5, consts.val1, 3);
+  tmp2 = vmlal_lane_s16(tmp2, col3, consts.val2, 0);
+  tmp2 = vmlal_lane_s16(tmp2, col1, consts.val2, 1);
+
+  /* Final output stage: descale and clamp to range 0-255. */
+  int16x8_t output_cols_02 = vcombine_s16(vaddhn_s32(tmp10, tmp2),
+                                          vsubhn_s32(tmp12, tmp0));
+  int16x8_t output_cols_13 = vcombine_s16(vaddhn_s32(tmp12, tmp0),
+                                          vsubhn_s32(tmp10, tmp2));
+  output_cols_02 = vrsraq_n_s16(vdupq_n_s16(CENTERJSAMPLE), output_cols_02,
+                                CONST_BITS + PASS1_BITS + 3 + 1 - 16);
+  output_cols_13 = vrsraq_n_s16(vdupq_n_s16(CENTERJSAMPLE), output_cols_13,
+                                CONST_BITS + PASS1_BITS + 3 + 1 - 16);
+  /* Narrow to 8-bit and convert to unsigned while zipping 8-bit elements.
+   * An interleaving store completes the transpose.
+   */
+  uint8x8x2_t output_0123 = vzip_u8(vqmovun_s16(output_cols_02),
+                                    vqmovun_s16(output_cols_13));
+  uint16x4x2_t output_01_23 = { {
+    vreinterpret_u16_u8(output_0123.val0),
+    vreinterpret_u16_u8(output_0123.val1)
+  } };
+
+  /* Store 4x4 block to memory. */
+  JSAMPROW outptr0 = output_buf0 + output_col;
+  JSAMPROW outptr1 = output_buf1 + output_col;
+  JSAMPROW outptr2 = output_buf2 + output_col;
+  JSAMPROW outptr3 = output_buf3 + output_col;
+  vst2_lane_u16((uint16_t *)outptr0, output_01_23, 0);
+  vst2_lane_u16((uint16_t *)outptr1, output_01_23, 1);
+  vst2_lane_u16((uint16_t *)outptr2, output_01_23, 2);
+  vst2_lane_u16((uint16_t *)outptr3, output_01_23, 3);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/jquanti-neon.c Added

@@ -0,0 +1,190 @@
+/*
+ * jquanti-neon.c - sample data conversion and quantization (Arm Neon)
+ *
+ * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+
+#include <arm_neon.h>
+
+
+/* After downsampling, the resulting sample values are in the range 0, 255,
+ * but the Discrete Cosine Transform (DCT) operates on values centered around
+ * 0.
+ *
+ * To prepare sample values for the DCT, load samples into a DCT workspace,
+ * subtracting CENTERJSAMPLE (128).  The samples, now in the range -128, 127,
+ * are also widened from 8- to 16-bit.
+ *
+ * The equivalent scalar C function convsamp() can be found in jcdctmgr.c.
+ */
+
+void jsimd_convsamp_neon(JSAMPARRAY sample_data, JDIMENSION start_col,
+                         DCTELEM *workspace)
+{
+  uint8x8_t samp_row0 = vld1_u8(sample_data0 + start_col);
+  uint8x8_t samp_row1 = vld1_u8(sample_data1 + start_col);
+  uint8x8_t samp_row2 = vld1_u8(sample_data2 + start_col);
+  uint8x8_t samp_row3 = vld1_u8(sample_data3 + start_col);
+  uint8x8_t samp_row4 = vld1_u8(sample_data4 + start_col);
+  uint8x8_t samp_row5 = vld1_u8(sample_data5 + start_col);
+  uint8x8_t samp_row6 = vld1_u8(sample_data6 + start_col);
+  uint8x8_t samp_row7 = vld1_u8(sample_data7 + start_col);
+
+  int16x8_t row0 =
+    vreinterpretq_s16_u16(vsubl_u8(samp_row0, vdup_n_u8(CENTERJSAMPLE)));
+  int16x8_t row1 =
+    vreinterpretq_s16_u16(vsubl_u8(samp_row1, vdup_n_u8(CENTERJSAMPLE)));
+  int16x8_t row2 =
+    vreinterpretq_s16_u16(vsubl_u8(samp_row2, vdup_n_u8(CENTERJSAMPLE)));
+  int16x8_t row3 =
+    vreinterpretq_s16_u16(vsubl_u8(samp_row3, vdup_n_u8(CENTERJSAMPLE)));
+  int16x8_t row4 =
+    vreinterpretq_s16_u16(vsubl_u8(samp_row4, vdup_n_u8(CENTERJSAMPLE)));
+  int16x8_t row5 =
+    vreinterpretq_s16_u16(vsubl_u8(samp_row5, vdup_n_u8(CENTERJSAMPLE)));
+  int16x8_t row6 =
+    vreinterpretq_s16_u16(vsubl_u8(samp_row6, vdup_n_u8(CENTERJSAMPLE)));
+  int16x8_t row7 =
+    vreinterpretq_s16_u16(vsubl_u8(samp_row7, vdup_n_u8(CENTERJSAMPLE)));
+
+  vst1q_s16(workspace + 0 * DCTSIZE, row0);
+  vst1q_s16(workspace + 1 * DCTSIZE, row1);
+  vst1q_s16(workspace + 2 * DCTSIZE, row2);
+  vst1q_s16(workspace + 3 * DCTSIZE, row3);
+  vst1q_s16(workspace + 4 * DCTSIZE, row4);
+  vst1q_s16(workspace + 5 * DCTSIZE, row5);
+  vst1q_s16(workspace + 6 * DCTSIZE, row6);
+  vst1q_s16(workspace + 7 * DCTSIZE, row7);
+}
+
+
+/* After the DCT, the resulting array of coefficient values needs to be divided
+ * by an array of quantization values.
+ *
+ * To avoid a slow division operation, the DCT coefficients are multiplied by
+ * the (scaled) reciprocals of the quantization values and then right-shifted.
+ *
+ * The equivalent scalar C function quantize() can be found in jcdctmgr.c.
+ */
+
+void jsimd_quantize_neon(JCOEFPTR coef_block, DCTELEM *divisors,
+                         DCTELEM *workspace)
+{
+  JCOEFPTR out_ptr = coef_block;
+  UDCTELEM *recip_ptr = (UDCTELEM *)divisors;
+  UDCTELEM *corr_ptr = (UDCTELEM *)divisors + DCTSIZE2;
+  DCTELEM *shift_ptr = divisors + 3 * DCTSIZE2;
+  int i;
+
+  for (i = 0; i < DCTSIZE; i += DCTSIZE / 2) {
+    /* Load DCT coefficients. */
+    int16x8_t row0 = vld1q_s16(workspace + (i + 0) * DCTSIZE);
+    int16x8_t row1 = vld1q_s16(workspace + (i + 1) * DCTSIZE);
+    int16x8_t row2 = vld1q_s16(workspace + (i + 2) * DCTSIZE);
+    int16x8_t row3 = vld1q_s16(workspace + (i + 3) * DCTSIZE);
+    /* Load reciprocals of quantization values. */
+    uint16x8_t recip0 = vld1q_u16(recip_ptr + (i + 0) * DCTSIZE);
+    uint16x8_t recip1 = vld1q_u16(recip_ptr + (i + 1) * DCTSIZE);
+    uint16x8_t recip2 = vld1q_u16(recip_ptr + (i + 2) * DCTSIZE);
+    uint16x8_t recip3 = vld1q_u16(recip_ptr + (i + 3) * DCTSIZE);
+    uint16x8_t corr0 = vld1q_u16(corr_ptr + (i + 0) * DCTSIZE);
+    uint16x8_t corr1 = vld1q_u16(corr_ptr + (i + 1) * DCTSIZE);
+    uint16x8_t corr2 = vld1q_u16(corr_ptr + (i + 2) * DCTSIZE);
+    uint16x8_t corr3 = vld1q_u16(corr_ptr + (i + 3) * DCTSIZE);
+    int16x8_t shift0 = vld1q_s16(shift_ptr + (i + 0) * DCTSIZE);
+    int16x8_t shift1 = vld1q_s16(shift_ptr + (i + 1) * DCTSIZE);
+    int16x8_t shift2 = vld1q_s16(shift_ptr + (i + 2) * DCTSIZE);
+    int16x8_t shift3 = vld1q_s16(shift_ptr + (i + 3) * DCTSIZE);
+
+    /* Extract sign from coefficients. */
+    int16x8_t sign_row0 = vshrq_n_s16(row0, 15);
+    int16x8_t sign_row1 = vshrq_n_s16(row1, 15);
+    int16x8_t sign_row2 = vshrq_n_s16(row2, 15);
+    int16x8_t sign_row3 = vshrq_n_s16(row3, 15);
+    /* Get absolute value of DCT coefficients. */
+    uint16x8_t abs_row0 = vreinterpretq_u16_s16(vabsq_s16(row0));
+    uint16x8_t abs_row1 = vreinterpretq_u16_s16(vabsq_s16(row1));
+    uint16x8_t abs_row2 = vreinterpretq_u16_s16(vabsq_s16(row2));
+    uint16x8_t abs_row3 = vreinterpretq_u16_s16(vabsq_s16(row3));
+    /* Add correction. */
+    abs_row0 = vaddq_u16(abs_row0, corr0);
+    abs_row1 = vaddq_u16(abs_row1, corr1);
+    abs_row2 = vaddq_u16(abs_row2, corr2);
+    abs_row3 = vaddq_u16(abs_row3, corr3);
+
+    /* Multiply DCT coefficients by quantization reciprocals. */
+    int32x4_t row0_l = vreinterpretq_s32_u32(vmull_u16(vget_low_u16(abs_row0),
+                                                       vget_low_u16(recip0)));
+    int32x4_t row0_h = vreinterpretq_s32_u32(vmull_u16(vget_high_u16(abs_row0),
+                                                       vget_high_u16(recip0)));
+    int32x4_t row1_l = vreinterpretq_s32_u32(vmull_u16(vget_low_u16(abs_row1),
+                                                       vget_low_u16(recip1)));
+    int32x4_t row1_h = vreinterpretq_s32_u32(vmull_u16(vget_high_u16(abs_row1),
+                                                       vget_high_u16(recip1)));
+    int32x4_t row2_l = vreinterpretq_s32_u32(vmull_u16(vget_low_u16(abs_row2),
+                                                       vget_low_u16(recip2)));
+    int32x4_t row2_h = vreinterpretq_s32_u32(vmull_u16(vget_high_u16(abs_row2),
+                                                       vget_high_u16(recip2)));
+    int32x4_t row3_l = vreinterpretq_s32_u32(vmull_u16(vget_low_u16(abs_row3),
+                                                       vget_low_u16(recip3)));
+    int32x4_t row3_h = vreinterpretq_s32_u32(vmull_u16(vget_high_u16(abs_row3),
+                                                       vget_high_u16(recip3)));
+    /* Narrow back to 16-bit. */
+    row0 = vcombine_s16(vshrn_n_s32(row0_l, 16), vshrn_n_s32(row0_h, 16));
+    row1 = vcombine_s16(vshrn_n_s32(row1_l, 16), vshrn_n_s32(row1_h, 16));
+    row2 = vcombine_s16(vshrn_n_s32(row2_l, 16), vshrn_n_s32(row2_h, 16));
+    row3 = vcombine_s16(vshrn_n_s32(row3_l, 16), vshrn_n_s32(row3_h, 16));
+
+    /* Since VSHR only supports an immediate as its second argument, negate the
+     * shift value and shift left.
+     */
+    row0 = vreinterpretq_s16_u16(vshlq_u16(vreinterpretq_u16_s16(row0),
+                                           vnegq_s16(shift0)));
+    row1 = vreinterpretq_s16_u16(vshlq_u16(vreinterpretq_u16_s16(row1),
+                                           vnegq_s16(shift1)));
+    row2 = vreinterpretq_s16_u16(vshlq_u16(vreinterpretq_u16_s16(row2),
+                                           vnegq_s16(shift2)));
+    row3 = vreinterpretq_s16_u16(vshlq_u16(vreinterpretq_u16_s16(row3),
+                                           vnegq_s16(shift3)));
+
+    /* Restore sign to original product. */
+    row0 = veorq_s16(row0, sign_row0);
+    row0 = vsubq_s16(row0, sign_row0);
+    row1 = veorq_s16(row1, sign_row1);
+    row1 = vsubq_s16(row1, sign_row1);
+    row2 = veorq_s16(row2, sign_row2);
+    row2 = vsubq_s16(row2, sign_row2);
+    row3 = veorq_s16(row3, sign_row3);
+    row3 = vsubq_s16(row3, sign_row3);
+
+    /* Store quantized coefficients to memory. */
+    vst1q_s16(out_ptr + (i + 0) * DCTSIZE, row0);
+    vst1q_s16(out_ptr + (i + 1) * DCTSIZE, row1);
+    vst1q_s16(out_ptr + (i + 2) * DCTSIZE, row2);
+    vst1q_s16(out_ptr + (i + 3) * DCTSIZE, row3);
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/arm/neon-compat.h.in Added

@@ -0,0 +1,37 @@
+/*
+ * Copyright (C) 2020, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2020-2021, Arm Limited.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#cmakedefine HAVE_VLD1_S16_X3
+#cmakedefine HAVE_VLD1_U16_X2
+#cmakedefine HAVE_VLD1Q_U8_X4
+
+/* Define compiler-independent count-leading-zeros and byte-swap macros */
+#if defined(_MSC_VER) && !defined(__clang__)
+#define BUILTIN_CLZ(x)  _CountLeadingZeros(x)
+#define BUILTIN_CLZLL(x)  _CountLeadingZeros64(x)
+#define BUILTIN_BSWAP64(x)  _byteswap_uint64(x)
+#elif defined(__clang__) || defined(__GNUC__)
+#define BUILTIN_CLZ(x)  __builtin_clz(x)
+#define BUILTIN_CLZLL(x)  __builtin_clzll(x)
+#define BUILTIN_BSWAP64(x)  __builtin_bswap64(x)
+#else
+#error "Unknown compiler"
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386 Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jccolext-avx2.asm Added

@@ -0,0 +1,578 @@
+;
+; jccolext.asm - colorspace conversion (AVX2)
+;
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_avx2(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                            JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                            int num_rows);
+;
+
+%define img_width(b)   (b) + 8          ; JDIMENSION img_width
+%define input_buf(b)   (b) + 12         ; JSAMPARRAY input_buf
+%define output_buf(b)  (b) + 16         ; JSAMPIMAGE output_buf
+%define output_row(b)  (b) + 20         ; JDIMENSION output_row
+%define num_rows(b)    (b) + 24         ; int num_rows
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_YMMWORD
+                                        ; ymmword wkWK_NUM
+%define WK_NUM         8
+%define gotptr         wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_ycc_convert_avx2)
+
+EXTN(jsimd_rgb_ycc_convert_avx2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION img_width(eax)
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         esi, JSAMPIMAGE output_buf(eax)
+    mov         ecx, JDIMENSION output_row(eax)
+    mov         edi, JSAMPARRAY esi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY esi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY esi+2*SIZEOF_JSAMPARRAY
+    lea         edi, edi+ecx*SIZEOF_JSAMPROW
+    lea         ebx, ebx+ecx*SIZEOF_JSAMPROW
+    lea         edx, edx+ecx*SIZEOF_JSAMPROW
+
+    pop         ecx
+
+    mov         esi, JSAMPARRAY input_buf(eax)
+    mov         eax, INT num_rows(eax)
+    test        eax, eax
+    jle         near .return
+    alignx      16, 7
+.rowloop:
+    pushpic     eax
+    push        edx
+    push        ebx
+    push        edi
+    push        esi
+    push        ecx                     ; col
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr0
+    mov         ebx, JSAMPROW ebx     ; outptr1
+    mov         edx, JSAMPROW edx     ; outptr2
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+    alignx      16, 7
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        eax
+    push        edx
+    lea         ecx, ecx+ecx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_BYTE
+    movzx       eax, byte esi+ecx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_WORD
+    movzx       edx, word esi+ecx
+    shl         eax, WORD_BIT
+    or          eax, edx
+.column_ld4:
+    vmovd       xmmA, eax
+    pop         edx
+    pop         eax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_DWORD
+    vmovd       xmmF, XMM_DWORD esi+ecx
+    vpslldq     xmmA, xmmA, SIZEOF_DWORD
+    vpor        xmmA, xmmA, xmmF
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    sub         ecx, byte SIZEOF_MMWORD
+    vmovq       xmmB, XMM_MMWORD esi+ecx
+    vpslldq     xmmA, xmmA, SIZEOF_MMWORD
+    vpor        xmmA, xmmA, xmmB
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    jz          short .column_ld32
+    sub         ecx, byte SIZEOF_XMMWORD
+    vmovdqu     xmmB, XMM_MMWORD esi+ecx
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    vpor        ymmA, ymmB
+.column_ld32:
+    test        cl, SIZEOF_YMMWORD
+    jz          short .column_ld64
+    sub         ecx, byte SIZEOF_YMMWORD
+    vmovdqa     ymmF, ymmA
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+.column_ld64:
+    test        cl, 2*SIZEOF_YMMWORD
+    mov         ecx, SIZEOF_YMMWORD
+    jz          short .rgb_ycc_cnv
+    vmovdqa     ymmB, ymmA
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD esi+1*SIZEOF_YMMWORD
+    jmp         short .rgb_ycc_cnv
+    alignx      16, 7
+
+.columnloop:
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD esi+1*SIZEOF_YMMWORD
+    vmovdqu     ymmB, YMMWORD esi+2*SIZEOF_YMMWORD
+
+.rgb_ycc_cnv:
+    ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+    ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    ; ymmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+    ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    ; ymmB=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+    ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    vmovdqu     ymmC, ymmA
+    vinserti128 ymmA, ymmF, xmmA, 0  ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                     ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vinserti128 ymmC, ymmC, xmmB, 0  ; ymmC=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+                                     ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    vinserti128 ymmB, ymmB, xmmF, 0  ; ymmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                     ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+    vperm2i128  ymmF, ymmC, ymmC, 1  ; ymmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A
+                                     ;       1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q)
+
+    vmovdqa     ymmG, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12
+                                  ;       22 03 13 23 04 14 24 05 0G 1G 2G 0H 1H 2H 0I 1I)
+    vpsrldq     ymmG, ymmG, 8     ; ymmG=(22 03 13 23 04 14 24 05 0G 1G 2G 0H 1H 2H 0I 1I
+                                  ;       2I 0J 1J 2J 0K 1K 2K 0L -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmF  ; ymmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A
+                                  ;       0G 0O 1G 1O 2G 2O 0H 0P 1H 1P 2H 2P 0I 0Q 1I 1Q)
+    vpslldq     ymmF, ymmF, 8     ; ymmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27
+                                  ;       08 18 28 09 19 29 0A 1A 1L 2L 0M 1M 2M 0N 1N 2N)
+
+    vpunpcklbw  ymmG, ymmG, ymmB  ; ymmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D
+                                  ;       2I 2Q 0J 0R 1J 1R 2J 2R 0K 0S 1K 1S 2K 2S 0L 0T)
+    vpunpckhbw  ymmF, ymmF, ymmB  ; ymmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F
+                                  ;       1L 1T 2L 2T 0M 0U 1M 1U 2M 2U 0N 0V 1N 1V 2N 2V)
+
+    vmovdqa     ymmD, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09
+                                  ;       11 19 21 29 02 0A 12 1A 0G 0O 1G 1O 2G 2O 0H 0P)
+    vpsrldq     ymmD, ymmD, 8     ; ymmD=(11 19 21 29 02 0A 12 1A 0G 0O 1G 1O 2G 2O 0H 0P
+                                  ;       1H 1P 2H 2P 0I 0Q 1I 1Q -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmG  ; ymmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D
+                                  ;       0G 0K 0O 0S 1G 1K 1O 1S 2G 2K 2O 2S 0H 0L 0P 0T)
+    vpslldq     ymmG, ymmG, 8     ; ymmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B
+                                  ;       04 0C 14 1C 24 2C 05 0D 2I 2Q 0J 0R 1J 1R 2J 2R)
+
+    vpunpcklbw  ymmD, ymmD, ymmF  ; ymmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E
+                                  ;       1H 1L 1P 1T 2H 2L 2P 2T 0I 0M 0Q 0U 1I 1M 1Q 1U)
+    vpunpckhbw  ymmG, ymmG, ymmF  ; ymmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F
+                                  ;       2I 2M 2Q 2U 0J 0N 0R 0V 1J 1N 1R 1V 2J 2N 2R 2V)
+
+    vmovdqa     ymmE, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C
+                                  ;       20 24 28 2C 01 05 09 0D 0G 0K 0O 0S 1G 1K 1O 1S)
+    vpsrldq     ymmE, ymmE, 8     ; ymmE=(20 24 28 2C 01 05 09 0D 0G 0K 0O 0S 1G 1K 1O 1S
+                                  ;       2G 2K 2O 2S 0H 0L 0P 0T -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmD  ; ymmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E
+                                  ;       0G 0I 0K 0M 0O 0Q 0S 0U 1G 1I 1K 1M 1O 1Q 1S 1U)
+    vpslldq     ymmD, ymmD, 8     ; ymmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D
+                                  ;       02 06 0A 0E 12 16 1A 1E 1H 1L 1P 1T 2H 2L 2P 2T)
+
+    vpunpcklbw  ymmE, ymmE, ymmG  ; ymmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F
+                                  ;       2G 2I 2K 2M 2O 2Q 2S 2U 0H 0J 0L 0N 0P 0R 0T 0V)
+    vpunpckhbw  ymmD, ymmD, ymmG  ; ymmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F
+                                  ;       1H 1J 1L 1N 1P 1R 1T 1V 2H 2J 2L 2N 2P 2R 2T 2V)
+
+    vpxor       ymmH, ymmH, ymmH
+
+    vmovdqa     ymmC, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmH  ; ymmA=(00 02 04 06 08 0A 0C 0E 0G 0I 0K 0M 0O 0Q 0S 0U)
+    vpunpckhbw  ymmC, ymmC, ymmH  ; ymmC=(10 12 14 16 18 1A 1C 1E 1G 1I 1K 1M 1O 1Q 1S 1U)
+
+    vmovdqa     ymmB, ymmE
+    vpunpcklbw  ymmE, ymmE, ymmH  ; ymmE=(20 22 24 26 28 2A 2C 2E 2G 2I 2K 2M 2O 2Q 2S 2U)
+    vpunpckhbw  ymmB, ymmB, ymmH  ; ymmB=(01 03 05 07 09 0B 0D 0F 0H 0J 0L 0N 0P 0R 0T 0V)
+
+    vmovdqa     ymmF, ymmD
+    vpunpcklbw  ymmD, ymmD, ymmH  ; ymmD=(11 13 15 17 19 1B 1D 1F 1H 1J 1L 1N 1P 1R 1T 1V)
+    vpunpckhbw  ymmF, ymmF, ymmH  ; ymmF=(21 23 25 27 29 2B 2D 2F 2H 2J 2L 2N 2P 2R 2T 2V)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_XMMWORD/16
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_XMMWORD/16
+    vmovd       xmmA, XMM_DWORD esi+ecx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_XMMWORD/8
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_XMMWORD/8
+    vmovq       xmmF, XMM_MMWORD esi+ecx*RGB_PIXELSIZE
+    vpslldq     xmmA, xmmA, SIZEOF_MMWORD
+    vpor        xmmA, xmmA, xmmF
+.column_ld4:
+    test        cl, SIZEOF_XMMWORD/4
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_XMMWORD/4
+    vmovdqa     xmmF, xmmA
+    vperm2i128  ymmF, ymmF, ymmF, 1
+    vmovdqu     xmmA, XMMWORD esi+ecx*RGB_PIXELSIZE
+    vpor        ymmA, ymmA, ymmF
+.column_ld8:
+    test        cl, SIZEOF_XMMWORD/2
+    jz          short .column_ld16
+    sub         ecx, byte SIZEOF_XMMWORD/2
+    vmovdqa     ymmF, ymmA
+    vmovdqu     ymmA, YMMWORD esi+ecx*RGB_PIXELSIZE
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    mov         ecx, SIZEOF_YMMWORD
+    jz          short .rgb_ycc_cnv
+    vmovdqa     ymmE, ymmA
+    vmovdqa     ymmH, ymmF
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD esi+1*SIZEOF_YMMWORD
+    jmp         short .rgb_ycc_cnv
+    alignx      16, 7
+
+.columnloop:
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD esi+1*SIZEOF_YMMWORD
+    vmovdqu     ymmE, YMMWORD esi+2*SIZEOF_YMMWORD
+    vmovdqu     ymmH, YMMWORD esi+3*SIZEOF_YMMWORD
+
+.rgb_ycc_cnv:
+    ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+    ;       04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    ; ymmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+    ;       0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+    ; ymmE=(0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J
+    ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    ; ymmH=(0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R
+    ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    vmovdqa     ymmB, ymmA
+    vinserti128 ymmA, ymmA, xmmE, 1     ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J)
+    vperm2i128  ymmE, ymmB, ymmE, 0x31  ; ymmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+
+    vmovdqa     ymmB, ymmF
+    vinserti128 ymmF, ymmF, xmmH, 1     ; ymmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R)
+    vperm2i128  ymmH, ymmB, ymmH, 0x31  ; ymmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    vmovdqa     ymmD, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmE      ; ymmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35
+                                      ;       0G 0K 1G 1K 2G 2K 3G 3K 0H 0L 1H 1L 2H 2L 3H 3L)
+    vpunpckhbw  ymmD, ymmD, ymmE      ; ymmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37
+                                      ;       0I 0M 1I 1M 2I 2M 3I 3M 0J 0N 1J 1N 2J 2N 3J 3N)
+
+    vmovdqa     ymmC, ymmF
+    vpunpcklbw  ymmF, ymmF, ymmH      ; ymmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D
+                                      ;       0O 0S 1O 1S 2O 2S 3O 3S 0P 0T 1P 1T 2P 2T 3P 3T)
+    vpunpckhbw  ymmC, ymmC, ymmH      ; ymmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F
+                                      ;       0Q 0U 1Q 1U 2Q 2U 3Q 3U 0R 0V 1R 1V 2R 2V 3R 3V)
+
+    vmovdqa     ymmB, ymmA
+    vpunpcklwd  ymmA, ymmA, ymmF      ; ymmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C
+                                      ;       0G 0K 0O 0S 1G 1K 1O 1S 2G 2K 2O 2S 3G 3K 3O 3S)
+    vpunpckhwd  ymmB, ymmB, ymmF      ; ymmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D
+                                      ;       0H 0L 0P 0T 1H 1L 1P 1T 2H 2L 2P 2T 3H 3L 3P 3T)
+
+    vmovdqa     ymmG, ymmD
+    vpunpcklwd  ymmD, ymmD, ymmC      ; ymmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E
+                                      ;       0I 0M 0Q 0U 1I 1M 1Q 1U 2I 2M 2Q 2U 3I 3M 3Q 3U)
+    vpunpckhwd  ymmG, ymmG, ymmC      ; ymmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F
+                                      ;       0J 0N 0R 0V 1J 1N 1R 1V 2J 2N 2R 2V 3J 3N 3R 3V)
+
+    vmovdqa     ymmE, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmD      ; ymmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E
+                                      ;       0G 0I 0K 0M 0O 0Q 0S 0U 1G 1I 1K 1M 1O 1Q 1S 1U)
+    vpunpckhbw  ymmE, ymmE, ymmD      ; ymmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E
+                                      ;       2G 2I 2K 2M 2O 2Q 2S 2U 3G 3I 3K 3M 3O 3Q 3S 3U)
+
+    vmovdqa     ymmH, ymmB
+    vpunpcklbw  ymmB, ymmB, ymmG      ; ymmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F
+                                      ;       0H 0J 0L 0N 0P 0R 0T 0V 1H 1J 1L 1N 1P 1R 1T 1V)
+    vpunpckhbw  ymmH, ymmH, ymmG      ; ymmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F
+                                      ;       2H 2J 2L 2N 2P 2R 2T 2V 3H 3J 3L 3N 3P 3R 3T 3V)
+
+    vpxor       ymmF, ymmF, ymmF
+
+    vmovdqa     ymmC, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmF      ; ymmA=(00 02 04 06 08 0A 0C 0E 0G 0I 0K 0M 0O 0Q 0S 0U)
+    vpunpckhbw  ymmC, ymmC, ymmF      ; ymmC=(10 12 14 16 18 1A 1C 1E 1G 1I 1K 1M 1O 1Q 1S 1U)
+
+    vmovdqa     ymmD, ymmB
+    vpunpcklbw  ymmB, ymmB, ymmF      ; ymmB=(01 03 05 07 09 0B 0D 0F 0H 0J 0L 0N 0P 0R 0T 0V)
+    vpunpckhbw  ymmD, ymmD, ymmF      ; ymmD=(11 13 15 17 19 1B 1D 1F 1H 1J 1L 1N 1P 1R 1T 1V)
+
+    vmovdqa     ymmG, ymmE
+    vpunpcklbw  ymmE, ymmE, ymmF      ; ymmE=(20 22 24 26 28 2A 2C 2E 2G 2I 2K 2M 2O 2Q 2S 2U)
+    vpunpckhbw  ymmG, ymmG, ymmF      ; ymmG=(30 32 34 36 38 3A 3C 3E 3G 3I 3K 3M 3O 3Q 3S 3U)
+
+    vpunpcklbw  ymmF, ymmF, ymmH
+    vpunpckhbw  ymmH, ymmH, ymmH
+    vpsrlw      ymmF, ymmF, BYTE_BIT  ; ymmF=(21 23 25 27 29 2B 2D 2F 2H 2J 2L 2N 2P 2R 2T 2V)
+    vpsrlw      ymmH, ymmH, BYTE_BIT  ; ymmH=(31 33 35 37 39 3B 3D 3F 3H 3J 3L 3N 3P 3R 3T 3V)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; ymm0=R(02468ACEGIKMOQSU)=RE, ymm2=G(02468ACEGIKMOQSU)=GE, ymm4=B(02468ACEGIKMOQSU)=BE
+    ; ymm1=R(13579BDFHJLNPRTV)=RO, ymm3=G(13579BDFHJLNPRTV)=GO, ymm5=B(13579BDFHJLNPRTV)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+    vmovdqa     YMMWORD wk(0), ymm0   ; wk(0)=RE
+    vmovdqa     YMMWORD wk(1), ymm1   ; wk(1)=RO
+    vmovdqa     YMMWORD wk(2), ymm4   ; wk(2)=BE
+    vmovdqa     YMMWORD wk(3), ymm5   ; wk(3)=BO
+
+    vmovdqa     ymm6, ymm1
+    vpunpcklwd  ymm1, ymm1, ymm3
+    vpunpckhwd  ymm6, ymm6, ymm3
+    vmovdqa     ymm7, ymm1
+    vmovdqa     ymm4, ymm6
+    vpmaddwd    ymm1, ymm1, GOTOFF(eax,PW_F0299_F0337)  ; ymm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    vpmaddwd    ymm6, ymm6, GOTOFF(eax,PW_F0299_F0337)  ; ymm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+    vpmaddwd    ymm7, ymm7, GOTOFF(eax,PW_MF016_MF033)  ; ymm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+    vpmaddwd    ymm4, ymm4, GOTOFF(eax,PW_MF016_MF033)  ; ymm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+    vmovdqa     YMMWORD wk(4), ymm1   ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+    vmovdqa     YMMWORD wk(5), ymm6   ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    vpxor       ymm1, ymm1, ymm1
+    vpxor       ymm6, ymm6, ymm6
+    vpunpcklwd  ymm1, ymm1, ymm5        ; ymm1=BOL
+    vpunpckhwd  ymm6, ymm6, ymm5        ; ymm6=BOH
+    vpsrld      ymm1, ymm1, 1           ; ymm1=BOL*FIX(0.500)
+    vpsrld      ymm6, ymm6, 1           ; ymm6=BOH*FIX(0.500)
+
+    vmovdqa     ymm5, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; ymm5=PD_ONEHALFM1_CJ
+
+    vpaddd      ymm7, ymm7, ymm1
+    vpaddd      ymm4, ymm4, ymm6
+    vpaddd      ymm7, ymm7, ymm5
+    vpaddd      ymm4, ymm4, ymm5
+    vpsrld      ymm7, ymm7, SCALEBITS   ; ymm7=CbOL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=CbOH
+    vpackssdw   ymm7, ymm7, ymm4        ; ymm7=CbO
+
+    vmovdqa     ymm1, YMMWORD wk(2)   ; ymm1=BE
+
+    vmovdqa     ymm6, ymm0
+    vpunpcklwd  ymm0, ymm0, ymm2
+    vpunpckhwd  ymm6, ymm6, ymm2
+    vmovdqa     ymm5, ymm0
+    vmovdqa     ymm4, ymm6
+    vpmaddwd    ymm0, ymm0, GOTOFF(eax,PW_F0299_F0337)  ; ymm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    vpmaddwd    ymm6, ymm6, GOTOFF(eax,PW_F0299_F0337)  ; ymm6=REH*FIX(0.299)+GEH*FIX(0.337)
+    vpmaddwd    ymm5, ymm5, GOTOFF(eax,PW_MF016_MF033)  ; ymm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+    vpmaddwd    ymm4, ymm4, GOTOFF(eax,PW_MF016_MF033)  ; ymm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+    vmovdqa     YMMWORD wk(6), ymm0   ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+    vmovdqa     YMMWORD wk(7), ymm6   ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    vpxor       ymm0, ymm0, ymm0
+    vpxor       ymm6, ymm6, ymm6
+    vpunpcklwd  ymm0, ymm0, ymm1        ; ymm0=BEL
+    vpunpckhwd  ymm6, ymm6, ymm1        ; ymm6=BEH
+    vpsrld      ymm0, ymm0, 1           ; ymm0=BEL*FIX(0.500)
+    vpsrld      ymm6, ymm6, 1           ; ymm6=BEH*FIX(0.500)
+
+    vmovdqa     ymm1, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; ymm1=PD_ONEHALFM1_CJ
+
+    vpaddd      ymm5, ymm5, ymm0
+    vpaddd      ymm4, ymm4, ymm6
+    vpaddd      ymm5, ymm5, ymm1
+    vpaddd      ymm4, ymm4, ymm1
+    vpsrld      ymm5, ymm5, SCALEBITS   ; ymm5=CbEL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=CbEH
+    vpackssdw   ymm5, ymm5, ymm4        ; ymm5=CbE
+
+    vpsllw      ymm7, ymm7, BYTE_BIT
+    vpor        ymm5, ymm5, ymm7        ; ymm5=Cb
+    vmovdqu     YMMWORD ebx, ymm5     ; Save Cb
+
+    vmovdqa     ymm0, YMMWORD wk(3)   ; ymm0=BO
+    vmovdqa     ymm6, YMMWORD wk(2)   ; ymm6=BE
+    vmovdqa     ymm1, YMMWORD wk(1)   ; ymm1=RO
+
+    vmovdqa     ymm4, ymm0
+    vpunpcklwd  ymm0, ymm0, ymm3
+    vpunpckhwd  ymm4, ymm4, ymm3
+    vmovdqa     ymm7, ymm0
+    vmovdqa     ymm5, ymm4
+    vpmaddwd    ymm0, ymm0, GOTOFF(eax,PW_F0114_F0250)  ; ymm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    vpmaddwd    ymm4, ymm4, GOTOFF(eax,PW_F0114_F0250)  ; ymm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+    vpmaddwd    ymm7, ymm7, GOTOFF(eax,PW_MF008_MF041)  ; ymm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+    vpmaddwd    ymm5, ymm5, GOTOFF(eax,PW_MF008_MF041)  ; ymm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+    vmovdqa     ymm3, GOTOFF(eax,PD_ONEHALF)            ; ymm3=PD_ONEHALF
+
+    vpaddd      ymm0, ymm0, YMMWORD wk(4)
+    vpaddd      ymm4, ymm4, YMMWORD wk(5)
+    vpaddd      ymm0, ymm0, ymm3
+    vpaddd      ymm4, ymm4, ymm3
+    vpsrld      ymm0, ymm0, SCALEBITS   ; ymm0=YOL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=YOH
+    vpackssdw   ymm0, ymm0, ymm4        ; ymm0=YO
+
+    vpxor       ymm3, ymm3, ymm3
+    vpxor       ymm4, ymm4, ymm4
+    vpunpcklwd  ymm3, ymm3, ymm1        ; ymm3=ROL
+    vpunpckhwd  ymm4, ymm4, ymm1        ; ymm4=ROH
+    vpsrld      ymm3, ymm3, 1           ; ymm3=ROL*FIX(0.500)
+    vpsrld      ymm4, ymm4, 1           ; ymm4=ROH*FIX(0.500)
+
+    vmovdqa     ymm1, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; ymm1=PD_ONEHALFM1_CJ
+
+    vpaddd      ymm7, ymm7, ymm3
+    vpaddd      ymm5, ymm5, ymm4
+    vpaddd      ymm7, ymm7, ymm1
+    vpaddd      ymm5, ymm5, ymm1
+    vpsrld      ymm7, ymm7, SCALEBITS   ; ymm7=CrOL
+    vpsrld      ymm5, ymm5, SCALEBITS   ; ymm5=CrOH
+    vpackssdw   ymm7, ymm7, ymm5        ; ymm7=CrO
+
+    vmovdqa     ymm3, YMMWORD wk(0)   ; ymm3=RE
+
+    vmovdqa     ymm4, ymm6
+    vpunpcklwd  ymm6, ymm6, ymm2
+    vpunpckhwd  ymm4, ymm4, ymm2
+    vmovdqa     ymm1, ymm6
+    vmovdqa     ymm5, ymm4
+    vpmaddwd    ymm6, ymm6, GOTOFF(eax,PW_F0114_F0250)  ; ymm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    vpmaddwd    ymm4, ymm4, GOTOFF(eax,PW_F0114_F0250)  ; ymm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+    vpmaddwd    ymm1, ymm1, GOTOFF(eax,PW_MF008_MF041)  ; ymm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+    vpmaddwd    ymm5, ymm5, GOTOFF(eax,PW_MF008_MF041)  ; ymm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+    vmovdqa     ymm2, GOTOFF(eax,PD_ONEHALF)            ; ymm2=PD_ONEHALF
+
+    vpaddd      ymm6, ymm6, YMMWORD wk(6)
+    vpaddd      ymm4, ymm4, YMMWORD wk(7)
+    vpaddd      ymm6, ymm6, ymm2
+    vpaddd      ymm4, ymm4, ymm2
+    vpsrld      ymm6, ymm6, SCALEBITS   ; ymm6=YEL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=YEH
+    vpackssdw   ymm6, ymm6, ymm4        ; ymm6=YE
+
+    vpsllw      ymm0, ymm0, BYTE_BIT
+    vpor        ymm6, ymm6, ymm0        ; ymm6=Y
+    vmovdqu     YMMWORD edi, ymm6     ; Save Y
+
+    vpxor       ymm2, ymm2, ymm2
+    vpxor       ymm4, ymm4, ymm4
+    vpunpcklwd  ymm2, ymm2, ymm3        ; ymm2=REL
+    vpunpckhwd  ymm4, ymm4, ymm3        ; ymm4=REH
+    vpsrld      ymm2, ymm2, 1           ; ymm2=REL*FIX(0.500)
+    vpsrld      ymm4, ymm4, 1           ; ymm4=REH*FIX(0.500)
+
+    vmovdqa     ymm0, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; ymm0=PD_ONEHALFM1_CJ
+
+    vpaddd      ymm1, ymm1, ymm2
+    vpaddd      ymm5, ymm5, ymm4
+    vpaddd      ymm1, ymm1, ymm0
+    vpaddd      ymm5, ymm5, ymm0
+    vpsrld      ymm1, ymm1, SCALEBITS   ; ymm1=CrEL
+    vpsrld      ymm5, ymm5, SCALEBITS   ; ymm5=CrEH
+    vpackssdw   ymm1, ymm1, ymm5        ; ymm1=CrE
+
+    vpsllw      ymm7, ymm7, BYTE_BIT
+    vpor        ymm1, ymm1, ymm7        ; ymm1=Cr
+    vmovdqu     YMMWORD edx, ymm1     ; Save Cr
+
+    sub         ecx, byte SIZEOF_YMMWORD
+    add         esi, RGB_PIXELSIZE*SIZEOF_YMMWORD  ; inptr
+    add         edi, byte SIZEOF_YMMWORD           ; outptr0
+    add         ebx, byte SIZEOF_YMMWORD           ; outptr1
+    add         edx, byte SIZEOF_YMMWORD           ; outptr2
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+    test        ecx, ecx
+    jnz         near .column_ld1
+
+    pop         ecx                     ; col
+    pop         esi
+    pop         edi
+    pop         ebx
+    pop         edx
+    poppic      eax
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         edi, byte SIZEOF_JSAMPROW
+    add         ebx, byte SIZEOF_JSAMPROW
+    add         edx, byte SIZEOF_JSAMPROW
+    dec         eax                        ; num_rows
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jccolext-mmx.asm Added

@@ -0,0 +1,476 @@
+;
+; jccolext.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_mmx(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                           JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                           int num_rows);
+;
+
+%define img_width(b)   (b) + 8          ; JDIMENSION img_width
+%define input_buf(b)   (b) + 12         ; JSAMPARRAY input_buf
+%define output_buf(b)  (b) + 16         ; JSAMPIMAGE output_buf
+%define output_row(b)  (b) + 20         ; JDIMENSION output_row
+%define num_rows(b)    (b) + 24         ; int num_rows
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_MMWORD
+                                        ; mmword wkWK_NUM
+%define WK_NUM         8
+%define gotptr         wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_ycc_convert_mmx)
+
+EXTN(jsimd_rgb_ycc_convert_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION img_width(eax)  ; num_cols
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         esi, JSAMPIMAGE output_buf(eax)
+    mov         ecx, JDIMENSION output_row(eax)
+    mov         edi, JSAMPARRAY esi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY esi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY esi+2*SIZEOF_JSAMPARRAY
+    lea         edi, edi+ecx*SIZEOF_JSAMPROW
+    lea         ebx, ebx+ecx*SIZEOF_JSAMPROW
+    lea         edx, edx+ecx*SIZEOF_JSAMPROW
+
+    pop         ecx
+
+    mov         esi, JSAMPARRAY input_buf(eax)
+    mov         eax, INT num_rows(eax)
+    test        eax, eax
+    jle         near .return
+    alignx      16, 7
+.rowloop:
+    pushpic     eax
+    push        edx
+    push        ebx
+    push        edi
+    push        esi
+    push        ecx                     ; col
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr0
+    mov         ebx, JSAMPROW ebx     ; outptr1
+    mov         edx, JSAMPROW edx     ; outptr2
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+
+    cmp         ecx, byte SIZEOF_MMWORD
+    jae         short .columnloop
+    alignx      16, 7
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        eax
+    push        edx
+    lea         ecx, ecx+ecx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_BYTE
+    xor         eax, eax
+    mov         al, byte esi+ecx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_WORD
+    xor         edx, edx
+    mov         dx, word esi+ecx
+    shl         eax, WORD_BIT
+    or          eax, edx
+.column_ld4:
+    movd        mmA, eax
+    pop         edx
+    pop         eax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_DWORD
+    movd        mmG, dword esi+ecx
+    psllq       mmA, DWORD_BIT
+    por         mmA, mmG
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    movq        mmG, mmA
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    mov         ecx, SIZEOF_MMWORD
+    jmp         short .rgb_ycc_cnv
+.column_ld16:
+    test        cl, 2*SIZEOF_MMWORD
+    mov         ecx, SIZEOF_MMWORD
+    jz          short .rgb_ycc_cnv
+    movq        mmF, mmA
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mmG, MMWORD esi+1*SIZEOF_MMWORD
+    jmp         short .rgb_ycc_cnv
+    alignx      16, 7
+
+.columnloop:
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mmG, MMWORD esi+1*SIZEOF_MMWORD
+    movq        mmF, MMWORD esi+2*SIZEOF_MMWORD
+
+.rgb_ycc_cnv:
+    ; mmA=(00 10 20 01 11 21 02 12)
+    ; mmG=(22 03 13 23 04 14 24 05)
+    ; mmF=(15 25 06 16 26 07 17 27)
+
+    movq        mmD, mmA
+    psllq       mmA, 4*BYTE_BIT         ; mmA=(-- -- -- -- 00 10 20 01)
+    psrlq       mmD, 4*BYTE_BIT         ; mmD=(11 21 02 12 -- -- -- --)
+
+    punpckhbw   mmA, mmG                ; mmA=(00 04 10 14 20 24 01 05)
+    psllq       mmG, 4*BYTE_BIT         ; mmG=(-- -- -- -- 22 03 13 23)
+
+    punpcklbw   mmD, mmF                ; mmD=(11 15 21 25 02 06 12 16)
+    punpckhbw   mmG, mmF                ; mmG=(22 26 03 07 13 17 23 27)
+
+    movq        mmE, mmA
+    psllq       mmA, 4*BYTE_BIT         ; mmA=(-- -- -- -- 00 04 10 14)
+    psrlq       mmE, 4*BYTE_BIT         ; mmE=(20 24 01 05 -- -- -- --)
+
+    punpckhbw   mmA, mmD                ; mmA=(00 02 04 06 10 12 14 16)
+    psllq       mmD, 4*BYTE_BIT         ; mmD=(-- -- -- -- 11 15 21 25)
+
+    punpcklbw   mmE, mmG                ; mmE=(20 22 24 26 01 03 05 07)
+    punpckhbw   mmD, mmG                ; mmD=(11 13 15 17 21 23 25 27)
+
+    pxor        mmH, mmH
+
+    movq        mmC, mmA
+    punpcklbw   mmA, mmH                ; mmA=(00 02 04 06)
+    punpckhbw   mmC, mmH                ; mmC=(10 12 14 16)
+
+    movq        mmB, mmE
+    punpcklbw   mmE, mmH                ; mmE=(20 22 24 26)
+    punpckhbw   mmB, mmH                ; mmB=(01 03 05 07)
+
+    movq        mmF, mmD
+    punpcklbw   mmD, mmH                ; mmD=(11 13 15 17)
+    punpckhbw   mmF, mmH                ; mmF=(21 23 25 27)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_MMWORD/8
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_MMWORD/8
+    movd        mmA, dword esi+ecx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_MMWORD/4
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_MMWORD/4
+    movq        mmF, mmA
+    movq        mmA, MMWORD esi+ecx*RGB_PIXELSIZE
+.column_ld4:
+    test        cl, SIZEOF_MMWORD/2
+    mov         ecx, SIZEOF_MMWORD
+    jz          short .rgb_ycc_cnv
+    movq        mmD, mmA
+    movq        mmC, mmF
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mmF, MMWORD esi+1*SIZEOF_MMWORD
+    jmp         short .rgb_ycc_cnv
+    alignx      16, 7
+
+.columnloop:
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mmF, MMWORD esi+1*SIZEOF_MMWORD
+    movq        mmD, MMWORD esi+2*SIZEOF_MMWORD
+    movq        mmC, MMWORD esi+3*SIZEOF_MMWORD
+
+.rgb_ycc_cnv:
+    ; mmA=(00 10 20 30 01 11 21 31)
+    ; mmF=(02 12 22 32 03 13 23 33)
+    ; mmD=(04 14 24 34 05 15 25 35)
+    ; mmC=(06 16 26 36 07 17 27 37)
+
+    movq        mmB, mmA
+    punpcklbw   mmA, mmF                ; mmA=(00 02 10 12 20 22 30 32)
+    punpckhbw   mmB, mmF                ; mmB=(01 03 11 13 21 23 31 33)
+
+    movq        mmG, mmD
+    punpcklbw   mmD, mmC                ; mmD=(04 06 14 16 24 26 34 36)
+    punpckhbw   mmG, mmC                ; mmG=(05 07 15 17 25 27 35 37)
+
+    movq        mmE, mmA
+    punpcklwd   mmA, mmD                ; mmA=(00 02 04 06 10 12 14 16)
+    punpckhwd   mmE, mmD                ; mmE=(20 22 24 26 30 32 34 36)
+
+    movq        mmH, mmB
+    punpcklwd   mmB, mmG                ; mmB=(01 03 05 07 11 13 15 17)
+    punpckhwd   mmH, mmG                ; mmH=(21 23 25 27 31 33 35 37)
+
+    pxor        mmF, mmF
+
+    movq        mmC, mmA
+    punpcklbw   mmA, mmF                ; mmA=(00 02 04 06)
+    punpckhbw   mmC, mmF                ; mmC=(10 12 14 16)
+
+    movq        mmD, mmB
+    punpcklbw   mmB, mmF                ; mmB=(01 03 05 07)
+    punpckhbw   mmD, mmF                ; mmD=(11 13 15 17)
+
+    movq        mmG, mmE
+    punpcklbw   mmE, mmF                ; mmE=(20 22 24 26)
+    punpckhbw   mmG, mmF                ; mmG=(30 32 34 36)
+
+    punpcklbw   mmF, mmH
+    punpckhbw   mmH, mmH
+    psrlw       mmF, BYTE_BIT           ; mmF=(21 23 25 27)
+    psrlw       mmH, BYTE_BIT           ; mmH=(31 33 35 37)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; mm0=(R0 R2 R4 R6)=RE, mm2=(G0 G2 G4 G6)=GE, mm4=(B0 B2 B4 B6)=BE
+    ; mm1=(R1 R3 R5 R7)=RO, mm3=(G1 G3 G5 G7)=GO, mm5=(B1 B3 B5 B7)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+    movq        MMWORD wk(0), mm0     ; wk(0)=RE
+    movq        MMWORD wk(1), mm1     ; wk(1)=RO
+    movq        MMWORD wk(2), mm4     ; wk(2)=BE
+    movq        MMWORD wk(3), mm5     ; wk(3)=BO
+
+    movq        mm6, mm1
+    punpcklwd   mm1, mm3
+    punpckhwd   mm6, mm3
+    movq        mm7, mm1
+    movq        mm4, mm6
+    pmaddwd     mm1, GOTOFF(eax,PW_F0299_F0337)  ; mm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    pmaddwd     mm6, GOTOFF(eax,PW_F0299_F0337)  ; mm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+    pmaddwd     mm7, GOTOFF(eax,PW_MF016_MF033)  ; mm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+    pmaddwd     mm4, GOTOFF(eax,PW_MF016_MF033)  ; mm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+    movq        MMWORD wk(4), mm1     ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+    movq        MMWORD wk(5), mm6     ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    pxor        mm1, mm1
+    pxor        mm6, mm6
+    punpcklwd   mm1, mm5                ; mm1=BOL
+    punpckhwd   mm6, mm5                ; mm6=BOH
+    psrld       mm1, 1                  ; mm1=BOL*FIX(0.500)
+    psrld       mm6, 1                  ; mm6=BOH*FIX(0.500)
+
+    movq        mm5, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; mm5=PD_ONEHALFM1_CJ
+
+    paddd       mm7, mm1
+    paddd       mm4, mm6
+    paddd       mm7, mm5
+    paddd       mm4, mm5
+    psrld       mm7, SCALEBITS          ; mm7=CbOL
+    psrld       mm4, SCALEBITS          ; mm4=CbOH
+    packssdw    mm7, mm4                ; mm7=CbO
+
+    movq        mm1, MMWORD wk(2)     ; mm1=BE
+
+    movq        mm6, mm0
+    punpcklwd   mm0, mm2
+    punpckhwd   mm6, mm2
+    movq        mm5, mm0
+    movq        mm4, mm6
+    pmaddwd     mm0, GOTOFF(eax,PW_F0299_F0337)  ; mm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    pmaddwd     mm6, GOTOFF(eax,PW_F0299_F0337)  ; mm6=REH*FIX(0.299)+GEH*FIX(0.337)
+    pmaddwd     mm5, GOTOFF(eax,PW_MF016_MF033)  ; mm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+    pmaddwd     mm4, GOTOFF(eax,PW_MF016_MF033)  ; mm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+    movq        MMWORD wk(6), mm0     ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+    movq        MMWORD wk(7), mm6     ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    pxor        mm0, mm0
+    pxor        mm6, mm6
+    punpcklwd   mm0, mm1                ; mm0=BEL
+    punpckhwd   mm6, mm1                ; mm6=BEH
+    psrld       mm0, 1                  ; mm0=BEL*FIX(0.500)
+    psrld       mm6, 1                  ; mm6=BEH*FIX(0.500)
+
+    movq        mm1, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; mm1=PD_ONEHALFM1_CJ
+
+    paddd       mm5, mm0
+    paddd       mm4, mm6
+    paddd       mm5, mm1
+    paddd       mm4, mm1
+    psrld       mm5, SCALEBITS          ; mm5=CbEL
+    psrld       mm4, SCALEBITS          ; mm4=CbEH
+    packssdw    mm5, mm4                ; mm5=CbE
+
+    psllw       mm7, BYTE_BIT
+    por         mm5, mm7                ; mm5=Cb
+    movq        MMWORD ebx, mm5       ; Save Cb
+
+    movq        mm0, MMWORD wk(3)     ; mm0=BO
+    movq        mm6, MMWORD wk(2)     ; mm6=BE
+    movq        mm1, MMWORD wk(1)     ; mm1=RO
+
+    movq        mm4, mm0
+    punpcklwd   mm0, mm3
+    punpckhwd   mm4, mm3
+    movq        mm7, mm0
+    movq        mm5, mm4
+    pmaddwd     mm0, GOTOFF(eax,PW_F0114_F0250)  ; mm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    pmaddwd     mm4, GOTOFF(eax,PW_F0114_F0250)  ; mm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+    pmaddwd     mm7, GOTOFF(eax,PW_MF008_MF041)  ; mm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+    pmaddwd     mm5, GOTOFF(eax,PW_MF008_MF041)  ; mm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+    movq        mm3, GOTOFF(eax,PD_ONEHALF)  ; mm3=PD_ONEHALF
+
+    paddd       mm0, MMWORD wk(4)
+    paddd       mm4, MMWORD wk(5)
+    paddd       mm0, mm3
+    paddd       mm4, mm3
+    psrld       mm0, SCALEBITS          ; mm0=YOL
+    psrld       mm4, SCALEBITS          ; mm4=YOH
+    packssdw    mm0, mm4                ; mm0=YO
+
+    pxor        mm3, mm3
+    pxor        mm4, mm4
+    punpcklwd   mm3, mm1                ; mm3=ROL
+    punpckhwd   mm4, mm1                ; mm4=ROH
+    psrld       mm3, 1                  ; mm3=ROL*FIX(0.500)
+    psrld       mm4, 1                  ; mm4=ROH*FIX(0.500)
+
+    movq        mm1, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; mm1=PD_ONEHALFM1_CJ
+
+    paddd       mm7, mm3
+    paddd       mm5, mm4
+    paddd       mm7, mm1
+    paddd       mm5, mm1
+    psrld       mm7, SCALEBITS          ; mm7=CrOL
+    psrld       mm5, SCALEBITS          ; mm5=CrOH
+    packssdw    mm7, mm5                ; mm7=CrO
+
+    movq        mm3, MMWORD wk(0)     ; mm3=RE
+
+    movq        mm4, mm6
+    punpcklwd   mm6, mm2
+    punpckhwd   mm4, mm2
+    movq        mm1, mm6
+    movq        mm5, mm4
+    pmaddwd     mm6, GOTOFF(eax,PW_F0114_F0250)  ; mm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    pmaddwd     mm4, GOTOFF(eax,PW_F0114_F0250)  ; mm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+    pmaddwd     mm1, GOTOFF(eax,PW_MF008_MF041)  ; mm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+    pmaddwd     mm5, GOTOFF(eax,PW_MF008_MF041)  ; mm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+    movq        mm2, GOTOFF(eax,PD_ONEHALF)      ; mm2=PD_ONEHALF
+
+    paddd       mm6, MMWORD wk(6)
+    paddd       mm4, MMWORD wk(7)
+    paddd       mm6, mm2
+    paddd       mm4, mm2
+    psrld       mm6, SCALEBITS          ; mm6=YEL
+    psrld       mm4, SCALEBITS          ; mm4=YEH
+    packssdw    mm6, mm4                ; mm6=YE
+
+    psllw       mm0, BYTE_BIT
+    por         mm6, mm0                ; mm6=Y
+    movq        MMWORD edi, mm6       ; Save Y
+
+    pxor        mm2, mm2
+    pxor        mm4, mm4
+    punpcklwd   mm2, mm3                ; mm2=REL
+    punpckhwd   mm4, mm3                ; mm4=REH
+    psrld       mm2, 1                  ; mm2=REL*FIX(0.500)
+    psrld       mm4, 1                  ; mm4=REH*FIX(0.500)
+
+    movq        mm0, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; mm0=PD_ONEHALFM1_CJ
+
+    paddd       mm1, mm2
+    paddd       mm5, mm4
+    paddd       mm1, mm0
+    paddd       mm5, mm0
+    psrld       mm1, SCALEBITS          ; mm1=CrEL
+    psrld       mm5, SCALEBITS          ; mm5=CrEH
+    packssdw    mm1, mm5                ; mm1=CrE
+
+    psllw       mm7, BYTE_BIT
+    por         mm1, mm7                ; mm1=Cr
+    movq        MMWORD edx, mm1       ; Save Cr
+
+    sub         ecx, byte SIZEOF_MMWORD
+    add         esi, byte RGB_PIXELSIZE*SIZEOF_MMWORD  ; inptr
+    add         edi, byte SIZEOF_MMWORD                ; outptr0
+    add         ebx, byte SIZEOF_MMWORD                ; outptr1
+    add         edx, byte SIZEOF_MMWORD                ; outptr2
+    cmp         ecx, byte SIZEOF_MMWORD
+    jae         near .columnloop
+    test        ecx, ecx
+    jnz         near .column_ld1
+
+    pop         ecx                     ; col
+    pop         esi
+    pop         edi
+    pop         ebx
+    pop         edx
+    poppic      eax
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         edi, byte SIZEOF_JSAMPROW
+    add         ebx, byte SIZEOF_JSAMPROW
+    add         edx, byte SIZEOF_JSAMPROW
+    dec         eax                        ; num_rows
+    jg          near .rowloop
+
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jccolext-sse2.asm Added

@@ -0,0 +1,503 @@
+;
+; jccolext.asm - colorspace conversion (SSE2)
+;
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_sse2(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                            JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                            int num_rows);
+;
+
+%define img_width(b)   (b) + 8          ; JDIMENSION img_width
+%define input_buf(b)   (b) + 12         ; JSAMPARRAY input_buf
+%define output_buf(b)  (b) + 16         ; JSAMPIMAGE output_buf
+%define output_row(b)  (b) + 20         ; JDIMENSION output_row
+%define num_rows(b)    (b) + 24         ; int num_rows
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM         8
+%define gotptr         wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_ycc_convert_sse2)
+
+EXTN(jsimd_rgb_ycc_convert_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION img_width(eax)
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         esi, JSAMPIMAGE output_buf(eax)
+    mov         ecx, JDIMENSION output_row(eax)
+    mov         edi, JSAMPARRAY esi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY esi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY esi+2*SIZEOF_JSAMPARRAY
+    lea         edi, edi+ecx*SIZEOF_JSAMPROW
+    lea         ebx, ebx+ecx*SIZEOF_JSAMPROW
+    lea         edx, edx+ecx*SIZEOF_JSAMPROW
+
+    pop         ecx
+
+    mov         esi, JSAMPARRAY input_buf(eax)
+    mov         eax, INT num_rows(eax)
+    test        eax, eax
+    jle         near .return
+    alignx      16, 7
+.rowloop:
+    pushpic     eax
+    push        edx
+    push        ebx
+    push        edi
+    push        esi
+    push        ecx                     ; col
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr0
+    mov         ebx, JSAMPROW ebx     ; outptr1
+    mov         edx, JSAMPROW edx     ; outptr2
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+    alignx      16, 7
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        eax
+    push        edx
+    lea         ecx, ecx+ecx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_BYTE
+    movzx       eax, byte esi+ecx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_WORD
+    movzx       edx, word esi+ecx
+    shl         eax, WORD_BIT
+    or          eax, edx
+.column_ld4:
+    movd        xmmA, eax
+    pop         edx
+    pop         eax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_DWORD
+    movd        xmmF, XMM_DWORD esi+ecx
+    pslldq      xmmA, SIZEOF_DWORD
+    por         xmmA, xmmF
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    sub         ecx, byte SIZEOF_MMWORD
+    movq        xmmB, XMM_MMWORD esi+ecx
+    pslldq      xmmA, SIZEOF_MMWORD
+    por         xmmA, xmmB
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    jz          short .column_ld32
+    movdqa      xmmF, xmmA
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    mov         ecx, SIZEOF_XMMWORD
+    jmp         short .rgb_ycc_cnv
+.column_ld32:
+    test        cl, 2*SIZEOF_XMMWORD
+    mov         ecx, SIZEOF_XMMWORD
+    jz          short .rgb_ycc_cnv
+    movdqa      xmmB, xmmA
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD esi+1*SIZEOF_XMMWORD
+    jmp         short .rgb_ycc_cnv
+    alignx      16, 7
+
+.columnloop:
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD esi+1*SIZEOF_XMMWORD
+    movdqu      xmmB, XMMWORD esi+2*SIZEOF_XMMWORD
+
+.rgb_ycc_cnv:
+    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+    ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+    movdqa      xmmG, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+    psrldq      xmmG, 8     ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmF  ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+    pslldq      xmmF, 8     ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+    punpcklbw   xmmG, xmmB  ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+    punpckhbw   xmmF, xmmB  ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+    movdqa      xmmD, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+    psrldq      xmmD, 8     ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmG  ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+    pslldq      xmmG, 8     ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+    punpcklbw   xmmD, xmmF  ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+    punpckhbw   xmmG, xmmF  ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+    movdqa      xmmE, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+    psrldq      xmmE, 8     ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmD  ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+    pslldq      xmmD, 8     ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+    punpcklbw   xmmE, xmmG  ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+    punpckhbw   xmmD, xmmG  ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+    pxor        xmmH, xmmH
+
+    movdqa      xmmC, xmmA
+    punpcklbw   xmmA, xmmH  ; xmmA=(00 02 04 06 08 0A 0C 0E)
+    punpckhbw   xmmC, xmmH  ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+    movdqa      xmmB, xmmE
+    punpcklbw   xmmE, xmmH  ; xmmE=(20 22 24 26 28 2A 2C 2E)
+    punpckhbw   xmmB, xmmH  ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+    movdqa      xmmF, xmmD
+    punpcklbw   xmmD, xmmH  ; xmmD=(11 13 15 17 19 1B 1D 1F)
+    punpckhbw   xmmF, xmmH  ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_XMMWORD/16
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_XMMWORD/16
+    movd        xmmA, XMM_DWORD esi+ecx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_XMMWORD/8
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_XMMWORD/8
+    movq        xmmE, XMM_MMWORD esi+ecx*RGB_PIXELSIZE
+    pslldq      xmmA, SIZEOF_MMWORD
+    por         xmmA, xmmE
+.column_ld4:
+    test        cl, SIZEOF_XMMWORD/4
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_XMMWORD/4
+    movdqa      xmmE, xmmA
+    movdqu      xmmA, XMMWORD esi+ecx*RGB_PIXELSIZE
+.column_ld8:
+    test        cl, SIZEOF_XMMWORD/2
+    mov         ecx, SIZEOF_XMMWORD
+    jz          short .rgb_ycc_cnv
+    movdqa      xmmF, xmmA
+    movdqa      xmmH, xmmE
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqu      xmmE, XMMWORD esi+1*SIZEOF_XMMWORD
+    jmp         short .rgb_ycc_cnv
+    alignx      16, 7
+
+.columnloop:
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqu      xmmE, XMMWORD esi+1*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD esi+2*SIZEOF_XMMWORD
+    movdqu      xmmH, XMMWORD esi+3*SIZEOF_XMMWORD
+
+.rgb_ycc_cnv:
+    ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+    ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+    ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+    movdqa      xmmD, xmmA
+    punpcklbw   xmmA, xmmE      ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+    punpckhbw   xmmD, xmmE      ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+    movdqa      xmmC, xmmF
+    punpcklbw   xmmF, xmmH      ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+    punpckhbw   xmmC, xmmH      ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+    movdqa      xmmB, xmmA
+    punpcklwd   xmmA, xmmF      ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+    punpckhwd   xmmB, xmmF      ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+    movdqa      xmmG, xmmD
+    punpcklwd   xmmD, xmmC      ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+    punpckhwd   xmmG, xmmC      ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+    movdqa      xmmE, xmmA
+    punpcklbw   xmmA, xmmD      ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+    punpckhbw   xmmE, xmmD      ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+    movdqa      xmmH, xmmB
+    punpcklbw   xmmB, xmmG      ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+    punpckhbw   xmmH, xmmG      ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+    pxor        xmmF, xmmF
+
+    movdqa      xmmC, xmmA
+    punpcklbw   xmmA, xmmF      ; xmmA=(00 02 04 06 08 0A 0C 0E)
+    punpckhbw   xmmC, xmmF      ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+    movdqa      xmmD, xmmB
+    punpcklbw   xmmB, xmmF      ; xmmB=(01 03 05 07 09 0B 0D 0F)
+    punpckhbw   xmmD, xmmF      ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+    movdqa      xmmG, xmmE
+    punpcklbw   xmmE, xmmF      ; xmmE=(20 22 24 26 28 2A 2C 2E)
+    punpckhbw   xmmG, xmmF      ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+    punpcklbw   xmmF, xmmH
+    punpckhbw   xmmH, xmmH
+    psrlw       xmmF, BYTE_BIT  ; xmmF=(21 23 25 27 29 2B 2D 2F)
+    psrlw       xmmH, BYTE_BIT  ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+    ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+    movdqa      XMMWORD wk(0), xmm0   ; wk(0)=RE
+    movdqa      XMMWORD wk(1), xmm1   ; wk(1)=RO
+    movdqa      XMMWORD wk(2), xmm4   ; wk(2)=BE
+    movdqa      XMMWORD wk(3), xmm5   ; wk(3)=BO
+
+    movdqa      xmm6, xmm1
+    punpcklwd   xmm1, xmm3
+    punpckhwd   xmm6, xmm3
+    movdqa      xmm7, xmm1
+    movdqa      xmm4, xmm6
+    pmaddwd     xmm1, GOTOFF(eax,PW_F0299_F0337)  ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    pmaddwd     xmm6, GOTOFF(eax,PW_F0299_F0337)  ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+    pmaddwd     xmm7, GOTOFF(eax,PW_MF016_MF033)  ; xmm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+    pmaddwd     xmm4, GOTOFF(eax,PW_MF016_MF033)  ; xmm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+    movdqa      XMMWORD wk(4), xmm1   ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+    movdqa      XMMWORD wk(5), xmm6   ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    pxor        xmm1, xmm1
+    pxor        xmm6, xmm6
+    punpcklwd   xmm1, xmm5              ; xmm1=BOL
+    punpckhwd   xmm6, xmm5              ; xmm6=BOH
+    psrld       xmm1, 1                 ; xmm1=BOL*FIX(0.500)
+    psrld       xmm6, 1                 ; xmm6=BOH*FIX(0.500)
+
+    movdqa      xmm5, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; xmm5=PD_ONEHALFM1_CJ
+
+    paddd       xmm7, xmm1
+    paddd       xmm4, xmm6
+    paddd       xmm7, xmm5
+    paddd       xmm4, xmm5
+    psrld       xmm7, SCALEBITS         ; xmm7=CbOL
+    psrld       xmm4, SCALEBITS         ; xmm4=CbOH
+    packssdw    xmm7, xmm4              ; xmm7=CbO
+
+    movdqa      xmm1, XMMWORD wk(2)   ; xmm1=BE
+
+    movdqa      xmm6, xmm0
+    punpcklwd   xmm0, xmm2
+    punpckhwd   xmm6, xmm2
+    movdqa      xmm5, xmm0
+    movdqa      xmm4, xmm6
+    pmaddwd     xmm0, GOTOFF(eax,PW_F0299_F0337)  ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    pmaddwd     xmm6, GOTOFF(eax,PW_F0299_F0337)  ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+    pmaddwd     xmm5, GOTOFF(eax,PW_MF016_MF033)  ; xmm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+    pmaddwd     xmm4, GOTOFF(eax,PW_MF016_MF033)  ; xmm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+    movdqa      XMMWORD wk(6), xmm0   ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+    movdqa      XMMWORD wk(7), xmm6   ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    pxor        xmm0, xmm0
+    pxor        xmm6, xmm6
+    punpcklwd   xmm0, xmm1              ; xmm0=BEL
+    punpckhwd   xmm6, xmm1              ; xmm6=BEH
+    psrld       xmm0, 1                 ; xmm0=BEL*FIX(0.500)
+    psrld       xmm6, 1                 ; xmm6=BEH*FIX(0.500)
+
+    movdqa      xmm1, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; xmm1=PD_ONEHALFM1_CJ
+
+    paddd       xmm5, xmm0
+    paddd       xmm4, xmm6
+    paddd       xmm5, xmm1
+    paddd       xmm4, xmm1
+    psrld       xmm5, SCALEBITS         ; xmm5=CbEL
+    psrld       xmm4, SCALEBITS         ; xmm4=CbEH
+    packssdw    xmm5, xmm4              ; xmm5=CbE
+
+    psllw       xmm7, BYTE_BIT
+    por         xmm5, xmm7              ; xmm5=Cb
+    movdqa      XMMWORD ebx, xmm5     ; Save Cb
+
+    movdqa      xmm0, XMMWORD wk(3)   ; xmm0=BO
+    movdqa      xmm6, XMMWORD wk(2)   ; xmm6=BE
+    movdqa      xmm1, XMMWORD wk(1)   ; xmm1=RO
+
+    movdqa      xmm4, xmm0
+    punpcklwd   xmm0, xmm3
+    punpckhwd   xmm4, xmm3
+    movdqa      xmm7, xmm0
+    movdqa      xmm5, xmm4
+    pmaddwd     xmm0, GOTOFF(eax,PW_F0114_F0250)  ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    pmaddwd     xmm4, GOTOFF(eax,PW_F0114_F0250)  ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+    pmaddwd     xmm7, GOTOFF(eax,PW_MF008_MF041)  ; xmm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+    pmaddwd     xmm5, GOTOFF(eax,PW_MF008_MF041)  ; xmm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+    movdqa      xmm3, GOTOFF(eax,PD_ONEHALF)  ; xmm3=PD_ONEHALF
+
+    paddd       xmm0, XMMWORD wk(4)
+    paddd       xmm4, XMMWORD wk(5)
+    paddd       xmm0, xmm3
+    paddd       xmm4, xmm3
+    psrld       xmm0, SCALEBITS         ; xmm0=YOL
+    psrld       xmm4, SCALEBITS         ; xmm4=YOH
+    packssdw    xmm0, xmm4              ; xmm0=YO
+
+    pxor        xmm3, xmm3
+    pxor        xmm4, xmm4
+    punpcklwd   xmm3, xmm1              ; xmm3=ROL
+    punpckhwd   xmm4, xmm1              ; xmm4=ROH
+    psrld       xmm3, 1                 ; xmm3=ROL*FIX(0.500)
+    psrld       xmm4, 1                 ; xmm4=ROH*FIX(0.500)
+
+    movdqa      xmm1, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; xmm1=PD_ONEHALFM1_CJ
+
+    paddd       xmm7, xmm3
+    paddd       xmm5, xmm4
+    paddd       xmm7, xmm1
+    paddd       xmm5, xmm1
+    psrld       xmm7, SCALEBITS         ; xmm7=CrOL
+    psrld       xmm5, SCALEBITS         ; xmm5=CrOH
+    packssdw    xmm7, xmm5              ; xmm7=CrO
+
+    movdqa      xmm3, XMMWORD wk(0)   ; xmm3=RE
+
+    movdqa      xmm4, xmm6
+    punpcklwd   xmm6, xmm2
+    punpckhwd   xmm4, xmm2
+    movdqa      xmm1, xmm6
+    movdqa      xmm5, xmm4
+    pmaddwd     xmm6, GOTOFF(eax,PW_F0114_F0250)  ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    pmaddwd     xmm4, GOTOFF(eax,PW_F0114_F0250)  ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+    pmaddwd     xmm1, GOTOFF(eax,PW_MF008_MF041)  ; xmm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+    pmaddwd     xmm5, GOTOFF(eax,PW_MF008_MF041)  ; xmm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+    movdqa      xmm2, GOTOFF(eax,PD_ONEHALF)      ; xmm2=PD_ONEHALF
+
+    paddd       xmm6, XMMWORD wk(6)
+    paddd       xmm4, XMMWORD wk(7)
+    paddd       xmm6, xmm2
+    paddd       xmm4, xmm2
+    psrld       xmm6, SCALEBITS         ; xmm6=YEL
+    psrld       xmm4, SCALEBITS         ; xmm4=YEH
+    packssdw    xmm6, xmm4              ; xmm6=YE
+
+    psllw       xmm0, BYTE_BIT
+    por         xmm6, xmm0              ; xmm6=Y
+    movdqa      XMMWORD edi, xmm6     ; Save Y
+
+    pxor        xmm2, xmm2
+    pxor        xmm4, xmm4
+    punpcklwd   xmm2, xmm3              ; xmm2=REL
+    punpckhwd   xmm4, xmm3              ; xmm4=REH
+    psrld       xmm2, 1                 ; xmm2=REL*FIX(0.500)
+    psrld       xmm4, 1                 ; xmm4=REH*FIX(0.500)
+
+    movdqa      xmm0, GOTOFF(eax,PD_ONEHALFM1_CJ)  ; xmm0=PD_ONEHALFM1_CJ
+
+    paddd       xmm1, xmm2
+    paddd       xmm5, xmm4
+    paddd       xmm1, xmm0
+    paddd       xmm5, xmm0
+    psrld       xmm1, SCALEBITS         ; xmm1=CrEL
+    psrld       xmm5, SCALEBITS         ; xmm5=CrEH
+    packssdw    xmm1, xmm5              ; xmm1=CrE
+
+    psllw       xmm7, BYTE_BIT
+    por         xmm1, xmm7              ; xmm1=Cr
+    movdqa      XMMWORD edx, xmm1     ; Save Cr
+
+    sub         ecx, byte SIZEOF_XMMWORD
+    add         esi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+    add         edi, byte SIZEOF_XMMWORD                ; outptr0
+    add         ebx, byte SIZEOF_XMMWORD                ; outptr1
+    add         edx, byte SIZEOF_XMMWORD                ; outptr2
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+    test        ecx, ecx
+    jnz         near .column_ld1
+
+    pop         ecx                     ; col
+    pop         esi
+    pop         edi
+    pop         ebx
+    pop         edx
+    poppic      eax
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         edi, byte SIZEOF_JSAMPROW
+    add         ebx, byte SIZEOF_JSAMPROW
+    add         edx, byte SIZEOF_JSAMPROW
+    dec         eax                        ; num_rows
+    jg          near .rowloop
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jccolor-avx2.asm Added

@@ -0,0 +1,121 @@
+;
+; jccolor.asm - colorspace conversion (AVX2)
+;
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_081 equ  5329                ; FIX(0.08131)
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_168 equ 11059                ; FIX(0.16874)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_331 equ 21709                ; FIX(0.33126)
+F_0_418 equ 27439                ; FIX(0.41869)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_ycc_convert_avx2)
+
+EXTN(jconst_rgb_ycc_convert_avx2):
+
+PW_F0299_F0337  times 8 dw  F_0_299,  F_0_337
+PW_F0114_F0250  times 8 dw  F_0_114,  F_0_250
+PW_MF016_MF033  times 8 dw -F_0_168, -F_0_331
+PW_MF008_MF041  times 8 dw -F_0_081, -F_0_418
+PD_ONEHALFM1_CJ times 8 dd  (1 << (SCALEBITS - 1)) - 1 + \
+                            (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 8 dd  (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extrgb_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extrgbx_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extbgr_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extbgrx_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extxbgr_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extxrgb_ycc_convert_avx2
+%include "jccolext-avx2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jccolor-mmx.asm Added

@@ -0,0 +1,121 @@
+;
+; jccolor.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_081 equ  5329                ; FIX(0.08131)
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_168 equ 11059                ; FIX(0.16874)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_331 equ 21709                ; FIX(0.33126)
+F_0_418 equ 27439                ; FIX(0.41869)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_ycc_convert_mmx)
+
+EXTN(jconst_rgb_ycc_convert_mmx):
+
+PW_F0299_F0337  times 2 dw  F_0_299,  F_0_337
+PW_F0114_F0250  times 2 dw  F_0_114,  F_0_250
+PW_MF016_MF033  times 2 dw -F_0_168, -F_0_331
+PW_MF008_MF041  times 2 dw -F_0_081, -F_0_418
+PD_ONEHALFM1_CJ times 2 dd  (1 << (SCALEBITS - 1)) - 1 + \
+                            (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 2 dd  (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx  jsimd_extrgb_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx  jsimd_extrgbx_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx  jsimd_extbgr_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx  jsimd_extbgrx_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx  jsimd_extxbgr_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx  jsimd_extxrgb_ycc_convert_mmx
+%include "jccolext-mmx.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jccolor-sse2.asm Added

@@ -0,0 +1,120 @@
+;
+; jccolor.asm - colorspace conversion (SSE2)
+;
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_081 equ  5329                ; FIX(0.08131)
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_168 equ 11059                ; FIX(0.16874)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_331 equ 21709                ; FIX(0.33126)
+F_0_418 equ 27439                ; FIX(0.41869)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_ycc_convert_sse2)
+
+EXTN(jconst_rgb_ycc_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299,  F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114,  F_0_250
+PW_MF016_MF033  times 4 dw -F_0_168, -F_0_331
+PW_MF008_MF041  times 4 dw -F_0_081, -F_0_418
+PD_ONEHALFM1_CJ times 4 dd  (1 << (SCALEBITS - 1)) - 1 + \
+                            (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extrgb_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extrgbx_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extbgr_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extbgrx_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extxbgr_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extxrgb_ycc_convert_sse2
+%include "jccolext-sse2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcgray-avx2.asm Added

@@ -0,0 +1,113 @@
+;
+; jcgray.asm - grayscale colorspace conversion (AVX2)
+;
+; Copyright (C) 2011, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_gray_convert_avx2)
+
+EXTN(jconst_rgb_gray_convert_avx2):
+
+PW_F0299_F0337 times 8 dw F_0_299, F_0_337
+PW_F0114_F0250 times 8 dw F_0_114, F_0_250
+PD_ONEHALF     times 8 dd (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extrgb_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extrgbx_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extbgr_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extbgrx_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extxbgr_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extxrgb_gray_convert_avx2
+%include "jcgryext-avx2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcgray-mmx.asm Added

@@ -0,0 +1,113 @@
+;
+; jcgray.asm - grayscale colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2011, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_gray_convert_mmx)
+
+EXTN(jconst_rgb_gray_convert_mmx):
+
+PW_F0299_F0337 times 2 dw F_0_299, F_0_337
+PW_F0114_F0250 times 2 dw F_0_114, F_0_250
+PD_ONEHALF     times 2 dd (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx  jsimd_extrgb_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx  jsimd_extrgbx_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx  jsimd_extbgr_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx  jsimd_extbgrx_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx  jsimd_extxbgr_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx  jsimd_extxrgb_gray_convert_mmx
+%include "jcgryext-mmx.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcgray-sse2.asm Added

@@ -0,0 +1,112 @@
+;
+; jcgray.asm - grayscale colorspace conversion (SSE2)
+;
+; Copyright (C) 2011, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_gray_convert_sse2)
+
+EXTN(jconst_rgb_gray_convert_sse2):
+
+PW_F0299_F0337 times 4 dw F_0_299, F_0_337
+PW_F0114_F0250 times 4 dw F_0_114, F_0_250
+PD_ONEHALF     times 4 dd (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extrgb_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extrgbx_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extbgr_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extbgrx_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extxbgr_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extxrgb_gray_convert_sse2
+%include "jcgryext-sse2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcgryext-avx2.asm Added

@@ -0,0 +1,457 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (AVX2)
+;
+; Copyright (C) 2011, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_avx2(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                             JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                             int num_rows);
+;
+
+%define img_width(b)   (b) + 8          ; JDIMENSION img_width
+%define input_buf(b)   (b) + 12         ; JSAMPARRAY input_buf
+%define output_buf(b)  (b) + 16         ; JSAMPIMAGE output_buf
+%define output_row(b)  (b) + 20         ; JDIMENSION output_row
+%define num_rows(b)    (b) + 24         ; int num_rows
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_YMMWORD
+                                        ; ymmword wkWK_NUM
+%define WK_NUM         2
+%define gotptr         wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_gray_convert_avx2)
+
+EXTN(jsimd_rgb_gray_convert_avx2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION img_width(eax)
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         esi, JSAMPIMAGE output_buf(eax)
+    mov         ecx, JDIMENSION output_row(eax)
+    mov         edi, JSAMPARRAY esi+0*SIZEOF_JSAMPARRAY
+    lea         edi, edi+ecx*SIZEOF_JSAMPROW
+
+    pop         ecx
+
+    mov         esi, JSAMPARRAY input_buf(eax)
+    mov         eax, INT num_rows(eax)
+    test        eax, eax
+    jle         near .return
+    alignx      16, 7
+.rowloop:
+    pushpic     eax
+    push        edi
+    push        esi
+    push        ecx                     ; col
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr0
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+    alignx      16, 7
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        eax
+    push        edx
+    lea         ecx, ecx+ecx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_BYTE
+    movzx       eax, byte esi+ecx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_WORD
+    movzx       edx, word esi+ecx
+    shl         eax, WORD_BIT
+    or          eax, edx
+.column_ld4:
+    vmovd       xmmA, eax
+    pop         edx
+    pop         eax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_DWORD
+    vmovd       xmmF, XMM_DWORD esi+ecx
+    vpslldq     xmmA, xmmA, SIZEOF_DWORD
+    vpor        xmmA, xmmA, xmmF
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    sub         ecx, byte SIZEOF_MMWORD
+    vmovq       xmmB, XMM_MMWORD esi+ecx
+    vpslldq     xmmA, xmmA, SIZEOF_MMWORD
+    vpor        xmmA, xmmA, xmmB
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    jz          short .column_ld32
+    sub         ecx, byte SIZEOF_XMMWORD
+    vmovdqu     xmmB, XMM_MMWORD esi+ecx
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    vpor        ymmA, ymmB
+.column_ld32:
+    test        cl, SIZEOF_YMMWORD
+    jz          short .column_ld64
+    sub         ecx, byte SIZEOF_YMMWORD
+    vmovdqa     ymmF, ymmA
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+.column_ld64:
+    test        cl, 2*SIZEOF_YMMWORD
+    mov         ecx, SIZEOF_YMMWORD
+    jz          short .rgb_gray_cnv
+    vmovdqa     ymmB, ymmA
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD esi+1*SIZEOF_YMMWORD
+    jmp         short .rgb_gray_cnv
+    alignx      16, 7
+
+.columnloop:
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD esi+1*SIZEOF_YMMWORD
+    vmovdqu     ymmB, YMMWORD esi+2*SIZEOF_YMMWORD
+
+.rgb_gray_cnv:
+    ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+    ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    ; ymmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+    ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    ; ymmB=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+    ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    vmovdqu     ymmC, ymmA
+    vinserti128 ymmA, ymmF, xmmA, 0  ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                     ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vinserti128 ymmC, ymmC, xmmB, 0  ; ymmC=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+                                     ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    vinserti128 ymmB, ymmB, xmmF, 0  ; ymmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                     ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+    vperm2i128  ymmF, ymmC, ymmC, 1  ; ymmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A
+                                     ;       1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q)
+
+    vmovdqa     ymmG, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12
+                                  ;       22 03 13 23 04 14 24 05 0G 1G 2G 0H 1H 2H 0I 1I)
+    vpsrldq     ymmG, ymmG, 8     ; ymmG=(22 03 13 23 04 14 24 05 0G 1G 2G 0H 1H 2H 0I 1I
+                                  ;       2I 0J 1J 2J 0K 1K 2K 0L -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmF  ; ymmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A
+                                  ;       0G 0O 1G 1O 2G 2O 0H 0P 1H 1P 2H 2P 0I 0Q 1I 1Q)
+    vpslldq     ymmF, ymmF, 8     ; ymmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27
+                                  ;       08 18 28 09 19 29 0A 1A 1L 2L 0M 1M 2M 0N 1N 2N)
+
+    vpunpcklbw  ymmG, ymmG, ymmB  ; ymmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D
+                                  ;       2I 2Q 0J 0R 1J 1R 2J 2R 0K 0S 1K 1S 2K 2S 0L 0T)
+    vpunpckhbw  ymmF, ymmF, ymmB  ; ymmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F
+                                  ;       1L 1T 2L 2T 0M 0U 1M 1U 2M 2U 0N 0V 1N 1V 2N 2V)
+
+    vmovdqa     ymmD, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09
+                                  ;       11 19 21 29 02 0A 12 1A 0G 0O 1G 1O 2G 2O 0H 0P)
+    vpsrldq     ymmD, ymmD, 8     ; ymmD=(11 19 21 29 02 0A 12 1A 0G 0O 1G 1O 2G 2O 0H 0P
+                                  ;       1H 1P 2H 2P 0I 0Q 1I 1Q -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmG  ; ymmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D
+                                  ;       0G 0K 0O 0S 1G 1K 1O 1S 2G 2K 2O 2S 0H 0L 0P 0T)
+    vpslldq     ymmG, ymmG, 8     ; ymmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B
+                                  ;       04 0C 14 1C 24 2C 05 0D 2I 2Q 0J 0R 1J 1R 2J 2R)
+
+    vpunpcklbw  ymmD, ymmD, ymmF  ; ymmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E
+                                  ;       1H 1L 1P 1T 2H 2L 2P 2T 0I 0M 0Q 0U 1I 1M 1Q 1U)
+    vpunpckhbw  ymmG, ymmG, ymmF  ; ymmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F
+                                  ;       2I 2M 2Q 2U 0J 0N 0R 0V 1J 1N 1R 1V 2J 2N 2R 2V)
+
+    vmovdqa     ymmE, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C
+                                  ;       20 24 28 2C 01 05 09 0D 0G 0K 0O 0S 1G 1K 1O 1S)
+    vpsrldq     ymmE, ymmE, 8     ; ymmE=(20 24 28 2C 01 05 09 0D 0G 0K 0O 0S 1G 1K 1O 1S
+                                  ;       2G 2K 2O 2S 0H 0L 0P 0T -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmD  ; ymmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E
+                                  ;       0G 0I 0K 0M 0O 0Q 0S 0U 1G 1I 1K 1M 1O 1Q 1S 1U)
+    vpslldq     ymmD, ymmD, 8     ; ymmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D
+                                  ;       02 06 0A 0E 12 16 1A 1E 1H 1L 1P 1T 2H 2L 2P 2T)
+
+    vpunpcklbw  ymmE, ymmE, ymmG  ; ymmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F
+                                  ;       2G 2I 2K 2M 2O 2Q 2S 2U 0H 0J 0L 0N 0P 0R 0T 0V)
+    vpunpckhbw  ymmD, ymmD, ymmG  ; ymmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F
+                                  ;       1H 1J 1L 1N 1P 1R 1T 1V 2H 2J 2L 2N 2P 2R 2T 2V)
+
+    vpxor       ymmH, ymmH, ymmH
+
+    vmovdqa     ymmC, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmH  ; ymmA=(00 02 04 06 08 0A 0C 0E 0G 0I 0K 0M 0O 0Q 0S 0U)
+    vpunpckhbw  ymmC, ymmC, ymmH  ; ymmC=(10 12 14 16 18 1A 1C 1E 1G 1I 1K 1M 1O 1Q 1S 1U)
+
+    vmovdqa     ymmB, ymmE
+    vpunpcklbw  ymmE, ymmE, ymmH  ; ymmE=(20 22 24 26 28 2A 2C 2E 2G 2I 2K 2M 2O 2Q 2S 2U)
+    vpunpckhbw  ymmB, ymmB, ymmH  ; ymmB=(01 03 05 07 09 0B 0D 0F 0H 0J 0L 0N 0P 0R 0T 0V)
+
+    vmovdqa     ymmF, ymmD
+    vpunpcklbw  ymmD, ymmD, ymmH  ; ymmD=(11 13 15 17 19 1B 1D 1F 1H 1J 1L 1N 1P 1R 1T 1V)
+    vpunpckhbw  ymmF, ymmF, ymmH  ; ymmF=(21 23 25 27 29 2B 2D 2F 2H 2J 2L 2N 2P 2R 2T 2V)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_XMMWORD/16
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_XMMWORD/16
+    vmovd       xmmA, XMM_DWORD esi+ecx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_XMMWORD/8
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_XMMWORD/8
+    vmovq       xmmF, XMM_MMWORD esi+ecx*RGB_PIXELSIZE
+    vpslldq     xmmA, xmmA, SIZEOF_MMWORD
+    vpor        xmmA, xmmA, xmmF
+.column_ld4:
+    test        cl, SIZEOF_XMMWORD/4
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_XMMWORD/4
+    vmovdqa     xmmF, xmmA
+    vperm2i128  ymmF, ymmF, ymmF, 1
+    vmovdqu     xmmA, XMMWORD esi+ecx*RGB_PIXELSIZE
+    vpor        ymmA, ymmA, ymmF
+.column_ld8:
+    test        cl, SIZEOF_XMMWORD/2
+    jz          short .column_ld16
+    sub         ecx, byte SIZEOF_XMMWORD/2
+    vmovdqa     ymmF, ymmA
+    vmovdqu     ymmA, YMMWORD esi+ecx*RGB_PIXELSIZE
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    mov         ecx, SIZEOF_YMMWORD
+    jz          short .rgb_gray_cnv
+    vmovdqa     ymmE, ymmA
+    vmovdqa     ymmH, ymmF
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD esi+1*SIZEOF_YMMWORD
+    jmp         short .rgb_gray_cnv
+    alignx      16, 7
+
+.columnloop:
+    vmovdqu     ymmA, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD esi+1*SIZEOF_YMMWORD
+    vmovdqu     ymmE, YMMWORD esi+2*SIZEOF_YMMWORD
+    vmovdqu     ymmH, YMMWORD esi+3*SIZEOF_YMMWORD
+
+.rgb_gray_cnv:
+    ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+    ;       04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    ; ymmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+    ;       0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+    ; ymmE=(0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J
+    ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    ; ymmH=(0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R
+    ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    vmovdqa     ymmB, ymmA
+    vinserti128 ymmA, ymmA, xmmE, 1     ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J)
+    vperm2i128  ymmE, ymmB, ymmE, 0x31  ; ymmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+
+    vmovdqa     ymmB, ymmF
+    vinserti128 ymmF, ymmF, xmmH, 1     ; ymmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R)
+    vperm2i128  ymmH, ymmB, ymmH, 0x31  ; ymmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    vmovdqa     ymmD, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmE      ; ymmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35
+                                      ;       0G 0K 1G 1K 2G 2K 3G 3K 0H 0L 1H 1L 2H 2L 3H 3L)
+    vpunpckhbw  ymmD, ymmD, ymmE      ; ymmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37
+                                      ;       0I 0M 1I 1M 2I 2M 3I 3M 0J 0N 1J 1N 2J 2N 3J 3N)
+
+    vmovdqa     ymmC, ymmF
+    vpunpcklbw  ymmF, ymmF, ymmH      ; ymmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D
+                                      ;       0O 0S 1O 1S 2O 2S 3O 3S 0P 0T 1P 1T 2P 2T 3P 3T)
+    vpunpckhbw  ymmC, ymmC, ymmH      ; ymmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F
+                                      ;       0Q 0U 1Q 1U 2Q 2U 3Q 3U 0R 0V 1R 1V 2R 2V 3R 3V)
+
+    vmovdqa     ymmB, ymmA
+    vpunpcklwd  ymmA, ymmA, ymmF      ; ymmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C
+                                      ;       0G 0K 0O 0S 1G 1K 1O 1S 2G 2K 2O 2S 3G 3K 3O 3S)
+    vpunpckhwd  ymmB, ymmB, ymmF      ; ymmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D
+                                      ;       0H 0L 0P 0T 1H 1L 1P 1T 2H 2L 2P 2T 3H 3L 3P 3T)
+
+    vmovdqa     ymmG, ymmD
+    vpunpcklwd  ymmD, ymmD, ymmC      ; ymmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E
+                                      ;       0I 0M 0Q 0U 1I 1M 1Q 1U 2I 2M 2Q 2U 3I 3M 3Q 3U)
+    vpunpckhwd  ymmG, ymmG, ymmC      ; ymmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F
+                                      ;       0J 0N 0R 0V 1J 1N 1R 1V 2J 2N 2R 2V 3J 3N 3R 3V)
+
+    vmovdqa     ymmE, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmD      ; ymmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E
+                                      ;       0G 0I 0K 0M 0O 0Q 0S 0U 1G 1I 1K 1M 1O 1Q 1S 1U)
+    vpunpckhbw  ymmE, ymmE, ymmD      ; ymmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E
+                                      ;       2G 2I 2K 2M 2O 2Q 2S 2U 3G 3I 3K 3M 3O 3Q 3S 3U)
+
+    vmovdqa     ymmH, ymmB
+    vpunpcklbw  ymmB, ymmB, ymmG      ; ymmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F
+                                      ;       0H 0J 0L 0N 0P 0R 0T 0V 1H 1J 1L 1N 1P 1R 1T 1V)
+    vpunpckhbw  ymmH, ymmH, ymmG      ; ymmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F
+                                      ;       2H 2J 2L 2N 2P 2R 2T 2V 3H 3J 3L 3N 3P 3R 3T 3V)
+
+    vpxor       ymmF, ymmF, ymmF
+
+    vmovdqa     ymmC, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmF      ; ymmA=(00 02 04 06 08 0A 0C 0E 0G 0I 0K 0M 0O 0Q 0S 0U)
+    vpunpckhbw  ymmC, ymmC, ymmF      ; ymmC=(10 12 14 16 18 1A 1C 1E 1G 1I 1K 1M 1O 1Q 1S 1U)
+
+    vmovdqa     ymmD, ymmB
+    vpunpcklbw  ymmB, ymmB, ymmF      ; ymmB=(01 03 05 07 09 0B 0D 0F 0H 0J 0L 0N 0P 0R 0T 0V)
+    vpunpckhbw  ymmD, ymmD, ymmF      ; ymmD=(11 13 15 17 19 1B 1D 1F 1H 1J 1L 1N 1P 1R 1T 1V)
+
+    vmovdqa     ymmG, ymmE
+    vpunpcklbw  ymmE, ymmE, ymmF      ; ymmE=(20 22 24 26 28 2A 2C 2E 2G 2I 2K 2M 2O 2Q 2S 2U)
+    vpunpckhbw  ymmG, ymmG, ymmF      ; ymmG=(30 32 34 36 38 3A 3C 3E 3G 3I 3K 3M 3O 3Q 3S 3U)
+
+    vpunpcklbw  ymmF, ymmF, ymmH
+    vpunpckhbw  ymmH, ymmH, ymmH
+    vpsrlw      ymmF, ymmF, BYTE_BIT  ; ymmF=(21 23 25 27 29 2B 2D 2F 2H 2J 2L 2N 2P 2R 2T 2V)
+    vpsrlw      ymmH, ymmH, BYTE_BIT  ; ymmH=(31 33 35 37 39 3B 3D 3F 3H 3J 3L 3N 3P 3R 3T 3V)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; ymm0=R(02468ACEGIKMOQSU)=RE, ymm2=G(02468ACEGIKMOQSU)=GE, ymm4=B(02468ACEGIKMOQSU)=BE
+    ; ymm1=R(13579BDFHJLNPRTV)=RO, ymm3=G(13579BDFHJLNPRTV)=GO, ymm5=B(13579BDFHJLNPRTV)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+    vmovdqa     ymm6, ymm1
+    vpunpcklwd  ymm1, ymm1, ymm3
+    vpunpckhwd  ymm6, ymm6, ymm3
+    vpmaddwd    ymm1, ymm1, GOTOFF(eax,PW_F0299_F0337)  ; ymm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    vpmaddwd    ymm6, ymm6, GOTOFF(eax,PW_F0299_F0337)  ; ymm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    vmovdqa     ymm7, ymm6              ; ymm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    vmovdqa     ymm6, ymm0
+    vpunpcklwd  ymm0, ymm0, ymm2
+    vpunpckhwd  ymm6, ymm6, ymm2
+    vpmaddwd    ymm0, ymm0, GOTOFF(eax,PW_F0299_F0337)  ; ymm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    vpmaddwd    ymm6, ymm6, GOTOFF(eax,PW_F0299_F0337)  ; ymm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    vmovdqa     YMMWORD wk(0), ymm0   ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+    vmovdqa     YMMWORD wk(1), ymm6   ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    vmovdqa     ymm0, ymm5              ; ymm0=BO
+    vmovdqa     ymm6, ymm4              ; ymm6=BE
+
+    vmovdqa     ymm4, ymm0
+    vpunpcklwd  ymm0, ymm0, ymm3
+    vpunpckhwd  ymm4, ymm4, ymm3
+    vpmaddwd    ymm0, ymm0, GOTOFF(eax,PW_F0114_F0250)  ; ymm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    vpmaddwd    ymm4, ymm4, GOTOFF(eax,PW_F0114_F0250)  ; ymm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+    vmovdqa     ymm3, GOTOFF(eax,PD_ONEHALF)            ; ymm3=PD_ONEHALF
+
+    vpaddd      ymm0, ymm0, ymm1
+    vpaddd      ymm4, ymm4, ymm7
+    vpaddd      ymm0, ymm0, ymm3
+    vpaddd      ymm4, ymm4, ymm3
+    vpsrld      ymm0, ymm0, SCALEBITS   ; ymm0=YOL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=YOH
+    vpackssdw   ymm0, ymm0, ymm4        ; ymm0=YO
+
+    vmovdqa     ymm4, ymm6
+    vpunpcklwd  ymm6, ymm6, ymm2
+    vpunpckhwd  ymm4, ymm4, ymm2
+    vpmaddwd    ymm6, ymm6, GOTOFF(eax,PW_F0114_F0250)  ; ymm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    vpmaddwd    ymm4, ymm4, GOTOFF(eax,PW_F0114_F0250)  ; ymm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+    vmovdqa     ymm2, GOTOFF(eax,PD_ONEHALF)            ; ymm2=PD_ONEHALF
+
+    vpaddd      ymm6, ymm6, YMMWORD wk(0)
+    vpaddd      ymm4, ymm4, YMMWORD wk(1)
+    vpaddd      ymm6, ymm6, ymm2
+    vpaddd      ymm4, ymm4, ymm2
+    vpsrld      ymm6, ymm6, SCALEBITS   ; ymm6=YEL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=YEH
+    vpackssdw   ymm6, ymm6, ymm4        ; ymm6=YE
+
+    vpsllw      ymm0, ymm0, BYTE_BIT
+    vpor        ymm6, ymm6, ymm0        ; ymm6=Y
+    vmovdqu     YMMWORD edi, ymm6     ; Save Y
+
+    sub         ecx, byte SIZEOF_YMMWORD
+    add         esi, RGB_PIXELSIZE*SIZEOF_YMMWORD  ; inptr
+    add         edi, byte SIZEOF_YMMWORD           ; outptr0
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+    test        ecx, ecx
+    jnz         near .column_ld1
+
+    pop         ecx                     ; col
+    pop         esi
+    pop         edi
+    poppic      eax
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         edi, byte SIZEOF_JSAMPROW
+    dec         eax                        ; num_rows
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcgryext-mmx.asm Added

@@ -0,0 +1,355 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2011, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_mmx(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                            JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                            int num_rows);
+;
+
+%define img_width(b)   (b) + 8          ; JDIMENSION img_width
+%define input_buf(b)   (b) + 12         ; JSAMPARRAY input_buf
+%define output_buf(b)  (b) + 16         ; JSAMPIMAGE output_buf
+%define output_row(b)  (b) + 20         ; JDIMENSION output_row
+%define num_rows(b)    (b) + 24         ; int num_rows
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_MMWORD
+                                        ; mmword wkWK_NUM
+%define WK_NUM         2
+%define gotptr         wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_gray_convert_mmx)
+
+EXTN(jsimd_rgb_gray_convert_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION img_width(eax)  ; num_cols
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         esi, JSAMPIMAGE output_buf(eax)
+    mov         ecx, JDIMENSION output_row(eax)
+    mov         edi, JSAMPARRAY esi+0*SIZEOF_JSAMPARRAY
+    lea         edi, edi+ecx*SIZEOF_JSAMPROW
+
+    pop         ecx
+
+    mov         esi, JSAMPARRAY input_buf(eax)
+    mov         eax, INT num_rows(eax)
+    test        eax, eax
+    jle         near .return
+    alignx      16, 7
+.rowloop:
+    pushpic     eax
+    push        edi
+    push        esi
+    push        ecx                     ; col
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr0
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+
+    cmp         ecx, byte SIZEOF_MMWORD
+    jae         short .columnloop
+    alignx      16, 7
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        eax
+    push        edx
+    lea         ecx, ecx+ecx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_BYTE
+    xor         eax, eax
+    mov         al, byte esi+ecx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_WORD
+    xor         edx, edx
+    mov         dx, word esi+ecx
+    shl         eax, WORD_BIT
+    or          eax, edx
+.column_ld4:
+    movd        mmA, eax
+    pop         edx
+    pop         eax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_DWORD
+    movd        mmG, dword esi+ecx
+    psllq       mmA, DWORD_BIT
+    por         mmA, mmG
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    movq        mmG, mmA
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    mov         ecx, SIZEOF_MMWORD
+    jmp         short .rgb_gray_cnv
+.column_ld16:
+    test        cl, 2*SIZEOF_MMWORD
+    mov         ecx, SIZEOF_MMWORD
+    jz          short .rgb_gray_cnv
+    movq        mmF, mmA
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mmG, MMWORD esi+1*SIZEOF_MMWORD
+    jmp         short .rgb_gray_cnv
+    alignx      16, 7
+
+.columnloop:
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mmG, MMWORD esi+1*SIZEOF_MMWORD
+    movq        mmF, MMWORD esi+2*SIZEOF_MMWORD
+
+.rgb_gray_cnv:
+    ; mmA=(00 10 20 01 11 21 02 12)
+    ; mmG=(22 03 13 23 04 14 24 05)
+    ; mmF=(15 25 06 16 26 07 17 27)
+
+    movq        mmD, mmA
+    psllq       mmA, 4*BYTE_BIT         ; mmA=(-- -- -- -- 00 10 20 01)
+    psrlq       mmD, 4*BYTE_BIT         ; mmD=(11 21 02 12 -- -- -- --)
+
+    punpckhbw   mmA, mmG                ; mmA=(00 04 10 14 20 24 01 05)
+    psllq       mmG, 4*BYTE_BIT         ; mmG=(-- -- -- -- 22 03 13 23)
+
+    punpcklbw   mmD, mmF                ; mmD=(11 15 21 25 02 06 12 16)
+    punpckhbw   mmG, mmF                ; mmG=(22 26 03 07 13 17 23 27)
+
+    movq        mmE, mmA
+    psllq       mmA, 4*BYTE_BIT         ; mmA=(-- -- -- -- 00 04 10 14)
+    psrlq       mmE, 4*BYTE_BIT         ; mmE=(20 24 01 05 -- -- -- --)
+
+    punpckhbw   mmA, mmD                ; mmA=(00 02 04 06 10 12 14 16)
+    psllq       mmD, 4*BYTE_BIT         ; mmD=(-- -- -- -- 11 15 21 25)
+
+    punpcklbw   mmE, mmG                ; mmE=(20 22 24 26 01 03 05 07)
+    punpckhbw   mmD, mmG                ; mmD=(11 13 15 17 21 23 25 27)
+
+    pxor        mmH, mmH
+
+    movq        mmC, mmA
+    punpcklbw   mmA, mmH                ; mmA=(00 02 04 06)
+    punpckhbw   mmC, mmH                ; mmC=(10 12 14 16)
+
+    movq        mmB, mmE
+    punpcklbw   mmE, mmH                ; mmE=(20 22 24 26)
+    punpckhbw   mmB, mmH                ; mmB=(01 03 05 07)
+
+    movq        mmF, mmD
+    punpcklbw   mmD, mmH                ; mmD=(11 13 15 17)
+    punpckhbw   mmF, mmH                ; mmF=(21 23 25 27)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_MMWORD/8
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_MMWORD/8
+    movd        mmA, dword esi+ecx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_MMWORD/4
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_MMWORD/4
+    movq        mmF, mmA
+    movq        mmA, MMWORD esi+ecx*RGB_PIXELSIZE
+.column_ld4:
+    test        cl, SIZEOF_MMWORD/2
+    mov         ecx, SIZEOF_MMWORD
+    jz          short .rgb_gray_cnv
+    movq        mmD, mmA
+    movq        mmC, mmF
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mmF, MMWORD esi+1*SIZEOF_MMWORD
+    jmp         short .rgb_gray_cnv
+    alignx      16, 7
+
+.columnloop:
+    movq        mmA, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mmF, MMWORD esi+1*SIZEOF_MMWORD
+    movq        mmD, MMWORD esi+2*SIZEOF_MMWORD
+    movq        mmC, MMWORD esi+3*SIZEOF_MMWORD
+
+.rgb_gray_cnv:
+    ; mmA=(00 10 20 30 01 11 21 31)
+    ; mmF=(02 12 22 32 03 13 23 33)
+    ; mmD=(04 14 24 34 05 15 25 35)
+    ; mmC=(06 16 26 36 07 17 27 37)
+
+    movq        mmB, mmA
+    punpcklbw   mmA, mmF                ; mmA=(00 02 10 12 20 22 30 32)
+    punpckhbw   mmB, mmF                ; mmB=(01 03 11 13 21 23 31 33)
+
+    movq        mmG, mmD
+    punpcklbw   mmD, mmC                ; mmD=(04 06 14 16 24 26 34 36)
+    punpckhbw   mmG, mmC                ; mmG=(05 07 15 17 25 27 35 37)
+
+    movq        mmE, mmA
+    punpcklwd   mmA, mmD                ; mmA=(00 02 04 06 10 12 14 16)
+    punpckhwd   mmE, mmD                ; mmE=(20 22 24 26 30 32 34 36)
+
+    movq        mmH, mmB
+    punpcklwd   mmB, mmG                ; mmB=(01 03 05 07 11 13 15 17)
+    punpckhwd   mmH, mmG                ; mmH=(21 23 25 27 31 33 35 37)
+
+    pxor        mmF, mmF
+
+    movq        mmC, mmA
+    punpcklbw   mmA, mmF                ; mmA=(00 02 04 06)
+    punpckhbw   mmC, mmF                ; mmC=(10 12 14 16)
+
+    movq        mmD, mmB
+    punpcklbw   mmB, mmF                ; mmB=(01 03 05 07)
+    punpckhbw   mmD, mmF                ; mmD=(11 13 15 17)
+
+    movq        mmG, mmE
+    punpcklbw   mmE, mmF                ; mmE=(20 22 24 26)
+    punpckhbw   mmG, mmF                ; mmG=(30 32 34 36)
+
+    punpcklbw   mmF, mmH
+    punpckhbw   mmH, mmH
+    psrlw       mmF, BYTE_BIT           ; mmF=(21 23 25 27)
+    psrlw       mmH, BYTE_BIT           ; mmH=(31 33 35 37)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; mm0=(R0 R2 R4 R6)=RE, mm2=(G0 G2 G4 G6)=GE, mm4=(B0 B2 B4 B6)=BE
+    ; mm1=(R1 R3 R5 R7)=RO, mm3=(G1 G3 G5 G7)=GO, mm5=(B1 B3 B5 B7)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+    movq        mm6, mm1
+    punpcklwd   mm1, mm3
+    punpckhwd   mm6, mm3
+    pmaddwd     mm1, GOTOFF(eax,PW_F0299_F0337)  ; mm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    pmaddwd     mm6, GOTOFF(eax,PW_F0299_F0337)  ; mm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    movq        mm7,  mm6               ; mm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    movq        mm6, mm0
+    punpcklwd   mm0, mm2
+    punpckhwd   mm6, mm2
+    pmaddwd     mm0, GOTOFF(eax,PW_F0299_F0337)  ; mm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    pmaddwd     mm6, GOTOFF(eax,PW_F0299_F0337)  ; mm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    movq        MMWORD wk(0), mm0     ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+    movq        MMWORD wk(1), mm6     ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    movq        mm0, mm5                ; mm0=BO
+    movq        mm6, mm4                ; mm6=BE
+
+    movq        mm4, mm0
+    punpcklwd   mm0, mm3
+    punpckhwd   mm4, mm3
+    pmaddwd     mm0, GOTOFF(eax,PW_F0114_F0250)  ; mm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    pmaddwd     mm4, GOTOFF(eax,PW_F0114_F0250)  ; mm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+    movq        mm3, GOTOFF(eax,PD_ONEHALF)  ; mm3=PD_ONEHALF
+
+    paddd       mm0, mm1
+    paddd       mm4, mm7
+    paddd       mm0, mm3
+    paddd       mm4, mm3
+    psrld       mm0, SCALEBITS          ; mm0=YOL
+    psrld       mm4, SCALEBITS          ; mm4=YOH
+    packssdw    mm0, mm4                ; mm0=YO
+
+    movq        mm4, mm6
+    punpcklwd   mm6, mm2
+    punpckhwd   mm4, mm2
+    pmaddwd     mm6, GOTOFF(eax,PW_F0114_F0250)  ; mm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    pmaddwd     mm4, GOTOFF(eax,PW_F0114_F0250)  ; mm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+    movq        mm2, GOTOFF(eax,PD_ONEHALF)      ; mm2=PD_ONEHALF
+
+    paddd       mm6, MMWORD wk(0)
+    paddd       mm4, MMWORD wk(1)
+    paddd       mm6, mm2
+    paddd       mm4, mm2
+    psrld       mm6, SCALEBITS          ; mm6=YEL
+    psrld       mm4, SCALEBITS          ; mm4=YEH
+    packssdw    mm6, mm4                ; mm6=YE
+
+    psllw       mm0, BYTE_BIT
+    por         mm6, mm0                ; mm6=Y
+    movq        MMWORD edi, mm6       ; Save Y
+
+    sub         ecx, byte SIZEOF_MMWORD
+    add         esi, byte RGB_PIXELSIZE*SIZEOF_MMWORD  ; inptr
+    add         edi, byte SIZEOF_MMWORD                ; outptr0
+    cmp         ecx, byte SIZEOF_MMWORD
+    jae         near .columnloop
+    test        ecx, ecx
+    jnz         near .column_ld1
+
+    pop         ecx                     ; col
+    pop         esi
+    pop         edi
+    poppic      eax
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         edi, byte SIZEOF_JSAMPROW
+    dec         eax                        ; num_rows
+    jg          near .rowloop
+
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcgryext-sse2.asm Added

@@ -0,0 +1,382 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (SSE2)
+;
+; Copyright (C) 2011, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_sse2(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                             JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                             int num_rows);
+;
+
+%define img_width(b)   (b) + 8          ; JDIMENSION img_width
+%define input_buf(b)   (b) + 12         ; JSAMPARRAY input_buf
+%define output_buf(b)  (b) + 16         ; JSAMPIMAGE output_buf
+%define output_row(b)  (b) + 20         ; JDIMENSION output_row
+%define num_rows(b)    (b) + 24         ; int num_rows
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM         2
+%define gotptr         wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_gray_convert_sse2)
+
+EXTN(jsimd_rgb_gray_convert_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION img_width(eax)
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         esi, JSAMPIMAGE output_buf(eax)
+    mov         ecx, JDIMENSION output_row(eax)
+    mov         edi, JSAMPARRAY esi+0*SIZEOF_JSAMPARRAY
+    lea         edi, edi+ecx*SIZEOF_JSAMPROW
+
+    pop         ecx
+
+    mov         esi, JSAMPARRAY input_buf(eax)
+    mov         eax, INT num_rows(eax)
+    test        eax, eax
+    jle         near .return
+    alignx      16, 7
+.rowloop:
+    pushpic     eax
+    push        edi
+    push        esi
+    push        ecx                     ; col
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr0
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+    alignx      16, 7
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        eax
+    push        edx
+    lea         ecx, ecx+ecx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_BYTE
+    movzx       eax, byte esi+ecx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_WORD
+    movzx       edx, word esi+ecx
+    shl         eax, WORD_BIT
+    or          eax, edx
+.column_ld4:
+    movd        xmmA, eax
+    pop         edx
+    pop         eax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_DWORD
+    movd        xmmF, XMM_DWORD esi+ecx
+    pslldq      xmmA, SIZEOF_DWORD
+    por         xmmA, xmmF
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    sub         ecx, byte SIZEOF_MMWORD
+    movq        xmmB, XMM_MMWORD esi+ecx
+    pslldq      xmmA, SIZEOF_MMWORD
+    por         xmmA, xmmB
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    jz          short .column_ld32
+    movdqa      xmmF, xmmA
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    mov         ecx, SIZEOF_XMMWORD
+    jmp         short .rgb_gray_cnv
+.column_ld32:
+    test        cl, 2*SIZEOF_XMMWORD
+    mov         ecx, SIZEOF_XMMWORD
+    jz          short .rgb_gray_cnv
+    movdqa      xmmB, xmmA
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD esi+1*SIZEOF_XMMWORD
+    jmp         short .rgb_gray_cnv
+    alignx      16, 7
+
+.columnloop:
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD esi+1*SIZEOF_XMMWORD
+    movdqu      xmmB, XMMWORD esi+2*SIZEOF_XMMWORD
+
+.rgb_gray_cnv:
+    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+    ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+    movdqa      xmmG, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+    psrldq      xmmG, 8     ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmF  ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+    pslldq      xmmF, 8     ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+    punpcklbw   xmmG, xmmB  ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+    punpckhbw   xmmF, xmmB  ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+    movdqa      xmmD, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+    psrldq      xmmD, 8     ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmG  ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+    pslldq      xmmG, 8     ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+    punpcklbw   xmmD, xmmF  ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+    punpckhbw   xmmG, xmmF  ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+    movdqa      xmmE, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+    psrldq      xmmE, 8     ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmD  ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+    pslldq      xmmD, 8     ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+    punpcklbw   xmmE, xmmG  ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+    punpckhbw   xmmD, xmmG  ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+    pxor        xmmH, xmmH
+
+    movdqa      xmmC, xmmA
+    punpcklbw   xmmA, xmmH  ; xmmA=(00 02 04 06 08 0A 0C 0E)
+    punpckhbw   xmmC, xmmH  ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+    movdqa      xmmB, xmmE
+    punpcklbw   xmmE, xmmH  ; xmmE=(20 22 24 26 28 2A 2C 2E)
+    punpckhbw   xmmB, xmmH  ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+    movdqa      xmmF, xmmD
+    punpcklbw   xmmD, xmmH  ; xmmD=(11 13 15 17 19 1B 1D 1F)
+    punpckhbw   xmmF, xmmH  ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_XMMWORD/16
+    jz          short .column_ld2
+    sub         ecx, byte SIZEOF_XMMWORD/16
+    movd        xmmA, XMM_DWORD esi+ecx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_XMMWORD/8
+    jz          short .column_ld4
+    sub         ecx, byte SIZEOF_XMMWORD/8
+    movq        xmmE, XMM_MMWORD esi+ecx*RGB_PIXELSIZE
+    pslldq      xmmA, SIZEOF_MMWORD
+    por         xmmA, xmmE
+.column_ld4:
+    test        cl, SIZEOF_XMMWORD/4
+    jz          short .column_ld8
+    sub         ecx, byte SIZEOF_XMMWORD/4
+    movdqa      xmmE, xmmA
+    movdqu      xmmA, XMMWORD esi+ecx*RGB_PIXELSIZE
+.column_ld8:
+    test        cl, SIZEOF_XMMWORD/2
+    mov         ecx, SIZEOF_XMMWORD
+    jz          short .rgb_gray_cnv
+    movdqa      xmmF, xmmA
+    movdqa      xmmH, xmmE
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqu      xmmE, XMMWORD esi+1*SIZEOF_XMMWORD
+    jmp         short .rgb_gray_cnv
+    alignx      16, 7
+
+.columnloop:
+    movdqu      xmmA, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqu      xmmE, XMMWORD esi+1*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD esi+2*SIZEOF_XMMWORD
+    movdqu      xmmH, XMMWORD esi+3*SIZEOF_XMMWORD
+
+.rgb_gray_cnv:
+    ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+    ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+    ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+    movdqa      xmmD, xmmA
+    punpcklbw   xmmA, xmmE      ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+    punpckhbw   xmmD, xmmE      ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+    movdqa      xmmC, xmmF
+    punpcklbw   xmmF, xmmH      ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+    punpckhbw   xmmC, xmmH      ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+    movdqa      xmmB, xmmA
+    punpcklwd   xmmA, xmmF      ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+    punpckhwd   xmmB, xmmF      ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+    movdqa      xmmG, xmmD
+    punpcklwd   xmmD, xmmC      ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+    punpckhwd   xmmG, xmmC      ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+    movdqa      xmmE, xmmA
+    punpcklbw   xmmA, xmmD      ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+    punpckhbw   xmmE, xmmD      ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+    movdqa      xmmH, xmmB
+    punpcklbw   xmmB, xmmG      ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+    punpckhbw   xmmH, xmmG      ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+    pxor        xmmF, xmmF
+
+    movdqa      xmmC, xmmA
+    punpcklbw   xmmA, xmmF      ; xmmA=(00 02 04 06 08 0A 0C 0E)
+    punpckhbw   xmmC, xmmF      ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+    movdqa      xmmD, xmmB
+    punpcklbw   xmmB, xmmF      ; xmmB=(01 03 05 07 09 0B 0D 0F)
+    punpckhbw   xmmD, xmmF      ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+    movdqa      xmmG, xmmE
+    punpcklbw   xmmE, xmmF      ; xmmE=(20 22 24 26 28 2A 2C 2E)
+    punpckhbw   xmmG, xmmF      ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+    punpcklbw   xmmF, xmmH
+    punpckhbw   xmmH, xmmH
+    psrlw       xmmF, BYTE_BIT  ; xmmF=(21 23 25 27 29 2B 2D 2F)
+    psrlw       xmmH, BYTE_BIT  ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+    ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+    movdqa      xmm6, xmm1
+    punpcklwd   xmm1, xmm3
+    punpckhwd   xmm6, xmm3
+    pmaddwd     xmm1, GOTOFF(eax,PW_F0299_F0337)  ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    pmaddwd     xmm6, GOTOFF(eax,PW_F0299_F0337)  ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    movdqa      xmm7, xmm6              ; xmm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    movdqa      xmm6, xmm0
+    punpcklwd   xmm0, xmm2
+    punpckhwd   xmm6, xmm2
+    pmaddwd     xmm0, GOTOFF(eax,PW_F0299_F0337)  ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    pmaddwd     xmm6, GOTOFF(eax,PW_F0299_F0337)  ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    movdqa      XMMWORD wk(0), xmm0   ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+    movdqa      XMMWORD wk(1), xmm6   ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    movdqa      xmm0, xmm5              ; xmm0=BO
+    movdqa      xmm6, xmm4              ; xmm6=BE
+
+    movdqa      xmm4, xmm0
+    punpcklwd   xmm0, xmm3
+    punpckhwd   xmm4, xmm3
+    pmaddwd     xmm0, GOTOFF(eax,PW_F0114_F0250)  ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    pmaddwd     xmm4, GOTOFF(eax,PW_F0114_F0250)  ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+    movdqa      xmm3, GOTOFF(eax,PD_ONEHALF)      ; xmm3=PD_ONEHALF
+
+    paddd       xmm0, xmm1
+    paddd       xmm4, xmm7
+    paddd       xmm0, xmm3
+    paddd       xmm4, xmm3
+    psrld       xmm0, SCALEBITS         ; xmm0=YOL
+    psrld       xmm4, SCALEBITS         ; xmm4=YOH
+    packssdw    xmm0, xmm4              ; xmm0=YO
+
+    movdqa      xmm4, xmm6
+    punpcklwd   xmm6, xmm2
+    punpckhwd   xmm4, xmm2
+    pmaddwd     xmm6, GOTOFF(eax,PW_F0114_F0250)  ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    pmaddwd     xmm4, GOTOFF(eax,PW_F0114_F0250)  ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+    movdqa      xmm2, GOTOFF(eax,PD_ONEHALF)      ; xmm2=PD_ONEHALF
+
+    paddd       xmm6, XMMWORD wk(0)
+    paddd       xmm4, XMMWORD wk(1)
+    paddd       xmm6, xmm2
+    paddd       xmm4, xmm2
+    psrld       xmm6, SCALEBITS         ; xmm6=YEL
+    psrld       xmm4, SCALEBITS         ; xmm4=YEH
+    packssdw    xmm6, xmm4              ; xmm6=YE
+
+    psllw       xmm0, BYTE_BIT
+    por         xmm6, xmm0              ; xmm6=Y
+    movdqa      XMMWORD edi, xmm6     ; Save Y
+
+    sub         ecx, byte SIZEOF_XMMWORD
+    add         esi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+    add         edi, byte SIZEOF_XMMWORD                ; outptr0
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+    test        ecx, ecx
+    jnz         near .column_ld1
+
+    pop         ecx                     ; col
+    pop         esi
+    pop         edi
+    poppic      eax
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         edi, byte SIZEOF_JSAMPROW
+    dec         eax                        ; num_rows
+    jg          near .rowloop
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jchuff-sse2.asm Added

@@ -0,0 +1,761 @@
+;
+; jchuff-sse2.asm - Huffman entropy encoding (SSE2)
+;
+; Copyright (C) 2009-2011, 2014-2017, 2019, D. R. Commander.
+; Copyright (C) 2015, Matthieu Darbois.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains an SSE2 implementation for Huffman coding of one block.
+; The following code is based on jchuff.c; see jchuff.c for more details.
+
+%include "jsimdext.inc"
+
+struc working_state
+.next_output_byte:   resp 1     ; => next byte to write in buffer
+.free_in_buffer:     resp 1     ; # of byte spaces remaining in buffer
+.cur.put_buffer.simd resq 1     ; current bit accumulation buffer
+.cur.free_bits       resd 1     ; # of bits available in it
+.cur.last_dc_val     resd 4     ; last DC coef for each component
+.cinfo:              resp 1     ; dump_buffer needs access to this
+endstruc
+
+struc c_derived_tbl
+.ehufco:             resd 256   ; code for each symbol
+.ehufsi:             resb 256   ; length of code for each symbol
+; If no code has been allocated for a symbol S, ehufsiS contains 0
+endstruc
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    GLOBAL_DATA(jconst_huff_encode_one_block)
+
+EXTN(jconst_huff_encode_one_block):
+
+    alignz      32
+
+jpeg_mask_bits dq 0x0000, 0x0001, 0x0003, 0x0007
+               dq 0x000f, 0x001f, 0x003f, 0x007f
+               dq 0x00ff, 0x01ff, 0x03ff, 0x07ff
+               dq 0x0fff, 0x1fff, 0x3fff, 0x7fff
+
+times 1 << 14 db 15
+times 1 << 13 db 14
+times 1 << 12 db 13
+times 1 << 11 db 12
+times 1 << 10 db 11
+times 1 <<  9 db 10
+times 1 <<  8 db  9
+times 1 <<  7 db  8
+times 1 <<  6 db  7
+times 1 <<  5 db  6
+times 1 <<  4 db  5
+times 1 <<  3 db  4
+times 1 <<  2 db  3
+times 1 <<  1 db  2
+times 1 <<  0 db  1
+times 1       db  0
+jpeg_nbits_table:
+times 1       db  0
+times 1 <<  0 db  1
+times 1 <<  1 db  2
+times 1 <<  2 db  3
+times 1 <<  3 db  4
+times 1 <<  4 db  5
+times 1 <<  5 db  6
+times 1 <<  6 db  7
+times 1 <<  7 db  8
+times 1 <<  8 db  9
+times 1 <<  9 db 10
+times 1 << 10 db 11
+times 1 << 11 db 12
+times 1 << 12 db 13
+times 1 << 13 db 14
+times 1 << 14 db 15
+
+    alignz      32
+
+%ifdef PIC
+%define NBITS(x)      nbits_base + x
+%else
+%define NBITS(x)      jpeg_nbits_table + x
+%endif
+%define MASK_BITS(x)  NBITS((x) * 8) + (jpeg_mask_bits - jpeg_nbits_table)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%define mm_put_buffer     mm0
+%define mm_all_0xff       mm1
+%define mm_temp           mm2
+%define mm_nbits          mm3
+%define mm_code_bits      mm3
+%define mm_code           mm4
+%define mm_overflow_bits  mm5
+%define mm_save_nbits     mm6
+
+; Shorthand used to describe SIMD operations:
+; wN:  xmmN treated as eight signed 16-bit values
+; wNi:  perform the same operation on all eight signed 16-bit values, i=0..7
+; bN:  xmmN treated as 16 unsigned 8-bit values, or
+;      mmN treated as eight unsigned 8-bit values
+; bNi:  perform the same operation on all unsigned 8-bit values,
+;         i=0..15 (SSE register) or i=0..7 (MMX register)
+; Contents of SIMD registers are shown in memory order.
+
+; Fill the bit buffer to capacity with the leading bits from code, then output
+; the bit buffer and put the remaining bits from code into the bit buffer.
+;
+; Usage:
+; code - contains the bits to shift into the bit buffer (LSB-aligned)
+; %1 - temp register
+; %2 - low byte of temp register
+; %3 - second byte of temp register
+; %4-%8 (optional) - extra instructions to execute before the macro completes
+; %9 - the label to which to jump when the macro completes
+;
+; Upon completion, free_bits will be set to the number of remaining bits from
+; code, and put_buffer will contain those remaining bits.  temp and code will
+; be clobbered.
+;
+; This macro encodes any 0xFF bytes as 0xFF 0x00, as does the EMIT_BYTE()
+; macro in jchuff.c.
+
+%macro EMIT_QWORD 9
+%define %%temp   %1
+%define %%tempb  %2
+%define %%temph  %3
+    add         nbits, free_bits             ; nbits += free_bits;
+    neg         free_bits                    ; free_bits = -free_bits;
+    movq        mm_temp, mm_code             ; temp = code;
+    movd        mm_nbits, nbits              ; nbits --> MMX register
+    movd        mm_overflow_bits, free_bits  ; overflow_bits (temp register) = free_bits;
+    neg         free_bits                    ; free_bits = -free_bits;
+    psllq       mm_put_buffer, mm_nbits      ; put_buffer <<= nbits;
+    psrlq       mm_temp, mm_overflow_bits    ; temp >>= overflow_bits;
+    add         free_bits, 64                ; free_bits += 64;
+    por         mm_temp, mm_put_buffer       ; temp |= put_buffer;
+%ifidn %%temp, nbits_base
+    movd        mm_save_nbits, nbits_base    ; save nbits_base
+%endif
+    movq        mm_code_bits, mm_temp        ; code_bits (temp register) = temp;
+    movq        mm_put_buffer, mm_code       ; put_buffer = code;
+    pcmpeqb     mm_temp, mm_all_0xff         ; b_tempi = (b_tempi == 0xFF ? 0xFF : 0);
+    movq        mm_code, mm_code_bits        ; code = code_bits;
+    psrlq       mm_code_bits, 32             ; code_bits >>= 32;
+    pmovmskb    nbits, mm_temp               ; nbits = 0;  nbits |= ((b_tempi >> 7) << i);
+    movd        %%temp, mm_code_bits         ; temp = code_bits;
+    bswap       %%temp                       ; temp = htonl(temp);
+    test        nbits, nbits                 ; if (nbits != 0)  /* Some 0xFF bytes */
+    jnz         %%.SLOW                      ;   goto %%.SLOW
+    mov         dword buffer, %%temp       ; *(uint32_t)buffer = temp;
+%ifidn %%temp, nbits_base
+    movd        nbits_base, mm_save_nbits    ; restore nbits_base
+%endif
+    %4
+    movd        nbits, mm_code               ; nbits = (uint32_t)(code);
+    %5
+    bswap       nbits                        ; nbits = htonl(nbits);
+    mov         dword buffer + 4, nbits    ; *(uint32_t)(buffer + 4) = nbits;
+    lea         buffer, buffer + 8         ; buffer += 8;
+    %6
+    %7
+    %8
+    jmp %9                                   ; return
+%%.SLOW:
+    ; Execute the equivalent of the EMIT_BYTE() macro in jchuff.c for all 8
+    ; bytes in the qword.
+    mov         byte buffer, %%tempb     ; buffer0 = temp0;
+    cmp         %%tempb, 0xFF              ; Set CF if temp0 < 0xFF
+    mov         byte buffer+1, 0         ; buffer1 = 0;
+    sbb         buffer, -2                 ; buffer -= (-2 + (temp0 < 0xFF ? 1 : 0));
+    mov         byte buffer, %%temph     ; buffer0 = temp1;
+    cmp         %%temph, 0xFF              ; Set CF if temp1 < 0xFF
+    mov         byte buffer+1, 0         ; buffer1 = 0;
+    sbb         buffer, -2                 ; buffer -= (-2 + (temp1 < 0xFF ? 1 : 0));
+    shr         %%temp, 16                 ; temp >>= 16;
+    mov         byte buffer, %%tempb     ; buffer0 = temp0;
+    cmp         %%tempb, 0xFF              ; Set CF if temp0 < 0xFF
+    mov         byte buffer+1, 0         ; buffer1 = 0;
+    sbb         buffer, -2                 ; buffer -= (-2 + (temp0 < 0xFF ? 1 : 0));
+    mov         byte buffer, %%temph     ; buffer0 = temp1;
+    cmp         %%temph, 0xFF              ; Set CF if temp1 < 0xFF
+    mov         byte buffer+1, 0         ; buffer1 = 0;
+    sbb         buffer, -2                 ; buffer -= (-2 + (temp1 < 0xFF ? 1 : 0));
+    movd        nbits, mm_code             ; nbits (temp register) = (uint32_t)(code)
+%ifidn %%temp, nbits_base
+    movd        nbits_base, mm_save_nbits  ; restore nbits_base
+%endif
+    bswap       nbits                      ; nbits = htonl(nbits)
+    mov         byte buffer, nbitsb      ; buffer0 = nbits0;
+    cmp         nbitsb, 0xFF               ; Set CF if nbits0 < 0xFF
+    mov         byte buffer+1, 0         ; buffer1 = 0;
+    sbb         buffer, -2                 ; buffer -= (-2 + (nbits0 < 0xFF ? 1 : 0));
+    mov         byte buffer, nbitsh      ; buffer0 = nbits1;
+    cmp         nbitsh, 0xFF               ; Set CF if nbits1 < 0xFF
+    mov         byte buffer+1, 0         ; buffer1 = 0;
+    sbb         buffer, -2                 ; buffer -= (-2 + (nbits1 < 0xFF ? 1 : 0));
+    shr         nbits, 16                  ; nbits >>= 16;
+    mov         byte buffer, nbitsb      ; buffer0 = nbits0;
+    cmp         nbitsb, 0xFF               ; Set CF if nbits0 < 0xFF
+    mov         byte buffer+1, 0         ; buffer1 = 0;
+    sbb         buffer, -2                 ; buffer -= (-2 + (nbits0 < 0xFF ? 1 : 0));
+    mov         byte buffer, nbitsh      ; buffer0 = nbits1;
+    %4
+    cmp         nbitsh, 0xFF               ; Set CF if nbits1 < 0xFF
+    mov         byte buffer+1, 0         ; buffer1 = 0;
+    sbb         buffer, -2                 ; buffer -= (-2 + (nbits1 < 0xFF ? 1 : 0));
+    %5
+    %6
+    %7
+    %8
+    jmp %9                                 ; return;
+%endmacro
+
+%macro PUSH 1
+    push        %1
+%assign stack_offset  stack_offset + 4
+%endmacro
+
+%macro POP 1
+    pop         %1
+%assign stack_offset  stack_offset - 4
+%endmacro
+
+; If PIC is defined, load the address of a symbol defined in this file into a
+; register.  Equivalent to
+;   get_GOT     %1
+;   lea         %1, GOTOFF(%1, %2)
+; without using the GOT.
+;
+; Usage:
+; %1 - register into which to load the address of the symbol
+; %2 - symbol whose address should be loaded
+; %3 - optional multi-line macro to execute before the symbol address is loaded
+; %4 - optional multi-line macro to execute after the symbol address is loaded
+;
+; If PIC is not defined, then %3 and %4 are executed in order.
+
+%macro GET_SYM 2-4
+%ifdef PIC
+    call        %%.geteip
+%%.ref:
+    %4
+    add         %1, %2 - %%.ref
+    jmp         short %%.done
+    align       32
+%%.geteip:
+    %3          4               ; must adjust stack pointer because of call
+    mov         %1, POINTER esp
+    ret
+    align       32
+%%.done:
+%else
+    %3          0
+    %4
+%endif
+%endmacro
+
+;
+; Encode a single block's worth of coefficients.
+;
+; GLOBAL(JOCTET *)
+; jsimd_huff_encode_one_block_sse2(working_state *state, JOCTET *buffer,
+;                                  JCOEFPTR block, int last_dc_val,
+;                                  c_derived_tbl *dctbl, c_derived_tbl *actbl)
+;
+; Stack layout:
+; Function args
+; Return address
+; Saved ebx
+; Saved ebp
+; Saved esi
+; Saved edi <-- esp_save
+; ...
+; esp_save
+; t_ 64*2 bytes (aligned to 128 bytes)
+;
+; esp is used (as t) to point into t_ (data in lower indices is not used once
+; esp passes over them, so this is signal-safe.)  Aligning to 128 bytes allows
+; us to find the rest of the data again.
+;
+; NOTES:
+; When shuffling data, we try to avoid pinsrw as much as possible, since it is
+; slow on many CPUs.  Its reciprocal throughput (issue latency) is 1 even on
+; modern CPUs, so chains of pinsrw instructions (even with different outputs)
+; can limit performance.  pinsrw is a VectorPath instruction on AMD K8 and
+; requires 2 µops (with memory operand) on Intel.  In either case, only one
+; pinsrw instruction can be decoded per cycle (and nothing else if they are
+; back-to-back), so out-of-order execution cannot be used to work around long
+; pinsrw chains (though for Sandy Bridge and later, this may be less of a
+; problem if the code runs from the µop cache.)
+;
+; We use tzcnt instead of bsf without checking for support.  The instruction is
+; executed as bsf on CPUs that don't support tzcnt (encoding is equivalent to
+; rep bsf.)  The destination (first) operand of bsf (and tzcnt on some CPUs) is
+; an input dependency (although the behavior is not formally defined, Intel
+; CPUs usually leave the destination unmodified if the source is zero.)  This
+; can prevent out-of-order execution, so we clear the destination before
+; invoking tzcnt.
+;
+; Initial register allocation
+; eax - frame --> buffer
+; ebx - nbits_base (PIC) / emit_temp
+; ecx - dctbl --> size --> state
+; edx - block --> nbits
+; esi - code_temp --> state --> actbl
+; edi - index_temp --> free_bits
+; esp - t
+; ebp - index
+
+%define frame       eax
+%ifdef PIC
+%define nbits_base  ebx
+%endif
+%define emit_temp   ebx
+%define emit_tempb  bl
+%define emit_temph  bh
+%define dctbl       ecx
+%define block       edx
+%define code_temp   esi
+%define index_temp  edi
+%define t           esp
+%define index       ebp
+
+%assign save_frame  DCTSIZE2 * SIZEOF_WORD
+
+; Step 1: Re-arrange input data according to jpeg_natural_order
+; xx 01 02 03 04 05 06 07      xx 01 08 16 09 02 03 10
+; 08 09 10 11 12 13 14 15      17 24 32 25 18 11 04 05
+; 16 17 18 19 20 21 22 23      12 19 26 33 40 48 41 34
+; 24 25 26 27 28 29 30 31 ==>  27 20 13 06 07 14 21 28
+; 32 33 34 35 36 37 38 39      35 42 49 56 57 50 43 36
+; 40 41 42 43 44 45 46 47      29 22 15 23 30 37 44 51
+; 48 49 50 51 52 53 54 55      58 59 52 45 38 31 39 46
+; 56 57 58 59 60 61 62 63      53 60 61 54 47 55 62 63
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_huff_encode_one_block_sse2)
+
+EXTN(jsimd_huff_encode_one_block_sse2):
+
+%assign stack_offset      0
+%define arg_state         4 + stack_offset
+%define arg_buffer        8 + stack_offset
+%define arg_block        12 + stack_offset
+%define arg_last_dc_val  16 + stack_offset
+%define arg_dctbl        20 + stack_offset
+%define arg_actbl        24 + stack_offset
+
+                                                          ;X: X = code stream
+    mov         block, esp + arg_block
+    PUSH        ebx
+    PUSH        ebp
+    movups      xmm3, XMMWORD block + 0 * SIZEOF_WORD   ;D: w3 = xx 01 02 03 04 05 06 07
+    PUSH        esi
+    PUSH        edi
+    movdqa      xmm0, xmm3                                ;A: w0 = xx 01 02 03 04 05 06 07
+    mov         frame, esp
+    lea         t, frame - (save_frame + 4)
+    movups      xmm1, XMMWORD block + 8 * SIZEOF_WORD   ;B: w1 = 08 09 10 11 12 13 14 15
+    and         t, -DCTSIZE2 * SIZEOF_WORD                                             ; t = &t_0
+    mov         t + save_frame, frame
+    pxor        xmm4, xmm4                                ;A: w4i = 0;
+    punpckldq   xmm0, xmm1                                ;A: w0 = xx 01 08 09 02 03 10 11
+    pshuflw     xmm0, xmm0, 11001001b                     ;A: w0 = 01 08 xx 09 02 03 10 11
+    pinsrw      xmm0, word block + 16 * SIZEOF_WORD, 2  ;A: w0 = 01 08 16 09 02 03 10 11
+    punpckhdq   xmm3, xmm1                                ;D: w3 = 04 05 12 13 06 07 14 15
+    punpcklqdq  xmm1, xmm3                                ;B: w1 = 08 09 10 11 04 05 12 13
+    pinsrw      xmm0, word block + 17 * SIZEOF_WORD, 7  ;A: w0 = 01 08 16 09 02 03 10 17
+                                                          ;A:      (Row 0, offset 1)
+    pcmpgtw     xmm4, xmm0                                ;A: w4i = (w0i < 0 ? -1 : 0);
+    paddw       xmm0, xmm4                                ;A: w0i += w4i;
+    movaps      XMMWORD t + 0 * SIZEOF_WORD, xmm0       ;A: ti = w0i;
+
+    movq        xmm2, qword block + 24 * SIZEOF_WORD    ;B: w2 = 24 25 26 27 -- -- -- --
+    pshuflw     xmm2, xmm2, 11011000b                     ;B: w2 = 24 26 25 27 -- -- -- --
+    pslldq      xmm1, 1 * SIZEOF_WORD                     ;B: w1 = -- 08 09 10 11 04 05 12
+    movups      xmm5, XMMWORD block + 48 * SIZEOF_WORD  ;H: w5 = 48 49 50 51 52 53 54 55
+    movsd       xmm1, xmm2                                ;B: w1 = 24 26 25 27 11 04 05 12
+    punpcklqdq  xmm2, xmm5                                ;C: w2 = 24 26 25 27 48 49 50 51
+    pinsrw      xmm1, word block + 32 * SIZEOF_WORD, 1  ;B: w1 = 24 32 25 27 11 04 05 12
+    pxor        xmm4, xmm4                                ;A: w4i = 0;
+    psrldq      xmm3, 2 * SIZEOF_WORD                     ;D: w3 = 12 13 06 07 14 15 -- --
+    pcmpeqw     xmm0, xmm4                                ;A: w0i = (w0i == 0 ? -1 : 0);
+    pinsrw      xmm1, word block + 18 * SIZEOF_WORD, 3  ;B: w1 = 24 32 25 18 11 04 05 12
+                                                          ;        (Row 1, offset 1)
+    pcmpgtw     xmm4, xmm1                                ;B: w4i = (w1i < 0 ? -1 : 0);
+    paddw       xmm1, xmm4                                ;B: w1i += w4i;
+    movaps      XMMWORD t + 8 * SIZEOF_WORD, xmm1       ;B: ti+8 = w1i;
+    pxor        xmm4, xmm4                                ;B: w4i = 0;
+    pcmpeqw     xmm1, xmm4                                ;B: w1i = (w1i == 0 ? -1 : 0);
+
+    packsswb    xmm0, xmm1                                ;AB: b0i = w0i, b0i+8 = w1i
+                                                          ;    w/ signed saturation
+
+    pinsrw      xmm3, word block + 20 * SIZEOF_WORD, 0  ;D: w3 = 20 13 06 07 14 15 -- --
+    pinsrw      xmm3, word block + 21 * SIZEOF_WORD, 5  ;D: w3 = 20 13 06 07 14 21 -- --
+    pinsrw      xmm3, word block + 28 * SIZEOF_WORD, 6  ;D: w3 = 20 13 06 07 14 21 28 --
+    pinsrw      xmm3, word block + 35 * SIZEOF_WORD, 7  ;D: w3 = 20 13 06 07 14 21 28 35
+                                                          ;        (Row 3, offset 1)
+    pcmpgtw     xmm4, xmm3                                ;D: w4i = (w3i < 0 ? -1 : 0);
+    paddw       xmm3, xmm4                                ;D: w3i += w4i;
+    movaps      XMMWORD t + 24 * SIZEOF_WORD, xmm3      ;D: ti+24 = w3i;
+    pxor        xmm4, xmm4                                ;D: w4i = 0;
+    pcmpeqw     xmm3, xmm4                                ;D: w3i = (w3i == 0 ? -1 : 0);
+
+    pinsrw      xmm2, word block + 19 * SIZEOF_WORD, 0  ;C: w2 = 19 26 25 27 48 49 50 51
+    pinsrw      xmm2, word block + 33 * SIZEOF_WORD, 2  ;C: w2 = 19 26 33 27 48 49 50 51
+    pinsrw      xmm2, word block + 40 * SIZEOF_WORD, 3  ;C: w2 = 19 26 33 40 48 49 50 51
+    pinsrw      xmm2, word block + 41 * SIZEOF_WORD, 5  ;C: w2 = 19 26 33 40 48 41 50 51
+    pinsrw      xmm2, word block + 34 * SIZEOF_WORD, 6  ;C: w2 = 19 26 33 40 48 41 34 51
+    pinsrw      xmm2, word block + 27 * SIZEOF_WORD, 7  ;C: w2 = 19 26 33 40 48 41 34 27
+                                                          ;        (Row 2, offset 1)
+    pcmpgtw     xmm4, xmm2                                ;C: w4i = (w2i < 0 ? -1 : 0);
+    paddw       xmm2, xmm4                                ;C: w2i += w4i;
+    movsx       code_temp, word block                   ;Z:     code_temp = block0;
+
+; %1 - stack pointer adjustment
+%macro GET_SYM_BEFORE 1
+    movaps      XMMWORD t + 16 * SIZEOF_WORD + %1, xmm2
+                                                          ;C: ti+16 = w2i;
+    pxor        xmm4, xmm4                                ;C: w4i = 0;
+    pcmpeqw     xmm2, xmm4                                ;C: w2i = (w2i == 0 ? -1 : 0);
+    sub         code_temp, frame + arg_last_dc_val      ;Z:     code_temp -= last_dc_val;
+
+    packsswb    xmm2, xmm3                                ;CD: b2i = w2i, b2i+8 = w3i
+                                                          ;    w/ signed saturation
+
+    movdqa      xmm3, xmm5                                ;H: w3 = 48 49 50 51 52 53 54 55
+    pmovmskb    index_temp, xmm2                          ;Z:     index_temp = 0;  index_temp |= ((b2i >> 7) << i);
+    pmovmskb    index, xmm0                               ;Z:     index = 0;  index |= ((b0i >> 7) << i);
+    movups      xmm0, XMMWORD block + 56 * SIZEOF_WORD  ;H: w0 = 56 57 58 59 60 61 62 63
+    punpckhdq   xmm3, xmm0                                ;H: w3 = 52 53 60 61 54 55 62 63
+    shl         index_temp, 16                            ;Z:     index_temp <<= 16;
+    psrldq      xmm3, 1 * SIZEOF_WORD                     ;H: w3 = 53 60 61 54 55 62 63 --
+    pxor        xmm2, xmm2                                ;H: w2i = 0;
+    pshuflw     xmm3, xmm3, 00111001b                     ;H: w3 = 60 61 54 53 55 62 63 --
+    or          index, index_temp                         ;Z:     index |= index_temp;
+%undef index_temp
+%define free_bits  edi
+%endmacro
+
+%macro GET_SYM_AFTER 0
+    movq        xmm1, qword block + 44 * SIZEOF_WORD    ;G: w1 = 44 45 46 47 -- -- -- --
+    unpcklps    xmm5, xmm0                                ;E: w5 = 48 49 56 57 50 51 58 59
+    pxor        xmm0, xmm0                                ;H: w0i = 0;
+    not         index                                     ;Z:     index = ~index;
+    pinsrw      xmm3, word block + 47 * SIZEOF_WORD, 3  ;H: w3 = 60 61 54 47 55 62 63 --
+                                                          ;        (Row 7, offset 1)
+    pcmpgtw     xmm2, xmm3                                ;H: w2i = (w3i < 0 ? -1 : 0);
+    mov         dctbl, frame + arg_dctbl
+    paddw       xmm3, xmm2                                ;H: w3i += w2i;
+    movaps      XMMWORD t + 56 * SIZEOF_WORD, xmm3      ;H: ti+56 = w3i;
+    movq        xmm4, qword block + 36 * SIZEOF_WORD    ;G: w4 = 36 37 38 39 -- -- -- --
+    pcmpeqw     xmm3, xmm0                                ;H: w3i = (w3i == 0 ? -1 : 0);
+    punpckldq   xmm4, xmm1                                ;G: w4 = 36 37 44 45 38 39 46 47
+    movdqa      xmm1, xmm4                                ;F: w1 = 36 37 44 45 38 39 46 47
+    pcmpeqw     mm_all_0xff, mm_all_0xff                  ;Z:     all_0xffi = 0xFF;
+%endmacro
+
+    GET_SYM     nbits_base, jpeg_nbits_table, GET_SYM_BEFORE, GET_SYM_AFTER
+
+    psrldq      xmm4, 1 * SIZEOF_WORD                     ;G: w4 = 37 44 45 38 39 46 47 --
+    shufpd      xmm1, xmm5, 10b                           ;F: w1 = 36 37 44 45 50 51 58 59
+    pshufhw     xmm4, xmm4, 11010011b                     ;G: w4 = 37 44 45 38 -- 39 46 --
+    pslldq      xmm1, 1 * SIZEOF_WORD                     ;F: w1 = -- 36 37 44 45 50 51 58
+    pinsrw      xmm4, word block + 59 * SIZEOF_WORD, 0  ;G: w4 = 59 44 45 38 -- 39 46 --
+    pshufd      xmm1, xmm1, 11011000b                     ;F: w1 = -- 36 45 50 37 44 51 58
+    cmp         code_temp, 1 << 31                        ;Z:     Set CF if code_temp < 0x80000000,
+                                                          ;Z:     i.e. if code_temp is positive
+    pinsrw      xmm4, word block + 52 * SIZEOF_WORD, 1  ;G: w4 = 59 52 45 38 -- 39 46 --
+    movlps      xmm1, qword block + 20 * SIZEOF_WORD    ;F: w1 = 20 21 22 23 37 44 51 58
+    pinsrw      xmm4, word block + 31 * SIZEOF_WORD, 4  ;G: w4 = 59 52 45 38 31 39 46 --
+    pshuflw     xmm1, xmm1, 01110010b                     ;F: w1 = 22 20 23 21 37 44 51 58
+    pinsrw      xmm4, word block + 53 * SIZEOF_WORD, 7  ;G: w4 = 59 52 45 38 31 39 46 53
+                                                          ;        (Row 6, offset 1)
+    adc         code_temp, -1                             ;Z:     code_temp += -1 + (code_temp >= 0 ? 1 : 0);
+    pxor        xmm2, xmm2                                ;G: w2i = 0;
+    pcmpgtw     xmm0, xmm4                                ;G: w0i = (w4i < 0 ? -1 : 0);
+    pinsrw      xmm1, word block + 15 * SIZEOF_WORD, 1  ;F: w1 = 22 15 23 21 37 44 51 58
+    paddw       xmm4, xmm0                                ;G: w4i += w0i;
+    movaps      XMMWORD t + 48 * SIZEOF_WORD, xmm4      ;G: t48+i = w4i;
+    movd        mm_temp, code_temp                        ;Z:     temp = code_temp
+    pinsrw      xmm1, word block + 30 * SIZEOF_WORD, 3  ;F: w1 = 22 15 23 30 37 44 51 58
+                                                          ;        (Row 5, offset 1)
+    pcmpeqw     xmm4, xmm2                                ;G: w4i = (w4i == 0 ? -1 : 0);
+
+    packsswb    xmm4, xmm3                                ;GH: b4i = w4i, b4i+8 = w3i
+                                                          ;    w/ signed saturation
+
+    lea         t, t - SIZEOF_WORD                      ;Z:     t = &t-1
+    pxor        xmm0, xmm0                                ;F: w0i = 0;
+    pcmpgtw     xmm2, xmm1                                ;F: w2i = (w1i < 0 ? -1 : 0);
+    paddw       xmm1, xmm2                                ;F: w1i += w2i;
+    movaps      XMMWORD t + (40+1) * SIZEOF_WORD, xmm1  ;F: t40+i = w1i;
+    pcmpeqw     xmm1, xmm0                                ;F: w1i = (w1i == 0 ? -1 : 0);
+    pinsrw      xmm5, word block + 42 * SIZEOF_WORD, 0  ;E: w5 = 42 49 56 57 50 51 58 59
+    pinsrw      xmm5, word block + 43 * SIZEOF_WORD, 5  ;E: w5 = 42 49 56 57 50 43 58 59
+    pinsrw      xmm5, word block + 36 * SIZEOF_WORD, 6  ;E: w5 = 42 49 56 57 50 43 36 59
+    pinsrw      xmm5, word block + 29 * SIZEOF_WORD, 7  ;E: w5 = 42 49 56 57 50 43 36 29
+                                                          ;        (Row 4, offset 1)
+%undef block
+%define nbits  edx
+%define nbitsb  dl
+%define nbitsh  dh
+    movzx       nbits, byte NBITS(code_temp)            ;Z:     nbits = JPEG_NBITS(code_temp);
+%undef code_temp
+%define state  esi
+    pxor        xmm2, xmm2                                ;E: w2i = 0;
+    mov         state, frame + arg_state
+    movd        mm_nbits, nbits                           ;Z:     nbits --> MMX register
+    pcmpgtw     xmm0, xmm5                                ;E: w0i = (w5i < 0 ? -1 : 0);
+    movd        mm_code, dword dctbl + c_derived_tbl.ehufco + nbits * 4
+                                                          ;Z:     code = dctbl->ehufconbits;
+%define size  ecx
+%define sizeb  cl
+%define sizeh  ch
+    paddw       xmm5, xmm0                                ;E: w5i += w0i;
+    movaps      XMMWORD t + (32+1) * SIZEOF_WORD, xmm5  ;E: t32+i = w5i;
+    movzx       size, byte dctbl + c_derived_tbl.ehufsi + nbits
+                                                          ;Z:     size = dctbl->ehufsinbits;
+%undef dctbl
+    pcmpeqw     xmm5, xmm2                                ;E: w5i = (w5i == 0 ? -1 : 0);
+
+    packsswb    xmm5, xmm1                                ;EF: b5i = w5i, b5i+8 = w1i
+                                                          ;    w/ signed saturation
+
+    movq        mm_put_buffer, state + working_state.cur.put_buffer.simd
+                                                          ;Z:     put_buffer = state->cur.put_buffer.simd;
+    mov         free_bits, state + working_state.cur.free_bits
+                                                          ;Z:     free_bits = state->cur.free_bits;
+%undef state
+%define actbl  esi
+    mov         actbl, frame + arg_actbl
+%define buffer  eax
+    mov         buffer, frame + arg_buffer
+%undef frame
+    jmp        .BEGIN
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align       16
+; size <= 32, so this is not really a loop
+.BRLOOP1:                                                 ; .BRLOOP1:
+    movzx       nbits, byte actbl + c_derived_tbl.ehufsi + 0xf0
+                                                          ; nbits = actbl->ehufsi0xf0;
+    movd        mm_code, dword actbl + c_derived_tbl.ehufco + 0xf0 * 4
+                                                          ; code = actbl->ehufco0xf0;
+    and         index, 0x7ffffff                          ; clear index if size == 32
+    sub         size, 16                                  ; size -= 16;
+    sub         free_bits, nbits                          ; if ((free_bits -= nbits) <= 0)
+    jle         .EMIT_BRLOOP1                             ;   goto .EMIT_BRLOOP1;
+    movd        mm_nbits, nbits                           ; nbits --> MMX register
+    psllq       mm_put_buffer, mm_nbits                   ; put_buffer <<= nbits;
+    por         mm_put_buffer, mm_code                    ; put_buffer |= code;
+    jmp         .ERLOOP1                                  ; goto .ERLOOP1;
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align       16
+%ifdef PIC
+    times 6     nop
+%else
+    times 2     nop
+%endif
+.BLOOP1:                                                  ; do {  /* size = # of zero bits/elements to skip */
+; if size == 32, index remains unchanged.  Correct in .BRLOOP.
+    shr         index, sizeb                              ;   index >>= size;
+    lea         t, t + size * SIZEOF_WORD               ;   t += size;
+    cmp         size, 16                                  ;   if (size > 16)
+    jg          .BRLOOP1                                  ;     goto .BRLOOP1;
+.ERLOOP1:                                                 ; .ERLOOP1:
+    movsx       nbits, word t                           ;   nbits = *t;
+%ifdef PIC
+    add         size, size                                ;   size += size;
+%else
+    lea         size, size * 2                          ;   size += size;
+%endif
+    movd        mm_temp, nbits                            ;   temp = nbits;
+    movzx       nbits, byte NBITS(nbits)                ;   nbits = JPEG_NBITS(nbits);
+    lea         size, size * 8 + nbits                  ;   size = size * 8 + nbits;
+    movd        mm_nbits, nbits                           ;   nbits --> MMX register
+    movd        mm_code, dword actbl + c_derived_tbl.ehufco + (size - 16) * 4
+                                                          ;   code = actbl->ehufcosize-16;
+    movzx       size, byte actbl + c_derived_tbl.ehufsi + (size - 16)
+                                                          ;   size = actbl->ehufsisize-16;
+.BEGIN:                                                   ; .BEGIN:
+    pand        mm_temp, MASK_BITS(nbits)               ;   temp &= (1 << nbits) - 1;
+    psllq       mm_code, mm_nbits                         ;   code <<= nbits;
+    add         nbits, size                               ;   nbits += size;
+    por         mm_code, mm_temp                          ;   code |= temp;
+    sub         free_bits, nbits                          ;   if ((free_bits -= nbits) <= 0)
+    jle         .EMIT_ERLOOP1                             ;     insert code, flush buffer, init size, goto .BLOOP1
+    xor         size, size                                ;   size = 0;  /* kill tzcnt input dependency */
+    tzcnt       size, index                               ;   size = # of trailing 0 bits in index
+    movd        mm_nbits, nbits                           ;   nbits --> MMX register
+    psllq       mm_put_buffer, mm_nbits                   ;   put_buffer <<= nbits;
+    inc         size                                      ;   ++size;
+    por         mm_put_buffer, mm_code                    ;   put_buffer |= code;
+    test        index, index
+    jnz         .BLOOP1                                   ; } while (index != 0);
+; Round 2
+; t points to the last used word, possibly below t_ if the previous index had 32 zero bits.
+.ELOOP1:                                                  ; .ELOOP1:
+    pmovmskb    size, xmm4                                ; size = 0;  size |= ((b4i >> 7) << i);
+    pmovmskb    index, xmm5                               ; index = 0;  index |= ((b5i >> 7) << i);
+    shl         size, 16                                  ; size <<= 16;
+    or          index, size                               ; index |= size;
+    not         index                                     ; index = ~index;
+    lea         nbits, t + (1 + DCTSIZE2) * SIZEOF_WORD
+                                                          ; nbits = t + 1 + 64;
+    and         nbits, -DCTSIZE2 * SIZEOF_WORD            ; nbits &= -128;  /* now points to &t_64 */
+    sub         nbits, t                                  ; nbits -= t;
+    shr         nbits, 1                                  ; nbits >>= 1;  /* # of leading 0 bits in old index + 33 */
+    tzcnt       size, index                               ; size = # of trailing 0 bits in index
+    inc         size                                      ; ++size;
+    test        index, index                              ; if (index == 0)
+    jz          .ELOOP2                                   ;   goto .ELOOP2;
+; NOTE: size == 32 cannot happen, since the last element is always 0.
+    shr         index, sizeb                              ; index >>= size;
+    lea         size, size + nbits - 33                 ; size = size + nbits - 33;
+    lea         t, t + size * SIZEOF_WORD               ; t += size;
+    cmp         size, 16                                  ; if (size <= 16)
+    jle         .ERLOOP2                                  ;   goto .ERLOOP2;
+.BRLOOP2:                                                 ; do {
+    movzx       nbits, byte actbl + c_derived_tbl.ehufsi + 0xf0
+                                                          ;   nbits = actbl->ehufsi0xf0;
+    sub         size, 16                                  ;   size -= 16;
+    movd        mm_code, dword actbl + c_derived_tbl.ehufco + 0xf0 * 4
+                                                          ;   code = actbl->ehufco0xf0;
+    sub         free_bits, nbits                          ;   if ((free_bits -= nbits) <= 0)
+    jle         .EMIT_BRLOOP2                             ;     insert code and flush put_buffer
+    movd        mm_nbits, nbits                           ;   else { nbits --> MMX register
+    psllq       mm_put_buffer, mm_nbits                   ;     put_buffer <<= nbits;
+    por         mm_put_buffer, mm_code                    ;     put_buffer |= code;
+    cmp         size, 16                                  ;     if (size <= 16)
+    jle        .ERLOOP2                                   ;       goto .ERLOOP2;
+    jmp        .BRLOOP2                                   ; } while (1);
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align      16
+.BLOOP2:                                                  ; do {  /* size = # of zero bits/elements to skip */
+    shr         index, sizeb                              ;   index >>= size;
+    lea         t, t + size * SIZEOF_WORD               ;   t += size;
+    cmp         size, 16                                  ;   if (size > 16)
+    jg          .BRLOOP2                                  ;     goto .BRLOOP2;
+.ERLOOP2:                                                 ; .ERLOOP2:
+    movsx       nbits, word t                           ;   nbits = *t;
+    add         size, size                                ;   size += size;
+    movd        mm_temp, nbits                            ;   temp = nbits;
+    movzx       nbits, byte NBITS(nbits)                ;   nbits = JPEG_NBITS(nbits);
+    movd        mm_nbits, nbits                           ;   nbits --> MMX register
+    lea         size, size * 8 + nbits                  ;   size = size * 8 + nbits;
+    movd        mm_code, dword actbl + c_derived_tbl.ehufco + (size - 16) * 4
+                                                          ;   code = actbl->ehufcosize-16;
+    movzx       size, byte actbl + c_derived_tbl.ehufsi + (size - 16)
+                                                          ;   size = actbl->ehufsisize-16;
+    psllq       mm_code, mm_nbits                         ;   code <<= nbits;
+    pand        mm_temp, MASK_BITS(nbits)               ;   temp &= (1 << nbits) - 1;
+    lea         nbits, nbits + size                     ;   nbits += size;
+    por         mm_code, mm_temp                          ;   code |= temp;
+    xor         size, size                                ;   size = 0;  /* kill tzcnt input dependency */
+    sub         free_bits, nbits                          ;   if ((free_bits -= nbits) <= 0)
+    jle         .EMIT_ERLOOP2                             ;     insert code, flush buffer, init size, goto .BLOOP2
+    tzcnt       size, index                               ;   size = # of trailing 0 bits in index
+    movd        mm_nbits, nbits                           ;   nbits --> MMX register
+    psllq       mm_put_buffer, mm_nbits                   ;   put_buffer <<= nbits;
+    inc         size                                      ;   ++size;
+    por         mm_put_buffer, mm_code                    ;   put_buffer |= code;
+    test        index, index
+    jnz         .BLOOP2                                   ; } while (index != 0);
+.ELOOP2:                                                  ; .ELOOP2:
+    mov         nbits, t                                  ; nbits = t;
+    lea         t, t + SIZEOF_WORD                      ; t = &t1;
+    and         nbits, DCTSIZE2 * SIZEOF_WORD - 1         ; nbits &= 127;
+    and         t, -DCTSIZE2 * SIZEOF_WORD                ; t &= -128;  /* t = &t_0; */
+    cmp         nbits, (DCTSIZE2 - 2) * SIZEOF_WORD       ; if (nbits != 62 * 2)
+    je          .EFN                                      ; {
+    movd        mm_code, dword actbl + c_derived_tbl.ehufco + 0
+                                                          ;   code = actbl->ehufco0;
+    movzx       nbits, byte actbl + c_derived_tbl.ehufsi + 0
+                                                          ;   nbits = actbl->ehufsi0;
+    sub         free_bits, nbits                          ;   if ((free_bits -= nbits) <= 0)
+    jg          .EFN_SKIP_EMIT_CODE                       ;   {
+    EMIT_QWORD  size, sizeb, sizeh, , , , , , .EFN        ;     insert code, flush put_buffer
+    align       16
+.EFN_SKIP_EMIT_CODE:                                      ;   } else {
+    movd        mm_nbits, nbits                           ;     nbits --> MMX register
+    psllq       mm_put_buffer, mm_nbits                   ;     put_buffer <<= nbits;
+    por         mm_put_buffer, mm_code                    ;     put_buffer |= code;
+.EFN:                                                     ; } }
+%define frame  esp
+    mov         frame, t + save_frame
+%define state  ecx
+    mov         state, frame + arg_state
+    movq        state + working_state.cur.put_buffer.simd, mm_put_buffer
+                                                          ; state->cur.put_buffer.simd = put_buffer;
+    emms
+    mov         state + working_state.cur.free_bits, free_bits
+                                                          ; state->cur.free_bits = free_bits;
+    POP         edi
+    POP         esi
+    POP         ebp
+    POP         ebx
+    ret
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align       16
+.EMIT_BRLOOP1:
+    EMIT_QWORD  emit_temp, emit_tempb, emit_temph, , , , , , \
+      .ERLOOP1
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align       16
+.EMIT_ERLOOP1:
+    EMIT_QWORD  size, sizeb, sizeh, \
+      { xor     size, size }, \
+      { tzcnt   size, index }, \
+      { inc     size }, \
+      { test    index, index }, \
+      { jnz     .BLOOP1 }, \
+      .ELOOP1
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align       16
+.EMIT_BRLOOP2:
+    EMIT_QWORD  emit_temp, emit_tempb, emit_temph, , , , \
+      { cmp     size, 16 }, \
+      { jle     .ERLOOP2 }, \
+      .BRLOOP2
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align       16
+.EMIT_ERLOOP2:
+    EMIT_QWORD  size, sizeb, sizeh, \
+      { xor     size, size }, \
+      { tzcnt   size, index }, \
+      { inc     size }, \
+      { test    index, index }, \
+      { jnz     .BLOOP2 }, \
+      .ELOOP2
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcphuff-sse2.asm Added

@@ -0,0 +1,662 @@
+;
+; jcphuff-sse2.asm - prepare data for progressive Huffman encoding (SSE2)
+;
+; Copyright (C) 2016, 2018, Matthieu Darbois
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains an SSE2 implementation of data preparation for progressive
+; Huffman encoding.  See jcphuff.c for more details.
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+; --------------------------------------------------------------------------
+; Macros to load data for jsimd_encode_mcu_AC_first_prepare_sse2() and
+; jsimd_encode_mcu_AC_refine_prepare_sse2()
+
+%macro LOAD16 0
+    pxor        N0, N0
+    pxor        N1, N1
+
+    mov         T0, INT LUT +  0*SIZEOF_INT
+    mov         T1, INT LUT +  8*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 0
+    pinsrw      X1, word BLOCK + T1 * 2, 0
+
+    mov         T0, INT LUT +  1*SIZEOF_INT
+    mov         T1, INT LUT +  9*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 1
+    pinsrw      X1, word BLOCK + T1 * 2, 1
+
+    mov         T0, INT LUT +  2*SIZEOF_INT
+    mov         T1, INT LUT + 10*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 2
+    pinsrw      X1, word BLOCK + T1 * 2, 2
+
+    mov         T0, INT LUT +  3*SIZEOF_INT
+    mov         T1, INT LUT + 11*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 3
+    pinsrw      X1, word BLOCK + T1 * 2, 3
+
+    mov         T0, INT LUT +  4*SIZEOF_INT
+    mov         T1, INT LUT + 12*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 4
+    pinsrw      X1, word BLOCK + T1 * 2, 4
+
+    mov         T0, INT LUT +  5*SIZEOF_INT
+    mov         T1, INT LUT + 13*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 5
+    pinsrw      X1, word BLOCK + T1 * 2, 5
+
+    mov         T0, INT LUT +  6*SIZEOF_INT
+    mov         T1, INT LUT + 14*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 6
+    pinsrw      X1, word BLOCK + T1 * 2, 6
+
+    mov         T0, INT LUT +  7*SIZEOF_INT
+    mov         T1, INT LUT + 15*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 7
+    pinsrw      X1, word BLOCK + T1 * 2, 7
+%endmacro
+
+%macro LOAD15 0
+    pxor        N0, N0
+    pxor        N1, N1
+    pxor        X1, X1
+
+    mov         T0, INT LUT +  0*SIZEOF_INT
+    mov         T1, INT LUT +  8*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 0
+    pinsrw      X1, word BLOCK + T1 * 2, 0
+
+    mov         T0, INT LUT +  1*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 1
+
+    mov         T0, INT LUT +  2*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 2
+
+    mov         T0, INT LUT +  3*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 3
+
+    mov         T0, INT LUT +  4*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 4
+
+    mov         T0, INT LUT +  5*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 5
+
+    mov         T0, INT LUT +  6*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 6
+
+    mov         T0, INT LUT +  7*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 7
+
+    cmp         LENEND, 2
+    jl          %%.ELOAD15
+    mov         T1, INT LUT +  9*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 1
+
+    cmp         LENEND, 3
+    jl          %%.ELOAD15
+    mov         T1, INT LUT + 10*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 2
+
+    cmp         LENEND, 4
+    jl          %%.ELOAD15
+    mov         T1, INT LUT + 11*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 3
+
+    cmp         LENEND, 5
+    jl          %%.ELOAD15
+    mov         T1, INT LUT + 12*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 4
+
+    cmp         LENEND, 6
+    jl          %%.ELOAD15
+    mov         T1, INT LUT + 13*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 5
+
+    cmp         LENEND, 7
+    jl          %%.ELOAD15
+    mov         T1, INT LUT + 14*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 6
+%%.ELOAD15:
+%endmacro
+
+%macro LOAD8 0
+    pxor        N0, N0
+
+    mov         T0, INT LUT +  0*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 0
+
+    mov         T0, INT LUT +  1*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 1
+
+    mov         T0, INT LUT +  2*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 2
+
+    mov         T0, INT LUT +  3*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 3
+
+    mov         T0, INT LUT +  4*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 4
+
+    mov         T0, INT LUT +  5*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 5
+
+    mov         T0, INT LUT +  6*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 6
+
+    mov         T0, INT LUT +  7*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 7
+%endmacro
+
+%macro LOAD7 0
+    pxor        N0, N0
+    pxor        X0, X0
+
+    mov         T1, INT LUT +  0*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 0
+
+    cmp         LENEND, 2
+    jl          %%.ELOAD7
+    mov         T1, INT LUT +  1*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 1
+
+    cmp         LENEND, 3
+    jl          %%.ELOAD7
+    mov         T1, INT LUT +  2*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 2
+
+    cmp         LENEND, 4
+    jl          %%.ELOAD7
+    mov         T1, INT LUT +  3*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 3
+
+    cmp         LENEND, 5
+    jl          %%.ELOAD7
+    mov         T1, INT LUT +  4*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 4
+
+    cmp         LENEND, 6
+    jl          %%.ELOAD7
+    mov         T1, INT LUT +  5*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 5
+
+    cmp         LENEND, 7
+    jl          %%.ELOAD7
+    mov         T1, INT LUT +  6*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 6
+%%.ELOAD7:
+%endmacro
+
+%macro REDUCE0 0
+    movdqa      xmm0, XMMWORD VALUES + ( 0*2)
+    movdqa      xmm1, XMMWORD VALUES + ( 8*2)
+    movdqa      xmm2, XMMWORD VALUES + (16*2)
+    movdqa      xmm3, XMMWORD VALUES + (24*2)
+    movdqa      xmm4, XMMWORD VALUES + (32*2)
+    movdqa      xmm5, XMMWORD VALUES + (40*2)
+    movdqa      xmm6, XMMWORD VALUES + (48*2)
+
+    pcmpeqw     xmm0, ZERO
+    pcmpeqw     xmm1, ZERO
+    pcmpeqw     xmm2, ZERO
+    pcmpeqw     xmm3, ZERO
+    pcmpeqw     xmm4, ZERO
+    pcmpeqw     xmm5, ZERO
+    pcmpeqw     xmm6, ZERO
+    pcmpeqw     xmm7, XMMWORD VALUES + (56*2)
+
+    packsswb    xmm0, xmm1
+    packsswb    xmm2, xmm3
+    packsswb    xmm4, xmm5
+    packsswb    xmm6, xmm7
+
+    pmovmskb    eax, xmm0
+    pmovmskb    ecx, xmm2
+    pmovmskb    edx, xmm4
+    pmovmskb    esi, xmm6
+
+    shl         ecx, 16
+    shl         esi, 16
+
+    or          eax, ecx
+    or          edx, esi
+
+    not         eax
+    not         edx
+
+    mov         edi, ZEROBITS
+
+    mov         INT edi, eax
+    mov         INT edi+SIZEOF_INT, edx
+%endmacro
+
+;
+; Prepare data for jsimd_encode_mcu_AC_first().
+;
+; GLOBAL(void)
+; jsimd_encode_mcu_AC_first_prepare_sse2(const JCOEF *block,
+;                                        const int *jpeg_natural_order_start,
+;                                        int Sl, int Al, JCOEF *values,
+;                                        size_t *zerobits)
+;
+; eax + 8 = const JCOEF *block
+; eax + 12 = const int *jpeg_natural_order_start
+; eax + 16 = int Sl
+; eax + 20 = int Al
+; eax + 24 = JCOEF *values
+; eax + 28 = size_t *zerobits
+
+%define ZERO    xmm7
+%define X0      xmm0
+%define X1      xmm1
+%define N0      xmm2
+%define N1      xmm3
+%define AL      xmm4
+%define K       eax
+%define LENEND  eax
+%define LUT     ebx
+%define T0      ecx
+%define T1      edx
+%define BLOCK   esi
+%define VALUES  edi
+%define LEN     ebp
+
+%define ZEROBITS  INT esp + 5 * 4
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_encode_mcu_AC_first_prepare_sse2)
+
+EXTN(jsimd_encode_mcu_AC_first_prepare_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    sub         esp, 4
+    push        ebx
+    push        ecx
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+    push        ebp
+
+    mov         BLOCK, INT eax + 8
+    mov         LUT, INT eax + 12
+    mov         VALUES, INT eax + 24
+    movd        AL, INT eax + 20
+    mov         T0, INT eax + 28
+    mov         ZEROBITS, T0
+    mov         LEN, INT eax + 16
+    pxor        ZERO, ZERO
+    mov         K, LEN
+    and         K, -16
+    shr         K, 4
+    jz          .ELOOP16
+.BLOOP16:
+    LOAD16
+    pcmpgtw     N0, X0
+    pcmpgtw     N1, X1
+    paddw       X0, N0
+    paddw       X1, N1
+    pxor        X0, N0
+    pxor        X1, N1
+    psrlw       X0, AL
+    psrlw       X1, AL
+    pxor        N0, X0
+    pxor        N1, X1
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (8) * 2, X1
+    movdqa      XMMWORD VALUES + (0 + DCTSIZE2) * 2, N0
+    movdqa      XMMWORD VALUES + (8 + DCTSIZE2) * 2, N1
+    add         VALUES, 16*2
+    add         LUT, 16*SIZEOF_INT
+    dec         K
+    jnz         .BLOOP16
+    test        LEN, 15
+    je          .PADDING
+.ELOOP16:
+    mov         LENEND, LEN
+    and         LENEND, 7
+
+    test        LEN, 8
+    jz          .TRY7
+    test        LEN, 7
+    jz          .TRY8
+
+    LOAD15
+    pcmpgtw     N0, X0
+    pcmpgtw     N1, X1
+    paddw       X0, N0
+    paddw       X1, N1
+    pxor        X0, N0
+    pxor        X1, N1
+    psrlw       X0, AL
+    psrlw       X1, AL
+    pxor        N0, X0
+    pxor        N1, X1
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (8) * 2, X1
+    movdqa      XMMWORD VALUES + (0 + DCTSIZE2) * 2, N0
+    movdqa      XMMWORD VALUES + (8 + DCTSIZE2) * 2, N1
+    add         VALUES, 16*2
+    jmp         .PADDING
+.TRY8:
+    LOAD8
+    pcmpgtw     N0, X0
+    paddw       X0, N0
+    pxor        X0, N0
+    psrlw       X0, AL
+    pxor        N0, X0
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (0 + DCTSIZE2) * 2, N0
+    add         VALUES, 8*2
+    jmp         .PADDING
+.TRY7:
+    LOAD7
+    pcmpgtw     N0, X0
+    paddw       X0, N0
+    pxor        X0, N0
+    psrlw       X0, AL
+    pxor        N0, X0
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (0 + DCTSIZE2) * 2, N0
+    add         VALUES, 8*2
+.PADDING:
+    mov         K, LEN
+    add         K, 7
+    and         K, -8
+    shr         K, 3
+    sub         K, DCTSIZE2/8
+    jz          .EPADDING
+    align       16
+.ZEROLOOP:
+    movdqa      XMMWORD VALUES + 0, ZERO
+    add         VALUES, 8*2
+    inc         K
+    jnz         .ZEROLOOP
+.EPADDING:
+    sub         VALUES, DCTSIZE2*2
+
+    REDUCE0
+
+    pop         ebp
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+    pop         ecx
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+%undef ZERO
+%undef X0
+%undef X1
+%undef N0
+%undef N1
+%undef AL
+%undef K
+%undef LUT
+%undef T0
+%undef T1
+%undef BLOCK
+%undef VALUES
+%undef LEN
+
+;
+; Prepare data for jsimd_encode_mcu_AC_refine().
+;
+; GLOBAL(int)
+; jsimd_encode_mcu_AC_refine_prepare_sse2(const JCOEF *block,
+;                                         const int *jpeg_natural_order_start,
+;                                         int Sl, int Al, JCOEF *absvalues,
+;                                         size_t *bits)
+;
+; eax + 8 = const JCOEF *block
+; eax + 12 = const int *jpeg_natural_order_start
+; eax + 16 = int Sl
+; eax + 20 = int Al
+; eax + 24 = JCOEF *values
+; eax + 28 = size_t *bits
+
+%define ZERO    xmm7
+%define ONE     xmm5
+%define X0      xmm0
+%define X1      xmm1
+%define N0      xmm2
+%define N1      xmm3
+%define AL      xmm4
+%define K       eax
+%define LENEND  eax
+%define LUT     ebx
+%define T0      ecx
+%define T0w      cx
+%define T1      edx
+%define BLOCK   esi
+%define VALUES  edi
+%define KK      ebp
+
+%define ZEROBITS  INT esp + 5 * 4
+%define EOB       INT esp + 5 * 4 + 4
+%define LEN       INT esp + 5 * 4 + 8
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_encode_mcu_AC_refine_prepare_sse2)
+
+EXTN(jsimd_encode_mcu_AC_refine_prepare_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    sub         esp, 16
+    push        ebx
+    push        ecx
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+    push        ebp
+
+    pcmpeqw     ONE, ONE
+    psrlw       ONE, 15
+    mov         BLOCK, INT eax + 8
+    mov         LUT, INT eax + 12
+    mov         VALUES, INT eax + 24
+    movd        AL, INT eax + 20
+    mov         T0, INT eax + 28
+    mov         K,  INT eax + 16
+    mov         INT T0 + 2 * SIZEOF_INT, -1
+    mov         INT T0 + 3 * SIZEOF_INT, -1
+    mov         ZEROBITS, T0
+    mov         LEN, K
+    pxor        ZERO, ZERO
+    and         K, -16
+    mov         EOB, 0
+    xor         KK, KK
+    shr         K, 4
+    jz          .ELOOPR16
+.BLOOPR16:
+    LOAD16
+    pcmpgtw     N0, X0
+    pcmpgtw     N1, X1
+    paddw       X0, N0
+    paddw       X1, N1
+    pxor        X0, N0
+    pxor        X1, N1
+    psrlw       X0, AL
+    psrlw       X1, AL
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (8) * 2, X1
+    pcmpeqw     X0, ONE
+    pcmpeqw     X1, ONE
+    packsswb    N0, N1
+    packsswb    X0, X1
+    pmovmskb    T0, N0                  ; lsignbits.val16uk>>4 = _mm_movemask_epi8(neg);
+    mov         T1, ZEROBITS
+    not         T0
+    mov         word T1 + 2 * SIZEOF_INT + KK, T0w
+    pmovmskb    T1, X0                  ; idx = _mm_movemask_epi8(x1);
+    bsr         T1, T1                  ;  idx = 16 - (__builtin_clz(idx)>>1);
+    jz          .CONTINUER16            ; if (idx) {
+    lea         T1, T1+KK*8
+    mov         EOB, T1                 ; EOB = k + idx;
+.CONTINUER16:
+    add         VALUES, 16*2
+    add         LUT, 16*SIZEOF_INT
+    add         KK, 2
+    dec         K
+    jnz         .BLOOPR16
+    test        LEN, 15
+    je          .PADDINGR
+.ELOOPR16:
+    mov         LENEND, LEN
+
+    test        LENEND, 8
+    jz          .TRYR7
+    test        LENEND, 7
+    jz          .TRYR8
+
+    and         LENEND, 7
+    LOAD15
+    pcmpgtw     N0, X0
+    pcmpgtw     N1, X1
+    paddw       X0, N0
+    paddw       X1, N1
+    pxor        X0, N0
+    pxor        X1, N1
+    psrlw       X0, AL
+    psrlw       X1, AL
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (8) * 2, X1
+    pcmpeqw     X0, ONE
+    pcmpeqw     X1, ONE
+    packsswb    N0, N1
+    packsswb    X0, X1
+    pmovmskb    T0, N0                  ; lsignbits.val16uk>>4 = _mm_movemask_epi8(neg);
+    mov         T1, ZEROBITS
+    not         T0
+    mov         word T1 + 2 * SIZEOF_INT + KK, T0w
+    pmovmskb    T1, X0                  ; idx = _mm_movemask_epi8(x1);
+    bsr         T1, T1                  ;  idx = 16 - (__builtin_clz(idx)>>1);
+    jz          .CONTINUER15            ; if (idx) {
+    lea         T1, T1+KK*8
+    mov         EOB, T1                 ; EOB = k + idx;
+.CONTINUER15:
+    add         VALUES, 16*2
+    jmp         .PADDINGR
+.TRYR8:
+    LOAD8
+
+    pcmpgtw     N0, X0
+    paddw       X0, N0
+    pxor        X0, N0
+    psrlw       X0, AL
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    pcmpeqw     X0, ONE
+    packsswb    N0, ZERO
+    packsswb    X0, ZERO
+    pmovmskb    T0, N0                  ; lsignbits.val16uk>>4 = _mm_movemask_epi8(neg);
+    mov         T1, ZEROBITS
+    not         T0
+    mov         word T1 + 2 * SIZEOF_INT + KK, T0w
+    pmovmskb    T1, X0                  ; idx = _mm_movemask_epi8(x1);
+    bsr         T1, T1                  ;  idx = 16 - (__builtin_clz(idx)>>1);
+    jz          .CONTINUER8             ; if (idx) {
+    lea         T1, T1+KK*8
+    mov         EOB, T1                 ; EOB = k + idx;
+.CONTINUER8:
+    add         VALUES, 8*2
+    jmp         .PADDINGR
+.TRYR7:
+    and         LENEND, 7
+    LOAD7
+
+    pcmpgtw     N0, X0
+    paddw       X0, N0
+    pxor        X0, N0
+    psrlw       X0, AL
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    pcmpeqw     X0, ONE
+    packsswb    N0, ZERO
+    packsswb    X0, ZERO
+    pmovmskb    T0, N0                  ; lsignbits.val16uk>>4 = _mm_movemask_epi8(neg);
+    mov         T1, ZEROBITS
+    not         T0
+    mov         word T1 + 2 * SIZEOF_INT + KK, T0w
+    pmovmskb    T1, X0                  ; idx = _mm_movemask_epi8(x1);
+    bsr         T1, T1                  ;  idx = 16 - (__builtin_clz(idx)>>1);
+    jz          .CONTINUER7             ; if (idx) {
+    lea         T1, T1+KK*8
+    mov         EOB, T1                 ; EOB = k + idx;
+.CONTINUER7:
+    add         VALUES, 8*2
+.PADDINGR:
+    mov         K, LEN
+    add         K, 7
+    and         K, -8
+    shr         K, 3
+    sub         K, DCTSIZE2/8
+    jz          .EPADDINGR
+    align       16
+.ZEROLOOPR:
+    movdqa      XMMWORD VALUES + 0, ZERO
+    add         VALUES, 8*2
+    inc         K
+    jnz         .ZEROLOOPR
+.EPADDINGR:
+    sub         VALUES, DCTSIZE2*2
+
+    REDUCE0
+
+    mov         eax, EOB
+
+    pop         ebp
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+    pop         ecx
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+%undef ZERO
+%undef ONE
+%undef X0
+%undef X1
+%undef N0
+%undef N1
+%undef AL
+%undef K
+%undef KK
+%undef EOB
+%undef SIGN
+%undef LUT
+%undef T0
+%undef T1
+%undef BLOCK
+%undef VALUES
+%undef LEN
+%undef LENEND
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcsample-avx2.asm Added

@@ -0,0 +1,388 @@
+;
+; jcsample.asm - downsampling (AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_avx2(JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor,
+;                            JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                            JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b) + 8         ; JDIMENSION image_width
+%define max_v_samp(b)   (b) + 12        ; int max_v_samp_factor
+%define v_samp(b)       (b) + 16        ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b) + 20        ; JDIMENSION width_in_blocks
+%define input_data(b)   (b) + 24        ; JSAMPARRAY input_data
+%define output_data(b)  (b) + 28        ; JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_downsample_avx2)
+
+EXTN(jsimd_h2v1_downsample_avx2):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         ecx, JDIMENSION width_blks(ebp)
+    shl         ecx, 3                  ; imul ecx,DCTSIZE (ecx = output_cols)
+    jz          near .return
+
+    mov         edx, JDIMENSION img_width(ebp)
+
+    ; -- expand_right_edge
+
+    push        ecx
+    shl         ecx, 1                  ; output_cols * 2
+    sub         ecx, edx
+    jle         short .expand_end
+
+    mov         eax, INT max_v_samp(ebp)
+    test        eax, eax
+    jle         short .expand_end
+
+    cld
+    mov         esi, JSAMPARRAY input_data(ebp)  ; input_data
+    alignx      16, 7
+.expandloop:
+    push        eax
+    push        ecx
+
+    mov         edi, JSAMPROW esi
+    add         edi, edx
+    mov         al, JSAMPLE edi-1
+
+    rep stosb
+
+    pop         ecx
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW
+    dec         eax
+    jg          short .expandloop
+
+.expand_end:
+    pop         ecx                     ; output_cols
+
+    ; -- h2v1_downsample
+
+    mov         eax, JDIMENSION v_samp(ebp)  ; rowctr
+    test        eax, eax
+    jle         near .return
+
+    mov         edx, 0x00010000         ; bias pattern
+    vmovd       xmm7, edx
+    vpshufd     xmm7, xmm7, 0x00        ; xmm7={0, 1, 0, 1, 0, 1, 0, 1}
+    vperm2i128  ymm7, ymm7, ymm7, 0     ; ymm7={xmm7, xmm7}
+    vpcmpeqw    ymm6, ymm6, ymm6
+    vpsrlw      ymm6, ymm6, BYTE_BIT    ; ymm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         esi, JSAMPARRAY input_data(ebp)   ; input_data
+    mov         edi, JSAMPARRAY output_data(ebp)  ; output_data
+    alignx      16, 7
+.rowloop:
+    push        ecx
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr
+
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jae         short .columnloop
+    alignx      16, 7
+
+.columnloop_r24:
+    ; ecx can possibly be 8, 16, 24
+    cmp         ecx, 24
+    jne         .columnloop_r16
+    vmovdqu     ymm0, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     xmm1, XMMWORD esi+1*SIZEOF_YMMWORD
+    mov         ecx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop_r16:
+    cmp         ecx, 16
+    jne         .columnloop_r8
+    vmovdqu     ymm0, YMMWORD esi+0*SIZEOF_YMMWORD
+    vpxor       ymm1, ymm1, ymm1
+    mov         ecx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop_r8:
+    vmovdqu     xmm0, XMMWORDesi+0*SIZEOF_YMMWORD
+    vpxor       ymm1, ymm1, ymm1
+    mov         ecx, SIZEOF_YMMWORD
+    jmp         short .downsample
+    alignx      16, 7
+
+.columnloop:
+    vmovdqu     ymm0, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymm1, YMMWORD esi+1*SIZEOF_YMMWORD
+
+.downsample:
+    vpsrlw      ymm2, ymm0, BYTE_BIT
+    vpand       ymm0, ymm0, ymm6
+    vpsrlw      ymm3, ymm1, BYTE_BIT
+    vpand       ymm1, ymm1, ymm6
+
+    vpaddw      ymm0, ymm0, ymm2
+    vpaddw      ymm1, ymm1, ymm3
+    vpaddw      ymm0, ymm0, ymm7
+    vpaddw      ymm1, ymm1, ymm7
+    vpsrlw      ymm0, ymm0, 1
+    vpsrlw      ymm1, ymm1, 1
+
+    vpackuswb   ymm0, ymm0, ymm1
+    vpermq      ymm0, ymm0, 0xd8
+
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymm0
+
+    sub         ecx, byte SIZEOF_YMMWORD    ; outcol
+    add         esi, byte 2*SIZEOF_YMMWORD  ; inptr
+    add         edi, byte 1*SIZEOF_YMMWORD  ; outptr
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jae         short .columnloop
+    test        ecx, ecx
+    jnz         near .columnloop_r24
+
+    pop         esi
+    pop         edi
+    pop         ecx
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         eax                        ; rowctr
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_avx2(JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor,
+;                            JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                            JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b) + 8         ; JDIMENSION image_width
+%define max_v_samp(b)   (b) + 12        ; int max_v_samp_factor
+%define v_samp(b)       (b) + 16        ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b) + 20        ; JDIMENSION width_in_blocks
+%define input_data(b)   (b) + 24        ; JSAMPARRAY input_data
+%define output_data(b)  (b) + 28        ; JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_downsample_avx2)
+
+EXTN(jsimd_h2v2_downsample_avx2):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         ecx, JDIMENSION width_blks(ebp)
+    shl         ecx, 3                  ; imul ecx,DCTSIZE (ecx = output_cols)
+    jz          near .return
+
+    mov         edx, JDIMENSION img_width(ebp)
+
+    ; -- expand_right_edge
+
+    push        ecx
+    shl         ecx, 1                  ; output_cols * 2
+    sub         ecx, edx
+    jle         short .expand_end
+
+    mov         eax, INT max_v_samp(ebp)
+    test        eax, eax
+    jle         short .expand_end
+
+    cld
+    mov         esi, JSAMPARRAY input_data(ebp)  ; input_data
+    alignx      16, 7
+.expandloop:
+    push        eax
+    push        ecx
+
+    mov         edi, JSAMPROW esi
+    add         edi, edx
+    mov         al, JSAMPLE edi-1
+
+    rep stosb
+
+    pop         ecx
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW
+    dec         eax
+    jg          short .expandloop
+
+.expand_end:
+    pop         ecx                     ; output_cols
+
+    ; -- h2v2_downsample
+
+    mov         eax, JDIMENSION v_samp(ebp)  ; rowctr
+    test        eax, eax
+    jle         near .return
+
+    mov         edx, 0x00020001         ; bias pattern
+    vmovd       xmm7, edx
+    vpcmpeqw    ymm6, ymm6, ymm6
+    vpshufd     xmm7, xmm7, 0x00        ; ymm7={1, 2, 1, 2, 1, 2, 1, 2}
+    vperm2i128  ymm7, ymm7, ymm7, 0
+    vpsrlw      ymm6, ymm6, BYTE_BIT    ; ymm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         esi, JSAMPARRAY input_data(ebp)   ; input_data
+    mov         edi, JSAMPARRAY output_data(ebp)  ; output_data
+    alignx      16, 7
+.rowloop:
+    push        ecx
+    push        edi
+    push        esi
+
+    mov         edx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         esi, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; inptr1
+    mov         edi, JSAMPROW edi                    ; outptr
+
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jae         short .columnloop
+    alignx      16, 7
+
+.columnloop_r24:
+    cmp         ecx, 24
+    jne         .columnloop_r16
+    vmovdqu     ymm0, YMMWORD edx+0*SIZEOF_YMMWORD
+    vmovdqu     ymm1, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     xmm2, XMMWORD edx+1*SIZEOF_YMMWORD
+    vmovdqu     xmm3, XMMWORD esi+1*SIZEOF_YMMWORD
+    mov         ecx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop_r16:
+    cmp         ecx, 16
+    jne         .columnloop_r8
+    vmovdqu     ymm0, YMMWORD edx+0*SIZEOF_YMMWORD
+    vmovdqu     ymm1, YMMWORD esi+0*SIZEOF_YMMWORD
+    vpxor       ymm2, ymm2, ymm2
+    vpxor       ymm3, ymm3, ymm3
+    mov         ecx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop_r8:
+    vmovdqu     xmm0, XMMWORD edx+0*SIZEOF_XMMWORD
+    vmovdqu     xmm1, XMMWORD esi+0*SIZEOF_XMMWORD
+    vpxor       ymm2, ymm2, ymm2
+    vpxor       ymm3, ymm3, ymm3
+    mov         ecx, SIZEOF_YMMWORD
+    jmp         short .downsample
+    alignx      16, 7
+
+.columnloop:
+    vmovdqu     ymm0, YMMWORD edx+0*SIZEOF_YMMWORD
+    vmovdqu     ymm1, YMMWORD esi+0*SIZEOF_YMMWORD
+    vmovdqu     ymm2, YMMWORD edx+1*SIZEOF_YMMWORD
+    vmovdqu     ymm3, YMMWORD esi+1*SIZEOF_YMMWORD
+
+.downsample:
+    vpand       ymm4, ymm0, ymm6
+    vpsrlw      ymm0, ymm0, BYTE_BIT
+    vpand       ymm5, ymm1, ymm6
+    vpsrlw      ymm1, ymm1, BYTE_BIT
+    vpaddw      ymm0, ymm0, ymm4
+    vpaddw      ymm1, ymm1, ymm5
+
+    vpand       ymm4, ymm2, ymm6
+    vpsrlw      ymm2, ymm2, BYTE_BIT
+    vpand       ymm5, ymm3, ymm6
+    vpsrlw      ymm3, ymm3, BYTE_BIT
+    vpaddw      ymm2, ymm2, ymm4
+    vpaddw      ymm3, ymm3, ymm5
+
+    vpaddw      ymm0, ymm0, ymm1
+    vpaddw      ymm2, ymm2, ymm3
+    vpaddw      ymm0, ymm0, ymm7
+    vpaddw      ymm2, ymm2, ymm7
+    vpsrlw      ymm0, ymm0, 2
+    vpsrlw      ymm2, ymm2, 2
+
+    vpackuswb   ymm0, ymm0, ymm2
+    vpermq      ymm0, ymm0, 0xd8
+
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymm0
+
+    sub         ecx, byte SIZEOF_YMMWORD    ; outcol
+    add         edx, byte 2*SIZEOF_YMMWORD  ; inptr0
+    add         esi, byte 2*SIZEOF_YMMWORD  ; inptr1
+    add         edi, byte 1*SIZEOF_YMMWORD  ; outptr
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+    test        ecx, ecx
+    jnz         near .columnloop_r24
+
+    pop         esi
+    pop         edi
+    pop         ecx
+
+    add         esi, byte 2*SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte 1*SIZEOF_JSAMPROW  ; output_data
+    dec         eax                          ; rowctr
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcsample-mmx.asm Added

@@ -0,0 +1,324 @@
+;
+; jcsample.asm - downsampling (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_mmx(JDIMENSION image_width, int max_v_samp_factor,
+;                           JDIMENSION v_samp_factor,
+;                           JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                           JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b) + 8         ; JDIMENSION image_width
+%define max_v_samp(b)   (b) + 12        ; int max_v_samp_factor
+%define v_samp(b)       (b) + 16        ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b) + 20        ; JDIMENSION width_in_blocks
+%define input_data(b)   (b) + 24        ; JSAMPARRAY input_data
+%define output_data(b)  (b) + 28        ; JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_downsample_mmx)
+
+EXTN(jsimd_h2v1_downsample_mmx):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         ecx, JDIMENSION width_blks(ebp)
+    shl         ecx, 3                  ; imul ecx,DCTSIZE (ecx = output_cols)
+    jz          near .return
+
+    mov         edx, JDIMENSION img_width(ebp)
+
+    ; -- expand_right_edge
+
+    push        ecx
+    shl         ecx, 1                  ; output_cols * 2
+    sub         ecx, edx
+    jle         short .expand_end
+
+    mov         eax, INT max_v_samp(ebp)
+    test        eax, eax
+    jle         short .expand_end
+
+    cld
+    mov         esi, JSAMPARRAY input_data(ebp)  ; input_data
+    alignx      16, 7
+.expandloop:
+    push        eax
+    push        ecx
+
+    mov         edi, JSAMPROW esi
+    add         edi, edx
+    mov         al, JSAMPLE edi-1
+
+    rep stosb
+
+    pop         ecx
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW
+    dec         eax
+    jg          short .expandloop
+
+.expand_end:
+    pop         ecx                     ; output_cols
+
+    ; -- h2v1_downsample
+
+    mov         eax, JDIMENSION v_samp(ebp)  ; rowctr
+    test        eax, eax
+    jle         near .return
+
+    mov         edx, 0x00010000         ; bias pattern
+    movd        mm7, edx
+    pcmpeqw     mm6, mm6
+    punpckldq   mm7, mm7                ; mm7={0, 1, 0, 1}
+    psrlw       mm6, BYTE_BIT           ; mm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         esi, JSAMPARRAY input_data(ebp)   ; input_data
+    mov         edi, JSAMPARRAY output_data(ebp)  ; output_data
+    alignx      16, 7
+.rowloop:
+    push        ecx
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr
+    alignx      16, 7
+.columnloop:
+
+    movq        mm0, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mm1, MMWORD esi+1*SIZEOF_MMWORD
+    movq        mm2, mm0
+    movq        mm3, mm1
+
+    pand        mm0, mm6
+    psrlw       mm2, BYTE_BIT
+    pand        mm1, mm6
+    psrlw       mm3, BYTE_BIT
+
+    paddw       mm0, mm2
+    paddw       mm1, mm3
+    paddw       mm0, mm7
+    paddw       mm1, mm7
+    psrlw       mm0, 1
+    psrlw       mm1, 1
+
+    packuswb    mm0, mm1
+
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mm0
+
+    add         esi, byte 2*SIZEOF_MMWORD  ; inptr
+    add         edi, byte 1*SIZEOF_MMWORD  ; outptr
+    sub         ecx, byte SIZEOF_MMWORD    ; outcol
+    jnz         short .columnloop
+
+    pop         esi
+    pop         edi
+    pop         ecx
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         eax                        ; rowctr
+    jg          short .rowloop
+
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_mmx(JDIMENSION image_width, int max_v_samp_factor,
+;                           JDIMENSION v_samp_factor,
+;                           JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                           JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b) + 8         ; JDIMENSION image_width
+%define max_v_samp(b)   (b) + 12        ; int max_v_samp_factor
+%define v_samp(b)       (b) + 16        ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b) + 20        ; JDIMENSION width_in_blocks
+%define input_data(b)   (b) + 24        ; JSAMPARRAY input_data
+%define output_data(b)  (b) + 28        ; JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_downsample_mmx)
+
+EXTN(jsimd_h2v2_downsample_mmx):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         ecx, JDIMENSION width_blks(ebp)
+    shl         ecx, 3                  ; imul ecx,DCTSIZE (ecx = output_cols)
+    jz          near .return
+
+    mov         edx, JDIMENSION img_width(ebp)
+
+    ; -- expand_right_edge
+
+    push        ecx
+    shl         ecx, 1                  ; output_cols * 2
+    sub         ecx, edx
+    jle         short .expand_end
+
+    mov         eax, INT max_v_samp(ebp)
+    test        eax, eax
+    jle         short .expand_end
+
+    cld
+    mov         esi, JSAMPARRAY input_data(ebp)  ; input_data
+    alignx      16, 7
+.expandloop:
+    push        eax
+    push        ecx
+
+    mov         edi, JSAMPROW esi
+    add         edi, edx
+    mov         al, JSAMPLE edi-1
+
+    rep stosb
+
+    pop         ecx
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW
+    dec         eax
+    jg          short .expandloop
+
+.expand_end:
+    pop         ecx                     ; output_cols
+
+    ; -- h2v2_downsample
+
+    mov         eax, JDIMENSION v_samp(ebp)  ; rowctr
+    test        eax, eax
+    jle         near .return
+
+    mov         edx, 0x00020001         ; bias pattern
+    movd        mm7, edx
+    pcmpeqw     mm6, mm6
+    punpckldq   mm7, mm7                ; mm7={1, 2, 1, 2}
+    psrlw       mm6, BYTE_BIT           ; mm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         esi, JSAMPARRAY input_data(ebp)   ; input_data
+    mov         edi, JSAMPARRAY output_data(ebp)  ; output_data
+    alignx      16, 7
+.rowloop:
+    push        ecx
+    push        edi
+    push        esi
+
+    mov         edx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         esi, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; inptr1
+    mov         edi, JSAMPROW edi                    ; outptr
+    alignx      16, 7
+.columnloop:
+
+    movq        mm0, MMWORD edx+0*SIZEOF_MMWORD
+    movq        mm1, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mm2, MMWORD edx+1*SIZEOF_MMWORD
+    movq        mm3, MMWORD esi+1*SIZEOF_MMWORD
+
+    movq        mm4, mm0
+    movq        mm5, mm1
+    pand        mm0, mm6
+    psrlw       mm4, BYTE_BIT
+    pand        mm1, mm6
+    psrlw       mm5, BYTE_BIT
+    paddw       mm0, mm4
+    paddw       mm1, mm5
+
+    movq        mm4, mm2
+    movq        mm5, mm3
+    pand        mm2, mm6
+    psrlw       mm4, BYTE_BIT
+    pand        mm3, mm6
+    psrlw       mm5, BYTE_BIT
+    paddw       mm2, mm4
+    paddw       mm3, mm5
+
+    paddw       mm0, mm1
+    paddw       mm2, mm3
+    paddw       mm0, mm7
+    paddw       mm2, mm7
+    psrlw       mm0, 2
+    psrlw       mm2, 2
+
+    packuswb    mm0, mm2
+
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mm0
+
+    add         edx, byte 2*SIZEOF_MMWORD  ; inptr0
+    add         esi, byte 2*SIZEOF_MMWORD  ; inptr1
+    add         edi, byte 1*SIZEOF_MMWORD  ; outptr
+    sub         ecx, byte SIZEOF_MMWORD    ; outcol
+    jnz         near .columnloop
+
+    pop         esi
+    pop         edi
+    pop         ecx
+
+    add         esi, byte 2*SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte 1*SIZEOF_JSAMPROW  ; output_data
+    dec         eax                          ; rowctr
+    jg          near .rowloop
+
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jcsample-sse2.asm Added

@@ -0,0 +1,351 @@
+;
+; jcsample.asm - downsampling (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_sse2(JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor,
+;                            JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                            JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b) + 8         ; JDIMENSION image_width
+%define max_v_samp(b)   (b) + 12        ; int max_v_samp_factor
+%define v_samp(b)       (b) + 16        ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b) + 20        ; JDIMENSION width_in_blocks
+%define input_data(b)   (b) + 24        ; JSAMPARRAY input_data
+%define output_data(b)  (b) + 28        ; JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_downsample_sse2)
+
+EXTN(jsimd_h2v1_downsample_sse2):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         ecx, JDIMENSION width_blks(ebp)
+    shl         ecx, 3                  ; imul ecx,DCTSIZE (ecx = output_cols)
+    jz          near .return
+
+    mov         edx, JDIMENSION img_width(ebp)
+
+    ; -- expand_right_edge
+
+    push        ecx
+    shl         ecx, 1                  ; output_cols * 2
+    sub         ecx, edx
+    jle         short .expand_end
+
+    mov         eax, INT max_v_samp(ebp)
+    test        eax, eax
+    jle         short .expand_end
+
+    cld
+    mov         esi, JSAMPARRAY input_data(ebp)  ; input_data
+    alignx      16, 7
+.expandloop:
+    push        eax
+    push        ecx
+
+    mov         edi, JSAMPROW esi
+    add         edi, edx
+    mov         al, JSAMPLE edi-1
+
+    rep stosb
+
+    pop         ecx
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW
+    dec         eax
+    jg          short .expandloop
+
+.expand_end:
+    pop         ecx                     ; output_cols
+
+    ; -- h2v1_downsample
+
+    mov         eax, JDIMENSION v_samp(ebp)  ; rowctr
+    test        eax, eax
+    jle         near .return
+
+    mov         edx, 0x00010000         ; bias pattern
+    movd        xmm7, edx
+    pcmpeqw     xmm6, xmm6
+    pshufd      xmm7, xmm7, 0x00        ; xmm7={0, 1, 0, 1, 0, 1, 0, 1}
+    psrlw       xmm6, BYTE_BIT          ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         esi, JSAMPARRAY input_data(ebp)   ; input_data
+    mov         edi, JSAMPARRAY output_data(ebp)  ; output_data
+    alignx      16, 7
+.rowloop:
+    push        ecx
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr
+
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jae         short .columnloop
+    alignx      16, 7
+
+.columnloop_r8:
+    movdqa      xmm0, XMMWORD esi+0*SIZEOF_XMMWORD
+    pxor        xmm1, xmm1
+    mov         ecx, SIZEOF_XMMWORD
+    jmp         short .downsample
+    alignx      16, 7
+
+.columnloop:
+    movdqa      xmm0, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqa      xmm1, XMMWORD esi+1*SIZEOF_XMMWORD
+
+.downsample:
+    movdqa      xmm2, xmm0
+    movdqa      xmm3, xmm1
+
+    pand        xmm0, xmm6
+    psrlw       xmm2, BYTE_BIT
+    pand        xmm1, xmm6
+    psrlw       xmm3, BYTE_BIT
+
+    paddw       xmm0, xmm2
+    paddw       xmm1, xmm3
+    paddw       xmm0, xmm7
+    paddw       xmm1, xmm7
+    psrlw       xmm0, 1
+    psrlw       xmm1, 1
+
+    packuswb    xmm0, xmm1
+
+    movdqa      XMMWORD edi+0*SIZEOF_XMMWORD, xmm0
+
+    sub         ecx, byte SIZEOF_XMMWORD    ; outcol
+    add         esi, byte 2*SIZEOF_XMMWORD  ; inptr
+    add         edi, byte 1*SIZEOF_XMMWORD  ; outptr
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jae         short .columnloop
+    test        ecx, ecx
+    jnz         short .columnloop_r8
+
+    pop         esi
+    pop         edi
+    pop         ecx
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         eax                        ; rowctr
+    jg          near .rowloop
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_sse2(JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor,
+;                            JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                            JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b) + 8         ; JDIMENSION image_width
+%define max_v_samp(b)   (b) + 12        ; int max_v_samp_factor
+%define v_samp(b)       (b) + 16        ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b) + 20        ; JDIMENSION width_in_blocks
+%define input_data(b)   (b) + 24        ; JSAMPARRAY input_data
+%define output_data(b)  (b) + 28        ; JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_downsample_sse2)
+
+EXTN(jsimd_h2v2_downsample_sse2):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         ecx, JDIMENSION width_blks(ebp)
+    shl         ecx, 3                  ; imul ecx,DCTSIZE (ecx = output_cols)
+    jz          near .return
+
+    mov         edx, JDIMENSION img_width(ebp)
+
+    ; -- expand_right_edge
+
+    push        ecx
+    shl         ecx, 1                  ; output_cols * 2
+    sub         ecx, edx
+    jle         short .expand_end
+
+    mov         eax, INT max_v_samp(ebp)
+    test        eax, eax
+    jle         short .expand_end
+
+    cld
+    mov         esi, JSAMPARRAY input_data(ebp)  ; input_data
+    alignx      16, 7
+.expandloop:
+    push        eax
+    push        ecx
+
+    mov         edi, JSAMPROW esi
+    add         edi, edx
+    mov         al, JSAMPLE edi-1
+
+    rep stosb
+
+    pop         ecx
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW
+    dec         eax
+    jg          short .expandloop
+
+.expand_end:
+    pop         ecx                     ; output_cols
+
+    ; -- h2v2_downsample
+
+    mov         eax, JDIMENSION v_samp(ebp)  ; rowctr
+    test        eax, eax
+    jle         near .return
+
+    mov         edx, 0x00020001         ; bias pattern
+    movd        xmm7, edx
+    pcmpeqw     xmm6, xmm6
+    pshufd      xmm7, xmm7, 0x00        ; xmm7={1, 2, 1, 2, 1, 2, 1, 2}
+    psrlw       xmm6, BYTE_BIT          ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         esi, JSAMPARRAY input_data(ebp)   ; input_data
+    mov         edi, JSAMPARRAY output_data(ebp)  ; output_data
+    alignx      16, 7
+.rowloop:
+    push        ecx
+    push        edi
+    push        esi
+
+    mov         edx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         esi, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; inptr1
+    mov         edi, JSAMPROW edi                    ; outptr
+
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jae         short .columnloop
+    alignx      16, 7
+
+.columnloop_r8:
+    movdqa      xmm0, XMMWORD edx+0*SIZEOF_XMMWORD
+    movdqa      xmm1, XMMWORD esi+0*SIZEOF_XMMWORD
+    pxor        xmm2, xmm2
+    pxor        xmm3, xmm3
+    mov         ecx, SIZEOF_XMMWORD
+    jmp         short .downsample
+    alignx      16, 7
+
+.columnloop:
+    movdqa      xmm0, XMMWORD edx+0*SIZEOF_XMMWORD
+    movdqa      xmm1, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqa      xmm2, XMMWORD edx+1*SIZEOF_XMMWORD
+    movdqa      xmm3, XMMWORD esi+1*SIZEOF_XMMWORD
+
+.downsample:
+    movdqa      xmm4, xmm0
+    movdqa      xmm5, xmm1
+    pand        xmm0, xmm6
+    psrlw       xmm4, BYTE_BIT
+    pand        xmm1, xmm6
+    psrlw       xmm5, BYTE_BIT
+    paddw       xmm0, xmm4
+    paddw       xmm1, xmm5
+
+    movdqa      xmm4, xmm2
+    movdqa      xmm5, xmm3
+    pand        xmm2, xmm6
+    psrlw       xmm4, BYTE_BIT
+    pand        xmm3, xmm6
+    psrlw       xmm5, BYTE_BIT
+    paddw       xmm2, xmm4
+    paddw       xmm3, xmm5
+
+    paddw       xmm0, xmm1
+    paddw       xmm2, xmm3
+    paddw       xmm0, xmm7
+    paddw       xmm2, xmm7
+    psrlw       xmm0, 2
+    psrlw       xmm2, 2
+
+    packuswb    xmm0, xmm2
+
+    movdqa      XMMWORD edi+0*SIZEOF_XMMWORD, xmm0
+
+    sub         ecx, byte SIZEOF_XMMWORD    ; outcol
+    add         edx, byte 2*SIZEOF_XMMWORD  ; inptr0
+    add         esi, byte 2*SIZEOF_XMMWORD  ; inptr1
+    add         edi, byte 1*SIZEOF_XMMWORD  ; outptr
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+    test        ecx, ecx
+    jnz         near .columnloop_r8
+
+    pop         esi
+    pop         edi
+    pop         ecx
+
+    add         esi, byte 2*SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte 1*SIZEOF_JSAMPROW  ; output_data
+    dec         eax                          ; rowctr
+    jg          near .rowloop
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdcolext-avx2.asm Added

@@ -0,0 +1,515 @@
+;
+; jdcolext.asm - colorspace conversion (AVX2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2012, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_avx2(JDIMENSION out_width, JSAMPIMAGE input_buf,
+;                            JDIMENSION input_row, JSAMPARRAY output_buf,
+;                            int num_rows)
+;
+
+%define out_width(b)   (b) + 8          ; JDIMENSION out_width
+%define input_buf(b)   (b) + 12         ; JSAMPIMAGE input_buf
+%define input_row(b)   (b) + 16         ; JDIMENSION input_row
+%define output_buf(b)  (b) + 20         ; JSAMPARRAY output_buf
+%define num_rows(b)    (b) + 24         ; int num_rows
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_YMMWORD
+                                        ; ymmword wkWK_NUM
+%define WK_NUM         2
+%define gotptr         wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_ycc_rgb_convert_avx2)
+
+EXTN(jsimd_ycc_rgb_convert_avx2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION out_width(eax)  ; num_cols
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         edi, JSAMPIMAGE input_buf(eax)
+    mov         ecx, JDIMENSION input_row(eax)
+    mov         esi, JSAMPARRAY edi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY edi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY edi+2*SIZEOF_JSAMPARRAY
+    lea         esi, esi+ecx*SIZEOF_JSAMPROW
+    lea         ebx, ebx+ecx*SIZEOF_JSAMPROW
+    lea         edx, edx+ecx*SIZEOF_JSAMPROW
+
+    pop         ecx
+
+    mov         edi, JSAMPARRAY output_buf(eax)
+    mov         eax, INT num_rows(eax)
+    test        eax, eax
+    jle         near .return
+    alignx      16, 7
+.rowloop:
+    push        eax
+    push        edi
+    push        edx
+    push        ebx
+    push        esi
+    push        ecx                     ; col
+
+    mov         esi, JSAMPROW esi     ; inptr0
+    mov         ebx, JSAMPROW ebx     ; inptr1
+    mov         edx, JSAMPROW edx     ; inptr2
+    mov         edi, JSAMPROW edi     ; outptr
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+    alignx      16, 7
+.columnloop:
+
+    vmovdqu     ymm5, YMMWORD ebx     ; ymm5=Cb(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+    vmovdqu     ymm1, YMMWORD edx     ; ymm1=Cr(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+
+    vpcmpeqw    ymm0, ymm0, ymm0
+    vpcmpeqw    ymm7, ymm7, ymm7
+    vpsrlw      ymm0, ymm0, BYTE_BIT    ; ymm0={0xFF 0x00 0xFF 0x00 ..}
+    vpsllw      ymm7, ymm7, 7           ; ymm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    vpand       ymm4, ymm0, ymm5        ; ymm4=Cb(02468ACEGIKMOQSU)=CbE
+    vpsrlw      ymm5, ymm5, BYTE_BIT    ; ymm5=Cb(13579BDFHJLNPRTV)=CbO
+    vpand       ymm0, ymm0, ymm1        ; ymm0=Cr(02468ACEGIKMOQSU)=CrE
+    vpsrlw      ymm1, ymm1, BYTE_BIT    ; ymm1=Cr(13579BDFHJLNPRTV)=CrO
+
+    vpaddw      ymm2, ymm4, ymm7
+    vpaddw      ymm3, ymm5, ymm7
+    vpaddw      ymm6, ymm0, ymm7
+    vpaddw      ymm7, ymm1, ymm7
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    vpaddw      ymm4, ymm2, ymm2                     ; ymm4=2*CbE
+    vpaddw      ymm5, ymm3, ymm3                     ; ymm5=2*CbO
+    vpaddw      ymm0, ymm6, ymm6                     ; ymm0=2*CrE
+    vpaddw      ymm1, ymm7, ymm7                     ; ymm1=2*CrO
+
+    vpmulhw     ymm4, ymm4, GOTOFF(eax,PW_MF0228)  ; ymm4=(2*CbE * -FIX(0.22800))
+    vpmulhw     ymm5, ymm5, GOTOFF(eax,PW_MF0228)  ; ymm5=(2*CbO * -FIX(0.22800))
+    vpmulhw     ymm0, ymm0, GOTOFF(eax,PW_F0402)   ; ymm0=(2*CrE * FIX(0.40200))
+    vpmulhw     ymm1, ymm1, GOTOFF(eax,PW_F0402)   ; ymm1=(2*CrO * FIX(0.40200))
+
+    vpaddw      ymm4, ymm4, GOTOFF(eax,PW_ONE)
+    vpaddw      ymm5, ymm5, GOTOFF(eax,PW_ONE)
+    vpsraw      ymm4, ymm4, 1                        ; ymm4=(CbE * -FIX(0.22800))
+    vpsraw      ymm5, ymm5, 1                        ; ymm5=(CbO * -FIX(0.22800))
+    vpaddw      ymm0, ymm0, GOTOFF(eax,PW_ONE)
+    vpaddw      ymm1, ymm1, GOTOFF(eax,PW_ONE)
+    vpsraw      ymm0, ymm0, 1                        ; ymm0=(CrE * FIX(0.40200))
+    vpsraw      ymm1, ymm1, 1                        ; ymm1=(CrO * FIX(0.40200))
+
+    vpaddw      ymm4, ymm4, ymm2
+    vpaddw      ymm5, ymm5, ymm3
+    vpaddw      ymm4, ymm4, ymm2                     ; ymm4=(CbE * FIX(1.77200))=(B-Y)E
+    vpaddw      ymm5, ymm5, ymm3                     ; ymm5=(CbO * FIX(1.77200))=(B-Y)O
+    vpaddw      ymm0, ymm0, ymm6                     ; ymm0=(CrE * FIX(1.40200))=(R-Y)E
+    vpaddw      ymm1, ymm1, ymm7                     ; ymm1=(CrO * FIX(1.40200))=(R-Y)O
+
+    vmovdqa     YMMWORD wk(0), ymm4                ; wk(0)=(B-Y)E
+    vmovdqa     YMMWORD wk(1), ymm5                ; wk(1)=(B-Y)O
+
+    vpunpckhwd  ymm4, ymm2, ymm6
+    vpunpcklwd  ymm2, ymm2, ymm6
+    vpmaddwd    ymm2, ymm2, GOTOFF(eax,PW_MF0344_F0285)
+    vpmaddwd    ymm4, ymm4, GOTOFF(eax,PW_MF0344_F0285)
+    vpunpckhwd  ymm5, ymm3, ymm7
+    vpunpcklwd  ymm3, ymm3, ymm7
+    vpmaddwd    ymm3, ymm3, GOTOFF(eax,PW_MF0344_F0285)
+    vpmaddwd    ymm5, ymm5, GOTOFF(eax,PW_MF0344_F0285)
+
+    vpaddd      ymm2, ymm2, GOTOFF(eax,PD_ONEHALF)
+    vpaddd      ymm4, ymm4, GOTOFF(eax,PD_ONEHALF)
+    vpsrad      ymm2, ymm2, SCALEBITS
+    vpsrad      ymm4, ymm4, SCALEBITS
+    vpaddd      ymm3, ymm3, GOTOFF(eax,PD_ONEHALF)
+    vpaddd      ymm5, ymm5, GOTOFF(eax,PD_ONEHALF)
+    vpsrad      ymm3, ymm3, SCALEBITS
+    vpsrad      ymm5, ymm5, SCALEBITS
+
+    vpackssdw   ymm2, ymm2, ymm4             ; ymm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+    vpackssdw   ymm3, ymm3, ymm5             ; ymm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+    vpsubw      ymm2, ymm2, ymm6             ; ymm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+    vpsubw      ymm3, ymm3, ymm7             ; ymm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+    vmovdqu     ymm5, YMMWORD esi          ; ymm5=Y(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+
+    vpcmpeqw    ymm4, ymm4, ymm4
+    vpsrlw      ymm4, ymm4, BYTE_BIT         ; ymm4={0xFF 0x00 0xFF 0x00 ..}
+    vpand       ymm4, ymm4, ymm5             ; ymm4=Y(02468ACEGIKMOQSU)=YE
+    vpsrlw      ymm5, ymm5, BYTE_BIT         ; ymm5=Y(13579BDFHJLNPRTV)=YO
+
+    vpaddw      ymm0, ymm0, ymm4             ; ymm0=((R-Y)E+YE)=RE=R(02468ACEGIKMOQSU)
+    vpaddw      ymm1, ymm1, ymm5             ; ymm1=((R-Y)O+YO)=RO=R(13579BDFHJLNPRTV)
+    vpackuswb   ymm0, ymm0, ymm0             ; ymm0=R(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm1, ymm1, ymm1             ; ymm1=R(13579BDF********HJLNPRTV********)
+
+    vpaddw      ymm2, ymm2, ymm4             ; ymm2=((G-Y)E+YE)=GE=G(02468ACEGIKMOQSU)
+    vpaddw      ymm3, ymm3, ymm5             ; ymm3=((G-Y)O+YO)=GO=G(13579BDFHJLNPRTV)
+    vpackuswb   ymm2, ymm2, ymm2             ; ymm2=G(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm3, ymm3, ymm3             ; ymm3=G(13579BDF********HJLNPRTV********)
+
+    vpaddw      ymm4, ymm4, YMMWORD wk(0)  ; ymm4=(YE+(B-Y)E)=BE=B(02468ACEGIKMOQSU)
+    vpaddw      ymm5, ymm5, YMMWORD wk(1)  ; ymm5=(YO+(B-Y)O)=BO=B(13579BDFHJLNPRTV)
+    vpackuswb   ymm4, ymm4, ymm4             ; ymm4=B(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm5, ymm5, ymm5             ; ymm5=B(13579BDF********HJLNPRTV********)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; ymmA=(00 02 04 06 08 0A 0C 0E ** 0G 0I 0K 0M 0O 0Q 0S 0U **)
+    ; ymmB=(01 03 05 07 09 0B 0D 0F ** 0H 0J 0L 0N 0P 0R 0T 0V **)
+    ; ymmC=(10 12 14 16 18 1A 1C 1E ** 1G 1I 1K 1M 1O 1Q 1S 1U **)
+    ; ymmD=(11 13 15 17 19 1B 1D 1F ** 1H 1J 1L 1N 1P 1R 1T 1V **)
+    ; ymmE=(20 22 24 26 28 2A 2C 2E ** 2G 2I 2K 2M 2O 2Q 2S 2U **)
+    ; ymmF=(21 23 25 27 29 2B 2D 2F ** 2H 2J 2L 2N 2P 2R 2T 2V **)
+    ; ymmG=(** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** **)
+    ; ymmH=(** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** **)
+
+    vpunpcklbw  ymmA, ymmA, ymmC        ; ymmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E
+                                        ;       0G 1G 0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U)
+    vpunpcklbw  ymmE, ymmE, ymmB        ; ymmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F
+                                        ;       2G 0H 2I 0J 2K 0L 2M 0N 2O 0P 2Q 0R 2S 0T 2U 0V)
+    vpunpcklbw  ymmD, ymmD, ymmF        ; ymmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F
+                                        ;       1H 2H 1J 2J 1L 2L 1N 2N 1P 2P 1R 2R 1T 2T 1V 2V)
+
+    vpsrldq     ymmH, ymmA, 2           ; ymmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E 0G 1G
+                                        ;       0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U -- --)
+    vpunpckhwd  ymmG, ymmA, ymmE        ; ymmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F
+                                        ;       0O 1O 2O 0P 0Q 1Q 2Q 0R 0S 1S 2S 0T 0U 1U 2U 0V)
+    vpunpcklwd  ymmA, ymmA, ymmE        ; ymmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07
+                                        ;       0G 1G 2G 0H 0I 1I 2I 0J 0K 1K 2K 0L 0M 1M 2M 0N)
+
+    vpsrldq     ymmE, ymmE, 2           ; ymmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F 2G 0H
+                                        ;       2I 0J 2K 0L 2M 0N 2O 0P 2Q 0R 2S 0T 2U 0V -- --)
+
+    vpsrldq     ymmB, ymmD, 2           ; ymmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F 1H 2H
+                                        ;       1J 2J 1L 2L 1N 2N 1P 2P 1R 2R 1T 2T 1V 2V -- --)
+    vpunpckhwd  ymmC, ymmD, ymmH        ; ymmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F 0G 1G
+                                        ;       1P 2P 0Q 1Q 1R 2R 0S 1S 1T 2T 0U 1U 1V 2V -- --)
+    vpunpcklwd  ymmD, ymmD, ymmH        ; ymmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18
+                                        ;       1H 2H 0I 1I 1J 2J 0K 1K 1L 2L 0M 1M 1N 2N 0O 1O)
+
+    vpunpckhwd  ymmF, ymmE, ymmB        ; ymmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F 2G 0H 1H 2H
+                                        ;       2Q 0R 1R 2R 2S 0T 1T 2T 2U 0V 1V 2V -- -- -- --)
+    vpunpcklwd  ymmE, ymmE, ymmB        ; ymmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29
+                                        ;       2I 0J 1J 2J 2K 0L 1L 2L 2M 0N 1N 2N 2O 0P 1P 2P)
+
+    vpshufd     ymmH, ymmA, 0x4E        ; ymmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03
+                                        ;       0K 1K 2K 0L 0M 1M 2M 0N 0G 1G 2G 0H 0I 1I 2I 0J)
+    vpunpckldq  ymmA, ymmA, ymmD        ; ymmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 0I 1I 2I 0J 1J 2J 0K 1K)
+    vpunpckhdq  ymmD, ymmD, ymmE        ; ymmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29
+                                        ;       1L 2L 0M 1M 2M 0N 1N 2N 1N 2N 0O 1O 2O 0P 1P 2P)
+    vpunpckldq  ymmE, ymmE, ymmH        ; ymmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07
+                                        ;       2I 0J 1J 2J 0K 1K 2K 0L 2K 0L 1L 2L 0M 1M 2M 0N)
+
+    vpshufd     ymmH, ymmG, 0x4E        ; ymmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B
+                                        ;       0S 1S 2S 0T 0U 1U 2U 0V 0O 1O 2O 0P 0Q 1Q 2Q 0R)
+    vpunpckldq  ymmG, ymmG, ymmC        ; ymmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C
+                                        ;       0O 1O 2O 0P 1P 2P 0Q 1Q 0Q 1Q 2Q 0R 1R 2R 0S 1S)
+    vpunpckhdq  ymmC, ymmC, ymmF        ; ymmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F 0G 1G 2G 0H 1H 2H
+                                        ;       1T 2T 0U 1U 2U 0V 1V 2V 1V 2V -- -- -- -- -- --)
+    vpunpckldq  ymmF, ymmF, ymmH        ; ymmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 2S 0T 1T 2T 0U 1U 2U 0V)
+
+    vpunpcklqdq ymmH, ymmA, ymmE        ; ymmH=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vpunpcklqdq ymmG, ymmD, ymmG        ; ymmG=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A
+                                        ;       1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q)
+    vpunpcklqdq ymmC, ymmF, ymmC        ; ymmC=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    vperm2i128  ymmA, ymmH, ymmG, 0x20  ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                        ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    vperm2i128  ymmD, ymmC, ymmH, 0x30  ; ymmD=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vperm2i128  ymmF, ymmG, ymmC, 0x31  ; ymmF=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jb          short .column_st64
+
+    test        edi, SIZEOF_YMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    vmovntdq    YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovntdq    YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    vmovntdq    YMMWORD edi+2*SIZEOF_YMMWORD, ymmF
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    vmovdqu     YMMWORD edi+2*SIZEOF_YMMWORD, ymmF
+.out0:
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_YMMWORD  ; outptr
+    sub         ecx, byte SIZEOF_YMMWORD
+    jz          near .nextrow
+
+    add         esi, byte SIZEOF_YMMWORD  ; inptr0
+    add         ebx, byte SIZEOF_YMMWORD  ; inptr1
+    add         edx, byte SIZEOF_YMMWORD  ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st64:
+    lea         ecx, ecx+ecx*2            ; imul ecx, RGB_PIXELSIZE
+    cmp         ecx, byte 2*SIZEOF_YMMWORD
+    jb          short .column_st32
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    add         edi, byte 2*SIZEOF_YMMWORD  ; outptr
+    vmovdqa     ymmA, ymmF
+    sub         ecx, byte 2*SIZEOF_YMMWORD
+    jmp         short .column_st31
+.column_st32:
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jb          short .column_st31
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    add         edi, byte SIZEOF_YMMWORD    ; outptr
+    vmovdqa     ymmA, ymmD
+    sub         ecx, byte SIZEOF_YMMWORD
+    jmp         short .column_st31
+.column_st31:
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jb          short .column_st15
+    vmovdqu     XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    add         edi, byte SIZEOF_XMMWORD    ; outptr
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    sub         ecx, byte SIZEOF_XMMWORD
+.column_st15:
+    ; Store the lower 8 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st7
+    vmovq       XMM_MMWORD edi, xmmA
+    add         edi, byte SIZEOF_MMWORD
+    sub         ecx, byte SIZEOF_MMWORD
+    vpsrldq     xmmA, xmmA, SIZEOF_MMWORD
+.column_st7:
+    ; Store the lower 4 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_DWORD
+    jb          short .column_st3
+    vmovd       XMM_DWORD edi, xmmA
+    add         edi, byte SIZEOF_DWORD
+    sub         ecx, byte SIZEOF_DWORD
+    vpsrldq     xmmA, xmmA, SIZEOF_DWORD
+.column_st3:
+    ; Store the lower 2 bytes of eax to the output when it has enough
+    ; space.
+    vmovd       eax, xmmA
+    cmp         ecx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word edi, ax
+    add         edi, byte SIZEOF_WORD
+    sub         ecx, byte SIZEOF_WORD
+    shr         eax, 16
+.column_st1:
+    ; Store the lower 1 byte of eax to the output when it has enough
+    ; space.
+    test        ecx, ecx
+    jz          short .nextrow
+    mov         byte edi, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    vpcmpeqb    ymm6, ymm6, ymm6        ; ymm6=XE=X(02468ACE********GIKMOQSU********)
+    vpcmpeqb    ymm7, ymm7, ymm7        ; ymm7=XO=X(13579BDF********HJLNPRTV********)
+%else
+    vpxor       ymm6, ymm6, ymm6        ; ymm6=XE=X(02468ACE********GIKMOQSU********)
+    vpxor       ymm7, ymm7, ymm7        ; ymm7=XO=X(13579BDF********HJLNPRTV********)
+%endif
+    ; ymmA=(00 02 04 06 08 0A 0C 0E ** 0G 0I 0K 0M 0O 0Q 0S 0U **)
+    ; ymmB=(01 03 05 07 09 0B 0D 0F ** 0H 0J 0L 0N 0P 0R 0T 0V **)
+    ; ymmC=(10 12 14 16 18 1A 1C 1E ** 1G 1I 1K 1M 1O 1Q 1S 1U **)
+    ; ymmD=(11 13 15 17 19 1B 1D 1F ** 1H 1J 1L 1N 1P 1R 1T 1V **)
+    ; ymmE=(20 22 24 26 28 2A 2C 2E ** 2G 2I 2K 2M 2O 2Q 2S 2U **)
+    ; ymmF=(21 23 25 27 29 2B 2D 2F ** 2H 2J 2L 2N 2P 2R 2T 2V **)
+    ; ymmG=(30 32 34 36 38 3A 3C 3E ** 3G 3I 3K 3M 3O 3Q 3S 3U **)
+    ; ymmH=(31 33 35 37 39 3B 3D 3F ** 3H 3J 3L 3N 3P 3R 3T 3V **)
+
+    vpunpcklbw  ymmA, ymmA, ymmC        ; ymmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E
+                                        ;       0G 1G 0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U)
+    vpunpcklbw  ymmE, ymmE, ymmG        ; ymmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E
+                                        ;       2G 3G 2I 3I 2K 3K 2M 3M 2O 3O 2Q 3Q 2S 3S 2U 3U)
+    vpunpcklbw  ymmB, ymmB, ymmD        ; ymmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F
+                                        ;       0H 1H 0J 1J 0L 1L 0N 1N 0P 1P 0R 1R 0T 1T 0V 1V)
+    vpunpcklbw  ymmF, ymmF, ymmH        ; ymmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F
+                                        ;       2H 3H 2J 3J 2L 3L 2N 3N 2P 3P 2R 3R 2T 3T 2V 3V)
+
+    vpunpckhwd  ymmC, ymmA, ymmE        ; ymmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E
+                                        ;       0O 1O 2O 3O 0Q 1Q 2Q 3Q 0S 1S 2S 3S 0U 1U 2U 3U)
+    vpunpcklwd  ymmA, ymmA, ymmE        ; ymmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36
+                                        ;       0G 1G 2G 3G 0I 1I 2I 3I 0K 1K 2K 3K 0M 1M 2M 3M)
+    vpunpckhwd  ymmG, ymmB, ymmF        ; ymmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F
+                                        ;       0P 1P 2P 3P 0R 1R 2R 3R 0T 1T 2T 3T 0V 1V 2V 3V)
+    vpunpcklwd  ymmB, ymmB, ymmF        ; ymmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37
+                                        ;       0H 1H 2H 3H 0J 1J 2J 3J 0L 1L 2L 3L 0N 1N 2N 3N)
+
+    vpunpckhdq  ymmE, ymmA, ymmB        ; ymmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    vpunpckldq  ymmB, ymmA, ymmB        ; ymmB=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J)
+    vpunpckhdq  ymmF, ymmC, ymmG        ; ymmF=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+    vpunpckldq  ymmG, ymmC, ymmG        ; ymmG=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R)
+
+    vperm2i128  ymmA, ymmB, ymmE, 0x20  ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    vperm2i128  ymmD, ymmG, ymmF, 0x20  ; ymmD=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+    vperm2i128  ymmC, ymmB, ymmE, 0x31  ; ymmC=(0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    vperm2i128  ymmH, ymmG, ymmF, 0x31  ; ymmH=(0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jb          short .column_st64
+
+    test        edi, SIZEOF_YMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    vmovntdq    YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovntdq    YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    vmovntdq    YMMWORD edi+2*SIZEOF_YMMWORD, ymmC
+    vmovntdq    YMMWORD edi+3*SIZEOF_YMMWORD, ymmH
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    vmovdqu     YMMWORD edi+2*SIZEOF_YMMWORD, ymmC
+    vmovdqu     YMMWORD edi+3*SIZEOF_YMMWORD, ymmH
+.out0:
+    add         edi, RGB_PIXELSIZE*SIZEOF_YMMWORD  ; outptr
+    sub         ecx, byte SIZEOF_YMMWORD
+    jz          near .nextrow
+
+    add         esi, byte SIZEOF_YMMWORD  ; inptr0
+    add         ebx, byte SIZEOF_YMMWORD  ; inptr1
+    add         edx, byte SIZEOF_YMMWORD  ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st64:
+    cmp         ecx, byte SIZEOF_YMMWORD/2
+    jb          short .column_st32
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    add         edi, byte 2*SIZEOF_YMMWORD  ; outptr
+    vmovdqa     ymmA, ymmC
+    vmovdqa     ymmD, ymmH
+    sub         ecx, byte SIZEOF_YMMWORD/2
+.column_st32:
+    cmp         ecx, byte SIZEOF_YMMWORD/4
+    jb          short .column_st16
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    add         edi, byte SIZEOF_YMMWORD    ; outptr
+    vmovdqa     ymmA, ymmD
+    sub         ecx, byte SIZEOF_YMMWORD/4
+.column_st16:
+    cmp         ecx, byte SIZEOF_YMMWORD/8
+    jb          short .column_st15
+    vmovdqu     XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    add         edi, byte SIZEOF_XMMWORD    ; outptr
+    sub         ecx, byte SIZEOF_YMMWORD/8
+.column_st15:
+    ; Store two pixels (8 bytes) of ymmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_YMMWORD/16
+    jb          short .column_st7
+    vmovq       MMWORD edi, xmmA
+    add         edi, byte SIZEOF_YMMWORD/16*4
+    sub         ecx, byte SIZEOF_YMMWORD/16
+    vpsrldq     xmmA, SIZEOF_YMMWORD/16*4
+.column_st7:
+    ; Store one pixel (4 bytes) of ymmA to the output when it has enough
+    ; space.
+    test        ecx, ecx
+    jz          short .nextrow
+    vmovd       XMM_DWORD edi, xmmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    alignx      16, 7
+
+.nextrow:
+    pop         ecx
+    pop         esi
+    pop         ebx
+    pop         edx
+    pop         edi
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW
+    add         ebx, byte SIZEOF_JSAMPROW
+    add         edx, byte SIZEOF_JSAMPROW
+    add         edi, byte SIZEOF_JSAMPROW  ; output_buf
+    dec         eax                        ; num_rows
+    jg          near .rowloop
+
+    sfence                              ; flush the write buffer
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdcolext-mmx.asm Added

@@ -0,0 +1,404 @@
+;
+; jdcolext.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_mmx(JDIMENSION out_width, JSAMPIMAGE input_buf,
+;                           JDIMENSION input_row, JSAMPARRAY output_buf,
+;                           int num_rows)
+;
+
+%define out_width(b)   (b) + 8          ; JDIMENSION out_width
+%define input_buf(b)   (b) + 12         ; JSAMPIMAGE input_buf
+%define input_row(b)   (b) + 16         ; JDIMENSION input_row
+%define output_buf(b)  (b) + 20         ; JSAMPARRAY output_buf
+%define num_rows(b)    (b) + 24         ; int num_rows
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_MMWORD
+                                        ; mmword wkWK_NUM
+%define WK_NUM         2
+%define gotptr         wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_ycc_rgb_convert_mmx)
+
+EXTN(jsimd_ycc_rgb_convert_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION out_width(eax)  ; num_cols
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         edi, JSAMPIMAGE input_buf(eax)
+    mov         ecx, JDIMENSION input_row(eax)
+    mov         esi, JSAMPARRAY edi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY edi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY edi+2*SIZEOF_JSAMPARRAY
+    lea         esi, esi+ecx*SIZEOF_JSAMPROW
+    lea         ebx, ebx+ecx*SIZEOF_JSAMPROW
+    lea         edx, edx+ecx*SIZEOF_JSAMPROW
+
+    pop         ecx
+
+    mov         edi, JSAMPARRAY output_buf(eax)
+    mov         eax, INT num_rows(eax)
+    test        eax, eax
+    jle         near .return
+    alignx      16, 7
+.rowloop:
+    push        eax
+    push        edi
+    push        edx
+    push        ebx
+    push        esi
+    push        ecx                     ; col
+
+    mov         esi, JSAMPROW esi     ; inptr0
+    mov         ebx, JSAMPROW ebx     ; inptr1
+    mov         edx, JSAMPROW edx     ; inptr2
+    mov         edi, JSAMPROW edi     ; outptr
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+    alignx      16, 7
+.columnloop:
+
+    movq        mm5, MMWORD ebx       ; mm5=Cb(01234567)
+    movq        mm1, MMWORD edx       ; mm1=Cr(01234567)
+
+    pcmpeqw     mm4, mm4
+    pcmpeqw     mm7, mm7
+    psrlw       mm4, BYTE_BIT
+    psllw       mm7, 7                  ; mm7={0xFF80 0xFF80 0xFF80 0xFF80}
+    movq        mm0, mm4                ; mm0=mm4={0xFF 0x00 0xFF 0x00 ..}
+
+    pand        mm4, mm5                ; mm4=Cb(0246)=CbE
+    psrlw       mm5, BYTE_BIT           ; mm5=Cb(1357)=CbO
+    pand        mm0, mm1                ; mm0=Cr(0246)=CrE
+    psrlw       mm1, BYTE_BIT           ; mm1=Cr(1357)=CrO
+
+    paddw       mm4, mm7
+    paddw       mm5, mm7
+    paddw       mm0, mm7
+    paddw       mm1, mm7
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    movq        mm2, mm4                ; mm2=CbE
+    movq        mm3, mm5                ; mm3=CbO
+    paddw       mm4, mm4                ; mm4=2*CbE
+    paddw       mm5, mm5                ; mm5=2*CbO
+    movq        mm6, mm0                ; mm6=CrE
+    movq        mm7, mm1                ; mm7=CrO
+    paddw       mm0, mm0                ; mm0=2*CrE
+    paddw       mm1, mm1                ; mm1=2*CrO
+
+    pmulhw      mm4, GOTOFF(eax,PW_MF0228)  ; mm4=(2*CbE * -FIX(0.22800))
+    pmulhw      mm5, GOTOFF(eax,PW_MF0228)  ; mm5=(2*CbO * -FIX(0.22800))
+    pmulhw      mm0, GOTOFF(eax,PW_F0402)   ; mm0=(2*CrE * FIX(0.40200))
+    pmulhw      mm1, GOTOFF(eax,PW_F0402)   ; mm1=(2*CrO * FIX(0.40200))
+
+    paddw       mm4, GOTOFF(eax,PW_ONE)
+    paddw       mm5, GOTOFF(eax,PW_ONE)
+    psraw       mm4, 1                  ; mm4=(CbE * -FIX(0.22800))
+    psraw       mm5, 1                  ; mm5=(CbO * -FIX(0.22800))
+    paddw       mm0, GOTOFF(eax,PW_ONE)
+    paddw       mm1, GOTOFF(eax,PW_ONE)
+    psraw       mm0, 1                  ; mm0=(CrE * FIX(0.40200))
+    psraw       mm1, 1                  ; mm1=(CrO * FIX(0.40200))
+
+    paddw       mm4, mm2
+    paddw       mm5, mm3
+    paddw       mm4, mm2                ; mm4=(CbE * FIX(1.77200))=(B-Y)E
+    paddw       mm5, mm3                ; mm5=(CbO * FIX(1.77200))=(B-Y)O
+    paddw       mm0, mm6                ; mm0=(CrE * FIX(1.40200))=(R-Y)E
+    paddw       mm1, mm7                ; mm1=(CrO * FIX(1.40200))=(R-Y)O
+
+    movq        MMWORD wk(0), mm4     ; wk(0)=(B-Y)E
+    movq        MMWORD wk(1), mm5     ; wk(1)=(B-Y)O
+
+    movq        mm4, mm2
+    movq        mm5, mm3
+    punpcklwd   mm2, mm6
+    punpckhwd   mm4, mm6
+    pmaddwd     mm2, GOTOFF(eax,PW_MF0344_F0285)
+    pmaddwd     mm4, GOTOFF(eax,PW_MF0344_F0285)
+    punpcklwd   mm3, mm7
+    punpckhwd   mm5, mm7
+    pmaddwd     mm3, GOTOFF(eax,PW_MF0344_F0285)
+    pmaddwd     mm5, GOTOFF(eax,PW_MF0344_F0285)
+
+    paddd       mm2, GOTOFF(eax,PD_ONEHALF)
+    paddd       mm4, GOTOFF(eax,PD_ONEHALF)
+    psrad       mm2, SCALEBITS
+    psrad       mm4, SCALEBITS
+    paddd       mm3, GOTOFF(eax,PD_ONEHALF)
+    paddd       mm5, GOTOFF(eax,PD_ONEHALF)
+    psrad       mm3, SCALEBITS
+    psrad       mm5, SCALEBITS
+
+    packssdw    mm2, mm4                ; mm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+    packssdw    mm3, mm5                ; mm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+    psubw       mm2, mm6                ; mm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+    psubw       mm3, mm7                ; mm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+    movq        mm5, MMWORD esi       ; mm5=Y(01234567)
+
+    pcmpeqw     mm4, mm4
+    psrlw       mm4, BYTE_BIT           ; mm4={0xFF 0x00 0xFF 0x00 ..}
+    pand        mm4, mm5                ; mm4=Y(0246)=YE
+    psrlw       mm5, BYTE_BIT           ; mm5=Y(1357)=YO
+
+    paddw       mm0, mm4                ; mm0=((R-Y)E+YE)=RE=(R0 R2 R4 R6)
+    paddw       mm1, mm5                ; mm1=((R-Y)O+YO)=RO=(R1 R3 R5 R7)
+    packuswb    mm0, mm0                ; mm0=(R0 R2 R4 R6 ** ** ** **)
+    packuswb    mm1, mm1                ; mm1=(R1 R3 R5 R7 ** ** ** **)
+
+    paddw       mm2, mm4                ; mm2=((G-Y)E+YE)=GE=(G0 G2 G4 G6)
+    paddw       mm3, mm5                ; mm3=((G-Y)O+YO)=GO=(G1 G3 G5 G7)
+    packuswb    mm2, mm2                ; mm2=(G0 G2 G4 G6 ** ** ** **)
+    packuswb    mm3, mm3                ; mm3=(G1 G3 G5 G7 ** ** ** **)
+
+    paddw       mm4,  MMWORD wk(0)    ; mm4=(YE+(B-Y)E)=BE=(B0 B2 B4 B6)
+    paddw       mm5,  MMWORD wk(1)    ; mm5=(YO+(B-Y)O)=BO=(B1 B3 B5 B7)
+    packuswb    mm4, mm4                ; mm4=(B0 B2 B4 B6 ** ** ** **)
+    packuswb    mm5, mm5                ; mm5=(B1 B3 B5 B7 ** ** ** **)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+    ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+    ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+    ; mmG=(** ** ** ** ** ** ** **), mmH=(** ** ** ** ** ** ** **)
+
+    punpcklbw   mmA, mmC                ; mmA=(00 10 02 12 04 14 06 16)
+    punpcklbw   mmE, mmB                ; mmE=(20 01 22 03 24 05 26 07)
+    punpcklbw   mmD, mmF                ; mmD=(11 21 13 23 15 25 17 27)
+
+    movq        mmG, mmA
+    movq        mmH, mmA
+    punpcklwd   mmA, mmE                ; mmA=(00 10 20 01 02 12 22 03)
+    punpckhwd   mmG, mmE                ; mmG=(04 14 24 05 06 16 26 07)
+
+    psrlq       mmH, 2*BYTE_BIT         ; mmH=(02 12 04 14 06 16 -- --)
+    psrlq       mmE, 2*BYTE_BIT         ; mmE=(22 03 24 05 26 07 -- --)
+
+    movq        mmC, mmD
+    movq        mmB, mmD
+    punpcklwd   mmD, mmH                ; mmD=(11 21 02 12 13 23 04 14)
+    punpckhwd   mmC, mmH                ; mmC=(15 25 06 16 17 27 -- --)
+
+    psrlq       mmB, 2*BYTE_BIT         ; mmB=(13 23 15 25 17 27 -- --)
+
+    movq        mmF, mmE
+    punpcklwd   mmE, mmB                ; mmE=(22 03 13 23 24 05 15 25)
+    punpckhwd   mmF, mmB                ; mmF=(26 07 17 27 -- -- -- --)
+
+    punpckldq   mmA, mmD                ; mmA=(00 10 20 01 11 21 02 12)
+    punpckldq   mmE, mmG                ; mmE=(22 03 13 23 04 14 24 05)
+    punpckldq   mmC, mmF                ; mmC=(15 25 06 16 26 07 17 27)
+
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st16
+
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mmE
+    movq        MMWORD edi+2*SIZEOF_MMWORD, mmC
+
+    sub         ecx, byte SIZEOF_MMWORD
+    jz          short .nextrow
+
+    add         esi, byte SIZEOF_MMWORD                ; inptr0
+    add         ebx, byte SIZEOF_MMWORD                ; inptr1
+    add         edx, byte SIZEOF_MMWORD                ; inptr2
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD  ; outptr
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st16:
+    lea         ecx, ecx+ecx*2        ; imul ecx, RGB_PIXELSIZE
+    cmp         ecx, byte 2*SIZEOF_MMWORD
+    jb          short .column_st8
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mmE
+    movq        mmA, mmC
+    sub         ecx, byte 2*SIZEOF_MMWORD
+    add         edi, byte 2*SIZEOF_MMWORD
+    jmp         short .column_st4
+.column_st8:
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st4
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        mmA, mmE
+    sub         ecx, byte SIZEOF_MMWORD
+    add         edi, byte SIZEOF_MMWORD
+.column_st4:
+    movd        eax, mmA
+    cmp         ecx, byte SIZEOF_DWORD
+    jb          short .column_st2
+    mov         dword edi+0*SIZEOF_DWORD, eax
+    psrlq       mmA, DWORD_BIT
+    movd        eax, mmA
+    sub         ecx, byte SIZEOF_DWORD
+    add         edi, byte SIZEOF_DWORD
+.column_st2:
+    cmp         ecx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word edi+0*SIZEOF_WORD, ax
+    shr         eax, WORD_BIT
+    sub         ecx, byte SIZEOF_WORD
+    add         edi, byte SIZEOF_WORD
+.column_st1:
+    cmp         ecx, byte SIZEOF_BYTE
+    jb          short .nextrow
+    mov         byte edi+0*SIZEOF_BYTE, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    pcmpeqb     mm6, mm6                ; mm6=(X0 X2 X4 X6 ** ** ** **)
+    pcmpeqb     mm7, mm7                ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%else
+    pxor        mm6, mm6                ; mm6=(X0 X2 X4 X6 ** ** ** **)
+    pxor        mm7, mm7                ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%endif
+    ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+    ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+    ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+    ; mmG=(30 32 34 36 ** ** ** **), mmH=(31 33 35 37 ** ** ** **)
+
+    punpcklbw   mmA, mmC                ; mmA=(00 10 02 12 04 14 06 16)
+    punpcklbw   mmE, mmG                ; mmE=(20 30 22 32 24 34 26 36)
+    punpcklbw   mmB, mmD                ; mmB=(01 11 03 13 05 15 07 17)
+    punpcklbw   mmF, mmH                ; mmF=(21 31 23 33 25 35 27 37)
+
+    movq        mmC, mmA
+    punpcklwd   mmA, mmE                ; mmA=(00 10 20 30 02 12 22 32)
+    punpckhwd   mmC, mmE                ; mmC=(04 14 24 34 06 16 26 36)
+    movq        mmG, mmB
+    punpcklwd   mmB, mmF                ; mmB=(01 11 21 31 03 13 23 33)
+    punpckhwd   mmG, mmF                ; mmG=(05 15 25 35 07 17 27 37)
+
+    movq        mmD, mmA
+    punpckldq   mmA, mmB                ; mmA=(00 10 20 30 01 11 21 31)
+    punpckhdq   mmD, mmB                ; mmD=(02 12 22 32 03 13 23 33)
+    movq        mmH, mmC
+    punpckldq   mmC, mmG                ; mmC=(04 14 24 34 05 15 25 35)
+    punpckhdq   mmH, mmG                ; mmH=(06 16 26 36 07 17 27 37)
+
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st16
+
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mmD
+    movq        MMWORD edi+2*SIZEOF_MMWORD, mmC
+    movq        MMWORD edi+3*SIZEOF_MMWORD, mmH
+
+    sub         ecx, byte SIZEOF_MMWORD
+    jz          short .nextrow
+
+    add         esi, byte SIZEOF_MMWORD                ; inptr0
+    add         ebx, byte SIZEOF_MMWORD                ; inptr1
+    add         edx, byte SIZEOF_MMWORD                ; inptr2
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD  ; outptr
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st16:
+    cmp         ecx, byte SIZEOF_MMWORD/2
+    jb          short .column_st8
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mmD
+    movq        mmA, mmC
+    movq        mmD, mmH
+    sub         ecx, byte SIZEOF_MMWORD/2
+    add         edi, byte 2*SIZEOF_MMWORD
+.column_st8:
+    cmp         ecx, byte SIZEOF_MMWORD/4
+    jb          short .column_st4
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        mmA, mmD
+    sub         ecx, byte SIZEOF_MMWORD/4
+    add         edi, byte 1*SIZEOF_MMWORD
+.column_st4:
+    cmp         ecx, byte SIZEOF_MMWORD/8
+    jb          short .nextrow
+    movd        dword edi+0*SIZEOF_DWORD, mmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    alignx      16, 7
+
+.nextrow:
+    pop         ecx
+    pop         esi
+    pop         ebx
+    pop         edx
+    pop         edi
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW
+    add         ebx, byte SIZEOF_JSAMPROW
+    add         edx, byte SIZEOF_JSAMPROW
+    add         edi, byte SIZEOF_JSAMPROW  ; output_buf
+    dec         eax                        ; num_rows
+    jg          near .rowloop
+
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdcolext-sse2.asm Added

@@ -0,0 +1,458 @@
+;
+; jdcolext.asm - colorspace conversion (SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2012, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_sse2(JDIMENSION out_width, JSAMPIMAGE input_buf,
+;                            JDIMENSION input_row, JSAMPARRAY output_buf,
+;                            int num_rows)
+;
+
+%define out_width(b)   (b) + 8          ; JDIMENSION out_width
+%define input_buf(b)   (b) + 12         ; JSAMPIMAGE input_buf
+%define input_row(b)   (b) + 16         ; JDIMENSION input_row
+%define output_buf(b)  (b) + 20         ; JSAMPARRAY output_buf
+%define num_rows(b)    (b) + 24         ; int num_rows
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM         2
+%define gotptr         wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_ycc_rgb_convert_sse2)
+
+EXTN(jsimd_ycc_rgb_convert_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION out_width(eax)  ; num_cols
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         edi, JSAMPIMAGE input_buf(eax)
+    mov         ecx, JDIMENSION input_row(eax)
+    mov         esi, JSAMPARRAY edi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY edi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY edi+2*SIZEOF_JSAMPARRAY
+    lea         esi, esi+ecx*SIZEOF_JSAMPROW
+    lea         ebx, ebx+ecx*SIZEOF_JSAMPROW
+    lea         edx, edx+ecx*SIZEOF_JSAMPROW
+
+    pop         ecx
+
+    mov         edi, JSAMPARRAY output_buf(eax)
+    mov         eax, INT num_rows(eax)
+    test        eax, eax
+    jle         near .return
+    alignx      16, 7
+.rowloop:
+    push        eax
+    push        edi
+    push        edx
+    push        ebx
+    push        esi
+    push        ecx                     ; col
+
+    mov         esi, JSAMPROW esi     ; inptr0
+    mov         ebx, JSAMPROW ebx     ; inptr1
+    mov         edx, JSAMPROW edx     ; inptr2
+    mov         edi, JSAMPROW edi     ; outptr
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+    alignx      16, 7
+.columnloop:
+
+    movdqa      xmm5, XMMWORD ebx     ; xmm5=Cb(0123456789ABCDEF)
+    movdqa      xmm1, XMMWORD edx     ; xmm1=Cr(0123456789ABCDEF)
+
+    pcmpeqw     xmm4, xmm4
+    pcmpeqw     xmm7, xmm7
+    psrlw       xmm4, BYTE_BIT
+    psllw       xmm7, 7                 ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+    movdqa      xmm0, xmm4              ; xmm0=xmm4={0xFF 0x00 0xFF 0x00 ..}
+
+    pand        xmm4, xmm5              ; xmm4=Cb(02468ACE)=CbE
+    psrlw       xmm5, BYTE_BIT          ; xmm5=Cb(13579BDF)=CbO
+    pand        xmm0, xmm1              ; xmm0=Cr(02468ACE)=CrE
+    psrlw       xmm1, BYTE_BIT          ; xmm1=Cr(13579BDF)=CrO
+
+    paddw       xmm4, xmm7
+    paddw       xmm5, xmm7
+    paddw       xmm0, xmm7
+    paddw       xmm1, xmm7
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    movdqa      xmm2, xmm4              ; xmm2=CbE
+    movdqa      xmm3, xmm5              ; xmm3=CbO
+    paddw       xmm4, xmm4              ; xmm4=2*CbE
+    paddw       xmm5, xmm5              ; xmm5=2*CbO
+    movdqa      xmm6, xmm0              ; xmm6=CrE
+    movdqa      xmm7, xmm1              ; xmm7=CrO
+    paddw       xmm0, xmm0              ; xmm0=2*CrE
+    paddw       xmm1, xmm1              ; xmm1=2*CrO
+
+    pmulhw      xmm4, GOTOFF(eax,PW_MF0228)  ; xmm4=(2*CbE * -FIX(0.22800))
+    pmulhw      xmm5, GOTOFF(eax,PW_MF0228)  ; xmm5=(2*CbO * -FIX(0.22800))
+    pmulhw      xmm0, GOTOFF(eax,PW_F0402)   ; xmm0=(2*CrE * FIX(0.40200))
+    pmulhw      xmm1, GOTOFF(eax,PW_F0402)   ; xmm1=(2*CrO * FIX(0.40200))
+
+    paddw       xmm4, GOTOFF(eax,PW_ONE)
+    paddw       xmm5, GOTOFF(eax,PW_ONE)
+    psraw       xmm4, 1                 ; xmm4=(CbE * -FIX(0.22800))
+    psraw       xmm5, 1                 ; xmm5=(CbO * -FIX(0.22800))
+    paddw       xmm0, GOTOFF(eax,PW_ONE)
+    paddw       xmm1, GOTOFF(eax,PW_ONE)
+    psraw       xmm0, 1                 ; xmm0=(CrE * FIX(0.40200))
+    psraw       xmm1, 1                 ; xmm1=(CrO * FIX(0.40200))
+
+    paddw       xmm4, xmm2
+    paddw       xmm5, xmm3
+    paddw       xmm4, xmm2              ; xmm4=(CbE * FIX(1.77200))=(B-Y)E
+    paddw       xmm5, xmm3              ; xmm5=(CbO * FIX(1.77200))=(B-Y)O
+    paddw       xmm0, xmm6              ; xmm0=(CrE * FIX(1.40200))=(R-Y)E
+    paddw       xmm1, xmm7              ; xmm1=(CrO * FIX(1.40200))=(R-Y)O
+
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=(B-Y)E
+    movdqa      XMMWORD wk(1), xmm5   ; wk(1)=(B-Y)O
+
+    movdqa      xmm4, xmm2
+    movdqa      xmm5, xmm3
+    punpcklwd   xmm2, xmm6
+    punpckhwd   xmm4, xmm6
+    pmaddwd     xmm2, GOTOFF(eax,PW_MF0344_F0285)
+    pmaddwd     xmm4, GOTOFF(eax,PW_MF0344_F0285)
+    punpcklwd   xmm3, xmm7
+    punpckhwd   xmm5, xmm7
+    pmaddwd     xmm3, GOTOFF(eax,PW_MF0344_F0285)
+    pmaddwd     xmm5, GOTOFF(eax,PW_MF0344_F0285)
+
+    paddd       xmm2, GOTOFF(eax,PD_ONEHALF)
+    paddd       xmm4, GOTOFF(eax,PD_ONEHALF)
+    psrad       xmm2, SCALEBITS
+    psrad       xmm4, SCALEBITS
+    paddd       xmm3, GOTOFF(eax,PD_ONEHALF)
+    paddd       xmm5, GOTOFF(eax,PD_ONEHALF)
+    psrad       xmm3, SCALEBITS
+    psrad       xmm5, SCALEBITS
+
+    packssdw    xmm2, xmm4              ; xmm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+    packssdw    xmm3, xmm5              ; xmm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+    psubw       xmm2, xmm6              ; xmm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+    psubw       xmm3, xmm7              ; xmm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+    movdqa      xmm5, XMMWORD esi     ; xmm5=Y(0123456789ABCDEF)
+
+    pcmpeqw     xmm4, xmm4
+    psrlw       xmm4, BYTE_BIT          ; xmm4={0xFF 0x00 0xFF 0x00 ..}
+    pand        xmm4, xmm5              ; xmm4=Y(02468ACE)=YE
+    psrlw       xmm5, BYTE_BIT          ; xmm5=Y(13579BDF)=YO
+
+    paddw       xmm0, xmm4              ; xmm0=((R-Y)E+YE)=RE=R(02468ACE)
+    paddw       xmm1, xmm5              ; xmm1=((R-Y)O+YO)=RO=R(13579BDF)
+    packuswb    xmm0, xmm0              ; xmm0=R(02468ACE********)
+    packuswb    xmm1, xmm1              ; xmm1=R(13579BDF********)
+
+    paddw       xmm2, xmm4              ; xmm2=((G-Y)E+YE)=GE=G(02468ACE)
+    paddw       xmm3, xmm5              ; xmm3=((G-Y)O+YO)=GO=G(13579BDF)
+    packuswb    xmm2, xmm2              ; xmm2=G(02468ACE********)
+    packuswb    xmm3, xmm3              ; xmm3=G(13579BDF********)
+
+    paddw       xmm4, XMMWORD wk(0)   ; xmm4=(YE+(B-Y)E)=BE=B(02468ACE)
+    paddw       xmm5, XMMWORD wk(1)   ; xmm5=(YO+(B-Y)O)=BO=B(13579BDF)
+    packuswb    xmm4, xmm4              ; xmm4=B(02468ACE********)
+    packuswb    xmm5, xmm5              ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+    ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+    ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+    ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+    punpcklbw   xmmA, xmmC        ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+    punpcklbw   xmmE, xmmB        ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+    punpcklbw   xmmD, xmmF        ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+    movdqa      xmmG, xmmA
+    movdqa      xmmH, xmmA
+    punpcklwd   xmmA, xmmE        ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+    punpckhwd   xmmG, xmmE        ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+    psrldq      xmmH, 2           ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+    psrldq      xmmE, 2           ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+    movdqa      xmmC, xmmD
+    movdqa      xmmB, xmmD
+    punpcklwd   xmmD, xmmH        ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+    punpckhwd   xmmC, xmmH        ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+    psrldq      xmmB, 2           ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+    movdqa      xmmF, xmmE
+    punpcklwd   xmmE, xmmB        ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+    punpckhwd   xmmF, xmmB        ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+    pshufd      xmmH, xmmA, 0x4E  ; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+    movdqa      xmmB, xmmE
+    punpckldq   xmmA, xmmD        ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+    punpckldq   xmmE, xmmH        ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+    punpckhdq   xmmD, xmmB        ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+    pshufd      xmmH, xmmG, 0x4E  ; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+    movdqa      xmmB, xmmF
+    punpckldq   xmmG, xmmC        ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+    punpckldq   xmmF, xmmH        ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+    punpckhdq   xmmC, xmmB        ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+    punpcklqdq  xmmA, xmmE        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+    punpcklqdq  xmmD, xmmG        ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    punpcklqdq  xmmF, xmmC        ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jb          short .column_st32
+
+    test        edi, SIZEOF_XMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    movntdq     XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movntdq     XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    movntdq     XMMWORD edi+2*SIZEOF_XMMWORD, xmmF
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    movdqu      XMMWORD edi+2*SIZEOF_XMMWORD, xmmF
+.out0:
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+    sub         ecx, byte SIZEOF_XMMWORD
+    jz          near .nextrow
+
+    add         esi, byte SIZEOF_XMMWORD  ; inptr0
+    add         ebx, byte SIZEOF_XMMWORD  ; inptr1
+    add         edx, byte SIZEOF_XMMWORD  ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st32:
+    lea         ecx, ecx+ecx*2        ; imul ecx, RGB_PIXELSIZE
+    cmp         ecx, byte 2*SIZEOF_XMMWORD
+    jb          short .column_st16
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    add         edi, byte 2*SIZEOF_XMMWORD  ; outptr
+    movdqa      xmmA, xmmF
+    sub         ecx, byte 2*SIZEOF_XMMWORD
+    jmp         short .column_st15
+.column_st16:
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jb          short .column_st15
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    add         edi, byte SIZEOF_XMMWORD    ; outptr
+    movdqa      xmmA, xmmD
+    sub         ecx, byte SIZEOF_XMMWORD
+.column_st15:
+    ; Store the lower 8 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st7
+    movq        XMM_MMWORD edi, xmmA
+    add         edi, byte SIZEOF_MMWORD
+    sub         ecx, byte SIZEOF_MMWORD
+    psrldq      xmmA, SIZEOF_MMWORD
+.column_st7:
+    ; Store the lower 4 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_DWORD
+    jb          short .column_st3
+    movd        XMM_DWORD edi, xmmA
+    add         edi, byte SIZEOF_DWORD
+    sub         ecx, byte SIZEOF_DWORD
+    psrldq      xmmA, SIZEOF_DWORD
+.column_st3:
+    ; Store the lower 2 bytes of eax to the output when it has enough
+    ; space.
+    movd        eax, xmmA
+    cmp         ecx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word edi, ax
+    add         edi, byte SIZEOF_WORD
+    sub         ecx, byte SIZEOF_WORD
+    shr         eax, 16
+.column_st1:
+    ; Store the lower 1 byte of eax to the output when it has enough
+    ; space.
+    test        ecx, ecx
+    jz          short .nextrow
+    mov         byte edi, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    pcmpeqb     xmm6, xmm6              ; xmm6=XE=X(02468ACE********)
+    pcmpeqb     xmm7, xmm7              ; xmm7=XO=X(13579BDF********)
+%else
+    pxor        xmm6, xmm6              ; xmm6=XE=X(02468ACE********)
+    pxor        xmm7, xmm7              ; xmm7=XO=X(13579BDF********)
+%endif
+    ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+    ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+    ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+    ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+    punpcklbw   xmmA, xmmC  ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+    punpcklbw   xmmE, xmmG  ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+    punpcklbw   xmmB, xmmD  ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+    punpcklbw   xmmF, xmmH  ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+    movdqa      xmmC, xmmA
+    punpcklwd   xmmA, xmmE  ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+    punpckhwd   xmmC, xmmE  ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+    movdqa      xmmG, xmmB
+    punpcklwd   xmmB, xmmF  ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+    punpckhwd   xmmG, xmmF  ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+    movdqa      xmmD, xmmA
+    punpckldq   xmmA, xmmB  ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+    punpckhdq   xmmD, xmmB  ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    movdqa      xmmH, xmmC
+    punpckldq   xmmC, xmmG  ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+    punpckhdq   xmmH, xmmG  ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jb          short .column_st32
+
+    test        edi, SIZEOF_XMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    movntdq     XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movntdq     XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    movntdq     XMMWORD edi+2*SIZEOF_XMMWORD, xmmC
+    movntdq     XMMWORD edi+3*SIZEOF_XMMWORD, xmmH
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    movdqu      XMMWORD edi+2*SIZEOF_XMMWORD, xmmC
+    movdqu      XMMWORD edi+3*SIZEOF_XMMWORD, xmmH
+.out0:
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+    sub         ecx, byte SIZEOF_XMMWORD
+    jz          near .nextrow
+
+    add         esi, byte SIZEOF_XMMWORD  ; inptr0
+    add         ebx, byte SIZEOF_XMMWORD  ; inptr1
+    add         edx, byte SIZEOF_XMMWORD  ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st32:
+    cmp         ecx, byte SIZEOF_XMMWORD/2
+    jb          short .column_st16
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    add         edi, byte 2*SIZEOF_XMMWORD  ; outptr
+    movdqa      xmmA, xmmC
+    movdqa      xmmD, xmmH
+    sub         ecx, byte SIZEOF_XMMWORD/2
+.column_st16:
+    cmp         ecx, byte SIZEOF_XMMWORD/4
+    jb          short .column_st15
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    add         edi, byte SIZEOF_XMMWORD    ; outptr
+    movdqa      xmmA, xmmD
+    sub         ecx, byte SIZEOF_XMMWORD/4
+.column_st15:
+    ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_XMMWORD/8
+    jb          short .column_st7
+    movq        XMM_MMWORD edi, xmmA
+    add         edi, byte SIZEOF_XMMWORD/8*4
+    sub         ecx, byte SIZEOF_XMMWORD/8
+    psrldq      xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+    ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+    ; space.
+    test        ecx, ecx
+    jz          short .nextrow
+    movd        XMM_DWORD edi, xmmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    alignx      16, 7
+
+.nextrow:
+    pop         ecx
+    pop         esi
+    pop         ebx
+    pop         edx
+    pop         edi
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW
+    add         ebx, byte SIZEOF_JSAMPROW
+    add         edx, byte SIZEOF_JSAMPROW
+    add         edi, byte SIZEOF_JSAMPROW  ; output_buf
+    dec         eax                        ; num_rows
+    jg          near .rowloop
+
+    sfence                              ; flush the write buffer
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdcolor-avx2.asm Added

@@ -0,0 +1,118 @@
+;
+; jdcolor.asm - colorspace conversion (AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_ycc_rgb_convert_avx2)
+
+EXTN(jconst_ycc_rgb_convert_avx2):
+
+PW_F0402        times 16 dw  F_0_402
+PW_MF0228       times 16 dw -F_0_228
+PW_MF0344_F0285 times 8  dw -F_0_344, F_0_285
+PW_ONE          times 16 dw  1
+PD_ONEHALF      times 8  dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extrgb_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extrgbx_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extbgr_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extbgrx_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extxbgr_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extxrgb_convert_avx2
+%include "jdcolext-avx2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdcolor-mmx.asm Added

@@ -0,0 +1,117 @@
+;
+; jdcolor.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_ycc_rgb_convert_mmx)
+
+EXTN(jconst_ycc_rgb_convert_mmx):
+
+PW_F0402        times 4 dw  F_0_402
+PW_MF0228       times 4 dw -F_0_228
+PW_MF0344_F0285 times 2 dw -F_0_344, F_0_285
+PW_ONE          times 4 dw  1
+PD_ONEHALF      times 2 dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx  jsimd_ycc_extrgb_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx  jsimd_ycc_extrgbx_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx  jsimd_ycc_extbgr_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx  jsimd_ycc_extbgrx_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx  jsimd_ycc_extxbgr_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx  jsimd_ycc_extxrgb_convert_mmx
+%include "jdcolext-mmx.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdcolor-sse2.asm Added

@@ -0,0 +1,117 @@
+;
+; jdcolor.asm - colorspace conversion (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_ycc_rgb_convert_sse2)
+
+EXTN(jconst_ycc_rgb_convert_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extrgb_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extrgbx_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extbgr_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extbgrx_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extxbgr_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extxrgb_convert_sse2
+%include "jdcolext-sse2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdmerge-avx2.asm Added

@@ -0,0 +1,136 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_merged_upsample_avx2)
+
+EXTN(jconst_merged_upsample_avx2):
+
+PW_F0402        times 16 dw  F_0_402
+PW_MF0228       times 16 dw -F_0_228
+PW_MF0344_F0285 times 8  dw -F_0_344, F_0_285
+PW_ONE          times 16 dw  1
+PD_ONEHALF      times 8  dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extrgb_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extrgb_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extrgbx_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extrgbx_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extbgr_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extbgr_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extbgrx_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extbgrx_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extxbgr_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extxbgr_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extxrgb_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extxrgb_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdmerge-mmx.asm Added

@@ -0,0 +1,123 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_merged_upsample_mmx)
+
+EXTN(jconst_merged_upsample_mmx):
+
+PW_F0402        times 4 dw  F_0_402
+PW_MF0228       times 4 dw -F_0_228
+PW_MF0344_F0285 times 2 dw -F_0_344, F_0_285
+PW_ONE          times 4 dw  1
+PD_ONEHALF      times 2 dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx  jsimd_h2v1_extrgb_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx  jsimd_h2v2_extrgb_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx  jsimd_h2v1_extrgbx_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx  jsimd_h2v2_extrgbx_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx  jsimd_h2v1_extbgr_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx  jsimd_h2v2_extbgr_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx  jsimd_h2v1_extbgrx_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx  jsimd_h2v2_extbgrx_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx  jsimd_h2v1_extxbgr_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx  jsimd_h2v2_extxbgr_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx  jsimd_h2v1_extxrgb_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx  jsimd_h2v2_extxrgb_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdmerge-sse2.asm Added

@@ -0,0 +1,135 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_merged_upsample_sse2)
+
+EXTN(jconst_merged_upsample_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extrgb_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extrgbx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extrgbx_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extbgr_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extbgrx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extbgrx_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extxbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extxbgr_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extxrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extxrgb_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdmrgext-avx2.asm Added

@@ -0,0 +1,575 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (AVX2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2012, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_avx2(JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+%define output_width(b)      (b) + 8    ; JDIMENSION output_width
+%define input_buf(b)         (b) + 12   ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)  (b) + 16   ; JDIMENSION in_row_group_ctr
+%define output_buf(b)        (b) + 20   ; JSAMPARRAY output_buf
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_YMMWORD
+                                        ; ymmword wkWK_NUM
+%define WK_NUM        3
+%define gotptr        wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_merged_upsample_avx2)
+
+EXTN(jsimd_h2v1_merged_upsample_avx2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION output_width(eax)  ; col
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         edi, JSAMPIMAGE input_buf(eax)
+    mov         ecx, JDIMENSION in_row_group_ctr(eax)
+    mov         esi, JSAMPARRAY edi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY edi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY edi+2*SIZEOF_JSAMPARRAY
+    mov         edi, JSAMPARRAY output_buf(eax)
+    mov         esi, JSAMPROW esi+ecx*SIZEOF_JSAMPROW  ; inptr0
+    mov         ebx, JSAMPROW ebx+ecx*SIZEOF_JSAMPROW  ; inptr1
+    mov         edx, JSAMPROW edx+ecx*SIZEOF_JSAMPROW  ; inptr2
+    mov         edi, JSAMPROW edi                      ; outptr
+
+    pop         ecx                     ; col
+
+    alignx      16, 7
+.columnloop:
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+
+    vmovdqu     ymm6, YMMWORD ebx     ; ymm6=Cb(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+    vmovdqu     ymm7, YMMWORD edx     ; ymm7=Cr(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+
+    vpxor       ymm1, ymm1, ymm1        ; ymm1=(all 0's)
+    vpcmpeqw    ymm3, ymm3, ymm3
+    vpsllw      ymm3, ymm3, 7           ; ymm3={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    vpermq      ymm6, ymm6, 0xd8        ; ymm6=Cb(01234567GHIJKLMN89ABCDEFOPQRSTUV)
+    vpermq      ymm7, ymm7, 0xd8        ; ymm7=Cr(01234567GHIJKLMN89ABCDEFOPQRSTUV)
+    vpunpcklbw  ymm4, ymm6, ymm1        ; ymm4=Cb(0123456789ABCDEF)=CbL
+    vpunpckhbw  ymm6, ymm6, ymm1        ; ymm6=Cb(GHIJKLMNOPQRSTUV)=CbH
+    vpunpcklbw  ymm0, ymm7, ymm1        ; ymm0=Cr(0123456789ABCDEF)=CrL
+    vpunpckhbw  ymm7, ymm7, ymm1        ; ymm7=Cr(GHIJKLMNOPQRSTUV)=CrH
+
+    vpaddw      ymm5, ymm6, ymm3
+    vpaddw      ymm2, ymm4, ymm3
+    vpaddw      ymm1, ymm7, ymm3
+    vpaddw      ymm3, ymm0, ymm3
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    vpaddw      ymm6, ymm5, ymm5             ; ymm6=2*CbH
+    vpaddw      ymm4, ymm2, ymm2             ; ymm4=2*CbL
+    vpaddw      ymm7, ymm1, ymm1             ; ymm7=2*CrH
+    vpaddw      ymm0, ymm3, ymm3             ; ymm0=2*CrL
+
+    vpmulhw     ymm6, ymm6, GOTOFF(eax,PW_MF0228)  ; ymm6=(2*CbH * -FIX(0.22800))
+    vpmulhw     ymm4, ymm4, GOTOFF(eax,PW_MF0228)  ; ymm4=(2*CbL * -FIX(0.22800))
+    vpmulhw     ymm7, ymm7, GOTOFF(eax,PW_F0402)   ; ymm7=(2*CrH * FIX(0.40200))
+    vpmulhw     ymm0, ymm0, GOTOFF(eax,PW_F0402)   ; ymm0=(2*CrL * FIX(0.40200))
+
+    vpaddw      ymm6, ymm6, GOTOFF(eax,PW_ONE)
+    vpaddw      ymm4, ymm4, GOTOFF(eax,PW_ONE)
+    vpsraw      ymm6, ymm6, 1                     ; ymm6=(CbH * -FIX(0.22800))
+    vpsraw      ymm4, ymm4, 1                     ; ymm4=(CbL * -FIX(0.22800))
+    vpaddw      ymm7, ymm7, GOTOFF(eax,PW_ONE)
+    vpaddw      ymm0, ymm0, GOTOFF(eax,PW_ONE)
+    vpsraw      ymm7, ymm7, 1                     ; ymm7=(CrH * FIX(0.40200))
+    vpsraw      ymm0, ymm0, 1                     ; ymm0=(CrL * FIX(0.40200))
+
+    vpaddw      ymm6, ymm6, ymm5
+    vpaddw      ymm4, ymm4, ymm2
+    vpaddw      ymm6, ymm6, ymm5                  ; ymm6=(CbH * FIX(1.77200))=(B-Y)H
+    vpaddw      ymm4, ymm4, ymm2                  ; ymm4=(CbL * FIX(1.77200))=(B-Y)L
+    vpaddw      ymm7, ymm7, ymm1                  ; ymm7=(CrH * FIX(1.40200))=(R-Y)H
+    vpaddw      ymm0, ymm0, ymm3                  ; ymm0=(CrL * FIX(1.40200))=(R-Y)L
+
+    vmovdqa     YMMWORD wk(0), ymm6             ; wk(0)=(B-Y)H
+    vmovdqa     YMMWORD wk(1), ymm7             ; wk(1)=(R-Y)H
+
+    vpunpckhwd  ymm6, ymm5, ymm1
+    vpunpcklwd  ymm5, ymm5, ymm1
+    vpmaddwd    ymm5, ymm5, GOTOFF(eax,PW_MF0344_F0285)
+    vpmaddwd    ymm6, ymm6, GOTOFF(eax,PW_MF0344_F0285)
+    vpunpckhwd  ymm7, ymm2, ymm3
+    vpunpcklwd  ymm2, ymm2, ymm3
+    vpmaddwd    ymm2, ymm2, GOTOFF(eax,PW_MF0344_F0285)
+    vpmaddwd    ymm7, ymm7, GOTOFF(eax,PW_MF0344_F0285)
+
+    vpaddd      ymm5, ymm5, GOTOFF(eax,PD_ONEHALF)
+    vpaddd      ymm6, ymm6, GOTOFF(eax,PD_ONEHALF)
+    vpsrad      ymm5, ymm5, SCALEBITS
+    vpsrad      ymm6, ymm6, SCALEBITS
+    vpaddd      ymm2, ymm2, GOTOFF(eax,PD_ONEHALF)
+    vpaddd      ymm7, ymm7, GOTOFF(eax,PD_ONEHALF)
+    vpsrad      ymm2, ymm2, SCALEBITS
+    vpsrad      ymm7, ymm7, SCALEBITS
+
+    vpackssdw   ymm5, ymm5, ymm6        ; ymm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+    vpackssdw   ymm2, ymm2, ymm7        ; ymm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+    vpsubw      ymm5, ymm5, ymm1        ; ymm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+    vpsubw      ymm2, ymm2, ymm3        ; ymm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+    vmovdqa     YMMWORD wk(2), ymm5   ; wk(2)=(G-Y)H
+
+    mov         al, 2                   ; Yctr
+    jmp         short .Yloop_1st
+    alignx      16, 7
+
+.Yloop_2nd:
+    vmovdqa     ymm0, YMMWORD wk(1)   ; ymm0=(R-Y)H
+    vmovdqa     ymm2, YMMWORD wk(2)   ; ymm2=(G-Y)H
+    vmovdqa     ymm4, YMMWORD wk(0)   ; ymm4=(B-Y)H
+    alignx      16, 7
+
+.Yloop_1st:
+    vmovdqu     ymm7, YMMWORD esi     ; ymm7=Y(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+
+    vpcmpeqw    ymm6, ymm6, ymm6
+    vpsrlw      ymm6, ymm6, BYTE_BIT    ; ymm6={0xFF 0x00 0xFF 0x00 ..}
+    vpand       ymm6, ymm6, ymm7        ; ymm6=Y(02468ACEGIKMOQSU)=YE
+    vpsrlw      ymm7, ymm7, BYTE_BIT    ; ymm7=Y(13579BDFHJLNPRTV)=YO
+
+    vmovdqa     ymm1, ymm0              ; ymm1=ymm0=(R-Y)(L/H)
+    vmovdqa     ymm3, ymm2              ; ymm3=ymm2=(G-Y)(L/H)
+    vmovdqa     ymm5, ymm4              ; ymm5=ymm4=(B-Y)(L/H)
+
+    vpaddw      ymm0, ymm0, ymm6        ; ymm0=((R-Y)+YE)=RE=R(02468ACEGIKMOQSU)
+    vpaddw      ymm1, ymm1, ymm7        ; ymm1=((R-Y)+YO)=RO=R(13579BDFHJLNPRTV)
+    vpackuswb   ymm0, ymm0, ymm0        ; ymm0=R(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm1, ymm1, ymm1        ; ymm1=R(13579BDF********HJLNPRTV********)
+
+    vpaddw      ymm2, ymm2, ymm6        ; ymm2=((G-Y)+YE)=GE=G(02468ACEGIKMOQSU)
+    vpaddw      ymm3, ymm3, ymm7        ; ymm3=((G-Y)+YO)=GO=G(13579BDFHJLNPRTV)
+    vpackuswb   ymm2, ymm2, ymm2        ; ymm2=G(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm3, ymm3, ymm3        ; ymm3=G(13579BDF********HJLNPRTV********)
+
+    vpaddw      ymm4, ymm4, ymm6        ; ymm4=((B-Y)+YE)=BE=B(02468ACEGIKMOQSU)
+    vpaddw      ymm5, ymm5, ymm7        ; ymm5=((B-Y)+YO)=BO=B(13579BDFHJLNPRTV)
+    vpackuswb   ymm4, ymm4, ymm4        ; ymm4=B(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm5, ymm5, ymm5        ; ymm5=B(13579BDF********HJLNPRTV********)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; ymmA=(00 02 04 06 08 0A 0C 0E ** 0G 0I 0K 0M 0O 0Q 0S 0U **)
+    ; ymmB=(01 03 05 07 09 0B 0D 0F ** 0H 0J 0L 0N 0P 0R 0T 0V **)
+    ; ymmC=(10 12 14 16 18 1A 1C 1E ** 1G 1I 1K 1M 1O 1Q 1S 1U **)
+    ; ymmD=(11 13 15 17 19 1B 1D 1F ** 1H 1J 1L 1N 1P 1R 1T 1V **)
+    ; ymmE=(20 22 24 26 28 2A 2C 2E ** 2G 2I 2K 2M 2O 2Q 2S 2U **)
+    ; ymmF=(21 23 25 27 29 2B 2D 2F ** 2H 2J 2L 2N 2P 2R 2T 2V **)
+    ; ymmG=(** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** **)
+    ; ymmH=(** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** **)
+
+    vpunpcklbw  ymmA, ymmA, ymmC        ; ymmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E
+                                        ;       0G 1G 0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U)
+    vpunpcklbw  ymmE, ymmE, ymmB        ; ymmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F
+                                        ;       2G 0H 2I 0J 2K 0L 2M 0N 2O 0P 2Q 0R 2S 0T 2U 0V)
+    vpunpcklbw  ymmD, ymmD, ymmF        ; ymmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F
+                                        ;       1H 2H 1J 2J 1L 2L 1N 2N 1P 2P 1R 2R 1T 2T 1V 2V)
+
+    vpsrldq     ymmH, ymmA, 2           ; ymmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E 0G 1G
+                                        ;       0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U -- --)
+    vpunpckhwd  ymmG, ymmA, ymmE        ; ymmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F
+                                        ;       0O 1O 2O 0P 0Q 1Q 2Q 0R 0S 1S 2S 0T 0U 1U 2U 0V)
+    vpunpcklwd  ymmA, ymmA, ymmE        ; ymmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07
+                                        ;       0G 1G 2G 0H 0I 1I 2I 0J 0K 1K 2K 0L 0M 1M 2M 0N)
+
+    vpsrldq     ymmE, ymmE, 2           ; ymmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F 2G 0H
+                                        ;       2I 0J 2K 0L 2M 0N 2O 0P 2Q 0R 2S 0T 2U 0V -- --)
+
+    vpsrldq     ymmB, ymmD, 2           ; ymmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F 1H 2H
+                                        ;       1J 2J 1L 2L 1N 2N 1P 2P 1R 2R 1T 2T 1V 2V -- --)
+    vpunpckhwd  ymmC, ymmD, ymmH        ; ymmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F 0G 1G
+                                        ;       1P 2P 0Q 1Q 1R 2R 0S 1S 1T 2T 0U 1U 1V 2V -- --)
+    vpunpcklwd  ymmD, ymmD, ymmH        ; ymmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18
+                                        ;       1H 2H 0I 1I 1J 2J 0K 1K 1L 2L 0M 1M 1N 2N 0O 1O)
+
+    vpunpckhwd  ymmF, ymmE, ymmB        ; ymmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F 2G 0H 1H 2H
+                                        ;       2Q 0R 1R 2R 2S 0T 1T 2T 2U 0V 1V 2V -- -- -- --)
+    vpunpcklwd  ymmE, ymmE, ymmB        ; ymmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29
+                                        ;       2I 0J 1J 2J 2K 0L 1L 2L 2M 0N 1N 2N 2O 0P 1P 2P)
+
+    vpshufd     ymmH, ymmA, 0x4E        ; ymmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03
+                                        ;       0K 1K 2K 0L 0M 1M 2M 0N 0G 1G 2G 0H 0I 1I 2I 0J)
+    vpunpckldq  ymmA, ymmA, ymmD        ; ymmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 0I 1I 2I 0J 1J 2J 0K 1K)
+    vpunpckhdq  ymmD, ymmD, ymmE        ; ymmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29
+                                        ;       1L 2L 0M 1M 2M 0N 1N 2N 1N 2N 0O 1O 2O 0P 1P 2P)
+    vpunpckldq  ymmE, ymmE, ymmH        ; ymmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07
+                                        ;       2I 0J 1J 2J 0K 1K 2K 0L 2K 0L 1L 2L 0M 1M 2M 0N)
+
+    vpshufd     ymmH, ymmG, 0x4E        ; ymmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B
+                                        ;       0S 1S 2S 0T 0U 1U 2U 0V 0O 1O 2O 0P 0Q 1Q 2Q 0R)
+    vpunpckldq  ymmG, ymmG, ymmC        ; ymmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C
+                                        ;       0O 1O 2O 0P 1P 2P 0Q 1Q 0Q 1Q 2Q 0R 1R 2R 0S 1S)
+    vpunpckhdq  ymmC, ymmC, ymmF        ; ymmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F 0G 1G 2G 0H 1H 2H
+                                        ;       1T 2T 0U 1U 2U 0V 1V 2V 1V 2V -- -- -- -- -- --)
+    vpunpckldq  ymmF, ymmF, ymmH        ; ymmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 2S 0T 1T 2T 0U 1U 2U 0V)
+
+    vpunpcklqdq ymmH, ymmA, ymmE        ; ymmH=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vpunpcklqdq ymmG, ymmD, ymmG        ; ymmG=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A
+                                        ;       1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q)
+    vpunpcklqdq ymmC, ymmF, ymmC        ; ymmC=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    vperm2i128  ymmA, ymmH, ymmG, 0x20  ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                        ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    vperm2i128  ymmD, ymmC, ymmH, 0x30  ; ymmD=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vperm2i128  ymmF, ymmG, ymmC, 0x31  ; ymmF=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jb          short .column_st64
+
+    test        edi, SIZEOF_YMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    vmovntdq    YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovntdq    YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    vmovntdq    YMMWORD edi+2*SIZEOF_YMMWORD, ymmF
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    vmovdqu     YMMWORD edi+2*SIZEOF_YMMWORD, ymmF
+.out0:
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_YMMWORD  ; outptr
+    sub         ecx, byte SIZEOF_YMMWORD
+    jz          near .endcolumn
+
+    add         esi, byte SIZEOF_YMMWORD  ; inptr0
+    dec         al                        ; Yctr
+    jnz         near .Yloop_2nd
+
+    add         ebx, byte SIZEOF_YMMWORD  ; inptr1
+    add         edx, byte SIZEOF_YMMWORD  ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st64:
+    lea         ecx, ecx+ecx*2            ; imul ecx, RGB_PIXELSIZE
+    cmp         ecx, byte 2*SIZEOF_YMMWORD
+    jb          short .column_st32
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    add         edi, byte 2*SIZEOF_YMMWORD  ; outptr
+    vmovdqa     ymmA, ymmF
+    sub         ecx, byte 2*SIZEOF_YMMWORD
+    jmp         short .column_st31
+.column_st32:
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jb          short .column_st31
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    add         edi, byte SIZEOF_YMMWORD    ; outptr
+    vmovdqa     ymmA, ymmD
+    sub         ecx, byte SIZEOF_YMMWORD
+    jmp         short .column_st31
+.column_st31:
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jb          short .column_st15
+    vmovdqu     XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    add         edi, byte SIZEOF_XMMWORD    ; outptr
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    sub         ecx, byte SIZEOF_XMMWORD
+.column_st15:
+    ; Store the lower 8 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st7
+    vmovq       XMM_MMWORD edi, xmmA
+    add         edi, byte SIZEOF_MMWORD
+    sub         ecx, byte SIZEOF_MMWORD
+    vpsrldq     xmmA, xmmA, SIZEOF_MMWORD
+.column_st7:
+    ; Store the lower 4 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_DWORD
+    jb          short .column_st3
+    vmovd       XMM_DWORD edi, xmmA
+    add         edi, byte SIZEOF_DWORD
+    sub         ecx, byte SIZEOF_DWORD
+    vpsrldq     xmmA, xmmA, SIZEOF_DWORD
+.column_st3:
+    ; Store the lower 2 bytes of eax to the output when it has enough
+    ; space.
+    vmovd       eax, xmmA
+    cmp         ecx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word edi, ax
+    add         edi, byte SIZEOF_WORD
+    sub         ecx, byte SIZEOF_WORD
+    shr         eax, 16
+.column_st1:
+    ; Store the lower 1 byte of eax to the output when it has enough
+    ; space.
+    test        ecx, ecx
+    jz          short .endcolumn
+    mov         byte edi, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    vpcmpeqb    ymm6, ymm6, ymm6        ; ymm6=XE=X(02468ACE********GIKMOQSU********)
+    vpcmpeqb    ymm7, ymm7, ymm7        ; ymm7=XO=X(13579BDF********HJLNPRTV********)
+%else
+    vpxor       ymm6, ymm6, ymm6        ; ymm6=XE=X(02468ACE********GIKMOQSU********)
+    vpxor       ymm7, ymm7, ymm7        ; ymm7=XO=X(13579BDF********HJLNPRTV********)
+%endif
+    ; ymmA=(00 02 04 06 08 0A 0C 0E ** 0G 0I 0K 0M 0O 0Q 0S 0U **)
+    ; ymmB=(01 03 05 07 09 0B 0D 0F ** 0H 0J 0L 0N 0P 0R 0T 0V **)
+    ; ymmC=(10 12 14 16 18 1A 1C 1E ** 1G 1I 1K 1M 1O 1Q 1S 1U **)
+    ; ymmD=(11 13 15 17 19 1B 1D 1F ** 1H 1J 1L 1N 1P 1R 1T 1V **)
+    ; ymmE=(20 22 24 26 28 2A 2C 2E ** 2G 2I 2K 2M 2O 2Q 2S 2U **)
+    ; ymmF=(21 23 25 27 29 2B 2D 2F ** 2H 2J 2L 2N 2P 2R 2T 2V **)
+    ; ymmG=(30 32 34 36 38 3A 3C 3E ** 3G 3I 3K 3M 3O 3Q 3S 3U **)
+    ; ymmH=(31 33 35 37 39 3B 3D 3F ** 3H 3J 3L 3N 3P 3R 3T 3V **)
+
+    vpunpcklbw  ymmA, ymmA, ymmC        ; ymmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E
+                                        ;       0G 1G 0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U)
+    vpunpcklbw  ymmE, ymmE, ymmG        ; ymmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E
+                                        ;       2G 3G 2I 3I 2K 3K 2M 3M 2O 3O 2Q 3Q 2S 3S 2U 3U)
+    vpunpcklbw  ymmB, ymmB, ymmD        ; ymmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F
+                                        ;       0H 1H 0J 1J 0L 1L 0N 1N 0P 1P 0R 1R 0T 1T 0V 1V)
+    vpunpcklbw  ymmF, ymmF, ymmH        ; ymmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F
+                                        ;       2H 3H 2J 3J 2L 3L 2N 3N 2P 3P 2R 3R 2T 3T 2V 3V)
+
+    vpunpckhwd  ymmC, ymmA, ymmE        ; ymmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E
+                                        ;       0O 1O 2O 3O 0Q 1Q 2Q 3Q 0S 1S 2S 3S 0U 1U 2U 3U)
+    vpunpcklwd  ymmA, ymmA, ymmE        ; ymmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36
+                                        ;       0G 1G 2G 3G 0I 1I 2I 3I 0K 1K 2K 3K 0M 1M 2M 3M)
+    vpunpckhwd  ymmG, ymmB, ymmF        ; ymmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F
+                                        ;       0P 1P 2P 3P 0R 1R 2R 3R 0T 1T 2T 3T 0V 1V 2V 3V)
+    vpunpcklwd  ymmB, ymmB, ymmF        ; ymmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37
+                                        ;       0H 1H 2H 3H 0J 1J 2J 3J 0L 1L 2L 3L 0N 1N 2N 3N)
+
+    vpunpckhdq  ymmE, ymmA, ymmB        ; ymmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    vpunpckldq  ymmB, ymmA, ymmB        ; ymmB=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J)
+    vpunpckhdq  ymmF, ymmC, ymmG        ; ymmF=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+    vpunpckldq  ymmG, ymmC, ymmG        ; ymmG=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R)
+
+    vperm2i128  ymmA, ymmB, ymmE, 0x20  ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    vperm2i128  ymmD, ymmG, ymmF, 0x20  ; ymmD=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+    vperm2i128  ymmC, ymmB, ymmE, 0x31  ; ymmC=(0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    vperm2i128  ymmH, ymmG, ymmF, 0x31  ; ymmH=(0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    cmp         ecx, byte SIZEOF_YMMWORD
+    jb          short .column_st64
+
+    test        edi, SIZEOF_YMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    vmovntdq    YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovntdq    YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    vmovntdq    YMMWORD edi+2*SIZEOF_YMMWORD, ymmC
+    vmovntdq    YMMWORD edi+3*SIZEOF_YMMWORD, ymmH
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    vmovdqu     YMMWORD edi+2*SIZEOF_YMMWORD, ymmC
+    vmovdqu     YMMWORD edi+3*SIZEOF_YMMWORD, ymmH
+.out0:
+    add         edi, RGB_PIXELSIZE*SIZEOF_YMMWORD  ; outptr
+    sub         ecx, byte SIZEOF_YMMWORD
+    jz          near .endcolumn
+
+    add         esi, byte SIZEOF_YMMWORD  ; inptr0
+    dec         al
+    jnz         near .Yloop_2nd
+
+    add         ebx, byte SIZEOF_YMMWORD  ; inptr1
+    add         edx, byte SIZEOF_YMMWORD  ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st64:
+    cmp         ecx, byte SIZEOF_YMMWORD/2
+    jb          short .column_st32
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymmD
+    add         edi, byte 2*SIZEOF_YMMWORD  ; outptr
+    vmovdqa     ymmA, ymmC
+    vmovdqa     ymmD, ymmH
+    sub         ecx, byte SIZEOF_YMMWORD/2
+.column_st32:
+    cmp         ecx, byte SIZEOF_YMMWORD/4
+    jb          short .column_st16
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymmA
+    add         edi, byte SIZEOF_YMMWORD    ; outptr
+    vmovdqa     ymmA, ymmD
+    sub         ecx, byte SIZEOF_YMMWORD/4
+.column_st16:
+    cmp         ecx, byte SIZEOF_YMMWORD/8
+    jb          short .column_st15
+    vmovdqu     XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    add         edi, byte SIZEOF_XMMWORD    ; outptr
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    sub         ecx, byte SIZEOF_YMMWORD/8
+.column_st15:
+    ; Store two pixels (8 bytes) of ymmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_YMMWORD/16
+    jb          short .column_st7
+    vmovq       MMWORD edi, xmmA
+    add         edi, byte SIZEOF_YMMWORD/16*4
+    sub         ecx, byte SIZEOF_YMMWORD/16
+    vpsrldq     xmmA, SIZEOF_YMMWORD/16*4
+.column_st7:
+    ; Store one pixel (4 bytes) of ymmA to the output when it has enough
+    ; space.
+    test        ecx, ecx
+    jz          short .endcolumn
+    vmovd       XMM_DWORD edi, xmmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+    sfence                              ; flush the write buffer
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_avx2(JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+%define output_width(b)      (b) + 8    ; JDIMENSION output_width
+%define input_buf(b)         (b) + 12   ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)  (b) + 16   ; JDIMENSION in_row_group_ctr
+%define output_buf(b)        (b) + 20   ; JSAMPARRAY output_buf
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_merged_upsample_avx2)
+
+EXTN(jsimd_h2v2_merged_upsample_avx2):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         eax, POINTER output_width(ebp)
+
+    mov         edi, JSAMPIMAGE input_buf(ebp)
+    mov         ecx, JDIMENSION in_row_group_ctr(ebp)
+    mov         esi, JSAMPARRAY edi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY edi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY edi+2*SIZEOF_JSAMPARRAY
+    mov         edi, JSAMPARRAY output_buf(ebp)
+    lea         esi, esi+ecx*SIZEOF_JSAMPROW
+
+    push        edx                     ; inptr2
+    push        ebx                     ; inptr1
+    push        esi                     ; inptr00
+    mov         ebx, esp
+
+    push        edi                     ; output_buf (outptr0)
+    push        ecx                     ; in_row_group_ctr
+    push        ebx                     ; input_buf
+    push        eax                     ; output_width
+
+    call        near EXTN(jsimd_h2v1_merged_upsample_avx2)
+
+    add         esi, byte SIZEOF_JSAMPROW  ; inptr01
+    add         edi, byte SIZEOF_JSAMPROW  ; outptr1
+    mov         POINTER ebx+0*SIZEOF_POINTER, esi
+    mov         POINTER ebx-1*SIZEOF_POINTER, edi
+
+    call        near EXTN(jsimd_h2v1_merged_upsample_avx2)
+
+    add         esp, byte 7*SIZEOF_DWORD
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdmrgext-mmx.asm Added

@@ -0,0 +1,460 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_mmx(JDIMENSION output_width, JSAMPIMAGE input_buf,
+;                                JDIMENSION in_row_group_ctr,
+;                                JSAMPARRAY output_buf);
+;
+
+%define output_width(b)      (b) + 8    ; JDIMENSION output_width
+%define input_buf(b)         (b) + 12   ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)  (b) + 16   ; JDIMENSION in_row_group_ctr
+%define output_buf(b)        (b) + 20   ; JSAMPARRAY output_buf
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_MMWORD  ; mmword wkWK_NUM
+%define WK_NUM        3
+%define gotptr        wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_merged_upsample_mmx)
+
+EXTN(jsimd_h2v1_merged_upsample_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION output_width(eax)  ; col
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         edi, JSAMPIMAGE input_buf(eax)
+    mov         ecx, JDIMENSION in_row_group_ctr(eax)
+    mov         esi, JSAMPARRAY edi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY edi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY edi+2*SIZEOF_JSAMPARRAY
+    mov         edi, JSAMPARRAY output_buf(eax)
+    mov         esi, JSAMPROW esi+ecx*SIZEOF_JSAMPROW  ; inptr0
+    mov         ebx, JSAMPROW ebx+ecx*SIZEOF_JSAMPROW  ; inptr1
+    mov         edx, JSAMPROW edx+ecx*SIZEOF_JSAMPROW  ; inptr2
+    mov         edi, JSAMPROW edi                      ; outptr
+
+    pop         ecx                     ; col
+
+    alignx      16, 7
+.columnloop:
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+
+    movq        mm6, MMWORD ebx       ; mm6=Cb(01234567)
+    movq        mm7, MMWORD edx       ; mm7=Cr(01234567)
+
+    pxor        mm1, mm1                ; mm1=(all 0's)
+    pcmpeqw     mm3, mm3
+    psllw       mm3, 7                  ; mm3={0xFF80 0xFF80 0xFF80 0xFF80}
+
+    movq        mm4, mm6
+    punpckhbw   mm6, mm1                ; mm6=Cb(4567)=CbH
+    punpcklbw   mm4, mm1                ; mm4=Cb(0123)=CbL
+    movq        mm0, mm7
+    punpckhbw   mm7, mm1                ; mm7=Cr(4567)=CrH
+    punpcklbw   mm0, mm1                ; mm0=Cr(0123)=CrL
+
+    paddw       mm6, mm3
+    paddw       mm4, mm3
+    paddw       mm7, mm3
+    paddw       mm0, mm3
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    movq        mm5, mm6                ; mm5=CbH
+    movq        mm2, mm4                ; mm2=CbL
+    paddw       mm6, mm6                ; mm6=2*CbH
+    paddw       mm4, mm4                ; mm4=2*CbL
+    movq        mm1, mm7                ; mm1=CrH
+    movq        mm3, mm0                ; mm3=CrL
+    paddw       mm7, mm7                ; mm7=2*CrH
+    paddw       mm0, mm0                ; mm0=2*CrL
+
+    pmulhw      mm6, GOTOFF(eax,PW_MF0228)  ; mm6=(2*CbH * -FIX(0.22800))
+    pmulhw      mm4, GOTOFF(eax,PW_MF0228)  ; mm4=(2*CbL * -FIX(0.22800))
+    pmulhw      mm7, GOTOFF(eax,PW_F0402)   ; mm7=(2*CrH * FIX(0.40200))
+    pmulhw      mm0, GOTOFF(eax,PW_F0402)   ; mm0=(2*CrL * FIX(0.40200))
+
+    paddw       mm6, GOTOFF(eax,PW_ONE)
+    paddw       mm4, GOTOFF(eax,PW_ONE)
+    psraw       mm6, 1                  ; mm6=(CbH * -FIX(0.22800))
+    psraw       mm4, 1                  ; mm4=(CbL * -FIX(0.22800))
+    paddw       mm7, GOTOFF(eax,PW_ONE)
+    paddw       mm0, GOTOFF(eax,PW_ONE)
+    psraw       mm7, 1                  ; mm7=(CrH * FIX(0.40200))
+    psraw       mm0, 1                  ; mm0=(CrL * FIX(0.40200))
+
+    paddw       mm6, mm5
+    paddw       mm4, mm2
+    paddw       mm6, mm5                ; mm6=(CbH * FIX(1.77200))=(B-Y)H
+    paddw       mm4, mm2                ; mm4=(CbL * FIX(1.77200))=(B-Y)L
+    paddw       mm7, mm1                ; mm7=(CrH * FIX(1.40200))=(R-Y)H
+    paddw       mm0, mm3                ; mm0=(CrL * FIX(1.40200))=(R-Y)L
+
+    movq        MMWORD wk(0), mm6     ; wk(0)=(B-Y)H
+    movq        MMWORD wk(1), mm7     ; wk(1)=(R-Y)H
+
+    movq        mm6, mm5
+    movq        mm7, mm2
+    punpcklwd   mm5, mm1
+    punpckhwd   mm6, mm1
+    pmaddwd     mm5, GOTOFF(eax,PW_MF0344_F0285)
+    pmaddwd     mm6, GOTOFF(eax,PW_MF0344_F0285)
+    punpcklwd   mm2, mm3
+    punpckhwd   mm7, mm3
+    pmaddwd     mm2, GOTOFF(eax,PW_MF0344_F0285)
+    pmaddwd     mm7, GOTOFF(eax,PW_MF0344_F0285)
+
+    paddd       mm5, GOTOFF(eax,PD_ONEHALF)
+    paddd       mm6, GOTOFF(eax,PD_ONEHALF)
+    psrad       mm5, SCALEBITS
+    psrad       mm6, SCALEBITS
+    paddd       mm2, GOTOFF(eax,PD_ONEHALF)
+    paddd       mm7, GOTOFF(eax,PD_ONEHALF)
+    psrad       mm2, SCALEBITS
+    psrad       mm7, SCALEBITS
+
+    packssdw    mm5, mm6                ; mm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+    packssdw    mm2, mm7                ; mm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+    psubw       mm5, mm1                ; mm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+    psubw       mm2, mm3                ; mm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+    movq        MMWORD wk(2), mm5     ; wk(2)=(G-Y)H
+
+    mov         al, 2                   ; Yctr
+    jmp         short .Yloop_1st
+    alignx      16, 7
+
+.Yloop_2nd:
+    movq        mm0, MMWORD wk(1)     ; mm0=(R-Y)H
+    movq        mm2, MMWORD wk(2)     ; mm2=(G-Y)H
+    movq        mm4, MMWORD wk(0)     ; mm4=(B-Y)H
+    alignx      16, 7
+
+.Yloop_1st:
+    movq        mm7, MMWORD esi       ; mm7=Y(01234567)
+
+    pcmpeqw     mm6, mm6
+    psrlw       mm6, BYTE_BIT           ; mm6={0xFF 0x00 0xFF 0x00 ..}
+    pand        mm6, mm7                ; mm6=Y(0246)=YE
+    psrlw       mm7, BYTE_BIT           ; mm7=Y(1357)=YO
+
+    movq        mm1, mm0                ; mm1=mm0=(R-Y)(L/H)
+    movq        mm3, mm2                ; mm3=mm2=(G-Y)(L/H)
+    movq        mm5, mm4                ; mm5=mm4=(B-Y)(L/H)
+
+    paddw       mm0, mm6                ; mm0=((R-Y)+YE)=RE=(R0 R2 R4 R6)
+    paddw       mm1, mm7                ; mm1=((R-Y)+YO)=RO=(R1 R3 R5 R7)
+    packuswb    mm0, mm0                ; mm0=(R0 R2 R4 R6 ** ** ** **)
+    packuswb    mm1, mm1                ; mm1=(R1 R3 R5 R7 ** ** ** **)
+
+    paddw       mm2, mm6                ; mm2=((G-Y)+YE)=GE=(G0 G2 G4 G6)
+    paddw       mm3, mm7                ; mm3=((G-Y)+YO)=GO=(G1 G3 G5 G7)
+    packuswb    mm2, mm2                ; mm2=(G0 G2 G4 G6 ** ** ** **)
+    packuswb    mm3, mm3                ; mm3=(G1 G3 G5 G7 ** ** ** **)
+
+    paddw       mm4, mm6                ; mm4=((B-Y)+YE)=BE=(B0 B2 B4 B6)
+    paddw       mm5, mm7                ; mm5=((B-Y)+YO)=BO=(B1 B3 B5 B7)
+    packuswb    mm4, mm4                ; mm4=(B0 B2 B4 B6 ** ** ** **)
+    packuswb    mm5, mm5                ; mm5=(B1 B3 B5 B7 ** ** ** **)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+    ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+    ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+    ; mmG=(** ** ** ** ** ** ** **), mmH=(** ** ** ** ** ** ** **)
+
+    punpcklbw   mmA, mmC                ; mmA=(00 10 02 12 04 14 06 16)
+    punpcklbw   mmE, mmB                ; mmE=(20 01 22 03 24 05 26 07)
+    punpcklbw   mmD, mmF                ; mmD=(11 21 13 23 15 25 17 27)
+
+    movq        mmG, mmA
+    movq        mmH, mmA
+    punpcklwd   mmA, mmE                ; mmA=(00 10 20 01 02 12 22 03)
+    punpckhwd   mmG, mmE                ; mmG=(04 14 24 05 06 16 26 07)
+
+    psrlq       mmH, 2*BYTE_BIT         ; mmH=(02 12 04 14 06 16 -- --)
+    psrlq       mmE, 2*BYTE_BIT         ; mmE=(22 03 24 05 26 07 -- --)
+
+    movq        mmC, mmD
+    movq        mmB, mmD
+    punpcklwd   mmD, mmH                ; mmD=(11 21 02 12 13 23 04 14)
+    punpckhwd   mmC, mmH                ; mmC=(15 25 06 16 17 27 -- --)
+
+    psrlq       mmB, 2*BYTE_BIT         ; mmB=(13 23 15 25 17 27 -- --)
+
+    movq        mmF, mmE
+    punpcklwd   mmE, mmB                ; mmE=(22 03 13 23 24 05 15 25)
+    punpckhwd   mmF, mmB                ; mmF=(26 07 17 27 -- -- -- --)
+
+    punpckldq   mmA, mmD                ; mmA=(00 10 20 01 11 21 02 12)
+    punpckldq   mmE, mmG                ; mmE=(22 03 13 23 04 14 24 05)
+    punpckldq   mmC, mmF                ; mmC=(15 25 06 16 26 07 17 27)
+
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st16
+
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mmE
+    movq        MMWORD edi+2*SIZEOF_MMWORD, mmC
+
+    sub         ecx, byte SIZEOF_MMWORD
+    jz          near .endcolumn
+
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD  ; outptr
+    add         esi, byte SIZEOF_MMWORD                ; inptr0
+    dec         al                                     ; Yctr
+    jnz         near .Yloop_2nd
+
+    add         ebx, byte SIZEOF_MMWORD                ; inptr1
+    add         edx, byte SIZEOF_MMWORD                ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st16:
+    lea         ecx, ecx+ecx*2        ; imul ecx, RGB_PIXELSIZE
+    cmp         ecx, byte 2*SIZEOF_MMWORD
+    jb          short .column_st8
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mmE
+    movq        mmA, mmC
+    sub         ecx, byte 2*SIZEOF_MMWORD
+    add         edi, byte 2*SIZEOF_MMWORD
+    jmp         short .column_st4
+.column_st8:
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st4
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        mmA, mmE
+    sub         ecx, byte SIZEOF_MMWORD
+    add         edi, byte SIZEOF_MMWORD
+.column_st4:
+    movd        eax, mmA
+    cmp         ecx, byte SIZEOF_DWORD
+    jb          short .column_st2
+    mov         dword edi+0*SIZEOF_DWORD, eax
+    psrlq       mmA, DWORD_BIT
+    movd        eax, mmA
+    sub         ecx, byte SIZEOF_DWORD
+    add         edi, byte SIZEOF_DWORD
+.column_st2:
+    cmp         ecx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word edi+0*SIZEOF_WORD, ax
+    shr         eax, WORD_BIT
+    sub         ecx, byte SIZEOF_WORD
+    add         edi, byte SIZEOF_WORD
+.column_st1:
+    cmp         ecx, byte SIZEOF_BYTE
+    jb          short .endcolumn
+    mov         byte edi+0*SIZEOF_BYTE, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    pcmpeqb     mm6, mm6                ; mm6=(X0 X2 X4 X6 ** ** ** **)
+    pcmpeqb     mm7, mm7                ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%else
+    pxor        mm6, mm6                ; mm6=(X0 X2 X4 X6 ** ** ** **)
+    pxor        mm7, mm7                ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%endif
+    ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+    ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+    ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+    ; mmG=(30 32 34 36 ** ** ** **), mmH=(31 33 35 37 ** ** ** **)
+
+    punpcklbw   mmA, mmC                ; mmA=(00 10 02 12 04 14 06 16)
+    punpcklbw   mmE, mmG                ; mmE=(20 30 22 32 24 34 26 36)
+    punpcklbw   mmB, mmD                ; mmB=(01 11 03 13 05 15 07 17)
+    punpcklbw   mmF, mmH                ; mmF=(21 31 23 33 25 35 27 37)
+
+    movq        mmC, mmA
+    punpcklwd   mmA, mmE                ; mmA=(00 10 20 30 02 12 22 32)
+    punpckhwd   mmC, mmE                ; mmC=(04 14 24 34 06 16 26 36)
+    movq        mmG, mmB
+    punpcklwd   mmB, mmF                ; mmB=(01 11 21 31 03 13 23 33)
+    punpckhwd   mmG, mmF                ; mmG=(05 15 25 35 07 17 27 37)
+
+    movq        mmD, mmA
+    punpckldq   mmA, mmB                ; mmA=(00 10 20 30 01 11 21 31)
+    punpckhdq   mmD, mmB                ; mmD=(02 12 22 32 03 13 23 33)
+    movq        mmH, mmC
+    punpckldq   mmC, mmG                ; mmC=(04 14 24 34 05 15 25 35)
+    punpckhdq   mmH, mmG                ; mmH=(06 16 26 36 07 17 27 37)
+
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st16
+
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mmD
+    movq        MMWORD edi+2*SIZEOF_MMWORD, mmC
+    movq        MMWORD edi+3*SIZEOF_MMWORD, mmH
+
+    sub         ecx, byte SIZEOF_MMWORD
+    jz          short .endcolumn
+
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD  ; outptr
+    add         esi, byte SIZEOF_MMWORD                ; inptr0
+    dec         al                                     ; Yctr
+    jnz         near .Yloop_2nd
+
+    add         ebx, byte SIZEOF_MMWORD                ; inptr1
+    add         edx, byte SIZEOF_MMWORD                ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st16:
+    cmp         ecx, byte SIZEOF_MMWORD/2
+    jb          short .column_st8
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mmD
+    movq        mmA, mmC
+    movq        mmD, mmH
+    sub         ecx, byte SIZEOF_MMWORD/2
+    add         edi, byte 2*SIZEOF_MMWORD
+.column_st8:
+    cmp         ecx, byte SIZEOF_MMWORD/4
+    jb          short .column_st4
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mmA
+    movq        mmA, mmD
+    sub         ecx, byte SIZEOF_MMWORD/4
+    add         edi, byte 1*SIZEOF_MMWORD
+.column_st4:
+    cmp         ecx, byte SIZEOF_MMWORD/8
+    jb          short .endcolumn
+    movd        dword edi+0*SIZEOF_DWORD, mmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_mmx(JDIMENSION output_width, JSAMPIMAGE input_buf,
+;                                JDIMENSION in_row_group_ctr,
+;                                JSAMPARRAY output_buf);
+;
+
+%define output_width(b)      (b) + 8    ; JDIMENSION output_width
+%define input_buf(b)         (b) + 12   ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)  (b) + 16   ; JDIMENSION in_row_group_ctr
+%define output_buf(b)        (b) + 20   ; JSAMPARRAY output_buf
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_merged_upsample_mmx)
+
+EXTN(jsimd_h2v2_merged_upsample_mmx):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         eax, JDIMENSION output_width(ebp)
+
+    mov         edi, JSAMPIMAGE input_buf(ebp)
+    mov         ecx, JDIMENSION in_row_group_ctr(ebp)
+    mov         esi, JSAMPARRAY edi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY edi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY edi+2*SIZEOF_JSAMPARRAY
+    mov         edi, JSAMPARRAY output_buf(ebp)
+    lea         esi, esi+ecx*SIZEOF_JSAMPROW
+
+    push        edx                     ; inptr2
+    push        ebx                     ; inptr1
+    push        esi                     ; inptr00
+    mov         ebx, esp
+
+    push        edi                     ; output_buf (outptr0)
+    push        ecx                     ; in_row_group_ctr
+    push        ebx                     ; input_buf
+    push        eax                     ; output_width
+
+    call        near EXTN(jsimd_h2v1_merged_upsample_mmx)
+
+    add         esi, byte SIZEOF_JSAMPROW  ; inptr01
+    add         edi, byte SIZEOF_JSAMPROW  ; outptr1
+    mov         POINTER ebx+0*SIZEOF_POINTER, esi
+    mov         POINTER ebx-1*SIZEOF_POINTER, edi
+
+    call        near EXTN(jsimd_h2v1_merged_upsample_mmx)
+
+    add         esp, byte 7*SIZEOF_DWORD
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdmrgext-sse2.asm Added

@@ -0,0 +1,517 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2012, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_sse2(JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+%define output_width(b)      (b) + 8    ; JDIMENSION output_width
+%define input_buf(b)         (b) + 12   ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)  (b) + 16   ; JDIMENSION in_row_group_ctr
+%define output_buf(b)        (b) + 20   ; JSAMPARRAY output_buf
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM        3
+%define gotptr        wk(0) - SIZEOF_POINTER  ; void * gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_merged_upsample_sse2)
+
+EXTN(jsimd_h2v1_merged_upsample_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         ecx, JDIMENSION output_width(eax)  ; col
+    test        ecx, ecx
+    jz          near .return
+
+    push        ecx
+
+    mov         edi, JSAMPIMAGE input_buf(eax)
+    mov         ecx, JDIMENSION in_row_group_ctr(eax)
+    mov         esi, JSAMPARRAY edi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY edi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY edi+2*SIZEOF_JSAMPARRAY
+    mov         edi, JSAMPARRAY output_buf(eax)
+    mov         esi, JSAMPROW esi+ecx*SIZEOF_JSAMPROW  ; inptr0
+    mov         ebx, JSAMPROW ebx+ecx*SIZEOF_JSAMPROW  ; inptr1
+    mov         edx, JSAMPROW edx+ecx*SIZEOF_JSAMPROW  ; inptr2
+    mov         edi, JSAMPROW edi                      ; outptr
+
+    pop         ecx                     ; col
+
+    alignx      16, 7
+.columnloop:
+    movpic      eax, POINTER gotptr   ; load GOT address (eax)
+
+    movdqa      xmm6, XMMWORD ebx     ; xmm6=Cb(0123456789ABCDEF)
+    movdqa      xmm7, XMMWORD edx     ; xmm7=Cr(0123456789ABCDEF)
+
+    pxor        xmm1, xmm1              ; xmm1=(all 0's)
+    pcmpeqw     xmm3, xmm3
+    psllw       xmm3, 7                 ; xmm3={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    movdqa      xmm4, xmm6
+    punpckhbw   xmm6, xmm1              ; xmm6=Cb(89ABCDEF)=CbH
+    punpcklbw   xmm4, xmm1              ; xmm4=Cb(01234567)=CbL
+    movdqa      xmm0, xmm7
+    punpckhbw   xmm7, xmm1              ; xmm7=Cr(89ABCDEF)=CrH
+    punpcklbw   xmm0, xmm1              ; xmm0=Cr(01234567)=CrL
+
+    paddw       xmm6, xmm3
+    paddw       xmm4, xmm3
+    paddw       xmm7, xmm3
+    paddw       xmm0, xmm3
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    movdqa      xmm5, xmm6              ; xmm5=CbH
+    movdqa      xmm2, xmm4              ; xmm2=CbL
+    paddw       xmm6, xmm6              ; xmm6=2*CbH
+    paddw       xmm4, xmm4              ; xmm4=2*CbL
+    movdqa      xmm1, xmm7              ; xmm1=CrH
+    movdqa      xmm3, xmm0              ; xmm3=CrL
+    paddw       xmm7, xmm7              ; xmm7=2*CrH
+    paddw       xmm0, xmm0              ; xmm0=2*CrL
+
+    pmulhw      xmm6, GOTOFF(eax,PW_MF0228)  ; xmm6=(2*CbH * -FIX(0.22800))
+    pmulhw      xmm4, GOTOFF(eax,PW_MF0228)  ; xmm4=(2*CbL * -FIX(0.22800))
+    pmulhw      xmm7, GOTOFF(eax,PW_F0402)   ; xmm7=(2*CrH * FIX(0.40200))
+    pmulhw      xmm0, GOTOFF(eax,PW_F0402)   ; xmm0=(2*CrL * FIX(0.40200))
+
+    paddw       xmm6, GOTOFF(eax,PW_ONE)
+    paddw       xmm4, GOTOFF(eax,PW_ONE)
+    psraw       xmm6, 1                 ; xmm6=(CbH * -FIX(0.22800))
+    psraw       xmm4, 1                 ; xmm4=(CbL * -FIX(0.22800))
+    paddw       xmm7, GOTOFF(eax,PW_ONE)
+    paddw       xmm0, GOTOFF(eax,PW_ONE)
+    psraw       xmm7, 1                 ; xmm7=(CrH * FIX(0.40200))
+    psraw       xmm0, 1                 ; xmm0=(CrL * FIX(0.40200))
+
+    paddw       xmm6, xmm5
+    paddw       xmm4, xmm2
+    paddw       xmm6, xmm5              ; xmm6=(CbH * FIX(1.77200))=(B-Y)H
+    paddw       xmm4, xmm2              ; xmm4=(CbL * FIX(1.77200))=(B-Y)L
+    paddw       xmm7, xmm1              ; xmm7=(CrH * FIX(1.40200))=(R-Y)H
+    paddw       xmm0, xmm3              ; xmm0=(CrL * FIX(1.40200))=(R-Y)L
+
+    movdqa      XMMWORD wk(0), xmm6   ; wk(0)=(B-Y)H
+    movdqa      XMMWORD wk(1), xmm7   ; wk(1)=(R-Y)H
+
+    movdqa      xmm6, xmm5
+    movdqa      xmm7, xmm2
+    punpcklwd   xmm5, xmm1
+    punpckhwd   xmm6, xmm1
+    pmaddwd     xmm5, GOTOFF(eax,PW_MF0344_F0285)
+    pmaddwd     xmm6, GOTOFF(eax,PW_MF0344_F0285)
+    punpcklwd   xmm2, xmm3
+    punpckhwd   xmm7, xmm3
+    pmaddwd     xmm2, GOTOFF(eax,PW_MF0344_F0285)
+    pmaddwd     xmm7, GOTOFF(eax,PW_MF0344_F0285)
+
+    paddd       xmm5, GOTOFF(eax,PD_ONEHALF)
+    paddd       xmm6, GOTOFF(eax,PD_ONEHALF)
+    psrad       xmm5, SCALEBITS
+    psrad       xmm6, SCALEBITS
+    paddd       xmm2, GOTOFF(eax,PD_ONEHALF)
+    paddd       xmm7, GOTOFF(eax,PD_ONEHALF)
+    psrad       xmm2, SCALEBITS
+    psrad       xmm7, SCALEBITS
+
+    packssdw    xmm5, xmm6              ; xmm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+    packssdw    xmm2, xmm7              ; xmm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+    psubw       xmm5, xmm1              ; xmm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+    psubw       xmm2, xmm3              ; xmm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+    movdqa      XMMWORD wk(2), xmm5   ; wk(2)=(G-Y)H
+
+    mov         al, 2                   ; Yctr
+    jmp         short .Yloop_1st
+    alignx      16, 7
+
+.Yloop_2nd:
+    movdqa      xmm0, XMMWORD wk(1)   ; xmm0=(R-Y)H
+    movdqa      xmm2, XMMWORD wk(2)   ; xmm2=(G-Y)H
+    movdqa      xmm4, XMMWORD wk(0)   ; xmm4=(B-Y)H
+    alignx      16, 7
+
+.Yloop_1st:
+    movdqa      xmm7, XMMWORD esi     ; xmm7=Y(0123456789ABCDEF)
+
+    pcmpeqw     xmm6, xmm6
+    psrlw       xmm6, BYTE_BIT          ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+    pand        xmm6, xmm7              ; xmm6=Y(02468ACE)=YE
+    psrlw       xmm7, BYTE_BIT          ; xmm7=Y(13579BDF)=YO
+
+    movdqa      xmm1, xmm0              ; xmm1=xmm0=(R-Y)(L/H)
+    movdqa      xmm3, xmm2              ; xmm3=xmm2=(G-Y)(L/H)
+    movdqa      xmm5, xmm4              ; xmm5=xmm4=(B-Y)(L/H)
+
+    paddw       xmm0, xmm6              ; xmm0=((R-Y)+YE)=RE=R(02468ACE)
+    paddw       xmm1, xmm7              ; xmm1=((R-Y)+YO)=RO=R(13579BDF)
+    packuswb    xmm0, xmm0              ; xmm0=R(02468ACE********)
+    packuswb    xmm1, xmm1              ; xmm1=R(13579BDF********)
+
+    paddw       xmm2, xmm6              ; xmm2=((G-Y)+YE)=GE=G(02468ACE)
+    paddw       xmm3, xmm7              ; xmm3=((G-Y)+YO)=GO=G(13579BDF)
+    packuswb    xmm2, xmm2              ; xmm2=G(02468ACE********)
+    packuswb    xmm3, xmm3              ; xmm3=G(13579BDF********)
+
+    paddw       xmm4, xmm6              ; xmm4=((B-Y)+YE)=BE=B(02468ACE)
+    paddw       xmm5, xmm7              ; xmm5=((B-Y)+YO)=BO=B(13579BDF)
+    packuswb    xmm4, xmm4              ; xmm4=B(02468ACE********)
+    packuswb    xmm5, xmm5              ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+    ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+    ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+    ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+    punpcklbw   xmmA, xmmC        ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+    punpcklbw   xmmE, xmmB        ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+    punpcklbw   xmmD, xmmF        ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+    movdqa      xmmG, xmmA
+    movdqa      xmmH, xmmA
+    punpcklwd   xmmA, xmmE        ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+    punpckhwd   xmmG, xmmE        ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+    psrldq      xmmH, 2           ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+    psrldq      xmmE, 2           ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+    movdqa      xmmC, xmmD
+    movdqa      xmmB, xmmD
+    punpcklwd   xmmD, xmmH        ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+    punpckhwd   xmmC, xmmH        ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+    psrldq      xmmB, 2           ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+    movdqa      xmmF, xmmE
+    punpcklwd   xmmE, xmmB        ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+    punpckhwd   xmmF, xmmB        ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+    pshufd      xmmH, xmmA, 0x4E  ; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+    movdqa      xmmB, xmmE
+    punpckldq   xmmA, xmmD        ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+    punpckldq   xmmE, xmmH        ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+    punpckhdq   xmmD, xmmB        ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+    pshufd      xmmH, xmmG, 0x4E  ; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+    movdqa      xmmB, xmmF
+    punpckldq   xmmG, xmmC        ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+    punpckldq   xmmF, xmmH        ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+    punpckhdq   xmmC, xmmB        ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+    punpcklqdq  xmmA, xmmE        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+    punpcklqdq  xmmD, xmmG        ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    punpcklqdq  xmmF, xmmC        ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jb          short .column_st32
+
+    test        edi, SIZEOF_XMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    movntdq     XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movntdq     XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    movntdq     XMMWORD edi+2*SIZEOF_XMMWORD, xmmF
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    movdqu      XMMWORD edi+2*SIZEOF_XMMWORD, xmmF
+.out0:
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+    sub         ecx, byte SIZEOF_XMMWORD
+    jz          near .endcolumn
+
+    add         esi, byte SIZEOF_XMMWORD  ; inptr0
+    dec         al                        ; Yctr
+    jnz         near .Yloop_2nd
+
+    add         ebx, byte SIZEOF_XMMWORD  ; inptr1
+    add         edx, byte SIZEOF_XMMWORD  ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st32:
+    lea         ecx, ecx+ecx*2            ; imul ecx, RGB_PIXELSIZE
+    cmp         ecx, byte 2*SIZEOF_XMMWORD
+    jb          short .column_st16
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    add         edi, byte 2*SIZEOF_XMMWORD  ; outptr
+    movdqa      xmmA, xmmF
+    sub         ecx, byte 2*SIZEOF_XMMWORD
+    jmp         short .column_st15
+.column_st16:
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jb          short .column_st15
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    add         edi, byte SIZEOF_XMMWORD    ; outptr
+    movdqa      xmmA, xmmD
+    sub         ecx, byte SIZEOF_XMMWORD
+.column_st15:
+    ; Store the lower 8 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_MMWORD
+    jb          short .column_st7
+    movq        XMM_MMWORD edi, xmmA
+    add         edi, byte SIZEOF_MMWORD
+    sub         ecx, byte SIZEOF_MMWORD
+    psrldq      xmmA, SIZEOF_MMWORD
+.column_st7:
+    ; Store the lower 4 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_DWORD
+    jb          short .column_st3
+    movd        XMM_DWORD edi, xmmA
+    add         edi, byte SIZEOF_DWORD
+    sub         ecx, byte SIZEOF_DWORD
+    psrldq      xmmA, SIZEOF_DWORD
+.column_st3:
+    ; Store the lower 2 bytes of eax to the output when it has enough
+    ; space.
+    movd        eax, xmmA
+    cmp         ecx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word edi, ax
+    add         edi, byte SIZEOF_WORD
+    sub         ecx, byte SIZEOF_WORD
+    shr         eax, 16
+.column_st1:
+    ; Store the lower 1 byte of eax to the output when it has enough
+    ; space.
+    test        ecx, ecx
+    jz          short .endcolumn
+    mov         byte edi, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    pcmpeqb     xmm6, xmm6              ; xmm6=XE=X(02468ACE********)
+    pcmpeqb     xmm7, xmm7              ; xmm7=XO=X(13579BDF********)
+%else
+    pxor        xmm6, xmm6              ; xmm6=XE=X(02468ACE********)
+    pxor        xmm7, xmm7              ; xmm7=XO=X(13579BDF********)
+%endif
+    ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+    ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+    ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+    ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+    punpcklbw   xmmA, xmmC  ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+    punpcklbw   xmmE, xmmG  ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+    punpcklbw   xmmB, xmmD  ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+    punpcklbw   xmmF, xmmH  ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+    movdqa      xmmC, xmmA
+    punpcklwd   xmmA, xmmE  ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+    punpckhwd   xmmC, xmmE  ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+    movdqa      xmmG, xmmB
+    punpcklwd   xmmB, xmmF  ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+    punpckhwd   xmmG, xmmF  ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+    movdqa      xmmD, xmmA
+    punpckldq   xmmA, xmmB  ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+    punpckhdq   xmmD, xmmB  ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    movdqa      xmmH, xmmC
+    punpckldq   xmmC, xmmG  ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+    punpckhdq   xmmH, xmmG  ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+    cmp         ecx, byte SIZEOF_XMMWORD
+    jb          short .column_st32
+
+    test        edi, SIZEOF_XMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    movntdq     XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movntdq     XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    movntdq     XMMWORD edi+2*SIZEOF_XMMWORD, xmmC
+    movntdq     XMMWORD edi+3*SIZEOF_XMMWORD, xmmH
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    movdqu      XMMWORD edi+2*SIZEOF_XMMWORD, xmmC
+    movdqu      XMMWORD edi+3*SIZEOF_XMMWORD, xmmH
+.out0:
+    add         edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+    sub         ecx, byte SIZEOF_XMMWORD
+    jz          near .endcolumn
+
+    add         esi, byte SIZEOF_XMMWORD  ; inptr0
+    dec         al                        ; Yctr
+    jnz         near .Yloop_2nd
+
+    add         ebx, byte SIZEOF_XMMWORD  ; inptr1
+    add         edx, byte SIZEOF_XMMWORD  ; inptr2
+    jmp         near .columnloop
+    alignx      16, 7
+
+.column_st32:
+    cmp         ecx, byte SIZEOF_XMMWORD/2
+    jb          short .column_st16
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD edi+1*SIZEOF_XMMWORD, xmmD
+    add         edi, byte 2*SIZEOF_XMMWORD  ; outptr
+    movdqa      xmmA, xmmC
+    movdqa      xmmD, xmmH
+    sub         ecx, byte SIZEOF_XMMWORD/2
+.column_st16:
+    cmp         ecx, byte SIZEOF_XMMWORD/4
+    jb          short .column_st15
+    movdqu      XMMWORD edi+0*SIZEOF_XMMWORD, xmmA
+    add         edi, byte SIZEOF_XMMWORD    ; outptr
+    movdqa      xmmA, xmmD
+    sub         ecx, byte SIZEOF_XMMWORD/4
+.column_st15:
+    ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+    ; space.
+    cmp         ecx, byte SIZEOF_XMMWORD/8
+    jb          short .column_st7
+    movq        XMM_MMWORD edi, xmmA
+    add         edi, byte SIZEOF_XMMWORD/8*4
+    sub         ecx, byte SIZEOF_XMMWORD/8
+    psrldq      xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+    ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+    ; space.
+    test        ecx, ecx
+    jz          short .endcolumn
+    movd        XMM_DWORD edi, xmmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+    sfence                              ; flush the write buffer
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_sse2(JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+%define output_width(b)      (b) + 8    ; JDIMENSION output_width
+%define input_buf(b)         (b) + 12   ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)  (b) + 16   ; JDIMENSION in_row_group_ctr
+%define output_buf(b)        (b) + 20   ; JSAMPARRAY output_buf
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_merged_upsample_sse2)
+
+EXTN(jsimd_h2v2_merged_upsample_sse2):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         eax, POINTER output_width(ebp)
+
+    mov         edi, JSAMPIMAGE input_buf(ebp)
+    mov         ecx, JDIMENSION in_row_group_ctr(ebp)
+    mov         esi, JSAMPARRAY edi+0*SIZEOF_JSAMPARRAY
+    mov         ebx, JSAMPARRAY edi+1*SIZEOF_JSAMPARRAY
+    mov         edx, JSAMPARRAY edi+2*SIZEOF_JSAMPARRAY
+    mov         edi, JSAMPARRAY output_buf(ebp)
+    lea         esi, esi+ecx*SIZEOF_JSAMPROW
+
+    push        edx                     ; inptr2
+    push        ebx                     ; inptr1
+    push        esi                     ; inptr00
+    mov         ebx, esp
+
+    push        edi                     ; output_buf (outptr0)
+    push        ecx                     ; in_row_group_ctr
+    push        ebx                     ; input_buf
+    push        eax                     ; output_width
+
+    call        near EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+    add         esi, byte SIZEOF_JSAMPROW  ; inptr01
+    add         edi, byte SIZEOF_JSAMPROW  ; outptr1
+    mov         POINTER ebx+0*SIZEOF_POINTER, esi
+    mov         POINTER ebx-1*SIZEOF_POINTER, edi
+
+    call        near EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+    add         esp, byte 7*SIZEOF_DWORD
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdsample-avx2.asm Added

@@ -0,0 +1,760 @@
+;
+; jdsample.asm - upsampling (AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fancy_upsample_avx2)
+
+EXTN(jconst_fancy_upsample_avx2):
+
+PW_ONE   times 16 dw 1
+PW_TWO   times 16 dw 2
+PW_THREE times 16 dw 3
+PW_SEVEN times 16 dw 7
+PW_EIGHT times 16 dw 8
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_avx2(int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define downsamp_width(b)   (b) + 12    ; JDIMENSION downsampled_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_fancy_upsample_avx2)
+
+EXTN(jsimd_h2v1_fancy_upsample_avx2):
+    push        ebp
+    mov         ebp, esp
+    pushpic     ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    mov         eax, JDIMENSION downsamp_width(ebp)  ; colctr
+    test        eax, eax
+    jz          near .return
+
+    mov         ecx, INT max_v_samp(ebp)  ; rowctr
+    test        ecx, ecx
+    jz          near .return
+
+    mov         esi, JSAMPARRAY input_data(ebp)    ; input_data
+    mov         edi, POINTER output_data_ptr(ebp)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        eax                     ; colctr
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr
+
+    test        eax, SIZEOF_YMMWORD-1
+    jz          short .skip
+    mov         dl, JSAMPLE esi+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE esi+eax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+.skip:
+    vpxor       ymm0, ymm0, ymm0                ; ymm0=(all 0's)
+    vpcmpeqb    xmm7, xmm7, xmm7
+    vpsrldq     xmm7, xmm7, (SIZEOF_XMMWORD-1)  ; (ff -- -- -- ... -- --) LSB is ff
+    vpand       ymm7, ymm7, YMMWORD esi+0*SIZEOF_YMMWORD
+
+    add         eax, byte SIZEOF_YMMWORD-1
+    and         eax, byte -SIZEOF_YMMWORD
+    cmp         eax, byte SIZEOF_YMMWORD
+    ja          short .columnloop
+    alignx      16, 7
+
+.columnloop_last:
+    vpcmpeqb    xmm6, xmm6, xmm6
+    vpslldq     xmm6, xmm6, (SIZEOF_XMMWORD-1)
+    vperm2i128  ymm6, ymm6, ymm6, 1             ; (---- ---- ... ---- ---- ff) MSB is ff
+    vpand       ymm6, ymm6, YMMWORD esi+0*SIZEOF_YMMWORD
+    jmp         short .upsample
+    alignx      16, 7
+
+.columnloop:
+    vmovdqu     ymm6, YMMWORD esi+1*SIZEOF_YMMWORD
+    vperm2i128  ymm6, ymm0, ymm6, 0x20
+    vpslldq     ymm6, ymm6, 15
+
+.upsample:
+    vmovdqu     ymm1, YMMWORD esi+0*SIZEOF_YMMWORD  ; ymm1=( 0  1  2 ... 29 30 31)
+
+    vperm2i128  ymm2, ymm0, ymm1, 0x20
+    vpalignr    ymm2, ymm1, ymm2, 15            ; ymm2=(--  0  1 ... 28 29 30)
+    vperm2i128  ymm4, ymm0, ymm1, 0x03
+    vpalignr    ymm3, ymm4, ymm1, 1             ; ymm3=( 1  2  3 ... 30 31 --)
+
+    vpor        ymm2, ymm2, ymm7                ; ymm2=(-1  0  1 ... 28 29 30)
+    vpor        ymm3, ymm3, ymm6                ; ymm3=( 1  2  3 ... 30 31 32)
+
+    vpsrldq     ymm7, ymm4, (SIZEOF_XMMWORD-1)  ; ymm7=(31 -- -- ... -- -- --)
+
+    vpunpckhbw  ymm4, ymm1, ymm0                ; ymm4=( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm5, ymm1, ymm0                ; ymm5=( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm1, ymm5, ymm4, 0x20          ; ymm1=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm4, ymm5, ymm4, 0x31          ; ymm4=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm5, ymm2, ymm0                ; ymm5=( 7  8  9 10 11 12 13 14 23 24 25 26 27 28 29 30)
+    vpunpcklbw  ymm6, ymm2, ymm0                ; ymm6=(-1  0  1  2  3  4  5  6 15 16 17 18 19 20 21 22)
+    vperm2i128  ymm2, ymm6, ymm5, 0x20          ; ymm2=(-1  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+    vperm2i128  ymm5, ymm6, ymm5, 0x31          ; ymm5=(15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vpunpckhbw  ymm6, ymm3, ymm0                ; ymm6=( 1  2  3  4  5  6  7  8 17 18 19 20 21 22 23 24)
+    vpunpcklbw  ymm0, ymm3, ymm0                ; ymm0=( 9 10 11 12 13 14 15 16 25 26 27 28 29 30 31 32)
+    vperm2i128  ymm3, ymm0, ymm6, 0x20          ; ymm3=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16)
+    vperm2i128  ymm6, ymm0, ymm6, 0x31          ; ymm6=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32)
+
+    vpxor       ymm0, ymm0, ymm0                ; ymm0=(all 0's)
+
+    vpmullw     ymm1, ymm1, GOTOFF(ebx,PW_THREE)
+    vpmullw     ymm4, ymm4, GOTOFF(ebx,PW_THREE)
+    vpaddw      ymm2, ymm2, GOTOFF(ebx,PW_ONE)
+    vpaddw      ymm5, ymm5, GOTOFF(ebx,PW_ONE)
+    vpaddw      ymm3, ymm3, GOTOFF(ebx,PW_TWO)
+    vpaddw      ymm6, ymm6, GOTOFF(ebx,PW_TWO)
+
+    vpaddw      ymm2, ymm2, ymm1
+    vpaddw      ymm5, ymm5, ymm4
+    vpsrlw      ymm2, ymm2, 2                   ; ymm2=OutLE=( 0  2  4  6  8 10 12 14 16 18 20 22 24 26 28 30)
+    vpsrlw      ymm5, ymm5, 2                   ; ymm5=OutHE=(32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62)
+    vpaddw      ymm3, ymm3, ymm1
+    vpaddw      ymm6, ymm6, ymm4
+    vpsrlw      ymm3, ymm3, 2                   ; ymm3=OutLO=( 1  3  5  7  9 11 13 15 17 19 21 23 25 27 29 31)
+    vpsrlw      ymm6, ymm6, 2                   ; ymm6=OutHO=(33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63)
+
+    vpsllw      ymm3, ymm3, BYTE_BIT
+    vpsllw      ymm6, ymm6, BYTE_BIT
+    vpor        ymm2, ymm2, ymm3                ; ymm2=OutL=( 0  1  2 ... 29 30 31)
+    vpor        ymm5, ymm5, ymm6                ; ymm5=OutH=(32 33 34 ... 61 62 63)
+
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymm2
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymm5
+
+    sub         eax, byte SIZEOF_YMMWORD
+    add         esi, byte 1*SIZEOF_YMMWORD  ; inptr
+    add         edi, byte 2*SIZEOF_YMMWORD  ; outptr
+    cmp         eax, byte SIZEOF_YMMWORD
+    ja          near .columnloop
+    test        eax, eax
+    jnz         near .columnloop_last
+
+    pop         esi
+    pop         edi
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         ecx                        ; rowctr
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    poppic      ebx
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_avx2(int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define downsamp_width(b)   (b) + 12    ; JDIMENSION downsampled_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_YMMWORD
+                                        ; ymmword wkWK_NUM
+%define WK_NUM        4
+%define gotptr        wk(0) - SIZEOF_POINTER  ; void *gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_fancy_upsample_avx2)
+
+EXTN(jsimd_h2v2_fancy_upsample_avx2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         edx, eax                ; edx = original ebp
+    mov         eax, JDIMENSION downsamp_width(edx)  ; colctr
+    test        eax, eax
+    jz          near .return
+
+    mov         ecx, INT max_v_samp(edx)  ; rowctr
+    test        ecx, ecx
+    jz          near .return
+
+    mov         esi, JSAMPARRAY input_data(edx)    ; input_data
+    mov         edi, POINTER output_data_ptr(edx)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        eax                     ; colctr
+    push        ecx
+    push        edi
+    push        esi
+
+    mov         ecx, JSAMPROW esi-1*SIZEOF_JSAMPROW  ; inptr1(above)
+    mov         ebx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         esi, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; inptr1(below)
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW  ; outptr0
+    mov         edi, JSAMPROW edi+1*SIZEOF_JSAMPROW  ; outptr1
+
+    test        eax, SIZEOF_YMMWORD-1
+    jz          short .skip
+    push        edx
+    mov         dl, JSAMPLE ecx+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE ecx+eax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE ebx+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE ebx+eax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE esi+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE esi+eax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+    pop         edx
+.skip:
+    ; -- process the first column block
+
+    vmovdqu     ymm0, YMMWORD ebx+0*SIZEOF_YMMWORD  ; ymm0=row 00
+    vmovdqu     ymm1, YMMWORD ecx+0*SIZEOF_YMMWORD  ; ymm1=row-10
+    vmovdqu     ymm2, YMMWORD esi+0*SIZEOF_YMMWORD  ; ymm2=row+10
+
+    pushpic     ebx
+    movpic      ebx, POINTER gotptr   ; load GOT address
+
+    vpxor       ymm3, ymm3, ymm3        ; ymm3=(all 0's)
+
+    vpunpckhbw  ymm4, ymm0, ymm3        ; ymm4=row 0( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm5, ymm0, ymm3        ; ymm5=row 0( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm0, ymm5, ymm4, 0x20  ; ymm0=row 0( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm4, ymm5, ymm4, 0x31  ; ymm4=row 0(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm5, ymm1, ymm3        ; ymm5=row-1( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm6, ymm1, ymm3        ; ymm6=row-1( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm1, ymm6, ymm5, 0x20  ; ymm1=row-1( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm5, ymm6, ymm5, 0x31  ; ymm5=row-1(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm6, ymm2, ymm3        ; ymm6=row+1( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm3, ymm2, ymm3        ; ymm3=row+1( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm2, ymm3, ymm6, 0x20  ; ymm2=row+1( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm6, ymm3, ymm6, 0x31  ; ymm6=row+1(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpmullw     ymm0, ymm0, GOTOFF(ebx,PW_THREE)
+    vpmullw     ymm4, ymm4, GOTOFF(ebx,PW_THREE)
+
+    vpcmpeqb    xmm7, xmm7, xmm7
+    vpsrldq     xmm7, xmm7, (SIZEOF_XMMWORD-2)  ; (ffff ---- ---- ... ---- ----) LSB is ffff
+
+    vpaddw      ymm1, ymm1, ymm0        ; ymm1=Int0L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vpaddw      ymm5, ymm5, ymm4        ; ymm5=Int0H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+    vpaddw      ymm2, ymm2, ymm0        ; ymm2=Int1L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vpaddw      ymm6, ymm6, ymm4        ; ymm6=Int1H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vmovdqu     YMMWORD edx+0*SIZEOF_YMMWORD, ymm1  ; temporarily save
+    vmovdqu     YMMWORD edx+1*SIZEOF_YMMWORD, ymm5  ; the intermediate data
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymm2
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymm6
+
+    vpand       ymm1, ymm1, ymm7        ; ymm1=( 0 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vpand       ymm2, ymm2, ymm7        ; ymm2=( 0 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+
+    vmovdqa     YMMWORD wk(0), ymm1
+    vmovdqa     YMMWORD wk(1), ymm2
+
+    poppic      ebx
+
+    add         eax, byte SIZEOF_YMMWORD-1
+    and         eax, byte -SIZEOF_YMMWORD
+    cmp         eax, byte SIZEOF_YMMWORD
+    ja          short .columnloop
+    alignx      16, 7
+
+.columnloop_last:
+    ; -- process the last column block
+
+    pushpic     ebx
+    movpic      ebx, POINTER gotptr   ; load GOT address
+
+    vpcmpeqb    xmm1, xmm1, xmm1
+    vpslldq     xmm1, xmm1, (SIZEOF_XMMWORD-2)
+    vperm2i128  ymm1, ymm1, ymm1, 1             ; (---- ---- ... ---- ---- ffff) MSB is ffff
+
+    vpand       ymm2, ymm1, YMMWORD edi+1*SIZEOF_YMMWORD
+    vpand       ymm1, ymm1, YMMWORD edx+1*SIZEOF_YMMWORD
+
+    vmovdqa     YMMWORD wk(2), ymm1          ; ymm1=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 31)
+    vmovdqa     YMMWORD wk(3), ymm2          ; ymm2=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 31)
+
+    jmp         near .upsample
+    alignx      16, 7
+
+.columnloop:
+    ; -- process the next column block
+
+    vmovdqu     ymm0, YMMWORD ebx+1*SIZEOF_YMMWORD  ; ymm0=row 01
+    vmovdqu     ymm1, YMMWORD ecx+1*SIZEOF_YMMWORD  ; ymm1=row-11
+    vmovdqu     ymm2, YMMWORD esi+1*SIZEOF_YMMWORD  ; ymm2=row+11
+
+    pushpic     ebx
+    movpic      ebx, POINTER gotptr   ; load GOT address
+
+    vpxor       ymm3, ymm3, ymm3        ; ymm3=(all 0's)
+
+    vpunpckhbw  ymm4, ymm0, ymm3        ; ymm4=row 0( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm5, ymm0, ymm3        ; ymm5=row 0( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm0, ymm5, ymm4, 0x20  ; ymm0=row 0( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm4, ymm5, ymm4, 0x31  ; ymm4=row 0(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm5, ymm1, ymm3        ; ymm5=row-1( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm6, ymm1, ymm3        ; ymm6=row-1( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm1, ymm6, ymm5, 0x20  ; ymm1=row-1( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm5, ymm6, ymm5, 0x31  ; ymm5=row-1(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm6, ymm2, ymm3        ; ymm6=row+1( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm7, ymm2, ymm3        ; ymm7=row+1( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm2, ymm7, ymm6, 0x20  ; ymm2=row+1( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm6, ymm7, ymm6, 0x31  ; ymm6=row+1(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpmullw     ymm0, ymm0, GOTOFF(ebx,PW_THREE)
+    vpmullw     ymm4, ymm4, GOTOFF(ebx,PW_THREE)
+
+    vpaddw      ymm1, ymm1, ymm0        ; ymm1=Int0L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vpaddw      ymm5, ymm5, ymm4        ; ymm5=Int0H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+    vpaddw      ymm2, ymm2, ymm0        ; ymm2=Int1L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vpaddw      ymm6, ymm6, ymm4        ; ymm6=Int1H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vmovdqu     YMMWORD edx+2*SIZEOF_YMMWORD, ymm1  ; temporarily save
+    vmovdqu     YMMWORD edx+3*SIZEOF_YMMWORD, ymm5  ; the intermediate data
+    vmovdqu     YMMWORD edi+2*SIZEOF_YMMWORD, ymm2
+    vmovdqu     YMMWORD edi+3*SIZEOF_YMMWORD, ymm6
+
+    vperm2i128  ymm1, ymm3, ymm1, 0x20
+    vpslldq     ymm1, ymm1, 14          ; ymm1=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- --  0)
+    vperm2i128  ymm2, ymm3, ymm2, 0x20
+    vpslldq     ymm2, ymm2, 14          ; ymm2=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- --  0)
+
+    vmovdqa     YMMWORD wk(2), ymm1
+    vmovdqa     YMMWORD wk(3), ymm2
+
+.upsample:
+    ; -- process the upper row
+
+    vmovdqu     ymm7, YMMWORD edx+0*SIZEOF_YMMWORD  ; ymm7=Int0L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vmovdqu     ymm3, YMMWORD edx+1*SIZEOF_YMMWORD  ; ymm3=Int0H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpxor       ymm1, ymm1, ymm1        ; ymm1=(all 0's)
+
+    vperm2i128  ymm0, ymm1, ymm7, 0x03
+    vpalignr    ymm0, ymm0, ymm7, 2     ; ymm0=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 --)
+    vperm2i128  ymm4, ymm1, ymm3, 0x20
+    vpslldq     ymm4, ymm4, 14          ; ymm4=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 16)
+
+    vperm2i128  ymm5, ymm1, ymm7, 0x03
+    vpsrldq     ymm5, ymm5, 14          ; ymm5=(15 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vperm2i128  ymm6, ymm1, ymm3, 0x20
+    vpalignr    ymm6, ymm3, ymm6, 14    ; ymm6=(-- 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vpor        ymm0, ymm0, ymm4        ; ymm0=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16)
+    vpor        ymm5, ymm5, ymm6        ; ymm5=(15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vperm2i128  ymm2, ymm1, ymm3, 0x03
+    vpalignr    ymm2, ymm2, ymm3, 2     ; ymm2=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 --)
+    vperm2i128  ymm4, ymm1, ymm3, 0x03
+    vpsrldq     ymm4, ymm4, 14          ; ymm4=(31 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vperm2i128  ymm1, ymm1, ymm7, 0x20
+    vpalignr    ymm1, ymm7, ymm1, 14    ; ymm1=(--  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+
+    vpor        ymm1, ymm1, YMMWORD wk(0)  ; ymm1=(-1  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+    vpor        ymm2, ymm2, YMMWORD wk(2)  ; ymm2=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32)
+
+    vmovdqa     YMMWORD wk(0), ymm4
+
+    vpmullw     ymm7, ymm7, GOTOFF(ebx,PW_THREE)
+    vpmullw     ymm3, ymm3, GOTOFF(ebx,PW_THREE)
+    vpaddw      ymm1, ymm1, GOTOFF(ebx,PW_EIGHT)
+    vpaddw      ymm5, ymm5, GOTOFF(ebx,PW_EIGHT)
+    vpaddw      ymm0, ymm0, GOTOFF(ebx,PW_SEVEN)
+    vpaddw      ymm2, GOTOFF(ebx,PW_SEVEN)
+
+    vpaddw      ymm1, ymm1, ymm7
+    vpaddw      ymm5, ymm5, ymm3
+    vpsrlw      ymm1, ymm1, 4           ; ymm1=Out0LE=( 0  2  4  6  8 10 12 14 16 18 20 22 24 26 28 30)
+    vpsrlw      ymm5, ymm5, 4           ; ymm5=Out0HE=(32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62)
+    vpaddw      ymm0, ymm0, ymm7
+    vpaddw      ymm2, ymm2, ymm3
+    vpsrlw      ymm0, ymm0, 4           ; ymm0=Out0LO=( 1  3  5  7  9 11 13 15 17 19 21 23 25 27 29 31)
+    vpsrlw      ymm2, ymm2, 4           ; ymm2=Out0HO=(33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63)
+
+    vpsllw      ymm0, ymm0, BYTE_BIT
+    vpsllw      ymm2, ymm2, BYTE_BIT
+    vpor        ymm1, ymm1, ymm0        ; ymm1=Out0L=( 0  1  2 ... 29 30 31)
+    vpor        ymm5, ymm5, ymm2        ; ymm5=Out0H=(32 33 34 ... 61 62 63)
+
+    vmovdqu     YMMWORD edx+0*SIZEOF_YMMWORD, ymm1
+    vmovdqu     YMMWORD edx+1*SIZEOF_YMMWORD, ymm5
+
+    ; -- process the lower row
+
+    vmovdqu     ymm6, YMMWORD edi+0*SIZEOF_YMMWORD  ; ymm6=Int1L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vmovdqu     ymm4, YMMWORD edi+1*SIZEOF_YMMWORD  ; ymm4=Int1H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpxor       ymm1, ymm1, ymm1        ; ymm1=(all 0's)
+
+    vperm2i128  ymm7, ymm1, ymm6, 0x03
+    vpalignr    ymm7, ymm7, ymm6, 2     ; ymm7=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 --)
+    vperm2i128  ymm3, ymm1, ymm4, 0x20
+    vpslldq     ymm3, ymm3, 14          ; ymm3=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 16)
+
+    vperm2i128  ymm0, ymm1, ymm6, 0x03
+    vpsrldq     ymm0, ymm0, 14          ; ymm0=(15 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vperm2i128  ymm2, ymm1, ymm4, 0x20
+    vpalignr    ymm2, ymm4, ymm2, 14    ; ymm2=(-- 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vpor        ymm7, ymm7, ymm3        ; ymm7=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16)
+    vpor        ymm0, ymm0, ymm2        ; ymm0=(15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vperm2i128  ymm5, ymm1, ymm4, 0x03
+    vpalignr    ymm5, ymm5, ymm4, 2     ; ymm5=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 --)
+    vperm2i128  ymm3, ymm1, ymm4, 0x03
+    vpsrldq     ymm3, ymm3, 14          ; ymm3=(31 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vperm2i128  ymm1, ymm1, ymm6, 0x20
+    vpalignr    ymm1, ymm6, ymm1, 14    ; ymm1=(--  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+
+    vpor        ymm1, ymm1, YMMWORD wk(1)  ; ymm1=(-1  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+    vpor        ymm5, ymm5, YMMWORD wk(3)  ; ymm5=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32)
+
+    vmovdqa     YMMWORD wk(1), ymm3
+
+    vpmullw     ymm6, ymm6, GOTOFF(ebx,PW_THREE)
+    vpmullw     ymm4, ymm4, GOTOFF(ebx,PW_THREE)
+    vpaddw      ymm1, ymm1, GOTOFF(ebx,PW_EIGHT)
+    vpaddw      ymm0, ymm0, GOTOFF(ebx,PW_EIGHT)
+    vpaddw      ymm7, ymm7, GOTOFF(ebx,PW_SEVEN)
+    vpaddw      ymm5, ymm5, GOTOFF(ebx,PW_SEVEN)
+
+    vpaddw      ymm1, ymm1, ymm6
+    vpaddw      ymm0, ymm0, ymm4
+    vpsrlw      ymm1, ymm1, 4           ; ymm1=Out1LE=( 0  2  4  6  8 10 12 14 16 18 20 22 24 26 28 30)
+    vpsrlw      ymm0, ymm0, 4           ; ymm0=Out1HE=(32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62)
+    vpaddw      ymm7, ymm7, ymm6
+    vpaddw      ymm5, ymm5, ymm4
+    vpsrlw      ymm7, ymm7, 4           ; ymm7=Out1LO=( 1  3  5  7  9 11 13 15 17 19 21 23 25 27 29 31)
+    vpsrlw      ymm5, ymm5, 4           ; ymm5=Out1HO=(33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63)
+
+    vpsllw      ymm7, ymm7, BYTE_BIT
+    vpsllw      ymm5, ymm5, BYTE_BIT
+    vpor        ymm1, ymm1, ymm7        ; ymm1=Out1L=( 0  1  2 ... 29 30 31)
+    vpor        ymm0, ymm0, ymm5        ; ymm0=Out1H=(32 33 34 ... 61 62 63)
+
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymm1
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymm0
+
+    poppic      ebx
+
+    sub         eax, byte SIZEOF_YMMWORD
+    add         ecx, byte 1*SIZEOF_YMMWORD  ; inptr1(above)
+    add         ebx, byte 1*SIZEOF_YMMWORD  ; inptr0
+    add         esi, byte 1*SIZEOF_YMMWORD  ; inptr1(below)
+    add         edx, byte 2*SIZEOF_YMMWORD  ; outptr0
+    add         edi, byte 2*SIZEOF_YMMWORD  ; outptr1
+    cmp         eax, byte SIZEOF_YMMWORD
+    ja          near .columnloop
+    test        eax, eax
+    jnz         near .columnloop_last
+
+    pop         esi
+    pop         edi
+    pop         ecx
+    pop         eax
+
+    add         esi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         ecx, byte 2                  ; rowctr
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_avx2(int max_v_samp_factor, JDIMENSION output_width,
+;                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define output_width(b)     (b) + 12    ; JDIMENSION output_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_upsample_avx2)
+
+EXTN(jsimd_h2v1_upsample_avx2):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         edx, JDIMENSION output_width(ebp)
+    add         edx, byte (SIZEOF_YMMWORD-1)
+    and         edx, -SIZEOF_YMMWORD
+    jz          short .return
+
+    mov         ecx, INT max_v_samp(ebp)  ; rowctr
+    test        ecx, ecx
+    jz          short .return
+
+    mov         esi, JSAMPARRAY input_data(ebp)    ; input_data
+    mov         edi, POINTER output_data_ptr(ebp)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr
+    mov         eax, edx                ; colctr
+    alignx      16, 7
+.columnloop:
+
+    cmp         eax, byte SIZEOF_YMMWORD
+    ja          near .above_16
+
+    vmovdqu     xmm0, XMMWORD esi+0*SIZEOF_YMMWORD
+    vpunpckhbw  xmm1, xmm0, xmm0
+    vpunpcklbw  xmm0, xmm0, xmm0
+
+    vmovdqu     XMMWORD edi+0*SIZEOF_XMMWORD, xmm0
+    vmovdqu     XMMWORD edi+1*SIZEOF_XMMWORD, xmm1
+
+    jmp         short .nextrow
+
+.above_16:
+    vmovdqu     ymm0, YMMWORD esi+0*SIZEOF_YMMWORD
+
+    vpermq      ymm0, ymm0, 0xd8
+    vpunpckhbw  ymm1, ymm0, ymm0
+    vpunpcklbw  ymm0, ymm0, ymm0
+
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymm0
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymm1
+
+    sub         eax, byte 2*SIZEOF_YMMWORD
+    jz          short .nextrow
+
+    add         esi, byte SIZEOF_YMMWORD    ; inptr
+    add         edi, byte 2*SIZEOF_YMMWORD  ; outptr
+    jmp         short .columnloop
+    alignx      16, 7
+
+.nextrow:
+    pop         esi
+    pop         edi
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         ecx                        ; rowctr
+    jg          short .rowloop
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_avx2(int max_v_samp_factor, JDIMENSION output_width,
+;                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define output_width(b)     (b) + 12    ; JDIMENSION output_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_upsample_avx2)
+
+EXTN(jsimd_h2v2_upsample_avx2):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         edx, JDIMENSION output_width(ebp)
+    add         edx, byte (SIZEOF_YMMWORD-1)
+    and         edx, -SIZEOF_YMMWORD
+    jz          near .return
+
+    mov         ecx, INT max_v_samp(ebp)  ; rowctr
+    test        ecx, ecx
+    jz          near .return
+
+    mov         esi, JSAMPARRAY input_data(ebp)    ; input_data
+    mov         edi, POINTER output_data_ptr(ebp)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi                    ; inptr
+    mov         ebx, JSAMPROW edi+0*SIZEOF_JSAMPROW  ; outptr0
+    mov         edi, JSAMPROW edi+1*SIZEOF_JSAMPROW  ; outptr1
+    mov         eax, edx                               ; colctr
+    alignx      16, 7
+.columnloop:
+
+    cmp         eax, byte SIZEOF_YMMWORD
+    ja          short .above_16
+
+    vmovdqu     xmm0, XMMWORD esi+0*SIZEOF_XMMWORD
+    vpunpckhbw  xmm1, xmm0, xmm0
+    vpunpcklbw  xmm0, xmm0, xmm0
+
+    vmovdqu     XMMWORD ebx+0*SIZEOF_XMMWORD, xmm0
+    vmovdqu     XMMWORD ebx+1*SIZEOF_XMMWORD, xmm1
+    vmovdqu     XMMWORD edi+0*SIZEOF_XMMWORD, xmm0
+    vmovdqu     XMMWORD edi+1*SIZEOF_XMMWORD, xmm1
+
+    jmp         near .nextrow
+
+.above_16:
+    vmovdqu     ymm0, YMMWORD esi+0*SIZEOF_YMMWORD
+
+    vpermq      ymm0, ymm0, 0xd8
+    vpunpckhbw  ymm1, ymm0, ymm0
+    vpunpcklbw  ymm0, ymm0, ymm0
+
+    vmovdqu     YMMWORD ebx+0*SIZEOF_YMMWORD, ymm0
+    vmovdqu     YMMWORD ebx+1*SIZEOF_YMMWORD, ymm1
+    vmovdqu     YMMWORD edi+0*SIZEOF_YMMWORD, ymm0
+    vmovdqu     YMMWORD edi+1*SIZEOF_YMMWORD, ymm1
+
+    sub         eax, byte 2*SIZEOF_YMMWORD
+    jz          short .nextrow
+
+    add         esi, byte SIZEOF_YMMWORD  ; inptr
+    add         ebx, 2*SIZEOF_YMMWORD     ; outptr0
+    add         edi, 2*SIZEOF_YMMWORD     ; outptr1
+    jmp         short .columnloop
+    alignx      16, 7
+
+.nextrow:
+    pop         esi
+    pop         edi
+
+    add         esi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         ecx, byte 2                  ; rowctr
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdsample-mmx.asm Added

@@ -0,0 +1,731 @@
+;
+; jdsample.asm - upsampling (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fancy_upsample_mmx)
+
+EXTN(jconst_fancy_upsample_mmx):
+
+PW_ONE   times 4 dw 1
+PW_TWO   times 4 dw 2
+PW_THREE times 4 dw 3
+PW_SEVEN times 4 dw 7
+PW_EIGHT times 4 dw 8
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_mmx(int max_v_samp_factor,
+;                               JDIMENSION downsampled_width,
+;                               JSAMPARRAY input_data,
+;                               JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define downsamp_width(b)   (b) + 12    ; JDIMENSION downsampled_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_fancy_upsample_mmx)
+
+EXTN(jsimd_h2v1_fancy_upsample_mmx):
+    push        ebp
+    mov         ebp, esp
+    pushpic     ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    mov         eax, JDIMENSION downsamp_width(ebp)  ; colctr
+    test        eax, eax
+    jz          near .return
+
+    mov         ecx, INT max_v_samp(ebp)  ; rowctr
+    test        ecx, ecx
+    jz          near .return
+
+    mov         esi, JSAMPARRAY input_data(ebp)    ; input_data
+    mov         edi, POINTER output_data_ptr(ebp)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        eax                     ; colctr
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr
+
+    test        eax, SIZEOF_MMWORD-1
+    jz          short .skip
+    mov         dl, JSAMPLE esi+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE esi+eax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+.skip:
+    pxor        mm0, mm0                ; mm0=(all 0's)
+    pcmpeqb     mm7, mm7
+    psrlq       mm7, (SIZEOF_MMWORD-1)*BYTE_BIT
+    pand        mm7,  MMWORD esi+0*SIZEOF_MMWORD
+
+    add         eax, byte SIZEOF_MMWORD-1
+    and         eax, byte -SIZEOF_MMWORD
+    cmp         eax, byte SIZEOF_MMWORD
+    ja          short .columnloop
+    alignx      16, 7
+
+.columnloop_last:
+    pcmpeqb     mm6, mm6
+    psllq       mm6, (SIZEOF_MMWORD-1)*BYTE_BIT
+    pand        mm6, MMWORD esi+0*SIZEOF_MMWORD
+    jmp         short .upsample
+    alignx      16, 7
+
+.columnloop:
+    movq        mm6, MMWORD esi+1*SIZEOF_MMWORD
+    psllq       mm6, (SIZEOF_MMWORD-1)*BYTE_BIT
+
+.upsample:
+    movq        mm1, MMWORD esi+0*SIZEOF_MMWORD
+    movq        mm2, mm1
+    movq        mm3, mm1                ; mm1=( 0 1 2 3 4 5 6 7)
+    psllq       mm2, BYTE_BIT           ; mm2=( - 0 1 2 3 4 5 6)
+    psrlq       mm3, BYTE_BIT           ; mm3=( 1 2 3 4 5 6 7 -)
+
+    por         mm2, mm7                ; mm2=(-1 0 1 2 3 4 5 6)
+    por         mm3, mm6                ; mm3=( 1 2 3 4 5 6 7 8)
+
+    movq        mm7, mm1
+    psrlq       mm7, (SIZEOF_MMWORD-1)*BYTE_BIT  ; mm7=( 7 - - - - - - -)
+
+    movq        mm4, mm1
+    punpcklbw   mm1, mm0                ; mm1=( 0 1 2 3)
+    punpckhbw   mm4, mm0                ; mm4=( 4 5 6 7)
+    movq        mm5, mm2
+    punpcklbw   mm2, mm0                ; mm2=(-1 0 1 2)
+    punpckhbw   mm5, mm0                ; mm5=( 3 4 5 6)
+    movq        mm6, mm3
+    punpcklbw   mm3, mm0                ; mm3=( 1 2 3 4)
+    punpckhbw   mm6, mm0                ; mm6=( 5 6 7 8)
+
+    pmullw      mm1, GOTOFF(ebx,PW_THREE)
+    pmullw      mm4, GOTOFF(ebx,PW_THREE)
+    paddw       mm2, GOTOFF(ebx,PW_ONE)
+    paddw       mm5, GOTOFF(ebx,PW_ONE)
+    paddw       mm3, GOTOFF(ebx,PW_TWO)
+    paddw       mm6, GOTOFF(ebx,PW_TWO)
+
+    paddw       mm2, mm1
+    paddw       mm5, mm4
+    psrlw       mm2, 2                  ; mm2=OutLE=( 0  2  4  6)
+    psrlw       mm5, 2                  ; mm5=OutHE=( 8 10 12 14)
+    paddw       mm3, mm1
+    paddw       mm6, mm4
+    psrlw       mm3, 2                  ; mm3=OutLO=( 1  3  5  7)
+    psrlw       mm6, 2                  ; mm6=OutHO=( 9 11 13 15)
+
+    psllw       mm3, BYTE_BIT
+    psllw       mm6, BYTE_BIT
+    por         mm2, mm3                ; mm2=OutL=( 0  1  2  3  4  5  6  7)
+    por         mm5, mm6                ; mm5=OutH=( 8  9 10 11 12 13 14 15)
+
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mm2
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mm5
+
+    sub         eax, byte SIZEOF_MMWORD
+    add         esi, byte 1*SIZEOF_MMWORD  ; inptr
+    add         edi, byte 2*SIZEOF_MMWORD  ; outptr
+    cmp         eax, byte SIZEOF_MMWORD
+    ja          near .columnloop
+    test        eax, eax
+    jnz         near .columnloop_last
+
+    pop         esi
+    pop         edi
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         ecx                        ; rowctr
+    jg          near .rowloop
+
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    poppic      ebx
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_mmx(int max_v_samp_factor,
+;                               JDIMENSION downsampled_width,
+;                               JSAMPARRAY input_data,
+;                               JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define downsamp_width(b)   (b) + 12    ; JDIMENSION downsampled_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_MMWORD  ; mmword wkWK_NUM
+%define WK_NUM        4
+%define gotptr        wk(0) - SIZEOF_POINTER  ; void *gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_fancy_upsample_mmx)
+
+EXTN(jsimd_h2v2_fancy_upsample_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         edx, eax                ; edx = original ebp
+    mov         eax, JDIMENSION downsamp_width(edx)  ; colctr
+    test        eax, eax
+    jz          near .return
+
+    mov         ecx, INT max_v_samp(edx)  ; rowctr
+    test        ecx, ecx
+    jz          near .return
+
+    mov         esi, JSAMPARRAY input_data(edx)    ; input_data
+    mov         edi, POINTER output_data_ptr(edx)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        eax                     ; colctr
+    push        ecx
+    push        edi
+    push        esi
+
+    mov         ecx, JSAMPROW esi-1*SIZEOF_JSAMPROW  ; inptr1(above)
+    mov         ebx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         esi, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; inptr1(below)
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW  ; outptr0
+    mov         edi, JSAMPROW edi+1*SIZEOF_JSAMPROW  ; outptr1
+
+    test        eax, SIZEOF_MMWORD-1
+    jz          short .skip
+    push        edx
+    mov         dl, JSAMPLE ecx+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE ecx+eax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE ebx+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE ebx+eax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE esi+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE esi+eax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+    pop         edx
+.skip:
+    ; -- process the first column block
+
+    movq        mm0, MMWORD ebx+0*SIZEOF_MMWORD  ; mm0=row 00
+    movq        mm1, MMWORD ecx+0*SIZEOF_MMWORD  ; mm1=row-10
+    movq        mm2, MMWORD esi+0*SIZEOF_MMWORD  ; mm2=row+10
+
+    pushpic     ebx
+    movpic      ebx, POINTER gotptr   ; load GOT address
+
+    pxor        mm3, mm3                ; mm3=(all 0's)
+    movq        mm4, mm0
+    punpcklbw   mm0, mm3                ; mm0=row 00( 0 1 2 3)
+    punpckhbw   mm4, mm3                ; mm4=row 00( 4 5 6 7)
+    movq        mm5, mm1
+    punpcklbw   mm1, mm3                ; mm1=row-10( 0 1 2 3)
+    punpckhbw   mm5, mm3                ; mm5=row-10( 4 5 6 7)
+    movq        mm6, mm2
+    punpcklbw   mm2, mm3                ; mm2=row+10( 0 1 2 3)
+    punpckhbw   mm6, mm3                ; mm6=row+10( 4 5 6 7)
+
+    pmullw      mm0, GOTOFF(ebx,PW_THREE)
+    pmullw      mm4, GOTOFF(ebx,PW_THREE)
+
+    pcmpeqb     mm7, mm7
+    psrlq       mm7, (SIZEOF_MMWORD-2)*BYTE_BIT
+
+    paddw       mm1, mm0                ; mm1=Int0L=( 0 1 2 3)
+    paddw       mm5, mm4                ; mm5=Int0H=( 4 5 6 7)
+    paddw       mm2, mm0                ; mm2=Int1L=( 0 1 2 3)
+    paddw       mm6, mm4                ; mm6=Int1H=( 4 5 6 7)
+
+    movq        MMWORD edx+0*SIZEOF_MMWORD, mm1  ; temporarily save
+    movq        MMWORD edx+1*SIZEOF_MMWORD, mm5  ; the intermediate data
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mm2
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mm6
+
+    pand        mm1, mm7                ; mm1=( 0 - - -)
+    pand        mm2, mm7                ; mm2=( 0 - - -)
+
+    movq        MMWORD wk(0), mm1
+    movq        MMWORD wk(1), mm2
+
+    poppic      ebx
+
+    add         eax, byte SIZEOF_MMWORD-1
+    and         eax, byte -SIZEOF_MMWORD
+    cmp         eax, byte SIZEOF_MMWORD
+    ja          short .columnloop
+    alignx      16, 7
+
+.columnloop_last:
+    ; -- process the last column block
+
+    pushpic     ebx
+    movpic      ebx, POINTER gotptr   ; load GOT address
+
+    pcmpeqb     mm1, mm1
+    psllq       mm1, (SIZEOF_MMWORD-2)*BYTE_BIT
+    movq        mm2, mm1
+
+    pand        mm1, MMWORD edx+1*SIZEOF_MMWORD  ; mm1=( - - - 7)
+    pand        mm2, MMWORD edi+1*SIZEOF_MMWORD  ; mm2=( - - - 7)
+
+    movq        MMWORD wk(2), mm1
+    movq        MMWORD wk(3), mm2
+
+    jmp         short .upsample
+    alignx      16, 7
+
+.columnloop:
+    ; -- process the next column block
+
+    movq        mm0, MMWORD ebx+1*SIZEOF_MMWORD  ; mm0=row 01
+    movq        mm1, MMWORD ecx+1*SIZEOF_MMWORD  ; mm1=row-11
+    movq        mm2, MMWORD esi+1*SIZEOF_MMWORD  ; mm2=row+11
+
+    pushpic     ebx
+    movpic      ebx, POINTER gotptr   ; load GOT address
+
+    pxor        mm3, mm3                ; mm3=(all 0's)
+    movq        mm4, mm0
+    punpcklbw   mm0, mm3                ; mm0=row 01( 0 1 2 3)
+    punpckhbw   mm4, mm3                ; mm4=row 01( 4 5 6 7)
+    movq        mm5, mm1
+    punpcklbw   mm1, mm3                ; mm1=row-11( 0 1 2 3)
+    punpckhbw   mm5, mm3                ; mm5=row-11( 4 5 6 7)
+    movq        mm6, mm2
+    punpcklbw   mm2, mm3                ; mm2=row+11( 0 1 2 3)
+    punpckhbw   mm6, mm3                ; mm6=row+11( 4 5 6 7)
+
+    pmullw      mm0, GOTOFF(ebx,PW_THREE)
+    pmullw      mm4, GOTOFF(ebx,PW_THREE)
+
+    paddw       mm1, mm0                ; mm1=Int0L=( 0 1 2 3)
+    paddw       mm5, mm4                ; mm5=Int0H=( 4 5 6 7)
+    paddw       mm2, mm0                ; mm2=Int1L=( 0 1 2 3)
+    paddw       mm6, mm4                ; mm6=Int1H=( 4 5 6 7)
+
+    movq        MMWORD edx+2*SIZEOF_MMWORD, mm1  ; temporarily save
+    movq        MMWORD edx+3*SIZEOF_MMWORD, mm5  ; the intermediate data
+    movq        MMWORD edi+2*SIZEOF_MMWORD, mm2
+    movq        MMWORD edi+3*SIZEOF_MMWORD, mm6
+
+    psllq       mm1, (SIZEOF_MMWORD-2)*BYTE_BIT  ; mm1=( - - - 0)
+    psllq       mm2, (SIZEOF_MMWORD-2)*BYTE_BIT  ; mm2=( - - - 0)
+
+    movq        MMWORD wk(2), mm1
+    movq        MMWORD wk(3), mm2
+
+.upsample:
+    ; -- process the upper row
+
+    movq        mm7, MMWORD edx+0*SIZEOF_MMWORD  ; mm7=Int0L=( 0 1 2 3)
+    movq        mm3, MMWORD edx+1*SIZEOF_MMWORD  ; mm3=Int0H=( 4 5 6 7)
+
+    movq        mm0, mm7
+    movq        mm4, mm3
+    psrlq       mm0, 2*BYTE_BIT                  ; mm0=( 1 2 3 -)
+    psllq       mm4, (SIZEOF_MMWORD-2)*BYTE_BIT  ; mm4=( - - - 4)
+    movq        mm5, mm7
+    movq        mm6, mm3
+    psrlq       mm5, (SIZEOF_MMWORD-2)*BYTE_BIT  ; mm5=( 3 - - -)
+    psllq       mm6, 2*BYTE_BIT                  ; mm6=( - 4 5 6)
+
+    por         mm0, mm4                         ; mm0=( 1 2 3 4)
+    por         mm5, mm6                         ; mm5=( 3 4 5 6)
+
+    movq        mm1, mm7
+    movq        mm2, mm3
+    psllq       mm1, 2*BYTE_BIT                  ; mm1=( - 0 1 2)
+    psrlq       mm2, 2*BYTE_BIT                  ; mm2=( 5 6 7 -)
+    movq        mm4, mm3
+    psrlq       mm4, (SIZEOF_MMWORD-2)*BYTE_BIT  ; mm4=( 7 - - -)
+
+    por         mm1, MMWORD wk(0)              ; mm1=(-1 0 1 2)
+    por         mm2, MMWORD wk(2)              ; mm2=( 5 6 7 8)
+
+    movq        MMWORD wk(0), mm4
+
+    pmullw      mm7, GOTOFF(ebx,PW_THREE)
+    pmullw      mm3, GOTOFF(ebx,PW_THREE)
+    paddw       mm1, GOTOFF(ebx,PW_EIGHT)
+    paddw       mm5, GOTOFF(ebx,PW_EIGHT)
+    paddw       mm0, GOTOFF(ebx,PW_SEVEN)
+    paddw       mm2, GOTOFF(ebx,PW_SEVEN)
+
+    paddw       mm1, mm7
+    paddw       mm5, mm3
+    psrlw       mm1, 4                  ; mm1=Out0LE=( 0  2  4  6)
+    psrlw       mm5, 4                  ; mm5=Out0HE=( 8 10 12 14)
+    paddw       mm0, mm7
+    paddw       mm2, mm3
+    psrlw       mm0, 4                  ; mm0=Out0LO=( 1  3  5  7)
+    psrlw       mm2, 4                  ; mm2=Out0HO=( 9 11 13 15)
+
+    psllw       mm0, BYTE_BIT
+    psllw       mm2, BYTE_BIT
+    por         mm1, mm0                ; mm1=Out0L=( 0  1  2  3  4  5  6  7)
+    por         mm5, mm2                ; mm5=Out0H=( 8  9 10 11 12 13 14 15)
+
+    movq        MMWORD edx+0*SIZEOF_MMWORD, mm1
+    movq        MMWORD edx+1*SIZEOF_MMWORD, mm5
+
+    ; -- process the lower row
+
+    movq        mm6, MMWORD edi+0*SIZEOF_MMWORD  ; mm6=Int1L=( 0 1 2 3)
+    movq        mm4, MMWORD edi+1*SIZEOF_MMWORD  ; mm4=Int1H=( 4 5 6 7)
+
+    movq        mm7, mm6
+    movq        mm3, mm4
+    psrlq       mm7, 2*BYTE_BIT                  ; mm7=( 1 2 3 -)
+    psllq       mm3, (SIZEOF_MMWORD-2)*BYTE_BIT  ; mm3=( - - - 4)
+    movq        mm0, mm6
+    movq        mm2, mm4
+    psrlq       mm0, (SIZEOF_MMWORD-2)*BYTE_BIT  ; mm0=( 3 - - -)
+    psllq       mm2, 2*BYTE_BIT                  ; mm2=( - 4 5 6)
+
+    por         mm7, mm3                         ; mm7=( 1 2 3 4)
+    por         mm0, mm2                         ; mm0=( 3 4 5 6)
+
+    movq        mm1, mm6
+    movq        mm5, mm4
+    psllq       mm1, 2*BYTE_BIT                  ; mm1=( - 0 1 2)
+    psrlq       mm5, 2*BYTE_BIT                  ; mm5=( 5 6 7 -)
+    movq        mm3, mm4
+    psrlq       mm3, (SIZEOF_MMWORD-2)*BYTE_BIT  ; mm3=( 7 - - -)
+
+    por         mm1, MMWORD wk(1)              ; mm1=(-1 0 1 2)
+    por         mm5, MMWORD wk(3)              ; mm5=( 5 6 7 8)
+
+    movq        MMWORD wk(1), mm3
+
+    pmullw      mm6, GOTOFF(ebx,PW_THREE)
+    pmullw      mm4, GOTOFF(ebx,PW_THREE)
+    paddw       mm1, GOTOFF(ebx,PW_EIGHT)
+    paddw       mm0, GOTOFF(ebx,PW_EIGHT)
+    paddw       mm7, GOTOFF(ebx,PW_SEVEN)
+    paddw       mm5, GOTOFF(ebx,PW_SEVEN)
+
+    paddw       mm1, mm6
+    paddw       mm0, mm4
+    psrlw       mm1, 4                  ; mm1=Out1LE=( 0  2  4  6)
+    psrlw       mm0, 4                  ; mm0=Out1HE=( 8 10 12 14)
+    paddw       mm7, mm6
+    paddw       mm5, mm4
+    psrlw       mm7, 4                  ; mm7=Out1LO=( 1  3  5  7)
+    psrlw       mm5, 4                  ; mm5=Out1HO=( 9 11 13 15)
+
+    psllw       mm7, BYTE_BIT
+    psllw       mm5, BYTE_BIT
+    por         mm1, mm7                ; mm1=Out1L=( 0  1  2  3  4  5  6  7)
+    por         mm0, mm5                ; mm0=Out1H=( 8  9 10 11 12 13 14 15)
+
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mm1
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mm0
+
+    poppic      ebx
+
+    sub         eax, byte SIZEOF_MMWORD
+    add         ecx, byte 1*SIZEOF_MMWORD  ; inptr1(above)
+    add         ebx, byte 1*SIZEOF_MMWORD  ; inptr0
+    add         esi, byte 1*SIZEOF_MMWORD  ; inptr1(below)
+    add         edx, byte 2*SIZEOF_MMWORD  ; outptr0
+    add         edi, byte 2*SIZEOF_MMWORD  ; outptr1
+    cmp         eax, byte SIZEOF_MMWORD
+    ja          near .columnloop
+    test        eax, eax
+    jnz         near .columnloop_last
+
+    pop         esi
+    pop         edi
+    pop         ecx
+    pop         eax
+
+    add         esi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         ecx, byte 2                  ; rowctr
+    jg          near .rowloop
+
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_mmx(int max_v_samp_factor, JDIMENSION output_width,
+;                         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define output_width(b)     (b) + 12    ; JDIMENSION output_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_upsample_mmx)
+
+EXTN(jsimd_h2v1_upsample_mmx):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         edx, JDIMENSION output_width(ebp)
+    add         edx, byte (2*SIZEOF_MMWORD)-1
+    and         edx, byte -(2*SIZEOF_MMWORD)
+    jz          short .return
+
+    mov         ecx, INT max_v_samp(ebp)  ; rowctr
+    test        ecx, ecx
+    jz          short .return
+
+    mov         esi, JSAMPARRAY input_data(ebp)    ; input_data
+    mov         edi, POINTER output_data_ptr(ebp)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr
+    mov         eax, edx                ; colctr
+    alignx      16, 7
+.columnloop:
+
+    movq        mm0, MMWORD esi+0*SIZEOF_MMWORD
+
+    movq        mm1, mm0
+    punpcklbw   mm0, mm0
+    punpckhbw   mm1, mm1
+
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mm0
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mm1
+
+    sub         eax, byte 2*SIZEOF_MMWORD
+    jz          short .nextrow
+
+    movq        mm2, MMWORD esi+1*SIZEOF_MMWORD
+
+    movq        mm3, mm2
+    punpcklbw   mm2, mm2
+    punpckhbw   mm3, mm3
+
+    movq        MMWORD edi+2*SIZEOF_MMWORD, mm2
+    movq        MMWORD edi+3*SIZEOF_MMWORD, mm3
+
+    sub         eax, byte 2*SIZEOF_MMWORD
+    jz          short .nextrow
+
+    add         esi, byte 2*SIZEOF_MMWORD  ; inptr
+    add         edi, byte 4*SIZEOF_MMWORD  ; outptr
+    jmp         short .columnloop
+    alignx      16, 7
+
+.nextrow:
+    pop         esi
+    pop         edi
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         ecx                        ; rowctr
+    jg          short .rowloop
+
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_mmx(int max_v_samp_factor, JDIMENSION output_width,
+;                         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define output_width(b)     (b) + 12    ; JDIMENSION output_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_upsample_mmx)
+
+EXTN(jsimd_h2v2_upsample_mmx):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         edx, JDIMENSION output_width(ebp)
+    add         edx, byte (2*SIZEOF_MMWORD)-1
+    and         edx, byte -(2*SIZEOF_MMWORD)
+    jz          near .return
+
+    mov         ecx, INT max_v_samp(ebp)  ; rowctr
+    test        ecx, ecx
+    jz          short .return
+
+    mov         esi, JSAMPARRAY input_data(ebp)    ; input_data
+    mov         edi, POINTER output_data_ptr(ebp)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi                    ; inptr
+    mov         ebx, JSAMPROW edi+0*SIZEOF_JSAMPROW  ; outptr0
+    mov         edi, JSAMPROW edi+1*SIZEOF_JSAMPROW  ; outptr1
+    mov         eax, edx                               ; colctr
+    alignx      16, 7
+.columnloop:
+
+    movq        mm0, MMWORD esi+0*SIZEOF_MMWORD
+
+    movq        mm1, mm0
+    punpcklbw   mm0, mm0
+    punpckhbw   mm1, mm1
+
+    movq        MMWORD ebx+0*SIZEOF_MMWORD, mm0
+    movq        MMWORD ebx+1*SIZEOF_MMWORD, mm1
+    movq        MMWORD edi+0*SIZEOF_MMWORD, mm0
+    movq        MMWORD edi+1*SIZEOF_MMWORD, mm1
+
+    sub         eax, byte 2*SIZEOF_MMWORD
+    jz          short .nextrow
+
+    movq        mm2, MMWORD esi+1*SIZEOF_MMWORD
+
+    movq        mm3, mm2
+    punpcklbw   mm2, mm2
+    punpckhbw   mm3, mm3
+
+    movq        MMWORD ebx+2*SIZEOF_MMWORD, mm2
+    movq        MMWORD ebx+3*SIZEOF_MMWORD, mm3
+    movq        MMWORD edi+2*SIZEOF_MMWORD, mm2
+    movq        MMWORD edi+3*SIZEOF_MMWORD, mm3
+
+    sub         eax, byte 2*SIZEOF_MMWORD
+    jz          short .nextrow
+
+    add         esi, byte 2*SIZEOF_MMWORD  ; inptr
+    add         ebx, byte 4*SIZEOF_MMWORD  ; outptr0
+    add         edi, byte 4*SIZEOF_MMWORD  ; outptr1
+    jmp         short .columnloop
+    alignx      16, 7
+
+.nextrow:
+    pop         esi
+    pop         edi
+
+    add         esi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         ecx, byte 2                  ; rowctr
+    jg          short .rowloop
+
+    emms                                ; empty MMX state
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jdsample-sse2.asm Added

@@ -0,0 +1,724 @@
+;
+; jdsample.asm - upsampling (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fancy_upsample_sse2)
+
+EXTN(jconst_fancy_upsample_sse2):
+
+PW_ONE   times 8 dw 1
+PW_TWO   times 8 dw 2
+PW_THREE times 8 dw 3
+PW_SEVEN times 8 dw 7
+PW_EIGHT times 8 dw 8
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_sse2(int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define downsamp_width(b)   (b) + 12    ; JDIMENSION downsampled_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_fancy_upsample_sse2)
+
+EXTN(jsimd_h2v1_fancy_upsample_sse2):
+    push        ebp
+    mov         ebp, esp
+    pushpic     ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    mov         eax, JDIMENSION downsamp_width(ebp)  ; colctr
+    test        eax, eax
+    jz          near .return
+
+    mov         ecx, INT max_v_samp(ebp)  ; rowctr
+    test        ecx, ecx
+    jz          near .return
+
+    mov         esi, JSAMPARRAY input_data(ebp)    ; input_data
+    mov         edi, POINTER output_data_ptr(ebp)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        eax                     ; colctr
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr
+
+    test        eax, SIZEOF_XMMWORD-1
+    jz          short .skip
+    mov         dl, JSAMPLE esi+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE esi+eax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+.skip:
+    pxor        xmm0, xmm0              ; xmm0=(all 0's)
+    pcmpeqb     xmm7, xmm7
+    psrldq      xmm7, (SIZEOF_XMMWORD-1)
+    pand        xmm7, XMMWORD esi+0*SIZEOF_XMMWORD
+
+    add         eax, byte SIZEOF_XMMWORD-1
+    and         eax, byte -SIZEOF_XMMWORD
+    cmp         eax, byte SIZEOF_XMMWORD
+    ja          short .columnloop
+    alignx      16, 7
+
+.columnloop_last:
+    pcmpeqb     xmm6, xmm6
+    pslldq      xmm6, (SIZEOF_XMMWORD-1)
+    pand        xmm6, XMMWORD esi+0*SIZEOF_XMMWORD
+    jmp         short .upsample
+    alignx      16, 7
+
+.columnloop:
+    movdqa      xmm6, XMMWORD esi+1*SIZEOF_XMMWORD
+    pslldq      xmm6, (SIZEOF_XMMWORD-1)
+
+.upsample:
+    movdqa      xmm1, XMMWORD esi+0*SIZEOF_XMMWORD
+    movdqa      xmm2, xmm1
+    movdqa      xmm3, xmm1                ; xmm1=( 0  1  2 ... 13 14 15)
+    pslldq      xmm2, 1                   ; xmm2=(--  0  1 ... 12 13 14)
+    psrldq      xmm3, 1                   ; xmm3=( 1  2  3 ... 14 15 --)
+
+    por         xmm2, xmm7                ; xmm2=(-1  0  1 ... 12 13 14)
+    por         xmm3, xmm6                ; xmm3=( 1  2  3 ... 14 15 16)
+
+    movdqa      xmm7, xmm1
+    psrldq      xmm7, (SIZEOF_XMMWORD-1)  ; xmm7=(15 -- -- ... -- -- --)
+
+    movdqa      xmm4, xmm1
+    punpcklbw   xmm1, xmm0                ; xmm1=( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm4, xmm0                ; xmm4=( 8  9 10 11 12 13 14 15)
+    movdqa      xmm5, xmm2
+    punpcklbw   xmm2, xmm0                ; xmm2=(-1  0  1  2  3  4  5  6)
+    punpckhbw   xmm5, xmm0                ; xmm5=( 7  8  9 10 11 12 13 14)
+    movdqa      xmm6, xmm3
+    punpcklbw   xmm3, xmm0                ; xmm3=( 1  2  3  4  5  6  7  8)
+    punpckhbw   xmm6, xmm0                ; xmm6=( 9 10 11 12 13 14 15 16)
+
+    pmullw      xmm1, GOTOFF(ebx,PW_THREE)
+    pmullw      xmm4, GOTOFF(ebx,PW_THREE)
+    paddw       xmm2, GOTOFF(ebx,PW_ONE)
+    paddw       xmm5, GOTOFF(ebx,PW_ONE)
+    paddw       xmm3, GOTOFF(ebx,PW_TWO)
+    paddw       xmm6, GOTOFF(ebx,PW_TWO)
+
+    paddw       xmm2, xmm1
+    paddw       xmm5, xmm4
+    psrlw       xmm2, 2                 ; xmm2=OutLE=( 0  2  4  6  8 10 12 14)
+    psrlw       xmm5, 2                 ; xmm5=OutHE=(16 18 20 22 24 26 28 30)
+    paddw       xmm3, xmm1
+    paddw       xmm6, xmm4
+    psrlw       xmm3, 2                 ; xmm3=OutLO=( 1  3  5  7  9 11 13 15)
+    psrlw       xmm6, 2                 ; xmm6=OutHO=(17 19 21 23 25 27 29 31)
+
+    psllw       xmm3, BYTE_BIT
+    psllw       xmm6, BYTE_BIT
+    por         xmm2, xmm3              ; xmm2=OutL=( 0  1  2 ... 13 14 15)
+    por         xmm5, xmm6              ; xmm5=OutH=(16 17 18 ... 29 30 31)
+
+    movdqa      XMMWORD edi+0*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD edi+1*SIZEOF_XMMWORD, xmm5
+
+    sub         eax, byte SIZEOF_XMMWORD
+    add         esi, byte 1*SIZEOF_XMMWORD  ; inptr
+    add         edi, byte 2*SIZEOF_XMMWORD  ; outptr
+    cmp         eax, byte SIZEOF_XMMWORD
+    ja          near .columnloop
+    test        eax, eax
+    jnz         near .columnloop_last
+
+    pop         esi
+    pop         edi
+    pop         eax
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         ecx                        ; rowctr
+    jg          near .rowloop
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    poppic      ebx
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_sse2(int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define downsamp_width(b)   (b) + 12    ; JDIMENSION downsampled_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM        4
+%define gotptr        wk(0) - SIZEOF_POINTER  ; void *gotptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_fancy_upsample_sse2)
+
+EXTN(jsimd_h2v2_fancy_upsample_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     eax                     ; make a room for GOT address
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+    movpic      POINTER gotptr, ebx   ; save GOT address
+
+    mov         edx, eax                ; edx = original ebp
+    mov         eax, JDIMENSION downsamp_width(edx)  ; colctr
+    test        eax, eax
+    jz          near .return
+
+    mov         ecx, INT max_v_samp(edx)  ; rowctr
+    test        ecx, ecx
+    jz          near .return
+
+    mov         esi, JSAMPARRAY input_data(edx)    ; input_data
+    mov         edi, POINTER output_data_ptr(edx)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        eax                     ; colctr
+    push        ecx
+    push        edi
+    push        esi
+
+    mov         ecx, JSAMPROW esi-1*SIZEOF_JSAMPROW  ; inptr1(above)
+    mov         ebx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         esi, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; inptr1(below)
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW  ; outptr0
+    mov         edi, JSAMPROW edi+1*SIZEOF_JSAMPROW  ; outptr1
+
+    test        eax, SIZEOF_XMMWORD-1
+    jz          short .skip
+    push        edx
+    mov         dl, JSAMPLE ecx+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE ecx+eax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE ebx+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE ebx+eax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE esi+(eax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE esi+eax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+    pop         edx
+.skip:
+    ; -- process the first column block
+
+    movdqa      xmm0, XMMWORD ebx+0*SIZEOF_XMMWORD  ; xmm0=row 00
+    movdqa      xmm1, XMMWORD ecx+0*SIZEOF_XMMWORD  ; xmm1=row-10
+    movdqa      xmm2, XMMWORD esi+0*SIZEOF_XMMWORD  ; xmm2=row+10
+
+    pushpic     ebx
+    movpic      ebx, POINTER gotptr   ; load GOT address
+
+    pxor        xmm3, xmm3              ; xmm3=(all 0's)
+    movdqa      xmm4, xmm0
+    punpcklbw   xmm0, xmm3              ; xmm0=row 0( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm4, xmm3              ; xmm4=row 0( 8  9 10 11 12 13 14 15)
+    movdqa      xmm5, xmm1
+    punpcklbw   xmm1, xmm3              ; xmm1=row-1( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm5, xmm3              ; xmm5=row-1( 8  9 10 11 12 13 14 15)
+    movdqa      xmm6, xmm2
+    punpcklbw   xmm2, xmm3              ; xmm2=row+1( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm6, xmm3              ; xmm6=row+1( 8  9 10 11 12 13 14 15)
+
+    pmullw      xmm0, GOTOFF(ebx,PW_THREE)
+    pmullw      xmm4, GOTOFF(ebx,PW_THREE)
+
+    pcmpeqb     xmm7, xmm7
+    psrldq      xmm7, (SIZEOF_XMMWORD-2)
+
+    paddw       xmm1, xmm0              ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+    paddw       xmm5, xmm4              ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+    paddw       xmm2, xmm0              ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+    paddw       xmm6, xmm4              ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+    movdqa      XMMWORD edx+0*SIZEOF_XMMWORD, xmm1  ; temporarily save
+    movdqa      XMMWORD edx+1*SIZEOF_XMMWORD, xmm5  ; the intermediate data
+    movdqa      XMMWORD edi+0*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD edi+1*SIZEOF_XMMWORD, xmm6
+
+    pand        xmm1, xmm7              ; xmm1=( 0 -- -- -- -- -- -- --)
+    pand        xmm2, xmm7              ; xmm2=( 0 -- -- -- -- -- -- --)
+
+    movdqa      XMMWORD wk(0), xmm1
+    movdqa      XMMWORD wk(1), xmm2
+
+    poppic      ebx
+
+    add         eax, byte SIZEOF_XMMWORD-1
+    and         eax, byte -SIZEOF_XMMWORD
+    cmp         eax, byte SIZEOF_XMMWORD
+    ja          short .columnloop
+    alignx      16, 7
+
+.columnloop_last:
+    ; -- process the last column block
+
+    pushpic     ebx
+    movpic      ebx, POINTER gotptr   ; load GOT address
+
+    pcmpeqb     xmm1, xmm1
+    pslldq      xmm1, (SIZEOF_XMMWORD-2)
+    movdqa      xmm2, xmm1
+
+    pand        xmm1, XMMWORD edx+1*SIZEOF_XMMWORD
+    pand        xmm2, XMMWORD edi+1*SIZEOF_XMMWORD
+
+    movdqa      XMMWORD wk(2), xmm1   ; xmm1=(-- -- -- -- -- -- -- 15)
+    movdqa      XMMWORD wk(3), xmm2   ; xmm2=(-- -- -- -- -- -- -- 15)
+
+    jmp         near .upsample
+    alignx      16, 7
+
+.columnloop:
+    ; -- process the next column block
+
+    movdqa      xmm0, XMMWORD ebx+1*SIZEOF_XMMWORD  ; xmm0=row 01
+    movdqa      xmm1, XMMWORD ecx+1*SIZEOF_XMMWORD  ; xmm1=row-11
+    movdqa      xmm2, XMMWORD esi+1*SIZEOF_XMMWORD  ; xmm2=row+11
+
+    pushpic     ebx
+    movpic      ebx, POINTER gotptr   ; load GOT address
+
+    pxor        xmm3, xmm3              ; xmm3=(all 0's)
+    movdqa      xmm4, xmm0
+    punpcklbw   xmm0, xmm3              ; xmm0=row 0( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm4, xmm3              ; xmm4=row 0( 8  9 10 11 12 13 14 15)
+    movdqa      xmm5, xmm1
+    punpcklbw   xmm1, xmm3              ; xmm1=row-1( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm5, xmm3              ; xmm5=row-1( 8  9 10 11 12 13 14 15)
+    movdqa      xmm6, xmm2
+    punpcklbw   xmm2, xmm3              ; xmm2=row+1( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm6, xmm3              ; xmm6=row+1( 8  9 10 11 12 13 14 15)
+
+    pmullw      xmm0, GOTOFF(ebx,PW_THREE)
+    pmullw      xmm4, GOTOFF(ebx,PW_THREE)
+
+    paddw       xmm1, xmm0              ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+    paddw       xmm5, xmm4              ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+    paddw       xmm2, xmm0              ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+    paddw       xmm6, xmm4              ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+    movdqa      XMMWORD edx+2*SIZEOF_XMMWORD, xmm1  ; temporarily save
+    movdqa      XMMWORD edx+3*SIZEOF_XMMWORD, xmm5  ; the intermediate data
+    movdqa      XMMWORD edi+2*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD edi+3*SIZEOF_XMMWORD, xmm6
+
+    pslldq      xmm1, (SIZEOF_XMMWORD-2)  ; xmm1=(-- -- -- -- -- -- --  0)
+    pslldq      xmm2, (SIZEOF_XMMWORD-2)  ; xmm2=(-- -- -- -- -- -- --  0)
+
+    movdqa      XMMWORD wk(2), xmm1
+    movdqa      XMMWORD wk(3), xmm2
+
+.upsample:
+    ; -- process the upper row
+
+    movdqa      xmm7, XMMWORD edx+0*SIZEOF_XMMWORD
+    movdqa      xmm3, XMMWORD edx+1*SIZEOF_XMMWORD
+
+    movdqa      xmm0, xmm7                ; xmm7=Int0L=( 0  1  2  3  4  5  6  7)
+    movdqa      xmm4, xmm3                ; xmm3=Int0H=( 8  9 10 11 12 13 14 15)
+    psrldq      xmm0, 2                   ; xmm0=( 1  2  3  4  5  6  7 --)
+    pslldq      xmm4, (SIZEOF_XMMWORD-2)  ; xmm4=(-- -- -- -- -- -- --  8)
+    movdqa      xmm5, xmm7
+    movdqa      xmm6, xmm3
+    psrldq      xmm5, (SIZEOF_XMMWORD-2)  ; xmm5=( 7 -- -- -- -- -- -- --)
+    pslldq      xmm6, 2                   ; xmm6=(--  8  9 10 11 12 13 14)
+
+    por         xmm0, xmm4                ; xmm0=( 1  2  3  4  5  6  7  8)
+    por         xmm5, xmm6                ; xmm5=( 7  8  9 10 11 12 13 14)
+
+    movdqa      xmm1, xmm7
+    movdqa      xmm2, xmm3
+    pslldq      xmm1, 2                   ; xmm1=(--  0  1  2  3  4  5  6)
+    psrldq      xmm2, 2                   ; xmm2=( 9 10 11 12 13 14 15 --)
+    movdqa      xmm4, xmm3
+    psrldq      xmm4, (SIZEOF_XMMWORD-2)  ; xmm4=(15 -- -- -- -- -- -- --)
+
+    por         xmm1, XMMWORD wk(0)     ; xmm1=(-1  0  1  2  3  4  5  6)
+    por         xmm2, XMMWORD wk(2)     ; xmm2=( 9 10 11 12 13 14 15 16)
+
+    movdqa      XMMWORD wk(0), xmm4
+
+    pmullw      xmm7, GOTOFF(ebx,PW_THREE)
+    pmullw      xmm3, GOTOFF(ebx,PW_THREE)
+    paddw       xmm1, GOTOFF(ebx,PW_EIGHT)
+    paddw       xmm5, GOTOFF(ebx,PW_EIGHT)
+    paddw       xmm0, GOTOFF(ebx,PW_SEVEN)
+    paddw       xmm2, GOTOFF(ebx,PW_SEVEN)
+
+    paddw       xmm1, xmm7
+    paddw       xmm5, xmm3
+    psrlw       xmm1, 4                 ; xmm1=Out0LE=( 0  2  4  6  8 10 12 14)
+    psrlw       xmm5, 4                 ; xmm5=Out0HE=(16 18 20 22 24 26 28 30)
+    paddw       xmm0, xmm7
+    paddw       xmm2, xmm3
+    psrlw       xmm0, 4                 ; xmm0=Out0LO=( 1  3  5  7  9 11 13 15)
+    psrlw       xmm2, 4                 ; xmm2=Out0HO=(17 19 21 23 25 27 29 31)
+
+    psllw       xmm0, BYTE_BIT
+    psllw       xmm2, BYTE_BIT
+    por         xmm1, xmm0              ; xmm1=Out0L=( 0  1  2 ... 13 14 15)
+    por         xmm5, xmm2              ; xmm5=Out0H=(16 17 18 ... 29 30 31)
+
+    movdqa      XMMWORD edx+0*SIZEOF_XMMWORD, xmm1
+    movdqa      XMMWORD edx+1*SIZEOF_XMMWORD, xmm5
+
+    ; -- process the lower row
+
+    movdqa      xmm6, XMMWORD edi+0*SIZEOF_XMMWORD
+    movdqa      xmm4, XMMWORD edi+1*SIZEOF_XMMWORD
+
+    movdqa      xmm7, xmm6                ; xmm6=Int1L=( 0  1  2  3  4  5  6  7)
+    movdqa      xmm3, xmm4                ; xmm4=Int1H=( 8  9 10 11 12 13 14 15)
+    psrldq      xmm7, 2                   ; xmm7=( 1  2  3  4  5  6  7 --)
+    pslldq      xmm3, (SIZEOF_XMMWORD-2)  ; xmm3=(-- -- -- -- -- -- --  8)
+    movdqa      xmm0, xmm6
+    movdqa      xmm2, xmm4
+    psrldq      xmm0, (SIZEOF_XMMWORD-2)  ; xmm0=( 7 -- -- -- -- -- -- --)
+    pslldq      xmm2, 2                   ; xmm2=(--  8  9 10 11 12 13 14)
+
+    por         xmm7, xmm3                ; xmm7=( 1  2  3  4  5  6  7  8)
+    por         xmm0, xmm2                ; xmm0=( 7  8  9 10 11 12 13 14)
+
+    movdqa      xmm1, xmm6
+    movdqa      xmm5, xmm4
+    pslldq      xmm1, 2                   ; xmm1=(--  0  1  2  3  4  5  6)
+    psrldq      xmm5, 2                   ; xmm5=( 9 10 11 12 13 14 15 --)
+    movdqa      xmm3, xmm4
+    psrldq      xmm3, (SIZEOF_XMMWORD-2)  ; xmm3=(15 -- -- -- -- -- -- --)
+
+    por         xmm1, XMMWORD wk(1)     ; xmm1=(-1  0  1  2  3  4  5  6)
+    por         xmm5, XMMWORD wk(3)     ; xmm5=( 9 10 11 12 13 14 15 16)
+
+    movdqa      XMMWORD wk(1), xmm3
+
+    pmullw      xmm6, GOTOFF(ebx,PW_THREE)
+    pmullw      xmm4, GOTOFF(ebx,PW_THREE)
+    paddw       xmm1, GOTOFF(ebx,PW_EIGHT)
+    paddw       xmm0, GOTOFF(ebx,PW_EIGHT)
+    paddw       xmm7, GOTOFF(ebx,PW_SEVEN)
+    paddw       xmm5, GOTOFF(ebx,PW_SEVEN)
+
+    paddw       xmm1, xmm6
+    paddw       xmm0, xmm4
+    psrlw       xmm1, 4                 ; xmm1=Out1LE=( 0  2  4  6  8 10 12 14)
+    psrlw       xmm0, 4                 ; xmm0=Out1HE=(16 18 20 22 24 26 28 30)
+    paddw       xmm7, xmm6
+    paddw       xmm5, xmm4
+    psrlw       xmm7, 4                 ; xmm7=Out1LO=( 1  3  5  7  9 11 13 15)
+    psrlw       xmm5, 4                 ; xmm5=Out1HO=(17 19 21 23 25 27 29 31)
+
+    psllw       xmm7, BYTE_BIT
+    psllw       xmm5, BYTE_BIT
+    por         xmm1, xmm7              ; xmm1=Out1L=( 0  1  2 ... 13 14 15)
+    por         xmm0, xmm5              ; xmm0=Out1H=(16 17 18 ... 29 30 31)
+
+    movdqa      XMMWORD edi+0*SIZEOF_XMMWORD, xmm1
+    movdqa      XMMWORD edi+1*SIZEOF_XMMWORD, xmm0
+
+    poppic      ebx
+
+    sub         eax, byte SIZEOF_XMMWORD
+    add         ecx, byte 1*SIZEOF_XMMWORD  ; inptr1(above)
+    add         ebx, byte 1*SIZEOF_XMMWORD  ; inptr0
+    add         esi, byte 1*SIZEOF_XMMWORD  ; inptr1(below)
+    add         edx, byte 2*SIZEOF_XMMWORD  ; outptr0
+    add         edi, byte 2*SIZEOF_XMMWORD  ; outptr1
+    cmp         eax, byte SIZEOF_XMMWORD
+    ja          near .columnloop
+    test        eax, eax
+    jnz         near .columnloop_last
+
+    pop         esi
+    pop         edi
+    pop         ecx
+    pop         eax
+
+    add         esi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         ecx, byte 2                  ; rowctr
+    jg          near .rowloop
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_sse2(int max_v_samp_factor, JDIMENSION output_width,
+;                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define output_width(b)     (b) + 12    ; JDIMENSION output_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_upsample_sse2)
+
+EXTN(jsimd_h2v1_upsample_sse2):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         edx, JDIMENSION output_width(ebp)
+    add         edx, byte (2*SIZEOF_XMMWORD)-1
+    and         edx, byte -(2*SIZEOF_XMMWORD)
+    jz          short .return
+
+    mov         ecx, INT max_v_samp(ebp)  ; rowctr
+    test        ecx, ecx
+    jz          short .return
+
+    mov         esi, JSAMPARRAY input_data(ebp)    ; input_data
+    mov         edi, POINTER output_data_ptr(ebp)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi     ; inptr
+    mov         edi, JSAMPROW edi     ; outptr
+    mov         eax, edx                ; colctr
+    alignx      16, 7
+.columnloop:
+
+    movdqa      xmm0, XMMWORD esi+0*SIZEOF_XMMWORD
+
+    movdqa      xmm1, xmm0
+    punpcklbw   xmm0, xmm0
+    punpckhbw   xmm1, xmm1
+
+    movdqa      XMMWORD edi+0*SIZEOF_XMMWORD, xmm0
+    movdqa      XMMWORD edi+1*SIZEOF_XMMWORD, xmm1
+
+    sub         eax, byte 2*SIZEOF_XMMWORD
+    jz          short .nextrow
+
+    movdqa      xmm2, XMMWORD esi+1*SIZEOF_XMMWORD
+
+    movdqa      xmm3, xmm2
+    punpcklbw   xmm2, xmm2
+    punpckhbw   xmm3, xmm3
+
+    movdqa      XMMWORD edi+2*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD edi+3*SIZEOF_XMMWORD, xmm3
+
+    sub         eax, byte 2*SIZEOF_XMMWORD
+    jz          short .nextrow
+
+    add         esi, byte 2*SIZEOF_XMMWORD  ; inptr
+    add         edi, byte 4*SIZEOF_XMMWORD  ; outptr
+    jmp         short .columnloop
+    alignx      16, 7
+
+.nextrow:
+    pop         esi
+    pop         edi
+
+    add         esi, byte SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         ecx                        ; rowctr
+    jg          short .rowloop
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_sse2(int max_v_samp_factor, JDIMENSION output_width,
+;                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)       (b) + 8     ; int max_v_samp_factor
+%define output_width(b)     (b) + 12    ; JDIMENSION output_width
+%define input_data(b)       (b) + 16    ; JSAMPARRAY input_data
+%define output_data_ptr(b)  (b) + 20    ; JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_upsample_sse2)
+
+EXTN(jsimd_h2v2_upsample_sse2):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         edx, JDIMENSION output_width(ebp)
+    add         edx, byte (2*SIZEOF_XMMWORD)-1
+    and         edx, byte -(2*SIZEOF_XMMWORD)
+    jz          near .return
+
+    mov         ecx, INT max_v_samp(ebp)  ; rowctr
+    test        ecx, ecx
+    jz          near .return
+
+    mov         esi, JSAMPARRAY input_data(ebp)    ; input_data
+    mov         edi, POINTER output_data_ptr(ebp)
+    mov         edi, JSAMPARRAY edi                ; output_data
+    alignx      16, 7
+.rowloop:
+    push        edi
+    push        esi
+
+    mov         esi, JSAMPROW esi                    ; inptr
+    mov         ebx, JSAMPROW edi+0*SIZEOF_JSAMPROW  ; outptr0
+    mov         edi, JSAMPROW edi+1*SIZEOF_JSAMPROW  ; outptr1
+    mov         eax, edx                               ; colctr
+    alignx      16, 7
+.columnloop:
+
+    movdqa      xmm0, XMMWORD esi+0*SIZEOF_XMMWORD
+
+    movdqa      xmm1, xmm0
+    punpcklbw   xmm0, xmm0
+    punpckhbw   xmm1, xmm1
+
+    movdqa      XMMWORD ebx+0*SIZEOF_XMMWORD, xmm0
+    movdqa      XMMWORD ebx+1*SIZEOF_XMMWORD, xmm1
+    movdqa      XMMWORD edi+0*SIZEOF_XMMWORD, xmm0
+    movdqa      XMMWORD edi+1*SIZEOF_XMMWORD, xmm1
+
+    sub         eax, byte 2*SIZEOF_XMMWORD
+    jz          short .nextrow
+
+    movdqa      xmm2, XMMWORD esi+1*SIZEOF_XMMWORD
+
+    movdqa      xmm3, xmm2
+    punpcklbw   xmm2, xmm2
+    punpckhbw   xmm3, xmm3
+
+    movdqa      XMMWORD ebx+2*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD ebx+3*SIZEOF_XMMWORD, xmm3
+    movdqa      XMMWORD edi+2*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD edi+3*SIZEOF_XMMWORD, xmm3
+
+    sub         eax, byte 2*SIZEOF_XMMWORD
+    jz          short .nextrow
+
+    add         esi, byte 2*SIZEOF_XMMWORD  ; inptr
+    add         ebx, byte 4*SIZEOF_XMMWORD  ; outptr0
+    add         edi, byte 4*SIZEOF_XMMWORD  ; outptr1
+    jmp         short .columnloop
+    alignx      16, 7
+
+.nextrow:
+    pop         esi
+    pop         edi
+
+    add         esi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         edi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         ecx, byte 2                  ; rowctr
+    jg          short .rowloop
+
+.return:
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jfdctflt-3dn.asm Added

@@ -0,0 +1,318 @@
+;
+; jfdctflt.asm - floating-point FDCT (3DNow!)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the forward DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jfdctflt.c; see the jfdctflt.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_float_3dnow)
+
+EXTN(jconst_fdct_float_3dnow):
+
+PD_0_382 times 2 dd 0.382683432365089771728460
+PD_0_707 times 2 dd 0.707106781186547524400844
+PD_0_541 times 2 dd 0.541196100146196984399723
+PD_1_306 times 2 dd 1.306562964876376527856643
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_float_3dnow(FAST_FLOAT *data)
+;
+
+%define data(b)       (b) + 8           ; FAST_FLOAT *data
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_MMWORD  ; mmword wkWK_NUM
+%define WK_NUM        2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_float_3dnow)
+
+EXTN(jsimd_fdct_float_3dnow):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+;   push        esi                     ; unused
+;   push        edi                     ; unused
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process rows.
+
+    mov         edx, POINTER data(eax)  ; (FAST_FLOAT *)
+    mov         ecx, DCTSIZE/2
+    alignx      16, 7
+.rowloop:
+
+    movq        mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)
+    movq        mm1, MMWORD MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)
+    movq        mm2, MMWORD MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)
+    movq        mm3, MMWORD MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)
+
+    ; mm0=(00 01), mm1=(10 11), mm2=(06 07), mm3=(16 17)
+
+    movq        mm4, mm0                ; transpose coefficients
+    punpckldq   mm0, mm1                ; mm0=(00 10)=data0
+    punpckhdq   mm4, mm1                ; mm4=(01 11)=data1
+    movq        mm5, mm2                ; transpose coefficients
+    punpckldq   mm2, mm3                ; mm2=(06 16)=data6
+    punpckhdq   mm5, mm3                ; mm5=(07 17)=data7
+
+    movq        mm6, mm4
+    movq        mm7, mm0
+    pfsub       mm4, mm2                ; mm4=data1-data6=tmp6
+    pfsub       mm0, mm5                ; mm0=data0-data7=tmp7
+    pfadd       mm6, mm2                ; mm6=data1+data6=tmp1
+    pfadd       mm7, mm5                ; mm7=data0+data7=tmp0
+
+    movq        mm1, MMWORD MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)
+    movq        mm3, MMWORD MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)
+    movq        mm2, MMWORD MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)
+    movq        mm5, MMWORD MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)
+
+    ; mm1=(02 03), mm3=(12 13), mm2=(04 05), mm5=(14 15)
+
+    movq        MMWORD wk(0), mm4     ; wk(0)=tmp6
+    movq        MMWORD wk(1), mm0     ; wk(1)=tmp7
+
+    movq        mm4, mm1                ; transpose coefficients
+    punpckldq   mm1, mm3                ; mm1=(02 12)=data2
+    punpckhdq   mm4, mm3                ; mm4=(03 13)=data3
+    movq        mm0, mm2                ; transpose coefficients
+    punpckldq   mm2, mm5                ; mm2=(04 14)=data4
+    punpckhdq   mm0, mm5                ; mm0=(05 15)=data5
+
+    movq        mm3, mm4
+    movq        mm5, mm1
+    pfadd       mm4, mm2                ; mm4=data3+data4=tmp3
+    pfadd       mm1, mm0                ; mm1=data2+data5=tmp2
+    pfsub       mm3, mm2                ; mm3=data3-data4=tmp4
+    pfsub       mm5, mm0                ; mm5=data2-data5=tmp5
+
+    ; -- Even part
+
+    movq        mm2, mm7
+    movq        mm0, mm6
+    pfsub       mm7, mm4                ; mm7=tmp13
+    pfsub       mm6, mm1                ; mm6=tmp12
+    pfadd       mm2, mm4                ; mm2=tmp10
+    pfadd       mm0, mm1                ; mm0=tmp11
+
+    pfadd       mm6, mm7
+    pfmul       mm6, GOTOFF(ebx,PD_0_707)  ; mm6=z1
+
+    movq        mm4, mm2
+    movq        mm1, mm7
+    pfsub       mm2, mm0                ; mm2=data4
+    pfsub       mm7, mm6                ; mm7=data6
+    pfadd       mm4, mm0                ; mm4=data0
+    pfadd       mm1, mm6                ; mm1=data2
+
+    movq        MMWORD MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT), mm2
+    movq        MMWORD MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT), mm7
+    movq        MMWORD MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT), mm4
+    movq        MMWORD MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT), mm1
+
+    ; -- Odd part
+
+    movq        mm0, MMWORD wk(0)     ; mm0=tmp6
+    movq        mm6, MMWORD wk(1)     ; mm6=tmp7
+
+    pfadd       mm3, mm5                ; mm3=tmp10
+    pfadd       mm5, mm0                ; mm5=tmp11
+    pfadd       mm0, mm6                ; mm0=tmp12, mm6=tmp7
+
+    pfmul       mm5, GOTOFF(ebx,PD_0_707)  ; mm5=z3
+
+    movq        mm2, mm3                     ; mm2=tmp10
+    pfsub       mm3, mm0
+    pfmul       mm3, GOTOFF(ebx,PD_0_382)  ; mm3=z5
+    pfmul       mm2, GOTOFF(ebx,PD_0_541)  ; mm2=MULTIPLY(tmp10,FIX_0_54119610)
+    pfmul       mm0, GOTOFF(ebx,PD_1_306)  ; mm0=MULTIPLY(tmp12,FIX_1_30656296)
+    pfadd       mm2, mm3                     ; mm2=z2
+    pfadd       mm0, mm3                     ; mm0=z4
+
+    movq        mm7, mm6
+    pfsub       mm6, mm5                ; mm6=z13
+    pfadd       mm7, mm5                ; mm7=z11
+
+    movq        mm4, mm6
+    movq        mm1, mm7
+    pfsub       mm6, mm2                ; mm6=data3
+    pfsub       mm7, mm0                ; mm7=data7
+    pfadd       mm4, mm2                ; mm4=data5
+    pfadd       mm1, mm0                ; mm1=data1
+
+    movq        MMWORD MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT), mm6
+    movq        MMWORD MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT), mm7
+    movq        MMWORD MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT), mm4
+    movq        MMWORD MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT), mm1
+
+    add         edx, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+    dec         ecx
+    jnz         near .rowloop
+
+    ; ---- Pass 2: process columns.
+
+    mov         edx, POINTER data(eax)  ; (FAST_FLOAT *)
+    mov         ecx, DCTSIZE/2
+    alignx      16, 7
+.columnloop:
+
+    movq        mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)
+    movq        mm1, MMWORD MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)
+    movq        mm2, MMWORD MMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)
+    movq        mm3, MMWORD MMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)
+
+    ; mm0=(00 10), mm1=(01 11), mm2=(60 70), mm3=(61 71)
+
+    movq        mm4, mm0                ; transpose coefficients
+    punpckldq   mm0, mm1                ; mm0=(00 01)=data0
+    punpckhdq   mm4, mm1                ; mm4=(10 11)=data1
+    movq        mm5, mm2                ; transpose coefficients
+    punpckldq   mm2, mm3                ; mm2=(60 61)=data6
+    punpckhdq   mm5, mm3                ; mm5=(70 71)=data7
+
+    movq        mm6, mm4
+    movq        mm7, mm0
+    pfsub       mm4, mm2                ; mm4=data1-data6=tmp6
+    pfsub       mm0, mm5                ; mm0=data0-data7=tmp7
+    pfadd       mm6, mm2                ; mm6=data1+data6=tmp1
+    pfadd       mm7, mm5                ; mm7=data0+data7=tmp0
+
+    movq        mm1, MMWORD MMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)
+    movq        mm3, MMWORD MMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)
+    movq        mm2, MMWORD MMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)
+    movq        mm5, MMWORD MMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)
+
+    ; mm1=(20 30), mm3=(21 31), mm2=(40 50), mm5=(41 51)
+
+    movq        MMWORD wk(0), mm4     ; wk(0)=tmp6
+    movq        MMWORD wk(1), mm0     ; wk(1)=tmp7
+
+    movq        mm4, mm1                ; transpose coefficients
+    punpckldq   mm1, mm3                ; mm1=(20 21)=data2
+    punpckhdq   mm4, mm3                ; mm4=(30 31)=data3
+    movq        mm0, mm2                ; transpose coefficients
+    punpckldq   mm2, mm5                ; mm2=(40 41)=data4
+    punpckhdq   mm0, mm5                ; mm0=(50 51)=data5
+
+    movq        mm3, mm4
+    movq        mm5, mm1
+    pfadd       mm4, mm2                ; mm4=data3+data4=tmp3
+    pfadd       mm1, mm0                ; mm1=data2+data5=tmp2
+    pfsub       mm3, mm2                ; mm3=data3-data4=tmp4
+    pfsub       mm5, mm0                ; mm5=data2-data5=tmp5
+
+    ; -- Even part
+
+    movq        mm2, mm7
+    movq        mm0, mm6
+    pfsub       mm7, mm4                ; mm7=tmp13
+    pfsub       mm6, mm1                ; mm6=tmp12
+    pfadd       mm2, mm4                ; mm2=tmp10
+    pfadd       mm0, mm1                ; mm0=tmp11
+
+    pfadd       mm6, mm7
+    pfmul       mm6, GOTOFF(ebx,PD_0_707)  ; mm6=z1
+
+    movq        mm4, mm2
+    movq        mm1, mm7
+    pfsub       mm2, mm0                ; mm2=data4
+    pfsub       mm7, mm6                ; mm7=data6
+    pfadd       mm4, mm0                ; mm4=data0
+    pfadd       mm1, mm6                ; mm1=data2
+
+    movq        MMWORD MMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT), mm2
+    movq        MMWORD MMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT), mm7
+    movq        MMWORD MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT), mm4
+    movq        MMWORD MMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT), mm1
+
+    ; -- Odd part
+
+    movq        mm0, MMWORD wk(0)     ; mm0=tmp6
+    movq        mm6, MMWORD wk(1)     ; mm6=tmp7
+
+    pfadd       mm3, mm5                ; mm3=tmp10
+    pfadd       mm5, mm0                ; mm5=tmp11
+    pfadd       mm0, mm6                ; mm0=tmp12, mm6=tmp7
+
+    pfmul       mm5, GOTOFF(ebx,PD_0_707)  ; mm5=z3
+
+    movq        mm2, mm3                     ; mm2=tmp10
+    pfsub       mm3, mm0
+    pfmul       mm3, GOTOFF(ebx,PD_0_382)  ; mm3=z5
+    pfmul       mm2, GOTOFF(ebx,PD_0_541)  ; mm2=MULTIPLY(tmp10,FIX_0_54119610)
+    pfmul       mm0, GOTOFF(ebx,PD_1_306)  ; mm0=MULTIPLY(tmp12,FIX_1_30656296)
+    pfadd       mm2, mm3                     ; mm2=z2
+    pfadd       mm0, mm3                     ; mm0=z4
+
+    movq        mm7, mm6
+    pfsub       mm6, mm5                ; mm6=z13
+    pfadd       mm7, mm5                ; mm7=z11
+
+    movq        mm4, mm6
+    movq        mm1, mm7
+    pfsub       mm6, mm2                ; mm6=data3
+    pfsub       mm7, mm0                ; mm7=data7
+    pfadd       mm4, mm2                ; mm4=data5
+    pfadd       mm1, mm0                ; mm1=data1
+
+    movq        MMWORD MMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT), mm6
+    movq        MMWORD MMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT), mm7
+    movq        MMWORD MMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT), mm4
+    movq        MMWORD MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT), mm1
+
+    add         edx, byte 2*SIZEOF_FAST_FLOAT
+    dec         ecx
+    jnz         near .columnloop
+
+    femms                               ; empty MMX/3DNow! state
+
+;   pop         edi                     ; unused
+;   pop         esi                     ; unused
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jfdctflt-sse.asm Added

@@ -0,0 +1,369 @@
+;
+; jfdctflt.asm - floating-point FDCT (SSE)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the forward DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jfdctflt.c; see the jfdctflt.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+    shufps      %1, %2, 0x44
+%endmacro
+
+%macro  unpckhps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+    shufps      %1, %2, 0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_float_sse)
+
+EXTN(jconst_fdct_float_sse):
+
+PD_0_382 times 4 dd 0.382683432365089771728460
+PD_0_707 times 4 dd 0.707106781186547524400844
+PD_0_541 times 4 dd 0.541196100146196984399723
+PD_1_306 times 4 dd 1.306562964876376527856643
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_float_sse(FAST_FLOAT *data)
+;
+
+%define data(b)       (b) + 8           ; FAST_FLOAT *data
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM        2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_float_sse)
+
+EXTN(jsimd_fdct_float_sse):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+;   push        esi                     ; unused
+;   push        edi                     ; unused
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process rows.
+
+    mov         edx, POINTER data(eax)  ; (FAST_FLOAT *)
+    mov         ecx, DCTSIZE/4
+    alignx      16, 7
+.rowloop:
+
+    movaps      xmm0, XMMWORD XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(2,1,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(3,1,edx,SIZEOF_FAST_FLOAT)
+
+    ; xmm0=(20 21 22 23), xmm2=(24 25 26 27)
+    ; xmm1=(30 31 32 33), xmm3=(34 35 36 37)
+
+    movaps      xmm4, xmm0              ; transpose coefficients(phase 1)
+    unpcklps    xmm0, xmm1              ; xmm0=(20 30 21 31)
+    unpckhps    xmm4, xmm1              ; xmm4=(22 32 23 33)
+    movaps      xmm5, xmm2              ; transpose coefficients(phase 1)
+    unpcklps    xmm2, xmm3              ; xmm2=(24 34 25 35)
+    unpckhps    xmm5, xmm3              ; xmm5=(26 36 27 37)
+
+    movaps      xmm6, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm7, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)
+
+    ; xmm6=(00 01 02 03), xmm1=(04 05 06 07)
+    ; xmm7=(10 11 12 13), xmm3=(14 15 16 17)
+
+    movaps      XMMWORD wk(0), xmm4   ; wk(0)=(22 32 23 33)
+    movaps      XMMWORD wk(1), xmm2   ; wk(1)=(24 34 25 35)
+
+    movaps      xmm4, xmm6              ; transpose coefficients(phase 1)
+    unpcklps    xmm6, xmm7              ; xmm6=(00 10 01 11)
+    unpckhps    xmm4, xmm7              ; xmm4=(02 12 03 13)
+    movaps      xmm2, xmm1              ; transpose coefficients(phase 1)
+    unpcklps    xmm1, xmm3              ; xmm1=(04 14 05 15)
+    unpckhps    xmm2, xmm3              ; xmm2=(06 16 07 17)
+
+    movaps      xmm7, xmm6              ; transpose coefficients(phase 2)
+    unpcklps2   xmm6, xmm0              ; xmm6=(00 10 20 30)=data0
+    unpckhps2   xmm7, xmm0              ; xmm7=(01 11 21 31)=data1
+    movaps      xmm3, xmm2              ; transpose coefficients(phase 2)
+    unpcklps2   xmm2, xmm5              ; xmm2=(06 16 26 36)=data6
+    unpckhps2   xmm3, xmm5              ; xmm3=(07 17 27 37)=data7
+
+    movaps      xmm0, xmm7
+    movaps      xmm5, xmm6
+    subps       xmm7, xmm2              ; xmm7=data1-data6=tmp6
+    subps       xmm6, xmm3              ; xmm6=data0-data7=tmp7
+    addps       xmm0, xmm2              ; xmm0=data1+data6=tmp1
+    addps       xmm5, xmm3              ; xmm5=data0+data7=tmp0
+
+    movaps      xmm2, XMMWORD wk(0)   ; xmm2=(22 32 23 33)
+    movaps      xmm3, XMMWORD wk(1)   ; xmm3=(24 34 25 35)
+    movaps      XMMWORD wk(0), xmm7   ; wk(0)=tmp6
+    movaps      XMMWORD wk(1), xmm6   ; wk(1)=tmp7
+
+    movaps      xmm7, xmm4              ; transpose coefficients(phase 2)
+    unpcklps2   xmm4, xmm2              ; xmm4=(02 12 22 32)=data2
+    unpckhps2   xmm7, xmm2              ; xmm7=(03 13 23 33)=data3
+    movaps      xmm6, xmm1              ; transpose coefficients(phase 2)
+    unpcklps2   xmm1, xmm3              ; xmm1=(04 14 24 34)=data4
+    unpckhps2   xmm6, xmm3              ; xmm6=(05 15 25 35)=data5
+
+    movaps      xmm2, xmm7
+    movaps      xmm3, xmm4
+    addps       xmm7, xmm1              ; xmm7=data3+data4=tmp3
+    addps       xmm4, xmm6              ; xmm4=data2+data5=tmp2
+    subps       xmm2, xmm1              ; xmm2=data3-data4=tmp4
+    subps       xmm3, xmm6              ; xmm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movaps      xmm1, xmm5
+    movaps      xmm6, xmm0
+    subps       xmm5, xmm7              ; xmm5=tmp13
+    subps       xmm0, xmm4              ; xmm0=tmp12
+    addps       xmm1, xmm7              ; xmm1=tmp10
+    addps       xmm6, xmm4              ; xmm6=tmp11
+
+    addps       xmm0, xmm5
+    mulps       xmm0, GOTOFF(ebx,PD_0_707)  ; xmm0=z1
+
+    movaps      xmm7, xmm1
+    movaps      xmm4, xmm5
+    subps       xmm1, xmm6              ; xmm1=data4
+    subps       xmm5, xmm0              ; xmm5=data6
+    addps       xmm7, xmm6              ; xmm7=data0
+    addps       xmm4, xmm0              ; xmm4=data2
+
+    movaps      XMMWORD XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(2,1,edx,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT), xmm7
+    movaps      XMMWORD XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT), xmm4
+
+    ; -- Odd part
+
+    movaps      xmm6, XMMWORD wk(0)   ; xmm6=tmp6
+    movaps      xmm0, XMMWORD wk(1)   ; xmm0=tmp7
+
+    addps       xmm2, xmm3              ; xmm2=tmp10
+    addps       xmm3, xmm6              ; xmm3=tmp11
+    addps       xmm6, xmm0              ; xmm6=tmp12, xmm0=tmp7
+
+    mulps       xmm3, GOTOFF(ebx,PD_0_707)  ; xmm3=z3
+
+    movaps      xmm1, xmm2                    ; xmm1=tmp10
+    subps       xmm2, xmm6
+    mulps       xmm2, GOTOFF(ebx,PD_0_382)  ; xmm2=z5
+    mulps       xmm1, GOTOFF(ebx,PD_0_541)  ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+    mulps       xmm6, GOTOFF(ebx,PD_1_306)  ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+    addps       xmm1, xmm2                    ; xmm1=z2
+    addps       xmm6, xmm2                    ; xmm6=z4
+
+    movaps      xmm5, xmm0
+    subps       xmm0, xmm3              ; xmm0=z13
+    addps       xmm5, xmm3              ; xmm5=z11
+
+    movaps      xmm7, xmm0
+    movaps      xmm4, xmm5
+    subps       xmm0, xmm1              ; xmm0=data3
+    subps       xmm5, xmm6              ; xmm5=data7
+    addps       xmm7, xmm1              ; xmm7=data5
+    addps       xmm4, xmm6              ; xmm4=data1
+
+    movaps      XMMWORD XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(3,1,edx,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT), xmm7
+    movaps      XMMWORD XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT), xmm4
+
+    add         edx, 4*DCTSIZE*SIZEOF_FAST_FLOAT
+    dec         ecx
+    jnz         near .rowloop
+
+    ; ---- Pass 2: process columns.
+
+    mov         edx, POINTER data(eax)  ; (FAST_FLOAT *)
+    mov         ecx, DCTSIZE/4
+    alignx      16, 7
+.columnloop:
+
+    movaps      xmm0, XMMWORD XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)
+
+    ; xmm0=(02 12 22 32), xmm2=(42 52 62 72)
+    ; xmm1=(03 13 23 33), xmm3=(43 53 63 73)
+
+    movaps      xmm4, xmm0              ; transpose coefficients(phase 1)
+    unpcklps    xmm0, xmm1              ; xmm0=(02 03 12 13)
+    unpckhps    xmm4, xmm1              ; xmm4=(22 23 32 33)
+    movaps      xmm5, xmm2              ; transpose coefficients(phase 1)
+    unpcklps    xmm2, xmm3              ; xmm2=(42 43 52 53)
+    unpckhps    xmm5, xmm3              ; xmm5=(62 63 72 73)
+
+    movaps      xmm6, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm7, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)
+
+    ; xmm6=(00 10 20 30), xmm1=(40 50 60 70)
+    ; xmm7=(01 11 21 31), xmm3=(41 51 61 71)
+
+    movaps      XMMWORD wk(0), xmm4   ; wk(0)=(22 23 32 33)
+    movaps      XMMWORD wk(1), xmm2   ; wk(1)=(42 43 52 53)
+
+    movaps      xmm4, xmm6              ; transpose coefficients(phase 1)
+    unpcklps    xmm6, xmm7              ; xmm6=(00 01 10 11)
+    unpckhps    xmm4, xmm7              ; xmm4=(20 21 30 31)
+    movaps      xmm2, xmm1              ; transpose coefficients(phase 1)
+    unpcklps    xmm1, xmm3              ; xmm1=(40 41 50 51)
+    unpckhps    xmm2, xmm3              ; xmm2=(60 61 70 71)
+
+    movaps      xmm7, xmm6              ; transpose coefficients(phase 2)
+    unpcklps2   xmm6, xmm0              ; xmm6=(00 01 02 03)=data0
+    unpckhps2   xmm7, xmm0              ; xmm7=(10 11 12 13)=data1
+    movaps      xmm3, xmm2              ; transpose coefficients(phase 2)
+    unpcklps2   xmm2, xmm5              ; xmm2=(60 61 62 63)=data6
+    unpckhps2   xmm3, xmm5              ; xmm3=(70 71 72 73)=data7
+
+    movaps      xmm0, xmm7
+    movaps      xmm5, xmm6
+    subps       xmm7, xmm2              ; xmm7=data1-data6=tmp6
+    subps       xmm6, xmm3              ; xmm6=data0-data7=tmp7
+    addps       xmm0, xmm2              ; xmm0=data1+data6=tmp1
+    addps       xmm5, xmm3              ; xmm5=data0+data7=tmp0
+
+    movaps      xmm2, XMMWORD wk(0)   ; xmm2=(22 23 32 33)
+    movaps      xmm3, XMMWORD wk(1)   ; xmm3=(42 43 52 53)
+    movaps      XMMWORD wk(0), xmm7   ; wk(0)=tmp6
+    movaps      XMMWORD wk(1), xmm6   ; wk(1)=tmp7
+
+    movaps      xmm7, xmm4              ; transpose coefficients(phase 2)
+    unpcklps2   xmm4, xmm2              ; xmm4=(20 21 22 23)=data2
+    unpckhps2   xmm7, xmm2              ; xmm7=(30 31 32 33)=data3
+    movaps      xmm6, xmm1              ; transpose coefficients(phase 2)
+    unpcklps2   xmm1, xmm3              ; xmm1=(40 41 42 43)=data4
+    unpckhps2   xmm6, xmm3              ; xmm6=(50 51 52 53)=data5
+
+    movaps      xmm2, xmm7
+    movaps      xmm3, xmm4
+    addps       xmm7, xmm1              ; xmm7=data3+data4=tmp3
+    addps       xmm4, xmm6              ; xmm4=data2+data5=tmp2
+    subps       xmm2, xmm1              ; xmm2=data3-data4=tmp4
+    subps       xmm3, xmm6              ; xmm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movaps      xmm1, xmm5
+    movaps      xmm6, xmm0
+    subps       xmm5, xmm7              ; xmm5=tmp13
+    subps       xmm0, xmm4              ; xmm0=tmp12
+    addps       xmm1, xmm7              ; xmm1=tmp10
+    addps       xmm6, xmm4              ; xmm6=tmp11
+
+    addps       xmm0, xmm5
+    mulps       xmm0, GOTOFF(ebx,PD_0_707)  ; xmm0=z1
+
+    movaps      xmm7, xmm1
+    movaps      xmm4, xmm5
+    subps       xmm1, xmm6              ; xmm1=data4
+    subps       xmm5, xmm0              ; xmm5=data6
+    addps       xmm7, xmm6              ; xmm7=data0
+    addps       xmm4, xmm0              ; xmm4=data2
+
+    movaps      XMMWORD XMMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT), xmm7
+    movaps      XMMWORD XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT), xmm4
+
+    ; -- Odd part
+
+    movaps      xmm6, XMMWORD wk(0)   ; xmm6=tmp6
+    movaps      xmm0, XMMWORD wk(1)   ; xmm0=tmp7
+
+    addps       xmm2, xmm3              ; xmm2=tmp10
+    addps       xmm3, xmm6              ; xmm3=tmp11
+    addps       xmm6, xmm0              ; xmm6=tmp12, xmm0=tmp7
+
+    mulps       xmm3, GOTOFF(ebx,PD_0_707)  ; xmm3=z3
+
+    movaps      xmm1, xmm2                    ; xmm1=tmp10
+    subps       xmm2, xmm6
+    mulps       xmm2, GOTOFF(ebx,PD_0_382)  ; xmm2=z5
+    mulps       xmm1, GOTOFF(ebx,PD_0_541)  ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+    mulps       xmm6, GOTOFF(ebx,PD_1_306)  ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+    addps       xmm1, xmm2                    ; xmm1=z2
+    addps       xmm6, xmm2                    ; xmm6=z4
+
+    movaps      xmm5, xmm0
+    subps       xmm0, xmm3              ; xmm0=z13
+    addps       xmm5, xmm3              ; xmm5=z11
+
+    movaps      xmm7, xmm0
+    movaps      xmm4, xmm5
+    subps       xmm0, xmm1              ; xmm0=data3
+    subps       xmm5, xmm6              ; xmm5=data7
+    addps       xmm7, xmm1              ; xmm7=data5
+    addps       xmm4, xmm6              ; xmm4=data1
+
+    movaps      XMMWORD XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT), xmm7
+    movaps      XMMWORD XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT), xmm4
+
+    add         edx, byte 4*SIZEOF_FAST_FLOAT
+    dec         ecx
+    jnz         near .columnloop
+
+;   pop         edi                     ; unused
+;   pop         esi                     ; unused
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jfdctfst-mmx.asm Added

@@ -0,0 +1,395 @@
+;
+; jfdctfst.asm - fast integer FDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the forward DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
+; for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  8  ; 14 is also OK.
+
+%if CONST_BITS == 8
+F_0_382 equ  98  ; FIX(0.382683433)
+F_0_541 equ 139  ; FIX(0.541196100)
+F_0_707 equ 181  ; FIX(0.707106781)
+F_1_306 equ 334  ; FIX(1.306562965)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_382 equ DESCALE( 410903207, 30 - CONST_BITS)  ; FIX(0.382683433)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_707 equ DESCALE( 759250124, 30 - CONST_BITS)  ; FIX(0.707106781)
+F_1_306 equ DESCALE(1402911301, 30 - CONST_BITS)  ; FIX(1.306562965)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS  2
+%define CONST_SHIFT              (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_ifast_mmx)
+
+EXTN(jconst_fdct_ifast_mmx):
+
+PW_F0707 times 4 dw F_0_707 << CONST_SHIFT
+PW_F0382 times 4 dw F_0_382 << CONST_SHIFT
+PW_F0541 times 4 dw F_0_541 << CONST_SHIFT
+PW_F1306 times 4 dw F_1_306 << CONST_SHIFT
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_ifast_mmx(DCTELEM *data)
+;
+
+%define data(b)       (b) + 8           ; DCTELEM *data
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_MMWORD  ; mmword wkWK_NUM
+%define WK_NUM        2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_ifast_mmx)
+
+EXTN(jsimd_fdct_ifast_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+;   push        esi                     ; unused
+;   push        edi                     ; unused
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process rows.
+
+    mov         edx, POINTER data(eax)  ; (DCTELEM *)
+    mov         ecx, DCTSIZE/4
+    alignx      16, 7
+.rowloop:
+
+    movq        mm0, MMWORD MMBLOCK(2,0,edx,SIZEOF_DCTELEM)
+    movq        mm1, MMWORD MMBLOCK(3,0,edx,SIZEOF_DCTELEM)
+    movq        mm2, MMWORD MMBLOCK(2,1,edx,SIZEOF_DCTELEM)
+    movq        mm3, MMWORD MMBLOCK(3,1,edx,SIZEOF_DCTELEM)
+
+    ; mm0=(20 21 22 23), mm2=(24 25 26 27)
+    ; mm1=(30 31 32 33), mm3=(34 35 36 37)
+
+    movq        mm4, mm0                ; transpose coefficients(phase 1)
+    punpcklwd   mm0, mm1                ; mm0=(20 30 21 31)
+    punpckhwd   mm4, mm1                ; mm4=(22 32 23 33)
+    movq        mm5, mm2                ; transpose coefficients(phase 1)
+    punpcklwd   mm2, mm3                ; mm2=(24 34 25 35)
+    punpckhwd   mm5, mm3                ; mm5=(26 36 27 37)
+
+    movq        mm6, MMWORD MMBLOCK(0,0,edx,SIZEOF_DCTELEM)
+    movq        mm7, MMWORD MMBLOCK(1,0,edx,SIZEOF_DCTELEM)
+    movq        mm1, MMWORD MMBLOCK(0,1,edx,SIZEOF_DCTELEM)
+    movq        mm3, MMWORD MMBLOCK(1,1,edx,SIZEOF_DCTELEM)
+
+    ; mm6=(00 01 02 03), mm1=(04 05 06 07)
+    ; mm7=(10 11 12 13), mm3=(14 15 16 17)
+
+    movq        MMWORD wk(0), mm4     ; wk(0)=(22 32 23 33)
+    movq        MMWORD wk(1), mm2     ; wk(1)=(24 34 25 35)
+
+    movq        mm4, mm6                ; transpose coefficients(phase 1)
+    punpcklwd   mm6, mm7                ; mm6=(00 10 01 11)
+    punpckhwd   mm4, mm7                ; mm4=(02 12 03 13)
+    movq        mm2, mm1                ; transpose coefficients(phase 1)
+    punpcklwd   mm1, mm3                ; mm1=(04 14 05 15)
+    punpckhwd   mm2, mm3                ; mm2=(06 16 07 17)
+
+    movq        mm7, mm6                ; transpose coefficients(phase 2)
+    punpckldq   mm6, mm0                ; mm6=(00 10 20 30)=data0
+    punpckhdq   mm7, mm0                ; mm7=(01 11 21 31)=data1
+    movq        mm3, mm2                ; transpose coefficients(phase 2)
+    punpckldq   mm2, mm5                ; mm2=(06 16 26 36)=data6
+    punpckhdq   mm3, mm5                ; mm3=(07 17 27 37)=data7
+
+    movq        mm0, mm7
+    movq        mm5, mm6
+    psubw       mm7, mm2                ; mm7=data1-data6=tmp6
+    psubw       mm6, mm3                ; mm6=data0-data7=tmp7
+    paddw       mm0, mm2                ; mm0=data1+data6=tmp1
+    paddw       mm5, mm3                ; mm5=data0+data7=tmp0
+
+    movq        mm2, MMWORD wk(0)     ; mm2=(22 32 23 33)
+    movq        mm3, MMWORD wk(1)     ; mm3=(24 34 25 35)
+    movq        MMWORD wk(0), mm7     ; wk(0)=tmp6
+    movq        MMWORD wk(1), mm6     ; wk(1)=tmp7
+
+    movq        mm7, mm4                ; transpose coefficients(phase 2)
+    punpckldq   mm4, mm2                ; mm4=(02 12 22 32)=data2
+    punpckhdq   mm7, mm2                ; mm7=(03 13 23 33)=data3
+    movq        mm6, mm1                ; transpose coefficients(phase 2)
+    punpckldq   mm1, mm3                ; mm1=(04 14 24 34)=data4
+    punpckhdq   mm6, mm3                ; mm6=(05 15 25 35)=data5
+
+    movq        mm2, mm7
+    movq        mm3, mm4
+    paddw       mm7, mm1                ; mm7=data3+data4=tmp3
+    paddw       mm4, mm6                ; mm4=data2+data5=tmp2
+    psubw       mm2, mm1                ; mm2=data3-data4=tmp4
+    psubw       mm3, mm6                ; mm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movq        mm1, mm5
+    movq        mm6, mm0
+    psubw       mm5, mm7                ; mm5=tmp13
+    psubw       mm0, mm4                ; mm0=tmp12
+    paddw       mm1, mm7                ; mm1=tmp10
+    paddw       mm6, mm4                ; mm6=tmp11
+
+    paddw       mm0, mm5
+    psllw       mm0, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      mm0, GOTOFF(ebx,PW_F0707)  ; mm0=z1
+
+    movq        mm7, mm1
+    movq        mm4, mm5
+    psubw       mm1, mm6                ; mm1=data4
+    psubw       mm5, mm0                ; mm5=data6
+    paddw       mm7, mm6                ; mm7=data0
+    paddw       mm4, mm0                ; mm4=data2
+
+    movq        MMWORD MMBLOCK(0,1,edx,SIZEOF_DCTELEM), mm1
+    movq        MMWORD MMBLOCK(2,1,edx,SIZEOF_DCTELEM), mm5
+    movq        MMWORD MMBLOCK(0,0,edx,SIZEOF_DCTELEM), mm7
+    movq        MMWORD MMBLOCK(2,0,edx,SIZEOF_DCTELEM), mm4
+
+    ; -- Odd part
+
+    movq        mm6, MMWORD wk(0)     ; mm6=tmp6
+    movq        mm0, MMWORD wk(1)     ; mm0=tmp7
+
+    paddw       mm2, mm3                ; mm2=tmp10
+    paddw       mm3, mm6                ; mm3=tmp11
+    paddw       mm6, mm0                ; mm6=tmp12, mm0=tmp7
+
+    psllw       mm2, PRE_MULTIPLY_SCALE_BITS
+    psllw       mm6, PRE_MULTIPLY_SCALE_BITS
+
+    psllw       mm3, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      mm3, GOTOFF(ebx,PW_F0707)  ; mm3=z3
+
+    movq        mm1, mm2                     ; mm1=tmp10
+    psubw       mm2, mm6
+    pmulhw      mm2, GOTOFF(ebx,PW_F0382)  ; mm2=z5
+    pmulhw      mm1, GOTOFF(ebx,PW_F0541)  ; mm1=MULTIPLY(tmp10,FIX_0_54119610)
+    pmulhw      mm6, GOTOFF(ebx,PW_F1306)  ; mm6=MULTIPLY(tmp12,FIX_1_30656296)
+    paddw       mm1, mm2                     ; mm1=z2
+    paddw       mm6, mm2                     ; mm6=z4
+
+    movq        mm5, mm0
+    psubw       mm0, mm3                ; mm0=z13
+    paddw       mm5, mm3                ; mm5=z11
+
+    movq        mm7, mm0
+    movq        mm4, mm5
+    psubw       mm0, mm1                ; mm0=data3
+    psubw       mm5, mm6                ; mm5=data7
+    paddw       mm7, mm1                ; mm7=data5
+    paddw       mm4, mm6                ; mm4=data1
+
+    movq        MMWORD MMBLOCK(3,0,edx,SIZEOF_DCTELEM), mm0
+    movq        MMWORD MMBLOCK(3,1,edx,SIZEOF_DCTELEM), mm5
+    movq        MMWORD MMBLOCK(1,1,edx,SIZEOF_DCTELEM), mm7
+    movq        MMWORD MMBLOCK(1,0,edx,SIZEOF_DCTELEM), mm4
+
+    add         edx, byte 4*DCTSIZE*SIZEOF_DCTELEM
+    dec         ecx
+    jnz         near .rowloop
+
+    ; ---- Pass 2: process columns.
+
+    mov         edx, POINTER data(eax)  ; (DCTELEM *)
+    mov         ecx, DCTSIZE/4
+    alignx      16, 7
+.columnloop:
+
+    movq        mm0, MMWORD MMBLOCK(2,0,edx,SIZEOF_DCTELEM)
+    movq        mm1, MMWORD MMBLOCK(3,0,edx,SIZEOF_DCTELEM)
+    movq        mm2, MMWORD MMBLOCK(6,0,edx,SIZEOF_DCTELEM)
+    movq        mm3, MMWORD MMBLOCK(7,0,edx,SIZEOF_DCTELEM)
+
+    ; mm0=(02 12 22 32), mm2=(42 52 62 72)
+    ; mm1=(03 13 23 33), mm3=(43 53 63 73)
+
+    movq        mm4, mm0                ; transpose coefficients(phase 1)
+    punpcklwd   mm0, mm1                ; mm0=(02 03 12 13)
+    punpckhwd   mm4, mm1                ; mm4=(22 23 32 33)
+    movq        mm5, mm2                ; transpose coefficients(phase 1)
+    punpcklwd   mm2, mm3                ; mm2=(42 43 52 53)
+    punpckhwd   mm5, mm3                ; mm5=(62 63 72 73)
+
+    movq        mm6, MMWORD MMBLOCK(0,0,edx,SIZEOF_DCTELEM)
+    movq        mm7, MMWORD MMBLOCK(1,0,edx,SIZEOF_DCTELEM)
+    movq        mm1, MMWORD MMBLOCK(4,0,edx,SIZEOF_DCTELEM)
+    movq        mm3, MMWORD MMBLOCK(5,0,edx,SIZEOF_DCTELEM)
+
+    ; mm6=(00 10 20 30), mm1=(40 50 60 70)
+    ; mm7=(01 11 21 31), mm3=(41 51 61 71)
+
+    movq        MMWORD wk(0), mm4     ; wk(0)=(22 23 32 33)
+    movq        MMWORD wk(1), mm2     ; wk(1)=(42 43 52 53)
+
+    movq        mm4, mm6                ; transpose coefficients(phase 1)
+    punpcklwd   mm6, mm7                ; mm6=(00 01 10 11)
+    punpckhwd   mm4, mm7                ; mm4=(20 21 30 31)
+    movq        mm2, mm1                ; transpose coefficients(phase 1)
+    punpcklwd   mm1, mm3                ; mm1=(40 41 50 51)
+    punpckhwd   mm2, mm3                ; mm2=(60 61 70 71)
+
+    movq        mm7, mm6                ; transpose coefficients(phase 2)
+    punpckldq   mm6, mm0                ; mm6=(00 01 02 03)=data0
+    punpckhdq   mm7, mm0                ; mm7=(10 11 12 13)=data1
+    movq        mm3, mm2                ; transpose coefficients(phase 2)
+    punpckldq   mm2, mm5                ; mm2=(60 61 62 63)=data6
+    punpckhdq   mm3, mm5                ; mm3=(70 71 72 73)=data7
+
+    movq        mm0, mm7
+    movq        mm5, mm6
+    psubw       mm7, mm2                ; mm7=data1-data6=tmp6
+    psubw       mm6, mm3                ; mm6=data0-data7=tmp7
+    paddw       mm0, mm2                ; mm0=data1+data6=tmp1
+    paddw       mm5, mm3                ; mm5=data0+data7=tmp0
+
+    movq        mm2, MMWORD wk(0)     ; mm2=(22 23 32 33)
+    movq        mm3, MMWORD wk(1)     ; mm3=(42 43 52 53)
+    movq        MMWORD wk(0), mm7     ; wk(0)=tmp6
+    movq        MMWORD wk(1), mm6     ; wk(1)=tmp7
+
+    movq        mm7, mm4                ; transpose coefficients(phase 2)
+    punpckldq   mm4, mm2                ; mm4=(20 21 22 23)=data2
+    punpckhdq   mm7, mm2                ; mm7=(30 31 32 33)=data3
+    movq        mm6, mm1                ; transpose coefficients(phase 2)
+    punpckldq   mm1, mm3                ; mm1=(40 41 42 43)=data4
+    punpckhdq   mm6, mm3                ; mm6=(50 51 52 53)=data5
+
+    movq        mm2, mm7
+    movq        mm3, mm4
+    paddw       mm7, mm1                ; mm7=data3+data4=tmp3
+    paddw       mm4, mm6                ; mm4=data2+data5=tmp2
+    psubw       mm2, mm1                ; mm2=data3-data4=tmp4
+    psubw       mm3, mm6                ; mm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movq        mm1, mm5
+    movq        mm6, mm0
+    psubw       mm5, mm7                ; mm5=tmp13
+    psubw       mm0, mm4                ; mm0=tmp12
+    paddw       mm1, mm7                ; mm1=tmp10
+    paddw       mm6, mm4                ; mm6=tmp11
+
+    paddw       mm0, mm5
+    psllw       mm0, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      mm0, GOTOFF(ebx,PW_F0707)  ; mm0=z1
+
+    movq        mm7, mm1
+    movq        mm4, mm5
+    psubw       mm1, mm6                ; mm1=data4
+    psubw       mm5, mm0                ; mm5=data6
+    paddw       mm7, mm6                ; mm7=data0
+    paddw       mm4, mm0                ; mm4=data2
+
+    movq        MMWORD MMBLOCK(4,0,edx,SIZEOF_DCTELEM), mm1
+    movq        MMWORD MMBLOCK(6,0,edx,SIZEOF_DCTELEM), mm5
+    movq        MMWORD MMBLOCK(0,0,edx,SIZEOF_DCTELEM), mm7
+    movq        MMWORD MMBLOCK(2,0,edx,SIZEOF_DCTELEM), mm4
+
+    ; -- Odd part
+
+    movq        mm6, MMWORD wk(0)     ; mm6=tmp6
+    movq        mm0, MMWORD wk(1)     ; mm0=tmp7
+
+    paddw       mm2, mm3                ; mm2=tmp10
+    paddw       mm3, mm6                ; mm3=tmp11
+    paddw       mm6, mm0                ; mm6=tmp12, mm0=tmp7
+
+    psllw       mm2, PRE_MULTIPLY_SCALE_BITS
+    psllw       mm6, PRE_MULTIPLY_SCALE_BITS
+
+    psllw       mm3, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      mm3, GOTOFF(ebx,PW_F0707)  ; mm3=z3
+
+    movq        mm1, mm2                     ; mm1=tmp10
+    psubw       mm2, mm6
+    pmulhw      mm2, GOTOFF(ebx,PW_F0382)  ; mm2=z5
+    pmulhw      mm1, GOTOFF(ebx,PW_F0541)  ; mm1=MULTIPLY(tmp10,FIX_0_54119610)
+    pmulhw      mm6, GOTOFF(ebx,PW_F1306)  ; mm6=MULTIPLY(tmp12,FIX_1_30656296)
+    paddw       mm1, mm2                     ; mm1=z2
+    paddw       mm6, mm2                     ; mm6=z4
+
+    movq        mm5, mm0
+    psubw       mm0, mm3                ; mm0=z13
+    paddw       mm5, mm3                ; mm5=z11
+
+    movq        mm7, mm0
+    movq        mm4, mm5
+    psubw       mm0, mm1                ; mm0=data3
+    psubw       mm5, mm6                ; mm5=data7
+    paddw       mm7, mm1                ; mm7=data5
+    paddw       mm4, mm6                ; mm4=data1
+
+    movq        MMWORD MMBLOCK(3,0,edx,SIZEOF_DCTELEM), mm0
+    movq        MMWORD MMBLOCK(7,0,edx,SIZEOF_DCTELEM), mm5
+    movq        MMWORD MMBLOCK(5,0,edx,SIZEOF_DCTELEM), mm7
+    movq        MMWORD MMBLOCK(1,0,edx,SIZEOF_DCTELEM), mm4
+
+    add         edx, byte 4*SIZEOF_DCTELEM
+    dec         ecx
+    jnz         near .columnloop
+
+    emms                                ; empty MMX state
+
+;   pop         edi                     ; unused
+;   pop         esi                     ; unused
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jfdctfst-sse2.asm Added

@@ -0,0 +1,403 @@
+;
+; jfdctfst.asm - fast integer FDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the forward DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
+; for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  8  ; 14 is also OK.
+
+%if CONST_BITS == 8
+F_0_382 equ  98  ; FIX(0.382683433)
+F_0_541 equ 139  ; FIX(0.541196100)
+F_0_707 equ 181  ; FIX(0.707106781)
+F_1_306 equ 334  ; FIX(1.306562965)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_382 equ DESCALE( 410903207, 30 - CONST_BITS)  ; FIX(0.382683433)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_707 equ DESCALE( 759250124, 30 - CONST_BITS)  ; FIX(0.707106781)
+F_1_306 equ DESCALE(1402911301, 30 - CONST_BITS)  ; FIX(1.306562965)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS  2
+%define CONST_SHIFT              (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_ifast_sse2)
+
+EXTN(jconst_fdct_ifast_sse2):
+
+PW_F0707 times 8 dw F_0_707 << CONST_SHIFT
+PW_F0382 times 8 dw F_0_382 << CONST_SHIFT
+PW_F0541 times 8 dw F_0_541 << CONST_SHIFT
+PW_F1306 times 8 dw F_1_306 << CONST_SHIFT
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_ifast_sse2(DCTELEM *data)
+;
+
+%define data(b)       (b) + 8           ; DCTELEM *data
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM        2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_ifast_sse2)
+
+EXTN(jsimd_fdct_ifast_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+;   push        esi                     ; unused
+;   push        edi                     ; unused
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process rows.
+
+    mov         edx, POINTER data(eax)  ; (DCTELEM *)
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm1, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm2, XMMWORD XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm3, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)
+
+    ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+    ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+    movdqa      xmm4, xmm0              ; transpose coefficients(phase 1)
+    punpcklwd   xmm0, xmm1              ; xmm0=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm4, xmm1              ; xmm4=(04 14 05 15 06 16 07 17)
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 1)
+    punpcklwd   xmm2, xmm3              ; xmm2=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm5, xmm3              ; xmm5=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm6, XMMWORD XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm7, XMMWORD XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm1, XMMWORD XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)
+
+    ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+    ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+    movdqa      XMMWORD wk(1), xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm2, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm7              ; xmm6=(40 50 41 51 42 52 43 53)
+    punpckhwd   xmm2, xmm7              ; xmm2=(44 54 45 55 46 56 47 57)
+    movdqa      xmm5, xmm1              ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm3              ; xmm1=(60 70 61 71 62 72 63 73)
+    punpckhwd   xmm5, xmm3              ; xmm5=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm7, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm1              ; xmm6=(40 50 60 70 41 51 61 71)
+    punpckhdq   xmm7, xmm1              ; xmm7=(42 52 62 72 43 53 63 73)
+    movdqa      xmm3, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm5              ; xmm2=(44 54 64 74 45 55 65 75)
+    punpckhdq   xmm3, xmm5              ; xmm3=(46 56 66 76 47 57 67 77)
+
+    movdqa      xmm1, XMMWORD wk(0)   ; xmm1=(20 30 21 31 22 32 23 33)
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=(24 34 25 35 26 36 27 37)
+    movdqa      XMMWORD wk(0), xmm7   ; wk(0)=(42 52 62 72 43 53 63 73)
+    movdqa      XMMWORD wk(1), xmm2   ; wk(1)=(44 54 64 74 45 55 65 75)
+
+    movdqa      xmm7, xmm0              ; transpose coefficients(phase 2)
+    punpckldq   xmm0, xmm1              ; xmm0=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm7, xmm1              ; xmm7=(02 12 22 32 03 13 23 33)
+    movdqa      xmm2, xmm4              ; transpose coefficients(phase 2)
+    punpckldq   xmm4, xmm5              ; xmm4=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm2, xmm5              ; xmm2=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm1, xmm0              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm0, xmm6              ; xmm0=(00 10 20 30 40 50 60 70)=data0
+    punpckhqdq  xmm1, xmm6              ; xmm1=(01 11 21 31 41 51 61 71)=data1
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm2, xmm3              ; xmm2=(06 16 26 36 46 56 66 76)=data6
+    punpckhqdq  xmm5, xmm3              ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+    movdqa      xmm6, xmm1
+    movdqa      xmm3, xmm0
+    psubw       xmm1, xmm2              ; xmm1=data1-data6=tmp6
+    psubw       xmm0, xmm5              ; xmm0=data0-data7=tmp7
+    paddw       xmm6, xmm2              ; xmm6=data1+data6=tmp1
+    paddw       xmm3, xmm5              ; xmm3=data0+data7=tmp0
+
+    movdqa      xmm2, XMMWORD wk(0)   ; xmm2=(42 52 62 72 43 53 63 73)
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=(44 54 64 74 45 55 65 75)
+    movdqa      XMMWORD wk(0), xmm1   ; wk(0)=tmp6
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=tmp7
+
+    movdqa      xmm1, xmm7              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm7, xmm2              ; xmm7=(02 12 22 32 42 52 62 72)=data2
+    punpckhqdq  xmm1, xmm2              ; xmm1=(03 13 23 33 43 53 63 73)=data3
+    movdqa      xmm0, xmm4              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm4, xmm5              ; xmm4=(04 14 24 34 44 54 64 74)=data4
+    punpckhqdq  xmm0, xmm5              ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+    movdqa      xmm2, xmm1
+    movdqa      xmm5, xmm7
+    paddw       xmm1, xmm4              ; xmm1=data3+data4=tmp3
+    paddw       xmm7, xmm0              ; xmm7=data2+data5=tmp2
+    psubw       xmm2, xmm4              ; xmm2=data3-data4=tmp4
+    psubw       xmm5, xmm0              ; xmm5=data2-data5=tmp5
+
+    ; -- Even part
+
+    movdqa      xmm4, xmm3
+    movdqa      xmm0, xmm6
+    psubw       xmm3, xmm1              ; xmm3=tmp13
+    psubw       xmm6, xmm7              ; xmm6=tmp12
+    paddw       xmm4, xmm1              ; xmm4=tmp10
+    paddw       xmm0, xmm7              ; xmm0=tmp11
+
+    paddw       xmm6, xmm3
+    psllw       xmm6, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm6, GOTOFF(ebx,PW_F0707)  ; xmm6=z1
+
+    movdqa      xmm1, xmm4
+    movdqa      xmm7, xmm3
+    psubw       xmm4, xmm0              ; xmm4=data4
+    psubw       xmm3, xmm6              ; xmm3=data6
+    paddw       xmm1, xmm0              ; xmm1=data0
+    paddw       xmm7, xmm6              ; xmm7=data2
+
+    movdqa      xmm0, XMMWORD wk(0)   ; xmm0=tmp6
+    movdqa      xmm6, XMMWORD wk(1)   ; xmm6=tmp7
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=data4
+    movdqa      XMMWORD wk(1), xmm3   ; wk(1)=data6
+
+    ; -- Odd part
+
+    paddw       xmm2, xmm5              ; xmm2=tmp10
+    paddw       xmm5, xmm0              ; xmm5=tmp11
+    paddw       xmm0, xmm6              ; xmm0=tmp12, xmm6=tmp7
+
+    psllw       xmm2, PRE_MULTIPLY_SCALE_BITS
+    psllw       xmm0, PRE_MULTIPLY_SCALE_BITS
+
+    psllw       xmm5, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm5, GOTOFF(ebx,PW_F0707)  ; xmm5=z3
+
+    movdqa      xmm4, xmm2                    ; xmm4=tmp10
+    psubw       xmm2, xmm0
+    pmulhw      xmm2, GOTOFF(ebx,PW_F0382)  ; xmm2=z5
+    pmulhw      xmm4, GOTOFF(ebx,PW_F0541)  ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+    pmulhw      xmm0, GOTOFF(ebx,PW_F1306)  ; xmm0=MULTIPLY(tmp12,FIX_1_306562)
+    paddw       xmm4, xmm2                    ; xmm4=z2
+    paddw       xmm0, xmm2                    ; xmm0=z4
+
+    movdqa      xmm3, xmm6
+    psubw       xmm6, xmm5              ; xmm6=z13
+    paddw       xmm3, xmm5              ; xmm3=z11
+
+    movdqa      xmm2, xmm6
+    movdqa      xmm5, xmm3
+    psubw       xmm6, xmm4              ; xmm6=data3
+    psubw       xmm3, xmm0              ; xmm3=data7
+    paddw       xmm2, xmm4              ; xmm2=data5
+    paddw       xmm5, xmm0              ; xmm5=data1
+
+    ; ---- Pass 2: process columns.
+
+;   mov         edx, POINTER data(eax)  ; (DCTELEM *)
+
+    ; xmm1=(00 10 20 30 40 50 60 70), xmm7=(02 12 22 32 42 52 62 72)
+    ; xmm5=(01 11 21 31 41 51 61 71), xmm6=(03 13 23 33 43 53 63 73)
+
+    movdqa      xmm4, xmm1              ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm5              ; xmm1=(00 01 10 11 20 21 30 31)
+    punpckhwd   xmm4, xmm5              ; xmm4=(40 41 50 51 60 61 70 71)
+    movdqa      xmm0, xmm7              ; transpose coefficients(phase 1)
+    punpcklwd   xmm7, xmm6              ; xmm7=(02 03 12 13 22 23 32 33)
+    punpckhwd   xmm0, xmm6              ; xmm0=(42 43 52 53 62 63 72 73)
+
+    movdqa      xmm5, XMMWORD wk(0)   ; xmm5=col4
+    movdqa      xmm6, XMMWORD wk(1)   ; xmm6=col6
+
+    ; xmm5=(04 14 24 34 44 54 64 74), xmm6=(06 16 26 36 46 56 66 76)
+    ; xmm2=(05 15 25 35 45 55 65 75), xmm3=(07 17 27 37 47 57 67 77)
+
+    movdqa      XMMWORD wk(0), xmm7   ; wk(0)=(02 03 12 13 22 23 32 33)
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+    movdqa      xmm7, xmm5              ; transpose coefficients(phase 1)
+    punpcklwd   xmm5, xmm2              ; xmm5=(04 05 14 15 24 25 34 35)
+    punpckhwd   xmm7, xmm2              ; xmm7=(44 45 54 55 64 65 74 75)
+    movdqa      xmm0, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm3              ; xmm6=(06 07 16 17 26 27 36 37)
+    punpckhwd   xmm0, xmm3              ; xmm0=(46 47 56 57 66 67 76 77)
+
+    movdqa      xmm2, xmm5              ; transpose coefficients(phase 2)
+    punpckldq   xmm5, xmm6              ; xmm5=(04 05 06 07 14 15 16 17)
+    punpckhdq   xmm2, xmm6              ; xmm2=(24 25 26 27 34 35 36 37)
+    movdqa      xmm3, xmm7              ; transpose coefficients(phase 2)
+    punpckldq   xmm7, xmm0              ; xmm7=(44 45 46 47 54 55 56 57)
+    punpckhdq   xmm3, xmm0              ; xmm3=(64 65 66 67 74 75 76 77)
+
+    movdqa      xmm6, XMMWORD wk(0)   ; xmm6=(02 03 12 13 22 23 32 33)
+    movdqa      xmm0, XMMWORD wk(1)   ; xmm0=(42 43 52 53 62 63 72 73)
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=(24 25 26 27 34 35 36 37)
+    movdqa      XMMWORD wk(1), xmm7   ; wk(1)=(44 45 46 47 54 55 56 57)
+
+    movdqa      xmm2, xmm1              ; transpose coefficients(phase 2)
+    punpckldq   xmm1, xmm6              ; xmm1=(00 01 02 03 10 11 12 13)
+    punpckhdq   xmm2, xmm6              ; xmm2=(20 21 22 23 30 31 32 33)
+    movdqa      xmm7, xmm4              ; transpose coefficients(phase 2)
+    punpckldq   xmm4, xmm0              ; xmm4=(40 41 42 43 50 51 52 53)
+    punpckhdq   xmm7, xmm0              ; xmm7=(60 61 62 63 70 71 72 73)
+
+    movdqa      xmm6, xmm1              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm1, xmm5              ; xmm1=(00 01 02 03 04 05 06 07)=data0
+    punpckhqdq  xmm6, xmm5              ; xmm6=(10 11 12 13 14 15 16 17)=data1
+    movdqa      xmm0, xmm7              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm7, xmm3              ; xmm7=(60 61 62 63 64 65 66 67)=data6
+    punpckhqdq  xmm0, xmm3              ; xmm0=(70 71 72 73 74 75 76 77)=data7
+
+    movdqa      xmm5, xmm6
+    movdqa      xmm3, xmm1
+    psubw       xmm6, xmm7              ; xmm6=data1-data6=tmp6
+    psubw       xmm1, xmm0              ; xmm1=data0-data7=tmp7
+    paddw       xmm5, xmm7              ; xmm5=data1+data6=tmp1
+    paddw       xmm3, xmm0              ; xmm3=data0+data7=tmp0
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=(24 25 26 27 34 35 36 37)
+    movdqa      xmm0, XMMWORD wk(1)   ; xmm0=(44 45 46 47 54 55 56 57)
+    movdqa      XMMWORD wk(0), xmm6   ; wk(0)=tmp6
+    movdqa      XMMWORD wk(1), xmm1   ; wk(1)=tmp7
+
+    movdqa      xmm6, xmm2              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm2, xmm7              ; xmm2=(20 21 22 23 24 25 26 27)=data2
+    punpckhqdq  xmm6, xmm7              ; xmm6=(30 31 32 33 34 35 36 37)=data3
+    movdqa      xmm1, xmm4              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm4, xmm0              ; xmm4=(40 41 42 43 44 45 46 47)=data4
+    punpckhqdq  xmm1, xmm0              ; xmm1=(50 51 52 53 54 55 56 57)=data5
+
+    movdqa      xmm7, xmm6
+    movdqa      xmm0, xmm2
+    paddw       xmm6, xmm4              ; xmm6=data3+data4=tmp3
+    paddw       xmm2, xmm1              ; xmm2=data2+data5=tmp2
+    psubw       xmm7, xmm4              ; xmm7=data3-data4=tmp4
+    psubw       xmm0, xmm1              ; xmm0=data2-data5=tmp5
+
+    ; -- Even part
+
+    movdqa      xmm4, xmm3
+    movdqa      xmm1, xmm5
+    psubw       xmm3, xmm6              ; xmm3=tmp13
+    psubw       xmm5, xmm2              ; xmm5=tmp12
+    paddw       xmm4, xmm6              ; xmm4=tmp10
+    paddw       xmm1, xmm2              ; xmm1=tmp11
+
+    paddw       xmm5, xmm3
+    psllw       xmm5, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm5, GOTOFF(ebx,PW_F0707)  ; xmm5=z1
+
+    movdqa      xmm6, xmm4
+    movdqa      xmm2, xmm3
+    psubw       xmm4, xmm1              ; xmm4=data4
+    psubw       xmm3, xmm5              ; xmm3=data6
+    paddw       xmm6, xmm1              ; xmm6=data0
+    paddw       xmm2, xmm5              ; xmm2=data2
+
+    movdqa      XMMWORD XMMBLOCK(4,0,edx,SIZEOF_DCTELEM), xmm4
+    movdqa      XMMWORD XMMBLOCK(6,0,edx,SIZEOF_DCTELEM), xmm3
+    movdqa      XMMWORD XMMBLOCK(0,0,edx,SIZEOF_DCTELEM), xmm6
+    movdqa      XMMWORD XMMBLOCK(2,0,edx,SIZEOF_DCTELEM), xmm2
+
+    ; -- Odd part
+
+    movdqa      xmm1, XMMWORD wk(0)   ; xmm1=tmp6
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=tmp7
+
+    paddw       xmm7, xmm0              ; xmm7=tmp10
+    paddw       xmm0, xmm1              ; xmm0=tmp11
+    paddw       xmm1, xmm5              ; xmm1=tmp12, xmm5=tmp7
+
+    psllw       xmm7, PRE_MULTIPLY_SCALE_BITS
+    psllw       xmm1, PRE_MULTIPLY_SCALE_BITS
+
+    psllw       xmm0, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm0, GOTOFF(ebx,PW_F0707)  ; xmm0=z3
+
+    movdqa      xmm4, xmm7                    ; xmm4=tmp10
+    psubw       xmm7, xmm1
+    pmulhw      xmm7, GOTOFF(ebx,PW_F0382)  ; xmm7=z5
+    pmulhw      xmm4, GOTOFF(ebx,PW_F0541)  ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+    pmulhw      xmm1, GOTOFF(ebx,PW_F1306)  ; xmm1=MULTIPLY(tmp12,FIX_1_306562)
+    paddw       xmm4, xmm7                    ; xmm4=z2
+    paddw       xmm1, xmm7                    ; xmm1=z4
+
+    movdqa      xmm3, xmm5
+    psubw       xmm5, xmm0              ; xmm5=z13
+    paddw       xmm3, xmm0              ; xmm3=z11
+
+    movdqa      xmm6, xmm5
+    movdqa      xmm2, xmm3
+    psubw       xmm5, xmm4              ; xmm5=data3
+    psubw       xmm3, xmm1              ; xmm3=data7
+    paddw       xmm6, xmm4              ; xmm6=data5
+    paddw       xmm2, xmm1              ; xmm2=data1
+
+    movdqa      XMMWORD XMMBLOCK(3,0,edx,SIZEOF_DCTELEM), xmm5
+    movdqa      XMMWORD XMMBLOCK(7,0,edx,SIZEOF_DCTELEM), xmm3
+    movdqa      XMMWORD XMMBLOCK(5,0,edx,SIZEOF_DCTELEM), xmm6
+    movdqa      XMMWORD XMMBLOCK(1,0,edx,SIZEOF_DCTELEM), xmm2
+
+;   pop         edi                     ; unused
+;   pop         esi                     ; unused
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jfdctint-avx2.asm Added

@@ -0,0 +1,331 @@
+;
+; jfdctint.asm - accurate integer FDCT (AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, 2018, 2020, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+; In-place 8x8x16-bit matrix transpose using AVX2 instructions
+; %1-%4: Input/output registers
+; %5-%8: Temp registers
+
+%macro dotranspose 8
+    ; %1=(00 01 02 03 04 05 06 07  40 41 42 43 44 45 46 47)
+    ; %2=(10 11 12 13 14 15 16 17  50 51 52 53 54 55 56 57)
+    ; %3=(20 21 22 23 24 25 26 27  60 61 62 63 64 65 66 67)
+    ; %4=(30 31 32 33 34 35 36 37  70 71 72 73 74 75 76 77)
+
+    vpunpcklwd  %5, %1, %2
+    vpunpckhwd  %6, %1, %2
+    vpunpcklwd  %7, %3, %4
+    vpunpckhwd  %8, %3, %4
+    ; transpose coefficients(phase 1)
+    ; %5=(00 10 01 11 02 12 03 13  40 50 41 51 42 52 43 53)
+    ; %6=(04 14 05 15 06 16 07 17  44 54 45 55 46 56 47 57)
+    ; %7=(20 30 21 31 22 32 23 33  60 70 61 71 62 72 63 73)
+    ; %8=(24 34 25 35 26 36 27 37  64 74 65 75 66 76 67 77)
+
+    vpunpckldq  %1, %5, %7
+    vpunpckhdq  %2, %5, %7
+    vpunpckldq  %3, %6, %8
+    vpunpckhdq  %4, %6, %8
+    ; transpose coefficients(phase 2)
+    ; %1=(00 10 20 30 01 11 21 31  40 50 60 70 41 51 61 71)
+    ; %2=(02 12 22 32 03 13 23 33  42 52 62 72 43 53 63 73)
+    ; %3=(04 14 24 34 05 15 25 35  44 54 64 74 45 55 65 75)
+    ; %4=(06 16 26 36 07 17 27 37  46 56 66 76 47 57 67 77)
+
+    vpermq      %1, %1, 0x8D
+    vpermq      %2, %2, 0x8D
+    vpermq      %3, %3, 0xD8
+    vpermq      %4, %4, 0xD8
+    ; transpose coefficients(phase 3)
+    ; %1=(01 11 21 31 41 51 61 71  00 10 20 30 40 50 60 70)
+    ; %2=(03 13 23 33 43 53 63 73  02 12 22 32 42 52 62 72)
+    ; %3=(04 14 24 34 44 54 64 74  05 15 25 35 45 55 65 75)
+    ; %4=(06 16 26 36 46 56 66 76  07 17 27 37 47 57 67 77)
+%endmacro
+
+; --------------------------------------------------------------------------
+; In-place 8x8x16-bit accurate integer forward DCT using AVX2 instructions
+; %1-%4: Input/output registers
+; %5-%8: Temp registers
+; %9:    Pass (1 or 2)
+
+%macro dodct 9
+    vpsubw      %5, %1, %4              ; %5=data1_0-data6_7=tmp6_7
+    vpaddw      %6, %1, %4              ; %6=data1_0+data6_7=tmp1_0
+    vpaddw      %7, %2, %3              ; %7=data3_2+data4_5=tmp3_2
+    vpsubw      %8, %2, %3              ; %8=data3_2-data4_5=tmp4_5
+
+    ; -- Even part
+
+    vperm2i128  %6, %6, %6, 0x01        ; %6=tmp0_1
+    vpaddw      %1, %6, %7              ; %1=tmp0_1+tmp3_2=tmp10_11
+    vpsubw      %6, %6, %7              ; %6=tmp0_1-tmp3_2=tmp13_12
+
+    vperm2i128  %7, %1, %1, 0x01        ; %7=tmp11_10
+    vpsignw     %1, %1, GOTOFF(ebx, PW_1_NEG1)  ; %1=tmp10_neg11
+    vpaddw      %7, %7, %1              ; %7=(tmp10+tmp11)_(tmp10-tmp11)
+%if %9 == 1
+    vpsllw      %1, %7, PASS1_BITS      ; %1=data0_4
+%else
+    vpaddw      %7, %7, GOTOFF(ebx, PW_DESCALE_P2X)
+    vpsraw      %1, %7, PASS1_BITS      ; %1=data0_4
+%endif
+
+    ; (Original)
+    ; z1 = (tmp12 + tmp13) * 0.541196100;
+    ; data2 = z1 + tmp13 * 0.765366865;
+    ; data6 = z1 + tmp12 * -1.847759065;
+    ;
+    ; (This implementation)
+    ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+    ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+    vperm2i128  %7, %6, %6, 0x01        ; %7=tmp12_13
+    vpunpcklwd  %2, %6, %7
+    vpunpckhwd  %6, %6, %7
+    vpmaddwd    %2, %2, GOTOFF(ebx, PW_F130_F054_MF130_F054)  ; %2=data2_6L
+    vpmaddwd    %6, %6, GOTOFF(ebx, PW_F130_F054_MF130_F054)  ; %6=data2_6H
+
+    vpaddd      %2, %2, GOTOFF(ebx, PD_DESCALE_P %+ %9)
+    vpaddd      %6, %6, GOTOFF(ebx, PD_DESCALE_P %+ %9)
+    vpsrad      %2, %2, DESCALE_P %+ %9
+    vpsrad      %6, %6, DESCALE_P %+ %9
+
+    vpackssdw   %3, %2, %6              ; %6=data2_6
+
+    ; -- Odd part
+
+    vpaddw      %7, %8, %5              ; %7=tmp4_5+tmp6_7=z3_4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    vperm2i128  %2, %7, %7, 0x01        ; %2=z4_3
+    vpunpcklwd  %6, %7, %2
+    vpunpckhwd  %7, %7, %2
+    vpmaddwd    %6, %6, GOTOFF(ebx, PW_MF078_F117_F078_F117)  ; %6=z3_4L
+    vpmaddwd    %7, %7, GOTOFF(ebx, PW_MF078_F117_F078_F117)  ; %7=z3_4H
+
+    ; (Original)
+    ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+    ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+    ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+    ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+    ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+    ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+    ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+    ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+    ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+    vperm2i128  %4, %5, %5, 0x01        ; %4=tmp7_6
+    vpunpcklwd  %2, %8, %4
+    vpunpckhwd  %4, %8, %4
+    vpmaddwd    %2, %2, GOTOFF(ebx, PW_MF060_MF089_MF050_MF256)  ; %2=tmp4_5L
+    vpmaddwd    %4, %4, GOTOFF(ebx, PW_MF060_MF089_MF050_MF256)  ; %4=tmp4_5H
+
+    vpaddd      %2, %2, %6              ; %2=data7_5L
+    vpaddd      %4, %4, %7              ; %4=data7_5H
+
+    vpaddd      %2, %2, GOTOFF(ebx, PD_DESCALE_P %+ %9)
+    vpaddd      %4, %4, GOTOFF(ebx, PD_DESCALE_P %+ %9)
+    vpsrad      %2, %2, DESCALE_P %+ %9
+    vpsrad      %4, %4, DESCALE_P %+ %9
+
+    vpackssdw   %4, %2, %4              ; %4=data7_5
+
+    vperm2i128  %2, %8, %8, 0x01        ; %2=tmp5_4
+    vpunpcklwd  %8, %5, %2
+    vpunpckhwd  %5, %5, %2
+    vpmaddwd    %8, %8, GOTOFF(ebx, PW_F050_MF256_F060_MF089)  ; %8=tmp6_7L
+    vpmaddwd    %5, %5, GOTOFF(ebx, PW_F050_MF256_F060_MF089)  ; %5=tmp6_7H
+
+    vpaddd      %8, %8, %6              ; %8=data3_1L
+    vpaddd      %5, %5, %7              ; %5=data3_1H
+
+    vpaddd      %8, %8, GOTOFF(ebx, PD_DESCALE_P %+ %9)
+    vpaddd      %5, %5, GOTOFF(ebx, PD_DESCALE_P %+ %9)
+    vpsrad      %8, %8, DESCALE_P %+ %9
+    vpsrad      %5, %5, DESCALE_P %+ %9
+
+    vpackssdw   %2, %8, %5              ; %2=data3_1
+%endmacro
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_islow_avx2)
+
+EXTN(jconst_fdct_islow_avx2):
+
+PW_F130_F054_MF130_F054    times 4  dw  (F_0_541 + F_0_765),  F_0_541
+                           times 4  dw  (F_0_541 - F_1_847),  F_0_541
+PW_MF078_F117_F078_F117    times 4  dw  (F_1_175 - F_1_961),  F_1_175
+                           times 4  dw  (F_1_175 - F_0_390),  F_1_175
+PW_MF060_MF089_MF050_MF256 times 4  dw  (F_0_298 - F_0_899), -F_0_899
+                           times 4  dw  (F_2_053 - F_2_562), -F_2_562
+PW_F050_MF256_F060_MF089   times 4  dw  (F_3_072 - F_2_562), -F_2_562
+                           times 4  dw  (F_1_501 - F_0_899), -F_0_899
+PD_DESCALE_P1              times 8  dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2              times 8  dd  1 << (DESCALE_P2 - 1)
+PW_DESCALE_P2X             times 16 dw  1 << (PASS1_BITS - 1)
+PW_1_NEG1                  times 8  dw  1
+                           times 8  dw -1
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_avx2(DCTELEM *data)
+;
+
+%define data(b)       (b) + 8           ; DCTELEM *data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_islow_avx2)
+
+EXTN(jsimd_fdct_islow_avx2):
+    push        ebp
+    mov         ebp, esp
+    pushpic     ebx
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+;   push        esi                     ; unused
+;   push        edi                     ; unused
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process rows.
+
+    mov         edx, POINTER data(ebp)  ; (DCTELEM *)
+
+    vmovdqu     ymm4, YMMWORD YMMBLOCK(0,0,edx,SIZEOF_DCTELEM)
+    vmovdqu     ymm5, YMMWORD YMMBLOCK(2,0,edx,SIZEOF_DCTELEM)
+    vmovdqu     ymm6, YMMWORD YMMBLOCK(4,0,edx,SIZEOF_DCTELEM)
+    vmovdqu     ymm7, YMMWORD YMMBLOCK(6,0,edx,SIZEOF_DCTELEM)
+    ; ymm4=(00 01 02 03 04 05 06 07  10 11 12 13 14 15 16 17)
+    ; ymm5=(20 21 22 23 24 25 26 27  30 31 32 33 34 35 36 37)
+    ; ymm6=(40 41 42 43 44 45 46 47  50 51 52 53 54 55 56 57)
+    ; ymm7=(60 61 62 63 64 65 66 67  70 71 72 73 74 75 76 77)
+
+    vperm2i128  ymm0, ymm4, ymm6, 0x20
+    vperm2i128  ymm1, ymm4, ymm6, 0x31
+    vperm2i128  ymm2, ymm5, ymm7, 0x20
+    vperm2i128  ymm3, ymm5, ymm7, 0x31
+    ; ymm0=(00 01 02 03 04 05 06 07  40 41 42 43 44 45 46 47)
+    ; ymm1=(10 11 12 13 14 15 16 17  50 51 52 53 54 55 56 57)
+    ; ymm2=(20 21 22 23 24 25 26 27  60 61 62 63 64 65 66 67)
+    ; ymm3=(30 31 32 33 34 35 36 37  70 71 72 73 74 75 76 77)
+
+    dotranspose ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7
+
+    dodct       ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7, 1
+    ; ymm0=data0_4, ymm1=data3_1, ymm2=data2_6, ymm3=data7_5
+
+    ; ---- Pass 2: process columns.
+
+    vperm2i128  ymm4, ymm1, ymm3, 0x20  ; ymm4=data3_7
+    vperm2i128  ymm1, ymm1, ymm3, 0x31  ; ymm1=data1_5
+
+    dotranspose ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7
+
+    dodct       ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7, 2
+    ; ymm0=data0_4, ymm1=data3_1, ymm2=data2_6, ymm4=data7_5
+
+    vperm2i128 ymm3, ymm0, ymm1, 0x30   ; ymm3=data0_1
+    vperm2i128 ymm5, ymm2, ymm1, 0x20   ; ymm5=data2_3
+    vperm2i128 ymm6, ymm0, ymm4, 0x31   ; ymm6=data4_5
+    vperm2i128 ymm7, ymm2, ymm4, 0x21   ; ymm7=data6_7
+
+    vmovdqu     YMMWORD YMMBLOCK(0,0,edx,SIZEOF_DCTELEM), ymm3
+    vmovdqu     YMMWORD YMMBLOCK(2,0,edx,SIZEOF_DCTELEM), ymm5
+    vmovdqu     YMMWORD YMMBLOCK(4,0,edx,SIZEOF_DCTELEM), ymm6
+    vmovdqu     YMMWORD YMMBLOCK(6,0,edx,SIZEOF_DCTELEM), ymm7
+
+    vzeroupper
+;   pop         edi                     ; unused
+;   pop         esi                     ; unused
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+    poppic      ebx
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jfdctint-mmx.asm Added

@@ -0,0 +1,620 @@
+;
+; jfdctint.asm - accurate integer FDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, 2020, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_islow_mmx)
+
+EXTN(jconst_fdct_islow_mmx):
+
+PW_F130_F054   times 2 dw  (F_0_541 + F_0_765),  F_0_541
+PW_F054_MF130  times 2 dw  F_0_541, (F_0_541 - F_1_847)
+PW_MF078_F117  times 2 dw  (F_1_175 - F_1_961),  F_1_175
+PW_F117_F078   times 2 dw  F_1_175, (F_1_175 - F_0_390)
+PW_MF060_MF089 times 2 dw  (F_0_298 - F_0_899), -F_0_899
+PW_MF089_F060  times 2 dw -F_0_899, (F_1_501 - F_0_899)
+PW_MF050_MF256 times 2 dw  (F_2_053 - F_2_562), -F_2_562
+PW_MF256_F050  times 2 dw -F_2_562, (F_3_072 - F_2_562)
+PD_DESCALE_P1  times 2 dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2  times 2 dd  1 << (DESCALE_P2 - 1)
+PW_DESCALE_P2X times 4 dw  1 << (PASS1_BITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_mmx(DCTELEM *data)
+;
+
+%define data(b)       (b) + 8           ; DCTELEM *data
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_MMWORD  ; mmword wkWK_NUM
+%define WK_NUM        2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_islow_mmx)
+
+EXTN(jsimd_fdct_islow_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+;   push        esi                     ; unused
+;   push        edi                     ; unused
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process rows.
+
+    mov         edx, POINTER data(eax)  ; (DCTELEM *)
+    mov         ecx, DCTSIZE/4
+    alignx      16, 7
+.rowloop:
+
+    movq        mm0, MMWORD MMBLOCK(2,0,edx,SIZEOF_DCTELEM)
+    movq        mm1, MMWORD MMBLOCK(3,0,edx,SIZEOF_DCTELEM)
+    movq        mm2, MMWORD MMBLOCK(2,1,edx,SIZEOF_DCTELEM)
+    movq        mm3, MMWORD MMBLOCK(3,1,edx,SIZEOF_DCTELEM)
+
+    ; mm0=(20 21 22 23), mm2=(24 25 26 27)
+    ; mm1=(30 31 32 33), mm3=(34 35 36 37)
+
+    movq        mm4, mm0                ; transpose coefficients(phase 1)
+    punpcklwd   mm0, mm1                ; mm0=(20 30 21 31)
+    punpckhwd   mm4, mm1                ; mm4=(22 32 23 33)
+    movq        mm5, mm2                ; transpose coefficients(phase 1)
+    punpcklwd   mm2, mm3                ; mm2=(24 34 25 35)
+    punpckhwd   mm5, mm3                ; mm5=(26 36 27 37)
+
+    movq        mm6, MMWORD MMBLOCK(0,0,edx,SIZEOF_DCTELEM)
+    movq        mm7, MMWORD MMBLOCK(1,0,edx,SIZEOF_DCTELEM)
+    movq        mm1, MMWORD MMBLOCK(0,1,edx,SIZEOF_DCTELEM)
+    movq        mm3, MMWORD MMBLOCK(1,1,edx,SIZEOF_DCTELEM)
+
+    ; mm6=(00 01 02 03), mm1=(04 05 06 07)
+    ; mm7=(10 11 12 13), mm3=(14 15 16 17)
+
+    movq        MMWORD wk(0), mm4     ; wk(0)=(22 32 23 33)
+    movq        MMWORD wk(1), mm2     ; wk(1)=(24 34 25 35)
+
+    movq        mm4, mm6                ; transpose coefficients(phase 1)
+    punpcklwd   mm6, mm7                ; mm6=(00 10 01 11)
+    punpckhwd   mm4, mm7                ; mm4=(02 12 03 13)
+    movq        mm2, mm1                ; transpose coefficients(phase 1)
+    punpcklwd   mm1, mm3                ; mm1=(04 14 05 15)
+    punpckhwd   mm2, mm3                ; mm2=(06 16 07 17)
+
+    movq        mm7, mm6                ; transpose coefficients(phase 2)
+    punpckldq   mm6, mm0                ; mm6=(00 10 20 30)=data0
+    punpckhdq   mm7, mm0                ; mm7=(01 11 21 31)=data1
+    movq        mm3, mm2                ; transpose coefficients(phase 2)
+    punpckldq   mm2, mm5                ; mm2=(06 16 26 36)=data6
+    punpckhdq   mm3, mm5                ; mm3=(07 17 27 37)=data7
+
+    movq        mm0, mm7
+    movq        mm5, mm6
+    psubw       mm7, mm2                ; mm7=data1-data6=tmp6
+    psubw       mm6, mm3                ; mm6=data0-data7=tmp7
+    paddw       mm0, mm2                ; mm0=data1+data6=tmp1
+    paddw       mm5, mm3                ; mm5=data0+data7=tmp0
+
+    movq        mm2, MMWORD wk(0)     ; mm2=(22 32 23 33)
+    movq        mm3, MMWORD wk(1)     ; mm3=(24 34 25 35)
+    movq        MMWORD wk(0), mm7     ; wk(0)=tmp6
+    movq        MMWORD wk(1), mm6     ; wk(1)=tmp7
+
+    movq        mm7, mm4                ; transpose coefficients(phase 2)
+    punpckldq   mm4, mm2                ; mm4=(02 12 22 32)=data2
+    punpckhdq   mm7, mm2                ; mm7=(03 13 23 33)=data3
+    movq        mm6, mm1                ; transpose coefficients(phase 2)
+    punpckldq   mm1, mm3                ; mm1=(04 14 24 34)=data4
+    punpckhdq   mm6, mm3                ; mm6=(05 15 25 35)=data5
+
+    movq        mm2, mm7
+    movq        mm3, mm4
+    paddw       mm7, mm1                ; mm7=data3+data4=tmp3
+    paddw       mm4, mm6                ; mm4=data2+data5=tmp2
+    psubw       mm2, mm1                ; mm2=data3-data4=tmp4
+    psubw       mm3, mm6                ; mm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movq        mm1, mm5
+    movq        mm6, mm0
+    paddw       mm5, mm7                ; mm5=tmp10
+    paddw       mm0, mm4                ; mm0=tmp11
+    psubw       mm1, mm7                ; mm1=tmp13
+    psubw       mm6, mm4                ; mm6=tmp12
+
+    movq        mm7, mm5
+    paddw       mm5, mm0                ; mm5=tmp10+tmp11
+    psubw       mm7, mm0                ; mm7=tmp10-tmp11
+
+    psllw       mm5, PASS1_BITS         ; mm5=data0
+    psllw       mm7, PASS1_BITS         ; mm7=data4
+
+    movq        MMWORD MMBLOCK(0,0,edx,SIZEOF_DCTELEM), mm5
+    movq        MMWORD MMBLOCK(0,1,edx,SIZEOF_DCTELEM), mm7
+
+    ; (Original)
+    ; z1 = (tmp12 + tmp13) * 0.541196100;
+    ; data2 = z1 + tmp13 * 0.765366865;
+    ; data6 = z1 + tmp12 * -1.847759065;
+    ;
+    ; (This implementation)
+    ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+    ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+    movq        mm4, mm1                ; mm1=tmp13
+    movq        mm0, mm1
+    punpcklwd   mm4, mm6                ; mm6=tmp12
+    punpckhwd   mm0, mm6
+    movq        mm1, mm4
+    movq        mm6, mm0
+    pmaddwd     mm4, GOTOFF(ebx,PW_F130_F054)   ; mm4=data2L
+    pmaddwd     mm0, GOTOFF(ebx,PW_F130_F054)   ; mm0=data2H
+    pmaddwd     mm1, GOTOFF(ebx,PW_F054_MF130)  ; mm1=data6L
+    pmaddwd     mm6, GOTOFF(ebx,PW_F054_MF130)  ; mm6=data6H
+
+    paddd       mm4, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       mm0, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       mm4, DESCALE_P1
+    psrad       mm0, DESCALE_P1
+    paddd       mm1, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       mm6, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       mm1, DESCALE_P1
+    psrad       mm6, DESCALE_P1
+
+    packssdw    mm4, mm0                ; mm4=data2
+    packssdw    mm1, mm6                ; mm1=data6
+
+    movq        MMWORD MMBLOCK(2,0,edx,SIZEOF_DCTELEM), mm4
+    movq        MMWORD MMBLOCK(2,1,edx,SIZEOF_DCTELEM), mm1
+
+    ; -- Odd part
+
+    movq        mm5, MMWORD wk(0)     ; mm5=tmp6
+    movq        mm7, MMWORD wk(1)     ; mm7=tmp7
+
+    movq        mm0, mm2                ; mm2=tmp4
+    movq        mm6, mm3                ; mm3=tmp5
+    paddw       mm0, mm5                ; mm0=z3
+    paddw       mm6, mm7                ; mm6=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movq        mm4, mm0
+    movq        mm1, mm0
+    punpcklwd   mm4, mm6
+    punpckhwd   mm1, mm6
+    movq        mm0, mm4
+    movq        mm6, mm1
+    pmaddwd     mm4, GOTOFF(ebx,PW_MF078_F117)  ; mm4=z3L
+    pmaddwd     mm1, GOTOFF(ebx,PW_MF078_F117)  ; mm1=z3H
+    pmaddwd     mm0, GOTOFF(ebx,PW_F117_F078)   ; mm0=z4L
+    pmaddwd     mm6, GOTOFF(ebx,PW_F117_F078)   ; mm6=z4H
+
+    movq        MMWORD wk(0), mm4     ; wk(0)=z3L
+    movq        MMWORD wk(1), mm1     ; wk(1)=z3H
+
+    ; (Original)
+    ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+    ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+    ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+    ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+    ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+    ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+    ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+    ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+    ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+    movq        mm4, mm2
+    movq        mm1, mm2
+    punpcklwd   mm4, mm7
+    punpckhwd   mm1, mm7
+    movq        mm2, mm4
+    movq        mm7, mm1
+    pmaddwd     mm4, GOTOFF(ebx,PW_MF060_MF089)  ; mm4=tmp4L
+    pmaddwd     mm1, GOTOFF(ebx,PW_MF060_MF089)  ; mm1=tmp4H
+    pmaddwd     mm2, GOTOFF(ebx,PW_MF089_F060)   ; mm2=tmp7L
+    pmaddwd     mm7, GOTOFF(ebx,PW_MF089_F060)   ; mm7=tmp7H
+
+    paddd       mm4, MMWORD wk(0)     ; mm4=data7L
+    paddd       mm1, MMWORD wk(1)     ; mm1=data7H
+    paddd       mm2, mm0                ; mm2=data1L
+    paddd       mm7, mm6                ; mm7=data1H
+
+    paddd       mm4, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       mm1, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       mm4, DESCALE_P1
+    psrad       mm1, DESCALE_P1
+    paddd       mm2, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       mm7, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       mm2, DESCALE_P1
+    psrad       mm7, DESCALE_P1
+
+    packssdw    mm4, mm1                ; mm4=data7
+    packssdw    mm2, mm7                ; mm2=data1
+
+    movq        MMWORD MMBLOCK(3,1,edx,SIZEOF_DCTELEM), mm4
+    movq        MMWORD MMBLOCK(1,0,edx,SIZEOF_DCTELEM), mm2
+
+    movq        mm1, mm3
+    movq        mm7, mm3
+    punpcklwd   mm1, mm5
+    punpckhwd   mm7, mm5
+    movq        mm3, mm1
+    movq        mm5, mm7
+    pmaddwd     mm1, GOTOFF(ebx,PW_MF050_MF256)  ; mm1=tmp5L
+    pmaddwd     mm7, GOTOFF(ebx,PW_MF050_MF256)  ; mm7=tmp5H
+    pmaddwd     mm3, GOTOFF(ebx,PW_MF256_F050)   ; mm3=tmp6L
+    pmaddwd     mm5, GOTOFF(ebx,PW_MF256_F050)   ; mm5=tmp6H
+
+    paddd       mm1, mm0                ; mm1=data5L
+    paddd       mm7, mm6                ; mm7=data5H
+    paddd       mm3, MMWORD wk(0)     ; mm3=data3L
+    paddd       mm5, MMWORD wk(1)     ; mm5=data3H
+
+    paddd       mm1, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       mm7, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       mm1, DESCALE_P1
+    psrad       mm7, DESCALE_P1
+    paddd       mm3, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       mm5, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       mm3, DESCALE_P1
+    psrad       mm5, DESCALE_P1
+
+    packssdw    mm1, mm7                ; mm1=data5
+    packssdw    mm3, mm5                ; mm3=data3
+
+    movq        MMWORD MMBLOCK(1,1,edx,SIZEOF_DCTELEM), mm1
+    movq        MMWORD MMBLOCK(3,0,edx,SIZEOF_DCTELEM), mm3
+
+    add         edx, byte 4*DCTSIZE*SIZEOF_DCTELEM
+    dec         ecx
+    jnz         near .rowloop
+
+    ; ---- Pass 2: process columns.
+
+    mov         edx, POINTER data(eax)  ; (DCTELEM *)
+    mov         ecx, DCTSIZE/4
+    alignx      16, 7
+.columnloop:
+
+    movq        mm0, MMWORD MMBLOCK(2,0,edx,SIZEOF_DCTELEM)
+    movq        mm1, MMWORD MMBLOCK(3,0,edx,SIZEOF_DCTELEM)
+    movq        mm2, MMWORD MMBLOCK(6,0,edx,SIZEOF_DCTELEM)
+    movq        mm3, MMWORD MMBLOCK(7,0,edx,SIZEOF_DCTELEM)
+
+    ; mm0=(02 12 22 32), mm2=(42 52 62 72)
+    ; mm1=(03 13 23 33), mm3=(43 53 63 73)
+
+    movq        mm4, mm0                ; transpose coefficients(phase 1)
+    punpcklwd   mm0, mm1                ; mm0=(02 03 12 13)
+    punpckhwd   mm4, mm1                ; mm4=(22 23 32 33)
+    movq        mm5, mm2                ; transpose coefficients(phase 1)
+    punpcklwd   mm2, mm3                ; mm2=(42 43 52 53)
+    punpckhwd   mm5, mm3                ; mm5=(62 63 72 73)
+
+    movq        mm6, MMWORD MMBLOCK(0,0,edx,SIZEOF_DCTELEM)
+    movq        mm7, MMWORD MMBLOCK(1,0,edx,SIZEOF_DCTELEM)
+    movq        mm1, MMWORD MMBLOCK(4,0,edx,SIZEOF_DCTELEM)
+    movq        mm3, MMWORD MMBLOCK(5,0,edx,SIZEOF_DCTELEM)
+
+    ; mm6=(00 10 20 30), mm1=(40 50 60 70)
+    ; mm7=(01 11 21 31), mm3=(41 51 61 71)
+
+    movq        MMWORD wk(0), mm4     ; wk(0)=(22 23 32 33)
+    movq        MMWORD wk(1), mm2     ; wk(1)=(42 43 52 53)
+
+    movq        mm4, mm6                ; transpose coefficients(phase 1)
+    punpcklwd   mm6, mm7                ; mm6=(00 01 10 11)
+    punpckhwd   mm4, mm7                ; mm4=(20 21 30 31)
+    movq        mm2, mm1                ; transpose coefficients(phase 1)
+    punpcklwd   mm1, mm3                ; mm1=(40 41 50 51)
+    punpckhwd   mm2, mm3                ; mm2=(60 61 70 71)
+
+    movq        mm7, mm6                ; transpose coefficients(phase 2)
+    punpckldq   mm6, mm0                ; mm6=(00 01 02 03)=data0
+    punpckhdq   mm7, mm0                ; mm7=(10 11 12 13)=data1
+    movq        mm3, mm2                ; transpose coefficients(phase 2)
+    punpckldq   mm2, mm5                ; mm2=(60 61 62 63)=data6
+    punpckhdq   mm3, mm5                ; mm3=(70 71 72 73)=data7
+
+    movq        mm0, mm7
+    movq        mm5, mm6
+    psubw       mm7, mm2                ; mm7=data1-data6=tmp6
+    psubw       mm6, mm3                ; mm6=data0-data7=tmp7
+    paddw       mm0, mm2                ; mm0=data1+data6=tmp1
+    paddw       mm5, mm3                ; mm5=data0+data7=tmp0
+
+    movq        mm2, MMWORD wk(0)     ; mm2=(22 23 32 33)
+    movq        mm3, MMWORD wk(1)     ; mm3=(42 43 52 53)
+    movq        MMWORD wk(0), mm7     ; wk(0)=tmp6
+    movq        MMWORD wk(1), mm6     ; wk(1)=tmp7
+
+    movq        mm7, mm4                ; transpose coefficients(phase 2)
+    punpckldq   mm4, mm2                ; mm4=(20 21 22 23)=data2
+    punpckhdq   mm7, mm2                ; mm7=(30 31 32 33)=data3
+    movq        mm6, mm1                ; transpose coefficients(phase 2)
+    punpckldq   mm1, mm3                ; mm1=(40 41 42 43)=data4
+    punpckhdq   mm6, mm3                ; mm6=(50 51 52 53)=data5
+
+    movq        mm2, mm7
+    movq        mm3, mm4
+    paddw       mm7, mm1                ; mm7=data3+data4=tmp3
+    paddw       mm4, mm6                ; mm4=data2+data5=tmp2
+    psubw       mm2, mm1                ; mm2=data3-data4=tmp4
+    psubw       mm3, mm6                ; mm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movq        mm1, mm5
+    movq        mm6, mm0
+    paddw       mm5, mm7                ; mm5=tmp10
+    paddw       mm0, mm4                ; mm0=tmp11
+    psubw       mm1, mm7                ; mm1=tmp13
+    psubw       mm6, mm4                ; mm6=tmp12
+
+    movq        mm7, mm5
+    paddw       mm5, mm0                ; mm5=tmp10+tmp11
+    psubw       mm7, mm0                ; mm7=tmp10-tmp11
+
+    paddw       mm5, GOTOFF(ebx,PW_DESCALE_P2X)
+    paddw       mm7, GOTOFF(ebx,PW_DESCALE_P2X)
+    psraw       mm5, PASS1_BITS         ; mm5=data0
+    psraw       mm7, PASS1_BITS         ; mm7=data4
+
+    movq        MMWORD MMBLOCK(0,0,edx,SIZEOF_DCTELEM), mm5
+    movq        MMWORD MMBLOCK(4,0,edx,SIZEOF_DCTELEM), mm7
+
+    ; (Original)
+    ; z1 = (tmp12 + tmp13) * 0.541196100;
+    ; data2 = z1 + tmp13 * 0.765366865;
+    ; data6 = z1 + tmp12 * -1.847759065;
+    ;
+    ; (This implementation)
+    ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+    ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+    movq        mm4, mm1                ; mm1=tmp13
+    movq        mm0, mm1
+    punpcklwd   mm4, mm6                ; mm6=tmp12
+    punpckhwd   mm0, mm6
+    movq        mm1, mm4
+    movq        mm6, mm0
+    pmaddwd     mm4, GOTOFF(ebx,PW_F130_F054)   ; mm4=data2L
+    pmaddwd     mm0, GOTOFF(ebx,PW_F130_F054)   ; mm0=data2H
+    pmaddwd     mm1, GOTOFF(ebx,PW_F054_MF130)  ; mm1=data6L
+    pmaddwd     mm6, GOTOFF(ebx,PW_F054_MF130)  ; mm6=data6H
+
+    paddd       mm4, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       mm0, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       mm4, DESCALE_P2
+    psrad       mm0, DESCALE_P2
+    paddd       mm1, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       mm6, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       mm1, DESCALE_P2
+    psrad       mm6, DESCALE_P2
+
+    packssdw    mm4, mm0                ; mm4=data2
+    packssdw    mm1, mm6                ; mm1=data6
+
+    movq        MMWORD MMBLOCK(2,0,edx,SIZEOF_DCTELEM), mm4
+    movq        MMWORD MMBLOCK(6,0,edx,SIZEOF_DCTELEM), mm1
+
+    ; -- Odd part
+
+    movq        mm5, MMWORD wk(0)     ; mm5=tmp6
+    movq        mm7, MMWORD wk(1)     ; mm7=tmp7
+
+    movq        mm0, mm2                ; mm2=tmp4
+    movq        mm6, mm3                ; mm3=tmp5
+    paddw       mm0, mm5                ; mm0=z3
+    paddw       mm6, mm7                ; mm6=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movq        mm4, mm0
+    movq        mm1, mm0
+    punpcklwd   mm4, mm6
+    punpckhwd   mm1, mm6
+    movq        mm0, mm4
+    movq        mm6, mm1
+    pmaddwd     mm4, GOTOFF(ebx,PW_MF078_F117)  ; mm4=z3L
+    pmaddwd     mm1, GOTOFF(ebx,PW_MF078_F117)  ; mm1=z3H
+    pmaddwd     mm0, GOTOFF(ebx,PW_F117_F078)   ; mm0=z4L
+    pmaddwd     mm6, GOTOFF(ebx,PW_F117_F078)   ; mm6=z4H
+
+    movq        MMWORD wk(0), mm4     ; wk(0)=z3L
+    movq        MMWORD wk(1), mm1     ; wk(1)=z3H
+
+    ; (Original)
+    ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+    ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+    ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+    ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+    ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+    ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+    ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+    ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+    ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+    movq        mm4, mm2
+    movq        mm1, mm2
+    punpcklwd   mm4, mm7
+    punpckhwd   mm1, mm7
+    movq        mm2, mm4
+    movq        mm7, mm1
+    pmaddwd     mm4, GOTOFF(ebx,PW_MF060_MF089)  ; mm4=tmp4L
+    pmaddwd     mm1, GOTOFF(ebx,PW_MF060_MF089)  ; mm1=tmp4H
+    pmaddwd     mm2, GOTOFF(ebx,PW_MF089_F060)   ; mm2=tmp7L
+    pmaddwd     mm7, GOTOFF(ebx,PW_MF089_F060)   ; mm7=tmp7H
+
+    paddd       mm4, MMWORD wk(0)     ; mm4=data7L
+    paddd       mm1, MMWORD wk(1)     ; mm1=data7H
+    paddd       mm2, mm0                ; mm2=data1L
+    paddd       mm7, mm6                ; mm7=data1H
+
+    paddd       mm4, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       mm1, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       mm4, DESCALE_P2
+    psrad       mm1, DESCALE_P2
+    paddd       mm2, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       mm7, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       mm2, DESCALE_P2
+    psrad       mm7, DESCALE_P2
+
+    packssdw    mm4, mm1                ; mm4=data7
+    packssdw    mm2, mm7                ; mm2=data1
+
+    movq        MMWORD MMBLOCK(7,0,edx,SIZEOF_DCTELEM), mm4
+    movq        MMWORD MMBLOCK(1,0,edx,SIZEOF_DCTELEM), mm2
+
+    movq        mm1, mm3
+    movq        mm7, mm3
+    punpcklwd   mm1, mm5
+    punpckhwd   mm7, mm5
+    movq        mm3, mm1
+    movq        mm5, mm7
+    pmaddwd     mm1, GOTOFF(ebx,PW_MF050_MF256)  ; mm1=tmp5L
+    pmaddwd     mm7, GOTOFF(ebx,PW_MF050_MF256)  ; mm7=tmp5H
+    pmaddwd     mm3, GOTOFF(ebx,PW_MF256_F050)   ; mm3=tmp6L
+    pmaddwd     mm5, GOTOFF(ebx,PW_MF256_F050)   ; mm5=tmp6H
+
+    paddd       mm1, mm0                ; mm1=data5L
+    paddd       mm7, mm6                ; mm7=data5H
+    paddd       mm3, MMWORD wk(0)     ; mm3=data3L
+    paddd       mm5, MMWORD wk(1)     ; mm5=data3H
+
+    paddd       mm1, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       mm7, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       mm1, DESCALE_P2
+    psrad       mm7, DESCALE_P2
+    paddd       mm3, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       mm5, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       mm3, DESCALE_P2
+    psrad       mm5, DESCALE_P2
+
+    packssdw    mm1, mm7                ; mm1=data5
+    packssdw    mm3, mm5                ; mm3=data3
+
+    movq        MMWORD MMBLOCK(5,0,edx,SIZEOF_DCTELEM), mm1
+    movq        MMWORD MMBLOCK(3,0,edx,SIZEOF_DCTELEM), mm3
+
+    add         edx, byte 4*SIZEOF_DCTELEM
+    dec         ecx
+    jnz         near .columnloop
+
+    emms                                ; empty MMX state
+
+;   pop         edi                     ; unused
+;   pop         esi                     ; unused
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jfdctint-sse2.asm Added

@@ -0,0 +1,633 @@
+;
+; jfdctint.asm - accurate integer FDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, 2020, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_islow_sse2)
+
+EXTN(jconst_fdct_islow_sse2):
+
+PW_F130_F054   times 4 dw  (F_0_541 + F_0_765),  F_0_541
+PW_F054_MF130  times 4 dw  F_0_541, (F_0_541 - F_1_847)
+PW_MF078_F117  times 4 dw  (F_1_175 - F_1_961),  F_1_175
+PW_F117_F078   times 4 dw  F_1_175, (F_1_175 - F_0_390)
+PW_MF060_MF089 times 4 dw  (F_0_298 - F_0_899), -F_0_899
+PW_MF089_F060  times 4 dw -F_0_899, (F_1_501 - F_0_899)
+PW_MF050_MF256 times 4 dw  (F_2_053 - F_2_562), -F_2_562
+PW_MF256_F050  times 4 dw -F_2_562, (F_3_072 - F_2_562)
+PD_DESCALE_P1  times 4 dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2  times 4 dd  1 << (DESCALE_P2 - 1)
+PW_DESCALE_P2X times 8 dw  1 << (PASS1_BITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_sse2(DCTELEM *data)
+;
+
+%define data(b)       (b) + 8           ; DCTELEM *data
+
+%define original_ebp  ebp + 0
+%define wk(i)         ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM        6
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_islow_sse2)
+
+EXTN(jsimd_fdct_islow_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+;   push        esi                     ; unused
+;   push        edi                     ; unused
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process rows.
+
+    mov         edx, POINTER data(eax)  ; (DCTELEM *)
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm1, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm2, XMMWORD XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm3, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)
+
+    ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+    ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+    movdqa      xmm4, xmm0              ; transpose coefficients(phase 1)
+    punpcklwd   xmm0, xmm1              ; xmm0=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm4, xmm1              ; xmm4=(04 14 05 15 06 16 07 17)
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 1)
+    punpcklwd   xmm2, xmm3              ; xmm2=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm5, xmm3              ; xmm5=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm6, XMMWORD XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm7, XMMWORD XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm1, XMMWORD XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)
+
+    ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+    ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+    movdqa      XMMWORD wk(1), xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm2, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm7              ; xmm6=(40 50 41 51 42 52 43 53)
+    punpckhwd   xmm2, xmm7              ; xmm2=(44 54 45 55 46 56 47 57)
+    movdqa      xmm5, xmm1              ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm3              ; xmm1=(60 70 61 71 62 72 63 73)
+    punpckhwd   xmm5, xmm3              ; xmm5=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm7, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm1              ; xmm6=(40 50 60 70 41 51 61 71)
+    punpckhdq   xmm7, xmm1              ; xmm7=(42 52 62 72 43 53 63 73)
+    movdqa      xmm3, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm5              ; xmm2=(44 54 64 74 45 55 65 75)
+    punpckhdq   xmm3, xmm5              ; xmm3=(46 56 66 76 47 57 67 77)
+
+    movdqa      xmm1, XMMWORD wk(0)   ; xmm1=(20 30 21 31 22 32 23 33)
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=(24 34 25 35 26 36 27 37)
+    movdqa      XMMWORD wk(2), xmm7   ; wk(2)=(42 52 62 72 43 53 63 73)
+    movdqa      XMMWORD wk(3), xmm2   ; wk(3)=(44 54 64 74 45 55 65 75)
+
+    movdqa      xmm7, xmm0              ; transpose coefficients(phase 2)
+    punpckldq   xmm0, xmm1              ; xmm0=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm7, xmm1              ; xmm7=(02 12 22 32 03 13 23 33)
+    movdqa      xmm2, xmm4              ; transpose coefficients(phase 2)
+    punpckldq   xmm4, xmm5              ; xmm4=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm2, xmm5              ; xmm2=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm1, xmm0              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm0, xmm6              ; xmm0=(00 10 20 30 40 50 60 70)=data0
+    punpckhqdq  xmm1, xmm6              ; xmm1=(01 11 21 31 41 51 61 71)=data1
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm2, xmm3              ; xmm2=(06 16 26 36 46 56 66 76)=data6
+    punpckhqdq  xmm5, xmm3              ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+    movdqa      xmm6, xmm1
+    movdqa      xmm3, xmm0
+    psubw       xmm1, xmm2              ; xmm1=data1-data6=tmp6
+    psubw       xmm0, xmm5              ; xmm0=data0-data7=tmp7
+    paddw       xmm6, xmm2              ; xmm6=data1+data6=tmp1
+    paddw       xmm3, xmm5              ; xmm3=data0+data7=tmp0
+
+    movdqa      xmm2, XMMWORD wk(2)   ; xmm2=(42 52 62 72 43 53 63 73)
+    movdqa      xmm5, XMMWORD wk(3)   ; xmm5=(44 54 64 74 45 55 65 75)
+    movdqa      XMMWORD wk(0), xmm1   ; wk(0)=tmp6
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=tmp7
+
+    movdqa      xmm1, xmm7              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm7, xmm2              ; xmm7=(02 12 22 32 42 52 62 72)=data2
+    punpckhqdq  xmm1, xmm2              ; xmm1=(03 13 23 33 43 53 63 73)=data3
+    movdqa      xmm0, xmm4              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm4, xmm5              ; xmm4=(04 14 24 34 44 54 64 74)=data4
+    punpckhqdq  xmm0, xmm5              ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+    movdqa      xmm2, xmm1
+    movdqa      xmm5, xmm7
+    paddw       xmm1, xmm4              ; xmm1=data3+data4=tmp3
+    paddw       xmm7, xmm0              ; xmm7=data2+data5=tmp2
+    psubw       xmm2, xmm4              ; xmm2=data3-data4=tmp4
+    psubw       xmm5, xmm0              ; xmm5=data2-data5=tmp5
+
+    ; -- Even part
+
+    movdqa      xmm4, xmm3
+    movdqa      xmm0, xmm6
+    paddw       xmm3, xmm1              ; xmm3=tmp10
+    paddw       xmm6, xmm7              ; xmm6=tmp11
+    psubw       xmm4, xmm1              ; xmm4=tmp13
+    psubw       xmm0, xmm7              ; xmm0=tmp12
+
+    movdqa      xmm1, xmm3
+    paddw       xmm3, xmm6              ; xmm3=tmp10+tmp11
+    psubw       xmm1, xmm6              ; xmm1=tmp10-tmp11
+
+    psllw       xmm3, PASS1_BITS        ; xmm3=data0
+    psllw       xmm1, PASS1_BITS        ; xmm1=data4
+
+    movdqa      XMMWORD wk(2), xmm3   ; wk(2)=data0
+    movdqa      XMMWORD wk(3), xmm1   ; wk(3)=data4
+
+    ; (Original)
+    ; z1 = (tmp12 + tmp13) * 0.541196100;
+    ; data2 = z1 + tmp13 * 0.765366865;
+    ; data6 = z1 + tmp12 * -1.847759065;
+    ;
+    ; (This implementation)
+    ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+    ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+    movdqa      xmm7, xmm4              ; xmm4=tmp13
+    movdqa      xmm6, xmm4
+    punpcklwd   xmm7, xmm0              ; xmm0=tmp12
+    punpckhwd   xmm6, xmm0
+    movdqa      xmm4, xmm7
+    movdqa      xmm0, xmm6
+    pmaddwd     xmm7, GOTOFF(ebx,PW_F130_F054)   ; xmm7=data2L
+    pmaddwd     xmm6, GOTOFF(ebx,PW_F130_F054)   ; xmm6=data2H
+    pmaddwd     xmm4, GOTOFF(ebx,PW_F054_MF130)  ; xmm4=data6L
+    pmaddwd     xmm0, GOTOFF(ebx,PW_F054_MF130)  ; xmm0=data6H
+
+    paddd       xmm7, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       xmm6, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       xmm7, DESCALE_P1
+    psrad       xmm6, DESCALE_P1
+    paddd       xmm4, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       xmm0, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       xmm4, DESCALE_P1
+    psrad       xmm0, DESCALE_P1
+
+    packssdw    xmm7, xmm6              ; xmm7=data2
+    packssdw    xmm4, xmm0              ; xmm4=data6
+
+    movdqa      XMMWORD wk(4), xmm7   ; wk(4)=data2
+    movdqa      XMMWORD wk(5), xmm4   ; wk(5)=data6
+
+    ; -- Odd part
+
+    movdqa      xmm3, XMMWORD wk(0)   ; xmm3=tmp6
+    movdqa      xmm1, XMMWORD wk(1)   ; xmm1=tmp7
+
+    movdqa      xmm6, xmm2              ; xmm2=tmp4
+    movdqa      xmm0, xmm5              ; xmm5=tmp5
+    paddw       xmm6, xmm3              ; xmm6=z3
+    paddw       xmm0, xmm1              ; xmm0=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movdqa      xmm7, xmm6
+    movdqa      xmm4, xmm6
+    punpcklwd   xmm7, xmm0
+    punpckhwd   xmm4, xmm0
+    movdqa      xmm6, xmm7
+    movdqa      xmm0, xmm4
+    pmaddwd     xmm7, GOTOFF(ebx,PW_MF078_F117)  ; xmm7=z3L
+    pmaddwd     xmm4, GOTOFF(ebx,PW_MF078_F117)  ; xmm4=z3H
+    pmaddwd     xmm6, GOTOFF(ebx,PW_F117_F078)   ; xmm6=z4L
+    pmaddwd     xmm0, GOTOFF(ebx,PW_F117_F078)   ; xmm0=z4H
+
+    movdqa      XMMWORD wk(0), xmm7   ; wk(0)=z3L
+    movdqa      XMMWORD wk(1), xmm4   ; wk(1)=z3H
+
+    ; (Original)
+    ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+    ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+    ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+    ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+    ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+    ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+    ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+    ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+    ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+    movdqa      xmm7, xmm2
+    movdqa      xmm4, xmm2
+    punpcklwd   xmm7, xmm1
+    punpckhwd   xmm4, xmm1
+    movdqa      xmm2, xmm7
+    movdqa      xmm1, xmm4
+    pmaddwd     xmm7, GOTOFF(ebx,PW_MF060_MF089)  ; xmm7=tmp4L
+    pmaddwd     xmm4, GOTOFF(ebx,PW_MF060_MF089)  ; xmm4=tmp4H
+    pmaddwd     xmm2, GOTOFF(ebx,PW_MF089_F060)   ; xmm2=tmp7L
+    pmaddwd     xmm1, GOTOFF(ebx,PW_MF089_F060)   ; xmm1=tmp7H
+
+    paddd       xmm7, XMMWORD wk(0)   ; xmm7=data7L
+    paddd       xmm4, XMMWORD wk(1)   ; xmm4=data7H
+    paddd       xmm2, xmm6              ; xmm2=data1L
+    paddd       xmm1, xmm0              ; xmm1=data1H
+
+    paddd       xmm7, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       xmm4, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       xmm7, DESCALE_P1
+    psrad       xmm4, DESCALE_P1
+    paddd       xmm2, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       xmm1, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       xmm2, DESCALE_P1
+    psrad       xmm1, DESCALE_P1
+
+    packssdw    xmm7, xmm4              ; xmm7=data7
+    packssdw    xmm2, xmm1              ; xmm2=data1
+
+    movdqa      xmm4, xmm5
+    movdqa      xmm1, xmm5
+    punpcklwd   xmm4, xmm3
+    punpckhwd   xmm1, xmm3
+    movdqa      xmm5, xmm4
+    movdqa      xmm3, xmm1
+    pmaddwd     xmm4, GOTOFF(ebx,PW_MF050_MF256)  ; xmm4=tmp5L
+    pmaddwd     xmm1, GOTOFF(ebx,PW_MF050_MF256)  ; xmm1=tmp5H
+    pmaddwd     xmm5, GOTOFF(ebx,PW_MF256_F050)   ; xmm5=tmp6L
+    pmaddwd     xmm3, GOTOFF(ebx,PW_MF256_F050)   ; xmm3=tmp6H
+
+    paddd       xmm4, xmm6              ; xmm4=data5L
+    paddd       xmm1, xmm0              ; xmm1=data5H
+    paddd       xmm5, XMMWORD wk(0)   ; xmm5=data3L
+    paddd       xmm3, XMMWORD wk(1)   ; xmm3=data3H
+
+    paddd       xmm4, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       xmm1, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       xmm4, DESCALE_P1
+    psrad       xmm1, DESCALE_P1
+    paddd       xmm5, GOTOFF(ebx,PD_DESCALE_P1)
+    paddd       xmm3, GOTOFF(ebx,PD_DESCALE_P1)
+    psrad       xmm5, DESCALE_P1
+    psrad       xmm3, DESCALE_P1
+
+    packssdw    xmm4, xmm1              ; xmm4=data5
+    packssdw    xmm5, xmm3              ; xmm5=data3
+
+    ; ---- Pass 2: process columns.
+
+;   mov         edx, POINTER data(eax)  ; (DCTELEM *)
+
+    movdqa      xmm6, XMMWORD wk(2)   ; xmm6=col0
+    movdqa      xmm0, XMMWORD wk(4)   ; xmm0=col2
+
+    ; xmm6=(00 10 20 30 40 50 60 70), xmm0=(02 12 22 32 42 52 62 72)
+    ; xmm2=(01 11 21 31 41 51 61 71), xmm5=(03 13 23 33 43 53 63 73)
+
+    movdqa      xmm1, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm2              ; xmm6=(00 01 10 11 20 21 30 31)
+    punpckhwd   xmm1, xmm2              ; xmm1=(40 41 50 51 60 61 70 71)
+    movdqa      xmm3, xmm0              ; transpose coefficients(phase 1)
+    punpcklwd   xmm0, xmm5              ; xmm0=(02 03 12 13 22 23 32 33)
+    punpckhwd   xmm3, xmm5              ; xmm3=(42 43 52 53 62 63 72 73)
+
+    movdqa      xmm2, XMMWORD wk(3)   ; xmm2=col4
+    movdqa      xmm5, XMMWORD wk(5)   ; xmm5=col6
+
+    ; xmm2=(04 14 24 34 44 54 64 74), xmm5=(06 16 26 36 46 56 66 76)
+    ; xmm4=(05 15 25 35 45 55 65 75), xmm7=(07 17 27 37 47 57 67 77)
+
+    movdqa      XMMWORD wk(0), xmm0   ; wk(0)=(02 03 12 13 22 23 32 33)
+    movdqa      XMMWORD wk(1), xmm3   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+    movdqa      xmm0, xmm2              ; transpose coefficients(phase 1)
+    punpcklwd   xmm2, xmm4              ; xmm2=(04 05 14 15 24 25 34 35)
+    punpckhwd   xmm0, xmm4              ; xmm0=(44 45 54 55 64 65 74 75)
+    movdqa      xmm3, xmm5              ; transpose coefficients(phase 1)
+    punpcklwd   xmm5, xmm7              ; xmm5=(06 07 16 17 26 27 36 37)
+    punpckhwd   xmm3, xmm7              ; xmm3=(46 47 56 57 66 67 76 77)
+
+    movdqa      xmm4, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm5              ; xmm2=(04 05 06 07 14 15 16 17)
+    punpckhdq   xmm4, xmm5              ; xmm4=(24 25 26 27 34 35 36 37)
+    movdqa      xmm7, xmm0              ; transpose coefficients(phase 2)
+    punpckldq   xmm0, xmm3              ; xmm0=(44 45 46 47 54 55 56 57)
+    punpckhdq   xmm7, xmm3              ; xmm7=(64 65 66 67 74 75 76 77)
+
+    movdqa      xmm5, XMMWORD wk(0)   ; xmm5=(02 03 12 13 22 23 32 33)
+    movdqa      xmm3, XMMWORD wk(1)   ; xmm3=(42 43 52 53 62 63 72 73)
+    movdqa      XMMWORD wk(2), xmm4   ; wk(2)=(24 25 26 27 34 35 36 37)
+    movdqa      XMMWORD wk(3), xmm0   ; wk(3)=(44 45 46 47 54 55 56 57)
+
+    movdqa      xmm4, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm5              ; xmm6=(00 01 02 03 10 11 12 13)
+    punpckhdq   xmm4, xmm5              ; xmm4=(20 21 22 23 30 31 32 33)
+    movdqa      xmm0, xmm1              ; transpose coefficients(phase 2)
+    punpckldq   xmm1, xmm3              ; xmm1=(40 41 42 43 50 51 52 53)
+    punpckhdq   xmm0, xmm3              ; xmm0=(60 61 62 63 70 71 72 73)
+
+    movdqa      xmm5, xmm6              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm6, xmm2              ; xmm6=(00 01 02 03 04 05 06 07)=data0
+    punpckhqdq  xmm5, xmm2              ; xmm5=(10 11 12 13 14 15 16 17)=data1
+    movdqa      xmm3, xmm0              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm0, xmm7              ; xmm0=(60 61 62 63 64 65 66 67)=data6
+    punpckhqdq  xmm3, xmm7              ; xmm3=(70 71 72 73 74 75 76 77)=data7
+
+    movdqa      xmm2, xmm5
+    movdqa      xmm7, xmm6
+    psubw       xmm5, xmm0              ; xmm5=data1-data6=tmp6
+    psubw       xmm6, xmm3              ; xmm6=data0-data7=tmp7
+    paddw       xmm2, xmm0              ; xmm2=data1+data6=tmp1
+    paddw       xmm7, xmm3              ; xmm7=data0+data7=tmp0
+
+    movdqa      xmm0, XMMWORD wk(2)   ; xmm0=(24 25 26 27 34 35 36 37)
+    movdqa      xmm3, XMMWORD wk(3)   ; xmm3=(44 45 46 47 54 55 56 57)
+    movdqa      XMMWORD wk(0), xmm5   ; wk(0)=tmp6
+    movdqa      XMMWORD wk(1), xmm6   ; wk(1)=tmp7
+
+    movdqa      xmm5, xmm4              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm4, xmm0              ; xmm4=(20 21 22 23 24 25 26 27)=data2
+    punpckhqdq  xmm5, xmm0              ; xmm5=(30 31 32 33 34 35 36 37)=data3
+    movdqa      xmm6, xmm1              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm1, xmm3              ; xmm1=(40 41 42 43 44 45 46 47)=data4
+    punpckhqdq  xmm6, xmm3              ; xmm6=(50 51 52 53 54 55 56 57)=data5
+
+    movdqa      xmm0, xmm5
+    movdqa      xmm3, xmm4
+    paddw       xmm5, xmm1              ; xmm5=data3+data4=tmp3
+    paddw       xmm4, xmm6              ; xmm4=data2+data5=tmp2
+    psubw       xmm0, xmm1              ; xmm0=data3-data4=tmp4
+    psubw       xmm3, xmm6              ; xmm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movdqa      xmm1, xmm7
+    movdqa      xmm6, xmm2
+    paddw       xmm7, xmm5              ; xmm7=tmp10
+    paddw       xmm2, xmm4              ; xmm2=tmp11
+    psubw       xmm1, xmm5              ; xmm1=tmp13
+    psubw       xmm6, xmm4              ; xmm6=tmp12
+
+    movdqa      xmm5, xmm7
+    paddw       xmm7, xmm2              ; xmm7=tmp10+tmp11
+    psubw       xmm5, xmm2              ; xmm5=tmp10-tmp11
+
+    paddw       xmm7, GOTOFF(ebx,PW_DESCALE_P2X)
+    paddw       xmm5, GOTOFF(ebx,PW_DESCALE_P2X)
+    psraw       xmm7, PASS1_BITS        ; xmm7=data0
+    psraw       xmm5, PASS1_BITS        ; xmm5=data4
+
+    movdqa      XMMWORD XMMBLOCK(0,0,edx,SIZEOF_DCTELEM), xmm7
+    movdqa      XMMWORD XMMBLOCK(4,0,edx,SIZEOF_DCTELEM), xmm5
+
+    ; (Original)
+    ; z1 = (tmp12 + tmp13) * 0.541196100;
+    ; data2 = z1 + tmp13 * 0.765366865;
+    ; data6 = z1 + tmp12 * -1.847759065;
+    ;
+    ; (This implementation)
+    ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+    ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+    movdqa      xmm4, xmm1              ; xmm1=tmp13
+    movdqa      xmm2, xmm1
+    punpcklwd   xmm4, xmm6              ; xmm6=tmp12
+    punpckhwd   xmm2, xmm6
+    movdqa      xmm1, xmm4
+    movdqa      xmm6, xmm2
+    pmaddwd     xmm4, GOTOFF(ebx,PW_F130_F054)   ; xmm4=data2L
+    pmaddwd     xmm2, GOTOFF(ebx,PW_F130_F054)   ; xmm2=data2H
+    pmaddwd     xmm1, GOTOFF(ebx,PW_F054_MF130)  ; xmm1=data6L
+    pmaddwd     xmm6, GOTOFF(ebx,PW_F054_MF130)  ; xmm6=data6H
+
+    paddd       xmm4, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       xmm2, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       xmm4, DESCALE_P2
+    psrad       xmm2, DESCALE_P2
+    paddd       xmm1, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       xmm6, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       xmm1, DESCALE_P2
+    psrad       xmm6, DESCALE_P2
+
+    packssdw    xmm4, xmm2              ; xmm4=data2
+    packssdw    xmm1, xmm6              ; xmm1=data6
+
+    movdqa      XMMWORD XMMBLOCK(2,0,edx,SIZEOF_DCTELEM), xmm4
+    movdqa      XMMWORD XMMBLOCK(6,0,edx,SIZEOF_DCTELEM), xmm1
+
+    ; -- Odd part
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=tmp6
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=tmp7
+
+    movdqa      xmm2, xmm0              ; xmm0=tmp4
+    movdqa      xmm6, xmm3              ; xmm3=tmp5
+    paddw       xmm2, xmm7              ; xmm2=z3
+    paddw       xmm6, xmm5              ; xmm6=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movdqa      xmm4, xmm2
+    movdqa      xmm1, xmm2
+    punpcklwd   xmm4, xmm6
+    punpckhwd   xmm1, xmm6
+    movdqa      xmm2, xmm4
+    movdqa      xmm6, xmm1
+    pmaddwd     xmm4, GOTOFF(ebx,PW_MF078_F117)  ; xmm4=z3L
+    pmaddwd     xmm1, GOTOFF(ebx,PW_MF078_F117)  ; xmm1=z3H
+    pmaddwd     xmm2, GOTOFF(ebx,PW_F117_F078)   ; xmm2=z4L
+    pmaddwd     xmm6, GOTOFF(ebx,PW_F117_F078)   ; xmm6=z4H
+
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=z3L
+    movdqa      XMMWORD wk(1), xmm1   ; wk(1)=z3H
+
+    ; (Original)
+    ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+    ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+    ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+    ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+    ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+    ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+    ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+    ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+    ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+    movdqa      xmm4, xmm0
+    movdqa      xmm1, xmm0
+    punpcklwd   xmm4, xmm5
+    punpckhwd   xmm1, xmm5
+    movdqa      xmm0, xmm4
+    movdqa      xmm5, xmm1
+    pmaddwd     xmm4, GOTOFF(ebx,PW_MF060_MF089)  ; xmm4=tmp4L
+    pmaddwd     xmm1, GOTOFF(ebx,PW_MF060_MF089)  ; xmm1=tmp4H
+    pmaddwd     xmm0, GOTOFF(ebx,PW_MF089_F060)   ; xmm0=tmp7L
+    pmaddwd     xmm5, GOTOFF(ebx,PW_MF089_F060)   ; xmm5=tmp7H
+
+    paddd       xmm4,  XMMWORD wk(0)  ; xmm4=data7L
+    paddd       xmm1,  XMMWORD wk(1)  ; xmm1=data7H
+    paddd       xmm0, xmm2              ; xmm0=data1L
+    paddd       xmm5, xmm6              ; xmm5=data1H
+
+    paddd       xmm4, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       xmm1, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       xmm4, DESCALE_P2
+    psrad       xmm1, DESCALE_P2
+    paddd       xmm0, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       xmm5, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       xmm0, DESCALE_P2
+    psrad       xmm5, DESCALE_P2
+
+    packssdw    xmm4, xmm1              ; xmm4=data7
+    packssdw    xmm0, xmm5              ; xmm0=data1
+
+    movdqa      XMMWORD XMMBLOCK(7,0,edx,SIZEOF_DCTELEM), xmm4
+    movdqa      XMMWORD XMMBLOCK(1,0,edx,SIZEOF_DCTELEM), xmm0
+
+    movdqa      xmm1, xmm3
+    movdqa      xmm5, xmm3
+    punpcklwd   xmm1, xmm7
+    punpckhwd   xmm5, xmm7
+    movdqa      xmm3, xmm1
+    movdqa      xmm7, xmm5
+    pmaddwd     xmm1, GOTOFF(ebx,PW_MF050_MF256)  ; xmm1=tmp5L
+    pmaddwd     xmm5, GOTOFF(ebx,PW_MF050_MF256)  ; xmm5=tmp5H
+    pmaddwd     xmm3, GOTOFF(ebx,PW_MF256_F050)   ; xmm3=tmp6L
+    pmaddwd     xmm7, GOTOFF(ebx,PW_MF256_F050)   ; xmm7=tmp6H
+
+    paddd       xmm1, xmm2              ; xmm1=data5L
+    paddd       xmm5, xmm6              ; xmm5=data5H
+    paddd       xmm3, XMMWORD wk(0)   ; xmm3=data3L
+    paddd       xmm7, XMMWORD wk(1)   ; xmm7=data3H
+
+    paddd       xmm1, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       xmm5, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       xmm1, DESCALE_P2
+    psrad       xmm5, DESCALE_P2
+    paddd       xmm3, GOTOFF(ebx,PD_DESCALE_P2)
+    paddd       xmm7, GOTOFF(ebx,PD_DESCALE_P2)
+    psrad       xmm3, DESCALE_P2
+    psrad       xmm7, DESCALE_P2
+
+    packssdw    xmm1, xmm5              ; xmm1=data5
+    packssdw    xmm3, xmm7              ; xmm3=data3
+
+    movdqa      XMMWORD XMMBLOCK(5,0,edx,SIZEOF_DCTELEM), xmm1
+    movdqa      XMMWORD XMMBLOCK(3,0,edx,SIZEOF_DCTELEM), xmm3
+
+;   pop         edi                     ; unused
+;   pop         esi                     ; unused
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctflt-3dn.asm Added

@@ -0,0 +1,451 @@
+;
+; jidctflt.asm - floating-point IDCT (3DNow! & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_float_3dnow)
+
+EXTN(jconst_idct_float_3dnow):
+
+PD_1_414        times 2 dd 1.414213562373095048801689
+PD_1_847        times 2 dd 1.847759065022573512256366
+PD_1_082        times 2 dd 1.082392200292393968799446
+PD_2_613        times 2 dd 2.613125929752753055713286
+PD_RNDINT_MAGIC times 2 dd 100663296.0  ; (float)(0x00C00000 << 3)
+PB_CENTERJSAMP  times 8 db CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_3dnow(void *dct_table, JCOEFPTR coef_block,
+;                        JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; void *dct_table
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_MMWORD
+                                        ; mmword wkWK_NUM
+%define WK_NUM         2
+%define workspace      wk(0) - DCTSIZE2 * SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspaceDCTSIZE2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_float_3dnow)
+
+EXTN(jsimd_idct_float_3dnow):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, workspace
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input, store into work array.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+    lea         edi, workspace                 ; FAST_FLOAT *wsptr
+    mov         ecx, DCTSIZE/2                   ; ctr
+    alignx      16, 7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_3DNOW
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         short .columnDCT
+
+    pushpic     ebx                     ; save GOT address
+    mov         ebx, dword DWBLOCK(3,0,esi,SIZEOF_JCOEF)
+    mov         eax, dword DWBLOCK(4,0,esi,SIZEOF_JCOEF)
+    or          ebx, dword DWBLOCK(5,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(6,0,esi,SIZEOF_JCOEF)
+    or          ebx, dword DWBLOCK(7,0,esi,SIZEOF_JCOEF)
+    or          eax, ebx
+    poppic      ebx                     ; restore GOT address
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movd        mm0, dword DWBLOCK(0,0,esi,SIZEOF_JCOEF)
+
+    punpcklwd   mm0, mm0
+    psrad       mm0, (DWORD_BIT-WORD_BIT)
+    pi2fd       mm0, mm0
+
+    pfmul       mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movq        mm1, mm0
+    punpckldq   mm0, mm0
+    punpckhdq   mm1, mm1
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT), mm0
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT), mm0
+    movq        MMWORD MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT), mm0
+    movq        MMWORD MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT), mm0
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT), mm1
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT), mm1
+    movq        MMWORD MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT), mm1
+    movq        MMWORD MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT), mm1
+    jmp         near .nextcolumn
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movd        mm0, dword DWBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movd        mm1, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    movd        mm2, dword DWBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movd        mm3, dword DWBLOCK(6,0,esi,SIZEOF_JCOEF)
+
+    punpcklwd   mm0, mm0
+    punpcklwd   mm1, mm1
+    psrad       mm0, (DWORD_BIT-WORD_BIT)
+    psrad       mm1, (DWORD_BIT-WORD_BIT)
+    pi2fd       mm0, mm0
+    pi2fd       mm1, mm1
+
+    pfmul       mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    pfmul       mm1, MMWORD MMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    punpcklwd   mm2, mm2
+    punpcklwd   mm3, mm3
+    psrad       mm2, (DWORD_BIT-WORD_BIT)
+    psrad       mm3, (DWORD_BIT-WORD_BIT)
+    pi2fd       mm2, mm2
+    pi2fd       mm3, mm3
+
+    pfmul       mm2, MMWORD MMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    pfmul       mm3, MMWORD MMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movq        mm4, mm0
+    movq        mm5, mm1
+    pfsub       mm0, mm2                ; mm0=tmp11
+    pfsub       mm1, mm3
+    pfadd       mm4, mm2                ; mm4=tmp10
+    pfadd       mm5, mm3                ; mm5=tmp13
+
+    pfmul       mm1, GOTOFF(ebx,PD_1_414)
+    pfsub       mm1, mm5                ; mm1=tmp12
+
+    movq        mm6, mm4
+    movq        mm7, mm0
+    pfsub       mm4, mm5                ; mm4=tmp3
+    pfsub       mm0, mm1                ; mm0=tmp2
+    pfadd       mm6, mm5                ; mm6=tmp0
+    pfadd       mm7, mm1                ; mm7=tmp1
+
+    movq        MMWORD wk(1), mm4     ; tmp3
+    movq        MMWORD wk(0), mm0     ; tmp2
+
+    ; -- Odd part
+
+    movd        mm2, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movd        mm3, dword DWBLOCK(3,0,esi,SIZEOF_JCOEF)
+    movd        mm5, dword DWBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movd        mm1, dword DWBLOCK(7,0,esi,SIZEOF_JCOEF)
+
+    punpcklwd   mm2, mm2
+    punpcklwd   mm3, mm3
+    psrad       mm2, (DWORD_BIT-WORD_BIT)
+    psrad       mm3, (DWORD_BIT-WORD_BIT)
+    pi2fd       mm2, mm2
+    pi2fd       mm3, mm3
+
+    pfmul       mm2, MMWORD MMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    pfmul       mm3, MMWORD MMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    punpcklwd   mm5, mm5
+    punpcklwd   mm1, mm1
+    psrad       mm5, (DWORD_BIT-WORD_BIT)
+    psrad       mm1, (DWORD_BIT-WORD_BIT)
+    pi2fd       mm5, mm5
+    pi2fd       mm1, mm1
+
+    pfmul       mm5, MMWORD MMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    pfmul       mm1, MMWORD MMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movq        mm4, mm2
+    movq        mm0, mm5
+    pfadd       mm2, mm1                ; mm2=z11
+    pfadd       mm5, mm3                ; mm5=z13
+    pfsub       mm4, mm1                ; mm4=z12
+    pfsub       mm0, mm3                ; mm0=z10
+
+    movq        mm1, mm2
+    pfsub       mm2, mm5
+    pfadd       mm1, mm5                ; mm1=tmp7
+
+    pfmul       mm2, GOTOFF(ebx,PD_1_414)  ; mm2=tmp11
+
+    movq        mm3, mm0
+    pfadd       mm0, mm4
+    pfmul       mm0, GOTOFF(ebx,PD_1_847)  ; mm0=z5
+    pfmul       mm3, GOTOFF(ebx,PD_2_613)  ; mm3=(z10 * 2.613125930)
+    pfmul       mm4, GOTOFF(ebx,PD_1_082)  ; mm4=(z12 * 1.082392200)
+    pfsubr      mm3, mm0                     ; mm3=tmp12
+    pfsub       mm4, mm0                     ; mm4=tmp10
+
+    ; -- Final output stage
+
+    pfsub       mm3, mm1                ; mm3=tmp6
+    movq        mm5, mm6
+    movq        mm0, mm7
+    pfadd       mm6, mm1                ; mm6=data0=(00 01)
+    pfadd       mm7, mm3                ; mm7=data1=(10 11)
+    pfsub       mm5, mm1                ; mm5=data7=(70 71)
+    pfsub       mm0, mm3                ; mm0=data6=(60 61)
+    pfsub       mm2, mm3                ; mm2=tmp5
+
+    movq        mm1, mm6                ; transpose coefficients
+    punpckldq   mm6, mm7                ; mm6=(00 10)
+    punpckhdq   mm1, mm7                ; mm1=(01 11)
+    movq        mm3, mm0                ; transpose coefficients
+    punpckldq   mm0, mm5                ; mm0=(60 70)
+    punpckhdq   mm3, mm5                ; mm3=(61 71)
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT), mm6
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT), mm1
+    movq        MMWORD MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT), mm0
+    movq        MMWORD MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT), mm3
+
+    movq        mm7, MMWORD wk(0)     ; mm7=tmp2
+    movq        mm5, MMWORD wk(1)     ; mm5=tmp3
+
+    pfadd       mm4, mm2                ; mm4=tmp4
+    movq        mm6, mm7
+    movq        mm1, mm5
+    pfadd       mm7, mm2                ; mm7=data2=(20 21)
+    pfadd       mm5, mm4                ; mm5=data4=(40 41)
+    pfsub       mm6, mm2                ; mm6=data5=(50 51)
+    pfsub       mm1, mm4                ; mm1=data3=(30 31)
+
+    movq        mm0, mm7                ; transpose coefficients
+    punpckldq   mm7, mm1                ; mm7=(20 30)
+    punpckhdq   mm0, mm1                ; mm0=(21 31)
+    movq        mm3, mm5                ; transpose coefficients
+    punpckldq   mm5, mm6                ; mm5=(40 50)
+    punpckhdq   mm3, mm6                ; mm3=(41 51)
+
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT), mm7
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT), mm0
+    movq        MMWORD MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT), mm5
+    movq        MMWORD MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT), mm3
+
+.nextcolumn:
+    add         esi, byte 2*SIZEOF_JCOEF               ; coef_block
+    add         edx, byte 2*SIZEOF_FLOAT_MULT_TYPE     ; quantptr
+    add         edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT  ; wsptr
+    dec         ecx                                    ; ctr
+    jnz         near .columnloop
+
+    ; -- Prefetch the next coefficient block
+
+    prefetch esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32
+    prefetch esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32
+    prefetch esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32
+    prefetch esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         eax, original_ebp
+    lea         esi, workspace                   ; FAST_FLOAT *wsptr
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+    mov         ecx, DCTSIZE/2                     ; ctr
+    alignx      16, 7
+.rowloop:
+
+    ; -- Even part
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)
+    movq        mm2, MMWORD MMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)
+    movq        mm3, MMWORD MMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)
+
+    movq        mm4, mm0
+    movq        mm5, mm1
+    pfsub       mm0, mm2                ; mm0=tmp11
+    pfsub       mm1, mm3
+    pfadd       mm4, mm2                ; mm4=tmp10
+    pfadd       mm5, mm3                ; mm5=tmp13
+
+    pfmul       mm1, GOTOFF(ebx,PD_1_414)
+    pfsub       mm1, mm5                ; mm1=tmp12
+
+    movq        mm6, mm4
+    movq        mm7, mm0
+    pfsub       mm4, mm5                ; mm4=tmp3
+    pfsub       mm0, mm1                ; mm0=tmp2
+    pfadd       mm6, mm5                ; mm6=tmp0
+    pfadd       mm7, mm1                ; mm7=tmp1
+
+    movq        MMWORD wk(1), mm4     ; tmp3
+    movq        MMWORD wk(0), mm0     ; tmp2
+
+    ; -- Odd part
+
+    movq        mm2, MMWORD MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)
+    movq        mm3, MMWORD MMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)
+    movq        mm5, MMWORD MMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)
+    movq        mm1, MMWORD MMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)
+
+    movq        mm4, mm2
+    movq        mm0, mm5
+    pfadd       mm2, mm1                ; mm2=z11
+    pfadd       mm5, mm3                ; mm5=z13
+    pfsub       mm4, mm1                ; mm4=z12
+    pfsub       mm0, mm3                ; mm0=z10
+
+    movq        mm1, mm2
+    pfsub       mm2, mm5
+    pfadd       mm1, mm5                ; mm1=tmp7
+
+    pfmul       mm2, GOTOFF(ebx,PD_1_414)  ; mm2=tmp11
+
+    movq        mm3, mm0
+    pfadd       mm0, mm4
+    pfmul       mm0, GOTOFF(ebx,PD_1_847)  ; mm0=z5
+    pfmul       mm3, GOTOFF(ebx,PD_2_613)  ; mm3=(z10 * 2.613125930)
+    pfmul       mm4, GOTOFF(ebx,PD_1_082)  ; mm4=(z12 * 1.082392200)
+    pfsubr      mm3, mm0                     ; mm3=tmp12
+    pfsub       mm4, mm0                     ; mm4=tmp10
+
+    ; -- Final output stage
+
+    pfsub       mm3, mm1                ; mm3=tmp6
+    movq        mm5, mm6
+    movq        mm0, mm7
+    pfadd       mm6, mm1                ; mm6=data0=(00 10)
+    pfadd       mm7, mm3                ; mm7=data1=(01 11)
+    pfsub       mm5, mm1                ; mm5=data7=(07 17)
+    pfsub       mm0, mm3                ; mm0=data6=(06 16)
+    pfsub       mm2, mm3                ; mm2=tmp5
+
+    movq        mm1, GOTOFF(ebx,PD_RNDINT_MAGIC)  ; mm1=PD_RNDINT_MAGIC
+    pcmpeqd     mm3, mm3
+    psrld       mm3, WORD_BIT           ; mm3={0xFFFF 0x0000 0xFFFF 0x0000}
+
+    pfadd       mm6, mm1                ; mm6=roundint(data0/8)=(00 ** 10 **)
+    pfadd       mm7, mm1                ; mm7=roundint(data1/8)=(01 ** 11 **)
+    pfadd       mm0, mm1                ; mm0=roundint(data6/8)=(06 ** 16 **)
+    pfadd       mm5, mm1                ; mm5=roundint(data7/8)=(07 ** 17 **)
+
+    pand        mm6, mm3                ; mm6=(00 -- 10 --)
+    pslld       mm7, WORD_BIT           ; mm7=(-- 01 -- 11)
+    pand        mm0, mm3                ; mm0=(06 -- 16 --)
+    pslld       mm5, WORD_BIT           ; mm5=(-- 07 -- 17)
+    por         mm6, mm7                ; mm6=(00 01 10 11)
+    por         mm0, mm5                ; mm0=(06 07 16 17)
+
+    movq        mm1, MMWORD wk(0)     ; mm1=tmp2
+    movq        mm3, MMWORD wk(1)     ; mm3=tmp3
+
+    pfadd       mm4, mm2                ; mm4=tmp4
+    movq        mm7, mm1
+    movq        mm5, mm3
+    pfadd       mm1, mm2                ; mm1=data2=(02 12)
+    pfadd       mm3, mm4                ; mm3=data4=(04 14)
+    pfsub       mm7, mm2                ; mm7=data5=(05 15)
+    pfsub       mm5, mm4                ; mm5=data3=(03 13)
+
+    movq        mm2, GOTOFF(ebx,PD_RNDINT_MAGIC)  ; mm2=PD_RNDINT_MAGIC
+    pcmpeqd     mm4, mm4
+    psrld       mm4, WORD_BIT           ; mm4={0xFFFF 0x0000 0xFFFF 0x0000}
+
+    pfadd       mm3, mm2                ; mm3=roundint(data4/8)=(04 ** 14 **)
+    pfadd       mm7, mm2                ; mm7=roundint(data5/8)=(05 ** 15 **)
+    pfadd       mm1, mm2                ; mm1=roundint(data2/8)=(02 ** 12 **)
+    pfadd       mm5, mm2                ; mm5=roundint(data3/8)=(03 ** 13 **)
+
+    pand        mm3, mm4                ; mm3=(04 -- 14 --)
+    pslld       mm7, WORD_BIT           ; mm7=(-- 05 -- 15)
+    pand        mm1, mm4                ; mm1=(02 -- 12 --)
+    pslld       mm5, WORD_BIT           ; mm5=(-- 03 -- 13)
+    por         mm3, mm7                ; mm3=(04 05 14 15)
+    por         mm1, mm5                ; mm1=(02 03 12 13)
+
+    movq        mm2, GOTOFF(ebx,PB_CENTERJSAMP)  ; mm2=PB_CENTERJSAMP
+
+    packsswb    mm6, mm3                ; mm6=(00 01 10 11 04 05 14 15)
+    packsswb    mm1, mm0                ; mm1=(02 03 12 13 06 07 16 17)
+    paddb       mm6, mm2
+    paddb       mm1, mm2
+
+    movq        mm4, mm6                ; transpose coefficients(phase 2)
+    punpcklwd   mm6, mm1                ; mm6=(00 01 02 03 10 11 12 13)
+    punpckhwd   mm4, mm1                ; mm4=(04 05 06 07 14 15 16 17)
+
+    movq        mm7, mm6                ; transpose coefficients(phase 3)
+    punpckldq   mm6, mm4                ; mm6=(00 01 02 03 04 05 06 07)
+    punpckhdq   mm7, mm4                ; mm7=(10 11 12 13 14 15 16 17)
+
+    pushpic     ebx                     ; save GOT address
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         ebx, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    movq        MMWORD edx+eax*SIZEOF_JSAMPLE, mm6
+    movq        MMWORD ebx+eax*SIZEOF_JSAMPLE, mm7
+
+    poppic      ebx                     ; restore GOT address
+
+    add         esi, byte 2*SIZEOF_FAST_FLOAT  ; wsptr
+    add         edi, byte 2*SIZEOF_JSAMPROW
+    dec         ecx                            ; ctr
+    jnz         near .rowloop
+
+    femms                               ; empty MMX/3DNow! state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctflt-sse.asm Added

@@ -0,0 +1,571 @@
+;
+; jidctflt.asm - floating-point IDCT (SSE & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro unpcklps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+    shufps      %1, %2, 0x44
+%endmacro
+
+%macro unpckhps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+    shufps      %1, %2, 0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_float_sse)
+
+EXTN(jconst_idct_float_sse):
+
+PD_1_414       times 4 dd  1.414213562373095048801689
+PD_1_847       times 4 dd  1.847759065022573512256366
+PD_1_082       times 4 dd  1.082392200292393968799446
+PD_M2_613      times 4 dd -2.613125929752753055713286
+PD_0_125       times 4 dd  0.125        ; 1/8
+PB_CENTERJSAMP times 8 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_sse(void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; void *dct_table
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM         2
+%define workspace      wk(0) - DCTSIZE2 * SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspaceDCTSIZE2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_float_sse)
+
+EXTN(jsimd_idct_float_sse):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, workspace
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input, store into work array.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+    lea         edi, workspace                 ; FAST_FLOAT *wsptr
+    mov         ecx, DCTSIZE/4                   ; ctr
+    alignx      16, 7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         near .columnDCT
+
+    movq        mm0, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    por         mm1, MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    por         mm1, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    por         mm1, mm0
+    packsswb    mm1, mm1
+    movd        eax, mm1
+    test        eax, eax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+
+    punpckhwd   mm1, mm0                   ; mm1=(** 02 ** 03)
+    punpcklwd   mm0, mm0                   ; mm0=(00 00 01 01)
+    psrad       mm1, (DWORD_BIT-WORD_BIT)  ; mm1=in0H=(02 03)
+    psrad       mm0, (DWORD_BIT-WORD_BIT)  ; mm0=in0L=(00 01)
+    cvtpi2ps    xmm3, mm1                  ; xmm3=(02 03 ** **)
+    cvtpi2ps    xmm0, mm0                  ; xmm0=(00 01 ** **)
+    movlhps     xmm0, xmm3                 ; xmm0=in0=(00 01 02 03)
+
+    mulps       xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movaps      xmm1, xmm0
+    movaps      xmm2, xmm0
+    movaps      xmm3, xmm0
+
+    shufps      xmm0, xmm0, 0x00        ; xmm0=(00 00 00 00)
+    shufps      xmm1, xmm1, 0x55        ; xmm1=(01 01 01 01)
+    shufps      xmm2, xmm2, 0xAA        ; xmm2=(02 02 02 02)
+    shufps      xmm3, xmm3, 0xFF        ; xmm3=(03 03 03 03)
+
+    movaps      XMMWORD XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT), xmm2
+    movaps      XMMWORD XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT), xmm2
+    movaps      XMMWORD XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT), xmm3
+    movaps      XMMWORD XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT), xmm3
+    jmp         near .nextcolumn
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    movq        mm2, MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+
+    punpckhwd   mm4, mm0                ; mm4=(** 02 ** 03)
+    punpcklwd   mm0, mm0                ; mm0=(00 00 01 01)
+    punpckhwd   mm5, mm1                ; mm5=(** 22 ** 23)
+    punpcklwd   mm1, mm1                ; mm1=(20 20 21 21)
+
+    psrad       mm4, (DWORD_BIT-WORD_BIT)  ; mm4=in0H=(02 03)
+    psrad       mm0, (DWORD_BIT-WORD_BIT)  ; mm0=in0L=(00 01)
+    cvtpi2ps    xmm4, mm4                  ; xmm4=(02 03 ** **)
+    cvtpi2ps    xmm0, mm0                  ; xmm0=(00 01 ** **)
+    psrad       mm5, (DWORD_BIT-WORD_BIT)  ; mm5=in2H=(22 23)
+    psrad       mm1, (DWORD_BIT-WORD_BIT)  ; mm1=in2L=(20 21)
+    cvtpi2ps    xmm5, mm5                  ; xmm5=(22 23 ** **)
+    cvtpi2ps    xmm1, mm1                  ; xmm1=(20 21 ** **)
+
+    punpckhwd   mm6, mm2                ; mm6=(** 42 ** 43)
+    punpcklwd   mm2, mm2                ; mm2=(40 40 41 41)
+    punpckhwd   mm7, mm3                ; mm7=(** 62 ** 63)
+    punpcklwd   mm3, mm3                ; mm3=(60 60 61 61)
+
+    psrad       mm6, (DWORD_BIT-WORD_BIT)  ; mm6=in4H=(42 43)
+    psrad       mm2, (DWORD_BIT-WORD_BIT)  ; mm2=in4L=(40 41)
+    cvtpi2ps    xmm6, mm6                  ; xmm6=(42 43 ** **)
+    cvtpi2ps    xmm2, mm2                  ; xmm2=(40 41 ** **)
+    psrad       mm7, (DWORD_BIT-WORD_BIT)  ; mm7=in6H=(62 63)
+    psrad       mm3, (DWORD_BIT-WORD_BIT)  ; mm3=in6L=(60 61)
+    cvtpi2ps    xmm7, mm7                  ; xmm7=(62 63 ** **)
+    cvtpi2ps    xmm3, mm3                  ; xmm3=(60 61 ** **)
+
+    movlhps     xmm0, xmm4              ; xmm0=in0=(00 01 02 03)
+    movlhps     xmm1, xmm5              ; xmm1=in2=(20 21 22 23)
+    mulps       xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm1, XMMWORD XMMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movlhps     xmm2, xmm6              ; xmm2=in4=(40 41 42 43)
+    movlhps     xmm3, xmm7              ; xmm3=in6=(60 61 62 63)
+    mulps       xmm2, XMMWORD XMMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm3, XMMWORD XMMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movaps      xmm4, xmm0
+    movaps      xmm5, xmm1
+    subps       xmm0, xmm2              ; xmm0=tmp11
+    subps       xmm1, xmm3
+    addps       xmm4, xmm2              ; xmm4=tmp10
+    addps       xmm5, xmm3              ; xmm5=tmp13
+
+    mulps       xmm1, GOTOFF(ebx,PD_1_414)
+    subps       xmm1, xmm5              ; xmm1=tmp12
+
+    movaps      xmm6, xmm4
+    movaps      xmm7, xmm0
+    subps       xmm4, xmm5              ; xmm4=tmp3
+    subps       xmm0, xmm1              ; xmm0=tmp2
+    addps       xmm6, xmm5              ; xmm6=tmp0
+    addps       xmm7, xmm1              ; xmm7=tmp1
+
+    movaps      XMMWORD wk(1), xmm4   ; tmp3
+    movaps      XMMWORD wk(0), xmm0   ; tmp2
+
+    ; -- Odd part
+
+    movq        mm4, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm0, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    movq        mm5, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+
+    punpckhwd   mm6, mm4                ; mm6=(** 12 ** 13)
+    punpcklwd   mm4, mm4                ; mm4=(10 10 11 11)
+    punpckhwd   mm2, mm0                ; mm2=(** 32 ** 33)
+    punpcklwd   mm0, mm0                ; mm0=(30 30 31 31)
+
+    psrad       mm6, (DWORD_BIT-WORD_BIT)  ; mm6=in1H=(12 13)
+    psrad       mm4, (DWORD_BIT-WORD_BIT)  ; mm4=in1L=(10 11)
+    cvtpi2ps    xmm4, mm6                  ; xmm4=(12 13 ** **)
+    cvtpi2ps    xmm2, mm4                  ; xmm2=(10 11 ** **)
+    psrad       mm2, (DWORD_BIT-WORD_BIT)  ; mm2=in3H=(32 33)
+    psrad       mm0, (DWORD_BIT-WORD_BIT)  ; mm0=in3L=(30 31)
+    cvtpi2ps    xmm0, mm2                  ; xmm0=(32 33 ** **)
+    cvtpi2ps    xmm3, mm0                  ; xmm3=(30 31 ** **)
+
+    punpckhwd   mm7, mm5                ; mm7=(** 52 ** 53)
+    punpcklwd   mm5, mm5                ; mm5=(50 50 51 51)
+    punpckhwd   mm3, mm1                ; mm3=(** 72 ** 73)
+    punpcklwd   mm1, mm1                ; mm1=(70 70 71 71)
+
+    movlhps     xmm2, xmm4              ; xmm2=in1=(10 11 12 13)
+    movlhps     xmm3, xmm0              ; xmm3=in3=(30 31 32 33)
+
+    psrad       mm7, (DWORD_BIT-WORD_BIT)  ; mm7=in5H=(52 53)
+    psrad       mm5, (DWORD_BIT-WORD_BIT)  ; mm5=in5L=(50 51)
+    cvtpi2ps    xmm4, mm7                  ; xmm4=(52 53 ** **)
+    cvtpi2ps    xmm5, mm5                  ; xmm5=(50 51 ** **)
+    psrad       mm3, (DWORD_BIT-WORD_BIT)  ; mm3=in7H=(72 73)
+    psrad       mm1, (DWORD_BIT-WORD_BIT)  ; mm1=in7L=(70 71)
+    cvtpi2ps    xmm0, mm3                  ; xmm0=(72 73 ** **)
+    cvtpi2ps    xmm1, mm1                  ; xmm1=(70 71 ** **)
+
+    mulps       xmm2, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm3, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movlhps     xmm5, xmm4              ; xmm5=in5=(50 51 52 53)
+    movlhps     xmm1, xmm0              ; xmm1=in7=(70 71 72 73)
+    mulps       xmm5, XMMWORD XMMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm1, XMMWORD XMMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movaps      xmm4, xmm2
+    movaps      xmm0, xmm5
+    addps       xmm2, xmm1              ; xmm2=z11
+    addps       xmm5, xmm3              ; xmm5=z13
+    subps       xmm4, xmm1              ; xmm4=z12
+    subps       xmm0, xmm3              ; xmm0=z10
+
+    movaps      xmm1, xmm2
+    subps       xmm2, xmm5
+    addps       xmm1, xmm5              ; xmm1=tmp7
+
+    mulps       xmm2, GOTOFF(ebx,PD_1_414)  ; xmm2=tmp11
+
+    movaps      xmm3, xmm0
+    addps       xmm0, xmm4
+    mulps       xmm0, GOTOFF(ebx,PD_1_847)   ; xmm0=z5
+    mulps       xmm3, GOTOFF(ebx,PD_M2_613)  ; xmm3=(z10 * -2.613125930)
+    mulps       xmm4, GOTOFF(ebx,PD_1_082)   ; xmm4=(z12 * 1.082392200)
+    addps       xmm3, xmm0                     ; xmm3=tmp12
+    subps       xmm4, xmm0                     ; xmm4=tmp10
+
+    ; -- Final output stage
+
+    subps       xmm3, xmm1              ; xmm3=tmp6
+    movaps      xmm5, xmm6
+    movaps      xmm0, xmm7
+    addps       xmm6, xmm1              ; xmm6=data0=(00 01 02 03)
+    addps       xmm7, xmm3              ; xmm7=data1=(10 11 12 13)
+    subps       xmm5, xmm1              ; xmm5=data7=(70 71 72 73)
+    subps       xmm0, xmm3              ; xmm0=data6=(60 61 62 63)
+    subps       xmm2, xmm3              ; xmm2=tmp5
+
+    movaps      xmm1, xmm6              ; transpose coefficients(phase 1)
+    unpcklps    xmm6, xmm7              ; xmm6=(00 10 01 11)
+    unpckhps    xmm1, xmm7              ; xmm1=(02 12 03 13)
+    movaps      xmm3, xmm0              ; transpose coefficients(phase 1)
+    unpcklps    xmm0, xmm5              ; xmm0=(60 70 61 71)
+    unpckhps    xmm3, xmm5              ; xmm3=(62 72 63 73)
+
+    movaps      xmm7, XMMWORD wk(0)   ; xmm7=tmp2
+    movaps      xmm5, XMMWORD wk(1)   ; xmm5=tmp3
+
+    movaps      XMMWORD wk(0), xmm0   ; wk(0)=(60 70 61 71)
+    movaps      XMMWORD wk(1), xmm3   ; wk(1)=(62 72 63 73)
+
+    addps       xmm4, xmm2              ; xmm4=tmp4
+    movaps      xmm0, xmm7
+    movaps      xmm3, xmm5
+    addps       xmm7, xmm2              ; xmm7=data2=(20 21 22 23)
+    addps       xmm5, xmm4              ; xmm5=data4=(40 41 42 43)
+    subps       xmm0, xmm2              ; xmm0=data5=(50 51 52 53)
+    subps       xmm3, xmm4              ; xmm3=data3=(30 31 32 33)
+
+    movaps      xmm2, xmm7              ; transpose coefficients(phase 1)
+    unpcklps    xmm7, xmm3              ; xmm7=(20 30 21 31)
+    unpckhps    xmm2, xmm3              ; xmm2=(22 32 23 33)
+    movaps      xmm4, xmm5              ; transpose coefficients(phase 1)
+    unpcklps    xmm5, xmm0              ; xmm5=(40 50 41 51)
+    unpckhps    xmm4, xmm0              ; xmm4=(42 52 43 53)
+
+    movaps      xmm3, xmm6              ; transpose coefficients(phase 2)
+    unpcklps2   xmm6, xmm7              ; xmm6=(00 10 20 30)
+    unpckhps2   xmm3, xmm7              ; xmm3=(01 11 21 31)
+    movaps      xmm0, xmm1              ; transpose coefficients(phase 2)
+    unpcklps2   xmm1, xmm2              ; xmm1=(02 12 22 32)
+    unpckhps2   xmm0, xmm2              ; xmm0=(03 13 23 33)
+
+    movaps      xmm7, XMMWORD wk(0)   ; xmm7=(60 70 61 71)
+    movaps      xmm2, XMMWORD wk(1)   ; xmm2=(62 72 63 73)
+
+    movaps      XMMWORD XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT), xmm6
+    movaps      XMMWORD XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT), xmm3
+    movaps      XMMWORD XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT), xmm0
+
+    movaps      xmm6, xmm5              ; transpose coefficients(phase 2)
+    unpcklps2   xmm5, xmm7              ; xmm5=(40 50 60 70)
+    unpckhps2   xmm6, xmm7              ; xmm6=(41 51 61 71)
+    movaps      xmm3, xmm4              ; transpose coefficients(phase 2)
+    unpcklps2   xmm4, xmm2              ; xmm4=(42 52 62 72)
+    unpckhps2   xmm3, xmm2              ; xmm3=(43 53 63 73)
+
+    movaps      XMMWORD XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT), xmm6
+    movaps      XMMWORD XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT), xmm4
+    movaps      XMMWORD XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT), xmm3
+
+.nextcolumn:
+    add         esi, byte 4*SIZEOF_JCOEF               ; coef_block
+    add         edx, byte 4*SIZEOF_FLOAT_MULT_TYPE     ; quantptr
+    add         edi,      4*DCTSIZE*SIZEOF_FAST_FLOAT  ; wsptr
+    dec         ecx                                    ; ctr
+    jnz         near .columnloop
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32
+    prefetchnta esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32
+    prefetchnta esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32
+    prefetchnta esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         eax, original_ebp
+    lea         esi, workspace                   ; FAST_FLOAT *wsptr
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+    mov         ecx, DCTSIZE/4                     ; ctr
+    alignx      16, 7
+.rowloop:
+
+    ; -- Even part
+
+    movaps      xmm0, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)
+
+    movaps      xmm4, xmm0
+    movaps      xmm5, xmm1
+    subps       xmm0, xmm2              ; xmm0=tmp11
+    subps       xmm1, xmm3
+    addps       xmm4, xmm2              ; xmm4=tmp10
+    addps       xmm5, xmm3              ; xmm5=tmp13
+
+    mulps       xmm1, GOTOFF(ebx,PD_1_414)
+    subps       xmm1, xmm5              ; xmm1=tmp12
+
+    movaps      xmm6, xmm4
+    movaps      xmm7, xmm0
+    subps       xmm4, xmm5              ; xmm4=tmp3
+    subps       xmm0, xmm1              ; xmm0=tmp2
+    addps       xmm6, xmm5              ; xmm6=tmp0
+    addps       xmm7, xmm1              ; xmm7=tmp1
+
+    movaps      XMMWORD wk(1), xmm4   ; tmp3
+    movaps      XMMWORD wk(0), xmm0   ; tmp2
+
+    ; -- Odd part
+
+    movaps      xmm2, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm5, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)
+
+    movaps      xmm4, xmm2
+    movaps      xmm0, xmm5
+    addps       xmm2, xmm1              ; xmm2=z11
+    addps       xmm5, xmm3              ; xmm5=z13
+    subps       xmm4, xmm1              ; xmm4=z12
+    subps       xmm0, xmm3              ; xmm0=z10
+
+    movaps      xmm1, xmm2
+    subps       xmm2, xmm5
+    addps       xmm1, xmm5              ; xmm1=tmp7
+
+    mulps       xmm2, GOTOFF(ebx,PD_1_414)  ; xmm2=tmp11
+
+    movaps      xmm3, xmm0
+    addps       xmm0, xmm4
+    mulps       xmm0, GOTOFF(ebx,PD_1_847)   ; xmm0=z5
+    mulps       xmm3, GOTOFF(ebx,PD_M2_613)  ; xmm3=(z10 * -2.613125930)
+    mulps       xmm4, GOTOFF(ebx,PD_1_082)   ; xmm4=(z12 * 1.082392200)
+    addps       xmm3, xmm0                     ; xmm3=tmp12
+    subps       xmm4, xmm0                     ; xmm4=tmp10
+
+    ; -- Final output stage
+
+    subps       xmm3, xmm1              ; xmm3=tmp6
+    movaps      xmm5, xmm6
+    movaps      xmm0, xmm7
+    addps       xmm6, xmm1              ; xmm6=data0=(00 10 20 30)
+    addps       xmm7, xmm3              ; xmm7=data1=(01 11 21 31)
+    subps       xmm5, xmm1              ; xmm5=data7=(07 17 27 37)
+    subps       xmm0, xmm3              ; xmm0=data6=(06 16 26 36)
+    subps       xmm2, xmm3              ; xmm2=tmp5
+
+    movaps      xmm1, GOTOFF(ebx,PD_0_125)  ; xmm1=PD_0_125
+
+    mulps       xmm6, xmm1              ; descale(1/8)
+    mulps       xmm7, xmm1              ; descale(1/8)
+    mulps       xmm5, xmm1              ; descale(1/8)
+    mulps       xmm0, xmm1              ; descale(1/8)
+
+    movhlps     xmm3, xmm6
+    movhlps     xmm1, xmm7
+    cvtps2pi    mm0, xmm6               ; round to int32, mm0=data0L=(00 10)
+    cvtps2pi    mm1, xmm7               ; round to int32, mm1=data1L=(01 11)
+    cvtps2pi    mm2, xmm3               ; round to int32, mm2=data0H=(20 30)
+    cvtps2pi    mm3, xmm1               ; round to int32, mm3=data1H=(21 31)
+    packssdw    mm0, mm2                ; mm0=data0=(00 10 20 30)
+    packssdw    mm1, mm3                ; mm1=data1=(01 11 21 31)
+
+    movhlps     xmm6, xmm5
+    movhlps     xmm7, xmm0
+    cvtps2pi    mm4, xmm5               ; round to int32, mm4=data7L=(07 17)
+    cvtps2pi    mm5, xmm0               ; round to int32, mm5=data6L=(06 16)
+    cvtps2pi    mm6, xmm6               ; round to int32, mm6=data7H=(27 37)
+    cvtps2pi    mm7, xmm7               ; round to int32, mm7=data6H=(26 36)
+    packssdw    mm4, mm6                ; mm4=data7=(07 17 27 37)
+    packssdw    mm5, mm7                ; mm5=data6=(06 16 26 36)
+
+    packsswb    mm0, mm5                ; mm0=(00 10 20 30 06 16 26 36)
+    packsswb    mm1, mm4                ; mm1=(01 11 21 31 07 17 27 37)
+
+    movaps      xmm3, XMMWORD wk(0)   ; xmm3=tmp2
+    movaps      xmm1, XMMWORD wk(1)   ; xmm1=tmp3
+
+    movaps      xmm6, GOTOFF(ebx,PD_0_125)  ; xmm6=PD_0_125
+
+    addps       xmm4, xmm2              ; xmm4=tmp4
+    movaps      xmm5, xmm3
+    movaps      xmm0, xmm1
+    addps       xmm3, xmm2              ; xmm3=data2=(02 12 22 32)
+    addps       xmm1, xmm4              ; xmm1=data4=(04 14 24 34)
+    subps       xmm5, xmm2              ; xmm5=data5=(05 15 25 35)
+    subps       xmm0, xmm4              ; xmm0=data3=(03 13 23 33)
+
+    mulps       xmm3, xmm6              ; descale(1/8)
+    mulps       xmm1, xmm6              ; descale(1/8)
+    mulps       xmm5, xmm6              ; descale(1/8)
+    mulps       xmm0, xmm6              ; descale(1/8)
+
+    movhlps     xmm7, xmm3
+    movhlps     xmm2, xmm1
+    cvtps2pi    mm2, xmm3               ; round to int32, mm2=data2L=(02 12)
+    cvtps2pi    mm3, xmm1               ; round to int32, mm3=data4L=(04 14)
+    cvtps2pi    mm6, xmm7               ; round to int32, mm6=data2H=(22 32)
+    cvtps2pi    mm7, xmm2               ; round to int32, mm7=data4H=(24 34)
+    packssdw    mm2, mm6                ; mm2=data2=(02 12 22 32)
+    packssdw    mm3, mm7                ; mm3=data4=(04 14 24 34)
+
+    movhlps     xmm4, xmm5
+    movhlps     xmm6, xmm0
+    cvtps2pi    mm5, xmm5               ; round to int32, mm5=data5L=(05 15)
+    cvtps2pi    mm4, xmm0               ; round to int32, mm4=data3L=(03 13)
+    cvtps2pi    mm6, xmm4               ; round to int32, mm6=data5H=(25 35)
+    cvtps2pi    mm7, xmm6               ; round to int32, mm7=data3H=(23 33)
+    packssdw    mm5, mm6                ; mm5=data5=(05 15 25 35)
+    packssdw    mm4, mm7                ; mm4=data3=(03 13 23 33)
+
+    movq        mm6, GOTOFF(ebx,PB_CENTERJSAMP)  ; mm6=PB_CENTERJSAMP
+
+    packsswb    mm2, mm3                ; mm2=(02 12 22 32 04 14 24 34)
+    packsswb    mm4, mm5                ; mm4=(03 13 23 33 05 15 25 35)
+
+    paddb       mm0, mm6
+    paddb       mm1, mm6
+    paddb       mm2, mm6
+    paddb       mm4, mm6
+
+    movq        mm7, mm0                ; transpose coefficients(phase 1)
+    punpcklbw   mm0, mm1                ; mm0=(00 01 10 11 20 21 30 31)
+    punpckhbw   mm7, mm1                ; mm7=(06 07 16 17 26 27 36 37)
+    movq        mm3, mm2                ; transpose coefficients(phase 1)
+    punpcklbw   mm2, mm4                ; mm2=(02 03 12 13 22 23 32 33)
+    punpckhbw   mm3, mm4                ; mm3=(04 05 14 15 24 25 34 35)
+
+    movq        mm5, mm0                ; transpose coefficients(phase 2)
+    punpcklwd   mm0, mm2                ; mm0=(00 01 02 03 10 11 12 13)
+    punpckhwd   mm5, mm2                ; mm5=(20 21 22 23 30 31 32 33)
+    movq        mm6, mm3                ; transpose coefficients(phase 2)
+    punpcklwd   mm3, mm7                ; mm3=(04 05 06 07 14 15 16 17)
+    punpckhwd   mm6, mm7                ; mm6=(24 25 26 27 34 35 36 37)
+
+    movq        mm1, mm0                ; transpose coefficients(phase 3)
+    punpckldq   mm0, mm3                ; mm0=(00 01 02 03 04 05 06 07)
+    punpckhdq   mm1, mm3                ; mm1=(10 11 12 13 14 15 16 17)
+    movq        mm4, mm5                ; transpose coefficients(phase 3)
+    punpckldq   mm5, mm6                ; mm5=(20 21 22 23 24 25 26 27)
+    punpckhdq   mm4, mm6                ; mm4=(30 31 32 33 34 35 36 37)
+
+    pushpic     ebx                     ; save GOT address
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         ebx, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    movq        MMWORD edx+eax*SIZEOF_JSAMPLE, mm0
+    movq        MMWORD ebx+eax*SIZEOF_JSAMPLE, mm1
+    mov         edx, JSAMPROW edi+2*SIZEOF_JSAMPROW
+    mov         ebx, JSAMPROW edi+3*SIZEOF_JSAMPROW
+    movq        MMWORD edx+eax*SIZEOF_JSAMPLE, mm5
+    movq        MMWORD ebx+eax*SIZEOF_JSAMPLE, mm4
+
+    poppic      ebx                     ; restore GOT address
+
+    add         esi, byte 4*SIZEOF_FAST_FLOAT  ; wsptr
+    add         edi, byte 4*SIZEOF_JSAMPROW
+    dec         ecx                            ; ctr
+    jnz         near .rowloop
+
+    emms                                ; empty MMX state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctflt-sse2.asm Added

@@ -0,0 +1,497 @@
+;
+; jidctflt.asm - floating-point IDCT (SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro unpcklps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+    shufps      %1, %2, 0x44
+%endmacro
+
+%macro unpckhps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+    shufps      %1, %2, 0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_float_sse2)
+
+EXTN(jconst_idct_float_sse2):
+
+PD_1_414        times 4  dd  1.414213562373095048801689
+PD_1_847        times 4  dd  1.847759065022573512256366
+PD_1_082        times 4  dd  1.082392200292393968799446
+PD_M2_613       times 4  dd -2.613125929752753055713286
+PD_RNDINT_MAGIC times 4  dd  100663296.0  ; (float)(0x00C00000 << 3)
+PB_CENTERJSAMP  times 16 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_sse2(void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; void *dct_table
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM         2
+%define workspace      wk(0) - DCTSIZE2 * SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspaceDCTSIZE2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_float_sse2)
+
+EXTN(jsimd_idct_float_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, workspace
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input, store into work array.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+    lea         edi, workspace                 ; FAST_FLOAT *wsptr
+    mov         ecx, DCTSIZE/4                   ; ctr
+    alignx      16, 7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         near .columnDCT
+
+    movq        xmm1, XMM_MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        xmm2, XMM_MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    movq        xmm3, XMM_MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    movq        xmm4, XMM_MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movq        xmm5, XMM_MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        xmm6, XMM_MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    movq        xmm7, XMM_MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    por         xmm1, xmm2
+    por         xmm3, xmm4
+    por         xmm5, xmm6
+    por         xmm1, xmm3
+    por         xmm5, xmm7
+    por         xmm1, xmm5
+    packsswb    xmm1, xmm1
+    movd        eax, xmm1
+    test        eax, eax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movq        xmm0, XMM_MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+
+    punpcklwd   xmm0, xmm0                  ; xmm0=(00 00 01 01 02 02 03 03)
+    psrad       xmm0, (DWORD_BIT-WORD_BIT)  ; xmm0=in0=(00 01 02 03)
+    cvtdq2ps    xmm0, xmm0                  ; xmm0=in0=(00 01 02 03)
+
+    mulps       xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movaps      xmm1, xmm0
+    movaps      xmm2, xmm0
+    movaps      xmm3, xmm0
+
+    shufps      xmm0, xmm0, 0x00        ; xmm0=(00 00 00 00)
+    shufps      xmm1, xmm1, 0x55        ; xmm1=(01 01 01 01)
+    shufps      xmm2, xmm2, 0xAA        ; xmm2=(02 02 02 02)
+    shufps      xmm3, xmm3, 0xFF        ; xmm3=(03 03 03 03)
+
+    movaps      XMMWORD XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT), xmm2
+    movaps      XMMWORD XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT), xmm2
+    movaps      XMMWORD XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT), xmm3
+    movaps      XMMWORD XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT), xmm3
+    jmp         near .nextcolumn
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movq        xmm0, XMM_MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movq        xmm1, XMM_MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    movq        xmm2, XMM_MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movq        xmm3, XMM_MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+
+    punpcklwd   xmm0, xmm0                  ; xmm0=(00 00 01 01 02 02 03 03)
+    punpcklwd   xmm1, xmm1                  ; xmm1=(20 20 21 21 22 22 23 23)
+    psrad       xmm0, (DWORD_BIT-WORD_BIT)  ; xmm0=in0=(00 01 02 03)
+    psrad       xmm1, (DWORD_BIT-WORD_BIT)  ; xmm1=in2=(20 21 22 23)
+    cvtdq2ps    xmm0, xmm0                  ; xmm0=in0=(00 01 02 03)
+    cvtdq2ps    xmm1, xmm1                  ; xmm1=in2=(20 21 22 23)
+
+    punpcklwd   xmm2, xmm2                  ; xmm2=(40 40 41 41 42 42 43 43)
+    punpcklwd   xmm3, xmm3                  ; xmm3=(60 60 61 61 62 62 63 63)
+    psrad       xmm2, (DWORD_BIT-WORD_BIT)  ; xmm2=in4=(40 41 42 43)
+    psrad       xmm3, (DWORD_BIT-WORD_BIT)  ; xmm3=in6=(60 61 62 63)
+    cvtdq2ps    xmm2, xmm2                  ; xmm2=in4=(40 41 42 43)
+    cvtdq2ps    xmm3, xmm3                  ; xmm3=in6=(60 61 62 63)
+
+    mulps       xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm1, XMMWORD XMMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm2, XMMWORD XMMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm3, XMMWORD XMMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movaps      xmm4, xmm0
+    movaps      xmm5, xmm1
+    subps       xmm0, xmm2              ; xmm0=tmp11
+    subps       xmm1, xmm3
+    addps       xmm4, xmm2              ; xmm4=tmp10
+    addps       xmm5, xmm3              ; xmm5=tmp13
+
+    mulps       xmm1, GOTOFF(ebx,PD_1_414)
+    subps       xmm1, xmm5              ; xmm1=tmp12
+
+    movaps      xmm6, xmm4
+    movaps      xmm7, xmm0
+    subps       xmm4, xmm5              ; xmm4=tmp3
+    subps       xmm0, xmm1              ; xmm0=tmp2
+    addps       xmm6, xmm5              ; xmm6=tmp0
+    addps       xmm7, xmm1              ; xmm7=tmp1
+
+    movaps      XMMWORD wk(1), xmm4   ; tmp3
+    movaps      XMMWORD wk(0), xmm0   ; tmp2
+
+    ; -- Odd part
+
+    movq        xmm2, XMM_MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        xmm3, XMM_MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    movq        xmm5, XMM_MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        xmm1, XMM_MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+
+    punpcklwd   xmm2, xmm2                  ; xmm2=(10 10 11 11 12 12 13 13)
+    punpcklwd   xmm3, xmm3                  ; xmm3=(30 30 31 31 32 32 33 33)
+    psrad       xmm2, (DWORD_BIT-WORD_BIT)  ; xmm2=in1=(10 11 12 13)
+    psrad       xmm3, (DWORD_BIT-WORD_BIT)  ; xmm3=in3=(30 31 32 33)
+    cvtdq2ps    xmm2, xmm2                  ; xmm2=in1=(10 11 12 13)
+    cvtdq2ps    xmm3, xmm3                  ; xmm3=in3=(30 31 32 33)
+
+    punpcklwd   xmm5, xmm5                  ; xmm5=(50 50 51 51 52 52 53 53)
+    punpcklwd   xmm1, xmm1                  ; xmm1=(70 70 71 71 72 72 73 73)
+    psrad       xmm5, (DWORD_BIT-WORD_BIT)  ; xmm5=in5=(50 51 52 53)
+    psrad       xmm1, (DWORD_BIT-WORD_BIT)  ; xmm1=in7=(70 71 72 73)
+    cvtdq2ps    xmm5, xmm5                  ; xmm5=in5=(50 51 52 53)
+    cvtdq2ps    xmm1, xmm1                  ; xmm1=in7=(70 71 72 73)
+
+    mulps       xmm2, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm3, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm5, XMMWORD XMMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm1, XMMWORD XMMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movaps      xmm4, xmm2
+    movaps      xmm0, xmm5
+    addps       xmm2, xmm1              ; xmm2=z11
+    addps       xmm5, xmm3              ; xmm5=z13
+    subps       xmm4, xmm1              ; xmm4=z12
+    subps       xmm0, xmm3              ; xmm0=z10
+
+    movaps      xmm1, xmm2
+    subps       xmm2, xmm5
+    addps       xmm1, xmm5              ; xmm1=tmp7
+
+    mulps       xmm2, GOTOFF(ebx,PD_1_414)  ; xmm2=tmp11
+
+    movaps      xmm3, xmm0
+    addps       xmm0, xmm4
+    mulps       xmm0, GOTOFF(ebx,PD_1_847)   ; xmm0=z5
+    mulps       xmm3, GOTOFF(ebx,PD_M2_613)  ; xmm3=(z10 * -2.613125930)
+    mulps       xmm4, GOTOFF(ebx,PD_1_082)   ; xmm4=(z12 * 1.082392200)
+    addps       xmm3, xmm0                     ; xmm3=tmp12
+    subps       xmm4, xmm0                     ; xmm4=tmp10
+
+    ; -- Final output stage
+
+    subps       xmm3, xmm1              ; xmm3=tmp6
+    movaps      xmm5, xmm6
+    movaps      xmm0, xmm7
+    addps       xmm6, xmm1              ; xmm6=data0=(00 01 02 03)
+    addps       xmm7, xmm3              ; xmm7=data1=(10 11 12 13)
+    subps       xmm5, xmm1              ; xmm5=data7=(70 71 72 73)
+    subps       xmm0, xmm3              ; xmm0=data6=(60 61 62 63)
+    subps       xmm2, xmm3              ; xmm2=tmp5
+
+    movaps      xmm1, xmm6              ; transpose coefficients(phase 1)
+    unpcklps    xmm6, xmm7              ; xmm6=(00 10 01 11)
+    unpckhps    xmm1, xmm7              ; xmm1=(02 12 03 13)
+    movaps      xmm3, xmm0              ; transpose coefficients(phase 1)
+    unpcklps    xmm0, xmm5              ; xmm0=(60 70 61 71)
+    unpckhps    xmm3, xmm5              ; xmm3=(62 72 63 73)
+
+    movaps      xmm7, XMMWORD wk(0)   ; xmm7=tmp2
+    movaps      xmm5, XMMWORD wk(1)   ; xmm5=tmp3
+
+    movaps      XMMWORD wk(0), xmm0   ; wk(0)=(60 70 61 71)
+    movaps      XMMWORD wk(1), xmm3   ; wk(1)=(62 72 63 73)
+
+    addps       xmm4, xmm2              ; xmm4=tmp4
+    movaps      xmm0, xmm7
+    movaps      xmm3, xmm5
+    addps       xmm7, xmm2              ; xmm7=data2=(20 21 22 23)
+    addps       xmm5, xmm4              ; xmm5=data4=(40 41 42 43)
+    subps       xmm0, xmm2              ; xmm0=data5=(50 51 52 53)
+    subps       xmm3, xmm4              ; xmm3=data3=(30 31 32 33)
+
+    movaps      xmm2, xmm7              ; transpose coefficients(phase 1)
+    unpcklps    xmm7, xmm3              ; xmm7=(20 30 21 31)
+    unpckhps    xmm2, xmm3              ; xmm2=(22 32 23 33)
+    movaps      xmm4, xmm5              ; transpose coefficients(phase 1)
+    unpcklps    xmm5, xmm0              ; xmm5=(40 50 41 51)
+    unpckhps    xmm4, xmm0              ; xmm4=(42 52 43 53)
+
+    movaps      xmm3, xmm6              ; transpose coefficients(phase 2)
+    unpcklps2   xmm6, xmm7              ; xmm6=(00 10 20 30)
+    unpckhps2   xmm3, xmm7              ; xmm3=(01 11 21 31)
+    movaps      xmm0, xmm1              ; transpose coefficients(phase 2)
+    unpcklps2   xmm1, xmm2              ; xmm1=(02 12 22 32)
+    unpckhps2   xmm0, xmm2              ; xmm0=(03 13 23 33)
+
+    movaps      xmm7, XMMWORD wk(0)   ; xmm7=(60 70 61 71)
+    movaps      xmm2, XMMWORD wk(1)   ; xmm2=(62 72 63 73)
+
+    movaps      XMMWORD XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT), xmm6
+    movaps      XMMWORD XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT), xmm3
+    movaps      XMMWORD XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT), xmm0
+
+    movaps      xmm6, xmm5              ; transpose coefficients(phase 2)
+    unpcklps2   xmm5, xmm7              ; xmm5=(40 50 60 70)
+    unpckhps2   xmm6, xmm7              ; xmm6=(41 51 61 71)
+    movaps      xmm3, xmm4              ; transpose coefficients(phase 2)
+    unpcklps2   xmm4, xmm2              ; xmm4=(42 52 62 72)
+    unpckhps2   xmm3, xmm2              ; xmm3=(43 53 63 73)
+
+    movaps      XMMWORD XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT), xmm6
+    movaps      XMMWORD XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT), xmm4
+    movaps      XMMWORD XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT), xmm3
+
+.nextcolumn:
+    add         esi, byte 4*SIZEOF_JCOEF               ; coef_block
+    add         edx, byte 4*SIZEOF_FLOAT_MULT_TYPE     ; quantptr
+    add         edi,      4*DCTSIZE*SIZEOF_FAST_FLOAT  ; wsptr
+    dec         ecx                                    ; ctr
+    jnz         near .columnloop
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32
+    prefetchnta esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32
+    prefetchnta esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32
+    prefetchnta esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         eax, original_ebp
+    lea         esi, workspace                   ; FAST_FLOAT *wsptr
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+    mov         ecx, DCTSIZE/4                     ; ctr
+    alignx      16, 7
+.rowloop:
+
+    ; -- Even part
+
+    movaps      xmm0, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)
+
+    movaps      xmm4, xmm0
+    movaps      xmm5, xmm1
+    subps       xmm0, xmm2              ; xmm0=tmp11
+    subps       xmm1, xmm3
+    addps       xmm4, xmm2              ; xmm4=tmp10
+    addps       xmm5, xmm3              ; xmm5=tmp13
+
+    mulps       xmm1, GOTOFF(ebx,PD_1_414)
+    subps       xmm1, xmm5              ; xmm1=tmp12
+
+    movaps      xmm6, xmm4
+    movaps      xmm7, xmm0
+    subps       xmm4, xmm5              ; xmm4=tmp3
+    subps       xmm0, xmm1              ; xmm0=tmp2
+    addps       xmm6, xmm5              ; xmm6=tmp0
+    addps       xmm7, xmm1              ; xmm7=tmp1
+
+    movaps      XMMWORD wk(1), xmm4   ; tmp3
+    movaps      XMMWORD wk(0), xmm0   ; tmp2
+
+    ; -- Odd part
+
+    movaps      xmm2, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm5, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)
+
+    movaps      xmm4, xmm2
+    movaps      xmm0, xmm5
+    addps       xmm2, xmm1              ; xmm2=z11
+    addps       xmm5, xmm3              ; xmm5=z13
+    subps       xmm4, xmm1              ; xmm4=z12
+    subps       xmm0, xmm3              ; xmm0=z10
+
+    movaps      xmm1, xmm2
+    subps       xmm2, xmm5
+    addps       xmm1, xmm5              ; xmm1=tmp7
+
+    mulps       xmm2, GOTOFF(ebx,PD_1_414)  ; xmm2=tmp11
+
+    movaps      xmm3, xmm0
+    addps       xmm0, xmm4
+    mulps       xmm0, GOTOFF(ebx,PD_1_847)   ; xmm0=z5
+    mulps       xmm3, GOTOFF(ebx,PD_M2_613)  ; xmm3=(z10 * -2.613125930)
+    mulps       xmm4, GOTOFF(ebx,PD_1_082)   ; xmm4=(z12 * 1.082392200)
+    addps       xmm3, xmm0                     ; xmm3=tmp12
+    subps       xmm4, xmm0                     ; xmm4=tmp10
+
+    ; -- Final output stage
+
+    subps       xmm3, xmm1              ; xmm3=tmp6
+    movaps      xmm5, xmm6
+    movaps      xmm0, xmm7
+    addps       xmm6, xmm1              ; xmm6=data0=(00 10 20 30)
+    addps       xmm7, xmm3              ; xmm7=data1=(01 11 21 31)
+    subps       xmm5, xmm1              ; xmm5=data7=(07 17 27 37)
+    subps       xmm0, xmm3              ; xmm0=data6=(06 16 26 36)
+    subps       xmm2, xmm3              ; xmm2=tmp5
+
+    movaps      xmm1, GOTOFF(ebx,PD_RNDINT_MAGIC)  ; xmm1=PD_RNDINT_MAGIC
+    pcmpeqd     xmm3, xmm3
+    psrld       xmm3, WORD_BIT          ; xmm3={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+    addps       xmm6, xmm1              ; xmm6=roundint(data0/8)=(00 ** 10 ** 20 ** 30 **)
+    addps       xmm7, xmm1              ; xmm7=roundint(data1/8)=(01 ** 11 ** 21 ** 31 **)
+    addps       xmm0, xmm1              ; xmm0=roundint(data6/8)=(06 ** 16 ** 26 ** 36 **)
+    addps       xmm5, xmm1              ; xmm5=roundint(data7/8)=(07 ** 17 ** 27 ** 37 **)
+
+    pand        xmm6, xmm3              ; xmm6=(00 -- 10 -- 20 -- 30 --)
+    pslld       xmm7, WORD_BIT          ; xmm7=(-- 01 -- 11 -- 21 -- 31)
+    pand        xmm0, xmm3              ; xmm0=(06 -- 16 -- 26 -- 36 --)
+    pslld       xmm5, WORD_BIT          ; xmm5=(-- 07 -- 17 -- 27 -- 37)
+    por         xmm6, xmm7              ; xmm6=(00 01 10 11 20 21 30 31)
+    por         xmm0, xmm5              ; xmm0=(06 07 16 17 26 27 36 37)
+
+    movaps      xmm1, XMMWORD wk(0)   ; xmm1=tmp2
+    movaps      xmm3, XMMWORD wk(1)   ; xmm3=tmp3
+
+    addps       xmm4, xmm2              ; xmm4=tmp4
+    movaps      xmm7, xmm1
+    movaps      xmm5, xmm3
+    addps       xmm1, xmm2              ; xmm1=data2=(02 12 22 32)
+    addps       xmm3, xmm4              ; xmm3=data4=(04 14 24 34)
+    subps       xmm7, xmm2              ; xmm7=data5=(05 15 25 35)
+    subps       xmm5, xmm4              ; xmm5=data3=(03 13 23 33)
+
+    movaps      xmm2, GOTOFF(ebx,PD_RNDINT_MAGIC)  ; xmm2=PD_RNDINT_MAGIC
+    pcmpeqd     xmm4, xmm4
+    psrld       xmm4, WORD_BIT          ; xmm4={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+    addps       xmm3, xmm2              ; xmm3=roundint(data4/8)=(04 ** 14 ** 24 ** 34 **)
+    addps       xmm7, xmm2              ; xmm7=roundint(data5/8)=(05 ** 15 ** 25 ** 35 **)
+    addps       xmm1, xmm2              ; xmm1=roundint(data2/8)=(02 ** 12 ** 22 ** 32 **)
+    addps       xmm5, xmm2              ; xmm5=roundint(data3/8)=(03 ** 13 ** 23 ** 33 **)
+
+    pand        xmm3, xmm4              ; xmm3=(04 -- 14 -- 24 -- 34 --)
+    pslld       xmm7, WORD_BIT          ; xmm7=(-- 05 -- 15 -- 25 -- 35)
+    pand        xmm1, xmm4              ; xmm1=(02 -- 12 -- 22 -- 32 --)
+    pslld       xmm5, WORD_BIT          ; xmm5=(-- 03 -- 13 -- 23 -- 33)
+    por         xmm3, xmm7              ; xmm3=(04 05 14 15 24 25 34 35)
+    por         xmm1, xmm5              ; xmm1=(02 03 12 13 22 23 32 33)
+
+    movdqa      xmm2, GOTOFF(ebx,PB_CENTERJSAMP)  ; xmm2=PB_CENTERJSAMP
+
+    packsswb    xmm6, xmm3        ; xmm6=(00 01 10 11 20 21 30 31 04 05 14 15 24 25 34 35)
+    packsswb    xmm1, xmm0        ; xmm1=(02 03 12 13 22 23 32 33 06 07 16 17 26 27 36 37)
+    paddb       xmm6, xmm2
+    paddb       xmm1, xmm2
+
+    movdqa      xmm4, xmm6        ; transpose coefficients(phase 2)
+    punpcklwd   xmm6, xmm1        ; xmm6=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+    punpckhwd   xmm4, xmm1        ; xmm4=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+
+    movdqa      xmm7, xmm6        ; transpose coefficients(phase 3)
+    punpckldq   xmm6, xmm4        ; xmm6=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+    punpckhdq   xmm7, xmm4        ; xmm7=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+
+    pshufd      xmm5, xmm6, 0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+    pshufd      xmm3, xmm7, 0x4E  ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+
+    pushpic     ebx                     ; save GOT address
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         ebx, JSAMPROW edi+2*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm6
+    movq        XMM_MMWORD ebx+eax*SIZEOF_JSAMPLE, xmm7
+    mov         edx, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    mov         ebx, JSAMPROW edi+3*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm5
+    movq        XMM_MMWORD ebx+eax*SIZEOF_JSAMPLE, xmm3
+
+    poppic      ebx                     ; restore GOT address
+
+    add         esi, byte 4*SIZEOF_FAST_FLOAT  ; wsptr
+    add         edi, byte 4*SIZEOF_JSAMPROW
+    dec         ecx                            ; ctr
+    jnz         near .rowloop
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctfst-mmx.asm Added

@@ -0,0 +1,499 @@
+;
+; jidctfst.asm - fast integer IDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the inverse DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jidctfst.c; see the jidctfst.c
+; for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  8  ; 14 is also OK.
+%define PASS1_BITS  2
+
+%if IFAST_SCALE_BITS != PASS1_BITS
+%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'."
+%endif
+
+%if CONST_BITS == 8
+F_1_082 equ 277              ; FIX(1.082392200)
+F_1_414 equ 362              ; FIX(1.414213562)
+F_1_847 equ 473              ; FIX(1.847759065)
+F_2_613 equ 669              ; FIX(2.613125930)
+F_1_613 equ (F_2_613 - 256)  ; FIX(2.613125930) - FIX(1)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_1_082 equ DESCALE(1162209775, 30 - CONST_BITS)  ; FIX(1.082392200)
+F_1_414 equ DESCALE(1518500249, 30 - CONST_BITS)  ; FIX(1.414213562)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_2_613 equ DESCALE(2805822602, 30 - CONST_BITS)  ; FIX(2.613125930)
+F_1_613 equ (F_2_613 - (1 << CONST_BITS))       ; FIX(2.613125930) - FIX(1)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS  2
+%define CONST_SHIFT              (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_ifast_mmx)
+
+EXTN(jconst_idct_ifast_mmx):
+
+PW_F1414       times 4 dw  F_1_414 << CONST_SHIFT
+PW_F1847       times 4 dw  F_1_847 << CONST_SHIFT
+PW_MF1613      times 4 dw -F_1_613 << CONST_SHIFT
+PW_F1082       times 4 dw  F_1_082 << CONST_SHIFT
+PB_CENTERJSAMP times 8 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_ifast_mmx(void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; jpeg_component_info *compptr
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_MMWORD
+                                        ; mmword wkWK_NUM
+%define WK_NUM         2
+%define workspace      wk(0) - DCTSIZE2 * SIZEOF_JCOEF
+                                        ; JCOEF workspaceDCTSIZE2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_ifast_mmx)
+
+EXTN(jsimd_idct_ifast_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, workspace
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input, store into work array.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+    lea         edi, workspace                 ; JCOEF *wsptr
+    mov         ecx, DCTSIZE/4                   ; ctr
+    alignx      16, 7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_IFAST_MMX
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         short .columnDCT
+
+    movq        mm0, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    por         mm1, MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    por         mm1, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    por         mm1, mm0
+    packsswb    mm1, mm1
+    movd        eax, mm1
+    test        eax, eax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    pmullw      mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)
+
+    movq        mm2, mm0                ; mm0=in0=(00 01 02 03)
+    punpcklwd   mm0, mm0                ; mm0=(00 00 01 01)
+    punpckhwd   mm2, mm2                ; mm2=(02 02 03 03)
+
+    movq        mm1, mm0
+    punpckldq   mm0, mm0                ; mm0=(00 00 00 00)
+    punpckhdq   mm1, mm1                ; mm1=(01 01 01 01)
+    movq        mm3, mm2
+    punpckldq   mm2, mm2                ; mm2=(02 02 02 02)
+    punpckhdq   mm3, mm3                ; mm3=(03 03 03 03)
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_JCOEF), mm1
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_JCOEF), mm1
+    movq        MMWORD MMBLOCK(2,0,edi,SIZEOF_JCOEF), mm2
+    movq        MMWORD MMBLOCK(2,1,edi,SIZEOF_JCOEF), mm2
+    movq        MMWORD MMBLOCK(3,0,edi,SIZEOF_JCOEF), mm3
+    movq        MMWORD MMBLOCK(3,1,edi,SIZEOF_JCOEF), mm3
+    jmp         near .nextcolumn
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    pmullw      mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      mm1, MMWORD MMBLOCK(2,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    movq        mm2, MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    pmullw      mm2, MMWORD MMBLOCK(4,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      mm3, MMWORD MMBLOCK(6,0,edx,SIZEOF_IFAST_MULT_TYPE)
+
+    movq        mm4, mm0
+    movq        mm5, mm1
+    psubw       mm0, mm2                ; mm0=tmp11
+    psubw       mm1, mm3
+    paddw       mm4, mm2                ; mm4=tmp10
+    paddw       mm5, mm3                ; mm5=tmp13
+
+    psllw       mm1, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      mm1, GOTOFF(ebx,PW_F1414)
+    psubw       mm1, mm5                ; mm1=tmp12
+
+    movq        mm6, mm4
+    movq        mm7, mm0
+    psubw       mm4, mm5                ; mm4=tmp3
+    psubw       mm0, mm1                ; mm0=tmp2
+    paddw       mm6, mm5                ; mm6=tmp0
+    paddw       mm7, mm1                ; mm7=tmp1
+
+    movq        MMWORD wk(1), mm4     ; wk(1)=tmp3
+    movq        MMWORD wk(0), mm0     ; wk(0)=tmp2
+
+    ; -- Odd part
+
+    movq        mm2, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    pmullw      mm2, MMWORD MMBLOCK(1,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      mm3, MMWORD MMBLOCK(3,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    movq        mm5, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    pmullw      mm5, MMWORD MMBLOCK(5,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      mm1, MMWORD MMBLOCK(7,0,edx,SIZEOF_IFAST_MULT_TYPE)
+
+    movq        mm4, mm2
+    movq        mm0, mm5
+    psubw       mm2, mm1                ; mm2=z12
+    psubw       mm5, mm3                ; mm5=z10
+    paddw       mm4, mm1                ; mm4=z11
+    paddw       mm0, mm3                ; mm0=z13
+
+    movq        mm1, mm5                ; mm1=z10(unscaled)
+    psllw       mm2, PRE_MULTIPLY_SCALE_BITS
+    psllw       mm5, PRE_MULTIPLY_SCALE_BITS
+
+    movq        mm3, mm4
+    psubw       mm4, mm0
+    paddw       mm3, mm0                ; mm3=tmp7
+
+    psllw       mm4, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      mm4, GOTOFF(ebx,PW_F1414)  ; mm4=tmp11
+
+    ; To avoid overflow...
+    ;
+    ; (Original)
+    ; tmp12 = -2.613125930 * z10 + z5;
+    ;
+    ; (This implementation)
+    ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+    ;       = -1.613125930 * z10 - z10 + z5;
+
+    movq        mm0, mm5
+    paddw       mm5, mm2
+    pmulhw      mm5, GOTOFF(ebx,PW_F1847)   ; mm5=z5
+    pmulhw      mm0, GOTOFF(ebx,PW_MF1613)
+    pmulhw      mm2, GOTOFF(ebx,PW_F1082)
+    psubw       mm0, mm1
+    psubw       mm2, mm5                ; mm2=tmp10
+    paddw       mm0, mm5                ; mm0=tmp12
+
+    ; -- Final output stage
+
+    psubw       mm0, mm3                ; mm0=tmp6
+    movq        mm1, mm6
+    movq        mm5, mm7
+    paddw       mm6, mm3                ; mm6=data0=(00 01 02 03)
+    paddw       mm7, mm0                ; mm7=data1=(10 11 12 13)
+    psubw       mm1, mm3                ; mm1=data7=(70 71 72 73)
+    psubw       mm5, mm0                ; mm5=data6=(60 61 62 63)
+    psubw       mm4, mm0                ; mm4=tmp5
+
+    movq        mm3, mm6                ; transpose coefficients(phase 1)
+    punpcklwd   mm6, mm7                ; mm6=(00 10 01 11)
+    punpckhwd   mm3, mm7                ; mm3=(02 12 03 13)
+    movq        mm0, mm5                ; transpose coefficients(phase 1)
+    punpcklwd   mm5, mm1                ; mm5=(60 70 61 71)
+    punpckhwd   mm0, mm1                ; mm0=(62 72 63 73)
+
+    movq        mm7, MMWORD wk(0)     ; mm7=tmp2
+    movq        mm1, MMWORD wk(1)     ; mm1=tmp3
+
+    movq        MMWORD wk(0), mm5     ; wk(0)=(60 70 61 71)
+    movq        MMWORD wk(1), mm0     ; wk(1)=(62 72 63 73)
+
+    paddw       mm2, mm4                ; mm2=tmp4
+    movq        mm5, mm7
+    movq        mm0, mm1
+    paddw       mm7, mm4                ; mm7=data2=(20 21 22 23)
+    paddw       mm1, mm2                ; mm1=data4=(40 41 42 43)
+    psubw       mm5, mm4                ; mm5=data5=(50 51 52 53)
+    psubw       mm0, mm2                ; mm0=data3=(30 31 32 33)
+
+    movq        mm4, mm7                ; transpose coefficients(phase 1)
+    punpcklwd   mm7, mm0                ; mm7=(20 30 21 31)
+    punpckhwd   mm4, mm0                ; mm4=(22 32 23 33)
+    movq        mm2, mm1                ; transpose coefficients(phase 1)
+    punpcklwd   mm1, mm5                ; mm1=(40 50 41 51)
+    punpckhwd   mm2, mm5                ; mm2=(42 52 43 53)
+
+    movq        mm0, mm6                ; transpose coefficients(phase 2)
+    punpckldq   mm6, mm7                ; mm6=(00 10 20 30)
+    punpckhdq   mm0, mm7                ; mm0=(01 11 21 31)
+    movq        mm5, mm3                ; transpose coefficients(phase 2)
+    punpckldq   mm3, mm4                ; mm3=(02 12 22 32)
+    punpckhdq   mm5, mm4                ; mm5=(03 13 23 33)
+
+    movq        mm7, MMWORD wk(0)     ; mm7=(60 70 61 71)
+    movq        mm4, MMWORD wk(1)     ; mm4=(62 72 63 73)
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_JCOEF), mm6
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(2,0,edi,SIZEOF_JCOEF), mm3
+    movq        MMWORD MMBLOCK(3,0,edi,SIZEOF_JCOEF), mm5
+
+    movq        mm6, mm1                ; transpose coefficients(phase 2)
+    punpckldq   mm1, mm7                ; mm1=(40 50 60 70)
+    punpckhdq   mm6, mm7                ; mm6=(41 51 61 71)
+    movq        mm0, mm2                ; transpose coefficients(phase 2)
+    punpckldq   mm2, mm4                ; mm2=(42 52 62 72)
+    punpckhdq   mm0, mm4                ; mm0=(43 53 63 73)
+
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_JCOEF), mm1
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_JCOEF), mm6
+    movq        MMWORD MMBLOCK(2,1,edi,SIZEOF_JCOEF), mm2
+    movq        MMWORD MMBLOCK(3,1,edi,SIZEOF_JCOEF), mm0
+
+.nextcolumn:
+    add         esi, byte 4*SIZEOF_JCOEF            ; coef_block
+    add         edx, byte 4*SIZEOF_IFAST_MULT_TYPE  ; quantptr
+    add         edi, byte 4*DCTSIZE*SIZEOF_JCOEF    ; wsptr
+    dec         ecx                                 ; ctr
+    jnz         near .columnloop
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         eax, original_ebp
+    lea         esi, workspace                   ; JCOEF *wsptr
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+    mov         ecx, DCTSIZE/4                     ; ctr
+    alignx      16, 7
+.rowloop:
+
+    ; -- Even part
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    movq        mm2, MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+
+    movq        mm4, mm0
+    movq        mm5, mm1
+    psubw       mm0, mm2                ; mm0=tmp11
+    psubw       mm1, mm3
+    paddw       mm4, mm2                ; mm4=tmp10
+    paddw       mm5, mm3                ; mm5=tmp13
+
+    psllw       mm1, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      mm1, GOTOFF(ebx,PW_F1414)
+    psubw       mm1, mm5                ; mm1=tmp12
+
+    movq        mm6, mm4
+    movq        mm7, mm0
+    psubw       mm4, mm5                ; mm4=tmp3
+    psubw       mm0, mm1                ; mm0=tmp2
+    paddw       mm6, mm5                ; mm6=tmp0
+    paddw       mm7, mm1                ; mm7=tmp1
+
+    movq        MMWORD wk(1), mm4     ; wk(1)=tmp3
+    movq        MMWORD wk(0), mm0     ; wk(0)=tmp2
+
+    ; -- Odd part
+
+    movq        mm2, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    movq        mm5, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+
+    movq        mm4, mm2
+    movq        mm0, mm5
+    psubw       mm2, mm1                ; mm2=z12
+    psubw       mm5, mm3                ; mm5=z10
+    paddw       mm4, mm1                ; mm4=z11
+    paddw       mm0, mm3                ; mm0=z13
+
+    movq        mm1, mm5                ; mm1=z10(unscaled)
+    psllw       mm2, PRE_MULTIPLY_SCALE_BITS
+    psllw       mm5, PRE_MULTIPLY_SCALE_BITS
+
+    movq        mm3, mm4
+    psubw       mm4, mm0
+    paddw       mm3, mm0                ; mm3=tmp7
+
+    psllw       mm4, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      mm4, GOTOFF(ebx,PW_F1414)  ; mm4=tmp11
+
+    ; To avoid overflow...
+    ;
+    ; (Original)
+    ; tmp12 = -2.613125930 * z10 + z5;
+    ;
+    ; (This implementation)
+    ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+    ;       = -1.613125930 * z10 - z10 + z5;
+
+    movq        mm0, mm5
+    paddw       mm5, mm2
+    pmulhw      mm5, GOTOFF(ebx,PW_F1847)   ; mm5=z5
+    pmulhw      mm0, GOTOFF(ebx,PW_MF1613)
+    pmulhw      mm2, GOTOFF(ebx,PW_F1082)
+    psubw       mm0, mm1
+    psubw       mm2, mm5                ; mm2=tmp10
+    paddw       mm0, mm5                ; mm0=tmp12
+
+    ; -- Final output stage
+
+    psubw       mm0, mm3                ; mm0=tmp6
+    movq        mm1, mm6
+    movq        mm5, mm7
+    paddw       mm6, mm3                ; mm6=data0=(00 10 20 30)
+    paddw       mm7, mm0                ; mm7=data1=(01 11 21 31)
+    psraw       mm6, (PASS1_BITS+3)     ; descale
+    psraw       mm7, (PASS1_BITS+3)     ; descale
+    psubw       mm1, mm3                ; mm1=data7=(07 17 27 37)
+    psubw       mm5, mm0                ; mm5=data6=(06 16 26 36)
+    psraw       mm1, (PASS1_BITS+3)     ; descale
+    psraw       mm5, (PASS1_BITS+3)     ; descale
+    psubw       mm4, mm0                ; mm4=tmp5
+
+    packsswb    mm6, mm5                ; mm6=(00 10 20 30 06 16 26 36)
+    packsswb    mm7, mm1                ; mm7=(01 11 21 31 07 17 27 37)
+
+    movq        mm3, MMWORD wk(0)     ; mm3=tmp2
+    movq        mm0, MMWORD wk(1)     ; mm0=tmp3
+
+    paddw       mm2, mm4                ; mm2=tmp4
+    movq        mm5, mm3
+    movq        mm1, mm0
+    paddw       mm3, mm4                ; mm3=data2=(02 12 22 32)
+    paddw       mm0, mm2                ; mm0=data4=(04 14 24 34)
+    psraw       mm3, (PASS1_BITS+3)     ; descale
+    psraw       mm0, (PASS1_BITS+3)     ; descale
+    psubw       mm5, mm4                ; mm5=data5=(05 15 25 35)
+    psubw       mm1, mm2                ; mm1=data3=(03 13 23 33)
+    psraw       mm5, (PASS1_BITS+3)     ; descale
+    psraw       mm1, (PASS1_BITS+3)     ; descale
+
+    movq        mm4, GOTOFF(ebx,PB_CENTERJSAMP)  ; mm4=PB_CENTERJSAMP
+
+    packsswb    mm3, mm0                ; mm3=(02 12 22 32 04 14 24 34)
+    packsswb    mm1, mm5                ; mm1=(03 13 23 33 05 15 25 35)
+
+    paddb       mm6, mm4
+    paddb       mm7, mm4
+    paddb       mm3, mm4
+    paddb       mm1, mm4
+
+    movq        mm2, mm6                ; transpose coefficients(phase 1)
+    punpcklbw   mm6, mm7                ; mm6=(00 01 10 11 20 21 30 31)
+    punpckhbw   mm2, mm7                ; mm2=(06 07 16 17 26 27 36 37)
+    movq        mm0, mm3                ; transpose coefficients(phase 1)
+    punpcklbw   mm3, mm1                ; mm3=(02 03 12 13 22 23 32 33)
+    punpckhbw   mm0, mm1                ; mm0=(04 05 14 15 24 25 34 35)
+
+    movq        mm5, mm6                ; transpose coefficients(phase 2)
+    punpcklwd   mm6, mm3                ; mm6=(00 01 02 03 10 11 12 13)
+    punpckhwd   mm5, mm3                ; mm5=(20 21 22 23 30 31 32 33)
+    movq        mm4, mm0                ; transpose coefficients(phase 2)
+    punpcklwd   mm0, mm2                ; mm0=(04 05 06 07 14 15 16 17)
+    punpckhwd   mm4, mm2                ; mm4=(24 25 26 27 34 35 36 37)
+
+    movq        mm7, mm6                ; transpose coefficients(phase 3)
+    punpckldq   mm6, mm0                ; mm6=(00 01 02 03 04 05 06 07)
+    punpckhdq   mm7, mm0                ; mm7=(10 11 12 13 14 15 16 17)
+    movq        mm1, mm5                ; transpose coefficients(phase 3)
+    punpckldq   mm5, mm4                ; mm5=(20 21 22 23 24 25 26 27)
+    punpckhdq   mm1, mm4                ; mm1=(30 31 32 33 34 35 36 37)
+
+    pushpic     ebx                     ; save GOT address
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         ebx, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    movq        MMWORD edx+eax*SIZEOF_JSAMPLE, mm6
+    movq        MMWORD ebx+eax*SIZEOF_JSAMPLE, mm7
+    mov         edx, JSAMPROW edi+2*SIZEOF_JSAMPROW
+    mov         ebx, JSAMPROW edi+3*SIZEOF_JSAMPROW
+    movq        MMWORD edx+eax*SIZEOF_JSAMPLE, mm5
+    movq        MMWORD ebx+eax*SIZEOF_JSAMPLE, mm1
+
+    poppic      ebx                     ; restore GOT address
+
+    add         esi, byte 4*SIZEOF_JCOEF     ; wsptr
+    add         edi, byte 4*SIZEOF_JSAMPROW
+    dec         ecx                          ; ctr
+    jnz         near .rowloop
+
+    emms                                ; empty MMX state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctfst-sse2.asm Added

@@ -0,0 +1,501 @@
+;
+; jidctfst.asm - fast integer IDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the inverse DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jidctfst.c; see the jidctfst.c
+; for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  8  ; 14 is also OK.
+%define PASS1_BITS  2
+
+%if IFAST_SCALE_BITS != PASS1_BITS
+%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'."
+%endif
+
+%if CONST_BITS == 8
+F_1_082 equ 277              ; FIX(1.082392200)
+F_1_414 equ 362              ; FIX(1.414213562)
+F_1_847 equ 473              ; FIX(1.847759065)
+F_2_613 equ 669              ; FIX(2.613125930)
+F_1_613 equ (F_2_613 - 256)  ; FIX(2.613125930) - FIX(1)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_1_082 equ DESCALE(1162209775, 30 - CONST_BITS)  ; FIX(1.082392200)
+F_1_414 equ DESCALE(1518500249, 30 - CONST_BITS)  ; FIX(1.414213562)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_2_613 equ DESCALE(2805822602, 30 - CONST_BITS)  ; FIX(2.613125930)
+F_1_613 equ (F_2_613 - (1 << CONST_BITS))       ; FIX(2.613125930) - FIX(1)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS  2
+%define CONST_SHIFT              (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_ifast_sse2)
+
+EXTN(jconst_idct_ifast_sse2):
+
+PW_F1414       times 8  dw  F_1_414 << CONST_SHIFT
+PW_F1847       times 8  dw  F_1_847 << CONST_SHIFT
+PW_MF1613      times 8  dw -F_1_613 << CONST_SHIFT
+PW_F1082       times 8  dw  F_1_082 << CONST_SHIFT
+PB_CENTERJSAMP times 16 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_ifast_sse2(void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; jpeg_component_info *compptr
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM         2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_ifast_sse2)
+
+EXTN(jsimd_idct_ifast_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_IFAST_SSE2
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         near .columnDCT
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    por         xmm1, xmm0
+    packsswb    xmm1, xmm1
+    packsswb    xmm1, xmm1
+    movd        eax, xmm1
+    test        eax, eax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    movdqa      xmm7, xmm0              ; xmm0=in0=(00 01 02 03 04 05 06 07)
+    punpcklwd   xmm0, xmm0              ; xmm0=(00 00 01 01 02 02 03 03)
+    punpckhwd   xmm7, xmm7              ; xmm7=(04 04 05 05 06 06 07 07)
+
+    pshufd      xmm6, xmm0, 0x00        ; xmm6=col0=(00 00 00 00 00 00 00 00)
+    pshufd      xmm2, xmm0, 0x55        ; xmm2=col1=(01 01 01 01 01 01 01 01)
+    pshufd      xmm5, xmm0, 0xAA        ; xmm5=col2=(02 02 02 02 02 02 02 02)
+    pshufd      xmm0, xmm0, 0xFF        ; xmm0=col3=(03 03 03 03 03 03 03 03)
+    pshufd      xmm1, xmm7, 0x00        ; xmm1=col4=(04 04 04 04 04 04 04 04)
+    pshufd      xmm4, xmm7, 0x55        ; xmm4=col5=(05 05 05 05 05 05 05 05)
+    pshufd      xmm3, xmm7, 0xAA        ; xmm3=col6=(06 06 06 06 06 06 06 06)
+    pshufd      xmm7, xmm7, 0xFF        ; xmm7=col7=(07 07 07 07 07 07 07 07)
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=col1
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=col3
+    jmp         near .column_end
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(2,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    movdqa      xmm2, XMMWORD XMMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(4,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(6,0,edx,SIZEOF_IFAST_MULT_TYPE)
+
+    movdqa      xmm4, xmm0
+    movdqa      xmm5, xmm1
+    psubw       xmm0, xmm2              ; xmm0=tmp11
+    psubw       xmm1, xmm3
+    paddw       xmm4, xmm2              ; xmm4=tmp10
+    paddw       xmm5, xmm3              ; xmm5=tmp13
+
+    psllw       xmm1, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm1, GOTOFF(ebx,PW_F1414)
+    psubw       xmm1, xmm5              ; xmm1=tmp12
+
+    movdqa      xmm6, xmm4
+    movdqa      xmm7, xmm0
+    psubw       xmm4, xmm5              ; xmm4=tmp3
+    psubw       xmm0, xmm1              ; xmm0=tmp2
+    paddw       xmm6, xmm5              ; xmm6=tmp0
+    paddw       xmm7, xmm1              ; xmm7=tmp1
+
+    movdqa      XMMWORD wk(1), xmm4   ; wk(1)=tmp3
+    movdqa      XMMWORD wk(0), xmm0   ; wk(0)=tmp2
+
+    ; -- Odd part
+
+    movdqa      xmm2, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    movdqa      xmm5, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm5, XMMWORD XMMBLOCK(5,0,edx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(7,0,edx,SIZEOF_IFAST_MULT_TYPE)
+
+    movdqa      xmm4, xmm2
+    movdqa      xmm0, xmm5
+    psubw       xmm2, xmm1              ; xmm2=z12
+    psubw       xmm5, xmm3              ; xmm5=z10
+    paddw       xmm4, xmm1              ; xmm4=z11
+    paddw       xmm0, xmm3              ; xmm0=z13
+
+    movdqa      xmm1, xmm5              ; xmm1=z10(unscaled)
+    psllw       xmm2, PRE_MULTIPLY_SCALE_BITS
+    psllw       xmm5, PRE_MULTIPLY_SCALE_BITS
+
+    movdqa      xmm3, xmm4
+    psubw       xmm4, xmm0
+    paddw       xmm3, xmm0              ; xmm3=tmp7
+
+    psllw       xmm4, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm4, GOTOFF(ebx,PW_F1414)  ; xmm4=tmp11
+
+    ; To avoid overflow...
+    ;
+    ; (Original)
+    ; tmp12 = -2.613125930 * z10 + z5;
+    ;
+    ; (This implementation)
+    ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+    ;       = -1.613125930 * z10 - z10 + z5;
+
+    movdqa      xmm0, xmm5
+    paddw       xmm5, xmm2
+    pmulhw      xmm5, GOTOFF(ebx,PW_F1847)   ; xmm5=z5
+    pmulhw      xmm0, GOTOFF(ebx,PW_MF1613)
+    pmulhw      xmm2, GOTOFF(ebx,PW_F1082)
+    psubw       xmm0, xmm1
+    psubw       xmm2, xmm5              ; xmm2=tmp10
+    paddw       xmm0, xmm5              ; xmm0=tmp12
+
+    ; -- Final output stage
+
+    psubw       xmm0, xmm3              ; xmm0=tmp6
+    movdqa      xmm1, xmm6
+    movdqa      xmm5, xmm7
+    paddw       xmm6, xmm3              ; xmm6=data0=(00 01 02 03 04 05 06 07)
+    paddw       xmm7, xmm0              ; xmm7=data1=(10 11 12 13 14 15 16 17)
+    psubw       xmm1, xmm3              ; xmm1=data7=(70 71 72 73 74 75 76 77)
+    psubw       xmm5, xmm0              ; xmm5=data6=(60 61 62 63 64 65 66 67)
+    psubw       xmm4, xmm0              ; xmm4=tmp5
+
+    movdqa      xmm3, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm7              ; xmm6=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm3, xmm7              ; xmm3=(04 14 05 15 06 16 07 17)
+    movdqa      xmm0, xmm5              ; transpose coefficients(phase 1)
+    punpcklwd   xmm5, xmm1              ; xmm5=(60 70 61 71 62 72 63 73)
+    punpckhwd   xmm0, xmm1              ; xmm0=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=tmp2
+    movdqa      xmm1, XMMWORD wk(1)   ; xmm1=tmp3
+
+    movdqa      XMMWORD wk(0), xmm5   ; wk(0)=(60 70 61 71 62 72 63 73)
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=(64 74 65 75 66 76 67 77)
+
+    paddw       xmm2, xmm4              ; xmm2=tmp4
+    movdqa      xmm5, xmm7
+    movdqa      xmm0, xmm1
+    paddw       xmm7, xmm4              ; xmm7=data2=(20 21 22 23 24 25 26 27)
+    paddw       xmm1, xmm2              ; xmm1=data4=(40 41 42 43 44 45 46 47)
+    psubw       xmm5, xmm4              ; xmm5=data5=(50 51 52 53 54 55 56 57)
+    psubw       xmm0, xmm2              ; xmm0=data3=(30 31 32 33 34 35 36 37)
+
+    movdqa      xmm4, xmm7              ; transpose coefficients(phase 1)
+    punpcklwd   xmm7, xmm0              ; xmm7=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm4, xmm0              ; xmm4=(24 34 25 35 26 36 27 37)
+    movdqa      xmm2, xmm1              ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm5              ; xmm1=(40 50 41 51 42 52 43 53)
+    punpckhwd   xmm2, xmm5              ; xmm2=(44 54 45 55 46 56 47 57)
+
+    movdqa      xmm0, xmm3              ; transpose coefficients(phase 2)
+    punpckldq   xmm3, xmm4              ; xmm3=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm0, xmm4              ; xmm0=(06 16 26 36 07 17 27 37)
+    movdqa      xmm5, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm7              ; xmm6=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm5, xmm7              ; xmm5=(02 12 22 32 03 13 23 33)
+
+    movdqa      xmm4, XMMWORD wk(0)   ; xmm4=(60 70 61 71 62 72 63 73)
+    movdqa      xmm7, XMMWORD wk(1)   ; xmm7=(64 74 65 75 66 76 67 77)
+
+    movdqa      XMMWORD wk(0), xmm3   ; wk(0)=(04 14 24 34 05 15 25 35)
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm3, xmm1              ; transpose coefficients(phase 2)
+    punpckldq   xmm1, xmm4              ; xmm1=(40 50 60 70 41 51 61 71)
+    punpckhdq   xmm3, xmm4              ; xmm3=(42 52 62 72 43 53 63 73)
+    movdqa      xmm0, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm7              ; xmm2=(44 54 64 74 45 55 65 75)
+    punpckhdq   xmm0, xmm7              ; xmm0=(46 56 66 76 47 57 67 77)
+
+    movdqa      xmm4, xmm6              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm6, xmm1              ; xmm6=col0=(00 10 20 30 40 50 60 70)
+    punpckhqdq  xmm4, xmm1              ; xmm4=col1=(01 11 21 31 41 51 61 71)
+    movdqa      xmm7, xmm5              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm5, xmm3              ; xmm5=col2=(02 12 22 32 42 52 62 72)
+    punpckhqdq  xmm7, xmm3              ; xmm7=col3=(03 13 23 33 43 53 63 73)
+
+    movdqa      xmm1, XMMWORD wk(0)   ; xmm1=(04 14 24 34 05 15 25 35)
+    movdqa      xmm3, XMMWORD wk(1)   ; xmm3=(06 16 26 36 07 17 27 37)
+
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=col1
+    movdqa      XMMWORD wk(1), xmm7   ; wk(1)=col3
+
+    movdqa      xmm4, xmm1              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm1, xmm2              ; xmm1=col4=(04 14 24 34 44 54 64 74)
+    punpckhqdq  xmm4, xmm2              ; xmm4=col5=(05 15 25 35 45 55 65 75)
+    movdqa      xmm7, xmm3              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm3, xmm0              ; xmm3=col6=(06 16 26 36 46 56 66 76)
+    punpckhqdq  xmm7, xmm0              ; xmm7=col7=(07 17 27 37 47 57 67 77)
+.column_end:
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         eax, original_ebp
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+
+    ; -- Even part
+
+    ; xmm6=col0, xmm5=col2, xmm1=col4, xmm3=col6
+
+    movdqa      xmm2, xmm6
+    movdqa      xmm0, xmm5
+    psubw       xmm6, xmm1              ; xmm6=tmp11
+    psubw       xmm5, xmm3
+    paddw       xmm2, xmm1              ; xmm2=tmp10
+    paddw       xmm0, xmm3              ; xmm0=tmp13
+
+    psllw       xmm5, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm5, GOTOFF(ebx,PW_F1414)
+    psubw       xmm5, xmm0              ; xmm5=tmp12
+
+    movdqa      xmm1, xmm2
+    movdqa      xmm3, xmm6
+    psubw       xmm2, xmm0              ; xmm2=tmp3
+    psubw       xmm6, xmm5              ; xmm6=tmp2
+    paddw       xmm1, xmm0              ; xmm1=tmp0
+    paddw       xmm3, xmm5              ; xmm3=tmp1
+
+    movdqa      xmm0, XMMWORD wk(0)   ; xmm0=col1
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=col3
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=tmp3
+    movdqa      XMMWORD wk(1), xmm6   ; wk(1)=tmp2
+
+    ; -- Odd part
+
+    ; xmm0=col1, xmm5=col3, xmm4=col5, xmm7=col7
+
+    movdqa      xmm2, xmm0
+    movdqa      xmm6, xmm4
+    psubw       xmm0, xmm7              ; xmm0=z12
+    psubw       xmm4, xmm5              ; xmm4=z10
+    paddw       xmm2, xmm7              ; xmm2=z11
+    paddw       xmm6, xmm5              ; xmm6=z13
+
+    movdqa      xmm7, xmm4              ; xmm7=z10(unscaled)
+    psllw       xmm0, PRE_MULTIPLY_SCALE_BITS
+    psllw       xmm4, PRE_MULTIPLY_SCALE_BITS
+
+    movdqa      xmm5, xmm2
+    psubw       xmm2, xmm6
+    paddw       xmm5, xmm6              ; xmm5=tmp7
+
+    psllw       xmm2, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm2, GOTOFF(ebx,PW_F1414)  ; xmm2=tmp11
+
+    ; To avoid overflow...
+    ;
+    ; (Original)
+    ; tmp12 = -2.613125930 * z10 + z5;
+    ;
+    ; (This implementation)
+    ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+    ;       = -1.613125930 * z10 - z10 + z5;
+
+    movdqa      xmm6, xmm4
+    paddw       xmm4, xmm0
+    pmulhw      xmm4, GOTOFF(ebx,PW_F1847)   ; xmm4=z5
+    pmulhw      xmm6, GOTOFF(ebx,PW_MF1613)
+    pmulhw      xmm0, GOTOFF(ebx,PW_F1082)
+    psubw       xmm6, xmm7
+    psubw       xmm0, xmm4              ; xmm0=tmp10
+    paddw       xmm6, xmm4              ; xmm6=tmp12
+
+    ; -- Final output stage
+
+    psubw       xmm6, xmm5              ; xmm6=tmp6
+    movdqa      xmm7, xmm1
+    movdqa      xmm4, xmm3
+    paddw       xmm1, xmm5              ; xmm1=data0=(00 10 20 30 40 50 60 70)
+    paddw       xmm3, xmm6              ; xmm3=data1=(01 11 21 31 41 51 61 71)
+    psraw       xmm1, (PASS1_BITS+3)    ; descale
+    psraw       xmm3, (PASS1_BITS+3)    ; descale
+    psubw       xmm7, xmm5              ; xmm7=data7=(07 17 27 37 47 57 67 77)
+    psubw       xmm4, xmm6              ; xmm4=data6=(06 16 26 36 46 56 66 76)
+    psraw       xmm7, (PASS1_BITS+3)    ; descale
+    psraw       xmm4, (PASS1_BITS+3)    ; descale
+    psubw       xmm2, xmm6              ; xmm2=tmp5
+
+    packsswb    xmm1, xmm4        ; xmm1=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+    packsswb    xmm3, xmm7        ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=tmp2
+    movdqa      xmm6, XMMWORD wk(0)   ; xmm6=tmp3
+
+    paddw       xmm0, xmm2              ; xmm0=tmp4
+    movdqa      xmm4, xmm5
+    movdqa      xmm7, xmm6
+    paddw       xmm5, xmm2              ; xmm5=data2=(02 12 22 32 42 52 62 72)
+    paddw       xmm6, xmm0              ; xmm6=data4=(04 14 24 34 44 54 64 74)
+    psraw       xmm5, (PASS1_BITS+3)    ; descale
+    psraw       xmm6, (PASS1_BITS+3)    ; descale
+    psubw       xmm4, xmm2              ; xmm4=data5=(05 15 25 35 45 55 65 75)
+    psubw       xmm7, xmm0              ; xmm7=data3=(03 13 23 33 43 53 63 73)
+    psraw       xmm4, (PASS1_BITS+3)    ; descale
+    psraw       xmm7, (PASS1_BITS+3)    ; descale
+
+    movdqa      xmm2, GOTOFF(ebx,PB_CENTERJSAMP)  ; xmm2=PB_CENTERJSAMP
+
+    packsswb    xmm5, xmm6        ; xmm5=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+    packsswb    xmm7, xmm4        ; xmm7=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+    paddb       xmm1, xmm2
+    paddb       xmm3, xmm2
+    paddb       xmm5, xmm2
+    paddb       xmm7, xmm2
+
+    movdqa      xmm0, xmm1        ; transpose coefficients(phase 1)
+    punpcklbw   xmm1, xmm3        ; xmm1=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+    punpckhbw   xmm0, xmm3        ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+    movdqa      xmm6, xmm5        ; transpose coefficients(phase 1)
+    punpcklbw   xmm5, xmm7        ; xmm5=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+    punpckhbw   xmm6, xmm7        ; xmm6=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+    movdqa      xmm4, xmm1        ; transpose coefficients(phase 2)
+    punpcklwd   xmm1, xmm5        ; xmm1=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+    punpckhwd   xmm4, xmm5        ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+    movdqa      xmm2, xmm6        ; transpose coefficients(phase 2)
+    punpcklwd   xmm6, xmm0        ; xmm6=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+    punpckhwd   xmm2, xmm0        ; xmm2=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+    movdqa      xmm3, xmm1        ; transpose coefficients(phase 3)
+    punpckldq   xmm1, xmm6        ; xmm1=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+    punpckhdq   xmm3, xmm6        ; xmm3=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+    movdqa      xmm7, xmm4        ; transpose coefficients(phase 3)
+    punpckldq   xmm4, xmm2        ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+    punpckhdq   xmm7, xmm2        ; xmm7=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+    pshufd      xmm5, xmm1, 0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+    pshufd      xmm0, xmm3, 0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+    pshufd      xmm6, xmm4, 0x4E  ; xmm6=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+    pshufd      xmm2, xmm7, 0x4E  ; xmm2=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+2*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm1
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm3
+    mov         edx, JSAMPROW edi+4*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+6*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm4
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm7
+
+    mov         edx, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+3*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm5
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm0
+    mov         edx, JSAMPROW edi+5*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+7*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm6
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm2
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctint-avx2.asm Added

@@ -0,0 +1,453 @@
+;
+; jidctint.asm - accurate integer IDCT (AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, 2018, 2020, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS + 3)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+; In-place 8x8x16-bit inverse matrix transpose using AVX2 instructions
+; %1-%4: Input/output registers
+; %5-%8: Temp registers
+
+%macro dotranspose 8
+    ; %5=(00 10 20 30 40 50 60 70  01 11 21 31 41 51 61 71)
+    ; %6=(03 13 23 33 43 53 63 73  02 12 22 32 42 52 62 72)
+    ; %7=(04 14 24 34 44 54 64 74  05 15 25 35 45 55 65 75)
+    ; %8=(07 17 27 37 47 57 67 77  06 16 26 36 46 56 66 76)
+
+    vpermq      %5, %1, 0xD8
+    vpermq      %6, %2, 0x72
+    vpermq      %7, %3, 0xD8
+    vpermq      %8, %4, 0x72
+    ; transpose coefficients(phase 1)
+    ; %5=(00 10 20 30 01 11 21 31  40 50 60 70 41 51 61 71)
+    ; %6=(02 12 22 32 03 13 23 33  42 52 62 72 43 53 63 73)
+    ; %7=(04 14 24 34 05 15 25 35  44 54 64 74 45 55 65 75)
+    ; %8=(06 16 26 36 07 17 27 37  46 56 66 76 47 57 67 77)
+
+    vpunpcklwd  %1, %5, %6
+    vpunpckhwd  %2, %5, %6
+    vpunpcklwd  %3, %7, %8
+    vpunpckhwd  %4, %7, %8
+    ; transpose coefficients(phase 2)
+    ; %1=(00 02 10 12 20 22 30 32  40 42 50 52 60 62 70 72)
+    ; %2=(01 03 11 13 21 23 31 33  41 43 51 53 61 63 71 73)
+    ; %3=(04 06 14 16 24 26 34 36  44 46 54 56 64 66 74 76)
+    ; %4=(05 07 15 17 25 27 35 37  45 47 55 57 65 67 75 77)
+
+    vpunpcklwd  %5, %1, %2
+    vpunpcklwd  %6, %3, %4
+    vpunpckhwd  %7, %1, %2
+    vpunpckhwd  %8, %3, %4
+    ; transpose coefficients(phase 3)
+    ; %5=(00 01 02 03 10 11 12 13  40 41 42 43 50 51 52 53)
+    ; %6=(04 05 06 07 14 15 16 17  44 45 46 47 54 55 56 57)
+    ; %7=(20 21 22 23 30 31 32 33  60 61 62 63 70 71 72 73)
+    ; %8=(24 25 26 27 34 35 36 37  64 65 66 67 74 75 76 77)
+
+    vpunpcklqdq %1, %5, %6
+    vpunpckhqdq %2, %5, %6
+    vpunpcklqdq %3, %7, %8
+    vpunpckhqdq %4, %7, %8
+    ; transpose coefficients(phase 4)
+    ; %1=(00 01 02 03 04 05 06 07  40 41 42 43 44 45 46 47)
+    ; %2=(10 11 12 13 14 15 16 17  50 51 52 53 54 55 56 57)
+    ; %3=(20 21 22 23 24 25 26 27  60 61 62 63 64 65 66 67)
+    ; %4=(30 31 32 33 34 35 36 37  70 71 72 73 74 75 76 77)
+%endmacro
+
+; --------------------------------------------------------------------------
+; In-place 8x8x16-bit accurate integer inverse DCT using AVX2 instructions
+; %1-%4:  Input/output registers
+; %5-%12: Temp registers
+; %9:     Pass (1 or 2)
+
+%macro dodct 13
+    ; -- Even part
+
+    ; (Original)
+    ; z1 = (z2 + z3) * 0.541196100;
+    ; tmp2 = z1 + z3 * -1.847759065;
+    ; tmp3 = z1 + z2 * 0.765366865;
+    ;
+    ; (This implementation)
+    ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+    ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+    vperm2i128  %6, %3, %3, 0x01        ; %6=in6_2
+    vpunpcklwd  %5, %3, %6              ; %5=in26_62L
+    vpunpckhwd  %6, %3, %6              ; %6=in26_62H
+    vpmaddwd    %5, %5, GOTOFF(ebx,PW_F130_F054_MF130_F054)  ; %5=tmp3_2L
+    vpmaddwd    %6, %6, GOTOFF(ebx,PW_F130_F054_MF130_F054)  ; %6=tmp3_2H
+
+    vperm2i128  %7, %1, %1, 0x01        ; %7=in4_0
+    vpsignw     %1, %1, GOTOFF(ebx,PW_1_NEG1)
+    vpaddw      %7, %7, %1              ; %7=(in0+in4)_(in0-in4)
+
+    vpxor       %1, %1, %1
+    vpunpcklwd  %8, %1, %7              ; %8=tmp0_1L
+    vpunpckhwd  %1, %1, %7              ; %1=tmp0_1H
+    vpsrad      %8, %8, (16-CONST_BITS)  ; vpsrad %8,16 & vpslld %8,CONST_BITS
+    vpsrad      %1, %1, (16-CONST_BITS)  ; vpsrad %1,16 & vpslld %1,CONST_BITS
+
+    vpsubd      %3, %8, %5
+    vmovdqu     %11, %3                 ; %11=tmp0_1L-tmp3_2L=tmp13_12L
+    vpaddd      %3, %8, %5
+    vmovdqu     %9, %3                  ; %9=tmp0_1L+tmp3_2L=tmp10_11L
+    vpsubd      %3, %1, %6
+    vmovdqu     %12, %3                 ; %12=tmp0_1H-tmp3_2H=tmp13_12H
+    vpaddd      %3, %1, %6
+    vmovdqu     %10, %3                 ; %10=tmp0_1H+tmp3_2H=tmp10_11H
+
+    ; -- Odd part
+
+    vpaddw      %1, %4, %2              ; %1=in7_5+in3_1=z3_4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    vperm2i128  %8, %1, %1, 0x01        ; %8=z4_3
+    vpunpcklwd  %7, %1, %8              ; %7=z34_43L
+    vpunpckhwd  %8, %1, %8              ; %8=z34_43H
+    vpmaddwd    %7, %7, GOTOFF(ebx,PW_MF078_F117_F078_F117)  ; %7=z3_4L
+    vpmaddwd    %8, %8, GOTOFF(ebx,PW_MF078_F117_F078_F117)  ; %8=z3_4H
+
+    ; (Original)
+    ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+    ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+    ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+    ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+    ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+    ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+    ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+    ; tmp0 += z3;  tmp1 += z4;
+    ; tmp2 += z3;  tmp3 += z4;
+
+    vperm2i128  %2, %2, %2, 0x01        ; %2=in1_3
+    vpunpcklwd  %3, %4, %2              ; %3=in71_53L
+    vpunpckhwd  %4, %4, %2              ; %4=in71_53H
+
+    vpmaddwd    %5, %3, GOTOFF(ebx,PW_MF060_MF089_MF050_MF256)  ; %5=tmp0_1L
+    vpmaddwd    %6, %4, GOTOFF(ebx,PW_MF060_MF089_MF050_MF256)  ; %6=tmp0_1H
+    vpaddd      %5, %5, %7              ; %5=tmp0_1L+z3_4L=tmp0_1L
+    vpaddd      %6, %6, %8              ; %6=tmp0_1H+z3_4H=tmp0_1H
+
+    vpmaddwd    %3, %3, GOTOFF(ebx,PW_MF089_F060_MF256_F050)  ; %3=tmp3_2L
+    vpmaddwd    %4, %4, GOTOFF(ebx,PW_MF089_F060_MF256_F050)  ; %4=tmp3_2H
+    vperm2i128  %7, %7, %7, 0x01        ; %7=z4_3L
+    vperm2i128  %8, %8, %8, 0x01        ; %8=z4_3H
+    vpaddd      %7, %3, %7              ; %7=tmp3_2L+z4_3L=tmp3_2L
+    vpaddd      %8, %4, %8              ; %8=tmp3_2H+z4_3H=tmp3_2H
+
+    ; -- Final output stage
+
+    vmovdqu     %3, %9
+    vmovdqu     %4, %10
+
+    vpaddd      %1, %3, %7              ; %1=tmp10_11L+tmp3_2L=data0_1L
+    vpaddd      %2, %4, %8              ; %2=tmp10_11H+tmp3_2H=data0_1H
+    vpaddd      %1, %1, GOTOFF(ebx,PD_DESCALE_P %+ %13)
+    vpaddd      %2, %2, GOTOFF(ebx,PD_DESCALE_P %+ %13)
+    vpsrad      %1, %1, DESCALE_P %+ %13
+    vpsrad      %2, %2, DESCALE_P %+ %13
+    vpackssdw   %1, %1, %2              ; %1=data0_1
+
+    vpsubd      %3, %3, %7              ; %3=tmp10_11L-tmp3_2L=data7_6L
+    vpsubd      %4, %4, %8              ; %4=tmp10_11H-tmp3_2H=data7_6H
+    vpaddd      %3, %3, GOTOFF(ebx,PD_DESCALE_P %+ %13)
+    vpaddd      %4, %4, GOTOFF(ebx,PD_DESCALE_P %+ %13)
+    vpsrad      %3, %3, DESCALE_P %+ %13
+    vpsrad      %4, %4, DESCALE_P %+ %13
+    vpackssdw   %4, %3, %4              ; %4=data7_6
+
+    vmovdqu     %7, %11
+    vmovdqu     %8, %12
+
+    vpaddd      %2, %7, %5              ; %7=tmp13_12L+tmp0_1L=data3_2L
+    vpaddd      %3, %8, %6              ; %8=tmp13_12H+tmp0_1H=data3_2H
+    vpaddd      %2, %2, GOTOFF(ebx,PD_DESCALE_P %+ %13)
+    vpaddd      %3, %3, GOTOFF(ebx,PD_DESCALE_P %+ %13)
+    vpsrad      %2, %2, DESCALE_P %+ %13
+    vpsrad      %3, %3, DESCALE_P %+ %13
+    vpackssdw   %2, %2, %3              ; %2=data3_2
+
+    vpsubd      %3, %7, %5              ; %7=tmp13_12L-tmp0_1L=data4_5L
+    vpsubd      %6, %8, %6              ; %8=tmp13_12H-tmp0_1H=data4_5H
+    vpaddd      %3, %3, GOTOFF(ebx,PD_DESCALE_P %+ %13)
+    vpaddd      %6, %6, GOTOFF(ebx,PD_DESCALE_P %+ %13)
+    vpsrad      %3, %3, DESCALE_P %+ %13
+    vpsrad      %6, %6, DESCALE_P %+ %13
+    vpackssdw   %3, %3, %6              ; %3=data4_5
+%endmacro
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_islow_avx2)
+
+EXTN(jconst_idct_islow_avx2):
+
+PW_F130_F054_MF130_F054    times 4  dw  (F_0_541 + F_0_765),  F_0_541
+                           times 4  dw  (F_0_541 - F_1_847),  F_0_541
+PW_MF078_F117_F078_F117    times 4  dw  (F_1_175 - F_1_961),  F_1_175
+                           times 4  dw  (F_1_175 - F_0_390),  F_1_175
+PW_MF060_MF089_MF050_MF256 times 4  dw  (F_0_298 - F_0_899), -F_0_899
+                           times 4  dw  (F_2_053 - F_2_562), -F_2_562
+PW_MF089_F060_MF256_F050   times 4  dw -F_0_899, (F_1_501 - F_0_899)
+                           times 4  dw -F_2_562, (F_3_072 - F_2_562)
+PD_DESCALE_P1              times 8  dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2              times 8  dd  1 << (DESCALE_P2 - 1)
+PB_CENTERJSAMP             times 32 db  CENTERJSAMPLE
+PW_1_NEG1                  times 8  dw  1
+                           times 8  dw -1
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_avx2(void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; jpeg_component_info *compptr
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_YMMWORD
+                                        ; ymmword wkWK_NUM
+%define WK_NUM         4
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_islow_avx2)
+
+EXTN(jsimd_idct_islow_avx2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_AVX2
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         near .columnDCT
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    vpor        xmm0, xmm0, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    vpor        xmm1, xmm1, XMMWORD XMMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    vpor        xmm0, xmm0, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    vpor        xmm1, xmm1, XMMWORD XMMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    vpor        xmm0, xmm0, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    vpor        xmm1, xmm1, xmm0
+    vpacksswb   xmm1, xmm1, xmm1
+    vpacksswb   xmm1, xmm1, xmm1
+    movd        eax, xmm1
+    test        eax, eax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movdqa      xmm5, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    vpmullw     xmm5, xmm5, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    vpsllw      xmm5, xmm5, PASS1_BITS
+
+    vpunpcklwd  xmm4, xmm5, xmm5        ; xmm4=(00 00 01 01 02 02 03 03)
+    vpunpckhwd  xmm5, xmm5, xmm5        ; xmm5=(04 04 05 05 06 06 07 07)
+    vinserti128 ymm4, ymm4, xmm5, 1
+
+    vpshufd     ymm0, ymm4, 0x00        ; ymm0=col0_4=(00 00 00 00 00 00 00 00  04 04 04 04 04 04 04 04)
+    vpshufd     ymm1, ymm4, 0x55        ; ymm1=col1_5=(01 01 01 01 01 01 01 01  05 05 05 05 05 05 05 05)
+    vpshufd     ymm2, ymm4, 0xAA        ; ymm2=col2_6=(02 02 02 02 02 02 02 02  06 06 06 06 06 06 06 06)
+    vpshufd     ymm3, ymm4, 0xFF        ; ymm3=col3_7=(03 03 03 03 03 03 03 03  07 07 07 07 07 07 07 07)
+
+    jmp         near .column_end
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    vmovdqu     ymm4, YMMWORD YMMBLOCK(0,0,esi,SIZEOF_JCOEF)  ; ymm4=in0_1
+    vmovdqu     ymm5, YMMWORD YMMBLOCK(2,0,esi,SIZEOF_JCOEF)  ; ymm5=in2_3
+    vmovdqu     ymm6, YMMWORD YMMBLOCK(4,0,esi,SIZEOF_JCOEF)  ; ymm6=in4_5
+    vmovdqu     ymm7, YMMWORD YMMBLOCK(6,0,esi,SIZEOF_JCOEF)  ; ymm7=in6_7
+    vpmullw     ymm4, ymm4, YMMWORD YMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    vpmullw     ymm5, ymm5, YMMWORD YMMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    vpmullw     ymm6, ymm6, YMMWORD YMMBLOCK(4,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    vpmullw     ymm7, ymm7, YMMWORD YMMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    vperm2i128  ymm0, ymm4, ymm6, 0x20  ; ymm0=in0_4
+    vperm2i128  ymm1, ymm5, ymm4, 0x31  ; ymm1=in3_1
+    vperm2i128  ymm2, ymm5, ymm7, 0x20  ; ymm2=in2_6
+    vperm2i128  ymm3, ymm7, ymm6, 0x31  ; ymm3=in7_5
+
+    dodct ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7, XMMWORD wk(0), XMMWORD wk(1), XMMWORD wk(2), XMMWORD wk(3), 1
+    ; ymm0=data0_1, ymm1=data3_2, ymm2=data4_5, ymm3=data7_6
+
+    dotranspose ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7
+    ; ymm0=data0_4, ymm1=data1_5, ymm2=data2_6, ymm3=data3_7
+
+.column_end:
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows.
+
+    mov         eax, original_ebp
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+
+    vperm2i128  ymm4, ymm3, ymm1, 0x31  ; ymm3=in7_5
+    vperm2i128  ymm1, ymm3, ymm1, 0x20  ; ymm1=in3_1
+
+    dodct ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7, XMMWORD wk(0), XMMWORD wk(1), XMMWORD wk(2), XMMWORD wk(3), 2
+    ; ymm0=data0_1, ymm1=data3_2, ymm2=data4_5, ymm4=data7_6
+
+    dotranspose ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7
+    ; ymm0=data0_4, ymm1=data1_5, ymm2=data2_6, ymm4=data3_7
+
+    vpacksswb   ymm0, ymm0, ymm1        ; ymm0=data01_45
+    vpacksswb   ymm1, ymm2, ymm4        ; ymm1=data23_67
+    vpaddb      ymm0, ymm0, GOTOFF(ebx,PB_CENTERJSAMP)
+    vpaddb      ymm1, ymm1, GOTOFF(ebx,PB_CENTERJSAMP)
+
+    vextracti128 xmm6, ymm1, 1          ; xmm3=data67
+    vextracti128 xmm4, ymm0, 1          ; xmm2=data45
+    vextracti128 xmm2, ymm1, 0          ; xmm1=data23
+    vextracti128 xmm0, ymm0, 0          ; xmm0=data01
+
+    vpshufd     xmm1, xmm0, 0x4E  ; xmm1=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+    vpshufd     xmm3, xmm2, 0x4E  ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+    vpshufd     xmm5, xmm4, 0x4E  ; xmm5=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+    vpshufd     xmm7, xmm6, 0x4E  ; xmm7=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+    vzeroupper
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         esi, JSAMPROW edi+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm0
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm1
+
+    mov         edx, JSAMPROW edi+2*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         esi, JSAMPROW edi+3*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm2
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm3
+
+    mov         edx, JSAMPROW edi+4*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         esi, JSAMPROW edi+5*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm4
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm5
+
+    mov         edx, JSAMPROW edi+6*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         esi, JSAMPROW edi+7*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm6
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm7
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctint-mmx.asm Added

@@ -0,0 +1,851 @@
+;
+; jidctint.asm - accurate integer IDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, 2020, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS + 3)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_islow_mmx)
+
+EXTN(jconst_idct_islow_mmx):
+
+PW_F130_F054   times 2 dw  (F_0_541 + F_0_765),  F_0_541
+PW_F054_MF130  times 2 dw  F_0_541, (F_0_541 - F_1_847)
+PW_MF078_F117  times 2 dw  (F_1_175 - F_1_961),  F_1_175
+PW_F117_F078   times 2 dw  F_1_175, (F_1_175 - F_0_390)
+PW_MF060_MF089 times 2 dw  (F_0_298 - F_0_899), -F_0_899
+PW_MF089_F060  times 2 dw -F_0_899, (F_1_501 - F_0_899)
+PW_MF050_MF256 times 2 dw  (F_2_053 - F_2_562), -F_2_562
+PW_MF256_F050  times 2 dw -F_2_562, (F_3_072 - F_2_562)
+PD_DESCALE_P1  times 2 dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2  times 2 dd  1 << (DESCALE_P2 - 1)
+PB_CENTERJSAMP times 8 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_mmx(void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; jpeg_component_info *compptr
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_MMWORD
+                                        ; mmword wkWK_NUM
+%define WK_NUM         12
+%define workspace      wk(0) - DCTSIZE2 * SIZEOF_JCOEF
+                                        ; JCOEF workspaceDCTSIZE2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_islow_mmx)
+
+EXTN(jsimd_idct_islow_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, workspace
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input, store into work array.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+    lea         edi, workspace                 ; JCOEF *wsptr
+    mov         ecx, DCTSIZE/4                   ; ctr
+    alignx      16, 7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_MMX
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         short .columnDCT
+
+    movq        mm0, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    por         mm1, MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    por         mm1, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    por         mm1, mm0
+    packsswb    mm1, mm1
+    movd        eax, mm1
+    test        eax, eax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    pmullw      mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    psllw       mm0, PASS1_BITS
+
+    movq        mm2, mm0                ; mm0=in0=(00 01 02 03)
+    punpcklwd   mm0, mm0                ; mm0=(00 00 01 01)
+    punpckhwd   mm2, mm2                ; mm2=(02 02 03 03)
+
+    movq        mm1, mm0
+    punpckldq   mm0, mm0                ; mm0=(00 00 00 00)
+    punpckhdq   mm1, mm1                ; mm1=(01 01 01 01)
+    movq        mm3, mm2
+    punpckldq   mm2, mm2                ; mm2=(02 02 02 02)
+    punpckhdq   mm3, mm3                ; mm3=(03 03 03 03)
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_JCOEF), mm1
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_JCOEF), mm1
+    movq        MMWORD MMBLOCK(2,0,edi,SIZEOF_JCOEF), mm2
+    movq        MMWORD MMBLOCK(2,1,edi,SIZEOF_JCOEF), mm2
+    movq        MMWORD MMBLOCK(3,0,edi,SIZEOF_JCOEF), mm3
+    movq        MMWORD MMBLOCK(3,1,edi,SIZEOF_JCOEF), mm3
+    jmp         near .nextcolumn
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    pmullw      mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm1, MMWORD MMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    movq        mm2, MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    pmullw      mm2, MMWORD MMBLOCK(4,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm3, MMWORD MMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; (Original)
+    ; z1 = (z2 + z3) * 0.541196100;
+    ; tmp2 = z1 + z3 * -1.847759065;
+    ; tmp3 = z1 + z2 * 0.765366865;
+    ;
+    ; (This implementation)
+    ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+    ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+    movq        mm4, mm1                ; mm1=in2=z2
+    movq        mm5, mm1
+    punpcklwd   mm4, mm3                ; mm3=in6=z3
+    punpckhwd   mm5, mm3
+    movq        mm1, mm4
+    movq        mm3, mm5
+    pmaddwd     mm4, GOTOFF(ebx,PW_F130_F054)   ; mm4=tmp3L
+    pmaddwd     mm5, GOTOFF(ebx,PW_F130_F054)   ; mm5=tmp3H
+    pmaddwd     mm1, GOTOFF(ebx,PW_F054_MF130)  ; mm1=tmp2L
+    pmaddwd     mm3, GOTOFF(ebx,PW_F054_MF130)  ; mm3=tmp2H
+
+    movq        mm6, mm0
+    paddw       mm0, mm2                ; mm0=in0+in4
+    psubw       mm6, mm2                ; mm6=in0-in4
+
+    pxor        mm7, mm7
+    pxor        mm2, mm2
+    punpcklwd   mm7, mm0                ; mm7=tmp0L
+    punpckhwd   mm2, mm0                ; mm2=tmp0H
+    psrad       mm7, (16-CONST_BITS)    ; psrad mm7,16 & pslld mm7,CONST_BITS
+    psrad       mm2, (16-CONST_BITS)    ; psrad mm2,16 & pslld mm2,CONST_BITS
+
+    movq        mm0, mm7
+    paddd       mm7, mm4                ; mm7=tmp10L
+    psubd       mm0, mm4                ; mm0=tmp13L
+    movq        mm4, mm2
+    paddd       mm2, mm5                ; mm2=tmp10H
+    psubd       mm4, mm5                ; mm4=tmp13H
+
+    movq        MMWORD wk(0), mm7     ; wk(0)=tmp10L
+    movq        MMWORD wk(1), mm2     ; wk(1)=tmp10H
+    movq        MMWORD wk(2), mm0     ; wk(2)=tmp13L
+    movq        MMWORD wk(3), mm4     ; wk(3)=tmp13H
+
+    pxor        mm5, mm5
+    pxor        mm7, mm7
+    punpcklwd   mm5, mm6                ; mm5=tmp1L
+    punpckhwd   mm7, mm6                ; mm7=tmp1H
+    psrad       mm5, (16-CONST_BITS)    ; psrad mm5,16 & pslld mm5,CONST_BITS
+    psrad       mm7, (16-CONST_BITS)    ; psrad mm7,16 & pslld mm7,CONST_BITS
+
+    movq        mm2, mm5
+    paddd       mm5, mm1                ; mm5=tmp11L
+    psubd       mm2, mm1                ; mm2=tmp12L
+    movq        mm0, mm7
+    paddd       mm7, mm3                ; mm7=tmp11H
+    psubd       mm0, mm3                ; mm0=tmp12H
+
+    movq        MMWORD wk(4), mm5     ; wk(4)=tmp11L
+    movq        MMWORD wk(5), mm7     ; wk(5)=tmp11H
+    movq        MMWORD wk(6), mm2     ; wk(6)=tmp12L
+    movq        MMWORD wk(7), mm0     ; wk(7)=tmp12H
+
+    ; -- Odd part
+
+    movq        mm4, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm6, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    pmullw      mm4, MMWORD MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm6, MMWORD MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    movq        mm1, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    pmullw      mm1, MMWORD MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm3, MMWORD MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    movq        mm5, mm6
+    movq        mm7, mm4
+    paddw       mm5, mm3                ; mm5=z3
+    paddw       mm7, mm1                ; mm7=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movq        mm2, mm5
+    movq        mm0, mm5
+    punpcklwd   mm2, mm7
+    punpckhwd   mm0, mm7
+    movq        mm5, mm2
+    movq        mm7, mm0
+    pmaddwd     mm2, GOTOFF(ebx,PW_MF078_F117)  ; mm2=z3L
+    pmaddwd     mm0, GOTOFF(ebx,PW_MF078_F117)  ; mm0=z3H
+    pmaddwd     mm5, GOTOFF(ebx,PW_F117_F078)   ; mm5=z4L
+    pmaddwd     mm7, GOTOFF(ebx,PW_F117_F078)   ; mm7=z4H
+
+    movq        MMWORD wk(10), mm2    ; wk(10)=z3L
+    movq        MMWORD wk(11), mm0    ; wk(11)=z3H
+
+    ; (Original)
+    ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+    ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+    ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+    ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+    ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+    ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+    ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+    ; tmp0 += z3;  tmp1 += z4;
+    ; tmp2 += z3;  tmp3 += z4;
+
+    movq        mm2, mm3
+    movq        mm0, mm3
+    punpcklwd   mm2, mm4
+    punpckhwd   mm0, mm4
+    movq        mm3, mm2
+    movq        mm4, mm0
+    pmaddwd     mm2, GOTOFF(ebx,PW_MF060_MF089)  ; mm2=tmp0L
+    pmaddwd     mm0, GOTOFF(ebx,PW_MF060_MF089)  ; mm0=tmp0H
+    pmaddwd     mm3, GOTOFF(ebx,PW_MF089_F060)   ; mm3=tmp3L
+    pmaddwd     mm4, GOTOFF(ebx,PW_MF089_F060)   ; mm4=tmp3H
+
+    paddd       mm2, MMWORD wk(10)    ; mm2=tmp0L
+    paddd       mm0, MMWORD wk(11)    ; mm0=tmp0H
+    paddd       mm3, mm5                ; mm3=tmp3L
+    paddd       mm4, mm7                ; mm4=tmp3H
+
+    movq        MMWORD wk(8), mm2     ; wk(8)=tmp0L
+    movq        MMWORD wk(9), mm0     ; wk(9)=tmp0H
+
+    movq        mm2, mm1
+    movq        mm0, mm1
+    punpcklwd   mm2, mm6
+    punpckhwd   mm0, mm6
+    movq        mm1, mm2
+    movq        mm6, mm0
+    pmaddwd     mm2, GOTOFF(ebx,PW_MF050_MF256)  ; mm2=tmp1L
+    pmaddwd     mm0, GOTOFF(ebx,PW_MF050_MF256)  ; mm0=tmp1H
+    pmaddwd     mm1, GOTOFF(ebx,PW_MF256_F050)   ; mm1=tmp2L
+    pmaddwd     mm6, GOTOFF(ebx,PW_MF256_F050)   ; mm6=tmp2H
+
+    paddd       mm2, mm5                ; mm2=tmp1L
+    paddd       mm0, mm7                ; mm0=tmp1H
+    paddd       mm1, MMWORD wk(10)    ; mm1=tmp2L
+    paddd       mm6, MMWORD wk(11)    ; mm6=tmp2H
+
+    movq        MMWORD wk(10), mm2    ; wk(10)=tmp1L
+    movq        MMWORD wk(11), mm0    ; wk(11)=tmp1H
+
+    ; -- Final output stage
+
+    movq        mm5, MMWORD wk(0)     ; mm5=tmp10L
+    movq        mm7, MMWORD wk(1)     ; mm7=tmp10H
+
+    movq        mm2, mm5
+    movq        mm0, mm7
+    paddd       mm5, mm3                ; mm5=data0L
+    paddd       mm7, mm4                ; mm7=data0H
+    psubd       mm2, mm3                ; mm2=data7L
+    psubd       mm0, mm4                ; mm0=data7H
+
+    movq        mm3, GOTOFF(ebx,PD_DESCALE_P1)  ; mm3=PD_DESCALE_P1
+
+    paddd       mm5, mm3
+    paddd       mm7, mm3
+    psrad       mm5, DESCALE_P1
+    psrad       mm7, DESCALE_P1
+    paddd       mm2, mm3
+    paddd       mm0, mm3
+    psrad       mm2, DESCALE_P1
+    psrad       mm0, DESCALE_P1
+
+    packssdw    mm5, mm7                ; mm5=data0=(00 01 02 03)
+    packssdw    mm2, mm0                ; mm2=data7=(70 71 72 73)
+
+    movq        mm4, MMWORD wk(4)     ; mm4=tmp11L
+    movq        mm3, MMWORD wk(5)     ; mm3=tmp11H
+
+    movq        mm7, mm4
+    movq        mm0, mm3
+    paddd       mm4, mm1                ; mm4=data1L
+    paddd       mm3, mm6                ; mm3=data1H
+    psubd       mm7, mm1                ; mm7=data6L
+    psubd       mm0, mm6                ; mm0=data6H
+
+    movq        mm1, GOTOFF(ebx,PD_DESCALE_P1)  ; mm1=PD_DESCALE_P1
+
+    paddd       mm4, mm1
+    paddd       mm3, mm1
+    psrad       mm4, DESCALE_P1
+    psrad       mm3, DESCALE_P1
+    paddd       mm7, mm1
+    paddd       mm0, mm1
+    psrad       mm7, DESCALE_P1
+    psrad       mm0, DESCALE_P1
+
+    packssdw    mm4, mm3                ; mm4=data1=(10 11 12 13)
+    packssdw    mm7, mm0                ; mm7=data6=(60 61 62 63)
+
+    movq        mm6, mm5                ; transpose coefficients(phase 1)
+    punpcklwd   mm5, mm4                ; mm5=(00 10 01 11)
+    punpckhwd   mm6, mm4                ; mm6=(02 12 03 13)
+    movq        mm1, mm7                ; transpose coefficients(phase 1)
+    punpcklwd   mm7, mm2                ; mm7=(60 70 61 71)
+    punpckhwd   mm1, mm2                ; mm1=(62 72 63 73)
+
+    movq        mm3, MMWORD wk(6)     ; mm3=tmp12L
+    movq        mm0, MMWORD wk(7)     ; mm0=tmp12H
+    movq        mm4, MMWORD wk(10)    ; mm4=tmp1L
+    movq        mm2, MMWORD wk(11)    ; mm2=tmp1H
+
+    movq        MMWORD wk(0), mm5     ; wk(0)=(00 10 01 11)
+    movq        MMWORD wk(1), mm6     ; wk(1)=(02 12 03 13)
+    movq        MMWORD wk(4), mm7     ; wk(4)=(60 70 61 71)
+    movq        MMWORD wk(5), mm1     ; wk(5)=(62 72 63 73)
+
+    movq        mm5, mm3
+    movq        mm6, mm0
+    paddd       mm3, mm4                ; mm3=data2L
+    paddd       mm0, mm2                ; mm0=data2H
+    psubd       mm5, mm4                ; mm5=data5L
+    psubd       mm6, mm2                ; mm6=data5H
+
+    movq        mm7, GOTOFF(ebx,PD_DESCALE_P1)  ; mm7=PD_DESCALE_P1
+
+    paddd       mm3, mm7
+    paddd       mm0, mm7
+    psrad       mm3, DESCALE_P1
+    psrad       mm0, DESCALE_P1
+    paddd       mm5, mm7
+    paddd       mm6, mm7
+    psrad       mm5, DESCALE_P1
+    psrad       mm6, DESCALE_P1
+
+    packssdw    mm3, mm0                ; mm3=data2=(20 21 22 23)
+    packssdw    mm5, mm6                ; mm5=data5=(50 51 52 53)
+
+    movq        mm1, MMWORD wk(2)     ; mm1=tmp13L
+    movq        mm4, MMWORD wk(3)     ; mm4=tmp13H
+    movq        mm2, MMWORD wk(8)     ; mm2=tmp0L
+    movq        mm7, MMWORD wk(9)     ; mm7=tmp0H
+
+    movq        mm0, mm1
+    movq        mm6, mm4
+    paddd       mm1, mm2                ; mm1=data3L
+    paddd       mm4, mm7                ; mm4=data3H
+    psubd       mm0, mm2                ; mm0=data4L
+    psubd       mm6, mm7                ; mm6=data4H
+
+    movq        mm2, GOTOFF(ebx,PD_DESCALE_P1)  ; mm2=PD_DESCALE_P1
+
+    paddd       mm1, mm2
+    paddd       mm4, mm2
+    psrad       mm1, DESCALE_P1
+    psrad       mm4, DESCALE_P1
+    paddd       mm0, mm2
+    paddd       mm6, mm2
+    psrad       mm0, DESCALE_P1
+    psrad       mm6, DESCALE_P1
+
+    packssdw    mm1, mm4                ; mm1=data3=(30 31 32 33)
+    packssdw    mm0, mm6                ; mm0=data4=(40 41 42 43)
+
+    movq        mm7, MMWORD wk(0)     ; mm7=(00 10 01 11)
+    movq        mm2, MMWORD wk(1)     ; mm2=(02 12 03 13)
+
+    movq        mm4, mm3                ; transpose coefficients(phase 1)
+    punpcklwd   mm3, mm1                ; mm3=(20 30 21 31)
+    punpckhwd   mm4, mm1                ; mm4=(22 32 23 33)
+    movq        mm6, mm0                ; transpose coefficients(phase 1)
+    punpcklwd   mm0, mm5                ; mm0=(40 50 41 51)
+    punpckhwd   mm6, mm5                ; mm6=(42 52 43 53)
+
+    movq        mm1, mm7                ; transpose coefficients(phase 2)
+    punpckldq   mm7, mm3                ; mm7=(00 10 20 30)
+    punpckhdq   mm1, mm3                ; mm1=(01 11 21 31)
+    movq        mm5, mm2                ; transpose coefficients(phase 2)
+    punpckldq   mm2, mm4                ; mm2=(02 12 22 32)
+    punpckhdq   mm5, mm4                ; mm5=(03 13 23 33)
+
+    movq        mm3, MMWORD wk(4)     ; mm3=(60 70 61 71)
+    movq        mm4, MMWORD wk(5)     ; mm4=(62 72 63 73)
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_JCOEF), mm7
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_JCOEF), mm1
+    movq        MMWORD MMBLOCK(2,0,edi,SIZEOF_JCOEF), mm2
+    movq        MMWORD MMBLOCK(3,0,edi,SIZEOF_JCOEF), mm5
+
+    movq        mm7, mm0                ; transpose coefficients(phase 2)
+    punpckldq   mm0, mm3                ; mm0=(40 50 60 70)
+    punpckhdq   mm7, mm3                ; mm7=(41 51 61 71)
+    movq        mm1, mm6                ; transpose coefficients(phase 2)
+    punpckldq   mm6, mm4                ; mm6=(42 52 62 72)
+    punpckhdq   mm1, mm4                ; mm1=(43 53 63 73)
+
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_JCOEF), mm7
+    movq        MMWORD MMBLOCK(2,1,edi,SIZEOF_JCOEF), mm6
+    movq        MMWORD MMBLOCK(3,1,edi,SIZEOF_JCOEF), mm1
+
+.nextcolumn:
+    add         esi, byte 4*SIZEOF_JCOEF            ; coef_block
+    add         edx, byte 4*SIZEOF_ISLOW_MULT_TYPE  ; quantptr
+    add         edi, byte 4*DCTSIZE*SIZEOF_JCOEF    ; wsptr
+    dec         ecx                                 ; ctr
+    jnz         near .columnloop
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         eax, original_ebp
+    lea         esi, workspace                   ; JCOEF *wsptr
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+    mov         ecx, DCTSIZE/4                     ; ctr
+    alignx      16, 7
+.rowloop:
+
+    ; -- Even part
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    movq        mm2, MMWORD MMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+
+    ; (Original)
+    ; z1 = (z2 + z3) * 0.541196100;
+    ; tmp2 = z1 + z3 * -1.847759065;
+    ; tmp3 = z1 + z2 * 0.765366865;
+    ;
+    ; (This implementation)
+    ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+    ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+    movq        mm4, mm1                ; mm1=in2=z2
+    movq        mm5, mm1
+    punpcklwd   mm4, mm3                ; mm3=in6=z3
+    punpckhwd   mm5, mm3
+    movq        mm1, mm4
+    movq        mm3, mm5
+    pmaddwd     mm4, GOTOFF(ebx,PW_F130_F054)   ; mm4=tmp3L
+    pmaddwd     mm5, GOTOFF(ebx,PW_F130_F054)   ; mm5=tmp3H
+    pmaddwd     mm1, GOTOFF(ebx,PW_F054_MF130)  ; mm1=tmp2L
+    pmaddwd     mm3, GOTOFF(ebx,PW_F054_MF130)  ; mm3=tmp2H
+
+    movq        mm6, mm0
+    paddw       mm0, mm2                ; mm0=in0+in4
+    psubw       mm6, mm2                ; mm6=in0-in4
+
+    pxor        mm7, mm7
+    pxor        mm2, mm2
+    punpcklwd   mm7, mm0                ; mm7=tmp0L
+    punpckhwd   mm2, mm0                ; mm2=tmp0H
+    psrad       mm7, (16-CONST_BITS)    ; psrad mm7,16 & pslld mm7,CONST_BITS
+    psrad       mm2, (16-CONST_BITS)    ; psrad mm2,16 & pslld mm2,CONST_BITS
+
+    movq        mm0, mm7
+    paddd       mm7, mm4                ; mm7=tmp10L
+    psubd       mm0, mm4                ; mm0=tmp13L
+    movq        mm4, mm2
+    paddd       mm2, mm5                ; mm2=tmp10H
+    psubd       mm4, mm5                ; mm4=tmp13H
+
+    movq        MMWORD wk(0), mm7     ; wk(0)=tmp10L
+    movq        MMWORD wk(1), mm2     ; wk(1)=tmp10H
+    movq        MMWORD wk(2), mm0     ; wk(2)=tmp13L
+    movq        MMWORD wk(3), mm4     ; wk(3)=tmp13H
+
+    pxor        mm5, mm5
+    pxor        mm7, mm7
+    punpcklwd   mm5, mm6                ; mm5=tmp1L
+    punpckhwd   mm7, mm6                ; mm7=tmp1H
+    psrad       mm5, (16-CONST_BITS)    ; psrad mm5,16 & pslld mm5,CONST_BITS
+    psrad       mm7, (16-CONST_BITS)    ; psrad mm7,16 & pslld mm7,CONST_BITS
+
+    movq        mm2, mm5
+    paddd       mm5, mm1                ; mm5=tmp11L
+    psubd       mm2, mm1                ; mm2=tmp12L
+    movq        mm0, mm7
+    paddd       mm7, mm3                ; mm7=tmp11H
+    psubd       mm0, mm3                ; mm0=tmp12H
+
+    movq        MMWORD wk(4), mm5     ; wk(4)=tmp11L
+    movq        MMWORD wk(5), mm7     ; wk(5)=tmp11H
+    movq        MMWORD wk(6), mm2     ; wk(6)=tmp12L
+    movq        MMWORD wk(7), mm0     ; wk(7)=tmp12H
+
+    ; -- Odd part
+
+    movq        mm4, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm6, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+
+    movq        mm5, mm6
+    movq        mm7, mm4
+    paddw       mm5, mm3                ; mm5=z3
+    paddw       mm7, mm1                ; mm7=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movq        mm2, mm5
+    movq        mm0, mm5
+    punpcklwd   mm2, mm7
+    punpckhwd   mm0, mm7
+    movq        mm5, mm2
+    movq        mm7, mm0
+    pmaddwd     mm2, GOTOFF(ebx,PW_MF078_F117)  ; mm2=z3L
+    pmaddwd     mm0, GOTOFF(ebx,PW_MF078_F117)  ; mm0=z3H
+    pmaddwd     mm5, GOTOFF(ebx,PW_F117_F078)   ; mm5=z4L
+    pmaddwd     mm7, GOTOFF(ebx,PW_F117_F078)   ; mm7=z4H
+
+    movq        MMWORD wk(10), mm2    ; wk(10)=z3L
+    movq        MMWORD wk(11), mm0    ; wk(11)=z3H
+
+    ; (Original)
+    ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+    ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+    ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+    ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+    ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+    ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+    ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+    ; tmp0 += z3;  tmp1 += z4;
+    ; tmp2 += z3;  tmp3 += z4;
+
+    movq        mm2, mm3
+    movq        mm0, mm3
+    punpcklwd   mm2, mm4
+    punpckhwd   mm0, mm4
+    movq        mm3, mm2
+    movq        mm4, mm0
+    pmaddwd     mm2, GOTOFF(ebx,PW_MF060_MF089)  ; mm2=tmp0L
+    pmaddwd     mm0, GOTOFF(ebx,PW_MF060_MF089)  ; mm0=tmp0H
+    pmaddwd     mm3, GOTOFF(ebx,PW_MF089_F060)   ; mm3=tmp3L
+    pmaddwd     mm4, GOTOFF(ebx,PW_MF089_F060)   ; mm4=tmp3H
+
+    paddd       mm2, MMWORD wk(10)    ; mm2=tmp0L
+    paddd       mm0, MMWORD wk(11)    ; mm0=tmp0H
+    paddd       mm3, mm5                ; mm3=tmp3L
+    paddd       mm4, mm7                ; mm4=tmp3H
+
+    movq        MMWORD wk(8), mm2     ; wk(8)=tmp0L
+    movq        MMWORD wk(9), mm0     ; wk(9)=tmp0H
+
+    movq        mm2, mm1
+    movq        mm0, mm1
+    punpcklwd   mm2, mm6
+    punpckhwd   mm0, mm6
+    movq        mm1, mm2
+    movq        mm6, mm0
+    pmaddwd     mm2, GOTOFF(ebx,PW_MF050_MF256)  ; mm2=tmp1L
+    pmaddwd     mm0, GOTOFF(ebx,PW_MF050_MF256)  ; mm0=tmp1H
+    pmaddwd     mm1, GOTOFF(ebx,PW_MF256_F050)   ; mm1=tmp2L
+    pmaddwd     mm6, GOTOFF(ebx,PW_MF256_F050)   ; mm6=tmp2H
+
+    paddd       mm2, mm5                ; mm2=tmp1L
+    paddd       mm0, mm7                ; mm0=tmp1H
+    paddd       mm1, MMWORD wk(10)    ; mm1=tmp2L
+    paddd       mm6, MMWORD wk(11)    ; mm6=tmp2H
+
+    movq        MMWORD wk(10), mm2    ; wk(10)=tmp1L
+    movq        MMWORD wk(11), mm0    ; wk(11)=tmp1H
+
+    ; -- Final output stage
+
+    movq        mm5, MMWORD wk(0)     ; mm5=tmp10L
+    movq        mm7, MMWORD wk(1)     ; mm7=tmp10H
+
+    movq        mm2, mm5
+    movq        mm0, mm7
+    paddd       mm5, mm3                ; mm5=data0L
+    paddd       mm7, mm4                ; mm7=data0H
+    psubd       mm2, mm3                ; mm2=data7L
+    psubd       mm0, mm4                ; mm0=data7H
+
+    movq        mm3, GOTOFF(ebx,PD_DESCALE_P2)  ; mm3=PD_DESCALE_P2
+
+    paddd       mm5, mm3
+    paddd       mm7, mm3
+    psrad       mm5, DESCALE_P2
+    psrad       mm7, DESCALE_P2
+    paddd       mm2, mm3
+    paddd       mm0, mm3
+    psrad       mm2, DESCALE_P2
+    psrad       mm0, DESCALE_P2
+
+    packssdw    mm5, mm7                ; mm5=data0=(00 10 20 30)
+    packssdw    mm2, mm0                ; mm2=data7=(07 17 27 37)
+
+    movq        mm4, MMWORD wk(4)     ; mm4=tmp11L
+    movq        mm3, MMWORD wk(5)     ; mm3=tmp11H
+
+    movq        mm7, mm4
+    movq        mm0, mm3
+    paddd       mm4, mm1                ; mm4=data1L
+    paddd       mm3, mm6                ; mm3=data1H
+    psubd       mm7, mm1                ; mm7=data6L
+    psubd       mm0, mm6                ; mm0=data6H
+
+    movq        mm1, GOTOFF(ebx,PD_DESCALE_P2)  ; mm1=PD_DESCALE_P2
+
+    paddd       mm4, mm1
+    paddd       mm3, mm1
+    psrad       mm4, DESCALE_P2
+    psrad       mm3, DESCALE_P2
+    paddd       mm7, mm1
+    paddd       mm0, mm1
+    psrad       mm7, DESCALE_P2
+    psrad       mm0, DESCALE_P2
+
+    packssdw    mm4, mm3                ; mm4=data1=(01 11 21 31)
+    packssdw    mm7, mm0                ; mm7=data6=(06 16 26 36)
+
+    packsswb    mm5, mm7                ; mm5=(00 10 20 30 06 16 26 36)
+    packsswb    mm4, mm2                ; mm4=(01 11 21 31 07 17 27 37)
+
+    movq        mm6, MMWORD wk(6)     ; mm6=tmp12L
+    movq        mm1, MMWORD wk(7)     ; mm1=tmp12H
+    movq        mm3, MMWORD wk(10)    ; mm3=tmp1L
+    movq        mm0, MMWORD wk(11)    ; mm0=tmp1H
+
+    movq        MMWORD wk(0), mm5     ; wk(0)=(00 10 20 30 06 16 26 36)
+    movq        MMWORD wk(1), mm4     ; wk(1)=(01 11 21 31 07 17 27 37)
+
+    movq        mm7, mm6
+    movq        mm2, mm1
+    paddd       mm6, mm3                ; mm6=data2L
+    paddd       mm1, mm0                ; mm1=data2H
+    psubd       mm7, mm3                ; mm7=data5L
+    psubd       mm2, mm0                ; mm2=data5H
+
+    movq        mm5, GOTOFF(ebx,PD_DESCALE_P2)  ; mm5=PD_DESCALE_P2
+
+    paddd       mm6, mm5
+    paddd       mm1, mm5
+    psrad       mm6, DESCALE_P2
+    psrad       mm1, DESCALE_P2
+    paddd       mm7, mm5
+    paddd       mm2, mm5
+    psrad       mm7, DESCALE_P2
+    psrad       mm2, DESCALE_P2
+
+    packssdw    mm6, mm1                ; mm6=data2=(02 12 22 32)
+    packssdw    mm7, mm2                ; mm7=data5=(05 15 25 35)
+
+    movq        mm4, MMWORD wk(2)     ; mm4=tmp13L
+    movq        mm3, MMWORD wk(3)     ; mm3=tmp13H
+    movq        mm0, MMWORD wk(8)     ; mm0=tmp0L
+    movq        mm5, MMWORD wk(9)     ; mm5=tmp0H
+
+    movq        mm1, mm4
+    movq        mm2, mm3
+    paddd       mm4, mm0                ; mm4=data3L
+    paddd       mm3, mm5                ; mm3=data3H
+    psubd       mm1, mm0                ; mm1=data4L
+    psubd       mm2, mm5                ; mm2=data4H
+
+    movq        mm0, GOTOFF(ebx,PD_DESCALE_P2)  ; mm0=PD_DESCALE_P2
+
+    paddd       mm4, mm0
+    paddd       mm3, mm0
+    psrad       mm4, DESCALE_P2
+    psrad       mm3, DESCALE_P2
+    paddd       mm1, mm0
+    paddd       mm2, mm0
+    psrad       mm1, DESCALE_P2
+    psrad       mm2, DESCALE_P2
+
+    movq        mm5, GOTOFF(ebx,PB_CENTERJSAMP)  ; mm5=PB_CENTERJSAMP
+
+    packssdw    mm4, mm3                ; mm4=data3=(03 13 23 33)
+    packssdw    mm1, mm2                ; mm1=data4=(04 14 24 34)
+
+    movq        mm0, MMWORD wk(0)     ; mm0=(00 10 20 30 06 16 26 36)
+    movq        mm3, MMWORD wk(1)     ; mm3=(01 11 21 31 07 17 27 37)
+
+    packsswb    mm6, mm1                ; mm6=(02 12 22 32 04 14 24 34)
+    packsswb    mm4, mm7                ; mm4=(03 13 23 33 05 15 25 35)
+
+    paddb       mm0, mm5
+    paddb       mm3, mm5
+    paddb       mm6, mm5
+    paddb       mm4, mm5
+
+    movq        mm2, mm0                ; transpose coefficients(phase 1)
+    punpcklbw   mm0, mm3                ; mm0=(00 01 10 11 20 21 30 31)
+    punpckhbw   mm2, mm3                ; mm2=(06 07 16 17 26 27 36 37)
+    movq        mm1, mm6                ; transpose coefficients(phase 1)
+    punpcklbw   mm6, mm4                ; mm6=(02 03 12 13 22 23 32 33)
+    punpckhbw   mm1, mm4                ; mm1=(04 05 14 15 24 25 34 35)
+
+    movq        mm7, mm0                ; transpose coefficients(phase 2)
+    punpcklwd   mm0, mm6                ; mm0=(00 01 02 03 10 11 12 13)
+    punpckhwd   mm7, mm6                ; mm7=(20 21 22 23 30 31 32 33)
+    movq        mm5, mm1                ; transpose coefficients(phase 2)
+    punpcklwd   mm1, mm2                ; mm1=(04 05 06 07 14 15 16 17)
+    punpckhwd   mm5, mm2                ; mm5=(24 25 26 27 34 35 36 37)
+
+    movq        mm3, mm0                ; transpose coefficients(phase 3)
+    punpckldq   mm0, mm1                ; mm0=(00 01 02 03 04 05 06 07)
+    punpckhdq   mm3, mm1                ; mm3=(10 11 12 13 14 15 16 17)
+    movq        mm4, mm7                ; transpose coefficients(phase 3)
+    punpckldq   mm7, mm5                ; mm7=(20 21 22 23 24 25 26 27)
+    punpckhdq   mm4, mm5                ; mm4=(30 31 32 33 34 35 36 37)
+
+    pushpic     ebx                     ; save GOT address
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         ebx, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    movq        MMWORD edx+eax*SIZEOF_JSAMPLE, mm0
+    movq        MMWORD ebx+eax*SIZEOF_JSAMPLE, mm3
+    mov         edx, JSAMPROW edi+2*SIZEOF_JSAMPROW
+    mov         ebx, JSAMPROW edi+3*SIZEOF_JSAMPROW
+    movq        MMWORD edx+eax*SIZEOF_JSAMPLE, mm7
+    movq        MMWORD ebx+eax*SIZEOF_JSAMPLE, mm4
+
+    poppic      ebx                     ; restore GOT address
+
+    add         esi, byte 4*SIZEOF_JCOEF     ; wsptr
+    add         edi, byte 4*SIZEOF_JSAMPROW
+    dec         ecx                          ; ctr
+    jnz         near .rowloop
+
+    emms                                ; empty MMX state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctint-sse2.asm Added

@@ -0,0 +1,858 @@
+;
+; jidctint.asm - accurate integer IDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, 2020, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS + 3)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_islow_sse2)
+
+EXTN(jconst_idct_islow_sse2):
+
+PW_F130_F054   times 4  dw  (F_0_541 + F_0_765),  F_0_541
+PW_F054_MF130  times 4  dw  F_0_541, (F_0_541 - F_1_847)
+PW_MF078_F117  times 4  dw  (F_1_175 - F_1_961),  F_1_175
+PW_F117_F078   times 4  dw  F_1_175, (F_1_175 - F_0_390)
+PW_MF060_MF089 times 4  dw  (F_0_298 - F_0_899), -F_0_899
+PW_MF089_F060  times 4  dw -F_0_899, (F_1_501 - F_0_899)
+PW_MF050_MF256 times 4  dw  (F_2_053 - F_2_562), -F_2_562
+PW_MF256_F050  times 4  dw -F_2_562, (F_3_072 - F_2_562)
+PD_DESCALE_P1  times 4  dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2  times 4  dd  1 << (DESCALE_P2 - 1)
+PB_CENTERJSAMP times 16 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_sse2(void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; jpeg_component_info *compptr
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM         12
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_islow_sse2)
+
+EXTN(jsimd_idct_islow_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_SSE2
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         near .columnDCT
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    por         xmm1, xmm0
+    packsswb    xmm1, xmm1
+    packsswb    xmm1, xmm1
+    movd        eax, xmm1
+    test        eax, eax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movdqa      xmm5, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm5, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    psllw       xmm5, PASS1_BITS
+
+    movdqa      xmm4, xmm5              ; xmm5=in0=(00 01 02 03 04 05 06 07)
+    punpcklwd   xmm5, xmm5              ; xmm5=(00 00 01 01 02 02 03 03)
+    punpckhwd   xmm4, xmm4              ; xmm4=(04 04 05 05 06 06 07 07)
+
+    pshufd      xmm7, xmm5, 0x00        ; xmm7=col0=(00 00 00 00 00 00 00 00)
+    pshufd      xmm6, xmm5, 0x55        ; xmm6=col1=(01 01 01 01 01 01 01 01)
+    pshufd      xmm1, xmm5, 0xAA        ; xmm1=col2=(02 02 02 02 02 02 02 02)
+    pshufd      xmm5, xmm5, 0xFF        ; xmm5=col3=(03 03 03 03 03 03 03 03)
+    pshufd      xmm0, xmm4, 0x00        ; xmm0=col4=(04 04 04 04 04 04 04 04)
+    pshufd      xmm3, xmm4, 0x55        ; xmm3=col5=(05 05 05 05 05 05 05 05)
+    pshufd      xmm2, xmm4, 0xAA        ; xmm2=col6=(06 06 06 06 06 06 06 06)
+    pshufd      xmm4, xmm4, 0xFF        ; xmm4=col7=(07 07 07 07 07 07 07 07)
+
+    movdqa      XMMWORD wk(8), xmm6   ; wk(8)=col1
+    movdqa      XMMWORD wk(9), xmm5   ; wk(9)=col3
+    movdqa      XMMWORD wk(10), xmm3  ; wk(10)=col5
+    movdqa      XMMWORD wk(11), xmm4  ; wk(11)=col7
+    jmp         near .column_end
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    movdqa      xmm2, XMMWORD XMMBLOCK(4,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(4,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; (Original)
+    ; z1 = (z2 + z3) * 0.541196100;
+    ; tmp2 = z1 + z3 * -1.847759065;
+    ; tmp3 = z1 + z2 * 0.765366865;
+    ;
+    ; (This implementation)
+    ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+    ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+    movdqa      xmm4, xmm1              ; xmm1=in2=z2
+    movdqa      xmm5, xmm1
+    punpcklwd   xmm4, xmm3              ; xmm3=in6=z3
+    punpckhwd   xmm5, xmm3
+    movdqa      xmm1, xmm4
+    movdqa      xmm3, xmm5
+    pmaddwd     xmm4, GOTOFF(ebx,PW_F130_F054)   ; xmm4=tmp3L
+    pmaddwd     xmm5, GOTOFF(ebx,PW_F130_F054)   ; xmm5=tmp3H
+    pmaddwd     xmm1, GOTOFF(ebx,PW_F054_MF130)  ; xmm1=tmp2L
+    pmaddwd     xmm3, GOTOFF(ebx,PW_F054_MF130)  ; xmm3=tmp2H
+
+    movdqa      xmm6, xmm0
+    paddw       xmm0, xmm2              ; xmm0=in0+in4
+    psubw       xmm6, xmm2              ; xmm6=in0-in4
+
+    pxor        xmm7, xmm7
+    pxor        xmm2, xmm2
+    punpcklwd   xmm7, xmm0              ; xmm7=tmp0L
+    punpckhwd   xmm2, xmm0              ; xmm2=tmp0H
+    psrad       xmm7, (16-CONST_BITS)   ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+    psrad       xmm2, (16-CONST_BITS)   ; psrad xmm2,16 & pslld xmm2,CONST_BITS
+
+    movdqa      xmm0, xmm7
+    paddd       xmm7, xmm4              ; xmm7=tmp10L
+    psubd       xmm0, xmm4              ; xmm0=tmp13L
+    movdqa      xmm4, xmm2
+    paddd       xmm2, xmm5              ; xmm2=tmp10H
+    psubd       xmm4, xmm5              ; xmm4=tmp13H
+
+    movdqa      XMMWORD wk(0), xmm7   ; wk(0)=tmp10L
+    movdqa      XMMWORD wk(1), xmm2   ; wk(1)=tmp10H
+    movdqa      XMMWORD wk(2), xmm0   ; wk(2)=tmp13L
+    movdqa      XMMWORD wk(3), xmm4   ; wk(3)=tmp13H
+
+    pxor        xmm5, xmm5
+    pxor        xmm7, xmm7
+    punpcklwd   xmm5, xmm6              ; xmm5=tmp1L
+    punpckhwd   xmm7, xmm6              ; xmm7=tmp1H
+    psrad       xmm5, (16-CONST_BITS)   ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+    psrad       xmm7, (16-CONST_BITS)   ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+
+    movdqa      xmm2, xmm5
+    paddd       xmm5, xmm1              ; xmm5=tmp11L
+    psubd       xmm2, xmm1              ; xmm2=tmp12L
+    movdqa      xmm0, xmm7
+    paddd       xmm7, xmm3              ; xmm7=tmp11H
+    psubd       xmm0, xmm3              ; xmm0=tmp12H
+
+    movdqa      XMMWORD wk(4), xmm5   ; wk(4)=tmp11L
+    movdqa      XMMWORD wk(5), xmm7   ; wk(5)=tmp11H
+    movdqa      XMMWORD wk(6), xmm2   ; wk(6)=tmp12L
+    movdqa      XMMWORD wk(7), xmm0   ; wk(7)=tmp12H
+
+    ; -- Odd part
+
+    movdqa      xmm4, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm6, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm4, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm6, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    movdqa      xmm1, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm1, XMMWORD XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    movdqa      xmm5, xmm6
+    movdqa      xmm7, xmm4
+    paddw       xmm5, xmm3              ; xmm5=z3
+    paddw       xmm7, xmm1              ; xmm7=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movdqa      xmm2, xmm5
+    movdqa      xmm0, xmm5
+    punpcklwd   xmm2, xmm7
+    punpckhwd   xmm0, xmm7
+    movdqa      xmm5, xmm2
+    movdqa      xmm7, xmm0
+    pmaddwd     xmm2, GOTOFF(ebx,PW_MF078_F117)  ; xmm2=z3L
+    pmaddwd     xmm0, GOTOFF(ebx,PW_MF078_F117)  ; xmm0=z3H
+    pmaddwd     xmm5, GOTOFF(ebx,PW_F117_F078)   ; xmm5=z4L
+    pmaddwd     xmm7, GOTOFF(ebx,PW_F117_F078)   ; xmm7=z4H
+
+    movdqa      XMMWORD wk(10), xmm2  ; wk(10)=z3L
+    movdqa      XMMWORD wk(11), xmm0  ; wk(11)=z3H
+
+    ; (Original)
+    ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+    ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+    ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+    ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+    ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+    ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+    ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+    ; tmp0 += z3;  tmp1 += z4;
+    ; tmp2 += z3;  tmp3 += z4;
+
+    movdqa      xmm2, xmm3
+    movdqa      xmm0, xmm3
+    punpcklwd   xmm2, xmm4
+    punpckhwd   xmm0, xmm4
+    movdqa      xmm3, xmm2
+    movdqa      xmm4, xmm0
+    pmaddwd     xmm2, GOTOFF(ebx,PW_MF060_MF089)  ; xmm2=tmp0L
+    pmaddwd     xmm0, GOTOFF(ebx,PW_MF060_MF089)  ; xmm0=tmp0H
+    pmaddwd     xmm3, GOTOFF(ebx,PW_MF089_F060)   ; xmm3=tmp3L
+    pmaddwd     xmm4, GOTOFF(ebx,PW_MF089_F060)   ; xmm4=tmp3H
+
+    paddd       xmm2, XMMWORD wk(10)  ; xmm2=tmp0L
+    paddd       xmm0, XMMWORD wk(11)  ; xmm0=tmp0H
+    paddd       xmm3, xmm5              ; xmm3=tmp3L
+    paddd       xmm4, xmm7              ; xmm4=tmp3H
+
+    movdqa      XMMWORD wk(8), xmm2   ; wk(8)=tmp0L
+    movdqa      XMMWORD wk(9), xmm0   ; wk(9)=tmp0H
+
+    movdqa      xmm2, xmm1
+    movdqa      xmm0, xmm1
+    punpcklwd   xmm2, xmm6
+    punpckhwd   xmm0, xmm6
+    movdqa      xmm1, xmm2
+    movdqa      xmm6, xmm0
+    pmaddwd     xmm2, GOTOFF(ebx,PW_MF050_MF256)  ; xmm2=tmp1L
+    pmaddwd     xmm0, GOTOFF(ebx,PW_MF050_MF256)  ; xmm0=tmp1H
+    pmaddwd     xmm1, GOTOFF(ebx,PW_MF256_F050)   ; xmm1=tmp2L
+    pmaddwd     xmm6, GOTOFF(ebx,PW_MF256_F050)   ; xmm6=tmp2H
+
+    paddd       xmm2, xmm5              ; xmm2=tmp1L
+    paddd       xmm0, xmm7              ; xmm0=tmp1H
+    paddd       xmm1, XMMWORD wk(10)  ; xmm1=tmp2L
+    paddd       xmm6, XMMWORD wk(11)  ; xmm6=tmp2H
+
+    movdqa      XMMWORD wk(10), xmm2  ; wk(10)=tmp1L
+    movdqa      XMMWORD wk(11), xmm0  ; wk(11)=tmp1H
+
+    ; -- Final output stage
+
+    movdqa      xmm5, XMMWORD wk(0)   ; xmm5=tmp10L
+    movdqa      xmm7, XMMWORD wk(1)   ; xmm7=tmp10H
+
+    movdqa      xmm2, xmm5
+    movdqa      xmm0, xmm7
+    paddd       xmm5, xmm3              ; xmm5=data0L
+    paddd       xmm7, xmm4              ; xmm7=data0H
+    psubd       xmm2, xmm3              ; xmm2=data7L
+    psubd       xmm0, xmm4              ; xmm0=data7H
+
+    movdqa      xmm3, GOTOFF(ebx,PD_DESCALE_P1)  ; xmm3=PD_DESCALE_P1
+
+    paddd       xmm5, xmm3
+    paddd       xmm7, xmm3
+    psrad       xmm5, DESCALE_P1
+    psrad       xmm7, DESCALE_P1
+    paddd       xmm2, xmm3
+    paddd       xmm0, xmm3
+    psrad       xmm2, DESCALE_P1
+    psrad       xmm0, DESCALE_P1
+
+    packssdw    xmm5, xmm7              ; xmm5=data0=(00 01 02 03 04 05 06 07)
+    packssdw    xmm2, xmm0              ; xmm2=data7=(70 71 72 73 74 75 76 77)
+
+    movdqa      xmm4, XMMWORD wk(4)   ; xmm4=tmp11L
+    movdqa      xmm3, XMMWORD wk(5)   ; xmm3=tmp11H
+
+    movdqa      xmm7, xmm4
+    movdqa      xmm0, xmm3
+    paddd       xmm4, xmm1              ; xmm4=data1L
+    paddd       xmm3, xmm6              ; xmm3=data1H
+    psubd       xmm7, xmm1              ; xmm7=data6L
+    psubd       xmm0, xmm6              ; xmm0=data6H
+
+    movdqa      xmm1, GOTOFF(ebx,PD_DESCALE_P1)  ; xmm1=PD_DESCALE_P1
+
+    paddd       xmm4, xmm1
+    paddd       xmm3, xmm1
+    psrad       xmm4, DESCALE_P1
+    psrad       xmm3, DESCALE_P1
+    paddd       xmm7, xmm1
+    paddd       xmm0, xmm1
+    psrad       xmm7, DESCALE_P1
+    psrad       xmm0, DESCALE_P1
+
+    packssdw    xmm4, xmm3              ; xmm4=data1=(10 11 12 13 14 15 16 17)
+    packssdw    xmm7, xmm0              ; xmm7=data6=(60 61 62 63 64 65 66 67)
+
+    movdqa      xmm6, xmm5              ; transpose coefficients(phase 1)
+    punpcklwd   xmm5, xmm4              ; xmm5=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm6, xmm4              ; xmm6=(04 14 05 15 06 16 07 17)
+    movdqa      xmm1, xmm7              ; transpose coefficients(phase 1)
+    punpcklwd   xmm7, xmm2              ; xmm7=(60 70 61 71 62 72 63 73)
+    punpckhwd   xmm1, xmm2              ; xmm1=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm3, XMMWORD wk(6)   ; xmm3=tmp12L
+    movdqa      xmm0, XMMWORD wk(7)   ; xmm0=tmp12H
+    movdqa      xmm4, XMMWORD wk(10)  ; xmm4=tmp1L
+    movdqa      xmm2, XMMWORD wk(11)  ; xmm2=tmp1H
+
+    movdqa      XMMWORD wk(0), xmm5   ; wk(0)=(00 10 01 11 02 12 03 13)
+    movdqa      XMMWORD wk(1), xmm6   ; wk(1)=(04 14 05 15 06 16 07 17)
+    movdqa      XMMWORD wk(4), xmm7   ; wk(4)=(60 70 61 71 62 72 63 73)
+    movdqa      XMMWORD wk(5), xmm1   ; wk(5)=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm5, xmm3
+    movdqa      xmm6, xmm0
+    paddd       xmm3, xmm4              ; xmm3=data2L
+    paddd       xmm0, xmm2              ; xmm0=data2H
+    psubd       xmm5, xmm4              ; xmm5=data5L
+    psubd       xmm6, xmm2              ; xmm6=data5H
+
+    movdqa      xmm7, GOTOFF(ebx,PD_DESCALE_P1)  ; xmm7=PD_DESCALE_P1
+
+    paddd       xmm3, xmm7
+    paddd       xmm0, xmm7
+    psrad       xmm3, DESCALE_P1
+    psrad       xmm0, DESCALE_P1
+    paddd       xmm5, xmm7
+    paddd       xmm6, xmm7
+    psrad       xmm5, DESCALE_P1
+    psrad       xmm6, DESCALE_P1
+
+    packssdw    xmm3, xmm0              ; xmm3=data2=(20 21 22 23 24 25 26 27)
+    packssdw    xmm5, xmm6              ; xmm5=data5=(50 51 52 53 54 55 56 57)
+
+    movdqa      xmm1, XMMWORD wk(2)   ; xmm1=tmp13L
+    movdqa      xmm4, XMMWORD wk(3)   ; xmm4=tmp13H
+    movdqa      xmm2, XMMWORD wk(8)   ; xmm2=tmp0L
+    movdqa      xmm7, XMMWORD wk(9)   ; xmm7=tmp0H
+
+    movdqa      xmm0, xmm1
+    movdqa      xmm6, xmm4
+    paddd       xmm1, xmm2              ; xmm1=data3L
+    paddd       xmm4, xmm7              ; xmm4=data3H
+    psubd       xmm0, xmm2              ; xmm0=data4L
+    psubd       xmm6, xmm7              ; xmm6=data4H
+
+    movdqa      xmm2, GOTOFF(ebx,PD_DESCALE_P1)  ; xmm2=PD_DESCALE_P1
+
+    paddd       xmm1, xmm2
+    paddd       xmm4, xmm2
+    psrad       xmm1, DESCALE_P1
+    psrad       xmm4, DESCALE_P1
+    paddd       xmm0, xmm2
+    paddd       xmm6, xmm2
+    psrad       xmm0, DESCALE_P1
+    psrad       xmm6, DESCALE_P1
+
+    packssdw    xmm1, xmm4              ; xmm1=data3=(30 31 32 33 34 35 36 37)
+    packssdw    xmm0, xmm6              ; xmm0=data4=(40 41 42 43 44 45 46 47)
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=(00 10 01 11 02 12 03 13)
+    movdqa      xmm2, XMMWORD wk(1)   ; xmm2=(04 14 05 15 06 16 07 17)
+
+    movdqa      xmm4, xmm3              ; transpose coefficients(phase 1)
+    punpcklwd   xmm3, xmm1              ; xmm3=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm4, xmm1              ; xmm4=(24 34 25 35 26 36 27 37)
+    movdqa      xmm6, xmm0              ; transpose coefficients(phase 1)
+    punpcklwd   xmm0, xmm5              ; xmm0=(40 50 41 51 42 52 43 53)
+    punpckhwd   xmm6, xmm5              ; xmm6=(44 54 45 55 46 56 47 57)
+
+    movdqa      xmm1, xmm7              ; transpose coefficients(phase 2)
+    punpckldq   xmm7, xmm3              ; xmm7=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm1, xmm3              ; xmm1=(02 12 22 32 03 13 23 33)
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm4              ; xmm2=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm5, xmm4              ; xmm5=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm3, XMMWORD wk(4)   ; xmm3=(60 70 61 71 62 72 63 73)
+    movdqa      xmm4, XMMWORD wk(5)   ; xmm4=(64 74 65 75 66 76 67 77)
+
+    movdqa      XMMWORD wk(6), xmm2   ; wk(6)=(04 14 24 34 05 15 25 35)
+    movdqa      XMMWORD wk(7), xmm5   ; wk(7)=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm2, xmm0              ; transpose coefficients(phase 2)
+    punpckldq   xmm0, xmm3              ; xmm0=(40 50 60 70 41 51 61 71)
+    punpckhdq   xmm2, xmm3              ; xmm2=(42 52 62 72 43 53 63 73)
+    movdqa      xmm5, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm4              ; xmm6=(44 54 64 74 45 55 65 75)
+    punpckhdq   xmm5, xmm4              ; xmm5=(46 56 66 76 47 57 67 77)
+
+    movdqa      xmm3, xmm7              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm7, xmm0              ; xmm7=col0=(00 10 20 30 40 50 60 70)
+    punpckhqdq  xmm3, xmm0              ; xmm3=col1=(01 11 21 31 41 51 61 71)
+    movdqa      xmm4, xmm1              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm1, xmm2              ; xmm1=col2=(02 12 22 32 42 52 62 72)
+    punpckhqdq  xmm4, xmm2              ; xmm4=col3=(03 13 23 33 43 53 63 73)
+
+    movdqa      xmm0, XMMWORD wk(6)   ; xmm0=(04 14 24 34 05 15 25 35)
+    movdqa      xmm2, XMMWORD wk(7)   ; xmm2=(06 16 26 36 07 17 27 37)
+
+    movdqa      XMMWORD wk(8), xmm3   ; wk(8)=col1
+    movdqa      XMMWORD wk(9), xmm4   ; wk(9)=col3
+
+    movdqa      xmm3, xmm0              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm0, xmm6              ; xmm0=col4=(04 14 24 34 44 54 64 74)
+    punpckhqdq  xmm3, xmm6              ; xmm3=col5=(05 15 25 35 45 55 65 75)
+    movdqa      xmm4, xmm2              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm2, xmm5              ; xmm2=col6=(06 16 26 36 46 56 66 76)
+    punpckhqdq  xmm4, xmm5              ; xmm4=col7=(07 17 27 37 47 57 67 77)
+
+    movdqa      XMMWORD wk(10), xmm3  ; wk(10)=col5
+    movdqa      XMMWORD wk(11), xmm4  ; wk(11)=col7
+.column_end:
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         eax, original_ebp
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+
+    ; -- Even part
+
+    ; xmm7=col0, xmm1=col2, xmm0=col4, xmm2=col6
+
+    ; (Original)
+    ; z1 = (z2 + z3) * 0.541196100;
+    ; tmp2 = z1 + z3 * -1.847759065;
+    ; tmp3 = z1 + z2 * 0.765366865;
+    ;
+    ; (This implementation)
+    ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+    ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+    movdqa      xmm6, xmm1              ; xmm1=in2=z2
+    movdqa      xmm5, xmm1
+    punpcklwd   xmm6, xmm2              ; xmm2=in6=z3
+    punpckhwd   xmm5, xmm2
+    movdqa      xmm1, xmm6
+    movdqa      xmm2, xmm5
+    pmaddwd     xmm6, GOTOFF(ebx,PW_F130_F054)   ; xmm6=tmp3L
+    pmaddwd     xmm5, GOTOFF(ebx,PW_F130_F054)   ; xmm5=tmp3H
+    pmaddwd     xmm1, GOTOFF(ebx,PW_F054_MF130)  ; xmm1=tmp2L
+    pmaddwd     xmm2, GOTOFF(ebx,PW_F054_MF130)  ; xmm2=tmp2H
+
+    movdqa      xmm3, xmm7
+    paddw       xmm7, xmm0              ; xmm7=in0+in4
+    psubw       xmm3, xmm0              ; xmm3=in0-in4
+
+    pxor        xmm4, xmm4
+    pxor        xmm0, xmm0
+    punpcklwd   xmm4, xmm7              ; xmm4=tmp0L
+    punpckhwd   xmm0, xmm7              ; xmm0=tmp0H
+    psrad       xmm4, (16-CONST_BITS)   ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+    psrad       xmm0, (16-CONST_BITS)   ; psrad xmm0,16 & pslld xmm0,CONST_BITS
+
+    movdqa      xmm7, xmm4
+    paddd       xmm4, xmm6              ; xmm4=tmp10L
+    psubd       xmm7, xmm6              ; xmm7=tmp13L
+    movdqa      xmm6, xmm0
+    paddd       xmm0, xmm5              ; xmm0=tmp10H
+    psubd       xmm6, xmm5              ; xmm6=tmp13H
+
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=tmp10L
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=tmp10H
+    movdqa      XMMWORD wk(2), xmm7   ; wk(2)=tmp13L
+    movdqa      XMMWORD wk(3), xmm6   ; wk(3)=tmp13H
+
+    pxor        xmm5, xmm5
+    pxor        xmm4, xmm4
+    punpcklwd   xmm5, xmm3              ; xmm5=tmp1L
+    punpckhwd   xmm4, xmm3              ; xmm4=tmp1H
+    psrad       xmm5, (16-CONST_BITS)   ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+    psrad       xmm4, (16-CONST_BITS)   ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+
+    movdqa      xmm0, xmm5
+    paddd       xmm5, xmm1              ; xmm5=tmp11L
+    psubd       xmm0, xmm1              ; xmm0=tmp12L
+    movdqa      xmm7, xmm4
+    paddd       xmm4, xmm2              ; xmm4=tmp11H
+    psubd       xmm7, xmm2              ; xmm7=tmp12H
+
+    movdqa      XMMWORD wk(4), xmm5   ; wk(4)=tmp11L
+    movdqa      XMMWORD wk(5), xmm4   ; wk(5)=tmp11H
+    movdqa      XMMWORD wk(6), xmm0   ; wk(6)=tmp12L
+    movdqa      XMMWORD wk(7), xmm7   ; wk(7)=tmp12H
+
+    ; -- Odd part
+
+    movdqa      xmm6, XMMWORD wk(9)   ; xmm6=col3
+    movdqa      xmm3, XMMWORD wk(8)   ; xmm3=col1
+    movdqa      xmm1, XMMWORD wk(11)  ; xmm1=col7
+    movdqa      xmm2, XMMWORD wk(10)  ; xmm2=col5
+
+    movdqa      xmm5, xmm6
+    movdqa      xmm4, xmm3
+    paddw       xmm5, xmm1              ; xmm5=z3
+    paddw       xmm4, xmm2              ; xmm4=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movdqa      xmm0, xmm5
+    movdqa      xmm7, xmm5
+    punpcklwd   xmm0, xmm4
+    punpckhwd   xmm7, xmm4
+    movdqa      xmm5, xmm0
+    movdqa      xmm4, xmm7
+    pmaddwd     xmm0, GOTOFF(ebx,PW_MF078_F117)  ; xmm0=z3L
+    pmaddwd     xmm7, GOTOFF(ebx,PW_MF078_F117)  ; xmm7=z3H
+    pmaddwd     xmm5, GOTOFF(ebx,PW_F117_F078)   ; xmm5=z4L
+    pmaddwd     xmm4, GOTOFF(ebx,PW_F117_F078)   ; xmm4=z4H
+
+    movdqa      XMMWORD wk(10), xmm0  ; wk(10)=z3L
+    movdqa      XMMWORD wk(11), xmm7  ; wk(11)=z3H
+
+    ; (Original)
+    ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+    ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+    ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+    ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+    ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+    ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+    ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+    ; tmp0 += z3;  tmp1 += z4;
+    ; tmp2 += z3;  tmp3 += z4;
+
+    movdqa      xmm0, xmm1
+    movdqa      xmm7, xmm1
+    punpcklwd   xmm0, xmm3
+    punpckhwd   xmm7, xmm3
+    movdqa      xmm1, xmm0
+    movdqa      xmm3, xmm7
+    pmaddwd     xmm0, GOTOFF(ebx,PW_MF060_MF089)  ; xmm0=tmp0L
+    pmaddwd     xmm7, GOTOFF(ebx,PW_MF060_MF089)  ; xmm7=tmp0H
+    pmaddwd     xmm1, GOTOFF(ebx,PW_MF089_F060)   ; xmm1=tmp3L
+    pmaddwd     xmm3, GOTOFF(ebx,PW_MF089_F060)   ; xmm3=tmp3H
+
+    paddd       xmm0, XMMWORD wk(10)  ; xmm0=tmp0L
+    paddd       xmm7, XMMWORD wk(11)  ; xmm7=tmp0H
+    paddd       xmm1, xmm5              ; xmm1=tmp3L
+    paddd       xmm3, xmm4              ; xmm3=tmp3H
+
+    movdqa      XMMWORD wk(8), xmm0   ; wk(8)=tmp0L
+    movdqa      XMMWORD wk(9), xmm7   ; wk(9)=tmp0H
+
+    movdqa      xmm0, xmm2
+    movdqa      xmm7, xmm2
+    punpcklwd   xmm0, xmm6
+    punpckhwd   xmm7, xmm6
+    movdqa      xmm2, xmm0
+    movdqa      xmm6, xmm7
+    pmaddwd     xmm0, GOTOFF(ebx,PW_MF050_MF256)  ; xmm0=tmp1L
+    pmaddwd     xmm7, GOTOFF(ebx,PW_MF050_MF256)  ; xmm7=tmp1H
+    pmaddwd     xmm2, GOTOFF(ebx,PW_MF256_F050)   ; xmm2=tmp2L
+    pmaddwd     xmm6, GOTOFF(ebx,PW_MF256_F050)   ; xmm6=tmp2H
+
+    paddd       xmm0, xmm5              ; xmm0=tmp1L
+    paddd       xmm7, xmm4              ; xmm7=tmp1H
+    paddd       xmm2, XMMWORD wk(10)  ; xmm2=tmp2L
+    paddd       xmm6, XMMWORD wk(11)  ; xmm6=tmp2H
+
+    movdqa      XMMWORD wk(10), xmm0  ; wk(10)=tmp1L
+    movdqa      XMMWORD wk(11), xmm7  ; wk(11)=tmp1H
+
+    ; -- Final output stage
+
+    movdqa      xmm5, XMMWORD wk(0)   ; xmm5=tmp10L
+    movdqa      xmm4, XMMWORD wk(1)   ; xmm4=tmp10H
+
+    movdqa      xmm0, xmm5
+    movdqa      xmm7, xmm4
+    paddd       xmm5, xmm1              ; xmm5=data0L
+    paddd       xmm4, xmm3              ; xmm4=data0H
+    psubd       xmm0, xmm1              ; xmm0=data7L
+    psubd       xmm7, xmm3              ; xmm7=data7H
+
+    movdqa      xmm1, GOTOFF(ebx,PD_DESCALE_P2)  ; xmm1=PD_DESCALE_P2
+
+    paddd       xmm5, xmm1
+    paddd       xmm4, xmm1
+    psrad       xmm5, DESCALE_P2
+    psrad       xmm4, DESCALE_P2
+    paddd       xmm0, xmm1
+    paddd       xmm7, xmm1
+    psrad       xmm0, DESCALE_P2
+    psrad       xmm7, DESCALE_P2
+
+    packssdw    xmm5, xmm4              ; xmm5=data0=(00 10 20 30 40 50 60 70)
+    packssdw    xmm0, xmm7              ; xmm0=data7=(07 17 27 37 47 57 67 77)
+
+    movdqa      xmm3, XMMWORD wk(4)   ; xmm3=tmp11L
+    movdqa      xmm1, XMMWORD wk(5)   ; xmm1=tmp11H
+
+    movdqa      xmm4, xmm3
+    movdqa      xmm7, xmm1
+    paddd       xmm3, xmm2              ; xmm3=data1L
+    paddd       xmm1, xmm6              ; xmm1=data1H
+    psubd       xmm4, xmm2              ; xmm4=data6L
+    psubd       xmm7, xmm6              ; xmm7=data6H
+
+    movdqa      xmm2, GOTOFF(ebx,PD_DESCALE_P2)  ; xmm2=PD_DESCALE_P2
+
+    paddd       xmm3, xmm2
+    paddd       xmm1, xmm2
+    psrad       xmm3, DESCALE_P2
+    psrad       xmm1, DESCALE_P2
+    paddd       xmm4, xmm2
+    paddd       xmm7, xmm2
+    psrad       xmm4, DESCALE_P2
+    psrad       xmm7, DESCALE_P2
+
+    packssdw    xmm3, xmm1              ; xmm3=data1=(01 11 21 31 41 51 61 71)
+    packssdw    xmm4, xmm7              ; xmm4=data6=(06 16 26 36 46 56 66 76)
+
+    packsswb    xmm5, xmm4              ; xmm5=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+    packsswb    xmm3, xmm0              ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+    movdqa      xmm6, XMMWORD wk(6)   ; xmm6=tmp12L
+    movdqa      xmm2, XMMWORD wk(7)   ; xmm2=tmp12H
+    movdqa      xmm1, XMMWORD wk(10)  ; xmm1=tmp1L
+    movdqa      xmm7, XMMWORD wk(11)  ; xmm7=tmp1H
+
+    movdqa      XMMWORD wk(0), xmm5   ; wk(0)=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+    movdqa      XMMWORD wk(1), xmm3   ; wk(1)=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+    movdqa      xmm4, xmm6
+    movdqa      xmm0, xmm2
+    paddd       xmm6, xmm1              ; xmm6=data2L
+    paddd       xmm2, xmm7              ; xmm2=data2H
+    psubd       xmm4, xmm1              ; xmm4=data5L
+    psubd       xmm0, xmm7              ; xmm0=data5H
+
+    movdqa      xmm5, GOTOFF(ebx,PD_DESCALE_P2)  ; xmm5=PD_DESCALE_P2
+
+    paddd       xmm6, xmm5
+    paddd       xmm2, xmm5
+    psrad       xmm6, DESCALE_P2
+    psrad       xmm2, DESCALE_P2
+    paddd       xmm4, xmm5
+    paddd       xmm0, xmm5
+    psrad       xmm4, DESCALE_P2
+    psrad       xmm0, DESCALE_P2
+
+    packssdw    xmm6, xmm2              ; xmm6=data2=(02 12 22 32 42 52 62 72)
+    packssdw    xmm4, xmm0              ; xmm4=data5=(05 15 25 35 45 55 65 75)
+
+    movdqa      xmm3, XMMWORD wk(2)   ; xmm3=tmp13L
+    movdqa      xmm1, XMMWORD wk(3)   ; xmm1=tmp13H
+    movdqa      xmm7, XMMWORD wk(8)   ; xmm7=tmp0L
+    movdqa      xmm5, XMMWORD wk(9)   ; xmm5=tmp0H
+
+    movdqa      xmm2, xmm3
+    movdqa      xmm0, xmm1
+    paddd       xmm3, xmm7              ; xmm3=data3L
+    paddd       xmm1, xmm5              ; xmm1=data3H
+    psubd       xmm2, xmm7              ; xmm2=data4L
+    psubd       xmm0, xmm5              ; xmm0=data4H
+
+    movdqa      xmm7, GOTOFF(ebx,PD_DESCALE_P2)  ; xmm7=PD_DESCALE_P2
+
+    paddd       xmm3, xmm7
+    paddd       xmm1, xmm7
+    psrad       xmm3, DESCALE_P2
+    psrad       xmm1, DESCALE_P2
+    paddd       xmm2, xmm7
+    paddd       xmm0, xmm7
+    psrad       xmm2, DESCALE_P2
+    psrad       xmm0, DESCALE_P2
+
+    movdqa      xmm5, GOTOFF(ebx,PB_CENTERJSAMP)  ; xmm5=PB_CENTERJSAMP
+
+    packssdw    xmm3, xmm1             ; xmm3=data3=(03 13 23 33 43 53 63 73)
+    packssdw    xmm2, xmm0             ; xmm2=data4=(04 14 24 34 44 54 64 74)
+
+    movdqa      xmm7, XMMWORD wk(0)  ; xmm7=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+    movdqa      xmm1, XMMWORD wk(1)  ; xmm1=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+    packsswb    xmm6, xmm2             ; xmm6=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+    packsswb    xmm3, xmm4             ; xmm3=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+    paddb       xmm7, xmm5
+    paddb       xmm1, xmm5
+    paddb       xmm6, xmm5
+    paddb       xmm3, xmm5
+
+    movdqa      xmm0, xmm7        ; transpose coefficients(phase 1)
+    punpcklbw   xmm7, xmm1        ; xmm7=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+    punpckhbw   xmm0, xmm1        ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+    movdqa      xmm2, xmm6        ; transpose coefficients(phase 1)
+    punpcklbw   xmm6, xmm3        ; xmm6=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+    punpckhbw   xmm2, xmm3        ; xmm2=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+    movdqa      xmm4, xmm7        ; transpose coefficients(phase 2)
+    punpcklwd   xmm7, xmm6        ; xmm7=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+    punpckhwd   xmm4, xmm6        ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+    movdqa      xmm5, xmm2        ; transpose coefficients(phase 2)
+    punpcklwd   xmm2, xmm0        ; xmm2=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+    punpckhwd   xmm5, xmm0        ; xmm5=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+    movdqa      xmm1, xmm7        ; transpose coefficients(phase 3)
+    punpckldq   xmm7, xmm2        ; xmm7=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+    punpckhdq   xmm1, xmm2        ; xmm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+    movdqa      xmm3, xmm4        ; transpose coefficients(phase 3)
+    punpckldq   xmm4, xmm5        ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+    punpckhdq   xmm3, xmm5        ; xmm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+    pshufd      xmm6, xmm7, 0x4E  ; xmm6=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+    pshufd      xmm0, xmm1, 0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+    pshufd      xmm2, xmm4, 0x4E  ; xmm2=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+    pshufd      xmm5, xmm3, 0x4E  ; xmm5=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+2*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm7
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm1
+    mov         edx, JSAMPROW edi+4*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+6*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm4
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm3
+
+    mov         edx, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+3*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm6
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm0
+    mov         edx, JSAMPROW edi+5*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+7*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD edx+eax*SIZEOF_JSAMPLE, xmm2
+    movq        XMM_MMWORD esi+eax*SIZEOF_JSAMPLE, xmm5
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctred-mmx.asm Added

@@ -0,0 +1,704 @@
+;
+; jidctred.asm - reduced-size IDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains inverse-DCT routines that produce reduced-size
+; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.
+; The following code is based directly on the IJG's original jidctred.c;
+; see the jidctred.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS    13
+%define PASS1_BITS    2
+
+%define DESCALE_P1_4  (CONST_BITS - PASS1_BITS + 1)
+%define DESCALE_P2_4  (CONST_BITS + PASS1_BITS + 3 + 1)
+%define DESCALE_P1_2  (CONST_BITS - PASS1_BITS + 2)
+%define DESCALE_P2_2  (CONST_BITS + PASS1_BITS + 3 + 2)
+
+%if CONST_BITS == 13
+F_0_211 equ  1730  ; FIX(0.211164243)
+F_0_509 equ  4176  ; FIX(0.509795579)
+F_0_601 equ  4926  ; FIX(0.601344887)
+F_0_720 equ  5906  ; FIX(0.720959822)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_850 equ  6967  ; FIX(0.850430095)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_061 equ  8697  ; FIX(1.061594337)
+F_1_272 equ 10426  ; FIX(1.272758580)
+F_1_451 equ 11893  ; FIX(1.451774981)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_2_172 equ 17799  ; FIX(2.172734803)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_624 equ 29692  ; FIX(3.624509785)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_211 equ DESCALE( 226735879, 30 - CONST_BITS)  ; FIX(0.211164243)
+F_0_509 equ DESCALE( 547388834, 30 - CONST_BITS)  ; FIX(0.509795579)
+F_0_601 equ DESCALE( 645689155, 30 - CONST_BITS)  ; FIX(0.601344887)
+F_0_720 equ DESCALE( 774124714, 30 - CONST_BITS)  ; FIX(0.720959822)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_850 equ DESCALE( 913142361, 30 - CONST_BITS)  ; FIX(0.850430095)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_061 equ DESCALE(1139878239, 30 - CONST_BITS)  ; FIX(1.061594337)
+F_1_272 equ DESCALE(1366614119, 30 - CONST_BITS)  ; FIX(1.272758580)
+F_1_451 equ DESCALE(1558831516, 30 - CONST_BITS)  ; FIX(1.451774981)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_2_172 equ DESCALE(2332956230, 30 - CONST_BITS)  ; FIX(2.172734803)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_624 equ DESCALE(3891787747, 30 - CONST_BITS)  ; FIX(3.624509785)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_red_mmx)
+
+EXTN(jconst_idct_red_mmx):
+
+PW_F184_MF076   times 2 dw  F_1_847, -F_0_765
+PW_F256_F089    times 2 dw  F_2_562,  F_0_899
+PW_F106_MF217   times 2 dw  F_1_061, -F_2_172
+PW_MF060_MF050  times 2 dw -F_0_601, -F_0_509
+PW_F145_MF021   times 2 dw  F_1_451, -F_0_211
+PW_F362_MF127   times 2 dw  F_3_624, -F_1_272
+PW_F085_MF072   times 2 dw  F_0_850, -F_0_720
+PD_DESCALE_P1_4 times 2 dd  1 << (DESCALE_P1_4 - 1)
+PD_DESCALE_P2_4 times 2 dd  1 << (DESCALE_P2_4 - 1)
+PD_DESCALE_P1_2 times 2 dd  1 << (DESCALE_P1_2 - 1)
+PD_DESCALE_P2_2 times 2 dd  1 << (DESCALE_P2_2 - 1)
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 4x4 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_4x4_mmx(void *dct_table, JCOEFPTR coef_block,
+;                    JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; void *dct_table
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_MMWORD
+                                        ; mmword wkWK_NUM
+%define WK_NUM         2
+%define workspace      wk(0) - DCTSIZE2 * SIZEOF_JCOEF
+                                        ; JCOEF workspaceDCTSIZE2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_4x4_mmx)
+
+EXTN(jsimd_idct_4x4_mmx):
+    push        ebp
+    mov         eax, esp                    ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_MMWORD)  ; align to 64 bits
+    mov         esp, eax
+    mov         ebp, esp                    ; ebp = aligned ebp
+    lea         esp, workspace
+    pushpic     ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input, store into work array.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+    lea         edi, workspace                 ; JCOEF *wsptr
+    mov         ecx, DCTSIZE/4                   ; ctr
+    alignx      16, 7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_4X4_MMX
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         short .columnDCT
+
+    movq        mm0, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    por         mm1, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    por         mm0, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    por         mm1, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    por         mm0, mm1
+    packsswb    mm0, mm0
+    movd        eax, mm0
+    test        eax, eax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    pmullw      mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    psllw       mm0, PASS1_BITS
+
+    movq        mm2, mm0                ; mm0=in0=(00 01 02 03)
+    punpcklwd   mm0, mm0                ; mm0=(00 00 01 01)
+    punpckhwd   mm2, mm2                ; mm2=(02 02 03 03)
+
+    movq        mm1, mm0
+    punpckldq   mm0, mm0                ; mm0=(00 00 00 00)
+    punpckhdq   mm1, mm1                ; mm1=(01 01 01 01)
+    movq        mm3, mm2
+    punpckldq   mm2, mm2                ; mm2=(02 02 02 02)
+    punpckhdq   mm3, mm3                ; mm3=(03 03 03 03)
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_JCOEF), mm1
+    movq        MMWORD MMBLOCK(2,0,edi,SIZEOF_JCOEF), mm2
+    movq        MMWORD MMBLOCK(3,0,edi,SIZEOF_JCOEF), mm3
+    jmp         near .nextcolumn
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    ; -- Odd part
+
+    movq        mm0, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    pmullw      mm0, MMWORD MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm1, MMWORD MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    movq        mm2, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    pmullw      mm2, MMWORD MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm3, MMWORD MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    movq        mm4, mm0
+    movq        mm5, mm0
+    punpcklwd   mm4, mm1
+    punpckhwd   mm5, mm1
+    movq        mm0, mm4
+    movq        mm1, mm5
+    pmaddwd     mm4, GOTOFF(ebx,PW_F256_F089)   ; mm4=(tmp2L)
+    pmaddwd     mm5, GOTOFF(ebx,PW_F256_F089)   ; mm5=(tmp2H)
+    pmaddwd     mm0, GOTOFF(ebx,PW_F106_MF217)  ; mm0=(tmp0L)
+    pmaddwd     mm1, GOTOFF(ebx,PW_F106_MF217)  ; mm1=(tmp0H)
+
+    movq        mm6, mm2
+    movq        mm7, mm2
+    punpcklwd   mm6, mm3
+    punpckhwd   mm7, mm3
+    movq        mm2, mm6
+    movq        mm3, mm7
+    pmaddwd     mm6, GOTOFF(ebx,PW_MF060_MF050)  ; mm6=(tmp2L)
+    pmaddwd     mm7, GOTOFF(ebx,PW_MF060_MF050)  ; mm7=(tmp2H)
+    pmaddwd     mm2, GOTOFF(ebx,PW_F145_MF021)   ; mm2=(tmp0L)
+    pmaddwd     mm3, GOTOFF(ebx,PW_F145_MF021)   ; mm3=(tmp0H)
+
+    paddd       mm6, mm4                ; mm6=tmp2L
+    paddd       mm7, mm5                ; mm7=tmp2H
+    paddd       mm2, mm0                ; mm2=tmp0L
+    paddd       mm3, mm1                ; mm3=tmp0H
+
+    movq        MMWORD wk(0), mm2     ; wk(0)=tmp0L
+    movq        MMWORD wk(1), mm3     ; wk(1)=tmp0H
+
+    ; -- Even part
+
+    movq        mm4, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movq        mm5, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    movq        mm0, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    pmullw      mm4, MMWORD MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm5, MMWORD MMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm0, MMWORD MMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    pxor        mm1, mm1
+    pxor        mm2, mm2
+    punpcklwd   mm1, mm4                ; mm1=tmp0L
+    punpckhwd   mm2, mm4                ; mm2=tmp0H
+    psrad       mm1, (16-CONST_BITS-1)  ; psrad mm1,16 & pslld mm1,CONST_BITS+1
+    psrad       mm2, (16-CONST_BITS-1)  ; psrad mm2,16 & pslld mm2,CONST_BITS+1
+
+    movq        mm3, mm5                ; mm5=in2=z2
+    punpcklwd   mm5, mm0                ; mm0=in6=z3
+    punpckhwd   mm3, mm0
+    pmaddwd     mm5, GOTOFF(ebx,PW_F184_MF076)  ; mm5=tmp2L
+    pmaddwd     mm3, GOTOFF(ebx,PW_F184_MF076)  ; mm3=tmp2H
+
+    movq        mm4, mm1
+    movq        mm0, mm2
+    paddd       mm1, mm5                ; mm1=tmp10L
+    paddd       mm2, mm3                ; mm2=tmp10H
+    psubd       mm4, mm5                ; mm4=tmp12L
+    psubd       mm0, mm3                ; mm0=tmp12H
+
+    ; -- Final output stage
+
+    movq        mm5, mm1
+    movq        mm3, mm2
+    paddd       mm1, mm6                ; mm1=data0L
+    paddd       mm2, mm7                ; mm2=data0H
+    psubd       mm5, mm6                ; mm5=data3L
+    psubd       mm3, mm7                ; mm3=data3H
+
+    movq        mm6, GOTOFF(ebx,PD_DESCALE_P1_4)  ; mm6=PD_DESCALE_P1_4
+
+    paddd       mm1, mm6
+    paddd       mm2, mm6
+    psrad       mm1, DESCALE_P1_4
+    psrad       mm2, DESCALE_P1_4
+    paddd       mm5, mm6
+    paddd       mm3, mm6
+    psrad       mm5, DESCALE_P1_4
+    psrad       mm3, DESCALE_P1_4
+
+    packssdw    mm1, mm2                ; mm1=data0=(00 01 02 03)
+    packssdw    mm5, mm3                ; mm5=data3=(30 31 32 33)
+
+    movq        mm7, MMWORD wk(0)     ; mm7=tmp0L
+    movq        mm6, MMWORD wk(1)     ; mm6=tmp0H
+
+    movq        mm2, mm4
+    movq        mm3, mm0
+    paddd       mm4, mm7                ; mm4=data1L
+    paddd       mm0, mm6                ; mm0=data1H
+    psubd       mm2, mm7                ; mm2=data2L
+    psubd       mm3, mm6                ; mm3=data2H
+
+    movq        mm7, GOTOFF(ebx,PD_DESCALE_P1_4)  ; mm7=PD_DESCALE_P1_4
+
+    paddd       mm4, mm7
+    paddd       mm0, mm7
+    psrad       mm4, DESCALE_P1_4
+    psrad       mm0, DESCALE_P1_4
+    paddd       mm2, mm7
+    paddd       mm3, mm7
+    psrad       mm2, DESCALE_P1_4
+    psrad       mm3, DESCALE_P1_4
+
+    packssdw    mm4, mm0                ; mm4=data1=(10 11 12 13)
+    packssdw    mm2, mm3                ; mm2=data2=(20 21 22 23)
+
+    movq        mm6, mm1                ; transpose coefficients(phase 1)
+    punpcklwd   mm1, mm4                ; mm1=(00 10 01 11)
+    punpckhwd   mm6, mm4                ; mm6=(02 12 03 13)
+    movq        mm7, mm2                ; transpose coefficients(phase 1)
+    punpcklwd   mm2, mm5                ; mm2=(20 30 21 31)
+    punpckhwd   mm7, mm5                ; mm7=(22 32 23 33)
+
+    movq        mm0, mm1                ; transpose coefficients(phase 2)
+    punpckldq   mm1, mm2                ; mm1=(00 10 20 30)
+    punpckhdq   mm0, mm2                ; mm0=(01 11 21 31)
+    movq        mm3, mm6                ; transpose coefficients(phase 2)
+    punpckldq   mm6, mm7                ; mm6=(02 12 22 32)
+    punpckhdq   mm3, mm7                ; mm3=(03 13 23 33)
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_JCOEF), mm1
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(2,0,edi,SIZEOF_JCOEF), mm6
+    movq        MMWORD MMBLOCK(3,0,edi,SIZEOF_JCOEF), mm3
+
+.nextcolumn:
+    add         esi, byte 4*SIZEOF_JCOEF            ; coef_block
+    add         edx, byte 4*SIZEOF_ISLOW_MULT_TYPE  ; quantptr
+    add         edi, byte 4*DCTSIZE*SIZEOF_JCOEF    ; wsptr
+    dec         ecx                                 ; ctr
+    jnz         near .columnloop
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         eax, original_ebp
+    lea         esi, workspace                   ; JCOEF *wsptr
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+
+    ; -- Odd part
+
+    movq        mm0, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    movq        mm2, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+
+    movq        mm4, mm0
+    movq        mm5, mm0
+    punpcklwd   mm4, mm1
+    punpckhwd   mm5, mm1
+    movq        mm0, mm4
+    movq        mm1, mm5
+    pmaddwd     mm4, GOTOFF(ebx,PW_F256_F089)   ; mm4=(tmp2L)
+    pmaddwd     mm5, GOTOFF(ebx,PW_F256_F089)   ; mm5=(tmp2H)
+    pmaddwd     mm0, GOTOFF(ebx,PW_F106_MF217)  ; mm0=(tmp0L)
+    pmaddwd     mm1, GOTOFF(ebx,PW_F106_MF217)  ; mm1=(tmp0H)
+
+    movq        mm6, mm2
+    movq        mm7, mm2
+    punpcklwd   mm6, mm3
+    punpckhwd   mm7, mm3
+    movq        mm2, mm6
+    movq        mm3, mm7
+    pmaddwd     mm6, GOTOFF(ebx,PW_MF060_MF050)  ; mm6=(tmp2L)
+    pmaddwd     mm7, GOTOFF(ebx,PW_MF060_MF050)  ; mm7=(tmp2H)
+    pmaddwd     mm2, GOTOFF(ebx,PW_F145_MF021)   ; mm2=(tmp0L)
+    pmaddwd     mm3, GOTOFF(ebx,PW_F145_MF021)   ; mm3=(tmp0H)
+
+    paddd       mm6, mm4                ; mm6=tmp2L
+    paddd       mm7, mm5                ; mm7=tmp2H
+    paddd       mm2, mm0                ; mm2=tmp0L
+    paddd       mm3, mm1                ; mm3=tmp0H
+
+    movq        MMWORD wk(0), mm2     ; wk(0)=tmp0L
+    movq        MMWORD wk(1), mm3     ; wk(1)=tmp0H
+
+    ; -- Even part
+
+    movq        mm4, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movq        mm5, MMWORD MMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    movq        mm0, MMWORD MMBLOCK(6,0,esi,SIZEOF_JCOEF)
+
+    pxor        mm1, mm1
+    pxor        mm2, mm2
+    punpcklwd   mm1, mm4                ; mm1=tmp0L
+    punpckhwd   mm2, mm4                ; mm2=tmp0H
+    psrad       mm1, (16-CONST_BITS-1)  ; psrad mm1,16 & pslld mm1,CONST_BITS+1
+    psrad       mm2, (16-CONST_BITS-1)  ; psrad mm2,16 & pslld mm2,CONST_BITS+1
+
+    movq        mm3, mm5                ; mm5=in2=z2
+    punpcklwd   mm5, mm0                ; mm0=in6=z3
+    punpckhwd   mm3, mm0
+    pmaddwd     mm5, GOTOFF(ebx,PW_F184_MF076)  ; mm5=tmp2L
+    pmaddwd     mm3, GOTOFF(ebx,PW_F184_MF076)  ; mm3=tmp2H
+
+    movq        mm4, mm1
+    movq        mm0, mm2
+    paddd       mm1, mm5                ; mm1=tmp10L
+    paddd       mm2, mm3                ; mm2=tmp10H
+    psubd       mm4, mm5                ; mm4=tmp12L
+    psubd       mm0, mm3                ; mm0=tmp12H
+
+    ; -- Final output stage
+
+    movq        mm5, mm1
+    movq        mm3, mm2
+    paddd       mm1, mm6                ; mm1=data0L
+    paddd       mm2, mm7                ; mm2=data0H
+    psubd       mm5, mm6                ; mm5=data3L
+    psubd       mm3, mm7                ; mm3=data3H
+
+    movq        mm6, GOTOFF(ebx,PD_DESCALE_P2_4)  ; mm6=PD_DESCALE_P2_4
+
+    paddd       mm1, mm6
+    paddd       mm2, mm6
+    psrad       mm1, DESCALE_P2_4
+    psrad       mm2, DESCALE_P2_4
+    paddd       mm5, mm6
+    paddd       mm3, mm6
+    psrad       mm5, DESCALE_P2_4
+    psrad       mm3, DESCALE_P2_4
+
+    packssdw    mm1, mm2                ; mm1=data0=(00 10 20 30)
+    packssdw    mm5, mm3                ; mm5=data3=(03 13 23 33)
+
+    movq        mm7, MMWORD wk(0)     ; mm7=tmp0L
+    movq        mm6, MMWORD wk(1)     ; mm6=tmp0H
+
+    movq        mm2, mm4
+    movq        mm3, mm0
+    paddd       mm4, mm7                ; mm4=data1L
+    paddd       mm0, mm6                ; mm0=data1H
+    psubd       mm2, mm7                ; mm2=data2L
+    psubd       mm3, mm6                ; mm3=data2H
+
+    movq        mm7, GOTOFF(ebx,PD_DESCALE_P2_4)  ; mm7=PD_DESCALE_P2_4
+
+    paddd       mm4, mm7
+    paddd       mm0, mm7
+    psrad       mm4, DESCALE_P2_4
+    psrad       mm0, DESCALE_P2_4
+    paddd       mm2, mm7
+    paddd       mm3, mm7
+    psrad       mm2, DESCALE_P2_4
+    psrad       mm3, DESCALE_P2_4
+
+    packssdw    mm4, mm0                ; mm4=data1=(01 11 21 31)
+    packssdw    mm2, mm3                ; mm2=data2=(02 12 22 32)
+
+    movq        mm6, GOTOFF(ebx,PB_CENTERJSAMP)  ; mm6=PB_CENTERJSAMP
+
+    packsswb    mm1, mm2                ; mm1=(00 10 20 30 02 12 22 32)
+    packsswb    mm4, mm5                ; mm4=(01 11 21 31 03 13 23 33)
+    paddb       mm1, mm6
+    paddb       mm4, mm6
+
+    movq        mm7, mm1                ; transpose coefficients(phase 1)
+    punpcklbw   mm1, mm4                ; mm1=(00 01 10 11 20 21 30 31)
+    punpckhbw   mm7, mm4                ; mm7=(02 03 12 13 22 23 32 33)
+
+    movq        mm0, mm1                ; transpose coefficients(phase 2)
+    punpcklwd   mm1, mm7                ; mm1=(00 01 02 03 10 11 12 13)
+    punpckhwd   mm0, mm7                ; mm0=(20 21 22 23 30 31 32 33)
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+2*SIZEOF_JSAMPROW
+    movd        dword edx+eax*SIZEOF_JSAMPLE, mm1
+    movd        dword esi+eax*SIZEOF_JSAMPLE, mm0
+
+    psrlq       mm1, 4*BYTE_BIT
+    psrlq       mm0, 4*BYTE_BIT
+
+    mov         edx, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+3*SIZEOF_JSAMPROW
+    movd        dword edx+eax*SIZEOF_JSAMPLE, mm1
+    movd        dword esi+eax*SIZEOF_JSAMPLE, mm0
+
+    emms                                ; empty MMX state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 2x2 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_2x2_mmx(void *dct_table, JCOEFPTR coef_block,
+;                    JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; void *dct_table
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_2x2_mmx)
+
+EXTN(jsimd_idct_2x2_mmx):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input.
+
+    mov         edx, POINTER dct_table(ebp)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(ebp)  ; inptr
+
+    ; | input:                  | result:        |
+    ; | 00 01 ** 03 ** 05 ** 07 |                |
+    ; | 10 11 ** 13 ** 15 ** 17 |                |
+    ; | ** ** ** ** ** ** ** ** |                |
+    ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 |
+    ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 |
+    ; | 50 51 ** 53 ** 55 ** 57 |                |
+    ; | ** ** ** ** ** ** ** ** |                |
+    ; | 70 71 ** 73 ** 75 ** 77 |                |
+
+    ; -- Odd part
+
+    movq        mm0, MMWORD MMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    pmullw      mm0, MMWORD MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm1, MMWORD MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    movq        mm2, MMWORD MMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movq        mm3, MMWORD MMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    pmullw      mm2, MMWORD MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm3, MMWORD MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; mm0=(10 11 ** 13), mm1=(30 31 ** 33)
+    ; mm2=(50 51 ** 53), mm3=(70 71 ** 73)
+
+    pcmpeqd     mm7, mm7
+    pslld       mm7, WORD_BIT           ; mm7={0x0000 0xFFFF 0x0000 0xFFFF}
+
+    movq        mm4, mm0                ; mm4=(10 11 ** 13)
+    movq        mm5, mm2                ; mm5=(50 51 ** 53)
+    punpcklwd   mm4, mm1                ; mm4=(10 30 11 31)
+    punpcklwd   mm5, mm3                ; mm5=(50 70 51 71)
+    pmaddwd     mm4, GOTOFF(ebx,PW_F362_MF127)
+    pmaddwd     mm5, GOTOFF(ebx,PW_F085_MF072)
+
+    psrld       mm0, WORD_BIT           ; mm0=(11 -- 13 --)
+    pand        mm1, mm7                ; mm1=(-- 31 -- 33)
+    psrld       mm2, WORD_BIT           ; mm2=(51 -- 53 --)
+    pand        mm3, mm7                ; mm3=(-- 71 -- 73)
+    por         mm0, mm1                ; mm0=(11 31 13 33)
+    por         mm2, mm3                ; mm2=(51 71 53 73)
+    pmaddwd     mm0, GOTOFF(ebx,PW_F362_MF127)
+    pmaddwd     mm2, GOTOFF(ebx,PW_F085_MF072)
+
+    paddd       mm4, mm5                ; mm4=tmp0col0 col1
+
+    movq        mm6, MMWORD MMBLOCK(1,1,esi,SIZEOF_JCOEF)
+    movq        mm1, MMWORD MMBLOCK(3,1,esi,SIZEOF_JCOEF)
+    pmullw      mm6, MMWORD MMBLOCK(1,1,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm1, MMWORD MMBLOCK(3,1,edx,SIZEOF_ISLOW_MULT_TYPE)
+    movq        mm3, MMWORD MMBLOCK(5,1,esi,SIZEOF_JCOEF)
+    movq        mm5, MMWORD MMBLOCK(7,1,esi,SIZEOF_JCOEF)
+    pmullw      mm3, MMWORD MMBLOCK(5,1,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm5, MMWORD MMBLOCK(7,1,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; mm6=(** 15 ** 17), mm1=(** 35 ** 37)
+    ; mm3=(** 55 ** 57), mm5=(** 75 ** 77)
+
+    psrld       mm6, WORD_BIT           ; mm6=(15 -- 17 --)
+    pand        mm1, mm7                ; mm1=(-- 35 -- 37)
+    psrld       mm3, WORD_BIT           ; mm3=(55 -- 57 --)
+    pand        mm5, mm7                ; mm5=(-- 75 -- 77)
+    por         mm6, mm1                ; mm6=(15 35 17 37)
+    por         mm3, mm5                ; mm3=(55 75 57 77)
+    pmaddwd     mm6, GOTOFF(ebx,PW_F362_MF127)
+    pmaddwd     mm3, GOTOFF(ebx,PW_F085_MF072)
+
+    paddd       mm0, mm2                ; mm0=tmp0col1 col3
+    paddd       mm6, mm3                ; mm6=tmp0col5 col7
+
+    ; -- Even part
+
+    movq        mm1, MMWORD MMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movq        mm5, MMWORD MMBLOCK(0,1,esi,SIZEOF_JCOEF)
+    pmullw      mm1, MMWORD MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      mm5, MMWORD MMBLOCK(0,1,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; mm1=(00 01 ** 03), mm5=(** 05 ** 07)
+
+    movq        mm2, mm1                      ; mm2=(00 01 ** 03)
+    pslld       mm1, WORD_BIT                 ; mm1=(-- 00 -- **)
+    psrad       mm1, (WORD_BIT-CONST_BITS-2)  ; mm1=tmp10col0 ****
+
+    pand        mm2, mm7                      ; mm2=(-- 01 -- 03)
+    pand        mm5, mm7                      ; mm5=(-- 05 -- 07)
+    psrad       mm2, (WORD_BIT-CONST_BITS-2)  ; mm2=tmp10col1 col3
+    psrad       mm5, (WORD_BIT-CONST_BITS-2)  ; mm5=tmp10col5 col7
+
+    ; -- Final output stage
+
+    movq        mm3, mm1
+    paddd       mm1, mm4                ; mm1=data0col0 ****=(A0 **)
+    psubd       mm3, mm4                ; mm3=data1col0 ****=(B0 **)
+    punpckldq   mm1, mm3                ; mm1=(A0 B0)
+
+    movq        mm7, GOTOFF(ebx,PD_DESCALE_P1_2)  ; mm7=PD_DESCALE_P1_2
+
+    movq        mm4, mm2
+    movq        mm3, mm5
+    paddd       mm2, mm0                ; mm2=data0col1 col3=(A1 A3)
+    paddd       mm5, mm6                ; mm5=data0col5 col7=(A5 A7)
+    psubd       mm4, mm0                ; mm4=data1col1 col3=(B1 B3)
+    psubd       mm3, mm6                ; mm3=data1col5 col7=(B5 B7)
+
+    paddd       mm1, mm7
+    psrad       mm1, DESCALE_P1_2
+
+    paddd       mm2, mm7
+    paddd       mm5, mm7
+    psrad       mm2, DESCALE_P1_2
+    psrad       mm5, DESCALE_P1_2
+    paddd       mm4, mm7
+    paddd       mm3, mm7
+    psrad       mm4, DESCALE_P1_2
+    psrad       mm3, DESCALE_P1_2
+
+    ; ---- Pass 2: process rows, store into output array.
+
+    mov         edi, JSAMPARRAY output_buf(ebp)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(ebp)
+
+    ; | input:| result:|
+    ; | A0 B0 |        |
+    ; | A1 B1 | C0 C1  |
+    ; | A3 B3 | D0 D1  |
+    ; | A5 B5 |        |
+    ; | A7 B7 |        |
+
+    ; -- Odd part
+
+    packssdw    mm2, mm4                ; mm2=(A1 A3 B1 B3)
+    packssdw    mm5, mm3                ; mm5=(A5 A7 B5 B7)
+    pmaddwd     mm2, GOTOFF(ebx,PW_F362_MF127)
+    pmaddwd     mm5, GOTOFF(ebx,PW_F085_MF072)
+
+    paddd       mm2, mm5                ; mm2=tmp0row0 row1
+
+    ; -- Even part
+
+    pslld       mm1, (CONST_BITS+2)     ; mm1=tmp10row0 row1
+
+    ; -- Final output stage
+
+    movq        mm0, GOTOFF(ebx,PD_DESCALE_P2_2)  ; mm0=PD_DESCALE_P2_2
+
+    movq        mm6, mm1
+    paddd       mm1, mm2                ; mm1=data0row0 row1=(C0 C1)
+    psubd       mm6, mm2                ; mm6=data1row0 row1=(D0 D1)
+
+    paddd       mm1, mm0
+    paddd       mm6, mm0
+    psrad       mm1, DESCALE_P2_2
+    psrad       mm6, DESCALE_P2_2
+
+    movq        mm7, mm1                ; transpose coefficients
+    punpckldq   mm1, mm6                ; mm1=(C0 D0)
+    punpckhdq   mm7, mm6                ; mm7=(C1 D1)
+
+    packssdw    mm1, mm7                ; mm1=(C0 D0 C1 D1)
+    packsswb    mm1, mm1                ; mm1=(C0 D0 C1 D1 C0 D0 C1 D1)
+    paddb       mm1, GOTOFF(ebx,PB_CENTERJSAMP)
+
+    movd        ecx, mm1
+    movd        ebx, mm1                ; ebx=(C0 D0 C1 D1)
+    shr         ecx, 2*BYTE_BIT         ; ecx=(C1 D1 -- --)
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    mov         word edx+eax*SIZEOF_JSAMPLE, bx
+    mov         word esi+eax*SIZEOF_JSAMPLE, cx
+
+    emms                                ; empty MMX state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jidctred-sse2.asm Added

@@ -0,0 +1,592 @@
+;
+; jidctred.asm - reduced-size IDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains inverse-DCT routines that produce reduced-size
+; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.
+; The following code is based directly on the IJG's original jidctred.c;
+; see the jidctred.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS    13
+%define PASS1_BITS    2
+
+%define DESCALE_P1_4  (CONST_BITS - PASS1_BITS + 1)
+%define DESCALE_P2_4  (CONST_BITS + PASS1_BITS + 3 + 1)
+%define DESCALE_P1_2  (CONST_BITS - PASS1_BITS + 2)
+%define DESCALE_P2_2  (CONST_BITS + PASS1_BITS + 3 + 2)
+
+%if CONST_BITS == 13
+F_0_211 equ  1730  ; FIX(0.211164243)
+F_0_509 equ  4176  ; FIX(0.509795579)
+F_0_601 equ  4926  ; FIX(0.601344887)
+F_0_720 equ  5906  ; FIX(0.720959822)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_850 equ  6967  ; FIX(0.850430095)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_061 equ  8697  ; FIX(1.061594337)
+F_1_272 equ 10426  ; FIX(1.272758580)
+F_1_451 equ 11893  ; FIX(1.451774981)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_2_172 equ 17799  ; FIX(2.172734803)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_624 equ 29692  ; FIX(3.624509785)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_211 equ DESCALE( 226735879, 30 - CONST_BITS)  ; FIX(0.211164243)
+F_0_509 equ DESCALE( 547388834, 30 - CONST_BITS)  ; FIX(0.509795579)
+F_0_601 equ DESCALE( 645689155, 30 - CONST_BITS)  ; FIX(0.601344887)
+F_0_720 equ DESCALE( 774124714, 30 - CONST_BITS)  ; FIX(0.720959822)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_850 equ DESCALE( 913142361, 30 - CONST_BITS)  ; FIX(0.850430095)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_061 equ DESCALE(1139878239, 30 - CONST_BITS)  ; FIX(1.061594337)
+F_1_272 equ DESCALE(1366614119, 30 - CONST_BITS)  ; FIX(1.272758580)
+F_1_451 equ DESCALE(1558831516, 30 - CONST_BITS)  ; FIX(1.451774981)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_2_172 equ DESCALE(2332956230, 30 - CONST_BITS)  ; FIX(2.172734803)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_624 equ DESCALE(3891787747, 30 - CONST_BITS)  ; FIX(3.624509785)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_red_sse2)
+
+EXTN(jconst_idct_red_sse2):
+
+PW_F184_MF076   times 4  dw  F_1_847, -F_0_765
+PW_F256_F089    times 4  dw  F_2_562,  F_0_899
+PW_F106_MF217   times 4  dw  F_1_061, -F_2_172
+PW_MF060_MF050  times 4  dw -F_0_601, -F_0_509
+PW_F145_MF021   times 4  dw  F_1_451, -F_0_211
+PW_F362_MF127   times 4  dw  F_3_624, -F_1_272
+PW_F085_MF072   times 4  dw  F_0_850, -F_0_720
+PD_DESCALE_P1_4 times 4  dd  1 << (DESCALE_P1_4 - 1)
+PD_DESCALE_P2_4 times 4  dd  1 << (DESCALE_P2_4 - 1)
+PD_DESCALE_P1_2 times 4  dd  1 << (DESCALE_P1_2 - 1)
+PD_DESCALE_P2_2 times 4  dd  1 << (DESCALE_P2_2 - 1)
+PB_CENTERJSAMP  times 16 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 4x4 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_4x4_sse2(void *dct_table, JCOEFPTR coef_block,
+;                     JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; void *dct_table
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+%define original_ebp   ebp + 0
+%define wk(i)          ebp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM         2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_4x4_sse2)
+
+EXTN(jsimd_idct_4x4_sse2):
+    push        ebp
+    mov         eax, esp                     ; eax = original ebp
+    sub         esp, byte 4
+    and         esp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         esp, eax
+    mov         ebp, esp                     ; ebp = aligned ebp
+    lea         esp, wk(0)
+    pushpic     ebx
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input.
+
+;   mov         eax, original_ebp
+    mov         edx, POINTER dct_table(eax)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(eax)  ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_4X4_SSE2
+    mov         eax, dword DWBLOCK(1,0,esi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,esi,SIZEOF_JCOEF)
+    jnz         short .columnDCT
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    por         xmm0, xmm1
+    packsswb    xmm0, xmm0
+    packsswb    xmm0, xmm0
+    movd        eax, xmm0
+    test        eax, eax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    psllw       xmm0, PASS1_BITS
+
+    movdqa      xmm3, xmm0        ; xmm0=in0=(00 01 02 03 04 05 06 07)
+    punpcklwd   xmm0, xmm0        ; xmm0=(00 00 01 01 02 02 03 03)
+    punpckhwd   xmm3, xmm3        ; xmm3=(04 04 05 05 06 06 07 07)
+
+    pshufd      xmm1, xmm0, 0x50  ; xmm1=col0 col1=(00 00 00 00 01 01 01 01)
+    pshufd      xmm0, xmm0, 0xFA  ; xmm0=col2 col3=(02 02 02 02 03 03 03 03)
+    pshufd      xmm6, xmm3, 0x50  ; xmm6=col4 col5=(04 04 04 04 05 05 05 05)
+    pshufd      xmm3, xmm3, 0xFA  ; xmm3=col6 col7=(06 06 06 06 07 07 07 07)
+
+    jmp         near .column_end
+    alignx      16, 7
+%endif
+.columnDCT:
+
+    ; -- Odd part
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    movdqa      xmm2, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    movdqa      xmm4, xmm0
+    movdqa      xmm5, xmm0
+    punpcklwd   xmm4, xmm1
+    punpckhwd   xmm5, xmm1
+    movdqa      xmm0, xmm4
+    movdqa      xmm1, xmm5
+    pmaddwd     xmm4, GOTOFF(ebx,PW_F256_F089)   ; xmm4=(tmp2L)
+    pmaddwd     xmm5, GOTOFF(ebx,PW_F256_F089)   ; xmm5=(tmp2H)
+    pmaddwd     xmm0, GOTOFF(ebx,PW_F106_MF217)  ; xmm0=(tmp0L)
+    pmaddwd     xmm1, GOTOFF(ebx,PW_F106_MF217)  ; xmm1=(tmp0H)
+
+    movdqa      xmm6, xmm2
+    movdqa      xmm7, xmm2
+    punpcklwd   xmm6, xmm3
+    punpckhwd   xmm7, xmm3
+    movdqa      xmm2, xmm6
+    movdqa      xmm3, xmm7
+    pmaddwd     xmm6, GOTOFF(ebx,PW_MF060_MF050)  ; xmm6=(tmp2L)
+    pmaddwd     xmm7, GOTOFF(ebx,PW_MF060_MF050)  ; xmm7=(tmp2H)
+    pmaddwd     xmm2, GOTOFF(ebx,PW_F145_MF021)   ; xmm2=(tmp0L)
+    pmaddwd     xmm3, GOTOFF(ebx,PW_F145_MF021)   ; xmm3=(tmp0H)
+
+    paddd       xmm6, xmm4              ; xmm6=tmp2L
+    paddd       xmm7, xmm5              ; xmm7=tmp2H
+    paddd       xmm2, xmm0              ; xmm2=tmp0L
+    paddd       xmm3, xmm1              ; xmm3=tmp0H
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=tmp0L
+    movdqa      XMMWORD wk(1), xmm3   ; wk(1)=tmp0H
+
+    ; -- Even part
+
+    movdqa      xmm4, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm5, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm0, XMMWORD XMMBLOCK(6,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm4, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm5, XMMWORD XMMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm0, XMMWORD XMMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    pxor        xmm1, xmm1
+    pxor        xmm2, xmm2
+    punpcklwd   xmm1, xmm4               ; xmm1=tmp0L
+    punpckhwd   xmm2, xmm4               ; xmm2=tmp0H
+    psrad       xmm1, (16-CONST_BITS-1)  ; psrad xmm1,16 & pslld xmm1,CONST_BITS+1
+    psrad       xmm2, (16-CONST_BITS-1)  ; psrad xmm2,16 & pslld xmm2,CONST_BITS+1
+
+    movdqa      xmm3, xmm5              ; xmm5=in2=z2
+    punpcklwd   xmm5, xmm0              ; xmm0=in6=z3
+    punpckhwd   xmm3, xmm0
+    pmaddwd     xmm5, GOTOFF(ebx,PW_F184_MF076)  ; xmm5=tmp2L
+    pmaddwd     xmm3, GOTOFF(ebx,PW_F184_MF076)  ; xmm3=tmp2H
+
+    movdqa      xmm4, xmm1
+    movdqa      xmm0, xmm2
+    paddd       xmm1, xmm5              ; xmm1=tmp10L
+    paddd       xmm2, xmm3              ; xmm2=tmp10H
+    psubd       xmm4, xmm5              ; xmm4=tmp12L
+    psubd       xmm0, xmm3              ; xmm0=tmp12H
+
+    ; -- Final output stage
+
+    movdqa      xmm5, xmm1
+    movdqa      xmm3, xmm2
+    paddd       xmm1, xmm6              ; xmm1=data0L
+    paddd       xmm2, xmm7              ; xmm2=data0H
+    psubd       xmm5, xmm6              ; xmm5=data3L
+    psubd       xmm3, xmm7              ; xmm3=data3H
+
+    movdqa      xmm6, GOTOFF(ebx,PD_DESCALE_P1_4)  ; xmm6=PD_DESCALE_P1_4
+
+    paddd       xmm1, xmm6
+    paddd       xmm2, xmm6
+    psrad       xmm1, DESCALE_P1_4
+    psrad       xmm2, DESCALE_P1_4
+    paddd       xmm5, xmm6
+    paddd       xmm3, xmm6
+    psrad       xmm5, DESCALE_P1_4
+    psrad       xmm3, DESCALE_P1_4
+
+    packssdw    xmm1, xmm2              ; xmm1=data0=(00 01 02 03 04 05 06 07)
+    packssdw    xmm5, xmm3              ; xmm5=data3=(30 31 32 33 34 35 36 37)
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=tmp0L
+    movdqa      xmm6, XMMWORD wk(1)   ; xmm6=tmp0H
+
+    movdqa      xmm2, xmm4
+    movdqa      xmm3, xmm0
+    paddd       xmm4, xmm7              ; xmm4=data1L
+    paddd       xmm0, xmm6              ; xmm0=data1H
+    psubd       xmm2, xmm7              ; xmm2=data2L
+    psubd       xmm3, xmm6              ; xmm3=data2H
+
+    movdqa      xmm7, GOTOFF(ebx,PD_DESCALE_P1_4)  ; xmm7=PD_DESCALE_P1_4
+
+    paddd       xmm4, xmm7
+    paddd       xmm0, xmm7
+    psrad       xmm4, DESCALE_P1_4
+    psrad       xmm0, DESCALE_P1_4
+    paddd       xmm2, xmm7
+    paddd       xmm3, xmm7
+    psrad       xmm2, DESCALE_P1_4
+    psrad       xmm3, DESCALE_P1_4
+
+    packssdw    xmm4, xmm0        ; xmm4=data1=(10 11 12 13 14 15 16 17)
+    packssdw    xmm2, xmm3        ; xmm2=data2=(20 21 22 23 24 25 26 27)
+
+    movdqa      xmm6, xmm1        ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm4        ; xmm1=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm6, xmm4        ; xmm6=(04 14 05 15 06 16 07 17)
+    movdqa      xmm7, xmm2        ; transpose coefficients(phase 1)
+    punpcklwd   xmm2, xmm5        ; xmm2=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm7, xmm5        ; xmm7=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm0, xmm1        ; transpose coefficients(phase 2)
+    punpckldq   xmm1, xmm2        ; xmm1=col0 col1=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm0, xmm2        ; xmm0=col2 col3=(02 12 22 32 03 13 23 33)
+    movdqa      xmm3, xmm6        ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm7        ; xmm6=col4 col5=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm3, xmm7        ; xmm3=col6 col7=(06 16 26 36 07 17 27 37)
+.column_end:
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows, store into output array.
+
+    mov         eax, original_ebp
+    mov         edi, JSAMPARRAY output_buf(eax)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(eax)
+
+    ; -- Even part
+
+    pxor        xmm4, xmm4
+    punpcklwd   xmm4, xmm1               ; xmm4=tmp0
+    psrad       xmm4, (16-CONST_BITS-1)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS+1
+
+    ; -- Odd part
+
+    punpckhwd   xmm1, xmm0
+    punpckhwd   xmm6, xmm3
+    movdqa      xmm5, xmm1
+    movdqa      xmm2, xmm6
+    pmaddwd     xmm1, GOTOFF(ebx,PW_F256_F089)    ; xmm1=(tmp2)
+    pmaddwd     xmm6, GOTOFF(ebx,PW_MF060_MF050)  ; xmm6=(tmp2)
+    pmaddwd     xmm5, GOTOFF(ebx,PW_F106_MF217)   ; xmm5=(tmp0)
+    pmaddwd     xmm2, GOTOFF(ebx,PW_F145_MF021)   ; xmm2=(tmp0)
+
+    paddd       xmm6, xmm1              ; xmm6=tmp2
+    paddd       xmm2, xmm5              ; xmm2=tmp0
+
+    ; -- Even part
+
+    punpcklwd   xmm0, xmm3
+    pmaddwd     xmm0, GOTOFF(ebx,PW_F184_MF076)  ; xmm0=tmp2
+
+    movdqa      xmm7, xmm4
+    paddd       xmm4, xmm0              ; xmm4=tmp10
+    psubd       xmm7, xmm0              ; xmm7=tmp12
+
+    ; -- Final output stage
+
+    movdqa      xmm1, GOTOFF(ebx,PD_DESCALE_P2_4)  ; xmm1=PD_DESCALE_P2_4
+
+    movdqa      xmm5, xmm4
+    movdqa      xmm3, xmm7
+    paddd       xmm4, xmm6              ; xmm4=data0=(00 10 20 30)
+    paddd       xmm7, xmm2              ; xmm7=data1=(01 11 21 31)
+    psubd       xmm5, xmm6              ; xmm5=data3=(03 13 23 33)
+    psubd       xmm3, xmm2              ; xmm3=data2=(02 12 22 32)
+
+    paddd       xmm4, xmm1
+    paddd       xmm7, xmm1
+    psrad       xmm4, DESCALE_P2_4
+    psrad       xmm7, DESCALE_P2_4
+    paddd       xmm5, xmm1
+    paddd       xmm3, xmm1
+    psrad       xmm5, DESCALE_P2_4
+    psrad       xmm3, DESCALE_P2_4
+
+    packssdw    xmm4, xmm3              ; xmm4=(00 10 20 30 02 12 22 32)
+    packssdw    xmm7, xmm5              ; xmm7=(01 11 21 31 03 13 23 33)
+
+    movdqa      xmm0, xmm4              ; transpose coefficients(phase 1)
+    punpcklwd   xmm4, xmm7              ; xmm4=(00 01 10 11 20 21 30 31)
+    punpckhwd   xmm0, xmm7              ; xmm0=(02 03 12 13 22 23 32 33)
+
+    movdqa      xmm6, xmm4              ; transpose coefficients(phase 2)
+    punpckldq   xmm4, xmm0              ; xmm4=(00 01 02 03 10 11 12 13)
+    punpckhdq   xmm6, xmm0              ; xmm6=(20 21 22 23 30 31 32 33)
+
+    packsswb    xmm4, xmm6              ; xmm4=(00 01 02 03 10 11 12 13 20 ..)
+    paddb       xmm4, GOTOFF(ebx,PB_CENTERJSAMP)
+
+    pshufd      xmm2, xmm4, 0x39        ; xmm2=(10 11 12 13 20 21 22 23 30 ..)
+    pshufd      xmm1, xmm4, 0x4E        ; xmm1=(20 21 22 23 30 31 32 33 00 ..)
+    pshufd      xmm3, xmm4, 0x93        ; xmm3=(30 31 32 33 00 01 02 03 10 ..)
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    movd        XMM_DWORD edx+eax*SIZEOF_JSAMPLE, xmm4
+    movd        XMM_DWORD esi+eax*SIZEOF_JSAMPLE, xmm2
+    mov         edx, JSAMPROW edi+2*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+3*SIZEOF_JSAMPROW
+    movd        XMM_DWORD edx+eax*SIZEOF_JSAMPLE, xmm1
+    movd        XMM_DWORD esi+eax*SIZEOF_JSAMPLE, xmm3
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+    poppic      ebx
+    mov         esp, ebp                ; esp <- aligned ebp
+    pop         esp                     ; esp <- original ebp
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 2x2 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_2x2_sse2(void *dct_table, JCOEFPTR coef_block,
+;                     JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)   (b) + 8          ; void *dct_table
+%define coef_block(b)  (b) + 12         ; JCOEFPTR coef_block
+%define output_buf(b)  (b) + 16         ; JSAMPARRAY output_buf
+%define output_col(b)  (b) + 20         ; JDIMENSION output_col
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_2x2_sse2)
+
+EXTN(jsimd_idct_2x2_sse2):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    get_GOT     ebx                     ; get GOT address
+
+    ; ---- Pass 1: process columns from input.
+
+    mov         edx, POINTER dct_table(ebp)    ; quantptr
+    mov         esi, JCOEFPTR coef_block(ebp)  ; inptr
+
+    ; | input:                  | result:        |
+    ; | 00 01 ** 03 ** 05 ** 07 |                |
+    ; | 10 11 ** 13 ** 15 ** 17 |                |
+    ; | ** ** ** ** ** ** ** ** |                |
+    ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 |
+    ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 |
+    ; | 50 51 ** 53 ** 55 ** 57 |                |
+    ; | ** ** ** ** ** ** ** ** |                |
+    ; | 70 71 ** 73 ** 75 ** 77 |                |
+
+    ; -- Odd part
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    movdqa      xmm2, XMMWORD XMMBLOCK(5,0,esi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; xmm0=(10 11 ** 13 ** 15 ** 17), xmm1=(30 31 ** 33 ** 35 ** 37)
+    ; xmm2=(50 51 ** 53 ** 55 ** 57), xmm3=(70 71 ** 73 ** 75 ** 77)
+
+    pcmpeqd     xmm7, xmm7
+    pslld       xmm7, WORD_BIT          ; xmm7={0x0000 0xFFFF 0x0000 0xFFFF ..}
+
+    movdqa      xmm4, xmm0              ; xmm4=(10 11 ** 13 ** 15 ** 17)
+    movdqa      xmm5, xmm2              ; xmm5=(50 51 ** 53 ** 55 ** 57)
+    punpcklwd   xmm4, xmm1              ; xmm4=(10 30 11 31 ** ** 13 33)
+    punpcklwd   xmm5, xmm3              ; xmm5=(50 70 51 71 ** ** 53 73)
+    pmaddwd     xmm4, GOTOFF(ebx,PW_F362_MF127)
+    pmaddwd     xmm5, GOTOFF(ebx,PW_F085_MF072)
+
+    psrld       xmm0, WORD_BIT          ; xmm0=(11 -- 13 -- 15 -- 17 --)
+    pand        xmm1, xmm7              ; xmm1=(-- 31 -- 33 -- 35 -- 37)
+    psrld       xmm2, WORD_BIT          ; xmm2=(51 -- 53 -- 55 -- 57 --)
+    pand        xmm3, xmm7              ; xmm3=(-- 71 -- 73 -- 75 -- 77)
+    por         xmm0, xmm1              ; xmm0=(11 31 13 33 15 35 17 37)
+    por         xmm2, xmm3              ; xmm2=(51 71 53 73 55 75 57 77)
+    pmaddwd     xmm0, GOTOFF(ebx,PW_F362_MF127)
+    pmaddwd     xmm2, GOTOFF(ebx,PW_F085_MF072)
+
+    paddd       xmm4, xmm5              ; xmm4=tmp0col0 col1 **** col3
+    paddd       xmm0, xmm2              ; xmm0=tmp0col1 col3 col5 col7
+
+    ; -- Even part
+
+    movdqa      xmm6, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_JCOEF)
+    pmullw      xmm6, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; xmm6=(00 01 ** 03 ** 05 ** 07)
+
+    movdqa      xmm1, xmm6              ; xmm1=(00 01 ** 03 ** 05 ** 07)
+    pslld       xmm6, WORD_BIT          ; xmm6=(-- 00 -- ** -- ** -- **)
+    pand        xmm1, xmm7              ; xmm1=(-- 01 -- 03 -- 05 -- 07)
+    psrad       xmm6, (WORD_BIT-CONST_BITS-2)  ; xmm6=tmp10col0 **** **** ****
+    psrad       xmm1, (WORD_BIT-CONST_BITS-2)  ; xmm1=tmp10col1 col3 col5 col7
+
+    ; -- Final output stage
+
+    movdqa      xmm3, xmm6
+    movdqa      xmm5, xmm1
+    paddd       xmm6, xmm4      ; xmm6=data0col0 **** **** ****=(A0 ** ** **)
+    paddd       xmm1, xmm0      ; xmm1=data0col1 col3 col5 col7=(A1 A3 A5 A7)
+    psubd       xmm3, xmm4      ; xmm3=data1col0 **** **** ****=(B0 ** ** **)
+    psubd       xmm5, xmm0      ; xmm5=data1col1 col3 col5 col7=(B1 B3 B5 B7)
+
+    movdqa      xmm2, GOTOFF(ebx,PD_DESCALE_P1_2)  ; xmm2=PD_DESCALE_P1_2
+
+    punpckldq   xmm6, xmm3              ; xmm6=(A0 B0 ** **)
+
+    movdqa      xmm7, xmm1
+    punpcklqdq  xmm1, xmm5              ; xmm1=(A1 A3 B1 B3)
+    punpckhqdq  xmm7, xmm5              ; xmm7=(A5 A7 B5 B7)
+
+    paddd       xmm6, xmm2
+    psrad       xmm6, DESCALE_P1_2
+
+    paddd       xmm1, xmm2
+    paddd       xmm7, xmm2
+    psrad       xmm1, DESCALE_P1_2
+    psrad       xmm7, DESCALE_P1_2
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta esi + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows, store into output array.
+
+    mov         edi, JSAMPARRAY output_buf(ebp)  ; (JSAMPROW *)
+    mov         eax, JDIMENSION output_col(ebp)
+
+    ; | input:| result:|
+    ; | A0 B0 |        |
+    ; | A1 B1 | C0 C1  |
+    ; | A3 B3 | D0 D1  |
+    ; | A5 B5 |        |
+    ; | A7 B7 |        |
+
+    ; -- Odd part
+
+    packssdw    xmm1, xmm1              ; xmm1=(A1 A3 B1 B3 A1 A3 B1 B3)
+    packssdw    xmm7, xmm7              ; xmm7=(A5 A7 B5 B7 A5 A7 B5 B7)
+    pmaddwd     xmm1, GOTOFF(ebx,PW_F362_MF127)
+    pmaddwd     xmm7, GOTOFF(ebx,PW_F085_MF072)
+
+    paddd       xmm1, xmm7              ; xmm1=tmp0row0 row1 row0 row1
+
+    ; -- Even part
+
+    pslld       xmm6, (CONST_BITS+2)    ; xmm6=tmp10row0 row1 **** ****
+
+    ; -- Final output stage
+
+    movdqa      xmm4, xmm6
+    paddd       xmm6, xmm1     ; xmm6=data0row0 row1 **** ****=(C0 C1 ** **)
+    psubd       xmm4, xmm1     ; xmm4=data1row0 row1 **** ****=(D0 D1 ** **)
+
+    punpckldq   xmm6, xmm4     ; xmm6=(C0 D0 C1 D1)
+
+    paddd       xmm6, GOTOFF(ebx,PD_DESCALE_P2_2)
+    psrad       xmm6, DESCALE_P2_2
+
+    packssdw    xmm6, xmm6              ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1)
+    packsswb    xmm6, xmm6              ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1 ..)
+    paddb       xmm6, GOTOFF(ebx,PB_CENTERJSAMP)
+
+    pextrw      ebx, xmm6, 0x00         ; ebx=(C0 D0 -- --)
+    pextrw      ecx, xmm6, 0x01         ; ecx=(C1 D1 -- --)
+
+    mov         edx, JSAMPROW edi+0*SIZEOF_JSAMPROW
+    mov         esi, JSAMPROW edi+1*SIZEOF_JSAMPROW
+    mov         word edx+eax*SIZEOF_JSAMPLE, bx
+    mov         word esi+eax*SIZEOF_JSAMPLE, cx
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jquant-3dn.asm Added

@@ -0,0 +1,230 @@
+;
+; jquant.asm - sample data conversion and quantization (3DNow! & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_3dnow(JSAMPARRAY sample_data, JDIMENSION start_col,
+;                            FAST_FLOAT *workspace);
+;
+
+%define sample_data  ebp + 8            ; JSAMPARRAY sample_data
+%define start_col    ebp + 12           ; JDIMENSION start_col
+%define workspace    ebp + 16           ; FAST_FLOAT *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_convsamp_float_3dnow)
+
+EXTN(jsimd_convsamp_float_3dnow):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    pcmpeqw     mm7, mm7
+    psllw       mm7, 7
+    packsswb    mm7, mm7                ; mm7 = PB_CENTERJSAMPLE (0x808080..)
+
+    mov         esi, JSAMPARRAY sample_data  ; (JSAMPROW *)
+    mov         eax, JDIMENSION start_col
+    mov         edi, POINTER workspace       ; (DCTELEM *)
+    mov         ecx, DCTSIZE/2
+    alignx      16, 7
+.convloop:
+    mov         ebx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        mm0, MMWORD ebx+eax*SIZEOF_JSAMPLE
+    movq        mm1, MMWORD edx+eax*SIZEOF_JSAMPLE
+
+    psubb       mm0, mm7                ; mm0=(01234567)
+    psubb       mm1, mm7                ; mm1=(89ABCDEF)
+
+    punpcklbw   mm2, mm0                ; mm2=(*0*1*2*3)
+    punpckhbw   mm0, mm0                ; mm0=(*4*5*6*7)
+    punpcklbw   mm3, mm1                ; mm3=(*8*9*A*B)
+    punpckhbw   mm1, mm1                ; mm1=(*C*D*E*F)
+
+    punpcklwd   mm4, mm2                ; mm4=(***0***1)
+    punpckhwd   mm2, mm2                ; mm2=(***2***3)
+    punpcklwd   mm5, mm0                ; mm5=(***4***5)
+    punpckhwd   mm0, mm0                ; mm0=(***6***7)
+
+    psrad       mm4, (DWORD_BIT-BYTE_BIT)  ; mm4=(01)
+    psrad       mm2, (DWORD_BIT-BYTE_BIT)  ; mm2=(23)
+    pi2fd       mm4, mm4
+    pi2fd       mm2, mm2
+    psrad       mm5, (DWORD_BIT-BYTE_BIT)  ; mm5=(45)
+    psrad       mm0, (DWORD_BIT-BYTE_BIT)  ; mm0=(67)
+    pi2fd       mm5, mm5
+    pi2fd       mm0, mm0
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT), mm4
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT), mm2
+    movq        MMWORD MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT), mm5
+    movq        MMWORD MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT), mm0
+
+    punpcklwd   mm6, mm3                ; mm6=(***8***9)
+    punpckhwd   mm3, mm3                ; mm3=(***A***B)
+    punpcklwd   mm4, mm1                ; mm4=(***C***D)
+    punpckhwd   mm1, mm1                ; mm1=(***E***F)
+
+    psrad       mm6, (DWORD_BIT-BYTE_BIT)  ; mm6=(89)
+    psrad       mm3, (DWORD_BIT-BYTE_BIT)  ; mm3=(AB)
+    pi2fd       mm6, mm6
+    pi2fd       mm3, mm3
+    psrad       mm4, (DWORD_BIT-BYTE_BIT)  ; mm4=(CD)
+    psrad       mm1, (DWORD_BIT-BYTE_BIT)  ; mm1=(EF)
+    pi2fd       mm4, mm4
+    pi2fd       mm1, mm1
+
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT), mm6
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT), mm3
+    movq        MMWORD MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT), mm4
+    movq        MMWORD MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT), mm1
+
+    add         esi, byte 2*SIZEOF_JSAMPROW
+    add         edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+    dec         ecx
+    jnz         near .convloop
+
+    femms                               ; empty MMX/3DNow! state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_3dnow(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                            FAST_FLOAT *workspace);
+;
+
+%define coef_block  ebp + 8             ; JCOEFPTR coef_block
+%define divisors    ebp + 12            ; FAST_FLOAT *divisors
+%define workspace   ebp + 16            ; FAST_FLOAT *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_quantize_float_3dnow)
+
+EXTN(jsimd_quantize_float_3dnow):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         eax, 0x4B400000         ; (float)0x00C00000 (rndint_magic)
+    movd        mm7, eax
+    punpckldq   mm7, mm7                ; mm7={12582912.0F 12582912.0F}
+
+    mov         esi, POINTER workspace
+    mov         edx, POINTER divisors
+    mov         edi, JCOEFPTR coef_block
+    mov         eax, DCTSIZE2/16
+    alignx      16, 7
+.quantloop:
+    movq        mm0, MMWORD MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)
+    movq        mm1, MMWORD MMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)
+    pfmul       mm0, MMWORD MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)
+    pfmul       mm1, MMWORD MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)
+    movq        mm2, MMWORD MMBLOCK(0,2,esi,SIZEOF_FAST_FLOAT)
+    movq        mm3, MMWORD MMBLOCK(0,3,esi,SIZEOF_FAST_FLOAT)
+    pfmul       mm2, MMWORD MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)
+    pfmul       mm3, MMWORD MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)
+
+    pfadd       mm0, mm7                ; mm0=(00 ** 01 **)
+    pfadd       mm1, mm7                ; mm1=(02 ** 03 **)
+    pfadd       mm2, mm7                ; mm0=(04 ** 05 **)
+    pfadd       mm3, mm7                ; mm1=(06 ** 07 **)
+
+    movq        mm4, mm0
+    punpcklwd   mm0, mm1                ; mm0=(00 02 ** **)
+    punpckhwd   mm4, mm1                ; mm4=(01 03 ** **)
+    movq        mm5, mm2
+    punpcklwd   mm2, mm3                ; mm2=(04 06 ** **)
+    punpckhwd   mm5, mm3                ; mm5=(05 07 ** **)
+
+    punpcklwd   mm0, mm4                ; mm0=(00 01 02 03)
+    punpcklwd   mm2, mm5                ; mm2=(04 05 06 07)
+
+    movq        mm6, MMWORD MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)
+    movq        mm1, MMWORD MMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)
+    pfmul       mm6, MMWORD MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)
+    pfmul       mm1, MMWORD MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)
+    movq        mm3, MMWORD MMBLOCK(1,2,esi,SIZEOF_FAST_FLOAT)
+    movq        mm4, MMWORD MMBLOCK(1,3,esi,SIZEOF_FAST_FLOAT)
+    pfmul       mm3, MMWORD MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)
+    pfmul       mm4, MMWORD MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)
+
+    pfadd       mm6, mm7                ; mm0=(10 ** 11 **)
+    pfadd       mm1, mm7                ; mm4=(12 ** 13 **)
+    pfadd       mm3, mm7                ; mm0=(14 ** 15 **)
+    pfadd       mm4, mm7                ; mm4=(16 ** 17 **)
+
+    movq        mm5, mm6
+    punpcklwd   mm6, mm1                ; mm6=(10 12 ** **)
+    punpckhwd   mm5, mm1                ; mm5=(11 13 ** **)
+    movq        mm1, mm3
+    punpcklwd   mm3, mm4                ; mm3=(14 16 ** **)
+    punpckhwd   mm1, mm4                ; mm1=(15 17 ** **)
+
+    punpcklwd   mm6, mm5                ; mm6=(10 11 12 13)
+    punpcklwd   mm3, mm1                ; mm3=(14 15 16 17)
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_JCOEF), mm2
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_JCOEF), mm6
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_JCOEF), mm3
+
+    add         esi, byte 16*SIZEOF_FAST_FLOAT
+    add         edx, byte 16*SIZEOF_FAST_FLOAT
+    add         edi, byte 16*SIZEOF_JCOEF
+    dec         eax
+    jnz         near .quantloop
+
+    femms                               ; empty MMX/3DNow! state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jquant-mmx.asm Added

@@ -0,0 +1,276 @@
+;
+; jquant.asm - sample data conversion and quantization (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_mmx(JSAMPARRAY sample_data, JDIMENSION start_col,
+;                    DCTELEM *workspace);
+;
+
+%define sample_data  ebp + 8            ; JSAMPARRAY sample_data
+%define start_col    ebp + 12           ; JDIMENSION start_col
+%define workspace    ebp + 16           ; DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_convsamp_mmx)
+
+EXTN(jsimd_convsamp_mmx):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    pxor        mm6, mm6                ; mm6=(all 0's)
+    pcmpeqw     mm7, mm7
+    psllw       mm7, 7                  ; mm7={0xFF80 0xFF80 0xFF80 0xFF80}
+
+    mov         esi, JSAMPARRAY sample_data  ; (JSAMPROW *)
+    mov         eax, JDIMENSION start_col
+    mov         edi, POINTER workspace       ; (DCTELEM *)
+    mov         ecx, DCTSIZE/4
+    alignx      16, 7
+.convloop:
+    mov         ebx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        mm0, MMWORD ebx+eax*SIZEOF_JSAMPLE   ; mm0=(01234567)
+    movq        mm1, MMWORD edx+eax*SIZEOF_JSAMPLE   ; mm1=(89ABCDEF)
+
+    mov         ebx, JSAMPROW esi+2*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+3*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        mm2, MMWORD ebx+eax*SIZEOF_JSAMPLE   ; mm2=(GHIJKLMN)
+    movq        mm3, MMWORD edx+eax*SIZEOF_JSAMPLE   ; mm3=(OPQRSTUV)
+
+    movq        mm4, mm0
+    punpcklbw   mm0, mm6                ; mm0=(0123)
+    punpckhbw   mm4, mm6                ; mm4=(4567)
+    movq        mm5, mm1
+    punpcklbw   mm1, mm6                ; mm1=(89AB)
+    punpckhbw   mm5, mm6                ; mm5=(CDEF)
+
+    paddw       mm0, mm7
+    paddw       mm4, mm7
+    paddw       mm1, mm7
+    paddw       mm5, mm7
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_DCTELEM), mm0
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_DCTELEM), mm4
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_DCTELEM), mm1
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_DCTELEM), mm5
+
+    movq        mm0, mm2
+    punpcklbw   mm2, mm6                ; mm2=(GHIJ)
+    punpckhbw   mm0, mm6                ; mm0=(KLMN)
+    movq        mm4, mm3
+    punpcklbw   mm3, mm6                ; mm3=(OPQR)
+    punpckhbw   mm4, mm6                ; mm4=(STUV)
+
+    paddw       mm2, mm7
+    paddw       mm0, mm7
+    paddw       mm3, mm7
+    paddw       mm4, mm7
+
+    movq        MMWORD MMBLOCK(2,0,edi,SIZEOF_DCTELEM), mm2
+    movq        MMWORD MMBLOCK(2,1,edi,SIZEOF_DCTELEM), mm0
+    movq        MMWORD MMBLOCK(3,0,edi,SIZEOF_DCTELEM), mm3
+    movq        MMWORD MMBLOCK(3,1,edi,SIZEOF_DCTELEM), mm4
+
+    add         esi, byte 4*SIZEOF_JSAMPROW
+    add         edi, byte 4*DCTSIZE*SIZEOF_DCTELEM
+    dec         ecx
+    jnz         short .convloop
+
+    emms                                ; empty MMX state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_mmx(JCOEFPTR coef_block, DCTELEM *divisors,
+;                    DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m, n, b) \
+  MMBLOCK(DCTSIZE * 0 + (m), (n), (b), SIZEOF_DCTELEM)
+%define CORRECTION(m, n, b) \
+  MMBLOCK(DCTSIZE * 1 + (m), (n), (b), SIZEOF_DCTELEM)
+%define SCALE(m, n, b) \
+  MMBLOCK(DCTSIZE * 2 + (m), (n), (b), SIZEOF_DCTELEM)
+%define SHIFT(m, n, b) \
+  MMBLOCK(DCTSIZE * 3 + (m), (n), (b), SIZEOF_DCTELEM)
+
+%define coef_block  ebp + 8             ; JCOEFPTR coef_block
+%define divisors    ebp + 12            ; DCTELEM *divisors
+%define workspace   ebp + 16            ; DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_quantize_mmx)
+
+EXTN(jsimd_quantize_mmx):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         esi, POINTER workspace
+    mov         edx, POINTER divisors
+    mov         edi, JCOEFPTR coef_block
+    mov         ah, 2
+    alignx      16, 7
+.quantloop1:
+    mov         al, DCTSIZE2/8/2
+    alignx      16, 7
+.quantloop2:
+    movq        mm2, MMWORD MMBLOCK(0,0,esi,SIZEOF_DCTELEM)
+    movq        mm3, MMWORD MMBLOCK(0,1,esi,SIZEOF_DCTELEM)
+
+    movq        mm0, mm2
+    movq        mm1, mm3
+
+    psraw       mm2, (WORD_BIT-1)       ; -1 if value < 0, 0 otherwise
+    psraw       mm3, (WORD_BIT-1)
+
+    pxor        mm0, mm2                ; val = -val
+    pxor        mm1, mm3
+    psubw       mm0, mm2
+    psubw       mm1, mm3
+
+    ;
+    ; MMX is an annoyingly crappy instruction set. It has two
+    ; misfeatures that are causing problems here:
+    ;
+    ; - All multiplications are signed.
+    ;
+    ; - The second operand for the shifts is not treated as packed.
+    ;
+    ;
+    ; We work around the first problem by implementing this algorithm:
+    ;
+    ; unsigned long unsigned_multiply(unsigned short x, unsigned short y)
+    ; {
+    ;   enum { SHORT_BIT = 16 };
+    ;   signed short sx = (signed short)x;
+    ;   signed short sy = (signed short)y;
+    ;   signed long sz;
+    ;
+    ;   sz = (long)sx * (long)sy;    /* signed multiply */
+    ;
+    ;   if (sx < 0) sz += (long)sy << SHORT_BIT;
+    ;   if (sy < 0) sz += (long)sx << SHORT_BIT;
+    ;
+    ;   return (unsigned long)sz;
+    ; }
+    ;
+    ; (note that a negative sx adds _sy_ and vice versa)
+    ;
+    ; For the second problem, we replace the shift by a multiplication.
+    ; Unfortunately that means we have to deal with the signed issue again.
+    ;
+
+    paddw       mm0, MMWORD CORRECTION(0,0,edx)  ; correction + roundfactor
+    paddw       mm1, MMWORD CORRECTION(0,1,edx)
+
+    movq        mm4, mm0                ; store current value for later
+    movq        mm5, mm1
+    pmulhw      mm0, MMWORD RECIPROCAL(0,0,edx)  ; reciprocal
+    pmulhw      mm1, MMWORD RECIPROCAL(0,1,edx)
+    paddw       mm0, mm4  ; reciprocal is always negative (MSB=1),
+    paddw       mm1, mm5  ; so we always need to add the initial value
+                          ; (input value is never negative as we
+                          ; inverted it at the start of this routine)
+
+    ; here it gets a bit tricky as both scale
+    ; and mm0/mm1 can be negative
+    movq        mm6, MMWORD SCALE(0,0,edx)  ; scale
+    movq        mm7, MMWORD SCALE(0,1,edx)
+    movq        mm4, mm0
+    movq        mm5, mm1
+    pmulhw      mm0, mm6
+    pmulhw      mm1, mm7
+
+    psraw       mm6, (WORD_BIT-1)       ; determine if scale is negative
+    psraw       mm7, (WORD_BIT-1)
+
+    pand        mm6, mm4                ; and add input if it is
+    pand        mm7, mm5
+    paddw       mm0, mm6
+    paddw       mm1, mm7
+
+    psraw       mm4, (WORD_BIT-1)       ; then check if negative input
+    psraw       mm5, (WORD_BIT-1)
+
+    pand        mm4, MMWORD SCALE(0,0,edx)  ; and add scale if it is
+    pand        mm5, MMWORD SCALE(0,1,edx)
+    paddw       mm0, mm4
+    paddw       mm1, mm5
+
+    pxor        mm0, mm2                ; val = -val
+    pxor        mm1, mm3
+    psubw       mm0, mm2
+    psubw       mm1, mm3
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_DCTELEM), mm0
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_DCTELEM), mm1
+
+    add         esi, byte 8*SIZEOF_DCTELEM
+    add         edx, byte 8*SIZEOF_DCTELEM
+    add         edi, byte 8*SIZEOF_JCOEF
+    dec         al
+    jnz         near .quantloop2
+    dec         ah
+    jnz         near .quantloop1        ; to avoid branch misprediction
+
+    emms                                ; empty MMX state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jquant-sse.asm Added

@@ -0,0 +1,208 @@
+;
+; jquant.asm - sample data conversion and quantization (SSE & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_sse(JSAMPARRAY sample_data, JDIMENSION start_col,
+;                          FAST_FLOAT *workspace);
+;
+
+%define sample_data  ebp + 8            ; JSAMPARRAY sample_data
+%define start_col    ebp + 12           ; JDIMENSION start_col
+%define workspace    ebp + 16           ; FAST_FLOAT *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_convsamp_float_sse)
+
+EXTN(jsimd_convsamp_float_sse):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    pcmpeqw     mm7, mm7
+    psllw       mm7, 7
+    packsswb    mm7, mm7                ; mm7 = PB_CENTERJSAMPLE (0x808080..)
+
+    mov         esi, JSAMPARRAY sample_data  ; (JSAMPROW *)
+    mov         eax, JDIMENSION start_col
+    mov         edi, POINTER workspace       ; (DCTELEM *)
+    mov         ecx, DCTSIZE/2
+    alignx      16, 7
+.convloop:
+    mov         ebx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        mm0, MMWORD ebx+eax*SIZEOF_JSAMPLE
+    movq        mm1, MMWORD edx+eax*SIZEOF_JSAMPLE
+
+    psubb       mm0, mm7                ; mm0=(01234567)
+    psubb       mm1, mm7                ; mm1=(89ABCDEF)
+
+    punpcklbw   mm2, mm0                ; mm2=(*0*1*2*3)
+    punpckhbw   mm0, mm0                ; mm0=(*4*5*6*7)
+    punpcklbw   mm3, mm1                ; mm3=(*8*9*A*B)
+    punpckhbw   mm1, mm1                ; mm1=(*C*D*E*F)
+
+    punpcklwd   mm4, mm2                ; mm4=(***0***1)
+    punpckhwd   mm2, mm2                ; mm2=(***2***3)
+    punpcklwd   mm5, mm0                ; mm5=(***4***5)
+    punpckhwd   mm0, mm0                ; mm0=(***6***7)
+
+    psrad       mm4, (DWORD_BIT-BYTE_BIT)  ; mm4=(01)
+    psrad       mm2, (DWORD_BIT-BYTE_BIT)  ; mm2=(23)
+    cvtpi2ps    xmm0, mm4                  ; xmm0=(01**)
+    cvtpi2ps    xmm1, mm2                  ; xmm1=(23**)
+    psrad       mm5, (DWORD_BIT-BYTE_BIT)  ; mm5=(45)
+    psrad       mm0, (DWORD_BIT-BYTE_BIT)  ; mm0=(67)
+    cvtpi2ps    xmm2, mm5                  ; xmm2=(45**)
+    cvtpi2ps    xmm3, mm0                  ; xmm3=(67**)
+
+    punpcklwd   mm6, mm3                ; mm6=(***8***9)
+    punpckhwd   mm3, mm3                ; mm3=(***A***B)
+    punpcklwd   mm4, mm1                ; mm4=(***C***D)
+    punpckhwd   mm1, mm1                ; mm1=(***E***F)
+
+    psrad       mm6, (DWORD_BIT-BYTE_BIT)  ; mm6=(89)
+    psrad       mm3, (DWORD_BIT-BYTE_BIT)  ; mm3=(AB)
+    cvtpi2ps    xmm4, mm6                  ; xmm4=(89**)
+    cvtpi2ps    xmm5, mm3                  ; xmm5=(AB**)
+    psrad       mm4, (DWORD_BIT-BYTE_BIT)  ; mm4=(CD)
+    psrad       mm1, (DWORD_BIT-BYTE_BIT)  ; mm1=(EF)
+    cvtpi2ps    xmm6, mm4                  ; xmm6=(CD**)
+    cvtpi2ps    xmm7, mm1                  ; xmm7=(EF**)
+
+    movlhps     xmm0, xmm1              ; xmm0=(0123)
+    movlhps     xmm2, xmm3              ; xmm2=(4567)
+    movlhps     xmm4, xmm5              ; xmm4=(89AB)
+    movlhps     xmm6, xmm7              ; xmm6=(CDEF)
+
+    movaps      XMMWORD XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT), xmm2
+    movaps      XMMWORD XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT), xmm4
+    movaps      XMMWORD XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT), xmm6
+
+    add         esi, byte 2*SIZEOF_JSAMPROW
+    add         edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+    dec         ecx
+    jnz         near .convloop
+
+    emms                                ; empty MMX state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_sse(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                          FAST_FLOAT *workspace);
+;
+
+%define coef_block  ebp + 8             ; JCOEFPTR coef_block
+%define divisors    ebp + 12            ; FAST_FLOAT *divisors
+%define workspace   ebp + 16            ; FAST_FLOAT *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_quantize_float_sse)
+
+EXTN(jsimd_quantize_float_sse):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         esi, POINTER workspace
+    mov         edx, POINTER divisors
+    mov         edi, JCOEFPTR coef_block
+    mov         eax, DCTSIZE2/16
+    alignx      16, 7
+.quantloop:
+    movaps      xmm0, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)
+    mulps       xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)
+    mulps       xmm1, XMMWORD XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)
+    mulps       xmm2, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)
+    mulps       xmm3, XMMWORD XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)
+
+    movhlps     xmm4, xmm0
+    movhlps     xmm5, xmm1
+
+    cvtps2pi    mm0, xmm0
+    cvtps2pi    mm1, xmm1
+    cvtps2pi    mm4, xmm4
+    cvtps2pi    mm5, xmm5
+
+    movhlps     xmm6, xmm2
+    movhlps     xmm7, xmm3
+
+    cvtps2pi    mm2, xmm2
+    cvtps2pi    mm3, xmm3
+    cvtps2pi    mm6, xmm6
+    cvtps2pi    mm7, xmm7
+
+    packssdw    mm0, mm4
+    packssdw    mm1, mm5
+    packssdw    mm2, mm6
+    packssdw    mm3, mm7
+
+    movq        MMWORD MMBLOCK(0,0,edi,SIZEOF_JCOEF), mm0
+    movq        MMWORD MMBLOCK(0,1,edi,SIZEOF_JCOEF), mm1
+    movq        MMWORD MMBLOCK(1,0,edi,SIZEOF_JCOEF), mm2
+    movq        MMWORD MMBLOCK(1,1,edi,SIZEOF_JCOEF), mm3
+
+    add         esi, byte 16*SIZEOF_FAST_FLOAT
+    add         edx, byte 16*SIZEOF_FAST_FLOAT
+    add         edi, byte 16*SIZEOF_JCOEF
+    dec         eax
+    jnz         short .quantloop
+
+    emms                                ; empty MMX state
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jquantf-sse2.asm Added

@@ -0,0 +1,168 @@
+;
+; jquantf.asm - sample data conversion and quantization (SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_sse2(JSAMPARRAY sample_data, JDIMENSION start_col,
+;                           FAST_FLOAT *workspace);
+;
+
+%define sample_data  ebp + 8            ; JSAMPARRAY sample_data
+%define start_col    ebp + 12           ; JDIMENSION start_col
+%define workspace    ebp + 16           ; FAST_FLOAT *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_convsamp_float_sse2)
+
+EXTN(jsimd_convsamp_float_sse2):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    pcmpeqw     xmm7, xmm7
+    psllw       xmm7, 7
+    packsswb    xmm7, xmm7              ; xmm7 = PB_CENTERJSAMPLE (0x808080..)
+
+    mov         esi, JSAMPARRAY sample_data  ; (JSAMPROW *)
+    mov         eax, JDIMENSION start_col
+    mov         edi, POINTER workspace       ; (DCTELEM *)
+    mov         ecx, DCTSIZE/2
+    alignx      16, 7
+.convloop:
+    mov         ebx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        xmm0, XMM_MMWORD ebx+eax*SIZEOF_JSAMPLE
+    movq        xmm1, XMM_MMWORD edx+eax*SIZEOF_JSAMPLE
+
+    psubb       xmm0, xmm7              ; xmm0=(01234567)
+    psubb       xmm1, xmm7              ; xmm1=(89ABCDEF)
+
+    punpcklbw   xmm0, xmm0              ; xmm0=(*0*1*2*3*4*5*6*7)
+    punpcklbw   xmm1, xmm1              ; xmm1=(*8*9*A*B*C*D*E*F)
+
+    punpcklwd   xmm2, xmm0              ; xmm2=(***0***1***2***3)
+    punpckhwd   xmm0, xmm0              ; xmm0=(***4***5***6***7)
+    punpcklwd   xmm3, xmm1              ; xmm3=(***8***9***A***B)
+    punpckhwd   xmm1, xmm1              ; xmm1=(***C***D***E***F)
+
+    psrad       xmm2, (DWORD_BIT-BYTE_BIT)  ; xmm2=(0123)
+    psrad       xmm0, (DWORD_BIT-BYTE_BIT)  ; xmm0=(4567)
+    cvtdq2ps    xmm2, xmm2                  ; xmm2=(0123)
+    cvtdq2ps    xmm0, xmm0                  ; xmm0=(4567)
+    psrad       xmm3, (DWORD_BIT-BYTE_BIT)  ; xmm3=(89AB)
+    psrad       xmm1, (DWORD_BIT-BYTE_BIT)  ; xmm1=(CDEF)
+    cvtdq2ps    xmm3, xmm3                  ; xmm3=(89AB)
+    cvtdq2ps    xmm1, xmm1                  ; xmm1=(CDEF)
+
+    movaps      XMMWORD XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT), xmm2
+    movaps      XMMWORD XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT), xmm3
+    movaps      XMMWORD XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT), xmm1
+
+    add         esi, byte 2*SIZEOF_JSAMPROW
+    add         edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+    dec         ecx
+    jnz         short .convloop
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_sse2(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                           FAST_FLOAT *workspace);
+;
+
+%define coef_block  ebp + 8             ; JCOEFPTR coef_block
+%define divisors    ebp + 12            ; FAST_FLOAT *divisors
+%define workspace   ebp + 16            ; FAST_FLOAT *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_quantize_float_sse2)
+
+EXTN(jsimd_quantize_float_sse2):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         esi, POINTER workspace
+    mov         edx, POINTER divisors
+    mov         edi, JCOEFPTR coef_block
+    mov         eax, DCTSIZE2/16
+    alignx      16, 7
+.quantloop:
+    movaps      xmm0, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)
+    mulps       xmm0, XMMWORD XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)
+    mulps       xmm1, XMMWORD XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)
+    mulps       xmm2, XMMWORD XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)
+    mulps       xmm3, XMMWORD XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)
+
+    cvtps2dq    xmm0, xmm0
+    cvtps2dq    xmm1, xmm1
+    cvtps2dq    xmm2, xmm2
+    cvtps2dq    xmm3, xmm3
+
+    packssdw    xmm0, xmm1
+    packssdw    xmm2, xmm3
+
+    movdqa      XMMWORD XMMBLOCK(0,0,edi,SIZEOF_JCOEF), xmm0
+    movdqa      XMMWORD XMMBLOCK(1,0,edi,SIZEOF_JCOEF), xmm2
+
+    add         esi, byte 16*SIZEOF_FAST_FLOAT
+    add         edx, byte 16*SIZEOF_FAST_FLOAT
+    add         edi, byte 16*SIZEOF_JCOEF
+    dec         eax
+    jnz         short .quantloop
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jquanti-avx2.asm Added

@@ -0,0 +1,188 @@
+;
+; jquanti.asm - sample data conversion and quantization (AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, 2018, D. R. Commander.
+; Copyright (C) 2016, Matthieu Darbois.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_avx2(JSAMPARRAY sample_data, JDIMENSION start_col,
+;                     DCTELEM *workspace);
+;
+
+%define sample_data  ebp + 8            ; JSAMPARRAY sample_data
+%define start_col    ebp + 12           ; JDIMENSION start_col
+%define workspace    ebp + 16           ; DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_convsamp_avx2)
+
+EXTN(jsimd_convsamp_avx2):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         esi, JSAMPARRAY sample_data  ; (JSAMPROW *)
+    mov         eax, JDIMENSION start_col
+    mov         edi, POINTER workspace       ; (DCTELEM *)
+
+    mov         ebx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        xmm0, XMM_MMWORD ebx+eax*SIZEOF_JSAMPLE
+    movq        xmm1, XMM_MMWORD edx+eax*SIZEOF_JSAMPLE
+
+    mov         ebx, JSAMPROW esi+2*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+3*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        xmm2, XMM_MMWORD ebx+eax*SIZEOF_JSAMPLE
+    movq        xmm3, XMM_MMWORD edx+eax*SIZEOF_JSAMPLE
+
+    mov         ebx, JSAMPROW esi+4*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+5*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        xmm4, XMM_MMWORD ebx+eax*SIZEOF_JSAMPLE
+    movq        xmm5, XMM_MMWORD edx+eax*SIZEOF_JSAMPLE
+
+    mov         ebx, JSAMPROW esi+6*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+7*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        xmm6, XMM_MMWORD ebx+eax*SIZEOF_JSAMPLE
+    movq        xmm7, XMM_MMWORD edx+eax*SIZEOF_JSAMPLE
+
+    vinserti128 ymm0, ymm0, xmm1, 1
+    vinserti128 ymm2, ymm2, xmm3, 1
+    vinserti128 ymm4, ymm4, xmm5, 1
+    vinserti128 ymm6, ymm6, xmm7, 1
+
+    vpxor       ymm1, ymm1, ymm1        ; ymm1=(all 0's)
+    vpunpcklbw  ymm0, ymm0, ymm1
+    vpunpcklbw  ymm2, ymm2, ymm1
+    vpunpcklbw  ymm4, ymm4, ymm1
+    vpunpcklbw  ymm6, ymm6, ymm1
+
+    vpcmpeqw    ymm7, ymm7, ymm7
+    vpsllw      ymm7, ymm7, 7           ; ymm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    vpaddw      ymm0, ymm0, ymm7
+    vpaddw      ymm2, ymm2, ymm7
+    vpaddw      ymm4, ymm4, ymm7
+    vpaddw      ymm6, ymm6, ymm7
+
+    vmovdqu     YMMWORD YMMBLOCK(0,0,edi,SIZEOF_DCTELEM), ymm0
+    vmovdqu     YMMWORD YMMBLOCK(2,0,edi,SIZEOF_DCTELEM), ymm2
+    vmovdqu     YMMWORD YMMBLOCK(4,0,edi,SIZEOF_DCTELEM), ymm4
+    vmovdqu     YMMWORD YMMBLOCK(6,0,edi,SIZEOF_DCTELEM), ymm6
+
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_avx2(JCOEFPTR coef_block, DCTELEM *divisors,
+;                     DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m, n, b) \
+  YMMBLOCK(DCTSIZE * 0 + (m), (n), (b), SIZEOF_DCTELEM)
+%define CORRECTION(m, n, b) \
+  YMMBLOCK(DCTSIZE * 1 + (m), (n), (b), SIZEOF_DCTELEM)
+%define SCALE(m, n, b) \
+  YMMBLOCK(DCTSIZE * 2 + (m), (n), (b), SIZEOF_DCTELEM)
+
+%define coef_block  ebp + 8             ; JCOEFPTR coef_block
+%define divisors    ebp + 12            ; DCTELEM *divisors
+%define workspace   ebp + 16            ; DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_quantize_avx2)
+
+EXTN(jsimd_quantize_avx2):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         esi, POINTER workspace
+    mov         edx, POINTER divisors
+    mov         edi, JCOEFPTR coef_block
+
+    vmovdqu     ymm4, YMMBLOCK(0,0,esi,SIZEOF_DCTELEM)
+    vmovdqu     ymm5, YMMBLOCK(2,0,esi,SIZEOF_DCTELEM)
+    vmovdqu     ymm6, YMMBLOCK(4,0,esi,SIZEOF_DCTELEM)
+    vmovdqu     ymm7, YMMBLOCK(6,0,esi,SIZEOF_DCTELEM)
+    vpabsw      ymm0, ymm4
+    vpabsw      ymm1, ymm5
+    vpabsw      ymm2, ymm6
+    vpabsw      ymm3, ymm7
+
+    vpaddw      ymm0, YMMWORD CORRECTION(0,0,edx)  ; correction + roundfactor
+    vpaddw      ymm1, YMMWORD CORRECTION(2,0,edx)
+    vpaddw      ymm2, YMMWORD CORRECTION(4,0,edx)
+    vpaddw      ymm3, YMMWORD CORRECTION(6,0,edx)
+    vpmulhuw    ymm0, YMMWORD RECIPROCAL(0,0,edx)  ; reciprocal
+    vpmulhuw    ymm1, YMMWORD RECIPROCAL(2,0,edx)
+    vpmulhuw    ymm2, YMMWORD RECIPROCAL(4,0,edx)
+    vpmulhuw    ymm3, YMMWORD RECIPROCAL(6,0,edx)
+    vpmulhuw    ymm0, YMMWORD SCALE(0,0,edx)       ; scale
+    vpmulhuw    ymm1, YMMWORD SCALE(2,0,edx)
+    vpmulhuw    ymm2, YMMWORD SCALE(4,0,edx)
+    vpmulhuw    ymm3, YMMWORD SCALE(6,0,edx)
+
+    vpsignw     ymm0, ymm0, ymm4
+    vpsignw     ymm1, ymm1, ymm5
+    vpsignw     ymm2, ymm2, ymm6
+    vpsignw     ymm3, ymm3, ymm7
+
+    vmovdqu     YMMBLOCK(0,0,edi,SIZEOF_DCTELEM), ymm0
+    vmovdqu     YMMBLOCK(2,0,edi,SIZEOF_DCTELEM), ymm1
+    vmovdqu     YMMBLOCK(4,0,edi,SIZEOF_DCTELEM), ymm2
+    vmovdqu     YMMBLOCK(6,0,edi,SIZEOF_DCTELEM), ymm3
+
+    vzeroupper
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jquanti-sse2.asm Added

@@ -0,0 +1,201 @@
+;
+; jquanti.asm - sample data conversion and quantization (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_sse2(JSAMPARRAY sample_data, JDIMENSION start_col,
+;                     DCTELEM *workspace);
+;
+
+%define sample_data  ebp + 8            ; JSAMPARRAY sample_data
+%define start_col    ebp + 12           ; JDIMENSION start_col
+%define workspace    ebp + 16           ; DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_convsamp_sse2)
+
+EXTN(jsimd_convsamp_sse2):
+    push        ebp
+    mov         ebp, esp
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    pxor        xmm6, xmm6              ; xmm6=(all 0's)
+    pcmpeqw     xmm7, xmm7
+    psllw       xmm7, 7                 ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    mov         esi, JSAMPARRAY sample_data  ; (JSAMPROW *)
+    mov         eax, JDIMENSION start_col
+    mov         edi, POINTER workspace       ; (DCTELEM *)
+    mov         ecx, DCTSIZE/4
+    alignx      16, 7
+.convloop:
+    mov         ebx, JSAMPROW esi+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        xmm0, XMM_MMWORD ebx+eax*SIZEOF_JSAMPLE  ; xmm0=(01234567)
+    movq        xmm1, XMM_MMWORD edx+eax*SIZEOF_JSAMPLE  ; xmm1=(89ABCDEF)
+
+    mov         ebx, JSAMPROW esi+2*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         edx, JSAMPROW esi+3*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        xmm2, XMM_MMWORD ebx+eax*SIZEOF_JSAMPLE  ; xmm2=(GHIJKLMN)
+    movq        xmm3, XMM_MMWORD edx+eax*SIZEOF_JSAMPLE  ; xmm3=(OPQRSTUV)
+
+    punpcklbw   xmm0, xmm6              ; xmm0=(01234567)
+    punpcklbw   xmm1, xmm6              ; xmm1=(89ABCDEF)
+    paddw       xmm0, xmm7
+    paddw       xmm1, xmm7
+    punpcklbw   xmm2, xmm6              ; xmm2=(GHIJKLMN)
+    punpcklbw   xmm3, xmm6              ; xmm3=(OPQRSTUV)
+    paddw       xmm2, xmm7
+    paddw       xmm3, xmm7
+
+    movdqa      XMMWORD XMMBLOCK(0,0,edi,SIZEOF_DCTELEM), xmm0
+    movdqa      XMMWORD XMMBLOCK(1,0,edi,SIZEOF_DCTELEM), xmm1
+    movdqa      XMMWORD XMMBLOCK(2,0,edi,SIZEOF_DCTELEM), xmm2
+    movdqa      XMMWORD XMMBLOCK(3,0,edi,SIZEOF_DCTELEM), xmm3
+
+    add         esi, byte 4*SIZEOF_JSAMPROW
+    add         edi, byte 4*DCTSIZE*SIZEOF_DCTELEM
+    dec         ecx
+    jnz         short .convloop
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    pop         ebp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_sse2(JCOEFPTR coef_block, DCTELEM *divisors,
+;                     DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m, n, b) \
+  XMMBLOCK(DCTSIZE * 0 + (m), (n), (b), SIZEOF_DCTELEM)
+%define CORRECTION(m, n, b) \
+  XMMBLOCK(DCTSIZE * 1 + (m), (n), (b), SIZEOF_DCTELEM)
+%define SCALE(m, n, b) \
+  XMMBLOCK(DCTSIZE * 2 + (m), (n), (b), SIZEOF_DCTELEM)
+
+%define coef_block  ebp + 8             ; JCOEFPTR coef_block
+%define divisors    ebp + 12            ; DCTELEM *divisors
+%define workspace   ebp + 16            ; DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_quantize_sse2)
+
+EXTN(jsimd_quantize_sse2):
+    push        ebp
+    mov         ebp, esp
+;   push        ebx                     ; unused
+;   push        ecx                     ; unused
+;   push        edx                     ; need not be preserved
+    push        esi
+    push        edi
+
+    mov         esi, POINTER workspace
+    mov         edx, POINTER divisors
+    mov         edi, JCOEFPTR coef_block
+    mov         eax, DCTSIZE2/32
+    alignx      16, 7
+.quantloop:
+    movdqa      xmm4, XMMWORD XMMBLOCK(0,0,esi,SIZEOF_DCTELEM)
+    movdqa      xmm5, XMMWORD XMMBLOCK(1,0,esi,SIZEOF_DCTELEM)
+    movdqa      xmm6, XMMWORD XMMBLOCK(2,0,esi,SIZEOF_DCTELEM)
+    movdqa      xmm7, XMMWORD XMMBLOCK(3,0,esi,SIZEOF_DCTELEM)
+    movdqa      xmm0, xmm4
+    movdqa      xmm1, xmm5
+    movdqa      xmm2, xmm6
+    movdqa      xmm3, xmm7
+    psraw       xmm4, (WORD_BIT-1)
+    psraw       xmm5, (WORD_BIT-1)
+    psraw       xmm6, (WORD_BIT-1)
+    psraw       xmm7, (WORD_BIT-1)
+    pxor        xmm0, xmm4
+    pxor        xmm1, xmm5
+    pxor        xmm2, xmm6
+    pxor        xmm3, xmm7
+    psubw       xmm0, xmm4              ; if (xmm0 < 0) xmm0 = -xmm0;
+    psubw       xmm1, xmm5              ; if (xmm1 < 0) xmm1 = -xmm1;
+    psubw       xmm2, xmm6              ; if (xmm2 < 0) xmm2 = -xmm2;
+    psubw       xmm3, xmm7              ; if (xmm3 < 0) xmm3 = -xmm3;
+
+    paddw       xmm0, XMMWORD CORRECTION(0,0,edx)  ; correction + roundfactor
+    paddw       xmm1, XMMWORD CORRECTION(1,0,edx)
+    paddw       xmm2, XMMWORD CORRECTION(2,0,edx)
+    paddw       xmm3, XMMWORD CORRECTION(3,0,edx)
+    pmulhuw     xmm0, XMMWORD RECIPROCAL(0,0,edx)  ; reciprocal
+    pmulhuw     xmm1, XMMWORD RECIPROCAL(1,0,edx)
+    pmulhuw     xmm2, XMMWORD RECIPROCAL(2,0,edx)
+    pmulhuw     xmm3, XMMWORD RECIPROCAL(3,0,edx)
+    pmulhuw     xmm0, XMMWORD SCALE(0,0,edx)       ; scale
+    pmulhuw     xmm1, XMMWORD SCALE(1,0,edx)
+    pmulhuw     xmm2, XMMWORD SCALE(2,0,edx)
+    pmulhuw     xmm3, XMMWORD SCALE(3,0,edx)
+
+    pxor        xmm0, xmm4
+    pxor        xmm1, xmm5
+    pxor        xmm2, xmm6
+    pxor        xmm3, xmm7
+    psubw       xmm0, xmm4
+    psubw       xmm1, xmm5
+    psubw       xmm2, xmm6
+    psubw       xmm3, xmm7
+    movdqa      XMMWORD XMMBLOCK(0,0,edi,SIZEOF_DCTELEM), xmm0
+    movdqa      XMMWORD XMMBLOCK(1,0,edi,SIZEOF_DCTELEM), xmm1
+    movdqa      XMMWORD XMMBLOCK(2,0,edi,SIZEOF_DCTELEM), xmm2
+    movdqa      XMMWORD XMMBLOCK(3,0,edi,SIZEOF_DCTELEM), xmm3
+
+    add         esi, byte 32*SIZEOF_DCTELEM
+    add         edx, byte 32*SIZEOF_DCTELEM
+    add         edi, byte 32*SIZEOF_JCOEF
+    dec         eax
+    jnz         near .quantloop
+
+    pop         edi
+    pop         esi
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; unused
+;   pop         ebx                     ; unused
+    pop         ebp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jsimd.c Added

@@ -0,0 +1,1253 @@
+/*
+ * jsimd_i386.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2009-2011, 2013-2014, 2016, 2018, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 32-bit x86 architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "jconfigint.h"
+
+/*
+ * In the PIC cases, we have no guarantee that constants will keep
+ * their alignment. This macro allows us to verify it at runtime.
+ */
+#define IS_ALIGNED(ptr, order)  (((unsigned)ptr & ((1 << order) - 1)) == 0)
+
+#define IS_ALIGNED_SSE(ptr)  (IS_ALIGNED(ptr, 4)) /* 16 byte alignment */
+#define IS_ALIGNED_AVX(ptr)  (IS_ALIGNED(ptr, 5)) /* 32 byte alignment */
+
+static unsigned int simd_support = (unsigned int)(~0);
+static unsigned int simd_huffman = 1;
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd(void)
+{
+#ifndef NO_GETENV
+  char *env = NULL;
+#endif
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = jpeg_simd_cpu_support();
+
+#ifndef NO_GETENV
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCEMMX");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_MMX;
+  env = getenv("JSIMD_FORCE3DNOW");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_3DNOW | JSIMD_MMX;
+  env = getenv("JSIMD_FORCESSE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_SSE | JSIMD_MMX;
+  env = getenv("JSIMD_FORCESSE2");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_SSE2;
+  env = getenv("JSIMD_FORCEAVX2");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_AVX2;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+#endif
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_rgb_ycc_convert_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_rgb_gray_convert_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_ycc_rgb_convert_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                      JSAMPIMAGE output_buf, JDIMENSION output_row,
+                      int num_rows)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*sse2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*mmxfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_extrgb_ycc_convert_avx2;
+    sse2fct = jsimd_extrgb_ycc_convert_sse2;
+    mmxfct = jsimd_extrgb_ycc_convert_mmx;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_extrgbx_ycc_convert_avx2;
+    sse2fct = jsimd_extrgbx_ycc_convert_sse2;
+    mmxfct = jsimd_extrgbx_ycc_convert_mmx;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_extbgr_ycc_convert_avx2;
+    sse2fct = jsimd_extbgr_ycc_convert_sse2;
+    mmxfct = jsimd_extbgr_ycc_convert_mmx;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_extbgrx_ycc_convert_avx2;
+    sse2fct = jsimd_extbgrx_ycc_convert_sse2;
+    mmxfct = jsimd_extbgrx_ycc_convert_mmx;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_extxbgr_ycc_convert_avx2;
+    sse2fct = jsimd_extxbgr_ycc_convert_sse2;
+    mmxfct = jsimd_extxbgr_ycc_convert_mmx;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_extxrgb_ycc_convert_avx2;
+    sse2fct = jsimd_extxrgb_ycc_convert_sse2;
+    mmxfct = jsimd_extxrgb_ycc_convert_mmx;
+    break;
+  default:
+    avx2fct = jsimd_rgb_ycc_convert_avx2;
+    sse2fct = jsimd_rgb_ycc_convert_sse2;
+    mmxfct = jsimd_rgb_ycc_convert_mmx;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else if (simd_support & JSIMD_SSE2)
+    sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else
+    mmxfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                       JSAMPIMAGE output_buf, JDIMENSION output_row,
+                       int num_rows)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*sse2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*mmxfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_extrgb_gray_convert_avx2;
+    sse2fct = jsimd_extrgb_gray_convert_sse2;
+    mmxfct = jsimd_extrgb_gray_convert_mmx;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_extrgbx_gray_convert_avx2;
+    sse2fct = jsimd_extrgbx_gray_convert_sse2;
+    mmxfct = jsimd_extrgbx_gray_convert_mmx;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_extbgr_gray_convert_avx2;
+    sse2fct = jsimd_extbgr_gray_convert_sse2;
+    mmxfct = jsimd_extbgr_gray_convert_mmx;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_extbgrx_gray_convert_avx2;
+    sse2fct = jsimd_extbgrx_gray_convert_sse2;
+    mmxfct = jsimd_extbgrx_gray_convert_mmx;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_extxbgr_gray_convert_avx2;
+    sse2fct = jsimd_extxbgr_gray_convert_sse2;
+    mmxfct = jsimd_extxbgr_gray_convert_mmx;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_extxrgb_gray_convert_avx2;
+    sse2fct = jsimd_extxrgb_gray_convert_sse2;
+    mmxfct = jsimd_extxrgb_gray_convert_mmx;
+    break;
+  default:
+    avx2fct = jsimd_rgb_gray_convert_avx2;
+    sse2fct = jsimd_rgb_gray_convert_sse2;
+    mmxfct = jsimd_rgb_gray_convert_mmx;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else if (simd_support & JSIMD_SSE2)
+    sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else
+    mmxfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION input_row, JSAMPARRAY output_buf,
+                      int num_rows)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+  void (*sse2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+  void (*mmxfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_ycc_extrgb_convert_avx2;
+    sse2fct = jsimd_ycc_extrgb_convert_sse2;
+    mmxfct = jsimd_ycc_extrgb_convert_mmx;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_ycc_extrgbx_convert_avx2;
+    sse2fct = jsimd_ycc_extrgbx_convert_sse2;
+    mmxfct = jsimd_ycc_extrgbx_convert_mmx;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_ycc_extbgr_convert_avx2;
+    sse2fct = jsimd_ycc_extbgr_convert_sse2;
+    mmxfct = jsimd_ycc_extbgr_convert_mmx;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_ycc_extbgrx_convert_avx2;
+    sse2fct = jsimd_ycc_extbgrx_convert_sse2;
+    mmxfct = jsimd_ycc_extbgrx_convert_mmx;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_ycc_extxbgr_convert_avx2;
+    sse2fct = jsimd_ycc_extxbgr_convert_sse2;
+    mmxfct = jsimd_ycc_extxbgr_convert_mmx;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_ycc_extxrgb_convert_avx2;
+    sse2fct = jsimd_ycc_extxrgb_convert_sse2;
+    mmxfct = jsimd_ycc_extxrgb_convert_mmx;
+    break;
+  default:
+    avx2fct = jsimd_ycc_rgb_convert_avx2;
+    sse2fct = jsimd_ycc_rgb_convert_sse2;
+    mmxfct = jsimd_ycc_rgb_convert_mmx;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+  else if (simd_support & JSIMD_SSE2)
+    sse2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+  else
+    mmxfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v2_downsample_avx2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_h2v2_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else
+    jsimd_h2v2_downsample_mmx(cinfo->image_width, cinfo->max_v_samp_factor,
+                              compptr->v_samp_factor, compptr->width_in_blocks,
+                              input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v1_downsample_avx2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_h2v1_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else
+    jsimd_h2v1_downsample_mmx(cinfo->image_width, cinfo->max_v_samp_factor,
+                              compptr->v_samp_factor, compptr->width_in_blocks,
+                              input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v2_upsample_avx2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_h2v2_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else
+    jsimd_h2v2_upsample_mmx(cinfo->max_v_samp_factor, cinfo->output_width,
+                            input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v1_upsample_avx2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_h2v1_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else
+    jsimd_h2v1_upsample_mmx(cinfo->max_v_samp_factor, cinfo->output_width,
+                            input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_fancy_upsample_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_fancy_upsample_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v2_fancy_upsample_avx2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_h2v2_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else
+    jsimd_h2v2_fancy_upsample_mmx(cinfo->max_v_samp_factor,
+                                  compptr->downsampled_width, input_data,
+                                  output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v1_fancy_upsample_avx2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_h2v1_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else
+    jsimd_h2v1_fancy_upsample_mmx(cinfo->max_v_samp_factor,
+                                  compptr->downsampled_width, input_data,
+                                  output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_merged_upsample_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_merged_upsample_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*sse2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*mmxfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_h2v2_extrgb_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extrgb_merged_upsample_sse2;
+    mmxfct = jsimd_h2v2_extrgb_merged_upsample_mmx;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_h2v2_extrgbx_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extrgbx_merged_upsample_sse2;
+    mmxfct = jsimd_h2v2_extrgbx_merged_upsample_mmx;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_h2v2_extbgr_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extbgr_merged_upsample_sse2;
+    mmxfct = jsimd_h2v2_extbgr_merged_upsample_mmx;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_h2v2_extbgrx_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extbgrx_merged_upsample_sse2;
+    mmxfct = jsimd_h2v2_extbgrx_merged_upsample_mmx;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_h2v2_extxbgr_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extxbgr_merged_upsample_sse2;
+    mmxfct = jsimd_h2v2_extxbgr_merged_upsample_mmx;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_h2v2_extxrgb_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extxrgb_merged_upsample_sse2;
+    mmxfct = jsimd_h2v2_extxrgb_merged_upsample_mmx;
+    break;
+  default:
+    avx2fct = jsimd_h2v2_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_merged_upsample_sse2;
+    mmxfct = jsimd_h2v2_merged_upsample_mmx;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else if (simd_support & JSIMD_SSE2)
+    sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else
+    mmxfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*sse2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*mmxfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_h2v1_extrgb_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extrgb_merged_upsample_sse2;
+    mmxfct = jsimd_h2v1_extrgb_merged_upsample_mmx;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_h2v1_extrgbx_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extrgbx_merged_upsample_sse2;
+    mmxfct = jsimd_h2v1_extrgbx_merged_upsample_mmx;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_h2v1_extbgr_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extbgr_merged_upsample_sse2;
+    mmxfct = jsimd_h2v1_extbgr_merged_upsample_mmx;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_h2v1_extbgrx_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extbgrx_merged_upsample_sse2;
+    mmxfct = jsimd_h2v1_extbgrx_merged_upsample_mmx;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_h2v1_extxbgr_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extxbgr_merged_upsample_sse2;
+    mmxfct = jsimd_h2v1_extxbgr_merged_upsample_mmx;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_h2v1_extxrgb_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extxrgb_merged_upsample_sse2;
+    mmxfct = jsimd_h2v1_extxrgb_merged_upsample_mmx;
+    break;
+  default:
+    avx2fct = jsimd_h2v1_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_merged_upsample_sse2;
+    mmxfct = jsimd_h2v1_merged_upsample_mmx;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else if (simd_support & JSIMD_SSE2)
+    sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else
+    mmxfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_SSE)
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
+               DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_convsamp_avx2(sample_data, start_col, workspace);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_convsamp_sse2(sample_data, start_col, workspace);
+  else
+    jsimd_convsamp_mmx(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+                     FAST_FLOAT *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_convsamp_float_sse2(sample_data, start_col, workspace);
+  else if (simd_support & JSIMD_SSE)
+    jsimd_convsamp_float_sse(sample_data, start_col, workspace);
+  else
+    jsimd_convsamp_float_3dnow(sample_data, start_col, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) && IS_ALIGNED_AVX(jconst_fdct_islow_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_ifast_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow(DCTELEM *data)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_fdct_islow_avx2(data);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_fdct_islow_sse2(data);
+  else
+    jsimd_fdct_islow_mmx(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast(DCTELEM *data)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    jsimd_fdct_ifast_sse2(data);
+  else
+    jsimd_fdct_ifast_mmx(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float(FAST_FLOAT *data)
+{
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
+    jsimd_fdct_float_sse(data);
+  else if (simd_support & JSIMD_3DNOW)
+    jsimd_fdct_float_3dnow(data);
+}
+
+GLOBAL(int)
+jsimd_can_quantize(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_SSE)
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_quantize_avx2(coef_block, divisors, workspace);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_quantize_sse2(coef_block, divisors, workspace);
+  else
+    jsimd_quantize_mmx(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                     FAST_FLOAT *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_quantize_float_sse2(coef_block, divisors, workspace);
+  else if (simd_support & JSIMD_SSE)
+    jsimd_quantize_float_sse(coef_block, divisors, workspace);
+  else
+    jsimd_quantize_float_3dnow(coef_block, divisors, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    jsimd_idct_2x2_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+  else
+    jsimd_idct_2x2_mmx(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    jsimd_idct_4x4_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+  else
+    jsimd_idct_4x4_mmx(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) && IS_ALIGNED_AVX(jconst_idct_islow_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+  if (sizeof(FLOAT_MULT_TYPE) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
+    return 1;
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_idct_float_sse))
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_idct_islow_avx2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else if (simd_support & JSIMD_SSE2)
+    jsimd_idct_islow_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else
+    jsimd_idct_islow_mmx(compptr->dct_table, coef_block, output_buf,
+                         output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
+    jsimd_idct_ifast_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else
+    jsimd_idct_ifast_mmx(compptr->dct_table, coef_block, output_buf,
+                         output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
+    jsimd_idct_float_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_idct_float_sse))
+    jsimd_idct_float_sse(compptr->dct_table, coef_block, output_buf,
+                         output_col);
+  else
+    jsimd_idct_float_3dnow(compptr->dct_table, coef_block, output_buf,
+                           output_col);
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && simd_huffman &&
+      IS_ALIGNED_SSE(jconst_huff_encode_one_block))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET *)
+jsimd_huff_encode_one_block(void *state, JOCTET *buffer, JCOEFPTR block,
+                            int last_dc_val, c_derived_tbl *dctbl,
+                            c_derived_tbl *actbl)
+{
+  return jsimd_huff_encode_one_block_sse2(state, buffer, block, last_dc_val,
+                                          dctbl, actbl);
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_first_prepare(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (SIZEOF_SIZE_T != 4)
+    return 0;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_encode_mcu_AC_first_prepare(const JCOEF *block,
+                                  const int *jpeg_natural_order_start, int Sl,
+                                  int Al, JCOEF *values, size_t *zerobits)
+{
+  jsimd_encode_mcu_AC_first_prepare_sse2(block, jpeg_natural_order_start,
+                                         Sl, Al, values, zerobits);
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_refine_prepare(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (SIZEOF_SIZE_T != 4)
+    return 0;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block,
+                                   const int *jpeg_natural_order_start, int Sl,
+                                   int Al, JCOEF *absvalues, size_t *bits)
+{
+  return jsimd_encode_mcu_AC_refine_prepare_sse2(block,
+                                                 jpeg_natural_order_start,
+                                                 Sl, Al, absvalues, bits);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/i386/jsimdcpu.asm Added

@@ -0,0 +1,135 @@
+;
+; jsimdcpu.asm - SIMD instruction support check
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        32
+;
+; Check if the CPU supports SIMD instructions
+;
+; GLOBAL(unsigned int)
+; jpeg_simd_cpu_support(void)
+;
+
+    align       32
+    GLOBAL_FUNCTION(jpeg_simd_cpu_support)
+
+EXTN(jpeg_simd_cpu_support):
+    push        ebx
+;   push        ecx                     ; need not be preserved
+;   push        edx                     ; need not be preserved
+;   push        esi                     ; unused
+    push        edi
+
+    xor         edi, edi                ; simd support flag
+
+    pushfd
+    pop         eax
+    mov         edx, eax
+    xor         eax, 1<<21              ; flip ID bit in EFLAGS
+    push        eax
+    popfd
+    pushfd
+    pop         eax
+    xor         eax, edx
+    jz          near .return            ; CPUID is not supported
+
+    ; Check whether CPUID leaf 07H is supported
+    ; (leaf 07H is used to check for AVX2 instruction support)
+    xor         eax, eax
+    cpuid
+    test        eax, eax
+    jz          near .return
+    cmp         eax, 7
+    jl          short .no_avx2          ; Maximum leaf < 07H
+
+    ; Check for AVX2 instruction support
+    mov         eax, 7
+    xor         ecx, ecx
+    cpuid
+    mov         eax, ebx
+    test        eax, 1<<5               ; bit5:AVX2
+    jz          short .no_avx2
+
+    ; Check for AVX2 O/S support
+    mov         eax, 1
+    xor         ecx, ecx
+    cpuid
+    test        ecx, 1<<27
+    jz          short .no_avx2          ; O/S does not support XSAVE
+    test        ecx, 1<<28
+    jz          short .no_avx2          ; CPU does not support AVX2
+
+    xor         ecx, ecx
+    xgetbv
+    and         eax, 6
+    cmp         eax, 6                  ; O/S does not manage XMM/YMM state
+                                        ; using XSAVE
+    jnz         short .no_avx2
+
+    or          edi, JSIMD_AVX2
+.no_avx2:
+
+    ; Check CPUID leaf 01H for MMX, SSE, and SSE2 support
+    xor         eax, eax
+    inc         eax
+    cpuid
+    mov         eax, edx                ; eax = Standard feature flags
+
+    ; Check for MMX instruction support
+    test        eax, 1<<23              ; bit23:MMX
+    jz          short .no_mmx
+    or          edi, byte JSIMD_MMX
+.no_mmx:
+    test        eax, 1<<25              ; bit25:SSE
+    jz          short .no_sse
+    or          edi, byte JSIMD_SSE
+.no_sse:
+    test        eax, 1<<26              ; bit26:SSE2
+    jz          short .no_sse2
+    or          edi, byte JSIMD_SSE2
+.no_sse2:
+
+    ; Check for 3DNow! instruction support
+    mov         eax, 0x80000000
+    cpuid
+    cmp         eax, 0x80000000
+    jbe         short .return
+
+    mov         eax, 0x80000001
+    cpuid
+    mov         eax, edx                ; eax = Extended feature flags
+
+    test        eax, 1<<31              ; bit31:3DNow!(vendor independent)
+    jz          short .no_3dnow
+    or          edi, byte JSIMD_3DNOW
+.no_3dnow:
+
+.return:
+    mov         eax, edi
+
+    pop         edi
+;   pop         esi                     ; unused
+;   pop         edx                     ; need not be preserved
+;   pop         ecx                     ; need not be preserved
+    pop         ebx
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/jsimd.h Added

@@ -0,0 +1,1258 @@
+/*
+ * simd/jsimd.h
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2011, 2014-2016, 2018, 2020, D. R. Commander.
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California.
+ * Copyright (C) 2014, Linaro Limited.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ * Copyright (C) 2020, Arm Limited.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ */
+
+/* Bitmask for supported acceleration methods */
+
+#define JSIMD_NONE     0x00
+#define JSIMD_MMX      0x01
+#define JSIMD_3DNOW    0x02
+#define JSIMD_SSE      0x04
+#define JSIMD_SSE2     0x08
+#define JSIMD_NEON     0x10
+#define JSIMD_DSPR2    0x20
+#define JSIMD_ALTIVEC  0x40
+#define JSIMD_AVX2     0x80
+#define JSIMD_MMI      0x100
+
+/* SIMD Ext: retrieve SIMD/CPU information */
+EXTERN(unsigned int) jpeg_simd_cpu_support(void);
+
+/* RGB & extended RGB --> YCC Colorspace Conversion */
+EXTERN(void) jsimd_rgb_ycc_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+extern const int jconst_rgb_ycc_convert_sse2;
+EXTERN(void) jsimd_rgb_ycc_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+extern const int jconst_rgb_ycc_convert_avx2;
+EXTERN(void) jsimd_rgb_ycc_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_ycc_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+#ifndef NEON_INTRINSICS
+
+EXTERN(void) jsimd_extrgb_ycc_convert_neon_slowld3
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_neon_slowld3
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+#endif
+
+EXTERN(void) jsimd_rgb_ycc_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_ycc_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_ycc_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+/* RGB & extended RGB --> Grayscale Colorspace Conversion */
+EXTERN(void) jsimd_rgb_gray_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_mmx
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+extern const int jconst_rgb_gray_convert_sse2;
+EXTERN(void) jsimd_rgb_gray_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_sse2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+extern const int jconst_rgb_gray_convert_avx2;
+EXTERN(void) jsimd_rgb_gray_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_avx2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_gray_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_neon
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_gray_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_gray_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_mmi
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_gray_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_altivec
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows);
+
+/* YCC --> RGB & extended RGB Colorspace Conversion */
+EXTERN(void) jsimd_ycc_rgb_convert_mmx
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_mmx
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_mmx
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_mmx
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_mmx
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_mmx
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_mmx
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+
+extern const int jconst_ycc_rgb_convert_sse2;
+EXTERN(void) jsimd_ycc_rgb_convert_sse2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_sse2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_sse2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_sse2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_sse2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_sse2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_sse2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+
+extern const int jconst_ycc_rgb_convert_avx2;
+EXTERN(void) jsimd_ycc_rgb_convert_avx2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_avx2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_avx2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_avx2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_avx2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_avx2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_avx2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+
+EXTERN(void) jsimd_ycc_rgb_convert_neon
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_neon
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_neon
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_neon
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_neon
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_neon
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_neon
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_rgb565_convert_neon
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+
+#ifndef NEON_INTRINSICS
+
+EXTERN(void) jsimd_ycc_extrgb_convert_neon_slowst3
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_neon_slowst3
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+
+#endif
+
+EXTERN(void) jsimd_ycc_rgb_convert_dspr2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_dspr2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_dspr2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_dspr2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_dspr2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_dspr2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_dspr2
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+
+EXTERN(void) jsimd_ycc_rgb_convert_mmi
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_mmi
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_mmi
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_mmi
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_mmi
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_mmi
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_mmi
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+
+EXTERN(void) jsimd_ycc_rgb_convert_altivec
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_altivec
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_altivec
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_altivec
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_altivec
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_altivec
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_altivec
+  (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+   JSAMPARRAY output_buf, int num_rows);
+
+/* NULL Colorspace Conversion */
+EXTERN(void) jsimd_c_null_convert_dspr2
+  (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+   JDIMENSION output_row, int num_rows, int num_components);
+
+/* h2v1 Downsampling */
+EXTERN(void) jsimd_h2v1_downsample_mmx
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_sse2
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_avx2
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_neon
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_dspr2
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_altivec
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+/* h2v2 Downsampling */
+EXTERN(void) jsimd_h2v2_downsample_mmx
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_sse2
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_avx2
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_neon
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_dspr2
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_mmi
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_altivec
+  (JDIMENSION image_width, int max_v_samp_factor, JDIMENSION v_samp_factor,
+   JDIMENSION width_in_blocks, JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+/* h2v2 Smooth Downsampling */
+EXTERN(void) jsimd_h2v2_smooth_downsample_dspr2
+  (JSAMPARRAY input_data, JSAMPARRAY output_data, JDIMENSION v_samp_factor,
+   int max_v_samp_factor, int smoothing_factor, JDIMENSION width_in_blocks,
+   JDIMENSION image_width);
+
+
+/* Upsampling */
+EXTERN(void) jsimd_h2v1_upsample_mmx
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_mmx
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_upsample_sse2
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_sse2
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_upsample_avx2
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_avx2
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_upsample_neon
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_neon
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_upsample_dspr2
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_dspr2
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_int_upsample_dspr2
+  (UINT8 h_expand, UINT8 v_expand, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr, JDIMENSION output_width,
+   int max_v_samp_factor);
+
+EXTERN(void) jsimd_h2v1_upsample_altivec
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_altivec
+  (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+/* Fancy Upsampling */
+EXTERN(void) jsimd_h2v1_fancy_upsample_mmx
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_mmx
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+extern const int jconst_fancy_upsample_sse2;
+EXTERN(void) jsimd_h2v1_fancy_upsample_sse2
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_sse2
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+extern const int jconst_fancy_upsample_avx2;
+EXTERN(void) jsimd_h2v1_fancy_upsample_avx2
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_avx2
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_neon
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_neon
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h1v2_fancy_upsample_neon
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_dspr2
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_dspr2
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_mmi
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_mmi
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_altivec
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_altivec
+  (int max_v_samp_factor, JDIMENSION downsampled_width, JSAMPARRAY input_data,
+   JSAMPARRAY *output_data_ptr);
+
+/* Merged Upsampling */
+EXTERN(void) jsimd_h2v1_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mmx
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+extern const int jconst_merged_upsample_sse2;
+EXTERN(void) jsimd_h2v1_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_sse2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+extern const int jconst_merged_upsample_avx2;
+EXTERN(void) jsimd_h2v1_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_avx2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v1_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_neon
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v1_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_dspr2
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf, JSAMPLE *range);
+
+EXTERN(void) jsimd_h2v1_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mmi
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v1_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_altivec
+  (JDIMENSION output_width, JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+   JSAMPARRAY output_buf);
+
+/* Sample Conversion */
+EXTERN(void) jsimd_convsamp_mmx
+  (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_sse2
+  (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_avx2
+  (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_neon
+  (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_dspr2
+  (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_altivec
+  (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+/* Floating Point Sample Conversion */
+EXTERN(void) jsimd_convsamp_float_3dnow
+  (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_convsamp_float_sse
+  (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_convsamp_float_sse2
+  (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_convsamp_float_dspr2
+  (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+/* Accurate Integer Forward DCT */
+EXTERN(void) jsimd_fdct_islow_mmx(DCTELEM *data);
+
+extern const int jconst_fdct_islow_sse2;
+EXTERN(void) jsimd_fdct_islow_sse2(DCTELEM *data);
+
+extern const int jconst_fdct_islow_avx2;
+EXTERN(void) jsimd_fdct_islow_avx2(DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_neon(DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_dspr2(DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_mmi(DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_altivec(DCTELEM *data);
+
+/* Fast Integer Forward DCT */
+EXTERN(void) jsimd_fdct_ifast_mmx(DCTELEM *data);
+
+extern const int jconst_fdct_ifast_sse2;
+EXTERN(void) jsimd_fdct_ifast_sse2(DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_neon(DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_dspr2(DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_mmi(DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_altivec(DCTELEM *data);
+
+/* Floating Point Forward DCT */
+EXTERN(void) jsimd_fdct_float_3dnow(FAST_FLOAT *data);
+
+extern const int jconst_fdct_float_sse;
+EXTERN(void) jsimd_fdct_float_sse(FAST_FLOAT *data);
+
+/* Quantization */
+EXTERN(void) jsimd_quantize_mmx
+  (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_sse2
+  (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_avx2
+  (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_neon
+  (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_dspr2
+  (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_mmi
+  (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_altivec
+  (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+/* Floating Point Quantization */
+EXTERN(void) jsimd_quantize_float_3dnow
+  (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_quantize_float_sse
+  (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_quantize_float_sse2
+  (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_quantize_float_dspr2
+  (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+/* Scaled Inverse DCT */
+EXTERN(void) jsimd_idct_2x2_mmx
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_mmx
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+extern const int jconst_idct_red_sse2;
+EXTERN(void) jsimd_idct_2x2_sse2
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_sse2
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_2x2_neon
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_neon
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_2x2_dspr2
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_dspr2
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col, int *workspace);
+EXTERN(void) jsimd_idct_6x6_dspr2
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+EXTERN(void) jsimd_idct_12x12_pass1_dspr2
+  (JCOEFPTR coef_block, void *dct_table, int *workspace);
+EXTERN(void) jsimd_idct_12x12_pass2_dspr2
+  (int *workspace, int *output);
+
+/* Accurate Integer Inverse DCT */
+EXTERN(void) jsimd_idct_islow_mmx
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+extern const int jconst_idct_islow_sse2;
+EXTERN(void) jsimd_idct_islow_sse2
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+extern const int jconst_idct_islow_avx2;
+EXTERN(void) jsimd_idct_islow_avx2
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_islow_neon
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_islow_dspr2
+  (void *dct_table, JCOEFPTR coef_block, int *output_buf, JSAMPLE *output_col);
+
+EXTERN(void) jsimd_idct_islow_mmi
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_islow_altivec
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+/* Fast Integer Inverse DCT */
+EXTERN(void) jsimd_idct_ifast_mmx
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+extern const int jconst_idct_ifast_sse2;
+EXTERN(void) jsimd_idct_ifast_sse2
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_ifast_neon
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_ifast_cols_dspr2
+  (JCOEF *inptr, IFAST_MULT_TYPE *quantptr, DCTELEM *wsptr,
+   const int *idct_coefs);
+EXTERN(void) jsimd_idct_ifast_rows_dspr2
+  (DCTELEM *wsptr, JSAMPARRAY output_buf, JDIMENSION output_col,
+   const int *idct_coefs);
+
+EXTERN(void) jsimd_idct_ifast_mmi
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_ifast_altivec
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+/* Floating Point Inverse DCT */
+EXTERN(void) jsimd_idct_float_3dnow
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+extern const int jconst_idct_float_sse;
+EXTERN(void) jsimd_idct_float_sse
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+extern const int jconst_idct_float_sse2;
+EXTERN(void) jsimd_idct_float_sse2
+  (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+   JDIMENSION output_col);
+
+/* Huffman coding */
+extern const int jconst_huff_encode_one_block;
+EXTERN(JOCTET *) jsimd_huff_encode_one_block_sse2
+  (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+   c_derived_tbl *dctbl, c_derived_tbl *actbl);
+
+EXTERN(JOCTET *) jsimd_huff_encode_one_block_neon
+  (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+   c_derived_tbl *dctbl, c_derived_tbl *actbl);
+
+#ifndef NEON_INTRINSICS
+
+EXTERN(JOCTET *) jsimd_huff_encode_one_block_neon_slowtbl
+  (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+   c_derived_tbl *dctbl, c_derived_tbl *actbl);
+
+#endif
+
+/* Progressive Huffman encoding */
+EXTERN(void) jsimd_encode_mcu_AC_first_prepare_sse2
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *values, size_t *zerobits);
+
+EXTERN(void) jsimd_encode_mcu_AC_first_prepare_neon
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *values, size_t *zerobits);
+
+EXTERN(int) jsimd_encode_mcu_AC_refine_prepare_sse2
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *absvalues, size_t *bits);
+
+EXTERN(int) jsimd_encode_mcu_AC_refine_prepare_neon
+  (const JCOEF *block, const int *jpeg_natural_order_start, int Sl, int Al,
+   JCOEF *absvalues, size_t *bits);

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips/jsimd.c Added

@@ -0,0 +1,1147 @@
+/*
+ * jsimd_mips.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2009-2011, 2014, 2016, 2018, 2020, D. R. Commander.
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * MIPS architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+
+static unsigned int simd_support = ~0;
+
+#if !(defined(__mips_dsp) && (__mips_dsp_rev >= 2)) && defined(__linux__)
+
+LOCAL(void)
+parse_proc_cpuinfo(const char *search_string)
+{
+  const char *file_name = "/proc/cpuinfo";
+  char cpuinfo_line256;
+  FILE *f = NULL;
+
+  simd_support = 0;
+
+  if ((f = fopen(file_name, "r")) != NULL) {
+    while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f) != NULL) {
+      if (strstr(cpuinfo_line, search_string) != NULL) {
+        fclose(f);
+        simd_support |= JSIMD_DSPR2;
+        return;
+      }
+    }
+    fclose(f);
+  }
+  /* Did not find string in the proc file, or not Linux ELF. */
+}
+
+#endif
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd(void)
+{
+#ifndef NO_GETENV
+  char *env = NULL;
+#endif
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = 0;
+
+#if defined(__mips_dsp) && (__mips_dsp_rev >= 2)
+  simd_support |= JSIMD_DSPR2;
+#elif defined(__linux__)
+  /* We still have a chance to use MIPS DSPR2 regardless of globally used
+   * -mdspr2 options passed to gcc by performing runtime detection via
+   * /proc/cpuinfo parsing on linux */
+  parse_proc_cpuinfo("MIPS 74K");
+#endif
+
+#ifndef NO_GETENV
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCEDSPR2");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = JSIMD_DSPR2;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+#endif
+}
+
+static const int mips_idct_ifast_coefs4 = {
+  0x45404540,           /* FIX( 1.082392200 / 2) =  17734 = 0x4546 */
+  0x5A805A80,           /* FIX( 1.414213562 / 2) =  23170 = 0x5A82 */
+  0x76407640,           /* FIX( 1.847759065 / 2) =  30274 = 0x7642 */
+  0xAC60AC60            /* FIX(-2.613125930 / 4) = -21407 = 0xAC61 */
+};
+
+/* The following struct is borrowed from jdsample.c */
+typedef void (*upsample1_ptr) (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JSAMPARRAY input_data,
+                               JSAMPARRAY *output_data_ptr);
+typedef struct {
+  struct jpeg_upsampler pub;
+  JSAMPARRAY color_bufMAX_COMPONENTS;
+  upsample1_ptr methodsMAX_COMPONENTS;
+  int next_row_out;
+  JDIMENSION rows_to_go;
+  int rowgroup_heightMAX_COMPONENTS;
+  UINT8 h_expandMAX_COMPONENTS;
+  UINT8 v_expandMAX_COMPONENTS;
+} my_upsampler;
+
+typedef my_upsampler *my_upsample_ptr;
+
+GLOBAL(int)
+jsimd_can_rgb_ycc(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_c_can_null_convert(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                      JSAMPIMAGE output_buf, JDIMENSION output_row,
+                      int num_rows)
+{
+  void (*dspr2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    dspr2fct = jsimd_extrgb_ycc_convert_dspr2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    dspr2fct = jsimd_extrgbx_ycc_convert_dspr2;
+    break;
+  case JCS_EXT_BGR:
+    dspr2fct = jsimd_extbgr_ycc_convert_dspr2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    dspr2fct = jsimd_extbgrx_ycc_convert_dspr2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    dspr2fct = jsimd_extxbgr_ycc_convert_dspr2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    dspr2fct = jsimd_extxrgb_ycc_convert_dspr2;
+    break;
+  default:
+    dspr2fct = jsimd_extrgb_ycc_convert_dspr2;
+    break;
+  }
+
+  dspr2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                       JSAMPIMAGE output_buf, JDIMENSION output_row,
+                       int num_rows)
+{
+  void (*dspr2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    dspr2fct = jsimd_extrgb_gray_convert_dspr2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    dspr2fct = jsimd_extrgbx_gray_convert_dspr2;
+    break;
+  case JCS_EXT_BGR:
+    dspr2fct = jsimd_extbgr_gray_convert_dspr2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    dspr2fct = jsimd_extbgrx_gray_convert_dspr2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    dspr2fct = jsimd_extxbgr_gray_convert_dspr2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    dspr2fct = jsimd_extxrgb_gray_convert_dspr2;
+    break;
+  default:
+    dspr2fct = jsimd_extrgb_gray_convert_dspr2;
+    break;
+  }
+
+  dspr2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION input_row, JSAMPARRAY output_buf,
+                      int num_rows)
+{
+  void (*dspr2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    dspr2fct = jsimd_ycc_extrgb_convert_dspr2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    dspr2fct = jsimd_ycc_extrgbx_convert_dspr2;
+    break;
+  case JCS_EXT_BGR:
+    dspr2fct = jsimd_ycc_extbgr_convert_dspr2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    dspr2fct = jsimd_ycc_extbgrx_convert_dspr2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    dspr2fct = jsimd_ycc_extxbgr_convert_dspr2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    dspr2fct = jsimd_ycc_extxrgb_convert_dspr2;
+    break;
+  default:
+    dspr2fct = jsimd_ycc_extrgb_convert_dspr2;
+    break;
+  }
+
+  dspr2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_c_null_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                     JSAMPIMAGE output_buf, JDIMENSION output_row,
+                     int num_rows)
+{
+  jsimd_c_null_convert_dspr2(cinfo->image_width, input_buf, output_buf,
+                             output_row, num_rows, cinfo->num_components);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  /* FIXME: jsimd_h2v2_downsample_dspr2() fails some of the TJBench tiling
+   * regression tests, probably because the DSPr2 SIMD implementation predates
+   * those tests. */
+#if 0
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_smooth_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  /* FIXME: jsimd_h2v1_downsample_dspr2() fails some of the TJBench tiling
+   * regression tests, probably because the DSPr2 SIMD implementation predates
+   * those tests. */
+#if 0
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_dspr2(cinfo->image_width, cinfo->max_v_samp_factor,
+                              compptr->v_samp_factor, compptr->width_in_blocks,
+                              input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v2_smooth_downsample(j_compress_ptr cinfo,
+                             jpeg_component_info *compptr,
+                             JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_smooth_downsample_dspr2(input_data, output_data,
+                                     compptr->v_samp_factor,
+                                     cinfo->max_v_samp_factor,
+                                     cinfo->smoothing_factor,
+                                     compptr->width_in_blocks,
+                                     cinfo->image_width);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_dspr2(cinfo->image_width, cinfo->max_v_samp_factor,
+                              compptr->v_samp_factor, compptr->width_in_blocks,
+                              input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+#if defined(__MIPSEL__)
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_int_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_upsample_dspr2(cinfo->max_v_samp_factor, cinfo->output_width,
+                            input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_upsample_dspr2(cinfo->max_v_samp_factor, cinfo->output_width,
+                            input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_int_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                   JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
+
+  jsimd_int_upsample_dspr2(upsample->h_expandcompptr->component_index,
+                           upsample->v_expandcompptr->component_index,
+                           input_data, output_data_ptr, cinfo->output_width,
+                           cinfo->max_v_samp_factor);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+#if defined(__MIPSEL__)
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+#if defined(__MIPSEL__)
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_fancy_upsample_dspr2(cinfo->max_v_samp_factor,
+                                  compptr->downsampled_width, input_data,
+                                  output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_fancy_upsample_dspr2(cinfo->max_v_samp_factor,
+                                  compptr->downsampled_width, input_data,
+                                  output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*dspr2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, JSAMPLE *);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    dspr2fct = jsimd_h2v2_extrgb_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    dspr2fct = jsimd_h2v2_extrgbx_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_BGR:
+    dspr2fct = jsimd_h2v2_extbgr_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    dspr2fct = jsimd_h2v2_extbgrx_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    dspr2fct = jsimd_h2v2_extxbgr_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    dspr2fct = jsimd_h2v2_extxrgb_merged_upsample_dspr2;
+    break;
+  default:
+    dspr2fct = jsimd_h2v2_extrgb_merged_upsample_dspr2;
+    break;
+  }
+
+  dspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf,
+           cinfo->sample_range_limit);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*dspr2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, JSAMPLE *);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    dspr2fct = jsimd_h2v1_extrgb_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    dspr2fct = jsimd_h2v1_extrgbx_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_BGR:
+    dspr2fct = jsimd_h2v1_extbgr_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    dspr2fct = jsimd_h2v1_extbgrx_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    dspr2fct = jsimd_h2v1_extxbgr_merged_upsample_dspr2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    dspr2fct = jsimd_h2v1_extxrgb_merged_upsample_dspr2;
+    break;
+  default:
+    dspr2fct = jsimd_h2v1_extrgb_merged_upsample_dspr2;
+    break;
+  }
+
+  dspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf,
+           cinfo->sample_range_limit);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+#if defined(__MIPSEL__)
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+#ifndef __mips_soft_float
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
+               DCTELEM *workspace)
+{
+  jsimd_convsamp_dspr2(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+                     FAST_FLOAT *workspace)
+{
+#ifndef __mips_soft_float
+  jsimd_convsamp_float_dspr2(sample_data, start_col, workspace);
+#endif
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+#if defined(__MIPSEL__)
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+#if defined(__MIPSEL__)
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow(DCTELEM *data)
+{
+  jsimd_fdct_islow_dspr2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast(DCTELEM *data)
+{
+  jsimd_fdct_ifast_dspr2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float(FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+#ifndef __mips_soft_float
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
+{
+  jsimd_quantize_dspr2(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                     FAST_FLOAT *workspace)
+{
+#ifndef __mips_soft_float
+  jsimd_quantize_float_dspr2(coef_block, divisors, workspace);
+#endif
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+#if defined(__MIPSEL__)
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_6x6(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_12x12(void)
+{
+  init_simd();
+
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  jsimd_idct_2x2_dspr2(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  int workspaceDCTSIZE * 4; /* buffers data between passes */
+
+  jsimd_idct_4x4_dspr2(compptr->dct_table, coef_block, output_buf, output_col,
+                       workspace);
+}
+
+GLOBAL(void)
+jsimd_idct_6x6(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  jsimd_idct_6x6_dspr2(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_12x12(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  int workspace96;
+  int output12 = {
+    (int)(output_buf0 + output_col),
+    (int)(output_buf1 + output_col),
+    (int)(output_buf2 + output_col),
+    (int)(output_buf3 + output_col),
+    (int)(output_buf4 + output_col),
+    (int)(output_buf5 + output_col),
+    (int)(output_buf6 + output_col),
+    (int)(output_buf7 + output_col),
+    (int)(output_buf8 + output_col),
+    (int)(output_buf9 + output_col),
+    (int)(output_buf10 + output_col),
+    (int)(output_buf11 + output_col)
+  };
+
+  jsimd_idct_12x12_pass1_dspr2(coef_block, compptr->dct_table, workspace);
+  jsimd_idct_12x12_pass2_dspr2(workspace, output);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+#if defined(__MIPSEL__)
+  if (simd_support & JSIMD_DSPR2)
+    return 1;
+#endif
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  int output8 = {
+    (int)(output_buf0 + output_col),
+    (int)(output_buf1 + output_col),
+    (int)(output_buf2 + output_col),
+    (int)(output_buf3 + output_col),
+    (int)(output_buf4 + output_col),
+    (int)(output_buf5 + output_col),
+    (int)(output_buf6 + output_col),
+    (int)(output_buf7 + output_col)
+  };
+
+  jsimd_idct_islow_dspr2(coef_block, compptr->dct_table, output,
+                         IDCT_range_limit(cinfo));
+}
+
+GLOBAL(void)
+jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  JCOEFPTR inptr;
+  IFAST_MULT_TYPE *quantptr;
+  DCTELEM workspaceDCTSIZE2;  /* buffers data between passes */
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (IFAST_MULT_TYPE *)compptr->dct_table;
+
+  jsimd_idct_ifast_cols_dspr2(inptr, quantptr, workspace,
+                              mips_idct_ifast_coefs);
+
+  /* Pass 2: process rows from work array, store into output array. */
+  /* Note that we must descale the results by a factor of 8 == 2**3, */
+  /* and also undo the PASS1_BITS scaling. */
+
+  jsimd_idct_ifast_rows_dspr2(workspace, output_buf, output_col,
+                              mips_idct_ifast_coefs);
+}
+
+GLOBAL(void)
+jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block(void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET *)
+jsimd_huff_encode_one_block(void *state, JOCTET *buffer, JCOEFPTR block,
+                            int last_dc_val, c_derived_tbl *dctbl,
+                            c_derived_tbl *actbl)
+{
+  return NULL;
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_first_prepare(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_encode_mcu_AC_first_prepare(const JCOEF *block,
+                                  const int *jpeg_natural_order_start, int Sl,
+                                  int Al, JCOEF *values, size_t *zerobits)
+{
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_refine_prepare(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block,
+                                   const int *jpeg_natural_order_start, int Sl,
+                                   int Al, JCOEF *absvalues, size_t *bits)
+{
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips/jsimd_dspr2.S Added

@@ -0,0 +1,4543 @@
+/*
+ * MIPS DSPr2 optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California.
+ *                          All Rights Reserved.
+ * Authors:  Teodora Novkovic <teodora.novkovic@imgtec.com>
+ *           Darko Laus       <darko.laus@imgtec.com>
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#include "jsimd_dspr2_asm.h"
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_c_null_convert_dspr2)
+/*
+ * a0     = cinfo->image_width
+ * a1     = input_buf
+ * a2     = output_buf
+ * a3     = output_row
+ * 16(sp) = num_rows
+ * 20(sp) = cinfo->num_components
+ *
+ * Null conversion for compression
+ */
+    SAVE_REGS_ON_STACK 8, s0, s1
+
+    lw          t9, 24(sp)      /* t9 = num_rows */
+    lw          s0, 28(sp)      /* s0 = cinfo->num_components */
+    andi        t0, a0, 3       /* t0 = cinfo->image_width & 3 */
+    beqz        t0, 4f          /* no residual */
+     nop
+0:
+    addiu       t9, t9, -1
+    bltz        t9, 7f
+     li         t1, 0
+1:
+    sll         t3, t1, 2
+    lwx         t5, t3(a2)      /* t5 = outptr = output_bufci */
+    lw          t2, 0(a1)       /* t2 = inptr = *input_buf */
+    sll         t4, a3, 2
+    lwx         t5, t4(t5)      /* t5 = outptr = output_bufcioutput_row */
+    addu        t2, t2, t1
+    addu        s1, t5, a0
+    addu        t6, t5, t0
+2:
+    lbu         t3, 0(t2)
+    addiu       t5, t5, 1
+    sb          t3, -1(t5)
+    bne         t6, t5, 2b
+     addu       t2, t2, s0
+3:
+    lbu         t3, 0(t2)
+    addu        t4, t2, s0
+    addu        t7, t4, s0
+    addu        t8, t7, s0
+    addu        t2, t8, s0
+    lbu         t4, 0(t4)
+    lbu         t7, 0(t7)
+    lbu         t8, 0(t8)
+    addiu       t5, t5, 4
+    sb          t3, -4(t5)
+    sb          t4, -3(t5)
+    sb          t7, -2(t5)
+    bne         s1, t5, 3b
+     sb         t8, -1(t5)
+    addiu       t1, t1, 1
+    bne         t1, s0, 1b
+     nop
+    addiu       a1, a1, 4
+    bgez        t9, 0b
+     addiu      a3, a3, 1
+    b           7f
+     nop
+4:
+    addiu       t9, t9, -1
+    bltz        t9, 7f
+     li         t1, 0
+5:
+    sll         t3, t1, 2
+    lwx         t5, t3(a2)      /* t5 = outptr = output_bufci */
+    lw          t2, 0(a1)       /* t2 = inptr = *input_buf */
+    sll         t4, a3, 2
+    lwx         t5, t4(t5)      /* t5 = outptr = output_bufcioutput_row */
+    addu        t2, t2, t1
+    addu        s1, t5, a0
+    addu        t6, t5, t0
+6:
+    lbu         t3, 0(t2)
+    addu        t4, t2, s0
+    addu        t7, t4, s0
+    addu        t8, t7, s0
+    addu        t2, t8, s0
+    lbu         t4, 0(t4)
+    lbu         t7, 0(t7)
+    lbu         t8, 0(t8)
+    addiu       t5, t5, 4
+    sb          t3, -4(t5)
+    sb          t4, -3(t5)
+    sb          t7, -2(t5)
+    bne         s1, t5, 6b
+     sb         t8, -1(t5)
+    addiu       t1, t1, 1
+    bne         t1, s0, 5b
+     nop
+    addiu       a1, a1, 4
+    bgez        t9, 4b
+     addiu      a3, a3, 1
+7:
+    RESTORE_REGS_FROM_STACK 8, s0, s1
+
+    j           ra
+     nop
+
+END(jsimd_c_null_convert_dspr2)
+
+
+/*****************************************************************************/
+/*
+ * jsimd_extrgb_ycc_convert_dspr2
+ * jsimd_extbgr_ycc_convert_dspr2
+ * jsimd_extrgbx_ycc_convert_dspr2
+ * jsimd_extbgrx_ycc_convert_dspr2
+ * jsimd_extxbgr_ycc_convert_dspr2
+ * jsimd_extxrgb_ycc_convert_dspr2
+ *
+ * Colorspace conversion RGB -> YCbCr
+ */
+
+.macro GENERATE_JSIMD_RGB_YCC_CONVERT_DSPR2  colorid, pixel_size, \
+                                             r_offs, g_offs, b_offs
+
+.macro DO_RGB_TO_YCC  r, g, b, inptr
+    lbu         \r, \r_offs(\inptr)
+    lbu         \g, \g_offs(\inptr)
+    lbu         \b, \b_offs(\inptr)
+    addiu       \inptr, \pixel_size
+.endm
+
+LEAF_DSPR2(jsimd_\colorid\()_ycc_convert_dspr2)
+/*
+ * a0     = cinfo->image_width
+ * a1     = input_buf
+ * a2     = output_buf
+ * a3     = output_row
+ * 16(sp) = num_rows
+ */
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw          t7, 48(sp)      /* t7 = num_rows */
+    li          s0, 0x4c8b      /* FIX(0.29900) */
+    li          s1, 0x9646      /* FIX(0.58700) */
+    li          s2, 0x1d2f      /* FIX(0.11400) */
+    li          s3, 0xffffd4cd  /* -FIX(0.16874) */
+    li          s4, 0xffffab33  /* -FIX(0.33126) */
+    li          s5, 0x8000      /* FIX(0.50000) */
+    li          s6, 0xffff94d1  /* -FIX(0.41869) */
+    li          s7, 0xffffeb2f  /* -FIX(0.08131) */
+    li          t8, 0x807fff    /* CBCR_OFFSET + ONE_HALF-1 */
+
+0:
+    addiu       t7, -1          /* --num_rows */
+    lw          t6, 0(a1)       /* t6 = input_buf0 */
+    lw          t0, 0(a2)
+    lw          t1, 4(a2)
+    lw          t2, 8(a2)
+    sll         t3, a3, 2
+    lwx         t0, t3(t0)      /* t0 = output_buf0output_row */
+    lwx         t1, t3(t1)      /* t1 = output_buf1output_row */
+    lwx         t2, t3(t2)      /* t2 = output_buf2output_row */
+
+    addu        t9, t2, a0      /* t9 = end address */
+    addiu       a3, 1
+
+1:
+    DO_RGB_TO_YCC t3, t4, t5, t6
+
+    mtlo        s5, $ac0
+    mtlo        t8, $ac1
+    mtlo        t8, $ac2
+    maddu       $ac0, s2, t5
+    maddu       $ac1, s5, t5
+    maddu       $ac2, s5, t3
+    maddu       $ac0, s0, t3
+    maddu       $ac1, s3, t3
+    maddu       $ac2, s6, t4
+    maddu       $ac0, s1, t4
+    maddu       $ac1, s4, t4
+    maddu       $ac2, s7, t5
+    extr.w      t3, $ac0, 16
+    extr.w      t4, $ac1, 16
+    extr.w      t5, $ac2, 16
+    sb          t3, 0(t0)
+    sb          t4, 0(t1)
+    sb          t5, 0(t2)
+    addiu       t0, 1
+    addiu       t2, 1
+    bne         t2, t9, 1b
+     addiu      t1, 1
+    bgtz        t7, 0b
+     addiu      a1, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+END(jsimd_\colorid\()_ycc_convert_dspr2)
+
+.purgem DO_RGB_TO_YCC
+
+.endm
+
+/*-------------------------------------id -- pix R  G  B */
+GENERATE_JSIMD_RGB_YCC_CONVERT_DSPR2 extrgb,  3, 0, 1, 2
+GENERATE_JSIMD_RGB_YCC_CONVERT_DSPR2 extbgr,  3, 2, 1, 0
+GENERATE_JSIMD_RGB_YCC_CONVERT_DSPR2 extrgbx, 4, 0, 1, 2
+GENERATE_JSIMD_RGB_YCC_CONVERT_DSPR2 extbgrx, 4, 2, 1, 0
+GENERATE_JSIMD_RGB_YCC_CONVERT_DSPR2 extxbgr, 4, 3, 2, 1
+GENERATE_JSIMD_RGB_YCC_CONVERT_DSPR2 extxrgb, 4, 1, 2, 3
+
+
+/*****************************************************************************/
+/*
+ * jsimd_ycc_extrgb_convert_dspr2
+ * jsimd_ycc_extbgr_convert_dspr2
+ * jsimd_ycc_extrgbx_convert_dspr2
+ * jsimd_ycc_extbgrx_convert_dspr2
+ * jsimd_ycc_extxbgr_convert_dspr2
+ * jsimd_ycc_extxrgb_convert_dspr2
+ *
+ * Colorspace conversion YCbCr -> RGB
+ */
+
+.macro GENERATE_JSIMD_YCC_RGB_CONVERT_DSPR2  colorid, pixel_size, \
+                                             r_offs, g_offs, b_offs, a_offs
+
+.macro STORE_YCC_TO_RGB  scratch0 scratch1 scratch2 outptr
+    sb          \scratch0, \r_offs(\outptr)
+    sb          \scratch1, \g_offs(\outptr)
+    sb          \scratch2, \b_offs(\outptr)
+.if (\pixel_size == 4)
+    li          t0, 0xFF
+    sb          t0, \a_offs(\outptr)
+.endif
+    addiu       \outptr, \pixel_size
+.endm
+
+LEAF_DSPR2(jsimd_ycc_\colorid\()_convert_dspr2)
+/*
+ * a0     = cinfo->image_width
+ * a1     = input_buf
+ * a2     = input_row
+ * a3     = output_buf
+ * 16(sp) = num_rows
+ */
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw          s1, 48(sp)
+    li          t3, 0x8000
+    li          t4, 0x166e9     /* FIX(1.40200) */
+    li          t5, 0x1c5a2     /* FIX(1.77200) */
+    li          t6, 0xffff492e  /* -FIX(0.71414) */
+    li          t7, 0xffffa7e6  /* -FIX(0.34414) */
+    repl.ph     t8, 128
+
+0:
+    lw          s0, 0(a3)
+    lw          t0, 0(a1)
+    lw          t1, 4(a1)
+    lw          t2, 8(a1)
+    sll         s5, a2, 2
+    addiu       s1, -1
+    lwx         s2, s5(t0)
+    lwx         s3, s5(t1)
+    lwx         s4, s5(t2)
+    addu        t9, s2, a0
+    addiu       a2, 1
+
+1:
+    lbu         s7, 0(s4)       /* cr */
+    lbu         s6, 0(s3)       /* cb */
+    lbu         s5, 0(s2)       /* y */
+    addiu       s2, 1
+    addiu       s4, 1
+    addiu       s7, -128
+    addiu       s6, -128
+    mul         t2, t7, s6
+    mul         t0, t6, s7      /* Crgtabcr */
+    sll         s7, 15
+    mulq_rs.w   t1, t4, s7      /* Crrtabcr */
+    sll         s6, 15
+    addu        t2, t3          /* Cbgtabcb */
+    addu        t2, t0
+
+    mulq_rs.w   t0, t5, s6      /* Cbbtabcb */
+    sra         t2, 16
+    addu        t1, s5
+    addu        t2, s5          /* add y */
+    ins         t2, t1, 16, 16
+    subu.ph     t2, t2, t8
+    addu        t0, s5
+    shll_s.ph   t2, t2, 8
+    subu        t0, 128
+    shra.ph     t2, t2, 8
+    shll_s.w    t0, t0, 24
+    addu.ph     t2, t2, t8      /* clip & store */
+    sra         t0, t0, 24
+    sra         t1, t2, 16
+    addiu       t0, 128
+
+    STORE_YCC_TO_RGB t1, t2, t0, s0
+
+    bne         s2, t9, 1b
+     addiu      s3, 1
+    bgtz        s1, 0b
+     addiu      a3, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+END(jsimd_ycc_\colorid\()_convert_dspr2)
+
+.purgem STORE_YCC_TO_RGB
+
+.endm
+
+/*-------------------------------------id -- pix R  G  B  A */
+GENERATE_JSIMD_YCC_RGB_CONVERT_DSPR2 extrgb,  3, 0, 1, 2, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_DSPR2 extbgr,  3, 2, 1, 0, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_DSPR2 extrgbx, 4, 0, 1, 2, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_DSPR2 extbgrx, 4, 2, 1, 0, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_DSPR2 extxbgr, 4, 3, 2, 1, 0
+GENERATE_JSIMD_YCC_RGB_CONVERT_DSPR2 extxrgb, 4, 1, 2, 3, 0
+
+
+/*****************************************************************************/
+/*
+ * jsimd_extrgb_gray_convert_dspr2
+ * jsimd_extbgr_gray_convert_dspr2
+ * jsimd_extrgbx_gray_convert_dspr2
+ * jsimd_extbgrx_gray_convert_dspr2
+ * jsimd_extxbgr_gray_convert_dspr2
+ * jsimd_extxrgb_gray_convert_dspr2
+ *
+ * Colorspace conversion RGB -> GRAY
+ */
+
+.macro GENERATE_JSIMD_RGB_GRAY_CONVERT_DSPR2  colorid, pixel_size, \
+                                              r_offs, g_offs, b_offs
+
+.macro DO_RGB_TO_GRAY  r, g, b, inptr
+    lbu         \r, \r_offs(\inptr)
+    lbu         \g, \g_offs(\inptr)
+    lbu         \b, \b_offs(\inptr)
+    addiu       \inptr, \pixel_size
+.endm
+
+LEAF_DSPR2(jsimd_\colorid\()_gray_convert_dspr2)
+/*
+ * a0     = cinfo->image_width
+ * a1     = input_buf
+ * a2     = output_buf
+ * a3     = output_row
+ * 16(sp) = num_rows
+ */
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    li          s0, 0x4c8b      /* s0 = FIX(0.29900) */
+    li          s1, 0x9646      /* s1 = FIX(0.58700) */
+    li          s2, 0x1d2f      /* s2 = FIX(0.11400) */
+    li          s7, 0x8000      /* s7 = FIX(0.50000) */
+    lw          s6, 48(sp)
+    andi        t7, a0, 3
+
+0:
+    addiu       s6, -1          /* s6 = num_rows */
+    lw          t0, 0(a1)
+    lw          t1, 0(a2)
+    sll         t3, a3, 2
+    lwx         t1, t3(t1)
+    addiu       a3, 1
+    addu        t9, t1, a0
+    subu        t8, t9, t7
+    beq         t1, t8, 2f
+     nop
+
+1:
+    DO_RGB_TO_GRAY t3, t4, t5, t0
+    DO_RGB_TO_GRAY s3, s4, s5, t0
+
+    mtlo        s7, $ac0
+    maddu       $ac0, s2, t5
+    maddu       $ac0, s1, t4
+    maddu       $ac0, s0, t3
+    mtlo        s7, $ac1
+    maddu       $ac1, s2, s5
+    maddu       $ac1, s1, s4
+    maddu       $ac1, s0, s3
+    extr.w      t6, $ac0, 16
+
+    DO_RGB_TO_GRAY t3, t4, t5, t0
+    DO_RGB_TO_GRAY s3, s4, s5, t0
+
+    mtlo        s7, $ac0
+    maddu       $ac0, s2, t5
+    maddu       $ac0, s1, t4
+    extr.w      t2, $ac1, 16
+    maddu       $ac0, s0, t3
+    mtlo        s7, $ac1
+    maddu       $ac1, s2, s5
+    maddu       $ac1, s1, s4
+    maddu       $ac1, s0, s3
+    extr.w      t5, $ac0, 16
+    sb          t6, 0(t1)
+    sb          t2, 1(t1)
+    extr.w      t3, $ac1, 16
+    addiu       t1, 4
+    sb          t5, -2(t1)
+    sb          t3, -1(t1)
+    bne         t1, t8, 1b
+     nop
+
+2:
+    beqz        t7, 4f
+     nop
+
+3:
+    DO_RGB_TO_GRAY t3, t4, t5, t0
+
+    mtlo        s7, $ac0
+    maddu       $ac0, s2, t5
+    maddu       $ac0, s1, t4
+    maddu       $ac0, s0, t3
+    extr.w      t6, $ac0, 16
+    sb          t6, 0(t1)
+    addiu       t1, 1
+    bne         t1, t9, 3b
+     nop
+
+4:
+    bgtz        s6, 0b
+     addiu      a1, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+END(jsimd_\colorid\()_gray_convert_dspr2)
+
+.purgem DO_RGB_TO_GRAY
+
+.endm
+
+/*-------------------------------------id --  pix R  G  B */
+GENERATE_JSIMD_RGB_GRAY_CONVERT_DSPR2 extrgb,  3, 0, 1, 2
+GENERATE_JSIMD_RGB_GRAY_CONVERT_DSPR2 extbgr,  3, 2, 1, 0
+GENERATE_JSIMD_RGB_GRAY_CONVERT_DSPR2 extrgbx, 4, 0, 1, 2
+GENERATE_JSIMD_RGB_GRAY_CONVERT_DSPR2 extbgrx, 4, 2, 1, 0
+GENERATE_JSIMD_RGB_GRAY_CONVERT_DSPR2 extxbgr, 4, 3, 2, 1
+GENERATE_JSIMD_RGB_GRAY_CONVERT_DSPR2 extxrgb, 4, 1, 2, 3
+
+
+/*****************************************************************************/
+/*
+ * jsimd_h2v2_merged_upsample_dspr2
+ * jsimd_h2v2_extrgb_merged_upsample_dspr2
+ * jsimd_h2v2_extrgbx_merged_upsample_dspr2
+ * jsimd_h2v2_extbgr_merged_upsample_dspr2
+ * jsimd_h2v2_extbgrx_merged_upsample_dspr2
+ * jsimd_h2v2_extxbgr_merged_upsample_dspr2
+ * jsimd_h2v2_extxrgb_merged_upsample_dspr2
+ *
+ * Merged h2v2 upsample routines
+ */
+.macro GENERATE_H2V2_MERGED_UPSAMPLE_DSPR2  colorid, pixel_size, \
+                                            r1_offs, g1_offs, \
+                                            b1_offs, a1_offs, \
+                                            r2_offs, g2_offs, \
+                                            b2_offs, a2_offs
+
+.macro STORE_H2V2_2_PIXELS  scratch0 scratch1 scratch2 scratch3 scratch4 \
+                            scratch5 outptr
+    sb          \scratch0, \r1_offs(\outptr)
+    sb          \scratch1, \g1_offs(\outptr)
+    sb          \scratch2, \b1_offs(\outptr)
+    sb          \scratch3, \r2_offs(\outptr)
+    sb          \scratch4, \g2_offs(\outptr)
+    sb          \scratch5, \b2_offs(\outptr)
+.if (\pixel_size == 8)
+    li          \scratch0, 0xFF
+    sb          \scratch0, \a1_offs(\outptr)
+    sb          \scratch0, \a2_offs(\outptr)
+.endif
+    addiu       \outptr, \pixel_size
+.endm
+
+.macro STORE_H2V2_1_PIXEL  scratch0 scratch1 scratch2 outptr
+    sb          \scratch0, \r1_offs(\outptr)
+    sb          \scratch1, \g1_offs(\outptr)
+    sb          \scratch2, \b1_offs(\outptr)
+
+.if (\pixel_size == 8)
+    li          t0, 0xFF
+    sb          t0, \a1_offs(\outptr)
+.endif
+.endm
+
+LEAF_DSPR2(jsimd_h2v2_\colorid\()_merged_upsample_dspr2)
+/*
+ * a0     = cinfo->output_width
+ * a1     = input_buf
+ * a2     = in_row_group_ctr
+ * a3     = output_buf
+ * 16(sp) = cinfo->sample_range_limit
+ */
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    lw          t9, 56(sp)      /* cinfo->sample_range_limit */
+    lw          v0, 0(a1)
+    lw          v1, 4(a1)
+    lw          t0, 8(a1)
+    sll         t1, a2, 3
+    addiu       t2, t1, 4
+    sll         t3, a2, 2
+    lw          t4, 0(a3)       /* t4 = output_buf0 */
+    lwx         t1, t1(v0)      /* t1 = input_buf0in_row_group_ctr*2 */
+    lwx         t2, t2(v0)      /* t2 = input_buf0in_row_group_ctr*2 + 1 */
+    lwx         t5, t3(v1)      /* t5 = input_buf1in_row_group_ctr */
+    lwx         t6, t3(t0)      /* t6 = input_buf2in_row_group_ctr */
+    lw          t7, 4(a3)       /* t7 = output_buf1 */
+    li          s1, 0xe6ea
+    addiu       t8, s1, 0x7fff    /* t8 = 0x166e9 FIX(1.40200) */
+    addiu       s0, t8, 0x5eb9    /* s0 = 0x1c5a2 FIX(1.77200) */
+    addiu       s1, zero, 0xa7e6  /* s4 = 0xffffa7e6 -FIX(0.34414) */
+    xori        s2, s1, 0xeec8    /* s3 = 0xffff492e -FIX(0.71414) */
+    srl         t3, a0, 1
+    blez        t3, 2f
+     addu       t0, t5, t3      /* t0 = end address */
+ 1:
+    lbu         t3, 0(t5)
+    lbu         s3, 0(t6)
+    addiu       t5, t5, 1
+    addiu       t3, t3, -128    /* (cb - 128) */
+    addiu       s3, s3, -128    /* (cr - 128) */
+    mult        $ac1, s1, t3
+    madd        $ac1, s2, s3
+    sll         s3, s3, 15
+    sll         t3, t3, 15
+    mulq_rs.w   s4, t8, s3      /* s4 = (C1 * cr + ONE_HALF)>> SCALEBITS */
+    extr_r.w    s5, $ac1, 16
+    mulq_rs.w   s6, s0, t3      /* s6 = (C2 * cb + ONE_HALF)>> SCALEBITS */
+    lbu         v0, 0(t1)
+    addiu       t6, t6, 1
+    addiu       t1, t1, 2
+    addu        t3, v0, s4      /* y+cred */
+    addu        s3, v0, s5      /* y+cgreen */
+    addu        v1, v0, s6      /* y+cblue */
+    addu        t3, t9, t3      /* y+cred */
+    addu        s3, t9, s3      /* y+cgreen */
+    addu        v1, t9, v1      /* y+cblue */
+    lbu         AT, 0(t3)
+    lbu         s7, 0(s3)
+    lbu         ra, 0(v1)
+    lbu         v0, -1(t1)
+    addu        t3, v0, s4      /* y+cred */
+    addu        s3, v0, s5      /* y+cgreen */
+    addu        v1, v0, s6      /* y+cblue */
+    addu        t3, t9, t3      /* y+cred */
+    addu        s3, t9, s3      /* y+cgreen */
+    addu        v1, t9, v1      /* y+cblue */
+    lbu         t3, 0(t3)
+    lbu         s3, 0(s3)
+    lbu         v1, 0(v1)
+    lbu         v0, 0(t2)
+
+    STORE_H2V2_2_PIXELS AT, s7, ra, t3, s3, v1, t4
+
+    addu        t3, v0, s4      /* y+cred */
+    addu        s3, v0, s5      /* y+cgreen */
+    addu        v1, v0, s6      /* y+cblue */
+    addu        t3, t9, t3      /* y+cred */
+    addu        s3, t9, s3      /* y+cgreen */
+    addu        v1, t9, v1      /* y+cblue */
+    lbu         AT, 0(t3)
+    lbu         s7, 0(s3)
+    lbu         ra, 0(v1)
+    lbu         v0, 1(t2)
+    addiu       t2, t2, 2
+    addu        t3, v0, s4      /* y+cred */
+    addu        s3, v0, s5      /* y+cgreen */
+    addu        v1, v0, s6      /* y+cblue */
+    addu        t3, t9, t3      /* y+cred */
+    addu        s3, t9, s3      /* y+cgreen */
+    addu        v1, t9, v1      /* y+cblue */
+    lbu         t3, 0(t3)
+    lbu         s3, 0(s3)
+    lbu         v1, 0(v1)
+
+    STORE_H2V2_2_PIXELS AT, s7, ra, t3, s3, v1, t7
+
+    bne         t0, t5, 1b
+     nop
+2:
+    andi        t0, a0, 1
+    beqz        t0, 4f
+     lbu        t3, 0(t5)
+    lbu         s3, 0(t6)
+    addiu       t3, t3, -128    /* (cb - 128) */
+    addiu       s3, s3, -128    /* (cr - 128) */
+    mult        $ac1, s1, t3
+    madd        $ac1, s2, s3
+    sll         s3, s3, 15
+    sll         t3, t3, 15
+    lbu         v0, 0(t1)
+    extr_r.w    s5, $ac1, 16
+    mulq_rs.w   s4, t8, s3      /* s4 = (C1 * cr + ONE_HALF)>> SCALEBITS */
+    mulq_rs.w   s6, s0, t3      /* s6 = (C2 * cb + ONE_HALF)>> SCALEBITS */
+    addu        t3, v0, s4      /* y+cred */
+    addu        s3, v0, s5      /* y+cgreen */
+    addu        v1, v0, s6      /* y+cblue */
+    addu        t3, t9, t3      /* y+cred */
+    addu        s3, t9, s3      /* y+cgreen */
+    addu        v1, t9, v1      /* y+cblue */
+    lbu         t3, 0(t3)
+    lbu         s3, 0(s3)
+    lbu         v1, 0(v1)
+    lbu         v0, 0(t2)
+
+    STORE_H2V2_1_PIXEL t3, s3, v1, t4
+
+    addu        t3, v0, s4      /* y+cred */
+    addu        s3, v0, s5      /* y+cgreen */
+    addu        v1, v0, s6      /* y+cblue */
+    addu        t3, t9, t3      /* y+cred */
+    addu        s3, t9, s3      /* y+cgreen */
+    addu        v1, t9, v1      /* y+cblue */
+    lbu         t3, 0(t3)
+    lbu         s3, 0(s3)
+    lbu         v1, 0(v1)
+
+    STORE_H2V2_1_PIXEL t3, s3, v1, t7
+4:
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    j           ra
+     nop
+
+END(jsimd_h2v2_\colorid\()_merged_upsample_dspr2)
+
+.purgem STORE_H2V2_1_PIXEL
+.purgem STORE_H2V2_2_PIXELS
+.endm
+
+/*------------------------------------id -- pix R1 G1 B1 A1 R2 G2 B2 A2 */
+GENERATE_H2V2_MERGED_UPSAMPLE_DSPR2 extrgb,  6, 0, 1, 2, 6, 3, 4, 5, 6
+GENERATE_H2V2_MERGED_UPSAMPLE_DSPR2 extbgr,  6, 2, 1, 0, 3, 5, 4, 3, 6
+GENERATE_H2V2_MERGED_UPSAMPLE_DSPR2 extrgbx, 8, 0, 1, 2, 3, 4, 5, 6, 7
+GENERATE_H2V2_MERGED_UPSAMPLE_DSPR2 extbgrx, 8, 2, 1, 0, 3, 6, 5, 4, 7
+GENERATE_H2V2_MERGED_UPSAMPLE_DSPR2 extxbgr, 8, 3, 2, 1, 0, 7, 6, 5, 4
+GENERATE_H2V2_MERGED_UPSAMPLE_DSPR2 extxrgb, 8, 1, 2, 3, 0, 5, 6, 7, 4
+
+
+/*****************************************************************************/
+/*
+ * jsimd_h2v1_merged_upsample_dspr2
+ * jsimd_h2v1_extrgb_merged_upsample_dspr2
+ * jsimd_h2v1_extrgbx_merged_upsample_dspr2
+ * jsimd_h2v1_extbgr_merged_upsample_dspr2
+ * jsimd_h2v1_extbgrx_merged_upsample_dspr2
+ * jsimd_h2v1_extxbgr_merged_upsample_dspr2
+ * jsimd_h2v1_extxrgb_merged_upsample_dspr2
+ *
+ * Merged h2v1 upsample routines
+ */
+
+.macro GENERATE_H2V1_MERGED_UPSAMPLE_DSPR2  colorid, pixel_size, \
+                                            r1_offs, g1_offs, \
+                                            b1_offs, a1_offs, \
+                                            r2_offs, g2_offs, \
+                                            b2_offs, a2_offs
+
+.macro STORE_H2V1_2_PIXELS  scratch0 scratch1 scratch2 scratch3 scratch4 \
+                            scratch5 outptr
+    sb          \scratch0, \r1_offs(\outptr)
+    sb          \scratch1, \g1_offs(\outptr)
+    sb          \scratch2, \b1_offs(\outptr)
+    sb          \scratch3, \r2_offs(\outptr)
+    sb          \scratch4, \g2_offs(\outptr)
+    sb          \scratch5, \b2_offs(\outptr)
+.if (\pixel_size == 8)
+    li          t0, 0xFF
+    sb          t0, \a1_offs(\outptr)
+    sb          t0, \a2_offs(\outptr)
+.endif
+    addiu       \outptr, \pixel_size
+.endm
+
+.macro STORE_H2V1_1_PIXEL  scratch0 scratch1 scratch2 outptr
+    sb          \scratch0, \r1_offs(\outptr)
+    sb          \scratch1, \g1_offs(\outptr)
+    sb          \scratch2, \b1_offs(\outptr)
+.if (\pixel_size == 8)
+    li          t0, 0xFF
+    sb          t0, \a1_offs(\outptr)
+.endif
+.endm
+
+LEAF_DSPR2(jsimd_h2v1_\colorid\()_merged_upsample_dspr2)
+/*
+ * a0     = cinfo->output_width
+ * a1     = input_buf
+ * a2     = in_row_group_ctr
+ * a3     = output_buf
+ * 16(sp) = range_limit
+ */
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    li          t0, 0xe6ea
+    lw          t1, 0(a1)         /* t1 = input_buf0 */
+    lw          t2, 4(a1)         /* t2 = input_buf1 */
+    lw          t3, 8(a1)         /* t3 = input_buf2 */
+    lw          t8, 56(sp)        /* t8 = range_limit */
+    addiu       s1, t0, 0x7fff    /* s1 = 0x166e9 FIX(1.40200) */
+    addiu       s2, s1, 0x5eb9    /* s2 = 0x1c5a2 FIX(1.77200) */
+    addiu       s0, t0, 0x9916    /* s0 = 0x8000 */
+    addiu       s4, zero, 0xa7e6  /* s4 = 0xffffa7e6 -FIX(0.34414) */
+    xori        s3, s4, 0xeec8    /* s3 = 0xffff492e -FIX(0.71414) */
+    srl         t0, a0, 1
+    sll         t4, a2, 2
+    lwx         s5, t4(t1)      /* s5 = inptr0 */
+    lwx         s6, t4(t2)      /* s6 = inptr1 */
+    lwx         s7, t4(t3)      /* s7 = inptr2 */
+    lw          t7, 0(a3)       /* t7 = outptr */
+    blez        t0, 2f
+     addu       t9, s6, t0      /* t9 = end address */
+1:
+    lbu         t2, 0(s6)       /* t2 = cb */
+    lbu         t0, 0(s7)       /* t0 = cr */
+    lbu         t1, 0(s5)       /* t1 = y */
+    addiu       t2, t2, -128    /* t2 = cb - 128 */
+    addiu       t0, t0, -128    /* t0 = cr - 128 */
+    mult        $ac1, s4, t2
+    madd        $ac1, s3, t0
+    sll         t0, t0, 15
+    sll         t2, t2, 15
+    mulq_rs.w   t0, s1, t0      /* t0 = (C1*cr + ONE_HALF)>> SCALEBITS */
+    extr_r.w    t5, $ac1, 16
+    mulq_rs.w   t6, s2, t2      /* t6 = (C2*cb + ONE_HALF)>> SCALEBITS */
+    addiu       s7, s7, 1
+    addiu       s6, s6, 1
+    addu        t2, t1, t0      /* t2 = y + cred */
+    addu        t3, t1, t5      /* t3 = y + cgreen */
+    addu        t4, t1, t6      /* t4 = y + cblue */
+    addu        t2, t8, t2
+    addu        t3, t8, t3
+    addu        t4, t8, t4
+    lbu         t1, 1(s5)
+    lbu         v0, 0(t2)
+    lbu         v1, 0(t3)
+    lbu         ra, 0(t4)
+    addu        t2, t1, t0
+    addu        t3, t1, t5
+    addu        t4, t1, t6
+    addu        t2, t8, t2
+    addu        t3, t8, t3
+    addu        t4, t8, t4
+    lbu         t2, 0(t2)
+    lbu         t3, 0(t3)
+    lbu         t4, 0(t4)
+
+    STORE_H2V1_2_PIXELS v0, v1, ra, t2, t3, t4, t7
+
+    bne         t9, s6, 1b
+     addiu      s5, s5, 2
+2:
+    andi        t0, a0, 1
+    beqz        t0, 4f
+     nop
+3:
+    lbu         t2, 0(s6)
+    lbu         t0, 0(s7)
+    lbu         t1, 0(s5)
+    addiu       t2, t2, -128    /* (cb - 128) */
+    addiu       t0, t0, -128    /* (cr - 128) */
+    mul         t3, s4, t2
+    mul         t4, s3, t0
+    sll         t0, t0, 15
+    sll         t2, t2, 15
+    mulq_rs.w   t0, s1, t0      /* (C1*cr + ONE_HALF)>> SCALEBITS */
+    mulq_rs.w   t6, s2, t2      /* (C2*cb + ONE_HALF)>> SCALEBITS */
+    addu        t3, t3, s0
+    addu        t3, t4, t3
+    sra         t5, t3, 16      /* (C4*cb + ONE_HALF + C3*cr)>> SCALEBITS */
+    addu        t2, t1, t0      /* y + cred */
+    addu        t3, t1, t5      /* y + cgreen */
+    addu        t4, t1, t6      /* y + cblue */
+    addu        t2, t8, t2
+    addu        t3, t8, t3
+    addu        t4, t8, t4
+    lbu         t2, 0(t2)
+    lbu         t3, 0(t3)
+    lbu         t4, 0(t4)
+
+    STORE_H2V1_1_PIXEL t2, t3, t4, t7
+4:
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    j           ra
+     nop
+
+END(jsimd_h2v1_\colorid\()_merged_upsample_dspr2)
+
+.purgem STORE_H2V1_1_PIXEL
+.purgem STORE_H2V1_2_PIXELS
+.endm
+
+/*------------------------------------id -- pix R1 G1 B1 A1 R2 G2 B2 A2 */
+GENERATE_H2V1_MERGED_UPSAMPLE_DSPR2 extrgb,  6, 0, 1, 2, 6, 3, 4, 5, 6
+GENERATE_H2V1_MERGED_UPSAMPLE_DSPR2 extbgr,  6, 2, 1, 0, 3, 5, 4, 3, 6
+GENERATE_H2V1_MERGED_UPSAMPLE_DSPR2 extrgbx, 8, 0, 1, 2, 3, 4, 5, 6, 7
+GENERATE_H2V1_MERGED_UPSAMPLE_DSPR2 extbgrx, 8, 2, 1, 0, 3, 6, 5, 4, 7
+GENERATE_H2V1_MERGED_UPSAMPLE_DSPR2 extxbgr, 8, 3, 2, 1, 0, 7, 6, 5, 4
+GENERATE_H2V1_MERGED_UPSAMPLE_DSPR2 extxrgb, 8, 1, 2, 3, 0, 5, 6, 7, 4
+
+
+/*****************************************************************************/
+/*
+ * jsimd_h2v2_fancy_upsample_dspr2
+ *
+ * Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+ */
+LEAF_DSPR2(jsimd_h2v2_fancy_upsample_dspr2)
+/*
+ * a0 = cinfo->max_v_samp_factor
+ * a1 = downsampled_width
+ * a2 = input_data
+ * a3 = output_data_ptr
+ */
+    SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4, s5
+
+    li            s4, 0
+    lw            s2, 0(a3)       /* s2 = *output_data_ptr */
+0:
+    li            t9, 2
+    lw            s1, -4(a2)      /* s1 = inptr1 */
+
+1:
+    lw            s0, 0(a2)       /* s0 = inptr0 */
+    lwx           s3, s4(s2)
+    addiu         s5, a1, -2      /* s5 = downsampled_width - 2 */
+    srl           t4, s5, 1
+    sll           t4, t4, 1
+    lbu           t0, 0(s0)
+    lbu           t1, 1(s0)
+    lbu           t2, 0(s1)
+    lbu           t3, 1(s1)
+    addiu         s0, 2
+    addiu         s1, 2
+    addu          t8, s0, t4      /* t8 = end address */
+    andi          s5, s5, 1       /* s5 = residual */
+    sll           t4, t0, 1
+    sll           t6, t1, 1
+    addu          t0, t0, t4      /* t0 = (*inptr0++) * 3 */
+    addu          t1, t1, t6      /* t1 = (*inptr0++) * 3 */
+    addu          t7, t0, t2      /* t7 = thiscolsum */
+    addu          t6, t1, t3      /* t5 = nextcolsum */
+    sll           t0, t7, 2       /* t0 = thiscolsum * 4 */
+    subu          t1, t0, t7      /* t1 = thiscolsum * 3 */
+    shra_r.w      t0, t0, 4
+    addiu         t1, 7
+    addu          t1, t1, t6
+    srl           t1, t1, 4
+    sb            t0, 0(s3)
+    sb            t1, 1(s3)
+    beq           t8, s0, 22f     /* skip to final iteration if width == 3 */
+     addiu        s3, 2
+2:
+    lh            t0, 0(s0)       /* t0 = A3|A2 */
+    lh            t2, 0(s1)       /* t2 = B3|B2 */
+    addiu         s0, 2
+    addiu         s1, 2
+    preceu.ph.qbr t0, t0          /* t0 = 0|A3|0|A2 */
+    preceu.ph.qbr t2, t2          /* t2 = 0|B3|0|B2 */
+    shll.ph       t1, t0, 1
+    sll           t3, t6, 1
+    addu.ph       t0, t1, t0      /* t0 = A3*3|A2*3 */
+    addu          t3, t3, t6      /* t3 = this * 3 */
+    addu.ph       t0, t0, t2      /* t0 = next2|next1 */
+    addu          t1, t3, t7
+    andi          t7, t0, 0xFFFF  /* t7 = next1 */
+    sll           t2, t7, 1
+    addu          t2, t7, t2      /* t2 = next1*3 */
+    addu          t4, t2, t6
+    srl           t6, t0, 16      /* t6 = next2 */
+    shra_r.w      t1, t1, 4       /* t1 = (this*3 + last + 8) >> 4 */
+    addu          t0, t3, t7
+    addiu         t0, 7
+    srl           t0, t0, 4       /* t0 = (this*3 + next1 + 7) >> 4 */
+    shra_r.w      t4, t4, 4       /* t3 = (next1*3 + this + 8) >> 4 */
+    addu          t2, t2, t6
+    addiu         t2, 7
+    srl           t2, t2, 4       /* t2 = (next1*3 + next2 + 7) >> 4 */
+    sb            t1, 0(s3)
+    sb            t0, 1(s3)
+    sb            t4, 2(s3)
+    sb            t2, 3(s3)
+    bne           t8, s0, 2b
+     addiu        s3, 4
+22:
+    beqz          s5, 4f
+     addu         t8, s0, s5
+3:
+    lbu           t0, 0(s0)
+    lbu           t2, 0(s1)
+    addiu         s0, 1
+    addiu         s1, 1
+    sll           t3, t6, 1
+    sll           t1, t0, 1
+    addu          t1, t0, t1      /* t1 = inptr0 * 3 */
+    addu          t3, t3, t6      /* t3 = thiscolsum * 3 */
+    addu          t5, t1, t2
+    addu          t1, t3, t7
+    shra_r.w      t1, t1, 4
+    addu          t0, t3, t5
+    addiu         t0, 7
+    srl           t0, t0, 4
+    sb            t1, 0(s3)
+    sb            t0, 1(s3)
+    addiu         s3, 2
+    move          t7, t6
+    bne           t8, s0, 3b
+     move         t6, t5
+4:
+    sll           t0, t6, 2       /* t0 = thiscolsum * 4 */
+    subu          t1, t0, t6      /* t1 = thiscolsum * 3 */
+    addu          t1, t1, t7
+    addiu         s4, 4
+    shra_r.w      t1, t1, 4
+    addiu         t0, 7
+    srl           t0, t0, 4
+    sb            t1, 0(s3)
+    sb            t0, 1(s3)
+    addiu         t9, -1
+    addiu         s3, 2
+    bnez          t9, 1b
+     lw           s1, 4(a2)
+    srl           t0, s4, 2
+    subu          t0, a0, t0
+    bgtz          t0, 0b
+     addiu        a2, 4
+
+    RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4, s5
+
+    j             ra
+     nop
+END(jsimd_h2v2_fancy_upsample_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_h2v1_fancy_upsample_dspr2)
+/*
+ * a0 = cinfo->max_v_samp_factor
+ * a1 = downsampled_width
+ * a2 = input_data
+ * a3 = output_data_ptr
+ */
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    .set at
+
+    beqz          a0, 3f
+     sll          t0, a0, 2
+    lw            s1, 0(a3)
+    li            s3, 0x10001
+    addu          s0, s1, t0
+0:
+    addiu         t8, a1, -2
+    srl           t9, t8, 2
+    lw            t7, 0(a2)
+    lw            s2, 0(s1)
+    lbu           t0, 0(t7)
+    lbu           t1, 1(t7)       /* t1 = inptr1 */
+    sll           t2, t0, 1
+    addu          t2, t2, t0      /* t2 = invalue*3 */
+    addu          t2, t2, t1
+    shra_r.w      t2, t2, 2
+    sb            t0, 0(s2)
+    sb            t2, 1(s2)
+    beqz          t9, 11f
+     addiu        s2, 2
+1:
+    ulw           t0, 0(t7)       /* t0 = |P3|P2|P1|P0| */
+    ulw           t1, 1(t7)
+    ulh           t2, 4(t7)       /* t2 = |0|0|P5|P4| */
+    preceu.ph.qbl t3, t0          /* t3 = |0|P3|0|P2| */
+    preceu.ph.qbr t0, t0          /* t0 = |0|P1|0|P0| */
+    preceu.ph.qbr t2, t2          /* t2 = |0|P5|0|P4| */
+    preceu.ph.qbl t4, t1          /* t4 = |0|P4|0|P3| */
+    preceu.ph.qbr t1, t1          /* t1 = |0|P2|0|P1| */
+    shll.ph       t5, t4, 1
+    shll.ph       t6, t1, 1
+    addu.ph       t5, t5, t4      /* t5 = |P4*3|P3*3| */
+    addu.ph       t6, t6, t1      /* t6 = |P2*3|P1*3| */
+    addu.ph       t4, t3, s3
+    addu.ph       t0, t0, s3
+    addu.ph       t4, t4, t5
+    addu.ph       t0, t0, t6
+    shrl.ph       t4, t4, 2       /* t4 = |0|P3|0|P2| */
+    shrl.ph       t0, t0, 2       /* t0 = |0|P1|0|P0| */
+    addu.ph       t2, t2, t5
+    addu.ph       t3, t3, t6
+    shra_r.ph     t2, t2, 2       /* t2 = |0|P5|0|P4| */
+    shra_r.ph     t3, t3, 2       /* t3 = |0|P3|0|P2| */
+    shll.ph       t2, t2, 8
+    shll.ph       t3, t3, 8
+    or            t2, t4, t2
+    or            t3, t3, t0
+    addiu         t9, -1
+    usw           t3, 0(s2)
+    usw           t2, 4(s2)
+    addiu         s2, 8
+    bgtz          t9, 1b
+     addiu        t7, 4
+11:
+    andi          t8, 3
+    beqz          t8, 22f
+     addiu        t7, 1
+
+2:
+    lbu           t0, 0(t7)
+    addiu         t7, 1
+    sll           t1, t0, 1
+    addu          t2, t0, t1      /* t2 = invalue */
+    lbu           t3, -2(t7)
+    lbu           t4, 0(t7)
+    addiu         t3, 1
+    addiu         t4, 2
+    addu          t3, t3, t2
+    addu          t4, t4, t2
+    srl           t3, 2
+    srl           t4, 2
+    sb            t3, 0(s2)
+    sb            t4, 1(s2)
+    addiu         t8, -1
+    bgtz          t8, 2b
+     addiu        s2, 2
+
+22:
+    lbu           t0, 0(t7)
+    lbu           t2, -1(t7)
+    sll           t1, t0, 1
+    addu          t1, t1, t0      /* t1 = invalue * 3 */
+    addu          t1, t1, t2
+    addiu         t1, 1
+    srl           t1, t1, 2
+    sb            t1, 0(s2)
+    sb            t0, 1(s2)
+    addiu         s1, 4
+    bne           s1, s0, 0b
+     addiu        a2, 4
+3:
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    j             ra
+     nop
+END(jsimd_h2v1_fancy_upsample_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_h2v1_downsample_dspr2)
+/*
+ * a0     = cinfo->image_width
+ * a1     = cinfo->max_v_samp_factor
+ * a2     = compptr->v_samp_factor
+ * a3     = compptr->width_in_blocks
+ * 16(sp) = input_data
+ * 20(sp) = output_data
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4
+
+    beqz        a2, 7f
+     lw         s1, 44(sp)      /* s1 = output_data */
+    lw          s0, 40(sp)      /* s0 = input_data */
+    srl         s2, a0, 2
+    andi        t9, a0, 2
+    srl         t7, t9, 1
+    addu        s2, t7, s2
+    sll         t0, a3, 3       /* t0 = width_in_blocks*DCT */
+    srl         t7, t0, 1
+    subu        s2, t7, s2
+0:
+    andi        t6, a0, 1       /* t6 = temp_index */
+    addiu       t6, -1
+    lw          t4, 0(s1)       /* t4 = outptr */
+    lw          t5, 0(s0)       /* t5 = inptr0 */
+    li          s3, 0           /* s3 = bias */
+    srl         t7, a0, 1       /* t7 = image_width1 */
+    srl         s4, t7, 2
+    andi        t8, t7, 3
+1:
+    ulhu        t0, 0(t5)
+    ulhu        t1, 2(t5)
+    ulhu        t2, 4(t5)
+    ulhu        t3, 6(t5)
+    raddu.w.qb  t0, t0
+    raddu.w.qb  t1, t1
+    raddu.w.qb  t2, t2
+    raddu.w.qb  t3, t3
+    shra.ph     t0, t0, 1
+    shra_r.ph   t1, t1, 1
+    shra.ph     t2, t2, 1
+    shra_r.ph   t3, t3, 1
+    sb          t0, 0(t4)
+    sb          t1, 1(t4)
+    sb          t2, 2(t4)
+    sb          t3, 3(t4)
+    addiu       s4, -1
+    addiu       t4, 4
+    bgtz        s4, 1b
+     addiu      t5, 8
+    beqz        t8, 3f
+     addu       s4, t4, t8
+2:
+    ulhu        t0, 0(t5)
+    raddu.w.qb  t0, t0
+    addqh.w     t0, t0, s3
+    xori        s3, s3, 1
+    sb          t0, 0(t4)
+    addiu       t4, 1
+    bne         t4, s4, 2b
+     addiu      t5, 2
+3:
+    lbux        t1, t6(t5)
+    sll         t1, 1
+    addqh.w     t2, t1, s3      /* t2 = pixval1 */
+    xori        s3, s3, 1
+    addqh.w     t3, t1, s3      /* t3 = pixval2 */
+    blez        s2, 5f
+     append     t3, t2,  8
+    addu        t5, t4, s2      /* t5 = loop_end2 */
+4:
+    ush         t3, 0(t4)
+    addiu       s2, -1
+    bgtz        s2, 4b
+     addiu      t4,  2
+5:
+    beqz        t9, 6f
+     nop
+    sb          t2, 0(t4)
+6:
+    addiu       s1, 4
+    addiu       a2, -1
+    bnez        a2, 0b
+     addiu      s0, 4
+7:
+    RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4
+
+    j           ra
+    nop
+END(jsimd_h2v1_downsample_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_h2v2_downsample_dspr2)
+/*
+ * a0     = cinfo->image_width
+ * a1     = cinfo->max_v_samp_factor
+ * a2     = compptr->v_samp_factor
+ * a3     = compptr->width_in_blocks
+ * 16(sp) = input_data
+ * 20(sp) = output_data
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    beqz        a2, 8f
+     lw         s1, 52(sp)      /* s1 = output_data */
+    lw          s0, 48(sp)      /* s0 = input_data */
+
+    andi        t6, a0, 1       /* t6 = temp_index */
+    addiu       t6, -1
+    srl         t7, a0, 1       /* t7 = image_width1 */
+    srl         s4, t7, 2
+    andi        t8, t7, 3
+    andi        t9, a0, 2
+    srl         s2, a0, 2
+    srl         t7, t9, 1
+    addu        s2, t7, s2
+    sll         t0, a3, 3       /* s2 = width_in_blocks*DCT */
+    srl         t7, t0, 1
+    subu        s2, t7, s2
+0:
+    lw          t4, 0(s1)       /* t4 = outptr */
+    lw          t5, 0(s0)       /* t5 = inptr0 */
+    lw          s7, 4(s0)       /* s7 = inptr1 */
+    li          s6, 1           /* s6 = bias */
+2:
+    ulw         t0, 0(t5)       /* t0 = |P3|P2|P1|P0| */
+    ulw         t1, 0(s7)       /* t1 = |Q3|Q2|Q1|Q0| */
+    ulw         t2, 4(t5)
+    ulw         t3, 4(s7)
+    precrq.ph.w t7, t0, t1      /* t2 = |P3|P2|Q3|Q2| */
+    ins         t0, t1, 16, 16  /* t0 = |Q1|Q0|P1|P0| */
+    raddu.w.qb  t1, t7
+    raddu.w.qb  t0, t0
+    shra_r.w    t1, t1, 2
+    addiu       t0, 1
+    srl         t0, 2
+    precrq.ph.w t7, t2, t3
+    ins         t2, t3, 16, 16
+    raddu.w.qb  t7, t7
+    raddu.w.qb  t2, t2
+    shra_r.w    t7, t7, 2
+    addiu       t2, 1
+    srl         t2, 2
+    sb          t0, 0(t4)
+    sb          t1, 1(t4)
+    sb          t2, 2(t4)
+    sb          t7, 3(t4)
+    addiu       t4, 4
+    addiu       t5, 8
+    addiu       s4, s4, -1
+    bgtz        s4, 2b
+     addiu      s7, 8
+    beqz        t8, 4f
+     addu       t8, t4, t8
+3:
+    ulhu        t0, 0(t5)
+    ulhu        t1, 0(s7)
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t0, t0
+    addu        t0, t0, s6
+    srl         t0, 2
+    xori        s6, s6, 3
+    sb          t0, 0(t4)
+    addiu       t5, 2
+    addiu       t4, 1
+    bne         t8, t4, 3b
+     addiu      s7, 2
+4:
+    lbux        t1, t6(t5)
+    sll         t1, 1
+    lbux        t0, t6(s7)
+    sll         t0, 1
+    addu        t1, t1, t0
+    addu        t3, t1, s6
+    srl         t0, t3, 2       /* t2 = pixval1 */
+    xori        s6, s6, 3
+    addu        t2, t1, s6
+    srl         t1, t2, 2       /* t3 = pixval2 */
+    blez        s2, 6f
+     append     t1, t0, 8
+5:
+    ush         t1, 0(t4)
+    addiu       s2, -1
+    bgtz        s2, 5b
+     addiu      t4, 2
+6:
+    beqz        t9, 7f
+     nop
+    sb          t0, 0(t4)
+7:
+    addiu       s1, 4
+    addiu       a2, -1
+    bnez        a2, 0b
+     addiu      s0, 8
+8:
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+END(jsimd_h2v2_downsample_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_h2v2_smooth_downsample_dspr2)
+/*
+ * a0     = input_data
+ * a1     = output_data
+ * a2     = compptr->v_samp_factor
+ * a3     = cinfo->max_v_samp_factor
+ * 16(sp) = cinfo->smoothing_factor
+ * 20(sp) = compptr->width_in_blocks
+ * 24(sp) = cinfo->image_width
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw          s7, 52(sp)      /* compptr->width_in_blocks */
+    lw          s0, 56(sp)      /* cinfo->image_width */
+    lw          s6, 48(sp)      /* cinfo->smoothing_factor */
+    sll         s7, 3           /* output_cols = width_in_blocks * DCTSIZE */
+    sll         v0, s7, 1
+    subu        v0, v0, s0
+    blez        v0, 2f
+    move        v1, zero
+    addiu       t0, a3, 2       /* t0 = cinfo->max_v_samp_factor + 2 */
+0:
+    addiu       t1, a0, -4
+    sll         t2, v1, 2
+    lwx         t1, t2(t1)
+    move        t3, v0
+    addu        t1, t1, s0
+    lbu         t2, -1(t1)
+1:
+    addiu       t3, t3, -1
+    sb          t2, 0(t1)
+    bgtz        t3, 1b
+    addiu       t1, t1, 1
+    addiu       v1, v1, 1
+    bne         v1, t0, 0b
+    nop
+2:
+    li          v0, 80
+    mul         v0, s6, v0
+    li          v1, 16384
+    move        t4, zero
+    move        t5, zero
+    subu        t6, v1, v0      /* t6 = 16384 - tmp_smoot_f * 80 */
+    sll         t7, s6, 4       /* t7 = tmp_smoot_f * 16 */
+3:
+/* Special case for first column: pretend column -1 is same as column 0 */
+    sll         v0, t4, 2
+    lwx         t8, v0(a1)      /*  outptr = output_dataoutrow */
+    sll         v1, t5, 2
+    addiu       t9, v1, 4
+    addiu       s0, v1, -4
+    addiu       s1, v1, 8
+    lwx         s2, v1(a0)      /* inptr0 = input_datainrow */
+    lwx         t9, t9(a0)      /* inptr1 = input_datainrow+1 */
+    lwx         s0, s0(a0)      /* above_ptr = input_datainrow-1 */
+    lwx         s1, s1(a0)      /* below_ptr = input_datainrow+2 */
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, 0(s2)
+    lbu         v1, 2(s2)
+    lbu         t0, 0(t9)
+    lbu         t1, 2(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 2(s0)
+    addu        t0, t0, v0
+    lbu         t3, 2(s1)
+    addu        s3, t0, s3
+    lbu         v0, 0(s0)
+    lbu         t0, 0(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    addu        t0, t0, v0
+    addu        s3, t0, s3
+    madd        $ac1, s3, t7
+    extr_r.w    v0, $ac1, 16
+    addiu       t8, t8, 1
+    addiu       s2, s2, 2
+    addiu       t9, t9, 2
+    addiu       s0, s0, 2
+    addiu       s1, s1, 2
+    sb          v0, -1(t8)
+    addiu       s4, s7, -2
+    and         s4, s4, 3
+    addu        s5, s4, t8      /* end address */
+4:
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, -1(s2)
+    lbu         v1, 2(s2)
+    lbu         t0, -1(t9)
+    lbu         t1, 2(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 2(s0)
+    addu        t0, t0, v0
+    lbu         t3, 2(s1)
+    addu        s3, t0, s3
+    lbu         v0, -1(s0)
+    lbu         t0, -1(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    addu        t0, t0, v0
+    addu        s3, t0, s3
+    madd        $ac1, s3, t7
+    extr_r.w    t2, $ac1, 16
+    addiu       t8, t8, 1
+    addiu       s2, s2, 2
+    addiu       t9, t9, 2
+    addiu       s0, s0, 2
+    sb          t2, -1(t8)
+    bne         s5, t8, 4b
+    addiu       s1, s1, 2
+    addiu       s5, s7, -2
+    subu        s5, s5, s4
+    addu        s5, s5, t8      /* end address */
+5:
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, -1(s2)
+    lbu         v1, 2(s2)
+    lbu         t0, -1(t9)
+    lbu         t1, 2(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 2(s0)
+    addu        t0, t0, v0
+    lbu         t3, 2(s1)
+    addu        s3, t0, s3
+    lbu         v0, -1(s0)
+    lbu         t0, -1(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    lh          v1, 2(t9)
+    addu        t0, t0, v0
+    lh          v0, 2(s2)
+    addu        s3, t0, s3
+    lh          t0, 2(s0)
+    lh          t1, 2(s1)
+    madd        $ac1, s3, t7
+    extr_r.w    t2, $ac1, 16
+    ins         t0, t1, 16, 16
+    ins         v0, v1, 16, 16
+    raddu.w.qb  s3, t0
+    lbu         v1, 4(s2)
+    lbu         t0, 1(t9)
+    lbu         t1, 4(t9)
+    sb          t2, 0(t8)
+    raddu.w.qb  t3, v0
+    lbu         v0, 1(s2)
+    addu        t0, t0, t1
+    mult        $ac1, t3, t6
+    addu        v0, v0, v1
+    lbu         t2, 4(s0)
+    addu        t0, t0, v0
+    lbu         v0, 1(s0)
+    addu        s3, t0, s3
+    lbu         t0, 1(s1)
+    lbu         t3, 4(s1)
+    addu        v0, v0, t2
+    sll         s3, s3, 1
+    addu        t0, t0, t3
+    lh          v1, 4(t9)
+    addu        t0, t0, v0
+    lh          v0, 4(s2)
+    addu        s3, t0, s3
+    lh          t0, 4(s0)
+    lh          t1, 4(s1)
+    madd        $ac1, s3, t7
+    extr_r.w    t2, $ac1, 16
+    ins         t0, t1, 16, 16
+    ins         v0, v1, 16, 16
+    raddu.w.qb  s3, t0
+    lbu         v1, 6(s2)
+    lbu         t0, 3(t9)
+    lbu         t1, 6(t9)
+    sb          t2, 1(t8)
+    raddu.w.qb  t3, v0
+    lbu         v0, 3(s2)
+    addu        t0, t0, t1
+    mult        $ac1, t3, t6
+    addu        v0, v0, v1
+    lbu         t2, 6(s0)
+    addu        t0, t0, v0
+    lbu         v0, 3(s0)
+    addu        s3, t0, s3
+    lbu         t0, 3(s1)
+    lbu         t3, 6(s1)
+    addu        v0, v0, t2
+    sll         s3, s3, 1
+    addu        t0, t0, t3
+    lh          v1, 6(t9)
+    addu        t0, t0, v0
+    lh          v0, 6(s2)
+    addu        s3, t0, s3
+    lh          t0, 6(s0)
+    lh          t1, 6(s1)
+    madd        $ac1, s3, t7
+    extr_r.w    t3, $ac1, 16
+    ins         t0, t1, 16, 16
+    ins         v0, v1, 16, 16
+    raddu.w.qb  s3, t0
+    lbu         v1, 8(s2)
+    lbu         t0, 5(t9)
+    lbu         t1, 8(t9)
+    sb          t3, 2(t8)
+    raddu.w.qb  t2, v0
+    lbu         v0, 5(s2)
+    addu        t0, t0, t1
+    mult        $ac1, t2, t6
+    addu        v0, v0, v1
+    lbu         t2, 8(s0)
+    addu        t0, t0, v0
+    lbu         v0, 5(s0)
+    addu        s3, t0, s3
+    lbu         t0, 5(s1)
+    lbu         t3, 8(s1)
+    addu        v0, v0, t2
+    sll         s3, s3, 1
+    addu        t0, t0, t3
+    addiu       t8, t8, 4
+    addu        t0, t0, v0
+    addiu       s2, s2, 8
+    addu        s3, t0, s3
+    addiu       t9, t9, 8
+    madd        $ac1, s3, t7
+    extr_r.w    t1, $ac1, 16
+    addiu       s0, s0, 8
+    addiu       s1, s1, 8
+    bne         s5, t8, 5b
+    sb          t1, -1(t8)
+/* Special case for last column */
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, -1(s2)
+    lbu         v1, 1(s2)
+    lbu         t0, -1(t9)
+    lbu         t1, 1(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 1(s0)
+    addu        t0, t0, v0
+    lbu         t3, 1(s1)
+    addu        s3, t0, s3
+    lbu         v0, -1(s0)
+    lbu         t0, -1(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    addu        t0, t0, v0
+    addu        s3, t0, s3
+    madd        $ac1, s3, t7
+    extr_r.w    t0, $ac1, 16
+    addiu       t5, t5, 2
+    sb          t0, 0(t8)
+    addiu       t4, t4, 1
+    bne         t4, a2, 3b
+    addiu       t5, t5, 2
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+
+END(jsimd_h2v2_smooth_downsample_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_int_upsample_dspr2)
+/*
+ * a0     = upsample->h_expandcompptr->component_index
+ * a1     = upsample->v_expandcompptr->component_index
+ * a2     = input_data
+ * a3     = output_data_ptr
+ * 16(sp) = cinfo->output_width
+ * 20(sp) = cinfo->max_v_samp_factor
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    lw          s0, 0(a3)       /* s0 = output_data */
+    lw          s1, 32(sp)      /* s1 = cinfo->output_width */
+    lw          s2, 36(sp)      /* s2 = cinfo->max_v_samp_factor */
+    li          t6, 0           /* t6 = inrow */
+    beqz        s2, 10f
+     li         s3, 0           /* s3 = outrow */
+0:
+    addu        t0, a2, t6
+    addu        t7, s0, s3
+    lw          t3, 0(t0)       /* t3 = inptr */
+    lw          t8, 0(t7)       /* t8 = outptr */
+    beqz        s1, 4f
+     addu       t5, t8, s1      /* t5 = outend */
+1:
+    lb          t2, 0(t3)       /* t2 = invalue = *inptr++ */
+    addiu       t3, 1
+    beqz        a0, 3f
+     move       t0, a0          /* t0 = h_expand */
+2:
+    sb          t2, 0(t8)
+    addiu       t0, -1
+    bgtz        t0, 2b
+     addiu      t8, 1
+3:
+    bgt         t5, t8, 1b
+     nop
+4:
+    addiu       t9, a1, -1      /* t9 = v_expand - 1 */
+    blez        t9, 9f
+     nop
+5:
+    lw          t3, 0(s0)
+    lw          t4, 4(s0)
+    subu        t0, s1, 0xF
+    blez        t0, 7f
+     addu       t5, t3, s1      /* t5 = end address */
+    andi        t7, s1, 0xF     /* t7 = residual */
+    subu        t8, t5, t7
+6:
+    ulw         t0, 0(t3)
+    ulw         t1, 4(t3)
+    ulw         t2, 8(t3)
+    usw         t0, 0(t4)
+    ulw         t0, 12(t3)
+    usw         t1, 4(t4)
+    usw         t2, 8(t4)
+    usw         t0, 12(t4)
+    addiu       t3, 16
+    bne         t3, t8, 6b
+     addiu      t4, 16
+    beqz        t7, 8f
+     nop
+7:
+    lbu         t0, 0(t3)
+    sb          t0, 0(t4)
+    addiu       t3, 1
+    bne         t3, t5, 7b
+     addiu      t4, 1
+8:
+    addiu       t9, -1
+    bgtz        t9, 5b
+     addiu      s0, 8
+9:
+    addu        s3, s3, a1
+    bne         s3, s2, 0b
+     addiu      t6, 1
+10:
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    j           ra
+     nop
+END(jsimd_int_upsample_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_h2v1_upsample_dspr2)
+/*
+ * a0 = cinfo->max_v_samp_factor
+ * a1 = cinfo->output_width
+ * a2 = input_data
+ * a3 = output_data_ptr
+ */
+    lw          t7, 0(a3)       /* t7 = output_data */
+    andi        t8, a1, 0xf     /* t8 = residual */
+    sll         t0, a0, 2
+    blez        a0, 4f
+     addu       t9, t7, t0      /* t9 = output_data end address */
+0:
+    lw          t5, 0(t7)       /* t5 = outptr */
+    lw          t6, 0(a2)       /* t6 = inptr */
+    addu        t3, t5, a1      /* t3 = outptr + output_width (end address) */
+    subu        t3, t8          /* t3 = end address - residual */
+    beq         t5, t3, 2f
+     move       t4, t8
+1:
+    ulw         t0, 0(t6)       /* t0 = |P3|P2|P1|P0| */
+    ulw         t2, 4(t6)       /* t2 = |P7|P6|P5|P4| */
+    srl         t1, t0, 16      /* t1 = |X|X|P3|P2| */
+    ins         t0, t0, 16, 16  /* t0 = |P1|P0|P1|P0| */
+    ins         t1, t1, 16, 16  /* t1 = |P3|P2|P3|P2| */
+    ins         t0, t0, 8, 16   /* t0 = |P1|P1|P0|P0| */
+    ins         t1, t1, 8, 16   /* t1 = |P3|P3|P2|P2| */
+    usw         t0, 0(t5)
+    usw         t1, 4(t5)
+    srl         t0, t2, 16      /* t0 = |X|X|P7|P6| */
+    ins         t2, t2, 16, 16  /* t2 = |P5|P4|P5|P4| */
+    ins         t0, t0, 16, 16  /* t0 = |P7|P6|P7|P6| */
+    ins         t2, t2, 8, 16   /* t2 = |P5|P5|P4|P4| */
+    ins         t0, t0, 8, 16   /* t0 = |P7|P7|P6|P6| */
+    usw         t2, 8(t5)
+    usw         t0, 12(t5)
+    addiu       t5, 16
+    bne         t5, t3, 1b
+     addiu      t6, 8
+    beqz        t8, 3f
+     move       t4, t8
+2:
+    lbu         t1, 0(t6)
+    sb          t1, 0(t5)
+    sb          t1, 1(t5)
+    addiu       t4, -2
+    addiu       t6, 1
+    bgtz        t4, 2b
+     addiu      t5, 2
+3:
+    addiu       t7, 4
+    bne         t9, t7, 0b
+     addiu      a2, 4
+4:
+    j           ra
+     nop
+END(jsimd_h2v1_upsample_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_h2v2_upsample_dspr2)
+/*
+ * a0 = cinfo->max_v_samp_factor
+ * a1 = cinfo->output_width
+ * a2 = input_data
+ * a3 = output_data_ptr
+ */
+    lw          t7, 0(a3)
+    blez        a0, 7f
+     andi       t9, a1, 0xf     /* t9 = residual */
+0:
+    lw          t6, 0(a2)       /* t6 = inptr */
+    lw          t5, 0(t7)       /* t5 = outptr */
+    addu        t8, t5, a1      /* t8 = outptr end address */
+    subu        t8, t9          /* t8 = end address - residual */
+    beq         t5, t8, 2f
+     move       t4, t9
+1:
+    ulw         t0, 0(t6)
+    srl         t1, t0, 16
+    ins         t0, t0, 16, 16
+    ins         t0, t0, 8, 16
+    ins         t1, t1, 16, 16
+    ins         t1, t1, 8, 16
+    ulw         t2, 4(t6)
+    usw         t0, 0(t5)
+    usw         t1, 4(t5)
+    srl         t3, t2, 16
+    ins         t2, t2, 16, 16
+    ins         t2, t2, 8, 16
+    ins         t3, t3, 16, 16
+    ins         t3, t3, 8, 16
+    usw         t2, 8(t5)
+    usw         t3, 12(t5)
+    addiu       t5, 16
+    bne         t5, t8, 1b
+     addiu      t6, 8
+    beqz        t9, 3f
+     move       t4, t9
+2:
+    lbu         t0, 0(t6)
+    sb          t0, 0(t5)
+    sb          t0, 1(t5)
+    addiu       t4, -2
+    addiu       t6, 1
+    bgtz        t4, 2b
+     addiu      t5, 2
+3:
+    lw          t6, 0(t7)       /* t6 = outptr0 */
+    lw          t5, 4(t7)       /* t5 = outptr1 */
+    addu        t4, t6, a1      /* t4 = new end address */
+    beq         a1, t9, 5f
+     subu       t8, t4, t9
+4:
+    ulw         t0, 0(t6)
+    ulw         t1, 4(t6)
+    ulw         t2, 8(t6)
+    usw         t0, 0(t5)
+    ulw         t0, 12(t6)
+    usw         t1, 4(t5)
+    usw         t2, 8(t5)
+    usw         t0, 12(t5)
+    addiu       t6, 16
+    bne         t6, t8, 4b
+     addiu      t5, 16
+    beqz        t9, 6f
+     nop
+5:
+    lbu         t0, 0(t6)
+    sb          t0, 0(t5)
+    addiu       t6, 1
+    bne         t6, t4, 5b
+     addiu      t5, 1
+6:
+    addiu       t7, 8
+    addiu       a0, -2
+    bgtz        a0, 0b
+     addiu      a2, 4
+7:
+    j           ra
+     nop
+END(jsimd_h2v2_upsample_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_idct_islow_dspr2)
+/*
+ * a0 = coef_block
+ * a1 = compptr->dcttable
+ * a2 = output
+ * a3 = range_limit
+ */
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    addiu       sp, sp, -256
+    move        v0, sp
+    addiu       v1, zero, 8     /* v1 = DCTSIZE = 8 */
+1:
+    lh          s4, 32(a0)      /* s4 = inptr16 */
+    lh          s5, 64(a0)      /* s5 = inptr32 */
+    lh          s6, 96(a0)      /* s6 = inptr48 */
+    lh          t1, 112(a0)     /* t1 = inptr56 */
+    lh          t7, 16(a0)      /* t7 = inptr8 */
+    lh          t5, 80(a0)      /* t5 = inptr40 */
+    lh          t3, 48(a0)      /* t3 = inptr24 */
+    or          s4, s4, t1
+    or          s4, s4, t3
+    or          s4, s4, t5
+    or          s4, s4, t7
+    or          s4, s4, s5
+    or          s4, s4, s6
+    bnez        s4, 2f
+     addiu      v1, v1, -1
+    lh          s5, 0(a1)       /* quantptrDCTSIZE*0 */
+    lh          s6, 0(a0)       /* inptrDCTSIZE*0 */
+    mul         s5, s5, s6      /* DEQUANTIZE(inptr0, quantptr0) */
+    sll         s5, s5, 2
+    sw          s5, 0(v0)
+    sw          s5, 32(v0)
+    sw          s5, 64(v0)
+    sw          s5, 96(v0)
+    sw          s5, 128(v0)
+    sw          s5, 160(v0)
+    sw          s5, 192(v0)
+    b           3f
+     sw         s5, 224(v0)
+2:
+    lh          t0, 112(a1)
+    lh          t2, 48(a1)
+    lh          t4, 80(a1)
+    lh          t6, 16(a1)
+    mul         t0, t0, t1      /* DEQUANTIZE(inptrDCTSIZE*7,
+                                              quantptrDCTSIZE*7) */
+    mul         t1, t2, t3      /* DEQUANTIZE(inptrDCTSIZE*3,
+                                              quantptrDCTSIZE*3) */
+    mul         t2, t4, t5      /* DEQUANTIZE(inptrDCTSIZE*5,
+                                              quantptrDCTSIZE*5) */
+    mul         t3, t6, t7      /* DEQUANTIZE(inptrDCTSIZE*1,
+                                              quantptrDCTSIZE*1) */
+    lh          t4, 32(a1)
+    lh          t5, 32(a0)
+    lh          t6, 96(a1)
+    lh          t7, 96(a0)
+    addu        s0, t0, t1       /* z3 = tmp0 + tmp2 */
+    addu        s1, t1, t2       /* z2 = tmp1 + tmp2 */
+    addu        s2, t2, t3       /* z4 = tmp1 + tmp3 */
+    addu        s3, s0, s2       /* z3 + z4 */
+    addiu       t9, zero, 9633   /* FIX_1_175875602 */
+    mul         s3, s3, t9       /* z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    addu        t8, t0, t3       /* z1 = tmp0 + tmp3 */
+    addiu       t9, zero, 2446   /* FIX_0_298631336 */
+    mul         t0, t0, t9       /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    addiu       t9, zero, 16819  /* FIX_2_053119869 */
+    mul         t2, t2, t9       /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    addiu       t9, zero, 25172  /* FIX_3_072711026 */
+    mul         t1, t1, t9       /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    addiu       t9, zero, 12299  /* FIX_1_501321110 */
+    mul         t3, t3, t9       /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    addiu       t9, zero, 16069  /* FIX_1_961570560 */
+    mul         s0, s0, t9       /* -z3 = MULTIPLY(z3, FIX_1_961570560) */
+    addiu       t9, zero, 3196   /* FIX_0_390180644 */
+    mul         s2, s2, t9       /* -z4 = MULTIPLY(z4, FIX_0_390180644) */
+    addiu       t9, zero, 7373   /* FIX_0_899976223 */
+    mul         t8, t8, t9       /* -z1 = MULTIPLY(z1, FIX_0_899976223) */
+    addiu       t9, zero, 20995  /* FIX_2_562915447 */
+    mul         s1, s1, t9       /* -z2 = MULTIPLY(z2, FIX_2_562915447) */
+    subu        s0, s3, s0       /* z3 += z5 */
+    addu        t0, t0, s0       /* tmp0 += z3 */
+    addu        t1, t1, s0       /* tmp2 += z3 */
+    subu        s2, s3, s2       /* z4 += z5 */
+    addu        t2, t2, s2       /* tmp1 += z4 */
+    addu        t3, t3, s2       /* tmp3 += z4 */
+    subu        t0, t0, t8       /* tmp0 += z1 */
+    subu        t1, t1, s1       /* tmp2 += z2 */
+    subu        t2, t2, s1       /* tmp1 += z2 */
+    subu        t3, t3, t8       /* tmp3 += z1 */
+    mul         s0, t4, t5       /* DEQUANTIZE(inptrDCTSIZE*2,
+                                               quantptrDCTSIZE*2) */
+    addiu       t9, zero, 6270   /* FIX_0_765366865 */
+    mul         s1, t6, t7       /* DEQUANTIZE(inptrDCTSIZE*6,
+                                               quantptrDCTSIZE*6) */
+    lh          t4, 0(a1)
+    lh          t5, 0(a0)
+    lh          t6, 64(a1)
+    lh          t7, 64(a0)
+    mul         s2, t9, s0       /* MULTIPLY(z2, FIX_0_765366865) */
+    mul         t5, t4, t5       /* DEQUANTIZE(inptrDCTSIZE*0,
+                                               quantptrDCTSIZE*0) */
+    mul         t6, t6, t7       /* DEQUANTIZE(inptrDCTSIZE*4,
+                                               quantptrDCTSIZE*4) */
+    addiu       t9, zero, 4433   /* FIX_0_541196100 */
+    addu        s3, s0, s1       /* z2 + z3 */
+    mul         s3, s3, t9       /* z1 = MULTIPLY(z2 + z3, FIX_0_541196100) */
+    addiu       t9, zero, 15137  /* FIX_1_847759065 */
+    mul         t8, s1, t9       /* MULTIPLY(z3, FIX_1_847759065) */
+    addu        t4, t5, t6
+    subu        t5, t5, t6
+    sll         t4, t4, 13      /* tmp0 = (z2 + z3) << CONST_BITS */
+    sll         t5, t5, 13      /* tmp1 = (z2 - z3) << CONST_BITS */
+    addu        t7, s3, s2      /* tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865) */
+    subu        t6, s3, t8      /* tmp2 =
+                                     z1 + MULTIPLY(z3, -FIX_1_847759065) */
+    addu        s0, t4, t7
+    subu        s1, t4, t7
+    addu        s2, t5, t6
+    subu        s3, t5, t6
+    addu        t4, s0, t3
+    subu        s0, s0, t3
+    addu        t3, s2, t1
+    subu        s2, s2, t1
+    addu        t1, s3, t2
+    subu        s3, s3, t2
+    addu        t2, s1, t0
+    subu        s1, s1, t0
+    shra_r.w    t4, t4, 11
+    shra_r.w    t3, t3, 11
+    shra_r.w    t1, t1, 11
+    shra_r.w    t2, t2, 11
+    shra_r.w    s1, s1, 11
+    shra_r.w    s3, s3, 11
+    shra_r.w    s2, s2, 11
+    shra_r.w    s0, s0, 11
+    sw          t4, 0(v0)
+    sw          t3, 32(v0)
+    sw          t1, 64(v0)
+    sw          t2, 96(v0)
+    sw          s1, 128(v0)
+    sw          s3, 160(v0)
+    sw          s2, 192(v0)
+    sw          s0, 224(v0)
+3:
+    addiu       a1, a1, 2
+    addiu       a0, a0, 2
+    bgtz        v1, 1b
+     addiu      v0, v0, 4
+    move        v0, sp
+    addiu       v1, zero, 8
+4:
+    lw          t0, 8(v0)       /* z2 = (JLONG)wsptr2 */
+    lw          t1, 24(v0)      /* z3 = (JLONG)wsptr6 */
+    lw          t2, 0(v0)       /* (JLONG)wsptr0 */
+    lw          t3, 16(v0)      /* (JLONG)wsptr4 */
+    lw          s4, 4(v0)       /* (JLONG)wsptr1 */
+    lw          s5, 12(v0)      /* (JLONG)wsptr3 */
+    lw          s6, 20(v0)      /* (JLONG)wsptr5 */
+    lw          s7, 28(v0)      /* (JLONG)wsptr7 */
+    or          s4, s4, t0
+    or          s4, s4, t1
+    or          s4, s4, t3
+    or          s4, s4, s7
+    or          s4, s4, s5
+    or          s4, s4, s6
+    bnez        s4, 5f
+     addiu      v1, v1, -1
+    shra_r.w    s5, t2, 5
+    andi        s5, s5, 0x3ff
+    lbux        s5, s5(a3)
+    lw          s1, 0(a2)
+    replv.qb    s5, s5
+    usw         s5, 0(s1)
+    usw         s5, 4(s1)
+    b           6f
+     nop
+5:
+    addu        t4, t0, t1       /* z2 + z3 */
+    addiu       t8, zero, 4433   /* FIX_0_541196100 */
+    mul         t5, t4, t8       /* z1 = MULTIPLY(z2 + z3, FIX_0_541196100) */
+    addiu       t8, zero, 15137  /* FIX_1_847759065 */
+    mul         t1, t1, t8       /* MULTIPLY(z3, FIX_1_847759065) */
+    addiu       t8, zero, 6270   /* FIX_0_765366865 */
+    mul         t0, t0, t8       /* MULTIPLY(z2, FIX_0_765366865) */
+    addu        t4, t2, t3       /* (JLONG)wsptr0 + (JLONG)wsptr4 */
+    subu        t2, t2, t3       /* (JLONG)wsptr0 - (JLONG)wsptr4 */
+    sll         t4, t4, 13       /* tmp0 =
+                                      (wsptr0 + wsptr4) << CONST_BITS */
+    sll         t2, t2, 13       /* tmp1 =
+                                      (wsptr0 - wsptr4) << CONST_BITS */
+    subu        t1, t5, t1       /* tmp2 =
+                                      z1 + MULTIPLY(z3, -FIX_1_847759065) */
+    subu        t3, t2, t1       /* tmp12 = tmp1 - tmp2 */
+    addu        t2, t2, t1       /* tmp11 = tmp1 + tmp2 */
+    addu        t5, t5, t0       /* tmp3 =
+                                      z1 + MULTIPLY(z2, FIX_0_765366865) */
+    subu        t1, t4, t5       /* tmp13 = tmp0 - tmp3 */
+    addu        t0, t4, t5       /* tmp10 = tmp0 + tmp3 */
+    lw          t4, 28(v0)       /* tmp0 = (JLONG)wsptr7 */
+    lw          t6, 12(v0)       /* tmp2 = (JLONG)wsptr3 */
+    lw          t5, 20(v0)       /* tmp1 = (JLONG)wsptr5 */
+    lw          t7, 4(v0)        /* tmp3 = (JLONG)wsptr1 */
+    addu        s0, t4, t6       /* z3 = tmp0 + tmp2 */
+    addiu       t8, zero, 9633   /* FIX_1_175875602 */
+    addu        s1, t5, t7       /* z4 = tmp1 + tmp3 */
+    addu        s2, s0, s1       /* z3 + z4 */
+    mul         s2, s2, t8       /* z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    addu        s3, t4, t7       /* z1 = tmp0 + tmp3 */
+    addu        t9, t5, t6       /* z2 = tmp1 + tmp2 */
+    addiu       t8, zero, 16069  /* FIX_1_961570560 */
+    mul         s0, s0, t8       /* -z3 = MULTIPLY(z3, FIX_1_961570560) */
+    addiu       t8, zero, 3196   /* FIX_0_390180644 */
+    mul         s1, s1, t8       /* -z4 = MULTIPLY(z4, FIX_0_390180644) */
+    addiu       t8, zero, 2446   /* FIX_0_298631336 */
+    mul         t4, t4, t8       /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    addiu       t8, zero, 7373   /* FIX_0_899976223 */
+    mul         s3, s3, t8       /* -z1 = MULTIPLY(z1, FIX_0_899976223) */
+    addiu       t8, zero, 16819  /* FIX_2_053119869 */
+    mul         t5, t5, t8       /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    addiu       t8, zero, 20995  /* FIX_2_562915447 */
+    mul         t9, t9, t8       /* -z2 = MULTIPLY(z2, FIX_2_562915447) */
+    addiu       t8, zero, 25172  /* FIX_3_072711026 */
+    mul         t6, t6, t8       /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    addiu       t8, zero, 12299  /* FIX_1_501321110 */
+    mul         t7, t7, t8       /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    subu        s0, s2, s0       /* z3 += z5 */
+    subu        s1, s2, s1       /* z4 += z5 */
+    addu        t4, t4, s0
+    subu        t4, t4, s3      /* tmp0 */
+    addu        t5, t5, s1
+    subu        t5, t5, t9      /* tmp1 */
+    addu        t6, t6, s0
+    subu        t6, t6, t9      /* tmp2 */
+    addu        t7, t7, s1
+    subu        t7, t7, s3      /* tmp3 */
+    addu        s0, t0, t7
+    subu        t0, t0, t7
+    addu        t7, t2, t6
+    subu        t2, t2, t6
+    addu        t6, t3, t5
+    subu        t3, t3, t5
+    addu        t5, t1, t4
+    subu        t1, t1, t4
+    shra_r.w    s0, s0, 18
+    shra_r.w    t7, t7, 18
+    shra_r.w    t6, t6, 18
+    shra_r.w    t5, t5, 18
+    shra_r.w    t1, t1, 18
+    shra_r.w    t3, t3, 18
+    shra_r.w    t2, t2, 18
+    shra_r.w    t0, t0, 18
+    andi        s0, s0, 0x3ff
+    andi        t7, t7, 0x3ff
+    andi        t6, t6, 0x3ff
+    andi        t5, t5, 0x3ff
+    andi        t1, t1, 0x3ff
+    andi        t3, t3, 0x3ff
+    andi        t2, t2, 0x3ff
+    andi        t0, t0, 0x3ff
+    lw          s1, 0(a2)
+    lbux        s0, s0(a3)
+    lbux        t7, t7(a3)
+    lbux        t6, t6(a3)
+    lbux        t5, t5(a3)
+    lbux        t1, t1(a3)
+    lbux        t3, t3(a3)
+    lbux        t2, t2(a3)
+    lbux        t0, t0(a3)
+    sb          s0, 0(s1)
+    sb          t7, 1(s1)
+    sb          t6, 2(s1)
+    sb          t5, 3(s1)
+    sb          t1, 4(s1)
+    sb          t3, 5(s1)
+    sb          t2, 6(s1)
+    sb          t0, 7(s1)
+6:
+    addiu       v0, v0, 32
+    bgtz        v1, 4b
+     addiu      a2, a2, 4
+    addiu       sp, sp, 256
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+
+END(jsimd_idct_islow_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_idct_ifast_cols_dspr2)
+/*
+ * a0 = inptr
+ * a1 = quantptr
+ * a2 = wsptr
+ * a3 = mips_idct_ifast_coefs
+ */
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    addiu         t9, a0, 16      /* end address */
+    or            AT, a3, zero
+
+0:
+    lw            s0, 0(a1)       /* quantptrDCTSIZE*0 */
+    lw            t0, 0(a0)       /* inptrDCTSIZE*0 */
+    lw            t1, 16(a0)      /* inptrDCTSIZE*1 */
+    muleq_s.w.phl v0, t0, s0      /* tmp0 ... */
+    lw            t2, 32(a0)      /* inptrDCTSIZE*2 */
+    lw            t3, 48(a0)      /* inptrDCTSIZE*3 */
+    lw            t4, 64(a0)      /* inptrDCTSIZE*4 */
+    lw            t5, 80(a0)      /* inptrDCTSIZE*5 */
+    muleq_s.w.phr t0, t0, s0      /* ... tmp0 ... */
+    lw            t6, 96(a0)      /* inptrDCTSIZE*6 */
+    lw            t7, 112(a0)     /* inptrDCTSIZE*7 */
+    or            s4, t1, t2
+    or            s5, t3, t4
+    bnez          s4, 1f
+     ins          t0, v0, 16, 16  /* ... tmp0 */
+    bnez          s5, 1f
+     or           s6, t5, t6
+    or            s6, s6, t7
+    bnez          s6, 1f
+     sw           t0, 0(a2)       /* wsptrDCTSIZE*0 */
+    sw            t0, 16(a2)      /* wsptrDCTSIZE*1 */
+    sw            t0, 32(a2)      /* wsptrDCTSIZE*2 */
+    sw            t0, 48(a2)      /* wsptrDCTSIZE*3 */
+    sw            t0, 64(a2)      /* wsptrDCTSIZE*4 */
+    sw            t0, 80(a2)      /* wsptrDCTSIZE*5 */
+    sw            t0, 96(a2)      /* wsptrDCTSIZE*6 */
+    sw            t0, 112(a2)     /* wsptrDCTSIZE*7 */
+    addiu         a0, a0, 4
+    b             2f
+     addiu        a1, a1, 4
+
+1:
+    lw            s1, 32(a1)      /* quantptrDCTSIZE*2 */
+    lw            s2, 64(a1)      /* quantptrDCTSIZE*4 */
+    muleq_s.w.phl v0, t2, s1      /* tmp1 ... */
+    muleq_s.w.phr t2, t2, s1      /* ... tmp1 ... */
+    lw            s0, 16(a1)      /* quantptrDCTSIZE*1 */
+    lw            s1, 48(a1)      /* quantptrDCTSIZE*3 */
+    lw            s3, 96(a1)      /* quantptrDCTSIZE*6 */
+    muleq_s.w.phl v1, t4, s2      /* tmp2 ... */
+    muleq_s.w.phr t4, t4, s2      /* ... tmp2 ... */
+    lw            s2, 80(a1)      /* quantptrDCTSIZE*5 */
+    lw            t8, 4(AT)       /* FIX(1.414213562) */
+    ins           t2, v0, 16, 16  /* ... tmp1 */
+    muleq_s.w.phl v0, t6, s3      /* tmp3 ... */
+    muleq_s.w.phr t6, t6, s3      /* ... tmp3 ... */
+    ins           t4, v1, 16, 16  /* ... tmp2 */
+    addq.ph       s4, t0, t4      /* tmp10 */
+    subq.ph       s5, t0, t4      /* tmp11 */
+    ins           t6, v0, 16, 16  /* ... tmp3 */
+    subq.ph       s6, t2, t6      /* tmp12 ... */
+    addq.ph       s7, t2, t6      /* tmp13 */
+    mulq_s.ph     s6, s6, t8      /* ... tmp12 ... */
+    addq.ph       t0, s4, s7      /* tmp0 */
+    subq.ph       t6, s4, s7      /* tmp3 */
+    muleq_s.w.phl v0, t1, s0      /* tmp4 ... */
+    muleq_s.w.phr t1, t1, s0      /* ... tmp4 ... */
+    shll_s.ph     s6, s6, 1       /* x2 */
+    lw            s3, 112(a1)     /* quantptrDCTSIZE*7 */
+    subq.ph       s6, s6, s7      /* ... tmp12 */
+    muleq_s.w.phl v1, t7, s3      /* tmp7 ... */
+    muleq_s.w.phr t7, t7, s3      /* ... tmp7 ... */
+    ins           t1, v0, 16, 16  /* ... tmp4 */
+    addq.ph       t2, s5, s6      /* tmp1 */
+    subq.ph       t4, s5, s6      /* tmp2 */
+    muleq_s.w.phl v0, t5, s2      /* tmp6 ... */
+    muleq_s.w.phr t5, t5, s2      /* ... tmp6 ... */
+    ins           t7, v1, 16, 16  /* ... tmp7 */
+    addq.ph       s5, t1, t7      /* z11 */
+    subq.ph       s6, t1, t7      /* z12 */
+    muleq_s.w.phl v1, t3, s1      /* tmp5 ... */
+    muleq_s.w.phr t3, t3, s1      /* ... tmp5 ... */
+    ins           t5, v0, 16, 16  /* ... tmp6 */
+    ins           t3, v1, 16, 16  /* ... tmp5 */
+    addq.ph       s7, t5, t3      /* z13 */
+    subq.ph       v0, t5, t3      /* z10 */
+    addq.ph       t7, s5, s7      /* tmp7 */
+    subq.ph       s5, s5, s7      /* tmp11 ... */
+    addq.ph       v1, v0, s6      /* z5 ... */
+    mulq_s.ph     s5, s5, t8      /* ... tmp11 */
+    lw            t8, 8(AT)       /* FIX(1.847759065) */
+    lw            s4, 0(AT)       /* FIX(1.082392200) */
+    addq.ph       s0, t0, t7
+    subq.ph       s1, t0, t7
+    mulq_s.ph     v1, v1, t8      /* ... z5 */
+    shll_s.ph     s5, s5, 1       /* x2 */
+    lw            t8, 12(AT)      /* FIX(-2.613125930) */
+    sw            s0, 0(a2)       /* wsptrDCTSIZE*0 */
+    shll_s.ph     v0, v0, 1       /* x4 */
+    mulq_s.ph     v0, v0, t8      /* tmp12 ... */
+    mulq_s.ph     s4, s6, s4      /* tmp10 ... */
+    shll_s.ph     v1, v1, 1       /* x2 */
+    addiu         a0, a0, 4
+    addiu         a1, a1, 4
+    sw            s1, 112(a2)     /* wsptrDCTSIZE*7 */
+    shll_s.ph     s6, v0, 1       /* x4 */
+    shll_s.ph     s4, s4, 1       /* x2 */
+    addq.ph       s6, s6, v1      /* ... tmp12 */
+    subq.ph       t5, s6, t7      /* tmp6 */
+    subq.ph       s4, s4, v1      /* ... tmp10 */
+    subq.ph       t3, s5, t5      /* tmp5 */
+    addq.ph       s2, t2, t5
+    addq.ph       t1, s4, t3      /* tmp4 */
+    subq.ph       s3, t2, t5
+    sw            s2, 16(a2)      /* wsptrDCTSIZE*1 */
+    sw            s3, 96(a2)      /* wsptrDCTSIZE*6 */
+    addq.ph       v0, t4, t3
+    subq.ph       v1, t4, t3
+    sw            v0, 32(a2)      /* wsptrDCTSIZE*2 */
+    sw            v1, 80(a2)      /* wsptrDCTSIZE*5 */
+    addq.ph       v0, t6, t1
+    subq.ph       v1, t6, t1
+    sw            v0, 64(a2)      /* wsptrDCTSIZE*4 */
+    sw            v1, 48(a2)      /* wsptrDCTSIZE*3 */
+
+2:
+    bne           a0, t9, 0b
+     addiu        a2, a2, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j             ra
+     nop
+
+END(jsimd_idct_ifast_cols_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_idct_ifast_rows_dspr2)
+/*
+ * a0 = wsptr
+ * a1 = output_buf
+ * a2 = output_col
+ * a3 = mips_idct_ifast_coefs
+ */
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8, a3
+
+    addiu         t9, a0, 128     /* end address */
+    lui           s8, 0x8080
+    ori           s8, s8, 0x8080
+
+0:
+    lw            AT, 36(sp)      /* restore $a3 (mips_idct_ifast_coefs) */
+    lw            t0, 0(a0)       /* wsptrDCTSIZE*0+0/1  b a */
+    lw            s0, 16(a0)      /* wsptrDCTSIZE*1+0/1  B A */
+    lw            t2, 4(a0)       /* wsptrDCTSIZE*0+2/3  d c */
+    lw            s2, 20(a0)      /* wsptrDCTSIZE*1+2/3  D C */
+    lw            t4, 8(a0)       /* wsptrDCTSIZE*0+4/5  f e */
+    lw            s4, 24(a0)      /* wsptrDCTSIZE*1+4/5  F E */
+    lw            t6, 12(a0)      /* wsptrDCTSIZE*0+6/7  h g */
+    lw            s6, 28(a0)      /* wsptrDCTSIZE*1+6/7  H G */
+    precrq.ph.w   t1, s0, t0      /* B b */
+    ins           t0, s0, 16, 16  /* A a */
+    bnez          t1, 1f
+     or           s0, t2, s2
+    bnez          s0, 1f
+     or           s0, t4, s4
+    bnez          s0, 1f
+     or           s0, t6, s6
+    bnez          s0, 1f
+     shll_s.ph    s0, t0, 2       /* A a */
+    lw            a3, 0(a1)
+    lw            AT, 4(a1)
+    precrq.ph.w   t0, s0, s0      /* A A */
+    ins           s0, s0, 16, 16  /* a a */
+    addu          a3, a3, a2
+    addu          AT, AT, a2
+    precrq.qb.ph  t0, t0, t0      /* A A A A */
+    precrq.qb.ph  s0, s0, s0      /* a a a a */
+    addu.qb       s0, s0, s8
+    addu.qb       t0, t0, s8
+    sw            s0, 0(a3)
+    sw            s0, 4(a3)
+    sw            t0, 0(AT)
+    sw            t0, 4(AT)
+    addiu         a0, a0, 32
+    bne           a0, t9, 0b
+     addiu        a1, a1, 8
+    b             2f
+     nop
+
+1:
+    precrq.ph.w   t3, s2, t2
+    ins           t2, s2, 16, 16
+    precrq.ph.w   t5, s4, t4
+    ins           t4, s4, 16, 16
+    precrq.ph.w   t7, s6, t6
+    ins           t6, s6, 16, 16
+    lw            t8, 4(AT)       /* FIX(1.414213562) */
+    addq.ph       s4, t0, t4      /* tmp10 */
+    subq.ph       s5, t0, t4      /* tmp11 */
+    subq.ph       s6, t2, t6      /* tmp12 ... */
+    addq.ph       s7, t2, t6      /* tmp13 */
+    mulq_s.ph     s6, s6, t8      /* ... tmp12 ... */
+    addq.ph       t0, s4, s7      /* tmp0 */
+    subq.ph       t6, s4, s7      /* tmp3 */
+    shll_s.ph     s6, s6, 1       /* x2 */
+    subq.ph       s6, s6, s7      /* ... tmp12 */
+    addq.ph       t2, s5, s6      /* tmp1 */
+    subq.ph       t4, s5, s6      /* tmp2 */
+    addq.ph       s5, t1, t7      /* z11 */
+    subq.ph       s6, t1, t7      /* z12 */
+    addq.ph       s7, t5, t3      /* z13 */
+    subq.ph       v0, t5, t3      /* z10 */
+    addq.ph       t7, s5, s7      /* tmp7 */
+    subq.ph       s5, s5, s7      /* tmp11 ... */
+    addq.ph       v1, v0, s6      /* z5 ... */
+    mulq_s.ph     s5, s5, t8      /* ... tmp11 */
+    lw            t8, 8(AT)       /* FIX(1.847759065) */
+    lw            s4, 0(AT)       /* FIX(1.082392200) */
+    addq.ph       s0, t0, t7      /* tmp0 + tmp7 */
+    subq.ph       s7, t0, t7      /* tmp0 - tmp7 */
+    mulq_s.ph     v1, v1, t8      /* ... z5 */
+    lw            a3, 0(a1)
+    lw            t8, 12(AT)      /* FIX(-2.613125930) */
+    shll_s.ph     s5, s5, 1       /* x2 */
+    addu          a3, a3, a2
+    shll_s.ph     v0, v0, 1       /* x4 */
+    mulq_s.ph     v0, v0, t8      /* tmp12 ... */
+    mulq_s.ph     s4, s6, s4      /* tmp10 ... */
+    shll_s.ph     v1, v1, 1       /* x2 */
+    addiu         a0, a0, 32
+    addiu         a1, a1, 8
+    shll_s.ph     s6, v0, 1       /* x4 */
+    shll_s.ph     s4, s4, 1       /* x2 */
+    addq.ph       s6, s6, v1      /* ... tmp12 */
+    shll_s.ph     s0, s0, 2
+    subq.ph       t5, s6, t7      /* tmp6 */
+    subq.ph       s4, s4, v1      /* ... tmp10 */
+    subq.ph       t3, s5, t5      /* tmp5 */
+    shll_s.ph     s7, s7, 2
+    addq.ph       t1, s4, t3      /* tmp4 */
+    addq.ph       s1, t2, t5      /* tmp1 + tmp6 */
+    subq.ph       s6, t2, t5      /* tmp1 - tmp6 */
+    addq.ph       s2, t4, t3      /* tmp2 + tmp5 */
+    subq.ph       s5, t4, t3      /* tmp2 - tmp5 */
+    addq.ph       s4, t6, t1      /* tmp3 + tmp4 */
+    subq.ph       s3, t6, t1      /* tmp3 - tmp4 */
+    shll_s.ph     s1, s1, 2
+    shll_s.ph     s2, s2, 2
+    shll_s.ph     s3, s3, 2
+    shll_s.ph     s4, s4, 2
+    shll_s.ph     s5, s5, 2
+    shll_s.ph     s6, s6, 2
+    precrq.ph.w   t0, s1, s0      /* B A */
+    ins           s0, s1, 16, 16  /* b a */
+    precrq.ph.w   t2, s3, s2      /* D C */
+    ins           s2, s3, 16, 16  /* d c */
+    precrq.ph.w   t4, s5, s4      /* F E */
+    ins           s4, s5, 16, 16  /* f e */
+    precrq.ph.w   t6, s7, s6      /* H G */
+    ins           s6, s7, 16, 16  /* h g */
+    precrq.qb.ph  t0, t2, t0      /* D C B A */
+    precrq.qb.ph  s0, s2, s0      /* d c b a */
+    precrq.qb.ph  t4, t6, t4      /* H G F E */
+    precrq.qb.ph  s4, s6, s4      /* h g f e */
+    addu.qb       s0, s0, s8
+    addu.qb       s4, s4, s8
+    sw            s0, 0(a3)       /* outptr0/1/2/3       d c b a */
+    sw            s4, 4(a3)       /* outptr4/5/6/7       h g f e */
+    lw            a3, -4(a1)
+    addu.qb       t0, t0, s8
+    addu          a3, a3, a2
+    addu.qb       t4, t4, s8
+    sw            t0, 0(a3)       /* outptr0/1/2/3       D C B A */
+    bne           a0, t9, 0b
+     sw           t4, 4(a3)       /* outptr4/5/6/7       H G F E */
+
+2:
+
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8, a3
+
+    j             ra
+     nop
+
+END(jsimd_idct_ifast_rows_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_fdct_islow_dspr2)
+/*
+ * a0 = data
+ */
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8
+
+    lui         t0, 6437
+    ori         t0, 2260
+    lui         t1, 9633
+    ori         t1, 11363
+    lui         t2, 0xd39e
+    ori         t2, 0xe6dc
+    lui         t3, 0xf72d
+    ori         t3, 9633
+    lui         t4, 2261
+    ori         t4, 9633
+    lui         t5, 0xd39e
+    ori         t5, 6437
+    lui         t6, 9633
+    ori         t6, 0xd39d
+    lui         t7, 0xe6dc
+    ori         t7, 2260
+    lui         t8, 4433
+    ori         t8, 10703
+    lui         t9, 0xd630
+    ori         t9, 4433
+    li          s8, 8
+    move        a1, a0
+1:
+    lw          s0, 0(a1)       /* tmp0 = 1|0 */
+    lw          s1, 4(a1)       /* tmp1 = 3|2 */
+    lw          s2, 8(a1)       /* tmp2 = 5|4 */
+    lw          s3, 12(a1)      /* tmp3 = 7|6 */
+    packrl.ph   s1, s1, s1      /* tmp1 = 2|3 */
+    packrl.ph   s3, s3, s3      /* tmp3 = 6|7 */
+    subq.ph     s7, s1, s2      /* tmp7 = 2-5|3-4 = t5|t4 */
+    subq.ph     s5, s0, s3      /* tmp5 = 1-6|0-7 = t6|t7 */
+    mult        $0, $0          /* ac0  = 0 */
+    dpa.w.ph    $ac0, s7, t0    /* ac0 += t5*  6437 + t4*  2260 */
+    dpa.w.ph    $ac0, s5, t1    /* ac0 += t6*  9633 + t7* 11363 */
+    mult        $ac1, $0, $0    /* ac1  = 0 */
+    dpa.w.ph    $ac1, s7, t2    /* ac1 += t5*-11362 + t4* -6436 */
+    dpa.w.ph    $ac1, s5, t3    /* ac1 += t6* -2259 + t7*  9633 */
+    mult        $ac2, $0, $0    /* ac2  = 0 */
+    dpa.w.ph    $ac2, s7, t4    /* ac2 += t5*  2261 + t4*  9633 */
+    dpa.w.ph    $ac2, s5, t5    /* ac2 += t6*-11362 + t7*  6437 */
+    mult        $ac3, $0, $0    /* ac3  = 0 */
+    dpa.w.ph    $ac3, s7, t6    /* ac3 += t5*  9633 + t4*-11363 */
+    dpa.w.ph    $ac3, s5, t7    /* ac3 += t6* -6436 + t7*  2260 */
+    addq.ph     s6, s1, s2      /* tmp6 = 2+5|3+4 = t2|t3 */
+    addq.ph     s4, s0, s3      /* tmp4 = 1+6|0+7 = t1|t0 */
+    extr_r.w    s0, $ac0, 11    /* tmp0 = (ac0 + 1024) >> 11 */
+    extr_r.w    s1, $ac1, 11    /* tmp1 = (ac1 + 1024) >> 11 */
+    extr_r.w    s2, $ac2, 11    /* tmp2 = (ac2 + 1024) >> 11 */
+    extr_r.w    s3, $ac3, 11    /* tmp3 = (ac3 + 1024) >> 11 */
+    addq.ph     s5, s4, s6      /* tmp5 = t1+t2|t0+t3 = t11|t10 */
+    subq.ph     s7, s4, s6      /* tmp7 = t1-t2|t0-t3 = t12|t13 */
+    sh          s0, 2(a1)
+    sh          s1, 6(a1)
+    sh          s2, 10(a1)
+    sh          s3, 14(a1)
+    mult        $0, $0          /* ac0  = 0 */
+    dpa.w.ph    $ac0, s7, t8    /* ac0 += t12*  4433 + t13* 10703 */
+    mult        $ac1, $0, $0    /* ac1  = 0 */
+    dpa.w.ph    $ac1, s7, t9    /* ac1 += t12*-10704 + t13*  4433 */
+    sra         s4, s5, 16      /* tmp4 = t11 */
+    addiu       a1, a1, 16
+    addiu       s8, s8, -1
+    extr_r.w    s0, $ac0, 11    /* tmp0 = (ac0 + 1024) >> 11 */
+    extr_r.w    s1, $ac1, 11    /* tmp1 = (ac1 + 1024) >> 11 */
+    addu        s2, s5, s4      /* tmp2 = t10 + t11 */
+    subu        s3, s5, s4      /* tmp3 = t10 - t11 */
+    sll         s2, s2, 2       /* tmp2 = (t10 + t11) << 2 */
+    sll         s3, s3, 2       /* tmp3 = (t10 - t11) << 2 */
+    sh          s2, -16(a1)
+    sh          s3, -8(a1)
+    sh          s0, -12(a1)
+    bgtz        s8, 1b
+     sh         s1, -4(a1)
+    li          t0, 2260
+    li          t1, 11363
+    li          t2, 9633
+    li          t3, 6436
+    li          t4, 6437
+    li          t5, 2261
+    li          t6, 11362
+    li          t7, 2259
+    li          t8, 4433
+    li          t9, 10703
+    li          a1, 10704
+    li          s8, 8
+
+2:
+    lh          a2, 0(a0)       /* 0 */
+    lh          a3, 16(a0)      /* 8 */
+    lh          v0, 32(a0)      /* 16 */
+    lh          v1, 48(a0)      /* 24 */
+    lh          s4, 64(a0)      /* 32 */
+    lh          s5, 80(a0)      /* 40 */
+    lh          s6, 96(a0)      /* 48 */
+    lh          s7, 112(a0)     /* 56 */
+    addu        s2, v0, s5      /* tmp2 = 16 + 40 */
+    subu        s5, v0, s5      /* tmp5 = 16 - 40 */
+    addu        s3, v1, s4      /* tmp3 = 24 + 32 */
+    subu        s4, v1, s4      /* tmp4 = 24 - 32 */
+    addu        s0, a2, s7      /* tmp0 =  0 + 56 */
+    subu        s7, a2, s7      /* tmp7 =  0 - 56 */
+    addu        s1, a3, s6      /* tmp1 =  8 + 48 */
+    subu        s6, a3, s6      /* tmp6 =  8 - 48 */
+    addu        a2, s0, s3      /* tmp10 = tmp0 + tmp3 */
+    subu        v1, s0, s3      /* tmp13 = tmp0 - tmp3 */
+    addu        a3, s1, s2      /* tmp11 = tmp1 + tmp2 */
+    subu        v0, s1, s2      /* tmp12 = tmp1 - tmp2 */
+    mult        s7, t1          /* ac0  = tmp7 * c1 */
+    madd        s4, t0          /* ac0 += tmp4 * c0 */
+    madd        s5, t4          /* ac0 += tmp5 * c4 */
+    madd        s6, t2          /* ac0 += tmp6 * c2 */
+    mult        $ac1, s7, t2    /* ac1  = tmp7 * c2 */
+    msub        $ac1, s4, t3    /* ac1 -= tmp4 * c3 */
+    msub        $ac1, s5, t6    /* ac1 -= tmp5 * c6 */
+    msub        $ac1, s6, t7    /* ac1 -= tmp6 * c7 */
+    mult        $ac2, s7, t4    /* ac2  = tmp7 * c4 */
+    madd        $ac2, s4, t2    /* ac2 += tmp4 * c2 */
+    madd        $ac2, s5, t5    /* ac2 += tmp5 * c5 */
+    msub        $ac2, s6, t6    /* ac2 -= tmp6 * c6 */
+    mult        $ac3, s7, t0    /* ac3  = tmp7 * c0 */
+    msub        $ac3, s4, t1    /* ac3 -= tmp4 * c1 */
+    madd        $ac3, s5, t2    /* ac3 += tmp5 * c2 */
+    msub        $ac3, s6, t3    /* ac3 -= tmp6 * c3 */
+    extr_r.w    s0, $ac0, 15    /* tmp0 = (ac0 + 16384) >> 15 */
+    extr_r.w    s1, $ac1, 15    /* tmp1 = (ac1 + 16384) >> 15 */
+    extr_r.w    s2, $ac2, 15    /* tmp2 = (ac2 + 16384) >> 15 */
+    extr_r.w    s3, $ac3, 15    /* tmp3 = (ac3 + 16384) >> 15 */
+    addiu       s8, s8, -1
+    addu        s4, a2, a3      /* tmp4 = tmp10 + tmp11 */
+    subu        s5, a2, a3      /* tmp5 = tmp10 - tmp11 */
+    sh          s0, 16(a0)
+    sh          s1, 48(a0)
+    sh          s2, 80(a0)
+    sh          s3, 112(a0)
+    mult        v0, t8          /* ac0  = tmp12 * c8 */
+    madd        v1, t9          /* ac0 += tmp13 * c9 */
+    mult        $ac1, v1, t8    /* ac1  = tmp13 * c8 */
+    msub        $ac1, v0, a1    /* ac1 -= tmp12 * c10 */
+    addiu       a0, a0, 2
+    extr_r.w    s6, $ac0, 15    /* tmp6 = (ac0 + 16384) >> 15 */
+    extr_r.w    s7, $ac1, 15    /* tmp7 = (ac1 + 16384) >> 15 */
+    shra_r.w    s4, s4, 2       /* tmp4 = (tmp4 + 2) >> 2 */
+    shra_r.w    s5, s5, 2       /* tmp5 = (tmp5 + 2) >> 2 */
+    sh          s4, -2(a0)
+    sh          s5, 62(a0)
+    sh          s6, 30(a0)
+    bgtz        s8, 2b
+     sh         s7, 94(a0)
+
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8
+
+    jr          ra
+     nop
+
+END(jsimd_fdct_islow_dspr2)
+
+
+/**************************************************************************/
+LEAF_DSPR2(jsimd_fdct_ifast_dspr2)
+/*
+ * a0 = data
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 8, s0, s1
+
+    li          a1, 0x014e014e  /* FIX_1_306562965 (334 << 16) |
+                                                   (334 & 0xffff) */
+    li          a2, 0x008b008b  /* FIX_0_541196100 (139 << 16) |
+                                                   (139 & 0xffff) */
+    li          a3, 0x00620062  /* FIX_0_382683433 (98 << 16) |
+                                                   (98 & 0xffff) */
+    li          s1, 0x00b500b5  /* FIX_0_707106781 (181 << 16) |
+                                                   (181 & 0xffff) */
+
+    move        v0, a0
+    addiu       v1, v0, 128     /* end address */
+
+0:
+    lw          t0, 0(v0)       /* tmp0 = 1|0 */
+    lw          t1, 4(v0)       /* tmp1 = 3|2 */
+    lw          t2, 8(v0)       /* tmp2 = 5|4 */
+    lw          t3, 12(v0)      /* tmp3 = 7|6 */
+    packrl.ph   t1, t1, t1      /* tmp1 = 2|3 */
+    packrl.ph   t3, t3, t3      /* tmp3 = 6|7 */
+    subq.ph     t7, t1, t2      /* tmp7 = 2-5|3-4 = t5|t4 */
+    subq.ph     t5, t0, t3      /* tmp5 = 1-6|0-7 = t6|t7 */
+    addq.ph     t6, t1, t2      /* tmp6 = 2+5|3+4 = t2|t3 */
+    addq.ph     t4, t0, t3      /* tmp4 = 1+6|0+7 = t1|t0 */
+    addq.ph     t8, t4, t6      /* tmp5 = t1+t2|t0+t3 = t11|t10 */
+    subq.ph     t9, t4, t6      /* tmp7 = t1-t2|t0-t3 = t12|t13 */
+    sra         t4, t8, 16      /* tmp4 = t11 */
+    mult        $0, $0          /* ac0  = 0 */
+    dpa.w.ph    $ac0, t9, s1
+    mult        $ac1, $0, $0    /* ac1  = 0 */
+    dpa.w.ph    $ac1, t7, a3    /* ac1 += t4*98 + t5*98 */
+    dpsx.w.ph   $ac1, t5, a3    /* ac1 += t6*98 + t7*98 */
+    mult        $ac2, $0, $0    /* ac2  = 0 */
+    dpa.w.ph    $ac2, t7, a2    /* ac2 += t4*139 + t5*139 */
+    mult        $ac3, $0, $0    /* ac3  = 0 */
+    dpa.w.ph    $ac3, t5, a1    /* ac3 += t6*334 + t7*334 */
+    precrq.ph.w t0, t5, t7      /* t0 = t5|t6 */
+    addq.ph     t2, t8, t4      /* tmp2 = t10 + t11 */
+    subq.ph     t3, t8, t4      /* tmp3 = t10 - t11 */
+    extr.w      t4, $ac0, 8
+    mult        $0, $0          /* ac0  = 0 */
+    dpa.w.ph    $ac0, t0, s1    /* ac0 += t5*181 + t6*181 */
+    extr.w      t0, $ac1, 8     /* t0 = z5 */
+    extr.w      t1, $ac2, 8     /* t1 = MULTIPLY(tmp10, 139) */
+    extr.w      t7, $ac3, 8     /* t2 = MULTIPLY(tmp12, 334) */
+    extr.w      t8, $ac0, 8     /* t8 = z3 = MULTIPLY(tmp11, 181) */
+    add         t6, t1, t0      /* t6 = z2 */
+    add         t7, t7, t0      /* t7 = z4 */
+    subq.ph     t0, t5, t8      /* t0 = z13 = tmp7 - z3 */
+    addq.ph     t8, t5, t8      /* t9 = z11 = tmp7 + z3 */
+    addq.ph     t1, t0, t6      /* t1 = z13 + z2 */
+    subq.ph     t6, t0, t6      /* t6 = z13 - z2 */
+    addq.ph     t0, t8, t7      /* t0 = z11 + z4 */
+    subq.ph     t7, t8, t7      /* t7 = z11 - z4 */
+    addq.ph     t5, t4, t9
+    subq.ph     t4, t9, t4
+    sh          t2, 0(v0)
+    sh          t5, 4(v0)
+    sh          t3, 8(v0)
+    sh          t4, 12(v0)
+    sh          t1, 10(v0)
+    sh          t6, 6(v0)
+    sh          t0, 2(v0)
+    sh          t7, 14(v0)
+    addiu       v0, 16
+    bne         v1, v0, 0b
+     nop
+    move        v0, a0
+    addiu       v1, v0, 16
+
+1:
+    lh          t0, 0(v0)       /* 0 */
+    lh          t1, 16(v0)      /* 8 */
+    lh          t2, 32(v0)      /* 16 */
+    lh          t3, 48(v0)      /* 24 */
+    lh          t4, 64(v0)      /* 32 */
+    lh          t5, 80(v0)      /* 40 */
+    lh          t6, 96(v0)      /* 48 */
+    lh          t7, 112(v0)     /* 56 */
+    add         t8, t0, t7      /* t8 = tmp0 */
+    sub         t7, t0, t7      /* t7 = tmp7 */
+    add         t0, t1, t6      /* t0 = tmp1 */
+    sub         t1, t1, t6      /* t1 = tmp6 */
+    add         t6, t2, t5      /* t6 = tmp2 */
+    sub         t5, t2, t5      /* t5 = tmp5 */
+    add         t2, t3, t4      /* t2 = tmp3 */
+    sub         t3, t3, t4      /* t3 = tmp4 */
+    add         t4, t8, t2      /* t4 = tmp10 = tmp0 + tmp3 */
+    sub         t8, t8, t2      /* t8 = tmp13 = tmp0 - tmp3 */
+    sub         s0, t0, t6      /* s0 = tmp12 = tmp1 - tmp2 */
+    ins         t8, s0, 16, 16  /* t8 = tmp12|tmp13 */
+    add         t2, t0, t6      /* t2 = tmp11 = tmp1 + tmp2 */
+    mult        $0, $0          /* ac0  = 0 */
+    dpa.w.ph    $ac0, t8, s1    /* ac0 += t12*181 + t13*181 */
+    add         s0, t4, t2      /* t8 = tmp10+tmp11 */
+    sub         t4, t4, t2      /* t4 = tmp10-tmp11 */
+    sh          s0, 0(v0)
+    sh          t4, 64(v0)
+    extr.w      t2, $ac0, 8     /* z1 = MULTIPLY(tmp12+tmp13,
+                                                 FIX_0_707106781) */
+    addq.ph     t4, t8, t2      /* t9 = tmp13 + z1 */
+    subq.ph     t8, t8, t2      /* t2 = tmp13 - z1 */
+    sh          t4, 32(v0)
+    sh          t8, 96(v0)
+    add         t3, t3, t5      /* t3 = tmp10 = tmp4 + tmp5 */
+    add         t0, t5, t1      /* t0 = tmp11 = tmp5 + tmp6 */
+    add         t1, t1, t7      /* t1 = tmp12 = tmp6 + tmp7 */
+    andi        t4, a1, 0xffff
+    mul         s0, t1, t4
+    sra         s0, s0, 8       /* s0 = z4 =
+                                     MULTIPLY(tmp12, FIX_1_306562965) */
+    ins         t1, t3, 16, 16  /* t1 = tmp10|tmp12 */
+    mult        $0, $0          /* ac0  = 0 */
+    mulsa.w.ph  $ac0, t1, a3    /* ac0 += t10*98 - t12*98 */
+    extr.w      t8, $ac0, 8     /* z5 = MULTIPLY(tmp10-tmp12,
+                                                 FIX_0_382683433) */
+    add         t2, t7, t8      /* t2 = tmp7 + z5 */
+    sub         t7, t7, t8      /* t7 = tmp7 - z5 */
+    andi        t4, a2, 0xffff
+    mul         t8, t3, t4
+    sra         t8, t8, 8       /* t8 = z2 =
+                                     MULTIPLY(tmp10, FIX_0_541196100) */
+    andi        t4, s1, 0xffff
+    mul         t6, t0, t4
+    sra         t6, t6, 8       /* t6 = z3 =
+                                     MULTIPLY(tmp11, FIX_0_707106781) */
+    add         t0, t6, t8      /* t0 = z3 + z2 */
+    sub         t1, t6, t8      /* t1 = z3 - z2 */
+    add         t3, t6, s0      /* t3 = z3 + z4 */
+    sub         t4, t6, s0      /* t4 = z3 - z4 */
+    sub         t5, t2, t1      /* t5 = dataptr5 */
+    sub         t6, t7, t0      /* t6 = dataptr3 */
+    add         t3, t2, t3      /* t3 = dataptr1 */
+    add         t4, t7, t4      /* t4 = dataptr7 */
+    sh          t5, 80(v0)
+    sh          t6, 48(v0)
+    sh          t3, 16(v0)
+    sh          t4, 112(v0)
+    addiu       v0, 2
+    bne         v0, v1, 1b
+     nop
+
+    RESTORE_REGS_FROM_STACK 8, s0, s1
+
+    j           ra
+     nop
+END(jsimd_fdct_ifast_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_quantize_dspr2)
+/*
+ * a0 = coef_block
+ * a1 = divisors
+ * a2 = workspace
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2
+
+    addiu       v0, a2, 124     /* v0 = workspace_end */
+    lh          t0, 0(a2)
+    lh          t1, 0(a1)
+    lh          t2, 128(a1)
+    sra         t3, t0, 15
+    sll         t3, t3, 1
+    addiu       t3, t3, 1
+    mul         t0, t0, t3
+    lh          t4, 384(a1)
+    lh          t5, 130(a1)
+    lh          t6, 2(a2)
+    lh          t7, 2(a1)
+    lh          t8, 386(a1)
+
+1:
+    andi        t1, 0xffff
+    add         t9, t0, t2
+    andi        t9, 0xffff
+    mul         v1, t9, t1
+    sra         s0, t6, 15
+    sll         s0, s0, 1
+    addiu       s0, s0, 1
+    addiu       t9, t4, 16
+    srav        v1, v1, t9
+    mul         v1, v1, t3
+    mul         t6, t6, s0
+    andi        t7, 0xffff
+    addiu       a2, a2, 4
+    addiu       a1, a1, 4
+    add         s1, t6, t5
+    andi        s1, 0xffff
+    sh          v1, 0(a0)
+
+    mul         s2, s1, t7
+    addiu       s1, t8, 16
+    srav        s2, s2, s1
+    mul         s2, s2, s0
+    lh          t0, 0(a2)
+    lh          t1, 0(a1)
+    sra         t3, t0, 15
+    sll         t3, t3, 1
+    addiu       t3, t3, 1
+    mul         t0, t0, t3
+    lh          t2, 128(a1)
+    lh          t4, 384(a1)
+    lh          t5, 130(a1)
+    lh          t8, 386(a1)
+    lh          t6, 2(a2)
+    lh          t7, 2(a1)
+    sh          s2, 2(a0)
+    lh          t0, 0(a2)
+    sra         t3, t0, 15
+    sll         t3, t3, 1
+    addiu       t3, t3, 1
+    mul         t0, t0, t3
+    bne         a2, v0, 1b
+     addiu      a0, a0, 4
+
+    andi        t1, 0xffff
+    add         t9, t0, t2
+    andi        t9, 0xffff
+    mul         v1, t9, t1
+    sra         s0, t6, 15
+    sll         s0, s0, 1
+    addiu       s0, s0, 1
+    addiu       t9, t4, 16
+    srav        v1, v1, t9
+    mul         v1, v1, t3
+    mul         t6, t6, s0
+    andi        t7, 0xffff
+    sh          v1, 0(a0)
+    add         s1, t6, t5
+    andi        s1, 0xffff
+    mul         s2, s1, t7
+    addiu       s1, t8, 16
+    addiu       a2, a2, 4
+    addiu       a1, a1, 4
+    srav        s2, s2, s1
+    mul         s2, s2, s0
+    sh          s2, 2(a0)
+
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2
+
+    j           ra
+     nop
+
+END(jsimd_quantize_dspr2)
+
+
+#ifndef __mips_soft_float
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_quantize_float_dspr2)
+/*
+ * a0 = coef_block
+ * a1 = divisors
+ * a2 = workspace
+ */
+    .set at
+
+    li          t1, 0x46800100  /* integer representation 16384.5 */
+    mtc1        t1, f0
+    li          t0, 63
+0:
+    lwc1        f2, 0(a2)
+    lwc1        f10, 0(a1)
+    lwc1        f4, 4(a2)
+    lwc1        f12, 4(a1)
+    lwc1        f6, 8(a2)
+    lwc1        f14, 8(a1)
+    lwc1        f8, 12(a2)
+    lwc1        f16, 12(a1)
+    madd.s      f2, f0, f2, f10
+    madd.s      f4, f0, f4, f12
+    madd.s      f6, f0, f6, f14
+    madd.s      f8, f0, f8, f16
+    lwc1        f10, 16(a1)
+    lwc1        f12, 20(a1)
+    trunc.w.s   f2, f2
+    trunc.w.s   f4, f4
+    trunc.w.s   f6, f6
+    trunc.w.s   f8, f8
+    lwc1        f14, 24(a1)
+    lwc1        f16, 28(a1)
+    mfc1        t1, f2
+    mfc1        t2, f4
+    mfc1        t3, f6
+    mfc1        t4, f8
+    lwc1        f2, 16(a2)
+    lwc1        f4, 20(a2)
+    lwc1        f6, 24(a2)
+    lwc1        f8, 28(a2)
+    madd.s      f2, f0, f2, f10
+    madd.s      f4, f0, f4, f12
+    madd.s      f6, f0, f6, f14
+    madd.s      f8, f0, f8, f16
+    addiu       t1, t1, -16384
+    addiu       t2, t2, -16384
+    addiu       t3, t3, -16384
+    addiu       t4, t4, -16384
+    trunc.w.s   f2, f2
+    trunc.w.s   f4, f4
+    trunc.w.s   f6, f6
+    trunc.w.s   f8, f8
+    sh          t1, 0(a0)
+    sh          t2, 2(a0)
+    sh          t3, 4(a0)
+    sh          t4, 6(a0)
+    mfc1        t1, f2
+    mfc1        t2, f4
+    mfc1        t3, f6
+    mfc1        t4, f8
+    addiu       t0, t0, -8
+    addiu       a2, a2, 32
+    addiu       a1, a1, 32
+    addiu       t1, t1, -16384
+    addiu       t2, t2, -16384
+    addiu       t3, t3, -16384
+    addiu       t4, t4, -16384
+    sh          t1, 8(a0)
+    sh          t2, 10(a0)
+    sh          t3, 12(a0)
+    sh          t4, 14(a0)
+    bgez        t0, 0b
+     addiu      a0, a0, 16
+
+    j           ra
+     nop
+
+END(jsimd_quantize_float_dspr2)
+
+#endif
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_idct_2x2_dspr2)
+/*
+ * a0 = compptr->dct_table
+ * a1 = coef_block
+ * a2 = output_buf
+ * a3 = output_col
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4, s5
+
+    addiu       sp, sp, -40
+    move        v0, sp
+    addiu       s2, zero, 29692
+    addiu       s3, zero, -10426
+    addiu       s4, zero, 6967
+    addiu       s5, zero, -5906
+    lh          t0, 0(a1)       /* t0 = inptrDCTSIZE*0 */
+    lh          t5, 0(a0)       /* t5 = quantptrDCTSIZE*0 */
+    lh          t1, 48(a1)      /* t1 = inptrDCTSIZE*3 */
+    lh          t6, 48(a0)      /* t6 = quantptrDCTSIZE*3 */
+    mul         t4, t5, t0
+    lh          t0, 16(a1)      /* t0 = inptrDCTSIZE*1 */
+    lh          t5, 16(a0)      /* t5 = quantptrDCTSIZE*1 */
+    mul         t6, t6, t1
+    mul         t5, t5, t0
+    lh          t2, 80(a1)      /* t2 = inptrDCTSIZE*5 */
+    lh          t7, 80(a0)      /* t7 = quantptrDCTSIZE*5 */
+    lh          t3, 112(a1)     /* t3 = inptrDCTSIZE*7 */
+    lh          t8, 112(a0)     /* t8 = quantptrDCTSIZE*7 */
+    mul         t7, t7, t2
+    mult        zero, zero
+    mul         t8, t8, t3
+    li          s0, 0x73FCD746  /* s0 = (29692 << 16) | (-10426 & 0xffff) */
+    li          s1, 0x1B37E8EE  /* s1 = (6967 << 16) | (-5906 & 0xffff) */
+    ins         t6, t5, 16, 16  /* t6 = t5|t6 */
+    sll         t4, t4, 15
+    dpa.w.ph    $ac0, t6, s0
+    lh          t1, 2(a1)
+    lh          t6, 2(a0)
+    ins         t8, t7, 16, 16  /* t8 = t7|t8 */
+    dpa.w.ph    $ac0, t8, s1
+    mflo        t0, $ac0
+    mul         t5, t6, t1
+    lh          t1, 18(a1)
+    lh          t6, 18(a0)
+    lh          t2, 50(a1)
+    lh          t7, 50(a0)
+    mul         t6, t6, t1
+    subu        t8, t4, t0
+    mul         t7, t7, t2
+    addu        t0, t4, t0
+    shra_r.w    t0, t0, 13
+    lh          t1, 82(a1)
+    lh          t2, 82(a0)
+    lh          t3, 114(a1)
+    lh          t4, 114(a0)
+    shra_r.w    t8, t8, 13
+    mul         t1, t1, t2
+    mul         t3, t3, t4
+    sw          t0, 0(v0)
+    sw          t8, 20(v0)
+    sll         t4, t5, 15
+    ins         t7, t6, 16, 16
+    mult        zero, zero
+    dpa.w.ph    $ac0, t7, s0
+    ins         t3, t1, 16, 16
+    lh          t1, 6(a1)
+    lh          t6, 6(a0)
+    dpa.w.ph    $ac0, t3, s1
+    mflo        t0, $ac0
+    mul         t5, t6, t1
+    lh          t1, 22(a1)
+    lh          t6, 22(a0)
+    lh          t2, 54(a1)
+    lh          t7, 54(a0)
+    mul         t6, t6, t1
+    subu        t8, t4, t0
+    mul         t7, t7, t2
+    addu        t0, t4, t0
+    shra_r.w    t0, t0, 13
+    lh          t1, 86(a1)
+    lh          t2, 86(a0)
+    lh          t3, 118(a1)
+    lh          t4, 118(a0)
+    shra_r.w    t8, t8, 13
+    mul         t1, t1, t2
+    mul         t3, t3, t4
+    sw          t0, 4(v0)
+    sw          t8, 24(v0)
+    sll         t4, t5, 15
+    ins         t7, t6, 16, 16
+    mult        zero, zero
+    dpa.w.ph    $ac0, t7, s0
+    ins         t3, t1, 16, 16
+    lh          t1, 10(a1)
+    lh          t6, 10(a0)
+    dpa.w.ph    $ac0, t3, s1
+    mflo        t0, $ac0
+    mul         t5, t6, t1
+    lh          t1, 26(a1)
+    lh          t6, 26(a0)
+    lh          t2, 58(a1)
+    lh          t7, 58(a0)
+    mul         t6, t6, t1
+    subu        t8, t4, t0
+    mul         t7, t7, t2
+    addu        t0, t4, t0
+    shra_r.w    t0, t0, 13
+    lh          t1, 90(a1)
+    lh          t2, 90(a0)
+    lh          t3, 122(a1)
+    lh          t4, 122(a0)
+    shra_r.w    t8, t8, 13
+    mul         t1, t1, t2
+    mul         t3, t3, t4
+    sw          t0, 8(v0)
+    sw          t8, 28(v0)
+    sll         t4, t5, 15
+    ins         t7, t6, 16, 16
+    mult        zero, zero
+    dpa.w.ph    $ac0, t7, s0
+    ins         t3, t1, 16, 16
+    lh          t1, 14(a1)
+    lh          t6, 14(a0)
+    dpa.w.ph    $ac0, t3, s1
+    mflo        t0, $ac0
+    mul         t5, t6, t1
+    lh          t1, 30(a1)
+    lh          t6, 30(a0)
+    lh          t2, 62(a1)
+    lh          t7, 62(a0)
+    mul         t6, t6, t1
+    subu        t8, t4, t0
+    mul         t7, t7, t2
+    addu        t0, t4, t0
+    shra_r.w    t0, t0, 13
+    lh          t1, 94(a1)
+    lh          t2, 94(a0)
+    lh          t3, 126(a1)
+    lh          t4, 126(a0)
+    shra_r.w    t8, t8, 13
+    mul         t1, t1, t2
+    mul         t3, t3, t4
+    sw          t0, 12(v0)
+    sw          t8, 32(v0)
+    sll         t4, t5, 15
+    ins         t7, t6, 16, 16
+    mult        zero, zero
+    dpa.w.ph    $ac0, t7, s0
+    ins         t3, t1, 16, 16
+    dpa.w.ph    $ac0, t3, s1
+    mflo        t0, $ac0
+    lw          t9, 0(a2)
+    lw          t3, 0(v0)
+    lw          t7, 4(v0)
+    lw          t1, 8(v0)
+    addu        t9, t9, a3
+    sll         t3, t3, 15
+    subu        t8, t4, t0
+    addu        t0, t4, t0
+    shra_r.w    t0, t0, 13
+    shra_r.w    t8, t8, 13
+    sw          t0, 16(v0)
+    sw          t8, 36(v0)
+    lw          t5, 12(v0)
+    lw          t6, 16(v0)
+    mult        t7, s2
+    madd        t1, s3
+    madd        t5, s4
+    madd        t6, s5
+    lw          t5, 24(v0)
+    lw          t7, 28(v0)
+    mflo        t0, $ac0
+    lw          t8, 32(v0)
+    lw          t2, 36(v0)
+    mult        $ac1, t5, s2
+    madd        $ac1, t7, s3
+    madd        $ac1, t8, s4
+    madd        $ac1, t2, s5
+    addu        t1, t3, t0
+    subu        t6, t3, t0
+    shra_r.w    t1, t1, 20
+    shra_r.w    t6, t6, 20
+    mflo        t4, $ac1
+    shll_s.w    t1, t1, 24
+    shll_s.w    t6, t6, 24
+    sra         t1, t1, 24
+    sra         t6, t6, 24
+    addiu       t1, t1, 128
+    addiu       t6, t6, 128
+    lw          t0, 20(v0)
+    sb          t1, 0(t9)
+    sb          t6, 1(t9)
+    sll         t0, t0, 15
+    lw          t9, 4(a2)
+    addu        t1, t0, t4
+    subu        t6, t0, t4
+    addu        t9, t9, a3
+    shra_r.w    t1, t1, 20
+    shra_r.w    t6, t6, 20
+    shll_s.w    t1, t1, 24
+    shll_s.w    t6, t6, 24
+    sra         t1, t1, 24
+    sra         t6, t6, 24
+    addiu       t1, t1, 128
+    addiu       t6, t6, 128
+    sb          t1, 0(t9)
+    sb          t6, 1(t9)
+    addiu       sp, sp, 40
+
+    RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4, s5
+
+    j           ra
+     nop
+
+END(jsimd_idct_2x2_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_idct_4x4_dspr2)
+/*
+ * a0     = compptr->dct_table
+ * a1     = coef_block
+ * a2     = output_buf
+ * a3     = output_col
+ * 16(sp) = workspaceDCTSIZE*4  (buffers data between passes)
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw          v1, 48(sp)
+    move        t0, a1
+    move        t1, v1
+    li          t9, 4
+    li          s0, 0x2e75f93e
+    li          s1, 0x21f9ba79
+    li          s2, 0xecc2efb0
+    li          s3, 0x52031ccd
+
+0:
+    lh          s6, 32(t0)      /* inptrDCTSIZE*2 */
+    lh          t6, 32(a0)      /* quantptrDCTSIZE*2 */
+    lh          s7, 96(t0)      /* inptrDCTSIZE*6 */
+    lh          t7, 96(a0)      /* quantptrDCTSIZE*6 */
+    mul         t6, s6, t6      /* z2 = (inptrDCTSIZE*2 *
+                                         quantptrDCTSIZE*2) */
+    lh          s4, 0(t0)       /* inptrDCTSIZE*0 */
+    mul         t7, s7, t7      /* z3 = (inptrDCTSIZE*6 *
+                                         quantptrDCTSIZE*6) */
+    lh          s5, 0(a0)       /* quantptr0 */
+    li          s6, 15137
+    li          s7, 6270
+    mul         t2, s4, s5      /* tmp0 = (inptr0 * quantptr0) */
+    mul         t6, s6, t6      /* z2 = (inptrDCTSIZE*2 *
+                                         quantptrDCTSIZE*2) */
+    lh          t5, 112(t0)     /* inptrDCTSIZE*7 */
+    mul         t7, s7, t7      /* z3 = (inptrDCTSIZE*6 *
+                                         quantptrDCTSIZE*6) */
+    lh          s4, 112(a0)     /* quantptrDCTSIZE*7 */
+    lh          v0, 80(t0)      /* inptrDCTSIZE*5 */
+    lh          s5, 80(a0)      /* quantptrDCTSIZE*5 */
+    lh          s6, 48(a0)      /* quantptrDCTSIZE*3 */
+    sll         t2, t2, 14      /* tmp0 <<= (CONST_BITS+1) */
+    lh          s7, 16(a0)      /* quantptrDCTSIZE*1 */
+    lh          t8, 16(t0)      /* inptrDCTSIZE*1 */
+    subu        t6, t6, t7      /* tmp2 =
+                                     MULTIPLY(z2, t5) - MULTIPLY(z3, t6) */
+    lh          t7, 48(t0)      /* inptrDCTSIZE*3 */
+    mul         t5, s4, t5      /* z1 = (inptrDCTSIZE*7 *
+                                         quantptrDCTSIZE*7) */
+    mul         v0, s5, v0      /* z2 = (inptrDCTSIZE*5 *
+                                         quantptrDCTSIZE*5) */
+    mul         t7, s6, t7      /* z3 = (inptrDCTSIZE*3 *
+                                         quantptrDCTSIZE*3) */
+    mul         t8, s7, t8      /* z4 = (inptrDCTSIZE*1 *
+                                         quantptrDCTSIZE*1) */
+    addu        t3, t2, t6      /* tmp10 = tmp0 + z2 */
+    subu        t4, t2, t6      /* tmp10 = tmp0 - z2 */
+    mult        $ac0, zero, zero
+    mult        $ac1, zero, zero
+    ins         t5, v0, 16, 16
+    ins         t7, t8, 16, 16
+    addiu       t9, t9, -1
+    dpa.w.ph    $ac0, t5, s0
+    dpa.w.ph    $ac0, t7, s1
+    dpa.w.ph    $ac1, t5, s2
+    dpa.w.ph    $ac1, t7, s3
+    mflo        s4, $ac0
+    mflo        s5, $ac1
+    addiu       a0, a0, 2
+    addiu       t1, t1, 4
+    addiu       t0, t0, 2
+    addu        t6, t4, s4
+    subu        t5, t4, s4
+    addu        s6, t3, s5
+    subu        s7, t3, s5
+    shra_r.w    t6, t6, 12      /* DESCALE(tmp12 + temp1, 12) */
+    shra_r.w    t5, t5, 12      /* DESCALE(tmp12 - temp1, 12) */
+    shra_r.w    s6, s6, 12      /* DESCALE(tmp10 + temp2, 12) */
+    shra_r.w    s7, s7, 12      /* DESCALE(tmp10 - temp2, 12) */
+    sw          t6, 28(t1)
+    sw          t5, 60(t1)
+    sw          s6, -4(t1)
+    bgtz        t9, 0b
+     sw         s7, 92(t1)
+    /* second loop three pass */
+    li          t9, 3
+1:
+    lh          s6, 34(t0)      /* inptrDCTSIZE*2 */
+    lh          t6, 34(a0)      /* quantptrDCTSIZE*2 */
+    lh          s7, 98(t0)      /* inptrDCTSIZE*6 */
+    lh          t7, 98(a0)      /* quantptrDCTSIZE*6 */
+    mul         t6, s6, t6      /* z2 = (inptrDCTSIZE*2 *
+                                         quantptrDCTSIZE*2) */
+    lh          s4, 2(t0)       /* inptrDCTSIZE*0 */
+    mul         t7, s7, t7      /* z3 = (inptrDCTSIZE*6 *
+                                         quantptrDCTSIZE*6) */
+    lh          s5, 2(a0)       /* quantptrDCTSIZE*0 */
+    li          s6, 15137
+    li          s7, 6270
+    mul         t2, s4, s5      /* tmp0 = (inptr0 * quantptr0) */
+    mul         v0, s6, t6      /* z2 = (inptrDCTSIZE*2 *
+                                         quantptrDCTSIZE*2) */
+    lh          t5, 114(t0)     /* inptrDCTSIZE*7 */
+    mul         t7, s7, t7      /* z3 = (inptrDCTSIZE*6 *
+                                         quantptrDCTSIZE*6) */
+    lh          s4, 114(a0)     /* quantptrDCTSIZE*7 */
+    lh          s5, 82(a0)      /* quantptrDCTSIZE*5 */
+    lh          t6, 82(t0)      /* inptrDCTSIZE*5 */
+    sll         t2, t2, 14      /* tmp0 <<= (CONST_BITS+1) */
+    lh          s6, 50(a0)      /* quantptrDCTSIZE*3 */
+    lh          t8, 18(t0)      /* inptrDCTSIZE*1 */
+    subu        v0, v0, t7      /* tmp2 =
+                                     MULTIPLY(z2, t5) - MULTIPLY(z3, t6) */
+    lh          t7, 50(t0)      /* inptrDCTSIZE*3 */
+    lh          s7, 18(a0)      /* quantptrDCTSIZE*1 */
+    mul         t5, s4, t5      /* z1 = (inptrDCTSIZE*7 *
+                                         quantptrDCTSIZE*7) */
+    mul         t6, s5, t6      /* z2 = (inptrDCTSIZE*5 *
+                                         quantptrDCTSIZE*5) */
+    mul         t7, s6, t7      /* z3 = (inptrDCTSIZE*3 *
+                                         quantptrDCTSIZE*3) */
+    mul         t8, s7, t8      /* z4 = (inptrDCTSIZE*1 *
+                                         quantptrDCTSIZE*1) */
+    addu        t3, t2, v0      /* tmp10 = tmp0 + z2 */
+    subu        t4, t2, v0      /* tmp10 = tmp0 - z2 */
+    mult        $ac0, zero, zero
+    mult        $ac1, zero, zero
+    ins         t5, t6, 16, 16
+    ins         t7, t8, 16, 16
+    dpa.w.ph    $ac0, t5, s0
+    dpa.w.ph    $ac0, t7, s1
+    dpa.w.ph    $ac1, t5, s2
+    dpa.w.ph    $ac1, t7, s3
+    mflo        t5, $ac0
+    mflo        t6, $ac1
+    addiu       t9, t9, -1
+    addiu       t0, t0, 2
+    addiu       a0, a0, 2
+    addiu       t1, t1, 4
+    addu        s5, t4, t5
+    subu        s4, t4, t5
+    addu        s6, t3, t6
+    subu        s7, t3, t6
+    shra_r.w    s5, s5, 12      /* DESCALE(tmp12 + temp1, 12) */
+    shra_r.w    s4, s4, 12      /* DESCALE(tmp12 - temp1, 12) */
+    shra_r.w    s6, s6, 12      /* DESCALE(tmp10 + temp2, 12) */
+    shra_r.w    s7, s7, 12      /* DESCALE(tmp10 - temp2, 12) */
+    sw          s5, 32(t1)
+    sw          s4, 64(t1)
+    sw          s6, 0(t1)
+    bgtz        t9, 1b
+     sw         s7, 96(t1)
+    move        t1, v1
+    li          s4, 15137
+    lw          s6, 8(t1)       /* wsptr2 */
+    li          s5, 6270
+    lw          s7, 24(t1)      /* wsptr6 */
+    mul         s4, s4, s6      /* MULTIPLY((JLONG)wsptr2,
+                                            FIX_1_847759065) */
+    lw          t2, 0(t1)       /* wsptr0 */
+    mul         s5, s5, s7      /* MULTIPLY((JLONG)wsptr6,
+                                            -FIX_0_765366865) */
+    lh          t5, 28(t1)      /* wsptr7 */
+    lh          t6, 20(t1)      /* wsptr5 */
+    lh          t7, 12(t1)      /* wsptr3 */
+    lh          t8, 4(t1)       /* wsptr1 */
+    ins         t5, t6, 16, 16
+    ins         t7, t8, 16, 16
+    mult        $ac0, zero, zero
+    dpa.w.ph    $ac0, t5, s0
+    dpa.w.ph    $ac0, t7, s1
+    mult        $ac1, zero, zero
+    dpa.w.ph    $ac1, t5, s2
+    dpa.w.ph    $ac1, t7, s3
+    sll         t2, t2, 14      /* tmp0 =
+                                     ((JLONG)wsptr0) << (CONST_BITS+1) */
+    mflo        s6, $ac0
+    /* MULTIPLY(wsptr2, FIX_1_847759065) +
+       MULTIPLY(wsptr6, -FIX_0_765366865) */
+    subu        s4, s4, s5
+    addu        t3, t2, s4      /* tmp10 = tmp0 + z2 */
+    mflo        s7, $ac1
+    subu        t4, t2, s4      /* tmp10 = tmp0 - z2 */
+    addu        t7, t4, s6
+    subu        t8, t4, s6
+    addu        t5, t3, s7
+    subu        t6, t3, s7
+    shra_r.w    t5, t5, 19      /* DESCALE(tmp10 + temp2, 19) */
+    shra_r.w    t6, t6, 19      /* DESCALE(tmp10 - temp2, 19) */
+    shra_r.w    t7, t7, 19      /* DESCALE(tmp12 + temp1, 19) */
+    shra_r.w    t8, t8, 19      /* DESCALE(tmp12 - temp1, 19) */
+    sll         s4, t9, 2
+    lw          v0, 0(a2)       /* output_bufctr */
+    shll_s.w    t5, t5, 24
+    shll_s.w    t6, t6, 24
+    shll_s.w    t7, t7, 24
+    shll_s.w    t8, t8, 24
+    sra         t5, t5, 24
+    sra         t6, t6, 24
+    sra         t7, t7, 24
+    sra         t8, t8, 24
+    addu        v0, v0, a3      /* outptr = output_bufctr + output_col */
+    addiu       t5, t5, 128
+    addiu       t6, t6, 128
+    addiu       t7, t7, 128
+    addiu       t8, t8, 128
+    sb          t5, 0(v0)
+    sb          t7, 1(v0)
+    sb          t8, 2(v0)
+    sb          t6, 3(v0)
+    /* 2 */
+    li          s4, 15137
+    lw          s6, 40(t1)      /* wsptr2 */
+    li          s5, 6270
+    lw          s7, 56(t1)      /* wsptr6 */
+    mul         s4, s4, s6      /* MULTIPLY((JLONG)wsptr2,
+                                            FIX_1_847759065) */
+    lw          t2, 32(t1)      /* wsptr0 */
+    mul         s5, s5, s7      /* MULTIPLY((JLONG)wsptr6,
+                                            -FIX_0_765366865) */
+    lh          t5, 60(t1)      /* wsptr7 */
+    lh          t6, 52(t1)      /* wsptr5 */
+    lh          t7, 44(t1)      /* wsptr3 */
+    lh          t8, 36(t1)      /* wsptr1 */
+    ins         t5, t6, 16, 16
+    ins         t7, t8, 16, 16
+    mult        $ac0, zero, zero
+    dpa.w.ph    $ac0, t5, s0
+    dpa.w.ph    $ac0, t7, s1
+    mult        $ac1, zero, zero
+    dpa.w.ph    $ac1, t5, s2
+    dpa.w.ph    $ac1, t7, s3
+    sll         t2, t2, 14      /* tmp0 =
+                                     ((JLONG)wsptr0) << (CONST_BITS+1) */
+    mflo        s6, $ac0
+    /* MULTIPLY(wsptr2, FIX_1_847759065) +
+       MULTIPLY(wsptr6, -FIX_0_765366865) */
+    subu        s4, s4, s5
+    addu        t3, t2, s4      /* tmp10 = tmp0 + z2 */
+    mflo        s7, $ac1
+    subu        t4, t2, s4      /* tmp10 = tmp0 - z2 */
+    addu        t7, t4, s6
+    subu        t8, t4, s6
+    addu        t5, t3, s7
+    subu        t6, t3, s7
+    shra_r.w    t5, t5, 19      /* DESCALE(tmp10 + temp2,
+                                           CONST_BITS-PASS1_BITS+1) */
+    shra_r.w    t6, t6, 19      /* DESCALE(tmp10 - temp2,
+                                           CONST_BITS-PASS1_BITS+1) */
+    shra_r.w    t7, t7, 19      /* DESCALE(tmp12 + temp1,
+                                           CONST_BITS-PASS1_BITS+1) */
+    shra_r.w    t8, t8, 19      /* DESCALE(tmp12 - temp1,
+                                           CONST_BITS-PASS1_BITS+1) */
+    sll         s4, t9, 2
+    lw          v0, 4(a2)       /* output_bufctr */
+    shll_s.w    t5, t5, 24
+    shll_s.w    t6, t6, 24
+    shll_s.w    t7, t7, 24
+    shll_s.w    t8, t8, 24
+    sra         t5, t5, 24
+    sra         t6, t6, 24
+    sra         t7, t7, 24
+    sra         t8, t8, 24
+    addu        v0, v0, a3      /* outptr = output_bufctr + output_col */
+    addiu       t5, t5, 128
+    addiu       t6, t6, 128
+    addiu       t7, t7, 128
+    addiu       t8, t8, 128
+    sb          t5, 0(v0)
+    sb          t7, 1(v0)
+    sb          t8, 2(v0)
+    sb          t6, 3(v0)
+    /* 3 */
+    li          s4, 15137
+    lw          s6, 72(t1)      /* wsptr2 */
+    li          s5, 6270
+    lw          s7, 88(t1)      /* wsptr6 */
+    mul         s4, s4, s6      /* MULTIPLY((JLONG)wsptr2,
+                                            FIX_1_847759065) */
+    lw          t2, 64(t1)      /* wsptr0 */
+    mul         s5, s5, s7      /* MULTIPLY((JLONG)wsptr6,
+                                            -FIX_0_765366865) */
+    lh          t5, 92(t1)      /* wsptr7 */
+    lh          t6, 84(t1)      /* wsptr5 */
+    lh          t7, 76(t1)      /* wsptr3 */
+    lh          t8, 68(t1)      /* wsptr1 */
+    ins         t5, t6, 16, 16
+    ins         t7, t8, 16, 16
+    mult        $ac0, zero, zero
+    dpa.w.ph    $ac0, t5, s0
+    dpa.w.ph    $ac0, t7, s1
+    mult        $ac1, zero, zero
+    dpa.w.ph    $ac1, t5, s2
+    dpa.w.ph    $ac1, t7, s3
+    sll         t2, t2, 14      /* tmp0 =
+                                     ((JLONG)wsptr0) << (CONST_BITS+1) */
+    mflo        s6, $ac0
+    /* MULTIPLY(wsptr2, FIX_1_847759065) +
+       MULTIPLY(wsptr6, -FIX_0_765366865) */
+    subu        s4, s4, s5
+    addu        t3, t2, s4      /* tmp10 = tmp0 + z2 */
+    mflo        s7, $ac1
+    subu        t4, t2, s4      /* tmp10 = tmp0 - z2 */
+    addu        t7, t4, s6
+    subu        t8, t4, s6
+    addu        t5, t3, s7
+    subu        t6, t3, s7
+    shra_r.w    t5, t5, 19      /* DESCALE(tmp10 + temp2, 19) */
+    shra_r.w    t6, t6, 19      /* DESCALE(tmp10 - temp2, 19) */
+    shra_r.w    t7, t7, 19      /* DESCALE(tmp12 + temp1, 19) */
+    shra_r.w    t8, t8, 19      /* DESCALE(tmp12 - temp1, 19) */
+    sll         s4, t9, 2
+    lw          v0, 8(a2)       /* output_bufctr */
+    shll_s.w    t5, t5, 24
+    shll_s.w    t6, t6, 24
+    shll_s.w    t7, t7, 24
+    shll_s.w    t8, t8, 24
+    sra         t5, t5, 24
+    sra         t6, t6, 24
+    sra         t7, t7, 24
+    sra         t8, t8, 24
+    addu        v0, v0, a3      /* outptr = output_bufctr + output_col */
+    addiu       t5, t5, 128
+    addiu       t6, t6, 128
+    addiu       t7, t7, 128
+    addiu       t8, t8, 128
+    sb          t5, 0(v0)
+    sb          t7, 1(v0)
+    sb          t8, 2(v0)
+    sb          t6, 3(v0)
+    li          s4, 15137
+    lw          s6, 104(t1)     /* wsptr2 */
+    li          s5, 6270
+    lw          s7, 120(t1)     /* wsptr6 */
+    mul         s4, s4, s6      /* MULTIPLY((JLONG)wsptr2,
+                                            FIX_1_847759065) */
+    lw          t2, 96(t1)      /* wsptr0 */
+    mul         s5, s5, s7      /* MULTIPLY((JLONG)wsptr6,
+                                            -FIX_0_765366865) */
+    lh          t5, 124(t1)     /* wsptr7 */
+    lh          t6, 116(t1)     /* wsptr5 */
+    lh          t7, 108(t1)     /* wsptr3 */
+    lh          t8, 100(t1)     /* wsptr1 */
+    ins         t5, t6, 16, 16
+    ins         t7, t8, 16, 16
+    mult        $ac0, zero, zero
+    dpa.w.ph    $ac0, t5, s0
+    dpa.w.ph    $ac0, t7, s1
+    mult        $ac1, zero, zero
+    dpa.w.ph    $ac1, t5, s2
+    dpa.w.ph    $ac1, t7, s3
+    sll         t2, t2, 14      /* tmp0 =
+                                     ((JLONG)wsptr0) << (CONST_BITS+1) */
+    mflo        s6, $ac0
+    /* MULTIPLY(wsptr2, FIX_1_847759065) +
+       MULTIPLY(wsptr6, -FIX_0_765366865) */
+    subu        s4, s4, s5
+    addu        t3, t2, s4      /* tmp10 = tmp0 + z2; */
+    mflo        s7, $ac1
+    subu        t4, t2, s4      /* tmp10 = tmp0 - z2; */
+    addu        t7, t4, s6
+    subu        t8, t4, s6
+    addu        t5, t3, s7
+    subu        t6, t3, s7
+    shra_r.w    t5, t5, 19      /* DESCALE(tmp10 + temp2, 19) */
+    shra_r.w    t6, t6, 19      /* DESCALE(tmp10 - temp2, 19) */
+    shra_r.w    t7, t7, 19      /* DESCALE(tmp12 + temp1, 19) */
+    shra_r.w    t8, t8, 19      /* DESCALE(tmp12 - temp1, 19) */
+    sll         s4, t9, 2
+    lw          v0, 12(a2)      /* output_bufctr */
+    shll_s.w    t5, t5, 24
+    shll_s.w    t6, t6, 24
+    shll_s.w    t7, t7, 24
+    shll_s.w    t8, t8, 24
+    sra         t5, t5, 24
+    sra         t6, t6, 24
+    sra         t7, t7, 24
+    sra         t8, t8, 24
+    addu        v0, v0, a3      /* outptr = output_bufctr + output_col */
+    addiu       t5, t5, 128
+    addiu       t6, t6, 128
+    addiu       t7, t7, 128
+    addiu       t8, t8, 128
+    sb          t5, 0(v0)
+    sb          t7, 1(v0)
+    sb          t8, 2(v0)
+    sb          t6, 3(v0)
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+END(jsimd_idct_4x4_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_idct_6x6_dspr2)
+/*
+ * a0 = compptr->dct_table
+ * a1 = coef_block
+ * a2 = output_buf
+ * a3 = output_col
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    addiu       sp, sp, -144
+    move        v0, sp
+    addiu       v1, v0, 24
+    addiu       t9, zero, 5793
+    addiu       s0, zero, 10033
+    addiu       s1, zero, 2998
+
+1:
+    lh          s2, 0(a0)       /* q0 = quantptr 0 */
+    lh          s3, 32(a0)      /* q1 = quantptr16 */
+    lh          s4, 64(a0)      /* q2 = quantptr32 */
+    lh          t2, 64(a1)      /* tmp2 = inptr32 */
+    lh          t1, 32(a1)      /* tmp1 = inptr16 */
+    lh          t0, 0(a1)       /* tmp0 = inptr 0 */
+    mul         t2, t2, s4      /* tmp2 = tmp2 * q2 */
+    mul         t1, t1, s3      /* tmp1 = tmp1 * q1 */
+    mul         t0, t0, s2      /* tmp0 = tmp0 * q0 */
+    lh          t6, 16(a1)      /* z1 = inptr 8 */
+    lh          t8, 80(a1)      /* z3 = inptr40 */
+    lh          t7, 48(a1)      /* z2 = inptr24 */
+    lh          s2, 16(a0)      /* q0 = quantptr 8 */
+    lh          s4, 80(a0)      /* q2 = quantptr40 */
+    lh          s3, 48(a0)      /* q1 = quantptr24 */
+    mul         t2, t2, t9      /* tmp2 = tmp2 * 5793 */
+    mul         t1, t1, s0      /* tmp1 = tmp1 * 10033 */
+    sll         t0, t0, 13      /* tmp0 = tmp0 << 13 */
+    mul         t6, t6, s2      /* z1 = z1 * q0 */
+    mul         t8, t8, s4      /* z3 = z3 * q2 */
+    mul         t7, t7, s3      /* z2 = z2 * q1 */
+    addu        t3, t0, t2      /* tmp10 = tmp0 + tmp2 */
+    sll         t2, t2, 1       /* tmp2 = tmp2 << 2 */
+    subu        t4, t0, t2      /* tmp11 = tmp0 - tmp2; */
+    subu        t5, t3, t1      /* tmp12 = tmp10 - tmp1 */
+    addu        t3, t3, t1      /* tmp10 = tmp10 + tmp1 */
+    addu        t1, t6, t8      /* tmp1 = z1 + z3 */
+    mul         t1, t1, s1      /* tmp1 = tmp1 * 2998 */
+    shra_r.w    t4, t4, 11      /* tmp11 = (tmp11 + 1024) >> 11 */
+    subu        t2, t6, t8      /* tmp2 = z1 - z3 */
+    subu        t2, t2, t7      /* tmp2 = tmp2 - z2 */
+    sll         t2, t2, 2       /* tmp2 = tmp2 << 2 */
+    addu        t0, t6, t7      /* tmp0 = z1 + z2 */
+    sll         t0, t0, 13      /* tmp0 = tmp0 << 13 */
+    subu        s2, t8, t7      /* q0 = z3 - z2 */
+    sll         s2, s2, 13      /* q0 = q0 << 13 */
+    addu        t0, t0, t1      /* tmp0 = tmp0 + tmp1 */
+    addu        t1, s2, t1      /* tmp1 = q0 + tmp1 */
+    addu        s2, t4, t2      /* q0 = tmp11 + tmp2 */
+    subu        s3, t4, t2      /* q1 = tmp11 - tmp2 */
+    addu        t6, t3, t0      /* z1 = tmp10 + tmp0 */
+    subu        t7, t3, t0      /* z2 = tmp10 - tmp0 */
+    addu        t4, t5, t1      /* tmp11 = tmp12 + tmp1 */
+    subu        t5, t5, t1      /* tmp12 = tmp12 - tmp1 */
+    shra_r.w    t6, t6, 11      /* z1 = (z1 + 1024) >> 11 */
+    shra_r.w    t7, t7, 11      /* z2 = (z2 + 1024) >> 11 */
+    shra_r.w    t4, t4, 11      /* tmp11 = (tmp11 + 1024) >> 11 */
+    shra_r.w    t5, t5, 11      /* tmp12 = (tmp12 + 1024) >> 11 */
+    sw          s2, 24(v0)
+    sw          s3, 96(v0)
+    sw          t6, 0(v0)
+    sw          t7, 120(v0)
+    sw          t4, 48(v0)
+    sw          t5, 72(v0)
+    addiu       v0, v0, 4
+    addiu       a1, a1, 2
+    bne         v0, v1, 1b
+     addiu      a0, a0, 2
+
+    /* Pass 2: process 6 rows from work array, store into output array. */
+    move        v0, sp
+    addiu       v1, v0, 144
+
+2:
+    lw          t0, 0(v0)
+    lw          t2, 16(v0)
+    lw          s5, 0(a2)
+    addiu       t0, t0, 16
+    sll         t0, t0, 13
+    mul         t3, t2, t9
+    lw          t6, 4(v0)
+    lw          t8, 20(v0)
+    lw          t7, 12(v0)
+    addu        s5, s5, a3
+    addu        s6, t6, t8
+    mul         s6, s6, s1
+    addu        t1, t0, t3
+    subu        t4, t0, t3
+    subu        t4, t4, t3
+    lw          t3, 8(v0)
+    mul         t0, t3, s0
+    addu        s7, t6, t7
+    sll         s7, s7, 13
+    addu        s7, s6, s7
+    subu        t2, t8, t7
+    sll         t2, t2, 13
+    addu        t2, s6, t2
+    subu        s6, t6, t7
+    subu        s6, s6, t8
+    sll         s6, s6, 13
+    addu        t3, t1, t0
+    subu        t5, t1, t0
+    addu        t6, t3, s7
+    subu        t3, t3, s7
+    addu        t7, t4, s6
+    subu        t4, t4, s6
+    addu        t8, t5, t2
+    subu        t5, t5, t2
+    shll_s.w    t6, t6, 6
+    shll_s.w    t3, t3, 6
+    shll_s.w    t7, t7, 6
+    shll_s.w    t4, t4, 6
+    shll_s.w    t8, t8, 6
+    shll_s.w    t5, t5, 6
+    sra         t6, t6, 24
+    addiu       t6, t6, 128
+    sra         t3, t3, 24
+    addiu       t3, t3, 128
+    sb          t6, 0(s5)
+    sra         t7, t7, 24
+    addiu       t7, t7, 128
+    sb          t3, 5(s5)
+    sra         t4, t4, 24
+    addiu       t4, t4, 128
+    sb          t7, 1(s5)
+    sra         t8, t8, 24
+    addiu       t8, t8, 128
+    sb          t4, 4(s5)
+    addiu       v0, v0, 24
+    sra         t5, t5, 24
+    addiu       t5, t5, 128
+    sb          t8, 2(s5)
+    addiu       a2, a2,  4
+    bne         v0, v1, 2b
+     sb         t5, 3(s5)
+
+    addiu       sp, sp, 144
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+
+END(jsimd_idct_6x6_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_idct_12x12_pass1_dspr2)
+/*
+ * a0 = compptr->dct_table
+ * a1 = coef_block
+ * a2 = workspace
+ */
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    li          a3, 8
+
+1:
+    /* odd part */
+    lh          t0, 48(a1)
+    lh          t1, 48(a0)
+    lh          t2, 16(a1)
+    lh          t3, 16(a0)
+    lh          t4, 80(a1)
+    lh          t5, 80(a0)
+    lh          t6, 112(a1)
+    lh          t7, 112(a0)
+    mul         t0, t0, t1      /* z2 */
+    mul         t1, t2, t3      /* z1 */
+    mul         t2, t4, t5      /* z3 */
+    mul         t3, t6, t7      /* z4 */
+    li          t4, 10703       /* FIX(1.306562965) */
+    li          t5, 4433        /* FIX_0_541196100 */
+    li          t6, 7053        /* FIX(0.860918669) */
+    mul         t4, t0, t4      /* tmp11 */
+    mul         t5, t0, t5      /* -tmp14 */
+    addu        t7, t1, t2      /* tmp10 */
+    addu        t8, t7, t3      /* tmp10 + z4 */
+    mul         t6, t6, t8      /* tmp15 */
+    li          t8, 2139        /* FIX(0.261052384) */
+    mul         t8, t7, t8      /* MULTIPLY(tmp10, FIX(0.261052384)) */
+    li          t7, 2295        /* FIX(0.280143716) */
+    mul         t7, t1, t7      /* MULTIPLY(z1, FIX(0.280143716)) */
+    addu        t9, t2, t3      /* z3 + z4 */
+    li          s0, 8565        /* FIX(1.045510580) */
+    mul         t9, t9, s0      /* -tmp13 */
+    li          s0, 12112       /* FIX(1.478575242) */
+    mul         s0, t2, s0      /* MULTIPLY(z3, FIX(1.478575242) */
+    li          s1, 12998       /* FIX(1.586706681) */
+    mul         s1, t3, s1      /* MULTIPLY(z4, FIX(1.586706681)) */
+    li          s2, 5540        /* FIX(0.676326758) */
+    mul         s2, t1, s2      /* MULTIPLY(z1, FIX(0.676326758)) */
+    li          s3, 16244       /* FIX(1.982889723) */
+    mul         s3, t3, s3      /* MULTIPLY(z4, FIX(1.982889723)) */
+    subu        t1, t1, t3      /* z1-=z4 */
+    subu        t0, t0, t2      /* z2-=z3 */
+    addu        t2, t0, t1      /* z1+z2 */
+    li          t3, 4433        /* FIX_0_541196100 */
+    mul         t2, t2, t3      /* z3 */
+    li          t3, 6270        /* FIX_0_765366865 */
+    mul         t1, t1, t3      /* MULTIPLY(z1, FIX_0_765366865) */
+    li          t3, 15137       /* FIX_0_765366865 */
+    mul         t0, t0, t3      /* MULTIPLY(z2, FIX_1_847759065) */
+    addu        t8, t6, t8      /* tmp12 */
+    addu        t3, t8, t4      /* tmp12 + tmp11 */
+    addu        t3, t3, t7      /* tmp10 */
+    subu        t8, t8, t9      /* tmp12 + tmp13 */
+    addu        s0, t5, s0
+    subu        t8, t8, s0      /* tmp12 */
+    subu        t9, t6, t9
+    subu        s1, s1, t4
+    addu        t9, t9, s1      /* tmp13 */
+    subu        t6, t6, t5
+    subu        t6, t6, s2
+    subu        t6, t6, s3      /* tmp15 */
+    /* even part start */
+    lh          t4, 64(a1)
+    lh          t5, 64(a0)
+    lh          t7, 32(a1)
+    lh          s0, 32(a0)
+    lh          s1, 0(a1)
+    lh          s2, 0(a0)
+    lh          s3, 96(a1)
+    lh          v0, 96(a0)
+    mul         t4, t4, t5      /* DEQUANTIZE(inptrDCTSIZE*4,
+                                              quantptrDCTSIZE*4) */
+    mul         t5, t7, s0      /* DEQUANTIZE(inptrDCTSIZE*2,
+                                              quantptrDCTSIZE*2) */
+    mul         t7, s1, s2      /* DEQUANTIZE(inptrDCTSIZE*0,
+                                              quantptrDCTSIZE*0) */
+    mul         s0, s3, v0      /* DEQUANTIZE(inptrDCTSIZE*6,
+                                              quantptrDCTSIZE*6) */
+    /* odd part end */
+    addu        t1, t2, t1      /* tmp11 */
+    subu        t0, t2, t0      /* tmp14 */
+    /* update counter and pointers */
+    addiu       a3, a3, -1
+    addiu       a0, a0, 2
+    addiu       a1, a1, 2
+    /* even part rest */
+    li          s1, 10033
+    li          s2, 11190
+    mul         t4, t4, s1      /* z4 */
+    mul         s1, t5, s2      /* z4 */
+    sll         t5, t5, 13      /* z1 */
+    sll         t7, t7, 13
+    addiu       t7, t7, 1024    /* z3 */
+    sll         s0, s0, 13      /* z2 */
+    addu        s2, t7, t4      /* tmp10 */
+    subu        t4, t7, t4      /* tmp11 */
+    subu        s3, t5, s0      /* tmp12 */
+    addu        t2, t7, s3      /* tmp21 */
+    subu        s3, t7, s3      /* tmp24 */
+    addu        t7, s1, s0      /* tmp12 */
+    addu        v0, s2, t7      /* tmp20 */
+    subu        s2, s2, t7      /* tmp25 */
+    subu        s1, s1, t5      /* z4 - z1 */
+    subu        s1, s1, s0      /* tmp12 */
+    addu        s0, t4, s1      /* tmp22 */
+    subu        t4, t4, s1      /* tmp23 */
+    /* final output stage */
+    addu        t5, v0, t3
+    subu        v0, v0, t3
+    addu        t3, t2, t1
+    subu        t2, t2, t1
+    addu        t1, s0, t8
+    subu        s0, s0, t8
+    addu        t8, t4, t9
+    subu        t4, t4, t9
+    addu        t9, s3, t0
+    subu        s3, s3, t0
+    addu        t0, s2, t6
+    subu        s2, s2, t6
+    sra         t5, t5, 11
+    sra         t3, t3, 11
+    sra         t1, t1, 11
+    sra         t8, t8, 11
+    sra         t9, t9, 11
+    sra         t0, t0, 11
+    sra         s2, s2, 11
+    sra         s3, s3, 11
+    sra         t4, t4, 11
+    sra         s0, s0, 11
+    sra         t2, t2, 11
+    sra         v0, v0, 11
+    sw          t5, 0(a2)
+    sw          t3, 32(a2)
+    sw          t1, 64(a2)
+    sw          t8, 96(a2)
+    sw          t9, 128(a2)
+    sw          t0, 160(a2)
+    sw          s2, 192(a2)
+    sw          s3, 224(a2)
+    sw          t4, 256(a2)
+    sw          s0, 288(a2)
+    sw          t2, 320(a2)
+    sw          v0, 352(a2)
+    bgtz        a3, 1b
+     addiu      a2, a2, 4
+
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    j           ra
+     nop
+
+END(jsimd_idct_12x12_pass1_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_idct_12x12_pass2_dspr2)
+/*
+ * a0 = workspace
+ * a1 = output
+ */
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    li          a3, 12
+
+1:
+    /* Odd part */
+    lw          t0, 12(a0)
+    lw          t1, 4(a0)
+    lw          t2, 20(a0)
+    lw          t3, 28(a0)
+    li          t4, 10703       /* FIX(1.306562965) */
+    li          t5, 4433        /* FIX_0_541196100 */
+    mul         t4, t0, t4      /* tmp11 */
+    mul         t5, t0, t5      /* -tmp14 */
+    addu        t6, t1, t2      /* tmp10 */
+    li          t7, 2139        /* FIX(0.261052384) */
+    mul         t7, t6, t7      /* MULTIPLY(tmp10, FIX(0.261052384)) */
+    addu        t6, t6, t3      /* tmp10 + z4 */
+    li          t8, 7053        /* FIX(0.860918669) */
+    mul         t6, t6, t8      /* tmp15 */
+    li          t8, 2295        /* FIX(0.280143716) */
+    mul         t8, t1, t8      /* MULTIPLY(z1, FIX(0.280143716)) */
+    addu        t9, t2, t3      /* z3 + z4 */
+    li          s0, 8565        /* FIX(1.045510580) */
+    mul         t9, t9, s0      /* -tmp13 */
+    li          s0, 12112       /* FIX(1.478575242) */
+    mul         s0, t2, s0      /* MULTIPLY(z3, FIX(1.478575242)) */
+    li          s1, 12998       /* FIX(1.586706681) */
+    mul         s1, t3, s1      /* MULTIPLY(z4, FIX(1.586706681)) */
+    li          s2, 5540        /* FIX(0.676326758) */
+    mul         s2, t1, s2      /* MULTIPLY(z1, FIX(0.676326758)) */
+    li          s3, 16244       /* FIX(1.982889723) */
+    mul         s3, t3, s3      /* MULTIPLY(z4, FIX(1.982889723)) */
+    subu        t1, t1, t3      /* z1 -= z4 */
+    subu        t0, t0, t2      /* z2 -= z3 */
+    addu        t2, t1, t0      /* z1 + z2 */
+    li          t3, 4433        /* FIX_0_541196100 */
+    mul         t2, t2, t3      /* z3 */
+    li          t3, 6270        /* FIX_0_765366865 */
+    mul         t1, t1, t3      /* MULTIPLY(z1, FIX_0_765366865) */
+    li          t3, 15137       /* FIX_1_847759065 */
+    mul         t0, t0, t3      /* MULTIPLY(z2, FIX_1_847759065) */
+    addu        t3, t6, t7      /* tmp12 */
+    addu        t7, t3, t4
+    addu        t7, t7, t8      /* tmp10 */
+    subu        t3, t3, t9
+    subu        t3, t3, t5
+    subu        t3, t3, s0      /* tmp12 */
+    subu        t9, t6, t9
+    subu        t9, t9, t4
+    addu        t9, t9, s1      /* tmp13 */
+    subu        t6, t6, t5
+    subu        t6, t6, s2
+    subu        t6, t6, s3      /* tmp15 */
+    addu        t1, t2, t1      /* tmp11 */
+    subu        t0, t2, t0      /* tmp14 */
+    /* even part */
+    lw          t2, 16(a0)      /* z4 */
+    lw          t4, 8(a0)       /* z1 */
+    lw          t5, 0(a0)       /* z3 */
+    lw          t8, 24(a0)      /* z2 */
+    li          s0, 10033       /* FIX(1.224744871) */
+    li          s1, 11190       /* FIX(1.366025404) */
+    mul         t2, t2, s0      /* z4 */
+    mul         s0, t4, s1      /* z4 */
+    addiu       t5, t5, 0x10
+    sll         t5, t5, 13      /* z3 */
+    sll         t4, t4, 13      /* z1 */
+    sll         t8, t8, 13      /* z2 */
+    subu        s1, t4, t8      /* tmp12 */
+    addu        s2, t5, t2      /* tmp10 */
+    subu        t2, t5, t2      /* tmp11 */
+    addu        s3, t5, s1      /* tmp21 */
+    subu        s1, t5, s1      /* tmp24 */
+    addu        t5, s0, t8      /* tmp12 */
+    addu        v0, s2, t5      /* tmp20 */
+    subu        t5, s2, t5      /* tmp25 */
+    subu        t4, s0, t4
+    subu        t4, t4, t8      /* tmp12 */
+    addu        t8, t2, t4      /* tmp22 */
+    subu        t2, t2, t4      /* tmp23 */
+    /* increment counter and pointers */
+    addiu       a3, a3, -1
+    addiu       a0, a0, 32
+    /* Final stage */
+    addu        t4, v0, t7
+    subu        v0, v0, t7
+    addu        t7, s3, t1
+    subu        s3, s3, t1
+    addu        t1, t8, t3
+    subu        t8, t8, t3
+    addu        t3, t2, t9
+    subu        t2, t2, t9
+    addu        t9, s1, t0
+    subu        s1, s1, t0
+    addu        t0, t5, t6
+    subu        t5, t5, t6
+    sll         t4, t4, 4
+    sll         t7, t7, 4
+    sll         t1, t1, 4
+    sll         t3, t3, 4
+    sll         t9, t9, 4
+    sll         t0, t0, 4
+    sll         t5, t5, 4
+    sll         s1, s1, 4
+    sll         t2, t2, 4
+    sll         t8, t8, 4
+    sll         s3, s3, 4
+    sll         v0, v0, 4
+    shll_s.w    t4, t4, 2
+    shll_s.w    t7, t7, 2
+    shll_s.w    t1, t1, 2
+    shll_s.w    t3, t3, 2
+    shll_s.w    t9, t9, 2
+    shll_s.w    t0, t0, 2
+    shll_s.w    t5, t5, 2
+    shll_s.w    s1, s1, 2
+    shll_s.w    t2, t2, 2
+    shll_s.w    t8, t8, 2
+    shll_s.w    s3, s3, 2
+    shll_s.w    v0, v0, 2
+    srl         t4, t4, 24
+    srl         t7, t7, 24
+    srl         t1, t1, 24
+    srl         t3, t3, 24
+    srl         t9, t9, 24
+    srl         t0, t0, 24
+    srl         t5, t5, 24
+    srl         s1, s1, 24
+    srl         t2, t2, 24
+    srl         t8, t8, 24
+    srl         s3, s3, 24
+    srl         v0, v0, 24
+    lw          t6, 0(a1)
+    addiu       t4, t4, 0x80
+    addiu       t7, t7, 0x80
+    addiu       t1, t1, 0x80
+    addiu       t3, t3, 0x80
+    addiu       t9, t9, 0x80
+    addiu       t0, t0, 0x80
+    addiu       t5, t5, 0x80
+    addiu       s1, s1, 0x80
+    addiu       t2, t2, 0x80
+    addiu       t8, t8, 0x80
+    addiu       s3, s3, 0x80
+    addiu       v0, v0, 0x80
+    sb          t4, 0(t6)
+    sb          t7, 1(t6)
+    sb          t1, 2(t6)
+    sb          t3, 3(t6)
+    sb          t9, 4(t6)
+    sb          t0, 5(t6)
+    sb          t5, 6(t6)
+    sb          s1, 7(t6)
+    sb          t2, 8(t6)
+    sb          t8, 9(t6)
+    sb          s3, 10(t6)
+    sb          v0, 11(t6)
+    bgtz        a3, 1b
+     addiu      a1, a1, 4
+
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    jr          ra
+     nop
+
+END(jsimd_idct_12x12_pass2_dspr2)
+
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_convsamp_dspr2)
+/*
+ * a0 = sample_data
+ * a1 = start_col
+ * a2 = workspace
+ */
+    lw            t0, 0(a0)
+    li            t7, 0xff80ff80
+    addu          t0, t0, a1
+    ulw           t1, 0(t0)
+    ulw           t2, 4(t0)
+    preceu.ph.qbr t3, t1
+    preceu.ph.qbl t4, t1
+    lw            t0, 4(a0)
+    preceu.ph.qbr t5, t2
+    preceu.ph.qbl t6, t2
+    addu          t0, t0, a1
+    addu.ph       t3, t3, t7
+    addu.ph       t4, t4, t7
+    ulw           t1, 0(t0)
+    ulw           t2, 4(t0)
+    addu.ph       t5, t5, t7
+    addu.ph       t6, t6, t7
+    usw           t3, 0(a2)
+    usw           t4, 4(a2)
+    preceu.ph.qbr t3, t1
+    preceu.ph.qbl t4, t1
+    usw           t5, 8(a2)
+    usw           t6, 12(a2)
+
+    lw            t0, 8(a0)
+    preceu.ph.qbr t5, t2
+    preceu.ph.qbl t6, t2
+    addu          t0, t0, a1
+    addu.ph       t3, t3, t7
+    addu.ph       t4, t4, t7
+    ulw           t1, 0(t0)
+    ulw           t2, 4(t0)
+    addu.ph       t5, t5, t7
+    addu.ph       t6, t6, t7
+    usw           t3, 16(a2)
+    usw           t4, 20(a2)
+    preceu.ph.qbr t3, t1
+    preceu.ph.qbl t4, t1
+    usw           t5, 24(a2)
+    usw           t6, 28(a2)
+
+    lw            t0, 12(a0)
+    preceu.ph.qbr t5, t2
+    preceu.ph.qbl t6, t2
+    addu          t0, t0, a1
+    addu.ph       t3, t3, t7
+    addu.ph       t4, t4, t7
+    ulw           t1, 0(t0)
+    ulw           t2, 4(t0)
+    addu.ph       t5, t5, t7
+    addu.ph       t6, t6, t7
+    usw           t3, 32(a2)
+    usw           t4, 36(a2)
+    preceu.ph.qbr t3, t1
+    preceu.ph.qbl t4, t1
+    usw           t5, 40(a2)
+    usw           t6, 44(a2)
+
+    lw            t0, 16(a0)
+    preceu.ph.qbr t5, t2
+    preceu.ph.qbl t6, t2
+    addu          t0, t0, a1
+    addu.ph       t3, t3, t7
+    addu.ph       t4, t4, t7
+    ulw           t1, 0(t0)
+    ulw           t2, 4(t0)
+    addu.ph       t5, t5, t7
+    addu.ph       t6, t6, t7
+    usw           t3, 48(a2)
+    usw           t4, 52(a2)
+    preceu.ph.qbr t3, t1
+    preceu.ph.qbl t4, t1
+    usw           t5, 56(a2)
+    usw           t6, 60(a2)
+
+    lw            t0, 20(a0)
+    preceu.ph.qbr t5, t2
+    preceu.ph.qbl t6, t2
+    addu          t0, t0, a1
+    addu.ph       t3, t3, t7
+    addu.ph       t4, t4, t7
+    ulw           t1, 0(t0)
+    ulw           t2, 4(t0)
+    addu.ph       t5, t5, t7
+    addu.ph       t6, t6, t7
+    usw           t3, 64(a2)
+    usw           t4, 68(a2)
+    preceu.ph.qbr t3, t1
+    preceu.ph.qbl t4, t1
+    usw           t5, 72(a2)
+    usw           t6, 76(a2)
+
+    lw            t0, 24(a0)
+    preceu.ph.qbr t5, t2
+    preceu.ph.qbl t6, t2
+    addu          t0, t0, a1
+    addu.ph       t3, t3, t7
+    addu.ph       t4, t4, t7
+    ulw           t1, 0(t0)
+    ulw           t2, 4(t0)
+    addu.ph       t5, t5, t7
+    addu.ph       t6, t6, t7
+    usw           t3, 80(a2)
+    usw           t4, 84(a2)
+    preceu.ph.qbr t3, t1
+    preceu.ph.qbl t4, t1
+    usw           t5, 88(a2)
+    usw           t6, 92(a2)
+
+    lw            t0, 28(a0)
+    preceu.ph.qbr t5, t2
+    preceu.ph.qbl t6, t2
+    addu          t0, t0, a1
+    addu.ph       t3, t3, t7
+    addu.ph       t4, t4, t7
+    ulw           t1, 0(t0)
+    ulw           t2, 4(t0)
+    addu.ph       t5, t5, t7
+    addu.ph       t6, t6, t7
+    usw           t3, 96(a2)
+    usw           t4, 100(a2)
+    preceu.ph.qbr t3, t1
+    preceu.ph.qbl t4, t1
+    usw           t5, 104(a2)
+    usw           t6, 108(a2)
+    preceu.ph.qbr t5, t2
+    preceu.ph.qbl t6, t2
+    addu.ph       t3, t3, t7
+    addu.ph       t4, t4, t7
+    addu.ph       t5, t5, t7
+    addu.ph       t6, t6, t7
+    usw           t3, 112(a2)
+    usw           t4, 116(a2)
+    usw           t5, 120(a2)
+    usw           t6, 124(a2)
+
+    j             ra
+     nop
+
+END(jsimd_convsamp_dspr2)
+
+
+#ifndef __mips_soft_float
+
+/*****************************************************************************/
+LEAF_DSPR2(jsimd_convsamp_float_dspr2)
+/*
+ * a0 = sample_data
+ * a1 = start_col
+ * a2 = workspace
+ */
+    .set at
+
+    lw          t0, 0(a0)
+    addu        t0, t0, a1
+    lbu         t1, 0(t0)
+    lbu         t2, 1(t0)
+    lbu         t3, 2(t0)
+    lbu         t4, 3(t0)
+    lbu         t5, 4(t0)
+    lbu         t6, 5(t0)
+    lbu         t7, 6(t0)
+    lbu         t8, 7(t0)
+    addiu       t1, t1, -128
+    addiu       t2, t2, -128
+    addiu       t3, t3, -128
+    addiu       t4, t4, -128
+    addiu       t5, t5, -128
+    addiu       t6, t6, -128
+    addiu       t7, t7, -128
+    addiu       t8, t8, -128
+    mtc1        t1, f2
+    mtc1        t2, f4
+    mtc1        t3, f6
+    mtc1        t4, f8
+    mtc1        t5, f10
+    mtc1        t6, f12
+    mtc1        t7, f14
+    mtc1        t8, f16
+    cvt.s.w     f2, f2
+    cvt.s.w     f4, f4
+    cvt.s.w     f6, f6
+    cvt.s.w     f8, f8
+    cvt.s.w     f10, f10
+    cvt.s.w     f12, f12
+    cvt.s.w     f14, f14
+    cvt.s.w     f16, f16
+    lw          t0, 4(a0)
+    swc1        f2, 0(a2)
+    swc1        f4, 4(a2)
+    swc1        f6, 8(a2)
+    addu        t0, t0, a1
+    swc1        f8, 12(a2)
+    swc1        f10, 16(a2)
+    swc1        f12, 20(a2)
+    swc1        f14, 24(a2)
+    swc1        f16, 28(a2)
+    /* elemr 1 */
+    lbu         t1, 0(t0)
+    lbu         t2, 1(t0)
+    lbu         t3, 2(t0)
+    lbu         t4, 3(t0)
+    lbu         t5, 4(t0)
+    lbu         t6, 5(t0)
+    lbu         t7, 6(t0)
+    lbu         t8, 7(t0)
+    addiu       t1, t1, -128
+    addiu       t2, t2, -128
+    addiu       t3, t3, -128
+    addiu       t4, t4, -128
+    addiu       t5, t5, -128
+    addiu       t6, t6, -128
+    addiu       t7, t7, -128
+    addiu       t8, t8, -128
+    mtc1        t1, f2
+    mtc1        t2, f4
+    mtc1        t3, f6
+    mtc1        t4, f8
+    mtc1        t5, f10
+    mtc1        t6, f12
+    mtc1        t7, f14
+    mtc1        t8, f16
+    cvt.s.w     f2, f2
+    cvt.s.w     f4, f4
+    cvt.s.w     f6, f6
+    cvt.s.w     f8, f8
+    cvt.s.w     f10, f10
+    cvt.s.w     f12, f12
+    cvt.s.w     f14, f14
+    cvt.s.w     f16, f16
+    lw          t0, 8(a0)
+    swc1        f2, 32(a2)
+    swc1        f4, 36(a2)
+    swc1        f6, 40(a2)
+    addu        t0, t0, a1
+    swc1        f8, 44(a2)
+    swc1        f10, 48(a2)
+    swc1        f12, 52(a2)
+    swc1        f14, 56(a2)
+    swc1        f16, 60(a2)
+    /* elemr 2 */
+    lbu         t1, 0(t0)
+    lbu         t2, 1(t0)
+    lbu         t3, 2(t0)
+    lbu         t4, 3(t0)
+    lbu         t5, 4(t0)
+    lbu         t6, 5(t0)
+    lbu         t7, 6(t0)
+    lbu         t8, 7(t0)
+    addiu       t1, t1, -128
+    addiu       t2, t2, -128
+    addiu       t3, t3, -128
+    addiu       t4, t4, -128
+    addiu       t5, t5, -128
+    addiu       t6, t6, -128
+    addiu       t7, t7, -128
+    addiu       t8, t8, -128
+    mtc1        t1, f2
+    mtc1        t2, f4
+    mtc1        t3, f6
+    mtc1        t4, f8
+    mtc1        t5, f10
+    mtc1        t6, f12
+    mtc1        t7, f14
+    mtc1        t8, f16
+    cvt.s.w     f2, f2
+    cvt.s.w     f4, f4
+    cvt.s.w     f6, f6
+    cvt.s.w     f8, f8
+    cvt.s.w     f10, f10
+    cvt.s.w     f12, f12
+    cvt.s.w     f14, f14
+    cvt.s.w     f16, f16
+    lw          t0, 12(a0)
+    swc1        f2, 64(a2)
+    swc1        f4, 68(a2)
+    swc1        f6, 72(a2)
+    addu        t0, t0, a1
+    swc1        f8, 76(a2)
+    swc1        f10, 80(a2)
+    swc1        f12, 84(a2)
+    swc1        f14, 88(a2)
+    swc1        f16, 92(a2)
+    /*  elemr 3 */
+    lbu         t1, 0(t0)
+    lbu         t2, 1(t0)
+    lbu         t3, 2(t0)
+    lbu         t4, 3(t0)
+    lbu         t5, 4(t0)
+    lbu         t6, 5(t0)
+    lbu         t7, 6(t0)
+    lbu         t8, 7(t0)
+    addiu       t1, t1, -128
+    addiu       t2, t2, -128
+    addiu       t3, t3, -128
+    addiu       t4, t4, -128
+    addiu       t5, t5, -128
+    addiu       t6, t6, -128
+    addiu       t7, t7, -128
+    addiu       t8, t8, -128
+    mtc1        t1, f2
+    mtc1        t2, f4
+    mtc1        t3, f6
+    mtc1        t4, f8
+    mtc1        t5, f10
+    mtc1        t6, f12
+    mtc1        t7, f14
+    mtc1        t8, f16
+    cvt.s.w     f2, f2
+    cvt.s.w     f4, f4
+    cvt.s.w     f6, f6
+    cvt.s.w     f8, f8
+    cvt.s.w     f10, f10
+    cvt.s.w     f12, f12
+    cvt.s.w     f14, f14
+    cvt.s.w     f16, f16
+    lw          t0, 16(a0)
+    swc1        f2, 96(a2)
+    swc1        f4, 100(a2)
+    swc1        f6, 104(a2)
+    addu        t0, t0, a1
+    swc1        f8, 108(a2)
+    swc1        f10, 112(a2)
+    swc1        f12, 116(a2)
+    swc1        f14, 120(a2)
+    swc1        f16, 124(a2)
+    /* elemr 4 */
+    lbu         t1, 0(t0)
+    lbu         t2, 1(t0)
+    lbu         t3, 2(t0)
+    lbu         t4, 3(t0)
+    lbu         t5, 4(t0)
+    lbu         t6, 5(t0)
+    lbu         t7, 6(t0)
+    lbu         t8, 7(t0)
+    addiu       t1, t1, -128
+    addiu       t2, t2, -128
+    addiu       t3, t3, -128
+    addiu       t4, t4, -128
+    addiu       t5, t5, -128
+    addiu       t6, t6, -128
+    addiu       t7, t7, -128
+    addiu       t8, t8, -128
+    mtc1        t1, f2
+    mtc1        t2, f4
+    mtc1        t3, f6
+    mtc1        t4, f8
+    mtc1        t5, f10
+    mtc1        t6, f12
+    mtc1        t7, f14
+    mtc1        t8, f16
+    cvt.s.w     f2, f2
+    cvt.s.w     f4, f4
+    cvt.s.w     f6, f6
+    cvt.s.w     f8, f8
+    cvt.s.w     f10, f10
+    cvt.s.w     f12, f12
+    cvt.s.w     f14, f14
+    cvt.s.w     f16, f16
+    lw          t0, 20(a0)
+    swc1        f2, 128(a2)
+    swc1        f4, 132(a2)
+    swc1        f6, 136(a2)
+    addu        t0, t0, a1
+    swc1        f8, 140(a2)
+    swc1        f10, 144(a2)
+    swc1        f12, 148(a2)
+    swc1        f14, 152(a2)
+    swc1        f16, 156(a2)
+    /* elemr 5 */
+    lbu         t1, 0(t0)
+    lbu         t2, 1(t0)
+    lbu         t3, 2(t0)
+    lbu         t4, 3(t0)
+    lbu         t5, 4(t0)
+    lbu         t6, 5(t0)
+    lbu         t7, 6(t0)
+    lbu         t8, 7(t0)
+    addiu       t1, t1, -128
+    addiu       t2, t2, -128
+    addiu       t3, t3, -128
+    addiu       t4, t4, -128
+    addiu       t5, t5, -128
+    addiu       t6, t6, -128
+    addiu       t7, t7, -128
+    addiu       t8, t8, -128
+    mtc1        t1, f2
+    mtc1        t2, f4
+    mtc1        t3, f6
+    mtc1        t4, f8
+    mtc1        t5, f10
+    mtc1        t6, f12
+    mtc1        t7, f14
+    mtc1        t8, f16
+    cvt.s.w     f2, f2
+    cvt.s.w     f4, f4
+    cvt.s.w     f6, f6
+    cvt.s.w     f8, f8
+    cvt.s.w     f10, f10
+    cvt.s.w     f12, f12
+    cvt.s.w     f14, f14
+    cvt.s.w     f16, f16
+    lw          t0, 24(a0)
+    swc1        f2, 160(a2)
+    swc1        f4, 164(a2)
+    swc1        f6, 168(a2)
+    addu        t0, t0, a1
+    swc1        f8, 172(a2)
+    swc1        f10, 176(a2)
+    swc1        f12, 180(a2)
+    swc1        f14, 184(a2)
+    swc1        f16, 188(a2)
+    /* elemr 6 */
+    lbu         t1, 0(t0)
+    lbu         t2, 1(t0)
+    lbu         t3, 2(t0)
+    lbu         t4, 3(t0)
+    lbu         t5, 4(t0)
+    lbu         t6, 5(t0)
+    lbu         t7, 6(t0)
+    lbu         t8, 7(t0)
+    addiu       t1, t1, -128
+    addiu       t2, t2, -128
+    addiu       t3, t3, -128
+    addiu       t4, t4, -128
+    addiu       t5, t5, -128
+    addiu       t6, t6, -128
+    addiu       t7, t7, -128
+    addiu       t8, t8, -128
+    mtc1        t1, f2
+    mtc1        t2, f4
+    mtc1        t3, f6
+    mtc1        t4, f8
+    mtc1        t5, f10
+    mtc1        t6, f12
+    mtc1        t7, f14
+    mtc1        t8, f16
+    cvt.s.w     f2, f2
+    cvt.s.w     f4, f4
+    cvt.s.w     f6, f6
+    cvt.s.w     f8, f8
+    cvt.s.w     f10, f10
+    cvt.s.w     f12, f12
+    cvt.s.w     f14, f14
+    cvt.s.w     f16, f16
+    lw          t0, 28(a0)
+    swc1        f2, 192(a2)
+    swc1        f4, 196(a2)
+    swc1        f6, 200(a2)
+    addu        t0, t0, a1
+    swc1        f8, 204(a2)
+    swc1        f10, 208(a2)
+    swc1        f12, 212(a2)
+    swc1        f14, 216(a2)
+    swc1        f16, 220(a2)
+    /* elemr 7 */
+    lbu         t1, 0(t0)
+    lbu         t2, 1(t0)
+    lbu         t3, 2(t0)
+    lbu         t4, 3(t0)
+    lbu         t5, 4(t0)
+    lbu         t6, 5(t0)
+    lbu         t7, 6(t0)
+    lbu         t8, 7(t0)
+    addiu       t1, t1, -128
+    addiu       t2, t2, -128
+    addiu       t3, t3, -128
+    addiu       t4, t4, -128
+    addiu       t5, t5, -128
+    addiu       t6, t6, -128
+    addiu       t7, t7, -128
+    addiu       t8, t8, -128
+    mtc1        t1, f2
+    mtc1        t2, f4
+    mtc1        t3, f6
+    mtc1        t4, f8
+    mtc1        t5, f10
+    mtc1        t6, f12
+    mtc1        t7, f14
+    mtc1        t8, f16
+    cvt.s.w     f2, f2
+    cvt.s.w     f4, f4
+    cvt.s.w     f6, f6
+    cvt.s.w     f8, f8
+    cvt.s.w     f10, f10
+    cvt.s.w     f12, f12
+    cvt.s.w     f14, f14
+    cvt.s.w     f16, f16
+    swc1        f2, 224(a2)
+    swc1        f4, 228(a2)
+    swc1        f6, 232(a2)
+    swc1        f8, 236(a2)
+    swc1        f10, 240(a2)
+    swc1        f12, 244(a2)
+    swc1        f14, 248(a2)
+    swc1        f16, 252(a2)
+
+    j           ra
+     nop
+
+END(jsimd_convsamp_float_dspr2)
+
+#endif
+
+/*****************************************************************************/

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips/jsimd_dspr2_asm.h Added

@@ -0,0 +1,292 @@
+/*
+ * MIPS DSPr2 optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2013, MIPS Technologies, Inc., California.
+ * Copyright (C) 2018, Matthieu Darbois.
+ * All Rights Reserved.
+ * Authors:  Teodora Novkovic (teodora.novkovic@imgtec.com)
+ *           Darko Laus       (darko.laus@imgtec.com)
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define zero  $0
+#define AT    $1
+#define v0    $2
+#define v1    $3
+#define a0    $4
+#define a1    $5
+#define a2    $6
+#define a3    $7
+#define t0    $8
+#define t1    $9
+#define t2    $10
+#define t3    $11
+#define t4    $12
+#define t5    $13
+#define t6    $14
+#define t7    $15
+#define s0    $16
+#define s1    $17
+#define s2    $18
+#define s3    $19
+#define s4    $20
+#define s5    $21
+#define s6    $22
+#define s7    $23
+#define t8    $24
+#define t9    $25
+#define k0    $26
+#define k1    $27
+#define gp    $28
+#define sp    $29
+#define fp    $30
+#define s8    $30
+#define ra    $31
+
+#define f0    $f0
+#define f1    $f1
+#define f2    $f2
+#define f3    $f3
+#define f4    $f4
+#define f5    $f5
+#define f6    $f6
+#define f7    $f7
+#define f8    $f8
+#define f9    $f9
+#define f10   $f10
+#define f11   $f11
+#define f12   $f12
+#define f13   $f13
+#define f14   $f14
+#define f15   $f15
+#define f16   $f16
+#define f17   $f17
+#define f18   $f18
+#define f19   $f19
+#define f20   $f20
+#define f21   $f21
+#define f22   $f22
+#define f23   $f23
+#define f24   $f24
+#define f25   $f25
+#define f26   $f26
+#define f27   $f27
+#define f28   $f28
+#define f29   $f29
+#define f30   $f30
+#define f31   $f31
+
+#ifdef __ELF__
+#define HIDDEN_SYMBOL(symbol)  .hidden symbol;
+#else
+#define HIDDEN_SYMBOL(symbol)
+#endif
+
+/*
+ * LEAF_MIPS32R2 - declare leaf routine for MIPS32r2
+ */
+#define LEAF_MIPS32R2(symbol) \
+    .globl      symbol; \
+    HIDDEN_SYMBOL(symbol) \
+    .align      2; \
+    .type       symbol, @function; \
+    .ent        symbol, 0; \
+symbol: \
+    .frame      sp, 0, ra; \
+    .set        push; \
+    .set        arch = mips32r2; \
+    .set        noreorder; \
+    .set        noat;
+
+/*
+ * LEAF_DSPR2 - declare leaf routine for MIPS DSPr2
+ */
+#define LEAF_DSPR2(symbol) \
+LEAF_MIPS32R2(symbol) \
+    .set        dspr2;
+
+/*
+ * END - mark end of function
+ */
+#define END(function) \
+    .set        pop; \
+    .end        function; \
+    .size       function, .-function
+
+/*
+ * Checks if stack offset is big enough for storing/restoring regs_num
+ * number of register to/from stack. Stack offset must be greater than
+ * or equal to the number of bytes needed for storing registers (regs_num*4).
+ * Since MIPS ABI allows usage of first 16 bytes of stack frame (this is
+ * preserved for input arguments of the functions, already stored in a0-a3),
+ * stack size can be further optimized by utilizing this space.
+ */
+.macro CHECK_STACK_OFFSET regs_num, stack_offset
+.if \stack_offset < \regs_num * 4 - 16
+.error "Stack offset too small."
+.endif
+.endm
+
+/*
+ * Saves set of registers on stack. Maximum number of registers that
+ * can be saved on stack is limitted to 14 (a0-a3, v0-v1 and s0-s7).
+ * Stack offset is number of bytes that are added to stack pointer (sp)
+ * before registers are pushed in order to provide enough space on stack
+ * (offset must be multiple of 4, and must be big enough, as described by
+ * CHECK_STACK_OFFSET macro). This macro is intended to be used in
+ * combination with RESTORE_REGS_FROM_STACK macro. Example:
+ *  SAVE_REGS_ON_STACK      4, v0, v1, s0, s1
+ *  RESTORE_REGS_FROM_STACK 4, v0, v1, s0, s1
+ */
+.macro SAVE_REGS_ON_STACK  stack_offset = 0, r1, \
+                           r2  = 0, r3  = 0, r4  = 0, \
+                           r5  = 0, r6  = 0, r7  = 0, \
+                           r8  = 0, r9  = 0, r10 = 0, \
+                           r11 = 0, r12 = 0, r13 = 0, \
+                           r14 = 0
+.if (\stack_offset < 0) || (\stack_offset - (\stack_offset / 4) * 4)
+    .error "Stack offset must be pozitive and multiple of 4."
+.endif
+.if \stack_offset != 0
+    addiu       sp, sp, -\stack_offset
+.endif
+    sw          \r1, 0(sp)
+.if \r2 != 0
+    sw          \r2, 4(sp)
+.endif
+.if \r3 != 0
+    sw          \r3, 8(sp)
+.endif
+.if \r4 != 0
+    sw          \r4, 12(sp)
+.endif
+.if \r5 != 0
+    CHECK_STACK_OFFSET 5, \stack_offset
+    sw          \r5, 16(sp)
+.endif
+.if \r6 != 0
+    CHECK_STACK_OFFSET 6, \stack_offset
+    sw          \r6, 20(sp)
+.endif
+.if \r7 != 0
+    CHECK_STACK_OFFSET 7, \stack_offset
+    sw          \r7, 24(sp)
+.endif
+.if \r8 != 0
+    CHECK_STACK_OFFSET 8, \stack_offset
+    sw          \r8, 28(sp)
+.endif
+.if \r9 != 0
+    CHECK_STACK_OFFSET 9, \stack_offset
+    sw          \r9, 32(sp)
+.endif
+.if \r10 != 0
+    CHECK_STACK_OFFSET 10, \stack_offset
+    sw          \r10, 36(sp)
+.endif
+.if \r11 != 0
+    CHECK_STACK_OFFSET 11, \stack_offset
+    sw          \r11, 40(sp)
+.endif
+.if \r12 != 0
+    CHECK_STACK_OFFSET 12, \stack_offset
+    sw          \r12, 44(sp)
+.endif
+.if \r13 != 0
+    CHECK_STACK_OFFSET 13, \stack_offset
+    sw          \r13, 48(sp)
+.endif
+.if \r14 != 0
+    CHECK_STACK_OFFSET 14, \stack_offset
+    sw          \r14, 52(sp)
+.endif
+.endm
+
+/*
+ * Restores set of registers from stack. Maximum number of registers that
+ * can be restored from stack is limitted to 14 (a0-a3, v0-v1 and s0-s7).
+ * Stack offset is number of bytes that are added to stack pointer (sp)
+ * after registers are restored (offset must be multiple of 4, and must
+ * be big enough, as described by CHECK_STACK_OFFSET macro). This macro is
+ * intended to be used in combination with RESTORE_REGS_FROM_STACK macro.
+ * Example:
+ *  SAVE_REGS_ON_STACK      4, v0, v1, s0, s1
+ *  RESTORE_REGS_FROM_STACK 4, v0, v1, s0, s1
+ */
+.macro RESTORE_REGS_FROM_STACK  stack_offset = 0, r1, \
+                                r2  = 0, r3  = 0, r4  = 0, \
+                                r5  = 0, r6  = 0, r7  = 0, \
+                                r8  = 0, r9  = 0, r10 = 0, \
+                                r11 = 0, r12 = 0, r13 = 0, \
+                                r14 = 0
+.if (\stack_offset < 0) || (\stack_offset - (\stack_offset / 4) * 4)
+    .error "Stack offset must be pozitive and multiple of 4."
+.endif
+    lw          \r1, 0(sp)
+.if \r2 != 0
+    lw          \r2, 4(sp)
+.endif
+.if \r3 != 0
+    lw          \r3, 8(sp)
+.endif
+.if \r4 != 0
+    lw          \r4, 12(sp)
+.endif
+.if \r5 != 0
+    CHECK_STACK_OFFSET 5, \stack_offset
+    lw          \r5, 16(sp)
+.endif
+.if \r6 != 0
+    CHECK_STACK_OFFSET 6, \stack_offset
+    lw          \r6, 20(sp)
+.endif
+.if \r7 != 0
+    CHECK_STACK_OFFSET 7, \stack_offset
+    lw          \r7, 24(sp)
+.endif
+.if \r8 != 0
+    CHECK_STACK_OFFSET 8, \stack_offset
+    lw          \r8, 28(sp)
+.endif
+.if \r9 != 0
+    CHECK_STACK_OFFSET 9, \stack_offset
+    lw          \r9, 32(sp)
+.endif
+.if \r10 != 0
+    CHECK_STACK_OFFSET 10, \stack_offset
+    lw          \r10, 36(sp)
+.endif
+.if \r11 != 0
+    CHECK_STACK_OFFSET 11, \stack_offset
+    lw          \r11, 40(sp)
+.endif
+.if \r12 != 0
+    CHECK_STACK_OFFSET 12, \stack_offset
+    lw          \r12, 44(sp)
+.endif
+.if \r13 != 0
+    CHECK_STACK_OFFSET 13, \stack_offset
+    lw          \r13, 48(sp)
+.endif
+.if \r14 != 0
+    CHECK_STACK_OFFSET 14, \stack_offset
+    lw          \r14, 52(sp)
+.endif
+.if \stack_offset != 0
+    addiu       sp, sp, \stack_offset
+.endif
+.endm

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64 Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jccolext-mmi.c Added

@@ -0,0 +1,455 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2014-2015, 2019, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           ZhangLixia  <zhanglixia-hf@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jccolor-mmi.c */
+
+
+#if RGB_RED == 0
+#define mmA  re
+#define mmB  ro
+#elif RGB_GREEN == 0
+#define mmA  ge
+#define mmB  go
+#elif RGB_BLUE == 0
+#define mmA  be
+#define mmB  bo
+#else
+#define mmA  xe
+#define mmB  xo
+#endif
+
+#if RGB_RED == 1
+#define mmC  re
+#define mmD  ro
+#elif RGB_GREEN == 1
+#define mmC  ge
+#define mmD  go
+#elif RGB_BLUE == 1
+#define mmC  be
+#define mmD  bo
+#else
+#define mmC  xe
+#define mmD  xo
+#endif
+
+#if RGB_RED == 2
+#define mmE  re
+#define mmF  ro
+#elif RGB_GREEN == 2
+#define mmE  ge
+#define mmF  go
+#elif RGB_BLUE == 2
+#define mmE  be
+#define mmF  bo
+#else
+#define mmE  xe
+#define mmF  xo
+#endif
+
+#if RGB_RED == 3
+#define mmG  re
+#define mmH  ro
+#elif RGB_GREEN == 3
+#define mmG  ge
+#define mmH  go
+#elif RGB_BLUE == 3
+#define mmG  be
+#define mmH  bo
+#else
+#define mmG  xe
+#define mmH  xo
+#endif
+
+
+void jsimd_rgb_ycc_convert_mmi(JDIMENSION image_width, JSAMPARRAY input_buf,
+                               JSAMPIMAGE output_buf, JDIMENSION output_row,
+                               int num_rows)
+{
+  JSAMPROW inptr, outptr0, outptr1, outptr2;
+  int num_cols, col;
+  __m64 re, ro, ge, go, be, bo, xe;
+#if RGB_PIXELSIZE == 4
+  __m64 xo;
+#endif
+  __m64 rgle, rghe, rglo, rgho, bgle, bghe, bglo, bgho;
+  __m64 ble, halfble, bhe, halfbhe, blo, halfblo, bho, halfbho;
+  __m64 rle, halfrle, rhe, halfrhe, rlo, halfrlo, rho, halfrho;
+  __m64 yle_rg, yhe_rg, yle_bg, yhe_bg, yle, yhe, ye;
+  __m64 ylo_rg, yho_rg, ylo_bg, yho_bg, ylo, yho, yo, y;
+  __m64 cble, cbhe, cbe, cblo, cbho, cbo, cb;
+  __m64 crle, crhe, cre, crlo, crho, cro, cr;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf0output_row;
+    outptr1 = output_buf1output_row;
+    outptr2 = output_buf2output_row;
+    output_row++;
+
+    for (num_cols = image_width; num_cols > 0; num_cols -= 8,
+         outptr0 += 8, outptr1 += 8, outptr2 += 8) {
+
+#if RGB_PIXELSIZE == 3
+
+      if (num_cols < 8) {
+        col = num_cols * 3;
+        asm(".set noreorder\r\n"
+
+            "li       $8, 1\r\n"
+            "move     $9, %3\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 1f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 1\r\n"
+            "xor      $12, $12, $12\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $9\r\n"
+            "lbu      $12, 0($13)\r\n"
+
+            "1:       \r\n"
+            "li       $8, 2\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 2f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 2\r\n"
+            "xor      $11, $11, $11\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $9\r\n"
+            "lhu      $11, 0($13)\r\n"
+            "sll      $12, $12, 16\r\n"
+            "or       $12, $12, $11\r\n"
+
+            "2:       \r\n"
+            "dmtc1    $12, %0\r\n"
+            "li       $8, 4\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 3f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 4\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $9\r\n"
+            "lwu      $14, 0($13)\r\n"
+            "dmtc1    $14, %1\r\n"
+            "dsll32   $12, $12, 0\r\n"
+            "or       $12, $12, $14\r\n"
+            "dmtc1    $12, %0\r\n"
+
+            "3:       \r\n"
+            "li       $8, 8\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 4f\r\n"
+            "nop      \r\n"
+            "mov.s    %1, %0\r\n"
+            "ldc1     %0, 0(%5)\r\n"
+            "li       $9, 8\r\n"
+            "j        5f\r\n"
+            "nop      \r\n"
+
+            "4:       \r\n"
+            "li       $8, 16\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 5f\r\n"
+            "nop      \r\n"
+            "mov.s    %2, %0\r\n"
+            "ldc1     %0, 0(%5)\r\n"
+            "ldc1     %1, 8(%5)\r\n"
+
+            "5:       \r\n"
+            "nop      \r\n"
+            ".set reorder\r\n"
+
+            : "=f" (mmA), "=f" (mmG), "=f" (mmF)
+            : "r" (col), "r" (num_rows), "r" (inptr)
+            : "$f0", "$f2", "$f4", "$8", "$9", "$10", "$11", "$12", "$13",
+              "$14", "memory"
+           );
+      } else {
+        if (!(((long)inptr) & 7)) {
+          mmA = _mm_load_si64((__m64 *)&inptr0);
+          mmG = _mm_load_si64((__m64 *)&inptr8);
+          mmF = _mm_load_si64((__m64 *)&inptr16);
+        } else {
+          mmA = _mm_loadu_si64((__m64 *)&inptr0);
+          mmG = _mm_loadu_si64((__m64 *)&inptr8);
+          mmF = _mm_loadu_si64((__m64 *)&inptr16);
+        }
+        inptr += RGB_PIXELSIZE * 8;
+      }
+      mmD = _mm_srli_si64(mmA, 4 * BYTE_BIT);
+      mmA = _mm_slli_si64(mmA, 4 * BYTE_BIT);
+
+      mmA = _mm_unpackhi_pi8(mmA, mmG);
+      mmG = _mm_slli_si64(mmG, 4 * BYTE_BIT);
+
+      mmD = _mm_unpacklo_pi8(mmD, mmF);
+      mmG = _mm_unpackhi_pi8(mmG, mmF);
+
+      mmE = _mm_srli_si64(mmA, 4 * BYTE_BIT);
+      mmA = _mm_slli_si64(mmA, 4 * BYTE_BIT);
+
+      mmA = _mm_unpackhi_pi8(mmA, mmD);
+      mmD = _mm_slli_si64(mmD, 4 * BYTE_BIT);
+
+      mmE = _mm_unpacklo_pi8(mmE, mmG);
+      mmD = _mm_unpackhi_pi8(mmD, mmG);
+      mmC = _mm_loadhi_pi8_f(mmA);
+      mmA = _mm_loadlo_pi8_f(mmA);
+
+      mmB = _mm_loadhi_pi8_f(mmE);
+      mmE = _mm_loadlo_pi8_f(mmE);
+
+      mmF = _mm_loadhi_pi8_f(mmD);
+      mmD = _mm_loadlo_pi8_f(mmD);
+
+#else  /* RGB_PIXELSIZE == 4 */
+
+      if (num_cols < 8) {
+        col = num_cols;
+        asm(".set noreorder\r\n"
+
+            "li       $8, 1\r\n"
+            "move     $9, %4\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 1f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 1\r\n"
+            PTR_SLL   "$11, $9, 2\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $11\r\n"
+            "lwc1     %0, 0($13)\r\n"
+
+            "1:       \r\n"
+            "li       $8, 2\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 2f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 2\r\n"
+            PTR_SLL   "$11, $9, 2\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $11\r\n"
+            "mov.s    %1, %0\r\n"
+            "ldc1     %0, 0($13)\r\n"
+
+            "2:       \r\n"
+            "li       $8, 4\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 3f\r\n"
+            "nop      \r\n"
+            "mov.s    %2, %0\r\n"
+            "mov.s    %3, %1\r\n"
+            "ldc1     %0, 0(%5)\r\n"
+            "ldc1     %1, 8(%5)\r\n"
+
+            "3:       \r\n"
+            "nop      \r\n"
+            ".set reorder\r\n"
+
+            : "=f" (mmA), "=f" (mmF), "=f" (mmD), "=f" (mmC)
+            : "r" (col), "r" (inptr)
+            : "$f0", "$f2", "$8", "$9", "$10", "$11", "$13", "memory"
+           );
+      } else {
+        if (!(((long)inptr) & 7)) {
+          mmA = _mm_load_si64((__m64 *)&inptr0);
+          mmF = _mm_load_si64((__m64 *)&inptr8);
+          mmD = _mm_load_si64((__m64 *)&inptr16);
+          mmC = _mm_load_si64((__m64 *)&inptr24);
+        } else {
+          mmA = _mm_loadu_si64((__m64 *)&inptr0);
+          mmF = _mm_loadu_si64((__m64 *)&inptr8);
+          mmD = _mm_loadu_si64((__m64 *)&inptr16);
+          mmC = _mm_loadu_si64((__m64 *)&inptr24);
+        }
+        inptr += RGB_PIXELSIZE * 8;
+      }
+      mmB = _mm_unpackhi_pi8(mmA, mmF);
+      mmA = _mm_unpacklo_pi8(mmA, mmF);
+
+      mmG = _mm_unpackhi_pi8(mmD, mmC);
+      mmD = _mm_unpacklo_pi8(mmD, mmC);
+
+      mmE = _mm_unpackhi_pi16(mmA, mmD);
+      mmA = _mm_unpacklo_pi16(mmA, mmD);
+
+      mmH = _mm_unpackhi_pi16(mmB, mmG);
+      mmB = _mm_unpacklo_pi16(mmB, mmG);
+
+      mmC = _mm_loadhi_pi8_f(mmA);
+      mmA = _mm_loadlo_pi8_f(mmA);
+
+      mmD = _mm_loadhi_pi8_f(mmB);
+      mmB = _mm_loadlo_pi8_f(mmB);
+
+      mmG = _mm_loadhi_pi8_f(mmE);
+      mmE = _mm_loadlo_pi8_f(mmE);
+
+      mmF = _mm_unpacklo_pi8(mmH, mmH);
+      mmH = _mm_unpackhi_pi8(mmH, mmH);
+      mmF = _mm_srli_pi16(mmF, BYTE_BIT);
+      mmH = _mm_srli_pi16(mmH, BYTE_BIT);
+
+#endif
+
+      /* re=(R0 R2 R4 R6), ge=(G0 G2 G4 G6), be=(B0 B2 B4 B6)
+       * ro=(R1 R3 R5 R7), go=(G1 G3 G5 G7), bo=(B1 B3 B5 B7)
+       *
+       * (Original)
+       * Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       *
+       * (This implementation)
+       * Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       */
+
+      rglo = _mm_unpacklo_pi16(ro, go);
+      rgho = _mm_unpackhi_pi16(ro, go);
+      ylo_rg = _mm_madd_pi16(rglo, PW_F0299_F0337);
+      yho_rg = _mm_madd_pi16(rgho, PW_F0299_F0337);
+      cblo = _mm_madd_pi16(rglo, PW_MF016_MF033);
+      cbho = _mm_madd_pi16(rgho, PW_MF016_MF033);
+
+      blo = _mm_loadlo_pi16_f(bo);
+      bho = _mm_loadhi_pi16_f(bo);
+      halfblo = _mm_srli_pi32(blo, 1);
+      halfbho = _mm_srli_pi32(bho, 1);
+
+      cblo = _mm_add_pi32(cblo, halfblo);
+      cbho = _mm_add_pi32(cbho, halfbho);
+      cblo = _mm_add_pi32(cblo, PD_ONEHALFM1_CJ);
+      cbho = _mm_add_pi32(cbho, PD_ONEHALFM1_CJ);
+      cblo = _mm_srli_pi32(cblo, SCALEBITS);
+      cbho = _mm_srli_pi32(cbho, SCALEBITS);
+      cbo = _mm_packs_pi32(cblo, cbho);
+
+      rgle = _mm_unpacklo_pi16(re, ge);
+      rghe = _mm_unpackhi_pi16(re, ge);
+      yle_rg = _mm_madd_pi16(rgle, PW_F0299_F0337);
+      yhe_rg = _mm_madd_pi16(rghe, PW_F0299_F0337);
+      cble = _mm_madd_pi16(rgle, PW_MF016_MF033);
+      cbhe = _mm_madd_pi16(rghe, PW_MF016_MF033);
+
+      ble = _mm_loadlo_pi16_f(be);
+      bhe = _mm_loadhi_pi16_f(be);
+      halfble = _mm_srli_pi32(ble, 1);
+      halfbhe = _mm_srli_pi32(bhe, 1);
+
+      cble = _mm_add_pi32(cble, halfble);
+      cbhe = _mm_add_pi32(cbhe, halfbhe);
+      cble = _mm_add_pi32(cble, PD_ONEHALFM1_CJ);
+      cbhe = _mm_add_pi32(cbhe, PD_ONEHALFM1_CJ);
+      cble = _mm_srli_pi32(cble, SCALEBITS);
+      cbhe = _mm_srli_pi32(cbhe, SCALEBITS);
+      cbe = _mm_packs_pi32(cble, cbhe);
+
+      cbo = _mm_slli_pi16(cbo, BYTE_BIT);
+      cb = _mm_or_si64(cbe, cbo);
+
+      bglo = _mm_unpacklo_pi16(bo, go);
+      bgho = _mm_unpackhi_pi16(bo, go);
+      ylo_bg = _mm_madd_pi16(bglo, PW_F0114_F0250);
+      yho_bg = _mm_madd_pi16(bgho, PW_F0114_F0250);
+      crlo = _mm_madd_pi16(bglo, PW_MF008_MF041);
+      crho = _mm_madd_pi16(bgho, PW_MF008_MF041);
+
+      ylo = _mm_add_pi32(ylo_bg, ylo_rg);
+      yho = _mm_add_pi32(yho_bg, yho_rg);
+      ylo = _mm_add_pi32(ylo, PD_ONEHALF);
+      yho = _mm_add_pi32(yho, PD_ONEHALF);
+      ylo = _mm_srli_pi32(ylo, SCALEBITS);
+      yho = _mm_srli_pi32(yho, SCALEBITS);
+      yo = _mm_packs_pi32(ylo, yho);
+
+      rlo = _mm_loadlo_pi16_f(ro);
+      rho = _mm_loadhi_pi16_f(ro);
+      halfrlo = _mm_srli_pi32(rlo, 1);
+      halfrho = _mm_srli_pi32(rho, 1);
+
+      crlo = _mm_add_pi32(crlo, halfrlo);
+      crho = _mm_add_pi32(crho, halfrho);
+      crlo = _mm_add_pi32(crlo, PD_ONEHALFM1_CJ);
+      crho = _mm_add_pi32(crho, PD_ONEHALFM1_CJ);
+      crlo = _mm_srli_pi32(crlo, SCALEBITS);
+      crho = _mm_srli_pi32(crho, SCALEBITS);
+      cro = _mm_packs_pi32(crlo, crho);
+
+      bgle = _mm_unpacklo_pi16(be, ge);
+      bghe = _mm_unpackhi_pi16(be, ge);
+      yle_bg = _mm_madd_pi16(bgle, PW_F0114_F0250);
+      yhe_bg = _mm_madd_pi16(bghe, PW_F0114_F0250);
+      crle = _mm_madd_pi16(bgle, PW_MF008_MF041);
+      crhe = _mm_madd_pi16(bghe, PW_MF008_MF041);
+
+      yle = _mm_add_pi32(yle_bg, yle_rg);
+      yhe = _mm_add_pi32(yhe_bg, yhe_rg);
+      yle = _mm_add_pi32(yle, PD_ONEHALF);
+      yhe = _mm_add_pi32(yhe, PD_ONEHALF);
+      yle = _mm_srli_pi32(yle, SCALEBITS);
+      yhe = _mm_srli_pi32(yhe, SCALEBITS);
+      ye = _mm_packs_pi32(yle, yhe);
+
+      yo = _mm_slli_pi16(yo, BYTE_BIT);
+      y = _mm_or_si64(ye, yo);
+
+      rle = _mm_loadlo_pi16_f(re);
+      rhe = _mm_loadhi_pi16_f(re);
+      halfrle = _mm_srli_pi32(rle, 1);
+      halfrhe = _mm_srli_pi32(rhe, 1);
+
+      crle = _mm_add_pi32(crle, halfrle);
+      crhe = _mm_add_pi32(crhe, halfrhe);
+      crle = _mm_add_pi32(crle, PD_ONEHALFM1_CJ);
+      crhe = _mm_add_pi32(crhe, PD_ONEHALFM1_CJ);
+      crle = _mm_srli_pi32(crle, SCALEBITS);
+      crhe = _mm_srli_pi32(crhe, SCALEBITS);
+      cre = _mm_packs_pi32(crle, crhe);
+
+      cro = _mm_slli_pi16(cro, BYTE_BIT);
+      cr = _mm_or_si64(cre, cro);
+
+      _mm_store_si64((__m64 *)&outptr00, y);
+      _mm_store_si64((__m64 *)&outptr10, cb);
+      _mm_store_si64((__m64 *)&outptr20, cr);
+    }
+  }
+}
+
+#undef mmA
+#undef mmB
+#undef mmC
+#undef mmD
+#undef mmE
+#undef mmF
+#undef mmG
+#undef mmH

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jccolor-mmi.c Added

@@ -0,0 +1,148 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2011, 2014, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2017, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* RGB --> YCC CONVERSION */
+
+#include "jsimd_mmi.h"
+
+
+#define F_0_081  ((short)5329)                /* FIX(0.08131) */
+#define F_0_114  ((short)7471)                /* FIX(0.11400) */
+#define F_0_168  ((short)11059)               /* FIX(0.16874) */
+#define F_0_250  ((short)16384)               /* FIX(0.25000) */
+#define F_0_299  ((short)19595)               /* FIX(0.29900) */
+#define F_0_331  ((short)21709)               /* FIX(0.33126) */
+#define F_0_418  ((short)27439)               /* FIX(0.41869) */
+#define F_0_587  ((short)38470)               /* FIX(0.58700) */
+#define F_0_337  ((short)(F_0_587 - F_0_250)) /* FIX(0.58700) - FIX(0.25000) */
+
+enum const_index {
+  index_PD_ONEHALF,
+  index_PW_F0299_F0337,
+  index_PW_F0114_F0250,
+  index_PW_MF016_MF033,
+  index_PW_MF008_MF041,
+  index_PD_ONEHALFM1_CJ
+};
+
+static uint64_t const_value = {
+  _uint64_set_pi32((int)(1 << (SCALEBITS - 1)), (int)(1 << (SCALEBITS - 1))),
+  _uint64_set_pi16(F_0_337, F_0_299, F_0_337, F_0_299),
+  _uint64_set_pi16(F_0_250, F_0_114, F_0_250, F_0_114),
+  _uint64_set_pi16(-F_0_331, -F_0_168, -F_0_331, -F_0_168),
+  _uint64_set_pi16(-F_0_418, -F_0_081, -F_0_418, -F_0_081),
+  _uint64_set_pi32(((1 << (SCALEBITS - 1)) - 1 + (CENTERJSAMPLE << SCALEBITS)),
+                   ((1 << (SCALEBITS - 1)) - 1 + (CENTERJSAMPLE << SCALEBITS)))
+};
+
+#define get_const_value(index)  (*(__m64 *)&const_valueindex)
+
+#define PD_ONEHALF       get_const_value(index_PD_ONEHALF)
+#define PW_F0299_F0337   get_const_value(index_PW_F0299_F0337)
+#define PW_F0114_F0250   get_const_value(index_PW_F0114_F0250)
+#define PW_MF016_MF033   get_const_value(index_PW_MF016_MF033)
+#define PW_MF008_MF041   get_const_value(index_PW_MF008_MF041)
+#define PD_ONEHALFM1_CJ  get_const_value(index_PD_ONEHALFM1_CJ)
+
+
+#include "jccolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_rgb_ycc_convert_mmi  jsimd_extrgb_ycc_convert_mmi
+#include "jccolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_mmi
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define jsimd_rgb_ycc_convert_mmi  jsimd_extrgbx_ycc_convert_mmi
+#include "jccolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_mmi
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define jsimd_rgb_ycc_convert_mmi  jsimd_extbgr_ycc_convert_mmi
+#include "jccolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_mmi
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define jsimd_rgb_ycc_convert_mmi  jsimd_extbgrx_ycc_convert_mmi
+#include "jccolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_mmi
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define jsimd_rgb_ycc_convert_mmi  jsimd_extxbgr_ycc_convert_mmi
+#include "jccolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_mmi
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define jsimd_rgb_ycc_convert_mmi  jsimd_extxrgb_ycc_convert_mmi
+#include "jccolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_ycc_convert_mmi

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jcgray-mmi.c Added

@@ -0,0 +1,132 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2011, 2014, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhangLixia <zhanglixia-hf@loongson.cn>
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* RGB --> GRAYSCALE CONVERSION */
+
+#include "jsimd_mmi.h"
+
+
+#define F_0_114  ((short)7471)                /* FIX(0.11400) */
+#define F_0_250  ((short)16384)               /* FIX(0.25000) */
+#define F_0_299  ((short)19595)               /* FIX(0.29900) */
+#define F_0_587  ((short)38470)               /* FIX(0.58700) */
+#define F_0_337  ((short)(F_0_587 - F_0_250)) /* FIX(0.58700) - FIX(0.25000) */
+
+enum const_index {
+  index_PD_ONEHALF,
+  index_PW_F0299_F0337,
+  index_PW_F0114_F0250
+};
+
+static uint64_t const_value = {
+  _uint64_set_pi32((int)(1 << (SCALEBITS - 1)), (int)(1 << (SCALEBITS - 1))),
+  _uint64_set_pi16(F_0_337, F_0_299, F_0_337, F_0_299),
+  _uint64_set_pi16(F_0_250, F_0_114, F_0_250, F_0_114)
+};
+
+#define get_const_value(index)  (*(__m64 *)&const_valueindex)
+
+#define PD_ONEHALF       get_const_value(index_PD_ONEHALF)
+#define PW_F0299_F0337   get_const_value(index_PW_F0299_F0337)
+#define PW_F0114_F0250   get_const_value(index_PW_F0114_F0250)
+
+
+#include "jcgryext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_rgb_gray_convert_mmi  jsimd_extrgb_gray_convert_mmi
+#include "jcgryext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_mmi
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define jsimd_rgb_gray_convert_mmi  jsimd_extrgbx_gray_convert_mmi
+#include "jcgryext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_mmi
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define jsimd_rgb_gray_convert_mmi  jsimd_extbgr_gray_convert_mmi
+#include "jcgryext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_mmi
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define jsimd_rgb_gray_convert_mmi  jsimd_extbgrx_gray_convert_mmi
+#include "jcgryext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_mmi
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define jsimd_rgb_gray_convert_mmi  jsimd_extxbgr_gray_convert_mmi
+#include "jcgryext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_mmi
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define jsimd_rgb_gray_convert_mmi  jsimd_extxrgb_gray_convert_mmi
+#include "jcgryext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_rgb_gray_convert_mmi

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jcgryext-mmi.c Added

@@ -0,0 +1,374 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2014-2015, 2019, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhangLixia <zhanglixia-hf@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jcgray-mmi.c */
+
+
+#if RGB_RED == 0
+#define mmA  re
+#define mmB  ro
+#elif RGB_GREEN == 0
+#define mmA  ge
+#define mmB  go
+#elif RGB_BLUE == 0
+#define mmA  be
+#define mmB  bo
+#else
+#define mmA  xe
+#define mmB  xo
+#endif
+
+#if RGB_RED == 1
+#define mmC  re
+#define mmD  ro
+#elif RGB_GREEN == 1
+#define mmC  ge
+#define mmD  go
+#elif RGB_BLUE == 1
+#define mmC  be
+#define mmD  bo
+#else
+#define mmC  xe
+#define mmD  xo
+#endif
+
+#if RGB_RED == 2
+#define mmE  re
+#define mmF  ro
+#elif RGB_GREEN == 2
+#define mmE  ge
+#define mmF  go
+#elif RGB_BLUE == 2
+#define mmE  be
+#define mmF  bo
+#else
+#define mmE  xe
+#define mmF  xo
+#endif
+
+#if RGB_RED == 3
+#define mmG  re
+#define mmH  ro
+#elif RGB_GREEN == 3
+#define mmG  ge
+#define mmH  go
+#elif RGB_BLUE == 3
+#define mmG  be
+#define mmH  bo
+#else
+#define mmG  xe
+#define mmH  xo
+#endif
+
+
+void jsimd_rgb_gray_convert_mmi(JDIMENSION image_width, JSAMPARRAY input_buf,
+                                JSAMPIMAGE output_buf, JDIMENSION output_row,
+                                int num_rows)
+{
+  JSAMPROW inptr, outptr;
+  int num_cols, col;
+  __m64 re, ro, ge, go, be, bo, xe;
+#if RGB_PIXELSIZE == 4
+  __m64 xo;
+#endif
+  __m64 rgle, rghe, rglo, rgho, bgle, bghe, bglo, bgho;
+  __m64 yle_rg, yhe_rg, yle_bg, yhe_bg, yle, yhe, ye;
+  __m64 ylo_rg, yho_rg, ylo_bg, yho_bg, ylo, yho, yo, y;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf0output_row;
+    output_row++;
+
+    for (num_cols = image_width; num_cols > 0; num_cols -= 8,
+         outptr += 8) {
+
+#if RGB_PIXELSIZE == 3
+
+      if (num_cols < 8) {
+        col = num_cols * 3;
+        asm(".set noreorder\r\n"
+
+            "li       $8, 1\r\n"
+            "move     $9, %3\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 1f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 1\r\n"
+            "xor      $12, $12, $12\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $9\r\n"
+            "lbu      $12, 0($13)\r\n"
+
+            "1:       \r\n"
+            "li       $8, 2\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 2f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 2\r\n"
+            "xor      $11, $11, $11\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $9\r\n"
+            "lhu      $11, 0($13)\r\n"
+            "sll      $12, $12, 16\r\n"
+            "or       $12, $12, $11\r\n"
+
+            "2:       \r\n"
+            "dmtc1    $12, %0\r\n"
+            "li       $8, 4\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 3f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 4\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $9\r\n"
+            "lwu      $14, 0($13)\r\n"
+            "dmtc1    $14, %1\r\n"
+            "dsll32   $12, $12, 0\r\n"
+            "or       $12, $12, $14\r\n"
+            "dmtc1    $12, %0\r\n"
+
+            "3:       \r\n"
+            "li       $8, 8\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 4f\r\n"
+            "nop      \r\n"
+            "mov.s    %1, %0\r\n"
+            "ldc1     %0, 0(%5)\r\n"
+            "li       $9, 8\r\n"
+            "j        5f\r\n"
+            "nop      \r\n"
+
+            "4:       \r\n"
+            "li       $8, 16\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 5f\r\n"
+            "nop      \r\n"
+            "mov.s    %2, %0\r\n"
+            "ldc1     %0, 0(%5)\r\n"
+            "ldc1     %1, 8(%5)\r\n"
+
+            "5:       \r\n"
+            "nop      \r\n"
+            ".set reorder\r\n"
+
+            : "=f" (mmA), "=f" (mmG), "=f" (mmF)
+            : "r" (col), "r" (num_rows), "r" (inptr)
+            : "$f0", "$f2", "$f4", "$8", "$9", "$10", "$11", "$12", "$13",
+              "$14", "memory"
+           );
+      } else {
+        if (!(((long)inptr) & 7)) {
+          mmA = _mm_load_si64((__m64 *)&inptr0);
+          mmG = _mm_load_si64((__m64 *)&inptr8);
+          mmF = _mm_load_si64((__m64 *)&inptr16);
+        } else {
+          mmA = _mm_loadu_si64((__m64 *)&inptr0);
+          mmG = _mm_loadu_si64((__m64 *)&inptr8);
+          mmF = _mm_loadu_si64((__m64 *)&inptr16);
+        }
+        inptr += RGB_PIXELSIZE * 8;
+      }
+      mmD = _mm_srli_si64(mmA, 4 * BYTE_BIT);
+      mmA = _mm_slli_si64(mmA, 4 * BYTE_BIT);
+
+      mmA = _mm_unpackhi_pi8(mmA, mmG);
+      mmG = _mm_slli_si64(mmG, 4 * BYTE_BIT);
+
+      mmD = _mm_unpacklo_pi8(mmD, mmF);
+      mmG = _mm_unpackhi_pi8(mmG, mmF);
+
+      mmE = _mm_srli_si64(mmA, 4 * BYTE_BIT);
+      mmA = _mm_slli_si64(mmA, 4 * BYTE_BIT);
+
+      mmA = _mm_unpackhi_pi8(mmA, mmD);
+      mmD = _mm_slli_si64(mmD, 4 * BYTE_BIT);
+
+      mmE = _mm_unpacklo_pi8(mmE, mmG);
+      mmD = _mm_unpackhi_pi8(mmD, mmG);
+      mmC = _mm_loadhi_pi8_f(mmA);
+      mmA = _mm_loadlo_pi8_f(mmA);
+
+      mmB = _mm_loadhi_pi8_f(mmE);
+      mmE = _mm_loadlo_pi8_f(mmE);
+
+      mmF = _mm_loadhi_pi8_f(mmD);
+      mmD = _mm_loadlo_pi8_f(mmD);
+
+#else  /* RGB_PIXELSIZE == 4 */
+
+      if (num_cols < 8) {
+        col = num_cols;
+        asm(".set noreorder\r\n"
+
+            "li       $8, 1\r\n"
+            "move     $9, %4\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 1f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 1\r\n"
+            PTR_SLL   "$11, $9, 2\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $11\r\n"
+            "lwc1     %0, 0($13)\r\n"
+
+            "1:       \r\n"
+            "li       $8, 2\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 2f\r\n"
+            "nop      \r\n"
+            "subu     $9, $9, 2\r\n"
+            PTR_SLL   "$11, $9, 2\r\n"
+            "move     $13, %5\r\n"
+            PTR_ADDU  "$13, $13, $11\r\n"
+            "mov.s    %1, %0\r\n"
+            "ldc1     %0, 0($13)\r\n"
+
+            "2:       \r\n"
+            "li       $8, 4\r\n"
+            "and      $10, $9, $8\r\n"
+            "beqz     $10, 3f\r\n"
+            "nop      \r\n"
+            "mov.s    %2, %0\r\n"
+            "mov.s    %3, %1\r\n"
+            "ldc1     %0, 0(%5)\r\n"
+            "ldc1     %1, 8(%5)\r\n"
+
+            "3:       \r\n"
+            "nop      \r\n"
+            ".set reorder\r\n"
+
+            : "=f" (mmA), "=f" (mmF), "=f" (mmD), "=f" (mmC)
+            : "r" (col), "r" (inptr)
+            : "$f0", "$f2", "$8", "$9", "$10", "$11", "$13", "memory"
+           );
+      } else {
+        if (!(((long)inptr) & 7)) {
+          mmA = _mm_load_si64((__m64 *)&inptr0);
+          mmF = _mm_load_si64((__m64 *)&inptr8);
+          mmD = _mm_load_si64((__m64 *)&inptr16);
+          mmC = _mm_load_si64((__m64 *)&inptr24);
+        } else {
+          mmA = _mm_loadu_si64((__m64 *)&inptr0);
+          mmF = _mm_loadu_si64((__m64 *)&inptr8);
+          mmD = _mm_loadu_si64((__m64 *)&inptr16);
+          mmC = _mm_loadu_si64((__m64 *)&inptr24);
+        }
+        inptr += RGB_PIXELSIZE * 8;
+      }
+      mmB = _mm_unpackhi_pi8(mmA, mmF);
+      mmA = _mm_unpacklo_pi8(mmA, mmF);
+
+      mmG = _mm_unpackhi_pi8(mmD, mmC);
+      mmD = _mm_unpacklo_pi8(mmD, mmC);
+
+      mmE = _mm_unpackhi_pi16(mmA, mmD);
+      mmA = _mm_unpacklo_pi16(mmA, mmD);
+
+      mmH = _mm_unpackhi_pi16(mmB, mmG);
+      mmB = _mm_unpacklo_pi16(mmB, mmG);
+
+      mmC = _mm_loadhi_pi8_f(mmA);
+      mmA = _mm_loadlo_pi8_f(mmA);
+
+      mmD = _mm_loadhi_pi8_f(mmB);
+      mmB = _mm_loadlo_pi8_f(mmB);
+
+      mmG = _mm_loadhi_pi8_f(mmE);
+      mmE = _mm_loadlo_pi8_f(mmE);
+
+      mmF = _mm_unpacklo_pi8(mmH, mmH);
+      mmH = _mm_unpackhi_pi8(mmH, mmH);
+      mmF = _mm_srli_pi16(mmF, BYTE_BIT);
+      mmH = _mm_srli_pi16(mmH, BYTE_BIT);
+
+#endif
+
+      /* re=(R0 R2 R4 R6), ge=(G0 G2 G4 G6), be=(B0 B2 B4 B6)
+       * ro=(R1 R3 R5 R7), go=(G1 G3 G5 G7), bo=(B1 B3 B5 B7)
+       *
+       * (Original)
+       * Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+       *
+       * (This implementation)
+       * Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+       */
+
+      rglo = _mm_unpacklo_pi16(ro, go);
+      rgho = _mm_unpackhi_pi16(ro, go);
+      ylo_rg = _mm_madd_pi16(rglo, PW_F0299_F0337);
+      yho_rg = _mm_madd_pi16(rgho, PW_F0299_F0337);
+
+      rgle = _mm_unpacklo_pi16(re, ge);
+      rghe = _mm_unpackhi_pi16(re, ge);
+      yle_rg = _mm_madd_pi16(rgle, PW_F0299_F0337);
+      yhe_rg = _mm_madd_pi16(rghe, PW_F0299_F0337);
+
+      bglo = _mm_unpacklo_pi16(bo, go);
+      bgho = _mm_unpackhi_pi16(bo, go);
+      ylo_bg = _mm_madd_pi16(bglo, PW_F0114_F0250);
+      yho_bg = _mm_madd_pi16(bgho, PW_F0114_F0250);
+
+      ylo = _mm_add_pi32(ylo_bg, ylo_rg);
+      yho = _mm_add_pi32(yho_bg, yho_rg);
+      ylo = _mm_add_pi32(ylo, PD_ONEHALF);
+      yho = _mm_add_pi32(yho, PD_ONEHALF);
+      ylo = _mm_srli_pi32(ylo, SCALEBITS);
+      yho = _mm_srli_pi32(yho, SCALEBITS);
+      yo = _mm_packs_pi32(ylo, yho);
+
+      bgle = _mm_unpacklo_pi16(be, ge);
+      bghe = _mm_unpackhi_pi16(be, ge);
+      yle_bg = _mm_madd_pi16(bgle, PW_F0114_F0250);
+      yhe_bg = _mm_madd_pi16(bghe, PW_F0114_F0250);
+
+      yle = _mm_add_pi32(yle_bg, yle_rg);
+      yhe = _mm_add_pi32(yhe_bg, yhe_rg);
+      yle = _mm_add_pi32(yle, PD_ONEHALF);
+      yhe = _mm_add_pi32(yhe, PD_ONEHALF);
+      yle = _mm_srli_pi32(yle, SCALEBITS);
+      yhe = _mm_srli_pi32(yhe, SCALEBITS);
+      ye = _mm_packs_pi32(yle, yhe);
+
+      yo = _mm_slli_pi16(yo, BYTE_BIT);
+      y = _mm_or_si64(ye, yo);
+
+      _mm_store_si64((__m64 *)&outptr0, y);
+    }
+  }
+}
+
+#undef mmA
+#undef mmB
+#undef mmC
+#undef mmD
+#undef mmE
+#undef mmF
+#undef mmG
+#undef mmH

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jcsample-mmi.c Added

@@ -0,0 +1,98 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, 2018-2019, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2017, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* CHROMA DOWNSAMPLING */
+
+#include "jsimd_mmi.h"
+#include "jcsample.h"
+
+
+void jsimd_h2v2_downsample_mmi(JDIMENSION image_width, int max_v_samp_factor,
+                               JDIMENSION v_samp_factor,
+                               JDIMENSION width_in_blocks,
+                               JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow, outcol;
+  JDIMENSION output_cols = width_in_blocks * DCTSIZE;
+  JSAMPROW inptr0, inptr1, outptr;
+  __m64 bias, mask = 0.0, thisavg, nextavg, avg;
+  __m64 this0o, this0e, this0, this0sum, next0o, next0e, next0, next0sum;
+  __m64 this1o, this1e, this1, this1sum, next1o, next1e, next1, next1sum;
+
+  expand_right_edge(input_data, max_v_samp_factor, image_width,
+                    output_cols * 2);
+
+  bias = _mm_set1_pi32((1 << 17) + 1);   /* 0x00020001 (32-bit bias pattern) */
+                                         /* bias={1, 2, 1, 2} (16-bit) */
+  mask = _mm_cmpeq_pi16(mask, mask);
+  mask = _mm_srli_pi16(mask, BYTE_BIT);  /* {0xFF 0x00 0xFF 0x00 ..} */
+
+  for (inrow = 0, outrow = 0; outrow < v_samp_factor;
+       inrow += 2, outrow++) {
+
+    inptr0 = input_datainrow;
+    inptr1 = input_datainrow + 1;
+    outptr = output_dataoutrow;
+
+    for (outcol = output_cols; outcol > 0;
+         outcol -= 8, inptr0 += 16, inptr1 += 16, outptr += 8) {
+
+      this0 = _mm_load_si64((__m64 *)&inptr00);
+      this1 = _mm_load_si64((__m64 *)&inptr10);
+      next0 = _mm_load_si64((__m64 *)&inptr08);
+      next1 = _mm_load_si64((__m64 *)&inptr18);
+
+      this0o = _mm_and_si64(this0, mask);
+      this0e = _mm_srli_pi16(this0, BYTE_BIT);
+      this1o = _mm_and_si64(this1, mask);
+      this1e = _mm_srli_pi16(this1, BYTE_BIT);
+      this0sum = _mm_add_pi16(this0o, this0e);
+      this1sum = _mm_add_pi16(this1o, this1e);
+
+      next0o = _mm_and_si64(next0, mask);
+      next0e = _mm_srli_pi16(next0, BYTE_BIT);
+      next1o = _mm_and_si64(next1, mask);
+      next1e = _mm_srli_pi16(next1, BYTE_BIT);
+      next0sum = _mm_add_pi16(next0o, next0e);
+      next1sum = _mm_add_pi16(next1o, next1e);
+
+      thisavg = _mm_add_pi16(this0sum, this1sum);
+      nextavg = _mm_add_pi16(next0sum, next1sum);
+      thisavg = _mm_add_pi16(thisavg, bias);
+      nextavg = _mm_add_pi16(nextavg, bias);
+      thisavg = _mm_srli_pi16(thisavg, 2);
+      nextavg = _mm_srli_pi16(nextavg, 2);
+
+      avg = _mm_packs_pu16(thisavg, nextavg);
+
+      _mm_store_si64((__m64 *)&outptr0, avg);
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jcsample.h Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jdcolext-mmi.c Added

@@ -0,0 +1,415 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2015, 2019, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdcolor-mmi.c */
+
+
+#if RGB_RED == 0
+#define mmA  re
+#define mmB  ro
+#elif RGB_GREEN == 0
+#define mmA  ge
+#define mmB  go
+#elif RGB_BLUE == 0
+#define mmA  be
+#define mmB  bo
+#else
+#define mmA  xe
+#define mmB  xo
+#endif
+
+#if RGB_RED == 1
+#define mmC  re
+#define mmD  ro
+#elif RGB_GREEN == 1
+#define mmC  ge
+#define mmD  go
+#elif RGB_BLUE == 1
+#define mmC  be
+#define mmD  bo
+#else
+#define mmC  xe
+#define mmD  xo
+#endif
+
+#if RGB_RED == 2
+#define mmE  re
+#define mmF  ro
+#elif RGB_GREEN == 2
+#define mmE  ge
+#define mmF  go
+#elif RGB_BLUE == 2
+#define mmE  be
+#define mmF  bo
+#else
+#define mmE  xe
+#define mmF  xo
+#endif
+
+#if RGB_RED == 3
+#define mmG  re
+#define mmH  ro
+#elif RGB_GREEN == 3
+#define mmG  ge
+#define mmH  go
+#elif RGB_BLUE == 3
+#define mmG  be
+#define mmH  bo
+#else
+#define mmG  xe
+#define mmH  xo
+#endif
+
+
+void jsimd_ycc_rgb_convert_mmi(JDIMENSION out_width, JSAMPIMAGE input_buf,
+                               JDIMENSION input_row, JSAMPARRAY output_buf,
+                               int num_rows)
+{
+  JSAMPROW outptr, inptr0, inptr1, inptr2;
+  int num_cols, col;
+  __m64 ye, yo, y, cbe, cbe2, cbo, cbo2, cb, cre, cre2, cro, cro2, cr;
+  __m64 re, ro, gle, ghe, ge, glo, gho, go, be, bo, xe = 0.0, xo = 0.0;
+  __m64 decenter, mask;
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+
+    for (num_cols = out_width; num_cols > 0; num_cols -= 8,
+         inptr0 += 8, inptr1 += 8, inptr2 += 8) {
+
+      cb = _mm_load_si64((__m64 *)inptr1);
+      cr = _mm_load_si64((__m64 *)inptr2);
+      y = _mm_load_si64((__m64 *)inptr0);
+
+      mask = decenter = 0.0;
+      mask = _mm_cmpeq_pi16(mask, mask);
+      decenter = _mm_cmpeq_pi16(decenter, decenter);
+      mask = _mm_srli_pi16(mask, BYTE_BIT);   /* {0xFF 0x00 0xFF 0x00 ..} */
+      decenter = _mm_slli_pi16(decenter, 7);  /* {0xFF80 0xFF80 0xFF80 0xFF80} */
+
+      cbe = _mm_and_si64(mask, cb);           /* Cb(0246) */
+      cbo = _mm_srli_pi16(cb, BYTE_BIT);      /* Cb(1357) */
+      cre = _mm_and_si64(mask, cr);           /* Cr(0246) */
+      cro = _mm_srli_pi16(cr, BYTE_BIT);      /* Cr(1357) */
+      cbe = _mm_add_pi16(cbe, decenter);
+      cbo = _mm_add_pi16(cbo, decenter);
+      cre = _mm_add_pi16(cre, decenter);
+      cro = _mm_add_pi16(cro, decenter);
+
+      /* (Original)
+       * R = Y                + 1.40200 * Cr
+       * G = Y - 0.34414 * Cb - 0.71414 * Cr
+       * B = Y + 1.77200 * Cb
+       *
+       * (This implementation)
+       * R = Y                + 0.40200 * Cr + Cr
+       * G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+       * B = Y - 0.22800 * Cb + Cb + Cb
+       */
+
+      cbe2 = _mm_add_pi16(cbe, cbe);          /* 2*CbE */
+      cbo2 = _mm_add_pi16(cbo, cbo);          /* 2*CbO */
+      cre2 = _mm_add_pi16(cre, cre);          /* 2*CrE */
+      cro2 = _mm_add_pi16(cro, cro);          /* 2*CrO */
+
+      be = _mm_mulhi_pi16(cbe2, PW_MF0228);   /* (2*CbE * -FIX(0.22800) */
+      bo = _mm_mulhi_pi16(cbo2, PW_MF0228);   /* (2*CbO * -FIX(0.22800) */
+      re = _mm_mulhi_pi16(cre2, PW_F0402);    /* (2*CrE * FIX(0.40200)) */
+      ro = _mm_mulhi_pi16(cro2, PW_F0402);    /* (2*CrO * FIX(0.40200)) */
+
+      be = _mm_add_pi16(be, PW_ONE);
+      bo = _mm_add_pi16(bo, PW_ONE);
+      be = _mm_srai_pi16(be, 1);              /* (CbE * -FIX(0.22800)) */
+      bo = _mm_srai_pi16(bo, 1);              /* (CbO * -FIX(0.22800)) */
+      re = _mm_add_pi16(re, PW_ONE);
+      ro = _mm_add_pi16(ro, PW_ONE);
+      re = _mm_srai_pi16(re, 1);              /* (CrE * FIX(0.40200)) */
+      ro = _mm_srai_pi16(ro, 1);              /* (CrO * FIX(0.40200)) */
+
+      be = _mm_add_pi16(be, cbe);
+      bo = _mm_add_pi16(bo, cbo);
+      be = _mm_add_pi16(be, cbe);             /* (CbE * FIX(1.77200))=(B-Y)E */
+      bo = _mm_add_pi16(bo, cbo);             /* (CbO * FIX(1.77200))=(B-Y)O */
+      re = _mm_add_pi16(re, cre);             /* (CrE * FIX(1.40200))=(R-Y)E */
+      ro = _mm_add_pi16(ro, cro);             /* (CrO * FIX(1.40200))=(R-Y)O */
+
+      gle = _mm_unpacklo_pi16(cbe, cre);
+      ghe = _mm_unpackhi_pi16(cbe, cre);
+      gle = _mm_madd_pi16(gle, PW_MF0344_F0285);
+      ghe = _mm_madd_pi16(ghe, PW_MF0344_F0285);
+      glo = _mm_unpacklo_pi16(cbo, cro);
+      gho = _mm_unpackhi_pi16(cbo, cro);
+      glo = _mm_madd_pi16(glo, PW_MF0344_F0285);
+      gho = _mm_madd_pi16(gho, PW_MF0344_F0285);
+
+      gle = _mm_add_pi32(gle, PD_ONEHALF);
+      ghe = _mm_add_pi32(ghe, PD_ONEHALF);
+      gle = _mm_srai_pi32(gle, SCALEBITS);
+      ghe = _mm_srai_pi32(ghe, SCALEBITS);
+      glo = _mm_add_pi32(glo, PD_ONEHALF);
+      gho = _mm_add_pi32(gho, PD_ONEHALF);
+      glo = _mm_srai_pi32(glo, SCALEBITS);
+      gho = _mm_srai_pi32(gho, SCALEBITS);
+
+      ge = _mm_packs_pi32(gle, ghe);       /* CbE*-FIX(0.344)+CrE*FIX(0.285) */
+      go = _mm_packs_pi32(glo, gho);       /* CbO*-FIX(0.344)+CrO*FIX(0.285) */
+      ge = _mm_sub_pi16(ge, cre);  /* CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E */
+      go = _mm_sub_pi16(go, cro);  /* CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O */
+
+      ye = _mm_and_si64(mask, y);             /* Y(0246) */
+      yo = _mm_srli_pi16(y, BYTE_BIT);        /* Y(1357) */
+
+      re = _mm_add_pi16(re, ye);              /* ((R-Y)E+YE)=(R0 R2 R4 R6) */
+      ro = _mm_add_pi16(ro, yo);              /* ((R-Y)O+YO)=(R1 R3 R5 R7) */
+      re = _mm_packs_pu16(re, re);            /* (R0 R2 R4 R6 ** ** ** **) */
+      ro = _mm_packs_pu16(ro, ro);            /* (R1 R3 R5 R7 ** ** ** **) */
+
+      ge = _mm_add_pi16(ge, ye);              /* ((G-Y)E+YE)=(G0 G2 G4 G6) */
+      go = _mm_add_pi16(go, yo);              /* ((G-Y)O+YO)=(G1 G3 G5 G7) */
+      ge = _mm_packs_pu16(ge, ge);            /* (G0 G2 G4 G6 ** ** ** **) */
+      go = _mm_packs_pu16(go, go);            /* (G1 G3 G5 G7 ** ** ** **) */
+
+      be = _mm_add_pi16(be, ye);              /* (YE+(B-Y)E)=(B0 B2 B4 B6) */
+      bo = _mm_add_pi16(bo, yo);              /* (YO+(B-Y)O)=(B1 B3 B5 B7) */
+      be = _mm_packs_pu16(be, be);            /* (B0 B2 B4 B6 ** ** ** **) */
+      bo = _mm_packs_pu16(bo, bo);            /* (B1 B3 B5 B7 ** ** ** **) */
+
+#if RGB_PIXELSIZE == 3
+
+      /* mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **) */
+      /* mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **) */
+      mmA = _mm_unpacklo_pi8(mmA, mmC);       /* (00 10 02 12 04 14 06 16) */
+      mmE = _mm_unpacklo_pi8(mmE, mmB);       /* (20 01 22 03 24 05 26 07) */
+      mmD = _mm_unpacklo_pi8(mmD, mmF);       /* (11 21 13 23 15 25 17 27) */
+
+      mmH = _mm_srli_si64(mmA, 2 * BYTE_BIT);
+
+      mmG = _mm_unpackhi_pi16(mmA, mmE);      /* (04 14 24 05 06 16 26 07) */
+      mmA = _mm_unpacklo_pi16(mmA, mmE);      /* (00 10 20 01 02 12 22 03) */
+
+      mmE = _mm_srli_si64(mmE, 2 * BYTE_BIT);
+      mmB = _mm_srli_si64(mmD, 2 * BYTE_BIT);  /* (13 23 15 25 17 27 -- --) */
+
+      mmC = _mm_unpackhi_pi16(mmD, mmH);      /* (15 25 06 16 17 27 -- --) */
+      mmD = _mm_unpacklo_pi16(mmD, mmH);      /* (11 21 02 12 13 23 04 14) */
+
+      mmF = _mm_unpackhi_pi16(mmE, mmB);      /* (26 07 17 27 -- -- -- --) */
+      mmE = _mm_unpacklo_pi16(mmE, mmB);      /* (22 03 13 23 24 05 15 25) */
+
+      mmA = _mm_unpacklo_pi32(mmA, mmD);      /* (00 10 20 01 11 21 02 12) */
+      mmE = _mm_unpacklo_pi32(mmE, mmG);      /* (22 03 13 23 04 14 24 05) */
+      mmC = _mm_unpacklo_pi32(mmC, mmF);      /* (15 25 06 16 26 07 17 27) */
+
+      if (num_cols >= 8) {
+        if (!(((long)outptr) & 7)) {
+          _mm_store_si64((__m64 *)outptr, mmA);
+          _mm_store_si64((__m64 *)(outptr + 8), mmE);
+          _mm_store_si64((__m64 *)(outptr + 16), mmC);
+        } else {
+          _mm_storeu_si64((__m64 *)outptr, mmA);
+          _mm_storeu_si64((__m64 *)(outptr + 8), mmE);
+          _mm_storeu_si64((__m64 *)(outptr + 16), mmC);
+        }
+        outptr += RGB_PIXELSIZE * 8;
+      } else {
+        col = num_cols * 3;
+        asm(".set noreorder\r\n"
+
+            "li       $8, 16\r\n"
+            "move     $9, %4\r\n"
+            "mov.s    $f4, %1\r\n"
+            "mov.s    $f6, %3\r\n"
+            "move     $10, %5\r\n"
+            "bltu     $9, $8, 1f\r\n"
+            "nop      \r\n"
+            "gssdlc1  $f4, 7($10)\r\n"
+            "gssdrc1  $f4, 0($10)\r\n"
+            "gssdlc1  $f6, 7+8($10)\r\n"
+            "gssdrc1  $f6, 8($10)\r\n"
+            "mov.s    $f4, %2\r\n"
+            "subu     $9, $9, 16\r\n"
+            PTR_ADDU  "$10, $10, 16\r\n"
+            "b        2f\r\n"
+            "nop      \r\n"
+
+            "1:       \r\n"
+            "li       $8, 8\r\n"              /* st8 */
+            "bltu     $9, $8, 2f\r\n"
+            "nop      \r\n"
+            "gssdlc1  $f4, 7($10)\r\n"
+            "gssdrc1  $f4, 0($10)\r\n"
+            "mov.s    $f4, %3\r\n"
+            "subu     $9, $9, 8\r\n"
+            PTR_ADDU  "$10, $10, 8\r\n"
+
+            "2:       \r\n"
+            "li       $8, 4\r\n"              /* st4 */
+            "mfc1     $11, $f4\r\n"
+            "bltu     $9, $8, 3f\r\n"
+            "nop      \r\n"
+            "swl      $11, 3($10)\r\n"
+            "swr      $11, 0($10)\r\n"
+            "li       $8, 32\r\n"
+            "mtc1     $8, $f6\r\n"
+            "dsrl     $f4, $f4, $f6\r\n"
+            "mfc1     $11, $f4\r\n"
+            "subu     $9, $9, 4\r\n"
+            PTR_ADDU  "$10, $10, 4\r\n"
+
+            "3:       \r\n"
+            "li       $8, 2\r\n"              /* st2 */
+            "bltu     $9, $8, 4f\r\n"
+            "nop      \r\n"
+            "ush      $11, 0($10)\r\n"
+            "srl      $11, 16\r\n"
+            "subu     $9, $9, 2\r\n"
+            PTR_ADDU  "$10, $10, 2\r\n"
+
+            "4:       \r\n"
+            "li       $8, 1\r\n"              /* st1 */
+            "bltu     $9, $8, 5f\r\n"
+            "nop      \r\n"
+            "sb       $11, 0($10)\r\n"
+
+            "5:       \r\n"
+            "nop      \r\n"                   /* end */
+            : "=m" (*outptr)
+            : "f" (mmA), "f" (mmC), "f" (mmE), "r" (col), "r" (outptr)
+            : "$f4", "$f6", "$8", "$9", "$10", "$11", "memory"
+           );
+      }
+
+#else  /* RGB_PIXELSIZE == 4 */
+
+#ifdef RGBX_FILLER_0XFF
+      xe = _mm_cmpeq_pi8(xe, xe);
+      xo = _mm_cmpeq_pi8(xo, xo);
+#else
+      xe = _mm_xor_si64(xe, xe);
+      xo = _mm_xor_si64(xo, xo);
+#endif
+      /* mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **) */
+      /* mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **) */
+      /* mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **) */
+      /* mmG=(30 32 34 36 ** ** ** **), mmH=(31 33 35 37 ** ** ** **) */
+
+      mmA = _mm_unpacklo_pi8(mmA, mmC);       /* (00 10 02 12 04 14 06 16) */
+      mmE = _mm_unpacklo_pi8(mmE, mmG);       /* (20 30 22 32 24 34 26 36) */
+      mmB = _mm_unpacklo_pi8(mmB, mmD);       /* (01 11 03 13 05 15 07 17) */
+      mmF = _mm_unpacklo_pi8(mmF, mmH);       /* (21 31 23 33 25 35 27 37) */
+
+      mmC = _mm_unpackhi_pi16(mmA, mmE);      /* (04 14 24 34 06 16 26 36) */
+      mmA = _mm_unpacklo_pi16(mmA, mmE);      /* (00 10 20 30 02 12 22 32) */
+      mmG = _mm_unpackhi_pi16(mmB, mmF);      /* (05 15 25 35 07 17 27 37) */
+      mmB = _mm_unpacklo_pi16(mmB, mmF);      /* (01 11 21 31 03 13 23 33) */
+
+      mmD = _mm_unpackhi_pi32(mmA, mmB);      /* (02 12 22 32 03 13 23 33) */
+      mmA = _mm_unpacklo_pi32(mmA, mmB);      /* (00 10 20 30 01 11 21 31) */
+      mmH = _mm_unpackhi_pi32(mmC, mmG);      /* (06 16 26 36 07 17 27 37) */
+      mmC = _mm_unpacklo_pi32(mmC, mmG);      /* (04 14 24 34 05 15 25 35) */
+
+      if (num_cols >= 8) {
+        if (!(((long)outptr) & 7)) {
+          _mm_store_si64((__m64 *)outptr, mmA);
+          _mm_store_si64((__m64 *)(outptr + 8), mmD);
+          _mm_store_si64((__m64 *)(outptr + 16), mmC);
+          _mm_store_si64((__m64 *)(outptr + 24), mmH);
+        } else {
+          _mm_storeu_si64((__m64 *)outptr, mmA);
+          _mm_storeu_si64((__m64 *)(outptr + 8), mmD);
+          _mm_storeu_si64((__m64 *)(outptr + 16), mmC);
+          _mm_storeu_si64((__m64 *)(outptr + 24), mmH);
+        }
+        outptr += RGB_PIXELSIZE * 8;
+      } else {
+        col = num_cols;
+        asm(".set noreorder\r\n"              /* st16 */
+
+            "li       $8, 4\r\n"
+            "move     $9, %6\r\n"
+            "move     $10, %7\r\n"
+            "mov.s    $f4, %2\r\n"
+            "mov.s    $f6, %4\r\n"
+            "bltu     $9, $8, 1f\r\n"
+            "nop      \r\n"
+            "gssdlc1  $f4, 7($10)\r\n"
+            "gssdrc1  $f4, 0($10)\r\n"
+            "gssdlc1  $f6, 7+8($10)\r\n"
+            "gssdrc1  $f6, 8($10)\r\n"
+            "mov.s    $f4, %3\r\n"
+            "mov.s    $f6, %5\r\n"
+            "subu     $9, $9, 4\r\n"
+            PTR_ADDU  "$10, $10, 16\r\n"
+
+            "1:       \r\n"
+            "li       $8, 2\r\n"              /* st8 */
+            "bltu     $9, $8, 2f\r\n"
+            "nop      \r\n"
+            "gssdlc1  $f4, 7($10)\r\n"
+            "gssdrc1  $f4, 0($10)\r\n"
+            "mov.s    $f4, $f6\r\n"
+            "subu     $9, $9, 2\r\n"
+            PTR_ADDU  "$10, $10, 8\r\n"
+
+            "2:       \r\n"
+            "li       $8, 1\r\n"              /* st4 */
+            "bltu     $9, $8, 3f\r\n"
+            "nop      \r\n"
+            "gsswlc1  $f4, 3($10)\r\n"
+            "gsswrc1  $f4, 0($10)\r\n"
+
+            "3:       \r\n"
+            "li       %1, 0\r\n"              /* end */
+            : "=m" (*outptr), "=r" (col)
+            : "f" (mmA), "f" (mmC), "f" (mmD), "f" (mmH), "r" (col),
+              "r" (outptr)
+            : "$f4", "$f6", "$8", "$9", "$10", "memory"
+           );
+      }
+
+#endif
+
+    }
+  }
+}
+
+#undef mmA
+#undef mmB
+#undef mmC
+#undef mmD
+#undef mmE
+#undef mmF
+#undef mmG
+#undef mmH

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jdcolor-mmi.c Added

@@ -0,0 +1,139 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2011, 2015, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2017, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* YCC --> RGB CONVERSION */
+
+#include "jsimd_mmi.h"
+
+
+#define F_0_344  ((short)22554)  /* FIX(0.34414) */
+#define F_0_402  ((short)26345)  /* FIX(1.40200) - FIX(1) */
+#define F_0_285  ((short)18734)  /* FIX(1) - FIX(0.71414) */
+#define F_0_228  ((short)14942)  /* FIX(2) - FIX(1.77200) */
+
+enum const_index {
+  index_PW_ONE,
+  index_PW_F0402,
+  index_PW_MF0228,
+  index_PW_MF0344_F0285,
+  index_PD_ONEHALF
+};
+
+static uint64_t const_value = {
+  _uint64_set_pi16(1, 1, 1, 1),
+  _uint64_set_pi16(F_0_402, F_0_402, F_0_402, F_0_402),
+  _uint64_set_pi16(-F_0_228, -F_0_228, -F_0_228, -F_0_228),
+  _uint64_set_pi16(F_0_285, -F_0_344, F_0_285, -F_0_344),
+  _uint64_set_pi32((int)(1 << (SCALEBITS - 1)), (int)(1 << (SCALEBITS - 1)))
+};
+
+#define PW_ONE           get_const_value(index_PW_ONE)
+#define PW_F0402         get_const_value(index_PW_F0402)
+#define PW_MF0228        get_const_value(index_PW_MF0228)
+#define PW_MF0344_F0285  get_const_value(index_PW_MF0344_F0285)
+#define PD_ONEHALF       get_const_value(index_PD_ONEHALF)
+
+#define RGBX_FILLER_0XFF  1
+
+
+#include "jdcolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_ycc_rgb_convert_mmi  jsimd_ycc_extrgb_convert_mmi
+#include "jdcolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_mmi
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define jsimd_ycc_rgb_convert_mmi  jsimd_ycc_extrgbx_convert_mmi
+#include "jdcolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_mmi
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define jsimd_ycc_rgb_convert_mmi  jsimd_ycc_extbgr_convert_mmi
+#include "jdcolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_mmi
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define jsimd_ycc_rgb_convert_mmi  jsimd_ycc_extbgrx_convert_mmi
+#include "jdcolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_mmi
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define jsimd_ycc_rgb_convert_mmi  jsimd_ycc_extxbgr_convert_mmi
+#include "jdcolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_mmi
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define jsimd_ycc_rgb_convert_mmi  jsimd_ycc_extxrgb_convert_mmi
+#include "jdcolext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_ycc_rgb_convert_mmi

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jdmerge-mmi.c Added

@@ -0,0 +1,149 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2011, 2015, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhangLixia <zhanglixia-hf@loongson.cn>
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* YCC --> RGB CONVERSION */
+
+#include "jsimd_mmi.h"
+
+
+#define F_0_344  ((short)22554)  /* FIX(0.34414) */
+#define F_0_402  ((short)26345)  /* FIX(1.40200) - FIX(1) */
+#define F_0_285  ((short)18734)  /* FIX(1) - FIX(0.71414) */
+#define F_0_228  ((short)14942)  /* FIX(2) - FIX(1.77200) */
+
+enum const_index {
+  index_PW_ONE,
+  index_PW_F0402,
+  index_PW_MF0228,
+  index_PW_MF0344_F0285,
+  index_PD_ONEHALF
+};
+
+static uint64_t const_value = {
+  _uint64_set_pi16(1, 1, 1, 1),
+  _uint64_set_pi16(F_0_402, F_0_402, F_0_402, F_0_402),
+  _uint64_set_pi16(-F_0_228, -F_0_228, -F_0_228, -F_0_228),
+  _uint64_set_pi16(F_0_285, -F_0_344, F_0_285, -F_0_344),
+  _uint64_set_pi32((int)(1 << (SCALEBITS - 1)), (int)(1 << (SCALEBITS - 1)))
+};
+
+#define PW_ONE           get_const_value(index_PW_ONE)
+#define PW_F0402         get_const_value(index_PW_F0402)
+#define PW_MF0228        get_const_value(index_PW_MF0228)
+#define PW_MF0344_F0285  get_const_value(index_PW_MF0344_F0285)
+#define PD_ONEHALF       get_const_value(index_PD_ONEHALF)
+
+#define RGBX_FILLER_0XFF  1
+
+
+#include "jdmrgext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED  EXT_RGB_RED
+#define RGB_GREEN  EXT_RGB_GREEN
+#define RGB_BLUE  EXT_RGB_BLUE
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_mmi  jsimd_h2v1_extrgb_merged_upsample_mmi
+#define jsimd_h2v2_merged_upsample_mmi  jsimd_h2v2_extrgb_merged_upsample_mmi
+#include "jdmrgext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_mmi
+#undef jsimd_h2v2_merged_upsample_mmi
+
+#define RGB_RED  EXT_RGBX_RED
+#define RGB_GREEN  EXT_RGBX_GREEN
+#define RGB_BLUE  EXT_RGBX_BLUE
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_mmi  jsimd_h2v1_extrgbx_merged_upsample_mmi
+#define jsimd_h2v2_merged_upsample_mmi  jsimd_h2v2_extrgbx_merged_upsample_mmi
+#include "jdmrgext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_mmi
+#undef jsimd_h2v2_merged_upsample_mmi
+
+#define RGB_RED  EXT_BGR_RED
+#define RGB_GREEN  EXT_BGR_GREEN
+#define RGB_BLUE  EXT_BGR_BLUE
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_mmi  jsimd_h2v1_extbgr_merged_upsample_mmi
+#define jsimd_h2v2_merged_upsample_mmi  jsimd_h2v2_extbgr_merged_upsample_mmi
+#include "jdmrgext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_mmi
+#undef jsimd_h2v2_merged_upsample_mmi
+
+#define RGB_RED  EXT_BGRX_RED
+#define RGB_GREEN  EXT_BGRX_GREEN
+#define RGB_BLUE  EXT_BGRX_BLUE
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_mmi  jsimd_h2v1_extbgrx_merged_upsample_mmi
+#define jsimd_h2v2_merged_upsample_mmi  jsimd_h2v2_extbgrx_merged_upsample_mmi
+#include "jdmrgext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_mmi
+#undef jsimd_h2v2_merged_upsample_mmi
+
+#define RGB_RED  EXT_XBGR_RED
+#define RGB_GREEN  EXT_XBGR_GREEN
+#define RGB_BLUE  EXT_XBGR_BLUE
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_mmi  jsimd_h2v1_extxbgr_merged_upsample_mmi
+#define jsimd_h2v2_merged_upsample_mmi  jsimd_h2v2_extxbgr_merged_upsample_mmi
+#include "jdmrgext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_mmi
+#undef jsimd_h2v2_merged_upsample_mmi
+
+#define RGB_RED  EXT_XRGB_RED
+#define RGB_GREEN  EXT_XRGB_GREEN
+#define RGB_BLUE  EXT_XRGB_BLUE
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_mmi  jsimd_h2v1_extxrgb_merged_upsample_mmi
+#define jsimd_h2v2_merged_upsample_mmi  jsimd_h2v2_extxrgb_merged_upsample_mmi
+#include "jdmrgext-mmi.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef jsimd_h2v1_merged_upsample_mmi
+#undef jsimd_h2v2_merged_upsample_mmi

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jdmrgext-mmi.c Added

@@ -0,0 +1,615 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2015, 2019, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhangLixia <zhanglixia-hf@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdmerge-mmi.c */
+
+
+#if RGB_RED == 0
+#define mmA  re
+#define mmB  ro
+#elif RGB_GREEN == 0
+#define mmA  ge
+#define mmB  go
+#elif RGB_BLUE == 0
+#define mmA  be
+#define mmB  bo
+#else
+#define mmA  xe
+#define mmB  xo
+#endif
+
+#if RGB_RED == 1
+#define mmC  re
+#define mmD  ro
+#elif RGB_GREEN == 1
+#define mmC  ge
+#define mmD  go
+#elif RGB_BLUE == 1
+#define mmC  be
+#define mmD  bo
+#else
+#define mmC  xe
+#define mmD  xo
+#endif
+
+#if RGB_RED == 2
+#define mmE  re
+#define mmF  ro
+#elif RGB_GREEN == 2
+#define mmE  ge
+#define mmF  go
+#elif RGB_BLUE == 2
+#define mmE  be
+#define mmF  bo
+#else
+#define mmE  xe
+#define mmF  xo
+#endif
+
+#if RGB_RED == 3
+#define mmG  re
+#define mmH  ro
+#elif RGB_GREEN == 3
+#define mmG  ge
+#define mmH  go
+#elif RGB_BLUE == 3
+#define mmG  be
+#define mmH  bo
+#else
+#define mmG  xe
+#define mmH  xo
+#endif
+
+
+void jsimd_h2v1_merged_upsample_mmi(JDIMENSION output_width,
+                                    JSAMPIMAGE input_buf,
+                                    JDIMENSION in_row_group_ctr,
+                                    JSAMPARRAY output_buf)
+{
+  JSAMPROW outptr, inptr0, inptr1, inptr2;
+  int num_cols, col;
+  __m64 ythise, ythiso, ythis, ynexte, ynexto, ynext, yl, y;
+  __m64 cbl, cbl2, cbh, cbh2, cb, crl, crl2, crh, crh2, cr;
+  __m64 rle, rlo, rl, rhe, rho, rh, re, ro;
+  __m64 ga, gb, gle, glo, gl, gc, gd, ghe, gho, gh, ge, go;
+  __m64 ble, blo, bl, bhe, bho, bh, be, bo, xe = 0.0, xo = 0.0;
+  __m64 decenter, mask, zero = 0.0;
+#if RGB_PIXELSIZE == 4
+  __m64 mm8, mm9;
+#endif
+
+  inptr0 = input_buf0in_row_group_ctr;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr = output_buf0;
+
+  for (num_cols = output_width >> 1; num_cols > 0; num_cols -= 8,
+       inptr0 += 16, inptr1 += 8, inptr2 += 8) {
+
+    cb = _mm_load_si64((__m64 *)inptr1);
+    cr = _mm_load_si64((__m64 *)inptr2);
+    ythis = _mm_load_si64((__m64 *)inptr0);
+    ynext = _mm_load_si64((__m64 *)inptr0 + 1);
+
+    mask = decenter = 0.0;
+    mask = _mm_cmpeq_pi16(mask, mask);
+    decenter = _mm_cmpeq_pi16(decenter, decenter);
+    mask = _mm_srli_pi16(mask, BYTE_BIT);   /* {0xFF 0x00 0xFF 0x00 ..} */
+    decenter = _mm_slli_pi16(decenter, 7);  /* {0xFF80 0xFF80 0xFF80 0xFF80} */
+
+    cbl = _mm_unpacklo_pi8(cb, zero);         /* Cb(0123) */
+    cbh = _mm_unpackhi_pi8(cb, zero);         /* Cb(4567) */
+    crl = _mm_unpacklo_pi8(cr, zero);         /* Cr(0123) */
+    crh = _mm_unpackhi_pi8(cr, zero);         /* Cr(4567) */
+    cbl = _mm_add_pi16(cbl, decenter);
+    cbh = _mm_add_pi16(cbh, decenter);
+    crl = _mm_add_pi16(crl, decenter);
+    crh = _mm_add_pi16(crh, decenter);
+
+    /* (Original)
+     * R = Y                + 1.40200 * Cr
+     * G = Y - 0.34414 * Cb - 0.71414 * Cr
+     * B = Y + 1.77200 * Cb
+     *
+     * (This implementation)
+     * R = Y                + 0.40200 * Cr + Cr
+     * G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+     * B = Y - 0.22800 * Cb + Cb + Cb
+     */
+
+    cbl2 = _mm_add_pi16(cbl, cbl);            /* 2*CbL */
+    cbh2 = _mm_add_pi16(cbh, cbh);            /* 2*CbH */
+    crl2 = _mm_add_pi16(crl, crl);            /* 2*CrL */
+    crh2 = _mm_add_pi16(crh, crh);            /* 2*CrH */
+
+    bl = _mm_mulhi_pi16(cbl2, PW_MF0228);     /* (2*CbL * -FIX(0.22800) */
+    bh = _mm_mulhi_pi16(cbh2, PW_MF0228);     /* (2*CbH * -FIX(0.22800) */
+    rl = _mm_mulhi_pi16(crl2, PW_F0402);      /* (2*CrL * FIX(0.40200)) */
+    rh = _mm_mulhi_pi16(crh2, PW_F0402);      /* (2*CrH * FIX(0.40200)) */
+
+    bl = _mm_add_pi16(bl, PW_ONE);
+    bh = _mm_add_pi16(bh, PW_ONE);
+    bl = _mm_srai_pi16(bl, 1);                /* (CbL * -FIX(0.22800)) */
+    bh = _mm_srai_pi16(bh, 1);                /* (CbH * -FIX(0.22800)) */
+    rl = _mm_add_pi16(rl, PW_ONE);
+    rh = _mm_add_pi16(rh, PW_ONE);
+    rl = _mm_srai_pi16(rl, 1);                /* (CrL * FIX(0.40200)) */
+    rh = _mm_srai_pi16(rh, 1);                /* (CrH * FIX(0.40200)) */
+
+    bl = _mm_add_pi16(bl, cbl);
+    bh = _mm_add_pi16(bh, cbh);
+    bl = _mm_add_pi16(bl, cbl);               /* (CbL * FIX(1.77200))=(B-Y)L */
+    bh = _mm_add_pi16(bh, cbh);               /* (CbH * FIX(1.77200))=(B-Y)H */
+    rl = _mm_add_pi16(rl, crl);               /* (CrL * FIX(1.40200))=(R-Y)L */
+    rh = _mm_add_pi16(rh, crh);               /* (CrH * FIX(1.40200))=(R-Y)H */
+
+    ga = _mm_unpacklo_pi16(cbl, crl);
+    gb = _mm_unpackhi_pi16(cbl, crl);
+    ga = _mm_madd_pi16(ga, PW_MF0344_F0285);
+    gb = _mm_madd_pi16(gb, PW_MF0344_F0285);
+    gc = _mm_unpacklo_pi16(cbh, crh);
+    gd = _mm_unpackhi_pi16(cbh, crh);
+    gc = _mm_madd_pi16(gc, PW_MF0344_F0285);
+    gd = _mm_madd_pi16(gd, PW_MF0344_F0285);
+
+    ga = _mm_add_pi32(ga, PD_ONEHALF);
+    gb = _mm_add_pi32(gb, PD_ONEHALF);
+    ga = _mm_srai_pi32(ga, SCALEBITS);
+    gb = _mm_srai_pi32(gb, SCALEBITS);
+    gc = _mm_add_pi32(gc, PD_ONEHALF);
+    gd = _mm_add_pi32(gd, PD_ONEHALF);
+    gc = _mm_srai_pi32(gc, SCALEBITS);
+    gd = _mm_srai_pi32(gd, SCALEBITS);
+
+    gl = _mm_packs_pi32(ga, gb);           /* CbL*-FIX(0.344)+CrL*FIX(0.285) */
+    gh = _mm_packs_pi32(gc, gd);           /* CbH*-FIX(0.344)+CrH*FIX(0.285) */
+    gl = _mm_sub_pi16(gl, crl);    /* CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L */
+    gh = _mm_sub_pi16(gh, crh);    /* CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H */
+
+    ythise = _mm_and_si64(mask, ythis);       /* Y(0246) */
+    ythiso = _mm_srli_pi16(ythis, BYTE_BIT);  /* Y(1357) */
+    ynexte = _mm_and_si64(mask, ynext);       /* Y(8ACE) */
+    ynexto = _mm_srli_pi16(ynext, BYTE_BIT);  /* Y(9BDF) */
+
+    rle = _mm_add_pi16(rl, ythise);           /* (R0 R2 R4 R6) */
+    rlo = _mm_add_pi16(rl, ythiso);           /* (R1 R3 R5 R7) */
+    rhe = _mm_add_pi16(rh, ynexte);           /* (R8 RA RC RE) */
+    rho = _mm_add_pi16(rh, ynexto);           /* (R9 RB RD RF) */
+    re = _mm_packs_pu16(rle, rhe);            /* (R0 R2 R4 R6 R8 RA RC RE) */
+    ro = _mm_packs_pu16(rlo, rho);            /* (R1 R3 R5 R7 R9 RB RD RF) */
+
+    gle = _mm_add_pi16(gl, ythise);           /* (G0 G2 G4 G6) */
+    glo = _mm_add_pi16(gl, ythiso);           /* (G1 G3 G5 G7) */
+    ghe = _mm_add_pi16(gh, ynexte);           /* (G8 GA GC GE) */
+    gho = _mm_add_pi16(gh, ynexto);           /* (G9 GB GD GF) */
+    ge = _mm_packs_pu16(gle, ghe);            /* (G0 G2 G4 G6 G8 GA GC GE) */
+    go = _mm_packs_pu16(glo, gho);            /* (G1 G3 G5 G7 G9 GB GD GF) */
+
+    ble = _mm_add_pi16(bl, ythise);           /* (B0 B2 B4 B6) */
+    blo = _mm_add_pi16(bl, ythiso);           /* (B1 B3 B5 B7) */
+    bhe = _mm_add_pi16(bh, ynexte);           /* (B8 BA BC BE) */
+    bho = _mm_add_pi16(bh, ynexto);           /* (B9 BB BD BF) */
+    be = _mm_packs_pu16(ble, bhe);            /* (B0 B2 B4 B6 B8 BA BC BE) */
+    bo = _mm_packs_pu16(blo, bho);            /* (B1 B3 B5 B7 B9 BB BD BF) */
+
+#if RGB_PIXELSIZE == 3
+
+    /* mmA=(00 02 04 06 08 0A 0C 0E), mmB=(01 03 05 07 09 0B 0D 0F) */
+    /* mmC=(10 12 14 16 18 1A 1C 1E), mmD=(11 13 15 17 19 1B 1D 1F) */
+    /* mmE=(20 22 24 26 28 2A 2C 2E), mmF=(21 23 25 27 29 2B 2D 2F) */
+    mmG = _mm_unpacklo_pi8(mmA, mmC);         /* (00 10 02 12 04 14 06 16) */
+    mmA = _mm_unpackhi_pi8(mmA, mmC);         /* (08 18 0A 1A 0C 1C 0E 1E) */
+    mmH = _mm_unpacklo_pi8(mmE, mmB);         /* (20 01 22 03 24 05 26 07) */
+    mmE = _mm_unpackhi_pi8(mmE, mmB);         /* (28 09 2A 0B 2C 0D 2E 0F) */
+    mmC = _mm_unpacklo_pi8(mmD, mmF);         /* (11 21 13 23 15 25 17 27) */
+    mmD = _mm_unpackhi_pi8(mmD, mmF);         /* (19 29 1B 2B 1D 2D 1F 2F) */
+
+    mmB = _mm_unpacklo_pi16(mmG, mmA);        /* (00 10 08 18 02 12 0A 1A) */
+    mmA = _mm_unpackhi_pi16(mmG, mmA);        /* (04 14 0C 1C 06 16 0E 1E) */
+    mmF = _mm_unpacklo_pi16(mmH, mmE);        /* (20 01 28 09 22 03 2A 0B) */
+    mmE = _mm_unpackhi_pi16(mmH, mmE);        /* (24 05 2C 0D 26 07 2E 0F) */
+    mmH = _mm_unpacklo_pi16(mmC, mmD);        /* (11 21 19 29 13 23 1B 2B) */
+    mmG = _mm_unpackhi_pi16(mmC, mmD);        /* (15 25 1D 2D 17 27 1F 2F) */
+
+    mmC = _mm_unpacklo_pi16(mmB, mmF);        /* (00 10 20 01 08 18 28 09) */
+    mmB = _mm_srli_si64(mmB, 4 * BYTE_BIT);
+    mmB = _mm_unpacklo_pi16(mmH, mmB);        /* (11 21 02 12 19 29 0A 1A) */
+    mmD = _mm_unpackhi_pi16(mmF, mmH);        /* (22 03 13 23 2A 0B 1B 2B) */
+    mmF = _mm_unpacklo_pi16(mmA, mmE);        /* (04 14 24 05 0C 1C 2C 0D) */
+    mmA = _mm_srli_si64(mmA, 4 * BYTE_BIT);
+    mmH = _mm_unpacklo_pi16(mmG, mmA);        /* (15 25 06 16 1D 2D 0E 1E) */
+    mmG = _mm_unpackhi_pi16(mmE, mmG);        /* (26 07 17 27 2E 0F 1F 2F) */
+
+    mmA = _mm_unpacklo_pi32(mmC, mmB);        /* (00 10 20 01 11 21 02 12) */
+    mmE = _mm_unpackhi_pi32(mmC, mmB);        /* (08 18 28 09 19 29 0A 1A) */
+    mmB = _mm_unpacklo_pi32(mmD, mmF);        /* (22 03 13 23 04 14 24 05) */
+    mmF = _mm_unpackhi_pi32(mmD, mmF);        /* (2A 0B 1B 2B 0C 1C 2C 0D) */
+    mmC = _mm_unpacklo_pi32(mmH, mmG);        /* (15 25 06 16 26 07 17 27) */
+    mmG = _mm_unpackhi_pi32(mmH, mmG);        /* (1D 2D 0E 1E 2E 0F 1F 2F) */
+
+    if (num_cols >= 8) {
+      if (!(((long)outptr) & 7)) {
+        _mm_store_si64((__m64 *)outptr, mmA);
+        _mm_store_si64((__m64 *)(outptr + 8), mmB);
+        _mm_store_si64((__m64 *)(outptr + 16), mmC);
+        _mm_store_si64((__m64 *)(outptr + 24), mmE);
+        _mm_store_si64((__m64 *)(outptr + 32), mmF);
+        _mm_store_si64((__m64 *)(outptr + 40), mmG);
+      } else {
+        _mm_storeu_si64((__m64 *)outptr, mmA);
+        _mm_storeu_si64((__m64 *)(outptr + 8), mmB);
+        _mm_storeu_si64((__m64 *)(outptr + 16), mmC);
+        _mm_storeu_si64((__m64 *)(outptr + 24), mmE);
+        _mm_storeu_si64((__m64 *)(outptr + 32), mmF);
+        _mm_storeu_si64((__m64 *)(outptr + 40), mmG);
+      }
+      outptr += RGB_PIXELSIZE * 16;
+    } else {
+      if (output_width & 1)
+        col = num_cols * 6 + 3;
+      else
+        col = num_cols * 6;
+
+      asm(".set noreorder\r\n"                /* st24 */
+
+          "li       $8, 24\r\n"
+          "move     $9, %7\r\n"
+          "mov.s    $f4, %1\r\n"
+          "mov.s    $f6, %2\r\n"
+          "mov.s    $f8, %3\r\n"
+          "move     $10, %8\r\n"
+          "bltu     $9, $8, 1f\r\n"
+          "nop      \r\n"
+          "gssdlc1  $f4, 7($10)\r\n"
+          "gssdrc1  $f4, 0($10)\r\n"
+          "gssdlc1  $f6, 7+8($10)\r\n"
+          "gssdrc1  $f6, 8($10)\r\n"
+          "gssdlc1  $f8, 7+16($10)\r\n"
+          "gssdrc1  $f8, 16($10)\r\n"
+          "mov.s    $f4, %4\r\n"
+          "mov.s    $f6, %5\r\n"
+          "mov.s    $f8, %6\r\n"
+          "subu     $9, $9, 24\r\n"
+          PTR_ADDU  "$10, $10, 24\r\n"
+
+          "1:       \r\n"
+          "li       $8, 16\r\n"               /* st16 */
+          "bltu     $9, $8, 2f\r\n"
+          "nop      \r\n"
+          "gssdlc1  $f4, 7($10)\r\n"
+          "gssdrc1  $f4, 0($10)\r\n"
+          "gssdlc1  $f6, 7+8($10)\r\n"
+          "gssdrc1  $f6, 8($10)\r\n"
+          "mov.s    $f4, $f8\r\n"
+          "subu     $9, $9, 16\r\n"
+          PTR_ADDU  "$10, $10, 16\r\n"
+
+          "2:       \r\n"
+          "li       $8,  8\r\n"               /* st8 */
+          "bltu     $9, $8, 3f\r\n"
+          "nop      \r\n"
+          "gssdlc1  $f4, 7($10)\r\n"
+          "gssdrc1  $f4, 0($10)\r\n"
+          "mov.s    $f4, $f6\r\n"
+          "subu     $9, $9, 8\r\n"
+          PTR_ADDU  "$10, $10, 8\r\n"
+
+          "3:       \r\n"
+          "li       $8,  4\r\n"               /* st4 */
+          "mfc1     $11, $f4\r\n"
+          "bltu     $9, $8, 4f\r\n"
+          "nop      \r\n"
+          "swl      $11, 3($10)\r\n"
+          "swr      $11, 0($10)\r\n"
+          "li       $8, 32\r\n"
+          "mtc1     $8, $f6\r\n"
+          "dsrl     $f4, $f4, $f6\r\n"
+          "mfc1     $11, $f4\r\n"
+          "subu     $9, $9, 4\r\n"
+          PTR_ADDU  "$10, $10, 4\r\n"
+
+          "4:       \r\n"
+          "li       $8, 2\r\n"                /* st2 */
+          "bltu     $9, $8, 5f\r\n"
+          "nop      \r\n"
+          "ush      $11, 0($10)\r\n"
+          "srl      $11, 16\r\n"
+          "subu     $9, $9, 2\r\n"
+          PTR_ADDU  "$10, $10, 2\r\n"
+
+          "5:       \r\n"
+          "li       $8, 1\r\n"                /* st1 */
+          "bltu     $9, $8, 6f\r\n"
+          "nop      \r\n"
+          "sb       $11, 0($10)\r\n"
+
+          "6:       \r\n"
+          "nop      \r\n"                     /* end */
+          : "=m" (*outptr)
+          : "f" (mmA), "f" (mmB), "f" (mmC), "f" (mmE), "f" (mmF),
+            "f" (mmG), "r" (col), "r" (outptr)
+          : "$f4", "$f6", "$f8", "$8", "$9", "$10", "$11", "memory"
+         );
+    }
+
+#else  /* RGB_PIXELSIZE == 4 */
+
+#ifdef RGBX_FILLER_0XFF
+    xe = _mm_cmpeq_pi8(xe, xe);
+    xo = _mm_cmpeq_pi8(xo, xo);
+#else
+    xe = _mm_xor_si64(xe, xe);
+    xo = _mm_xor_si64(xo, xo);
+#endif
+    /* mmA=(00 02 04 06 08 0A 0C 0E), mmB=(01 03 05 07 09 0B 0D 0F) */
+    /* mmC=(10 12 14 16 18 1A 1C 1E), mmD=(11 13 15 17 19 1B 1D 1F) */
+    /* mmE=(20 22 24 26 28 2A 2C 2E), mmF=(21 23 25 27 29 2B 2D 2F) */
+    /* mmG=(30 32 34 36 38 3A 3C 3E), mmH=(31 33 35 37 39 3B 3D 3F) */
+
+    mm8 = _mm_unpacklo_pi8(mmA, mmC);         /* (00 10 02 12 04 14 06 16) */
+    mm9 = _mm_unpackhi_pi8(mmA, mmC);         /* (08 18 0A 1A 0C 1C 0E 1E) */
+    mmA = _mm_unpacklo_pi8(mmE, mmG);         /* (20 30 22 32 24 34 26 36) */
+    mmE = _mm_unpackhi_pi8(mmE, mmG);         /* (28 38 2A 3A 2C 3C 2E 3E) */
+
+    mmG = _mm_unpacklo_pi8(mmB, mmD);         /* (01 11 03 13 05 15 07 17) */
+    mmB = _mm_unpackhi_pi8(mmB, mmD);         /* (09 19 0B 1B 0D 1D 0F 1F) */
+    mmD = _mm_unpacklo_pi8(mmF, mmH);         /* (21 31 23 33 25 35 27 37) */
+    mmF = _mm_unpackhi_pi8(mmF, mmH);         /* (29 39 2B 3B 2D 3D 2F 3F) */
+
+    mmH = _mm_unpacklo_pi16(mm8, mmA);        /* (00 10 20 30 02 12 22 32) */
+    mm8 = _mm_unpackhi_pi16(mm8, mmA);        /* (04 14 24 34 06 16 26 36) */
+    mmA = _mm_unpacklo_pi16(mmG, mmD);        /* (01 11 21 31 03 13 23 33) */
+    mmD = _mm_unpackhi_pi16(mmG, mmD);        /* (05 15 25 35 07 17 27 37) */
+
+    mmG = _mm_unpackhi_pi16(mm9, mmE);        /* (0C 1C 2C 3C 0E 1E 2E 3E) */
+    mm9 = _mm_unpacklo_pi16(mm9, mmE);        /* (08 18 28 38 0A 1A 2A 3A) */
+    mmE = _mm_unpacklo_pi16(mmB, mmF);        /* (09 19 29 39 0B 1B 2B 3B) */
+    mmF = _mm_unpackhi_pi16(mmB, mmF);        /* (0D 1D 2D 3D 0F 1F 2F 3F) */
+
+    mmB = _mm_unpackhi_pi32(mmH, mmA);        /* (02 12 22 32 03 13 23 33) */
+    mmA = _mm_unpacklo_pi32(mmH, mmA);        /* (00 10 20 30 01 11 21 31) */
+    mmC = _mm_unpacklo_pi32(mm8, mmD);        /* (04 14 24 34 05 15 25 35) */
+    mmD = _mm_unpackhi_pi32(mm8, mmD);        /* (06 16 26 36 07 17 27 37) */
+
+    mmH = _mm_unpackhi_pi32(mmG, mmF);        /* (0E 1E 2E 3E 0F 1F 2F 3F) */
+    mmG = _mm_unpacklo_pi32(mmG, mmF);        /* (0C 1C 2C 3C 0D 1D 2D 3D) */
+    mmF = _mm_unpackhi_pi32(mm9, mmE);        /* (0A 1A 2A 3A 0B 1B 2B 3B) */
+    mmE = _mm_unpacklo_pi32(mm9, mmE);        /* (08 18 28 38 09 19 29 39) */
+
+    if (num_cols >= 8) {
+      if (!(((long)outptr) & 7)) {
+        _mm_store_si64((__m64 *)outptr, mmA);
+        _mm_store_si64((__m64 *)(outptr + 8), mmB);
+        _mm_store_si64((__m64 *)(outptr + 16), mmC);
+        _mm_store_si64((__m64 *)(outptr + 24), mmD);
+        _mm_store_si64((__m64 *)(outptr + 32), mmE);
+        _mm_store_si64((__m64 *)(outptr + 40), mmF);
+        _mm_store_si64((__m64 *)(outptr + 48), mmG);
+        _mm_store_si64((__m64 *)(outptr + 56), mmH);
+      } else {
+        _mm_storeu_si64((__m64 *)outptr, mmA);
+        _mm_storeu_si64((__m64 *)(outptr + 8), mmB);
+        _mm_storeu_si64((__m64 *)(outptr + 16), mmC);
+        _mm_storeu_si64((__m64 *)(outptr + 24), mmD);
+        _mm_storeu_si64((__m64 *)(outptr + 32), mmE);
+        _mm_storeu_si64((__m64 *)(outptr + 40), mmF);
+        _mm_storeu_si64((__m64 *)(outptr + 48), mmG);
+        _mm_storeu_si64((__m64 *)(outptr + 56), mmH);
+      }
+      outptr += RGB_PIXELSIZE * 16;
+    } else {
+      if (output_width & 1)
+        col = num_cols * 2 + 1;
+      else
+        col = num_cols * 2;
+      asm(".set noreorder\r\n"                /* st32 */
+
+          "li       $8, 8\r\n"
+          "move     $9, %10\r\n"
+          "move     $10, %11\r\n"
+          "mov.s    $f4, %2\r\n"
+          "mov.s    $f6, %3\r\n"
+          "mov.s    $f8, %4\r\n"
+          "mov.s    $f10, %5\r\n"
+          "bltu     $9, $8, 1f\r\n"
+          "nop      \r\n"
+          "gssdlc1  $f4, 7($10)\r\n"
+          "gssdrc1  $f4, 0($10)\r\n"
+          "gssdlc1  $f6, 7+8($10)\r\n"
+          "gssdrc1  $f6, 8($10)\r\n"
+          "gssdlc1  $f8, 7+16($10)\r\n"
+          "gssdrc1  $f8, 16($10)\r\n"
+          "gssdlc1  $f10, 7+24($10)\r\n"
+          "gssdrc1  $f10, 24($10)\r\n"
+          "mov.s    $f4, %6\r\n"
+          "mov.s    $f6, %7\r\n"
+          "mov.s    $f8, %8\r\n"
+          "mov.s    $f10, %9\r\n"
+          "subu     $9, $9, 8\r\n"
+          PTR_ADDU  "$10, $10, 32\r\n"
+
+          "1:       \r\n"
+          "li       $8, 4\r\n"                /* st16 */
+          "bltu     $9, $8, 2f\r\n"
+          "nop      \r\n"
+          "gssdlc1  $f4, 7($10)\r\n"
+          "gssdrc1  $f4, 0($10)\r\n"
+          "gssdlc1  $f6, 7+8($10)\r\n"
+          "gssdrc1  $f6, 8($10)\r\n"
+          "mov.s    $f4, $f8\r\n"
+          "mov.s    $f6, $f10\r\n"
+          "subu     $9, $9, 4\r\n"
+          PTR_ADDU  "$10, $10, 16\r\n"
+
+          "2:       \r\n"
+          "li       $8, 2\r\n"                /* st8 */
+          "bltu     $9, $8, 3f\r\n"
+          "nop      \r\n"
+          "gssdlc1  $f4, 7($10)\r\n"
+          "gssdrc1  $f4, 0($10)\r\n"
+          "mov.s    $f4, $f6\r\n"
+          "subu     $9, $9, 2\r\n"
+          PTR_ADDU  "$10, $10, 8\r\n"
+
+          "3:       \r\n"
+          "li       $8, 1\r\n"                /* st4 */
+          "bltu     $9, $8, 4f\r\n"
+          "nop      \r\n"
+          "gsswlc1  $f4, 3($10)\r\n"
+          "gsswrc1  $f4, 0($10)\r\n"
+
+          "4:       \r\n"
+          "li       %1, 0\r\n"                /* end */
+          : "=m" (*outptr), "=r" (col)
+          : "f" (mmA), "f" (mmB), "f" (mmC), "f" (mmD), "f" (mmE), "f" (mmF),
+            "f" (mmG), "f" (mmH), "r" (col), "r" (outptr)
+          : "$f4", "$f6", "$f8", "$f10", "$8", "$9", "$10", "memory"
+         );
+    }
+
+#endif
+
+  }
+
+  if (!((output_width >> 1) & 7)) {
+    if (output_width & 1) {
+      cb = _mm_load_si64((__m64 *)inptr1);
+      cr = _mm_load_si64((__m64 *)inptr2);
+      y = _mm_load_si64((__m64 *)inptr0);
+
+      decenter = 0.0;
+      decenter = _mm_cmpeq_pi16(decenter, decenter);
+      decenter = _mm_slli_pi16(decenter, 7);  /* {0xFF80 0xFF80 0xFF80 0xFF80} */
+
+      cbl = _mm_unpacklo_pi8(cb, zero);       /* Cb(0123) */
+      crl = _mm_unpacklo_pi8(cr, zero);       /* Cr(0123) */
+      cbl = _mm_add_pi16(cbl, decenter);
+      crl = _mm_add_pi16(crl, decenter);
+
+      cbl2 = _mm_add_pi16(cbl, cbl);          /* 2*CbL */
+      crl2 = _mm_add_pi16(crl, crl);          /* 2*CrL */
+      bl = _mm_mulhi_pi16(cbl2, PW_MF0228);   /* (2*CbL * -FIX(0.22800) */
+      rl = _mm_mulhi_pi16(crl2, PW_F0402);    /* (2*CrL * FIX(0.40200)) */
+
+      bl = _mm_add_pi16(bl, PW_ONE);
+      bl = _mm_srai_pi16(bl, 1);              /* (CbL * -FIX(0.22800)) */
+      rl = _mm_add_pi16(rl, PW_ONE);
+      rl = _mm_srai_pi16(rl, 1);              /* (CrL * FIX(0.40200)) */
+
+      bl = _mm_add_pi16(bl, cbl);
+      bl = _mm_add_pi16(bl, cbl);             /* (CbL * FIX(1.77200))=(B-Y)L */
+      rl = _mm_add_pi16(rl, crl);             /* (CrL * FIX(1.40200))=(R-Y)L */
+
+      gl = _mm_unpacklo_pi16(cbl, crl);
+      gl = _mm_madd_pi16(gl, PW_MF0344_F0285);
+      gl = _mm_add_pi32(gl, PD_ONEHALF);
+      gl = _mm_srai_pi32(gl, SCALEBITS);
+      gl = _mm_packs_pi32(gl, zero);       /* CbL*-FIX(0.344)+CrL*FIX(0.285) */
+      gl = _mm_sub_pi16(gl, crl);  /* CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L */
+
+      yl = _mm_unpacklo_pi8(y, zero);         /* Y(0123) */
+      rl = _mm_add_pi16(rl, yl);              /* (R0 R1 R2 R3) */
+      gl = _mm_add_pi16(gl, yl);              /* (G0 G1 G2 G3) */
+      bl = _mm_add_pi16(bl, yl);              /* (B0 B1 B2 B3) */
+      re = _mm_packs_pu16(rl, rl);
+      ge = _mm_packs_pu16(gl, gl);
+      be = _mm_packs_pu16(bl, bl);
+#if RGB_PIXELSIZE == 3
+      mmA = _mm_unpacklo_pi8(mmA, mmC);
+      mmA = _mm_unpacklo_pi16(mmA, mmE);
+      asm(".set noreorder\r\n"
+
+          "move    $8, %2\r\n"
+          "mov.s   $f4, %1\r\n"
+          "mfc1    $9, $f4\r\n"
+          "ush     $9, 0($8)\r\n"
+          "srl     $9, 16\r\n"
+          "sb      $9, 2($8)\r\n"
+          : "=m" (*outptr)
+          : "f" (mmA), "r" (outptr)
+          : "$f4", "$8", "$9", "memory"
+         );
+#else  /* RGB_PIXELSIZE == 4 */
+
+#ifdef RGBX_FILLER_0XFF
+      xe = _mm_cmpeq_pi8(xe, xe);
+#else
+      xe = _mm_xor_si64(xe, xe);
+#endif
+      mmA = _mm_unpacklo_pi8(mmA, mmC);
+      mmE = _mm_unpacklo_pi8(mmE, mmG);
+      mmA = _mm_unpacklo_pi16(mmA, mmE);
+      asm(".set noreorder\r\n"
+
+          "move    $8, %2\r\n"
+          "mov.s   $f4, %1\r\n"
+          "gsswlc1 $f4, 3($8)\r\n"
+          "gsswrc1 $f4, 0($8)\r\n"
+          : "=m" (*outptr)
+          : "f" (mmA), "r" (outptr)
+          : "$f4", "$8", "memory"
+         );
+#endif
+    }
+  }
+}
+
+
+void jsimd_h2v2_merged_upsample_mmi(JDIMENSION output_width,
+                                    JSAMPIMAGE input_buf,
+                                    JDIMENSION in_row_group_ctr,
+                                    JSAMPARRAY output_buf)
+{
+  JSAMPROW inptr, outptr;
+
+  inptr = input_buf0in_row_group_ctr;
+  outptr = output_buf0;
+
+  input_buf0in_row_group_ctr = input_buf0in_row_group_ctr * 2;
+  jsimd_h2v1_merged_upsample_mmi(output_width, input_buf, in_row_group_ctr,
+                                 output_buf);
+
+  input_buf0in_row_group_ctr = input_buf0in_row_group_ctr * 2 + 1;
+  output_buf0 = output_buf1;
+  jsimd_h2v1_merged_upsample_mmi(output_width, input_buf, in_row_group_ctr,
+                                 output_buf);
+
+  input_buf0in_row_group_ctr = inptr;
+  output_buf0 = outptr;
+}
+
+
+#undef mmA
+#undef mmB
+#undef mmC
+#undef mmD
+#undef mmE
+#undef mmF
+#undef mmG
+#undef mmH

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jdsample-mmi.c Added

@@ -0,0 +1,304 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, 2018-2019, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ *           ZhangLixia  <zhanglixia-hf@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* CHROMA UPSAMPLING */
+
+#include "jsimd_mmi.h"
+
+
+enum const_index {
+  index_PW_ONE,
+  index_PW_TWO,
+  index_PW_THREE,
+  index_PW_SEVEN,
+  index_PW_EIGHT,
+};
+
+static uint64_t const_value = {
+  _uint64_set_pi16(1, 1, 1, 1),
+  _uint64_set_pi16(2, 2, 2, 2),
+  _uint64_set_pi16(3, 3, 3, 3),
+  _uint64_set_pi16(7, 7, 7, 7),
+  _uint64_set_pi16(8, 8, 8, 8),
+};
+
+#define PW_ONE    get_const_value(index_PW_ONE)
+#define PW_TWO    get_const_value(index_PW_TWO)
+#define PW_THREE  get_const_value(index_PW_THREE)
+#define PW_SEVEN  get_const_value(index_PW_SEVEN)
+#define PW_EIGHT  get_const_value(index_PW_EIGHT)
+
+
+#define PROCESS_ROW(row, wkoffset, bias1, bias2, shift) { \
+  __m64 samp123X, samp3XXX, samp1234, sampX012, samp_1012; \
+  __m64 sampXXX4, sampX456, samp3456, samp567X, samp7XXX, samp5678; \
+  __m64 outle, outhe, outlo, outho, outl, outh; \
+  \
+  samp123X = _mm_srli_si64(samp0123, 2 * BYTE_BIT);  /* ( 1 2 3 -) */ \
+  sampXXX4 = _mm_slli_si64(samp4567, (SIZEOF_MMWORD - 2) * BYTE_BIT);  /* ( - - - 4) */ \
+  samp3XXX = _mm_srli_si64(samp0123, (SIZEOF_MMWORD - 2) * BYTE_BIT);  /* ( 3 - - -) */ \
+  sampX456 = _mm_slli_si64(samp4567, 2 * BYTE_BIT);  /* ( - 4 5 6) */ \
+  \
+  samp1234 = _mm_or_si64(samp123X, sampXXX4);  /* ( 1 2 3 4) */ \
+  samp3456 = _mm_or_si64(samp3XXX, sampX456);  /* ( 3 4 5 6) */ \
+  \
+  sampX012 = _mm_slli_si64(samp0123, 2 * BYTE_BIT);  /* ( - 0 1 2) */ \
+  samp567X = _mm_srli_si64(samp4567, 2 * BYTE_BIT);  /* ( 5 6 7 -) */ \
+  samp7XXX = _mm_srli_si64(samp4567, (SIZEOF_MMWORD - 2) * BYTE_BIT);  /* ( 7 - - -) */ \
+  \
+  samp_1012 = _mm_or_si64(sampX012, wkrow);            /* (-1 0 1 2) */ \
+  samp5678 = _mm_or_si64(samp567X, wkrow + wkoffset);  /* ( 5 6 7 8) */ \
+  \
+  wkrow = samp7XXX; \
+  \
+  samp0123 = _mm_mullo_pi16(samp0123, PW_THREE); \
+  samp4567 = _mm_mullo_pi16(samp4567, PW_THREE); \
+  samp_1012 = _mm_add_pi16(samp_1012, bias1); \
+  samp3456 = _mm_add_pi16(samp3456, bias1); \
+  samp1234 = _mm_add_pi16(samp1234, bias2); \
+  samp5678 = _mm_add_pi16(samp5678, bias2); \
+  \
+  outle = _mm_add_pi16(samp_1012, samp0123); \
+  outhe = _mm_add_pi16(samp3456, samp4567); \
+  outle = _mm_srli_pi16(outle, shift);        /* ( 0  2  4  6) */ \
+  outhe = _mm_srli_pi16(outhe, shift);        /* ( 8 10 12 14) */ \
+  outlo = _mm_add_pi16(samp1234, samp0123); \
+  outho = _mm_add_pi16(samp5678, samp4567); \
+  outlo = _mm_srli_pi16(outlo, shift);        /* ( 1  3  5  7) */ \
+  outho = _mm_srli_pi16(outho, shift);        /* ( 9 11 13 15) */ \
+  \
+  outlo = _mm_slli_pi16(outlo, BYTE_BIT); \
+  outho = _mm_slli_pi16(outho, BYTE_BIT); \
+  outl = _mm_or_si64(outle, outlo);           /* ( 0  1  2  3  4  5  6  7) */ \
+  outh = _mm_or_si64(outhe, outho);           /* ( 8  9 10 11 12 13 14 15) */ \
+  \
+  _mm_store_si64((__m64 *)outptr##row, outl); \
+  _mm_store_si64((__m64 *)outptr##row + 1, outh); \
+}
+
+void jsimd_h2v2_fancy_upsample_mmi(int max_v_samp_factor,
+                                   JDIMENSION downsampled_width,
+                                   JSAMPARRAY input_data,
+                                   JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr_1, inptr0, inptr1, outptr0, outptr1;
+  int inrow, outrow, incol, tmp, tmp1;
+  __m64 this_1l, this_1h, this_1, thiscolsum_1l, thiscolsum_1h;
+  __m64 this0l, this0h, this0;
+  __m64 this1l, this1h, this1, thiscolsum1l, thiscolsum1h;
+  __m64 next_1l, next_1h, next_1, nextcolsum_1l, nextcolsum_1h;
+  __m64 next0l, next0h, next0;
+  __m64 next1l, next1h, next1, nextcolsum1l, nextcolsum1h;
+  __m64 mask0 = 0.0, masklast, samp0123, samp4567, wk4, zero = 0.0;
+
+  mask0 = _mm_cmpeq_pi8(mask0, mask0);
+  masklast = _mm_slli_si64(mask0, (SIZEOF_MMWORD - 2) * BYTE_BIT);
+  mask0 = _mm_srli_si64(mask0, (SIZEOF_MMWORD - 2) * BYTE_BIT);
+
+  for (inrow = 0, outrow = 0; outrow < max_v_samp_factor; inrow++) {
+
+    inptr_1 = input_datainrow - 1;
+    inptr0 = input_datainrow;
+    inptr1 = input_datainrow + 1;
+    outptr0 = output_dataoutrow++;
+    outptr1 = output_dataoutrow++;
+
+    if (downsampled_width & 7) {
+      tmp = (downsampled_width - 1) * sizeof(JSAMPLE);
+      tmp1 = downsampled_width * sizeof(JSAMPLE);
+      asm(PTR_ADDU  "$8, %3, %6\r\n"
+          "lb       $9, ($8)\r\n"
+          PTR_ADDU  "$8, %3, %7\r\n"
+          "sb       $9, ($8)\r\n"
+          PTR_ADDU  "$8, %4, %6\r\n"
+          "lb       $9, ($8)\r\n"
+          PTR_ADDU  "$8, %4, %7\r\n"
+          "sb       $9, ($8)\r\n"
+          PTR_ADDU  "$8, %5, %6\r\n"
+          "lb       $9, ($8)\r\n"
+          PTR_ADDU  "$8, %5, %7\r\n"
+          "sb       $9, ($8)\r\n"
+          : "=m" (*inptr_1), "=m" (*inptr0), "=m" (*inptr1)
+          : "r" (inptr_1), "r" (inptr0), "r" (inptr1), "r" (tmp), "r" (tmp1)
+          : "$8", "$9"
+         );
+    }
+
+    /* process the first column block */
+    this0 = _mm_load_si64((__m64 *)inptr0);    /* row 00 */
+    this_1 = _mm_load_si64((__m64 *)inptr_1);  /* row-10 */
+    this1 = _mm_load_si64((__m64 *)inptr1);    /* row 10 */
+
+    this0l = _mm_unpacklo_pi8(this0, zero);    /* row 00( 0 1 2 3) */
+    this0h = _mm_unpackhi_pi8(this0, zero);    /* row 00( 4 5 6 7) */
+    this_1l = _mm_unpacklo_pi8(this_1, zero);  /* row-10( 0 1 2 3) */
+    this_1h = _mm_unpackhi_pi8(this_1, zero);  /* row-10( 4 5 6 7) */
+    this1l = _mm_unpacklo_pi8(this1, zero);    /* row+10( 0 1 2 3) */
+    this1h = _mm_unpackhi_pi8(this1, zero);    /* row+10( 4 5 6 7) */
+
+    this0l = _mm_mullo_pi16(this0l, PW_THREE);
+    this0h = _mm_mullo_pi16(this0h, PW_THREE);
+
+    thiscolsum_1l = _mm_add_pi16(this_1l, this0l);  /* ( 0 1 2 3) */
+    thiscolsum_1h = _mm_add_pi16(this_1h, this0h);  /* ( 4 5 6 7) */
+    thiscolsum1l = _mm_add_pi16(this0l, this1l);    /* ( 0 1 2 3) */
+    thiscolsum1h = _mm_add_pi16(this0h, this1h);    /* ( 4 5 6 7) */
+
+    /* temporarily save the intermediate data */
+    _mm_store_si64((__m64 *)outptr0, thiscolsum_1l);
+    _mm_store_si64((__m64 *)outptr0 + 1, thiscolsum_1h);
+    _mm_store_si64((__m64 *)outptr1, thiscolsum1l);
+    _mm_store_si64((__m64 *)outptr1 + 1, thiscolsum1h);
+
+    wk0 = _mm_and_si64(thiscolsum_1l, mask0);  /* ( 0 - - -) */
+    wk1 = _mm_and_si64(thiscolsum1l, mask0);   /* ( 0 - - -) */
+
+    for (incol = downsampled_width; incol > 0;
+         incol -= 8, inptr_1 += 8, inptr0 += 8, inptr1 += 8,
+         outptr0 += 16, outptr1 += 16) {
+
+      if (incol > 8) {
+        /* process the next column block */
+        next0 = _mm_load_si64((__m64 *)inptr0 + 1);    /* row 01 */
+        next_1 = _mm_load_si64((__m64 *)inptr_1 + 1);  /* row-11 */
+        next1 = _mm_load_si64((__m64 *)inptr1 + 1);    /* row+11 */
+
+        next0l = _mm_unpacklo_pi8(next0, zero);    /* row 01( 0 1 2 3) */
+        next0h = _mm_unpackhi_pi8(next0, zero);    /* row 01( 4 5 6 7) */
+        next_1l = _mm_unpacklo_pi8(next_1, zero);  /* row-11( 0 1 2 3) */
+        next_1h = _mm_unpackhi_pi8(next_1, zero);  /* row-11( 4 5 6 7) */
+        next1l = _mm_unpacklo_pi8(next1, zero);    /* row+11( 0 1 2 3) */
+        next1h = _mm_unpackhi_pi8(next1, zero);    /* row+11( 4 5 6 7) */
+
+        next0l = _mm_mullo_pi16(next0l, PW_THREE);
+        next0h = _mm_mullo_pi16(next0h, PW_THREE);
+
+        nextcolsum_1l = _mm_add_pi16(next_1l, next0l);  /* ( 0 1 2 3) */
+        nextcolsum_1h = _mm_add_pi16(next_1h, next0h);  /* ( 4 5 6 7) */
+        nextcolsum1l = _mm_add_pi16(next0l, next1l);    /* ( 0 1 2 3) */
+        nextcolsum1h = _mm_add_pi16(next0h, next1h);    /* ( 4 5 6 7) */
+
+        /* temporarily save the intermediate data */
+        _mm_store_si64((__m64 *)outptr0 + 2, nextcolsum_1l);
+        _mm_store_si64((__m64 *)outptr0 + 3, nextcolsum_1h);
+        _mm_store_si64((__m64 *)outptr1 + 2, nextcolsum1l);
+        _mm_store_si64((__m64 *)outptr1 + 3, nextcolsum1h);
+
+        wk2 = _mm_slli_si64(nextcolsum_1l, (SIZEOF_MMWORD - 2) * BYTE_BIT);  /* ( - - - 0) */
+        wk3 = _mm_slli_si64(nextcolsum1l, (SIZEOF_MMWORD - 2) * BYTE_BIT);   /* ( - - - 0) */
+      } else {
+        __m64 tmp;
+
+        /* process the last column block */
+        tmp = _mm_load_si64((__m64 *)outptr0 + 1);
+        wk2 = _mm_and_si64(masklast, tmp);        /* ( - - - 7) */
+        tmp = _mm_load_si64((__m64 *)outptr1 + 1);
+        wk3 = _mm_and_si64(masklast, tmp);        /* ( - - - 7) */
+      }
+
+      /* process the upper row */
+      samp0123 = _mm_load_si64((__m64 *)outptr0);      /* ( 0 1 2 3) */ \
+      samp4567 = _mm_load_si64((__m64 *)outptr0 + 1);  /* ( 4 5 6 7) */ \
+      PROCESS_ROW(0, 2, PW_EIGHT, PW_SEVEN, 4)
+
+      /* process the lower row */
+      samp0123 = _mm_load_si64((__m64 *)outptr1);      /* ( 0 1 2 3) */ \
+      samp4567 = _mm_load_si64((__m64 *)outptr1 + 1);  /* ( 4 5 6 7) */ \
+      PROCESS_ROW(1, 2, PW_EIGHT, PW_SEVEN, 4)
+    }
+  }
+}
+
+
+void jsimd_h2v1_fancy_upsample_mmi(int max_v_samp_factor,
+                                   JDIMENSION downsampled_width,
+                                   JSAMPARRAY input_data,
+                                   JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr0, outptr0;
+  int inrow, incol, tmp, tmp1;
+  __m64 thisl, this, nextl, next;
+  __m64 mask0 = 0.0, masklast, samp0123, samp4567, wk2, zero = 0.0;
+
+  mask0 = _mm_cmpeq_pi8(mask0, mask0);
+  masklast = _mm_slli_si64(mask0, (SIZEOF_MMWORD - 2) * BYTE_BIT);
+  mask0 = _mm_srli_si64(mask0, (SIZEOF_MMWORD - 2) * BYTE_BIT);
+
+  for (inrow = 0; inrow < max_v_samp_factor; inrow++) {
+
+    inptr0 = input_datainrow;
+    outptr0 = output_datainrow;
+
+    if (downsampled_width & 7) {
+      tmp = (downsampled_width - 1) * sizeof(JSAMPLE);
+      tmp1 = downsampled_width * sizeof(JSAMPLE);
+      asm(PTR_ADDU  "$8, %1, %2\r\n"
+          "lb       $9, ($8)\r\n"
+          PTR_ADDU  "$8, %1, %3\r\n"
+          "sb       $9, ($8)\r\n"
+          : "=m" (*inptr0)
+          : "r" (inptr0), "r" (tmp), "r" (tmp1)
+          : "$8", "$9"
+         );
+    }
+
+    /* process the first column block */
+    this = _mm_load_si64((__m64 *)inptr0);    /* row 00 */
+    thisl = _mm_unpacklo_pi8(this, zero);     /* row 00( 0 1 2 3) */
+    wk0 = _mm_and_si64(thisl, mask0);       /* ( 0 - - -) */
+
+    for (incol = downsampled_width; incol > 0;
+         incol -= 8, inptr0 += 8, outptr0 += 16) {
+
+      if (incol > 8) {
+        /* process the next column block */
+        next = _mm_load_si64((__m64 *)inptr0 + 1);  /* row 01 */
+        nextl = _mm_unpacklo_pi8(next, zero);       /* row 01( 0 1 2 3) */
+        wk1 = _mm_slli_si64(nextl, (SIZEOF_MMWORD - 2) * BYTE_BIT);  /* ( - - - 0) */
+      } else {
+        __m64 thish;
+
+        /* process the last column block */
+        this = _mm_load_si64((__m64 *)inptr0);  /* row 00 */
+        thish = _mm_unpackhi_pi8(this, zero);   /* row 01( 4 5 6 7) */
+        wk1 = _mm_and_si64(masklast, thish);  /* ( - - - 7) */
+      }
+
+      /* process the row */
+      this = _mm_load_si64((__m64 *)inptr0);    /* row 00 */
+      samp0123 = _mm_unpacklo_pi8(this, zero);  /* ( 0 1 2 3) */
+      samp4567 = _mm_unpackhi_pi8(this, zero);  /* ( 4 5 6 7) */
+      PROCESS_ROW(0, 1, PW_ONE, PW_TWO, 2)
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jfdctfst-mmi.c Added

@@ -0,0 +1,255 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, 2018-2019, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  LiuQingfa <liuqingfa-hf@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* FAST INTEGER FORWARD DCT */
+
+#include "jsimd_mmi.h"
+
+
+#define CONST_BITS  8
+
+#define F_0_382  ((short)98)   /* FIX(0.382683433) */
+#define F_0_541  ((short)139)  /* FIX(0.541196100) */
+#define F_0_707  ((short)181)  /* FIX(0.707106781) */
+#define F_1_306  ((short)334)  /* FIX(1.306562965) */
+
+#define PRE_MULTIPLY_SCALE_BITS  2
+#define CONST_SHIFT  (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+enum const_index {
+  index_PW_F0707,
+  index_PW_F0382,
+  index_PW_F0541,
+  index_PW_F1306
+};
+
+static uint64_t const_value = {
+  _uint64_set1_pi16(F_0_707),
+  _uint64_set1_pi16(F_0_382),
+  _uint64_set1_pi16(F_0_541),
+  _uint64_set1_pi16(F_1_306)
+};
+
+#define PW_F0707  get_const_value(index_PW_F0707)
+#define PW_F0382  get_const_value(index_PW_F0382)
+#define PW_F0541  get_const_value(index_PW_F0541)
+#define PW_F1306  get_const_value(index_PW_F1306)
+
+
+#define DO_FDCT_MULTIPLY(out, in, multiplier) { \
+  __m64 mulhi, mullo, mul12, mul34; \
+  \
+  mullo = _mm_mullo_pi16(in, multiplier); \
+  mulhi = _mm_mulhi_pi16(in, multiplier); \
+  mul12 = _mm_unpacklo_pi16(mullo, mulhi); \
+  mul34 = _mm_unpackhi_pi16(mullo, mulhi); \
+  mul12 = _mm_srai_pi32(mul12, CONST_BITS); \
+  mul34 = _mm_srai_pi32(mul34, CONST_BITS); \
+  out = _mm_packs_pi32(mul12, mul34); \
+}
+
+#define DO_FDCT_COMMON() { \
+  \
+  /* Even part */ \
+  \
+  tmp10 = _mm_add_pi16(tmp0, tmp3); \
+  tmp13 = _mm_sub_pi16(tmp0, tmp3); \
+  tmp11 = _mm_add_pi16(tmp1, tmp2); \
+  tmp12 = _mm_sub_pi16(tmp1, tmp2); \
+  \
+  out0 = _mm_add_pi16(tmp10, tmp11); \
+  out4 = _mm_sub_pi16(tmp10, tmp11); \
+  \
+  z1 = _mm_add_pi16(tmp12, tmp13); \
+  DO_FDCT_MULTIPLY(z1, z1, PW_F0707) \
+  \
+  out2 = _mm_add_pi16(tmp13, z1); \
+  out6 = _mm_sub_pi16(tmp13, z1); \
+  \
+  /* Odd part */ \
+  \
+  tmp10 = _mm_add_pi16(tmp4, tmp5); \
+  tmp11 = _mm_add_pi16(tmp5, tmp6); \
+  tmp12 = _mm_add_pi16(tmp6, tmp7); \
+  \
+  z5 = _mm_sub_pi16(tmp10, tmp12); \
+  DO_FDCT_MULTIPLY(z5, z5, PW_F0382) \
+  \
+  DO_FDCT_MULTIPLY(z2, tmp10, PW_F0541) \
+  z2 = _mm_add_pi16(z2, z5); \
+  \
+  DO_FDCT_MULTIPLY(z4, tmp12, PW_F1306) \
+  z4 = _mm_add_pi16(z4, z5); \
+  \
+  DO_FDCT_MULTIPLY(z3, tmp11, PW_F0707) \
+  \
+  z11 = _mm_add_pi16(tmp7, z3); \
+  z13 = _mm_sub_pi16(tmp7, z3); \
+  \
+  out5 = _mm_add_pi16(z13, z2); \
+  out3 = _mm_sub_pi16(z13, z2); \
+  out1 = _mm_add_pi16(z11, z4); \
+  out7 = _mm_sub_pi16(z11, z4); \
+}
+
+#define DO_FDCT_PASS1() { \
+  __m64 row0l, row0h, row1l, row1h, row2l, row2h, row3l, row3h; \
+  __m64 row01a, row01b, row01c, row01d, row23a, row23b, row23c, row23d; \
+  __m64 col0, col1, col2, col3, col4, col5, col6, col7; \
+  \
+  row0l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 0);     /* (00 01 02 03) */ \
+  row0h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 0 + 4); /* (04 05 06 07) */ \
+  row1l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 1);     /* (10 11 12 13) */ \
+  row1h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 1 + 4); /* (14 15 16 17) */ \
+  row2l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 2);     /* (20 21 22 23) */ \
+  row2h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 2 + 4); /* (24 25 26 27) */ \
+  row3l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 3);     /* (30 31 32 33) */ \
+  row3h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 3 + 4); /* (34 35 36 37) */ \
+  \
+  /* Transpose coefficients */ \
+  \
+  row23a = _mm_unpacklo_pi16(row2l, row3l);   /* row23a=(20 30 21 31) */ \
+  row23b = _mm_unpackhi_pi16(row2l, row3l);   /* row23b=(22 32 23 33) */ \
+  row23c = _mm_unpacklo_pi16(row2h, row3h);   /* row23c=(24 34 25 35) */ \
+  row23d = _mm_unpackhi_pi16(row2h, row3h);   /* row23d=(26 36 27 37) */ \
+  \
+  row01a = _mm_unpacklo_pi16(row0l, row1l);   /* row01a=(00 10 01 11) */ \
+  row01b = _mm_unpackhi_pi16(row0l, row1l);   /* row01b=(02 12 03 13) */ \
+  row01c = _mm_unpacklo_pi16(row0h, row1h);   /* row01c=(04 14 05 15) */ \
+  row01d = _mm_unpackhi_pi16(row0h, row1h);   /* row01d=(06 16 07 17) */ \
+  \
+  col0 = _mm_unpacklo_pi32(row01a, row23a);   /* col0=(00 10 20 30) */ \
+  col1 = _mm_unpackhi_pi32(row01a, row23a);   /* col1=(01 11 21 31) */ \
+  col6 = _mm_unpacklo_pi32(row01d, row23d);   /* col6=(06 16 26 36) */ \
+  col7 = _mm_unpackhi_pi32(row01d, row23d);   /* col7=(07 17 27 37) */ \
+  \
+  tmp6 = _mm_sub_pi16(col1, col6);            /* tmp6=col1-col6 */ \
+  tmp7 = _mm_sub_pi16(col0, col7);            /* tmp7=col0-col7 */ \
+  tmp1 = _mm_add_pi16(col1, col6);            /* tmp1=col1+col6 */ \
+  tmp0 = _mm_add_pi16(col0, col7);            /* tmp0=col0+col7 */ \
+  \
+  col2 = _mm_unpacklo_pi32(row01b, row23b);   /* col2=(02 12 22 32) */ \
+  col3 = _mm_unpackhi_pi32(row01b, row23b);   /* col3=(03 13 23 33) */ \
+  col4 = _mm_unpacklo_pi32(row01c, row23c);   /* col4=(04 14 24 34) */ \
+  col5 = _mm_unpackhi_pi32(row01c, row23c);   /* col5=(05 15 25 35) */ \
+  \
+  tmp3 = _mm_add_pi16(col3, col4);            /* tmp3=col3+col4 */ \
+  tmp2 = _mm_add_pi16(col2, col5);            /* tmp2=col2+col5 */ \
+  tmp4 = _mm_sub_pi16(col3, col4);            /* tmp4=col3-col4 */ \
+  tmp5 = _mm_sub_pi16(col2, col5);            /* tmp5=col2-col5 */ \
+  \
+  DO_FDCT_COMMON() \
+  \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 0, out0); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 0 + 4, out4); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 1, out1); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 1 + 4, out5); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 2, out2); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 2 + 4, out6); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 3, out3); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 3 + 4, out7); \
+}
+
+#define DO_FDCT_PASS2() { \
+  __m64 col0l, col0h, col1l, col1h, col2l, col2h, col3l, col3h; \
+  __m64 col01a, col01b, col01c, col01d, col23a, col23b, col23c, col23d; \
+  __m64 row0, row1, row2, row3, row4, row5, row6, row7; \
+  \
+  col0l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 0);  /* (00 10 20 30) */ \
+  col1l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 1);  /* (01 11 21 31) */ \
+  col2l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 2);  /* (02 12 22 32) */ \
+  col3l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 3);  /* (03 13 23 33) */ \
+  col0h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 4);  /* (40 50 60 70) */ \
+  col1h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 5);  /* (41 51 61 71) */ \
+  col2h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 6);  /* (42 52 62 72) */ \
+  col3h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 7);  /* (43 53 63 73) */ \
+  \
+  /* Transpose coefficients */ \
+  \
+  col23a = _mm_unpacklo_pi16(col2l, col3l);   /* col23a=(02 03 12 13) */ \
+  col23b = _mm_unpackhi_pi16(col2l, col3l);   /* col23b=(22 23 32 33) */ \
+  col23c = _mm_unpacklo_pi16(col2h, col3h);   /* col23c=(42 43 52 53) */ \
+  col23d = _mm_unpackhi_pi16(col2h, col3h);   /* col23d=(62 63 72 73) */ \
+  \
+  col01a = _mm_unpacklo_pi16(col0l, col1l);   /* col01a=(00 01 10 11) */ \
+  col01b = _mm_unpackhi_pi16(col0l, col1l);   /* col01b=(20 21 30 31) */ \
+  col01c = _mm_unpacklo_pi16(col0h, col1h);   /* col01c=(40 41 50 51) */ \
+  col01d = _mm_unpackhi_pi16(col0h, col1h);   /* col01d=(60 61 70 71) */ \
+  \
+  row0 = _mm_unpacklo_pi32(col01a, col23a);   /* row0=(00 01 02 03) */ \
+  row1 = _mm_unpackhi_pi32(col01a, col23a);   /* row1=(10 11 12 13) */ \
+  row6 = _mm_unpacklo_pi32(col01d, col23d);   /* row6=(60 61 62 63) */ \
+  row7 = _mm_unpackhi_pi32(col01d, col23d);   /* row7=(70 71 72 73) */ \
+  \
+  tmp6 = _mm_sub_pi16(row1, row6);            /* tmp6=row1-row6 */ \
+  tmp7 = _mm_sub_pi16(row0, row7);            /* tmp7=row0-row7 */ \
+  tmp1 = _mm_add_pi16(row1, row6);            /* tmp1=row1+row6 */ \
+  tmp0 = _mm_add_pi16(row0, row7);            /* tmp0=row0+row7 */ \
+  \
+  row2 = _mm_unpacklo_pi32(col01b, col23b);   /* row2=(20 21 22 23) */ \
+  row3 = _mm_unpackhi_pi32(col01b, col23b);   /* row3=(30 31 32 33) */ \
+  row4 = _mm_unpacklo_pi32(col01c, col23c);   /* row4=(40 41 42 43) */ \
+  row5 = _mm_unpackhi_pi32(col01c, col23c);   /* row5=(50 51 52 53) */ \
+  \
+  tmp3 = _mm_add_pi16(row3, row4);            /* tmp3=row3+row4 */ \
+  tmp2 = _mm_add_pi16(row2, row5);            /* tmp2=row2+row5 */ \
+  tmp4 = _mm_sub_pi16(row3, row4);            /* tmp4=row3-row4 */ \
+  tmp5 = _mm_sub_pi16(row2, row5);            /* tmp5=row2-row5 */ \
+  \
+  DO_FDCT_COMMON() \
+  \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 0, out0); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 1, out1); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 2, out2); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 3, out3); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 4, out4); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 5, out5); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 6, out6); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 7, out7); \
+}
+
+void jsimd_fdct_ifast_mmi(DCTELEM *data)
+{
+  __m64 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  __m64 out0, out1, out2, out3, out4, out5, out6, out7;
+  __m64 tmp10, tmp11, tmp12, tmp13, z1, z2, z3, z4, z5, z11, z13;
+  DCTELEM *dataptr = data;
+
+  /* Pass 1: process rows. */
+
+  DO_FDCT_PASS1()
+  dataptr += DCTSIZE * 4;
+  DO_FDCT_PASS1()
+
+  /* Pass 2: process columns. */
+
+  dataptr = data;
+  DO_FDCT_PASS2()
+  dataptr += 4;
+  DO_FDCT_PASS2()
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jfdctint-mmi.c Added

@@ -0,0 +1,398 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, 2018, 2020, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2017, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* ACCURATE INTEGER FORWARD DCT */
+
+#include "jsimd_mmi.h"
+
+
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2  (CONST_BITS + PASS1_BITS)
+
+#define FIX_0_298  ((short)2446)   /* FIX(0.298631336) */
+#define FIX_0_390  ((short)3196)   /* FIX(0.390180644) */
+#define FIX_0_541  ((short)4433)   /* FIX(0.541196100) */
+#define FIX_0_765  ((short)6270)   /* FIX(0.765366865) */
+#define FIX_0_899  ((short)7373)   /* FIX(0.899976223) */
+#define FIX_1_175  ((short)9633)   /* FIX(1.175875602) */
+#define FIX_1_501  ((short)12299)  /* FIX(1.501321110) */
+#define FIX_1_847  ((short)15137)  /* FIX(1.847759065) */
+#define FIX_1_961  ((short)16069)  /* FIX(1.961570560) */
+#define FIX_2_053  ((short)16819)  /* FIX(2.053119869) */
+#define FIX_2_562  ((short)20995)  /* FIX(2.562915447) */
+#define FIX_3_072  ((short)25172)  /* FIX(3.072711026) */
+
+enum const_index {
+  index_PW_F130_F054,
+  index_PW_F054_MF130,
+  index_PW_MF078_F117,
+  index_PW_F117_F078,
+  index_PW_MF060_MF089,
+  index_PW_MF089_F060,
+  index_PW_MF050_MF256,
+  index_PW_MF256_F050,
+  index_PD_DESCALE_P1,
+  index_PD_DESCALE_P2,
+  index_PW_DESCALE_P2X
+};
+
+static uint64_t const_value = {
+  _uint64_set_pi16(FIX_0_541, (FIX_0_541 + FIX_0_765),
+                   FIX_0_541, (FIX_0_541 + FIX_0_765)),
+  _uint64_set_pi16((FIX_0_541 - FIX_1_847), FIX_0_541,
+                   (FIX_0_541 - FIX_1_847), FIX_0_541),
+  _uint64_set_pi16(FIX_1_175, (FIX_1_175 - FIX_1_961),
+                   FIX_1_175, (FIX_1_175 - FIX_1_961)),
+  _uint64_set_pi16((FIX_1_175 - FIX_0_390), FIX_1_175,
+                   (FIX_1_175 - FIX_0_390), FIX_1_175),
+  _uint64_set_pi16(-FIX_0_899, (FIX_0_298 - FIX_0_899),
+                   -FIX_0_899, (FIX_0_298 - FIX_0_899)),
+  _uint64_set_pi16((FIX_1_501 - FIX_0_899), -FIX_0_899,
+                   (FIX_1_501 - FIX_0_899), -FIX_0_899),
+  _uint64_set_pi16(-FIX_2_562, (FIX_2_053 - FIX_2_562),
+                   -FIX_2_562, (FIX_2_053 - FIX_2_562)),
+  _uint64_set_pi16((FIX_3_072 - FIX_2_562), -FIX_2_562,
+                   (FIX_3_072 - FIX_2_562), -FIX_2_562),
+  _uint64_set_pi32((1 << (DESCALE_P1 - 1)), (1 << (DESCALE_P1 - 1))),
+  _uint64_set_pi32((1 << (DESCALE_P2 - 1)), (1 << (DESCALE_P2 - 1))),
+  _uint64_set_pi16((1 << (PASS1_BITS - 1)), (1 << (PASS1_BITS - 1)),
+                   (1 << (PASS1_BITS - 1)), (1 << (PASS1_BITS - 1)))
+};
+
+#define PW_F130_F054    get_const_value(index_PW_F130_F054)
+#define PW_F054_MF130   get_const_value(index_PW_F054_MF130)
+#define PW_MF078_F117   get_const_value(index_PW_MF078_F117)
+#define PW_F117_F078    get_const_value(index_PW_F117_F078)
+#define PW_MF060_MF089  get_const_value(index_PW_MF060_MF089)
+#define PW_MF089_F060   get_const_value(index_PW_MF089_F060)
+#define PW_MF050_MF256  get_const_value(index_PW_MF050_MF256)
+#define PW_MF256_F050   get_const_value(index_PW_MF256_F050)
+#define PD_DESCALE_P1   get_const_value(index_PD_DESCALE_P1)
+#define PD_DESCALE_P2   get_const_value(index_PD_DESCALE_P2)
+#define PW_DESCALE_P2X  get_const_value(index_PW_DESCALE_P2X)
+
+
+#define DO_FDCT_COMMON(PASS) { \
+  __m64 tmp1312l, tmp1312h, tmp47l, tmp47h, tmp4l, tmp4h, tmp7l, tmp7h; \
+  __m64 tmp56l, tmp56h, tmp5l, tmp5h, tmp6l, tmp6h; \
+  __m64 out1l, out1h, out2l, out2h, out3l, out3h; \
+  __m64 out5l, out5h, out6l, out6h, out7l, out7h; \
+  __m64 z34l, z34h, z3l, z3h, z4l, z4h, z3, z4; \
+  \
+  /* (Original) \
+   * z1 = (tmp12 + tmp13) * 0.541196100; \
+   * out2 = z1 + tmp13 * 0.765366865; \
+   * out6 = z1 + tmp12 * -1.847759065; \
+   * \
+   * (This implementation) \
+   * out2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; \
+   * out6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); \
+   */ \
+  \
+  tmp1312l = _mm_unpacklo_pi16(tmp13, tmp12); \
+  tmp1312h = _mm_unpackhi_pi16(tmp13, tmp12); \
+  \
+  out2l = _mm_madd_pi16(tmp1312l, PW_F130_F054); \
+  out2h = _mm_madd_pi16(tmp1312h, PW_F130_F054); \
+  out6l = _mm_madd_pi16(tmp1312l, PW_F054_MF130); \
+  out6h = _mm_madd_pi16(tmp1312h, PW_F054_MF130); \
+  \
+  out2l = _mm_add_pi32(out2l, PD_DESCALE_P##PASS); \
+  out2h = _mm_add_pi32(out2h, PD_DESCALE_P##PASS); \
+  out2l = _mm_srai_pi32(out2l, DESCALE_P##PASS); \
+  out2h = _mm_srai_pi32(out2h, DESCALE_P##PASS); \
+  \
+  out6l = _mm_add_pi32(out6l, PD_DESCALE_P##PASS); \
+  out6h = _mm_add_pi32(out6h, PD_DESCALE_P##PASS); \
+  out6l = _mm_srai_pi32(out6l, DESCALE_P##PASS); \
+  out6h = _mm_srai_pi32(out6h, DESCALE_P##PASS); \
+  \
+  out2 = _mm_packs_pi32(out2l, out2h); \
+  out6 = _mm_packs_pi32(out6l, out6h); \
+  \
+  /* Odd part */ \
+  \
+  z3 = _mm_add_pi16(tmp4, tmp6); \
+  z4 = _mm_add_pi16(tmp5, tmp7); \
+  \
+  /* (Original) \
+   * z5 = (z3 + z4) * 1.175875602; \
+   * z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644; \
+   * z3 += z5;  z4 += z5; \
+   * \
+   * (This implementation) \
+   * z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; \
+   * z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); \
+   */ \
+  \
+  z34l = _mm_unpacklo_pi16(z3, z4); \
+  z34h = _mm_unpackhi_pi16(z3, z4); \
+  z3l = _mm_madd_pi16(z34l, PW_MF078_F117); \
+  z3h = _mm_madd_pi16(z34h, PW_MF078_F117); \
+  z4l = _mm_madd_pi16(z34l, PW_F117_F078); \
+  z4h = _mm_madd_pi16(z34h, PW_F117_F078); \
+  \
+  /* (Original) \
+   * z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6; \
+   * tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869; \
+   * tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110; \
+   * z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447; \
+   * out7 = tmp4 + z1 + z3;  out5 = tmp5 + z2 + z4; \
+   * out3 = tmp6 + z2 + z3;  out1 = tmp7 + z1 + z4; \
+   * \
+   * (This implementation) \
+   * tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; \
+   * tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; \
+   * tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); \
+   * tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); \
+   * out7 = tmp4 + z3;  out5 = tmp5 + z4; \
+   * out3 = tmp6 + z3;  out1 = tmp7 + z4; \
+   */ \
+  \
+  tmp47l = _mm_unpacklo_pi16(tmp4, tmp7); \
+  tmp47h = _mm_unpackhi_pi16(tmp4, tmp7); \
+  \
+  tmp4l = _mm_madd_pi16(tmp47l, PW_MF060_MF089); \
+  tmp4h = _mm_madd_pi16(tmp47h, PW_MF060_MF089); \
+  tmp7l = _mm_madd_pi16(tmp47l, PW_MF089_F060); \
+  tmp7h = _mm_madd_pi16(tmp47h, PW_MF089_F060); \
+  \
+  out7l = _mm_add_pi32(tmp4l, z3l); \
+  out7h = _mm_add_pi32(tmp4h, z3h); \
+  out1l = _mm_add_pi32(tmp7l, z4l); \
+  out1h = _mm_add_pi32(tmp7h, z4h); \
+  \
+  out7l = _mm_add_pi32(out7l, PD_DESCALE_P##PASS); \
+  out7h = _mm_add_pi32(out7h, PD_DESCALE_P##PASS); \
+  out7l = _mm_srai_pi32(out7l, DESCALE_P##PASS); \
+  out7h = _mm_srai_pi32(out7h, DESCALE_P##PASS); \
+  \
+  out1l = _mm_add_pi32(out1l, PD_DESCALE_P##PASS); \
+  out1h = _mm_add_pi32(out1h, PD_DESCALE_P##PASS); \
+  out1l = _mm_srai_pi32(out1l, DESCALE_P##PASS); \
+  out1h = _mm_srai_pi32(out1h, DESCALE_P##PASS); \
+  \
+  out7 = _mm_packs_pi32(out7l, out7h); \
+  out1 = _mm_packs_pi32(out1l, out1h); \
+  \
+  tmp56l = _mm_unpacklo_pi16(tmp5, tmp6); \
+  tmp56h = _mm_unpackhi_pi16(tmp5, tmp6); \
+  \
+  tmp5l = _mm_madd_pi16(tmp56l, PW_MF050_MF256); \
+  tmp5h = _mm_madd_pi16(tmp56h, PW_MF050_MF256); \
+  tmp6l = _mm_madd_pi16(tmp56l, PW_MF256_F050); \
+  tmp6h = _mm_madd_pi16(tmp56h, PW_MF256_F050); \
+  \
+  out5l = _mm_add_pi32(tmp5l, z4l); \
+  out5h = _mm_add_pi32(tmp5h, z4h); \
+  out3l = _mm_add_pi32(tmp6l, z3l); \
+  out3h = _mm_add_pi32(tmp6h, z3h); \
+  \
+  out5l = _mm_add_pi32(out5l, PD_DESCALE_P##PASS); \
+  out5h = _mm_add_pi32(out5h, PD_DESCALE_P##PASS); \
+  out5l = _mm_srai_pi32(out5l, DESCALE_P##PASS); \
+  out5h = _mm_srai_pi32(out5h, DESCALE_P##PASS); \
+  \
+  out3l = _mm_add_pi32(out3l, PD_DESCALE_P##PASS); \
+  out3h = _mm_add_pi32(out3h, PD_DESCALE_P##PASS); \
+  out3l = _mm_srai_pi32(out3l, DESCALE_P##PASS); \
+  out3h = _mm_srai_pi32(out3h, DESCALE_P##PASS); \
+  \
+  out5 = _mm_packs_pi32(out5l, out5h); \
+  out3 = _mm_packs_pi32(out3l, out3h); \
+}
+
+#define DO_FDCT_PASS1() { \
+  __m64 row0l, row0h, row1l, row1h, row2l, row2h, row3l, row3h; \
+  __m64 row01a, row01b, row01c, row01d, row23a, row23b, row23c, row23d; \
+  __m64 col0, col1, col2, col3, col4, col5, col6, col7; \
+  __m64 tmp10, tmp11; \
+  \
+  row0l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 0);     /* (00 01 02 03) */ \
+  row0h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 0 + 4); /* (04 05 06 07) */ \
+  row1l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 1);     /* (10 11 12 13) */ \
+  row1h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 1 + 4); /* (14 15 16 17) */ \
+  row2l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 2);     /* (20 21 22 23) */ \
+  row2h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 2 + 4); /* (24 25 26 27) */ \
+  row3l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 3);     /* (30 31 32 33) */ \
+  row3h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 3 + 4); /* (34 35 36 37) */ \
+  \
+  /* Transpose coefficients */ \
+  \
+  row23a = _mm_unpacklo_pi16(row2l, row3l);   /* row23a=(20 30 21 31) */ \
+  row23b = _mm_unpackhi_pi16(row2l, row3l);   /* row23b=(22 32 23 33) */ \
+  row23c = _mm_unpacklo_pi16(row2h, row3h);   /* row23c=(24 34 25 35) */ \
+  row23d = _mm_unpackhi_pi16(row2h, row3h);   /* row23d=(26 36 27 37) */ \
+  \
+  row01a = _mm_unpacklo_pi16(row0l, row1l);   /* row01a=(00 10 01 11) */ \
+  row01b = _mm_unpackhi_pi16(row0l, row1l);   /* row01b=(02 12 03 13) */ \
+  row01c = _mm_unpacklo_pi16(row0h, row1h);   /* row01c=(04 14 05 15) */ \
+  row01d = _mm_unpackhi_pi16(row0h, row1h);   /* row01d=(06 16 07 17) */ \
+  \
+  col0 = _mm_unpacklo_pi32(row01a, row23a);   /* col0=(00 10 20 30) */ \
+  col1 = _mm_unpackhi_pi32(row01a, row23a);   /* col1=(01 11 21 31) */ \
+  col6 = _mm_unpacklo_pi32(row01d, row23d);   /* col6=(06 16 26 36) */ \
+  col7 = _mm_unpackhi_pi32(row01d, row23d);   /* col7=(07 17 27 37) */ \
+  \
+  tmp6 = _mm_sub_pi16(col1, col6);            /* tmp6=col1-col6 */ \
+  tmp7 = _mm_sub_pi16(col0, col7);            /* tmp7=col0-col7 */ \
+  tmp1 = _mm_add_pi16(col1, col6);            /* tmp1=col1+col6 */ \
+  tmp0 = _mm_add_pi16(col0, col7);            /* tmp0=col0+col7 */ \
+  \
+  col2 = _mm_unpacklo_pi32(row01b, row23b);   /* col2=(02 12 22 32) */ \
+  col3 = _mm_unpackhi_pi32(row01b, row23b);   /* col3=(03 13 23 33) */ \
+  col4 = _mm_unpacklo_pi32(row01c, row23c);   /* col4=(04 14 24 34) */ \
+  col5 = _mm_unpackhi_pi32(row01c, row23c);   /* col5=(05 15 25 35) */ \
+  \
+  tmp3 = _mm_add_pi16(col3, col4);            /* tmp3=col3+col4 */ \
+  tmp2 = _mm_add_pi16(col2, col5);            /* tmp2=col2+col5 */ \
+  tmp4 = _mm_sub_pi16(col3, col4);            /* tmp4=col3-col4 */ \
+  tmp5 = _mm_sub_pi16(col2, col5);            /* tmp5=col2-col5 */ \
+  \
+  /* Even part */ \
+  \
+  tmp10 = _mm_add_pi16(tmp0, tmp3);           /* tmp10=tmp0+tmp3 */ \
+  tmp13 = _mm_sub_pi16(tmp0, tmp3);           /* tmp13=tmp0-tmp3 */ \
+  tmp11 = _mm_add_pi16(tmp1, tmp2);           /* tmp11=tmp1+tmp2 */ \
+  tmp12 = _mm_sub_pi16(tmp1, tmp2);           /* tmp12=tmp1-tmp2 */ \
+  \
+  out0 = _mm_add_pi16(tmp10, tmp11);          /* out0=tmp10+tmp11 */ \
+  out4 = _mm_sub_pi16(tmp10, tmp11);          /* out4=tmp10-tmp11 */ \
+  out0 = _mm_slli_pi16(out0, PASS1_BITS); \
+  out4 = _mm_slli_pi16(out4, PASS1_BITS); \
+  \
+  DO_FDCT_COMMON(1) \
+  \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 0, out0); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 0 + 4, out4); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 1, out1); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 1 + 4, out5); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 2, out2); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 2 + 4, out6); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 3, out3); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 3 + 4, out7); \
+}
+
+#define DO_FDCT_PASS2() { \
+  __m64 col0l, col0h, col1l, col1h, col2l, col2h, col3l, col3h; \
+  __m64 col01a, col01b, col01c, col01d, col23a, col23b, col23c, col23d; \
+  __m64 row0, row1, row2, row3, row4, row5, row6, row7; \
+  __m64 tmp10, tmp11; \
+  \
+  col0l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 0);  /* (00 10 20 30) */ \
+  col1l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 1);  /* (01 11 21 31) */ \
+  col2l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 2);  /* (02 12 22 32) */ \
+  col3l = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 3);  /* (03 13 23 33) */ \
+  col0h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 4);  /* (40 50 60 70) */ \
+  col1h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 5);  /* (41 51 61 71) */ \
+  col2h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 6);  /* (42 52 62 72) */ \
+  col3h = _mm_load_si64((__m64 *)&dataptrDCTSIZE * 7);  /* (43 53 63 73) */ \
+  \
+  /* Transpose coefficients */ \
+  \
+  col23a = _mm_unpacklo_pi16(col2l, col3l);   /* col23a=(02 03 12 13) */ \
+  col23b = _mm_unpackhi_pi16(col2l, col3l);   /* col23b=(22 23 32 33) */ \
+  col23c = _mm_unpacklo_pi16(col2h, col3h);   /* col23c=(42 43 52 53) */ \
+  col23d = _mm_unpackhi_pi16(col2h, col3h);   /* col23d=(62 63 72 73) */ \
+  \
+  col01a = _mm_unpacklo_pi16(col0l, col1l);   /* col01a=(00 01 10 11) */ \
+  col01b = _mm_unpackhi_pi16(col0l, col1l);   /* col01b=(20 21 30 31) */ \
+  col01c = _mm_unpacklo_pi16(col0h, col1h);   /* col01c=(40 41 50 51) */ \
+  col01d = _mm_unpackhi_pi16(col0h, col1h);   /* col01d=(60 61 70 71) */ \
+  \
+  row0 = _mm_unpacklo_pi32(col01a, col23a);   /* row0=(00 01 02 03) */ \
+  row1 = _mm_unpackhi_pi32(col01a, col23a);   /* row1=(10 11 12 13) */ \
+  row6 = _mm_unpacklo_pi32(col01d, col23d);   /* row6=(60 61 62 63) */ \
+  row7 = _mm_unpackhi_pi32(col01d, col23d);   /* row7=(70 71 72 73) */ \
+  \
+  tmp6 = _mm_sub_pi16(row1, row6);            /* tmp6=row1-row6 */ \
+  tmp7 = _mm_sub_pi16(row0, row7);            /* tmp7=row0-row7 */ \
+  tmp1 = _mm_add_pi16(row1, row6);            /* tmp1=row1+row6 */ \
+  tmp0 = _mm_add_pi16(row0, row7);            /* tmp0=row0+row7 */ \
+  \
+  row2 = _mm_unpacklo_pi32(col01b, col23b);   /* row2=(20 21 22 23) */ \
+  row3 = _mm_unpackhi_pi32(col01b, col23b);   /* row3=(30 31 32 33) */ \
+  row4 = _mm_unpacklo_pi32(col01c, col23c);   /* row4=(40 41 42 43) */ \
+  row5 = _mm_unpackhi_pi32(col01c, col23c);   /* row5=(50 51 52 53) */ \
+  \
+  tmp3 = _mm_add_pi16(row3, row4);            /* tmp3=row3+row4 */ \
+  tmp2 = _mm_add_pi16(row2, row5);            /* tmp2=row2+row5 */ \
+  tmp4 = _mm_sub_pi16(row3, row4);            /* tmp4=row3-row4 */ \
+  tmp5 = _mm_sub_pi16(row2, row5);            /* tmp5=row2-row5 */ \
+  \
+  /* Even part */ \
+  \
+  tmp10 = _mm_add_pi16(tmp0, tmp3);           /* tmp10=tmp0+tmp3 */ \
+  tmp13 = _mm_sub_pi16(tmp0, tmp3);           /* tmp13=tmp0-tmp3 */ \
+  tmp11 = _mm_add_pi16(tmp1, tmp2);           /* tmp11=tmp1+tmp2 */ \
+  tmp12 = _mm_sub_pi16(tmp1, tmp2);           /* tmp12=tmp1-tmp2 */ \
+  \
+  out0 = _mm_add_pi16(tmp10, tmp11);          /* out0=tmp10+tmp11 */ \
+  out4 = _mm_sub_pi16(tmp10, tmp11);          /* out4=tmp10-tmp11 */ \
+  \
+  out0 = _mm_add_pi16(out0, PW_DESCALE_P2X); \
+  out4 = _mm_add_pi16(out4, PW_DESCALE_P2X); \
+  out0 = _mm_srai_pi16(out0, PASS1_BITS); \
+  out4 = _mm_srai_pi16(out4, PASS1_BITS); \
+  \
+  DO_FDCT_COMMON(2) \
+  \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 0, out0); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 1, out1); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 2, out2); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 3, out3); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 4, out4); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 5, out5); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 6, out6); \
+  _mm_store_si64((__m64 *)&dataptrDCTSIZE * 7, out7); \
+}
+
+void jsimd_fdct_islow_mmi(DCTELEM *data)
+{
+  __m64 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  __m64 out0, out1, out2, out3, out4, out5, out6, out7;
+  __m64 tmp12, tmp13;
+  DCTELEM *dataptr = data;
+
+  /* Pass 1: process rows. */
+
+  DO_FDCT_PASS1()
+  dataptr += DCTSIZE * 4;
+  DO_FDCT_PASS1()
+
+  /* Pass 2: process columns. */
+
+  dataptr = data;
+  DO_FDCT_PASS2()
+  dataptr += 4;
+  DO_FDCT_PASS2()
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jidctfst-mmi.c Added

@@ -0,0 +1,395 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, 2018-2019, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  LiuQingfa <liuqingfa-hf@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* FAST INTEGER INVERSE DCT */
+
+#include "jsimd_mmi.h"
+
+
+#define CONST_BITS  8
+#define PASS1_BITS  2
+
+#define FIX_1_082  ((short)277)                   /* FIX(1.082392200) */
+#define FIX_1_414  ((short)362)                   /* FIX(1.414213562) */
+#define FIX_1_847  ((short)473)                   /* FIX(1.847759065) */
+#define FIX_2_613  ((short)669)                   /* FIX(2.613125930) */
+#define FIX_1_613  ((short)(FIX_2_613 - 256 * 3)) /* FIX(2.613125930) - FIX(1) */
+
+#define PRE_MULTIPLY_SCALE_BITS  2
+#define CONST_SHIFT  (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+enum const_index {
+  index_PW_F1082,
+  index_PW_F1414,
+  index_PW_F1847,
+  index_PW_MF1613,
+  index_PB_CENTERJSAMP
+};
+
+static uint64_t const_value = {
+  _uint64_set1_pi16(FIX_1_082 << CONST_SHIFT),
+  _uint64_set1_pi16(FIX_1_414 << CONST_SHIFT),
+  _uint64_set1_pi16(FIX_1_847 << CONST_SHIFT),
+  _uint64_set1_pi16(-FIX_1_613 << CONST_SHIFT),
+  _uint64_set1_pi8(CENTERJSAMPLE)
+};
+
+#define PW_F1414        get_const_value(index_PW_F1414)
+#define PW_F1847        get_const_value(index_PW_F1847)
+#define PW_MF1613       get_const_value(index_PW_MF1613)
+#define PW_F1082        get_const_value(index_PW_F1082)
+#define PB_CENTERJSAMP  get_const_value(index_PB_CENTERJSAMP)
+
+
+#define test_m32_zero(mm32)  (!(*(uint32_t *)&mm32))
+#define test_m64_zero(mm64)  (!(*(uint64_t *)&mm64))
+
+
+#define DO_IDCT_COMMON() { \
+  tmp7 = _mm_add_pi16(z11, z13); \
+  \
+  tmp11 = _mm_sub_pi16(z11, z13); \
+  tmp11 = _mm_slli_pi16(tmp11, PRE_MULTIPLY_SCALE_BITS); \
+  tmp11 = _mm_mulhi_pi16(tmp11, PW_F1414); \
+  \
+  tmp10 = _mm_slli_pi16(z12, PRE_MULTIPLY_SCALE_BITS); \
+  tmp12 = _mm_slli_pi16(z10, PRE_MULTIPLY_SCALE_BITS); \
+  \
+  /* To avoid overflow... \
+   * \
+   * (Original) \
+   * tmp12 = -2.613125930 * z10 + z5; \
+   * \
+   * (This implementation) \
+   * tmp12 = (-1.613125930 - 1) * z10 + z5; \
+   *       = -1.613125930 * z10 - z10 + z5; \
+   */ \
+  \
+  z5 = _mm_add_pi16(tmp10, tmp12); \
+  z5 = _mm_mulhi_pi16(z5, PW_F1847); \
+  \
+  tmp10 = _mm_mulhi_pi16(tmp10, PW_F1082); \
+  tmp10 = _mm_sub_pi16(tmp10, z5); \
+  tmp12 = _mm_mulhi_pi16(tmp12, PW_MF1613); \
+  tmp12 = _mm_sub_pi16(tmp12, z10); \
+  tmp12 = _mm_sub_pi16(tmp12, z10); \
+  tmp12 = _mm_sub_pi16(tmp12, z10); \
+  tmp12 = _mm_add_pi16(tmp12, z5); \
+  \
+  /* Final output stage */ \
+  \
+  tmp6 = _mm_sub_pi16(tmp12, tmp7); \
+  tmp5 = _mm_sub_pi16(tmp11, tmp6); \
+  tmp4 = _mm_add_pi16(tmp10, tmp5); \
+  \
+  out0 = _mm_add_pi16(tmp0, tmp7); \
+  out7 = _mm_sub_pi16(tmp0, tmp7); \
+  out1 = _mm_add_pi16(tmp1, tmp6); \
+  out6 = _mm_sub_pi16(tmp1, tmp6); \
+  \
+  out2 = _mm_add_pi16(tmp2, tmp5); \
+  out5 = _mm_sub_pi16(tmp2, tmp5); \
+  out4 = _mm_add_pi16(tmp3, tmp4); \
+  out3 = _mm_sub_pi16(tmp3, tmp4); \
+}
+
+#define DO_IDCT_PASS1(iter) { \
+  __m64 col0l, col1l, col2l, col3l, col4l, col5l, col6l, col7l; \
+  __m64 quant0l, quant1l, quant2l, quant3l; \
+  __m64 quant4l, quant5l, quant6l, quant7l; \
+  __m64 row01a, row01b, row01c, row01d, row23a, row23b, row23c, row23d; \
+  __m64 row0l, row0h, row1l, row1h, row2l, row2h, row3l, row3h; \
+  __m32 col0a, col1a, mm0; \
+  \
+  col0a = _mm_load_si32((__m32 *)&inptrDCTSIZE * 1); \
+  col1a = _mm_load_si32((__m32 *)&inptrDCTSIZE * 2); \
+  mm0 = _mm_or_si32(col0a, col1a); \
+  \
+  if (test_m32_zero(mm0)) { \
+    __m64 mm1, mm2; \
+    \
+    col0l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 0); \
+    col1l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 1); \
+    col2l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 2); \
+    col3l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 3); \
+    col4l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 4); \
+    col5l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 5); \
+    col6l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 6); \
+    col7l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 7); \
+    \
+    mm1 = _mm_or_si64(col1l, col3l); \
+    mm2 = _mm_or_si64(col2l, col4l); \
+    mm1 = _mm_or_si64(mm1, col5l); \
+    mm2 = _mm_or_si64(mm2, col6l); \
+    mm1 = _mm_or_si64(mm1, col7l); \
+    mm1 = _mm_or_si64(mm1, mm2); \
+    \
+    if (test_m64_zero(mm1)) { \
+      __m64 dcval, dcvall, dcvalh, row0, row1, row2, row3; \
+      \
+      /* AC terms all zero */ \
+      \
+      quant0l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 0); \
+      \
+      dcval = _mm_mullo_pi16(col0l, quant0l);    /* dcval=(00 10 20 30) */ \
+      \
+      dcvall = _mm_unpacklo_pi16(dcval, dcval);  /* dcvall=(00 00 10 10) */ \
+      dcvalh = _mm_unpackhi_pi16(dcval, dcval);  /* dcvalh=(20 20 30 30) */ \
+      \
+      row0 = _mm_unpacklo_pi32(dcvall, dcvall);  /* row0=(00 00 00 00) */ \
+      row1 = _mm_unpackhi_pi32(dcvall, dcvall);  /* row1=(10 10 10 10) */ \
+      row2 = _mm_unpacklo_pi32(dcvalh, dcvalh);  /* row2=(20 20 20 20) */ \
+      row3 = _mm_unpackhi_pi32(dcvalh, dcvalh);  /* row3=(30 30 30 30) */ \
+      \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 0, row0); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 0 + 4, row0); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 1, row1); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 1 + 4, row1); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 2, row2); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 2 + 4, row2); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 3, row3); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 3 + 4, row3); \
+      \
+      goto nextcolumn##iter; \
+    } \
+  } \
+  \
+  /* Even part */ \
+  \
+  col0l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 0);  /* (00 10 20 30) */ \
+  col2l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 2);  /* (02 12 22 32) */ \
+  col4l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 4);  /* (04 14 24 34) */ \
+  col6l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 6);  /* (06 16 26 36) */ \
+  \
+  quant0l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 0); \
+  quant2l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 2); \
+  quant4l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 4); \
+  quant6l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 6); \
+  \
+  tmp0 = _mm_mullo_pi16(col0l, quant0l); \
+  tmp1 = _mm_mullo_pi16(col2l, quant2l); \
+  tmp2 = _mm_mullo_pi16(col4l, quant4l); \
+  tmp3 = _mm_mullo_pi16(col6l, quant6l); \
+  \
+  tmp10 = _mm_add_pi16(tmp0, tmp2); \
+  tmp11 = _mm_sub_pi16(tmp0, tmp2); \
+  tmp13 = _mm_add_pi16(tmp1, tmp3); \
+  \
+  tmp12 = _mm_sub_pi16(tmp1, tmp3); \
+  tmp12 = _mm_slli_pi16(tmp12, PRE_MULTIPLY_SCALE_BITS); \
+  tmp12 = _mm_mulhi_pi16(tmp12, PW_F1414); \
+  tmp12 = _mm_sub_pi16(tmp12, tmp13); \
+  \
+  tmp0 = _mm_add_pi16(tmp10, tmp13); \
+  tmp3 = _mm_sub_pi16(tmp10, tmp13); \
+  tmp1 = _mm_add_pi16(tmp11, tmp12); \
+  tmp2 = _mm_sub_pi16(tmp11, tmp12); \
+  \
+  /* Odd part */ \
+  \
+  col1l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 1);  /* (01 11 21 31) */ \
+  col3l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 3);  /* (03 13 23 33) */ \
+  col5l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 5);  /* (05 15 25 35) */ \
+  col7l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 7);  /* (07 17 27 37) */ \
+  \
+  quant1l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 1); \
+  quant3l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 3); \
+  quant5l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 5); \
+  quant7l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 7); \
+  \
+  tmp4 = _mm_mullo_pi16(col1l, quant1l); \
+  tmp5 = _mm_mullo_pi16(col3l, quant3l); \
+  tmp6 = _mm_mullo_pi16(col5l, quant5l); \
+  tmp7 = _mm_mullo_pi16(col7l, quant7l); \
+  \
+  z13 = _mm_add_pi16(tmp6, tmp5); \
+  z10 = _mm_sub_pi16(tmp6, tmp5); \
+  z11 = _mm_add_pi16(tmp4, tmp7); \
+  z12 = _mm_sub_pi16(tmp4, tmp7); \
+  \
+  DO_IDCT_COMMON() \
+  \
+  /* out0=(00 10 20 30), out1=(01 11 21 31) */ \
+  /* out2=(02 12 22 32), out3=(03 13 23 33) */ \
+  /* out4=(04 14 24 34), out5=(05 15 25 35) */ \
+  /* out6=(06 16 26 36), out7=(07 17 27 37) */ \
+  \
+  /* Transpose coefficients */ \
+  \
+  row01a = _mm_unpacklo_pi16(out0, out1);     /* row01a=(00 01 10 11) */ \
+  row23a = _mm_unpackhi_pi16(out0, out1);     /* row23a=(20 21 30 31) */ \
+  row01d = _mm_unpacklo_pi16(out6, out7);     /* row01d=(06 07 16 17) */ \
+  row23d = _mm_unpackhi_pi16(out6, out7);     /* row23d=(26 27 36 37) */ \
+  \
+  row01b = _mm_unpacklo_pi16(out2, out3);     /* row01b=(02 03 12 13) */ \
+  row23b = _mm_unpackhi_pi16(out2, out3);     /* row23b=(22 23 32 33) */ \
+  row01c = _mm_unpacklo_pi16(out4, out5);     /* row01c=(04 05 14 15) */ \
+  row23c = _mm_unpackhi_pi16(out4, out5);     /* row23c=(24 25 34 35) */ \
+  \
+  row0l = _mm_unpacklo_pi32(row01a, row01b);  /* row0l=(00 01 02 03) */ \
+  row1l = _mm_unpackhi_pi32(row01a, row01b);  /* row1l=(10 11 12 13) */ \
+  row2l = _mm_unpacklo_pi32(row23a, row23b);  /* row2l=(20 21 22 23) */ \
+  row3l = _mm_unpackhi_pi32(row23a, row23b);  /* row3l=(30 31 32 33) */ \
+  \
+  row0h = _mm_unpacklo_pi32(row01c, row01d);  /* row0h=(04 05 06 07) */ \
+  row1h = _mm_unpackhi_pi32(row01c, row01d);  /* row1h=(14 15 16 17) */ \
+  row2h = _mm_unpacklo_pi32(row23c, row23d);  /* row2h=(24 25 26 27) */ \
+  row3h = _mm_unpackhi_pi32(row23c, row23d);  /* row3h=(34 35 36 37) */ \
+  \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 0, row0l); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 0 + 4, row0h); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 1, row1l); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 1 + 4, row1h); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 2, row2l); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 2 + 4, row2h); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 3, row3l); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 3 + 4, row3h); \
+}
+
+#define DO_IDCT_PASS2(ctr) { \
+  __m64 row0l, row1l, row2l, row3l, row4l, row5l, row6l, row7l; \
+  __m64 col0123a, col0123b, col0123c, col0123d; \
+  __m64 col01l, col01h, col23l, col23h; \
+  __m64 col0, col1, col2, col3; \
+  __m64 row06, row17, row24, row35; \
+  \
+  row0l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 0);  /* (00 01 02 03) */ \
+  row1l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 1);  /* (10 11 12 13) */ \
+  row2l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 2);  /* (20 21 22 23) */ \
+  row3l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 3);  /* (30 31 32 33) */ \
+  row4l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 4);  /* (40 41 42 43) */ \
+  row5l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 5);  /* (50 51 52 53) */ \
+  row6l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 6);  /* (60 61 62 63) */ \
+  row7l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 7);  /* (70 71 72 73) */ \
+  \
+  /* Even part */ \
+  \
+  tmp10 = _mm_add_pi16(row0l, row4l); \
+  tmp11 = _mm_sub_pi16(row0l, row4l); \
+  tmp13 = _mm_add_pi16(row2l, row6l); \
+  \
+  tmp12 = _mm_sub_pi16(row2l, row6l); \
+  tmp12 = _mm_slli_pi16(tmp12, PRE_MULTIPLY_SCALE_BITS); \
+  tmp12 = _mm_mulhi_pi16(tmp12, PW_F1414); \
+  tmp12 = _mm_sub_pi16(tmp12, tmp13); \
+  \
+  tmp0 = _mm_add_pi16(tmp10, tmp13); \
+  tmp3 = _mm_sub_pi16(tmp10, tmp13); \
+  tmp1 = _mm_add_pi16(tmp11, tmp12); \
+  tmp2 = _mm_sub_pi16(tmp11, tmp12); \
+  \
+  /* Odd part */ \
+  \
+  z13 = _mm_add_pi16(row5l, row3l); \
+  z10 = _mm_sub_pi16(row5l, row3l); \
+  z11 = _mm_add_pi16(row1l, row7l); \
+  z12 = _mm_sub_pi16(row1l, row7l); \
+  \
+  DO_IDCT_COMMON() \
+  \
+  /* out0=(00 01 02 03), out1=(10 11 12 13) */ \
+  /* out2=(20 21 22 23), out3=(30 31 32 33) */ \
+  /* out4=(40 41 42 43), out5=(50 51 52 53) */ \
+  /* out6=(60 61 62 63), out7=(70 71 72 73) */ \
+  \
+  out0 = _mm_srai_pi16(out0, PASS1_BITS + 3); \
+  out1 = _mm_srai_pi16(out1, PASS1_BITS + 3); \
+  out2 = _mm_srai_pi16(out2, PASS1_BITS + 3); \
+  out3 = _mm_srai_pi16(out3, PASS1_BITS + 3); \
+  out4 = _mm_srai_pi16(out4, PASS1_BITS + 3); \
+  out5 = _mm_srai_pi16(out5, PASS1_BITS + 3); \
+  out6 = _mm_srai_pi16(out6, PASS1_BITS + 3); \
+  out7 = _mm_srai_pi16(out7, PASS1_BITS + 3); \
+  \
+  row06 = _mm_packs_pi16(out0, out6);  /* row06=(00 01 02 03 60 61 62 63) */ \
+  row17 = _mm_packs_pi16(out1, out7);  /* row17=(10 11 12 13 70 71 72 73) */ \
+  row24 = _mm_packs_pi16(out2, out4);  /* row24=(20 21 22 23 40 41 42 43) */ \
+  row35 = _mm_packs_pi16(out3, out5);  /* row35=(30 31 32 33 50 51 52 53) */ \
+  \
+  row06 = _mm_add_pi8(row06, PB_CENTERJSAMP); \
+  row17 = _mm_add_pi8(row17, PB_CENTERJSAMP); \
+  row24 = _mm_add_pi8(row24, PB_CENTERJSAMP); \
+  row35 = _mm_add_pi8(row35, PB_CENTERJSAMP); \
+  \
+  /* Transpose coefficients */ \
+  \
+  col0123a = _mm_unpacklo_pi8(row06, row17);  /* col0123a=(00 10 01 11 02 12 03 13) */ \
+  col0123d = _mm_unpackhi_pi8(row06, row17);  /* col0123d=(60 70 61 71 62 72 63 73) */ \
+  col0123b = _mm_unpacklo_pi8(row24, row35);  /* col0123b=(20 30 21 31 22 32 23 33) */ \
+  col0123c = _mm_unpackhi_pi8(row24, row35);  /* col0123c=(40 50 41 51 42 52 43 53) */ \
+  \
+  col01l = _mm_unpacklo_pi16(col0123a, col0123b);  /* col01l=(00 10 20 30 01 11 21 31) */ \
+  col23l = _mm_unpackhi_pi16(col0123a, col0123b);  /* col23l=(02 12 22 32 03 13 23 33) */ \
+  col01h = _mm_unpacklo_pi16(col0123c, col0123d);  /* col01h=(40 50 60 70 41 51 61 71) */ \
+  col23h = _mm_unpackhi_pi16(col0123c, col0123d);  /* col23h=(42 52 62 72 43 53 63 73) */ \
+  \
+  col0 = _mm_unpacklo_pi32(col01l, col01h);   /* col0=(00 10 20 30 40 50 60 70) */ \
+  col1 = _mm_unpackhi_pi32(col01l, col01h);   /* col1=(01 11 21 31 41 51 61 71) */ \
+  col2 = _mm_unpacklo_pi32(col23l, col23h);   /* col2=(02 12 22 32 42 52 62 72) */ \
+  col3 = _mm_unpackhi_pi32(col23l, col23h);   /* col3=(03 13 23 33 43 53 63 73) */ \
+  \
+  _mm_store_si64((__m64 *)(output_bufctr + 0 + output_col), col0); \
+  _mm_store_si64((__m64 *)(output_bufctr + 1 + output_col), col1); \
+  _mm_store_si64((__m64 *)(output_bufctr + 2 + output_col), col2); \
+  _mm_store_si64((__m64 *)(output_bufctr + 3 + output_col), col3); \
+}
+
+void jsimd_idct_ifast_mmi(void *dct_table, JCOEFPTR coef_block,
+                          JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  __m64 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  __m64 tmp10, tmp11, tmp12, tmp13;
+  __m64 out0, out1, out2, out3, out4, out5, out6, out7;
+  __m64 z5, z10, z11, z12, z13;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  JCOEF *wsptr;
+  JCOEF workspaceDCTSIZE2;  /* buffers data between passes */
+
+  /* Pass 1: process columns. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)dct_table;
+  wsptr = workspace;
+
+  DO_IDCT_PASS1(1)
+nextcolumn1:
+  inptr += 4;
+  quantptr += 4;
+  wsptr += DCTSIZE * 4;
+  DO_IDCT_PASS1(2)
+nextcolumn2:
+
+  /* Pass 2: process rows. */
+
+  wsptr = workspace;
+
+  DO_IDCT_PASS2(0)
+  wsptr += 4;
+  DO_IDCT_PASS2(4)
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jidctint-mmi.c Added

@@ -0,0 +1,571 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, 2018, 2020, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2016-2017, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* ACCUATE INTEGER INVERSE DCT */
+
+#include "jsimd_mmi.h"
+
+
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2  (CONST_BITS + PASS1_BITS + 3)
+#define CENTERJSAMPLE  128
+
+#define FIX_0_298  ((short)2446)  /* FIX(0.298631336) */
+#define FIX_0_390  ((short)3196)  /* FIX(0.390180644) */
+#define FIX_0_899  ((short)7373)  /* FIX(0.899976223) */
+#define FIX_0_541  ((short)4433)  /* FIX(0.541196100) */
+#define FIX_0_765  ((short)6270)  /* FIX(0.765366865) */
+#define FIX_1_175  ((short)9633)  /* FIX(1.175875602) */
+#define FIX_1_501  ((short)12299) /* FIX(1.501321110) */
+#define FIX_1_847  ((short)15137) /* FIX(1.847759065) */
+#define FIX_1_961  ((short)16069) /* FIX(1.961570560) */
+#define FIX_2_053  ((short)16819) /* FIX(2.053119869) */
+#define FIX_2_562  ((short)20995) /* FIX(2.562915447) */
+#define FIX_3_072  ((short)25172) /* FIX(3.072711026) */
+
+enum const_index {
+  index_PW_F130_F054,
+  index_PW_F054_MF130,
+  index_PW_MF078_F117,
+  index_PW_F117_F078,
+  index_PW_MF060_MF089,
+  index_PW_MF089_F060,
+  index_PW_MF050_MF256,
+  index_PW_MF256_F050,
+  index_PD_DESCALE_P1,
+  index_PD_DESCALE_P2,
+  index_PB_CENTERJSAMP
+};
+
+static uint64_t const_value = {
+  _uint64_set_pi16(FIX_0_541, (FIX_0_541 + FIX_0_765),
+                   FIX_0_541, (FIX_0_541 + FIX_0_765)),
+  _uint64_set_pi16((FIX_0_541 - FIX_1_847), FIX_0_541,
+                   (FIX_0_541 - FIX_1_847), FIX_0_541),
+  _uint64_set_pi16(FIX_1_175, (FIX_1_175 - FIX_1_961),
+                   FIX_1_175, (FIX_1_175 - FIX_1_961)),
+  _uint64_set_pi16((FIX_1_175 - FIX_0_390), FIX_1_175,
+                   (FIX_1_175 - FIX_0_390), FIX_1_175),
+  _uint64_set_pi16(-FIX_0_899, (FIX_0_298 - FIX_0_899),
+                   -FIX_0_899, (FIX_0_298 - FIX_0_899)),
+  _uint64_set_pi16((FIX_1_501 - FIX_0_899), -FIX_0_899,
+                   (FIX_1_501 - FIX_0_899), -FIX_0_899),
+  _uint64_set_pi16(-FIX_2_562, (FIX_2_053 - FIX_2_562),
+                   -FIX_2_562, (FIX_2_053 - FIX_2_562)),
+  _uint64_set_pi16((FIX_3_072 - FIX_2_562), -FIX_2_562,
+                   (FIX_3_072 - FIX_2_562), -FIX_2_562),
+  _uint64_set_pi32((1 << (DESCALE_P1 - 1)), (1 << (DESCALE_P1 - 1))),
+  _uint64_set_pi32((1 << (DESCALE_P2 - 1)), (1 << (DESCALE_P2 - 1))),
+  _uint64_set_pi8(CENTERJSAMPLE, CENTERJSAMPLE, CENTERJSAMPLE, CENTERJSAMPLE,
+                  CENTERJSAMPLE, CENTERJSAMPLE, CENTERJSAMPLE, CENTERJSAMPLE)
+};
+
+#define PW_F130_F054    get_const_value(index_PW_F130_F054)
+#define PW_F054_MF130   get_const_value(index_PW_F054_MF130)
+#define PW_MF078_F117   get_const_value(index_PW_MF078_F117)
+#define PW_F117_F078    get_const_value(index_PW_F117_F078)
+#define PW_MF060_MF089  get_const_value(index_PW_MF060_MF089)
+#define PW_MF089_F060   get_const_value(index_PW_MF089_F060)
+#define PW_MF050_MF256  get_const_value(index_PW_MF050_MF256)
+#define PW_MF256_F050   get_const_value(index_PW_MF256_F050)
+#define PD_DESCALE_P1   get_const_value(index_PD_DESCALE_P1)
+#define PD_DESCALE_P2   get_const_value(index_PD_DESCALE_P2)
+#define PB_CENTERJSAMP  get_const_value(index_PB_CENTERJSAMP)
+
+
+#define test_m32_zero(mm32)  (!(*(uint32_t *)&mm32))
+#define test_m64_zero(mm64)  (!(*(uint64_t *)&mm64))
+
+
+#define DO_IDCT_COMMON(PASS) { \
+  __m64 tmp0_3l, tmp0_3h, tmp1_2l, tmp1_2h; \
+  __m64 tmp0l, tmp0h, tmp1l, tmp1h, tmp2l, tmp2h, tmp3l, tmp3h; \
+  __m64 z34l, z34h, z3l, z3h, z4l, z4h, z3, z4; \
+  __m64 out0l, out0h, out1l, out1h, out2l, out2h, out3l, out3h; \
+  __m64 out4l, out4h, out5l, out5h, out6l, out6h, out7l, out7h; \
+  \
+  z3 = _mm_add_pi16(tmp0, tmp2); \
+  z4 = _mm_add_pi16(tmp1, tmp3); \
+  \
+  /* (Original) \
+   * z5 = (z3 + z4) * 1.175875602; \
+   * z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644; \
+   * z3 += z5;  z4 += z5; \
+   * \
+   * (This implementation) \
+   * z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; \
+   * z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); \
+   */ \
+  \
+  z34l = _mm_unpacklo_pi16(z3, z4); \
+  z34h = _mm_unpackhi_pi16(z3, z4); \
+  z3l = _mm_madd_pi16(z34l, PW_MF078_F117); \
+  z3h = _mm_madd_pi16(z34h, PW_MF078_F117); \
+  z4l = _mm_madd_pi16(z34l, PW_F117_F078); \
+  z4h = _mm_madd_pi16(z34h, PW_F117_F078); \
+  \
+  /* (Original) \
+   * z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2; \
+   * tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869; \
+   * tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110; \
+   * z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447; \
+   * tmp0 += z1 + z3;  tmp1 += z2 + z4; \
+   * tmp2 += z2 + z3;  tmp3 += z1 + z4; \
+   * \
+   * (This implementation) \
+   * tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223; \
+   * tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447; \
+   * tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447); \
+   * tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223); \
+   * tmp0 += z3;  tmp1 += z4; \
+   * tmp2 += z3;  tmp3 += z4; \
+   */ \
+  \
+  tmp0_3l = _mm_unpacklo_pi16(tmp0, tmp3); \
+  tmp0_3h = _mm_unpackhi_pi16(tmp0, tmp3); \
+  \
+  tmp0l = _mm_madd_pi16(tmp0_3l, PW_MF060_MF089); \
+  tmp0h = _mm_madd_pi16(tmp0_3h, PW_MF060_MF089); \
+  tmp3l = _mm_madd_pi16(tmp0_3l, PW_MF089_F060); \
+  tmp3h = _mm_madd_pi16(tmp0_3h, PW_MF089_F060); \
+  \
+  tmp0l = _mm_add_pi32(tmp0l, z3l); \
+  tmp0h = _mm_add_pi32(tmp0h, z3h); \
+  tmp3l = _mm_add_pi32(tmp3l, z4l); \
+  tmp3h = _mm_add_pi32(tmp3h, z4h); \
+  \
+  tmp1_2l = _mm_unpacklo_pi16(tmp1, tmp2); \
+  tmp1_2h = _mm_unpackhi_pi16(tmp1, tmp2); \
+  \
+  tmp1l = _mm_madd_pi16(tmp1_2l, PW_MF050_MF256); \
+  tmp1h = _mm_madd_pi16(tmp1_2h, PW_MF050_MF256); \
+  tmp2l = _mm_madd_pi16(tmp1_2l, PW_MF256_F050); \
+  tmp2h = _mm_madd_pi16(tmp1_2h, PW_MF256_F050); \
+  \
+  tmp1l = _mm_add_pi32(tmp1l, z4l); \
+  tmp1h = _mm_add_pi32(tmp1h, z4h); \
+  tmp2l = _mm_add_pi32(tmp2l, z3l); \
+  tmp2h = _mm_add_pi32(tmp2h, z3h); \
+  \
+  /* Final output stage */ \
+  \
+  out0l = _mm_add_pi32(tmp10l, tmp3l); \
+  out0h = _mm_add_pi32(tmp10h, tmp3h); \
+  out7l = _mm_sub_pi32(tmp10l, tmp3l); \
+  out7h = _mm_sub_pi32(tmp10h, tmp3h); \
+  \
+  out0l = _mm_add_pi32(out0l, PD_DESCALE_P##PASS); \
+  out0h = _mm_add_pi32(out0h, PD_DESCALE_P##PASS); \
+  out0l = _mm_srai_pi32(out0l, DESCALE_P##PASS); \
+  out0h = _mm_srai_pi32(out0h, DESCALE_P##PASS); \
+  \
+  out7l = _mm_add_pi32(out7l, PD_DESCALE_P##PASS); \
+  out7h = _mm_add_pi32(out7h, PD_DESCALE_P##PASS); \
+  out7l = _mm_srai_pi32(out7l, DESCALE_P##PASS); \
+  out7h = _mm_srai_pi32(out7h, DESCALE_P##PASS); \
+  \
+  out0 = _mm_packs_pi32(out0l, out0h); \
+  out7 = _mm_packs_pi32(out7l, out7h); \
+  \
+  out1l = _mm_add_pi32(tmp11l, tmp2l); \
+  out1h = _mm_add_pi32(tmp11h, tmp2h); \
+  out6l = _mm_sub_pi32(tmp11l, tmp2l); \
+  out6h = _mm_sub_pi32(tmp11h, tmp2h); \
+  \
+  out1l = _mm_add_pi32(out1l, PD_DESCALE_P##PASS); \
+  out1h = _mm_add_pi32(out1h, PD_DESCALE_P##PASS); \
+  out1l = _mm_srai_pi32(out1l, DESCALE_P##PASS); \
+  out1h = _mm_srai_pi32(out1h, DESCALE_P##PASS); \
+  \
+  out6l = _mm_add_pi32(out6l, PD_DESCALE_P##PASS); \
+  out6h = _mm_add_pi32(out6h, PD_DESCALE_P##PASS); \
+  out6l = _mm_srai_pi32(out6l, DESCALE_P##PASS); \
+  out6h = _mm_srai_pi32(out6h, DESCALE_P##PASS); \
+  \
+  out1 = _mm_packs_pi32(out1l, out1h); \
+  out6 = _mm_packs_pi32(out6l, out6h); \
+  \
+  out2l = _mm_add_pi32(tmp12l, tmp1l); \
+  out2h = _mm_add_pi32(tmp12h, tmp1h); \
+  out5l = _mm_sub_pi32(tmp12l, tmp1l); \
+  out5h = _mm_sub_pi32(tmp12h, tmp1h); \
+  \
+  out2l = _mm_add_pi32(out2l, PD_DESCALE_P##PASS); \
+  out2h = _mm_add_pi32(out2h, PD_DESCALE_P##PASS); \
+  out2l = _mm_srai_pi32(out2l, DESCALE_P##PASS); \
+  out2h = _mm_srai_pi32(out2h, DESCALE_P##PASS); \
+  \
+  out5l = _mm_add_pi32(out5l, PD_DESCALE_P##PASS); \
+  out5h = _mm_add_pi32(out5h, PD_DESCALE_P##PASS); \
+  out5l = _mm_srai_pi32(out5l, DESCALE_P##PASS); \
+  out5h = _mm_srai_pi32(out5h, DESCALE_P##PASS); \
+  \
+  out2 = _mm_packs_pi32(out2l, out2h); \
+  out5 = _mm_packs_pi32(out5l, out5h); \
+  \
+  out3l = _mm_add_pi32(tmp13l, tmp0l); \
+  out3h = _mm_add_pi32(tmp13h, tmp0h); \
+  \
+  out4l = _mm_sub_pi32(tmp13l, tmp0l); \
+  out4h = _mm_sub_pi32(tmp13h, tmp0h); \
+  \
+  out3l = _mm_add_pi32(out3l, PD_DESCALE_P##PASS); \
+  out3h = _mm_add_pi32(out3h, PD_DESCALE_P##PASS); \
+  out3l = _mm_srai_pi32(out3l, DESCALE_P##PASS); \
+  out3h = _mm_srai_pi32(out3h, DESCALE_P##PASS); \
+  \
+  out4l = _mm_add_pi32(out4l, PD_DESCALE_P##PASS); \
+  out4h = _mm_add_pi32(out4h, PD_DESCALE_P##PASS); \
+  out4l = _mm_srai_pi32(out4l, DESCALE_P##PASS); \
+  out4h = _mm_srai_pi32(out4h, DESCALE_P##PASS); \
+  \
+  out3 = _mm_packs_pi32(out3l, out3h); \
+  out4 = _mm_packs_pi32(out4l, out4h); \
+}
+
+#define DO_IDCT_PASS1(iter) { \
+  __m64 col0l, col1l, col2l, col3l, col4l, col5l, col6l, col7l; \
+  __m64 quant0l, quant1l, quant2l, quant3l; \
+  __m64 quant4l, quant5l, quant6l, quant7l; \
+  __m64 z23, z2, z3, z23l, z23h; \
+  __m64 row01a, row01b, row01c, row01d, row23a, row23b, row23c, row23d; \
+  __m64 row0l, row0h, row1l, row1h, row2l, row2h, row3l, row3h; \
+  __m64 tmp0l, tmp0h, tmp1l, tmp1h, tmp2l, tmp2h, tmp3l, tmp3h; \
+  __m64 tmp10l, tmp10h, tmp11l, tmp11h, tmp12l, tmp12h, tmp13l, tmp13h; \
+  __m32 col0a, col1a, mm0; \
+  \
+  col0a = _mm_load_si32((__m32 *)&inptrDCTSIZE * 1); \
+  col1a = _mm_load_si32((__m32 *)&inptrDCTSIZE * 2); \
+  mm0 = _mm_or_si32(col0a, col1a); \
+  \
+  if (test_m32_zero(mm0)) { \
+    __m64 mm1, mm2; \
+    \
+    col0l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 0); \
+    col1l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 1); \
+    col2l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 2); \
+    col3l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 3); \
+    col4l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 4); \
+    col5l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 5); \
+    col6l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 6); \
+    col7l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 7); \
+    \
+    mm1 = _mm_or_si64(col1l, col3l); \
+    mm2 = _mm_or_si64(col2l, col4l); \
+    mm1 = _mm_or_si64(mm1, col5l); \
+    mm2 = _mm_or_si64(mm2, col6l); \
+    mm1 = _mm_or_si64(mm1, col7l); \
+    mm1 = _mm_or_si64(mm1, mm2); \
+    \
+    if (test_m64_zero(mm1)) { \
+      __m64 dcval, dcvall, dcvalh, row0, row1, row2, row3; \
+      \
+      /* AC terms all zero */ \
+      \
+      quant0l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 0); \
+      \
+      dcval = _mm_mullo_pi16(col0l, quant0l); \
+      dcval = _mm_slli_pi16(dcval, PASS1_BITS);  /* dcval=(00 10 20 30) */ \
+      \
+      dcvall = _mm_unpacklo_pi16(dcval, dcval);  /* dcvall=(00 00 10 10) */ \
+      dcvalh = _mm_unpackhi_pi16(dcval, dcval);  /* dcvalh=(20 20 30 30) */ \
+      \
+      row0 = _mm_unpacklo_pi32(dcvall, dcvall);  /* row0=(00 00 00 00) */ \
+      row1 = _mm_unpackhi_pi32(dcvall, dcvall);  /* row1=(10 10 10 10) */ \
+      row2 = _mm_unpacklo_pi32(dcvalh, dcvalh);  /* row2=(20 20 20 20) */ \
+      row3 = _mm_unpackhi_pi32(dcvalh, dcvalh);  /* row3=(30 30 30 30) */ \
+      \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 0, row0); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 0 + 4, row0); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 1, row1); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 1 + 4, row1); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 2, row2); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 2 + 4, row2); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 3, row3); \
+      _mm_store_si64((__m64 *)&wsptrDCTSIZE * 3 + 4, row3); \
+      \
+      goto nextcolumn##iter; \
+    } \
+  } \
+  \
+  /* Even part \
+   * \
+   * (Original) \
+   * z1 = (z2 + z3) * 0.541196100; \
+   * tmp2 = z1 + z3 * -1.847759065; \
+   * tmp3 = z1 + z2 * 0.765366865; \
+   * \
+   * (This implementation) \
+   * tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065); \
+   * tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100; \
+   */ \
+  \
+  col0l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 0);  /* (00 10 20 30) */ \
+  col2l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 2);  /* (02 12 22 32) */ \
+  col4l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 4);  /* (04 14 24 34) */ \
+  col6l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 6);  /* (06 16 26 36) */ \
+  \
+  quant0l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 0); \
+  quant2l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 2); \
+  quant4l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 4); \
+  quant6l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 6); \
+  \
+  z2 = _mm_mullo_pi16(col2l, quant2l); \
+  z3 = _mm_mullo_pi16(col6l, quant6l); \
+  \
+  z23l = _mm_unpacklo_pi16(z2, z3); \
+  z23h = _mm_unpackhi_pi16(z2, z3); \
+  tmp3l = _mm_madd_pi16(z23l, PW_F130_F054); \
+  tmp3h = _mm_madd_pi16(z23h, PW_F130_F054); \
+  tmp2l = _mm_madd_pi16(z23l, PW_F054_MF130); \
+  tmp2h = _mm_madd_pi16(z23h, PW_F054_MF130); \
+  \
+  z2 = _mm_mullo_pi16(col0l, quant0l); \
+  z3 = _mm_mullo_pi16(col4l, quant4l); \
+  \
+  z23 = _mm_add_pi16(z2, z3); \
+  tmp0l = _mm_loadlo_pi16_f(z23); \
+  tmp0h = _mm_loadhi_pi16_f(z23); \
+  tmp0l = _mm_srai_pi32(tmp0l, (16 - CONST_BITS)); \
+  tmp0h = _mm_srai_pi32(tmp0h, (16 - CONST_BITS)); \
+  \
+  tmp10l = _mm_add_pi32(tmp0l, tmp3l); \
+  tmp10h = _mm_add_pi32(tmp0h, tmp3h); \
+  tmp13l = _mm_sub_pi32(tmp0l, tmp3l); \
+  tmp13h = _mm_sub_pi32(tmp0h, tmp3h); \
+  \
+  z23 = _mm_sub_pi16(z2, z3); \
+  tmp1l = _mm_loadlo_pi16_f(z23); \
+  tmp1h = _mm_loadhi_pi16_f(z23); \
+  tmp1l = _mm_srai_pi32(tmp1l, (16 - CONST_BITS)); \
+  tmp1h = _mm_srai_pi32(tmp1h, (16 - CONST_BITS)); \
+  \
+  tmp11l = _mm_add_pi32(tmp1l, tmp2l); \
+  tmp11h = _mm_add_pi32(tmp1h, tmp2h); \
+  tmp12l = _mm_sub_pi32(tmp1l, tmp2l); \
+  tmp12h = _mm_sub_pi32(tmp1h, tmp2h); \
+  \
+  /* Odd part */ \
+  \
+  col1l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 1);  /* (01 11 21 31) */ \
+  col3l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 3);  /* (03 13 23 33) */ \
+  col5l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 5);  /* (05 15 25 35) */ \
+  col7l = _mm_load_si64((__m64 *)&inptrDCTSIZE * 7);  /* (07 17 27 37) */ \
+  \
+  quant1l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 1); \
+  quant3l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 3); \
+  quant5l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 5); \
+  quant7l = _mm_load_si64((__m64 *)&quantptrDCTSIZE * 7); \
+  \
+  tmp0 = _mm_mullo_pi16(col7l, quant7l); \
+  tmp1 = _mm_mullo_pi16(col5l, quant5l); \
+  tmp2 = _mm_mullo_pi16(col3l, quant3l); \
+  tmp3 = _mm_mullo_pi16(col1l, quant1l); \
+  \
+  DO_IDCT_COMMON(1) \
+  \
+  /* out0=(00 10 20 30), out1=(01 11 21 31) */ \
+  /* out2=(02 12 22 32), out3=(03 13 23 33) */ \
+  /* out4=(04 14 24 34), out5=(05 15 25 35) */ \
+  /* out6=(06 16 26 36), out7=(07 17 27 37) */ \
+  \
+  /* Transpose coefficients */ \
+  \
+  row01a = _mm_unpacklo_pi16(out0, out1);     /* row01a=(00 01 10 11) */ \
+  row23a = _mm_unpackhi_pi16(out0, out1);     /* row23a=(20 21 30 31) */ \
+  row01d = _mm_unpacklo_pi16(out6, out7);     /* row01d=(06 07 16 17) */ \
+  row23d = _mm_unpackhi_pi16(out6, out7);     /* row23d=(26 27 36 37) */ \
+  \
+  row01b = _mm_unpacklo_pi16(out2, out3);     /* row01b=(02 03 12 13) */ \
+  row23b = _mm_unpackhi_pi16(out2, out3);     /* row23b=(22 23 32 33) */ \
+  row01c = _mm_unpacklo_pi16(out4, out5);     /* row01c=(04 05 14 15) */ \
+  row23c = _mm_unpackhi_pi16(out4, out5);     /* row23c=(24 25 34 35) */ \
+  \
+  row0l = _mm_unpacklo_pi32(row01a, row01b);  /* row0l=(00 01 02 03) */ \
+  row1l = _mm_unpackhi_pi32(row01a, row01b);  /* row1l=(10 11 12 13) */ \
+  row2l = _mm_unpacklo_pi32(row23a, row23b);  /* row2l=(20 21 22 23) */ \
+  row3l = _mm_unpackhi_pi32(row23a, row23b);  /* row3l=(30 31 32 33) */ \
+  \
+  row0h = _mm_unpacklo_pi32(row01c, row01d);  /* row0h=(04 05 06 07) */ \
+  row1h = _mm_unpackhi_pi32(row01c, row01d);  /* row1h=(14 15 16 17) */ \
+  row2h = _mm_unpacklo_pi32(row23c, row23d);  /* row2h=(24 25 26 27) */ \
+  row3h = _mm_unpackhi_pi32(row23c, row23d);  /* row3h=(34 35 36 37) */ \
+  \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 0, row0l); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 0 + 4, row0h); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 1, row1l); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 1 + 4, row1h); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 2, row2l); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 2 + 4, row2h); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 3, row3l); \
+  _mm_store_si64((__m64 *)&wsptrDCTSIZE * 3 + 4, row3h); \
+}
+
+#define DO_IDCT_PASS2(ctr) { \
+  __m64 row0l, row1l, row2l, row3l, row4l, row5l, row6l, row7l; \
+  __m64 z23, z23l, z23h; \
+  __m64 col0123a, col0123b, col0123c, col0123d; \
+  __m64 col01l, col01h, col23l, col23h, row06, row17, row24, row35; \
+  __m64 col0, col1, col2, col3; \
+  __m64 tmp0l, tmp0h, tmp1l, tmp1h, tmp2l, tmp2h, tmp3l, tmp3h; \
+  __m64 tmp10l, tmp10h, tmp11l, tmp11h, tmp12l, tmp12h, tmp13l, tmp13h; \
+  \
+  row0l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 0);  /* (00 01 02 03) */ \
+  row1l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 1);  /* (10 11 12 13) */ \
+  row2l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 2);  /* (20 21 22 23) */ \
+  row3l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 3);  /* (30 31 32 33) */ \
+  row4l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 4);  /* (40 41 42 43) */ \
+  row5l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 5);  /* (50 51 52 53) */ \
+  row6l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 6);  /* (60 61 62 63) */ \
+  row7l = _mm_load_si64((__m64 *)&wsptrDCTSIZE * 7);  /* (70 71 72 73) */ \
+  \
+  /* Even part \
+   * \
+   * (Original) \
+   * z1 = (z2 + z3) * 0.541196100; \
+   * tmp2 = z1 + z3 * -1.847759065; \
+   * tmp3 = z1 + z2 * 0.765366865; \
+   * \
+   * (This implementation) \
+   * tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065); \
+   * tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100; \
+   */ \
+  \
+  z23l = _mm_unpacklo_pi16(row2l, row6l); \
+  z23h = _mm_unpackhi_pi16(row2l, row6l); \
+  \
+  tmp3l = _mm_madd_pi16(z23l, PW_F130_F054); \
+  tmp3h = _mm_madd_pi16(z23h, PW_F130_F054); \
+  tmp2l = _mm_madd_pi16(z23l, PW_F054_MF130); \
+  tmp2h = _mm_madd_pi16(z23h, PW_F054_MF130); \
+  \
+  z23 = _mm_add_pi16(row0l, row4l); \
+  tmp0l = _mm_loadlo_pi16_f(z23); \
+  tmp0h = _mm_loadhi_pi16_f(z23); \
+  tmp0l = _mm_srai_pi32(tmp0l, (16 - CONST_BITS)); \
+  tmp0h = _mm_srai_pi32(tmp0h, (16 - CONST_BITS)); \
+  \
+  tmp10l = _mm_add_pi32(tmp0l, tmp3l); \
+  tmp10h = _mm_add_pi32(tmp0h, tmp3h); \
+  tmp13l = _mm_sub_pi32(tmp0l, tmp3l); \
+  tmp13h = _mm_sub_pi32(tmp0h, tmp3h); \
+  \
+  z23 = _mm_sub_pi16(row0l, row4l); \
+  tmp1l = _mm_loadlo_pi16_f(z23); \
+  tmp1h = _mm_loadhi_pi16_f(z23); \
+  tmp1l = _mm_srai_pi32(tmp1l, (16 - CONST_BITS)); \
+  tmp1h = _mm_srai_pi32(tmp1h, (16 - CONST_BITS)); \
+  \
+  tmp11l = _mm_add_pi32(tmp1l, tmp2l); \
+  tmp11h = _mm_add_pi32(tmp1h, tmp2h); \
+  tmp12l = _mm_sub_pi32(tmp1l, tmp2l); \
+  tmp12h = _mm_sub_pi32(tmp1h, tmp2h); \
+  \
+  /* Odd part */ \
+  \
+  tmp0 = row7l; \
+  tmp1 = row5l; \
+  tmp2 = row3l; \
+  tmp3 = row1l; \
+  \
+  DO_IDCT_COMMON(2) \
+  \
+  /* out0=(00 01 02 03), out1=(10 11 12 13) */ \
+  /* out2=(20 21 22 23), out3=(30 31 32 33) */ \
+  /* out4=(40 41 42 43), out5=(50 51 52 53) */ \
+  /* out6=(60 61 62 63), out7=(70 71 72 73) */ \
+  \
+  row06 = _mm_packs_pi16(out0, out6);  /* row06=(00 01 02 03 60 61 62 63) */ \
+  row17 = _mm_packs_pi16(out1, out7);  /* row17=(10 11 12 13 70 71 72 73) */ \
+  row24 = _mm_packs_pi16(out2, out4);  /* row24=(20 21 22 23 40 41 42 43) */ \
+  row35 = _mm_packs_pi16(out3, out5);  /* row35=(30 31 32 33 50 51 52 53) */ \
+  \
+  row06 = _mm_add_pi8(row06, PB_CENTERJSAMP); \
+  row17 = _mm_add_pi8(row17, PB_CENTERJSAMP); \
+  row24 = _mm_add_pi8(row24, PB_CENTERJSAMP); \
+  row35 = _mm_add_pi8(row35, PB_CENTERJSAMP); \
+  \
+  /* Transpose coefficients */ \
+  \
+  col0123a = _mm_unpacklo_pi8(row06, row17);  /* col0123a=(00 10 01 11 02 12 03 13) */ \
+  col0123d = _mm_unpackhi_pi8(row06, row17);  /* col0123d=(60 70 61 71 62 72 63 73) */ \
+  col0123b = _mm_unpacklo_pi8(row24, row35);  /* col0123b=(20 30 21 31 22 32 23 33) */ \
+  col0123c = _mm_unpackhi_pi8(row24, row35);  /* col0123c=(40 50 41 51 42 52 43 53) */ \
+  \
+  col01l = _mm_unpacklo_pi16(col0123a, col0123b);  /* col01l=(00 10 20 30 01 11 21 31) */ \
+  col23l = _mm_unpackhi_pi16(col0123a, col0123b);  /* col23l=(02 12 22 32 03 13 23 33) */ \
+  col01h = _mm_unpacklo_pi16(col0123c, col0123d);  /* col01h=(40 50 60 70 41 51 61 71) */ \
+  col23h = _mm_unpackhi_pi16(col0123c, col0123d);  /* col23h=(42 52 62 72 43 53 63 73) */ \
+  \
+  col0 = _mm_unpacklo_pi32(col01l, col01h);   /* col0=(00 10 20 30 40 50 60 70) */ \
+  col1 = _mm_unpackhi_pi32(col01l, col01h);   /* col1=(01 11 21 31 41 51 61 71) */ \
+  col2 = _mm_unpacklo_pi32(col23l, col23h);   /* col2=(02 12 22 32 42 52 62 72) */ \
+  col3 = _mm_unpackhi_pi32(col23l, col23h);   /* col3=(03 13 23 33 43 53 63 73) */ \
+  \
+  _mm_store_si64((__m64 *)(output_bufctr + 0 + output_col), col0); \
+  _mm_store_si64((__m64 *)(output_bufctr + 1 + output_col), col1); \
+  _mm_store_si64((__m64 *)(output_bufctr + 2 + output_col), col2); \
+  _mm_store_si64((__m64 *)(output_bufctr + 3 + output_col), col3); \
+}
+
+void jsimd_idct_islow_mmi(void *dct_table, JCOEFPTR coef_block,
+                          JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  __m64 tmp0, tmp1, tmp2, tmp3;
+  __m64 out0, out1, out2, out3, out4, out5, out6, out7;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  JCOEF *wsptr;
+  JCOEF workspaceDCTSIZE2;  /* buffers data between passes */
+
+  /* Pass 1: process columns. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *)dct_table;
+  wsptr = workspace;
+
+  DO_IDCT_PASS1(1)
+nextcolumn1:
+  inptr += 4;
+  quantptr += 4;
+  wsptr += DCTSIZE * 4;
+  DO_IDCT_PASS1(2)
+nextcolumn2:
+
+  /* Pass 2: process rows. */
+
+  wsptr = workspace;
+
+  DO_IDCT_PASS2(0)
+  wsptr += 4;
+  DO_IDCT_PASS2(4)
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jquanti-mmi.c Added

@@ -0,0 +1,124 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2016-2017, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ * Copyright (C) 2018-2019, D. R. Commander.  All Rights Reserved.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* INTEGER QUANTIZATION AND SAMPLE CONVERSION */
+
+#include "jsimd_mmi.h"
+
+
+#define DO_QUANT() { \
+  __m64 rowl, rowh, rowls, rowhs, rowlsave, rowhsave; \
+  __m64 corrl, corrh, recipl, reciph, scalel, scaleh; \
+  \
+  rowl = _mm_load_si64((__m64 *)&workspace0); \
+  rowh = _mm_load_si64((__m64 *)&workspace4); \
+  \
+  /* Branch-less absolute value */ \
+  rowls = _mm_srai_pi16(rowl, (WORD_BIT - 1));  /* -1 if value < 0, */ \
+                                                /* 0 otherwise */ \
+  rowhs = _mm_srai_pi16(rowh, (WORD_BIT - 1)); \
+  \
+  rowl = _mm_xor_si64(rowl, rowls);           /* val = -val */ \
+  rowh = _mm_xor_si64(rowh, rowhs); \
+  rowl = _mm_sub_pi16(rowl, rowls); \
+  rowh = _mm_sub_pi16(rowh, rowhs); \
+  \
+  corrl = _mm_load_si64((__m64 *)&divisorsDCTSIZE2 * 1);  /* correction */ \
+  corrh = _mm_load_si64((__m64 *)&divisorsDCTSIZE2 * 1 + 4); \
+  \
+  rowlsave = rowl = _mm_add_pi16(rowl, corrl);  /* correction + roundfactor */ \
+  rowhsave = rowh = _mm_add_pi16(rowh, corrh); \
+  \
+  recipl = _mm_load_si64((__m64 *)&divisorsDCTSIZE2 * 0);  /* reciprocal */ \
+  reciph = _mm_load_si64((__m64 *)&divisorsDCTSIZE2 * 0 + 4); \
+  \
+  rowl = _mm_mulhi_pi16(rowl, recipl); \
+  rowh = _mm_mulhi_pi16(rowh, reciph); \
+  \
+  /* reciprocal is always negative (MSB=1), so we always need to add the */ \
+  /* initial value (input value is never negative as we inverted it at the */ \
+  /* start of this routine) */ \
+  rowlsave = rowl = _mm_add_pi16(rowl, rowlsave); \
+  rowhsave = rowh = _mm_add_pi16(rowh, rowhsave); \
+  \
+  scalel = _mm_load_si64((__m64 *)&divisorsDCTSIZE2 * 2);  /* scale */ \
+  scaleh = _mm_load_si64((__m64 *)&divisorsDCTSIZE2 * 2 + 4); \
+  \
+  rowl = _mm_mulhi_pi16(rowl, scalel); \
+  rowh = _mm_mulhi_pi16(rowh, scaleh); \
+  \
+  /* determine if scale is negative */ \
+  scalel = _mm_srai_pi16(scalel, (WORD_BIT - 1)); \
+  scaleh = _mm_srai_pi16(scaleh, (WORD_BIT - 1)); \
+  \
+  /* and add input if it is */ \
+  scalel = _mm_and_si64(scalel, rowlsave); \
+  scaleh = _mm_and_si64(scaleh, rowhsave); \
+  rowl = _mm_add_pi16(rowl, scalel); \
+  rowh = _mm_add_pi16(rowh, scaleh); \
+  \
+  /* then check if negative input */ \
+  rowlsave = _mm_srai_pi16(rowlsave, (WORD_BIT - 1)); \
+  rowhsave = _mm_srai_pi16(rowhsave, (WORD_BIT - 1)); \
+  \
+  /* and add scale if it is */ \
+  rowlsave = _mm_and_si64(rowlsave, scalel); \
+  rowhsave = _mm_and_si64(rowhsave, scaleh); \
+  rowl = _mm_add_pi16(rowl, rowlsave); \
+  rowh = _mm_add_pi16(rowh, rowhsave); \
+  \
+  rowl = _mm_xor_si64(rowl, rowls);           /* val = -val */ \
+  rowh = _mm_xor_si64(rowh, rowhs); \
+  rowl = _mm_sub_pi16(rowl, rowls); \
+  rowh = _mm_sub_pi16(rowh, rowhs); \
+  \
+  _mm_store_si64((__m64 *)&output_ptr0, rowl); \
+  _mm_store_si64((__m64 *)&output_ptr4, rowh); \
+  \
+  workspace += DCTSIZE; \
+  divisors += DCTSIZE; \
+  output_ptr += DCTSIZE; \
+}
+
+
+void jsimd_quantize_mmi(JCOEFPTR coef_block, DCTELEM *divisors,
+                        DCTELEM *workspace)
+{
+  JCOEFPTR output_ptr = coef_block;
+
+  DO_QUANT()
+  DO_QUANT()
+  DO_QUANT()
+  DO_QUANT()
+  DO_QUANT()
+  DO_QUANT()
+  DO_QUANT()
+  DO_QUANT()
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jsimd.c Added

@@ -0,0 +1,870 @@
+/*
+ * jsimd_mips64.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2009-2011, 2014, 2016, 2018, D. R. Commander.
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California.
+ * Copyright (C) 2015, 2018, Matthieu Darbois.
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 64-bit MIPS architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+
+static unsigned int simd_support = ~0;
+
+#if defined(__linux__)
+
+#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT  (1024 * 1024)
+
+LOCAL(int)
+check_feature(char *buffer, char *feature)
+{
+  char *p;
+
+  if (*feature == 0)
+    return 0;
+  if (strncmp(buffer, "ASEs implemented", 16) != 0)
+    return 0;
+  buffer += 16;
+  while (isspace(*buffer))
+    buffer++;
+
+  /* Check if 'feature' is present in the buffer as a separate word */
+  while ((p = strstr(buffer, feature))) {
+    if (p > buffer && !isspace(*(p - 1))) {
+      buffer++;
+      continue;
+    }
+    p += strlen(feature);
+    if (*p != 0 && !isspace(*p)) {
+      buffer++;
+      continue;
+    }
+    return 1;
+  }
+  return 0;
+}
+
+LOCAL(int)
+parse_proc_cpuinfo(int bufsize)
+{
+  char *buffer = (char *)malloc(bufsize);
+  FILE *fd;
+
+  simd_support = 0;
+
+  if (!buffer)
+    return 0;
+
+  fd = fopen("/proc/cpuinfo", "r");
+  if (fd) {
+    while (fgets(buffer, bufsize, fd)) {
+      if (!strchr(buffer, '\n') && !feof(fd)) {
+        /* "impossible" happened - insufficient size of the buffer! */
+        fclose(fd);
+        free(buffer);
+        return 0;
+      }
+      if (check_feature(buffer, "loongson-mmi"))
+        simd_support |= JSIMD_MMI;
+    }
+    fclose(fd);
+  }
+  free(buffer);
+  return 1;
+}
+
+#endif
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd(void)
+{
+#ifndef NO_GETENV
+  char *env = NULL;
+#endif
+#if defined(__linux__)
+  int bufsize = 1024; /* an initial guess for the line buffer size limit */
+#endif
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = 0;
+
+#if defined(__linux__)
+  while (!parse_proc_cpuinfo(bufsize)) {
+    bufsize *= 2;
+    if (bufsize > SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT)
+      break;
+  }
+#elif defined(__mips_loongson_vector_rev)
+  /* Only enable MMI by default on non-Linux platforms when the compiler flags
+   * support it. */
+  simd_support |= JSIMD_MMI;
+#endif
+
+#ifndef NO_GETENV
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCEMMI");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = JSIMD_MMI;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+#endif
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_c_can_null_convert(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                      JSAMPIMAGE output_buf, JDIMENSION output_row,
+                      int num_rows)
+{
+  void (*mmifct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    mmifct = jsimd_extrgb_ycc_convert_mmi;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    mmifct = jsimd_extrgbx_ycc_convert_mmi;
+    break;
+  case JCS_EXT_BGR:
+    mmifct = jsimd_extbgr_ycc_convert_mmi;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    mmifct = jsimd_extbgrx_ycc_convert_mmi;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    mmifct = jsimd_extxbgr_ycc_convert_mmi;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    mmifct = jsimd_extxrgb_ycc_convert_mmi;
+    break;
+  default:
+    mmifct = jsimd_rgb_ycc_convert_mmi;
+    break;
+  }
+
+  mmifct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                       JSAMPIMAGE output_buf, JDIMENSION output_row,
+                       int num_rows)
+{
+  void (*mmifct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    mmifct = jsimd_extrgb_gray_convert_mmi;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    mmifct = jsimd_extrgbx_gray_convert_mmi;
+    break;
+  case JCS_EXT_BGR:
+    mmifct = jsimd_extbgr_gray_convert_mmi;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    mmifct = jsimd_extbgrx_gray_convert_mmi;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    mmifct = jsimd_extxbgr_gray_convert_mmi;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    mmifct = jsimd_extxrgb_gray_convert_mmi;
+    break;
+  default:
+    mmifct = jsimd_rgb_gray_convert_mmi;
+    break;
+  }
+
+  mmifct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION input_row, JSAMPARRAY output_buf,
+                      int num_rows)
+{
+  void (*mmifct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    mmifct = jsimd_ycc_extrgb_convert_mmi;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    mmifct = jsimd_ycc_extrgbx_convert_mmi;
+    break;
+  case JCS_EXT_BGR:
+    mmifct = jsimd_ycc_extbgr_convert_mmi;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    mmifct = jsimd_ycc_extbgrx_convert_mmi;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    mmifct = jsimd_ycc_extxbgr_convert_mmi;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    mmifct = jsimd_ycc_extxrgb_convert_mmi;
+    break;
+  default:
+    mmifct = jsimd_ycc_rgb_convert_mmi;
+    break;
+  }
+
+  mmifct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_c_null_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                     JSAMPIMAGE output_buf, JDIMENSION output_row,
+                     int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_smooth_downsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_mmi(cinfo->image_width, cinfo->max_v_samp_factor,
+                            compptr->v_samp_factor, compptr->width_in_blocks,
+                            input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v2_smooth_downsample(j_compress_ptr cinfo,
+                             jpeg_component_info *compptr,
+                             JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_int_upsample(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_int_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                   JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_fancy_upsample_mmi(cinfo->max_v_samp_factor,
+                                compptr->downsampled_width, input_data,
+                                output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_fancy_upsample_mmi(cinfo->max_v_samp_factor,
+                                compptr->downsampled_width, input_data,
+                                output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*mmifct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    mmifct = jsimd_h2v2_extrgb_merged_upsample_mmi;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    mmifct = jsimd_h2v2_extrgbx_merged_upsample_mmi;
+    break;
+  case JCS_EXT_BGR:
+    mmifct = jsimd_h2v2_extbgr_merged_upsample_mmi;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    mmifct = jsimd_h2v2_extbgrx_merged_upsample_mmi;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    mmifct = jsimd_h2v2_extxbgr_merged_upsample_mmi;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    mmifct = jsimd_h2v2_extxrgb_merged_upsample_mmi;
+    break;
+  default:
+    mmifct = jsimd_h2v2_merged_upsample_mmi;
+    break;
+  }
+
+  mmifct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*mmifct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    mmifct = jsimd_h2v1_extrgb_merged_upsample_mmi;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    mmifct = jsimd_h2v1_extrgbx_merged_upsample_mmi;
+    break;
+  case JCS_EXT_BGR:
+    mmifct = jsimd_h2v1_extbgr_merged_upsample_mmi;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    mmifct = jsimd_h2v1_extbgrx_merged_upsample_mmi;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    mmifct = jsimd_h2v1_extxbgr_merged_upsample_mmi;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    mmifct = jsimd_h2v1_extxrgb_merged_upsample_mmi;
+    break;
+  default:
+    mmifct = jsimd_h2v1_merged_upsample_mmi;
+    break;
+  }
+
+  mmifct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
+               DCTELEM *workspace)
+{
+}
+
+GLOBAL(void)
+jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow(DCTELEM *data)
+{
+  jsimd_fdct_islow_mmi(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast(DCTELEM *data)
+{
+  jsimd_fdct_ifast_mmi(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float(FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
+{
+  jsimd_quantize_mmi(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_6x6(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_12x12(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_6x6(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_12x12(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MMI)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_islow_mmi(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_ifast_mmi(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block(void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET *)
+jsimd_huff_encode_one_block(void *state, JOCTET *buffer, JCOEFPTR block,
+                            int last_dc_val, c_derived_tbl *dctbl,
+                            c_derived_tbl *actbl)
+{
+  return NULL;
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_first_prepare(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_encode_mcu_AC_first_prepare(const JCOEF *block,
+                                  const int *jpeg_natural_order_start, int Sl,
+                                  int Al, JCOEF *values, size_t *zerobits)
+{
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_refine_prepare(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block,
+                                   const int *jpeg_natural_order_start, int Sl,
+                                   int Al, JCOEF *absvalues, size_t *bits)
+{
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/jsimd_mmi.h Added

@@ -0,0 +1,69 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Authors:  ZhuChen     <zhuchen@loongson.cn>
+ *           CaiWanwei   <caiwanwei@loongson.cn>
+ *           SunZhangzhi <sunzhangzhi-cq@loongson.cn>
+ *           QingfaLiu   <liuqingfa-hf@loongson.cn>
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jdct.h"
+#include "loongson-mmintrin.h"
+
+
+/* Common code */
+#if defined(_ABI64) && _MIPS_SIM == _ABI64
+# define PTR_ADDU  "daddu "
+# define PTR_SLL   "dsll "
+#else
+# define PTR_ADDU  "addu "
+# define PTR_SLL   "sll "
+#endif
+
+#define SIZEOF_MMWORD  8
+#define BYTE_BIT  8
+#define WORD_BIT  16
+#define SCALEBITS  16
+
+#define _uint64_set_pi8(a, b, c, d, e, f, g, h) \
+  (((uint64_t)(uint8_t)a << 56) | \
+   ((uint64_t)(uint8_t)b << 48) | \
+   ((uint64_t)(uint8_t)c << 40) | \
+   ((uint64_t)(uint8_t)d << 32) | \
+   ((uint64_t)(uint8_t)e << 24) | \
+   ((uint64_t)(uint8_t)f << 16) | \
+   ((uint64_t)(uint8_t)g << 8)  | \
+   ((uint64_t)(uint8_t)h))
+#define _uint64_set1_pi8(a)  _uint64_set_pi8(a, a, a, a, a, a, a, a)
+#define _uint64_set_pi16(a, b, c, d) \
+  (((uint64_t)(uint16_t)a << 48) | \
+   ((uint64_t)(uint16_t)b << 32) | \
+   ((uint64_t)(uint16_t)c << 16) | \
+   ((uint64_t)(uint16_t)d))
+#define _uint64_set1_pi16(a)  _uint64_set_pi16(a, a, a, a)
+#define _uint64_set_pi32(a, b) \
+  (((uint64_t)(uint32_t)a << 32) | \
+   ((uint64_t)(uint32_t)b))
+
+#define get_const_value(index)  (*(__m64 *)&const_valueindex)

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/mips64/loongson-mmintrin.h Added

@@ -0,0 +1,1334 @@
+/*
+ * Loongson MMI optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2016-2018, Loongson Technology Corporation Limited, BeiJing.
+ *                          All Rights Reserved.
+ * Copyright (C) 2019, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#ifndef __LOONGSON_MMINTRIN_H__
+#define __LOONGSON_MMINTRIN_H__
+
+#include <stdint.h>
+
+
+#define FUNCTION_ATTRIBS \
+  __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+
+
+/* Vectors are stored in 64-bit floating-point registers. */
+typedef double __m64;
+
+/* Having a 32-bit datatype allows us to use 32-bit loads in places like
+   load8888. */
+typedef float __m32;
+
+
+/********** Set Operations **********/
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_setzero_si64(void)
+{
+  return 0.0;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_set_pi8(uint8_t __b7, uint8_t __b6, uint8_t __b5, uint8_t __b4,
+            uint8_t __b3, uint8_t __b2, uint8_t __b1, uint8_t __b0)
+{
+  __m64 ret;
+  uint32_t lo = ((uint32_t)__b6 << 24) |
+                ((uint32_t)__b4 << 16) |
+                ((uint32_t)__b2 << 8) |
+                (uint32_t)__b0;
+  uint32_t hi = ((uint32_t)__b7 << 24) |
+                ((uint32_t)__b5 << 16) |
+                ((uint32_t)__b3 << 8) |
+                (uint32_t)__b1;
+
+  asm("mtc1      %1, %0\n\t"
+      "mtc1      %2, $f0\n\t"
+      "punpcklbh %0, %0, $f0\n\t"
+      : "=f" (ret)
+      : "r" (lo), "r" (hi)
+      : "$f0"
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_set_pi16(uint16_t __h3, uint16_t __h2, uint16_t __h1, uint16_t __h0)
+{
+  __m64 ret;
+  uint32_t lo = ((uint32_t)__h2 << 16) | (uint32_t)__h0;
+  uint32_t hi = ((uint32_t)__h3 << 16) | (uint32_t)__h1;
+
+  asm("mtc1      %1, %0\n\t"
+      "mtc1      %2, $f0\n\t"
+      "punpcklhw %0, %0, $f0\n\t"
+      : "=f" (ret)
+      : "r" (lo), "r" (hi)
+      : "$f0"
+     );
+
+  return ret;
+}
+
+#define _MM_SHUFFLE(fp3, fp2, fp1, fp0) \
+  (((fp3) << 6) | ((fp2) << 4) | ((fp1) << 2) | (fp0))
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_set_pi32(uint32_t __i1, uint32_t __i0)
+{
+  if (__builtin_constant_p(__i1) && __builtin_constant_p(__i0)) {
+    uint64_t val = ((uint64_t)__i1 << 32) |
+                   ((uint64_t)__i0 <<  0);
+
+    return *(__m64 *)&val;
+  } else if (__i1 == __i0) {
+    uint64_t imm = _MM_SHUFFLE(1, 0, 1, 0);
+    __m64 ret;
+
+    asm("pshufh %0, %1, %2\n\t"
+        : "=f" (ret)
+        : "f" (*(__m32 *)&__i1), "f" (*(__m64 *)&imm)
+       );
+
+    return ret;
+  } else {
+    uint64_t val = ((uint64_t)__i1 << 32) |
+                   ((uint64_t)__i0 <<  0);
+
+    return *(__m64 *)&val;
+  }
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_set1_pi8(uint8_t __b0)
+{
+  __m64 ret;
+
+  asm("sll    $8, %1, 8\n\t"
+      "or     %1, %1, $8\n\t"
+      "mtc1   %1, %0\n\t"
+      "mtc1   $0, $f0\n\t"
+      "pshufh %0, %0, $f0\n\t"
+      : "=f" (ret)
+      : "r" (__b0)
+      : "$8", "$f0"
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_set1_pi16(uint16_t __h0)
+{
+  __m64 ret;
+
+  asm("mtc1   %1, %0\n\t"
+      "mtc1   $0, $f0\n\t"
+      "pshufh %0, %0, $f0\n\t"
+      : "=f" (ret)
+      : "r" (__h0)
+      : "$8", "$f0"
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_set1_pi32(unsigned __i0)
+{
+  return _mm_set_pi32(__i0, __i0);
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_setr_pi8(uint8_t __h0, uint8_t __h1, uint8_t __h2, uint8_t __h3,
+             uint8_t __h4, uint8_t __h5, uint8_t __h6, uint8_t __h7)
+{
+  return _mm_set_pi8(__h7, __h6, __h5, __h4,
+                     __h3, __h2, __h1, __h0);
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_setr_pi16(uint16_t __w0, uint16_t __w1, uint16_t __w2, uint16_t __w3)
+{
+  return _mm_set_pi16(__w3, __w2, __w1, __w0);
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_setr_pi32(uint32_t __i0, uint32_t __i1)
+{
+  return _mm_set_pi32(__i1, __i0);
+}
+
+
+/********** Arithmetic Operations **********/
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_add_pi8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("paddb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_add_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("paddh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_add_pi32(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("paddw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_add_si64(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("paddd %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_adds_pi8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("paddsb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_adds_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("paddsh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_adds_pu8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("paddusb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_adds_pu16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("paddush %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_avg_pu8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pavgb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_avg_pu16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pavgh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_madd_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pmaddhw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_max_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pmaxsh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_max_pu8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pmaxub %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_min_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pminsh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_min_pu8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pminub %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline int FUNCTION_ATTRIBS
+_mm_movemask_pi8(__m64 __m1)
+{
+  int ret;
+
+  asm("pmovmskb %0, %1\n\t"
+      : "=r" (ret)
+      : "y" (__m1)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_mulhi_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pmulhh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_mulhi_pu16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pmulhuh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_mullo_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pmullh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_mul_pu32(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pmuluw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_sad_pu8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("psadbh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_asub_pu8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pasubub %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_biadd_pu8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("biadd %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_sub_pi8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("psubb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_sub_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("psubh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_sub_pi32(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("psubw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_sub_si64(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("psubd %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_subs_pi8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("psubsb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_subs_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("psubsh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_subs_pu8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("psubusb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_subs_pu16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("psubush %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+
+/********** Logical Operations **********/
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_and_si64(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("and %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_andnot_si64(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("andn %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_or_si32(__m32 __m1, __m32 __m2)
+{
+  __m32 ret;
+
+  asm("or %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_or_si64(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("or %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_xor_si64(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("xor %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+
+/********** Shift Operations **********/
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_slli_pi16(__m64 __m, int64_t __count)
+{
+  __m64 ret;
+
+  asm("psllh  %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__count)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_slli_pi32(__m64 __m, int64_t __count)
+{
+  __m64 ret;
+
+  asm("psllw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__count)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_slli_si64(__m64 __m, int64_t __count)
+{
+  __m64 ret;
+
+  asm("dsll  %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__count)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_srli_pi16(__m64 __m, int64_t __count)
+{
+  __m64 ret;
+
+  asm("psrlh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__count)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_srli_pi32(__m64 __m, int64_t __count)
+{
+  __m64 ret;
+
+  asm("psrlw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__count)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_srli_si64(__m64 __m, int64_t __count)
+{
+  __m64 ret;
+
+  asm("dsrl  %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__count)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_srai_pi16(__m64 __m, int64_t __count)
+{
+  __m64 ret;
+
+  asm("psrah %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__count)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_srai_pi32(__m64 __m, int64_t __count)
+{
+  __m64 ret;
+
+  asm("psraw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__count)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_srai_si64(__m64 __m, int64_t __count)
+{
+  __m64 ret;
+
+  asm("dsra %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__count)
+     );
+
+  return ret;
+}
+
+
+/********** Conversion Intrinsics **********/
+
+extern __inline __m64 FUNCTION_ATTRIBS
+to_m64(uint64_t x)
+{
+  return *(__m64 *)&x;
+}
+
+extern __inline uint64_t FUNCTION_ATTRIBS
+to_uint64(__m64 x)
+{
+  return *(uint64_t *)&x;
+}
+
+
+/********** Comparison Intrinsics **********/
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_cmpeq_pi8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pcmpeqb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_cmpeq_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pcmpeqh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_cmpeq_pi32(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pcmpeqw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_cmpgt_pi8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pcmpgtb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_cmpgt_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pcmpgth %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_cmpgt_pi32(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pcmpgtw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_cmplt_pi8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pcmpltb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_cmplt_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pcmplth %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_cmplt_pi32(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("pcmpltw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+
+/********** Miscellaneous Operations **********/
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_packs_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("packsshb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_packs_pi32(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("packsswh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_packs_pi32_f(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("packsswh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_packs_pu16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("packushb %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_extract_pi16(__m64 __m, int64_t __pos)
+{
+  __m64 ret;
+
+  asm("pextrh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__pos)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_insert_pi16(__m64 __m1, __m64 __m2, int64_t __pos)
+{
+  __m64 ret;
+
+  switch (__pos) {
+  case 0:
+
+    asm("pinsrh_0 %0, %1, %2\n\t"
+        : "=f" (ret)
+        : "f" (__m1), "f" (__m2), "i" (__pos)
+       );
+
+    break;
+
+  case 1:
+
+    asm("pinsrh_1 %0, %1, %2\n\t"
+        : "=f" (ret)
+        : "f" (__m1), "f" (__m2), "i" (__pos)
+       );
+
+    break;
+  case 2:
+
+    asm("pinsrh_2 %0, %1, %2\n\t"
+        : "=f" (ret)
+        : "f" (__m1), "f" (__m2), "i" (__pos)
+       );
+
+    break;
+
+  case 3:
+
+    asm("pinsrh_3 %0, %1, %2\n\t"
+        : "=f" (ret)
+        : "f" (__m1), "f" (__m2), "i" (__pos)
+       );
+
+    break;
+  }
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_shuffle_pi16(__m64 __m, int64_t __n)
+{
+  __m64 ret;
+
+  asm("pshufh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m), "f" (*(__m64 *)&__n)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpackhi_pi8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpckhbh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpackhi_pi8_f(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpckhbh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpackhi_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpckhhw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpackhi_pi16_f(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpckhhw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpackhi_pi32(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpckhwd %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpacklo_pi8(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpcklbh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+/* Since punpcklbh cares about the high 32-bits, we use the __m64 datatype,
+   which preserves the data. */
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpacklo_pi8_f64(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpcklbh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+/* Since punpcklbh doesn't care about the high 32-bits, we use the __m32,
+   datatype, which allows load8888 to use 32-bit loads. */
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpacklo_pi8_f(__m32 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpcklbh %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpacklo_pi16(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpcklhw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpacklo_pi16_f(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpcklhw %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpacklo_pi32(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpcklwd %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_unpacklo_pi32_f(__m64 __m1, __m64 __m2)
+{
+  __m64 ret;
+
+  asm("punpcklwd %0, %1, %2\n\t"
+      : "=f" (ret)
+      : "f" (__m1), "f" (__m2)
+     );
+
+  return ret;
+}
+
+extern __inline void FUNCTION_ATTRIBS
+_mm_store_pi32(__m32 *dest, __m64 src)
+{
+  src = _mm_packs_pu16(src, _mm_setzero_si64());
+
+  asm("swc1 %1, %0\n\t"
+      : "=m" (*dest)
+      : "f" (src)
+      : "memory"
+     );
+}
+
+extern __inline void FUNCTION_ATTRIBS
+_mm_store_si64(__m64 *dest, __m64 src)
+{
+  asm("sdc1 %1, %0 \n\t"
+      : "=m" (*dest)
+      : "f" (src)
+      : "memory"
+     );
+}
+
+extern __inline void FUNCTION_ATTRIBS
+_mm_storeu_si64(__m64 *dest, __m64 src)
+{
+  asm("gssdlc1 %1, 7(%0) \n\t"
+      "gssdrc1 %1, 0(%0) \n\t"
+      :
+      : "r" (dest), "f" (src)
+      : "memory"
+     );
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_load_si32(const __m32 *src)
+{
+  __m32 ret;
+
+  asm("lwc1 %0, %1\n\t"
+      : "=f" (ret)
+      : "m" (*src)
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_load_si64(const __m64 *src)
+{
+  __m64 ret;
+
+  asm("ldc1 %0, %1\n\t"
+      : "=f" (ret)
+      : "m" (*src)
+      : "memory"
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_loadu_si64(const __m64 *src)
+{
+  __m64 ret;
+
+  asm("gsldlc1 %0,  7(%1)\n\t"
+      "gsldrc1 %0,  0(%1)\n\t"
+      : "=f" (ret)
+      : "r" (src)
+      : "memory"
+     );
+
+  return ret;
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_loadlo_pi8(const uint32_t *src)
+{
+  return _mm_unpacklo_pi8_f(*(__m32 *)src, _mm_setzero_si64());
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_loadlo_pi8_f(__m64 src)
+{
+  return _mm_unpacklo_pi8_f64(src, _mm_setzero_si64());
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_loadhi_pi8_f(__m64 src)
+{
+  return _mm_unpackhi_pi8_f(src, _mm_setzero_si64());
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_loadlo_pi16(__m64 src)
+{
+  return _mm_unpacklo_pi16(src, _mm_setzero_si64());
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_loadlo_pi16_f(__m64 src)
+{
+  return _mm_unpacklo_pi16_f(_mm_setzero_si64(), src);
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_loadhi_pi16(__m64 src)
+{
+  return _mm_unpackhi_pi16(src, _mm_setzero_si64());
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_loadhi_pi16_f(__m64 src)
+{
+  return _mm_unpackhi_pi16_f(_mm_setzero_si64(), src);
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_expand_alpha(__m64 pixel)
+{
+  return _mm_shuffle_pi16(pixel, _MM_SHUFFLE(3, 3, 3, 3));
+}
+
+extern __inline __m64 FUNCTION_ATTRIBS
+_mm_expand_alpha_rev(__m64 pixel)
+{
+  return _mm_shuffle_pi16(pixel, _MM_SHUFFLE(0, 0, 0, 0));
+}
+
+#endif  /* __LOONGSON_MMINTRIN_H__ */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/nasm Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/nasm/jcolsamp.inc Added

@@ -0,0 +1,135 @@
+;
+; jcolsamp.inc - private declarations for color conversion & up/downsampling
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+
+; --------------------------------------------------------------------------
+
+; pseudo-resisters to make ordering of RGB configurable
+;
+%if RGB_RED == 0
+%define mmA  mm0
+%define mmB  mm1
+%define xmmA  xmm0
+%define xmmB  xmm1
+%define ymmA  ymm0
+%define ymmB  ymm1
+%elif RGB_GREEN == 0
+%define mmA  mm2
+%define mmB  mm3
+%define xmmA  xmm2
+%define xmmB  xmm3
+%define ymmA  ymm2
+%define ymmB  ymm3
+%elif RGB_BLUE == 0
+%define mmA  mm4
+%define mmB  mm5
+%define xmmA  xmm4
+%define xmmB  xmm5
+%define ymmA  ymm4
+%define ymmB  ymm5
+%else
+%define mmA  mm6
+%define mmB  mm7
+%define xmmA  xmm6
+%define xmmB  xmm7
+%define ymmA  ymm6
+%define ymmB  ymm7
+%endif
+
+%if RGB_RED == 1
+%define mmC  mm0
+%define mmD  mm1
+%define xmmC  xmm0
+%define xmmD  xmm1
+%define ymmC  ymm0
+%define ymmD  ymm1
+%elif RGB_GREEN == 1
+%define mmC  mm2
+%define mmD  mm3
+%define xmmC  xmm2
+%define xmmD  xmm3
+%define ymmC  ymm2
+%define ymmD  ymm3
+%elif RGB_BLUE == 1
+%define mmC  mm4
+%define mmD  mm5
+%define xmmC  xmm4
+%define xmmD  xmm5
+%define ymmC  ymm4
+%define ymmD  ymm5
+%else
+%define mmC  mm6
+%define mmD  mm7
+%define xmmC  xmm6
+%define xmmD  xmm7
+%define ymmC  ymm6
+%define ymmD  ymm7
+%endif
+
+%if RGB_RED == 2
+%define mmE  mm0
+%define mmF  mm1
+%define xmmE  xmm0
+%define xmmF  xmm1
+%define ymmE  ymm0
+%define ymmF  ymm1
+%elif RGB_GREEN == 2
+%define mmE  mm2
+%define mmF  mm3
+%define xmmE  xmm2
+%define xmmF  xmm3
+%define ymmE  ymm2
+%define ymmF  ymm3
+%elif RGB_BLUE == 2
+%define mmE  mm4
+%define mmF  mm5
+%define xmmE  xmm4
+%define xmmF  xmm5
+%define ymmE  ymm4
+%define ymmF  ymm5
+%else
+%define mmE  mm6
+%define mmF  mm7
+%define xmmE  xmm6
+%define xmmF  xmm7
+%define ymmE  ymm6
+%define ymmF  ymm7
+%endif
+
+%if RGB_RED == 3
+%define mmG  mm0
+%define mmH  mm1
+%define xmmG  xmm0
+%define xmmH  xmm1
+%define ymmG  ymm0
+%define ymmH  ymm1
+%elif RGB_GREEN == 3
+%define mmG  mm2
+%define mmH  mm3
+%define xmmG  xmm2
+%define xmmH  xmm3
+%define ymmG  ymm2
+%define ymmH  ymm3
+%elif RGB_BLUE == 3
+%define mmG  mm4
+%define mmH  mm5
+%define xmmG  xmm4
+%define xmmH  xmm5
+%define ymmG  ymm4
+%define ymmH  ymm5
+%else
+%define mmG  mm6
+%define mmH  mm7
+%define xmmG  xmm6
+%define xmmH  xmm7
+%define ymmG  ymm6
+%define ymmH  ymm7
+%endif
+
+; --------------------------------------------------------------------------

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/nasm/jdct.inc Added

@@ -0,0 +1,31 @@
+;
+; jdct.inc - private declarations for forward & reverse DCT subsystems
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2018, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+
+; Each IDCT routine is responsible for range-limiting its results and
+; converting them to unsigned form (0..MAXJSAMPLE).  The raw outputs could
+; be quite far out of range if the input data is corrupt, so a bulletproof
+; range-limiting step is required.  We use a mask-and-table-lookup method
+; to do the combined operations quickly.
+;
+%define RANGE_MASK  (MAXJSAMPLE * 4 + 3)  ; 2 bits wider than legal samples
+
+%define ROW(n, b, s)  ((b) + (n) * (s))
+%define COL(n, b, s)  ((b) + (n) * (s) * DCTSIZE)
+
+%define DWBLOCK(m, n, b, s) \
+  ((b) + (m) * DCTSIZE * (s) + (n) * SIZEOF_DWORD)
+%define MMBLOCK(m, n, b, s) \
+  ((b) + (m) * DCTSIZE * (s) + (n) * SIZEOF_MMWORD)
+%define XMMBLOCK(m, n, b, s) \
+  ((b) + (m) * DCTSIZE * (s) + (n) * SIZEOF_XMMWORD)
+%define YMMBLOCK(m, n, b, s) \
+  ((b) + (m) * DCTSIZE * (s) + (n) * SIZEOF_YMMWORD)
+
+; --------------------------------------------------------------------------

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/nasm/jsimdcfg.inc Added

@@ -0,0 +1,93 @@
+;
+; Automatically generated include file from jsimdcfg.inc.h
+;
+;
+; -- jpeglib.h
+;
+%define DCTSIZE 8
+%define DCTSIZE2 64
+;
+; -- jmorecfg.h
+;
+%define RGB_RED 0
+%define RGB_GREEN 1
+%define RGB_BLUE 2
+%define RGB_PIXELSIZE 3
+%define EXT_RGB_RED 0
+%define EXT_RGB_GREEN 1
+%define EXT_RGB_BLUE 2
+%define EXT_RGB_PIXELSIZE 3
+%define EXT_RGBX_RED 0
+%define EXT_RGBX_GREEN 1
+%define EXT_RGBX_BLUE 2
+%define EXT_RGBX_PIXELSIZE 4
+%define EXT_BGR_RED 2
+%define EXT_BGR_GREEN 1
+%define EXT_BGR_BLUE 0
+%define EXT_BGR_PIXELSIZE 3
+%define EXT_BGRX_RED 2
+%define EXT_BGRX_GREEN 1
+%define EXT_BGRX_BLUE 0
+%define EXT_BGRX_PIXELSIZE 4
+%define EXT_XBGR_RED 3
+%define EXT_XBGR_GREEN 2
+%define EXT_XBGR_BLUE 1
+%define EXT_XBGR_PIXELSIZE 4
+%define EXT_XRGB_RED 1
+%define EXT_XRGB_GREEN 2
+%define EXT_XRGB_BLUE 3
+%define EXT_XRGB_PIXELSIZE 4
+%define RGBX_FILLER_0XFF 1
+; Representation of a single sample (pixel element value).
+; On this SIMD implementation, this must be 'unsigned char'.
+;
+%define JSAMPLE byte ; unsigned char
+%define SIZEOF_JSAMPLE SIZEOF_BYTE ; sizeof(JSAMPLE)
+%define CENTERJSAMPLE 128
+; Representation of a DCT frequency coefficient.
+; On this SIMD implementation, this must be 'short'.
+;
+%define JCOEF word ; short
+%define SIZEOF_JCOEF SIZEOF_WORD ; sizeof(JCOEF)
+; Datatype used for image dimensions.
+; On this SIMD implementation, this must be 'unsigned int'.
+;
+%define JDIMENSION dword ; unsigned int
+%define SIZEOF_JDIMENSION SIZEOF_DWORD ; sizeof(JDIMENSION)
+%define JSAMPROW POINTER ; JSAMPLE * (jpeglib.h)
+%define JSAMPARRAY POINTER ; JSAMPROW * (jpeglib.h)
+%define JSAMPIMAGE POINTER ; JSAMPARRAY * (jpeglib.h)
+%define JCOEFPTR POINTER ; JCOEF * (jpeglib.h)
+%define SIZEOF_JSAMPROW SIZEOF_POINTER ; sizeof(JSAMPROW)
+%define SIZEOF_JSAMPARRAY SIZEOF_POINTER ; sizeof(JSAMPARRAY)
+%define SIZEOF_JSAMPIMAGE SIZEOF_POINTER ; sizeof(JSAMPIMAGE)
+%define SIZEOF_JCOEFPTR SIZEOF_POINTER ; sizeof(JCOEFPTR)
+;
+; -- jdct.h
+;
+; A forward DCT routine is given a pointer to a work area of type DCTELEM;
+; the DCT is to be performed in-place in that buffer.
+; To maximize parallelism, Type DCTELEM is changed to short (originally, int).
+;
+%define DCTELEM word ; short
+%define SIZEOF_DCTELEM SIZEOF_WORD ; sizeof(DCTELEM)
+%define float FP32 ; float
+%define SIZEOF_FAST_FLOAT SIZEOF_FP32 ; sizeof(float)
+; To maximize parallelism, Type short is changed to short.
+;
+%define ISLOW_MULT_TYPE word ; must be short
+%define SIZEOF_ISLOW_MULT_TYPE SIZEOF_WORD ; sizeof(ISLOW_MULT_TYPE)
+%define IFAST_MULT_TYPE word ; must be short
+%define SIZEOF_IFAST_MULT_TYPE SIZEOF_WORD ; sizeof(IFAST_MULT_TYPE)
+%define IFAST_SCALE_BITS 2 ; fractional bits in scale factors
+%define FLOAT_MULT_TYPE FP32 ; must be float
+%define SIZEOF_FLOAT_MULT_TYPE SIZEOF_FP32 ; sizeof(FLOAT_MULT_TYPE)
+;
+; -- jsimd.h
+;
+%define JSIMD_NONE 0x00
+%define JSIMD_MMX 0x01
+%define JSIMD_3DNOW 0x02
+%define JSIMD_SSE 0x04
+%define JSIMD_SSE2 0x08
+%define JSIMD_AVX2 0x80

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/nasm/jsimdcfg.inc.h Added

@@ -0,0 +1,133 @@
+/*
+ * This file generates the include file for the assembly
+ * implementations by abusing the C preprocessor.
+ *
+ * Note: Some things are manually defined as they need to
+ * be mapped to NASM types.
+ */
+
+;
+; Automatically generated include file from jsimdcfg.inc.h
+;
+
+#define JPEG_INTERNALS
+
+#include "../jpeglib.h"
+#include "../jconfig.h"
+#include "../jmorecfg.h"
+#include "jsimd.h"
+
+;
+; -- jpeglib.h
+;
+
+%define _cpp_protection_DCTSIZE   DCTSIZE
+%define _cpp_protection_DCTSIZE2  DCTSIZE2
+
+;
+; -- jmorecfg.h
+;
+
+%define _cpp_protection_RGB_RED             RGB_RED
+%define _cpp_protection_RGB_GREEN           RGB_GREEN
+%define _cpp_protection_RGB_BLUE            RGB_BLUE
+%define _cpp_protection_RGB_PIXELSIZE       RGB_PIXELSIZE
+
+%define _cpp_protection_EXT_RGB_RED         EXT_RGB_RED
+%define _cpp_protection_EXT_RGB_GREEN       EXT_RGB_GREEN
+%define _cpp_protection_EXT_RGB_BLUE        EXT_RGB_BLUE
+%define _cpp_protection_EXT_RGB_PIXELSIZE   EXT_RGB_PIXELSIZE
+
+%define _cpp_protection_EXT_RGBX_RED        EXT_RGBX_RED
+%define _cpp_protection_EXT_RGBX_GREEN      EXT_RGBX_GREEN
+%define _cpp_protection_EXT_RGBX_BLUE       EXT_RGBX_BLUE
+%define _cpp_protection_EXT_RGBX_PIXELSIZE  EXT_RGBX_PIXELSIZE
+
+%define _cpp_protection_EXT_BGR_RED         EXT_BGR_RED
+%define _cpp_protection_EXT_BGR_GREEN       EXT_BGR_GREEN
+%define _cpp_protection_EXT_BGR_BLUE        EXT_BGR_BLUE
+%define _cpp_protection_EXT_BGR_PIXELSIZE   EXT_BGR_PIXELSIZE
+
+%define _cpp_protection_EXT_BGRX_RED        EXT_BGRX_RED
+%define _cpp_protection_EXT_BGRX_GREEN      EXT_BGRX_GREEN
+%define _cpp_protection_EXT_BGRX_BLUE       EXT_BGRX_BLUE
+%define _cpp_protection_EXT_BGRX_PIXELSIZE  EXT_BGRX_PIXELSIZE
+
+%define _cpp_protection_EXT_XBGR_RED        EXT_XBGR_RED
+%define _cpp_protection_EXT_XBGR_GREEN      EXT_XBGR_GREEN
+%define _cpp_protection_EXT_XBGR_BLUE       EXT_XBGR_BLUE
+%define _cpp_protection_EXT_XBGR_PIXELSIZE  EXT_XBGR_PIXELSIZE
+
+%define _cpp_protection_EXT_XRGB_RED        EXT_XRGB_RED
+%define _cpp_protection_EXT_XRGB_GREEN      EXT_XRGB_GREEN
+%define _cpp_protection_EXT_XRGB_BLUE       EXT_XRGB_BLUE
+%define _cpp_protection_EXT_XRGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+
+%define RGBX_FILLER_0XFF  1
+
+; Representation of a single sample (pixel element value).
+; On this SIMD implementation, this must be 'unsigned char'.
+;
+
+%define JSAMPLE            byte            ; unsigned char
+%define SIZEOF_JSAMPLE     SIZEOF_BYTE     ; sizeof(JSAMPLE)
+
+%define _cpp_protection_CENTERJSAMPLE  CENTERJSAMPLE
+
+; Representation of a DCT frequency coefficient.
+; On this SIMD implementation, this must be 'short'.
+;
+%define JCOEF              word            ; short
+%define SIZEOF_JCOEF       SIZEOF_WORD     ; sizeof(JCOEF)
+
+; Datatype used for image dimensions.
+; On this SIMD implementation, this must be 'unsigned int'.
+;
+%define JDIMENSION         dword           ; unsigned int
+%define SIZEOF_JDIMENSION  SIZEOF_DWORD    ; sizeof(JDIMENSION)
+
+%define JSAMPROW           POINTER         ; JSAMPLE *     (jpeglib.h)
+%define JSAMPARRAY         POINTER         ; JSAMPROW *    (jpeglib.h)
+%define JSAMPIMAGE         POINTER         ; JSAMPARRAY *  (jpeglib.h)
+%define JCOEFPTR           POINTER         ; JCOEF *       (jpeglib.h)
+%define SIZEOF_JSAMPROW    SIZEOF_POINTER  ; sizeof(JSAMPROW)
+%define SIZEOF_JSAMPARRAY  SIZEOF_POINTER  ; sizeof(JSAMPARRAY)
+%define SIZEOF_JSAMPIMAGE  SIZEOF_POINTER  ; sizeof(JSAMPIMAGE)
+%define SIZEOF_JCOEFPTR    SIZEOF_POINTER  ; sizeof(JCOEFPTR)
+
+;
+; -- jdct.h
+;
+
+; A forward DCT routine is given a pointer to a work area of type DCTELEM;
+; the DCT is to be performed in-place in that buffer.
+; To maximize parallelism, Type DCTELEM is changed to short (originally, int).
+;
+%define DCTELEM                 word         ; short
+%define SIZEOF_DCTELEM          SIZEOF_WORD  ; sizeof(DCTELEM)
+
+%define FAST_FLOAT              FP32         ; float
+%define SIZEOF_FAST_FLOAT       SIZEOF_FP32  ; sizeof(FAST_FLOAT)
+
+; To maximize parallelism, Type MULTIPLIER is changed to short.
+;
+%define ISLOW_MULT_TYPE         word         ; must be short
+%define SIZEOF_ISLOW_MULT_TYPE  SIZEOF_WORD  ; sizeof(ISLOW_MULT_TYPE)
+
+%define IFAST_MULT_TYPE         word         ; must be short
+%define SIZEOF_IFAST_MULT_TYPE  SIZEOF_WORD  ; sizeof(IFAST_MULT_TYPE)
+%define IFAST_SCALE_BITS        2            ; fractional bits in scale factors
+
+%define FLOAT_MULT_TYPE         FP32         ; must be float
+%define SIZEOF_FLOAT_MULT_TYPE  SIZEOF_FP32  ; sizeof(FLOAT_MULT_TYPE)
+
+;
+; -- jsimd.h
+;
+
+%define _cpp_protection_JSIMD_NONE   JSIMD_NONE
+%define _cpp_protection_JSIMD_MMX    JSIMD_MMX
+%define _cpp_protection_JSIMD_3DNOW  JSIMD_3DNOW
+%define _cpp_protection_JSIMD_SSE    JSIMD_SSE
+%define _cpp_protection_JSIMD_SSE2   JSIMD_SSE2
+%define _cpp_protection_JSIMD_AVX2   JSIMD_AVX2

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/nasm/jsimdext.inc Added

@@ -0,0 +1,520 @@
+;
+; jsimdext.inc - common declarations
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2010, 2016, 2018-2019, D. R. Commander.
+; Copyright (C) 2018, Matthieu Darbois.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library - version 1.02
+;
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+;
+; This software is provided 'as-is', without any express or implied
+; warranty.  In no event will the authors be held liable for any damages
+; arising from the use of this software.
+;
+; Permission is granted to anyone to use this software for any purpose,
+; including commercial applications, and to alter it and redistribute it
+; freely, subject to the following restrictions:
+;
+; 1. The origin of this software must not be misrepresented; you must not
+;    claim that you wrote the original software. If you use this software
+;    in a product, an acknowledgment in the product documentation would be
+;    appreciated but is not required.
+; 2. Altered source versions must be plainly marked as such, and must not be
+;    misrepresented as being the original software.
+; 3. This notice may not be removed or altered from any source distribution.
+
+; ==========================================================================
+;  System-dependent configurations
+
+%ifdef WIN32    ; ----(nasm -fwin32 -DWIN32 ...)--------
+; * Microsoft Visual C++
+; * MinGW (Minimalist GNU for Windows)
+; * CygWin
+; * LCC-Win32
+
+; -- segment definition --
+;
+%ifdef __YASM_VER__
+%define SEG_TEXT   .text  align=32
+%define SEG_CONST  .rdata align=32
+%else
+%define SEG_TEXT   .text  align=32 public use32 class=CODE
+%define SEG_CONST  .rdata align=32 public use32 class=CONST
+%endif
+
+%elifdef WIN64  ; ----(nasm -fwin64 -DWIN64 ...)--------
+; * Microsoft Visual C++
+
+; -- segment definition --
+;
+%ifdef __YASM_VER__
+%define SEG_TEXT    .text  align=32
+%define SEG_CONST   .rdata align=32
+%else
+%define SEG_TEXT    .text  align=32 public use64 class=CODE
+%define SEG_CONST   .rdata align=32 public use64 class=CONST
+%endif
+%define EXTN(name)  name                ; foo() -> foo
+
+%elifdef OBJ32  ; ----(nasm -fobj -DOBJ32 ...)----------
+; * Borland C++ (Win32)
+
+; -- segment definition --
+;
+%define SEG_TEXT   _text align=32 public use32 class=CODE
+%define SEG_CONST  _data align=32 public use32 class=DATA
+
+%elifdef ELF    ; ----(nasm -felf64 -DELF ...)------------
+; * Linux
+; * *BSD family Unix using elf format
+; * Unix System V, including Solaris x86, UnixWare and SCO Unix
+
+; mark stack as non-executable
+section .note.GNU-stack noalloc noexec nowrite progbits
+
+; -- segment definition --
+;
+%ifdef __x86_64__
+%define SEG_TEXT   .text   progbits align=32
+%define SEG_CONST  .rodata progbits align=32
+%else
+%define SEG_TEXT   .text   progbits alloc exec   nowrite align=32
+%define SEG_CONST  .rodata progbits alloc noexec nowrite align=32
+%endif
+
+; To make the code position-independent, append -DPIC to the commandline
+;
+%define GOT_SYMBOL  _GLOBAL_OFFSET_TABLE_  ; ELF supports PIC
+%define EXTN(name)  name                   ; foo() -> foo
+
+%elifdef AOUT   ; ----(nasm -faoutb/aout -DAOUT ...)----
+; * Older Linux using a.out format  (nasm -f aout -DAOUT ...)
+; * *BSD family Unix using a.out format  (nasm -f aoutb -DAOUT ...)
+
+; -- segment definition --
+;
+%define SEG_TEXT   .text
+%define SEG_CONST  .data
+
+; To make the code position-independent, append -DPIC to the commandline
+;
+%define GOT_SYMBOL  __GLOBAL_OFFSET_TABLE_  ; BSD-style a.out supports PIC
+
+%elifdef MACHO  ; ----(nasm -fmacho -DMACHO ...)--------
+; * NeXTstep/OpenStep/Rhapsody/Darwin/MacOS X (Mach-O format)
+
+; -- segment definition --
+;
+%define SEG_TEXT   .text  ;align=32     ; nasm doesn't accept align=32. why?
+%define SEG_CONST  .rodata align=32
+
+; The generation of position-independent code (PIC) is the default on Darwin.
+;
+%define PIC
+%define GOT_SYMBOL  _MACHO_PIC_         ; Mach-O style code-relative addressing
+
+%else           ; ----(Other case)----------------------
+
+; -- segment definition --
+;
+%define SEG_TEXT   .text
+%define SEG_CONST  .data
+
+%endif          ; ----------------------------------------------
+
+; ==========================================================================
+
+; --------------------------------------------------------------------------
+;  Common types
+;
+%ifdef __x86_64__
+%ifnidn __OUTPUT_FORMAT__, elfx32
+%define POINTER         qword           ; general pointer type
+%define SIZEOF_POINTER  SIZEOF_QWORD    ; sizeof(POINTER)
+%define POINTER_BIT     QWORD_BIT       ; sizeof(POINTER)*BYTE_BIT
+%define resp            resq
+%define dp              dq
+%define raxp            rax
+%define rbxp            rbx
+%define rcxp            rcx
+%define rdxp            rdx
+%define rsip            rsi
+%define rdip            rdi
+%define rbpp            rbp
+%define rspp            rsp
+%define r8p             r8
+%define r9p             r9
+%define r10p            r10
+%define r11p            r11
+%define r12p            r12
+%define r13p            r13
+%define r14p            r14
+%define r15p            r15
+%endif
+%endif
+%ifndef raxp
+%define POINTER         dword           ; general pointer type
+%define SIZEOF_POINTER  SIZEOF_DWORD    ; sizeof(POINTER)
+%define POINTER_BIT     DWORD_BIT       ; sizeof(POINTER)*BYTE_BIT
+%define resp            resd
+%define dp              dd
+; x86_64 ILP32 ABI (x32)
+%define raxp            eax
+%define rbxp            ebx
+%define rcxp            ecx
+%define rdxp            edx
+%define rsip            esi
+%define rdip            edi
+%define rbpp            ebp
+%define rspp            esp
+%define r8p             r8d
+%define r9p             r9d
+%define r10p            r10d
+%define r11p            r11d
+%define r12p            r12d
+%define r13p            r13d
+%define r14p            r14d
+%define r15p            r15d
+%endif
+
+%define INT             dword           ; signed integer type
+%define SIZEOF_INT      SIZEOF_DWORD    ; sizeof(INT)
+%define INT_BIT         DWORD_BIT       ; sizeof(INT)*BYTE_BIT
+
+%define FP32            dword           ; IEEE754 single
+%define SIZEOF_FP32     SIZEOF_DWORD    ; sizeof(FP32)
+%define FP32_BIT        DWORD_BIT       ; sizeof(FP32)*BYTE_BIT
+
+%define MMWORD          qword           ; int64  (MMX register)
+%define SIZEOF_MMWORD   SIZEOF_QWORD    ; sizeof(MMWORD)
+%define MMWORD_BIT      QWORD_BIT       ; sizeof(MMWORD)*BYTE_BIT
+
+; NASM is buggy and doesn't properly handle operand sizes for SSE
+; instructions, so for now we have to define XMMWORD as blank.
+%define XMMWORD                         ; int128 (SSE register)
+%define SIZEOF_XMMWORD  SIZEOF_OWORD    ; sizeof(XMMWORD)
+%define XMMWORD_BIT     OWORD_BIT       ; sizeof(XMMWORD)*BYTE_BIT
+
+%define YMMWORD                         ; int256 (AVX register)
+%define SIZEOF_YMMWORD  SIZEOF_YWORD    ; sizeof(YMMWORD)
+%define YMMWORD_BIT     YWORD_BIT       ; sizeof(YMMWORD)*BYTE_BIT
+
+; Similar hacks for when we load a dword or MMWORD into an xmm# register
+%define XMM_DWORD
+%define XMM_MMWORD
+
+%define SIZEOF_BYTE   1                 ; sizeof(byte)
+%define SIZEOF_WORD   2                 ; sizeof(word)
+%define SIZEOF_DWORD  4                 ; sizeof(dword)
+%define SIZEOF_QWORD  8                 ; sizeof(qword)
+%define SIZEOF_OWORD  16                ; sizeof(oword)
+%define SIZEOF_YWORD  32                ; sizeof(yword)
+
+%define BYTE_BIT      8                 ; CHAR_BIT in C
+%define WORD_BIT      16                ; sizeof(word)*BYTE_BIT
+%define DWORD_BIT     32                ; sizeof(dword)*BYTE_BIT
+%define QWORD_BIT     64                ; sizeof(qword)*BYTE_BIT
+%define OWORD_BIT     128               ; sizeof(oword)*BYTE_BIT
+%define YWORD_BIT     256               ; sizeof(yword)*BYTE_BIT
+
+; --------------------------------------------------------------------------
+;  External Symbol Name
+;
+%ifndef EXTN
+%define EXTN(name)  _ %+ name           ; foo() -> _foo
+%endif
+
+; --------------------------------------------------------------------------
+;  Hidden symbols
+;
+%ifdef ELF      ; ----(nasm -felf64 -DELF ...)--------
+%define GLOBAL_FUNCTION(name)  global EXTN(name):function hidden
+%define GLOBAL_DATA(name)      global EXTN(name):data hidden
+%elifdef MACHO  ; ----(nasm -fmacho -DMACHO ...)--------
+%ifdef __YASM_VER__
+%define GLOBAL_FUNCTION(name)  global EXTN(name):private_extern
+%define GLOBAL_DATA(name)      global EXTN(name):private_extern
+%else
+%if __NASM_VERSION_ID__ >= 0x020E0000
+%define GLOBAL_FUNCTION(name)  global EXTN(name):private_extern
+%define GLOBAL_DATA(name)      global EXTN(name):private_extern
+%endif
+%endif
+%endif
+
+%ifndef GLOBAL_FUNCTION
+%define GLOBAL_FUNCTION(name)  global EXTN(name)
+%endif
+%ifndef GLOBAL_DATA
+%define GLOBAL_DATA(name)      global EXTN(name)
+%endif
+
+; --------------------------------------------------------------------------
+;  Macros for position-independent code (PIC) support
+;
+%ifndef GOT_SYMBOL
+%undef PIC
+%endif
+
+%ifdef PIC  ; -------------------------------------------
+
+%ifidn GOT_SYMBOL, _MACHO_PIC_  ; --------------------
+
+; At present, nasm doesn't seem to support PIC generation for Mach-O.
+; The PIC support code below is a little tricky.
+
+    SECTION     SEG_CONST
+const_base:
+
+%define GOTOFF(got, sym)  (got) + (sym) - const_base
+
+%imacro get_GOT 1
+    ; NOTE: this macro destroys ecx resister.
+    call        %%geteip
+    add         ecx, byte (%%ref - $)
+    jmp         short %%adjust
+%%geteip:
+    mov         ecx, POINTER esp
+    ret
+%%adjust:
+    push        ebp
+    xor         ebp, ebp                ; ebp = 0
+%ifidni %1, ebx  ; (%1 == ebx)
+    ; db 0x8D,0x9C + jmp near const_base =
+    ;   lea ebx, ecx+ebp*8+(const_base-%%ref) ; 8D,9C,E9,(offset32)
+    db          0x8D, 0x9C              ; 8D,9C
+    jmp         near const_base         ; E9,(const_base-%%ref)
+%%ref:
+%else  ; (%1 != ebx)
+    ; db 0x8D,0x8C + jmp near const_base =
+    ;   lea ecx, ecx+ebp*8+(const_base-%%ref) ; 8D,8C,E9,(offset32)
+    db          0x8D, 0x8C              ; 8D,8C
+    jmp         near const_base         ; E9,(const_base-%%ref)
+%%ref:
+    mov         %1, ecx
+%endif  ; (%1 == ebx)
+    pop         ebp
+%endmacro
+
+%else     ; GOT_SYMBOL != _MACHO_PIC_ ----------------
+
+%define GOTOFF(got, sym)  (got) + (sym) wrt ..gotoff
+
+%imacro get_GOT 1
+    extern      GOT_SYMBOL
+    call        %%geteip
+    add         %1, GOT_SYMBOL + $$ - $ wrt ..gotpc
+    jmp         short %%done
+%%geteip:
+    mov         %1, POINTER esp
+    ret
+%%done:
+%endmacro
+
+%endif    ; GOT_SYMBOL == _MACHO_PIC_ ----------------
+
+%imacro pushpic 1.nolist
+    push        %1
+%endmacro
+%imacro poppic  1.nolist
+    pop         %1
+%endmacro
+%imacro movpic  2.nolist
+    mov         %1, %2
+%endmacro
+
+%else    ; !PIC -----------------------------------------
+
+%define GOTOFF(got, sym)  (sym)
+
+%imacro get_GOT 1.nolist
+%endmacro
+%imacro pushpic 1.nolist
+%endmacro
+%imacro poppic  1.nolist
+%endmacro
+%imacro movpic  2.nolist
+%endmacro
+
+%endif   ;  PIC -----------------------------------------
+
+; --------------------------------------------------------------------------
+;  Align the next instruction on {2,4,8,16,..}-byte boundary.
+;  ".balign n,,m" in GNU as
+;
+%define MSKLE(x, y)  (~(((y) & 0xFFFF) - ((x) & 0xFFFF)) >> 16)
+%define FILLB(b, n)  (($$-(b)) & ((n)-1))
+
+%imacro alignx 1-2.nolist 0xFFFF
+%%bs: \
+  times MSKLE(FILLB(%%bs, %1), %2) & MSKLE(16, FILLB($, %1)) & FILLB($, %1) \
+        db 0x90                                      ; nop
+  times MSKLE(FILLB(%%bs, %1), %2) & FILLB($, %1) / 9 \
+        db 0x8D, 0x9C, 0x23, 0x00, 0x00, 0x00, 0x00  ; lea ebx,ebx+0x00000000
+  times MSKLE(FILLB(%%bs, %1), %2) & FILLB($, %1) / 7 \
+        db 0x8D, 0xAC, 0x25, 0x00, 0x00, 0x00, 0x00  ; lea ebp,ebp+0x00000000
+  times MSKLE(FILLB(%%bs, %1), %2) & FILLB($, %1) / 6 \
+        db 0x8D, 0xAD, 0x00, 0x00, 0x00, 0x00        ; lea ebp,ebp+0x00000000
+  times MSKLE(FILLB(%%bs, %1), %2) & FILLB($, %1) / 4 \
+        db 0x8D, 0x6C, 0x25, 0x00                    ; lea ebp,ebp+0x00
+  times MSKLE(FILLB(%%bs, %1), %2) & FILLB($, %1) / 3 \
+        db 0x8D, 0x6D, 0x00                          ; lea ebp,ebp+0x00
+  times MSKLE(FILLB(%%bs, %1), %2) & FILLB($, %1) / 2 \
+        db 0x8B, 0xED                                ; mov ebp,ebp
+  times MSKLE(FILLB(%%bs, %1), %2) & FILLB($, %1) / 1 \
+        db 0x90                                      ; nop
+%endmacro
+
+; Align the next data on {2,4,8,16,..}-byte boundary.
+;
+%imacro alignz 1.nolist
+    align       %1, db 0                ; filling zeros
+%endmacro
+
+%ifdef __x86_64__
+
+%ifdef WIN64
+
+%imacro collect_args 1
+    sub         rsp, SIZEOF_XMMWORD
+    movaps      XMMWORD rsp, xmm6
+    sub         rsp, SIZEOF_XMMWORD
+    movaps      XMMWORD rsp, xmm7
+    mov         r10, rcx
+%if %1 > 1
+    mov         r11, rdx
+%endif
+%if %1 > 2
+    push        r12
+    mov         r12, r8
+%endif
+%if %1 > 3
+    push        r13
+    mov         r13, r9
+%endif
+%if %1 > 4
+    push        r14
+    mov         r14, rax+48
+%endif
+%if %1 > 5
+    push        r15
+    mov         r15, rax+56
+%endif
+    push        rsi
+    push        rdi
+%endmacro
+
+%imacro uncollect_args 1
+    pop         rdi
+    pop         rsi
+%if %1 > 5
+    pop         r15
+%endif
+%if %1 > 4
+    pop         r14
+%endif
+%if %1 > 3
+    pop         r13
+%endif
+%if %1 > 2
+    pop         r12
+%endif
+    movaps      xmm7, XMMWORD rsp
+    add         rsp, SIZEOF_XMMWORD
+    movaps      xmm6, XMMWORD rsp
+    add         rsp, SIZEOF_XMMWORD
+%endmacro
+
+%imacro push_xmm 1
+    sub         rsp, %1 * SIZEOF_XMMWORD
+    movaps      XMMWORD rsp+0*SIZEOF_XMMWORD, xmm8
+%if %1 > 1
+    movaps      XMMWORD rsp+1*SIZEOF_XMMWORD, xmm9
+%endif
+%if %1 > 2
+    movaps      XMMWORD rsp+2*SIZEOF_XMMWORD, xmm10
+%endif
+%if %1 > 3
+    movaps      XMMWORD rsp+3*SIZEOF_XMMWORD, xmm11
+%endif
+%endmacro
+
+%imacro pop_xmm 1
+    movaps      xmm8, XMMWORD rsp+0*SIZEOF_XMMWORD
+%if %1 > 1
+    movaps      xmm9, XMMWORD rsp+1*SIZEOF_XMMWORD
+%endif
+%if %1 > 2
+    movaps      xmm10, XMMWORD rsp+2*SIZEOF_XMMWORD
+%endif
+%if %1 > 3
+    movaps      xmm11, XMMWORD rsp+3*SIZEOF_XMMWORD
+%endif
+    add         rsp, %1 * SIZEOF_XMMWORD
+%endmacro
+
+%else
+
+%imacro collect_args 1
+    push        r10
+    mov         r10, rdi
+%if %1 > 1
+    push        r11
+    mov         r11, rsi
+%endif
+%if %1 > 2
+    push        r12
+    mov         r12, rdx
+%endif
+%if %1 > 3
+    push        r13
+    mov         r13, rcx
+%endif
+%if %1 > 4
+    push        r14
+    mov         r14, r8
+%endif
+%if %1 > 5
+    push        r15
+    mov         r15, r9
+%endif
+%endmacro
+
+%imacro uncollect_args 1
+%if %1 > 5
+    pop         r15
+%endif
+%if %1 > 4
+    pop         r14
+%endif
+%if %1 > 3
+    pop         r13
+%endif
+%if %1 > 2
+    pop         r12
+%endif
+%if %1 > 1
+    pop         r11
+%endif
+    pop         r10
+%endmacro
+
+%imacro push_xmm 1
+%endmacro
+
+%imacro pop_xmm 1
+%endmacro
+
+%endif
+
+%endif
+
+; --------------------------------------------------------------------------
+;  Defines picked up from the C headers
+;
+%include "jsimdcfg.inc"
+
+; --------------------------------------------------------------------------

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jccolext-altivec.c Added

@@ -0,0 +1,269 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2014, Jay Foad.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jccolor-altivec.c */
+
+
+void jsimd_rgb_ycc_convert_altivec(JDIMENSION img_width, JSAMPARRAY input_buf,
+                                   JSAMPIMAGE output_buf,
+                                   JDIMENSION output_row, int num_rows)
+{
+  JSAMPROW inptr, outptr0, outptr1, outptr2;
+  int pitch = img_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbufRGB_PIXELSIZE * 16;
+
+  __vector unsigned char rgb0, rgb1 = { 0 }, rgb2 = { 0 },
+    rgbg0, rgbg1, rgbg2, rgbg3, y, cb, cr;
+#if __BIG_ENDIAN__ || RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3 = { 0 };
+#endif
+#if __BIG_ENDIAN__ && RGB_PIXELSIZE == 4
+  __vector unsigned char rgb4 = { 0 };
+#endif
+  __vector short rg0, rg1, rg2, rg3, bg0, bg1, bg2, bg3;
+  __vector unsigned short yl, yh, crl, crh, cbl, cbh;
+  __vector int y0, y1, y2, y3, cr0, cr1, cr2, cr3, cb0, cb1, cb2, cb3;
+
+  /* Constants */
+  __vector short pw_f0299_f0337 = { __4X2(F_0_299, F_0_337) },
+    pw_f0114_f0250 = { __4X2(F_0_114, F_0_250) },
+    pw_mf016_mf033 = { __4X2(-F_0_168, -F_0_331) },
+    pw_mf008_mf041 = { __4X2(-F_0_081, -F_0_418) };
+  __vector unsigned short pw_f050_f000 = { __4X2(F_0_500, 0) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) },
+    pd_onehalfm1_cj = { __4X(ONE_HALF - 1 + (CENTERJSAMPLE << SCALEBITS)) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index =
+      {  0,  1,  4,  5,  8,  9, 12, 13, 16, 17, 20, 21, 24, 25, 28, 29 };
+#else
+    shift_pack_index =
+      {  2,  3,  6,  7, 10, 11, 14, 15, 18, 19, 22, 23, 26, 27, 30, 31 };
+#endif
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf0output_row;
+    outptr1 = output_buf1output_row;
+    outptr2 = output_buf2output_row;
+    output_row++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, inptr += RGB_PIXELSIZE * 16,
+         outptr0 += 16, outptr1 += 16, outptr2 += 16) {
+
+#if __BIG_ENDIAN__
+      /* Load 16 pixels == 48 or 64 bytes */
+      offset = (size_t)inptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overread.  Since there is no way to
+           * read a partial AltiVec register, overread would occur on the last
+           * chunk of the last image row if the right edge is not on a 16-byte
+           * boundary.  It could also occur on other rows if the bytes per row
+           * is low enough.  Since we can't determine whether we're on the last
+           * image row, we have to assume every row is the last.
+           */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (bytes > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (bytes > 32)
+            rgb2 = vec_ld(32, inptr);
+          if (bytes > 48)
+            rgb3 = vec_ld(48, inptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            rgb4 = vec_ld(64, inptr);
+#endif
+          unaligned_shift_index = vec_lvsl(0, inptr);
+          rgb0 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          rgb1 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+          rgb2 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_perm(rgb3, rgb4, unaligned_shift_index);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = VEC_LD(0, tmpbuf);
+          rgb1 = VEC_LD(16, tmpbuf);
+          rgb2 = VEC_LD(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = VEC_LD(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = VEC_LD(0, inptr);
+          if (num_cols > 16)
+            rgb1 = VEC_LD(16, inptr);
+          if (num_cols > 32)
+            rgb2 = VEC_LD(32, inptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            rgb3 = VEC_LD(48, inptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+
+#if RGB_PIXELSIZE == 3
+      /* rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX0);
+      rgbg1 = vec_perm(rgb0, rgb1, (__vector unsigned char)RGBG_INDEX1);
+      rgbg2 = vec_perm(rgb1, rgb2, (__vector unsigned char)RGBG_INDEX2);
+      rgbg3 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX3);
+#else
+      /* rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX);
+      rgbg1 = vec_perm(rgb1, rgb1, (__vector unsigned char)RGBG_INDEX);
+      rgbg2 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX);
+      rgbg3 = vec_perm(rgb3, rgb3, (__vector unsigned char)RGBG_INDEX);
+#endif
+
+      /* rg0 = R0 G0 R1 G1 R2 G2 R3 G3
+       * bg0 = B0 G0 B1 G1 B2 G2 B3 G3
+       * ...
+       *
+       * NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      rg0 = (__vector signed short)VEC_UNPACKHU(rgbg0);
+      bg0 = (__vector signed short)VEC_UNPACKLU(rgbg0);
+      rg1 = (__vector signed short)VEC_UNPACKHU(rgbg1);
+      bg1 = (__vector signed short)VEC_UNPACKLU(rgbg1);
+      rg2 = (__vector signed short)VEC_UNPACKHU(rgbg2);
+      bg2 = (__vector signed short)VEC_UNPACKLU(rgbg2);
+      rg3 = (__vector signed short)VEC_UNPACKHU(rgbg3);
+      bg3 = (__vector signed short)VEC_UNPACKLU(rgbg3);
+
+      /* (Original)
+       * Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       *
+       * (This implementation)
+       * Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       */
+
+      /* Calculate Y values */
+
+      y0 = vec_msums(rg0, pw_f0299_f0337, pd_onehalf);
+      y1 = vec_msums(rg1, pw_f0299_f0337, pd_onehalf);
+      y2 = vec_msums(rg2, pw_f0299_f0337, pd_onehalf);
+      y3 = vec_msums(rg3, pw_f0299_f0337, pd_onehalf);
+      y0 = vec_msums(bg0, pw_f0114_f0250, y0);
+      y1 = vec_msums(bg1, pw_f0114_f0250, y1);
+      y2 = vec_msums(bg2, pw_f0114_f0250, y2);
+      y3 = vec_msums(bg3, pw_f0114_f0250, y3);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      yl = vec_perm((__vector unsigned short)y0, (__vector unsigned short)y1,
+                    shift_pack_index);
+      yh = vec_perm((__vector unsigned short)y2, (__vector unsigned short)y3,
+                    shift_pack_index);
+      y = vec_pack(yl, yh);
+      vec_st(y, 0, outptr0);
+
+      /* Calculate Cb values */
+      cb0 = vec_msums(rg0, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb1 = vec_msums(rg1, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb2 = vec_msums(rg2, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb3 = vec_msums(rg3, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb0 = (__vector int)vec_msum((__vector unsigned short)bg0, pw_f050_f000,
+                                   (__vector unsigned int)cb0);
+      cb1 = (__vector int)vec_msum((__vector unsigned short)bg1, pw_f050_f000,
+                                   (__vector unsigned int)cb1);
+      cb2 = (__vector int)vec_msum((__vector unsigned short)bg2, pw_f050_f000,
+                                   (__vector unsigned int)cb2);
+      cb3 = (__vector int)vec_msum((__vector unsigned short)bg3, pw_f050_f000,
+                                   (__vector unsigned int)cb3);
+      cbl = vec_perm((__vector unsigned short)cb0,
+                     (__vector unsigned short)cb1, shift_pack_index);
+      cbh = vec_perm((__vector unsigned short)cb2,
+                     (__vector unsigned short)cb3, shift_pack_index);
+      cb = vec_pack(cbl, cbh);
+      vec_st(cb, 0, outptr1);
+
+      /* Calculate Cr values */
+      cr0 = vec_msums(bg0, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr1 = vec_msums(bg1, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr2 = vec_msums(bg2, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr3 = vec_msums(bg3, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr0 = (__vector int)vec_msum((__vector unsigned short)rg0, pw_f050_f000,
+                                   (__vector unsigned int)cr0);
+      cr1 = (__vector int)vec_msum((__vector unsigned short)rg1, pw_f050_f000,
+                                   (__vector unsigned int)cr1);
+      cr2 = (__vector int)vec_msum((__vector unsigned short)rg2, pw_f050_f000,
+                                   (__vector unsigned int)cr2);
+      cr3 = (__vector int)vec_msum((__vector unsigned short)rg3, pw_f050_f000,
+                                   (__vector unsigned int)cr3);
+      crl = vec_perm((__vector unsigned short)cr0,
+                     (__vector unsigned short)cr1, shift_pack_index);
+      crh = vec_perm((__vector unsigned short)cr2,
+                     (__vector unsigned short)cr3, shift_pack_index);
+      cr = vec_pack(crl, crh);
+      vec_st(cr, 0, outptr2);
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jccolor-altivec.c Added

@@ -0,0 +1,116 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* RGB --> YCC CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_081  5329                 /* FIX(0.08131) */
+#define F_0_114  7471                 /* FIX(0.11400) */
+#define F_0_168  11059                /* FIX(0.16874) */
+#define F_0_250  16384                /* FIX(0.25000) */
+#define F_0_299  19595                /* FIX(0.29900) */
+#define F_0_331  21709                /* FIX(0.33126) */
+#define F_0_418  27439                /* FIX(0.41869) */
+#define F_0_500  32768                /* FIX(0.50000) */
+#define F_0_587  38470                /* FIX(0.58700) */
+#define F_0_337  (F_0_587 - F_0_250)  /* FIX(0.58700) - FIX(0.25000) */
+
+#define SCALEBITS  16
+#define ONE_HALF  (1 << (SCALEBITS - 1))
+
+
+#define RGBG_INDEX0 \
+  {  0,  1,  3,  4,  6,  7,  9, 10,  2,  1,  5,  4,  8,  7, 11, 10 }
+#define RGBG_INDEX1 \
+  { 12, 13, 15, 16, 18, 19, 21, 22, 14, 13, 17, 16, 20, 19, 23, 22 }
+#define RGBG_INDEX2 \
+  {  8,  9, 11, 12, 14, 15, 17, 18, 10,  9, 13, 12, 16, 15, 19, 18 }
+#define RGBG_INDEX3 \
+  {  4,  5,  7,  8, 10, 11, 13, 14,  6,  5,  9,  8, 12, 11, 15, 14 }
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_rgb_ycc_convert_altivec  jsimd_extrgb_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define RGBG_INDEX \
+  {  0,  1,  4,  5,  8,  9, 12, 13,  2,  1,  6,  5, 10,  9, 14, 13 }
+#define jsimd_rgb_ycc_convert_altivec  jsimd_extrgbx_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define RGBG_INDEX0 \
+  {  2,  1,  5,  4,  8,  7, 11, 10,  0,  1,  3,  4,  6,  7,  9, 10 }
+#define RGBG_INDEX1 \
+  { 14, 13, 17, 16, 20, 19, 23, 22, 12, 13, 15, 16, 18, 19, 21, 22 }
+#define RGBG_INDEX2 \
+  { 10,  9, 13, 12, 16, 15, 19, 18,  8,  9, 11, 12, 14, 15, 17, 18 }
+#define RGBG_INDEX3 \
+  {  6,  5,  9,  8, 12, 11, 15, 14,  4,  5,  7,  8, 10, 11, 13, 14 }
+#define jsimd_rgb_ycc_convert_altivec  jsimd_extbgr_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define RGBG_INDEX \
+  {  2,  1,  6,  5, 10,  9, 14, 13,  0,  1,  4,  5,  8,  9, 12, 13 }
+#define jsimd_rgb_ycc_convert_altivec  jsimd_extbgrx_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define RGBG_INDEX \
+  {  3,  2,  7,  6, 11, 10, 15, 14,  1,  2,  5,  6,  9, 10, 13, 14 }
+#define jsimd_rgb_ycc_convert_altivec  jsimd_extxbgr_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define RGBG_INDEX \
+  {  1,  2,  5,  6,  9, 10, 13, 14,  3,  2,  7,  6, 11, 10, 15, 14 }
+#define jsimd_rgb_ycc_convert_altivec  jsimd_extxrgb_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jcgray-altivec.c Added

@@ -0,0 +1,111 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* RGB --> GRAYSCALE CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_114  7471                 /* FIX(0.11400) */
+#define F_0_250  16384                /* FIX(0.25000) */
+#define F_0_299  19595                /* FIX(0.29900) */
+#define F_0_587  38470                /* FIX(0.58700) */
+#define F_0_337  (F_0_587 - F_0_250)  /* FIX(0.58700) - FIX(0.25000) */
+
+#define SCALEBITS  16
+#define ONE_HALF  (1 << (SCALEBITS - 1))
+
+
+#define RGBG_INDEX0 \
+  {  0,  1,  3,  4,  6,  7,  9, 10,  2,  1,  5,  4,  8,  7, 11, 10 }
+#define RGBG_INDEX1 \
+  { 12, 13, 15, 16, 18, 19, 21, 22, 14, 13, 17, 16, 20, 19, 23, 22 }
+#define RGBG_INDEX2 \
+  {  8,  9, 11, 12, 14, 15, 17, 18, 10,  9, 13, 12, 16, 15, 19, 18 }
+#define RGBG_INDEX3 \
+  {  4,  5,  7,  8, 10, 11, 13, 14,  6,  5,  9,  8, 12, 11, 15, 14 }
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_rgb_gray_convert_altivec  jsimd_extrgb_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define RGBG_INDEX \
+  {  0,  1,  4,  5,  8,  9, 12, 13,  2,  1,  6,  5, 10,  9, 14, 13 }
+#define jsimd_rgb_gray_convert_altivec  jsimd_extrgbx_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define RGBG_INDEX0 \
+  {  2,  1,  5,  4,  8,  7, 11, 10,  0,  1,  3,  4,  6,  7,  9, 10 }
+#define RGBG_INDEX1 \
+  { 14, 13, 17, 16, 20, 19, 23, 22, 12, 13, 15, 16, 18, 19, 21, 22 }
+#define RGBG_INDEX2 \
+  { 10,  9, 13, 12, 16, 15, 19, 18,  8,  9, 11, 12, 14, 15, 17, 18 }
+#define RGBG_INDEX3 \
+  {  6,  5,  9,  8, 12, 11, 15, 14,  4,  5,  7,  8, 10, 11, 13, 14 }
+#define jsimd_rgb_gray_convert_altivec  jsimd_extbgr_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define RGBG_INDEX \
+  {  2,  1,  6,  5, 10,  9, 14, 13,  0,  1,  4,  5,  8,  9, 12, 13 }
+#define jsimd_rgb_gray_convert_altivec  jsimd_extbgrx_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define RGBG_INDEX \
+  {  3,  2,  7,  6, 11, 10, 15, 14,  1,  2,  5,  6,  9, 10, 13, 14 }
+#define jsimd_rgb_gray_convert_altivec  jsimd_extxbgr_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define RGBG_INDEX \
+  {  1,  2,  5,  6,  9, 10, 13, 14,  3,  2,  7,  6, 11, 10, 15, 14 }
+#define jsimd_rgb_gray_convert_altivec  jsimd_extxrgb_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jcgryext-altivec.c Added

@@ -0,0 +1,228 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2014, Jay Foad.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jcgray-altivec.c */
+
+
+void jsimd_rgb_gray_convert_altivec(JDIMENSION img_width, JSAMPARRAY input_buf,
+                                    JSAMPIMAGE output_buf,
+                                    JDIMENSION output_row, int num_rows)
+{
+  JSAMPROW inptr, outptr;
+  int pitch = img_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+  unsigned char __attribute__((aligned(16))) tmpbufRGB_PIXELSIZE * 16;
+#endif
+
+  __vector unsigned char rgb0, rgb1 = { 0 }, rgb2 = { 0 },
+    rgbg0, rgbg1, rgbg2, rgbg3, y;
+#if __BIG_ENDIAN__ || RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3 = { 0 };
+#endif
+#if __BIG_ENDIAN__ && RGB_PIXELSIZE == 4
+  __vector unsigned char rgb4 = { 0 };
+#endif
+  __vector short rg0, rg1, rg2, rg3, bg0, bg1, bg2, bg3;
+  __vector unsigned short yl, yh;
+  __vector int y0, y1, y2, y3;
+
+  /* Constants */
+  __vector short pw_f0299_f0337 = { __4X2(F_0_299, F_0_337) },
+    pw_f0114_f0250 = { __4X2(F_0_114, F_0_250) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index =
+      { 0, 1, 4, 5,  8,  9, 12, 13, 16, 17, 20, 21, 24, 25, 28, 29 };
+#else
+    shift_pack_index =
+      { 2, 3, 6, 7, 10, 11, 14, 15, 18, 19, 22, 23, 26, 27, 30, 31 };
+#endif
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf0output_row;
+    output_row++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, inptr += RGB_PIXELSIZE * 16,
+         outptr += 16) {
+
+#if __BIG_ENDIAN__
+      /* Load 16 pixels == 48 or 64 bytes */
+      offset = (size_t)inptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overread.  Since there is no way to
+           * read a partial AltiVec register, overread would occur on the last
+           * chunk of the last image row if the right edge is not on a 16-byte
+           * boundary.  It could also occur on other rows if the bytes per row
+           * is low enough.  Since we can't determine whether we're on the last
+           * image row, we have to assume every row is the last.
+           */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (bytes > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (bytes > 32)
+            rgb2 = vec_ld(32, inptr);
+          if (bytes > 48)
+            rgb3 = vec_ld(48, inptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            rgb4 = vec_ld(64, inptr);
+#endif
+          unaligned_shift_index = vec_lvsl(0, inptr);
+          rgb0 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          rgb1 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+          rgb2 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_perm(rgb3, rgb4, unaligned_shift_index);
+#endif
+        }
+      } else {
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (num_cols > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (num_cols > 32)
+            rgb2 = vec_ld(32, inptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            rgb3 = vec_ld(48, inptr);
+#endif
+        }
+      }
+#else
+      /* Little endian */
+      rgb0 = vec_vsx_ld(0, inptr);
+      if (num_cols > 16)
+        rgb1 = vec_vsx_ld(16, inptr);
+      if (num_cols > 32)
+        rgb2 = vec_vsx_ld(32, inptr);
+#if RGB_PIXELSIZE == 4
+      if (num_cols > 48)
+        rgb3 = vec_vsx_ld(48, inptr);
+#endif
+#endif
+
+#if RGB_PIXELSIZE == 3
+      /* rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX0);
+      rgbg1 = vec_perm(rgb0, rgb1, (__vector unsigned char)RGBG_INDEX1);
+      rgbg2 = vec_perm(rgb1, rgb2, (__vector unsigned char)RGBG_INDEX2);
+      rgbg3 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX3);
+#else
+      /* rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX);
+      rgbg1 = vec_perm(rgb1, rgb1, (__vector unsigned char)RGBG_INDEX);
+      rgbg2 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX);
+      rgbg3 = vec_perm(rgb3, rgb3, (__vector unsigned char)RGBG_INDEX);
+#endif
+
+      /* rg0 = R0 G0 R1 G1 R2 G2 R3 G3
+       * bg0 = B0 G0 B1 G1 B2 G2 B3 G3
+       * ...
+       *
+       * NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      rg0 = (__vector signed short)VEC_UNPACKHU(rgbg0);
+      bg0 = (__vector signed short)VEC_UNPACKLU(rgbg0);
+      rg1 = (__vector signed short)VEC_UNPACKHU(rgbg1);
+      bg1 = (__vector signed short)VEC_UNPACKLU(rgbg1);
+      rg2 = (__vector signed short)VEC_UNPACKHU(rgbg2);
+      bg2 = (__vector signed short)VEC_UNPACKLU(rgbg2);
+      rg3 = (__vector signed short)VEC_UNPACKHU(rgbg3);
+      bg3 = (__vector signed short)VEC_UNPACKLU(rgbg3);
+
+      /* (Original)
+       * Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+       *
+       * (This implementation)
+       * Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+       */
+
+      /* Calculate Y values */
+
+      y0 = vec_msums(rg0, pw_f0299_f0337, pd_onehalf);
+      y1 = vec_msums(rg1, pw_f0299_f0337, pd_onehalf);
+      y2 = vec_msums(rg2, pw_f0299_f0337, pd_onehalf);
+      y3 = vec_msums(rg3, pw_f0299_f0337, pd_onehalf);
+      y0 = vec_msums(bg0, pw_f0114_f0250, y0);
+      y1 = vec_msums(bg1, pw_f0114_f0250, y1);
+      y2 = vec_msums(bg2, pw_f0114_f0250, y2);
+      y3 = vec_msums(bg3, pw_f0114_f0250, y3);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      yl = vec_perm((__vector unsigned short)y0, (__vector unsigned short)y1,
+                    shift_pack_index);
+      yh = vec_perm((__vector unsigned short)y2, (__vector unsigned short)y3,
+                    shift_pack_index);
+      y = vec_pack(yl, yh);
+      vec_st(y, 0, outptr);
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jcsample-altivec.c Added

@@ -0,0 +1,159 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* CHROMA DOWNSAMPLING */
+
+#include "jsimd_altivec.h"
+#include "jcsample.h"
+
+
+void jsimd_h2v1_downsample_altivec(JDIMENSION image_width,
+                                   int max_v_samp_factor,
+                                   JDIMENSION v_samp_factor,
+                                   JDIMENSION width_in_blocks,
+                                   JSAMPARRAY input_data,
+                                   JSAMPARRAY output_data)
+{
+  int outrow, outcol;
+  JDIMENSION output_cols = width_in_blocks * DCTSIZE;
+  JSAMPROW inptr, outptr;
+
+  __vector unsigned char this0, next0, out;
+  __vector unsigned short this0e, this0o, next0e, next0o, outl, outh;
+
+  /* Constants */
+  __vector unsigned short pw_bias = { __4X2(0, 1) },
+    pw_one = { __8X(1) };
+  __vector unsigned char even_odd_index =
+    {  0,  2,  4,  6,  8, 10, 12, 14,  1,  3,  5,  7,  9, 11, 13, 15 },
+    pb_zero = { __16X(0) };
+
+  expand_right_edge(input_data, max_v_samp_factor, image_width,
+                    output_cols * 2);
+
+  for (outrow = 0; outrow < v_samp_factor; outrow++) {
+    outptr = output_dataoutrow;
+    inptr = input_dataoutrow;
+
+    for (outcol = output_cols; outcol > 0;
+         outcol -= 16, inptr += 32, outptr += 16) {
+
+      this0 = vec_ld(0, inptr);
+      this0 = vec_perm(this0, this0, even_odd_index);
+      this0e = (__vector unsigned short)VEC_UNPACKHU(this0);
+      this0o = (__vector unsigned short)VEC_UNPACKLU(this0);
+      outl = vec_add(this0e, this0o);
+      outl = vec_add(outl, pw_bias);
+      outl = vec_sr(outl, pw_one);
+
+      if (outcol > 8) {
+        next0 = vec_ld(16, inptr);
+        next0 = vec_perm(next0, next0, even_odd_index);
+        next0e = (__vector unsigned short)VEC_UNPACKHU(next0);
+        next0o = (__vector unsigned short)VEC_UNPACKLU(next0);
+        outh = vec_add(next0e, next0o);
+        outh = vec_add(outh, pw_bias);
+        outh = vec_sr(outh, pw_one);
+      } else
+        outh = vec_splat_u16(0);
+
+      out = vec_pack(outl, outh);
+      vec_st(out, 0, outptr);
+    }
+  }
+}
+
+
+void
+jsimd_h2v2_downsample_altivec(JDIMENSION image_width, int max_v_samp_factor,
+                              JDIMENSION v_samp_factor,
+                              JDIMENSION width_in_blocks,
+                              JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow, outcol;
+  JDIMENSION output_cols = width_in_blocks * DCTSIZE;
+  JSAMPROW inptr0, inptr1, outptr;
+
+  __vector unsigned char this0, next0, this1, next1, out;
+  __vector unsigned short this0e, this0o, next0e, next0o, this1e, this1o,
+    next1e, next1o, out0l, out0h, out1l, out1h, outl, outh;
+
+  /* Constants */
+  __vector unsigned short pw_bias = { __4X2(1, 2) },
+    pw_two = { __8X(2) };
+  __vector unsigned char even_odd_index =
+    {  0,  2,  4,  6,  8, 10, 12, 14,  1,  3,  5,  7,  9, 11, 13, 15 },
+    pb_zero = { __16X(0) };
+
+  expand_right_edge(input_data, max_v_samp_factor, image_width,
+                    output_cols * 2);
+
+  for (inrow = 0, outrow = 0; outrow < v_samp_factor;
+       inrow += 2, outrow++) {
+
+    inptr0 = input_datainrow;
+    inptr1 = input_datainrow + 1;
+    outptr = output_dataoutrow;
+
+    for (outcol = output_cols; outcol > 0;
+         outcol -= 16, inptr0 += 32, inptr1 += 32, outptr += 16) {
+
+      this0 = vec_ld(0, inptr0);
+      this0 = vec_perm(this0, this0, even_odd_index);
+      this0e = (__vector unsigned short)VEC_UNPACKHU(this0);
+      this0o = (__vector unsigned short)VEC_UNPACKLU(this0);
+      out0l = vec_add(this0e, this0o);
+
+      this1 = vec_ld(0, inptr1);
+      this1 = vec_perm(this1, this1, even_odd_index);
+      this1e = (__vector unsigned short)VEC_UNPACKHU(this1);
+      this1o = (__vector unsigned short)VEC_UNPACKLU(this1);
+      out1l = vec_add(this1e, this1o);
+
+      outl = vec_add(out0l, out1l);
+      outl = vec_add(outl, pw_bias);
+      outl = vec_sr(outl, pw_two);
+
+      if (outcol > 8) {
+        next0 = vec_ld(16, inptr0);
+        next0 = vec_perm(next0, next0, even_odd_index);
+        next0e = (__vector unsigned short)VEC_UNPACKHU(next0);
+        next0o = (__vector unsigned short)VEC_UNPACKLU(next0);
+        out0h = vec_add(next0e, next0o);
+
+        next1 = vec_ld(16, inptr1);
+        next1 = vec_perm(next1, next1, even_odd_index);
+        next1e = (__vector unsigned short)VEC_UNPACKHU(next1);
+        next1o = (__vector unsigned short)VEC_UNPACKLU(next1);
+        out1h = vec_add(next1e, next1o);
+
+        outh = vec_add(out0h, out1h);
+        outh = vec_add(outh, pw_bias);
+        outh = vec_sr(outh, pw_two);
+      } else
+        outh = vec_splat_u16(0);
+
+      out = vec_pack(outl, outh);
+      vec_st(out, 0, outptr);
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jcsample.h Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jdcolext-altivec.c Added

@@ -0,0 +1,276 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdcolor-altivec.c */
+
+
+void jsimd_ycc_rgb_convert_altivec(JDIMENSION out_width, JSAMPIMAGE input_buf,
+                                   JDIMENSION input_row, JSAMPARRAY output_buf,
+                                   int num_rows)
+{
+  JSAMPROW outptr, inptr0, inptr1, inptr2;
+  int pitch = out_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbufRGB_PIXELSIZE * 16;
+
+  __vector unsigned char rgb0, rgb1, rgb2, rgbx0, rgbx1, rgbx2, rgbx3,
+    y, cb, cr;
+#if __BIG_ENDIAN__
+  __vector unsigned char edgel, edgeh, edges, out0, out1, out2, out3;
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char out4;
+#endif
+#endif
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3;
+#endif
+  __vector short rg0, rg1, rg2, rg3, bx0, bx1, bx2, bx3, yl, yh, cbl, cbh,
+    crl, crh, rl, rh, gl, gh, bl, bh, g0w, g1w, g2w, g3w;
+  __vector int g0, g1, g2, g3;
+
+  /* Constants
+   * NOTE: The >> 1 is to compensate for the fact that vec_madds() returns 17
+   * high-order bits, not 16.
+   */
+  __vector short pw_f0402 = { __8X(F_0_402 >> 1) },
+    pw_mf0228 = { __8X(-F_0_228 >> 1) },
+    pw_mf0344_f0285 = { __4X2(-F_0_344, F_0_285) },
+    pw_one = { __8X(1) }, pw_255 = { __8X(255) },
+    pw_cj = { __8X(CENTERJSAMPLE) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index =
+      {  0,  1,  4,  5,  8,  9, 12, 13, 16, 17, 20, 21, 24, 25, 28, 29 };
+#else
+    shift_pack_index =
+      {  2,  3,  6,  7, 10, 11, 14, 15, 18, 19, 22, 23, 26, 27, 30, 31 };
+#endif
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf0input_row;
+    inptr1 = input_buf1input_row;
+    inptr2 = input_buf2input_row;
+    input_row++;
+    outptr = *output_buf++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, outptr += RGB_PIXELSIZE * 16,
+         inptr0 += 16, inptr1 += 16, inptr2 += 16) {
+
+      y = vec_ld(0, inptr0);
+      /* NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      yl = (__vector signed short)VEC_UNPACKHU(y);
+      yh = (__vector signed short)VEC_UNPACKLU(y);
+
+      cb = vec_ld(0, inptr1);
+      cbl = (__vector signed short)VEC_UNPACKHU(cb);
+      cbh = (__vector signed short)VEC_UNPACKLU(cb);
+      cbl = vec_sub(cbl, pw_cj);
+      cbh = vec_sub(cbh, pw_cj);
+
+      cr = vec_ld(0, inptr2);
+      crl = (__vector signed short)VEC_UNPACKHU(cr);
+      crh = (__vector signed short)VEC_UNPACKLU(cr);
+      crl = vec_sub(crl, pw_cj);
+      crh = vec_sub(crh, pw_cj);
+
+      /* (Original)
+       * R = Y                + 1.40200 * Cr
+       * G = Y - 0.34414 * Cb - 0.71414 * Cr
+       * B = Y + 1.77200 * Cb
+       *
+       * (This implementation)
+       * R = Y                + 0.40200 * Cr + Cr
+       * G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+       * B = Y - 0.22800 * Cb + Cb + Cb
+       */
+      bl = vec_add(cbl, cbl);
+      bh = vec_add(cbh, cbh);
+      bl = vec_madds(bl, pw_mf0228, pw_one);
+      bh = vec_madds(bh, pw_mf0228, pw_one);
+      bl = vec_sra(bl, (__vector unsigned short)pw_one);
+      bh = vec_sra(bh, (__vector unsigned short)pw_one);
+      bl = vec_add(bl, cbl);
+      bh = vec_add(bh, cbh);
+      bl = vec_add(bl, cbl);
+      bh = vec_add(bh, cbh);
+      bl = vec_add(bl, yl);
+      bh = vec_add(bh, yh);
+
+      rl = vec_add(crl, crl);
+      rh = vec_add(crh, crh);
+      rl = vec_madds(rl, pw_f0402, pw_one);
+      rh = vec_madds(rh, pw_f0402, pw_one);
+      rl = vec_sra(rl, (__vector unsigned short)pw_one);
+      rh = vec_sra(rh, (__vector unsigned short)pw_one);
+      rl = vec_add(rl, crl);
+      rh = vec_add(rh, crh);
+      rl = vec_add(rl, yl);
+      rh = vec_add(rh, yh);
+
+      g0w = vec_mergeh(cbl, crl);
+      g1w = vec_mergel(cbl, crl);
+      g0 = vec_msums(g0w, pw_mf0344_f0285, pd_onehalf);
+      g1 = vec_msums(g1w, pw_mf0344_f0285, pd_onehalf);
+      g2w = vec_mergeh(cbh, crh);
+      g3w = vec_mergel(cbh, crh);
+      g2 = vec_msums(g2w, pw_mf0344_f0285, pd_onehalf);
+      g3 = vec_msums(g3w, pw_mf0344_f0285, pd_onehalf);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      gl = vec_perm((__vector short)g0, (__vector short)g1, shift_pack_index);
+      gh = vec_perm((__vector short)g2, (__vector short)g3, shift_pack_index);
+      gl = vec_sub(gl, crl);
+      gh = vec_sub(gh, crh);
+      gl = vec_add(gl, yl);
+      gh = vec_add(gh, yh);
+
+      rg0 = vec_mergeh(rl, gl);
+      bx0 = vec_mergeh(bl, pw_255);
+      rg1 = vec_mergel(rl, gl);
+      bx1 = vec_mergel(bl, pw_255);
+      rg2 = vec_mergeh(rh, gh);
+      bx2 = vec_mergeh(bh, pw_255);
+      rg3 = vec_mergel(rh, gh);
+      bx3 = vec_mergel(bh, pw_255);
+
+      rgbx0 = vec_packsu(rg0, bx0);
+      rgbx1 = vec_packsu(rg1, bx1);
+      rgbx2 = vec_packsu(rg2, bx2);
+      rgbx3 = vec_packsu(rg3, bx3);
+
+#if RGB_PIXELSIZE == 3
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx1, (__vector unsigned char)RGB_INDEX0);
+      rgb1 = vec_perm(rgbx1, rgbx2, (__vector unsigned char)RGB_INDEX1);
+      rgb2 = vec_perm(rgbx2, rgbx3, (__vector unsigned char)RGB_INDEX2);
+#else
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx0, (__vector unsigned char)RGB_INDEX);
+      rgb1 = vec_perm(rgbx1, rgbx1, (__vector unsigned char)RGB_INDEX);
+      rgb2 = vec_perm(rgbx2, rgbx2, (__vector unsigned char)RGB_INDEX);
+      rgb3 = vec_perm(rgbx3, rgbx3, (__vector unsigned char)RGB_INDEX);
+#endif
+
+#if __BIG_ENDIAN__
+      offset = (size_t)outptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overwrite.  Since there is no way to
+           * write a partial AltiVec register, overwrite would occur on the
+           * last chunk of the last image row if the right edge is not on a
+           * 16-byte boundary.  It could also occur on other rows if the bytes
+           * per row is low enough.  Since we can't determine whether we're on
+           * the last image row, we have to assume every row is the last.
+           */
+          vec_st(rgb0, 0, tmpbuf);
+          vec_st(rgb1, 16, tmpbuf);
+          vec_st(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          vec_st(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          unaligned_shift_index = vec_lvsl(0, outptr);
+          edgel = vec_ld(0, outptr);
+          edgeh = vec_ld(min(num_cols - 1, RGB_PIXELSIZE * 16), outptr);
+          edges = vec_perm(edgeh, edgel, unaligned_shift_index);
+          unaligned_shift_index = vec_lvsr(0, outptr);
+          out0 = vec_perm(edges, rgb0, unaligned_shift_index);
+          out1 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          out2 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          out3 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+          out4 = vec_perm(rgb3, edges, unaligned_shift_index);
+#else
+          out3 = vec_perm(rgb2, edges, unaligned_shift_index);
+#endif
+          vec_st(out0, 0, outptr);
+          if (bytes > 16)
+            vec_st(out1, 16, outptr);
+          if (bytes > 32)
+            vec_st(out2, 32, outptr);
+          if (bytes > 48)
+            vec_st(out3, 48, outptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            vec_st(out4, 64, outptr);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          VEC_ST(rgb0, 0, tmpbuf);
+          VEC_ST(rgb1, 16, tmpbuf);
+          VEC_ST(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          VEC_ST(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          VEC_ST(rgb0, 0, outptr);
+          if (num_cols > 16)
+            VEC_ST(rgb1, 16, outptr);
+          if (num_cols > 32)
+            VEC_ST(rgb2, 32, outptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            VEC_ST(rgb3, 48, outptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jdcolor-altivec.c Added

@@ -0,0 +1,106 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* YCC --> RGB CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_344  22554              /* FIX(0.34414) */
+#define F_0_714  46802              /* FIX(0.71414) */
+#define F_1_402  91881              /* FIX(1.40200) */
+#define F_1_772  116130             /* FIX(1.77200) */
+#define F_0_402  (F_1_402 - 65536)  /* FIX(1.40200) - FIX(1) */
+#define F_0_285  (65536 - F_0_714)  /* FIX(1) - FIX(0.71414) */
+#define F_0_228  (131072 - F_1_772) /* FIX(2) - FIX(1.77200) */
+
+#define SCALEBITS  16
+#define ONE_HALF  (1 << (SCALEBITS - 1))
+
+#define RGB_INDEX0 \
+  {  0,  1,  8,  2,  3, 10,  4,  5, 12,  6,  7, 14, 16, 17, 24, 18 }
+#define RGB_INDEX1 \
+  {  3, 10,  4,  5, 12,  6,  7, 14, 16, 17, 24, 18, 19, 26, 20, 21 }
+#define RGB_INDEX2 \
+  { 12,  6,  7, 14, 16, 17, 24, 18, 19, 26, 20, 21, 28, 22, 23, 30 }
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_ycc_rgb_convert_altivec  jsimd_ycc_extrgb_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define RGB_INDEX \
+  {  0,  1,  8,  9,  2,  3, 10, 11,  4,  5, 12, 13,  6,  7, 14, 15 }
+#define jsimd_ycc_rgb_convert_altivec  jsimd_ycc_extrgbx_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define RGB_INDEX0 \
+  {  8,  1,  0, 10,  3,  2, 12,  5,  4, 14,  7,  6, 24, 17, 16, 26 }
+#define RGB_INDEX1 \
+  {  3,  2, 12,  5,  4, 14,  7,  6, 24, 17, 16, 26, 19, 18, 28, 21 }
+#define RGB_INDEX2 \
+  {  4, 14,  7,  6, 24, 17, 16, 26, 19, 18, 28, 21, 20, 30, 23, 22 }
+#define jsimd_ycc_rgb_convert_altivec  jsimd_ycc_extbgr_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define RGB_INDEX \
+  {  8,  1,  0,  9, 10,  3,  2, 11, 12,  5,  4, 13, 14,  7,  6, 15 }
+#define jsimd_ycc_rgb_convert_altivec  jsimd_ycc_extbgrx_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define RGB_INDEX \
+  {  9,  8,  1,  0, 11, 10,  3,  2, 13, 12,  5,  4, 15, 14,  7,  6 }
+#define jsimd_ycc_rgb_convert_altivec  jsimd_ycc_extxbgr_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define RGB_INDEX \
+  {  9,  0,  1,  8, 11,  2,  3, 10, 13,  4,  5, 12, 15,  6,  7, 14 }
+#define jsimd_ycc_rgb_convert_altivec  jsimd_ycc_extxrgb_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jdmerge-altivec.c Added

@@ -0,0 +1,130 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* MERGED YCC --> RGB CONVERSION AND UPSAMPLING */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_344  22554              /* FIX(0.34414) */
+#define F_0_714  46802              /* FIX(0.71414) */
+#define F_1_402  91881              /* FIX(1.40200) */
+#define F_1_772  116130             /* FIX(1.77200) */
+#define F_0_402  (F_1_402 - 65536)  /* FIX(1.40200) - FIX(1) */
+#define F_0_285  (65536 - F_0_714)  /* FIX(1) - FIX(0.71414) */
+#define F_0_228  (131072 - F_1_772) /* FIX(2) - FIX(1.77200) */
+
+#define SCALEBITS  16
+#define ONE_HALF  (1 << (SCALEBITS - 1))
+
+#define RGB_INDEX0 \
+  {  0,  1,  8,  2,  3, 10,  4,  5, 12,  6,  7, 14, 16, 17, 24, 18 }
+#define RGB_INDEX1 \
+  {  3, 10,  4,  5, 12,  6,  7, 14, 16, 17, 24, 18, 19, 26, 20, 21 }
+#define RGB_INDEX2 \
+  { 12,  6,  7, 14, 16, 17, 24, 18, 19, 26, 20, 21, 28, 22, 23, 30 }
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_altivec \
+  jsimd_h2v1_extrgb_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec \
+  jsimd_h2v2_extrgb_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+#define RGB_INDEX \
+  {  0,  1,  8,  9,  2,  3, 10, 11,  4,  5, 12, 13,  6,  7, 14, 15 }
+#define jsimd_h2v1_merged_upsample_altivec \
+  jsimd_h2v1_extrgbx_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec \
+  jsimd_h2v2_extrgbx_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+#define RGB_INDEX0 \
+  {  8,  1,  0, 10,  3,  2, 12,  5,  4, 14,  7,  6, 24, 17, 16, 26 }
+#define RGB_INDEX1 \
+  {  3,  2, 12,  5,  4, 14,  7,  6, 24, 17, 16, 26, 19, 18, 28, 21 }
+#define RGB_INDEX2 \
+  {  4, 14,  7,  6, 24, 17, 16, 26, 19, 18, 28, 21, 20, 30, 23, 22 }
+#define jsimd_h2v1_merged_upsample_altivec \
+  jsimd_h2v1_extbgr_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec \
+  jsimd_h2v2_extbgr_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+#define RGB_INDEX \
+  {  8,  1,  0,  9, 10,  3,  2, 11, 12,  5,  4, 13, 14,  7,  6, 15 }
+#define jsimd_h2v1_merged_upsample_altivec \
+  jsimd_h2v1_extbgrx_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec \
+  jsimd_h2v2_extbgrx_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+#define RGB_INDEX \
+  {  9,  8,  1,  0, 11, 10,  3,  2, 13, 12,  5,  4, 15, 14,  7,  6 }
+#define jsimd_h2v1_merged_upsample_altivec \
+  jsimd_h2v1_extxbgr_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec \
+  jsimd_h2v2_extxbgr_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+#define RGB_INDEX \
+  {  9,  0,  1,  8, 11,  2,  3, 10, 13,  4,  5, 12, 15,  6,  7, 14 }
+#define jsimd_h2v1_merged_upsample_altivec \
+  jsimd_h2v1_extxrgb_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec \
+  jsimd_h2v2_extxrgb_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jdmrgext-altivec.c Added

@@ -0,0 +1,329 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdmerge-altivec.c */
+
+
+void jsimd_h2v1_merged_upsample_altivec(JDIMENSION output_width,
+                                        JSAMPIMAGE input_buf,
+                                        JDIMENSION in_row_group_ctr,
+                                        JSAMPARRAY output_buf)
+{
+  JSAMPROW outptr, inptr0, inptr1, inptr2;
+  int pitch = output_width * RGB_PIXELSIZE, num_cols, yloop;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbufRGB_PIXELSIZE * 16;
+
+  __vector unsigned char rgb0, rgb1, rgb2, rgbx0, rgbx1, rgbx2, rgbx3,
+    y, cb, cr;
+#if __BIG_ENDIAN__
+  __vector unsigned char edgel, edgeh, edges, out0, out1, out2, out3;
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char out4;
+#endif
+#endif
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3;
+#endif
+  __vector short rg0, rg1, rg2, rg3, bx0, bx1, bx2, bx3, ye, yo, cbl, cbh,
+    crl, crh, r_yl, r_yh, g_yl, g_yh, b_yl, b_yh, g_y0w, g_y1w, g_y2w, g_y3w,
+    rl, rh, gl, gh, bl, bh, re, ro, ge, go, be, bo;
+  __vector int g_y0, g_y1, g_y2, g_y3;
+
+  /* Constants
+   * NOTE: The >> 1 is to compensate for the fact that vec_madds() returns 17
+   * high-order bits, not 16.
+   */
+  __vector short pw_f0402 = { __8X(F_0_402 >> 1) },
+    pw_mf0228 = { __8X(-F_0_228 >> 1) },
+    pw_mf0344_f0285 = { __4X2(-F_0_344, F_0_285) },
+    pw_one = { __8X(1) }, pw_255 = { __8X(255) },
+    pw_cj = { __8X(CENTERJSAMPLE) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index =
+      {  0,  1,  4,  5,  8,  9, 12, 13, 16, 17, 20, 21, 24, 25, 28, 29 },
+    even_index =
+      {  0, 16,  0, 18,  0, 20,  0, 22,  0, 24,  0, 26,  0, 28,  0, 30 },
+    odd_index =
+      {  0, 17,  0, 19,  0, 21,  0, 23,  0, 25,  0, 27,  0, 29,  0, 31 };
+#else
+    shift_pack_index =
+      {  2,  3,  6,  7, 10, 11, 14, 15, 18, 19, 22, 23, 26, 27, 30, 31 },
+    even_index =
+      { 16,  0, 18,  0, 20,  0, 22,  0, 24,  0, 26,  0, 28,  0, 30,  0 },
+    odd_index =
+      { 17,  0, 19,  0, 21,  0, 23,  0, 25,  0, 27,  0, 29,  0, 31,  0 };
+#endif
+
+  inptr0 = input_buf0in_row_group_ctr;
+  inptr1 = input_buf1in_row_group_ctr;
+  inptr2 = input_buf2in_row_group_ctr;
+  outptr = output_buf0;
+
+  for (num_cols = pitch; num_cols > 0; inptr1 += 16, inptr2 += 16) {
+
+    cb = vec_ld(0, inptr1);
+    /* NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+     * support unsigned vectors.
+     */
+    cbl = (__vector signed short)VEC_UNPACKHU(cb);
+    cbh = (__vector signed short)VEC_UNPACKLU(cb);
+    cbl = vec_sub(cbl, pw_cj);
+    cbh = vec_sub(cbh, pw_cj);
+
+    cr = vec_ld(0, inptr2);
+    crl = (__vector signed short)VEC_UNPACKHU(cr);
+    crh = (__vector signed short)VEC_UNPACKLU(cr);
+    crl = vec_sub(crl, pw_cj);
+    crh = vec_sub(crh, pw_cj);
+
+    /* (Original)
+     * R = Y                + 1.40200 * Cr
+     * G = Y - 0.34414 * Cb - 0.71414 * Cr
+     * B = Y + 1.77200 * Cb
+     *
+     * (This implementation)
+     * R = Y                + 0.40200 * Cr + Cr
+     * G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+     * B = Y - 0.22800 * Cb + Cb + Cb
+     */
+    b_yl = vec_add(cbl, cbl);
+    b_yh = vec_add(cbh, cbh);
+    b_yl = vec_madds(b_yl, pw_mf0228, pw_one);
+    b_yh = vec_madds(b_yh, pw_mf0228, pw_one);
+    b_yl = vec_sra(b_yl, (__vector unsigned short)pw_one);
+    b_yh = vec_sra(b_yh, (__vector unsigned short)pw_one);
+    b_yl = vec_add(b_yl, cbl);
+    b_yh = vec_add(b_yh, cbh);
+    b_yl = vec_add(b_yl, cbl);
+    b_yh = vec_add(b_yh, cbh);
+
+    r_yl = vec_add(crl, crl);
+    r_yh = vec_add(crh, crh);
+    r_yl = vec_madds(r_yl, pw_f0402, pw_one);
+    r_yh = vec_madds(r_yh, pw_f0402, pw_one);
+    r_yl = vec_sra(r_yl, (__vector unsigned short)pw_one);
+    r_yh = vec_sra(r_yh, (__vector unsigned short)pw_one);
+    r_yl = vec_add(r_yl, crl);
+    r_yh = vec_add(r_yh, crh);
+
+    g_y0w = vec_mergeh(cbl, crl);
+    g_y1w = vec_mergel(cbl, crl);
+    g_y0 = vec_msums(g_y0w, pw_mf0344_f0285, pd_onehalf);
+    g_y1 = vec_msums(g_y1w, pw_mf0344_f0285, pd_onehalf);
+    g_y2w = vec_mergeh(cbh, crh);
+    g_y3w = vec_mergel(cbh, crh);
+    g_y2 = vec_msums(g_y2w, pw_mf0344_f0285, pd_onehalf);
+    g_y3 = vec_msums(g_y3w, pw_mf0344_f0285, pd_onehalf);
+    /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+     * each dword into a new 16-bit vector, which is the equivalent of
+     * descaling the 32-bit results (right-shifting by 16 bits) and then
+     * packing them.
+     */
+    g_yl = vec_perm((__vector short)g_y0, (__vector short)g_y1,
+                    shift_pack_index);
+    g_yh = vec_perm((__vector short)g_y2, (__vector short)g_y3,
+                    shift_pack_index);
+    g_yl = vec_sub(g_yl, crl);
+    g_yh = vec_sub(g_yh, crh);
+
+    for (yloop = 0; yloop < 2 && num_cols > 0; yloop++,
+         num_cols -= RGB_PIXELSIZE * 16,
+         outptr += RGB_PIXELSIZE * 16, inptr0 += 16) {
+
+      y = vec_ld(0, inptr0);
+      ye = (__vector signed short)vec_perm(pb_zero, y, even_index);
+      yo = (__vector signed short)vec_perm(pb_zero, y, odd_index);
+
+      if (yloop == 0) {
+        be = vec_add(b_yl, ye);
+        bo = vec_add(b_yl, yo);
+        re = vec_add(r_yl, ye);
+        ro = vec_add(r_yl, yo);
+        ge = vec_add(g_yl, ye);
+        go = vec_add(g_yl, yo);
+      } else {
+        be = vec_add(b_yh, ye);
+        bo = vec_add(b_yh, yo);
+        re = vec_add(r_yh, ye);
+        ro = vec_add(r_yh, yo);
+        ge = vec_add(g_yh, ye);
+        go = vec_add(g_yh, yo);
+      }
+
+      rl = vec_mergeh(re, ro);
+      rh = vec_mergel(re, ro);
+      gl = vec_mergeh(ge, go);
+      gh = vec_mergel(ge, go);
+      bl = vec_mergeh(be, bo);
+      bh = vec_mergel(be, bo);
+
+      rg0 = vec_mergeh(rl, gl);
+      bx0 = vec_mergeh(bl, pw_255);
+      rg1 = vec_mergel(rl, gl);
+      bx1 = vec_mergel(bl, pw_255);
+      rg2 = vec_mergeh(rh, gh);
+      bx2 = vec_mergeh(bh, pw_255);
+      rg3 = vec_mergel(rh, gh);
+      bx3 = vec_mergel(bh, pw_255);
+
+      rgbx0 = vec_packsu(rg0, bx0);
+      rgbx1 = vec_packsu(rg1, bx1);
+      rgbx2 = vec_packsu(rg2, bx2);
+      rgbx3 = vec_packsu(rg3, bx3);
+
+#if RGB_PIXELSIZE == 3
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx1, (__vector unsigned char)RGB_INDEX0);
+      rgb1 = vec_perm(rgbx1, rgbx2, (__vector unsigned char)RGB_INDEX1);
+      rgb2 = vec_perm(rgbx2, rgbx3, (__vector unsigned char)RGB_INDEX2);
+#else
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx0, (__vector unsigned char)RGB_INDEX);
+      rgb1 = vec_perm(rgbx1, rgbx1, (__vector unsigned char)RGB_INDEX);
+      rgb2 = vec_perm(rgbx2, rgbx2, (__vector unsigned char)RGB_INDEX);
+      rgb3 = vec_perm(rgbx3, rgbx3, (__vector unsigned char)RGB_INDEX);
+#endif
+
+#if __BIG_ENDIAN__
+      offset = (size_t)outptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overwrite.  Since there is no way to
+           * write a partial AltiVec register, overwrite would occur on the
+           * last chunk of the last image row if the right edge is not on a
+           * 16-byte boundary.  It could also occur on other rows if the bytes
+           * per row is low enough.  Since we can't determine whether we're on
+           * the last image row, we have to assume every row is the last.
+           */
+          vec_st(rgb0, 0, tmpbuf);
+          vec_st(rgb1, 16, tmpbuf);
+          vec_st(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          vec_st(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          unaligned_shift_index = vec_lvsl(0, outptr);
+          edgel = vec_ld(0, outptr);
+          edgeh = vec_ld(min(num_cols - 1, RGB_PIXELSIZE * 16), outptr);
+          edges = vec_perm(edgeh, edgel, unaligned_shift_index);
+          unaligned_shift_index = vec_lvsr(0, outptr);
+          out0 = vec_perm(edges, rgb0, unaligned_shift_index);
+          out1 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          out2 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          out3 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+          out4 = vec_perm(rgb3, edges, unaligned_shift_index);
+#else
+          out3 = vec_perm(rgb2, edges, unaligned_shift_index);
+#endif
+          vec_st(out0, 0, outptr);
+          if (bytes > 16)
+            vec_st(out1, 16, outptr);
+          if (bytes > 32)
+            vec_st(out2, 32, outptr);
+          if (bytes > 48)
+            vec_st(out3, 48, outptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            vec_st(out4, 64, outptr);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          VEC_ST(rgb0, 0, tmpbuf);
+          VEC_ST(rgb1, 16, tmpbuf);
+          VEC_ST(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          VEC_ST(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          VEC_ST(rgb0, 0, outptr);
+          if (num_cols > 16)
+            VEC_ST(rgb1, 16, outptr);
+          if (num_cols > 32)
+            VEC_ST(rgb2, 32, outptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            VEC_ST(rgb3, 48, outptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+    }
+  }
+}
+
+
+void jsimd_h2v2_merged_upsample_altivec(JDIMENSION output_width,
+                                        JSAMPIMAGE input_buf,
+                                        JDIMENSION in_row_group_ctr,
+                                        JSAMPARRAY output_buf)
+{
+  JSAMPROW inptr, outptr;
+
+  inptr = input_buf0in_row_group_ctr;
+  outptr = output_buf0;
+
+  input_buf0in_row_group_ctr = input_buf0in_row_group_ctr * 2;
+  jsimd_h2v1_merged_upsample_altivec(output_width, input_buf, in_row_group_ctr,
+                                     output_buf);
+
+  input_buf0in_row_group_ctr = input_buf0in_row_group_ctr * 2 + 1;
+  output_buf0 = output_buf1;
+  jsimd_h2v1_merged_upsample_altivec(output_width, input_buf, in_row_group_ctr,
+                                     output_buf);
+
+  input_buf0in_row_group_ctr = inptr;
+  output_buf0 = outptr;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jdsample-altivec.c Added

@@ -0,0 +1,400 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* CHROMA UPSAMPLING */
+
+#include "jsimd_altivec.h"
+
+
+void jsimd_h2v1_fancy_upsample_altivec(int max_v_samp_factor,
+                                       JDIMENSION downsampled_width,
+                                       JSAMPARRAY input_data,
+                                       JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr;
+  int inrow, incol;
+
+  __vector unsigned char this0, last0, p_last0, next0 = { 0 }, p_next0,
+    out;
+  __vector short this0e, this0o, this0l, this0h, last0l, last0h,
+    next0l, next0h, outle, outhe, outlo, outho;
+
+  /* Constants */
+  __vector unsigned char pb_zero = { __16X(0) }, pb_three = { __16X(3) },
+    last_index_col0 =
+      {  0,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14 },
+    last_index =
+      { 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30 },
+    next_index =
+      {  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16 },
+    next_index_lastcol =
+      {  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 15 },
+#if __BIG_ENDIAN__
+    merge_pack_index =
+      {  1, 17,  3, 19,  5, 21,  7, 23,  9, 25, 11, 27, 13, 29, 15, 31 };
+#else
+    merge_pack_index =
+      {  0, 16,  2, 18,  4, 20,  6, 22,  8, 24, 10, 26, 12, 28, 14, 30 };
+#endif
+  __vector short pw_one = { __8X(1) }, pw_two = { __8X(2) };
+
+  for (inrow = 0; inrow < max_v_samp_factor; inrow++) {
+    inptr = input_datainrow;
+    outptr = output_datainrow;
+
+    if (downsampled_width & 15)
+      inptrdownsampled_width = inptrdownsampled_width - 1;
+
+    this0 = vec_ld(0, inptr);
+    p_last0 = vec_perm(this0, this0, last_index_col0);
+    last0 = this0;
+
+    for (incol = downsampled_width; incol > 0;
+         incol -= 16, inptr += 16, outptr += 32) {
+
+      if (downsampled_width - incol > 0) {
+        p_last0 = vec_perm(last0, this0, last_index);
+        last0 = this0;
+      }
+
+      if (incol <= 16)
+        p_next0 = vec_perm(this0, this0, next_index_lastcol);
+      else {
+        next0 = vec_ld(16, inptr);
+        p_next0 = vec_perm(this0, next0, next_index);
+      }
+
+      this0e = (__vector short)vec_mule(this0, pb_three);
+      this0o = (__vector short)vec_mulo(this0, pb_three);
+      this0l = vec_mergeh(this0e, this0o);
+      this0h = vec_mergel(this0e, this0o);
+
+      last0l = (__vector short)VEC_UNPACKHU(p_last0);
+      last0h = (__vector short)VEC_UNPACKLU(p_last0);
+      last0l = vec_add(last0l, pw_one);
+
+      next0l = (__vector short)VEC_UNPACKHU(p_next0);
+      next0h = (__vector short)VEC_UNPACKLU(p_next0);
+      next0l = vec_add(next0l, pw_two);
+
+      outle = vec_add(this0l, last0l);
+      outlo = vec_add(this0l, next0l);
+      outle = vec_sr(outle, (__vector unsigned short)pw_two);
+      outlo = vec_sr(outlo, (__vector unsigned short)pw_two);
+
+      out = vec_perm((__vector unsigned char)outle,
+                     (__vector unsigned char)outlo, merge_pack_index);
+      vec_st(out, 0, outptr);
+
+      if (incol > 8) {
+        last0h = vec_add(last0h, pw_one);
+        next0h = vec_add(next0h, pw_two);
+
+        outhe = vec_add(this0h, last0h);
+        outho = vec_add(this0h, next0h);
+        outhe = vec_sr(outhe, (__vector unsigned short)pw_two);
+        outho = vec_sr(outho, (__vector unsigned short)pw_two);
+
+        out = vec_perm((__vector unsigned char)outhe,
+                       (__vector unsigned char)outho, merge_pack_index);
+        vec_st(out, 16, outptr);
+      }
+
+      this0 = next0;
+    }
+  }
+}
+
+
+void jsimd_h2v2_fancy_upsample_altivec(int max_v_samp_factor,
+                                       JDIMENSION downsampled_width,
+                                       JSAMPARRAY input_data,
+                                       JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr_1, inptr0, inptr1, outptr0, outptr1;
+  int inrow, outrow, incol;
+
+  __vector unsigned char this_1, this0, this1, out;
+  __vector short this_1l, this_1h, this0l, this0h, this1l, this1h,
+    lastcolsum_1h, lastcolsum1h,
+    p_lastcolsum_1l, p_lastcolsum_1h, p_lastcolsum1l, p_lastcolsum1h,
+    thiscolsum_1l, thiscolsum_1h, thiscolsum1l, thiscolsum1h,
+    nextcolsum_1l = { 0 }, nextcolsum_1h = { 0 },
+    nextcolsum1l = { 0 }, nextcolsum1h = { 0 },
+    p_nextcolsum_1l, p_nextcolsum_1h, p_nextcolsum1l, p_nextcolsum1h,
+    tmpl, tmph, outle, outhe, outlo, outho;
+
+  /* Constants */
+  __vector unsigned char pb_zero = { __16X(0) },
+    last_index_col0 =
+      {  0,  1,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13 },
+    last_index =
+      { 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29 },
+    next_index =
+      {  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17 },
+    next_index_lastcol =
+      {  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 14, 15 },
+#if __BIG_ENDIAN__
+    merge_pack_index =
+      {  1, 17,  3, 19,  5, 21,  7, 23,  9, 25, 11, 27, 13, 29, 15, 31 };
+#else
+    merge_pack_index =
+      {  0, 16,  2, 18,  4, 20,  6, 22,  8, 24, 10, 26, 12, 28, 14, 30 };
+#endif
+  __vector short pw_zero = { __8X(0) }, pw_three = { __8X(3) },
+    pw_seven = { __8X(7) }, pw_eight = { __8X(8) };
+  __vector unsigned short pw_four = { __8X(4) };
+
+  for (inrow = 0, outrow = 0; outrow < max_v_samp_factor; inrow++) {
+
+    inptr_1 = input_datainrow - 1;
+    inptr0 = input_datainrow;
+    inptr1 = input_datainrow + 1;
+    outptr0 = output_dataoutrow++;
+    outptr1 = output_dataoutrow++;
+
+    if (downsampled_width & 15) {
+      inptr_1downsampled_width = inptr_1downsampled_width - 1;
+      inptr0downsampled_width = inptr0downsampled_width - 1;
+      inptr1downsampled_width = inptr1downsampled_width - 1;
+    }
+
+    this0 = vec_ld(0, inptr0);
+    this0l = (__vector short)VEC_UNPACKHU(this0);
+    this0h = (__vector short)VEC_UNPACKLU(this0);
+    this0l = vec_mladd(this0l, pw_three, pw_zero);
+    this0h = vec_mladd(this0h, pw_three, pw_zero);
+
+    this_1 = vec_ld(0, inptr_1);
+    this_1l = (__vector short)VEC_UNPACKHU(this_1);
+    this_1h = (__vector short)VEC_UNPACKLU(this_1);
+    thiscolsum_1l = vec_add(this0l, this_1l);
+    thiscolsum_1h = vec_add(this0h, this_1h);
+    lastcolsum_1h = thiscolsum_1h;
+    p_lastcolsum_1l = vec_perm(thiscolsum_1l, thiscolsum_1l, last_index_col0);
+    p_lastcolsum_1h = vec_perm(thiscolsum_1l, thiscolsum_1h, last_index);
+
+    this1 = vec_ld(0, inptr1);
+    this1l = (__vector short)VEC_UNPACKHU(this1);
+    this1h = (__vector short)VEC_UNPACKLU(this1);
+    thiscolsum1l = vec_add(this0l, this1l);
+    thiscolsum1h = vec_add(this0h, this1h);
+    lastcolsum1h = thiscolsum1h;
+    p_lastcolsum1l = vec_perm(thiscolsum1l, thiscolsum1l, last_index_col0);
+    p_lastcolsum1h = vec_perm(thiscolsum1l, thiscolsum1h, last_index);
+
+    for (incol = downsampled_width; incol > 0;
+         incol -= 16, inptr_1 += 16, inptr0 += 16, inptr1 += 16,
+         outptr0 += 32, outptr1 += 32) {
+
+      if (downsampled_width - incol > 0) {
+        p_lastcolsum_1l = vec_perm(lastcolsum_1h, thiscolsum_1l, last_index);
+        p_lastcolsum_1h = vec_perm(thiscolsum_1l, thiscolsum_1h, last_index);
+        p_lastcolsum1l = vec_perm(lastcolsum1h, thiscolsum1l, last_index);
+        p_lastcolsum1h = vec_perm(thiscolsum1l, thiscolsum1h, last_index);
+        lastcolsum_1h = thiscolsum_1h;  lastcolsum1h = thiscolsum1h;
+      }
+
+      if (incol <= 16) {
+        p_nextcolsum_1l = vec_perm(thiscolsum_1l, thiscolsum_1h, next_index);
+        p_nextcolsum_1h = vec_perm(thiscolsum_1h, thiscolsum_1h,
+                                   next_index_lastcol);
+        p_nextcolsum1l = vec_perm(thiscolsum1l, thiscolsum1h, next_index);
+        p_nextcolsum1h = vec_perm(thiscolsum1h, thiscolsum1h,
+                                  next_index_lastcol);
+      } else {
+        this0 = vec_ld(16, inptr0);
+        this0l = (__vector short)VEC_UNPACKHU(this0);
+        this0h = (__vector short)VEC_UNPACKLU(this0);
+        this0l = vec_mladd(this0l, pw_three, pw_zero);
+        this0h = vec_mladd(this0h, pw_three, pw_zero);
+
+        this_1 = vec_ld(16, inptr_1);
+        this_1l = (__vector short)VEC_UNPACKHU(this_1);
+        this_1h = (__vector short)VEC_UNPACKLU(this_1);
+        nextcolsum_1l = vec_add(this0l, this_1l);
+        nextcolsum_1h = vec_add(this0h, this_1h);
+        p_nextcolsum_1l = vec_perm(thiscolsum_1l, thiscolsum_1h, next_index);
+        p_nextcolsum_1h = vec_perm(thiscolsum_1h, nextcolsum_1l, next_index);
+
+        this1 = vec_ld(16, inptr1);
+        this1l = (__vector short)VEC_UNPACKHU(this1);
+        this1h = (__vector short)VEC_UNPACKLU(this1);
+        nextcolsum1l = vec_add(this0l, this1l);
+        nextcolsum1h = vec_add(this0h, this1h);
+        p_nextcolsum1l = vec_perm(thiscolsum1l, thiscolsum1h, next_index);
+        p_nextcolsum1h = vec_perm(thiscolsum1h, nextcolsum1l, next_index);
+      }
+
+      /* Process the upper row */
+
+      tmpl = vec_mladd(thiscolsum_1l, pw_three, pw_zero);
+      outle = vec_add(tmpl, p_lastcolsum_1l);
+      outle = vec_add(outle, pw_eight);
+      outle = vec_sr(outle, pw_four);
+
+      outlo = vec_add(tmpl, p_nextcolsum_1l);
+      outlo = vec_add(outlo, pw_seven);
+      outlo = vec_sr(outlo, pw_four);
+
+      out = vec_perm((__vector unsigned char)outle,
+                     (__vector unsigned char)outlo, merge_pack_index);
+      vec_st(out, 0, outptr0);
+
+      if (incol > 8) {
+        tmph = vec_mladd(thiscolsum_1h, pw_three, pw_zero);
+        outhe = vec_add(tmph, p_lastcolsum_1h);
+        outhe = vec_add(outhe, pw_eight);
+        outhe = vec_sr(outhe, pw_four);
+
+        outho = vec_add(tmph, p_nextcolsum_1h);
+        outho = vec_add(outho, pw_seven);
+        outho = vec_sr(outho, pw_four);
+
+        out = vec_perm((__vector unsigned char)outhe,
+                       (__vector unsigned char)outho, merge_pack_index);
+        vec_st(out, 16, outptr0);
+      }
+
+      /* Process the lower row */
+
+      tmpl = vec_mladd(thiscolsum1l, pw_three, pw_zero);
+      outle = vec_add(tmpl, p_lastcolsum1l);
+      outle = vec_add(outle, pw_eight);
+      outle = vec_sr(outle, pw_four);
+
+      outlo = vec_add(tmpl, p_nextcolsum1l);
+      outlo = vec_add(outlo, pw_seven);
+      outlo = vec_sr(outlo, pw_four);
+
+      out = vec_perm((__vector unsigned char)outle,
+                     (__vector unsigned char)outlo, merge_pack_index);
+      vec_st(out, 0, outptr1);
+
+      if (incol > 8) {
+        tmph = vec_mladd(thiscolsum1h, pw_three, pw_zero);
+        outhe = vec_add(tmph, p_lastcolsum1h);
+        outhe = vec_add(outhe, pw_eight);
+        outhe = vec_sr(outhe, pw_four);
+
+        outho = vec_add(tmph, p_nextcolsum1h);
+        outho = vec_add(outho, pw_seven);
+        outho = vec_sr(outho, pw_four);
+
+        out = vec_perm((__vector unsigned char)outhe,
+                       (__vector unsigned char)outho, merge_pack_index);
+        vec_st(out, 16, outptr1);
+      }
+
+      thiscolsum_1l = nextcolsum_1l;  thiscolsum_1h = nextcolsum_1h;
+      thiscolsum1l = nextcolsum1l;  thiscolsum1h = nextcolsum1h;
+    }
+  }
+}
+
+
+/* These are rarely used (mainly just for decompressing YCCK images) */
+
+void jsimd_h2v1_upsample_altivec(int max_v_samp_factor,
+                                 JDIMENSION output_width,
+                                 JSAMPARRAY input_data,
+                                 JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr;
+  int inrow, incol;
+
+  __vector unsigned char in, inl, inh;
+
+  for (inrow = 0; inrow < max_v_samp_factor; inrow++) {
+    inptr = input_datainrow;
+    outptr = output_datainrow;
+
+    for (incol = (output_width + 31) & (~31); incol > 0;
+         incol -= 64, inptr += 32, outptr += 64) {
+
+      in = vec_ld(0, inptr);
+      inl = vec_mergeh(in, in);
+      inh = vec_mergel(in, in);
+
+      vec_st(inl, 0, outptr);
+      vec_st(inh, 16, outptr);
+
+      if (incol > 32) {
+        in = vec_ld(16, inptr);
+        inl = vec_mergeh(in, in);
+        inh = vec_mergel(in, in);
+
+        vec_st(inl, 32, outptr);
+        vec_st(inh, 48, outptr);
+      }
+    }
+  }
+}
+
+
+void jsimd_h2v2_upsample_altivec(int max_v_samp_factor,
+                                 JDIMENSION output_width,
+                                 JSAMPARRAY input_data,
+                                 JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr0, outptr1;
+  int inrow, outrow, incol;
+
+  __vector unsigned char in, inl, inh;
+
+  for (inrow = 0, outrow = 0; outrow < max_v_samp_factor; inrow++) {
+
+    inptr = input_datainrow;
+    outptr0 = output_dataoutrow++;
+    outptr1 = output_dataoutrow++;
+
+    for (incol = (output_width + 31) & (~31); incol > 0;
+         incol -= 64, inptr += 32, outptr0 += 64, outptr1 += 64) {
+
+      in = vec_ld(0, inptr);
+      inl = vec_mergeh(in, in);
+      inh = vec_mergel(in, in);
+
+      vec_st(inl, 0, outptr0);
+      vec_st(inl, 0, outptr1);
+
+      vec_st(inh, 16, outptr0);
+      vec_st(inh, 16, outptr1);
+
+      if (incol > 32) {
+        in = vec_ld(16, inptr);
+        inl = vec_mergeh(in, in);
+        inh = vec_mergel(in, in);
+
+        vec_st(inl, 32, outptr0);
+        vec_st(inl, 32, outptr1);
+
+        vec_st(inh, 48, outptr0);
+        vec_st(inh, 48, outptr1);
+      }
+    }
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jfdctfst-altivec.c Added

@@ -0,0 +1,154 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* FAST INTEGER FORWARD DCT
+ *
+ * This is similar to the SSE2 implementation, except that we left-shift the
+ * constants by 1 less bit (the -1 in CONST_SHIFT.)  This is because
+ * vec_madds(arg1, arg2, arg3) generates the 16-bit saturated sum of:
+ *   the elements in arg3 + the most significant 17 bits of
+ *     (the elements in arg1 * the elements in arg2).
+ */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_382  98   /* FIX(0.382683433) */
+#define F_0_541  139  /* FIX(0.541196100) */
+#define F_0_707  181  /* FIX(0.707106781) */
+#define F_1_306  334  /* FIX(1.306562965) */
+
+#define CONST_BITS  8
+#define PRE_MULTIPLY_SCALE_BITS  2
+#define CONST_SHIFT  (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS - 1)
+
+
+#define DO_FDCT() { \
+  /* Even part */ \
+  \
+  tmp10 = vec_add(tmp0, tmp3); \
+  tmp13 = vec_sub(tmp0, tmp3); \
+  tmp11 = vec_add(tmp1, tmp2); \
+  tmp12 = vec_sub(tmp1, tmp2); \
+  \
+  out0  = vec_add(tmp10, tmp11); \
+  out4  = vec_sub(tmp10, tmp11); \
+  \
+  z1 = vec_add(tmp12, tmp13); \
+  z1 = vec_sl(z1, pre_multiply_scale_bits); \
+  z1 = vec_madds(z1, pw_0707, pw_zero); \
+  \
+  out2 = vec_add(tmp13, z1); \
+  out6 = vec_sub(tmp13, z1); \
+  \
+  /* Odd part */ \
+  \
+  tmp10 = vec_add(tmp4, tmp5); \
+  tmp11 = vec_add(tmp5, tmp6); \
+  tmp12 = vec_add(tmp6, tmp7); \
+  \
+  tmp10 = vec_sl(tmp10, pre_multiply_scale_bits); \
+  tmp12 = vec_sl(tmp12, pre_multiply_scale_bits); \
+  z5 = vec_sub(tmp10, tmp12); \
+  z5 = vec_madds(z5, pw_0382, pw_zero); \
+  \
+  z2 = vec_madds(tmp10, pw_0541, z5); \
+  z4 = vec_madds(tmp12, pw_1306, z5); \
+  \
+  tmp11 = vec_sl(tmp11, pre_multiply_scale_bits); \
+  z3 = vec_madds(tmp11, pw_0707, pw_zero); \
+  \
+  z11 = vec_add(tmp7, z3); \
+  z13 = vec_sub(tmp7, z3); \
+  \
+  out5 = vec_add(z13, z2); \
+  out3 = vec_sub(z13, z2); \
+  out1 = vec_add(z11, z4); \
+  out7 = vec_sub(z11, z4); \
+}
+
+
+void jsimd_fdct_ifast_altivec(DCTELEM *data)
+{
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
+    z1, z2, z3, z4, z5, z11, z13,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+
+  /* Constants */
+  __vector short pw_zero = { __8X(0) },
+    pw_0382 = { __8X(F_0_382 << CONST_SHIFT) },
+    pw_0541 = { __8X(F_0_541 << CONST_SHIFT) },
+    pw_0707 = { __8X(F_0_707 << CONST_SHIFT) },
+    pw_1306 = { __8X(F_1_306 << CONST_SHIFT) };
+  __vector unsigned short
+    pre_multiply_scale_bits = { __8X(PRE_MULTIPLY_SCALE_BITS) };
+
+  /* Pass 1: process rows */
+
+  row0 = vec_ld(0, data);
+  row1 = vec_ld(16, data);
+  row2 = vec_ld(32, data);
+  row3 = vec_ld(48, data);
+  row4 = vec_ld(64, data);
+  row5 = vec_ld(80, data);
+  row6 = vec_ld(96, data);
+  row7 = vec_ld(112, data);
+
+  TRANSPOSE(row, col);
+
+  tmp0 = vec_add(col0, col7);
+  tmp7 = vec_sub(col0, col7);
+  tmp1 = vec_add(col1, col6);
+  tmp6 = vec_sub(col1, col6);
+  tmp2 = vec_add(col2, col5);
+  tmp5 = vec_sub(col2, col5);
+  tmp3 = vec_add(col3, col4);
+  tmp4 = vec_sub(col3, col4);
+
+  DO_FDCT();
+
+  /* Pass 2: process columns */
+
+  TRANSPOSE(out, row);
+
+  tmp0 = vec_add(row0, row7);
+  tmp7 = vec_sub(row0, row7);
+  tmp1 = vec_add(row1, row6);
+  tmp6 = vec_sub(row1, row6);
+  tmp2 = vec_add(row2, row5);
+  tmp5 = vec_sub(row2, row5);
+  tmp3 = vec_add(row3, row4);
+  tmp4 = vec_sub(row3, row4);
+
+  DO_FDCT();
+
+  vec_st(out0, 0, data);
+  vec_st(out1, 16, data);
+  vec_st(out2, 32, data);
+  vec_st(out3, 48, data);
+  vec_st(out4, 64, data);
+  vec_st(out5, 80, data);
+  vec_st(out6, 96, data);
+  vec_st(out7, 112, data);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jfdctint-altivec.c Added

@@ -0,0 +1,258 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* ACCURATE INTEGER FORWARD DCT */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_298  2446   /* FIX(0.298631336) */
+#define F_0_390  3196   /* FIX(0.390180644) */
+#define F_0_541  4433   /* FIX(0.541196100) */
+#define F_0_765  6270   /* FIX(0.765366865) */
+#define F_0_899  7373   /* FIX(0.899976223) */
+#define F_1_175  9633   /* FIX(1.175875602) */
+#define F_1_501  12299  /* FIX(1.501321110) */
+#define F_1_847  15137  /* FIX(1.847759065) */
+#define F_1_961  16069  /* FIX(1.961570560) */
+#define F_2_053  16819  /* FIX(2.053119869) */
+#define F_2_562  20995  /* FIX(2.562915447) */
+#define F_3_072  25172  /* FIX(3.072711026) */
+
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2  (CONST_BITS + PASS1_BITS)
+
+
+#define DO_FDCT_COMMON(PASS) { \
+  /* (Original) \
+   * z1 = (tmp12 + tmp13) * 0.541196100; \
+   * data2 = z1 + tmp13 * 0.765366865; \
+   * data6 = z1 + tmp12 * -1.847759065; \
+   * \
+   * (This implementation) \
+   * data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; \
+   * data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); \
+   */ \
+  \
+  tmp1312l = vec_mergeh(tmp13, tmp12); \
+  tmp1312h = vec_mergel(tmp13, tmp12); \
+  \
+  out2l = vec_msums(tmp1312l, pw_f130_f054, pd_descale_p##PASS); \
+  out2h = vec_msums(tmp1312h, pw_f130_f054, pd_descale_p##PASS); \
+  out6l = vec_msums(tmp1312l, pw_f054_mf130, pd_descale_p##PASS); \
+  out6h = vec_msums(tmp1312h, pw_f054_mf130, pd_descale_p##PASS); \
+  \
+  out2l = vec_sra(out2l, descale_p##PASS); \
+  out2h = vec_sra(out2h, descale_p##PASS); \
+  out6l = vec_sra(out6l, descale_p##PASS); \
+  out6h = vec_sra(out6h, descale_p##PASS); \
+  \
+  out2 = vec_pack(out2l, out2h); \
+  out6 = vec_pack(out6l, out6h); \
+  \
+  /* Odd part */ \
+  \
+  z3 = vec_add(tmp4, tmp6); \
+  z4 = vec_add(tmp5, tmp7); \
+  \
+  /* (Original) \
+   * z5 = (z3 + z4) * 1.175875602; \
+   * z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644; \
+   * z3 += z5;  z4 += z5; \
+   * \
+   * (This implementation) \
+   * z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; \
+   * z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); \
+   */ \
+  \
+  z34l = vec_mergeh(z3, z4); \
+  z34h = vec_mergel(z3, z4); \
+  \
+  z3l = vec_msums(z34l, pw_mf078_f117, pd_descale_p##PASS); \
+  z3h = vec_msums(z34h, pw_mf078_f117, pd_descale_p##PASS); \
+  z4l = vec_msums(z34l, pw_f117_f078, pd_descale_p##PASS); \
+  z4h = vec_msums(z34h, pw_f117_f078, pd_descale_p##PASS); \
+  \
+  /* (Original) \
+   * z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6; \
+   * tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869; \
+   * tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110; \
+   * z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447; \
+   * data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4; \
+   * data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4; \
+   * \
+   * (This implementation) \
+   * tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; \
+   * tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; \
+   * tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); \
+   * tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); \
+   * data7 = tmp4 + z3;  data5 = tmp5 + z4; \
+   * data3 = tmp6 + z3;  data1 = tmp7 + z4; \
+   */ \
+  \
+  tmp47l = vec_mergeh(tmp4, tmp7); \
+  tmp47h = vec_mergel(tmp4, tmp7); \
+  \
+  out7l = vec_msums(tmp47l, pw_mf060_mf089, z3l); \
+  out7h = vec_msums(tmp47h, pw_mf060_mf089, z3h); \
+  out1l = vec_msums(tmp47l, pw_mf089_f060, z4l); \
+  out1h = vec_msums(tmp47h, pw_mf089_f060, z4h); \
+  \
+  out7l = vec_sra(out7l, descale_p##PASS); \
+  out7h = vec_sra(out7h, descale_p##PASS); \
+  out1l = vec_sra(out1l, descale_p##PASS); \
+  out1h = vec_sra(out1h, descale_p##PASS); \
+  \
+  out7 = vec_pack(out7l, out7h); \
+  out1 = vec_pack(out1l, out1h); \
+  \
+  tmp56l = vec_mergeh(tmp5, tmp6); \
+  tmp56h = vec_mergel(tmp5, tmp6); \
+  \
+  out5l = vec_msums(tmp56l, pw_mf050_mf256, z4l); \
+  out5h = vec_msums(tmp56h, pw_mf050_mf256, z4h); \
+  out3l = vec_msums(tmp56l, pw_mf256_f050, z3l); \
+  out3h = vec_msums(tmp56h, pw_mf256_f050, z3h); \
+  \
+  out5l = vec_sra(out5l, descale_p##PASS); \
+  out5h = vec_sra(out5h, descale_p##PASS); \
+  out3l = vec_sra(out3l, descale_p##PASS); \
+  out3h = vec_sra(out3h, descale_p##PASS); \
+  \
+  out5 = vec_pack(out5l, out5h); \
+  out3 = vec_pack(out3l, out3h); \
+}
+
+#define DO_FDCT_PASS1() { \
+  /* Even part */ \
+  \
+  tmp10 = vec_add(tmp0, tmp3); \
+  tmp13 = vec_sub(tmp0, tmp3); \
+  tmp11 = vec_add(tmp1, tmp2); \
+  tmp12 = vec_sub(tmp1, tmp2); \
+  \
+  out0  = vec_add(tmp10, tmp11); \
+  out0  = vec_sl(out0, pass1_bits); \
+  out4  = vec_sub(tmp10, tmp11); \
+  out4  = vec_sl(out4, pass1_bits); \
+  \
+  DO_FDCT_COMMON(1); \
+}
+
+#define DO_FDCT_PASS2() { \
+  /* Even part */ \
+  \
+  tmp10 = vec_add(tmp0, tmp3); \
+  tmp13 = vec_sub(tmp0, tmp3); \
+  tmp11 = vec_add(tmp1, tmp2); \
+  tmp12 = vec_sub(tmp1, tmp2); \
+  \
+  out0  = vec_add(tmp10, tmp11); \
+  out0  = vec_add(out0, pw_descale_p2x); \
+  out0  = vec_sra(out0, pass1_bits); \
+  out4  = vec_sub(tmp10, tmp11); \
+  out4  = vec_add(out4, pw_descale_p2x); \
+  out4  = vec_sra(out4, pass1_bits); \
+  \
+  DO_FDCT_COMMON(2); \
+}
+
+
+void jsimd_fdct_islow_altivec(DCTELEM *data)
+{
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
+    tmp47l, tmp47h, tmp56l, tmp56h, tmp1312l, tmp1312h,
+    z3, z4, z34l, z34h,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+  __vector int z3l, z3h, z4l, z4h,
+    out1l, out1h, out2l, out2h, out3l, out3h, out5l, out5h, out6l, out6h,
+    out7l, out7h;
+
+  /* Constants */
+  __vector short
+    pw_f130_f054 = { __4X2(F_0_541 + F_0_765, F_0_541) },
+    pw_f054_mf130 = { __4X2(F_0_541, F_0_541 - F_1_847) },
+    pw_mf078_f117 = { __4X2(F_1_175 - F_1_961, F_1_175) },
+    pw_f117_f078 = { __4X2(F_1_175, F_1_175 - F_0_390) },
+    pw_mf060_mf089 = { __4X2(F_0_298 - F_0_899, -F_0_899) },
+    pw_mf089_f060 = { __4X2(-F_0_899, F_1_501 - F_0_899) },
+    pw_mf050_mf256 = { __4X2(F_2_053 - F_2_562, -F_2_562) },
+    pw_mf256_f050 = { __4X2(-F_2_562, F_3_072 - F_2_562) },
+    pw_descale_p2x = { __8X(1 << (PASS1_BITS - 1)) };
+  __vector unsigned short pass1_bits = { __8X(PASS1_BITS) };
+  __vector int pd_descale_p1 = { __4X(1 << (DESCALE_P1 - 1)) },
+    pd_descale_p2 = { __4X(1 << (DESCALE_P2 - 1)) };
+  __vector unsigned int descale_p1 = { __4X(DESCALE_P1) },
+    descale_p2 = { __4X(DESCALE_P2) };
+
+  /* Pass 1: process rows */
+
+  row0 = vec_ld(0, data);
+  row1 = vec_ld(16, data);
+  row2 = vec_ld(32, data);
+  row3 = vec_ld(48, data);
+  row4 = vec_ld(64, data);
+  row5 = vec_ld(80, data);
+  row6 = vec_ld(96, data);
+  row7 = vec_ld(112, data);
+
+  TRANSPOSE(row, col);
+
+  tmp0 = vec_add(col0, col7);
+  tmp7 = vec_sub(col0, col7);
+  tmp1 = vec_add(col1, col6);
+  tmp6 = vec_sub(col1, col6);
+  tmp2 = vec_add(col2, col5);
+  tmp5 = vec_sub(col2, col5);
+  tmp3 = vec_add(col3, col4);
+  tmp4 = vec_sub(col3, col4);
+
+  DO_FDCT_PASS1();
+
+  /* Pass 2: process columns */
+
+  TRANSPOSE(out, row);
+
+  tmp0 = vec_add(row0, row7);
+  tmp7 = vec_sub(row0, row7);
+  tmp1 = vec_add(row1, row6);
+  tmp6 = vec_sub(row1, row6);
+  tmp2 = vec_add(row2, row5);
+  tmp5 = vec_sub(row2, row5);
+  tmp3 = vec_add(row3, row4);
+  tmp4 = vec_sub(row3, row4);
+
+  DO_FDCT_PASS2();
+
+  vec_st(out0, 0, data);
+  vec_st(out1, 16, data);
+  vec_st(out2, 32, data);
+  vec_st(out3, 48, data);
+  vec_st(out4, 64, data);
+  vec_st(out5, 80, data);
+  vec_st(out6, 96, data);
+  vec_st(out7, 112, data);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jidctfst-altivec.c Added

@@ -0,0 +1,255 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* FAST INTEGER INVERSE DCT
+ *
+ * This is similar to the SSE2 implementation, except that we left-shift the
+ * constants by 1 less bit (the -1 in CONST_SHIFT.)  This is because
+ * vec_madds(arg1, arg2, arg3) generates the 16-bit saturated sum of:
+ *   the elements in arg3 + the most significant 17 bits of
+ *     (the elements in arg1 * the elements in arg2).
+ */
+
+#include "jsimd_altivec.h"
+
+
+#define F_1_082  277              /* FIX(1.082392200) */
+#define F_1_414  362              /* FIX(1.414213562) */
+#define F_1_847  473              /* FIX(1.847759065) */
+#define F_2_613  669              /* FIX(2.613125930) */
+#define F_1_613  (F_2_613 - 256)  /* FIX(2.613125930) - FIX(1) */
+
+#define CONST_BITS  8
+#define PASS1_BITS  2
+#define PRE_MULTIPLY_SCALE_BITS  2
+#define CONST_SHIFT  (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS - 1)
+
+
+#define DO_IDCT(in) { \
+  /* Even part */ \
+  \
+  tmp10 = vec_add(in##0, in##4); \
+  tmp11 = vec_sub(in##0, in##4); \
+  tmp13 = vec_add(in##2, in##6); \
+  \
+  tmp12 = vec_sub(in##2, in##6); \
+  tmp12 = vec_sl(tmp12, pre_multiply_scale_bits); \
+  tmp12 = vec_madds(tmp12, pw_F1414, pw_zero); \
+  tmp12 = vec_sub(tmp12, tmp13); \
+  \
+  tmp0 = vec_add(tmp10, tmp13); \
+  tmp3 = vec_sub(tmp10, tmp13); \
+  tmp1 = vec_add(tmp11, tmp12); \
+  tmp2 = vec_sub(tmp11, tmp12); \
+  \
+  /* Odd part */ \
+  \
+  z13 = vec_add(in##5, in##3); \
+  z10 = vec_sub(in##5, in##3); \
+  z10s = vec_sl(z10, pre_multiply_scale_bits); \
+  z11 = vec_add(in##1, in##7); \
+  z12s = vec_sub(in##1, in##7); \
+  z12s = vec_sl(z12s, pre_multiply_scale_bits); \
+  \
+  tmp11 = vec_sub(z11, z13); \
+  tmp11 = vec_sl(tmp11, pre_multiply_scale_bits); \
+  tmp11 = vec_madds(tmp11, pw_F1414, pw_zero); \
+  \
+  tmp7 = vec_add(z11, z13); \
+  \
+  /* To avoid overflow... \
+   * \
+   * (Original) \
+   * tmp12 = -2.613125930 * z10 + z5; \
+   * \
+   * (This implementation) \
+   * tmp12 = (-1.613125930 - 1) * z10 + z5; \
+   *       = -1.613125930 * z10 - z10 + z5; \
+   */ \
+  \
+  z5 = vec_add(z10s, z12s); \
+  z5 = vec_madds(z5, pw_F1847, pw_zero); \
+  \
+  tmp10 = vec_madds(z12s, pw_F1082, pw_zero); \
+  tmp10 = vec_sub(tmp10, z5); \
+  tmp12 = vec_madds(z10s, pw_MF1613, z5); \
+  tmp12 = vec_sub(tmp12, z10); \
+  \
+  tmp6 = vec_sub(tmp12, tmp7); \
+  tmp5 = vec_sub(tmp11, tmp6); \
+  tmp4 = vec_add(tmp10, tmp5); \
+  \
+  out0 = vec_add(tmp0, tmp7); \
+  out1 = vec_add(tmp1, tmp6); \
+  out2 = vec_add(tmp2, tmp5); \
+  out3 = vec_sub(tmp3, tmp4); \
+  out4 = vec_add(tmp3, tmp4); \
+  out5 = vec_sub(tmp2, tmp5); \
+  out6 = vec_sub(tmp1, tmp6); \
+  out7 = vec_sub(tmp0, tmp7); \
+}
+
+
+void jsimd_idct_ifast_altivec(void *dct_table_, JCOEFPTR coef_block,
+                              JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  short *dct_table = (short *)dct_table_;
+  int *outptr;
+
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    quant0, quant1, quant2, quant3, quant4, quant5, quant6, quant7,
+    tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
+    z5, z10, z10s, z11, z12s, z13,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+  __vector signed char outb;
+
+  /* Constants */
+  __vector short pw_zero = { __8X(0) },
+    pw_F1414 = { __8X(F_1_414 << CONST_SHIFT) },
+    pw_F1847 = { __8X(F_1_847 << CONST_SHIFT) },
+    pw_MF1613 = { __8X(-F_1_613 << CONST_SHIFT) },
+    pw_F1082 = { __8X(F_1_082 << CONST_SHIFT) };
+  __vector unsigned short
+    pre_multiply_scale_bits = { __8X(PRE_MULTIPLY_SCALE_BITS) },
+    pass1_bits3 = { __8X(PASS1_BITS + 3) };
+  __vector signed char pb_centerjsamp = { __16X(CENTERJSAMPLE) };
+
+  /* Pass 1: process columns */
+
+  col0 = vec_ld(0, coef_block);
+  col1 = vec_ld(16, coef_block);
+  col2 = vec_ld(32, coef_block);
+  col3 = vec_ld(48, coef_block);
+  col4 = vec_ld(64, coef_block);
+  col5 = vec_ld(80, coef_block);
+  col6 = vec_ld(96, coef_block);
+  col7 = vec_ld(112, coef_block);
+
+  tmp1 = vec_or(col1, col2);
+  tmp2 = vec_or(col3, col4);
+  tmp1 = vec_or(tmp1, tmp2);
+  tmp3 = vec_or(col5, col6);
+  tmp3 = vec_or(tmp3, col7);
+  tmp1 = vec_or(tmp1, tmp3);
+
+  quant0 = vec_ld(0, dct_table);
+  col0 = vec_mladd(col0, quant0, pw_zero);
+
+  if (vec_all_eq(tmp1, pw_zero)) {
+    /* AC terms all zero */
+
+    row0 = vec_splat(col0, 0);
+    row1 = vec_splat(col0, 1);
+    row2 = vec_splat(col0, 2);
+    row3 = vec_splat(col0, 3);
+    row4 = vec_splat(col0, 4);
+    row5 = vec_splat(col0, 5);
+    row6 = vec_splat(col0, 6);
+    row7 = vec_splat(col0, 7);
+
+  } else {
+
+    quant1 = vec_ld(16, dct_table);
+    quant2 = vec_ld(32, dct_table);
+    quant3 = vec_ld(48, dct_table);
+    quant4 = vec_ld(64, dct_table);
+    quant5 = vec_ld(80, dct_table);
+    quant6 = vec_ld(96, dct_table);
+    quant7 = vec_ld(112, dct_table);
+
+    col1 = vec_mladd(col1, quant1, pw_zero);
+    col2 = vec_mladd(col2, quant2, pw_zero);
+    col3 = vec_mladd(col3, quant3, pw_zero);
+    col4 = vec_mladd(col4, quant4, pw_zero);
+    col5 = vec_mladd(col5, quant5, pw_zero);
+    col6 = vec_mladd(col6, quant6, pw_zero);
+    col7 = vec_mladd(col7, quant7, pw_zero);
+
+    DO_IDCT(col);
+
+    TRANSPOSE(out, row);
+  }
+
+  /* Pass 2: process rows */
+
+  DO_IDCT(row);
+
+  out0 = vec_sra(out0, pass1_bits3);
+  out1 = vec_sra(out1, pass1_bits3);
+  out2 = vec_sra(out2, pass1_bits3);
+  out3 = vec_sra(out3, pass1_bits3);
+  out4 = vec_sra(out4, pass1_bits3);
+  out5 = vec_sra(out5, pass1_bits3);
+  out6 = vec_sra(out6, pass1_bits3);
+  out7 = vec_sra(out7, pass1_bits3);
+
+  TRANSPOSE(out, col);
+
+  outb = vec_packs(col0, col0);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf0 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col1, col1);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf1 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col2, col2);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf2 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col3, col3);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf3 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col4, col4);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf4 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col5, col5);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf5 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col6, col6);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf6 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col7, col7);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf7 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jidctint-altivec.c Added

@@ -0,0 +1,357 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, 2020, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* ACCURATE INTEGER INVERSE DCT */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_298  2446   /* FIX(0.298631336) */
+#define F_0_390  3196   /* FIX(0.390180644) */
+#define F_0_541  4433   /* FIX(0.541196100) */
+#define F_0_765  6270   /* FIX(0.765366865) */
+#define F_0_899  7373   /* FIX(0.899976223) */
+#define F_1_175  9633   /* FIX(1.175875602) */
+#define F_1_501  12299  /* FIX(1.501321110) */
+#define F_1_847  15137  /* FIX(1.847759065) */
+#define F_1_961  16069  /* FIX(1.961570560) */
+#define F_2_053  16819  /* FIX(2.053119869) */
+#define F_2_562  20995  /* FIX(2.562915447) */
+#define F_3_072  25172  /* FIX(3.072711026) */
+
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2  (CONST_BITS + PASS1_BITS + 3)
+
+
+#define DO_IDCT(in, PASS) { \
+  /* Even part \
+   * \
+   * (Original) \
+   * z1 = (z2 + z3) * 0.541196100; \
+   * tmp2 = z1 + z3 * -1.847759065; \
+   * tmp3 = z1 + z2 * 0.765366865; \
+   * \
+   * (This implementation) \
+   * tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065); \
+   * tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100; \
+   */ \
+  \
+  in##26l = vec_mergeh(in##2, in##6); \
+  in##26h = vec_mergel(in##2, in##6); \
+  \
+  tmp3l = vec_msums(in##26l, pw_f130_f054, pd_zero); \
+  tmp3h = vec_msums(in##26h, pw_f130_f054, pd_zero); \
+  tmp2l = vec_msums(in##26l, pw_f054_mf130, pd_zero); \
+  tmp2h = vec_msums(in##26h, pw_f054_mf130, pd_zero); \
+  \
+  tmp0 = vec_add(in##0, in##4); \
+  tmp1 = vec_sub(in##0, in##4); \
+  \
+  tmp0l = vec_unpackh(tmp0); \
+  tmp0h = vec_unpackl(tmp0); \
+  tmp0l = vec_sl(tmp0l, const_bits); \
+  tmp0h = vec_sl(tmp0h, const_bits); \
+  tmp0l = vec_add(tmp0l, pd_descale_p##PASS); \
+  tmp0h = vec_add(tmp0h, pd_descale_p##PASS); \
+  \
+  tmp10l = vec_add(tmp0l, tmp3l); \
+  tmp10h = vec_add(tmp0h, tmp3h); \
+  tmp13l = vec_sub(tmp0l, tmp3l); \
+  tmp13h = vec_sub(tmp0h, tmp3h); \
+  \
+  tmp1l = vec_unpackh(tmp1); \
+  tmp1h = vec_unpackl(tmp1); \
+  tmp1l = vec_sl(tmp1l, const_bits); \
+  tmp1h = vec_sl(tmp1h, const_bits); \
+  tmp1l = vec_add(tmp1l, pd_descale_p##PASS); \
+  tmp1h = vec_add(tmp1h, pd_descale_p##PASS); \
+  \
+  tmp11l = vec_add(tmp1l, tmp2l); \
+  tmp11h = vec_add(tmp1h, tmp2h); \
+  tmp12l = vec_sub(tmp1l, tmp2l); \
+  tmp12h = vec_sub(tmp1h, tmp2h); \
+  \
+  /* Odd part */ \
+  \
+  z3 = vec_add(in##3, in##7); \
+  z4 = vec_add(in##1, in##5); \
+  \
+  /* (Original) \
+   * z5 = (z3 + z4) * 1.175875602; \
+   * z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644; \
+   * z3 += z5;  z4 += z5; \
+   * \
+   * (This implementation) \
+   * z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; \
+   * z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); \
+   */ \
+  \
+  z34l = vec_mergeh(z3, z4); \
+  z34h = vec_mergel(z3, z4); \
+  \
+  z3l = vec_msums(z34l, pw_mf078_f117, pd_zero); \
+  z3h = vec_msums(z34h, pw_mf078_f117, pd_zero); \
+  z4l = vec_msums(z34l, pw_f117_f078, pd_zero); \
+  z4h = vec_msums(z34h, pw_f117_f078, pd_zero); \
+  \
+  /* (Original) \
+   * z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2; \
+   * tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869; \
+   * tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110; \
+   * z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447; \
+   * tmp0 += z1 + z3;  tmp1 += z2 + z4; \
+   * tmp2 += z2 + z3;  tmp3 += z1 + z4; \
+   * \
+   * (This implementation) \
+   * tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223; \
+   * tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447; \
+   * tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447); \
+   * tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223); \
+   * tmp0 += z3;  tmp1 += z4; \
+   * tmp2 += z3;  tmp3 += z4; \
+   */ \
+  \
+  in##71l = vec_mergeh(in##7, in##1); \
+  in##71h = vec_mergel(in##7, in##1); \
+  \
+  tmp0l = vec_msums(in##71l, pw_mf060_mf089, z3l); \
+  tmp0h = vec_msums(in##71h, pw_mf060_mf089, z3h); \
+  tmp3l = vec_msums(in##71l, pw_mf089_f060, z4l); \
+  tmp3h = vec_msums(in##71h, pw_mf089_f060, z4h); \
+  \
+  in##53l = vec_mergeh(in##5, in##3); \
+  in##53h = vec_mergel(in##5, in##3); \
+  \
+  tmp1l = vec_msums(in##53l, pw_mf050_mf256, z4l); \
+  tmp1h = vec_msums(in##53h, pw_mf050_mf256, z4h); \
+  tmp2l = vec_msums(in##53l, pw_mf256_f050, z3l); \
+  tmp2h = vec_msums(in##53h, pw_mf256_f050, z3h); \
+  \
+  /* Final output stage */ \
+  \
+  out0l = vec_add(tmp10l, tmp3l); \
+  out0h = vec_add(tmp10h, tmp3h); \
+  out7l = vec_sub(tmp10l, tmp3l); \
+  out7h = vec_sub(tmp10h, tmp3h); \
+  \
+  out0l = vec_sra(out0l, descale_p##PASS); \
+  out0h = vec_sra(out0h, descale_p##PASS); \
+  out7l = vec_sra(out7l, descale_p##PASS); \
+  out7h = vec_sra(out7h, descale_p##PASS); \
+  \
+  out0 = vec_pack(out0l, out0h); \
+  out7 = vec_pack(out7l, out7h); \
+  \
+  out1l = vec_add(tmp11l, tmp2l); \
+  out1h = vec_add(tmp11h, tmp2h); \
+  out6l = vec_sub(tmp11l, tmp2l); \
+  out6h = vec_sub(tmp11h, tmp2h); \
+  \
+  out1l = vec_sra(out1l, descale_p##PASS); \
+  out1h = vec_sra(out1h, descale_p##PASS); \
+  out6l = vec_sra(out6l, descale_p##PASS); \
+  out6h = vec_sra(out6h, descale_p##PASS); \
+  \
+  out1 = vec_pack(out1l, out1h); \
+  out6 = vec_pack(out6l, out6h); \
+  \
+  out2l = vec_add(tmp12l, tmp1l); \
+  out2h = vec_add(tmp12h, tmp1h); \
+  out5l = vec_sub(tmp12l, tmp1l); \
+  out5h = vec_sub(tmp12h, tmp1h); \
+  \
+  out2l = vec_sra(out2l, descale_p##PASS); \
+  out2h = vec_sra(out2h, descale_p##PASS); \
+  out5l = vec_sra(out5l, descale_p##PASS); \
+  out5h = vec_sra(out5h, descale_p##PASS); \
+  \
+  out2 = vec_pack(out2l, out2h); \
+  out5 = vec_pack(out5l, out5h); \
+  \
+  out3l = vec_add(tmp13l, tmp0l); \
+  out3h = vec_add(tmp13h, tmp0h); \
+  out4l = vec_sub(tmp13l, tmp0l); \
+  out4h = vec_sub(tmp13h, tmp0h); \
+  \
+  out3l = vec_sra(out3l, descale_p##PASS); \
+  out3h = vec_sra(out3h, descale_p##PASS); \
+  out4l = vec_sra(out4l, descale_p##PASS); \
+  out4h = vec_sra(out4h, descale_p##PASS); \
+  \
+  out3 = vec_pack(out3l, out3h); \
+  out4 = vec_pack(out4l, out4h); \
+}
+
+
+void jsimd_idct_islow_altivec(void *dct_table_, JCOEFPTR coef_block,
+                              JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  short *dct_table = (short *)dct_table_;
+  int *outptr;
+
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    quant0, quant1, quant2, quant3, quant4, quant5, quant6, quant7,
+    tmp0, tmp1, tmp2, tmp3, z3, z4,
+    z34l, z34h, col71l, col71h, col26l, col26h, col53l, col53h,
+    row71l, row71h, row26l, row26h, row53l, row53h,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+  __vector int tmp0l, tmp0h, tmp1l, tmp1h, tmp2l, tmp2h, tmp3l, tmp3h,
+    tmp10l, tmp10h, tmp11l, tmp11h, tmp12l, tmp12h, tmp13l, tmp13h,
+    z3l, z3h, z4l, z4h,
+    out0l, out0h, out1l, out1h, out2l, out2h, out3l, out3h, out4l, out4h,
+    out5l, out5h, out6l, out6h, out7l, out7h;
+  __vector signed char outb;
+
+  /* Constants */
+  __vector short pw_zero = { __8X(0) },
+    pw_f130_f054 = { __4X2(F_0_541 + F_0_765, F_0_541) },
+    pw_f054_mf130 = { __4X2(F_0_541, F_0_541 - F_1_847) },
+    pw_mf078_f117 = { __4X2(F_1_175 - F_1_961, F_1_175) },
+    pw_f117_f078 = { __4X2(F_1_175, F_1_175 - F_0_390) },
+    pw_mf060_mf089 = { __4X2(F_0_298 - F_0_899, -F_0_899) },
+    pw_mf089_f060 = { __4X2(-F_0_899, F_1_501 - F_0_899) },
+    pw_mf050_mf256 = { __4X2(F_2_053 - F_2_562, -F_2_562) },
+    pw_mf256_f050 = { __4X2(-F_2_562, F_3_072 - F_2_562) };
+  __vector unsigned short pass1_bits = { __8X(PASS1_BITS) };
+  __vector int pd_zero = { __4X(0) },
+    pd_descale_p1 = { __4X(1 << (DESCALE_P1 - 1)) },
+    pd_descale_p2 = { __4X(1 << (DESCALE_P2 - 1)) };
+  __vector unsigned int descale_p1 = { __4X(DESCALE_P1) },
+    descale_p2 = { __4X(DESCALE_P2) },
+    const_bits = { __4X(CONST_BITS) };
+  __vector signed char pb_centerjsamp = { __16X(CENTERJSAMPLE) };
+
+  /* Pass 1: process columns */
+
+  col0 = vec_ld(0, coef_block);
+  col1 = vec_ld(16, coef_block);
+  col2 = vec_ld(32, coef_block);
+  col3 = vec_ld(48, coef_block);
+  col4 = vec_ld(64, coef_block);
+  col5 = vec_ld(80, coef_block);
+  col6 = vec_ld(96, coef_block);
+  col7 = vec_ld(112, coef_block);
+
+  tmp1 = vec_or(col1, col2);
+  tmp2 = vec_or(col3, col4);
+  tmp1 = vec_or(tmp1, tmp2);
+  tmp3 = vec_or(col5, col6);
+  tmp3 = vec_or(tmp3, col7);
+  tmp1 = vec_or(tmp1, tmp3);
+
+  quant0 = vec_ld(0, dct_table);
+  col0 = vec_mladd(col0, quant0, pw_zero);
+
+  if (vec_all_eq(tmp1, pw_zero)) {
+    /* AC terms all zero */
+
+    col0 = vec_sl(col0, pass1_bits);
+
+    row0 = vec_splat(col0, 0);
+    row1 = vec_splat(col0, 1);
+    row2 = vec_splat(col0, 2);
+    row3 = vec_splat(col0, 3);
+    row4 = vec_splat(col0, 4);
+    row5 = vec_splat(col0, 5);
+    row6 = vec_splat(col0, 6);
+    row7 = vec_splat(col0, 7);
+
+  } else {
+
+    quant1 = vec_ld(16, dct_table);
+    quant2 = vec_ld(32, dct_table);
+    quant3 = vec_ld(48, dct_table);
+    quant4 = vec_ld(64, dct_table);
+    quant5 = vec_ld(80, dct_table);
+    quant6 = vec_ld(96, dct_table);
+    quant7 = vec_ld(112, dct_table);
+
+    col1 = vec_mladd(col1, quant1, pw_zero);
+    col2 = vec_mladd(col2, quant2, pw_zero);
+    col3 = vec_mladd(col3, quant3, pw_zero);
+    col4 = vec_mladd(col4, quant4, pw_zero);
+    col5 = vec_mladd(col5, quant5, pw_zero);
+    col6 = vec_mladd(col6, quant6, pw_zero);
+    col7 = vec_mladd(col7, quant7, pw_zero);
+
+    DO_IDCT(col, 1);
+
+    TRANSPOSE(out, row);
+  }
+
+  /* Pass 2: process rows */
+
+  DO_IDCT(row, 2);
+
+  TRANSPOSE(out, col);
+
+  outb = vec_packs(col0, col0);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf0 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col1, col1);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf1 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col2, col2);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf2 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col3, col3);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf3 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col4, col4);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf4 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col5, col5);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf5 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col6, col6);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf6 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col7, col7);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf7 + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jquanti-altivec.c Added

@@ -0,0 +1,250 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* INTEGER QUANTIZATION AND SAMPLE CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+/* NOTE: The address will either be aligned or offset by 8 bytes, so we can
+ * always get the data we want by using a single vector load (although we may
+ * have to permute the result.)
+ */
+#if __BIG_ENDIAN__
+
+#define LOAD_ROW(row) { \
+  elemptr = sample_datarow + start_col; \
+  in##row = vec_ld(0, elemptr); \
+  if ((size_t)elemptr & 15) \
+    in##row = vec_perm(in##row, in##row, vec_lvsl(0, elemptr)); \
+}
+
+#else
+
+#define LOAD_ROW(row) { \
+  elemptr = sample_datarow + start_col; \
+  in##row = vec_vsx_ld(0, elemptr); \
+}
+
+#endif
+
+
+void jsimd_convsamp_altivec(JSAMPARRAY sample_data, JDIMENSION start_col,
+                            DCTELEM *workspace)
+{
+  JSAMPROW elemptr;
+
+  __vector unsigned char in0, in1, in2, in3, in4, in5, in6, in7;
+  __vector short out0, out1, out2, out3, out4, out5, out6, out7;
+
+  /* Constants */
+  __vector short pw_centerjsamp = { __8X(CENTERJSAMPLE) };
+  __vector unsigned char pb_zero = { __16X(0) };
+
+  LOAD_ROW(0);
+  LOAD_ROW(1);
+  LOAD_ROW(2);
+  LOAD_ROW(3);
+  LOAD_ROW(4);
+  LOAD_ROW(5);
+  LOAD_ROW(6);
+  LOAD_ROW(7);
+
+  out0 = (__vector short)VEC_UNPACKHU(in0);
+  out1 = (__vector short)VEC_UNPACKHU(in1);
+  out2 = (__vector short)VEC_UNPACKHU(in2);
+  out3 = (__vector short)VEC_UNPACKHU(in3);
+  out4 = (__vector short)VEC_UNPACKHU(in4);
+  out5 = (__vector short)VEC_UNPACKHU(in5);
+  out6 = (__vector short)VEC_UNPACKHU(in6);
+  out7 = (__vector short)VEC_UNPACKHU(in7);
+
+  out0 = vec_sub(out0, pw_centerjsamp);
+  out1 = vec_sub(out1, pw_centerjsamp);
+  out2 = vec_sub(out2, pw_centerjsamp);
+  out3 = vec_sub(out3, pw_centerjsamp);
+  out4 = vec_sub(out4, pw_centerjsamp);
+  out5 = vec_sub(out5, pw_centerjsamp);
+  out6 = vec_sub(out6, pw_centerjsamp);
+  out7 = vec_sub(out7, pw_centerjsamp);
+
+  vec_st(out0, 0, workspace);
+  vec_st(out1, 16, workspace);
+  vec_st(out2, 32, workspace);
+  vec_st(out3, 48, workspace);
+  vec_st(out4, 64, workspace);
+  vec_st(out5, 80, workspace);
+  vec_st(out6, 96, workspace);
+  vec_st(out7, 112, workspace);
+}
+
+
+#define WORD_BIT  16
+
+/* There is no AltiVec 16-bit unsigned multiply instruction, hence this.
+   We basically need an unsigned equivalent of vec_madds(). */
+
+#define MULTIPLY(vs0, vs1, out) { \
+  tmpe = vec_mule((__vector unsigned short)vs0, \
+                  (__vector unsigned short)vs1); \
+  tmpo = vec_mulo((__vector unsigned short)vs0, \
+                  (__vector unsigned short)vs1); \
+  out = (__vector short)vec_perm((__vector unsigned short)tmpe, \
+                                 (__vector unsigned short)tmpo, \
+                                 shift_pack_index); \
+}
+
+void jsimd_quantize_altivec(JCOEFPTR coef_block, DCTELEM *divisors,
+                            DCTELEM *workspace)
+{
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    row0s, row1s, row2s, row3s, row4s, row5s, row6s, row7s,
+    corr0, corr1, corr2, corr3, corr4, corr5, corr6, corr7,
+    recip0, recip1, recip2, recip3, recip4, recip5, recip6, recip7,
+    scale0, scale1, scale2, scale3, scale4, scale5, scale6, scale7;
+  __vector unsigned int tmpe, tmpo;
+
+  /* Constants */
+  __vector unsigned short pw_word_bit_m1 = { __8X(WORD_BIT - 1) };
+#if __BIG_ENDIAN__
+  __vector unsigned char shift_pack_index =
+    {  0,  1, 16, 17,  4,  5, 20, 21,  8,  9, 24, 25, 12, 13, 28, 29 };
+#else
+  __vector unsigned char shift_pack_index =
+    {  2,  3, 18, 19,  6,  7, 22, 23, 10, 11, 26, 27, 14, 15, 30, 31 };
+#endif
+
+  row0 = vec_ld(0, workspace);
+  row1 = vec_ld(16, workspace);
+  row2 = vec_ld(32, workspace);
+  row3 = vec_ld(48, workspace);
+  row4 = vec_ld(64, workspace);
+  row5 = vec_ld(80, workspace);
+  row6 = vec_ld(96, workspace);
+  row7 = vec_ld(112, workspace);
+
+  /* Branch-less absolute value */
+  row0s = vec_sra(row0, pw_word_bit_m1);
+  row1s = vec_sra(row1, pw_word_bit_m1);
+  row2s = vec_sra(row2, pw_word_bit_m1);
+  row3s = vec_sra(row3, pw_word_bit_m1);
+  row4s = vec_sra(row4, pw_word_bit_m1);
+  row5s = vec_sra(row5, pw_word_bit_m1);
+  row6s = vec_sra(row6, pw_word_bit_m1);
+  row7s = vec_sra(row7, pw_word_bit_m1);
+  row0 = vec_xor(row0, row0s);
+  row1 = vec_xor(row1, row1s);
+  row2 = vec_xor(row2, row2s);
+  row3 = vec_xor(row3, row3s);
+  row4 = vec_xor(row4, row4s);
+  row5 = vec_xor(row5, row5s);
+  row6 = vec_xor(row6, row6s);
+  row7 = vec_xor(row7, row7s);
+  row0 = vec_sub(row0, row0s);
+  row1 = vec_sub(row1, row1s);
+  row2 = vec_sub(row2, row2s);
+  row3 = vec_sub(row3, row3s);
+  row4 = vec_sub(row4, row4s);
+  row5 = vec_sub(row5, row5s);
+  row6 = vec_sub(row6, row6s);
+  row7 = vec_sub(row7, row7s);
+
+  corr0 = vec_ld(DCTSIZE2 * 2, divisors);
+  corr1 = vec_ld(DCTSIZE2 * 2 + 16, divisors);
+  corr2 = vec_ld(DCTSIZE2 * 2 + 32, divisors);
+  corr3 = vec_ld(DCTSIZE2 * 2 + 48, divisors);
+  corr4 = vec_ld(DCTSIZE2 * 2 + 64, divisors);
+  corr5 = vec_ld(DCTSIZE2 * 2 + 80, divisors);
+  corr6 = vec_ld(DCTSIZE2 * 2 + 96, divisors);
+  corr7 = vec_ld(DCTSIZE2 * 2 + 112, divisors);
+
+  row0 = vec_add(row0, corr0);
+  row1 = vec_add(row1, corr1);
+  row2 = vec_add(row2, corr2);
+  row3 = vec_add(row3, corr3);
+  row4 = vec_add(row4, corr4);
+  row5 = vec_add(row5, corr5);
+  row6 = vec_add(row6, corr6);
+  row7 = vec_add(row7, corr7);
+
+  recip0 = vec_ld(0, divisors);
+  recip1 = vec_ld(16, divisors);
+  recip2 = vec_ld(32, divisors);
+  recip3 = vec_ld(48, divisors);
+  recip4 = vec_ld(64, divisors);
+  recip5 = vec_ld(80, divisors);
+  recip6 = vec_ld(96, divisors);
+  recip7 = vec_ld(112, divisors);
+
+  MULTIPLY(row0, recip0, row0);
+  MULTIPLY(row1, recip1, row1);
+  MULTIPLY(row2, recip2, row2);
+  MULTIPLY(row3, recip3, row3);
+  MULTIPLY(row4, recip4, row4);
+  MULTIPLY(row5, recip5, row5);
+  MULTIPLY(row6, recip6, row6);
+  MULTIPLY(row7, recip7, row7);
+
+  scale0 = vec_ld(DCTSIZE2 * 4, divisors);
+  scale1 = vec_ld(DCTSIZE2 * 4 + 16, divisors);
+  scale2 = vec_ld(DCTSIZE2 * 4 + 32, divisors);
+  scale3 = vec_ld(DCTSIZE2 * 4 + 48, divisors);
+  scale4 = vec_ld(DCTSIZE2 * 4 + 64, divisors);
+  scale5 = vec_ld(DCTSIZE2 * 4 + 80, divisors);
+  scale6 = vec_ld(DCTSIZE2 * 4 + 96, divisors);
+  scale7 = vec_ld(DCTSIZE2 * 4 + 112, divisors);
+
+  MULTIPLY(row0, scale0, row0);
+  MULTIPLY(row1, scale1, row1);
+  MULTIPLY(row2, scale2, row2);
+  MULTIPLY(row3, scale3, row3);
+  MULTIPLY(row4, scale4, row4);
+  MULTIPLY(row5, scale5, row5);
+  MULTIPLY(row6, scale6, row6);
+  MULTIPLY(row7, scale7, row7);
+
+  row0 = vec_xor(row0, row0s);
+  row1 = vec_xor(row1, row1s);
+  row2 = vec_xor(row2, row2s);
+  row3 = vec_xor(row3, row3s);
+  row4 = vec_xor(row4, row4s);
+  row5 = vec_xor(row5, row5s);
+  row6 = vec_xor(row6, row6s);
+  row7 = vec_xor(row7, row7s);
+  row0 = vec_sub(row0, row0s);
+  row1 = vec_sub(row1, row1s);
+  row2 = vec_sub(row2, row2s);
+  row3 = vec_sub(row3, row3s);
+  row4 = vec_sub(row4, row4s);
+  row5 = vec_sub(row5, row5s);
+  row6 = vec_sub(row6, row6s);
+  row7 = vec_sub(row7, row7s);
+
+  vec_st(row0, 0, coef_block);
+  vec_st(row1, 16, coef_block);
+  vec_st(row2, 32, coef_block);
+  vec_st(row3, 48, coef_block);
+  vec_st(row4, 64, coef_block);
+  vec_st(row5, 80, coef_block);
+  vec_st(row6, 96, coef_block);
+  vec_st(row7, 112, coef_block);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jsimd.c Added

@@ -0,0 +1,872 @@
+/*
+ * jsimd_powerpc.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2009-2011, 2014-2016, 2018, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * PowerPC architecture.
+ */
+
+#ifdef __amigaos4__
+/* This must be defined first as it re-defines GLOBAL otherwise */
+#include <proto/exec.h>
+#endif
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+
+#if defined(__OpenBSD__)
+#include <sys/param.h>
+#include <sys/sysctl.h>
+#include <machine/cpu.h>
+#endif
+
+static unsigned int simd_support = ~0;
+
+#if !defined(__ALTIVEC__) && (defined(__linux__) || defined(ANDROID) || defined(__ANDROID__))
+
+#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT  (1024 * 1024)
+
+LOCAL(int)
+check_feature(char *buffer, char *feature)
+{
+  char *p;
+
+  if (*feature == 0)
+    return 0;
+  if (strncmp(buffer, "cpu", 3) != 0)
+    return 0;
+  buffer += 3;
+  while (isspace(*buffer))
+    buffer++;
+
+  /* Check if 'feature' is present in the buffer as a separate word */
+  while ((p = strstr(buffer, feature))) {
+    if (p > buffer && !isspace(*(p - 1))) {
+      buffer++;
+      continue;
+    }
+    p += strlen(feature);
+    if (*p != 0 && !isspace(*p)) {
+      buffer++;
+      continue;
+    }
+    return 1;
+  }
+  return 0;
+}
+
+LOCAL(int)
+parse_proc_cpuinfo(int bufsize)
+{
+  char *buffer = (char *)malloc(bufsize);
+  FILE *fd;
+
+  simd_support = 0;
+
+  if (!buffer)
+    return 0;
+
+  fd = fopen("/proc/cpuinfo", "r");
+  if (fd) {
+    while (fgets(buffer, bufsize, fd)) {
+      if (!strchr(buffer, '\n') && !feof(fd)) {
+        /* "impossible" happened - insufficient size of the buffer! */
+        fclose(fd);
+        free(buffer);
+        return 0;
+      }
+      if (check_feature(buffer, "altivec"))
+        simd_support |= JSIMD_ALTIVEC;
+    }
+    fclose(fd);
+  }
+  free(buffer);
+  return 1;
+}
+
+#endif
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd(void)
+{
+#ifndef NO_GETENV
+  char *env = NULL;
+#endif
+#if !defined(__ALTIVEC__) && (defined(__linux__) || defined(ANDROID) || defined(__ANDROID__))
+  int bufsize = 1024; /* an initial guess for the line buffer size limit */
+#elif defined(__amigaos4__)
+  uint32 altivec = 0;
+#elif defined(__OpenBSD__)
+  int mib2 = { CTL_MACHDEP, CPU_ALTIVEC };
+  int altivec;
+  size_t len = sizeof(altivec);
+#endif
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = 0;
+
+#if defined(__ALTIVEC__) || defined(__APPLE__)
+  simd_support |= JSIMD_ALTIVEC;
+#elif defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+  while (!parse_proc_cpuinfo(bufsize)) {
+    bufsize *= 2;
+    if (bufsize > SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT)
+      break;
+  }
+#elif defined(__amigaos4__)
+  IExec->GetCPUInfoTags(GCIT_VectorUnit, &altivec, TAG_DONE);
+  if (altivec == VECTORTYPE_ALTIVEC)
+    simd_support |= JSIMD_ALTIVEC;
+#elif defined(__OpenBSD__)
+  if (sysctl(mib, 2, &altivec, &len, NULL, 0) == 0 && altivec != 0)
+    simd_support |= JSIMD_ALTIVEC;
+#endif
+
+#ifndef NO_GETENV
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCEALTIVEC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = JSIMD_ALTIVEC;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+#endif
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                      JSAMPIMAGE output_buf, JDIMENSION output_row,
+                      int num_rows)
+{
+  void (*altivecfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    altivecfct = jsimd_extrgb_ycc_convert_altivec;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    altivecfct = jsimd_extrgbx_ycc_convert_altivec;
+    break;
+  case JCS_EXT_BGR:
+    altivecfct = jsimd_extbgr_ycc_convert_altivec;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    altivecfct = jsimd_extbgrx_ycc_convert_altivec;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    altivecfct = jsimd_extxbgr_ycc_convert_altivec;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    altivecfct = jsimd_extxrgb_ycc_convert_altivec;
+    break;
+  default:
+    altivecfct = jsimd_rgb_ycc_convert_altivec;
+    break;
+  }
+
+  altivecfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                       JSAMPIMAGE output_buf, JDIMENSION output_row,
+                       int num_rows)
+{
+  void (*altivecfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    altivecfct = jsimd_extrgb_gray_convert_altivec;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    altivecfct = jsimd_extrgbx_gray_convert_altivec;
+    break;
+  case JCS_EXT_BGR:
+    altivecfct = jsimd_extbgr_gray_convert_altivec;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    altivecfct = jsimd_extbgrx_gray_convert_altivec;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    altivecfct = jsimd_extxbgr_gray_convert_altivec;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    altivecfct = jsimd_extxrgb_gray_convert_altivec;
+    break;
+  default:
+    altivecfct = jsimd_rgb_gray_convert_altivec;
+    break;
+  }
+
+  altivecfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION input_row, JSAMPARRAY output_buf,
+                      int num_rows)
+{
+  void (*altivecfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    altivecfct = jsimd_ycc_extrgb_convert_altivec;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    altivecfct = jsimd_ycc_extrgbx_convert_altivec;
+    break;
+  case JCS_EXT_BGR:
+    altivecfct = jsimd_ycc_extbgr_convert_altivec;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    altivecfct = jsimd_ycc_extbgrx_convert_altivec;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    altivecfct = jsimd_ycc_extxbgr_convert_altivec;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    altivecfct = jsimd_ycc_extxrgb_convert_altivec;
+    break;
+  default:
+    altivecfct = jsimd_ycc_rgb_convert_altivec;
+    break;
+  }
+
+  altivecfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_altivec(cinfo->image_width, cinfo->max_v_samp_factor,
+                                compptr->v_samp_factor,
+                                compptr->width_in_blocks, input_data,
+                                output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_altivec(cinfo->image_width, cinfo->max_v_samp_factor,
+                                compptr->v_samp_factor,
+                                compptr->width_in_blocks, input_data,
+                                output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_upsample_altivec(cinfo->max_v_samp_factor, cinfo->output_width,
+                              input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_upsample_altivec(cinfo->max_v_samp_factor, cinfo->output_width,
+                              input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_fancy_upsample_altivec(cinfo->max_v_samp_factor,
+                                    compptr->downsampled_width, input_data,
+                                    output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_fancy_upsample_altivec(cinfo->max_v_samp_factor,
+                                    compptr->downsampled_width, input_data,
+                                    output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*altivecfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    altivecfct = jsimd_h2v2_extrgb_merged_upsample_altivec;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    altivecfct = jsimd_h2v2_extrgbx_merged_upsample_altivec;
+    break;
+  case JCS_EXT_BGR:
+    altivecfct = jsimd_h2v2_extbgr_merged_upsample_altivec;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    altivecfct = jsimd_h2v2_extbgrx_merged_upsample_altivec;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    altivecfct = jsimd_h2v2_extxbgr_merged_upsample_altivec;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    altivecfct = jsimd_h2v2_extxrgb_merged_upsample_altivec;
+    break;
+  default:
+    altivecfct = jsimd_h2v2_merged_upsample_altivec;
+    break;
+  }
+
+  altivecfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*altivecfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    altivecfct = jsimd_h2v1_extrgb_merged_upsample_altivec;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    altivecfct = jsimd_h2v1_extrgbx_merged_upsample_altivec;
+    break;
+  case JCS_EXT_BGR:
+    altivecfct = jsimd_h2v1_extbgr_merged_upsample_altivec;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    altivecfct = jsimd_h2v1_extbgrx_merged_upsample_altivec;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    altivecfct = jsimd_h2v1_extxbgr_merged_upsample_altivec;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    altivecfct = jsimd_h2v1_extxrgb_merged_upsample_altivec;
+    break;
+  default:
+    altivecfct = jsimd_h2v1_merged_upsample_altivec;
+    break;
+  }
+
+  altivecfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
+               DCTELEM *workspace)
+{
+  jsimd_convsamp_altivec(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow(DCTELEM *data)
+{
+  jsimd_fdct_islow_altivec(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast(DCTELEM *data)
+{
+  jsimd_fdct_ifast_altivec(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float(FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
+{
+  jsimd_quantize_altivec(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                     FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_islow_altivec(compptr->dct_table, coef_block, output_buf,
+                           output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_ifast_altivec(compptr->dct_table, coef_block, output_buf,
+                           output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block(void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET *)
+jsimd_huff_encode_one_block(void *state, JOCTET *buffer, JCOEFPTR block,
+                            int last_dc_val, c_derived_tbl *dctbl,
+                            c_derived_tbl *actbl)
+{
+  return NULL;
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_first_prepare(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_encode_mcu_AC_first_prepare(const JCOEF *block,
+                                  const int *jpeg_natural_order_start, int Sl,
+                                  int Al, JCOEF *values, size_t *zerobits)
+{
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_refine_prepare(void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block,
+                                   const int *jpeg_natural_order_start, int Sl,
+                                   int Al, JCOEF *absvalues, size_t *bits)
+{
+  return 0;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/powerpc/jsimd_altivec.h Added

@@ -0,0 +1,98 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include <altivec.h>
+
+
+/* Common code */
+
+#define __4X(a)      a, a, a, a
+#define __4X2(a, b)  a, b, a, b, a, b, a, b
+#define __8X(a)      __4X(a), __4X(a)
+#define __16X(a)     __8X(a), __8X(a)
+
+#define TRANSPOSE(row, col) { \
+  __vector short row04l, row04h, row15l, row15h, \
+                 row26l, row26h, row37l, row37h; \
+  __vector short col01e, col01o, col23e, col23o, \
+                 col45e, col45o, col67e, col67o; \
+  \
+                                       /* transpose coefficients (phase 1) */ \
+  row04l = vec_mergeh(row##0, row##4); /* row04l=(00 40 01 41 02 42 03 43) */ \
+  row04h = vec_mergel(row##0, row##4); /* row04h=(04 44 05 45 06 46 07 47) */ \
+  row15l = vec_mergeh(row##1, row##5); /* row15l=(10 50 11 51 12 52 13 53) */ \
+  row15h = vec_mergel(row##1, row##5); /* row15h=(14 54 15 55 16 56 17 57) */ \
+  row26l = vec_mergeh(row##2, row##6); /* row26l=(20 60 21 61 22 62 23 63) */ \
+  row26h = vec_mergel(row##2, row##6); /* row26h=(24 64 25 65 26 66 27 67) */ \
+  row37l = vec_mergeh(row##3, row##7); /* row37l=(30 70 31 71 32 72 33 73) */ \
+  row37h = vec_mergel(row##3, row##7); /* row37h=(34 74 35 75 36 76 37 77) */ \
+  \
+                                       /* transpose coefficients (phase 2) */ \
+  col01e = vec_mergeh(row04l, row26l); /* col01e=(00 20 40 60 01 21 41 61) */ \
+  col23e = vec_mergel(row04l, row26l); /* col23e=(02 22 42 62 03 23 43 63) */ \
+  col45e = vec_mergeh(row04h, row26h); /* col45e=(04 24 44 64 05 25 45 65) */ \
+  col67e = vec_mergel(row04h, row26h); /* col67e=(06 26 46 66 07 27 47 67) */ \
+  col01o = vec_mergeh(row15l, row37l); /* col01o=(10 30 50 70 11 31 51 71) */ \
+  col23o = vec_mergel(row15l, row37l); /* col23o=(12 32 52 72 13 33 53 73) */ \
+  col45o = vec_mergeh(row15h, row37h); /* col45o=(14 34 54 74 15 35 55 75) */ \
+  col67o = vec_mergel(row15h, row37h); /* col67o=(16 36 56 76 17 37 57 77) */ \
+  \
+                                       /* transpose coefficients (phase 3) */ \
+  col##0 = vec_mergeh(col01e, col01o); /* col0=(00 10 20 30 40 50 60 70) */ \
+  col##1 = vec_mergel(col01e, col01o); /* col1=(01 11 21 31 41 51 61 71) */ \
+  col##2 = vec_mergeh(col23e, col23o); /* col2=(02 12 22 32 42 52 62 72) */ \
+  col##3 = vec_mergel(col23e, col23o); /* col3=(03 13 23 33 43 53 63 73) */ \
+  col##4 = vec_mergeh(col45e, col45o); /* col4=(04 14 24 34 44 54 64 74) */ \
+  col##5 = vec_mergel(col45e, col45o); /* col5=(05 15 25 35 45 55 65 75) */ \
+  col##6 = vec_mergeh(col67e, col67o); /* col6=(06 16 26 36 46 56 66 76) */ \
+  col##7 = vec_mergel(col67e, col67o); /* col7=(07 17 27 37 47 57 67 77) */ \
+}
+
+#ifndef min
+#define min(a, b)  ((a) < (b) ? (a) : (b))
+#endif
+
+
+/* Macros to abstract big/little endian bit twiddling */
+
+#if __BIG_ENDIAN__
+
+#define VEC_LD(a, b)     vec_ld(a, b)
+#define VEC_ST(a, b, c)  vec_st(a, b, c)
+#define VEC_UNPACKHU(a)  vec_mergeh(pb_zero, a)
+#define VEC_UNPACKLU(a)  vec_mergel(pb_zero, a)
+
+#else
+
+#define VEC_LD(a, b)     vec_vsx_ld(a, b)
+#define VEC_ST(a, b, c)  vec_vsx_st(a, b, c)
+#define VEC_UNPACKHU(a)  vec_mergeh(a, pb_zero)
+#define VEC_UNPACKLU(a)  vec_mergel(a, pb_zero)
+
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64 Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jccolext-avx2.asm Added

@@ -0,0 +1,559 @@
+;
+; jccolext.asm - colorspace conversion (64-bit AVX2)
+;
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_avx2(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                            JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                            int num_rows);
+;
+
+; r10d = JDIMENSION img_width
+; r11 = JSAMPARRAY input_buf
+; r12 = JSAMPIMAGE output_buf
+; r13d = JDIMENSION output_row
+; r14d = int num_rows
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_YMMWORD  ; ymmword wkWK_NUM
+%define WK_NUM  8
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_ycc_convert_avx2)
+
+EXTN(jsimd_rgb_ycc_convert_avx2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 5
+    push        rbx
+
+    mov         ecx, r10d
+    test        rcx, rcx
+    jz          near .return
+
+    push        rcx
+
+    mov         rsi, r12
+    mov         ecx, r13d
+    mov         rdip, JSAMPARRAY rsi+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rsi+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rsi+2*SIZEOF_JSAMPARRAY
+    lea         rdi, rdi+rcx*SIZEOF_JSAMPROW
+    lea         rbx, rbx+rcx*SIZEOF_JSAMPROW
+    lea         rdx, rdx+rcx*SIZEOF_JSAMPROW
+
+    pop         rcx
+
+    mov         rsi, r11
+    mov         eax, r14d
+    test        rax, rax
+    jle         near .return
+.rowloop:
+    push        rdx
+    push        rbx
+    push        rdi
+    push        rsi
+    push        rcx                     ; col
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr0
+    mov         rbxp, JSAMPROW rbx    ; outptr1
+    mov         rdxp, JSAMPROW rdx    ; outptr2
+
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        rax
+    push        rdx
+    lea         rcx, rcx+rcx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         rcx, byte SIZEOF_BYTE
+    movzx       rax, byte rsi+rcx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         rcx, byte SIZEOF_WORD
+    movzx       rdx, word rsi+rcx
+    shl         rax, WORD_BIT
+    or          rax, rdx
+.column_ld4:
+    vmovd       xmmA, eax
+    pop         rdx
+    pop         rax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         rcx, byte SIZEOF_DWORD
+    vmovd       xmmF, XMM_DWORD rsi+rcx
+    vpslldq     xmmA, xmmA, SIZEOF_DWORD
+    vpor        xmmA, xmmA, xmmF
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    sub         rcx, byte SIZEOF_MMWORD
+    vmovq       xmmB, XMM_MMWORD rsi+rcx
+    vpslldq     xmmA, xmmA, SIZEOF_MMWORD
+    vpor        xmmA, xmmA, xmmB
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    jz          short .column_ld32
+    sub         rcx, byte SIZEOF_XMMWORD
+    vmovdqu     xmmB, XMM_MMWORD rsi+rcx
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    vpor        ymmA, ymmB
+.column_ld32:
+    test        cl, SIZEOF_YMMWORD
+    jz          short .column_ld64
+    sub         rcx, byte SIZEOF_YMMWORD
+    vmovdqa     ymmF, ymmA
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+.column_ld64:
+    test        cl, 2*SIZEOF_YMMWORD
+    mov         rcx, SIZEOF_YMMWORD
+    jz          short .rgb_ycc_cnv
+    vmovdqa     ymmB, ymmA
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD rsi+1*SIZEOF_YMMWORD
+    jmp         short .rgb_ycc_cnv
+
+.columnloop:
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD rsi+1*SIZEOF_YMMWORD
+    vmovdqu     ymmB, YMMWORD rsi+2*SIZEOF_YMMWORD
+
+.rgb_ycc_cnv:
+    ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+    ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    ; ymmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+    ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    ; ymmB=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+    ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    vmovdqu     ymmC, ymmA
+    vinserti128 ymmA, ymmF, xmmA, 0  ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                     ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vinserti128 ymmC, ymmC, xmmB, 0  ; ymmC=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+                                     ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    vinserti128 ymmB, ymmB, xmmF, 0  ; ymmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                     ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+    vperm2i128  ymmF, ymmC, ymmC, 1  ; ymmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A
+                                     ;       1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q)
+
+    vmovdqa     ymmG, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12
+                                  ;       22 03 13 23 04 14 24 05 0G 1G 2G 0H 1H 2H 0I 1I)
+    vpsrldq     ymmG, ymmG, 8     ; ymmG=(22 03 13 23 04 14 24 05 0G 1G 2G 0H 1H 2H 0I 1I
+                                  ;       2I 0J 1J 2J 0K 1K 2K 0L -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmF  ; ymmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A
+                                  ;       0G 0O 1G 1O 2G 2O 0H 0P 1H 1P 2H 2P 0I 0Q 1I 1Q)
+    vpslldq     ymmF, ymmF, 8     ; ymmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27
+                                  ;       08 18 28 09 19 29 0A 1A 1L 2L 0M 1M 2M 0N 1N 2N)
+
+    vpunpcklbw  ymmG, ymmG, ymmB  ; ymmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D
+                                  ;       2I 2Q 0J 0R 1J 1R 2J 2R 0K 0S 1K 1S 2K 2S 0L 0T)
+    vpunpckhbw  ymmF, ymmF, ymmB  ; ymmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F
+                                  ;       1L 1T 2L 2T 0M 0U 1M 1U 2M 2U 0N 0V 1N 1V 2N 2V)
+
+    vmovdqa     ymmD, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09
+                                  ;       11 19 21 29 02 0A 12 1A 0G 0O 1G 1O 2G 2O 0H 0P)
+    vpsrldq     ymmD, ymmD, 8     ; ymmD=(11 19 21 29 02 0A 12 1A 0G 0O 1G 1O 2G 2O 0H 0P
+                                  ;       1H 1P 2H 2P 0I 0Q 1I 1Q -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmG  ; ymmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D
+                                  ;       0G 0K 0O 0S 1G 1K 1O 1S 2G 2K 2O 2S 0H 0L 0P 0T)
+    vpslldq     ymmG, ymmG, 8     ; ymmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B
+                                  ;       04 0C 14 1C 24 2C 05 0D 2I 2Q 0J 0R 1J 1R 2J 2R)
+
+    vpunpcklbw  ymmD, ymmD, ymmF  ; ymmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E
+                                  ;       1H 1L 1P 1T 2H 2L 2P 2T 0I 0M 0Q 0U 1I 1M 1Q 1U)
+    vpunpckhbw  ymmG, ymmG, ymmF  ; ymmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F
+                                  ;       2I 2M 2Q 2U 0J 0N 0R 0V 1J 1N 1R 1V 2J 2N 2R 2V)
+
+    vmovdqa     ymmE, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C
+                                  ;       20 24 28 2C 01 05 09 0D 0G 0K 0O 0S 1G 1K 1O 1S)
+    vpsrldq     ymmE, ymmE, 8     ; ymmE=(20 24 28 2C 01 05 09 0D 0G 0K 0O 0S 1G 1K 1O 1S
+                                  ;       2G 2K 2O 2S 0H 0L 0P 0T -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmD  ; ymmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E
+                                  ;       0G 0I 0K 0M 0O 0Q 0S 0U 1G 1I 1K 1M 1O 1Q 1S 1U)
+    vpslldq     ymmD, ymmD, 8     ; ymmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D
+                                  ;       02 06 0A 0E 12 16 1A 1E 1H 1L 1P 1T 2H 2L 2P 2T)
+
+    vpunpcklbw  ymmE, ymmE, ymmG  ; ymmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F
+                                  ;       2G 2I 2K 2M 2O 2Q 2S 2U 0H 0J 0L 0N 0P 0R 0T 0V)
+    vpunpckhbw  ymmD, ymmD, ymmG  ; ymmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F
+                                  ;       1H 1J 1L 1N 1P 1R 1T 1V 2H 2J 2L 2N 2P 2R 2T 2V)
+
+    vpxor       ymmH, ymmH, ymmH
+
+    vmovdqa     ymmC, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmH  ; ymmA=(00 02 04 06 08 0A 0C 0E 0G 0I 0K 0M 0O 0Q 0S 0U)
+    vpunpckhbw  ymmC, ymmC, ymmH  ; ymmC=(10 12 14 16 18 1A 1C 1E 1G 1I 1K 1M 1O 1Q 1S 1U)
+
+    vmovdqa     ymmB, ymmE
+    vpunpcklbw  ymmE, ymmE, ymmH  ; ymmE=(20 22 24 26 28 2A 2C 2E 2G 2I 2K 2M 2O 2Q 2S 2U)
+    vpunpckhbw  ymmB, ymmB, ymmH  ; ymmB=(01 03 05 07 09 0B 0D 0F 0H 0J 0L 0N 0P 0R 0T 0V)
+
+    vmovdqa     ymmF, ymmD
+    vpunpcklbw  ymmD, ymmD, ymmH  ; ymmD=(11 13 15 17 19 1B 1D 1F 1H 1J 1L 1N 1P 1R 1T 1V)
+    vpunpckhbw  ymmF, ymmF, ymmH  ; ymmF=(21 23 25 27 29 2B 2D 2F 2H 2J 2L 2N 2P 2R 2T 2V)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_XMMWORD/16
+    jz          short .column_ld2
+    sub         rcx, byte SIZEOF_XMMWORD/16
+    vmovd       xmmA, XMM_DWORD rsi+rcx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_XMMWORD/8
+    jz          short .column_ld4
+    sub         rcx, byte SIZEOF_XMMWORD/8
+    vmovq       xmmF, XMM_MMWORD rsi+rcx*RGB_PIXELSIZE
+    vpslldq     xmmA, xmmA, SIZEOF_MMWORD
+    vpor        xmmA, xmmA, xmmF
+.column_ld4:
+    test        cl, SIZEOF_XMMWORD/4
+    jz          short .column_ld8
+    sub         rcx, byte SIZEOF_XMMWORD/4
+    vmovdqa     xmmF, xmmA
+    vperm2i128  ymmF, ymmF, ymmF, 1
+    vmovdqu     xmmA, XMMWORD rsi+rcx*RGB_PIXELSIZE
+    vpor        ymmA, ymmA, ymmF
+.column_ld8:
+    test        cl, SIZEOF_XMMWORD/2
+    jz          short .column_ld16
+    sub         rcx, byte SIZEOF_XMMWORD/2
+    vmovdqa     ymmF, ymmA
+    vmovdqu     ymmA, YMMWORD rsi+rcx*RGB_PIXELSIZE
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    mov         rcx, SIZEOF_YMMWORD
+    jz          short .rgb_ycc_cnv
+    vmovdqa     ymmE, ymmA
+    vmovdqa     ymmH, ymmF
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD rsi+1*SIZEOF_YMMWORD
+    jmp         short .rgb_ycc_cnv
+
+.columnloop:
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD rsi+1*SIZEOF_YMMWORD
+    vmovdqu     ymmE, YMMWORD rsi+2*SIZEOF_YMMWORD
+    vmovdqu     ymmH, YMMWORD rsi+3*SIZEOF_YMMWORD
+
+.rgb_ycc_cnv:
+    ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+    ;       04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    ; ymmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+    ;       0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+    ; ymmE=(0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J
+    ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    ; ymmH=(0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R
+    ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    vmovdqa     ymmB, ymmA
+    vinserti128 ymmA, ymmA, xmmE, 1     ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J)
+    vperm2i128  ymmE, ymmB, ymmE, 0x31  ; ymmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+
+    vmovdqa     ymmB, ymmF
+    vinserti128 ymmF, ymmF, xmmH, 1     ; ymmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R)
+    vperm2i128  ymmH, ymmB, ymmH, 0x31  ; ymmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    vmovdqa     ymmD, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmE      ; ymmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35
+                                      ;       0G 0K 1G 1K 2G 2K 3G 3K 0H 0L 1H 1L 2H 2L 3H 3L)
+    vpunpckhbw  ymmD, ymmD, ymmE      ; ymmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37
+                                      ;       0I 0M 1I 1M 2I 2M 3I 3M 0J 0N 1J 1N 2J 2N 3J 3N)
+
+    vmovdqa     ymmC, ymmF
+    vpunpcklbw  ymmF, ymmF, ymmH      ; ymmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D
+                                      ;       0O 0S 1O 1S 2O 2S 3O 3S 0P 0T 1P 1T 2P 2T 3P 3T)
+    vpunpckhbw  ymmC, ymmC, ymmH      ; ymmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F
+                                      ;       0Q 0U 1Q 1U 2Q 2U 3Q 3U 0R 0V 1R 1V 2R 2V 3R 3V)
+
+    vmovdqa     ymmB, ymmA
+    vpunpcklwd  ymmA, ymmA, ymmF      ; ymmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C
+                                      ;       0G 0K 0O 0S 1G 1K 1O 1S 2G 2K 2O 2S 3G 3K 3O 3S)
+    vpunpckhwd  ymmB, ymmB, ymmF      ; ymmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D
+                                      ;       0H 0L 0P 0T 1H 1L 1P 1T 2H 2L 2P 2T 3H 3L 3P 3T)
+
+    vmovdqa     ymmG, ymmD
+    vpunpcklwd  ymmD, ymmD, ymmC      ; ymmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E
+                                      ;       0I 0M 0Q 0U 1I 1M 1Q 1U 2I 2M 2Q 2U 3I 3M 3Q 3U)
+    vpunpckhwd  ymmG, ymmG, ymmC      ; ymmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F
+                                      ;       0J 0N 0R 0V 1J 1N 1R 1V 2J 2N 2R 2V 3J 3N 3R 3V)
+
+    vmovdqa     ymmE, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmD      ; ymmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E
+                                      ;       0G 0I 0K 0M 0O 0Q 0S 0U 1G 1I 1K 1M 1O 1Q 1S 1U)
+    vpunpckhbw  ymmE, ymmE, ymmD      ; ymmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E
+                                      ;       2G 2I 2K 2M 2O 2Q 2S 2U 3G 3I 3K 3M 3O 3Q 3S 3U)
+
+    vmovdqa     ymmH, ymmB
+    vpunpcklbw  ymmB, ymmB, ymmG      ; ymmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F
+                                      ;       0H 0J 0L 0N 0P 0R 0T 0V 1H 1J 1L 1N 1P 1R 1T 1V)
+    vpunpckhbw  ymmH, ymmH, ymmG      ; ymmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F
+                                      ;       2H 2J 2L 2N 2P 2R 2T 2V 3H 3J 3L 3N 3P 3R 3T 3V)
+
+    vpxor       ymmF, ymmF, ymmF
+
+    vmovdqa     ymmC, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmF      ; ymmA=(00 02 04 06 08 0A 0C 0E 0G 0I 0K 0M 0O 0Q 0S 0U)
+    vpunpckhbw  ymmC, ymmC, ymmF      ; ymmC=(10 12 14 16 18 1A 1C 1E 1G 1I 1K 1M 1O 1Q 1S 1U)
+
+    vmovdqa     ymmD, ymmB
+    vpunpcklbw  ymmB, ymmB, ymmF      ; ymmB=(01 03 05 07 09 0B 0D 0F 0H 0J 0L 0N 0P 0R 0T 0V)
+    vpunpckhbw  ymmD, ymmD, ymmF      ; ymmD=(11 13 15 17 19 1B 1D 1F 1H 1J 1L 1N 1P 1R 1T 1V)
+
+    vmovdqa     ymmG, ymmE
+    vpunpcklbw  ymmE, ymmE, ymmF      ; ymmE=(20 22 24 26 28 2A 2C 2E 2G 2I 2K 2M 2O 2Q 2S 2U)
+    vpunpckhbw  ymmG, ymmG, ymmF      ; ymmG=(30 32 34 36 38 3A 3C 3E 3G 3I 3K 3M 3O 3Q 3S 3U)
+
+    vpunpcklbw  ymmF, ymmF, ymmH
+    vpunpckhbw  ymmH, ymmH, ymmH
+    vpsrlw      ymmF, ymmF, BYTE_BIT  ; ymmF=(21 23 25 27 29 2B 2D 2F 2H 2J 2L 2N 2P 2R 2T 2V)
+    vpsrlw      ymmH, ymmH, BYTE_BIT  ; ymmH=(31 33 35 37 39 3B 3D 3F 3H 3J 3L 3N 3P 3R 3T 3V)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; ymm0=R(02468ACEGIKMOQSU)=RE, ymm2=G(02468ACEGIKMOQSU)=GE, ymm4=B(02468ACEGIKMOQSU)=BE
+    ; ymm1=R(13579BDFHJLNPRTV)=RO, ymm3=G(13579BDFHJLNPRTV)=GO, ymm5=B(13579BDFHJLNPRTV)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+    vmovdqa     YMMWORD wk(0), ymm0   ; wk(0)=RE
+    vmovdqa     YMMWORD wk(1), ymm1   ; wk(1)=RO
+    vmovdqa     YMMWORD wk(2), ymm4   ; wk(2)=BE
+    vmovdqa     YMMWORD wk(3), ymm5   ; wk(3)=BO
+
+    vmovdqa     ymm6, ymm1
+    vpunpcklwd  ymm1, ymm1, ymm3
+    vpunpckhwd  ymm6, ymm6, ymm3
+    vmovdqa     ymm7, ymm1
+    vmovdqa     ymm4, ymm6
+    vpmaddwd    ymm1, ymm1, rel PW_F0299_F0337  ; ymm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    vpmaddwd    ymm6, ymm6, rel PW_F0299_F0337  ; ymm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+    vpmaddwd    ymm7, ymm7, rel PW_MF016_MF033  ; ymm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+    vpmaddwd    ymm4, ymm4, rel PW_MF016_MF033  ; ymm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+    vmovdqa     YMMWORD wk(4), ymm1   ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+    vmovdqa     YMMWORD wk(5), ymm6   ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    vpxor       ymm1, ymm1, ymm1
+    vpxor       ymm6, ymm6, ymm6
+    vpunpcklwd  ymm1, ymm1, ymm5        ; ymm1=BOL
+    vpunpckhwd  ymm6, ymm6, ymm5        ; ymm6=BOH
+    vpsrld      ymm1, ymm1, 1           ; ymm1=BOL*FIX(0.500)
+    vpsrld      ymm6, ymm6, 1           ; ymm6=BOH*FIX(0.500)
+
+    vmovdqa     ymm5, rel PD_ONEHALFM1_CJ  ; ymm5=PD_ONEHALFM1_CJ
+
+    vpaddd      ymm7, ymm7, ymm1
+    vpaddd      ymm4, ymm4, ymm6
+    vpaddd      ymm7, ymm7, ymm5
+    vpaddd      ymm4, ymm4, ymm5
+    vpsrld      ymm7, ymm7, SCALEBITS   ; ymm7=CbOL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=CbOH
+    vpackssdw   ymm7, ymm7, ymm4        ; ymm7=CbO
+
+    vmovdqa     ymm1, YMMWORD wk(2)   ; ymm1=BE
+
+    vmovdqa     ymm6, ymm0
+    vpunpcklwd  ymm0, ymm0, ymm2
+    vpunpckhwd  ymm6, ymm6, ymm2
+    vmovdqa     ymm5, ymm0
+    vmovdqa     ymm4, ymm6
+    vpmaddwd    ymm0, ymm0, rel PW_F0299_F0337  ; ymm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    vpmaddwd    ymm6, ymm6, rel PW_F0299_F0337  ; ymm6=REH*FIX(0.299)+GEH*FIX(0.337)
+    vpmaddwd    ymm5, ymm5, rel PW_MF016_MF033  ; ymm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+    vpmaddwd    ymm4, ymm4, rel PW_MF016_MF033  ; ymm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+    vmovdqa     YMMWORD wk(6), ymm0   ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+    vmovdqa     YMMWORD wk(7), ymm6   ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    vpxor       ymm0, ymm0, ymm0
+    vpxor       ymm6, ymm6, ymm6
+    vpunpcklwd  ymm0, ymm0, ymm1        ; ymm0=BEL
+    vpunpckhwd  ymm6, ymm6, ymm1        ; ymm6=BEH
+    vpsrld      ymm0, ymm0, 1           ; ymm0=BEL*FIX(0.500)
+    vpsrld      ymm6, ymm6, 1           ; ymm6=BEH*FIX(0.500)
+
+    vmovdqa     ymm1, rel PD_ONEHALFM1_CJ  ; ymm1=PD_ONEHALFM1_CJ
+
+    vpaddd      ymm5, ymm5, ymm0
+    vpaddd      ymm4, ymm4, ymm6
+    vpaddd      ymm5, ymm5, ymm1
+    vpaddd      ymm4, ymm4, ymm1
+    vpsrld      ymm5, ymm5, SCALEBITS   ; ymm5=CbEL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=CbEH
+    vpackssdw   ymm5, ymm5, ymm4        ; ymm5=CbE
+
+    vpsllw      ymm7, ymm7, BYTE_BIT
+    vpor        ymm5, ymm5, ymm7        ; ymm5=Cb
+    vmovdqu     YMMWORD rbx, ymm5     ; Save Cb
+
+    vmovdqa     ymm0, YMMWORD wk(3)   ; ymm0=BO
+    vmovdqa     ymm6, YMMWORD wk(2)   ; ymm6=BE
+    vmovdqa     ymm1, YMMWORD wk(1)   ; ymm1=RO
+
+    vmovdqa     ymm4, ymm0
+    vpunpcklwd  ymm0, ymm0, ymm3
+    vpunpckhwd  ymm4, ymm4, ymm3
+    vmovdqa     ymm7, ymm0
+    vmovdqa     ymm5, ymm4
+    vpmaddwd    ymm0, ymm0, rel PW_F0114_F0250  ; ymm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    vpmaddwd    ymm4, ymm4, rel PW_F0114_F0250  ; ymm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+    vpmaddwd    ymm7, ymm7, rel PW_MF008_MF041  ; ymm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+    vpmaddwd    ymm5, ymm5, rel PW_MF008_MF041  ; ymm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+    vmovdqa     ymm3, rel PD_ONEHALF            ; ymm3=PD_ONEHALF
+
+    vpaddd      ymm0, ymm0, YMMWORD wk(4)
+    vpaddd      ymm4, ymm4, YMMWORD wk(5)
+    vpaddd      ymm0, ymm0, ymm3
+    vpaddd      ymm4, ymm4, ymm3
+    vpsrld      ymm0, ymm0, SCALEBITS   ; ymm0=YOL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=YOH
+    vpackssdw   ymm0, ymm0, ymm4        ; ymm0=YO
+
+    vpxor       ymm3, ymm3, ymm3
+    vpxor       ymm4, ymm4, ymm4
+    vpunpcklwd  ymm3, ymm3, ymm1        ; ymm3=ROL
+    vpunpckhwd  ymm4, ymm4, ymm1        ; ymm4=ROH
+    vpsrld      ymm3, ymm3, 1           ; ymm3=ROL*FIX(0.500)
+    vpsrld      ymm4, ymm4, 1           ; ymm4=ROH*FIX(0.500)
+
+    vmovdqa     ymm1, rel PD_ONEHALFM1_CJ  ; ymm1=PD_ONEHALFM1_CJ
+
+    vpaddd      ymm7, ymm7, ymm3
+    vpaddd      ymm5, ymm5, ymm4
+    vpaddd      ymm7, ymm7, ymm1
+    vpaddd      ymm5, ymm5, ymm1
+    vpsrld      ymm7, ymm7, SCALEBITS   ; ymm7=CrOL
+    vpsrld      ymm5, ymm5, SCALEBITS   ; ymm5=CrOH
+    vpackssdw   ymm7, ymm7, ymm5        ; ymm7=CrO
+
+    vmovdqa     ymm3, YMMWORD wk(0)   ; ymm3=RE
+
+    vmovdqa     ymm4, ymm6
+    vpunpcklwd  ymm6, ymm6, ymm2
+    vpunpckhwd  ymm4, ymm4, ymm2
+    vmovdqa     ymm1, ymm6
+    vmovdqa     ymm5, ymm4
+    vpmaddwd    ymm6, ymm6, rel PW_F0114_F0250  ; ymm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    vpmaddwd    ymm4, ymm4, rel PW_F0114_F0250  ; ymm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+    vpmaddwd    ymm1, ymm1, rel PW_MF008_MF041  ; ymm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+    vpmaddwd    ymm5, ymm5, rel PW_MF008_MF041  ; ymm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+    vmovdqa     ymm2, rel PD_ONEHALF            ; ymm2=PD_ONEHALF
+
+    vpaddd      ymm6, ymm6, YMMWORD wk(6)
+    vpaddd      ymm4, ymm4, YMMWORD wk(7)
+    vpaddd      ymm6, ymm6, ymm2
+    vpaddd      ymm4, ymm4, ymm2
+    vpsrld      ymm6, ymm6, SCALEBITS   ; ymm6=YEL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=YEH
+    vpackssdw   ymm6, ymm6, ymm4        ; ymm6=YE
+
+    vpsllw      ymm0, ymm0, BYTE_BIT
+    vpor        ymm6, ymm6, ymm0        ; ymm6=Y
+    vmovdqu     YMMWORD rdi, ymm6     ; Save Y
+
+    vpxor       ymm2, ymm2, ymm2
+    vpxor       ymm4, ymm4, ymm4
+    vpunpcklwd  ymm2, ymm2, ymm3        ; ymm2=REL
+    vpunpckhwd  ymm4, ymm4, ymm3        ; ymm4=REH
+    vpsrld      ymm2, ymm2, 1           ; ymm2=REL*FIX(0.500)
+    vpsrld      ymm4, ymm4, 1           ; ymm4=REH*FIX(0.500)
+
+    vmovdqa     ymm0, rel PD_ONEHALFM1_CJ  ; ymm0=PD_ONEHALFM1_CJ
+
+    vpaddd      ymm1, ymm1, ymm2
+    vpaddd      ymm5, ymm5, ymm4
+    vpaddd      ymm1, ymm1, ymm0
+    vpaddd      ymm5, ymm5, ymm0
+    vpsrld      ymm1, ymm1, SCALEBITS   ; ymm1=CrEL
+    vpsrld      ymm5, ymm5, SCALEBITS   ; ymm5=CrEH
+    vpackssdw   ymm1, ymm1, ymm5        ; ymm1=CrE
+
+    vpsllw      ymm7, ymm7, BYTE_BIT
+    vpor        ymm1, ymm1, ymm7        ; ymm1=Cr
+    vmovdqu     YMMWORD rdx, ymm1     ; Save Cr
+
+    sub         rcx, byte SIZEOF_YMMWORD
+    add         rsi, RGB_PIXELSIZE*SIZEOF_YMMWORD  ; inptr
+    add         rdi, byte SIZEOF_YMMWORD           ; outptr0
+    add         rbx, byte SIZEOF_YMMWORD           ; outptr1
+    add         rdx, byte SIZEOF_YMMWORD           ; outptr2
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+    test        rcx, rcx
+    jnz         near .column_ld1
+
+    pop         rcx                     ; col
+    pop         rsi
+    pop         rdi
+    pop         rbx
+    pop         rdx
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         rdi, byte SIZEOF_JSAMPROW
+    add         rbx, byte SIZEOF_JSAMPROW
+    add         rdx, byte SIZEOF_JSAMPROW
+    dec         rax                        ; num_rows
+    jg          near .rowloop
+
+.return:
+    pop         rbx
+    vzeroupper
+    uncollect_args 5
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jccolext-sse2.asm Added

@@ -0,0 +1,484 @@
+;
+; jccolext.asm - colorspace conversion (64-bit SSE2)
+;
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_sse2(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                            JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                            int num_rows);
+;
+
+; r10d = JDIMENSION img_width
+; r11 = JSAMPARRAY input_buf
+; r12 = JSAMPIMAGE output_buf
+; r13d = JDIMENSION output_row
+; r14d = int num_rows
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD  ; xmmword wkWK_NUM
+%define WK_NUM  8
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_ycc_convert_sse2)
+
+EXTN(jsimd_rgb_ycc_convert_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 5
+    push        rbx
+
+    mov         ecx, r10d
+    test        rcx, rcx
+    jz          near .return
+
+    push        rcx
+
+    mov         rsi, r12
+    mov         ecx, r13d
+    mov         rdip, JSAMPARRAY rsi+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rsi+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rsi+2*SIZEOF_JSAMPARRAY
+    lea         rdi, rdi+rcx*SIZEOF_JSAMPROW
+    lea         rbx, rbx+rcx*SIZEOF_JSAMPROW
+    lea         rdx, rdx+rcx*SIZEOF_JSAMPROW
+
+    pop         rcx
+
+    mov         rsi, r11
+    mov         eax, r14d
+    test        rax, rax
+    jle         near .return
+.rowloop:
+    push        rdx
+    push        rbx
+    push        rdi
+    push        rsi
+    push        rcx                     ; col
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr0
+    mov         rbxp, JSAMPROW rbx    ; outptr1
+    mov         rdxp, JSAMPROW rdx    ; outptr2
+
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        rax
+    push        rdx
+    lea         rcx, rcx+rcx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         rcx, byte SIZEOF_BYTE
+    movzx       rax, byte rsi+rcx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         rcx, byte SIZEOF_WORD
+    movzx       rdx, word rsi+rcx
+    shl         rax, WORD_BIT
+    or          rax, rdx
+.column_ld4:
+    movd        xmmA, eax
+    pop         rdx
+    pop         rax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         rcx, byte SIZEOF_DWORD
+    movd        xmmF, XMM_DWORD rsi+rcx
+    pslldq      xmmA, SIZEOF_DWORD
+    por         xmmA, xmmF
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    sub         rcx, byte SIZEOF_MMWORD
+    movq        xmmB, XMM_MMWORD rsi+rcx
+    pslldq      xmmA, SIZEOF_MMWORD
+    por         xmmA, xmmB
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    jz          short .column_ld32
+    movdqa      xmmF, xmmA
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    mov         rcx, SIZEOF_XMMWORD
+    jmp         short .rgb_ycc_cnv
+.column_ld32:
+    test        cl, 2*SIZEOF_XMMWORD
+    mov         rcx, SIZEOF_XMMWORD
+    jz          short .rgb_ycc_cnv
+    movdqa      xmmB, xmmA
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD rsi+1*SIZEOF_XMMWORD
+    jmp         short .rgb_ycc_cnv
+
+.columnloop:
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD rsi+1*SIZEOF_XMMWORD
+    movdqu      xmmB, XMMWORD rsi+2*SIZEOF_XMMWORD
+
+.rgb_ycc_cnv:
+    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+    ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+    movdqa      xmmG, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+    psrldq      xmmG, 8     ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmF  ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+    pslldq      xmmF, 8     ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+    punpcklbw   xmmG, xmmB  ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+    punpckhbw   xmmF, xmmB  ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+    movdqa      xmmD, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+    psrldq      xmmD, 8     ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmG  ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+    pslldq      xmmG, 8     ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+    punpcklbw   xmmD, xmmF  ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+    punpckhbw   xmmG, xmmF  ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+    movdqa      xmmE, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+    psrldq      xmmE, 8     ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmD  ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+    pslldq      xmmD, 8     ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+    punpcklbw   xmmE, xmmG  ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+    punpckhbw   xmmD, xmmG  ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+    pxor        xmmH, xmmH
+
+    movdqa      xmmC, xmmA
+    punpcklbw   xmmA, xmmH  ; xmmA=(00 02 04 06 08 0A 0C 0E)
+    punpckhbw   xmmC, xmmH  ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+    movdqa      xmmB, xmmE
+    punpcklbw   xmmE, xmmH  ; xmmE=(20 22 24 26 28 2A 2C 2E)
+    punpckhbw   xmmB, xmmH  ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+    movdqa      xmmF, xmmD
+    punpcklbw   xmmD, xmmH  ; xmmD=(11 13 15 17 19 1B 1D 1F)
+    punpckhbw   xmmF, xmmH  ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_XMMWORD/16
+    jz          short .column_ld2
+    sub         rcx, byte SIZEOF_XMMWORD/16
+    movd        xmmA, XMM_DWORD rsi+rcx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_XMMWORD/8
+    jz          short .column_ld4
+    sub         rcx, byte SIZEOF_XMMWORD/8
+    movq        xmmE, XMM_MMWORD rsi+rcx*RGB_PIXELSIZE
+    pslldq      xmmA, SIZEOF_MMWORD
+    por         xmmA, xmmE
+.column_ld4:
+    test        cl, SIZEOF_XMMWORD/4
+    jz          short .column_ld8
+    sub         rcx, byte SIZEOF_XMMWORD/4
+    movdqa      xmmE, xmmA
+    movdqu      xmmA, XMMWORD rsi+rcx*RGB_PIXELSIZE
+.column_ld8:
+    test        cl, SIZEOF_XMMWORD/2
+    mov         rcx, SIZEOF_XMMWORD
+    jz          short .rgb_ycc_cnv
+    movdqa      xmmF, xmmA
+    movdqa      xmmH, xmmE
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqu      xmmE, XMMWORD rsi+1*SIZEOF_XMMWORD
+    jmp         short .rgb_ycc_cnv
+
+.columnloop:
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqu      xmmE, XMMWORD rsi+1*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD rsi+2*SIZEOF_XMMWORD
+    movdqu      xmmH, XMMWORD rsi+3*SIZEOF_XMMWORD
+
+.rgb_ycc_cnv:
+    ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+    ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+    ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+    movdqa      xmmD, xmmA
+    punpcklbw   xmmA, xmmE      ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+    punpckhbw   xmmD, xmmE      ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+    movdqa      xmmC, xmmF
+    punpcklbw   xmmF, xmmH      ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+    punpckhbw   xmmC, xmmH      ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+    movdqa      xmmB, xmmA
+    punpcklwd   xmmA, xmmF      ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+    punpckhwd   xmmB, xmmF      ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+    movdqa      xmmG, xmmD
+    punpcklwd   xmmD, xmmC      ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+    punpckhwd   xmmG, xmmC      ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+    movdqa      xmmE, xmmA
+    punpcklbw   xmmA, xmmD      ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+    punpckhbw   xmmE, xmmD      ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+    movdqa      xmmH, xmmB
+    punpcklbw   xmmB, xmmG      ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+    punpckhbw   xmmH, xmmG      ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+    pxor        xmmF, xmmF
+
+    movdqa      xmmC, xmmA
+    punpcklbw   xmmA, xmmF      ; xmmA=(00 02 04 06 08 0A 0C 0E)
+    punpckhbw   xmmC, xmmF      ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+    movdqa      xmmD, xmmB
+    punpcklbw   xmmB, xmmF      ; xmmB=(01 03 05 07 09 0B 0D 0F)
+    punpckhbw   xmmD, xmmF      ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+    movdqa      xmmG, xmmE
+    punpcklbw   xmmE, xmmF      ; xmmE=(20 22 24 26 28 2A 2C 2E)
+    punpckhbw   xmmG, xmmF      ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+    punpcklbw   xmmF, xmmH
+    punpckhbw   xmmH, xmmH
+    psrlw       xmmF, BYTE_BIT  ; xmmF=(21 23 25 27 29 2B 2D 2F)
+    psrlw       xmmH, BYTE_BIT  ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+    ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+    ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+    ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+    movdqa      XMMWORD wk(0), xmm0   ; wk(0)=RE
+    movdqa      XMMWORD wk(1), xmm1   ; wk(1)=RO
+    movdqa      XMMWORD wk(2), xmm4   ; wk(2)=BE
+    movdqa      XMMWORD wk(3), xmm5   ; wk(3)=BO
+
+    movdqa      xmm6, xmm1
+    punpcklwd   xmm1, xmm3
+    punpckhwd   xmm6, xmm3
+    movdqa      xmm7, xmm1
+    movdqa      xmm4, xmm6
+    pmaddwd     xmm1, rel PW_F0299_F0337  ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    pmaddwd     xmm6, rel PW_F0299_F0337  ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+    pmaddwd     xmm7, rel PW_MF016_MF033  ; xmm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+    pmaddwd     xmm4, rel PW_MF016_MF033  ; xmm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+    movdqa      XMMWORD wk(4), xmm1   ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+    movdqa      XMMWORD wk(5), xmm6   ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    pxor        xmm1, xmm1
+    pxor        xmm6, xmm6
+    punpcklwd   xmm1, xmm5              ; xmm1=BOL
+    punpckhwd   xmm6, xmm5              ; xmm6=BOH
+    psrld       xmm1, 1                 ; xmm1=BOL*FIX(0.500)
+    psrld       xmm6, 1                 ; xmm6=BOH*FIX(0.500)
+
+    movdqa      xmm5, rel PD_ONEHALFM1_CJ  ; xmm5=PD_ONEHALFM1_CJ
+
+    paddd       xmm7, xmm1
+    paddd       xmm4, xmm6
+    paddd       xmm7, xmm5
+    paddd       xmm4, xmm5
+    psrld       xmm7, SCALEBITS         ; xmm7=CbOL
+    psrld       xmm4, SCALEBITS         ; xmm4=CbOH
+    packssdw    xmm7, xmm4              ; xmm7=CbO
+
+    movdqa      xmm1, XMMWORD wk(2)   ; xmm1=BE
+
+    movdqa      xmm6, xmm0
+    punpcklwd   xmm0, xmm2
+    punpckhwd   xmm6, xmm2
+    movdqa      xmm5, xmm0
+    movdqa      xmm4, xmm6
+    pmaddwd     xmm0, rel PW_F0299_F0337  ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    pmaddwd     xmm6, rel PW_F0299_F0337  ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+    pmaddwd     xmm5, rel PW_MF016_MF033  ; xmm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+    pmaddwd     xmm4, rel PW_MF016_MF033  ; xmm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+    movdqa      XMMWORD wk(6), xmm0   ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+    movdqa      XMMWORD wk(7), xmm6   ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    pxor        xmm0, xmm0
+    pxor        xmm6, xmm6
+    punpcklwd   xmm0, xmm1              ; xmm0=BEL
+    punpckhwd   xmm6, xmm1              ; xmm6=BEH
+    psrld       xmm0, 1                 ; xmm0=BEL*FIX(0.500)
+    psrld       xmm6, 1                 ; xmm6=BEH*FIX(0.500)
+
+    movdqa      xmm1, rel PD_ONEHALFM1_CJ  ; xmm1=PD_ONEHALFM1_CJ
+
+    paddd       xmm5, xmm0
+    paddd       xmm4, xmm6
+    paddd       xmm5, xmm1
+    paddd       xmm4, xmm1
+    psrld       xmm5, SCALEBITS         ; xmm5=CbEL
+    psrld       xmm4, SCALEBITS         ; xmm4=CbEH
+    packssdw    xmm5, xmm4              ; xmm5=CbE
+
+    psllw       xmm7, BYTE_BIT
+    por         xmm5, xmm7              ; xmm5=Cb
+    movdqa      XMMWORD rbx, xmm5     ; Save Cb
+
+    movdqa      xmm0, XMMWORD wk(3)   ; xmm0=BO
+    movdqa      xmm6, XMMWORD wk(2)   ; xmm6=BE
+    movdqa      xmm1, XMMWORD wk(1)   ; xmm1=RO
+
+    movdqa      xmm4, xmm0
+    punpcklwd   xmm0, xmm3
+    punpckhwd   xmm4, xmm3
+    movdqa      xmm7, xmm0
+    movdqa      xmm5, xmm4
+    pmaddwd     xmm0, rel PW_F0114_F0250  ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    pmaddwd     xmm4, rel PW_F0114_F0250  ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+    pmaddwd     xmm7, rel PW_MF008_MF041  ; xmm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+    pmaddwd     xmm5, rel PW_MF008_MF041  ; xmm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+    movdqa      xmm3, rel PD_ONEHALF      ; xmm3=PD_ONEHALF
+
+    paddd       xmm0, XMMWORD wk(4)
+    paddd       xmm4, XMMWORD wk(5)
+    paddd       xmm0, xmm3
+    paddd       xmm4, xmm3
+    psrld       xmm0, SCALEBITS         ; xmm0=YOL
+    psrld       xmm4, SCALEBITS         ; xmm4=YOH
+    packssdw    xmm0, xmm4              ; xmm0=YO
+
+    pxor        xmm3, xmm3
+    pxor        xmm4, xmm4
+    punpcklwd   xmm3, xmm1              ; xmm3=ROL
+    punpckhwd   xmm4, xmm1              ; xmm4=ROH
+    psrld       xmm3, 1                 ; xmm3=ROL*FIX(0.500)
+    psrld       xmm4, 1                 ; xmm4=ROH*FIX(0.500)
+
+    movdqa      xmm1, rel PD_ONEHALFM1_CJ  ; xmm1=PD_ONEHALFM1_CJ
+
+    paddd       xmm7, xmm3
+    paddd       xmm5, xmm4
+    paddd       xmm7, xmm1
+    paddd       xmm5, xmm1
+    psrld       xmm7, SCALEBITS         ; xmm7=CrOL
+    psrld       xmm5, SCALEBITS         ; xmm5=CrOH
+    packssdw    xmm7, xmm5              ; xmm7=CrO
+
+    movdqa      xmm3, XMMWORD wk(0)   ; xmm3=RE
+
+    movdqa      xmm4, xmm6
+    punpcklwd   xmm6, xmm2
+    punpckhwd   xmm4, xmm2
+    movdqa      xmm1, xmm6
+    movdqa      xmm5, xmm4
+    pmaddwd     xmm6, rel PW_F0114_F0250  ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    pmaddwd     xmm4, rel PW_F0114_F0250  ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+    pmaddwd     xmm1, rel PW_MF008_MF041  ; xmm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+    pmaddwd     xmm5, rel PW_MF008_MF041  ; xmm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+    movdqa      xmm2, rel PD_ONEHALF      ; xmm2=PD_ONEHALF
+
+    paddd       xmm6, XMMWORD wk(6)
+    paddd       xmm4, XMMWORD wk(7)
+    paddd       xmm6, xmm2
+    paddd       xmm4, xmm2
+    psrld       xmm6, SCALEBITS         ; xmm6=YEL
+    psrld       xmm4, SCALEBITS         ; xmm4=YEH
+    packssdw    xmm6, xmm4              ; xmm6=YE
+
+    psllw       xmm0, BYTE_BIT
+    por         xmm6, xmm0              ; xmm6=Y
+    movdqa      XMMWORD rdi, xmm6     ; Save Y
+
+    pxor        xmm2, xmm2
+    pxor        xmm4, xmm4
+    punpcklwd   xmm2, xmm3              ; xmm2=REL
+    punpckhwd   xmm4, xmm3              ; xmm4=REH
+    psrld       xmm2, 1                 ; xmm2=REL*FIX(0.500)
+    psrld       xmm4, 1                 ; xmm4=REH*FIX(0.500)
+
+    movdqa      xmm0, rel PD_ONEHALFM1_CJ  ; xmm0=PD_ONEHALFM1_CJ
+
+    paddd       xmm1, xmm2
+    paddd       xmm5, xmm4
+    paddd       xmm1, xmm0
+    paddd       xmm5, xmm0
+    psrld       xmm1, SCALEBITS         ; xmm1=CrEL
+    psrld       xmm5, SCALEBITS         ; xmm5=CrEH
+    packssdw    xmm1, xmm5              ; xmm1=CrE
+
+    psllw       xmm7, BYTE_BIT
+    por         xmm1, xmm7              ; xmm1=Cr
+    movdqa      XMMWORD rdx, xmm1     ; Save Cr
+
+    sub         rcx, byte SIZEOF_XMMWORD
+    add         rsi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+    add         rdi, byte SIZEOF_XMMWORD                ; outptr0
+    add         rbx, byte SIZEOF_XMMWORD                ; outptr1
+    add         rdx, byte SIZEOF_XMMWORD                ; outptr2
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+    test        rcx, rcx
+    jnz         near .column_ld1
+
+    pop         rcx                     ; col
+    pop         rsi
+    pop         rdi
+    pop         rbx
+    pop         rdx
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         rdi, byte SIZEOF_JSAMPROW
+    add         rbx, byte SIZEOF_JSAMPROW
+    add         rdx, byte SIZEOF_JSAMPROW
+    dec         rax                        ; num_rows
+    jg          near .rowloop
+
+.return:
+    pop         rbx
+    uncollect_args 5
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jccolor-avx2.asm Added

@@ -0,0 +1,121 @@
+;
+; jccolor.asm - colorspace conversion (64-bit AVX2)
+;
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_081 equ  5329                ; FIX(0.08131)
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_168 equ 11059                ; FIX(0.16874)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_331 equ 21709                ; FIX(0.33126)
+F_0_418 equ 27439                ; FIX(0.41869)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_ycc_convert_avx2)
+
+EXTN(jconst_rgb_ycc_convert_avx2):
+
+PW_F0299_F0337  times 8 dw  F_0_299,  F_0_337
+PW_F0114_F0250  times 8 dw  F_0_114,  F_0_250
+PW_MF016_MF033  times 8 dw -F_0_168, -F_0_331
+PW_MF008_MF041  times 8 dw -F_0_081, -F_0_418
+PD_ONEHALFM1_CJ times 8 dd  (1 << (SCALEBITS - 1)) - 1 + \
+                            (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 8 dd  (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extrgb_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extrgbx_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extbgr_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extbgrx_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extxbgr_ycc_convert_avx2
+%include "jccolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_avx2  jsimd_extxrgb_ycc_convert_avx2
+%include "jccolext-avx2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jccolor-sse2.asm Added

@@ -0,0 +1,120 @@
+;
+; jccolor.asm - colorspace conversion (64-bit SSE2)
+;
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_081 equ  5329                ; FIX(0.08131)
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_168 equ 11059                ; FIX(0.16874)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_331 equ 21709                ; FIX(0.33126)
+F_0_418 equ 27439                ; FIX(0.41869)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_ycc_convert_sse2)
+
+EXTN(jconst_rgb_ycc_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299,  F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114,  F_0_250
+PW_MF016_MF033  times 4 dw -F_0_168, -F_0_331
+PW_MF008_MF041  times 4 dw -F_0_081, -F_0_418
+PD_ONEHALFM1_CJ times 4 dd  (1 << (SCALEBITS - 1)) - 1 + \
+                            (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extrgb_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extrgbx_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extbgr_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extbgrx_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extxbgr_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2  jsimd_extxrgb_ycc_convert_sse2
+%include "jccolext-sse2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jcgray-avx2.asm Added

@@ -0,0 +1,113 @@
+;
+; jcgray.asm - grayscale colorspace conversion (64-bit AVX2)
+;
+; Copyright (C) 2011, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_gray_convert_avx2)
+
+EXTN(jconst_rgb_gray_convert_avx2):
+
+PW_F0299_F0337 times 8 dw F_0_299, F_0_337
+PW_F0114_F0250 times 8 dw F_0_114, F_0_250
+PD_ONEHALF     times 8 dd (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extrgb_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extrgbx_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extbgr_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extbgrx_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extxbgr_gray_convert_avx2
+%include "jcgryext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_avx2  jsimd_extxrgb_gray_convert_avx2
+%include "jcgryext-avx2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jcgray-sse2.asm Added

@@ -0,0 +1,112 @@
+;
+; jcgray.asm - grayscale colorspace conversion (64-bit SSE2)
+;
+; Copyright (C) 2011, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_114 equ  7471                ; FIX(0.11400)
+F_0_250 equ 16384                ; FIX(0.25000)
+F_0_299 equ 19595                ; FIX(0.29900)
+F_0_587 equ 38470                ; FIX(0.58700)
+F_0_337 equ (F_0_587 - F_0_250)  ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_rgb_gray_convert_sse2)
+
+EXTN(jconst_rgb_gray_convert_sse2):
+
+PW_F0299_F0337 times 4 dw F_0_299, F_0_337
+PW_F0114_F0250 times 4 dw F_0_114, F_0_250
+PD_ONEHALF     times 4 dd (1 << (SCALEBITS - 1))
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extrgb_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extrgbx_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extbgr_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extbgrx_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extxbgr_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2  jsimd_extxrgb_gray_convert_sse2
+%include "jcgryext-sse2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jcgryext-avx2.asm Added

@@ -0,0 +1,438 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (64-bit AVX2)
+;
+; Copyright (C) 2011, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_avx2(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                             JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                             int num_rows);
+;
+
+; r10d = JDIMENSION img_width
+; r11 = JSAMPARRAY input_buf
+; r12 = JSAMPIMAGE output_buf
+; r13d = JDIMENSION output_row
+; r14d = int num_rows
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_YMMWORD  ; ymmword wkWK_NUM
+%define WK_NUM  2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_gray_convert_avx2)
+
+EXTN(jsimd_rgb_gray_convert_avx2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 5
+    push        rbx
+
+    mov         ecx, r10d
+    test        rcx, rcx
+    jz          near .return
+
+    push        rcx
+
+    mov         rsi, r12
+    mov         ecx, r13d
+    mov         rdip, JSAMPARRAY rsi+0*SIZEOF_JSAMPARRAY
+    lea         rdi, rdi+rcx*SIZEOF_JSAMPROW
+
+    pop         rcx
+
+    mov         rsi, r11
+    mov         eax, r14d
+    test        rax, rax
+    jle         near .return
+.rowloop:
+    push        rdi
+    push        rsi
+    push        rcx                     ; col
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr0
+
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        rax
+    push        rdx
+    lea         rcx, rcx+rcx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         rcx, byte SIZEOF_BYTE
+    movzx       rax, byte rsi+rcx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         rcx, byte SIZEOF_WORD
+    movzx       rdx, word rsi+rcx
+    shl         rax, WORD_BIT
+    or          rax, rdx
+.column_ld4:
+    vmovd       xmmA, eax
+    pop         rdx
+    pop         rax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         rcx, byte SIZEOF_DWORD
+    vmovd       xmmF, XMM_DWORD rsi+rcx
+    vpslldq     xmmA, xmmA, SIZEOF_DWORD
+    vpor        xmmA, xmmA, xmmF
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    sub         rcx, byte SIZEOF_MMWORD
+    vmovq       xmmB, XMM_MMWORD rsi+rcx
+    vpslldq     xmmA, xmmA, SIZEOF_MMWORD
+    vpor        xmmA, xmmA, xmmB
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    jz          short .column_ld32
+    sub         rcx, byte SIZEOF_XMMWORD
+    vmovdqu     xmmB, XMM_MMWORD rsi+rcx
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    vpor        ymmA, ymmB
+.column_ld32:
+    test        cl, SIZEOF_YMMWORD
+    jz          short .column_ld64
+    sub         rcx, byte SIZEOF_YMMWORD
+    vmovdqa     ymmF, ymmA
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+.column_ld64:
+    test        cl, 2*SIZEOF_YMMWORD
+    mov         rcx, SIZEOF_YMMWORD
+    jz          short .rgb_gray_cnv
+    vmovdqa     ymmB, ymmA
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD rsi+1*SIZEOF_YMMWORD
+    jmp         short .rgb_gray_cnv
+
+.columnloop:
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD rsi+1*SIZEOF_YMMWORD
+    vmovdqu     ymmB, YMMWORD rsi+2*SIZEOF_YMMWORD
+
+.rgb_gray_cnv:
+    ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+    ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    ; ymmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+    ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    ; ymmB=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+    ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    vmovdqu     ymmC, ymmA
+    vinserti128 ymmA, ymmF, xmmA, 0  ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                     ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vinserti128 ymmC, ymmC, xmmB, 0  ; ymmC=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+                                     ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    vinserti128 ymmB, ymmB, xmmF, 0  ; ymmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                     ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+    vperm2i128  ymmF, ymmC, ymmC, 1  ; ymmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A
+                                     ;       1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q)
+
+    vmovdqa     ymmG, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12
+                                  ;       22 03 13 23 04 14 24 05 0G 1G 2G 0H 1H 2H 0I 1I)
+    vpsrldq     ymmG, ymmG, 8     ; ymmG=(22 03 13 23 04 14 24 05 0G 1G 2G 0H 1H 2H 0I 1I
+                                  ;       2I 0J 1J 2J 0K 1K 2K 0L -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmF  ; ymmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A
+                                  ;       0G 0O 1G 1O 2G 2O 0H 0P 1H 1P 2H 2P 0I 0Q 1I 1Q)
+    vpslldq     ymmF, ymmF, 8     ; ymmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27
+                                  ;       08 18 28 09 19 29 0A 1A 1L 2L 0M 1M 2M 0N 1N 2N)
+
+    vpunpcklbw  ymmG, ymmG, ymmB  ; ymmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D
+                                  ;       2I 2Q 0J 0R 1J 1R 2J 2R 0K 0S 1K 1S 2K 2S 0L 0T)
+    vpunpckhbw  ymmF, ymmF, ymmB  ; ymmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F
+                                  ;       1L 1T 2L 2T 0M 0U 1M 1U 2M 2U 0N 0V 1N 1V 2N 2V)
+
+    vmovdqa     ymmD, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09
+                                  ;       11 19 21 29 02 0A 12 1A 0G 0O 1G 1O 2G 2O 0H 0P)
+    vpsrldq     ymmD, ymmD, 8     ; ymmD=(11 19 21 29 02 0A 12 1A 0G 0O 1G 1O 2G 2O 0H 0P
+                                  ;       1H 1P 2H 2P 0I 0Q 1I 1Q -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmG  ; ymmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D
+                                  ;       0G 0K 0O 0S 1G 1K 1O 1S 2G 2K 2O 2S 0H 0L 0P 0T)
+    vpslldq     ymmG, ymmG, 8     ; ymmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B
+                                  ;       04 0C 14 1C 24 2C 05 0D 2I 2Q 0J 0R 1J 1R 2J 2R)
+
+    vpunpcklbw  ymmD, ymmD, ymmF  ; ymmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E
+                                  ;       1H 1L 1P 1T 2H 2L 2P 2T 0I 0M 0Q 0U 1I 1M 1Q 1U)
+    vpunpckhbw  ymmG, ymmG, ymmF  ; ymmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F
+                                  ;       2I 2M 2Q 2U 0J 0N 0R 0V 1J 1N 1R 1V 2J 2N 2R 2V)
+
+    vmovdqa     ymmE, ymmA
+    vpslldq     ymmA, ymmA, 8     ; ymmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C
+                                  ;       20 24 28 2C 01 05 09 0D 0G 0K 0O 0S 1G 1K 1O 1S)
+    vpsrldq     ymmE, ymmE, 8     ; ymmE=(20 24 28 2C 01 05 09 0D 0G 0K 0O 0S 1G 1K 1O 1S
+                                  ;       2G 2K 2O 2S 0H 0L 0P 0T -- -- -- -- -- -- -- --)
+
+    vpunpckhbw  ymmA, ymmA, ymmD  ; ymmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E
+                                  ;       0G 0I 0K 0M 0O 0Q 0S 0U 1G 1I 1K 1M 1O 1Q 1S 1U)
+    vpslldq     ymmD, ymmD, 8     ; ymmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D
+                                  ;       02 06 0A 0E 12 16 1A 1E 1H 1L 1P 1T 2H 2L 2P 2T)
+
+    vpunpcklbw  ymmE, ymmE, ymmG  ; ymmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F
+                                  ;       2G 2I 2K 2M 2O 2Q 2S 2U 0H 0J 0L 0N 0P 0R 0T 0V)
+    vpunpckhbw  ymmD, ymmD, ymmG  ; ymmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F
+                                  ;       1H 1J 1L 1N 1P 1R 1T 1V 2H 2J 2L 2N 2P 2R 2T 2V)
+
+    vpxor       ymmH, ymmH, ymmH
+
+    vmovdqa     ymmC, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmH  ; ymmA=(00 02 04 06 08 0A 0C 0E 0G 0I 0K 0M 0O 0Q 0S 0U)
+    vpunpckhbw  ymmC, ymmC, ymmH  ; ymmC=(10 12 14 16 18 1A 1C 1E 1G 1I 1K 1M 1O 1Q 1S 1U)
+
+    vmovdqa     ymmB, ymmE
+    vpunpcklbw  ymmE, ymmE, ymmH  ; ymmE=(20 22 24 26 28 2A 2C 2E 2G 2I 2K 2M 2O 2Q 2S 2U)
+    vpunpckhbw  ymmB, ymmB, ymmH  ; ymmB=(01 03 05 07 09 0B 0D 0F 0H 0J 0L 0N 0P 0R 0T 0V)
+
+    vmovdqa     ymmF, ymmD
+    vpunpcklbw  ymmD, ymmD, ymmH  ; ymmD=(11 13 15 17 19 1B 1D 1F 1H 1J 1L 1N 1P 1R 1T 1V)
+    vpunpckhbw  ymmF, ymmF, ymmH  ; ymmF=(21 23 25 27 29 2B 2D 2F 2H 2J 2L 2N 2P 2R 2T 2V)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_XMMWORD/16
+    jz          short .column_ld2
+    sub         rcx, byte SIZEOF_XMMWORD/16
+    vmovd       xmmA, XMM_DWORD rsi+rcx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_XMMWORD/8
+    jz          short .column_ld4
+    sub         rcx, byte SIZEOF_XMMWORD/8
+    vmovq       xmmF, XMM_MMWORD rsi+rcx*RGB_PIXELSIZE
+    vpslldq     xmmA, xmmA, SIZEOF_MMWORD
+    vpor        xmmA, xmmA, xmmF
+.column_ld4:
+    test        cl, SIZEOF_XMMWORD/4
+    jz          short .column_ld8
+    sub         rcx, byte SIZEOF_XMMWORD/4
+    vmovdqa     xmmF, xmmA
+    vperm2i128  ymmF, ymmF, ymmF, 1
+    vmovdqu     xmmA, XMMWORD rsi+rcx*RGB_PIXELSIZE
+    vpor        ymmA, ymmA, ymmF
+.column_ld8:
+    test        cl, SIZEOF_XMMWORD/2
+    jz          short .column_ld16
+    sub         rcx, byte SIZEOF_XMMWORD/2
+    vmovdqa     ymmF, ymmA
+    vmovdqu     ymmA, YMMWORD rsi+rcx*RGB_PIXELSIZE
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    mov         rcx, SIZEOF_YMMWORD
+    jz          short .rgb_gray_cnv
+    vmovdqa     ymmE, ymmA
+    vmovdqa     ymmH, ymmF
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD rsi+1*SIZEOF_YMMWORD
+    jmp         short .rgb_gray_cnv
+
+.columnloop:
+    vmovdqu     ymmA, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymmF, YMMWORD rsi+1*SIZEOF_YMMWORD
+    vmovdqu     ymmE, YMMWORD rsi+2*SIZEOF_YMMWORD
+    vmovdqu     ymmH, YMMWORD rsi+3*SIZEOF_YMMWORD
+
+.rgb_gray_cnv:
+    ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+    ;       04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    ; ymmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+    ;       0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+    ; ymmE=(0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J
+    ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    ; ymmH=(0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R
+    ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    vmovdqa     ymmB, ymmA
+    vinserti128 ymmA, ymmA, xmmE, 1     ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J)
+    vperm2i128  ymmE, ymmB, ymmE, 0x31  ; ymmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+
+    vmovdqa     ymmB, ymmF
+    vinserti128 ymmF, ymmF, xmmH, 1     ; ymmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R)
+    vperm2i128  ymmH, ymmB, ymmH, 0x31  ; ymmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    vmovdqa     ymmD, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmE      ; ymmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35
+                                      ;       0G 0K 1G 1K 2G 2K 3G 3K 0H 0L 1H 1L 2H 2L 3H 3L)
+    vpunpckhbw  ymmD, ymmD, ymmE      ; ymmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37
+                                      ;       0I 0M 1I 1M 2I 2M 3I 3M 0J 0N 1J 1N 2J 2N 3J 3N)
+
+    vmovdqa     ymmC, ymmF
+    vpunpcklbw  ymmF, ymmF, ymmH      ; ymmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D
+                                      ;       0O 0S 1O 1S 2O 2S 3O 3S 0P 0T 1P 1T 2P 2T 3P 3T)
+    vpunpckhbw  ymmC, ymmC, ymmH      ; ymmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F
+                                      ;       0Q 0U 1Q 1U 2Q 2U 3Q 3U 0R 0V 1R 1V 2R 2V 3R 3V)
+
+    vmovdqa     ymmB, ymmA
+    vpunpcklwd  ymmA, ymmA, ymmF      ; ymmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C
+                                      ;       0G 0K 0O 0S 1G 1K 1O 1S 2G 2K 2O 2S 3G 3K 3O 3S)
+    vpunpckhwd  ymmB, ymmB, ymmF      ; ymmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D
+                                      ;       0H 0L 0P 0T 1H 1L 1P 1T 2H 2L 2P 2T 3H 3L 3P 3T)
+
+    vmovdqa     ymmG, ymmD
+    vpunpcklwd  ymmD, ymmD, ymmC      ; ymmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E
+                                      ;       0I 0M 0Q 0U 1I 1M 1Q 1U 2I 2M 2Q 2U 3I 3M 3Q 3U)
+    vpunpckhwd  ymmG, ymmG, ymmC      ; ymmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F
+                                      ;       0J 0N 0R 0V 1J 1N 1R 1V 2J 2N 2R 2V 3J 3N 3R 3V)
+
+    vmovdqa     ymmE, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmD      ; ymmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E
+                                      ;       0G 0I 0K 0M 0O 0Q 0S 0U 1G 1I 1K 1M 1O 1Q 1S 1U)
+    vpunpckhbw  ymmE, ymmE, ymmD      ; ymmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E
+                                      ;       2G 2I 2K 2M 2O 2Q 2S 2U 3G 3I 3K 3M 3O 3Q 3S 3U)
+
+    vmovdqa     ymmH, ymmB
+    vpunpcklbw  ymmB, ymmB, ymmG      ; ymmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F
+                                      ;       0H 0J 0L 0N 0P 0R 0T 0V 1H 1J 1L 1N 1P 1R 1T 1V)
+    vpunpckhbw  ymmH, ymmH, ymmG      ; ymmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F
+                                      ;       2H 2J 2L 2N 2P 2R 2T 2V 3H 3J 3L 3N 3P 3R 3T 3V)
+
+    vpxor       ymmF, ymmF, ymmF
+
+    vmovdqa     ymmC, ymmA
+    vpunpcklbw  ymmA, ymmA, ymmF      ; ymmA=(00 02 04 06 08 0A 0C 0E 0G 0I 0K 0M 0O 0Q 0S 0U)
+    vpunpckhbw  ymmC, ymmC, ymmF      ; ymmC=(10 12 14 16 18 1A 1C 1E 1G 1I 1K 1M 1O 1Q 1S 1U)
+
+    vmovdqa     ymmD, ymmB
+    vpunpcklbw  ymmB, ymmB, ymmF      ; ymmB=(01 03 05 07 09 0B 0D 0F 0H 0J 0L 0N 0P 0R 0T 0V)
+    vpunpckhbw  ymmD, ymmD, ymmF      ; ymmD=(11 13 15 17 19 1B 1D 1F 1H 1J 1L 1N 1P 1R 1T 1V)
+
+    vmovdqa     ymmG, ymmE
+    vpunpcklbw  ymmE, ymmE, ymmF      ; ymmE=(20 22 24 26 28 2A 2C 2E 2G 2I 2K 2M 2O 2Q 2S 2U)
+    vpunpckhbw  ymmG, ymmG, ymmF      ; ymmG=(30 32 34 36 38 3A 3C 3E 3G 3I 3K 3M 3O 3Q 3S 3U)
+
+    vpunpcklbw  ymmF, ymmF, ymmH
+    vpunpckhbw  ymmH, ymmH, ymmH
+    vpsrlw      ymmF, ymmF, BYTE_BIT  ; ymmF=(21 23 25 27 29 2B 2D 2F 2H 2J 2L 2N 2P 2R 2T 2V)
+    vpsrlw      ymmH, ymmH, BYTE_BIT  ; ymmH=(31 33 35 37 39 3B 3D 3F 3H 3J 3L 3N 3P 3R 3T 3V)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; ymm0=R(02468ACEGIKMOQSU)=RE, ymm2=G(02468ACEGIKMOQSU)=GE, ymm4=B(02468ACEGIKMOQSU)=BE
+    ; ymm1=R(13579BDFHJLNPRTV)=RO, ymm3=G(13579BDFHJLNPRTV)=GO, ymm5=B(13579BDFHJLNPRTV)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+    vmovdqa     ymm6, ymm1
+    vpunpcklwd  ymm1, ymm1, ymm3
+    vpunpckhwd  ymm6, ymm6, ymm3
+    vpmaddwd    ymm1, ymm1, rel PW_F0299_F0337  ; ymm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    vpmaddwd    ymm6, ymm6, rel PW_F0299_F0337  ; ymm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    vmovdqa     ymm7, ymm6              ; ymm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    vmovdqa     ymm6, ymm0
+    vpunpcklwd  ymm0, ymm0, ymm2
+    vpunpckhwd  ymm6, ymm6, ymm2
+    vpmaddwd    ymm0, ymm0, rel PW_F0299_F0337  ; ymm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    vpmaddwd    ymm6, ymm6, rel PW_F0299_F0337  ; ymm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    vmovdqa     YMMWORD wk(0), ymm0   ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+    vmovdqa     YMMWORD wk(1), ymm6   ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    vmovdqa     ymm0, ymm5              ; ymm0=BO
+    vmovdqa     ymm6, ymm4              ; ymm6=BE
+
+    vmovdqa     ymm4, ymm0
+    vpunpcklwd  ymm0, ymm0, ymm3
+    vpunpckhwd  ymm4, ymm4, ymm3
+    vpmaddwd    ymm0, ymm0, rel PW_F0114_F0250  ; ymm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    vpmaddwd    ymm4, ymm4, rel PW_F0114_F0250  ; ymm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+    vmovdqa     ymm3, rel PD_ONEHALF            ; ymm3=PD_ONEHALF
+
+    vpaddd      ymm0, ymm0, ymm1
+    vpaddd      ymm4, ymm4, ymm7
+    vpaddd      ymm0, ymm0, ymm3
+    vpaddd      ymm4, ymm4, ymm3
+    vpsrld      ymm0, ymm0, SCALEBITS   ; ymm0=YOL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=YOH
+    vpackssdw   ymm0, ymm0, ymm4        ; ymm0=YO
+
+    vmovdqa     ymm4, ymm6
+    vpunpcklwd  ymm6, ymm6, ymm2
+    vpunpckhwd  ymm4, ymm4, ymm2
+    vpmaddwd    ymm6, ymm6, rel PW_F0114_F0250  ; ymm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    vpmaddwd    ymm4, ymm4, rel PW_F0114_F0250  ; ymm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+    vmovdqa     ymm2, rel PD_ONEHALF            ; ymm2=PD_ONEHALF
+
+    vpaddd      ymm6, ymm6, YMMWORD wk(0)
+    vpaddd      ymm4, ymm4, YMMWORD wk(1)
+    vpaddd      ymm6, ymm6, ymm2
+    vpaddd      ymm4, ymm4, ymm2
+    vpsrld      ymm6, ymm6, SCALEBITS   ; ymm6=YEL
+    vpsrld      ymm4, ymm4, SCALEBITS   ; ymm4=YEH
+    vpackssdw   ymm6, ymm6, ymm4        ; ymm6=YE
+
+    vpsllw      ymm0, ymm0, BYTE_BIT
+    vpor        ymm6, ymm6, ymm0        ; ymm6=Y
+    vmovdqu     YMMWORD rdi, ymm6     ; Save Y
+
+    sub         rcx, byte SIZEOF_YMMWORD
+    add         rsi, RGB_PIXELSIZE*SIZEOF_YMMWORD  ; inptr
+    add         rdi, byte SIZEOF_YMMWORD           ; outptr0
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+    test        rcx, rcx
+    jnz         near .column_ld1
+
+    pop         rcx                     ; col
+    pop         rsi
+    pop         rdi
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         rdi, byte SIZEOF_JSAMPROW
+    dec         rax                        ; num_rows
+    jg          near .rowloop
+
+.return:
+    pop         rbx
+    vzeroupper
+    uncollect_args 5
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jcgryext-sse2.asm Added

@@ -0,0 +1,363 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (64-bit SSE2)
+;
+; Copyright (C) 2011, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_sse2(JDIMENSION img_width, JSAMPARRAY input_buf,
+;                             JSAMPIMAGE output_buf, JDIMENSION output_row,
+;                             int num_rows);
+;
+
+; r10d = JDIMENSION img_width
+; r11 = JSAMPARRAY input_buf
+; r12 = JSAMPIMAGE output_buf
+; r13d = JDIMENSION output_row
+; r14d = int num_rows
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD  ; xmmword wkWK_NUM
+%define WK_NUM  2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_rgb_gray_convert_sse2)
+
+EXTN(jsimd_rgb_gray_convert_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 5
+    push        rbx
+
+    mov         ecx, r10d
+    test        rcx, rcx
+    jz          near .return
+
+    push        rcx
+
+    mov         rsi, r12
+    mov         ecx, r13d
+    mov         rdip, JSAMPARRAY rsi+0*SIZEOF_JSAMPARRAY
+    lea         rdi, rdi+rcx*SIZEOF_JSAMPROW
+
+    pop         rcx
+
+    mov         rsi, r11
+    mov         eax, r14d
+    test        rax, rax
+    jle         near .return
+.rowloop:
+    push        rdi
+    push        rsi
+    push        rcx                     ; col
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr0
+
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+.column_ld1:
+    push        rax
+    push        rdx
+    lea         rcx, rcx+rcx*2        ; imul ecx,RGB_PIXELSIZE
+    test        cl, SIZEOF_BYTE
+    jz          short .column_ld2
+    sub         rcx, byte SIZEOF_BYTE
+    movzx       rax, byte rsi+rcx
+.column_ld2:
+    test        cl, SIZEOF_WORD
+    jz          short .column_ld4
+    sub         rcx, byte SIZEOF_WORD
+    movzx       rdx, word rsi+rcx
+    shl         rax, WORD_BIT
+    or          rax, rdx
+.column_ld4:
+    movd        xmmA, eax
+    pop         rdx
+    pop         rax
+    test        cl, SIZEOF_DWORD
+    jz          short .column_ld8
+    sub         rcx, byte SIZEOF_DWORD
+    movd        xmmF, XMM_DWORD rsi+rcx
+    pslldq      xmmA, SIZEOF_DWORD
+    por         xmmA, xmmF
+.column_ld8:
+    test        cl, SIZEOF_MMWORD
+    jz          short .column_ld16
+    sub         rcx, byte SIZEOF_MMWORD
+    movq        xmmB, XMM_MMWORD rsi+rcx
+    pslldq      xmmA, SIZEOF_MMWORD
+    por         xmmA, xmmB
+.column_ld16:
+    test        cl, SIZEOF_XMMWORD
+    jz          short .column_ld32
+    movdqa      xmmF, xmmA
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    mov         rcx, SIZEOF_XMMWORD
+    jmp         short .rgb_gray_cnv
+.column_ld32:
+    test        cl, 2*SIZEOF_XMMWORD
+    mov         rcx, SIZEOF_XMMWORD
+    jz          short .rgb_gray_cnv
+    movdqa      xmmB, xmmA
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD rsi+1*SIZEOF_XMMWORD
+    jmp         short .rgb_gray_cnv
+
+.columnloop:
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD rsi+1*SIZEOF_XMMWORD
+    movdqu      xmmB, XMMWORD rsi+2*SIZEOF_XMMWORD
+
+.rgb_gray_cnv:
+    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+    ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+    movdqa      xmmG, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+    psrldq      xmmG, 8     ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmF  ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+    pslldq      xmmF, 8     ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+    punpcklbw   xmmG, xmmB  ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+    punpckhbw   xmmF, xmmB  ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+    movdqa      xmmD, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+    psrldq      xmmD, 8     ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmG  ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+    pslldq      xmmG, 8     ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+    punpcklbw   xmmD, xmmF  ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+    punpckhbw   xmmG, xmmF  ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+    movdqa      xmmE, xmmA
+    pslldq      xmmA, 8     ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+    psrldq      xmmE, 8     ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+    punpckhbw   xmmA, xmmD  ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+    pslldq      xmmD, 8     ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+    punpcklbw   xmmE, xmmG  ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+    punpckhbw   xmmD, xmmG  ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+    pxor        xmmH, xmmH
+
+    movdqa      xmmC, xmmA
+    punpcklbw   xmmA, xmmH  ; xmmA=(00 02 04 06 08 0A 0C 0E)
+    punpckhbw   xmmC, xmmH  ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+    movdqa      xmmB, xmmE
+    punpcklbw   xmmE, xmmH  ; xmmE=(20 22 24 26 28 2A 2C 2E)
+    punpckhbw   xmmB, xmmH  ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+    movdqa      xmmF, xmmD
+    punpcklbw   xmmD, xmmH  ; xmmD=(11 13 15 17 19 1B 1D 1F)
+    punpckhbw   xmmF, xmmH  ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+    test        cl, SIZEOF_XMMWORD/16
+    jz          short .column_ld2
+    sub         rcx, byte SIZEOF_XMMWORD/16
+    movd        xmmA, XMM_DWORD rsi+rcx*RGB_PIXELSIZE
+.column_ld2:
+    test        cl, SIZEOF_XMMWORD/8
+    jz          short .column_ld4
+    sub         rcx, byte SIZEOF_XMMWORD/8
+    movq        xmmE, XMM_MMWORD rsi+rcx*RGB_PIXELSIZE
+    pslldq      xmmA, SIZEOF_MMWORD
+    por         xmmA, xmmE
+.column_ld4:
+    test        cl, SIZEOF_XMMWORD/4
+    jz          short .column_ld8
+    sub         rcx, byte SIZEOF_XMMWORD/4
+    movdqa      xmmE, xmmA
+    movdqu      xmmA, XMMWORD rsi+rcx*RGB_PIXELSIZE
+.column_ld8:
+    test        cl, SIZEOF_XMMWORD/2
+    mov         rcx, SIZEOF_XMMWORD
+    jz          short .rgb_gray_cnv
+    movdqa      xmmF, xmmA
+    movdqa      xmmH, xmmE
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqu      xmmE, XMMWORD rsi+1*SIZEOF_XMMWORD
+    jmp         short .rgb_gray_cnv
+
+.columnloop:
+    movdqu      xmmA, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqu      xmmE, XMMWORD rsi+1*SIZEOF_XMMWORD
+    movdqu      xmmF, XMMWORD rsi+2*SIZEOF_XMMWORD
+    movdqu      xmmH, XMMWORD rsi+3*SIZEOF_XMMWORD
+
+.rgb_gray_cnv:
+    ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+    ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+    ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+    movdqa      xmmD, xmmA
+    punpcklbw   xmmA, xmmE      ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+    punpckhbw   xmmD, xmmE      ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+    movdqa      xmmC, xmmF
+    punpcklbw   xmmF, xmmH      ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+    punpckhbw   xmmC, xmmH      ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+    movdqa      xmmB, xmmA
+    punpcklwd   xmmA, xmmF      ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+    punpckhwd   xmmB, xmmF      ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+    movdqa      xmmG, xmmD
+    punpcklwd   xmmD, xmmC      ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+    punpckhwd   xmmG, xmmC      ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+    movdqa      xmmE, xmmA
+    punpcklbw   xmmA, xmmD      ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+    punpckhbw   xmmE, xmmD      ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+    movdqa      xmmH, xmmB
+    punpcklbw   xmmB, xmmG      ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+    punpckhbw   xmmH, xmmG      ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+    pxor        xmmF, xmmF
+
+    movdqa      xmmC, xmmA
+    punpcklbw   xmmA, xmmF      ; xmmA=(00 02 04 06 08 0A 0C 0E)
+    punpckhbw   xmmC, xmmF      ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+    movdqa      xmmD, xmmB
+    punpcklbw   xmmB, xmmF      ; xmmB=(01 03 05 07 09 0B 0D 0F)
+    punpckhbw   xmmD, xmmF      ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+    movdqa      xmmG, xmmE
+    punpcklbw   xmmE, xmmF      ; xmmE=(20 22 24 26 28 2A 2C 2E)
+    punpckhbw   xmmG, xmmF      ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+    punpcklbw   xmmF, xmmH
+    punpckhbw   xmmH, xmmH
+    psrlw       xmmF, BYTE_BIT  ; xmmF=(21 23 25 27 29 2B 2D 2F)
+    psrlw       xmmH, BYTE_BIT  ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+    ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+    ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+    ; (Original)
+    ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+    ;
+    ; (This implementation)
+    ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+    movdqa      xmm6, xmm1
+    punpcklwd   xmm1, xmm3
+    punpckhwd   xmm6, xmm3
+    pmaddwd     xmm1, rel PW_F0299_F0337  ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+    pmaddwd     xmm6, rel PW_F0299_F0337  ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    movdqa      xmm7, xmm6              ; xmm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+    movdqa      xmm6, xmm0
+    punpcklwd   xmm0, xmm2
+    punpckhwd   xmm6, xmm2
+    pmaddwd     xmm0, rel PW_F0299_F0337  ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+    pmaddwd     xmm6, rel PW_F0299_F0337  ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    movdqa      XMMWORD wk(0), xmm0   ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+    movdqa      XMMWORD wk(1), xmm6   ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+    movdqa      xmm0, xmm5              ; xmm0=BO
+    movdqa      xmm6, xmm4              ; xmm6=BE
+
+    movdqa      xmm4, xmm0
+    punpcklwd   xmm0, xmm3
+    punpckhwd   xmm4, xmm3
+    pmaddwd     xmm0, rel PW_F0114_F0250  ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+    pmaddwd     xmm4, rel PW_F0114_F0250  ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+    movdqa      xmm3, rel PD_ONEHALF      ; xmm3=PD_ONEHALF
+
+    paddd       xmm0, xmm1
+    paddd       xmm4, xmm7
+    paddd       xmm0, xmm3
+    paddd       xmm4, xmm3
+    psrld       xmm0, SCALEBITS         ; xmm0=YOL
+    psrld       xmm4, SCALEBITS         ; xmm4=YOH
+    packssdw    xmm0, xmm4              ; xmm0=YO
+
+    movdqa      xmm4, xmm6
+    punpcklwd   xmm6, xmm2
+    punpckhwd   xmm4, xmm2
+    pmaddwd     xmm6, rel PW_F0114_F0250  ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+    pmaddwd     xmm4, rel PW_F0114_F0250  ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+    movdqa      xmm2, rel PD_ONEHALF      ; xmm2=PD_ONEHALF
+
+    paddd       xmm6, XMMWORD wk(0)
+    paddd       xmm4, XMMWORD wk(1)
+    paddd       xmm6, xmm2
+    paddd       xmm4, xmm2
+    psrld       xmm6, SCALEBITS         ; xmm6=YEL
+    psrld       xmm4, SCALEBITS         ; xmm4=YEH
+    packssdw    xmm6, xmm4              ; xmm6=YE
+
+    psllw       xmm0, BYTE_BIT
+    por         xmm6, xmm0              ; xmm6=Y
+    movdqa      XMMWORD rdi, xmm6     ; Save Y
+
+    sub         rcx, byte SIZEOF_XMMWORD
+    add         rsi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+    add         rdi, byte SIZEOF_XMMWORD                ; outptr0
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+    test        rcx, rcx
+    jnz         near .column_ld1
+
+    pop         rcx                     ; col
+    pop         rsi
+    pop         rdi
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_buf
+    add         rdi, byte SIZEOF_JSAMPROW
+    dec         rax                        ; num_rows
+    jg          near .rowloop
+
+.return:
+    pop         rbx
+    uncollect_args 5
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jchuff-sse2.asm Added

@@ -0,0 +1,583 @@
+;
+; jchuff-sse2.asm - Huffman entropy encoding (64-bit SSE2)
+;
+; Copyright (C) 2009-2011, 2014-2016, 2019, 2021, D. R. Commander.
+; Copyright (C) 2015, Matthieu Darbois.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains an SSE2 implementation for Huffman coding of one block.
+; The following code is based on jchuff.c; see jchuff.c for more details.
+
+%include "jsimdext.inc"
+
+struc working_state
+.next_output_byte:   resp 1     ; => next byte to write in buffer
+.free_in_buffer:     resp 1     ; # of byte spaces remaining in buffer
+.cur.put_buffer.simd resq 1     ; current bit accumulation buffer
+.cur.free_bits       resd 1     ; # of bits available in it
+.cur.last_dc_val     resd 4     ; last DC coef for each component
+.cinfo:              resp 1     ; dump_buffer needs access to this
+endstruc
+
+struc c_derived_tbl
+.ehufco:             resd 256   ; code for each symbol
+.ehufsi:             resb 256   ; length of code for each symbol
+; If no code has been allocated for a symbol S, ehufsiS contains 0
+endstruc
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_huff_encode_one_block)
+
+EXTN(jconst_huff_encode_one_block):
+
+jpeg_mask_bits dd 0x0000, 0x0001, 0x0003, 0x0007
+               dd 0x000f, 0x001f, 0x003f, 0x007f
+               dd 0x00ff, 0x01ff, 0x03ff, 0x07ff
+               dd 0x0fff, 0x1fff, 0x3fff, 0x7fff
+
+    alignz      32
+
+times 1 << 14 db 15
+times 1 << 13 db 14
+times 1 << 12 db 13
+times 1 << 11 db 12
+times 1 << 10 db 11
+times 1 <<  9 db 10
+times 1 <<  8 db  9
+times 1 <<  7 db  8
+times 1 <<  6 db  7
+times 1 <<  5 db  6
+times 1 <<  4 db  5
+times 1 <<  3 db  4
+times 1 <<  2 db  3
+times 1 <<  1 db  2
+times 1 <<  0 db  1
+times 1       db  0
+jpeg_nbits_table:
+times 1       db  0
+times 1 <<  0 db  1
+times 1 <<  1 db  2
+times 1 <<  2 db  3
+times 1 <<  3 db  4
+times 1 <<  4 db  5
+times 1 <<  5 db  6
+times 1 <<  6 db  7
+times 1 <<  7 db  8
+times 1 <<  8 db  9
+times 1 <<  9 db 10
+times 1 << 10 db 11
+times 1 << 11 db 12
+times 1 << 12 db 13
+times 1 << 13 db 14
+times 1 << 14 db 15
+times 1 << 15 db 16
+
+    alignz      32
+
+%define NBITS(x)      nbits_base + x
+%define MASK_BITS(x)  NBITS((x) * 4) + (jpeg_mask_bits - jpeg_nbits_table)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+; Shorthand used to describe SIMD operations:
+; wN:  xmmN treated as eight signed 16-bit values
+; wNi:  perform the same operation on all eight signed 16-bit values, i=0..7
+; bN:  xmmN treated as 16 unsigned 8-bit values
+; bNi:  perform the same operation on all 16 unsigned 8-bit values, i=0..15
+; Contents of SIMD registers are shown in memory order.
+
+; Fill the bit buffer to capacity with the leading bits from code, then output
+; the bit buffer and put the remaining bits from code into the bit buffer.
+;
+; Usage:
+; code - contains the bits to shift into the bit buffer (LSB-aligned)
+; %1 - the label to which to jump when the macro completes
+; %2 (optional) - extra instructions to execute after nbits has been set
+;
+; Upon completion, free_bits will be set to the number of remaining bits from
+; code, and put_buffer will contain those remaining bits.  temp and code will
+; be clobbered.
+;
+; This macro encodes any 0xFF bytes as 0xFF 0x00, as does the EMIT_BYTE()
+; macro in jchuff.c.
+
+%macro EMIT_QWORD 1-2
+    add         nbitsb, free_bitsb      ; nbits += free_bits;
+    neg         free_bitsb              ; free_bits = -free_bits;
+    mov         tempd, code             ; temp = code;
+    shl         put_buffer, nbitsb      ; put_buffer <<= nbits;
+    mov         nbitsb, free_bitsb      ; nbits = free_bits;
+    neg         free_bitsb              ; free_bits = -free_bits;
+    shr         tempd, nbitsb           ; temp >>= nbits;
+    or          tempq, put_buffer       ; temp |= put_buffer;
+    movq        xmm0, tempq             ; xmm0.u64 = { temp, 0 };
+    bswap       tempq                   ; temp = htonl(temp);
+    mov         put_buffer, codeq       ; put_buffer = code;
+    pcmpeqb     xmm0, xmm1              ; b0i = (b0i == 0xFF ? 0xFF : 0);
+    %2
+    pmovmskb    code, xmm0              ; code = 0;  code |= ((b0i >> 7) << i);
+    mov         qword buffer, tempq   ; memcpy(buffer, &temp, 8);
+                                        ; (speculative; will be overwritten if
+                                        ; code contains any 0xFF bytes)
+    add         free_bitsb, 64          ; free_bits += 64;
+    add         bufferp, 8              ; buffer += 8;
+    test        code, code              ; if (code == 0)  /* No 0xFF bytes */
+    jz          %1                      ;   return;
+    ; Execute the equivalent of the EMIT_BYTE() macro in jchuff.c for all 8
+    ; bytes in the qword.
+    cmp         tempb, 0xFF             ; Set CF if temp0 < 0xFF
+    mov         byte buffer-7, 0      ; buffer-7 = 0;
+    sbb         bufferp, 6              ; buffer -= (6 + (temp0 < 0xFF ? 1 : 0));
+    mov         byte buffer, temph    ; buffer0 = temp1;
+    cmp         temph, 0xFF             ; Set CF if temp1 < 0xFF
+    mov         byte buffer+1, 0      ; buffer1 = 0;
+    sbb         bufferp, -2             ; buffer -= (-2 + (temp1 < 0xFF ? 1 : 0));
+    shr         tempq, 16               ; temp >>= 16;
+    mov         byte buffer, tempb    ; buffer0 = temp0;
+    cmp         tempb, 0xFF             ; Set CF if temp0 < 0xFF
+    mov         byte buffer+1, 0      ; buffer1 = 0;
+    sbb         bufferp, -2             ; buffer -= (-2 + (temp0 < 0xFF ? 1 : 0));
+    mov         byte buffer, temph    ; buffer0 = temp1;
+    cmp         temph, 0xFF             ; Set CF if temp1 < 0xFF
+    mov         byte buffer+1, 0      ; buffer1 = 0;
+    sbb         bufferp, -2             ; buffer -= (-2 + (temp1 < 0xFF ? 1 : 0));
+    shr         tempq, 16               ; temp >>= 16;
+    mov         byte buffer, tempb    ; buffer0 = temp0;
+    cmp         tempb, 0xFF             ; Set CF if temp0 < 0xFF
+    mov         byte buffer+1, 0      ; buffer1 = 0;
+    sbb         bufferp, -2             ; buffer -= (-2 + (temp0 < 0xFF ? 1 : 0));
+    mov         byte buffer, temph    ; buffer0 = temp1;
+    cmp         temph, 0xFF             ; Set CF if temp1 < 0xFF
+    mov         byte buffer+1, 0      ; buffer1 = 0;
+    sbb         bufferp, -2             ; buffer -= (-2 + (temp1 < 0xFF ? 1 : 0));
+    shr         tempd, 16               ; temp >>= 16;
+    mov         byte buffer, tempb    ; buffer0 = temp0;
+    cmp         tempb, 0xFF             ; Set CF if temp0 < 0xFF
+    mov         byte buffer+1, 0      ; buffer1 = 0;
+    sbb         bufferp, -2             ; buffer -= (-2 + (temp0 < 0xFF ? 1 : 0));
+    mov         byte buffer, temph    ; buffer0 = temp1;
+    cmp         temph, 0xFF             ; Set CF if temp1 < 0xFF
+    mov         byte buffer+1, 0      ; buffer1 = 0;
+    sbb         bufferp, -2             ; buffer -= (-2 + (temp1 < 0xFF ? 1 : 0));
+    jmp         %1                      ; return;
+%endmacro
+
+;
+; Encode a single block's worth of coefficients.
+;
+; GLOBAL(JOCTET *)
+; jsimd_huff_encode_one_block_sse2(working_state *state, JOCTET *buffer,
+;                                  JCOEFPTR block, int last_dc_val,
+;                                  c_derived_tbl *dctbl, c_derived_tbl *actbl)
+;
+; NOTES:
+; When shuffling data, we try to avoid pinsrw as much as possible, since it is
+; slow on many CPUs.  Its reciprocal throughput (issue latency) is 1 even on
+; modern CPUs, so chains of pinsrw instructions (even with different outputs)
+; can limit performance.  pinsrw is a VectorPath instruction on AMD K8 and
+; requires 2 µops (with memory operand) on Intel.  In either case, only one
+; pinsrw instruction can be decoded per cycle (and nothing else if they are
+; back-to-back), so out-of-order execution cannot be used to work around long
+; pinsrw chains (though for Sandy Bridge and later, this may be less of a
+; problem if the code runs from the µop cache.)
+;
+; We use tzcnt instead of bsf without checking for support.  The instruction is
+; executed as bsf on CPUs that don't support tzcnt (encoding is equivalent to
+; rep bsf.)  The destination (first) operand of bsf (and tzcnt on some CPUs) is
+; an input dependency (although the behavior is not formally defined, Intel
+; CPUs usually leave the destination unmodified if the source is zero.)  This
+; can prevent out-of-order execution, so we clear the destination before
+; invoking tzcnt.
+;
+; Initial register allocation
+; rax - buffer
+; rbx - temp
+; rcx - nbits
+; rdx - block --> free_bits
+; rsi - nbits_base
+; rdi - t
+; rbp - code
+; r8  - dctbl --> code_temp
+; r9  - actbl
+; r10 - state
+; r11 - index
+; r12 - put_buffer
+
+%define buffer       rax
+%ifdef WIN64
+%define bufferp      rax
+%else
+%define bufferp      raxp
+%endif
+%define tempq        rbx
+%define tempd        ebx
+%define tempb        bl
+%define temph        bh
+%define nbitsq       rcx
+%define nbits        ecx
+%define nbitsb       cl
+%define block        rdx
+%define nbits_base   rsi
+%define t            rdi
+%define td           edi
+%define codeq        rbp
+%define code         ebp
+%define dctbl        r8
+%define actbl        r9
+%define state        r10
+%define index        r11
+%define indexd       r11d
+%define put_buffer   r12
+%define put_bufferd  r12d
+
+; Step 1: Re-arrange input data according to jpeg_natural_order
+; xx 01 02 03 04 05 06 07      xx 01 08 16 09 02 03 10
+; 08 09 10 11 12 13 14 15      17 24 32 25 18 11 04 05
+; 16 17 18 19 20 21 22 23      12 19 26 33 40 48 41 34
+; 24 25 26 27 28 29 30 31 ==>  27 20 13 06 07 14 21 28
+; 32 33 34 35 36 37 38 39      35 42 49 56 57 50 43 36
+; 40 41 42 43 44 45 46 47      29 22 15 23 30 37 44 51
+; 48 49 50 51 52 53 54 55      58 59 52 45 38 31 39 46
+; 56 57 58 59 60 61 62 63      53 60 61 54 47 55 62 63
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_huff_encode_one_block_sse2)
+
+EXTN(jsimd_huff_encode_one_block_sse2):
+
+%ifdef WIN64
+
+; rcx = working_state *state
+; rdx = JOCTET *buffer
+; r8 = JCOEFPTR block
+; r9 = int last_dc_val
+; rax+48 = c_derived_tbl *dctbl
+; rax+56 = c_derived_tbl *actbl
+
+                                                          ;X: X = code stream
+    mov         buffer, rdx
+    mov         block, r8
+    movups      xmm3, XMMWORD block + 0 * SIZEOF_WORD   ;D: w3 = xx 01 02 03 04 05 06 07
+    push        rbx
+    push        rbp
+    movdqa      xmm0, xmm3                                ;A: w0 = xx 01 02 03 04 05 06 07
+    push        rsi
+    push        rdi
+    push        r12
+    movups      xmm1, XMMWORD block + 8 * SIZEOF_WORD   ;B: w1 = 08 09 10 11 12 13 14 15
+    mov         state, rcx
+    movsx       code, word block                        ;Z:     code = block0;
+    pxor        xmm4, xmm4                                ;A: w4i = 0;
+    sub         code, r9d                                 ;Z:     code -= last_dc_val;
+    mov         dctbl, POINTER rsp+6*8+4*8
+    mov         actbl, POINTER rsp+6*8+5*8
+    punpckldq   xmm0, xmm1                                ;A: w0 = xx 01 08 09 02 03 10 11
+    lea         nbits_base, rel jpeg_nbits_table
+    add         rsp, -DCTSIZE2 * SIZEOF_WORD
+    mov         t, rsp
+
+%else
+
+; rdi = working_state *state
+; rsi = JOCTET *buffer
+; rdx = JCOEFPTR block
+; rcx = int last_dc_val
+; r8 = c_derived_tbl *dctbl
+; r9 = c_derived_tbl *actbl
+
+                                                          ;X: X = code stream
+    movups      xmm3, XMMWORD block + 0 * SIZEOF_WORD   ;D: w3 = xx 01 02 03 04 05 06 07
+    push        rbx
+    push        rbp
+    movdqa      xmm0, xmm3                                ;A: w0 = xx 01 02 03 04 05 06 07
+    push        r12
+    mov         state, rdi
+    mov         buffer, rsi
+    movups      xmm1, XMMWORD block + 8 * SIZEOF_WORD   ;B: w1 = 08 09 10 11 12 13 14 15
+    movsx       codeq, word block                       ;Z:     code = block0;
+    lea         nbits_base, rel jpeg_nbits_table
+    pxor        xmm4, xmm4                                ;A: w4i = 0;
+    sub         codeq, rcx                                ;Z:     code -= last_dc_val;
+    punpckldq   xmm0, xmm1                                ;A: w0 = xx 01 08 09 02 03 10 11
+    lea         t, rsp - DCTSIZE2 * SIZEOF_WORD         ;   use red zone for t_
+
+%endif
+
+    pshuflw     xmm0, xmm0, 11001001b                     ;A: w0 = 01 08 xx 09 02 03 10 11
+    pinsrw      xmm0, word block + 16 * SIZEOF_WORD, 2  ;A: w0 = 01 08 16 09 02 03 10 11
+    punpckhdq   xmm3, xmm1                                ;D: w3 = 04 05 12 13 06 07 14 15
+    punpcklqdq  xmm1, xmm3                                ;B: w1 = 08 09 10 11 04 05 12 13
+    pinsrw      xmm0, word block + 17 * SIZEOF_WORD, 7  ;A: w0 = 01 08 16 09 02 03 10 17
+                                                          ;A:      (Row 0, offset 1)
+    pcmpgtw     xmm4, xmm0                                ;A: w4i = (w0i < 0 ? -1 : 0);
+    paddw       xmm0, xmm4                                ;A: w0i += w4i;
+    movaps      XMMWORD t + 0 * SIZEOF_WORD, xmm0       ;A: ti = w0i;
+
+    movq        xmm2, qword block + 24 * SIZEOF_WORD    ;B: w2 = 24 25 26 27 -- -- -- --
+    pshuflw     xmm2, xmm2, 11011000b                     ;B: w2 = 24 26 25 27 -- -- -- --
+    pslldq      xmm1, 1 * SIZEOF_WORD                     ;B: w1 = -- 08 09 10 11 04 05 12
+    movups      xmm5, XMMWORD block + 48 * SIZEOF_WORD  ;H: w5 = 48 49 50 51 52 53 54 55
+    movsd       xmm1, xmm2                                ;B: w1 = 24 26 25 27 11 04 05 12
+    punpcklqdq  xmm2, xmm5                                ;C: w2 = 24 26 25 27 48 49 50 51
+    pinsrw      xmm1, word block + 32 * SIZEOF_WORD, 1  ;B: w1 = 24 32 25 27 11 04 05 12
+    pxor        xmm4, xmm4                                ;A: w4i = 0;
+    psrldq      xmm3, 2 * SIZEOF_WORD                     ;D: w3 = 12 13 06 07 14 15 -- --
+    pcmpeqw     xmm0, xmm4                                ;A: w0i = (w0i == 0 ? -1 : 0);
+    pinsrw      xmm1, word block + 18 * SIZEOF_WORD, 3  ;B: w1 = 24 32 25 18 11 04 05 12
+                                                          ;        (Row 1, offset 1)
+    pcmpgtw     xmm4, xmm1                                ;B: w4i = (w1i < 0 ? -1 : 0);
+    paddw       xmm1, xmm4                                ;B: w1i += w4i;
+    movaps      XMMWORD t + 8 * SIZEOF_WORD, xmm1       ;B: ti+8 = w1i;
+    pxor        xmm4, xmm4                                ;B: w4i = 0;
+    pcmpeqw     xmm1, xmm4                                ;B: w1i = (w1i == 0 ? -1 : 0);
+
+    packsswb    xmm0, xmm1                                ;AB: b0i = w0i, b0i+8 = w1i
+                                                          ;    w/ signed saturation
+
+    pinsrw      xmm3, word block + 20 * SIZEOF_WORD, 0  ;D: w3 = 20 13 06 07 14 15 -- --
+    pinsrw      xmm3, word block + 21 * SIZEOF_WORD, 5  ;D: w3 = 20 13 06 07 14 21 -- --
+    pinsrw      xmm3, word block + 28 * SIZEOF_WORD, 6  ;D: w3 = 20 13 06 07 14 21 28 --
+    pinsrw      xmm3, word block + 35 * SIZEOF_WORD, 7  ;D: w3 = 20 13 06 07 14 21 28 35
+                                                          ;        (Row 3, offset 1)
+    pcmpgtw     xmm4, xmm3                                ;D: w4i = (w3i < 0 ? -1 : 0);
+    paddw       xmm3, xmm4                                ;D: w3i += w4i;
+    movaps      XMMWORD t + 24 * SIZEOF_WORD, xmm3      ;D: ti+24 = w3i;
+    pxor        xmm4, xmm4                                ;D: w4i = 0;
+    pcmpeqw     xmm3, xmm4                                ;D: w3i = (w3i == 0 ? -1 : 0);
+
+    pinsrw      xmm2, word block + 19 * SIZEOF_WORD, 0  ;C: w2 = 19 26 25 27 48 49 50 51
+    cmp         code, 1 << 31                             ;Z:     Set CF if code < 0x80000000,
+                                                          ;Z:     i.e. if code is positive
+    pinsrw      xmm2, word block + 33 * SIZEOF_WORD, 2  ;C: w2 = 19 26 33 27 48 49 50 51
+    pinsrw      xmm2, word block + 40 * SIZEOF_WORD, 3  ;C: w2 = 19 26 33 40 48 49 50 51
+    adc         code, -1                                  ;Z:     code += -1 + (code >= 0 ? 1 : 0);
+    pinsrw      xmm2, word block + 41 * SIZEOF_WORD, 5  ;C: w2 = 19 26 33 40 48 41 50 51
+    pinsrw      xmm2, word block + 34 * SIZEOF_WORD, 6  ;C: w2 = 19 26 33 40 48 41 34 51
+    movsxd      codeq, code                               ;Z:     sign extend code
+    pinsrw      xmm2, word block + 27 * SIZEOF_WORD, 7  ;C: w2 = 19 26 33 40 48 41 34 27
+                                                          ;        (Row 2, offset 1)
+    pcmpgtw     xmm4, xmm2                                ;C: w4i = (w2i < 0 ? -1 : 0);
+    paddw       xmm2, xmm4                                ;C: w2i += w4i;
+    movaps      XMMWORD t + 16 * SIZEOF_WORD, xmm2      ;C: ti+16 = w2i;
+    pxor        xmm4, xmm4                                ;C: w4i = 0;
+    pcmpeqw     xmm2, xmm4                                ;C: w2i = (w2i == 0 ? -1 : 0);
+
+    packsswb    xmm2, xmm3                                ;CD: b2i = w2i, b2i+8 = w3i
+                                                          ;    w/ signed saturation
+
+    movzx       nbitsq, byte NBITS(codeq)               ;Z:     nbits = JPEG_NBITS(code);
+    movdqa      xmm3, xmm5                                ;H: w3 = 48 49 50 51 52 53 54 55
+    pmovmskb    tempd, xmm2                               ;Z:     temp = 0;  temp |= ((b2i >> 7) << i);
+    pmovmskb    put_bufferd, xmm0                         ;Z:     put_buffer = 0;  put_buffer |= ((b0i >> 7) << i);
+    movups      xmm0, XMMWORD block + 56 * SIZEOF_WORD  ;H: w0 = 56 57 58 59 60 61 62 63
+    punpckhdq   xmm3, xmm0                                ;H: w3 = 52 53 60 61 54 55 62 63
+    shl         tempd, 16                                 ;Z:     temp <<= 16;
+    psrldq      xmm3, 1 * SIZEOF_WORD                     ;H: w3 = 53 60 61 54 55 62 63 --
+    pxor        xmm2, xmm2                                ;H: w2i = 0;
+    or          put_bufferd, tempd                        ;Z:     put_buffer |= temp;
+    pshuflw     xmm3, xmm3, 00111001b                     ;H: w3 = 60 61 54 53 55 62 63 --
+    movq        xmm1, qword block + 44 * SIZEOF_WORD    ;G: w1 = 44 45 46 47 -- -- -- --
+    unpcklps    xmm5, xmm0                                ;E: w5 = 48 49 56 57 50 51 58 59
+    pxor        xmm0, xmm0                                ;H: w0i = 0;
+    pinsrw      xmm3, word block + 47 * SIZEOF_WORD, 3  ;H: w3 = 60 61 54 47 55 62 63 --
+                                                          ;        (Row 7, offset 1)
+    pcmpgtw     xmm2, xmm3                                ;H: w2i = (w3i < 0 ? -1 : 0);
+    paddw       xmm3, xmm2                                ;H: w3i += w2i;
+    movaps      XMMWORD t + 56 * SIZEOF_WORD, xmm3      ;H: ti+56 = w3i;
+    movq        xmm4, qword block + 36 * SIZEOF_WORD    ;G: w4 = 36 37 38 39 -- -- -- --
+    pcmpeqw     xmm3, xmm0                                ;H: w3i = (w3i == 0 ? -1 : 0);
+    punpckldq   xmm4, xmm1                                ;G: w4 = 36 37 44 45 38 39 46 47
+    mov         tempd, dctbl + c_derived_tbl.ehufco + nbitsq * 4
+                                                          ;Z:     temp = dctbl->ehufconbits;
+    movdqa      xmm1, xmm4                                ;F: w1 = 36 37 44 45 38 39 46 47
+    psrldq      xmm4, 1 * SIZEOF_WORD                     ;G: w4 = 37 44 45 38 39 46 47 --
+    shufpd      xmm1, xmm5, 10b                           ;F: w1 = 36 37 44 45 50 51 58 59
+    and         code, dword MASK_BITS(nbitsq)           ;Z:     code &= (1 << nbits) - 1;
+    pshufhw     xmm4, xmm4, 11010011b                     ;G: w4 = 37 44 45 38 -- 39 46 --
+    pslldq      xmm1, 1 * SIZEOF_WORD                     ;F: w1 = -- 36 37 44 45 50 51 58
+    shl         tempq, nbitsb                             ;Z:     temp <<= nbits;
+    pinsrw      xmm4, word block + 59 * SIZEOF_WORD, 0  ;G: w4 = 59 44 45 38 -- 39 46 --
+    pshufd      xmm1, xmm1, 11011000b                     ;F: w1 = -- 36 45 50 37 44 51 58
+    pinsrw      xmm4, word block + 52 * SIZEOF_WORD, 1  ;G: w4 = 59 52 45 38 -- 39 46 --
+    or          code, tempd                               ;Z:     code |= temp;
+    movlps      xmm1, qword block + 20 * SIZEOF_WORD    ;F: w1 = 20 21 22 23 37 44 51 58
+    pinsrw      xmm4, word block + 31 * SIZEOF_WORD, 4  ;G: w4 = 59 52 45 38 31 39 46 --
+    pshuflw     xmm1, xmm1, 01110010b                     ;F: w1 = 22 20 23 21 37 44 51 58
+    pinsrw      xmm4, word block + 53 * SIZEOF_WORD, 7  ;G: w4 = 59 52 45 38 31 39 46 53
+                                                          ;        (Row 6, offset 1)
+    pxor        xmm2, xmm2                                ;G: w2i = 0;
+    pcmpgtw     xmm0, xmm4                                ;G: w0i = (w4i < 0 ? -1 : 0);
+    pinsrw      xmm1, word block + 15 * SIZEOF_WORD, 1  ;F: w1 = 22 15 23 21 37 44 51 58
+    paddw       xmm4, xmm0                                ;G: w4i += w0i;
+    movaps      XMMWORD t + 48 * SIZEOF_WORD, xmm4      ;G: t48+i = w4i;
+    pinsrw      xmm1, word block + 30 * SIZEOF_WORD, 3  ;F: w1 = 22 15 23 30 37 44 51 58
+                                                          ;        (Row 5, offset 1)
+    pcmpeqw     xmm4, xmm2                                ;G: w4i = (w4i == 0 ? -1 : 0);
+    pinsrw      xmm5, word block + 42 * SIZEOF_WORD, 0  ;E: w5 = 42 49 56 57 50 51 58 59
+
+    packsswb    xmm4, xmm3                                ;GH: b4i = w4i, b4i+8 = w3i
+                                                          ;    w/ signed saturation
+
+    pxor        xmm0, xmm0                                ;F: w0i = 0;
+    pinsrw      xmm5, word block + 43 * SIZEOF_WORD, 5  ;E: w5 = 42 49 56 57 50 43 58 59
+    pcmpgtw     xmm2, xmm1                                ;F: w2i = (w1i < 0 ? -1 : 0);
+    pmovmskb    tempd, xmm4                               ;Z:     temp = 0;  temp |= ((b4i >> 7) << i);
+    pinsrw      xmm5, word block + 36 * SIZEOF_WORD, 6  ;E: w5 = 42 49 56 57 50 43 36 59
+    paddw       xmm1, xmm2                                ;F: w1i += w2i;
+    movaps      XMMWORD t + 40 * SIZEOF_WORD, xmm1      ;F: t40+i = w1i;
+    pinsrw      xmm5, word block + 29 * SIZEOF_WORD, 7  ;E: w5 = 42 49 56 57 50 43 36 29
+                                                          ;        (Row 4, offset 1)
+%undef block
+%define free_bitsq  rdx
+%define free_bitsd  edx
+%define free_bitsb  dl
+    pcmpeqw     xmm1, xmm0                                ;F: w1i = (w1i == 0 ? -1 : 0);
+    shl         tempq, 48                                 ;Z:     temp <<= 48;
+    pxor        xmm2, xmm2                                ;E: w2i = 0;
+    pcmpgtw     xmm0, xmm5                                ;E: w0i = (w5i < 0 ? -1 : 0);
+    paddw       xmm5, xmm0                                ;E: w5i += w0i;
+    or          tempq, put_buffer                         ;Z:     temp |= put_buffer;
+    movaps      XMMWORD t + 32 * SIZEOF_WORD, xmm5      ;E: t32+i = w5i;
+    lea         t, dword t - 2                          ;Z:     t = &t-1;
+    pcmpeqw     xmm5, xmm2                                ;E: w5i = (w5i == 0 ? -1 : 0);
+
+    packsswb    xmm5, xmm1                                ;EF: b5i = w5i, b5i+8 = w1i
+                                                          ;    w/ signed saturation
+
+    add         nbitsb, byte dctbl + c_derived_tbl.ehufsi + nbitsq
+                                                          ;Z:     nbits += dctbl->ehufsinbits;
+%undef dctbl
+%define code_temp  r8d
+    pmovmskb    indexd, xmm5                              ;Z:     index = 0;  index |= ((b5i >> 7) << i);
+    mov         free_bitsd, state+working_state.cur.free_bits
+                                                          ;Z:     free_bits = state->cur.free_bits;
+    pcmpeqw     xmm1, xmm1                                ;Z:     b1i = 0xFF;
+    shl         index, 32                                 ;Z:     index <<= 32;
+    mov         put_buffer, state+working_state.cur.put_buffer.simd
+                                                          ;Z:     put_buffer = state->cur.put_buffer.simd;
+    or          index, tempq                              ;Z:     index |= temp;
+    not         index                                     ;Z:     index = ~index;
+    sub         free_bitsb, nbitsb                        ;Z:     if ((free_bits -= nbits) >= 0)
+    jnl         .ENTRY_SKIP_EMIT_CODE                     ;Z:       goto .ENTRY_SKIP_EMIT_CODE;
+    align       16
+.EMIT_CODE:                                               ;Z:     .EMIT_CODE:
+    EMIT_QWORD  .BLOOP_COND                               ;Z:     insert code, flush buffer, goto .BLOOP_COND
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align       16
+.BRLOOP:                                                  ; do {
+    lea         code_temp, nbitsq - 16                  ;   code_temp = nbits - 16;
+    movzx       nbits, byte actbl + c_derived_tbl.ehufsi + 0xf0
+                                                          ;   nbits = actbl->ehufsi0xf0;
+    mov         code, actbl + c_derived_tbl.ehufco + 0xf0 * 4
+                                                          ;   code = actbl->ehufco0xf0;
+    sub         free_bitsb, nbitsb                        ;   if ((free_bits -= nbits) <= 0)
+    jle         .EMIT_BRLOOP_CODE                         ;     goto .EMIT_BRLOOP_CODE;
+    shl         put_buffer, nbitsb                        ;   put_buffer <<= nbits;
+    mov         nbits, code_temp                          ;   nbits = code_temp;
+    or          put_buffer, codeq                         ;   put_buffer |= code;
+    cmp         nbits, 16                                 ;   if (nbits <= 16)
+    jle         .ERLOOP                                   ;     break;
+    jmp         .BRLOOP                                   ; } while (1);
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align       16
+    times 5     nop
+.ENTRY_SKIP_EMIT_CODE:                                    ; .ENTRY_SKIP_EMIT_CODE:
+    shl         put_buffer, nbitsb                        ; put_buffer <<= nbits;
+    or          put_buffer, codeq                         ; put_buffer |= code;
+.BLOOP_COND:                                              ; .BLOOP_COND:
+    test        index, index                              ; if (index != 0)
+    jz          .ELOOP                                    ; {
+.BLOOP:                                                   ;   do {
+    xor         nbits, nbits                              ;     nbits = 0;  /* kill tzcnt input dependency */
+    tzcnt       nbitsq, index                             ;     nbits = # of trailing 0 bits in index
+    inc         nbits                                     ;     ++nbits;
+    lea         t, t + nbitsq * 2                       ;     t = &tnbits;
+    shr         index, nbitsb                             ;     index >>= nbits;
+.EMIT_BRLOOP_CODE_END:                                    ; .EMIT_BRLOOP_CODE_END:
+    cmp         nbits, 16                                 ;     if (nbits > 16)
+    jg          .BRLOOP                                   ;       goto .BRLOOP;
+.ERLOOP:                                                  ; .ERLOOP:
+    movsx       codeq, word t                           ;     code = *t;
+    lea         tempd, nbitsq * 2                       ;     temp = nbits * 2;
+    movzx       nbits, byte NBITS(codeq)                ;     nbits = JPEG_NBITS(code);
+    lea         tempd, nbitsq + tempq * 8               ;     temp = temp * 8 + nbits;
+    mov         code_temp, actbl + c_derived_tbl.ehufco + (tempq - 16) * 4
+                                                          ;     code_temp = actbl->ehufcotemp-16;
+    shl         code_temp, nbitsb                         ;     code_temp <<= nbits;
+    and         code, dword MASK_BITS(nbitsq)           ;     code &= (1 << nbits) - 1;
+    add         nbitsb, actbl + c_derived_tbl.ehufsi + (tempq - 16)
+                                                          ;     free_bits -= actbl->ehufsitemp-16;
+    or          code, code_temp                           ;     code |= code_temp;
+    sub         free_bitsb, nbitsb                        ;     if ((free_bits -= nbits) <= 0)
+    jle         .EMIT_CODE                                ;       goto .EMIT_CODE;
+    shl         put_buffer, nbitsb                        ;     put_buffer <<= nbits;
+    or          put_buffer, codeq                         ;     put_buffer |= code;
+    test        index, index
+    jnz         .BLOOP                                    ;   } while (index != 0);
+.ELOOP:                                                   ; }  /* index != 0 */
+    sub         td, esp                                   ; t -= (WIN64: &t_0, UNIX: &t_64);
+%ifdef WIN64
+    cmp         td, (DCTSIZE2 - 2) * SIZEOF_WORD          ; if (t != 62)
+%else
+    cmp         td, -2 * SIZEOF_WORD                      ; if (t != -2)
+%endif
+    je          .EFN                                      ; {
+    movzx       nbits, byte actbl + c_derived_tbl.ehufsi + 0
+                                                          ;   nbits = actbl->ehufsi0;
+    mov         code, actbl + c_derived_tbl.ehufco + 0  ;   code = actbl->ehufco0;
+    sub         free_bitsb, nbitsb                        ;   if ((free_bits -= nbits) <= 0)
+    jg          .EFN_SKIP_EMIT_CODE                       ;   {
+    EMIT_QWORD  .EFN                                      ;     insert code, flush buffer
+    align       16
+.EFN_SKIP_EMIT_CODE:                                      ;   } else {
+    shl         put_buffer, nbitsb                        ;     put_buffer <<= nbits;
+    or          put_buffer, codeq                         ;     put_buffer |= code;
+.EFN:                                                     ; } }
+    mov         state + working_state.cur.put_buffer.simd, put_buffer
+                                                          ; state->cur.put_buffer.simd = put_buffer;
+    mov         byte state + working_state.cur.free_bits, free_bitsb
+                                                          ; state->cur.free_bits = free_bits;
+%ifdef WIN64
+    sub         rsp, -DCTSIZE2 * SIZEOF_WORD
+    pop         r12
+    pop         rdi
+    pop         rsi
+    pop         rbp
+    pop         rbx
+%else
+    pop         r12
+    pop         rbp
+    pop         rbx
+%endif
+    ret
+
+; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    align       16
+.EMIT_BRLOOP_CODE:
+    EMIT_QWORD  .EMIT_BRLOOP_CODE_END, { mov nbits, code_temp }
+                                                          ; insert code, flush buffer,
+                                                          ; nbits = code_temp, goto .EMIT_BRLOOP_CODE_END
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jcphuff-sse2.asm Added

@@ -0,0 +1,639 @@
+;
+; jcphuff-sse2.asm - prepare data for progressive Huffman encoding
+; (64-bit SSE2)
+;
+; Copyright (C) 2016, 2018, Matthieu Darbois
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains an SSE2 implementation of data preparation for progressive
+; Huffman encoding.  See jcphuff.c for more details.
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+; --------------------------------------------------------------------------
+; Macros to load data for jsimd_encode_mcu_AC_first_prepare_sse2() and
+; jsimd_encode_mcu_AC_refine_prepare_sse2()
+
+%macro LOAD16 0
+    pxor        N0, N0
+    pxor        N1, N1
+
+    mov         T0d, INT LUT +  0*SIZEOF_INT
+    mov         T1d, INT LUT +  8*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 0
+    pinsrw      X1, word BLOCK + T1 * 2, 0
+
+    mov         T0d, INT LUT +  1*SIZEOF_INT
+    mov         T1d, INT LUT +  9*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 1
+    pinsrw      X1, word BLOCK + T1 * 2, 1
+
+    mov         T0d, INT LUT +  2*SIZEOF_INT
+    mov         T1d, INT LUT + 10*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 2
+    pinsrw      X1, word BLOCK + T1 * 2, 2
+
+    mov         T0d, INT LUT +  3*SIZEOF_INT
+    mov         T1d, INT LUT + 11*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 3
+    pinsrw      X1, word BLOCK + T1 * 2, 3
+
+    mov         T0d, INT LUT +  4*SIZEOF_INT
+    mov         T1d, INT LUT + 12*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 4
+    pinsrw      X1, word BLOCK + T1 * 2, 4
+
+    mov         T0d, INT LUT +  5*SIZEOF_INT
+    mov         T1d, INT LUT + 13*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 5
+    pinsrw      X1, word BLOCK + T1 * 2, 5
+
+    mov         T0d, INT LUT +  6*SIZEOF_INT
+    mov         T1d, INT LUT + 14*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 6
+    pinsrw      X1, word BLOCK + T1 * 2, 6
+
+    mov         T0d, INT LUT +  7*SIZEOF_INT
+    mov         T1d, INT LUT + 15*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 7
+    pinsrw      X1, word BLOCK + T1 * 2, 7
+%endmacro
+
+%macro LOAD15 0
+    pxor        N0, N0
+    pxor        N1, N1
+    pxor        X1, X1
+
+    mov         T0d, INT LUT +  0*SIZEOF_INT
+    mov         T1d, INT LUT +  8*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 0
+    pinsrw      X1, word BLOCK + T1 * 2, 0
+
+    mov         T0d, INT LUT +  1*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 1
+
+    mov         T0d, INT LUT +  2*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 2
+
+    mov         T0d, INT LUT +  3*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 3
+
+    mov         T0d, INT LUT +  4*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 4
+
+    mov         T0d, INT LUT +  5*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 5
+
+    mov         T0d, INT LUT +  6*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 6
+
+    mov         T0d, INT LUT +  7*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 7
+
+    cmp         LENEND, 2
+    jl          %%.ELOAD15
+    mov         T1d, INT LUT +  9*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 1
+
+    cmp         LENEND, 3
+    jl          %%.ELOAD15
+    mov         T1d, INT LUT + 10*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 2
+
+    cmp         LENEND, 4
+    jl          %%.ELOAD15
+    mov         T1d, INT LUT + 11*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 3
+
+    cmp         LENEND, 5
+    jl          %%.ELOAD15
+    mov         T1d, INT LUT + 12*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 4
+
+    cmp         LENEND, 6
+    jl          %%.ELOAD15
+    mov         T1d, INT LUT + 13*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 5
+
+    cmp         LENEND, 7
+    jl          %%.ELOAD15
+    mov         T1d, INT LUT + 14*SIZEOF_INT
+    pinsrw      X1, word BLOCK + T1 * 2, 6
+%%.ELOAD15:
+%endmacro
+
+%macro LOAD8 0
+    pxor        N0, N0
+
+    mov         T0d, INT LUT +  0*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 0
+
+    mov         T0d, INT LUT +  1*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 1
+
+    mov         T0d, INT LUT +  2*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 2
+
+    mov         T0d, INT LUT +  3*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 3
+
+    mov         T0d, INT LUT +  4*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 4
+
+    mov         T0d, INT LUT +  5*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 5
+
+    mov         T0d, INT LUT +  6*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 6
+
+    mov         T0d, INT LUT +  7*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T0 * 2, 7
+%endmacro
+
+%macro LOAD7 0
+    pxor        N0, N0
+    pxor        X0, X0
+
+    mov         T1d, INT LUT +  0*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 0
+
+    cmp         LENEND, 2
+    jl          %%.ELOAD7
+    mov         T1d, INT LUT +  1*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 1
+
+    cmp         LENEND, 3
+    jl          %%.ELOAD7
+    mov         T1d, INT LUT +  2*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 2
+
+    cmp         LENEND, 4
+    jl          %%.ELOAD7
+    mov         T1d, INT LUT +  3*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 3
+
+    cmp         LENEND, 5
+    jl          %%.ELOAD7
+    mov         T1d, INT LUT +  4*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 4
+
+    cmp         LENEND, 6
+    jl          %%.ELOAD7
+    mov         T1d, INT LUT +  5*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 5
+
+    cmp         LENEND, 7
+    jl          %%.ELOAD7
+    mov         T1d, INT LUT +  6*SIZEOF_INT
+    pinsrw      X0, word BLOCK + T1 * 2, 6
+%%.ELOAD7:
+%endmacro
+
+%macro REDUCE0 0
+    movdqa      xmm0, XMMWORD VALUES + ( 0*2)
+    movdqa      xmm1, XMMWORD VALUES + ( 8*2)
+    movdqa      xmm2, XMMWORD VALUES + (16*2)
+    movdqa      xmm3, XMMWORD VALUES + (24*2)
+    movdqa      xmm4, XMMWORD VALUES + (32*2)
+    movdqa      xmm5, XMMWORD VALUES + (40*2)
+    movdqa      xmm6, XMMWORD VALUES + (48*2)
+    movdqa      xmm7, XMMWORD VALUES + (56*2)
+
+    pcmpeqw     xmm0, ZERO
+    pcmpeqw     xmm1, ZERO
+    pcmpeqw     xmm2, ZERO
+    pcmpeqw     xmm3, ZERO
+    pcmpeqw     xmm4, ZERO
+    pcmpeqw     xmm5, ZERO
+    pcmpeqw     xmm6, ZERO
+    pcmpeqw     xmm7, ZERO
+
+    packsswb    xmm0, xmm1
+    packsswb    xmm2, xmm3
+    packsswb    xmm4, xmm5
+    packsswb    xmm6, xmm7
+
+    pmovmskb    eax, xmm0
+    pmovmskb    ecx, xmm2
+    pmovmskb    edx, xmm4
+    pmovmskb    esi, xmm6
+
+    shl         rcx, 16
+    shl         rdx, 32
+    shl         rsi, 48
+
+    or          rax, rcx
+    or          rdx, rsi
+    or          rax, rdx
+
+    not         rax
+
+    mov         MMWORD r15, rax
+%endmacro
+
+;
+; Prepare data for jsimd_encode_mcu_AC_first().
+;
+; GLOBAL(void)
+; jsimd_encode_mcu_AC_first_prepare_sse2(const JCOEF *block,
+;                                        const int *jpeg_natural_order_start,
+;                                        int Sl, int Al, JCOEF *values,
+;                                        size_t *zerobits)
+;
+; r10 = const JCOEF *block
+; r11 = const int *jpeg_natural_order_start
+; r12 = int Sl
+; r13 = int Al
+; r14 = JCOEF *values
+; r15 = size_t *zerobits
+
+%define ZERO    xmm9
+%define X0      xmm0
+%define X1      xmm1
+%define N0      xmm2
+%define N1      xmm3
+%define AL      xmm4
+%define K       eax
+%define LUT     r11
+%define T0      rcx
+%define T0d     ecx
+%define T1      rdx
+%define T1d     edx
+%define BLOCK   r10
+%define VALUES  r14
+%define LEN     r12d
+%define LENEND  r13d
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_encode_mcu_AC_first_prepare_sse2)
+
+EXTN(jsimd_encode_mcu_AC_first_prepare_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, rbp - 16
+    collect_args 6
+
+    movdqa      XMMWORD rbp - 16, ZERO
+
+    movd        AL, r13d
+    pxor        ZERO, ZERO
+    mov         K, LEN
+    mov         LENEND, LEN
+    and         K, -16
+    and         LENEND, 7
+    shr         K, 4
+    jz          .ELOOP16
+.BLOOP16:
+    LOAD16
+    pcmpgtw     N0, X0
+    pcmpgtw     N1, X1
+    paddw       X0, N0
+    paddw       X1, N1
+    pxor        X0, N0
+    pxor        X1, N1
+    psrlw       X0, AL
+    psrlw       X1, AL
+    pxor        N0, X0
+    pxor        N1, X1
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (8) * 2, X1
+    movdqa      XMMWORD VALUES + (0 + DCTSIZE2) * 2, N0
+    movdqa      XMMWORD VALUES + (8 + DCTSIZE2) * 2, N1
+    add         VALUES, 16*2
+    add         LUT, 16*SIZEOF_INT
+    dec         K
+    jnz         .BLOOP16
+    test        LEN, 15
+    je          .PADDING
+.ELOOP16:
+    test        LEN, 8
+    jz          .TRY7
+    test        LEN, 7
+    jz          .TRY8
+
+    LOAD15
+    pcmpgtw     N0, X0
+    pcmpgtw     N1, X1
+    paddw       X0, N0
+    paddw       X1, N1
+    pxor        X0, N0
+    pxor        X1, N1
+    psrlw       X0, AL
+    psrlw       X1, AL
+    pxor        N0, X0
+    pxor        N1, X1
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (8) * 2, X1
+    movdqa      XMMWORD VALUES + (0 + DCTSIZE2) * 2, N0
+    movdqa      XMMWORD VALUES + (8 + DCTSIZE2) * 2, N1
+    add         VALUES, 16*2
+    jmp         .PADDING
+.TRY8:
+    LOAD8
+    pcmpgtw     N0, X0
+    paddw       X0, N0
+    pxor        X0, N0
+    psrlw       X0, AL
+    pxor        N0, X0
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (0 + DCTSIZE2) * 2, N0
+    add         VALUES, 8*2
+    jmp         .PADDING
+.TRY7:
+    LOAD7
+    pcmpgtw     N0, X0
+    paddw       X0, N0
+    pxor        X0, N0
+    psrlw       X0, AL
+    pxor        N0, X0
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (0 + DCTSIZE2) * 2, N0
+    add         VALUES, 8*2
+.PADDING:
+    mov         K, LEN
+    add         K, 7
+    and         K, -8
+    shr         K, 3
+    sub         K, DCTSIZE2/8
+    jz          .EPADDING
+    align       16
+.ZEROLOOP:
+    movdqa      XMMWORD VALUES + 0, ZERO
+    add         VALUES, 8*2
+    inc         K
+    jnz         .ZEROLOOP
+.EPADDING:
+    sub         VALUES, DCTSIZE2*2
+
+    REDUCE0
+
+    movdqa      ZERO, XMMWORD rbp - 16
+    uncollect_args 6
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+%undef ZERO
+%undef X0
+%undef X1
+%undef N0
+%undef N1
+%undef AL
+%undef K
+%undef LUT
+%undef T0
+%undef T0d
+%undef T1
+%undef T1d
+%undef BLOCK
+%undef VALUES
+%undef LEN
+%undef LENEND
+
+;
+; Prepare data for jsimd_encode_mcu_AC_refine().
+;
+; GLOBAL(int)
+; jsimd_encode_mcu_AC_refine_prepare_sse2(const JCOEF *block,
+;                                         const int *jpeg_natural_order_start,
+;                                         int Sl, int Al, JCOEF *absvalues,
+;                                         size_t *bits)
+;
+; r10 = const JCOEF *block
+; r11 = const int *jpeg_natural_order_start
+; r12 = int Sl
+; r13 = int Al
+; r14 = JCOEF *values
+; r15 = size_t *bits
+
+%define ZERO    xmm9
+%define ONE     xmm5
+%define X0      xmm0
+%define X1      xmm1
+%define N0      xmm2
+%define N1      xmm3
+%define AL      xmm4
+%define K       eax
+%define KK      r9d
+%define EOB     r8d
+%define SIGN    rdi
+%define LUT     r11
+%define T0      rcx
+%define T0d     ecx
+%define T1      rdx
+%define T1d     edx
+%define BLOCK   r10
+%define VALUES  r14
+%define LEN     r12d
+%define LENEND  r13d
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_encode_mcu_AC_refine_prepare_sse2)
+
+EXTN(jsimd_encode_mcu_AC_refine_prepare_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, rbp - 16
+    collect_args 6
+
+    movdqa      XMMWORD rbp - 16, ZERO
+
+    xor         SIGN, SIGN
+    xor         EOB, EOB
+    xor         KK, KK
+    movd        AL, r13d
+    pxor        ZERO, ZERO
+    pcmpeqw     ONE, ONE
+    psrlw       ONE, 15
+    mov         K, LEN
+    mov         LENEND, LEN
+    and         K, -16
+    and         LENEND, 7
+    shr         K, 4
+    jz          .ELOOPR16
+.BLOOPR16:
+    LOAD16
+    pcmpgtw     N0, X0
+    pcmpgtw     N1, X1
+    paddw       X0, N0
+    paddw       X1, N1
+    pxor        X0, N0
+    pxor        X1, N1
+    psrlw       X0, AL
+    psrlw       X1, AL
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (8) * 2, X1
+    pcmpeqw     X0, ONE
+    pcmpeqw     X1, ONE
+    packsswb    N0, N1
+    packsswb    X0, X1
+    pmovmskb    T0d, N0                 ; lsignbits.val16uk>>4 = _mm_movemask_epi8(neg);
+    pmovmskb    T1d, X0                 ; idx = _mm_movemask_epi8(x1);
+    shr         SIGN, 16                ; make room for sizebits
+    shl         T0, 48
+    or          SIGN, T0
+    bsr         T1d, T1d                ;  idx = 16 - (__builtin_clz(idx)>>1);
+    jz          .CONTINUER16            ; if (idx) {
+    mov         EOB, KK
+    add         EOB, T1d                ; EOB = k + idx;
+.CONTINUER16:
+    add         VALUES, 16*2
+    add         LUT, 16*SIZEOF_INT
+    add         KK, 16
+    dec         K
+    jnz         .BLOOPR16
+    test        LEN, 15
+    je          .PADDINGR
+.ELOOPR16:
+    test        LEN, 8
+    jz          .TRYR7
+    test        LEN, 7
+    jz          .TRYR8
+
+    LOAD15
+    pcmpgtw     N0, X0
+    pcmpgtw     N1, X1
+    paddw       X0, N0
+    paddw       X1, N1
+    pxor        X0, N0
+    pxor        X1, N1
+    psrlw       X0, AL
+    psrlw       X1, AL
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    movdqa      XMMWORD VALUES + (8) * 2, X1
+    pcmpeqw     X0, ONE
+    pcmpeqw     X1, ONE
+    packsswb    N0, N1
+    packsswb    X0, X1
+    pmovmskb    T0d, N0                 ; lsignbits.val16uk>>4 = _mm_movemask_epi8(neg);
+    pmovmskb    T1d, X0                 ; idx = _mm_movemask_epi8(x1);
+    shr         SIGN, 16                ; make room for sizebits
+    shl         T0, 48
+    or          SIGN, T0
+    bsr         T1d, T1d                ;  idx = 16 - (__builtin_clz(idx)>>1);
+    jz          .CONTINUER15            ; if (idx) {
+    mov         EOB, KK
+    add         EOB, T1d                ; EOB = k + idx;
+.CONTINUER15:
+    add         VALUES, 16*2
+    jmp         .PADDINGR
+.TRYR8:
+    LOAD8
+
+    pcmpgtw     N0, X0
+    paddw       X0, N0
+    pxor        X0, N0
+    psrlw       X0, AL
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    pcmpeqw     X0, ONE
+    packsswb    N0, ZERO
+    packsswb    X0, ZERO
+    pmovmskb    T0d, N0                 ; lsignbits.val16uk>>4 = _mm_movemask_epi8(neg);
+    pmovmskb    T1d, X0                 ; idx = _mm_movemask_epi8(x1);
+    shr         SIGN, 8                 ; make room for sizebits
+    shl         T0, 56
+    or          SIGN, T0
+    bsr         T1d, T1d                ;  idx = 16 - (__builtin_clz(idx)>>1);
+    jz          .CONTINUER8             ; if (idx) {
+    mov         EOB, KK
+    add         EOB, T1d                ; EOB = k + idx;
+.CONTINUER8:
+    add         VALUES, 8*2
+    jmp         .PADDINGR
+.TRYR7:
+    LOAD7
+
+    pcmpgtw     N0, X0
+    paddw       X0, N0
+    pxor        X0, N0
+    psrlw       X0, AL
+    movdqa      XMMWORD VALUES + (0) * 2, X0
+    pcmpeqw     X0, ONE
+    packsswb    N0, ZERO
+    packsswb    X0, ZERO
+    pmovmskb    T0d, N0                 ; lsignbits.val16uk>>4 = _mm_movemask_epi8(neg);
+    pmovmskb    T1d, X0                 ; idx = _mm_movemask_epi8(x1);
+    shr         SIGN, 8                 ; make room for sizebits
+    shl         T0, 56
+    or          SIGN, T0
+    bsr         T1d, T1d                ; idx = 16 - (__builtin_clz(idx)>>1);
+    jz          .CONTINUER7             ; if (idx) {
+    mov         EOB, KK
+    add         EOB, T1d                ; EOB = k + idx;
+.CONTINUER7:
+    add         VALUES, 8*2
+.PADDINGR:
+    mov         K, LEN
+    add         K, 7
+    and         K, -8
+    shr         K, 3
+    sub         K, DCTSIZE2/8
+    jz          .EPADDINGR
+    align       16
+.ZEROLOOPR:
+    movdqa      XMMWORD VALUES + 0, ZERO
+    shr         SIGN, 8
+    add         VALUES, 8*2
+    inc         K
+    jnz         .ZEROLOOPR
+.EPADDINGR:
+    not         SIGN
+    sub         VALUES, DCTSIZE2*2
+    mov         MMWORD r15+SIZEOF_MMWORD, SIGN
+
+    REDUCE0
+
+    mov         eax, EOB
+    movdqa      ZERO, XMMWORD rbp - 16
+    uncollect_args 6
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+%undef ZERO
+%undef ONE
+%undef X0
+%undef X1
+%undef N0
+%undef N1
+%undef AL
+%undef K
+%undef KK
+%undef EOB
+%undef SIGN
+%undef LUT
+%undef T0
+%undef T0d
+%undef T1
+%undef T1d
+%undef BLOCK
+%undef VALUES
+%undef LEN
+%undef LENEND
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jcsample-avx2.asm Added

@@ -0,0 +1,367 @@
+;
+; jcsample.asm - downsampling (64-bit AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_avx2(JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor,
+;                            JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                            JSAMPARRAY output_data);
+;
+
+; r10d = JDIMENSION image_width
+; r11 = int max_v_samp_factor
+; r12d = JDIMENSION v_samp_factor
+; r13d = JDIMENSION width_in_blocks
+; r14 = JSAMPARRAY input_data
+; r15 = JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_downsample_avx2)
+
+EXTN(jsimd_h2v1_downsample_avx2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 6
+
+    mov         ecx, r13d
+    shl         rcx, 3                  ; imul rcx,DCTSIZE (rcx = output_cols)
+    jz          near .return
+
+    mov         edx, r10d
+
+    ; -- expand_right_edge
+
+    push        rcx
+    shl         rcx, 1                  ; output_cols * 2
+    sub         rcx, rdx
+    jle         short .expand_end
+
+    mov         rax, r11
+    test        rax, rax
+    jle         short .expand_end
+
+    cld
+    mov         rsi, r14                ; input_data
+.expandloop:
+    push        rax
+    push        rcx
+
+    mov         rdip, JSAMPROW rsi
+    add         rdi, rdx
+    mov         al, JSAMPLE rdi-1
+
+    rep stosb
+
+    pop         rcx
+    pop         rax
+
+    add         rsi, byte SIZEOF_JSAMPROW
+    dec         rax
+    jg          short .expandloop
+
+.expand_end:
+    pop         rcx                     ; output_cols
+
+    ; -- h2v1_downsample
+
+    mov         eax, r12d               ; rowctr
+    test        eax, eax
+    jle         near .return
+
+    mov         rdx, 0x00010000         ; bias pattern
+    vmovd       xmm7, edx
+    vpshufd     xmm7, xmm7, 0x00        ; xmm7={0, 1, 0, 1, 0, 1, 0, 1}
+    vperm2i128  ymm7, ymm7, ymm7, 0     ; ymm7={xmm7, xmm7}
+    vpcmpeqw    ymm6, ymm6, ymm6
+    vpsrlw      ymm6, ymm6, BYTE_BIT    ; ymm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         rsi, r14                ; input_data
+    mov         rdi, r15                ; output_data
+.rowloop:
+    push        rcx
+    push        rdi
+    push        rsi
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr
+
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jae         short .columnloop
+
+.columnloop_r24:
+    ; rcx can possibly be 8, 16, 24
+    cmp         rcx, 24
+    jne         .columnloop_r16
+    vmovdqu     ymm0, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     xmm1, XMMWORD rsi+1*SIZEOF_YMMWORD
+    mov         rcx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop_r16:
+    cmp         rcx, 16
+    jne         .columnloop_r8
+    vmovdqu     ymm0, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vpxor       ymm1, ymm1, ymm1
+    mov         rcx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop_r8:
+    vmovdqu     xmm0, XMMWORDrsi+0*SIZEOF_YMMWORD
+    vpxor       ymm1, ymm1, ymm1
+    mov         rcx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop:
+    vmovdqu     ymm0, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymm1, YMMWORD rsi+1*SIZEOF_YMMWORD
+
+.downsample:
+    vpsrlw      ymm2, ymm0, BYTE_BIT
+    vpand       ymm0, ymm0, ymm6
+    vpsrlw      ymm3, ymm1, BYTE_BIT
+    vpand       ymm1, ymm1, ymm6
+
+    vpaddw      ymm0, ymm0, ymm2
+    vpaddw      ymm1, ymm1, ymm3
+    vpaddw      ymm0, ymm0, ymm7
+    vpaddw      ymm1, ymm1, ymm7
+    vpsrlw      ymm0, ymm0, 1
+    vpsrlw      ymm1, ymm1, 1
+
+    vpackuswb   ymm0, ymm0, ymm1
+    vpermq      ymm0, ymm0, 0xd8
+
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymm0
+
+    sub         rcx, byte SIZEOF_YMMWORD    ; outcol
+    add         rsi, byte 2*SIZEOF_YMMWORD  ; inptr
+    add         rdi, byte 1*SIZEOF_YMMWORD  ; outptr
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jae         short .columnloop
+    test        rcx, rcx
+    jnz         near .columnloop_r24
+
+    pop         rsi
+    pop         rdi
+    pop         rcx
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         rax                        ; rowctr
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    uncollect_args 6
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_avx2(JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor,
+;                            JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                            JSAMPARRAY output_data);
+;
+
+; r10d = JDIMENSION image_width
+; r11 = int max_v_samp_factor
+; r12d = JDIMENSION v_samp_factor
+; r13d = JDIMENSION width_in_blocks
+; r14 = JSAMPARRAY input_data
+; r15 = JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_downsample_avx2)
+
+EXTN(jsimd_h2v2_downsample_avx2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 6
+
+    mov         ecx, r13d
+    shl         rcx, 3                  ; imul rcx,DCTSIZE (rcx = output_cols)
+    jz          near .return
+
+    mov         edx, r10d
+
+    ; -- expand_right_edge
+
+    push        rcx
+    shl         rcx, 1                  ; output_cols * 2
+    sub         rcx, rdx
+    jle         short .expand_end
+
+    mov         rax, r11
+    test        rax, rax
+    jle         short .expand_end
+
+    cld
+    mov         rsi, r14                ; input_data
+.expandloop:
+    push        rax
+    push        rcx
+
+    mov         rdip, JSAMPROW rsi
+    add         rdi, rdx
+    mov         al, JSAMPLE rdi-1
+
+    rep stosb
+
+    pop         rcx
+    pop         rax
+
+    add         rsi, byte SIZEOF_JSAMPROW
+    dec         rax
+    jg          short .expandloop
+
+.expand_end:
+    pop         rcx                     ; output_cols
+
+    ; -- h2v2_downsample
+
+    mov         eax, r12d               ; rowctr
+    test        rax, rax
+    jle         near .return
+
+    mov         rdx, 0x00020001         ; bias pattern
+    vmovd       xmm7, edx
+    vpcmpeqw    ymm6, ymm6, ymm6
+    vpshufd     xmm7, xmm7, 0x00        ; ymm7={1, 2, 1, 2, 1, 2, 1, 2}
+    vperm2i128  ymm7, ymm7, ymm7, 0
+    vpsrlw      ymm6, ymm6, BYTE_BIT    ; ymm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         rsi, r14                ; input_data
+    mov         rdi, r15                ; output_data
+.rowloop:
+    push        rcx
+    push        rdi
+    push        rsi
+
+    mov         rdxp, JSAMPROW rsi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         rsip, JSAMPROW rsi+1*SIZEOF_JSAMPROW  ; inptr1
+    mov         rdip, JSAMPROW rdi                    ; outptr
+
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jae         short .columnloop
+
+.columnloop_r24:
+    cmp         rcx, 24
+    jne         .columnloop_r16
+    vmovdqu     ymm0, YMMWORD rdx+0*SIZEOF_YMMWORD
+    vmovdqu     ymm1, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     xmm2, XMMWORD rdx+1*SIZEOF_YMMWORD
+    vmovdqu     xmm3, XMMWORD rsi+1*SIZEOF_YMMWORD
+    mov         rcx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop_r16:
+    cmp         rcx, 16
+    jne         .columnloop_r8
+    vmovdqu     ymm0, YMMWORD rdx+0*SIZEOF_YMMWORD
+    vmovdqu     ymm1, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vpxor       ymm2, ymm2, ymm2
+    vpxor       ymm3, ymm3, ymm3
+    mov         rcx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop_r8:
+    vmovdqu     xmm0, XMMWORD rdx+0*SIZEOF_XMMWORD
+    vmovdqu     xmm1, XMMWORD rsi+0*SIZEOF_XMMWORD
+    vpxor       ymm2, ymm2, ymm2
+    vpxor       ymm3, ymm3, ymm3
+    mov         rcx, SIZEOF_YMMWORD
+    jmp         short .downsample
+
+.columnloop:
+    vmovdqu     ymm0, YMMWORD rdx+0*SIZEOF_YMMWORD
+    vmovdqu     ymm1, YMMWORD rsi+0*SIZEOF_YMMWORD
+    vmovdqu     ymm2, YMMWORD rdx+1*SIZEOF_YMMWORD
+    vmovdqu     ymm3, YMMWORD rsi+1*SIZEOF_YMMWORD
+
+.downsample:
+    vpand       ymm4, ymm0, ymm6
+    vpsrlw      ymm0, ymm0, BYTE_BIT
+    vpand       ymm5, ymm1, ymm6
+    vpsrlw      ymm1, ymm1, BYTE_BIT
+    vpaddw      ymm0, ymm0, ymm4
+    vpaddw      ymm1, ymm1, ymm5
+
+    vpand       ymm4, ymm2, ymm6
+    vpsrlw      ymm2, ymm2, BYTE_BIT
+    vpand       ymm5, ymm3, ymm6
+    vpsrlw      ymm3, ymm3, BYTE_BIT
+    vpaddw      ymm2, ymm2, ymm4
+    vpaddw      ymm3, ymm3, ymm5
+
+    vpaddw      ymm0, ymm0, ymm1
+    vpaddw      ymm2, ymm2, ymm3
+    vpaddw      ymm0, ymm0, ymm7
+    vpaddw      ymm2, ymm2, ymm7
+    vpsrlw      ymm0, ymm0, 2
+    vpsrlw      ymm2, ymm2, 2
+
+    vpackuswb   ymm0, ymm0, ymm2
+    vpermq      ymm0, ymm0, 0xd8
+
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymm0
+
+    sub         rcx, byte SIZEOF_YMMWORD    ; outcol
+    add         rdx, byte 2*SIZEOF_YMMWORD  ; inptr0
+    add         rsi, byte 2*SIZEOF_YMMWORD  ; inptr1
+    add         rdi, byte 1*SIZEOF_YMMWORD  ; outptr
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jae         near .columnloop
+    test        rcx, rcx
+    jnz         near .columnloop_r24
+
+    pop         rsi
+    pop         rdi
+    pop         rcx
+
+    add         rsi, byte 2*SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte 1*SIZEOF_JSAMPROW  ; output_data
+    dec         rax                          ; rowctr
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    uncollect_args 6
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jcsample-sse2.asm Added

@@ -0,0 +1,330 @@
+;
+; jcsample.asm - downsampling (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_sse2(JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor,
+;                            JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                            JSAMPARRAY output_data);
+;
+
+; r10d = JDIMENSION image_width
+; r11 = int max_v_samp_factor
+; r12d = JDIMENSION v_samp_factor
+; r13d = JDIMENSION width_in_blocks
+; r14 = JSAMPARRAY input_data
+; r15 = JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_downsample_sse2)
+
+EXTN(jsimd_h2v1_downsample_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 6
+
+    mov         ecx, r13d
+    shl         rcx, 3                  ; imul rcx,DCTSIZE (rcx = output_cols)
+    jz          near .return
+
+    mov         edx, r10d
+
+    ; -- expand_right_edge
+
+    push        rcx
+    shl         rcx, 1                  ; output_cols * 2
+    sub         rcx, rdx
+    jle         short .expand_end
+
+    mov         rax, r11
+    test        rax, rax
+    jle         short .expand_end
+
+    cld
+    mov         rsi, r14                ; input_data
+.expandloop:
+    push        rax
+    push        rcx
+
+    mov         rdip, JSAMPROW rsi
+    add         rdi, rdx
+    mov         al, JSAMPLE rdi-1
+
+    rep stosb
+
+    pop         rcx
+    pop         rax
+
+    add         rsi, byte SIZEOF_JSAMPROW
+    dec         rax
+    jg          short .expandloop
+
+.expand_end:
+    pop         rcx                     ; output_cols
+
+    ; -- h2v1_downsample
+
+    mov         eax, r12d               ; rowctr
+    test        eax, eax
+    jle         near .return
+
+    mov         rdx, 0x00010000         ; bias pattern
+    movd        xmm7, edx
+    pcmpeqw     xmm6, xmm6
+    pshufd      xmm7, xmm7, 0x00        ; xmm7={0, 1, 0, 1, 0, 1, 0, 1}
+    psrlw       xmm6, BYTE_BIT          ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         rsi, r14                ; input_data
+    mov         rdi, r15                ; output_data
+.rowloop:
+    push        rcx
+    push        rdi
+    push        rsi
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr
+
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jae         short .columnloop
+
+.columnloop_r8:
+    movdqa      xmm0, XMMWORD rsi+0*SIZEOF_XMMWORD
+    pxor        xmm1, xmm1
+    mov         rcx, SIZEOF_XMMWORD
+    jmp         short .downsample
+
+.columnloop:
+    movdqa      xmm0, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqa      xmm1, XMMWORD rsi+1*SIZEOF_XMMWORD
+
+.downsample:
+    movdqa      xmm2, xmm0
+    movdqa      xmm3, xmm1
+
+    pand        xmm0, xmm6
+    psrlw       xmm2, BYTE_BIT
+    pand        xmm1, xmm6
+    psrlw       xmm3, BYTE_BIT
+
+    paddw       xmm0, xmm2
+    paddw       xmm1, xmm3
+    paddw       xmm0, xmm7
+    paddw       xmm1, xmm7
+    psrlw       xmm0, 1
+    psrlw       xmm1, 1
+
+    packuswb    xmm0, xmm1
+
+    movdqa      XMMWORD rdi+0*SIZEOF_XMMWORD, xmm0
+
+    sub         rcx, byte SIZEOF_XMMWORD    ; outcol
+    add         rsi, byte 2*SIZEOF_XMMWORD  ; inptr
+    add         rdi, byte 1*SIZEOF_XMMWORD  ; outptr
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jae         short .columnloop
+    test        rcx, rcx
+    jnz         short .columnloop_r8
+
+    pop         rsi
+    pop         rdi
+    pop         rcx
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         rax                        ; rowctr
+    jg          near .rowloop
+
+.return:
+    uncollect_args 6
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_sse2(JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor,
+;                            JDIMENSION width_in_blocks, JSAMPARRAY input_data,
+;                            JSAMPARRAY output_data);
+;
+
+; r10d = JDIMENSION image_width
+; r11 = int max_v_samp_factor
+; r12d = JDIMENSION v_samp_factor
+; r13d = JDIMENSION width_in_blocks
+; r14 = JSAMPARRAY input_data
+; r15 = JSAMPARRAY output_data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_downsample_sse2)
+
+EXTN(jsimd_h2v2_downsample_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 6
+
+    mov         ecx, r13d
+    shl         rcx, 3                  ; imul rcx,DCTSIZE (rcx = output_cols)
+    jz          near .return
+
+    mov         edx, r10d
+
+    ; -- expand_right_edge
+
+    push        rcx
+    shl         rcx, 1                  ; output_cols * 2
+    sub         rcx, rdx
+    jle         short .expand_end
+
+    mov         rax, r11
+    test        rax, rax
+    jle         short .expand_end
+
+    cld
+    mov         rsi, r14                ; input_data
+.expandloop:
+    push        rax
+    push        rcx
+
+    mov         rdip, JSAMPROW rsi
+    add         rdi, rdx
+    mov         al, JSAMPLE rdi-1
+
+    rep stosb
+
+    pop         rcx
+    pop         rax
+
+    add         rsi, byte SIZEOF_JSAMPROW
+    dec         rax
+    jg          short .expandloop
+
+.expand_end:
+    pop         rcx                     ; output_cols
+
+    ; -- h2v2_downsample
+
+    mov         eax, r12d               ; rowctr
+    test        rax, rax
+    jle         near .return
+
+    mov         rdx, 0x00020001         ; bias pattern
+    movd        xmm7, edx
+    pcmpeqw     xmm6, xmm6
+    pshufd      xmm7, xmm7, 0x00        ; xmm7={1, 2, 1, 2, 1, 2, 1, 2}
+    psrlw       xmm6, BYTE_BIT          ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+    mov         rsi, r14                ; input_data
+    mov         rdi, r15                ; output_data
+.rowloop:
+    push        rcx
+    push        rdi
+    push        rsi
+
+    mov         rdxp, JSAMPROW rsi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         rsip, JSAMPROW rsi+1*SIZEOF_JSAMPROW  ; inptr1
+    mov         rdip, JSAMPROW rdi                    ; outptr
+
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jae         short .columnloop
+
+.columnloop_r8:
+    movdqa      xmm0, XMMWORD rdx+0*SIZEOF_XMMWORD
+    movdqa      xmm1, XMMWORD rsi+0*SIZEOF_XMMWORD
+    pxor        xmm2, xmm2
+    pxor        xmm3, xmm3
+    mov         rcx, SIZEOF_XMMWORD
+    jmp         short .downsample
+
+.columnloop:
+    movdqa      xmm0, XMMWORD rdx+0*SIZEOF_XMMWORD
+    movdqa      xmm1, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqa      xmm2, XMMWORD rdx+1*SIZEOF_XMMWORD
+    movdqa      xmm3, XMMWORD rsi+1*SIZEOF_XMMWORD
+
+.downsample:
+    movdqa      xmm4, xmm0
+    movdqa      xmm5, xmm1
+    pand        xmm0, xmm6
+    psrlw       xmm4, BYTE_BIT
+    pand        xmm1, xmm6
+    psrlw       xmm5, BYTE_BIT
+    paddw       xmm0, xmm4
+    paddw       xmm1, xmm5
+
+    movdqa      xmm4, xmm2
+    movdqa      xmm5, xmm3
+    pand        xmm2, xmm6
+    psrlw       xmm4, BYTE_BIT
+    pand        xmm3, xmm6
+    psrlw       xmm5, BYTE_BIT
+    paddw       xmm2, xmm4
+    paddw       xmm3, xmm5
+
+    paddw       xmm0, xmm1
+    paddw       xmm2, xmm3
+    paddw       xmm0, xmm7
+    paddw       xmm2, xmm7
+    psrlw       xmm0, 2
+    psrlw       xmm2, 2
+
+    packuswb    xmm0, xmm2
+
+    movdqa      XMMWORD rdi+0*SIZEOF_XMMWORD, xmm0
+
+    sub         rcx, byte SIZEOF_XMMWORD    ; outcol
+    add         rdx, byte 2*SIZEOF_XMMWORD  ; inptr0
+    add         rsi, byte 2*SIZEOF_XMMWORD  ; inptr1
+    add         rdi, byte 1*SIZEOF_XMMWORD  ; outptr
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jae         near .columnloop
+    test        rcx, rcx
+    jnz         near .columnloop_r8
+
+    pop         rsi
+    pop         rdi
+    pop         rcx
+
+    add         rsi, byte 2*SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte 1*SIZEOF_JSAMPROW  ; output_data
+    dec         rax                          ; rowctr
+    jg          near .rowloop
+
+.return:
+    uncollect_args 6
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdcolext-avx2.asm Added

@@ -0,0 +1,496 @@
+;
+; jdcolext.asm - colorspace conversion (64-bit AVX2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2012, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_avx2(JDIMENSION out_width, JSAMPIMAGE input_buf,
+;                            JDIMENSION input_row, JSAMPARRAY output_buf,
+;                            int num_rows)
+;
+
+; r10d = JDIMENSION out_width
+; r11 = JSAMPIMAGE input_buf
+; r12d = JDIMENSION input_row
+; r13 = JSAMPARRAY output_buf
+; r14d = int num_rows
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_YMMWORD  ; ymmword wkWK_NUM
+%define WK_NUM  2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_ycc_rgb_convert_avx2)
+
+EXTN(jsimd_ycc_rgb_convert_avx2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 5
+    push        rbx
+
+    mov         ecx, r10d               ; num_cols
+    test        rcx, rcx
+    jz          near .return
+
+    push        rcx
+
+    mov         rdi, r11
+    mov         ecx, r12d
+    mov         rsip, JSAMPARRAY rdi+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rdi+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rdi+2*SIZEOF_JSAMPARRAY
+    lea         rsi, rsi+rcx*SIZEOF_JSAMPROW
+    lea         rbx, rbx+rcx*SIZEOF_JSAMPROW
+    lea         rdx, rdx+rcx*SIZEOF_JSAMPROW
+
+    pop         rcx
+
+    mov         rdi, r13
+    mov         eax, r14d
+    test        rax, rax
+    jle         near .return
+.rowloop:
+    push        rax
+    push        rdi
+    push        rdx
+    push        rbx
+    push        rsi
+    push        rcx                     ; col
+
+    mov         rsip, JSAMPROW rsi    ; inptr0
+    mov         rbxp, JSAMPROW rbx    ; inptr1
+    mov         rdxp, JSAMPROW rdx    ; inptr2
+    mov         rdip, JSAMPROW rdi    ; outptr
+.columnloop:
+
+    vmovdqu     ymm5, YMMWORD rbx     ; ymm5=Cb(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+    vmovdqu     ymm1, YMMWORD rdx     ; ymm1=Cr(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+
+    vpcmpeqw    ymm0, ymm0, ymm0
+    vpcmpeqw    ymm7, ymm7, ymm7
+    vpsrlw      ymm0, ymm0, BYTE_BIT    ; ymm0={0xFF 0x00 0xFF 0x00 ..}
+    vpsllw      ymm7, ymm7, 7           ; ymm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    vpand       ymm4, ymm0, ymm5        ; ymm4=Cb(02468ACEGIKMOQSU)=CbE
+    vpsrlw      ymm5, ymm5, BYTE_BIT    ; ymm5=Cb(13579BDFHJLNPRTV)=CbO
+    vpand       ymm0, ymm0, ymm1        ; ymm0=Cr(02468ACEGIKMOQSU)=CrE
+    vpsrlw      ymm1, ymm1, BYTE_BIT    ; ymm1=Cr(13579BDFHJLNPRTV)=CrO
+
+    vpaddw      ymm2, ymm4, ymm7
+    vpaddw      ymm3, ymm5, ymm7
+    vpaddw      ymm6, ymm0, ymm7
+    vpaddw      ymm7, ymm1, ymm7
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    vpaddw      ymm4, ymm2, ymm2             ; ymm4=2*CbE
+    vpaddw      ymm5, ymm3, ymm3             ; ymm5=2*CbO
+    vpaddw      ymm0, ymm6, ymm6             ; ymm0=2*CrE
+    vpaddw      ymm1, ymm7, ymm7             ; ymm1=2*CrO
+
+    vpmulhw     ymm4, ymm4, rel PW_MF0228  ; ymm4=(2*CbE * -FIX(0.22800))
+    vpmulhw     ymm5, ymm5, rel PW_MF0228  ; ymm5=(2*CbO * -FIX(0.22800))
+    vpmulhw     ymm0, ymm0, rel PW_F0402   ; ymm0=(2*CrE * FIX(0.40200))
+    vpmulhw     ymm1, ymm1, rel PW_F0402   ; ymm1=(2*CrO * FIX(0.40200))
+
+    vpaddw      ymm4, ymm4, rel PW_ONE
+    vpaddw      ymm5, ymm5, rel PW_ONE
+    vpsraw      ymm4, ymm4, 1                ; ymm4=(CbE * -FIX(0.22800))
+    vpsraw      ymm5, ymm5, 1                ; ymm5=(CbO * -FIX(0.22800))
+    vpaddw      ymm0, ymm0, rel PW_ONE
+    vpaddw      ymm1, ymm1, rel PW_ONE
+    vpsraw      ymm0, ymm0, 1                ; ymm0=(CrE * FIX(0.40200))
+    vpsraw      ymm1, ymm1, 1                ; ymm1=(CrO * FIX(0.40200))
+
+    vpaddw      ymm4, ymm4, ymm2
+    vpaddw      ymm5, ymm5, ymm3
+    vpaddw      ymm4, ymm4, ymm2             ; ymm4=(CbE * FIX(1.77200))=(B-Y)E
+    vpaddw      ymm5, ymm5, ymm3             ; ymm5=(CbO * FIX(1.77200))=(B-Y)O
+    vpaddw      ymm0, ymm0, ymm6             ; ymm0=(CrE * FIX(1.40200))=(R-Y)E
+    vpaddw      ymm1, ymm1, ymm7             ; ymm1=(CrO * FIX(1.40200))=(R-Y)O
+
+    vmovdqa     YMMWORD wk(0), ymm4        ; wk(0)=(B-Y)E
+    vmovdqa     YMMWORD wk(1), ymm5        ; wk(1)=(B-Y)O
+
+    vpunpckhwd  ymm4, ymm2, ymm6
+    vpunpcklwd  ymm2, ymm2, ymm6
+    vpmaddwd    ymm2, ymm2, rel PW_MF0344_F0285
+    vpmaddwd    ymm4, ymm4, rel PW_MF0344_F0285
+    vpunpckhwd  ymm5, ymm3, ymm7
+    vpunpcklwd  ymm3, ymm3, ymm7
+    vpmaddwd    ymm3, ymm3, rel PW_MF0344_F0285
+    vpmaddwd    ymm5, ymm5, rel PW_MF0344_F0285
+
+    vpaddd      ymm2, ymm2, rel PD_ONEHALF
+    vpaddd      ymm4, ymm4, rel PD_ONEHALF
+    vpsrad      ymm2, ymm2, SCALEBITS
+    vpsrad      ymm4, ymm4, SCALEBITS
+    vpaddd      ymm3, ymm3, rel PD_ONEHALF
+    vpaddd      ymm5, ymm5, rel PD_ONEHALF
+    vpsrad      ymm3, ymm3, SCALEBITS
+    vpsrad      ymm5, ymm5, SCALEBITS
+
+    vpackssdw   ymm2, ymm2, ymm4             ; ymm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+    vpackssdw   ymm3, ymm3, ymm5             ; ymm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+    vpsubw      ymm2, ymm2, ymm6             ; ymm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+    vpsubw      ymm3, ymm3, ymm7             ; ymm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+    vmovdqu     ymm5, YMMWORD rsi          ; ymm5=Y(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+
+    vpcmpeqw    ymm4, ymm4, ymm4
+    vpsrlw      ymm4, ymm4, BYTE_BIT         ; ymm4={0xFF 0x00 0xFF 0x00 ..}
+    vpand       ymm4, ymm4, ymm5             ; ymm4=Y(02468ACEGIKMOQSU)=YE
+    vpsrlw      ymm5, ymm5, BYTE_BIT         ; ymm5=Y(13579BDFHJLNPRTV)=YO
+
+    vpaddw      ymm0, ymm0, ymm4             ; ymm0=((R-Y)E+YE)=RE=R(02468ACEGIKMOQSU)
+    vpaddw      ymm1, ymm1, ymm5             ; ymm1=((R-Y)O+YO)=RO=R(13579BDFHJLNPRTV)
+    vpackuswb   ymm0, ymm0, ymm0             ; ymm0=R(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm1, ymm1, ymm1             ; ymm1=R(13579BDF********HJLNPRTV********)
+
+    vpaddw      ymm2, ymm2, ymm4             ; ymm2=((G-Y)E+YE)=GE=G(02468ACEGIKMOQSU)
+    vpaddw      ymm3, ymm3, ymm5             ; ymm3=((G-Y)O+YO)=GO=G(13579BDFHJLNPRTV)
+    vpackuswb   ymm2, ymm2, ymm2             ; ymm2=G(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm3, ymm3, ymm3             ; ymm3=G(13579BDF********HJLNPRTV********)
+
+    vpaddw      ymm4, ymm4, YMMWORD wk(0)  ; ymm4=(YE+(B-Y)E)=BE=B(02468ACEGIKMOQSU)
+    vpaddw      ymm5, ymm5, YMMWORD wk(1)  ; ymm5=(YO+(B-Y)O)=BO=B(13579BDFHJLNPRTV)
+    vpackuswb   ymm4, ymm4, ymm4             ; ymm4=B(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm5, ymm5, ymm5             ; ymm5=B(13579BDF********HJLNPRTV********)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; ymmA=(00 02 04 06 08 0A 0C 0E ** 0G 0I 0K 0M 0O 0Q 0S 0U **)
+    ; ymmB=(01 03 05 07 09 0B 0D 0F ** 0H 0J 0L 0N 0P 0R 0T 0V **)
+    ; ymmC=(10 12 14 16 18 1A 1C 1E ** 1G 1I 1K 1M 1O 1Q 1S 1U **)
+    ; ymmD=(11 13 15 17 19 1B 1D 1F ** 1H 1J 1L 1N 1P 1R 1T 1V **)
+    ; ymmE=(20 22 24 26 28 2A 2C 2E ** 2G 2I 2K 2M 2O 2Q 2S 2U **)
+    ; ymmF=(21 23 25 27 29 2B 2D 2F ** 2H 2J 2L 2N 2P 2R 2T 2V **)
+    ; ymmG=(** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** **)
+    ; ymmH=(** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** **)
+
+    vpunpcklbw  ymmA, ymmA, ymmC        ; ymmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E
+                                        ;       0G 1G 0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U)
+    vpunpcklbw  ymmE, ymmE, ymmB        ; ymmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F
+                                        ;       2G 0H 2I 0J 2K 0L 2M 0N 2O 0P 2Q 0R 2S 0T 2U 0V)
+    vpunpcklbw  ymmD, ymmD, ymmF        ; ymmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F
+                                        ;       1H 2H 1J 2J 1L 2L 1N 2N 1P 2P 1R 2R 1T 2T 1V 2V)
+
+    vpsrldq     ymmH, ymmA, 2           ; ymmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E 0G 1G
+                                        ;       0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U -- --)
+    vpunpckhwd  ymmG, ymmA, ymmE        ; ymmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F
+                                        ;       0O 1O 2O 0P 0Q 1Q 2Q 0R 0S 1S 2S 0T 0U 1U 2U 0V)
+    vpunpcklwd  ymmA, ymmA, ymmE        ; ymmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07
+                                        ;       0G 1G 2G 0H 0I 1I 2I 0J 0K 1K 2K 0L 0M 1M 2M 0N)
+
+    vpsrldq     ymmE, ymmE, 2           ; ymmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F 2G 0H
+                                        ;       2I 0J 2K 0L 2M 0N 2O 0P 2Q 0R 2S 0T 2U 0V -- --)
+
+    vpsrldq     ymmB, ymmD, 2           ; ymmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F 1H 2H
+                                        ;       1J 2J 1L 2L 1N 2N 1P 2P 1R 2R 1T 2T 1V 2V -- --)
+    vpunpckhwd  ymmC, ymmD, ymmH        ; ymmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F 0G 1G
+                                        ;       1P 2P 0Q 1Q 1R 2R 0S 1S 1T 2T 0U 1U 1V 2V -- --)
+    vpunpcklwd  ymmD, ymmD, ymmH        ; ymmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18
+                                        ;       1H 2H 0I 1I 1J 2J 0K 1K 1L 2L 0M 1M 1N 2N 0O 1O)
+
+    vpunpckhwd  ymmF, ymmE, ymmB        ; ymmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F 2G 0H 1H 2H
+                                        ;       2Q 0R 1R 2R 2S 0T 1T 2T 2U 0V 1V 2V -- -- -- --)
+    vpunpcklwd  ymmE, ymmE, ymmB        ; ymmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29
+                                        ;       2I 0J 1J 2J 2K 0L 1L 2L 2M 0N 1N 2N 2O 0P 1P 2P)
+
+    vpshufd     ymmH, ymmA, 0x4E        ; ymmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03
+                                        ;       0K 1K 2K 0L 0M 1M 2M 0N 0G 1G 2G 0H 0I 1I 2I 0J)
+    vpunpckldq  ymmA, ymmA, ymmD        ; ymmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 0I 1I 2I 0J 1J 2J 0K 1K)
+    vpunpckhdq  ymmD, ymmD, ymmE        ; ymmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29
+                                        ;       1L 2L 0M 1M 2M 0N 1N 2N 1N 2N 0O 1O 2O 0P 1P 2P)
+    vpunpckldq  ymmE, ymmE, ymmH        ; ymmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07
+                                        ;       2I 0J 1J 2J 0K 1K 2K 0L 2K 0L 1L 2L 0M 1M 2M 0N)
+
+    vpshufd     ymmH, ymmG, 0x4E        ; ymmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B
+                                        ;       0S 1S 2S 0T 0U 1U 2U 0V 0O 1O 2O 0P 0Q 1Q 2Q 0R)
+    vpunpckldq  ymmG, ymmG, ymmC        ; ymmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C
+                                        ;       0O 1O 2O 0P 1P 2P 0Q 1Q 0Q 1Q 2Q 0R 1R 2R 0S 1S)
+    vpunpckhdq  ymmC, ymmC, ymmF        ; ymmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F 0G 1G 2G 0H 1H 2H
+                                        ;       1T 2T 0U 1U 2U 0V 1V 2V 1V 2V -- -- -- -- -- --)
+    vpunpckldq  ymmF, ymmF, ymmH        ; ymmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 2S 0T 1T 2T 0U 1U 2U 0V)
+
+    vpunpcklqdq ymmH, ymmA, ymmE        ; ymmH=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vpunpcklqdq ymmG, ymmD, ymmG        ; ymmG=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A
+                                        ;       1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q)
+    vpunpcklqdq ymmC, ymmF, ymmC        ; ymmC=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    vperm2i128  ymmA, ymmH, ymmG, 0x20  ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                        ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    vperm2i128  ymmD, ymmC, ymmH, 0x30  ; ymmD=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vperm2i128  ymmF, ymmG, ymmC, 0x31  ; ymmF=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jb          short .column_st64
+
+    test        rdi, SIZEOF_YMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    vmovntdq    YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovntdq    YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    vmovntdq    YMMWORD rdi+2*SIZEOF_YMMWORD, ymmF
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    vmovdqu     YMMWORD rdi+2*SIZEOF_YMMWORD, ymmF
+.out0:
+    add         rdi, byte RGB_PIXELSIZE*SIZEOF_YMMWORD  ; outptr
+    sub         rcx, byte SIZEOF_YMMWORD
+    jz          near .nextrow
+
+    add         rsi, byte SIZEOF_YMMWORD  ; inptr0
+    add         rbx, byte SIZEOF_YMMWORD  ; inptr1
+    add         rdx, byte SIZEOF_YMMWORD  ; inptr2
+    jmp         near .columnloop
+
+.column_st64:
+    lea         rcx, rcx+rcx*2            ; imul ecx, RGB_PIXELSIZE
+    cmp         rcx, byte 2*SIZEOF_YMMWORD
+    jb          short .column_st32
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    add         rdi, byte 2*SIZEOF_YMMWORD  ; outptr
+    vmovdqa     ymmA, ymmF
+    sub         rcx, byte 2*SIZEOF_YMMWORD
+    jmp         short .column_st31
+.column_st32:
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jb          short .column_st31
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    add         rdi, byte SIZEOF_YMMWORD    ; outptr
+    vmovdqa     ymmA, ymmD
+    sub         rcx, byte SIZEOF_YMMWORD
+    jmp         short .column_st31
+.column_st31:
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jb          short .column_st15
+    vmovdqu     XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    add         rdi, byte SIZEOF_XMMWORD    ; outptr
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    sub         rcx, byte SIZEOF_XMMWORD
+.column_st15:
+    ; Store the lower 8 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_MMWORD
+    jb          short .column_st7
+    vmovq       XMM_MMWORD rdi, xmmA
+    add         rdi, byte SIZEOF_MMWORD
+    sub         rcx, byte SIZEOF_MMWORD
+    vpsrldq     xmmA, xmmA, SIZEOF_MMWORD
+.column_st7:
+    ; Store the lower 4 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_DWORD
+    jb          short .column_st3
+    vmovd       XMM_DWORD rdi, xmmA
+    add         rdi, byte SIZEOF_DWORD
+    sub         rcx, byte SIZEOF_DWORD
+    vpsrldq     xmmA, xmmA, SIZEOF_DWORD
+.column_st3:
+    ; Store the lower 2 bytes of rax to the output when it has enough
+    ; space.
+    vmovd       eax, xmmA
+    cmp         rcx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word rdi, ax
+    add         rdi, byte SIZEOF_WORD
+    sub         rcx, byte SIZEOF_WORD
+    shr         rax, 16
+.column_st1:
+    ; Store the lower 1 byte of rax to the output when it has enough
+    ; space.
+    test        rcx, rcx
+    jz          short .nextrow
+    mov         byte rdi, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    vpcmpeqb    ymm6, ymm6, ymm6        ; ymm6=XE=X(02468ACE********GIKMOQSU********)
+    vpcmpeqb    ymm7, ymm7, ymm7        ; ymm7=XO=X(13579BDF********HJLNPRTV********)
+%else
+    vpxor       ymm6, ymm6, ymm6        ; ymm6=XE=X(02468ACE********GIKMOQSU********)
+    vpxor       ymm7, ymm7, ymm7        ; ymm7=XO=X(13579BDF********HJLNPRTV********)
+%endif
+    ; ymmA=(00 02 04 06 08 0A 0C 0E ** 0G 0I 0K 0M 0O 0Q 0S 0U **)
+    ; ymmB=(01 03 05 07 09 0B 0D 0F ** 0H 0J 0L 0N 0P 0R 0T 0V **)
+    ; ymmC=(10 12 14 16 18 1A 1C 1E ** 1G 1I 1K 1M 1O 1Q 1S 1U **)
+    ; ymmD=(11 13 15 17 19 1B 1D 1F ** 1H 1J 1L 1N 1P 1R 1T 1V **)
+    ; ymmE=(20 22 24 26 28 2A 2C 2E ** 2G 2I 2K 2M 2O 2Q 2S 2U **)
+    ; ymmF=(21 23 25 27 29 2B 2D 2F ** 2H 2J 2L 2N 2P 2R 2T 2V **)
+    ; ymmG=(30 32 34 36 38 3A 3C 3E ** 3G 3I 3K 3M 3O 3Q 3S 3U **)
+    ; ymmH=(31 33 35 37 39 3B 3D 3F ** 3H 3J 3L 3N 3P 3R 3T 3V **)
+
+    vpunpcklbw  ymmA, ymmA, ymmC        ; ymmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E
+                                        ;       0G 1G 0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U)
+    vpunpcklbw  ymmE, ymmE, ymmG        ; ymmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E
+                                        ;       2G 3G 2I 3I 2K 3K 2M 3M 2O 3O 2Q 3Q 2S 3S 2U 3U)
+    vpunpcklbw  ymmB, ymmB, ymmD        ; ymmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F
+                                        ;       0H 1H 0J 1J 0L 1L 0N 1N 0P 1P 0R 1R 0T 1T 0V 1V)
+    vpunpcklbw  ymmF, ymmF, ymmH        ; ymmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F
+                                        ;       2H 3H 2J 3J 2L 3L 2N 3N 2P 3P 2R 3R 2T 3T 2V 3V)
+
+    vpunpckhwd  ymmC, ymmA, ymmE        ; ymmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E
+                                        ;       0O 1O 2O 3O 0Q 1Q 2Q 3Q 0S 1S 2S 3S 0U 1U 2U 3U)
+    vpunpcklwd  ymmA, ymmA, ymmE        ; ymmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36
+                                        ;       0G 1G 2G 3G 0I 1I 2I 3I 0K 1K 2K 3K 0M 1M 2M 3M)
+    vpunpckhwd  ymmG, ymmB, ymmF        ; ymmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F
+                                        ;       0P 1P 2P 3P 0R 1R 2R 3R 0T 1T 2T 3T 0V 1V 2V 3V)
+    vpunpcklwd  ymmB, ymmB, ymmF        ; ymmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37
+                                        ;       0H 1H 2H 3H 0J 1J 2J 3J 0L 1L 2L 3L 0N 1N 2N 3N)
+
+    vpunpckhdq  ymmE, ymmA, ymmB        ; ymmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    vpunpckldq  ymmB, ymmA, ymmB        ; ymmB=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J)
+    vpunpckhdq  ymmF, ymmC, ymmG        ; ymmF=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+    vpunpckldq  ymmG, ymmC, ymmG        ; ymmG=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R)
+
+    vperm2i128  ymmA, ymmB, ymmE, 0x20  ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    vperm2i128  ymmD, ymmG, ymmF, 0x20  ; ymmD=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+    vperm2i128  ymmC, ymmB, ymmE, 0x31  ; ymmC=(0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    vperm2i128  ymmH, ymmG, ymmF, 0x31  ; ymmH=(0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jb          short .column_st64
+
+    test        rdi, SIZEOF_YMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    vmovntdq    YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovntdq    YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    vmovntdq    YMMWORD rdi+2*SIZEOF_YMMWORD, ymmC
+    vmovntdq    YMMWORD rdi+3*SIZEOF_YMMWORD, ymmH
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    vmovdqu     YMMWORD rdi+2*SIZEOF_YMMWORD, ymmC
+    vmovdqu     YMMWORD rdi+3*SIZEOF_YMMWORD, ymmH
+.out0:
+    add         rdi, RGB_PIXELSIZE*SIZEOF_YMMWORD  ; outptr
+    sub         rcx, byte SIZEOF_YMMWORD
+    jz          near .nextrow
+
+    add         rsi, byte SIZEOF_YMMWORD  ; inptr0
+    add         rbx, byte SIZEOF_YMMWORD  ; inptr1
+    add         rdx, byte SIZEOF_YMMWORD  ; inptr2
+    jmp         near .columnloop
+
+.column_st64:
+    cmp         rcx, byte SIZEOF_YMMWORD/2
+    jb          short .column_st32
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    add         rdi, byte 2*SIZEOF_YMMWORD  ; outptr
+    vmovdqa     ymmA, ymmC
+    vmovdqa     ymmD, ymmH
+    sub         rcx, byte SIZEOF_YMMWORD/2
+.column_st32:
+    cmp         rcx, byte SIZEOF_YMMWORD/4
+    jb          short .column_st16
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    add         rdi, byte SIZEOF_YMMWORD    ; outptr
+    vmovdqa     ymmA, ymmD
+    sub         rcx, byte SIZEOF_YMMWORD/4
+.column_st16:
+    cmp         rcx, byte SIZEOF_YMMWORD/8
+    jb          short .column_st15
+    vmovdqu     XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    add         rdi, byte SIZEOF_XMMWORD    ; outptr
+    sub         rcx, byte SIZEOF_YMMWORD/8
+.column_st15:
+    ; Store two pixels (8 bytes) of ymmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_YMMWORD/16
+    jb          short .column_st7
+    vmovq       MMWORD rdi, xmmA
+    add         rdi, byte SIZEOF_YMMWORD/16*4
+    sub         rcx, byte SIZEOF_YMMWORD/16
+    vpsrldq     xmmA, SIZEOF_YMMWORD/16*4
+.column_st7:
+    ; Store one pixel (4 bytes) of ymmA to the output when it has enough
+    ; space.
+    test        rcx, rcx
+    jz          short .nextrow
+    vmovd       XMM_DWORD rdi, xmmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+.nextrow:
+    pop         rcx
+    pop         rsi
+    pop         rbx
+    pop         rdx
+    pop         rdi
+    pop         rax
+
+    add         rsi, byte SIZEOF_JSAMPROW
+    add         rbx, byte SIZEOF_JSAMPROW
+    add         rdx, byte SIZEOF_JSAMPROW
+    add         rdi, byte SIZEOF_JSAMPROW  ; output_buf
+    dec         rax                        ; num_rows
+    jg          near .rowloop
+
+    sfence                              ; flush the write buffer
+
+.return:
+    pop         rbx
+    vzeroupper
+    uncollect_args 5
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdcolext-sse2.asm Added

@@ -0,0 +1,439 @@
+;
+; jdcolext.asm - colorspace conversion (64-bit SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2012, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_sse2(JDIMENSION out_width, JSAMPIMAGE input_buf,
+;                            JDIMENSION input_row, JSAMPARRAY output_buf,
+;                            int num_rows)
+;
+
+; r10d = JDIMENSION out_width
+; r11 = JSAMPIMAGE input_buf
+; r12d = JDIMENSION input_row
+; r13 = JSAMPARRAY output_buf
+; r14d = int num_rows
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD  ; xmmword wkWK_NUM
+%define WK_NUM  2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_ycc_rgb_convert_sse2)
+
+EXTN(jsimd_ycc_rgb_convert_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 5
+    push        rbx
+
+    mov         ecx, r10d               ; num_cols
+    test        rcx, rcx
+    jz          near .return
+
+    push        rcx
+
+    mov         rdi, r11
+    mov         ecx, r12d
+    mov         rsip, JSAMPARRAY rdi+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rdi+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rdi+2*SIZEOF_JSAMPARRAY
+    lea         rsi, rsi+rcx*SIZEOF_JSAMPROW
+    lea         rbx, rbx+rcx*SIZEOF_JSAMPROW
+    lea         rdx, rdx+rcx*SIZEOF_JSAMPROW
+
+    pop         rcx
+
+    mov         rdi, r13
+    mov         eax, r14d
+    test        rax, rax
+    jle         near .return
+.rowloop:
+    push        rax
+    push        rdi
+    push        rdx
+    push        rbx
+    push        rsi
+    push        rcx                     ; col
+
+    mov         rsip, JSAMPROW rsi    ; inptr0
+    mov         rbxp, JSAMPROW rbx    ; inptr1
+    mov         rdxp, JSAMPROW rdx    ; inptr2
+    mov         rdip, JSAMPROW rdi    ; outptr
+.columnloop:
+
+    movdqa      xmm5, XMMWORD rbx     ; xmm5=Cb(0123456789ABCDEF)
+    movdqa      xmm1, XMMWORD rdx     ; xmm1=Cr(0123456789ABCDEF)
+
+    pcmpeqw     xmm4, xmm4
+    pcmpeqw     xmm7, xmm7
+    psrlw       xmm4, BYTE_BIT
+    psllw       xmm7, 7                 ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+    movdqa      xmm0, xmm4              ; xmm0=xmm4={0xFF 0x00 0xFF 0x00 ..}
+
+    pand        xmm4, xmm5              ; xmm4=Cb(02468ACE)=CbE
+    psrlw       xmm5, BYTE_BIT          ; xmm5=Cb(13579BDF)=CbO
+    pand        xmm0, xmm1              ; xmm0=Cr(02468ACE)=CrE
+    psrlw       xmm1, BYTE_BIT          ; xmm1=Cr(13579BDF)=CrO
+
+    paddw       xmm4, xmm7
+    paddw       xmm5, xmm7
+    paddw       xmm0, xmm7
+    paddw       xmm1, xmm7
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    movdqa      xmm2, xmm4              ; xmm2=CbE
+    movdqa      xmm3, xmm5              ; xmm3=CbO
+    paddw       xmm4, xmm4              ; xmm4=2*CbE
+    paddw       xmm5, xmm5              ; xmm5=2*CbO
+    movdqa      xmm6, xmm0              ; xmm6=CrE
+    movdqa      xmm7, xmm1              ; xmm7=CrO
+    paddw       xmm0, xmm0              ; xmm0=2*CrE
+    paddw       xmm1, xmm1              ; xmm1=2*CrO
+
+    pmulhw      xmm4, rel PW_MF0228   ; xmm4=(2*CbE * -FIX(0.22800))
+    pmulhw      xmm5, rel PW_MF0228   ; xmm5=(2*CbO * -FIX(0.22800))
+    pmulhw      xmm0, rel PW_F0402    ; xmm0=(2*CrE * FIX(0.40200))
+    pmulhw      xmm1, rel PW_F0402    ; xmm1=(2*CrO * FIX(0.40200))
+
+    paddw       xmm4, rel PW_ONE
+    paddw       xmm5, rel PW_ONE
+    psraw       xmm4, 1                 ; xmm4=(CbE * -FIX(0.22800))
+    psraw       xmm5, 1                 ; xmm5=(CbO * -FIX(0.22800))
+    paddw       xmm0, rel PW_ONE
+    paddw       xmm1, rel PW_ONE
+    psraw       xmm0, 1                 ; xmm0=(CrE * FIX(0.40200))
+    psraw       xmm1, 1                 ; xmm1=(CrO * FIX(0.40200))
+
+    paddw       xmm4, xmm2
+    paddw       xmm5, xmm3
+    paddw       xmm4, xmm2              ; xmm4=(CbE * FIX(1.77200))=(B-Y)E
+    paddw       xmm5, xmm3              ; xmm5=(CbO * FIX(1.77200))=(B-Y)O
+    paddw       xmm0, xmm6              ; xmm0=(CrE * FIX(1.40200))=(R-Y)E
+    paddw       xmm1, xmm7              ; xmm1=(CrO * FIX(1.40200))=(R-Y)O
+
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=(B-Y)E
+    movdqa      XMMWORD wk(1), xmm5   ; wk(1)=(B-Y)O
+
+    movdqa      xmm4, xmm2
+    movdqa      xmm5, xmm3
+    punpcklwd   xmm2, xmm6
+    punpckhwd   xmm4, xmm6
+    pmaddwd     xmm2, rel PW_MF0344_F0285
+    pmaddwd     xmm4, rel PW_MF0344_F0285
+    punpcklwd   xmm3, xmm7
+    punpckhwd   xmm5, xmm7
+    pmaddwd     xmm3, rel PW_MF0344_F0285
+    pmaddwd     xmm5, rel PW_MF0344_F0285
+
+    paddd       xmm2, rel PD_ONEHALF
+    paddd       xmm4, rel PD_ONEHALF
+    psrad       xmm2, SCALEBITS
+    psrad       xmm4, SCALEBITS
+    paddd       xmm3, rel PD_ONEHALF
+    paddd       xmm5, rel PD_ONEHALF
+    psrad       xmm3, SCALEBITS
+    psrad       xmm5, SCALEBITS
+
+    packssdw    xmm2, xmm4              ; xmm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+    packssdw    xmm3, xmm5              ; xmm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+    psubw       xmm2, xmm6              ; xmm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+    psubw       xmm3, xmm7              ; xmm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+    movdqa      xmm5, XMMWORD rsi     ; xmm5=Y(0123456789ABCDEF)
+
+    pcmpeqw     xmm4, xmm4
+    psrlw       xmm4, BYTE_BIT          ; xmm4={0xFF 0x00 0xFF 0x00 ..}
+    pand        xmm4, xmm5              ; xmm4=Y(02468ACE)=YE
+    psrlw       xmm5, BYTE_BIT          ; xmm5=Y(13579BDF)=YO
+
+    paddw       xmm0, xmm4              ; xmm0=((R-Y)E+YE)=RE=R(02468ACE)
+    paddw       xmm1, xmm5              ; xmm1=((R-Y)O+YO)=RO=R(13579BDF)
+    packuswb    xmm0, xmm0              ; xmm0=R(02468ACE********)
+    packuswb    xmm1, xmm1              ; xmm1=R(13579BDF********)
+
+    paddw       xmm2, xmm4              ; xmm2=((G-Y)E+YE)=GE=G(02468ACE)
+    paddw       xmm3, xmm5              ; xmm3=((G-Y)O+YO)=GO=G(13579BDF)
+    packuswb    xmm2, xmm2              ; xmm2=G(02468ACE********)
+    packuswb    xmm3, xmm3              ; xmm3=G(13579BDF********)
+
+    paddw       xmm4, XMMWORD wk(0)   ; xmm4=(YE+(B-Y)E)=BE=B(02468ACE)
+    paddw       xmm5, XMMWORD wk(1)   ; xmm5=(YO+(B-Y)O)=BO=B(13579BDF)
+    packuswb    xmm4, xmm4              ; xmm4=B(02468ACE********)
+    packuswb    xmm5, xmm5              ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+    ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+    ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+    ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+    punpcklbw   xmmA, xmmC        ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+    punpcklbw   xmmE, xmmB        ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+    punpcklbw   xmmD, xmmF        ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+    movdqa      xmmG, xmmA
+    movdqa      xmmH, xmmA
+    punpcklwd   xmmA, xmmE        ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+    punpckhwd   xmmG, xmmE        ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+    psrldq      xmmH, 2           ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+    psrldq      xmmE, 2           ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+    movdqa      xmmC, xmmD
+    movdqa      xmmB, xmmD
+    punpcklwd   xmmD, xmmH        ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+    punpckhwd   xmmC, xmmH        ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+    psrldq      xmmB, 2           ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+    movdqa      xmmF, xmmE
+    punpcklwd   xmmE, xmmB        ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+    punpckhwd   xmmF, xmmB        ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+    pshufd      xmmH, xmmA, 0x4E  ; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+    movdqa      xmmB, xmmE
+    punpckldq   xmmA, xmmD        ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+    punpckldq   xmmE, xmmH        ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+    punpckhdq   xmmD, xmmB        ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+    pshufd      xmmH, xmmG, 0x4E  ; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+    movdqa      xmmB, xmmF
+    punpckldq   xmmG, xmmC        ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+    punpckldq   xmmF, xmmH        ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+    punpckhdq   xmmC, xmmB        ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+    punpcklqdq  xmmA, xmmE        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+    punpcklqdq  xmmD, xmmG        ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    punpcklqdq  xmmF, xmmC        ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jb          short .column_st32
+
+    test        rdi, SIZEOF_XMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    movntdq     XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movntdq     XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    movntdq     XMMWORD rdi+2*SIZEOF_XMMWORD, xmmF
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    movdqu      XMMWORD rdi+2*SIZEOF_XMMWORD, xmmF
+.out0:
+    add         rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+    sub         rcx, byte SIZEOF_XMMWORD
+    jz          near .nextrow
+
+    add         rsi, byte SIZEOF_XMMWORD  ; inptr0
+    add         rbx, byte SIZEOF_XMMWORD  ; inptr1
+    add         rdx, byte SIZEOF_XMMWORD  ; inptr2
+    jmp         near .columnloop
+
+.column_st32:
+    lea         rcx, rcx+rcx*2            ; imul ecx, RGB_PIXELSIZE
+    cmp         rcx, byte 2*SIZEOF_XMMWORD
+    jb          short .column_st16
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    add         rdi, byte 2*SIZEOF_XMMWORD  ; outptr
+    movdqa      xmmA, xmmF
+    sub         rcx, byte 2*SIZEOF_XMMWORD
+    jmp         short .column_st15
+.column_st16:
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jb          short .column_st15
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    add         rdi, byte SIZEOF_XMMWORD    ; outptr
+    movdqa      xmmA, xmmD
+    sub         rcx, byte SIZEOF_XMMWORD
+.column_st15:
+    ; Store the lower 8 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_MMWORD
+    jb          short .column_st7
+    movq        XMM_MMWORD rdi, xmmA
+    add         rdi, byte SIZEOF_MMWORD
+    sub         rcx, byte SIZEOF_MMWORD
+    psrldq      xmmA, SIZEOF_MMWORD
+.column_st7:
+    ; Store the lower 4 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_DWORD
+    jb          short .column_st3
+    movd        XMM_DWORD rdi, xmmA
+    add         rdi, byte SIZEOF_DWORD
+    sub         rcx, byte SIZEOF_DWORD
+    psrldq      xmmA, SIZEOF_DWORD
+.column_st3:
+    ; Store the lower 2 bytes of rax to the output when it has enough
+    ; space.
+    movd        eax, xmmA
+    cmp         rcx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word rdi, ax
+    add         rdi, byte SIZEOF_WORD
+    sub         rcx, byte SIZEOF_WORD
+    shr         rax, 16
+.column_st1:
+    ; Store the lower 1 byte of rax to the output when it has enough
+    ; space.
+    test        rcx, rcx
+    jz          short .nextrow
+    mov         byte rdi, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    pcmpeqb     xmm6, xmm6              ; xmm6=XE=X(02468ACE********)
+    pcmpeqb     xmm7, xmm7              ; xmm7=XO=X(13579BDF********)
+%else
+    pxor        xmm6, xmm6              ; xmm6=XE=X(02468ACE********)
+    pxor        xmm7, xmm7              ; xmm7=XO=X(13579BDF********)
+%endif
+    ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+    ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+    ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+    ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+    punpcklbw   xmmA, xmmC  ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+    punpcklbw   xmmE, xmmG  ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+    punpcklbw   xmmB, xmmD  ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+    punpcklbw   xmmF, xmmH  ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+    movdqa      xmmC, xmmA
+    punpcklwd   xmmA, xmmE  ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+    punpckhwd   xmmC, xmmE  ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+    movdqa      xmmG, xmmB
+    punpcklwd   xmmB, xmmF  ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+    punpckhwd   xmmG, xmmF  ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+    movdqa      xmmD, xmmA
+    punpckldq   xmmA, xmmB  ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+    punpckhdq   xmmD, xmmB  ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    movdqa      xmmH, xmmC
+    punpckldq   xmmC, xmmG  ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+    punpckhdq   xmmH, xmmG  ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jb          short .column_st32
+
+    test        rdi, SIZEOF_XMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    movntdq     XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movntdq     XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    movntdq     XMMWORD rdi+2*SIZEOF_XMMWORD, xmmC
+    movntdq     XMMWORD rdi+3*SIZEOF_XMMWORD, xmmH
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    movdqu      XMMWORD rdi+2*SIZEOF_XMMWORD, xmmC
+    movdqu      XMMWORD rdi+3*SIZEOF_XMMWORD, xmmH
+.out0:
+    add         rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+    sub         rcx, byte SIZEOF_XMMWORD
+    jz          near .nextrow
+
+    add         rsi, byte SIZEOF_XMMWORD  ; inptr0
+    add         rbx, byte SIZEOF_XMMWORD  ; inptr1
+    add         rdx, byte SIZEOF_XMMWORD  ; inptr2
+    jmp         near .columnloop
+
+.column_st32:
+    cmp         rcx, byte SIZEOF_XMMWORD/2
+    jb          short .column_st16
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    add         rdi, byte 2*SIZEOF_XMMWORD  ; outptr
+    movdqa      xmmA, xmmC
+    movdqa      xmmD, xmmH
+    sub         rcx, byte SIZEOF_XMMWORD/2
+.column_st16:
+    cmp         rcx, byte SIZEOF_XMMWORD/4
+    jb          short .column_st15
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    add         rdi, byte SIZEOF_XMMWORD    ; outptr
+    movdqa      xmmA, xmmD
+    sub         rcx, byte SIZEOF_XMMWORD/4
+.column_st15:
+    ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_XMMWORD/8
+    jb          short .column_st7
+    movq        MMWORD rdi, xmmA
+    add         rdi, byte SIZEOF_XMMWORD/8*4
+    sub         rcx, byte SIZEOF_XMMWORD/8
+    psrldq      xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+    ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+    ; space.
+    test        rcx, rcx
+    jz          short .nextrow
+    movd        XMM_DWORD rdi, xmmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+.nextrow:
+    pop         rcx
+    pop         rsi
+    pop         rbx
+    pop         rdx
+    pop         rdi
+    pop         rax
+
+    add         rsi, byte SIZEOF_JSAMPROW
+    add         rbx, byte SIZEOF_JSAMPROW
+    add         rdx, byte SIZEOF_JSAMPROW
+    add         rdi, byte SIZEOF_JSAMPROW  ; output_buf
+    dec         rax                        ; num_rows
+    jg          near .rowloop
+
+    sfence                              ; flush the write buffer
+
+.return:
+    pop         rbx
+    uncollect_args 5
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdcolor-avx2.asm Added

@@ -0,0 +1,118 @@
+;
+; jdcolor.asm - colorspace conversion (64-bit AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_ycc_rgb_convert_avx2)
+
+EXTN(jconst_ycc_rgb_convert_avx2):
+
+PW_F0402        times 16 dw  F_0_402
+PW_MF0228       times 16 dw -F_0_228
+PW_MF0344_F0285 times 8  dw -F_0_344, F_0_285
+PW_ONE          times 16 dw  1
+PD_ONEHALF      times 8  dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extrgb_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extrgbx_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extbgr_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extbgrx_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extxbgr_convert_avx2
+%include "jdcolext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_avx2  jsimd_ycc_extxrgb_convert_avx2
+%include "jdcolext-avx2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdcolor-sse2.asm Added

@@ -0,0 +1,117 @@
+;
+; jdcolor.asm - colorspace conversion (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_ycc_rgb_convert_sse2)
+
+EXTN(jconst_ycc_rgb_convert_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extrgb_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extrgbx_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extbgr_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extbgrx_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extxbgr_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2  jsimd_ycc_extxrgb_convert_sse2
+%include "jdcolext-sse2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdmerge-avx2.asm Added

@@ -0,0 +1,136 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (64-bit AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_merged_upsample_avx2)
+
+EXTN(jconst_merged_upsample_avx2):
+
+PW_F0402        times 16 dw  F_0_402
+PW_MF0228       times 16 dw -F_0_228
+PW_MF0344_F0285 times 8  dw -F_0_344, F_0_285
+PW_ONE          times 16 dw  1
+PD_ONEHALF      times 8  dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extrgb_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extrgb_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extrgbx_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extrgbx_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extbgr_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extbgr_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extbgrx_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extbgrx_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extxbgr_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extxbgr_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_avx2 \
+  jsimd_h2v1_extxrgb_merged_upsample_avx2
+%define jsimd_h2v2_merged_upsample_avx2 \
+  jsimd_h2v2_extxrgb_merged_upsample_avx2
+%include "jdmrgext-avx2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdmerge-sse2.asm Added

@@ -0,0 +1,135 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS  16
+
+F_0_344 equ  22554              ; FIX(0.34414)
+F_0_714 equ  46802              ; FIX(0.71414)
+F_1_402 equ  91881              ; FIX(1.40200)
+F_1_772 equ 116130              ; FIX(1.77200)
+F_0_402 equ (F_1_402 - 65536)   ; FIX(1.40200) - FIX(1)
+F_0_285 equ ( 65536 - F_0_714)  ; FIX(1) - FIX(0.71414)
+F_0_228 equ (131072 - F_1_772)  ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_merged_upsample_sse2)
+
+EXTN(jconst_merged_upsample_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGB_RED
+%define RGB_GREEN  EXT_RGB_GREEN
+%define RGB_BLUE  EXT_RGB_BLUE
+%define RGB_PIXELSIZE  EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extrgb_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_RGBX_RED
+%define RGB_GREEN  EXT_RGBX_GREEN
+%define RGB_BLUE  EXT_RGBX_BLUE
+%define RGB_PIXELSIZE  EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extrgbx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extrgbx_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGR_RED
+%define RGB_GREEN  EXT_BGR_GREEN
+%define RGB_BLUE  EXT_BGR_BLUE
+%define RGB_PIXELSIZE  EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extbgr_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_BGRX_RED
+%define RGB_GREEN  EXT_BGRX_GREEN
+%define RGB_BLUE  EXT_BGRX_BLUE
+%define RGB_PIXELSIZE  EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extbgrx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extbgrx_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XBGR_RED
+%define RGB_GREEN  EXT_XBGR_GREEN
+%define RGB_BLUE  EXT_XBGR_BLUE
+%define RGB_PIXELSIZE  EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extxbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extxbgr_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED  EXT_XRGB_RED
+%define RGB_GREEN  EXT_XRGB_GREEN
+%define RGB_BLUE  EXT_XRGB_BLUE
+%define RGB_PIXELSIZE  EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 \
+  jsimd_h2v1_extxrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 \
+  jsimd_h2v2_extxrgb_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdmrgext-avx2.asm Added

@@ -0,0 +1,596 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (64-bit AVX2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2012, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_avx2(JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+; r10d = JDIMENSION output_width
+; r11 = JSAMPIMAGE input_buf
+; r12d = JDIMENSION in_row_group_ctr
+; r13 = JSAMPARRAY output_buf
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_YMMWORD  ; ymmword wkWK_NUM
+%define WK_NUM  3
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_merged_upsample_avx2)
+
+EXTN(jsimd_h2v1_merged_upsample_avx2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 4
+    push        rbx
+
+    mov         ecx, r10d               ; col
+    test        rcx, rcx
+    jz          near .return
+
+    push        rcx
+
+    mov         rdi, r11
+    mov         ecx, r12d
+    mov         rsip, JSAMPARRAY rdi+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rdi+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rdi+2*SIZEOF_JSAMPARRAY
+    mov         rdi, r13
+    mov         rsip, JSAMPROW rsi+rcx*SIZEOF_JSAMPROW  ; inptr0
+    mov         rbxp, JSAMPROW rbx+rcx*SIZEOF_JSAMPROW  ; inptr1
+    mov         rdxp, JSAMPROW rdx+rcx*SIZEOF_JSAMPROW  ; inptr2
+    mov         rdip, JSAMPROW rdi                      ; outptr
+
+    pop         rcx                     ; col
+
+.columnloop:
+
+    vmovdqu     ymm6, YMMWORD rbx     ; ymm6=Cb(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+    vmovdqu     ymm7, YMMWORD rdx     ; ymm7=Cr(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+
+    vpxor       ymm1, ymm1, ymm1        ; ymm1=(all 0's)
+    vpcmpeqw    ymm3, ymm3, ymm3
+    vpsllw      ymm3, ymm3, 7           ; ymm3={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    vpermq      ymm6, ymm6, 0xd8        ; ymm6=Cb(01234567GHIJKLMN89ABCDEFOPQRSTUV)
+    vpermq      ymm7, ymm7, 0xd8        ; ymm7=Cr(01234567GHIJKLMN89ABCDEFOPQRSTUV)
+    vpunpcklbw  ymm4, ymm6, ymm1        ; ymm4=Cb(0123456789ABCDEF)=CbL
+    vpunpckhbw  ymm6, ymm6, ymm1        ; ymm6=Cb(GHIJKLMNOPQRSTUV)=CbH
+    vpunpcklbw  ymm0, ymm7, ymm1        ; ymm0=Cr(0123456789ABCDEF)=CrL
+    vpunpckhbw  ymm7, ymm7, ymm1        ; ymm7=Cr(GHIJKLMNOPQRSTUV)=CrH
+
+    vpaddw      ymm5, ymm6, ymm3
+    vpaddw      ymm2, ymm4, ymm3
+    vpaddw      ymm1, ymm7, ymm3
+    vpaddw      ymm3, ymm0, ymm3
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    vpaddw      ymm6, ymm5, ymm5             ; ymm6=2*CbH
+    vpaddw      ymm4, ymm2, ymm2             ; ymm4=2*CbL
+    vpaddw      ymm7, ymm1, ymm1             ; ymm7=2*CrH
+    vpaddw      ymm0, ymm3, ymm3             ; ymm0=2*CrL
+
+    vpmulhw     ymm6, ymm6, rel PW_MF0228  ; ymm6=(2*CbH * -FIX(0.22800))
+    vpmulhw     ymm4, ymm4, rel PW_MF0228  ; ymm4=(2*CbL * -FIX(0.22800))
+    vpmulhw     ymm7, ymm7, rel PW_F0402   ; ymm7=(2*CrH * FIX(0.40200))
+    vpmulhw     ymm0, ymm0, rel PW_F0402   ; ymm0=(2*CrL * FIX(0.40200))
+
+    vpaddw      ymm6, ymm6, rel PW_ONE
+    vpaddw      ymm4, ymm4, rel PW_ONE
+    vpsraw      ymm6, ymm6, 1                ; ymm6=(CbH * -FIX(0.22800))
+    vpsraw      ymm4, ymm4, 1                ; ymm4=(CbL * -FIX(0.22800))
+    vpaddw      ymm7, ymm7, rel PW_ONE
+    vpaddw      ymm0, ymm0, rel PW_ONE
+    vpsraw      ymm7, ymm7, 1                ; ymm7=(CrH * FIX(0.40200))
+    vpsraw      ymm0, ymm0, 1                ; ymm0=(CrL * FIX(0.40200))
+
+    vpaddw      ymm6, ymm6, ymm5
+    vpaddw      ymm4, ymm4, ymm2
+    vpaddw      ymm6, ymm6, ymm5             ; ymm6=(CbH * FIX(1.77200))=(B-Y)H
+    vpaddw      ymm4, ymm4, ymm2             ; ymm4=(CbL * FIX(1.77200))=(B-Y)L
+    vpaddw      ymm7, ymm7, ymm1             ; ymm7=(CrH * FIX(1.40200))=(R-Y)H
+    vpaddw      ymm0, ymm0, ymm3             ; ymm0=(CrL * FIX(1.40200))=(R-Y)L
+
+    vmovdqa     YMMWORD wk(0), ymm6        ; wk(0)=(B-Y)H
+    vmovdqa     YMMWORD wk(1), ymm7        ; wk(1)=(R-Y)H
+
+    vpunpckhwd  ymm6, ymm5, ymm1
+    vpunpcklwd  ymm5, ymm5, ymm1
+    vpmaddwd    ymm5, ymm5, rel PW_MF0344_F0285
+    vpmaddwd    ymm6, ymm6, rel PW_MF0344_F0285
+    vpunpckhwd  ymm7, ymm2, ymm3
+    vpunpcklwd  ymm2, ymm2, ymm3
+    vpmaddwd    ymm2, ymm2, rel PW_MF0344_F0285
+    vpmaddwd    ymm7, ymm7, rel PW_MF0344_F0285
+
+    vpaddd      ymm5, ymm5, rel PD_ONEHALF
+    vpaddd      ymm6, ymm6, rel PD_ONEHALF
+    vpsrad      ymm5, ymm5, SCALEBITS
+    vpsrad      ymm6, ymm6, SCALEBITS
+    vpaddd      ymm2, ymm2, rel PD_ONEHALF
+    vpaddd      ymm7, ymm7, rel PD_ONEHALF
+    vpsrad      ymm2, ymm2, SCALEBITS
+    vpsrad      ymm7, ymm7, SCALEBITS
+
+    vpackssdw   ymm5, ymm5, ymm6        ; ymm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+    vpackssdw   ymm2, ymm2, ymm7        ; ymm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+    vpsubw      ymm5, ymm5, ymm1        ; ymm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+    vpsubw      ymm2, ymm2, ymm3        ; ymm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+    vmovdqa     YMMWORD wk(2), ymm5   ; wk(2)=(G-Y)H
+
+    mov         al, 2                   ; Yctr
+    jmp         short .Yloop_1st
+
+.Yloop_2nd:
+    vmovdqa     ymm0, YMMWORD wk(1)   ; ymm0=(R-Y)H
+    vmovdqa     ymm2, YMMWORD wk(2)   ; ymm2=(G-Y)H
+    vmovdqa     ymm4, YMMWORD wk(0)   ; ymm4=(B-Y)H
+
+.Yloop_1st:
+    vmovdqu     ymm7, YMMWORD rsi     ; ymm7=Y(0123456789ABCDEFGHIJKLMNOPQRSTUV)
+
+    vpcmpeqw    ymm6, ymm6, ymm6
+    vpsrlw      ymm6, ymm6, BYTE_BIT    ; ymm6={0xFF 0x00 0xFF 0x00 ..}
+    vpand       ymm6, ymm6, ymm7        ; ymm6=Y(02468ACEGIKMOQSU)=YE
+    vpsrlw      ymm7, ymm7, BYTE_BIT    ; ymm7=Y(13579BDFHJLNPRTV)=YO
+
+    vmovdqa     ymm1, ymm0              ; ymm1=ymm0=(R-Y)(L/H)
+    vmovdqa     ymm3, ymm2              ; ymm3=ymm2=(G-Y)(L/H)
+    vmovdqa     ymm5, ymm4              ; ymm5=ymm4=(B-Y)(L/H)
+
+    vpaddw      ymm0, ymm0, ymm6        ; ymm0=((R-Y)+YE)=RE=R(02468ACEGIKMOQSU)
+    vpaddw      ymm1, ymm1, ymm7        ; ymm1=((R-Y)+YO)=RO=R(13579BDFHJLNPRTV)
+    vpackuswb   ymm0, ymm0, ymm0        ; ymm0=R(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm1, ymm1, ymm1        ; ymm1=R(13579BDF********HJLNPRTV********)
+
+    vpaddw      ymm2, ymm2, ymm6        ; ymm2=((G-Y)+YE)=GE=G(02468ACEGIKMOQSU)
+    vpaddw      ymm3, ymm3, ymm7        ; ymm3=((G-Y)+YO)=GO=G(13579BDFHJLNPRTV)
+    vpackuswb   ymm2, ymm2, ymm2        ; ymm2=G(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm3, ymm3, ymm3        ; ymm3=G(13579BDF********HJLNPRTV********)
+
+    vpaddw      ymm4, ymm4, ymm6        ; ymm4=((B-Y)+YE)=BE=B(02468ACEGIKMOQSU)
+    vpaddw      ymm5, ymm5, ymm7        ; ymm5=((B-Y)+YO)=BO=B(13579BDFHJLNPRTV)
+    vpackuswb   ymm4, ymm4, ymm4        ; ymm4=B(02468ACE********GIKMOQSU********)
+    vpackuswb   ymm5, ymm5, ymm5        ; ymm5=B(13579BDF********HJLNPRTV********)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; ymmA=(00 02 04 06 08 0A 0C 0E ** 0G 0I 0K 0M 0O 0Q 0S 0U **)
+    ; ymmB=(01 03 05 07 09 0B 0D 0F ** 0H 0J 0L 0N 0P 0R 0T 0V **)
+    ; ymmC=(10 12 14 16 18 1A 1C 1E ** 1G 1I 1K 1M 1O 1Q 1S 1U **)
+    ; ymmD=(11 13 15 17 19 1B 1D 1F ** 1H 1J 1L 1N 1P 1R 1T 1V **)
+    ; ymmE=(20 22 24 26 28 2A 2C 2E ** 2G 2I 2K 2M 2O 2Q 2S 2U **)
+    ; ymmF=(21 23 25 27 29 2B 2D 2F ** 2H 2J 2L 2N 2P 2R 2T 2V **)
+    ; ymmG=(** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** **)
+    ; ymmH=(** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** **)
+
+    vpunpcklbw  ymmA, ymmA, ymmC        ; ymmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E
+                                        ;       0G 1G 0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U)
+    vpunpcklbw  ymmE, ymmE, ymmB        ; ymmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F
+                                        ;       2G 0H 2I 0J 2K 0L 2M 0N 2O 0P 2Q 0R 2S 0T 2U 0V)
+    vpunpcklbw  ymmD, ymmD, ymmF        ; ymmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F
+                                        ;       1H 2H 1J 2J 1L 2L 1N 2N 1P 2P 1R 2R 1T 2T 1V 2V)
+
+    vpsrldq     ymmH, ymmA, 2           ; ymmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E 0G 1G
+                                        ;       0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U -- --)
+    vpunpckhwd  ymmG, ymmA, ymmE        ; ymmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F
+                                        ;       0O 1O 2O 0P 0Q 1Q 2Q 0R 0S 1S 2S 0T 0U 1U 2U 0V)
+    vpunpcklwd  ymmA, ymmA, ymmE        ; ymmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07
+                                        ;       0G 1G 2G 0H 0I 1I 2I 0J 0K 1K 2K 0L 0M 1M 2M 0N)
+
+    vpsrldq     ymmE, ymmE, 2           ; ymmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F 2G 0H
+                                        ;       2I 0J 2K 0L 2M 0N 2O 0P 2Q 0R 2S 0T 2U 0V -- --)
+
+    vpsrldq     ymmB, ymmD, 2           ; ymmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F 1H 2H
+                                        ;       1J 2J 1L 2L 1N 2N 1P 2P 1R 2R 1T 2T 1V 2V -- --)
+    vpunpckhwd  ymmC, ymmD, ymmH        ; ymmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F 0G 1G
+                                        ;       1P 2P 0Q 1Q 1R 2R 0S 1S 1T 2T 0U 1U 1V 2V -- --)
+    vpunpcklwd  ymmD, ymmD, ymmH        ; ymmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18
+                                        ;       1H 2H 0I 1I 1J 2J 0K 1K 1L 2L 0M 1M 1N 2N 0O 1O)
+
+    vpunpckhwd  ymmF, ymmE, ymmB        ; ymmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F 2G 0H 1H 2H
+                                        ;       2Q 0R 1R 2R 2S 0T 1T 2T 2U 0V 1V 2V -- -- -- --)
+    vpunpcklwd  ymmE, ymmE, ymmB        ; ymmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29
+                                        ;       2I 0J 1J 2J 2K 0L 1L 2L 2M 0N 1N 2N 2O 0P 1P 2P)
+
+    vpshufd     ymmH, ymmA, 0x4E        ; ymmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03
+                                        ;       0K 1K 2K 0L 0M 1M 2M 0N 0G 1G 2G 0H 0I 1I 2I 0J)
+    vpunpckldq  ymmA, ymmA, ymmD        ; ymmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 0I 1I 2I 0J 1J 2J 0K 1K)
+    vpunpckhdq  ymmD, ymmD, ymmE        ; ymmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29
+                                        ;       1L 2L 0M 1M 2M 0N 1N 2N 1N 2N 0O 1O 2O 0P 1P 2P)
+    vpunpckldq  ymmE, ymmE, ymmH        ; ymmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07
+                                        ;       2I 0J 1J 2J 0K 1K 2K 0L 2K 0L 1L 2L 0M 1M 2M 0N)
+
+    vpshufd     ymmH, ymmG, 0x4E        ; ymmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B
+                                        ;       0S 1S 2S 0T 0U 1U 2U 0V 0O 1O 2O 0P 0Q 1Q 2Q 0R)
+    vpunpckldq  ymmG, ymmG, ymmC        ; ymmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C
+                                        ;       0O 1O 2O 0P 1P 2P 0Q 1Q 0Q 1Q 2Q 0R 1R 2R 0S 1S)
+    vpunpckhdq  ymmC, ymmC, ymmF        ; ymmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F 0G 1G 2G 0H 1H 2H
+                                        ;       1T 2T 0U 1U 2U 0V 1V 2V 1V 2V -- -- -- -- -- --)
+    vpunpckldq  ymmF, ymmF, ymmH        ; ymmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 2S 0T 1T 2T 0U 1U 2U 0V)
+
+    vpunpcklqdq ymmH, ymmA, ymmE        ; ymmH=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vpunpcklqdq ymmG, ymmD, ymmG        ; ymmG=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A
+                                        ;       1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q)
+    vpunpcklqdq ymmC, ymmF, ymmC        ; ymmC=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    vperm2i128  ymmA, ymmH, ymmG, 0x20  ; ymmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05
+                                        ;       15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    vperm2i128  ymmD, ymmC, ymmH, 0x30  ; ymmD=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F
+                                        ;       0G 1G 2G 0H 1H 2H 0I 1I 2I 0J 1J 2J 0K 1K 2K 0L)
+    vperm2i128  ymmF, ymmG, ymmC, 0x31  ; ymmF=(1L 2L 0M 1M 2M 0N 1N 2N 0O 1O 2O 0P 1P 2P 0Q 1Q
+                                        ;       2Q 0R 1R 2R 0S 1S 2S 0T 1T 2T 0U 1U 2U 0V 1V 2V)
+
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jb          short .column_st64
+
+    test        rdi, SIZEOF_YMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    vmovntdq    YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovntdq    YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    vmovntdq    YMMWORD rdi+2*SIZEOF_YMMWORD, ymmF
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    vmovdqu     YMMWORD rdi+2*SIZEOF_YMMWORD, ymmF
+.out0:
+    add         rdi, byte RGB_PIXELSIZE*SIZEOF_YMMWORD  ; outptr
+    sub         rcx, byte SIZEOF_YMMWORD
+    jz          near .endcolumn
+
+    add         rsi, byte SIZEOF_YMMWORD  ; inptr0
+    dec         al                        ; Yctr
+    jnz         near .Yloop_2nd
+
+    add         rbx, byte SIZEOF_YMMWORD  ; inptr1
+    add         rdx, byte SIZEOF_YMMWORD  ; inptr2
+    jmp         near .columnloop
+
+.column_st64:
+    lea         rcx, rcx+rcx*2            ; imul ecx, RGB_PIXELSIZE
+    cmp         rcx, byte 2*SIZEOF_YMMWORD
+    jb          short .column_st32
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    add         rdi, byte 2*SIZEOF_YMMWORD  ; outptr
+    vmovdqa     ymmA, ymmF
+    sub         rcx, byte 2*SIZEOF_YMMWORD
+    jmp         short .column_st31
+.column_st32:
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jb          short .column_st31
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    add         rdi, byte SIZEOF_YMMWORD    ; outptr
+    vmovdqa     ymmA, ymmD
+    sub         rcx, byte SIZEOF_YMMWORD
+    jmp         short .column_st31
+.column_st31:
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jb          short .column_st15
+    vmovdqu     XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    add         rdi, byte SIZEOF_XMMWORD    ; outptr
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    sub         rcx, byte SIZEOF_XMMWORD
+.column_st15:
+    ; Store the lower 8 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_MMWORD
+    jb          short .column_st7
+    vmovq       XMM_MMWORD rdi, xmmA
+    add         rdi, byte SIZEOF_MMWORD
+    sub         rcx, byte SIZEOF_MMWORD
+    vpsrldq     xmmA, xmmA, SIZEOF_MMWORD
+.column_st7:
+    ; Store the lower 4 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_DWORD
+    jb          short .column_st3
+    vmovd       XMM_DWORD rdi, xmmA
+    add         rdi, byte SIZEOF_DWORD
+    sub         rcx, byte SIZEOF_DWORD
+    vpsrldq     xmmA, xmmA, SIZEOF_DWORD
+.column_st3:
+    ; Store the lower 2 bytes of rax to the output when it has enough
+    ; space.
+    vmovd       eax, xmmA
+    cmp         rcx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word rdi, ax
+    add         rdi, byte SIZEOF_WORD
+    sub         rcx, byte SIZEOF_WORD
+    shr         rax, 16
+.column_st1:
+    ; Store the lower 1 byte of rax to the output when it has enough
+    ; space.
+    test        rcx, rcx
+    jz          short .endcolumn
+    mov         byte rdi, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    vpcmpeqb    ymm6, ymm6, ymm6        ; ymm6=XE=X(02468ACE********GIKMOQSU********)
+    vpcmpeqb    ymm7, ymm7, ymm7        ; ymm7=XO=X(13579BDF********HJLNPRTV********)
+%else
+    vpxor       ymm6, ymm6, ymm6        ; ymm6=XE=X(02468ACE********GIKMOQSU********)
+    vpxor       ymm7, ymm7, ymm7        ; ymm7=XO=X(13579BDF********HJLNPRTV********)
+%endif
+    ; ymmA=(00 02 04 06 08 0A 0C 0E ** 0G 0I 0K 0M 0O 0Q 0S 0U **)
+    ; ymmB=(01 03 05 07 09 0B 0D 0F ** 0H 0J 0L 0N 0P 0R 0T 0V **)
+    ; ymmC=(10 12 14 16 18 1A 1C 1E ** 1G 1I 1K 1M 1O 1Q 1S 1U **)
+    ; ymmD=(11 13 15 17 19 1B 1D 1F ** 1H 1J 1L 1N 1P 1R 1T 1V **)
+    ; ymmE=(20 22 24 26 28 2A 2C 2E ** 2G 2I 2K 2M 2O 2Q 2S 2U **)
+    ; ymmF=(21 23 25 27 29 2B 2D 2F ** 2H 2J 2L 2N 2P 2R 2T 2V **)
+    ; ymmG=(30 32 34 36 38 3A 3C 3E ** 3G 3I 3K 3M 3O 3Q 3S 3U **)
+    ; ymmH=(31 33 35 37 39 3B 3D 3F ** 3H 3J 3L 3N 3P 3R 3T 3V **)
+
+    vpunpcklbw  ymmA, ymmA, ymmC        ; ymmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E
+                                        ;       0G 1G 0I 1I 0K 1K 0M 1M 0O 1O 0Q 1Q 0S 1S 0U 1U)
+    vpunpcklbw  ymmE, ymmE, ymmG        ; ymmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E
+                                        ;       2G 3G 2I 3I 2K 3K 2M 3M 2O 3O 2Q 3Q 2S 3S 2U 3U)
+    vpunpcklbw  ymmB, ymmB, ymmD        ; ymmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F
+                                        ;       0H 1H 0J 1J 0L 1L 0N 1N 0P 1P 0R 1R 0T 1T 0V 1V)
+    vpunpcklbw  ymmF, ymmF, ymmH        ; ymmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F
+                                        ;       2H 3H 2J 3J 2L 3L 2N 3N 2P 3P 2R 3R 2T 3T 2V 3V)
+
+    vpunpckhwd  ymmC, ymmA, ymmE        ; ymmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E
+                                        ;       0O 1O 2O 3O 0Q 1Q 2Q 3Q 0S 1S 2S 3S 0U 1U 2U 3U)
+    vpunpcklwd  ymmA, ymmA, ymmE        ; ymmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36
+                                        ;       0G 1G 2G 3G 0I 1I 2I 3I 0K 1K 2K 3K 0M 1M 2M 3M)
+    vpunpckhwd  ymmG, ymmB, ymmF        ; ymmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F
+                                        ;       0P 1P 2P 3P 0R 1R 2R 3R 0T 1T 2T 3T 0V 1V 2V 3V)
+    vpunpcklwd  ymmB, ymmB, ymmF        ; ymmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37
+                                        ;       0H 1H 2H 3H 0J 1J 2J 3J 0L 1L 2L 3L 0N 1N 2N 3N)
+
+    vpunpckhdq  ymmE, ymmA, ymmB        ; ymmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    vpunpckldq  ymmB, ymmA, ymmB        ; ymmB=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J)
+    vpunpckhdq  ymmF, ymmC, ymmG        ; ymmF=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+    vpunpckldq  ymmG, ymmC, ymmG        ; ymmG=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R)
+
+    vperm2i128  ymmA, ymmB, ymmE, 0x20  ; ymmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
+                                        ;       04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    vperm2i128  ymmD, ymmG, ymmF, 0x20  ; ymmD=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B
+                                        ;       0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+    vperm2i128  ymmC, ymmB, ymmE, 0x31  ; ymmC=(0G 1G 2G 3G 0H 1H 2H 3H 0I 1I 2I 3I 0J 1J 2J 3J
+                                        ;       0K 1K 2K 3K 0L 1L 2L 3L 0M 1M 2M 3M 0N 1N 2N 3N)
+    vperm2i128  ymmH, ymmG, ymmF, 0x31  ; ymmH=(0O 1O 2O 3O 0P 1P 2P 3P 0Q 1Q 2Q 3Q 0R 1R 2R 3R
+                                        ;       0S 1S 2S 3S 0T 1T 2T 3T 0U 1U 2U 3U 0V 1V 2V 3V)
+
+    cmp         rcx, byte SIZEOF_YMMWORD
+    jb          short .column_st64
+
+    test        rdi, SIZEOF_YMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    vmovntdq    YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovntdq    YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    vmovntdq    YMMWORD rdi+2*SIZEOF_YMMWORD, ymmC
+    vmovntdq    YMMWORD rdi+3*SIZEOF_YMMWORD, ymmH
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    vmovdqu     YMMWORD rdi+2*SIZEOF_YMMWORD, ymmC
+    vmovdqu     YMMWORD rdi+3*SIZEOF_YMMWORD, ymmH
+.out0:
+    add         rdi, RGB_PIXELSIZE*SIZEOF_YMMWORD  ; outptr
+    sub         rcx, byte SIZEOF_YMMWORD
+    jz          near .endcolumn
+
+    add         rsi, byte SIZEOF_YMMWORD  ; inptr0
+    dec         al
+    jnz         near .Yloop_2nd
+
+    add         rbx, byte SIZEOF_YMMWORD  ; inptr1
+    add         rdx, byte SIZEOF_YMMWORD  ; inptr2
+    jmp         near .columnloop
+
+.column_st64:
+    cmp         rcx, byte SIZEOF_YMMWORD/2
+    jb          short .column_st32
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymmD
+    add         rdi, byte 2*SIZEOF_YMMWORD  ; outptr
+    vmovdqa     ymmA, ymmC
+    vmovdqa     ymmD, ymmH
+    sub         rcx, byte SIZEOF_YMMWORD/2
+.column_st32:
+    cmp         rcx, byte SIZEOF_YMMWORD/4
+    jb          short .column_st16
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymmA
+    add         rdi, byte SIZEOF_YMMWORD    ; outptr
+    vmovdqa     ymmA, ymmD
+    sub         rcx, byte SIZEOF_YMMWORD/4
+.column_st16:
+    cmp         rcx, byte SIZEOF_YMMWORD/8
+    jb          short .column_st15
+    vmovdqu     XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    add         rdi, byte SIZEOF_XMMWORD    ; outptr
+    vperm2i128  ymmA, ymmA, ymmA, 1
+    sub         rcx, byte SIZEOF_YMMWORD/8
+.column_st15:
+    ; Store two pixels (8 bytes) of ymmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_YMMWORD/16
+    jb          short .column_st7
+    vmovq       MMWORD rdi, xmmA
+    add         rdi, byte SIZEOF_YMMWORD/16*4
+    sub         rcx, byte SIZEOF_YMMWORD/16
+    vpsrldq     xmmA, SIZEOF_YMMWORD/16*4
+.column_st7:
+    ; Store one pixel (4 bytes) of ymmA to the output when it has enough
+    ; space.
+    test        rcx, rcx
+    jz          short .endcolumn
+    vmovd       XMM_DWORD rdi, xmmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+    sfence                              ; flush the write buffer
+
+.return:
+    pop         rbx
+    vzeroupper
+    uncollect_args 4
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_avx2(JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+; r10d = JDIMENSION output_width
+; r11 = JSAMPIMAGE input_buf
+; r12d = JDIMENSION in_row_group_ctr
+; r13 = JSAMPARRAY output_buf
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_merged_upsample_avx2)
+
+EXTN(jsimd_h2v2_merged_upsample_avx2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 4
+    push        rbx
+
+    mov         eax, r10d
+
+    mov         rdi, r11
+    mov         ecx, r12d
+    mov         rsip, JSAMPARRAY rdi+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rdi+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rdi+2*SIZEOF_JSAMPARRAY
+    mov         rdi, r13
+    lea         rsi, rsi+rcx*SIZEOF_JSAMPROW
+
+    sub         rsp, SIZEOF_JSAMPARRAY*4
+    mov         JSAMPARRAY rsp+0*SIZEOF_JSAMPARRAY, rsip  ; intpr00
+    mov         JSAMPARRAY rsp+1*SIZEOF_JSAMPARRAY, rbxp  ; intpr1
+    mov         JSAMPARRAY rsp+2*SIZEOF_JSAMPARRAY, rdxp  ; intpr2
+    mov         rbx, rsp
+
+    push        rdi
+    push        rcx
+    push        rax
+
+    %ifdef WIN64
+    mov         r8, rcx
+    mov         r9, rdi
+    mov         rcx, rax
+    mov         rdx, rbx
+    %else
+    mov         rdx, rcx
+    mov         rcx, rdi
+    mov         rdi, rax
+    mov         rsi, rbx
+    %endif
+
+    call        EXTN(jsimd_h2v1_merged_upsample_avx2)
+
+    pop         rax
+    pop         rcx
+    pop         rdi
+    mov         rsip, JSAMPARRAY rsp+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rsp+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rsp+2*SIZEOF_JSAMPARRAY
+
+    add         rdi, byte SIZEOF_JSAMPROW  ; outptr1
+    add         rsi, byte SIZEOF_JSAMPROW  ; inptr01
+
+    mov         JSAMPARRAY rsp+0*SIZEOF_JSAMPARRAY, rsip  ; intpr00
+    mov         JSAMPARRAY rsp+1*SIZEOF_JSAMPARRAY, rbxp  ; intpr1
+    mov         JSAMPARRAY rsp+2*SIZEOF_JSAMPARRAY, rdxp  ; intpr2
+    mov         rbx, rsp
+
+    push        rdi
+    push        rcx
+    push        rax
+
+    %ifdef WIN64
+    mov         r8, rcx
+    mov         r9, rdi
+    mov         rcx, rax
+    mov         rdx, rbx
+    %else
+    mov         rdx, rcx
+    mov         rcx, rdi
+    mov         rdi, rax
+    mov         rsi, rbx
+    %endif
+
+    call        EXTN(jsimd_h2v1_merged_upsample_avx2)
+
+    pop         rax
+    pop         rcx
+    pop         rdi
+    mov         rsip, JSAMPARRAY rsp+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rsp+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rsp+2*SIZEOF_JSAMPARRAY
+    add         rsp, SIZEOF_JSAMPARRAY*4
+
+    pop         rbx
+    uncollect_args 4
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdmrgext-sse2.asm Added

@@ -0,0 +1,538 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (64-bit SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2012, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_sse2(JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+; r10d = JDIMENSION output_width
+; r11 = JSAMPIMAGE input_buf
+; r12d = JDIMENSION in_row_group_ctr
+; r13 = JSAMPARRAY output_buf
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD  ; xmmword wkWK_NUM
+%define WK_NUM  3
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_merged_upsample_sse2)
+
+EXTN(jsimd_h2v1_merged_upsample_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 4
+    push        rbx
+
+    mov         ecx, r10d               ; col
+    test        rcx, rcx
+    jz          near .return
+
+    push        rcx
+
+    mov         rdi, r11
+    mov         ecx, r12d
+    mov         rsip, JSAMPARRAY rdi+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rdi+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rdi+2*SIZEOF_JSAMPARRAY
+    mov         rdi, r13
+    mov         rsip, JSAMPROW rsi+rcx*SIZEOF_JSAMPROW  ; inptr0
+    mov         rbxp, JSAMPROW rbx+rcx*SIZEOF_JSAMPROW  ; inptr1
+    mov         rdxp, JSAMPROW rdx+rcx*SIZEOF_JSAMPROW  ; inptr2
+    mov         rdip, JSAMPROW rdi                      ; outptr
+
+    pop         rcx                     ; col
+
+.columnloop:
+
+    movdqa      xmm6, XMMWORD rbx     ; xmm6=Cb(0123456789ABCDEF)
+    movdqa      xmm7, XMMWORD rdx     ; xmm7=Cr(0123456789ABCDEF)
+
+    pxor        xmm1, xmm1              ; xmm1=(all 0's)
+    pcmpeqw     xmm3, xmm3
+    psllw       xmm3, 7                 ; xmm3={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    movdqa      xmm4, xmm6
+    punpckhbw   xmm6, xmm1              ; xmm6=Cb(89ABCDEF)=CbH
+    punpcklbw   xmm4, xmm1              ; xmm4=Cb(01234567)=CbL
+    movdqa      xmm0, xmm7
+    punpckhbw   xmm7, xmm1              ; xmm7=Cr(89ABCDEF)=CrH
+    punpcklbw   xmm0, xmm1              ; xmm0=Cr(01234567)=CrL
+
+    paddw       xmm6, xmm3
+    paddw       xmm4, xmm3
+    paddw       xmm7, xmm3
+    paddw       xmm0, xmm3
+
+    ; (Original)
+    ; R = Y                + 1.40200 * Cr
+    ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+    ; B = Y + 1.77200 * Cb
+    ;
+    ; (This implementation)
+    ; R = Y                + 0.40200 * Cr + Cr
+    ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+    ; B = Y - 0.22800 * Cb + Cb + Cb
+
+    movdqa      xmm5, xmm6              ; xmm5=CbH
+    movdqa      xmm2, xmm4              ; xmm2=CbL
+    paddw       xmm6, xmm6              ; xmm6=2*CbH
+    paddw       xmm4, xmm4              ; xmm4=2*CbL
+    movdqa      xmm1, xmm7              ; xmm1=CrH
+    movdqa      xmm3, xmm0              ; xmm3=CrL
+    paddw       xmm7, xmm7              ; xmm7=2*CrH
+    paddw       xmm0, xmm0              ; xmm0=2*CrL
+
+    pmulhw      xmm6, rel PW_MF0228   ; xmm6=(2*CbH * -FIX(0.22800))
+    pmulhw      xmm4, rel PW_MF0228   ; xmm4=(2*CbL * -FIX(0.22800))
+    pmulhw      xmm7, rel PW_F0402    ; xmm7=(2*CrH * FIX(0.40200))
+    pmulhw      xmm0, rel PW_F0402    ; xmm0=(2*CrL * FIX(0.40200))
+
+    paddw       xmm6, rel PW_ONE
+    paddw       xmm4, rel PW_ONE
+    psraw       xmm6, 1                 ; xmm6=(CbH * -FIX(0.22800))
+    psraw       xmm4, 1                 ; xmm4=(CbL * -FIX(0.22800))
+    paddw       xmm7, rel PW_ONE
+    paddw       xmm0, rel PW_ONE
+    psraw       xmm7, 1                 ; xmm7=(CrH * FIX(0.40200))
+    psraw       xmm0, 1                 ; xmm0=(CrL * FIX(0.40200))
+
+    paddw       xmm6, xmm5
+    paddw       xmm4, xmm2
+    paddw       xmm6, xmm5              ; xmm6=(CbH * FIX(1.77200))=(B-Y)H
+    paddw       xmm4, xmm2              ; xmm4=(CbL * FIX(1.77200))=(B-Y)L
+    paddw       xmm7, xmm1              ; xmm7=(CrH * FIX(1.40200))=(R-Y)H
+    paddw       xmm0, xmm3              ; xmm0=(CrL * FIX(1.40200))=(R-Y)L
+
+    movdqa      XMMWORD wk(0), xmm6   ; wk(0)=(B-Y)H
+    movdqa      XMMWORD wk(1), xmm7   ; wk(1)=(R-Y)H
+
+    movdqa      xmm6, xmm5
+    movdqa      xmm7, xmm2
+    punpcklwd   xmm5, xmm1
+    punpckhwd   xmm6, xmm1
+    pmaddwd     xmm5, rel PW_MF0344_F0285
+    pmaddwd     xmm6, rel PW_MF0344_F0285
+    punpcklwd   xmm2, xmm3
+    punpckhwd   xmm7, xmm3
+    pmaddwd     xmm2, rel PW_MF0344_F0285
+    pmaddwd     xmm7, rel PW_MF0344_F0285
+
+    paddd       xmm5, rel PD_ONEHALF
+    paddd       xmm6, rel PD_ONEHALF
+    psrad       xmm5, SCALEBITS
+    psrad       xmm6, SCALEBITS
+    paddd       xmm2, rel PD_ONEHALF
+    paddd       xmm7, rel PD_ONEHALF
+    psrad       xmm2, SCALEBITS
+    psrad       xmm7, SCALEBITS
+
+    packssdw    xmm5, xmm6              ; xmm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+    packssdw    xmm2, xmm7              ; xmm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+    psubw       xmm5, xmm1              ; xmm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+    psubw       xmm2, xmm3              ; xmm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+    movdqa      XMMWORD wk(2), xmm5   ; wk(2)=(G-Y)H
+
+    mov         al, 2                   ; Yctr
+    jmp         short .Yloop_1st
+
+.Yloop_2nd:
+    movdqa      xmm0, XMMWORD wk(1)   ; xmm0=(R-Y)H
+    movdqa      xmm2, XMMWORD wk(2)   ; xmm2=(G-Y)H
+    movdqa      xmm4, XMMWORD wk(0)   ; xmm4=(B-Y)H
+
+.Yloop_1st:
+    movdqa      xmm7, XMMWORD rsi     ; xmm7=Y(0123456789ABCDEF)
+
+    pcmpeqw     xmm6, xmm6
+    psrlw       xmm6, BYTE_BIT          ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+    pand        xmm6, xmm7              ; xmm6=Y(02468ACE)=YE
+    psrlw       xmm7, BYTE_BIT          ; xmm7=Y(13579BDF)=YO
+
+    movdqa      xmm1, xmm0              ; xmm1=xmm0=(R-Y)(L/H)
+    movdqa      xmm3, xmm2              ; xmm3=xmm2=(G-Y)(L/H)
+    movdqa      xmm5, xmm4              ; xmm5=xmm4=(B-Y)(L/H)
+
+    paddw       xmm0, xmm6              ; xmm0=((R-Y)+YE)=RE=R(02468ACE)
+    paddw       xmm1, xmm7              ; xmm1=((R-Y)+YO)=RO=R(13579BDF)
+    packuswb    xmm0, xmm0              ; xmm0=R(02468ACE********)
+    packuswb    xmm1, xmm1              ; xmm1=R(13579BDF********)
+
+    paddw       xmm2, xmm6              ; xmm2=((G-Y)+YE)=GE=G(02468ACE)
+    paddw       xmm3, xmm7              ; xmm3=((G-Y)+YO)=GO=G(13579BDF)
+    packuswb    xmm2, xmm2              ; xmm2=G(02468ACE********)
+    packuswb    xmm3, xmm3              ; xmm3=G(13579BDF********)
+
+    paddw       xmm4, xmm6              ; xmm4=((B-Y)+YE)=BE=B(02468ACE)
+    paddw       xmm5, xmm7              ; xmm5=((B-Y)+YO)=BO=B(13579BDF)
+    packuswb    xmm4, xmm4              ; xmm4=B(02468ACE********)
+    packuswb    xmm5, xmm5              ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3  ; ---------------
+
+    ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+    ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+    ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+    ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+    punpcklbw   xmmA, xmmC        ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+    punpcklbw   xmmE, xmmB        ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+    punpcklbw   xmmD, xmmF        ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+    movdqa      xmmG, xmmA
+    movdqa      xmmH, xmmA
+    punpcklwd   xmmA, xmmE        ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+    punpckhwd   xmmG, xmmE        ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+    psrldq      xmmH, 2           ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+    psrldq      xmmE, 2           ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+    movdqa      xmmC, xmmD
+    movdqa      xmmB, xmmD
+    punpcklwd   xmmD, xmmH        ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+    punpckhwd   xmmC, xmmH        ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+    psrldq      xmmB, 2           ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+    movdqa      xmmF, xmmE
+    punpcklwd   xmmE, xmmB        ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+    punpckhwd   xmmF, xmmB        ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+    pshufd      xmmH, xmmA, 0x4E  ; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+    movdqa      xmmB, xmmE
+    punpckldq   xmmA, xmmD        ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+    punpckldq   xmmE, xmmH        ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+    punpckhdq   xmmD, xmmB        ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+    pshufd      xmmH, xmmG, 0x4E  ; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+    movdqa      xmmB, xmmF
+    punpckldq   xmmG, xmmC        ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+    punpckldq   xmmF, xmmH        ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+    punpckhdq   xmmC, xmmB        ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+    punpcklqdq  xmmA, xmmE        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+    punpcklqdq  xmmD, xmmG        ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+    punpcklqdq  xmmF, xmmC        ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jb          short .column_st32
+
+    test        rdi, SIZEOF_XMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    movntdq     XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movntdq     XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    movntdq     XMMWORD rdi+2*SIZEOF_XMMWORD, xmmF
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    movdqu      XMMWORD rdi+2*SIZEOF_XMMWORD, xmmF
+.out0:
+    add         rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+    sub         rcx, byte SIZEOF_XMMWORD
+    jz          near .endcolumn
+
+    add         rsi, byte SIZEOF_XMMWORD  ; inptr0
+    dec         al                        ; Yctr
+    jnz         near .Yloop_2nd
+
+    add         rbx, byte SIZEOF_XMMWORD  ; inptr1
+    add         rdx, byte SIZEOF_XMMWORD  ; inptr2
+    jmp         near .columnloop
+
+.column_st32:
+    lea         rcx, rcx+rcx*2            ; imul ecx, RGB_PIXELSIZE
+    cmp         rcx, byte 2*SIZEOF_XMMWORD
+    jb          short .column_st16
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    add         rdi, byte 2*SIZEOF_XMMWORD  ; outptr
+    movdqa      xmmA, xmmF
+    sub         rcx, byte 2*SIZEOF_XMMWORD
+    jmp         short .column_st15
+.column_st16:
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jb          short .column_st15
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    add         rdi, byte SIZEOF_XMMWORD    ; outptr
+    movdqa      xmmA, xmmD
+    sub         rcx, byte SIZEOF_XMMWORD
+.column_st15:
+    ; Store the lower 8 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_MMWORD
+    jb          short .column_st7
+    movq        XMM_MMWORD rdi, xmmA
+    add         rdi, byte SIZEOF_MMWORD
+    sub         rcx, byte SIZEOF_MMWORD
+    psrldq      xmmA, SIZEOF_MMWORD
+.column_st7:
+    ; Store the lower 4 bytes of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_DWORD
+    jb          short .column_st3
+    movd        XMM_DWORD rdi, xmmA
+    add         rdi, byte SIZEOF_DWORD
+    sub         rcx, byte SIZEOF_DWORD
+    psrldq      xmmA, SIZEOF_DWORD
+.column_st3:
+    ; Store the lower 2 bytes of rax to the output when it has enough
+    ; space.
+    movd        eax, xmmA
+    cmp         rcx, byte SIZEOF_WORD
+    jb          short .column_st1
+    mov         word rdi, ax
+    add         rdi, byte SIZEOF_WORD
+    sub         rcx, byte SIZEOF_WORD
+    shr         rax, 16
+.column_st1:
+    ; Store the lower 1 byte of rax to the output when it has enough
+    ; space.
+    test        rcx, rcx
+    jz          short .endcolumn
+    mov         byte rdi, al
+
+%else  ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+    pcmpeqb     xmm6, xmm6              ; xmm6=XE=X(02468ACE********)
+    pcmpeqb     xmm7, xmm7              ; xmm7=XO=X(13579BDF********)
+%else
+    pxor        xmm6, xmm6              ; xmm6=XE=X(02468ACE********)
+    pxor        xmm7, xmm7              ; xmm7=XO=X(13579BDF********)
+%endif
+    ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+    ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+    ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+    ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+    punpcklbw   xmmA, xmmC  ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+    punpcklbw   xmmE, xmmG  ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+    punpcklbw   xmmB, xmmD  ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+    punpcklbw   xmmF, xmmH  ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+    movdqa      xmmC, xmmA
+    punpcklwd   xmmA, xmmE  ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+    punpckhwd   xmmC, xmmE  ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+    movdqa      xmmG, xmmB
+    punpcklwd   xmmB, xmmF  ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+    punpckhwd   xmmG, xmmF  ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+    movdqa      xmmD, xmmA
+    punpckldq   xmmA, xmmB  ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+    punpckhdq   xmmD, xmmB  ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+    movdqa      xmmH, xmmC
+    punpckldq   xmmC, xmmG  ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+    punpckhdq   xmmH, xmmG  ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+    cmp         rcx, byte SIZEOF_XMMWORD
+    jb          short .column_st32
+
+    test        rdi, SIZEOF_XMMWORD-1
+    jnz         short .out1
+    ; --(aligned)-------------------
+    movntdq     XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movntdq     XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    movntdq     XMMWORD rdi+2*SIZEOF_XMMWORD, xmmC
+    movntdq     XMMWORD rdi+3*SIZEOF_XMMWORD, xmmH
+    jmp         short .out0
+.out1:  ; --(unaligned)-----------------
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    movdqu      XMMWORD rdi+2*SIZEOF_XMMWORD, xmmC
+    movdqu      XMMWORD rdi+3*SIZEOF_XMMWORD, xmmH
+.out0:
+    add         rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+    sub         rcx, byte SIZEOF_XMMWORD
+    jz          near .endcolumn
+
+    add         rsi, byte SIZEOF_XMMWORD  ; inptr0
+    dec         al                        ; Yctr
+    jnz         near .Yloop_2nd
+
+    add         rbx, byte SIZEOF_XMMWORD  ; inptr1
+    add         rdx, byte SIZEOF_XMMWORD  ; inptr2
+    jmp         near .columnloop
+
+.column_st32:
+    cmp         rcx, byte SIZEOF_XMMWORD/2
+    jb          short .column_st16
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    movdqu      XMMWORD rdi+1*SIZEOF_XMMWORD, xmmD
+    add         rdi, byte 2*SIZEOF_XMMWORD  ; outptr
+    movdqa      xmmA, xmmC
+    movdqa      xmmD, xmmH
+    sub         rcx, byte SIZEOF_XMMWORD/2
+.column_st16:
+    cmp         rcx, byte SIZEOF_XMMWORD/4
+    jb          short .column_st15
+    movdqu      XMMWORD rdi+0*SIZEOF_XMMWORD, xmmA
+    add         rdi, byte SIZEOF_XMMWORD    ; outptr
+    movdqa      xmmA, xmmD
+    sub         rcx, byte SIZEOF_XMMWORD/4
+.column_st15:
+    ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+    ; space.
+    cmp         rcx, byte SIZEOF_XMMWORD/8
+    jb          short .column_st7
+    movq        XMM_MMWORD rdi, xmmA
+    add         rdi, byte SIZEOF_XMMWORD/8*4
+    sub         rcx, byte SIZEOF_XMMWORD/8
+    psrldq      xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+    ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+    ; space.
+    test        rcx, rcx
+    jz          short .endcolumn
+    movd        XMM_DWORD rdi, xmmA
+
+%endif  ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+    sfence                              ; flush the write buffer
+
+.return:
+    pop         rbx
+    uncollect_args 4
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_sse2(JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+; r10d = JDIMENSION output_width
+; r11 = JSAMPIMAGE input_buf
+; r12d = JDIMENSION in_row_group_ctr
+; r13 = JSAMPARRAY output_buf
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_merged_upsample_sse2)
+
+EXTN(jsimd_h2v2_merged_upsample_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 4
+    push        rbx
+
+    mov         eax, r10d
+
+    mov         rdi, r11
+    mov         ecx, r12d
+    mov         rsip, JSAMPARRAY rdi+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rdi+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rdi+2*SIZEOF_JSAMPARRAY
+    mov         rdi, r13
+    lea         rsi, rsi+rcx*SIZEOF_JSAMPROW
+
+    sub         rsp, SIZEOF_JSAMPARRAY*4
+    mov         JSAMPARRAY rsp+0*SIZEOF_JSAMPARRAY, rsip  ; intpr00
+    mov         JSAMPARRAY rsp+1*SIZEOF_JSAMPARRAY, rbxp  ; intpr1
+    mov         JSAMPARRAY rsp+2*SIZEOF_JSAMPARRAY, rdxp  ; intpr2
+    mov         rbx, rsp
+
+    push        rdi
+    push        rcx
+    push        rax
+
+    %ifdef WIN64
+    mov         r8, rcx
+    mov         r9, rdi
+    mov         rcx, rax
+    mov         rdx, rbx
+    %else
+    mov         rdx, rcx
+    mov         rcx, rdi
+    mov         rdi, rax
+    mov         rsi, rbx
+    %endif
+
+    call        EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+    pop         rax
+    pop         rcx
+    pop         rdi
+    mov         rsip, JSAMPARRAY rsp+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rsp+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rsp+2*SIZEOF_JSAMPARRAY
+
+    add         rdi, byte SIZEOF_JSAMPROW  ; outptr1
+    add         rsi, byte SIZEOF_JSAMPROW  ; inptr01
+
+    mov         JSAMPARRAY rsp+0*SIZEOF_JSAMPARRAY, rsip  ; intpr00
+    mov         JSAMPARRAY rsp+1*SIZEOF_JSAMPARRAY, rbxp  ; intpr1
+    mov         JSAMPARRAY rsp+2*SIZEOF_JSAMPARRAY, rdxp  ; intpr2
+    mov         rbx, rsp
+
+    push        rdi
+    push        rcx
+    push        rax
+
+    %ifdef WIN64
+    mov         r8, rcx
+    mov         r9, rdi
+    mov         rcx, rax
+    mov         rdx, rbx
+    %else
+    mov         rdx, rcx
+    mov         rcx, rdi
+    mov         rdi, rax
+    mov         rsi, rbx
+    %endif
+
+    call        EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+    pop         rax
+    pop         rcx
+    pop         rdi
+    mov         rsip, JSAMPARRAY rsp+0*SIZEOF_JSAMPARRAY
+    mov         rbxp, JSAMPARRAY rsp+1*SIZEOF_JSAMPARRAY
+    mov         rdxp, JSAMPARRAY rsp+2*SIZEOF_JSAMPARRAY
+    add         rsp, SIZEOF_JSAMPARRAY*4
+
+    pop         rbx
+    uncollect_args 4
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdsample-avx2.asm Added

@@ -0,0 +1,696 @@
+;
+; jdsample.asm - upsampling (64-bit AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2015, Intel Corporation.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fancy_upsample_avx2)
+
+EXTN(jconst_fancy_upsample_avx2):
+
+PW_ONE   times 16 dw 1
+PW_TWO   times 16 dw 2
+PW_THREE times 16 dw 3
+PW_SEVEN times 16 dw 7
+PW_EIGHT times 16 dw 8
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_avx2(int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11d = JDIMENSION downsampled_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_fancy_upsample_avx2)
+
+EXTN(jsimd_h2v1_fancy_upsample_avx2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    push_xmm    3
+    collect_args 4
+
+    mov         eax, r11d               ; colctr
+    test        rax, rax
+    jz          near .return
+
+    mov         rcx, r10                ; rowctr
+    test        rcx, rcx
+    jz          near .return
+
+    mov         rsi, r12                ; input_data
+    mov         rdi, r13
+    mov         rdip, JSAMPARRAY rdi  ; output_data
+
+    vpxor       ymm0, ymm0, ymm0                 ; ymm0=(all 0's)
+    vpcmpeqb    xmm9, xmm9, xmm9
+    vpsrldq     xmm10, xmm9, (SIZEOF_XMMWORD-1)  ; (ff -- -- -- ... -- --) LSB is ff
+
+    vpslldq     xmm9, xmm9, (SIZEOF_XMMWORD-1)
+    vperm2i128  ymm9, ymm9, ymm9, 1              ; (---- ---- ... ---- ---- ff) MSB is ff
+
+.rowloop:
+    push        rax                     ; colctr
+    push        rdi
+    push        rsi
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr
+
+    test        rax, SIZEOF_YMMWORD-1
+    jz          short .skip
+    mov         dl, JSAMPLE rsi+(rax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE rsi+rax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+.skip:
+    vpand       ymm7, ymm10, YMMWORD rsi+0*SIZEOF_YMMWORD
+
+    add         rax, byte SIZEOF_YMMWORD-1
+    and         rax, byte -SIZEOF_YMMWORD
+    cmp         rax, byte SIZEOF_YMMWORD
+    ja          short .columnloop
+
+.columnloop_last:
+    vpand       ymm6, ymm9, YMMWORD rsi+0*SIZEOF_YMMWORD
+    jmp         short .upsample
+
+.columnloop:
+    vmovdqu     ymm6, YMMWORD rsi+1*SIZEOF_YMMWORD
+    vperm2i128  ymm6, ymm0, ymm6, 0x20
+    vpslldq     ymm6, ymm6, 15
+
+.upsample:
+    vmovdqu     ymm1, YMMWORD rsi+0*SIZEOF_YMMWORD  ; ymm1=( 0  1  2 ... 29 30 31)
+
+    vperm2i128  ymm2, ymm0, ymm1, 0x20
+    vpalignr    ymm2, ymm1, ymm2, 15            ; ymm2=(--  0  1 ... 28 29 30)
+    vperm2i128  ymm4, ymm0, ymm1, 0x03
+    vpalignr    ymm3, ymm4, ymm1, 1             ; ymm3=( 1  2  3 ... 30 31 --)
+
+    vpor        ymm2, ymm2, ymm7                ; ymm2=(-1  0  1 ... 28 29 30)
+    vpor        ymm3, ymm3, ymm6                ; ymm3=( 1  2  3 ... 30 31 32)
+
+    vpsrldq     ymm7, ymm4, (SIZEOF_XMMWORD-1)  ; ymm7=(31 -- -- ... -- -- --)
+
+    vpunpckhbw  ymm4, ymm1, ymm0                ; ymm4=( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm5, ymm1, ymm0                ; ymm5=( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm1, ymm5, ymm4, 0x20          ; ymm1=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm4, ymm5, ymm4, 0x31          ; ymm4=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm5, ymm2, ymm0                ; ymm5=( 7  8  9 10 11 12 13 14 23 24 25 26 27 28 29 30)
+    vpunpcklbw  ymm6, ymm2, ymm0                ; ymm6=(-1  0  1  2  3  4  5  6 15 16 17 18 19 20 21 22)
+    vperm2i128  ymm2, ymm6, ymm5, 0x20          ; ymm2=(-1  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+    vperm2i128  ymm5, ymm6, ymm5, 0x31          ; ymm5=(15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vpunpckhbw  ymm6, ymm3, ymm0                ; ymm6=( 1  2  3  4  5  6  7  8 17 18 19 20 21 22 23 24)
+    vpunpcklbw  ymm8, ymm3, ymm0                ; ymm8=( 9 10 11 12 13 14 15 16 25 26 27 28 29 30 31 32)
+    vperm2i128  ymm3, ymm8, ymm6, 0x20          ; ymm3=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16)
+    vperm2i128  ymm6, ymm8, ymm6, 0x31          ; ymm6=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32)
+
+    vpmullw     ymm1, ymm1, rel PW_THREE
+    vpmullw     ymm4, ymm4, rel PW_THREE
+    vpaddw      ymm2, ymm2, rel PW_ONE
+    vpaddw      ymm5, ymm5, rel PW_ONE
+    vpaddw      ymm3, ymm3, rel PW_TWO
+    vpaddw      ymm6, ymm6, rel PW_TWO
+
+    vpaddw      ymm2, ymm2, ymm1
+    vpaddw      ymm5, ymm5, ymm4
+    vpsrlw      ymm2, ymm2, 2                   ; ymm2=OutLE=( 0  2  4  6  8 10 12 14 16 18 20 22 24 26 28 30)
+    vpsrlw      ymm5, ymm5, 2                   ; ymm5=OutHE=(32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62)
+    vpaddw      ymm3, ymm3, ymm1
+    vpaddw      ymm6, ymm6, ymm4
+    vpsrlw      ymm3, ymm3, 2                   ; ymm3=OutLO=( 1  3  5  7  9 11 13 15 17 19 21 23 25 27 29 31)
+    vpsrlw      ymm6, ymm6, 2                   ; ymm6=OutHO=(33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63)
+
+    vpsllw      ymm3, ymm3, BYTE_BIT
+    vpsllw      ymm6, ymm6, BYTE_BIT
+    vpor        ymm2, ymm2, ymm3                ; ymm2=OutL=( 0  1  2 ... 29 30 31)
+    vpor        ymm5, ymm5, ymm6                ; ymm5=OutH=(32 33 34 ... 61 62 63)
+
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymm2
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymm5
+
+    sub         rax, byte SIZEOF_YMMWORD
+    add         rsi, byte 1*SIZEOF_YMMWORD  ; inptr
+    add         rdi, byte 2*SIZEOF_YMMWORD  ; outptr
+    cmp         rax, byte SIZEOF_YMMWORD
+    ja          near .columnloop
+    test        eax, eax
+    jnz         near .columnloop_last
+
+    pop         rsi
+    pop         rdi
+    pop         rax
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         rcx                        ; rowctr
+    jg          near .rowloop
+
+.return:
+    vzeroupper
+    uncollect_args 4
+    pop_xmm     3
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_avx2(int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11d = JDIMENSION downsampled_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_YMMWORD  ; ymmword wkWK_NUM
+%define WK_NUM  4
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_fancy_upsample_avx2)
+
+EXTN(jsimd_h2v2_fancy_upsample_avx2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_YMMWORD)  ; align to 256 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    push_xmm    3
+    collect_args 4
+    push        rbx
+
+    mov         eax, r11d               ; colctr
+    test        rax, rax
+    jz          near .return
+
+    mov         rcx, r10                ; rowctr
+    test        rcx, rcx
+    jz          near .return
+
+    mov         rsi, r12                ; input_data
+    mov         rdi, r13
+    mov         rdip, JSAMPARRAY rdi  ; output_data
+.rowloop:
+    push        rax                     ; colctr
+    push        rcx
+    push        rdi
+    push        rsi
+
+    mov         rcxp, JSAMPROW rsi-1*SIZEOF_JSAMPROW  ; inptr1(above)
+    mov         rbxp, JSAMPROW rsi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         rsip, JSAMPROW rsi+1*SIZEOF_JSAMPROW  ; inptr1(below)
+    mov         rdxp, JSAMPROW rdi+0*SIZEOF_JSAMPROW  ; outptr0
+    mov         rdip, JSAMPROW rdi+1*SIZEOF_JSAMPROW  ; outptr1
+
+    vpxor       ymm8, ymm8, ymm8                 ; ymm8=(all 0's)
+    vpcmpeqb    xmm9, xmm9, xmm9
+    vpsrldq     xmm10, xmm9, (SIZEOF_XMMWORD-2)  ; (ffff ---- ---- ... ---- ----) LSB is ffff
+    vpslldq     xmm9, xmm9, (SIZEOF_XMMWORD-2)
+    vperm2i128  ymm9, ymm9, ymm9, 1              ; (---- ---- ... ---- ---- ffff) MSB is ffff
+
+    test        rax, SIZEOF_YMMWORD-1
+    jz          short .skip
+    push        rdx
+    mov         dl, JSAMPLE rcx+(rax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE rcx+rax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE rbx+(rax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE rbx+rax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE rsi+(rax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE rsi+rax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+    pop         rdx
+.skip:
+    ; -- process the first column block
+
+    vmovdqu     ymm0, YMMWORD rbx+0*SIZEOF_YMMWORD  ; ymm0=row 00
+    vmovdqu     ymm1, YMMWORD rcx+0*SIZEOF_YMMWORD  ; ymm1=row-10
+    vmovdqu     ymm2, YMMWORD rsi+0*SIZEOF_YMMWORD  ; ymm2=row+10
+
+    vpunpckhbw  ymm4, ymm0, ymm8        ; ymm4=row 0( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm5, ymm0, ymm8        ; ymm5=row 0( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm0, ymm5, ymm4, 0x20  ; ymm0=row 0( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm4, ymm5, ymm4, 0x31  ; ymm4=row 0(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm5, ymm1, ymm8        ; ymm5=row-1( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm6, ymm1, ymm8        ; ymm6=row-1( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm1, ymm6, ymm5, 0x20  ; ymm1=row-1( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm5, ymm6, ymm5, 0x31  ; ymm5=row-1(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm6, ymm2, ymm8        ; ymm6=row+1( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm3, ymm2, ymm8        ; ymm3=row+1( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm2, ymm3, ymm6, 0x20  ; ymm2=row+1( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm6, ymm3, ymm6, 0x31  ; ymm6=row+1(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpmullw     ymm0, ymm0, rel PW_THREE
+    vpmullw     ymm4, ymm4, rel PW_THREE
+
+    vpaddw      ymm1, ymm1, ymm0        ; ymm1=Int0L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vpaddw      ymm5, ymm5, ymm4        ; ymm5=Int0H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+    vpaddw      ymm2, ymm2, ymm0        ; ymm2=Int1L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vpaddw      ymm6, ymm6, ymm4        ; ymm6=Int1H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vmovdqu     YMMWORD rdx+0*SIZEOF_YMMWORD, ymm1  ; temporarily save
+    vmovdqu     YMMWORD rdx+1*SIZEOF_YMMWORD, ymm5  ; the intermediate data
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymm2
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymm6
+
+    vpand       ymm1, ymm1, ymm10       ; ymm1=( 0 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vpand       ymm2, ymm2, ymm10       ; ymm2=( 0 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+
+    vmovdqa     YMMWORD wk(0), ymm1
+    vmovdqa     YMMWORD wk(1), ymm2
+
+    add         rax, byte SIZEOF_YMMWORD-1
+    and         rax, byte -SIZEOF_YMMWORD
+    cmp         rax, byte SIZEOF_YMMWORD
+    ja          short .columnloop
+
+.columnloop_last:
+    ; -- process the last column block
+
+    vpand       ymm1, ymm9, YMMWORD rdx+1*SIZEOF_YMMWORD
+    vpand       ymm2, ymm9, YMMWORD rdi+1*SIZEOF_YMMWORD
+
+    vmovdqa     YMMWORD wk(2), ymm1   ; ymm1=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 31)
+    vmovdqa     YMMWORD wk(3), ymm2   ; ymm2=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 31)
+
+    jmp         near .upsample
+
+.columnloop:
+    ; -- process the next column block
+
+    vmovdqu     ymm0, YMMWORD rbx+1*SIZEOF_YMMWORD  ; ymm0=row 01
+    vmovdqu     ymm1, YMMWORD rcx+1*SIZEOF_YMMWORD  ; ymm1=row-11
+    vmovdqu     ymm2, YMMWORD rsi+1*SIZEOF_YMMWORD  ; ymm2=row+11
+
+    vpunpckhbw  ymm4, ymm0, ymm8        ; ymm4=row 0( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm5, ymm0, ymm8        ; ymm5=row 0( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm0, ymm5, ymm4, 0x20  ; ymm0=row 0( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm4, ymm5, ymm4, 0x31  ; ymm4=row 0(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm5, ymm1, ymm8        ; ymm5=row-1( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm6, ymm1, ymm8        ; ymm6=row-1( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm1, ymm6, ymm5, 0x20  ; ymm1=row-1( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm5, ymm6, ymm5, 0x31  ; ymm5=row-1(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpunpckhbw  ymm6, ymm2, ymm8        ; ymm6=row+1( 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31)
+    vpunpcklbw  ymm7, ymm2, ymm8        ; ymm7=row+1( 0  1  2  3  4  5  6  7 16 17 18 19 20 21 22 23)
+    vperm2i128  ymm2, ymm7, ymm6, 0x20  ; ymm2=row+1( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vperm2i128  ymm6, ymm7, ymm6, 0x31  ; ymm6=row+1(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vpmullw     ymm0, ymm0, rel PW_THREE
+    vpmullw     ymm4, ymm4, rel PW_THREE
+
+    vpaddw      ymm1, ymm1, ymm0        ; ymm1=Int0L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vpaddw      ymm5, ymm5, ymm4        ; ymm5=Int0H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+    vpaddw      ymm2, ymm2, ymm0        ; ymm2=Int1L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vpaddw      ymm6, ymm6, ymm4        ; ymm6=Int1H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vmovdqu     YMMWORD rdx+2*SIZEOF_YMMWORD, ymm1  ; temporarily save
+    vmovdqu     YMMWORD rdx+3*SIZEOF_YMMWORD, ymm5  ; the intermediate data
+    vmovdqu     YMMWORD rdi+2*SIZEOF_YMMWORD, ymm2
+    vmovdqu     YMMWORD rdi+3*SIZEOF_YMMWORD, ymm6
+
+    vperm2i128  ymm1, ymm8, ymm1, 0x20
+    vpslldq     ymm1, ymm1, 14          ; ymm1=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- --  0)
+    vperm2i128  ymm2, ymm8, ymm2, 0x20
+    vpslldq     ymm2, ymm2, 14          ; ymm2=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- --  0)
+
+    vmovdqa     YMMWORD wk(2), ymm1
+    vmovdqa     YMMWORD wk(3), ymm2
+
+.upsample:
+    ; -- process the upper row
+
+    vmovdqu     ymm7, YMMWORD rdx+0*SIZEOF_YMMWORD  ; ymm7=Int0L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vmovdqu     ymm3, YMMWORD rdx+1*SIZEOF_YMMWORD  ; ymm3=Int0H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vperm2i128  ymm0, ymm8, ymm7, 0x03
+    vpalignr    ymm0, ymm0, ymm7, 2     ; ymm0=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 --)
+    vperm2i128  ymm4, ymm8, ymm3, 0x20
+    vpslldq     ymm4, ymm4, 14          ; ymm4=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 16)
+
+    vperm2i128  ymm5, ymm8, ymm7, 0x03
+    vpsrldq     ymm5, ymm5, 14          ; ymm5=(15 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vperm2i128  ymm6, ymm8, ymm3, 0x20
+    vpalignr    ymm6, ymm3, ymm6, 14    ; ymm6=(-- 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vpor        ymm0, ymm0, ymm4        ; ymm0=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16)
+    vpor        ymm5, ymm5, ymm6        ; ymm5=(15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vperm2i128  ymm2, ymm8, ymm3, 0x03
+    vpalignr    ymm2, ymm2, ymm3, 2     ; ymm2=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 --)
+    vperm2i128  ymm4, ymm8, ymm3, 0x03
+    vpsrldq     ymm4, ymm4, 14          ; ymm4=(31 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vperm2i128  ymm1, ymm8, ymm7, 0x20
+    vpalignr    ymm1, ymm7, ymm1, 14    ; ymm1=(--  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+
+    vpor        ymm1, ymm1, YMMWORD wk(0)  ; ymm1=(-1  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+    vpor        ymm2, ymm2, YMMWORD wk(2)  ; ymm2=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32)
+
+    vmovdqa     YMMWORD wk(0), ymm4
+
+    vpmullw     ymm7, ymm7, rel PW_THREE
+    vpmullw     ymm3, ymm3, rel PW_THREE
+    vpaddw      ymm1, ymm1, rel PW_EIGHT
+    vpaddw      ymm5, ymm5, rel PW_EIGHT
+    vpaddw      ymm0, ymm0, rel PW_SEVEN
+    vpaddw      ymm2, rel PW_SEVEN
+
+    vpaddw      ymm1, ymm1, ymm7
+    vpaddw      ymm5, ymm5, ymm3
+    vpsrlw      ymm1, ymm1, 4           ; ymm1=Out0LE=( 0  2  4  6  8 10 12 14 16 18 20 22 24 26 28 30)
+    vpsrlw      ymm5, ymm5, 4           ; ymm5=Out0HE=(32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62)
+    vpaddw      ymm0, ymm0, ymm7
+    vpaddw      ymm2, ymm2, ymm3
+    vpsrlw      ymm0, ymm0, 4           ; ymm0=Out0LO=( 1  3  5  7  9 11 13 15 17 19 21 23 25 27 29 31)
+    vpsrlw      ymm2, ymm2, 4           ; ymm2=Out0HO=(33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63)
+
+    vpsllw      ymm0, ymm0, BYTE_BIT
+    vpsllw      ymm2, ymm2, BYTE_BIT
+    vpor        ymm1, ymm1, ymm0        ; ymm1=Out0L=( 0  1  2 ... 29 30 31)
+    vpor        ymm5, ymm5, ymm2        ; ymm5=Out0H=(32 33 34 ... 61 62 63)
+
+    vmovdqu     YMMWORD rdx+0*SIZEOF_YMMWORD, ymm1
+    vmovdqu     YMMWORD rdx+1*SIZEOF_YMMWORD, ymm5
+
+    ; -- process the lower row
+
+    vmovdqu     ymm6, YMMWORD rdi+0*SIZEOF_YMMWORD  ; ymm6=Int1L=( 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15)
+    vmovdqu     ymm4, YMMWORD rdi+1*SIZEOF_YMMWORD  ; ymm4=Int1H=(16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31)
+
+    vperm2i128  ymm7, ymm8, ymm6, 0x03
+    vpalignr    ymm7, ymm7, ymm6, 2     ; ymm7=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 --)
+    vperm2i128  ymm3, ymm8, ymm4, 0x20
+    vpslldq     ymm3, ymm3, 14          ; ymm3=(-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 16)
+
+    vperm2i128  ymm0, ymm8, ymm6, 0x03
+    vpsrldq     ymm0, ymm0, 14          ; ymm0=(15 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vperm2i128  ymm2, ymm8, ymm4, 0x20
+    vpalignr    ymm2, ymm4, ymm2, 14    ; ymm2=(-- 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vpor        ymm7, ymm7, ymm3        ; ymm7=( 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16)
+    vpor        ymm0, ymm0, ymm2        ; ymm0=(15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30)
+
+    vperm2i128  ymm5, ymm8, ymm4, 0x03
+    vpalignr    ymm5, ymm5, ymm4, 2     ; ymm5=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 --)
+    vperm2i128  ymm3, ymm8, ymm4, 0x03
+    vpsrldq     ymm3, ymm3, 14          ; ymm3=(31 -- -- -- -- -- -- -- -- -- -- -- -- -- -- --)
+    vperm2i128  ymm1, ymm8, ymm6, 0x20
+    vpalignr    ymm1, ymm6, ymm1, 14    ; ymm1=(--  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+
+    vpor        ymm1, ymm1, YMMWORD wk(1)  ; ymm1=(-1  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14)
+    vpor        ymm5, ymm5, YMMWORD wk(3)  ; ymm5=(17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32)
+
+    vmovdqa     YMMWORD wk(1), ymm3
+
+    vpmullw     ymm6, ymm6, rel PW_THREE
+    vpmullw     ymm4, ymm4, rel PW_THREE
+    vpaddw      ymm1, ymm1, rel PW_EIGHT
+    vpaddw      ymm0, ymm0, rel PW_EIGHT
+    vpaddw      ymm7, ymm7, rel PW_SEVEN
+    vpaddw      ymm5, ymm5, rel PW_SEVEN
+
+    vpaddw      ymm1, ymm1, ymm6
+    vpaddw      ymm0, ymm0, ymm4
+    vpsrlw      ymm1, ymm1, 4           ; ymm1=Out1LE=( 0  2  4  6  8 10 12 14 16 18 20 22 24 26 28 30)
+    vpsrlw      ymm0, ymm0, 4           ; ymm0=Out1HE=(32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62)
+    vpaddw      ymm7, ymm7, ymm6
+    vpaddw      ymm5, ymm5, ymm4
+    vpsrlw      ymm7, ymm7, 4           ; ymm7=Out1LO=( 1  3  5  7  9 11 13 15 17 19 21 23 25 27 29 31)
+    vpsrlw      ymm5, ymm5, 4           ; ymm5=Out1HO=(33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63)
+
+    vpsllw      ymm7, ymm7, BYTE_BIT
+    vpsllw      ymm5, ymm5, BYTE_BIT
+    vpor        ymm1, ymm1, ymm7        ; ymm1=Out1L=( 0  1  2 ... 29 30 31)
+    vpor        ymm0, ymm0, ymm5        ; ymm0=Out1H=(32 33 34 ... 61 62 63)
+
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymm1
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymm0
+
+    sub         rax, byte SIZEOF_YMMWORD
+    add         rcx, byte 1*SIZEOF_YMMWORD  ; inptr1(above)
+    add         rbx, byte 1*SIZEOF_YMMWORD  ; inptr0
+    add         rsi, byte 1*SIZEOF_YMMWORD  ; inptr1(below)
+    add         rdx, byte 2*SIZEOF_YMMWORD  ; outptr0
+    add         rdi, byte 2*SIZEOF_YMMWORD  ; outptr1
+    cmp         rax, byte SIZEOF_YMMWORD
+    ja          near .columnloop
+    test        rax, rax
+    jnz         near .columnloop_last
+
+    pop         rsi
+    pop         rdi
+    pop         rcx
+    pop         rax
+
+    add         rsi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         rcx, byte 2                  ; rowctr
+    jg          near .rowloop
+
+.return:
+    pop         rbx
+    vzeroupper
+    uncollect_args 4
+    pop_xmm     3
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_avx2(int max_v_samp_factor, JDIMENSION output_width,
+;                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11d = JDIMENSION output_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_upsample_avx2)
+
+EXTN(jsimd_h2v1_upsample_avx2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 4
+
+    mov         edx, r11d
+    add         rdx, byte (SIZEOF_YMMWORD-1)
+    and         rdx, -SIZEOF_YMMWORD
+    jz          near .return
+
+    mov         rcx, r10                ; rowctr
+    test        rcx, rcx
+    jz          short .return
+
+    mov         rsi, r12                ; input_data
+    mov         rdi, r13
+    mov         rdip, JSAMPARRAY rdi  ; output_data
+.rowloop:
+    push        rdi
+    push        rsi
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr
+    mov         rax, rdx                ; colctr
+.columnloop:
+
+    cmp         rax, byte SIZEOF_YMMWORD
+    ja          near .above_16
+
+    vmovdqu     xmm0, XMMWORD rsi+0*SIZEOF_YMMWORD
+    vpunpckhbw  xmm1, xmm0, xmm0
+    vpunpcklbw  xmm0, xmm0, xmm0
+
+    vmovdqu     XMMWORD rdi+0*SIZEOF_XMMWORD, xmm0
+    vmovdqu     XMMWORD rdi+1*SIZEOF_XMMWORD, xmm1
+
+    jmp         short .nextrow
+
+.above_16:
+    vmovdqu     ymm0, YMMWORD rsi+0*SIZEOF_YMMWORD
+
+    vpermq      ymm0, ymm0, 0xd8
+    vpunpckhbw  ymm1, ymm0, ymm0
+    vpunpcklbw  ymm0, ymm0, ymm0
+
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymm0
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymm1
+
+    sub         rax, byte 2*SIZEOF_YMMWORD
+    jz          short .nextrow
+
+    add         rsi, byte SIZEOF_YMMWORD    ; inptr
+    add         rdi, byte 2*SIZEOF_YMMWORD  ; outptr
+    jmp         short .columnloop
+
+.nextrow:
+    pop         rsi
+    pop         rdi
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         rcx                        ; rowctr
+    jg          short .rowloop
+
+.return:
+    vzeroupper
+    uncollect_args 4
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_avx2(int max_v_samp_factor, JDIMENSION output_width,
+;                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11d = JDIMENSION output_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_upsample_avx2)
+
+EXTN(jsimd_h2v2_upsample_avx2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 4
+    push        rbx
+
+    mov         edx, r11d
+    add         rdx, byte (SIZEOF_YMMWORD-1)
+    and         rdx, -SIZEOF_YMMWORD
+    jz          near .return
+
+    mov         rcx, r10                ; rowctr
+    test        rcx, rcx
+    jz          near .return
+
+    mov         rsi, r12                ; input_data
+    mov         rdi, r13
+    mov         rdip, JSAMPARRAY rdi  ; output_data
+.rowloop:
+    push        rdi
+    push        rsi
+
+    mov         rsip, JSAMPROW rsi                   ; inptr
+    mov         rbxp, JSAMPROW rdi+0*SIZEOF_JSAMPROW ; outptr0
+    mov         rdip, JSAMPROW rdi+1*SIZEOF_JSAMPROW ; outptr1
+    mov         rax, rdx                               ; colctr
+.columnloop:
+
+    cmp         rax, byte SIZEOF_YMMWORD
+    ja          short .above_16
+
+    vmovdqu     xmm0, XMMWORD rsi+0*SIZEOF_XMMWORD
+    vpunpckhbw  xmm1, xmm0, xmm0
+    vpunpcklbw  xmm0, xmm0, xmm0
+
+    vmovdqu     XMMWORD rbx+0*SIZEOF_XMMWORD, xmm0
+    vmovdqu     XMMWORD rbx+1*SIZEOF_XMMWORD, xmm1
+    vmovdqu     XMMWORD rdi+0*SIZEOF_XMMWORD, xmm0
+    vmovdqu     XMMWORD rdi+1*SIZEOF_XMMWORD, xmm1
+
+    jmp         near .nextrow
+
+.above_16:
+    vmovdqu     ymm0, YMMWORD rsi+0*SIZEOF_YMMWORD
+
+    vpermq      ymm0, ymm0, 0xd8
+    vpunpckhbw  ymm1, ymm0, ymm0
+    vpunpcklbw  ymm0, ymm0, ymm0
+
+    vmovdqu     YMMWORD rbx+0*SIZEOF_YMMWORD, ymm0
+    vmovdqu     YMMWORD rbx+1*SIZEOF_YMMWORD, ymm1
+    vmovdqu     YMMWORD rdi+0*SIZEOF_YMMWORD, ymm0
+    vmovdqu     YMMWORD rdi+1*SIZEOF_YMMWORD, ymm1
+
+    sub         rax, byte 2*SIZEOF_YMMWORD
+    jz          short .nextrow
+
+    add         rsi, byte SIZEOF_YMMWORD  ; inptr
+    add         rbx, 2*SIZEOF_YMMWORD     ; outptr0
+    add         rdi, 2*SIZEOF_YMMWORD     ; outptr1
+    jmp         short .columnloop
+
+.nextrow:
+    pop         rsi
+    pop         rdi
+
+    add         rsi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         rcx, byte 2                  ; rowctr
+    jg          near .rowloop
+
+.return:
+    pop         rbx
+    vzeroupper
+    uncollect_args 4
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jdsample-sse2.asm Added

@@ -0,0 +1,665 @@
+;
+; jdsample.asm - upsampling (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fancy_upsample_sse2)
+
+EXTN(jconst_fancy_upsample_sse2):
+
+PW_ONE   times 8 dw 1
+PW_TWO   times 8 dw 2
+PW_THREE times 8 dw 3
+PW_SEVEN times 8 dw 7
+PW_EIGHT times 8 dw 8
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_sse2(int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11d = JDIMENSION downsampled_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_fancy_upsample_sse2)
+
+EXTN(jsimd_h2v1_fancy_upsample_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 4
+
+    mov         eax, r11d               ; colctr
+    test        rax, rax
+    jz          near .return
+
+    mov         rcx, r10                ; rowctr
+    test        rcx, rcx
+    jz          near .return
+
+    mov         rsi, r12                ; input_data
+    mov         rdi, r13
+    mov         rdip, JSAMPARRAY rdi  ; output_data
+.rowloop:
+    push        rax                     ; colctr
+    push        rdi
+    push        rsi
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr
+
+    test        rax, SIZEOF_XMMWORD-1
+    jz          short .skip
+    mov         dl, JSAMPLE rsi+(rax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE rsi+rax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+.skip:
+    pxor        xmm0, xmm0              ; xmm0=(all 0's)
+    pcmpeqb     xmm7, xmm7
+    psrldq      xmm7, (SIZEOF_XMMWORD-1)
+    pand        xmm7, XMMWORD rsi+0*SIZEOF_XMMWORD
+
+    add         rax, byte SIZEOF_XMMWORD-1
+    and         rax, byte -SIZEOF_XMMWORD
+    cmp         rax, byte SIZEOF_XMMWORD
+    ja          short .columnloop
+
+.columnloop_last:
+    pcmpeqb     xmm6, xmm6
+    pslldq      xmm6, (SIZEOF_XMMWORD-1)
+    pand        xmm6, XMMWORD rsi+0*SIZEOF_XMMWORD
+    jmp         short .upsample
+
+.columnloop:
+    movdqa      xmm6, XMMWORD rsi+1*SIZEOF_XMMWORD
+    pslldq      xmm6, (SIZEOF_XMMWORD-1)
+
+.upsample:
+    movdqa      xmm1, XMMWORD rsi+0*SIZEOF_XMMWORD
+    movdqa      xmm2, xmm1
+    movdqa      xmm3, xmm1                ; xmm1=( 0  1  2 ... 13 14 15)
+    pslldq      xmm2, 1                   ; xmm2=(--  0  1 ... 12 13 14)
+    psrldq      xmm3, 1                   ; xmm3=( 1  2  3 ... 14 15 --)
+
+    por         xmm2, xmm7                ; xmm2=(-1  0  1 ... 12 13 14)
+    por         xmm3, xmm6                ; xmm3=( 1  2  3 ... 14 15 16)
+
+    movdqa      xmm7, xmm1
+    psrldq      xmm7, (SIZEOF_XMMWORD-1)  ; xmm7=(15 -- -- ... -- -- --)
+
+    movdqa      xmm4, xmm1
+    punpcklbw   xmm1, xmm0                ; xmm1=( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm4, xmm0                ; xmm4=( 8  9 10 11 12 13 14 15)
+    movdqa      xmm5, xmm2
+    punpcklbw   xmm2, xmm0                ; xmm2=(-1  0  1  2  3  4  5  6)
+    punpckhbw   xmm5, xmm0                ; xmm5=( 7  8  9 10 11 12 13 14)
+    movdqa      xmm6, xmm3
+    punpcklbw   xmm3, xmm0                ; xmm3=( 1  2  3  4  5  6  7  8)
+    punpckhbw   xmm6, xmm0                ; xmm6=( 9 10 11 12 13 14 15 16)
+
+    pmullw      xmm1, rel PW_THREE
+    pmullw      xmm4, rel PW_THREE
+    paddw       xmm2, rel PW_ONE
+    paddw       xmm5, rel PW_ONE
+    paddw       xmm3, rel PW_TWO
+    paddw       xmm6, rel PW_TWO
+
+    paddw       xmm2, xmm1
+    paddw       xmm5, xmm4
+    psrlw       xmm2, 2                 ; xmm2=OutLE=( 0  2  4  6  8 10 12 14)
+    psrlw       xmm5, 2                 ; xmm5=OutHE=(16 18 20 22 24 26 28 30)
+    paddw       xmm3, xmm1
+    paddw       xmm6, xmm4
+    psrlw       xmm3, 2                 ; xmm3=OutLO=( 1  3  5  7  9 11 13 15)
+    psrlw       xmm6, 2                 ; xmm6=OutHO=(17 19 21 23 25 27 29 31)
+
+    psllw       xmm3, BYTE_BIT
+    psllw       xmm6, BYTE_BIT
+    por         xmm2, xmm3              ; xmm2=OutL=( 0  1  2 ... 13 14 15)
+    por         xmm5, xmm6              ; xmm5=OutH=(16 17 18 ... 29 30 31)
+
+    movdqa      XMMWORD rdi+0*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD rdi+1*SIZEOF_XMMWORD, xmm5
+
+    sub         rax, byte SIZEOF_XMMWORD
+    add         rsi, byte 1*SIZEOF_XMMWORD  ; inptr
+    add         rdi, byte 2*SIZEOF_XMMWORD  ; outptr
+    cmp         rax, byte SIZEOF_XMMWORD
+    ja          near .columnloop
+    test        eax, eax
+    jnz         near .columnloop_last
+
+    pop         rsi
+    pop         rdi
+    pop         rax
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         rcx                        ; rowctr
+    jg          near .rowloop
+
+.return:
+    uncollect_args 4
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_sse2(int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11d = JDIMENSION downsampled_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD  ; xmmword wkWK_NUM
+%define WK_NUM  4
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_fancy_upsample_sse2)
+
+EXTN(jsimd_h2v2_fancy_upsample_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 4
+    push        rbx
+
+    mov         eax, r11d               ; colctr
+    test        rax, rax
+    jz          near .return
+
+    mov         rcx, r10                ; rowctr
+    test        rcx, rcx
+    jz          near .return
+
+    mov         rsi, r12                ; input_data
+    mov         rdi, r13
+    mov         rdip, JSAMPARRAY rdi  ; output_data
+.rowloop:
+    push        rax                     ; colctr
+    push        rcx
+    push        rdi
+    push        rsi
+
+    mov         rcxp, JSAMPROW rsi-1*SIZEOF_JSAMPROW  ; inptr1(above)
+    mov         rbxp, JSAMPROW rsi+0*SIZEOF_JSAMPROW  ; inptr0
+    mov         rsip, JSAMPROW rsi+1*SIZEOF_JSAMPROW  ; inptr1(below)
+    mov         rdxp, JSAMPROW rdi+0*SIZEOF_JSAMPROW  ; outptr0
+    mov         rdip, JSAMPROW rdi+1*SIZEOF_JSAMPROW  ; outptr1
+
+    test        rax, SIZEOF_XMMWORD-1
+    jz          short .skip
+    push        rdx
+    mov         dl, JSAMPLE rcx+(rax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE rcx+rax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE rbx+(rax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE rbx+rax*SIZEOF_JSAMPLE, dl
+    mov         dl, JSAMPLE rsi+(rax-1)*SIZEOF_JSAMPLE
+    mov         JSAMPLE rsi+rax*SIZEOF_JSAMPLE, dl    ; insert a dummy sample
+    pop         rdx
+.skip:
+    ; -- process the first column block
+
+    movdqa      xmm0, XMMWORD rbx+0*SIZEOF_XMMWORD  ; xmm0=row 00
+    movdqa      xmm1, XMMWORD rcx+0*SIZEOF_XMMWORD  ; xmm1=row-10
+    movdqa      xmm2, XMMWORD rsi+0*SIZEOF_XMMWORD  ; xmm2=row+10
+
+    pxor        xmm3, xmm3              ; xmm3=(all 0's)
+    movdqa      xmm4, xmm0
+    punpcklbw   xmm0, xmm3              ; xmm0=row 0( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm4, xmm3              ; xmm4=row 0( 8  9 10 11 12 13 14 15)
+    movdqa      xmm5, xmm1
+    punpcklbw   xmm1, xmm3              ; xmm1=row-1( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm5, xmm3              ; xmm5=row-1( 8  9 10 11 12 13 14 15)
+    movdqa      xmm6, xmm2
+    punpcklbw   xmm2, xmm3              ; xmm2=row+1( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm6, xmm3              ; xmm6=row+1( 8  9 10 11 12 13 14 15)
+
+    pmullw      xmm0, rel PW_THREE
+    pmullw      xmm4, rel PW_THREE
+
+    pcmpeqb     xmm7, xmm7
+    psrldq      xmm7, (SIZEOF_XMMWORD-2)
+
+    paddw       xmm1, xmm0              ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+    paddw       xmm5, xmm4              ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+    paddw       xmm2, xmm0              ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+    paddw       xmm6, xmm4              ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+    movdqa      XMMWORD rdx+0*SIZEOF_XMMWORD, xmm1  ; temporarily save
+    movdqa      XMMWORD rdx+1*SIZEOF_XMMWORD, xmm5  ; the intermediate data
+    movdqa      XMMWORD rdi+0*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD rdi+1*SIZEOF_XMMWORD, xmm6
+
+    pand        xmm1, xmm7              ; xmm1=( 0 -- -- -- -- -- -- --)
+    pand        xmm2, xmm7              ; xmm2=( 0 -- -- -- -- -- -- --)
+
+    movdqa      XMMWORD wk(0), xmm1
+    movdqa      XMMWORD wk(1), xmm2
+
+    add         rax, byte SIZEOF_XMMWORD-1
+    and         rax, byte -SIZEOF_XMMWORD
+    cmp         rax, byte SIZEOF_XMMWORD
+    ja          short .columnloop
+
+.columnloop_last:
+    ; -- process the last column block
+
+    pcmpeqb     xmm1, xmm1
+    pslldq      xmm1, (SIZEOF_XMMWORD-2)
+    movdqa      xmm2, xmm1
+
+    pand        xmm1, XMMWORD rdx+1*SIZEOF_XMMWORD
+    pand        xmm2, XMMWORD rdi+1*SIZEOF_XMMWORD
+
+    movdqa      XMMWORD wk(2), xmm1   ; xmm1=(-- -- -- -- -- -- -- 15)
+    movdqa      XMMWORD wk(3), xmm2   ; xmm2=(-- -- -- -- -- -- -- 15)
+
+    jmp         near .upsample
+
+.columnloop:
+    ; -- process the next column block
+
+    movdqa      xmm0, XMMWORD rbx+1*SIZEOF_XMMWORD  ; xmm0=row 01
+    movdqa      xmm1, XMMWORD rcx+1*SIZEOF_XMMWORD  ; xmm1=row-11
+    movdqa      xmm2, XMMWORD rsi+1*SIZEOF_XMMWORD  ; xmm2=row+11
+
+    pxor        xmm3, xmm3              ; xmm3=(all 0's)
+    movdqa      xmm4, xmm0
+    punpcklbw   xmm0, xmm3              ; xmm0=row 0( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm4, xmm3              ; xmm4=row 0( 8  9 10 11 12 13 14 15)
+    movdqa      xmm5, xmm1
+    punpcklbw   xmm1, xmm3              ; xmm1=row-1( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm5, xmm3              ; xmm5=row-1( 8  9 10 11 12 13 14 15)
+    movdqa      xmm6, xmm2
+    punpcklbw   xmm2, xmm3              ; xmm2=row+1( 0  1  2  3  4  5  6  7)
+    punpckhbw   xmm6, xmm3              ; xmm6=row+1( 8  9 10 11 12 13 14 15)
+
+    pmullw      xmm0, rel PW_THREE
+    pmullw      xmm4, rel PW_THREE
+
+    paddw       xmm1, xmm0              ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+    paddw       xmm5, xmm4              ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+    paddw       xmm2, xmm0              ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+    paddw       xmm6, xmm4              ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+    movdqa      XMMWORD rdx+2*SIZEOF_XMMWORD, xmm1  ; temporarily save
+    movdqa      XMMWORD rdx+3*SIZEOF_XMMWORD, xmm5  ; the intermediate data
+    movdqa      XMMWORD rdi+2*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD rdi+3*SIZEOF_XMMWORD, xmm6
+
+    pslldq      xmm1, (SIZEOF_XMMWORD-2)  ; xmm1=(-- -- -- -- -- -- --  0)
+    pslldq      xmm2, (SIZEOF_XMMWORD-2)  ; xmm2=(-- -- -- -- -- -- --  0)
+
+    movdqa      XMMWORD wk(2), xmm1
+    movdqa      XMMWORD wk(3), xmm2
+
+.upsample:
+    ; -- process the upper row
+
+    movdqa      xmm7, XMMWORD rdx+0*SIZEOF_XMMWORD
+    movdqa      xmm3, XMMWORD rdx+1*SIZEOF_XMMWORD
+
+    movdqa      xmm0, xmm7                ; xmm7=Int0L=( 0  1  2  3  4  5  6  7)
+    movdqa      xmm4, xmm3                ; xmm3=Int0H=( 8  9 10 11 12 13 14 15)
+    psrldq      xmm0, 2                   ; xmm0=( 1  2  3  4  5  6  7 --)
+    pslldq      xmm4, (SIZEOF_XMMWORD-2)  ; xmm4=(-- -- -- -- -- -- --  8)
+    movdqa      xmm5, xmm7
+    movdqa      xmm6, xmm3
+    psrldq      xmm5, (SIZEOF_XMMWORD-2)  ; xmm5=( 7 -- -- -- -- -- -- --)
+    pslldq      xmm6, 2                   ; xmm6=(--  8  9 10 11 12 13 14)
+
+    por         xmm0, xmm4                ; xmm0=( 1  2  3  4  5  6  7  8)
+    por         xmm5, xmm6                ; xmm5=( 7  8  9 10 11 12 13 14)
+
+    movdqa      xmm1, xmm7
+    movdqa      xmm2, xmm3
+    pslldq      xmm1, 2                   ; xmm1=(--  0  1  2  3  4  5  6)
+    psrldq      xmm2, 2                   ; xmm2=( 9 10 11 12 13 14 15 --)
+    movdqa      xmm4, xmm3
+    psrldq      xmm4, (SIZEOF_XMMWORD-2)  ; xmm4=(15 -- -- -- -- -- -- --)
+
+    por         xmm1, XMMWORD wk(0)     ; xmm1=(-1  0  1  2  3  4  5  6)
+    por         xmm2, XMMWORD wk(2)     ; xmm2=( 9 10 11 12 13 14 15 16)
+
+    movdqa      XMMWORD wk(0), xmm4
+
+    pmullw      xmm7, rel PW_THREE
+    pmullw      xmm3, rel PW_THREE
+    paddw       xmm1, rel PW_EIGHT
+    paddw       xmm5, rel PW_EIGHT
+    paddw       xmm0, rel PW_SEVEN
+    paddw       xmm2, rel PW_SEVEN
+
+    paddw       xmm1, xmm7
+    paddw       xmm5, xmm3
+    psrlw       xmm1, 4                 ; xmm1=Out0LE=( 0  2  4  6  8 10 12 14)
+    psrlw       xmm5, 4                 ; xmm5=Out0HE=(16 18 20 22 24 26 28 30)
+    paddw       xmm0, xmm7
+    paddw       xmm2, xmm3
+    psrlw       xmm0, 4                 ; xmm0=Out0LO=( 1  3  5  7  9 11 13 15)
+    psrlw       xmm2, 4                 ; xmm2=Out0HO=(17 19 21 23 25 27 29 31)
+
+    psllw       xmm0, BYTE_BIT
+    psllw       xmm2, BYTE_BIT
+    por         xmm1, xmm0              ; xmm1=Out0L=( 0  1  2 ... 13 14 15)
+    por         xmm5, xmm2              ; xmm5=Out0H=(16 17 18 ... 29 30 31)
+
+    movdqa      XMMWORD rdx+0*SIZEOF_XMMWORD, xmm1
+    movdqa      XMMWORD rdx+1*SIZEOF_XMMWORD, xmm5
+
+    ; -- process the lower row
+
+    movdqa      xmm6, XMMWORD rdi+0*SIZEOF_XMMWORD
+    movdqa      xmm4, XMMWORD rdi+1*SIZEOF_XMMWORD
+
+    movdqa      xmm7, xmm6                ; xmm6=Int1L=( 0  1  2  3  4  5  6  7)
+    movdqa      xmm3, xmm4                ; xmm4=Int1H=( 8  9 10 11 12 13 14 15)
+    psrldq      xmm7, 2                   ; xmm7=( 1  2  3  4  5  6  7 --)
+    pslldq      xmm3, (SIZEOF_XMMWORD-2)  ; xmm3=(-- -- -- -- -- -- --  8)
+    movdqa      xmm0, xmm6
+    movdqa      xmm2, xmm4
+    psrldq      xmm0, (SIZEOF_XMMWORD-2)  ; xmm0=( 7 -- -- -- -- -- -- --)
+    pslldq      xmm2, 2                   ; xmm2=(--  8  9 10 11 12 13 14)
+
+    por         xmm7, xmm3                ; xmm7=( 1  2  3  4  5  6  7  8)
+    por         xmm0, xmm2                ; xmm0=( 7  8  9 10 11 12 13 14)
+
+    movdqa      xmm1, xmm6
+    movdqa      xmm5, xmm4
+    pslldq      xmm1, 2                   ; xmm1=(--  0  1  2  3  4  5  6)
+    psrldq      xmm5, 2                   ; xmm5=( 9 10 11 12 13 14 15 --)
+    movdqa      xmm3, xmm4
+    psrldq      xmm3, (SIZEOF_XMMWORD-2)  ; xmm3=(15 -- -- -- -- -- -- --)
+
+    por         xmm1, XMMWORD wk(1)     ; xmm1=(-1  0  1  2  3  4  5  6)
+    por         xmm5, XMMWORD wk(3)     ; xmm5=( 9 10 11 12 13 14 15 16)
+
+    movdqa      XMMWORD wk(1), xmm3
+
+    pmullw      xmm6, rel PW_THREE
+    pmullw      xmm4, rel PW_THREE
+    paddw       xmm1, rel PW_EIGHT
+    paddw       xmm0, rel PW_EIGHT
+    paddw       xmm7, rel PW_SEVEN
+    paddw       xmm5, rel PW_SEVEN
+
+    paddw       xmm1, xmm6
+    paddw       xmm0, xmm4
+    psrlw       xmm1, 4                 ; xmm1=Out1LE=( 0  2  4  6  8 10 12 14)
+    psrlw       xmm0, 4                 ; xmm0=Out1HE=(16 18 20 22 24 26 28 30)
+    paddw       xmm7, xmm6
+    paddw       xmm5, xmm4
+    psrlw       xmm7, 4                 ; xmm7=Out1LO=( 1  3  5  7  9 11 13 15)
+    psrlw       xmm5, 4                 ; xmm5=Out1HO=(17 19 21 23 25 27 29 31)
+
+    psllw       xmm7, BYTE_BIT
+    psllw       xmm5, BYTE_BIT
+    por         xmm1, xmm7              ; xmm1=Out1L=( 0  1  2 ... 13 14 15)
+    por         xmm0, xmm5              ; xmm0=Out1H=(16 17 18 ... 29 30 31)
+
+    movdqa      XMMWORD rdi+0*SIZEOF_XMMWORD, xmm1
+    movdqa      XMMWORD rdi+1*SIZEOF_XMMWORD, xmm0
+
+    sub         rax, byte SIZEOF_XMMWORD
+    add         rcx, byte 1*SIZEOF_XMMWORD  ; inptr1(above)
+    add         rbx, byte 1*SIZEOF_XMMWORD  ; inptr0
+    add         rsi, byte 1*SIZEOF_XMMWORD  ; inptr1(below)
+    add         rdx, byte 2*SIZEOF_XMMWORD  ; outptr0
+    add         rdi, byte 2*SIZEOF_XMMWORD  ; outptr1
+    cmp         rax, byte SIZEOF_XMMWORD
+    ja          near .columnloop
+    test        rax, rax
+    jnz         near .columnloop_last
+
+    pop         rsi
+    pop         rdi
+    pop         rcx
+    pop         rax
+
+    add         rsi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         rcx, byte 2                  ; rowctr
+    jg          near .rowloop
+
+.return:
+    pop         rbx
+    uncollect_args 4
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_sse2(int max_v_samp_factor, JDIMENSION output_width,
+;                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11d = JDIMENSION output_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v1_upsample_sse2)
+
+EXTN(jsimd_h2v1_upsample_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 4
+
+    mov         edx, r11d
+    add         rdx, byte (2*SIZEOF_XMMWORD)-1
+    and         rdx, byte -(2*SIZEOF_XMMWORD)
+    jz          near .return
+
+    mov         rcx, r10                ; rowctr
+    test        rcx, rcx
+    jz          short .return
+
+    mov         rsi, r12                ; input_data
+    mov         rdi, r13
+    mov         rdip, JSAMPARRAY rdi  ; output_data
+.rowloop:
+    push        rdi
+    push        rsi
+
+    mov         rsip, JSAMPROW rsi    ; inptr
+    mov         rdip, JSAMPROW rdi    ; outptr
+    mov         rax, rdx                ; colctr
+.columnloop:
+
+    movdqa      xmm0, XMMWORD rsi+0*SIZEOF_XMMWORD
+
+    movdqa      xmm1, xmm0
+    punpcklbw   xmm0, xmm0
+    punpckhbw   xmm1, xmm1
+
+    movdqa      XMMWORD rdi+0*SIZEOF_XMMWORD, xmm0
+    movdqa      XMMWORD rdi+1*SIZEOF_XMMWORD, xmm1
+
+    sub         rax, byte 2*SIZEOF_XMMWORD
+    jz          short .nextrow
+
+    movdqa      xmm2, XMMWORD rsi+1*SIZEOF_XMMWORD
+
+    movdqa      xmm3, xmm2
+    punpcklbw   xmm2, xmm2
+    punpckhbw   xmm3, xmm3
+
+    movdqa      XMMWORD rdi+2*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD rdi+3*SIZEOF_XMMWORD, xmm3
+
+    sub         rax, byte 2*SIZEOF_XMMWORD
+    jz          short .nextrow
+
+    add         rsi, byte 2*SIZEOF_XMMWORD  ; inptr
+    add         rdi, byte 4*SIZEOF_XMMWORD  ; outptr
+    jmp         short .columnloop
+
+.nextrow:
+    pop         rsi
+    pop         rdi
+
+    add         rsi, byte SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte SIZEOF_JSAMPROW  ; output_data
+    dec         rcx                        ; rowctr
+    jg          short .rowloop
+
+.return:
+    uncollect_args 4
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_sse2(int max_v_samp_factor, JDIMENSION output_width,
+;                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11d = JDIMENSION output_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_h2v2_upsample_sse2)
+
+EXTN(jsimd_h2v2_upsample_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 4
+    push        rbx
+
+    mov         edx, r11d
+    add         rdx, byte (2*SIZEOF_XMMWORD)-1
+    and         rdx, byte -(2*SIZEOF_XMMWORD)
+    jz          near .return
+
+    mov         rcx, r10                ; rowctr
+    test        rcx, rcx
+    jz          near .return
+
+    mov         rsi, r12                ; input_data
+    mov         rdi, r13
+    mov         rdip, JSAMPARRAY rdi  ; output_data
+.rowloop:
+    push        rdi
+    push        rsi
+
+    mov         rsip, JSAMPROW rsi                   ; inptr
+    mov         rbxp, JSAMPROW rdi+0*SIZEOF_JSAMPROW ; outptr0
+    mov         rdip, JSAMPROW rdi+1*SIZEOF_JSAMPROW ; outptr1
+    mov         rax, rdx                               ; colctr
+.columnloop:
+
+    movdqa      xmm0, XMMWORD rsi+0*SIZEOF_XMMWORD
+
+    movdqa      xmm1, xmm0
+    punpcklbw   xmm0, xmm0
+    punpckhbw   xmm1, xmm1
+
+    movdqa      XMMWORD rbx+0*SIZEOF_XMMWORD, xmm0
+    movdqa      XMMWORD rbx+1*SIZEOF_XMMWORD, xmm1
+    movdqa      XMMWORD rdi+0*SIZEOF_XMMWORD, xmm0
+    movdqa      XMMWORD rdi+1*SIZEOF_XMMWORD, xmm1
+
+    sub         rax, byte 2*SIZEOF_XMMWORD
+    jz          short .nextrow
+
+    movdqa      xmm2, XMMWORD rsi+1*SIZEOF_XMMWORD
+
+    movdqa      xmm3, xmm2
+    punpcklbw   xmm2, xmm2
+    punpckhbw   xmm3, xmm3
+
+    movdqa      XMMWORD rbx+2*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD rbx+3*SIZEOF_XMMWORD, xmm3
+    movdqa      XMMWORD rdi+2*SIZEOF_XMMWORD, xmm2
+    movdqa      XMMWORD rdi+3*SIZEOF_XMMWORD, xmm3
+
+    sub         rax, byte 2*SIZEOF_XMMWORD
+    jz          short .nextrow
+
+    add         rsi, byte 2*SIZEOF_XMMWORD  ; inptr
+    add         rbx, byte 4*SIZEOF_XMMWORD  ; outptr0
+    add         rdi, byte 4*SIZEOF_XMMWORD  ; outptr1
+    jmp         short .columnloop
+
+.nextrow:
+    pop         rsi
+    pop         rdi
+
+    add         rsi, byte 1*SIZEOF_JSAMPROW  ; input_data
+    add         rdi, byte 2*SIZEOF_JSAMPROW  ; output_data
+    sub         rcx, byte 2                  ; rowctr
+    jg          near .rowloop
+
+.return:
+    pop         rbx
+    uncollect_args 4
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jfdctflt-sse.asm Added

@@ -0,0 +1,355 @@
+;
+; jfdctflt.asm - floating-point FDCT (64-bit SSE)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the forward DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jfdctflt.c; see the jfdctflt.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+    shufps      %1, %2, 0x44
+%endmacro
+
+%macro  unpckhps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+    shufps      %1, %2, 0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_float_sse)
+
+EXTN(jconst_fdct_float_sse):
+
+PD_0_382 times 4 dd 0.382683432365089771728460
+PD_0_707 times 4 dd 0.707106781186547524400844
+PD_0_541 times 4 dd 0.541196100146196984399723
+PD_1_306 times 4 dd 1.306562964876376527856643
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_float_sse(FAST_FLOAT *data)
+;
+
+; r10 = FAST_FLOAT *data
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD  ; xmmword wkWK_NUM
+%define WK_NUM  2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_float_sse)
+
+EXTN(jsimd_fdct_float_sse):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 1
+
+    ; ---- Pass 1: process rows.
+
+    mov         rdx, r10                ; (FAST_FLOAT *)
+    mov         rcx, DCTSIZE/4
+.rowloop:
+
+    movaps      xmm0, XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(2,1,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(3,1,rdx,SIZEOF_FAST_FLOAT)
+
+    ; xmm0=(20 21 22 23), xmm2=(24 25 26 27)
+    ; xmm1=(30 31 32 33), xmm3=(34 35 36 37)
+
+    movaps      xmm4, xmm0              ; transpose coefficients(phase 1)
+    unpcklps    xmm0, xmm1              ; xmm0=(20 30 21 31)
+    unpckhps    xmm4, xmm1              ; xmm4=(22 32 23 33)
+    movaps      xmm5, xmm2              ; transpose coefficients(phase 1)
+    unpcklps    xmm2, xmm3              ; xmm2=(24 34 25 35)
+    unpckhps    xmm5, xmm3              ; xmm5=(26 36 27 37)
+
+    movaps      xmm6, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm7, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)
+
+    ; xmm6=(00 01 02 03), xmm1=(04 05 06 07)
+    ; xmm7=(10 11 12 13), xmm3=(14 15 16 17)
+
+    movaps      XMMWORD wk(0), xmm4   ; wk(0)=(22 32 23 33)
+    movaps      XMMWORD wk(1), xmm2   ; wk(1)=(24 34 25 35)
+
+    movaps      xmm4, xmm6              ; transpose coefficients(phase 1)
+    unpcklps    xmm6, xmm7              ; xmm6=(00 10 01 11)
+    unpckhps    xmm4, xmm7              ; xmm4=(02 12 03 13)
+    movaps      xmm2, xmm1              ; transpose coefficients(phase 1)
+    unpcklps    xmm1, xmm3              ; xmm1=(04 14 05 15)
+    unpckhps    xmm2, xmm3              ; xmm2=(06 16 07 17)
+
+    movaps      xmm7, xmm6              ; transpose coefficients(phase 2)
+    unpcklps2   xmm6, xmm0              ; xmm6=(00 10 20 30)=data0
+    unpckhps2   xmm7, xmm0              ; xmm7=(01 11 21 31)=data1
+    movaps      xmm3, xmm2              ; transpose coefficients(phase 2)
+    unpcklps2   xmm2, xmm5              ; xmm2=(06 16 26 36)=data6
+    unpckhps2   xmm3, xmm5              ; xmm3=(07 17 27 37)=data7
+
+    movaps      xmm0, xmm7
+    movaps      xmm5, xmm6
+    subps       xmm7, xmm2              ; xmm7=data1-data6=tmp6
+    subps       xmm6, xmm3              ; xmm6=data0-data7=tmp7
+    addps       xmm0, xmm2              ; xmm0=data1+data6=tmp1
+    addps       xmm5, xmm3              ; xmm5=data0+data7=tmp0
+
+    movaps      xmm2, XMMWORD wk(0)   ; xmm2=(22 32 23 33)
+    movaps      xmm3, XMMWORD wk(1)   ; xmm3=(24 34 25 35)
+    movaps      XMMWORD wk(0), xmm7   ; wk(0)=tmp6
+    movaps      XMMWORD wk(1), xmm6   ; wk(1)=tmp7
+
+    movaps      xmm7, xmm4              ; transpose coefficients(phase 2)
+    unpcklps2   xmm4, xmm2              ; xmm4=(02 12 22 32)=data2
+    unpckhps2   xmm7, xmm2              ; xmm7=(03 13 23 33)=data3
+    movaps      xmm6, xmm1              ; transpose coefficients(phase 2)
+    unpcklps2   xmm1, xmm3              ; xmm1=(04 14 24 34)=data4
+    unpckhps2   xmm6, xmm3              ; xmm6=(05 15 25 35)=data5
+
+    movaps      xmm2, xmm7
+    movaps      xmm3, xmm4
+    addps       xmm7, xmm1              ; xmm7=data3+data4=tmp3
+    addps       xmm4, xmm6              ; xmm4=data2+data5=tmp2
+    subps       xmm2, xmm1              ; xmm2=data3-data4=tmp4
+    subps       xmm3, xmm6              ; xmm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movaps      xmm1, xmm5
+    movaps      xmm6, xmm0
+    subps       xmm5, xmm7              ; xmm5=tmp13
+    subps       xmm0, xmm4              ; xmm0=tmp12
+    addps       xmm1, xmm7              ; xmm1=tmp10
+    addps       xmm6, xmm4              ; xmm6=tmp11
+
+    addps       xmm0, xmm5
+    mulps       xmm0, rel PD_0_707    ; xmm0=z1
+
+    movaps      xmm7, xmm1
+    movaps      xmm4, xmm5
+    subps       xmm1, xmm6              ; xmm1=data4
+    subps       xmm5, xmm0              ; xmm5=data6
+    addps       xmm7, xmm6              ; xmm7=data0
+    addps       xmm4, xmm0              ; xmm4=data2
+
+    movaps      XMMWORD XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(2,1,rdx,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT), xmm7
+    movaps      XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT), xmm4
+
+    ; -- Odd part
+
+    movaps      xmm6, XMMWORD wk(0)   ; xmm6=tmp6
+    movaps      xmm0, XMMWORD wk(1)   ; xmm0=tmp7
+
+    addps       xmm2, xmm3              ; xmm2=tmp10
+    addps       xmm3, xmm6              ; xmm3=tmp11
+    addps       xmm6, xmm0              ; xmm6=tmp12, xmm0=tmp7
+
+    mulps       xmm3, rel PD_0_707    ; xmm3=z3
+
+    movaps      xmm1, xmm2              ; xmm1=tmp10
+    subps       xmm2, xmm6
+    mulps       xmm2, rel PD_0_382    ; xmm2=z5
+    mulps       xmm1, rel PD_0_541    ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+    mulps       xmm6, rel PD_1_306    ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+    addps       xmm1, xmm2              ; xmm1=z2
+    addps       xmm6, xmm2              ; xmm6=z4
+
+    movaps      xmm5, xmm0
+    subps       xmm0, xmm3              ; xmm0=z13
+    addps       xmm5, xmm3              ; xmm5=z11
+
+    movaps      xmm7, xmm0
+    movaps      xmm4, xmm5
+    subps       xmm0, xmm1              ; xmm0=data3
+    subps       xmm5, xmm6              ; xmm5=data7
+    addps       xmm7, xmm1              ; xmm7=data5
+    addps       xmm4, xmm6              ; xmm4=data1
+
+    movaps      XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(3,1,rdx,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT), xmm7
+    movaps      XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT), xmm4
+
+    add         rdx, 4*DCTSIZE*SIZEOF_FAST_FLOAT
+    dec         rcx
+    jnz         near .rowloop
+
+    ; ---- Pass 2: process columns.
+
+    mov         rdx, r10                ; (FAST_FLOAT *)
+    mov         rcx, DCTSIZE/4
+.columnloop:
+
+    movaps      xmm0, XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_FAST_FLOAT)
+
+    ; xmm0=(02 12 22 32), xmm2=(42 52 62 72)
+    ; xmm1=(03 13 23 33), xmm3=(43 53 63 73)
+
+    movaps      xmm4, xmm0              ; transpose coefficients(phase 1)
+    unpcklps    xmm0, xmm1              ; xmm0=(02 03 12 13)
+    unpckhps    xmm4, xmm1              ; xmm4=(22 23 32 33)
+    movaps      xmm5, xmm2              ; transpose coefficients(phase 1)
+    unpcklps    xmm2, xmm3              ; xmm2=(42 43 52 53)
+    unpckhps    xmm5, xmm3              ; xmm5=(62 63 72 73)
+
+    movaps      xmm6, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm7, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(4,0,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_FAST_FLOAT)
+
+    ; xmm6=(00 10 20 30), xmm1=(40 50 60 70)
+    ; xmm7=(01 11 21 31), xmm3=(41 51 61 71)
+
+    movaps      XMMWORD wk(0), xmm4   ; wk(0)=(22 23 32 33)
+    movaps      XMMWORD wk(1), xmm2   ; wk(1)=(42 43 52 53)
+
+    movaps      xmm4, xmm6              ; transpose coefficients(phase 1)
+    unpcklps    xmm6, xmm7              ; xmm6=(00 01 10 11)
+    unpckhps    xmm4, xmm7              ; xmm4=(20 21 30 31)
+    movaps      xmm2, xmm1              ; transpose coefficients(phase 1)
+    unpcklps    xmm1, xmm3              ; xmm1=(40 41 50 51)
+    unpckhps    xmm2, xmm3              ; xmm2=(60 61 70 71)
+
+    movaps      xmm7, xmm6              ; transpose coefficients(phase 2)
+    unpcklps2   xmm6, xmm0              ; xmm6=(00 01 02 03)=data0
+    unpckhps2   xmm7, xmm0              ; xmm7=(10 11 12 13)=data1
+    movaps      xmm3, xmm2              ; transpose coefficients(phase 2)
+    unpcklps2   xmm2, xmm5              ; xmm2=(60 61 62 63)=data6
+    unpckhps2   xmm3, xmm5              ; xmm3=(70 71 72 73)=data7
+
+    movaps      xmm0, xmm7
+    movaps      xmm5, xmm6
+    subps       xmm7, xmm2              ; xmm7=data1-data6=tmp6
+    subps       xmm6, xmm3              ; xmm6=data0-data7=tmp7
+    addps       xmm0, xmm2              ; xmm0=data1+data6=tmp1
+    addps       xmm5, xmm3              ; xmm5=data0+data7=tmp0
+
+    movaps      xmm2, XMMWORD wk(0)   ; xmm2=(22 23 32 33)
+    movaps      xmm3, XMMWORD wk(1)   ; xmm3=(42 43 52 53)
+    movaps      XMMWORD wk(0), xmm7   ; wk(0)=tmp6
+    movaps      XMMWORD wk(1), xmm6   ; wk(1)=tmp7
+
+    movaps      xmm7, xmm4              ; transpose coefficients(phase 2)
+    unpcklps2   xmm4, xmm2              ; xmm4=(20 21 22 23)=data2
+    unpckhps2   xmm7, xmm2              ; xmm7=(30 31 32 33)=data3
+    movaps      xmm6, xmm1              ; transpose coefficients(phase 2)
+    unpcklps2   xmm1, xmm3              ; xmm1=(40 41 42 43)=data4
+    unpckhps2   xmm6, xmm3              ; xmm6=(50 51 52 53)=data5
+
+    movaps      xmm2, xmm7
+    movaps      xmm3, xmm4
+    addps       xmm7, xmm1              ; xmm7=data3+data4=tmp3
+    addps       xmm4, xmm6              ; xmm4=data2+data5=tmp2
+    subps       xmm2, xmm1              ; xmm2=data3-data4=tmp4
+    subps       xmm3, xmm6              ; xmm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movaps      xmm1, xmm5
+    movaps      xmm6, xmm0
+    subps       xmm5, xmm7              ; xmm5=tmp13
+    subps       xmm0, xmm4              ; xmm0=tmp12
+    addps       xmm1, xmm7              ; xmm1=tmp10
+    addps       xmm6, xmm4              ; xmm6=tmp11
+
+    addps       xmm0, xmm5
+    mulps       xmm0, rel PD_0_707    ; xmm0=z1
+
+    movaps      xmm7, xmm1
+    movaps      xmm4, xmm5
+    subps       xmm1, xmm6              ; xmm1=data4
+    subps       xmm5, xmm0              ; xmm5=data6
+    addps       xmm7, xmm6              ; xmm7=data0
+    addps       xmm4, xmm0              ; xmm4=data2
+
+    movaps      XMMWORD XMMBLOCK(4,0,rdx,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT), xmm7
+    movaps      XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT), xmm4
+
+    ; -- Odd part
+
+    movaps      xmm6, XMMWORD wk(0)   ; xmm6=tmp6
+    movaps      xmm0, XMMWORD wk(1)   ; xmm0=tmp7
+
+    addps       xmm2, xmm3              ; xmm2=tmp10
+    addps       xmm3, xmm6              ; xmm3=tmp11
+    addps       xmm6, xmm0              ; xmm6=tmp12, xmm0=tmp7
+
+    mulps       xmm3, rel PD_0_707    ; xmm3=z3
+
+    movaps      xmm1, xmm2              ; xmm1=tmp10
+    subps       xmm2, xmm6
+    mulps       xmm2, rel PD_0_382    ; xmm2=z5
+    mulps       xmm1, rel PD_0_541    ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+    mulps       xmm6, rel PD_1_306    ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+    addps       xmm1, xmm2              ; xmm1=z2
+    addps       xmm6, xmm2              ; xmm6=z4
+
+    movaps      xmm5, xmm0
+    subps       xmm0, xmm3              ; xmm0=z13
+    addps       xmm5, xmm3              ; xmm5=z11
+
+    movaps      xmm7, xmm0
+    movaps      xmm4, xmm5
+    subps       xmm0, xmm1              ; xmm0=data3
+    subps       xmm5, xmm6              ; xmm5=data7
+    addps       xmm7, xmm1              ; xmm7=data5
+    addps       xmm4, xmm6              ; xmm4=data1
+
+    movaps      XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_FAST_FLOAT), xmm7
+    movaps      XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT), xmm4
+
+    add         rdx, byte 4*SIZEOF_FAST_FLOAT
+    dec         rcx
+    jnz         near .columnloop
+
+    uncollect_args 1
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jfdctfst-sse2.asm Added

@@ -0,0 +1,389 @@
+;
+; jfdctfst.asm - fast integer FDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the forward DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
+; for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  8  ; 14 is also OK.
+
+%if CONST_BITS == 8
+F_0_382 equ  98  ; FIX(0.382683433)
+F_0_541 equ 139  ; FIX(0.541196100)
+F_0_707 equ 181  ; FIX(0.707106781)
+F_1_306 equ 334  ; FIX(1.306562965)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_382 equ DESCALE( 410903207, 30 - CONST_BITS)  ; FIX(0.382683433)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_707 equ DESCALE( 759250124, 30 - CONST_BITS)  ; FIX(0.707106781)
+F_1_306 equ DESCALE(1402911301, 30 - CONST_BITS)  ; FIX(1.306562965)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS  2
+%define CONST_SHIFT              (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_ifast_sse2)
+
+EXTN(jconst_fdct_ifast_sse2):
+
+PW_F0707 times 8 dw F_0_707 << CONST_SHIFT
+PW_F0382 times 8 dw F_0_382 << CONST_SHIFT
+PW_F0541 times 8 dw F_0_541 << CONST_SHIFT
+PW_F1306 times 8 dw F_1_306 << CONST_SHIFT
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_ifast_sse2(DCTELEM *data)
+;
+
+; r10 = DCTELEM *data
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD  ; xmmword wkWK_NUM
+%define WK_NUM  2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_ifast_sse2)
+
+EXTN(jsimd_fdct_ifast_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 1
+
+    ; ---- Pass 1: process rows.
+
+    mov         rdx, r10                ; (DCTELEM *)
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm1, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm2, XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm3, XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)
+
+    ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+    ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+    movdqa      xmm4, xmm0              ; transpose coefficients(phase 1)
+    punpcklwd   xmm0, xmm1              ; xmm0=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm4, xmm1              ; xmm4=(04 14 05 15 06 16 07 17)
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 1)
+    punpcklwd   xmm2, xmm3              ; xmm2=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm5, xmm3              ; xmm5=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm6, XMMWORD XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm7, XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm1, XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)
+
+    ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+    ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+    movdqa      XMMWORD wk(1), xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm2, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm7              ; xmm6=(40 50 41 51 42 52 43 53)
+    punpckhwd   xmm2, xmm7              ; xmm2=(44 54 45 55 46 56 47 57)
+    movdqa      xmm5, xmm1              ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm3              ; xmm1=(60 70 61 71 62 72 63 73)
+    punpckhwd   xmm5, xmm3              ; xmm5=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm7, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm1              ; xmm6=(40 50 60 70 41 51 61 71)
+    punpckhdq   xmm7, xmm1              ; xmm7=(42 52 62 72 43 53 63 73)
+    movdqa      xmm3, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm5              ; xmm2=(44 54 64 74 45 55 65 75)
+    punpckhdq   xmm3, xmm5              ; xmm3=(46 56 66 76 47 57 67 77)
+
+    movdqa      xmm1, XMMWORD wk(0)   ; xmm1=(20 30 21 31 22 32 23 33)
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=(24 34 25 35 26 36 27 37)
+    movdqa      XMMWORD wk(0), xmm7   ; wk(0)=(42 52 62 72 43 53 63 73)
+    movdqa      XMMWORD wk(1), xmm2   ; wk(1)=(44 54 64 74 45 55 65 75)
+
+    movdqa      xmm7, xmm0              ; transpose coefficients(phase 2)
+    punpckldq   xmm0, xmm1              ; xmm0=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm7, xmm1              ; xmm7=(02 12 22 32 03 13 23 33)
+    movdqa      xmm2, xmm4              ; transpose coefficients(phase 2)
+    punpckldq   xmm4, xmm5              ; xmm4=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm2, xmm5              ; xmm2=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm1, xmm0              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm0, xmm6              ; xmm0=(00 10 20 30 40 50 60 70)=data0
+    punpckhqdq  xmm1, xmm6              ; xmm1=(01 11 21 31 41 51 61 71)=data1
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm2, xmm3              ; xmm2=(06 16 26 36 46 56 66 76)=data6
+    punpckhqdq  xmm5, xmm3              ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+    movdqa      xmm6, xmm1
+    movdqa      xmm3, xmm0
+    psubw       xmm1, xmm2              ; xmm1=data1-data6=tmp6
+    psubw       xmm0, xmm5              ; xmm0=data0-data7=tmp7
+    paddw       xmm6, xmm2              ; xmm6=data1+data6=tmp1
+    paddw       xmm3, xmm5              ; xmm3=data0+data7=tmp0
+
+    movdqa      xmm2, XMMWORD wk(0)   ; xmm2=(42 52 62 72 43 53 63 73)
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=(44 54 64 74 45 55 65 75)
+    movdqa      XMMWORD wk(0), xmm1   ; wk(0)=tmp6
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=tmp7
+
+    movdqa      xmm1, xmm7              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm7, xmm2              ; xmm7=(02 12 22 32 42 52 62 72)=data2
+    punpckhqdq  xmm1, xmm2              ; xmm1=(03 13 23 33 43 53 63 73)=data3
+    movdqa      xmm0, xmm4              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm4, xmm5              ; xmm4=(04 14 24 34 44 54 64 74)=data4
+    punpckhqdq  xmm0, xmm5              ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+    movdqa      xmm2, xmm1
+    movdqa      xmm5, xmm7
+    paddw       xmm1, xmm4              ; xmm1=data3+data4=tmp3
+    paddw       xmm7, xmm0              ; xmm7=data2+data5=tmp2
+    psubw       xmm2, xmm4              ; xmm2=data3-data4=tmp4
+    psubw       xmm5, xmm0              ; xmm5=data2-data5=tmp5
+
+    ; -- Even part
+
+    movdqa      xmm4, xmm3
+    movdqa      xmm0, xmm6
+    psubw       xmm3, xmm1              ; xmm3=tmp13
+    psubw       xmm6, xmm7              ; xmm6=tmp12
+    paddw       xmm4, xmm1              ; xmm4=tmp10
+    paddw       xmm0, xmm7              ; xmm0=tmp11
+
+    paddw       xmm6, xmm3
+    psllw       xmm6, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm6, rel PW_F0707    ; xmm6=z1
+
+    movdqa      xmm1, xmm4
+    movdqa      xmm7, xmm3
+    psubw       xmm4, xmm0              ; xmm4=data4
+    psubw       xmm3, xmm6              ; xmm3=data6
+    paddw       xmm1, xmm0              ; xmm1=data0
+    paddw       xmm7, xmm6              ; xmm7=data2
+
+    movdqa      xmm0, XMMWORD wk(0)   ; xmm0=tmp6
+    movdqa      xmm6, XMMWORD wk(1)   ; xmm6=tmp7
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=data4
+    movdqa      XMMWORD wk(1), xmm3   ; wk(1)=data6
+
+    ; -- Odd part
+
+    paddw       xmm2, xmm5              ; xmm2=tmp10
+    paddw       xmm5, xmm0              ; xmm5=tmp11
+    paddw       xmm0, xmm6              ; xmm0=tmp12, xmm6=tmp7
+
+    psllw       xmm2, PRE_MULTIPLY_SCALE_BITS
+    psllw       xmm0, PRE_MULTIPLY_SCALE_BITS
+
+    psllw       xmm5, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm5, rel PW_F0707    ; xmm5=z3
+
+    movdqa      xmm4, xmm2              ; xmm4=tmp10
+    psubw       xmm2, xmm0
+    pmulhw      xmm2, rel PW_F0382    ; xmm2=z5
+    pmulhw      xmm4, rel PW_F0541    ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+    pmulhw      xmm0, rel PW_F1306    ; xmm0=MULTIPLY(tmp12,FIX_1_306562)
+    paddw       xmm4, xmm2              ; xmm4=z2
+    paddw       xmm0, xmm2              ; xmm0=z4
+
+    movdqa      xmm3, xmm6
+    psubw       xmm6, xmm5              ; xmm6=z13
+    paddw       xmm3, xmm5              ; xmm3=z11
+
+    movdqa      xmm2, xmm6
+    movdqa      xmm5, xmm3
+    psubw       xmm6, xmm4              ; xmm6=data3
+    psubw       xmm3, xmm0              ; xmm3=data7
+    paddw       xmm2, xmm4              ; xmm2=data5
+    paddw       xmm5, xmm0              ; xmm5=data1
+
+    ; ---- Pass 2: process columns.
+
+    ; xmm1=(00 10 20 30 40 50 60 70), xmm7=(02 12 22 32 42 52 62 72)
+    ; xmm5=(01 11 21 31 41 51 61 71), xmm6=(03 13 23 33 43 53 63 73)
+
+    movdqa      xmm4, xmm1              ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm5              ; xmm1=(00 01 10 11 20 21 30 31)
+    punpckhwd   xmm4, xmm5              ; xmm4=(40 41 50 51 60 61 70 71)
+    movdqa      xmm0, xmm7              ; transpose coefficients(phase 1)
+    punpcklwd   xmm7, xmm6              ; xmm7=(02 03 12 13 22 23 32 33)
+    punpckhwd   xmm0, xmm6              ; xmm0=(42 43 52 53 62 63 72 73)
+
+    movdqa      xmm5, XMMWORD wk(0)   ; xmm5=col4
+    movdqa      xmm6, XMMWORD wk(1)   ; xmm6=col6
+
+    ; xmm5=(04 14 24 34 44 54 64 74), xmm6=(06 16 26 36 46 56 66 76)
+    ; xmm2=(05 15 25 35 45 55 65 75), xmm3=(07 17 27 37 47 57 67 77)
+
+    movdqa      XMMWORD wk(0), xmm7   ; wk(0)=(02 03 12 13 22 23 32 33)
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+    movdqa      xmm7, xmm5              ; transpose coefficients(phase 1)
+    punpcklwd   xmm5, xmm2              ; xmm5=(04 05 14 15 24 25 34 35)
+    punpckhwd   xmm7, xmm2              ; xmm7=(44 45 54 55 64 65 74 75)
+    movdqa      xmm0, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm3              ; xmm6=(06 07 16 17 26 27 36 37)
+    punpckhwd   xmm0, xmm3              ; xmm0=(46 47 56 57 66 67 76 77)
+
+    movdqa      xmm2, xmm5              ; transpose coefficients(phase 2)
+    punpckldq   xmm5, xmm6              ; xmm5=(04 05 06 07 14 15 16 17)
+    punpckhdq   xmm2, xmm6              ; xmm2=(24 25 26 27 34 35 36 37)
+    movdqa      xmm3, xmm7              ; transpose coefficients(phase 2)
+    punpckldq   xmm7, xmm0              ; xmm7=(44 45 46 47 54 55 56 57)
+    punpckhdq   xmm3, xmm0              ; xmm3=(64 65 66 67 74 75 76 77)
+
+    movdqa      xmm6, XMMWORD wk(0)   ; xmm6=(02 03 12 13 22 23 32 33)
+    movdqa      xmm0, XMMWORD wk(1)   ; xmm0=(42 43 52 53 62 63 72 73)
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=(24 25 26 27 34 35 36 37)
+    movdqa      XMMWORD wk(1), xmm7   ; wk(1)=(44 45 46 47 54 55 56 57)
+
+    movdqa      xmm2, xmm1              ; transpose coefficients(phase 2)
+    punpckldq   xmm1, xmm6              ; xmm1=(00 01 02 03 10 11 12 13)
+    punpckhdq   xmm2, xmm6              ; xmm2=(20 21 22 23 30 31 32 33)
+    movdqa      xmm7, xmm4              ; transpose coefficients(phase 2)
+    punpckldq   xmm4, xmm0              ; xmm4=(40 41 42 43 50 51 52 53)
+    punpckhdq   xmm7, xmm0              ; xmm7=(60 61 62 63 70 71 72 73)
+
+    movdqa      xmm6, xmm1              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm1, xmm5              ; xmm1=(00 01 02 03 04 05 06 07)=data0
+    punpckhqdq  xmm6, xmm5              ; xmm6=(10 11 12 13 14 15 16 17)=data1
+    movdqa      xmm0, xmm7              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm7, xmm3              ; xmm7=(60 61 62 63 64 65 66 67)=data6
+    punpckhqdq  xmm0, xmm3              ; xmm0=(70 71 72 73 74 75 76 77)=data7
+
+    movdqa      xmm5, xmm6
+    movdqa      xmm3, xmm1
+    psubw       xmm6, xmm7              ; xmm6=data1-data6=tmp6
+    psubw       xmm1, xmm0              ; xmm1=data0-data7=tmp7
+    paddw       xmm5, xmm7              ; xmm5=data1+data6=tmp1
+    paddw       xmm3, xmm0              ; xmm3=data0+data7=tmp0
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=(24 25 26 27 34 35 36 37)
+    movdqa      xmm0, XMMWORD wk(1)   ; xmm0=(44 45 46 47 54 55 56 57)
+    movdqa      XMMWORD wk(0), xmm6   ; wk(0)=tmp6
+    movdqa      XMMWORD wk(1), xmm1   ; wk(1)=tmp7
+
+    movdqa      xmm6, xmm2              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm2, xmm7              ; xmm2=(20 21 22 23 24 25 26 27)=data2
+    punpckhqdq  xmm6, xmm7              ; xmm6=(30 31 32 33 34 35 36 37)=data3
+    movdqa      xmm1, xmm4              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm4, xmm0              ; xmm4=(40 41 42 43 44 45 46 47)=data4
+    punpckhqdq  xmm1, xmm0              ; xmm1=(50 51 52 53 54 55 56 57)=data5
+
+    movdqa      xmm7, xmm6
+    movdqa      xmm0, xmm2
+    paddw       xmm6, xmm4              ; xmm6=data3+data4=tmp3
+    paddw       xmm2, xmm1              ; xmm2=data2+data5=tmp2
+    psubw       xmm7, xmm4              ; xmm7=data3-data4=tmp4
+    psubw       xmm0, xmm1              ; xmm0=data2-data5=tmp5
+
+    ; -- Even part
+
+    movdqa      xmm4, xmm3
+    movdqa      xmm1, xmm5
+    psubw       xmm3, xmm6              ; xmm3=tmp13
+    psubw       xmm5, xmm2              ; xmm5=tmp12
+    paddw       xmm4, xmm6              ; xmm4=tmp10
+    paddw       xmm1, xmm2              ; xmm1=tmp11
+
+    paddw       xmm5, xmm3
+    psllw       xmm5, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm5, rel PW_F0707    ; xmm5=z1
+
+    movdqa      xmm6, xmm4
+    movdqa      xmm2, xmm3
+    psubw       xmm4, xmm1              ; xmm4=data4
+    psubw       xmm3, xmm5              ; xmm3=data6
+    paddw       xmm6, xmm1              ; xmm6=data0
+    paddw       xmm2, xmm5              ; xmm2=data2
+
+    movdqa      XMMWORD XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM), xmm4
+    movdqa      XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM), xmm3
+    movdqa      XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM), xmm6
+    movdqa      XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM), xmm2
+
+    ; -- Odd part
+
+    movdqa      xmm1, XMMWORD wk(0)   ; xmm1=tmp6
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=tmp7
+
+    paddw       xmm7, xmm0              ; xmm7=tmp10
+    paddw       xmm0, xmm1              ; xmm0=tmp11
+    paddw       xmm1, xmm5              ; xmm1=tmp12, xmm5=tmp7
+
+    psllw       xmm7, PRE_MULTIPLY_SCALE_BITS
+    psllw       xmm1, PRE_MULTIPLY_SCALE_BITS
+
+    psllw       xmm0, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm0, rel PW_F0707    ; xmm0=z3
+
+    movdqa      xmm4, xmm7              ; xmm4=tmp10
+    psubw       xmm7, xmm1
+    pmulhw      xmm7, rel PW_F0382    ; xmm7=z5
+    pmulhw      xmm4, rel PW_F0541    ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+    pmulhw      xmm1, rel PW_F1306    ; xmm1=MULTIPLY(tmp12,FIX_1_306562)
+    paddw       xmm4, xmm7              ; xmm4=z2
+    paddw       xmm1, xmm7              ; xmm1=z4
+
+    movdqa      xmm3, xmm5
+    psubw       xmm5, xmm0              ; xmm5=z13
+    paddw       xmm3, xmm0              ; xmm3=z11
+
+    movdqa      xmm6, xmm5
+    movdqa      xmm2, xmm3
+    psubw       xmm5, xmm4              ; xmm5=data3
+    psubw       xmm3, xmm1              ; xmm3=data7
+    paddw       xmm6, xmm4              ; xmm6=data5
+    paddw       xmm2, xmm1              ; xmm2=data1
+
+    movdqa      XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM), xmm5
+    movdqa      XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM), xmm3
+    movdqa      XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM), xmm6
+    movdqa      XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM), xmm2
+
+    uncollect_args 1
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jfdctint-avx2.asm Added

@@ -0,0 +1,320 @@
+;
+; jfdctint.asm - accurate integer FDCT (64-bit AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, 2018, 2020, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+; In-place 8x8x16-bit matrix transpose using AVX2 instructions
+; %1-%4: Input/output registers
+; %5-%8: Temp registers
+
+%macro dotranspose 8
+    ; %1=(00 01 02 03 04 05 06 07  40 41 42 43 44 45 46 47)
+    ; %2=(10 11 12 13 14 15 16 17  50 51 52 53 54 55 56 57)
+    ; %3=(20 21 22 23 24 25 26 27  60 61 62 63 64 65 66 67)
+    ; %4=(30 31 32 33 34 35 36 37  70 71 72 73 74 75 76 77)
+
+    vpunpcklwd  %5, %1, %2
+    vpunpckhwd  %6, %1, %2
+    vpunpcklwd  %7, %3, %4
+    vpunpckhwd  %8, %3, %4
+    ; transpose coefficients(phase 1)
+    ; %5=(00 10 01 11 02 12 03 13  40 50 41 51 42 52 43 53)
+    ; %6=(04 14 05 15 06 16 07 17  44 54 45 55 46 56 47 57)
+    ; %7=(20 30 21 31 22 32 23 33  60 70 61 71 62 72 63 73)
+    ; %8=(24 34 25 35 26 36 27 37  64 74 65 75 66 76 67 77)
+
+    vpunpckldq  %1, %5, %7
+    vpunpckhdq  %2, %5, %7
+    vpunpckldq  %3, %6, %8
+    vpunpckhdq  %4, %6, %8
+    ; transpose coefficients(phase 2)
+    ; %1=(00 10 20 30 01 11 21 31  40 50 60 70 41 51 61 71)
+    ; %2=(02 12 22 32 03 13 23 33  42 52 62 72 43 53 63 73)
+    ; %3=(04 14 24 34 05 15 25 35  44 54 64 74 45 55 65 75)
+    ; %4=(06 16 26 36 07 17 27 37  46 56 66 76 47 57 67 77)
+
+    vpermq      %1, %1, 0x8D
+    vpermq      %2, %2, 0x8D
+    vpermq      %3, %3, 0xD8
+    vpermq      %4, %4, 0xD8
+    ; transpose coefficients(phase 3)
+    ; %1=(01 11 21 31 41 51 61 71  00 10 20 30 40 50 60 70)
+    ; %2=(03 13 23 33 43 53 63 73  02 12 22 32 42 52 62 72)
+    ; %3=(04 14 24 34 44 54 64 74  05 15 25 35 45 55 65 75)
+    ; %4=(06 16 26 36 46 56 66 76  07 17 27 37 47 57 67 77)
+%endmacro
+
+; --------------------------------------------------------------------------
+; In-place 8x8x16-bit accurate integer forward DCT using AVX2 instructions
+; %1-%4: Input/output registers
+; %5-%8: Temp registers
+; %9:    Pass (1 or 2)
+
+%macro dodct 9
+    vpsubw      %5, %1, %4              ; %5=data1_0-data6_7=tmp6_7
+    vpaddw      %6, %1, %4              ; %6=data1_0+data6_7=tmp1_0
+    vpaddw      %7, %2, %3              ; %7=data3_2+data4_5=tmp3_2
+    vpsubw      %8, %2, %3              ; %8=data3_2-data4_5=tmp4_5
+
+    ; -- Even part
+
+    vperm2i128  %6, %6, %6, 0x01        ; %6=tmp0_1
+    vpaddw      %1, %6, %7              ; %1=tmp0_1+tmp3_2=tmp10_11
+    vpsubw      %6, %6, %7              ; %6=tmp0_1-tmp3_2=tmp13_12
+
+    vperm2i128  %7, %1, %1, 0x01        ; %7=tmp11_10
+    vpsignw     %1, %1, rel PW_1_NEG1  ; %1=tmp10_neg11
+    vpaddw      %7, %7, %1              ; %7=(tmp10+tmp11)_(tmp10-tmp11)
+%if %9 == 1
+    vpsllw      %1, %7, PASS1_BITS      ; %1=data0_4
+%else
+    vpaddw      %7, %7, rel PW_DESCALE_P2X
+    vpsraw      %1, %7, PASS1_BITS      ; %1=data0_4
+%endif
+
+    ; (Original)
+    ; z1 = (tmp12 + tmp13) * 0.541196100;
+    ; data2 = z1 + tmp13 * 0.765366865;
+    ; data6 = z1 + tmp12 * -1.847759065;
+    ;
+    ; (This implementation)
+    ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+    ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+    vperm2i128  %7, %6, %6, 0x01        ; %7=tmp12_13
+    vpunpcklwd  %2, %6, %7
+    vpunpckhwd  %6, %6, %7
+    vpmaddwd    %2, %2, rel PW_F130_F054_MF130_F054  ; %2=data2_6L
+    vpmaddwd    %6, %6, rel PW_F130_F054_MF130_F054  ; %6=data2_6H
+
+    vpaddd      %2, %2, rel PD_DESCALE_P %+ %9
+    vpaddd      %6, %6, rel PD_DESCALE_P %+ %9
+    vpsrad      %2, %2, DESCALE_P %+ %9
+    vpsrad      %6, %6, DESCALE_P %+ %9
+
+    vpackssdw   %3, %2, %6              ; %6=data2_6
+
+    ; -- Odd part
+
+    vpaddw      %7, %8, %5              ; %7=tmp4_5+tmp6_7=z3_4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    vperm2i128  %2, %7, %7, 0x01        ; %2=z4_3
+    vpunpcklwd  %6, %7, %2
+    vpunpckhwd  %7, %7, %2
+    vpmaddwd    %6, %6, rel PW_MF078_F117_F078_F117  ; %6=z3_4L
+    vpmaddwd    %7, %7, rel PW_MF078_F117_F078_F117  ; %7=z3_4H
+
+    ; (Original)
+    ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+    ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+    ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+    ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+    ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+    ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+    ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+    ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+    ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+    vperm2i128  %4, %5, %5, 0x01        ; %4=tmp7_6
+    vpunpcklwd  %2, %8, %4
+    vpunpckhwd  %4, %8, %4
+    vpmaddwd    %2, %2, rel PW_MF060_MF089_MF050_MF256  ; %2=tmp4_5L
+    vpmaddwd    %4, %4, rel PW_MF060_MF089_MF050_MF256  ; %4=tmp4_5H
+
+    vpaddd      %2, %2, %6              ; %2=data7_5L
+    vpaddd      %4, %4, %7              ; %4=data7_5H
+
+    vpaddd      %2, %2, rel PD_DESCALE_P %+ %9
+    vpaddd      %4, %4, rel PD_DESCALE_P %+ %9
+    vpsrad      %2, %2, DESCALE_P %+ %9
+    vpsrad      %4, %4, DESCALE_P %+ %9
+
+    vpackssdw   %4, %2, %4              ; %4=data7_5
+
+    vperm2i128  %2, %8, %8, 0x01        ; %2=tmp5_4
+    vpunpcklwd  %8, %5, %2
+    vpunpckhwd  %5, %5, %2
+    vpmaddwd    %8, %8, rel PW_F050_MF256_F060_MF089  ; %8=tmp6_7L
+    vpmaddwd    %5, %5, rel PW_F050_MF256_F060_MF089  ; %5=tmp6_7H
+
+    vpaddd      %8, %8, %6              ; %8=data3_1L
+    vpaddd      %5, %5, %7              ; %5=data3_1H
+
+    vpaddd      %8, %8, rel PD_DESCALE_P %+ %9
+    vpaddd      %5, %5, rel PD_DESCALE_P %+ %9
+    vpsrad      %8, %8, DESCALE_P %+ %9
+    vpsrad      %5, %5, DESCALE_P %+ %9
+
+    vpackssdw   %2, %8, %5              ; %2=data3_1
+%endmacro
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_islow_avx2)
+
+EXTN(jconst_fdct_islow_avx2):
+
+PW_F130_F054_MF130_F054    times 4  dw  (F_0_541 + F_0_765),  F_0_541
+                           times 4  dw  (F_0_541 - F_1_847),  F_0_541
+PW_MF078_F117_F078_F117    times 4  dw  (F_1_175 - F_1_961),  F_1_175
+                           times 4  dw  (F_1_175 - F_0_390),  F_1_175
+PW_MF060_MF089_MF050_MF256 times 4  dw  (F_0_298 - F_0_899), -F_0_899
+                           times 4  dw  (F_2_053 - F_2_562), -F_2_562
+PW_F050_MF256_F060_MF089   times 4  dw  (F_3_072 - F_2_562), -F_2_562
+                           times 4  dw  (F_1_501 - F_0_899), -F_0_899
+PD_DESCALE_P1              times 8  dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2              times 8  dd  1 << (DESCALE_P2 - 1)
+PW_DESCALE_P2X             times 16 dw  1 << (PASS1_BITS - 1)
+PW_1_NEG1                  times 8  dw  1
+                           times 8  dw -1
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_avx2(DCTELEM *data)
+;
+
+; r10 = DCTELEM *data
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_islow_avx2)
+
+EXTN(jsimd_fdct_islow_avx2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 1
+
+    ; ---- Pass 1: process rows.
+
+    vmovdqu     ymm4, YMMWORD YMMBLOCK(0,0,r10,SIZEOF_DCTELEM)
+    vmovdqu     ymm5, YMMWORD YMMBLOCK(2,0,r10,SIZEOF_DCTELEM)
+    vmovdqu     ymm6, YMMWORD YMMBLOCK(4,0,r10,SIZEOF_DCTELEM)
+    vmovdqu     ymm7, YMMWORD YMMBLOCK(6,0,r10,SIZEOF_DCTELEM)
+    ; ymm4=(00 01 02 03 04 05 06 07  10 11 12 13 14 15 16 17)
+    ; ymm5=(20 21 22 23 24 25 26 27  30 31 32 33 34 35 36 37)
+    ; ymm6=(40 41 42 43 44 45 46 47  50 51 52 53 54 55 56 57)
+    ; ymm7=(60 61 62 63 64 65 66 67  70 71 72 73 74 75 76 77)
+
+    vperm2i128  ymm0, ymm4, ymm6, 0x20
+    vperm2i128  ymm1, ymm4, ymm6, 0x31
+    vperm2i128  ymm2, ymm5, ymm7, 0x20
+    vperm2i128  ymm3, ymm5, ymm7, 0x31
+    ; ymm0=(00 01 02 03 04 05 06 07  40 41 42 43 44 45 46 47)
+    ; ymm1=(10 11 12 13 14 15 16 17  50 51 52 53 54 55 56 57)
+    ; ymm2=(20 21 22 23 24 25 26 27  60 61 62 63 64 65 66 67)
+    ; ymm3=(30 31 32 33 34 35 36 37  70 71 72 73 74 75 76 77)
+
+    dotranspose ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7
+
+    dodct       ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7, 1
+    ; ymm0=data0_4, ymm1=data3_1, ymm2=data2_6, ymm3=data7_5
+
+    ; ---- Pass 2: process columns.
+
+    vperm2i128  ymm4, ymm1, ymm3, 0x20  ; ymm4=data3_7
+    vperm2i128  ymm1, ymm1, ymm3, 0x31  ; ymm1=data1_5
+
+    dotranspose ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7
+
+    dodct       ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7, 2
+    ; ymm0=data0_4, ymm1=data3_1, ymm2=data2_6, ymm4=data7_5
+
+    vperm2i128 ymm3, ymm0, ymm1, 0x30   ; ymm3=data0_1
+    vperm2i128 ymm5, ymm2, ymm1, 0x20   ; ymm5=data2_3
+    vperm2i128 ymm6, ymm0, ymm4, 0x31   ; ymm6=data4_5
+    vperm2i128 ymm7, ymm2, ymm4, 0x21   ; ymm7=data6_7
+
+    vmovdqu     YMMWORD YMMBLOCK(0,0,r10,SIZEOF_DCTELEM), ymm3
+    vmovdqu     YMMWORD YMMBLOCK(2,0,r10,SIZEOF_DCTELEM), ymm5
+    vmovdqu     YMMWORD YMMBLOCK(4,0,r10,SIZEOF_DCTELEM), ymm6
+    vmovdqu     YMMWORD YMMBLOCK(6,0,r10,SIZEOF_DCTELEM), ymm7
+
+    vzeroupper
+    uncollect_args 1
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jfdctint-sse2.asm Added

@@ -0,0 +1,619 @@
+;
+; jfdctint.asm - accurate integer FDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, 2020, D. R. Commander.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_fdct_islow_sse2)
+
+EXTN(jconst_fdct_islow_sse2):
+
+PW_F130_F054   times 4 dw  (F_0_541 + F_0_765),  F_0_541
+PW_F054_MF130  times 4 dw  F_0_541, (F_0_541 - F_1_847)
+PW_MF078_F117  times 4 dw  (F_1_175 - F_1_961),  F_1_175
+PW_F117_F078   times 4 dw  F_1_175, (F_1_175 - F_0_390)
+PW_MF060_MF089 times 4 dw  (F_0_298 - F_0_899), -F_0_899
+PW_MF089_F060  times 4 dw -F_0_899, (F_1_501 - F_0_899)
+PW_MF050_MF256 times 4 dw  (F_2_053 - F_2_562), -F_2_562
+PW_MF256_F050  times 4 dw -F_2_562, (F_3_072 - F_2_562)
+PD_DESCALE_P1  times 4 dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2  times 4 dd  1 << (DESCALE_P2 - 1)
+PW_DESCALE_P2X times 8 dw  1 << (PASS1_BITS - 1)
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_sse2(DCTELEM *data)
+;
+
+; r10 = DCTELEM *data
+
+%define wk(i)   rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD  ; xmmword wkWK_NUM
+%define WK_NUM  6
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_fdct_islow_sse2)
+
+EXTN(jsimd_fdct_islow_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 1
+
+    ; ---- Pass 1: process rows.
+
+    mov         rdx, r10                ; (DCTELEM *)
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm1, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm2, XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm3, XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)
+
+    ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+    ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+    movdqa      xmm4, xmm0              ; transpose coefficients(phase 1)
+    punpcklwd   xmm0, xmm1              ; xmm0=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm4, xmm1              ; xmm4=(04 14 05 15 06 16 07 17)
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 1)
+    punpcklwd   xmm2, xmm3              ; xmm2=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm5, xmm3              ; xmm5=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm6, XMMWORD XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm7, XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm1, XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)
+
+    ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+    ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+    movdqa      XMMWORD wk(1), xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm2, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm7              ; xmm6=(40 50 41 51 42 52 43 53)
+    punpckhwd   xmm2, xmm7              ; xmm2=(44 54 45 55 46 56 47 57)
+    movdqa      xmm5, xmm1              ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm3              ; xmm1=(60 70 61 71 62 72 63 73)
+    punpckhwd   xmm5, xmm3              ; xmm5=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm7, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm1              ; xmm6=(40 50 60 70 41 51 61 71)
+    punpckhdq   xmm7, xmm1              ; xmm7=(42 52 62 72 43 53 63 73)
+    movdqa      xmm3, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm5              ; xmm2=(44 54 64 74 45 55 65 75)
+    punpckhdq   xmm3, xmm5              ; xmm3=(46 56 66 76 47 57 67 77)
+
+    movdqa      xmm1, XMMWORD wk(0)   ; xmm1=(20 30 21 31 22 32 23 33)
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=(24 34 25 35 26 36 27 37)
+    movdqa      XMMWORD wk(2), xmm7   ; wk(2)=(42 52 62 72 43 53 63 73)
+    movdqa      XMMWORD wk(3), xmm2   ; wk(3)=(44 54 64 74 45 55 65 75)
+
+    movdqa      xmm7, xmm0              ; transpose coefficients(phase 2)
+    punpckldq   xmm0, xmm1              ; xmm0=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm7, xmm1              ; xmm7=(02 12 22 32 03 13 23 33)
+    movdqa      xmm2, xmm4              ; transpose coefficients(phase 2)
+    punpckldq   xmm4, xmm5              ; xmm4=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm2, xmm5              ; xmm2=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm1, xmm0              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm0, xmm6              ; xmm0=(00 10 20 30 40 50 60 70)=data0
+    punpckhqdq  xmm1, xmm6              ; xmm1=(01 11 21 31 41 51 61 71)=data1
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm2, xmm3              ; xmm2=(06 16 26 36 46 56 66 76)=data6
+    punpckhqdq  xmm5, xmm3              ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+    movdqa      xmm6, xmm1
+    movdqa      xmm3, xmm0
+    psubw       xmm1, xmm2              ; xmm1=data1-data6=tmp6
+    psubw       xmm0, xmm5              ; xmm0=data0-data7=tmp7
+    paddw       xmm6, xmm2              ; xmm6=data1+data6=tmp1
+    paddw       xmm3, xmm5              ; xmm3=data0+data7=tmp0
+
+    movdqa      xmm2, XMMWORD wk(2)   ; xmm2=(42 52 62 72 43 53 63 73)
+    movdqa      xmm5, XMMWORD wk(3)   ; xmm5=(44 54 64 74 45 55 65 75)
+    movdqa      XMMWORD wk(0), xmm1   ; wk(0)=tmp6
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=tmp7
+
+    movdqa      xmm1, xmm7              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm7, xmm2              ; xmm7=(02 12 22 32 42 52 62 72)=data2
+    punpckhqdq  xmm1, xmm2              ; xmm1=(03 13 23 33 43 53 63 73)=data3
+    movdqa      xmm0, xmm4              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm4, xmm5              ; xmm4=(04 14 24 34 44 54 64 74)=data4
+    punpckhqdq  xmm0, xmm5              ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+    movdqa      xmm2, xmm1
+    movdqa      xmm5, xmm7
+    paddw       xmm1, xmm4              ; xmm1=data3+data4=tmp3
+    paddw       xmm7, xmm0              ; xmm7=data2+data5=tmp2
+    psubw       xmm2, xmm4              ; xmm2=data3-data4=tmp4
+    psubw       xmm5, xmm0              ; xmm5=data2-data5=tmp5
+
+    ; -- Even part
+
+    movdqa      xmm4, xmm3
+    movdqa      xmm0, xmm6
+    paddw       xmm3, xmm1              ; xmm3=tmp10
+    paddw       xmm6, xmm7              ; xmm6=tmp11
+    psubw       xmm4, xmm1              ; xmm4=tmp13
+    psubw       xmm0, xmm7              ; xmm0=tmp12
+
+    movdqa      xmm1, xmm3
+    paddw       xmm3, xmm6              ; xmm3=tmp10+tmp11
+    psubw       xmm1, xmm6              ; xmm1=tmp10-tmp11
+
+    psllw       xmm3, PASS1_BITS        ; xmm3=data0
+    psllw       xmm1, PASS1_BITS        ; xmm1=data4
+
+    movdqa      XMMWORD wk(2), xmm3   ; wk(2)=data0
+    movdqa      XMMWORD wk(3), xmm1   ; wk(3)=data4
+
+    ; (Original)
+    ; z1 = (tmp12 + tmp13) * 0.541196100;
+    ; data2 = z1 + tmp13 * 0.765366865;
+    ; data6 = z1 + tmp12 * -1.847759065;
+    ;
+    ; (This implementation)
+    ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+    ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+    movdqa      xmm7, xmm4              ; xmm4=tmp13
+    movdqa      xmm6, xmm4
+    punpcklwd   xmm7, xmm0              ; xmm0=tmp12
+    punpckhwd   xmm6, xmm0
+    movdqa      xmm4, xmm7
+    movdqa      xmm0, xmm6
+    pmaddwd     xmm7, rel PW_F130_F054   ; xmm7=data2L
+    pmaddwd     xmm6, rel PW_F130_F054   ; xmm6=data2H
+    pmaddwd     xmm4, rel PW_F054_MF130  ; xmm4=data6L
+    pmaddwd     xmm0, rel PW_F054_MF130  ; xmm0=data6H
+
+    paddd       xmm7, rel PD_DESCALE_P1
+    paddd       xmm6, rel PD_DESCALE_P1
+    psrad       xmm7, DESCALE_P1
+    psrad       xmm6, DESCALE_P1
+    paddd       xmm4, rel PD_DESCALE_P1
+    paddd       xmm0, rel PD_DESCALE_P1
+    psrad       xmm4, DESCALE_P1
+    psrad       xmm0, DESCALE_P1
+
+    packssdw    xmm7, xmm6              ; xmm7=data2
+    packssdw    xmm4, xmm0              ; xmm4=data6
+
+    movdqa      XMMWORD wk(4), xmm7   ; wk(4)=data2
+    movdqa      XMMWORD wk(5), xmm4   ; wk(5)=data6
+
+    ; -- Odd part
+
+    movdqa      xmm3, XMMWORD wk(0)   ; xmm3=tmp6
+    movdqa      xmm1, XMMWORD wk(1)   ; xmm1=tmp7
+
+    movdqa      xmm6, xmm2              ; xmm2=tmp4
+    movdqa      xmm0, xmm5              ; xmm5=tmp5
+    paddw       xmm6, xmm3              ; xmm6=z3
+    paddw       xmm0, xmm1              ; xmm0=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movdqa      xmm7, xmm6
+    movdqa      xmm4, xmm6
+    punpcklwd   xmm7, xmm0
+    punpckhwd   xmm4, xmm0
+    movdqa      xmm6, xmm7
+    movdqa      xmm0, xmm4
+    pmaddwd     xmm7, rel PW_MF078_F117  ; xmm7=z3L
+    pmaddwd     xmm4, rel PW_MF078_F117  ; xmm4=z3H
+    pmaddwd     xmm6, rel PW_F117_F078   ; xmm6=z4L
+    pmaddwd     xmm0, rel PW_F117_F078   ; xmm0=z4H
+
+    movdqa      XMMWORD wk(0), xmm7   ; wk(0)=z3L
+    movdqa      XMMWORD wk(1), xmm4   ; wk(1)=z3H
+
+    ; (Original)
+    ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+    ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+    ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+    ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+    ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+    ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+    ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+    ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+    ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+    movdqa      xmm7, xmm2
+    movdqa      xmm4, xmm2
+    punpcklwd   xmm7, xmm1
+    punpckhwd   xmm4, xmm1
+    movdqa      xmm2, xmm7
+    movdqa      xmm1, xmm4
+    pmaddwd     xmm7, rel PW_MF060_MF089  ; xmm7=tmp4L
+    pmaddwd     xmm4, rel PW_MF060_MF089  ; xmm4=tmp4H
+    pmaddwd     xmm2, rel PW_MF089_F060   ; xmm2=tmp7L
+    pmaddwd     xmm1, rel PW_MF089_F060   ; xmm1=tmp7H
+
+    paddd       xmm7, XMMWORD wk(0)   ; xmm7=data7L
+    paddd       xmm4, XMMWORD wk(1)   ; xmm4=data7H
+    paddd       xmm2, xmm6              ; xmm2=data1L
+    paddd       xmm1, xmm0              ; xmm1=data1H
+
+    paddd       xmm7, rel PD_DESCALE_P1
+    paddd       xmm4, rel PD_DESCALE_P1
+    psrad       xmm7, DESCALE_P1
+    psrad       xmm4, DESCALE_P1
+    paddd       xmm2, rel PD_DESCALE_P1
+    paddd       xmm1, rel PD_DESCALE_P1
+    psrad       xmm2, DESCALE_P1
+    psrad       xmm1, DESCALE_P1
+
+    packssdw    xmm7, xmm4              ; xmm7=data7
+    packssdw    xmm2, xmm1              ; xmm2=data1
+
+    movdqa      xmm4, xmm5
+    movdqa      xmm1, xmm5
+    punpcklwd   xmm4, xmm3
+    punpckhwd   xmm1, xmm3
+    movdqa      xmm5, xmm4
+    movdqa      xmm3, xmm1
+    pmaddwd     xmm4, rel PW_MF050_MF256  ; xmm4=tmp5L
+    pmaddwd     xmm1, rel PW_MF050_MF256  ; xmm1=tmp5H
+    pmaddwd     xmm5, rel PW_MF256_F050   ; xmm5=tmp6L
+    pmaddwd     xmm3, rel PW_MF256_F050   ; xmm3=tmp6H
+
+    paddd       xmm4, xmm6              ; xmm4=data5L
+    paddd       xmm1, xmm0              ; xmm1=data5H
+    paddd       xmm5, XMMWORD wk(0)   ; xmm5=data3L
+    paddd       xmm3, XMMWORD wk(1)   ; xmm3=data3H
+
+    paddd       xmm4, rel PD_DESCALE_P1
+    paddd       xmm1, rel PD_DESCALE_P1
+    psrad       xmm4, DESCALE_P1
+    psrad       xmm1, DESCALE_P1
+    paddd       xmm5, rel PD_DESCALE_P1
+    paddd       xmm3, rel PD_DESCALE_P1
+    psrad       xmm5, DESCALE_P1
+    psrad       xmm3, DESCALE_P1
+
+    packssdw    xmm4, xmm1              ; xmm4=data5
+    packssdw    xmm5, xmm3              ; xmm5=data3
+
+    ; ---- Pass 2: process columns.
+
+    movdqa      xmm6, XMMWORD wk(2)   ; xmm6=col0
+    movdqa      xmm0, XMMWORD wk(4)   ; xmm0=col2
+
+    ; xmm6=(00 10 20 30 40 50 60 70), xmm0=(02 12 22 32 42 52 62 72)
+    ; xmm2=(01 11 21 31 41 51 61 71), xmm5=(03 13 23 33 43 53 63 73)
+
+    movdqa      xmm1, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm2              ; xmm6=(00 01 10 11 20 21 30 31)
+    punpckhwd   xmm1, xmm2              ; xmm1=(40 41 50 51 60 61 70 71)
+    movdqa      xmm3, xmm0              ; transpose coefficients(phase 1)
+    punpcklwd   xmm0, xmm5              ; xmm0=(02 03 12 13 22 23 32 33)
+    punpckhwd   xmm3, xmm5              ; xmm3=(42 43 52 53 62 63 72 73)
+
+    movdqa      xmm2, XMMWORD wk(3)   ; xmm2=col4
+    movdqa      xmm5, XMMWORD wk(5)   ; xmm5=col6
+
+    ; xmm2=(04 14 24 34 44 54 64 74), xmm5=(06 16 26 36 46 56 66 76)
+    ; xmm4=(05 15 25 35 45 55 65 75), xmm7=(07 17 27 37 47 57 67 77)
+
+    movdqa      XMMWORD wk(0), xmm0   ; wk(0)=(02 03 12 13 22 23 32 33)
+    movdqa      XMMWORD wk(1), xmm3   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+    movdqa      xmm0, xmm2              ; transpose coefficients(phase 1)
+    punpcklwd   xmm2, xmm4              ; xmm2=(04 05 14 15 24 25 34 35)
+    punpckhwd   xmm0, xmm4              ; xmm0=(44 45 54 55 64 65 74 75)
+    movdqa      xmm3, xmm5              ; transpose coefficients(phase 1)
+    punpcklwd   xmm5, xmm7              ; xmm5=(06 07 16 17 26 27 36 37)
+    punpckhwd   xmm3, xmm7              ; xmm3=(46 47 56 57 66 67 76 77)
+
+    movdqa      xmm4, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm5              ; xmm2=(04 05 06 07 14 15 16 17)
+    punpckhdq   xmm4, xmm5              ; xmm4=(24 25 26 27 34 35 36 37)
+    movdqa      xmm7, xmm0              ; transpose coefficients(phase 2)
+    punpckldq   xmm0, xmm3              ; xmm0=(44 45 46 47 54 55 56 57)
+    punpckhdq   xmm7, xmm3              ; xmm7=(64 65 66 67 74 75 76 77)
+
+    movdqa      xmm5, XMMWORD wk(0)   ; xmm5=(02 03 12 13 22 23 32 33)
+    movdqa      xmm3, XMMWORD wk(1)   ; xmm3=(42 43 52 53 62 63 72 73)
+    movdqa      XMMWORD wk(2), xmm4   ; wk(2)=(24 25 26 27 34 35 36 37)
+    movdqa      XMMWORD wk(3), xmm0   ; wk(3)=(44 45 46 47 54 55 56 57)
+
+    movdqa      xmm4, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm5              ; xmm6=(00 01 02 03 10 11 12 13)
+    punpckhdq   xmm4, xmm5              ; xmm4=(20 21 22 23 30 31 32 33)
+    movdqa      xmm0, xmm1              ; transpose coefficients(phase 2)
+    punpckldq   xmm1, xmm3              ; xmm1=(40 41 42 43 50 51 52 53)
+    punpckhdq   xmm0, xmm3              ; xmm0=(60 61 62 63 70 71 72 73)
+
+    movdqa      xmm5, xmm6              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm6, xmm2              ; xmm6=(00 01 02 03 04 05 06 07)=data0
+    punpckhqdq  xmm5, xmm2              ; xmm5=(10 11 12 13 14 15 16 17)=data1
+    movdqa      xmm3, xmm0              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm0, xmm7              ; xmm0=(60 61 62 63 64 65 66 67)=data6
+    punpckhqdq  xmm3, xmm7              ; xmm3=(70 71 72 73 74 75 76 77)=data7
+
+    movdqa      xmm2, xmm5
+    movdqa      xmm7, xmm6
+    psubw       xmm5, xmm0              ; xmm5=data1-data6=tmp6
+    psubw       xmm6, xmm3              ; xmm6=data0-data7=tmp7
+    paddw       xmm2, xmm0              ; xmm2=data1+data6=tmp1
+    paddw       xmm7, xmm3              ; xmm7=data0+data7=tmp0
+
+    movdqa      xmm0, XMMWORD wk(2)   ; xmm0=(24 25 26 27 34 35 36 37)
+    movdqa      xmm3, XMMWORD wk(3)   ; xmm3=(44 45 46 47 54 55 56 57)
+    movdqa      XMMWORD wk(0), xmm5   ; wk(0)=tmp6
+    movdqa      XMMWORD wk(1), xmm6   ; wk(1)=tmp7
+
+    movdqa      xmm5, xmm4              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm4, xmm0              ; xmm4=(20 21 22 23 24 25 26 27)=data2
+    punpckhqdq  xmm5, xmm0              ; xmm5=(30 31 32 33 34 35 36 37)=data3
+    movdqa      xmm6, xmm1              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm1, xmm3              ; xmm1=(40 41 42 43 44 45 46 47)=data4
+    punpckhqdq  xmm6, xmm3              ; xmm6=(50 51 52 53 54 55 56 57)=data5
+
+    movdqa      xmm0, xmm5
+    movdqa      xmm3, xmm4
+    paddw       xmm5, xmm1              ; xmm5=data3+data4=tmp3
+    paddw       xmm4, xmm6              ; xmm4=data2+data5=tmp2
+    psubw       xmm0, xmm1              ; xmm0=data3-data4=tmp4
+    psubw       xmm3, xmm6              ; xmm3=data2-data5=tmp5
+
+    ; -- Even part
+
+    movdqa      xmm1, xmm7
+    movdqa      xmm6, xmm2
+    paddw       xmm7, xmm5              ; xmm7=tmp10
+    paddw       xmm2, xmm4              ; xmm2=tmp11
+    psubw       xmm1, xmm5              ; xmm1=tmp13
+    psubw       xmm6, xmm4              ; xmm6=tmp12
+
+    movdqa      xmm5, xmm7
+    paddw       xmm7, xmm2              ; xmm7=tmp10+tmp11
+    psubw       xmm5, xmm2              ; xmm5=tmp10-tmp11
+
+    paddw       xmm7, rel PW_DESCALE_P2X
+    paddw       xmm5, rel PW_DESCALE_P2X
+    psraw       xmm7, PASS1_BITS        ; xmm7=data0
+    psraw       xmm5, PASS1_BITS        ; xmm5=data4
+
+    movdqa      XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM), xmm7
+    movdqa      XMMWORD XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM), xmm5
+
+    ; (Original)
+    ; z1 = (tmp12 + tmp13) * 0.541196100;
+    ; data2 = z1 + tmp13 * 0.765366865;
+    ; data6 = z1 + tmp12 * -1.847759065;
+    ;
+    ; (This implementation)
+    ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+    ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+    movdqa      xmm4, xmm1              ; xmm1=tmp13
+    movdqa      xmm2, xmm1
+    punpcklwd   xmm4, xmm6              ; xmm6=tmp12
+    punpckhwd   xmm2, xmm6
+    movdqa      xmm1, xmm4
+    movdqa      xmm6, xmm2
+    pmaddwd     xmm4, rel PW_F130_F054   ; xmm4=data2L
+    pmaddwd     xmm2, rel PW_F130_F054   ; xmm2=data2H
+    pmaddwd     xmm1, rel PW_F054_MF130  ; xmm1=data6L
+    pmaddwd     xmm6, rel PW_F054_MF130  ; xmm6=data6H
+
+    paddd       xmm4, rel PD_DESCALE_P2
+    paddd       xmm2, rel PD_DESCALE_P2
+    psrad       xmm4, DESCALE_P2
+    psrad       xmm2, DESCALE_P2
+    paddd       xmm1, rel PD_DESCALE_P2
+    paddd       xmm6, rel PD_DESCALE_P2
+    psrad       xmm1, DESCALE_P2
+    psrad       xmm6, DESCALE_P2
+
+    packssdw    xmm4, xmm2              ; xmm4=data2
+    packssdw    xmm1, xmm6              ; xmm1=data6
+
+    movdqa      XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM), xmm4
+    movdqa      XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM), xmm1
+
+    ; -- Odd part
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=tmp6
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=tmp7
+
+    movdqa      xmm2, xmm0              ; xmm0=tmp4
+    movdqa      xmm6, xmm3              ; xmm3=tmp5
+    paddw       xmm2, xmm7              ; xmm2=z3
+    paddw       xmm6, xmm5              ; xmm6=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movdqa      xmm4, xmm2
+    movdqa      xmm1, xmm2
+    punpcklwd   xmm4, xmm6
+    punpckhwd   xmm1, xmm6
+    movdqa      xmm2, xmm4
+    movdqa      xmm6, xmm1
+    pmaddwd     xmm4, rel PW_MF078_F117  ; xmm4=z3L
+    pmaddwd     xmm1, rel PW_MF078_F117  ; xmm1=z3H
+    pmaddwd     xmm2, rel PW_F117_F078   ; xmm2=z4L
+    pmaddwd     xmm6, rel PW_F117_F078   ; xmm6=z4H
+
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=z3L
+    movdqa      XMMWORD wk(1), xmm1   ; wk(1)=z3H
+
+    ; (Original)
+    ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+    ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+    ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+    ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+    ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+    ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+    ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+    ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+    ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+    movdqa      xmm4, xmm0
+    movdqa      xmm1, xmm0
+    punpcklwd   xmm4, xmm5
+    punpckhwd   xmm1, xmm5
+    movdqa      xmm0, xmm4
+    movdqa      xmm5, xmm1
+    pmaddwd     xmm4, rel PW_MF060_MF089  ; xmm4=tmp4L
+    pmaddwd     xmm1, rel PW_MF060_MF089  ; xmm1=tmp4H
+    pmaddwd     xmm0, rel PW_MF089_F060   ; xmm0=tmp7L
+    pmaddwd     xmm5, rel PW_MF089_F060   ; xmm5=tmp7H
+
+    paddd       xmm4,  XMMWORD wk(0)  ; xmm4=data7L
+    paddd       xmm1,  XMMWORD wk(1)  ; xmm1=data7H
+    paddd       xmm0, xmm2              ; xmm0=data1L
+    paddd       xmm5, xmm6              ; xmm5=data1H
+
+    paddd       xmm4, rel PD_DESCALE_P2
+    paddd       xmm1, rel PD_DESCALE_P2
+    psrad       xmm4, DESCALE_P2
+    psrad       xmm1, DESCALE_P2
+    paddd       xmm0, rel PD_DESCALE_P2
+    paddd       xmm5, rel PD_DESCALE_P2
+    psrad       xmm0, DESCALE_P2
+    psrad       xmm5, DESCALE_P2
+
+    packssdw    xmm4, xmm1              ; xmm4=data7
+    packssdw    xmm0, xmm5              ; xmm0=data1
+
+    movdqa      XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM), xmm4
+    movdqa      XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM), xmm0
+
+    movdqa      xmm1, xmm3
+    movdqa      xmm5, xmm3
+    punpcklwd   xmm1, xmm7
+    punpckhwd   xmm5, xmm7
+    movdqa      xmm3, xmm1
+    movdqa      xmm7, xmm5
+    pmaddwd     xmm1, rel PW_MF050_MF256  ; xmm1=tmp5L
+    pmaddwd     xmm5, rel PW_MF050_MF256  ; xmm5=tmp5H
+    pmaddwd     xmm3, rel PW_MF256_F050   ; xmm3=tmp6L
+    pmaddwd     xmm7, rel PW_MF256_F050   ; xmm7=tmp6H
+
+    paddd       xmm1, xmm2              ; xmm1=data5L
+    paddd       xmm5, xmm6              ; xmm5=data5H
+    paddd       xmm3, XMMWORD wk(0)   ; xmm3=data3L
+    paddd       xmm7, XMMWORD wk(1)   ; xmm7=data3H
+
+    paddd       xmm1, rel PD_DESCALE_P2
+    paddd       xmm5, rel PD_DESCALE_P2
+    psrad       xmm1, DESCALE_P2
+    psrad       xmm5, DESCALE_P2
+    paddd       xmm3, rel PD_DESCALE_P2
+    paddd       xmm7, rel PD_DESCALE_P2
+    psrad       xmm3, DESCALE_P2
+    psrad       xmm7, DESCALE_P2
+
+    packssdw    xmm1, xmm5              ; xmm1=data5
+    packssdw    xmm3, xmm7              ; xmm3=data3
+
+    movdqa      XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM), xmm1
+    movdqa      XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM), xmm3
+
+    uncollect_args 1
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jidctflt-sse2.asm Added

@@ -0,0 +1,482 @@
+;
+; jidctflt.asm - floating-point IDCT (64-bit SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro unpcklps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+    shufps      %1, %2, 0x44
+%endmacro
+
+%macro unpckhps2 2  ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+    shufps      %1, %2, 0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_float_sse2)
+
+EXTN(jconst_idct_float_sse2):
+
+PD_1_414        times 4  dd  1.414213562373095048801689
+PD_1_847        times 4  dd  1.847759065022573512256366
+PD_1_082        times 4  dd  1.082392200292393968799446
+PD_M2_613       times 4  dd -2.613125929752753055713286
+PD_RNDINT_MAGIC times 4  dd  100663296.0  ; (float)(0x00C00000 << 3)
+PB_CENTERJSAMP  times 16 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_sse2(void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = void *dct_table
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13d = JDIMENSION output_col
+
+%define original_rbp  rbp + 0
+%define wk(i)         rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM        2
+%define workspace     wk(0) - DCTSIZE2 * SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspaceDCTSIZE2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_float_sse2)
+
+EXTN(jsimd_idct_float_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, workspace
+    collect_args 4
+    push        rbx
+
+    ; ---- Pass 1: process columns from input, store into work array.
+
+    mov         rdx, r10                ; quantptr
+    mov         rsi, r11                ; inptr
+    lea         rdi, workspace        ; FAST_FLOAT *wsptr
+    mov         rcx, DCTSIZE/4          ; ctr
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE
+    mov         eax, dword DWBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    jnz         near .columnDCT
+
+    movq        xmm1, XMM_MMWORD MMBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    movq        xmm2, XMM_MMWORD MMBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    movq        xmm3, XMM_MMWORD MMBLOCK(3,0,rsi,SIZEOF_JCOEF)
+    movq        xmm4, XMM_MMWORD MMBLOCK(4,0,rsi,SIZEOF_JCOEF)
+    movq        xmm5, XMM_MMWORD MMBLOCK(5,0,rsi,SIZEOF_JCOEF)
+    movq        xmm6, XMM_MMWORD MMBLOCK(6,0,rsi,SIZEOF_JCOEF)
+    movq        xmm7, XMM_MMWORD MMBLOCK(7,0,rsi,SIZEOF_JCOEF)
+    por         xmm1, xmm2
+    por         xmm3, xmm4
+    por         xmm5, xmm6
+    por         xmm1, xmm3
+    por         xmm5, xmm7
+    por         xmm1, xmm5
+    packsswb    xmm1, xmm1
+    movd        eax, xmm1
+    test        rax, rax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movq        xmm0, XMM_MMWORD MMBLOCK(0,0,rsi,SIZEOF_JCOEF)
+
+    punpcklwd   xmm0, xmm0                  ; xmm0=(00 00 01 01 02 02 03 03)
+    psrad       xmm0, (DWORD_BIT-WORD_BIT)  ; xmm0=in0=(00 01 02 03)
+    cvtdq2ps    xmm0, xmm0                  ; xmm0=in0=(00 01 02 03)
+
+    mulps       xmm0, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movaps      xmm1, xmm0
+    movaps      xmm2, xmm0
+    movaps      xmm3, xmm0
+
+    shufps      xmm0, xmm0, 0x00        ; xmm0=(00 00 00 00)
+    shufps      xmm1, xmm1, 0x55        ; xmm1=(01 01 01 01)
+    shufps      xmm2, xmm2, 0xAA        ; xmm2=(02 02 02 02)
+    shufps      xmm3, xmm3, 0xFF        ; xmm3=(03 03 03 03)
+
+    movaps      XMMWORD XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(2,0,rdi,SIZEOF_FAST_FLOAT), xmm2
+    movaps      XMMWORD XMMBLOCK(2,1,rdi,SIZEOF_FAST_FLOAT), xmm2
+    movaps      XMMWORD XMMBLOCK(3,0,rdi,SIZEOF_FAST_FLOAT), xmm3
+    movaps      XMMWORD XMMBLOCK(3,1,rdi,SIZEOF_FAST_FLOAT), xmm3
+    jmp         near .nextcolumn
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movq        xmm0, XMM_MMWORD MMBLOCK(0,0,rsi,SIZEOF_JCOEF)
+    movq        xmm1, XMM_MMWORD MMBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    movq        xmm2, XMM_MMWORD MMBLOCK(4,0,rsi,SIZEOF_JCOEF)
+    movq        xmm3, XMM_MMWORD MMBLOCK(6,0,rsi,SIZEOF_JCOEF)
+
+    punpcklwd   xmm0, xmm0                  ; xmm0=(00 00 01 01 02 02 03 03)
+    punpcklwd   xmm1, xmm1                  ; xmm1=(20 20 21 21 22 22 23 23)
+    psrad       xmm0, (DWORD_BIT-WORD_BIT)  ; xmm0=in0=(00 01 02 03)
+    psrad       xmm1, (DWORD_BIT-WORD_BIT)  ; xmm1=in2=(20 21 22 23)
+    cvtdq2ps    xmm0, xmm0                  ; xmm0=in0=(00 01 02 03)
+    cvtdq2ps    xmm1, xmm1                  ; xmm1=in2=(20 21 22 23)
+
+    punpcklwd   xmm2, xmm2                  ; xmm2=(40 40 41 41 42 42 43 43)
+    punpcklwd   xmm3, xmm3                  ; xmm3=(60 60 61 61 62 62 63 63)
+    psrad       xmm2, (DWORD_BIT-WORD_BIT)  ; xmm2=in4=(40 41 42 43)
+    psrad       xmm3, (DWORD_BIT-WORD_BIT)  ; xmm3=in6=(60 61 62 63)
+    cvtdq2ps    xmm2, xmm2                  ; xmm2=in4=(40 41 42 43)
+    cvtdq2ps    xmm3, xmm3                  ; xmm3=in6=(60 61 62 63)
+
+    mulps       xmm0, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm1, XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm2, XMMWORD XMMBLOCK(4,0,rdx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm3, XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movaps      xmm4, xmm0
+    movaps      xmm5, xmm1
+    subps       xmm0, xmm2              ; xmm0=tmp11
+    subps       xmm1, xmm3
+    addps       xmm4, xmm2              ; xmm4=tmp10
+    addps       xmm5, xmm3              ; xmm5=tmp13
+
+    mulps       xmm1, rel PD_1_414
+    subps       xmm1, xmm5              ; xmm1=tmp12
+
+    movaps      xmm6, xmm4
+    movaps      xmm7, xmm0
+    subps       xmm4, xmm5              ; xmm4=tmp3
+    subps       xmm0, xmm1              ; xmm0=tmp2
+    addps       xmm6, xmm5              ; xmm6=tmp0
+    addps       xmm7, xmm1              ; xmm7=tmp1
+
+    movaps      XMMWORD wk(1), xmm4   ; tmp3
+    movaps      XMMWORD wk(0), xmm0   ; tmp2
+
+    ; -- Odd part
+
+    movq        xmm2, XMM_MMWORD MMBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    movq        xmm3, XMM_MMWORD MMBLOCK(3,0,rsi,SIZEOF_JCOEF)
+    movq        xmm5, XMM_MMWORD MMBLOCK(5,0,rsi,SIZEOF_JCOEF)
+    movq        xmm1, XMM_MMWORD MMBLOCK(7,0,rsi,SIZEOF_JCOEF)
+
+    punpcklwd   xmm2, xmm2                  ; xmm2=(10 10 11 11 12 12 13 13)
+    punpcklwd   xmm3, xmm3                  ; xmm3=(30 30 31 31 32 32 33 33)
+    psrad       xmm2, (DWORD_BIT-WORD_BIT)  ; xmm2=in1=(10 11 12 13)
+    psrad       xmm3, (DWORD_BIT-WORD_BIT)  ; xmm3=in3=(30 31 32 33)
+    cvtdq2ps    xmm2, xmm2                  ; xmm2=in1=(10 11 12 13)
+    cvtdq2ps    xmm3, xmm3                  ; xmm3=in3=(30 31 32 33)
+
+    punpcklwd   xmm5, xmm5                  ; xmm5=(50 50 51 51 52 52 53 53)
+    punpcklwd   xmm1, xmm1                  ; xmm1=(70 70 71 71 72 72 73 73)
+    psrad       xmm5, (DWORD_BIT-WORD_BIT)  ; xmm5=in5=(50 51 52 53)
+    psrad       xmm1, (DWORD_BIT-WORD_BIT)  ; xmm1=in7=(70 71 72 73)
+    cvtdq2ps    xmm5, xmm5                  ; xmm5=in5=(50 51 52 53)
+    cvtdq2ps    xmm1, xmm1                  ; xmm1=in7=(70 71 72 73)
+
+    mulps       xmm2, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm3, XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm5, XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_FLOAT_MULT_TYPE)
+    mulps       xmm1, XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_FLOAT_MULT_TYPE)
+
+    movaps      xmm4, xmm2
+    movaps      xmm0, xmm5
+    addps       xmm2, xmm1              ; xmm2=z11
+    addps       xmm5, xmm3              ; xmm5=z13
+    subps       xmm4, xmm1              ; xmm4=z12
+    subps       xmm0, xmm3              ; xmm0=z10
+
+    movaps      xmm1, xmm2
+    subps       xmm2, xmm5
+    addps       xmm1, xmm5              ; xmm1=tmp7
+
+    mulps       xmm2, rel PD_1_414    ; xmm2=tmp11
+
+    movaps      xmm3, xmm0
+    addps       xmm0, xmm4
+    mulps       xmm0, rel PD_1_847    ; xmm0=z5
+    mulps       xmm3, rel PD_M2_613   ; xmm3=(z10 * -2.613125930)
+    mulps       xmm4, rel PD_1_082    ; xmm4=(z12 * 1.082392200)
+    addps       xmm3, xmm0              ; xmm3=tmp12
+    subps       xmm4, xmm0              ; xmm4=tmp10
+
+    ; -- Final output stage
+
+    subps       xmm3, xmm1              ; xmm3=tmp6
+    movaps      xmm5, xmm6
+    movaps      xmm0, xmm7
+    addps       xmm6, xmm1              ; xmm6=data0=(00 01 02 03)
+    addps       xmm7, xmm3              ; xmm7=data1=(10 11 12 13)
+    subps       xmm5, xmm1              ; xmm5=data7=(70 71 72 73)
+    subps       xmm0, xmm3              ; xmm0=data6=(60 61 62 63)
+    subps       xmm2, xmm3              ; xmm2=tmp5
+
+    movaps      xmm1, xmm6              ; transpose coefficients(phase 1)
+    unpcklps    xmm6, xmm7              ; xmm6=(00 10 01 11)
+    unpckhps    xmm1, xmm7              ; xmm1=(02 12 03 13)
+    movaps      xmm3, xmm0              ; transpose coefficients(phase 1)
+    unpcklps    xmm0, xmm5              ; xmm0=(60 70 61 71)
+    unpckhps    xmm3, xmm5              ; xmm3=(62 72 63 73)
+
+    movaps      xmm7, XMMWORD wk(0)   ; xmm7=tmp2
+    movaps      xmm5, XMMWORD wk(1)   ; xmm5=tmp3
+
+    movaps      XMMWORD wk(0), xmm0   ; wk(0)=(60 70 61 71)
+    movaps      XMMWORD wk(1), xmm3   ; wk(1)=(62 72 63 73)
+
+    addps       xmm4, xmm2              ; xmm4=tmp4
+    movaps      xmm0, xmm7
+    movaps      xmm3, xmm5
+    addps       xmm7, xmm2              ; xmm7=data2=(20 21 22 23)
+    addps       xmm5, xmm4              ; xmm5=data4=(40 41 42 43)
+    subps       xmm0, xmm2              ; xmm0=data5=(50 51 52 53)
+    subps       xmm3, xmm4              ; xmm3=data3=(30 31 32 33)
+
+    movaps      xmm2, xmm7              ; transpose coefficients(phase 1)
+    unpcklps    xmm7, xmm3              ; xmm7=(20 30 21 31)
+    unpckhps    xmm2, xmm3              ; xmm2=(22 32 23 33)
+    movaps      xmm4, xmm5              ; transpose coefficients(phase 1)
+    unpcklps    xmm5, xmm0              ; xmm5=(40 50 41 51)
+    unpckhps    xmm4, xmm0              ; xmm4=(42 52 43 53)
+
+    movaps      xmm3, xmm6              ; transpose coefficients(phase 2)
+    unpcklps2   xmm6, xmm7              ; xmm6=(00 10 20 30)
+    unpckhps2   xmm3, xmm7              ; xmm3=(01 11 21 31)
+    movaps      xmm0, xmm1              ; transpose coefficients(phase 2)
+    unpcklps2   xmm1, xmm2              ; xmm1=(02 12 22 32)
+    unpckhps2   xmm0, xmm2              ; xmm0=(03 13 23 33)
+
+    movaps      xmm7, XMMWORD wk(0)   ; xmm7=(60 70 61 71)
+    movaps      xmm2, XMMWORD wk(1)   ; xmm2=(62 72 63 73)
+
+    movaps      XMMWORD XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT), xmm6
+    movaps      XMMWORD XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT), xmm3
+    movaps      XMMWORD XMMBLOCK(2,0,rdi,SIZEOF_FAST_FLOAT), xmm1
+    movaps      XMMWORD XMMBLOCK(3,0,rdi,SIZEOF_FAST_FLOAT), xmm0
+
+    movaps      xmm6, xmm5              ; transpose coefficients(phase 2)
+    unpcklps2   xmm5, xmm7              ; xmm5=(40 50 60 70)
+    unpckhps2   xmm6, xmm7              ; xmm6=(41 51 61 71)
+    movaps      xmm3, xmm4              ; transpose coefficients(phase 2)
+    unpcklps2   xmm4, xmm2              ; xmm4=(42 52 62 72)
+    unpckhps2   xmm3, xmm2              ; xmm3=(43 53 63 73)
+
+    movaps      XMMWORD XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT), xmm5
+    movaps      XMMWORD XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT), xmm6
+    movaps      XMMWORD XMMBLOCK(2,1,rdi,SIZEOF_FAST_FLOAT), xmm4
+    movaps      XMMWORD XMMBLOCK(3,1,rdi,SIZEOF_FAST_FLOAT), xmm3
+
+.nextcolumn:
+    add         rsi, byte 4*SIZEOF_JCOEF               ; coef_block
+    add         rdx, byte 4*SIZEOF_FLOAT_MULT_TYPE     ; quantptr
+    add         rdi,      4*DCTSIZE*SIZEOF_FAST_FLOAT  ; wsptr
+    dec         rcx                                    ; ctr
+    jnz         near .columnloop
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32
+    prefetchnta rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32
+    prefetchnta rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32
+    prefetchnta rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         rax, original_rbp
+    lea         rsi, workspace        ; FAST_FLOAT *wsptr
+    mov         rdi, r12                ; (JSAMPROW *)
+    mov         eax, r13d
+    mov         rcx, DCTSIZE/4          ; ctr
+.rowloop:
+
+    ; -- Even part
+
+    movaps      xmm0, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(2,0,rsi,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(4,0,rsi,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(6,0,rsi,SIZEOF_FAST_FLOAT)
+
+    movaps      xmm4, xmm0
+    movaps      xmm5, xmm1
+    subps       xmm0, xmm2              ; xmm0=tmp11
+    subps       xmm1, xmm3
+    addps       xmm4, xmm2              ; xmm4=tmp10
+    addps       xmm5, xmm3              ; xmm5=tmp13
+
+    mulps       xmm1, rel PD_1_414
+    subps       xmm1, xmm5              ; xmm1=tmp12
+
+    movaps      xmm6, xmm4
+    movaps      xmm7, xmm0
+    subps       xmm4, xmm5              ; xmm4=tmp3
+    subps       xmm0, xmm1              ; xmm0=tmp2
+    addps       xmm6, xmm5              ; xmm6=tmp0
+    addps       xmm7, xmm1              ; xmm7=tmp1
+
+    movaps      XMMWORD wk(1), xmm4   ; tmp3
+    movaps      XMMWORD wk(0), xmm0   ; tmp2
+
+    ; -- Odd part
+
+    movaps      xmm2, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(3,0,rsi,SIZEOF_FAST_FLOAT)
+    movaps      xmm5, XMMWORD XMMBLOCK(5,0,rsi,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(7,0,rsi,SIZEOF_FAST_FLOAT)
+
+    movaps      xmm4, xmm2
+    movaps      xmm0, xmm5
+    addps       xmm2, xmm1              ; xmm2=z11
+    addps       xmm5, xmm3              ; xmm5=z13
+    subps       xmm4, xmm1              ; xmm4=z12
+    subps       xmm0, xmm3              ; xmm0=z10
+
+    movaps      xmm1, xmm2
+    subps       xmm2, xmm5
+    addps       xmm1, xmm5              ; xmm1=tmp7
+
+    mulps       xmm2, rel PD_1_414    ; xmm2=tmp11
+
+    movaps      xmm3, xmm0
+    addps       xmm0, xmm4
+    mulps       xmm0, rel PD_1_847    ; xmm0=z5
+    mulps       xmm3, rel PD_M2_613   ; xmm3=(z10 * -2.613125930)
+    mulps       xmm4, rel PD_1_082    ; xmm4=(z12 * 1.082392200)
+    addps       xmm3, xmm0              ; xmm3=tmp12
+    subps       xmm4, xmm0              ; xmm4=tmp10
+
+    ; -- Final output stage
+
+    subps       xmm3, xmm1              ; xmm3=tmp6
+    movaps      xmm5, xmm6
+    movaps      xmm0, xmm7
+    addps       xmm6, xmm1              ; xmm6=data0=(00 10 20 30)
+    addps       xmm7, xmm3              ; xmm7=data1=(01 11 21 31)
+    subps       xmm5, xmm1              ; xmm5=data7=(07 17 27 37)
+    subps       xmm0, xmm3              ; xmm0=data6=(06 16 26 36)
+    subps       xmm2, xmm3              ; xmm2=tmp5
+
+    movaps      xmm1, rel PD_RNDINT_MAGIC  ; xmm1=rel PD_RNDINT_MAGIC
+    pcmpeqd     xmm3, xmm3
+    psrld       xmm3, WORD_BIT          ; xmm3={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+    addps       xmm6, xmm1              ; xmm6=roundint(data0/8)=(00 ** 10 ** 20 ** 30 **)
+    addps       xmm7, xmm1              ; xmm7=roundint(data1/8)=(01 ** 11 ** 21 ** 31 **)
+    addps       xmm0, xmm1              ; xmm0=roundint(data6/8)=(06 ** 16 ** 26 ** 36 **)
+    addps       xmm5, xmm1              ; xmm5=roundint(data7/8)=(07 ** 17 ** 27 ** 37 **)
+
+    pand        xmm6, xmm3              ; xmm6=(00 -- 10 -- 20 -- 30 --)
+    pslld       xmm7, WORD_BIT          ; xmm7=(-- 01 -- 11 -- 21 -- 31)
+    pand        xmm0, xmm3              ; xmm0=(06 -- 16 -- 26 -- 36 --)
+    pslld       xmm5, WORD_BIT          ; xmm5=(-- 07 -- 17 -- 27 -- 37)
+    por         xmm6, xmm7              ; xmm6=(00 01 10 11 20 21 30 31)
+    por         xmm0, xmm5              ; xmm0=(06 07 16 17 26 27 36 37)
+
+    movaps      xmm1,  XMMWORD wk(0)  ; xmm1=tmp2
+    movaps      xmm3,  XMMWORD wk(1)  ; xmm3=tmp3
+
+    addps       xmm4, xmm2              ; xmm4=tmp4
+    movaps      xmm7, xmm1
+    movaps      xmm5, xmm3
+    addps       xmm1, xmm2              ; xmm1=data2=(02 12 22 32)
+    addps       xmm3, xmm4              ; xmm3=data4=(04 14 24 34)
+    subps       xmm7, xmm2              ; xmm7=data5=(05 15 25 35)
+    subps       xmm5, xmm4              ; xmm5=data3=(03 13 23 33)
+
+    movaps      xmm2, rel PD_RNDINT_MAGIC  ; xmm2=rel PD_RNDINT_MAGIC
+    pcmpeqd     xmm4, xmm4
+    psrld       xmm4, WORD_BIT          ; xmm4={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+    addps       xmm3, xmm2              ; xmm3=roundint(data4/8)=(04 ** 14 ** 24 ** 34 **)
+    addps       xmm7, xmm2              ; xmm7=roundint(data5/8)=(05 ** 15 ** 25 ** 35 **)
+    addps       xmm1, xmm2              ; xmm1=roundint(data2/8)=(02 ** 12 ** 22 ** 32 **)
+    addps       xmm5, xmm2              ; xmm5=roundint(data3/8)=(03 ** 13 ** 23 ** 33 **)
+
+    pand        xmm3, xmm4              ; xmm3=(04 -- 14 -- 24 -- 34 --)
+    pslld       xmm7, WORD_BIT          ; xmm7=(-- 05 -- 15 -- 25 -- 35)
+    pand        xmm1, xmm4              ; xmm1=(02 -- 12 -- 22 -- 32 --)
+    pslld       xmm5, WORD_BIT          ; xmm5=(-- 03 -- 13 -- 23 -- 33)
+    por         xmm3, xmm7              ; xmm3=(04 05 14 15 24 25 34 35)
+    por         xmm1, xmm5              ; xmm1=(02 03 12 13 22 23 32 33)
+
+    movdqa      xmm2, rel PB_CENTERJSAMP  ; xmm2=rel PB_CENTERJSAMP
+
+    packsswb    xmm6, xmm3        ; xmm6=(00 01 10 11 20 21 30 31 04 05 14 15 24 25 34 35)
+    packsswb    xmm1, xmm0        ; xmm1=(02 03 12 13 22 23 32 33 06 07 16 17 26 27 36 37)
+    paddb       xmm6, xmm2
+    paddb       xmm1, xmm2
+
+    movdqa      xmm4, xmm6        ; transpose coefficients(phase 2)
+    punpcklwd   xmm6, xmm1        ; xmm6=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+    punpckhwd   xmm4, xmm1        ; xmm4=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+
+    movdqa      xmm7, xmm6        ; transpose coefficients(phase 3)
+    punpckldq   xmm6, xmm4        ; xmm6=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+    punpckhdq   xmm7, xmm4        ; xmm7=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+
+    pshufd      xmm5, xmm6, 0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+    pshufd      xmm3, xmm7, 0x4E  ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+
+    mov         rdxp, JSAMPROW rdi+0*SIZEOF_JSAMPROW
+    mov         rbxp, JSAMPROW rdi+2*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm6
+    movq        XMM_MMWORD rbx+rax*SIZEOF_JSAMPLE, xmm7
+    mov         rdxp, JSAMPROW rdi+1*SIZEOF_JSAMPROW
+    mov         rbxp, JSAMPROW rdi+3*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm5
+    movq        XMM_MMWORD rbx+rax*SIZEOF_JSAMPLE, xmm3
+
+    add         rsi, byte 4*SIZEOF_FAST_FLOAT  ; wsptr
+    add         rdi, byte 4*SIZEOF_JSAMPROW
+    dec         rcx                            ; ctr
+    jnz         near .rowloop
+
+    pop         rbx
+    uncollect_args 4
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jidctfst-sse2.asm Added

@@ -0,0 +1,491 @@
+;
+; jidctfst.asm - fast integer IDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the inverse DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jidctfst.c; see the jidctfst.c
+; for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  8  ; 14 is also OK.
+%define PASS1_BITS  2
+
+%if IFAST_SCALE_BITS != PASS1_BITS
+%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'."
+%endif
+
+%if CONST_BITS == 8
+F_1_082 equ 277              ; FIX(1.082392200)
+F_1_414 equ 362              ; FIX(1.414213562)
+F_1_847 equ 473              ; FIX(1.847759065)
+F_2_613 equ 669              ; FIX(2.613125930)
+F_1_613 equ (F_2_613 - 256)  ; FIX(2.613125930) - FIX(1)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_1_082 equ DESCALE(1162209775, 30 - CONST_BITS)  ; FIX(1.082392200)
+F_1_414 equ DESCALE(1518500249, 30 - CONST_BITS)  ; FIX(1.414213562)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_2_613 equ DESCALE(2805822602, 30 - CONST_BITS)  ; FIX(2.613125930)
+F_1_613 equ (F_2_613 - (1 << CONST_BITS))         ; FIX(2.613125930) - FIX(1)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS  2
+%define CONST_SHIFT              (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_ifast_sse2)
+
+EXTN(jconst_idct_ifast_sse2):
+
+PW_F1414       times 8  dw  F_1_414 << CONST_SHIFT
+PW_F1847       times 8  dw  F_1_847 << CONST_SHIFT
+PW_MF1613      times 8  dw -F_1_613 << CONST_SHIFT
+PW_F1082       times 8  dw  F_1_082 << CONST_SHIFT
+PB_CENTERJSAMP times 16 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_ifast_sse2(void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = jpeg_component_info *compptr
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13d = JDIMENSION output_col
+
+%define original_rbp  rbp + 0
+%define wk(i)         rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM        2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_ifast_sse2)
+
+EXTN(jsimd_idct_ifast_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 4
+
+    ; ---- Pass 1: process columns from input.
+
+    mov         rdx, r10                ; quantptr
+    mov         rsi, r11                ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_IFAST_SSE2
+    mov         eax, dword DWBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    jnz         near .columnDCT
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)
+    por         xmm1, xmm0
+    packsswb    xmm1, xmm1
+    packsswb    xmm1, xmm1
+    movd        eax, xmm1
+    test        rax, rax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+
+    movdqa      xmm7, xmm0              ; xmm0=in0=(00 01 02 03 04 05 06 07)
+    punpcklwd   xmm0, xmm0              ; xmm0=(00 00 01 01 02 02 03 03)
+    punpckhwd   xmm7, xmm7              ; xmm7=(04 04 05 05 06 06 07 07)
+
+    pshufd      xmm6, xmm0, 0x00        ; xmm6=col0=(00 00 00 00 00 00 00 00)
+    pshufd      xmm2, xmm0, 0x55        ; xmm2=col1=(01 01 01 01 01 01 01 01)
+    pshufd      xmm5, xmm0, 0xAA        ; xmm5=col2=(02 02 02 02 02 02 02 02)
+    pshufd      xmm0, xmm0, 0xFF        ; xmm0=col3=(03 03 03 03 03 03 03 03)
+    pshufd      xmm1, xmm7, 0x00        ; xmm1=col4=(04 04 04 04 04 04 04 04)
+    pshufd      xmm4, xmm7, 0x55        ; xmm4=col5=(05 05 05 05 05 05 05 05)
+    pshufd      xmm3, xmm7, 0xAA        ; xmm3=col6=(06 06 06 06 06 06 06 06)
+    pshufd      xmm7, xmm7, 0xFF        ; xmm7=col7=(07 07 07 07 07 07 07 07)
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=col1
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=col3
+    jmp         near .column_end
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_IFAST_MULT_TYPE)
+    movdqa      xmm2, XMMWORD XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(4,0,rdx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_IFAST_MULT_TYPE)
+
+    movdqa      xmm4, xmm0
+    movdqa      xmm5, xmm1
+    psubw       xmm0, xmm2              ; xmm0=tmp11
+    psubw       xmm1, xmm3
+    paddw       xmm4, xmm2              ; xmm4=tmp10
+    paddw       xmm5, xmm3              ; xmm5=tmp13
+
+    psllw       xmm1, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm1, rel PW_F1414
+    psubw       xmm1, xmm5              ; xmm1=tmp12
+
+    movdqa      xmm6, xmm4
+    movdqa      xmm7, xmm0
+    psubw       xmm4, xmm5              ; xmm4=tmp3
+    psubw       xmm0, xmm1              ; xmm0=tmp2
+    paddw       xmm6, xmm5              ; xmm6=tmp0
+    paddw       xmm7, xmm1              ; xmm7=tmp1
+
+    movdqa      XMMWORD wk(1), xmm4   ; wk(1)=tmp3
+    movdqa      XMMWORD wk(0), xmm0   ; wk(0)=tmp2
+
+    ; -- Odd part
+
+    movdqa      xmm2, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_IFAST_MULT_TYPE)
+    movdqa      xmm5, XMMWORD XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm5, XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_IFAST_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_IFAST_MULT_TYPE)
+
+    movdqa      xmm4, xmm2
+    movdqa      xmm0, xmm5
+    psubw       xmm2, xmm1              ; xmm2=z12
+    psubw       xmm5, xmm3              ; xmm5=z10
+    paddw       xmm4, xmm1              ; xmm4=z11
+    paddw       xmm0, xmm3              ; xmm0=z13
+
+    movdqa      xmm1, xmm5              ; xmm1=z10(unscaled)
+    psllw       xmm2, PRE_MULTIPLY_SCALE_BITS
+    psllw       xmm5, PRE_MULTIPLY_SCALE_BITS
+
+    movdqa      xmm3, xmm4
+    psubw       xmm4, xmm0
+    paddw       xmm3, xmm0              ; xmm3=tmp7
+
+    psllw       xmm4, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm4, rel PW_F1414    ; xmm4=tmp11
+
+    ; To avoid overflow...
+    ;
+    ; (Original)
+    ; tmp12 = -2.613125930 * z10 + z5;
+    ;
+    ; (This implementation)
+    ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+    ;       = -1.613125930 * z10 - z10 + z5;
+
+    movdqa      xmm0, xmm5
+    paddw       xmm5, xmm2
+    pmulhw      xmm5, rel PW_F1847    ; xmm5=z5
+    pmulhw      xmm0, rel PW_MF1613
+    pmulhw      xmm2, rel PW_F1082
+    psubw       xmm0, xmm1
+    psubw       xmm2, xmm5              ; xmm2=tmp10
+    paddw       xmm0, xmm5              ; xmm0=tmp12
+
+    ; -- Final output stage
+
+    psubw       xmm0, xmm3              ; xmm0=tmp6
+    movdqa      xmm1, xmm6
+    movdqa      xmm5, xmm7
+    paddw       xmm6, xmm3              ; xmm6=data0=(00 01 02 03 04 05 06 07)
+    paddw       xmm7, xmm0              ; xmm7=data1=(10 11 12 13 14 15 16 17)
+    psubw       xmm1, xmm3              ; xmm1=data7=(70 71 72 73 74 75 76 77)
+    psubw       xmm5, xmm0              ; xmm5=data6=(60 61 62 63 64 65 66 67)
+    psubw       xmm4, xmm0              ; xmm4=tmp5
+
+    movdqa      xmm3, xmm6              ; transpose coefficients(phase 1)
+    punpcklwd   xmm6, xmm7              ; xmm6=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm3, xmm7              ; xmm3=(04 14 05 15 06 16 07 17)
+    movdqa      xmm0, xmm5              ; transpose coefficients(phase 1)
+    punpcklwd   xmm5, xmm1              ; xmm5=(60 70 61 71 62 72 63 73)
+    punpckhwd   xmm0, xmm1              ; xmm0=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=tmp2
+    movdqa      xmm1, XMMWORD wk(1)   ; xmm1=tmp3
+
+    movdqa      XMMWORD wk(0), xmm5   ; wk(0)=(60 70 61 71 62 72 63 73)
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=(64 74 65 75 66 76 67 77)
+
+    paddw       xmm2, xmm4              ; xmm2=tmp4
+    movdqa      xmm5, xmm7
+    movdqa      xmm0, xmm1
+    paddw       xmm7, xmm4              ; xmm7=data2=(20 21 22 23 24 25 26 27)
+    paddw       xmm1, xmm2              ; xmm1=data4=(40 41 42 43 44 45 46 47)
+    psubw       xmm5, xmm4              ; xmm5=data5=(50 51 52 53 54 55 56 57)
+    psubw       xmm0, xmm2              ; xmm0=data3=(30 31 32 33 34 35 36 37)
+
+    movdqa      xmm4, xmm7              ; transpose coefficients(phase 1)
+    punpcklwd   xmm7, xmm0              ; xmm7=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm4, xmm0              ; xmm4=(24 34 25 35 26 36 27 37)
+    movdqa      xmm2, xmm1              ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm5              ; xmm1=(40 50 41 51 42 52 43 53)
+    punpckhwd   xmm2, xmm5              ; xmm2=(44 54 45 55 46 56 47 57)
+
+    movdqa      xmm0, xmm3              ; transpose coefficients(phase 2)
+    punpckldq   xmm3, xmm4              ; xmm3=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm0, xmm4              ; xmm0=(06 16 26 36 07 17 27 37)
+    movdqa      xmm5, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm7              ; xmm6=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm5, xmm7              ; xmm5=(02 12 22 32 03 13 23 33)
+
+    movdqa      xmm4, XMMWORD wk(0)   ; xmm4=(60 70 61 71 62 72 63 73)
+    movdqa      xmm7, XMMWORD wk(1)   ; xmm7=(64 74 65 75 66 76 67 77)
+
+    movdqa      XMMWORD wk(0), xmm3   ; wk(0)=(04 14 24 34 05 15 25 35)
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm3, xmm1              ; transpose coefficients(phase 2)
+    punpckldq   xmm1, xmm4              ; xmm1=(40 50 60 70 41 51 61 71)
+    punpckhdq   xmm3, xmm4              ; xmm3=(42 52 62 72 43 53 63 73)
+    movdqa      xmm0, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm7              ; xmm2=(44 54 64 74 45 55 65 75)
+    punpckhdq   xmm0, xmm7              ; xmm0=(46 56 66 76 47 57 67 77)
+
+    movdqa      xmm4, xmm6              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm6, xmm1              ; xmm6=col0=(00 10 20 30 40 50 60 70)
+    punpckhqdq  xmm4, xmm1              ; xmm4=col1=(01 11 21 31 41 51 61 71)
+    movdqa      xmm7, xmm5              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm5, xmm3              ; xmm5=col2=(02 12 22 32 42 52 62 72)
+    punpckhqdq  xmm7, xmm3              ; xmm7=col3=(03 13 23 33 43 53 63 73)
+
+    movdqa      xmm1, XMMWORD wk(0)   ; xmm1=(04 14 24 34 05 15 25 35)
+    movdqa      xmm3, XMMWORD wk(1)   ; xmm3=(06 16 26 36 07 17 27 37)
+
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=col1
+    movdqa      XMMWORD wk(1), xmm7   ; wk(1)=col3
+
+    movdqa      xmm4, xmm1              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm1, xmm2              ; xmm1=col4=(04 14 24 34 44 54 64 74)
+    punpckhqdq  xmm4, xmm2              ; xmm4=col5=(05 15 25 35 45 55 65 75)
+    movdqa      xmm7, xmm3              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm3, xmm0              ; xmm3=col6=(06 16 26 36 46 56 66 76)
+    punpckhqdq  xmm7, xmm0              ; xmm7=col7=(07 17 27 37 47 57 67 77)
+.column_end:
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         rax, original_rbp
+    mov         rdi, r12                ; (JSAMPROW *)
+    mov         eax, r13d
+
+    ; -- Even part
+
+    ; xmm6=col0, xmm5=col2, xmm1=col4, xmm3=col6
+
+    movdqa      xmm2, xmm6
+    movdqa      xmm0, xmm5
+    psubw       xmm6, xmm1              ; xmm6=tmp11
+    psubw       xmm5, xmm3
+    paddw       xmm2, xmm1              ; xmm2=tmp10
+    paddw       xmm0, xmm3              ; xmm0=tmp13
+
+    psllw       xmm5, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm5, rel PW_F1414
+    psubw       xmm5, xmm0              ; xmm5=tmp12
+
+    movdqa      xmm1, xmm2
+    movdqa      xmm3, xmm6
+    psubw       xmm2, xmm0              ; xmm2=tmp3
+    psubw       xmm6, xmm5              ; xmm6=tmp2
+    paddw       xmm1, xmm0              ; xmm1=tmp0
+    paddw       xmm3, xmm5              ; xmm3=tmp1
+
+    movdqa      xmm0, XMMWORD wk(0)   ; xmm0=col1
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=col3
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=tmp3
+    movdqa      XMMWORD wk(1), xmm6   ; wk(1)=tmp2
+
+    ; -- Odd part
+
+    ; xmm0=col1, xmm5=col3, xmm4=col5, xmm7=col7
+
+    movdqa      xmm2, xmm0
+    movdqa      xmm6, xmm4
+    psubw       xmm0, xmm7              ; xmm0=z12
+    psubw       xmm4, xmm5              ; xmm4=z10
+    paddw       xmm2, xmm7              ; xmm2=z11
+    paddw       xmm6, xmm5              ; xmm6=z13
+
+    movdqa      xmm7, xmm4              ; xmm7=z10(unscaled)
+    psllw       xmm0, PRE_MULTIPLY_SCALE_BITS
+    psllw       xmm4, PRE_MULTIPLY_SCALE_BITS
+
+    movdqa      xmm5, xmm2
+    psubw       xmm2, xmm6
+    paddw       xmm5, xmm6              ; xmm5=tmp7
+
+    psllw       xmm2, PRE_MULTIPLY_SCALE_BITS
+    pmulhw      xmm2, rel PW_F1414    ; xmm2=tmp11
+
+    ; To avoid overflow...
+    ;
+    ; (Original)
+    ; tmp12 = -2.613125930 * z10 + z5;
+    ;
+    ; (This implementation)
+    ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+    ;       = -1.613125930 * z10 - z10 + z5;
+
+    movdqa      xmm6, xmm4
+    paddw       xmm4, xmm0
+    pmulhw      xmm4, rel PW_F1847    ; xmm4=z5
+    pmulhw      xmm6, rel PW_MF1613
+    pmulhw      xmm0, rel PW_F1082
+    psubw       xmm6, xmm7
+    psubw       xmm0, xmm4              ; xmm0=tmp10
+    paddw       xmm6, xmm4              ; xmm6=tmp12
+
+    ; -- Final output stage
+
+    psubw       xmm6, xmm5              ; xmm6=tmp6
+    movdqa      xmm7, xmm1
+    movdqa      xmm4, xmm3
+    paddw       xmm1, xmm5              ; xmm1=data0=(00 10 20 30 40 50 60 70)
+    paddw       xmm3, xmm6              ; xmm3=data1=(01 11 21 31 41 51 61 71)
+    psraw       xmm1, (PASS1_BITS+3)    ; descale
+    psraw       xmm3, (PASS1_BITS+3)    ; descale
+    psubw       xmm7, xmm5              ; xmm7=data7=(07 17 27 37 47 57 67 77)
+    psubw       xmm4, xmm6              ; xmm4=data6=(06 16 26 36 46 56 66 76)
+    psraw       xmm7, (PASS1_BITS+3)    ; descale
+    psraw       xmm4, (PASS1_BITS+3)    ; descale
+    psubw       xmm2, xmm6              ; xmm2=tmp5
+
+    packsswb    xmm1, xmm4        ; xmm1=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+    packsswb    xmm3, xmm7        ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+    movdqa      xmm5, XMMWORD wk(1)   ; xmm5=tmp2
+    movdqa      xmm6, XMMWORD wk(0)   ; xmm6=tmp3
+
+    paddw       xmm0, xmm2              ; xmm0=tmp4
+    movdqa      xmm4, xmm5
+    movdqa      xmm7, xmm6
+    paddw       xmm5, xmm2              ; xmm5=data2=(02 12 22 32 42 52 62 72)
+    paddw       xmm6, xmm0              ; xmm6=data4=(04 14 24 34 44 54 64 74)
+    psraw       xmm5, (PASS1_BITS+3)    ; descale
+    psraw       xmm6, (PASS1_BITS+3)    ; descale
+    psubw       xmm4, xmm2              ; xmm4=data5=(05 15 25 35 45 55 65 75)
+    psubw       xmm7, xmm0              ; xmm7=data3=(03 13 23 33 43 53 63 73)
+    psraw       xmm4, (PASS1_BITS+3)    ; descale
+    psraw       xmm7, (PASS1_BITS+3)    ; descale
+
+    movdqa      xmm2, rel PB_CENTERJSAMP  ; xmm2=rel PB_CENTERJSAMP
+
+    packsswb    xmm5, xmm6        ; xmm5=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+    packsswb    xmm7, xmm4        ; xmm7=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+    paddb       xmm1, xmm2
+    paddb       xmm3, xmm2
+    paddb       xmm5, xmm2
+    paddb       xmm7, xmm2
+
+    movdqa      xmm0, xmm1        ; transpose coefficients(phase 1)
+    punpcklbw   xmm1, xmm3        ; xmm1=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+    punpckhbw   xmm0, xmm3        ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+    movdqa      xmm6, xmm5        ; transpose coefficients(phase 1)
+    punpcklbw   xmm5, xmm7        ; xmm5=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+    punpckhbw   xmm6, xmm7        ; xmm6=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+    movdqa      xmm4, xmm1        ; transpose coefficients(phase 2)
+    punpcklwd   xmm1, xmm5        ; xmm1=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+    punpckhwd   xmm4, xmm5        ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+    movdqa      xmm2, xmm6        ; transpose coefficients(phase 2)
+    punpcklwd   xmm6, xmm0        ; xmm6=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+    punpckhwd   xmm2, xmm0        ; xmm2=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+    movdqa      xmm3, xmm1        ; transpose coefficients(phase 3)
+    punpckldq   xmm1, xmm6        ; xmm1=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+    punpckhdq   xmm3, xmm6        ; xmm3=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+    movdqa      xmm7, xmm4        ; transpose coefficients(phase 3)
+    punpckldq   xmm4, xmm2        ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+    punpckhdq   xmm7, xmm2        ; xmm7=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+    pshufd      xmm5, xmm1, 0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+    pshufd      xmm0, xmm3, 0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+    pshufd      xmm6, xmm4, 0x4E  ; xmm6=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+    pshufd      xmm2, xmm7, 0x4E  ; xmm2=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+    mov         rdxp, JSAMPROW rdi+0*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+2*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm1
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm3
+    mov         rdxp, JSAMPROW rdi+4*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+6*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm4
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm7
+
+    mov         rdxp, JSAMPROW rdi+1*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+3*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm5
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm0
+    mov         rdxp, JSAMPROW rdi+5*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+7*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm6
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm2
+
+    uncollect_args 4
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jidctint-avx2.asm Added

@@ -0,0 +1,418 @@
+;
+; jidctint.asm - accurate integer IDCT (64-bit AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, 2018, 2020, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS + 3)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+; In-place 8x8x16-bit inverse matrix transpose using AVX2 instructions
+; %1-%4: Input/output registers
+; %5-%8: Temp registers
+
+%macro dotranspose 8
+    ; %5=(00 10 20 30 40 50 60 70  01 11 21 31 41 51 61 71)
+    ; %6=(03 13 23 33 43 53 63 73  02 12 22 32 42 52 62 72)
+    ; %7=(04 14 24 34 44 54 64 74  05 15 25 35 45 55 65 75)
+    ; %8=(07 17 27 37 47 57 67 77  06 16 26 36 46 56 66 76)
+
+    vpermq      %5, %1, 0xD8
+    vpermq      %6, %2, 0x72
+    vpermq      %7, %3, 0xD8
+    vpermq      %8, %4, 0x72
+    ; transpose coefficients(phase 1)
+    ; %5=(00 10 20 30 01 11 21 31  40 50 60 70 41 51 61 71)
+    ; %6=(02 12 22 32 03 13 23 33  42 52 62 72 43 53 63 73)
+    ; %7=(04 14 24 34 05 15 25 35  44 54 64 74 45 55 65 75)
+    ; %8=(06 16 26 36 07 17 27 37  46 56 66 76 47 57 67 77)
+
+    vpunpcklwd  %1, %5, %6
+    vpunpckhwd  %2, %5, %6
+    vpunpcklwd  %3, %7, %8
+    vpunpckhwd  %4, %7, %8
+    ; transpose coefficients(phase 2)
+    ; %1=(00 02 10 12 20 22 30 32  40 42 50 52 60 62 70 72)
+    ; %2=(01 03 11 13 21 23 31 33  41 43 51 53 61 63 71 73)
+    ; %3=(04 06 14 16 24 26 34 36  44 46 54 56 64 66 74 76)
+    ; %4=(05 07 15 17 25 27 35 37  45 47 55 57 65 67 75 77)
+
+    vpunpcklwd  %5, %1, %2
+    vpunpcklwd  %6, %3, %4
+    vpunpckhwd  %7, %1, %2
+    vpunpckhwd  %8, %3, %4
+    ; transpose coefficients(phase 3)
+    ; %5=(00 01 02 03 10 11 12 13  40 41 42 43 50 51 52 53)
+    ; %6=(04 05 06 07 14 15 16 17  44 45 46 47 54 55 56 57)
+    ; %7=(20 21 22 23 30 31 32 33  60 61 62 63 70 71 72 73)
+    ; %8=(24 25 26 27 34 35 36 37  64 65 66 67 74 75 76 77)
+
+    vpunpcklqdq %1, %5, %6
+    vpunpckhqdq %2, %5, %6
+    vpunpcklqdq %3, %7, %8
+    vpunpckhqdq %4, %7, %8
+    ; transpose coefficients(phase 4)
+    ; %1=(00 01 02 03 04 05 06 07  40 41 42 43 44 45 46 47)
+    ; %2=(10 11 12 13 14 15 16 17  50 51 52 53 54 55 56 57)
+    ; %3=(20 21 22 23 24 25 26 27  60 61 62 63 64 65 66 67)
+    ; %4=(30 31 32 33 34 35 36 37  70 71 72 73 74 75 76 77)
+%endmacro
+
+; --------------------------------------------------------------------------
+; In-place 8x8x16-bit accurate integer inverse DCT using AVX2 instructions
+; %1-%4:  Input/output registers
+; %5-%12: Temp registers
+; %9:     Pass (1 or 2)
+
+%macro dodct 13
+    ; -- Even part
+
+    ; (Original)
+    ; z1 = (z2 + z3) * 0.541196100;
+    ; tmp2 = z1 + z3 * -1.847759065;
+    ; tmp3 = z1 + z2 * 0.765366865;
+    ;
+    ; (This implementation)
+    ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+    ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+    vperm2i128  %6, %3, %3, 0x01        ; %6=in6_2
+    vpunpcklwd  %5, %3, %6              ; %5=in26_62L
+    vpunpckhwd  %6, %3, %6              ; %6=in26_62H
+    vpmaddwd    %5, %5, rel PW_F130_F054_MF130_F054  ; %5=tmp3_2L
+    vpmaddwd    %6, %6, rel PW_F130_F054_MF130_F054  ; %6=tmp3_2H
+
+    vperm2i128  %7, %1, %1, 0x01        ; %7=in4_0
+    vpsignw     %1, %1, rel PW_1_NEG1
+    vpaddw      %7, %7, %1              ; %7=(in0+in4)_(in0-in4)
+
+    vpxor       %1, %1, %1
+    vpunpcklwd  %8, %1, %7              ; %8=tmp0_1L
+    vpunpckhwd  %1, %1, %7              ; %1=tmp0_1H
+    vpsrad      %8, %8, (16-CONST_BITS)  ; vpsrad %8,16 & vpslld %8,CONST_BITS
+    vpsrad      %1, %1, (16-CONST_BITS)  ; vpsrad %1,16 & vpslld %1,CONST_BITS
+
+    vpsubd      %11, %8, %5             ; %11=tmp0_1L-tmp3_2L=tmp13_12L
+    vpaddd      %9, %8, %5              ; %9=tmp0_1L+tmp3_2L=tmp10_11L
+    vpsubd      %12, %1, %6             ; %12=tmp0_1H-tmp3_2H=tmp13_12H
+    vpaddd      %10, %1, %6             ; %10=tmp0_1H+tmp3_2H=tmp10_11H
+
+    ; -- Odd part
+
+    vpaddw      %1, %4, %2              ; %1=in7_5+in3_1=z3_4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    vperm2i128  %8, %1, %1, 0x01        ; %8=z4_3
+    vpunpcklwd  %7, %1, %8              ; %7=z34_43L
+    vpunpckhwd  %8, %1, %8              ; %8=z34_43H
+    vpmaddwd    %7, %7, rel PW_MF078_F117_F078_F117  ; %7=z3_4L
+    vpmaddwd    %8, %8, rel PW_MF078_F117_F078_F117  ; %8=z3_4H
+
+    ; (Original)
+    ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+    ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+    ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+    ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+    ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+    ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+    ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+    ; tmp0 += z3;  tmp1 += z4;
+    ; tmp2 += z3;  tmp3 += z4;
+
+    vperm2i128  %2, %2, %2, 0x01        ; %2=in1_3
+    vpunpcklwd  %3, %4, %2              ; %3=in71_53L
+    vpunpckhwd  %4, %4, %2              ; %4=in71_53H
+
+    vpmaddwd    %5, %3, rel PW_MF060_MF089_MF050_MF256  ; %5=tmp0_1L
+    vpmaddwd    %6, %4, rel PW_MF060_MF089_MF050_MF256  ; %6=tmp0_1H
+    vpaddd      %5, %5, %7              ; %5=tmp0_1L+z3_4L=tmp0_1L
+    vpaddd      %6, %6, %8              ; %6=tmp0_1H+z3_4H=tmp0_1H
+
+    vpmaddwd    %3, %3, rel PW_MF089_F060_MF256_F050  ; %3=tmp3_2L
+    vpmaddwd    %4, %4, rel PW_MF089_F060_MF256_F050  ; %4=tmp3_2H
+    vperm2i128  %7, %7, %7, 0x01        ; %7=z4_3L
+    vperm2i128  %8, %8, %8, 0x01        ; %8=z4_3H
+    vpaddd      %7, %3, %7              ; %7=tmp3_2L+z4_3L=tmp3_2L
+    vpaddd      %8, %4, %8              ; %8=tmp3_2H+z4_3H=tmp3_2H
+
+    ; -- Final output stage
+
+    vpaddd      %1, %9, %7              ; %1=tmp10_11L+tmp3_2L=data0_1L
+    vpaddd      %2, %10, %8             ; %2=tmp10_11H+tmp3_2H=data0_1H
+    vpaddd      %1, %1, rel PD_DESCALE_P %+ %13
+    vpaddd      %2, %2, rel PD_DESCALE_P %+ %13
+    vpsrad      %1, %1, DESCALE_P %+ %13
+    vpsrad      %2, %2, DESCALE_P %+ %13
+    vpackssdw   %1, %1, %2              ; %1=data0_1
+
+    vpsubd      %3, %9, %7              ; %3=tmp10_11L-tmp3_2L=data7_6L
+    vpsubd      %4, %10, %8             ; %4=tmp10_11H-tmp3_2H=data7_6H
+    vpaddd      %3, %3, rel PD_DESCALE_P %+ %13
+    vpaddd      %4, %4, rel PD_DESCALE_P %+ %13
+    vpsrad      %3, %3, DESCALE_P %+ %13
+    vpsrad      %4, %4, DESCALE_P %+ %13
+    vpackssdw   %4, %3, %4              ; %4=data7_6
+
+    vpaddd      %7, %11, %5             ; %7=tmp13_12L+tmp0_1L=data3_2L
+    vpaddd      %8, %12, %6             ; %8=tmp13_12H+tmp0_1H=data3_2H
+    vpaddd      %7, %7, rel PD_DESCALE_P %+ %13
+    vpaddd      %8, %8, rel PD_DESCALE_P %+ %13
+    vpsrad      %7, %7, DESCALE_P %+ %13
+    vpsrad      %8, %8, DESCALE_P %+ %13
+    vpackssdw   %2, %7, %8              ; %2=data3_2
+
+    vpsubd      %7, %11, %5             ; %7=tmp13_12L-tmp0_1L=data4_5L
+    vpsubd      %8, %12, %6             ; %8=tmp13_12H-tmp0_1H=data4_5H
+    vpaddd      %7, %7, rel PD_DESCALE_P %+ %13
+    vpaddd      %8, %8, rel PD_DESCALE_P %+ %13
+    vpsrad      %7, %7, DESCALE_P %+ %13
+    vpsrad      %8, %8, DESCALE_P %+ %13
+    vpackssdw   %3, %7, %8              ; %3=data4_5
+%endmacro
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_islow_avx2)
+
+EXTN(jconst_idct_islow_avx2):
+
+PW_F130_F054_MF130_F054    times 4  dw  (F_0_541 + F_0_765),  F_0_541
+                           times 4  dw  (F_0_541 - F_1_847),  F_0_541
+PW_MF078_F117_F078_F117    times 4  dw  (F_1_175 - F_1_961),  F_1_175
+                           times 4  dw  (F_1_175 - F_0_390),  F_1_175
+PW_MF060_MF089_MF050_MF256 times 4  dw  (F_0_298 - F_0_899), -F_0_899
+                           times 4  dw  (F_2_053 - F_2_562), -F_2_562
+PW_MF089_F060_MF256_F050   times 4  dw -F_0_899, (F_1_501 - F_0_899)
+                           times 4  dw -F_2_562, (F_3_072 - F_2_562)
+PD_DESCALE_P1              times 8  dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2              times 8  dd  1 << (DESCALE_P2 - 1)
+PB_CENTERJSAMP             times 32 db  CENTERJSAMPLE
+PW_1_NEG1                  times 8  dw  1
+                           times 8  dw -1
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_avx2(void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = jpeg_component_info *compptr
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13d = JDIMENSION output_col
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_islow_avx2)
+
+EXTN(jsimd_idct_islow_avx2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    push_xmm    4
+    collect_args 4
+
+    ; ---- Pass 1: process columns.
+
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_AVX2
+    mov         eax, dword DWBLOCK(1,0,r11,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,r11,SIZEOF_JCOEF)
+    jnz         near .columnDCT
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,r11,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,r11,SIZEOF_JCOEF)
+    vpor        xmm0, xmm0, XMMWORD XMMBLOCK(3,0,r11,SIZEOF_JCOEF)
+    vpor        xmm1, xmm1, XMMWORD XMMBLOCK(4,0,r11,SIZEOF_JCOEF)
+    vpor        xmm0, xmm0, XMMWORD XMMBLOCK(5,0,r11,SIZEOF_JCOEF)
+    vpor        xmm1, xmm1, XMMWORD XMMBLOCK(6,0,r11,SIZEOF_JCOEF)
+    vpor        xmm0, xmm0, XMMWORD XMMBLOCK(7,0,r11,SIZEOF_JCOEF)
+    vpor        xmm1, xmm1, xmm0
+    vpacksswb   xmm1, xmm1, xmm1
+    vpacksswb   xmm1, xmm1, xmm1
+    movd        eax, xmm1
+    test        rax, rax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movdqa      xmm5, XMMWORD XMMBLOCK(0,0,r11,SIZEOF_JCOEF)
+    vpmullw     xmm5, xmm5, XMMWORD XMMBLOCK(0,0,r10,SIZEOF_ISLOW_MULT_TYPE)
+
+    vpsllw      xmm5, xmm5, PASS1_BITS
+
+    vpunpcklwd  xmm4, xmm5, xmm5        ; xmm4=(00 00 01 01 02 02 03 03)
+    vpunpckhwd  xmm5, xmm5, xmm5        ; xmm5=(04 04 05 05 06 06 07 07)
+    vinserti128 ymm4, ymm4, xmm5, 1
+
+    vpshufd     ymm0, ymm4, 0x00        ; ymm0=col0_4=(00 00 00 00 00 00 00 00  04 04 04 04 04 04 04 04)
+    vpshufd     ymm1, ymm4, 0x55        ; ymm1=col1_5=(01 01 01 01 01 01 01 01  05 05 05 05 05 05 05 05)
+    vpshufd     ymm2, ymm4, 0xAA        ; ymm2=col2_6=(02 02 02 02 02 02 02 02  06 06 06 06 06 06 06 06)
+    vpshufd     ymm3, ymm4, 0xFF        ; ymm3=col3_7=(03 03 03 03 03 03 03 03  07 07 07 07 07 07 07 07)
+
+    jmp         near .column_end
+%endif
+.columnDCT:
+
+    vmovdqu     ymm4, YMMWORD YMMBLOCK(0,0,r11,SIZEOF_JCOEF)  ; ymm4=in0_1
+    vmovdqu     ymm5, YMMWORD YMMBLOCK(2,0,r11,SIZEOF_JCOEF)  ; ymm5=in2_3
+    vmovdqu     ymm6, YMMWORD YMMBLOCK(4,0,r11,SIZEOF_JCOEF)  ; ymm6=in4_5
+    vmovdqu     ymm7, YMMWORD YMMBLOCK(6,0,r11,SIZEOF_JCOEF)  ; ymm7=in6_7
+    vpmullw     ymm4, ymm4, YMMWORD YMMBLOCK(0,0,r10,SIZEOF_ISLOW_MULT_TYPE)
+    vpmullw     ymm5, ymm5, YMMWORD YMMBLOCK(2,0,r10,SIZEOF_ISLOW_MULT_TYPE)
+    vpmullw     ymm6, ymm6, YMMWORD YMMBLOCK(4,0,r10,SIZEOF_ISLOW_MULT_TYPE)
+    vpmullw     ymm7, ymm7, YMMWORD YMMBLOCK(6,0,r10,SIZEOF_ISLOW_MULT_TYPE)
+
+    vperm2i128  ymm0, ymm4, ymm6, 0x20  ; ymm0=in0_4
+    vperm2i128  ymm1, ymm5, ymm4, 0x31  ; ymm1=in3_1
+    vperm2i128  ymm2, ymm5, ymm7, 0x20  ; ymm2=in2_6
+    vperm2i128  ymm3, ymm7, ymm6, 0x31  ; ymm3=in7_5
+
+    dodct ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7, ymm8, ymm9, ymm10, ymm11, 1
+    ; ymm0=data0_1, ymm1=data3_2, ymm2=data4_5, ymm3=data7_6
+
+    dotranspose ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7
+    ; ymm0=data0_4, ymm1=data1_5, ymm2=data2_6, ymm3=data3_7
+
+.column_end:
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta r11 + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta r11 + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta r11 + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta r11 + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows.
+
+    vperm2i128  ymm4, ymm3, ymm1, 0x31  ; ymm3=in7_5
+    vperm2i128  ymm1, ymm3, ymm1, 0x20  ; ymm1=in3_1
+
+    dodct ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7, ymm8, ymm9, ymm10, ymm11, 2
+    ; ymm0=data0_1, ymm1=data3_2, ymm2=data4_5, ymm4=data7_6
+
+    dotranspose ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7
+    ; ymm0=data0_4, ymm1=data1_5, ymm2=data2_6, ymm4=data3_7
+
+    vpacksswb   ymm0, ymm0, ymm1        ; ymm0=data01_45
+    vpacksswb   ymm1, ymm2, ymm4        ; ymm1=data23_67
+    vpaddb      ymm0, ymm0, rel PB_CENTERJSAMP
+    vpaddb      ymm1, ymm1, rel PB_CENTERJSAMP
+
+    vextracti128 xmm6, ymm1, 1          ; xmm3=data67
+    vextracti128 xmm4, ymm0, 1          ; xmm2=data45
+    vextracti128 xmm2, ymm1, 0          ; xmm1=data23
+    vextracti128 xmm0, ymm0, 0          ; xmm0=data01
+
+    vpshufd     xmm1, xmm0, 0x4E  ; xmm1=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+    vpshufd     xmm3, xmm2, 0x4E  ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+    vpshufd     xmm5, xmm4, 0x4E  ; xmm5=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+    vpshufd     xmm7, xmm6, 0x4E  ; xmm7=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+    vzeroupper
+
+    mov         eax, r13d
+
+    mov         rdxp, JSAMPROW r12+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rsip, JSAMPROW r12+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm0
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm1
+
+    mov         rdxp, JSAMPROW r12+2*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rsip, JSAMPROW r12+3*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm2
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm3
+
+    mov         rdxp, JSAMPROW r12+4*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rsip, JSAMPROW r12+5*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm4
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm5
+
+    mov         rdxp, JSAMPROW r12+6*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rsip, JSAMPROW r12+7*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm6
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm7
+
+    uncollect_args 4
+    pop_xmm     4
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jidctint-sse2.asm Added

@@ -0,0 +1,847 @@
+;
+; jidctint.asm - accurate integer IDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, 2020, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slower but more accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS  13
+%define PASS1_BITS  2
+
+%define DESCALE_P1  (CONST_BITS - PASS1_BITS)
+%define DESCALE_P2  (CONST_BITS + PASS1_BITS + 3)
+
+%if CONST_BITS == 13
+F_0_298 equ  2446  ; FIX(0.298631336)
+F_0_390 equ  3196  ; FIX(0.390180644)
+F_0_541 equ  4433  ; FIX(0.541196100)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_175 equ  9633  ; FIX(1.175875602)
+F_1_501 equ 12299  ; FIX(1.501321110)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_1_961 equ 16069  ; FIX(1.961570560)
+F_2_053 equ 16819  ; FIX(2.053119869)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_072 equ 25172  ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS)  ; FIX(0.298631336)
+F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS)  ; FIX(0.390180644)
+F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS)  ; FIX(0.541196100)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS)  ; FIX(1.175875602)
+F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS)  ; FIX(1.501321110)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS)  ; FIX(1.961570560)
+F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS)  ; FIX(2.053119869)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS)  ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_islow_sse2)
+
+EXTN(jconst_idct_islow_sse2):
+
+PW_F130_F054   times 4  dw  (F_0_541 + F_0_765),  F_0_541
+PW_F054_MF130  times 4  dw  F_0_541, (F_0_541 - F_1_847)
+PW_MF078_F117  times 4  dw  (F_1_175 - F_1_961),  F_1_175
+PW_F117_F078   times 4  dw  F_1_175, (F_1_175 - F_0_390)
+PW_MF060_MF089 times 4  dw  (F_0_298 - F_0_899), -F_0_899
+PW_MF089_F060  times 4  dw -F_0_899, (F_1_501 - F_0_899)
+PW_MF050_MF256 times 4  dw  (F_2_053 - F_2_562), -F_2_562
+PW_MF256_F050  times 4  dw -F_2_562, (F_3_072 - F_2_562)
+PD_DESCALE_P1  times 4  dd  1 << (DESCALE_P1 - 1)
+PD_DESCALE_P2  times 4  dd  1 << (DESCALE_P2 - 1)
+PB_CENTERJSAMP times 16 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_sse2(void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = jpeg_component_info *compptr
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13d = JDIMENSION output_col
+
+%define original_rbp  rbp + 0
+%define wk(i)         rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM        12
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_islow_sse2)
+
+EXTN(jsimd_idct_islow_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 4
+
+    ; ---- Pass 1: process columns from input.
+
+    mov         rdx, r10                ; quantptr
+    mov         rsi, r11                ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_SSE2
+    mov         eax, dword DWBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    jnz         near .columnDCT
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)
+    por         xmm1, xmm0
+    packsswb    xmm1, xmm1
+    packsswb    xmm1, xmm1
+    movd        eax, xmm1
+    test        rax, rax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movdqa      xmm5, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm5, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+
+    psllw       xmm5, PASS1_BITS
+
+    movdqa      xmm4, xmm5              ; xmm5=in0=(00 01 02 03 04 05 06 07)
+    punpcklwd   xmm5, xmm5              ; xmm5=(00 00 01 01 02 02 03 03)
+    punpckhwd   xmm4, xmm4              ; xmm4=(04 04 05 05 06 06 07 07)
+
+    pshufd      xmm7, xmm5, 0x00        ; xmm7=col0=(00 00 00 00 00 00 00 00)
+    pshufd      xmm6, xmm5, 0x55        ; xmm6=col1=(01 01 01 01 01 01 01 01)
+    pshufd      xmm1, xmm5, 0xAA        ; xmm1=col2=(02 02 02 02 02 02 02 02)
+    pshufd      xmm5, xmm5, 0xFF        ; xmm5=col3=(03 03 03 03 03 03 03 03)
+    pshufd      xmm0, xmm4, 0x00        ; xmm0=col4=(04 04 04 04 04 04 04 04)
+    pshufd      xmm3, xmm4, 0x55        ; xmm3=col5=(05 05 05 05 05 05 05 05)
+    pshufd      xmm2, xmm4, 0xAA        ; xmm2=col6=(06 06 06 06 06 06 06 06)
+    pshufd      xmm4, xmm4, 0xFF        ; xmm4=col7=(07 07 07 07 07 07 07 07)
+
+    movdqa      XMMWORD wk(8), xmm6   ; wk(8)=col1
+    movdqa      XMMWORD wk(9), xmm5   ; wk(9)=col3
+    movdqa      XMMWORD wk(10), xmm3  ; wk(10)=col5
+    movdqa      XMMWORD wk(11), xmm4  ; wk(11)=col7
+    jmp         near .column_end
+%endif
+.columnDCT:
+
+    ; -- Even part
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    movdqa      xmm2, XMMWORD XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(4,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; (Original)
+    ; z1 = (z2 + z3) * 0.541196100;
+    ; tmp2 = z1 + z3 * -1.847759065;
+    ; tmp3 = z1 + z2 * 0.765366865;
+    ;
+    ; (This implementation)
+    ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+    ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+    movdqa      xmm4, xmm1              ; xmm1=in2=z2
+    movdqa      xmm5, xmm1
+    punpcklwd   xmm4, xmm3              ; xmm3=in6=z3
+    punpckhwd   xmm5, xmm3
+    movdqa      xmm1, xmm4
+    movdqa      xmm3, xmm5
+    pmaddwd     xmm4, rel PW_F130_F054   ; xmm4=tmp3L
+    pmaddwd     xmm5, rel PW_F130_F054   ; xmm5=tmp3H
+    pmaddwd     xmm1, rel PW_F054_MF130  ; xmm1=tmp2L
+    pmaddwd     xmm3, rel PW_F054_MF130  ; xmm3=tmp2H
+
+    movdqa      xmm6, xmm0
+    paddw       xmm0, xmm2              ; xmm0=in0+in4
+    psubw       xmm6, xmm2              ; xmm6=in0-in4
+
+    pxor        xmm7, xmm7
+    pxor        xmm2, xmm2
+    punpcklwd   xmm7, xmm0              ; xmm7=tmp0L
+    punpckhwd   xmm2, xmm0              ; xmm2=tmp0H
+    psrad       xmm7, (16-CONST_BITS)   ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+    psrad       xmm2, (16-CONST_BITS)   ; psrad xmm2,16 & pslld xmm2,CONST_BITS
+
+    movdqa      xmm0, xmm7
+    paddd       xmm7, xmm4              ; xmm7=tmp10L
+    psubd       xmm0, xmm4              ; xmm0=tmp13L
+    movdqa      xmm4, xmm2
+    paddd       xmm2, xmm5              ; xmm2=tmp10H
+    psubd       xmm4, xmm5              ; xmm4=tmp13H
+
+    movdqa      XMMWORD wk(0), xmm7   ; wk(0)=tmp10L
+    movdqa      XMMWORD wk(1), xmm2   ; wk(1)=tmp10H
+    movdqa      XMMWORD wk(2), xmm0   ; wk(2)=tmp13L
+    movdqa      XMMWORD wk(3), xmm4   ; wk(3)=tmp13H
+
+    pxor        xmm5, xmm5
+    pxor        xmm7, xmm7
+    punpcklwd   xmm5, xmm6              ; xmm5=tmp1L
+    punpckhwd   xmm7, xmm6              ; xmm7=tmp1H
+    psrad       xmm5, (16-CONST_BITS)   ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+    psrad       xmm7, (16-CONST_BITS)   ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+
+    movdqa      xmm2, xmm5
+    paddd       xmm5, xmm1              ; xmm5=tmp11L
+    psubd       xmm2, xmm1              ; xmm2=tmp12L
+    movdqa      xmm0, xmm7
+    paddd       xmm7, xmm3              ; xmm7=tmp11H
+    psubd       xmm0, xmm3              ; xmm0=tmp12H
+
+    movdqa      XMMWORD wk(4), xmm5   ; wk(4)=tmp11L
+    movdqa      XMMWORD wk(5), xmm7   ; wk(5)=tmp11H
+    movdqa      XMMWORD wk(6), xmm2   ; wk(6)=tmp12L
+    movdqa      XMMWORD wk(7), xmm0   ; wk(7)=tmp12H
+
+    ; -- Odd part
+
+    movdqa      xmm4, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm6, XMMWORD XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm4, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm6, XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    movdqa      xmm1, XMMWORD XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm1, XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+
+    movdqa      xmm5, xmm6
+    movdqa      xmm7, xmm4
+    paddw       xmm5, xmm3              ; xmm5=z3
+    paddw       xmm7, xmm1              ; xmm7=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movdqa      xmm2, xmm5
+    movdqa      xmm0, xmm5
+    punpcklwd   xmm2, xmm7
+    punpckhwd   xmm0, xmm7
+    movdqa      xmm5, xmm2
+    movdqa      xmm7, xmm0
+    pmaddwd     xmm2, rel PW_MF078_F117  ; xmm2=z3L
+    pmaddwd     xmm0, rel PW_MF078_F117  ; xmm0=z3H
+    pmaddwd     xmm5, rel PW_F117_F078   ; xmm5=z4L
+    pmaddwd     xmm7, rel PW_F117_F078   ; xmm7=z4H
+
+    movdqa      XMMWORD wk(10), xmm2  ; wk(10)=z3L
+    movdqa      XMMWORD wk(11), xmm0  ; wk(11)=z3H
+
+    ; (Original)
+    ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+    ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+    ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+    ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+    ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+    ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+    ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+    ; tmp0 += z3;  tmp1 += z4;
+    ; tmp2 += z3;  tmp3 += z4;
+
+    movdqa      xmm2, xmm3
+    movdqa      xmm0, xmm3
+    punpcklwd   xmm2, xmm4
+    punpckhwd   xmm0, xmm4
+    movdqa      xmm3, xmm2
+    movdqa      xmm4, xmm0
+    pmaddwd     xmm2, rel PW_MF060_MF089  ; xmm2=tmp0L
+    pmaddwd     xmm0, rel PW_MF060_MF089  ; xmm0=tmp0H
+    pmaddwd     xmm3, rel PW_MF089_F060   ; xmm3=tmp3L
+    pmaddwd     xmm4, rel PW_MF089_F060   ; xmm4=tmp3H
+
+    paddd       xmm2, XMMWORD wk(10)  ; xmm2=tmp0L
+    paddd       xmm0, XMMWORD wk(11)  ; xmm0=tmp0H
+    paddd       xmm3, xmm5              ; xmm3=tmp3L
+    paddd       xmm4, xmm7              ; xmm4=tmp3H
+
+    movdqa      XMMWORD wk(8), xmm2   ; wk(8)=tmp0L
+    movdqa      XMMWORD wk(9), xmm0   ; wk(9)=tmp0H
+
+    movdqa      xmm2, xmm1
+    movdqa      xmm0, xmm1
+    punpcklwd   xmm2, xmm6
+    punpckhwd   xmm0, xmm6
+    movdqa      xmm1, xmm2
+    movdqa      xmm6, xmm0
+    pmaddwd     xmm2, rel PW_MF050_MF256  ; xmm2=tmp1L
+    pmaddwd     xmm0, rel PW_MF050_MF256  ; xmm0=tmp1H
+    pmaddwd     xmm1, rel PW_MF256_F050   ; xmm1=tmp2L
+    pmaddwd     xmm6, rel PW_MF256_F050   ; xmm6=tmp2H
+
+    paddd       xmm2, xmm5              ; xmm2=tmp1L
+    paddd       xmm0, xmm7              ; xmm0=tmp1H
+    paddd       xmm1, XMMWORD wk(10)  ; xmm1=tmp2L
+    paddd       xmm6, XMMWORD wk(11)  ; xmm6=tmp2H
+
+    movdqa      XMMWORD wk(10), xmm2  ; wk(10)=tmp1L
+    movdqa      XMMWORD wk(11), xmm0  ; wk(11)=tmp1H
+
+    ; -- Final output stage
+
+    movdqa      xmm5, XMMWORD wk(0)   ; xmm5=tmp10L
+    movdqa      xmm7, XMMWORD wk(1)   ; xmm7=tmp10H
+
+    movdqa      xmm2, xmm5
+    movdqa      xmm0, xmm7
+    paddd       xmm5, xmm3              ; xmm5=data0L
+    paddd       xmm7, xmm4              ; xmm7=data0H
+    psubd       xmm2, xmm3              ; xmm2=data7L
+    psubd       xmm0, xmm4              ; xmm0=data7H
+
+    movdqa      xmm3, rel PD_DESCALE_P1  ; xmm3=rel PD_DESCALE_P1
+
+    paddd       xmm5, xmm3
+    paddd       xmm7, xmm3
+    psrad       xmm5, DESCALE_P1
+    psrad       xmm7, DESCALE_P1
+    paddd       xmm2, xmm3
+    paddd       xmm0, xmm3
+    psrad       xmm2, DESCALE_P1
+    psrad       xmm0, DESCALE_P1
+
+    packssdw    xmm5, xmm7              ; xmm5=data0=(00 01 02 03 04 05 06 07)
+    packssdw    xmm2, xmm0              ; xmm2=data7=(70 71 72 73 74 75 76 77)
+
+    movdqa      xmm4, XMMWORD wk(4)   ; xmm4=tmp11L
+    movdqa      xmm3, XMMWORD wk(5)   ; xmm3=tmp11H
+
+    movdqa      xmm7, xmm4
+    movdqa      xmm0, xmm3
+    paddd       xmm4, xmm1              ; xmm4=data1L
+    paddd       xmm3, xmm6              ; xmm3=data1H
+    psubd       xmm7, xmm1              ; xmm7=data6L
+    psubd       xmm0, xmm6              ; xmm0=data6H
+
+    movdqa      xmm1, rel PD_DESCALE_P1  ; xmm1=rel PD_DESCALE_P1
+
+    paddd       xmm4, xmm1
+    paddd       xmm3, xmm1
+    psrad       xmm4, DESCALE_P1
+    psrad       xmm3, DESCALE_P1
+    paddd       xmm7, xmm1
+    paddd       xmm0, xmm1
+    psrad       xmm7, DESCALE_P1
+    psrad       xmm0, DESCALE_P1
+
+    packssdw    xmm4, xmm3              ; xmm4=data1=(10 11 12 13 14 15 16 17)
+    packssdw    xmm7, xmm0              ; xmm7=data6=(60 61 62 63 64 65 66 67)
+
+    movdqa      xmm6, xmm5              ; transpose coefficients(phase 1)
+    punpcklwd   xmm5, xmm4              ; xmm5=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm6, xmm4              ; xmm6=(04 14 05 15 06 16 07 17)
+    movdqa      xmm1, xmm7              ; transpose coefficients(phase 1)
+    punpcklwd   xmm7, xmm2              ; xmm7=(60 70 61 71 62 72 63 73)
+    punpckhwd   xmm1, xmm2              ; xmm1=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm3, XMMWORD wk(6)   ; xmm3=tmp12L
+    movdqa      xmm0, XMMWORD wk(7)   ; xmm0=tmp12H
+    movdqa      xmm4, XMMWORD wk(10)  ; xmm4=tmp1L
+    movdqa      xmm2, XMMWORD wk(11)  ; xmm2=tmp1H
+
+    movdqa      XMMWORD wk(0), xmm5   ; wk(0)=(00 10 01 11 02 12 03 13)
+    movdqa      XMMWORD wk(1), xmm6   ; wk(1)=(04 14 05 15 06 16 07 17)
+    movdqa      XMMWORD wk(4), xmm7   ; wk(4)=(60 70 61 71 62 72 63 73)
+    movdqa      XMMWORD wk(5), xmm1   ; wk(5)=(64 74 65 75 66 76 67 77)
+
+    movdqa      xmm5, xmm3
+    movdqa      xmm6, xmm0
+    paddd       xmm3, xmm4              ; xmm3=data2L
+    paddd       xmm0, xmm2              ; xmm0=data2H
+    psubd       xmm5, xmm4              ; xmm5=data5L
+    psubd       xmm6, xmm2              ; xmm6=data5H
+
+    movdqa      xmm7, rel PD_DESCALE_P1  ; xmm7=rel PD_DESCALE_P1
+
+    paddd       xmm3, xmm7
+    paddd       xmm0, xmm7
+    psrad       xmm3, DESCALE_P1
+    psrad       xmm0, DESCALE_P1
+    paddd       xmm5, xmm7
+    paddd       xmm6, xmm7
+    psrad       xmm5, DESCALE_P1
+    psrad       xmm6, DESCALE_P1
+
+    packssdw    xmm3, xmm0              ; xmm3=data2=(20 21 22 23 24 25 26 27)
+    packssdw    xmm5, xmm6              ; xmm5=data5=(50 51 52 53 54 55 56 57)
+
+    movdqa      xmm1, XMMWORD wk(2)   ; xmm1=tmp13L
+    movdqa      xmm4, XMMWORD wk(3)   ; xmm4=tmp13H
+    movdqa      xmm2, XMMWORD wk(8)   ; xmm2=tmp0L
+    movdqa      xmm7, XMMWORD wk(9)   ; xmm7=tmp0H
+
+    movdqa      xmm0, xmm1
+    movdqa      xmm6, xmm4
+    paddd       xmm1, xmm2              ; xmm1=data3L
+    paddd       xmm4, xmm7              ; xmm4=data3H
+    psubd       xmm0, xmm2              ; xmm0=data4L
+    psubd       xmm6, xmm7              ; xmm6=data4H
+
+    movdqa      xmm2, rel PD_DESCALE_P1  ; xmm2=rel PD_DESCALE_P1
+
+    paddd       xmm1, xmm2
+    paddd       xmm4, xmm2
+    psrad       xmm1, DESCALE_P1
+    psrad       xmm4, DESCALE_P1
+    paddd       xmm0, xmm2
+    paddd       xmm6, xmm2
+    psrad       xmm0, DESCALE_P1
+    psrad       xmm6, DESCALE_P1
+
+    packssdw    xmm1, xmm4              ; xmm1=data3=(30 31 32 33 34 35 36 37)
+    packssdw    xmm0, xmm6              ; xmm0=data4=(40 41 42 43 44 45 46 47)
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=(00 10 01 11 02 12 03 13)
+    movdqa      xmm2, XMMWORD wk(1)   ; xmm2=(04 14 05 15 06 16 07 17)
+
+    movdqa      xmm4, xmm3              ; transpose coefficients(phase 1)
+    punpcklwd   xmm3, xmm1              ; xmm3=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm4, xmm1              ; xmm4=(24 34 25 35 26 36 27 37)
+    movdqa      xmm6, xmm0              ; transpose coefficients(phase 1)
+    punpcklwd   xmm0, xmm5              ; xmm0=(40 50 41 51 42 52 43 53)
+    punpckhwd   xmm6, xmm5              ; xmm6=(44 54 45 55 46 56 47 57)
+
+    movdqa      xmm1, xmm7              ; transpose coefficients(phase 2)
+    punpckldq   xmm7, xmm3              ; xmm7=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm1, xmm3              ; xmm1=(02 12 22 32 03 13 23 33)
+    movdqa      xmm5, xmm2              ; transpose coefficients(phase 2)
+    punpckldq   xmm2, xmm4              ; xmm2=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm5, xmm4              ; xmm5=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm3, XMMWORD wk(4)   ; xmm3=(60 70 61 71 62 72 63 73)
+    movdqa      xmm4, XMMWORD wk(5)   ; xmm4=(64 74 65 75 66 76 67 77)
+
+    movdqa      XMMWORD wk(6), xmm2   ; wk(6)=(04 14 24 34 05 15 25 35)
+    movdqa      XMMWORD wk(7), xmm5   ; wk(7)=(06 16 26 36 07 17 27 37)
+
+    movdqa      xmm2, xmm0              ; transpose coefficients(phase 2)
+    punpckldq   xmm0, xmm3              ; xmm0=(40 50 60 70 41 51 61 71)
+    punpckhdq   xmm2, xmm3              ; xmm2=(42 52 62 72 43 53 63 73)
+    movdqa      xmm5, xmm6              ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm4              ; xmm6=(44 54 64 74 45 55 65 75)
+    punpckhdq   xmm5, xmm4              ; xmm5=(46 56 66 76 47 57 67 77)
+
+    movdqa      xmm3, xmm7              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm7, xmm0              ; xmm7=col0=(00 10 20 30 40 50 60 70)
+    punpckhqdq  xmm3, xmm0              ; xmm3=col1=(01 11 21 31 41 51 61 71)
+    movdqa      xmm4, xmm1              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm1, xmm2              ; xmm1=col2=(02 12 22 32 42 52 62 72)
+    punpckhqdq  xmm4, xmm2              ; xmm4=col3=(03 13 23 33 43 53 63 73)
+
+    movdqa      xmm0, XMMWORD wk(6)   ; xmm0=(04 14 24 34 05 15 25 35)
+    movdqa      xmm2, XMMWORD wk(7)   ; xmm2=(06 16 26 36 07 17 27 37)
+
+    movdqa      XMMWORD wk(8), xmm3   ; wk(8)=col1
+    movdqa      XMMWORD wk(9), xmm4   ; wk(9)=col3
+
+    movdqa      xmm3, xmm0              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm0, xmm6              ; xmm0=col4=(04 14 24 34 44 54 64 74)
+    punpckhqdq  xmm3, xmm6              ; xmm3=col5=(05 15 25 35 45 55 65 75)
+    movdqa      xmm4, xmm2              ; transpose coefficients(phase 3)
+    punpcklqdq  xmm2, xmm5              ; xmm2=col6=(06 16 26 36 46 56 66 76)
+    punpckhqdq  xmm4, xmm5              ; xmm4=col7=(07 17 27 37 47 57 67 77)
+
+    movdqa      XMMWORD wk(10), xmm3  ; wk(10)=col5
+    movdqa      XMMWORD wk(11), xmm4  ; wk(11)=col7
+.column_end:
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows from work array, store into output array.
+
+    mov         rax, original_rbp
+    mov         rdi, r12                ; (JSAMPROW *)
+    mov         eax, r13d
+
+    ; -- Even part
+
+    ; xmm7=col0, xmm1=col2, xmm0=col4, xmm2=col6
+
+    ; (Original)
+    ; z1 = (z2 + z3) * 0.541196100;
+    ; tmp2 = z1 + z3 * -1.847759065;
+    ; tmp3 = z1 + z2 * 0.765366865;
+    ;
+    ; (This implementation)
+    ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+    ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+    movdqa      xmm6, xmm1              ; xmm1=in2=z2
+    movdqa      xmm5, xmm1
+    punpcklwd   xmm6, xmm2              ; xmm2=in6=z3
+    punpckhwd   xmm5, xmm2
+    movdqa      xmm1, xmm6
+    movdqa      xmm2, xmm5
+    pmaddwd     xmm6, rel PW_F130_F054   ; xmm6=tmp3L
+    pmaddwd     xmm5, rel PW_F130_F054   ; xmm5=tmp3H
+    pmaddwd     xmm1, rel PW_F054_MF130  ; xmm1=tmp2L
+    pmaddwd     xmm2, rel PW_F054_MF130  ; xmm2=tmp2H
+
+    movdqa      xmm3, xmm7
+    paddw       xmm7, xmm0              ; xmm7=in0+in4
+    psubw       xmm3, xmm0              ; xmm3=in0-in4
+
+    pxor        xmm4, xmm4
+    pxor        xmm0, xmm0
+    punpcklwd   xmm4, xmm7              ; xmm4=tmp0L
+    punpckhwd   xmm0, xmm7              ; xmm0=tmp0H
+    psrad       xmm4, (16-CONST_BITS)   ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+    psrad       xmm0, (16-CONST_BITS)   ; psrad xmm0,16 & pslld xmm0,CONST_BITS
+
+    movdqa      xmm7, xmm4
+    paddd       xmm4, xmm6              ; xmm4=tmp10L
+    psubd       xmm7, xmm6              ; xmm7=tmp13L
+    movdqa      xmm6, xmm0
+    paddd       xmm0, xmm5              ; xmm0=tmp10H
+    psubd       xmm6, xmm5              ; xmm6=tmp13H
+
+    movdqa      XMMWORD wk(0), xmm4   ; wk(0)=tmp10L
+    movdqa      XMMWORD wk(1), xmm0   ; wk(1)=tmp10H
+    movdqa      XMMWORD wk(2), xmm7   ; wk(2)=tmp13L
+    movdqa      XMMWORD wk(3), xmm6   ; wk(3)=tmp13H
+
+    pxor        xmm5, xmm5
+    pxor        xmm4, xmm4
+    punpcklwd   xmm5, xmm3              ; xmm5=tmp1L
+    punpckhwd   xmm4, xmm3              ; xmm4=tmp1H
+    psrad       xmm5, (16-CONST_BITS)   ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+    psrad       xmm4, (16-CONST_BITS)   ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+
+    movdqa      xmm0, xmm5
+    paddd       xmm5, xmm1              ; xmm5=tmp11L
+    psubd       xmm0, xmm1              ; xmm0=tmp12L
+    movdqa      xmm7, xmm4
+    paddd       xmm4, xmm2              ; xmm4=tmp11H
+    psubd       xmm7, xmm2              ; xmm7=tmp12H
+
+    movdqa      XMMWORD wk(4), xmm5   ; wk(4)=tmp11L
+    movdqa      XMMWORD wk(5), xmm4   ; wk(5)=tmp11H
+    movdqa      XMMWORD wk(6), xmm0   ; wk(6)=tmp12L
+    movdqa      XMMWORD wk(7), xmm7   ; wk(7)=tmp12H
+
+    ; -- Odd part
+
+    movdqa      xmm6, XMMWORD wk(9)   ; xmm6=col3
+    movdqa      xmm3, XMMWORD wk(8)   ; xmm3=col1
+    movdqa      xmm1, XMMWORD wk(11)  ; xmm1=col7
+    movdqa      xmm2, XMMWORD wk(10)  ; xmm2=col5
+
+    movdqa      xmm5, xmm6
+    movdqa      xmm4, xmm3
+    paddw       xmm5, xmm1              ; xmm5=z3
+    paddw       xmm4, xmm2              ; xmm4=z4
+
+    ; (Original)
+    ; z5 = (z3 + z4) * 1.175875602;
+    ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+    ; z3 += z5;  z4 += z5;
+    ;
+    ; (This implementation)
+    ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+    ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+    movdqa      xmm0, xmm5
+    movdqa      xmm7, xmm5
+    punpcklwd   xmm0, xmm4
+    punpckhwd   xmm7, xmm4
+    movdqa      xmm5, xmm0
+    movdqa      xmm4, xmm7
+    pmaddwd     xmm0, rel PW_MF078_F117  ; xmm0=z3L
+    pmaddwd     xmm7, rel PW_MF078_F117  ; xmm7=z3H
+    pmaddwd     xmm5, rel PW_F117_F078   ; xmm5=z4L
+    pmaddwd     xmm4, rel PW_F117_F078   ; xmm4=z4H
+
+    movdqa      XMMWORD wk(10), xmm0  ; wk(10)=z3L
+    movdqa      XMMWORD wk(11), xmm7  ; wk(11)=z3H
+
+    ; (Original)
+    ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+    ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+    ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+    ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+    ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+    ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+    ;
+    ; (This implementation)
+    ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+    ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+    ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+    ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+    ; tmp0 += z3;  tmp1 += z4;
+    ; tmp2 += z3;  tmp3 += z4;
+
+    movdqa      xmm0, xmm1
+    movdqa      xmm7, xmm1
+    punpcklwd   xmm0, xmm3
+    punpckhwd   xmm7, xmm3
+    movdqa      xmm1, xmm0
+    movdqa      xmm3, xmm7
+    pmaddwd     xmm0, rel PW_MF060_MF089  ; xmm0=tmp0L
+    pmaddwd     xmm7, rel PW_MF060_MF089  ; xmm7=tmp0H
+    pmaddwd     xmm1, rel PW_MF089_F060   ; xmm1=tmp3L
+    pmaddwd     xmm3, rel PW_MF089_F060   ; xmm3=tmp3H
+
+    paddd       xmm0, XMMWORD wk(10)  ; xmm0=tmp0L
+    paddd       xmm7, XMMWORD wk(11)  ; xmm7=tmp0H
+    paddd       xmm1, xmm5              ; xmm1=tmp3L
+    paddd       xmm3, xmm4              ; xmm3=tmp3H
+
+    movdqa      XMMWORD wk(8), xmm0   ; wk(8)=tmp0L
+    movdqa      XMMWORD wk(9), xmm7   ; wk(9)=tmp0H
+
+    movdqa      xmm0, xmm2
+    movdqa      xmm7, xmm2
+    punpcklwd   xmm0, xmm6
+    punpckhwd   xmm7, xmm6
+    movdqa      xmm2, xmm0
+    movdqa      xmm6, xmm7
+    pmaddwd     xmm0, rel PW_MF050_MF256  ; xmm0=tmp1L
+    pmaddwd     xmm7, rel PW_MF050_MF256  ; xmm7=tmp1H
+    pmaddwd     xmm2, rel PW_MF256_F050   ; xmm2=tmp2L
+    pmaddwd     xmm6, rel PW_MF256_F050   ; xmm6=tmp2H
+
+    paddd       xmm0, xmm5              ; xmm0=tmp1L
+    paddd       xmm7, xmm4              ; xmm7=tmp1H
+    paddd       xmm2, XMMWORD wk(10)  ; xmm2=tmp2L
+    paddd       xmm6, XMMWORD wk(11)  ; xmm6=tmp2H
+
+    movdqa      XMMWORD wk(10), xmm0  ; wk(10)=tmp1L
+    movdqa      XMMWORD wk(11), xmm7  ; wk(11)=tmp1H
+
+    ; -- Final output stage
+
+    movdqa      xmm5, XMMWORD wk(0)   ; xmm5=tmp10L
+    movdqa      xmm4, XMMWORD wk(1)   ; xmm4=tmp10H
+
+    movdqa      xmm0, xmm5
+    movdqa      xmm7, xmm4
+    paddd       xmm5, xmm1              ; xmm5=data0L
+    paddd       xmm4, xmm3              ; xmm4=data0H
+    psubd       xmm0, xmm1              ; xmm0=data7L
+    psubd       xmm7, xmm3              ; xmm7=data7H
+
+    movdqa      xmm1, rel PD_DESCALE_P2  ; xmm1=rel PD_DESCALE_P2
+
+    paddd       xmm5, xmm1
+    paddd       xmm4, xmm1
+    psrad       xmm5, DESCALE_P2
+    psrad       xmm4, DESCALE_P2
+    paddd       xmm0, xmm1
+    paddd       xmm7, xmm1
+    psrad       xmm0, DESCALE_P2
+    psrad       xmm7, DESCALE_P2
+
+    packssdw    xmm5, xmm4              ; xmm5=data0=(00 10 20 30 40 50 60 70)
+    packssdw    xmm0, xmm7              ; xmm0=data7=(07 17 27 37 47 57 67 77)
+
+    movdqa      xmm3, XMMWORD wk(4)   ; xmm3=tmp11L
+    movdqa      xmm1, XMMWORD wk(5)   ; xmm1=tmp11H
+
+    movdqa      xmm4, xmm3
+    movdqa      xmm7, xmm1
+    paddd       xmm3, xmm2              ; xmm3=data1L
+    paddd       xmm1, xmm6              ; xmm1=data1H
+    psubd       xmm4, xmm2              ; xmm4=data6L
+    psubd       xmm7, xmm6              ; xmm7=data6H
+
+    movdqa      xmm2, rel PD_DESCALE_P2  ; xmm2=rel PD_DESCALE_P2
+
+    paddd       xmm3, xmm2
+    paddd       xmm1, xmm2
+    psrad       xmm3, DESCALE_P2
+    psrad       xmm1, DESCALE_P2
+    paddd       xmm4, xmm2
+    paddd       xmm7, xmm2
+    psrad       xmm4, DESCALE_P2
+    psrad       xmm7, DESCALE_P2
+
+    packssdw    xmm3, xmm1              ; xmm3=data1=(01 11 21 31 41 51 61 71)
+    packssdw    xmm4, xmm7              ; xmm4=data6=(06 16 26 36 46 56 66 76)
+
+    packsswb    xmm5, xmm4              ; xmm5=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+    packsswb    xmm3, xmm0              ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+    movdqa      xmm6, XMMWORD wk(6)   ; xmm6=tmp12L
+    movdqa      xmm2, XMMWORD wk(7)   ; xmm2=tmp12H
+    movdqa      xmm1, XMMWORD wk(10)  ; xmm1=tmp1L
+    movdqa      xmm7, XMMWORD wk(11)  ; xmm7=tmp1H
+
+    movdqa      XMMWORD wk(0), xmm5   ; wk(0)=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+    movdqa      XMMWORD wk(1), xmm3   ; wk(1)=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+    movdqa      xmm4, xmm6
+    movdqa      xmm0, xmm2
+    paddd       xmm6, xmm1              ; xmm6=data2L
+    paddd       xmm2, xmm7              ; xmm2=data2H
+    psubd       xmm4, xmm1              ; xmm4=data5L
+    psubd       xmm0, xmm7              ; xmm0=data5H
+
+    movdqa      xmm5, rel PD_DESCALE_P2  ; xmm5=rel PD_DESCALE_P2
+
+    paddd       xmm6, xmm5
+    paddd       xmm2, xmm5
+    psrad       xmm6, DESCALE_P2
+    psrad       xmm2, DESCALE_P2
+    paddd       xmm4, xmm5
+    paddd       xmm0, xmm5
+    psrad       xmm4, DESCALE_P2
+    psrad       xmm0, DESCALE_P2
+
+    packssdw    xmm6, xmm2              ; xmm6=data2=(02 12 22 32 42 52 62 72)
+    packssdw    xmm4, xmm0              ; xmm4=data5=(05 15 25 35 45 55 65 75)
+
+    movdqa      xmm3, XMMWORD wk(2)   ; xmm3=tmp13L
+    movdqa      xmm1, XMMWORD wk(3)   ; xmm1=tmp13H
+    movdqa      xmm7, XMMWORD wk(8)   ; xmm7=tmp0L
+    movdqa      xmm5, XMMWORD wk(9)   ; xmm5=tmp0H
+
+    movdqa      xmm2, xmm3
+    movdqa      xmm0, xmm1
+    paddd       xmm3, xmm7              ; xmm3=data3L
+    paddd       xmm1, xmm5              ; xmm1=data3H
+    psubd       xmm2, xmm7              ; xmm2=data4L
+    psubd       xmm0, xmm5              ; xmm0=data4H
+
+    movdqa      xmm7, rel PD_DESCALE_P2  ; xmm7=rel PD_DESCALE_P2
+
+    paddd       xmm3, xmm7
+    paddd       xmm1, xmm7
+    psrad       xmm3, DESCALE_P2
+    psrad       xmm1, DESCALE_P2
+    paddd       xmm2, xmm7
+    paddd       xmm0, xmm7
+    psrad       xmm2, DESCALE_P2
+    psrad       xmm0, DESCALE_P2
+
+    movdqa      xmm5, rel PB_CENTERJSAMP  ; xmm5=rel PB_CENTERJSAMP
+
+    packssdw    xmm3, xmm1             ; xmm3=data3=(03 13 23 33 43 53 63 73)
+    packssdw    xmm2, xmm0             ; xmm2=data4=(04 14 24 34 44 54 64 74)
+
+    movdqa      xmm7, XMMWORD wk(0)  ; xmm7=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+    movdqa      xmm1, XMMWORD wk(1)  ; xmm1=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+    packsswb    xmm6, xmm2             ; xmm6=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+    packsswb    xmm3, xmm4             ; xmm3=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+    paddb       xmm7, xmm5
+    paddb       xmm1, xmm5
+    paddb       xmm6, xmm5
+    paddb       xmm3, xmm5
+
+    movdqa      xmm0, xmm7        ; transpose coefficients(phase 1)
+    punpcklbw   xmm7, xmm1        ; xmm7=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+    punpckhbw   xmm0, xmm1        ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+    movdqa      xmm2, xmm6        ; transpose coefficients(phase 1)
+    punpcklbw   xmm6, xmm3        ; xmm6=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+    punpckhbw   xmm2, xmm3        ; xmm2=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+    movdqa      xmm4, xmm7        ; transpose coefficients(phase 2)
+    punpcklwd   xmm7, xmm6        ; xmm7=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+    punpckhwd   xmm4, xmm6        ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+    movdqa      xmm5, xmm2        ; transpose coefficients(phase 2)
+    punpcklwd   xmm2, xmm0        ; xmm2=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+    punpckhwd   xmm5, xmm0        ; xmm5=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+    movdqa      xmm1, xmm7        ; transpose coefficients(phase 3)
+    punpckldq   xmm7, xmm2        ; xmm7=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+    punpckhdq   xmm1, xmm2        ; xmm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+    movdqa      xmm3, xmm4        ; transpose coefficients(phase 3)
+    punpckldq   xmm4, xmm5        ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+    punpckhdq   xmm3, xmm5        ; xmm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+    pshufd      xmm6, xmm7, 0x4E  ; xmm6=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+    pshufd      xmm0, xmm1, 0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+    pshufd      xmm2, xmm4, 0x4E  ; xmm2=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+    pshufd      xmm5, xmm3, 0x4E  ; xmm5=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+    mov         rdxp, JSAMPROW rdi+0*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+2*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm7
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm1
+    mov         rdxp, JSAMPROW rdi+4*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+6*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm4
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm3
+
+    mov         rdxp, JSAMPROW rdi+1*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+3*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm6
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm0
+    mov         rdxp, JSAMPROW rdi+5*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+7*SIZEOF_JSAMPROW
+    movq        XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE, xmm2
+    movq        XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE, xmm5
+
+    uncollect_args 4
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jidctred-sse2.asm Added

@@ -0,0 +1,574 @@
+;
+; jidctred.asm - reduced-size IDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains inverse-DCT routines that produce reduced-size
+; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.
+; The following code is based directly on the IJG's original jidctred.c;
+; see the jidctred.c for more details.
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS    13
+%define PASS1_BITS    2
+
+%define DESCALE_P1_4  (CONST_BITS - PASS1_BITS + 1)
+%define DESCALE_P2_4  (CONST_BITS + PASS1_BITS + 3 + 1)
+%define DESCALE_P1_2  (CONST_BITS - PASS1_BITS + 2)
+%define DESCALE_P2_2  (CONST_BITS + PASS1_BITS + 3 + 2)
+
+%if CONST_BITS == 13
+F_0_211 equ  1730  ; FIX(0.211164243)
+F_0_509 equ  4176  ; FIX(0.509795579)
+F_0_601 equ  4926  ; FIX(0.601344887)
+F_0_720 equ  5906  ; FIX(0.720959822)
+F_0_765 equ  6270  ; FIX(0.765366865)
+F_0_850 equ  6967  ; FIX(0.850430095)
+F_0_899 equ  7373  ; FIX(0.899976223)
+F_1_061 equ  8697  ; FIX(1.061594337)
+F_1_272 equ 10426  ; FIX(1.272758580)
+F_1_451 equ 11893  ; FIX(1.451774981)
+F_1_847 equ 15137  ; FIX(1.847759065)
+F_2_172 equ 17799  ; FIX(2.172734803)
+F_2_562 equ 20995  ; FIX(2.562915447)
+F_3_624 equ 29692  ; FIX(3.624509785)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x, n)  (((x) + (1 << ((n) - 1))) >> (n))
+F_0_211 equ DESCALE( 226735879, 30 - CONST_BITS)  ; FIX(0.211164243)
+F_0_509 equ DESCALE( 547388834, 30 - CONST_BITS)  ; FIX(0.509795579)
+F_0_601 equ DESCALE( 645689155, 30 - CONST_BITS)  ; FIX(0.601344887)
+F_0_720 equ DESCALE( 774124714, 30 - CONST_BITS)  ; FIX(0.720959822)
+F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS)  ; FIX(0.765366865)
+F_0_850 equ DESCALE( 913142361, 30 - CONST_BITS)  ; FIX(0.850430095)
+F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS)  ; FIX(0.899976223)
+F_1_061 equ DESCALE(1139878239, 30 - CONST_BITS)  ; FIX(1.061594337)
+F_1_272 equ DESCALE(1366614119, 30 - CONST_BITS)  ; FIX(1.272758580)
+F_1_451 equ DESCALE(1558831516, 30 - CONST_BITS)  ; FIX(1.451774981)
+F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS)  ; FIX(1.847759065)
+F_2_172 equ DESCALE(2332956230, 30 - CONST_BITS)  ; FIX(2.172734803)
+F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS)  ; FIX(2.562915447)
+F_3_624 equ DESCALE(3891787747, 30 - CONST_BITS)  ; FIX(3.624509785)
+%endif
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_CONST
+
+    alignz      32
+    GLOBAL_DATA(jconst_idct_red_sse2)
+
+EXTN(jconst_idct_red_sse2):
+
+PW_F184_MF076   times 4  dw  F_1_847, -F_0_765
+PW_F256_F089    times 4  dw  F_2_562,  F_0_899
+PW_F106_MF217   times 4  dw  F_1_061, -F_2_172
+PW_MF060_MF050  times 4  dw -F_0_601, -F_0_509
+PW_F145_MF021   times 4  dw  F_1_451, -F_0_211
+PW_F362_MF127   times 4  dw  F_3_624, -F_1_272
+PW_F085_MF072   times 4  dw  F_0_850, -F_0_720
+PD_DESCALE_P1_4 times 4  dd  1 << (DESCALE_P1_4 - 1)
+PD_DESCALE_P2_4 times 4  dd  1 << (DESCALE_P2_4 - 1)
+PD_DESCALE_P1_2 times 4  dd  1 << (DESCALE_P1_2 - 1)
+PD_DESCALE_P2_2 times 4  dd  1 << (DESCALE_P2_2 - 1)
+PB_CENTERJSAMP  times 16 db  CENTERJSAMPLE
+
+    alignz      32
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 4x4 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_4x4_sse2(void *dct_table, JCOEFPTR coef_block,
+;                     JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = void *dct_table
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13d = JDIMENSION output_col
+
+%define original_rbp  rbp + 0
+%define wk(i)         rbp - (WK_NUM - (i)) * SIZEOF_XMMWORD
+                                        ; xmmword wkWK_NUM
+%define WK_NUM        2
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_4x4_sse2)
+
+EXTN(jsimd_idct_4x4_sse2):
+    push        rbp
+    mov         rax, rsp                     ; rax = original rbp
+    sub         rsp, byte 4
+    and         rsp, byte (-SIZEOF_XMMWORD)  ; align to 128 bits
+    mov         rsp, rax
+    mov         rbp, rsp                     ; rbp = aligned rbp
+    lea         rsp, wk(0)
+    collect_args 4
+
+    ; ---- Pass 1: process columns from input.
+
+    mov         rdx, r10                ; quantptr
+    mov         rsi, r11                ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_4X4_SSE2
+    mov         eax, dword DWBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    or          eax, dword DWBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    jnz         short .columnDCT
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)
+    por         xmm0, XMMWORD XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)
+    por         xmm1, XMMWORD XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)
+    por         xmm0, xmm1
+    packsswb    xmm0, xmm0
+    packsswb    xmm0, xmm0
+    movd        eax, xmm0
+    test        rax, rax
+    jnz         short .columnDCT
+
+    ; -- AC terms all zero
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+
+    psllw       xmm0, PASS1_BITS
+
+    movdqa      xmm3, xmm0        ; xmm0=in0=(00 01 02 03 04 05 06 07)
+    punpcklwd   xmm0, xmm0        ; xmm0=(00 00 01 01 02 02 03 03)
+    punpckhwd   xmm3, xmm3        ; xmm3=(04 04 05 05 06 06 07 07)
+
+    pshufd      xmm1, xmm0, 0x50  ; xmm1=col0 col1=(00 00 00 00 01 01 01 01)
+    pshufd      xmm0, xmm0, 0xFA  ; xmm0=col2 col3=(02 02 02 02 03 03 03 03)
+    pshufd      xmm6, xmm3, 0x50  ; xmm6=col4 col5=(04 04 04 04 05 05 05 05)
+    pshufd      xmm3, xmm3, 0xFA  ; xmm3=col6 col7=(06 06 06 06 07 07 07 07)
+
+    jmp         near .column_end
+%endif
+.columnDCT:
+
+    ; -- Odd part
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    movdqa      xmm2, XMMWORD XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+
+    movdqa      xmm4, xmm0
+    movdqa      xmm5, xmm0
+    punpcklwd   xmm4, xmm1
+    punpckhwd   xmm5, xmm1
+    movdqa      xmm0, xmm4
+    movdqa      xmm1, xmm5
+    pmaddwd     xmm4, rel PW_F256_F089   ; xmm4=(tmp2L)
+    pmaddwd     xmm5, rel PW_F256_F089   ; xmm5=(tmp2H)
+    pmaddwd     xmm0, rel PW_F106_MF217  ; xmm0=(tmp0L)
+    pmaddwd     xmm1, rel PW_F106_MF217  ; xmm1=(tmp0H)
+
+    movdqa      xmm6, xmm2
+    movdqa      xmm7, xmm2
+    punpcklwd   xmm6, xmm3
+    punpckhwd   xmm7, xmm3
+    movdqa      xmm2, xmm6
+    movdqa      xmm3, xmm7
+    pmaddwd     xmm6, rel PW_MF060_MF050  ; xmm6=(tmp2L)
+    pmaddwd     xmm7, rel PW_MF060_MF050  ; xmm7=(tmp2H)
+    pmaddwd     xmm2, rel PW_F145_MF021   ; xmm2=(tmp0L)
+    pmaddwd     xmm3, rel PW_F145_MF021   ; xmm3=(tmp0H)
+
+    paddd       xmm6, xmm4              ; xmm6=tmp2L
+    paddd       xmm7, xmm5              ; xmm7=tmp2H
+    paddd       xmm2, xmm0              ; xmm2=tmp0L
+    paddd       xmm3, xmm1              ; xmm3=tmp0H
+
+    movdqa      XMMWORD wk(0), xmm2   ; wk(0)=tmp0L
+    movdqa      XMMWORD wk(1), xmm3   ; wk(1)=tmp0H
+
+    ; -- Even part
+
+    movdqa      xmm4, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm5, XMMWORD XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm0, XMMWORD XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm4, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm5, XMMWORD XMMBLOCK(2,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm0, XMMWORD XMMBLOCK(6,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+
+    pxor        xmm1, xmm1
+    pxor        xmm2, xmm2
+    punpcklwd   xmm1, xmm4               ; xmm1=tmp0L
+    punpckhwd   xmm2, xmm4               ; xmm2=tmp0H
+    psrad       xmm1, (16-CONST_BITS-1)  ; psrad xmm1,16 & pslld xmm1,CONST_BITS+1
+    psrad       xmm2, (16-CONST_BITS-1)  ; psrad xmm2,16 & pslld xmm2,CONST_BITS+1
+
+    movdqa      xmm3, xmm5              ; xmm5=in2=z2
+    punpcklwd   xmm5, xmm0              ; xmm0=in6=z3
+    punpckhwd   xmm3, xmm0
+    pmaddwd     xmm5, rel PW_F184_MF076  ; xmm5=tmp2L
+    pmaddwd     xmm3, rel PW_F184_MF076  ; xmm3=tmp2H
+
+    movdqa      xmm4, xmm1
+    movdqa      xmm0, xmm2
+    paddd       xmm1, xmm5              ; xmm1=tmp10L
+    paddd       xmm2, xmm3              ; xmm2=tmp10H
+    psubd       xmm4, xmm5              ; xmm4=tmp12L
+    psubd       xmm0, xmm3              ; xmm0=tmp12H
+
+    ; -- Final output stage
+
+    movdqa      xmm5, xmm1
+    movdqa      xmm3, xmm2
+    paddd       xmm1, xmm6              ; xmm1=data0L
+    paddd       xmm2, xmm7              ; xmm2=data0H
+    psubd       xmm5, xmm6              ; xmm5=data3L
+    psubd       xmm3, xmm7              ; xmm3=data3H
+
+    movdqa      xmm6, rel PD_DESCALE_P1_4  ; xmm6=rel PD_DESCALE_P1_4
+
+    paddd       xmm1, xmm6
+    paddd       xmm2, xmm6
+    psrad       xmm1, DESCALE_P1_4
+    psrad       xmm2, DESCALE_P1_4
+    paddd       xmm5, xmm6
+    paddd       xmm3, xmm6
+    psrad       xmm5, DESCALE_P1_4
+    psrad       xmm3, DESCALE_P1_4
+
+    packssdw    xmm1, xmm2              ; xmm1=data0=(00 01 02 03 04 05 06 07)
+    packssdw    xmm5, xmm3              ; xmm5=data3=(30 31 32 33 34 35 36 37)
+
+    movdqa      xmm7, XMMWORD wk(0)   ; xmm7=tmp0L
+    movdqa      xmm6, XMMWORD wk(1)   ; xmm6=tmp0H
+
+    movdqa      xmm2, xmm4
+    movdqa      xmm3, xmm0
+    paddd       xmm4, xmm7              ; xmm4=data1L
+    paddd       xmm0, xmm6              ; xmm0=data1H
+    psubd       xmm2, xmm7              ; xmm2=data2L
+    psubd       xmm3, xmm6              ; xmm3=data2H
+
+    movdqa      xmm7, rel PD_DESCALE_P1_4  ; xmm7=rel PD_DESCALE_P1_4
+
+    paddd       xmm4, xmm7
+    paddd       xmm0, xmm7
+    psrad       xmm4, DESCALE_P1_4
+    psrad       xmm0, DESCALE_P1_4
+    paddd       xmm2, xmm7
+    paddd       xmm3, xmm7
+    psrad       xmm2, DESCALE_P1_4
+    psrad       xmm3, DESCALE_P1_4
+
+    packssdw    xmm4, xmm0        ; xmm4=data1=(10 11 12 13 14 15 16 17)
+    packssdw    xmm2, xmm3        ; xmm2=data2=(20 21 22 23 24 25 26 27)
+
+    movdqa      xmm6, xmm1        ; transpose coefficients(phase 1)
+    punpcklwd   xmm1, xmm4        ; xmm1=(00 10 01 11 02 12 03 13)
+    punpckhwd   xmm6, xmm4        ; xmm6=(04 14 05 15 06 16 07 17)
+    movdqa      xmm7, xmm2        ; transpose coefficients(phase 1)
+    punpcklwd   xmm2, xmm5        ; xmm2=(20 30 21 31 22 32 23 33)
+    punpckhwd   xmm7, xmm5        ; xmm7=(24 34 25 35 26 36 27 37)
+
+    movdqa      xmm0, xmm1        ; transpose coefficients(phase 2)
+    punpckldq   xmm1, xmm2        ; xmm1=col0 col1=(00 10 20 30 01 11 21 31)
+    punpckhdq   xmm0, xmm2        ; xmm0=col2 col3=(02 12 22 32 03 13 23 33)
+    movdqa      xmm3, xmm6        ; transpose coefficients(phase 2)
+    punpckldq   xmm6, xmm7        ; xmm6=col4 col5=(04 14 24 34 05 15 25 35)
+    punpckhdq   xmm3, xmm7        ; xmm3=col6 col7=(06 16 26 36 07 17 27 37)
+.column_end:
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows, store into output array.
+
+    mov         rax, original_rbp
+    mov         rdi, r12                ; (JSAMPROW *)
+    mov         eax, r13d
+
+    ; -- Even part
+
+    pxor        xmm4, xmm4
+    punpcklwd   xmm4, xmm1               ; xmm4=tmp0
+    psrad       xmm4, (16-CONST_BITS-1)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS+1
+
+    ; -- Odd part
+
+    punpckhwd   xmm1, xmm0
+    punpckhwd   xmm6, xmm3
+    movdqa      xmm5, xmm1
+    movdqa      xmm2, xmm6
+    pmaddwd     xmm1, rel PW_F256_F089    ; xmm1=(tmp2)
+    pmaddwd     xmm6, rel PW_MF060_MF050  ; xmm6=(tmp2)
+    pmaddwd     xmm5, rel PW_F106_MF217   ; xmm5=(tmp0)
+    pmaddwd     xmm2, rel PW_F145_MF021   ; xmm2=(tmp0)
+
+    paddd       xmm6, xmm1              ; xmm6=tmp2
+    paddd       xmm2, xmm5              ; xmm2=tmp0
+
+    ; -- Even part
+
+    punpcklwd   xmm0, xmm3
+    pmaddwd     xmm0, rel PW_F184_MF076  ; xmm0=tmp2
+
+    movdqa      xmm7, xmm4
+    paddd       xmm4, xmm0              ; xmm4=tmp10
+    psubd       xmm7, xmm0              ; xmm7=tmp12
+
+    ; -- Final output stage
+
+    movdqa      xmm1, rel PD_DESCALE_P2_4  ; xmm1=rel PD_DESCALE_P2_4
+
+    movdqa      xmm5, xmm4
+    movdqa      xmm3, xmm7
+    paddd       xmm4, xmm6              ; xmm4=data0=(00 10 20 30)
+    paddd       xmm7, xmm2              ; xmm7=data1=(01 11 21 31)
+    psubd       xmm5, xmm6              ; xmm5=data3=(03 13 23 33)
+    psubd       xmm3, xmm2              ; xmm3=data2=(02 12 22 32)
+
+    paddd       xmm4, xmm1
+    paddd       xmm7, xmm1
+    psrad       xmm4, DESCALE_P2_4
+    psrad       xmm7, DESCALE_P2_4
+    paddd       xmm5, xmm1
+    paddd       xmm3, xmm1
+    psrad       xmm5, DESCALE_P2_4
+    psrad       xmm3, DESCALE_P2_4
+
+    packssdw    xmm4, xmm3              ; xmm4=(00 10 20 30 02 12 22 32)
+    packssdw    xmm7, xmm5              ; xmm7=(01 11 21 31 03 13 23 33)
+
+    movdqa      xmm0, xmm4              ; transpose coefficients(phase 1)
+    punpcklwd   xmm4, xmm7              ; xmm4=(00 01 10 11 20 21 30 31)
+    punpckhwd   xmm0, xmm7              ; xmm0=(02 03 12 13 22 23 32 33)
+
+    movdqa      xmm6, xmm4              ; transpose coefficients(phase 2)
+    punpckldq   xmm4, xmm0              ; xmm4=(00 01 02 03 10 11 12 13)
+    punpckhdq   xmm6, xmm0              ; xmm6=(20 21 22 23 30 31 32 33)
+
+    packsswb    xmm4, xmm6              ; xmm4=(00 01 02 03 10 11 12 13 20 ..)
+    paddb       xmm4, rel PB_CENTERJSAMP
+
+    pshufd      xmm2, xmm4, 0x39        ; xmm2=(10 11 12 13 20 21 22 23 30 ..)
+    pshufd      xmm1, xmm4, 0x4E        ; xmm1=(20 21 22 23 30 31 32 33 00 ..)
+    pshufd      xmm3, xmm4, 0x93        ; xmm3=(30 31 32 33 00 01 02 03 10 ..)
+
+    mov         rdxp, JSAMPROW rdi+0*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+1*SIZEOF_JSAMPROW
+    movd        XMM_DWORD rdx+rax*SIZEOF_JSAMPLE, xmm4
+    movd        XMM_DWORD rsi+rax*SIZEOF_JSAMPLE, xmm2
+    mov         rdxp, JSAMPROW rdi+2*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+3*SIZEOF_JSAMPROW
+    movd        XMM_DWORD rdx+rax*SIZEOF_JSAMPLE, xmm1
+    movd        XMM_DWORD rsi+rax*SIZEOF_JSAMPLE, xmm3
+
+    uncollect_args 4
+    mov         rsp, rbp                ; rsp <- aligned rbp
+    pop         rsp                     ; rsp <- original rbp
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 2x2 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_2x2_sse2(void *dct_table, JCOEFPTR coef_block,
+;                     JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = void *dct_table
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13d = JDIMENSION output_col
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_idct_2x2_sse2)
+
+EXTN(jsimd_idct_2x2_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 4
+    push        rbx
+
+    ; ---- Pass 1: process columns from input.
+
+    mov         rdx, r10                ; quantptr
+    mov         rsi, r11                ; inptr
+
+    ; | input:                  | result:        |
+    ; | 00 01 ** 03 ** 05 ** 07 |                |
+    ; | 10 11 ** 13 ** 15 ** 17 |                |
+    ; | ** ** ** ** ** ** ** ** |                |
+    ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 |
+    ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 |
+    ; | 50 51 ** 53 ** 55 ** 57 |                |
+    ; | ** ** ** ** ** ** ** ** |                |
+    ; | 70 71 ** 73 ** 75 ** 77 |                |
+
+    ; -- Odd part
+
+    movdqa      xmm0, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm1, XMMWORD XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm0, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm1, XMMWORD XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    movdqa      xmm2, XMMWORD XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)
+    movdqa      xmm3, XMMWORD XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm2, XMMWORD XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+    pmullw      xmm3, XMMWORD XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; xmm0=(10 11 ** 13 ** 15 ** 17), xmm1=(30 31 ** 33 ** 35 ** 37)
+    ; xmm2=(50 51 ** 53 ** 55 ** 57), xmm3=(70 71 ** 73 ** 75 ** 77)
+
+    pcmpeqd     xmm7, xmm7
+    pslld       xmm7, WORD_BIT          ; xmm7={0x0000 0xFFFF 0x0000 0xFFFF ..}
+
+    movdqa      xmm4, xmm0              ; xmm4=(10 11 ** 13 ** 15 ** 17)
+    movdqa      xmm5, xmm2              ; xmm5=(50 51 ** 53 ** 55 ** 57)
+    punpcklwd   xmm4, xmm1              ; xmm4=(10 30 11 31 ** ** 13 33)
+    punpcklwd   xmm5, xmm3              ; xmm5=(50 70 51 71 ** ** 53 73)
+    pmaddwd     xmm4, rel PW_F362_MF127
+    pmaddwd     xmm5, rel PW_F085_MF072
+
+    psrld       xmm0, WORD_BIT          ; xmm0=(11 -- 13 -- 15 -- 17 --)
+    pand        xmm1, xmm7              ; xmm1=(-- 31 -- 33 -- 35 -- 37)
+    psrld       xmm2, WORD_BIT          ; xmm2=(51 -- 53 -- 55 -- 57 --)
+    pand        xmm3, xmm7              ; xmm3=(-- 71 -- 73 -- 75 -- 77)
+    por         xmm0, xmm1              ; xmm0=(11 31 13 33 15 35 17 37)
+    por         xmm2, xmm3              ; xmm2=(51 71 53 73 55 75 57 77)
+    pmaddwd     xmm0, rel PW_F362_MF127
+    pmaddwd     xmm2, rel PW_F085_MF072
+
+    paddd       xmm4, xmm5              ; xmm4=tmp0col0 col1 **** col3
+    paddd       xmm0, xmm2              ; xmm0=tmp0col1 col3 col5 col7
+
+    ; -- Even part
+
+    movdqa      xmm6, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)
+    pmullw      xmm6, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)
+
+    ; xmm6=(00 01 ** 03 ** 05 ** 07)
+
+    movdqa      xmm1, xmm6              ; xmm1=(00 01 ** 03 ** 05 ** 07)
+    pslld       xmm6, WORD_BIT          ; xmm6=(-- 00 -- ** -- ** -- **)
+    pand        xmm1, xmm7              ; xmm1=(-- 01 -- 03 -- 05 -- 07)
+    psrad       xmm6, (WORD_BIT-CONST_BITS-2)  ; xmm6=tmp10col0 **** **** ****
+    psrad       xmm1, (WORD_BIT-CONST_BITS-2)  ; xmm1=tmp10col1 col3 col5 col7
+
+    ; -- Final output stage
+
+    movdqa      xmm3, xmm6
+    movdqa      xmm5, xmm1
+    paddd       xmm6, xmm4      ; xmm6=data0col0 **** **** ****=(A0 ** ** **)
+    paddd       xmm1, xmm0      ; xmm1=data0col1 col3 col5 col7=(A1 A3 A5 A7)
+    psubd       xmm3, xmm4      ; xmm3=data1col0 **** **** ****=(B0 ** ** **)
+    psubd       xmm5, xmm0      ; xmm5=data1col1 col3 col5 col7=(B1 B3 B5 B7)
+
+    movdqa      xmm2, rel PD_DESCALE_P1_2  ; xmm2=rel PD_DESCALE_P1_2
+
+    punpckldq   xmm6, xmm3              ; xmm6=(A0 B0 ** **)
+
+    movdqa      xmm7, xmm1
+    punpcklqdq  xmm1, xmm5              ; xmm1=(A1 A3 B1 B3)
+    punpckhqdq  xmm7, xmm5              ; xmm7=(A5 A7 B5 B7)
+
+    paddd       xmm6, xmm2
+    psrad       xmm6, DESCALE_P1_2
+
+    paddd       xmm1, xmm2
+    paddd       xmm7, xmm2
+    psrad       xmm1, DESCALE_P1_2
+    psrad       xmm7, DESCALE_P1_2
+
+    ; -- Prefetch the next coefficient block
+
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32
+    prefetchnta rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32
+
+    ; ---- Pass 2: process rows, store into output array.
+
+    mov         rdi, r12                ; (JSAMPROW *)
+    mov         eax, r13d
+
+    ; | input:| result:|
+    ; | A0 B0 |        |
+    ; | A1 B1 | C0 C1  |
+    ; | A3 B3 | D0 D1  |
+    ; | A5 B5 |        |
+    ; | A7 B7 |        |
+
+    ; -- Odd part
+
+    packssdw    xmm1, xmm1              ; xmm1=(A1 A3 B1 B3 A1 A3 B1 B3)
+    packssdw    xmm7, xmm7              ; xmm7=(A5 A7 B5 B7 A5 A7 B5 B7)
+    pmaddwd     xmm1, rel PW_F362_MF127
+    pmaddwd     xmm7, rel PW_F085_MF072
+
+    paddd       xmm1, xmm7              ; xmm1=tmp0row0 row1 row0 row1
+
+    ; -- Even part
+
+    pslld       xmm6, (CONST_BITS+2)    ; xmm6=tmp10row0 row1 **** ****
+
+    ; -- Final output stage
+
+    movdqa      xmm4, xmm6
+    paddd       xmm6, xmm1     ; xmm6=data0row0 row1 **** ****=(C0 C1 ** **)
+    psubd       xmm4, xmm1     ; xmm4=data1row0 row1 **** ****=(D0 D1 ** **)
+
+    punpckldq   xmm6, xmm4     ; xmm6=(C0 D0 C1 D1)
+
+    paddd       xmm6, rel PD_DESCALE_P2_2
+    psrad       xmm6, DESCALE_P2_2
+
+    packssdw    xmm6, xmm6              ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1)
+    packsswb    xmm6, xmm6              ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1 ..)
+    paddb       xmm6, rel PB_CENTERJSAMP
+
+    pextrw      ebx, xmm6, 0x00         ; ebx=(C0 D0 -- --)
+    pextrw      ecx, xmm6, 0x01         ; ecx=(C1 D1 -- --)
+
+    mov         rdxp, JSAMPROW rdi+0*SIZEOF_JSAMPROW
+    mov         rsip, JSAMPROW rdi+1*SIZEOF_JSAMPROW
+    mov         word rdx+rax*SIZEOF_JSAMPLE, bx
+    mov         word rsi+rax*SIZEOF_JSAMPLE, cx
+
+    pop         rbx
+    uncollect_args 4
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jquantf-sse2.asm Added

@@ -0,0 +1,155 @@
+;
+; jquantf.asm - sample data conversion and quantization (64-bit SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_sse2(JSAMPARRAY sample_data, JDIMENSION start_col,
+;                           FAST_FLOAT *workspace);
+;
+
+; r10 = JSAMPARRAY sample_data
+; r11d = JDIMENSION start_col
+; r12 = FAST_FLOAT *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_convsamp_float_sse2)
+
+EXTN(jsimd_convsamp_float_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 3
+    push        rbx
+
+    pcmpeqw     xmm7, xmm7
+    psllw       xmm7, 7
+    packsswb    xmm7, xmm7              ; xmm7 = PB_CENTERJSAMPLE (0x808080..)
+
+    mov         rsi, r10
+    mov         eax, r11d
+    mov         rdi, r12
+    mov         rcx, DCTSIZE/2
+.convloop:
+    mov         rbxp, JSAMPROW rsi+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rdxp, JSAMPROW rsi+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        xmm0, XMM_MMWORD rbx+rax*SIZEOF_JSAMPLE
+    movq        xmm1, XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE
+
+    psubb       xmm0, xmm7              ; xmm0=(01234567)
+    psubb       xmm1, xmm7              ; xmm1=(89ABCDEF)
+
+    punpcklbw   xmm0, xmm0              ; xmm0=(*0*1*2*3*4*5*6*7)
+    punpcklbw   xmm1, xmm1              ; xmm1=(*8*9*A*B*C*D*E*F)
+
+    punpcklwd   xmm2, xmm0              ; xmm2=(***0***1***2***3)
+    punpckhwd   xmm0, xmm0              ; xmm0=(***4***5***6***7)
+    punpcklwd   xmm3, xmm1              ; xmm3=(***8***9***A***B)
+    punpckhwd   xmm1, xmm1              ; xmm1=(***C***D***E***F)
+
+    psrad       xmm2, (DWORD_BIT-BYTE_BIT)  ; xmm2=(0123)
+    psrad       xmm0, (DWORD_BIT-BYTE_BIT)  ; xmm0=(4567)
+    cvtdq2ps    xmm2, xmm2                  ; xmm2=(0123)
+    cvtdq2ps    xmm0, xmm0                  ; xmm0=(4567)
+    psrad       xmm3, (DWORD_BIT-BYTE_BIT)  ; xmm3=(89AB)
+    psrad       xmm1, (DWORD_BIT-BYTE_BIT)  ; xmm1=(CDEF)
+    cvtdq2ps    xmm3, xmm3                  ; xmm3=(89AB)
+    cvtdq2ps    xmm1, xmm1                  ; xmm1=(CDEF)
+
+    movaps      XMMWORD XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT), xmm2
+    movaps      XMMWORD XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT), xmm0
+    movaps      XMMWORD XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT), xmm3
+    movaps      XMMWORD XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT), xmm1
+
+    add         rsi, byte 2*SIZEOF_JSAMPROW
+    add         rdi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+    dec         rcx
+    jnz         short .convloop
+
+    pop         rbx
+    uncollect_args 3
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_sse2(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                           FAST_FLOAT *workspace);
+;
+
+; r10 = JCOEFPTR coef_block
+; r11 = FAST_FLOAT *divisors
+; r12 = FAST_FLOAT *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_quantize_float_sse2)
+
+EXTN(jsimd_quantize_float_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 3
+
+    mov         rsi, r12
+    mov         rdx, r11
+    mov         rdi, r10
+    mov         rax, DCTSIZE2/16
+.quantloop:
+    movaps      xmm0, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_FAST_FLOAT)
+    movaps      xmm1, XMMWORD XMMBLOCK(0,1,rsi,SIZEOF_FAST_FLOAT)
+    mulps       xmm0, XMMWORD XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)
+    mulps       xmm1, XMMWORD XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)
+    movaps      xmm2, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_FAST_FLOAT)
+    movaps      xmm3, XMMWORD XMMBLOCK(1,1,rsi,SIZEOF_FAST_FLOAT)
+    mulps       xmm2, XMMWORD XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)
+    mulps       xmm3, XMMWORD XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)
+
+    cvtps2dq    xmm0, xmm0
+    cvtps2dq    xmm1, xmm1
+    cvtps2dq    xmm2, xmm2
+    cvtps2dq    xmm3, xmm3
+
+    packssdw    xmm0, xmm1
+    packssdw    xmm2, xmm3
+
+    movdqa      XMMWORD XMMBLOCK(0,0,rdi,SIZEOF_JCOEF), xmm0
+    movdqa      XMMWORD XMMBLOCK(1,0,rdi,SIZEOF_JCOEF), xmm2
+
+    add         rsi, byte 16*SIZEOF_FAST_FLOAT
+    add         rdx, byte 16*SIZEOF_FAST_FLOAT
+    add         rdi, byte 16*SIZEOF_JCOEF
+    dec         rax
+    jnz         short .quantloop
+
+    uncollect_args 3
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jquanti-avx2.asm Added

@@ -0,0 +1,163 @@
+;
+; jquanti.asm - sample data conversion and quantization (64-bit AVX2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, 2018, D. R. Commander.
+; Copyright (C) 2016, Matthieu Darbois.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_avx2(JSAMPARRAY sample_data, JDIMENSION start_col,
+;                     DCTELEM *workspace);
+;
+
+; r10 = JSAMPARRAY sample_data
+; r11d = JDIMENSION start_col
+; r12 = DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_convsamp_avx2)
+
+EXTN(jsimd_convsamp_avx2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 3
+
+    mov         eax, r11d
+
+    mov         rsip, JSAMPROW r10+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rdip, JSAMPROW r10+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        xmm0, XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE
+    pinsrq      xmm0, XMM_MMWORD rdi+rax*SIZEOF_JSAMPLE, 1
+
+    mov         rsip, JSAMPROW r10+2*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rdip, JSAMPROW r10+3*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        xmm1, XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE
+    pinsrq      xmm1, XMM_MMWORD rdi+rax*SIZEOF_JSAMPLE, 1
+
+    mov         rsip, JSAMPROW r10+4*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rdip, JSAMPROW r10+5*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        xmm2, XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE
+    pinsrq      xmm2, XMM_MMWORD rdi+rax*SIZEOF_JSAMPLE, 1
+
+    mov         rsip, JSAMPROW r10+6*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rdip, JSAMPROW r10+7*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    movq        xmm3, XMM_MMWORD rsi+rax*SIZEOF_JSAMPLE
+    pinsrq      xmm3, XMM_MMWORD rdi+rax*SIZEOF_JSAMPLE, 1
+
+    vpmovzxbw   ymm0, xmm0              ; ymm0=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+    vpmovzxbw   ymm1, xmm1              ; ymm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+    vpmovzxbw   ymm2, xmm2              ; ymm2=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+    vpmovzxbw   ymm3, xmm3              ; ymm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+    vpcmpeqw    ymm7, ymm7, ymm7
+    vpsllw      ymm7, ymm7, 7           ; ymm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    vpaddw      ymm0, ymm0, ymm7
+    vpaddw      ymm1, ymm1, ymm7
+    vpaddw      ymm2, ymm2, ymm7
+    vpaddw      ymm3, ymm3, ymm7
+
+    vmovdqu     YMMWORD YMMBLOCK(0,0,r12,SIZEOF_DCTELEM), ymm0
+    vmovdqu     YMMWORD YMMBLOCK(2,0,r12,SIZEOF_DCTELEM), ymm1
+    vmovdqu     YMMWORD YMMBLOCK(4,0,r12,SIZEOF_DCTELEM), ymm2
+    vmovdqu     YMMWORD YMMBLOCK(6,0,r12,SIZEOF_DCTELEM), ymm3
+
+    vzeroupper
+    uncollect_args 3
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_avx2(JCOEFPTR coef_block, DCTELEM *divisors,
+;                     DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m, n, b) \
+  YMMBLOCK(DCTSIZE * 0 + (m), (n), (b), SIZEOF_DCTELEM)
+%define CORRECTION(m, n, b) \
+  YMMBLOCK(DCTSIZE * 1 + (m), (n), (b), SIZEOF_DCTELEM)
+%define SCALE(m, n, b) \
+  YMMBLOCK(DCTSIZE * 2 + (m), (n), (b), SIZEOF_DCTELEM)
+
+; r10 = JCOEFPTR coef_block
+; r11 = DCTELEM *divisors
+; r12 = DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_quantize_avx2)
+
+EXTN(jsimd_quantize_avx2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 3
+
+    vmovdqu     ymm4, YMMBLOCK(0,0,r12,SIZEOF_DCTELEM)
+    vmovdqu     ymm5, YMMBLOCK(2,0,r12,SIZEOF_DCTELEM)
+    vmovdqu     ymm6, YMMBLOCK(4,0,r12,SIZEOF_DCTELEM)
+    vmovdqu     ymm7, YMMBLOCK(6,0,r12,SIZEOF_DCTELEM)
+    vpabsw      ymm0, ymm4
+    vpabsw      ymm1, ymm5
+    vpabsw      ymm2, ymm6
+    vpabsw      ymm3, ymm7
+
+    vpaddw      ymm0, YMMWORD CORRECTION(0,0,r11)  ; correction + roundfactor
+    vpaddw      ymm1, YMMWORD CORRECTION(2,0,r11)
+    vpaddw      ymm2, YMMWORD CORRECTION(4,0,r11)
+    vpaddw      ymm3, YMMWORD CORRECTION(6,0,r11)
+    vpmulhuw    ymm0, YMMWORD RECIPROCAL(0,0,r11)  ; reciprocal
+    vpmulhuw    ymm1, YMMWORD RECIPROCAL(2,0,r11)
+    vpmulhuw    ymm2, YMMWORD RECIPROCAL(4,0,r11)
+    vpmulhuw    ymm3, YMMWORD RECIPROCAL(6,0,r11)
+    vpmulhuw    ymm0, YMMWORD SCALE(0,0,r11)       ; scale
+    vpmulhuw    ymm1, YMMWORD SCALE(2,0,r11)
+    vpmulhuw    ymm2, YMMWORD SCALE(4,0,r11)
+    vpmulhuw    ymm3, YMMWORD SCALE(6,0,r11)
+
+    vpsignw     ymm0, ymm0, ymm4
+    vpsignw     ymm1, ymm1, ymm5
+    vpsignw     ymm2, ymm2, ymm6
+    vpsignw     ymm3, ymm3, ymm7
+
+    vmovdqu     YMMBLOCK(0,0,r10,SIZEOF_DCTELEM), ymm0
+    vmovdqu     YMMBLOCK(2,0,r10,SIZEOF_DCTELEM), ymm1
+    vmovdqu     YMMBLOCK(4,0,r10,SIZEOF_DCTELEM), ymm2
+    vmovdqu     YMMBLOCK(6,0,r10,SIZEOF_DCTELEM), ymm3
+
+    vzeroupper
+    uncollect_args 3
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jquanti-sse2.asm Added

@@ -0,0 +1,188 @@
+;
+; jquanti.asm - sample data conversion and quantization (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2009, 2016, D. R. Commander.
+; Copyright (C) 2018, Matthias Räncker.
+;
+; Based on the x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_sse2(JSAMPARRAY sample_data, JDIMENSION start_col,
+;                     DCTELEM *workspace);
+;
+
+; r10 = JSAMPARRAY sample_data
+; r11d = JDIMENSION start_col
+; r12 = DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_convsamp_sse2)
+
+EXTN(jsimd_convsamp_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 3
+    push        rbx
+
+    pxor        xmm6, xmm6              ; xmm6=(all 0's)
+    pcmpeqw     xmm7, xmm7
+    psllw       xmm7, 7                 ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+    mov         rsi, r10
+    mov         eax, r11d
+    mov         rdi, r12
+    mov         rcx, DCTSIZE/4
+.convloop:
+    mov         rbxp, JSAMPROW rsi+0*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rdxp, JSAMPROW rsi+1*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        xmm0, XMM_MMWORD rbx+rax*SIZEOF_JSAMPLE  ; xmm0=(01234567)
+    movq        xmm1, XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE  ; xmm1=(89ABCDEF)
+
+    mov         rbxp, JSAMPROW rsi+2*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+    mov         rdxp, JSAMPROW rsi+3*SIZEOF_JSAMPROW  ; (JSAMPLE *)
+
+    movq        xmm2, XMM_MMWORD rbx+rax*SIZEOF_JSAMPLE  ; xmm2=(GHIJKLMN)
+    movq        xmm3, XMM_MMWORD rdx+rax*SIZEOF_JSAMPLE  ; xmm3=(OPQRSTUV)
+
+    punpcklbw   xmm0, xmm6              ; xmm0=(01234567)
+    punpcklbw   xmm1, xmm6              ; xmm1=(89ABCDEF)
+    paddw       xmm0, xmm7
+    paddw       xmm1, xmm7
+    punpcklbw   xmm2, xmm6              ; xmm2=(GHIJKLMN)
+    punpcklbw   xmm3, xmm6              ; xmm3=(OPQRSTUV)
+    paddw       xmm2, xmm7
+    paddw       xmm3, xmm7
+
+    movdqa      XMMWORD XMMBLOCK(0,0,rdi,SIZEOF_DCTELEM), xmm0
+    movdqa      XMMWORD XMMBLOCK(1,0,rdi,SIZEOF_DCTELEM), xmm1
+    movdqa      XMMWORD XMMBLOCK(2,0,rdi,SIZEOF_DCTELEM), xmm2
+    movdqa      XMMWORD XMMBLOCK(3,0,rdi,SIZEOF_DCTELEM), xmm3
+
+    add         rsi, byte 4*SIZEOF_JSAMPROW
+    add         rdi, byte 4*DCTSIZE*SIZEOF_DCTELEM
+    dec         rcx
+    jnz         short .convloop
+
+    pop         rbx
+    uncollect_args 3
+    pop         rbp
+    ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_sse2(JCOEFPTR coef_block, DCTELEM *divisors,
+;                     DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m, n, b) \
+  XMMBLOCK(DCTSIZE * 0 + (m), (n), (b), SIZEOF_DCTELEM)
+%define CORRECTION(m, n, b) \
+  XMMBLOCK(DCTSIZE * 1 + (m), (n), (b), SIZEOF_DCTELEM)
+%define SCALE(m, n, b) \
+  XMMBLOCK(DCTSIZE * 2 + (m), (n), (b), SIZEOF_DCTELEM)
+
+; r10 = JCOEFPTR coef_block
+; r11 = DCTELEM *divisors
+; r12 = DCTELEM *workspace
+
+    align       32
+    GLOBAL_FUNCTION(jsimd_quantize_sse2)
+
+EXTN(jsimd_quantize_sse2):
+    push        rbp
+    mov         rax, rsp
+    mov         rbp, rsp
+    collect_args 3
+
+    mov         rsi, r12
+    mov         rdx, r11
+    mov         rdi, r10
+    mov         rax, DCTSIZE2/32
+.quantloop:
+    movdqa      xmm4, XMMWORD XMMBLOCK(0,0,rsi,SIZEOF_DCTELEM)
+    movdqa      xmm5, XMMWORD XMMBLOCK(1,0,rsi,SIZEOF_DCTELEM)
+    movdqa      xmm6, XMMWORD XMMBLOCK(2,0,rsi,SIZEOF_DCTELEM)
+    movdqa      xmm7, XMMWORD XMMBLOCK(3,0,rsi,SIZEOF_DCTELEM)
+    movdqa      xmm0, xmm4
+    movdqa      xmm1, xmm5
+    movdqa      xmm2, xmm6
+    movdqa      xmm3, xmm7
+    psraw       xmm4, (WORD_BIT-1)
+    psraw       xmm5, (WORD_BIT-1)
+    psraw       xmm6, (WORD_BIT-1)
+    psraw       xmm7, (WORD_BIT-1)
+    pxor        xmm0, xmm4
+    pxor        xmm1, xmm5
+    pxor        xmm2, xmm6
+    pxor        xmm3, xmm7
+    psubw       xmm0, xmm4              ; if (xmm0 < 0) xmm0 = -xmm0;
+    psubw       xmm1, xmm5              ; if (xmm1 < 0) xmm1 = -xmm1;
+    psubw       xmm2, xmm6              ; if (xmm2 < 0) xmm2 = -xmm2;
+    psubw       xmm3, xmm7              ; if (xmm3 < 0) xmm3 = -xmm3;
+
+    paddw       xmm0, XMMWORD CORRECTION(0,0,rdx)  ; correction + roundfactor
+    paddw       xmm1, XMMWORD CORRECTION(1,0,rdx)
+    paddw       xmm2, XMMWORD CORRECTION(2,0,rdx)
+    paddw       xmm3, XMMWORD CORRECTION(3,0,rdx)
+    pmulhuw     xmm0, XMMWORD RECIPROCAL(0,0,rdx)  ; reciprocal
+    pmulhuw     xmm1, XMMWORD RECIPROCAL(1,0,rdx)
+    pmulhuw     xmm2, XMMWORD RECIPROCAL(2,0,rdx)
+    pmulhuw     xmm3, XMMWORD RECIPROCAL(3,0,rdx)
+    pmulhuw     xmm0, XMMWORD SCALE(0,0,rdx)       ; scale
+    pmulhuw     xmm1, XMMWORD SCALE(1,0,rdx)
+    pmulhuw     xmm2, XMMWORD SCALE(2,0,rdx)
+    pmulhuw     xmm3, XMMWORD SCALE(3,0,rdx)
+
+    pxor        xmm0, xmm4
+    pxor        xmm1, xmm5
+    pxor        xmm2, xmm6
+    pxor        xmm3, xmm7
+    psubw       xmm0, xmm4
+    psubw       xmm1, xmm5
+    psubw       xmm2, xmm6
+    psubw       xmm3, xmm7
+    movdqa      XMMWORD XMMBLOCK(0,0,rdi,SIZEOF_DCTELEM), xmm0
+    movdqa      XMMWORD XMMBLOCK(1,0,rdi,SIZEOF_DCTELEM), xmm1
+    movdqa      XMMWORD XMMBLOCK(2,0,rdi,SIZEOF_DCTELEM), xmm2
+    movdqa      XMMWORD XMMBLOCK(3,0,rdi,SIZEOF_DCTELEM), xmm3
+
+    add         rsi, byte 32*SIZEOF_DCTELEM
+    add         rdx, byte 32*SIZEOF_DCTELEM
+    add         rdi, byte 32*SIZEOF_JCOEF
+    dec         rax
+    jnz         near .quantloop
+
+    uncollect_args 3
+    pop         rbp
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jsimd.c Added

@@ -0,0 +1,1072 @@
+/*
+ * jsimd_x86_64.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2009-2011, 2014, 2016, 2018, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 64-bit x86 architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../../jinclude.h"
+#include "../../jpeglib.h"
+#include "../../jsimd.h"
+#include "../../jdct.h"
+#include "../../jsimddct.h"
+#include "../jsimd.h"
+#include "jconfigint.h"
+
+/*
+ * In the PIC cases, we have no guarantee that constants will keep
+ * their alignment. This macro allows us to verify it at runtime.
+ */
+#define IS_ALIGNED(ptr, order)  (((size_t)ptr & ((1 << order) - 1)) == 0)
+
+#define IS_ALIGNED_SSE(ptr)  (IS_ALIGNED(ptr, 4)) /* 16 byte alignment */
+#define IS_ALIGNED_AVX(ptr)  (IS_ALIGNED(ptr, 5)) /* 32 byte alignment */
+
+static unsigned int simd_support = (unsigned int)(~0);
+static unsigned int simd_huffman = 1;
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd(void)
+{
+#ifndef NO_GETENV
+  char *env = NULL;
+#endif
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = jpeg_simd_cpu_support();
+
+#ifndef NO_GETENV
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCESSE2");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_SSE2;
+  env = getenv("JSIMD_FORCEAVX2");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_AVX2;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+#endif
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_rgb_ycc_convert_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_rgb_gray_convert_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_ycc_rgb_convert_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565(void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                      JSAMPIMAGE output_buf, JDIMENSION output_row,
+                      int num_rows)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*sse2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_extrgb_ycc_convert_avx2;
+    sse2fct = jsimd_extrgb_ycc_convert_sse2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_extrgbx_ycc_convert_avx2;
+    sse2fct = jsimd_extrgbx_ycc_convert_sse2;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_extbgr_ycc_convert_avx2;
+    sse2fct = jsimd_extbgr_ycc_convert_sse2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_extbgrx_ycc_convert_avx2;
+    sse2fct = jsimd_extbgrx_ycc_convert_sse2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_extxbgr_ycc_convert_avx2;
+    sse2fct = jsimd_extxbgr_ycc_convert_sse2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_extxrgb_ycc_convert_avx2;
+    sse2fct = jsimd_extxrgb_ycc_convert_sse2;
+    break;
+  default:
+    avx2fct = jsimd_rgb_ycc_convert_avx2;
+    sse2fct = jsimd_rgb_ycc_convert_sse2;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else
+    sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                       JSAMPIMAGE output_buf, JDIMENSION output_row,
+                       int num_rows)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*sse2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch (cinfo->in_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_extrgb_gray_convert_avx2;
+    sse2fct = jsimd_extrgb_gray_convert_sse2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_extrgbx_gray_convert_avx2;
+    sse2fct = jsimd_extrgbx_gray_convert_sse2;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_extbgr_gray_convert_avx2;
+    sse2fct = jsimd_extbgr_gray_convert_sse2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_extbgrx_gray_convert_avx2;
+    sse2fct = jsimd_extbgrx_gray_convert_sse2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_extxbgr_gray_convert_avx2;
+    sse2fct = jsimd_extxbgr_gray_convert_sse2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_extxrgb_gray_convert_avx2;
+    sse2fct = jsimd_extxrgb_gray_convert_sse2;
+    break;
+  default:
+    avx2fct = jsimd_rgb_gray_convert_avx2;
+    sse2fct = jsimd_rgb_gray_convert_sse2;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else
+    sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION input_row, JSAMPARRAY output_buf,
+                      int num_rows)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+  void (*sse2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_ycc_extrgb_convert_avx2;
+    sse2fct = jsimd_ycc_extrgb_convert_sse2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_ycc_extrgbx_convert_avx2;
+    sse2fct = jsimd_ycc_extrgbx_convert_sse2;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_ycc_extbgr_convert_avx2;
+    sse2fct = jsimd_ycc_extbgr_convert_sse2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_ycc_extbgrx_convert_avx2;
+    sse2fct = jsimd_ycc_extbgrx_convert_sse2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_ycc_extxbgr_convert_avx2;
+    sse2fct = jsimd_ycc_extxbgr_convert_sse2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_ycc_extxrgb_convert_avx2;
+    sse2fct = jsimd_ycc_extxrgb_convert_sse2;
+    break;
+  default:
+    avx2fct = jsimd_ycc_rgb_convert_avx2;
+    sse2fct = jsimd_ycc_rgb_convert_sse2;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+  else
+    sse2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v2_downsample_avx2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else
+    jsimd_h2v2_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v1_downsample_avx2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else
+    jsimd_h2v1_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v2_upsample_avx2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else
+    jsimd_h2v2_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v1_upsample_avx2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else
+    jsimd_h2v1_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_fancy_upsample_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_fancy_upsample_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v2_fancy_upsample_avx2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else
+    jsimd_h2v2_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_h2v1_fancy_upsample_avx2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else
+    jsimd_h2v1_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_merged_upsample_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) &&
+      IS_ALIGNED_AVX(jconst_merged_upsample_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*sse2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_h2v2_extrgb_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extrgb_merged_upsample_sse2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_h2v2_extrgbx_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extrgbx_merged_upsample_sse2;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_h2v2_extbgr_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extbgr_merged_upsample_sse2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_h2v2_extbgrx_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extbgrx_merged_upsample_sse2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_h2v2_extxbgr_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extxbgr_merged_upsample_sse2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_h2v2_extxrgb_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_extxrgb_merged_upsample_sse2;
+    break;
+  default:
+    avx2fct = jsimd_h2v2_merged_upsample_avx2;
+    sse2fct = jsimd_h2v2_merged_upsample_sse2;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else
+    sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                           JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf)
+{
+  void (*avx2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*sse2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch (cinfo->out_color_space) {
+  case JCS_EXT_RGB:
+    avx2fct = jsimd_h2v1_extrgb_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extrgb_merged_upsample_sse2;
+    break;
+  case JCS_EXT_RGBX:
+  case JCS_EXT_RGBA:
+    avx2fct = jsimd_h2v1_extrgbx_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extrgbx_merged_upsample_sse2;
+    break;
+  case JCS_EXT_BGR:
+    avx2fct = jsimd_h2v1_extbgr_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extbgr_merged_upsample_sse2;
+    break;
+  case JCS_EXT_BGRX:
+  case JCS_EXT_BGRA:
+    avx2fct = jsimd_h2v1_extbgrx_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extbgrx_merged_upsample_sse2;
+    break;
+  case JCS_EXT_XBGR:
+  case JCS_EXT_ABGR:
+    avx2fct = jsimd_h2v1_extxbgr_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extxbgr_merged_upsample_sse2;
+    break;
+  case JCS_EXT_XRGB:
+  case JCS_EXT_ARGB:
+    avx2fct = jsimd_h2v1_extxrgb_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_extxrgb_merged_upsample_sse2;
+    break;
+  default:
+    avx2fct = jsimd_h2v1_merged_upsample_avx2;
+    sse2fct = jsimd_h2v1_merged_upsample_sse2;
+    break;
+  }
+
+  if (simd_support & JSIMD_AVX2)
+    avx2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else
+    sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
+               DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_convsamp_avx2(sample_data, start_col, workspace);
+  else
+    jsimd_convsamp_sse2(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
+                     FAST_FLOAT *workspace)
+{
+  jsimd_convsamp_float_sse2(sample_data, start_col, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) && IS_ALIGNED_AVX(jconst_fdct_islow_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_ifast_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow(DCTELEM *data)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_fdct_islow_avx2(data);
+  else
+    jsimd_fdct_islow_sse2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast(DCTELEM *data)
+{
+  jsimd_fdct_ifast_sse2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float(FAST_FLOAT *data)
+{
+  jsimd_fdct_float_sse(data);
+}
+
+GLOBAL(int)
+jsimd_can_quantize(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_AVX2)
+    return 1;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_quantize_avx2(coef_block, divisors, workspace);
+  else
+    jsimd_quantize_sse2(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                     FAST_FLOAT *workspace)
+{
+  jsimd_quantize_float_sse2(coef_block, divisors, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  jsimd_idct_2x2_sse2(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+               JDIMENSION output_col)
+{
+  jsimd_idct_4x4_sse2(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_AVX2) && IS_ALIGNED_AVX(jconst_idct_islow_avx2))
+    return 1;
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast(void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+  if (sizeof(FLOAT_MULT_TYPE) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_AVX2)
+    jsimd_idct_islow_avx2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else
+    jsimd_idct_islow_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_ifast_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                 JDIMENSION output_col)
+{
+  jsimd_idct_float_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && simd_huffman &&
+      IS_ALIGNED_SSE(jconst_huff_encode_one_block))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET *)
+jsimd_huff_encode_one_block(void *state, JOCTET *buffer, JCOEFPTR block,
+                            int last_dc_val, c_derived_tbl *dctbl,
+                            c_derived_tbl *actbl)
+{
+  return jsimd_huff_encode_one_block_sse2(state, buffer, block, last_dc_val,
+                                          dctbl, actbl);
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_first_prepare(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_encode_mcu_AC_first_prepare(const JCOEF *block,
+                                  const int *jpeg_natural_order_start, int Sl,
+                                  int Al, JCOEF *values, size_t *zerobits)
+{
+  jsimd_encode_mcu_AC_first_prepare_sse2(block, jpeg_natural_order_start,
+                                         Sl, Al, values, zerobits);
+}
+
+GLOBAL(int)
+jsimd_can_encode_mcu_AC_refine_prepare(void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block,
+                                   const int *jpeg_natural_order_start, int Sl,
+                                   int Al, JCOEF *absvalues, size_t *bits)
+{
+  return jsimd_encode_mcu_AC_refine_prepare_sse2(block,
+                                                 jpeg_natural_order_start,
+                                                 Sl, Al, absvalues, bits);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/simd/x86_64/jsimdcpu.asm Added

@@ -0,0 +1,86 @@
+;
+; jsimdcpu.asm - SIMD instruction support check
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright (C) 2016, D. R. Commander.
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+    SECTION     SEG_TEXT
+    BITS        64
+;
+; Check if the CPU supports SIMD instructions
+;
+; GLOBAL(unsigned int)
+; jpeg_simd_cpu_support(void)
+;
+
+    align       32
+    GLOBAL_FUNCTION(jpeg_simd_cpu_support)
+
+EXTN(jpeg_simd_cpu_support):
+    push        rbx
+    push        rdi
+
+    xor         rdi, rdi                ; simd support flag
+
+    ; Assume that all x86-64 processors support SSE & SSE2 instructions
+    or          rdi, JSIMD_SSE2
+    or          rdi, JSIMD_SSE
+
+    ; Check whether CPUID leaf 07H is supported
+    ; (leaf 07H is used to check for AVX2 instruction support)
+    mov         rax, 0
+    cpuid
+    cmp         rax, 7
+    jl          short .return           ; Maximum leaf < 07H
+
+    ; Check for AVX2 instruction support
+    mov         rax, 7
+    xor         rcx, rcx
+    cpuid
+    mov         rax, rbx                ; rax = Extended feature flags
+
+    test        rax, 1<<5               ; bit5:AVX2
+    jz          short .return
+
+    ; Check for AVX2 O/S support
+    mov         rax, 1
+    xor         rcx, rcx
+    cpuid
+    test        rcx, 1<<27
+    jz          short .return           ; O/S does not support XSAVE
+    test        rcx, 1<<28
+    jz          short .return           ; CPU does not support AVX2
+
+    xor         rcx, rcx
+    xgetbv
+    and         rax, 6
+    cmp         rax, 6                  ; O/S does not manage XMM/YMM state
+                                        ; using XSAVE
+    jnz         short .return
+
+    or          rdi, JSIMD_AVX2
+
+.return:
+    mov         rax, rdi
+
+    pop         rdi
+    pop         rbx
+    ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+    align       32

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/structure.txt Added

@@ -0,0 +1,900 @@
+IJG JPEG LIBRARY:  SYSTEM ARCHITECTURE
+
+This file was part of the Independent JPEG Group's software:
+Copyright (C) 1991-2012, Thomas G. Lane, Guido Vollbeding.
+It was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo.
+For conditions of distribution and use, see the accompanying README.ijg file.
+
+
+This file provides an overview of the architecture of the IJG JPEG software;
+that is, the functions of the various modules in the system and the interfaces
+between modules.  For more precise details about any data structure or calling
+convention, see the include files and comments in the source code.
+
+We assume that the reader is already somewhat familiar with the JPEG standard.
+The README.ijg file includes references for learning about JPEG.  The file
+libjpeg.txt describes the library from the viewpoint of an application
+programmer using the library; it's best to read that file before this one.
+Also, the file coderules.txt describes the coding style conventions we use.
+
+In this document, JPEG-specific terminology follows the JPEG standard:
+  A "component" means a color channel, e.g., Red or Luminance.
+  A "sample" is a single component value (i.e., one number in the image data).
+  A "coefficient" is a frequency coefficient (a DCT transform output number).
+  A "block" is an 8x8 group of samples or coefficients.
+  An "MCU" (minimum coded unit) is an interleaved set of blocks of size
+        determined by the sampling factors, or a single block in a
+        noninterleaved scan.
+We do not use the terms "pixel" and "sample" interchangeably.  When we say
+pixel, we mean an element of the full-size image, while a sample is an element
+of the downsampled image.  Thus the number of samples may vary across
+components while the number of pixels does not.  (This terminology is not used
+rigorously throughout the code, but it is used in places where confusion would
+otherwise result.)
+
+
+*** System features ***
+
+The IJG distribution contains two parts:
+  * A subroutine library for JPEG compression and decompression.
+  * cjpeg/djpeg, two sample applications that use the library to transform
+    JFIF JPEG files to and from several other image formats.
+cjpeg/djpeg are of no great intellectual complexity: they merely add a simple
+command-line user interface and I/O routines for several uncompressed image
+formats.  This document concentrates on the library itself.
+
+We desire the library to be capable of supporting all JPEG baseline, extended
+sequential, and progressive DCT processes.  Hierarchical processes are not
+supported.
+
+The library does not support the lossless (spatial) JPEG process.  Lossless
+JPEG shares little or no code with lossy JPEG, and would normally be used
+without the extensive pre- and post-processing provided by this library.
+We feel that lossless JPEG is better handled by a separate library.
+
+Within these limits, any set of compression parameters allowed by the JPEG
+spec should be readable for decompression.  (We can be more restrictive about
+what formats we can generate.)  Although the system design allows for all
+parameter values, some uncommon settings are not yet implemented and may
+never be; nonintegral sampling ratios are the prime example.  Furthermore,
+we treat 8-bit vs. 12-bit data precision as a compile-time switch, not a
+run-time option, because most machines can store 8-bit pixels much more
+compactly than 12-bit.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, libtiff uses this
+library to implement JPEG compression within the TIFF file format.)
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  This code can be omitted if not needed.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+
+*** System overview ***
+
+The compressor and decompressor are each divided into two main sections:
+the JPEG compressor or decompressor proper, and the preprocessing or
+postprocessing functions.  The interface between these two sections is the
+image data that Rec. ITU-T T.81 | ISO/IEC 10918-1 regards as its input or
+output: this data is in the colorspace to be used for compression, and it is
+downsampled to the sampling factors to be used.  The preprocessing and
+postprocessing steps are responsible for converting a normal image
+representation to or from this form.  (Those few applications that want to deal
+with YCbCr downsampled data can skip the preprocessing or postprocessing step.)
+
+Looking more closely, the compressor library contains the following main
+elements:
+
+  Preprocessing:
+    * Color space conversion (e.g., RGB to YCbCr).
+    * Edge expansion and downsampling.  Optionally, this step can do simple
+      smoothing --- this is often helpful for low-quality source data.
+  JPEG proper:
+    * MCU assembly, DCT, quantization.
+    * Entropy coding (sequential or progressive, Huffman or arithmetic).
+
+In addition to these modules we need overall control, marker generation,
+and support code (memory management & error handling).  There is also a
+module responsible for physically writing the output data --- typically
+this is just an interface to fwrite(), but some applications may need to
+do something else with the data.
+
+The decompressor library contains the following main elements:
+
+  JPEG proper:
+    * Entropy decoding (sequential or progressive, Huffman or arithmetic).
+    * Dequantization, inverse DCT, MCU disassembly.
+  Postprocessing:
+    * Upsampling.  Optionally, this step may be able to do more general
+      rescaling of the image.
+    * Color space conversion (e.g., YCbCr to RGB).  This step may also
+      provide gamma adjustment  currently it does not .
+    * Optional color quantization (e.g., reduction to 256 colors).
+    * Optional color precision reduction (e.g., 24-bit to 15-bit color).
+      This feature is not currently implemented.
+
+We also need overall control, marker parsing, and a data source module.
+The support code (memory management & error handling) can be shared with
+the compression half of the library.
+
+There may be several implementations of each of these elements, particularly
+in the decompressor, where a wide range of speed/quality tradeoffs is very
+useful.  It must be understood that some of the best speedups involve
+merging adjacent steps in the pipeline.  For example, upsampling, color space
+conversion, and color quantization might all be done at once when using a
+low-quality ordered-dither technique.  The system architecture is designed to
+allow such merging where appropriate.
+
+
+Note: it is convenient to regard edge expansion (padding to block boundaries)
+as a preprocessing/postprocessing function, even though
+Rec. ITU-T T.81 | ISO/IEC 10918-1 includes it in compression/decompression.  We
+do this because downsampling/upsampling can be simplified a little if they work
+on padded data: it's not necessary to have special cases at the right and
+bottom edges.  Therefore the interface buffer is always an integral number of
+blocks wide and high, and we expect compression preprocessing to pad the source
+data properly.  Padding will occur only to the next block (8-sample) boundary.
+In an interleaved-scan situation, additional dummy blocks may be used to fill
+out MCUs, but the MCU assembly and disassembly logic will create or discard
+these blocks internally.  (This is advantageous for speed reasons, since we
+avoid DCTing the dummy blocks.  It also permits a small reduction in file size,
+because the compressor can choose dummy block contents so as to minimize their
+size in compressed form.  Finally, it makes the interface buffer specification
+independent of whether the file is actually interleaved or not.)  Applications
+that wish to deal directly with the downsampled data must provide similar
+buffering and padding for odd-sized images.
+
+
+*** Poor man's object-oriented programming ***
+
+It should be clear by now that we have a lot of quasi-independent processing
+steps, many of which have several possible behaviors.  To avoid cluttering the
+code with lots of switch statements, we use a simple form of object-style
+programming to separate out the different possibilities.
+
+For example, two different color quantization algorithms could be implemented
+as two separate modules that present the same external interface; at runtime,
+the calling code will access the proper module indirectly through an "object".
+
+We can get the limited features we need while staying within portable C.
+The basic tool is a function pointer.  An "object" is just a struct
+containing one or more function pointer fields, each of which corresponds to
+a method name in real object-oriented languages.  During initialization we
+fill in the function pointers with references to whichever module we have
+determined we need to use in this run.  Then invocation of the module is done
+by indirecting through a function pointer; on most machines this is no more
+expensive than a switch statement, which would be the only other way of
+making the required run-time choice.  The really significant benefit, of
+course, is keeping the source code clean and well structured.
+
+We can also arrange to have private storage that varies between different
+implementations of the same kind of object.  We do this by making all the
+module-specific object structs be separately allocated entities, which will
+be accessed via pointers in the master compression or decompression struct.
+The "public" fields or methods for a given kind of object are specified by
+a commonly known struct.  But a module's initialization code can allocate
+a larger struct that contains the common struct as its first member, plus
+additional private fields.  With appropriate pointer casting, the module's
+internal functions can access these private fields.  (For a simple example,
+see jdatadst.c, which implements the external interface specified by struct
+jpeg_destination_mgr, but adds extra fields.)
+
+(Of course this would all be a lot easier if we were using C++, but we are
+not yet prepared to assume that everyone has a C++ compiler.)
+
+An important benefit of this scheme is that it is easy to provide multiple
+versions of any method, each tuned to a particular case.  While a lot of
+precalculation might be done to select an optimal implementation of a method,
+the cost per invocation is constant.  For example, the upsampling step might
+have a "generic" method, plus one or more "hardwired" methods for the most
+popular sampling factors; the hardwired methods would be faster because they'd
+use straight-line code instead of for-loops.  The cost to determine which
+method to use is paid only once, at startup, and the selection criteria are
+hidden from the callers of the method.
+
+This plan differs a little bit from usual object-oriented structures, in that
+only one instance of each object class will exist during execution.  The
+reason for having the class structure is that on different runs we may create
+different instances (choose to execute different modules).  You can think of
+the term "method" as denoting the common interface presented by a particular
+set of interchangeable functions, and "object" as denoting a group of related
+methods, or the total shared interface behavior of a group of modules.
+
+
+*** Overall control structure ***
+
+We previously mentioned the need for overall control logic in the compression
+and decompression libraries.  In IJG implementations prior to v5, overall
+control was mostly provided by "pipeline control" modules, which proved to be
+large, unwieldy, and hard to understand.  To improve the situation, the
+control logic has been subdivided into multiple modules.  The control modules
+consist of:
+
+1. Master control for module selection and initialization.  This has two
+responsibilities:
+
+   1A.  Startup initialization at the beginning of image processing.
+        The individual processing modules to be used in this run are selected
+        and given initialization calls.
+
+   1B.  Per-pass control.  This determines how many passes will be performed
+        and calls each active processing module to configure itself
+        appropriately at the beginning of each pass.  End-of-pass processing,
+        where necessary, is also invoked from the master control module.
+
+   Method selection is partially distributed, in that a particular processing
+   module may contain several possible implementations of a particular method,
+   which it will select among when given its initialization call.  The master
+   control code need only be concerned with decisions that affect more than
+   one module.
+
+2. Data buffering control.  A separate control module exists for each
+   inter-processing-step data buffer.  This module is responsible for
+   invoking the processing steps that write or read that data buffer.
+
+Each buffer controller sees the world as follows:
+
+input data => processing step A => buffer => processing step B => output data
+                      |              |               |
+              ------------------ controller ------------------
+
+The controller knows the dataflow requirements of steps A and B: how much data
+they want to accept in one chunk and how much they output in one chunk.  Its
+function is to manage its buffer and call A and B at the proper times.
+
+A data buffer control module may itself be viewed as a processing step by a
+higher-level control module; thus the control modules form a binary tree with
+elementary processing steps at the leaves of the tree.
+
+The control modules are objects.  A considerable amount of flexibility can
+be had by replacing implementations of a control module.  For example:
+* Merging of adjacent steps in the pipeline is done by replacing a control
+  module and its pair of processing-step modules with a single processing-
+  step module.  (Hence the possible merges are determined by the tree of
+  control modules.)
+* In some processing modes, a given interstep buffer need only be a "strip"
+  buffer large enough to accommodate the desired data chunk sizes.  In other
+  modes, a full-image buffer is needed and several passes are required.
+  The control module determines which kind of buffer is used and manipulates
+  virtual array buffers as needed.  One or both processing steps may be
+  unaware of the multi-pass behavior.
+
+In theory, we might be able to make all of the data buffer controllers
+interchangeable and provide just one set of implementations for all.  In
+practice, each one contains considerable special-case processing for its
+particular job.  The buffer controller concept should be regarded as an
+overall system structuring principle, not as a complete description of the
+task performed by any one controller.
+
+
+*** Compression object structure ***
+
+Here is a sketch of the logical structure of the JPEG compression library:
+
+                                                 |-- Colorspace conversion
+                  |-- Preprocessing controller --|
+                  |                              |-- Downsampling
+Main controller --|
+                  |                            |-- Forward DCT, quantize
+                  |-- Coefficient controller --|
+                                               |-- Entropy encoding
+
+This sketch also describes the flow of control (subroutine calls) during
+typical image data processing.  Each of the components shown in the diagram is
+an "object" which may have several different implementations available.  One
+or more source code files contain the actual implementation(s) of each object.
+
+The objects shown above are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the preprocessed input data.  This controller invokes preprocessing to
+  fill the subsampled-data buffer, and JPEG compression to empty it.  There is
+  usually no need for a full-image buffer here; a strip buffer is adequate.
+
+* Preprocessing controller: buffer controller for the downsampling input data
+  buffer, which lies between colorspace conversion and downsampling.  Note
+  that a unified conversion/downsampling module would probably replace this
+  controller entirely.
+
+* Colorspace conversion: converts application image data into the desired
+  JPEG color space; also changes the data from pixel-interleaved layout to
+  separate component planes.  Processes one pixel row at a time.
+
+* Downsampling: performs reduction of chroma components as required.
+  Optionally may perform pixel-level smoothing as well.  Processes a "row
+  group" at a time, where a row group is defined as Vmax pixel rows of each
+  component before downsampling, and Vk sample rows afterwards (remember Vk
+  differs across components).  Some downsampling or smoothing algorithms may
+  require context rows above and below the current row group; the
+  preprocessing controller is responsible for supplying these rows via proper
+  buffering.  The downsampler is responsible for edge expansion at the right
+  edge (i.e., extending each sample row to a multiple of 8 samples); but the
+  preprocessing controller is responsible for vertical edge expansion (i.e.,
+  duplicating the bottom sample row as needed to make a multiple of 8 rows).
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU assembly, including insertion of dummy DCT
+  blocks when needed at the right or bottom edge.  When performing
+  Huffman-code optimization or emitting a multiscan JPEG file, this
+  controller is responsible for buffering the full image.  The equivalent of
+  one fully interleaved MCU row of subsampled data is processed per call,
+  even when the JPEG file is noninterleaved.
+
+* Forward DCT and quantization: Perform DCT, quantize, and emit coefficients.
+  Works on one or more DCT blocks at a time.  (Note: the coefficients are now
+  emitted in normal array order, which the entropy encoder is expected to
+  convert to zigzag order as necessary.  Prior versions of the IJG code did
+  the conversion to zigzag order within the quantization step.)
+
+* Entropy encoding: Perform Huffman or arithmetic entropy coding and emit the
+  coded data to the data destination module.  Works on one MCU per call.
+  For progressive JPEG, the same DCT blocks are fed to the entropy coder
+  during each pass, and the coder must emit the appropriate subset of
+  coefficients.
+
+In addition to the above objects, the compression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.
+
+* Marker writing: generates JPEG markers (except for RSTn, which is emitted
+  by the entropy encoder when needed).
+
+* Data destination manager: writes the output JPEG datastream to its final
+  destination (e.g., a file).  The destination manager supplied with the
+  library knows how to write to a stdio stream or to a memory buffer;
+  for other behaviors, the surrounding application may provide its own
+  destination manager.
+
+* Memory manager: allocates and releases memory, controls virtual arrays
+  (with backing store management, where required).
+
+* Error handler: performs formatting and output of error and trace messages;
+  determines handling of nonfatal errors.  The surrounding application may
+  override some or all of this object's methods to change error handling.
+
+* Progress monitor: supports output of "percent-done" progress reports.
+  This object represents an optional callback to the surrounding application:
+  if wanted, it must be supplied by the application.
+
+The error handler, destination manager, and progress monitor objects are
+defined as separate objects in order to simplify application-specific
+customization of the JPEG library.  A surrounding application may override
+individual methods or supply its own all-new implementation of one of these
+objects.  The object interfaces for these objects are therefore treated as
+part of the application interface of the library, whereas the other objects
+are internal to the library.
+
+The error handler and memory manager are shared by JPEG compression and
+decompression; the progress monitor, if used, may be shared as well.
+
+
+*** Decompression object structure ***
+
+Here is a sketch of the logical structure of the JPEG decompression library:
+
+                                               |-- Entropy decoding
+                  |-- Coefficient controller --|
+                  |                            |-- Dequantize, Inverse DCT
+Main controller --|
+                  |                               |-- Upsampling
+                  |-- Postprocessing controller --|   |-- Colorspace conversion
+                                                  |-- Color quantization
+                                                  |-- Color precision reduction
+
+As before, this diagram also represents typical control flow.  The objects
+shown are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the output of JPEG decompression proper.  This controller's primary
+  task is to feed the postprocessing procedure.  Some upsampling algorithms
+  may require context rows above and below the current row group; when this
+  is true, the main controller is responsible for managing its buffer so as
+  to make context rows available.  In the current design, the main buffer is
+  always a strip buffer; a full-image buffer is never required.
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU disassembly, including deletion of any dummy
+  DCT blocks at the right or bottom edge.  When reading a multiscan JPEG
+  file, this controller is responsible for buffering the full image.
+  (Buffering DCT coefficients, rather than samples, is necessary to support
+  progressive JPEG.)  The equivalent of one fully interleaved MCU row of
+  subsampled data is processed per call, even when the source JPEG file is
+  noninterleaved.
+
+* Entropy decoding: Read coded data from the data source module and perform
+  Huffman or arithmetic entropy decoding.  Works on one MCU per call.
+  For progressive JPEG decoding, the coefficient controller supplies the prior
+  coefficients of each MCU (initially all zeroes), which the entropy decoder
+  modifies in each scan.
+
+* Dequantization and inverse DCT: like it says.  Note that the coefficients
+  buffered by the coefficient controller have NOT been dequantized; we
+  merge dequantization and inverse DCT into a single step for speed reasons.
+  When scaled-down output is asked for, simplified DCT algorithms may be used
+  that emit fewer samples per DCT block, not the full 8x8.  Works on one DCT
+  block at a time.
+
+* Postprocessing controller: buffer controller for the color quantization
+  input buffer, when quantization is in use.  (Without quantization, this
+  controller just calls the upsampler.)  For two-pass quantization, this
+  controller is responsible for buffering the full-image data.
+
+* Upsampling: restores chroma components to full size.  (May support more
+  general output rescaling, too.  Note that if undersized DCT outputs have
+  been emitted by the DCT module, this module must adjust so that properly
+  sized outputs are created.)  Works on one row group at a time.  This module
+  also calls the color conversion module, so its top level is effectively a
+  buffer controller for the upsampling->color conversion buffer.  However, in
+  all but the highest-quality operating modes, upsampling and color
+  conversion are likely to be merged into a single step.
+
+* Colorspace conversion: convert from JPEG color space to output color space,
+  and change data layout from separate component planes to pixel-interleaved.
+  Works on one pixel row at a time.
+
+* Color quantization: reduce the data to colormapped form, using either an
+  externally specified colormap or an internally generated one.  This module
+  is not used for full-color output.  Works on one pixel row at a time; may
+  require two passes to generate a color map.  Note that the output will
+  always be a single component representing colormap indexes.  In the current
+  design, the output values are JSAMPLEs, so an 8-bit compilation cannot
+  quantize to more than 256 colors.  This is unlikely to be a problem in
+  practice.
+
+* Color reduction: this module handles color precision reduction, e.g.,
+  generating 15-bit color (5 bits/primary) from JPEG's 24-bit output.
+  Not quite clear yet how this should be handled... should we merge it with
+  colorspace conversion???
+
+Note that some high-speed operating modes might condense the entire
+postprocessing sequence to a single module (upsample, color convert, and
+quantize in one step).
+
+In addition to the above objects, the decompression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.  This is subdivided into
+  input and output control: jdinput.c controls only input-side processing,
+  while jdmaster.c handles overall initialization and output-side control.
+
+* Marker reading: decodes JPEG markers (except for RSTn).
+
+* Data source manager: supplies the input JPEG datastream.  The source
+  manager supplied with the library knows how to read from a stdio stream
+  or from a memory buffer;  for other behaviors, the surrounding application
+  may provide its own source manager.
+
+* Memory manager: same as for compression library.
+
+* Error handler: same as for compression library.
+
+* Progress monitor: same as for compression library.
+
+As with compression, the data source manager, error handler, and progress
+monitor are candidates for replacement by a surrounding application.
+
+
+*** Decompression input and output separation ***
+
+To support efficient incremental display of progressive JPEG files, the
+decompressor is divided into two sections that can run independently:
+
+1. Data input includes marker parsing, entropy decoding, and input into the
+   coefficient controller's DCT coefficient buffer.  Note that this
+   processing is relatively cheap and fast.
+
+2. Data output reads from the DCT coefficient buffer and performs the IDCT
+   and all postprocessing steps.
+
+For a progressive JPEG file, the data input processing is allowed to get
+arbitrarily far ahead of the data output processing.  (This occurs only
+if the application calls jpeg_consume_input(); otherwise input and output
+run in lockstep, since the input section is called only when the output
+section needs more data.)  In this way the application can avoid making
+extra display passes when data is arriving faster than the display pass
+can run.  Furthermore, it is possible to abort an output pass without
+losing anything, since the coefficient buffer is read-only as far as the
+output section is concerned.  See libjpeg.txt for more detail.
+
+A full-image coefficient array is only created if the JPEG file has multiple
+scans (or if the application specifies buffered-image mode anyway).  When
+reading a single-scan file, the coefficient controller normally creates only
+a one-MCU buffer, so input and output processing must run in lockstep in this
+case.  jpeg_consume_input() is effectively a no-op in this situation.
+
+The main impact of dividing the decompressor in this fashion is that we must
+be very careful with shared variables in the cinfo data structure.  Each
+variable that can change during the course of decompression must be
+classified as belonging to data input or data output, and each section must
+look only at its own variables.  For example, the data output section may not
+depend on any of the variables that describe the current scan in the JPEG
+file, because these may change as the data input section advances into a new
+scan.
+
+The progress monitor is (somewhat arbitrarily) defined to treat input of the
+file as one pass when buffered-image mode is not used, and to ignore data
+input work completely when buffered-image mode is used.  Note that the
+library has no reliable way to predict the number of passes when dealing
+with a progressive JPEG file, nor can it predict the number of output passes
+in buffered-image mode.  So the work estimate is inherently bogus anyway.
+
+No comparable division is currently made in the compression library, because
+there isn't any real need for it.
+
+
+*** Data formats ***
+
+Arrays of pixel sample values use the following data structure:
+
+    typedef something JSAMPLE;          a pixel component value, 0..MAXJSAMPLE
+    typedef JSAMPLE *JSAMPROW;          ptr to a row of samples
+    typedef JSAMPROW *JSAMPARRAY;       ptr to a list of rows
+    typedef JSAMPARRAY *JSAMPIMAGE;     ptr to a list of color-component arrays
+
+The basic element type JSAMPLE will be one of unsigned char or short.  Short
+will be used if samples wider than 8 bits are to be supported (this is a
+compile-time option).  Otherwise, unsigned char is used.
+
+With these conventions, JSAMPLE values can be assumed to be >= 0.  This helps
+simplify correct rounding during downsampling, etc.  The JPEG standard's
+specification that sample values run from -128..127 is accommodated by
+subtracting 128 from the sample value in the DCT step.  Similarly, during
+decompression the output of the IDCT step will be immediately shifted back to
+0..255.  (NB: different values are required when 12-bit samples are in use.
+The code is written in terms of MAXJSAMPLE and CENTERJSAMPLE, which will be
+defined as 255 and 128 respectively in an 8-bit implementation, and as 4095
+and 2048 in a 12-bit implementation.)
+
+We use a pointer per row, rather than a two-dimensional JSAMPLE array.  This
+choice costs only a small amount of memory and has several benefits:
+* Code using the data structure doesn't need to know the allocated width of
+  the rows.  This simplifies edge expansion/compression, since we can work
+  in an array that's wider than the logical picture width.
+* Indexing doesn't require multiplication; this is a performance win on many
+  machines.
+* Arrays with more than 64K total elements can be supported even on machines
+  where malloc() cannot allocate chunks larger than 64K.
+* The rows forming a component array may be allocated at different times
+  without extra copying.  This trick allows some speedups in smoothing steps
+  that need access to the previous and next rows.
+
+Note that each color component is stored in a separate array; we don't use the
+traditional layout in which the components of a pixel are stored together.
+This simplifies coding of modules that work on each component independently,
+because they don't need to know how many components there are.  Furthermore,
+we can read or write each component to a temporary file independently, which
+is helpful when dealing with noninterleaved JPEG files.
+
+In general, a specific sample value is accessed by code such as
+        imagecolorcomponentrowcol
+where col is measured from the image left edge, but row is measured from the
+first sample row currently in memory.  Either of the first two indexings can
+be precomputed by copying the relevant pointer.
+
+
+Since most image-processing applications prefer to work on images in which
+the components of a pixel are stored together, the data passed to or from the
+surrounding application uses the traditional convention: a single pixel is
+represented by N consecutive JSAMPLE values, and an image row is an array of
+(# of color components)*(image width) JSAMPLEs.  One or more rows of data can
+be represented by a pointer of type JSAMPARRAY in this scheme.  This scheme is
+converted to component-wise storage inside the JPEG library.  (Applications
+that want to skip JPEG preprocessing or postprocessing will have to contend
+with component-wise storage.)
+
+
+Arrays of DCT-coefficient values use the following data structure:
+
+    typedef short JCOEF;                a 16-bit signed integer
+    typedef JCOEF JBLOCKDCTSIZE2;     an 8x8 block of coefficients
+    typedef JBLOCK *JBLOCKROW;          ptr to one horizontal row of 8x8 blocks
+    typedef JBLOCKROW *JBLOCKARRAY;     ptr to a list of such rows
+    typedef JBLOCKARRAY *JBLOCKIMAGE;   ptr to a list of color component arrays
+
+The underlying type is at least a 16-bit signed integer; while "short" is big
+enough on all machines of interest, on some machines it is preferable to use
+"int" for speed reasons, despite the storage cost.  Coefficients are grouped
+into 8x8 blocks (but we always use #defines DCTSIZE and DCTSIZE2 rather than
+"8" and "64").
+
+The contents of a coefficient block may be in either "natural" or zigzagged
+order, and may be true values or divided by the quantization coefficients,
+depending on where the block is in the processing pipeline.  In the current
+library, coefficient blocks are kept in natural order everywhere; the entropy
+codecs zigzag or dezigzag the data as it is written or read.  The blocks
+contain quantized coefficients everywhere outside the DCT/IDCT subsystems.
+(This latter decision may need to be revisited to support variable
+quantization a la JPEG Part 3.)
+
+Notice that the allocation unit is now a row of 8x8 blocks, corresponding to
+eight rows of samples.  Otherwise the structure is much the same as for
+samples, and for the same reasons.
+
+
+*** Suspendable processing ***
+
+In some applications it is desirable to use the JPEG library as an
+incremental, memory-to-memory filter.  In this situation the data source or
+destination may be a limited-size buffer, and we can't rely on being able to
+empty or refill the buffer at arbitrary times.  Instead the application would
+like to have control return from the library at buffer overflow/underrun, and
+then resume compression or decompression at a later time.
+
+This scenario is supported for simple cases.  (For anything more complex, we
+recommend that the application "bite the bullet" and develop real multitasking
+capability.)  The libjpeg.txt file goes into more detail about the usage and
+limitations of this capability; here we address the implications for library
+structure.
+
+The essence of the problem is that the entropy codec (coder or decoder) must
+be prepared to stop at arbitrary times.  In turn, the controllers that call
+the entropy codec must be able to stop before having produced or consumed all
+the data that they normally would handle in one call.  That part is reasonably
+straightforward: we make the controller call interfaces include "progress
+counters" which indicate the number of data chunks successfully processed, and
+we require callers to test the counter rather than just assume all of the data
+was processed.
+
+Rather than trying to restart at an arbitrary point, the current Huffman
+codecs are designed to restart at the beginning of the current MCU after a
+suspension due to buffer overflow/underrun.  At the start of each call, the
+codec's internal state is loaded from permanent storage (in the JPEG object
+structures) into local variables.  On successful completion of the MCU, the
+permanent state is updated.  (This copying is not very expensive, and may even
+lead to *improved* performance if the local variables can be registerized.)
+If a suspension occurs, the codec simply returns without updating the state,
+thus effectively reverting to the start of the MCU.  Note that this implies
+leaving some data unprocessed in the source/destination buffer (ie, the
+compressed partial MCU).  The data source/destination module interfaces are
+specified so as to make this possible.  This also implies that the data buffer
+must be large enough to hold a worst-case compressed MCU; a couple thousand
+bytes should be enough.
+
+In a successive-approximation AC refinement scan, the progressive Huffman
+decoder has to be able to undo assignments of newly nonzero coefficients if it
+suspends before the MCU is complete, since decoding requires distinguishing
+previously-zero and previously-nonzero coefficients.  This is a bit tedious
+but probably won't have much effect on performance.  Other variants of Huffman
+decoding need not worry about this, since they will just store the same values
+again if forced to repeat the MCU.
+
+This approach would probably not work for an arithmetic codec, since its
+modifiable state is quite large and couldn't be copied cheaply.  Instead it
+would have to suspend and resume exactly at the point of the buffer end.
+
+The JPEG marker reader is designed to cope with suspension at an arbitrary
+point.  It does so by backing up to the start of the marker parameter segment,
+so the data buffer must be big enough to hold the largest marker of interest.
+Again, a couple KB should be adequate.  (A special "skip" convention is used
+to bypass COM and APPn markers, so these can be larger than the buffer size
+without causing problems; otherwise a 64K buffer would be needed in the worst
+case.)
+
+The JPEG marker writer currently does *not* cope with suspension.
+We feel that this is not necessary; it is much easier simply to require
+the application to ensure there is enough buffer space before starting.  (An
+empty 2K buffer is more than sufficient for the header markers; and ensuring
+there are a dozen or two bytes available before calling jpeg_finish_compress()
+will suffice for the trailer.)  This would not work for writing multi-scan
+JPEG files, but we simply do not intend to support that capability with
+suspension.
+
+
+*** Memory manager services ***
+
+The JPEG library's memory manager controls allocation and deallocation of
+memory, and it manages large "virtual" data arrays on machines where the
+operating system does not provide virtual memory.  Note that the same
+memory manager serves both compression and decompression operations.
+
+In all cases, allocated objects are tied to a particular compression or
+decompression master record, and they will be released when that master
+record is destroyed.
+
+The memory manager does not provide explicit deallocation of objects.
+Instead, objects are created in "pools" of free storage, and a whole pool
+can be freed at once.  This approach helps prevent storage-leak bugs, and
+it speeds up operations whenever malloc/free are slow (as they often are).
+The pools can be regarded as lifetime identifiers for objects.  Two
+pools/lifetimes are defined:
+  * JPOOL_PERMANENT     lasts until master record is destroyed
+  * JPOOL_IMAGE         lasts until done with image (JPEG datastream)
+Permanent lifetime is used for parameters and tables that should be carried
+across from one datastream to another; this includes all application-visible
+parameters.  Image lifetime is used for everything else.  (A third lifetime,
+JPOOL_PASS = one processing pass, was originally planned.  However it was
+dropped as not being worthwhile.  The actual usage patterns are such that the
+peak memory usage would be about the same anyway; and having per-pass storage
+substantially complicates the virtual memory allocation rules --- see below.)
+
+The memory manager deals with three kinds of object:
+1. "Small" objects.  Typically these require no more than 10K-20K total.
+2. "Large" objects.  These may require tens to hundreds of K depending on
+   image size.  Semantically they behave the same as small objects, but we
+   distinguish them because pool allocation heuristics may differ for large and
+   small objects (historically, large objects were also referenced by far
+   pointers on MS-DOS machines.)  Note that individual "large" objects cannot
+   exceed the size allowed by type size_t, which may be 64K or less on some
+   machines.
+3. "Virtual" objects.  These are large 2-D arrays of JSAMPLEs or JBLOCKs
+   (typically large enough for the entire image being processed).  The
+   memory manager provides stripwise access to these arrays.  On machines
+   without virtual memory, the rest of the array may be swapped out to a
+   temporary file.
+
+(Note: JSAMPARRAY and JBLOCKARRAY data structures are a combination of large
+objects for the data proper and small objects for the row pointers.  For
+convenience and speed, the memory manager provides single routines to create
+these structures.  Similarly, virtual arrays include a small control block
+and a JSAMPARRAY or JBLOCKARRAY working buffer, all created with one call.)
+
+In the present implementation, virtual arrays are only permitted to have image
+lifespan.  (Permanent lifespan would not be reasonable, and pass lifespan is
+not very useful since a virtual array's raison d'etre is to store data for
+multiple passes through the image.)  We also expect that only "small" objects
+will be given permanent lifespan, though this restriction is not required by
+the memory manager.
+
+In a non-virtual-memory machine, some performance benefit can be gained by
+making the in-memory buffers for virtual arrays be as large as possible.
+(For small images, the buffers might fit entirely in memory, so blind
+swapping would be very wasteful.)  The memory manager will adjust the height
+of the buffers to fit within a prespecified maximum memory usage.  In order
+to do this in a reasonably optimal fashion, the manager needs to allocate all
+of the virtual arrays at once.  Therefore, there isn't a one-step allocation
+routine for virtual arrays; instead, there is a "request" routine that simply
+allocates the control block, and a "realize" routine (called just once) that
+determines space allocation and creates all of the actual buffers.  The
+realize routine must allow for space occupied by non-virtual large objects.
+(We don't bother to factor in the space needed for small objects, on the
+grounds that it isn't worth the trouble.)
+
+To support all this, we establish the following protocol for doing business
+with the memory manager:
+  1. Modules must request virtual arrays (which may have only image lifespan)
+     during the initial setup phase, i.e., in their jinit_xxx routines.
+  2. All "large" objects (including JSAMPARRAYs and JBLOCKARRAYs) must also be
+     allocated during initial setup.
+  3. realize_virt_arrays will be called at the completion of initial setup.
+     The above conventions ensure that sufficient information is available
+     for it to choose a good size for virtual array buffers.
+Small objects of any lifespan may be allocated at any time.  We expect that
+the total space used for small objects will be small enough to be negligible
+in the realize_virt_arrays computation.
+
+In a virtual-memory machine, we simply pretend that the available space is
+infinite, thus causing realize_virt_arrays to decide that it can allocate all
+the virtual arrays as full-size in-memory buffers.  The overhead of the
+virtual-array access protocol is very small when no swapping occurs.
+
+A virtual array can be specified to be "pre-zeroed"; when this flag is set,
+never-yet-written sections of the array are set to zero before being made
+available to the caller.  If this flag is not set, never-written sections
+of the array contain garbage.  (This feature exists primarily because the
+equivalent logic would otherwise be needed in jdcoefct.c for progressive
+JPEG mode; we may as well make it available for possible other uses.)
+
+The first write pass on a virtual array is required to occur in top-to-bottom
+order; read passes, as well as any write passes after the first one, may
+access the array in any order.  This restriction exists partly to simplify
+the virtual array control logic, and partly because some file systems may not
+support seeking beyond the current end-of-file in a temporary file.  The main
+implication of this restriction is that rearrangement of rows (such as
+converting top-to-bottom data order to bottom-to-top) must be handled while
+reading data out of the virtual array, not while putting it in.
+
+
+*** Memory manager internal structure ***
+
+To isolate system dependencies as much as possible, we have broken the
+memory manager into two parts.  There is a reasonably system-independent
+"front end" (jmemmgr.c) and a "back end" that contains only the code
+likely to change across systems.  All of the memory management methods
+outlined above are implemented by the front end.  The back end provides
+the following routines for use by the front end (none of these routines
+are known to the rest of the JPEG code):
+
+jpeg_mem_init, jpeg_mem_term    system-dependent initialization/shutdown
+
+jpeg_get_small, jpeg_free_small interface to malloc and free library routines
+                                (or their equivalents)
+
+jpeg_get_large, jpeg_free_large historically was used to interface with
+                                FAR malloc/free on MS-DOS machines;  now the
+                                same as jpeg_get_small/jpeg_free_small
+
+jpeg_mem_available              estimate available memory
+
+jpeg_open_backing_store         create a backing-store object
+
+read_backing_store,             manipulate a backing-store object
+write_backing_store,
+close_backing_store
+
+On some systems there will be more than one type of backing-store object.
+jpeg_open_backing_store is responsible for choosing how to implement a given
+object.  The read/write/close routines are method pointers in the structure
+that describes a given object; this lets them be different for different object
+types.
+
+It may be necessary to ensure that backing store objects are explicitly
+released upon abnormal program termination.  To support this, we will expect
+the main program or surrounding application to arrange to call self_destruct
+(typically via jpeg_destroy) upon abnormal termination.  This may require a
+SIGINT signal handler or equivalent.  We don't want to have the back end module
+install its own signal handler, because that would pre-empt the surrounding
+application's ability to control signal handling.
+
+The IJG distribution includes several memory manager back end implementations.
+Usually the same back end should be suitable for all applications on a given
+system, but it is possible for an application to supply its own back end at
+need.
+
+
+*** Implications of DNL marker ***
+
+Some JPEG files may use a DNL marker to postpone definition of the image
+height (this would be useful for a fax-like scanner's output, for instance).
+In these files the SOF marker claims the image height is 0, and you only
+find out the true image height at the end of the first scan.
+
+We could read these files as follows:
+1. Upon seeing zero image height, replace it by 65535 (the maximum allowed).
+2. When the DNL is found, update the image height in the global image
+   descriptor.
+This implies that control modules must avoid making copies of the image
+height, and must re-test for termination after each MCU row.  This would
+be easy enough to do.
+
+In cases where image-size data structures are allocated, this approach will
+result in very inefficient use of virtual memory or much-larger-than-necessary
+temporary files.  This seems acceptable for something that probably won't be a
+mainstream usage.  People might have to forgo use of memory-hogging options
+(such as two-pass color quantization or noninterleaved JPEG files) if they
+want efficient conversion of such files.  (One could improve efficiency by
+demanding a user-supplied upper bound for the height, less than 65536; in most
+cases it could be much less.)
+
+The standard also permits the SOF marker to overestimate the image height,
+with a DNL to give the true, smaller height at the end of the first scan.
+This would solve the space problems if the overestimate wasn't too great.
+However, it implies that you don't even know whether DNL will be used.
+
+This leads to a couple of very serious objections:
+1. Testing for a DNL marker must occur in the inner loop of the decompressor's
+   Huffman decoder; this implies a speed penalty whether the feature is used
+   or not.
+2. There is no way to hide the last-minute change in image height from an
+   application using the decoder.  Thus *every* application using the IJG
+   library would suffer a complexity penalty whether it cared about DNL or
+   not.
+We currently do not support DNL because of these problems.
+
+A different approach is to insist that DNL-using files be preprocessed by a
+separate program that reads ahead to the DNL, then goes back and fixes the SOF
+marker.  This is a much simpler solution and is probably far more efficient.
+Even if one wants piped input, buffering the first scan of the JPEG file needs
+a lot smaller temp file than is implied by the maximum-height method.  For
+this approach we'd simply treat DNL as a no-op in the decompressor (at most,
+check that it matches the SOF image height).
+
+We will not worry about making the compressor capable of outputting DNL.
+Something similar to the first scheme above could be applied if anyone ever
+wants to make that work.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/nightshot_iso_100.bmp Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/nightshot_iso_100.txt Added

@@ -0,0 +1,25 @@
+libjpeg-turbo note:  This image was extracted from the 8-bit nightshot_iso_100
+image.  The original can be downloaded at the link below.
+
+The New Image Compression Test Set - Jan 2008
+http://www.imagecompression.info/test_images
+
+The images historically used for compression research (lena, barbra, pepper etc...) have outlived their useful life and its about time they become a part of history only. They are too small, come from data sources too old and are available in only 8-bit precision.
+
+These images have been carefully selected to aid in image compression algorithm research and evaluation. These are photographic images chosen to come from a wide variety of sources and each one picked to stress different aspects of algorithms. Images are available in 8-bit, 16-bit and 16-bit linear variations, RGB and gray.
+
+Images are available without any prohibitive copyright restrictions.
+
+These images are (c) there respective owners. You are granted full redistribution and publication rights on these images provided:
+
+1. The origin of the pictures must not be misrepresented; you must not claim that you took the original pictures. If you use, publish or redistribute them, an acknowledgment would be appreciated but is not required.
+2. Altered versions must be plainly marked as such, and must not be misinterpreted as being the originals.
+3. No payment is required for distribution this material, it must be available freely under the conditions stated here. That is, it is prohibited to sell the material.
+4. This notice may not be removed or altered from any distribution.
+
+Acknowledgments: A lot of people contributed a lot of time and effort in making this test set possible. Thanks to everyone who voiced their opinion in any of the discussions online. Thanks to Axel Becker, Thomas Richter and Niels Fröhling for their extensive help in picking images, running all the various tests etc... Thanks to Pete Fraser, Tony Story, Wayne J. Cosshall, David Coffin, Bruce Lindbloom and raw.fotosite.pl for the images which make up this set.
+
+Sachin Garg India
+sachingarg@c10n.info
+
+www.sachingarg.com | www.c10n.info | www.imagecompression.info

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/test.scan Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/test1.icc Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/test1.icc.txt Added

@@ -0,0 +1,20 @@
+Little CMS
+Copyright (c) 1998-2011 Marti Maria Saguer
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal in
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/test2.icc Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/test2.icc.txt Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/testimgari.jpg Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/testimgint.jpg Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/testorig.jpg Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/testorig.ppm Added

@@ -0,0 +1,4 @@
+P6
+227 149
+255
+0/-0/-10.21/51.51.62/62/83/83/:2/:2/:3-:3-:3-:3-:2/:2/91.80-80-91.:2/:2/80-80-80-80-80-80-80-80-6.+6.+6.+5-*5-*5-*4,)4,)4,)4,)4,)4,)4,)4,)4,)2-)/*$/,%0-&0-&1.'2/(30)30)63,63,74-85.96/96/:70:7.A:0B<0D>2F@4IA4JB5KC6KC6NE6MD5OC3NB2OC3OC3PD4RE5R?1Y?2b@4nB5}E6H8G9£F7°H;¸F;¿F;ÅF=ÇG>ËH@ËH@ÐEBçFLíCLìEMëEIîCIïBDò?Cô=Aø;A÷:@ô:?ð<?é?Bâ?@×?<ËA7»=/µ@.µ@.´?-´?-³@-²?-¯@-?.ªA.¦A-¢B, A-@+A+A,>-?/?/>.>,>,<+<+>->-=*=*>+>+?,@/?6>5=4?3©B3³D3¼D4¿D4¹?0¶B3¬F:H;G;oA2U9+C3&=52:659548437116005//5//72/72/72/61.61-61-50,41,//-.0-//-//-0/-0/-2.-2.-5,-4+,4*+3)*7(+=.1E69P:<jAE|HJNOOQWªdnªot{£&#x7f;®º~½tz®`aTKvQEiSJgPH_MH^TQbfdo|}£©¢¨ {|lfgWYXFQNEURUQbUQb0/-0/-10.10.40-51.62/62/83/83/:2/:2/:3-:3-:3-:3-:2/91.80-80-80-80-91.:2/80-80-80-80-80-80-80-80-6.+6.+6.+5-*5-*4,)4,)4,)5-*5-*5-*5-*5-*5-*5-*3.*1,(0-&0-&1.'2/(30)41*41*63,63,74-85.85.96/:70:7.@9/B<0C=1E?3H@3IA4JB5JB5LC4LC4MA3MA1MA1NB2OC3QD4P>0U?1^A3jC4xD6F5E5C3§C4¯A4µ@6¼B7ÀD:ÄE<ÅF=ÍC@áEIçBIèCIêDHíDGðCEó@Cö?Cø;A÷:@ô:?ð<?é?@à@@Õ@<Ê@6¹>/µ@.´?-´?-´?-²?,°?-¯@-?.ªA.¦A-¢B,@,@+@*A,>-?/?/>.>,=+<+<+>-=,=*=*=*>+?,@1A6?6>5¡?4«A3µD4½C4¿D5»A2·C6¬F:H=G:l@3S9*B4)>63:65:6584382271160060072/72/72/61.61-61-50,41,//-.0-//-//-0/-0/-2.-2.-4..5,-5+,3)*5)+<-0C47N8:d=>vDCJIMNTV¤aj¥l}r{¢®&#x7f;¹{»ou©RIvOCiOFePH`PH_RN_Yfnot ~&#x7f;qjkLVSJXSZVRaXQa/.,0/-0/-10.40/40/51.51.72.72.72.72.92,92,92,92,91.80.80.7/-7/-80.80.91/80.80.80.80.80.80.80.80.6.,6.,5-+5-+5-+4,*4,*4,*6.,6.,6.,6.,6.,6.,6.,4/+2-)1.)2/*30+30+41,52-52-63.63.74/74/850961961:70?8.@:.B<0D>2G?4H@5H@3H@3J@4I@1K?1K?1K?1L@2MA1NB2MA3QA2YB4dC4qC4|C2B2A0<-¡;,§;.¯=2µ@6ºD:¿F=ÅD>ÙCEá@FãBGèCGêDFðCEôADø?Dú;@ù:?õ;@ð=@è@@ÝB>Ñ@;Æ@5·=.³@-³@-³@-²?-°?-¯>,@,ª?-§@-¥@.¡@-A,@+@*@+>-?/>.>.>.=-=+=+>,>,<+<+=,>-?.?0A6?5>3¤?3¯A4¹C5¿D5ÁC5ÀD8¹G<®I=J=G;h@4Q:,B5,?74=77<66;5594183072/72/62/62/62/51.52-52-41,21,/1.-2./1./1.00.00.10.3/.5//4..5,-4*+4*+9-/>24I5697l?:}FAIDOM`fvnw}}¹u~·gl¤UUMEvLAkMAeOFcQHcNI_NK\esty|xzlfhZZMVSLZUZT`S`.-+/.,/.,0/-3/.40/40-51.61-61-61-61-81+81+81+80-80.7/-6.,6.,6.,6.,7/-80.80.80.80.80.80.80.80.80.5-+5-+5-+5-+4,*4,*3+)3+)6.,6.,6.,6.,6.,6.,6.,4/,30+30+30+41,41,52-52-52-52-63.63.74/85096196196/>7-?9-A;/B<0E=2E=2F>3F>1G=1G=1H<.I=/I=/J>0L@0JA0LE5NE4VE5^D3iD2sB1~A/?-9)9'9*¤=.¬@3³E8¸H<ÁF>ÒDCÚACÞBEâDEèDEìBCó@C÷?Aú;@ù:?ö<Aî>@åA@ÚB=Í@9Â@3¶>.°@,°@,°@,¯>,®?,®?,¬?+©@-¦?,£@- ?,@+@*@)?*>,>.>.>.=-=-<*<*=+=+=+=+<+>,>-?/A6@6>2¦@4²B6¼C8ÁC7ÂB7ÂF<ºJ?¬KBJ@|F:b@4L:.A7-@85>88=77<66:5294183083062/62/62/32.52-32-21,12--2.-2./1./1.00.00.10.10.5106005//5,-4+,6,-:01D22T71c;3rB8{E;IERU_lis}y«x}µowµae¢SRMDyL@pL@hPEgQFfLC^HCWNLZ^^fjnquyxy~xzvwzokoa`bUWYLTTLWYVV^------/.,/.,0/-10.3/,40-40-40-50,50,50,50,7/,7/,4/,4/,3.+3.+3.+3.+4/,4/,50-50-50-50-50-50-50-50-3.+3.+3.+2-*2-*1,)1,)1,)4/,4/,4/,4/,4/,4/,4/,4/,41,41,41,41,52-52-52-52-52-52-63.74/74/85096196/<5-=6,?8.@9/B90C;0C;0C;0E;/D:.F:.G;/H<.I=/J>0I@1JG6MH5RG5YF5bE3jD1uB/|?,;):):*=,£B2¬F8²J=¼J@ÌGBÔDCØDDÝEDãCCéAAð=@ô<>ù:?ù;=ô<>í?>áB>ÓC:ÅA5¹?0²?-¯@-®?,®?,®?.¬>-¬>-ª?-¨>.¤?- ?,?+?,?+>*?+>,?.?.>->-=,=,=,=,=,=,=,<,>->.@/B4A4@4¨@3¶A7¿C9ÅB:ÄA9¾C;·H?¦LCKCtE;Z>2E9-=6,A96@86?75>64=53<4294183062/43/43/23.32.23.12-02--2.,2.-2.-2./1./1.00.10.3205105104..2,,5,-7./>0/N5.Y9.e=1oA4tC<yKKZeh&#x7f;pzxªu{·ltµ_d¦TTOGLBwNAmNBhMAeJA`GBYHEXKKWMPU^bc`fbbia`f\Z`VWZOUXMXXP^YV`WX`WZ,,,,,,.-+/.,/.,0/-3/.3/,2.+2.+3.*3.*3.*3.*5-*5-*3.+3.+2-*2-*2-*2-*3.+3.+3.+3.+3.+3.+3.+3.+3.+3.+3.+2-*2-*2-*1,)1,)1,)0+(4/,4/,4/,4/,4/,4/,4/,4/,40-41,41,41,41,41,41,41,41,52-52-63.74/85085096/;4,<5+=6,?8.@7.A8/A8/A9.C9/C9-E9-F:.G;/I=/J>0HA1JG6JI7NG5VF6\E3dC2lA0t?-|=,<*;*=,@1£F5ªJ:´J=ÄH@ÌEAÑFAÖE@ÞCAåA?ì>?ò;=÷;<ô:;ð<=é@=ÜC=ÍC8¾@2²?-®?,«@,«@,ª?+ª?-©>,©>,¨?,¥>-£@- ?,?-?+>*>)?*?,>->->-=,=,=,<+=,<+<+<+<-=,>/?0B4B3¡A3¬B5¹C9ÂC:ÅB:ÃB<»B:±HB£NGNEpH>T@5A;/96-@85A75?75>63=5394194173043/43/34/23.23.13.02-.3--3/-3/.3/.3/02/02/11/11/21/32032040/2.-1-,4..8.,G4-O4)X8+`<0e?6mGFyYdku||²w|¼nu»dh¯¡SLLB~OArL@hI=cH>`HB^ECX@BO<AGCHKDMJJQJJQIIPHKQGOUKVWO^YS`YS`XU+++,,,,,,---/-.0/-0/-0/-1-*1-*1-*1-*2-)2-)2-)2-)2-*2,,1++1++1++1++2,,2,,1++1++1++1++1++1++1++1++2,,2,,2,,1++1++1++0**0**3--3--3--3--3--3--3--3.+40-40-40-40-3/,3/,3/,3/,40-40-51.62/730730841850:3-;4,<5-=6.?6/?6-?6-?6-C90C9/E8/F90G:1I=1J>2H@3HE6GE6KE5QD4YD3_B2g@/n=,v=,|:*9+:,=.B2¢F7¬F8¼G=ÂF>ÉF>ÐE>ÙD@âC?ê@@ð>>ò::ñ;:ì<<äA<ÖC;ÆD6µ@/ª=)ª?-©@-©@-©@-¨>.¨>.§=-¥>-£=. ?.?-?-?,?+?*>)>+>->-=,=.=.<-<-<-=-;,<,</>.>1?/C4B3¡A3®B6¼C:ÃD=ÄC=ÀC=ºGB²QK¦YSXQsRIWI>CC7<?6>93@72>63=60:5194083/63.43.43.34/23.13.13.02-.3--3/-3/.3/.3/.3/.3/02/02/00.11/22021/10./.,2.-4/,?0+D0)K3)T8-Z<4eGGujs«¾~ÇtzÆmp½ee¯VSLCK?qJ=hG;cE>_FBDBW?AN;?H:BE>HGDMHGQIGQHJRGNVKUXM^ZOaYNaXO++++++,,,---.,-/-.0/-0/-1-,1-*1-*1-*2-)2-)2-)2-*2,,1++1++0**0**1++1++2,,0**0**0**0**0**0**0**0**2,,2,,2,,1++1++0**0**0**2,,2,,2,,2,,2,,2,,2,,2,,3/,3/,3/,3/,3/,3/,3/,3/,40-40-51.62/62/73084185092,:3+;4,<5->5.>5.>5.>5,B8/B8/E80F90G:1I<3J=4I?5FB6FB6JB5OA4UB3\@2c?1j<-q<.w9*}8)7*:,>0B4¤B5²F:ºE;ÁF>ÊG?ÔG@ÞFAçCAîB@í;;ë;;ç>;ßB;ÑD:¿D4°A.¤>(¦A-¦A-¦A-¥@.¥@.¤?-¤?-¤>/¢>.@.?.?->+?*>)>+>->->-=,=.<-<-=-=-=/<.<.</=0>1?1D4C3¢B4C6ºC;ÁD>ÁD@»EA¹PL²T¦ffu_TUIHNBCI?<92?82>71;6094.74-63.43.43.43.34/23.13.13..3-.3-.3/-3/.3/.3/.3/.3/02/02///-11/22022010.0/-0/-3/,8,,<-*C0*K70S<6^HJtbnz¶ÆÏzÌtwÆjl·YW IDE=nG<dD<aC>^CAY@CV@DP>EKGQRKWUQ^WU`XT`VSTT^SY_S^LaZJaZJ,-/,-/--/--/------.,-.-+/.,/.,1-*0,)0,)0,)/+(/+(/+*/+*/+*/+*/+*/+*0,+0,+/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*0,+0,+1-,1-,2.-2.-2.-1-,1-,1-,1-,1-,1-,0,)1-*2.+3/,3/,3/,3/,3/,3/,3/,3/,3/,40-51.62/73080-92,:3-;4.=4/>5.>5.>5.@5/@6-B5-C6.D7/F91H;3G=4G>5H@5J@6P?5T>3X<1^90c7,m9.t8-|8-9.;/=0?1>0§A3A4µC8¾E:ÊG=ÔG>ÞE?åC@è@?êBAæDAÚE>ÈD8·B1ªA.¢B,¢A.¡@-¢?,¢>.¡=-¡=-¢>. ?/<,=,<->-=,=+=*=,<+<+=-<,<,<,:-;-=/=0=0=0>3@4A5A4E5D4£E9±I>ºG@»D>»EA¸MGµS¯f_£qfsh~rdjj^V^SJRGLLBJF=B>5=90:6-74+63,33+54/34.34/23.02-/1,.0-,1--2.-2.-2.-2./1./1./1./1.02/02/11/11/11/11/11/40/4+0;/3A32C41J;8NQym{»ÏÙÓtzÆjn·`c¨ZLItHBdA>>>X?BUAIVLU\U`bbqno~yv|svlyohth_k_W_P^Ib\FcG,-/,-/--/--/------.,-.,-/.,/.,0,+0,)0,)/+(/+(/+(.*).*).*)/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*/+*0,+0,+1-,1-,1-,1-,1-,1-,1-,1-,1-,1-,1-,0,+1-*2.+2.+3/,3/,3/,2.+2.+2.+2.+2.+3/,40-62/62/80.91.:2/;4.=4/>50>50>5.?4.?4.B5/B5-D7/E80G:2H;3H>5H>5L=6O>6R>5V;2Z90_7/i81p7.x8.8/:/<1<1=1¢@3§A3C6´D8¾E:ÉF<ÔE=ÛC>ßD@àEAßGBÔG>ÄF:³D3¥B/B-C.A/@.?-?-?->.?/=,>,=-=,=+>+=,=,<+=+<,<,<,;+;-<-</=/=1=1?3@6B8B6G9F8£G<¯JB¸HD¸GC·KH³TN±d\ªqf~o&#x7f;o&#x7f;|mmseZfZNXMLNCKI=EC7A>5=:188.44,11)23-23-12-01,/1,/1,.0-.0-/1.-2./1./1./1./1./1./1.02/02/11/11/11/11/11/2015+49-7<15?54I?=^UVys}¹ÊÔÏyÂqy¸kthn`XZqSUjRWjT^hZgmfvvr&#x7f;t&#x7f;~~wzrqftgZeTZE`ZBb\D-.0-.0-.0-.0-.0-.0.......,-.-+.-+-,*/+*.*).*'.*',+),*+,*+,*+,*+,*++)*+)*-+,-+,-+,-+,-+,-+,-+,-+,,*+-+,-+,-+,-+,.,-.,-.,-.,-/-./-./-./-./-./-./.,0,+0,+1-,2.-2.-2.-2.-1-,1-,1-,1-,1-,2.-3/.40/51.80.91/:20<41=31>42=31=4/?40?4.A4.A4.C60D71F93G:4H;5J;6K<7N=6P;6S:5W836.c60k6.t5,}7/9/;0<2=2@6 @4¢@3¨@3±C6ºD8ÅE<ÍD<ÕF@×HBÔIBÌI?¾E:®C3¡B0A.B/A/?.>->->->/?0<-=-<.=.>-=,=.=-<,<,<,<,;-<-:-;-<0<0<2>3?4A8C:D9J=H; H>¬KD²KF³LG²SM®`Vsg¦qz{wsocqbXcUNRDMN@HI;DD8@@4:</46+/1&12*01+/0+/0+./*/0+//-//-//-.0-//-//-//-//-//-//-00.00.00.00.00.00.00.1/26+98,:8/4;63HE@_^Yzy~±¿ÇÃ}»{¶|²}«}yq~o~n~~p&#x7f;y¡¨¥|{j{k\hTXX@Y>_@-.0-.0-.0-.0-.0-.0.......,-.,--,*-,*/+*.*).*'.*),*+,*++)*+)*+)*+)**()*(),*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+-+,-+,-+,-+,.,-.,-.,-.,-.,-/-./-./-./+*0,+1-,1-,1-,1-,0,+0,+0,+0,+0,+0,+1-,2.-3/.40/91/:20;31<42=31=31=31=31>3/>3/@2/@3-A4.C60D71E82F93H94I:5J;6L:6N94Q83T50^72e60o6/x8/90;2<2=3@7?4?2?1¥A2®B5¸C9¿E:ÈH?ËH@ÊJAÃH@¶F:ªB5A0@.A/@.?->,>,>,=.?/=-=,=.=-=,=+=-<,<,;+;+;+;,;,;-;,;/<1=2>5@7C:D<E<KBH>G@§JE®LI®QLVªj^§o¡yzxm{lbm_SZJQUFKO@EI:@D6;=057,13(01)/0(./*.-).-)/.*0/-0/-0/-0/-0/-0/-0/-//-0/-//-10.00.10.00.00.00.00.3.27,:6*83-1961HJ?bfX{z£°¶²¯²¤¶¨³§«££ }£°±£~}k|iXfOSU=ZV;^Z?+/2+/2-.2-.2-.0-.0..0..0------.-+-,*-,*-,*,+),+),*+,*+,*++)*+)**()*()*(),*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*+,*++)*+)*-+,-+,-+,.,-.,-/-./-./-./+*0,+0,+1-,0,+0,+/+*/+*0,+/+*/+*0,+0,+2.-3/.40/:12:12;31<42=32=32<20<20>31=2.?1.?1.@2/A30B41C52D63C84D95E:6G96H94K84N50X72_60i70s80}:1<1>2>2@6?5?5?3@3£C5«C6³E8ºE;½G=¾H>¹G=°D8¦A5@1@/@-@-?.>-=,=-=/=/<.;-</<.<-=-;.<.;+;+:*:*;,:+:,:.:0<1=4&#x7f;?6B9D=F@H?KCH?GA£KG¨PL©YR©f¦vh¢w¤£~}tqjue\eTV_LNUCEL:?F6<@27:/58-12*/0(.-)/+(/+*0,+1-,2.-1-,1-,1-,1-,1-,/.,1-,/.,2.-0/-2.-0/-0/-0/-0/-2-18,:5)51++66,GL8aiRzp ¥£¢ª©¹±¡¿³¶¨¶§·¨·¥±««±³´¢{xgzdTbIQT9VU9YX<,03,03./3./3./1./1//1//1....../.,.-+.-+.-+-,*-,*.,--+,-+,,*++)*+)**()*(),*+,*+,*+,*+,*+,*+,*+,*+,*+,*++)*+)*+)**()*()*(),*+,*+,*+-+,.,-/-./-./-./+*/+*0,+0,+0,+/+*.*)-)(0,+0,+0,+0,+1-,2.-3/.40/:12;23;31<42=32=32<21<20=20=20>0->0-?1.@2/A30?40@51@72@93A:4B94C84F74H5/Q51X5/a6/l8-v:/</>0=1?5>5?5?4B4B4 C4¥D4¬B5°D8´E:±E;ªB7¢@5>1>0?.@,>-=,<+<+<.<.;-<-<.;-<-<,;-;-:*:*:*:):+9+&#x7f;9-&#x7f;9-<1&#x7f;<3&#x7f;?6A8C<F?IBICIBG@HD OK§VS§`Z©pe¤r} ¨§ ~~zuq|kdp\iUR^JJS@BK:>E5<@29<134,22*1-*/+(/))0**1++2,,1++1++1++1++1++0,+0,+0,+1-,1-,1-,1-,/.,/.,/.,2,.8*75(13+(56&EK1\gGudy¥´¢Å¥©Î¥É§É°§Ë±¤È¬¾³°²²²¬¡{wrdxR`FNR7QQ5SR6,03,03,03,03./1./1./1./1000//////////.,/.,/.,.-+/-./-..,--+,,*++)*+)**()+)*+)*+)*+)*+)*+)*+)*+)*,*++)*+)*+)**()*())'()'(+)*+)*,*+-+,.,-/-./-.0.//+*/+*/+*/+*/+*.*)-)(,('0,+0,+0,+0,+1-,3/.40/510:12;23<34<34=34<21<21;10<1/<1/>0/=/.>0-?1.@2/>3/=52;62;83<94=:5>93@72C60G4.O4-Y4,d5+n8,x:-&#x7f;;.;.<4<5>3@3A2B1A2A1 >1¦@4ªB7ªB9¦A7>5>3?2?-?->,=+;,<-<-<-;.;.<.;-;.:-;-;,9)9):):)9*9+~8,~8,=2=4@7B9E>HCKFMFICGAJH¡SO¨Z©hb©{n¤y§¬ª¡~wxnjwcdrZgSQIKTCEL<@D6<?467/44,3/,1,)1()1()3*+4+,2)*2)*2)*2)*2)*0**0**0**1++1++0,+0,+0,+0,+0,+1+-6)25)-4-%46!BI'Wc;qWm} ¡ ¨¸ È¤Ð©Ó«¨Ò®§Ñ¡Ë¥½¯¬°®±«{{tqk_sWMBIO3JM0KK/./1,01./1./1.0/.0/.0////00.00.00.10.0/-0/+/.,1-,2.-1-,1-,0,+/+,,*++)*+)*+)*+)*+)*+)*+)*+)*+),+)*+),+)*-(,*(),())'()'()'(+)*+)*,*+-,*.,-.-+/.,0/--,*-,*/+*/+*/+*.*),(',('1-,2-*2-*2-*3.+4/,50-61.;31;31<42<42=32<21;10;1/<1/<1/=/.=/.=/,>0-?1.=2.=4/=60;81::29:49:4;81?61C2+J1,T1+^3,g7-o9-u=.{=0=4=3?2@2@3@1?1=/</>2¢B6¡C7B7?4>2?2>/=.=-<,<.=.=.>/<.<.</;.;.:-:,;,:+:+;+~;+~;+|:,}9,|:.>2>5@7C:G?KCOHOGHEHFOLVg`uiu¡ª¬¨¡~&#x7f;|ytoj~ccz^qXWfOO\HIRAAH8>@399/85.7/,3+)2()2()3)*4*+0*,/+*0*,0**0*,0*,2),2),3*-1+-1+-1+-0,-0,-0,-1+/5*05+,4-%46!?F%T`8n}Rgtz¡~¡z§{¶ÄËÉÇÆÁ·®µ«&#x7f;®©~xvolfZnSJZ@GM3FJ1DF./0+.0+/0+01,01,12-12-21-43/43/43/62/51.41,3/,4/,50-50-4/,3.+3--1-,0,+0,+.,-.,-..0--/,,.++-*).))+.)/.)-/(/.)-/)-.)-.*+.*+/+*/+*-,*.-)--+./*./*./*------.,-/-./.,0/-2.+2.+2-*4,)5-*6.+90+:1,;2+;2+=4-=4->50>50>50=4/<3.<3.=2.<1-<1/;0.=/.>0/>0/@1.A0)@1*;4*77-39/39/560:3-?-+F*'L)'S*&Z/(`5,d<0k@0yA2A0A2B3@3=0;/8-;/>/A1B2C2A0>.;.:2:2;1>1?0?/?.>.@1?0>1<0;/:.:-;-{9+~@3w<.q7)w=/w=/v;-}?2{;/@4E<JALBMBPENEKISRboc}jq|¤¦§§¤~u{ouiib_zXZsUmSWeNPWEJK=C?6@93;0.6**4(*3'+3'+1&*,*-)+***,*(),'+.(,1(-2'-3(.3(.3(02(00)00)0/*1/*0/)-1++0-(//#<?,V\BpzX&#x7f;gus¡v¦w¨vªu´{¿Á¾º³®}«|ª{ª~¨}¦~|yurmh|cYkUNGIP>CH4?B/01+01+01+01+12,12,21,32-43.43.74/74/63.52-50,50,7/,7/,6.,6.,3.+2-*1-,0,+/-./-./-0..0-,1+*/)*.)(-.)//(//(//(//)-/)-.*+.*+/+*0,+.-).-)./*./*./)/0+.....0....../.,/.,2.+3.+5-*5-*7.)8/(:/)<1+<2)=3*>5,>5,>5.?6/>5.=4-=4/<3.=2.<1-<1/;0.=/.>0/>00@1.C0*C0)A2+>3-:5/94.:2/<1/?-+D*)I*(N+'T/)Z5-`;2e=1pA1x@/A0@2?4>4;3;19.</=0@0A1A1B2@2;2;2<2=0?1@/?/?.@1>/=0<.;.;,;-;,x8,x<1t:.n9+t?/s>.r8*u:,}=1A5E<H?KBMBPCOFJHSRfwipw&#x7f;¥§§§§¤w}oveiaaZY{X\wXZnSSaJNUCFH;C@7<737/-3*+2)*1(+.(*,*-**,+),+),-(,/)-2(02(04)13)13)21*20+2.+2.+4.+22-13/.0,)--%8:-SXDoxioln£p¤p¥pw¶¹¶³¯|«x¨w¨y¨|¦|¤&#x7f;&#x7f;{xuqnh{eYjWNIEL<@D5;=/12,12,12,23-23-23-43.43.54/65085085085074/72.72.80-80-7/,7/,4/,3.+1-,1-,1-.0./0.1/-0..0,,.+*/+)./(/1'//)-/)-/)-/)-.*+.*+0,+0,+/.*/.*0/*0/*/0*/0+//-///0./0./0/-/.,1-*2-*6.+7/,90+:1*<1+=3*>4+?5,?6-@7.@7.@7.?6/>5.=4/=4/=2.<1-<1/;0.;0.<1/<1/?1.C2+E0+H/+K--L,/K+.I*/E+.A-,@.*A.(F/)N/*X1*b3-g5.j:,o;-x<1=4<6<5:4:4;4=2>2?3@4B6B5C6?2=/>1>.>/?-=.<,;.<,:-;+;,&#x7f;<)~=+{<+}=1z<1v:/v<0x@1x@1v<.v;-&#x7f;?5C9F=H?JALBNENFKITQf\whnt&#x7f;{¤¨ªª©¥}yximae\a\`^a|UrVQgPHYGBPA:D93:2.3,+0**,)**(1(+1&*1&*1&*0'*1(+2).1+//*.-+.,+0+,0)-0(-0(-1)-01/23/./+(**"57*QVBmu|ik~j&#x7f;lo o~ m¤q«x¯|zªx©v¦v¦u¦w¥y£y¢}}ywtpmh{eYkUNIDH9>@399-23-23-34.34.45/45/54/54/761761:72:72:72961:51940:2/:2/91.80-50-50-3/.3/.3/03/01/01/20.1..0--/-+.0)02).0*.0*.0*.0*,/+,/+*1-,2.-0/+0/+10+10+01+12-11/1111/010.10.2.+3.+3.*91.92,;2+<3,?5,@6-A7.A7.B8/A8/B90A8/A81@70>50>50=2.=2.<1/<1/<1/<1/=20=2.B3.E2.M//R+0W(0U&.P'/I).C/.<1+;2+?2)H1)S0*_.*d/)i9/k;/u<3}<6<9;9;7:6>9?9B9B9B9B9A7B6@1A/A/@.>-<,<,:+9+9*:*:)&#x7f;<){=(x>(x>*>3{7.z7.z<1v;-w=/|A3{@2B7E:H=I@KAMBOFQHSK\Qhtb&#x7f;gm~v&#x7f;|§©ªª¦ ~}yqojifgfeggde_~_\vPhRK^KBPA8D62:/.4*,/(+*%2&(5%(4%(2&(1&*/)+/+,.,-++-*+-*+-(,-(,/',/',/*+-/+,1++0+(/,%99-PTCiqYve{hyh|k&#x7f;n}oyk{ m¥r§t¥q}£p}£p~¤s~¥v~¥x}¤x|¡x}zzwuqnje|bXlSNGDF9?=1:8,45/45/45/560560671761761872983;83<94<94<94<73<73<41<41;30;3083072/61.61.5106216213123121/00./1-.2,04+.4+.4+.2,.2,,2,,2,,3/.3/,3/,40-21,21,32-32.22022032032051051.61.61-;30<5/>5.?6-A7.B8/E8/C9/E;2E;2E;2D:1C90B8/A60@5/>3/>3/=2.=2.=20=20>31>31@51F42M02T,4X)3W(2R)1K,1B30:6-77-:6*B4)M2)X/)^/)f:1j;3s<7z=:<:<;;;::;7>8@:C<C:C:@9@7C2C1B0?/>/<.:-9-8,8+9+:+~=+x>*v?*x=+9-|/'8/>4w4+t6+}A6}C7&#x7f;E:G<I>KBKBMDRHWJ}`N{iQpYu|biu|¢~¦ª¬©&#x7f;¥{}zwzp{lylwkumsmmgk&#x7f;d`rZhTRJIM>@@4;7,70&5*$6('5''3''1'&.((,+)++)+-*(,+(,+*,+*,+++-,*-,*-.*+,#$3+(50*85,BC5UZFfpWn}^teqbuexivjshtiwkz¢nx lwmx n{¢s{¥u{¥w|£wyuzwwtsomhd{_WkPNGCC7>:195,560671671671782782872983983:94=:5>;6>;6>;6>95>95?74?74>63>63;63:5294194184195495484384343151240/6-06-.6-06-.4..4..4..4/,40-40-40-51.32-32-43.43.43/442542540841850:51:5/>71>7/@7.A8/C90D:0G:1H;2G=4G=4F<3F<3E;2C90B71A60@51@51?40?40?42?42@53?53@72D63I35P16T/6S.5P05J22C52=90<:.=9-C7)I6(Q3)W2)2+d3,l50v64}77877777~75}:4}<6~?8?9@8?;A8C4C1B1@0=/;.:.9-9,9,:,&#x7f;<,|=,y>,x?,|=,4,2,¡LE¨XQA8|90F;K=yH:zJ<{M@|NA}OBQEUJ_LrkOptS|uXwyb&#x7f;is zz|£&#x7f;¨§¤~{y~vpopp&#x7f;r{pwmtlj~cfv\_hSVGOM>GA3A7+=0';,'9+(6+)3+(/,',-'+-().().(+-(-,*/+*3)*4(*7'*7'(3($<3,E>4IG:QR@^cMgqVjyZo^k\lpbqeodoerfumrjqitnx¢rz¤vy¢vyvvtwvurokkfc|^UlON\ECC7@91;4,782782782782893893:94:94:94;:5>;6?<7@=8@=8@;7@;7B:7B:7A96@85=84=84<73<73<73<74<74<74;639529338308/09/.8/080.80.80.61.61-61-61-52-63.63.63.74/54/540651841952;74<94=84@93@70A8/C90D:0G:1H<0I=1J>2J=4J=4J=4I<3F<3D:1C90B8/A81A81@72?61?61@72A83A83?74@85B86E:8G96I:7H96H94E80E8/E9-E9+G9,I9*K9+Q7*Z/&d/'n3-z6398;9;:=;A>@:>7&#x7f;<4&#x7f;:3<4=8@9A5B3@2?2=1<0;/</&#x7f;=/}>/|>/|>/|>/}>/&#x7f;=/9/1+£;8ÎkfÛ~yZTD=LA&#x7f;N@tJ<rN>pQ?qR@tSB{WGJ~eOmsOpzUyu_©oc¶mf¾qk½wo»w±z§~¡¡~yvrrrrt}sxptpmiiedxkRV^GMP;ED0B;+@3+?2,;0,70*30)00(./)./)01+30+7-,;+,?),D',F%,D'+F5-LC4VP@XE`bLgmSjxWj{WlgWgZl^ocnanbrfskpjoirnvsy¢xxwvvssuvsrnlkdb{UmMN\CGH:E<5@707827828938938939:4:94:94;:5<;6?<7@=8@=8@=8A<8A<8D<9C;8C;8B:7?:6>95>95=84>95>95>95>95=85<73:52<41:0/:0.91/91.91/91.91.72.61-61-63.63.63.74/74/74/540651952;83<94?:4B;5C<4A8/B:/D:0E;/H<0I=1J>2J>2K>5K>5K>5J=4F<3E;2C90B8/B92B92A83A83A83A83B94A:4@85A96B;5D=7F=8G<6K<5N;4M6.O7-Q6+Q6+R8+P9+P9)W7(f6*r6,~;3@9D@HDJGLF QJLDD;<17-8.91=6>5>7=6<5<5=4}=3z>3x@3vA3x@3z>3&#x7f;<3:38442(&³=;í{zþÅhcG@K@xH:nM<jQ=fT>hV@lWBtE&#x7f;`LhP{rQvUt\´l^É__ÓV\ÙQ×T\äouÙzxÌ~Àµ«}£xvrtuuuusr|nyju{anpZgbN_TBUE5R</O4-N2.I0,D/*>/(9/&7.'6/'92*;0*>/,B-,G*,I),L'.I)*TB6YQ<d^HgeLilOnuVm{Xl~XnhVgXmpbqbrdvitlqkojrmwvzxxytuuuuusrokjeb{UmMMBMN@KB;F=69:49:49:49:4:;5;<6=<7=<7=<7=<7@=8@=8A>9B?:D?;D?;E@<E@<E@<D?;D?9C>8E>8E>8B;5B;5B;5B;5B:7B;5B:7A:4A75A83A75@72@64>71>71>71<71<71;60:5/96185074/74/961961961:72<73>93A:2B;1C:1D<1F<2J>2L@4L@2N@3N@5N@7N@7L?6K>5J=4H;2F<3E;2B90B90@91?80?82?82@93@93<5/MD?M@:K:3T?:S73O0+a<6\3-`5.`4+^/'^/%f6,m=1q=0J>G;I?KAF>>7>7D;¦ND¥QF©WK«YK¥OBA5;0>698:9655284?8|@6r>1rB4oA2q=0w;17243-1',ÈHIèbaÅEDÍWUáyv«VOF<VFnS>iAibFjeHj`En^D|dLjTbL´lÑnhÛXà;Lç-Eò+Hõ0Lö@YìI\äeånrâ{|Û}Û&#x7f;ØÃ~w½wºx¶x²xy¨y¥xo~mxiocbYUM{IBxA>u:<i.2d04b14P&'G%#L0-M51G4.E2,D1-F1.F.,G+*K-/Q51YH6`V;f_CgdEkoNu|ZuoYlWlZp_ratcudwfxhqisotqroutyx{|yzxysrmlkgidazZTmMO_DNOASJAQH?9:49:49:49:4:;5;<6>=8?>9>=8>=8A>9A>9B?:C@;D?;D?;FA=FA=E@<E@<D?9D?9E>8E>8E>8E>8E>8D=5D=7C<4D;4C:3B92B92B92B92B94A83@93@93>71<71<71;6096185085074/961961961:70<71>:1A:0B<0C;0E;/I=1J>0L@2L@2N@3N@3N@5N@7M?6K>5I<3H;2E;2E;2C:1B90A:2@91@93@93A:4B94E:4J71L.,V..j68u9;{;;ECE@IAIAH?IBNH¡PL¢OGD7B4B6 D9 >37-7-=2£C5@1@0 D5¤G6§F6®H:¶NE´LM¯HL¥@D9;98>:@8x>3o>0zI;LA?8-,).³9DÒLWÜKNßMMÊ>=À@?Ð`\³YQ@5UEyaIe@_^@ihIslOvgJ~dKcN¼dXÖ_æUZëANò,Eÿ&Eÿ&Hÿ'Iÿ.Nù3Nò<RðIëP^åQãSàXbÛbkÕflÔioÒmqÏqrÍutÉxwÇywÊ||ÉyzÈvxÇqtÆkrÅfnÅakÃ_kÅ`n³Sa¦LXGQELEIm69H&#x1c;&#x1b;T1-M0*I2,J6/L92K:2O>7UE8O7cZ;hbBifEnrO{^zbt^r_p_p`r`udwdwdwfskxt|x{x&#x7f;}z|rqmjhd`yTmMP`ERSCVPDUOC8938938939:4;<6=>8?>9@?:@?:@?:C@;C@;C@;C@;E@<E@<FA;FA;FA;FA;E@:E@:G@8G@8JA:I@9I@9H?6F=6E<3F<3E;2D:1D:1D:1D:1D;4D;4D;4C<4?80?80<71;60;60:5/94.94.96/96/:5/;7.>7/@9/C:1C;0E;1F<0I=1K?1M?2NA1P@1P@3P?5M?4L>3K=2I<3H;2E;1E;1D;2C:1B;1A:0?;2@<3A=4D;4VE>U4/j25AGLWÃUbËfÆ_cB>I?LAND¬QL¾URÌOSÄLM¥@6B0 @0¥?1¥;.£7+¤8+§=/«E6¡=-7$:'¥>-¬A1²C2·E;§02¯:B¹DL½LRºMRKLA>71{8/v6,x2*1-37¹=GÕGWèMâFIÏ53Ð:9½31ÈNKÆcK@J}cLk`DgdEtoO&#x7f;pSgLhN°hYäbbõQZôDQö7Iþ2Kÿ2Nÿ/Mÿ+Iÿ2Oü0K÷1Jõ5Nó5Nð2Kó3Lô8Qò@ZïCîF_íIaëNcêQfèTjèVkåSjäRiäPjäNiçNlëPoòSsóVuþgñ`}åuÕUlÎZmÉdr¨Rr-2^&'Y0,W:4Q?5C;.:8)>B3LN9VO3aV6e_?heDquT&#x7f;eixdxiuetdrcveyhzizjv&#x7f; £¨©¦¢zypmh&#x7f;c_vYTmMPcGTWDYVGYVG7827828939:4;<6=>8@?:A@;BA<BA<DA<DA<DA<DA<E@<E@<GB<GB<GB<GB<FA;FA;HA9HA9LC<KB9JA8I@7H>5G=3F<2E;1D:0D:0E;1E;2D;2E<3E<5E<5@91?80=82<71;60;60:5/:5/:70:70;7.;7.?8.@9/C;0D</F<0G>/J>0K?/M@0NA1P@1O?0O?2O>4L>3K=2J<3H;2E;1D:0D;2D;2B;1@<1@<3A=4B>5G=4S81*&;@¾T`ÙTeæTgãUcËJO 41;0?1A3®F=ÅIGÕBHÎ?C±B7§F5©B3¬@3¯?3°>3³B4µE7²G7©B1£<)¤;(©;*«:*®6&3(¶97º9=½7>¼7<¿<BÆJLÄTS¾XS?74*,%61ÃHKÜOWãBRÛ3@Þ:9Í.*Ó84È74ÆE@ÈU©WKXGwW@veI~pSqRfLcL©oÕuiîSWüDP÷;Jù:Jþ=Nÿ@Qÿ;Où4Hÿ?Sÿ9Pû4Ký3Ký1Jÿ/Iÿ0Nÿ6Sÿ3Tý4Vý6Wú7Wù:ø=\÷>`õ@aøCf÷Agõ?eõ?gø?hüCmÿErÿIsÿOwÿQuþUvòSqçTpåaxÖfv·Wb~17j0/Y2+P8,G>/>A.;D/?G/QL/_T8g`CokN||`rv~q|q|t|szqzm}ou|£©£®¦±¡¬´©±¸°°·°«µ¡«¢¡sqh|ctXSlNQdHRZCWYDYZH671671782893:;5<=7?>9@?:BA<BA<DA<DA<EB=EB=FA=FA=FC<GD=HC=HC=HD;HD;JC;JC;LC:KB9KA8J@6K>5I=1H<0H<0E;/E;/E;/E;1E;1E;1D;2D;2A:2@91@91?80<71;60;60:5/;7.;7.;7.<8/?8.A:0C;0D</G=1G>/K?1M@0NA1NA0P@0O?0O?2O>4N=3K=2I;0G;/D:0D:0D<1D<1B;1@<1A=2A?3B?6K=4\50w32LSÛiæM_å?S×6E·&+¨1+7(;'<)¬@3ÃD=Ô>?Í;<±?5¨B4ª@3>3²@5¸C9¼G=»I>¬>/ª?/¨=+¦;)ª9)°:,·;/¼<1ÊFAÏEBÐ>?Ð79Ö7<ÞBEâLNÞSPÄC>ÇLDÍTLÔSMÙKJÜ@Cà5>á27Û4.Ú7.Ð1+Ï:4¿84³@9µPTD{N9bIkPaGF®gUËrdçmhêEKò:Dê9Cç<DçBHçFKèGMêFMõKTóAOó8Iö5Hù2Gþ0Gÿ3Mÿ8Rÿ9Vÿ9Xÿ:Xý:Zþ;ÿ=_ÿ?dÿAeÿCiþDkþDmüCmüCoüCoýCrþCpÿGpÿ@hÿGmÿMpòIjéOkårÙcs½Ye?D`&$Q+"V@3VO=IL7@D-RJ3eYCvkW{gy¤¨²¶¨¸¼®¿Á¶ÀÁ¹ÂÃ¾ÅÅÃÃÃÃ¾¾¾°µ±¢©¢yvi{e\sYTmORfJQBUZCW\F560671671782893:;5=<7>=8@?:@?:C@;DA<EB=EB=GB>GB>GD=GD=HC=ID>IE<IE<LE=LE;LC:LD9LB8KA5L@4K?3J>2J>2G=1G=1F<0F<0E;1E;1D;2C:1A:2A:2@91@91=82<71<71;60;7.;7.<8/<8-@9/A;/D</E=0H?0J>0L?/M@0NA0NA0O?/O?/O?2N=3M<2L;1I;0G;/F90C9/C;0B<0B;1@<1@>2A?3B?6N;4n95FEÄY_ÛWbßCQÞ8FÔ3;½++1'¡8%:!9 ¤;&¶>.Ã:2¾71§7,¢:-¥9/§7,¯:1¸A9ºE<·E:®>2?0§<,¤6'ª6)¶>0ÃD;ÊG=¾8-Î@6âE@ìBBô=Aó9>ë27Þ.0Û75âGBèSLåNGÜ=9Ù10ã/2ê67Ù1(Þ</Í,"Ð71¾3,°7/Ég\¸l^WCZCV?R=©\LÎlaágbåUUðEKñ=Fä;@Ù=>ÒA>ÒGDÝOMëWWíNSïDMò9Gö5Fø1Fø.Dû1Gþ7Nþ8Sý8Tþ9Wÿ:Xÿ;\ÿ>aÿAeÿCkÿ<fý<hý>jý?mú>nø>oô=mö:jÿHrÿ;aÿ?dÿJpÿKn÷NmïUoãYpèrÄ`jINt:8\6-P9+PC2UM:^QAreU|n¦§¡¦§¤¦¯¬¦¬ª¢§£¦¨£µ´°ÂÁ½ÌÉÂÑÎÇÕÐÌÕÐÍÔÎÐÕÏÓÑËÏËÆÊ¿½À°²¯~m~kau\XoSVjNS`FV`EX`H560560560671782893:94;:5>=8>=8A>9C@;DA<FC>HC?HC?HE>HE>ID>JE?JF=JF=MF<MF<MD;ME:MC9MC7NB6NB4N@3MA3JA2JA2I@1H?0G=1F<0D<1D<1C:1C:1A:2@91?82?82<71<71<8/<8-<8-=9.@9/A;/D</E=0H?0K?1M@0NA1P@0P@0O@-O?/O=1O=1M<2L;1I;0F:.F90C9/B:/A;/A:0?;0?=1@>2@@4Q:2r1/©KLÁRX»>D½06Â03Á//º1)ª2"£9#>";&#x1f;< ¡=#¥;%¤6%6) 8-£6/¥4,®92·@:¸C<²@6±B7A4¨</¦8+¯:0¼C8ÄE>Å@7Ã>-Ï>-Ù5+á*&í"%ú%+ÿ+2ÿ.4í)*ç.+á3,ß4,à3,ä3-é0-ç2+Ú2%Ö6&Ð2&Ï7,Å6.½>7È`UÒ&#x7f;q¦eSVBM:£WIÈfßd_çRTèCIõFMîAEß??Ô@<É@8Æ?9ÑFAãMLïJPô@Kù9Hý7Hü5Hø3E÷6Gú;Mû9Qü7Rû6Rý6Uÿ6Xÿ9ÿ;aÿ<gÿ>kÿ?mÿ@qÿBtÿBuþBuüBtùBpÿFmúAaþEeÿJlþKkÿVtÿ_zõvòj~æp~Ómx¯aCAg6/hB7sUJye\}v££¡²®«¶¶¶²¶¹ª¯³°·½¶½Å¾ÁÊÂÂÊÄÃËÉÆÍÕÎÖÝ×ÛâÚØçÝÛéÝÝæÚÜåØßä×ààÔÞÙÏØÊÅËººº¢§£tqexb\sYoSjM\iKgL561561561560561671983983<;6=<7@=8B?:DA<FC<GD?HE>HE>HE>HE<JF=KG<KG;KG;NH:MG;MG;NF;MD;MD;MD=NC?NC=MC7MD5LC4JA2H>2G=1E=2D;2C:1C:3A:2A:4@93?74<73<71>7/>7/<8/=90>:1B;3C<2D=3H@5I?3L@2MA1NA1P@0O@-O@-L?/L>1M<2L;1J91I81E80C90@9/>:/;;/<<0=>0>?1A?0S8-<7Ä\ÊZY®86«.*²1+¯/&2#¬9&¬B*ªF,¦D)£A&£A&£=$9#¡:+£;0¦:0¦7.<4¶C<·E;±A6«=0©=0©;.=1»E9ÇMBÅE<¼8,¿6$Î;)Ý;.æ3,ñ+*ù(+ý&+û%'ø**õ.+ï2,ë4,è3*å3)ä2(à3%à:*Ñ2&#x1f;Ü?0Ï7*È8-Â?5³A6Ñl`¿gY¥RB¨NCËcZçhbæRRêAFôCKë>Dá;=Ö<<Í@9Á>4¶8,¹6.Ä94âHHé@Eð:F÷;Iú:Iø8G÷;Iû?NþAUÿ?Vþ<Tþ9Tÿ8Uÿ9Zÿ:^ÿ;cÿ>iÿ?mÿ>oÿ>pþ?rú@rø@t÷Aq÷CjÿMmÿSs÷JhìAaýTsÿgÿe÷wï_xèg{Üp&#x7f;Æqx§gg^X{ZQ~wª¾°Ä¼ºÆÂÃÈÇÌÇÇÏÃÃÏËÊØÍÉØÑÊÚ×ÎßßÔäæÛéêàëíãëïåæòèæôèèðäæîáèíàéêÝçäØâÑÊÑÁ¿Â««©zxlzibu_`rZcqWcpTcmT21/320542651875984984984;:5<;6==5??7AA9CC9EE=EE=HH@HH>HH<JH;KJ8LK7ML8OL9NK:JF:HE<JFCNIMNIPICOE>ENE>LC4J@4KA5MC9LB8F=4B90E<5C:5@93@85@85@86<74<42C97B94A96>95?:7>:7?;8@<9B?:D@7G@6J@4L@0O@-O@+L?,G@.G@0I;0K81M53L76J88F;9C@;6904</9D4:A/69$?=(bC1¼j^¹JA¦7,¥5)§7+¥5'¥7&¨=+£8$¥<&ª@*@+¬=):'®6%«6%¤6'¡7)£9+¦<.ª@2¬B4¬B2¬A1©>.«=.<,³=/¾D7ÅG;Á?2º2&Ç9+Ì8*Ð8+Ö8-Ü8.â5.é3/î1-ò/-ô.-ó0,ñ1,ê5,ä8,Ý:)Û:(Ü9&Ô3&#x1f;Ð3"Ï8'Æ7'¾6(ÂB5ÏSIÈMEÐTLáXî\\ðPRí>Cò9?÷?GêAFßCDÖBBÈ?9»<3±;/°</²<0ÊKBÔKCÜFEá>?ë<A÷?Gý?Iù;GþDQüCQþDRÿDUý@Vü<Uÿ<ZÿBdÿFkÿAjú;gõ:g÷=nùCtýH{ÿK{ÿMyÿMtøPsðOpçMiêMjüYxÿgÿ_õGl÷Vxè_{Ã^n½za^xp§»§ÑÄ»ÝÖÐÝÜÚÜÛàÞÜçâÛëæ×ìëØîòØñöÝó÷âó÷çôõëôôîòùôñùôðùñïöíîóèìîãééÞäæÝâÖÐÔÍÉÊ»º¸£¥ yxr}op{k}sxlv}k0./1/0320542653873873984;:5;:5==5??7AA7CC9DD:EE;HH@HH>HH<JH9KJ6LK6ML7NM8LK9KI=KJEPPRWU`YXjVTjSO`QHKMD?H?:F=8G>9H?:F=8B;5D:8A96?74?74@85@85=85<73=31=31=52=85=96>:9>=;>=9B?:D?9G@6J@4M@0O@-P?+L@*B?,B?.F<0H:1J65I56F35@65A=::=6;>5@@4E<-J9)^B4L>®QB«;- 2#¡6&£:'7$9%£>*¡;%§<(®?,³@.¶=,·9*¹7)¶8*ª9)¥:*¥:*¦<,©?/«A1«B/«@.®?.«:*²=,ÀF7ÆH:Á?2¼8+À6)Ë7+Ï7,Ð9.Ñ;-Ô</Ø:.à8/ç4-î1-ô.-õ/,ò1,ê3+á7*×:'Õ:&Þ;(Ù6#Õ8%Ô=,Ì;*Á5&Ä:/ÏF<ÑF?ÚIFéOOóPQôGKñ<Aõ:AøCHåBEÛEDÒBAÅ>8·;1®:-ª<-«=.¶E5ÆL?ÔNEÜGCä@Aï?Bû@Iþ@JùDM÷CLúDPÿFTþCTú?Tü>XÿBaÿGiÿCiü?iù@jùBpüFvþJzÿK|ÿJzÿHtõKpñQsõ\zù_{úWvüNqýEkúDlïEjÚKiÈ^r±nukLI~t°Ì²¥äÓÉìãÜëçæêéîêçòíâóöãùûáúÿâýÿçÿÿìÿÿñýÿöþÿùûþú÷ýúõü÷ôúõòøïòôëîñèíïæëæÝàÜÖØÊÆÅ´³± ~/.,0/-10.21/43/651762761:94:94<<4>>6@@6BB8CC9DD:IF=JH<JH;LI:MJ7NK8OL9OL9KI:NKBTRS_^fihxmmlkkh`YiYQ\OGRH@KH>GG=EE<AA;=D;<?:7=85<73=82>93>:1>:194.:5/;60<92>;4?<5?<5@<3E>4G?4I?3L@2O?0P?/P?-L?,<;&:=(?<+A;-B71A62>42;30=84B;5H94N2.Y,)l/.<>¢GF£;0¢2$1"¤9)¥<)8%7$¢;(¤;&«>*±A-¶A/¹>.¼=.¾:-¾</²<.¬<.ª:,©9+«<+>->-<,²?-°8(¹>/ËK>ËG:¼6*º2&É9.Ð6,Ô6+Ô8,Ö:.Ø:/Ý9/ã6/é4-ð0-ô.+ô.+ð0+ç2'Ý6&Ô7$Ò7#Ý8%Ü5#Ú9'Ú>/Ó=.È6)Å9,ËA7É<5Î?9ÙEAåKIêJJéCCí@BëCCÞCAÕD?ÍB=Â=4¶:.®8*«:*¬;+¬8)ÀD8ÕNHÞJHâ?Bê=A÷BKÿHRô@Iô@IùCOýGTþEUú@Uû?XþA_ÿEgÿCiÿCkÿFpÿIwÿK{ÿL|ÿJzÿFxÿHwÿKtûOuÿZ|ÿaýUvñCdûIkÿTvÜ?\ÍI`Ø{¸||aC;{p®Õ³§óÛÑûìåúòðù÷ú÷ôýôìûúëÿÿêÿÿìÿÿîÿÿóÿÿ÷ÿÿûÿÿýüÿþûÿýùÿúøÿøöÿôøýòöûðôùðóðçêåßáÓÏÎ¾½»®©¢¥£££«¦£/.,/.,0/-10,21-43/54/650983983;;3==5??5AA7BB8CC9IF=JF;LI:MJ9PK8OL9QL9OL=LJ>QPL^bmlzzz¡¨¦{urke_{YSmTLdMEZG@PB<HD?F@;?<87;63<71=90=90;8/:70991;;3>;2?=1@<1?;/A;-F?/H?0K?1M@0O?0O?/P>0L?/@=*?>,@=.?;/?;2>93=:5:94<94E:6M51V-+j)-188C¬=C§8/¥7(¥:*ª?/¨?, 9&¡8%¤;&¥:&«<)´?-¹@/¾?0Á=0Å=1Ä>2½?3¶=2²9.®8*¯9+°:,²:,³9*¶;,º<.ÄB5ËE9È>3¿2(À2(Í9/Ö5+Ù5+Ú6-Û7-Þ7.â7-ç4-ì1*ó0,õ/,ô.+ï0(ä2&Û4$Ò5"Ð5!×2&#x1f;Ú0&#x1f;Ú4&Ü<.Ö=/Ñ:/Í=2ÒD:Ä5-Å60Ê;5ÙEAãNJåKIÞC?Ö=8Ó@9ÎA8È?5¿=0µ9-±9+¯9+±;-¬4&½?3ÑHBÙHEßABæ@DôGMýNUð?Gð?GöBMþHUþHWüBWú@XüA^ÿAcÿCfÿFmÿIrÿMyÿO|ÿL{ÿIyÿDvÿN~ÿR}þKrøMoÿVuÿUu÷MjñGdðLgÒ>Vádtþ§Ód92lJ@¥ypÐ¥õÔËÿéãÿôòÿýÿÿüÿùóÿüïÿÿïÿÿðÿÿòÿÿôÿÿ÷ÿÿûýÿüûÿýúÿüøÿúøÿù÷ÿö÷ÿôøÿôøþó÷öëïìãäÚÔÔÉÅÂº¹´³´¬±´©°¶ª¶¾³°¸¬´©10,10,0/+0/+10,21,43.54/77/880::2;;3>>4??5AA7DA8IE:LF:NH:PJ:RK9RM:SL<RL@OJDVTUfdqwx©¹ÃÆÀ¶{w©pmid_~UPnNJaKGXFCN?=B<8995296/85,85.671783891;;1></@=,B=*C<)G?,J?-K@.L@0M@0M?2L@4K>5M@7K>5F<3B92=82:946;47<59<5>:1H4-Y2-x89?D¦>E©8<ª;0«=,«@.¬A/©@-¥<)¦;'«>*§8%®;(¶=,¼=.Á<-Å;.Ê<0Ê=3Å?6¾>3º:/·7,¸8-¸8+º8+º8+»7+ÇA5ÍC8Æ9/Á1(Æ3)Ì8.Ï5+Û4+ß3)á4-â6,ä6-ç4-í2-ð/*õ/,ö/*ò/)í1(ã3&Ù4!Ð5&#x1f;Ï4&#x1e;×2&#x1f;Ù/&#x1e;Ø2$Ø6)×9-Õ;1Ö?6ØE=ÖF>Í@7É<5Ë@9ÖKDÙNGÒE>Å;1Å=1Ã=1À</º;,·9+´:+¶<-¹?0µ9-¿<2É@:ÒC?ÛBDãDHíJOöOVí@Fí>EòBLûIUÿJYýFXúCYûCû@_ûBdÿElÿJsÿNzÿO|ÿLyþIvÿJwÿTÿT}ýJqùImÿStÿVtøNiëD^äFêXkÿ¢ÿµ¼óHFk.)¦keÍóÈÁÿãÞÿñðÿüýÿýÿü÷þþöÿÿöÿÿøÿÿøÿÿúÿÿüÿÿýÿÿýüÿüùÿûøÿùøÿø÷ÿöøÿõ÷ÿôøÿôøýñóóéêäÜÚÖÑÍÍÊÃÉÉ¿ÊÊÀÊÍÂÉÐÈ¾È¿¸Â¹65154/32.21,10+21,32-43.66.77/880::2<<2>>4@@6CA5JD8ME8OH8RK;TK:TM=SK>RKAQLIZW^li|&#x7f;~¸ÌØÞÚÔÉ¿~µxv§nkfe^xWTiMKYDAJ><?;:8;76875557457664872<:.?<+C>*F@*G?*H@+IA.IB0IA4HB6HA9G@:K=<I;;C9:=77875384/83/917<574+C1'80~C=GC>?52¨<0ª<+ª?-ª?-ª?-©>*¬?+°?-«8%²9(¹:+¿9-Ã9,È:.Í:0Ï<4Ë>5Æ;4Â91Á8.Á8.Â9/Â8.Ã6,Æ8.ÑC9ÑA8Ã0&Ã,#Ð7/Ö<4Ó2*à3,ä1*ç2+è3,ê3-í2+ð/*ó-*ö/*ö/*ò/)ì1(á4$Ú5"Ñ6 Ð5&#x1f;Ø7#Ù3#Ö3$Õ3&Õ5)Ô8,Ô<1Ö?6ìYQáRJÔIBÌC;ËF=ÍJ@ÈE;À>1¼:*½;+¼:*¹:)·9*¹;,¼>0¿A3ÁA6Â=4Æ=7ÏA=×EEÝGHäIMìMQèBFê@CðBKùKTþNüJZüF\üF_øA_ùBaýEiÿJpÿNwÿOzÿNxÿLvÿQ{ÿR{þOvýOtÿVyÿ{ÿSoîD^úTlîQd÷dtÿÿôÇ`d§MMµjgÎï¼¸ÿÞÙÿïîÿúúÿüþýûÿýúÿþûÿÿûÿÿûÿÿüÿÿûýÿûûÿûøÿûøÿüùÿûúÿûúÿøúÿöøÿó÷ýñóùíïòææçÝÛßØÒÛ×ÎÝÛÏàÞÒßáÖÜãÛÏØÓÅÎÉ=:5;8185063,41*41*52+63,74-85.:70;81>;2?<3A>5C?4KC8MC7PG8SJ;WK;UL=UK?SJCSJKWcom¦Å Ú¦ê¢©ñ¡¥ï ëåÝÖÊ½±zy¡rqfc~ZWlPN\IGRFCNCAL??K==G;:@;9:=:5?;/C=-E@,F@*FA+EB/EC4CC7BC;AC>@ACB<FD>LD@NEEOCHNAIL?IJ?HECHAA=2L:.gF7OAJ=?48+¥>/©>,«=,¬>-@,¯@-°?-²=+±8'¶8)½9*Ã9,È8-Ì8.Ò91Ô;5Ï:4Í:3Ë81Ë81Ì92Í:2Ì70Ë6/Ô=6Õ>5Ò91Ì2(Î4*Ù<3Ú<3Ø4+ä1*é1)ì1*î1+ð0+ò/)ô-(ö,(ø.*ö/*ñ0)ê2(â5'Ù6%Ó6#Ï7"Ô9%Õ8%Ó6%Ò5&Ò4(Ó7+Ò8.Ð7/ãNGèXPçZSÚQIÌG>ÄD9¿@7¹>/¹<*¹<(¹<(¹:'¸9(º;*À>0ÃA4ÉD;Æ@7É@:ÑFCÕIHÖHGØGJßIKæFHèBDíDKõLSüP\ûOûLûK`öD^÷DaùFfýIlÿNuÿOxÿOxÿPy÷UzóUxôTvüYxÿ{ÿZwûQkòI`ýWköUgêM^ö^mâKZæTaåTc×XcÁefÈ~å«§ÿÔÑÿëéÿóôÿøùÿþÿûüÿûüÿûüÿüûÿüúýýùúý÷÷ü÷ôÿú÷ÿûøÿüûÿûúÿøúÿóõýîñúëîðáäéÝÝäÙ×âÙÔæßÕëçÛñíáññåíòìÜåâÏØÕ@=6>;4:7074-52+52+52+63,74-85.96/;81=:1?<3@=4B>3JB7MC7SG9VJ<WK;XL>WK?TICSJM^Xfpm«Ë¤ã¤¬õ§°ý«°ÿ¦¬ú¡§ó ¤î£ê¡áÖÌÄ´zxji`|XUpRPhOMeOKdKG^FBS@=H?:>?:6@<1A>/C@/CB0BC3BD7?D=>D@<ED;BJ>CVDHaMRhVn\br`dobbje`dfXWaJDlG>TFXFJ7@->+¨A0®@/®@/¯@-°A.´A/²=+±8'µ7(¼8+Â8+È8-Í6-Ñ7/Õ81Ø;4Ò72Ñ83Ð72Ò93Ô94Ô94Ó83Ó6/ÞA:Õ8/Ñ3*Ö8/Ü>5Þ=5Ú91Ý6.ç2+ë0)ï/*ñ0+ô/)÷-)÷,(÷,(÷-)ô/)ð1)ê4)á5'Ú7&Ô7$Ð8#Î7"Ï8%Ï8'Ï6&Ð7)Ò9+Ñ7+Ï5+Ì5,ãNGód\ë`YÕOFÄA7º;2µ9-¸>)¸>'º='¹;%º9&½9*À</Ä>2ÊD;ÊA9ÎE?ÕKHÔMJÎGDÍEEÖHGãIIæCDéFIòMSøSZùR\÷O^úOaôH^õG`øGdûIiÿNrÿQxÿRyÿTzðUuïZxù^}ý^|ùTrõMgùOiÿYnÿmÿctðM^øUfôO`üUgõL_äR_À^_»vq×úÉÅÿåãÿííÿôôÿÿýüÿÿûÿÿûÿÿûÿÿýýýýûüþøøýøõÿøõÿùöÿúùÿúùÿöøüðòøéìôææëÝÝèÚÙäÙÕèÝ×ðèÝùóåÿúìþþòõúöãìëÔÝÜB>5A=4@<3>:1<8/:6-84+84+62)62)72,83-;60>93A<6E>6I@7MC7RF8UI;WJ:XK;ZLANHTIMXR`gd|~§Ê ã ©ö¥°ÿª³ÿ«²ÿ¬´ÿ¬²üª±ùª¯ó¨¬ì©«è¡ÚÒÁ&#x7f;±ur¡heZXR~XQ{YQxVOpNI`C?M?:@?;:B?8>>4AC6BE:?D=<B>:CB>FH?LUEWoSfbqistvxo_mcXZWU¦VM¦M?@.:% :$¨=)¬=*®?,±@.³@.´?-¶=,¹;,º6'¿7)Ä6*Ë7-Ñ7-Ö8/Ú91Ú83×84Ø95Ù:6Ø93Ø61Ø61Ú83Þ;4çD=à=6Ú70ã@7þRÿlcÿTç?6å3)í2)ñ2*ô/)ô*&õ(%ø((ø**÷-+ó0,í2+ã1'Ú0#Ô1"Ô4$Ó:(Ì;&Ê<(Í<)Ï<*Ò:,Ò9+Ò8,Ñ7-Ú@8Ï81ÜGAôc^åXQËB<ÆA<·4*·:&¶<$¹<&º<&½:(¿:+Ä<0Æ>2Ç>6ÌC;ÙPJßXTÔNKÃ?;À?:ÏHDßIHâGEæJKïPTñRWïNVîLYñL\ùRføOf÷Jf÷JhüMnÿQuÿRvýRvôWvîXuõXwûwÿ\xÿZsÿWnÿVkÿbuýVgÿ\mÿlñFXôDXÿQeîVeÆfgªieÊç¶²øÒÏÿêéýïîüø÷ùýüùÿÿùÿÿúÿþüþýÿýþÿúúþöôþùõÿû÷ÿùöÿöôÿööÿõõöèèæØØãÕÔÞÐÍáÖÐñçÞüôçþøèÿüéÿÿóúÿùèñðØáàC?6B>3A=4?;2=90;7.:6-95,73*73*72,83-:5/>93A<6D=5H?6LB8OE9TH8VI8WJ:YK@ZMGUJNXP_fa&#x7f;y{¤Éâ¦õ£®þ«´ÿ¬µÿ®¶ÿ¯·ÿ°·ÿ¯¶ü¯´ø¯²õ©ªë¢£ãØÉ&#x7f;»uq®jf¡d_`W`VT&#x7f;VPtMIbFCTBAIAAC<<:AB=FEACD?A?@DBEOMRWVdchqu&#x7f;}¢}| |¥p¤`u¬Yi¶VaºQU¸HF°@4¬;)¬<&®>(¨8$¨7%¬9'®9(±9(²9(µ7(¸6(»3%Â4(Ê7-Ó:2Ù<3Þ=5ß<5ß<5á>9ß<7ß<7á>9åB=èC=èC=æA;å@:æB9ä@7Ü8/Ô1(Ø5,ëH?ÿZQñ@6î6,ë,$î)#ø-)ÿ//þ,-ö((ú0.õ4/ï61ç6.Þ3)Ö2&Ô2%Ï6&Ë:'È;'Ê<(Í:(Ï9*Ð7)Ñ5)Ñ5)Ô:0Ï5-Ô=6åPJåTQÛLHÍB?·1(¼;(»=&½<'¾;'¾9(Á9+Â:.Ä:/È>4ËB:ÕNHÝXQÓRLÃD=¿C;ÌHCÛLHßJFåMLëSRíTWëPVéNVìOZ÷VføUhùSkúRlþQoÿRsÿStÿRuòMmøUtÿ\|ÿ_}ÿ\vÿUoûRiûReüUg÷RbÿZjÿZjóN^øScþYißS^¶`_^W¶}Ù©¥îÈÅýáÞùééýøõúüùùÿÿøÿÿ÷ÿýúþýÿÿÿÿûûÿö÷ýøõþùõþùõþöóÿ÷öÿõõöêêêÜÛäÖÓÝÐÊßÕÌïåÛüõåÿùçÿýéÿÿïúÿøèñîÛáßFB7EA5C?4A=2?;0=9.<8-;7,84+84+83-83-:5/=82?:4A=4F?5JB7ND8RF6TG6WG7WI<XKCWKKXPb^ysuÂÜ¤ð «úª³þµþ¯·ÿ±¹ÿ³ºÿ´»ÿµ¹ÿµ¹ÿ¯²ù¬¬ô££ëàÓÇxu¼sm³k`¤icZWVRwNKhEDV@@L85<A<@KABL@BQ>BFKpWcus¤{§®&#x7f;«²{£¸vÀpÂcÀSpÂF^ÐK\ÑHRÄ<<º5.¸9*µ<'°<%°;)°;*²:)´;*·<,»<-À<-Ä<.Ä6*Ë8.Ó<1Ù?5ßA8á>7á=4ß:4Ý84Ù40Ù40Þ93çB<ìE?éB<ä=5ã<4Ü5-Ø4+Û7-Ô2'Ì, Ö6*éG:õK>ò@6ï4-ð-)ö,*û-/ü+.ø*,ð*)ì/+é3/ã5.Ü4+Ù2)Õ1'Ñ3'Í7(Ë:)Í:(Ð:)Ò;*Ó:*Ô8+Ô8,Ñ4+Ö<4Õ:5Õ<7åONíYWÙEEÂ3/Á<-¿<(À;*À;*Á9+Á9+À8,À8,Æ=3ÇA6ÏJAØUKÓSJÆG>ÀD:ÇG>ÖKDÙJDÝLIäSPåSTâPSâMSåNWòXdöZhýpþZrýUoûQnýPnÿQrýGmÿPvÿZ}ÿ{ÿVrýSmþVmÿ\mûXiøZhû`nö`lígïfnìfmÉ\_ W{OFmfÁÝº¸ðÔÓôàßÿõôùù÷ùÿýùÿÿ÷ÿÿúþýÿÿÿÿýÿÿøùüøõüùôüùôýøôÿùöÿøöùîìïäâåÚÖÛÐÊÜÒÈîäØýöäÿûèÿýçÿÿíúýôêðìÞãßIE9HD8FB6EA5B>3A=2?;0>:/:6-:6-94.94.:5/<71?:4@;5C?6HA7LD7OF5SF5UE5UH8VJ>YNLXNW^ZqppºÔ¡é©õ¨°ù©±ù³ý°·ÿ±·ÿ²¹ÿ´¸ÿ´¸ÿ¯±ü«ø¥§òêâ×Í}yÅum¶pgªf`_ZYXQQuGGcB?R=5DH8BT=C`@EnCJOU^gªgx½kÉnËoÉiÈ`ÉWxÆIgÄ;UÆ2HÙAPàDOÑ;=Ç82Ä?0½@,±:$5$5%¯6%³5&¶7(½9,Ã;-È</Ó@6×@7Ú@6Ý@7Þ=5Þ93Ý6.Û4,Ó.(Ò-'Ó.(Ø3-á:4ã<4á81Ü5-Ü5-Ð,"Î* Ù7,âB6Ü>2Ñ5&Ë.&#x1f;éF7óI<ýH?ü<9ô,,ñ#%õ')û/2ñ-.ì0/ä2.Þ3,Ú1*×0*Ö/)Ó0'Ó7+Ò9+Ö:-Ø<-Ú>/Û?0Ü@3Ü@3Ò6*æLBåJEÎ50ßGFóZ\ßIKÕA?È</Ä;+Ã:*Ã:*Â:,Â:,Á9-¾8,Â<1Â>2ÈF9ÏOBÎPDÆH<¿C7ÁC7ÎH?ÑF?ÕJEÛPKÝROÙMLÚKMÝLQëX`ógüaqÿauûWoõNhõMgÿMmÿOuÿRyÿUwÿTqüTmýpÿcuÿizÿguúboõamídnãflÙhjÎghµfa^TeG<&#x7f;^U©Í«©æÈÈïÚÙÿóóùøöùýüøÿÿ÷ÿÿûÿÿÿþÿÿüÿÿùüþú÷ûúõûúõþûöÿýùÿûøþôòøíéêßÙÞÔËÝÓÉíæÖÿøæÿþèÿÿæÿÿëøúïêïèãæßNH:MG9LF8JD6HB6F@4D=3D=3?80>7/=60=60=60?82@85?:4C>8EA8LD9NE6RE4RE2VF6VH;ZMGVMRWhlk~±Îã¦î§÷©¯ù°ý¯³ý°³ÿ°´þ°³ÿ¯³ý«®û¨«ø¤§ô ¢ïéâÙÑ&#x7f;É|v¼mk¬de_aYRQsNHbWFY_CQmBLFPLV¬U^¼dÊZhÙRpâRußTuÙPlÒJbÍBWÇ8JÅ.?Ô8EãCMæGLØ>>Í>8ÌF:ÃH6·@,·>-¸=-º<-½>/Ã?0ÊB4ÐD7ÕE:ÙE9ÚB7Û>5Ù;2Ù6-Ù5,Ú3+Ù2*Õ0*×4-Û60Ý90ß80Þ7/Ý6.Ý6-Ö/&Ú6,Ø8,Ï2#Î2#Ò9)Õ>-Ô;+Ø<-èD8øJAýDAù56õ*-ö*-÷/2÷67í55ã41Ú1,×0*×/,Ù0-Ú1.ã81ä91ã;2â;2á=1ß=0Ü?0Ú>1Ñ7+ïWLøaZÓ;6Ñ;:æPQÜFHæPQÏ<4Ê:/Æ8,Å9,Ä:-Ä<.Ä<0À:.¾:-¾<.ÃA3ÇH9ÇI;ÂD6¾@2¼>0ÉF<ÊE<ÎIBÕPI×RMÕNKÔLL×LOéZ`ð^hüetÿhzüqóOgóMeýNkÿVzÿTxÿRrøTmöqûhxûjw÷erüetó_mì`kêktßruÃjfªd\la}fVYI:kXJ{t½ ÞÂÁîÕØþîïûõõûûûûÿÿùÿÿýþÿÿþÿÿûÿÿúýÿþúûþõûüöÿþùÿÿúÿþúÿûõÿ÷òñèáäÜÑáÙÌðé×ÿùæÿþèÿþåÿÿê÷ùëíðçèéáQK=PJ<OI;MG9KE9IC7G@6G@6B;3A:2@93?82?82@93A96@;7B?8EB9KE9MF6PE3RE2VG4UH7YLDUKLZScii|«ÈÞ¤ë¦¬ø©¬û¬ý°ÿ®±ÿ®±þ°ý°ý©¬ù§ª÷¤¨ò¢¦ð£íéãÞÚÏ|Àty³ou©mpigj_}y_xWjP_£O\»S^ÍTÖPWÛHRå@Qè?RäCRÜDPÖCKÏ@DÊ9<É46äJLçIJáAAÖ87Î95ÎA8ÈD8¾>1ÃE6ÃE6ÅF7ÉE8ÎF8ÑG:×G<ÚG=Ö>3Õ;1Ö8/Ö5-Ö3*Ø4+Û4,Û7.Û81Þ;4à=4à<3Ý90Ü5-Ü5,Û7-Ò.$Ý;0Ú=.Í1"Ç.&#x1e;Î7&Ò<+Í:(Î8'Ô8)â:1ó=9ÿ@@ÿ<>þ37ó,/ê,.á+*Ø*)Ô+(×/,Þ44ä88é9;ð88ð75î73è71ã7-Ü6*×5(Ñ5&Í5(çPEÿmcÛHAÈ42×CCÔ?CêVVÙA<Ò>4Ë80Ç7,Æ:-Æ</Ã=1Á=0¼:,»<-¾?0ÀA2¿A2½@.¼=.»<-ÄD7ÄD9ÈH?ÐPGÓRLÑPJÒNLÕONç_cîaiûitÿm|ÿdwõVjõQiÿUoÿWvÿVtúXpñ^pônyöwíksÞXañ`mñ`mìboïs}ìÌ|ªthyjskXON<_VGrh¯ÜÀ¿ïÕØüéëþó÷ýûüýþÿúþÿþýÿÿýÿÿûÿÿùþÿÿûùÿõúýöÿÿúÿÿúÿÿúÿý÷ÿþöùðçíåÚéáÔôïÜÿûçÿþåÿÿãÿÿèùúêòôçîðåSM=SM=QK=OI;MG;KE9JC9IB8E>6D=5B;5A:4@93A:4B:7@;7A@;CC;IE:MG7OG4RF0VG2UH5WJASIHXR^jg}§ÄÚ¢ç¤§ö¦§ú«©ü«¬þ¬ÿ®þ¬®û¬®û¨¬ö¦ªô¤¨ñ£§ð£§î §í¤è¢æâÛÒÊÂ·¨ymcz©Xi»Q_ÐNZàIRèAIê=Aé=;ä?9ÝC9ÖH:ÎI8ÆG4ÂC0Å@1äTIàG?Ú;7Ø64Ø88Ù;<Ñ98È74À:/¾<.¿;.À:.Â8+Ã7*Æ3)Æ2(Ï5+Ð3*Ñ3(Ô3)Ö3*Ù5,Ý6.Ü8/ß>4Ý<2Ü;1Û8/Û7.Ú6-Ø4*×3)Õ3&Ó3%Ð4%Ð7'Ò=)Ñ>*Ê9$À2&#x1e;È:&Ê4%Õ1(ç51ú<<ÿ=@ÿ8>÷37ê-1á-.Ù--Ø0/Ü43ã9:é;=ð9=ö26ø03ô01í1/å2+Þ2(Õ1%Ï3$É1$Ñ=1új_äTLÈ95Ð@?Ë:=äRSäJHÜC=Ñ94É6.Ç9/Å;0Â<0¿=/º;*¼?-¾A/¼?-º;*º;(½;+¾<,¼>0¼>2ÀD:ÈLBÌPHÊNFÌLIÐNLç`dìahøhrÿo{ÿhx÷Zk÷UjþZrÿYsûtîaræjtë}ñèwyÚ_dñgtþm|óeuëlwöì¢Ä¢|ikVHQ<WUFtg^¤Û¾ÀòÕÙùãæÿóøÿúüþþÿüýÿþûÿÿüÿÿúÿÿøýÿÿûùÿõ÷üõýÿùÿÿúÿýøÿý÷ÿÿöÿùïõïáòìÜúõáÿýçÿýäÿþâÿÿçýþìùúì÷÷ëVO?TN>SK>PJ<OG<MG;LD9JC9H?6F?5E<5B;3B92A:2C:5A<6C@9DD:JF;MG7OG2QF0TH2UH5UH?QGFWQjh~¨ÅØ¢ç ò£ ÷¦£ú§§ý¨ªÿ©«þ©¬ý§û¥«õ£ªò£§î¢§ë¤©í¦©î¥¨í£§î è åß×Í¾ ¯ª ¥e&#x7f;´pÃRdÐIWÜBNå>Fë<Aì<<ë@9ãA4ÜE4ÔI4ÊI3ÂF.¿B,Â?-âSEÞD<Ý97æ>>êDFå@DÙ8=Î65ÍB;ÉE9ÉC8ÇA5Æ>2Ã9.Â5+Â2'Æ2(Ë3(Î4*Ó5,Ø5.Ù4.Û4.Û4,äA8Þ>2Ù7,Ú6,ß8/à:.Û5)Õ1%Ø8(Í2 Ê1!Ï9(Ï9(Ç4"Ê7%ÔC0È7&Î8*Ø7-Þ5.ç40î53ö:9û?>ÿLKúHFòBBì>?ê<=è8;å26æ+2ô+1ø)/ô,.ï/.è1+à4*Ù5)Ó7(Æ0!À.&#x1f;ôdYëSÏ@:ÔDCË;;ßMMêRQàHEÔ<9É6/Æ8.Æ9/Ã;/¾:+¾<,Á@-ÀA.¼=*·8%¸9&½<)¾?.µ7)´8,¸>3ÁG<ÅK@ÅIAÆJBËJEåa_æ^`ñdjþnwþkuóiðXeømú^sõbtèdqãlrì~&#x7f;ùöéqsøryÿyõaqàXföÿ´´Ð®¢ jPDR9QTAjbUÚÀ¿òØÛöàãÿô÷ÿúýÿþÿýüÿþüÿÿýÿÿûÿþùýûÿúôÿôòþòùÿöüÿöûüôÿýöÿÿöÿÿóûõç÷ñáþùåÿýçþüãþýáÿÿêÿÿñýýóüüòXNBWOBWMAUM@TJ>RJ=RH<PH=MC9KC8KA7H@5H>4G?4H>4F?5GC:GE9JF:NI6RJ5UJ4VK5VK9XNEULMZTbkh{~§Ã×ç£¤ö¥¢ù¤¡ü¡¢ü¡¥ÿ¢¨ÿ£«ÿ¤ü¡ªõ¡©ñ¡¨ì¥¨íª©ï©ó±©ö¬©ø ©ø§ò¡âÖ¦£Î¶¾¹¸gz¿LÑERß>Mç;Gç<Dæ=Bà@@à@@â>?ß<=ÝEBÉ:4ÊD;ÊE<º1)ÜKFèJIéCEë>Bé<@æ<?á=>Ú<=Ô<;ÏB;È?7ÊD;ÑMAÍK>¿=0¸6)½9,¿7+Ã6,È5-Ë2*Ñ/*Õ0,Û2/Þ50ç@7ß9-Ú2%Ý3&å9+ç;-ß8(Õ2&#x1f;Ï2&#x1f;Ì4&#x1f;Ê5!Ç4"Ç4"Ê4%Í5(Ï5)Ê0&Í3)Ñ5)Ò6*Ó5)Õ3&×3'Ü4)à2)æ3.ë52ë33é/2ç-2è-4í.5ö-3ö+/ï+,í0.ê5.ã9,Ù7(Ð4%Ã-&#x1c;Ä3"çXHéYNË;3ßNIÅ41ÚHHáOPÜJJÔBBÍ<7È80Æ8.È:.È<-Æ=+Â;(¾9&¼;(»<)»>,º=+·<,´8,µ;0¹=1»?3»?3¾@4À@5ÄA7ïjaõpiêc_ômjûqqìadôgmÿpyÿlzûhxðdoébiëdjîlnîqoîqoûy{ÿ}ùZlücwßbpû¬¯¶ª}urQJY:LR8snZ¥Ë·°íÕÓÿîðÿöùÿ÷ûýøüüüþÿÿÿÿþÿÿþÿûÿþðÿôåþëæûêïÿðøÿöüÿôþÿóÿþñÿýñÿûîÿúêÿùæÿùãÿùáÿùáüúåÿÿõÿÿûÿÿûYOCYOCXNBVL@UK?TJ>SI=RH<ND8ND8LB6KA5J@4J@4J@4IA6GC8HD9LF8OH6RJ5SK4VK7TK:XNEULMYUckh{}¦ÀÓáì ò ôù¢ý¦ÿ ªÿ¡¬û ¬ô «í§è¢¥è¥¤ê©£í« î¤ï §÷¦ð¡á£Ð±À¿ªÂqÄTbÑCOâ<Fì8Aï7?ê:=å<?Ü@AÛACà<CÛ:@ÚADÌ<<ÍCAÊ@>Á31åOPçDGé@Eê;Bç8?â8;Ü9:Õ:8Ï;7ÔE?Æ=5Æ=5ÌG>ÌLAÂD6¸:,·9+¹7*¿7+Æ6-Ì5.Ò3/Ø3/á55ä84å>5á;-Þ6)ß5&â6(á5'Û6#Ó4&#x1f;Ñ9$Ë9"Ç9%Ç:&É<+Ë<,Ì8.Î4,Ó4.Ö5-Ó6-Ï7)Ë5&È5#Ë4!Ï4"Ø2$Þ2(ã0+æ.,æ,-æ,/æ-2é/4ì+.í,-ì0/è2.á5+Û5'Õ8'Ò9'Õ?.»*&#x19;ÖD5Ñ?2ÙF>ëVPÚB?ØBAçUVâRRÜJJÔC@Î>6Ê:/È8-Ç8*É<+Ä;(¿:'½<)½>-º?-¹@/¶>.²:,´;0·=0¸>1º>2½?1ÁA4ÅC5ëi\òmdçbðkfõolèbañklütxÿq~ÿo|öhtífêY`ñbf÷qpþzxûvwÿx~ùRdþpänö¦©©¦jksOO_;SZ;us\©ÖÄºöáÜÿôóÿúûÿúûÿûüüüüüþýûÿÿúþÿõÿúíÿóáÿéáýæéÿëóÿñøÿòýÿòÿÿñÿýïÿýíÿúéÿöåûóàúòÝüôßü÷äýúóüüúÿÿý\PBOAOAZN@XL>WK=WK=VJ<SG9RF8QE7PD6OC5OC5OC5MC7KD:JF;MG9PI7SK6VK5VK7TK:WMCVMNZWbkiz|¢¹ËÖÞãéî ô£÷§ø ©ô ªï§è¤ã¢ààáãâßÚ¤Ð±À¿¬ÇrÉYqÈCT×ALß>Dã=?å==å<?ã=?á>Cá=Dà<C×8=Ö@BÏA@ÊC?Ã<8Ä96éWWàBCä>Bè;Aã8>Ü68Ö66Ñ96Í<7ÕHAÄ;1¾5+ÅA5ÍK>ÆH:»=/·8)º6)¿7)Ç7,Ï6.×50Þ71æ87é=9×3*×5*Ú6,Ü6*Ý5*Ý7)Ú:*×>,Å4!¿4&#x1f;¹4!º7#¼8)¾9*¾4*À-%ã?=å>8ß>6Õ<.Î8)È7$È7"Í8$Ò7%Õ3&×/&Ù.'Ü/+Ý/.Û//Ü./Û,)ß0+à5.Ü5,Õ3&Ð4%Ð9&Ñ>*ÔA/Â1 çSEÒ;0ÚA;ÞC?éKJðTUíTVèTTãOMÝIEÖC<Ï<2Ê7-Å6(Å8'À7%¼7&¹:'º=+·>-µ=-²=,°:,±;/µ<1µ=/·=0¹?0¾C4ÃE7Ü\QêgäaYðlgöpoçabìfgójqàR^ï^kõgsòemñ`gõfjùqqþxwútuÿnuóJøRhämñ¡£b&#x7f;`VoHRd>W`Axx^¥áÑÄÿîæÿøñÿüøÿþúÿþúýÿüûÿýûÿÿùÿýóÿ÷èÿîÜþãÚøÞáùáéüæïüèöýëýþîýúëÿúêÿøçúñàôèØòçÕöëÙøñáþúñýüøÿþû\PB\PBOAOAZN@YM?XL>XL>UI;TH:SG9RF8RF8RF8RF8QG;LE;KG<OI;RK9TL7WL6WL8WK;VLBUMK\V`jh~yy²ÂÎÒÖÛãèì¢í£ë¥è¡ßÚÕÕÖÖÒ½¶¥}±·s¢ÈfÓW{ÕGaÔ<KÞAJàBCÞCAàB?áAAä?Cç=Fç=Fã?FÖ:>Ñ?@ÑEDÆA<º61Å>:êZYÝABã@Cæ=Bã:?Û89Ó97Ð<8ÎA8ÏE;Â:.»2(À</ÇE8ÄE6½>/¸9(½8)Â9)Ì9/Ò:/Ú91à91è;7ë>8Ù2,Ù2*Ü3,Ý5,à5-ß7.Ü:/Ö=/É7(Æ:)Á<+¾<,¿;/Á;0Å<6Î95ëCBîA=ä@7Ú<1Ï9*É8%È:&É;%Ð=)Ð7'Ï1%Ñ1%Ô1(Ö3,Ô1,Ñ/*Ó2*Ó5,Ó7+Ð6*Í5'Ê7%Ì;(Ì>*Ì;(Ï<,ô^PàF<Ò3/Ó/-æ@@ûWXÿtuÿrqÿjgö^YçPIÕA7Ç5(½. Ä8'À8(»8&¸;)·<,µ=-²=,°<-¯;.®=/°</°</±;-¶>0¼B3ÀF9ËMAß`Wâaôpløstæ_cäaædÖHTçYeógrôgoöelükpþsvþvvûpsûenñDXôKbåXkë _zYRkCSh?\gEy|_°ªçÜÊÿõèÿùðÿûñÿýôýÿ÷ûÿúùÿúöÿøòüôêúíÚøÞÎóÒÊëÌÏìÎ×îÔàïØéòÝò÷ãöõãûöãüõãõìÛíáÑéÝÍíáÑñèÙüõíü÷ñþùó\PB\PBOAOAZN@YM?YM?YM?WK=VJ<VJ<UI;UI;UI;UI;TJ>OG<OI=QI<SL:UM:XM9YL9XL<UKAVLJ\Uhexut|~¥·ÂÃÈÏÔÜàßÞÜÖÎÇÅÃ}Á~&#x7f;¸}tp¨gÁ\ÖNvâAcè:Uê9KæAGàECÜGAÛGCßEEæAGî<Jí=JæBI×>AÐ@?ÐIEÀ?9´3-ÈC<àUPÜDCáACä@Aá>?Ú?=Ô@<ÑD;ÏF<Ç=2À8*½5)½9*À>0¿=-½;+½<)Â:*Ç;*Ï;/Õ;/Û8/á90ç83é:5ç:6å63ä20æ21è43æ40Ý2+Ô0'Ì2(Ë7-É;1Å8/Á3/Ã40Í;<Ý?@ë8;í76ä71Ø5,Ì4&Æ5"Ä7#Å:%Ë=)Ë8&Ì4&Î5'Ô8,Õ;1Ó9/Ð8-Ð>1Ê;-Å8'Å8'Ç:(Ç<)Ç<)È;)Í>-Í:*Ø>2áC:Ô/-ä::à24à67×53Ö;7ÞC?æKFêQIëTIèTHäUGË?0Ä?.¾<,¸=-·=.´>0°?/®>0¬<.¬<0=/¬<.<,±=.·A3ºD6ÇMBÝbZàcïolôrrå`cå`eç`g÷kvûoz÷kví`hð_fýlqÿy|ÿy|úmsø^jóAWõD^èUhæ¤h^tJ_vJhwPf³²èâÌÿúéÿýíÿþïþÿòûÿôøÿôôÿñíûêåñãÚíÙÆèÇ»ä¼ºÞº¿Þ¼ÇàÀÍâÃØæÌãëÓìîØóðÝ÷òßóêÛìàÒèÚÍêÜÏïáÖòèßñèáòéâOAOAOAZN@ZN@YM?YM?YM?XL>XL>WK=VJ<VJ<WK=WK=XL>QI>RJ?SK>UL;XM;YN:ZM:YL;XL@ULGYSWe^nnjut~}¥±´¹ÁÈÏÔÕÓÓÊÁ|¸u|²ovio«ljkdz®YvËPoãBdð7Vö4Lø9Iï@EåFBÝJCÛJEáGGèCJò=Nð>NäCI×BDÌA>ÎID»<5°4,ÊJAÕLFØBAÚ>?Û;;Û=<ÙA>ÔE?ÍG<ËG:À8*Á7*À8*¾9(¼8)¼9'½:(À;(Ä;)Ê=,Ò<.×;.Ü8.á6.ä6/è50é32è./ê-1î02õ47õ79î45æ21Ú.,×4/Õ63Ñ32Ï/1Ò/4Ü7>ê:Dí06î02æ21Ü3,Ñ5)Ê7'È9(É<(Æ8$É6$Ï7)Ô;-Ú>2ÝA5Ý@7ÙB7ÓG8ÉA1Ã<)Á:'Ã<)Æ=*Å<*Å8&Ì:+Î8*Í0'æC<Û2/ï?Aã.3Ú*,Õ1/Ó6/Õ83Ö92Ò8.Ì5*Æ2&À1#ÎE5ÆA2¼=.µ:+±9+¯9+«:,ª:,¬<0«=0¬>/¬<.«<+®=-³?0µA4ÉSIàg_ßd_èkiînoæchðjqõoxúp}üs}ömwîbködnÿrzÿx~ÿquúipøWfùAYô@êQfâ}zknoZy`r¶·ßÞÂùôÞüúåÿÿïûÿïõÿïïÿëçúäÜïÙÑãÍÅÞÁ±Õ¯¨Ó¦ªÐ§°Ó«¶Ô®¼Ö±ÉÚºÒßÁàæÌèêÔòíÚóêÛîâÖìÛÑêÙÑìÛÓéÚÓæÙÑä×ÏN>N>N>N>ZM=ZM=ZM=ZM=ZM=YL<YL<XK;XK;YL<YL<YM?SI?TJ>UL=XL<YN<ZM:ZM:ZM<ZN@VKEXOR_Wbd_sjfsozz|~¤~©²¼ÆËÌÊÈ¿|}µvwqr¨kl¤fg l`_sV_³N^ÒJ^ì@Xø7Lþ3Fþ9Có?BéD@áHBÞICâFIéBJò=Nï>NÝAEÔFDÈA=ÉHB´:/±7,ÏPGÊC=Ï>9Ð64Ð21Ð53Ó>8ÒE<ÉG:ÃD5À8*Ã:*Â:*¿:'½8'¾9&¿:'Á:&Å:'Ë:)Ó:,Ù9+Ý7+á5+å4,ê3-ë*+ñ*-ø-3ü/4ÿ17ÿ39ÿ49ý58ù8;õ8<ò9?ò9Aô9Dø8Gû6Hþ5Gô+5ô-2í12ã4/Ù6-Ð8+Î:,Î=,Ë5&Ñ8*Ù;/à>3â>4ä=4å>6ßA6ÙJ<ÐH8ÉA1Ä<,Ã:*Â9'Æ8*È9)É2'Ó9/á@8ëD>Þ0/à..ì59è58Ø3/Õ7.×82×:1Õ;1Ò;0Í:0É;/ÖL?ÌF:¿@1´9*°6)¬6*«7*©8*¬;-=/®>0=/¬=,«<+<.±=0ÁLBÚcÞc^èkiðosêinõrzüvîfr÷oyùq{öjsþlvÿw&#x7f;ÿu|ùhoùcn÷RbþB÷<YêKaÝt{µz}k{fhu¬²ËÌààÄéìÑîöÞåóÙÙíÑÏæÉÅÞÀ¼Õ·´Ë®¬Ç¤ÁÂÃ¡È§Ë¬Ë¡·Í§ÁÒ°ÎÙ»ÙÝÄåâÏìåÕïáÖëÚÒçÒÍãÐÊàÏÈÚÌÃÕÇ¾N>N>ZM=ZM=ZM=ZM=YL<YL<ZM=YL<YL<XK;XK;YL<ZM=ZL?VJ>VJ>XL>YN<ZM<N;ZM:ZM<OAWMDWML\RZ^Xfbqjf&#x7f;spwvxx|{§~µ&#x7f;¿ÅÈÇ}¿&#x7f;y·zuxq§vo¥tk¢pg zaP\HG³EHÓHOîBNø:Fþ6Aÿ:Aõ==îA=æDAãEBäCHèAKî=Më?MØ?BÑGDÅA<ÄH@²9.³:/ÑUKÃ>7É83Ç/,Å*(È/*Î;4ÍC9ÃC6¼>/Á<-Ä;+Ã<)Á:'Â;(Ã<)Â;'Ã8%Ä5$Ê7%Ô8)Ù7*Þ6+á3*å3)ë3+õ33ÿ58ÿ8<ÿ4:ÿ-2ÿ(-ÿ(/þ+1ü-3ö)0ò&1÷)6ÿ,?ÿ*Aý!:ó&#x1a;/ö"1ó'0í,1æ0/Û2-Ô3+Ñ4+Ð6*Ô6+Ý90æ=6ë>7ë:4è71è51â92ßH=×K<ÓE7Ë=/Ä6(Â3%È5+Ï8-Î4,Ô6-æC<à93à21Ò  ì89å55Õ0*Ò4+Ò4+Ï5+Î6+Ê6*È5+Å7+ãYNØRFÊH;¾@4·;/µ;0µ<1µ<1¯9-±;/²>1±=.¯;,9*9,®:-±<2ÐYQÜbìppôvyîmrõq|ör}ÿzÿ&#x7f;ÿxôhq÷epÿq|ÿu&#x7f;ÿkvø^jöOaÿA^÷8WéG^Ûntº©n}fgq¡©´ºÈÍÕÜ½ÌÙ»ÁÔ´²Ì©§Â¼¶²°³µ&#x7f;¹ÀÂ¡Ã¨Å²È¡¾ÎªËÑµÚÙÄæßÍêÞÒéÖÏáÌÉÛÆÃÔÁ»Î½µÈ·°ZM=ZM=ZM=ZM=ZM=ZM=ZM=ZM=XK;XK;YL<YL<YL<ZM=ZM=ZL?ZLAZLAM@N=M=^O<\O<\O>YM=YLCXMIXNOQYUb`Yia\rjfokspvq§{u³}ÁÉÇ&#x7f;À~¼y²r¨s¨t«l£\GN¤B9¶@<ÍC@Þ@?è:;õ=?ÿDE÷><ô>=í?@ê@CëBIêCMëANãALÑ??Å@9¾?6·>3°</²<0ÀD8ÒMDÈ72Ô:8Í2.Ì3.ÖC<É?4µ6'»>,»6%Á:'Ã<)Â;(À9%Â9&Æ;(Ç:(Ë:'Ð9(Ö6(Ø2$Ú."á0&ì7.õ=5õ82÷40÷0-ø**û&(û%'û%'û$)û&,÷!)ø".ÿ'7ÿ(>ÿ!;ÿ&#x19;7ý&#x17;1ÿ'8ï#.æ#+ã+-Ý--Ô+(Ó.*Ü41â64è64ï77ò65ó23ï./î,,å/,Ü92Ñ:1Ò91ìSK¼#&#x1b;Ä+#ßF@È.&Í.*Þ<7èE@â:7Ø/*Ú.*â51å<7ß>4Õ;/Ï5)Ë3&Ì5*Í;.Ë;0Å8.ìbXàZOÈC:¹6,º:/¸8-±3'´6*±1&°2&®2&°5&°6'²:*´<.³=1°:0ºC;Ö\Wìppñsvöx|ûyõq|ÿ|ýwút}üs{ÿr~ÿr~ÿmyÿgtü`nôI\ÿ>\ÿ>ä@XÔelµ |m}f~cjw¦ §§£&#x7f;£|vzqyo|r v¤uw³y¸&#x7f;¼¾¾¾¢½§¼±¼¾Â§ÏÌ¹ßÓÇãÒÊÜÇÄÔ¿¼Ìº¶Á²«¹ª¥ZM=ZM=ZM=ZM=ZM=ZM=ZM=ZM=XK;XK;YL<YL<YL<ZM=ZM=ZM=ZL?ZL?M>M=N;^O<\O<\O>P@OCNFYNJZPQ\SX_U^_Wfd\tibngqjxp¬y¼&#x7f;ÄÄ¿~ºt«n¢q¤¡u¨kW}<Cª5+²6,Â<3Ó@8Ý>8è?:óA=õ=;õ;<ô;@ñ>Dî@IèAIâ@KÛBGÇ<9½=4¶<1±;/«:,®:-½A5ÐJAÚFBÜA=Ô63Ð51ÖA:ÍC8¼:,²5#½8'Â;(Ä=*Ä=*Ä;(Æ;(È;)Ë:)Ï9(Ï6&Ô2%Ú2%ã5*ê8.ð91ð91ç1&è0&ê.%ï,&ö+'û+)þ,+ÿ-.ü)/ú$,ú$0ÿ'7ÿ$;ÿ&#x1d;8ÿ&#x14;3ü&#x13;0õ&#x1e;0ö.9õ3<æ+2Û(,Ü.0Þ02Û+.ñ:>ô7=÷4:ö26÷05ø03ù25ô87Û2-Ú;5×82äE?äEAÎ/+Ñ2.Á"&#x1e;Ô51á?<èE@à=8Ø3-Ù2,Ý60Ý:1Ô:.Ï9*É5'Ç5(È9+Ê<0Ê<2Æ9/ícYáXNÉ@8º4+Â91Ã:0¿6,Â9/¿5+¿5+½4*»5)¸6)µ5(²7(°6)´;0¼C:ÓYTèljïqtöx|þ|øvþ|üyùv~þu&#x7f;ÿs&#x7f;ÿp}ÿkwüdqý_nöIÿ>\ÿ>\êDZÓdj®yqzdu\x&#x7f;ckppmr`mig^lbqeyizk¬t±v·|»}¼&#x7f;¼¼»µ¢²¶¾½¨ÏÆ·ØÇ¿ÖÁ¼Ï¼¸Æ·´»ª²¤£\L<\L<\L<\L<\L<\L<\L<\L<ZJ:ZJ:K;K;K;\L<\L<\L=\L=\L=^L>M=N;^O<^O<\O<_RAQA\PDNFZOKPNQSRZ_SgcXvhkbsk¤&#x7f;vµ|¾}»¿·v©p®u¢ºv¥µi°Rv¸<D¾5/º4+À<0ËC7ÓE9ÛC8ä?9ñ=<÷:>ù:Aö=Eï@GæAHÙ@EÏ@BÀ;6¸90±8-9,ª9)9*»?3ÏF>ÙD>Ø:7Ù75Ú;7ÜC=ÞPFÎH<µ3#¾9(Á:'Ä;)Æ=+É=,È<+É:)Ì9)Ï7)Ð3$Ó/#à5+î=5õ@9ò:2é4+à4&ß5&â2%å/$ì-%î+%ñ+(ó,)ó+-ð(+ò&/ø(6þ%:ý&#x1d;7ü&#x16;3ø&#x18;3ô%7ò0;ó5Aò8Cñ<Eð<Eê6?ä,6ð2<ó/;ö.9ö-7õ.3õ.1ó/1í42Ü0,â=9Ú64ß;9ÿmkåCAÒ0.Î/,Ú;8ß@<àA;Û<6Õ7.Ö5-Ö5+Ò6*È6'Ä7&Â6'Ä8)Å;.Æ<1Æ<1Å;1ë`YáTMÉ<3¿0(Ë;3Ñ>6Î;3Ò=6Ø=8×<7Õ>7Ð=3Ê<2Â:.»7*¶6)¹;/¼@6ÏTMãgeíorøz~ÿû|û|ûyûxÿwÿsÿn}ÿgvû`pümõG^ÿ<Zÿ>\ìFZÎ`c~gyiQhOmRtX|^}`x\qXazP_&#x7f;PaUfZnaug} j~£m¬r²t¶x¹y¹y¹zº~º±&#x7f;¬©©ºµ¢Ç¹®Ë¸²È¶´¸¬¬ª¡¢ K;K;K;K;K;K;K;K;ZJ:ZJ:K;K;K;\L<\L<\L<K=K=^M=^M=^M;^O<^O<^O<^Q@^Q@OBOC\NE\OGPJOO^P_bTmdY{i^pg}u®{·{³¸¥´¯ª¸z¡ÆxÑuÊdÆMlÜFRßA@ÕA=ÏE;ÌH;ÍG;ÓE9ÞA:î@Aõ<Aú=Dø?GñBIãBGÒ@AÆ?<¿?6¶<1°:.¬;-ª9)¯9+¼>2ÎD:Õ<7Õ31à;9åA?áD?èSLßUJ¿:+Â:*¿8%À7%Ä;)É=,Ê=,Ë9*Î8)Î5'Ò2&Ù2)å9/ñ>7õ>8ð50ä/&à5+Þ6)à4*ã1'ç.)ê-)í+)í++ê),è',ë'1ò)9÷'=ø"<ü&#x1f;>û%Aû6Hä(6ç.<ÿP\ÿ`lÿP\ò<Ió9Gè)8í*8ò.:ô0:ñ27ê01á+*Ø)&Ø0-Ú72Ô2/åC@ÿspõVSÝ==åGFÛ=<Ú<9Õ:5Ò80Ñ7-Ð6*Í4&Æ3#¾5#¹6$»8&¿;,À</¿9.À:/Å<4åXQÜMGÊ70Ç0)×<7ÞA:Û<6ß<5ç>9é=9æ?9á@8Ù?5Ï<2È:.Á9-»9,»<3ÌOIàc_ìnoú|&#x7f;ÿý~øyøyüyÿxÿtÿn}ýetù^n÷VhóE\ý8Vÿ<ZîH\ÇYZmScrGuC\yCdzIm~QtVvXsYoWmZl\obuh{£n¨qªq«pq±t´u·v·v·v·x¸|´~«{¤~¤«¾²¤È·¯Ç¸µ´«°¡¥K;K;K;K;K;K;K;K;ZJ:ZJ:K;K;K;\L<\L<\L<K=^K=_L=^M=^M;_N<^O<^O<^O<^N>^N>\O?_OBOB^PE^OJbNYcRecTqdZ}lexr¢&#x7f;w¨v¤}§¬«Â©ÎÚwálÛXwÚD_ôCUûEQïJPáIHÑF?ËE<ÐE>ÚEAéCCð@Cõ@GõBHíDIßCDÏA?Ã@8¿@7µ=/®:+«:*ª9)®8*»;.Ì?6áFBâ=;îBBëAAÞ;6äKCæXLÕK>ÉA3À8(¼3#Á8(È<-É:*Ë7)Ð8+Î0%×3)à8/é;4í:5í62ë/-å,)ã0,ã1-ç10ì31ó25ö37÷48÷6;ì-5ê,6í.=ñ0Có-Fö)Hý+Nÿ3Rô9Lå3?ð@Mÿ_lÿkyÿWfõCSõ>Pî3Dí3Aí3@ê7=ã99Û83Î4*Ç/$Ë2*Ç.(Ð72òYTö\ZòXVÛAAâHFÜB@Ô<7Î70Í6-Ð9.Ï9+Ê7'Â5#¹8%³8&µ:(¹>.¹;-¶6+¼7.Ç>6ÛNGÙEAÎ50Ð1-á>9ç@:ã81ç51ë20î21ë52ç83á:2Ø:/Ð9.È:.¾5+º7-ÈIBÜ_Yënlû}ÿý&#x7f;õv}öw~üyÿxÿsÿk}ýdvù^pôRgôF_ý8Vþ=ZïLÀTR{`CWh<Yq?vCfyKn~QtVxZy_wayexhz m|§qªr«r«p©or°t³tµtµt·r·v¸w¶{¬x£x¤À¹§ÎÀµÒÄÁ¼¹À¨ª¶¨K;K;K;K;K;K;K;K;ZJ:ZJ:K;K;K;\L<\L<L<^K=^K=_L=_L=^M;_N<^O<^O<N;N;^O<_O?`P@`P@aQAbPFbNPcN_Pe`Usgato{t~rv®Ì¡ÛærëduéPdé;Rù3Lÿ<QûERéDKÕ?@Î@>ÑCAÖEBãEFèBDîAEíCFçDGÚDCËB<À@5»?3±;-©8(©:)ª9)°8*¼:-Ê;3èIEê@@ë=>å78Ú2/ßB;îZPôhÛQDÉA3¾4'Â6'È:,È9+Ë7+Ò8,Ñ0&Ú6-æ;4é;4ê40è/,ê-+ë--ó49õ3;ø3=ü3=ÿ1>ÿ1>ý1=ù1>õ0Aò1Dô4Kõ5Nõ/Pö-Qÿ2Yÿ>_æ3HùO\ÿanÿ_nÿWfûP`ôDXé8JóAQê:GÝ2;Ò/2Ê2-Â8-½;+½;+Ä:/À2(ÙJBúkcÝLGáPKÏ;9Ï;9àLJÔ@<Ë81Ê7-Î</Î=,Ë:)Ä;(´;&®;&¯<)³>-±9+¯3'º:1ÊC=ÖGAÙD>Ô72×2.ç;7ì:6è2.í2-ò,-ô,,ó/0ï31è71â:1Ú<1Ò>2Ã6,º4)ÅB:ØYSçjhú|}ÿþõv}÷x&#x7f;ýzÿxÿqÿi{ûbtøoòPe÷Ibÿ:Xþ@\ðO_ºQNqX:Sf8uEa|InSvY{~`gll}m|¤o}¨p©oªn©l©l«q®q°r³r³oµp·r¸tµv¬t¥w¨°´ÈÃ¯ØÌÀÛÑÏÂÂÌ«±Á¡±ZK8ZK8ZK8ZK8ZK8ZK8ZK8ZK8ZK8ZK8L9L9L9\M:\M:L:^K<`J<_L=_L=^M=_N>^O<^O<^O<_P=aP>aP>aP>aP>aP>bOAeMMdMW`O_aUmgbuq|vsuÎß~îqyùclúPú;Mÿ*Fÿ2Qÿ<Sñ<Mà<E×@EÕCDÔBCÜCEàBCäADãCEÞDBÔC>É@8¾>3¸>1¯9+¨7'©:)<*´<,Á=1Ï?7à>;å99ã03â/2à42Û94çNFülaòhÙQCÆ</Æ:-È:.É7*Í6+Õ8/Õ4,Þ7/æ93é73é1/è/-î1/õ15ÿ2Cÿ0Eÿ-Fÿ*Bÿ&Aÿ#=ü :ö :÷&Cõ*Gõ.Mõ/Pó+Rô)Tÿ2`ÿAhóCXÿ`mÿhuøVeíKZñL\ëBUÚ2Cë@RÝ7CÉ,3º((±.$«8&¥?&¨@'¸?.¼:,ÝYMáQÍG>ÍD>Á63É;7äUQÕF@É91È8-Ê;-É:)È9(Ä=)±>)¦<&¨;'¬=*«7(¬3(¼=4ÓLFÖGAÝE@Û96Ü30é73í62ë0+ô1-ý,/ÿ+-ú./ô1/í4/æ81ß;1Ø>2É9.½3(À;2ÏNHáa^÷yzÿÿùzû|ÿ{ÿxÿn~þew÷^põZlëL`õIaü9Wû=YîM³KHhQ1Qd6\tDf~Nv&#x7f;chjns¤t¦t&#x7f;§r}¨p&#x7f;¨n©m©l©l¨n«no¯o°l²k¶n·r³tt§w¬²¹ÈÆ¯×Î¿ÙÒÌ»¿Ê¤¬¿J8ZK8J8ZK8J8ZK8J8ZK8J8ZK8\K9L9\K9\M:L:L:^K<^K<_L=^M=^M=^N>^N>\O<aR?aR?aR?bQ?bQ?aP>aN=bL?hOKgNRbP\cWkjf&#x7f;wv}|xz«ÆÚ}ðuzýklÿYýDLÿ0Jÿ5Tÿ<Uô=Qé?LãCMÝBHÖ=@ÛACÜ@CÞ@AÛA?ØC?ÐA9Ç>4½=0µ?1¬;+©8(<,´?.»@1ÊA7ØC=Ù74á85ß//â30ç;7Ó0'Ñ5)ëUGÿvlëPÑC7Ê<0Ë;0Ê8+Í6+Ô:0Ú70ß82å63ç32ê01í12ô36ý4<ÿ&7ÿ#9ÿ"9ÿ"9ÿ!8ÿ&#x1f;8þ&#x1e;6ø&#x1e;6ï&#x1b;4î&#x1f;9ð%Bð)Hî&Jñ'Mÿ3^ÿEhÿbuÿ`l÷WcñQòPïMZâ@MÔ2=Ú=FÌ7;»/.®/(¦5'¢;(@'>%ª9'²8+ÒVJ»;2Ä?8¾41½2/ÎC@åWSÕHAÈ91Ä6,Æ7)Ä5%Æ5$Ã:(±<(§:&¨7%9*¬4&¯1%Â?5ÛRJÚIDáHCÞ;6Þ2.ê41í1/ï-+ü22þ,/ý+.ù+-ó++í-*æ1*ß4,Ö8,Ñ=1¿2(¼7.ÈIBÙ\Xówwÿÿûþ&#x7f;ÿ}ÿvÿj|þ_s÷XlöTiëEõD^õ6Uñ:VàLZ¥HC^K-Mc5Wo?b}Ju\gln r¥v¬z«wªr~§m|¥i}¥g¥h¦i¥k¨m«m¬l&#x7f;®h~°i³j¶o¶s¯r¬x°»¿Å©ËÊµËÍÂ¶½Åªº§K7\K7K7\K7K7\K7K7\K7^L8L8^L8L8^L8L8^L8L:J<K=^L>M>\L=N=P?^SA`SB_RA`P@bQAcRBdQBdNAdLBjMIhKMdPYflnkx{²©¥©¾ Ðây~ðsqújbÿa^ùJWñ>Që:Lî?PïCQê@Mç=Hå>Fâ=Cß<AÚ<=Ö<:Ï>9Ê=4Á;0¹;-¨7'¯A0<,¬6(¿A5ÌF=Í>6Ð72Ú85Ü71Þ5.Þ6-Þ8*Ý<*Ù:'Ó6%Ù;2ðSLòYQÙE9Ç8*Í>.ÔB3Ò:-Ô3+ðGBá//î5:ó6=î,7ÿBMñ'3ú&2ÿ)2ÿ(2ù&#x1d;'ý!+ÿ'0ÿ(0ñ&#x1e;'ð&#x1e;)ù+7ê .ì&7ñ-Cç#;ÿXsè.Eÿlÿ_iþZcùU\õPWëHMÝ?@Î95Ë>7ÄA7·?1¦7&3"6#;'<)§8-¯80ÂE?¿;7½31Å75Å54ÔDCÛLHÚKEÑD;È:0Ä4)Ç5(Ì5*Å5*¼</³9,/#«(&#x1e;³-$Ã9/ÒC;ØH?äOHßE=Ú70Ý0*ç/-ò21ú24ý14÷/2ö/2ö01ó0.ò0.î3.æ5-Ý9/Õ=0¾2%À>1¹=3Ö_Yésqÿûûÿÿxÿh}ÿ_wÿuÿUoøLføEcñ:Yô;æ?YÛi@;VF-L^6ZuBhNw}c&#x7f;g¡j¦p§q§n&#x7f;¦m~¥l}¥i}£h&#x7f;£f&#x7f;£f&#x7f;£f~¤k&#x7f;¦mªn&#x7f;¬k~g&#x7f;®f±g³iµm´r°v®|³¨¼±Á¤²Á®¦²°zK7K7K7K7K7K7K7K7^L8^L8^L8^L8^L8^L8^L8^K:K=^L>`N@^N?N>N>\P@^RB_SCQCOB_NDaOEbPFcOHeNHiMJgNQeS_h_tony~¨»ÁÊ£»«&#x7f;£¹xÓyê|{ôvjöi`øaóOZìFRòHUùJYöFSï?Jë;Eä9Aà:>Ú::Õ=8Î?7È?5À</·<,«<+ª?-ª9)²:,ÁA6Æ=5È72Ô<7Ò50Ö5-×5*Ù5)Ú7&Û:&Û;%Ú9'Ý6.æA;ëMDâJ=Ñ?0Ç8(Ë9*Ô</Þ;4öJFå12ð6;ö6Añ.<ÿ@Nï'4ó$,÷%(û),ý+.ý+.ø((ö(*÷+,ï$'í%(ë&-þ<EÙ&#x1a;)ð4CüATØ&6ÿjuÿbjþZaüWôOSåBCØ=9Ñ>6¾8,¿D5¹H6©>,/&#x1f;)&#x1a;(&#x19;(&#x1b;-%±;7ÑTRÎHGÃ54Ç56Ñ==ëWUØGBÖGAÐA9Ç9/Æ3)Ê6,Ð7/Ï81Ç=3Â91Â5.Ä3.É4.Ñ83Ø?9ßF@äJBÞA8Ú6-Ý0*ç/-ò12û03û03ô04ó12ô02ó/0ó0.ï2.ç4-ß9-×=1¿3$¾</·=2Ó^Wévsÿùøÿÿtÿf|ÿ\vÿXtÿQoþIhýAbó:Zñ?_âJaÁS\|@8SG1M_9vCgKsYx_wb{¡f}¤k}¦l|¥i|¥i|¤h{£g{¡d}¡d~ c~¢e|£j}¦n}©l}ªi|«e}c&#x7f;¯c°e°h²k¯p«s«|¬ª¢xdppR^^L:L:L:L:L:L:L:L:^M;^M;^M;^M;^M;^M;^M;^M;^K<_N>aP@`P@P?\O?\P@RC\RFQG\OG\OI_PMcQOfTTgUUhVVhWhnmhsv£|ºËÔ× Êª´¹}Ïäî&#x7f;vósjüfh÷S\íGQðFQöGT÷EQò@Lî=Gå8>à7:Ù99Ô<7Í@7ÅA4¾?0¶>->-ª<+«:,¸B6ÇH?À;2¼1*Ê:2Í6/Ð6.Õ7,Ø8,Ù7(Ù8&Ü9&à:*Ý4-Û4.âA9éOCÞH:È5%É3$Ù@2â>5õHDè13ò5;ù7@ø4@ÿESø2?ð)0ë#&ì$'ù13õ-/ï''ì&'õ12ë)*ö8:è+1â)1ì4@ÿR^ï=Mÿ\iþ^fñV\îOTñNSêEIÝ:;Õ74Ô?9À6,Á?2¼B5³=1¬7-¥3)¡,%' %&#x1e;ª0+ÇECÂ::½--È35×ABô^Ñ@;ÒC=ÐA9Ë=3Ì9/Ô=4ÛA9Û@;Û@<Ø=;Ý==ã@Aä>>à::ã?=ìJGàA;Ý<4Ù5,Ý2+ç1-ð31õ12ø02ô04ô02ô02ó/0ó0.ï2.ç4-ß9-Ø>2À4%º8+µ;0ÐTízwÿûúý~ÿrÿdyÿtÿTqÿNmÿFgû=aô<^ïFcßTg¢AHp;3SH2Qa<^yFhNrZu^t_xc{£g|¤hy¢fy¢fy¡cx byb{b}b| cz¡h{¤j{§h{¨e{©a{©`}«`&#x7f;b²jµn²s¬t¦wuynpmTdYCOM4@>^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^M;^L8`N:aP<`Q>P?ZN>YOCXPEXOHQO_UTbXYeX_iYcl\fk^glaimeroktty~µ&#x7f;ÊÚá×ÓÄ³©½yÌ{ÞzòwÿsrÿcgùU\ôMTõGQôCMò>Iï;Då6;à88Û97Ô<7Í@6ÆB5¼A1·?.«:*>-±=0¼F:ÍSHÇH?º5.º1)È91Í81Ô<1Ù=1Û;-Ù6'Ü6&à8+à5-Û2+á>5ìNBäL>Ò:,Ñ8*àB6ß7.ð>:é/0ð16÷4<ü8BÿNZÿLUý>Eò38â$&í/1ê,,ð22è**ë/.ò::Ò&#x1b;&#x1d;æ37Ý,2è8BðENÿXcòNWØ=CÛBEáBGá>Cß:>Ý8<Ø88Ó97É83Â91½:2ÀA:ÍNHÖVSÐNNÃCB²72¶95ÏKIÙMLäRSïYZëSRñYVÉ51Î>6ÒB9Ð@7Ó?5ÙB9àC<ãA>é?@ç8=ì9?ò?Cð;@è58ì>?öMJÞ93Û60Ú3+Þ3,å4.í41ò21ô01ô02ö/2ö/2õ/0õ/.ñ1.é4-à8-Ù?3Â6'·5(´:/ËVOï|yÿüýþzýo&#x7f;ücwÿXqÿOlÿFhÿ@cù9^ô?`éKdØZh35g;0WO8XhDa{KjPsYt\u^wby¡ez¢fw dv awav`xaxa{a| cz¢fz£g|¦fz§dz¨`z¨|«~_²g´n³ts¦uqtiidRbUFSL:G@_N<_N<_N<_N<_N<_N<_N<_N<^M;^M;^M;^M;^M;^M;^M;^M9K5_M5`P9_P;\O?ZNBXOFXQKYPQ`W\g`hnfsshysg{sg}ri~pm~pqsvx|«|ÁÒàæãÚ ¿u l¬nq¼vlÒviòsmÿllÿgiþ^`ûRYõFMï;Dê7=ç6<â89Ý98Ö=7ÎA7ÆB5½B0¸@/7)µA4°;1°;1ÈRHÕ\SÉME»;2¾90Á7-Ì9/×?4Û=1Ú6*Ü4'â6*à3,â70ä=5åE9âF9ÞB3ÞA2â>2Û0&ê72ì0/ï.1ñ.4ú7?ÿPXÿdÿZ_ÿOSá.1ã03à,/÷DGè66Ü,,Û-.ë>@Ý36Ù37ÿbhà<CË*2Ê-4Á&*Õ;=á@EÛ6:Ù37ß9=Ü7;Í/0Ì43Í;;ÕGFÞRSì`cõhnôekêcÄ@>ÆE@å^øhgÿopÿjjåKK×=;É40Ñ>7ØE=ÖC;Õ>5Õ<4Ø93Ü43ï6<ñ-9ð,6ò.8ð,6ë,3ó9>þJKß3/Ü3.Ú3+Ü3,ã5.ê40ï4/ò21ô02ö/2÷/2ö.0õ/.ñ1.ê3-â7-Ù?3Æ8*´2%²8-ÃNGï|yÿÿÿ&#x7f;ÿwúl|øatûTnýHgÿ>bÿ:_ú6\ôBbàNeÂWaq0,_?0W>^lIf~NkSt\u_v_xcy¡ex dv`v`u_v_w_x`{a| bz¢d{¥e|¦d|§by§^y§\{ª\}¬^¬a¯g®mªp¥s r{nrjfwek^RbU_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N<_N:^L4_N4`N8^O<N>YOEZSMZTT_Zahbprmzu}t{qzpvpqtrxu|¨y¶{Ä&#x7f;ÏÚÞÜÍ¶¢£ª«~ylR?¼MBÔQIâWRëWUñSTñKOí>Cæ5;é9<æ9;à;9Ù>9Ó@8ÊB6ÁB1º?/´:-ºA6«5+£.$¾I?Ùg\ÙdZËUK¹=3¸5+À3)Î:0Ù;0Ú6,Þ3)ä6-ß1*å82ã;2Ý9/Þ>0ãF7äB5ß9-ß1(ê5.ð31ð./í*.ô17ÿEJÿVÿbÿ\aâ9<ã9<Õ+.ôJMâ89Ø00Û57Ì()Ì'+øW\Ý>CÖ:>Ò7=º&#x1f;#Å),Ü<>çBFÝ49Ù,2Þ17Ú16Í(.Ò37äKNöaeûhnùcl÷_kú`løbkÖLLÕNJê\ZêVTçMMãHFÏ42Î3/ÙA<àKDäOHßJCÙ@8×:3×50Ü.-õ1;û,<ø*7õ'4õ)4ø0:ÿ>CÿJLà1.Þ3,Ü3,Ý5,á4-ç4/í4/ñ42ô02÷/2ù.2ø-0ö..ò0.ë2-â7-Ù=1É;-´2%±7,»D>ìyvÿÿÿ|ÿqöhxö_rùRlùDcÿ9^ÿ6ú8ñGdÑN`¢IMe5+VD0Y@^lIe|NkRtw_x`{dz ev`ss\s\tv^w_{az¡bz¡b{¤b|§b{¦_x¦x§Yz©Y{ªZ~ª_e¬k©m§s¥uu|rvpn&#x7f;lfwe_O?_O?_O?_O?_O?_O?_O?_O?^N>^N>^N>^N>^N>^N>^N>_N:aO7aP6aO9^O<OCRK\VV_Zakgvso}z~¡|¥}w£{u¥xv§sy©s}°z¸}¿ÄÉÎÍÇ ½«µ¼¦²É©Å¦ªmfJE2¯A4·A5ÃC:×IEéOOîJKì?Cë=?é;=ä<;ß=:Ù@:Ð@7È>1Á=0½:2¼@8¯4,§1'¼J@ÑbWÖg\ÕeYÃOBº>2º4)Ç7.Ö90Ù5,ß2+ç60ã1-ä50â7/ß8/à>1æD7ä>0Þ4'æ4*é0+ð3/ð..ï-.ò14ó78öBEöMRüY^éFKöQWÙ48ëFJÝ8<æAEÚ7:Ï/1Ñ05äHKÓ8<Ð7:¹#%Õ:>Û8;å;>ë<Aè7=â-6Ý)2á1;ç<DðLSýeÿksÿjsú`löZhùXh÷hôddèZVèWTØ@=Ï10Ô44Ó33áC@ãJEéRKêSLáJAÙ?7Ø93Û81ã33ó.8ÿ0>ÿ0>ú*8ü.:ÿ;EÿCHûADã1/ß4-Ý4-Ý5,â4-æ3,í4/ò53ô02÷/2ù.2ø-0ö..ò0.ë2-ã7-Ø:/Í>0µ1%±7,²;5ésqÿÿÿvþi}óatóp÷PjúCbÿ8`ÿ8aù;aìLh¿O~86=0NF1VX@ZeEcxMjRrv^ya{dxatpZpYqZqXt\uy_x`y _y¢^z¥^y¤\x¤Yx¤Wy¥V{§X&#x7f;©_¬d«h¨l¥o¢qoxjril~ddv^_O?_O?_O?_O?_O?_O?_O?_O?^N>^N>^N>^N>^N>^N>^N>_N<cQ9cQ9aP<P@ZPG\TR_Zaa_mtrzy¨°&#x7f;}¯{x¯zv±wwµw|¾xÁ}ÄÄÁ¿¾ ¼¬§»¶¬µ¾³±È¸«Ð¾¦ÖÁ¢Ñ»Ð¬¯oT¯H¥H6¨>0ÂG?àWQïWVñNOé@Cê=?æ<=ã><ß=:Ø=8Ñ:1É90Á80À<7¹:3·>5ÂPFÇXMÇPÑeXÓcUÃM?¼9/Å8.Ò91×4-ß2,ç60æ40â3.à5.ã<3çC7æB6â:-ß3'ê5,æ.&ð0-ð0-ð31ñ54ç-.ã35çBF÷XöW\ÿouåDJëHMá=Dþ`ÿouß>Cÿ}ÿmqêQTÔ>@Æ02Ö:=óILî9@ñ8@ø=Fò7Bè-8ô>Jÿhÿmwÿjrÿcn÷gú^kÿapÿnùYeò\ãOKâJGÕ74Ó01Ù54Ò0.Ø95ßD?ãJDãJBÙ@8Ó6/Ö4/Ü71é77ê'/ü0<þ2>ø,7ù0:ÿ>Eý>Cì44â3.à5.Þ6-ß4,á3*è3,ï4/ô43ö/2÷/2ù.2ø-0ø..ó0.ë2-ã7-×9.Ñ?2¶0%²6,°61çonÿÿÿpücxñ\pôXn÷OiûBbÿ9aÿ<döAdäRi°QWd/'WD3GG/RT<U^?cuMiQqv_zb|dx_rZnWnUnUoVrZt\wwwx¡y£w£Xv¢Wv¢Uw£Ty¥Xz¤Z~¦_¦c&#x7f;¡d~f{fr_kZ`vRnPSfH\L<^N>`P@`P@_O?_O?`P@aQAbRB_O?M=^N>aQAcSCbRB`Q>dS?aP>P@^QIQPZSZeapsqyxz|¥~±¹~~ºyy·sr´nn¶twÈv{ÌÊÆÃ¤¤¾²°»¼¹´ËÆ³ÓÌ°ÙÏ¬ÛÏ§ÜÏ£ÝÐ£ÙËÙÂÑ¨¸g¡_IËzgÖtg½H?×QN÷ccéKLå?Aâ89ä:;â:9Ü75Ù53Ö;6Ë73À50ÇB;¹=3¿I?ÆXK¹OAÎdVÑcTÌXIÅG;Ä:/Í6-Ø7/á6/ã4/ß0-ß2,à70ä=4á=1Ý7+ß5(å7,å0'ê/(ì/+î1-î20ë40è51ã75×79æJNõYîOT×8=×6<ùX`ÿ~ÿ~ÿouÿgoÿipÿflõ\añXûZ_ÿU\ÿDLü8Bû7C÷3Aï,:ì0>ñ=IÿS_ÿ`iÿgrÿepþbmÿbpÿ`oýgíSSßFAÖ85Ö41×11×/.Ø31Û94áD=ÜB:Ù?7Ö<4×90Û81à93æ66ñ7<ô5<ö4<ö4<÷5=ò59ë35ã1/Þ3,Þ5.ß7.á6.ä3+é2,ï2,ö31÷03ø03ú/3ù.1ù//ô1/ì3.å7.à?5Ì8,¿6,¯/&¸;7ÒXWÿþ{ÿmÿe|ø\tðPjõHfüCeÿ;dù9`õNlÓRd=?R,!C;&CG.KM5Z`DZjEcxMqZxbx`t\rYqYlSmRnSoTrXtZuutXtWv Xw¡Wv¢Wv¢Uu¡Tt S{¥{£z^umUb~MXrCSj>AW0?U/<R,\L<^N>`P@aQA`P@_O?`P@aQAcSC`P@_O?`P@bRBcSCbRB`P@fVG_QFPH^TSaZad`omkvwxz£{}®}&#x7f;¸x|¹rv¶lo´mp·oq¾uvÐ|~ÕÓÎ¨§Ç¹¶ÁÉÅºÓÎ±Û×ªâÛ¥çÞ¥æÝ¢äÛ¤âØ£ÛÑ ÔÇÓ¼»©|eÅwÊ}m´VJÒaörpìZçKLã?@è>?ë>@è<<â66Û55Ö=8É83Ê?8¹6.¸>3¾LA¶H;ÈMÝo`Ô`QÆH:½5)Å1'Ô6-Ü5-Û0)â51å95ä=7à<3Ü8.Û5)ß4*ã5*ì7.í5-í2-ë0+ç0,ã2,á4.Ý52Ö87Ø>@åJNóWúdü\dþeÿaiôSïNVñQÿ_gÿjqÿflõZ`óNTí4:ñ,5ô*6÷-9ñ'5å!/æ'6í7DþO\ÿZdÿblÿdmÿenÿcmýhõQZåEEÞ<7Û64Þ63ß55Ü30Ú2/Û62à>9Û>7Ù<5×:3Û81ß82å95ç:6ë8;ë7:í6:ï6;ï79ì57å31ß2,Ý4-Ý6-ß7.â6,å2+ê1,ò1,õ20÷03÷03ú/3ù.1÷//ô1/ì3.å7.ß<3Ô=2Â8.¶1*·74ØZÿûs}ÿfzÿ^vùVq÷NkúGgü@cü<cõAdåOhºLWy<9L2#<:#<B(EG/PT9VdA`rJnXu_v_tZqWpTkOkPlQmRpVrXssYuYuWuWuWtTrQqRqQoPlNiLbFXu?Ni6F/@W+@W-AX.AX.YL<N>^QA^QA^QA^QA_RB`SCaTD`SC_RB`SCbUEbUE`SC^PCbUM\QMQRcYbkdtqmvuz|¥y{¬y~¶x|»ot¶ei°bf¯jn¸ruÆ}}×ÝÛ©§Ø»¹ÏÍÉÈÜØ¿åá´éäªíè¦ðé¥íå¤çá§âÛ§×Ï¡ËÃ¸¯©l¨p§o`RGºUÝmi÷uuï__äJLæAEëADì>@è8;ã68ß=;Ô;6ÔA:Ã81·7,·>3®=/·G9Ûj\ßiÚ\NÊB6Æ3)Ñ4+Ù6-Ù2,Þ52éA>ìGCà=6Õ2)Õ1'Û3(Þ0'ç4-è1+ç.)æ/)ã2,ã80á=4àB9Æ-(Ê43ÜCEðVXú^b÷ZaóSñQéGRêHSïOü\fÿmvÿntôY_ã>Dé.5ô-4û2<ü2>õ+9í'6ö6EÿKW÷ISõPWøW_þ^fÿckÿ`i÷RYéBIÞ97Þ71ã75é<8ì<<æ95á51Ý60Ý:3Ù;2×90×90Ü71â94è;7é<8ã99ã99æ87è88è88æ66á40Û2+Ú6-Û7-Þ7.â6,ç2+ë0)ó0,÷10õ13õ13ø03÷/1÷//ó1/î3.ç6.Þ:1ÜD9È91»4.µ/.ä_bÿølwþbxþWsúPmþLlÿBh÷:`ô<`ìIfÔSgDJm<5O>,=<':?(ED/FH0Q\<YkEg|Qrt\qWmRlPiMiKjMkPnToUpXqYuYuWtVrTpRmMkMjLfHaE\~AUt;Li3E`-AX*>U'D/E_2Ic6WJ:ZM=P@^QA^QA^QA_RBaTD`SC`SCaTDaTDaTDaTD_RBQEZOMRWbXcjcsqmvuxz£z|w|´sy·mr´dh¯d¬ag±lr¾x{Êßå¤¡ä·´áÇÄÙÕÑÎáÝÄçä¹ëæ¯îêîè¬éâ«ãÛªÚÓ©ËÂ¡»¶¥¥vqrcbW³i`ÏrmûõutêæLNéDHí@Dì;Aé;=ß:8Û<8ßF@ÓC;Ã:0º:/°6)§1#¿I;ÛaRêjÛSGÊ:/Ì4)Ô6+×4-Ù42ñMKþZXíKFÚ91×4+Û4+Þ2(ß.&â-&á,%Þ-%Ü1)Ú6,Ù;0×?4Â/'Ä4,Î:8ÙCBßFHàEIäGNêMVñQöVbøXdù\gÿgrÿmuôYaÞ;@ë28ñ.4ó.7ò*4ë%2í)5û<KÿS_í?IèAIêGLôQVÿY`ÿXôGMå79á53å84ë<9ò@>ô@?ï=;æ95à72Ý82Ú91Ù80Ù80Ü71ä84è96ê;8â:7à;9ã:7æ:8ç98å84à5.Ú3+Ú6,Ü8.Þ8,â6,ç2)ë0)ò/+ö0/ô23ô23÷03ö01ö0/ó1/î3.ç6.Ü8/àF<É91¿4/¶,,îfjÿyøgtÿbyýTqþKkÿGjÿ=eó6\ë=^ãQhÄXeBBgC7N;GF1AC-JH3EE-LU6Sc>btLlUpXoTmPiKiIhJhKjMlRnToXpXtXsUqSoQkNhKeHdG`DAWy=Rq8Mi6Lf6Oe7Of8Ne7Oi9Sm>UH8XK;N>\O?P@P@_RBaTD_RB`SCaTDaTD`SC`SC_RB^QIVMR_Xhnf{uqxtxw¡xz«y{´tx·lq³bf\c«`f°kq»v|ÆÎàä®ç½¼æËÉßÔÓØÝÛÎáßÆåá¾æâ¼äß¹à×¶ÚÐµÎÆ±¾µ¦§y|qvksi¨meºoiæî~ósrñedñWYðMNîDGì?Aã:7à;7äB=ÞE?Ì<3Æ=3ÄB5°2$®0"ÈJ<ÜXKÖN@Ì=/Î7,Ñ7+Ñ3*Ö42õUUÿkhúXãD>Ü;3ß;2ã80â4-ã2,â1+Ý0)Ô0'Í/$Æ/$Â0#Å9,À4'Á2*Ì;6×CAÛEGÛBGÚ@HàDOìPòVcòVaú^iÿgoóYaáBGê:<í57ï28ï28ò2=÷9CÿHUÿVaå7@Ý4;ß6;ëADùJOüIMõ<Bé13é54ë95ð=9ó=:ó;9ð:7é73ã81ß;2Ü;1Û:2Ü92ß82ä73è64é75æ:8ä;8æ:8è:9è;7æ95á6/Ü3,Û7-Ü8.ß9-á6,å3)é1)ñ0)ô1-ó23ñ33ô23ô01ô1/ó1/î3.ç6.Þ7/àC:Ê70À1-À13økqÿqzúftûtúPmÿBhÿ>gÿ9cõ8^æEdÙZm«VwE>dJ=aYFTO;KI4PI7GE0GP3M9ZlFf{PmSmRlMhJiIgHgJiLkRlSnWoWtXsUnRkOfKcHaF_FCY}@St;Pn8Ok:Rl=Xn@ZqCWo?Vq>WqARE5UH8XK;ZM=N>\O?^QA`SC`SCaTDaTDaTD`SC`SCbUEbWQ^Vckf}yu~zzy¡xy©vy°txµlq³cj°\c«^d®io¹w}ÇÑÓÝ£¥à°²ã¼¾åÆÇãÏÎÞÔÓØ×ÖÑÙ×ËÙÖÇØÐÅÒÉÀÌÃ¾Å¹¹´¨¬£yztyvqule&#x7f;lepha`ÀsmØyuð~}ü|yÿrqýccöSTðHHðDBê>:á<6áB<Ñ:3ÓC:ÞUKÂ>2¸6)½9,Â=.È?/ÐB4×C5Ö@1Ñ7-Ñ30êLMú_òWSßD?Ù:4Û81á81æ93ç92ç92â:1Û:0Ñ9,Ê8+Å9*±)&#x19;¸3"È@2ÓI>ÒD@Ë;:Ì7;Ï:@Ò8BâHRíUaðXdõ^g÷`iêUÜAEå@>ì?;ó@CûGJÿLTÿQYÿTÿV`á3<Ú/5Ý04è8;õ>Bø;?÷48ó04í42ï96ñ97î53ë20ê20ç40á4.ã<4à=4ß<5Þ;4á83ä73è43è43é99ç98è88ë99ë97é75â5/Þ3,Ü8.Ý9-Ý9-ß7*ä3)è1)ï0(ò1,ñ33ð43ó23ó11ó1/ò2/ì3.ç6.à91Ù<3Ì71Á0-Ñ@CÿnuýkuùbqðRiõHfÿ;cÿ5aÿ6bü?eäNiËjGFjF:^M=^WDYR?RJ7OF5ID0DJ.GU2Sc>^sHg&#x7f;MjOjKgGgFfGfIhKiNkRmVmUqUpRkOgKcH`E_F_G^E~DYwAVr?Uo@YpBrG^uGsCYtAXs@NB2QE5UI9WK;XL<ZN>\P@^RBaUEbVFbVFaUE`TDaUEfZJi^Zngwup|x{y wx¦vwpu¯jo¯ch¬`g¯`f°em¶qxÀ~ÍØØ Ö¡¨Öª±Û´¹ßº¿ßÀÁÝÃÃÛÇÆØÆÃÔÅÁÐÂºÉ¼³Ä»°Áµ¨¼¦¯qqquqppfewhenj_Yd_ZT¹f`Ùsoðzvÿ{yÿvuÿhhüZXþRNõHBæ=8èE>Ù<5ÛG=ê\RÅ;0Ä<0½7+¿7)Ç>.ÑC5ÔB3ÔA1ØB4Ò86Ø>>ÝCCÜB@Ö=7Õ81Ø7/Þ7/å:3æ93å:3â;3Û>5ÖB6ÑE8ÌG6®-&#x18;¼;%ÍJ8ÑL=Æ=5»0+À00Ê9>×AJâLWëU`ðZeðeìYaßLTÔ>@Ï4/Ø7/à;9ê@@òCHõDJñBIî?Fâ5;ß26ã36ï8<ö8:ø14û/2ü14ï20ð95ñ85ì0.è,*ç0,æ3/á4.ä=5á>5à=4à<3ã:5æ95ê65ì65ì57ì57í57î68î66ë54ä2.Þ1+Þ7.Ý9-Ý9-ß7*á3(æ1&í1(ñ1,ð43î53ñ33ò21ó1/ò2/ì3.ç6.â;3Ò4+Ò:5Æ20èVYÿpxùcnòZiçI`óEfÿ8bÿ0^ÿ5bÿBhÜLe±PYk60^G7XK:UN;O?VI9M@0JC0@F,AO.L\7Xj@bxGfJhIfFeDcBdFeHgLhOjSkSnRmQhNeKbHaG`H`HaH_H_}IyHwH`wKcxOczNuCvAvAJC3MF6RI:TK<WK=YM?\O?^QAdWFeXGdXHbVF_UIbXNg^Ulcfwrzvzutsss¥tulp¯bg©\`©ae¯fl¸ntÀx{ÈÓÝÞ¡Õ §Ó©Ú®³Ý±´Ý²´Û´´Ú¶´Ûµ±Ö³Ñ®§É«¡ÄªÂ¥½¯~|xdbgiefdZjX{c_yVPcNF¢UM½`YÖkeñwrÿ&#x7f;{ÿzwÿnjþa\÷TOéGDîOKÝB=ÞG@êWOº,"¾1'¾4'È?/ÑE6Ë>-Á2"Ê7'ÛE7ÞE?Ó97Ì3.Î61Ô;5×=5Ý<4á=4ä;4ä92Þ5.×4+Ï5)Ê8+Å<,Â?-ÕT?ÈG2½<)¿;.Ã=2Ä;5Â74Â35ÜKPÞMRáPUåTYèYåVXÜMQÓCBÉ5+Ñ7+Ù80Ý52â24å26æ18ç/9ì4>ì3;ó5?û9Bý4>ú+5ÿ(3ÿ0:ø-3ù59ù59ð./ë*+ê1/è43â51ã<6à=4à=4á=4ä;6ç:6ì87ï77ì46ë35ì25í34í55ê41ã1-Û0(Þ8,Ý:+Ý:+Þ8(à4&ã3$é3&ì4,î52î53ð42ð3/ò2/ð3-ì3.ç6.ä=4Í/&×>9Ì35ùbgÿoy÷iðNcêCùEhÿ:eÿ0ÿ6aöFjÉJBHS-$WJ9SL:OH6O?UI9F?/HF1>B'@J(IU1Sd:_rEe}KgJeHcCaBcDcEeJgMhPiQlPkMfJcIaGbFbHcJ_G^&#x7f;H_}I`|Ka{Kd{Mf|Nf~NxCz@^{CDB6FD8KG<PG>RH>XJ?M>`P@`Q>aR?`SB_VGcZSgb_ojnso~wqzt ~x¨yw©pp¦gg£`a¢\_¤`bfhµopÀyzÌÖßçæ Ý¡¡×££Ù¤¤Ú¦£Ø¦¡×¥¡Ô¦Ó¥Ñ¤Í¢É¢ÈÃ»°|rn}igld^`gkYUrWP}WNWORI©QµXPÁZSØkdìyrù|ÿùuqÿusí_àQMÙECïZTÌ71Å2*Å7+Ã7(À7'Ã7&Æ9(Ê;+Ñ=/Õ?1Ò8.Ö<2Ø@5Ô<1Í5(Î4(Ø:/ãA6à90á90Þ7.×7+Ï7)É:*Æ?,ÄA/Â=.Æ@4ËE:ÍG<ÌC;Å<4¼3-¶+&Ç<9ÐEBÚPMàVSãYVáXRØNKÐC:Ì;*Ó:(Ü8,â5.è./ì*2ð(3ô'8ú->û,@û*?ý)?ÿ(?ÿ&?ÿ%=ÿ%=ÿ#:ÿ':ÿ+<þ,9ô,6í.5è45ç;9Ý84Ý:3Þ;4à;5ã:5æ95é73ë54ë35ì46ë54é54ç53â5/Ý5,Û5)Ü8,Û9*Ü9(Þ9&ß8&â7%ã6%æ4&ì5/í41ð50ñ4.ñ4.í5+ë6-å7,à90Ø7/Õ:6Û?BÿmwÿbsûSjõIcÿKjö6ÿ6_ÿ<gø>cçNj¯IV^$#E-!?:'HE4PM<PI9EB1>B1?G/BH&EK%KS.S8kDavKd~NdJcF`Ba@cBeFjJlPnRlLjIgGfHfJfIcHaGeKdIb&#x7f;Id&#x7f;Jg&#x7f;KgIe~Ga}C\|=\&#x7f;=_B=?4@B7FC<JE?PE?TG>K>^N>aP>`Q<^SA^VIcXhcinlysouovptp¢pm¢gfaa^¡¥ed°kjºusÆ}{ÏÙàæåÙÕÖ×ÖÕÔÑÑÍÊÇÄ»&#x7f;®xtk|mdif^fYiZSoWM{UJVKVK±S»ZSÁXRÓfaàqjãvoí~wýÿ÷uuîgdãUSëZUÇ2,Â/%Å7)Ã7&Á8&Á8&Ä8'È;*Ï;-Ó=/Õ9,Õ9,Ó:,Ò:,Ñ9+Ñ9+Ô8+Ø8,ß;1à90Ü8.Ö8,Ï9*È9(Á:&¾9&¾6*Á80Æ=5É@8ÊA9È?7Å<4Ã:2¿4-Ç>6ÒIAØOGÚQI×NFÌC;Ã7*Ë8&Ó8$Ý7)å4,í//ô+3ù)7þ);ÿ+Bÿ*Aÿ)Aÿ'Bÿ#@ÿ!<ÿ&#x1e;=ÿ&#x1e;;ÿ&#x1f;:ÿ ;ÿ%;ÿ);÷-9ð19ê7:å;;Ü94Ü:5Þ;4à;5ã:5æ95é73ë54ì46ì46ì65è64ä71ß6/Ü5,Ù5)Û9*Û9*Ü9(Ü9&Þ9&ß8%á8%ã6&ç5+ê3-ë4,ë4,ë5*ê5*ä7)â8+à<2Õ3.Ó54óTYÿesÿpöHcÿIhÿAdÿ;`þ4\ú<bòMmÍLb8?S*&A5)7:);<,B@1F@0@?-=B.?H-=C&#x1d;KP'\b<fpKhvRf{Rf}QeNaFaAa@c@gFjJkMmOmLjGhHgGgJgJfJeIfKdId&#x7f;Hd&#x7f;HgGgGf&#x7f;Eb~A^?_?bB7:/;=2@@8EB;KB=OE<VH=YL<^N>P?\SD^WMcifmom{sqsoqkjfc`\ZZ\ \_¦hjµno¿wxÊ&#x7f;ÒØÜàÝÓÐÐÐÑÐÏÎÍÇÄÀ¾³z¥r|ewu^fie\WdRgYLrVHTH£ZQ½`YÉ\YÐZXÛcbákißokãvqûÿÿÿ}y÷nhîaXÅ5,À2&Å9*Á:'À9&À9%Â9'Æ:)Ì:+Ð:,Ö:-Ô6*Î5'Ð8*Ô>/Ô>/Ï7)Ë/"Ú:.Ú8-Ù7,Ö8,Ð:+É:)Â9&¾7$¾6*À6,Â8.Å;1Ç=3Ê@6ËA7ÌB8Ç:1ËA7ÏE;ÐF<ÒH>ÑG=Ê@6Å7+Ì6%Ö6&ß7,è50ñ03ö-7û+9ÿ*<ü'=ü'=û%=ù#;û!:ú&#x1e;7ü&#x1c;7þ&#x1b;6ÿ&#x1e;8ÿ&#x1f;6ÿ$8ý*;÷2<ñ6=ë8;ä::Ü94Ü:5Þ;6à;7å95ç85é75ë54í36î47í55ê65æ72á6/Ü5,Ù5)Ú8)Ù9)Ú9'Û8%Ý8%Þ7$à7$á6%á5)â4)ã5*ã6(ä7)á7(ß7*Û7+Ý<2Ð1+Ø88ÿkrÿnþPiñ<ÿHlÿ7ÿ>eù5î>bæVqª?Qa#(L/+;7+18(37&:;)B=*A<(@>)BB&?F&#x1d;Xa4s~Tdw`iTb|Ma&#x7f;I`E`@a@dAhGjJkMlNnMkHiIgGhKhKhLgKfKeJcHdIfHfHeFbBcDeEhH25*69.;<4A>7G@:KB=SF>VJ>XL>YM?YQD^WOc^bigronrormjga_WWTUWX\_¢acln»rsÃ{|ÎÕÙÛÜ×ÒÎÏÎÏÐÏÎÊÆÁ¼¶¬x&#x7f;o_p|Y`lXYdT^_Q_\IjWF}SEUL¾\YÑZ\ÙX\äcgçkkáplåxsì|xízuþ~ÿÿ{õkaÉ<2Ä8)Â;(À;(¿;&¾:%Á:'Å9(È9)Î8*Ù;/Ô6*Î5'Ï9*Ñ>.Ñ>.Ë8(Ç/!Ò6)Ó5)Ó5)Ó7*Ñ9+Í<+È=*Ä;)Ä:-Ä:/Æ90Æ9/Ç:1È;1É<3Ë=3ÔE=ÓE;ÐA9Ë=3Ì=5ÏA7Ï@8Î;1Ï6(Ù5)ã6/ë31ò/5÷+6ú*:ü)<ù&9÷&;ø'<ø'<û&<ý%:ÿ$:ÿ#:ÿ#8þ#7û&8ù,;÷5>ò9?è8:à87Ü94Ü:5Þ;6à;7å95ç85é75ë54î47ï58î66ê65æ72á6/Ü5,×5*Ù9+Ù9)Ù9)Ú9'Û8'Ý8%Þ7%Þ7%Û5%Ü6(Ý7'Ü9(Ü9(Û:(Ù9)×:+×:1Ñ4/ìILÿoyÿVjùE`ø<_ÿAgÿ3\ÿ8bø>cèMlÅOe0=J&#x1d; C1-11'/7(5<,>A.B@+B<$E>$IF%U\0p~Mr¨|mjW_{J~E_CaAcBfCkJmMmOmOoNlKiIhHiLiLjNiMgLeJdIeHgIhIgHdDgHiIlL.4(36+891<;6B=9H@=MD=QG=SI=SK>UNDZUQa^eigumklkkib`XVPPQQXY`cªfj´orÁuxÉ|&#x7f;ÐÖÙØÙÖÏÌÍÐÑÑÑÐÏÈÁº°¥~w&#x7f;mYjSZnVTc\RZaOZ_HeZDxUBME¸TRÐSWÛRYå^dèejãklæuqäunâoh÷~vÿ{ÿ~ôl`É?2Á8(À;(¿<(¿='À<'Â;'Å:'È9)Í7)Ø:/Ù8.Ò8,Ï9+Ì:+É:*È9)É7(Ï9+Ñ7+Ò8,Ñ7+Ï7*Ì8*Ë<,Ë>-Ê<0É=0Ê<2É;/È:0Ç9-Å7-Å5*Ð@7Ð@5Í=4É9.Ë80Î;1Í:2Ì5,Ð/%Ù0)ä20ì25ô/8ù-9û+;ü+>ü-?ü/@ü/@ü/@þ/Aÿ/?ÿ.@ÿ.@ÿ)9ü)8ø,8õ0:ñ6=í8=ã77Ü43Ü86Ü:7Þ;6â:7å97ç77ê65ì44ñ48ñ48ð67í76æ74á6/Ú6,×5*Ø8*Ö9(Ö9(Ø9&Ø9&Ù8&Ú7&Ù8&Õ8%Õ:&Ö;'Õ=(Ô=(Ô=*Ô=,Ô<.Ñ7/Û<9ÿ^eÿaqÿPi÷<ÿAiÿ4_ÿ5aû1YôKlÛYq>NZ%-C),:20.1*08-8A0=B,<<"A<&#x1e;UH(`5|T¢nª¾¨¾¥woa&#x7f;K^&#x7f;F`DcBgDlGoNpPoSoSoOlKiIgGiLiLiOhNgKeIdHeIhJhHgGeFiKiKlN+1%.4*470894>:7B=:HA;KD<NG=NG=PKEWSR_^ffeugg&#x7f;gfbaYRRNORS^£ei³lo¾uwÊz|Ñ&#x7f;ÖØÙ×ÕÓËÈËÎÐÒ&#x7f;ÒÐÒËÁ¸«&#x7f;{v}k{VcPVnUQd^PeMYcJb^EsXCOE´WRÐUZÛTç\cèagågjèpoìyræqh÷|tósjþxmæ\QÄ;+¾5#Ã<)Á=(Á=(Ã<(Ä;(È;)Ë9*Ï7*×6,Ø7-Ô:.Î:,Ç:)Ä8'Ç;*Ê=,Ï=.Ö@2ÙA4Ö>1Î8*É5'Ê8)Ë<,É:,È:,É;/Ê<.Ê:/È9+È5+Ç5(Ê6,Ï;/Ò>4Ó?3Ö?6Õ>3Î7.É,#Ò+%Ü-*ç02ð39ö1;ú0>ý0Aþ1Bú1A÷1@÷1@ö0=÷/<÷-9ú,9ú*7ü0<ù/;ô0:ñ4;í6;æ69ß55Ú53Ü86Ý97ß:6â:7å97ç77ê65ì44ò59ò59ð67í76æ74à72Ú6-Ö6*Õ7+Ô8)Ô8)Ô9'Õ8'Õ8%×7'Ô9%Ð9$Î<%Ï=&Ï?'Î@(Î@*Ð?.Ó<1Ñ61ëHIÿfrÿOdÿHfû7ÿAmý/ÿ7aò4XäQk¼Ufm19E(,E697325426926=-5:#38&#x18;CC&#x1f;ja:T®w©ÁµÍ¯Ç®}vbeQ`F`BcBiFoJrOrRqUpToOlKhHfFhKiLjOiOgKfJeIfIiIjJiIgFjJkMnP(0#+1'14-561764<87@<9C>8IE<HE<KHCRPQ\Zecbtbb|``_^ZXTTTT\ gi³psÄvxÍ{&#x7f;Ø~ÛÞÝØÖÔÑËÈËÍÑÔÕÓÒÈ¾²¤|wpwfp~U|PQnUNf_O^fNZdIaaGqETFTÈZÔVZâaædéaeîllõzsírj÷xoÞYPåPÒD8Ã5'Ç:)Å9(Ä;(Å<)È=(Ê;*Î;+Ñ9,Ô6+Ó0'Õ4*Ò8,Í;,Ç;*Â;(À;(Ã<)È;*ÕC4áK=ÞH:Ó=/Ë5'Ê6(Í;,È9)È9)È9+È9)Ê8+Ë9*Ì8,Ë7)Í6+Ò<.×?4Ø@3ÛA7ÛA5×:1Ò1)Ü0.å14ï5:õ6>ù4>ú2?ú0@ù0@ò.<ï/<í/9í07ï-5ï,2ò+2ò+2ö3;ó4;ò5;í6:å57á55Þ65Ü75Ü86Ý97ß:8ã99æ87é77ê67ì46ô5:ô5:ñ7:î79ç85à72Ú6-Ô6+Ó7+Ñ8*Ð8*Ð9(Ð9(Ð9&Ñ8(Ð9&Ë9"È:"È=&Ç?'È@(È@*Ì=-Ï;1×96üU\ÿ^qÿGcÿ?aÿ;bÿ7gþ3bû8`íFdÅNa@IT15A55@:<:46?56>6389+6<"9C&#x1e;SZ.Q¢§q±Å¯Ë°Ë¦Ã®}{ehR_E^@c@iDoJsPsSqUoSoNlJhGfEgJiLiNiNfKdIdIeJiJjIiHfEjLlNoQ&.!)/%-2+13.333764:97=<7GD=DD<HGCPPRZZd_ap``z\\~`_YXggrtÁz|Ñ&#x7f;ÜáäãáÛÕÒÎÌÉÌÎÓÖ×ÔÐÄ·«&#x7f;xsnuqbewXVtSLjVKg`NbfO_eIcaHp^F{R@\N·ZSÆTS×VZâW\êY`òce÷qnðme÷qhÑD;ÔA7Å1%È4&ÔC2Ç8'Ç:(É;'Í<)Ñ;,Ó:,Ø7-Ù6-Ï,#Ï1&Î6)Ë<,Ç@-Â?+¾;'¼7$À4#ÑB2ãOAåOAØB4Ë7)Ë7)Ð>/Ê;+É:*Ç8(Æ7'È6'Ê6(Î8*Ð8+Ó;.Õ;/Õ9-Ò6*Õ7,Ú<1Ý<2Ü71è88ñ8=ø;Bû9Bú4Aö0=ò,;ï,:í/;ê19ê2:ë48í49ñ48ô36ó57ð37ï6;í9<é69â45Þ44ß76à;9Ý86Þ97á98ã99æ87é77ê67ì46ô5:ô5:ñ7:ì89æ95ß82Ø7/Ó7+Ð8+Í9+Í9+Ì;*Í:*Í:(Í:*Ì;(Å9"Ã;#Ã=$Â>'Ã?(Æ?,Ê<.Ð92ß<?ÿ\fÿQhÿIfÿ9ÿ=hù.ý<hé;\å\p¦LV^-0G85BC><89?48F37C41=:)>D(GV-bs?a½&#x7f;±Ë«É¦Ãº¬z{efPZ|@^?a>iDoJrOrRpTnRoNlJgFeDfIhKiNiNfKeJdIeHgHhGjGgEkLmOpR)/#)/%(-&+-(///3317759:4==5?@8DE@KKKQQYVXes^}YXY^ee£nn¶wxÈ&#x7f;ØâîðìäÜÖÓÎÈÇËÎÐÓÒÏÅ¼¯}xwozkegfXmQmMi\Lf^Kd_IeaHh`Io_FoP;UA¦^PÀbZÑYXÝQTîSø\`ùggæYRÔD<Î70Ñ7/×:1Ö<0Ô;-Ï9*Í:(Ð:)Ñ:)Ô8+Ø8,Ü5,Ü5-Ð/%Í3)Ì8,Ê>/Å@/ÂA.¾=*½:(Â9)¾0"éWJàL@Ã/#Ñ=1Î</Ç8(Ê;+É:)È9(Ç8'È7&É6&Ë5&Ì4&Ð7)×;.Ü>2Ü<0Ù7,Ú8-â>4ìC<õBEò9?ñ3=ö4?ö2@ñ-;í-:í2=ì7@æ5;ä5:å78ë8;ð9;ò89ï77ò;=î<<ê::æ87á77ß76Ü75Ü75Þ97á98ã99æ::é9;ì9<í9<ï8<ô7;ó6:ð8:í9:ç;9à;5×90Ï7,Ê8+È9+Ç9+Æ:)Ç:)È;)È;*Ç<)Á9#Ä='¾:#¶4&#x1e;¼9%ÉD3Ï?6Ê0.ÿ^eÿVgÿIbÿ<\ø1X÷0Yû6cõBiçYo¬FQo33N3,B?6>C<A89@-1L/3K2.B9(DI)Zm?~_¸w¨Å«É¾¶²~ nfRYwCY{?^?cAiFnImLlNkQlPjIfDcBcBfIhKgLfKhNhNgKfHeGgGkImLqSrUtW*0$)/#).',.)//-220561782;<4>?7CD>IIGOPTSUaXYku\Z&#x7f;^dbkj¤utº&#x7f;~ÎßéïðëâØ~Ò&#x7f;ÏËÉÈËÎÏÎÌÈ¼&#x7f;±|¡x{stknlb`eRe_Oe^Lf^Kf^KhIj^Hk_Io_Fw_G&#x7f;CYFZLÇVPÞRSõQXÿX_ðUSãKFÙ>9Õ60Ù6/Ü90Û9.Ö9*Ó7(Ñ8(Ò9)Ô8)×7+Ü5,ß4,Ý4-Î0'É5)È9+Ç>.ÄA/Á@-¿<*¾9(Ä8)Ë<.äREÜH<È4(Ì:-È:,Â4&Ê=,Ë<+Ê;*Ê;*Ë:)Ì9)Î8)Ð7)Ó7*Ö8,Ù7,Ø4*Õ1'Ø1(Þ7.ç<5øDGñ8>í/9ï/:ð0=î.;ë1<ë6?å6=Ü36Ø24ß:8îDDùIIøDCó?>é:7ç98å97â96á77à87á98á98à87â88å99ç9:é9;ì9<í9<ï8<ó6:ò59î68ê88ä;8Ü:5Ó9/Ë7+Æ8*Ã:*Â:*À;*Á<+Á<)Á<+Á<)½9$Â@*»:%³5&#x1f;º;(Á;/Ê94Ø<?ÿiÿPdÿAù8Wö6õ;`ô>dãIe«?L?>`;3G;/:>08>2F<:R>=N0.S8/VK7ciEx°pÂ}¡Æ¹}¹¶©xr^YuETp=\{B_&#x7f;@b@iEmHlKkMjPiOgGeCcBdEgJiLiNhMiOiMiKhHgGhGlImLpSqTtW-1#-1#,/&,.)//-11/34/56.:<1<>1@B7FGAKMLQRWVVbXWi\Yt^&#x7f;fdpn zx·Îßêïíçß&#x7f;Ó&#x7f;|Ë&#x7f;}È&#x7f;ÅÇÆÊÉÉÅ¿·|¡v{pujpoeci_VcM_aL_aKc^Jg\JlIn\HoIr^FwaIxX?S=£VDÇXOãUSøOTþMSâ>=Þ<7Ý84Þ71á6/â7/Ý7+Ú7(Ô7&Ó8&Ô9'Õ8'Ø6)Ý5*à3,Ý4-Ë3(Ä6*Â8+Á<-Â>/Â>/Á<-À8*Ã5)ÜI?ÚF<ÕA7Ï<2Ä6*Ä:-¾6&Æ:)Ç:)È;*É:*Ë9*Ë7)Ì6(Î4(Õ9-Õ7,×4+×3*Ú3+ß6/å:3é<8ð<=í49ë07í29î3<í4<ë7@ê=CÝ7;áAAìNMøYUýYWúQNïB>å84â62à72à74à74á85â96ä::ä::á77ä88æ8:é9<ë8<ì9=ë8<ì8;ï58ï58ì57ç77â:7Ù:4Ï8/Ç7,Â8+¿:+½;+»<+»<+»<)»<+»<)¸9&½@*³9$±6$¶;+µ0'Ã40ìPTÿiÿJ`õ:Uð7Vñ@ïHfçHdÍNa89e?6R@2DA09=,57)?7,L:0F.$R=,gB_ªt»{Áz¿{³z¸´mvLHb5Nh8\wB_}?c@hDlGkJiJgLgNdDdCcDeFhKkNkOkOkOkOiMhJgHiHlKoMu¡Vt¢Wv¤Y24&04&12*01+12-23.34.46+9;-;>-?A3DF9JKENPORQWTT^XVd\Zoda~nlzw¬ÃÖãêëâÙ&#x7f;Ì&#x7f;|Ã&#x7f;|¿~¿ÁÁÂÃ¿¸&#x7f;®}¤{vyn~rgmmaak^ViPc\J^aLaJc^JhJmZKqZJtZIvFsWAxR;Q=«ZGÍPãTPðFIñ>AÛ2/Ü5/á51ã60ã5.á5+à6)Û8)Ö6&Ó8&Ò9'Ô9'Ø6)Ü6*à3,Ú6-É7*À8*¼7(½9*¿=-Â>/Ã;-Ä8+Æ3)éUKÔ=4Ì8.ÑA6¾4'À<-À<-À8(Ä8'Å9(Ç:)È9)Ë7)Ì6(Í3'Ò6*Ô3)Õ2)Ü5-ã:3é>7ì?9ì=:è45ì59ï6<î5;ê2:æ39ä7=â<@óUVübbÿmjÿidóVQäA<ß63Þ50ß61Þ71Þ63ß74á85â96å99æ::å78æ89è8;é9<ë8<ë8<ë8<ì7<ï6;î68ê67å97ß<7Ö<4Ì9/Å9,¾9*º;*·<*µ<)µ<+´=)´<+³<(³:'·@,8&8'³;-«+"Ã54ý`iÿTdúGð<Uë=XèH`áOdÒO_´SZvGAXH9KD2FE1BE0=?*:8#;3&#x1e;?3&#x1d;PG*nkHf«v³t·sºxº¡¸¦&#x7f;lK_:BW0Mb7Yq?`|Ad@iClFiHgIeKeJbBcBdEgHjMlOmQnRlPjNjLiJiHiGmKpOv¤Yw¥Zy§\78(68*77-56056167267/68+;>-=@-@C.DG4IK>MNFQPNRRRVUSYXa_lkius}²ÇÓáâÜÓÄ}º&#x7f;|µ&#x7f;|³~·º¹·²&#x7f;¨zxtznzrfjm_^jTiPjZKfZJb_Lb_Le^Li\KnLrKw\K|H&#x7f;VBS>RA¶VFÏSIÚG@â88å33Ý1-ß3/â51ã4/á3,ß3)Þ6)Ü9(Ó8$Ð9$Ð;'Ñ:'Õ8)Ù7*Ý5,Ø7-Æ:+»9)µ6%·8'½;+Ã>/Ç;.Ç7,Ð90õ\TÖ=5É5+ÏA5¹4%·:(¼?-À;*Ã:*Æ:+È:,Ê;-Ì:-Ï8-Ñ7-Ð3*Ò1)Ö1+Þ71å<7ê>:é<8æ74é77ï;<ð<?ë6;å28ä5:èBFíMOÿxwÿqnñ`ÝJCÑ83Ñ2.Ü75ç?<à93à93à72à72â64ã75å76å76ç79ç79ê7;ë8<ë8>ë8>è7=ê7;ï8<î7;ê7:ä::ß=:Ö=7Ë;2Ã;/º8*µ:*³;*¯<)¯<*¬<(¬;)¬;)«<)>+§9(§9*ª8-¬/)Ë@CÿhrôM^óHîEZåHÙO^ÉT·TWXUdM?PM:LG4KF2FD-@C(>E$?I$KV.\g<xTh¤m¨n²r¹}¶§udRcC?P0BS1Oa9Wm>b{BeBiCkEiHfHbHcGbBfCgHiJkNmPnRoSkOkOkMiKiIkIoMrQv¤Yv¦Zw§?=.=>0==3==5=<7<=7;=2;=/?B/@D-CG.FJ3KL:NOAQQIRRHYWJZZN`_gerqn{y¡·ÄÑÓÏÇ&#x7f;º~y¯zw¦yv¥{v¬|y®|®}©}£&#x7f;x|tzrvksrffn`j\SkNlZLlKkKiMg^MiMjMo^Ns^MzMJ`OVG§N@¸J=Å@7Ï60Ø1+Þ1+á51â70â5/â4-à2)Þ4'Ü6&Ú;(Ñ9$Í;$Í<'Î;'Ñ:)Ö9*Ù7,Ó9-Â=,µ:(°5#³6$»9)Ä<.É;/Ë7-Ù?7øTãIAÎ:0Ì@3¸6&¬3 µ<)½:(¿7'À7'Ã7(Å7)É7*Î7,Ñ7-Ù<3Ü;3ß:4å<7ê>:ë>:ç85ã41è96ë;;ì::è8:ì>@øORÿbeÿppùheãUQÊ=6À1+É40Ö;7ß=;à;9â;5â;5â94á83ã75ä65å55æ66é69ê7:ê7;ë8<ë8>ë8>è7=ê7=ï8=ì8;è8:ã;:Ý>:Ó>7Ê=3À</¶8)²:)®;)«<)©<(¨;'¦;'¥:(¥<) 9( ;) 9*¢6*µ>8ÛTXûepîK\ïIéJ^ÛMÉQZ´VV ZR`RPI7IN:NI5H@+;5&#x1b;78&#x18;>O%Lf6lTw\edf¢l¬t°|mn}\Q`C<J09F,AP1O_;Wj=d{EfCkDlFiHeFaFcGcCiFjKlMlOlOmQnRmQmOmPlLlLnLrOs¢R{¬{«_zª^EB1DB3DB6CC;CC;CC;BB6BC3DF0EH-GJ-JM2ON9RQ?TREUSD\ZC\\D\Oaaaihxrq{z¦~µ¿ÃÁº&#x7f;{®ys¡toqlqmrn¡vpwrwrtoskzrinmb`l_Wj\QkLm\Lo\MpLqMqNoPmNl_Om`Os`O{`OZM_S¥QG®A:º;4Æ6.Î1*Ö2)Ý5,â70á6/á4.á4-à4*Ý5(Ù8&Ô9%Ï:"Ê="Ê>%Ê>'Í<)Ò;*Õ9,Ï;-¿@-±<(¬5!®3!º8(Ä<.Ê:/Î5-Ø93ðNIñTMÙB9ÌB5¾?.¦1&#x1d;´=)¿=-Á9+Á7*Ã7*Å7+É9.Ï;1Õ<4âE>ãA<ä?;ç>;é=;é;:è88ç85è96æ95æ66é;<ôJKÿ\ÿdeøbaÌ=9Á82½4.Ä92ÕA=ÞE@ß=;Ù42â;5â;3â94â94ä84å84ç77ê88ê69ë7:ì7<ë8<ë8>è7=è7?é6<ë6;ê7:æ89à;9Û>9Ñ>6Æ<2¾</³8)®;)ª;(¦;'¤;(£:'¡:':&¡>+7%=+9)2'ÃPKíkmî\fìP^éJ\ßHYÎLV¹RS¤RcQjUJM8DL7JB/H9$B7&#x1b;FH#Up=gV~²p³q¥h}YYcd|_XfCHU9:F.7C+:D,=J0JX7Wh>e|FiDlEmGiGdGaGaGfFlImNnOlOkNkOmQpTqSpSpPoPqPv¢Sw¦V|^z^y¬IC3JD6IE9JF;IF=IG;HF9HG5IH3JJ2LL2NN4RP;TQ>XRDYTAY>Z>\YH_\Udahnk~xs}x {§}«&#x7f;|¦}wvooilfiejfmhmimh|lfrkbgj`^gTh\NhKk\Io^Lq_Kq^Mq^MpOpOpOsOwN~N\LXL¢VI¤F<«9/º7/Ê7/Ô7.Ú6-Þ6-ß6/à5.ß4,à5-ß7,Þ8*Ø6'Ñ6$Ï:$Ì<$Ë='Ë='Î;)Ñ;*Ó:,Î<-¾A/¯<'©4"¬3 ¶9'Á=.É;/Ì8.Í3+àA;øTàI@ÐD7ÊF7«2&#x1f;¹>,ÊF7ÌB5Ê@5Ë?2Î@4ÔA7ÜE<ãF?â@;â=9ã:7ã75ã54ä65æ:8å<9ä?9à=8Ü:5ß@<êLIêQLÓ?;¸)#¸-&¿6.É>7Ñ@;Ö>9Ú;8ã;;ç==â96â96â96ä86å95ç:6é99ê::ê88ê88ë8;ë8;é9<è8;ç6<ç6<é6:æ68â89Þ:8Ö=8Í>6Â<1¹;-®8*ª;*¦;)¢;( ;'<);(;'?-5"=.8*/&Í\X÷y|ßX_æXdÛQ^ÊKT¹LO©VR`UhVnVPW8?J*<;&#x1c;C=&#x1b;UQ+lvD¢fÁ~¼v¸s|bnMv&#x7f;R^u}Vb@;D)4<%5=(=D2<E29C+CP2Vf?g{HiFmHmGjGfGbGcGhJlMoPoPlMiKjJjLsUsUsSrQrQs¢Ry¥V{ªZv¦Zu¥s£YRK;RK;RJ=RJ=RJ=RJ=RJ=RK;UN<VO=WP>XQ?YR@TD\UE\UEc\IbIaYLbYRekaiqftuj{zn|p&#x7f;s&#x7f;s|puj{mbrf^kc`k`_g_\c_Z^^ZY^YUaZRbZOeOgKi\Kk\IlHmFm\Hj^HibOqfTqZJuOBWJ\P¤ZM¬ZL¥M?¦E5§<,±7(Â8-Ñ80Ü41Ý1-ß6/Þ7.Þ7.Þ7.Þ7.Û7+Ù7*×7)Õ8)Ô8)Ò9)Ò9)Ò9+Ò9+Ô8+Í;,½;+¶>-®9(©1 «2!¸:+À</Ã9.É6,Î5-Ô:2ÛB:ÝK>ÔH9Â=,·2#ÝPFÙKAÔE=ÕE<ÚG?ÞG@àC>á=;à74æ87í;;ì::ç77ã75à85Ü=7ÙF>ÏB9Ä:0¼3)·1&º6*¿;/Ä@4Ã:0Ê=4Ó@8Û@;â=;ç;;ì8;í9<ç7:å8:å8:å99å97å97å95å95á51á51ä65å76æ89ç9:ç9;ç9;ë;>æ9;ß9;Ü=:ÖA=ÍB;ÁA6·?1©9+£<+ =*<(:&9'8&9&A0<-/#0'¼LJämoçloÍVZ½LNµNOªPO¢UOXPRaTjSWZ/:M&#x16;7F&#xf;Ta)~P°o¡¼y¼yµs~ahN\mCYdBTBHP9?D0:=,@C2=A2;?1<C3:C0AN2Rd>sBgHpKoKlGjIiKgLlPmSoSnPmMkJmJmIqMqNrNr¡Pt£Rv§Vz©Y{©^y¦cmZcPWN?WN?WN?WN?WN?WN?WN?WN?XO@YPAZQBRC\SDTE^UF_VGdJdLcYMcYOfUj_nbdpdhreltgpvirvirsfmm`gg_bXY^VZZRZWPXULXTIXTH\VH^WGcZKcZIfIg\HiGj^Fk_IiaJcbMngTv\MSFWN«VO°RJ¸RF¯G:¬B2¬=,±9)¾8,Ë80Õ62Ù40Ý6.Ü8.Ü8.Ü8.Ü8.Ú8-Ù7,×7+×7+Õ7+Õ7+Õ7+Ô8+Ô8+Ô8+Î:,¾<.¶>.®9(¦1 §2 ²:)¹>.¾<.Æ:-È5+Ë4+Ò;2ÜE:áOBßQCÝOCÛF?Ø@;Ó;6Ò:5Ö=7Ú=8Ü86Þ44è8:ë78î79ë78ç77å97â=9ÛB<ÊA7À@5¸:,±6'¯5&²8)¸=-¿@1ÅA4Í@6×@7ß=8ã:7é77ï58ñ7<é6<ç6<ç7:æ89æ87å95å95ã:5â92ã:3ã:5ä;6ä;8ä;8æ::å8:ä:=ã:=ß<=Ù?=ÐA;Å@7·=2<.¥;+ =*=+<);(:&8&9'9'8)5*<5¿QP×eeÑ^a¸NN¬NL¤SOXQ\R^S_T_VfQ`_/Sg*^r3zO¯n¥¾|²p bjHbvCTe;KY8FP7@I69?158-:</AB4=?4:=2>B4:C0@M3Rd>^tEhIqMpLnHmJkLjNmSnToSoQlMkHlHmHpJpJp Ls¢Qu¦Tx©Xx©Xx¨^mXbQYHTETETETETETETETE\SD\SDTE^UF_VG`WHaXIaXGeJeJdJeOeQfUhYi^\j^^j^`k_ak__i^\fYbWU_VQZUOWTKUQHRNCQMBSOCWQCXRB^WG_XFaYFcFg\Hi^Hk`LicMbaMngUy_PXN¢U±VQ²IE¸E@¹@8¸>3·;/¸8+¼8+Ã9.É;/Ñ:/Ù8.Û7-Ü8.Û9.Ú8-Ú8-Ù7,×7+×7+Õ7+Õ7+Ô8+Ô8+Ô8+Ò9+Í;,À>0·?/:(£2 £2 ª9'³>,º?/Ä?0Ä8+Å2(È4*Ð<0ØF9ÝN@âNDÙ:6Ø43Ô20Ö42Û97á=<æ<=é;<í9<î5:ë27ê37è58å99á<:Ù@:¾8-µ:+±6'®4%5%°8(µ<+»=.ÈD7ÐC9ÖB8Ý>8â96è66î66ï79ë6;ê7;ê7:è88è88ç:6æ:6æ;4ã:3ä;4ä;6ã<6ã;8â:9á99Þ88Ý9:Ü<<Ú@>ÓB=È?9º;27+¥7(¡;,<+<+;):(9&8%9&z6#y6%9, F=¹SOÀWT¸PO«IF¢MHSLZQ^SaWaW_VePsrFzV¦n¤¹«À³{|o<@S%AS+BQ0?L2:C06<.69058/9;.@A3=?4:=2=A3:C0@M3Qc=`vGiJsOrNpJoLmNlPpTqUoSnPlMlIlHnIpLpLpNu¤Sx©Xyªu¦Wq X_JV&#x7f;GOx@aXIaXIaXIaXIaXIaXIaXIaXI_VG`WH`WHaXIbYJcZKdLdJeHf^If^Kg^Mg^Of\Pf\Rf\ShWhWhWhWgTf\RdZPcPYNZVKVRFRNBPL@PM>TN@UO?XRBYTAVCXDaZGdJf_LfaMdcQleUv\OQ©b\¸ZX¶FD¹<:Ã<9Æ;6Ä92Â8-¾8,½;+¾?.Ç?/Ó9-Ù7,Ú8-Ù9-Ú8-Ù9-Ø8,Õ7+Õ7+Ô8+Ô8+Ô8+Ô8+Ô8+Ò9+Í;,Á?1·?/:(¢3 3&#x1f;¢7#¬=)µ@.¼?-¿:+Ã7*Å5*Ë7+Î:.Î</Ó:2Þ66à24Û12Ý34à88æ<=ê=?ì<?î7<ì38é06æ25å58á77Ú65Î61·5(¯7'®6&¬7&¬7&®9(²:)µ:*ÃA3ÊB6ÔA7Ü?8ã;8æ87ë76î79í6:ë7:ë78ê86ê86è94ç:4æ;4â92á:2ß:4à;7Þ:8Ý;9Ü:8Ù99Ö<:ÖA=ÒC?ÊA;¼<3¯6+¤6'7(;*;););(:&9%7#9%=*w6$5'E:ªPH¬MG©LG¦QLPJULZO^RaVbWaWgT_©z½¦¸¡tsYN`8/A&#x1b;->&#x1c;4C&:F.=F3;A3:=2:=4<=5::.?@2<>39<1<@29B/?L2Qc=awHjKtPsOrLqNoPnPrTsUpRoPmLmJnJpKqMrOt£Sy¨X|«w§mSeMHS|DNw?dLdLdLdLdLdLdLdLcZKcZKdLdLe\MfNfNg^Mf^Ig_Hh`IiaLiaNi`Oh_Ph_Pj`Tj`Ti`Qh_Ph_Nh_Nh`MhaOd^N`ZLUGVPBPM>NK:NK:NK:QN=RO>TQ@VS@XWCZF\H^Kb`Qf^QmWJRjd¾c`ÁMMÈBCÐ;=Ô89Ó84Ì70Á9-º=+µB-¼B+Î;+×7)Ø8*×:+Ù9+Ö9*Ö9*Ô8)Ô8)Ô8)Ô8)Ò9)Ò9)Ò9)Ò9)Í;,Á?1¶>0«:(¡6"4&#x1e;7 £=&¯B-³;*º;,Â:,Ç9-Ë8.Î:0Ñ=3Ú;5é9<ì4<æ39ã28â38â59â38â17é6<é49ç4:ç7:ã9<Û76Ð21À/*´6*9*:(«:(«:(«:(¬9&¯7&¸:+Á;/Î>3Ø?7á>9ä;8è96ì87î68î66î66í74ê84è:3ç:3ä<3á:2ß;2ß<5Ü=7Û=:Ù><×=;Ô>=Î@<ÊC=ÅA<º>67-£5(7(:*;)<);*:(9&8%7$8%@-u4"{3%G:¥RJ¡NFNG¡WNVMYMZNOR`U`YhWh¤|©jkHDS42@&$2&#x19;-:&0;*5=.9?3<?4=>6;<4::099-?@2;=28;0;?18A.>K1Pb<bxIkLuQtPsMsPqRpRrSrSrSpOmMmKpLsOt Qv¢Uz¦Y{©^z¨_qVdJZDJW~GRyBe\Me\Me\Me\Me\Me\Me\Me\Me\MfNfNfNg^Og^Oh_Ph_Nh`Kh`IiaLjbMjbOjbOjaPjaPjaPiaNh`Mh`Kh`Kh`IiaJjbKf_Mc^K^XHXRBSM=MJ9KH7IH6LK9LK9LM;NO=PQ?QTCSVETUEYL^VIcQEzXN¡ha¶eaÄVU×RSÙ@Cß:>Ü87Õ83Ç;.¼?-±C*µA(Ê;*Õ8)Õ8)Õ9*×:+Õ9*Õ9*Õ9*Ô8)Ò9)Ò9)Ò9)Ò9)Ò9)Ò9)Í;,Á?1¶>0«<)£:%6&#x1f;5&#x1d;;#§A*²B.¹@/¿;,Ã7*Å2(Ë4+Ò;2ß=:ê5:î3<ë6=é8>ç:>ä;>ã:=â9<ß58á7:â9>á=>Û=>Ñ96Ã2/¶0'¯9+«=,«=,©>*ª=)©<(©:'«8%¯6%¹7)Å;0Ð=3Ù<5ß<7ä;8é:7î87ï75ï75î85ë:4é;4ç<4ä=4â>5à?7ÞA:ÛB<ÙA>ÕA?ÒA>ÍB?Á@:»@8³=3ª8-¡5(5'9)<*;*<)<(;'9'7%6$7&}:)t3!~9*OA£YNRGOEWKYNMZMM\Q_T`YgY~wZy`r|ZVaC;E,/;%0;*0<.3=25=27=3<?6@A9?A6<<277+89+>?1:<17:/;?17@->K1Oa;bxIkLuQuQtNsPrSqRsRsPrOpMnLpNsOv Ty¥X{§\|¨z¥qWgOEXBaOYISzCg^Og^Og^Og^Og^Og^Og^Og^Og^Og^Og^Oh_Ph_Ph_Pi`Qi`OjaPjbOjbOjbOi`Oi`OiaNiaNiaNh`Kh`Kh`Ih`IhaGhaGhaGg`Me^KbI\WDWR?PM:MJ9IH6IG8GH8GJ9GK:HL;IM>JN?KM?PPDXRFOCiRDZNVL°SLÔZYÛIJâ?Bá99Û75Ï;1Ã@.´B*³<$É:)Ó7(Ô8)Ó:*Õ9*Ó:*Ó:*Ó:*Ò9)Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Í;,Á=0µ=/¬=,£<)8"4&#x1b;9 ¡A)C-¶A/¿=/À6+Â/%Ç0'Ð6.Þ:8è59î6>ì;Cê?EæAEàBCÜ@AÚ@@Í12Ð66Ó:<Ñ=;É;9¿82µ4.4)ª<-¥>+¦?,¥@,¥>+¥?)§<(©:'ª7%²7'¼8+Å8.Ì70Ô94Þ<9æ>;ë97ì95ì95ë:4é;4è;4å=4â>4àB7ÝC9ÚC<ÕB;ÏA=Ê?<Ç=;¿>9±<3©:/¢8+6(7'8):):(<+<*<):'9(7&&#x7f;6%}6$x3#x5%C3UEWKPCPDZLYMZN\Q\QT\U\WbUkbEWa>JS4=G,6?*3>-4>35?66=69@9=B;CF=EG<DE7@>/;9*78*=>0:<17:/:>07@-=J0N`:awHkLuQuQtNtQrSqRqOrNqMpLoKrNu¡Ty¥X&#x7f;«`~©az¥^qWfL_GE^IbPYKSzEh_Ph_Ph_Ph_Ph_Ph_Ph_Ph_Ph_Ph_Ph_Ph_Ph_Pi`Qi`Qi`QmdUlcTj`Ti`Qh_Pg^Oh_Nh`MiaNiaLiaLiaLiaJh`IhaGh`IgaKg`Me^KbH\WDUR?QN=ML:EF6EH7DH7DH7CJ:CJ:CI;EI;IK>SQDSPAYM=eK:tD6I?ÂYSÙQQàDEà::Ý86Ö<4Ê>/º=)µ8"É:)Ð9(Ñ:)Ò;*Ò;*Ò;*Ò;*Ò;*Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Ñ:)Í;,Á;0¶:.<,§>+9#3&#x1c;7&#x1e;?'¢9$¯<)À>0Ç=2Ì8.Ð7/Ù<5ã?=ì=Bî?DêDHâDEØ@?Ê;7À50»1.¹/,¼51¾:6»<6³:2ª8.¤8,¢;,¡=->+ ?,¡@-¡@-¡A+¤?+¦=*©;*:(²7(¹5)¿5+É90Ö>9àC>å<7è;5è;5ç<4å<5ä=4á>5Þ@5ÜD9ØD:ÒC;ËB:Ä?:½<7¸85°93¢7-8(5'6':)<,:)8&=*<+<*<)~9)}9&{6&{6&z2$8*G8SCQCL=RE_SWLYNRT^W\VZU_RaX;EM(6@&#x1e;8A&>G2=F53=2-7.1817>6CF?GJ?HI;BC1?>*;:&78(=>09;069.:>06?,=J0N`:awHjKuQtPtNtQrSqRqMqLqLpKpLsQw£V{§\e«dx£^jPEYA^HePcSZLSzEi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi_Sj`Wj`Wj_Yj`Wj`Vj`Vj`TjaRjaRjaPjaPjbOjbOjbOjbMjbMg_Jg_Jf^Ie_IbIZXCSP=ML:HI9EH7AE6@D5?E7?G8?G:?E9GK=IK=BF5KJ6\Q;aE0J9ÀjÜg`ãUSäFEåA@àA=Õ>5É=0Æ=-Æ4%Ë5&Ì6'Í7(Í7(Ï9*Ñ;,Ó=.Ò<-Ò<-Ò<-Ò<-Ò<-Ñ;,Ñ;,Î<-»2(¶7.±;/¨:+7"5&#x1f;7!9"¬A-²=,º6)Á4*Í6/×<7âC?êHEèDEéFIÝCEÉ98»61´80®8.©4+«9/©9.¥9/ 8-9,:+<+<+=+=*>+?,@-@. A/£@-¢;*¥:(«7(°6)·7,¾8-Ä;3Ì<4Ú=6à=6â?8âA7âA9ÞB6Ú@6×@5ØH=ÒE;ÇA8»;2²5/ª3-¦0,2+9-;*9+:*:);+;+<+:(~9)|9(|9(|9){8'y6&x3$w- A6PDMAMARFUJRG\R\RUMTLZUXSWTgZT5IQ*BL)BK,>H/7@+2=-4?14<1HPCZRX\MMO:DF.AA'==%:;)>?1;=2:=2<@24=*;H.Pb<cyJlMuQsOrLrOrSrSrNqKrNrNsPv¢U{§\&#x7f;ªbfx£^jO`G^GaJfPiUdSLTyEi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`Qi_Sj_Yj_j_j_j_Yj`Wj`Vj`Tj`TjaRjaRjaRjaRjaRjaPjaPh`Mh`Kg_Jg_JdJ^YEUR?ON<IJ:FI8BD6@D5?E7?G8>F9=E6@F8>E5>F1HL5MI0U?(&#x7f;P<¼sdËcXà_ZëUTçGGâ@>àA>Ø?9Î:0Í9-Î<-Ï=.Ï=.Ï;-Î:,Î:,Î:,Î:,Ð:,Ð:,Ð:,Ð:,Ð:,Ð:,Í;.É?5ÅB:¼B7®=/ 7$2&#x1e;1&#x1d;2&#x1f;¦5#²7(Ã:0ÑA8ßF@æIDéJFêKHÞCAÛEDÐB@¼;5¯9/©=0¥?1=0?1>2?1>/=/>.@.?,=*;*<+=,>-?,>,?-@.¡=-¤:,¨8,¬8+±8-¶:0¼9/Ê;3Ð;4Ó?5ÖB8×C9ÖC9ÔD9ÑE8ÊB6ÃA4¹=3°:0©6/¢5. 5/7.</=,;,;+:*;*;,<,~;*|:*{9)y:)x9(w8'v7(w3&3)C9NCL@K?QDTIRGSJXOUOWQXWRSOcTW5Xc9_jBZfBIT66C)4@*<H4OXGYbQ_fTW\HJN7BE*??#<<"9:(=>0:<1:=2<@25>+;H.Pb<dzKlMuQtPrLsPrSqQsPrLpLqNt Sx¤Y|§_&#x7f;ªc{¦aoUcJDaJeNeNcOdSLTyEi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QjaRjaRjaRjaRjaRjaRjaRj`Vk`\k__k__k`^k`\k`ZkaXkaWkaWkaUkaUkaUkaUkaUkaWkaUlaOk`LiaNiaNf_Ma\IYVESR@NL=IJ:DF8AE6AE6?F6>D6<D5=F57@-@H1IK3F?%W?'ZE·rb¶VJÙbZðebìTSæFHëGHæCDØ:7Ë7-É:,É:,É:,Ë9,Ê8+É7*È6)Ê8+Ë7+Ë7+Ì8,Ì8,Ì8,Ì8,Ë8.Â3-¾5/¶6-3(¡0". ¡2!§3$¼@4ÇC7ÖG?áLFèMIéJGåFBßD@×CAÏDAÁ@:±;1¤</¡A1E4E3D3D4B2A0@0A/~A.&#x7f;@.:*9+:);*<+<+=,<*?.>.;+9* 8+¥9-©:/®9/¸90¾90Á;0Ä>3Å?4ÅA5ÄB4ÁA4¶:.²9.ª8-£7+6+8-8/;1</=-;-;,:+;+;,&#x7f;<,};+{<-z;,w9*v8)u7(t6)v4&<2G=MDJ?H<OBTIRIOITOSQWSZYUTl}X~^alxRLX46D#=J.LZ@eqYeoWeMPU>EI0?C(=@#:<$89';<.9;09<1=A36?,<I/Pb<dzKmNvRtPsMsPrSpPu¡Rp LnLpOw£V{§\}¨a|§bpUfM\E\EeMjRfN`IcRZ&#x7f;LSxEi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QjaRjaRjaRjaRjaRjaRjaRj`Vk`\k__k`^k`\k`\kaXkaXkaWkaUkaUkaUkaUkaWkaWkaXkaWnbRmbPjbOjbOibPe^L_YIYVERO@NL=GH:DF8CE7BD6@D5<C3<H4:C.FJ3MG1P;&qL9bQª^P¡A5ÆSLçc_ð^^íSUëJOéDJáACÌ92Ç9-Å7-Å7+È8/Ê:/Ë;2Í:0Ñ>6Ñ>4Ò=6Ò>4Ó>7Ó?5Ó>7Ó>7Ô@<ÑB>ËD>ÆF=ÂF<ÃI>ÇM@ÏQEÙSJÞQJâMGäIEäFCâC@àA=ÛB=ÕFBÉD?º?7«=0 ?/C2H4I6G5E3C3B2A0~A/zA0}@.:,8,9+:,:,;-;+;+:+9+9*9+9,;/=1¢<0ª:/®9/°:0°:.°:.°:.¯9+®8*¨7)¥7*6)6*7+8-9/<0<.<-;,:+&#x7f;:+~9*~;+};+|:,z;,y;,w9,s8*q6(r7)u5)C:I>KAH<I<ODTKTLQLTPQOSO\X_Xh^~jpfuWYh?CR+?O*P`<crQetU\iKOY>CL1?E+?C*=A(;='78(:;-79.9<1=A37@-<I/Oa;bxIlMvRuQtNtQrSpRsPoMoMs¢R|¨~©ay¤_tZcJ^ED_HgOjPfL`HbQY~KRwDi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QjaRjaRjaRjaRjaRjaRjaRj`Tk`Zk`\k`ZkaXkaXkaUkaUkbSkbSkbSkbSkaUkaWkaWkaXkaWocUnbRkbQlcRkbShaQdM`ZJXRBSP?NK<IG8GE6DE5BC3>B1=I1?G/LG1S>+g=-ZL_S>63+¶?9ØXUóigödeëRUæJNéOQÚIDÓF<Í@7Ë>4Î?7ÒD:ÕF>×G>ÚJBÚJAÝJCÝJBÞKDÞKCàKEàKEëSRèTTãSRÝSPØSLØSLÛTNàUPàLHáIFãEBâC@äB@âC@äEBáGEÙIHÊDAº?8¬=2¡?2B1E3G6E3D4B3A1?1~?0z?1|>/:.9.9.:/;-;-;-;-9*:+9+9,;,;-<.;/:19/9- :.¡9,£9,¢8*¢8*8)8)8(8)9*:,:-;-:-:,9+8*}7+}7+|8+{9+{9-y9-x:-v8-q5*o5)q7+u:,H=J?I=I;L@RFWMXOQMUQSPVQ`YbWg{enrQZi@JY0BS)IZ0WiAdvNj|VXiEL:?L.:D)<D,>C-<A-:=,89+:;-79.9<1>B47@-;H.L^8_uFiJtPtPtNtQsTqSnMmLp¡Pw¨Y|ªaz¨`rXjRZC\EaIeMgMgKhJgMcPZ|JRtBi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QkbSkbSkbSkbSkbSkbSkbSkaUlbYlbYlbYlbVlbVlcRlcRldQldQldQlcRlcTlbVlbXlbXlbXocUnbRkbQlcRlcTkbSg`PdM^WGYSCTN@MJ;KH9GE6FC4@C0;H.@I.OC-a=-F;¬YS´NJ,+²:9¹=;ÓSRòpnþvvñehëX^ð^_ãUQÚPFÔG@ÐC:ÑD=ÔG>ÖGAÖG?ÙHCÚJBÜHDÝJCÞJFÞKDàKGàKGâHHáGIÜHHÙGGÖHD×HDØGDÛEDàBAäABæBAèBBçCBäB@ßA>ÛA?ÛIIÊC@¹<6®>3¢@3@2B3D3A2@1?1=1<2<1&#x7f;=1&#x7f;=1;1;1:3;1;1</<1&#x7f;=/</=-;/;,:,9*8+7*:/8.9-:/;.=.</>.9+:+:*:+;,;+;,:+:,9+8*~6(|6*{5){5){7*x6*x8,x:/s7,p4)o5)r:-w=1M@K>J<M=QCVHXMWNNITOVRVe\|_QgWGa_HEO->N)?O(J\4_qGk}SewMXj@N`8@Q-6E&7D*=F1=D4:>07;-8:,:;-68-9<1?C57@-9F,I5qBeFrNsOtNtQsTqSmLo Os¤Ux¨\z¨`s iQ`HYB^GeMgMgKeGgGhLaNXzHQsAi`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QkbSkbSkbSkbSkbSkbSkbSkbSlbXlbVlbVlcRlcRldOldOldMldMldMldOldQlcTlbVlbVlbVocUnbTkbSlcTmdUlcTjaRf_ObK^WGXQARL<OI9JG6IC3AC.:G+AG+U@+xG9£PLµII¶8<º39ÈAEÄ?@ÊJIßb`ñqpðnpçbeâZ\ÛQNÕLDÐE@ÍB;ÏD?ÐE>ÐB>ÎA:ÒC?ÒC=ÔC@ÕD?ØDBØD@ÙECÚDCÝADÝADÞDFÝGHßIHàJIãIIåGHçACêADìBEìBEçAAâ@>Û=:Õ=:ÚFFÉ=<¹83±<3¦@4?1>0A1=0<0;1:08193:4;4;3<3:4<3&#x7f;<3}=1}=3}=1|<0|=.|<0|=.};-{9)z8*{7*:.9.9-9.9.;.<.=-9+9+;+;*;+<*;+&#x7f;<+8*8,~6*}5)|3*z4*z4*y5*u2)v6,w9.s7,p4*o5*s:/x@3QCN?N=Q@VFXIUJSJPJSPTR\Ve\nLPK7EK1AN0KZ9WgC`rLgyQdyNWm?G\1AV-7H$1@!6B*=F5<D98=67:3:</;<.79.9<1?C57@-7D*FX2Vl=b~CoKqMsMtQtUrToPq¢Qu¦Wv¦Zs¡YjRbJE^H`IdLfLhJhHfFeIaNXzHPr@i`Qi`Qi`Qi`Qi`Qi`Qi`Qi`QkbSkbSkbSkbSkbSkbSkbSkbSlbVlcTlcTlcRldQldOldOldMldMldOldOldQlcRlcTlbVlcTnbTmaSmaSmaSmdUlcTjaRg`PdM_YIZSCTN>PK8MH5ID1DC.@F*DD(B,TG¯\X«>A¬/5ÎHOÆ?EÀ;>»;<ÄFGØZãefÛYÌGHÑJGÐEBÍB?ÐB>ÔFBÔFBÐB>Ê?8ÏD?ÏD?ÒD@ÔEAÖEBÙECÚDEÝDFß@Dß@DÞBEßEGáGGàFFàBCá>?èBDé@Cè>Aç=>ä>>ãA?áC@ßEC×CCÅ98¶50°;4¦?6>3=/?2</:0:081839494;5:3;3:2&#x7f;;2~<0~<0|<0|<0x8,y9-z:.z:.{<-{<-{<-z;,};/&#x7f;9/7.7,6*5*7*8)8,9*:-<-<-=,~<,~<,}9,}7-|6,z6+z4*x4)y5,x5,r2(u5+v8-r6+n4)n5*t;0xB6UGPBO@SDWIVJRGNGTNUPRKXO&#x7f;cUfGGK2=J,IX9ZkIhyUcxQZoFSh?Pb:L^6;M'0?&#x1e;,9&#x1d;5>)=C5<B8:<79<5;=2;=079.9<1?C58?-8B)EU1Uh;azCnMrNrPtQtUrTt Uu£Xv¤t¢ZoUfN^HDdNaK`HcIhJiKeEaE`MWyGPr@i`Qi`QjaRjaRjaRkbSkbSkbSjaRjaRjaRjaRjaRjaRjaRjaRkbSkbSkbSkbSkbQkbQkbQkbQlcRlcRlcRlcRlcTlcTlcTnbTkPqaRtdWrdWnbVj`TiaThbTgaSebQcMVCVQ=TO9NH2GA)KG,I?$bI3gW¶wn»lg¸YWºPR¶BE½BEÂDGÈFHÍHKÐJKÒLMÓKMÏEEÐDEÓEDÖFEÕFBÒC?ÏD=ÏF<ÈC:ÉD;ÊE<ÌE?ÒD@×CCÚADÜ@DáBFâBDáCDàDEàDEáEFãEDåEEãCCäB@ä@?ä@?ã?>â@=ãA>ÞC?ÓB?Æ?;¶93ª70:0=2>3>2?4?3=2<493839292;0<.<.;-;/;/:.:.~:-~:-~:/~:/};/};/};/};/}90}90~80~8.7.7.8/&#x7f;9-7.~8,}9.|8-{9-z8,x8,x8,x8.x8.w7-v8-v6,t6+s4+q5+m1'r6,n4)m3(o6+m4)q8-&#x7f;I=SHSJRJQJQIQJQKRLTMTKNcQo_HYU:MV7O_;j~YbxQZpIUjCRg@Pa=IY5DR17D&4>#2:#6;'<>0@B5@@6<>358-39-4:.7:/9=/;?.=B,=F)S_;arFoRtUtStSsSsUuZ~§e~§epYbK^G^H^G`IbKdMfNeMcK`H^G_LTvDKm;i`Qi`QjaRjaRjaRkbSkbSkbSjaRjaRjaRjaRjaRjaRjaRjaRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTnbTrbSqaRrbUrdWrfZogZmeXjdVgcWliZmjYjgThcOc^HU=RL4PJ0I?&Q?)kP=&#x7f;XINDE=EA±WV¸TT»QQÀPOÀNMÁMMÃMKÃKJÐTTÈHIÂ=>Ä=:ÐB@ÓEAÏB;Ç>4ÊD9ÊF:ËF=ÎH?ÔFBØDDÜAEÝAEáBFáCDáCDâDEâDEáEFáFDáFDàEAßD@àB?ßA>àA=àA=àA=ÞC?ÔE?ÈC<º>6;1 90:/<0=1@4~@5~>4=4:3919290;.<-<-;,;-:,:,:,&#x7f;9-&#x7f;9-&#x7f;9/&#x7f;9/~:/~:/~:/~:/}90}90}90}90}90}90|90|90|90z:0z:0z:0y9/y9/x:/w9.w8/w8/t8.s7-r6,r6,p6+p6+l2'q7,m4)l3(o6+l3(p7,~H>SJSLQLQMQLPMQNSOOGSHK{ePsiPilMfqQgxT^tMXnGPf@Ma<M^<L:HU7CP4<F-8@)6;'7:)<=/?@2>>2;=04:04<15;16<0:</<?.>A,>D*Q9^mDnSsUrSsSsSsVx^|¢exakT`IFF\E`IaJdMeNeNbK`I^I&#x7f;LRtBIk9i`Qi`QjaRjaRjaRkbSkbSkbSjaRjaRjaRjaRjaRjaRjaRjaRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTnbTqcVo_Rm\Rm_Tqdog\ldYhbVfbWqma{xi~{jzj{vcqjWe`JTT<FF.C>(MB.S@/U8*7*m>4n70u91~;5@:IBPJ¥XPZT»^YºRO·GE¼B?ÈDBÑEDÓD@Ï@:ÐA;ÐC<ÒE>ÔG@ÖGCÚFFÛEFÞDFßCDßCDßCDàDEàDEßEEßECàFDßECÞDBÜD?ÛC>ÜC>ÜC>ÜC=ÛC>ÖE@ÏF@ÃB<³>5¢:18-9.;/@2&#x7f;?3~>2>2:1919290;.<-:-:-;-:,:.:.&#x7f;9-&#x7f;9-~:/~:/~:/~:/~:/~:/}90}90}90}90|90|90|90{8/{;1{;1z:0y;0y;0x:/x:/x:/t8.t8.s7-s7-r6,q5+p6+o5*j1&p7,m4)k2'l6*i3'm7+}G=PIQJQKQLRMTOUPVPWOP~^OtbNoiQorUlwYfwUQeBK_:EY6DU3EV6GV9GT8ER8@J2<D/9>*7;*:<.<>0;=/9;.6<25=26<26<0:</<>0=@-=C)MW5Zi@h{MnPpQsRuUtWz¡`w`nWcL\E\E\ED_HaJcLdMcLaJ_HH}JPr@Gh9i`Qi`QjaRjaRjaRkbSkbSkbSjaRjaRjaRjaRjaRjaRjaRjaRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTnbTrdWpbUm_Tl_VkaWlcZlejfjfyujx~r}zg`cNOT>DG2GF2HC0F=,L<,VC4P9+R6*V4*Y4+\6+a9/e=3n>4KBRL®UQ·OL¼FFÆABÒDCÙGGÔ@>ÕA?ÖB>ÖE@ÙECÚFDÛEDÝEDÞDDßCDßCDÞDDÞDDÝEDÞFCÞFCÛFBÛFBÚE?ØE>×D=×D=ÙD=ØE>ÔD<ÓF?ÉF>ºA8§;18,8-;.@0&#x7f;@1~?0>0:1:/:0:.:-;,:-:-:,:,9-&#x7f;9-&#x7f;9-&#x7f;9-~:/~:/~:/~:/~:/~:/}90|90}90{8/{8/{8/{8/y9/{;1z<1y;0y;0x:/x:/x:/v:/t8.s7-s7-r6,q5+o5*o5*m4)j1&o6+l3(h2&k5)h2&l6*|F<OGOJPKRMTOVPWPZQRaUy_PjHd_IdhO\fKN@EU8AQ4>M0=L/?M3DQ7FR:GS;BK6>G2:A/9=,9=.9=.8<-7;-7=36=56;47=3:<1;=/<?,<A*JS4Ve>dwIlLpQuTw¡Wx¢Z{¢areN\EZC\E\EZC_H`IaJbKaJ`I^GHZ|INp>Ef7haQhaQibRibRibRjcSjcSjcSibRibRibRibRibRibRibRibRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTlcTrfZth\sfndZjaXlesnh{xq~w ¨¥«§¨¤£z{kdgTOR?GJ7EH5BE2BE2EH5KL<JH9GD5D>0B:-A7+@6*G5)S5*qB8PHPK¯HI¿GIÌEIÓBEÖ@BÙ@BÚBAÜDCÚEAÛFBÝEBÜDAÞDDÞDFÞDFÞDFÞDDÝEDÝEDÛEDÚFB×FAÖF>ÔE=ÔE=ÓF=ÔE=ÕF>ÔA:ÓC;ÎE=ÁB9®>3:.;.=0?0?/>.>.;/:-;.;.:-;-;-:,:,9+&#x7f;9-&#x7f;9-&#x7f;9/~:/~:/~:/}:1}:1}:1}:1|91z:1|91y90y90x8/x8/w8/y:1y:1x90v:0v:0u9/u9/t:/r7/q6.q6.p5-o4,o4,n3+m4+j1(m7-j4*h2(j4*f2'j6+}D;NGNHQJTLUOWPYR\R{YMwPkZJUBYXDY^HOYA?K3:H/7E.5A+5A+8D.?H3CL7EN9BK8@I6<E4:A19@07>.5<,4:,5<45<56;56;49;0:</;>-;@)GP1Sb;buGkKqRwVy£Yy£x`nW`IZ}C\&#x7f;E^G^GF`IaJbKbKbK`I_H^IY{INo@Ef9haQhaQibRibRibRjcSjcSjcSibRibRibRibRibRibRibRibRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTlcTkaUoepf\lcZkd\rmg|§£ µ±®ÄÁ¼ÉÆÁÉÄ¾ÅÀºÀ¹±¹²¨¤rgdUPQ?FI8>E3;D1:F28E38E38C3:B3<B6@C8CD<HE<>4*P:/e?6EA¡PO¿XËRWÊCIÕFJ×EFÚFFÜFEÞFCÞDBÞE@ÞCAÞDDÞDFÞDFÞDFÜDCÜDCÜDCÚDCÙEA×FAÕF@ÓF=ÓF=ÑG=ÑG=ÔE=Ô?8Õ@9ÐC:ÆC9¶A7¥=0=/=.>/?/>.>.<-;-;.;.;-:,:,:,9+&#x7f;:+&#x7f;9-~8,~:/~:/~:/~:/}:1}:1}:1}:1z:1z:1z:1y90x8/v7.v7.t8.w8/u9/u9/t:/t8.s9.s9.r8-q6.q6.p5-o4,o4,m4+l3*j4*h2(m7-i3)f2'i5*e1&i5*{B9MGMHPJTLVOXNZP|\Qs\Nl\M`WFVSBQRBJQ?@I64@,3<+2;*09(09(2;*5>-9B1<E4?H7?H7>G6;F5:C25@/3>-1<,3:33954954928919;.;>-;@*CL/O^7`sEkMrSyYy£Yw¡YqYfOZ}CX{A\&#x7f;E^G^G^G`I`I`I`I`I_H^G^KXzHLm@Cd7haQhaQibRibRibRjcSjcSjcSibRibRibRibRibRibRibRibRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTlbVj`VjaXkdZlh_upl°°®ÃÂÀÓÒÐâáßçæäçãàâÞÛÜ×ÓØÏÈÎ¼²¸¥zym_YIKL:AF2<E0?K7>I8>F7;C6;>5:;3983872==3C<2N71a84HH¯_Å`dÊWZÍQSÓNO×MKÛKJÞHGáFDãDAâBBßCFÞCGÞDFÝCEÝCEÝCEÜBBÛCBÚDCØD@ÕD?ÓF?ÒE>ÐE>ÐG=ÔE=Ù@:Ù@:ÑB:ÉD;¼C:¬@4;/9+=/>.>->-<,;,<-<-:,:,:.9-&#x7f;9-&#x7f;9-~8.}9.~:/~:/}:1}:1}:1}:1}:2}:2{;2y:1y:1x90w8/t8.s7-r8-t8.s9.r8-q8-r8-p7,p7,p7,o6-o6-o6-n5,m4+l3*l3*i3)h2(k7,h4)f2'h4)b0%f4)zA8MGNIPJULXNYOxZOr\Nk^N_XHRPAKN=EI:9B13;,08)17+17-06,06,06*17+2:+4<-9D4;F5=H7<I7;F56C13@,1>-.800621622717829;0:<.:?)?H+LZ6pCjLsVyw¡YsVgP^IVyAX{C~F~F~FH_J_J_J_J_J^IHJUvIJk@Ab7haQhaQibRibRibRjcSjcSjcSibRibRibRibRibRibRibRibRkbSkbSkbSkbSkbSkbSkbSkbSlcTlcTlcTlcTlcTlcTlcTlbVqh_ng_mhbtqj¡ ¼¼ºÍÏÎÏÏÏàààòòòúøùû÷ø÷óòñëëîãßèÐÆÝÁµÁ«¢}ub^\GKM7CH2AI2AH6@G7BD9BB:B=9C97?:6:=6>?7C82J.+e77QR´hjÆnmÀ\\ÇZWÐVS×QNÝMLâHHæCDåBEàDGÝDGÝDFÝCEÝCEÜBDÜBBÜBBÜDCÙCB×C?ÓD>ÒE>ÐE>ÑG=ÔE=ÜC=ÛA9ÒB:ÊD;¿E:°A6:-6'</=.=->-<,<,<-<-:,:,:.&#x7f;9-&#x7f;9-~8,}9.}9.~:/~:/}:1}:1}:1}:1}:2{;2z;2w;1x90v:0u9/s9.r8-r8-r8-q8-p7,n8,p7,m7+o6+o6+o6-o6-n5,m4+m4+j4*i3)h4)f2'k7,h4)c1&f4)b0%e3(xB8NHOIRMVNXOzZOqMjMZTDIJ:@D5>D69A219,/7*6<247058157247025.06,08-2:-6A3:E5=J9>K9<I77F33B-1>-,6..400511606718:/:<.:?+<E*IW4nAiMrVx\uXpU`IX{CSv>WzB~FZ}EY|D\&#x7f;G_J_J_J^I^IHH&#x7f;LStIHhA@^8icSicSicSicSicSicSicSicSicSicSicSicSicSicSicSjcSkbSkbSlcTlcTlcTmdUmdUmdUmdUmdUmdUmdUmdUmdUmdUldWjcgdvsn¯¯ÄÆÅÚÜÛêîïóôöö÷ùüüþÿÿÿÿþÿÿûúüööýòî÷áÔòØÇàÉ·Æµ¡ª ibbFEH-EI0CG0FE3GC7F<3C60G53I;:>=9>?:B:7I75cGFjjÅè©ÒÇtnÀXÆSPÕONàKMçFKéFKßDHßIKÛEGÕ<?Ø>@áGIàDGÖ:=Ú@@Ú@@ÙA@ÖB@ÔC>ÒC=ÐA;ÑA9ÞC>ÛA9Ó@9ÊA9¾B8°@5:.8*=0>/>.>-=,=,<,<,&#x7f;:+&#x7f;9-&#x7f;9-~8,~8.}9.|8-|8-|90|90|90|90z:1z:1z:1z:1y=5w<4w;3u:2t91q8/p7.p7.o6-m7-m7-l8-m7-l8-m7-m7-i2+m6/p92p92l5.g2*g2*h3+e0(i7.d2)^.$`0&_/%b2(s>6OJQLQLRJ}XOxRhXITN>DE5@D5:B37?24</3;04:049279478366457246116//6./7,.9+2=-6C2:G6:I68G44C.1?.*4+,2.-2./4.45/79.:</:?+6>&ES2YkCjRuy_rYkR~DXyDTu@Tu@XyDZ{F|GZ{F^&#x7f;JbNbN^&#x7f;J\}H_K`L_&#x7f;NTsJFd@<W6icSicSicSicSicSicSicSicSicSicSicSicSicSicSicSicSkbSkbSlcTlcTlcTmdUmdUmdUmdUmdUmdUmdUmdUmdUmdUldYgdonj¤¤¢ÁÃÂÖÚÛêîï÷ûü÷ûþøüÿüýÿþþþÿþÿÿþüÿüûÿúóÿðàüêÔíÞÇ×Í²À»¤£cgjKLQ3GJ/EC.GA1H>4I:3L95K<9?:7CB@NIFRPyjg£Î²®èÅ¿ÿ×Ñë²©Í|Àc^ÊVVØRSßHMÚADÖDEÕEEØHHÜJKÛEFÕ<>Ö<>ÛACÙ??Ù??ÙA@×A@ÖB>ÓB=ÒA<ÔA:ÜA<Û@;ÔA:ËB:¿C;±A6 ;18-=0>/>.>-=,=,<,<-&#x7f;9-&#x7f;9-&#x7f;9-~8,}9.}9.|8-|8-|90|90|90|90{;2{;2{;2z;2w<4u<3u:2s:1r90n8.p7.m7-m7-l8-l8-l8-l8-j8-l8-l8-k6.n70p92n91l7/i4,g2*e3*c1(d4*_/%^.$b2(^.$`0&p>7KFRNXR~XOrVKdRFRI:A?0<@18@15@23>04<14<15:37:379677577557246116/07//7,.9+1<.5B19F59H58G44E24B1.8//6/.3--2+23+46+8:-9=,7?(DQ3YkEjUv`ybpZgQY|DWxCTtBTu@VvDYzEYyGYzE{I_K`N\}H{I^&#x7f;J^~L^{MSnKF_A9R5icSicSicSicSicSicSicSicSicSicSicSicSicSicSicSicSkbSkbSlcTlcTlcTmdUmdUmdUmdUmdUmdUmdUneVneVneVlfZed_y{x²¶·ÍÑÒäéìôùüúÿÿûÿÿüÿÿýþÿþþþÿþüÿÿûÿÿúÿÿôþüçúùÝðïÑáåÄÒØ´½Å  ¨l`gEQX9EH-DC.JD4LD9KA8J?;C;9SJKia_wv¾³¯ÙÎÈêÜÓþèÝÿóéÿäÛð©£ÃecµBEÇDIÚRVÛUTÑMKÊDAÍCAÔFE×EFØDDÙCDÖ=?×>@Ø@?Ö@?ÖB@ÓB?ÓB?ÕA=Ù@;Ù@:ÓC;ËE<ÀD<²B7¡<29.<0=0>/>.&#x7f;=-&#x7f;=-=-=-&#x7f;9-&#x7f;9/~:/}9.}9.}9.{8/{8/}:1}:1{;2{;2{;2{;2z;4x<4t;2q;1r90o9/n8.l8-l6,k7,l8-j8-j8-j8-j8-i9-j8-j8-n91m82m82m82m82i70f4-c1*e3,a0),%_0(c4,\-%_0(sD<}PKVP~YQpSK^J?OC7@=.46(3:*1<,1<.2=/5=26<27:37:379679668368357227018008-.9+0;-2?.6C27E48G49H59F54>33:205.,2(/0(13(57*7;*6>'ER6lJmZyfyflZ`NWwEWuCUrBUsAWtDXvDYvFYwEZwG^|J_|L{IzJ_}K^{KZwKPgJAW@3I2icSicSicSicSicSicSicSicSicSicSicSicSicSicSicSicSjcSkbSlcTlcTlcTmdUmdUmdUmdUmdUmdUneVneVneVofWmgjkf£§¨¹¾ÁÐÕØåíïóûýõýÿûÿÿûÿÿüÿÿüþûýþùþýøÿþ÷ÿÿóùýæöýÞî÷ØçñÎÞèÃÎÚ´¶À¡«efmLNR7FH0JI7KI<HD9D@7OGDj`_¤·¯¬ËÆÂÞÛÔêæÝúñèÿûñÿ÷íÿÝÕûµ³åÊcf°@?ÃPKÉTMÉPHÄGAÈGBÔMJÖKHÐ@?Ò>>Ó??Õ?>ÕA?ÕA?ÔC@ÓB?ÓB=Ö>9Õ@:ÒC=ÌE?ÀE>²A9¢=5;2<1=0>/>.&#x7f;=/&#x7f;=-=-</&#x7f;9/&#x7f;9/~:/}9.}9.}9.{8/{8/~;2~;2|<3|<3{;2{;2z;4x<4r90o9/o9/m9.l8-k7,j6+h6+i7,i7,i7,h8,h8,h8,h8,h8.m;4l71j5/i70j81j81f4-a0)e4-a2*.&a2*b3+Y*"a2*yNG\VvWRjOHXD;I<3@:.:;-8<.08)/:*3;.4</6<26<27:169279479468349238139/19.19..9+/:,1<,2?.5B17E4;H6<I89C87?428.-3'./'/1&24'48'4<'ER8_pPtd}nyifXU|GSsBTqASp@Sp@TqAUrBVsCVsCXuExHzJzJ^{K_|L\yIWrIK_F<M;.?-gdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSicSjcSkbSlcTlcTlcTmdUmdUmdUmdUmdUneVneVofWofWofWngqqo¤©¬·¿ÂÍÕØãíïòüþôþÿøÿÿûÿÿûÿüýÿúüýõúüñøøìöøê÷ÿèóÿäòüáîùÙêôÒÞèÅÊÑ¯¶½£}bYBIM6GJ9FJ;CF;AB:c^X}z«¡Á·µÌÂÁ×ÏÍæáÝñîéú÷ðüõíÿñèÿòéÿóíÿåßé¨¤¸id§MBµPDÁYNÈ\RÆQH¿D=ÂA<ËDAÌA>Î@>Ð@?ÑA@ÒC?ÑC?ÑC?ÒC?Ó?;ÒA<ÐE@ÉE@½C>¯@9 =7<4<1<2=1=0&#x7f;=1&#x7f;=/&#x7f;=/&#x7f;=/~:/~:/~:/}9.|90|90{8/{8/|<3|<3|<3{;2z;4y:3y:3w;3n70m80m80l7/k6.i7.h6-h6-i7.h8.h8.g8.g8.g8.g8.h8.k92h6/f4-g5.i70h70e4-b1*c2+c4,_0(1(1(W+"c7.{ULtXTaNHN?8A7.=6,;9-9;.9<139-3;.5;/6<06<07:/68-47.69049238139/39/2:/19.19,08+.9+/:,1<.3>.6C2:E5<G7=E8:B55;/06(02'02%13%26%2:%GS;dtWyk~rsgRJp?Oo>Qn>Qn>Qn>Qn>Ro?Sp@TqAWtDxHzJzJ_|L`}MZwGSmFCU?6D7(6)gdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSicSjcSkbSlcTlcTlcTmdUmdUmdUmdUmdUneVneVofWpgXpgXoh^lnmz «µ·ÄÎÐÜæèí÷ùôþÿõþýöÿüúÿúûÿùûþóøúíõöèñôãõüêôýèôüåóùßòöÛéìÏØ×¹ÅÄ¦±°rggONP;GK:FL>HPEKQG{{sÅ¾¸ØÎÌßÓÓçÛÛóëéü÷ôðïëüûöÿÿöÿýôÿøñÿóêÿçÞÿ×ËØ¸eS¤K;´TF½WIµF;·@8ÈKEÇC>ÊC?ÍC@ÎC@ÎC>ÎC>ÍD>ÍB=ÑB<ÒC?ÎDAÅD?¸A=ª=8;6<6;3<4=3=2&#x7f;=1&#x7f;=1&#x7f;=1&#x7f;=/~:/~:/~:/}9.|90|90{8/{8/|<3|<3{;2z:1y:3x92w81u91n70m80l7/j8/i7.i7.h6-g7-h8.h8.g8.g8.g8.g8.g8.g8.h70h6/g5.g5.h70h70g6/f5.a2*e6.^2)\0'`4+a5,i=4uSJXG@E>6<5/:6-;8199/57,36+7:/5;/7:/7:/69.58-57,36+28,19,28.19,19.19,19,19,08+08+.9+/:,0;+3>.6A18C3<D5:C28>05;-57*35(13%04%18&HT>gw\wmvlh_RwKBh9Kk<Nk;Pm=Pm=Ol<Nk;Pm=Qn>WtDZwGzJ^{K`}M_|LVsCNg@<J90;3%/'gdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSicSjcSkbSlcTlcTlcTmdUmdUmdUmdUmdUneVofWofWpgXqhYoh^hikmtz£¬±¿ÈÍÓÞâåðòðüüòüûõþùøÿ÷úÿöüÿòúüîøùçö÷çööêù÷ëú÷èüöæýõâ÷ìØçØÃÕÆ¯À±}vnZWFPQCSVK\cfme²²¨ÕÎÈèÝÛóãäúêëÿñôÿ÷ùÿýýþýùùúôøùñüüòÿÿóÿÿñÿûêÿæÑñª¼o©VD®RCµPD»LA¿JAÂG@ÆE@ÉE@ÊE@ÊE>ÉD=ÈC<ÉB<ÑC?ÐE@ÌEAÂC=³=9¦;5;5<6<5<6<5=4&#x7f;<3&#x7f;=1&#x7f;=1&#x7f;=1~:/~:/~:/}9.|90|90{80{80{;2{;2y:3x92w81v70v70t80n70k90j8/j8/j8/h8.h8.h8.g7-f7-f7-f7-d8-d8-d8-f7-g6/h70i81i81h70i81j;3l=5g80l=5c7.a5,oC:xLCtI@nMDD:167/45/891:;556.14+25,58/58/57,46+46+46+46)48*08).9)08+08)08+08)08)08)19,08+08+08+08+2:-3;.4<-9B1:C0;B2:A1:<.68*35'/3$29'IU?euoghbYyTInE?d8Hg;Li;Ol>Nk=Mj<Li;Mj<Ol>UrDYvHxJzL^{M\yKQn@G_;6B4+4/#)%gdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSgdSicSjcSjcSkcVlbVlbVmcWmdUmdUmdUleUmfVngWogZoioh^mjekormtz¥®³ÃÌÑÕßáæîñôüþôýú÷ýùùþ÷úÿöýÿòýÿñüÿîýþðûøïþúñÿùïÿúíÿùéÿòßïàËÝÎ·Á²¢|ubb_NXZL^dXmwn|}¦¬¢¾¿·ÚÕÏîäâüîîÿôôÿô÷þôõþøøûúøûýøüÿúøÿõôüñ÷ýñÿÿñÿùèÿôãÿÌ»½&#x7f;pRE©SF¶TI²C:ÀIAÄH@ÇF@ÉF>ÉF>ÈE=ÆC;ÆC;ÍG>ÌG>ÇG>¼C;>5 ;3;4>6=5=6=5=4<3<1;1;1~:/~:/~:1}90|90|90{80{80z:1z:1x92w81v70v70u6/q6.m80k90k90j8/i9/i9/i9/i9/g7-f7-g7-f7-f7-f7-f7-f7-f5.j81m;4k:3j92j:0m>4oC8mA6oD;e=3d>3zVJdZyWMbLA>:13814927<569214-25,9<336-25,24)03(13(25*46+47,/7*-8*/7*/7*/7*/7*/7*/7*19.08-08-/7,/7,/7,08-08+7?09B1<E4=D4;A38<.26'-4$2;*JVBcri{a^uXRmLGeAAa:Hd;Kh<Nk?Nk?Kh<Jg9Li=Nk=TqEXuIZwKxLzNZvMNjAC;2>4*00!''heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjdVjdVkeYldYlcZmeZmeXmfVg`NkfSnkXolhe\feasqr¡«³¶ÃÈËØÝàêîïö÷ùùûúüþýþþüÿÿúýþöûüôúüñûýòüþóùúòúúòùùíüúëþüçù÷àéåÊÕÔ¶º¹|qtY_dMdkYr}l´¹²ËÌÇãâÞòîë÷óòû÷öýùøüø÷ÿþüþþüþþüýÿúýÿúýÿúýÿúýÿúúü÷ÿÿúÿûöÿêåØ§ £`XE>µNG¿MCÂF>ÄD9ËF=ÎI@ÎKAËH@ÇG<ÆG8ÃH8½J8³H8¦E5B4B5C5B7B6A5>3<2:19191:2~:1|91|91{80z7/w7.w7.w7.w7.w8/w8/v7.u6-q5+o4,l7/i81i81h70h70g6/g6/g6/f5.g6/i70i81i81g6/f5.e3,h3-j5/m80k90j:0i;.j>1kA3gB2jI:dH:v_O~k\xj&#x7f;oRJ=79.4:04:039/39/39/28.28.17-17-36-06,25,06,14+/5+/4-.5-/4-/4-/4-/4-/4-/4-05.05./4-.3,.3,/4-05./6.1;23=26@59C89D67B45@03>-4?.DQ=XfOkQRbHEX<BU9BW6Jd?Ke>Ke>Ke>Je<Je:Mh?Oj?UpGYtK\vOwPwRWqNHaA;N8&1+&+.$),heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVkdZkdZlc\mdmeZmfVidQjhSmjWjkgg_lll~&#x7f;¨¯µ¸ÂÄÑÖÙâçêòóõûüþÿþÿÿþÿþýûÿþúÿÿúÿþùûüôùúòøûòùüóøùñøùñøúí÷úéùûåòõÚßâÅÌÏ°´·vkqUZbJbkVuo§¿Á¼ÓÓÑççåóóñ÷÷õûûùüüúûûùþþüþþüþþüþþüþþüþþüþþüûÿþõÿÿõÿÿþÿÿÿûúÿíêð¿ºÂ~uKA¯MB¹J?ÂL@ÅK>ÃE9¼=4¾?6ÃE9ÄF7ÃJ7½L:³I9¦F6C4C6E7C7B8A7>5<4:39193&#x7f;:3}:2{;2z:1y90x8/v7.v7.v7.v7.u9/u9/t8.r6,p6+m4+l7/i81j81h70i70h70g6/g6/e3,f4-g5.h6/j81j81k92k92k60l71k90k;1l<0j>1j@0iD2gH6lQ>gRAvgTuesrKL<69.39/39/39/28.28.28.28.17-17-17-06,06,06,/5+/5+/4./4./4./4./4./4./4./4.05//4./4..3-.3-/4./4./6/.80/:22>44@66B66B66B45B16C1CP>PIR`IIW@AO6BP7FV;Jb@Ke@LfAKe@Ic<Hc:Id;Je<SnEXrK\vQwT\uUTmOC\?6I5&1-',0%*.heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVkdZkdZlc\mdmeZmfVmhUljUmjWjkkkcwww§¬²¿ÆÌÒÜÞìñô÷üÿþÿÿþÿÿÿþÿÿþÿþýûÿþúÿÿúþýøúûóö÷ïóöëóöëõ÷ìöøíöøêõøåõ÷áîîÔÛÛ¿ÉÊ«°±yuy`kpZt{i§®µÍÏÊÞÞÜïïíøøöûûùþþüþþüýýûþþüþþüþþüþþüþþüþþüþþüýÿþûÿÿûÿÿþþþÿûùÿúôÿ÷íýÑÆØM?¥RD©L=§@1·G;ÏYMÏUJ¼@4ÄD7ÆH:ÄK@¹I=«C8?5@7D=@9B:@:&#x7f;>8~>5}=4&#x7f;<4}=4z;4x<4x<4w;3u:2t91t91t91r90r90r90r90o9/n8.l6,k5+m80j81l71j81k60i70i70h6/g5.g5.f5.g6/h70i81k:3l;4l;4j;3i:0h<1k?2jC4iD2fE2gJ8lVAjYGujV&#x7f;m{&#x7f;mDG658-39/28.28.17-17-17-17-17-17-17-06,06,06,/5+/5+/4./4./4./4./4./4./4./4./4./4..3--2,-2,.3-/4..5.+5-+6.-9//;12>25A56B47D38E3?L:ER>CQ:;I28F->L3DT9J_@Jb@LdBKc?Ia=G`9F_8G`9RkDXpLuS^vVZqTPgK=T8.A.&1-',0%*.heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVkdZkdZlc\mdmeZmfVniVljUlkWlmpqi¡¦¹¾ÄÄËÑÛåçõúýûÿÿüýÿüýÿþüýüûùÿþüÿÿûÿþùüüôõöîïñæëîãéìáîðãñóåñôãñôßòñÜêêÐØØ¼ÇÇ«²±¡¡}ª¢¶»´ÀÅ¿ÜÜÚééçööôûûùýýûÿÿýÿÿýýýûþþüþþüþþüþþüþþüþþüþþüþþþûûýÿþÿÿþÿ÷òïüóìÿüóÿúíÿôäìÂ²¹pP@¢O?¸XJ·M?µE9ÀG<ÁA6ÈE;ÅIA½F>¯@9¡;6=8B??;@<@<}@;z?9y@9y@9y@9v=4v=4v=4u<3s=3r<2r<2r<2o;0o;0o;0n:/k9.k9.j8-i7.m82n72n72l71l71k60i70i70l:3j92h70f7/f7/g80h91g;2j?6f>4e=1f>2hC3iE5fG3cG2cJ6hV@i\IskVozor_9=,28,28.28.17-17-06,06,06,17-17-17-06,06,06,/5+/5+.3-.3-.3-.3-.3-.3-.3-.3-.3-.3--2,-2,-2,-2,.3--4-*4,)4,*6,+7-.:01=14@46B46C2:G5<I78E12?+2@)8F/>N4DX<E=K`AJb@K`?G_;H^:F^:ShGXpN`uV^uXoTLcI9M4):(%.+%*.$)-heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVkdZkdZlc\mdmeZmfVlgTljUnmYop`tum¤§¬¿ÄÊÎÕÛæðòûÿÿûÿÿûüþýþÿÿþÿþýûÿÿýÿÿûþýøøøðïðèèêßâåÚàäÖçëÜêïÛíðÛìðÙîîÖèèÐÚØ¿ÌÊ±Â½§·µ ±®¶´§Á¾µÉÈÃÒÓÎÙÙ×èèæòòðúúøüüúýýûÿÿýÿÿýüüúþþüþþüþþüþþüþþüþþüþþüÿþüÿüÿÿüÿÿûúüû÷ýÿùþÿøÿÿôÿþíÿþìøßËÅbPJ:¥L<µQA¼L>¿@7ÆC;ÅF@¿D?±>;¦:8::>==;?<@=}@=xA<tB;rC;rC;q?6s>6s>6r=5r=5r=5p>5p>5n>4m=3l<2k;1j;1i:0i:0j:0n72p62p62n72m61l71i70i70l;4k:3h91e90e90e90f;2d<2fB6dB6cC4dD5dG5dI6bI3_I2^L6eV?jbMrmWor\aM3:(28,28.28.17-17-06,06,06,17-17-17-06,06,06,/5+/5+.3-.3-.3-.3-.3-.3-.3-.3-.3--2,,1+,1+,1+,1+-2,-4-,6.+5-*4,*4+,6-.8-0:/2=/2=-5@/7B15A-2>*1=)5A-8F/=O7@T9FX>H\@J?I^=K^>K`?TgIoScvZauZZlTJG6G4'5&$-*$),"'*heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVkdZkdZlc\mdmeZlfVjgTkjUlnYorasvm¤¨¸¿ÇÖßäí÷ùûÿÿûÿÿûüÿþÿÿÿþÿþýûÿÿûÿÿúüüôôôêëíâäæÙßãÕÞâÓãèÔçìÖéíÖêìÔëëÓèæÏßÚÄÓÎ¸ÓÎºÌÆ¶ËÃ¸ÐÉÁ×ÒÎàÛØêæåðïíóòðùù÷ýýûýýûýýûÿÿýþþüüüúþþüþþüþþüþþüþþüþþüþþüÿýüÿûüÿøùÿþýþÿýùÿýôÿøðÿõ÷ÿöð÷çÿÿíÿîÛÍ§¤kX¢GªWEL;¹D:ÀD<ÀE>¼E?´A>ª=:¢:9:;=<?=?=~A<wB<sD<oD;oD;p?8q?6q?8q?6q?8q?6p?8o?5o@8n?5m>6k<2i=4h<1h<3l;4o83r73p62p62n72l71j81i81h70g80e90e:1d<2e?4f@5dB6_C5aG8cL:dM;cM8`K6^L4M4^R:bX?mhRssnv{eFO:4=*39-39/28.28.17-17-17-17-17-17-17-06,06,06,/5+/5+.3-.3-.3-.3-.3-.3-.3-.3--2,,1+,1++0*+0*,1+,1+,3,/6/,6.+5-*4,*4++5,-7,.8--8*0;+4?/5@/4?.3?+4?.3@,7F19K3@O8CU;GW<H?MBM`DUeJ\nTdtZ`rZXgRGXE4B1$2%$-*$),"'*heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjdVjdVkdZkdZlc\mdlfZlfVliVlkVkmXlo^lrhz~ ®·¾ÒÛàêóøûÿÿùþÿùúþþÿÿÿþÿúù÷ÿÿûþýøùùñòòèéëÞäæØáåÖâçÓäéÒçíÓèìÓçéÑèèÐçåÎàÛÇ×Ð¾ÚÓÃÙÑÆÞÔËæÝØîäãóéêúñôÿùûúù÷þþüÿÿýýýûýýûÿÿýÿÿýüüúþþüþþüþþüþþüþþüþþüþþüÿýüÿùøÿúùÿþýüÿý÷ÿþîÿùñÿÿòÿý÷ÿúõùëÿýíÿúèôÎ»«wbQ:ª\H²N>¸J=¹I>¸I@´G@¯D>¥=::7=9>:?9A:yB;tD:pF:pE<p?8q?8q>:q?8q>:q?8p?:p?8qB<pA9n?9l=5j=7j>5j=7n=6o83r73q73p62n72l71j81i81d8/d90d<2d>3d@4cA5cA5_C5ZE4_N<gVBhWCcS<^O8O5\R7V<Y>nlUtv^|kcmU2>(6B.3;.39/39/39/28.28.28.28.17-17-17-06,06,06,/5+/5+-2,-2,-2,-2,-2,-2,-2,-2,,1+,1++0*+0*+0*+0*,1++2+-4-+5-*4,*4,+5,+5,,6--7,,6+/:,2=/6A17B27B15@04?.2?-4C.:G3=L5AO8DT:JX?L\BTbI\kTcp\^mZUbQDRC2>0#/%'0-&,,$**heVheVheVheVheVheVheVheVheVheVheVheVheVheVheVheVjdVjdVkdZkdZlc\mdlfZlfVolYmlWjlWgkZgmcr{x¨±¸ÊÓØãìñøýÿ÷üÿúûÿþÿÿÿþÿúùõÿÿúýýõøøîññåéëÝæéØåéØçìÖæìÒéíÒêíÒèèÎèæÏèãÏáÚÈØÐÃØÎÄÝÒÌéÞÚøíëÿôöÿõùÿõûÿùüÿýþÿÿýÿÿýþþüþþüÿÿýÿÿýýýûþþüþþüþþüþþüþþüþþüþþüÿþúÿýûÿýûýüúùýüøÿÿõÿÿóÿÿíýúóÿûüÿúÿÿôÿöæÿóßÿãÍ×©oY¨WD®Q@¬O>¯OA±OBK@¤C<<5>7?6@9A7{B7uD6qE8qE:p?8q>:s>:q>:q>:q>:r?;q@;sB=qB<p?:m>8l=7k>8m>8o>9o83q73o83m82m82j92h91f:1f;2d>3d@4bB5`C5A3\@2WB1TG4\T?f^Gj_IbX?ZS9ZS7\W:W<XW;kmUsw^u}fUaI&2&#x1c;8D03;.4:04:039/39/39/28.28.17-17-17-06,06,06,/5+/5+-2,-2,-2,-2,-2,-2,-2,-2,,1++0*+0**/)*/)+0*+0*,1++2+)3+*4,+5-,6-,6--7.-7.-7./9.2<16A39D69D47B46A10;+2?-6C19F2<I5AO8FT=JX?R`IhTan\jYQOBN@0<0",#+1/(..&,,ifWheVheVheVheVheVheVheVifWifWifWifWifWifWifWifWkeWkeWlelemdne\mgmgWqnkmWknYkr`kshpyvª°»ÄÉÐÙÞêïóöûþüýÿþÿÿÿþÿþýùÿÿúþþöúúð÷÷ëôöèñôãìðßéîØñõÚíòÔëîÓììÔèæÑáÜÉÝÕÈÞÔËáÔÎèÚÙòääúëîþòôÿ÷ûÿùÿÿüÿÿýþþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþÿúÿþúÿþúþþüýÿþûÿÿúÿÿúÿÿûÿÿþÿÿÿÿýÿûõÿûñÿüìÿûåÿóÛÿêÑºi¦bKO9T?¨ZF£RAK; M?<0=0>2&#x7f;=/x=/t?/sB3uE7r=7s>:t=:t?;u@<u@<t?;q>:s@<rA<r?;q@;p?:o@:o>9o>9k92m82l:3l;4l;4h<3g<3d>3c?3^>1dG9cG9T=-N9(M8'@1&#x1e;XR<PN7KI0OK2Y>ieJeaDVU7XX<^aDptzfdlU>J2-9#2>*4</5;15;14:04:039/39/39/39/39/39/28.28.17-17-06,16016005/05/05//4./4./4.,1+,1+,1+,1+,1+,1+,1++0*.5..5.-4--4--4--4-.5./6.+2*-4,1805=29A6=E8@H=?J<7B26A03>-1=)3?+9E/BN8IU?OGWcOhWZeULWI<F;/90)3+)/+(.,'-+heVheVheVheVheVheVheVheVifWifWifWifWifWifWifWifWkeWkeWlelemdne\mgkhWonZjlVknYlsajtiq|x£®´¾ÇÌÒÛàëðô÷üÿýþÿþÿÿÿýþþýùÿÿúþþôûûïøùëô÷æñôáìñÝèíÖêïÑæëËäçÊææÎãàÍßÙËÞÔËßÔÎçÙØíßßöçêûïóÿôøÿ÷ûÿúþÿüÿÿþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüýÿúýÿúýÿüýþÿþýÿÿüÿÿüÿÿûÿÿüÿÿüÿÿûùÿüõÿþñÿÿíÿûåÿòÙÿëÑà±§u\\DZBYCVCSAG5F6E4D3|C2wB0s@/o>/u@8t?9u>;u@:t?;p>7p=9tB;q>:p?8p?:o>7o>9m>6n=8m>6k<4l=5m>6k?6j?6gA6eA5bB5dG9A2^G7_J9O>,F7$G8%?7"LL4HL3DG,CF+MM1?baC`aBZZ>dgJsw^sy_X`I:D,/;%7C/6>17=36<26<26<25;15;15;15;15;15;14:04:04:039/39/27127127116016016005/05/.3-.3--2,-2,,1+,1++0*+0**1**1*)0))0)*1**1*+2+,3,.5..5-.5-07/4;3:B7?F>BJ?;F8:E57B14?.3?+6B.<H2@L6LXDS_KYdTWbRLVK=G<1;2+5-+1-*0,)/-heVheVheVheVheVheVheVheVifWifWifWifWifWifWifWifWkeWkeWmdmdmdne\mgkhWnmYimVjoYktaiuiq}y¥²¸¿ÊÐÔÝâíòö÷üÿüýÿýÿþÿýþþýùÿÿ÷ýýóúúîøùëô÷äðóÞéîØåëÑâçÉÞáÂÝÝÁÞÞÆÞÛÈÝ×ËàÕÏäÙ×îâäóçëúîòÿôøÿ÷ûÿùüÿúýÿýýÿþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüýÿüúÿüúÿüûÿþþþþÿüÿÿúÿÿùÿÿøÿÿùÿÿúÿÿúúÿýøÿÿöÿÿóýÿïÿúæÿóÝÿøàÿìÕã¿§¬jYBR<FIS?&#x7f;F2u>)r>)tA,tD0vE6q?4n<3q<6tB9tB;o?5q@9xH>n=6m>4m>6m>4l=5j>3l=5j>5g?5gA6hB7gC7eC7cC6aD6^D5bK;UB1ZI7`S@RG3D<'E=(FA+BF-BH.AE*<@%?B%LO2_`AijKaaEnqTvzagmSJO95=&5=(=F39?38>48>48>47=37=37=37=38>48>48>47=37=37=36<26<25:45:449349338238238238205/05//4..3--2,,1++0*+0*).*).*).*).*).**/++0,,1-162/4.,1+,1+05/6;4=B<AH@@K=>K:<I78E34A-2?+3A*5C,DQ=KXDQ^MQ^MHTH<H<2>4.8/-4--4-,2.gdUgdUheVheVheVifWifWifWifWifWifWifWifWifWifWifWkeWkeWmdmdmdmf\mgjiWkmXimVjoYjs`hthn}x¦³¹¿ÊÐÓÜãëðööûÿüýÿýÿþÿÿýÿþúþþöýýñúûíøùéô÷äïóÜçìÕãçÎÜßÀ×Ú»ÖÖ¼Ù×ÂÝ×ÇÞ×ÍåÛÙìààøìðûðöÿõûÿùþÿúýÿûûÿûûÿýüÿþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüýÿüúÿþøÿþûÿþþþþÿüþÿúÿÿùÿÿøþÿúýÿùüÿùùÿýúÿÿúÿÿøûÿõüþðÿÿïÿûëÿñßÿðÝÿïÚñÒ½´zyS>tJ4uH3vI4zM8~O;yL7nA.e7'{M@tE;qB8sD:oC8g;0d8-g=1i?3i?3i?3g?3g?3f>2f>2e?2cA5bB5`C5`D6_C5D4ZE4WD3XI6OB/\T?oiSc^HMK4IG0IK3>D*?H-@F*<B&<A#EJ,XY:deFmmQxx\tv^_FAD/6;%:?+>B19?39?59?58>48>48>47=37=39?59?58>48>48>48>48>48>47<67<67<66;56;56;55:45:438238227105//4.-2,,1+,1++0,+0,+0,+0,+0,,1--2.-2.051.3/+0,+0*-2.3828=9;B:AL>@M;@M;=J68E14A-1?(1?(=J6BO;IVDJWFDPB;G;2>4.:0,6..5.-4-gdUgdUgdUheVheVifWifWifWifWifWifWifWifWifWifWifWkeWldWmdmdmdmf\mgjiWlnYinWiqZgs_drejyt£²·½ÈÎÒÛâêïõôùýûüÿýÿþÿÿýÿÿúÿÿøÿÿóþÿñüýë÷úåñõÞèíÖãçÌÙÜ¿ÕÕ¹ÒÐ¹ÔÑ¾ÚÔÈàÙÑêàßôéíþóùÿ÷üÿûÿÿýÿÿþÿÿþüÿþúÿþúþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüýÿþúÿÿúÿÿûÿÿþþþÿýþÿüüÿûüÿûúÿúøÿù÷ÿúøÿüùÿÿûÿÿûýÿúýþùùôîÿþöÿýôÿüóÿûîÿ÷çûæÕéÒÀÏ´¡¶uawVClI6mH6rM=uP@mF7jC4gB2kF6nI9nJ:oK;pL<eA1cB1cB1cB1cB1cB1bC1`C3F6\G6H7ZG6ZG6WH5TG4RG3PH3OI3fdM~~frrZVX@JL4GM3<E*<F+?H+AG+AF(CH*MN/TU6xw{z^nnVUU=CB.?A,?@.;>-;>39?59?58>48>47=37=37=38>48>48>48>48>48>48>48>49>89>89>89>88=78=77<67<66;56;55:438227105//4./4.-2.-2.,1-,1-,1-,1--2.-2.-10,0/,0/-2..211623764;4;G9=L9@O<@O:<K67F/4C,2A*6E.:I4@O<BP??M>8F71?2-9-,6--4,,3+fcTgdUgdUheVheVifWjgXjgXifWifWifWifWifWifWifWifWldWldWmdmdmdmf\mgjiWkmXinWiqZgs_bpcgxr£²·ÀËÑÔÝäëðöõúþûüÿýÿþÿÿýÿÿúÿÿøÿÿôÿÿóÿÿïúýèó÷àéïÕãçÌ×Ú½ÑÑµÍË´ÎËºÖÏÅÞ×Ñìâã÷îóÿöþÿúÿÿýÿÿþÿÿÿýþÿúþÿùþÿúþÿúþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþþýþÿýþÿýþÿýÿþýÿüþÿúþÿùÿÿ÷ýýóýýõþýøþýùÿþüþþþÿýÿÿþÿþùÿÿüÿÿûüûöóÿøòÿýôÿþñÿþíÿúéÿóà÷äÓçÒ¿Ôº©ºurd|K}^LtUCbE3X;)Y<*>,X=*`E2`E2`E2`G3`G3`G3`G3^I4\K7\M:\M:ZM:XM9UM8SM7QL6NL5QQ9jlT|gmqXSY?FL2>G,<F+<F+?H+DJ.EJ,DG*HI*LM.yx\tsWfdMTR;LI6KJ6FE3<=-<?4:@69?59?58>48>48>48>47=37=38>48>48>49?59?59?5;@::?9:?9:?99>89>89>89>89>88=77<66;55:4493382382/40.3/-2.,1-+0,+0,+0,+0,+/.,0/-10.21/32/32/32.5.4@27F3=L7@O:?N7<K49H18G05D-8G0<K6>M:<J97E61?2-9-+5,-4,,3+fcTfcTgdUheVifWifWjgXjgXifWifWifWifWifWifWifWifWldWldWmdmdmdmf\mgjiWjlWhmVhrZfs_cqdhys¥·»ÇÒØÚãêðõûøýÿüýÿýÿþÿÿýÿþùÿÿ÷ÿÿóÿÿðýþìøûæðôÝäêÐÞâÇÕÕ¹ÎÌ³ÉÄ°ÊÄ´ÐÉ¿ÚÒÏéàãöíòÿ÷ÿÿúÿÿýÿÿþÿÿÿýýÿúüÿøüÿøýÿúþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþþÿüÿÿüÿþýÿýÿþúÿúøÿùøÿõøÿóøÿóøÿõüÿøýÿúÿýþÿüÿÿüÿÿûÿý÷ÿÿúÿÿüÿÿýÿÿþÿþýûûûóúøìÿÿñÿÿïÿÿïÿþíÿýíÿöçÿîâýêÛúãÑÿæÐãÊ´¥xt^GeO8cM6\H0^J2\J2\J2K3K3^L4^L4\M6YN8XP;XP;VP:UO9SN8PN7NN6MO7PT;`fLfoTU^CEO4AK0<F+?I.?I.CL/HN2HM/GJ-OP1YX:qmRkgLa\FV@YTATQ>MJ9BC3>A6;A7;A7;A7:@6:@6:@69?58>48>48>49?59?5:@6:@6:@6;@:;@:;@::?9:?99>89>89>8:?9:?9:?99>88=78=78=77<6495384162/40-2.,1-,0/+/.,0/-10-10.21.23.23-12,2.+9*/@-6H2<N8=O7<N6:L2:L25G/7I1:L6<M:;L:6G70@3-;.,6--4,,3+cdTcdRdeUgfThfWigXigXjhYhfWifWifWifWifWifWifWkeWkeWkeWmeZlemdmf\kgjiWikVhmWisgt`dqgk|v¬»ÀÍØÞàçíôùÿúÿÿûÿÿýÿüþÿúþþöúúðúûíúûëøùçôöáëí×ßãÊÙÜÁÓÑ¸ÌÇ±Å¾¬Æ¾±ÌÃ¼ÖÎËæÝàôêòþ÷ÿÿúÿÿþÿÿþÿþþþýÿúüÿúüÿúýÿüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüþþüÿþüÿþüÿþüÿýþÿüÿÿüÿÿýÿýÿþûÿüøÿù÷ÿ÷÷ÿõøÿöùÿöüÿøÿÿûÿþýÿüÿÿûÿÿúÿÿúÿÿûÿÿüÿÿýÿþýÿûýüüþûþÿúùúòûûñýúñþúïÿùìÿûìÿýïÿýìÿðÛÿûâìÜÃu`P7RB)XH/XH.M3M3ZN4ZN4O5O5\P6Q8UO7TO9TO9QO8PN7NN6KM5IM4JN5LR8S\APY>@I,<E(AL.?J,DM0EN1JP4MR4KN1LL0YY=kjNhdIb^E^ZA_ZD`GZUAQO:KJ8AB4@B5@B5>A6>A6>A6=@7=@79?59?59?5:@6:@6:B7:B7;C8;@9;@9;@::?8:?9:?89>89>7;@:;@9;@:;@9:?9:?8:?9:?98?87=94;4382160/4..3/.3/.21.21-2.,2.,3,,3+,3+)6%$6&#x1e;*>"2F+8L3;O6<N6;M5;M58J29K3;M5=O7<N87H51B0.<--7..5--4,eP^fO_gRbgQdiUfiVgjWijXfgUhgUheVheVjdVkeWlfZlfXlfXkhYmgkg\mf^jfieZgeVghVfkUenYbodqht¨¬¸ÇÎÒÛâäéïò÷û÷üÿøüûüþùüÿøüþó÷÷ëùúêø÷åòñßîíÙëêÖáßÊ×Ò¾×Ï¼ËÀ®½´¥½³©Â·±ÌÂÁÛÒ×êàèúóûüöÿþûÿÿýÿþýÿüýÿüþýüþýþþþþþþþþþþþþþþþþþþÿþüÿþüÿþüÿþüþýùþýùÿüùÿýúÿþûÿþûÿþÿÿýþÿüýÿýþÿÿÿþÿýýÿüûýøÿÿûüý÷ÿýøÿþúÿýøÿúöÿ÷ôÿùùÿúÿÿüÿÿýÿÿþÿÿÿÿþþüýÿþüþýÿÿÿÿÿÿÿþüÿþùÿþõÿþðÿýêÿýæúôÜÿúàçàÆnUN2WM2TK.SJ+WN1WN1UN1UN1TO2TO2TN4SO4QM4RN5QO6PP6MO7KO6HM6GL5JP6FO4AJ-=F)@G(DK*FM+EL*PV4PS4KN/IL/TT8bbFccIZZ@Y@\ZABBY@XV=US:SQ8HF1GD1EB1B@1A?2>>2??5>@5;>59?59?59A69A48C58C59C8:B79A69@88@58?79A69@8:B7:A9:B7:A9:B7;B:<D9=D<>E=8B:8B:8B98@56>14</39/271.40/51070.8-.;)0@&6F)7M';V+:W+;U0;T4;R8=P<>O<>O<:L6@R8EW=FY=DX=@T;5H2+<*,6+-4,+2*ZfNgO^hP_iQbjSckTglXhkXhiWghVifWifWkeWlfXmeZmgYmgYkhYkgkg\je_jfhfZfgWhiWglVfo\erajwnz«²ºÈÑÐÙàãèìò÷ûøþþùþúúÿøûþõùýïøùë÷øèõôâíìÚéæÓåâÏÜ×ÄÒË¹Ê¿Á´¤¹¡¼¯¦Á¶²ËÁÀÛÐÖæßæûóþý÷ÿþûÿÿýÿþýÿýüÿüýÿüýÿþþþþþþþþþþþþþþþþþþÿþüÿþüÿþüÿþüÿýúÿüùÿüùÿýúÿýùÿþúÿþýÿýüÿüýÿýþÿþÿÿþÿÿýþýûüÿþýÿþûÿüúÿøóøíçòåÝöèßûðêÿúýÿüÿÿýÿÿÿýÿÿýþþüýÿþüýÿþýÿþýÿÿýÿÿþüÿý÷ÿÿðÿþèÿÿäúöÝýûâäàÇlRM0SN1RK.QK+UO/UO/TO1TO1SO2SO2QP4QO6QM2PO3PN5NQ6MO7JP6HM6EM5HQ6FO4BK.?H)CH(GM+JP,KQ-OS0PT1VZ9^aB_bCZ@VV<UU=VV>WU>XV=YW>YW>XW;WV:XT9QK3PJ4LG4ID1DA2B@3A?3>@5<=59?59?59A67B48C57C57C5:B7:B79A68@58@59A6:B7:B7:B7:B7:B7:B7;C8<D9=E:<F;=G?<H><F;;F6<D5:B39?38=67>76=63=52>04C,9L,BY/Ga1Qo;Mm;Li=Fa>BY?<R=<M=:K9=O9DV<K^BNaCL`DH\A:Q71D0+5*+2*)0(ZfNgO\hP`jRbjSdlUglXilYijXijXkhYkhYmgYmgYog\nhZmgYkhYkgjfje_jfhfZfgWhiWejTdmZgtco|s&#x7f;¬³¶ÄÍÍÖÝÞæéñöúøþþúþýùþøøûòôøêóôæðñáêé×áàÎÜÙÆ×ÔÁÐË¸ÇÀ®¿³£½° ¼°¤Ã¶®ËÀ¾ÕËÌãØÞëäëüôÿý÷ÿþûÿÿýÿþýÿýüÿüýÿýÿþþþþþþþþþþþþþþþþþþþÿþüÿþüÿþüÿþüÿýúÿýúÿüùÿüùÿüøÿüøÿýüÿüûÿûüÿüýÿþÿÿþÿÿýþþüýÿþýÿþûÿýûüóîêßÙâÕÍëÝÔ÷ìæÿúýÿüÿÿýÿÿÿýÿÿýþþüýÿþüýÿþýÿþýÿÿþÿÿÿýÿþõÿþíÿýçÿþãû÷ÞþüããßÆjPK.RM0QJ-QJ-UO/UO/UP2UP2TP3SO2QP4PO3QM2ON2PN5MP5MO7JP6IN7GO7HQ6FO4CL/CL-GL,KQ/OU1SY5QY2X`9jpLy&#x7f;qvVY^@LO4MP5QQ9QQ9RP9SQ8TR9VT;YU<ZV=XR<VP:RM:NI6HE6DB5CA5?A6<=59?5:@69A67B48C57C57C5:B7:B79A69A69A69A6:B7:B7:B7:B7:B7:B7;C8<D9>F;<F;?IA>J@?I>?J:@H9>F7>D8=B;>E><C<7A95A39H1BU5Pg=XrB`~J\|JXuINiFDA9O:6G74E3<N8EW=NaESfHSgKNbG@W=6I5/9./6.-4,ZfNgO\hP^jRblTemVhmYinZklZklZljliZnhZnhZphoiliZkhYkgjfid^ie\hfZghXfgUbgQajWerao|s~¥¬«¹ÂÇÐ×Úáçîó÷öüüøüûõúôóöíîòäéêÜãäÔÜÛÉÓÒÀÎË¸ÉÆ³Ã¾«½¶¤Ã·§Ç¹¬ÍÁµÖÉÁßÔÒèÞßóèîøñøýõÿþøÿÿüÿÿüÿýüÿýýÿüþýýÿþþþþþþþþþþþþþþþþþþþÿþüÿþüþýûÿþüÿýúÿýúÿýúÿüùÿû÷ÿû÷ÿýüÿüûþúûÿûüÿýþÿþÿÿýþþüýÿüûÿþûÿýûýôïíâÜæÙÑðâÙüñëÿúýÿüÿÿýþÿÿýÿÿûþÿúýÿþüýÿþýÿþýÿÿþÿÿÿýÿþõÿþíÿüæÿüãþùãÿþçäßÉmPJ0RM0SL/SL/UO/UO/UP2VQ3TP3SO2PO3ON2PL1ON2OM4LO4MO7KQ7JO8HP8JS8GP5GP3IR3NS3PV4W9\d=grHtUjulnuTZaBSW<MO7NN6MM5NL5OM6RP9WR<YT>ZT>XR<TO<QL9KH9GE8EC7AC8=>6:@6:@69A67B47B46B46B4;C8:B7:B79A69A6:B7:B7;C8;C8:B7:B7;C8<D9=E:>F;=G<=G?=I?>H=>I9?G8>F7>D8=B;=D=;B;6@85A3:I2DW7Ri?uE\zFXxFTqEIdA>U;2H3/@0.?-6H2@R8L_CReGSgKOcH@W=6I51;007/-4,YgNZhO\hP^jRblTemVinZjomn\lmmk\mjoioioioiliZkhYjfZieZid^ie\igghXghVchRajWerao|s{¦¢°¹ÁËÔÔÝäéñôòúüôúúñöòíðçèìÞâãÕÚÛËÒÑ¿ÌÉ¸ÉÄ±Ã¾«¿¸¦¼³¢Ê¾®ÓÅ¸ÜÐÄæÙÑíâàöìíÿôúÿùÿÿ÷ÿÿúÿÿüÿÿüÿýüÿüüþüþýþÿÿþþþþþþþþþþþþþþþþþþÿþüÿþüþýûþýûÿýúÿýúÿýúÿüùÿúöþùõÿýüÿûúýùúþúûþüýÿýþÿþÿÿýþÿüûÿýúÿüúÿøó÷ìæòåÝøêáÿôîÿúýÿüÿÿýþÿÿýÿÿûþÿúýÿþüýÿýüÿþýÿÿþÿÿÿýÿþõÿýìÿûåþúáÿüæÿÿéçâÌoRL2TO2TM0UN1UO/UO/VQ3VQ3UQ4SO2ON2NM1PL1NM1NL3KN3LN6KQ7LQ:JR:LU:IR7JS6OX9TY9U9c?fnGc¦s°ª&#x7f; zpr{\\bFLP7LN6LK6LJ5LJ5QL8TM:VP:WQ;VP:SN;QL9MJ;IG:GE9CE:=>6:@6:@69A67B47B46B46B4;C8;C8:B7:B7:B7:B7;C8;C8;C8;C8;C8;C8<D9=E:>F;=G<=G?<H>=G<=H8?G8=E6=C7=B;:A:9@96@85A39H1BU5Ne;Vp@Vt@SsAPmAE`=9P6-C.+<,+<*1C-<N4H?PcERfJNbG@W=5H4,6++2*'.&
\ No newline at end of file

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/testorig12.jpg Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/vgl_5674_0098.bmp Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/vgl_6434_0018a.bmp Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/testimages/vgl_6548_0026a.bmp Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/tjbench.c Added

@@ -0,0 +1,1031 @@
+/*
+ * Copyright (C)2009-2019, 2021 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <ctype.h>
+#include <math.h>
+#include <errno.h>
+#include <limits.h>
+#include <cdjpeg.h>
+#include "./tjutil.h"
+#include "./turbojpeg.h"
+
+
+#define THROW(op, err) { \
+  printf("ERROR in line %d while %s:\n%s\n", __LINE__, op, err); \
+  retval = -1;  goto bailout; \
+}
+#define THROW_UNIX(m)  THROW(m, strerror(errno))
+
+char tjErrorStrJMSG_LENGTH_MAX = "\0", tjErrorMsgJMSG_LENGTH_MAX = "\0";
+int tjErrorLine = -1, tjErrorCode = -1;
+
+#define THROW_TJG(m) { \
+  printf("ERROR in line %d while %s:\n%s\n", __LINE__, m, \
+         tjGetErrorStr2(NULL)); \
+  retval = -1;  goto bailout; \
+}
+
+#define THROW_TJ(m) { \
+  int _tjErrorCode = tjGetErrorCode(handle); \
+  char *_tjErrorStr = tjGetErrorStr2(handle); \
+  \
+  if (!(flags & TJFLAG_STOPONWARNING) && _tjErrorCode == TJERR_WARNING) { \
+    if (strncmp(tjErrorStr, _tjErrorStr, JMSG_LENGTH_MAX) || \
+        strncmp(tjErrorMsg, m, JMSG_LENGTH_MAX) || \
+        tjErrorCode != _tjErrorCode || tjErrorLine != __LINE__) { \
+      strncpy(tjErrorStr, _tjErrorStr, JMSG_LENGTH_MAX - 1); \
+      strncpy(tjErrorMsg, m, JMSG_LENGTH_MAX - 1); \
+      tjErrorCode = _tjErrorCode; \
+      tjErrorLine = __LINE__; \
+      printf("WARNING in line %d while %s:\n%s\n", __LINE__, m, _tjErrorStr); \
+    } \
+  } else { \
+    printf("%s in line %d while %s:\n%s\n", \
+           _tjErrorCode == TJERR_WARNING ? "WARNING" : "ERROR", __LINE__, m, \
+           _tjErrorStr); \
+    retval = -1;  goto bailout; \
+  } \
+}
+
+int flags = TJFLAG_NOREALLOC, compOnly = 0, decompOnly = 0, doYUV = 0,
+  quiet = 0, doTile = 0, pf = TJPF_BGR, yuvPad = 1, doWrite = 1;
+char *ext = "ppm";
+const char *pixFormatStrTJ_NUMPF = {
+  "RGB", "BGR", "RGBX", "BGRX", "XBGR", "XRGB", "GRAY", "", "", "", "", "CMYK"
+};
+const char *subNameLongTJ_NUMSAMP = {
+  "4:4:4", "4:2:2", "4:2:0", "GRAY", "4:4:0", "4:1:1"
+};
+const char *csNameTJ_NUMCS = {
+  "RGB", "YCbCr", "GRAY", "CMYK", "YCCK"
+};
+const char *subNameTJ_NUMSAMP = {
+  "444", "422", "420", "GRAY", "440", "411"
+};
+tjscalingfactor *scalingFactors = NULL, sf = { 1, 1 };
+int nsf = 0, xformOp = TJXOP_NONE, xformOpt = 0;
+int (*customFilter) (short *, tjregion, tjregion, int, int, tjtransform *);
+double benchTime = 5.0, warmup = 1.0;
+
+
+static char *formatName(int subsamp, int cs, char *buf)
+{
+  if (cs == TJCS_YCbCr)
+    return (char *)subNameLongsubsamp;
+  else if (cs == TJCS_YCCK || cs == TJCS_CMYK) {
+    snprintf(buf, 80, "%s %s", csNamecs, subNameLongsubsamp);
+    return buf;
+  } else
+    return (char *)csNamecs;
+}
+
+
+static char *sigfig(double val, int figs, char *buf, int len)
+{
+  char format80;
+  int digitsAfterDecimal = figs - (int)ceil(log10(fabs(val)));
+
+  if (digitsAfterDecimal < 1)
+    snprintf(format, 80, "%%.0f");
+  else
+    snprintf(format, 80, "%%.%df", digitsAfterDecimal);
+  snprintf(buf, len, format, val);
+  return buf;
+}
+
+
+/* Custom DCT filter which produces a negative of the image */
+static int dummyDCTFilter(short *coeffs, tjregion arrayRegion,
+                          tjregion planeRegion, int componentIndex,
+                          int transformIndex, tjtransform *transform)
+{
+  int i;
+
+  for (i = 0; i < arrayRegion.w * arrayRegion.h; i++)
+    coeffsi = -coeffsi;
+  return 0;
+}
+
+
+/* Decompression test */
+static int decomp(unsigned char *srcBuf, unsigned char **jpegBuf,
+                  unsigned long *jpegSize, unsigned char *dstBuf, int w, int h,
+                  int subsamp, int jpegQual, char *fileName, int tilew,
+                  int tileh)
+{
+  char tempStr1024, sizeStr24 = "\0", qualStr13 = "\0", *ptr;
+  FILE *file = NULL;
+  tjhandle handle = NULL;
+  int row, col, iter = 0, dstBufAlloc = 0, retval = 0;
+  double elapsed, elapsedDecode;
+  int ps = tjPixelSizepf;
+  int scaledw = TJSCALED(w, sf);
+  int scaledh = TJSCALED(h, sf);
+  int pitch = scaledw * ps;
+  int ntilesw = (w + tilew - 1) / tilew, ntilesh = (h + tileh - 1) / tileh;
+  unsigned char *dstPtr, *dstPtr2, *yuvBuf = NULL;
+
+  if (jpegQual > 0) {
+    snprintf(qualStr, 13, "_Q%d", jpegQual);
+    qualStr12 = 0;
+  }
+
+  if ((handle = tjInitDecompress()) == NULL)
+    THROW_TJ("executing tjInitDecompress()");
+
+  if (dstBuf == NULL) {
+    if ((unsigned long long)pitch * (unsigned long long)scaledh >
+        (unsigned long long)((size_t)-1))
+      THROW("allocating destination buffer", "Image is too large");
+    if ((dstBuf = (unsigned char *)malloc((size_t)pitch * scaledh)) == NULL)
+      THROW_UNIX("allocating destination buffer");
+    dstBufAlloc = 1;
+  }
+  /* Set the destination buffer to gray so we know whether the decompressor
+     attempted to write to it */
+  memset(dstBuf, 127, (size_t)pitch * scaledh);
+
+  if (doYUV) {
+    int width = doTile ? tilew : scaledw;
+    int height = doTile ? tileh : scaledh;
+    unsigned long yuvSize = tjBufSizeYUV2(width, yuvPad, height, subsamp);
+
+    if (yuvSize == (unsigned long)-1)
+      THROW_TJ("allocating YUV buffer");
+    if ((yuvBuf = (unsigned char *)malloc(yuvSize)) == NULL)
+      THROW_UNIX("allocating YUV buffer");
+    memset(yuvBuf, 127, yuvSize);
+  }
+
+  /* Benchmark */
+  iter = -1;
+  elapsed = elapsedDecode = 0.;
+  while (1) {
+    int tile = 0;
+    double start = getTime();
+
+    for (row = 0, dstPtr = dstBuf; row < ntilesh;
+         row++, dstPtr += (size_t)pitch * tileh) {
+      for (col = 0, dstPtr2 = dstPtr; col < ntilesw;
+           col++, tile++, dstPtr2 += ps * tilew) {
+        int width = doTile ? min(tilew, w - col * tilew) : scaledw;
+        int height = doTile ? min(tileh, h - row * tileh) : scaledh;
+
+        if (doYUV) {
+          double startDecode;
+
+          if (tjDecompressToYUV2(handle, jpegBuftile, jpegSizetile, yuvBuf,
+                                 width, yuvPad, height, flags) == -1)
+            THROW_TJ("executing tjDecompressToYUV2()");
+          startDecode = getTime();
+          if (tjDecodeYUV(handle, yuvBuf, yuvPad, subsamp, dstPtr2, width,
+                          pitch, height, pf, flags) == -1)
+            THROW_TJ("executing tjDecodeYUV()");
+          if (iter >= 0) elapsedDecode += getTime() - startDecode;
+        } else if (tjDecompress2(handle, jpegBuftile, jpegSizetile,
+                                 dstPtr2, width, pitch, height, pf,
+                                 flags) == -1)
+          THROW_TJ("executing tjDecompress2()");
+      }
+    }
+    elapsed += getTime() - start;
+    if (iter >= 0) {
+      iter++;
+      if (elapsed >= benchTime) break;
+    } else if (elapsed >= warmup) {
+      iter = 0;
+      elapsed = elapsedDecode = 0.;
+    }
+  }
+  if (doYUV) elapsed -= elapsedDecode;
+
+  if (tjDestroy(handle) == -1) THROW_TJ("executing tjDestroy()");
+  handle = NULL;
+
+  if (quiet) {
+    printf("%-6s%s",
+           sigfig((double)(w * h) / 1000000. * (double)iter / elapsed, 4,
+                  tempStr, 1024),
+           quiet == 2 ? "\n" : "  ");
+    if (doYUV)
+      printf("%s\n",
+             sigfig((double)(w * h) / 1000000. * (double)iter / elapsedDecode,
+                    4, tempStr, 1024));
+    else if (quiet != 2) printf("\n");
+  } else {
+    printf("%s --> Frame rate:         %f fps\n",
+           doYUV ? "Decomp to YUV" : "Decompress   ", (double)iter / elapsed);
+    printf("                  Throughput:         %f Megapixels/sec\n",
+           (double)(w * h) / 1000000. * (double)iter / elapsed);
+    if (doYUV) {
+      printf("YUV Decode    --> Frame rate:         %f fps\n",
+             (double)iter / elapsedDecode);
+      printf("                  Throughput:         %f Megapixels/sec\n",
+             (double)(w * h) / 1000000. * (double)iter / elapsedDecode);
+    }
+  }
+
+  if (!doWrite) goto bailout;
+
+  if (sf.num != 1 || sf.denom != 1)
+    snprintf(sizeStr, 24, "%d_%d", sf.num, sf.denom);
+  else if (tilew != w || tileh != h)
+    snprintf(sizeStr, 24, "%dx%d", tilew, tileh);
+  else snprintf(sizeStr, 24, "full");
+  if (decompOnly)
+    snprintf(tempStr, 1024, "%s_%s.%s", fileName, sizeStr, ext);
+  else
+    snprintf(tempStr, 1024, "%s_%s%s_%s.%s", fileName, subNamesubsamp,
+             qualStr, sizeStr, ext);
+
+  if (tjSaveImage(tempStr, dstBuf, scaledw, 0, scaledh, pf, flags) == -1)
+    THROW_TJG("saving bitmap");
+  ptr = strrchr(tempStr, '.');
+  snprintf(ptr, 1024 - (ptr - tempStr), "-err.%s", ext);
+  if (srcBuf && sf.num == 1 && sf.denom == 1) {
+    if (!quiet) printf("Compression error written to %s.\n", tempStr);
+    if (subsamp == TJ_GRAYSCALE) {
+      unsigned long index, index2;
+
+      for (row = 0, index = 0; row < h; row++, index += pitch) {
+        for (col = 0, index2 = index; col < w; col++, index2 += ps) {
+          unsigned long rindex = index2 + tjRedOffsetpf;
+          unsigned long gindex = index2 + tjGreenOffsetpf;
+          unsigned long bindex = index2 + tjBlueOffsetpf;
+          int y = (int)((double)srcBufrindex * 0.299 +
+                        (double)srcBufgindex * 0.587 +
+                        (double)srcBufbindex * 0.114 + 0.5);
+
+          if (y > 255) y = 255;
+          if (y < 0) y = 0;
+          dstBufrindex = abs(dstBufrindex - y);
+          dstBufgindex = abs(dstBufgindex - y);
+          dstBufbindex = abs(dstBufbindex - y);
+        }
+      }
+    } else {
+      for (row = 0; row < h; row++)
+        for (col = 0; col < w * ps; col++)
+          dstBufpitch * row + col =
+            abs(dstBufpitch * row + col - srcBufpitch * row + col);
+    }
+    if (tjSaveImage(tempStr, dstBuf, w, 0, h, pf, flags) == -1)
+      THROW_TJG("saving bitmap");
+  }
+
+bailout:
+  if (file) fclose(file);
+  if (handle) tjDestroy(handle);
+  if (dstBufAlloc) free(dstBuf);
+  free(yuvBuf);
+  return retval;
+}
+
+
+static int fullTest(unsigned char *srcBuf, int w, int h, int subsamp,
+                    int jpegQual, char *fileName)
+{
+  char tempStr1024, tempStr280;
+  FILE *file = NULL;
+  tjhandle handle = NULL;
+  unsigned char **jpegBuf = NULL, *yuvBuf = NULL, *tmpBuf = NULL, *srcPtr,
+    *srcPtr2;
+  double start, elapsed, elapsedEncode;
+  int totalJpegSize = 0, row, col, i, tilew = w, tileh = h, retval = 0;
+  int iter;
+  unsigned long *jpegSize = NULL, yuvSize = 0;
+  int ps = tjPixelSizepf;
+  int ntilesw = 1, ntilesh = 1, pitch = w * ps;
+  const char *pfStr = pixFormatStrpf;
+
+  if ((unsigned long long)pitch * (unsigned long long)h >
+      (unsigned long long)((size_t)-1))
+    THROW("allocating temporary image buffer", "Image is too large");
+  if ((tmpBuf = (unsigned char *)malloc((size_t)pitch * h)) == NULL)
+    THROW_UNIX("allocating temporary image buffer");
+
+  if (!quiet)
+    printf(">>>>>  %s (%s) <--> JPEG %s Q%d  <<<<<\n", pfStr,
+           (flags & TJFLAG_BOTTOMUP) ? "Bottom-up" : "Top-down",
+           subNameLongsubsamp, jpegQual);
+
+  for (tilew = doTile ? 8 : w, tileh = doTile ? 8 : h; ;
+       tilew *= 2, tileh *= 2) {
+    if (tilew > w) tilew = w;
+    if (tileh > h) tileh = h;
+    ntilesw = (w + tilew - 1) / tilew;
+    ntilesh = (h + tileh - 1) / tileh;
+
+    if ((jpegBuf = (unsigned char **)malloc(sizeof(unsigned char *) *
+                                            ntilesw * ntilesh)) == NULL)
+      THROW_UNIX("allocating JPEG tile array");
+    memset(jpegBuf, 0, sizeof(unsigned char *) * ntilesw * ntilesh);
+    if ((jpegSize = (unsigned long *)malloc(sizeof(unsigned long) *
+                                            ntilesw * ntilesh)) == NULL)
+      THROW_UNIX("allocating JPEG size array");
+    memset(jpegSize, 0, sizeof(unsigned long) * ntilesw * ntilesh);
+
+    if ((flags & TJFLAG_NOREALLOC) != 0)
+      for (i = 0; i < ntilesw * ntilesh; i++) {
+        if (tjBufSize(tilew, tileh, subsamp) > (unsigned long)INT_MAX)
+          THROW("getting buffer size", "Image is too large");
+        if ((jpegBufi = (unsigned char *)
+                          tjAlloc(tjBufSize(tilew, tileh, subsamp))) == NULL)
+          THROW_UNIX("allocating JPEG tiles");
+      }
+
+    /* Compression test */
+    if (quiet == 1)
+      printf("%-4s (%s)  %-5s    %-3d   ", pfStr,
+             (flags & TJFLAG_BOTTOMUP) ? "BU" : "TD", subNameLongsubsamp,
+             jpegQual);
+    for (i = 0; i < h; i++)
+      memcpy(&tmpBufpitch * i, &srcBufw * ps * i, w * ps);
+    if ((handle = tjInitCompress()) == NULL)
+      THROW_TJ("executing tjInitCompress()");
+
+    if (doYUV) {
+      yuvSize = tjBufSizeYUV2(tilew, yuvPad, tileh, subsamp);
+      if (yuvSize == (unsigned long)-1)
+        THROW_TJ("allocating YUV buffer");
+      if ((yuvBuf = (unsigned char *)malloc(yuvSize)) == NULL)
+        THROW_UNIX("allocating YUV buffer");
+      memset(yuvBuf, 127, yuvSize);
+    }
+
+    /* Benchmark */
+    iter = -1;
+    elapsed = elapsedEncode = 0.;
+    while (1) {
+      int tile = 0;
+
+      totalJpegSize = 0;
+      start = getTime();
+      for (row = 0, srcPtr = srcBuf; row < ntilesh;
+           row++, srcPtr += pitch * tileh) {
+        for (col = 0, srcPtr2 = srcPtr; col < ntilesw;
+             col++, tile++, srcPtr2 += ps * tilew) {
+          int width = min(tilew, w - col * tilew);
+          int height = min(tileh, h - row * tileh);
+
+          if (doYUV) {
+            double startEncode = getTime();
+
+            if (tjEncodeYUV3(handle, srcPtr2, width, pitch, height, pf, yuvBuf,
+                             yuvPad, subsamp, flags) == -1)
+              THROW_TJ("executing tjEncodeYUV3()");
+            if (iter >= 0) elapsedEncode += getTime() - startEncode;
+            if (tjCompressFromYUV(handle, yuvBuf, width, yuvPad, height,
+                                  subsamp, &jpegBuftile, &jpegSizetile,
+                                  jpegQual, flags) == -1)
+              THROW_TJ("executing tjCompressFromYUV()");
+          } else {
+            if (tjCompress2(handle, srcPtr2, width, pitch, height, pf,
+                            &jpegBuftile, &jpegSizetile, subsamp, jpegQual,
+                            flags) == -1)
+              THROW_TJ("executing tjCompress2()");
+          }
+          totalJpegSize += jpegSizetile;
+        }
+      }
+      elapsed += getTime() - start;
+      if (iter >= 0) {
+        iter++;
+        if (elapsed >= benchTime) break;
+      } else if (elapsed >= warmup) {
+        iter = 0;
+        elapsed = elapsedEncode = 0.;
+      }
+    }
+    if (doYUV) elapsed -= elapsedEncode;
+
+    if (tjDestroy(handle) == -1) THROW_TJ("executing tjDestroy()");
+    handle = NULL;
+
+    if (quiet == 1) printf("%-5d  %-5d   ", tilew, tileh);
+    if (quiet) {
+      if (doYUV)
+        printf("%-6s%s",
+               sigfig((double)(w * h) / 1000000. *
+                      (double)iter / elapsedEncode, 4, tempStr, 1024),
+               quiet == 2 ? "\n" : "  ");
+      printf("%-6s%s",
+             sigfig((double)(w * h) / 1000000. * (double)iter / elapsed, 4,
+                    tempStr, 1024),
+             quiet == 2 ? "\n" : "  ");
+      printf("%-6s%s",
+             sigfig((double)(w * h * ps) / (double)totalJpegSize, 4, tempStr2,
+                    80),
+             quiet == 2 ? "\n" : "  ");
+    } else {
+      printf("\n%s size: %d x %d\n", doTile ? "Tile" : "Image", tilew, tileh);
+      if (doYUV) {
+        printf("Encode YUV    --> Frame rate:         %f fps\n",
+               (double)iter / elapsedEncode);
+        printf("                  Output image size:  %lu bytes\n", yuvSize);
+        printf("                  Compression ratio:  %f:1\n",
+               (double)(w * h * ps) / (double)yuvSize);
+        printf("                  Throughput:         %f Megapixels/sec\n",
+               (double)(w * h) / 1000000. * (double)iter / elapsedEncode);
+        printf("                  Output bit stream:  %f Megabits/sec\n",
+               (double)yuvSize * 8. / 1000000. * (double)iter / elapsedEncode);
+      }
+      printf("%s --> Frame rate:         %f fps\n",
+             doYUV ? "Comp from YUV" : "Compress     ",
+             (double)iter / elapsed);
+      printf("                  Output image size:  %d bytes\n",
+             totalJpegSize);
+      printf("                  Compression ratio:  %f:1\n",
+             (double)(w * h * ps) / (double)totalJpegSize);
+      printf("                  Throughput:         %f Megapixels/sec\n",
+             (double)(w * h) / 1000000. * (double)iter / elapsed);
+      printf("                  Output bit stream:  %f Megabits/sec\n",
+             (double)totalJpegSize * 8. / 1000000. * (double)iter / elapsed);
+    }
+    if (tilew == w && tileh == h && doWrite) {
+      snprintf(tempStr, 1024, "%s_%s_Q%d.jpg", fileName, subNamesubsamp,
+               jpegQual);
+      if ((file = fopen(tempStr, "wb")) == NULL)
+        THROW_UNIX("opening reference image");
+      if (fwrite(jpegBuf0, jpegSize0, 1, file) != 1)
+        THROW_UNIX("writing reference image");
+      fclose(file);  file = NULL;
+      if (!quiet) printf("Reference image written to %s\n", tempStr);
+    }
+
+    /* Decompression test */
+    if (!compOnly) {
+      if (decomp(srcBuf, jpegBuf, jpegSize, tmpBuf, w, h, subsamp, jpegQual,
+                 fileName, tilew, tileh) == -1)
+        goto bailout;
+    } else if (quiet == 1) printf("N/A\n");
+
+    for (i = 0; i < ntilesw * ntilesh; i++) {
+      tjFree(jpegBufi);
+      jpegBufi = NULL;
+    }
+    free(jpegBuf);  jpegBuf = NULL;
+    free(jpegSize);  jpegSize = NULL;
+    if (doYUV) {
+      free(yuvBuf);  yuvBuf = NULL;
+    }
+
+    if (tilew == w && tileh == h) break;
+  }
+
+bailout:
+  if (file) fclose(file);
+  if (jpegBuf) {
+    for (i = 0; i < ntilesw * ntilesh; i++)
+      tjFree(jpegBufi);
+  }
+  free(jpegBuf);
+  free(yuvBuf);
+  free(jpegSize);
+  free(tmpBuf);
+  if (handle) tjDestroy(handle);
+  return retval;
+}
+
+
+static int decompTest(char *fileName)
+{
+  FILE *file = NULL;
+  tjhandle handle = NULL;
+  unsigned char **jpegBuf = NULL, *srcBuf = NULL;
+  unsigned long *jpegSize = NULL, srcSize, totalJpegSize;
+  tjtransform *t = NULL;
+  double start, elapsed;
+  int ps = tjPixelSizepf, tile, row, col, i, iter, retval = 0, decompsrc = 0;
+  char *temp = NULL, tempStr80, tempStr280;
+  /* Original image */
+  int w = 0, h = 0, tilew, tileh, ntilesw = 1, ntilesh = 1, subsamp = -1,
+    cs = -1;
+  /* Transformed image */
+  int tw, th, ttilew, ttileh, tntilesw, tntilesh, tsubsamp;
+
+  if ((file = fopen(fileName, "rb")) == NULL)
+    THROW_UNIX("opening file");
+  if (fseek(file, 0, SEEK_END) < 0 ||
+      (srcSize = ftell(file)) == (unsigned long)-1)
+    THROW_UNIX("determining file size");
+  if ((srcBuf = (unsigned char *)malloc(srcSize)) == NULL)
+    THROW_UNIX("allocating memory");
+  if (fseek(file, 0, SEEK_SET) < 0)
+    THROW_UNIX("setting file position");
+  if (fread(srcBuf, srcSize, 1, file) < 1)
+    THROW_UNIX("reading JPEG data");
+  fclose(file);  file = NULL;
+
+  temp = strrchr(fileName, '.');
+  if (temp != NULL) *temp = '\0';
+
+  if ((handle = tjInitTransform()) == NULL)
+    THROW_TJ("executing tjInitTransform()");
+  if (tjDecompressHeader3(handle, srcBuf, srcSize, &w, &h, &subsamp,
+                          &cs) == -1)
+    THROW_TJ("executing tjDecompressHeader3()");
+  if (w < 1 || h < 1)
+    THROW("reading JPEG header", "Invalid image dimensions");
+  if (cs == TJCS_YCCK || cs == TJCS_CMYK) {
+    pf = TJPF_CMYK;  ps = tjPixelSizepf;
+  }
+
+  if (quiet == 1) {
+    printf("All performance values in Mpixels/sec\n\n");
+    printf("Bitmap     JPEG   JPEG     %s  %s   Xform   Comp    Decomp  ",
+           doTile ? "Tile " : "Image", doTile ? "Tile " : "Image");
+    if (doYUV) printf("Decode");
+    printf("\n");
+    printf("Format     CS     Subsamp  Width  Height  Perf    Ratio   Perf    ");
+    if (doYUV) printf("Perf");
+    printf("\n\n");
+  } else if (!quiet)
+    printf(">>>>>  JPEG %s --> %s (%s)  <<<<<\n",
+           formatName(subsamp, cs, tempStr), pixFormatStrpf,
+           (flags & TJFLAG_BOTTOMUP) ? "Bottom-up" : "Top-down");
+
+  for (tilew = doTile ? 16 : w, tileh = doTile ? 16 : h; ;
+       tilew *= 2, tileh *= 2) {
+    if (tilew > w) tilew = w;
+    if (tileh > h) tileh = h;
+    ntilesw = (w + tilew - 1) / tilew;
+    ntilesh = (h + tileh - 1) / tileh;
+
+    if ((jpegBuf = (unsigned char **)malloc(sizeof(unsigned char *) *
+                                            ntilesw * ntilesh)) == NULL)
+      THROW_UNIX("allocating JPEG tile array");
+    memset(jpegBuf, 0, sizeof(unsigned char *) * ntilesw * ntilesh);
+    if ((jpegSize = (unsigned long *)malloc(sizeof(unsigned long) *
+                                            ntilesw * ntilesh)) == NULL)
+      THROW_UNIX("allocating JPEG size array");
+    memset(jpegSize, 0, sizeof(unsigned long) * ntilesw * ntilesh);
+
+    if ((flags & TJFLAG_NOREALLOC) != 0 &&
+        (doTile || xformOp != TJXOP_NONE || xformOpt != 0 || customFilter))
+      for (i = 0; i < ntilesw * ntilesh; i++) {
+        if (tjBufSize(tilew, tileh, subsamp) > (unsigned long)INT_MAX)
+          THROW("getting buffer size", "Image is too large");
+        if ((jpegBufi = (unsigned char *)
+                          tjAlloc(tjBufSize(tilew, tileh, subsamp))) == NULL)
+          THROW_UNIX("allocating JPEG tiles");
+      }
+
+    tw = w;  th = h;  ttilew = tilew;  ttileh = tileh;
+    if (!quiet) {
+      printf("\n%s size: %d x %d", doTile ? "Tile" : "Image", ttilew, ttileh);
+      if (sf.num != 1 || sf.denom != 1)
+        printf(" --> %d x %d", TJSCALED(tw, sf), TJSCALED(th, sf));
+      printf("\n");
+    } else if (quiet == 1) {
+      printf("%-4s (%s)  %-5s  %-5s    ", pixFormatStrpf,
+             (flags & TJFLAG_BOTTOMUP) ? "BU" : "TD", csNamecs,
+             subNameLongsubsamp);
+      printf("%-5d  %-5d   ", tilew, tileh);
+    }
+
+    tsubsamp = subsamp;
+    if (doTile || xformOp != TJXOP_NONE || xformOpt != 0 || customFilter) {
+      if ((t = (tjtransform *)malloc(sizeof(tjtransform) * ntilesw *
+                                     ntilesh)) == NULL)
+        THROW_UNIX("allocating image transform array");
+
+      if (xformOp == TJXOP_TRANSPOSE || xformOp == TJXOP_TRANSVERSE ||
+          xformOp == TJXOP_ROT90 || xformOp == TJXOP_ROT270) {
+        tw = h;  th = w;  ttilew = tileh;  ttileh = tilew;
+      }
+
+      if (xformOpt & TJXOPT_GRAY) tsubsamp = TJ_GRAYSCALE;
+      if (xformOp == TJXOP_HFLIP || xformOp == TJXOP_ROT180)
+        tw = tw - (tw % tjMCUWidthtsubsamp);
+      if (xformOp == TJXOP_VFLIP || xformOp == TJXOP_ROT180)
+        th = th - (th % tjMCUHeighttsubsamp);
+      if (xformOp == TJXOP_TRANSVERSE || xformOp == TJXOP_ROT90)
+        tw = tw - (tw % tjMCUHeighttsubsamp);
+      if (xformOp == TJXOP_TRANSVERSE || xformOp == TJXOP_ROT270)
+        th = th - (th % tjMCUWidthtsubsamp);
+      tntilesw = (tw + ttilew - 1) / ttilew;
+      tntilesh = (th + ttileh - 1) / ttileh;
+
+      if (xformOp == TJXOP_TRANSPOSE || xformOp == TJXOP_TRANSVERSE ||
+          xformOp == TJXOP_ROT90 || xformOp == TJXOP_ROT270) {
+        if (tsubsamp == TJSAMP_422) tsubsamp = TJSAMP_440;
+        else if (tsubsamp == TJSAMP_440) tsubsamp = TJSAMP_422;
+      }
+
+      for (row = 0, tile = 0; row < tntilesh; row++) {
+        for (col = 0; col < tntilesw; col++, tile++) {
+          ttile.r.w = min(ttilew, tw - col * ttilew);
+          ttile.r.h = min(ttileh, th - row * ttileh);
+          ttile.r.x = col * ttilew;
+          ttile.r.y = row * ttileh;
+          ttile.op = xformOp;
+          ttile.options = xformOpt | TJXOPT_TRIM;
+          ttile.customFilter = customFilter;
+          if (ttile.options & TJXOPT_NOOUTPUT && jpegBuftile) {
+            tjFree(jpegBuftile);  jpegBuftile = NULL;
+          }
+        }
+      }
+
+      iter = -1;
+      elapsed = 0.;
+      while (1) {
+        start = getTime();
+        if (tjTransform(handle, srcBuf, srcSize, tntilesw * tntilesh, jpegBuf,
+                        jpegSize, t, flags) == -1)
+          THROW_TJ("executing tjTransform()");
+        elapsed += getTime() - start;
+        if (iter >= 0) {
+          iter++;
+          if (elapsed >= benchTime) break;
+        } else if (elapsed >= warmup) {
+          iter = 0;
+          elapsed = 0.;
+        }
+      }
+
+      free(t);  t = NULL;
+
+      for (tile = 0, totalJpegSize = 0; tile < tntilesw * tntilesh; tile++)
+        totalJpegSize += jpegSizetile;
+
+      if (quiet) {
+        printf("%-6s%s%-6s%s",
+               sigfig((double)(w * h) / 1000000. / elapsed, 4, tempStr, 80),
+               quiet == 2 ? "\n" : "  ",
+               sigfig((double)(w * h * ps) / (double)totalJpegSize, 4,
+                      tempStr2, 80),
+               quiet == 2 ? "\n" : "  ");
+      } else if (!quiet) {
+        printf("Transform     --> Frame rate:         %f fps\n",
+               1.0 / elapsed);
+        printf("                  Output image size:  %lu bytes\n",
+               totalJpegSize);
+        printf("                  Compression ratio:  %f:1\n",
+               (double)(w * h * ps) / (double)totalJpegSize);
+        printf("                  Throughput:         %f Megapixels/sec\n",
+               (double)(w * h) / 1000000. / elapsed);
+        printf("                  Output bit stream:  %f Megabits/sec\n",
+               (double)totalJpegSize * 8. / 1000000. / elapsed);
+      }
+    } else {
+      if (quiet == 1) printf("N/A     N/A     ");
+      tjFree(jpegBuf0);
+      jpegBuf0 = NULL;
+      decompsrc = 1;
+    }
+
+    if (w == tilew) ttilew = tw;
+    if (h == tileh) ttileh = th;
+    if (!(xformOpt & TJXOPT_NOOUTPUT)) {
+      if (decomp(NULL, decompsrc ? &srcBuf : jpegBuf,
+                 decompsrc ? &srcSize : jpegSize, NULL, tw, th, tsubsamp, 0,
+                 fileName, ttilew, ttileh) == -1)
+        goto bailout;
+    } else if (quiet == 1) printf("N/A\n");
+
+    for (i = 0; i < ntilesw * ntilesh; i++) {
+      tjFree(jpegBufi);
+      jpegBufi = NULL;
+    }
+    free(jpegBuf);  jpegBuf = NULL;
+    free(jpegSize);  jpegSize = NULL;
+
+    if (tilew == w && tileh == h) break;
+  }
+
+bailout:
+  if (file) fclose(file);
+  if (jpegBuf) {
+    for (i = 0; i < ntilesw * ntilesh; i++)
+      tjFree(jpegBufi);
+  }
+  free(jpegBuf);
+  free(jpegSize);
+  free(srcBuf);
+  free(t);
+  if (handle) { tjDestroy(handle);  handle = NULL; }
+  return retval;
+}
+
+
+static void usage(char *progName)
+{
+  int i;
+
+  printf("USAGE: %s\n", progName);
+  printf("       <Inputfile (BMP|PPM)> <Quality> options\n\n");
+  printf("       %s\n", progName);
+  printf("       <Inputfile (JPG)> options\n\n");
+  printf("Options:\n\n");
+  printf("-alloc = Dynamically allocate JPEG image buffers\n");
+  printf("-bmp = Generate output images in Windows Bitmap format (default = PPM)\n");
+  printf("-bottomup = Test bottom-up compression/decompression\n");
+  printf("-tile = Test performance of the codec when the image is encoded as separate\n");
+  printf("     tiles of varying sizes.\n");
+  printf("-rgb, -bgr, -rgbx, -bgrx, -xbgr, -xrgb =\n");
+  printf("     Test the specified color conversion path in the codec (default = BGR)\n");
+  printf("-cmyk = Indirectly test YCCK JPEG compression/decompression (the source\n");
+  printf("     and destination bitmaps are still RGB.  The conversion is done\n");
+  printf("     internally prior to compression or after decompression.)\n");
+  printf("-fastupsample = Use the fastest chrominance upsampling algorithm available in\n");
+  printf("     the underlying codec\n");
+  printf("-fastdct = Use the fastest DCT/IDCT algorithms available in the underlying\n");
+  printf("     codec\n");
+  printf("-accuratedct = Use the most accurate DCT/IDCT algorithms available in the\n");
+  printf("     underlying codec\n");
+  printf("-progressive = Use progressive entropy coding in JPEG images generated by\n");
+  printf("     compression and transform operations.\n");
+  printf("-subsamp <s> = When testing JPEG compression, this option specifies the level\n");
+  printf("     of chrominance subsampling to use (<s> = 444, 422, 440, 420, 411, or\n");
+  printf("     GRAY).  The default is to test Grayscale, 4:2:0, 4:2:2, and 4:4:4 in\n");
+  printf("     sequence.\n");
+  printf("-quiet = Output results in tabular rather than verbose format\n");
+  printf("-yuv = Test YUV encoding/decoding functions\n");
+  printf("-yuvpad <p> = If testing YUV encoding/decoding, this specifies the number of\n");
+  printf("     bytes to which each row of each plane in the intermediate YUV image is\n");
+  printf("     padded (default = 1)\n");
+  printf("-scale M/N = Scale down the width/height of the decompressed JPEG image by a\n");
+  printf("     factor of M/N (M/N = ");
+  for (i = 0; i < nsf; i++) {
+    printf("%d/%d", scalingFactorsi.num, scalingFactorsi.denom);
+    if (nsf == 2 && i != nsf - 1) printf(" or ");
+    else if (nsf > 2) {
+      if (i != nsf - 1) printf(", ");
+      if (i == nsf - 2) printf("or ");
+    }
+    if (i % 8 == 0 && i != 0) printf("\n     ");
+  }
+  printf(")\n");
+  printf("-hflip, -vflip, -transpose, -transverse, -rot90, -rot180, -rot270 =\n");
+  printf("     Perform the corresponding lossless transform prior to\n");
+  printf("     decompression (these options are mutually exclusive)\n");
+  printf("-grayscale = Perform lossless grayscale conversion prior to decompression\n");
+  printf("     test (can be combined with the other transforms above)\n");
+  printf("-copynone = Do not copy any extra markers (including EXIF and ICC profile data)\n");
+  printf("     when transforming the image.\n");
+  printf("-benchtime <t> = Run each benchmark for at least <t> seconds (default = 5.0)\n");
+  printf("-warmup <t> = Run each benchmark for <t> seconds (default = 1.0) prior to\n");
+  printf("     starting the timer, in order to prime the caches and thus improve the\n");
+  printf("     consistency of the results.\n");
+  printf("-componly = Stop after running compression tests.  Do not test decompression.\n");
+  printf("-nowrite = Do not write reference or output images (improves consistency of\n");
+  printf("     performance measurements.)\n");
+  printf("-limitscans = Refuse to decompress or transform progressive JPEG images that\n");
+  printf("     have an unreasonably large number of scans\n");
+  printf("-stoponwarning = Immediately discontinue the current\n");
+  printf("     compression/decompression/transform operation if the underlying codec\n");
+  printf("     throws a warning (non-fatal error)\n\n");
+  printf("NOTE:  If the quality is specified as a range (e.g. 90-100), a separate\n");
+  printf("test will be performed for all quality values in the range.\n\n");
+  exit(1);
+}
+
+
+int main(int argc, char *argv)
+{
+  unsigned char *srcBuf = NULL;
+  int w = 0, h = 0, i, j, minQual = -1, maxQual = -1;
+  char *temp;
+  int minArg = 2, retval = 0, subsamp = -1;
+
+  if ((scalingFactors = tjGetScalingFactors(&nsf)) == NULL || nsf == 0)
+    THROW("executing tjGetScalingFactors()", tjGetErrorStr());
+
+  if (argc < minArg) usage(argv0);
+
+  temp = strrchr(argv1, '.');
+  if (temp != NULL) {
+    if (!strcasecmp(temp, ".bmp")) ext = "bmp";
+    if (!strcasecmp(temp, ".jpg") || !strcasecmp(temp, ".jpeg"))
+      decompOnly = 1;
+  }
+
+  printf("\n");
+
+  if (!decompOnly) {
+    minArg = 3;
+    if (argc < minArg) usage(argv0);
+    if ((minQual = atoi(argv2)) < 1 || minQual > 100) {
+      puts("ERROR: Quality must be between 1 and 100.");
+      exit(1);
+    }
+    if ((temp = strchr(argv2, '-')) != NULL && strlen(temp) > 1 &&
+        sscanf(&temp1, "%d", &maxQual) == 1 && maxQual > minQual &&
+        maxQual >= 1 && maxQual <= 100) {}
+    else maxQual = minQual;
+  }
+
+  if (argc > minArg) {
+    for (i = minArg; i < argc; i++) {
+      if (!strcasecmp(argvi, "-tile")) {
+        doTile = 1;  xformOpt |= TJXOPT_CROP;
+      } else if (!strcasecmp(argvi, "-fastupsample")) {
+        printf("Using fast upsampling code\n\n");
+        flags |= TJFLAG_FASTUPSAMPLE;
+      } else if (!strcasecmp(argvi, "-fastdct")) {
+        printf("Using fastest DCT/IDCT algorithm\n\n");
+        flags |= TJFLAG_FASTDCT;
+      } else if (!strcasecmp(argvi, "-accuratedct")) {
+        printf("Using most accurate DCT/IDCT algorithm\n\n");
+        flags |= TJFLAG_ACCURATEDCT;
+      } else if (!strcasecmp(argvi, "-progressive")) {
+        printf("Using progressive entropy coding\n\n");
+        flags |= TJFLAG_PROGRESSIVE;
+      } else if (!strcasecmp(argvi, "-rgb"))
+        pf = TJPF_RGB;
+      else if (!strcasecmp(argvi, "-rgbx"))
+        pf = TJPF_RGBX;
+      else if (!strcasecmp(argvi, "-bgr"))
+        pf = TJPF_BGR;
+      else if (!strcasecmp(argvi, "-bgrx"))
+        pf = TJPF_BGRX;
+      else if (!strcasecmp(argvi, "-xbgr"))
+        pf = TJPF_XBGR;
+      else if (!strcasecmp(argvi, "-xrgb"))
+        pf = TJPF_XRGB;
+      else if (!strcasecmp(argvi, "-cmyk"))
+        pf = TJPF_CMYK;
+      else if (!strcasecmp(argvi, "-bottomup"))
+        flags |= TJFLAG_BOTTOMUP;
+      else if (!strcasecmp(argvi, "-quiet"))
+        quiet = 1;
+      else if (!strcasecmp(argvi, "-qq"))
+        quiet = 2;
+      else if (!strcasecmp(argvi, "-scale") && i < argc - 1) {
+        int temp1 = 0, temp2 = 0, match = 0;
+
+        if (sscanf(argv++i, "%d/%d", &temp1, &temp2) == 2) {
+          for (j = 0; j < nsf; j++) {
+            if ((double)temp1 / (double)temp2 ==
+                (double)scalingFactorsj.num /
+                (double)scalingFactorsj.denom) {
+              sf = scalingFactorsj;
+              match = 1;  break;
+            }
+          }
+          if (!match) usage(argv0);
+        } else usage(argv0);
+      } else if (!strcasecmp(argvi, "-hflip"))
+        xformOp = TJXOP_HFLIP;
+      else if (!strcasecmp(argvi, "-vflip"))
+        xformOp = TJXOP_VFLIP;
+      else if (!strcasecmp(argvi, "-transpose"))
+        xformOp = TJXOP_TRANSPOSE;
+      else if (!strcasecmp(argvi, "-transverse"))
+        xformOp = TJXOP_TRANSVERSE;
+      else if (!strcasecmp(argvi, "-rot90"))
+        xformOp = TJXOP_ROT90;
+      else if (!strcasecmp(argvi, "-rot180"))
+        xformOp = TJXOP_ROT180;
+      else if (!strcasecmp(argvi, "-rot270"))
+        xformOp = TJXOP_ROT270;
+      else if (!strcasecmp(argvi, "-grayscale"))
+        xformOpt |= TJXOPT_GRAY;
+      else if (!strcasecmp(argvi, "-custom"))
+        customFilter = dummyDCTFilter;
+      else if (!strcasecmp(argvi, "-nooutput"))
+        xformOpt |= TJXOPT_NOOUTPUT;
+      else if (!strcasecmp(argvi, "-copynone"))
+        xformOpt |= TJXOPT_COPYNONE;
+      else if (!strcasecmp(argvi, "-benchtime") && i < argc - 1) {
+        double tempd = atof(argv++i);
+
+        if (tempd > 0.0) benchTime = tempd;
+        else usage(argv0);
+      } else if (!strcasecmp(argvi, "-warmup") && i < argc - 1) {
+        double tempd = atof(argv++i);
+
+        if (tempd >= 0.0) warmup = tempd;
+        else usage(argv0);
+        printf("Warmup time = %.1f seconds\n\n", warmup);
+      } else if (!strcasecmp(argvi, "-alloc"))
+        flags &= (~TJFLAG_NOREALLOC);
+      else if (!strcasecmp(argvi, "-bmp"))
+        ext = "bmp";
+      else if (!strcasecmp(argvi, "-yuv")) {
+        printf("Testing YUV planar encoding/decoding\n\n");
+        doYUV = 1;
+      } else if (!strcasecmp(argvi, "-yuvpad") && i < argc - 1) {
+        int tempi = atoi(argv++i);
+
+        if (tempi >= 1) yuvPad = tempi;
+      } else if (!strcasecmp(argvi, "-subsamp") && i < argc - 1) {
+        i++;
+        if (toupper(argvi0) == 'G') subsamp = TJSAMP_GRAY;
+        else {
+          int tempi = atoi(argvi);
+
+          switch (tempi) {
+          case 444:  subsamp = TJSAMP_444;  break;
+          case 422:  subsamp = TJSAMP_422;  break;
+          case 440:  subsamp = TJSAMP_440;  break;
+          case 420:  subsamp = TJSAMP_420;  break;
+          case 411:  subsamp = TJSAMP_411;  break;
+          }
+        }
+      } else if (!strcasecmp(argvi, "-componly"))
+        compOnly = 1;
+      else if (!strcasecmp(argvi, "-nowrite"))
+        doWrite = 0;
+      else if (!strcasecmp(argvi, "-limitscans"))
+        flags |= TJFLAG_LIMITSCANS;
+      else if (!strcasecmp(argvi, "-stoponwarning"))
+        flags |= TJFLAG_STOPONWARNING;
+      else usage(argv0);
+    }
+  }
+
+  if ((sf.num != 1 || sf.denom != 1) && doTile) {
+    printf("Disabling tiled compression/decompression tests, because those tests do not\n");
+    printf("work when scaled decompression is enabled.\n");
+    doTile = 0;
+  }
+
+  if ((flags & TJFLAG_NOREALLOC) == 0 && doTile) {
+    printf("Disabling tiled compression/decompression tests, because those tests do not\n");
+    printf("work when dynamic JPEG buffer allocation is enabled.\n\n");
+    doTile = 0;
+  }
+
+  if (!decompOnly) {
+    if ((srcBuf = tjLoadImage(argv1, &w, 1, &h, &pf, flags)) == NULL)
+      THROW_TJG("loading bitmap");
+    temp = strrchr(argv1, '.');
+    if (temp != NULL) *temp = '\0';
+  }
+
+  if (quiet == 1 && !decompOnly) {
+    printf("All performance values in Mpixels/sec\n\n");
+    printf("Bitmap     JPEG     JPEG  %s  %s   ",
+           doTile ? "Tile " : "Image", doTile ? "Tile " : "Image");
+    if (doYUV) printf("Encode  ");
+    printf("Comp    Comp    Decomp  ");
+    if (doYUV) printf("Decode");
+    printf("\n");
+    printf("Format     Subsamp  Qual  Width  Height  ");
+    if (doYUV) printf("Perf    ");
+    printf("Perf    Ratio   Perf    ");
+    if (doYUV) printf("Perf");
+    printf("\n\n");
+  }
+
+  if (decompOnly) {
+    decompTest(argv1);
+    printf("\n");
+    goto bailout;
+  }
+  if (subsamp >= 0 && subsamp < TJ_NUMSAMP) {
+    for (i = maxQual; i >= minQual; i--)
+      fullTest(srcBuf, w, h, subsamp, i, argv1);
+    printf("\n");
+  } else {
+    if (pf != TJPF_CMYK) {
+      for (i = maxQual; i >= minQual; i--)
+        fullTest(srcBuf, w, h, TJSAMP_GRAY, i, argv1);
+      printf("\n");
+    }
+    for (i = maxQual; i >= minQual; i--)
+      fullTest(srcBuf, w, h, TJSAMP_420, i, argv1);
+    printf("\n");
+    for (i = maxQual; i >= minQual; i--)
+      fullTest(srcBuf, w, h, TJSAMP_422, i, argv1);
+    printf("\n");
+    for (i = maxQual; i >= minQual; i--)
+      fullTest(srcBuf, w, h, TJSAMP_444, i, argv1);
+    printf("\n");
+  }
+
+bailout:
+  tjFree(srcBuf);
+  return retval;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/tjbenchtest.in Added

@@ -0,0 +1,256 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if  -d $OUTDIR ; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	$*
+}
+
+EXT=bmp
+IMAGES="vgl_5674_0098.${EXT} vgl_6434_0018a.${EXT} vgl_6548_0026a.${EXT} nightshot_iso_100.${EXT}"
+IMGDIR=@CMAKE_CURRENT_SOURCE_DIR@/testimages
+OUTDIR=`mktemp -d /tmp/__tjbenchtest_output.XXXXXX`
+EXEDIR=@CMAKE_CURRENT_BINARY_DIR@
+BMPARG=
+NSARG=
+YUVARG=
+ALLOC=0
+ALLOCARG=
+PROGARG=
+if  "$EXT" = "bmp" ; then BMPARG=-bmp; fi
+
+if  -d $OUTDIR ; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+while  $# -gt 0 ; do
+	case "$1" in
+	-yuv)
+		NSARG=-nosmooth
+		YUVARG=-yuv
+
+# NOTE: The combination of tjEncodeYUV*() and tjCompressFromYUV*() does not
+# always produce bitwise-identical results to tjCompress*() if subsampling is
+# enabled.  In both cases, if the image width or height are not evenly
+# divisible by the MCU width/height, then the bottom and/or right edge are
+# expanded.  However, the libjpeg code performs this expansion prior to
+# downsampling, and TurboJPEG performs it in tjCompressFromYUV*(), which is
+# after downsampling.  Thus, the two will agree only if the width/height along
+# each downsampled dimension is an odd number or is evenly divisible by the MCU
+# width/height.  This disagreement basically amounts to a round-off error, but
+# there is no easy way around it, so for now, we just test the only image that
+# works.  (NOTE: nightshot_iso_100 does not suffer from the above issue, but
+# it suffers from an unrelated problem whereby the combination of
+# tjDecompressToYUV*() and tjDecodeYUV*() do not produce bitwise-identical
+# results to tjDecompress*() if decompression scaling is enabled.  This latter
+# phenomenon is not yet fully understood but is also believed to be some sort
+# of round-off error.)
+		IMAGES="vgl_6548_0026a.${EXT}"
+		;;
+	-alloc)
+		ALLOCARG=-alloc
+		ALLOC=1
+		;;
+	-progressive)
+		PROGARG=-progressive
+		;;
+	esac
+	shift
+done
+
+exec >$EXEDIR/tjbenchtest$YUVARG$ALLOCARG$PROGARG.log
+
+# Standard tests
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .${EXT}`
+	runme $EXEDIR/cjpeg -quality 95 -dct fast $PROGARG -grayscale -outfile $OUTDIR/${basename}_GRAY_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct fast $PROGARG -sample 2x2 -outfile $OUTDIR/${basename}_420_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct fast $PROGARG -sample 2x1 -outfile $OUTDIR/${basename}_422_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct fast $PROGARG -sample 1x1 -outfile $OUTDIR/${basename}_444_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int $PROGARG -grayscale -outfile $OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int $PROGARG -sample 2x2 -outfile $OUTDIR/${basename}_420_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int $PROGARG -sample 2x1 -outfile $OUTDIR/${basename}_422_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int $PROGARG -sample 1x1 -outfile $OUTDIR/${basename}_444_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/djpeg -rgb $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_default_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -rgb $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_fast_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -rgb $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_accurate_djpeg.${EXT} $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+	for samp in 420 422; do
+		runme $EXEDIR/djpeg -nosmooth $BMPARG -outfile $OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -nosmooth $BMPARG -outfile $OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -nosmooth $BMPARG -outfile $OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.${EXT} $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+
+	# Compression
+	for dct in accurate fast; do
+		runme $EXEDIR/tjbench $OUTDIR/$image 95 -rgb -quiet -benchtime 0.01 -warmup 0 -${dct}dct $YUVARG $ALLOCARG $PROGARG
+		for samp in GRAY 420 422 444; do
+			runme cmp $OUTDIR/${basename}_${samp}_Q95.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	for dct in fast accurate default; do
+		dctarg=-${dct}dct
+		if  "${dct}" = "default" ; then
+			dctarg=
+		fi
+
+		# Tiled compression & decompression
+		runme $EXEDIR/tjbench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG $ALLOCARG $PROGARG
+		for samp in GRAY 444; do
+			if  $ALLOC = 1 ; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+		runme $EXEDIR/tjbench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG $ALLOCARG $PROGARG
+		for samp in 420 422; do
+			if  $ALLOC = 1 ; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+
+		# Tiled decompression
+		for samp in GRAY 444; do
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG $ALLOCARG $PROGARG
+			if  $ALLOC = 1 ; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG $ALLOCARG $PROGARG
+			if  $ALLOC = 1 ; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed 's/\_/\//g'`
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_${scale}_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG $ALLOCARG $PROGARG
+			runme cmp $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT} $OUTDIR/${basename}_${samp}_${scale}_djpeg.${EXT}
+			rm $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT}
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/jpegtran -flip horizontal -trim -outfile $OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -flip vertical -trim -outfile $OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transpose -trim -outfile $OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transverse -trim -outfile $OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 90 -trim -outfile $OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 180 -trim -outfile $OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 270 -trim -outfile $OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444; do
+			runme $EXEDIR/djpeg -rgb $BMPARG -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -tile -quiet -benchtime 0.01 -warmup 0 $YUVARG $ALLOCARG $PROGARG
+			if  $ALLOC = 1 ; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+					rm $i
+				done
+			fi
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/djpeg -nosmooth -rgb $BMPARG -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample $YUVARG $ALLOCARG $PROGARG
+			if  $ALLOC = 1 ; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+					rm $i
+				done
+			fi
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -tile -quiet -benchtime 0.01 -warmup 0 -grayscale $YUVARG $ALLOCARG $PROGARG
+			if  $ALLOC = 1 ; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_GRAY_${xform}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_GRAY_${xform}_jpegtran.${EXT}
+					rm $i
+				done
+			fi
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed 's/\_/\//g'`
+				runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+				runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG $ALLOCARG $PROGARG
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT} $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT}
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/tjbenchtest.java.in Added

@@ -0,0 +1,215 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if  -d $OUTDIR ; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	"$@"
+}
+
+IMAGES="vgl_5674_0098.bmp vgl_6434_0018a.bmp vgl_6548_0026a.bmp nightshot_iso_100.bmp"
+IMGDIR=@CMAKE_CURRENT_SOURCE_DIR@/testimages
+OUTDIR=`mktemp -d /tmp/__tjbenchtest_java_output.XXXXXX`
+EXEDIR=@CMAKE_CURRENT_BINARY_DIR@
+JAVA="@Java_JAVA_EXECUTABLE@"
+JAVAARGS="-cp $EXEDIR/java/turbojpeg.jar -Djava.library.path=$EXEDIR"
+BMPARG=
+NSARG=
+YUVARG=
+PROGARG=
+
+if  -d $OUTDIR ; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+while  $# -gt 0 ; do
+	case "$1" in
+	-yuv)
+		NSARG=-nosmooth
+		YUVARG=-yuv
+
+# NOTE: The combination of tjEncodeYUV*() and tjCompressFromYUV*() does not
+# always produce bitwise-identical results to tjCompress*() if subsampling is
+# enabled.  In both cases, if the image width or height are not evenly
+# divisible by the MCU width/height, then the bottom and/or right edge are
+# expanded.  However, the libjpeg code performs this expansion prior to
+# downsampling, and TurboJPEG performs it in tjCompressFromYUV*(), which is
+# after downsampling.  Thus, the two will agree only if the width/height along
+# each downsampled dimension is an odd number or is evenly divisible by the MCU
+# width/height.  This disagreement basically amounts to a round-off error, but
+# there is no easy way around it, so for now, we just test the only image that
+# works.  (NOTE: nightshot_iso_100 does not suffer from the above issue, but
+# it suffers from an unrelated problem whereby the combination of
+# tjDecompressToYUV*() and tjDecodeYUV*() do not produce bitwise-identical
+# results to tjDecompress*() if decompression scaling is enabled.  This latter
+# phenomenon is not yet fully understood but is also believed to be some sort
+# of round-off error.)
+		IMAGES="vgl_6548_0026a.bmp"
+		;;
+	-progressive)
+		PROGARG=-progressive
+		;;
+	esac
+	shift
+done
+
+exec >$EXEDIR/tjbenchtest-java$YUVARG$PROGARG.log
+
+# Standard tests
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .bmp`
+	runme $EXEDIR/cjpeg -quality 95 -dct fast $PROGARG -grayscale -outfile $OUTDIR/${basename}_GRAY_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast $PROGARG -sample 2x2 -outfile $OUTDIR/${basename}_420_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast $PROGARG -sample 2x1 -outfile $OUTDIR/${basename}_422_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast $PROGARG -sample 1x1 -outfile $OUTDIR/${basename}_444_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int $PROGARG -grayscale -outfile $OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int $PROGARG -sample 2x2 -outfile $OUTDIR/${basename}_420_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int $PROGARG -sample 2x1 -outfile $OUTDIR/${basename}_422_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int $PROGARG -sample 1x1 -outfile $OUTDIR/${basename}_444_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_default_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_fast_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+	for samp in 420 422; do
+		runme $EXEDIR/djpeg -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+
+	# Compression
+	for dct in accurate fast; do
+		runme "$JAVA" $JAVAARGS TJBench $OUTDIR/$image 95 -rgb -quiet -benchtime 0.01 -warmup 0 -${dct}dct $YUVARG $PROGARG
+		for samp in GRAY 420 422 444; do
+			runme cmp $OUTDIR/${basename}_${samp}_Q95.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	for dct in fast accurate default; do
+		dctarg=-${dct}dct
+		if  "${dct}" = "default" ; then
+			dctarg=
+		fi
+
+		# Tiled compression & decompression
+		runme "$JAVA" $JAVAARGS TJBench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG $PROGARG
+		for samp in GRAY 444; do
+			for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+				rm $i
+			done
+		done
+		runme "$JAVA" $JAVAARGS TJBench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG $PROGARG
+		for samp in 420 422; do
+			for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+				rm $i
+			done
+		done
+
+		# Tiled decompression
+		for samp in GRAY 444; do
+			runme "$JAVA" $JAVAARGS TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG $PROGARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+				rm $i
+			done
+		done
+		for samp in 420 422; do
+			runme "$JAVA" $JAVAARGS TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG $PROGARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp $i -i 54:54 $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+				rm $i
+			done
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed 's/\_/\//g'`
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG -bmp -outfile $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+			runme "$JAVA" $JAVAARGS TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG $PROGARG
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/jpegtran -flip horizontal -trim -outfile $OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -flip vertical -trim -outfile $OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transpose -trim -outfile $OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transverse -trim -outfile $OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 90 -trim -outfile $OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 180 -trim -outfile $OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 270 -trim -outfile $OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444; do
+			runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme "$JAVA" $JAVAARGS TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -tile -quiet -benchtime 0.01 -warmup 0 $YUVARG $PROGARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+				rm $i
+			done
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/djpeg -nosmooth -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme "$JAVA" $JAVAARGS TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample $YUVARG $PROGARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+				rm $i
+			done
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme "$JAVA" $JAVAARGS TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -tile -quiet -benchtime 0.01 -warmup 0 -grayscale $YUVARG $PROGARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_0-9*0-9x0-9*0-9.bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_GRAY_${xform}_jpegtran.bmp
+				rm $i
+			done
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed 's/\_/\//g'`
+				runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+				runme "$JAVA" $JAVAARGS TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG $PROGARG
+				runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp
+				rm $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/tjexample.c Added

@@ -0,0 +1,398 @@
+/*
+ * Copyright (C)2011-2012, 2014-2015, 2017, 2019, 2021 D. R. Commander.
+ *                                                     All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * This program demonstrates how to compress, decompress, and transform JPEG
+ * images using the TurboJPEG C API
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <turbojpeg.h>
+
+
+#ifdef _WIN32
+#define strcasecmp  stricmp
+#define strncasecmp  strnicmp
+#endif
+
+#define THROW(action, message) { \
+  printf("ERROR in line %d while %s:\n%s\n", __LINE__, action, message); \
+  retval = -1;  goto bailout; \
+}
+
+#define THROW_TJ(action)  THROW(action, tjGetErrorStr2(tjInstance))
+
+#define THROW_UNIX(action)  THROW(action, strerror(errno))
+
+#define DEFAULT_SUBSAMP  TJSAMP_444
+#define DEFAULT_QUALITY  95
+
+
+const char *subsampNameTJ_NUMSAMP = {
+  "4:4:4", "4:2:2", "4:2:0", "Grayscale", "4:4:0", "4:1:1"
+};
+
+const char *colorspaceNameTJ_NUMCS = {
+  "RGB", "YCbCr", "GRAY", "CMYK", "YCCK"
+};
+
+tjscalingfactor *scalingFactors = NULL;
+int numScalingFactors = 0;
+
+
+/* DCT filter example.  This produces a negative of the image. */
+
+static int customFilter(short *coeffs, tjregion arrayRegion,
+                        tjregion planeRegion, int componentIndex,
+                        int transformIndex, tjtransform *transform)
+{
+  int i;
+
+  for (i = 0; i < arrayRegion.w * arrayRegion.h; i++)
+    coeffsi = -coeffsi;
+
+  return 0;
+}
+
+
+static void usage(char *programName)
+{
+  int i;
+
+  printf("\nUSAGE: %s <Input image> <Output image> options\n\n",
+         programName);
+
+  printf("Input and output images can be in Windows BMP or PBMPLUS (PPM/PGM) format.  If\n");
+  printf("either filename ends in a .jpg extension, then the TurboJPEG API will be used\n");
+  printf("to compress or decompress the image.\n\n");
+
+  printf("Compression Options (used if the output image is a JPEG image)\n");
+  printf("--------------------------------------------------------------\n\n");
+
+  printf("-subsamp <444|422|420|gray> = Apply this level of chrominance subsampling when\n");
+  printf("     compressing the output image.  The default is to use the same level of\n");
+  printf("     subsampling as in the input image, if the input image is also a JPEG\n");
+  printf("     image, or to use grayscale if the input image is a grayscale non-JPEG\n");
+  printf("     image, or to use %s subsampling otherwise.\n\n",
+         subsampNameDEFAULT_SUBSAMP);
+
+  printf("-q <1-100> = Compress the output image with this JPEG quality level\n");
+  printf("     (default = %d).\n\n", DEFAULT_QUALITY);
+
+  printf("Decompression Options (used if the input image is a JPEG image)\n");
+  printf("---------------------------------------------------------------\n\n");
+
+  printf("-scale M/N = Scale the input image by a factor of M/N when decompressing it.\n");
+  printf("(M/N = ");
+  for (i = 0; i < numScalingFactors; i++) {
+    printf("%d/%d", scalingFactorsi.num, scalingFactorsi.denom);
+    if (numScalingFactors == 2 && i != numScalingFactors - 1)
+      printf(" or ");
+    else if (numScalingFactors > 2) {
+      if (i != numScalingFactors - 1)
+        printf(", ");
+      if (i == numScalingFactors - 2)
+        printf("or ");
+    }
+  }
+  printf(")\n\n");
+
+  printf("-hflip, -vflip, -transpose, -transverse, -rot90, -rot180, -rot270 =\n");
+  printf("     Perform one of these lossless transform operations on the input image\n");
+  printf("     prior to decompressing it (these options are mutually exclusive.)\n\n");
+
+  printf("-grayscale = Perform lossless grayscale conversion on the input image prior\n");
+  printf("     to decompressing it (can be combined with the other transform operations\n");
+  printf("     above.)\n\n");
+
+  printf("-crop WxH+X+Y = Perform lossless cropping on the input image prior to\n");
+  printf("     decompressing it.  X and Y specify the upper left corner of the cropping\n");
+  printf("     region, and W and H specify the width and height of the cropping region.\n");
+  printf("     X and Y must be evenly divible by the MCU block size (8x8 if the input\n");
+  printf("     image was compressed using no subsampling or grayscale, 16x8 if it was\n");
+  printf("     compressed using 4:2:2 subsampling, or 16x16 if it was compressed using\n");
+  printf("     4:2:0 subsampling.)\n\n");
+
+  printf("General Options\n");
+  printf("---------------\n\n");
+
+  printf("-fastupsample = Use the fastest chrominance upsampling algorithm available in\n");
+  printf("     the underlying codec.\n\n");
+
+  printf("-fastdct = Use the fastest DCT/IDCT algorithms available in the underlying\n");
+  printf("     codec.\n\n");
+
+  printf("-accuratedct = Use the most accurate DCT/IDCT algorithms available in the\n");
+  printf("     underlying codec.\n\n");
+
+  exit(1);
+}
+
+
+int main(int argc, char **argv)
+{
+  tjscalingfactor scalingFactor = { 1, 1 };
+  int outSubsamp = -1, outQual = -1;
+  tjtransform xform;
+  int flags = 0;
+  int width, height;
+  char *inFormat, *outFormat;
+  FILE *jpegFile = NULL;
+  unsigned char *imgBuf = NULL, *jpegBuf = NULL;
+  int retval = 0, i, pixelFormat = TJPF_UNKNOWN;
+  tjhandle tjInstance = NULL;
+
+  if ((scalingFactors = tjGetScalingFactors(&numScalingFactors)) == NULL)
+    THROW_TJ("getting scaling factors");
+  memset(&xform, 0, sizeof(tjtransform));
+
+  if (argc < 3)
+    usage(argv0);
+
+  /* Parse arguments. */
+  for (i = 3; i < argc; i++) {
+    if (!strncasecmp(argvi, "-sc", 3) && i < argc - 1) {
+      int match = 0, temp1 = 0, temp2 = 0, j;
+
+      if (sscanf(argv++i, "%d/%d", &temp1, &temp2) < 2)
+        usage(argv0);
+      for (j = 0; j < numScalingFactors; j++) {
+        if ((double)temp1 / (double)temp2 == (double)scalingFactorsj.num /
+                                             (double)scalingFactorsj.denom) {
+          scalingFactor = scalingFactorsj;
+          match = 1;
+          break;
+        }
+      }
+      if (match != 1)
+        usage(argv0);
+    } else if (!strncasecmp(argvi, "-su", 3) && i < argc - 1) {
+      i++;
+      if (!strncasecmp(argvi, "g", 1))
+        outSubsamp = TJSAMP_GRAY;
+      else if (!strcasecmp(argvi, "444"))
+        outSubsamp = TJSAMP_444;
+      else if (!strcasecmp(argvi, "422"))
+        outSubsamp = TJSAMP_422;
+      else if (!strcasecmp(argvi, "420"))
+        outSubsamp = TJSAMP_420;
+      else
+        usage(argv0);
+    } else if (!strncasecmp(argvi, "-q", 2) && i < argc - 1) {
+      outQual = atoi(argv++i);
+      if (outQual < 1 || outQual > 100)
+        usage(argv0);
+    } else if (!strncasecmp(argvi, "-g", 2))
+      xform.options |= TJXOPT_GRAY;
+    else if (!strcasecmp(argvi, "-hflip"))
+      xform.op = TJXOP_HFLIP;
+    else if (!strcasecmp(argvi, "-vflip"))
+      xform.op = TJXOP_VFLIP;
+    else if (!strcasecmp(argvi, "-transpose"))
+      xform.op = TJXOP_TRANSPOSE;
+    else if (!strcasecmp(argvi, "-transverse"))
+      xform.op = TJXOP_TRANSVERSE;
+    else if (!strcasecmp(argvi, "-rot90"))
+      xform.op = TJXOP_ROT90;
+    else if (!strcasecmp(argvi, "-rot180"))
+      xform.op = TJXOP_ROT180;
+    else if (!strcasecmp(argvi, "-rot270"))
+      xform.op = TJXOP_ROT270;
+    else if (!strcasecmp(argvi, "-custom"))
+      xform.customFilter = customFilter;
+    else if (!strncasecmp(argvi, "-c", 2) && i < argc - 1) {
+      if (sscanf(argv++i, "%dx%d+%d+%d", &xform.r.w, &xform.r.h, &xform.r.x,
+                 &xform.r.y) < 4 ||
+          xform.r.x < 0 || xform.r.y < 0 || xform.r.w < 1 || xform.r.h < 1)
+        usage(argv0);
+      xform.options |= TJXOPT_CROP;
+    } else if (!strcasecmp(argvi, "-fastupsample")) {
+      printf("Using fast upsampling code\n");
+      flags |= TJFLAG_FASTUPSAMPLE;
+    } else if (!strcasecmp(argvi, "-fastdct")) {
+      printf("Using fastest DCT/IDCT algorithm\n");
+      flags |= TJFLAG_FASTDCT;
+    } else if (!strcasecmp(argvi, "-accuratedct")) {
+      printf("Using most accurate DCT/IDCT algorithm\n");
+      flags |= TJFLAG_ACCURATEDCT;
+    } else usage(argv0);
+  }
+
+  /* Determine input and output image formats based on file extensions. */
+  inFormat = strrchr(argv1, '.');
+  outFormat = strrchr(argv2, '.');
+  if (inFormat == NULL || outFormat == NULL || strlen(inFormat) < 2 ||
+      strlen(outFormat) < 2)
+    usage(argv0);
+  inFormat = &inFormat1;
+  outFormat = &outFormat1;
+
+  if (!strcasecmp(inFormat, "jpg")) {
+    /* Input image is a JPEG image.  Decompress and/or transform it. */
+    long size;
+    int inSubsamp, inColorspace;
+    int doTransform = (xform.op != TJXOP_NONE || xform.options != 0 ||
+                       xform.customFilter != NULL);
+    unsigned long jpegSize;
+
+    /* Read the JPEG file into memory. */
+    if ((jpegFile = fopen(argv1, "rb")) == NULL)
+      THROW_UNIX("opening input file");
+    if (fseek(jpegFile, 0, SEEK_END) < 0 || ((size = ftell(jpegFile)) < 0) ||
+        fseek(jpegFile, 0, SEEK_SET) < 0)
+      THROW_UNIX("determining input file size");
+    if (size == 0)
+      THROW("determining input file size", "Input file contains no data");
+    jpegSize = (unsigned long)size;
+    if ((jpegBuf = (unsigned char *)tjAlloc(jpegSize)) == NULL)
+      THROW_UNIX("allocating JPEG buffer");
+    if (fread(jpegBuf, jpegSize, 1, jpegFile) < 1)
+      THROW_UNIX("reading input file");
+    fclose(jpegFile);  jpegFile = NULL;
+
+    if (doTransform) {
+      /* Transform it. */
+      unsigned char *dstBuf = NULL;  /* Dynamically allocate the JPEG buffer */
+      unsigned long dstSize = 0;
+
+      if ((tjInstance = tjInitTransform()) == NULL)
+        THROW_TJ("initializing transformer");
+      xform.options |= TJXOPT_TRIM;
+      if (tjTransform(tjInstance, jpegBuf, jpegSize, 1, &dstBuf, &dstSize,
+                      &xform, flags) < 0) {
+        tjFree(dstBuf);
+        THROW_TJ("transforming input image");
+      }
+      tjFree(jpegBuf);
+      jpegBuf = dstBuf;
+      jpegSize = dstSize;
+    } else {
+      if ((tjInstance = tjInitDecompress()) == NULL)
+        THROW_TJ("initializing decompressor");
+    }
+
+    if (tjDecompressHeader3(tjInstance, jpegBuf, jpegSize, &width, &height,
+                            &inSubsamp, &inColorspace) < 0)
+      THROW_TJ("reading JPEG header");
+
+    printf("%s Image:  %d x %d pixels, %s subsampling, %s colorspace\n",
+           (doTransform ? "Transformed" : "Input"), width, height,
+           subsampNameinSubsamp, colorspaceNameinColorspace);
+
+    if (!strcasecmp(outFormat, "jpg") && doTransform &&
+        scalingFactor.num == 1 && scalingFactor.denom == 1 && outSubsamp < 0 &&
+        outQual < 0) {
+      /* Input image has been transformed, and no re-compression options
+         have been selected.  Write the transformed image to disk and exit. */
+      if ((jpegFile = fopen(argv2, "wb")) == NULL)
+        THROW_UNIX("opening output file");
+      if (fwrite(jpegBuf, jpegSize, 1, jpegFile) < 1)
+        THROW_UNIX("writing output file");
+      fclose(jpegFile);  jpegFile = NULL;
+      goto bailout;
+    }
+
+    /* Scaling and/or a non-JPEG output image format and/or compression options
+       have been selected, so we need to decompress the input/transformed
+       image. */
+    width = TJSCALED(width, scalingFactor);
+    height = TJSCALED(height, scalingFactor);
+    if (outSubsamp < 0)
+      outSubsamp = inSubsamp;
+
+    pixelFormat = TJPF_BGRX;
+    if ((imgBuf = (unsigned char *)tjAlloc(width * height *
+                                           tjPixelSizepixelFormat)) == NULL)
+      THROW_UNIX("allocating uncompressed image buffer");
+
+    if (tjDecompress2(tjInstance, jpegBuf, jpegSize, imgBuf, width, 0, height,
+                      pixelFormat, flags) < 0)
+      THROW_TJ("decompressing JPEG image");
+    tjFree(jpegBuf);  jpegBuf = NULL;
+    tjDestroy(tjInstance);  tjInstance = NULL;
+  } else {
+    /* Input image is not a JPEG image.  Load it into memory. */
+    if ((imgBuf = tjLoadImage(argv1, &width, 1, &height, &pixelFormat,
+                              0)) == NULL)
+      THROW_TJ("loading input image");
+    if (outSubsamp < 0) {
+      if (pixelFormat == TJPF_GRAY)
+        outSubsamp = TJSAMP_GRAY;
+      else
+        outSubsamp = TJSAMP_444;
+    }
+    printf("Input Image:  %d x %d pixels\n", width, height);
+  }
+
+  printf("Output Image (%s):  %d x %d pixels", outFormat, width, height);
+
+  if (!strcasecmp(outFormat, "jpg")) {
+    /* Output image format is JPEG.  Compress the uncompressed image. */
+    unsigned long jpegSize = 0;
+
+    jpegBuf = NULL;  /* Dynamically allocate the JPEG buffer */
+
+    if (outQual < 0)
+      outQual = DEFAULT_QUALITY;
+    printf(", %s subsampling, quality = %d\n", subsampNameoutSubsamp,
+           outQual);
+
+    if ((tjInstance = tjInitCompress()) == NULL)
+      THROW_TJ("initializing compressor");
+    if (tjCompress2(tjInstance, imgBuf, width, 0, height, pixelFormat,
+                    &jpegBuf, &jpegSize, outSubsamp, outQual, flags) < 0)
+      THROW_TJ("compressing image");
+    tjDestroy(tjInstance);  tjInstance = NULL;
+
+    /* Write the JPEG image to disk. */
+    if ((jpegFile = fopen(argv2, "wb")) == NULL)
+      THROW_UNIX("opening output file");
+    if (fwrite(jpegBuf, jpegSize, 1, jpegFile) < 1)
+      THROW_UNIX("writing output file");
+    tjDestroy(tjInstance);  tjInstance = NULL;
+    fclose(jpegFile);  jpegFile = NULL;
+    tjFree(jpegBuf);  jpegBuf = NULL;
+  } else {
+    /* Output image format is not JPEG.  Save the uncompressed image
+       directly to disk. */
+    printf("\n");
+    if (tjSaveImage(argv2, imgBuf, width, 0, height, pixelFormat, 0) < 0)
+      THROW_TJ("saving output image");
+  }
+
+bailout:
+  tjFree(imgBuf);
+  if (tjInstance) tjDestroy(tjInstance);
+  tjFree(jpegBuf);
+  if (jpegFile) fclose(jpegFile);
+  return retval;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/tjexampletest.in Added

@@ -0,0 +1,149 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if  -d $OUTDIR ; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	$*
+}
+
+IMAGES="vgl_5674_0098.bmp vgl_6434_0018a.bmp vgl_6548_0026a.bmp nightshot_iso_100.bmp"
+IMGDIR=@CMAKE_CURRENT_SOURCE_DIR@/testimages
+OUTDIR=`mktemp -d /tmp/__tjexampletest_output.XXXXXX`
+EXEDIR=@CMAKE_CURRENT_BINARY_DIR@
+
+if  -d $OUTDIR ; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+exec >$EXEDIR/tjexampletest.log
+
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .bmp`
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -grayscale -outfile $OUTDIR/${basename}_GRAY_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x2 -outfile $OUTDIR/${basename}_420_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x1 -outfile $OUTDIR/${basename}_422_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 1x1 -outfile $OUTDIR/${basename}_444_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -grayscale -outfile $OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x2 -outfile $OUTDIR/${basename}_420_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x1 -outfile $OUTDIR/${basename}_422_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 1x1 -outfile $OUTDIR/${basename}_444_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_default_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_fast_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+	for samp in 420 422; do
+		runme $EXEDIR/djpeg -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+
+	# Compression
+	for dct in fast accurate; do
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/tjexample $OUTDIR/$image $OUTDIR/${basename}_${samp}_${dct}.jpg -q 95 -subsamp ${samp} -${dct}dct
+			runme cmp $OUTDIR/${basename}_${samp}_${dct}.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	# Decompression
+	for dct in fast accurate default; do
+		srcdct=${dct}
+		dctarg=-${dct}dct
+		if  "${dct}" = "default" ; then
+			srcdct=fast
+			dctarg=
+		fi
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/tjexample $OUTDIR/${basename}_${samp}_${srcdct}.jpg $OUTDIR/${basename}_${samp}_${dct}.bmp ${dctarg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${dct}.bmp $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${dct}.bmp
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/tjexample $OUTDIR/${basename}_${samp}_${srcdct}.jpg $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp -fastupsample ${dctarg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed 's/\_/\//g'`
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/djpeg -rgb -bmp -scale ${scalearg} -outfile $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+			runme $EXEDIR/tjexample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${scale}.bmp -scale ${scalearg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${scale}.bmp $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${scale}.bmp
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -flip horizontal -trim -outfile $OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -flip vertical -trim -outfile $OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -transpose -trim -outfile $OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -transverse -trim -outfile $OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -rotate 90 -trim -outfile $OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -rotate 180 -trim -outfile $OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -rotate 270 -trim -outfile $OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/tjexample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.jpg -$xform -crop 70x60+16+16
+			runme cmp $OUTDIR/${basename}_${samp}_${xform}.jpg $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/tjexample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -crop 70x60+16+16
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/djpeg -nosmooth -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/tjexample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -crop 70x60+16+16 -fastupsample
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme $EXEDIR/tjexample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.jpg -$xform -grayscale -crop 70x60+16+16
+			runme cmp $OUTDIR/${basename}_${samp}_${xform}.jpg $OUTDIR/${basename}_GRAY_${xform}_jpegtran.jpg
+			runme $EXEDIR/tjexample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -grayscale -crop 70x60+16+16
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_GRAY_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed 's/\_/\//g'`
+				runme $EXEDIR/djpeg -rgb -bmp -scale ${scalearg} -outfile $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+				runme $EXEDIR/tjexample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp -$xform -scale ${scalearg} -crop 70x60+16+16
+				runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp
+				rm $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/tjexampletest.java.in Added

@@ -0,0 +1,151 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if  -d $OUTDIR ; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	"$@"
+}
+
+IMAGES="vgl_5674_0098.bmp vgl_6434_0018a.bmp vgl_6548_0026a.bmp nightshot_iso_100.bmp"
+IMGDIR=@CMAKE_CURRENT_SOURCE_DIR@/testimages
+OUTDIR=`mktemp -d /tmp/__tjexampletest_java_output.XXXXXX`
+EXEDIR=@CMAKE_CURRENT_BINARY_DIR@
+JAVA="@Java_JAVA_EXECUTABLE@"
+JAVAARGS="-cp $EXEDIR/java/turbojpeg.jar -Djava.library.path=$EXEDIR"
+
+if  -d $OUTDIR ; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+exec >$EXEDIR/tjexampletest-java.log
+
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .bmp`
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -grayscale -outfile $OUTDIR/${basename}_GRAY_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x2 -outfile $OUTDIR/${basename}_420_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x1 -outfile $OUTDIR/${basename}_422_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 1x1 -outfile $OUTDIR/${basename}_444_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -grayscale -outfile $OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x2 -outfile $OUTDIR/${basename}_420_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x1 -outfile $OUTDIR/${basename}_422_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 1x1 -outfile $OUTDIR/${basename}_444_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_default_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_fast_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+	for samp in 420 422; do
+		runme $EXEDIR/djpeg -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+
+	# Compression
+	for dct in fast accurate; do
+		for samp in GRAY 420 422 444; do
+			runme "$JAVA" $JAVAARGS TJExample $OUTDIR/$image $OUTDIR/${basename}_${samp}_${dct}.jpg -q 95 -subsamp ${samp} -${dct}dct
+			runme cmp $OUTDIR/${basename}_${samp}_${dct}.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	# Decompression
+	for dct in fast accurate default; do
+		srcdct=${dct}
+		dctarg=-${dct}dct
+		if  "${dct}" = "default" ; then
+			srcdct=fast
+			dctarg=
+		fi
+		for samp in GRAY 420 422 444; do
+			runme "$JAVA" $JAVAARGS TJExample $OUTDIR/${basename}_${samp}_${srcdct}.jpg $OUTDIR/${basename}_${samp}_${dct}.bmp ${dctarg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${dct}.bmp $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${dct}.bmp
+		done
+		for samp in 420 422; do
+			runme "$JAVA" $JAVAARGS TJExample $OUTDIR/${basename}_${samp}_${srcdct}.jpg $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp -fastupsample ${dctarg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed 's/\_/\//g'`
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/djpeg -rgb -bmp -scale ${scalearg} -outfile $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+			runme "$JAVA" $JAVAARGS TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${scale}.bmp -scale ${scalearg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${scale}.bmp $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${scale}.bmp
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -flip horizontal -trim -outfile $OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -flip vertical -trim -outfile $OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -transpose -trim -outfile $OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -transverse -trim -outfile $OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -rotate 90 -trim -outfile $OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -rotate 180 -trim -outfile $OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+		runme $EXEDIR/jpegtran -crop 70x60+16+16 -rotate 270 -trim -outfile $OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg $OUTDIR/${basename}_${samp}_fast.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 420 422 444; do
+			runme "$JAVA" $JAVAARGS TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.jpg -$xform -crop 70x60+16+16
+			runme cmp $OUTDIR/${basename}_${samp}_${xform}.jpg $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme "$JAVA" $JAVAARGS TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -crop 70x60+16+16
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/djpeg -nosmooth -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme "$JAVA" $JAVAARGS TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -crop 70x60+16+16 -fastupsample
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme "$JAVA" $JAVAARGS TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.jpg -$xform -grayscale -crop 70x60+16+16
+			runme cmp $OUTDIR/${basename}_${samp}_${xform}.jpg $OUTDIR/${basename}_GRAY_${xform}_jpegtran.jpg
+			runme "$JAVA" $JAVAARGS TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -grayscale -crop 70x60+16+16
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_GRAY_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed 's/\_/\//g'`
+				runme $EXEDIR/djpeg -rgb -bmp -scale ${scalearg} -outfile $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+				runme "$JAVA" $JAVAARGS TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp -$xform -scale ${scalearg} -crop 70x60+16+16
+				runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp
+				rm $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/tjunittest.c Added

@@ -0,0 +1,931 @@
+/*
+ * Copyright (C)2009-2014, 2017-2019 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * This program tests the various code paths in the TurboJPEG C Wrapper
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include "tjutil.h"
+#include "turbojpeg.h"
+#include "md5/md5.h"
+#include "cmyk.h"
+#ifdef _WIN32
+#include <time.h>
+#define random()  rand()
+#else
+#include <unistd.h>
+#endif
+
+
+static void usage(char *progName)
+{
+  printf("\nUSAGE: %s options\n\n", progName);
+  printf("Options:\n");
+  printf("-yuv = test YUV encoding/decoding support\n");
+  printf("-noyuvpad = do not pad each line of each Y, U, and V plane to the nearest\n");
+  printf("            4-byte boundary\n");
+  printf("-alloc = test automatic buffer allocation\n");
+  printf("-bmp = tjLoadImage()/tjSaveImage() unit test\n\n");
+  exit(1);
+}
+
+
+#define THROW_TJ() { \
+  printf("TurboJPEG ERROR:\n%s\n", tjGetErrorStr()); \
+  BAILOUT() \
+}
+#define TRY_TJ(f) { if ((f) == -1) THROW_TJ(); }
+#define THROW(m) { printf("ERROR: %s\n", m);  BAILOUT() }
+#define THROW_MD5(filename, md5sum, ref) { \
+  printf("\n%s has an MD5 sum of %s.\n   Should be %s.\n", filename, md5sum, \
+         ref); \
+  BAILOUT() \
+}
+
+const char *subNameLongTJ_NUMSAMP = {
+  "4:4:4", "4:2:2", "4:2:0", "GRAY", "4:4:0", "4:1:1"
+};
+const char *subNameTJ_NUMSAMP = {
+  "444", "422", "420", "GRAY", "440", "411"
+};
+
+const char *pixFormatStrTJ_NUMPF = {
+  "RGB", "BGR", "RGBX", "BGRX", "XBGR", "XRGB", "Grayscale",
+  "RGBA", "BGRA", "ABGR", "ARGB", "CMYK"
+};
+
+const int _3byteFormats = { TJPF_RGB, TJPF_BGR };
+const int _4byteFormats = {
+  TJPF_RGBX, TJPF_BGRX, TJPF_XBGR, TJPF_XRGB, TJPF_CMYK
+};
+const int _onlyGray = { TJPF_GRAY };
+const int _onlyRGB = { TJPF_RGB };
+
+int doYUV = 0, alloc = 0, pad = 4;
+
+int exitStatus = 0;
+#define BAILOUT() { exitStatus = -1;  goto bailout; }
+
+
+static void initBuf(unsigned char *buf, int w, int h, int pf, int flags)
+{
+  int roffset = tjRedOffsetpf;
+  int goffset = tjGreenOffsetpf;
+  int boffset = tjBlueOffsetpf;
+  int ps = tjPixelSizepf;
+  int index, row, col, halfway = 16;
+
+  if (pf == TJPF_GRAY) {
+    memset(buf, 0, w * h * ps);
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        if (flags & TJFLAG_BOTTOMUP) index = (h - row - 1) * w + col;
+        else index = row * w + col;
+        if (((row / 8) + (col / 8)) % 2 == 0)
+          bufindex = (row < halfway) ? 255 : 0;
+        else bufindex = (row < halfway) ? 76 : 226;
+      }
+    }
+  } else if (pf == TJPF_CMYK) {
+    memset(buf, 255, w * h * ps);
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        if (flags & TJFLAG_BOTTOMUP) index = (h - row - 1) * w + col;
+        else index = row * w + col;
+        if (((row / 8) + (col / 8)) % 2 == 0) {
+          if (row >= halfway) bufindex * ps + 3 = 0;
+        } else {
+          bufindex * ps + 2 = 0;
+          if (row < halfway) bufindex * ps + 1 = 0;
+        }
+      }
+    }
+  } else {
+    memset(buf, 0, w * h * ps);
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        if (flags & TJFLAG_BOTTOMUP) index = (h - row - 1) * w + col;
+        else index = row * w + col;
+        if (((row / 8) + (col / 8)) % 2 == 0) {
+          if (row < halfway) {
+            bufindex * ps + roffset = 255;
+            bufindex * ps + goffset = 255;
+            bufindex * ps + boffset = 255;
+          }
+        } else {
+          bufindex * ps + roffset = 255;
+          if (row >= halfway) bufindex * ps + goffset = 255;
+        }
+      }
+    }
+  }
+}
+
+
+#define CHECKVAL(v, cv) { \
+  if (v < cv - 1 || v > cv + 1) { \
+    printf("\nComp. %s at %d,%d should be %d, not %d\n", #v, row, col, cv, \
+           v); \
+    retval = 0;  exitStatus = -1;  goto bailout; \
+  } \
+}
+
+#define CHECKVAL0(v) { \
+  if (v > 1) { \
+    printf("\nComp. %s at %d,%d should be 0, not %d\n", #v, row, col, v); \
+    retval = 0;  exitStatus = -1;  goto bailout; \
+  } \
+}
+
+#define CHECKVAL255(v) { \
+  if (v < 254) { \
+    printf("\nComp. %s at %d,%d should be 255, not %d\n", #v, row, col, v); \
+    retval = 0;  exitStatus = -1;  goto bailout; \
+  } \
+}
+
+
+static int checkBuf(unsigned char *buf, int w, int h, int pf, int subsamp,
+                    tjscalingfactor sf, int flags)
+{
+  int roffset = tjRedOffsetpf;
+  int goffset = tjGreenOffsetpf;
+  int boffset = tjBlueOffsetpf;
+  int aoffset = tjAlphaOffsetpf;
+  int ps = tjPixelSizepf;
+  int index, row, col, retval = 1;
+  int halfway = 16 * sf.num / sf.denom;
+  int blocksize = 8 * sf.num / sf.denom;
+
+  if (pf == TJPF_GRAY) roffset = goffset = boffset = 0;
+
+  if (pf == TJPF_CMYK) {
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        unsigned char c, m, y, k;
+
+        if (flags & TJFLAG_BOTTOMUP) index = (h - row - 1) * w + col;
+        else index = row * w + col;
+        c = bufindex * ps;
+        m = bufindex * ps + 1;
+        y = bufindex * ps + 2;
+        k = bufindex * ps + 3;
+        if (((row / blocksize) + (col / blocksize)) % 2 == 0) {
+          CHECKVAL255(c);  CHECKVAL255(m);  CHECKVAL255(y);
+          if (row < halfway) CHECKVAL255(k)
+          else CHECKVAL0(k)
+        } else {
+          CHECKVAL255(c);  CHECKVAL0(y);  CHECKVAL255(k);
+          if (row < halfway) CHECKVAL0(m)
+          else CHECKVAL255(m)
+        }
+      }
+    }
+    return 1;
+  }
+
+  for (row = 0; row < h; row++) {
+    for (col = 0; col < w; col++) {
+      unsigned char r, g, b, a;
+
+      if (flags & TJFLAG_BOTTOMUP) index = (h - row - 1) * w + col;
+      else index = row * w + col;
+      r = bufindex * ps + roffset;
+      g = bufindex * ps + goffset;
+      b = bufindex * ps + boffset;
+      a = aoffset >= 0 ? bufindex * ps + aoffset : 0xFF;
+      if (((row / blocksize) + (col / blocksize)) % 2 == 0) {
+        if (row < halfway) {
+          CHECKVAL255(r);  CHECKVAL255(g);  CHECKVAL255(b);
+        } else {
+          CHECKVAL0(r);  CHECKVAL0(g);  CHECKVAL0(b);
+        }
+      } else {
+        if (subsamp == TJSAMP_GRAY) {
+          if (row < halfway) {
+            CHECKVAL(r, 76);  CHECKVAL(g, 76);  CHECKVAL(b, 76);
+          } else {
+            CHECKVAL(r, 226);  CHECKVAL(g, 226);  CHECKVAL(b, 226);
+          }
+        } else {
+          if (row < halfway) {
+            CHECKVAL255(r);  CHECKVAL0(g);  CHECKVAL0(b);
+          } else {
+            CHECKVAL255(r);  CHECKVAL255(g);  CHECKVAL0(b);
+          }
+        }
+      }
+      CHECKVAL255(a);
+    }
+  }
+
+bailout:
+  if (retval == 0) {
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        if (pf == TJPF_CMYK)
+          printf("%.3d/%.3d/%.3d/%.3d ", buf(row * w + col) * ps,
+                 buf(row * w + col) * ps + 1, buf(row * w + col) * ps + 2,
+                 buf(row * w + col) * ps + 3);
+        else
+          printf("%.3d/%.3d/%.3d ", buf(row * w + col) * ps + roffset,
+                 buf(row * w + col) * ps + goffset,
+                 buf(row * w + col) * ps + boffset);
+      }
+      printf("\n");
+    }
+  }
+  return retval;
+}
+
+
+#define PAD(v, p)  ((v + (p) - 1) & (~((p) - 1)))
+
+static int checkBufYUV(unsigned char *buf, int w, int h, int subsamp,
+                       tjscalingfactor sf)
+{
+  int row, col;
+  int hsf = tjMCUWidthsubsamp / 8, vsf = tjMCUHeightsubsamp / 8;
+  int pw = PAD(w, hsf), ph = PAD(h, vsf);
+  int cw = pw / hsf, ch = ph / vsf;
+  int ypitch = PAD(pw, pad), uvpitch = PAD(cw, pad);
+  int retval = 1;
+  int halfway = 16 * sf.num / sf.denom;
+  int blocksize = 8 * sf.num / sf.denom;
+
+  for (row = 0; row < ph; row++) {
+    for (col = 0; col < pw; col++) {
+      unsigned char y = bufypitch * row + col;
+
+      if (((row / blocksize) + (col / blocksize)) % 2 == 0) {
+        if (row < halfway) CHECKVAL255(y)
+        else CHECKVAL0(y);
+      } else {
+        if (row < halfway) CHECKVAL(y, 76)
+        else CHECKVAL(y, 226);
+      }
+    }
+  }
+  if (subsamp != TJSAMP_GRAY) {
+    halfway = 16 / vsf * sf.num / sf.denom;
+
+    for (row = 0; row < ch; row++) {
+      for (col = 0; col < cw; col++) {
+        unsigned char u = bufypitch * ph + (uvpitch * row + col),
+          v = bufypitch * ph + uvpitch * ch + (uvpitch * row + col);
+
+        if (((row * vsf / blocksize) + (col * hsf / blocksize)) % 2 == 0) {
+          CHECKVAL(u, 128);  CHECKVAL(v, 128);
+        } else {
+          if (row < halfway) {
+            CHECKVAL(u, 85);  CHECKVAL255(v);
+          } else {
+            CHECKVAL0(u);  CHECKVAL(v, 149);
+          }
+        }
+      }
+    }
+  }
+
+bailout:
+  if (retval == 0) {
+    for (row = 0; row < ph; row++) {
+      for (col = 0; col < pw; col++)
+        printf("%.3d ", bufypitch * row + col);
+      printf("\n");
+    }
+    printf("\n");
+    for (row = 0; row < ch; row++) {
+      for (col = 0; col < cw; col++)
+        printf("%.3d ", bufypitch * ph + (uvpitch * row + col));
+      printf("\n");
+    }
+    printf("\n");
+    for (row = 0; row < ch; row++) {
+      for (col = 0; col < cw; col++)
+        printf("%.3d ",
+               bufypitch * ph + uvpitch * ch + (uvpitch * row + col));
+      printf("\n");
+    }
+  }
+
+  return retval;
+}
+
+
+static void writeJPEG(unsigned char *jpegBuf, unsigned long jpegSize,
+                      char *filename)
+{
+  FILE *file = fopen(filename, "wb");
+
+  if (!file || fwrite(jpegBuf, jpegSize, 1, file) != 1) {
+    printf("ERROR: Could not write to %s.\n%s\n", filename, strerror(errno));
+    BAILOUT()
+  }
+
+bailout:
+  if (file) fclose(file);
+}
+
+
+static void compTest(tjhandle handle, unsigned char **dstBuf,
+                     unsigned long *dstSize, int w, int h, int pf,
+                     char *basename, int subsamp, int jpegQual, int flags)
+{
+  char tempStr1024;
+  unsigned char *srcBuf = NULL, *yuvBuf = NULL;
+  const char *pfStr = pixFormatStrpf;
+  const char *buStrLong =
+    (flags & TJFLAG_BOTTOMUP) ? "Bottom-Up" : "Top-Down ";
+  const char *buStr = (flags & TJFLAG_BOTTOMUP) ? "BU" : "TD";
+
+  if ((srcBuf = (unsigned char *)malloc(w * h * tjPixelSizepf)) == NULL)
+    THROW("Memory allocation failure");
+  initBuf(srcBuf, w, h, pf, flags);
+
+  if (*dstBuf && *dstSize > 0) memset(*dstBuf, 0, *dstSize);
+
+  if (!alloc) flags |= TJFLAG_NOREALLOC;
+  if (doYUV) {
+    unsigned long yuvSize = tjBufSizeYUV2(w, pad, h, subsamp);
+    tjscalingfactor sf = { 1, 1 };
+    tjhandle handle2 = tjInitCompress();
+
+    if (!handle2) THROW_TJ();
+
+    if ((yuvBuf = (unsigned char *)malloc(yuvSize)) == NULL)
+      THROW("Memory allocation failure");
+    memset(yuvBuf, 0, yuvSize);
+
+    printf("%s %s -> YUV %s ... ", pfStr, buStrLong, subNameLongsubsamp);
+    TRY_TJ(tjEncodeYUV3(handle2, srcBuf, w, 0, h, pf, yuvBuf, pad, subsamp,
+                        flags));
+    tjDestroy(handle2);
+    if (checkBufYUV(yuvBuf, w, h, subsamp, sf)) printf("Passed.\n");
+    else printf("FAILED!\n");
+
+    printf("YUV %s %s -> JPEG Q%d ... ", subNameLongsubsamp, buStrLong,
+           jpegQual);
+    TRY_TJ(tjCompressFromYUV(handle, yuvBuf, w, pad, h, subsamp, dstBuf,
+                             dstSize, jpegQual, flags));
+  } else {
+    printf("%s %s -> %s Q%d ... ", pfStr, buStrLong, subNameLongsubsamp,
+           jpegQual);
+    TRY_TJ(tjCompress2(handle, srcBuf, w, 0, h, pf, dstBuf, dstSize, subsamp,
+                       jpegQual, flags));
+  }
+
+  snprintf(tempStr, 1024, "%s_enc_%s_%s_%s_Q%d.jpg", basename, pfStr, buStr,
+           subNamesubsamp, jpegQual);
+  writeJPEG(*dstBuf, *dstSize, tempStr);
+  printf("Done.\n  Result in %s\n", tempStr);
+
+bailout:
+  free(yuvBuf);
+  free(srcBuf);
+}
+
+
+static void _decompTest(tjhandle handle, unsigned char *jpegBuf,
+                        unsigned long jpegSize, int w, int h, int pf,
+                        char *basename, int subsamp, int flags,
+                        tjscalingfactor sf)
+{
+  unsigned char *dstBuf = NULL, *yuvBuf = NULL;
+  int _hdrw = 0, _hdrh = 0, _hdrsubsamp = -1;
+  int scaledWidth = TJSCALED(w, sf);
+  int scaledHeight = TJSCALED(h, sf);
+  unsigned long dstSize = 0;
+
+  TRY_TJ(tjDecompressHeader2(handle, jpegBuf, jpegSize, &_hdrw, &_hdrh,
+                             &_hdrsubsamp));
+  if (_hdrw != w || _hdrh != h || _hdrsubsamp != subsamp)
+    THROW("Incorrect JPEG header");
+
+  dstSize = scaledWidth * scaledHeight * tjPixelSizepf;
+  if ((dstBuf = (unsigned char *)malloc(dstSize)) == NULL)
+    THROW("Memory allocation failure");
+  memset(dstBuf, 0, dstSize);
+
+  if (doYUV) {
+    unsigned long yuvSize = tjBufSizeYUV2(scaledWidth, pad, scaledHeight,
+                                          subsamp);
+    tjhandle handle2 = tjInitDecompress();
+
+    if (!handle2) THROW_TJ();
+
+    if ((yuvBuf = (unsigned char *)malloc(yuvSize)) == NULL)
+      THROW("Memory allocation failure");
+    memset(yuvBuf, 0, yuvSize);
+
+    printf("JPEG -> YUV %s ", subNameLongsubsamp);
+    if (sf.num != 1 || sf.denom != 1)
+      printf("%d/%d ... ", sf.num, sf.denom);
+    else printf("... ");
+    TRY_TJ(tjDecompressToYUV2(handle, jpegBuf, jpegSize, yuvBuf, scaledWidth,
+                              pad, scaledHeight, flags));
+    if (checkBufYUV(yuvBuf, scaledWidth, scaledHeight, subsamp, sf))
+      printf("Passed.\n");
+    else printf("FAILED!\n");
+
+    printf("YUV %s -> %s %s ... ", subNameLongsubsamp, pixFormatStrpf,
+           (flags & TJFLAG_BOTTOMUP) ? "Bottom-Up" : "Top-Down ");
+    TRY_TJ(tjDecodeYUV(handle2, yuvBuf, pad, subsamp, dstBuf, scaledWidth, 0,
+                       scaledHeight, pf, flags));
+    tjDestroy(handle2);
+  } else {
+    printf("JPEG -> %s %s ", pixFormatStrpf,
+           (flags & TJFLAG_BOTTOMUP) ? "Bottom-Up" : "Top-Down ");
+    if (sf.num != 1 || sf.denom != 1)
+      printf("%d/%d ... ", sf.num, sf.denom);
+    else printf("... ");
+    TRY_TJ(tjDecompress2(handle, jpegBuf, jpegSize, dstBuf, scaledWidth, 0,
+                         scaledHeight, pf, flags));
+  }
+
+  if (checkBuf(dstBuf, scaledWidth, scaledHeight, pf, subsamp, sf, flags))
+    printf("Passed.");
+  else printf("FAILED!");
+  printf("\n");
+
+bailout:
+  free(yuvBuf);
+  free(dstBuf);
+}
+
+
+static void decompTest(tjhandle handle, unsigned char *jpegBuf,
+                       unsigned long jpegSize, int w, int h, int pf,
+                       char *basename, int subsamp, int flags)
+{
+  int i, n = 0;
+  tjscalingfactor *sf = tjGetScalingFactors(&n);
+
+  if (!sf || !n) THROW_TJ();
+
+  for (i = 0; i < n; i++) {
+    if (subsamp == TJSAMP_444 || subsamp == TJSAMP_GRAY ||
+        (subsamp == TJSAMP_411 && sfi.num == 1 &&
+         (sfi.denom == 2 || sfi.denom == 1)) ||
+        (subsamp != TJSAMP_411 && sfi.num == 1 &&
+         (sfi.denom == 4 || sfi.denom == 2 || sfi.denom == 1)))
+      _decompTest(handle, jpegBuf, jpegSize, w, h, pf, basename, subsamp,
+                  flags, sfi);
+  }
+
+bailout:
+  return;
+}
+
+
+static void doTest(int w, int h, const int *formats, int nformats, int subsamp,
+                   char *basename)
+{
+  tjhandle chandle = NULL, dhandle = NULL;
+  unsigned char *dstBuf = NULL;
+  unsigned long size = 0;
+  int pfi, pf, i;
+
+  if (!alloc)
+    size = tjBufSize(w, h, subsamp);
+  if (size != 0)
+    if ((dstBuf = (unsigned char *)tjAlloc(size)) == NULL)
+      THROW("Memory allocation failure.");
+
+  if ((chandle = tjInitCompress()) == NULL ||
+      (dhandle = tjInitDecompress()) == NULL)
+    THROW_TJ();
+
+  for (pfi = 0; pfi < nformats; pfi++) {
+    for (i = 0; i < 2; i++) {
+      int flags = 0;
+
+      if (subsamp == TJSAMP_422 || subsamp == TJSAMP_420 ||
+          subsamp == TJSAMP_440 || subsamp == TJSAMP_411)
+        flags |= TJFLAG_FASTUPSAMPLE;
+      if (i == 1) flags |= TJFLAG_BOTTOMUP;
+      pf = formatspfi;
+      compTest(chandle, &dstBuf, &size, w, h, pf, basename, subsamp, 100,
+               flags);
+      decompTest(dhandle, dstBuf, size, w, h, pf, basename, subsamp, flags);
+      if (pf >= TJPF_RGBX && pf <= TJPF_XRGB) {
+        printf("\n");
+        decompTest(dhandle, dstBuf, size, w, h, pf + (TJPF_RGBA - TJPF_RGBX),
+                   basename, subsamp, flags);
+      }
+      printf("\n");
+    }
+  }
+  printf("--------------------\n\n");
+
+bailout:
+  if (chandle) tjDestroy(chandle);
+  if (dhandle) tjDestroy(dhandle);
+  tjFree(dstBuf);
+}
+
+
+#if SIZEOF_SIZE_T == 8
+#define CHECKSIZE(function) { \
+  if ((unsigned long long)size < (unsigned long long)0xFFFFFFFF) \
+    THROW(#function " overflow"); \
+}
+#else
+#define CHECKSIZE(function) { \
+  if (size != (unsigned long)(-1) || \
+      !strcmp(tjGetErrorStr2(NULL), "No error")) \
+    THROW(#function " overflow"); \
+}
+#endif
+
+static void overflowTest(void)
+{
+  /* Ensure that the various buffer size functions don't overflow */
+  unsigned long size;
+
+  size = tjBufSize(26755, 26755, TJSAMP_444);
+  CHECKSIZE(tjBufSize());
+  size = TJBUFSIZE(26755, 26755);
+  CHECKSIZE(TJBUFSIZE());
+  size = tjBufSizeYUV2(37838, 1, 37838, TJSAMP_444);
+  CHECKSIZE(tjBufSizeYUV2());
+  size = TJBUFSIZEYUV(37838, 37838, TJSAMP_444);
+  CHECKSIZE(TJBUFSIZEYUV());
+  size = tjBufSizeYUV(37838, 37838, TJSAMP_444);
+  CHECKSIZE(tjBufSizeYUV());
+  size = tjPlaneSizeYUV(0, 65536, 0, 65536, TJSAMP_444);
+  CHECKSIZE(tjPlaneSizeYUV());
+
+bailout:
+  return;
+}
+
+
+static void bufSizeTest(void)
+{
+  int w, h, i, subsamp;
+  unsigned char *srcBuf = NULL, *dstBuf = NULL;
+  tjhandle handle = NULL;
+  unsigned long dstSize = 0;
+
+  if ((handle = tjInitCompress()) == NULL) THROW_TJ();
+
+  printf("Buffer size regression test\n");
+  for (subsamp = 0; subsamp < TJ_NUMSAMP; subsamp++) {
+    for (w = 1; w < 48; w++) {
+      int maxh = (w == 1) ? 2048 : 48;
+
+      for (h = 1; h < maxh; h++) {
+        if (h % 100 == 0) printf("%.4d x %.4d\b\b\b\b\b\b\b\b\b\b\b", w, h);
+        if ((srcBuf = (unsigned char *)malloc(w * h * 4)) == NULL)
+          THROW("Memory allocation failure");
+        if (!alloc || doYUV) {
+          if (doYUV) dstSize = tjBufSizeYUV2(w, pad, h, subsamp);
+          else dstSize = tjBufSize(w, h, subsamp);
+          if ((dstBuf = (unsigned char *)tjAlloc(dstSize)) == NULL)
+            THROW("Memory allocation failure");
+        }
+
+        for (i = 0; i < w * h * 4; i++) {
+          if (random() < RAND_MAX / 2) srcBufi = 0;
+          else srcBufi = 255;
+        }
+
+        if (doYUV) {
+          TRY_TJ(tjEncodeYUV3(handle, srcBuf, w, 0, h, TJPF_BGRX, dstBuf, pad,
+                              subsamp, 0));
+        } else {
+          TRY_TJ(tjCompress2(handle, srcBuf, w, 0, h, TJPF_BGRX, &dstBuf,
+                             &dstSize, subsamp, 100,
+                             alloc ? 0 : TJFLAG_NOREALLOC));
+        }
+        free(srcBuf);  srcBuf = NULL;
+        if (!alloc || doYUV) {
+          tjFree(dstBuf);  dstBuf = NULL;
+        }
+
+        if ((srcBuf = (unsigned char *)malloc(h * w * 4)) == NULL)
+          THROW("Memory allocation failure");
+        if (!alloc || doYUV) {
+          if (doYUV) dstSize = tjBufSizeYUV2(h, pad, w, subsamp);
+          else dstSize = tjBufSize(h, w, subsamp);
+          if ((dstBuf = (unsigned char *)tjAlloc(dstSize)) == NULL)
+            THROW("Memory allocation failure");
+        }
+
+        for (i = 0; i < h * w * 4; i++) {
+          if (random() < RAND_MAX / 2) srcBufi = 0;
+          else srcBufi = 255;
+        }
+
+        if (doYUV) {
+          TRY_TJ(tjEncodeYUV3(handle, srcBuf, h, 0, w, TJPF_BGRX, dstBuf, pad,
+                              subsamp, 0));
+        } else {
+          TRY_TJ(tjCompress2(handle, srcBuf, h, 0, w, TJPF_BGRX, &dstBuf,
+                             &dstSize, subsamp, 100,
+                             alloc ? 0 : TJFLAG_NOREALLOC));
+        }
+        free(srcBuf);  srcBuf = NULL;
+        if (!alloc || doYUV) {
+          tjFree(dstBuf);  dstBuf = NULL;
+        }
+      }
+    }
+  }
+  printf("Done.      \n");
+
+bailout:
+  free(srcBuf);
+  tjFree(dstBuf);
+  if (handle) tjDestroy(handle);
+}
+
+
+static void initBitmap(unsigned char *buf, int width, int pitch, int height,
+                       int pf, int flags)
+{
+  int roffset = tjRedOffsetpf;
+  int goffset = tjGreenOffsetpf;
+  int boffset = tjBlueOffsetpf;
+  int ps = tjPixelSizepf;
+  int i, j;
+
+  for (j = 0; j < height; j++) {
+    int row = (flags & TJFLAG_BOTTOMUP) ? height - j - 1 : j;
+
+    for (i = 0; i < width; i++) {
+      unsigned char r = (i * 256 / width) % 256;
+      unsigned char g = (j * 256 / height) % 256;
+      unsigned char b = (j * 256 / height + i * 256 / width) % 256;
+
+      memset(&bufrow * pitch + i * ps, 0, ps);
+      if (pf == TJPF_GRAY) bufrow * pitch + i * ps = b;
+      else if (pf == TJPF_CMYK)
+        rgb_to_cmyk(r, g, b, &bufrow * pitch + i * ps + 0,
+                    &bufrow * pitch + i * ps + 1,
+                    &bufrow * pitch + i * ps + 2,
+                    &bufrow * pitch + i * ps + 3);
+      else {
+        bufrow * pitch + i * ps + roffset = r;
+        bufrow * pitch + i * ps + goffset = g;
+        bufrow * pitch + i * ps + boffset = b;
+      }
+    }
+  }
+}
+
+
+static int cmpBitmap(unsigned char *buf, int width, int pitch, int height,
+                     int pf, int flags, int gray2rgb)
+{
+  int roffset = tjRedOffsetpf;
+  int goffset = tjGreenOffsetpf;
+  int boffset = tjBlueOffsetpf;
+  int aoffset = tjAlphaOffsetpf;
+  int ps = tjPixelSizepf;
+  int i, j;
+
+  for (j = 0; j < height; j++) {
+    int row = (flags & TJFLAG_BOTTOMUP) ? height - j - 1 : j;
+
+    for (i = 0; i < width; i++) {
+      unsigned char r = (i * 256 / width) % 256;
+      unsigned char g = (j * 256 / height) % 256;
+      unsigned char b = (j * 256 / height + i * 256 / width) % 256;
+
+      if (pf == TJPF_GRAY) {
+        if (bufrow * pitch + i * ps != b)
+          return 0;
+      } else if (pf == TJPF_CMYK) {
+        unsigned char rf, gf, bf;
+
+        cmyk_to_rgb(bufrow * pitch + i * ps + 0,
+                    bufrow * pitch + i * ps + 1,
+                    bufrow * pitch + i * ps + 2,
+                    bufrow * pitch + i * ps + 3, &rf, &gf, &bf);
+        if (gray2rgb) {
+          if (rf != b || gf != b || bf != b)
+            return 0;
+        } else if (rf != r || gf != g || bf != b) return 0;
+      } else {
+        if (gray2rgb) {
+          if (bufrow * pitch + i * ps + roffset != b ||
+              bufrow * pitch + i * ps + goffset != b ||
+              bufrow * pitch + i * ps + boffset != b)
+            return 0;
+        } else if (bufrow * pitch + i * ps + roffset != r ||
+                   bufrow * pitch + i * ps + goffset != g ||
+                   bufrow * pitch + i * ps + boffset != b)
+          return 0;
+        if (aoffset >= 0 && bufrow * pitch + i * ps + aoffset != 0xFF)
+          return 0;
+      }
+    }
+  }
+  return 1;
+}
+
+
+static int doBmpTest(const char *ext, int width, int align, int height, int pf,
+                     int flags)
+{
+  char filename80, *md5sum, md5buf65;
+  int ps = tjPixelSizepf, pitch = PAD(width * ps, align), loadWidth = 0,
+    loadHeight = 0, retval = 0, pixelFormat = pf;
+  unsigned char *buf = NULL;
+  char *md5ref;
+
+  if (pf == TJPF_GRAY) {
+    md5ref = !strcasecmp(ext, "ppm") ? "112c682e82ce5de1cca089e20d60000b" :
+                                       "51976530acf75f02beddf5d21149101d";
+  } else {
+    md5ref = !strcasecmp(ext, "ppm") ? "c0c9f772b464d1896326883a5c79c545" :
+                                       "6d659071b9bfcdee2def22cb58ddadca";
+  }
+
+  if ((buf = (unsigned char *)tjAlloc(pitch * height)) == NULL)
+    THROW("Could not allocate memory");
+  initBitmap(buf, width, pitch, height, pf, flags);
+
+  snprintf(filename, 80, "test_bmp_%s_%d_%s.%s", pixFormatStrpf, align,
+           (flags & TJFLAG_BOTTOMUP) ? "bu" : "td", ext);
+  TRY_TJ(tjSaveImage(filename, buf, width, pitch, height, pf, flags));
+  md5sum = MD5File(filename, md5buf);
+  if (strcasecmp(md5sum, md5ref))
+    THROW_MD5(filename, md5sum, md5ref);
+
+  tjFree(buf);  buf = NULL;
+  if ((buf = tjLoadImage(filename, &loadWidth, align, &loadHeight, &pf,
+                         flags)) == NULL)
+    THROW_TJ();
+  if (width != loadWidth || height != loadHeight) {
+    printf("\n   Image dimensions of %s are bogus\n", filename);
+    retval = -1;  goto bailout;
+  }
+  if (!cmpBitmap(buf, width, pitch, height, pf, flags, 0)) {
+    printf("\n   Pixel data in %s is bogus\n", filename);
+    retval = -1;  goto bailout;
+  }
+  if (pf == TJPF_GRAY) {
+    tjFree(buf);  buf = NULL;
+    pf = TJPF_XBGR;
+    if ((buf = tjLoadImage(filename, &loadWidth, align, &loadHeight, &pf,
+                           flags)) == NULL)
+      THROW_TJ();
+    pitch = PAD(width * tjPixelSizepf, align);
+    if (!cmpBitmap(buf, width, pitch, height, pf, flags, 1)) {
+      printf("\n   Converting %s to RGB failed\n", filename);
+      retval = -1;  goto bailout;
+    }
+
+    tjFree(buf);  buf = NULL;
+    pf = TJPF_CMYK;
+    if ((buf = tjLoadImage(filename, &loadWidth, align, &loadHeight, &pf,
+                           flags)) == NULL)
+      THROW_TJ();
+    pitch = PAD(width * tjPixelSizepf, align);
+    if (!cmpBitmap(buf, width, pitch, height, pf, flags, 1)) {
+      printf("\n   Converting %s to CMYK failed\n", filename);
+      retval = -1;  goto bailout;
+    }
+  }
+  /* Verify that tjLoadImage() returns the proper "preferred" pixel format for
+     the file type. */
+  tjFree(buf);  buf = NULL;
+  pf = pixelFormat;
+  pixelFormat = TJPF_UNKNOWN;
+  if ((buf = tjLoadImage(filename, &loadWidth, align, &loadHeight,
+                         &pixelFormat, flags)) == NULL)
+    THROW_TJ();
+  if ((pf == TJPF_GRAY && pixelFormat != TJPF_GRAY) ||
+      (pf != TJPF_GRAY && !strcasecmp(ext, "bmp") &&
+       pixelFormat != TJPF_BGR) ||
+      (pf != TJPF_GRAY && !strcasecmp(ext, "ppm") &&
+       pixelFormat != TJPF_RGB)) {
+    printf("\n   tjLoadImage() returned unexpected pixel format: %s\n",
+           pixFormatStrpixelFormat);
+    retval = -1;
+  }
+  unlink(filename);
+
+bailout:
+  tjFree(buf);
+  if (exitStatus < 0) return exitStatus;
+  return retval;
+}
+
+
+static int bmpTest(void)
+{
+  int align, width = 35, height = 39, format;
+
+  for (align = 1; align <= 8; align *= 2) {
+    for (format = 0; format < TJ_NUMPF; format++) {
+      printf("%s Top-Down BMP (row alignment = %d bytes)  ...  ",
+             pixFormatStrformat, align);
+      if (doBmpTest("bmp", width, align, height, format, 0) == -1)
+        return -1;
+      printf("OK.\n");
+
+      printf("%s Top-Down PPM (row alignment = %d bytes)  ...  ",
+             pixFormatStrformat, align);
+      if (doBmpTest("ppm", width, align, height, format,
+                    TJFLAG_BOTTOMUP) == -1)
+        return -1;
+      printf("OK.\n");
+
+      printf("%s Bottom-Up BMP (row alignment = %d bytes)  ...  ",
+             pixFormatStrformat, align);
+      if (doBmpTest("bmp", width, align, height, format, 0) == -1)
+        return -1;
+      printf("OK.\n");
+
+      printf("%s Bottom-Up PPM (row alignment = %d bytes)  ...  ",
+             pixFormatStrformat, align);
+      if (doBmpTest("ppm", width, align, height, format,
+                    TJFLAG_BOTTOMUP) == -1)
+        return -1;
+      printf("OK.\n");
+    }
+  }
+
+  return 0;
+}
+
+
+int main(int argc, char *argv)
+{
+  int i, num4bf = 5;
+
+#ifdef _WIN32
+  srand((unsigned int)time(NULL));
+#endif
+  if (argc > 1) {
+    for (i = 1; i < argc; i++) {
+      if (!strcasecmp(argvi, "-yuv")) doYUV = 1;
+      else if (!strcasecmp(argvi, "-noyuvpad")) pad = 1;
+      else if (!strcasecmp(argvi, "-alloc")) alloc = 1;
+      else if (!strcasecmp(argvi, "-bmp")) return bmpTest();
+      else usage(argv0);
+    }
+  }
+  if (alloc) printf("Testing automatic buffer allocation\n");
+  if (doYUV) num4bf = 4;
+  overflowTest();
+  doTest(35, 39, _3byteFormats, 2, TJSAMP_444, "test");
+  doTest(39, 41, _4byteFormats, num4bf, TJSAMP_444, "test");
+  doTest(41, 35, _3byteFormats, 2, TJSAMP_422, "test");
+  doTest(35, 39, _4byteFormats, num4bf, TJSAMP_422, "test");
+  doTest(39, 41, _3byteFormats, 2, TJSAMP_420, "test");
+  doTest(41, 35, _4byteFormats, num4bf, TJSAMP_420, "test");
+  doTest(35, 39, _3byteFormats, 2, TJSAMP_440, "test");
+  doTest(39, 41, _4byteFormats, num4bf, TJSAMP_440, "test");
+  doTest(41, 35, _3byteFormats, 2, TJSAMP_411, "test");
+  doTest(35, 39, _4byteFormats, num4bf, TJSAMP_411, "test");
+  doTest(39, 41, _onlyGray, 1, TJSAMP_GRAY, "test");
+  doTest(41, 35, _3byteFormats, 2, TJSAMP_GRAY, "test");
+  doTest(35, 39, _4byteFormats, 4, TJSAMP_GRAY, "test");
+  bufSizeTest();
+  if (doYUV) {
+    printf("\n--------------------\n\n");
+    doTest(48, 48, _onlyRGB, 1, TJSAMP_444, "test_yuv0");
+    doTest(48, 48, _onlyRGB, 1, TJSAMP_422, "test_yuv0");
+    doTest(48, 48, _onlyRGB, 1, TJSAMP_420, "test_yuv0");
+    doTest(48, 48, _onlyRGB, 1, TJSAMP_440, "test_yuv0");
+    doTest(48, 48, _onlyRGB, 1, TJSAMP_411, "test_yuv0");
+    doTest(48, 48, _onlyRGB, 1, TJSAMP_GRAY, "test_yuv0");
+    doTest(48, 48, _onlyGray, 1, TJSAMP_GRAY, "test_yuv0");
+  }
+
+  return exitStatus;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/tjutil.c Added

@@ -0,0 +1,70 @@
+/*
+ * Copyright (C)2011, 2019 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifdef _WIN32
+
+#include <windows.h>
+#include "tjutil.h"
+
+static double getFreq(void)
+{
+  LARGE_INTEGER freq;
+
+  if (!QueryPerformanceFrequency(&freq)) return 0.0;
+  return (double)freq.QuadPart;
+}
+
+static double f = -1.0;
+
+double getTime(void)
+{
+  LARGE_INTEGER t;
+
+  if (f < 0.0) f = getFreq();
+  if (f == 0.0) return (double)GetTickCount() / 1000.;
+  else {
+    QueryPerformanceCounter(&t);
+    return (double)t.QuadPart / f;
+  }
+}
+
+#else
+
+#include <stdlib.h>
+#include <sys/time.h>
+#include "tjutil.h"
+
+double getTime(void)
+{
+  struct timeval tv;
+
+  if (gettimeofday(&tv, NULL) < 0) return 0.0;
+  else return (double)tv.tv_sec + ((double)tv.tv_usec / 1000000.);
+}
+
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/tjutil.h Added

@@ -0,0 +1,47 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifdef _WIN32
+#ifndef __MINGW32__
+#include <stdio.h>
+#define snprintf(str, n, format, ...) \
+  _snprintf_s(str, n, _TRUNCATE, format, ##__VA_ARGS__)
+#endif
+#define strcasecmp  stricmp
+#define strncasecmp  strnicmp
+#endif
+
+#ifndef min
+#define min(a, b)  ((a) < (b) ? (a) : (b))
+#endif
+
+#ifndef max
+#define max(a, b)  ((a) > (b) ? (a) : (b))
+#endif
+
+extern double getTime(void);

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/transupp.c Added

@@ -0,0 +1,2371 @@
+/*
+ * transupp.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1997-2019, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2017, 2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains image transformation routines and other utility code
+ * used by the jpegtran sample application.  These are NOT part of the core
+ * JPEG library.  But we keep these routines separate from jpegtran.c to
+ * ease the task of maintaining jpegtran-like programs that have other user
+ * interfaces.
+ */
+
+/* Although this file really shouldn't have access to the library internals,
+ * it's helpful to let it call jround_up() and jcopy_block_row().
+ */
+#define JPEG_INTERNALS
+
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "transupp.h"           /* My own external interface */
+#include "jpegcomp.h"
+#include <ctype.h>              /* to declare isdigit() */
+
+
+#if JPEG_LIB_VERSION >= 70
+#define dstinfo_min_DCT_h_scaled_size  dstinfo->min_DCT_h_scaled_size
+#define dstinfo_min_DCT_v_scaled_size  dstinfo->min_DCT_v_scaled_size
+#else
+#define dstinfo_min_DCT_h_scaled_size  DCTSIZE
+#define dstinfo_min_DCT_v_scaled_size  DCTSIZE
+#endif
+
+
+#if TRANSFORMS_SUPPORTED
+
+/*
+ * Lossless image transformation routines.  These routines work on DCT
+ * coefficient arrays and thus do not require any lossy decompression
+ * or recompression of the image.
+ * Thanks to Guido Vollbeding for the initial design and code of this feature,
+ * and to Ben Jackson for introducing the cropping feature.
+ *
+ * Horizontal flipping is done in-place, using a single top-to-bottom
+ * pass through the virtual source array.  It will thus be much the
+ * fastest option for images larger than main memory.
+ *
+ * The other routines require a set of destination virtual arrays, so they
+ * need twice as much memory as jpegtran normally does.  The destination
+ * arrays are always written in normal scan order (top to bottom) because
+ * the virtual array manager expects this.  The source arrays will be scanned
+ * in the corresponding order, which means multiple passes through the source
+ * arrays for most of the transforms.  That could result in much thrashing
+ * if the image is larger than main memory.
+ *
+ * If cropping or trimming is involved, the destination arrays may be smaller
+ * than the source arrays.  Note it is not possible to do horizontal flip
+ * in-place when a nonzero Y crop offset is specified, since we'd have to move
+ * data from one block row to another but the virtual array manager doesn't
+ * guarantee we can touch more than one row at a time.  So in that case,
+ * we have to use a separate destination array.
+ *
+ * Some notes about the operating environment of the individual transform
+ * routines:
+ * 1. Both the source and destination virtual arrays are allocated from the
+ *    source JPEG object, and therefore should be manipulated by calling the
+ *    source's memory manager.
+ * 2. The destination's component count should be used.  It may be smaller
+ *    than the source's when forcing to grayscale.
+ * 3. Likewise the destination's sampling factors should be used.  When
+ *    forcing to grayscale the destination's sampling factors will be all 1,
+ *    and we may as well take that as the effective iMCU size.
+ * 4. When "trim" is in effect, the destination's dimensions will be the
+ *    trimmed values but the source's will be untrimmed.
+ * 5. When "crop" is in effect, the destination's dimensions will be the
+ *    cropped values but the source's will be uncropped.  Each transform
+ *    routine is responsible for picking up source data starting at the
+ *    correct X and Y offset for the crop region.  (The X and Y offsets
+ *    passed to the transform routines are measured in iMCU blocks of the
+ *    destination.)
+ * 6. All the routines assume that the source and destination buffers are
+ *    padded out to a full iMCU boundary.  This is true, although for the
+ *    source buffer it is an undocumented property of jdcoefct.c.
+ */
+
+
+LOCAL(void)
+dequant_comp(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+             jvirt_barray_ptr coef_array, JQUANT_TBL *qtblptr1)
+{
+  JDIMENSION blk_x, blk_y;
+  int offset_y, k;
+  JQUANT_TBL *qtblptr;
+  JBLOCKARRAY buffer;
+  JBLOCKROW block;
+  JCOEFPTR ptr;
+
+  qtblptr = compptr->quant_table;
+  for (blk_y = 0; blk_y < compptr->height_in_blocks;
+       blk_y += compptr->v_samp_factor) {
+    buffer = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr)cinfo, coef_array, blk_y,
+       (JDIMENSION)compptr->v_samp_factor, TRUE);
+    for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+      block = bufferoffset_y;
+      for (blk_x = 0; blk_x < compptr->width_in_blocks; blk_x++) {
+        ptr = blockblk_x;
+        for (k = 0; k < DCTSIZE2; k++)
+          if (qtblptr->quantvalk != qtblptr1->quantvalk)
+            ptrk *= qtblptr->quantvalk / qtblptr1->quantvalk;
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+requant_comp(j_decompress_ptr cinfo, jpeg_component_info *compptr,
+             jvirt_barray_ptr coef_array, JQUANT_TBL *qtblptr1)
+{
+  JDIMENSION blk_x, blk_y;
+  int offset_y, k;
+  JQUANT_TBL *qtblptr;
+  JBLOCKARRAY buffer;
+  JBLOCKROW block;
+  JCOEFPTR ptr;
+  JCOEF temp, qval;
+
+  qtblptr = compptr->quant_table;
+  for (blk_y = 0; blk_y < compptr->height_in_blocks;
+       blk_y += compptr->v_samp_factor) {
+    buffer = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr)cinfo, coef_array, blk_y,
+       (JDIMENSION)compptr->v_samp_factor, TRUE);
+    for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+      block = bufferoffset_y;
+      for (blk_x = 0; blk_x < compptr->width_in_blocks; blk_x++) {
+        ptr = blockblk_x;
+        for (k = 0; k < DCTSIZE2; k++) {
+          temp = qtblptr->quantvalk;
+          qval = qtblptr1->quantvalk;
+          if (temp != qval) {
+            temp *= ptrk;
+            /* The following quantization code is copied from jcdctmgr.c */
+#ifdef FAST_DIVIDE
+#define DIVIDE_BY(a, b)  a /= b
+#else
+#define DIVIDE_BY(a, b)  if (a >= b) a /= b;  else a = 0
+#endif
+            if (temp < 0) {
+              temp = -temp;
+              temp += qval >> 1; /* for rounding */
+              DIVIDE_BY(temp, qval);
+              temp = -temp;
+            } else {
+              temp += qval >> 1; /* for rounding */
+              DIVIDE_BY(temp, qval);
+            }
+            ptrk = temp;
+          }
+        }
+      }
+    }
+  }
+}
+
+
+/*
+ * Calculate largest common denominator using Euclid's algorithm.
+ */
+LOCAL(JCOEF)
+largest_common_denominator(JCOEF a, JCOEF b)
+{
+  JCOEF c;
+
+  do {
+    c = a % b;
+    a = b;
+    b = c;
+  } while (c);
+
+  return a;
+}
+
+
+LOCAL(void)
+adjust_quant(j_decompress_ptr srcinfo, jvirt_barray_ptr *src_coef_arrays,
+             j_decompress_ptr dropinfo, jvirt_barray_ptr *drop_coef_arrays,
+             boolean trim, j_compress_ptr dstinfo)
+{
+  jpeg_component_info *compptr1, *compptr2;
+  JQUANT_TBL *qtblptr1, *qtblptr2, *qtblptr3;
+  int ci, k;
+
+  for (ci = 0; ci < dstinfo->num_components && ci < dropinfo->num_components;
+       ci++) {
+    compptr1 = srcinfo->comp_info + ci;
+    compptr2 = dropinfo->comp_info + ci;
+    qtblptr1 = compptr1->quant_table;
+    qtblptr2 = compptr2->quant_table;
+    for (k = 0; k < DCTSIZE2; k++) {
+      if (qtblptr1->quantvalk != qtblptr2->quantvalk) {
+        if (trim)
+          requant_comp(dropinfo, compptr2, drop_coef_arraysci, qtblptr1);
+        else {
+          qtblptr3 = dstinfo->quant_tbl_ptrscompptr1->quant_tbl_no;
+          for (k = 0; k < DCTSIZE2; k++)
+            if (qtblptr1->quantvalk != qtblptr2->quantvalk)
+              qtblptr3->quantvalk =
+                largest_common_denominator(qtblptr1->quantvalk,
+                                           qtblptr2->quantvalk);
+          dequant_comp(srcinfo, compptr1, src_coef_arraysci, qtblptr3);
+          dequant_comp(dropinfo, compptr2, drop_coef_arraysci, qtblptr3);
+        }
+        break;
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_drop(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+        JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+        jvirt_barray_ptr *src_coef_arrays,
+        j_decompress_ptr dropinfo, jvirt_barray_ptr *drop_coef_arrays,
+        JDIMENSION drop_width, JDIMENSION drop_height)
+/* Drop (insert) the contents of another image into the source image.  If the
+ * number of components in the drop image is smaller than the number of
+ * components in the destination image, then we fill in the remaining
+ * components with zero.  This allows for dropping the contents of grayscale
+ * images into (arbitrarily sampled) color images.
+ */
+{
+  JDIMENSION comp_width, comp_height;
+  JDIMENSION blk_y, x_drop_blocks, y_drop_blocks;
+  int ci, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  jpeg_component_info *compptr;
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = drop_width * compptr->h_samp_factor;
+    comp_height = drop_height * compptr->v_samp_factor;
+    x_drop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_drop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (blk_y = 0; blk_y < comp_height; blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, src_coef_arraysci, blk_y + y_drop_blocks,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      if (ci < dropinfo->num_components) {
+        src_buffer = (*dropinfo->mem->access_virt_barray)
+          ((j_common_ptr)dropinfo, drop_coef_arraysci, blk_y,
+           (JDIMENSION)compptr->v_samp_factor, FALSE);
+        for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+          jcopy_block_row(src_bufferoffset_y,
+                          dst_bufferoffset_y + x_drop_blocks, comp_width);
+        }
+      } else {
+        for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+          MEMZERO(dst_bufferoffset_y + x_drop_blocks,
+                  comp_width * sizeof(JBLOCK));
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_crop(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+        JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+        jvirt_barray_ptr *src_coef_arrays,
+        jvirt_barray_ptr *dst_coef_arrays)
+/* Crop.  This is only used when no rotate/flip is requested with the crop. */
+{
+  JDIMENSION dst_blk_y, x_crop_blocks, y_crop_blocks;
+  int ci, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  jpeg_component_info *compptr;
+
+  /* We simply have to copy the right amount of data (the destination's
+   * image size) starting at the given X and Y offsets in the source.
+   */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      src_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, src_coef_arraysci, dst_blk_y + y_crop_blocks,
+         (JDIMENSION)compptr->v_samp_factor, FALSE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        jcopy_block_row(src_bufferoffset_y + x_crop_blocks,
+                        dst_bufferoffset_y, compptr->width_in_blocks);
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_crop_ext_zero(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                 JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+                 jvirt_barray_ptr *src_coef_arrays,
+                 jvirt_barray_ptr *dst_coef_arrays)
+/* Crop.  This is only used when no rotate/flip is requested with the crop.
+ * Extension: If the destination size is larger than the source, we fill in the
+ * expanded region with zero (neutral gray).  Note that we also have to zero
+ * partial iMCUs at the right and bottom edge of the source image area in this
+ * case.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height;
+  JDIMENSION dst_blk_y, x_crop_blocks, y_crop_blocks;
+  int ci, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  jpeg_component_info *compptr;
+
+  MCU_cols = srcinfo->output_width /
+             (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+  MCU_rows = srcinfo->output_height /
+             (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      if (dstinfo->_jpeg_height > srcinfo->output_height) {
+        if (dst_blk_y < y_crop_blocks ||
+            dst_blk_y >= y_crop_blocks + comp_height) {
+          for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+            MEMZERO(dst_bufferoffset_y,
+                    compptr->width_in_blocks * sizeof(JBLOCK));
+          }
+          continue;
+        }
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           dst_blk_y - y_crop_blocks, (JDIMENSION)compptr->v_samp_factor,
+           FALSE);
+      } else {
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           dst_blk_y + y_crop_blocks, (JDIMENSION)compptr->v_samp_factor,
+           FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        if (dstinfo->_jpeg_width > srcinfo->output_width) {
+          if (x_crop_blocks > 0) {
+            MEMZERO(dst_bufferoffset_y, x_crop_blocks * sizeof(JBLOCK));
+          }
+          jcopy_block_row(src_bufferoffset_y,
+                          dst_bufferoffset_y + x_crop_blocks, comp_width);
+          if (compptr->width_in_blocks > x_crop_blocks + comp_width) {
+            MEMZERO(dst_bufferoffset_y + x_crop_blocks + comp_width,
+                    (compptr->width_in_blocks - x_crop_blocks - comp_width) *
+                    sizeof(JBLOCK));
+          }
+        } else {
+          jcopy_block_row(src_bufferoffset_y + x_crop_blocks,
+                          dst_bufferoffset_y, compptr->width_in_blocks);
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_crop_ext_flat(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                 JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+                 jvirt_barray_ptr *src_coef_arrays,
+                 jvirt_barray_ptr *dst_coef_arrays)
+/* Crop.  This is only used when no rotate/flip is requested with the crop.
+ * Extension: The destination width is larger than the source, and we fill in
+ * the expanded region with the DC coefficient of the adjacent block.  Note
+ * that we also have to fill partial iMCUs at the right and bottom edge of the
+ * source image area in this case.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height;
+  JDIMENSION dst_blk_x, dst_blk_y, x_crop_blocks, y_crop_blocks;
+  int ci, offset_y;
+  JCOEF dc;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  jpeg_component_info *compptr;
+
+  MCU_cols = srcinfo->output_width /
+             (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+  MCU_rows = srcinfo->output_height /
+             (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      if (dstinfo->_jpeg_height > srcinfo->output_height) {
+        if (dst_blk_y < y_crop_blocks ||
+            dst_blk_y >= y_crop_blocks + comp_height) {
+          for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+            MEMZERO(dst_bufferoffset_y,
+                    compptr->width_in_blocks * sizeof(JBLOCK));
+          }
+          continue;
+        }
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           dst_blk_y - y_crop_blocks, (JDIMENSION)compptr->v_samp_factor,
+           FALSE);
+      } else {
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           dst_blk_y + y_crop_blocks, (JDIMENSION)compptr->v_samp_factor,
+          FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        if (x_crop_blocks > 0) {
+          MEMZERO(dst_bufferoffset_y, x_crop_blocks * sizeof(JBLOCK));
+          dc = src_bufferoffset_y00;
+          for (dst_blk_x = 0; dst_blk_x < x_crop_blocks; dst_blk_x++) {
+            dst_bufferoffset_ydst_blk_x0 = dc;
+          }
+        }
+        jcopy_block_row(src_bufferoffset_y,
+                        dst_bufferoffset_y + x_crop_blocks, comp_width);
+        if (compptr->width_in_blocks > x_crop_blocks + comp_width) {
+          MEMZERO(dst_bufferoffset_y + x_crop_blocks + comp_width,
+                  (compptr->width_in_blocks - x_crop_blocks - comp_width) *
+                  sizeof(JBLOCK));
+          dc = src_bufferoffset_ycomp_width - 10;
+          for (dst_blk_x = x_crop_blocks + comp_width;
+               dst_blk_x < compptr->width_in_blocks; dst_blk_x++) {
+            dst_bufferoffset_ydst_blk_x0 = dc;
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_crop_ext_reflect(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                    JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+                    jvirt_barray_ptr *src_coef_arrays,
+                    jvirt_barray_ptr *dst_coef_arrays)
+/* Crop.  This is only used when no rotate/flip is requested with the crop.
+ * Extension: The destination width is larger than the source, and we fill in
+ * the expanded region with repeated reflections of the source image.  Note
+ * that we also have to fill partial iMCUs at the right and bottom edge of the
+ * source image area in this case.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height, src_blk_x;
+  JDIMENSION dst_blk_x, dst_blk_y, x_crop_blocks, y_crop_blocks;
+  int ci, k, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  MCU_cols = srcinfo->output_width /
+             (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+  MCU_rows = srcinfo->output_height /
+             (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      if (dstinfo->_jpeg_height > srcinfo->output_height) {
+        if (dst_blk_y < y_crop_blocks ||
+            dst_blk_y >= y_crop_blocks + comp_height) {
+          for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+            MEMZERO(dst_bufferoffset_y,
+                    compptr->width_in_blocks * sizeof(JBLOCK));
+          }
+          continue;
+        }
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           dst_blk_y - y_crop_blocks, (JDIMENSION)compptr->v_samp_factor,
+           FALSE);
+      } else {
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           dst_blk_y + y_crop_blocks, (JDIMENSION)compptr->v_samp_factor,
+           FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        /* Copy source region */
+        jcopy_block_row(src_bufferoffset_y,
+                        dst_bufferoffset_y + x_crop_blocks, comp_width);
+        if (x_crop_blocks > 0) {
+          /* Reflect to left */
+          dst_row_ptr = dst_bufferoffset_y + x_crop_blocks;
+          for (dst_blk_x = x_crop_blocks; dst_blk_x > 0;) {
+            src_row_ptr = dst_row_ptr;      /* (re)set axis of reflection */
+            for (src_blk_x = comp_width; src_blk_x > 0 && dst_blk_x > 0;
+                 src_blk_x--, dst_blk_x--) {
+              dst_ptr = *(--dst_row_ptr);   /* destination goes left */
+              src_ptr = *src_row_ptr++;     /* source goes right */
+              /* This unrolled loop doesn't need to know which row it's on. */
+              for (k = 0; k < DCTSIZE2; k += 2) {
+                *dst_ptr++ = *src_ptr++;    /* copy even column */
+                *dst_ptr++ = -(*src_ptr++); /* copy odd column with sign
+                                               change */
+              }
+            }
+          }
+        }
+        if (compptr->width_in_blocks > x_crop_blocks + comp_width) {
+          /* Reflect to right */
+          dst_row_ptr = dst_bufferoffset_y + x_crop_blocks + comp_width;
+          for (dst_blk_x = compptr->width_in_blocks - x_crop_blocks - comp_width;
+               dst_blk_x > 0;) {
+            src_row_ptr = dst_row_ptr;      /* (re)set axis of reflection */
+            for (src_blk_x = comp_width; src_blk_x > 0 && dst_blk_x > 0;
+                 src_blk_x--, dst_blk_x--) {
+              dst_ptr = *dst_row_ptr++;     /* destination goes right */
+              src_ptr = *(--src_row_ptr);   /* source goes left */
+              /* This unrolled loop doesn't need to know which row it's on. */
+              for (k = 0; k < DCTSIZE2; k += 2) {
+                *dst_ptr++ = *src_ptr++;    /* copy even column */
+                *dst_ptr++ = -(*src_ptr++); /* copy odd column with sign
+                                               change */
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_wipe(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+        JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+        jvirt_barray_ptr *src_coef_arrays,
+        JDIMENSION drop_width, JDIMENSION drop_height)
+/* Wipe - discard image contents of specified region and fill with zero
+ * (neutral gray)
+ */
+{
+  JDIMENSION x_wipe_blocks, wipe_width;
+  JDIMENSION y_wipe_blocks, wipe_bottom;
+  int ci, offset_y;
+  JBLOCKARRAY buffer;
+  jpeg_component_info *compptr;
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    x_wipe_blocks = x_crop_offset * compptr->h_samp_factor;
+    wipe_width = drop_width * compptr->h_samp_factor;
+    y_wipe_blocks = y_crop_offset * compptr->v_samp_factor;
+    wipe_bottom = drop_height * compptr->v_samp_factor + y_wipe_blocks;
+    for (; y_wipe_blocks < wipe_bottom;
+         y_wipe_blocks += compptr->v_samp_factor) {
+      buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, src_coef_arraysci, y_wipe_blocks,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        MEMZERO(bufferoffset_y + x_wipe_blocks, wipe_width * sizeof(JBLOCK));
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flatten(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+           jvirt_barray_ptr *src_coef_arrays,
+           JDIMENSION drop_width, JDIMENSION drop_height)
+/* Flatten - discard image contents of specified region, similarly to wipe,
+ * but fill with the average of adjacent blocks instead of zero.
+ */
+{
+  JDIMENSION x_wipe_blocks, wipe_width, wipe_right;
+  JDIMENSION y_wipe_blocks, wipe_bottom, blk_x;
+  int ci, offset_y, dc_left_value, dc_right_value, average;
+  JBLOCKARRAY buffer;
+  jpeg_component_info *compptr;
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    x_wipe_blocks = x_crop_offset * compptr->h_samp_factor;
+    wipe_width = drop_width * compptr->h_samp_factor;
+    wipe_right = wipe_width + x_wipe_blocks;
+    y_wipe_blocks = y_crop_offset * compptr->v_samp_factor;
+    wipe_bottom = drop_height * compptr->v_samp_factor + y_wipe_blocks;
+    for (; y_wipe_blocks < wipe_bottom;
+         y_wipe_blocks += compptr->v_samp_factor) {
+      buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, src_coef_arraysci, y_wipe_blocks,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        MEMZERO(bufferoffset_y + x_wipe_blocks, wipe_width * sizeof(JBLOCK));
+        if (x_wipe_blocks > 0) {
+          dc_left_value = bufferoffset_yx_wipe_blocks - 10;
+          if (wipe_right < compptr->width_in_blocks) {
+            dc_right_value = bufferoffset_ywipe_right0;
+            average = (dc_left_value + dc_right_value) >> 1;
+          } else {
+            average = dc_left_value;
+          }
+        } else if (wipe_right < compptr->width_in_blocks) {
+          average = bufferoffset_ywipe_right0;
+        } else continue;
+        for (blk_x = x_wipe_blocks; blk_x < wipe_right; blk_x++) {
+          bufferoffset_yblk_x0 = (JCOEF)average;
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_reflect(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, jvirt_barray_ptr *src_coef_arrays,
+           JDIMENSION drop_width, JDIMENSION drop_height)
+/* Reflect - discard image contents of specified region, similarly to wipe,
+ * but fill with repeated reflections of the outside region instead of zero.
+ * NB: y_crop_offset is assumed to be zero.
+ */
+{
+  JDIMENSION x_wipe_blocks, wipe_width;
+  JDIMENSION y_wipe_blocks, wipe_bottom;
+  JDIMENSION src_blk_x, dst_blk_x;
+  int ci, k, offset_y;
+  JBLOCKARRAY buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    x_wipe_blocks = x_crop_offset * compptr->h_samp_factor;
+    wipe_width = drop_width * compptr->h_samp_factor;
+    wipe_bottom = drop_height * compptr->v_samp_factor;
+    for (y_wipe_blocks = 0; y_wipe_blocks < wipe_bottom;
+         y_wipe_blocks += compptr->v_samp_factor) {
+      buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, src_coef_arraysci, y_wipe_blocks,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        if (x_wipe_blocks > 0) {
+          /* Reflect from left */
+          dst_row_ptr = bufferoffset_y + x_wipe_blocks;
+          for (dst_blk_x = wipe_width; dst_blk_x > 0;) {
+            src_row_ptr = dst_row_ptr;     /* (re)set axis of reflection */
+            for (src_blk_x = x_wipe_blocks;
+                 src_blk_x > 0 && dst_blk_x > 0; src_blk_x--, dst_blk_x--) {
+              dst_ptr = *dst_row_ptr++;    /* destination goes right */
+              src_ptr = *(--src_row_ptr);  /* source goes left */
+              /* this unrolled loop doesn't need to know which row it's on... */
+              for (k = 0; k < DCTSIZE2; k += 2) {
+                *dst_ptr++ = *src_ptr++;   /* copy even column */
+                *dst_ptr++ = -(*src_ptr++); /* copy odd column with sign change */
+              }
+            }
+          }
+        } else if (compptr->width_in_blocks > x_wipe_blocks + wipe_width) {
+          /* Reflect from right */
+          dst_row_ptr = bufferoffset_y + x_wipe_blocks + wipe_width;
+          for (dst_blk_x = wipe_width; dst_blk_x > 0;) {
+            src_row_ptr = dst_row_ptr;     /* (re)set axis of reflection */
+            src_blk_x = compptr->width_in_blocks - x_wipe_blocks - wipe_width;
+            for (; src_blk_x > 0 && dst_blk_x > 0; src_blk_x--, dst_blk_x--) {
+              dst_ptr = *(--dst_row_ptr);  /* destination goes left */
+              src_ptr = *src_row_ptr++;    /* source goes right */
+              /* this unrolled loop doesn't need to know which row it's on... */
+              for (k = 0; k < DCTSIZE2; k += 2) {
+                *dst_ptr++ = *src_ptr++;   /* copy even column */
+                *dst_ptr++ = -(*src_ptr++); /* copy odd column with sign change */
+              }
+            }
+          }
+        } else {
+          MEMZERO(bufferoffset_y + x_wipe_blocks,
+                  wipe_width * sizeof(JBLOCK));
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_h_no_crop(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                  JDIMENSION x_crop_offset, jvirt_barray_ptr *src_coef_arrays)
+/* Horizontal flip; done in-place, so no separate dest array is required.
+ * NB: this only works when y_crop_offset is zero.
+ */
+{
+  JDIMENSION MCU_cols, comp_width, blk_x, blk_y, x_crop_blocks;
+  int ci, k, offset_y;
+  JBLOCKARRAY buffer;
+  JCOEFPTR ptr1, ptr2;
+  JCOEF temp1, temp2;
+  jpeg_component_info *compptr;
+
+  /* Horizontal mirroring of DCT blocks is accomplished by swapping
+   * pairs of blocks in-place.  Within a DCT block, we perform horizontal
+   * mirroring by changing the signs of odd-numbered columns.
+   * Partial iMCUs at the right edge are left untouched.
+   */
+  MCU_cols = srcinfo->output_width /
+             (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    for (blk_y = 0; blk_y < compptr->height_in_blocks;
+         blk_y += compptr->v_samp_factor) {
+      buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, src_coef_arraysci, blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        /* Do the mirroring */
+        for (blk_x = 0; blk_x * 2 < comp_width; blk_x++) {
+          ptr1 = bufferoffset_yblk_x;
+          ptr2 = bufferoffset_ycomp_width - blk_x - 1;
+          /* this unrolled loop doesn't need to know which row it's on... */
+          for (k = 0; k < DCTSIZE2; k += 2) {
+            temp1 = *ptr1;      /* swap even column */
+            temp2 = *ptr2;
+            *ptr1++ = temp2;
+            *ptr2++ = temp1;
+            temp1 = *ptr1;      /* swap odd column with sign change */
+            temp2 = *ptr2;
+            *ptr1++ = -temp2;
+            *ptr2++ = -temp1;
+          }
+        }
+        if (x_crop_blocks > 0) {
+          /* Now left-justify the portion of the data to be kept.
+           * We can't use a single jcopy_block_row() call because that routine
+           * depends on memcpy(), whose behavior is unspecified for overlapping
+           * source and destination areas.  Sigh.
+           */
+          for (blk_x = 0; blk_x < compptr->width_in_blocks; blk_x++) {
+            jcopy_block_row(bufferoffset_y + blk_x + x_crop_blocks,
+                            bufferoffset_y + blk_x, (JDIMENSION)1);
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_h(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+          JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+          jvirt_barray_ptr *src_coef_arrays,
+          jvirt_barray_ptr *dst_coef_arrays)
+/* Horizontal flip in general cropping case */
+{
+  JDIMENSION MCU_cols, comp_width, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, k, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Here we must output into a separate array because we can't touch
+   * different rows of a single virtual array simultaneously.  Otherwise,
+   * this is essentially the same as the routine above.
+   */
+  MCU_cols = srcinfo->output_width /
+             (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      src_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, src_coef_arraysci, dst_blk_y + y_crop_blocks,
+         (JDIMENSION)compptr->v_samp_factor, FALSE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        dst_row_ptr = dst_bufferoffset_y;
+        src_row_ptr = src_bufferoffset_y;
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x++) {
+          if (x_crop_blocks + dst_blk_x < comp_width) {
+            /* Do the mirrorable blocks */
+            dst_ptr = dst_row_ptrdst_blk_x;
+            src_ptr = src_row_ptrcomp_width - x_crop_blocks - dst_blk_x - 1;
+            /* this unrolled loop doesn't need to know which row it's on... */
+            for (k = 0; k < DCTSIZE2; k += 2) {
+              *dst_ptr++ = *src_ptr++;    /* copy even column */
+              *dst_ptr++ = -(*src_ptr++); /* copy odd column with sign
+                                             change */
+            }
+          } else {
+            /* Copy last partial block(s) verbatim */
+            jcopy_block_row(src_row_ptr + dst_blk_x + x_crop_blocks,
+                            dst_row_ptr + dst_blk_x, (JDIMENSION)1);
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_v(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+          JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+          jvirt_barray_ptr *src_coef_arrays,
+          jvirt_barray_ptr *dst_coef_arrays)
+/* Vertical flip */
+{
+  JDIMENSION MCU_rows, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* We output into a separate array because we can't touch different
+   * rows of the source virtual array simultaneously.  Otherwise, this
+   * is a pretty straightforward analog of horizontal flip.
+   * Within a DCT block, vertical mirroring is done by changing the signs
+   * of odd-numbered rows.
+   * Partial iMCUs at the bottom edge are copied verbatim.
+   */
+  MCU_rows = srcinfo->output_height /
+             (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      if (y_crop_blocks + dst_blk_y < comp_height) {
+        /* Row is within the mirrorable area. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           comp_height - y_crop_blocks - dst_blk_y -
+           (JDIMENSION)compptr->v_samp_factor,
+           (JDIMENSION)compptr->v_samp_factor, FALSE);
+      } else {
+        /* Bottom-edge blocks will be copied verbatim. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           dst_blk_y + y_crop_blocks,
+           (JDIMENSION)compptr->v_samp_factor, FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        if (y_crop_blocks + dst_blk_y < comp_height) {
+          /* Row is within the mirrorable area. */
+          dst_row_ptr = dst_bufferoffset_y;
+          src_row_ptr = src_buffercompptr->v_samp_factor - offset_y - 1;
+          src_row_ptr += x_crop_blocks;
+          for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+               dst_blk_x++) {
+            dst_ptr = dst_row_ptrdst_blk_x;
+            src_ptr = src_row_ptrdst_blk_x;
+            for (i = 0; i < DCTSIZE; i += 2) {
+              /* copy even row */
+              for (j = 0; j < DCTSIZE; j++)
+                *dst_ptr++ = *src_ptr++;
+              /* copy odd row with sign change */
+              for (j = 0; j < DCTSIZE; j++)
+                *dst_ptr++ = -(*src_ptr++);
+            }
+          }
+        } else {
+          /* Just copy row verbatim. */
+          jcopy_block_row(src_bufferoffset_y + x_crop_blocks,
+                          dst_bufferoffset_y, compptr->width_in_blocks);
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_transpose(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+             JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+             jvirt_barray_ptr *src_coef_arrays,
+             jvirt_barray_ptr *dst_coef_arrays)
+/* Transpose source into destination */
+{
+  JDIMENSION dst_blk_x, dst_blk_y, x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Transposing pixels within a block just requires transposing the
+   * DCT coefficients.
+   * Partial iMCUs at the edges require no special treatment; we simply
+   * process all the available DCT blocks for every component.
+   */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          src_buffer = (*srcinfo->mem->access_virt_barray)
+            ((j_common_ptr)srcinfo, src_coef_arraysci,
+             dst_blk_x + x_crop_blocks,
+             (JDIMENSION)compptr->h_samp_factor, FALSE);
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_bufferoffset_ydst_blk_x + offset_x;
+            src_ptr =
+              src_bufferoffset_xdst_blk_y + offset_y + y_crop_blocks;
+            for (i = 0; i < DCTSIZE; i++)
+              for (j = 0; j < DCTSIZE; j++)
+                dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_90(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+          JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+          jvirt_barray_ptr *src_coef_arrays,
+          jvirt_barray_ptr *dst_coef_arrays)
+/* 90 degree rotation is equivalent to
+ *   1. Transposing the image;
+ *   2. Horizontal mirroring.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, comp_width, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Because of the horizontal mirror step, we can't process partial iMCUs
+   * at the (output) right edge properly.  They just get transposed and
+   * not mirrored.
+   */
+  MCU_cols = srcinfo->output_height /
+             (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          if (x_crop_blocks + dst_blk_x < comp_width) {
+            /* Block is within the mirrorable area. */
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr)srcinfo, src_coef_arraysci,
+               comp_width - x_crop_blocks - dst_blk_x -
+               (JDIMENSION)compptr->h_samp_factor,
+               (JDIMENSION)compptr->h_samp_factor, FALSE);
+          } else {
+            /* Edge blocks are transposed but not mirrored. */
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr)srcinfo, src_coef_arraysci,
+               dst_blk_x + x_crop_blocks,
+               (JDIMENSION)compptr->h_samp_factor, FALSE);
+          }
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_bufferoffset_ydst_blk_x + offset_x;
+            if (x_crop_blocks + dst_blk_x < comp_width) {
+              /* Block is within the mirrorable area. */
+              src_ptr = src_buffercompptr->h_samp_factor - offset_x - 1
+                dst_blk_y + offset_y + y_crop_blocks;
+              for (i = 0; i < DCTSIZE; i++) {
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+                i++;
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptrj * DCTSIZE + i = -src_ptri * DCTSIZE + j;
+              }
+            } else {
+              /* Edge blocks are transposed but not mirrored. */
+              src_ptr = src_bufferoffset_x
+                dst_blk_y + offset_y + y_crop_blocks;
+              for (i = 0; i < DCTSIZE; i++)
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_270(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+           jvirt_barray_ptr *src_coef_arrays,
+           jvirt_barray_ptr *dst_coef_arrays)
+/* 270 degree rotation is equivalent to
+ *   1. Horizontal mirroring;
+ *   2. Transposing the image.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_rows, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Because of the horizontal mirror step, we can't process partial iMCUs
+   * at the (output) bottom edge properly.  They just get transposed and
+   * not mirrored.
+   */
+  MCU_rows = srcinfo->output_width /
+             (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          src_buffer = (*srcinfo->mem->access_virt_barray)
+            ((j_common_ptr)srcinfo, src_coef_arraysci,
+             dst_blk_x + x_crop_blocks,
+             (JDIMENSION)compptr->h_samp_factor, FALSE);
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_bufferoffset_ydst_blk_x + offset_x;
+            if (y_crop_blocks + dst_blk_y < comp_height) {
+              /* Block is within the mirrorable area. */
+              src_ptr = src_bufferoffset_x
+                comp_height - y_crop_blocks - dst_blk_y - offset_y - 1;
+              for (i = 0; i < DCTSIZE; i++) {
+                for (j = 0; j < DCTSIZE; j++) {
+                  dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+                  j++;
+                  dst_ptrj * DCTSIZE + i = -src_ptri * DCTSIZE + j;
+                }
+              }
+            } else {
+              /* Edge blocks are transposed but not mirrored. */
+              src_ptr = src_bufferoffset_x
+                dst_blk_y + offset_y + y_crop_blocks;
+              for (i = 0; i < DCTSIZE; i++)
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_180(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+           jvirt_barray_ptr *src_coef_arrays,
+           jvirt_barray_ptr *dst_coef_arrays)
+/* 180 degree rotation is equivalent to
+ *   1. Vertical mirroring;
+ *   2. Horizontal mirroring.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  MCU_cols = srcinfo->output_width /
+             (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+  MCU_rows = srcinfo->output_height /
+             (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      if (y_crop_blocks + dst_blk_y < comp_height) {
+        /* Row is within the vertically mirrorable area. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           comp_height - y_crop_blocks - dst_blk_y -
+           (JDIMENSION)compptr->v_samp_factor,
+           (JDIMENSION)compptr->v_samp_factor, FALSE);
+      } else {
+        /* Bottom-edge rows are only mirrored horizontally. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr)srcinfo, src_coef_arraysci,
+           dst_blk_y + y_crop_blocks,
+           (JDIMENSION)compptr->v_samp_factor, FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        dst_row_ptr = dst_bufferoffset_y;
+        if (y_crop_blocks + dst_blk_y < comp_height) {
+          /* Row is within the mirrorable area. */
+          src_row_ptr = src_buffercompptr->v_samp_factor - offset_y - 1;
+          for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+               dst_blk_x++) {
+            dst_ptr = dst_row_ptrdst_blk_x;
+            if (x_crop_blocks + dst_blk_x < comp_width) {
+              /* Process the blocks that can be mirrored both ways. */
+              src_ptr =
+                src_row_ptrcomp_width - x_crop_blocks - dst_blk_x - 1;
+              for (i = 0; i < DCTSIZE; i += 2) {
+                /* For even row, negate every odd column. */
+                for (j = 0; j < DCTSIZE; j += 2) {
+                  *dst_ptr++ = *src_ptr++;
+                  *dst_ptr++ = -(*src_ptr++);
+                }
+                /* For odd row, negate every even column. */
+                for (j = 0; j < DCTSIZE; j += 2) {
+                  *dst_ptr++ = -(*src_ptr++);
+                  *dst_ptr++ = *src_ptr++;
+                }
+              }
+            } else {
+              /* Any remaining right-edge blocks are only mirrored vertically. */
+              src_ptr = src_row_ptrx_crop_blocks + dst_blk_x;
+              for (i = 0; i < DCTSIZE; i += 2) {
+                for (j = 0; j < DCTSIZE; j++)
+                  *dst_ptr++ = *src_ptr++;
+                for (j = 0; j < DCTSIZE; j++)
+                  *dst_ptr++ = -(*src_ptr++);
+              }
+            }
+          }
+        } else {
+          /* Remaining rows are just mirrored horizontally. */
+          src_row_ptr = src_bufferoffset_y;
+          for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+               dst_blk_x++) {
+            if (x_crop_blocks + dst_blk_x < comp_width) {
+              /* Process the blocks that can be mirrored. */
+              dst_ptr = dst_row_ptrdst_blk_x;
+              src_ptr =
+                src_row_ptrcomp_width - x_crop_blocks - dst_blk_x - 1;
+              for (i = 0; i < DCTSIZE2; i += 2) {
+                *dst_ptr++ = *src_ptr++;
+                *dst_ptr++ = -(*src_ptr++);
+              }
+            } else {
+              /* Any remaining right-edge blocks are only copied. */
+              jcopy_block_row(src_row_ptr + dst_blk_x + x_crop_blocks,
+                              dst_row_ptr + dst_blk_x, (JDIMENSION)1);
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_transverse(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+              JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+              jvirt_barray_ptr *src_coef_arrays,
+              jvirt_barray_ptr *dst_coef_arrays)
+/* Transverse transpose is equivalent to
+ *   1. 180 degree rotation;
+ *   2. Transposition;
+ * or
+ *   1. Horizontal mirroring;
+ *   2. Transposition;
+ *   3. Horizontal mirroring.
+ * These steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  MCU_cols = srcinfo->output_height /
+             (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+  MCU_rows = srcinfo->output_width /
+             (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr)srcinfo, dst_coef_arraysci, dst_blk_y,
+         (JDIMENSION)compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          if (x_crop_blocks + dst_blk_x < comp_width) {
+            /* Block is within the mirrorable area. */
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr)srcinfo, src_coef_arraysci,
+               comp_width - x_crop_blocks - dst_blk_x -
+               (JDIMENSION)compptr->h_samp_factor,
+               (JDIMENSION)compptr->h_samp_factor, FALSE);
+          } else {
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr)srcinfo, src_coef_arraysci,
+               dst_blk_x + x_crop_blocks,
+               (JDIMENSION)compptr->h_samp_factor, FALSE);
+          }
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_bufferoffset_ydst_blk_x + offset_x;
+            if (y_crop_blocks + dst_blk_y < comp_height) {
+              if (x_crop_blocks + dst_blk_x < comp_width) {
+                /* Block is within the mirrorable area. */
+                src_ptr = src_buffercompptr->h_samp_factor - offset_x - 1
+                  comp_height - y_crop_blocks - dst_blk_y - offset_y - 1;
+                for (i = 0; i < DCTSIZE; i++) {
+                  for (j = 0; j < DCTSIZE; j++) {
+                    dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+                    j++;
+                    dst_ptrj * DCTSIZE + i = -src_ptri * DCTSIZE + j;
+                  }
+                  i++;
+                  for (j = 0; j < DCTSIZE; j++) {
+                    dst_ptrj * DCTSIZE + i = -src_ptri * DCTSIZE + j;
+                    j++;
+                    dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+                  }
+                }
+              } else {
+                /* Right-edge blocks are mirrored in y only */
+                src_ptr = src_bufferoffset_x
+                  comp_height - y_crop_blocks - dst_blk_y - offset_y - 1;
+                for (i = 0; i < DCTSIZE; i++) {
+                  for (j = 0; j < DCTSIZE; j++) {
+                    dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+                    j++;
+                    dst_ptrj * DCTSIZE + i = -src_ptri * DCTSIZE + j;
+                  }
+                }
+              }
+            } else {
+              if (x_crop_blocks + dst_blk_x < comp_width) {
+                /* Bottom-edge blocks are mirrored in x only */
+                src_ptr = src_buffercompptr->h_samp_factor - offset_x - 1
+                  dst_blk_y + offset_y + y_crop_blocks;
+                for (i = 0; i < DCTSIZE; i++) {
+                  for (j = 0; j < DCTSIZE; j++)
+                    dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+                  i++;
+                  for (j = 0; j < DCTSIZE; j++)
+                    dst_ptrj * DCTSIZE + i = -src_ptri * DCTSIZE + j;
+                }
+              } else {
+                /* At lower right corner, just transpose, no mirroring */
+                src_ptr = src_bufferoffset_x
+                  dst_blk_y + offset_y + y_crop_blocks;
+                for (i = 0; i < DCTSIZE; i++)
+                  for (j = 0; j < DCTSIZE; j++)
+                    dst_ptrj * DCTSIZE + i = src_ptri * DCTSIZE + j;
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+/* Parse an unsigned integer: subroutine for jtransform_parse_crop_spec.
+ * Returns TRUE if valid integer found, FALSE if not.
+ * *strptr is advanced over the digit string, and *result is set to its value.
+ */
+
+LOCAL(boolean)
+jt_read_integer(const char **strptr, JDIMENSION *result)
+{
+  const char *ptr = *strptr;
+  JDIMENSION val = 0;
+
+  for (; isdigit(*ptr); ptr++) {
+    val = val * 10 + (JDIMENSION)(*ptr - '0');
+  }
+  *result = val;
+  if (ptr == *strptr)
+    return FALSE;               /* oops, no digits */
+  *strptr = ptr;
+  return TRUE;
+}
+
+
+/* Parse a crop specification (written in X11 geometry style).
+ * The routine returns TRUE if the spec string is valid, FALSE if not.
+ *
+ * The crop spec string should have the format
+ *      <width>{fr}x<height>{fr}{+-}<xoffset>{+-}<yoffset>
+ * where width, height, xoffset, and yoffset are unsigned integers.
+ * Each of the elements can be omitted to indicate a default value.
+ * (A weakness of this style is that it is not possible to omit xoffset
+ * while specifying yoffset, since they look alike.)
+ *
+ * This code is loosely based on XParseGeometry from the X11 distribution.
+ */
+
+GLOBAL(boolean)
+jtransform_parse_crop_spec(jpeg_transform_info *info, const char *spec)
+{
+  info->crop = FALSE;
+  info->crop_width_set = JCROP_UNSET;
+  info->crop_height_set = JCROP_UNSET;
+  info->crop_xoffset_set = JCROP_UNSET;
+  info->crop_yoffset_set = JCROP_UNSET;
+
+  if (isdigit(*spec)) {
+    /* fetch width */
+    if (!jt_read_integer(&spec, &info->crop_width))
+      return FALSE;
+    if (*spec == 'f' || *spec == 'F') {
+      spec++;
+      info->crop_width_set = JCROP_FORCE;
+    } else if (*spec == 'r' || *spec == 'R') {
+      spec++;
+      info->crop_width_set = JCROP_REFLECT;
+    } else
+      info->crop_width_set = JCROP_POS;
+  }
+  if (*spec == 'x' || *spec == 'X') {
+    /* fetch height */
+    spec++;
+    if (!jt_read_integer(&spec, &info->crop_height))
+      return FALSE;
+    if (*spec == 'f' || *spec == 'F') {
+      spec++;
+      info->crop_height_set = JCROP_FORCE;
+    } else if (*spec == 'r' || *spec == 'R') {
+      spec++;
+      info->crop_height_set = JCROP_REFLECT;
+    } else
+      info->crop_height_set = JCROP_POS;
+  }
+  if (*spec == '+' || *spec == '-') {
+    /* fetch xoffset */
+    info->crop_xoffset_set = (*spec == '-') ? JCROP_NEG : JCROP_POS;
+    spec++;
+    if (!jt_read_integer(&spec, &info->crop_xoffset))
+      return FALSE;
+  }
+  if (*spec == '+' || *spec == '-') {
+    /* fetch yoffset */
+    info->crop_yoffset_set = (*spec == '-') ? JCROP_NEG : JCROP_POS;
+    spec++;
+    if (!jt_read_integer(&spec, &info->crop_yoffset))
+      return FALSE;
+  }
+  /* We had better have gotten to the end of the string. */
+  if (*spec != '\0')
+    return FALSE;
+  info->crop = TRUE;
+  return TRUE;
+}
+
+
+/* Trim off any partial iMCUs on the indicated destination edge */
+
+LOCAL(void)
+trim_right_edge(jpeg_transform_info *info, JDIMENSION full_width)
+{
+  JDIMENSION MCU_cols;
+
+  MCU_cols = info->output_width / info->iMCU_sample_width;
+  if (MCU_cols > 0 && info->x_crop_offset + MCU_cols ==
+      full_width / info->iMCU_sample_width)
+    info->output_width = MCU_cols * info->iMCU_sample_width;
+}
+
+LOCAL(void)
+trim_bottom_edge(jpeg_transform_info *info, JDIMENSION full_height)
+{
+  JDIMENSION MCU_rows;
+
+  MCU_rows = info->output_height / info->iMCU_sample_height;
+  if (MCU_rows > 0 && info->y_crop_offset + MCU_rows ==
+      full_height / info->iMCU_sample_height)
+    info->output_height = MCU_rows * info->iMCU_sample_height;
+}
+
+
+/* Request any required workspace.
+ *
+ * This routine figures out the size that the output image will be
+ * (which implies that all the transform parameters must be set before
+ * it is called).
+ *
+ * We allocate the workspace virtual arrays from the source decompression
+ * object, so that all the arrays (both the original data and the workspace)
+ * will be taken into account while making memory management decisions.
+ * Hence, this routine must be called after jpeg_read_header (which reads
+ * the image dimensions) and before jpeg_read_coefficients (which realizes
+ * the source's virtual arrays).
+ *
+ * This function returns FALSE right away if -perfect is given
+ * and transformation is not perfect.  Otherwise returns TRUE.
+ */
+
+GLOBAL(boolean)
+jtransform_request_workspace(j_decompress_ptr srcinfo,
+                             jpeg_transform_info *info)
+{
+  jvirt_barray_ptr *coef_arrays;
+  boolean need_workspace, transpose_it;
+  jpeg_component_info *compptr;
+  JDIMENSION xoffset, yoffset, dtemp;
+  JDIMENSION width_in_iMCUs, height_in_iMCUs;
+  JDIMENSION width_in_blocks, height_in_blocks;
+  int itemp, ci, h_samp_factor, v_samp_factor;
+
+  /* Determine number of components in output image */
+  if (info->force_grayscale &&
+      srcinfo->jpeg_color_space == JCS_YCbCr &&
+      srcinfo->num_components == 3)
+    /* We'll only process the first component */
+    info->num_components = 1;
+  else
+    /* Process all the components */
+    info->num_components = srcinfo->num_components;
+
+  /* Compute output image dimensions and related values. */
+#if JPEG_LIB_VERSION >= 80
+  jpeg_core_output_dimensions(srcinfo);
+#else
+  srcinfo->output_width = srcinfo->image_width;
+  srcinfo->output_height = srcinfo->image_height;
+#endif
+
+  /* Return right away if -perfect is given and transformation is not perfect.
+   */
+  if (info->perfect) {
+    if (info->num_components == 1) {
+      if (!jtransform_perfect_transform(srcinfo->output_width,
+          srcinfo->output_height,
+          srcinfo->_min_DCT_h_scaled_size,
+          srcinfo->_min_DCT_v_scaled_size,
+          info->transform))
+        return FALSE;
+    } else {
+      if (!jtransform_perfect_transform(srcinfo->output_width,
+          srcinfo->output_height,
+          srcinfo->max_h_samp_factor * srcinfo->_min_DCT_h_scaled_size,
+          srcinfo->max_v_samp_factor * srcinfo->_min_DCT_v_scaled_size,
+          info->transform))
+        return FALSE;
+    }
+  }
+
+  /* If there is only one output component, force the iMCU size to be 1;
+   * else use the source iMCU size.  (This allows us to do the right thing
+   * when reducing color to grayscale, and also provides a handy way of
+   * cleaning up "funny" grayscale images whose sampling factors are not 1x1.)
+   */
+  switch (info->transform) {
+  case JXFORM_TRANSPOSE:
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_90:
+  case JXFORM_ROT_270:
+    info->output_width = srcinfo->output_height;
+    info->output_height = srcinfo->output_width;
+    if (info->num_components == 1) {
+      info->iMCU_sample_width = srcinfo->_min_DCT_v_scaled_size;
+      info->iMCU_sample_height = srcinfo->_min_DCT_h_scaled_size;
+    } else {
+      info->iMCU_sample_width =
+        srcinfo->max_v_samp_factor * srcinfo->_min_DCT_v_scaled_size;
+      info->iMCU_sample_height =
+        srcinfo->max_h_samp_factor * srcinfo->_min_DCT_h_scaled_size;
+    }
+    break;
+  default:
+    info->output_width = srcinfo->output_width;
+    info->output_height = srcinfo->output_height;
+    if (info->num_components == 1) {
+      info->iMCU_sample_width = srcinfo->_min_DCT_h_scaled_size;
+      info->iMCU_sample_height = srcinfo->_min_DCT_v_scaled_size;
+    } else {
+      info->iMCU_sample_width =
+        srcinfo->max_h_samp_factor * srcinfo->_min_DCT_h_scaled_size;
+      info->iMCU_sample_height =
+        srcinfo->max_v_samp_factor * srcinfo->_min_DCT_v_scaled_size;
+    }
+    break;
+  }
+
+  /* If cropping has been requested, compute the crop area's position and
+   * dimensions, ensuring that its upper left corner falls at an iMCU boundary.
+   */
+  if (info->crop) {
+    /* Insert default values for unset crop parameters */
+    if (info->crop_xoffset_set == JCROP_UNSET)
+      info->crop_xoffset = 0;   /* default to +0 */
+    if (info->crop_yoffset_set == JCROP_UNSET)
+      info->crop_yoffset = 0;   /* default to +0 */
+    if (info->crop_width_set == JCROP_UNSET) {
+      if (info->crop_xoffset >= info->output_width)
+        ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+      info->crop_width = info->output_width - info->crop_xoffset;
+    } else {
+      /* Check for crop extension */
+      if (info->crop_width > info->output_width) {
+        /* Crop extension does not work when transforming! */
+        if (info->transform != JXFORM_NONE ||
+            info->crop_xoffset >= info->crop_width ||
+            info->crop_xoffset > info->crop_width - info->output_width)
+          ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+      } else {
+        if (info->crop_xoffset >= info->output_width ||
+            info->crop_width <= 0 ||
+            info->crop_xoffset > info->output_width - info->crop_width)
+          ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+      }
+    }
+    if (info->crop_height_set == JCROP_UNSET) {
+      if (info->crop_yoffset >= info->output_height)
+        ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+      info->crop_height = info->output_height - info->crop_yoffset;
+    } else {
+      /* Check for crop extension */
+      if (info->crop_height > info->output_height) {
+        /* Crop extension does not work when transforming! */
+        if (info->transform != JXFORM_NONE ||
+            info->crop_yoffset >= info->crop_height ||
+            info->crop_yoffset > info->crop_height - info->output_height)
+          ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+      } else {
+        if (info->crop_yoffset >= info->output_height ||
+            info->crop_height <= 0 ||
+            info->crop_yoffset > info->output_height - info->crop_height)
+          ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+      }
+    }
+    /* Convert negative crop offsets into regular offsets */
+    if (info->crop_xoffset_set != JCROP_NEG)
+      xoffset = info->crop_xoffset;
+    else if (info->crop_width > info->output_width) /* crop extension */
+      xoffset = info->crop_width - info->output_width - info->crop_xoffset;
+    else
+      xoffset = info->output_width - info->crop_width - info->crop_xoffset;
+    if (info->crop_yoffset_set != JCROP_NEG)
+      yoffset = info->crop_yoffset;
+    else if (info->crop_height > info->output_height) /* crop extension */
+      yoffset = info->crop_height - info->output_height - info->crop_yoffset;
+    else
+      yoffset = info->output_height - info->crop_height - info->crop_yoffset;
+    /* Now adjust so that upper left corner falls at an iMCU boundary */
+    switch (info->transform) {
+    case JXFORM_DROP:
+      /* Ensure the effective drop region will not exceed the requested */
+      itemp = info->iMCU_sample_width;
+      dtemp = itemp - 1 - ((xoffset + itemp - 1) % itemp);
+      xoffset += dtemp;
+      if (info->crop_width <= dtemp)
+        info->drop_width = 0;
+      else if (xoffset + info->crop_width - dtemp == info->output_width)
+        /* Matching right edge: include partial iMCU */
+        info->drop_width = (info->crop_width - dtemp + itemp - 1) / itemp;
+      else
+        info->drop_width = (info->crop_width - dtemp) / itemp;
+      itemp = info->iMCU_sample_height;
+      dtemp = itemp - 1 - ((yoffset + itemp - 1) % itemp);
+      yoffset += dtemp;
+      if (info->crop_height <= dtemp)
+        info->drop_height = 0;
+      else if (yoffset + info->crop_height - dtemp == info->output_height)
+        /* Matching bottom edge: include partial iMCU */
+        info->drop_height = (info->crop_height - dtemp + itemp - 1) / itemp;
+      else
+        info->drop_height = (info->crop_height - dtemp) / itemp;
+      /* Check if sampling factors match for dropping */
+      if (info->drop_width != 0 && info->drop_height != 0)
+        for (ci = 0; ci < info->num_components &&
+                     ci < info->drop_ptr->num_components; ci++) {
+          if (info->drop_ptr->comp_infoci.h_samp_factor *
+              srcinfo->max_h_samp_factor !=
+              srcinfo->comp_infoci.h_samp_factor *
+              info->drop_ptr->max_h_samp_factor)
+            ERREXIT6(srcinfo, JERR_BAD_DROP_SAMPLING, ci,
+              info->drop_ptr->comp_infoci.h_samp_factor,
+              info->drop_ptr->max_h_samp_factor,
+              srcinfo->comp_infoci.h_samp_factor,
+              srcinfo->max_h_samp_factor, 'h');
+          if (info->drop_ptr->comp_infoci.v_samp_factor *
+              srcinfo->max_v_samp_factor !=
+              srcinfo->comp_infoci.v_samp_factor *
+              info->drop_ptr->max_v_samp_factor)
+            ERREXIT6(srcinfo, JERR_BAD_DROP_SAMPLING, ci,
+              info->drop_ptr->comp_infoci.v_samp_factor,
+              info->drop_ptr->max_v_samp_factor,
+              srcinfo->comp_infoci.v_samp_factor,
+              srcinfo->max_v_samp_factor, 'v');
+        }
+      break;
+    case JXFORM_WIPE:
+      /* Ensure the effective wipe region will cover the requested */
+      info->drop_width = (JDIMENSION)jdiv_round_up
+        ((long)(info->crop_width + (xoffset % info->iMCU_sample_width)),
+         (long)info->iMCU_sample_width);
+      info->drop_height = (JDIMENSION)jdiv_round_up
+        ((long)(info->crop_height + (yoffset % info->iMCU_sample_height)),
+         (long)info->iMCU_sample_height);
+      break;
+    default:
+      /* Ensure the effective crop region will cover the requested */
+      if (info->crop_width_set == JCROP_FORCE ||
+          info->crop_width > info->output_width)
+        info->output_width = info->crop_width;
+      else
+        info->output_width =
+          info->crop_width + (xoffset % info->iMCU_sample_width);
+      if (info->crop_height_set == JCROP_FORCE ||
+          info->crop_height > info->output_height)
+        info->output_height = info->crop_height;
+      else
+        info->output_height =
+          info->crop_height + (yoffset % info->iMCU_sample_height);
+    }
+    /* Save x/y offsets measured in iMCUs */
+    info->x_crop_offset = xoffset / info->iMCU_sample_width;
+    info->y_crop_offset = yoffset / info->iMCU_sample_height;
+  } else {
+    info->x_crop_offset = 0;
+    info->y_crop_offset = 0;
+  }
+
+  /* Figure out whether we need workspace arrays,
+   * and if so whether they are transposed relative to the source.
+   */
+  need_workspace = FALSE;
+  transpose_it = FALSE;
+  switch (info->transform) {
+  case JXFORM_NONE:
+    if (info->x_crop_offset != 0 || info->y_crop_offset != 0 ||
+        info->output_width > srcinfo->output_width ||
+        info->output_height > srcinfo->output_height)
+      need_workspace = TRUE;
+    /* No workspace needed if neither cropping nor transforming */
+    break;
+  case JXFORM_FLIP_H:
+    if (info->trim)
+      trim_right_edge(info, srcinfo->output_width);
+    if (info->y_crop_offset != 0 || info->slow_hflip)
+      need_workspace = TRUE;
+    /* do_flip_h_no_crop doesn't need a workspace array */
+    break;
+  case JXFORM_FLIP_V:
+    if (info->trim)
+      trim_bottom_edge(info, srcinfo->output_height);
+    /* Need workspace arrays having same dimensions as source image. */
+    need_workspace = TRUE;
+    break;
+  case JXFORM_TRANSPOSE:
+    /* transpose does NOT have to trim anything */
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_TRANSVERSE:
+    if (info->trim) {
+      trim_right_edge(info, srcinfo->output_height);
+      trim_bottom_edge(info, srcinfo->output_width);
+    }
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_ROT_90:
+    if (info->trim)
+      trim_right_edge(info, srcinfo->output_height);
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_ROT_180:
+    if (info->trim) {
+      trim_right_edge(info, srcinfo->output_width);
+      trim_bottom_edge(info, srcinfo->output_height);
+    }
+    /* Need workspace arrays having same dimensions as source image. */
+    need_workspace = TRUE;
+    break;
+  case JXFORM_ROT_270:
+    if (info->trim)
+      trim_bottom_edge(info, srcinfo->output_width);
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_WIPE:
+    break;
+  case JXFORM_DROP:
+    break;
+  }
+
+  /* Allocate workspace if needed.
+   * Note that we allocate arrays padded out to the next iMCU boundary,
+   * so that transform routines need not worry about missing edge blocks.
+   */
+  if (need_workspace) {
+    coef_arrays = (jvirt_barray_ptr *)
+      (*srcinfo->mem->alloc_small) ((j_common_ptr)srcinfo, JPOOL_IMAGE,
+                sizeof(jvirt_barray_ptr) * info->num_components);
+    width_in_iMCUs = (JDIMENSION)
+      jdiv_round_up((long)info->output_width, (long)info->iMCU_sample_width);
+    height_in_iMCUs = (JDIMENSION)
+      jdiv_round_up((long)info->output_height, (long)info->iMCU_sample_height);
+    for (ci = 0; ci < info->num_components; ci++) {
+      compptr = srcinfo->comp_info + ci;
+      if (info->num_components == 1) {
+        /* we're going to force samp factors to 1x1 in this case */
+        h_samp_factor = v_samp_factor = 1;
+      } else if (transpose_it) {
+        h_samp_factor = compptr->v_samp_factor;
+        v_samp_factor = compptr->h_samp_factor;
+      } else {
+        h_samp_factor = compptr->h_samp_factor;
+        v_samp_factor = compptr->v_samp_factor;
+      }
+      width_in_blocks = width_in_iMCUs * h_samp_factor;
+      height_in_blocks = height_in_iMCUs * v_samp_factor;
+      coef_arraysci = (*srcinfo->mem->request_virt_barray)
+        ((j_common_ptr)srcinfo, JPOOL_IMAGE, FALSE,
+         width_in_blocks, height_in_blocks, (JDIMENSION)v_samp_factor);
+    }
+    info->workspace_coef_arrays = coef_arrays;
+  } else
+    info->workspace_coef_arrays = NULL;
+
+  return TRUE;
+}
+
+
+/* Transpose destination image parameters */
+
+LOCAL(void)
+transpose_critical_parameters(j_compress_ptr dstinfo)
+{
+  int tblno, i, j, ci, itemp;
+  jpeg_component_info *compptr;
+  JQUANT_TBL *qtblptr;
+  JDIMENSION jtemp;
+  UINT16 qtemp;
+
+  /* Transpose image dimensions */
+  jtemp = dstinfo->image_width;
+  dstinfo->image_width = dstinfo->image_height;
+  dstinfo->image_height = jtemp;
+#if JPEG_LIB_VERSION >= 70
+  itemp = dstinfo->min_DCT_h_scaled_size;
+  dstinfo->min_DCT_h_scaled_size = dstinfo->min_DCT_v_scaled_size;
+  dstinfo->min_DCT_v_scaled_size = itemp;
+#endif
+
+  /* Transpose sampling factors */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    itemp = compptr->h_samp_factor;
+    compptr->h_samp_factor = compptr->v_samp_factor;
+    compptr->v_samp_factor = itemp;
+  }
+
+  /* Transpose quantization tables */
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    qtblptr = dstinfo->quant_tbl_ptrstblno;
+    if (qtblptr != NULL) {
+      for (i = 0; i < DCTSIZE; i++) {
+        for (j = 0; j < i; j++) {
+          qtemp = qtblptr->quantvali * DCTSIZE + j;
+          qtblptr->quantvali * DCTSIZE + j =
+            qtblptr->quantvalj * DCTSIZE + i;
+          qtblptr->quantvalj * DCTSIZE + i = qtemp;
+        }
+      }
+    }
+  }
+}
+
+
+/* Adjust Exif image parameters.
+ *
+ * We try to adjust the Tags ExifImageWidth and ExifImageHeight if possible.
+ */
+
+LOCAL(void)
+adjust_exif_parameters(JOCTET *data, unsigned int length, JDIMENSION new_width,
+                       JDIMENSION new_height)
+{
+  boolean is_motorola; /* Flag for byte order */
+  unsigned int number_of_tags, tagnum;
+  unsigned int firstoffset, offset;
+  JDIMENSION new_value;
+
+  if (length < 12) return; /* Length of an IFD entry */
+
+  /* Discover byte order */
+  if (data0 == 0x49 && data1 == 0x49)
+    is_motorola = FALSE;
+  else if (data0 == 0x4D && data1 == 0x4D)
+    is_motorola = TRUE;
+  else
+    return;
+
+  /* Check Tag Mark */
+  if (is_motorola) {
+    if (data2 != 0) return;
+    if (data3 != 0x2A) return;
+  } else {
+    if (data3 != 0) return;
+    if (data2 != 0x2A) return;
+  }
+
+  /* Get first IFD offset (offset to IFD0) */
+  if (is_motorola) {
+    if (data4 != 0) return;
+    if (data5 != 0) return;
+    firstoffset = data6;
+    firstoffset <<= 8;
+    firstoffset += data7;
+  } else {
+    if (data7 != 0) return;
+    if (data6 != 0) return;
+    firstoffset = data5;
+    firstoffset <<= 8;
+    firstoffset += data4;
+  }
+  if (firstoffset > length - 2) return; /* check end of data segment */
+
+  /* Get the number of directory entries contained in this IFD */
+  if (is_motorola) {
+    number_of_tags = datafirstoffset;
+    number_of_tags <<= 8;
+    number_of_tags += datafirstoffset + 1;
+  } else {
+    number_of_tags = datafirstoffset + 1;
+    number_of_tags <<= 8;
+    number_of_tags += datafirstoffset;
+  }
+  if (number_of_tags == 0) return;
+  firstoffset += 2;
+
+  /* Search for ExifSubIFD offset Tag in IFD0 */
+  for (;;) {
+    if (firstoffset > length - 12) return; /* check end of data segment */
+    /* Get Tag number */
+    if (is_motorola) {
+      tagnum = datafirstoffset;
+      tagnum <<= 8;
+      tagnum += datafirstoffset + 1;
+    } else {
+      tagnum = datafirstoffset + 1;
+      tagnum <<= 8;
+      tagnum += datafirstoffset;
+    }
+    if (tagnum == 0x8769) break; /* found ExifSubIFD offset Tag */
+    if (--number_of_tags == 0) return;
+    firstoffset += 12;
+  }
+
+  /* Get the ExifSubIFD offset */
+  if (is_motorola) {
+    if (datafirstoffset + 8 != 0) return;
+    if (datafirstoffset + 9 != 0) return;
+    offset = datafirstoffset + 10;
+    offset <<= 8;
+    offset += datafirstoffset + 11;
+  } else {
+    if (datafirstoffset + 11 != 0) return;
+    if (datafirstoffset + 10 != 0) return;
+    offset = datafirstoffset + 9;
+    offset <<= 8;
+    offset += datafirstoffset + 8;
+  }
+  if (offset > length - 2) return; /* check end of data segment */
+
+  /* Get the number of directory entries contained in this SubIFD */
+  if (is_motorola) {
+    number_of_tags = dataoffset;
+    number_of_tags <<= 8;
+    number_of_tags += dataoffset + 1;
+  } else {
+    number_of_tags = dataoffset + 1;
+    number_of_tags <<= 8;
+    number_of_tags += dataoffset;
+  }
+  if (number_of_tags < 2) return;
+  offset += 2;
+
+  /* Search for ExifImageWidth and ExifImageHeight Tags in this SubIFD */
+  do {
+    if (offset > length - 12) return; /* check end of data segment */
+    /* Get Tag number */
+    if (is_motorola) {
+      tagnum = dataoffset;
+      tagnum <<= 8;
+      tagnum += dataoffset + 1;
+    } else {
+      tagnum = dataoffset + 1;
+      tagnum <<= 8;
+      tagnum += dataoffset;
+    }
+    if (tagnum == 0xA002 || tagnum == 0xA003) {
+      if (tagnum == 0xA002)
+        new_value = new_width; /* ExifImageWidth Tag */
+      else
+        new_value = new_height; /* ExifImageHeight Tag */
+      if (is_motorola) {
+        dataoffset + 2 = 0; /* Format = unsigned long (4 octets) */
+        dataoffset + 3 = 4;
+        dataoffset + 4 = 0; /* Number Of Components = 1 */
+        dataoffset + 5 = 0;
+        dataoffset + 6 = 0;
+        dataoffset + 7 = 1;
+        dataoffset + 8 = 0;
+        dataoffset + 9 = 0;
+        dataoffset + 10 = (JOCTET)((new_value >> 8) & 0xFF);
+        dataoffset + 11 = (JOCTET)(new_value & 0xFF);
+      } else {
+        dataoffset + 2 = 4; /* Format = unsigned long (4 octets) */
+        dataoffset + 3 = 0;
+        dataoffset + 4 = 1; /* Number Of Components = 1 */
+        dataoffset + 5 = 0;
+        dataoffset + 6 = 0;
+        dataoffset + 7 = 0;
+        dataoffset + 8 = (JOCTET)(new_value & 0xFF);
+        dataoffset + 9 = (JOCTET)((new_value >> 8) & 0xFF);
+        dataoffset + 10 = 0;
+        dataoffset + 11 = 0;
+      }
+    }
+    offset += 12;
+  } while (--number_of_tags);
+}
+
+
+/* Adjust output image parameters as needed.
+ *
+ * This must be called after jpeg_copy_critical_parameters()
+ * and before jpeg_write_coefficients().
+ *
+ * The return value is the set of virtual coefficient arrays to be written
+ * (either the ones allocated by jtransform_request_workspace, or the
+ * original source data arrays).  The caller will need to pass this value
+ * to jpeg_write_coefficients().
+ */
+
+GLOBAL(jvirt_barray_ptr *)
+jtransform_adjust_parameters(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                             jvirt_barray_ptr *src_coef_arrays,
+                             jpeg_transform_info *info)
+{
+  /* If force-to-grayscale is requested, adjust destination parameters */
+  if (info->force_grayscale) {
+    /* First, ensure we have YCbCr or grayscale data, and that the source's
+     * Y channel is full resolution.  (No reasonable person would make Y
+     * be less than full resolution, so actually coping with that case
+     * isn't worth extra code space.  But we check it to avoid crashing.)
+     */
+    if (((dstinfo->jpeg_color_space == JCS_YCbCr &&
+          dstinfo->num_components == 3) ||
+         (dstinfo->jpeg_color_space == JCS_GRAYSCALE &&
+          dstinfo->num_components == 1)) &&
+        srcinfo->comp_info0.h_samp_factor == srcinfo->max_h_samp_factor &&
+        srcinfo->comp_info0.v_samp_factor == srcinfo->max_v_samp_factor) {
+      /* We use jpeg_set_colorspace to make sure subsidiary settings get fixed
+       * properly.  Among other things, it sets the target h_samp_factor &
+       * v_samp_factor to 1, which typically won't match the source.
+       * We have to preserve the source's quantization table number, however.
+       */
+      int sv_quant_tbl_no = dstinfo->comp_info0.quant_tbl_no;
+      jpeg_set_colorspace(dstinfo, JCS_GRAYSCALE);
+      dstinfo->comp_info0.quant_tbl_no = sv_quant_tbl_no;
+    } else {
+      /* Sorry, can't do it */
+      ERREXIT(dstinfo, JERR_CONVERSION_NOTIMPL);
+    }
+  } else if (info->num_components == 1) {
+    /* For a single-component source, we force the destination sampling factors
+     * to 1x1, with or without force_grayscale.  This is useful because some
+     * decoders choke on grayscale images with other sampling factors.
+     */
+    dstinfo->comp_info0.h_samp_factor = 1;
+    dstinfo->comp_info0.v_samp_factor = 1;
+  }
+
+  /* Correct the destination's image dimensions as necessary
+   * for rotate/flip, resize, and crop operations.
+   */
+#if JPEG_LIB_VERSION >= 80
+  dstinfo->jpeg_width = info->output_width;
+  dstinfo->jpeg_height = info->output_height;
+#endif
+
+  /* Transpose destination image parameters, adjust quantization */
+  switch (info->transform) {
+  case JXFORM_TRANSPOSE:
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_90:
+  case JXFORM_ROT_270:
+#if JPEG_LIB_VERSION < 80
+    dstinfo->image_width = info->output_height;
+    dstinfo->image_height = info->output_width;
+#endif
+    transpose_critical_parameters(dstinfo);
+    break;
+  case JXFORM_DROP:
+    if (info->drop_width != 0 && info->drop_height != 0)
+      adjust_quant(srcinfo, src_coef_arrays,
+                   info->drop_ptr, info->drop_coef_arrays,
+                   info->trim, dstinfo);
+    break;
+  default:
+#if JPEG_LIB_VERSION < 80
+    dstinfo->image_width = info->output_width;
+    dstinfo->image_height = info->output_height;
+#endif
+    break;
+  }
+
+  /* Adjust Exif properties */
+  if (srcinfo->marker_list != NULL &&
+      srcinfo->marker_list->marker == JPEG_APP0 + 1 &&
+      srcinfo->marker_list->data_length >= 6 &&
+      srcinfo->marker_list->data0 == 0x45 &&
+      srcinfo->marker_list->data1 == 0x78 &&
+      srcinfo->marker_list->data2 == 0x69 &&
+      srcinfo->marker_list->data3 == 0x66 &&
+      srcinfo->marker_list->data4 == 0 &&
+      srcinfo->marker_list->data5 == 0) {
+    /* Suppress output of JFIF marker */
+    dstinfo->write_JFIF_header = FALSE;
+    /* Adjust Exif image parameters */
+#if JPEG_LIB_VERSION >= 80
+    if (dstinfo->jpeg_width != srcinfo->image_width ||
+        dstinfo->jpeg_height != srcinfo->image_height)
+      /* Align data segment to start of TIFF structure for parsing */
+      adjust_exif_parameters(srcinfo->marker_list->data + 6,
+                             srcinfo->marker_list->data_length - 6,
+                             dstinfo->jpeg_width, dstinfo->jpeg_height);
+#else
+    if (dstinfo->image_width != srcinfo->image_width ||
+        dstinfo->image_height != srcinfo->image_height)
+      /* Align data segment to start of TIFF structure for parsing */
+      adjust_exif_parameters(srcinfo->marker_list->data + 6,
+                             srcinfo->marker_list->data_length - 6,
+                             dstinfo->image_width, dstinfo->image_height);
+#endif
+  }
+
+  /* Return the appropriate output data set */
+  if (info->workspace_coef_arrays != NULL)
+    return info->workspace_coef_arrays;
+  return src_coef_arrays;
+}
+
+
+/* Execute the actual transformation, if any.
+ *
+ * This must be called *after* jpeg_write_coefficients, because it depends
+ * on jpeg_write_coefficients to have computed subsidiary values such as
+ * the per-component width and height fields in the destination object.
+ *
+ * Note that some transformations will modify the source data arrays!
+ */
+
+GLOBAL(void)
+jtransform_execute_transform(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                             jvirt_barray_ptr *src_coef_arrays,
+                             jpeg_transform_info *info)
+{
+  jvirt_barray_ptr *dst_coef_arrays = info->workspace_coef_arrays;
+
+  /* Note: conditions tested here should match those in switch statement
+   * in jtransform_request_workspace()
+   */
+  switch (info->transform) {
+  case JXFORM_NONE:
+    if (info->output_width > srcinfo->output_width ||
+        info->output_height > srcinfo->output_height) {
+      if (info->output_width > srcinfo->output_width &&
+          info->crop_width_set == JCROP_REFLECT)
+        do_crop_ext_reflect(srcinfo, dstinfo,
+                            info->x_crop_offset, info->y_crop_offset,
+                            src_coef_arrays, dst_coef_arrays);
+      else if (info->output_width > srcinfo->output_width &&
+               info->crop_width_set == JCROP_FORCE)
+        do_crop_ext_flat(srcinfo, dstinfo,
+                         info->x_crop_offset, info->y_crop_offset,
+                         src_coef_arrays, dst_coef_arrays);
+      else
+        do_crop_ext_zero(srcinfo, dstinfo,
+                         info->x_crop_offset, info->y_crop_offset,
+                         src_coef_arrays, dst_coef_arrays);
+    } else if (info->x_crop_offset != 0 || info->y_crop_offset != 0)
+      do_crop(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_FLIP_H:
+    if (info->y_crop_offset != 0 || info->slow_hflip)
+      do_flip_h(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                src_coef_arrays, dst_coef_arrays);
+    else
+      do_flip_h_no_crop(srcinfo, dstinfo, info->x_crop_offset,
+                        src_coef_arrays);
+    break;
+  case JXFORM_FLIP_V:
+    do_flip_v(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_TRANSPOSE:
+    do_transpose(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                 src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_TRANSVERSE:
+    do_transverse(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                  src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_90:
+    do_rot_90(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_180:
+    do_rot_180(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+               src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_270:
+    do_rot_270(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+               src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_WIPE:
+    if (info->crop_width_set == JCROP_REFLECT &&
+        info->y_crop_offset == 0 && info->drop_height ==
+        (JDIMENSION)jdiv_round_up
+          ((long)info->output_height, (long)info->iMCU_sample_height) &&
+        (info->x_crop_offset == 0 ||
+         info->x_crop_offset + info->drop_width ==
+         (JDIMENSION)jdiv_round_up
+           ((long)info->output_width, (long)info->iMCU_sample_width)))
+      do_reflect(srcinfo, dstinfo, info->x_crop_offset,
+                 src_coef_arrays, info->drop_width, info->drop_height);
+    else if (info->crop_width_set == JCROP_FORCE)
+      do_flatten(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                 src_coef_arrays, info->drop_width, info->drop_height);
+    else
+      do_wipe(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, info->drop_width, info->drop_height);
+    break;
+  case JXFORM_DROP:
+    if (info->drop_width != 0 && info->drop_height != 0)
+      do_drop(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, info->drop_ptr, info->drop_coef_arrays,
+              info->drop_width, info->drop_height);
+    break;
+  }
+}
+
+/* jtransform_perfect_transform
+ *
+ * Determine whether lossless transformation is perfectly
+ * possible for a specified image and transformation.
+ *
+ * Inputs:
+ *   image_width, image_height: source image dimensions.
+ *   MCU_width, MCU_height: pixel dimensions of MCU.
+ *   transform: transformation identifier.
+ * Parameter sources from initialized jpeg_struct
+ * (after reading source header):
+ *   image_width = cinfo.image_width
+ *   image_height = cinfo.image_height
+ *   MCU_width = cinfo.max_h_samp_factor * cinfo.block_size
+ *   MCU_height = cinfo.max_v_samp_factor * cinfo.block_size
+ * Result:
+ *   TRUE = perfect transformation possible
+ *   FALSE = perfect transformation not possible
+ *           (may use custom action then)
+ */
+
+GLOBAL(boolean)
+jtransform_perfect_transform(JDIMENSION image_width, JDIMENSION image_height,
+                             int MCU_width, int MCU_height,
+                             JXFORM_CODE transform)
+{
+  boolean result = TRUE; /* initialize TRUE */
+
+  switch (transform) {
+  case JXFORM_FLIP_H:
+  case JXFORM_ROT_270:
+    if (image_width % (JDIMENSION)MCU_width)
+      result = FALSE;
+    break;
+  case JXFORM_FLIP_V:
+  case JXFORM_ROT_90:
+    if (image_height % (JDIMENSION)MCU_height)
+      result = FALSE;
+    break;
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_180:
+    if (image_width % (JDIMENSION)MCU_width)
+      result = FALSE;
+    if (image_height % (JDIMENSION)MCU_height)
+      result = FALSE;
+    break;
+  default:
+    break;
+  }
+
+  return result;
+}
+
+#endif /* TRANSFORMS_SUPPORTED */
+
+
+/* Setup decompression object to save desired markers in memory.
+ * This must be called before jpeg_read_header() to have the desired effect.
+ */
+
+GLOBAL(void)
+jcopy_markers_setup(j_decompress_ptr srcinfo, JCOPY_OPTION option)
+{
+#ifdef SAVE_MARKERS_SUPPORTED
+  int m;
+
+  /* Save comments except under NONE option */
+  if (option != JCOPYOPT_NONE && option != JCOPYOPT_ICC) {
+    jpeg_save_markers(srcinfo, JPEG_COM, 0xFFFF);
+  }
+  /* Save all types of APPn markers iff ALL option */
+  if (option == JCOPYOPT_ALL || option == JCOPYOPT_ALL_EXCEPT_ICC) {
+    for (m = 0; m < 16; m++) {
+      if (option == JCOPYOPT_ALL_EXCEPT_ICC && m == 2)
+        continue;
+      jpeg_save_markers(srcinfo, JPEG_APP0 + m, 0xFFFF);
+    }
+  }
+  /* Save only APP2 markers if ICC option selected */
+  if (option == JCOPYOPT_ICC) {
+    jpeg_save_markers(srcinfo, JPEG_APP0 + 2, 0xFFFF);
+  }
+#endif /* SAVE_MARKERS_SUPPORTED */
+}
+
+/* Copy markers saved in the given source object to the destination object.
+ * This should be called just after jpeg_start_compress() or
+ * jpeg_write_coefficients().
+ * Note that those routines will have written the SOI, and also the
+ * JFIF APP0 or Adobe APP14 markers if selected.
+ */
+
+GLOBAL(void)
+jcopy_markers_execute(j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                      JCOPY_OPTION option)
+{
+  jpeg_saved_marker_ptr marker;
+
+  /* In the current implementation, we don't actually need to examine the
+   * option flag here; we just copy everything that got saved.
+   * But to avoid confusion, we do not output JFIF and Adobe APP14 markers
+   * if the encoder library already wrote one.
+   */
+  for (marker = srcinfo->marker_list; marker != NULL; marker = marker->next) {
+    if (dstinfo->write_JFIF_header &&
+        marker->marker == JPEG_APP0 &&
+        marker->data_length >= 5 &&
+        marker->data0 == 0x4A &&
+        marker->data1 == 0x46 &&
+        marker->data2 == 0x49 &&
+        marker->data3 == 0x46 &&
+        marker->data4 == 0)
+      continue;                 /* reject duplicate JFIF */
+    if (dstinfo->write_Adobe_marker &&
+        marker->marker == JPEG_APP0 + 14 &&
+        marker->data_length >= 5 &&
+        marker->data0 == 0x41 &&
+        marker->data1 == 0x64 &&
+        marker->data2 == 0x6F &&
+        marker->data3 == 0x62 &&
+        marker->data4 == 0x65)
+      continue;                 /* reject duplicate Adobe */
+    jpeg_write_marker(dstinfo, marker->marker,
+                      marker->data, marker->data_length);
+  }
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/transupp.h Added

@@ -0,0 +1,231 @@
+/*
+ * transupp.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1997-2019, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2017, 2021, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains declarations for image transformation routines and
+ * other utility code used by the jpegtran sample application.  These are
+ * NOT part of the core JPEG library.  But we keep these routines separate
+ * from jpegtran.c to ease the task of maintaining jpegtran-like programs
+ * that have other user interfaces.
+ *
+ * NOTE: all the routines declared here have very specific requirements
+ * about when they are to be executed during the reading and writing of the
+ * source and destination files.  See the comments in transupp.c, or see
+ * jpegtran.c for an example of correct usage.
+ */
+
+/* If you happen not to want the image transform support, disable it here */
+#ifndef TRANSFORMS_SUPPORTED
+#define TRANSFORMS_SUPPORTED  1         /* 0 disables transform code */
+#endif
+
+/*
+ * Although rotating and flipping data expressed as DCT coefficients is not
+ * hard, there is an asymmetry in the JPEG format specification for images
+ * whose dimensions aren't multiples of the iMCU size.  The right and bottom
+ * image edges are padded out to the next iMCU boundary with junk data; but
+ * no padding is possible at the top and left edges.  If we were to flip
+ * the whole image including the pad data, then pad garbage would become
+ * visible at the top and/or left, and real pixels would disappear into the
+ * pad margins --- perhaps permanently, since encoders & decoders may not
+ * bother to preserve DCT blocks that appear to be completely outside the
+ * nominal image area.  So, we have to exclude any partial iMCUs from the
+ * basic transformation.
+ *
+ * Transpose is the only transformation that can handle partial iMCUs at the
+ * right and bottom edges completely cleanly.  flip_h can flip partial iMCUs
+ * at the bottom, but leaves any partial iMCUs at the right edge untouched.
+ * Similarly flip_v leaves any partial iMCUs at the bottom edge untouched.
+ * The other transforms are defined as combinations of these basic transforms
+ * and process edge blocks in a way that preserves the equivalence.
+ *
+ * The "trim" option causes untransformable partial iMCUs to be dropped;
+ * this is not strictly lossless, but it usually gives the best-looking
+ * result for odd-size images.  Note that when this option is active,
+ * the expected mathematical equivalences between the transforms may not hold.
+ * (For example, -rot 270 -trim trims only the bottom edge, but -rot 90 -trim
+ * followed by -rot 180 -trim trims both edges.)
+ *
+ * We also offer a lossless-crop option, which discards data outside a given
+ * image region but losslessly preserves what is inside.  Like the rotate and
+ * flip transforms, lossless crop is restricted by the JPEG format: the upper
+ * left corner of the selected region must fall on an iMCU boundary.  If this
+ * does not hold for the given crop parameters, we silently move the upper left
+ * corner up and/or left to make it so, simultaneously increasing the region
+ * dimensions to keep the lower right crop corner unchanged.  (Thus, the
+ * output image covers at least the requested region, but may cover more.)
+ * The adjustment of the region dimensions may be optionally disabled.
+ *
+ * A complementary lossless wipe option is provided to discard (gray out) data
+ * inside a given image region while losslessly preserving what is outside.
+ * A lossless drop option is also provided, which allows another JPEG image to
+ * be inserted ("dropped") into the source image data at a given position,
+ * replacing the existing image data at that position.  Both the source image
+ * and the drop image must have the same subsampling level.  It is best if they
+ * also have the same quantization (quality.)  Otherwise, the quantization of
+ * the output image will be adapted to accommodate the higher of the source
+ * image quality and the drop image quality.  The trim option can be used with
+ * the drop option to requantize the drop image to match the source image.
+ *
+ * We also provide a lossless-resize option, which is kind of a lossless-crop
+ * operation in the DCT coefficient block domain - it discards higher-order
+ * coefficients and losslessly preserves lower-order coefficients of a
+ * sub-block.
+ *
+ * Rotate/flip transform, resize, and crop can be requested together in a
+ * single invocation.  The crop is applied last --- that is, the crop region
+ * is specified in terms of the destination image after transform/resize.
+ *
+ * We also offer a "force to grayscale" option, which simply discards the
+ * chrominance channels of a YCbCr image.  This is lossless in the sense that
+ * the luminance channel is preserved exactly.  It's not the same kind of
+ * thing as the rotate/flip transformations, but it's convenient to handle it
+ * as part of this package, mainly because the transformation routines have to
+ * be aware of the option to know how many components to work on.
+ */
+
+
+/*
+ * Codes for supported types of image transformations.
+ */
+
+typedef enum {
+  JXFORM_NONE,            /* no transformation */
+  JXFORM_FLIP_H,          /* horizontal flip */
+  JXFORM_FLIP_V,          /* vertical flip */
+  JXFORM_TRANSPOSE,       /* transpose across UL-to-LR axis */
+  JXFORM_TRANSVERSE,      /* transpose across UR-to-LL axis */
+  JXFORM_ROT_90,          /* 90-degree clockwise rotation */
+  JXFORM_ROT_180,         /* 180-degree rotation */
+  JXFORM_ROT_270,         /* 270-degree clockwise (or 90 ccw) */
+  JXFORM_WIPE,            /* wipe */
+  JXFORM_DROP             /* drop */
+} JXFORM_CODE;
+
+/*
+ * Codes for crop parameters, which can individually be unspecified,
+ * positive or negative for xoffset or yoffset,
+ * positive or force or reflect for width or height.
+ */
+
+typedef enum {
+  JCROP_UNSET,
+  JCROP_POS,
+  JCROP_NEG,
+  JCROP_FORCE,
+  JCROP_REFLECT
+} JCROP_CODE;
+
+/*
+ * Transform parameters struct.
+ * NB: application must not change any elements of this struct after
+ * calling jtransform_request_workspace.
+ */
+
+typedef struct {
+  /* Options: set by caller */
+  JXFORM_CODE transform;        /* image transform operator */
+  boolean perfect;              /* if TRUE, fail if partial MCUs are requested */
+  boolean trim;                 /* if TRUE, trim partial MCUs as needed */
+  boolean force_grayscale;      /* if TRUE, convert color image to grayscale */
+  boolean crop;                 /* if TRUE, crop or wipe source image, or drop */
+  boolean slow_hflip;  /* For best performance, the JXFORM_FLIP_H transform
+                          normally modifies the source coefficients in place.
+                          Setting this to TRUE will instead use a slower,
+                          double-buffered algorithm, which leaves the source
+                          coefficients in tact (necessary if other transformed
+                          images must be generated from the same set of
+                          coefficients. */
+
+  /* Crop parameters: application need not set these unless crop is TRUE.
+   * These can be filled in by jtransform_parse_crop_spec().
+   */
+  JDIMENSION crop_width;        /* Width of selected region */
+  JCROP_CODE crop_width_set;    /* (force-disables adjustment) */
+  JDIMENSION crop_height;       /* Height of selected region */
+  JCROP_CODE crop_height_set;   /* (force-disables adjustment) */
+  JDIMENSION crop_xoffset;      /* X offset of selected region */
+  JCROP_CODE crop_xoffset_set;  /* (negative measures from right edge) */
+  JDIMENSION crop_yoffset;      /* Y offset of selected region */
+  JCROP_CODE crop_yoffset_set;  /* (negative measures from bottom edge) */
+
+  /* Drop parameters: set by caller for drop request */
+  j_decompress_ptr drop_ptr;
+  jvirt_barray_ptr *drop_coef_arrays;
+
+  /* Internal workspace: caller should not touch these */
+  int num_components;           /* # of components in workspace */
+  jvirt_barray_ptr *workspace_coef_arrays; /* workspace for transformations */
+  JDIMENSION output_width;      /* cropped destination dimensions */
+  JDIMENSION output_height;
+  JDIMENSION x_crop_offset;     /* destination crop offsets measured in iMCUs */
+  JDIMENSION y_crop_offset;
+  JDIMENSION drop_width;        /* drop/wipe dimensions measured in iMCUs */
+  JDIMENSION drop_height;
+  int iMCU_sample_width;        /* destination iMCU size */
+  int iMCU_sample_height;
+} jpeg_transform_info;
+
+
+#if TRANSFORMS_SUPPORTED
+
+/* Parse a crop specification (written in X11 geometry style) */
+EXTERN(boolean) jtransform_parse_crop_spec(jpeg_transform_info *info,
+                                           const char *spec);
+/* Request any required workspace */
+EXTERN(boolean) jtransform_request_workspace(j_decompress_ptr srcinfo,
+                                             jpeg_transform_info *info);
+/* Adjust output image parameters */
+EXTERN(jvirt_barray_ptr *) jtransform_adjust_parameters
+  (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+   jvirt_barray_ptr *src_coef_arrays, jpeg_transform_info *info);
+/* Execute the actual transformation, if any */
+EXTERN(void) jtransform_execute_transform(j_decompress_ptr srcinfo,
+                                          j_compress_ptr dstinfo,
+                                          jvirt_barray_ptr *src_coef_arrays,
+                                          jpeg_transform_info *info);
+/* Determine whether lossless transformation is perfectly
+ * possible for a specified image and transformation.
+ */
+EXTERN(boolean) jtransform_perfect_transform(JDIMENSION image_width,
+                                             JDIMENSION image_height,
+                                             int MCU_width, int MCU_height,
+                                             JXFORM_CODE transform);
+
+/* jtransform_execute_transform used to be called
+ * jtransform_execute_transformation, but some compilers complain about
+ * routine names that long.  This macro is here to avoid breaking any
+ * old source code that uses the original name...
+ */
+#define jtransform_execute_transformation       jtransform_execute_transform
+
+#endif /* TRANSFORMS_SUPPORTED */
+
+
+/*
+ * Support for copying optional markers from source to destination file.
+ */
+
+typedef enum {
+  JCOPYOPT_NONE,           /* copy no optional markers */
+  JCOPYOPT_COMMENTS,       /* copy only comment (COM) markers */
+  JCOPYOPT_ALL,            /* copy all optional markers */
+  JCOPYOPT_ALL_EXCEPT_ICC, /* copy all optional markers except APP2 */
+  JCOPYOPT_ICC             /* copy only ICC profile (APP2) markers */
+} JCOPY_OPTION;
+
+#define JCOPYOPT_DEFAULT  JCOPYOPT_COMMENTS     /* recommended default */
+
+/* Setup decompression object to save desired markers in memory */
+EXTERN(void) jcopy_markers_setup(j_decompress_ptr srcinfo,
+                                 JCOPY_OPTION option);
+/* Copy markers saved in the given source object to the destination object */
+EXTERN(void) jcopy_markers_execute(j_decompress_ptr srcinfo,
+                                   j_compress_ptr dstinfo,
+                                   JCOPY_OPTION option);

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/turbojpeg-jni.c Added

@@ -0,0 +1,1228 @@
+/*
+ * Copyright (C)2011-2020 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "turbojpeg.h"
+#ifdef WIN32
+#include "tjutil.h"
+#endif
+#include <jni.h>
+#include "java/org_libjpegturbo_turbojpeg_TJCompressor.h"
+#include "java/org_libjpegturbo_turbojpeg_TJDecompressor.h"
+#include "java/org_libjpegturbo_turbojpeg_TJTransformer.h"
+#include "java/org_libjpegturbo_turbojpeg_TJ.h"
+
+#define BAILIF0(f) { \
+  if (!(f) || (*env)->ExceptionCheck(env)) { \
+    goto bailout; \
+  } \
+}
+
+#define THROW(msg, exceptionClass) { \
+  jclass _exccls = (*env)->FindClass(env, exceptionClass); \
+  \
+  BAILIF0(_exccls); \
+  (*env)->ThrowNew(env, _exccls, msg); \
+  goto bailout; \
+}
+
+#define THROW_TJ() { \
+  jclass _exccls; \
+  jmethodID _excid; \
+  jobject _excobj; \
+  jstring _errstr; \
+  \
+  BAILIF0(_errstr = (*env)->NewStringUTF(env, tjGetErrorStr2(handle))); \
+  BAILIF0(_exccls = (*env)->FindClass(env, \
+    "org/libjpegturbo/turbojpeg/TJException")); \
+  BAILIF0(_excid = (*env)->GetMethodID(env, _exccls, "<init>", \
+                                       "(Ljava/lang/String;I)V")); \
+  BAILIF0(_excobj = (*env)->NewObject(env, _exccls, _excid, _errstr, \
+                                      tjGetErrorCode(handle))); \
+  (*env)->Throw(env, _excobj); \
+  goto bailout; \
+}
+
+#define THROW_ARG(msg)  THROW(msg, "java/lang/IllegalArgumentException")
+
+#define THROW_MEM() \
+  THROW("Memory allocation failure", "java/lang/OutOfMemoryError");
+
+#define GET_HANDLE() \
+  jclass _cls = (*env)->GetObjectClass(env, obj); \
+  jfieldID _fid; \
+  \
+  BAILIF0(_cls); \
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "handle", "J")); \
+  handle = (tjhandle)(size_t)(*env)->GetLongField(env, obj, _fid);
+
+#ifdef _WIN32
+#define setenv(envvar, value, dummy)  _putenv_s(envvar, value)
+#endif
+
+#define PROP2ENV(property, envvar) { \
+  if ((jName = (*env)->NewStringUTF(env, property)) != NULL && \
+      (jValue = (*env)->CallStaticObjectMethod(env, cls, mid, \
+                                               jName)) != NULL) { \
+    if ((value = (*env)->GetStringUTFChars(env, jValue, 0)) != NULL) { \
+      setenv(envvar, value, 1); \
+      (*env)->ReleaseStringUTFChars(env, jValue, value); \
+    } \
+  } \
+}
+
+#define SAFE_RELEASE(javaArray, cArray) { \
+  if (javaArray && cArray) \
+    (*env)->ReleasePrimitiveArrayCritical(env, javaArray, (void *)cArray, 0); \
+  cArray = NULL; \
+}
+
+static int ProcessSystemProperties(JNIEnv *env)
+{
+  jclass cls;
+  jmethodID mid;
+  jstring jName, jValue;
+  const char *value;
+
+  BAILIF0(cls = (*env)->FindClass(env, "java/lang/System"));
+  BAILIF0(mid = (*env)->GetStaticMethodID(env, cls, "getProperty",
+    "(Ljava/lang/String;)Ljava/lang/String;"));
+
+  PROP2ENV("turbojpeg.optimize", "TJ_OPTIMIZE");
+  PROP2ENV("turbojpeg.arithmetic", "TJ_ARITHMETIC");
+  PROP2ENV("turbojpeg.restart", "TJ_RESTART");
+  PROP2ENV("turbojpeg.progressive", "TJ_PROGRESSIVE");
+  return 0;
+
+bailout:
+  return -1;
+}
+
+/* TurboJPEG 1.2.x: TJ::bufSize() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSize
+  (JNIEnv *env, jclass cls, jint width, jint height, jint jpegSubsamp)
+{
+  jint retval = (jint)tjBufSize(width, height, jpegSubsamp);
+
+  if (retval == -1) THROW_ARG(tjGetErrorStr());
+
+bailout:
+  return retval;
+}
+
+/* TurboJPEG 1.4.x: TJ::bufSizeYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII
+  (JNIEnv *env, jclass cls, jint width, jint pad, jint height, jint subsamp)
+{
+  jint retval = (jint)tjBufSizeYUV2(width, pad, height, subsamp);
+
+  if (retval == -1) THROW_ARG(tjGetErrorStr());
+
+bailout:
+  return retval;
+}
+
+/* TurboJPEG 1.2.x: TJ::bufSizeYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__III
+  (JNIEnv *env, jclass cls, jint width, jint height, jint subsamp)
+{
+  return Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII(env, cls, width,
+                                                             4, height,
+                                                             subsamp);
+}
+
+/* TurboJPEG 1.4.x: TJ::planeSizeYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeSizeYUV__IIIII
+  (JNIEnv *env, jclass cls, jint componentID, jint width, jint stride,
+   jint height, jint subsamp)
+{
+  jint retval = (jint)tjPlaneSizeYUV(componentID, width, stride, height,
+                                     subsamp);
+
+  if (retval == -1) THROW_ARG(tjGetErrorStr());
+
+bailout:
+  return retval;
+}
+
+/* TurboJPEG 1.4.x: TJ::planeWidth() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeWidth__III
+  (JNIEnv *env, jclass cls, jint componentID, jint width, jint subsamp)
+{
+  jint retval = (jint)tjPlaneWidth(componentID, width, subsamp);
+
+  if (retval == -1) THROW_ARG(tjGetErrorStr());
+
+bailout:
+  return retval;
+}
+
+/* TurboJPEG 1.4.x: TJ::planeHeight() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeHeight__III
+  (JNIEnv *env, jclass cls, jint componentID, jint height, jint subsamp)
+{
+  jint retval = (jint)tjPlaneHeight(componentID, height, subsamp);
+
+  if (retval == -1) THROW_ARG(tjGetErrorStr());
+
+bailout:
+  return retval;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::init() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_init
+  (JNIEnv *env, jobject obj)
+{
+  jclass cls;
+  jfieldID fid;
+  tjhandle handle;
+
+  if ((handle = tjInitCompress()) == NULL)
+    THROW(tjGetErrorStr(), "org/libjpegturbo/turbojpeg/TJException");
+
+  BAILIF0(cls = (*env)->GetObjectClass(env, obj));
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "handle", "J"));
+  (*env)->SetLongField(env, obj, fid, (size_t)handle);
+
+bailout:
+  return;
+}
+
+static jint TJCompressor_compress
+  (JNIEnv *env, jobject obj, jarray src, jint srcElementSize, jint x, jint y,
+   jint width, jint pitch, jint height, jint pf, jbyteArray dst,
+   jint jpegSubsamp, jint jpegQual, jint flags)
+{
+  tjhandle handle = 0;
+  unsigned long jpegSize = 0;
+  jsize arraySize = 0, actualPitch;
+  unsigned char *srcBuf = NULL, *jpegBuf = NULL;
+
+  GET_HANDLE();
+
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF || width < 1 ||
+      height < 1 || pitch < 0)
+    THROW_ARG("Invalid argument in compress()");
+  if (org_libjpegturbo_turbojpeg_TJ_NUMPF != TJ_NUMPF)
+    THROW_ARG("Mismatch between Java and C API");
+
+  actualPitch = (pitch == 0) ? width * tjPixelSizepf : pitch;
+  arraySize = (y + height - 1) * actualPitch + (x + width) * tjPixelSizepf;
+  if ((*env)->GetArrayLength(env, src) * srcElementSize < arraySize)
+    THROW_ARG("Source buffer is not large enough");
+  jpegSize = tjBufSize(width, height, jpegSubsamp);
+  if ((*env)->GetArrayLength(env, dst) < (jsize)jpegSize)
+    THROW_ARG("Destination buffer is not large enough");
+
+  if (ProcessSystemProperties(env) < 0) goto bailout;
+
+  BAILIF0(srcBuf = (*env)->GetPrimitiveArrayCritical(env, src, 0));
+  BAILIF0(jpegBuf = (*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+  if (tjCompress2(handle, &srcBufy * actualPitch + x * tjPixelSizepf,
+                  width, pitch, height, pf, &jpegBuf, &jpegSize, jpegSubsamp,
+                  jpegQual, flags | TJFLAG_NOREALLOC) == -1) {
+    SAFE_RELEASE(dst, jpegBuf);
+    SAFE_RELEASE(src, srcBuf);
+    THROW_TJ();
+  }
+
+bailout:
+  SAFE_RELEASE(dst, jpegBuf);
+  SAFE_RELEASE(src, srcBuf);
+  return (jint)jpegSize;
+}
+
+/* TurboJPEG 1.3.x: TJCompressor::compress() byte source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIIIII_3BIII
+  (JNIEnv *env, jobject obj, jbyteArray src, jint x, jint y, jint width,
+   jint pitch, jint height, jint pf, jbyteArray dst, jint jpegSubsamp,
+   jint jpegQual, jint flags)
+{
+  return TJCompressor_compress(env, obj, src, 1, x, y, width, pitch, height,
+                               pf, dst, jpegSubsamp, jpegQual, flags);
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::compress() byte source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIII_3BIII
+  (JNIEnv *env, jobject obj, jbyteArray src, jint width, jint pitch,
+   jint height, jint pf, jbyteArray dst, jint jpegSubsamp, jint jpegQual,
+   jint flags)
+{
+  return TJCompressor_compress(env, obj, src, 1, 0, 0, width, pitch, height,
+                               pf, dst, jpegSubsamp, jpegQual, flags);
+}
+
+/* TurboJPEG 1.3.x: TJCompressor::compress() int source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIIIII_3BIII
+  (JNIEnv *env, jobject obj, jintArray src, jint x, jint y, jint width,
+   jint stride, jint height, jint pf, jbyteArray dst, jint jpegSubsamp,
+   jint jpegQual, jint flags)
+{
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF)
+    THROW_ARG("Invalid argument in compress()");
+  if (tjPixelSizepf != sizeof(jint))
+    THROW_ARG("Pixel format must be 32-bit when compressing from an integer buffer.");
+
+  return TJCompressor_compress(env, obj, src, sizeof(jint), x, y, width,
+                               stride * sizeof(jint), height, pf, dst,
+                               jpegSubsamp, jpegQual, flags);
+
+bailout:
+  return 0;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::compress() int source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIII_3BIII
+  (JNIEnv *env, jobject obj, jintArray src, jint width, jint stride,
+   jint height, jint pf, jbyteArray dst, jint jpegSubsamp, jint jpegQual,
+   jint flags)
+{
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF)
+    THROW_ARG("Invalid argument in compress()");
+  if (tjPixelSizepf != sizeof(jint))
+    THROW_ARG("Pixel format must be 32-bit when compressing from an integer buffer.");
+
+  return TJCompressor_compress(env, obj, src, sizeof(jint), 0, 0, width,
+                               stride * sizeof(jint), height, pf, dst,
+                               jpegSubsamp, jpegQual, flags);
+
+bailout:
+  return 0;
+}
+
+/* TurboJPEG 1.4.x: TJCompressor::compressFromYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compressFromYUV___3_3B_3II_3III_3BII
+  (JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+   jint width, jintArray jSrcStrides, jint height, jint subsamp,
+   jbyteArray dst, jint jpegQual, jint flags)
+{
+  tjhandle handle = 0;
+  unsigned long jpegSize = 0;
+  jbyteArray jSrcPlanes3 = { NULL, NULL, NULL };
+  const unsigned char *srcPlanesTmp3 = { NULL, NULL, NULL };
+  const unsigned char *srcPlanes3 = { NULL, NULL, NULL };
+  int *srcOffsetsTmp = NULL, srcOffsets3 = { 0, 0, 0 };
+  int *srcStridesTmp = NULL, srcStrides3 = { 0, 0, 0 };
+  unsigned char *jpegBuf = NULL;
+  int nc = (subsamp == org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY ? 1 : 3), i;
+
+  GET_HANDLE();
+
+  if (subsamp < 0 || subsamp >= org_libjpegturbo_turbojpeg_TJ_NUMSAMP)
+    THROW_ARG("Invalid argument in compressFromYUV()");
+  if (org_libjpegturbo_turbojpeg_TJ_NUMSAMP != TJ_NUMSAMP)
+    THROW_ARG("Mismatch between Java and C API");
+
+  if ((*env)->GetArrayLength(env, srcobjs) < nc)
+    THROW_ARG("Planes array is too small for the subsampling type");
+  if ((*env)->GetArrayLength(env, jSrcOffsets) < nc)
+    THROW_ARG("Offsets array is too small for the subsampling type");
+  if ((*env)->GetArrayLength(env, jSrcStrides) < nc)
+    THROW_ARG("Strides array is too small for the subsampling type");
+
+  jpegSize = tjBufSize(width, height, subsamp);
+  if ((*env)->GetArrayLength(env, dst) < (jsize)jpegSize)
+    THROW_ARG("Destination buffer is not large enough");
+
+  if (ProcessSystemProperties(env) < 0) goto bailout;
+
+  BAILIF0(srcOffsetsTmp =
+          (*env)->GetPrimitiveArrayCritical(env, jSrcOffsets, 0));
+  for (i = 0; i < nc; i++) srcOffsetsi = srcOffsetsTmpi;
+  SAFE_RELEASE(jSrcOffsets, srcOffsetsTmp);
+
+  BAILIF0(srcStridesTmp =
+          (*env)->GetPrimitiveArrayCritical(env, jSrcStrides, 0));
+  for (i = 0; i < nc; i++) srcStridesi = srcStridesTmpi;
+  SAFE_RELEASE(jSrcStrides, srcStridesTmp);
+
+  for (i = 0; i < nc; i++) {
+    int planeSize = tjPlaneSizeYUV(i, width, srcStridesi, height, subsamp);
+    int pw = tjPlaneWidth(i, width, subsamp);
+
+    if (planeSize < 0 || pw < 0)
+      THROW_ARG(tjGetErrorStr());
+
+    if (srcOffsetsi < 0)
+      THROW_ARG("Invalid argument in compressFromYUV()");
+    if (srcStridesi < 0 && srcOffsetsi - planeSize + pw < 0)
+      THROW_ARG("Negative plane stride would cause memory to be accessed below plane boundary");
+
+    BAILIF0(jSrcPlanesi = (*env)->GetObjectArrayElement(env, srcobjs, i));
+    if ((*env)->GetArrayLength(env, jSrcPlanesi) <
+        srcOffsetsi + planeSize)
+      THROW_ARG("Source plane is not large enough");
+
+    BAILIF0(srcPlanesTmpi =
+            (*env)->GetPrimitiveArrayCritical(env, jSrcPlanesi, 0));
+    srcPlanesi = &srcPlanesTmpisrcOffsetsi;
+    SAFE_RELEASE(jSrcPlanesi, srcPlanesTmpi);
+  }
+  BAILIF0(jpegBuf = (*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+  if (tjCompressFromYUVPlanes(handle, srcPlanes, width, srcStrides, height,
+                              subsamp, &jpegBuf, &jpegSize, jpegQual,
+                              flags | TJFLAG_NOREALLOC) == -1) {
+    SAFE_RELEASE(dst, jpegBuf);
+    THROW_TJ();
+  }
+
+bailout:
+  SAFE_RELEASE(dst, jpegBuf);
+  return (jint)jpegSize;
+}
+
+static void TJCompressor_encodeYUV
+  (JNIEnv *env, jobject obj, jarray src, jint srcElementSize, jint x, jint y,
+   jint width, jint pitch, jint height, jint pf, jobjectArray dstobjs,
+   jintArray jDstOffsets, jintArray jDstStrides, jint subsamp, jint flags)
+{
+  tjhandle handle = 0;
+  jsize arraySize = 0, actualPitch;
+  unsigned char *srcBuf = NULL;
+  jbyteArray jDstPlanes3 = { NULL, NULL, NULL };
+  unsigned char *dstPlanesTmp3 = { NULL, NULL, NULL };
+  unsigned char *dstPlanes3 = { NULL, NULL, NULL };
+  int *dstOffsetsTmp = NULL, dstOffsets3 = { 0, 0, 0 };
+  int *dstStridesTmp = NULL, dstStrides3 = { 0, 0, 0 };
+  int nc = (subsamp == org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY ? 1 : 3), i;
+
+  GET_HANDLE();
+
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF || width < 1 ||
+      height < 1 || pitch < 0 || subsamp < 0 ||
+      subsamp >= org_libjpegturbo_turbojpeg_TJ_NUMSAMP)
+    THROW_ARG("Invalid argument in encodeYUV()");
+  if (org_libjpegturbo_turbojpeg_TJ_NUMPF != TJ_NUMPF ||
+      org_libjpegturbo_turbojpeg_TJ_NUMSAMP != TJ_NUMSAMP)
+    THROW_ARG("Mismatch between Java and C API");
+
+  if ((*env)->GetArrayLength(env, dstobjs) < nc)
+    THROW_ARG("Planes array is too small for the subsampling type");
+  if ((*env)->GetArrayLength(env, jDstOffsets) < nc)
+    THROW_ARG("Offsets array is too small for the subsampling type");
+  if ((*env)->GetArrayLength(env, jDstStrides) < nc)
+    THROW_ARG("Strides array is too small for the subsampling type");
+
+  actualPitch = (pitch == 0) ? width * tjPixelSizepf : pitch;
+  arraySize = (y + height - 1) * actualPitch + (x + width) * tjPixelSizepf;
+  if ((*env)->GetArrayLength(env, src) * srcElementSize < arraySize)
+    THROW_ARG("Source buffer is not large enough");
+
+  BAILIF0(dstOffsetsTmp =
+          (*env)->GetPrimitiveArrayCritical(env, jDstOffsets, 0));
+  for (i = 0; i < nc; i++) dstOffsetsi = dstOffsetsTmpi;
+  SAFE_RELEASE(jDstOffsets, dstOffsetsTmp);
+
+  BAILIF0(dstStridesTmp =
+          (*env)->GetPrimitiveArrayCritical(env, jDstStrides, 0));
+  for (i = 0; i < nc; i++) dstStridesi = dstStridesTmpi;
+  SAFE_RELEASE(jDstStrides, dstStridesTmp);
+
+  for (i = 0; i < nc; i++) {
+    int planeSize = tjPlaneSizeYUV(i, width, dstStridesi, height, subsamp);
+    int pw = tjPlaneWidth(i, width, subsamp);
+
+    if (planeSize < 0 || pw < 0)
+      THROW_ARG(tjGetErrorStr());
+
+    if (dstOffsetsi < 0)
+      THROW_ARG("Invalid argument in encodeYUV()");
+    if (dstStridesi < 0 && dstOffsetsi - planeSize + pw < 0)
+      THROW_ARG("Negative plane stride would cause memory to be accessed below plane boundary");
+
+    BAILIF0(jDstPlanesi = (*env)->GetObjectArrayElement(env, dstobjs, i));
+    if ((*env)->GetArrayLength(env, jDstPlanesi) <
+        dstOffsetsi + planeSize)
+      THROW_ARG("Destination plane is not large enough");
+
+    BAILIF0(dstPlanesTmpi =
+            (*env)->GetPrimitiveArrayCritical(env, jDstPlanesi, 0));
+    dstPlanesi = &dstPlanesTmpidstOffsetsi;
+    SAFE_RELEASE(jDstPlanesi, dstPlanesTmpi);
+  }
+  BAILIF0(srcBuf = (*env)->GetPrimitiveArrayCritical(env, src, 0));
+
+  if (tjEncodeYUVPlanes(handle, &srcBufy * actualPitch + x * tjPixelSizepf,
+                        width, pitch, height, pf, dstPlanes, dstStrides,
+                        subsamp, flags) == -1) {
+    SAFE_RELEASE(src, srcBuf);
+    THROW_TJ();
+  }
+
+bailout:
+  SAFE_RELEASE(src, srcBuf);
+}
+
+/* TurboJPEG 1.4.x: TJCompressor::encodeYUV() byte source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIIIII_3_3B_3I_3III
+  (JNIEnv *env, jobject obj, jbyteArray src, jint x, jint y, jint width,
+   jint pitch, jint height, jint pf, jobjectArray dstobjs,
+   jintArray jDstOffsets, jintArray jDstStrides, jint subsamp, jint flags)
+{
+  TJCompressor_encodeYUV(env, obj, src, 1, x, y, width, pitch, height, pf,
+                         dstobjs, jDstOffsets, jDstStrides, subsamp, flags);
+}
+
+/* TurboJPEG 1.4.x: TJCompressor::encodeYUV() int source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIIIII_3_3B_3I_3III
+  (JNIEnv *env, jobject obj, jintArray src, jint x, jint y, jint width,
+   jint stride, jint height, jint pf, jobjectArray dstobjs,
+   jintArray jDstOffsets, jintArray jDstStrides, jint subsamp, jint flags)
+{
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF)
+    THROW_ARG("Invalid argument in encodeYUV()");
+  if (tjPixelSizepf != sizeof(jint))
+    THROW_ARG("Pixel format must be 32-bit when encoding from an integer buffer.");
+
+  TJCompressor_encodeYUV(env, obj, src, sizeof(jint), x, y, width,
+                         stride * sizeof(jint), height, pf, dstobjs,
+                         jDstOffsets, jDstStrides, subsamp, flags);
+
+bailout:
+  return;
+}
+
+static void JNICALL TJCompressor_encodeYUV_12
+  (JNIEnv *env, jobject obj, jarray src, jint srcElementSize, jint width,
+   jint pitch, jint height, jint pf, jbyteArray dst, jint subsamp, jint flags)
+{
+  tjhandle handle = 0;
+  jsize arraySize = 0;
+  unsigned char *srcBuf = NULL, *dstBuf = NULL;
+
+  GET_HANDLE();
+
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF || width < 1 ||
+      height < 1 || pitch < 0)
+    THROW_ARG("Invalid argument in encodeYUV()");
+  if (org_libjpegturbo_turbojpeg_TJ_NUMPF != TJ_NUMPF)
+    THROW_ARG("Mismatch between Java and C API");
+
+  arraySize = (pitch == 0) ? width * tjPixelSizepf * height : pitch * height;
+  if ((*env)->GetArrayLength(env, src) * srcElementSize < arraySize)
+    THROW_ARG("Source buffer is not large enough");
+  if ((*env)->GetArrayLength(env, dst) <
+      (jsize)tjBufSizeYUV(width, height, subsamp))
+    THROW_ARG("Destination buffer is not large enough");
+
+  BAILIF0(srcBuf = (*env)->GetPrimitiveArrayCritical(env, src, 0));
+  BAILIF0(dstBuf = (*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+  if (tjEncodeYUV2(handle, srcBuf, width, pitch, height, pf, dstBuf, subsamp,
+                   flags) == -1) {
+    SAFE_RELEASE(dst, dstBuf);
+    SAFE_RELEASE(src, srcBuf);
+    THROW_TJ();
+  }
+
+bailout:
+  SAFE_RELEASE(dst, dstBuf);
+  SAFE_RELEASE(src, srcBuf);
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::encodeYUV() byte source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIII_3BII
+  (JNIEnv *env, jobject obj, jbyteArray src, jint width, jint pitch,
+   jint height, jint pf, jbyteArray dst, jint subsamp, jint flags)
+{
+  TJCompressor_encodeYUV_12(env, obj, src, 1, width, pitch, height, pf, dst,
+                            subsamp, flags);
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::encodeYUV() int source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIII_3BII
+  (JNIEnv *env, jobject obj, jintArray src, jint width, jint stride,
+   jint height, jint pf, jbyteArray dst, jint subsamp, jint flags)
+{
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF)
+    THROW_ARG("Invalid argument in encodeYUV()");
+  if (tjPixelSizepf != sizeof(jint))
+    THROW_ARG("Pixel format must be 32-bit when encoding from an integer buffer.");
+
+  TJCompressor_encodeYUV_12(env, obj, src, sizeof(jint), width,
+                            stride * sizeof(jint), height, pf, dst, subsamp,
+                            flags);
+
+bailout:
+  return;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::destroy() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy
+  (JNIEnv *env, jobject obj)
+{
+  tjhandle handle = 0;
+
+  GET_HANDLE();
+
+  if (tjDestroy(handle) == -1) THROW_TJ();
+  (*env)->SetLongField(env, obj, _fid, 0);
+
+bailout:
+  return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::init() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_init
+  (JNIEnv *env, jobject obj)
+{
+  jclass cls;
+  jfieldID fid;
+  tjhandle handle;
+
+  if ((handle = tjInitDecompress()) == NULL)
+    THROW(tjGetErrorStr(), "org/libjpegturbo/turbojpeg/TJException");
+
+  BAILIF0(cls = (*env)->GetObjectClass(env, obj));
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "handle", "J"));
+  (*env)->SetLongField(env, obj, fid, (size_t)handle);
+
+bailout:
+  return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::getScalingFactors() */
+JNIEXPORT jobjectArray JNICALL Java_org_libjpegturbo_turbojpeg_TJ_getScalingFactors
+  (JNIEnv *env, jclass cls)
+{
+  jclass sfcls = NULL;
+  jfieldID fid = 0;
+  tjscalingfactor *sf = NULL;
+  int n = 0, i;
+  jobject sfobj = NULL;
+  jobjectArray sfjava = NULL;
+
+  if ((sf = tjGetScalingFactors(&n)) == NULL || n == 0)
+    THROW_ARG(tjGetErrorStr());
+
+  BAILIF0(sfcls = (*env)->FindClass(env,
+    "org/libjpegturbo/turbojpeg/TJScalingFactor"));
+  BAILIF0(sfjava = (jobjectArray)(*env)->NewObjectArray(env, n, sfcls, 0));
+
+  for (i = 0; i < n; i++) {
+    BAILIF0(sfobj = (*env)->AllocObject(env, sfcls));
+    BAILIF0(fid = (*env)->GetFieldID(env, sfcls, "num", "I"));
+    (*env)->SetIntField(env, sfobj, fid, sfi.num);
+    BAILIF0(fid = (*env)->GetFieldID(env, sfcls, "denom", "I"));
+    (*env)->SetIntField(env, sfobj, fid, sfi.denom);
+    (*env)->SetObjectArrayElement(env, sfjava, i, sfobj);
+  }
+
+bailout:
+  return sfjava;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompressHeader() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressHeader
+  (JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize)
+{
+  tjhandle handle = 0;
+  unsigned char *jpegBuf = NULL;
+  int width = 0, height = 0, jpegSubsamp = -1, jpegColorspace = -1;
+
+  GET_HANDLE();
+
+  if ((*env)->GetArrayLength(env, src) < jpegSize)
+    THROW_ARG("Source buffer is not large enough");
+
+  BAILIF0(jpegBuf = (*env)->GetPrimitiveArrayCritical(env, src, 0));
+
+  if (tjDecompressHeader3(handle, jpegBuf, (unsigned long)jpegSize, &width,
+                          &height, &jpegSubsamp, &jpegColorspace) == -1) {
+    SAFE_RELEASE(src, jpegBuf);
+    THROW_TJ();
+  }
+
+  SAFE_RELEASE(src, jpegBuf);
+
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+  (*env)->SetIntField(env, obj, _fid, jpegSubsamp);
+  if ((_fid = (*env)->GetFieldID(env, _cls, "jpegColorspace", "I")) == 0)
+    (*env)->ExceptionClear(env);
+  else
+    (*env)->SetIntField(env, obj, _fid, jpegColorspace);
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+  (*env)->SetIntField(env, obj, _fid, width);
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+  (*env)->SetIntField(env, obj, _fid, height);
+
+bailout:
+  SAFE_RELEASE(src, jpegBuf);
+}
+
+static void TJDecompressor_decompress
+  (JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jarray dst,
+   jint dstElementSize, jint x, jint y, jint width, jint pitch, jint height,
+   jint pf, jint flags)
+{
+  tjhandle handle = 0;
+  jsize arraySize = 0, actualPitch;
+  unsigned char *jpegBuf = NULL, *dstBuf = NULL;
+
+  GET_HANDLE();
+
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF)
+    THROW_ARG("Invalid argument in decompress()");
+  if (org_libjpegturbo_turbojpeg_TJ_NUMPF != TJ_NUMPF)
+    THROW_ARG("Mismatch between Java and C API");
+
+  if ((*env)->GetArrayLength(env, src) < jpegSize)
+    THROW_ARG("Source buffer is not large enough");
+  actualPitch = (pitch == 0) ? width * tjPixelSizepf : pitch;
+  arraySize = (y + height - 1) * actualPitch + (x + width) * tjPixelSizepf;
+  if ((*env)->GetArrayLength(env, dst) * dstElementSize < arraySize)
+    THROW_ARG("Destination buffer is not large enough");
+
+  BAILIF0(jpegBuf = (*env)->GetPrimitiveArrayCritical(env, src, 0));
+  BAILIF0(dstBuf = (*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+  if (tjDecompress2(handle, jpegBuf, (unsigned long)jpegSize,
+                    &dstBufy * actualPitch + x * tjPixelSizepf, width,
+                    pitch, height, pf, flags) == -1) {
+    SAFE_RELEASE(dst, dstBuf);
+    SAFE_RELEASE(src, jpegBuf);
+    THROW_TJ();
+  }
+
+bailout:
+  SAFE_RELEASE(dst, dstBuf);
+  SAFE_RELEASE(src, jpegBuf);
+}
+
+/* TurboJPEG 1.3.x: TJDecompressor::decompress() byte destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIIIII
+  (JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jbyteArray dst,
+   jint x, jint y, jint width, jint pitch, jint height, jint pf, jint flags)
+{
+  TJDecompressor_decompress(env, obj, src, jpegSize, dst, 1, x, y, width,
+                            pitch, height, pf, flags);
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompress() byte destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIII
+  (JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jbyteArray dst,
+   jint width, jint pitch, jint height, jint pf, jint flags)
+{
+  TJDecompressor_decompress(env, obj, src, jpegSize, dst, 1, 0, 0, width,
+                            pitch, height, pf, flags);
+}
+
+/* TurboJPEG 1.3.x: TJDecompressor::decompress() int destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIIIII
+  (JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jintArray dst,
+   jint x, jint y, jint width, jint stride, jint height, jint pf, jint flags)
+{
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF)
+    THROW_ARG("Invalid argument in decompress()");
+  if (tjPixelSizepf != sizeof(jint))
+    THROW_ARG("Pixel format must be 32-bit when decompressing to an integer buffer.");
+
+  TJDecompressor_decompress(env, obj, src, jpegSize, dst, sizeof(jint), x, y,
+                            width, stride * sizeof(jint), height, pf, flags);
+
+bailout:
+  return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompress() int destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIII
+  (JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jintArray dst,
+   jint width, jint stride, jint height, jint pf, jint flags)
+{
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF)
+    THROW_ARG("Invalid argument in decompress()");
+  if (tjPixelSizepf != sizeof(jint))
+    THROW_ARG("Pixel format must be 32-bit when decompressing to an integer buffer.");
+
+  TJDecompressor_decompress(env, obj, src, jpegSize, dst, sizeof(jint), 0, 0,
+                            width, stride * sizeof(jint), height, pf, flags);
+
+bailout:
+  return;
+}
+
+/* TurboJPEG 1.4.x: TJDecompressor::decompressToYUV() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3_3B_3II_3III
+  (JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize,
+   jobjectArray dstobjs, jintArray jDstOffsets, jint desiredWidth,
+   jintArray jDstStrides, jint desiredHeight, jint flags)
+{
+  tjhandle handle = 0;
+  unsigned char *jpegBuf = NULL;
+  jbyteArray jDstPlanes3 = { NULL, NULL, NULL };
+  unsigned char *dstPlanesTmp3 = { NULL, NULL, NULL };
+  unsigned char *dstPlanes3 = { NULL, NULL, NULL };
+  int *dstOffsetsTmp = NULL, dstOffsets3 = { 0, 0, 0 };
+  int *dstStridesTmp = NULL, dstStrides3 = { 0, 0, 0 };
+  int jpegSubsamp = -1, jpegWidth = 0, jpegHeight = 0;
+  int nc = 0, i, width, height, scaledWidth, scaledHeight, nsf = 0;
+  tjscalingfactor *sf;
+
+  GET_HANDLE();
+
+  if ((*env)->GetArrayLength(env, src) < jpegSize)
+    THROW_ARG("Source buffer is not large enough");
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+  jpegSubsamp = (int)(*env)->GetIntField(env, obj, _fid);
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+  jpegWidth = (int)(*env)->GetIntField(env, obj, _fid);
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+  jpegHeight = (int)(*env)->GetIntField(env, obj, _fid);
+
+  nc = (jpegSubsamp == org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY ? 1 : 3);
+
+  width = desiredWidth;
+  height = desiredHeight;
+  if (width == 0) width = jpegWidth;
+  if (height == 0) height = jpegHeight;
+  sf = tjGetScalingFactors(&nsf);
+  if (!sf || nsf < 1)
+    THROW_ARG(tjGetErrorStr());
+  for (i = 0; i < nsf; i++) {
+    scaledWidth = TJSCALED(jpegWidth, sfi);
+    scaledHeight = TJSCALED(jpegHeight, sfi);
+    if (scaledWidth <= width && scaledHeight <= height)
+      break;
+  }
+  if (i >= nsf)
+    THROW_ARG("Could not scale down to desired image dimensions");
+
+  BAILIF0(dstOffsetsTmp =
+          (*env)->GetPrimitiveArrayCritical(env, jDstOffsets, 0));
+  for (i = 0; i < nc; i++) dstOffsetsi = dstOffsetsTmpi;
+  SAFE_RELEASE(jDstOffsets, dstOffsetsTmp);
+
+  BAILIF0(dstStridesTmp =
+          (*env)->GetPrimitiveArrayCritical(env, jDstStrides, 0));
+  for (i = 0; i < nc; i++) dstStridesi = dstStridesTmpi;
+  SAFE_RELEASE(jDstStrides, dstStridesTmp);
+
+  for (i = 0; i < nc; i++) {
+    int planeSize = tjPlaneSizeYUV(i, scaledWidth, dstStridesi, scaledHeight,
+                                   jpegSubsamp);
+    int pw = tjPlaneWidth(i, scaledWidth, jpegSubsamp);
+
+    if (planeSize < 0 || pw < 0)
+      THROW_ARG(tjGetErrorStr());
+
+    if (dstOffsetsi < 0)
+      THROW_ARG("Invalid argument in decompressToYUV()");
+    if (dstStridesi < 0 && dstOffsetsi - planeSize + pw < 0)
+      THROW_ARG("Negative plane stride would cause memory to be accessed below plane boundary");
+
+    BAILIF0(jDstPlanesi = (*env)->GetObjectArrayElement(env, dstobjs, i));
+    if ((*env)->GetArrayLength(env, jDstPlanesi) <
+        dstOffsetsi + planeSize)
+      THROW_ARG("Destination plane is not large enough");
+
+    BAILIF0(dstPlanesTmpi =
+            (*env)->GetPrimitiveArrayCritical(env, jDstPlanesi, 0));
+    dstPlanesi = &dstPlanesTmpidstOffsetsi;
+    SAFE_RELEASE(jDstPlanesi, dstPlanesTmpi);
+  }
+  BAILIF0(jpegBuf = (*env)->GetPrimitiveArrayCritical(env, src, 0));
+
+  if (tjDecompressToYUVPlanes(handle, jpegBuf, (unsigned long)jpegSize,
+                              dstPlanes, desiredWidth, dstStrides,
+                              desiredHeight, flags) == -1) {
+    SAFE_RELEASE(src, jpegBuf);
+    THROW_TJ();
+  }
+
+bailout:
+  SAFE_RELEASE(src, jpegBuf);
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompressToYUV() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3BI
+  (JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jbyteArray dst,
+   jint flags)
+{
+  tjhandle handle = 0;
+  unsigned char *jpegBuf = NULL, *dstBuf = NULL;
+  int jpegSubsamp = -1, jpegWidth = 0, jpegHeight = 0;
+
+  GET_HANDLE();
+
+  if ((*env)->GetArrayLength(env, src) < jpegSize)
+    THROW_ARG("Source buffer is not large enough");
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+  jpegSubsamp = (int)(*env)->GetIntField(env, obj, _fid);
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+  jpegWidth = (int)(*env)->GetIntField(env, obj, _fid);
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+  jpegHeight = (int)(*env)->GetIntField(env, obj, _fid);
+  if ((*env)->GetArrayLength(env, dst) <
+      (jsize)tjBufSizeYUV(jpegWidth, jpegHeight, jpegSubsamp))
+    THROW_ARG("Destination buffer is not large enough");
+
+  BAILIF0(jpegBuf = (*env)->GetPrimitiveArrayCritical(env, src, 0));
+  BAILIF0(dstBuf = (*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+  if (tjDecompressToYUV(handle, jpegBuf, (unsigned long)jpegSize, dstBuf,
+                        flags) == -1) {
+    SAFE_RELEASE(dst, dstBuf);
+    SAFE_RELEASE(src, jpegBuf);
+    THROW_TJ();
+  }
+
+bailout:
+  SAFE_RELEASE(dst, dstBuf);
+  SAFE_RELEASE(src, jpegBuf);
+}
+
+static void TJDecompressor_decodeYUV
+  (JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+   jintArray jSrcStrides, jint subsamp, jarray dst, jint dstElementSize,
+   jint x, jint y, jint width, jint pitch, jint height, jint pf, jint flags)
+{
+  tjhandle handle = 0;
+  jsize arraySize = 0, actualPitch;
+  jbyteArray jSrcPlanes3 = { NULL, NULL, NULL };
+  const unsigned char *srcPlanesTmp3 = { NULL, NULL, NULL };
+  const unsigned char *srcPlanes3 = { NULL, NULL, NULL };
+  int *srcOffsetsTmp = NULL, srcOffsets3 = { 0, 0, 0 };
+  int *srcStridesTmp = NULL, srcStrides3 = { 0, 0, 0 };
+  unsigned char *dstBuf = NULL;
+  int nc = (subsamp == org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY ? 1 : 3), i;
+
+  GET_HANDLE();
+
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF || subsamp < 0 ||
+      subsamp >= org_libjpegturbo_turbojpeg_TJ_NUMSAMP)
+    THROW_ARG("Invalid argument in decodeYUV()");
+  if (org_libjpegturbo_turbojpeg_TJ_NUMPF != TJ_NUMPF ||
+      org_libjpegturbo_turbojpeg_TJ_NUMSAMP != TJ_NUMSAMP)
+    THROW_ARG("Mismatch between Java and C API");
+
+  if ((*env)->GetArrayLength(env, srcobjs) < nc)
+    THROW_ARG("Planes array is too small for the subsampling type");
+  if ((*env)->GetArrayLength(env, jSrcOffsets) < nc)
+    THROW_ARG("Offsets array is too small for the subsampling type");
+  if ((*env)->GetArrayLength(env, jSrcStrides) < nc)
+    THROW_ARG("Strides array is too small for the subsampling type");
+
+  actualPitch = (pitch == 0) ? width * tjPixelSizepf : pitch;
+  arraySize = (y + height - 1) * actualPitch + (x + width) * tjPixelSizepf;
+  if ((*env)->GetArrayLength(env, dst) * dstElementSize < arraySize)
+    THROW_ARG("Destination buffer is not large enough");
+
+  BAILIF0(srcOffsetsTmp =
+          (*env)->GetPrimitiveArrayCritical(env, jSrcOffsets, 0));
+  for (i = 0; i < nc; i++) srcOffsetsi = srcOffsetsTmpi;
+  SAFE_RELEASE(jSrcOffsets, srcOffsetsTmp);
+
+  BAILIF0(srcStridesTmp =
+          (*env)->GetPrimitiveArrayCritical(env, jSrcStrides, 0));
+  for (i = 0; i < nc; i++) srcStridesi = srcStridesTmpi;
+  SAFE_RELEASE(jSrcStrides, srcStridesTmp);
+
+  for (i = 0; i < nc; i++) {
+    int planeSize = tjPlaneSizeYUV(i, width, srcStridesi, height, subsamp);
+    int pw = tjPlaneWidth(i, width, subsamp);
+
+    if (planeSize < 0 || pw < 0)
+      THROW_ARG(tjGetErrorStr());
+
+    if (srcOffsetsi < 0)
+      THROW_ARG("Invalid argument in decodeYUV()");
+    if (srcStridesi < 0 && srcOffsetsi - planeSize + pw < 0)
+      THROW_ARG("Negative plane stride would cause memory to be accessed below plane boundary");
+
+    BAILIF0(jSrcPlanesi = (*env)->GetObjectArrayElement(env, srcobjs, i));
+    if ((*env)->GetArrayLength(env, jSrcPlanesi) <
+        srcOffsetsi + planeSize)
+      THROW_ARG("Source plane is not large enough");
+
+    BAILIF0(srcPlanesTmpi =
+            (*env)->GetPrimitiveArrayCritical(env, jSrcPlanesi, 0));
+    srcPlanesi = &srcPlanesTmpisrcOffsetsi;
+    SAFE_RELEASE(jSrcPlanesi, srcPlanesTmpi);
+  }
+  BAILIF0(dstBuf = (*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+  if (tjDecodeYUVPlanes(handle, srcPlanes, srcStrides, subsamp,
+                        &dstBufy * actualPitch + x * tjPixelSizepf, width,
+                        pitch, height, pf, flags) == -1) {
+    SAFE_RELEASE(dst, dstBuf);
+    THROW_TJ();
+  }
+
+bailout:
+  SAFE_RELEASE(dst, dstBuf);
+}
+
+/* TurboJPEG 1.4.x: TJDecompressor::decodeYUV() byte destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3BIIIIIII
+  (JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+   jintArray jSrcStrides, jint subsamp, jbyteArray dst, jint x, jint y,
+   jint width, jint pitch, jint height, jint pf, jint flags)
+{
+  TJDecompressor_decodeYUV(env, obj, srcobjs, jSrcOffsets, jSrcStrides,
+                           subsamp, dst, 1, x, y, width, pitch, height, pf,
+                           flags);
+}
+
+/* TurboJPEG 1.4.x: TJDecompressor::decodeYUV() int destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3IIIIIIII
+  (JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+   jintArray jSrcStrides, jint subsamp, jintArray dst, jint x, jint y,
+   jint width, jint stride, jint height, jint pf, jint flags)
+{
+  if (pf < 0 || pf >= org_libjpegturbo_turbojpeg_TJ_NUMPF)
+    THROW_ARG("Invalid argument in decodeYUV()");
+  if (tjPixelSizepf != sizeof(jint))
+    THROW_ARG("Pixel format must be 32-bit when decoding to an integer buffer.");
+
+  TJDecompressor_decodeYUV(env, obj, srcobjs, jSrcOffsets, jSrcStrides,
+                           subsamp, dst, sizeof(jint), x, y, width,
+                           stride * sizeof(jint), height, pf, flags);
+
+bailout:
+  return;
+}
+
+/* TurboJPEG 1.2.x: TJTransformer::init() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJTransformer_init
+  (JNIEnv *env, jobject obj)
+{
+  jclass cls;
+  jfieldID fid;
+  tjhandle handle;
+
+  if ((handle = tjInitTransform()) == NULL)
+    THROW(tjGetErrorStr(), "org/libjpegturbo/turbojpeg/TJException");
+
+  BAILIF0(cls = (*env)->GetObjectClass(env, obj));
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "handle", "J"));
+  (*env)->SetLongField(env, obj, fid, (size_t)handle);
+
+bailout:
+  return;
+}
+
+typedef struct _JNICustomFilterParams {
+  JNIEnv *env;
+  jobject tobj;
+  jobject cfobj;
+} JNICustomFilterParams;
+
+static int JNICustomFilter(short *coeffs, tjregion arrayRegion,
+                           tjregion planeRegion, int componentIndex,
+                           int transformIndex, tjtransform *transform)
+{
+  JNICustomFilterParams *params = (JNICustomFilterParams *)transform->data;
+  JNIEnv *env = params->env;
+  jobject tobj = params->tobj, cfobj = params->cfobj;
+  jobject arrayRegionObj, planeRegionObj, bufobj, borobj;
+  jclass cls;
+  jmethodID mid;
+  jfieldID fid;
+
+  BAILIF0(bufobj = (*env)->NewDirectByteBuffer(env, coeffs,
+    sizeof(short) * arrayRegion.w * arrayRegion.h));
+  BAILIF0(cls = (*env)->FindClass(env, "java/nio/ByteOrder"));
+  BAILIF0(mid = (*env)->GetStaticMethodID(env, cls, "nativeOrder",
+                                          "()Ljava/nio/ByteOrder;"));
+  BAILIF0(borobj = (*env)->CallStaticObjectMethod(env, cls, mid));
+  BAILIF0(cls = (*env)->GetObjectClass(env, bufobj));
+  BAILIF0(mid = (*env)->GetMethodID(env, cls, "order",
+    "(Ljava/nio/ByteOrder;)Ljava/nio/ByteBuffer;"));
+  (*env)->CallObjectMethod(env, bufobj, mid, borobj);
+  BAILIF0(mid = (*env)->GetMethodID(env, cls, "asShortBuffer",
+                                    "()Ljava/nio/ShortBuffer;"));
+  BAILIF0(bufobj = (*env)->CallObjectMethod(env, bufobj, mid));
+
+  BAILIF0(cls = (*env)->FindClass(env, "java/awt/Rectangle"));
+  BAILIF0(arrayRegionObj = (*env)->AllocObject(env, cls));
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "x", "I"));
+  (*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.x);
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "y", "I"));
+  (*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.y);
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "width", "I"));
+  (*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.w);
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "height", "I"));
+  (*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.h);
+
+  BAILIF0(planeRegionObj = (*env)->AllocObject(env, cls));
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "x", "I"));
+  (*env)->SetIntField(env, planeRegionObj, fid, planeRegion.x);
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "y", "I"));
+  (*env)->SetIntField(env, planeRegionObj, fid, planeRegion.y);
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "width", "I"));
+  (*env)->SetIntField(env, planeRegionObj, fid, planeRegion.w);
+  BAILIF0(fid = (*env)->GetFieldID(env, cls, "height", "I"));
+  (*env)->SetIntField(env, planeRegionObj, fid, planeRegion.h);
+
+  BAILIF0(cls = (*env)->GetObjectClass(env, cfobj));
+  BAILIF0(mid = (*env)->GetMethodID(env, cls, "customFilter",
+    "(Ljava/nio/ShortBuffer;Ljava/awt/Rectangle;Ljava/awt/Rectangle;IILorg/libjpegturbo/turbojpeg/TJTransform;)V"));
+  (*env)->CallVoidMethod(env, cfobj, mid, bufobj, arrayRegionObj,
+                         planeRegionObj, componentIndex, transformIndex, tobj);
+
+  return 0;
+
+bailout:
+  return -1;
+}
+
+/* TurboJPEG 1.2.x: TJTransformer::transform() */
+JNIEXPORT jintArray JNICALL Java_org_libjpegturbo_turbojpeg_TJTransformer_transform
+  (JNIEnv *env, jobject obj, jbyteArray jsrcBuf, jint jpegSize,
+   jobjectArray dstobjs, jobjectArray tobjs, jint flags)
+{
+  tjhandle handle = 0;
+  unsigned char *jpegBuf = NULL, **dstBufs = NULL;
+  jsize n = 0;
+  unsigned long *dstSizes = NULL;
+  tjtransform *t = NULL;
+  jbyteArray *jdstBufs = NULL;
+  int i, jpegWidth = 0, jpegHeight = 0, jpegSubsamp;
+  jintArray jdstSizes = 0;
+  jint *dstSizesi = NULL;
+  JNICustomFilterParams *params = NULL;
+
+  GET_HANDLE();
+
+  if ((*env)->GetArrayLength(env, jsrcBuf) < jpegSize)
+    THROW_ARG("Source buffer is not large enough");
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+  jpegWidth = (int)(*env)->GetIntField(env, obj, _fid);
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+  jpegHeight = (int)(*env)->GetIntField(env, obj, _fid);
+  BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+  jpegSubsamp = (int)(*env)->GetIntField(env, obj, _fid);
+
+  n = (*env)->GetArrayLength(env, dstobjs);
+  if (n != (*env)->GetArrayLength(env, tobjs))
+    THROW_ARG("Mismatch between size of transforms array and destination buffers array");
+
+  if ((dstBufs =
+       (unsigned char **)malloc(sizeof(unsigned char *) * n)) == NULL)
+    THROW_MEM();
+  if ((jdstBufs = (jbyteArray *)malloc(sizeof(jbyteArray) * n)) == NULL)
+    THROW_MEM();
+  if ((dstSizes = (unsigned long *)malloc(sizeof(unsigned long) * n)) == NULL)
+    THROW_MEM();
+  if ((t = (tjtransform *)malloc(sizeof(tjtransform) * n)) == NULL)
+    THROW_MEM();
+  if ((params = (JNICustomFilterParams *)malloc(sizeof(JNICustomFilterParams) *
+                                                n)) == NULL)
+    THROW_MEM();
+  for (i = 0; i < n; i++) {
+    dstBufsi = NULL;  jdstBufsi = NULL;  dstSizesi = 0;
+    memset(&ti, 0, sizeof(tjtransform));
+    memset(&paramsi, 0, sizeof(JNICustomFilterParams));
+  }
+
+  for (i = 0; i < n; i++) {
+    jobject tobj, cfobj;
+
+    BAILIF0(tobj = (*env)->GetObjectArrayElement(env, tobjs, i));
+    BAILIF0(_cls = (*env)->GetObjectClass(env, tobj));
+    BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "op", "I"));
+    ti.op = (*env)->GetIntField(env, tobj, _fid);
+    BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "options", "I"));
+    ti.options = (*env)->GetIntField(env, tobj, _fid);
+    BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "x", "I"));
+    ti.r.x = (*env)->GetIntField(env, tobj, _fid);
+    BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "y", "I"));
+    ti.r.y = (*env)->GetIntField(env, tobj, _fid);
+    BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "width", "I"));
+    ti.r.w = (*env)->GetIntField(env, tobj, _fid);
+    BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "height", "I"));
+    ti.r.h = (*env)->GetIntField(env, tobj, _fid);
+
+    BAILIF0(_fid = (*env)->GetFieldID(env, _cls, "cf",
+      "Lorg/libjpegturbo/turbojpeg/TJCustomFilter;"));
+    cfobj = (*env)->GetObjectField(env, tobj, _fid);
+    if (cfobj) {
+      paramsi.env = env;
+      paramsi.tobj = tobj;
+      paramsi.cfobj = cfobj;
+      ti.customFilter = JNICustomFilter;
+      ti.data = (void *)&paramsi;
+    }
+  }
+
+  for (i = 0; i < n; i++) {
+    int w = jpegWidth, h = jpegHeight;
+
+    if (ti.r.w != 0) w = ti.r.w;
+    if (ti.r.h != 0) h = ti.r.h;
+    BAILIF0(jdstBufsi = (*env)->GetObjectArrayElement(env, dstobjs, i));
+    if ((unsigned long)(*env)->GetArrayLength(env, jdstBufsi) <
+        tjBufSize(w, h, jpegSubsamp))
+      THROW_ARG("Destination buffer is not large enough");
+  }
+  BAILIF0(jpegBuf = (*env)->GetPrimitiveArrayCritical(env, jsrcBuf, 0));
+  for (i = 0; i < n; i++)
+    BAILIF0(dstBufsi =
+            (*env)->GetPrimitiveArrayCritical(env, jdstBufsi, 0));
+
+  if (tjTransform(handle, jpegBuf, jpegSize, n, dstBufs, dstSizes, t,
+                  flags | TJFLAG_NOREALLOC) == -1) {
+    for (i = 0; i < n; i++)
+      SAFE_RELEASE(jdstBufsi, dstBufsi);
+    SAFE_RELEASE(jsrcBuf, jpegBuf);
+    THROW_TJ();
+  }
+
+  for (i = 0; i < n; i++)
+    SAFE_RELEASE(jdstBufsi, dstBufsi);
+  SAFE_RELEASE(jsrcBuf, jpegBuf);
+
+  jdstSizes = (*env)->NewIntArray(env, n);
+  BAILIF0(dstSizesi = (*env)->GetIntArrayElements(env, jdstSizes, 0));
+  for (i = 0; i < n; i++) dstSizesii = (int)dstSizesi;
+
+bailout:
+  if (dstSizesi) (*env)->ReleaseIntArrayElements(env, jdstSizes, dstSizesi, 0);
+  if (dstBufs) {
+    for (i = 0; i < n; i++) {
+      if (dstBufsi && jdstBufs && jdstBufsi)
+        (*env)->ReleasePrimitiveArrayCritical(env, jdstBufsi, dstBufsi, 0);
+    }
+    free(dstBufs);
+  }
+  SAFE_RELEASE(jsrcBuf, jpegBuf);
+  free(jdstBufs);
+  free(dstSizes);
+  free(t);
+  return jdstSizes;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::destroy() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_destroy
+  (JNIEnv *env, jobject obj)
+{
+  Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy(env, obj);
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/turbojpeg-mapfile Added

@@ -0,0 +1,65 @@
+TURBOJPEG_1.0
+{
+  global:
+    tjInitCompress;
+    tjCompress;
+    TJBUFSIZE;
+    tjInitDecompress;
+    tjDecompressHeader;
+    tjDecompress;
+    tjDestroy;
+    tjGetErrorStr;
+  local:
+    *;
+};
+
+TURBOJPEG_1.1
+{
+  global:
+    TJBUFSIZEYUV;
+    tjDecompressHeader2;
+    tjDecompressToYUV;
+    tjEncodeYUV;
+} TURBOJPEG_1.0;
+
+TURBOJPEG_1.2
+{
+  global:
+    tjAlloc;
+    tjBufSize;
+    tjBufSizeYUV;
+    tjCompress2;
+    tjDecompress2;
+    tjEncodeYUV2;
+    tjFree;
+    tjGetScalingFactors;
+    tjInitTransform;
+    tjTransform;
+} TURBOJPEG_1.1;
+
+TURBOJPEG_1.4
+{
+  global:
+    tjBufSizeYUV2;
+    tjCompressFromYUV;
+    tjCompressFromYUVPlanes;
+    tjDecodeYUV;
+    tjDecodeYUVPlanes;
+    tjDecompressHeader3;
+    tjDecompressToYUV2;
+    tjDecompressToYUVPlanes;
+    tjEncodeYUV3;
+    tjEncodeYUVPlanes;
+    tjPlaneHeight;
+    tjPlaneSizeYUV;
+    tjPlaneWidth;
+} TURBOJPEG_1.2;
+
+TURBOJPEG_2.0
+{
+  global:
+    tjGetErrorCode;
+    tjGetErrorStr2;
+    tjLoadImage;
+    tjSaveImage;
+} TURBOJPEG_1.4;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/turbojpeg-mapfile.jni Added

@@ -0,0 +1,101 @@
+TURBOJPEG_1.0
+{
+  global:
+    tjInitCompress;
+    tjCompress;
+    TJBUFSIZE;
+    tjInitDecompress;
+    tjDecompressHeader;
+    tjDecompress;
+    tjDestroy;
+    tjGetErrorStr;
+  local:
+    *;
+};
+
+TURBOJPEG_1.1
+{
+  global:
+    TJBUFSIZEYUV;
+    tjDecompressHeader2;
+    tjDecompressToYUV;
+    tjEncodeYUV;
+} TURBOJPEG_1.0;
+
+TURBOJPEG_1.2
+{
+  global:
+    tjAlloc;
+    tjBufSize;
+    tjBufSizeYUV;
+    tjCompress2;
+    tjDecompress2;
+    tjEncodeYUV2;
+    tjFree;
+    tjGetScalingFactors;
+    tjInitTransform;
+    tjTransform;
+    Java_org_libjpegturbo_turbojpeg_TJ_bufSize;
+    Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__III;
+    Java_org_libjpegturbo_turbojpeg_TJ_getScalingFactors;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_init;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIII_3BIII;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIII_3BIII;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIII_3BII;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIII_3BII;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_init;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressHeader;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIII;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIII;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3BI;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_destroy;
+    Java_org_libjpegturbo_turbojpeg_TJTransformer_init;
+    Java_org_libjpegturbo_turbojpeg_TJTransformer_transform;
+} TURBOJPEG_1.1;
+
+TURBOJPEG_1.3
+{
+  global:
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIIIII_3BIII;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIIIII_3BIII;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIIIII;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIIIII;
+} TURBOJPEG_1.2;
+
+TURBOJPEG_1.4
+{
+  global:
+    tjBufSizeYUV2;
+    tjCompressFromYUV;
+    tjCompressFromYUVPlanes;
+    tjDecodeYUV;
+    tjDecodeYUVPlanes;
+    tjDecompressHeader3;
+    tjDecompressToYUV2;
+    tjDecompressToYUVPlanes;
+    tjEncodeYUV3;
+    tjEncodeYUVPlanes;
+    tjPlaneHeight;
+    tjPlaneSizeYUV;
+    tjPlaneWidth;
+    Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_compressFromYUV___3_3B_3II_3III_3BII;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIIIII_3_3B_3I_3III;
+    Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIIIII_3_3B_3I_3III;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3_3B_3II_3III;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3BIIIIIII;
+    Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3IIIIIIII;
+    Java_org_libjpegturbo_turbojpeg_TJ_planeHeight__III;
+    Java_org_libjpegturbo_turbojpeg_TJ_planeSizeYUV__IIIII;
+    Java_org_libjpegturbo_turbojpeg_TJ_planeWidth__III;
+} TURBOJPEG_1.3;
+
+TURBOJPEG_2.0
+{
+  global:
+    tjGetErrorCode;
+    tjGetErrorStr2;
+    tjLoadImage;
+    tjSaveImage;
+} TURBOJPEG_1.4;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/turbojpeg.c Added

@@ -0,0 +1,2222 @@
+/*
+ * Copyright (C)2009-2021 D. R. Commander.  All Rights Reserved.
+ * Copyright (C)2021 Alex Richardson.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* TurboJPEG/LJT:  this implements the TurboJPEG API using libjpeg or
+   libjpeg-turbo */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <ctype.h>
+#include <jinclude.h>
+#define JPEG_INTERNALS
+#include <jpeglib.h>
+#include <jerror.h>
+#include <setjmp.h>
+#include <errno.h>
+#include "./turbojpeg.h"
+#include "./tjutil.h"
+#include "transupp.h"
+#include "./jpegcomp.h"
+#include "./cdjpeg.h"
+#include "jconfigint.h"
+
+extern void jpeg_mem_dest_tj(j_compress_ptr, unsigned char **, unsigned long *,
+                             boolean);
+extern void jpeg_mem_src_tj(j_decompress_ptr, const unsigned char *,
+                            unsigned long);
+
+#define PAD(v, p)  ((v + (p) - 1) & (~((p) - 1)))
+#define IS_POW2(x)  (((x) & (x - 1)) == 0)
+
+
+/* Error handling (based on example in example.txt) */
+
+static THREAD_LOCAL char errStrJMSG_LENGTH_MAX = "No error";
+
+struct my_error_mgr {
+  struct jpeg_error_mgr pub;
+  jmp_buf setjmp_buffer;
+  void (*emit_message) (j_common_ptr, int);
+  boolean warning, stopOnWarning;
+};
+typedef struct my_error_mgr *my_error_ptr;
+
+#define JMESSAGE(code, string)  string,
+static const char *turbojpeg_message_table = {
+#include "cderror.h"
+  NULL
+};
+
+static void my_error_exit(j_common_ptr cinfo)
+{
+  my_error_ptr myerr = (my_error_ptr)cinfo->err;
+
+  (*cinfo->err->output_message) (cinfo);
+  longjmp(myerr->setjmp_buffer, 1);
+}
+
+/* Based on output_message() in jerror.c */
+
+static void my_output_message(j_common_ptr cinfo)
+{
+  (*cinfo->err->format_message) (cinfo, errStr);
+}
+
+static void my_emit_message(j_common_ptr cinfo, int msg_level)
+{
+  my_error_ptr myerr = (my_error_ptr)cinfo->err;
+
+  myerr->emit_message(cinfo, msg_level);
+  if (msg_level < 0) {
+    myerr->warning = TRUE;
+    if (myerr->stopOnWarning) longjmp(myerr->setjmp_buffer, 1);
+  }
+}
+
+
+/* Global structures, macros, etc. */
+
+enum { COMPRESS = 1, DECOMPRESS = 2 };
+
+typedef struct _tjinstance {
+  struct jpeg_compress_struct cinfo;
+  struct jpeg_decompress_struct dinfo;
+  struct my_error_mgr jerr;
+  int init, headerRead;
+  char errStrJMSG_LENGTH_MAX;
+  boolean isInstanceError;
+} tjinstance;
+
+struct my_progress_mgr {
+  struct jpeg_progress_mgr pub;
+  tjinstance *this;
+};
+typedef struct my_progress_mgr *my_progress_ptr;
+
+static void my_progress_monitor(j_common_ptr dinfo)
+{
+  my_error_ptr myerr = (my_error_ptr)dinfo->err;
+  my_progress_ptr myprog = (my_progress_ptr)dinfo->progress;
+
+  if (dinfo->is_decompressor) {
+    int scan_no = ((j_decompress_ptr)dinfo)->input_scan_number;
+
+    if (scan_no > 500) {
+      snprintf(myprog->this->errStr, JMSG_LENGTH_MAX,
+               "Progressive JPEG image has more than 500 scans");
+      snprintf(errStr, JMSG_LENGTH_MAX,
+               "Progressive JPEG image has more than 500 scans");
+      myprog->this->isInstanceError = TRUE;
+      myerr->warning = FALSE;
+      longjmp(myerr->setjmp_buffer, 1);
+    }
+  }
+}
+
+static const int pixelsizeTJ_NUMSAMP = { 3, 3, 3, 1, 3, 3 };
+
+static const JXFORM_CODE xformtypesTJ_NUMXOP = {
+  JXFORM_NONE, JXFORM_FLIP_H, JXFORM_FLIP_V, JXFORM_TRANSPOSE,
+  JXFORM_TRANSVERSE, JXFORM_ROT_90, JXFORM_ROT_180, JXFORM_ROT_270
+};
+
+#define NUMSF  16
+static const tjscalingfactor sfNUMSF = {
+  { 2, 1 },
+  { 15, 8 },
+  { 7, 4 },
+  { 13, 8 },
+  { 3, 2 },
+  { 11, 8 },
+  { 5, 4 },
+  { 9, 8 },
+  { 1, 1 },
+  { 7, 8 },
+  { 3, 4 },
+  { 5, 8 },
+  { 1, 2 },
+  { 3, 8 },
+  { 1, 4 },
+  { 1, 8 }
+};
+
+static J_COLOR_SPACE pf2csTJ_NUMPF = {
+  JCS_EXT_RGB, JCS_EXT_BGR, JCS_EXT_RGBX, JCS_EXT_BGRX, JCS_EXT_XBGR,
+  JCS_EXT_XRGB, JCS_GRAYSCALE, JCS_EXT_RGBA, JCS_EXT_BGRA, JCS_EXT_ABGR,
+  JCS_EXT_ARGB, JCS_CMYK
+};
+
+static int cs2pfJPEG_NUMCS = {
+  TJPF_UNKNOWN, TJPF_GRAY,
+#if RGB_RED == 0 && RGB_GREEN == 1 && RGB_BLUE == 2 && RGB_PIXELSIZE == 3
+  TJPF_RGB,
+#elif RGB_RED == 2 && RGB_GREEN == 1 && RGB_BLUE == 0 && RGB_PIXELSIZE == 3
+  TJPF_BGR,
+#elif RGB_RED == 0 && RGB_GREEN == 1 && RGB_BLUE == 2 && RGB_PIXELSIZE == 4
+  TJPF_RGBX,
+#elif RGB_RED == 2 && RGB_GREEN == 1 && RGB_BLUE == 0 && RGB_PIXELSIZE == 4
+  TJPF_BGRX,
+#elif RGB_RED == 3 && RGB_GREEN == 2 && RGB_BLUE == 1 && RGB_PIXELSIZE == 4
+  TJPF_XBGR,
+#elif RGB_RED == 1 && RGB_GREEN == 2 && RGB_BLUE == 3 && RGB_PIXELSIZE == 4
+  TJPF_XRGB,
+#endif
+  TJPF_UNKNOWN, TJPF_CMYK, TJPF_UNKNOWN, TJPF_RGB, TJPF_RGBX, TJPF_BGR,
+  TJPF_BGRX, TJPF_XBGR, TJPF_XRGB, TJPF_RGBA, TJPF_BGRA, TJPF_ABGR, TJPF_ARGB,
+  TJPF_UNKNOWN
+};
+
+#define THROWG(m) { \
+  snprintf(errStr, JMSG_LENGTH_MAX, "%s", m); \
+  retval = -1;  goto bailout; \
+}
+#define THROW_UNIX(m) { \
+  snprintf(errStr, JMSG_LENGTH_MAX, "%s\n%s", m, strerror(errno)); \
+  retval = -1;  goto bailout; \
+}
+#define THROW(m) { \
+  snprintf(this->errStr, JMSG_LENGTH_MAX, "%s", m); \
+  this->isInstanceError = TRUE;  THROWG(m) \
+}
+
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+/* Private flag that triggers different TurboJPEG API behavior when fuzzing */
+#define TJFLAG_FUZZING  (1 << 30)
+#endif
+
+#define GET_INSTANCE(handle) \
+  tjinstance *this = (tjinstance *)handle; \
+  j_compress_ptr cinfo = NULL; \
+  j_decompress_ptr dinfo = NULL; \
+  \
+  if (!this) { \
+    snprintf(errStr, JMSG_LENGTH_MAX, "Invalid handle"); \
+    return -1; \
+  } \
+  cinfo = &this->cinfo;  dinfo = &this->dinfo; \
+  this->jerr.warning = FALSE; \
+  this->isInstanceError = FALSE;
+
+#define GET_CINSTANCE(handle) \
+  tjinstance *this = (tjinstance *)handle; \
+  j_compress_ptr cinfo = NULL; \
+  \
+  if (!this) { \
+    snprintf(errStr, JMSG_LENGTH_MAX, "Invalid handle"); \
+    return -1; \
+  } \
+  cinfo = &this->cinfo; \
+  this->jerr.warning = FALSE; \
+  this->isInstanceError = FALSE;
+
+#define GET_DINSTANCE(handle) \
+  tjinstance *this = (tjinstance *)handle; \
+  j_decompress_ptr dinfo = NULL; \
+  \
+  if (!this) { \
+    snprintf(errStr, JMSG_LENGTH_MAX, "Invalid handle"); \
+    return -1; \
+  } \
+  dinfo = &this->dinfo; \
+  this->jerr.warning = FALSE; \
+  this->isInstanceError = FALSE;
+
+static int getPixelFormat(int pixelSize, int flags)
+{
+  if (pixelSize == 1) return TJPF_GRAY;
+  if (pixelSize == 3) {
+    if (flags & TJ_BGR) return TJPF_BGR;
+    else return TJPF_RGB;
+  }
+  if (pixelSize == 4) {
+    if (flags & TJ_ALPHAFIRST) {
+      if (flags & TJ_BGR) return TJPF_XBGR;
+      else return TJPF_XRGB;
+    } else {
+      if (flags & TJ_BGR) return TJPF_BGRX;
+      else return TJPF_RGBX;
+    }
+  }
+  return -1;
+}
+
+static void setCompDefaults(struct jpeg_compress_struct *cinfo,
+                            int pixelFormat, int subsamp, int jpegQual,
+                            int flags)
+{
+#ifndef NO_GETENV
+  char *env = NULL;
+#endif
+
+  cinfo->in_color_space = pf2cspixelFormat;
+  cinfo->input_components = tjPixelSizepixelFormat;
+  jpeg_set_defaults(cinfo);
+
+#ifndef NO_GETENV
+  if ((env = getenv("TJ_OPTIMIZE")) != NULL && strlen(env) > 0 &&
+      !strcmp(env, "1"))
+    cinfo->optimize_coding = TRUE;
+  if ((env = getenv("TJ_ARITHMETIC")) != NULL && strlen(env) > 0 &&
+      !strcmp(env, "1"))
+    cinfo->arith_code = TRUE;
+  if ((env = getenv("TJ_RESTART")) != NULL && strlen(env) > 0) {
+    int temp = -1;
+    char tempc = 0;
+
+    if (sscanf(env, "%d%c", &temp, &tempc) >= 1 && temp >= 0 &&
+        temp <= 65535) {
+      if (toupper(tempc) == 'B') {
+        cinfo->restart_interval = temp;
+        cinfo->restart_in_rows = 0;
+      } else
+        cinfo->restart_in_rows = temp;
+    }
+  }
+#endif
+
+  if (jpegQual >= 0) {
+    jpeg_set_quality(cinfo, jpegQual, TRUE);
+    if (jpegQual >= 96 || flags & TJFLAG_ACCURATEDCT)
+      cinfo->dct_method = JDCT_ISLOW;
+    else
+      cinfo->dct_method = JDCT_FASTEST;
+  }
+  if (subsamp == TJSAMP_GRAY)
+    jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+  else if (pixelFormat == TJPF_CMYK)
+    jpeg_set_colorspace(cinfo, JCS_YCCK);
+  else
+    jpeg_set_colorspace(cinfo, JCS_YCbCr);
+
+  if (flags & TJFLAG_PROGRESSIVE)
+    jpeg_simple_progression(cinfo);
+#ifndef NO_GETENV
+  else if ((env = getenv("TJ_PROGRESSIVE")) != NULL && strlen(env) > 0 &&
+           !strcmp(env, "1"))
+    jpeg_simple_progression(cinfo);
+#endif
+
+  cinfo->comp_info0.h_samp_factor = tjMCUWidthsubsamp / 8;
+  cinfo->comp_info1.h_samp_factor = 1;
+  cinfo->comp_info2.h_samp_factor = 1;
+  if (cinfo->num_components > 3)
+    cinfo->comp_info3.h_samp_factor = tjMCUWidthsubsamp / 8;
+  cinfo->comp_info0.v_samp_factor = tjMCUHeightsubsamp / 8;
+  cinfo->comp_info1.v_samp_factor = 1;
+  cinfo->comp_info2.v_samp_factor = 1;
+  if (cinfo->num_components > 3)
+    cinfo->comp_info3.v_samp_factor = tjMCUHeightsubsamp / 8;
+}
+
+
+static int getSubsamp(j_decompress_ptr dinfo)
+{
+  int retval = -1, i, k;
+
+  /* The sampling factors actually have no meaning with grayscale JPEG files,
+     and in fact it's possible to generate grayscale JPEGs with sampling
+     factors > 1 (even though those sampling factors are ignored by the
+     decompressor.)  Thus, we need to treat grayscale as a special case. */
+  if (dinfo->num_components == 1 && dinfo->jpeg_color_space == JCS_GRAYSCALE)
+    return TJSAMP_GRAY;
+
+  for (i = 0; i < NUMSUBOPT; i++) {
+    if (dinfo->num_components == pixelsizei ||
+        ((dinfo->jpeg_color_space == JCS_YCCK ||
+          dinfo->jpeg_color_space == JCS_CMYK) &&
+         pixelsizei == 3 && dinfo->num_components == 4)) {
+      if (dinfo->comp_info0.h_samp_factor == tjMCUWidthi / 8 &&
+          dinfo->comp_info0.v_samp_factor == tjMCUHeighti / 8) {
+        int match = 0;
+
+        for (k = 1; k < dinfo->num_components; k++) {
+          int href = 1, vref = 1;
+
+          if ((dinfo->jpeg_color_space == JCS_YCCK ||
+               dinfo->jpeg_color_space == JCS_CMYK) && k == 3) {
+            href = tjMCUWidthi / 8;  vref = tjMCUHeighti / 8;
+          }
+          if (dinfo->comp_infok.h_samp_factor == href &&
+              dinfo->comp_infok.v_samp_factor == vref)
+            match++;
+        }
+        if (match == dinfo->num_components - 1) {
+          retval = i;  break;
+        }
+      }
+      /* Handle 4:2:2 and 4:4:0 images whose sampling factors are specified
+         in non-standard ways. */
+      if (dinfo->comp_info0.h_samp_factor == 2 &&
+          dinfo->comp_info0.v_samp_factor == 2 &&
+          (i == TJSAMP_422 || i == TJSAMP_440)) {
+        int match = 0;
+
+        for (k = 1; k < dinfo->num_components; k++) {
+          int href = tjMCUHeighti / 8, vref = tjMCUWidthi / 8;
+
+          if ((dinfo->jpeg_color_space == JCS_YCCK ||
+               dinfo->jpeg_color_space == JCS_CMYK) && k == 3) {
+            href = vref = 2;
+          }
+          if (dinfo->comp_infok.h_samp_factor == href &&
+              dinfo->comp_infok.v_samp_factor == vref)
+            match++;
+        }
+        if (match == dinfo->num_components - 1) {
+          retval = i;  break;
+        }
+      }
+      /* Handle 4:4:4 images whose sampling factors are specified in
+         non-standard ways. */
+      if (dinfo->comp_info0.h_samp_factor *
+          dinfo->comp_info0.v_samp_factor <=
+          D_MAX_BLOCKS_IN_MCU / pixelsizei && i == TJSAMP_444) {
+        int match = 0;
+        for (k = 1; k < dinfo->num_components; k++) {
+          if (dinfo->comp_infok.h_samp_factor ==
+              dinfo->comp_info0.h_samp_factor &&
+              dinfo->comp_infok.v_samp_factor ==
+              dinfo->comp_info0.v_samp_factor)
+            match++;
+          if (match == dinfo->num_components - 1) {
+            retval = i;  break;
+          }
+        }
+      }
+    }
+  }
+  return retval;
+}
+
+
+/* General API functions */
+
+DLLEXPORT char *tjGetErrorStr2(tjhandle handle)
+{
+  tjinstance *this = (tjinstance *)handle;
+
+  if (this && this->isInstanceError) {
+    this->isInstanceError = FALSE;
+    return this->errStr;
+  } else
+    return errStr;
+}
+
+
+DLLEXPORT char *tjGetErrorStr(void)
+{
+  return errStr;
+}
+
+
+DLLEXPORT int tjGetErrorCode(tjhandle handle)
+{
+  tjinstance *this = (tjinstance *)handle;
+
+  if (this && this->jerr.warning) return TJERR_WARNING;
+  else return TJERR_FATAL;
+}
+
+
+DLLEXPORT int tjDestroy(tjhandle handle)
+{
+  GET_INSTANCE(handle);
+
+  if (setjmp(this->jerr.setjmp_buffer)) return -1;
+  if (this->init & COMPRESS) jpeg_destroy_compress(cinfo);
+  if (this->init & DECOMPRESS) jpeg_destroy_decompress(dinfo);
+  free(this);
+  return 0;
+}
+
+
+/* These are exposed mainly because Windows can't malloc() and free() across
+   DLL boundaries except when the CRT DLL is used, and we don't use the CRT DLL
+   with turbojpeg.dll for compatibility reasons.  However, these functions
+   can potentially be used for other purposes by different implementations. */
+
+DLLEXPORT void tjFree(unsigned char *buf)
+{
+  free(buf);
+}
+
+
+DLLEXPORT unsigned char *tjAlloc(int bytes)
+{
+  return (unsigned char *)malloc(bytes);
+}
+
+
+/* Compressor  */
+
+static tjhandle _tjInitCompress(tjinstance *this)
+{
+  static unsigned char buffer1;
+  unsigned char *buf = buffer;
+  unsigned long size = 1;
+
+  /* This is also straight out of example.txt */
+  this->cinfo.err = jpeg_std_error(&this->jerr.pub);
+  this->jerr.pub.error_exit = my_error_exit;
+  this->jerr.pub.output_message = my_output_message;
+  this->jerr.emit_message = this->jerr.pub.emit_message;
+  this->jerr.pub.emit_message = my_emit_message;
+  this->jerr.pub.addon_message_table = turbojpeg_message_table;
+  this->jerr.pub.first_addon_message = JMSG_FIRSTADDONCODE;
+  this->jerr.pub.last_addon_message = JMSG_LASTADDONCODE;
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    free(this);
+    return NULL;
+  }
+
+  jpeg_create_compress(&this->cinfo);
+  /* Make an initial call so it will create the destination manager */
+  jpeg_mem_dest_tj(&this->cinfo, &buf, &size, 0);
+
+  this->init |= COMPRESS;
+  return (tjhandle)this;
+}
+
+DLLEXPORT tjhandle tjInitCompress(void)
+{
+  tjinstance *this = NULL;
+
+  if ((this = (tjinstance *)malloc(sizeof(tjinstance))) == NULL) {
+    snprintf(errStr, JMSG_LENGTH_MAX,
+             "tjInitCompress(): Memory allocation failure");
+    return NULL;
+  }
+  MEMZERO(this, sizeof(tjinstance));
+  snprintf(this->errStr, JMSG_LENGTH_MAX, "No error");
+  return _tjInitCompress(this);
+}
+
+
+DLLEXPORT unsigned long tjBufSize(int width, int height, int jpegSubsamp)
+{
+  unsigned long long retval = 0;
+  int mcuw, mcuh, chromasf;
+
+  if (width < 1 || height < 1 || jpegSubsamp < 0 || jpegSubsamp >= NUMSUBOPT)
+    THROWG("tjBufSize(): Invalid argument");
+
+  /* This allows for rare corner cases in which a JPEG image can actually be
+     larger than the uncompressed input (we wouldn't mention it if it hadn't
+     happened before.) */
+  mcuw = tjMCUWidthjpegSubsamp;
+  mcuh = tjMCUHeightjpegSubsamp;
+  chromasf = jpegSubsamp == TJSAMP_GRAY ? 0 : 4 * 64 / (mcuw * mcuh);
+  retval = PAD(width, mcuw) * PAD(height, mcuh) * (2ULL + chromasf) + 2048ULL;
+  if (retval > (unsigned long long)((unsigned long)-1))
+    THROWG("tjBufSize(): Image is too large");
+
+bailout:
+  return (unsigned long)retval;
+}
+
+DLLEXPORT unsigned long TJBUFSIZE(int width, int height)
+{
+  unsigned long long retval = 0;
+
+  if (width < 1 || height < 1)
+    THROWG("TJBUFSIZE(): Invalid argument");
+
+  /* This allows for rare corner cases in which a JPEG image can actually be
+     larger than the uncompressed input (we wouldn't mention it if it hadn't
+     happened before.) */
+  retval = PAD(width, 16) * PAD(height, 16) * 6ULL + 2048ULL;
+  if (retval > (unsigned long long)((unsigned long)-1))
+    THROWG("TJBUFSIZE(): Image is too large");
+
+bailout:
+  return (unsigned long)retval;
+}
+
+
+DLLEXPORT unsigned long tjBufSizeYUV2(int width, int pad, int height,
+                                      int subsamp)
+{
+  unsigned long long retval = 0;
+  int nc, i;
+
+  if (subsamp < 0 || subsamp >= NUMSUBOPT)
+    THROWG("tjBufSizeYUV2(): Invalid argument");
+
+  nc = (subsamp == TJSAMP_GRAY ? 1 : 3);
+  for (i = 0; i < nc; i++) {
+    int pw = tjPlaneWidth(i, width, subsamp);
+    int stride = PAD(pw, pad);
+    int ph = tjPlaneHeight(i, height, subsamp);
+
+    if (pw < 0 || ph < 0) return -1;
+    else retval += (unsigned long long)stride * ph;
+  }
+  if (retval > (unsigned long long)((unsigned long)-1))
+    THROWG("tjBufSizeYUV2(): Image is too large");
+
+bailout:
+  return (unsigned long)retval;
+}
+
+DLLEXPORT unsigned long tjBufSizeYUV(int width, int height, int subsamp)
+{
+  return tjBufSizeYUV2(width, 4, height, subsamp);
+}
+
+DLLEXPORT unsigned long TJBUFSIZEYUV(int width, int height, int subsamp)
+{
+  return tjBufSizeYUV(width, height, subsamp);
+}
+
+
+DLLEXPORT int tjPlaneWidth(int componentID, int width, int subsamp)
+{
+  int pw, nc, retval = 0;
+
+  if (width < 1 || subsamp < 0 || subsamp >= TJ_NUMSAMP)
+    THROWG("tjPlaneWidth(): Invalid argument");
+  nc = (subsamp == TJSAMP_GRAY ? 1 : 3);
+  if (componentID < 0 || componentID >= nc)
+    THROWG("tjPlaneWidth(): Invalid argument");
+
+  pw = PAD(width, tjMCUWidthsubsamp / 8);
+  if (componentID == 0)
+    retval = pw;
+  else
+    retval = pw * 8 / tjMCUWidthsubsamp;
+
+bailout:
+  return retval;
+}
+
+
+DLLEXPORT int tjPlaneHeight(int componentID, int height, int subsamp)
+{
+  int ph, nc, retval = 0;
+
+  if (height < 1 || subsamp < 0 || subsamp >= TJ_NUMSAMP)
+    THROWG("tjPlaneHeight(): Invalid argument");
+  nc = (subsamp == TJSAMP_GRAY ? 1 : 3);
+  if (componentID < 0 || componentID >= nc)
+    THROWG("tjPlaneHeight(): Invalid argument");
+
+  ph = PAD(height, tjMCUHeightsubsamp / 8);
+  if (componentID == 0)
+    retval = ph;
+  else
+    retval = ph * 8 / tjMCUHeightsubsamp;
+
+bailout:
+  return retval;
+}
+
+
+DLLEXPORT unsigned long tjPlaneSizeYUV(int componentID, int width, int stride,
+                                       int height, int subsamp)
+{
+  unsigned long long retval = 0;
+  int pw, ph;
+
+  if (width < 1 || height < 1 || subsamp < 0 || subsamp >= NUMSUBOPT)
+    THROWG("tjPlaneSizeYUV(): Invalid argument");
+
+  pw = tjPlaneWidth(componentID, width, subsamp);
+  ph = tjPlaneHeight(componentID, height, subsamp);
+  if (pw < 0 || ph < 0) return -1;
+
+  if (stride == 0) stride = pw;
+  else stride = abs(stride);
+
+  retval = (unsigned long long)stride * (ph - 1) + pw;
+  if (retval > (unsigned long long)((unsigned long)-1))
+    THROWG("tjPlaneSizeYUV(): Image is too large");
+
+bailout:
+  return (unsigned long)retval;
+}
+
+
+DLLEXPORT int tjCompress2(tjhandle handle, const unsigned char *srcBuf,
+                          int width, int pitch, int height, int pixelFormat,
+                          unsigned char **jpegBuf, unsigned long *jpegSize,
+                          int jpegSubsamp, int jpegQual, int flags)
+{
+  int i, retval = 0, alloc = 1;
+  JSAMPROW *row_pointer = NULL;
+
+  GET_CINSTANCE(handle)
+  this->jerr.stopOnWarning = (flags & TJFLAG_STOPONWARNING) ? TRUE : FALSE;
+  if ((this->init & COMPRESS) == 0)
+    THROW("tjCompress2(): Instance has not been initialized for compression");
+
+  if (srcBuf == NULL || width <= 0 || pitch < 0 || height <= 0 ||
+      pixelFormat < 0 || pixelFormat >= TJ_NUMPF || jpegBuf == NULL ||
+      jpegSize == NULL || jpegSubsamp < 0 || jpegSubsamp >= NUMSUBOPT ||
+      jpegQual < 0 || jpegQual > 100)
+    THROW("tjCompress2(): Invalid argument");
+
+  if (pitch == 0) pitch = width * tjPixelSizepixelFormat;
+
+  if ((row_pointer = (JSAMPROW *)malloc(sizeof(JSAMPROW) * height)) == NULL)
+    THROW("tjCompress2(): Memory allocation failure");
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  cinfo->image_width = width;
+  cinfo->image_height = height;
+
+#ifndef NO_PUTENV
+  if (flags & TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+  else if (flags & TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+  else if (flags & TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+#endif
+
+  if (flags & TJFLAG_NOREALLOC) {
+    alloc = 0;  *jpegSize = tjBufSize(width, height, jpegSubsamp);
+  }
+  jpeg_mem_dest_tj(cinfo, jpegBuf, jpegSize, alloc);
+  setCompDefaults(cinfo, pixelFormat, jpegSubsamp, jpegQual, flags);
+
+  jpeg_start_compress(cinfo, TRUE);
+  for (i = 0; i < height; i++) {
+    if (flags & TJFLAG_BOTTOMUP)
+      row_pointeri = (JSAMPROW)&srcBuf(height - i - 1) * (size_t)pitch;
+    else
+      row_pointeri = (JSAMPROW)&srcBufi * (size_t)pitch;
+  }
+  while (cinfo->next_scanline < cinfo->image_height)
+    jpeg_write_scanlines(cinfo, &row_pointercinfo->next_scanline,
+                         cinfo->image_height - cinfo->next_scanline);
+  jpeg_finish_compress(cinfo);
+
+bailout:
+  if (cinfo->global_state > CSTATE_START) {
+    if (alloc) (*cinfo->dest->term_destination) (cinfo);
+    jpeg_abort_compress(cinfo);
+  }
+  free(row_pointer);
+  if (this->jerr.warning) retval = -1;
+  this->jerr.stopOnWarning = FALSE;
+  return retval;
+}
+
+DLLEXPORT int tjCompress(tjhandle handle, unsigned char *srcBuf, int width,
+                         int pitch, int height, int pixelSize,
+                         unsigned char *jpegBuf, unsigned long *jpegSize,
+                         int jpegSubsamp, int jpegQual, int flags)
+{
+  int retval = 0;
+  unsigned long size;
+
+  if (flags & TJ_YUV) {
+    size = tjBufSizeYUV(width, height, jpegSubsamp);
+    retval = tjEncodeYUV2(handle, srcBuf, width, pitch, height,
+                          getPixelFormat(pixelSize, flags), jpegBuf,
+                          jpegSubsamp, flags);
+  } else {
+    retval = tjCompress2(handle, srcBuf, width, pitch, height,
+                         getPixelFormat(pixelSize, flags), &jpegBuf, &size,
+                         jpegSubsamp, jpegQual, flags | TJFLAG_NOREALLOC);
+  }
+  *jpegSize = size;
+  return retval;
+}
+
+
+DLLEXPORT int tjEncodeYUVPlanes(tjhandle handle, const unsigned char *srcBuf,
+                                int width, int pitch, int height,
+                                int pixelFormat, unsigned char **dstPlanes,
+                                int *strides, int subsamp, int flags)
+{
+  JSAMPROW *row_pointer = NULL;
+  JSAMPLE *_tmpbufMAX_COMPONENTS, *_tmpbuf2MAX_COMPONENTS;
+  JSAMPROW *tmpbufMAX_COMPONENTS, *tmpbuf2MAX_COMPONENTS;
+  JSAMPROW *outbufMAX_COMPONENTS;
+  int i, retval = 0, row, pw0, ph0, pwMAX_COMPONENTS, phMAX_COMPONENTS;
+  JSAMPLE *ptr;
+  jpeg_component_info *compptr;
+
+  GET_CINSTANCE(handle);
+  this->jerr.stopOnWarning = (flags & TJFLAG_STOPONWARNING) ? TRUE : FALSE;
+
+  for (i = 0; i < MAX_COMPONENTS; i++) {
+    tmpbufi = NULL;  _tmpbufi = NULL;
+    tmpbuf2i = NULL;  _tmpbuf2i = NULL;  outbufi = NULL;
+  }
+
+  if ((this->init & COMPRESS) == 0)
+    THROW("tjEncodeYUVPlanes(): Instance has not been initialized for compression");
+
+  if (srcBuf == NULL || width <= 0 || pitch < 0 || height <= 0 ||
+      pixelFormat < 0 || pixelFormat >= TJ_NUMPF || !dstPlanes ||
+      !dstPlanes0 || subsamp < 0 || subsamp >= NUMSUBOPT)
+    THROW("tjEncodeYUVPlanes(): Invalid argument");
+  if (subsamp != TJSAMP_GRAY && (!dstPlanes1 || !dstPlanes2))
+    THROW("tjEncodeYUVPlanes(): Invalid argument");
+
+  if (pixelFormat == TJPF_CMYK)
+    THROW("tjEncodeYUVPlanes(): Cannot generate YUV images from CMYK pixels");
+
+  if (pitch == 0) pitch = width * tjPixelSizepixelFormat;
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  cinfo->image_width = width;
+  cinfo->image_height = height;
+
+#ifndef NO_PUTENV
+  if (flags & TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+  else if (flags & TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+  else if (flags & TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+#endif
+
+  setCompDefaults(cinfo, pixelFormat, subsamp, -1, flags);
+
+  /* Execute only the parts of jpeg_start_compress() that we need.  If we
+     were to call the whole jpeg_start_compress() function, then it would try
+     to write the file headers, which could overflow the output buffer if the
+     YUV image were very small. */
+  if (cinfo->global_state != CSTATE_START)
+    THROW("tjEncodeYUVPlanes(): libjpeg API is in the wrong state");
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr)cinfo);
+  jinit_c_master_control(cinfo, FALSE);
+  jinit_color_converter(cinfo);
+  jinit_downsampler(cinfo);
+  (*cinfo->cconvert->start_pass) (cinfo);
+
+  pw0 = PAD(width, cinfo->max_h_samp_factor);
+  ph0 = PAD(height, cinfo->max_v_samp_factor);
+
+  if ((row_pointer = (JSAMPROW *)malloc(sizeof(JSAMPROW) * ph0)) == NULL)
+    THROW("tjEncodeYUVPlanes(): Memory allocation failure");
+  for (i = 0; i < height; i++) {
+    if (flags & TJFLAG_BOTTOMUP)
+      row_pointeri = (JSAMPROW)&srcBuf(height - i - 1) * (size_t)pitch;
+    else
+      row_pointeri = (JSAMPROW)&srcBufi * (size_t)pitch;
+  }
+  if (height < ph0)
+    for (i = height; i < ph0; i++) row_pointeri = row_pointerheight - 1;
+
+  for (i = 0; i < cinfo->num_components; i++) {
+    compptr = &cinfo->comp_infoi;
+    _tmpbufi = (JSAMPLE *)malloc(
+      PAD((compptr->width_in_blocks * cinfo->max_h_samp_factor * DCTSIZE) /
+          compptr->h_samp_factor, 32) *
+      cinfo->max_v_samp_factor + 32);
+    if (!_tmpbufi)
+      THROW("tjEncodeYUVPlanes(): Memory allocation failure");
+    tmpbufi =
+      (JSAMPROW *)malloc(sizeof(JSAMPROW) * cinfo->max_v_samp_factor);
+    if (!tmpbufi)
+      THROW("tjEncodeYUVPlanes(): Memory allocation failure");
+    for (row = 0; row < cinfo->max_v_samp_factor; row++) {
+      unsigned char *_tmpbuf_aligned =
+        (unsigned char *)PAD((JUINTPTR)_tmpbufi, 32);
+
+      tmpbufirow = &_tmpbuf_aligned
+        PAD((compptr->width_in_blocks * cinfo->max_h_samp_factor * DCTSIZE) /
+            compptr->h_samp_factor, 32) * row;
+    }
+    _tmpbuf2i =
+      (JSAMPLE *)malloc(PAD(compptr->width_in_blocks * DCTSIZE, 32) *
+                        compptr->v_samp_factor + 32);
+    if (!_tmpbuf2i)
+      THROW("tjEncodeYUVPlanes(): Memory allocation failure");
+    tmpbuf2i = (JSAMPROW *)malloc(sizeof(JSAMPROW) * compptr->v_samp_factor);
+    if (!tmpbuf2i)
+      THROW("tjEncodeYUVPlanes(): Memory allocation failure");
+    for (row = 0; row < compptr->v_samp_factor; row++) {
+      unsigned char *_tmpbuf2_aligned =
+        (unsigned char *)PAD((JUINTPTR)_tmpbuf2i, 32);
+
+      tmpbuf2irow =
+        &_tmpbuf2_alignedPAD(compptr->width_in_blocks * DCTSIZE, 32) * row;
+    }
+    pwi = pw0 * compptr->h_samp_factor / cinfo->max_h_samp_factor;
+    phi = ph0 * compptr->v_samp_factor / cinfo->max_v_samp_factor;
+    outbufi = (JSAMPROW *)malloc(sizeof(JSAMPROW) * phi);
+    if (!outbufi)
+      THROW("tjEncodeYUVPlanes(): Memory allocation failure");
+    ptr = dstPlanesi;
+    for (row = 0; row < phi; row++) {
+      outbufirow = ptr;
+      ptr += (strides && stridesi != 0) ? stridesi : pwi;
+    }
+  }
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  for (row = 0; row < ph0; row += cinfo->max_v_samp_factor) {
+    (*cinfo->cconvert->color_convert) (cinfo, &row_pointerrow, tmpbuf, 0,
+                                       cinfo->max_v_samp_factor);
+    (cinfo->downsample->downsample) (cinfo, tmpbuf, 0, tmpbuf2, 0);
+    for (i = 0, compptr = cinfo->comp_info; i < cinfo->num_components;
+         i++, compptr++)
+      jcopy_sample_rows(tmpbuf2i, 0, outbufi,
+        row * compptr->v_samp_factor / cinfo->max_v_samp_factor,
+        compptr->v_samp_factor, pwi);
+  }
+  cinfo->next_scanline += height;
+  jpeg_abort_compress(cinfo);
+
+bailout:
+  if (cinfo->global_state > CSTATE_START) jpeg_abort_compress(cinfo);
+  free(row_pointer);
+  for (i = 0; i < MAX_COMPONENTS; i++) {
+    free(tmpbufi);
+    free(_tmpbufi);
+    free(tmpbuf2i);
+    free(_tmpbuf2i);
+    free(outbufi);
+  }
+  if (this->jerr.warning) retval = -1;
+  this->jerr.stopOnWarning = FALSE;
+  return retval;
+}
+
+DLLEXPORT int tjEncodeYUV3(tjhandle handle, const unsigned char *srcBuf,
+                           int width, int pitch, int height, int pixelFormat,
+                           unsigned char *dstBuf, int pad, int subsamp,
+                           int flags)
+{
+  unsigned char *dstPlanes3;
+  int pw0, ph0, strides3, retval = -1;
+  tjinstance *this = (tjinstance *)handle;
+
+  if (!this) THROWG("tjEncodeYUV3(): Invalid handle");
+  this->isInstanceError = FALSE;
+
+  if (width <= 0 || height <= 0 || dstBuf == NULL || pad < 0 ||
+      !IS_POW2(pad) || subsamp < 0 || subsamp >= NUMSUBOPT)
+    THROW("tjEncodeYUV3(): Invalid argument");
+
+  pw0 = tjPlaneWidth(0, width, subsamp);
+  ph0 = tjPlaneHeight(0, height, subsamp);
+  dstPlanes0 = dstBuf;
+  strides0 = PAD(pw0, pad);
+  if (subsamp == TJSAMP_GRAY) {
+    strides1 = strides2 = 0;
+    dstPlanes1 = dstPlanes2 = NULL;
+  } else {
+    int pw1 = tjPlaneWidth(1, width, subsamp);
+    int ph1 = tjPlaneHeight(1, height, subsamp);
+
+    strides1 = strides2 = PAD(pw1, pad);
+    dstPlanes1 = dstPlanes0 + strides0 * ph0;
+    dstPlanes2 = dstPlanes1 + strides1 * ph1;
+  }
+
+  return tjEncodeYUVPlanes(handle, srcBuf, width, pitch, height, pixelFormat,
+                           dstPlanes, strides, subsamp, flags);
+
+bailout:
+  return retval;
+}
+
+DLLEXPORT int tjEncodeYUV2(tjhandle handle, unsigned char *srcBuf, int width,
+                           int pitch, int height, int pixelFormat,
+                           unsigned char *dstBuf, int subsamp, int flags)
+{
+  return tjEncodeYUV3(handle, srcBuf, width, pitch, height, pixelFormat,
+                      dstBuf, 4, subsamp, flags);
+}
+
+DLLEXPORT int tjEncodeYUV(tjhandle handle, unsigned char *srcBuf, int width,
+                          int pitch, int height, int pixelSize,
+                          unsigned char *dstBuf, int subsamp, int flags)
+{
+  return tjEncodeYUV2(handle, srcBuf, width, pitch, height,
+                      getPixelFormat(pixelSize, flags), dstBuf, subsamp,
+                      flags);
+}
+
+
+DLLEXPORT int tjCompressFromYUVPlanes(tjhandle handle,
+                                      const unsigned char **srcPlanes,
+                                      int width, const int *strides,
+                                      int height, int subsamp,
+                                      unsigned char **jpegBuf,
+                                      unsigned long *jpegSize, int jpegQual,
+                                      int flags)
+{
+  int i, row, retval = 0, alloc = 1;
+  int pwMAX_COMPONENTS, phMAX_COMPONENTS, iwMAX_COMPONENTS,
+    tmpbufsize = 0, usetmpbuf = 0, thMAX_COMPONENTS;
+  JSAMPLE *_tmpbuf = NULL, *ptr;
+  JSAMPROW *inbufMAX_COMPONENTS, *tmpbufMAX_COMPONENTS;
+
+  GET_CINSTANCE(handle)
+  this->jerr.stopOnWarning = (flags & TJFLAG_STOPONWARNING) ? TRUE : FALSE;
+
+  for (i = 0; i < MAX_COMPONENTS; i++) {
+    tmpbufi = NULL;  inbufi = NULL;
+  }
+
+  if ((this->init & COMPRESS) == 0)
+    THROW("tjCompressFromYUVPlanes(): Instance has not been initialized for compression");
+
+  if (!srcPlanes || !srcPlanes0 || width <= 0 || height <= 0 ||
+      subsamp < 0 || subsamp >= NUMSUBOPT || jpegBuf == NULL ||
+      jpegSize == NULL || jpegQual < 0 || jpegQual > 100)
+    THROW("tjCompressFromYUVPlanes(): Invalid argument");
+  if (subsamp != TJSAMP_GRAY && (!srcPlanes1 || !srcPlanes2))
+    THROW("tjCompressFromYUVPlanes(): Invalid argument");
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  cinfo->image_width = width;
+  cinfo->image_height = height;
+
+#ifndef NO_PUTENV
+  if (flags & TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+  else if (flags & TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+  else if (flags & TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+#endif
+
+  if (flags & TJFLAG_NOREALLOC) {
+    alloc = 0;  *jpegSize = tjBufSize(width, height, subsamp);
+  }
+  jpeg_mem_dest_tj(cinfo, jpegBuf, jpegSize, alloc);
+  setCompDefaults(cinfo, TJPF_RGB, subsamp, jpegQual, flags);
+  cinfo->raw_data_in = TRUE;
+
+  jpeg_start_compress(cinfo, TRUE);
+  for (i = 0; i < cinfo->num_components; i++) {
+    jpeg_component_info *compptr = &cinfo->comp_infoi;
+    int ih;
+
+    iwi = compptr->width_in_blocks * DCTSIZE;
+    ih = compptr->height_in_blocks * DCTSIZE;
+    pwi = PAD(cinfo->image_width, cinfo->max_h_samp_factor) *
+            compptr->h_samp_factor / cinfo->max_h_samp_factor;
+    phi = PAD(cinfo->image_height, cinfo->max_v_samp_factor) *
+            compptr->v_samp_factor / cinfo->max_v_samp_factor;
+    if (iwi != pwi || ih != phi) usetmpbuf = 1;
+    thi = compptr->v_samp_factor * DCTSIZE;
+    tmpbufsize += iwi * thi;
+    if ((inbufi = (JSAMPROW *)malloc(sizeof(JSAMPROW) * phi)) == NULL)
+      THROW("tjCompressFromYUVPlanes(): Memory allocation failure");
+    ptr = (JSAMPLE *)srcPlanesi;
+    for (row = 0; row < phi; row++) {
+      inbufirow = ptr;
+      ptr += (strides && stridesi != 0) ? stridesi : pwi;
+    }
+  }
+  if (usetmpbuf) {
+    if ((_tmpbuf = (JSAMPLE *)malloc(sizeof(JSAMPLE) * tmpbufsize)) == NULL)
+      THROW("tjCompressFromYUVPlanes(): Memory allocation failure");
+    ptr = _tmpbuf;
+    for (i = 0; i < cinfo->num_components; i++) {
+      if ((tmpbufi = (JSAMPROW *)malloc(sizeof(JSAMPROW) * thi)) == NULL)
+        THROW("tjCompressFromYUVPlanes(): Memory allocation failure");
+      for (row = 0; row < thi; row++) {
+        tmpbufirow = ptr;
+        ptr += iwi;
+      }
+    }
+  }
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  for (row = 0; row < (int)cinfo->image_height;
+       row += cinfo->max_v_samp_factor * DCTSIZE) {
+    JSAMPARRAY yuvptrMAX_COMPONENTS;
+    int crowMAX_COMPONENTS;
+
+    for (i = 0; i < cinfo->num_components; i++) {
+      jpeg_component_info *compptr = &cinfo->comp_infoi;
+
+      crowi = row * compptr->v_samp_factor / cinfo->max_v_samp_factor;
+      if (usetmpbuf) {
+        int j, k;
+
+        for (j = 0; j < MIN(thi, phi - crowi); j++) {
+          memcpy(tmpbufij, inbuficrowi + j, pwi);
+          /* Duplicate last sample in row to fill out MCU */
+          for (k = pwi; k < iwi; k++)
+            tmpbufijk = tmpbufijpwi - 1;
+        }
+        /* Duplicate last row to fill out MCU */
+        for (j = phi - crowi; j < thi; j++)
+          memcpy(tmpbufij, tmpbufiphi - crowi - 1, iwi);
+        yuvptri = tmpbufi;
+      } else
+        yuvptri = &inbuficrowi;
+    }
+    jpeg_write_raw_data(cinfo, yuvptr, cinfo->max_v_samp_factor * DCTSIZE);
+  }
+  jpeg_finish_compress(cinfo);
+
+bailout:
+  if (cinfo->global_state > CSTATE_START) {
+    if (alloc) (*cinfo->dest->term_destination) (cinfo);
+    jpeg_abort_compress(cinfo);
+  }
+  for (i = 0; i < MAX_COMPONENTS; i++) {
+    free(tmpbufi);
+    free(inbufi);
+  }
+  free(_tmpbuf);
+  if (this->jerr.warning) retval = -1;
+  this->jerr.stopOnWarning = FALSE;
+  return retval;
+}
+
+DLLEXPORT int tjCompressFromYUV(tjhandle handle, const unsigned char *srcBuf,
+                                int width, int pad, int height, int subsamp,
+                                unsigned char **jpegBuf,
+                                unsigned long *jpegSize, int jpegQual,
+                                int flags)
+{
+  const unsigned char *srcPlanes3;
+  int pw0, ph0, strides3, retval = -1;
+  tjinstance *this = (tjinstance *)handle;
+
+  if (!this) THROWG("tjCompressFromYUV(): Invalid handle");
+  this->isInstanceError = FALSE;
+
+  if (srcBuf == NULL || width <= 0 || pad < 1 || height <= 0 || subsamp < 0 ||
+      subsamp >= NUMSUBOPT)
+    THROW("tjCompressFromYUV(): Invalid argument");
+
+  pw0 = tjPlaneWidth(0, width, subsamp);
+  ph0 = tjPlaneHeight(0, height, subsamp);
+  srcPlanes0 = srcBuf;
+  strides0 = PAD(pw0, pad);
+  if (subsamp == TJSAMP_GRAY) {
+    strides1 = strides2 = 0;
+    srcPlanes1 = srcPlanes2 = NULL;
+  } else {
+    int pw1 = tjPlaneWidth(1, width, subsamp);
+    int ph1 = tjPlaneHeight(1, height, subsamp);
+
+    strides1 = strides2 = PAD(pw1, pad);
+    srcPlanes1 = srcPlanes0 + strides0 * ph0;
+    srcPlanes2 = srcPlanes1 + strides1 * ph1;
+  }
+
+  return tjCompressFromYUVPlanes(handle, srcPlanes, width, strides, height,
+                                 subsamp, jpegBuf, jpegSize, jpegQual, flags);
+
+bailout:
+  return retval;
+}
+
+
+/* Decompressor */
+
+static tjhandle _tjInitDecompress(tjinstance *this)
+{
+  static unsigned char buffer1;
+
+  /* This is also straight out of example.txt */
+  this->dinfo.err = jpeg_std_error(&this->jerr.pub);
+  this->jerr.pub.error_exit = my_error_exit;
+  this->jerr.pub.output_message = my_output_message;
+  this->jerr.emit_message = this->jerr.pub.emit_message;
+  this->jerr.pub.emit_message = my_emit_message;
+  this->jerr.pub.addon_message_table = turbojpeg_message_table;
+  this->jerr.pub.first_addon_message = JMSG_FIRSTADDONCODE;
+  this->jerr.pub.last_addon_message = JMSG_LASTADDONCODE;
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    free(this);
+    return NULL;
+  }
+
+  jpeg_create_decompress(&this->dinfo);
+  /* Make an initial call so it will create the source manager */
+  jpeg_mem_src_tj(&this->dinfo, buffer, 1);
+
+  this->init |= DECOMPRESS;
+  return (tjhandle)this;
+}
+
+DLLEXPORT tjhandle tjInitDecompress(void)
+{
+  tjinstance *this;
+
+  if ((this = (tjinstance *)malloc(sizeof(tjinstance))) == NULL) {
+    snprintf(errStr, JMSG_LENGTH_MAX,
+             "tjInitDecompress(): Memory allocation failure");
+    return NULL;
+  }
+  MEMZERO(this, sizeof(tjinstance));
+  snprintf(this->errStr, JMSG_LENGTH_MAX, "No error");
+  return _tjInitDecompress(this);
+}
+
+
+DLLEXPORT int tjDecompressHeader3(tjhandle handle,
+                                  const unsigned char *jpegBuf,
+                                  unsigned long jpegSize, int *width,
+                                  int *height, int *jpegSubsamp,
+                                  int *jpegColorspace)
+{
+  int retval = 0;
+
+  GET_DINSTANCE(handle);
+  if ((this->init & DECOMPRESS) == 0)
+    THROW("tjDecompressHeader3(): Instance has not been initialized for decompression");
+
+  if (jpegBuf == NULL || jpegSize <= 0 || width == NULL || height == NULL ||
+      jpegSubsamp == NULL || jpegColorspace == NULL)
+    THROW("tjDecompressHeader3(): Invalid argument");
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    return -1;
+  }
+
+  jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+  jpeg_read_header(dinfo, TRUE);
+
+  *width = dinfo->image_width;
+  *height = dinfo->image_height;
+  *jpegSubsamp = getSubsamp(dinfo);
+  switch (dinfo->jpeg_color_space) {
+  case JCS_GRAYSCALE:  *jpegColorspace = TJCS_GRAY;  break;
+  case JCS_RGB:        *jpegColorspace = TJCS_RGB;  break;
+  case JCS_YCbCr:      *jpegColorspace = TJCS_YCbCr;  break;
+  case JCS_CMYK:       *jpegColorspace = TJCS_CMYK;  break;
+  case JCS_YCCK:       *jpegColorspace = TJCS_YCCK;  break;
+  default:             *jpegColorspace = -1;  break;
+  }
+
+  jpeg_abort_decompress(dinfo);
+
+  if (*jpegSubsamp < 0)
+    THROW("tjDecompressHeader3(): Could not determine subsampling type for JPEG image");
+  if (*jpegColorspace < 0)
+    THROW("tjDecompressHeader3(): Could not determine colorspace of JPEG image");
+  if (*width < 1 || *height < 1)
+    THROW("tjDecompressHeader3(): Invalid data returned in header");
+
+bailout:
+  if (this->jerr.warning) retval = -1;
+  return retval;
+}
+
+DLLEXPORT int tjDecompressHeader2(tjhandle handle, unsigned char *jpegBuf,
+                                  unsigned long jpegSize, int *width,
+                                  int *height, int *jpegSubsamp)
+{
+  int jpegColorspace;
+
+  return tjDecompressHeader3(handle, jpegBuf, jpegSize, width, height,
+                             jpegSubsamp, &jpegColorspace);
+}
+
+DLLEXPORT int tjDecompressHeader(tjhandle handle, unsigned char *jpegBuf,
+                                 unsigned long jpegSize, int *width,
+                                 int *height)
+{
+  int jpegSubsamp;
+
+  return tjDecompressHeader2(handle, jpegBuf, jpegSize, width, height,
+                             &jpegSubsamp);
+}
+
+
+DLLEXPORT tjscalingfactor *tjGetScalingFactors(int *numscalingfactors)
+{
+  if (numscalingfactors == NULL) {
+    snprintf(errStr, JMSG_LENGTH_MAX,
+             "tjGetScalingFactors(): Invalid argument");
+    return NULL;
+  }
+
+  *numscalingfactors = NUMSF;
+  return (tjscalingfactor *)sf;
+}
+
+
+DLLEXPORT int tjDecompress2(tjhandle handle, const unsigned char *jpegBuf,
+                            unsigned long jpegSize, unsigned char *dstBuf,
+                            int width, int pitch, int height, int pixelFormat,
+                            int flags)
+{
+  JSAMPROW *row_pointer = NULL;
+  int i, retval = 0, jpegwidth, jpegheight, scaledw, scaledh;
+  struct my_progress_mgr progress;
+
+  GET_DINSTANCE(handle);
+  this->jerr.stopOnWarning = (flags & TJFLAG_STOPONWARNING) ? TRUE : FALSE;
+  if ((this->init & DECOMPRESS) == 0)
+    THROW("tjDecompress2(): Instance has not been initialized for decompression");
+
+  if (jpegBuf == NULL || jpegSize <= 0 || dstBuf == NULL || width < 0 ||
+      pitch < 0 || height < 0 || pixelFormat < 0 || pixelFormat >= TJ_NUMPF)
+    THROW("tjDecompress2(): Invalid argument");
+
+#ifndef NO_PUTENV
+  if (flags & TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+  else if (flags & TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+  else if (flags & TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+#endif
+
+  if (flags & TJFLAG_LIMITSCANS) {
+    MEMZERO(&progress, sizeof(struct my_progress_mgr));
+    progress.pub.progress_monitor = my_progress_monitor;
+    progress.this = this;
+    dinfo->progress = &progress.pub;
+  } else
+    dinfo->progress = NULL;
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+  jpeg_read_header(dinfo, TRUE);
+  this->dinfo.out_color_space = pf2cspixelFormat;
+  if (flags & TJFLAG_FASTDCT) this->dinfo.dct_method = JDCT_FASTEST;
+  if (flags & TJFLAG_FASTUPSAMPLE) dinfo->do_fancy_upsampling = FALSE;
+
+  jpegwidth = dinfo->image_width;  jpegheight = dinfo->image_height;
+  if (width == 0) width = jpegwidth;
+  if (height == 0) height = jpegheight;
+  for (i = 0; i < NUMSF; i++) {
+    scaledw = TJSCALED(jpegwidth, sfi);
+    scaledh = TJSCALED(jpegheight, sfi);
+    if (scaledw <= width && scaledh <= height)
+      break;
+  }
+  if (i >= NUMSF)
+    THROW("tjDecompress2(): Could not scale down to desired image dimensions");
+  width = scaledw;  height = scaledh;
+  dinfo->scale_num = sfi.num;
+  dinfo->scale_denom = sfi.denom;
+
+  jpeg_start_decompress(dinfo);
+  if (pitch == 0) pitch = dinfo->output_width * tjPixelSizepixelFormat;
+
+  if ((row_pointer =
+       (JSAMPROW *)malloc(sizeof(JSAMPROW) * dinfo->output_height)) == NULL)
+    THROW("tjDecompress2(): Memory allocation failure");
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+  for (i = 0; i < (int)dinfo->output_height; i++) {
+    if (flags & TJFLAG_BOTTOMUP)
+      row_pointeri = &dstBuf(dinfo->output_height - i - 1) * (size_t)pitch;
+    else
+      row_pointeri = &dstBufi * (size_t)pitch;
+  }
+  while (dinfo->output_scanline < dinfo->output_height)
+    jpeg_read_scanlines(dinfo, &row_pointerdinfo->output_scanline,
+                        dinfo->output_height - dinfo->output_scanline);
+  jpeg_finish_decompress(dinfo);
+
+bailout:
+  if (dinfo->global_state > DSTATE_START) jpeg_abort_decompress(dinfo);
+  free(row_pointer);
+  if (this->jerr.warning) retval = -1;
+  this->jerr.stopOnWarning = FALSE;
+  return retval;
+}
+
+DLLEXPORT int tjDecompress(tjhandle handle, unsigned char *jpegBuf,
+                           unsigned long jpegSize, unsigned char *dstBuf,
+                           int width, int pitch, int height, int pixelSize,
+                           int flags)
+{
+  if (flags & TJ_YUV)
+    return tjDecompressToYUV(handle, jpegBuf, jpegSize, dstBuf, flags);
+  else
+    return tjDecompress2(handle, jpegBuf, jpegSize, dstBuf, width, pitch,
+                         height, getPixelFormat(pixelSize, flags), flags);
+}
+
+
+static int setDecodeDefaults(struct jpeg_decompress_struct *dinfo,
+                             int pixelFormat, int subsamp, int flags)
+{
+  int i;
+
+  dinfo->scale_num = dinfo->scale_denom = 1;
+
+  if (subsamp == TJSAMP_GRAY) {
+    dinfo->num_components = dinfo->comps_in_scan = 1;
+    dinfo->jpeg_color_space = JCS_GRAYSCALE;
+  } else {
+    dinfo->num_components = dinfo->comps_in_scan = 3;
+    dinfo->jpeg_color_space = JCS_YCbCr;
+  }
+
+  dinfo->comp_info = (jpeg_component_info *)
+    (*dinfo->mem->alloc_small) ((j_common_ptr)dinfo, JPOOL_IMAGE,
+                                dinfo->num_components *
+                                sizeof(jpeg_component_info));
+
+  for (i = 0; i < dinfo->num_components; i++) {
+    jpeg_component_info *compptr = &dinfo->comp_infoi;
+
+    compptr->h_samp_factor = (i == 0) ? tjMCUWidthsubsamp / 8 : 1;
+    compptr->v_samp_factor = (i == 0) ? tjMCUHeightsubsamp / 8 : 1;
+    compptr->component_index = i;
+    compptr->component_id = i + 1;
+    compptr->quant_tbl_no = compptr->dc_tbl_no =
+      compptr->ac_tbl_no = (i == 0) ? 0 : 1;
+    dinfo->cur_comp_infoi = compptr;
+  }
+  dinfo->data_precision = 8;
+  for (i = 0; i < 2; i++) {
+    if (dinfo->quant_tbl_ptrsi == NULL)
+      dinfo->quant_tbl_ptrsi = jpeg_alloc_quant_table((j_common_ptr)dinfo);
+  }
+
+  return 0;
+}
+
+
+static int my_read_markers(j_decompress_ptr dinfo)
+{
+  return JPEG_REACHED_SOS;
+}
+
+static void my_reset_marker_reader(j_decompress_ptr dinfo)
+{
+}
+
+DLLEXPORT int tjDecodeYUVPlanes(tjhandle handle,
+                                const unsigned char **srcPlanes,
+                                const int *strides, int subsamp,
+                                unsigned char *dstBuf, int width, int pitch,
+                                int height, int pixelFormat, int flags)
+{
+  JSAMPROW *row_pointer = NULL;
+  JSAMPLE *_tmpbufMAX_COMPONENTS;
+  JSAMPROW *tmpbufMAX_COMPONENTS, *inbufMAX_COMPONENTS;
+  int i, retval = 0, row, pw0, ph0, pwMAX_COMPONENTS, phMAX_COMPONENTS;
+  JSAMPLE *ptr;
+  jpeg_component_info *compptr;
+  int (*old_read_markers) (j_decompress_ptr);
+  void (*old_reset_marker_reader) (j_decompress_ptr);
+
+  GET_DINSTANCE(handle);
+  this->jerr.stopOnWarning = (flags & TJFLAG_STOPONWARNING) ? TRUE : FALSE;
+
+  for (i = 0; i < MAX_COMPONENTS; i++) {
+    tmpbufi = NULL;  _tmpbufi = NULL;  inbufi = NULL;
+  }
+
+  if ((this->init & DECOMPRESS) == 0)
+    THROW("tjDecodeYUVPlanes(): Instance has not been initialized for decompression");
+
+  if (!srcPlanes || !srcPlanes0 || subsamp < 0 || subsamp >= NUMSUBOPT ||
+      dstBuf == NULL || width <= 0 || pitch < 0 || height <= 0 ||
+      pixelFormat < 0 || pixelFormat >= TJ_NUMPF)
+    THROW("tjDecodeYUVPlanes(): Invalid argument");
+  if (subsamp != TJSAMP_GRAY && (!srcPlanes1 || !srcPlanes2))
+    THROW("tjDecodeYUVPlanes(): Invalid argument");
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  if (pixelFormat == TJPF_CMYK)
+    THROW("tjDecodeYUVPlanes(): Cannot decode YUV images into CMYK pixels.");
+
+  if (pitch == 0) pitch = width * tjPixelSizepixelFormat;
+  dinfo->image_width = width;
+  dinfo->image_height = height;
+
+#ifndef NO_PUTENV
+  if (flags & TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+  else if (flags & TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+  else if (flags & TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+#endif
+
+  dinfo->progressive_mode = dinfo->inputctl->has_multiple_scans = FALSE;
+  dinfo->Ss = dinfo->Ah = dinfo->Al = 0;
+  dinfo->Se = DCTSIZE2 - 1;
+  if (setDecodeDefaults(dinfo, pixelFormat, subsamp, flags) == -1) {
+    retval = -1;  goto bailout;
+  }
+  old_read_markers = dinfo->marker->read_markers;
+  dinfo->marker->read_markers = my_read_markers;
+  old_reset_marker_reader = dinfo->marker->reset_marker_reader;
+  dinfo->marker->reset_marker_reader = my_reset_marker_reader;
+  jpeg_read_header(dinfo, TRUE);
+  dinfo->marker->read_markers = old_read_markers;
+  dinfo->marker->reset_marker_reader = old_reset_marker_reader;
+
+  this->dinfo.out_color_space = pf2cspixelFormat;
+  if (flags & TJFLAG_FASTDCT) this->dinfo.dct_method = JDCT_FASTEST;
+  dinfo->do_fancy_upsampling = FALSE;
+  dinfo->Se = DCTSIZE2 - 1;
+  jinit_master_decompress(dinfo);
+  (*dinfo->upsample->start_pass) (dinfo);
+
+  pw0 = PAD(width, dinfo->max_h_samp_factor);
+  ph0 = PAD(height, dinfo->max_v_samp_factor);
+
+  if (pitch == 0) pitch = dinfo->output_width * tjPixelSizepixelFormat;
+
+  if ((row_pointer = (JSAMPROW *)malloc(sizeof(JSAMPROW) * ph0)) == NULL)
+    THROW("tjDecodeYUVPlanes(): Memory allocation failure");
+  for (i = 0; i < height; i++) {
+    if (flags & TJFLAG_BOTTOMUP)
+      row_pointeri = &dstBuf(height - i - 1) * (size_t)pitch;
+    else
+      row_pointeri = &dstBufi * (size_t)pitch;
+  }
+  if (height < ph0)
+    for (i = height; i < ph0; i++) row_pointeri = row_pointerheight - 1;
+
+  for (i = 0; i < dinfo->num_components; i++) {
+    compptr = &dinfo->comp_infoi;
+    _tmpbufi =
+      (JSAMPLE *)malloc(PAD(compptr->width_in_blocks * DCTSIZE, 32) *
+                        compptr->v_samp_factor + 32);
+    if (!_tmpbufi)
+      THROW("tjDecodeYUVPlanes(): Memory allocation failure");
+    tmpbufi = (JSAMPROW *)malloc(sizeof(JSAMPROW) * compptr->v_samp_factor);
+    if (!tmpbufi)
+      THROW("tjDecodeYUVPlanes(): Memory allocation failure");
+    for (row = 0; row < compptr->v_samp_factor; row++) {
+      unsigned char *_tmpbuf_aligned =
+        (unsigned char *)PAD((JUINTPTR)_tmpbufi, 32);
+
+      tmpbufirow =
+        &_tmpbuf_alignedPAD(compptr->width_in_blocks * DCTSIZE, 32) * row;
+    }
+    pwi = pw0 * compptr->h_samp_factor / dinfo->max_h_samp_factor;
+    phi = ph0 * compptr->v_samp_factor / dinfo->max_v_samp_factor;
+    inbufi = (JSAMPROW *)malloc(sizeof(JSAMPROW) * phi);
+    if (!inbufi)
+      THROW("tjDecodeYUVPlanes(): Memory allocation failure");
+    ptr = (JSAMPLE *)srcPlanesi;
+    for (row = 0; row < phi; row++) {
+      inbufirow = ptr;
+      ptr += (strides && stridesi != 0) ? stridesi : pwi;
+    }
+  }
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  for (row = 0; row < ph0; row += dinfo->max_v_samp_factor) {
+    JDIMENSION inrow = 0, outrow = 0;
+
+    for (i = 0, compptr = dinfo->comp_info; i < dinfo->num_components;
+         i++, compptr++)
+      jcopy_sample_rows(inbufi,
+        row * compptr->v_samp_factor / dinfo->max_v_samp_factor, tmpbufi, 0,
+        compptr->v_samp_factor, pwi);
+    (dinfo->upsample->upsample) (dinfo, tmpbuf, &inrow,
+                                 dinfo->max_v_samp_factor, &row_pointerrow,
+                                 &outrow, dinfo->max_v_samp_factor);
+  }
+  jpeg_abort_decompress(dinfo);
+
+bailout:
+  if (dinfo->global_state > DSTATE_START) jpeg_abort_decompress(dinfo);
+  free(row_pointer);
+  for (i = 0; i < MAX_COMPONENTS; i++) {
+    free(tmpbufi);
+    free(_tmpbufi);
+    free(inbufi);
+  }
+  if (this->jerr.warning) retval = -1;
+  this->jerr.stopOnWarning = FALSE;
+  return retval;
+}
+
+DLLEXPORT int tjDecodeYUV(tjhandle handle, const unsigned char *srcBuf,
+                          int pad, int subsamp, unsigned char *dstBuf,
+                          int width, int pitch, int height, int pixelFormat,
+                          int flags)
+{
+  const unsigned char *srcPlanes3;
+  int pw0, ph0, strides3, retval = -1;
+  tjinstance *this = (tjinstance *)handle;
+
+  if (!this) THROWG("tjDecodeYUV(): Invalid handle");
+  this->isInstanceError = FALSE;
+
+  if (srcBuf == NULL || pad < 0 || !IS_POW2(pad) || subsamp < 0 ||
+      subsamp >= NUMSUBOPT || width <= 0 || height <= 0)
+    THROW("tjDecodeYUV(): Invalid argument");
+
+  pw0 = tjPlaneWidth(0, width, subsamp);
+  ph0 = tjPlaneHeight(0, height, subsamp);
+  srcPlanes0 = srcBuf;
+  strides0 = PAD(pw0, pad);
+  if (subsamp == TJSAMP_GRAY) {
+    strides1 = strides2 = 0;
+    srcPlanes1 = srcPlanes2 = NULL;
+  } else {
+    int pw1 = tjPlaneWidth(1, width, subsamp);
+    int ph1 = tjPlaneHeight(1, height, subsamp);
+
+    strides1 = strides2 = PAD(pw1, pad);
+    srcPlanes1 = srcPlanes0 + strides0 * ph0;
+    srcPlanes2 = srcPlanes1 + strides1 * ph1;
+  }
+
+  return tjDecodeYUVPlanes(handle, srcPlanes, strides, subsamp, dstBuf, width,
+                           pitch, height, pixelFormat, flags);
+
+bailout:
+  return retval;
+}
+
+DLLEXPORT int tjDecompressToYUVPlanes(tjhandle handle,
+                                      const unsigned char *jpegBuf,
+                                      unsigned long jpegSize,
+                                      unsigned char **dstPlanes, int width,
+                                      int *strides, int height, int flags)
+{
+  int i, sfi, row, retval = 0;
+  int jpegwidth, jpegheight, jpegSubsamp, scaledw, scaledh;
+  int pwMAX_COMPONENTS, phMAX_COMPONENTS, iwMAX_COMPONENTS,
+    tmpbufsize = 0, usetmpbuf = 0, thMAX_COMPONENTS;
+  JSAMPLE *_tmpbuf = NULL, *ptr;
+  JSAMPROW *outbufMAX_COMPONENTS, *tmpbufMAX_COMPONENTS;
+  int dctsize;
+  struct my_progress_mgr progress;
+
+  GET_DINSTANCE(handle);
+  this->jerr.stopOnWarning = (flags & TJFLAG_STOPONWARNING) ? TRUE : FALSE;
+
+  for (i = 0; i < MAX_COMPONENTS; i++) {
+    tmpbufi = NULL;  outbufi = NULL;
+  }
+
+  if ((this->init & DECOMPRESS) == 0)
+    THROW("tjDecompressToYUVPlanes(): Instance has not been initialized for decompression");
+
+  if (jpegBuf == NULL || jpegSize <= 0 || !dstPlanes || !dstPlanes0 ||
+      width < 0 || height < 0)
+    THROW("tjDecompressToYUVPlanes(): Invalid argument");
+
+#ifndef NO_PUTENV
+  if (flags & TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+  else if (flags & TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+  else if (flags & TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+#endif
+
+  if (flags & TJFLAG_LIMITSCANS) {
+    MEMZERO(&progress, sizeof(struct my_progress_mgr));
+    progress.pub.progress_monitor = my_progress_monitor;
+    progress.this = this;
+    dinfo->progress = &progress.pub;
+  } else
+    dinfo->progress = NULL;
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  if (!this->headerRead) {
+    jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+    jpeg_read_header(dinfo, TRUE);
+  }
+  this->headerRead = 0;
+  jpegSubsamp = getSubsamp(dinfo);
+  if (jpegSubsamp < 0)
+    THROW("tjDecompressToYUVPlanes(): Could not determine subsampling type for JPEG image");
+
+  if (jpegSubsamp != TJSAMP_GRAY && (!dstPlanes1 || !dstPlanes2))
+    THROW("tjDecompressToYUVPlanes(): Invalid argument");
+
+  jpegwidth = dinfo->image_width;  jpegheight = dinfo->image_height;
+  if (width == 0) width = jpegwidth;
+  if (height == 0) height = jpegheight;
+  for (i = 0; i < NUMSF; i++) {
+    scaledw = TJSCALED(jpegwidth, sfi);
+    scaledh = TJSCALED(jpegheight, sfi);
+    if (scaledw <= width && scaledh <= height)
+      break;
+  }
+  if (i >= NUMSF)
+    THROW("tjDecompressToYUVPlanes(): Could not scale down to desired image dimensions");
+  if (dinfo->num_components > 3)
+    THROW("tjDecompressToYUVPlanes(): JPEG image must have 3 or fewer components");
+
+  width = scaledw;  height = scaledh;
+  dinfo->scale_num = sfi.num;
+  dinfo->scale_denom = sfi.denom;
+  sfi = i;
+  jpeg_calc_output_dimensions(dinfo);
+
+  dctsize = DCTSIZE * sfsfi.num / sfsfi.denom;
+
+  for (i = 0; i < dinfo->num_components; i++) {
+    jpeg_component_info *compptr = &dinfo->comp_infoi;
+    int ih;
+
+    iwi = compptr->width_in_blocks * dctsize;
+    ih = compptr->height_in_blocks * dctsize;
+    pwi = tjPlaneWidth(i, dinfo->output_width, jpegSubsamp);
+    phi = tjPlaneHeight(i, dinfo->output_height, jpegSubsamp);
+    if (iwi != pwi || ih != phi) usetmpbuf = 1;
+    thi = compptr->v_samp_factor * dctsize;
+    tmpbufsize += iwi * thi;
+    if ((outbufi = (JSAMPROW *)malloc(sizeof(JSAMPROW) * phi)) == NULL)
+      THROW("tjDecompressToYUVPlanes(): Memory allocation failure");
+    ptr = dstPlanesi;
+    for (row = 0; row < phi; row++) {
+      outbufirow = ptr;
+      ptr += (strides && stridesi != 0) ? stridesi : pwi;
+    }
+  }
+  if (usetmpbuf) {
+    if ((_tmpbuf = (JSAMPLE *)malloc(sizeof(JSAMPLE) * tmpbufsize)) == NULL)
+      THROW("tjDecompressToYUVPlanes(): Memory allocation failure");
+    ptr = _tmpbuf;
+    for (i = 0; i < dinfo->num_components; i++) {
+      if ((tmpbufi = (JSAMPROW *)malloc(sizeof(JSAMPROW) * thi)) == NULL)
+        THROW("tjDecompressToYUVPlanes(): Memory allocation failure");
+      for (row = 0; row < thi; row++) {
+        tmpbufirow = ptr;
+        ptr += iwi;
+      }
+    }
+  }
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  if (flags & TJFLAG_FASTUPSAMPLE) dinfo->do_fancy_upsampling = FALSE;
+  if (flags & TJFLAG_FASTDCT) dinfo->dct_method = JDCT_FASTEST;
+  dinfo->raw_data_out = TRUE;
+
+  jpeg_start_decompress(dinfo);
+  for (row = 0; row < (int)dinfo->output_height;
+       row += dinfo->max_v_samp_factor * dinfo->_min_DCT_scaled_size) {
+    JSAMPARRAY yuvptrMAX_COMPONENTS;
+    int crowMAX_COMPONENTS;
+
+    for (i = 0; i < dinfo->num_components; i++) {
+      jpeg_component_info *compptr = &dinfo->comp_infoi;
+
+      if (jpegSubsamp == TJ_420) {
+        /* When 4:2:0 subsampling is used with IDCT scaling, libjpeg will try
+           to be clever and use the IDCT to perform upsampling on the U and V
+           planes.  For instance, if the output image is to be scaled by 1/2
+           relative to the JPEG image, then the scaling factor and upsampling
+           effectively cancel each other, so a normal 8x8 IDCT can be used.
+           However, this is not desirable when using the decompress-to-YUV
+           functionality in TurboJPEG, since we want to output the U and V
+           planes in their subsampled form.  Thus, we have to override some
+           internal libjpeg parameters to force it to use the "scaled" IDCT
+           functions on the U and V planes. */
+        compptr->_DCT_scaled_size = dctsize;
+        compptr->MCU_sample_width = tjMCUWidthjpegSubsamp *
+          sfsfi.num / sfsfi.denom *
+          compptr->v_samp_factor / dinfo->max_v_samp_factor;
+        dinfo->idct->inverse_DCTi = dinfo->idct->inverse_DCT0;
+      }
+      crowi = row * compptr->v_samp_factor / dinfo->max_v_samp_factor;
+      if (usetmpbuf) yuvptri = tmpbufi;
+      else yuvptri = &outbuficrowi;
+    }
+    jpeg_read_raw_data(dinfo, yuvptr,
+                       dinfo->max_v_samp_factor * dinfo->_min_DCT_scaled_size);
+    if (usetmpbuf) {
+      int j;
+
+      for (i = 0; i < dinfo->num_components; i++) {
+        for (j = 0; j < MIN(thi, phi - crowi); j++) {
+          memcpy(outbuficrowi + j, tmpbufij, pwi);
+        }
+      }
+    }
+  }
+  jpeg_finish_decompress(dinfo);
+
+bailout:
+  if (dinfo->global_state > DSTATE_START) jpeg_abort_decompress(dinfo);
+  for (i = 0; i < MAX_COMPONENTS; i++) {
+    free(tmpbufi);
+    free(outbufi);
+  }
+  free(_tmpbuf);
+  if (this->jerr.warning) retval = -1;
+  this->jerr.stopOnWarning = FALSE;
+  return retval;
+}
+
+DLLEXPORT int tjDecompressToYUV2(tjhandle handle, const unsigned char *jpegBuf,
+                                 unsigned long jpegSize, unsigned char *dstBuf,
+                                 int width, int pad, int height, int flags)
+{
+  unsigned char *dstPlanes3;
+  int pw0, ph0, strides3, retval = -1, jpegSubsamp = -1;
+  int i, jpegwidth, jpegheight, scaledw, scaledh;
+
+  GET_DINSTANCE(handle);
+  this->jerr.stopOnWarning = (flags & TJFLAG_STOPONWARNING) ? TRUE : FALSE;
+
+  if (jpegBuf == NULL || jpegSize <= 0 || dstBuf == NULL || width < 0 ||
+      pad < 1 || !IS_POW2(pad) || height < 0)
+    THROW("tjDecompressToYUV2(): Invalid argument");
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    return -1;
+  }
+
+  jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+  jpeg_read_header(dinfo, TRUE);
+  jpegSubsamp = getSubsamp(dinfo);
+  if (jpegSubsamp < 0)
+    THROW("tjDecompressToYUV2(): Could not determine subsampling type for JPEG image");
+
+  jpegwidth = dinfo->image_width;  jpegheight = dinfo->image_height;
+  if (width == 0) width = jpegwidth;
+  if (height == 0) height = jpegheight;
+
+  for (i = 0; i < NUMSF; i++) {
+    scaledw = TJSCALED(jpegwidth, sfi);
+    scaledh = TJSCALED(jpegheight, sfi);
+    if (scaledw <= width && scaledh <= height)
+      break;
+  }
+  if (i >= NUMSF)
+    THROW("tjDecompressToYUV2(): Could not scale down to desired image dimensions");
+
+  pw0 = tjPlaneWidth(0, width, jpegSubsamp);
+  ph0 = tjPlaneHeight(0, height, jpegSubsamp);
+  dstPlanes0 = dstBuf;
+  strides0 = PAD(pw0, pad);
+  if (jpegSubsamp == TJSAMP_GRAY) {
+    strides1 = strides2 = 0;
+    dstPlanes1 = dstPlanes2 = NULL;
+  } else {
+    int pw1 = tjPlaneWidth(1, width, jpegSubsamp);
+    int ph1 = tjPlaneHeight(1, height, jpegSubsamp);
+
+    strides1 = strides2 = PAD(pw1, pad);
+    dstPlanes1 = dstPlanes0 + strides0 * ph0;
+    dstPlanes2 = dstPlanes1 + strides1 * ph1;
+  }
+
+  this->headerRead = 1;
+  return tjDecompressToYUVPlanes(handle, jpegBuf, jpegSize, dstPlanes, width,
+                                 strides, height, flags);
+
+bailout:
+  this->jerr.stopOnWarning = FALSE;
+  return retval;
+}
+
+DLLEXPORT int tjDecompressToYUV(tjhandle handle, unsigned char *jpegBuf,
+                                unsigned long jpegSize, unsigned char *dstBuf,
+                                int flags)
+{
+  return tjDecompressToYUV2(handle, jpegBuf, jpegSize, dstBuf, 0, 4, 0, flags);
+}
+
+
+/* Transformer */
+
+DLLEXPORT tjhandle tjInitTransform(void)
+{
+  tjinstance *this = NULL;
+  tjhandle handle = NULL;
+
+  if ((this = (tjinstance *)malloc(sizeof(tjinstance))) == NULL) {
+    snprintf(errStr, JMSG_LENGTH_MAX,
+             "tjInitTransform(): Memory allocation failure");
+    return NULL;
+  }
+  MEMZERO(this, sizeof(tjinstance));
+  snprintf(this->errStr, JMSG_LENGTH_MAX, "No error");
+  handle = _tjInitCompress(this);
+  if (!handle) return NULL;
+  handle = _tjInitDecompress(this);
+  return handle;
+}
+
+
+DLLEXPORT int tjTransform(tjhandle handle, const unsigned char *jpegBuf,
+                          unsigned long jpegSize, int n,
+                          unsigned char **dstBufs, unsigned long *dstSizes,
+                          tjtransform *t, int flags)
+{
+  jpeg_transform_info *xinfo = NULL;
+  jvirt_barray_ptr *srccoefs, *dstcoefs;
+  int retval = 0, alloc = 1, i, jpegSubsamp, saveMarkers = 0;
+  struct my_progress_mgr progress;
+
+  GET_INSTANCE(handle);
+  this->jerr.stopOnWarning = (flags & TJFLAG_STOPONWARNING) ? TRUE : FALSE;
+  if ((this->init & COMPRESS) == 0 || (this->init & DECOMPRESS) == 0)
+    THROW("tjTransform(): Instance has not been initialized for transformation");
+
+  if (jpegBuf == NULL || jpegSize <= 0 || n < 1 || dstBufs == NULL ||
+      dstSizes == NULL || t == NULL || flags < 0)
+    THROW("tjTransform(): Invalid argument");
+
+#ifndef NO_PUTENV
+  if (flags & TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+  else if (flags & TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+  else if (flags & TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+#endif
+
+  if (flags & TJFLAG_LIMITSCANS) {
+    MEMZERO(&progress, sizeof(struct my_progress_mgr));
+    progress.pub.progress_monitor = my_progress_monitor;
+    progress.this = this;
+    dinfo->progress = &progress.pub;
+  } else
+    dinfo->progress = NULL;
+
+  if ((xinfo =
+       (jpeg_transform_info *)malloc(sizeof(jpeg_transform_info) * n)) == NULL)
+    THROW("tjTransform(): Memory allocation failure");
+  MEMZERO(xinfo, sizeof(jpeg_transform_info) * n);
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+
+  for (i = 0; i < n; i++) {
+    xinfoi.transform = xformtypesti.op;
+    xinfoi.perfect = (ti.options & TJXOPT_PERFECT) ? 1 : 0;
+    xinfoi.trim = (ti.options & TJXOPT_TRIM) ? 1 : 0;
+    xinfoi.force_grayscale = (ti.options & TJXOPT_GRAY) ? 1 : 0;
+    xinfoi.crop = (ti.options & TJXOPT_CROP) ? 1 : 0;
+    if (n != 1 && ti.op == TJXOP_HFLIP) xinfoi.slow_hflip = 1;
+    else xinfoi.slow_hflip = 0;
+
+    if (xinfoi.crop) {
+      xinfoi.crop_xoffset = ti.r.x;  xinfoi.crop_xoffset_set = JCROP_POS;
+      xinfoi.crop_yoffset = ti.r.y;  xinfoi.crop_yoffset_set = JCROP_POS;
+      if (ti.r.w != 0) {
+        xinfoi.crop_width = ti.r.w;  xinfoi.crop_width_set = JCROP_POS;
+      } else
+        xinfoi.crop_width = JCROP_UNSET;
+      if (ti.r.h != 0) {
+        xinfoi.crop_height = ti.r.h;  xinfoi.crop_height_set = JCROP_POS;
+      } else
+        xinfoi.crop_height = JCROP_UNSET;
+    }
+    if (!(ti.options & TJXOPT_COPYNONE)) saveMarkers = 1;
+  }
+
+  jcopy_markers_setup(dinfo, saveMarkers ? JCOPYOPT_ALL : JCOPYOPT_NONE);
+  jpeg_read_header(dinfo, TRUE);
+  jpegSubsamp = getSubsamp(dinfo);
+  if (jpegSubsamp < 0)
+    THROW("tjTransform(): Could not determine subsampling type for JPEG image");
+
+  for (i = 0; i < n; i++) {
+    if (!jtransform_request_workspace(dinfo, &xinfoi))
+      THROW("tjTransform(): Transform is not perfect");
+
+    if (xinfoi.crop) {
+      if ((ti.r.x % xinfoi.iMCU_sample_width) != 0 ||
+          (ti.r.y % xinfoi.iMCU_sample_height) != 0) {
+        snprintf(this->errStr, JMSG_LENGTH_MAX,
+                 "To crop this JPEG image, x must be a multiple of %d\n"
+                 "and y must be a multiple of %d.\n",
+                 xinfoi.iMCU_sample_width, xinfoi.iMCU_sample_height);
+        this->isInstanceError = TRUE;
+        retval = -1;  goto bailout;
+      }
+    }
+  }
+
+  srccoefs = jpeg_read_coefficients(dinfo);
+
+  for (i = 0; i < n; i++) {
+    int w, h;
+
+    if (!xinfoi.crop) {
+      w = dinfo->image_width;  h = dinfo->image_height;
+    } else {
+      w = xinfoi.crop_width;  h = xinfoi.crop_height;
+    }
+    if (flags & TJFLAG_NOREALLOC) {
+      alloc = 0;  dstSizesi = tjBufSize(w, h, jpegSubsamp);
+    }
+    if (!(ti.options & TJXOPT_NOOUTPUT))
+      jpeg_mem_dest_tj(cinfo, &dstBufsi, &dstSizesi, alloc);
+    jpeg_copy_critical_parameters(dinfo, cinfo);
+    dstcoefs = jtransform_adjust_parameters(dinfo, cinfo, srccoefs, &xinfoi);
+    if (flags & TJFLAG_PROGRESSIVE || ti.options & TJXOPT_PROGRESSIVE)
+      jpeg_simple_progression(cinfo);
+    if (!(ti.options & TJXOPT_NOOUTPUT)) {
+      jpeg_write_coefficients(cinfo, dstcoefs);
+      jcopy_markers_execute(dinfo, cinfo, ti.options & TJXOPT_COPYNONE ?
+                                          JCOPYOPT_NONE : JCOPYOPT_ALL);
+    } else
+      jinit_c_master_control(cinfo, TRUE);
+    jtransform_execute_transformation(dinfo, cinfo, srccoefs, &xinfoi);
+    if (ti.customFilter) {
+      int ci, y;
+      JDIMENSION by;
+
+      for (ci = 0; ci < cinfo->num_components; ci++) {
+        jpeg_component_info *compptr = &cinfo->comp_infoci;
+        tjregion arrayRegion = {
+          0, 0, compptr->width_in_blocks * DCTSIZE, DCTSIZE
+        };
+        tjregion planeRegion = {
+          0, 0, compptr->width_in_blocks * DCTSIZE,
+          compptr->height_in_blocks * DCTSIZE
+        };
+
+        for (by = 0; by < compptr->height_in_blocks;
+             by += compptr->v_samp_factor) {
+          JBLOCKARRAY barray = (dinfo->mem->access_virt_barray)
+            ((j_common_ptr)dinfo, dstcoefsci, by, compptr->v_samp_factor,
+             TRUE);
+
+          for (y = 0; y < compptr->v_samp_factor; y++) {
+            if (ti.customFilter(barrayy0, arrayRegion, planeRegion, ci,
+                                  i, &ti) == -1)
+              THROW("tjTransform(): Error in custom filter");
+            arrayRegion.y += DCTSIZE;
+          }
+        }
+      }
+    }
+    if (!(ti.options & TJXOPT_NOOUTPUT)) jpeg_finish_compress(cinfo);
+  }
+
+  jpeg_finish_decompress(dinfo);
+
+bailout:
+  if (cinfo->global_state > CSTATE_START) {
+    if (alloc) (*cinfo->dest->term_destination) (cinfo);
+    jpeg_abort_compress(cinfo);
+  }
+  if (dinfo->global_state > DSTATE_START) jpeg_abort_decompress(dinfo);
+  free(xinfo);
+  if (this->jerr.warning) retval = -1;
+  this->jerr.stopOnWarning = FALSE;
+  return retval;
+}
+
+
+DLLEXPORT unsigned char *tjLoadImage(const char *filename, int *width,
+                                     int align, int *height, int *pixelFormat,
+                                     int flags)
+{
+  int retval = 0, tempc;
+  size_t pitch;
+  tjhandle handle = NULL;
+  tjinstance *this;
+  j_compress_ptr cinfo = NULL;
+  cjpeg_source_ptr src;
+  unsigned char *dstBuf = NULL;
+  FILE *file = NULL;
+  boolean invert;
+
+  if (!filename || !width || align < 1 || !height || !pixelFormat ||
+      *pixelFormat < TJPF_UNKNOWN || *pixelFormat >= TJ_NUMPF)
+    THROWG("tjLoadImage(): Invalid argument");
+  if ((align & (align - 1)) != 0)
+    THROWG("tjLoadImage(): Alignment must be a power of 2");
+
+  if ((handle = tjInitCompress()) == NULL) return NULL;
+  this = (tjinstance *)handle;
+  cinfo = &this->cinfo;
+
+  if ((file = fopen(filename, "rb")) == NULL)
+    THROW_UNIX("tjLoadImage(): Cannot open input file");
+
+  if ((tempc = getc(file)) < 0 || ungetc(tempc, file) == EOF)
+    THROW_UNIX("tjLoadImage(): Could not read input file")
+  else if (tempc == EOF)
+    THROWG("tjLoadImage(): Input file contains no data");
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  if (*pixelFormat == TJPF_UNKNOWN) cinfo->in_color_space = JCS_UNKNOWN;
+  else cinfo->in_color_space = pf2cs*pixelFormat;
+  if (tempc == 'B') {
+    if ((src = jinit_read_bmp(cinfo, FALSE)) == NULL)
+      THROWG("tjLoadImage(): Could not initialize bitmap loader");
+    invert = (flags & TJFLAG_BOTTOMUP) == 0;
+  } else if (tempc == 'P') {
+    if ((src = jinit_read_ppm(cinfo)) == NULL)
+      THROWG("tjLoadImage(): Could not initialize bitmap loader");
+    invert = (flags & TJFLAG_BOTTOMUP) != 0;
+  } else
+    THROWG("tjLoadImage(): Unsupported file type");
+
+  src->input_file = file;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+  /* Refuse to load images larger than 1 Megapixel when fuzzing. */
+  if (flags & TJFLAG_FUZZING)
+    src->max_pixels = 1048576;
+#endif
+  (*src->start_input) (cinfo, src);
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr)cinfo);
+
+  *width = cinfo->image_width;  *height = cinfo->image_height;
+  *pixelFormat = cs2pfcinfo->in_color_space;
+
+  pitch = PAD((*width) * tjPixelSize*pixelFormat, align);
+  if ((unsigned long long)pitch * (unsigned long long)(*height) >
+      (unsigned long long)((size_t)-1) ||
+      (dstBuf = (unsigned char *)malloc(pitch * (*height))) == NULL)
+    THROWG("tjLoadImage(): Memory allocation failure");
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  while (cinfo->next_scanline < cinfo->image_height) {
+    int i, nlines = (*src->get_pixel_rows) (cinfo, src);
+
+    for (i = 0; i < nlines; i++) {
+      unsigned char *dstptr;
+      int row;
+
+      row = cinfo->next_scanline + i;
+      if (invert) dstptr = &dstBuf((*height) - row - 1) * pitch;
+      else dstptr = &dstBufrow * pitch;
+      memcpy(dstptr, src->bufferi, (*width) * tjPixelSize*pixelFormat);
+    }
+    cinfo->next_scanline += nlines;
+  }
+
+  (*src->finish_input) (cinfo, src);
+
+bailout:
+  if (handle) tjDestroy(handle);
+  if (file) fclose(file);
+  if (retval < 0) { free(dstBuf);  dstBuf = NULL; }
+  return dstBuf;
+}
+
+
+DLLEXPORT int tjSaveImage(const char *filename, unsigned char *buffer,
+                          int width, int pitch, int height, int pixelFormat,
+                          int flags)
+{
+  int retval = 0;
+  tjhandle handle = NULL;
+  tjinstance *this;
+  j_decompress_ptr dinfo = NULL;
+  djpeg_dest_ptr dst;
+  FILE *file = NULL;
+  char *ptr = NULL;
+  boolean invert;
+
+  if (!filename || !buffer || width < 1 || pitch < 0 || height < 1 ||
+      pixelFormat < 0 || pixelFormat >= TJ_NUMPF)
+    THROWG("tjSaveImage(): Invalid argument");
+
+  if ((handle = tjInitDecompress()) == NULL)
+    return -1;
+  this = (tjinstance *)handle;
+  dinfo = &this->dinfo;
+
+  if ((file = fopen(filename, "wb")) == NULL)
+    THROW_UNIX("tjSaveImage(): Cannot open output file");
+
+  if (setjmp(this->jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error. */
+    retval = -1;  goto bailout;
+  }
+
+  this->dinfo.out_color_space = pf2cspixelFormat;
+  dinfo->image_width = width;  dinfo->image_height = height;
+  dinfo->global_state = DSTATE_READY;
+  dinfo->scale_num = dinfo->scale_denom = 1;
+
+  ptr = strrchr(filename, '.');
+  if (ptr && !strcasecmp(ptr, ".bmp")) {
+    if ((dst = jinit_write_bmp(dinfo, FALSE, FALSE)) == NULL)
+      THROWG("tjSaveImage(): Could not initialize bitmap writer");
+    invert = (flags & TJFLAG_BOTTOMUP) == 0;
+  } else {
+    if ((dst = jinit_write_ppm(dinfo)) == NULL)
+      THROWG("tjSaveImage(): Could not initialize PPM writer");
+    invert = (flags & TJFLAG_BOTTOMUP) != 0;
+  }
+
+  dst->output_file = file;
+  (*dst->start_output) (dinfo, dst);
+  (*dinfo->mem->realize_virt_arrays) ((j_common_ptr)dinfo);
+
+  if (pitch == 0) pitch = width * tjPixelSizepixelFormat;
+
+  while (dinfo->output_scanline < dinfo->output_height) {
+    unsigned char *rowptr;
+
+    if (invert)
+      rowptr = &buffer(height - dinfo->output_scanline - 1) * pitch;
+    else
+      rowptr = &bufferdinfo->output_scanline * pitch;
+    memcpy(dst->buffer0, rowptr, width * tjPixelSizepixelFormat);
+    (*dst->put_pixel_rows) (dinfo, dst, 1);
+    dinfo->output_scanline++;
+  }
+
+  (*dst->finish_output) (dinfo, dst);
+
+bailout:
+  if (handle) tjDestroy(handle);
+  if (file) fclose(file);
+  return retval;
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/turbojpeg.h Added

@@ -0,0 +1,1756 @@
+/*
+ * Copyright (C)2009-2015, 2017, 2020-2021 D. R. Commander.
+ *                                         All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __TURBOJPEG_H__
+#define __TURBOJPEG_H__
+
+#if defined(_WIN32) && defined(DLLDEFINE)
+#define DLLEXPORT  __declspec(dllexport)
+#else
+#define DLLEXPORT
+#endif
+#define DLLCALL
+
+
+/**
+ * @addtogroup TurboJPEG
+ * TurboJPEG API.  This API provides an interface for generating, decoding, and
+ * transforming planar YUV and JPEG images in memory.
+ *
+ * @anchor YUVnotes
+ * YUV Image Format Notes
+ * ----------------------
+ * Technically, the JPEG format uses the YCbCr colorspace (which is technically
+ * not a colorspace but a color transform), but per the convention of the
+ * digital video community, the TurboJPEG API uses "YUV" to refer to an image
+ * format consisting of Y, Cb, and Cr image planes.
+ *
+ * Each plane is simply a 2D array of bytes, each byte representing the value
+ * of one of the components (Y, Cb, or Cr) at a particular location in the
+ * image.  The width and height of each plane are determined by the image
+ * width, height, and level of chrominance subsampling.   The luminance plane
+ * width is the image width padded to the nearest multiple of the horizontal
+ * subsampling factor (2 in the case of 4:2:0 and 4:2:2, 4 in the case of
+ * 4:1:1, 1 in the case of 4:4:4 or grayscale.)  Similarly, the luminance plane
+ * height is the image height padded to the nearest multiple of the vertical
+ * subsampling factor (2 in the case of 4:2:0 or 4:4:0, 1 in the case of 4:4:4
+ * or grayscale.)  This is irrespective of any additional padding that may be
+ * specified as an argument to the various YUV functions.  The chrominance
+ * plane width is equal to the luminance plane width divided by the horizontal
+ * subsampling factor, and the chrominance plane height is equal to the
+ * luminance plane height divided by the vertical subsampling factor.
+ *
+ * For example, if the source image is 35 x 35 pixels and 4:2:2 subsampling is
+ * used, then the luminance plane would be 36 x 35 bytes, and each of the
+ * chrominance planes would be 18 x 35 bytes.  If you specify a line padding of
+ * 4 bytes on top of this, then the luminance plane would be 36 x 35 bytes, and
+ * each of the chrominance planes would be 20 x 35 bytes.
+ *
+ * @{
+ */
+
+
+/**
+ * The number of chrominance subsampling options
+ */
+#define TJ_NUMSAMP  6
+
+/**
+ * Chrominance subsampling options.
+ * When pixels are converted from RGB to YCbCr (see #TJCS_YCbCr) or from CMYK
+ * to YCCK (see #TJCS_YCCK) as part of the JPEG compression process, some of
+ * the Cb and Cr (chrominance) components can be discarded or averaged together
+ * to produce a smaller image with little perceptible loss of image clarity
+ * (the human eye is more sensitive to small changes in brightness than to
+ * small changes in color.)  This is called "chrominance subsampling".
+ */
+enum TJSAMP {
+  /**
+   * 4:4:4 chrominance subsampling (no chrominance subsampling).  The JPEG or
+   * YUV image will contain one chrominance component for every pixel in the
+   * source image.
+   */
+  TJSAMP_444 = 0,
+  /**
+   * 4:2:2 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 2x1 block of pixels in the source image.
+   */
+  TJSAMP_422,
+  /**
+   * 4:2:0 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 2x2 block of pixels in the source image.
+   */
+  TJSAMP_420,
+  /**
+   * Grayscale.  The JPEG or YUV image will contain no chrominance components.
+   */
+  TJSAMP_GRAY,
+  /**
+   * 4:4:0 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 1x2 block of pixels in the source image.
+   *
+   * @note 4:4:0 subsampling is not fully accelerated in libjpeg-turbo.
+   */
+  TJSAMP_440,
+  /**
+   * 4:1:1 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 4x1 block of pixels in the source image.
+   * JPEG images compressed with 4:1:1 subsampling will be almost exactly the
+   * same size as those compressed with 4:2:0 subsampling, and in the
+   * aggregate, both subsampling methods produce approximately the same
+   * perceptual quality.  However, 4:1:1 is better able to reproduce sharp
+   * horizontal features.
+   *
+   * @note 4:1:1 subsampling is not fully accelerated in libjpeg-turbo.
+   */
+  TJSAMP_411
+};
+
+/**
+ * MCU block width (in pixels) for a given level of chrominance subsampling.
+ * MCU block sizes:
+ * - 8x8 for no subsampling or grayscale
+ * - 16x8 for 4:2:2
+ * - 8x16 for 4:4:0
+ * - 16x16 for 4:2:0
+ * - 32x8 for 4:1:1
+ */
+static const int tjMCUWidthTJ_NUMSAMP  = { 8, 16, 16, 8, 8, 32 };
+
+/**
+ * MCU block height (in pixels) for a given level of chrominance subsampling.
+ * MCU block sizes:
+ * - 8x8 for no subsampling or grayscale
+ * - 16x8 for 4:2:2
+ * - 8x16 for 4:4:0
+ * - 16x16 for 4:2:0
+ * - 32x8 for 4:1:1
+ */
+static const int tjMCUHeightTJ_NUMSAMP = { 8, 8, 16, 8, 16, 8 };
+
+
+/**
+ * The number of pixel formats
+ */
+#define TJ_NUMPF  12
+
+/**
+ * Pixel formats
+ */
+enum TJPF {
+  /**
+   * RGB pixel format.  The red, green, and blue components in the image are
+   * stored in 3-byte pixels in the order R, G, B from lowest to highest byte
+   * address within each pixel.
+   */
+  TJPF_RGB = 0,
+  /**
+   * BGR pixel format.  The red, green, and blue components in the image are
+   * stored in 3-byte pixels in the order B, G, R from lowest to highest byte
+   * address within each pixel.
+   */
+  TJPF_BGR,
+  /**
+   * RGBX pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order R, G, B from lowest to highest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_RGBX,
+  /**
+   * BGRX pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order B, G, R from lowest to highest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_BGRX,
+  /**
+   * XBGR pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order R, G, B from highest to lowest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_XBGR,
+  /**
+   * XRGB pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order B, G, R from highest to lowest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_XRGB,
+  /**
+   * Grayscale pixel format.  Each 1-byte pixel represents a luminance
+   * (brightness) level from 0 to 255.
+   */
+  TJPF_GRAY,
+  /**
+   * RGBA pixel format.  This is the same as @ref TJPF_RGBX, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_RGBA,
+  /**
+   * BGRA pixel format.  This is the same as @ref TJPF_BGRX, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_BGRA,
+  /**
+   * ABGR pixel format.  This is the same as @ref TJPF_XBGR, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_ABGR,
+  /**
+   * ARGB pixel format.  This is the same as @ref TJPF_XRGB, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_ARGB,
+  /**
+   * CMYK pixel format.  Unlike RGB, which is an additive color model used
+   * primarily for display, CMYK (Cyan/Magenta/Yellow/Key) is a subtractive
+   * color model used primarily for printing.  In the CMYK color model, the
+   * value of each color component typically corresponds to an amount of cyan,
+   * magenta, yellow, or black ink that is applied to a white background.  In
+   * order to convert between CMYK and RGB, it is necessary to use a color
+   * management system (CMS.)  A CMS will attempt to map colors within the
+   * printer's gamut to perceptually similar colors in the display's gamut and
+   * vice versa, but the mapping is typically not 1:1 or reversible, nor can it
+   * be defined with a simple formula.  Thus, such a conversion is out of scope
+   * for a codec library.  However, the TurboJPEG API allows for compressing
+   * CMYK pixels into a YCCK JPEG image (see #TJCS_YCCK) and decompressing YCCK
+   * JPEG images into CMYK pixels.
+   */
+  TJPF_CMYK,
+  /**
+   * Unknown pixel format.  Currently this is only used by #tjLoadImage().
+   */
+  TJPF_UNKNOWN = -1
+};
+
+/**
+ * Red offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the red component is offset from the start of the pixel.  For
+ * instance, if a pixel of format TJ_BGRX is stored in <tt>char pixel</tt>,
+ * then the red component will be <tt>pixeltjRedOffsetTJ_BGRX</tt>.  This
+ * will be -1 if the pixel format does not have a red component.
+ */
+static const int tjRedOffsetTJ_NUMPF = {
+  0, 2, 0, 2, 3, 1, -1, 0, 2, 3, 1, -1
+};
+/**
+ * Green offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the green component is offset from the start of the pixel.
+ * For instance, if a pixel of format TJ_BGRX is stored in
+ * <tt>char pixel</tt>, then the green component will be
+ * <tt>pixeltjGreenOffsetTJ_BGRX</tt>.  This will be -1 if the pixel format
+ * does not have a green component.
+ */
+static const int tjGreenOffsetTJ_NUMPF = {
+  1, 1, 1, 1, 2, 2, -1, 1, 1, 2, 2, -1
+};
+/**
+ * Blue offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the Blue component is offset from the start of the pixel.  For
+ * instance, if a pixel of format TJ_BGRX is stored in <tt>char pixel</tt>,
+ * then the blue component will be <tt>pixeltjBlueOffsetTJ_BGRX</tt>.  This
+ * will be -1 if the pixel format does not have a blue component.
+ */
+static const int tjBlueOffsetTJ_NUMPF = {
+  2, 0, 2, 0, 1, 3, -1, 2, 0, 1, 3, -1
+};
+/**
+ * Alpha offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the Alpha component is offset from the start of the pixel.
+ * For instance, if a pixel of format TJ_BGRA is stored in
+ * <tt>char pixel</tt>, then the alpha component will be
+ * <tt>pixeltjAlphaOffsetTJ_BGRA</tt>.  This will be -1 if the pixel format
+ * does not have an alpha component.
+ */
+static const int tjAlphaOffsetTJ_NUMPF = {
+  -1, -1, -1, -1, -1, -1, -1, 3, 3, 0, 0, -1
+};
+/**
+ * Pixel size (in bytes) for a given pixel format
+ */
+static const int tjPixelSizeTJ_NUMPF = {
+  3, 3, 4, 4, 4, 4, 1, 4, 4, 4, 4, 4
+};
+
+
+/**
+ * The number of JPEG colorspaces
+ */
+#define TJ_NUMCS  5
+
+/**
+ * JPEG colorspaces
+ */
+enum TJCS {
+  /**
+   * RGB colorspace.  When compressing the JPEG image, the R, G, and B
+   * components in the source image are reordered into image planes, but no
+   * colorspace conversion or subsampling is performed.  RGB JPEG images can be
+   * decompressed to any of the extended RGB pixel formats or grayscale, but
+   * they cannot be decompressed to YUV images.
+   */
+  TJCS_RGB = 0,
+  /**
+   * YCbCr colorspace.  YCbCr is not an absolute colorspace but rather a
+   * mathematical transformation of RGB designed solely for storage and
+   * transmission.  YCbCr images must be converted to RGB before they can
+   * actually be displayed.  In the YCbCr colorspace, the Y (luminance)
+   * component represents the black & white portion of the original image, and
+   * the Cb and Cr (chrominance) components represent the color portion of the
+   * original image.  Originally, the analog equivalent of this transformation
+   * allowed the same signal to drive both black & white and color televisions,
+   * but JPEG images use YCbCr primarily because it allows the color data to be
+   * optionally subsampled for the purposes of reducing bandwidth or disk
+   * space.  YCbCr is the most common JPEG colorspace, and YCbCr JPEG images
+   * can be compressed from and decompressed to any of the extended RGB pixel
+   * formats or grayscale, or they can be decompressed to YUV planar images.
+   */
+  TJCS_YCbCr,
+  /**
+   * Grayscale colorspace.  The JPEG image retains only the luminance data (Y
+   * component), and any color data from the source image is discarded.
+   * Grayscale JPEG images can be compressed from and decompressed to any of
+   * the extended RGB pixel formats or grayscale, or they can be decompressed
+   * to YUV planar images.
+   */
+  TJCS_GRAY,
+  /**
+   * CMYK colorspace.  When compressing the JPEG image, the C, M, Y, and K
+   * components in the source image are reordered into image planes, but no
+   * colorspace conversion or subsampling is performed.  CMYK JPEG images can
+   * only be decompressed to CMYK pixels.
+   */
+  TJCS_CMYK,
+  /**
+   * YCCK colorspace.  YCCK (AKA "YCbCrK") is not an absolute colorspace but
+   * rather a mathematical transformation of CMYK designed solely for storage
+   * and transmission.  It is to CMYK as YCbCr is to RGB.  CMYK pixels can be
+   * reversibly transformed into YCCK, and as with YCbCr, the chrominance
+   * components in the YCCK pixels can be subsampled without incurring major
+   * perceptual loss.  YCCK JPEG images can only be compressed from and
+   * decompressed to CMYK pixels.
+   */
+  TJCS_YCCK
+};
+
+
+/**
+ * The uncompressed source/destination image is stored in bottom-up (Windows,
+ * OpenGL) order, not top-down (X11) order.
+ */
+#define TJFLAG_BOTTOMUP  2
+/**
+ * When decompressing an image that was compressed using chrominance
+ * subsampling, use the fastest chrominance upsampling algorithm available in
+ * the underlying codec.  The default is to use smooth upsampling, which
+ * creates a smooth transition between neighboring chrominance components in
+ * order to reduce upsampling artifacts in the decompressed image.
+ */
+#define TJFLAG_FASTUPSAMPLE  256
+/**
+ * Disable buffer (re)allocation.  If passed to one of the JPEG compression or
+ * transform functions, this flag will cause those functions to generate an
+ * error if the JPEG image buffer is invalid or too small rather than
+ * attempting to allocate or reallocate that buffer.  This reproduces the
+ * behavior of earlier versions of TurboJPEG.
+ */
+#define TJFLAG_NOREALLOC  1024
+/**
+ * Use the fastest DCT/IDCT algorithm available in the underlying codec.  The
+ * default if this flag is not specified is implementation-specific.  For
+ * example, the implementation of TurboJPEG for libjpeg-turbo uses the fast
+ * algorithm by default when compressing, because this has been shown to have
+ * only a very slight effect on accuracy, but it uses the accurate algorithm
+ * when decompressing, because this has been shown to have a larger effect.
+ */
+#define TJFLAG_FASTDCT  2048
+/**
+ * Use the most accurate DCT/IDCT algorithm available in the underlying codec.
+ * The default if this flag is not specified is implementation-specific.  For
+ * example, the implementation of TurboJPEG for libjpeg-turbo uses the fast
+ * algorithm by default when compressing, because this has been shown to have
+ * only a very slight effect on accuracy, but it uses the accurate algorithm
+ * when decompressing, because this has been shown to have a larger effect.
+ */
+#define TJFLAG_ACCURATEDCT  4096
+/**
+ * Immediately discontinue the current compression/decompression/transform
+ * operation if the underlying codec throws a warning (non-fatal error).  The
+ * default behavior is to allow the operation to complete unless a fatal error
+ * is encountered.
+ */
+#define TJFLAG_STOPONWARNING  8192
+/**
+ * Use progressive entropy coding in JPEG images generated by the compression
+ * and transform functions.  Progressive entropy coding will generally improve
+ * compression relative to baseline entropy coding (the default), but it will
+ * reduce compression and decompression performance considerably.
+ */
+#define TJFLAG_PROGRESSIVE  16384
+/**
+ * Limit the number of progressive JPEG scans that the decompression and
+ * transform functions will process.  If a progressive JPEG image contains an
+ * unreasonably large number of scans, then this flag will cause the
+ * decompression and transform functions to return an error.  The primary
+ * purpose of this is to allow security-critical applications to guard against
+ * an exploit of the progressive JPEG format described in
+ * <a href="https://libjpeg-turbo.org/pmwiki/uploads/About/TwoIssueswiththeJPEGStandard.pdf" target="_blank">this report</a>.
+ */
+#define TJFLAG_LIMITSCANS  32768
+
+
+/**
+ * The number of error codes
+ */
+#define TJ_NUMERR  2
+
+/**
+ * Error codes
+ */
+enum TJERR {
+  /**
+   * The error was non-fatal and recoverable, but the image may still be
+   * corrupt.
+   */
+  TJERR_WARNING = 0,
+  /**
+   * The error was fatal and non-recoverable.
+   */
+  TJERR_FATAL
+};
+
+
+/**
+ * The number of transform operations
+ */
+#define TJ_NUMXOP  8
+
+/**
+ * Transform operations for #tjTransform()
+ */
+enum TJXOP {
+  /**
+   * Do not transform the position of the image pixels
+   */
+  TJXOP_NONE = 0,
+  /**
+   * Flip (mirror) image horizontally.  This transform is imperfect if there
+   * are any partial MCU blocks on the right edge (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_HFLIP,
+  /**
+   * Flip (mirror) image vertically.  This transform is imperfect if there are
+   * any partial MCU blocks on the bottom edge (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_VFLIP,
+  /**
+   * Transpose image (flip/mirror along upper left to lower right axis.)  This
+   * transform is always perfect.
+   */
+  TJXOP_TRANSPOSE,
+  /**
+   * Transverse transpose image (flip/mirror along upper right to lower left
+   * axis.)  This transform is imperfect if there are any partial MCU blocks in
+   * the image (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_TRANSVERSE,
+  /**
+   * Rotate image clockwise by 90 degrees.  This transform is imperfect if
+   * there are any partial MCU blocks on the bottom edge (see
+   * #TJXOPT_PERFECT.)
+   */
+  TJXOP_ROT90,
+  /**
+   * Rotate image 180 degrees.  This transform is imperfect if there are any
+   * partial MCU blocks in the image (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_ROT180,
+  /**
+   * Rotate image counter-clockwise by 90 degrees.  This transform is imperfect
+   * if there are any partial MCU blocks on the right edge (see
+   * #TJXOPT_PERFECT.)
+   */
+  TJXOP_ROT270
+};
+
+
+/**
+ * This option will cause #tjTransform() to return an error if the transform is
+ * not perfect.  Lossless transforms operate on MCU blocks, whose size depends
+ * on the level of chrominance subsampling used (see #tjMCUWidth
+ * and #tjMCUHeight.)  If the image's width or height is not evenly divisible
+ * by the MCU block size, then there will be partial MCU blocks on the right
+ * and/or bottom edges.  It is not possible to move these partial MCU blocks to
+ * the top or left of the image, so any transform that would require that is
+ * "imperfect."  If this option is not specified, then any partial MCU blocks
+ * that cannot be transformed will be left in place, which will create
+ * odd-looking strips on the right or bottom edge of the image.
+ */
+#define TJXOPT_PERFECT  1
+/**
+ * This option will cause #tjTransform() to discard any partial MCU blocks that
+ * cannot be transformed.
+ */
+#define TJXOPT_TRIM  2
+/**
+ * This option will enable lossless cropping.  See #tjTransform() for more
+ * information.
+ */
+#define TJXOPT_CROP  4
+/**
+ * This option will discard the color data in the input image and produce
+ * a grayscale output image.
+ */
+#define TJXOPT_GRAY  8
+/**
+ * This option will prevent #tjTransform() from outputting a JPEG image for
+ * this particular transform (this can be used in conjunction with a custom
+ * filter to capture the transformed DCT coefficients without transcoding
+ * them.)
+ */
+#define TJXOPT_NOOUTPUT  16
+/**
+ * This option will enable progressive entropy coding in the output image
+ * generated by this particular transform.  Progressive entropy coding will
+ * generally improve compression relative to baseline entropy coding (the
+ * default), but it will reduce compression and decompression performance
+ * considerably.
+ */
+#define TJXOPT_PROGRESSIVE  32
+/**
+ * This option will prevent #tjTransform() from copying any extra markers
+ * (including EXIF and ICC profile data) from the source image to the output
+ * image.
+ */
+#define TJXOPT_COPYNONE  64
+
+
+/**
+ * Scaling factor
+ */
+typedef struct {
+  /**
+   * Numerator
+   */
+  int num;
+  /**
+   * Denominator
+   */
+  int denom;
+} tjscalingfactor;
+
+/**
+ * Cropping region
+ */
+typedef struct {
+  /**
+   * The left boundary of the cropping region.  This must be evenly divisible
+   * by the MCU block width (see #tjMCUWidth.)
+   */
+  int x;
+  /**
+   * The upper boundary of the cropping region.  This must be evenly divisible
+   * by the MCU block height (see #tjMCUHeight.)
+   */
+  int y;
+  /**
+   * The width of the cropping region. Setting this to 0 is the equivalent of
+   * setting it to the width of the source JPEG image - x.
+   */
+  int w;
+  /**
+   * The height of the cropping region. Setting this to 0 is the equivalent of
+   * setting it to the height of the source JPEG image - y.
+   */
+  int h;
+} tjregion;
+
+/**
+ * Lossless transform
+ */
+typedef struct tjtransform {
+  /**
+   * Cropping region
+   */
+  tjregion r;
+  /**
+   * One of the @ref TJXOP "transform operations"
+   */
+  int op;
+  /**
+   * The bitwise OR of one of more of the @ref TJXOPT_CROP "transform options"
+   */
+  int options;
+  /**
+   * Arbitrary data that can be accessed within the body of the callback
+   * function
+   */
+  void *data;
+  /**
+   * A callback function that can be used to modify the DCT coefficients
+   * after they are losslessly transformed but before they are transcoded to a
+   * new JPEG image.  This allows for custom filters or other transformations
+   * to be applied in the frequency domain.
+   *
+   * @param coeffs pointer to an array of transformed DCT coefficients.  (NOTE:
+   * this pointer is not guaranteed to be valid once the callback returns, so
+   * applications wishing to hand off the DCT coefficients to another function
+   * or library should make a copy of them within the body of the callback.)
+   *
+   * @param arrayRegion #tjregion structure containing the width and height of
+   * the array pointed to by <tt>coeffs</tt> as well as its offset relative to
+   * the component plane.  TurboJPEG implementations may choose to split each
+   * component plane into multiple DCT coefficient arrays and call the callback
+   * function once for each array.
+   *
+   * @param planeRegion #tjregion structure containing the width and height of
+   * the component plane to which <tt>coeffs</tt> belongs
+   *
+   * @param componentID ID number of the component plane to which
+   * <tt>coeffs</tt> belongs (Y, Cb, and Cr have, respectively, ID's of 0, 1,
+   * and 2 in typical JPEG images.)
+   *
+   * @param transformID ID number of the transformed image to which
+   * <tt>coeffs</tt> belongs.  This is the same as the index of the transform
+   * in the <tt>transforms</tt> array that was passed to #tjTransform().
+   *
+   * @param transform a pointer to a #tjtransform structure that specifies the
+   * parameters and/or cropping region for this transform
+   *
+   * @return 0 if the callback was successful, or -1 if an error occurred.
+   */
+  int (*customFilter) (short *coeffs, tjregion arrayRegion,
+                       tjregion planeRegion, int componentIndex,
+                       int transformIndex, struct tjtransform *transform);
+} tjtransform;
+
+/**
+ * TurboJPEG instance handle
+ */
+typedef void *tjhandle;
+
+
+/**
+ * Pad the given width to the nearest 32-bit boundary
+ */
+#define TJPAD(width)  (((width) + 3) & (~3))
+
+/**
+ * Compute the scaled value of <tt>dimension</tt> using the given scaling
+ * factor.  This macro performs the integer equivalent of <tt>ceil(dimension *
+ * scalingFactor)</tt>.
+ */
+#define TJSCALED(dimension, scalingFactor) \
+  ((dimension * scalingFactor.num + scalingFactor.denom - 1) / \
+   scalingFactor.denom)
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/**
+ * Create a TurboJPEG compressor instance.
+ *
+ * @return a handle to the newly-created instance, or NULL if an error
+ * occurred (see #tjGetErrorStr2().)
+ */
+DLLEXPORT tjhandle tjInitCompress(void);
+
+
+/**
+ * Compress an RGB, grayscale, or CMYK image into a JPEG image.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing RGB, grayscale, or
+ * CMYK pixels to be compressed
+ *
+ * @param width width (in pixels) of the source image
+ *
+ * @param pitch bytes per line in the source image.  Normally, this should be
+ * <tt>width * #tjPixelSizepixelFormat</tt> if the image is unpadded, or
+ * <tt>#TJPAD(width * #tjPixelSizepixelFormat)</tt> if each line of the image
+ * is padded to the nearest 32-bit boundary, as is the case for Windows
+ * bitmaps.  You can also be clever and use this parameter to skip lines, etc.
+ * Setting this parameter to 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSizepixelFormat</tt>.
+ *
+ * @param height height (in pixels) of the source image
+ *
+ * @param pixelFormat pixel format of the source image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param jpegBuf address of a pointer to an image buffer that will receive the
+ * JPEG image.  TurboJPEG has the ability to reallocate the JPEG buffer
+ * to accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>*jpegBuf</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize().  This should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees that it won't be.)
+ * .
+ * If you choose option 1, <tt>*jpegSize</tt> should be set to the size of your
+ * pre-allocated buffer.  In any case, unless you have set #TJFLAG_NOREALLOC,
+ * you should always check <tt>*jpegBuf</tt> upon return from this function, as
+ * it may have changed.
+ *
+ * @param jpegSize pointer to an unsigned long variable that holds the size of
+ * the JPEG image buffer.  If <tt>*jpegBuf</tt> points to a pre-allocated
+ * buffer, then <tt>*jpegSize</tt> should be set to the size of the buffer.
+ * Upon return, <tt>*jpegSize</tt> will contain the size of the JPEG image (in
+ * bytes.)  If <tt>*jpegBuf</tt> points to a JPEG image buffer that is being
+ * reused from a previous call to one of the JPEG compression functions, then
+ * <tt>*jpegSize</tt> is ignored.
+ *
+ * @param jpegSubsamp the level of chrominance subsampling to be used when
+ * generating the JPEG image (see @ref TJSAMP
+ * "Chrominance subsampling options".)
+ *
+ * @param jpegQual the image quality of the generated JPEG image (1 = worst,
+ * 100 = best)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+*/
+DLLEXPORT int tjCompress2(tjhandle handle, const unsigned char *srcBuf,
+                          int width, int pitch, int height, int pixelFormat,
+                          unsigned char **jpegBuf, unsigned long *jpegSize,
+                          int jpegSubsamp, int jpegQual, int flags);
+
+
+/**
+ * Compress a YUV planar image into a JPEG image.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing a YUV planar image to be
+ * compressed.  The size of this buffer should match the value returned by
+ * #tjBufSizeYUV2() for the given image width, height, padding, and level of
+ * chrominance subsampling.  The Y, U (Cb), and V (Cr) image planes should be
+ * stored sequentially in the source buffer (refer to @ref YUVnotes
+ * "YUV Image Format Notes".)
+ *
+ * @param width width (in pixels) of the source image.  If the width is not an
+ * even multiple of the MCU block width (see #tjMCUWidth), then an intermediate
+ * buffer copy will be performed within TurboJPEG.
+ *
+ * @param pad the line padding used in the source image.  For instance, if each
+ * line in each plane of the YUV image is padded to the nearest multiple of 4
+ * bytes, then <tt>pad</tt> should be set to 4.
+ *
+ * @param height height (in pixels) of the source image.  If the height is not
+ * an even multiple of the MCU block height (see #tjMCUHeight), then an
+ * intermediate buffer copy will be performed within TurboJPEG.
+ *
+ * @param subsamp the level of chrominance subsampling used in the source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param jpegBuf address of a pointer to an image buffer that will receive the
+ * JPEG image.  TurboJPEG has the ability to reallocate the JPEG buffer to
+ * accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>*jpegBuf</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize().  This should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees that it won't be.)
+ * .
+ * If you choose option 1, <tt>*jpegSize</tt> should be set to the size of your
+ * pre-allocated buffer.  In any case, unless you have set #TJFLAG_NOREALLOC,
+ * you should always check <tt>*jpegBuf</tt> upon return from this function, as
+ * it may have changed.
+ *
+ * @param jpegSize pointer to an unsigned long variable that holds the size of
+ * the JPEG image buffer.  If <tt>*jpegBuf</tt> points to a pre-allocated
+ * buffer, then <tt>*jpegSize</tt> should be set to the size of the buffer.
+ * Upon return, <tt>*jpegSize</tt> will contain the size of the JPEG image (in
+ * bytes.)  If <tt>*jpegBuf</tt> points to a JPEG image buffer that is being
+ * reused from a previous call to one of the JPEG compression functions, then
+ * <tt>*jpegSize</tt> is ignored.
+ *
+ * @param jpegQual the image quality of the generated JPEG image (1 = worst,
+ * 100 = best)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+*/
+DLLEXPORT int tjCompressFromYUV(tjhandle handle, const unsigned char *srcBuf,
+                                int width, int pad, int height, int subsamp,
+                                unsigned char **jpegBuf,
+                                unsigned long *jpegSize, int jpegQual,
+                                int flags);
+
+
+/**
+ * Compress a set of Y, U (Cb), and V (Cr) image planes into a JPEG image.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if compressing a grayscale image) that contain a YUV
+ * image to be compressed.  These planes can be contiguous or non-contiguous in
+ * memory.  The size of each plane should match the value returned by
+ * #tjPlaneSizeYUV() for the given image width, height, strides, and level of
+ * chrominance subsampling.  Refer to @ref YUVnotes "YUV Image Format Notes"
+ * for more details.
+ *
+ * @param width width (in pixels) of the source image.  If the width is not an
+ * even multiple of the MCU block width (see #tjMCUWidth), then an intermediate
+ * buffer copy will be performed within TurboJPEG.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the YUV source image.  Setting the stride
+ * for any plane to 0 is the same as setting it to the plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective plane widths.
+ * You can adjust the strides in order to specify an arbitrary amount of line
+ * padding in each plane or to create a JPEG image from a subregion of a larger
+ * YUV planar image.
+ *
+ * @param height height (in pixels) of the source image.  If the height is not
+ * an even multiple of the MCU block height (see #tjMCUHeight), then an
+ * intermediate buffer copy will be performed within TurboJPEG.
+ *
+ * @param subsamp the level of chrominance subsampling used in the source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param jpegBuf address of a pointer to an image buffer that will receive the
+ * JPEG image.  TurboJPEG has the ability to reallocate the JPEG buffer to
+ * accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>*jpegBuf</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize().  This should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees that it won't be.)
+ * .
+ * If you choose option 1, <tt>*jpegSize</tt> should be set to the size of your
+ * pre-allocated buffer.  In any case, unless you have set #TJFLAG_NOREALLOC,
+ * you should always check <tt>*jpegBuf</tt> upon return from this function, as
+ * it may have changed.
+ *
+ * @param jpegSize pointer to an unsigned long variable that holds the size of
+ * the JPEG image buffer.  If <tt>*jpegBuf</tt> points to a pre-allocated
+ * buffer, then <tt>*jpegSize</tt> should be set to the size of the buffer.
+ * Upon return, <tt>*jpegSize</tt> will contain the size of the JPEG image (in
+ * bytes.)  If <tt>*jpegBuf</tt> points to a JPEG image buffer that is being
+ * reused from a previous call to one of the JPEG compression functions, then
+ * <tt>*jpegSize</tt> is ignored.
+ *
+ * @param jpegQual the image quality of the generated JPEG image (1 = worst,
+ * 100 = best)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+*/
+DLLEXPORT int tjCompressFromYUVPlanes(tjhandle handle,
+                                      const unsigned char **srcPlanes,
+                                      int width, const int *strides,
+                                      int height, int subsamp,
+                                      unsigned char **jpegBuf,
+                                      unsigned long *jpegSize, int jpegQual,
+                                      int flags);
+
+
+/**
+ * The maximum size of the buffer (in bytes) required to hold a JPEG image with
+ * the given parameters.  The number of bytes returned by this function is
+ * larger than the size of the uncompressed source image.  The reason for this
+ * is that the JPEG format uses 16-bit coefficients, and it is thus possible
+ * for a very high-quality JPEG image with very high-frequency content to
+ * expand rather than compress when converted to the JPEG format.  Such images
+ * represent a very rare corner case, but since there is no way to predict the
+ * size of a JPEG image prior to compression, the corner case has to be
+ * handled.
+ *
+ * @param width width (in pixels) of the image
+ *
+ * @param height height (in pixels) of the image
+ *
+ * @param jpegSubsamp the level of chrominance subsampling to be used when
+ * generating the JPEG image (see @ref TJSAMP
+ * "Chrominance subsampling options".)
+ *
+ * @return the maximum size of the buffer (in bytes) required to hold the
+ * image, or -1 if the arguments are out of bounds.
+ */
+DLLEXPORT unsigned long tjBufSize(int width, int height, int jpegSubsamp);
+
+
+/**
+ * The size of the buffer (in bytes) required to hold a YUV planar image with
+ * the given parameters.
+ *
+ * @param width width (in pixels) of the image
+ *
+ * @param pad the width of each line in each plane of the image is padded to
+ * the nearest multiple of this number of bytes (must be a power of 2.)
+ *
+ * @param height height (in pixels) of the image
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the size of the buffer (in bytes) required to hold the image, or
+ * -1 if the arguments are out of bounds.
+ */
+DLLEXPORT unsigned long tjBufSizeYUV2(int width, int pad, int height,
+                                      int subsamp);
+
+
+/**
+ * The size of the buffer (in bytes) required to hold a YUV image plane with
+ * the given parameters.
+ *
+ * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)
+ *
+ * @param width width (in pixels) of the YUV image.  NOTE: this is the width of
+ * the whole image, not the plane width.
+ *
+ * @param stride bytes per line in the image plane.  Setting this to 0 is the
+ * equivalent of setting it to the plane width.
+ *
+ * @param height height (in pixels) of the YUV image.  NOTE: this is the height
+ * of the whole image, not the plane height.
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the size of the buffer (in bytes) required to hold the YUV image
+ * plane, or -1 if the arguments are out of bounds.
+ */
+DLLEXPORT unsigned long tjPlaneSizeYUV(int componentID, int width, int stride,
+                                       int height, int subsamp);
+
+
+/**
+ * The plane width of a YUV image plane with the given parameters.  Refer to
+ * @ref YUVnotes "YUV Image Format Notes" for a description of plane width.
+ *
+ * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)
+ *
+ * @param width width (in pixels) of the YUV image
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the plane width of a YUV image plane with the given parameters, or
+ * -1 if the arguments are out of bounds.
+ */
+DLLEXPORT int tjPlaneWidth(int componentID, int width, int subsamp);
+
+
+/**
+ * The plane height of a YUV image plane with the given parameters.  Refer to
+ * @ref YUVnotes "YUV Image Format Notes" for a description of plane height.
+ *
+ * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)
+ *
+ * @param height height (in pixels) of the YUV image
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the plane height of a YUV image plane with the given parameters, or
+ * -1 if the arguments are out of bounds.
+ */
+DLLEXPORT int tjPlaneHeight(int componentID, int height, int subsamp);
+
+
+/**
+ * Encode an RGB or grayscale image into a YUV planar image.  This function
+ * uses the accelerated color conversion routines in the underlying
+ * codec but does not execute any of the other steps in the JPEG compression
+ * process.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing RGB or grayscale pixels
+ * to be encoded
+ *
+ * @param width width (in pixels) of the source image
+ *
+ * @param pitch bytes per line in the source image.  Normally, this should be
+ * <tt>width * #tjPixelSizepixelFormat</tt> if the image is unpadded, or
+ * <tt>#TJPAD(width * #tjPixelSizepixelFormat)</tt> if each line of the image
+ * is padded to the nearest 32-bit boundary, as is the case for Windows
+ * bitmaps.  You can also be clever and use this parameter to skip lines, etc.
+ * Setting this parameter to 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSizepixelFormat</tt>.
+ *
+ * @param height height (in pixels) of the source image
+ *
+ * @param pixelFormat pixel format of the source image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the YUV image.
+ * Use #tjBufSizeYUV2() to determine the appropriate size for this buffer based
+ * on the image width, height, padding, and level of chrominance subsampling.
+ * The Y, U (Cb), and V (Cr) image planes will be stored sequentially in the
+ * buffer (refer to @ref YUVnotes "YUV Image Format Notes".)
+ *
+ * @param pad the width of each line in each plane of the YUV image will be
+ * padded to the nearest multiple of this number of bytes (must be a power of
+ * 2.)  To generate images suitable for X Video, <tt>pad</tt> should be set to
+ * 4.
+ *
+ * @param subsamp the level of chrominance subsampling to be used when
+ * generating the YUV image (see @ref TJSAMP
+ * "Chrominance subsampling options".)  To generate images suitable for X
+ * Video, <tt>subsamp</tt> should be set to @ref TJSAMP_420.  This produces an
+ * image compatible with the I420 (AKA "YUV420P") format.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+*/
+DLLEXPORT int tjEncodeYUV3(tjhandle handle, const unsigned char *srcBuf,
+                           int width, int pitch, int height, int pixelFormat,
+                           unsigned char *dstBuf, int pad, int subsamp,
+                           int flags);
+
+
+/**
+ * Encode an RGB or grayscale image into separate Y, U (Cb), and V (Cr) image
+ * planes.  This function uses the accelerated color conversion routines in the
+ * underlying codec but does not execute any of the other steps in the JPEG
+ * compression process.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing RGB or grayscale pixels
+ * to be encoded
+ *
+ * @param width width (in pixels) of the source image
+ *
+ * @param pitch bytes per line in the source image.  Normally, this should be
+ * <tt>width * #tjPixelSizepixelFormat</tt> if the image is unpadded, or
+ * <tt>#TJPAD(width * #tjPixelSizepixelFormat)</tt> if each line of the image
+ * is padded to the nearest 32-bit boundary, as is the case for Windows
+ * bitmaps.  You can also be clever and use this parameter to skip lines, etc.
+ * Setting this parameter to 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSizepixelFormat</tt>.
+ *
+ * @param height height (in pixels) of the source image
+ *
+ * @param pixelFormat pixel format of the source image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param dstPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if generating a grayscale image) that will receive the
+ * encoded image.  These planes can be contiguous or non-contiguous in memory.
+ * Use #tjPlaneSizeYUV() to determine the appropriate size for each plane based
+ * on the image width, height, strides, and level of chrominance subsampling.
+ * Refer to @ref YUVnotes "YUV Image Format Notes" for more details.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the output image.  Setting the stride for
+ * any plane to 0 is the same as setting it to the plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective plane widths.
+ * You can adjust the strides in order to add an arbitrary amount of line
+ * padding to each plane or to encode an RGB or grayscale image into a
+ * subregion of a larger YUV planar image.
+ *
+ * @param subsamp the level of chrominance subsampling to be used when
+ * generating the YUV image (see @ref TJSAMP
+ * "Chrominance subsampling options".)  To generate images suitable for X
+ * Video, <tt>subsamp</tt> should be set to @ref TJSAMP_420.  This produces an
+ * image compatible with the I420 (AKA "YUV420P") format.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+*/
+DLLEXPORT int tjEncodeYUVPlanes(tjhandle handle, const unsigned char *srcBuf,
+                                int width, int pitch, int height,
+                                int pixelFormat, unsigned char **dstPlanes,
+                                int *strides, int subsamp, int flags);
+
+
+/**
+ * Create a TurboJPEG decompressor instance.
+ *
+ * @return a handle to the newly-created instance, or NULL if an error
+ * occurred (see #tjGetErrorStr2().)
+*/
+DLLEXPORT tjhandle tjInitDecompress(void);
+
+
+/**
+ * Retrieve information about a JPEG image without decompressing it.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing a JPEG image
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param width pointer to an integer variable that will receive the width (in
+ * pixels) of the JPEG image
+ *
+ * @param height pointer to an integer variable that will receive the height
+ * (in pixels) of the JPEG image
+ *
+ * @param jpegSubsamp pointer to an integer variable that will receive the
+ * level of chrominance subsampling used when the JPEG image was compressed
+ * (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param jpegColorspace pointer to an integer variable that will receive one
+ * of the JPEG colorspace constants, indicating the colorspace of the JPEG
+ * image (see @ref TJCS "JPEG colorspaces".)
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+*/
+DLLEXPORT int tjDecompressHeader3(tjhandle handle,
+                                  const unsigned char *jpegBuf,
+                                  unsigned long jpegSize, int *width,
+                                  int *height, int *jpegSubsamp,
+                                  int *jpegColorspace);
+
+
+/**
+ * Returns a list of fractional scaling factors that the JPEG decompressor in
+ * this implementation of TurboJPEG supports.
+ *
+ * @param numscalingfactors pointer to an integer variable that will receive
+ * the number of elements in the list
+ *
+ * @return a pointer to a list of fractional scaling factors, or NULL if an
+ * error is encountered (see #tjGetErrorStr2().)
+*/
+DLLEXPORT tjscalingfactor *tjGetScalingFactors(int *numscalingfactors);
+
+
+/**
+ * Decompress a JPEG image to an RGB, grayscale, or CMYK image.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG image to decompress
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the decompressed
+ * image.  This buffer should normally be <tt>pitch * scaledHeight</tt> bytes
+ * in size, where <tt>scaledHeight</tt> can be determined by calling
+ * #TJSCALED() with the JPEG image height and one of the scaling factors
+ * returned by #tjGetScalingFactors().  The <tt>dstBuf</tt> pointer may also be
+ * used to decompress into a specific region of a larger buffer.
+ *
+ * @param width desired width (in pixels) of the destination image.  If this is
+ * different than the width of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired width.  If <tt>width</tt> is
+ * set to 0, then only the height will be considered when determining the
+ * scaled image size.
+ *
+ * @param pitch bytes per line in the destination image.  Normally, this is
+ * <tt>scaledWidth * #tjPixelSizepixelFormat</tt> if the decompressed image
+ * is unpadded, else <tt>#TJPAD(scaledWidth * #tjPixelSizepixelFormat)</tt>
+ * if each line of the decompressed image is padded to the nearest 32-bit
+ * boundary, as is the case for Windows bitmaps.  (NOTE: <tt>scaledWidth</tt>
+ * can be determined by calling #TJSCALED() with the JPEG image width and one
+ * of the scaling factors returned by #tjGetScalingFactors().)  You can also be
+ * clever and use the pitch parameter to skip lines, etc.  Setting this
+ * parameter to 0 is the equivalent of setting it to
+ * <tt>scaledWidth * #tjPixelSizepixelFormat</tt>.
+ *
+ * @param height desired height (in pixels) of the destination image.  If this
+ * is different than the height of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired height.  If <tt>height</tt>
+ * is set to 0, then only the width will be considered when determining the
+ * scaled image size.
+ *
+ * @param pixelFormat pixel format of the destination image (see @ref
+ * TJPF "Pixel formats".)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+ */
+DLLEXPORT int tjDecompress2(tjhandle handle, const unsigned char *jpegBuf,
+                            unsigned long jpegSize, unsigned char *dstBuf,
+                            int width, int pitch, int height, int pixelFormat,
+                            int flags);
+
+
+/**
+ * Decompress a JPEG image to a YUV planar image.  This function performs JPEG
+ * decompression but leaves out the color conversion step, so a planar YUV
+ * image is generated instead of an RGB image.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG image to decompress
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the YUV image.
+ * Use #tjBufSizeYUV2() to determine the appropriate size for this buffer based
+ * on the image width, height, padding, and level of subsampling.  The Y,
+ * U (Cb), and V (Cr) image planes will be stored sequentially in the buffer
+ * (refer to @ref YUVnotes "YUV Image Format Notes".)
+ *
+ * @param width desired width (in pixels) of the YUV image.  If this is
+ * different than the width of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired width.  If <tt>width</tt> is
+ * set to 0, then only the height will be considered when determining the
+ * scaled image size.  If the scaled width is not an even multiple of the MCU
+ * block width (see #tjMCUWidth), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param pad the width of each line in each plane of the YUV image will be
+ * padded to the nearest multiple of this number of bytes (must be a power of
+ * 2.)  To generate images suitable for X Video, <tt>pad</tt> should be set to
+ * 4.
+ *
+ * @param height desired height (in pixels) of the YUV image.  If this is
+ * different than the height of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired height.  If <tt>height</tt>
+ * is set to 0, then only the width will be considered when determining the
+ * scaled image size.  If the scaled height is not an even multiple of the MCU
+ * block height (see #tjMCUHeight), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+ */
+DLLEXPORT int tjDecompressToYUV2(tjhandle handle, const unsigned char *jpegBuf,
+                                 unsigned long jpegSize, unsigned char *dstBuf,
+                                 int width, int pad, int height, int flags);
+
+
+/**
+ * Decompress a JPEG image into separate Y, U (Cb), and V (Cr) image
+ * planes.  This function performs JPEG decompression but leaves out the color
+ * conversion step, so a planar YUV image is generated instead of an RGB image.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG image to decompress
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param dstPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if decompressing a grayscale image) that will receive
+ * the YUV image.  These planes can be contiguous or non-contiguous in memory.
+ * Use #tjPlaneSizeYUV() to determine the appropriate size for each plane based
+ * on the scaled image width, scaled image height, strides, and level of
+ * chrominance subsampling.  Refer to @ref YUVnotes "YUV Image Format Notes"
+ * for more details.
+ *
+ * @param width desired width (in pixels) of the YUV image.  If this is
+ * different than the width of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired width.  If <tt>width</tt> is
+ * set to 0, then only the height will be considered when determining the
+ * scaled image size.  If the scaled width is not an even multiple of the MCU
+ * block width (see #tjMCUWidth), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the output image.  Setting the stride for
+ * any plane to 0 is the same as setting it to the scaled plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective scaled plane
+ * widths.  You can adjust the strides in order to add an arbitrary amount of
+ * line padding to each plane or to decompress the JPEG image into a subregion
+ * of a larger YUV planar image.
+ *
+ * @param height desired height (in pixels) of the YUV image.  If this is
+ * different than the height of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired height.  If <tt>height</tt>
+ * is set to 0, then only the width will be considered when determining the
+ * scaled image size.  If the scaled height is not an even multiple of the MCU
+ * block height (see #tjMCUHeight), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+ */
+DLLEXPORT int tjDecompressToYUVPlanes(tjhandle handle,
+                                      const unsigned char *jpegBuf,
+                                      unsigned long jpegSize,
+                                      unsigned char **dstPlanes, int width,
+                                      int *strides, int height, int flags);
+
+
+/**
+ * Decode a YUV planar image into an RGB or grayscale image.  This function
+ * uses the accelerated color conversion routines in the underlying
+ * codec but does not execute any of the other steps in the JPEG decompression
+ * process.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing a YUV planar image to be
+ * decoded.  The size of this buffer should match the value returned by
+ * #tjBufSizeYUV2() for the given image width, height, padding, and level of
+ * chrominance subsampling.  The Y, U (Cb), and V (Cr) image planes should be
+ * stored sequentially in the source buffer (refer to @ref YUVnotes
+ * "YUV Image Format Notes".)
+ *
+ * @param pad Use this parameter to specify that the width of each line in each
+ * plane of the YUV source image is padded to the nearest multiple of this
+ * number of bytes (must be a power of 2.)
+ *
+ * @param subsamp the level of chrominance subsampling used in the YUV source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the decoded
+ * image.  This buffer should normally be <tt>pitch * height</tt> bytes in
+ * size, but the <tt>dstBuf</tt> pointer can also be used to decode into a
+ * specific region of a larger buffer.
+ *
+ * @param width width (in pixels) of the source and destination images
+ *
+ * @param pitch bytes per line in the destination image.  Normally, this should
+ * be <tt>width * #tjPixelSizepixelFormat</tt> if the destination image is
+ * unpadded, or <tt>#TJPAD(width * #tjPixelSizepixelFormat)</tt> if each line
+ * of the destination image should be padded to the nearest 32-bit boundary, as
+ * is the case for Windows bitmaps.  You can also be clever and use the pitch
+ * parameter to skip lines, etc.  Setting this parameter to 0 is the equivalent
+ * of setting it to <tt>width * #tjPixelSizepixelFormat</tt>.
+ *
+ * @param height height (in pixels) of the source and destination images
+ *
+ * @param pixelFormat pixel format of the destination image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+ */
+DLLEXPORT int tjDecodeYUV(tjhandle handle, const unsigned char *srcBuf,
+                          int pad, int subsamp, unsigned char *dstBuf,
+                          int width, int pitch, int height, int pixelFormat,
+                          int flags);
+
+
+/**
+ * Decode a set of Y, U (Cb), and V (Cr) image planes into an RGB or grayscale
+ * image.  This function uses the accelerated color conversion routines in the
+ * underlying codec but does not execute any of the other steps in the JPEG
+ * decompression process.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param srcPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if decoding a grayscale image) that contain a YUV image
+ * to be decoded.  These planes can be contiguous or non-contiguous in memory.
+ * The size of each plane should match the value returned by #tjPlaneSizeYUV()
+ * for the given image width, height, strides, and level of chrominance
+ * subsampling.  Refer to @ref YUVnotes "YUV Image Format Notes" for more
+ * details.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the YUV source image.  Setting the stride
+ * for any plane to 0 is the same as setting it to the plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective plane widths.
+ * You can adjust the strides in order to specify an arbitrary amount of line
+ * padding in each plane or to decode a subregion of a larger YUV planar image.
+ *
+ * @param subsamp the level of chrominance subsampling used in the YUV source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the decoded
+ * image.  This buffer should normally be <tt>pitch * height</tt> bytes in
+ * size, but the <tt>dstBuf</tt> pointer can also be used to decode into a
+ * specific region of a larger buffer.
+ *
+ * @param width width (in pixels) of the source and destination images
+ *
+ * @param pitch bytes per line in the destination image.  Normally, this should
+ * be <tt>width * #tjPixelSizepixelFormat</tt> if the destination image is
+ * unpadded, or <tt>#TJPAD(width * #tjPixelSizepixelFormat)</tt> if each line
+ * of the destination image should be padded to the nearest 32-bit boundary, as
+ * is the case for Windows bitmaps.  You can also be clever and use the pitch
+ * parameter to skip lines, etc.  Setting this parameter to 0 is the equivalent
+ * of setting it to <tt>width * #tjPixelSizepixelFormat</tt>.
+ *
+ * @param height height (in pixels) of the source and destination images
+ *
+ * @param pixelFormat pixel format of the destination image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+ */
+DLLEXPORT int tjDecodeYUVPlanes(tjhandle handle,
+                                const unsigned char **srcPlanes,
+                                const int *strides, int subsamp,
+                                unsigned char *dstBuf, int width, int pitch,
+                                int height, int pixelFormat, int flags);
+
+
+/**
+ * Create a new TurboJPEG transformer instance.
+ *
+ * @return a handle to the newly-created instance, or NULL if an error
+ * occurred (see #tjGetErrorStr2().)
+ */
+DLLEXPORT tjhandle tjInitTransform(void);
+
+
+/**
+ * Losslessly transform a JPEG image into another JPEG image.  Lossless
+ * transforms work by moving the raw DCT coefficients from one JPEG image
+ * structure to another without altering the values of the coefficients.  While
+ * this is typically faster than decompressing the image, transforming it, and
+ * re-compressing it, lossless transforms are not free.  Each lossless
+ * transform requires reading and performing Huffman decoding on all of the
+ * coefficients in the source image, regardless of the size of the destination
+ * image.  Thus, this function provides a means of generating multiple
+ * transformed images from the same source or  applying multiple
+ * transformations simultaneously, in order to eliminate the need to read the
+ * source coefficients multiple times.
+ *
+ * @param handle a handle to a TurboJPEG transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG source image to
+ * transform
+ *
+ * @param jpegSize size of the JPEG source image (in bytes)
+ *
+ * @param n the number of transformed JPEG images to generate
+ *
+ * @param dstBufs pointer to an array of n image buffers.  <tt>dstBufsi</tt>
+ * will receive a JPEG image that has been transformed using the parameters in
+ * <tt>transformsi</tt>.  TurboJPEG has the ability to reallocate the JPEG
+ * buffer to accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>dstBufsi</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize() with the transformed or cropped width and height.  Under normal
+ * circumstances, this should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees that it won't be.)  Note,
+ * however, that there are some rare cases (such as transforming images with a
+ * large amount of embedded EXIF or ICC profile data) in which the output image
+ * will be larger than the worst-case size, and #TJFLAG_NOREALLOC cannot be
+ * used in those cases.
+ * .
+ * If you choose option 1, <tt>dstSizesi</tt> should be set to the size of
+ * your pre-allocated buffer.  In any case, unless you have set
+ * #TJFLAG_NOREALLOC, you should always check <tt>dstBufsi</tt> upon return
+ * from this function, as it may have changed.
+ *
+ * @param dstSizes pointer to an array of n unsigned long variables that will
+ * receive the actual sizes (in bytes) of each transformed JPEG image.  If
+ * <tt>dstBufsi</tt> points to a pre-allocated buffer, then
+ * <tt>dstSizesi</tt> should be set to the size of the buffer.  Upon return,
+ * <tt>dstSizesi</tt> will contain the size of the JPEG image (in bytes.)
+ *
+ * @param transforms pointer to an array of n #tjtransform structures, each of
+ * which specifies the transform parameters and/or cropping region for the
+ * corresponding transformed output image.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_ACCURATEDCT
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2()
+ * and #tjGetErrorCode().)
+ */
+DLLEXPORT int tjTransform(tjhandle handle, const unsigned char *jpegBuf,
+                          unsigned long jpegSize, int n,
+                          unsigned char **dstBufs, unsigned long *dstSizes,
+                          tjtransform *transforms, int flags);
+
+
+/**
+ * Destroy a TurboJPEG compressor, decompressor, or transformer instance.
+ *
+ * @param handle a handle to a TurboJPEG compressor, decompressor or
+ * transformer instance
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2().)
+ */
+DLLEXPORT int tjDestroy(tjhandle handle);
+
+
+/**
+ * Allocate an image buffer for use with TurboJPEG.  You should always use
+ * this function to allocate the JPEG destination buffer(s) for the compression
+ * and transform functions unless you are disabling automatic buffer
+ * (re)allocation (by setting #TJFLAG_NOREALLOC.)
+ *
+ * @param bytes the number of bytes to allocate
+ *
+ * @return a pointer to a newly-allocated buffer with the specified number of
+ * bytes.
+ *
+ * @sa tjFree()
+ */
+DLLEXPORT unsigned char *tjAlloc(int bytes);
+
+
+/**
+ * Load an uncompressed image from disk into memory.
+ *
+ * @param filename name of a file containing an uncompressed image in Windows
+ * BMP or PBMPLUS (PPM/PGM) format
+ *
+ * @param width pointer to an integer variable that will receive the width (in
+ * pixels) of the uncompressed image
+ *
+ * @param align row alignment of the image buffer to be returned (must be a
+ * power of 2.)  For instance, setting this parameter to 4 will cause all rows
+ * in the image buffer to be padded to the nearest 32-bit boundary, and setting
+ * this parameter to 1 will cause all rows in the image buffer to be unpadded.
+ *
+ * @param height pointer to an integer variable that will receive the height
+ * (in pixels) of the uncompressed image
+ *
+ * @param pixelFormat pointer to an integer variable that specifies or will
+ * receive the pixel format of the uncompressed image buffer.  The behavior of
+ * #tjLoadImage() will vary depending on the value of <tt>*pixelFormat</tt>
+ * passed to the function:
+ * - @ref TJPF_UNKNOWN : The uncompressed image buffer returned by the function
+ * will use the most optimal pixel format for the file type, and
+ * <tt>*pixelFormat</tt> will contain the ID of this pixel format upon
+ * successful return from the function.
+ * - @ref TJPF_GRAY : Only PGM files and 8-bit BMP files with a grayscale
+ * colormap can be loaded.
+ * - @ref TJPF_CMYK : The RGB or grayscale pixels stored in the file will be
+ * converted using a quick & dirty algorithm that is suitable only for testing
+ * purposes (proper conversion between CMYK and other formats requires a color
+ * management system.)
+ * - Other @ref TJPF "pixel formats" : The uncompressed image buffer will use
+ * the specified pixel format, and pixel format conversion will be performed if
+ * necessary.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags".
+ *
+ * @return a pointer to a newly-allocated buffer containing the uncompressed
+ * image, converted to the chosen pixel format and with the chosen row
+ * alignment, or NULL if an error occurred (see #tjGetErrorStr2().)  This
+ * buffer should be freed using #tjFree().
+ */
+DLLEXPORT unsigned char *tjLoadImage(const char *filename, int *width,
+                                     int align, int *height, int *pixelFormat,
+                                     int flags);
+
+
+/**
+ * Save an uncompressed image from memory to disk.
+ *
+ * @param filename name of a file to which to save the uncompressed image.
+ * The image will be stored in Windows BMP or PBMPLUS (PPM/PGM) format,
+ * depending on the file extension.
+ *
+ * @param buffer pointer to an image buffer containing RGB, grayscale, or
+ * CMYK pixels to be saved
+ *
+ * @param width width (in pixels) of the uncompressed image
+ *
+ * @param pitch bytes per line in the image buffer.  Setting this parameter to
+ * 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSizepixelFormat</tt>.
+ *
+ * @param height height (in pixels) of the uncompressed image
+ *
+ * @param pixelFormat pixel format of the image buffer (see @ref TJPF
+ * "Pixel formats".)  If this parameter is set to @ref TJPF_GRAY, then the
+ * image will be stored in PGM or 8-bit (indexed color) BMP format.  Otherwise,
+ * the image will be stored in PPM or 24-bit BMP format.  If this parameter
+ * is set to @ref TJPF_CMYK, then the CMYK pixels will be converted to RGB
+ * using a quick & dirty algorithm that is suitable only for testing (proper
+ * conversion between CMYK and other formats requires a color management
+ * system.)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags".
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr2().)
+ */
+DLLEXPORT int tjSaveImage(const char *filename, unsigned char *buffer,
+                          int width, int pitch, int height, int pixelFormat,
+                          int flags);
+
+
+/**
+ * Free an image buffer previously allocated by TurboJPEG.  You should always
+ * use this function to free JPEG destination buffer(s) that were automatically
+ * (re)allocated by the compression and transform functions or that were
+ * manually allocated using #tjAlloc().
+ *
+ * @param buffer address of the buffer to free.  If the address is NULL, then
+ * this function has no effect.
+ *
+ * @sa tjAlloc()
+ */
+DLLEXPORT void tjFree(unsigned char *buffer);
+
+
+/**
+ * Returns a descriptive error message explaining why the last command failed.
+ *
+ * @param handle a handle to a TurboJPEG compressor, decompressor, or
+ * transformer instance, or NULL if the error was generated by a global
+ * function (but note that retrieving the error message for a global function
+ * is thread-safe only on platforms that support thread-local storage.)
+ *
+ * @return a descriptive error message explaining why the last command failed.
+ */
+DLLEXPORT char *tjGetErrorStr2(tjhandle handle);
+
+
+/**
+ * Returns a code indicating the severity of the last error.  See
+ * @ref TJERR "Error codes".
+ *
+ * @param handle a handle to a TurboJPEG compressor, decompressor or
+ * transformer instance
+ *
+ * @return a code indicating the severity of the last error.  See
+ * @ref TJERR "Error codes".
+ */
+DLLEXPORT int tjGetErrorCode(tjhandle handle);
+
+
+/* Deprecated functions and macros */
+#define TJFLAG_FORCEMMX  8
+#define TJFLAG_FORCESSE  16
+#define TJFLAG_FORCESSE2  32
+#define TJFLAG_FORCESSE3  128
+
+
+/* Backward compatibility functions and macros (nothing to see here) */
+#define NUMSUBOPT  TJ_NUMSAMP
+#define TJ_444  TJSAMP_444
+#define TJ_422  TJSAMP_422
+#define TJ_420  TJSAMP_420
+#define TJ_411  TJSAMP_420
+#define TJ_GRAYSCALE  TJSAMP_GRAY
+
+#define TJ_BGR  1
+#define TJ_BOTTOMUP  TJFLAG_BOTTOMUP
+#define TJ_FORCEMMX  TJFLAG_FORCEMMX
+#define TJ_FORCESSE  TJFLAG_FORCESSE
+#define TJ_FORCESSE2  TJFLAG_FORCESSE2
+#define TJ_ALPHAFIRST  64
+#define TJ_FORCESSE3  TJFLAG_FORCESSE3
+#define TJ_FASTUPSAMPLE  TJFLAG_FASTUPSAMPLE
+#define TJ_YUV  512
+
+DLLEXPORT unsigned long TJBUFSIZE(int width, int height);
+
+DLLEXPORT unsigned long TJBUFSIZEYUV(int width, int height, int jpegSubsamp);
+
+DLLEXPORT unsigned long tjBufSizeYUV(int width, int height, int subsamp);
+
+DLLEXPORT int tjCompress(tjhandle handle, unsigned char *srcBuf, int width,
+                         int pitch, int height, int pixelSize,
+                         unsigned char *dstBuf, unsigned long *compressedSize,
+                         int jpegSubsamp, int jpegQual, int flags);
+
+DLLEXPORT int tjEncodeYUV(tjhandle handle, unsigned char *srcBuf, int width,
+                          int pitch, int height, int pixelSize,
+                          unsigned char *dstBuf, int subsamp, int flags);
+
+DLLEXPORT int tjEncodeYUV2(tjhandle handle, unsigned char *srcBuf, int width,
+                           int pitch, int height, int pixelFormat,
+                           unsigned char *dstBuf, int subsamp, int flags);
+
+DLLEXPORT int tjDecompressHeader(tjhandle handle, unsigned char *jpegBuf,
+                                 unsigned long jpegSize, int *width,
+                                 int *height);
+
+DLLEXPORT int tjDecompressHeader2(tjhandle handle, unsigned char *jpegBuf,
+                                  unsigned long jpegSize, int *width,
+                                  int *height, int *jpegSubsamp);
+
+DLLEXPORT int tjDecompress(tjhandle handle, unsigned char *jpegBuf,
+                           unsigned long jpegSize, unsigned char *dstBuf,
+                           int width, int pitch, int height, int pixelSize,
+                           int flags);
+
+DLLEXPORT int tjDecompressToYUV(tjhandle handle, unsigned char *jpegBuf,
+                                unsigned long jpegSize, unsigned char *dstBuf,
+                                int flags);
+
+DLLEXPORT char *tjGetErrorStr(void);
+
+
+/**
+ * @}
+ */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/usage.txt Added

@@ -0,0 +1,686 @@
+NOTE:  This file was modified by The libjpeg-turbo Project to include only
+information relevant to libjpeg-turbo and to wordsmith certain sections.
+
+USAGE instructions for the Independent JPEG Group's JPEG software
+=================================================================
+
+This file describes usage of the JPEG conversion programs cjpeg and djpeg,
+as well as the utility programs jpegtran, rdjpgcom and wrjpgcom.  (See
+the other documentation files if you wish to use the JPEG library within
+your own programs.)
+
+If you are on a Unix machine you may prefer to read the Unix-style manual
+pages in files cjpeg.1, djpeg.1, jpegtran.1, rdjpgcom.1, wrjpgcom.1.
+
+
+INTRODUCTION
+
+These programs implement JPEG image encoding, decoding, and transcoding.
+JPEG (pronounced "jay-peg") is a standardized compression method for
+full-color and grayscale images.
+
+
+GENERAL USAGE
+
+We provide two programs, cjpeg to compress an image file into JPEG format,
+and djpeg to decompress a JPEG file back into a conventional image format.
+
+On Unix-like systems, you say:
+        cjpeg switches imagefile >jpegfile
+or
+        djpeg switches jpegfile  >imagefile
+The programs read the specified input file, or standard input if none is
+named.  They always write to standard output (with trace/error messages to
+standard error).  These conventions are handy for piping images between
+programs.
+
+On most non-Unix systems, you say:
+        cjpeg switches imagefile jpegfile
+or
+        djpeg switches jpegfile  imagefile
+i.e., both the input and output files are named on the command line.  This
+style is a little more foolproof, and it loses no functionality if you don't
+have pipes.  (You can get this style on Unix too, if you prefer, by defining
+TWO_FILE_COMMANDLINE when you compile the programs; see install.txt.)
+
+You can also say:
+        cjpeg switches -outfile jpegfile  imagefile
+or
+        djpeg switches -outfile imagefile  jpegfile
+This syntax works on all systems, so it is useful for scripts.
+
+The currently supported image file formats are: PPM (PBMPLUS color format),
+PGM (PBMPLUS grayscale format), BMP, GIF, and Targa.  cjpeg recognizes the
+input image format automatically, with the exception of some Targa files.  You
+have to tell djpeg which format to generate.
+
+JPEG files are in the defacto standard JFIF file format.  There are other,
+less widely used JPEG-based file formats, but we don't support them.
+
+All switch names may be abbreviated; for example, -grayscale may be written
+-gray or -gr.  Most of the "basic" switches can be abbreviated to as little as
+one letter.  Upper and lower case are equivalent (-BMP is the same as -bmp).
+British spellings are also accepted (e.g., -greyscale), though for brevity
+these are not mentioned below.
+
+
+CJPEG DETAILS
+
+The basic command line switches for cjpeg are:
+
+        -quality N,...  Scale quantization tables to adjust image quality.
+                          Quality is 0 (worst) to 100 (best); default is 75.
+                          (See below for more info.)
+
+        -grayscale      Create monochrome JPEG file from color input.
+                        Be sure to use this switch when compressing a grayscale
+                        BMP or GIF file, because cjpeg isn't bright enough to
+                        notice whether a BMP or GIF file uses only shades of
+                        gray.  By saying -grayscale, you'll get a smaller JPEG
+                        file that takes less time to process.
+
+        -rgb            Create RGB JPEG file.
+                        Using this switch suppresses the conversion from RGB
+                        colorspace input to the default YCbCr JPEG colorspace.
+
+        -optimize       Perform optimization of entropy encoding parameters.
+                        Without this, default encoding parameters are used.
+                        -optimize usually makes the JPEG file a little smaller,
+                        but cjpeg runs somewhat slower and needs much more
+                        memory.  Image quality and speed of decompression are
+                        unaffected by -optimize.
+
+        -progressive    Create progressive JPEG file (see below).
+
+        -targa          Input file is Targa format.  Targa files that contain
+                        an "identification" field will not be automatically
+                        recognized by cjpeg; for such files you must specify
+                        -targa to make cjpeg treat the input as Targa format.
+                        For most Targa files, you won't need this switch.
+
+The -quality switch lets you trade off compressed file size against quality of
+the reconstructed image: the higher the quality setting, the larger the JPEG
+file, and the closer the output image will be to the original input.  Normally
+you want to use the lowest quality setting (smallest file) that decompresses
+into something visually indistinguishable from the original image.  For this
+purpose the quality setting should generally be between 50 and 95 (the default
+is 75) for photographic images.  If you see defects at -quality 75, then go up
+5 or 10 counts at a time until you are happy with the output image.  (The
+optimal setting will vary from one image to another.)
+
+-quality 100 will generate a quantization table of all 1's, minimizing loss
+in the quantization step (but there is still information loss in subsampling,
+as well as roundoff error.)  For most images, specifying a quality value above
+about 95 will increase the size of the compressed file dramatically, and while
+the quality gain from these higher quality values is measurable (using metrics
+such as PSNR or SSIM), it is rarely perceivable by human vision.
+
+In the other direction, quality values below 50 will produce very small files
+of low image quality.  Settings around 5 to 10 might be useful in preparing an
+index of a large image library, for example.  Try -quality 2 (or so) for some
+amusing Cubist effects.  (Note: quality values below about 25 generate 2-byte
+quantization tables, which are considered optional in the JPEG standard.
+cjpeg emits a warning message when you give such a quality value, because some
+other JPEG programs may be unable to decode the resulting file.  Use -baseline
+if you need to ensure compatibility at low quality values.)
+
+The -quality option has been extended in this version of cjpeg to support
+separate quality settings for luminance and chrominance (or, in general,
+separate settings for every quantization table slot.)  The principle is the
+same as chrominance subsampling:  since the human eye is more sensitive to
+spatial changes in brightness than spatial changes in color, the chrominance
+components can be quantized more than the luminance components without
+incurring any visible image quality loss.  However, unlike subsampling, this
+feature reduces data in the frequency domain instead of the spatial domain,
+which allows for more fine-grained control.  This option is useful in
+quality-sensitive applications, for which the artifacts generated by
+subsampling may be unacceptable.
+
+The -quality option accepts a comma-separated list of parameters, which
+respectively refer to the quality levels that should be assigned to the
+quantization table slots.  If there are more q-table slots than parameters,
+then the last parameter is replicated.  Thus, if only one quality parameter is
+given, this is used for both luminance and chrominance (slots 0 and 1,
+respectively), preserving the legacy behavior of cjpeg v6b and prior.  More (or
+customized) quantization tables can be set with the -qtables option and
+assigned to components with the -qslots option (see the "wizard" switches
+below.)
+
+JPEG  files  generated  with separate luminance and chrominance quality are
+fully compliant with standard JPEG decoders.
+
+CAUTION: For this setting to be useful, be sure to pass an argument of
+-sample 1x1 to cjpeg to disable chrominance subsampling.  Otherwise, the
+default subsampling level (2x2, AKA "4:2:0") will be used.
+
+The -progressive switch creates a "progressive JPEG" file.  In this type of
+JPEG file, the data is stored in multiple scans of increasing quality.  If the
+file is being transmitted over a slow communications link, the decoder can use
+the first scan to display a low-quality image very quickly, and can then
+improve the display with each subsequent scan.  The final image is exactly
+equivalent to a standard JPEG file of the same quality setting, and the total
+file size is about the same --- often a little smaller.
+
+Switches for advanced users:
+
+        -arithmetic     Use arithmetic coding.  CAUTION: arithmetic coded JPEG
+                        is not yet widely implemented, so many decoders will
+                        be unable to view an arithmetic coded JPEG file at
+                        all.
+
+        -dct int        Use accurate integer DCT method (default).
+        -dct fast       Use less accurate integer DCT method legacy feature.
+                        When the Independent JPEG Group's software was first
+                        released in 1991, the compression time for a
+                        1-megapixel JPEG image on a mainstream PC was measured
+                        in minutes.  Thus, the fast integer DCT algorithm
+                        provided noticeable performance benefits.  On modern
+                        CPUs running libjpeg-turbo, however, the compression
+                        time for a 1-megapixel JPEG image is measured in
+                        milliseconds, and thus the performance benefits of the
+                        fast algorithm are much less noticeable.  On modern
+                        x86/x86-64 CPUs that support AVX2 instructions, the
+                        fast and int methods have similar performance.  On
+                        other types of CPUs, the fast method is generally about
+                        5-15% faster than the int method.
+
+                        For quality levels of 90 and below, there should be
+                        little or no perceptible quality difference between the
+                        two algorithms.  For quality levels above 90, however,
+                        the difference between the fast and int methods becomes
+                        more pronounced.  With quality=97, for instance, the
+                        fast method incurs generally about a 1-3 dB loss in
+                        PSNR relative to the int method, but this can be larger
+                        for some images.  Do not use the fast method with
+                        quality levels above 97.  The algorithm often
+                        degenerates at quality=98 and above and can actually
+                        produce a more lossy image than if lower quality levels
+                        had been used.  Also, in libjpeg-turbo, the fast method
+                        is not fully accelerated for quality levels above 97,
+                        so it will be slower than the int method.
+        -dct float      Use floating-point DCT method legacy feature.
+                        The float method does not produce significantly more
+                        accurate results than the int method, and it is much
+                        slower.  The float method may also give different
+                        results on different machines due to varying roundoff
+                        behavior, whereas the integer methods should give the
+                        same results on all machines.
+
+        -restart N      Emit a JPEG restart marker every N MCU rows, or every
+                        N MCU blocks if "B" is attached to the number.
+                        -restart 0 (the default) means no restart markers.
+
+        -smooth N       Smooth the input image to eliminate dithering noise.
+                        N, ranging from 1 to 100, indicates the strength of
+                        smoothing.  0 (the default) means no smoothing.
+
+        -maxmemory N    Set limit for amount of memory to use in processing
+                        large images.  Value is in thousands of bytes, or
+                        millions of bytes if "M" is attached to the number.
+                        For example, -max 4m selects 4000000 bytes.  If more
+                        space is needed, an error will occur.
+
+        -verbose        Enable debug printout.  More -v's give more printout.
+        or  -debug      Also, version information is printed at startup.
+
+The -restart option inserts extra markers that allow a JPEG decoder to
+resynchronize after a transmission error.  Without restart markers, any damage
+to a compressed file will usually ruin the image from the point of the error
+to the end of the image; with restart markers, the damage is usually confined
+to the portion of the image up to the next restart marker.  Of course, the
+restart markers occupy extra space.  We recommend -restart 1 for images that
+will be transmitted across unreliable networks such as Usenet.
+
+The -smooth option filters the input to eliminate fine-scale noise.  This is
+often useful when converting dithered images to JPEG: a moderate smoothing
+factor of 10 to 50 gets rid of dithering patterns in the input file, resulting
+in a smaller JPEG file and a better-looking image.  Too large a smoothing
+factor will visibly blur the image, however.
+
+Switches for wizards:
+
+        -baseline       Force baseline-compatible quantization tables to be
+                        generated.  This clamps quantization values to 8 bits
+                        even at low quality settings.  (This switch is poorly
+                        named, since it does not ensure that the output is
+                        actually baseline JPEG.  For example, you can use
+                        -baseline and -progressive together.)
+
+        -qtables file   Use the quantization tables given in the specified
+                        text file.
+
+        -qslots N,... Select which quantization table to use for each color
+                        component.
+
+        -sample HxV,...  Set JPEG sampling factors for each color component.
+
+        -scans file     Use the scan script given in the specified text file.
+
+The "wizard" switches are intended for experimentation with JPEG.  If you
+don't know what you are doing, DON'T USE THEM.  These switches are documented
+further in the file wizard.txt.
+
+
+DJPEG DETAILS
+
+The basic command line switches for djpeg are:
+
+        -colors N       Reduce image to at most N colors.  This reduces the
+        or -quantize N  number of colors used in the output image, so that it
+                        can be displayed on a colormapped display or stored in
+                        a colormapped file format.  For example, if you have
+                        an 8-bit display, you'd need to reduce to 256 or fewer
+                        colors.  (-colors is the recommended name, -quantize
+                        is provided only for backwards compatibility.)
+
+        -fast           Select recommended processing options for fast, low
+                        quality output.  (The default options are chosen for
+                        highest quality output.)  Currently, this is equivalent
+                        to "-dct fast -nosmooth -onepass -dither ordered".
+
+        -grayscale      Force grayscale output even if JPEG file is color.
+                        Useful for viewing on monochrome displays; also,
+                        djpeg runs noticeably faster in this mode.
+
+        -rgb            Force RGB output even if JPEG file is grayscale.
+
+        -scale M/N      Scale the output image by a factor M/N.  Currently
+                        the scale factor must be M/8, where M is an integer
+                        between 1 and 16 inclusive, or any reduced fraction
+                        thereof (such as 1/2, 3/4, etc.  Scaling is handy if
+                        the image is larger than your screen; also, djpeg runs
+                        much faster when scaling down the output.
+
+        -bmp            Select BMP output format (Windows flavor).  8-bit
+                        colormapped format is emitted if -colors or -grayscale
+                        is specified, or if the JPEG file is grayscale;
+                        otherwise, 24-bit full-color format is emitted.
+
+        -gif            Select GIF output format (LZW-compressed).  Since GIF
+                        does not support more than 256 colors, -colors 256 is
+                        assumed (unless you specify a smaller number of
+                        colors).  If you specify -fast, the default number of
+                        colors is 216.
+
+        -gif0           Select GIF output format (uncompressed).  Since GIF
+                        does not support more than 256 colors, -colors 256 is
+                        assumed (unless you specify a smaller number of
+                        colors).  If you specify -fast, the default number of
+                        colors is 216.
+
+        -os2            Select BMP output format (OS/2 1.x flavor).  8-bit
+                        colormapped format is emitted if -colors or -grayscale
+                        is specified, or if the JPEG file is grayscale;
+                        otherwise, 24-bit full-color format is emitted.
+
+        -pnm            Select PBMPLUS (PPM/PGM) output format (this is the
+                        default format).  PGM is emitted if the JPEG file is
+                        grayscale or if -grayscale is specified; otherwise
+                        PPM is emitted.
+
+        -targa          Select Targa output format.  Grayscale format is
+                        emitted if the JPEG file is grayscale or if
+                        -grayscale is specified; otherwise, colormapped format
+                        is emitted if -colors is specified; otherwise, 24-bit
+                        full-color format is emitted.
+
+Switches for advanced users:
+
+        -dct int        Use accurate integer DCT method (default).
+        -dct fast       Use less accurate integer DCT method legacy feature.
+                        When the Independent JPEG Group's software was first
+                        released in 1991, the decompression time for a
+                        1-megapixel JPEG image on a mainstream PC was measured
+                        in minutes.  Thus, the fast integer DCT algorithm
+                        provided noticeable performance benefits.  On modern
+                        CPUs running libjpeg-turbo, however, the decompression
+                        time for a 1-megapixel JPEG image is measured in
+                        milliseconds, and thus the performance benefits of the
+                        fast algorithm are much less noticeable.  On modern
+                        x86/x86-64 CPUs that support AVX2 instructions, the
+                        fast and int methods have similar performance.  On
+                        other types of CPUs, the fast method is generally about
+                        5-15% faster than the int method.
+
+                        If the JPEG image was compressed using a quality level
+                        of 85 or below, then there should be little or no
+                        perceptible quality difference between the two
+                        algorithms.  When decompressing images that were
+                        compressed using quality levels above 85, however, the
+                        difference between the fast and int methods becomes
+                        more pronounced.  With images compressed using
+                        quality=97, for instance, the fast method incurs
+                        generally about a 4-6 dB loss in PSNR relative to the
+                        int method, but this can be larger for some images.  If
+                        you can avoid it, do not use the fast method when
+                        decompressing images that were compressed using quality
+                        levels above 97.  The algorithm often degenerates for
+                        such images and can actually produce a more lossy
+                        output image than if the JPEG image had been compressed
+                        using lower quality levels.
+        -dct float      Use floating-point DCT method legacy feature.
+                        The float method does not produce significantly more
+                        accurate results than the int method, and it is much
+                        slower.  The float method may also give different
+                        results on different machines due to varying roundoff
+                        behavior, whereas the integer methods should give the
+                        same results on all machines.
+
+        -dither fs      Use Floyd-Steinberg dithering in color quantization.
+        -dither ordered Use ordered dithering in color quantization.
+        -dither none    Do not use dithering in color quantization.
+                        By default, Floyd-Steinberg dithering is applied when
+                        quantizing colors; this is slow but usually produces
+                        the best results.  Ordered dither is a compromise
+                        between speed and quality; no dithering is fast but
+                        usually looks awful.  Note that these switches have
+                        no effect unless color quantization is being done.
+                        Ordered dither is only available in -onepass mode.
+
+        -map FILE       Quantize to the colors used in the specified image
+                        file.  This is useful for producing multiple files
+                        with identical color maps, or for forcing a predefined
+                        set of colors to be used.  The FILE must be a GIF
+                        or PPM file.  This option overrides -colors and
+                        -onepass.
+
+        -nosmooth       Use a faster, lower-quality upsampling routine.
+
+        -onepass        Use one-pass instead of two-pass color quantization.
+                        The one-pass method is faster and needs less memory,
+                        but it produces a lower-quality image.  -onepass is
+                        ignored unless you also say -colors N.  Also,
+                        the one-pass method is always used for grayscale
+                        output (the two-pass method is no improvement then).
+
+        -maxmemory N    Set limit for amount of memory to use in processing
+                        large images.  Value is in thousands of bytes, or
+                        millions of bytes if "M" is attached to the number.
+                        For example, -max 4m selects 4000000 bytes.  If more
+                        space is needed, an error will occur.
+
+        -verbose        Enable debug printout.  More -v's give more printout.
+        or  -debug      Also, version information is printed at startup.
+
+
+HINTS FOR CJPEG
+
+Color GIF files are not the ideal input for JPEG; JPEG is really intended for
+compressing full-color (24-bit) images.  In particular, don't try to convert
+cartoons, line drawings, and other images that have only a few distinct
+colors.  GIF works great on these, JPEG does not.  If you want to convert a
+GIF to JPEG, you should experiment with cjpeg's -quality and -smooth options
+to get a satisfactory conversion.  -smooth 10 or so is often helpful.
+
+Avoid running an image through a series of JPEG compression/decompression
+cycles.  Image quality loss will accumulate; after ten or so cycles the image
+may be noticeably worse than it was after one cycle.  It's best to use a
+lossless format while manipulating an image, then convert to JPEG format when
+you are ready to file the image away.
+
+The -optimize option to cjpeg is worth using when you are making a "final"
+version for posting or archiving.  It's also a win when you are using low
+quality settings to make very small JPEG files; the percentage improvement
+is often a lot more than it is on larger files.  (At present, -optimize
+mode is always selected when generating progressive JPEG files.)
+
+
+HINTS FOR DJPEG
+
+To get a quick preview of an image, use the -grayscale and/or -scale switches.
+"-grayscale -scale 1/8" is the fastest case.
+
+Several options are available that trade off image quality to gain speed.
+"-fast" turns on the recommended settings.
+
+"-dct fast" and/or "-nosmooth" gain speed at a small sacrifice in quality.
+When producing a color-quantized image, "-onepass -dither ordered" is fast but
+much lower quality than the default behavior.  "-dither none" may give
+acceptable results in two-pass mode, but is seldom tolerable in one-pass mode.
+
+
+HINTS FOR BOTH PROGRAMS
+
+If the memory needed by cjpeg or djpeg exceeds the limit specified by
+-maxmemory, an error will occur.  You can leave out -progressive and -optimize
+(for cjpeg) or specify -onepass (for djpeg) to reduce memory usage.
+
+On machines that have "environment" variables, you can define the environment
+variable JPEGMEM to set the default memory limit.  The value is specified as
+described for the -maxmemory switch.  JPEGMEM overrides the default value
+specified when the program was compiled, and itself is overridden by an
+explicit -maxmemory switch.
+
+
+JPEGTRAN
+
+jpegtran performs various useful transformations of JPEG files.
+It can translate the coded representation from one variant of JPEG to another,
+for example from baseline JPEG to progressive JPEG or vice versa.  It can also
+perform some rearrangements of the image data, for example turning an image
+from landscape to portrait format by rotation.  For EXIF files and JPEG files
+containing Exif data, you may prefer to use exiftran instead.
+
+jpegtran works by rearranging the compressed data (DCT coefficients), without
+ever fully decoding the image.  Therefore, its transformations are lossless:
+there is no image degradation at all, which would not be true if you used
+djpeg followed by cjpeg to accomplish the same conversion.  But by the same
+token, jpegtran cannot perform lossy operations such as changing the image
+quality.  However, while the image data is losslessly transformed, metadata
+can be removed.  See the -copy option for specifics.
+
+jpegtran uses a command line syntax similar to cjpeg or djpeg.
+On Unix-like systems, you say:
+        jpegtran switches inputfile >outputfile
+On most non-Unix systems, you say:
+        jpegtran switches inputfile outputfile
+where both the input and output files are JPEG files.
+
+To specify the coded JPEG representation used in the output file,
+jpegtran accepts a subset of the switches recognized by cjpeg:
+        -optimize       Perform optimization of entropy encoding parameters.
+        -progressive    Create progressive JPEG file.
+        -arithmetic     Use arithmetic coding.
+        -restart N      Emit a JPEG restart marker every N MCU rows, or every
+                        N MCU blocks if "B" is attached to the number.
+        -scans file     Use the scan script given in the specified text file.
+See the previous discussion of cjpeg for more details about these switches.
+If you specify none of these switches, you get a plain baseline-JPEG output
+file.  The quality setting and so forth are determined by the input file.
+
+The image can be losslessly transformed by giving one of these switches:
+        -flip horizontal        Mirror image horizontally (left-right).
+        -flip vertical          Mirror image vertically (top-bottom).
+        -rotate 90              Rotate image 90 degrees clockwise.
+        -rotate 180             Rotate image 180 degrees.
+        -rotate 270             Rotate image 270 degrees clockwise (or 90 ccw).
+        -transpose              Transpose image (across UL-to-LR axis).
+        -transverse             Transverse transpose (across UR-to-LL axis).
+
+The transpose transformation has no restrictions regarding image dimensions.
+The other transformations operate rather oddly if the image dimensions are not
+a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
+transform complete blocks of DCT coefficient data in the desired way.
+
+jpegtran's default behavior when transforming an odd-size image is designed
+to preserve exact reversibility and mathematical consistency of the
+transformation set.  As stated, transpose is able to flip the entire image
+area.  Horizontal mirroring leaves any partial iMCU column at the right edge
+untouched, but is able to flip all rows of the image.  Similarly, vertical
+mirroring leaves any partial iMCU row at the bottom edge untouched, but is
+able to flip all columns.  The other transforms can be built up as sequences
+of transpose and flip operations; for consistency, their actions on edge
+pixels are defined to be the same as the end result of the corresponding
+transpose-and-flip sequence.
+
+For practical use, you may prefer to discard any untransformable edge pixels
+rather than having a strange-looking strip along the right and/or bottom edges
+of a transformed image.  To do this, add the -trim switch:
+        -trim           Drop non-transformable edge blocks.
+Obviously, a transformation with -trim is not reversible, so strictly speaking
+jpegtran with this switch is not lossless.  Also, the expected mathematical
+equivalences between the transformations no longer hold.  For example,
+"-rot 270 -trim" trims only the bottom edge, but "-rot 90 -trim" followed by
+"-rot 180 -trim" trims both edges.
+
+If you are only interested in perfect transformations, add the -perfect switch:
+        -perfect        Fail with an error if the transformation is not
+                        perfect.
+For example, you may want to do
+  jpegtran -rot 90 -perfect foo.jpg || djpeg foo.jpg | pnmflip -r90 | cjpeg
+to do a perfect rotation, if available, or an approximated one if not.
+
+This version of jpegtran also offers a lossless crop option, which discards
+data outside of a given image region but losslessly preserves what is inside.
+Like the rotate and flip transforms, lossless crop is restricted by the current
+JPEG format; the upper left corner of the selected region must fall on an iMCU
+boundary.  If it doesn't, then it is silently moved up and/or left to the
+nearest iMCU boundary (the lower right corner is unchanged.)  Thus, the output
+image covers at least the requested region, but it may cover more.  The
+adjustment of the region dimensions may be optionally disabled by attaching an
+'f' character ("force") to the width or height number.
+
+The image can be losslessly cropped by giving the switch:
+        -crop WxH+X+Y   Crop to a rectangular region of width W and height H,
+                        starting at point X,Y.
+
+If W or H is larger than the width/height of the input image, then the output
+image is expanded in size, and the expanded region is filled in with zeros
+(neutral gray).  Attaching an 'f' character ("flatten") to the width number
+will cause each block in the expanded region to be filled in with the DC
+coefficient of the nearest block in the input image rather than grayed out.
+Attaching an 'r' character ("reflect") to the width number will cause the
+expanded region to be filled in with repeated reflections of the input image
+rather than grayed out.
+
+A complementary lossless wipe option is provided to discard (gray out) data
+inside a given image region while losslessly preserving what is outside:
+        -wipe WxH+X+Y   Wipe (gray out) a rectangular region of width W and
+                        height H from the input image, starting at point X,Y.
+
+Attaching an 'f' character ("flatten") to the width number will cause the
+region to be filled with the average of adjacent blocks rather than grayed out.
+If the wipe region and the region outside the wipe region, when adjusted to the
+nearest iMCU boundary, form two horizontally adjacent rectangles, then
+attaching an 'r' character ("reflect") to the width number will cause the wipe
+region to be filled with repeated reflections of the outside region rather than
+grayed out.
+
+A lossless drop option is also provided, which allows another JPEG image to be
+inserted ("dropped") into the input image data at a given position, replacing
+the existing image data at that position:
+        -drop +X+Y filename     Drop (insert) another image at point X,Y
+
+Both the input image and the drop image must have the same subsampling level.
+It is best if they also have the same quantization (quality.)  Otherwise, the
+quantization of the output image will be adapted to accommodate the higher of
+the input image quality and the drop image quality.  The trim option can be
+used with the drop option to requantize the drop image to match the input
+image.  Note that a grayscale image can be dropped into a full-color image or
+vice versa, as long as the full-color image has no vertical subsampling.  If
+the input image is grayscale and the drop image is full-color, then the
+chrominance channels from the drop image will be discarded.
+
+Other not-strictly-lossless transformation switches are:
+
+        -grayscale      Force grayscale output.
+This option discards the chrominance channels if the input image is YCbCr
+(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
+luminance channel is preserved exactly, so this is a better method of reducing
+to grayscale than decompression, conversion, and recompression.  This switch
+is particularly handy for fixing a monochrome picture that was mistakenly
+encoded as a color JPEG.  (In such a case, the space savings from getting rid
+of the near-empty chroma channels won't be large; but the decoding time for
+a grayscale JPEG is substantially less than that for a color JPEG.)
+
+jpegtran also recognizes these switches that control what to do with "extra"
+markers, such as comment blocks:
+        -copy none      Copy no extra markers from source file.  This setting
+                        suppresses all comments and other metadata in the
+                        source file.
+        -copy comments  Copy only comment markers.  This setting copies
+                        comments from the source file but discards any other
+                        metadata.
+        -copy icc       Copy only ICC profile markers.  This setting copies the
+                        ICC profile from the source file but discards any other
+                        metadata.
+        -copy all       Copy all extra markers.  This setting preserves
+                        miscellaneous markers found in the source file, such
+                        as JFIF thumbnails, Exif data, and Photoshop settings.
+                        In some files, these extra markers can be sizable.
+                        Note that this option will copy thumbnails as-is;
+                        they will not be transformed.
+The default behavior is -copy comments.  (Note: in IJG releases v6 and v6a,
+jpegtran always did the equivalent of -copy none.)
+
+Additional switches recognized by jpegtran are:
+        -outfile filename
+        -maxmemory N
+        -verbose
+        -debug
+These work the same as in cjpeg or djpeg.
+
+
+THE COMMENT UTILITIES
+
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+
+We provide two utility programs to display COM block contents and add COM
+blocks to a JPEG file.
+
+rdjpgcom searches a JPEG file and prints the contents of any COM blocks on
+standard output.  The command line syntax is
+        rdjpgcom -raw -verbose inputfilename
+The switch "-raw" (or just "-r") causes rdjpgcom to output non-printable
+characters in JPEG comments.  These characters are normally escaped for
+security reasons.
+The switch "-verbose" (or just "-v") causes rdjpgcom to also display the JPEG
+image dimensions.  If you omit the input file name from the command line,
+the JPEG file is read from standard input.  (This may not work on some
+operating systems, if binary data can't be read from stdin.)
+
+wrjpgcom adds a COM block, containing text you provide, to a JPEG file.
+Ordinarily, the COM block is added after any existing COM blocks, but you
+can delete the old COM blocks if you wish.  wrjpgcom produces a new JPEG
+file; it does not modify the input file.  DO NOT try to overwrite the input
+file by directing wrjpgcom's output back into it; on most systems this will
+just destroy your file.
+
+The command line syntax for wrjpgcom is similar to cjpeg's.  On Unix-like
+systems, it is
+        wrjpgcom switches inputfilename
+The output file is written to standard output.  The input file comes from
+the named file, or from standard input if no input file is named.
+
+On most non-Unix systems, the syntax is
+        wrjpgcom switches inputfilename outputfilename
+where both input and output file names must be given explicitly.
+
+wrjpgcom understands three switches:
+        -replace                 Delete any existing COM blocks from the file.
+        -comment "Comment text"  Supply new COM text on command line.
+        -cfile name              Read text for new COM block from named file.
+(Switch names can be abbreviated.)  If you have only one line of comment text
+to add, you can provide it on the command line with -comment.  The comment
+text must be surrounded with quotes so that it is treated as a single
+argument.  Longer comments can be read from a text file.
+
+If you give neither -comment nor -cfile, then wrjpgcom will read the comment
+text from standard input.  (In this case an input image file name MUST be
+supplied, so that the source JPEG file comes from somewhere else.)  You can
+enter multiple lines, up to 64KB worth.  Type an end-of-file indicator
+(usually control-D or control-Z) to terminate the comment text entry.
+
+wrjpgcom will not add a COM block if the provided comment string is empty.
+Therefore -replace -comment "" can be used to delete all COM blocks from a
+file.
+
+These utility programs do not depend on the IJG JPEG library.  In
+particular, the source code for rdjpgcom is intended as an illustration of
+the minimum amount of code required to parse a JPEG file header correctly.

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/gcc Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/gcc/projectTargets-release.cmake.in Added

@@ -0,0 +1,49 @@
+#----------------------------------------------------------------
+# Generated CMake target import file for configuration "Release".
+#----------------------------------------------------------------
+
+# Commands may need to know the format version.
+set(CMAKE_IMPORT_FILE_VERSION 1)
+
+# Import target "@CMAKE_PROJECT_NAME@::jpeg" for configuration "Release"
+set_property(TARGET @CMAKE_PROJECT_NAME@::jpeg APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+set_target_properties(@CMAKE_PROJECT_NAME@::jpeg PROPERTIES
+  IMPORTED_IMPLIB_RELEASE "${_IMPORT_PREFIX}/lib/libjpeg.dll.a"
+  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/bin/libjpeg-62.dll"
+  )
+
+list(APPEND _IMPORT_CHECK_TARGETS @CMAKE_PROJECT_NAME@::jpeg )
+list(APPEND _IMPORT_CHECK_FILES_FOR_@CMAKE_PROJECT_NAME@::jpeg "${_IMPORT_PREFIX}/lib/libjpeg.dll.a" "${_IMPORT_PREFIX}/bin/libjpeg-62.dll" )
+
+# Import target "@CMAKE_PROJECT_NAME@::turbojpeg" for configuration "Release"
+set_property(TARGET @CMAKE_PROJECT_NAME@::turbojpeg APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+set_target_properties(@CMAKE_PROJECT_NAME@::turbojpeg PROPERTIES
+  IMPORTED_IMPLIB_RELEASE "${_IMPORT_PREFIX}/lib/libturbojpeg.dll.a"
+  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/bin/libturbojpeg.dll"
+  )
+
+list(APPEND _IMPORT_CHECK_TARGETS @CMAKE_PROJECT_NAME@::turbojpeg )
+list(APPEND _IMPORT_CHECK_FILES_FOR_@CMAKE_PROJECT_NAME@::turbojpeg "${_IMPORT_PREFIX}/lib/libturbojpeg.dll.a" "${_IMPORT_PREFIX}/bin/libturbojpeg.dll" )
+
+# Import target "@CMAKE_PROJECT_NAME@::turbojpeg-static" for configuration "Release"
+set_property(TARGET @CMAKE_PROJECT_NAME@::turbojpeg-static APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+set_target_properties(@CMAKE_PROJECT_NAME@::turbojpeg-static PROPERTIES
+  IMPORTED_LINK_INTERFACE_LANGUAGES_RELEASE "ASM_NASM;C"
+  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/lib/libturbojpeg.a"
+  )
+
+list(APPEND _IMPORT_CHECK_TARGETS @CMAKE_PROJECT_NAME@::turbojpeg-static )
+list(APPEND _IMPORT_CHECK_FILES_FOR_@CMAKE_PROJECT_NAME@::turbojpeg-static "${_IMPORT_PREFIX}/lib/libturbojpeg.a" )
+
+# Import target "@CMAKE_PROJECT_NAME@::jpeg-static" for configuration "Release"
+set_property(TARGET @CMAKE_PROJECT_NAME@::jpeg-static APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+set_target_properties(@CMAKE_PROJECT_NAME@::jpeg-static PROPERTIES
+  IMPORTED_LINK_INTERFACE_LANGUAGES_RELEASE "ASM_NASM;C"
+  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/lib/libjpeg.a"
+  )
+
+list(APPEND _IMPORT_CHECK_TARGETS @CMAKE_PROJECT_NAME@::jpeg-static )
+list(APPEND _IMPORT_CHECK_FILES_FOR_@CMAKE_PROJECT_NAME@::jpeg-static "${_IMPORT_PREFIX}/lib/libjpeg.a" )
+
+# Commands beyond this point should not need to know the version.
+set(CMAKE_IMPORT_FILE_VERSION)

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/jconfig.h.in Added

@@ -0,0 +1,33 @@
+#define JPEG_LIB_VERSION  @JPEG_LIB_VERSION@
+#define LIBJPEG_TURBO_VERSION  @VERSION@
+#define LIBJPEG_TURBO_VERSION_NUMBER  @LIBJPEG_TURBO_VERSION_NUMBER@
+
+#cmakedefine C_ARITH_CODING_SUPPORTED
+#cmakedefine D_ARITH_CODING_SUPPORTED
+#cmakedefine MEM_SRCDST_SUPPORTED
+#cmakedefine WITH_SIMD
+
+#define BITS_IN_JSAMPLE  @BITS_IN_JSAMPLE@      /* use 8 or 12 */
+
+#define HAVE_STDDEF_H
+#define HAVE_STDLIB_H
+#undef NEED_SYS_TYPES_H
+#undef NEED_BSD_STRINGS
+
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+#undef INCOMPLETE_TYPES_BROKEN
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+/* Define "boolean" as unsigned char, not int, per Windows custom */
+#ifndef __RPCNDR_H__            /* don't conflict if rpcndr.h already read */
+typedef unsigned char boolean;
+#endif
+#define HAVE_BOOLEAN            /* prevent jmorecfg.h from redefining it */
+
+/* Define "INT32" as int, not long, per Windows custom */
+#if !(defined(_BASETSD_H_) || defined(_BASETSD_H))   /* don't conflict if basetsd.h already read */
+typedef short INT16;
+typedef signed int INT32;
+#endif
+#define XMD_H                   /* prevent jmorecfg.h from redefining it */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/jpeg62-memsrcdst.def Added

@@ -0,0 +1,108 @@
+EXPORTS
+  jcopy_block_row @ 1 ;
+  jcopy_sample_rows @ 2 ;
+  jdiv_round_up @ 3 ;
+  jinit_1pass_quantizer @ 4 ;
+  jinit_2pass_quantizer @ 5 ;
+  jinit_c_coef_controller @ 6 ;
+  jinit_c_main_controller @ 7 ;
+  jinit_c_master_control @ 8 ;
+  jinit_c_prep_controller @ 9 ;
+  jinit_color_converter @ 10 ;
+  jinit_color_deconverter @ 11 ;
+  jinit_compress_master @ 12 ;
+  jinit_d_coef_controller @ 13 ;
+  jinit_d_main_controller @ 14 ;
+  jinit_d_post_controller @ 15 ;
+  jinit_downsampler @ 16 ;
+  jinit_forward_dct @ 17 ;
+  jinit_huff_decoder @ 18 ;
+  jinit_huff_encoder @ 19 ;
+  jinit_input_controller @ 20 ;
+  jinit_inverse_dct @ 21 ;
+  jinit_marker_reader @ 22 ;
+  jinit_marker_writer @ 23 ;
+  jinit_master_decompress @ 24 ;
+  jinit_memory_mgr @ 25 ;
+  jinit_merged_upsampler @ 26 ;
+  jinit_phuff_decoder @ 27 ;
+  jinit_phuff_encoder @ 28 ;
+  jinit_upsampler @ 29 ;
+  jpeg_CreateCompress @ 30 ;
+  jpeg_CreateDecompress @ 31 ;
+  jpeg_abort @ 32 ;
+  jpeg_abort_compress @ 33 ;
+  jpeg_abort_decompress @ 34 ;
+  jpeg_add_quant_table @ 35 ;
+  jpeg_alloc_huff_table @ 36 ;
+  jpeg_alloc_quant_table @ 37 ;
+  jpeg_calc_output_dimensions @ 38 ;
+  jpeg_consume_input @ 39 ;
+  jpeg_copy_critical_parameters @ 40 ;
+  jpeg_default_colorspace @ 41 ;
+  jpeg_destroy @ 42 ;
+  jpeg_destroy_compress @ 43 ;
+  jpeg_destroy_decompress @ 44 ;
+  jpeg_fdct_float @ 45 ;
+  jpeg_fdct_ifast @ 46 ;
+  jpeg_fdct_islow @ 47 ;
+  jpeg_fill_bit_buffer @ 48 ;
+  jpeg_finish_compress @ 49 ;
+  jpeg_finish_decompress @ 50 ;
+  jpeg_finish_output @ 51 ;
+  jpeg_free_large @ 52 ;
+  jpeg_free_small @ 53 ;
+  jpeg_gen_optimal_table @ 54 ;
+  jpeg_get_large @ 55 ;
+  jpeg_get_small @ 56 ;
+  jpeg_has_multiple_scans @ 57 ;
+  jpeg_huff_decode @ 58 ;
+  jpeg_idct_1x1 @ 59 ;
+  jpeg_idct_2x2 @ 60 ;
+  jpeg_idct_4x4 @ 61 ;
+  jpeg_idct_float @ 62 ;
+  jpeg_idct_ifast @ 63 ;
+  jpeg_idct_islow @ 64 ;
+  jpeg_input_complete @ 65 ;
+  jpeg_make_c_derived_tbl @ 66 ;
+  jpeg_make_d_derived_tbl @ 67 ;
+  jpeg_mem_available @ 68 ;
+  jpeg_mem_init @ 69 ;
+  jpeg_mem_term @ 70 ;
+  jpeg_new_colormap @ 71 ;
+  jpeg_open_backing_store @ 72 ;
+  jpeg_quality_scaling @ 73 ;
+  jpeg_read_coefficients @ 74 ;
+  jpeg_read_header @ 75 ;
+  jpeg_read_raw_data @ 76 ;
+  jpeg_read_scanlines @ 77 ;
+  jpeg_resync_to_restart @ 78 ;
+  jpeg_save_markers @ 79 ;
+  jpeg_set_colorspace @ 80 ;
+  jpeg_set_defaults @ 81 ;
+  jpeg_set_linear_quality @ 82 ;
+  jpeg_set_marker_processor @ 83 ;
+  jpeg_set_quality @ 84 ;
+  jpeg_simple_progression @ 85 ;
+  jpeg_start_compress @ 86 ;
+  jpeg_start_decompress @ 87 ;
+  jpeg_start_output @ 88 ;
+  jpeg_std_error @ 89 ;
+  jpeg_stdio_dest @ 90 ;
+  jpeg_stdio_src @ 91 ;
+  jpeg_suppress_tables @ 92 ;
+  jpeg_write_coefficients @ 93 ;
+  jpeg_write_m_byte @ 94 ;
+  jpeg_write_m_header @ 95 ;
+  jpeg_write_marker @ 96 ;
+  jpeg_write_raw_data @ 97 ;
+  jpeg_write_scanlines @ 98 ;
+  jpeg_write_tables @ 99 ;
+  jround_up @ 100 ;
+  jzero_far @ 101 ;
+  jpeg_mem_dest @ 102 ;
+  jpeg_mem_src @ 103 ;
+  jpeg_skip_scanlines @ 104 ;
+  jpeg_crop_scanline @ 105 ;
+  jpeg_read_icc_profile @ 106 ;
+  jpeg_write_icc_profile @ 107 ;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/jpeg62.def Added

@@ -0,0 +1,106 @@
+EXPORTS
+  jcopy_block_row @ 1 ;
+  jcopy_sample_rows @ 2 ;
+  jdiv_round_up @ 3 ;
+  jinit_1pass_quantizer @ 4 ;
+  jinit_2pass_quantizer @ 5 ;
+  jinit_c_coef_controller @ 6 ;
+  jinit_c_main_controller @ 7 ;
+  jinit_c_master_control @ 8 ;
+  jinit_c_prep_controller @ 9 ;
+  jinit_color_converter @ 10 ;
+  jinit_color_deconverter @ 11 ;
+  jinit_compress_master @ 12 ;
+  jinit_d_coef_controller @ 13 ;
+  jinit_d_main_controller @ 14 ;
+  jinit_d_post_controller @ 15 ;
+  jinit_downsampler @ 16 ;
+  jinit_forward_dct @ 17 ;
+  jinit_huff_decoder @ 18 ;
+  jinit_huff_encoder @ 19 ;
+  jinit_input_controller @ 20 ;
+  jinit_inverse_dct @ 21 ;
+  jinit_marker_reader @ 22 ;
+  jinit_marker_writer @ 23 ;
+  jinit_master_decompress @ 24 ;
+  jinit_memory_mgr @ 25 ;
+  jinit_merged_upsampler @ 26 ;
+  jinit_phuff_decoder @ 27 ;
+  jinit_phuff_encoder @ 28 ;
+  jinit_upsampler @ 29 ;
+  jpeg_CreateCompress @ 30 ;
+  jpeg_CreateDecompress @ 31 ;
+  jpeg_abort @ 32 ;
+  jpeg_abort_compress @ 33 ;
+  jpeg_abort_decompress @ 34 ;
+  jpeg_add_quant_table @ 35 ;
+  jpeg_alloc_huff_table @ 36 ;
+  jpeg_alloc_quant_table @ 37 ;
+  jpeg_calc_output_dimensions @ 38 ;
+  jpeg_consume_input @ 39 ;
+  jpeg_copy_critical_parameters @ 40 ;
+  jpeg_default_colorspace @ 41 ;
+  jpeg_destroy @ 42 ;
+  jpeg_destroy_compress @ 43 ;
+  jpeg_destroy_decompress @ 44 ;
+  jpeg_fdct_float @ 45 ;
+  jpeg_fdct_ifast @ 46 ;
+  jpeg_fdct_islow @ 47 ;
+  jpeg_fill_bit_buffer @ 48 ;
+  jpeg_finish_compress @ 49 ;
+  jpeg_finish_decompress @ 50 ;
+  jpeg_finish_output @ 51 ;
+  jpeg_free_large @ 52 ;
+  jpeg_free_small @ 53 ;
+  jpeg_gen_optimal_table @ 54 ;
+  jpeg_get_large @ 55 ;
+  jpeg_get_small @ 56 ;
+  jpeg_has_multiple_scans @ 57 ;
+  jpeg_huff_decode @ 58 ;
+  jpeg_idct_1x1 @ 59 ;
+  jpeg_idct_2x2 @ 60 ;
+  jpeg_idct_4x4 @ 61 ;
+  jpeg_idct_float @ 62 ;
+  jpeg_idct_ifast @ 63 ;
+  jpeg_idct_islow @ 64 ;
+  jpeg_input_complete @ 65 ;
+  jpeg_make_c_derived_tbl @ 66 ;
+  jpeg_make_d_derived_tbl @ 67 ;
+  jpeg_mem_available @ 68 ;
+  jpeg_mem_init @ 69 ;
+  jpeg_mem_term @ 70 ;
+  jpeg_new_colormap @ 71 ;
+  jpeg_open_backing_store @ 72 ;
+  jpeg_quality_scaling @ 73 ;
+  jpeg_read_coefficients @ 74 ;
+  jpeg_read_header @ 75 ;
+  jpeg_read_raw_data @ 76 ;
+  jpeg_read_scanlines @ 77 ;
+  jpeg_resync_to_restart @ 78 ;
+  jpeg_save_markers @ 79 ;
+  jpeg_set_colorspace @ 80 ;
+  jpeg_set_defaults @ 81 ;
+  jpeg_set_linear_quality @ 82 ;
+  jpeg_set_marker_processor @ 83 ;
+  jpeg_set_quality @ 84 ;
+  jpeg_simple_progression @ 85 ;
+  jpeg_start_compress @ 86 ;
+  jpeg_start_decompress @ 87 ;
+  jpeg_start_output @ 88 ;
+  jpeg_std_error @ 89 ;
+  jpeg_stdio_dest @ 90 ;
+  jpeg_stdio_src @ 91 ;
+  jpeg_suppress_tables @ 92 ;
+  jpeg_write_coefficients @ 93 ;
+  jpeg_write_m_byte @ 94 ;
+  jpeg_write_m_header @ 95 ;
+  jpeg_write_marker @ 96 ;
+  jpeg_write_raw_data @ 97 ;
+  jpeg_write_scanlines @ 98 ;
+  jpeg_write_tables @ 99 ;
+  jround_up @ 100 ;
+  jzero_far @ 101 ;
+  jpeg_skip_scanlines @ 102 ;
+  jpeg_crop_scanline @ 103 ;
+  jpeg_read_icc_profile @ 104 ;
+  jpeg_write_icc_profile @ 105 ;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/jpeg7-memsrcdst.def Added

@@ -0,0 +1,110 @@
+EXPORTS
+  jcopy_block_row @ 1 ;
+  jcopy_sample_rows @ 2 ;
+  jdiv_round_up @ 3 ;
+  jinit_1pass_quantizer @ 4 ;
+  jinit_2pass_quantizer @ 5 ;
+  jinit_c_coef_controller @ 6 ;
+  jinit_c_main_controller @ 7 ;
+  jinit_c_master_control @ 8 ;
+  jinit_c_prep_controller @ 9 ;
+  jinit_color_converter @ 10 ;
+  jinit_color_deconverter @ 11 ;
+  jinit_compress_master @ 12 ;
+  jinit_d_coef_controller @ 13 ;
+  jinit_d_main_controller @ 14 ;
+  jinit_d_post_controller @ 15 ;
+  jinit_downsampler @ 16 ;
+  jinit_forward_dct @ 17 ;
+  jinit_huff_decoder @ 18 ;
+  jinit_huff_encoder @ 19 ;
+  jinit_input_controller @ 20 ;
+  jinit_inverse_dct @ 21 ;
+  jinit_marker_reader @ 22 ;
+  jinit_marker_writer @ 23 ;
+  jinit_master_decompress @ 24 ;
+  jinit_memory_mgr @ 25 ;
+  jinit_merged_upsampler @ 26 ;
+  jinit_phuff_decoder @ 27 ;
+  jinit_phuff_encoder @ 28 ;
+  jinit_upsampler @ 29 ;
+  jpeg_CreateCompress @ 30 ;
+  jpeg_CreateDecompress @ 31 ;
+  jpeg_abort @ 32 ;
+  jpeg_abort_compress @ 33 ;
+  jpeg_abort_decompress @ 34 ;
+  jpeg_add_quant_table @ 35 ;
+  jpeg_alloc_huff_table @ 36 ;
+  jpeg_alloc_quant_table @ 37 ;
+  jpeg_calc_jpeg_dimensions @ 38 ;
+  jpeg_calc_output_dimensions @ 39 ;
+  jpeg_consume_input @ 40 ;
+  jpeg_copy_critical_parameters @ 41 ;
+  jpeg_default_colorspace @ 42 ;
+  jpeg_default_qtables @ 43 ;
+  jpeg_destroy @ 44 ;
+  jpeg_destroy_compress @ 45 ;
+  jpeg_destroy_decompress @ 46 ;
+  jpeg_fdct_float @ 47 ;
+  jpeg_fdct_ifast @ 48 ;
+  jpeg_fdct_islow @ 49 ;
+  jpeg_fill_bit_buffer @ 50 ;
+  jpeg_finish_compress @ 51 ;
+  jpeg_finish_decompress @ 52 ;
+  jpeg_finish_output @ 53 ;
+  jpeg_free_large @ 54 ;
+  jpeg_free_small @ 55 ;
+  jpeg_gen_optimal_table @ 56 ;
+  jpeg_get_large @ 57 ;
+  jpeg_get_small @ 58 ;
+  jpeg_has_multiple_scans @ 59 ;
+  jpeg_huff_decode @ 60 ;
+  jpeg_idct_1x1 @ 61 ;
+  jpeg_idct_2x2 @ 62 ;
+  jpeg_idct_4x4 @ 63 ;
+  jpeg_idct_float @ 64 ;
+  jpeg_idct_ifast @ 65 ;
+  jpeg_idct_islow @ 66 ;
+  jpeg_input_complete @ 67 ;
+  jpeg_make_c_derived_tbl @ 68 ;
+  jpeg_make_d_derived_tbl @ 69 ;
+  jpeg_mem_available @ 70 ;
+  jpeg_mem_init @ 71 ;
+  jpeg_mem_term @ 72 ;
+  jpeg_new_colormap @ 73 ;
+  jpeg_open_backing_store @ 74 ;
+  jpeg_quality_scaling @ 75 ;
+  jpeg_read_coefficients @ 76 ;
+  jpeg_read_header @ 77 ;
+  jpeg_read_raw_data @ 78 ;
+  jpeg_read_scanlines @ 79 ;
+  jpeg_resync_to_restart @ 80 ;
+  jpeg_save_markers @ 81 ;
+  jpeg_set_colorspace @ 82 ;
+  jpeg_set_defaults @ 83 ;
+  jpeg_set_linear_quality @ 84 ;
+  jpeg_set_marker_processor @ 85 ;
+  jpeg_set_quality @ 86 ;
+  jpeg_simple_progression @ 87 ;
+  jpeg_start_compress @ 88 ;
+  jpeg_start_decompress @ 89 ;
+  jpeg_start_output @ 90 ;
+  jpeg_std_error @ 91 ;
+  jpeg_stdio_dest @ 92 ;
+  jpeg_stdio_src @ 93 ;
+  jpeg_suppress_tables @ 94 ;
+  jpeg_write_coefficients @ 95 ;
+  jpeg_write_m_byte @ 96 ;
+  jpeg_write_m_header @ 97 ;
+  jpeg_write_marker @ 98 ;
+  jpeg_write_raw_data @ 99 ;
+  jpeg_write_scanlines @ 100 ;
+  jpeg_write_tables @ 101 ;
+  jround_up @ 102 ;
+  jzero_far @ 103 ;
+  jpeg_mem_dest @ 104 ;
+  jpeg_mem_src @ 105 ;
+  jpeg_skip_scanlines @ 106 ;
+  jpeg_crop_scanline @ 107 ;
+  jpeg_read_icc_profile @ 108 ;
+  jpeg_write_icc_profile @ 109 ;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/jpeg7.def Added

@@ -0,0 +1,108 @@
+EXPORTS
+  jcopy_block_row @ 1 ;
+  jcopy_sample_rows @ 2 ;
+  jdiv_round_up @ 3 ;
+  jinit_1pass_quantizer @ 4 ;
+  jinit_2pass_quantizer @ 5 ;
+  jinit_c_coef_controller @ 6 ;
+  jinit_c_main_controller @ 7 ;
+  jinit_c_master_control @ 8 ;
+  jinit_c_prep_controller @ 9 ;
+  jinit_color_converter @ 10 ;
+  jinit_color_deconverter @ 11 ;
+  jinit_compress_master @ 12 ;
+  jinit_d_coef_controller @ 13 ;
+  jinit_d_main_controller @ 14 ;
+  jinit_d_post_controller @ 15 ;
+  jinit_downsampler @ 16 ;
+  jinit_forward_dct @ 17 ;
+  jinit_huff_decoder @ 18 ;
+  jinit_huff_encoder @ 19 ;
+  jinit_input_controller @ 20 ;
+  jinit_inverse_dct @ 21 ;
+  jinit_marker_reader @ 22 ;
+  jinit_marker_writer @ 23 ;
+  jinit_master_decompress @ 24 ;
+  jinit_memory_mgr @ 25 ;
+  jinit_merged_upsampler @ 26 ;
+  jinit_phuff_decoder @ 27 ;
+  jinit_phuff_encoder @ 28 ;
+  jinit_upsampler @ 29 ;
+  jpeg_CreateCompress @ 30 ;
+  jpeg_CreateDecompress @ 31 ;
+  jpeg_abort @ 32 ;
+  jpeg_abort_compress @ 33 ;
+  jpeg_abort_decompress @ 34 ;
+  jpeg_add_quant_table @ 35 ;
+  jpeg_alloc_huff_table @ 36 ;
+  jpeg_alloc_quant_table @ 37 ;
+  jpeg_calc_jpeg_dimensions @ 38 ;
+  jpeg_calc_output_dimensions @ 39 ;
+  jpeg_consume_input @ 40 ;
+  jpeg_copy_critical_parameters @ 41 ;
+  jpeg_default_colorspace @ 42 ;
+  jpeg_default_qtables @ 43 ;
+  jpeg_destroy @ 44 ;
+  jpeg_destroy_compress @ 45 ;
+  jpeg_destroy_decompress @ 46 ;
+  jpeg_fdct_float @ 47 ;
+  jpeg_fdct_ifast @ 48 ;
+  jpeg_fdct_islow @ 49 ;
+  jpeg_fill_bit_buffer @ 50 ;
+  jpeg_finish_compress @ 51 ;
+  jpeg_finish_decompress @ 52 ;
+  jpeg_finish_output @ 53 ;
+  jpeg_free_large @ 54 ;
+  jpeg_free_small @ 55 ;
+  jpeg_gen_optimal_table @ 56 ;
+  jpeg_get_large @ 57 ;
+  jpeg_get_small @ 58 ;
+  jpeg_has_multiple_scans @ 59 ;
+  jpeg_huff_decode @ 60 ;
+  jpeg_idct_1x1 @ 61 ;
+  jpeg_idct_2x2 @ 62 ;
+  jpeg_idct_4x4 @ 63 ;
+  jpeg_idct_float @ 64 ;
+  jpeg_idct_ifast @ 65 ;
+  jpeg_idct_islow @ 66 ;
+  jpeg_input_complete @ 67 ;
+  jpeg_make_c_derived_tbl @ 68 ;
+  jpeg_make_d_derived_tbl @ 69 ;
+  jpeg_mem_available @ 70 ;
+  jpeg_mem_init @ 71 ;
+  jpeg_mem_term @ 72 ;
+  jpeg_new_colormap @ 73 ;
+  jpeg_open_backing_store @ 74 ;
+  jpeg_quality_scaling @ 75 ;
+  jpeg_read_coefficients @ 76 ;
+  jpeg_read_header @ 77 ;
+  jpeg_read_raw_data @ 78 ;
+  jpeg_read_scanlines @ 79 ;
+  jpeg_resync_to_restart @ 80 ;
+  jpeg_save_markers @ 81 ;
+  jpeg_set_colorspace @ 82 ;
+  jpeg_set_defaults @ 83 ;
+  jpeg_set_linear_quality @ 84 ;
+  jpeg_set_marker_processor @ 85 ;
+  jpeg_set_quality @ 86 ;
+  jpeg_simple_progression @ 87 ;
+  jpeg_start_compress @ 88 ;
+  jpeg_start_decompress @ 89 ;
+  jpeg_start_output @ 90 ;
+  jpeg_std_error @ 91 ;
+  jpeg_stdio_dest @ 92 ;
+  jpeg_stdio_src @ 93 ;
+  jpeg_suppress_tables @ 94 ;
+  jpeg_write_coefficients @ 95 ;
+  jpeg_write_m_byte @ 96 ;
+  jpeg_write_m_header @ 97 ;
+  jpeg_write_marker @ 98 ;
+  jpeg_write_raw_data @ 99 ;
+  jpeg_write_scanlines @ 100 ;
+  jpeg_write_tables @ 101 ;
+  jround_up @ 102 ;
+  jzero_far @ 103 ;
+  jpeg_skip_scanlines @ 104 ;
+  jpeg_crop_scanline @ 105 ;
+  jpeg_read_icc_profile @ 106 ;
+  jpeg_write_icc_profile @ 107 ;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/jpeg8.def Added

@@ -0,0 +1,111 @@
+EXPORTS
+  jcopy_block_row @ 1 ;
+  jcopy_sample_rows @ 2 ;
+  jdiv_round_up @ 3 ;
+  jinit_1pass_quantizer @ 4 ;
+  jinit_2pass_quantizer @ 5 ;
+  jinit_c_coef_controller @ 6 ;
+  jinit_c_main_controller @ 7 ;
+  jinit_c_master_control @ 8 ;
+  jinit_c_prep_controller @ 9 ;
+  jinit_color_converter @ 10 ;
+  jinit_color_deconverter @ 11 ;
+  jinit_compress_master @ 12 ;
+  jinit_d_coef_controller @ 13 ;
+  jinit_d_main_controller @ 14 ;
+  jinit_d_post_controller @ 15 ;
+  jinit_downsampler @ 16 ;
+  jinit_forward_dct @ 17 ;
+  jinit_huff_decoder @ 18 ;
+  jinit_huff_encoder @ 19 ;
+  jinit_input_controller @ 20 ;
+  jinit_inverse_dct @ 21 ;
+  jinit_marker_reader @ 22 ;
+  jinit_marker_writer @ 23 ;
+  jinit_master_decompress @ 24 ;
+  jinit_memory_mgr @ 25 ;
+  jinit_merged_upsampler @ 26 ;
+  jinit_phuff_decoder @ 27 ;
+  jinit_phuff_encoder @ 28 ;
+  jinit_upsampler @ 29 ;
+  jpeg_CreateCompress @ 30 ;
+  jpeg_CreateDecompress @ 31 ;
+  jpeg_abort @ 32 ;
+  jpeg_abort_compress @ 33 ;
+  jpeg_abort_decompress @ 34 ;
+  jpeg_add_quant_table @ 35 ;
+  jpeg_alloc_huff_table @ 36 ;
+  jpeg_alloc_quant_table @ 37 ;
+  jpeg_calc_jpeg_dimensions @ 38 ;
+  jpeg_calc_output_dimensions @ 39 ;
+  jpeg_consume_input @ 40 ;
+  jpeg_copy_critical_parameters @ 41 ;
+  jpeg_core_output_dimensions @ 42 ;
+  jpeg_default_colorspace @ 43 ;
+  jpeg_default_qtables @ 44 ;
+  jpeg_destroy @ 45 ;
+  jpeg_destroy_compress @ 46 ;
+  jpeg_destroy_decompress @ 47 ;
+  jpeg_fdct_float @ 48 ;
+  jpeg_fdct_ifast @ 49 ;
+  jpeg_fdct_islow @ 50 ;
+  jpeg_fill_bit_buffer @ 51 ;
+  jpeg_finish_compress @ 52 ;
+  jpeg_finish_decompress @ 53 ;
+  jpeg_finish_output @ 54 ;
+  jpeg_free_large @ 55 ;
+  jpeg_free_small @ 56 ;
+  jpeg_gen_optimal_table @ 57 ;
+  jpeg_get_large @ 58 ;
+  jpeg_get_small @ 59 ;
+  jpeg_has_multiple_scans @ 60 ;
+  jpeg_huff_decode @ 61 ;
+  jpeg_idct_1x1 @ 62 ;
+  jpeg_idct_2x2 @ 63 ;
+  jpeg_idct_4x4 @ 64 ;
+  jpeg_idct_float @ 65 ;
+  jpeg_idct_ifast @ 66 ;
+  jpeg_idct_islow @ 67 ;
+  jpeg_input_complete @ 68 ;
+  jpeg_make_c_derived_tbl @ 69 ;
+  jpeg_make_d_derived_tbl @ 70 ;
+  jpeg_mem_available @ 71 ;
+  jpeg_mem_dest @ 72 ;
+  jpeg_mem_init @ 73 ;
+  jpeg_mem_src @ 74 ;
+  jpeg_mem_term @ 75 ;
+  jpeg_new_colormap @ 76 ;
+  jpeg_open_backing_store @ 77 ;
+  jpeg_quality_scaling @ 78 ;
+  jpeg_read_coefficients @ 79 ;
+  jpeg_read_header @ 80 ;
+  jpeg_read_raw_data @ 81 ;
+  jpeg_read_scanlines @ 82 ;
+  jpeg_resync_to_restart @ 83 ;
+  jpeg_save_markers @ 84 ;
+  jpeg_set_colorspace @ 85 ;
+  jpeg_set_defaults @ 86 ;
+  jpeg_set_linear_quality @ 87 ;
+  jpeg_set_marker_processor @ 88 ;
+  jpeg_set_quality @ 89 ;
+  jpeg_simple_progression @ 90 ;
+  jpeg_start_compress @ 91 ;
+  jpeg_start_decompress @ 92 ;
+  jpeg_start_output @ 93 ;
+  jpeg_std_error @ 94 ;
+  jpeg_stdio_dest @ 95 ;
+  jpeg_stdio_src @ 96 ;
+  jpeg_suppress_tables @ 97 ;
+  jpeg_write_coefficients @ 98 ;
+  jpeg_write_m_byte @ 99 ;
+  jpeg_write_m_header @ 100 ;
+  jpeg_write_marker @ 101 ;
+  jpeg_write_raw_data @ 102 ;
+  jpeg_write_scanlines @ 103 ;
+  jpeg_write_tables @ 104 ;
+  jround_up @ 105 ;
+  jzero_far @ 106 ;
+  jpeg_skip_scanlines @ 107 ;
+  jpeg_crop_scanline @ 108 ;
+  jpeg_read_icc_profile @ 109 ;
+  jpeg_write_icc_profile @ 110 ;

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/projectTargets.cmake.in Added

@@ -0,0 +1,115 @@
+# Generated by CMake
+
+if("${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION}" LESS 2.5)
+   message(FATAL_ERROR "CMake >= 2.6.0 required")
+endif()
+cmake_policy(PUSH)
+cmake_policy(VERSION 2.6)
+#----------------------------------------------------------------
+# Generated CMake target import file.
+#----------------------------------------------------------------
+
+# Commands may need to know the format version.
+set(CMAKE_IMPORT_FILE_VERSION 1)
+
+# Protect against multiple inclusion, which would fail when already imported targets are added once more.
+set(_targetsDefined)
+set(_targetsNotDefined)
+set(_expectedTargets)
+foreach(_expectedTarget @CMAKE_PROJECT_NAME@::jpeg @CMAKE_PROJECT_NAME@::turbojpeg @CMAKE_PROJECT_NAME@::turbojpeg-static @CMAKE_PROJECT_NAME@::jpeg-static)
+  list(APPEND _expectedTargets ${_expectedTarget})
+  if(NOT TARGET ${_expectedTarget})
+    list(APPEND _targetsNotDefined ${_expectedTarget})
+  endif()
+  if(TARGET ${_expectedTarget})
+    list(APPEND _targetsDefined ${_expectedTarget})
+  endif()
+endforeach()
+if("${_targetsDefined}" STREQUAL "${_expectedTargets}")
+  unset(_targetsDefined)
+  unset(_targetsNotDefined)
+  unset(_expectedTargets)
+  set(CMAKE_IMPORT_FILE_VERSION)
+  cmake_policy(POP)
+  return()
+endif()
+if(NOT "${_targetsDefined}" STREQUAL "")
+  message(FATAL_ERROR "Some (but not all) targets in this export set were already defined.\nTargets Defined: ${_targetsDefined}\nTargets not yet defined: ${_targetsNotDefined}\n")
+endif()
+unset(_targetsDefined)
+unset(_targetsNotDefined)
+unset(_expectedTargets)
+
+
+# Compute the installation prefix relative to this file.
+get_filename_component(_IMPORT_PREFIX "${CMAKE_CURRENT_LIST_FILE}" PATH)
+get_filename_component(_IMPORT_PREFIX "${_IMPORT_PREFIX}" PATH)
+get_filename_component(_IMPORT_PREFIX "${_IMPORT_PREFIX}" PATH)
+get_filename_component(_IMPORT_PREFIX "${_IMPORT_PREFIX}" PATH)
+if(_IMPORT_PREFIX STREQUAL "/")
+  set(_IMPORT_PREFIX "")
+endif()
+
+# Create imported target @CMAKE_PROJECT_NAME@::jpeg
+add_library(@CMAKE_PROJECT_NAME@::jpeg SHARED IMPORTED)
+
+set_target_properties(@CMAKE_PROJECT_NAME@::jpeg PROPERTIES
+  INTERFACE_INCLUDE_DIRECTORIES "${_IMPORT_PREFIX}/include"
+)
+
+# Create imported target @CMAKE_PROJECT_NAME@::turbojpeg
+add_library(@CMAKE_PROJECT_NAME@::turbojpeg SHARED IMPORTED)
+
+set_target_properties(@CMAKE_PROJECT_NAME@::turbojpeg PROPERTIES
+  INTERFACE_INCLUDE_DIRECTORIES "${_IMPORT_PREFIX}/include"
+)
+
+# Create imported target @CMAKE_PROJECT_NAME@::turbojpeg-static
+add_library(@CMAKE_PROJECT_NAME@::turbojpeg-static STATIC IMPORTED)
+
+set_target_properties(@CMAKE_PROJECT_NAME@::turbojpeg-static PROPERTIES
+  INTERFACE_INCLUDE_DIRECTORIES "${_IMPORT_PREFIX}/include"
+)
+
+# Create imported target @CMAKE_PROJECT_NAME@::jpeg-static
+add_library(@CMAKE_PROJECT_NAME@::jpeg-static STATIC IMPORTED)
+
+set_target_properties(@CMAKE_PROJECT_NAME@::jpeg-static PROPERTIES
+  INTERFACE_INCLUDE_DIRECTORIES "${_IMPORT_PREFIX}/include"
+)
+
+# Load information for each installed configuration.
+get_filename_component(_DIR "${CMAKE_CURRENT_LIST_FILE}" PATH)
+file(GLOB CONFIG_FILES "${_DIR}/@CMAKE_PROJECT_NAME@Targets-*.cmake")
+foreach(f ${CONFIG_FILES})
+  include(${f})
+endforeach()
+
+# Cleanup temporary variables.
+set(_IMPORT_PREFIX)
+
+# Loop over all imported files and verify that they actually exist
+foreach(target ${_IMPORT_CHECK_TARGETS} )
+  foreach(file ${_IMPORT_CHECK_FILES_FOR_${target}} )
+    if(NOT EXISTS "${file}" )
+      message(FATAL_ERROR "The imported target \"${target}\" references the file
+   \"${file}\"
+but this file does not exist.  Possible reasons include:
+* The file was deleted, renamed, or moved to another location.
+* An install or uninstall procedure did not complete successfully.
+* The installation package was faulty and contained
+   \"${CMAKE_CURRENT_LIST_FILE}\"
+but not all the files it references.
+")
+    endif()
+  endforeach()
+  unset(_IMPORT_CHECK_FILES_FOR_${target})
+endforeach()
+unset(_IMPORT_CHECK_TARGETS)
+
+# This file does not depend on other imported targets which have
+# been exported from the same project but in a separate export set.
+
+# Commands beyond this point should not need to know the version.
+set(CMAKE_IMPORT_FILE_VERSION)
+cmake_policy(POP)

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/vc Added

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/win/vc/projectTargets-release.cmake.in Added

@@ -0,0 +1,49 @@
+#----------------------------------------------------------------
+# Generated CMake target import file for configuration "Release".
+#----------------------------------------------------------------
+
+# Commands may need to know the format version.
+set(CMAKE_IMPORT_FILE_VERSION 1)
+
+# Import target "@CMAKE_PROJECT_NAME@::jpeg" for configuration "Release"
+set_property(TARGET @CMAKE_PROJECT_NAME@::jpeg APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+set_target_properties(@CMAKE_PROJECT_NAME@::jpeg PROPERTIES
+  IMPORTED_IMPLIB_RELEASE "${_IMPORT_PREFIX}/lib/jpeg.lib"
+  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/bin/jpeg62.dll"
+  )
+
+list(APPEND _IMPORT_CHECK_TARGETS @CMAKE_PROJECT_NAME@::jpeg )
+list(APPEND _IMPORT_CHECK_FILES_FOR_@CMAKE_PROJECT_NAME@::jpeg "${_IMPORT_PREFIX}/lib/jpeg.lib" "${_IMPORT_PREFIX}/bin/jpeg62.dll" )
+
+# Import target "@CMAKE_PROJECT_NAME@::turbojpeg" for configuration "Release"
+set_property(TARGET @CMAKE_PROJECT_NAME@::turbojpeg APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+set_target_properties(@CMAKE_PROJECT_NAME@::turbojpeg PROPERTIES
+  IMPORTED_IMPLIB_RELEASE "${_IMPORT_PREFIX}/lib/turbojpeg.lib"
+  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/bin/turbojpeg.dll"
+  )
+
+list(APPEND _IMPORT_CHECK_TARGETS @CMAKE_PROJECT_NAME@::turbojpeg )
+list(APPEND _IMPORT_CHECK_FILES_FOR_@CMAKE_PROJECT_NAME@::turbojpeg "${_IMPORT_PREFIX}/lib/turbojpeg.lib" "${_IMPORT_PREFIX}/bin/turbojpeg.dll" )
+
+# Import target "@CMAKE_PROJECT_NAME@::turbojpeg-static" for configuration "Release"
+set_property(TARGET @CMAKE_PROJECT_NAME@::turbojpeg-static APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+set_target_properties(@CMAKE_PROJECT_NAME@::turbojpeg-static PROPERTIES
+  IMPORTED_LINK_INTERFACE_LANGUAGES_RELEASE "ASM_NASM;C"
+  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/lib/turbojpeg-static.lib"
+  )
+
+list(APPEND _IMPORT_CHECK_TARGETS @CMAKE_PROJECT_NAME@::turbojpeg-static )
+list(APPEND _IMPORT_CHECK_FILES_FOR_@CMAKE_PROJECT_NAME@::turbojpeg-static "${_IMPORT_PREFIX}/lib/turbojpeg-static.lib" )
+
+# Import target "@CMAKE_PROJECT_NAME@::jpeg-static" for configuration "Release"
+set_property(TARGET @CMAKE_PROJECT_NAME@::jpeg-static APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
+set_target_properties(@CMAKE_PROJECT_NAME@::jpeg-static PROPERTIES
+  IMPORTED_LINK_INTERFACE_LANGUAGES_RELEASE "ASM_NASM;C"
+  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/lib/jpeg-static.lib"
+  )
+
+list(APPEND _IMPORT_CHECK_TARGETS @CMAKE_PROJECT_NAME@::jpeg-static )
+list(APPEND _IMPORT_CHECK_FILES_FOR_@CMAKE_PROJECT_NAME@::jpeg-static "${_IMPORT_PREFIX}/lib/jpeg-static.lib" )
+
+# Commands beyond this point should not need to know the version.
+set(CMAKE_IMPORT_FILE_VERSION)

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/wizard.txt Added

@@ -0,0 +1,212 @@
+Advanced usage instructions for the Independent JPEG Group's JPEG software
+==========================================================================
+
+This file describes cjpeg's "switches for wizards".
+
+The "wizard" switches are intended for experimentation with JPEG by persons
+who are reasonably knowledgeable about the JPEG standard.  If you don't know
+what you are doing, DON'T USE THESE SWITCHES.  You'll likely produce files
+with worse image quality and/or poorer compression than you'd get from the
+default settings.  Furthermore, these switches must be used with caution
+when making files intended for general use, because not all JPEG decoders
+will support unusual JPEG parameter settings.
+
+
+Quantization Table Adjustment
+-----------------------------
+
+Ordinarily, cjpeg starts with a default set of tables (the same ones given
+as examples in the JPEG standard) and scales them up or down according to
+the -quality setting.  The details of the scaling algorithm can be found in
+jcparam.c.  At very low quality settings, some quantization table entries
+can get scaled up to values exceeding 255.  Although 2-byte quantization
+values are supported by the IJG software, this feature is not in baseline
+JPEG and is not supported by all implementations.  If you need to ensure
+wide compatibility of low-quality files, you can constrain the scaled
+quantization values to no more than 255 by giving the -baseline switch.
+Note that use of -baseline will result in poorer quality for the same file
+size, since more bits than necessary are expended on higher AC coefficients.
+
+You can substitute a different set of quantization values by using the
+-qtables switch:
+
+        -qtables file   Use the quantization tables given in the named file.
+
+The specified file should be a text file containing decimal quantization
+values.  The file should contain one to four tables, each of 64 elements.
+The tables are implicitly numbered 0,1,etc. in order of appearance.  Table
+entries appear in normal array order (NOT in the zigzag order in which they
+will be stored in the JPEG file).
+
+Quantization table files are free format, in that arbitrary whitespace can
+appear between numbers.  Also, comments can be included: a comment starts
+with '#' and extends to the end of the line.  Here is an example file that
+duplicates the default quantization tables:
+
+        # Quantization tables given in Annex K (Clause K.1) of
+        # Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
+
+        # This is table 0 (the luminance table):
+          16  11  10  16  24  40  51  61
+          12  12  14  19  26  58  60  55
+          14  13  16  24  40  57  69  56
+          14  17  22  29  51  87  80  62
+          18  22  37  56  68 109 103  77
+          24  35  55  64  81 104 113  92
+          49  64  78  87 103 121 120 101
+          72  92  95  98 112 100 103  99
+
+        # This is table 1 (the chrominance table):
+          17  18  24  47  99  99  99  99
+          18  21  26  66  99  99  99  99
+          24  26  56  99  99  99  99  99
+          47  66  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+
+If the -qtables switch is used without -quality, then the specified tables
+are used exactly as-is.  If both -qtables and -quality are used, then the
+tables taken from the file are scaled in the same fashion that the default
+tables would be scaled for that quality setting.  If -baseline appears, then
+the quantization values are constrained to the range 1-255.
+
+By default, cjpeg will use quantization table 0 for luminance components and
+table 1 for chrominance components.  To override this choice, use the -qslots
+switch:
+
+        -qslots N,...         Select which quantization table to use for
+                                each color component.
+
+The -qslots switch specifies a quantization table number for each color
+component, in the order in which the components appear in the JPEG SOF marker.
+For example, to create a separate table for each of Y,Cb,Cr, you could
+provide a -qtables file that defines three quantization tables and say
+"-qslots 0,1,2".  If -qslots gives fewer table numbers than there are color
+components, then the last table number is repeated as necessary.
+
+
+Sampling Factor Adjustment
+--------------------------
+
+By default, cjpeg uses 2:1 horizontal and vertical downsampling when
+compressing YCbCr data, and no downsampling for all other color spaces.
+You can override this default with the -sample switch:
+
+        -sample HxV,...       Set JPEG sampling factors for each color
+                                component.
+
+The -sample switch specifies the JPEG sampling factors for each color
+component, in the order in which they appear in the JPEG SOF marker.
+If you specify fewer HxV pairs than there are components, the remaining
+components are set to 1x1 sampling.  For example, the default YCbCr setting
+is equivalent to "-sample 2x2,1x1,1x1", which can be abbreviated to
+"-sample 2x2".
+
+There are still some JPEG decoders in existence that support only 2x1
+sampling (also called 4:2:2 sampling).  Compatibility with such decoders can
+be achieved by specifying "-sample 2x1".  This is not recommended unless
+really necessary, since it increases file size and encoding/decoding time
+with very little quality gain.
+
+
+Multiple Scan / Progression Control
+-----------------------------------
+
+By default, cjpeg emits a single-scan sequential JPEG file.  The
+-progressive switch generates a progressive JPEG file using a default series
+of progression parameters.  You can create multiple-scan sequential JPEG
+files or progressive JPEG files with custom progression parameters by using
+the -scans switch:
+
+        -scans file     Use the scan sequence given in the named file.
+
+The specified file should be a text file containing a "scan script".
+The script specifies the contents and ordering of the scans to be emitted.
+Each entry in the script defines one scan.  A scan definition specifies
+the components to be included in the scan, and for progressive JPEG it also
+specifies the progression parameters Ss,Se,Ah,Al for the scan.  Scan
+definitions are separated by semicolons (';').  A semicolon after the last
+scan definition is optional.
+
+Each scan definition contains one to four component indexes, optionally
+followed by a colon (':') and the four progressive-JPEG parameters.  The
+component indexes denote which color component(s) are to be transmitted in
+the scan.  Components are numbered in the order in which they appear in the
+JPEG SOF marker, with the first component being numbered 0.  (Note that these
+indexes are not the "component ID" codes assigned to the components, just
+positional indexes.)
+
+The progression parameters for each scan are:
+        Ss      Zigzag index of first coefficient included in scan
+        Se      Zigzag index of last coefficient included in scan
+        Ah      Zero for first scan of a coefficient, else Al of prior scan
+        Al      Successive approximation low bit position for scan
+If the progression parameters are omitted, the values 0,63,0,0 are used,
+producing a sequential JPEG file.  cjpeg automatically determines whether
+the script represents a progressive or sequential file, by observing whether
+Ss and Se values other than 0 and 63 appear.  (The -progressive switch is
+not needed to specify this; in fact, it is ignored when -scans appears.)
+The scan script must meet the JPEG restrictions on progression sequences.
+(cjpeg checks that the spec's requirements are obeyed.)
+
+Scan script files are free format, in that arbitrary whitespace can appear
+between numbers and around punctuation.  Also, comments can be included: a
+comment starts with '#' and extends to the end of the line.  For additional
+legibility, commas or dashes can be placed between values.  (Actually, any
+single punctuation character other than ':' or ';' can be inserted.)  For
+example, the following two scan definitions are equivalent:
+        0 1 2: 0 63 0 0;
+        0,1,2 : 0-63, 0,0 ;
+
+Here is an example of a scan script that generates a partially interleaved
+sequential JPEG file:
+
+        0;                      # Y only in first scan
+        1 2;                    # Cb and Cr in second scan
+
+Here is an example of a progressive scan script using only spectral selection
+(no successive approximation):
+
+        # Interleaved DC scan for Y,Cb,Cr:
+        0,1,2: 0-0,   0, 0 ;
+        # AC scans:
+        0:     1-2,   0, 0 ;    # First two Y AC coefficients
+        0:     3-5,   0, 0 ;    # Three more
+        1:     1-63,  0, 0 ;    # All AC coefficients for Cb
+        2:     1-63,  0, 0 ;    # All AC coefficients for Cr
+        0:     6-9,   0, 0 ;    # More Y coefficients
+        0:     10-63, 0, 0 ;    # Remaining Y coefficients
+
+Here is an example of a successive-approximation script.  This is equivalent
+to the default script used by "cjpeg -progressive" for YCbCr images:
+
+        # Initial DC scan for Y,Cb,Cr (lowest bit not sent)
+        0,1,2: 0-0,   0, 1 ;
+        # First AC scan: send first 5 Y AC coefficients, minus 2 lowest bits:
+        0:     1-5,   0, 2 ;
+        # Send all Cr,Cb AC coefficients, minus lowest bit:
+        # (chroma data is usually too small to be worth subdividing further;
+        #  but note we send Cr first since eye is least sensitive to Cb)
+        2:     1-63,  0, 1 ;
+        1:     1-63,  0, 1 ;
+        # Send remaining Y AC coefficients, minus 2 lowest bits:
+        0:     6-63,  0, 2 ;
+        # Send next-to-lowest bit of all Y AC coefficients:
+        0:     1-63,  2, 1 ;
+        # At this point we've sent all but the lowest bit of all coefficients.
+        # Send lowest bit of DC coefficients
+        0,1,2: 0-0,   1, 0 ;
+        # Send lowest bit of AC coefficients
+        2:     1-63,  1, 0 ;
+        1:     1-63,  1, 0 ;
+        # Y AC lowest bit scan is last; it's usually the largest scan
+        0:     1-63,  1, 0 ;
+
+It may be worth pointing out that this script is tuned for quality settings
+of around 50 to 75.  For lower quality settings, you'd probably want to use
+a script with fewer stages of successive approximation (otherwise the
+initial scans will be really bad).  For higher quality settings, you might
+want to use more stages of successive approximation (so that the initial
+scans are not too large).

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/wrbmp.c Added

@@ -0,0 +1,552 @@
+/*
+ * wrbmp.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2014-2015, 2017, 2019, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in Microsoft "BMP"
+ * format (MS Windows 3.x and OS/2 1.x flavors).
+ * Either 8-bit colormapped or 24-bit full-color format can be written.
+ * No compression is supported.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * This code contributed by James Arthur Boucher.
+ */
+
+#include "cmyk.h"
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "jconfigint.h"
+
+#ifdef BMP_SUPPORTED
+
+
+/*
+ * To support 12-bit JPEG data, we'd have to scale output down to 8 bits.
+ * This is not yet implemented.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+/*
+ * Since BMP stores scanlines bottom-to-top, we have to invert the image
+ * from JPEG's top-to-bottom order.  To do this, we save the outgoing data
+ * in a virtual array during put_pixel_row calls, then actually emit the
+ * BMP file during finish_output.  The virtual array contains one JSAMPLE per
+ * pixel if the output is grayscale or colormapped, three if it is full color.
+ */
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  boolean is_os2;               /* saves the OS2 format request flag */
+
+  jvirt_sarray_ptr whole_image; /* needed to reverse row order */
+  JDIMENSION data_width;        /* JSAMPLEs per row */
+  JDIMENSION row_width;         /* physical width of one row in the BMP file */
+  int pad_bytes;                /* number of padding bytes needed per row */
+  JDIMENSION cur_output_row;    /* next row# to write to virtual array */
+
+  boolean use_inversion_array;  /* TRUE = buffer the whole image, which is
+                                   stored to disk in bottom-up order, and
+                                   receive rows from the calling program in
+                                   top-down order
+
+                                   FALSE = the calling program will maintain
+                                   its own image buffer and write the rows in
+                                   bottom-up order */
+
+  JSAMPLE *iobuffer;            /* I/O buffer (used to buffer a single row to
+                                   disk if use_inversion_array == FALSE) */
+} bmp_dest_struct;
+
+typedef bmp_dest_struct *bmp_dest_ptr;
+
+
+/* Forward declarations */
+LOCAL(void) write_colormap(j_decompress_ptr cinfo, bmp_dest_ptr dest,
+                           int map_colors, int map_entry_size);
+
+
+static INLINE boolean is_big_endian(void)
+{
+  int test_value = 1;
+  if (*(char *)&test_value != 1)
+    return TRUE;
+  return FALSE;
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+               JDIMENSION rows_supplied)
+/* This version is for writing 24-bit pixels */
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr)dinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  int pad;
+
+  if (dest->use_inversion_array) {
+    /* Access next row in virtual array */
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr)cinfo, dest->whole_image,
+       dest->cur_output_row, (JDIMENSION)1, TRUE);
+    dest->cur_output_row++;
+    outptr = image_ptr0;
+  } else {
+    outptr = dest->iobuffer;
+  }
+
+  /* Transfer data.  Note destination values must be in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  inptr = dest->pub.buffer0;
+
+  if (cinfo->out_color_space == JCS_EXT_BGR) {
+    MEMCOPY(outptr, inptr, dest->row_width);
+    outptr += cinfo->output_width * 3;
+  } else if (cinfo->out_color_space == JCS_RGB565) {
+    boolean big_endian = is_big_endian();
+    unsigned short *inptr2 = (unsigned short *)inptr;
+    for (col = cinfo->output_width; col > 0; col--) {
+      if (big_endian) {
+        outptr0 = (*inptr2 >> 5) & 0xF8;
+        outptr1 = ((*inptr2 << 5) & 0xE0) | ((*inptr2 >> 11) & 0x1C);
+        outptr2 = *inptr2 & 0xF8;
+      } else {
+        outptr0 = (*inptr2 << 3) & 0xF8;
+        outptr1 = (*inptr2 >> 3) & 0xFC;
+        outptr2 = (*inptr2 >> 8) & 0xF8;
+      }
+      outptr += 3;
+      inptr2++;
+    }
+  } else if (cinfo->out_color_space == JCS_CMYK) {
+    for (col = cinfo->output_width; col > 0; col--) {
+      JSAMPLE c = *inptr++, m = *inptr++, y = *inptr++, k = *inptr++;
+      cmyk_to_rgb(c, m, y, k, outptr + 2, outptr + 1, outptr);
+      outptr += 3;
+    }
+  } else {
+    register int rindex = rgb_redcinfo->out_color_space;
+    register int gindex = rgb_greencinfo->out_color_space;
+    register int bindex = rgb_bluecinfo->out_color_space;
+    register int ps = rgb_pixelsizecinfo->out_color_space;
+
+    for (col = cinfo->output_width; col > 0; col--) {
+      outptr0 = inptrbindex;
+      outptr1 = inptrgindex;
+      outptr2 = inptrrindex;
+      outptr += 3;  inptr += ps;
+    }
+  }
+
+  /* Zero out the pad bytes. */
+  pad = dest->pad_bytes;
+  while (--pad >= 0)
+    *outptr++ = 0;
+
+  if (!dest->use_inversion_array)
+    (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->row_width);
+}
+
+METHODDEF(void)
+put_gray_rows(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+              JDIMENSION rows_supplied)
+/* This version is for grayscale OR quantized color output */
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr)dinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  int pad;
+
+  if (dest->use_inversion_array) {
+    /* Access next row in virtual array */
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr)cinfo, dest->whole_image,
+       dest->cur_output_row, (JDIMENSION)1, TRUE);
+    dest->cur_output_row++;
+    outptr = image_ptr0;
+  } else {
+    outptr = dest->iobuffer;
+  }
+
+  /* Transfer data. */
+  inptr = dest->pub.buffer0;
+  MEMCOPY(outptr, inptr, cinfo->output_width);
+  outptr += cinfo->output_width;
+
+  /* Zero out the pad bytes. */
+  pad = dest->pad_bytes;
+  while (--pad >= 0)
+    *outptr++ = 0;
+
+  if (!dest->use_inversion_array)
+    (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->row_width);
+}
+
+
+/*
+ * Finish up at the end of the file.
+ *
+ * Here is where we really output the BMP file.
+ *
+ * First, routines to write the Windows and OS/2 variants of the file header.
+ */
+
+LOCAL(void)
+write_bmp_header(j_decompress_ptr cinfo, bmp_dest_ptr dest)
+/* Write a Windows-style BMP file header, including colormap if needed */
+{
+  char bmpfileheader14;
+  char bmpinfoheader40;
+
+#define PUT_2B(array, offset, value) \
+  (arrayoffset = (char)((value) & 0xFF), \
+   arrayoffset + 1 = (char)(((value) >> 8) & 0xFF))
+#define PUT_4B(array, offset, value) \
+  (arrayoffset = (char)((value) & 0xFF), \
+   arrayoffset + 1 = (char)(((value) >> 8) & 0xFF), \
+   arrayoffset + 2 = (char)(((value) >> 16) & 0xFF), \
+   arrayoffset + 3 = (char)(((value) >> 24) & 0xFF))
+
+  long headersize, bfSize;
+  int bits_per_pixel, cmap_entries;
+
+  /* Compute colormap size and total file size */
+  if (IsExtRGB(cinfo->out_color_space)) {
+    if (cinfo->quantize_colors) {
+      /* Colormapped RGB */
+      bits_per_pixel = 8;
+      cmap_entries = 256;
+    } else {
+      /* Unquantized, full color RGB */
+      bits_per_pixel = 24;
+      cmap_entries = 0;
+    }
+  } else if (cinfo->out_color_space == JCS_RGB565 ||
+             cinfo->out_color_space == JCS_CMYK) {
+    bits_per_pixel = 24;
+    cmap_entries   = 0;
+  } else {
+    /* Grayscale output.  We need to fake a 256-entry colormap. */
+    bits_per_pixel = 8;
+    cmap_entries = 256;
+  }
+  /* File size */
+  headersize = 14 + 40 + cmap_entries * 4; /* Header and colormap */
+  bfSize = headersize + (long)dest->row_width * (long)cinfo->output_height;
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(bmpfileheader, sizeof(bmpfileheader));
+  MEMZERO(bmpinfoheader, sizeof(bmpinfoheader));
+
+  /* Fill the file header */
+  bmpfileheader0 = 0x42;      /* first 2 bytes are ASCII 'B', 'M' */
+  bmpfileheader1 = 0x4D;
+  PUT_4B(bmpfileheader, 2, bfSize); /* bfSize */
+  /* we leave bfReserved1 & bfReserved2 = 0 */
+  PUT_4B(bmpfileheader, 10, headersize); /* bfOffBits */
+
+  /* Fill the info header (Microsoft calls this a BITMAPINFOHEADER) */
+  PUT_2B(bmpinfoheader, 0, 40); /* biSize */
+  PUT_4B(bmpinfoheader, 4, cinfo->output_width); /* biWidth */
+  PUT_4B(bmpinfoheader, 8, cinfo->output_height); /* biHeight */
+  PUT_2B(bmpinfoheader, 12, 1); /* biPlanes - must be 1 */
+  PUT_2B(bmpinfoheader, 14, bits_per_pixel); /* biBitCount */
+  /* we leave biCompression = 0, for none */
+  /* we leave biSizeImage = 0; this is correct for uncompressed data */
+  if (cinfo->density_unit == 2) { /* if have density in dots/cm, then */
+    PUT_4B(bmpinfoheader, 24, (long)(cinfo->X_density * 100)); /* XPels/M */
+    PUT_4B(bmpinfoheader, 28, (long)(cinfo->Y_density * 100)); /* XPels/M */
+  }
+  PUT_2B(bmpinfoheader, 32, cmap_entries); /* biClrUsed */
+  /* we leave biClrImportant = 0 */
+
+  if (JFWRITE(dest->pub.output_file, bmpfileheader, 14) != (size_t)14)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+  if (JFWRITE(dest->pub.output_file, bmpinfoheader, 40) != (size_t)40)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  if (cmap_entries > 0)
+    write_colormap(cinfo, dest, cmap_entries, 4);
+}
+
+
+LOCAL(void)
+write_os2_header(j_decompress_ptr cinfo, bmp_dest_ptr dest)
+/* Write an OS2-style BMP file header, including colormap if needed */
+{
+  char bmpfileheader14;
+  char bmpcoreheader12;
+  long headersize, bfSize;
+  int bits_per_pixel, cmap_entries;
+
+  /* Compute colormap size and total file size */
+  if (IsExtRGB(cinfo->out_color_space)) {
+    if (cinfo->quantize_colors) {
+      /* Colormapped RGB */
+      bits_per_pixel = 8;
+      cmap_entries = 256;
+    } else {
+      /* Unquantized, full color RGB */
+      bits_per_pixel = 24;
+      cmap_entries = 0;
+    }
+  } else if (cinfo->out_color_space == JCS_RGB565 ||
+             cinfo->out_color_space == JCS_CMYK) {
+    bits_per_pixel = 24;
+    cmap_entries   = 0;
+  } else {
+    /* Grayscale output.  We need to fake a 256-entry colormap. */
+    bits_per_pixel = 8;
+    cmap_entries = 256;
+  }
+  /* File size */
+  headersize = 14 + 12 + cmap_entries * 3; /* Header and colormap */
+  bfSize = headersize + (long)dest->row_width * (long)cinfo->output_height;
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(bmpfileheader, sizeof(bmpfileheader));
+  MEMZERO(bmpcoreheader, sizeof(bmpcoreheader));
+
+  /* Fill the file header */
+  bmpfileheader0 = 0x42;      /* first 2 bytes are ASCII 'B', 'M' */
+  bmpfileheader1 = 0x4D;
+  PUT_4B(bmpfileheader, 2, bfSize); /* bfSize */
+  /* we leave bfReserved1 & bfReserved2 = 0 */
+  PUT_4B(bmpfileheader, 10, headersize); /* bfOffBits */
+
+  /* Fill the info header (Microsoft calls this a BITMAPCOREHEADER) */
+  PUT_2B(bmpcoreheader, 0, 12); /* bcSize */
+  PUT_2B(bmpcoreheader, 4, cinfo->output_width); /* bcWidth */
+  PUT_2B(bmpcoreheader, 6, cinfo->output_height); /* bcHeight */
+  PUT_2B(bmpcoreheader, 8, 1);  /* bcPlanes - must be 1 */
+  PUT_2B(bmpcoreheader, 10, bits_per_pixel); /* bcBitCount */
+
+  if (JFWRITE(dest->pub.output_file, bmpfileheader, 14) != (size_t)14)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+  if (JFWRITE(dest->pub.output_file, bmpcoreheader, 12) != (size_t)12)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  if (cmap_entries > 0)
+    write_colormap(cinfo, dest, cmap_entries, 3);
+}
+
+
+/*
+ * Write the colormap.
+ * Windows uses BGR0 map entries; OS/2 uses BGR entries.
+ */
+
+LOCAL(void)
+write_colormap(j_decompress_ptr cinfo, bmp_dest_ptr dest, int map_colors,
+               int map_entry_size)
+{
+  JSAMPARRAY colormap = cinfo->colormap;
+  int num_colors = cinfo->actual_number_of_colors;
+  FILE *outfile = dest->pub.output_file;
+  int i;
+
+  if (colormap != NULL) {
+    if (cinfo->out_color_components == 3) {
+      /* Normal case with RGB colormap */
+      for (i = 0; i < num_colors; i++) {
+        putc(colormap2i, outfile);
+        putc(colormap1i, outfile);
+        putc(colormap0i, outfile);
+        if (map_entry_size == 4)
+          putc(0, outfile);
+      }
+    } else {
+      /* Grayscale colormap (only happens with grayscale quantization) */
+      for (i = 0; i < num_colors; i++) {
+        putc(colormap0i, outfile);
+        putc(colormap0i, outfile);
+        putc(colormap0i, outfile);
+        if (map_entry_size == 4)
+          putc(0, outfile);
+      }
+    }
+  } else {
+    /* If no colormap, must be grayscale data.  Generate a linear "map". */
+    for (i = 0; i < 256; i++) {
+      putc(i, outfile);
+      putc(i, outfile);
+      putc(i, outfile);
+      if (map_entry_size == 4)
+        putc(0, outfile);
+    }
+  }
+  /* Pad colormap with zeros to ensure specified number of colormap entries */
+  if (i > map_colors)
+    ERREXIT1(cinfo, JERR_TOO_MANY_COLORS, i);
+  for (; i < map_colors; i++) {
+    putc(0, outfile);
+    putc(0, outfile);
+    putc(0, outfile);
+    if (map_entry_size == 4)
+      putc(0, outfile);
+  }
+}
+
+
+/*
+ * Startup: write the file header unless the inversion array is being used.
+ */
+
+METHODDEF(void)
+start_output_bmp(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr)dinfo;
+
+  if (!dest->use_inversion_array) {
+    /* Write the header and colormap */
+    if (dest->is_os2)
+      write_os2_header(cinfo, dest);
+    else
+      write_bmp_header(cinfo, dest);
+  }
+}
+
+
+METHODDEF(void)
+finish_output_bmp(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr)dinfo;
+  register FILE *outfile = dest->pub.output_file;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW data_ptr;
+  JDIMENSION row;
+  cd_progress_ptr progress = (cd_progress_ptr)cinfo->progress;
+
+  if (dest->use_inversion_array) {
+    /* Write the header and colormap */
+    if (dest->is_os2)
+      write_os2_header(cinfo, dest);
+    else
+      write_bmp_header(cinfo, dest);
+
+    /* Write the file body from our virtual array */
+    for (row = cinfo->output_height; row > 0; row--) {
+      if (progress != NULL) {
+        progress->pub.pass_counter = (long)(cinfo->output_height - row);
+        progress->pub.pass_limit = (long)cinfo->output_height;
+        (*progress->pub.progress_monitor) ((j_common_ptr)cinfo);
+      }
+      image_ptr = (*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr)cinfo, dest->whole_image, row - 1, (JDIMENSION)1,
+         FALSE);
+      data_ptr = image_ptr0;
+      (void)JFWRITE(outfile, data_ptr, dest->row_width);
+    }
+    if (progress != NULL)
+      progress->completed_extra_passes++;
+  }
+
+  /* Make sure we wrote the output file OK */
+  fflush(outfile);
+  if (ferror(outfile))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for BMP format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_bmp(j_decompress_ptr cinfo, boolean is_os2,
+                boolean use_inversion_array)
+{
+  bmp_dest_ptr dest;
+  JDIMENSION row_width;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (bmp_dest_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(bmp_dest_struct));
+  dest->pub.start_output = start_output_bmp;
+  dest->pub.finish_output = finish_output_bmp;
+  dest->pub.calc_buffer_dimensions = NULL;
+  dest->is_os2 = is_os2;
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    dest->pub.put_pixel_rows = put_gray_rows;
+  } else if (IsExtRGB(cinfo->out_color_space)) {
+    if (cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = put_gray_rows;
+    else
+      dest->pub.put_pixel_rows = put_pixel_rows;
+  } else if (!cinfo->quantize_colors &&
+             (cinfo->out_color_space == JCS_RGB565 ||
+              cinfo->out_color_space == JCS_CMYK)) {
+    dest->pub.put_pixel_rows = put_pixel_rows;
+  } else {
+    ERREXIT(cinfo, JERR_BMP_COLORSPACE);
+  }
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Determine width of rows in the BMP file (padded to 4-byte boundary). */
+  if (cinfo->out_color_space == JCS_RGB565) {
+    row_width = cinfo->output_width * 2;
+    dest->row_width = dest->data_width = cinfo->output_width * 3;
+    while ((row_width & 3) != 0) row_width++;
+  } else if (!cinfo->quantize_colors &&
+             (IsExtRGB(cinfo->out_color_space) ||
+              cinfo->out_color_space == JCS_CMYK)) {
+    row_width = cinfo->output_width * cinfo->output_components;
+    dest->row_width = dest->data_width = cinfo->output_width * 3;
+  } else {
+    row_width = cinfo->output_width * cinfo->output_components;
+    dest->row_width = dest->data_width = row_width;
+  }
+  while ((dest->row_width & 3) != 0) dest->row_width++;
+  dest->pad_bytes = (int)(dest->row_width - dest->data_width);
+
+
+  if (use_inversion_array) {
+    /* Allocate space for inversion array, prepare for write pass */
+    dest->whole_image = (*cinfo->mem->request_virt_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE, FALSE,
+       dest->row_width, cinfo->output_height, (JDIMENSION)1);
+    dest->cur_output_row = 0;
+    if (cinfo->progress != NULL) {
+      cd_progress_ptr progress = (cd_progress_ptr)cinfo->progress;
+      progress->total_extra_passes++; /* count file input as separate pass */
+    }
+  } else {
+    dest->iobuffer = (JSAMPLE *)(*cinfo->mem->alloc_small)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE, dest->row_width);
+  }
+  dest->use_inversion_array = use_inversion_array;
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE, row_width, (JDIMENSION)1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr)dest;
+}
+
+#endif /* BMP_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/wrgif.c Added

@@ -0,0 +1,580 @@
+/*
+ * wrgif.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2015-2019 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, 2017, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in GIF format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ */
+
+/*
+ * This code is loosely based on ppmtogif from the PBMPLUS distribution
+ * of Feb. 1991.  That file contains the following copyright notice:
+ *    Based on GIFENCODE by David Rowley <mgardi@watdscu.waterloo.edu>.
+ *    Lempel-Ziv compression based on "compress" by Spencer W. Thomas et al.
+ *    Copyright (C) 1989 by Jef Poskanzer.
+ *    Permission to use, copy, modify, and distribute this software and its
+ *    documentation for any purpose and without fee is hereby granted, provided
+ *    that the above copyright notice appear in all copies and that both that
+ *    copyright notice and this permission notice appear in supporting
+ *    documentation.  This software is provided "as is" without express or
+ *    implied warranty.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef GIF_SUPPORTED
+
+
+#define MAX_LZW_BITS     12     /* maximum LZW code size (4096 symbols) */
+
+typedef INT16 code_int;         /* must hold -1 .. 2**MAX_LZW_BITS */
+
+#define LZW_TABLE_SIZE   ((code_int)1 << MAX_LZW_BITS)
+
+#define HSIZE            5003   /* hash table size for 80% occupancy */
+
+typedef int hash_int;           /* must hold -2*HSIZE..2*HSIZE */
+
+#define MAXCODE(n_bits)  (((code_int)1 << (n_bits)) - 1)
+
+
+/*
+ * The LZW hash table consists of two parallel arrays:
+ *   hash_codei       code of symbol in slot i, or 0 if empty slot
+ *   hash_valuei      symbol's value; undefined if empty slot
+ * where slot values (i) range from 0 to HSIZE-1.  The symbol value is
+ * its prefix symbol's code concatenated with its suffix character.
+ *
+ * Algorithm:  use open addressing double hashing (no chaining) on the
+ * prefix code / suffix character combination.  We do a variant of Knuth's
+ * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
+ * secondary probe.
+ */
+
+typedef int hash_entry;         /* must hold (code_int << 8) | byte */
+
+#define HASH_ENTRY(prefix, suffix)  ((((hash_entry)(prefix)) << 8) | (suffix))
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  j_decompress_ptr cinfo;       /* back link saves passing separate parm */
+
+  /* State for packing variable-width codes into a bitstream */
+  int n_bits;                   /* current number of bits/code */
+  code_int maxcode;             /* maximum code, given n_bits */
+  int init_bits;                /* initial n_bits ... restored after clear */
+  int cur_accum;                /* holds bits not yet output */
+  int cur_bits;                 /* # of bits in cur_accum */
+
+  /* LZW string construction */
+  code_int waiting_code;        /* symbol not yet output; may be extendable */
+  boolean first_byte;           /* if TRUE, waiting_code is not valid */
+
+  /* State for GIF code assignment */
+  code_int ClearCode;           /* clear code (doesn't change) */
+  code_int EOFCode;             /* EOF code (ditto) */
+  code_int free_code;           /* LZW: first not-yet-used symbol code */
+  code_int code_counter;        /* not LZW: counts output symbols */
+
+  /* LZW hash table */
+  code_int *hash_code;          /* => hash table of symbol codes */
+  hash_entry *hash_value;       /* => hash table of symbol values */
+
+  /* GIF data packet construction buffer */
+  int bytesinpkt;               /* # of bytes in current packet */
+  char packetbuf256;          /* workspace for accumulating packet */
+
+} gif_dest_struct;
+
+typedef gif_dest_struct *gif_dest_ptr;
+
+
+/*
+ * Routines to package finished data bytes into GIF data blocks.
+ * A data block consists of a count byte (1..255) and that many data bytes.
+ */
+
+LOCAL(void)
+flush_packet(gif_dest_ptr dinfo)
+/* flush any accumulated data */
+{
+  if (dinfo->bytesinpkt > 0) {  /* never write zero-length packet */
+    dinfo->packetbuf0 = (char)dinfo->bytesinpkt++;
+    if (JFWRITE(dinfo->pub.output_file, dinfo->packetbuf, dinfo->bytesinpkt) !=
+        (size_t)dinfo->bytesinpkt)
+      ERREXIT(dinfo->cinfo, JERR_FILE_WRITE);
+    dinfo->bytesinpkt = 0;
+  }
+}
+
+
+/* Add a character to current packet; flush to disk if necessary */
+#define CHAR_OUT(dinfo, c) { \
+  (dinfo)->packetbuf++(dinfo)->bytesinpkt = (char)(c); \
+  if ((dinfo)->bytesinpkt >= 255) \
+    flush_packet(dinfo); \
+}
+
+
+/* Routine to convert variable-width codes into a byte stream */
+
+LOCAL(void)
+output(gif_dest_ptr dinfo, code_int code)
+/* Emit a code of n_bits bits */
+/* Uses cur_accum and cur_bits to reblock into 8-bit bytes */
+{
+  dinfo->cur_accum |= ((long)code) << dinfo->cur_bits;
+  dinfo->cur_bits += dinfo->n_bits;
+
+  while (dinfo->cur_bits >= 8) {
+    CHAR_OUT(dinfo, dinfo->cur_accum & 0xFF);
+    dinfo->cur_accum >>= 8;
+    dinfo->cur_bits -= 8;
+  }
+
+  /*
+   * If the next entry is going to be too big for the code size,
+   * then increase it, if possible.  We do this here to ensure
+   * that it's done in sync with the decoder's codesize increases.
+   */
+  if (dinfo->free_code > dinfo->maxcode) {
+    dinfo->n_bits++;
+    if (dinfo->n_bits == MAX_LZW_BITS)
+      dinfo->maxcode = LZW_TABLE_SIZE; /* free_code will never exceed this */
+    else
+      dinfo->maxcode = MAXCODE(dinfo->n_bits);
+  }
+}
+
+
+/* Compression initialization & termination */
+
+
+LOCAL(void)
+clear_hash(gif_dest_ptr dinfo)
+/* Fill the hash table with empty entries */
+{
+  /* It's sufficient to zero hash_code */
+  MEMZERO(dinfo->hash_code, HSIZE * sizeof(code_int));
+}
+
+
+LOCAL(void)
+clear_block(gif_dest_ptr dinfo)
+/* Reset compressor and issue a Clear code */
+{
+  clear_hash(dinfo);                    /* delete all the symbols */
+  dinfo->free_code = dinfo->ClearCode + 2;
+  output(dinfo, dinfo->ClearCode);      /* inform decoder */
+  dinfo->n_bits = dinfo->init_bits;     /* reset code size */
+  dinfo->maxcode = MAXCODE(dinfo->n_bits);
+}
+
+
+LOCAL(void)
+compress_init(gif_dest_ptr dinfo, int i_bits)
+/* Initialize compressor */
+{
+  /* init all the state variables */
+  dinfo->n_bits = dinfo->init_bits = i_bits;
+  dinfo->maxcode = MAXCODE(dinfo->n_bits);
+  dinfo->ClearCode = ((code_int) 1 << (i_bits - 1));
+  dinfo->EOFCode = dinfo->ClearCode + 1;
+  dinfo->code_counter = dinfo->free_code = dinfo->ClearCode + 2;
+  dinfo->first_byte = TRUE;     /* no waiting symbol yet */
+  /* init output buffering vars */
+  dinfo->bytesinpkt = 0;
+  dinfo->cur_accum = 0;
+  dinfo->cur_bits = 0;
+  /* clear hash table */
+  if (dinfo->hash_code != NULL)
+    clear_hash(dinfo);
+  /* GIF specifies an initial Clear code */
+  output(dinfo, dinfo->ClearCode);
+}
+
+
+LOCAL(void)
+compress_term(gif_dest_ptr dinfo)
+/* Clean up at end */
+{
+  /* Flush out the buffered LZW code */
+  if (!dinfo->first_byte)
+    output(dinfo, dinfo->waiting_code);
+  /* Send an EOF code */
+  output(dinfo, dinfo->EOFCode);
+  /* Flush the bit-packing buffer */
+  if (dinfo->cur_bits > 0) {
+    CHAR_OUT(dinfo, dinfo->cur_accum & 0xFF);
+  }
+  /* Flush the packet buffer */
+  flush_packet(dinfo);
+}
+
+
+/* GIF header construction */
+
+
+LOCAL(void)
+put_word(gif_dest_ptr dinfo, unsigned int w)
+/* Emit a 16-bit word, LSB first */
+{
+  putc(w & 0xFF, dinfo->pub.output_file);
+  putc((w >> 8) & 0xFF, dinfo->pub.output_file);
+}
+
+
+LOCAL(void)
+put_3bytes(gif_dest_ptr dinfo, int val)
+/* Emit 3 copies of same byte value --- handy subr for colormap construction */
+{
+  putc(val, dinfo->pub.output_file);
+  putc(val, dinfo->pub.output_file);
+  putc(val, dinfo->pub.output_file);
+}
+
+
+LOCAL(void)
+emit_header(gif_dest_ptr dinfo, int num_colors, JSAMPARRAY colormap)
+/* Output the GIF file header, including color map */
+/* If colormap == NULL, synthesize a grayscale colormap */
+{
+  int BitsPerPixel, ColorMapSize, InitCodeSize, FlagByte;
+  int cshift = dinfo->cinfo->data_precision - 8;
+  int i;
+
+  if (num_colors > 256)
+    ERREXIT1(dinfo->cinfo, JERR_TOO_MANY_COLORS, num_colors);
+  /* Compute bits/pixel and related values */
+  BitsPerPixel = 1;
+  while (num_colors > (1 << BitsPerPixel))
+    BitsPerPixel++;
+  ColorMapSize = 1 << BitsPerPixel;
+  if (BitsPerPixel <= 1)
+    InitCodeSize = 2;
+  else
+    InitCodeSize = BitsPerPixel;
+  /*
+   * Write the GIF header.
+   * Note that we generate a plain GIF87 header for maximum compatibility.
+   */
+  putc('G', dinfo->pub.output_file);
+  putc('I', dinfo->pub.output_file);
+  putc('F', dinfo->pub.output_file);
+  putc('8', dinfo->pub.output_file);
+  putc('7', dinfo->pub.output_file);
+  putc('a', dinfo->pub.output_file);
+  /* Write the Logical Screen Descriptor */
+  put_word(dinfo, (unsigned int)dinfo->cinfo->output_width);
+  put_word(dinfo, (unsigned int)dinfo->cinfo->output_height);
+  FlagByte = 0x80;              /* Yes, there is a global color table */
+  FlagByte |= (BitsPerPixel - 1) << 4; /* color resolution */
+  FlagByte |= (BitsPerPixel - 1); /* size of global color table */
+  putc(FlagByte, dinfo->pub.output_file);
+  putc(0, dinfo->pub.output_file); /* Background color index */
+  putc(0, dinfo->pub.output_file); /* Reserved (aspect ratio in GIF89) */
+  /* Write the Global Color Map */
+  /* If the color map is more than 8 bits precision, */
+  /* we reduce it to 8 bits by shifting */
+  for (i = 0; i < ColorMapSize; i++) {
+    if (i < num_colors) {
+      if (colormap != NULL) {
+        if (dinfo->cinfo->out_color_space == JCS_RGB) {
+          /* Normal case: RGB color map */
+          putc(colormap0i >> cshift, dinfo->pub.output_file);
+          putc(colormap1i >> cshift, dinfo->pub.output_file);
+          putc(colormap2i >> cshift, dinfo->pub.output_file);
+        } else {
+          /* Grayscale "color map": possible if quantizing grayscale image */
+          put_3bytes(dinfo, colormap0i >> cshift);
+        }
+      } else {
+        /* Create a grayscale map of num_colors values, range 0..255 */
+        put_3bytes(dinfo, (i * 255 + (num_colors - 1) / 2) / (num_colors - 1));
+      }
+    } else {
+      /* fill out the map to a power of 2 */
+      put_3bytes(dinfo, CENTERJSAMPLE >> cshift);
+    }
+  }
+  /* Write image separator and Image Descriptor */
+  putc(',', dinfo->pub.output_file); /* separator */
+  put_word(dinfo, 0);           /* left/top offset */
+  put_word(dinfo, 0);
+  put_word(dinfo, (unsigned int)dinfo->cinfo->output_width); /* image size */
+  put_word(dinfo, (unsigned int)dinfo->cinfo->output_height);
+  /* flag byte: not interlaced, no local color map */
+  putc(0x00, dinfo->pub.output_file);
+  /* Write Initial Code Size byte */
+  putc(InitCodeSize, dinfo->pub.output_file);
+
+  /* Initialize for compression of image data */
+  compress_init(dinfo, InitCodeSize + 1);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_gif(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  gif_dest_ptr dest = (gif_dest_ptr)dinfo;
+
+  if (cinfo->quantize_colors)
+    emit_header(dest, cinfo->actual_number_of_colors, cinfo->colormap);
+  else
+    emit_header(dest, 256, (JSAMPARRAY)NULL);
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+
+/*
+ * The LZW algorithm proper
+ */
+
+METHODDEF(void)
+put_LZW_pixel_rows(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                   JDIMENSION rows_supplied)
+{
+  gif_dest_ptr dest = (gif_dest_ptr)dinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  code_int c;
+  register hash_int i;
+  register hash_int disp;
+  register hash_entry probe_value;
+
+  ptr = dest->pub.buffer0;
+  for (col = cinfo->output_width; col > 0; col--) {
+    /* Accept and compress one 8-bit byte */
+    c = (code_int)(*ptr++);
+
+    if (dest->first_byte) {     /* need to initialize waiting_code */
+      dest->waiting_code = c;
+      dest->first_byte = FALSE;
+      continue;
+    }
+
+    /* Probe hash table to see if a symbol exists for
+     * waiting_code followed by c.
+     * If so, replace waiting_code by that symbol and continue.
+     */
+    i = ((hash_int)c << (MAX_LZW_BITS - 8)) + dest->waiting_code;
+    /* i is less than twice 2**MAX_LZW_BITS, therefore less than twice HSIZE */
+    if (i >= HSIZE)
+      i -= HSIZE;
+
+    probe_value = HASH_ENTRY(dest->waiting_code, c);
+
+    if (dest->hash_codei == 0) {
+      /* hit empty slot; desired symbol not in table */
+      output(dest, dest->waiting_code);
+      if (dest->free_code < LZW_TABLE_SIZE) {
+        dest->hash_codei = dest->free_code++; /* add symbol to hashtable */
+        dest->hash_valuei = probe_value;
+      } else
+        clear_block(dest);
+      dest->waiting_code = c;
+      continue;
+    }
+    if (dest->hash_valuei == probe_value) {
+      dest->waiting_code = dest->hash_codei;
+      continue;
+    }
+
+    if (i == 0)                 /* secondary hash (after G. Knott) */
+      disp = 1;
+    else
+      disp = HSIZE - i;
+    for (;;) {
+      i -= disp;
+      if (i < 0)
+        i += HSIZE;
+      if (dest->hash_codei == 0) {
+        /* hit empty slot; desired symbol not in table */
+        output(dest, dest->waiting_code);
+        if (dest->free_code < LZW_TABLE_SIZE) {
+          dest->hash_codei = dest->free_code++; /* add symbol to hashtable */
+          dest->hash_valuei = probe_value;
+        } else
+          clear_block(dest);
+        dest->waiting_code = c;
+        break;
+      }
+      if (dest->hash_valuei == probe_value) {
+        dest->waiting_code = dest->hash_codei;
+        break;
+      }
+    }
+  }
+}
+
+
+/*
+ * The pseudo-compression algorithm.
+ *
+ * In this version we simply output each pixel value as a separate symbol;
+ * thus, no compression occurs.  In fact, there is expansion of one bit per
+ * pixel, because we use a symbol width one bit wider than the pixel width.
+ *
+ * GIF ordinarily uses variable-width symbols, and the decoder will expect
+ * to ratchet up the symbol width after a fixed number of symbols.
+ * To simplify the logic and keep the expansion penalty down, we emit a
+ * GIF Clear code to reset the decoder just before the width would ratchet up.
+ * Thus, all the symbols in the output file will have the same bit width.
+ * Note that emitting the Clear codes at the right times is a mere matter of
+ * counting output symbols and is in no way dependent on the LZW algorithm.
+ *
+ * With a small basic pixel width (low color count), Clear codes will be
+ * needed very frequently, causing the file to expand even more.  So this
+ * simplistic approach wouldn't work too well on bilevel images, for example.
+ * But for output of JPEG conversions the pixel width will usually be 8 bits
+ * (129 to 256 colors), so the overhead added by Clear symbols is only about
+ * one symbol in every 256.
+ */
+
+METHODDEF(void)
+put_raw_pixel_rows(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                   JDIMENSION rows_supplied)
+{
+  gif_dest_ptr dest = (gif_dest_ptr)dinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  code_int c;
+
+  ptr = dest->pub.buffer0;
+  for (col = cinfo->output_width; col > 0; col--) {
+    c = (code_int)(*ptr++);
+    /* Accept and output one pixel value.
+     * The given value must be less than n_bits wide.
+     */
+
+    /* Output the given pixel value as a symbol. */
+    output(dest, c);
+    /* Issue Clear codes often enough to keep the reader from ratcheting up
+     * its symbol size.
+     */
+    if (dest->code_counter < dest->maxcode) {
+      dest->code_counter++;
+    } else {
+      output(dest, dest->ClearCode);
+      dest->code_counter = dest->ClearCode + 2; /* reset the counter */
+    }
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_gif(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  gif_dest_ptr dest = (gif_dest_ptr)dinfo;
+
+  /* Flush compression mechanism */
+  compress_term(dest);
+  /* Write a zero-length data block to end the series */
+  putc(0, dest->pub.output_file);
+  /* Write the GIF terminator mark */
+  putc(';', dest->pub.output_file);
+  /* Make sure we wrote the output file OK */
+  fflush(dest->pub.output_file);
+  if (ferror(dest->pub.output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * Re-calculate buffer dimensions based on output dimensions.
+ */
+
+METHODDEF(void)
+calc_buffer_dimensions_gif(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+}
+
+
+/*
+ * The module selection routine for GIF format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_gif(j_decompress_ptr cinfo, boolean is_lzw)
+{
+  gif_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (gif_dest_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(gif_dest_struct));
+  dest->cinfo = cinfo;          /* make back link for subroutines */
+  dest->pub.start_output = start_output_gif;
+  dest->pub.finish_output = finish_output_gif;
+  dest->pub.calc_buffer_dimensions = calc_buffer_dimensions_gif;
+
+  if (cinfo->out_color_space != JCS_GRAYSCALE &&
+      cinfo->out_color_space != JCS_RGB)
+    ERREXIT(cinfo, JERR_GIF_COLORSPACE);
+
+  /* Force quantization if color or if > 8 bits input */
+  if (cinfo->out_color_space != JCS_GRAYSCALE || cinfo->data_precision > 8) {
+    /* Force quantization to at most 256 colors */
+    cinfo->quantize_colors = TRUE;
+    if (cinfo->desired_number_of_colors > 256)
+      cinfo->desired_number_of_colors = 256;
+  }
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  if (cinfo->output_components != 1) /* safety check: just one component? */
+    ERREXIT(cinfo, JERR_GIF_BUG);
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE, cinfo->output_width, (JDIMENSION)1);
+  dest->pub.buffer_height = 1;
+
+  if (is_lzw) {
+    dest->pub.put_pixel_rows = put_LZW_pixel_rows;
+    /* Allocate space for hash table */
+    dest->hash_code = (code_int *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  HSIZE * sizeof(code_int));
+    dest->hash_value = (hash_entry *)
+      (*cinfo->mem->alloc_large) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  HSIZE * sizeof(hash_entry));
+  } else {
+    dest->pub.put_pixel_rows = put_raw_pixel_rows;
+    /* Mark tables unused */
+    dest->hash_code = NULL;
+    dest->hash_value = NULL;
+  }
+
+  return (djpeg_dest_ptr)dest;
+}
+
+#endif /* GIF_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/wrjpgcom.1 Added

@@ -0,0 +1,103 @@
+.TH WRJPGCOM 1 "15 June 1995"
+.SH NAME
+wrjpgcom \- insert text comments into a JPEG file
+.SH SYNOPSIS
+.B wrjpgcom
+
+.B \-replace
+
+
+.BI \-comment " text"
+
+
+.BI \-cfile " name"
+
+
+.I filename
+
+.LP
+.SH DESCRIPTION
+.LP
+.B wrjpgcom
+reads the named JPEG/JFIF file, or the standard input if no file is named,
+and generates a new JPEG/JFIF file on standard output.  A comment block is
+added to the file.
+.PP
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+.PP
+.B wrjpgcom
+adds a COM block, containing text you provide, to a JPEG file.
+Ordinarily, the COM block is added after any existing COM blocks; but you
+can delete the old COM blocks if you wish.
+.SH OPTIONS
+Switch names may be abbreviated, and are not case sensitive.
+.TP
+.B \-replace
+Delete any existing COM blocks from the file.
+.TP
+.BI \-comment " text"
+Supply text for new COM block on command line.
+.TP
+.BI \-cfile " name"
+Read text for new COM block from named file.
+.PP
+If you have only one line of comment text to add, you can provide it on the
+command line with
+.BR \-comment .
+The comment text must be surrounded with quotes so that it is treated as a
+single argument.  Longer comments can be read from a text file.
+.PP
+If you give neither
+.B \-comment
+nor
+.BR \-cfile,
+then
+.B wrjpgcom
+will read the comment text from standard input.  (In this case an input image
+file name MUST be supplied, so that the source JPEG file comes from somewhere
+else.)  You can enter multiple lines, up to 64KB worth.  Type an end-of-file
+indicator (usually control-D) to terminate the comment text entry.
+.PP
+.B wrjpgcom
+will not add a COM block if the provided comment string is empty.  Therefore
+\fB\-replace \-comment ""\fR can be used to delete all COM blocks from a file.
+.SH EXAMPLES
+.LP
+Add a short comment to in.jpg, producing out.jpg:
+.IP
+.B wrjpgcom \-c
+\fI"View of my back yard" in.jpg
+.B >
+.I out.jpg
+.PP
+Attach a long comment previously stored in comment.txt:
+.IP
+.B wrjpgcom
+.I in.jpg
+.B <
+.I comment.txt
+.B >
+.I out.jpg
+.PP
+or equivalently
+.IP
+.B wrjpgcom
+.B -cfile
+.I comment.txt
+.B <
+.I in.jpg
+.B >
+.I out.jpg
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1)
+.SH AUTHOR
+Independent JPEG Group

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/wrjpgcom.c Added

@@ -0,0 +1,591 @@
+/*
+ * wrjpgcom.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a very simple stand-alone application that inserts
+ * user-supplied text as a COM (comment) marker in a JFIF file.
+ * This may be useful as an example of the minimum logic needed to parse
+ * JPEG markers.
+ */
+
+#define JPEG_CJPEG_DJPEG        /* to get the command-line config symbols */
+#include "jinclude.h"           /* get auto-config symbols, <stdio.h> */
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc() */
+extern void *malloc();
+#endif
+#include <ctype.h>              /* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>              /* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>                 /* to declare setmode() */
+#endif
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+#ifdef DONT_USE_B_MODE          /* define mode parameters for fopen() */
+#define READ_BINARY     "r"
+#define WRITE_BINARY    "w"
+#else
+#define READ_BINARY     "rb"
+#define WRITE_BINARY    "wb"
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#define EXIT_SUCCESS  0
+#endif
+
+/* Reduce this value if your malloc() can't allocate blocks up to 64K.
+ * On DOS, compiling in large model is usually a better solution.
+ */
+
+#ifndef MAX_COM_LENGTH
+#define MAX_COM_LENGTH  65000L  /* must be <= 65533 in any case */
+#endif
+
+
+/*
+ * These macros are used to read the input file and write the output file.
+ * To reuse this code in another application, you might need to change these.
+ */
+
+static FILE *infile;            /* input JPEG file */
+
+/* Return next input byte, or EOF if no more */
+#define NEXTBYTE()  getc(infile)
+
+static FILE *outfile;           /* output JPEG file */
+
+/* Emit an output byte */
+#define PUTBYTE(x)  putc((x), outfile)
+
+
+/* Error exit handler */
+#define ERREXIT(msg)  (fprintf(stderr, "%s\n", msg), exit(EXIT_FAILURE))
+
+
+/* Read one byte, testing for EOF */
+static int
+read_1_byte(void)
+{
+  int c;
+
+  c = NEXTBYTE();
+  if (c == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return c;
+}
+
+/* Read 2 bytes, convert to unsigned int */
+/* All 2-byte quantities in JPEG markers are MSB first */
+static unsigned int
+read_2_bytes(void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  if (c1 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  c2 = NEXTBYTE();
+  if (c2 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return (((unsigned int)c1) << 8) + ((unsigned int)c2);
+}
+
+
+/* Routines to write data to output file */
+
+static void
+write_1_byte(int c)
+{
+  PUTBYTE(c);
+}
+
+static void
+write_2_bytes(unsigned int val)
+{
+  PUTBYTE((val >> 8) & 0xFF);
+  PUTBYTE(val & 0xFF);
+}
+
+static void
+write_marker(int marker)
+{
+  PUTBYTE(0xFF);
+  PUTBYTE(marker);
+}
+
+static void
+copy_rest_of_file(void)
+{
+  int c;
+
+  while ((c = NEXTBYTE()) != EOF)
+    PUTBYTE(c);
+}
+
+
+/*
+ * JPEG markers consist of one or more 0xFF bytes, followed by a marker
+ * code byte (which is not an FF).  Here are the marker codes of interest
+ * in this program.  (See jdmarker.c for a more complete list.)
+ */
+
+#define M_SOF0   0xC0           /* Start Of Frame N */
+#define M_SOF1   0xC1           /* N indicates which compression process */
+#define M_SOF2   0xC2           /* Only SOF0-SOF2 are now in common use */
+#define M_SOF3   0xC3
+#define M_SOF5   0xC5           /* NB: codes C4 and CC are NOT SOF markers */
+#define M_SOF6   0xC6
+#define M_SOF7   0xC7
+#define M_SOF9   0xC9
+#define M_SOF10  0xCA
+#define M_SOF11  0xCB
+#define M_SOF13  0xCD
+#define M_SOF14  0xCE
+#define M_SOF15  0xCF
+#define M_SOI    0xD8           /* Start Of Image (beginning of datastream) */
+#define M_EOI    0xD9           /* End Of Image (end of datastream) */
+#define M_SOS    0xDA           /* Start Of Scan (begins compressed data) */
+#define M_COM    0xFE           /* COMment */
+
+
+/*
+ * Find the next JPEG marker and return its marker code.
+ * We expect at least one FF byte, possibly more if the compressor used FFs
+ * to pad the file.  (Padding FFs will NOT be replicated in the output file.)
+ * There could also be non-FF garbage between markers.  The treatment of such
+ * garbage is unspecified; we choose to skip over it but emit a warning msg.
+ * NB: this routine must not be used after seeing SOS marker, since it will
+ * not deal correctly with FF/00 sequences in the compressed image data...
+ */
+
+static int
+next_marker(void)
+{
+  int c;
+  int discarded_bytes = 0;
+
+  /* Find 0xFF byte; count and skip any non-FFs. */
+  c = read_1_byte();
+  while (c != 0xFF) {
+    discarded_bytes++;
+    c = read_1_byte();
+  }
+  /* Get marker code byte, swallowing any duplicate FF bytes.  Extra FFs
+   * are legal as pad bytes, so don't count them in discarded_bytes.
+   */
+  do {
+    c = read_1_byte();
+  } while (c == 0xFF);
+
+  if (discarded_bytes != 0) {
+    fprintf(stderr, "Warning: garbage data found in JPEG file\n");
+  }
+
+  return c;
+}
+
+
+/*
+ * Read the initial marker, which should be SOI.
+ * For a JFIF file, the first two bytes of the file should be literally
+ * 0xFF M_SOI.  To be more general, we could use next_marker, but if the
+ * input file weren't actually JPEG at all, next_marker might read the whole
+ * file and then return a misleading error message...
+ */
+
+static int
+first_marker(void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  c2 = NEXTBYTE();
+  if (c1 != 0xFF || c2 != M_SOI)
+    ERREXIT("Not a JPEG file");
+  return c2;
+}
+
+
+/*
+ * Most types of marker are followed by a variable-length parameter segment.
+ * This routine skips over the parameters for any marker we don't otherwise
+ * want to process.
+ * Note that we MUST skip the parameter segment explicitly in order not to
+ * be fooled by 0xFF bytes that might appear within the parameter segment;
+ * such bytes do NOT introduce new markers.
+ */
+
+static void
+copy_variable(void)
+/* Copy an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  write_2_bytes(length);
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Copy the remaining bytes */
+  while (length > 0) {
+    write_1_byte(read_1_byte());
+    length--;
+  }
+}
+
+static void
+skip_variable(void)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    (void)read_1_byte();
+    length--;
+  }
+}
+
+
+/*
+ * Parse the marker stream until SOFn or EOI is seen;
+ * copy data to output, but discard COM markers unless keep_COM is true.
+ */
+
+static int
+scan_JPEG_header(int keep_COM)
+{
+  int marker;
+
+  /* Expect SOI at start of file */
+  if (first_marker() != M_SOI)
+    ERREXIT("Expected SOI marker first");
+  write_marker(M_SOI);
+
+  /* Scan miscellaneous markers until we reach SOFn. */
+  for (;;) {
+    marker = next_marker();
+    switch (marker) {
+      /* Note that marker codes 0xC4, 0xC8, 0xCC are not, and must not be,
+       * treated as SOFn.  C4 in particular is actually DHT.
+       */
+    case M_SOF0:                /* Baseline */
+    case M_SOF1:                /* Extended sequential, Huffman */
+    case M_SOF2:                /* Progressive, Huffman */
+    case M_SOF3:                /* Lossless, Huffman */
+    case M_SOF5:                /* Differential sequential, Huffman */
+    case M_SOF6:                /* Differential progressive, Huffman */
+    case M_SOF7:                /* Differential lossless, Huffman */
+    case M_SOF9:                /* Extended sequential, arithmetic */
+    case M_SOF10:               /* Progressive, arithmetic */
+    case M_SOF11:               /* Lossless, arithmetic */
+    case M_SOF13:               /* Differential sequential, arithmetic */
+    case M_SOF14:               /* Differential progressive, arithmetic */
+    case M_SOF15:               /* Differential lossless, arithmetic */
+      return marker;
+
+    case M_SOS:                 /* should not see compressed data before SOF */
+      ERREXIT("SOS without prior SOFn");
+      break;
+
+    case M_EOI:                 /* in case it's a tables-only JPEG stream */
+      return marker;
+
+    case M_COM:                 /* Existing COM: conditionally discard */
+      if (keep_COM) {
+        write_marker(marker);
+        copy_variable();
+      } else {
+        skip_variable();
+      }
+      break;
+
+    default:                    /* Anything else just gets copied */
+      write_marker(marker);
+      copy_variable();          /* we assume it has a parameter count... */
+      break;
+    }
+  } /* end loop */
+}
+
+
+/* Command line parsing code */
+
+static const char *progname;    /* program name for error messages */
+
+
+static void
+usage(void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "wrjpgcom inserts a textual comment in a JPEG file.\n");
+  fprintf(stderr, "You can add to or replace any existing comment(s).\n");
+
+  fprintf(stderr, "Usage: %s switches ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "inputfile\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -replace         Delete any existing comments\n");
+  fprintf(stderr, "  -comment \"text\"  Insert comment with given text\n");
+  fprintf(stderr, "  -cfile name      Read comment from named file\n");
+  fprintf(stderr, "Notice that you must put quotes around the comment text\n");
+  fprintf(stderr, "when you use -comment.\n");
+  fprintf(stderr, "If you do not give either -comment or -cfile on the command line,\n");
+  fprintf(stderr, "then the comment text is read from standard input.\n");
+  fprintf(stderr, "It can be multiple lines, up to %u characters total.\n",
+          (unsigned int)MAX_COM_LENGTH);
+#ifndef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "You must specify an input JPEG file name when supplying\n");
+  fprintf(stderr, "comment text from standard input.\n");
+#endif
+
+  exit(EXIT_FAILURE);
+}
+
+
+static int
+keymatch(char *arg, const char *keyword, int minchars)
+/* Case-insensitive matching of (possibly abbreviated) keyword switches. */
+/* keyword is the constant keyword (must be lower case already), */
+/* minchars is length of minimum legal abbreviation. */
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return 0;                 /* arg longer than keyword, no good */
+    if (isupper(ca))            /* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return 0;                 /* no good */
+    nmatched++;                 /* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return 0;
+  return 1;                     /* A-OK */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main(int argc, char **argv)
+{
+  int argn;
+  char *arg;
+  int keep_COM = 1;
+  char *comment_arg = NULL;
+  FILE *comment_file = NULL;
+  unsigned int comment_length = 0;
+  int marker;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv0;
+  if (progname == NULL || progname0 == 0)
+    progname = "wrjpgcom";      /* in case C library doesn't provide it */
+
+  /* Parse switches, if any */
+  for (argn = 1; argn < argc; argn++) {
+    arg = argvargn;
+    if (arg0 != '-')
+      break;                    /* not switch, must be file name */
+    arg++;                      /* advance over '-' */
+    if (keymatch(arg, "replace", 1)) {
+      keep_COM = 0;
+    } else if (keymatch(arg, "cfile", 2)) {
+      if (++argn >= argc) usage();
+      if ((comment_file = fopen(argvargn, "r")) == NULL) {
+        fprintf(stderr, "%s: can't open %s\n", progname, argvargn);
+        exit(EXIT_FAILURE);
+      }
+    } else if (keymatch(arg, "comment", 1)) {
+      if (++argn >= argc) usage();
+      comment_arg = argvargn;
+      /* If the comment text starts with '"', then we are probably running
+       * under MS-DOG and must parse out the quoted string ourselves.  Sigh.
+       */
+      if (comment_arg0 == '"') {
+        comment_arg = (char *)malloc((size_t)MAX_COM_LENGTH);
+        if (comment_arg == NULL)
+          ERREXIT("Insufficient memory");
+        if (strlen(argvargn) + 2 >= (size_t)MAX_COM_LENGTH) {
+          fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                  (unsigned int)MAX_COM_LENGTH);
+          exit(EXIT_FAILURE);
+        }
+        strcpy(comment_arg, argvargn + 1);
+        for (;;) {
+          comment_length = (unsigned int)strlen(comment_arg);
+          if (comment_length > 0 && comment_argcomment_length - 1 == '"') {
+            comment_argcomment_length - 1 = '\0'; /* zap terminating quote */
+            break;
+          }
+          if (++argn >= argc)
+            ERREXIT("Missing ending quote mark");
+          if (strlen(comment_arg) + strlen(argvargn) + 2 >=
+              (size_t)MAX_COM_LENGTH) {
+            fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                    (unsigned int)MAX_COM_LENGTH);
+            exit(EXIT_FAILURE);
+          }
+          strcat(comment_arg, " ");
+          strcat(comment_arg, argvargn);
+        }
+      } else if (strlen(argvargn) >= (size_t)MAX_COM_LENGTH) {
+        fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                (unsigned int)MAX_COM_LENGTH);
+        exit(EXIT_FAILURE);
+      }
+      comment_length = (unsigned int)strlen(comment_arg);
+    } else
+      usage();
+  }
+
+  /* Cannot use both -comment and -cfile. */
+  if (comment_arg != NULL && comment_file != NULL)
+    usage();
+  /* If there is neither -comment nor -cfile, we will read the comment text
+   * from stdin; in this case there MUST be an input JPEG file name.
+   */
+  if (comment_arg == NULL && comment_file == NULL && argn >= argc)
+    usage();
+
+  /* Open the input file. */
+  if (argn < argc) {
+    if ((infile = fopen(argvargn, READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argvargn);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+#ifdef USE_SETMODE              /* need to hack file mode? */
+    setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+    if ((infile = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open stdin\n", progname);
+      exit(EXIT_FAILURE);
+    }
+#else
+    infile = stdin;
+#endif
+  }
+
+  /* Open the output file. */
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have explicit output file name */
+  if (argn != argc - 2) {
+    fprintf(stderr, "%s: must name one input and one output file\n", progname);
+    usage();
+  }
+  if ((outfile = fopen(argvargn + 1, WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "%s: can't open %s\n", progname, argvargn + 1);
+    exit(EXIT_FAILURE);
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (argn < argc - 1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+  /* default output file is stdout */
+#ifdef USE_SETMODE              /* need to hack file mode? */
+  setmode(fileno(stdout), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+  if ((outfile = fdopen(fileno(stdout), WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "%s: can't open stdout\n", progname);
+    exit(EXIT_FAILURE);
+  }
+#else
+  outfile = stdout;
+#endif
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Collect comment text from comment_file or stdin, if necessary */
+  if (comment_arg == NULL) {
+    FILE *src_file;
+    int c;
+
+    comment_arg = (char *)malloc((size_t)MAX_COM_LENGTH);
+    if (comment_arg == NULL)
+      ERREXIT("Insufficient memory");
+    comment_length = 0;
+    src_file = (comment_file != NULL ? comment_file : stdin);
+    while ((c = getc(src_file)) != EOF) {
+      if (comment_length >= (unsigned int)MAX_COM_LENGTH) {
+        fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                (unsigned int)MAX_COM_LENGTH);
+        exit(EXIT_FAILURE);
+      }
+      comment_argcomment_length++ = (char)c;
+    }
+    if (comment_file != NULL)
+      fclose(comment_file);
+  }
+
+  /* Copy JPEG headers until SOFn marker;
+   * we will insert the new comment marker just before SOFn.
+   * This (a) causes the new comment to appear after, rather than before,
+   * existing comments; and (b) ensures that comments come after any JFIF
+   * or JFXX markers, as required by the JFIF specification.
+   */
+  marker = scan_JPEG_header(keep_COM);
+  /* Insert the new COM marker, but only if nonempty text has been supplied */
+  if (comment_length > 0) {
+    write_marker(M_COM);
+    write_2_bytes(comment_length + 2);
+    while (comment_length > 0) {
+      write_1_byte(*comment_arg++);
+      comment_length--;
+    }
+  }
+  /* Duplicate the remainder of the source file.
+   * Note that any COM markers occurring after SOF will not be touched.
+   */
+  write_marker(marker);
+  copy_rest_of_file();
+
+  /* All done. */
+  exit(EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/wrppm.c Added

@@ -0,0 +1,366 @@
+/*
+ * wrppm.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * Modified 2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2017, 2019-2020, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in PPM/PGM format.
+ * The extended 2-byte-per-sample raw PPM/PGM formats are supported.
+ * The PBMPLUS library is NOT required to compile this software
+ * (but it is highly useful as a set of PPM image manipulation programs).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ */
+
+#include "cmyk.h"
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef PPM_SUPPORTED
+
+
+/*
+ * For 12-bit JPEG data, we either downscale the values to 8 bits
+ * (to write standard byte-per-sample PPM/PGM files), or output
+ * nonstandard word-per-sample PPM/PGM files.  Downscaling is done
+ * if PPM_NORAWWORD is defined (this can be done in the Makefile
+ * or in jconfig.h).
+ * (When the core library supports data precision reduction, a cleaner
+ * implementation will be to ask for that instead.)
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define PUTPPMSAMPLE(ptr, v)  *ptr++ = (char)(v)
+#define BYTESPERSAMPLE  1
+#define PPM_MAXVAL  255
+#else
+#ifdef PPM_NORAWWORD
+#define PUTPPMSAMPLE(ptr, v)  *ptr++ = (char)((v) >> (BITS_IN_JSAMPLE - 8))
+#define BYTESPERSAMPLE  1
+#define PPM_MAXVAL  255
+#else
+/* The word-per-sample format always puts the MSB first. */
+#define PUTPPMSAMPLE(ptr, v) { \
+  register int val_ = v; \
+  *ptr++ = (char)((val_ >> 8) & 0xFF); \
+  *ptr++ = (char)(val_ & 0xFF); \
+}
+#define BYTESPERSAMPLE  2
+#define PPM_MAXVAL  ((1 << BITS_IN_JSAMPLE) - 1)
+#endif
+#endif
+
+
+/*
+ * When JSAMPLE is the same size as char, we can just fwrite() the
+ * decompressed data to the PPM or PGM file.
+ */
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  /* Usually these two pointers point to the same place: */
+  char *iobuffer;               /* fwrite's I/O buffer */
+  JSAMPROW pixrow;              /* decompressor output buffer */
+  size_t buffer_width;          /* width of I/O buffer */
+  JDIMENSION samples_per_row;   /* JSAMPLEs per output row */
+} ppm_dest_struct;
+
+typedef ppm_dest_struct *ppm_dest_ptr;
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ *
+ * put_pixel_rows handles the "normal" 8-bit case where the decompressor
+ * output buffer is physically the same as the fwrite buffer.
+ */
+
+METHODDEF(void)
+put_pixel_rows(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+               JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr)dinfo;
+
+  (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * This code is used when we have to copy the data and apply a pixel
+ * format translation.  Typically this only happens in 12-bit mode.
+ */
+
+METHODDEF(void)
+copy_pixel_rows(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr)dinfo;
+  register char *bufferptr;
+  register JSAMPROW ptr;
+#if BITS_IN_JSAMPLE != 8
+  register JDIMENSION col;
+#endif
+
+  ptr = dest->pub.buffer0;
+  bufferptr = dest->iobuffer;
+#if BITS_IN_JSAMPLE == 8
+  MEMCOPY(bufferptr, ptr, dest->samples_per_row);
+#else
+  for (col = dest->samples_per_row; col > 0; col--) {
+    PUTPPMSAMPLE(bufferptr, *ptr++);
+  }
+#endif
+  (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Convert extended RGB to RGB.
+ */
+
+METHODDEF(void)
+put_rgb(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo, JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr)dinfo;
+  register char *bufferptr;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  register int rindex = rgb_redcinfo->out_color_space;
+  register int gindex = rgb_greencinfo->out_color_space;
+  register int bindex = rgb_bluecinfo->out_color_space;
+  register int ps = rgb_pixelsizecinfo->out_color_space;
+
+  ptr = dest->pub.buffer0;
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    PUTPPMSAMPLE(bufferptr, ptrrindex);
+    PUTPPMSAMPLE(bufferptr, ptrgindex);
+    PUTPPMSAMPLE(bufferptr, ptrbindex);
+    ptr += ps;
+  }
+  (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Convert CMYK to RGB.
+ */
+
+METHODDEF(void)
+put_cmyk(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+         JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr)dinfo;
+  register char *bufferptr;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer0;
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    JSAMPLE r, g, b, c = *ptr++, m = *ptr++, y = *ptr++, k = *ptr++;
+    cmyk_to_rgb(c, m, y, k, &r, &g, &b);
+    PUTPPMSAMPLE(bufferptr, r);
+    PUTPPMSAMPLE(bufferptr, g);
+    PUTPPMSAMPLE(bufferptr, b);
+  }
+  (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Write some pixel data when color quantization is in effect.
+ * We have to demap the color index values to straight data.
+ */
+
+METHODDEF(void)
+put_demapped_rgb(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                 JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr)dinfo;
+  register char *bufferptr;
+  register int pixval;
+  register JSAMPROW ptr;
+  register JSAMPROW color_map0 = cinfo->colormap0;
+  register JSAMPROW color_map1 = cinfo->colormap1;
+  register JSAMPROW color_map2 = cinfo->colormap2;
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer0;
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    pixval = *ptr++;
+    PUTPPMSAMPLE(bufferptr, color_map0pixval);
+    PUTPPMSAMPLE(bufferptr, color_map1pixval);
+    PUTPPMSAMPLE(bufferptr, color_map2pixval);
+  }
+  (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+METHODDEF(void)
+put_demapped_gray(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                  JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr)dinfo;
+  register char *bufferptr;
+  register JSAMPROW ptr;
+  register JSAMPROW color_map = cinfo->colormap0;
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer0;
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    PUTPPMSAMPLE(bufferptr, color_map*ptr++);
+  }
+  (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_ppm(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr)dinfo;
+
+  /* Emit file header */
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    /* emit header for raw PGM format */
+    fprintf(dest->pub.output_file, "P5\n%ld %ld\n%d\n",
+            (long)cinfo->output_width, (long)cinfo->output_height, PPM_MAXVAL);
+    break;
+  case JCS_RGB:
+  case JCS_EXT_RGB:
+  case JCS_EXT_RGBX:
+  case JCS_EXT_BGR:
+  case JCS_EXT_BGRX:
+  case JCS_EXT_XBGR:
+  case JCS_EXT_XRGB:
+  case JCS_EXT_RGBA:
+  case JCS_EXT_BGRA:
+  case JCS_EXT_ABGR:
+  case JCS_EXT_ARGB:
+  case JCS_CMYK:
+    if (!IsExtRGB(cinfo->out_color_space) && cinfo->quantize_colors)
+      ERREXIT(cinfo, JERR_PPM_COLORSPACE);
+    /* emit header for raw PPM format */
+    fprintf(dest->pub.output_file, "P6\n%ld %ld\n%d\n",
+            (long)cinfo->output_width, (long)cinfo->output_height, PPM_MAXVAL);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_PPM_COLORSPACE);
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_ppm(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* Make sure we wrote the output file OK */
+  fflush(dinfo->output_file);
+  if (ferror(dinfo->output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * Re-calculate buffer dimensions based on output dimensions.
+ */
+
+METHODDEF(void)
+calc_buffer_dimensions_ppm(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr)dinfo;
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE)
+    dest->samples_per_row = cinfo->output_width * cinfo->out_color_components;
+  else
+    dest->samples_per_row = cinfo->output_width * 3;
+  dest->buffer_width = dest->samples_per_row * (BYTESPERSAMPLE * sizeof(char));
+}
+
+
+/*
+ * The module selection routine for PPM format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_ppm(j_decompress_ptr cinfo)
+{
+  ppm_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (ppm_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                  sizeof(ppm_dest_struct));
+  dest->pub.start_output = start_output_ppm;
+  dest->pub.finish_output = finish_output_ppm;
+  dest->pub.calc_buffer_dimensions = calc_buffer_dimensions_ppm;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Create physical I/O buffer */
+  dest->pub.calc_buffer_dimensions(cinfo, (djpeg_dest_ptr)dest);
+  dest->iobuffer = (char *)(*cinfo->mem->alloc_small)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE, dest->buffer_width);
+
+  if (cinfo->quantize_colors || BITS_IN_JSAMPLE != 8 ||
+      sizeof(JSAMPLE) != sizeof(char) ||
+#if RGB_RED == 0 && RGB_GREEN == 1 && RGB_BLUE == 2 && RGB_PIXELSIZE == 3
+      (cinfo->out_color_space != JCS_EXT_RGB &&
+       cinfo->out_color_space != JCS_RGB)) {
+#else
+      cinfo->out_color_space != JCS_EXT_RGB) {
+#endif
+    /* When quantizing, we need an output buffer for colormap indexes
+     * that's separate from the physical I/O buffer.  We also need a
+     * separate buffer if pixel format translation must take place.
+     */
+    dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr)cinfo, JPOOL_IMAGE,
+       cinfo->output_width * cinfo->output_components, (JDIMENSION)1);
+    dest->pub.buffer_height = 1;
+    if (!cinfo->quantize_colors) {
+      if (IsExtRGB(cinfo->out_color_space))
+        dest->pub.put_pixel_rows = put_rgb;
+      else if (cinfo->out_color_space == JCS_CMYK)
+        dest->pub.put_pixel_rows = put_cmyk;
+      else
+        dest->pub.put_pixel_rows = copy_pixel_rows;
+    } else if (cinfo->out_color_space == JCS_GRAYSCALE)
+      dest->pub.put_pixel_rows = put_demapped_gray;
+    else
+      dest->pub.put_pixel_rows = put_demapped_rgb;
+  } else {
+    /* We will fwrite() directly from decompressor output buffer. */
+    /* Synthesize a JSAMPARRAY pointer structure */
+    dest->pixrow = (JSAMPROW)dest->iobuffer;
+    dest->pub.buffer = &dest->pixrow;
+    dest->pub.buffer_height = 1;
+    dest->pub.put_pixel_rows = put_pixel_rows;
+  }
+
+  return (djpeg_dest_ptr)dest;
+}
+
+#endif /* PPM_SUPPORTED */

_service:tar_scm:libjpeg-turbo-2.1.1.tar.gz/wrtarga.c Added

@@ -0,0 +1,258 @@
+/*
+ * wrtarga.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2017, 2019, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in Targa format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * Based on code contributed by Lee Daniel Crocker.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef TARGA_SUPPORTED
+
+
+/*
+ * To support 12-bit JPEG data, we'd have to scale output down to 8 bits.
+ * This is not yet implemented.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  char *iobuffer;               /* physical I/O buffer */
+  JDIMENSION buffer_width;      /* width of one row */
+} tga_dest_struct;
+
+typedef tga_dest_struct *tga_dest_ptr;
+
+
+LOCAL(void)
+write_header(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo, int num_colors)
+/* Create and write a Targa header */
+{
+  char targaheader18;
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(targaheader, sizeof(targaheader));
+
+  if (num_colors > 0) {
+    targaheader1 = 1;         /* color map type 1 */
+    targaheader5 = (char)(num_colors & 0xFF);
+    targaheader6 = (char)(num_colors >> 8);
+    targaheader7 = 24;        /* 24 bits per cmap entry */
+  }
+
+  targaheader12 = (char)(cinfo->output_width & 0xFF);
+  targaheader13 = (char)(cinfo->output_width >> 8);
+  targaheader14 = (char)(cinfo->output_height & 0xFF);
+  targaheader15 = (char)(cinfo->output_height >> 8);
+  targaheader17 = 0x20;       /* Top-down, non-interlaced */
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    targaheader2 = 3;         /* image type = uncompressed grayscale */
+    targaheader16 = 8;        /* bits per pixel */
+  } else {                      /* must be RGB */
+    if (num_colors > 0) {
+      targaheader2 = 1;       /* image type = colormapped RGB */
+      targaheader16 = 8;
+    } else {
+      targaheader2 = 2;       /* image type = uncompressed RGB */
+      targaheader16 = 24;
+    }
+  }
+
+  if (JFWRITE(dinfo->output_file, targaheader, 18) != (size_t)18)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+               JDIMENSION rows_supplied)
+/* used for unquantized full-color output */
+{
+  tga_dest_ptr dest = (tga_dest_ptr)dinfo;
+  register JSAMPROW inptr;
+  register char *outptr;
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer0;
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    outptr0 = inptr2; /* RGB to BGR order */
+    outptr1 = inptr1;
+    outptr2 = inptr0;
+    inptr += 3, outptr += 3;
+  }
+  (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+METHODDEF(void)
+put_gray_rows(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+              JDIMENSION rows_supplied)
+/* used for grayscale OR quantized color output */
+{
+  tga_dest_ptr dest = (tga_dest_ptr)dinfo;
+  register JSAMPROW inptr;
+  register char *outptr;
+
+  inptr = dest->pub.buffer0;
+  outptr = dest->iobuffer;
+  MEMCOPY(outptr, inptr, cinfo->output_width);
+  (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Write some demapped pixel data when color quantization is in effect.
+ * For Targa, this is only applied to grayscale data.
+ */
+
+METHODDEF(void)
+put_demapped_gray(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                  JDIMENSION rows_supplied)
+{
+  tga_dest_ptr dest = (tga_dest_ptr)dinfo;
+  register JSAMPROW inptr;
+  register char *outptr;
+  register JSAMPROW color_map0 = cinfo->colormap0;
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer0;
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = color_map0*inptr++;
+  }
+  (void)JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_tga(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  tga_dest_ptr dest = (tga_dest_ptr)dinfo;
+  int num_colors, i;
+  FILE *outfile;
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    /* Targa doesn't have a mapped grayscale format, so we will */
+    /* demap quantized gray output.  Never emit a colormap. */
+    write_header(cinfo, dinfo, 0);
+    if (cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = put_demapped_gray;
+    else
+      dest->pub.put_pixel_rows = put_gray_rows;
+  } else if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* We only support 8-bit colormap indexes, so only 256 colors */
+      num_colors = cinfo->actual_number_of_colors;
+      if (num_colors > 256)
+        ERREXIT1(cinfo, JERR_TOO_MANY_COLORS, num_colors);
+      write_header(cinfo, dinfo, num_colors);
+      /* Write the colormap.  Note Targa uses BGR byte order */
+      outfile = dest->pub.output_file;
+      for (i = 0; i < num_colors; i++) {
+        putc(cinfo->colormap2i, outfile);
+        putc(cinfo->colormap1i, outfile);
+        putc(cinfo->colormap0i, outfile);
+      }
+      dest->pub.put_pixel_rows = put_gray_rows;
+    } else {
+      write_header(cinfo, dinfo, 0);
+      dest->pub.put_pixel_rows = put_pixel_rows;
+    }
+  } else {
+    ERREXIT(cinfo, JERR_TGA_COLORSPACE);
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_tga(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* Make sure we wrote the output file OK */
+  fflush(dinfo->output_file);
+  if (ferror(dinfo->output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * Re-calculate buffer dimensions based on output dimensions.
+ */
+
+METHODDEF(void)
+calc_buffer_dimensions_tga(j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  tga_dest_ptr dest = (tga_dest_ptr)dinfo;
+
+  dest->buffer_width = cinfo->output_width * cinfo->output_components;
+}
+
+
+/*
+ * The module selection routine for Targa format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_targa(j_decompress_ptr cinfo)
+{
+  tga_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (tga_dest_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                sizeof(tga_dest_struct));
+  dest->pub.start_output = start_output_tga;
+  dest->pub.finish_output = finish_output_tga;
+  dest->pub.calc_buffer_dimensions = calc_buffer_dimensions_tga;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Create I/O buffer. */
+  dest->pub.calc_buffer_dimensions(cinfo, (djpeg_dest_ptr)dest);
+  dest->iobuffer = (char *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
+                                (size_t)(dest->buffer_width * sizeof(char)));
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr)cinfo, JPOOL_IMAGE, dest->buffer_width, (JDIMENSION)1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr)dest;
+}
+
+#endif /* TARGA_SUPPORTED */

_service:tar_scm:libjpeg-turbo.yaml Added