开源软件构建与测试

Changes of Revision 11

_service:tar_scm:rocmlir.spec Added

@@ -0,0 +1,176 @@
+%global upstreamname rocMLIR
+%define debug_package %{nil}
+
+%global rocm_release 6.2
+%global rocm_patch 4
+%global rocm_version %{rocm_release}.%{rocm_patch}
+
+%global toolchain rocm
+# hipcc does not support some clang flags
+%global build_cxxflags %(echo %{optflags} | sed -e 's/-fstack-protector-strong/-Xarch_host -fstack-protector-strong/' -e 's/-fcf-protection/-Xarch_host -fcf-protection/')
+
+# $gpu will be evaluated in the loops below
+%global _vpath_builddir %{_vendor}-%{_target_os}-build-${gpu}
+
+%bcond_with debug
+%if %{with debug}
+%global build_type DEBUG
+%else
+%global build_type RelWithDebInfo
+%endif
+
+# It is necessary to use this with a local build
+# export QA_RPATHS=0xff
+%bcond_with test
+%if %{with test}
+%global build_test ON
+%else
+%global build_test OFF
+%endif
+
+# Option to test suite for testing on real HW:
+%bcond_with check
+
+Name:           rocmlir
+Version:        %{rocm_version}
+Release:        1
+Summary:        rocMLIR is for a MLIR-based convolution and GEMM kernel generator targetting AMD hardware.
+
+Url:            https://github.com/ROCm/rocMLIR
+License:        MIT and BSD
+Source0:        %{url}/archive/rocm-%{version}.tar.gz#/%{upstreamname}-%{version}.tar.gz
+
+Patch0:         0001-Fix-build-on-riscv64.patch
+
+BuildRequires:  git
+BuildRequires:  cmake
+BuildRequires:  rocm-cmake
+BuildRequires:  rocm-comgr-devel
+BuildRequires:  rocm-hip-devel
+BuildRequires:  rocm-runtime-devel
+BuildRequires:  rocm-rpm-macros
+BuildRequires:  rocm-rpm-macros-modules
+BuildRequires:  rocblas-devel
+BuildRequires:  rocminfo
+BuildRequires:  ninja-build
+
+#Use compat if LLVM is not latest, 1 is enable
+%global compat_build 1
+%global maj_ver 18
+
+%if %{compat_build}
+BuildRequires:  llvm%{maj_ver}-devel
+BuildRequires:  llvm%{maj_ver}-test
+BuildRequires:  llvm%{maj_ver}-cmake-utils
+BuildRequires:  llvm%{maj_ver}-libs
+BuildRequires:  lld%{maj_ver}
+BuildRequires:  lld%{maj_ver}-devel
+BuildRequires:  clang%{maj_ver}
+BuildRequires:  clang%{maj_ver}-devel
+BuildRequires:  clang%{maj_ver}-libs
+BuildRequires:  clang%{maj_ver}-tools-extra
+%else
+BuildRequires:  llvm-devel = %{version}
+BuildRequires:  llvm-test = %{version}
+BuildRequires:  llvm-cmake-utils = %{version}
+%endif
+
+%if %{with test}
+BuildRequires:  gtest-devel
+%endif
+
+Requires:       rocm-rpm-macros-modules
+
+# Only x86_64 works right now:
+ExclusiveArch:  x86_64 riscv64
+
+%description
+rocMLIR is for a MLIR-based convolution and GEMM kernel generator
+targetting AMD hardware. This generator is mainly used from MIGraphX,
+but it can be used on a standalone basis. (The ability to use this code
+via torch-mlir is being investigated as well.)
+
+%package devel
+Summary:        The rocRAND development package
+Requires:       %{name}%{?_isa} = %{version}-%{release}
+
+%description devel
+The rocRAND development package.
+
+%if %{with test}
+%package test
+Summary:        Tests for %{name}
+Requires:       %{name}%{?_isa} = %{version}-%{release}
+
+%description test
+%{summary}
+%endif
+
+%prep
+%autosetup -p1 -n %{upstreamname}-rocm-%{version}
+
+%build
+for gpu in %{rocm_gpu_list}
+do
+    module load rocm/$gpu
+    mkdir %{_target_platform}-$gpu
+    pushd %{_target_platform}-$gpu
+    # We can't compile all the targets at once
+    export ROCM_GPUS='gfx1030;gfx1100'
+    %cmake -G Ninja .. \
+           -DCMAKE_BUILD_TYPE=%build_type \
+	   -DAMDGPU_TARGETS=${ROCM_GPUS} \
+	   -DCMAKE_INSTALL_LIBDIR=$ROCM_LIB \
+	   -DCMAKE_INSTALL_BINDIR=$ROCM_BIN \
+	   -DCMAKE_C_FLAGS="-Wno-gnu-line-marker" \
+       -DBUILD_FAT_LIBROCKCOMPILER=ON \
+       -DCMAKE_CXX_FLAGS="-Wno-gnu-line-marker -Wno-narrowing" \
+       -DBUILD_SHARED_LIBS:BOOL=OFF \
+       -DBUILD_TEST=%build_test
+    popd
+    ninja -C %{_target_platform}-$gpu %{?_smp_mflags} -v
+    module purge
+done
+
+%install
+for gpu in %{rocm_gpu_list}
+do
+    DESTDIR=%{buildroot} ninja -C %{_target_platform}-$gpu install
+done
+
+echo s@%{buildroot}@@ > br.sed
+find %{buildroot}/usr/lib -name '*.a' | sed -f br.sed >  %{name}.files
+find %{buildroot}/usr/lib -name '*.o'   | sed -f br.sed >> %{name}.files
+find %{buildroot}%{_libdir} -name '*.cmake'      | sed -f br.sed > %{name}.devel
+%if %{with test}
+find %{buildroot}           -name 'test_*'       | sed -f br.sed >  %{name}.test
+find %{buildroot}           -name '*RAND'        | sed -f br.sed >> %{name}.test
+%endif
+
+%check
+%if %{with test}
+%if %{with check}
+# Assumes default
+%ctest
+%endif
+%endif
+
+%files -f %{name}.files
+%license %{_docdir}/%{name}/LICENSE
+
+%files devel -f %{name}.devel
+%{_includedir}/%{name}
+
+%if %{with test}
+%files test -f %{name}.test
+%endif
+
+%changelog
+* Mon Dec 16 2024 QiuWenjian <wenjian.oerv@isrc.iscas.ac.cn> - 6.2.4-1
+- Upgraded to 6.2.4
+
+* Thu Dec 11 2024 QiuWenjian <wenjian.oerv@isrc.iscas.ac.cn> - 6.2.2-2
+- Fix compile error
+
+* Tue Nov 26 2024 misaka00251 <liuxin@iscas.ac.cn> - 6.2.2-1
+- Init package

_service:tar_scm:0001-Fix-build-on-riscv64.patch Added

@@ -0,0 +1,104 @@
+From 6bdfbddb00d5550902970c05c512c46c2fe5862e Mon Sep 17 00:00:00 2001
+From: misaka00251 <liuxin@iscas.ac.cn>
+Date: Tue, 26 Nov 2024 15:15:37 +0800
+Subject: PATCH Fix build on riscv64
+
+
+diff --git a/external/llvm-project/clang/lib/Basic/Targets/RISCV.h b/external/llvm-project/clang/lib/Basic/Targets/RISCV.h
+index 78580b5b1..720ee6989 100644
+--- a/external/llvm-project/clang/lib/Basic/Targets/RISCV.h
++++ b/external/llvm-project/clang/lib/Basic/Targets/RISCV.h
+@@ -49,6 +49,7 @@ public:
+     MCountName = "_mcount";
+     HasFloat16 = true;
+     HasStrictFP = true;
++    HasFloat128 = true;
+   }
+ 
+   bool setCPU(const std::string &Name) override {
+diff --git a/external/llvm-project/clang/lib/CodeGen/BackendUtil.cpp b/external/llvm-project/clang/lib/CodeGen/BackendUtil.cpp
+index 22c3f8642..a1d160411 100644
+--- a/external/llvm-project/clang/lib/CodeGen/BackendUtil.cpp
++++ b/external/llvm-project/clang/lib/CodeGen/BackendUtil.cpp
+@@ -302,6 +302,7 @@ getCodeModel(const CodeGenOptions &CodeGenOpts) {
+                            .Case("kernel", llvm::CodeModel::Kernel)
+                            .Case("medium", llvm::CodeModel::Medium)
+                            .Case("large", llvm::CodeModel::Large)
++                           .Case("extreme", llvm::CodeModel::Large)
+                            .Case("default", ~1u)
+                            .Default(~0u);
+   assert(CodeModel != ~0u && "invalid code model!");
+diff --git a/external/llvm-project/clang/lib/Driver/ToolChains/HIPUtility.cpp b/external/llvm-project/clang/lib/Driver/ToolChains/HIPUtility.cpp
+index 08c647dfc..42f91a4e1 100644
+--- a/external/llvm-project/clang/lib/Driver/ToolChains/HIPUtility.cpp
++++ b/external/llvm-project/clang/lib/Driver/ToolChains/HIPUtility.cpp
+@@ -36,7 +36,7 @@ using llvm::dyn_cast;
+ #endif
+ 
+ namespace {
+-const unsigned HIPCodeObjectAlign = 4096;
++const unsigned HIPCodeObjectAlign = 65536;
+ } // namespace
+ 
+ // Constructs a triple string for clang offload bundler.
+@@ -227,7 +227,12 @@ void HIP::constructHIPFatbinCommand(Compilation &C, const JobAction &JA,
+ 
+   // ToDo: Remove the dummy host binary entry which is required by
+   // clang-offload-bundler.
++#if defined(__x86_64)
+   std::string BundlerTargetArg = "-targets=host-x86_64-unknown-linux";
++#elif (defined(__riscv) && __riscv_xlen == 64)
++  std::string BundlerTargetArg = "-targets=host-riscv64-unknown-linux";
++#else
++#endif
+   // AMDGCN:
+   // For code object version 2 and 3, the offload kind in bundle ID is 'hip'
+   // for backward compatibility. For code object version 4 and greater, the
+diff --git a/external/llvm-project/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp b/external/llvm-project/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp
+index 1de1a30a7..718bc1d46 100644
+--- a/external/llvm-project/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp
++++ b/external/llvm-project/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp
+@@ -404,15 +404,23 @@ fatbinary(ArrayRef<std::pair<StringRef, StringRef>> InputFiles,
+   SmallVector<StringRef, 16> CmdArgs;
+   CmdArgs.push_back(*OffloadBundlerPath);
+   CmdArgs.push_back("-type=o");
++#if defined(__x86_64)
+   CmdArgs.push_back("-bundle-align=4096");
+-
++#elif (defined(__riscv) && __riscv_xlen == 64)
++  CmdArgs.push_back("-bundle-align=4096");
++#else
++#endif
+   if (Args.hasArg(OPT_compress))
+     CmdArgs.push_back("-compress");
+   if (auto *Arg = Args.getLastArg(OPT_compression_level_eq))
+     CmdArgs.push_back(
+         Args.MakeArgString(Twine("-compression-level=") + Arg->getValue()));
+-
++#if defined(__x86_64)
+   SmallVector<StringRef> Targets = {"-targets=host-x86_64-unknown-linux"};
++#elif (defined(__riscv) && __riscv_xlen == 64)
++  SmallVector<StringRef> Targets = {"-targets=host-riscv64-unknown-linux"};
++#else
++#endif
+   for (const auto &File, Arch : InputFiles)
+     Targets.push_back(Saver.save("hipv4-amdgcn-amd-amdhsa--" + Arch));
+   CmdArgs.push_back(Saver.save(llvm::join(Targets, ",")));
+diff --git a/mlir/tools/xmir-runner/CMakeLists.txt b/mlir/tools/xmir-runner/CMakeLists.txt
+index 81f484a51..bc88a0a7f 100644
+--- a/mlir/tools/xmir-runner/CMakeLists.txt
++++ b/mlir/tools/xmir-runner/CMakeLists.txt
+@@ -14,6 +14,10 @@ if(MLIR_ENABLE_ROCM_RUNNER)
+     ${rocmlir_conversion_libs}
+     ${rocmlir_test_libs}
+     LLVMX86AsmParser
++    LLVMX86Desc
++    LLVMX86CodeGen
++    LLVMX86Info
++    LLVMRISCVArchAsmParser
+     MLIRMHAL
+     MLIRMHALTransforms
+     MLIRMHALPipeline
+-- 
+2.47.0
+

_service Added

_service:tar_scm:README.md Added

_service:tar_scm:rocMLIR-6.2.4.tar.gz Added