Configure clang via cmake options rather than via modifying sources

diff --git a/Makefile b/Makefile
index 6caf7c0190e0269e5c7c89233c23b8591b4d7884..4ea5cee91a19597a8a68e6717f9baed215aa5c31 100644 (file)
--- a/Makefile
+++ b/Makefile
@@ -62,7 +62,8 @@ build/build.ninja: | build llvm/tools/clang llvm/projects/openmp
        cd $(@D) && cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX="$(PREFIX)" \
                -DLLVM_ENABLE_ASSERTIONS=ON -DLLVM_ENABLE_BACKTRACES=ON -DLLVM_ENABLE_WERROR=OFF \
                -DBUILD_SHARED_LIBS=OFF -DLLVM_ENABLE_RTTI=ON $(LLVM_CROSS_FLAGS) \
-               -DCMAKE_CXX_FLAGS='-DHERCULES_PASS_ROOT=\"$(PREFIX)/lib/hercules/src/passes/\"' \
+               -DCMAKE_CXX_FLAGS='-DHERCULES_PASS_ROOT=\"$(PREFIX)/lib/hercules/src/passes/\" -DOPENMP_NVPTX_COMPUTE_CAPABILITY=62' \
+               -DLIBOMPTARGET_NVPTX_COMPUTE_CAPABILITY="62" \
                -G"Ninja" $(CURDIR)/llvm
 
 $(TMP_DESTDIR)/$(PREFIX)/lib/cmake/llvm: | llvm-clang

diff --git a/debian/patches/cuda_arch.patch b/debian/patches/cuda_arch.patch
deleted file mode 100644 (file)
index 842e368..0000000
--- a/debian/patches/cuda_arch.patch
+++ /dev/null
@@ -1,33 +0,0 @@
-Description: Configuration for TX2 according to D33CompilerRuntime.docx 
-
---- a/openmp/libomptarget/deviceRTLs/nvptx/CMakeLists.txt
-+++ b/openmp/libomptarget/deviceRTLs/nvptx/CMakeLists.txt
-@@ -66,7 +66,7 @@
-       set(CUDA_ARCH ${CUDA_ARCH} -gencode arch=compute_${sm},code=sm_${sm})
-     endforeach()
-   else()
--    set(CUDA_ARCH -arch sm_35)
-+    set(CUDA_ARCH -arch sm_62)
-   endif()
-   
-   # Activate RTL message dumps if requested by the user.
-@@ -176,7 +176,7 @@
-           set(CUDA_ARCH ${CUDA_ARCH} --cuda-gpu-arch=sm_${sm})
-         endforeach()
-       else()
--        set(CUDA_ARCH --cuda-gpu-arch=sm_35)
-+        set(CUDA_ARCH --cuda-gpu-arch=sm_62)
-       endif()
- 
-       # Compile cuda files to bitcode.
---- a/clang/lib/Driver/ToolChains.cpp
-+++ b/clang/lib/Driver/ToolChains.cpp
-@@ -4967,7 +4967,7 @@
- // macro for it. Also, select the default PTX version to be used. We use 4.2 for
- // compute capabilities older than 6.0 and 5.0 otherwise.
- #ifndef OPENMP_NVPTX_COMPUTE_CAPABILITY
--#define OPENMP_NVPTX_COMPUTE_CAPABILITY 53
-+#define OPENMP_NVPTX_COMPUTE_CAPABILITY 62
- #endif
- 
- #if OPENMP_NVPTX_COMPUTE_CAPABILITY < 60

diff --git a/debian/patches/series b/debian/patches/series
index eb6eecf0e55daaba8acc3d1d25cce07ddbbf9350..c9cc75908e6df668e49659f2463ecd6a58f2d6e8 100644 (file)
--- a/debian/patches/series
+++ b/debian/patches/series
@@ -3,4 +3,3 @@ Guard-architecture-dependent-code-with-#ifdefs.patch
 require-cuda.patch
 configure-herculescompiler-via-cmake.patch
 use_gpuguard.patch
-cuda_arch.patch