jhuber6 created this revision.
jhuber6 added reviewers: jdoerfert, tra, yaxunl, JonChesterfield, MaskRay,
mgorny, tstellar.
Herald added subscribers: mattd, gchakrabarti, asavonic, guansong.
Herald added a project: All.
jhuber6 requested review of this revision.
Herald added subscribers: llvm-commits, cfe-commits, sstefan1.
Herald added projects: clang, LLVM.
The `CLANG_OPENMP_NVPTX_DEFAULT_ARCH` is a static build configuration to
set the default OpenMP value. This was replaced in D141708
<https://reviews.llvm.org/D141708> with the use
of the `nvptx-arch` tool which lets us query this at runtime instead.
This makes the behaviour between AMD and NVIDIA be more consisten and
allows users to have a default architecture that does not rely on
whoever configured the LLVM build.
Depends on D141708 <https://reviews.llvm.org/D141708>
Repository:
rG LLVM Github Monorepo
https://reviews.llvm.org/D141723
Files:
clang/CMakeLists.txt
clang/include/clang/Config/config.h.cmake
llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
utils/bazel/llvm-project-overlay/clang/include/clang/Config/config.h
Index: utils/bazel/llvm-project-overlay/clang/include/clang/Config/config.h
===================================================================
--- utils/bazel/llvm-project-overlay/clang/include/clang/Config/config.h
+++ utils/bazel/llvm-project-overlay/clang/include/clang/Config/config.h
@@ -43,9 +43,6 @@
/* Default OpenMP runtime used by -fopenmp. */
#define CLANG_DEFAULT_OPENMP_RUNTIME "libomp"
-/* Default architecture for OpenMP offloading to Nvidia GPUs. */
-#define CLANG_OPENMP_NVPTX_DEFAULT_ARCH "sm_35"
-
/* Default architecture for SystemZ. */
#define CLANG_SYSTEMZ_DEFAULT_ARCH "z10"
Index: llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
===================================================================
--- llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
+++ llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
@@ -17,7 +17,6 @@
"CLANG_DEFAULT_OPENMP_RUNTIME=libomp",
"CLANG_DEFAULT_UNWINDLIB=",
"CLANG_INSTALL_LIBDIR_BASENAME=lib",
- "CLANG_OPENMP_NVPTX_DEFAULT_ARCH=sm_35",
"CLANG_RESOURCE_DIR=",
"C_INCLUDE_DIRS=",
"CLANG_CONFIG_FILE_SYSTEM_DIR=",
Index: clang/include/clang/Config/config.h.cmake
===================================================================
--- clang/include/clang/Config/config.h.cmake
+++ clang/include/clang/Config/config.h.cmake
@@ -29,9 +29,6 @@
/* Default OpenMP runtime used by -fopenmp. */
#define CLANG_DEFAULT_OPENMP_RUNTIME "${CLANG_DEFAULT_OPENMP_RUNTIME}"
-/* Default architecture for OpenMP offloading to Nvidia GPUs. */
-#define CLANG_OPENMP_NVPTX_DEFAULT_ARCH "${CLANG_OPENMP_NVPTX_DEFAULT_ARCH}"
-
/* Default architecture for SystemZ. */
#define CLANG_SYSTEMZ_DEFAULT_ARCH "${CLANG_SYSTEMZ_DEFAULT_ARCH}"
Index: clang/CMakeLists.txt
===================================================================
--- clang/CMakeLists.txt
+++ clang/CMakeLists.txt
@@ -245,31 +245,6 @@
set(CLANG_DEFAULT_OPENMP_RUNTIME "libomp" CACHE STRING
"Default OpenMP runtime used by -fopenmp.")
-# OpenMP offloading requires at least sm_35 because we use shuffle instructions
-# to generate efficient code for reductions and the atomicMax instruction on
-# 64-bit integers in the implementation of conditional lastprivate.
-set(CUDA_ARCH_FLAGS "sm_35")
-
-# Try to find the highest Nvidia GPU architecture the system supports
-if (NOT DEFINED CLANG_OPENMP_NVPTX_DEFAULT_ARCH)
- find_package(CUDA QUIET)
- if (CUDA_FOUND)
- cuda_select_nvcc_arch_flags(CUDA_ARCH_FLAGS)
- endif()
-else()
- set(CUDA_ARCH_FLAGS ${CLANG_OPENMP_NVPTX_DEFAULT_ARCH})
-endif()
-
-string(REGEX MATCH "sm_([0-9]+)" CUDA_ARCH_MATCH ${CUDA_ARCH_FLAGS})
-if (NOT DEFINED CUDA_ARCH_MATCH OR "${CMAKE_MATCH_1}" LESS 35)
- set(CLANG_OPENMP_NVPTX_DEFAULT_ARCH "sm_35" CACHE STRING
- "Default architecture for OpenMP offloading to Nvidia GPUs." FORCE)
- message(WARNING "Resetting default architecture for OpenMP offloading to
Nvidia GPUs to sm_35")
-else()
- set(CLANG_OPENMP_NVPTX_DEFAULT_ARCH ${CUDA_ARCH_MATCH} CACHE STRING
- "Default architecture for OpenMP offloading to Nvidia GPUs.")
-endif()
-
set(CLANG_SYSTEMZ_DEFAULT_ARCH "z10" CACHE STRING "SystemZ Default Arch")
set(CLANG_VENDOR ${PACKAGE_VENDOR} CACHE STRING
Index: utils/bazel/llvm-project-overlay/clang/include/clang/Config/config.h
===================================================================
--- utils/bazel/llvm-project-overlay/clang/include/clang/Config/config.h
+++ utils/bazel/llvm-project-overlay/clang/include/clang/Config/config.h
@@ -43,9 +43,6 @@
/* Default OpenMP runtime used by -fopenmp. */
#define CLANG_DEFAULT_OPENMP_RUNTIME "libomp"
-/* Default architecture for OpenMP offloading to Nvidia GPUs. */
-#define CLANG_OPENMP_NVPTX_DEFAULT_ARCH "sm_35"
-
/* Default architecture for SystemZ. */
#define CLANG_SYSTEMZ_DEFAULT_ARCH "z10"
Index: llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
===================================================================
--- llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
+++ llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
@@ -17,7 +17,6 @@
"CLANG_DEFAULT_OPENMP_RUNTIME=libomp",
"CLANG_DEFAULT_UNWINDLIB=",
"CLANG_INSTALL_LIBDIR_BASENAME=lib",
- "CLANG_OPENMP_NVPTX_DEFAULT_ARCH=sm_35",
"CLANG_RESOURCE_DIR=",
"C_INCLUDE_DIRS=",
"CLANG_CONFIG_FILE_SYSTEM_DIR=",
Index: clang/include/clang/Config/config.h.cmake
===================================================================
--- clang/include/clang/Config/config.h.cmake
+++ clang/include/clang/Config/config.h.cmake
@@ -29,9 +29,6 @@
/* Default OpenMP runtime used by -fopenmp. */
#define CLANG_DEFAULT_OPENMP_RUNTIME "${CLANG_DEFAULT_OPENMP_RUNTIME}"
-/* Default architecture for OpenMP offloading to Nvidia GPUs. */
-#define CLANG_OPENMP_NVPTX_DEFAULT_ARCH "${CLANG_OPENMP_NVPTX_DEFAULT_ARCH}"
-
/* Default architecture for SystemZ. */
#define CLANG_SYSTEMZ_DEFAULT_ARCH "${CLANG_SYSTEMZ_DEFAULT_ARCH}"
Index: clang/CMakeLists.txt
===================================================================
--- clang/CMakeLists.txt
+++ clang/CMakeLists.txt
@@ -245,31 +245,6 @@
set(CLANG_DEFAULT_OPENMP_RUNTIME "libomp" CACHE STRING
"Default OpenMP runtime used by -fopenmp.")
-# OpenMP offloading requires at least sm_35 because we use shuffle instructions
-# to generate efficient code for reductions and the atomicMax instruction on
-# 64-bit integers in the implementation of conditional lastprivate.
-set(CUDA_ARCH_FLAGS "sm_35")
-
-# Try to find the highest Nvidia GPU architecture the system supports
-if (NOT DEFINED CLANG_OPENMP_NVPTX_DEFAULT_ARCH)
- find_package(CUDA QUIET)
- if (CUDA_FOUND)
- cuda_select_nvcc_arch_flags(CUDA_ARCH_FLAGS)
- endif()
-else()
- set(CUDA_ARCH_FLAGS ${CLANG_OPENMP_NVPTX_DEFAULT_ARCH})
-endif()
-
-string(REGEX MATCH "sm_([0-9]+)" CUDA_ARCH_MATCH ${CUDA_ARCH_FLAGS})
-if (NOT DEFINED CUDA_ARCH_MATCH OR "${CMAKE_MATCH_1}" LESS 35)
- set(CLANG_OPENMP_NVPTX_DEFAULT_ARCH "sm_35" CACHE STRING
- "Default architecture for OpenMP offloading to Nvidia GPUs." FORCE)
- message(WARNING "Resetting default architecture for OpenMP offloading to Nvidia GPUs to sm_35")
-else()
- set(CLANG_OPENMP_NVPTX_DEFAULT_ARCH ${CUDA_ARCH_MATCH} CACHE STRING
- "Default architecture for OpenMP offloading to Nvidia GPUs.")
-endif()
-
set(CLANG_SYSTEMZ_DEFAULT_ARCH "z10" CACHE STRING "SystemZ Default Arch")
set(CLANG_VENDOR ${PACKAGE_VENDOR} CACHE STRING
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
