OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

[ROCm] Use generated CK config.h rather than system (#147993)

Browse Source 
prevents pytorch from potentially using system version of config.h and instead prioritize the CK submodule's version

Pull Request resolved: https://github.com/pytorch/pytorch/pull/147993
Approved by: https://github.com/jeffdaily

Co-authored-by: Jeff Daily <jeff.daily@amd.com>
This commit is contained in:
Andy Lugo 2025-03-13 20:04:12 +00:00 committed by PyTorch MergeBot
parent 4f8391db55
commit 38e81a5332
1 changed files with 23 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
aten/src/ATen/CMakeLists.txt
View File

@ -343,9 +343,32 @@ if(USE_CUDA)
endif()
if(USE_ROCM)
# NOTE: The PyTorch build does not actually add_subdirectory
# third_party/composable_kernel or use it as a CMake library. What is used
# is header only, so this should be ok, except that the CMake build generates
# a ck/config.h. We just do that part here. Without this, the ck.h from the
# ROCM SDK may get accidentally used instead.
function(_pytorch_rocm_generate_ck_conf)
set(CK_ENABLE_INT8 "ON")
set(CK_ENABLE_FP16 "ON")
set(CK_ENABLE_FP32 "ON")
set(CK_ENABLE_FP64 "ON")
set(CK_ENABLE_BF16 "ON")
set(CK_ENABLE_FP8 "ON")
set(CK_ENABLE_BF8 "ON")
set(CK_USE_XDL "ON")
set(CK_USE_WMMA "ON")
configure_file(
"${Torch_SOURCE_DIR}/third_party/composable_kernel/include/ck/config.h.in"
"${CMAKE_CURRENT_BINARY_DIR}/composable_kernel/ck/config.h"
)
endfunction()
list(APPEND ATen_HIP_INCLUDE ${CMAKE_CURRENT_SOURCE_DIR}/hip)
list(APPEND ATen_HIP_INCLUDE ${CMAKE_CURRENT_SOURCE_DIR}/../../../third_party/composable_kernel/include)
list(APPEND ATen_HIP_INCLUDE ${CMAKE_CURRENT_SOURCE_DIR}/../../../third_party/composable_kernel/library/include)
list(APPEND ATen_HIP_INCLUDE ${CMAKE_CURRENT_BINARY_DIR}/composable_kernel)
_pytorch_rocm_generate_ck_conf()
# Next two lines are needed because TunableOp uses third-party/fmt
list(APPEND ATen_HIP_INCLUDE $<TARGET_PROPERTY:fmt::fmt-header-only,INTERFACE_INCLUDE_DIRECTORIES>)
list(APPEND ATen_HIP_DEPENDENCY_LIBS fmt::fmt-header-only)