dnrops
/
pytorch
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[CUDA][complex] Define `LIBCUDACXX_ENABLE_SIMPLIFIED_COMPLEX_OPERATIONS` in CUDA build (#117061) 

An upcoming CUDA release will migrate to CCCL, and we need this define to preserve current complex behavior: https://nvidia.github.io/libcudacxx/standard_api/numerics_library/complex.html

CC @miscco @ptrblck

Pull Request resolved: https://github.com/pytorch/pytorch/pull/117061
Approved by: https://github.com/ezyang, https://github.com/malfet
This commit is contained in:
eqy 2024-01-30 06:11:26 +00:00 committed by PyTorch MergeBot
parent c203d88795
commit 4a48899b6e
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
CMakeLists.txt
View File

@ -48,6 +48,10 @@ if(CMAKE_COMPILER_IS_GNUCXX AND CMAKE_CXX_COMPILER_VERSION VERSION_LESS 9.4)
message(FATAL "GCC-9.4 or newer is required to compile PyTorch, but found ${CMAKE_CXX_COMPILER_VERSION}")
endif()
# This define is needed to preserve behavior given anticpated changes to cccl/thrust
# https://nvidia.github.io/libcudacxx/standard_api/numerics_library/complex.html
string(APPEND CMAKE_CUDA_FLAGS "-DLIBCUDACXX_ENABLE_SIMPLIFIED_COMPLEX_OPERATIONS")
if(LINUX)
include(cmake/CheckAbi.cmake)
string(APPEND CMAKE_CXX_FLAGS " -D_GLIBCXX_USE_CXX11_ABI=${GLIBCXX_USE_CXX11_ABI}")