diff options
| author | Raul Tambre <raul@tambre.ee> | 2020-03-07 17:29:29 +0200 |
|---|---|---|
| committer | Raul Tambre <raul@tambre.ee> | 2020-05-15 17:46:51 +0300 |
| commit | 5df21adf46be14061fe0e8ea53213d905e834600 (patch) | |
| tree | 9e50ef0dc770d6ad68eede0ad04e5058d1775069 /Modules/CMakeTestCUDACompiler.cmake | |
| parent | dc2eae1f91f98019efe14c1398e10f23e93a7005 (diff) | |
| download | cmake-5df21adf46be14061fe0e8ea53213d905e834600.tar.gz | |
CUDA: Add support for Clang compiler
When crosscompiling we pass the sysroot.
We need to try various architecture flags. Clang doesn't automatically
select one that works. First try the ones that are more likely to work
for modern installations:
* <=sm_50 is deprecated since CUDA 10.2, try sm_52 first for
future compatibility.
* <=sm_20 is removed since CUDA 9.0, try sm_30.
Otherwise fallback to Clang's current default. Currently that's `sm_20`,
the lowest it supports.
Separable compilation isn't supported yet.
Fixes: #16586
Diffstat (limited to 'Modules/CMakeTestCUDACompiler.cmake')
| -rw-r--r-- | Modules/CMakeTestCUDACompiler.cmake | 9 |
1 files changed, 7 insertions, 2 deletions
diff --git a/Modules/CMakeTestCUDACompiler.cmake b/Modules/CMakeTestCUDACompiler.cmake index 5ad725702b..b3b62bdcb7 100644 --- a/Modules/CMakeTestCUDACompiler.cmake +++ b/Modules/CMakeTestCUDACompiler.cmake @@ -74,9 +74,14 @@ else() # - cudart_static # - cudadevrt # + # Additionally on Linux: + # - rt + # - pthread + # - dl + # # These are controlled by CMAKE_CUDA_RUNTIME_LIBRARY - list(REMOVE_ITEM CMAKE_CUDA_IMPLICIT_LINK_LIBRARIES cudart cudart_static cudadevrt) - list(REMOVE_ITEM CMAKE_CUDA_HOST_IMPLICIT_LINK_LIBRARIES cudart cudart_static cudadevrt) + list(REMOVE_ITEM CMAKE_CUDA_IMPLICIT_LINK_LIBRARIES cudart cudart_static cudadevrt rt pthread dl) + list(REMOVE_ITEM CMAKE_CUDA_HOST_IMPLICIT_LINK_LIBRARIES cudart cudart_static cudadevrt rt pthread dl) if(CMAKE_CUDA_COMPILER_ID STREQUAL "NVIDIA") # Remove the CUDA Toolkit include directories from the set of |