mirror of
https://github.com/Kitware/CMake.git
synced 2026-04-23 22:58:37 -05:00
Raul Tambre
c63fe01835
For NVCC the compiler takes care of device linking when passed the "-dlink" flag. Clang doesn't support such magic and requires the buildsystem to do the work that NVCC does behind the scenes. The implementation is based on Bazel's device linking documentation: https://github.com/tensorflow/tensorflow/blob/7cabcdf073abad8c46e9dda62bb8fa4682d2061e/third_party/nccl/build_defs.bzl.tpl#L259 Closes: #20726
35 lines
1.5 KiB
CMake
35 lines
1.5 KiB
CMake
macro (add_cuda_test_macro name)
|
|
add_test_macro("${name}" ${ARGN})
|
|
set_property(TEST "${name}" APPEND
|
|
PROPERTY LABELS "CUDA")
|
|
endmacro ()
|
|
|
|
add_cuda_test_macro(Cuda.ConsumeCompileFeatures CudaConsumeCompileFeatures)
|
|
add_cuda_test_macro(Cuda.CXXStandardSetTwice CXXStandardSetTwice)
|
|
add_cuda_test_macro(Cuda.ObjectLibrary CudaObjectLibrary)
|
|
add_cuda_test_macro(Cuda.MixedStandardLevels1 MixedStandardLevels1)
|
|
add_cuda_test_macro(Cuda.MixedStandardLevels2 MixedStandardLevels2)
|
|
add_cuda_test_macro(Cuda.MixedStandardLevels3 MixedStandardLevels3)
|
|
add_cuda_test_macro(Cuda.MixedStandardLevels4 MixedStandardLevels4)
|
|
add_cuda_test_macro(Cuda.MixedStandardLevels5 MixedStandardLevels5)
|
|
add_cuda_test_macro(Cuda.NotEnabled CudaNotEnabled)
|
|
add_cuda_test_macro(Cuda.SeparableCompCXXOnly SeparableCompCXXOnly)
|
|
add_cuda_test_macro(Cuda.Toolkit Toolkit)
|
|
add_cuda_test_macro(Cuda.IncludePathNoToolkit IncludePathNoToolkit)
|
|
add_cuda_test_macro(Cuda.SharedRuntimePlusToolkit SharedRuntimePlusToolkit)
|
|
add_cuda_test_macro(Cuda.Complex CudaComplex)
|
|
add_cuda_test_macro(Cuda.ProperLinkFlags ProperLinkFlags)
|
|
|
|
if(CMake_TEST_CUDA AND NOT CMake_TEST_CUDA STREQUAL "Clang")
|
|
# Clang lacks __CUDACC_VER*__ defines.
|
|
add_cuda_test_macro(Cuda.ProperDeviceLibraries ProperDeviceLibraries)
|
|
endif()
|
|
|
|
# The CUDA only ships the shared version of the toolkit libraries
|
|
# on windows
|
|
if(NOT WIN32)
|
|
add_cuda_test_macro(Cuda.StaticRuntimePlusToolkit StaticRuntimePlusToolkit)
|
|
endif()
|
|
|
|
add_cuda_test_macro(Cuda.WithC CudaWithC)
|