GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-11 11:19:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add -arch=<arch> argument for nvcc.

Browse Source 
Without this flag, when compiling with nvcc, if the compute architecture of a card does
not exactly match any of those listed for `-gencode arch=compute_<arch>,code=sm_<arch>`,
then the kernel will fail to run with:
```
cudaErrorNoKernelImageForDevice: no kernel image is available for execution on the device.
```
This can happen, for example, when compiling with an older cuda version
that does not support a newer architecture (e.g. T4 is `sm_75`, but cuda
9.2 only supports up to `sm_70`).

With the `-arch=<arch>` flag, the code will compile and run at the
supplied architecture.
This commit is contained in:
Antonio Sanchez 2021-09-24 20:48:01 -07:00
parent 846d34384a
commit de218b471d
2 changed files with 20 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
test/CMakeLists.txt
View File

@ -403,6 +403,16 @@ if(CUDA_FOUND AND EIGEN_TEST_CUDA)
else() else()
set(CUDA_PROPAGATE_HOST_FLAGS OFF) set(CUDA_PROPAGATE_HOST_FLAGS OFF)
set(NVCC_ARCH_FLAGS) set(NVCC_ARCH_FLAGS)
# Define an -arch=sm_<arch>, otherwise if GPU does not exactly match one of
# those in the arch list for -gencode, the kernels will fail to run with
# cudaErrorNoKernelImageForDevice
# This can happen with newer cards (e.g. sm_75) and compiling with older
# versions of nvcc (e.g. 9.2) that do not support their specific arch.
list(LENGTH EIGEN_CUDA_COMPUTE_ARCH EIGEN_CUDA_COMPUTE_ARCH_SIZE)
if(EIGEN_CUDA_COMPUTE_ARCH_SIZE)
list(GET EIGEN_CUDA_COMPUTE_ARCH 0 EIGEN_CUDA_COMPUTE_DEFAULT)
set(NVCC_ARCH_FLAGS " -arch=sm_${EIGEN_CUDA_COMPUTE_DEFAULT}")
endif()
foreach(ARCH IN LISTS EIGEN_CUDA_COMPUTE_ARCH) foreach(ARCH IN LISTS EIGEN_CUDA_COMPUTE_ARCH)
string(APPEND NVCC_ARCH_FLAGS " -gencode arch=compute_${ARCH},code=sm_${ARCH}") string(APPEND NVCC_ARCH_FLAGS " -gencode arch=compute_${ARCH},code=sm_${ARCH}")
endforeach() endforeach()

10
unsupported/test/CMakeLists.txt
View File

@ -303,6 +303,16 @@ if(CUDA_FOUND AND EIGEN_TEST_CUDA)
else() else()
set(CUDA_PROPAGATE_HOST_FLAGS OFF) set(CUDA_PROPAGATE_HOST_FLAGS OFF)
set(NVCC_ARCH_FLAGS) set(NVCC_ARCH_FLAGS)
# Define an -arch=sm_<arch>, otherwise if GPU does not exactly match one of
# those in the arch list for -gencode, the kernels will fail to run with
# cudaErrorNoKernelImageForDevice
# This can happen with newer cards (e.g. sm_75) and compiling with older
# versions of nvcc (e.g. 9.2) that do not support their specific arch.
list(LENGTH EIGEN_CUDA_COMPUTE_ARCH EIGEN_CUDA_COMPUTE_ARCH_SIZE)
if(EIGEN_CUDA_COMPUTE_ARCH_SIZE)
list(GET EIGEN_CUDA_COMPUTE_ARCH 0 EIGEN_CUDA_COMPUTE_DEFAULT)
set(NVCC_ARCH_FLAGS " -arch=sm_${EIGEN_CUDA_COMPUTE_DEFAULT}")
endif()
foreach(ARCH IN LISTS EIGEN_CUDA_COMPUTE_ARCH) foreach(ARCH IN LISTS EIGEN_CUDA_COMPUTE_ARCH)
string(APPEND NVCC_ARCH_FLAGS " -gencode arch=compute_${ARCH},code=sm_${ARCH}") string(APPEND NVCC_ARCH_FLAGS " -gencode arch=compute_${ARCH},code=sm_${ARCH}")
endforeach() endforeach()