CODEWITHSUNDEEP
c++visual-studio-2013cmakecuda
Eric
Eric

Reputation: 64

CMake: How to pass mode dependent compile flags to nvcc in visual studio environment

I am recently using CMake to compile my CUDA codes, but don't know how to pass different compile flags to nvcc, in Debug or Release mode. I wish I can specify something as follows,

set(CUDA_GENE_FLAGS_DEBUG "-gencode arch=compute_35,code=sm_35 -gencode arch=compute_37,code=sm_37")
set(CUDA_GENE_FLAGS_RELEASE "-Xptxas -O3 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_37,code=sm_37")

So when I specify Debug or Release mode in visual studio, these flags can be chosen automatically. Can anyone give me a hand on this? Thanks.

Upvotes: 4

Views: 4334

Answers (1)

Christophe
Christophe

Reputation: 305

I encountered a same problem while compiling linking OpenMP for a CUDA program.

According to the latest CMake document for FindCUDA here, you should try these three variables: CUDA_NVCC_FLAGS, CUDA_NVCC_FLAGS_DEBUG and CUDA_NVCC_FLAGS_RELEASE.

But it didn't work for me and then I found another reference.

If none of above variables work, try to use CMAKE_CUDA_FLAGS like this for a workaround:

set(CMAKE_CUDA_FLAGS ${YOUR_FLAGS})

Upvotes: 6

Related Questions