SLIC_CUDA
SLIC_CUDA copied to clipboard
Set CUDA_COMPUTE_CAPABILITY
Maybe it is not a good idea to set CUDA_COMPUTE_CAPABILITY manually, which may lead to some unexpected errors.
I change the code in CMake from
### CUDA CHECK
find_package(CUDA REQUIRED)
set(CUDA_COMPUTE_CAPABILITY "52")
set(CUDA_GENERATE_CODE "arch=compute_${CUDA_COMPUTE_CAPABILITY},code=sm_${CUDA_COMPUTE_CAPABILITY}" CACHE STRING "Which GPU architectures to generate code for (each arch/code pair will be passed as --generate-code option to nvcc, separate multiple pairs by ;)")
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -arch compute_${CUDA_COMPUTE_CAPABILITY})
to
find_package(CUDA REQUIRED)
I tested this code and works.