How do I force FindCuda to compile to compute 1.3?
There isn’t an option to directly support this feature. You must modify the arguments to NVCC. You can do this in one of two ways.
Modify CUDA_NVCC_FLAGS to include -arch sm_13.
Add OPTION -arch sm_13 to your CUDA_ADD_LIBRARY or CUDA_ADD_EXECUTABLE command. This will add these flags to the NVCC flags for this one target. See the documentation at the head of the file.