need to update CUDA compute capabilities in CMakeLists.txt

[djboersma]

Today I read on gate-users, mail by Bryan McIntosh:

I had a similar issue with my GTX 980, and the problem lies in the file CMakeLists.txt that is in your gate_v8.0 directory. Much like in GATE 7.2, you need to find the line CUDA_NVCC_FLAGS and search for "gencode arch=compute_XX,code=sm_XX" and replace the value represented by the XX characters (likely 21) with the compute capability for the Pascal architecture cards which is 61.

This sounds like something we should fix: apparently the list of supported compute capabilities is outdated. Do we have a benchmark to validate GPU-based simulations on new hardware or new "compute capabilities"?

[AndiH]

See also pull request #134 for a related thing.

[dsarrut]

Discussion are still pending inside the collaboration about the GPU parts in Gate. Not sure yet if it will be maintained/enhanced or removed.