Look into nvcc compiler options

[stiber]

For example, we should be using a virtual architecture switch for the -arch option. See Options for Steering GPU Code Generation in the CUDA documentation.

[stardriftfx]

@stiber I think this issue is related to #262 as the changes are being made to CMakeList.txt. Specifically:

        set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} 
                -gencode=arch=compute_${TARGET_ARCH},
                code=[sm_${TARGET_ARCH}, compute_${TARGET_ARCH}])    

(Added code=[sm_${TARGET_ARCH}, compute_${TARGET_ARCH} where sm_ = real architecture, compute_= virtual architecture)