allow compilation to optimize for CUDA compute cap. 3.5

allow compilation to optimize for CUDA compute cap. 3.5

Enabling optimizations targeting compute capability 3.5 devices
(GK110) slightly improves performance of both PME and RF kernels.
This requires a hint for the compiler optimization indicating
the maximum number of threads/block and minimum number of
blocks/multiprocessor. This change allows nvcc >=5.0 to generate
code for CC 3.5 devices and switches to including PTX 3.5 code
(instead of 3.0) in the binary.

Change-Id: If7e14d31165bc05859250db7468bf6bd8c186264