Apparently, how many optimization levels there are and what techniques each one adopts have not been well documented yet in here:http://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html#compilation-phases. Is it similar to gcc:http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html?
In addition, the sample makefile in sdk uses -fno-strict-aliasing as one of the nvcc flags, which is not mentioned in the nvcc manual. What else gcc options may apply to nvcc?
Thanks for clarifications!