CUDA_FT

Jun 26, 2012 at 2:18 AM
Edited Jun 26, 2012 at 2:18 AM

it's not convinient to set the variant when run the CUDA_FT code, its default setting is CLASS=A, it   seems that it compiled in default nodes too. so, it 's hard to see whether using CUDA is better than MPI, OMP or hybrid. how do you think about it, looking forwards to your reply.

Would  someone like to recommend some material about FT to me.  thanks in advance anyway.