I believe CUDA is the answer to my problem of a slow running software program. I think parallel processing is the way to go. However, management wants to see a proof of concept. It is my understanding that one can run CUDA compiled programs in emulation mode. The output will not only be the same as if your ran with CUDA enabled hardware, but with data on how this program would perform when it is run on CUDA enabled hardware - GPUs. Is this correct?
Of course the program run in emulation mode is slower.
Sorry, emulation mode is only applicable for some aspects of debugging - looking at variables etc.
You an however get a reasonably good handle on speedup by looking at examples and carefully studying your problem. The proof is in the pudding though …