"Perfect" will be great. Here in 2012 we don't have good performing GPUs with open documentation available for integration. This is what we can have now. It su

"Perfect" will be great. Here in 2012 we don't have good performing GPUs with open documentation available for integration. This is what we can have now. It sucks a bit. I don't see it changing anytime soon. That's most of the reason I supported Parallela. That promises to be open, it probably won't be as much power as a nice GPU, but I want to compute with that hardware, and more problems will fit better. Plus I won't get trapped in some flawed API that has weird black boxes bugs and unpredictable performance issues.