What’s so hard about doing non-graphics programming on a GPU? What do IBM's Cell, NVIDIA's CUDA, and AMD's CTM all have in common? They're … ...