I'm pessimistic. I doubt things will change unless all the major platform owners can come to an agreement on a common API for GPU compute. If NVIDIA were the on

I'm pessimistic. I doubt things will change unless all the major platform owners can come to an agreement on a common API for GPU compute. If NVIDIA were the only holdout, they might be forced to support it. As it stands, CUDA is the only API with a decent implementation available on the three major desktop OSes. It's ridiculous.