Comment by jorlow

Does this compete at all with openAI's triton (which is sort of a higher level cuda without the vendor lock in)?