A Thread-Register Decoupled GPU Execution Model for Efficient Tensor Computation

 (arxiv.org)

14 points | by matt_d 3 hours ago ago

No comments yet.