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

 (arxiv.org)

16 points | by matt_d 6 hours ago ago

No comments yet.