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

(arxiv.org)

12 points | by matt_d 2 hours ago ago

No comments yet.