Hzfengsy commented on PR #80: URL: https://github.com/apache/tvm-rfcs/pull/80#issuecomment-1159400036
My observation: async works well for fp16-16-16 (use fp16 accumulator) but helps little for fp16-16-32. On the other hand, the best cutlass kernel only uses 3 stages on my machine. I guess it is because of the shared memory usage -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
