jinhongyii commented on PR #20072: URL: https://github.com/apache/tvm/pull/20072#issuecomment-5113958623
Closing this in favor of the existing explicit single-lane guard form (`if mma_issue:`). On the current grouped FP8 kernel, a like-for-like B200 A/B benchmark found no measurable regression from the explicit guard, and the generated SASS still uses uniform registers for the MMA descriptor path. The additional `pred` API therefore does not appear justified by this use case. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
