zxybazh commented on PR #14402: URL: https://github.com/apache/tvm/pull/14402#issuecomment-1483729037
Yeah if we can have such attribute to mark the primfunc, it would be even more convenient and generic to targets outside cuda. Specifically we are looking into the case where we have a cumsum operator manually scheduled -- where we want to avoid doing that again in tuning pass. The is kinda new because when we design MetaSchedule we are not expecting scheduled workloads and now it's possible given MS is a pass in the TVM Unity. -- 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]
