tkonolige commented on pull request #7952: URL: https://github.com/apache/tvm/pull/7952#issuecomment-844369810
I agree that following precedent is useful for consistency, but it seems like it will cause a lot of problems here. What happens in this case: The first instrument run before the pass initializes some state and returns true. The second returns false. Now there is no matching after pass call for the first instrument. When the next before pass gets called, the instrument doesn't know if it is now more deeply nested or if it had actually been canceled. Also, given we already have a way to disable passes, wouldn't it make more sense to keep the existing interface? -- 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. For queries about this service, please contact Infrastructure at: [email protected]
