zxybazh commented on issue #13425: URL: https://github.com/apache/tvm/issues/13425#issuecomment-1320836531
Sorry for the confusion, to clarify, there're 2 issues with different targets here: 1. The vnni PR introduced a segfault when the vnni detect function is not imported on python side. This is caused when using llvm as target in the script shared here. related PR #13383 #13441 2. There's another issue triggered when using cuda target with this same script. Seems like an illegal reverse-compute-at. I've talked to @shingjan and he would provide some more context on which TIR & reverse-compute-at decision caused this issue as a follow up later. -- 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]
