masahi edited a comment on pull request #7154: URL: https://github.com/apache/tvm/pull/7154#issuecomment-749858279
@zhiics Sure updated the description. Unfortunately I cannot claim that this is perf improvement. The regression is only 200 us on CPU, so it may be just a measurement noise, though. I have no idea why I'm not getting good speed up. IOU tests, including memory access to boxes should be definitely reduced. The only additional overhead I think of is that the input to NMS is one column wider, due to storing class ids. Performance is not great, but I believe having access to class ids should not be a bad idea... ---------------------------------------------------------------- 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]
