trevor-m commented on pull request #7195: URL: https://github.com/apache/tvm/pull/7195#issuecomment-754151980
Nice! Have you also looked at CUB's [DeviceSegmentedRadixSort::SortPairsDescending](https://nvlabs.github.io/cub/structcub_1_1_device_segmented_radix_sort.html#abbba6639f928bebd19435f185ea10618) ? It sounds like it is exactly what you need with no tricks required. It's used by some fast NMS implementations such as [TensorRT](https://github.com/NVIDIA/TensorRT/tree/master/plugin/batchedNMSPlugin). ---------------------------------------------------------------- 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: us...@infra.apache.org
