lsy643 commented on pull request #6108: URL: https://github.com/apache/incubator-tvm/pull/6108#issuecomment-692416856
@yongwww The `rearrange_indices_out` is a newly part added to `non_max_suppression` from `topi/cuda/nms.py`, and so I compare the latency of `non_max_suppression` when `return_indices=True` and `return_indices=False`. For test data with shape [1, 20000, 6] When `return_indices=True`, the `rearrange_indices_out` is used - latency: 6342us When `return_indices=False`, the `rearrange_indices_out` is used - latency: 5645us ---------------------------------------------------------------- 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]
