[GH-ISSUE #342] How to test the performance #239

New issue

Closed

opened 2026-10-26 09:57:11 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-26 09:57:11 +00:00

Owner

Originally created by @DDDVE on GitHub (Jul 30, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/342

Hi, could you please tell me how you got the performance data in the README? Did you only test the cuda kernel, or do the test from host api? Thank you!

Originally created by @DDDVE on GitHub (Jul 30, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/342 Hi, could you please tell me how you got the performance data in the README? Did you only test the cuda kernel, or do the test from host api? Thank you!

bsakaguchi closed this issue

2026-10-26 09:57:12 +00:00

bsakaguchi commented

2026-10-26 09:57:13 +00:00

Author

Owner

@sphish commented on GitHub (Aug 5, 2025):

You can refer to the test scripts under the tests directory, such as tests/test_internode.py. We call CUDA kernels via the host API, and use CUDA events or the PyTorch profiler to measure the kernel execution time.

@sphish commented on GitHub (Aug 5, 2025): You can refer to the test scripts under the `tests` directory, such as `tests/test_internode.py`. We call CUDA kernels via the host API, and use CUDA events or the PyTorch profiler to measure the kernel execution time.

bsakaguchi commented

2026-10-26 09:57:13 +00:00

Author

Owner

@DDDVE commented on GitHub (Aug 5, 2025):

I get it, thank you!

@DDDVE commented on GitHub (Aug 5, 2025): I get it, thank you!

bsakaguchi referenced this issue

2026-10-26 10:01:14 +00:00

[PR #239] [MERGED] Support displaying separate send and recv time #478