[GH-ISSUE #631] [Question] about EPv2 hybrid mode get_theoretical_num_qps #377

Open
opened 2026-10-26 09:59:44 +00:00 by bsakaguchi · 0 comments
Owner

Originally created by @huakainomore on GitHub (May 11, 2026).
Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/631

hybrid mode set num_qps to num_sms * 16 + 1, But it seems only kNumChannelsPerSM were used and kNumMaxChannelsPerSM is 8, So is num_sms * 8 enough?

    const auto [qp_idx, sharing_mode] = comm::get_qp_mode<kNumSMs, kNumQPs, kNumChannelsPerSM, (kNumNotifyWarps > 0)>(
        sm_idx, (warp_idx - kNumNotifyWarps) % kNumChannelsPerSM, warp_idx < kNumNotifyWarps);
    const auto gin = handle::NCCLGin(nccl_dev_comm, nccl_window, qp_idx, sharing_mode);
Originally created by @huakainomore on GitHub (May 11, 2026). Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/631 hybrid mode set num_qps to num_sms * 16 + 1, But it seems only `kNumChannelsPerSM` were used and `kNumMaxChannelsPerSM` is 8, So is num_sms * 8 enough? ``` const auto [qp_idx, sharing_mode] = comm::get_qp_mode<kNumSMs, kNumQPs, kNumChannelsPerSM, (kNumNotifyWarps > 0)>( sm_idx, (warp_idx - kNumNotifyWarps) % kNumChannelsPerSM, warp_idx < kNumNotifyWarps); const auto gin = handle::NCCLGin(nccl_dev_comm, nccl_window, qp_idx, sharing_mode); ```
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepEP#377
No description provided.