NVIDIA / NVIDIA/cutlass

[QST]In tma_wasp producer's 128 threads, warp1 and warp3 are idle?

Open
#1,675 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

? - Needs Triage inactive-30d inactive-90d question
Dominant language
C++
Stars
10.5k
Forks
2.1k
Avg merge
3d 11h
Merged PRs (30d)
7

Description

sm90_gemm_tma_warpspecialized_cooperative

    enum class ProducerWarpRole {
      Mainloop = 0,
      Warp1 = 1,
      Epilogue = 2,
      Warp3 = 3
    };

I find usage of Mainloop and Epilogue, but no usage of warp1 and 3?

By the way, I noticed if we use 320 threads(do not use warp1 3) the occupancy will be 10, and if we use 384 threads(original cutlass method) the occupancy will be 12. Maybe because of this?

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with include/cutlass/gemm/kernel/sm90_gemm_tma_warpspecialized_cooperative.hpp around the ProducerWarpRole definition and inspect the usages of Mainloop, Warp1, Epilogue, and Warp3. Compare the 320-thread and 384-thread configurations mentioned in the issue and determine whether the apparently unused warp roles are intentional and how they affect occupancy.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.