[QST] Why we need prologue mma in hopper densegemm
Open
Nobody has claimed this yet.
? - Needs Triage
inactive-30d
question
- Dominant language
- C++
- Stars
- 10.5k
- Forks
- 2.1k
- Avg merge
- 3d 11h
- Merged PRs (30d)
- 7
Description
What is your question?
I noticed that in https://github.com/NVIDIA/cutlass/blob/main/examples/python/CuTeDSL/cute/hopper/kernel/dense_gemm/dense_gemm.py#L772-L824, there exist a prologue mma before mainloop, why we need to separate the first mma iter from mainloop?...
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Read examples/python/CuTeDSL/cute/hopper/kernel/dense_gemm/dense_gemm.py at lines 772-824, focusing on the prologue MMA and the mainloop boundary. Trace why the first MMA iteration is separated and document whether that separation is required; the issue is done when the rationale is clearly answered.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- hpc
- Issue type
- Documentation
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 42/100