[QST] Cutlass Profiler
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 10.5k
- Forks
- 2.1k
- Avg merge
- 3d 11h
- Merged PRs (30d)
- 7
Description
What is your question?
When I run
cutlass_profiler --operation=gemm --enable-best-kernel-for-fixed-shape --m=6144 --n=6144 --k=6144 --sort-results-flops-per-sec --output=gemm.csv
The cluster_m, cluster_n columns are always 1, even for gemms with procedural names like:
cutlass3x_sm90_tensorop_gemm_f16_f16_f32_f16_f16_256x128x64_1x2x1_0_tnn_align8_warpspecialized_cooperative_epi_tma
which have a static ClusterShape (1x2x1 in the given example) encoded in the CollectiveMainloop.
However, when I run the following for the same set of compiled gemms:
cutlass_profiler --operation=gemm --m=6144 --n=6144 --k=6144 --output=gemm.no_search.csv
the cluster_m and cluster_n columns match those of the template.
Using cutlass 4.0 where I compiled only a subset of gemms matching the pattern cutlass3x_sm90_tensorop_gemm_f16_f16_f32_f16_f16_*_tn*.
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
Start with the two cutlass_profiler invocations and compare the --enable-best-kernel-for-fixed-shape path with the ordinary GEMM reporting path. Reproduce the discrepancy using the stated SM90 procedural kernel pattern and fixed 6144x6144x6144 shape; done means cluster_m and cluster_n consistently reflect the compiled template in both outputs.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- performance, tooling
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100