NVIDIA / NVIDIA/cutlass

[QST] Error when run example 23 gemm_operand_reduction_fusion with splitk

Open
#1,861 2 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

What is your question?
I want to fuse gemm and reduction ops and when I run this example with parameters --parallel_split_k and split-k_slices it can not work.
I test this example on A10 and here is my command and the result:

./23_ampere_gemm_operand_reduction_fusion --save-workspace --split-k-slices=2 --parallel-split-k ERROR - results miscompared. Results written to '23_ampere_gemm_operand_reduction_fusion1024x1024x1024.dat'. ID,M,N,K,SplitK-Slices,Parallel-SplitK,Runtime gemm_1,1024,1024,1024,2,1,0

Would you please help me with this error? PS: If my inputs are tf32, can I use cutlass to fuse these two ops?
Thanks a lot!!

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 by reproducing entry point 23_ampere_gemm_operand_reduction_fusion with --split-k-slices=2 and --parallel-split-k on the reported A10 configuration, then inspect the generated 23_ampere_gemm_operand_reduction_fusion1024x1024x1024.dat results. Done means determining why the results miscompare and whether the example supports the reported split-K and TF32 fusion inputs.

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
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.