NVIDIA / NVIDIA/cutlass

[BUG] CuteDSL example hits IMA for large tensors due to strides in Int32

Open
#2,312 13 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

Tensors in Cute DSL uses strides in int32 by default. This causes IMA for large tensors. Is there a way to force strides to be int64?

Steps/Code to reproduce bug

python examples/python/CuTeDSL/ampere/elementwise_add.py --M=65384 --N=65384
RuntimeError: CUDA error: an illegal memory access was encountered

cc: @thakkarV

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 examples/python/CuTeDSL/ampere/elementwise_add.py and reproduce the issue using the reported --M=65384 --N=65384 command. Trace how Cute DSL assigns tensor strides and determine whether int64 strides can be selected; done means the large-tensor example no longer causes the reported illegal memory access and the supported configuration is covered.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp, python
Domain
backend, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
42/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.