Skip to content

Update the error message for cublas version check#2843

Merged
ptrendx merged 1 commit intoNVIDIA:mainfrom
yaox12:xiny/update_cublas_comment
Apr 7, 2026
Merged

Update the error message for cublas version check#2843
ptrendx merged 1 commit intoNVIDIA:mainfrom
yaox12:xiny/update_cublas_comment

Conversation

@yaox12
Copy link
Copy Markdown
Member

@yaox12 yaox12 commented Apr 7, 2026

Description

Now that cuBLAS has its own semantic versioning, update the error message to avoid confusion.

cc @ksivaman @ptrendx Please review.

Type of change

  • Documentation change (change only to the documentation, either a fix or a new content)
  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to not work as expected)
  • Infra/Build change
  • Code refactoring

Changes

Please list the changes introduced in this PR:

  • Change A
  • Change B

Checklist:

  • I have read and followed the contributing guidelines
  • The functionality is complete
  • I have commented my code, particularly in hard-to-understand areas
  • I have made corresponding changes to the documentation
  • My changes generate no new warnings
  • I have added tests that prove my fix is effective or that my feature works
  • New and existing unit tests pass locally with my changes

Signed-off-by: Xin Yao <xiny@nvidia.com>
@greptile-apps
Copy link
Copy Markdown
Contributor

greptile-apps Bot commented Apr 7, 2026

Greptile Summary

This PR corrects five error messages in the #else stub block of cublaslt_grouped_gemm.cu that fire when the compile-time cuBLAS version is below 13.3. The old messages instructed users to "upgrade to CUDA 13.3", which is misleading because cuBLAS now follows its own independent semantic versioning — cuBLAS 13.3 is actually shipped with CUDA 13.2, not a hypothetical CUDA 13.3.

Confidence Score: 5/5

Safe to merge — pure error message correction with no logic or behavioral changes.

All changes are string literals inside compile-time fallback stubs. No logic, API, or behavior is altered. The updated wording is factually accurate per cuBLAS's independent versioning scheme.

No files require special attention.

Important Files Changed

Filename Overview
transformer_engine/common/gemm/cublaslt_grouped_gemm.cu 5 error message strings updated in version-guard stubs to correctly reference cuBLAS 13.3 and its CUDA 13.2 shipping vehicle; no logic changes.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Compile-time check] --> B{CUBLAS_VERSION >= 130300?}
    B -- Yes --> C[Full grouped GEMM implementation compiled in]
    B -- No --> D[Stub functions compiled in]
    D --> E[NVTE_ERROR called at runtime]
    E --> F["Error: requires cuBLAS 13.3+, current version is CUBLAS_VERSION
Please upgrade to cuBLAS 13.3 (shipped with CUDA 13.2) or newer."]
Loading

Reviews (1): Last reviewed commit: "update the error message for cublas vers..." | Re-trigger Greptile

@pggPL pggPL self-requested a review April 7, 2026 06:35
Copy link
Copy Markdown
Collaborator

@pggPL pggPL left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@ptrendx ptrendx merged commit edf10bb into NVIDIA:main Apr 7, 2026
10 of 12 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants