Skip to content

fix moe fp4 dual prebuild issues#1775

Merged
valarLip merged 4 commits intomainfrom
prebuild_dual
Jan 12, 2026
Merged

fix moe fp4 dual prebuild issues#1775
valarLip merged 4 commits intomainfrom
prebuild_dual

Conversation

@zufayu
Copy link
Contributor

@zufayu zufayu commented Jan 6, 2026

Motivation

Technical Details

Test Plan

Test Result

Submission Checklist

Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

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

Pull request overview

This PR fixes issues with FP4 (4-bit floating point) instance generation for MoE (Mixture of Experts) GEMM operations by adding architecture-aware build filtering to prevent generating FP4 code for unsupported GPU architectures.

Key Changes:

  • Added arch-aware filtering to skip FP4 instance generation unless gfx950 GPU architecture is targeted
  • Wrapped FP4 instances with preprocessor directives (#if defined(__gfx950__)) for compile-time filtering
  • Fixed typo: renamed variable from intsance to body

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

@valarLip
Copy link
Collaborator

valarLip commented Jan 7, 2026

@zufayu ci failed

@valarLip valarLip merged commit 2946bda into main Jan 12, 2026
20 of 21 checks passed
@valarLip valarLip deleted the prebuild_dual branch January 12, 2026 14:27
@lalala-sh lalala-sh restored the prebuild_dual branch January 13, 2026 07:22
zhuyuhua-v pushed a commit that referenced this pull request Jan 14, 2026
* bug fix

* fix flag

* update

---------

Co-authored-by: zufayu <zufayu@amd.com>
Co-authored-by: lalala-sh <Jiaxing.Wen@amd.com>
valarLip pushed a commit that referenced this pull request Mar 18, 2026
* bug fix

* fix flag

* update

---------

Co-authored-by: zufayu <zufayu@amd.com>
Co-authored-by: lalala-sh <Jiaxing.Wen@amd.com>
valarLip pushed a commit that referenced this pull request Mar 18, 2026
* bug fix

* fix flag

* update

---------

Co-authored-by: zufayu <zufayu@amd.com>
Co-authored-by: lalala-sh <Jiaxing.Wen@amd.com>
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.

6 participants