[kainzhong]

Description

Add MXFP8 2D quantization CuTeDSL kernels

Fixes # (issue)

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:

• MXFP8 2D quantization CuTeDSL kernels

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

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; both previous findings are addressed and no new actionable issues were found.

Summary

Adds CuTeDSL MXFP8 quantization with one shared scale per 32×32 block.

• CuTeDSL now quantizes plain MXFP8 casts with one scale per 32-by-32 block.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[MXFP8 cast request] --> B{2D scaling requested?}
  B -->|Yes| C[2D tile math in general pipeline]
  B -->|No| D[Existing kernel selection]
  C --> E[FP8 values and repeated block scales]
  D --> F[FP8 values and directional scales]

Loading

_{Reviews (2) · Last reviewed commit: "[pre-commit.ci] auto fixes from pre-comm..." · Reviewed by Greptile}