[kainzhong]

Description

Add CuTeDSL port of CUDA's MXFP8 gated quantization kernel

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

• Add CuTeDSL MXFP8 gated kernel

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; no actionable defects were found.

Summary

Adds CuTeDSL kernels for gated MXFP8 forward and backward operations.

• MXFP8 gated activations can run through CuTeDSL.

No actionable issues were found.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Gated MXFP8 call] --> B{CuTeDSL enabled and inputs supported?}
  B -->|No| F[CUDA kernel]
  B -->|Yes| C[Find or compile gated kernel]
  C --> D{Kernel available?}
  D -->|No| F
  D -->|Yes| E[Load activation and gate tiles]
  E --> G[Compute gated output or gradients]
  G --> H[Quantize outputs and write scales]

Loading

_{Reviews (1) · Last reviewed commit: "[pre-commit.ci] auto fixes from pre-comm..." · Reviewed by Greptile}