[sraman-rgb]

Compute both operands' per-group NVFP4 global scales and their product (or its square root for dGLU) in one Triton launch instead of several eager elementwise kernels. Bit-identical to the eager ops; falls back to them for unsupported inputs.

Description

Please include a brief summary of the changes, relevant motivation and context.

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:

• 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

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no actionable defects were found.

Summary

Fuses NVFP4 operand scales and optional GEMM alpha into one Triton launch for the grouped MLP.

• Grouped MLP scale pairs and GEMM alpha are computed in one Triton launch.

_{Reviews (2) · Last reviewed commit: "Merge upstream main into sraman/nvfp4-fu..." · Reviewed by Greptile}