Support FP8 scale calculation with scalar and cleanup #2593
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
Follow up on D57263833 to support FP8 scale calculation with scalar and merge two FP8 tensorwise GEMMs into one
Note that besides
Sm90ScalarBroadcast
in CUTLASS, AMD CK f8f8bf16 GEMM also requires passing scales as scalar instead of tensor scalar. This support is required in both NV and AMD sidesDifferential Revision: D57367680