Repository navigation
Conversation
… when C or D is bfloat16
dkhaldi
marked this pull request as ready for review
October 23, 2026 19:42
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
When C or D is bfloat16, each DPAS rounds its result to bfloat16, and a
TK=32 joint_matrix_mad is two hardware DPAS with K=16. The reference computation on CPU
rounds only once, so with random fractional inputs the difference can
exceeds the 0.1 tolerance and made joint_matrix_bfloat16_accumulator.cpp fail
sporadically on PVC and BMG.
When C or D is bfloat16, this PR fills A and B with integers in [-1, 1] from a
fixed seed and compare exactly and the result no longer depends on where the rounding happens.
Fixes #21096