layernorm: enlarge the range for 2-pass reduction #2282
Conversation
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
EikanWang
reviewed
Nov 5, 2025
| N / 32 < syclGpuEuCount() / syclGpuEUCountPerSubslice() / 2) { | ||
| int subslice_count = syclGpuEuCount() / syclGpuEUCountPerSubslice(); | ||
| if (use_two_stage_col_reduction && M > subslice_count * 1024 && | ||
| N / 32 < subslice_count) { |
Contributor
There was a problem hiding this comment.
How could we correlate N / 32 < subslice_count to the performance insights? Does the 32 mean simd32?
Co-authored-by: Eikan Wang <[email protected]>
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Outdated suggestions cannot be applied.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.
From the OOB models, there are some shapes still below the performance expectation with large M but small N.
Simple shapes:
[128, 197, 384]
[64,784, 256]
[(64, 28, 28, 256]
[256, 197, 256]
[128, 196, 384]
After enlarging the range for 2-pass reduction, these models can benefit an average of 10-20ms model execution time and optimize the geomean performance of eager training in timm models from 0.835 to 0.842.