Repository navigation
perf(core): run few-row quantW matmuls on a K-split matmulQuantW with f16 and the epilogue instead of the tiled kernel #96
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merged
Merged
Changes from all commits
Commits
Show all changes
9 commits
Select commit
Hold shift + click to select a range
0f5f1f9
perf(core): split K across lanes in matmulQuantW and route quantW mat…
chmjkb 5cc4ef4
test(core): cover matmulQuantW's f16 and epilogue cases and both side…
chmjkb 3a1564c
test(core): add f16 and epilogue quantW cases at MiniLM and decode sh…
chmjkb c817a94
fix(core): refuse a quantW scaleBase that is not a multiple of 4, sin…
chmjkb f5ef008
refactor(core): share one config and one buffers type between matmulQ…
chmjkb cfc6d6f
refactor(core): add doc comments to matmulQuantW's workgroup shape co…
chmjkb d9a137e
refactor(core): look up the element type once in matmulQuantWSpec's e…
chmjkb 63e1363
fix(core): keep quantW matmuls under 32 rows on matmulQuantW, since a…
chmjkb fcbcf19
test(core): add wide-N quantW cases under 32 rows to the matmul A/B b…
chmjkb File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
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
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
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
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
Oops, something went wrong.
Oops, something went wrong.
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.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
With wide N a single row already passes m·n ≥ 48·1024, so few-row calls land on the 32-row tile again. [1,2048]×[2048,131072] takes 6505 µs tiled vs 1362 on matmulQuantW, [4,4096]×[4096,14336] 1522 vs 474. On N 384–1536 the threshold is right. Maybe also require m >= 32? That fits every case I measured.