Skip to content

perf(index): add four-chain native u2 and u6 ex-dot kernels - #9839

Merged
BubbleCal merged 2 commits into
mainfrom
yang/layered-rq-native-kernels
Oct 10, 2026
Merged

BubbleCal merged 2 commits into
mainfrom
yang/layered-rq-native-kernels

Conversation

@BubbleCal

@BubbleCal BubbleCal commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

Use four independent accumulation chains in the native AVX-512 u2/u6 ex-dot dispatch. Preserve the existing blocked code layouts and scalar/fallback behavior; fused layered-plane kernels are reviewed separately.

Dependencies

Independent PR targeting main; no other Layered RQ PR is a prerequisite.

The original validated base is retained. Current-main and internal-fork integration remains OSS-2559. Standalone benchmark/POC tools remain excluded.

Validation

Fresh cargo +1.97.0 fmt --all -- --check and full-workspace clippy (--all --tests --benches -- -D warnings) passed for this new branch base on macOS arm64. The full composed tree matches the previous submitted tip exactly. No new performance measurement is claimed; CI is not awaited.

Tracking

OSS-2346

@github-actions github-actions Bot added A-index Vector index, linalg, tokenizer performance labels Oct 9, 2026
@BubbleCal
BubbleCal force-pushed the yang/layered-rq-native-kernels branch from c74090f to de9ffbe Compare October 9, 2026 18:55
@BubbleCal
BubbleCal force-pushed the yang/layered-rq-cache-pins branch from ffb720a to e003e44 Compare October 9, 2026 18:55
@BubbleCal
BubbleCal changed the base branch from yang/layered-rq-cache-pins to main October 9, 2026 18:56
@BubbleCal
BubbleCal marked this pull request as ready for review October 10, 2026 16:19

@claude claude Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Claude Code Review

This repository is configured for manual code reviews. Comment @claude review for a one-time review, or @claude review always to subscribe this PR to a review on every future push.

Tip: disable this comment in your organization's Code Review settings.

@lance-gatekeeper lance-gatekeeper Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

✅ Gate recommendation: approve.

The existing format and fallback contracts remain intact. AVX-512 parity checks passed, and paired kernel timings improved for both widths at 1536 and 2048 dimensions. No blocking issues found.

@lance-gatekeeper lance-gatekeeper Bot added the K-approved Latest Gatekeeper recommendation permits acceptance. label Oct 10, 2026
@BubbleCal
BubbleCal merged commit fbec024 into main Oct 10, 2026
38 checks passed
@BubbleCal
BubbleCal deleted the yang/layered-rq-native-kernels branch October 10, 2026 17:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

A-index Vector index, linalg, tokenizer K-approved Latest Gatekeeper recommendation permits acceptance. performance

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants