docs(research): pretable scroll-with-render perf diagnostic (noise verdict)#133
Merged
Conversation
Three-phase research PR mirroring PR #124's pattern: high-repeat (n=20) re-run, conditional Playwright trace capture, research memo. Diagnoses whether the PR #130 cheap-render anomaly (16.4 ms vs 10.3 ms for format and heavy-render) is real or a low-sample artifact. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Seven-task plan mirroring PR #124's three-phase pattern: n=20 matrix re-run, conditional Playwright trace capture, research memo. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Verdict: noise. The PR #130 cheap-render anomaly (16.4 ms vs ~10.3 ms for format and heavy-render at n=3) was a sampling artifact. At higher repeats, scroll-with-render is at parity with (in fact marginally faster than) the other two: | Script | n | mean (ms) | σ (ms) | | -------------------------- | --: | --------: | -------: | | scroll-with-format | 8 | 9.36 | 0.80 | | scroll-with-render | 7 | 8.97 | 0.35 | | scroll-with-heavy-render | 6 | 9.15 | 0.13 | Both 2σ pairs (cheap-vs-format, cheap-vs-heavy) are well within the noise floor. Same shape as PR #124's finding at larger magnitude. The matrix run completed only ~36% of planned repeats (Playwright flake; not investigated) but the observed σ values make the verdict unambiguous — PR #130's 6 ms gap is ~21σ away from the observed distribution. No perf-fix PR needed. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Contributor
Vercel preview readyPreview: https://pretable-d07kph9m5-cacheplane.vercel.app Updated automatically by the |
This was referenced May 11, 2026
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
1 participant
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
Diagnoses whether pretable's `scroll-with-render` 6 ms gap from PR #130 (16.4 ms vs ~10.3 ms for format + heavy-render at n=3) is real or a low-sample artifact. Verdict: noise. Mirrors PR #124's pattern at larger magnitude.
Verdict
At higher repeats, `scroll-with-render` is at parity with (in fact marginally faster than) `scroll-with-format` and `scroll-with-heavy-render`:
Both 2σ pairs are well within the noise floor:
The PR #130 cheap-render value of 16.4 ms is ~21σ away from the observed distribution — statistically impossible at this sample size.
No perf-fix PR needed. No React-reconciliation cliff exists between single-text-child and multi-child spans in pretable's MemoizedCellContent path.
Sample size caveat
The matrix completed only ~36 % of planned repeats (Playwright flake; not investigated). Sample counts came out 6–8 per script instead of n=20. Documented in the memo; the observed σ values make the verdict unambiguous regardless — a 6 ms gap couldn't hide in this distribution.
What's NOT in this PR
Test plan
🤖 Generated with Claude Code