test: demonstrate a HybridFunction performance regression (DO NOT MERGE) - #1588
Draft
mrousavy wants to merge 3 commits into
Draft
test: demonstrate a HybridFunction performance regression (DO NOT MERGE)#1588mrousavy wants to merge 3 commits into
mrousavy wants to merge 3 commits into
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Nitro performance
Android
iOS
Base |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Performance regression canary — DO NOT MERGE
This draft PR intentionally adds 128 rounds of dependent unsigned arithmetic to
each HybridFunction native-state lookup. Volatile local storage keeps the work
in optimized Release builds without shared state, sleeps, allocations, or altered
return values. Closing this PR discards the entire experiment.
The base is the top of the infrastructure stack, including the version-alignment
PR. Both base and head already contain the same benchmark app and definitions.
CI must build both binaries independently and run the existing ABBA sequence.
The suite hash must remain identical; this is not a rebaseline or an A/A run.
Expected observations:
a detected regression does not imply that the build check should turn red.
End-to-end validation checks the real GitHub Actions artifacts, confirms base
and head SHAs and matching suite hashes, inspects the paired PR comment, and
verifies Bencher reports for both platform testbeds and this PR branch. Bencher
publishing requires the rotated project key in
BENCHER_KEYandNITRO_BENCHER_ENABLED=true; the paired PR comment is independent of that key.Reporting fixes belong in the infrastructure PR, not in this canary diff. Keep
this PR draft and never merge its intentional production slowdown.
First paired CI result
Run 33898772711
built both revisions independently and completed six base/head suites per
platform, including the one permitted noise retry. Both sides used Release
Nitro, Hermes, and the same benchmark-suite hash.
addNumbers: 173.6 ns/op → 621.4 ns/op (+258%, 95% CI+247%…+261%). The JS control changed -0.6%; the TurboModule control remained
inconclusive.
addNumbers: 161.5 ns/op → 450.5 ns/op (+179%, 95% CI+154%…+211%). The first reporter called every synchronous iOS case noisy
because its robust CV exceeded 5%, even when the paired interval was
decisive.
That result exposed a reporting-policy bug rather than a sampling failure. The
infrastructure PR now lets a change whose full paired confidence interval is
beyond the budget keep its regression/improvement verdict. CV still produces a
neutral result when the interval cannot decide. Rebuilding the report from the
same raw samples yields 29 Android and 25 iOS regressions while leaving the JS
and TurboModule controls neutral.