bench: re-measure the Windows reference profile after the dirty-index fix - #746
Merged
Merged
Conversation
Exoridus
enabled auto-merge (squash)
September 14, 2026 20:58
Bundle ReportChanges will increase total bundle size by 4.57kB (0.01%) ⬆️. This is within the configured threshold ✅ Detailed changes
Affected Assets, Files, and Routes:view changes for bundle: site-server-esmAssets Changed:
|
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
Re-measures the Windows reference profile (
rtx-5070-ti-windows-11-chromium.json) onnextafter #745, and fixes thecompositearm's figures, which every earlier published value predates.Three separate runs, pooled with
bench:compare --profile.verify:bench-resultsandgates siteare green, and the eight scenarios that carried no measurement on this profile now all carry values:dynamic-all,fill-layers,interaction-picking,particles-draw,particles-lifecycle,tilemap-edit,tilemap-scroll,ui-layout-update.Why it is a re-measurement
A first set of three runs on 2026-09-14 produced a profile that was formally valid and still not publishable. Two things were wrong with it, and both are gone from this one.
The engine carried the #741 regression fixed in #745, which cost
mask-clip-animatedabout four times its previous CPU time on the retained path. That cell now reads 0.143 ms where the discarded profile read 1.150 ms.The machine was also loaded during those runs. The contention probe — comparing both arms against the previous profile — showed 32 ExoJS cells and 8 competitor cells more than 25% slower, and nothing faster on either arm, which is a machine-level slowdown rather than a code effect. The same probe over this profile shows 4 ExoJS cells and 1 competitor cell outside the band. Those four are one cell counted across its competitors,
deep-hierarchyat 10000 nodes at 1.30x, which sits inside the drift two quiet runs are already known to produce.Known harness outcomes
Unchanged from earlier runs and labelled as such in the profile:
phaserwedges the browser onstatic-heavyat 100000 nodes (isolated asunavailable, the arm's remaining cells continue after a relaunch), andexcaliburruns into the frame ceiling at 100000 nodes onstatic-heavyanddynamic-all.https://claude.ai/code/session_013jqQeGU6gwsQhYDFrKiN1A