Skip to content

Nightly perf regression #10247

Closed
Closed

Description

@github-actions

Nightly perf run on 0081c11 found 1 metric(s) more than 10% worse than their 7-day median, as on the recorded run before it: run.

Scene Metric This run 7-day median Change
overlay-churn interactionLayouts 924 752 +23%

Commits since the last good run: b8f2a9f…0081c11

Activity

  1. cixzhang commented on Nov 7, 2026

    @cixzhang
    Contributor

    Reopening: the first nightly after #10368 (run on a19d45a, which includes it) measured overlay-churn interactionLayouts at 805 (samples 805, 817, 799, 820, 703) against the 752 baseline. #10368 measured 709 on its own base, and one nightly sample is at that level. So a change merged after it, in 2816569..a19d45a, likely adds layouts again, partly depending on timing. Candidates on the scene's path: #10240, #10398, #10290, #10351.

  2. cixzhang commented on Nov 8, 2026

    @cixzhang
    Contributor

    The first nightly on the settled scene (run on bb3a4b2, which includes #10368 and #10496) reads overlay-churn-settled interactionLayouts at 1100 on all five samples (1100, 1100, 1100, 1100, 1100), and the other scenes are unchanged (2/8/5/1). The count no longer depends on the runner. It is lower than the 2258 to 2272 measured on #10496 because #10561, merged the same day, plays anchored entrances on the surface, so they no longer lay out on every frame. Closing: #10368 fixed the extra layout on close, and #10496 made the scene's count independent of runner speed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions