2037332 - 55.72 - 4.29% pinterest largestContentfulPaint + 4 more (Linux) regression on Tue May 5 2026

Status: RESOLVED WONTFIX

(Core :: JavaScript Engine, defect)

Description

[Treeherder Bug Filer]

Perfherder has detected a browsertime performance regression from push 88491774a5332a7413b9f6b3f382304516bc9e02. As author of one of the patches included in that push, we need your help to address this regression.

Please acknowledge, and begin investigating this alert within 3 business days, or the patch(es) may be backed out in accordance with our regression policy.

Regression	Test	Platform	Options	Absolute values [old vs new]	Performance Profiles
56%	pinterest largestContentfulPaint (doc)	linux2404-64-shippable	fission warm webrender	384.37 -> 598.54	Before/After
34%	pinterest ContentfulSpeedIndex (doc)	linux2404-64-shippable	fission warm webrender	518.55 -> 694.72	Before/After
27%	pinterest SpeedIndex (doc)	linux2404-64-shippable	fission warm webrender	496.63 -> 631.32	Before/After
18%	pinterest fcp (doc)	linux2404-64-shippable	fission warm webrender	228.48 -> 269.18	Before/After
4%	speedometer3 Editor-CodeMirror/total (doc)	linux2404-64-shippable	fission webrender	38.29 -> 39.93	Before/After

Improvements	Test	Platform	Options	Absolute values [old vs new]	Performance Profiles
62%	google-search LastVisualChange (doc)	linux2404-64-shippable	fission warm webrender	756.02 ms -> 290.58 ms	Before/After
41%	tumblr largestContentfulPaint (doc)	linux2404-64-shippable	fission warm webrender	636.63 ms -> 377.18 ms	Before/After
34%	tumblr LastVisualChange (doc)	linux2404-64-shippable	fission warm webrender	719.28 ms -> 472.65 ms	Before/After
24%	pinterest loadtime (doc)	linux2404-64-shippable	fission warm webrender	801.08 ms -> 610.53 ms	Before/After
19%	tumblr ContentfulSpeedIndex (doc)	linux2404-64-shippable	fission warm webrender	258.10 ms -> 208.81 ms	Before/After
...	...	...	...	...	...
4%	pinterest LastVisualChange (doc)	linux2404-64-shippable	fission warm webrender	3,745.99 ms -> 3,587.62 ms	Before/After

Support & Actionable Steps

For questions, please contact [e-mail]. You can also find help on Slack (#perf-help) or Matrix (#perftest).

Action guide details:

• See the guide to handling regression bugs for investigation steps.
• Review the alert summary for affected tests, graphs, and comparisons.
• Check the PerfCompare results.
• Tests on Try: Run all tests using ./mach try perf --alert 50125 (See documentation link for details).
• Profiling jobs: Trigger jobs from Treeherder job view or request them from [e-mail].

Comment 1

[BugBot [:suhaib / :marco]]

Set release status flags based on info from the regressing bug 2027803

Comment 2

[Tooru Fujisawa [:arai]]

This is expected. The patch introduced both improvements and regressions for the pageload performance, especially the visual score.

The patch introduces per-process in-memory cache for JavaScript, which allows reusing the result of fetch+compilation across navigation and reloads.
While this improves the pageload performance in general, this also means that the JS resources becomes available earlier than before, which means the other resources and tasks get less chances of getting processed earlier.
This means results in the visual score regression in some cases, for example the placeholder elements getting removed earlier, or the styles and images getting displayed later. There's no single reliable way to determine which resources or tasks should be processed first, and the improvement for single task often result in regressing some score.

In the browsertime scores, most of the regressions come from the unpredictability around the scheduling, which depends on various things such as server, network, hardware, concurrency, etc, and the score on the automation is just single sample, and also the testing environment doesn't represent the real use case, for example around the network latency.
We're planning to perform an A/B testing experiment on the release channel once this change reaches the release channel (on June 16), to determine the actual impact on the real use cases.

Anyway, we're going to accept the regression.