1991726 - 6.5% regression on Kraken-imaging-gaussian blur on 16Sep2025

Status: RESOLVED WORKSFORME

(Core :: JavaScript Engine, defect)

Description

[Mayank Bansal]

https://treeherder.mozilla.org/perfherder/graphs?highlightAlerts=1&highlightChangelogData=1&highlightCommonAlerts=0&replicates=0&selected=5306404,2130402949&series=autoland,5306404,1,1&timerange=5184000&zoom=1758029765001,1758032104525,63.72594205980259,72.38914591668309

This regressed by bug 1986983, but then partially improved somewhat later (my best guess is bug 1991077)

Filing this regression for posterity. This is not a major regression.

Comment 1

[BugBot [:suhaib / :marco]]

Set release status flags based on info from the regressing bug 1986983

:jandem, since you are the author of the regressor, bug 1986983, could you take a look? Also, could you set the severity field?

For more information, please visit BugBot documentation.

Comment 2

[Jan de Mooij [:jandem]]

(In reply to Mayank Bansal from comment #0)

Filing this regression for posterity. This is not a major regression.

Thanks for reporting but I think we can close this.

Locally when running this test in the JS shell, we're a bit (a few %?) faster with objectfuse_for_js_builtin_ctors_protos=true than with objectfuse_for_js_builtin_ctors_protos=false. Our codegen is a little bit better too, but we hoisted the Math.abs guards out of the (inner) loop so eliminating those guards is not a huge win on this test.

There could be small differences in register allocation or alignment that might have affected CI. Skimming the machine code I didn't see anything obviously worse compared to before, but I also didn't compare in detail.