1638303 - 2.01 - 10.64% about_newtab_with_snippets / displaylist_mutate / perf_reftest_singletons[...] (linux64-shippable|-qr, windows10-64-shippable|-qr, windows7-32-shippable) regression on push 80ba3f3cfaf9b31341bca28e2e6aea383cec8acc (Thu May 14 2020)

Status: RESOLVED FIXED

(Firefox Build System :: Toolchains, defect)

Description

[Alexandru Ionescu (needinfo me) [:alexandrui]]

Perfherder has detected a talos performance regression from push 80ba3f3cfaf9b31341bca28e2e6aea383cec8acc. As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

11% about_newtab_with_snippets linux64-shippable opt e10s stylo 77.15 -> 85.36
9% perf_reftest_singletons id-getter-2.html windows10-64-shippable-qr opt e10s stylo 687.15 -> 748.96
8% perf_reftest_singletons id-getter-2.html windows10-64-shippable opt e10s stylo 689.70 -> 747.84
8% tp5o_webext responsiveness linux64-shippable opt e10s stylo 2.30 -> 2.48
8% tp5o_webext responsiveness linux64-shippable-qr opt e10s stylo 2.56 -> 2.76
7% perf_reftest_singletons external-string-pass.html windows10-64-shippable-qr opt e10s stylo 706.89 -> 758.64
6% perf_reftest_singletons external-string-pass.html windows10-64-shippable opt e10s stylo 714.52 -> 760.89
6% perf_reftest_singletons window-named-property-get.html linux64-shippable-qr opt e10s stylo 585.13 -> 622.21
6% tsvgr_opacity linux64-shippable-qr opt e10s stylo 103.92 -> 110.35
6% perf_reftest_singletons window-named-property-get.html linux64-shippable opt e10s stylo 561.28 -> 593.98
5% tsvgx windows7-32-shippable opt e10s stylo 123.39 -> 129.68
5% perf_reftest_singletons id-getter-2.html linux64-shippable-qr opt e10s stylo 647.35 -> 680.21
5% displaylist_mutate linux64-shippable opt e10s stylo 1,381.56 -> 1,450.21
4% tsvgr_opacity linux64-shippable opt e10s stylo 147.34 -> 153.56
4% tscrollx linux64-shippable opt e10s stylo 0.66 -> 0.68
4% sessionrestore linux64-shippable opt e10s stylo 524.96 -> 544.08
3% about_newtab_with_snippets linux64-shippable-qr opt e10s stylo 83.81 -> 86.48
3% tsvgx linux64-shippable opt e10s stylo 180.14 -> 185.08
3% tp5o_scroll linux64-shippable opt e10s stylo 1.24 -> 1.27
3% tp5n main_startup_fileio windows7-32-shippable opt e10s stylo 107,713,002.00 -> 110,475,998.75
2% tart linux64-shippable opt e10s stylo 2.00 -> 2.04
2% startup_about_home_paint_realworld_webextensions linux64-shippable opt e10s stylo 774.25 -> 790.58
2% tp5o_webext linux64-shippable opt e10s stylo 284.97 -> 290.71

Details of the alert can be found in the alert summary, including links to graphs and comparisons for each of the affected tests. Please follow our guide to handling regression bugs and let us know your plans within 3 business days, or the offending patch(es) will be backed out in accordance with our regression policy.

For more information on performance sheriffing please see our FAQ.

Comment 1

[(Away)]

These numbers are impressively bad.

And it's even worse when we look at individual subtests. I was hoping to start with the svgx subtest 6.xml on win32 since it's a consistent 25% regression, but I can't reproduce it locally. https://treeherder.mozilla.org/perf.html#/graphs?highlightAlerts=1&series=autoland,1922707,1,1&timerange=1209600&zoom=1589227478952,1589517920439,124.83055555555555,232.075

Comment 2

[(Away)]

OK, I can reproduce a 9% regression in win64 PGO perf_reftest_singletons id-getter-2.html. I'll need to continue from here on Monday.

Comment 3

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1616692

Comment 4

[Julien Cristau [:jcristau]]

The clang update was backed out.

Comment 5

[Daniel Holbert [:dholbert]]

Looks like this was fixed-by-backout per comment 4 (and we ended up skipping clang 10 entirely per bug 1616692 comment 37).