1643598 - 2.56 - 3.28% compiler_metrics num_static_constructors (linux64) regression on push f3a829091fef65cf49209ed374688cdd6d5753ed (Tue June 2 2020)

Status: RESOLVED DUPLICATE of bug 1644404

(Core :: Graphics: CanvasWebGL, defect)

Description

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

Perfherder has detected a build_metrics performance regression from push 801f886cf1b608a2cb697ffa02067ef1d42c9d1f(https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=48dbf3178e91a05a27b2e61b1a45ac9ff54f0135&tochange=801f886cf1b608a2cb697ffa02067ef1d42c9d1f). As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

3% compiler_metrics num_static_constructors linux64 debug base-toolchains-clang 183.00 -> 189.00
3% compiler_metrics num_static_constructors linux64 debug base-toolchains 292.50 -> 300.00

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

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

Looks like this is compiler metrics and can be backfilled. I will update the bug as the culprit reveals itself.

Comment 2

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

[deleted due to wrong bug]

Comment 3

[Mike Hommey [:glandium]]

The graph shows a regression in May that doesn't appear to have an associated regression. Similarly, I randomly picked another job (Linux64 debug on autoland), checked the num_static_constructors graph for it, and there are plenty of visible regressions that are not marked with alerts.
https://treeherder.mozilla.org/perf.html#/graphs?highlightAlerts=1&series=autoland,1468482,1,2&timerange=1209600

Comment 4

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

Bebe, Ionut, you should probably also see this.

Comment 5

[Ionuț Goldan [:igoldan]]

(In reply to Mike Hommey [:glandium] from comment #3)

The graph shows a regression in May that doesn't appear to have an associated regression. Similarly, I randomly picked another job (Linux64 debug on autoland), checked the num_static_constructors graph for it, and there are plenty of visible regressions that are not marked with alerts.
https://treeherder.mozilla.org/perf.html#/graphs?highlightAlerts=1&series=autoland,1468482,1,2&timerange=1209600

This series alerts when its baseline changes its absolute (not percentage) value by 3. Those baseline changes increased just by 1. Thus, they didn't alert.
Problem here is the zoomed in graph is confusing: those regressions seem serious, when in fact they're hardly noticeable. But I guess this is the risk when a test like compiler_metrics num_static_constructors applies for absolute changes.

Comment 6

[Florin Strugariu [:Bebe]]

agree with Ionut

Comment 7

[Ryan VanderMeulen [:RyanVM]]

(In reply to Alexandru Ionescu (needinfo me) [:alexandrui] from comment #1)

Looks like this is compiler metrics and can be backfilled. I will update the bug as the culprit reveals itself.

Did this backfilling ever reveal the source of the regression?

Comment 8

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

Yeah, looks like Bug 1635956 is the culprit.

Comment 9

[Ryan VanderMeulen [:RyanVM]]

Hi David, can you please take a look at this static constructors regression?

Comment 10

[David Parks [:handyman]]

:)
Looks like there was a mix-up back when it was filed, probably due to the wrong component?