1409411 - 13.32% tsvgr_opacity (osx-10-10) regression on push 2c77560532c96f99b60ae5b28aec4e597f1f30e7 (Mon Oct 16 2017)

Status: RESOLVED FIXED

(Core :: Web Painting, defect)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

Talos has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?changeset=2c77560532c96f99b60ae5b28aec4e597f1f30e7

As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

 13%  tsvgr_opacity summary osx-10-10 opt e10s     319.05 -> 361.54


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=10035

On the page above you can see an alert for each affected platform as well as a link to a graph showing the history of scores for this test. There is also a link to a treeherder page showing the Talos jobs in a pushlog format.

To learn more about the regressing test(s), please see: https://wiki.mozilla.org/Buildbot/Talos/Tests

For information on reproducing and debugging the regression, either on try or locally, see: https://wiki.mozilla.org/Buildbot/Talos/Running

*** Please let us know your plans within 3 business days, or the offending patch(es) will be backed out! ***

Our wiki page outlines the common responses and expectations: https://wiki.mozilla.org/Buildbot/Talos/RegressionBugsHandling

Comment 1

[Joel Maher ( :jmaher ) (UTC -8)]

Hi Matt, in bug 1406727  the build was broken, as the patch author I am asking you to help figure out which of the changes caused the perf regression:
https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&group_state=expanded&filter-searchStr=osx%20build&tochange=2c77560532c96f99b60ae5b28aec4e597f1f30e7&fromchange=1498abcee79d232125fb44846849160ec3c1fa51

Comment 2

[Adam Morris]

I'm unsure as to how to solve this regression. I have only patched two bugs. Any help in resolving this issue would be great.

Comment 3

[Matt Woodrow (:mattwoodrow)]

Did the regression go away again? Bug 1407815 got backed out, so if it's still present, then almost certainly bug 1406727.

Comment 4

[Matt Woodrow (:mattwoodrow)]

It doesn't look like it did, so bug 1406727 seems like the likely culprit.

This isn't overly surprising, since it's a diagnostic patch with some overhead. It's only enabled for development builds (so won't affect release), and I plan on backing it out once we have results.

Given that, I think we might just have to live with this for now.

Comment 5

[Ionuț Goldan [:igoldan]]

(In reply to Matt Woodrow (:mattwoodrow) from comment #4)
> It doesn't look like it did, so bug 1406727 seems like the likely culprit.
> 
> This isn't overly surprising, since it's a diagnostic patch with some
> overhead. It's only enabled for development builds (so won't affect
> release), and I plan on backing it out once we have results.

Very likely. These Linux regressions also showed up. Fortunately, their previous builds weren't broken.

== Change summary for alert #10035 (as of October 17 2017 03:21 UTC) ==

Regressions:

 11%  tsvgr_opacity summary linux64 pgo e10s       329.00 -> 366.77
 11%  tsvgr_opacity summary linux64 opt e10s       366.87 -> 406.34

They appeared after this patch - consisting of bug 1406727 and bug 1407815, landed:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=1498abcee79d232125fb44846849160ec3c1fa51&tochange=940a0c98e46c363d97e9cc8ef448faec9ff725f2

Comment 6

[Matt Woodrow (:mattwoodrow)]

This should be backed out now, have the numbers returned to their previous values?

Comment 7

[Joel Maher ( :jmaher ) (UTC -8)]

are you referring to these changes:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=393eddd4e496e2f067eab91262359c5aa959a9e5&tochange=0f783930e1b231009e96f7cfbbdb2f8324718cef

I had assumed that was related to bug 1406727.  Is there a timeline for when the backout occurred or a bug to reference?

Comment 8

[Matt Woodrow (:mattwoodrow)]

That one got backed out, fixed version is https://hg.mozilla.org/mozilla-central/rev/99b6831627b1

It was all related to bug 1406727, the regression was adding some diagnostic code, and the above changeset removes it again.

Comment 9

[Matt Woodrow (:mattwoodrow)]

Closing this since it looks like the numbers are fixed now, please re-open if that's not the case.