1410908 - 1.46 - 4.91% tart / tsvgx (linux64, osx-10-10, windows10-64, windows7-32) regression on push 63c91c79579b9992e674ec17b6008432589d0201 (Mon Oct 23 2017)

Status: RESOLVED WONTFIX

(Core :: Web Painting, defect)

Description

[Ionuț Goldan [:igoldan]]

Talos has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/autoland/pushloghtml?changeset=63c91c79579b9992e674ec17b6008432589d0201

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

Regressions:

  5%  tsvgx summary windows10-64 opt e10s     228.29 -> 239.51
  4%  tsvgx summary windows10-64 pgo e10s     212.16 -> 221.37
  4%  tsvgx summary osx-10-10 opt e10s        387.38 -> 403.08
  4%  tart summary linux64 opt e10s           4.95 -> 5.14
  3%  tsvgx summary windows7-32 opt e10s      462.60 -> 476.83
  3%  tsvgx summary linux64 opt e10s          366.41 -> 376.86
  2%  tsvgx summary linux64 pgo e10s          347.17 -> 355.63
  2%  tart summary windows7-32 opt e10s       4.89 -> 5.00
  2%  tart summary windows7-32 pgo e10s       3.91 -> 3.99
  1%  tart summary windows10-64 opt e10s      4.15 -> 4.21

Improvements:

  4%  tsvgr_opacity summary linux64 opt e10s     408.79 -> 391.63
  4%  tsvgr_opacity summary linux64 pgo e10s     367.54 -> 352.35
  3%  tsvgr_opacity summary osx-10-10 opt e10s   360.57 -> 350.35


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=10122

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

[Ionuț Goldan [:igoldan]]

:mattwoodrow Bug 1404181 caused tart & tsvgx regressions on all platforms. Can you look over fixing these regressions?

Comment 2

[Matt Woodrow (:mattwoodrow)]

This is an expected regression as part of landing retained display lists. We added extra book keeping code to handle the new functionality, but that hurts performance for first load (especially when the feature is turned off, as it is currently).

Please see the dev.platform intent to ship thread here: https://groups.google.com/forum/#!topic/mozilla.dev.platform/jSuLiuXqs8w

We're hoping to get this enabled properly within the next few weeks, and get big wins on real sites, so we believe this regression to be justified.

I'll take another look at the tsvgx regression too, since that's a bit higher than I would have liked.

Comment 3

[Mike Taylor [:miketaylr]]

It sounds like we want to live with the tart regression (ni? jet just to confirm).

Should we update the title of this bug to only be about tsvgx?

Comment 4

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

isn't this supposed to be preffed off, are we testing talos with the pref=true?

Comment 5

[Markus Stange [:mstange]]

The speedups are preffed off, but some behavior changed in all configurations (the "extra book keeping code"), and that's what's causing the slowdown.

Comment 6

[Jet Villegas (inactive)]

(In reply to Mike Taylor [:miketaylr] (58 Regression Engineering Owner) from comment #3)
> It sounds like we want to live with the tart regression (ni? jet just to
> confirm).

We expect that these first-paint regressions will be offset by nth-paint improvements once we ship.

> 
> Should we update the title of this bug to only be about tsvgx?

I also expected a small regression for tsvgx, but I did not expect any improvements to tsvgr with this change. 

Matt: did your investigation from comment 2 result in any changes?

Comment 7

[Matt Woodrow (:mattwoodrow)]

Unfortunately not. Miko had a good look and our initial idea (reverting the nsDisplayLayerEventRegions changes when the pref is disabled) didn't end up helping.

The tsvgr changes are indeed unexpected, but a nice bonus!

Comment 8

[Ionuț Goldan [:igoldan]]

Should we resolve this as WONTFIX or wait for extra investigations?

Comment 9

[Mike Taylor [:miketaylr]]

(This is enabled in Nightly only, updating affected releases)

Comment 10

[Matt Woodrow (:mattwoodrow)]

Yeah, I think we can WONTFIX this for now.