1466830 - 2.6 - 3.01% sessionrestore_many_windows (linux64-qr, windows10-64-qr) regression on push 5b15326286d466b5cf4889160cc09b59bdde08fd (Mon Jun 4 2018)

Status: RESOLVED WONTFIX

(Core :: Graphics: WebRender, defect)

Description

[Ionuț Goldan [:igoldan]]

Talos has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/autoland/pushloghtml?changeset=5b15326286d466b5cf4889160cc09b59bdde08fd

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

Regressions:

  3%  sessionrestore_many_windows windows10-64-qr opt e10s stylo     2,883.62 -> 2,970.42
  3%  sessionrestore_many_windows linux64-qr opt e10s stylo          1,483.83 -> 1,522.42


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=13635

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]]

Here are the Gecko profiles from Linux 64 QR:

before: https://perf-html.io/from-url/https%3A%2F%2Fqueue.taskcluster.net%2Fv1%2Ftask%2Ffl4sUCeORnuLBE8_Z469rg%2Fruns%2F0%2Fartifacts%2Fpublic%2Ftest_info%2Fprofile_sessionrestore_many_windows.zip
 
after: https://perf-html.io/from-url/https%3A%2F%2Fqueue.taskcluster.net%2Fv1%2Ftask%2FX84x1eLkTCu9nFZv_7jFRA%2Fruns%2F0%2Fartifacts%2Fpublic%2Ftest_info%2Fprofile_sessionrestore_many_windows.zip

Comment 2

[Kartikaya Gupta (email:[e-mail])]

Here's the data for the windows version from the try pushes in bug 1465058:

WR revision                                 data points
67f8715a719aeae408de76906a0798191ba6de70    3420 3457 3458
f13c425afaff909ade9de2482bdb280bb11d554e    3457
0d753e6b5682b9dcba96faa7210a1733dcb4b713    3479
5e4f257d9f3cdb5691ac62f3affe2e9189d66447    3728 3748 3766 3797 3804 2991 3002 2990
8e697f8cb1f1aab2e5f6b9b903eb7191340b10c5    2995 2989 2980 2974 2988 2995 2994

IT looks like there was a jump from the 3400's to the 3700's when going from 0d753e6b5682b9dcba96faa7210a1733dcb4b713 -> 5e4f257d9f3cdb5691ac62f3affe2e9189d66447. And then sometime on May 31 when we got a new version of m-c the numbers dropped down to the 3000's, so presumably that was the improvement from bug 1418202.

This means the WR regression range is this:

*   5e4f257d Auto merge of #2790 - pcwalton:pathfinder-crash-fix, r=gw3583
|\
| * aee31d68 Write `Blank` into the glyph key cache if glyph dimensions can't be fetched from the font.
* 72c09b7b Auto merge of #2784 - gw3583:border-groove-ridge-double, r=kvark
* 9985a998 Address review comments.
* abd6bed7 Defer building of instances unless render task is stale.
* 2f13b4c0 Support double, groove, ridge borders in new border brush path.

I think we can discount #2790 since pathfinder is not enabled with gecko, so it must be from #2784. Most likely the shader changes in 2f13b4c0 resulted in a slightly higher shader compilation startup cost.

So this is a wontfix, as we need the shader changes for correctness, and the perf hit is relatively small. Forwarding needinfo to glenn in case he has any comments, but otherwise I think this is a wontfix.

Comment 3

[Glenn Watson [:gw]]

Your analysis above sounds correct to me.