1319005 - 2.04 - 4.96% sessionrestore_no_auto_restore / tpaint / ts_paint (windows8-64) regression on push 964223d37a2e59a28c4ee1df5a808252442ea94f (Thu Nov 3 2016)

Status: RESOLVED WONTFIX

(Firefox :: Untriaged, defect)

Description

[Alison Shiue]

Talos has detected a Firefox performance regression from push 964223d37a2e59a28c4ee1df5a808252442ea94f. As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

  5%  tpaint windows8-64 pgo e10s                             217.36 -> 228.15
  2%  tpaint windows8-64 pgo                                  223.16 -> 228.11
  2%  ts_paint windows8-64 pgo e10s                           729.58 -> 744.58
  2%  sessionrestore_no_auto_restore windows8-64 pgo e10s     641.88 -> 655


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=3976

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

[Alison Shiue]

This regression only shows obviously on PGO build.

I did bisect on try server, and found this regression looks like caused by push 2cf0ccb1a85c.
https://treeherder.mozilla.org/#/jobs?repo=try&author=[e-mail]&fromchange=284b60fd7c67

Here is the compare result: https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=b954787b1e11a512bc2c7cb2270870c650d26e25&newProject=try&newRevision=6a5688412523f62a13de1044debb46c607232827&framework=1&showOnlyImportant=0

Aurora also shows some symptoms since this patch already merged into aurora.(https://treeherder.mozilla.org/perf.html#/alerts?id=4129)

Hi Till, as you are the patch author, and  , can you take a look at this and determine what is the root cause? Thanks!

Comment 2

[Till Schneidereit [:till]]

There really isn't anything I can do here, I'm afraid. The only non-test code that's changed by this patch is here: https://hg.mozilla.org/mozilla-central/rev/2cf0ccb1a85c#l2.16

That change causes strictly fewer instructions to be run for the same scenario. At least before compiler optimizations. That this only really shows up in PGO builds lets me conclude that the change triggers slightly changed code generation which for this test makes the outcome worse. As I said, there just isn't anything I (or anybody else) can do about this.

Backing out the change also isn't an option as it's a correctness fix.

Comment 3

[Alison Shiue]

Thanks for the reply. 
Since there is nothing we can do, mark this bug as wontfix.