1420369 - stylo: 6.21 - 6.54% remote-tsvg / remote-twitter (android-6-0-armv8-api16, android-7-1-armv8-api16) regression on push db56323cd08f4883e4824199b441a3141be655e5 (Thu Nov 23 2017)

Status: NEW

(Core :: CSS Parsing and Computation, defect, P2)

Description

[Ionuț Goldan [:igoldan]]

We have detected an autophone (Android) regression from push:

https://hg.mozilla.org/integration/autoland/pushloghtml?changeset=db56323cd08f4883e4824199b441a3141be655e5

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

Regressions:

  7%  remote-tsvg summary android-6-0-armv8-api16 opt      84.13 -> 89.63
  6%  remote-twitter summary android-7-1-armv8-api16 opt   774.07 -> 823.84
  6%  remote-tsvg summary android-7-1-armv8-api16 opt      102.84 -> 109.23


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=10710

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 jobs in a pushlog format.

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

Comment 1

[Ionuț Goldan [:igoldan]]

:m_kato Looks like enabling Stylo on Android brought some perf regressions. Were these expected?

Comment 2

[Makoto Kato [:m_kato]]

(In reply to Ionuț Goldan [:igoldan], Performance Sheriffing from comment #1)
> :m_kato Looks like enabling Stylo on Android brought some perf regressions.
> Were these expected?

Yes and no.  But we need figure out why remote-tsvg is slow.  And, why does remote-twitter by android-6-0-armv8-api16 doesn't hit this?

Comment 3

[Chris Peterson [:cpeterson]]

P2 because we should address this regression before we ship stylo-android.

Comment 4

[Makoto Kato [:m_kato]]

Gecko profiler doesn't support android...

Comment 5

[Makoto Kato [:m_kato]]

twitter and gearflowers.svg have a long style attribute.

Comment 6

[Liz Henry (:lizzard) (relman/hg->git project)]

Tracking for 59 since we're aiming to enable this feature on nightly 59.

Comment 7

[Andrew Overholt [:overholt]]

Are we still planning on shipping Stylo on Android in 59?

Comment 8

[Chris Peterson [:cpeterson]]

(In reply to Andrew Overholt [:overholt] from comment #7)
> Are we still planning on shipping Stylo on Android in 59?

Yes. Makoto thinks the NEON optimization patch in bug 586838 will fix this Android perf regression.

Comment 9

[Andrew Overholt [:overholt]]

[Tracking Requested - why for this release]:
Apparently stylo on android is slipping to 60

Comment 10

[Julien Cristau [:jcristau]]

AIUI we're not blocking stylo-android on this for 60, so I'll move this to fix-optional.

Comment 11

[James Willcox (:snorp) ([e-mail]) (he/him)]

Since bug 586838 landed, has this gotten better?

Comment 12

[Ionuț Goldan [:igoldan]]

(In reply to James Willcox (:snorp) ([e-mail]) from comment #11)
> Since bug 586838 landed, has this gotten better?

Not really.

Comment 13

[Makoto Kato [:m_kato]]

(In reply to James Willcox (:snorp) ([e-mail]) from comment #11)
> Since bug 586838 landed, has this gotten better?

not enough.  I guess that this is into cssparser and rust.

Comment 14

[David Bolter [:davidb] (NeedInfo me for attention)]

Looking at this during QF triage, Panos is this actionable? Maybe we can get a profile?

Comment 15

[Panos Astithas (he/him) [:past] (please ni?)]

(In reply to Makoto Kato [:m_kato] from comment #13)
> (In reply to James Willcox (:snorp) ([e-mail]) from comment #11)
> > Since bug 586838 landed, has this gotten better?
> 
> not enough.  I guess that this is into cssparser and rust.

Do you have a profile that shows this? If not, could you get one and share it here? Emilio or Xidorn could take a look if it is cssparser related.

Comment 16

[Makoto Kato [:m_kato]]

https://perfht.ml/2Ks8P1I

Since we don't use rust 1.27+ as build environment, walking stack on rust code isn't correct.  So profiler cannot get better data.

Comment 17

[Makoto Kato [:m_kato]]

(this data uses https://searchfox.org/mozilla-central/source/testing/talos/talos/tests/svg_static/gearflowers.svg)

Comment 18

[Liz Henry (:lizzard) (relman/hg->git project)]

Looks like this is still not actionable.

Comment 19

[Liz Henry (:lizzard) (relman/hg->git project)]

Letting this fall out of regression triage since it's had several wontfixes in a row.