Electrolysis/Release Criteria/Scrolling: Difference between revisions

(updated tests data)
 
(8 intermediate revisions by 4 users not shown)
Line 1: Line 1:
Scrolling responsiveness must not regress when we turn on e10s.
Scrolling responsiveness must not regress when we turn on e10s.


RASCI:
= RASCI =


* Responsible: ?
* Responsible: kats
* Accountable: RyanVM
* Accountable: RyanVM
* Supporting:  
* Supporting: avih, jmaher, graphics team
* Consulted: release management, jmaher
* Consulted: release management, jmaher
* Informed: cpeterson, elan
* Informed: cpeterson, elan


Criteria:
= Metrics =


=== FX_REFRESH_DRIVER_SYNC_SCROLL_FRAME_DELAY_MS ===
== FX_REFRESH_DRIVER_SYNC_SCROLL_FRAME_DELAY_MS ==


* "e10s appears to be much better here"
* Regressed, see {{bug|1255175}}
* from billm: Analysis was done in {{bug|1228147}} but this was not a valid dataset because it considers users from outside the experiment.


=== talos tp5o_scroll ===
== talos tp5o_scroll ==


* e10s comparison validated: kats
* e10s comparison validated: kats
Line 24: Line 23:
* {{bug|1179735}} - tp5o_scroll regression on windows compared to non-e10s
* {{bug|1179735}} - tp5o_scroll regression on windows compared to non-e10s
* Note: reflects synchronous scroll only, so meaningless where APZ kicks in. Currently APZ handles only mouse wheel, so the test currently reflects autoscroll/arrows/page-up-down/etc.
* Note: reflects synchronous scroll only, so meaningless where APZ kicks in. Currently APZ handles only mouse wheel, so the test currently reflects autoscroll/arrows/page-up-down/etc.
* [https://bugzilla.mozilla.org/show_bug.cgi?id=1253860#c9 kats comments on apz and this test]


=== talos tscrollx ===
== talos tscrollx ==


* e10s comparison validated: kats
* e10s comparison validated: kats
* Current e10s diff: worse except linux
* Current e10s diff: mixed
** Linux: 30% better
** Linux: 30% better
** Windows 7/8: 7% worse
** OS X: 17% better
** OS X: 17% worse
** Windows 8: 7% better
** Windows 7: 7% worse
** Windows XP: 50% worse
** Windows XP: 50% worse
* {{bug|1249976}} - tscrollx regressed 6% with e10s on December 25th
* {{bug|1249976}} - tscrollx regressed 6% with e10s on December 25th
* {{bug|1174772}} - 2.9% win7 tscrollx regression on e10s only
* {{bug|1174772}} - 2.9% win7 tscrollx regression on e10s only
* Note: reflects synchronous scroll only, so meaningless where APZ kicks in. Currently APZ handles only mouse wheel, so the test currently reflects autoscroll/arrows/page-up-down/etc.
* Note: reflects synchronous scroll only, so meaningless where APZ kicks in. Currently APZ handles only mouse wheel, so the test currently reflects autoscroll/arrows/page-up-down/etc.
* Note: tscrollx improved with e10s + no apz 6%-40% so this regression appears to be apz related.
= Bugs =
<bugzilla>
{
    "blocks": "1251388",
    "resolution": "---",
    "include_fields": "id, summary, whiteboard, keywords, assigned_to"
}
</bugzilla>

Latest revision as of 21:27, 9 March 2016

Scrolling responsiveness must not regress when we turn on e10s.

RASCI

  • Responsible: kats
  • Accountable: RyanVM
  • Supporting: avih, jmaher, graphics team
  • Consulted: release management, jmaher
  • Informed: cpeterson, elan

Metrics

FX_REFRESH_DRIVER_SYNC_SCROLL_FRAME_DELAY_MS

talos tp5o_scroll

  • e10s comparison validated: kats
  • Current e10s diff: worse except linux
    • Linux: 17% better
    • All Windows and OS X: 45-110% worse
  • bug 1179735 - tp5o_scroll regression on windows compared to non-e10s
  • Note: reflects synchronous scroll only, so meaningless where APZ kicks in. Currently APZ handles only mouse wheel, so the test currently reflects autoscroll/arrows/page-up-down/etc.
  • kats comments on apz and this test

talos tscrollx

  • e10s comparison validated: kats
  • Current e10s diff: mixed
    • Linux: 30% better
    • OS X: 17% better
    • Windows 8: 7% better
    • Windows 7: 7% worse
    • Windows XP: 50% worse
  • bug 1249976 - tscrollx regressed 6% with e10s on December 25th
  • bug 1174772 - 2.9% win7 tscrollx regression on e10s only
  • Note: reflects synchronous scroll only, so meaningless where APZ kicks in. Currently APZ handles only mouse wheel, so the test currently reflects autoscroll/arrows/page-up-down/etc.
  • Note: tscrollx improved with e10s + no apz 6%-40% so this regression appears to be apz related.

Bugs

No results.

0 Total; 0 Open (0%); 0 Resolved (0%); 0 Verified (0%);