New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 758281 link

Starred by 3 users

Issue metadata

Status: WontFix
Owner:
Closed: Sep 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

10.4% regression in smoothness.top_25_smooth at 494991:495262

Project Member Reported by nzolghadr@chromium.org, Aug 23 2017

Issue description

See the link to graphs below.
 
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Aug 23 2017

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=758281

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=00ac0cedab8d6eb41a47c4c1b984fc4a553f9a1cc7c0e71d2834f495c1a5b280


Bot(s) for this bug's original alert(s):

chromium-rel-mac-retina
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Aug 23 2017

Cc: sahel@chromium.org
Owner: sahel@chromium.org
Status: Assigned (was: Untriaged)

=== Auto-CCing suspected CL author sahel@chromium.org ===

Hi sahel@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : sahel
  Commit : 955c32907a2c9921065fa628e0b8b5b28a7afe6c
  Date   : Thu Aug 17 14:56:44 2017
  Subject: Change tests that fail with wheel scroll latching and async wheel on.

Bisect Details
  Configuration: mac_retina_perf_bisect
  Benchmark    : smoothness.top_25_smooth
  Metric       : first_gesture_scroll_update_latency/ESPN
  Change       : 10.74% | 1.7525 -> 1.94078571429

Revision             Result                  N
chromium@494990      1.7525 +- 1.47136       14      good
chromium@495126      1.68156 +- 1.09833      9       good
chromium@495160      1.65643 +- 1.03465      14      good
chromium@495169      1.68889 +- 1.1042       9       good
chromium@495171      1.68356 +- 1.08621      9       good
chromium@495172      1.64633 +- 1.08002      9       good
chromium@495173      1.76136 +- 1.29294      14      bad       <--
chromium@495177      1.99133 +- 1.51935      9       bad
chromium@495194      1.87378 +- 1.32693      9       bad
chromium@495262      1.94079 +- 1.97296      14      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=ESPN smoothness.top_25_smooth

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8970449006108581504


For feedback, file a bug with component Speed>Bisection

Comment 4 by sahel@chromium.org, Aug 25 2017

Status: WontFix (was: Assigned)
955c32907a2c9921065fa628e0b8b5b28a7afe6c changes SyntheticSmoothMoveGesture::ForwardMouseWheelInputEvents to not send wheel events with zero deltas. The smoothness benchmark uses GPU benchmarking which uses the changed method. This change filters out the first synthetic wheel event on Mac  and that's why there is a 10% regression on the first scroll update latency.

I mark it as won't fix since a)The variance of the data is a lot bigger than the regression. b)The new data is the more accurate one since measuring scroll update latency for an event with zero delta is useless.
Project Member

Comment 5 by 42576172...@developer.gserviceaccount.com, Aug 25 2017

Cc: kraynov@chromium.org
 Issue 758980  has been merged into this issue.
Project Member

Comment 6 by 42576172...@developer.gserviceaccount.com, Aug 29 2017

 Issue 759711  has been merged into this issue.
Project Member

Comment 7 by 42576172...@developer.gserviceaccount.com, Aug 29 2017

 Issue 759583  has been merged into this issue.

Comment 8 by sahel@chromium.org, Aug 29 2017

Status: Assigned (was: WontFix)
Delta values for wheel events are the floored_toward_zero values of the original float values which are calculated in SyntheticSmoothMoveGesture::ComputeNextMoveSegment.
Sometimes the discrete delta value is zero while the original delta is not. This is the cause for this regression since the duration for the event is calculated based on the original values and even though zero-delta events are not sent their non-zero durations causes the first sent wheel event to have a greater timestamp.
Project Member

Comment 9 by bugdroid1@chromium.org, Sep 8 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8f2fa60e3dff739013a59734c46207b511d0efad

commit 8f2fa60e3dff739013a59734c46207b511d0efad
Author: sahel <sahel@chromium.org>
Date: Fri Sep 08 21:32:35 2017

Make synthetic smooth wheel forwarding similar to touch.


Synthetic_smooth_move doesn't generate wheel with zero deltas, it assigns
proper timestamps to events which are used in generated ui events, and it
uses original float delta values rather than making them discrete.

Bug:  758281 
Change-Id: I3b0c46cb6a95f8b93097c1bbc28b41648b364977
Reviewed-on: https://chromium-review.googlesource.com/656088
Reviewed-by: Dave Tapuska <dtapuska@chromium.org>
Commit-Queue: Sahel Sharifymoghaddam <sahel@chromium.org>
Cr-Commit-Position: refs/heads/master@{#500691}
[modify] https://crrev.com/8f2fa60e3dff739013a59734c46207b511d0efad/content/browser/renderer_host/input/mouse_wheel_event_queue.cc
[modify] https://crrev.com/8f2fa60e3dff739013a59734c46207b511d0efad/content/browser/renderer_host/input/synthetic_gesture_controller_unittest.cc
[modify] https://crrev.com/8f2fa60e3dff739013a59734c46207b511d0efad/content/browser/renderer_host/input/synthetic_gesture_target_aura.cc
[modify] https://crrev.com/8f2fa60e3dff739013a59734c46207b511d0efad/content/browser/renderer_host/input/synthetic_smooth_move_gesture.cc
[modify] https://crrev.com/8f2fa60e3dff739013a59734c46207b511d0efad/content/browser/renderer_host/input/synthetic_smooth_move_gesture.h

Cc: sullivan@google.com
 Issue 763391  has been merged into this issue.

Comment 11 by sahel@chromium.org, Sep 14 2017

Status: WontFix (was: Assigned)

Sign in to add a comment