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

Issue 750869 link

Starred by 1 user
Status: WontFix
Owner:
wangxianzhu@chromium.org
Closed: Aug 2017
Cc:
█ fsam...@chromium.org
wangxianzhu@chromium.org
xhw...@chromium.org
blundell@chromium.org
dalecur...@chromium.org
█ a.suc...@samsung.com
majidvp@chromium.org
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

15.2%-40.4% regression in blink_perf.paint at 490260:490447

Project Member Reported by majidvp@google.com, Jul 31 2017 
See the link to graphs below.
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Jul 31 2017 

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=750869

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=24e9d2a40be081958a35f0b8a3f7bf232013f8b884da92978b87a782d48aa003


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

chromium-rel-win8-dual
win-high-dpi
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Aug 1 2017

Cc: blundell@chromium.org
Owner: blundell@chromium.org

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

Hi blundell@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 : Colin Blundell
  Commit : bf44268201f947d844fb2667ca2b618d1d4213d8
  Date   : Fri Jul 28 13:08:05 2017
  Subject: [Identity Service] Add GetAccounts() API

Bisect Details
  Configuration: win_8_perf_bisect
  Benchmark    : blink_perf.paint
  Metric       : large-table-background-change-with-visible-collapsed-borders/large-table-background-change-with-visible-collapsed-borders
  Change       : 4.12% | 41.7235 -> 40.0064166667

Revision             Result                   N
chromium@490325      41.7235 +- 0.503684      6      good
chromium@490356      43.9118 +- 0.465597      6      good
chromium@490371      44.1626 +- 0.683071      9      good
chromium@490375      44.2018 +- 0.316203      9      good
chromium@490376      40.189 +- 0.166954       6      bad       <--
chromium@490377      40.8844 +- 0.516239      6      bad
chromium@490379      40.3779 +- 0.120294      4      bad
chromium@490386      40.1252 +- 0.163116      6      bad
chromium@490447      40.0064 +- 0.354914      6      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 blink_perf.paint

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8972519328739097616


For feedback, file a bug with component Speed>Bisection
Project Member

Comment 5 by 42576172...@developer.gserviceaccount.com, Aug 2 2017

Cc: a.suc...@samsung.com
Owner: a.suc...@samsung.com

=== Auto-CCing suspected CL author a.suchit@samsung.com ===

Hi a.suchit@samsung.com, 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 : Suchit Agrawal
  Commit : 5b3329fdca90d0449fc8d4419bee11320e6c88c1
  Date   : Fri Jul 28 12:34:12 2017
  Subject: //media/renderers should depend on //media/base.

Bisect Details
  Configuration: win_8_perf_bisect
  Benchmark    : blink_perf.paint
  Metric       : large-table-background-change-with-invisible-collapsed-borders/large-table-background-change-with-invisible-collapsed-borders
  Change       : 0.72% | 89.9931666667 -> 89.3475833333

Revision             Result                   N
chromium@490325      89.9932 +- 0.83721       6      good
chromium@490356      96.5053 +- 0.731995      6      good
chromium@490364      99.4015 +- 0.33762       6      good
chromium@490368      97.2433 +- 0.487298      6      good
chromium@490369      89.255 +- 0.222461       6      bad       <--
chromium@490370      88.5131 +- 0.158848      6      bad
chromium@490371      88.8558 +- 0.360474      6      bad
chromium@490386      88.6052 +- 0.420125      6      bad
chromium@490447      89.3476 +- 0.326868      6      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 blink_perf.paint

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8972443652698675568


For feedback, file a bug with component Speed>Bisection

Comment 6 by a.suchit@chromium.org, Aug 2 2017

Cc: dalecur...@chromium.org xhw...@chromium.org
In this Commit : 5b3329fdca90d0449fc8d4419bee11320e6c88c1, 2 dependencies are added for //media/renderers target.

I do not feel that I would give any performance issue.

Comment 8 by sullivan@chromium.org, Aug 2 2017

Owner: ----
Results look pretty noisy; kicked off a bisect on a wider range.
Project Member

Comment 9 by 42576172...@developer.gserviceaccount.com, Aug 2 2017

Cc: fsam...@chromium.org
Owner: fsam...@chromium.org

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

Hi fsamuel@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 : Fady Samuel
  Commit : e65ab0a11b73c8eea0139c0757005432d6dbb0b6
  Date   : Fri Jul 28 17:47:25 2017
  Subject: viz: Move surface sync unit test dependencies to components/viz/test

Bisect Details
  Configuration: win_8_perf_bisect
  Benchmark    : blink_perf.paint
  Metric       : large-table-background-change-with-invisible-collapsed-borders/large-table-background-change-with-invisible-collapsed-borders
  Change       : 8.11% | 89.5181666667 -> 96.7755833333

Revision             Result                   N
chromium@489984      89.5182 +- 0.551354      6      good
chromium@490327      91.5148 +- 0.539936      6      good
chromium@490413      89.4597 +- 1.00991       6      good
chromium@490435      90.1308 +- 0.967417      6      good
chromium@490446      89.7718 +- 0.981441      6      good
chromium@490451      89.2589 +- 0.999408      6      good
chromium@490452      90.7858 +- 0.838665      6      good
chromium@490453      99.4465 +- 1.49315       6      bad       <--
chromium@490454      97.1149 +- 1.38495       6      bad
chromium@490456      99.0271 +- 0.716449      6      bad
chromium@490498      96.5151 +- 0.744086      6      bad
chromium@490669      96.7756 +- 0.740364      6      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 blink_perf.paint

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8972369576007345376


For feedback, file a bug with component Speed>Bisection

Comment 10 by fsam...@chromium.org, Aug 2 2017 

I don't see how moving unit tests could cause a regression.

Comment 11 by sullivan@chromium.org, Aug 2 2017

Owner: wangxianzhu@chromium.org
wangxianzhu, as owner of blink_perf.paint, any ideas why the test results are so unstable? The graph looks a bit noisy, but the bisects seem to be bimodal.
Project Member

Comment 14 by 42576172...@developer.gserviceaccount.com, Aug 17 2017

Cc: isandrk@google.com
Owner: isandrk@google.com
Status: Assigned (was: Untriaged)

=== Auto-CCing suspected CL author isandrk@google.com ===

Hi isandrk@google.com, 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 : Ivan Sandrk
  Commit : 34836cc78e7eb47c4f0fe708fe369171181184f8
  Date   : Fri Jul 28 11:02:32 2017
  Subject: US International (PC) Keyboard layout was missing resourceId

Bisect Details
  Configuration: win_8_perf_bisect
  Benchmark    : blink_perf.paint
  Metric       : large-table-background-change-with-zero-width-collapsed-borders/large-table-background-change-with-zero-width-collapsed-borders
  Change       : 1.85% | 90.1436666667 -> 88.4759166667

Revision             Result                   N
chromium@490325      90.1437 +- 0.411554      6      good
chromium@490341      90.7923 +- 1.11938       6      good
chromium@490349      89.8113 +- 1.25096       6      good
chromium@490353      89.3111 +- 0.562567      6      good
chromium@490355      89.6848 +- 0.702762      6      good
chromium@490356      88.948 +- 0.231424       6      bad       <--
chromium@490386      88.9681 +- 0.710253      6      bad
chromium@490447      88.4759 +- 0.525144      6      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 blink_perf.paint

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8970996731823357088


For feedback, file a bug with component Speed>Bisection

Comment 15 by wangxianzhu@chromium.org, Aug 17 2017

Cc: -isandrk@google.com
Owner: wangxianzhu@chromium.org
The #14 bisect seems not to reproduce the regression. Let's see other bisect results.

Comment 16 by wangxianzhu@chromium.org, Aug 21 2017

Status: WontFix (was: Assigned)
The other bisect job seems to fail to reproduce the regression or find the culprit. The bisect jobs show much smaller regression (if it exists) than those shown by the waterfall perf bots, so I wonder if any configuration change that happened on the waterfall perf bots (but not on try bots) caused the regression. Also the waterfall bots showed recovery around r495788:495808 of the benchmarks.
►

Sign in to add a comment