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

Issue 750822 link

Starred by 3 users

Issue metadata

Status: WontFix
Merged: issue 750923
Owner:
Closed: Aug 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

1.9%-471.1% regression in loading.desktop at 490023:490345

Project Member Reported by majidvp@google.com, Jul 31 2017

Issue description

Looking at the affected metrics it seems to be a paint/raster related regression which affects

loading.desktop/timeToFCP and timeToFMP
smoothness.tough_webgl_cases
rasterize_and_record_micro


 
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=750822

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=fa0c3ba57cbe0d11d38ed724689b0c5154105367c694c5d56fe35a5fa98c03c9


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

chromium-rel-mac-retina
chromium-rel-mac12
chromium-rel-mac12-mini-8gb
chromium-rel-win7-gpu-ati
chromium-rel-win7-gpu-nvidia
linux-release
win-high-dpi

=== BISECT JOB RESULTS ===
NO Perf regression found

Bisect Details
  Configuration: mac_10_12_mini_8gb_perf_bisect
  Benchmark    : loading.desktop
  Metric       : cpuTimeToFirstMeaningfulPaint_avg/pcv1-cold/FarsNews

Revision             Result                  N
chromium@490022      3242.39 +- 4874.49      20      good
chromium@490245      2190.84 +- 6761.8       21      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=FarsNews loading.desktop

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8972524225352733792


For feedback, file a bug with component Speed>Bisection

Comment 4 by zmo@chromium.org, Aug 3 2017

Cc: vmi...@chromium.org ericrk@chromium.org

Comment 6 by vmi...@chromium.org, Aug 24 2017

Cc: jbroman@chromium.org yukishiino@chromium.org
Components: Blink>Bindings
There is a significant regression in "blink_perf.bindings / set-attribute" which could regress things like the WebGL & Canvas benchmarks.

https://chromeperf.appspot.com/report?sid=084c8e63f64bf642f433530983e7159c09859fdc77d9d149d0a3aade00dabea3&rev=490345
Project Member

Comment 8 by 42576172...@developer.gserviceaccount.com, Aug 24 2017

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

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

Hi jmadill@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 : Jamie Madill
  Commit : 4c19a8a82b2e93f936ea38c4772ce0182cd3dc2a
  Date   : Wed Jul 26 16:44:57 2017
  Subject: D3D11: Update cached dynamically recompiled programs.

Bisect Details
  Configuration: winx64ati_perf_bisect
  Benchmark    : smoothness.tough_animation_cases
  Metric       : frame_times/css_value_type_shadow.html?api_css_animations_N_0316
  Change       : 30.73% | 25.5609501025 -> 33.4156938037

Revision                              Result                   N
chromium@490131                       25.561 +- 20.824         9       good
chromium@490148                       23.6293 +- 20.4612       9       good
chromium@490152                       23.9759 +- 20.0302       9       good
chromium@490154                       19.2911 +- 4.50028       6       good
chromium@490155                       25.2472 +- 19.4662       9       good
chromium@490155,angle@17d270311e      25.8422 +- 24.6809       14      good
chromium@490155,angle@4c56c607f4      28.2183 +- 17.7047       9       good
chromium@490155,angle@b6664925cb      31.1856 +- 10.873        9       good
chromium@490155,angle@4c19a8a82b      33.4225 +- 1.25829       21      bad       <--
chromium@490156                       33.418 +- 1.00013        14      bad
chromium@490164                       33.498 +- 0.703752       6       bad
chromium@490196                       33.4887 +- 0.42554       6       bad
chromium@490260                       33.4157 +- 0.856062      9       bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=css.value.type.shadow.html.api.css.animations.N.0316 smoothness.tough_animation_cases

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8970364620671134000


For feedback, file a bug with component Speed>Bisection
Uh oh. Question, why does the result go from bad to good with my CL?
Owner: vmi...@chromium.org
The ANGLE regression was fixed in https://chromium-review.googlesource.com/531798 and the graphs have already recovered. Needs further triage for the other regressions. 
Project Member

Comment 11 by 42576172...@developer.gserviceaccount.com, Aug 25 2017

Cc: agrieve@chromium.org
Owner: agrieve@chromium.org

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

Hi agrieve@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 : Andrew Grieve
  Commit : 767a6f5a890a6028a1d8a8f30cea242b568e6cee
  Date   : Fri Jul 28 18:11:29 2017
  Subject: Android: Add Zip Overhead breakdown to resource_sizes.py

Bisect Details
  Configuration: winx64nvidia_perf_bisect
  Benchmark    : blink_perf.bindings
  Metric       : set-attribute/set-attribute
  Change       : 13.04% | 852.127659 -> 740.976761564

Revision             Result                  N
chromium@490218      852.128 +- 3.80316      6      good
chromium@490383      855.891 +- 5.69566      6      good
chromium@490465      839.533 +- 8.00877      9      good
chromium@490466      849.744 +- 10.0413      6      bad       <--
chromium@490467      849.864 +- 14.8519      9      bad
chromium@490468      849.0 +- 4.22416        6      bad
chromium@490471      857.825 +- 4.1953       6      bad
chromium@490476      854.668 +- 7.91506      6      bad
chromium@490486      813.005 +- 358.247      9      bad
chromium@490506      736.029 +- 3.45797      6      bad
chromium@490547      740.977 +- 7.12472      6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests blink_perf.bindings

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8970364159315918672


For feedback, file a bug with component Speed>Bisection
Cc: sullivan@chromium.org
Owner: vmi...@chromium.org
#11 is a strange bisect result.  Un-assigning Andrew and running again on a narrower range.

sullivan: The large regression is the drop from ~854 to ~736, but the bisect narrowed in on a change which was within the noise +/- 10 noise.
Project Member

Comment 14 by 42576172...@developer.gserviceaccount.com, Aug 25 2017

Mergedinto: 750923
Status: Duplicate (was: Assigned)

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

Suspected Commit
  Author : Hans Wennborg
  Commit : d2c91228a51bdf37ae3b2e501fb53c0528f1629c
  Date   : Fri Jul 28 20:11:05 2017
  Subject: win: Set is_clang=true by default

Bisect Details
  Configuration: winx64nvidia_perf_bisect
  Benchmark    : blink_perf.bindings
  Metric       : set-attribute/set-attribute
  Change       : 13.37% | 853.013931152 -> 739.003396065

Revision             Result                  N
chromium@490476      853.014 +- 10.4747      6      good
chromium@490491      853.481 +- 2.3588       6      good
chromium@490493      861.817 +- 9.9443       6      good
chromium@490494      743.362 +- 8.10611      6      bad       <--
chromium@490495      738.962 +- 8.26932      6      bad
chromium@490499      739.211 +- 6.24375      6      bad
chromium@490506      739.003 +- 4.9182       6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests blink_perf.bindings

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8970330513512243568


For feedback, file a bug with component Speed>Bisection
Owner: h...@chromium.org
Status: Assigned (was: Duplicate)
Un-duping based on Hans' request in  Issue 750923 .  That issue was fixed, but this regression in blink_perf.bindings has not recovered.

Comment 16 by h...@chromium.org, Aug 25 2017

Owner: vmi...@chromium.org
The clang=1 patch was reverted last Friday. The "All graphs for this bug" link above doesn't show me any graphs. Can you confirm whether performance recovered or not? If not, it's probably not clang related.
Cc: dtu@chromium.org simonhatch@chromium.org
Re #16: I triaged the blink_perf alert that bisected in #14 back into this bug, the link should show it now. It does seem to have recovered with the revert on Friday.

Re #12: Thanks for the report, Victor! Simon, Dave, the bisect in #11 may be a good test case for pinpoint/multiple regressions.


Project Member

Comment 18 by 42576172...@developer.gserviceaccount.com, Aug 25 2017

Cc: vmiura@google.com k...@chromium.org
 Issue 759131  has been merged into this issue.
Status: WontFix (was: Assigned)
This has recovered, so closing.

Sign in to add a comment