New issue
Advanced search Search tips

Issue 727568 link

Starred by 1 user

Issue metadata

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



Sign in to add a comment

26.4% regression in blink_perf.layout at 475321:475333

Project Member Reported by toyoshim@chromium.org, May 30 2017

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=727568

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDghouM-QkM


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

android-webview-nexus5X
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, May 30 2017


=== BISECT JOB RESULTS ===
Bisect failed unexpectedly

Bisect was aborted with the following:
  Step('Bisecting revision.Post bisect results (3)') failed with return_code 1


Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.layout
  Metric       : floats_10_1000/floats_10_1000

Revision             Result                  N
chromium@475320      2636.34 +- 15.2529      6      good
chromium@475327      2635.33 +- 8.81174      6      good
chromium@475330      3413.72 +- 661.09       6      bad
chromium@475333      3642.86 +- 301.741      6      bad

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

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8978185421520562032

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5293039058157568


| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Speed>Bisection.  Thank you!

=== BISECT JOB RESULTS ===
Bisect failed unexpectedly

Bisect was aborted with the following:
  Step('Bisecting revision.Post bisect results (3)') failed with return_code 1


Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.layout
  Metric       : floats_10_1000/floats_10_1000

Revision             Result                  N
chromium@475320      2636.34 +- 15.2529      6      good
chromium@475327      2635.33 +- 8.81174      6      good
chromium@475330      3413.72 +- 661.09       6      bad
chromium@475333      3642.86 +- 301.741      6      bad

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

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8978136801834671312

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5293039058157568


| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Speed>Bisection.  Thank you!

=== BISECT JOB RESULTS ===
Bisect failed for unknown reasons

Please contact the team (see below) and report the error.


Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.layout
  Metric       : floats_10_1000/floats_10_1000


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

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8978087808613964192

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5863786122051584


| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Speed>Bisection.  Thank you!
Owner: yhirano@chromium.org
Status: Assigned (was: Untriaged)
chromium@475327      2635.33 +- 8.81174      6      good
chromium@475330      3413.72 +- 661.09       6      bad

475328 https://chromium-review.googlesource.com/512382
475329 https://codereview.chromium.org/2902113004
475330 https://chromium-review.googlesource.com/517447/

Bisect could not finish, but the last one, yhirano@'s change is most likely in the last possible range?
My change is a refactoring, so I don't think it regressed the performance, but let's wait for the bisect results...
Or this test is just unstable?
Looks performance gets back to the original level at later revision 475672, and I could say this was a noise.
https://chromeperf.appspot.com/group_report?bug_id=727568

Please feel free to close as WontFix if you feel this is a false alert.
Other changes are for histograms.xml and autofill feature, and should not affect layout performance too.
Bisects were failing due to  crbug.com/728594 , restarting this.
Cc: yhirano@chromium.org

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

Hi yhirano@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 : Yutaka Hirano
  Commit : 7d8a4d97b96a7e1f398424c840668fc17989f5d0
  Date   : Mon May 29 10:27:30 2017
  Subject: Remove FetchContext::LoadingTaskRunner

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.layout
  Metric       : floats_10_1000/floats_10_1000
  Change       : 28.54% | 2657.84633333 -> 3416.50616667

Revision             Result                  N
chromium@475327      2657.85 +- 11.0372      6      good
chromium@475329      2621.92 +- 11.8571      6      good
chromium@475330      3416.51 +- 586.219      6      bad       <--

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

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8978023549551306864

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5863786122051584


| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Speed>Bisection.  Thank you!

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

Suspected Commit
  Author : Yutaka Hirano
  Commit : 7d8a4d97b96a7e1f398424c840668fc17989f5d0
  Date   : Mon May 29 10:27:30 2017
  Subject: Remove FetchContext::LoadingTaskRunner

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.layout
  Metric       : floats_10_1000/floats_10_1000

Revision             Result                  N
chromium@475320      2651.24 +- 13.359       6      good
chromium@475327      2648.83 +- 11.4669      6      good
chromium@475329      2620.42 +- 15.3152      6      good
chromium@475330      3463.83 +- 681.477      6      bad       <--
chromium@475333      3557.72 +- 650.36       6      bad

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

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8977938072044364448

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5896005356093440


| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Speed>Bisection.  Thank you!
Status: WontFix (was: Assigned)
My change could affect loading performance (though I think it's unlikely), but the test is not related to loading. Closing as WontFix.
Owner: e...@chromium.org
Status: Assigned (was: WontFix)
Reopening and assigning to Emil, owner of blink_perf.layout tests. Emil, this is a strange bug. Here is the graph: https://chromeperf.appspot.com/group_report?bug_id=727568

* It only happens on WebView
* The graph seems to go up and down 30% quite often.
* The bisects consistently reproduce a 30% regression at r475330, which yhirano believes is unrelated. The subsequent dip in the graph back to normal levels does not seem to be related to yhirano's changes (no land/revert pattern).

Is it worth investigating further? Is this not a good test case for webview?

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

Suspected Commit
  Author : Yutaka Hirano
  Commit : 7d8a4d97b96a7e1f398424c840668fc17989f5d0
  Date   : Mon May 29 10:27:30 2017
  Subject: Remove FetchContext::LoadingTaskRunner

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.layout
  Metric       : floats_10_1000/floats_10_1000
  Change       : 14.09% | 2880.7505 -> 3286.58133333

Revision             Result                  N
chromium@475300      2880.75 +- 113.139      6      good
chromium@475325      2651.9 +- 11.3968       6      good
chromium@475329      2617.76 +- 11.2765      6      good
chromium@475330      5301.31 +- 4205.0       6      bad       <--
chromium@475331      3386.06 +- 288.66       6      bad
chromium@475332      3565.58 +- 711.265      6      bad
chromium@475338      3404.47 +- 546.176      6      bad
chromium@475350      3286.58 +- 539.92       6      bad

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

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8977641073937826128

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=4933570025488384


| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Speed>Bisection.  Thank you!

Comment 20 by e...@chromium.org, Jun 6 2017

Owner: ----
Status: Untriaged (was: Assigned)
As this is only affecting a single configuration, given the lack of reliable bisect information, and the extreme variability of the metric this really is something that the webview team needs to look into. No one on the layout team knows enough about how either android or the webview works to debug this further.

From our perspective this is a WontFix at the moment.
Owner: sullivan@chromium.org
Status: Assigned (was: Untriaged)
+1 for WontFix.

Also, I should say sorry about I completely forgot to add test owners to CC list in the last perf sheriff period.

Let me assign sullivan just in case.
Status: WontFix (was: Assigned)
Labels: Performance-Tradeoff

Sign in to add a comment