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

Issue 771253 link

Starred by 1 user

Issue metadata

Status: Duplicate
Merged: issue 769548
Owner:
Closed: Oct 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

33.4% regression in blink_perf.parser at 504350:504406

Project Member Reported by majidvp@chromium.org, Oct 3 2017

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=771253

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=68b493dd5ddc3255aa0bf67321742cffdff3fa43c820295310bbbd1a7b676ff5


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

android-webview-nexus5X
Cc: phoglund@chromium.org
Owner: phoglund@chromium.org
Status: Assigned (was: Untriaged)

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

Hi phoglund@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 : Patrik Höglund
  Commit : b34fe684c6a5cd0bc0efacafadf8e934dc392a0f
  Date   : Tue Sep 26 13:25:00 2017
  Subject: Revert "Re-enabling Firefox interop test."

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.parser
  Metric       : html-parser-threaded/html-parser-threaded
  Change       : 0.39% | 2744.56971429 -> 2733.8977619

Revision             Result                  N
chromium@504349      2744.57 +- 30.5525      21      good
chromium@504350      4973.12 +- 85.622       6       bad       <--
chromium@504351      4961.57 +- 38.6998      6       bad
chromium@504353      4960.12 +- 54.7914      6       bad
chromium@504356      4950.79 +- 39.9127      6       bad
chromium@504363      4662.49 +- 2962.15      14      bad
chromium@504378      3353.17 +- 3673.8       14      bad
chromium@504406      2733.9 +- 48.7027       21      bad

Please refer to the following doc on diagnosing blink_perf regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/speed/benchmark_harnesses/blink_perf.md

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.parser

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8966732990368817536


For feedback, file a bug with component Speed>Bisection
Status: WontFix (was: Assigned)
The above CL disabled a browser_test, so it's not me.

I think it looks like the graph is recovering, so WontFix?
Status: Untriaged (was: WontFix)
I agree that bisect is not pointing to the right CL.

But the graph does not look like that it is recovering to me but it is just noisier. I kicked off another bisect with a wider range to see if that helps finding it out.
Project Member

Comment 7 by 42576172...@developer.gserviceaccount.com, Oct 17 2017

Cc: r...@opera.com
Owner: r...@opera.com
Status: Assigned (was: Untriaged)

=== Auto-CCing suspected CL author rune@opera.com ===

Hi rune@opera.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 : Rune Lillesveen
  Commit : da85874e0ca9635522b92bdb13efe57947c1a317
  Date   : Tue Sep 26 08:07:36 2017
  Subject: Update counters after style recalc, not layout.

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.parser
  Metric       : html-parser-threaded/html-parser-threaded
  Change       : 2.00% | 2863.7325 -> 2806.59183333

Revision             Result                  N
chromium@504300      2863.73 +- 23.9322      6      good
chromium@504305      2818.62 +- 24.5324      6      good
chromium@504308      2850.5 +- 12.7129       6      good
chromium@504309      2847.27 +- 19.3339      6      good
chromium@504310      2752.81 +- 25.9293      6      bad       <--
chromium@504320      2809.27 +- 9.15052      6      bad
chromium@504339      2821.53 +- 12.2196      6      bad
chromium@504378      2780.53 +- 33.3207      6      bad
chromium@504456      2806.59 +- 20.9609      6      bad

Please refer to the following doc on diagnosing blink_perf regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/speed/benchmark_harnesses/blink_perf.md

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.parser

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8965542376921019376


For feedback, file a bug with component Speed>Bisection

Comment 8 by r...@opera.com, Oct 17 2017

Owner: majidvp@chromium.org
The bisect data shown above looks bogus. 504310 which is marked as the first bad is actually the best result above. It's also a commit in the stable part of the graph. Also, looking at the graph, this stabilized later before starting to get noisy again.

Cc: -yukishiino@chromium.org majidvp@chromium.org
Owner: yukishiino@chromium.org
Graph is too noisy for bisect or the metric to be of any use. Assigning to yukishiino@chromium.org to either try to improve the metric here or just close
this issue.
Cc: yukishiino@chromium.org
Owner: jbroman@chromium.org
Defer to Jeremy.
Project Member

Comment 12 by 42576172...@developer.gserviceaccount.com, Oct 18 2017

Mergedinto: 769548
Status: Duplicate (was: Assigned)

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

Suspected Commit
  Author : Andrew Grieve
  Commit : 8a218af5c68598645675d76980984777c8429bb3
  Date   : Tue Sep 26 04:31:21 2017
  Subject: Android: Roll ndk to add CHROMIUM_CXX_TWEAK_INLINES

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.parser
  Metric       : html-parser-threaded/html-parser-threaded
  Change       : 1.29% | 2721.2815 -> 2756.46633333

Revision             Result                  N
chromium@504200      2721.28 +- 11.0258      6      good
chromium@504250      2753.72 +- 24.2049      6      good
chromium@504275      2778.17 +- 11.3463      6      good
chromium@504282      2784.04 +- 39.5995      9      good
chromium@504283      2807.11 +- 23.1735      6      bad       <--
chromium@504284      2811.84 +- 12.3207      6      bad
chromium@504285      2808.47 +- 28.6678      6      bad
chromium@504288      2811.78 +- 24.1514      6      bad
chromium@504300      2834.43 +- 18.0789      6      bad
chromium@504400      2746.63 +- 26.2315      6      bad
chromium@504600      2756.47 +- 14.0318      6      bad

Please refer to the following doc on diagnosing blink_perf regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/speed/benchmark_harnesses/blink_perf.md

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.parser

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8965454997751463968


For feedback, file a bug with component Speed>Bisection

Sign in to add a comment