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

Issue 755765 link

Starred by 1 user

Issue metadata

Status: WontFix
Merged: issue 733502
Owner:
Last visit > 30 days ago
Closed: Jan 2018
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

10.1%-13.3% regression in blink_perf.layout at 492961:493377

Project Member Reported by ellenpli@google.com, Aug 15 2017

Issue description

See the link to graphs below.
 
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Aug 15 2017

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=755765

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=da46758e841882c50af283f02603ed68f882c331f6466e731bdeaa83e47e94e3


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

android-nexus6
android-webview-nexus6
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Aug 16 2017

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

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

Hi tyoshino@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 : Takeshi Yoshino
  Commit : e688f87c18ee1a1bdec979a1153edc74c087db24
  Date   : Thu Aug 10 06:10:10 2017
  Subject: Stop propagating unused data to WebAssociatedURLLoaderClient

Bisect Details
  Configuration: android_nexus6_perf_bisect
  Benchmark    : blink_perf.layout
  Metric       : fixed-grid-lots-of-stretched-data/fixed-grid-lots-of-stretched-data
  Change       : 7.64% | 208.062009162 -> 192.170651348

Revision             Result                  N
chromium@493262      208.062 +- 4.76982      6      good
chromium@493291      209.475 +- 3.50165      6      good
chromium@493299      211.059 +- 3.62791      6      good
chromium@493303      208.829 +- 4.832        6      good
chromium@493304      208.891 +- 3.88196      6      good
chromium@493305      195.435 +- 4.22689      6      bad       <--
chromium@493306      194.438 +- 5.70477      6      bad
chromium@493320      186.239 +- 5.5435       6      bad
chromium@493377      192.171 +- 8.0695       6      bad

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

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8971155374123720192


For feedback, file a bug with component Speed>Bisection
Owner: ----
Status: Available (was: Assigned)
Kicked the bisect again.

I don't think the patch caused such a regression. It just reduced unnecessary data propagation in a function call chain.
Project Member

Comment 6 by 42576172...@developer.gserviceaccount.com, Aug 18 2017

Mergedinto: 733502
Status: Duplicate (was: Available)

Comment 7 by nmduc@chromium.org, Aug 18 2017

Cc: nmduc@chromium.org
Status: Available (was: Duplicate)
Just testing bisect, and somehow this bug gets merge into my bug. Reopening this.
nmduc@: Sorry, I can't get it. What's the intention of reopening? It looked the original bisect result was false and the re-run caught revision 493319 which was actual cause.

Comment 9 by nmduc@chromium.org, Aug 22 2017

Because, I did not run the same bisect settings as your bisect. I dont know why they merge but I dont see blink_perf.layout->fixed-grid-lots-of-stretched-data/fixed-grid-lots-of-stretched-data in my bisect.
Pretty sure the bisect in #4 blamed nmduc's change, but didn't post the right comment when it duped bugs in #6. Re-running.
Project Member

Comment 12 by 42576172...@developer.gserviceaccount.com, Sep 22 2017

Cc: nmduc@google.com
Owner: nmduc@google.com
Status: Assigned (was: Available)

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

Hi nmduc@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 : Minh-Duc Nguyen
  Commit : ed4698554ab3fdfdc4da6f4e60d9994c2e57616c
  Date   : Thu Aug 10 07:38:46 2017
  Subject: Group optimization on Computed Style

Bisect Details
  Configuration: android_webview_nexus6_aosp_perf_bisect
  Benchmark    : blink_perf.layout
  Metric       : fixed-grid-lots-of-data/fixed-grid-lots-of-data
  Change       : 15.48% | 236.72152956 -> 200.078435905

Revision             Result                  N
chromium@492960      236.722 +- 5.63836      6      good
chromium@493151      231.26 +- 3.45053       6      good
chromium@493246      232.315 +- 6.87709      6      good
chromium@493294      223.804 +- 9.70574      6      good
chromium@493318      228.334 +- 10.5575      6      good
chromium@493319      195.331 +- 13.0491      6      bad       <--
chromium@493320      195.259 +- 10.6174      6      bad
chromium@493321      193.73 +- 7.54701       6      bad
chromium@493324      197.358 +- 6.43137      6      bad
chromium@493330      189.092 +- 7.94405      6      bad
chromium@493341      200.078 +- 8.04267      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

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8967812532098524848


For feedback, file a bug with component Speed>Bisection
Status: WontFix (was: Assigned)

Sign in to add a comment