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

Issue 758230 link

Starred by 2 users

Issue metadata

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



Sign in to add a comment

11.5% regression in thread_times.key_silk_cases at 494610:494704

Project Member Reported by nzolghadr@chromium.org, Aug 23 2017

Issue description

See the link to graphs below.
 
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Aug 23 2017

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=758230

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=73d91d98d971214a941ff36d727802979a5d98158982b453d8837a89352dfc3a


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

android-one
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Aug 23 2017

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

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

Hi xidachen@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 : Xida Chen
  Commit : 177134823106e782420b7ce5150c2aa137014ee2
  Date   : Wed Aug 16 02:35:04 2017
  Subject: Use half-width GPU tiles on low-end android devices

Bisect Details
  Configuration: android_one_perf_bisect
  Benchmark    : thread_times.key_silk_cases
  Metric       : thread_renderer_compositor_cpu_time_per_frame/http___jsfiddle.net_xLuvC_1_show_
  Change       : 14.09% | 3.56463790961 -> 4.06703160871

Revision             Result                    N
chromium@494609      3.56464 +- 0.0560376      6      good
chromium@494657      3.55386 +- 0.0440732      6      good
chromium@494669      3.51718 +- 0.0611333      6      good
chromium@494672      3.51131 +- 0.0498725      6      good
chromium@494673      4.06521 +- 0.082022       6      bad       <--
chromium@494674      4.09471 +- 0.0493758      6      bad
chromium@494675      4.1055 +- 0.0854747       6      bad
chromium@494681      4.08855 +- 0.0359438      6      bad
chromium@494704      4.06703 +- 0.0703297      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 --story-filter=http...jsfiddle.net.xLuvC.1.show. thread_times.key_silk_cases

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8970455728336223376


For feedback, file a bug with component Speed>Bisection
Cc: flackr@chromium.org ericrk@chromium.org
ericrk@: this is likely legitimate since we reduce the tile size of be half. It impact low-end android device only.

Comment 5 by ericrk@chromium.org, Aug 23 2017

Yeah, we're probably going to have to make a call on whether any benefits outweigh these downsides, once we have better UMA data.
ericrk@: do we have an effective UMA to measure this? Note this is caused by our decision to cut tile size by half, if the tile_width > screen_width.
Project Member

Comment 7 by 42576172...@developer.gserviceaccount.com, Aug 25 2017

Cc: kraynov@chromium.org
 Issue 759035  has been merged into this issue.
Project Member

Comment 8 by 42576172...@developer.gserviceaccount.com, Aug 28 2017

 Issue 759612  has been merged into this issue.
Project Member

Comment 9 by 42576172...@developer.gserviceaccount.com, Aug 28 2017

 Issue 759695  has been merged into this issue.
Status: WontFix (was: Assigned)
 Issue 759611  has been merged into this issue.

Sign in to add a comment