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

Issue 787206 link

Starred by 2 users

Issue metadata

Status: WontFix
Owner:
ooo
Closed: Jan 2018
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression

Blocking:
issue 731202



Sign in to add a comment

3.8%-3.9% regression in thread_times.key_mobile_sites_smooth at 517523:517669

Project Member Reported by benhenry@google.com, Nov 21 2017

Issue description

See the link to graphs below.
 
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Nov 21 2017

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=787206

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=817b75d2843ea6df1c0a11058a14ddae9b1f9dc708cbb205c3ed446b8df140f4


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

android-nexus5
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Nov 21 2017

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

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

Hi brianderson@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 : Brian Anderson
  Commit : feaa47c68a99a0817c2b945801fdc9d2873fc6b6
  Date   : Fri Nov 17 23:40:21 2017
  Subject: gpu: Remove gpu's dependency on ui/latency_info.

Bisect Details
  Configuration: android_nexus5_perf_bisect
  Benchmark    : thread_times.key_mobile_sites_smooth
  Metric       : thread_browser_cpu_time_per_frame/http___nytimes.com_
  Change       : 4.50% | 3.80125633426 -> 3.97214138609

Revision             Result                    N
chromium@517522      3.80126 +- 0.0623836      6      good
chromium@517596      3.80632 +- 0.0576942      6      good
chromium@517606      3.82165 +- 0.0348709      6      good
chromium@517611      3.82181 +- 0.0485262      6      good
chromium@517613      3.82575 +- 0.0375367      6      good
chromium@517614      3.97464 +- 0.0727921      6      bad       <--
chromium@517615      3.95755 +- 0.0821384      6      bad
chromium@517633      3.97006 +- 0.0424525      6      bad
chromium@517669      3.97214 +- 0.0891478      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...nytimes.com. thread_times.key_mobile_sites_smooth

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8962359527994049616


For feedback, file a bug with component Speed>Bisection
Project Member

Comment 4 by 42576172...@developer.gserviceaccount.com, Nov 21 2017

Cc: hjd@google.com
 Issue 787407  has been merged into this issue.
brianderson: can you take a look at this bug?
Blocking: 731202
Cc: sadrul@chromium.org nzolghadr@chromium.org tdres...@chromium.org
Status: WontFix (was: Assigned)
Looking at all alerts with the patch in the alert range:
https://chromeperf.appspot.com/group_report?rev=517614

There are more (5x) and bigger (2x) wins in the GPU service than the losses we see in the browser, so closing as WontFix.

It amazes me that there's a measurable perf impact of between .1 to .3 ms to update LatencyInfos. +tdresser, sadrul, and nzolghadr.


Sign in to add a comment