New issue
Advanced search Search tips

Issue 607961 link

Starred by 0 users

Issue metadata

Status: Duplicate
Merged: issue 607598
Owner:
Closed: May 2016
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

27.8% regression in blink_perf.canvas at 389513:389602

Project Member Reported by qyears...@chromium.org, Apr 29 2016

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=607961

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICglMiFswoM


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

android-nexus9
Project Member

Comment 2 by 42576172...@developer.gserviceaccount.com, Apr 29 2016

Cc: tedc...@chromium.org
Owner: tedc...@chromium.org

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

Hi tedchoc@chromium.org, the bisect results pointed to your CL below as possibly
causing a regression. Please have a look at this info and see whether
your CL be related.


===== BISECT JOB RESULTS =====
Status: completed


===== SUSPECTED CL(s) =====
Subject : Remove URL path fading from tabbed mode.
Author  : tedchoc
Commit description:
  
BUG= 521178 

Review URL: https://codereview.chromium.org/1916933002

Cr-Commit-Position: refs/heads/master@{#389560}
Commit  : caa39378ce5da0d3ff191a06c9e162e3b3db7948
Date    : Mon Apr 25 21:04:18 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@389512  2091.01  69.671   5  good
chromium@389557  2055.59  104.923  5  good
chromium@389559  2053.34  100.703  5  good
chromium@389560  1659.89  113.567  5  bad    <--
chromium@389563  1460.97  131.76   5  bad
chromium@389569  1563.78  183.79   5  bad
chromium@389580  1584.54  267.811  5  bad
chromium@389602  1745.55  136.904  5  bad

Bisect job ran on: android_nexus9_perf_bisect
Bug ID: 607961

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --also-run-disabled-tests blink_perf.canvas
Test Metric: draw-static-canvas-2d-to-hw-accelerated-canvas-2d/draw-static-canvas-2d-to-hw-accelerated-canvas-2d
Relative Change: 16.52%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus9_perf_bisect/builds/1765
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9014026336755444656


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5059408703258624

| 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 Tests>AutoBisect.  Thank you!
Owner: qyears...@chromium.org
The change linked above removed animations from the browser UI.  If anything, it should improve performance for anything that happens around page load.  I'm very suspect that is the source of any perf regressions.
Thanks for the feedback :-)

Currently it looks like this alert is only on android-nexus9, but the bisect bot did reproduce this, so it appears to be a real thing.

Next step: see if it's a thing on other platforms, and see if bisect can reproduce it multiple times.
Don't see any regressions for this test on similar android platforms:
https://chromeperf.appspot.com/report?sid=5e8cb68dbd69eec70b1d9d745e0e52c1107384033bed60437826e6eff427783b

Maybe this is just noise. Waiting for more bisect results before closing.
Last bisect fails, kick off another one in https://chromeperf.appspot.com/buildbucket_job_status/9013379009525917616

===== BISECT JOB RESULTS =====
Status: completed


===== SUSPECTED CL(s) =====
Subject : Remove URL path fading from tabbed mode.
Author  : tedchoc
Commit description:
  
BUG= 521178 

Review URL: https://codereview.chromium.org/1916933002

Cr-Commit-Position: refs/heads/master@{#389560}
Commit  : caa39378ce5da0d3ff191a06c9e162e3b3db7948
Date    : Mon Apr 25 21:04:18 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@389500  2155.13  58.8011  5  good
chromium@389551  2169.92  64.8991  5  good
chromium@389558  2116.85  79.1943  8  good
chromium@389559  2098.88  142.995  8  good
chromium@389560  1680.66  106.002  5  bad    <--
chromium@389561  1538.23  153.335  5  bad
chromium@389564  1594.1   79.0241  5  bad
chromium@389577  1583.33  116.068  5  bad
chromium@389602  1617.68  128.54   6  bad

Bisect job ran on: android_nexus9_perf_bisect
Bug ID: 607961

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --also-run-disabled-tests blink_perf.canvas
Test Metric: draw-static-canvas-2d-to-hw-accelerated-canvas-2d/draw-static-canvas-2d-to-hw-accelerated-canvas-2d
Relative Change: 24.81%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus9_perf_bisect/builds/1779
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9013379009525917616


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5855536256385024

| 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 Tests>AutoBisect.  Thank you!
Hi tedchoc, just noticed something I didn't notice before -- for the blink perf test blink_perf.canvas draw-static-canvas-2d-to-hw-accelerated-canvas-2d (the one for which we reproduced the regression with the bisect bots), the metric improved when you enabled URL path fading on Android (r383200). See: https://chromeperf.appspot.com/report?sid=f68cff9f80081310e5a659ea30d83caf121ec856f658b3e9b03fea341a20d3d7&start_rev=381996&end_rev=392749

So it looks like the results for that blink perf test are affected by URL path fading in tabbed mode.


The memory (vm_private_dirty_final_renderer) regression alerts associated with this bug may be unrelated or noise; started bisect jobs on those alerts.
Hmm...I'm inclined to mark this as won't fix then.

URL path fading does a non trivial amount of additional work in the browser animating off the path portion of the URL.  There is no way it makes rendering any faster.  There must be some weird interaction that this test is benefiting from, but I know it's not a general performance gain for traditional page loads. 
Status: WontFix (was: Assigned)
Alright -- then I agree with marking it as WontFix. Thanks!
Project Member

Comment 11 by 42576172...@developer.gserviceaccount.com, May 21 2016

Mergedinto: 607598
Status: Duplicate (was: WontFix)

===== BISECT JOB RESULTS =====
Status: completed


===== SUSPECTED CL(s) =====
Subject : Check for 0-length files in ResourceExtractor
Author  : agrieve
Commit description:
  
Also moves from an on-disk version timestamp file to SharedPreferences
(should be faster).

BUG= 606413 

Review URL: https://codereview.chromium.org/1920893003

Cr-Commit-Position: refs/heads/master@{#390068}
Commit  : b28607d4c72261657ea9e24ea9195bc23a472b3f
Date    : Wed Apr 27 14:44:16 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@390062  35085.2  51.6039  6  good
chromium@390065  35074.2  132.587  5  good
chromium@390067  35026.8  189.894  5  good
chromium@390068  35694.0  121.744  5  bad    <--
chromium@390073  35732.0  117.658  5  bad
chromium@390084  35668.4  80.4568  5  bad
chromium@390106  35739.4  144.808  5  bad

Bisect job ran on: android_nexus6_perf_bisect
Bug ID: 607961

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --also-run-disabled-tests page_cycler.intl_ar_fa_he
Test Metric: vm_private_dirty_final_browser/vm_private_dirty_final_browser
Relative Change: 1.92%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus6_perf_bisect/builds/2195
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9012114809444901136


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5783470647803904

| 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 Tests>AutoBisect.  Thank you!

Sign in to add a comment