All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=597648

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg0JyRrwoM


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

android-nexus6

=== Auto-CCing suspected CL author abhishek.ka@samsung.com ===

Hi abhishek.ka@samsung.com, 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 : Handling transition animation between non-interpolable values.
Author  : abhishek.ka
Commit description:
  
In current transition animations, for any lengthsize css property
irrespective of interpolable values animation does start.
This patch will check whether the values are interpolable and decides
as follows:
i) If values are interpoable, start animation,
ii) else, hard set the respective length property to the element .
BUG= 513127 

Review URL: https://codereview.chromium.org/1732323003

Cr-Commit-Position: refs/heads/master@{#382801}
Commit  : 4d344058522e01d85162fcd1a9e5f4b747db9383
Date    : Wed Mar 23 06:48:44 2016


===== TESTED REVISIONS =====
Revision                Mean Value  Std. Dev.   Num Values  Good?
chromium@382796         22.466513   0.351476    8           good
chromium@382799         22.519844   0.278691    8           good
chromium@382800         22.549257   0.135953    5           good
chromium@382801         23.917014   0.580893    5           bad         <-
chromium@382802         24.022036   0.811915    8           bad

Bisect job ran on: android_nexus5_perf_bisect
Bug ID: 597648

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --also-run-disabled-tests smoothness.key_silk_cases
Test Metric: mean_frame_time/mean_frame_time
Relative Change: 6.41%
Score: 99.5

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus5_perf_bisect/builds/3532
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9016740551450334112


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=597648

| 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!

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


===== SUSPECTED CL(s) =====
Subject : Handling transition animation between non-interpolable values.
Author  : abhishek.ka
Commit description:
  
In current transition animations, for any lengthsize css property
irrespective of interpolable values animation does start.
This patch will check whether the values are interpolable and decides
as follows:
i) If values are interpoable, start animation,
ii) else, hard set the respective length property to the element .
BUG= 513127 

Review URL: https://codereview.chromium.org/1732323003

Cr-Commit-Position: refs/heads/master@{#382801}
Commit  : 4d344058522e01d85162fcd1a9e5f4b747db9383
Date    : Wed Mar 23 06:48:44 2016


===== TESTED REVISIONS =====
Revision                Mean Value  Std. Dev.   Num Values  Good?
chromium@382796         21.472771   0.28817     5           good
chromium@382800         21.627679   0.38421     5           good
chromium@382801         25.418229   0.146991    5           bad         <-
chromium@382802         25.527629   0.282825    5           bad
chromium@382804         25.544293   0.382891    5           bad

Bisect job ran on: android_nexus7_perf_bisect
Bug ID: 597648

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --also-run-disabled-tests smoothness.key_silk_cases
Test Metric: mean_frame_time/mean_frame_time
Relative Change: 18.96%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus7_perf_bisect/builds/2868
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9016732565789198528


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=597648

| 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!

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


===== SUSPECTED CL(s) =====
Subject : Handling transition animation between non-interpolable values.
Author  : abhishek.ka
Commit description:
  
In current transition animations, for any lengthsize css property
irrespective of interpolable values animation does start.
This patch will check whether the values are interpolable and decides
as follows:
i) If values are interpoable, start animation,
ii) else, hard set the respective length property to the element .
BUG= 513127 

Review URL: https://codereview.chromium.org/1732323003

Cr-Commit-Position: refs/heads/master@{#382801}
Commit  : 4d344058522e01d85162fcd1a9e5f4b747db9383
Date    : Wed Mar 23 06:48:44 2016


===== TESTED REVISIONS =====
Revision                Mean Value  Std. Dev.   Num Values  Good?
chromium@382794         20.166336   0.139342    5           good
chromium@382799         20.517843   0.481617    5           good
chromium@382800         20.563193   0.474084    5           good
chromium@382801         23.708064   0.27903     5           bad         <-
chromium@382802         23.68895    0.414525    5           bad
chromium@382804         23.775943   0.349857    5           bad
chromium@382813         23.600079   0.308781    5           bad
chromium@382831         23.961421   0.455002    5           bad

Bisect job ran on: android_webview_aosp_perf_bisect
Bug ID: 597648

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --also-run-disabled-tests smoothness.key_silk_cases
Test Metric: mean_frame_time/mean_frame_time
Relative Change: 18.82%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_webview_aosp_perf_bisect/builds/84
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9016740565137039728


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=597648

| 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!

This caused 200% regression in ChromiumPerf/android-nexus9/smoothness.key_silk_cases/mean_frame_time!

Started looking into the regression.
Seems like http://jsfiddle.net/xLuvC/1/ is failing from key_silk_cases.
Will update frequently in this bug.

I find it hard to believe the code added in https://codereview.chromium.org/1732323003 adds a whole 3ms of computation per frame.
I think we need to compare traces (about:tracing) of the performance test with and without that patch to try to find out where that extra time is being spent.

When i tried running the key_silk_cases test suite. Only few were passing the test(around 11).
With and without patch i observe the following readings for mean_frame_time:
With patch: 40.8 ms
Without patch: 41.154 ms

Tested on:
Nexus 7 - Cyanogen mode.

Deleted: data.xlsx 8.5 KB

data.xlsx 8.5 KB Download

The data sheet of mean frame time for key_silk_cases are only for 10 tests out of 17 tests present in the key_silk_cases testsuite.
Remaining 7 tests requires special permission to run locally.(raises CloudStortageError).
Could you please point out which particular testcase is degraded so that i can debug further.

Alan could you respond to #10?

Alancutter@ ping. Could you respond to #10

Alancutter@ : Gentle ping..!

Ryan, can you help Abhishek per comment 10?

Following up on #15, assigning to Ryan to help regarding comment #10.

Please reassign back to abhishek.ka@samsung.com once you've provided the required information.

Apologies that I missed this. Abishek, I've granted you access to the test data.

Friendly perf-sheriff ping, any update on this?

Another perf sheriff ping: abhishek.ka, any updates?

My apologies, typo in the email address.
@alancutter, abhishek doesn't have edit bugs access. Can you help here?

abhishek.ka, it looks like the permission issue was resolved. Can you take another look at this regression, please?

Apologies for leaving this one for so long.

This drop in mean_frame_time is due to the transition animation no longer being triggered in the http://jsfiddle.net/xLuvC/1/ test case after https://codereview.chromium.org/1732323003. See  crbug.com/616072  for bug details.

This bug was fixed in https://codereview.chromium.org/2027933002 which corresponds with the recovery of the specific perf test case.

Deleted: Screenshot from 2016-10-06 09:23:15.png 270 KB

	Screenshot from 2016-10-06 09:23:15.png 270 KB View Download