All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=603257

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICguIXDqQoM


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

android-galaxy-s5

Bisect results did not post for some reason. Pasted below. r386456 is the skia deps roll. I'll re-run to narrow it down.

 ===== BISECT JOB RESULTS ===== 
 Status: started  
 ===== TESTED REVISIONS =====
 Revision Mean Value Std. Dev. Num Values Good?
 chromium@386441 147.998656 6.030655 5 good
 chromium@386453 151.136175 8.491447 5 good
 chromium@386455 149.617073 3.966814 5 good
 chromium@386456 210.036185 11.37898 5 bad
 chromium@386459 223.372568 10.342808 5 bad
 chromium@386464 210.512005 11.940161 5 bad
 chromium@386486 222.375296 11.934874 5 bad
 chromium@386530 223.357588 11.635013 5 bad

  Bisect job ran on: android_s5_perf_bisect

 Bug ID: 603257

  Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --also-run-disabled-tests smoothness.gpu_rasterization.tough_filters_cases

 Test Metric: frame_times/http___static.bobdo.net_Analog_Clock.svg

 Relative Change: 50.92% Score: 0

  Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_s5_perf_bisect/builds/597 Job details: https://chromeperf.appspot.com/buildbucket_job_status/9015372388711517760   Not what you expected? We'll investigate and get back to you! https://chromeperf.appspot.com/bad_bisect?try_job_id=603257  | 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!

New bisect over skia roll: https://chromeperf.appspot.com/buildbucket_job_status/9015306457640100896

Per comment #4, something in the Skia DEPS roll regressed performance here.  Assigning to borenet@ for further triage.

Moving this nonessential bug to the next milestone.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

This issue has been moved once and is lower than Pri-1. Removing the milestone.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Perf sheriff ping

borenet@: is there anything in that skia roll (r386456) that looks like it may have caused a performance regression in this benchmark?

Not sure how I missed this the first time around. +GPU folks. At a glance it looks like it might be one of:
https://chromium.googlesource.com/skia.git/+/29feef80de0af74eb24b703d6675aea1bc17e655
https://chromium.googlesource.com/skia.git/+/fef28606d2c51be627be3a656581882e31945c03

This is a pretty significant regression, and from current test values, it seems not to have recovered. Can someone please prioritize work on root-cause and resolution?

Perhaps robertphillips@ should own this issue?

For the Android Galaxy S5 only two tests in this block have data back that far: Analog_Clock and letmespellitoutforyou.

Only Analog_Clock test shows the performance regression.

Here is the status of the devices that have data for Analog_Clock back that far (and that I looked at - I skipped several of the desktop Windows configs):

G-S5  - shows the regression
N5    - no perf change
N6    - shows perf improvement
N9    - hard to tell but the times went back down so probably no change

rel-max-hdd - shows perf improvement
rel-mac-retina - no change
rel-mac10 - no change
rel-max11 - no change
rel-win8  - no change
rel-win10 - no change
linux-rel - no change

So, of the Android devices, 1 shows a performance regression, 1 shows a performance improvement and 2 show no change. Except for rel-mac-hdd all the desktops show no change and rel-mac-hdd shows a performance improvement.

Adding senorblanco who is listed as the test owner.

What are the next steps here? Does the improvement on some devices mean we want to live with the regression on S5?

Re: #15: thanks for digging into this! (BTW, it can very useful in cases like these to post the a link to the relevant graphs, for posterity.)

I think we can live with that regression on a single (old) device.