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

Issue 739300 link

Starred by 2 users
Status: Duplicate
Merged: issue 737623
Owner: ----
Closed: Jul 2017
Cc:
vmi...@chromium.org
rmcilroy@chromium.org
█ rnep...@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

5.5%-97.6% regression in thread_times.simple_mobile_sites at 482599:482711

Project Member Reported by rmcilroy@chromium.org, Jul 5 2017 
See the link to graphs below.
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Jul 6 2017 


=== BISECT JOB RESULTS ===
NO Perf regression found

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : v8.mobile_infinite_scroll_tbmv2
  Metric       : v8-gc-incremental-step_max/facebook

Revision             Result                  N
chromium@482610      4.79338 +- 2.16454      21      good
chromium@482669      4.78771 +- 2.50842      21      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=facebook v8.mobile_infinite_scroll_tbmv2

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8974918099965665328


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

Comment 7 by 42576172...@developer.gserviceaccount.com, Jul 7 2017 


=== BISECT JOB RESULTS ===
NO Perf regression found

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : v8.mobile_infinite_scroll_tbmv2
  Metric       : v8-gc-incremental-step_max/facebook

Revision             Result                  N
chromium@482610      4.78671 +- 2.50182      21      good
chromium@482669      4.63633 +- 2.4761       21      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=facebook v8.mobile_infinite_scroll_tbmv2

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8974772701582494272


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

Comment 8 by 42576172...@developer.gserviceaccount.com, Jul 8 2017 


=== BISECT JOB RESULTS ===
NO Perf regression found

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : v8.mobile_infinite_scroll_tbmv2
  Metric       : v8-gc-scavenger_max/tumblr

Revision             Result                  N
chromium@482610      28.401 +- 13.8151       21      good
chromium@482669      27.8721 +- 8.91085      21      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=tumblr v8.mobile_infinite_scroll_tbmv2

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8974772698736946112


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

Comment 9 by 42576172...@developer.gserviceaccount.com, Jul 8 2017 


=== BISECT JOB RESULTS ===
NO Perf regression found

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : thread_times.simple_mobile_sites
  Metric       : thread_raster_cpu_time_per_frame/thread_raster_cpu_time_per_frame

Revision             Result                     N
chromium@482617      0.364548 +- 0.182816       21      good
chromium@482635      0.384128 +- 0.0678957      20      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests thread_times.simple_mobile_sites

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8974709437049454208


For feedback, file a bug with component Speed>Bisection

Comment 10 by senorblanco@chromium.org, Jul 27 2017 

The worst offender (at 91.1%), v8.mobile_infinite_scroll_tbmv2, seems to be very noisy on this device, and the regression looks to be in the noise.

The second worst offender (at 13.6%) renderer_compositor_cpu_time_per_frame has no compositor changes in its regression range that I can see. It's also quite noisy, and seems to have briefly recovered for a few subsequent revision, although it then looks like it regressed for good.

Comment 13 by vmi...@chromium.org, Jul 28 2017

Cc: vmi...@chromium.org
Project Member

Comment 14 by 42576172...@developer.gserviceaccount.com, Jul 29 2017

Mergedinto: 737623
Status: Duplicate (was: Untriaged)

=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : Xianzhu Wang
  Commit : 2381016a46e881351c34f7b7c2e6692565222132
  Date   : Tue Jun 27 18:43:07 2017
  Subject: Add WritingModeUtils.h to convert logical/physical directions

Bisect Details
  Configuration: win_perf_bisect
  Benchmark    : blink_perf.paint
  Metric       : LocalFrameView::prePaint/LocalFrameView::prePaint
  Change       : 9.18% | 44.8399722222 -> 48.9554074074

Revision             Result                   N
chromium@482612      44.84 +- 1.02782         6      good
chromium@482662      45.8128 +- 0.83069       6      good
chromium@482675      45.4854 +- 0.547803      6      good
chromium@482681      45.9108 +- 0.633997      6      good
chromium@482682      45.8473 +- 1.82834       6      good
chromium@482683      48.8915 +- 0.760324      6      bad       <--
chromium@482684      49.0251 +- 1.07809       6      bad
chromium@482687      48.6382 +- 0.380932      6      bad
chromium@482711      48.9554 +- 0.910938      6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests blink_perf.paint

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8972794331649532880


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

Comment 15 by 42576172...@developer.gserviceaccount.com, Jul 30 2017

Cc: rnep...@chromium.org
Owner: rnep...@chromium.org

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

Hi rnephew@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 : rnephew
  Commit : 038bd90be90e8fb7ccc2e2b801166cb93a05b4ea
  Date   : Tue Jun 27 03:40:00 2017
  Subject: [Telemetry] Move smoothness.* benchmarks to story expectations.

Bisect Details
  Configuration: win_perf_bisect
  Benchmark    : smoothness.tough_canvas_cases
  Metric       : frame_times/frame_times
  Change       : 8.08% | 24.0184277906 -> 25.9599863668

Revision             Result                   N
chromium@482252      24.0184 +- 0.22361       6      good
chromium@482518      24.1027 +- 0.326085      6      good
chromium@482527      24.1419 +- 0.453039      6      good
chromium@482531      24.2999 +- 0.206405      6      good
chromium@482532      24.1138 +- 0.190726      6      good
chromium@482533      24.7956 +- 0.416442      6      bad       <--
chromium@482535      24.7526 +- 0.37415       6      bad
chromium@482552      24.868 +- 0.493664       6      bad
chromium@482585      24.9132 +- 0.250128      6      bad
chromium@482651      24.9249 +- 0.445116      6      bad
chromium@482784      25.1658 +- 0.316395      6      bad
chromium@483315      25.96 +- 0.256014        6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests smoothness.tough_canvas_cases

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8972794249949647312


For feedback, file a bug with component Speed>Bisection

Comment 16 by rnep...@chromium.org, Jul 30 2017

Owner: ----
Any 'regression' that was caused by this CL is intentional. A test was disabled on all platfroms because we did not have a way to disable stories by platform before. I added the ability to do so, and re-enabled a test on all platforms except android nexus 5, where the story was originally failing. 

Since the other bisect caught a different CL, I wont close as WontFix. Taking myself off as owner. If the other CL isn't the cause, then it seems safe to close as wont fix since my cl is likely to show up as a regression even though it does not introduce one since it adds another story to windows runs.

Comment 17 by sullivan@chromium.org, Jul 31 2017 

Note: I found a couple of tests where we were still monitoring the average and moved them to per-page thanks to Randy's comment.
►

Sign in to add a comment