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

Issue 766578 link

Starred by 2 users

Issue metadata

Status: WontFix
Owner:
Closed: Sep 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

1.7% regression in smoothness.gpu_rasterization.tough_filters_cases at 502413:502476

Project Member Reported by alexclarke@chromium.org, Sep 19 2017

Issue description

See the link to graphs below.
 
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Sep 19 2017

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=766578

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=5ec579b4bf5a30181b8308c1f181224f2d5c3e7fe0efa03f954b32b4a3d43660


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

android-nexus7v2
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Sep 19 2017

Cc: ethannicholas@google.com
Owner: ethannicholas@google.com
Status: Assigned (was: Untriaged)

=== Auto-CCing suspected CL author ethannicholas@google.com ===

Hi ethannicholas@google.com, 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 : Ethan Nicholas
  Commit : 05d5a13fea6246648de7e41358ed338d53c85ea2
  Date   : Fri Sep 15 18:50:54 2017
  Subject: Revert "Revert "Switched highp float to highfloat and mediump float to half.""

Bisect Details
  Configuration: android_nexus7_perf_bisect
  Benchmark    : smoothness.gpu_rasterization.tough_filters_cases
  Metric       : frame_times/IE PirateMark
  Change       : 2.00% | 29.3170197828 -> 29.9039252011

Revision                             Result                   N
chromium@502412                      29.317 +- 0.302337       6      good
chromium@502444                      29.3109 +- 0.323964      6      good
chromium@502446                      29.2787 +- 0.208846      6      good
chromium@502447                      29.3325 +- 0.432927      6      good
chromium@502447,skia@49f1f34438      29.2165 +- 0.186568      6      good
chromium@502447,skia@05d5a13fea      29.8565 +- 0.316534      6      bad       <--
chromium@502447,skia@451b7439f5      29.9565 +- 0.262992      6      bad
chromium@502447,skia@53cd6c4331      29.8949 +- 0.135004      6      bad
chromium@502447,skia@b2b28d1eba      29.8932 +- 0.346461      6      bad
chromium@502448                      29.8214 +- 0.303508      6      bad
chromium@502452                      29.8091 +- 0.22301       6      bad
chromium@502460                      29.8037 +- 0.223437      6      bad
chromium@502476                      29.9039 +- 0.263948      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=IE.PirateMark smoothness.gpu_rasterization.tough_filters_cases

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8968030400664632944


For feedback, file a bug with component Speed>Bisection
Status: WontFix (was: Assigned)
Some slowdown was expected as a result of this change. There were several spots where we were handling coordinates using insufficient precision, resulting in rendering errors with coordinates greater than 1024. Increasing the precision fixes this, at the cost of the calculations taking slightly longer.

Sign in to add a comment