New issue
Advanced search Search tips

Issue 893797 link

Starred by 1 user

Issue metadata

Status: WontFix
Merged: issue 893818
Owner:
Closed: Oct 19
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression

Blocked on:
issue 896542



Sign in to add a comment

2.2%-31.9% regression in rendering.desktop/tasks_per_frame_total_all at 596272:596513

Project Member Reported by m...@chromium.org, Oct 9

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=893797

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=406c61520f7822970a87510bbd3e850c781510f1f904a7b6898cf053a33bc93c


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

Win 7 Perf

rendering.desktop - Benchmark documentation link:
  https://bit.ly/rendering-benchmarks
Cc: tapted@chromium.org
Owner: tapted@chromium.org
Status: Assigned (was: Untriaged)
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/158a9e04e40000

Make ObserverList iteration 10~20 times faster. by tapted@chromium.org
https://chromium.googlesource.com/chromium/src/+/f59cdf0d304cd451e2507a13dd12da6f8fccaffb
thread_raster_cpu_time_per_frame: 0.03786 → 0.05389 (+0.01603)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Benchmark documentation link:
  https://bit.ly/rendering-benchmarks
Mergedinto: 893818
Status: Duplicate (was: Assigned)
Status: Assigned (was: Duplicate)
unduping in case there's something here.  Issue 893818  looks to be noise. See also https://bugs.chromium.org/p/chromium/issues/detail?id=866000#c14 - it may be a systemic problem with rendering.desktop
Blockedon: 896542
Status: WontFix (was: Assigned)
I've spent some time on this, but I can't see anything interesting. It's just noise -- see *Issue 894287 rather -- "rendering benchmark results is noisy"

I looked at the "All graphs for this bug". The one named in the pinpoint seemed least like noise, but the noise is evident in the pinpoint plot. Digging around, the regression in thread_raster_cpu_time_per_frame is wholly balanced by an equivalent perf-improve in thread_browser_cpu_time_per_frame. r596492 mostly affects browser threads, so that seems expected.

https://chromeperf.appspot.com/report?sid=237cf8f0b41eb9bbcf9f867e366f1d440cdbe0f35bd3f287599b38b003906365&start_rev=585672&end_rev=600886

I don't think there's anything to be gained by spending more time on this.

As an idle task, I'm looking at other changes around r596492 in Issue 896542, but we don't have good perf coverage of things the browser thread is doing.
Screenshot from 2018-10-19 15-34-38.png
147 KB View Download
Screenshot from 2018-10-19 15-36-27.png
20.3 KB View Download

Sign in to add a comment