New issue
Advanced search Search tips

Issue 897145 link

Starred by 1 user

Issue metadata

Status: Assigned
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

12%-14.8% regression in rendering.mobile/tasks_per_frame_total_all at 600568:600618

Project Member Reported by hjd@google.com, Oct 19

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=897145

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=db3cc817434877c31a7e5d32e1e89f3f68a11ffc2561ea18331b81f3fc2ba5c4


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

Android Nexus5 Perf

rendering.mobile - Benchmark documentation link:
  https://bit.ly/rendering-benchmarks
Cc: johnidel@chromium.org
Owner: johnidel@chromium.org
Status: Assigned (was: Untriaged)
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/11d08d7ae40000

Measure media bytes received for MSE/EME/SRC by johnidel@chromium.org
https://chromium.googlesource.com/chromium/src/+/b93339160a4cb20476f98229c98c9147e9816dd4
tasks_per_frame_total_all: 626 → 699.1 (+73.13)

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

Benchmark documentation link:
  https://bit.ly/rendering-benchmarks
Cc: dalecur...@chromium.org
+dalecurtis@

Does this seem acceptable? I have not verified locally on Android but this seems reasonable if we are sending a mojo everytime data is buffered. 

To decrease the frequency we could lower the resolution and only send an update every ~100k. Alternatively, we could aggregate and send on a timer which will lose some data to fast process shutdown.
Ah yeah these should be buffered -- I think time based is best, since rate based will influence the peaks of the histogram. I forgot we had this same problem in the past when we used to send buffering data over to chrome://media-internals. 
Project Member

Comment 6 by bugdroid1@chromium.org, Oct 31

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/c9abf4aeccc1ac05e0cf6ac3d3d3d32d5f094851

commit c9abf4aeccc1ac05e0cf6ac3d3d3d32d5f094851
Author: John Delaney <johnidel@chromium.org>
Date: Wed Oct 31 00:39:16 2018

Buffer bytes received updates from WebMediaPlayer to metrics

Currently reporting bytes to the MediaMetricsProvider is causing a
performance regression due to the frequency of IPCs being sent. Buffer
updates in the media player and send every 500 milliseconds to limit the
number of IPCs sent over the lifetime of the mediaplayer.

Bug: 897145
Change-Id: I7468a750e2a96106aa080559321902cf0636d3c0
Reviewed-on: https://chromium-review.googlesource.com/c/1308761
Commit-Queue: John Delaney <johnidel@chromium.org>
Reviewed-by: Dale Curtis <dalecurtis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#604079}
[modify] https://crrev.com/c9abf4aeccc1ac05e0cf6ac3d3d3d32d5f094851/media/blink/webmediaplayer_impl.cc
[modify] https://crrev.com/c9abf4aeccc1ac05e0cf6ac3d3d3d32d5f094851/media/blink/webmediaplayer_impl.h

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/17482489e40000

HTTP status code 401: Login Required

Sign in to add a comment