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

Issue 797060 link

Starred by 1 user
Status: Fixed
Owner:
japhet@chromium.org
Closed: Jan 2018
Cc:
perezju@chromium.org
japhet@chromium.org
█ briander...@chromium.org
ksakamoto@chromium.org
erikc...@chromium.org
hirosh...@chromium.org
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 1
Type: Bug-Regression



Sign in to add a comment

1.4%-159.4% regression in system_health.memory_desktop at 524735:525248

Project Member Reported by briander...@chromium.org, Dec 21 2017 
See the link to graphs below.
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Dec 21 2017 

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=797060

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=4d8c0a96e129585ee74f9df671c25b801beb4210dfb25260e08aba8bca8b453e


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

android-nexus5
android-nexus5X
android-one
android-webview-nexus5X
chromium-rel-mac12
chromium-rel-win10
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Dec 21 2017

Cc: hirosh...@chromium.org ksakamoto@chromium.org japhet@chromium.org
Owner: japhet@chromium.org
Status: Assigned (was: Untriaged)
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/12dbabc1040000

SetResource() in ResourceFetcher for FontResources
By japhet@chromium.org · Mon Dec 18 21:42:32 2017
chromium @ a1bfc9f9cb531a1d5dd1e28b51a38b728ece1235

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

Comment 4 by sullivan@chromium.org, Jan 23 2018

Cc: perezju@chromium.org erikc...@chromium.org
Labels: -Pri-2 Pri-1
japhet: this is a pretty huge memory regression; the bisect is repro-ing 100MiB increase. Here are the traces it spit out, click the (M) in the upper right for a full memory dump:

Before your CL: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/browse_news_nytimes_2017-12-21_14-53-35_6934.html
After your CL: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/browse_news_nytimes_2017-12-21_15-00-34_32369.html

You can see in the attached screenshot that after your CL, we use a lot more memory in malloc in the renderer process (skia->sk_glyph_cache)

cc-ing desktop and android memory benchmark owners.
Screen Shot 2018-01-23 at 2.05.08 PM.png
262 KB View Download

Comment 5 by japhet@chromium.org, Jan 23 2018

Status: Fixed (was: Assigned)
Sorry, I apparently didn't include this bug in my fix. https://chromium-review.googlesource.com/c/chromium/src/+/848014 partially reverted the offending CL and landed as r527372.  The graphs on https://chromeperf.appspot.com/group_report?bug_id=797060 appears to show full recovery of all regressions around r527372, so I think this is fixed.
►

Sign in to add a comment