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

Issue 809446 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner:
Last visit > 30 days ago
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

10.2%-56.2% regression in blink_perf.shadow_dom at 533538:533655

Project Member Reported by kraynov@chromium.org, Feb 6 2018

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=809446

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=1cd06cba6c0d4c82b85ccfec249c1c1a1a563e49a61ec034e702c96ceafb0785


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

android-nexus5
chromium-rel-mac-retina
chromium-rel-mac11
chromium-rel-mac11-air
chromium-rel-mac11-pro
chromium-rel-win10
chromium-rel-win7-dual
chromium-rel-win7-gpu-ati
chromium-rel-win7-gpu-intel
chromium-rel-win7-gpu-nvidia
chromium-rel-win7-x64-dual
chromium-rel-win8-dual
linux-release
win-high-dpi
Cc: haraken@chromium.org hpayer@chromium.org mlippautz@chromium.org
Owner: hpayer@chromium.org
Status: Assigned (was: Untriaged)
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/14c70429840000

Oilpan: Remove Coalesce() logic. by hpayer@chromium.org
chromium @ fa293678b3eceaa1ed646ada4865bb875926aef6

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions
Hmm. This looks like a real regression.

Hannes: Would you mind taking a look?

Comment 5 by hpayer@google.com, Feb 20 2018

Yes, will take a look ASAP.
Hannes: Any update on this? This looks like a real regression.

Comment 7 by hpayer@chromium.org, Mar 23 2018

This regressions is already open too long.

I would close it with won't fix. This was the only regression visible on micro-benchmarks that triggered the old coalescing logic. The benefit of the change is simpler GC logic and less real-world jank.

Improvements on other perf bots of the regression range:
https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICQ7IOxgAgM
https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICQ9IXEogkM

The alternative is to revert this change and/or implement coalescing during idle time but I do not think it is worth the effort. kentaro WDYT?

Comment 8 by hpayer@chromium.org, Mar 23 2018

 Issue 809445  has been merged into this issue.

Comment 9 by haraken@google.com, Mar 23 2018

Fair enough :)

The coalescing would make sense only when we implement StackHeapVector / StackHeapHashTable (then many backing stores will be released promptly). We don't have a plan to do this now.

Status: Started (was: Assigned)
Status: Wontfix (was: Started)
OK, I am closing this one for now. I am pretty happy with the removed jank on UMA.
Cc: keishi@chromium.org

Sign in to add a comment