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

Issue 868246 link

Starred by 1 user

Issue metadata

Status: Verified
Owner:
Closed: Jan 9
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Feature



Sign in to add a comment

experiment with shorter yield times in HTML parser

Project Member Reported by fergal@chromium.org, Jul 27

Issue description

Feature description: Currently it yields after 500ms, try some other values and measure impact

Eng owner: fergal@chromium.org
Product owner: kenjibaheux@chromium.org

Design doc: https://docs.google.com/document/d/1BCFYMOyoZcQ7pGc76DZATIp3AHHJfnIG7PFBl0xsr0w/edit?disco=AAAACB5jzeE&ts=5b58ced9

Are you planning on experimenting before launch? yes
Any new strings? no
Any implications for Google webservices (i.e. sync, translate)? no
Binary size? no
Do the existing perf tests exercise all aspects of your new feature(s)?

 
Project Member

Comment 1 by bugdroid1@chromium.org, Nov 14

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/e2fadf24cd0c2c40009b1eb60e38f5ff3e3c859e

commit e2fadf24cd0c2c40009b1eb60e38f5ff3e3c859e
Author: Fergal Daly <fergal@chromium.org>
Date: Wed Nov 14 22:08:38 2018

Add a testing configuration for HTMLParsingYieldTime study.

Corresponds to http://cl/221021676

Bug:  868246 
Change-Id: I26f7fe58f07ecfcbe1b540b7c805f575b8b73320
Reviewed-on: https://chromium-review.googlesource.com/c/1198363
Reviewed-by: Mark Pearson <mpearson@chromium.org>
Commit-Queue: Fergal Daly <fergal@chromium.org>
Cr-Commit-Position: refs/heads/master@{#608141}
[modify] https://crrev.com/e2fadf24cd0c2c40009b1eb60e38f5ff3e3c859e/testing/variations/fieldtrial_testing_config.json

Project Member

Comment 2 by bugdroid1@chromium.org, Jan 7

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/2d7da2e9289a81e9ba92ba54af8887cdc20e7fd4

commit 2d7da2e9289a81e9ba92ba54af8887cdc20e7fd4
Author: Fergal Daly <fergal@chromium.org>
Date: Mon Jan 07 05:52:32 2019

Revert https://crrev.com/c/1148172 to control the HTML parser yield
interval via field trial."

No significant difference was observed in experiments.

Change-Id: I908a0b2d725512bb46d46e2dc69fdffa32185e0c
Bug:  868246 
Reviewed-on: https://chromium-review.googlesource.com/c/1388350
Reviewed-by: Kouhei Ueno <kouhei@chromium.org>
Commit-Queue: Fergal Daly <fergal@chromium.org>
Cr-Commit-Position: refs/heads/master@{#620257}
[modify] https://crrev.com/2d7da2e9289a81e9ba92ba54af8887cdc20e7fd4/third_party/blink/renderer/core/html/parser/html_parser_scheduler.cc
[modify] https://crrev.com/2d7da2e9289a81e9ba92ba54af8887cdc20e7fd4/third_party/blink/renderer/core/html/parser/html_parser_scheduler.h

Project Member

Comment 3 by bugdroid1@chromium.org, Jan 8

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/e550d1af5cbd5a21dd90ebc450335736024ca496

commit e550d1af5cbd5a21dd90ebc450335736024ca496
Author: Fergal Daly <fergal@chromium.org>
Date: Tue Jan 08 03:09:18 2019

Remove HTMLParsingYieldTime study for testing, it has expired and will
not be run again.

Code was removed in https://chromium-review.googlesource.com/c/1388350

Bug:  868246 
Change-Id: Iaaa6e908c9d215bef1246a54b2a11dcf8b1d1045
Reviewed-on: https://chromium-review.googlesource.com/c/1388347
Commit-Queue: Fergal Daly <fergal@chromium.org>
Reviewed-by: Kouhei Ueno <kouhei@chromium.org>
Reviewed-by: Mark Pearson <mpearson@chromium.org>
Cr-Commit-Position: refs/heads/master@{#620596}
[modify] https://crrev.com/e550d1af5cbd5a21dd90ebc450335736024ca496/testing/variations/fieldtrial_testing_config.json

Status: Verified (was: Assigned)
Closing this bug. I couldn't produce an effect visible in finch even extending the trial to BETA

https://docs.google.com/spreadsheets/d/1NyPX57R24SVn3bN-jkCMA2x2FodUtagBkoSSKjHF7D4/edit#gid=0

and

https://uma.googleplex.com/variations?sid=aa4697c9a759aef8c82c412d16845d0f

It also seemed to produce a bunch of unexpected memory regressions, all merged into http://crbug/905844

Finally, when I reverted it, perf regression bugs were filed

https://crbug.com/919537
https://crbug.com/919919

E.g. timeToFirstPaint on baidu.com went from 4s to 7s on Android go. So maybe this has some positive impact but it's either not significant in the real world or we're not measuring it right.

Sign in to add a comment