Bug 16503 - match limit takes at least 13% of the time on the SunSpider regexp-dna test
Summary: match limit takes at least 13% of the time on the SunSpider regexp-dna test
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: 528+ (Nightly build)
Hardware: Mac OS X 10.4
: P3 Enhancement
Assignee: Nobody
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2007-12-18 10:54 PST by Darin Adler
Modified: 2010-06-11 17:36 PDT (History)
6 users (show)

See Also:


Attachments
2.3% improvement (1.49 KB, patch)
2008-05-29 08:02 PDT, Alexey Proskuryakov
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Darin Adler 2007-12-18 10:54:28 PST
The match limit inside the regular expression execution engine slows down the regexp-dna test by about 13%. If we can find some more efficient way of preventing complex regular expressions from causing a hang, then we could get a substantial speed-up.

See bug 16458 for some history.
Comment 1 Alexey Proskuryakov 2008-05-29 08:02:00 PDT
Created attachment 21408 [details]
2.3% improvement

Removing the check altogether is a 9.5% win on my machine.

I guess this micro-optimization is unnecessary if someone has a real fix in mind, but I thought it was worth posting anyway.
Comment 2 Darin Adler 2008-05-29 09:33:37 PDT
Comment on attachment 21408 [details]
2.3% improvement

r=me

I'm not sure anyone has a handle on a "real fix".
Comment 3 Cameron Zwarich (cpst) 2008-06-08 22:36:51 PDT
Alexey, are you going to land this?
Comment 4 Alexey Proskuryakov 2008-06-09 04:28:33 PDT
I'm trying to land this periodically, but random code generation changes elsewhere cause it to be a regression overall now. One of these days, I'll get lucky and have it landed.
Comment 5 Alexey Proskuryakov 2008-06-10 07:56:39 PDT
Comment on attachment 21408 [details]
2.3% improvement

Landed in r34480, clearing review flag. Today, the win on regexp-dna was only 0.6%, but the total win was huge (0.5%), apparently due to random code generation changes.
Comment 6 Alexey Proskuryakov 2010-06-11 17:36:13 PDT
Something was landed here, so marking fixed. The rest of the bug is likely long obsolete.