Bug 20974

Summary: JSC::fastIsNumber() should have ALWAYS_INLINE modifier
Product: WebKit Reporter: Csaba Osztrogonác <ossy>
Component: JavaScriptCoreAssignee: Geoffrey Garen <ggaren>
Status: RESOLVED FIXED    
Severity: Normal CC: abarth, zwarich
Priority: P2    
Version: 528+ (Nightly build)   
Hardware: PC   
OS: Linux   
Bug Depends on: 20914    
Bug Blocks:    
Attachments:
Description Flags
proposed patch ggaren: review+

Csaba Osztrogonác
Reported 2008-09-21 14:16:15 PDT
I created a profile with GNU gprof, and founded out that JSC::fastIsNumber() takes 4.96%/6.03% of SunSpider runtime on Linux platform at r36540 (non-CTI/CTI). I suprised, because this frequent function hasn't ALWAYS_INLINE modifier. With ALWAYS_INLINE modifier It looks good: 1.40% speedup on Sunspider, 1.44% on V8 (Linux, non-CTI version, r36738) 2.91% speedup on SunSpider, 0.75% on V8 (Linux, CTI verison, r36540) (Unfortunately CTI not works currently on linux platform.)
Attachments
proposed patch (1.18 KB, patch)
2008-09-21 14:21 PDT, Csaba Osztrogonác 		ggaren: review+
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Csaba Osztrogonác
Comment 1 2008-09-21 14:21:57 PDT
Created attachment 23631 [details] proposed patch I'm interested in results on Mac too. Anybody?
Csaba Osztrogonác
Comment 2 2008-09-21 14:38:24 PDT
This patch depends on bug20914. Without this patch, jsc aborts with segmentation fault.
Geoffrey Garen
Comment 3 2008-09-24 15:27:54 PDT
Comment on attachment 23631 [details] proposed patch r=me we should verify the numbers on mac before landing.
Adam Barth
Comment 4 2008-10-14 01:27:31 PDT
Is this good to land? Its been weeks and no one seems interested in running the numbers on the Mac. I would do it, but I know nothing about performance.
Geoffrey Garen
Comment 5 2008-10-15 09:55:00 PDT
I'll try to land this today.
Geoffrey Garen
Comment 6 2008-10-15 19:36:34 PDT
Committed revision 37625.
Note You need to log in before you can comment on or make changes to this bug.