Bug 46537 - SVG: Avoid calling pow() in genericParseNumber() unless necessary
Summary: SVG: Avoid calling pow() in genericParseNumber() unless necessary
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: SVG (show other bugs)
Version: 528+ (Nightly build)
Hardware: All All
: P2 Normal
Assignee: Nobody
URL:
Keywords: Performance
Depends on:
Blocks:
 
Reported: 2010-09-24 16:26 PDT by Andreas Kling
Modified: 2010-09-26 10:46 PDT (History)
3 users (show)

See Also:

Attachments
Proposed patch (1.24 KB, patch)
2010-09-24 16:28 PDT, Andreas Kling 		no flags Details | Formatted Diff | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Andreas Kling 2010-09-24 16:26:49 PDT 
We don't need to call pow() if there's no exponent.

0.4% speedup on <http://themaninblue.com/experiment/AnimationBenchmark/svg/>
Comment 1 Andreas Kling 2010-09-24 16:28:13 PDT 
Created attachment 68787 [details]
Proposed patch
Comment 2 Kenneth Rohde Christiansen 2010-09-24 16:36:58 PDT 
Comment on attachment 68787 [details]
Proposed patch

Nice !
Comment 3 WebKit Commit Bot 2010-09-24 18:57:24 PDT 
Comment on attachment 68787 [details]
Proposed patch

Clearing flags on attachment: 68787

Committed r68321: <http://trac.webkit.org/changeset/68321>
Comment 4 WebKit Commit Bot 2010-09-24 18:57:29 PDT 
All reviewed patches have been landed.  Closing bug.
Comment 5 Matthew Delaney 2010-09-26 10:37:01 PDT 
Just out of curiosity, how were you able to accurately/reliably measure a 0.4% speedup? Did you instrument the fps meter from that test to display some sort of longer run - and thus more steady - average?
Comment 6 Andreas Kling 2010-09-26 10:46:56 PDT 
(In reply to comment #5)
> Just out of curiosity, how were you able to accurately/reliably measure a 0.4% speedup? Did you instrument the fps meter from that test to display some sort of longer run - and thus more steady - average?

Using an internal tool that runs something for X frames and measures wall/CPU time spent. Also, testing on a slow device helps a lot with the stability of timings.