Bug 14717

Summary: REGRESSION: 1% JavaScript performance regression fro Window refactoring
Product: WebKit Reporter: Maciej Stachowiak <mjs>
Component: WebCore JavaScriptAssignee: Maciej Stachowiak <mjs>
Status: RESOLVED FIXED    
Severity: Normal Keywords: InRadar
Priority: P1    
Version: 523.x (Safari 3)   
Hardware: Mac   
OS: OS X 10.4   
Attachments:
Description Flags
the fix - inline some new methods aroben: review+

Maciej Stachowiak
Reported 2007-07-22 19:32:45 PDT
Recent refactoring for the Window object's JS bindings has caused a 1% speed regression in JavaScript iBench and the "Celtic Kane" JavaScript speed test <http://celtickane.com/projects/jsspeed.php>. It shows up as 3% on the math test of that benchmark specifically. Inlining Window::impl() and DOMWindow::frame() fixes the regression.
Attachments
the fix - inline some new methods (2.34 KB, patch)
2007-07-22 19:40 PDT, Maciej Stachowiak 		aroben: review+
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Adam Roben (:aroben)
Comment 1 2007-07-22 19:37:03 PDT
<rdar://problem/5353174>
Maciej Stachowiak
Comment 2 2007-07-22 19:40:57 PDT
Created attachment 15639 [details] the fix - inline some new methods
Adam Roben (:aroben)
Comment 3 2007-07-22 19:44:38 PDT
Comment on attachment 15639 [details] the fix - inline some new methods + Frame* frame() { return m_frame; } Seems like this should be a const method -- I wonder if anything depends on it not being one? r=me
Maciej Stachowiak
Comment 4 2007-07-22 22:49:02 PDT
Fixed - I did not try to make any tweaks to const-correctness for now.
Note You need to log in before you can comment on or make changes to this bug.