Bug 29323 - canvas demo stutters half way through
Summary: canvas demo stutters half way through
Status: NEW
Alias: None
Product: WebKit
Classification: Unclassified
Component: DOM (show other bugs)
Version: 528+ (Nightly build)
Hardware: PC OS X 10.5
: P2 Normal
Assignee: Nobody
URL: http://www.nihilogic.dk/labs/pocket_f...
Keywords:
Depends on:
Blocks:
 
Reported: 2009-09-16 19:06 PDT by Eric Seidel (no email)
Modified: 2009-09-22 11:05 PDT (History)
1 user (show)

See Also:

Attachments
Shark sample from Safari 4.0.3, in case it's useful. (3.06 MB, application/octet-stream)
2009-09-17 01:02 PDT, Eric Seidel (no email) 		no flags Details
possibly useful sample trace? (610.75 KB, application/octet-stream)
2009-09-17 01:06 PDT, Eric Seidel (no email) 		no flags Details
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 Eric Seidel (no email) 2009-09-16 19:06:27 PDT 
canvas demo stutters half way through

http://www.nihilogic.dk/labs/pocket_full_of_canvas/demo.php

It also seems to peg Safari, but that may be expected.

CCing oliver as he's had interest in this type of perf bug in the past.
Comment 1 Oliver Hunt 2009-09-16 21:39:01 PDT 
I don't see any stuttering -- could you sample it when you see the stutter?
Comment 2 Eric Seidel (no email) 2009-09-17 00:57:03 PDT 
The stutter happens for me when the big white blobs come on screen.  Attempting to get a sample now.
Comment 3 Eric Seidel (no email) 2009-09-17 01:01:03 PDT 
The big white blobs right after the swirly colored dots (rather early in the demo).

I found resizing the window while the demo was going made it obvious when it started to stutter.

I was testing in a Safari 4.0.3 release build, so symbols are mostly missing.  It seems it was spending most of its time in JSObject::mark() according to Shark heavy view:

	25.7%	25.7%	JavaScriptCore	JSC::JSObject::mark()	
	12.0%	12.0%	JavaScriptCore	JSC::JSArray::mark()	
	9.9%	9.9%	JavaScriptCore	JSC::JSFunction::mark()	
	6.0%	6.0%	JavaScriptCore	JSC::CodeBlock::mark()	
	5.6%	5.6%	JavaScriptCore	unsigned long JSC::Heap::sweep<(JSC::HeapType)0>()	
	3.3%	3.3%	JavaScriptCore	JSC::JSActivation::mark()	
	3.1%	3.1%	JavaScriptCore	JSC::JSStaticScopeObject::mark()	

Still working on getting you a better sample.
Comment 4 Eric Seidel (no email) 2009-09-17 01:02:30 PDT 
Created attachment 39683 [details]
Shark sample from Safari 4.0.3, in case it's useful.
Comment 5 Eric Seidel (no email) 2009-09-17 01:06:28 PDT 
Created attachment 39684 [details]
possibly useful sample trace?

I can't really tell how to use Sample.app, it seems more complicated than the command line version. :)
Comment 6 Oliver Hunt 2009-09-17 11:15:23 PDT 
Can you please test in a nightly?  We've made a number of improvements to JSC since 4.0.3
Comment 7 Oliver Hunt 2009-09-17 11:15:48 PDT 
Are you testing in 32bit or 64bit safari?
Comment 8 Eric Seidel (no email) 2009-09-22 10:47:18 PDT 
It looks like my Safari only has 2 architectures.  Looks like I'm using 32-bit.  Safari 4.0.3 on 10.5.8.
file /Applications/Safari.app/Contents/MacOS/Safari
/Applications/Safari.app/Contents/MacOS/Safari: Mach-O universal binary with 2 architectures
/Applications/Safari.app/Contents/MacOS/Safari (for architecture ppc7400):	Mach-O executable ppc
/Applications/Safari.app/Contents/MacOS/Safari (for architecture i386):	Mach-O executable i386
Comment 9 Oliver Hunt 2009-09-22 11:01:56 PDT 
Eric, can you do me a favour and try with a nightly? (i assume you're on a 32bit only machine, or haven't upgraded to SL :P )
Comment 10 Eric Seidel (no email) 2009-09-22 11:05:54 PDT 
Happy to! :)  My SL copy is still in the mail (personal), and work won't have SL for a while yet (due to some Radar's on file). :(