Summary: | optimize jsarray setlength | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Peng Xinchao <xinchao.peng> | ||||
Component: | JavaScriptCore | Assignee: | Nobody <webkit-unassigned> | ||||
Status: | RESOLVED INVALID | ||||||
Severity: | Normal | CC: | ggaren | ||||
Priority: | P2 | ||||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | Unspecified | ||||||
OS: | Unspecified | ||||||
Attachments: |
|
Description
Peng Xinchao
2013-09-29 19:33:54 PDT
Created attachment 212944 [details]
optimze jsarray setlength
JSC Engine Optimize 'var ret =[] ; ret.length=1024' .When type of ret is undefine and it is setted length , its only public length is setted ,but its vector is not setted . the memory of variable ret is not changed .When the type of variable ret is confirm , its will allocate memory by public length.
Can you provide a benchmark that demonstrates why you did this? sunsprider and v8benchmark is not changed . It is to improve Drameo/JavaScript Arrays . The first item of Arrays is "Array Construction, []". its source code : for ( var j = 0; j < i * 15; j++ ) { ret = []; ret.length = 1024; } My purpose is here. I'm pretty skeptical about a change just for this benchmark. Real programs typically use their arrays. Have you seen behavior like this in any real programs? I don't argree with you. When type of array is undefine and array is setted length , That allocate memory directly is not necessary . I think that Memory will be allocated when it is really require. > I don't argree with you.
This isn't a matter of opinion. If a real program or website uses this idiom, you should be able to provide an empirical example.
Comment on attachment 212944 [details]
optimze jsarray setlength
Marking r- because we don't have an example of a website that uses this idiom.
|