Bug 103862 – [V8] Make v8::Null(isolate) faster by caching the value in V8 bindings

• Status: RESOLVED
• Resolution: FIXED
• Component: WebCore JavaScript

[Kentaro Hara]

Performance comparison between V8 and JSC revealed that DOM attributes that return v8::Null(isolate) are much slower than JSC. (https://docs.google.com/a/chromium.org/spreadsheet/ccc?key=0AlobCOyvTnPKdDg5S0dMdGRGRTRSaW53V1ppVzh6eXc#gid=2)

We can optimize v8::Null(isolate) by caching the value to V8PerIsolateData. (Although I don't like the optimization, it looks difficult to optimize v8::Null(isolate) more.)

[Kentaro Hara]

Created attachment 177196 [details]
Patch

[Adam Barth]

Comment on attachment 177196 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=177196&action=review

> Source/WebCore/bindings/v8/V8PerIsolateData.h:143
> +    v8::Handle<v8::Value> m_v8Null;

Do we need to use a ScopedPersistent here?  It's unclear to me whether V8 promises to return a persistent handel to null via v8::Null

[Adam Barth]

Comment on attachment 177196 [details]
Patch

This patch seems to be causing lots of crashes on the ews bots.

[Kentaro Hara]

(In reply to comment #3)
> (From update of attachment 177196 [details])
> This patch seems to be causing lots of crashes on the ews bots.

The problem will be fixed in bug 103949.

[Kentaro Hara]

Created attachment 177427 [details]
Patch

[WebKit Review Bot]

Comment on attachment 177427 [details]
Patch

Clearing flags on attachment: 177427

Committed r136526: <http://trac.webkit.org/changeset/136526>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.