Summary: | REGRESSION: d3 Bullet Charts demo doesn't work (call with argument assignment is broken) | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Geoffrey Garen <ggaren> | ||||
Component: | WebKit Misc. | Assignee: | Geoffrey Garen <ggaren> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | fpizlo | ||||
Priority: | P2 | Keywords: | InRadar | ||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | Unspecified | ||||||
OS: | Unspecified | ||||||
Attachments: |
|
Description
Geoffrey Garen
2012-01-09 16:41:16 PST
Created attachment 121766 [details]
Patch
Comment on attachment 121766 [details]
Patch
Yay! r=me
Any performance impact? > Any performance impact?
Bencher says 1.002x worse on SunSpider and v8 -- seems reasonable to call that "no change".
Bytecode generation for f.apply is definitely worse by one op_mov. This is theoretically fixable with more robust optimization information in the AST, but I think that's probably the wrong direction -- long-term, we just want to compile f.apply in the DFG, which will elide the op_mov.
(In reply to comment #5) > > Any performance impact? > > Bencher says 1.002x worse on SunSpider and v8 -- seems reasonable to call that "no change". > > Bytecode generation for f.apply is definitely worse by one op_mov. This is theoretically fixable with more robust optimization information in the AST, but I think that's probably the wrong direction -- long-term, we just want to compile f.apply in the DFG, which will elide the op_mov. Agree! Committed r104762: <http://trac.webkit.org/changeset/104762> |