[webkit-reviews] review cancelled: [Bug 19180] speed up SunSpider by optimizing immediate number cases : [Attachment 21430] Patch for << and >>>

[bugzilla-daemon at webkit.org]

Alexey Proskuryakov <ap at webkit.org> has cancelled 's request for review:
Bug 19180: speed up SunSpider by optimizing immediate number cases
http://bugs.webkit.org/show_bug.cgi?id=19180

Attachment 21430: Patch for << and >>>
http://bugs.webkit.org/attachment.cgi?id=21430&action=edit

------- Additional Comments from Alexey Proskuryakov <ap at webkit.org>
Clearing review flag, since I landed this.

(In reply to comment #27)
> If you fixed a regression on PowerPC, was there a regression test indicating
> it?

Yes, it's one of the three Mozilla tests that I claimed to have fixed in the
previous patch. Only two were fixed, in fact.

> I'm concerned about the "to" vs. "get" in the name here. This class seems we
> use "get" to mean "convert with no possibility of error" and "to" to mean
> "convert with a possible exception or error".

I'm rather confused about the naming in JSImmediate.h - some "get" functions
also return a bool to indicate success. That said, this new function just
implements "ToUInt32" algorithm from ECMA-232 for immediates as far as I can
tell, and is not really a "get" in the sense that -1 turns into 4294967296. The
"get" functions that we have do not cast the result like that. I wasn't sure
what the word "truncated" meant here, but since this function does almost the
same work as getTruncatedInt32, I kept it.