[webkit-reviews] review granted: [Bug 35233] Optimize Latin-1 decoding in TextCodecLatin1::decode() : [Attachment 49275] Same patch, templates instead of ifdefs

[bugzilla-daemon at webkit.org]

Darin Adler <darin at apple.com> has granted Andreas Kling
<andreas.kling at nokia.com>'s request for review:
Bug 35233: Optimize Latin-1 decoding in TextCodecLatin1::decode()
https://bugs.webkit.org/show_bug.cgi?id=35233

Attachment 49275: Same patch, templates instead of ifdefs
https://bugs.webkit.org/attachment.cgi?id=49275&action=review

------- Additional Comments from Darin Adler <darin at apple.com>
> +	   * platform/text/TextCodecLatin1.cpp:
> +	   (WebCore::):
> +	   (WebCore::TextCodecLatin1::decode):

You should delete partial lines like that "WebCore::" one there. They're
created due to a bug in the script.

> +template<> struct NonASCIIMask<8> {
> +    static uintptr_t value() { return 0x8080808080808080UL; }
> +};

I was concerned that compiling this might yield warnings on 32-bit platforms.
That's why I used the type "unsigned long long" in my suggested code. I guess
we're OK because the "UL" suffix makes it an unsigned long, not unsigned long
long. I meant to use "ULL".

Please investigate writing full words as well and applying this same
optimization to the increasingly-often-used UTF-8 decoder as well.