317281 - Make UCS4 to High/Low surrogate a little bit more efficient

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[Jungshik Shin]

Macros for UCS4 to high/low surrogate code points of UTF-16 in nsCharTrait.h can be made a bit more efficient.

Comment 1

[Jungshik Shin]

Attachment: patch

This patch would save us a few cycles. Two formulae were circulated on the Unicode mailing list a few years ago and were used in our code (nsUTF32toUnicode.cpp)

Comment 2

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 203791 [details] [diff] [review]
patch

>Index: xpcom/string/public/nsCharTraits.h
>+// It's trivial to derive the following formulae from the Unicode standard
>+// 4.0 section 3.7

I just looked at the standard, and isn't this actually section 3.9?

>+// ((c - 0x10000) >> 10) + 0xd800
>+#define H_SURROGATE(c) PRUnichar(PRUnichar(PRUint32(c) >> 10) + \
>+                                 PRUnichar(0xD7C0))

Please document that this _MUST_ be '+' and not '|'.  I'd be even happier with that _and_ a comment like:

  (c - 0x10000) >> 10 == (c >> 10) - 0x0080.  And 0xD7C0 == 0xD800 - 0x0080.

to make it clear what's going on.

>+// ((c - 0x10000) & 0x03ff) | 0xdc00

And here, a comment like:

   (c - 0x10000) & 0x03FF == c & 0x03FF, since 0x10000 & 0x03FF == 0.

With those comments, sr=bzbarsky

Comment 3

[Jungshik Shin]

thanks for quick reviews. I added comment lines suggested by bz