[Webkit-unassigned] [Bug 13227] StringImpl::isLower incorrectly assumes islower returns 1 (it can return any non-0)

[bugzilla-daemon at webkit.org]

http://bugs.webkit.org/show_bug.cgi?id=13227





------- Comment #7 from KwhiteRight at gmail.com  2007-04-10 11:58 PDT -------
Created an attachment (id=13998)
 --> (http://bugs.webkit.org/attachment.cgi?id=13998&action=view)
benchmark source file

(In reply to comment #4)
> (In reply to comment #3)
> > An alternative to 
> >   allLower &= islower(c & 0x7F) != 0;
> > would be
> >   allLower = allLower && islower(c & 0x7F);
> > if that's faster.
> 
> No, that's way slower.

Are you sure?  I ran a simple (and completely unoptimized) benchmark that
indicates it's noticeably faster than either the original version or the new
boolean one, even for an all-lower-case string.  And for one with any
upper-case, it's faster still, since logical && short-circuits the islower()
call as soon as the first upper-case letter is encountered.

Lower-case string:
Original: 17.819 seconds (wall time)
Boolean: 19.257
Logical: 10.185

Upper-case string:
Original: 17.229
Boolean: 18.081
Logical: 7.698

User and kernel times from tcsh 'time' are analogous.  The benchmark source is
attached.


-- 
Configure bugmail: http://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.