[PATCH] Microoptimize regex, take 2

Jakub Jelinek jakub@redhat.com
Wed Mar 24 12:14:00 GMT 2004


On Wed, Mar 24, 2004 at 06:05:59AM -0500, Bonzini wrote:
> Hi.  This is the same series of speedups that I sent earlier in
> March, without the changes for expecting mb_cur_max to be 1.
> 
> Note that I did not refer to linguistic minorities, but to
> a relative minority of UTF8 regexes that can actually be optimized,
> for example regexes with character ranges cannot be reduced to
> mb_cur_max to be 1.  Anyway no problem, I take your point and
> here is the updated patch you requested.
> 
> Paolo
> 
> 2004-03-24  Paolo Bonzini  <bonzini@gnu.org>
> 
> 	* posix/regex_internal.c (free_state): Free the
> 	word_trtable if it is used.
> 	(re_string_reconstruct, re_string_context_at): Add
> 	several branch predictions for mb_cur_max == 1,
> 	case-sensitive matching and no transition table being used.

Then you shouldn't mention it in the ChangeLog either ;).

	Jakub



More information about the Libc-alpha mailing list