From owner-freebsd-bugs  Fri Apr  4 12:59:18 1997
Return-Path: <owner-bugs>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id MAA03072
          for bugs-outgoing; Fri, 4 Apr 1997 12:59:18 -0800 (PST)
Received: from gemini.yars.free.net (gemini.yars.free.net [193.233.192.17])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id MAA03019
          for <bugs@freebsd.org>; Fri, 4 Apr 1997 12:57:48 -0800 (PST)
Received: (from lav@localhost)
          by gemini.yars.free.net (8.8.5/8.8.4)
	  id AAA18942; Sat, 5 Apr 1997 00:57:12 +0400 (MSD)
Date: Sat, 5 Apr 1997 00:57:12 +0400 (MSD)
From: "Alexander V. Lukyanov" <lav@yars.free.net>
Message-Id: <199704042057.AAA18942@gemini.yars.free.net>
To: ache@nagual.ru, peter@spinner.DIALix.COM
Cc: bugs@freebsd.org, kev@l321.omsk.net.ru
Subject: Re: sscanf is slow on 2.2.1
Sender: owner-bugs@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

> From: =?KOI8-R?B?4c7E0sXKIP7F0s7P1w==?= <ache@nagual.ru>

> I just add speedup in case locale is not used. Nothing can be speeduped
> when locale is used, however...

But I noticed fstat being called inside sscanf... I don't know why it is, but
if cache something (locale info?) it could be speed up.

As about building the character set with locale, it could be very efficient
if there were two prebuilt 256 character tables: map character to linear code 
and map linear code to character. So we just get two linear codes for bound
characters and then fill the set with characters in the second table between
those two bound linear codes. Very simple.

   Alexander.