From owner-freebsd-hackers  Tue Feb  4 21:50:51 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id VAA05508
          for hackers-outgoing; Tue, 4 Feb 1997 21:50:51 -0800 (PST)
Received: from dg-rtp.dg.com (dg-rtp.rtp.dg.com [128.222.1.2])
          by freefall.freebsd.org (8.8.5/8.8.5) with SMTP id VAA05503
          for <freebsd-hackers@freebsd.org>; Tue, 4 Feb 1997 21:50:48 -0800 (PST)
Received: by dg-rtp.dg.com (5.4R3.10/dg-rtp-v02)
	id AA07471; Wed, 5 Feb 1997 00:50:12 -0500
Received: from ponds by dg-rtp.dg.com.rtp.dg.com; Wed,  5 Feb 1997 00:50 EST
Received: from lakes.water.net (lakes [10.0.0.3]) by ponds.water.net (8.8.3/8.7.3) with ESMTP id VAA25506; Tue, 4 Feb 1997 21:43:06 -0500 (EST)
Received: (from rivers@localhost) by lakes.water.net (8.8.3/8.6.9) id VAA23395; Tue, 4 Feb 1997 21:47:32 -0500 (EST)
Date: Tue, 4 Feb 1997 21:47:32 -0500 (EST)
From: Thomas David Rivers <ponds!rivers@dg-rtp.dg.com>
Message-Id: <199702050247.VAA23395@lakes.water.net>
To: patrick@xinside.com, ponds!freebsd.org!freebsd-hackers
Subject: Re: [Fwd: freebsd performance.]
Content-Type: text
Sender: owner-hackers@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk


Patrick Giagnocavo writes:

> Julian Elischer writes:
> 
>  > The posix regex library is VERY VERY slow.
>  > 
>  > I have a program that uses a large regex to parse some input.
>  > 
>  > I have a version in perl and a version in C++ using the freebsd posix
>  > regex library.
>  > 
>  > The perl version is 100X faster that the C++ version.
>  > 
>  > gprof on the C++ version shows 99% of the spend in:
>  > 
>  >  91.53     46.17    46.17  1152366     0.04     0.04  lstep
>  >   6.52     49.46     3.29    98497     0.03     0.47  lslow
> 
> I am surprised that some of our more erudite members on the list have
> not jumped on this.  So, a definitely non-erudite person will.
> 
	...
> 
> There are two different 'engines' - NFA (nondeterministic finite
> automaton) and DFA (deterministic finite automaton).  Perl is
> 'traditional NFA' according to Mr. Friedl, while POSIX leans towards
> DFA-like behavior in all cases (always returns 'leftmost-longest' that
> matches - perl returns I believe the first part that matches).
 
 Ah, but you should remember that all NFAs are convertable to DFAs.

> 
> Also, Perl does not 'do' POSIX IIRC; so the results can actually be
> different when using the same regex string.  Perl does however have
> some very powerful features for its regex - covered in the book. 

 This could be part of the reason for a performance difference;
if they are not performing the same operation - you're comparing apples
and oranges.


	- Dave Rivers -