X-pop3-spooler:	POP3MAIL 2.1.0 b 4 980420 -bs-
Message-ID:	<19980909212933.55079@cerebro.laendle>
Date:	Wed, 9 Sep 1998 21:29:33 +0200
From:	Marc Lehmann <pcg AT goof DOT com>
To:	Steven Snyder <ssnyder AT indy DOT net>,
	pgcc mailing list <beastium-list AT Desk DOT nl>
Subject:	Re: Is -fstrength-reduce unstable?
Mail-Followup-To:	Steven Snyder <ssnyder AT indy DOT net>,
	pgcc mailing list <beastium-list AT Desk DOT nl>
References:	<Pine DOT SUN DOT 3 DOT 96 DOT 980909131604 DOT 9304A-100000 AT indy1>
Mime-Version:	1.0
In-Reply-To:	<Pine.SUN.3.96.980909131604.9304A-100000@indy1>; from Steven Snyder on Wed, Sep 09, 1998 at 01:22:27PM -0500
X-Operating-System:	Linux version 2.1.120 (root AT cerebro) (gcc version pgcc-2.91.57 19980901 (egcs-1.1 release))
Status:	RO
Lines:	24

On Wed, Sep 09, 1998 at 01:22:27PM -0500, Steven Snyder wrote:
> I see a lot of software packages whose compilation uses the
> -fno-strength-reduce switch.  (I'm reminded of this because I'm building
> glibc right now and it uses that switch as well.)  
> 
> Is strength reduction known to be buggy?  And if so, why is it enabled by
> default?

Its not perfect, but its as safe as almost every other part of the compiler.
Back in linux-1.3, the kernel hit a bug with -fstrength-reduce, and all people
suddenly though "see how buggy gcc is", even when, in fact, they work fine.

But there is a reason why (some!) software packages use this switch, as it
sometimes (rare) decreases performance.

But normally you wouldn't want to disable strength reduction.

      -----==-                                              |
      ----==-- _                                            |
      ---==---(_)__  __ ____  __       Marc Lehmann       +--
      --==---/ / _ \/ // /\ \/ /       pcg AT goof DOT com       |e|
      -=====/_/_//_/\_,_/ /_/\_\                          --+
    The choice of a GNU generation                        |
                                                          |

- Raw text -