Date: Fri, 12 Mar 1999 22:22:41 +0100
To: pgcc AT delorie DOT com
Subject: Re: AMDK6 optimized kernel and others
Message-ID: <19990312222241.F9148@cerebro.laendle>
Mail-Followup-To: pgcc AT delorie DOT com
References: <Pine DOT GSO DOT 3 DOT 95 DOT 990312082710 DOT 13124A-100000 AT c2m2-17 DOT nada DOT kth DOT se> <Pine DOT GSO DOT 4 DOT 05 DOT 9903121429040 DOT 12114-100000 AT legolas DOT mdh DOT se>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <Pine.GSO.4.05.9903121429040.12114-100000@legolas.mdh.se>; from Henrik Berglund SdU on Fri, Mar 12, 1999 at 02:36:40PM +0100
X-Operating-System: Linux version 2.2.3 (root AT cerebro) (gcc driver version pgcc-2.93.09 19990221 (gcc2 ss-980929 experimental) executing gcc version 2.7.2.3) 
From: Marc Lehmann <pcg AT goof DOT com>
Reply-To: pgcc AT delorie DOT com
X-Mailing-List: pgcc AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

On Fri, Mar 12, 1999 at 02:36:40PM +0100, Henrik Berglund SdU wrote:
> I have notised that the 1.1.1 release of pgcc optimises better when it
> comes to float than the snapshot but maybe a bit slower at int.

That might be caused by bugs that were fixed, but sometimes made a program
faster. Often you have bugs that are easy to fix: you know that the
transformation you make is safe in 90% of the cases, but you can't check
for the remaining 10%, so you have to disable them all.

> the best flags for good float are 
> -O6 -march=amdk6 -funroll-all-loops -fforce-addr   

-fforce-addr is included in -O2 and up anyway, but apart from that I'd say
yes. Esp. loop unrolling is critical.

One might experiment with -fschedule-insns. It sometimes helps and sometimes
makes your code much slower...

--  
      -----==-                                             |
      ----==-- _                                           |
      ---==---(_)__  __ ____  __       Marc Lehmann      +--
      --==---/ / _ \/ // /\ \/ /       pcg AT goof DOT com      |e|
      -=====/_/_//_/\_,_/ /_/\_\       XX11-RIPE         --+
    The choice of a GNU generation                       |
                                                         |