Sender: root AT squid DOT netplus DOT net
Message-ID: <36E15E7E.C77773EF@netplus.net>
Date: Sat, 06 Mar 1999 10:57:34 -0600
From: Steve Bergman <steve AT netplus DOT net>
X-Mailer: Mozilla 4.07 [en] (X11; I; Linux 2.2.2 i586)
MIME-Version: 1.0
To: pgcc AT delorie DOT com
Subject: How well do the -mcpu=amdk6 -march=amdk6 options work
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: pgcc AT delorie DOT com

Hi,

First of all, Thank You, pgcc/egcs developers.  I appreciate your efforts.

I am a long time amdk6 owner having had a K6/233 and now a K6-2/300 and about to
get a K6-3/400-450.  I have tried pgcc off and on for a year or so and have
always been disappointed by the results I get on my chip.  For example, today I
downloaded and compiled egcs-1.1.1/pgcc-1.1.1 and recompiled gzip with it, using
-mcpu=amdk6, -march=amdk6.  The result is larger and very slightly *slower* than
the stock redhat rpm.  This is about what I have found in past attempts.  Am I
missing something?  Anyone with K6 optimization ideas need someone to help test?

Also, what is the best way to lock the compiler into amdk6 optimization.  If I
can get a speed up, I would like to recompile all the SRPM's in my RH5.2
distribution with PGCC, but the RPMS all have -m486 build in.  For now, I have
hard coded ix86_arch_string and ix86_cpu_string in gcc/config/i386/i386.c but is
there a better way?

Thanks for any information or assistance,
Steve

 
----------
If this had been a real emergency, we would all have fled in terror, 
and you would not have been notified.