From owner-freebsd-current  Mon Apr  5 21:59:33 1999
Delivered-To: freebsd-current@freebsd.org
Received: from apollo.backplane.com (apollo.backplane.com [209.157.86.2])
	by hub.freebsd.org (Postfix) with ESMTP id 185811518C
	for <current@FreeBSD.ORG>; Mon,  5 Apr 1999 21:59:31 -0700 (PDT)
	(envelope-from dillon@apollo.backplane.com)
Received: (from dillon@localhost)
	by apollo.backplane.com (8.9.3/8.9.1) id VAA01999;
	Mon, 5 Apr 1999 21:57:34 -0700 (PDT)
	(envelope-from dillon)
Date: Mon, 5 Apr 1999 21:57:34 -0700 (PDT)
From: Matthew Dillon <dillon@apollo.backplane.com>
Message-Id: <199904060457.VAA01999@apollo.backplane.com>
To: current@FreeBSD.ORG
Subject: EGCS optimizations
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

    Well, I played around with egcs a bit.  I had blown away my original gcc
    install so I couldn't compare egcs w/ gcc, but I did mess around with
    egcs's optimization options.

    My conclusion:  Don't bother with -mpentiumpro or -march=pentiumpro.
    Not only do they not result in better performance, -march=pentiumpro
    will not run on a K6-2.  I dunno about a K6-3.  -m<cpu> does not change
    the assembly output at all.  -march=<cpu> does change the assembly output,
    but does not appear to result in any noticeable improvement in performance
    over not using -m at all.

    However, -Os does seem to produce slightly smaller binaries compared with
    -O2 and actually does not appear to cost performance on my PPro-200.

	before

	-r--r--r--  1 root  wheel  543096 Apr  5 21:29 /usr/lib/libc.so.3
	-r-xr-xr-x  1 root  wheel  1517032 Apr  5 20:37 /usr/libexec/cc1
	-r-xr-xr-x  1 root  wheel  1580456 Apr  5 20:37 /usr/libexec/cc1obj
	-r-xr-xr-x  1 root  wheel  1928112 Apr  5 20:37 /usr/libexec/cc1plus

	after

	-r--r--r--  1 root  wheel  538872 Apr  5 21:35 /usr/lib/libc.so.3
	-r-xr-xr-x  1 root  wheel  1479240 Apr  5 21:47 /usr/libexec/cc1
	-r-xr-xr-x  1 root  wheel  1538376 Apr  5 21:47 /usr/libexec/cc1obj
	-r-xr-xr-x  1 root  wheel  1876464 Apr  5 21:47 /usr/libexec/cc1plus

    Compiling up /usr/src/usr.sbin with egcs and libc compiled with:

	-O2				160 seconds
	-O2 -march=pentiumpro		162 seconds
	-Os				161 seconds

    Which leads me to believe that using -Os might be beneficial.

						-Matt



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message