From owner-freebsd-hackers  Thu Feb 27 08:12:14 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id IAA04557
          for hackers-outgoing; Thu, 27 Feb 1997 08:12:14 -0800 (PST)
Received: from usr11.primenet.com (root@usr11.primenet.com [206.165.5.111])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id IAA04552
          for <freebsd-hackers@freebsd.org>; Thu, 27 Feb 1997 08:12:10 -0800 (PST)
Received: from primenet.com (root@mailhost01.primenet.com [206.165.5.52])
	by usr11.primenet.com (8.8.5/8.8.5) with ESMTP id JAA11245;
	Thu, 27 Feb 1997 09:11:20 -0700 (MST)
Received: from conceptual.com (consys.com [207.218.17.187])
	by primenet.com (8.8.5/8.8.5) with ESMTP id JAA07965;
	Thu, 27 Feb 1997 09:11:15 -0700 (MST)
Received: from conceptual.com (localhost [127.0.0.1]) by conceptual.com (8.8.5/8.6.9) with ESMTP id JAA15971; Thu, 27 Feb 1997 09:11:08 -0700 (MST)
Message-Id: <199702271611.JAA15971@conceptual.com>
X-Mailer: exmh version 2.0gamma 1/27/96
To: mika ruohotie <bsdhack@shadows.aeon.net>
cc: hfwirth@ping.at (Helmut F. Wirth), freebsd-hackers@freebsd.org
Subject: Re: Pentium MMX Extensions, Changes to (g)as and gdb 
In-reply-to: Your message of "Thu, 27 Feb 1997 11:23:10 +0200."
             <199702270923.LAA07796@shadows.aeon.net> 
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Thu, 27 Feb 1997 09:11:08 -0700
From: "Russell L. Carter" <rcarter@consys.com>
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

> > Hello,
> > I made some changes to gas and gdb to support the new MMX extensions
> > for the Pentium P55C.
> 
> gee, cool. i didnt think we'd something about that this fast...
> 
> what i'm interested in is, does the mmx cpu run normal freebsd any
> faster? the 'make world' is probably the only thing one could use
> for measuring it in the "reality" level... i would assume just coz of
> larger L1 cache it is already faster.
> 

The MMX instructions themselves will not affect the speed of anything
resembling a "normal" code.  Intel's P5 cpus with MMX run faster because
they have a 2x bigger level1 cache and a few of the scheduling enhancements
that first showed up in P6.  Interestingly, the P6++ cpus with MMX (so-called
Pentium II) run slower than P6 at the same clock rate because of the
psuedo-on-chip level2 cache.

It's conceivable that pd libraries could
start showing up that use these instructions in specific assembly coded
routines.

I like the idea of putting the MMX instructions in though.

Russell