From owner-freebsd-hackers  Fri Dec 22 23:44:51 1995
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id XAA23793
          for hackers-outgoing; Fri, 22 Dec 1995 23:44:51 -0800 (PST)
Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.211])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id XAA23787
          Fri, 22 Dec 1995 23:44:46 -0800 (PST)
Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id AAA21670; Sat, 23 Dec 1995 00:40:17 -0700
From: Terry Lambert <terry@lambert.org>
Message-Id: <199512230740.AAA21670@phaeton.artisoft.com>
Subject: Re: SMP Motherboards
To: gclarkii@FreeBSD.ORG
Date: Sat, 23 Dec 1995 00:40:17 -0700 (MST)
Cc: freebsd-hackers@FreeBSD.ORG
In-Reply-To: <199512222121.PAA18590@main.gbdata.com> from "Gary Clark II" at Dec 22, 95 03:21:09 pm
X-Mailer: ELM [version 2.4 PL24]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-hackers@FreeBSD.ORG
Precedence: bulk

> Does anyone here have an idea on what would be the cheapest SMP compatible
> motherboard?  This board will only be for hacking on so a 486 based system
> would be just fine (I have axp 4 486DX-33 here).  ISA and VLB is also fine.

Jack Vogel, who made patches to the 28 Oct 94 tree for SMP, has an ASUS
dual P90 EISA/PCI.

I've got an ASUS dual P90 ISA/PCI (my only intent was to hack SMP stuff,
so I went budgo -- total system cost without monitor was in the $1990
range).

If you do a 486, you should be aware that the SMP patches that Jack Vogel
passed around definitely assume Pentiums.  Further, theyy assum internal
APIC's.  Finally, they assume a BIOS memory area and an ID that is not
the ID in the spec.

So if you go hacking, be aware that the patches you can get your hands
on right now are far from general.

There is a nice problem between L1 cache proficiency and scheduling
anonymity in the current code, since there is no way for a process to
express a processor preference.  This is both good and bad, since it
trades extreme scheduler complexity for basically a stack hack and
the lieklihood of increased overhead on process rescheduling, especially
when a preeemption occurs because of a quantum expiration rather than
a blocking event.

If you are planning on hacking locore.s and mpcore.s code, then a 486
might be OK.  Otherwise, you'll want to go for a P90, and then most
likely an ASUS.

Finally, if you want to run right now without some agregious hacks, you
will need to sup a CVS tree and check out the code from 28 Oct 94 and
apply Jack's patches, plus the mpasm.h that Jack forgot initially, and
the small ordering hacks for the binary code #included in the mpcore.s
file (it is generated from post-assembly of "mpboot").


					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.