Emulators

Warren Toomey wkt at cs.adfa.edu.au
Fri Jan 21 12:16:40 AEST 2000


In article by emanuel stiebler:
> Hi,
> Anybody here, who made some benchmarks of the different simulators (supnik,
> apout, ...)
> 
> What I'm looking for is something like:
> supnik version xxx on pentium 2 350 MHz using linux, is xxx times faster
> than a 11/73.
> 
> cheers & thanks,
> emanuel

Here's my no-numbers-just-gut-feelings of the various PDP-11 emulators.
John Wilson's Ersatz is probably the fastest; it's written in assembly
code, and so gains a fair bit that way.

Second would be the Begemot emulator. They've unrolled the instruction
decode loop heavily, and that helps a lot.

Bob Supnik's emulator would be the slowest of the three. However, it's
still not that slow, may 1/3 the speed of Ersatz.

Apout can't be compared to the above 3 emulators, because it doesn't emulate
peripherals nor supervisor mode. User-mode instructions run at about the
same speed as Supnik's emulator, but system calls are done by native code.

The sole benchmark I have is: FreeBSD identifies my desktop box as
Pentium II/Xeon/Celeron (348.93-MHz 686-class CPU). Using Apout, I can
compile the 2.11BSD GENERIC kernel in 4 minutes 15 seconds.

I'll try building Supnik and Begemot and getting comparative results.

Last comment: all the simulators have strengths & shortcomings, and that
applies not just to ease of use but also to CPU, I/O performance etc. You
really have to try them all & find the one that suits you.

	Warren

Received: (from major at localhost)
	by minnie.cs.adfa.edu.au (8.9.3/8.9.3) id NAA35744
	for pups-liszt; Fri, 21 Jan 2000 13:44:42 +1100 (EST)
Received: from begemot.org (negara.cs.waikato.ac.nz [130.217.248.112])
	by minnie.cs.adfa.edu.au (8.9.3/8.9.3) with ESMTP id NAA35740
	for <pups at minnie.cs.adfa.edu.au>; Fri, 21 Jan 2000 13:44:20 +1100 (EST)
Received: (from joerg at localhost)
	by begemot.org (8.9.3/8.9.2) id PAA71802;
	Fri, 21 Jan 2000 15:45:24 +1300 (NZDT)
	(envelope-from joerg)
Date: Fri, 21 Jan 2000 15:45:24 +1300
From: "Joerg B. Micheel" <joerg at begemot.org>
To: Warren Toomey <wkt at cs.adfa.edu.au>
Cc: Unix Heritage Society <pups at minnie.cs.adfa.edu.au>, joerg at begemot.org,
        "Hartmut B. Brandt" <brandt at fokus.gmd.de>
Subject: Re: Emulators
Message-ID: <20000121154524.A71774 at begemot.org>
References: <034201bf63ad$8ad414b0$5d01a8c0 at p2350> <200001210216.NAA53527 at henry.cs.adfa.edu.au>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <200001210216.NAA53527 at henry.cs.adfa.edu.au>; from Warren Toomey on Fri, Jan 21, 2000 at 01:16:40PM +1100
Organization: Begemot Computer Associates
Operating-System: ... powered by FreeBSD
Sender: owner-pups at minnie.cs.adfa.edu.au
Precedence: bulk

On Fri, Jan 21, 2000 at 01:16:40PM +1100, Warren Toomey wrote:
> In article by emanuel stiebler:
> > Hi,
> > Anybody here, who made some benchmarks of the different simulators (supnik,
> > apout, ...)
> > 
> > What I'm looking for is something like:
> > supnik version xxx on pentium 2 350 MHz using linux, is xxx times faster
> > than a 11/73.

Us too! :-)

The problem is that it doesn't scale that simple. Each and every
instruction has the parsing overhead. Next comes execution overhead.
You'll find that the parsing is pretty constant, no matter whether
it is a NOP or some sophisticated MUL command. The execution speed
varies heavily, very often it is alot faster than the original
hardware. IO has seen a tremendous speedup, we can benefit here
from todays hardware alot. Just remeber how long it took to get
a prompt or echo when hitting the keyboard. As a result, the
original feeling of the real machine is lost, very unfortunate.

As a rough summary, simple commands do not improve (much), whereas
everything complex speeds up with the emulator. Harti has done quite
a bit of testing on different instructions and compared them to an
LSI11/73 (KDJ11A). The emulator was run on a i486 at the time. Have
a look at the p11 distribution, it should be in Tests somewhere.

Regards,
	Joerg
-- 
Joerg B. Micheel			Email: <joerg at begemot.org>
Begemot Computer Associates		Phone: +64 7 8562148
40 Masters Avenue, Hillcrest		Fax:   +64 7 8562148
Hamilton, New Zealand			Pager: +64 868 38222

Received: (from major at localhost)
	by minnie.cs.adfa.edu.au (8.9.3/8.9.3) id DAA39169
	for pups-liszt; Sat, 22 Jan 2000 03:08:07 +1100 (EST)
Received: from uni02du.unity.ncsu.edu (uni02du.unity.ncsu.edu [152.1.2.66])
	by minnie.cs.adfa.edu.au (8.9.3/8.9.3) with ESMTP id DAA39165
	for <pups at minnie.cs.adfa.edu.au>; Sat, 22 Jan 2000 03:07:42 +1100 (EST)
From: rdkeys at unity.ncsu.edu
Received: (from rdkeys at localhost)
          by uni02du.unity.ncsu.edu (8.8.4/EC02Jan97)
	  id LAA12512; Fri, 21 Jan 2000 11:07:37 -0500 (EST)
Message-Id: <200001211607.LAA12512 at uni02du.unity.ncsu.edu>
Subject: Anyone know what a Masscomp 5400 (54S-01) is?
To: pups at minnie.cs.adfa.edu.au
Date: Fri, 21 Jan 2000 11:07:37 -0500 (EST)
Cc: rdkeys at unity.ncsu.edu
X-Mailer: ELM [version 2.5 TKL/POP PL2]
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-pups at minnie.cs.adfa.edu.au
Precedence: bulk

On a surplus junket, today, I ran across a 2 dollar chassis that was
listed as a Masscomp 5400 (54S-01) computer.  It looked rather like
a DECish based thing with dual height cards, like some sort of laboratory
digital aquisition machine.  On the long-shot that it is some sort of
PDP-11ish thing, anyone have any recollection or pointers to any info
on that kind of a Masscomp machine? 

Thanks

Bob

Received: (from major at localhost)
	by minnie.cs.adfa.edu.au (8.9.3/8.9.3) id DAA42516
	for pups-liszt; Sat, 22 Jan 2000 03:45:56 +1100 (EST)
Received: from lostwithiel.cley.com (lostwithiel.cley.com [212.240.242.98])
	by minnie.cs.adfa.edu.au (8.9.3/8.9.3) with ESMTP id DAA42512
	for <pups at minnie.cs.adfa.edu.au>; Sat, 22 Jan 2000 03:45:33 +1100 (EST)
Received: (from tfb at localhost)
	by lostwithiel.cley.com (8.9.1b+Sun/8.9.1) id QAA04637;
	Fri, 21 Jan 2000 16:44:35 GMT
X-Mailer: 21.1 (patch 8) "Bryce Canyon" XEmacs Lucid (via feedmail 8 I);
	VM 6.72 under 21.1 (patch 8) "Bryce Canyon" XEmacs Lucid
From: Tim Bradshaw <tfb at cley.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Message-ID: <14472.36082.530024.331321 at cley.com>
Date: Fri, 21 Jan 2000 16:44:34 +0000 (GMT)
To: rdkeys at unity.ncsu.edu
Cc: pups at minnie.cs.adfa.edu.au
Subject: Re: Anyone know what a Masscomp 5400 (54S-01) is?
In-Reply-To: <200001211607.LAA12512 at uni02du.unity.ncsu.edu>
References: <200001211607.LAA12512 at uni02du.unity.ncsu.edu>
Content-Type: text/plain; charset=us-ascii
Sender: owner-pups at minnie.cs.adfa.edu.au
Precedence: bulk

* rdkeys  wrote:
> On a surplus junket, today, I ran across a 2 dollar chassis that was
> listed as a Masscomp 5400 (54S-01) computer.  It looked rather like
> a DECish based thing with dual height cards, like some sort of laboratory
> digital aquisition machine.  On the long-shot that it is some sort of
> PDP-11ish thing, anyone have any recollection or pointers to any info
> on that kind of a Masscomp machine? 

Masscomps were 68k based machines, they had a whole bunch of stuff for
real-time and data-acquistion type stuff.  They ran something called
RTU -- real-time Unix -- which was a weirdo sysv / BSD hybrid, not fun
to use.  If the 5400 is the machine I remember it's a 68020 machine
but it may have a lot of cards for other stuff in.  If it *is* the
machine we had it was deeply unreliable.  Masscomp got bought by
someone else later on but I forget who, so they sometimes get badged
as some other make.

I would run away, fast.

--tim




More information about the TUHS mailing list