From: William Annis
Subject: Lisp Machine Question (and some geekery).
Date: 
Message-ID: <4elb0d$1ns8@news.doit.wisc.edu>
	Section 2-20 of the Lisp FAQ says the following:

>                                                          The
>TI and Symbolics Lisp machines are currently available as cards that
>fit into Macintosh computers (the so-called "Lisp on a chip").

Then procedes to tell me nothing about how I would get such a thing, and
what sort of costs one might expect.  Anyone know?  Are these things
even still available?

	Now for geekery: 

	Last week a friend of mine informed me (in remarkably respectful
tones, considering his usual dislike of Lisp) that our Engnineering department
had taken a Sybolics machine that hadn't been turned on in something like
six years and "put it in the hall."  Being put in the hall is certain
death to a machine, and the noble 3600 was savaged almost immediately.  It
was lost by the time I heard about it, it's components scattered across
the land like Osiris' body parts.

	However, the 36-bit spirit survived: I liberated about 2 feet of
the manuals.  The architecture of the thing is fascinating.  It has left
me with the desire to use MACH to create a Virtual Lisp Machine, sitting
on top of MACH, with a bcode interpreter which more or less mimics the
3600 (or something).  Has anyone ever tried this?  None of the manuals
I could find had a complete 3600 Instruction set.  Does anyone know if such
a thing existed (external to Symbolics, at any rate)?

	I never imagined that my obsession with this language would lead
me to seriously contemplate mucking about with OS programming.


--
William S. Annis                               SysAdmin
Ofc: 608 265-5343 	      ·······@neurosim.wisc.edu
Department of Neurology             Neurosimulation Lab

From: Dan Rabin
Subject: Re: Lisp Machine Question (and some geekery).
Date: 
Message-ID: <danrabin-3001961349290001@a101021.sfo1.as.crl.com>
In article <···········@news.doit.wisc.edu>, ·······@neurosim.wisc.edu
(William S. Annis) wrote:

>         However, the 36-bit spirit survived: I liberated about 2 feet of
> the manuals.  The architecture of the thing is fascinating.  It has left
> me with the desire to use MACH to create a Virtual Lisp Machine, sitting
> on top of MACH, with a bcode interpreter which more or less mimics the
> 3600 (or something).  Has anyone ever tried this?  None of the manuals
> I could find had a complete 3600 Instruction set.  Does anyone know if such
> a thing existed (external to Symbolics, at any rate)?
> 
>         I never imagined that my obsession with this language would lead
> me to seriously contemplate mucking about with OS programming.


The 3600 series were microcoded machines with a writable control store. 
At boot time the machine would load the current microcode (stored on the
disk in a partition administered by the FEP file system).  This microcode
defined the macro instruction set, and it changed from release to release
of the system.  The tools and documentation for writing and editing
microcode were not part of the system released to normal users.  For some
optional features, such as Prolog or (I seem to recall) hardware add-ons,
there were different microcode files.  I also seem to recall that the
Prolog release's microcode was so big that customers had to buy more
writable control store in order to accommodate it. 

To give an example, I recall once disassembling some graphics code, and
finding that there were macro-instructions such as "bitblt-long-row" and
"bitblt-short-row".  There were specialized macro-instructions for
floating point, array access, and possibly GC.  The collection of such
specialized instructions would be tuned for performance from release to
release.   

The bottom line is that Mr. Annis has some serious archaeology to do.  It
might be easier just to emulate the documented subprimitives (many of
which corresponded to a single macro-instruction).  

Alas.

  -- Dan Rabin
From: William D. Gooch
Subject: Re: Lisp Machine Question (and some geekery).
Date: 
Message-ID: <goochb.266.0013C03F@rwi.com>
In article <···········@news.doit.wisc.edu> ·······@jackson (William Annis) writes:

>        Section 2-20 of the Lisp FAQ says the following:

>>                                                          The
>>TI and Symbolics Lisp machines are currently available as cards that
>>fit into Macintosh computers (the so-called "Lisp on a chip").

>Then procedes to tell me nothing about how I would get such a thing, and
>what sort of costs one might expect.  Anyone know?  Are these things
>even still available?

If you had asked 10 days ago, I could have offered to sell you 
a Symbolics MacIvory, but the several I had are now all sold.
Symbolics does however still exist, and sells these.

>....    However, the 36-bit spirit survived: I liberated about 2 feet of
>the manuals.  The architecture of the thing is fascinating.  It has left
>me with the desire to use MACH to create a Virtual Lisp Machine, sitting
>on top of MACH, with a bcode interpreter which more or less mimics the
>3600 (or something).  Has anyone ever tried this?  None of the manuals
>I could find had a complete 3600 Instruction set.  Does anyone know if such
>a thing existed (external to Symbolics, at any rate)?

Sort of.  There is a version of Genera for the DEC Alpha; in fact
they are talking of a new release being worked on.  It wouldn't
really make much sense to emulate the instruction set or to try
to carry over 32+ bit word capabilities to a 32-bit architecture,
since you'd surely lose to Common Lisps written more directly 
for conventional hardware.

Better you should just buy a MacIvory, or even a used 36xx. 
The latter are generally not expensive, and Symbolics is 
apparently offering software upgrades for old machines at 
a reasonable price nowadays.