From: Michael J. Ferrador N2KRA
Subject: Modern Workstations -vs- Old Lisp Machines
Date:
Message-ID: <D50ADr.L8r@news2.new-york.net>
How do Modern (300Mhz Alpha, and 100Mhz 601,R4X00,Pentium) machines
compare with Classic Lisp Workstations? Are there any modern 0.5 micron
100Mhz revamps of Lisp Processors?
What other Lisp Workstations were there, I'm only heard of Symbolics
and TI Explorer.
What was a typical configuration? I was almost given a Symbolics, but
the .5Gig HD was removed, and the owner said that it was really meant to
be hooked up to a server with a Gig. Was that Gig all the shared Lisp OS,
Utils, and Environment? Then the local storage your own apps?
--
Q: What's the best strategy for the PowerPC Alliance to topple Intel?
A: FDIV (Divide) and Conquer
From: Barry Margolin
Subject: Re: Modern Workstations -vs- Old Lisp Machines
Date:
Message-ID: <3jeg14$8ir@tools.near.net>
In article <··········@news2.new-york.net> ·····@starcomm.overleaf.com (Michael J. Ferrador N2KRA) writes:
> How do Modern (300Mhz Alpha, and 100Mhz 601,R4X00,Pentium) machines
>compare with Classic Lisp Workstations? Are there any modern 0.5 micron
>100Mhz revamps of Lisp Processors?
The most recent Symbolics workstations, the Ivory-based XL1200 that came
out about 3-4 years ago, could run Lisp a little faster than a Sparcstation
1+ running Lucid CL. Since then, conventional processors have increased
their speed by an order of magnitude.
Shortly before they went bankrupt, Symbolics implemented an Ivory emulator
for the Alpha, and ported their system to it. I think it was about as fast
as the native Ivory implementation.
> What other Lisp Workstations were there, I'm only heard of Symbolics
>and TI Explorer.
Lisp Machines, Inc.'s Lambdas.
> What was a typical configuration? I was almost given a Symbolics, but
>the .5Gig HD was removed, and the owner said that it was really meant to
>be hooked up to a server with a Gig. Was that Gig all the shared Lisp OS,
>Utils, and Environment? Then the local storage your own apps?
The local disk normally contained the Lisp image, swap space, microcode,
FEP code, and boot scripts. Source files and loadable files were generally
stored on a file server (which could be a Lisp machine, but NFS and FTP
servers could also be used). For machines with limited disk space, the
Lisp image could be loaded over the network (this is only useful if you
need to be able to boot different images, since netbooting copies the
loaded image into swap space so you need enough space for the largest image
you might boot).
--
Barry Margolin
BBN Internet Services Corp.
······@near.net
In article <··········@tools.near.net> ······@nic.near.net (Barry Margolin) writes:
> What other Lisp Workstations were there, I'm only heard of Symbolics
>and TI Explorer.
Lisp Machines, Inc.'s Lambdas.
Xerox D-machines running Interlisp.
Bill
--
Bill Janssen <·······@parc.xerox.com> (415) 812-4763 FAX: (415) 812-4777
Xerox Palo Alto Research Center, 3333 Coyote Hill Rd, Palo Alto, CA 94304
URL: ftp://parcftp.parc.xerox.com/pub/ilu/misc/janssen.html
In article <··········@tools.near.net> ······@nic.near.net (Barry Margolin) writes:
> What other Lisp Workstations were there, I'm only heard of Symbolics
>and TI Explorer.
Lisp Machines, Inc.'s Lambdas.
Xerox D-machines running Interlisp.
Bill
--
Bill Janssen <·······@parc.xerox.com> (415) 812-4763 FAX: (415) 812-4777
Xerox Palo Alto Research Center, 3333 Coyote Hill Rd, Palo Alto, CA 94304
URL: ftp://parcftp.parc.xerox.com/pub/ilu/misc/janssen.html
Date: Mon, 6 Mar 1995 06:37:01 GMT
From: ·····@starcomm.overleaf.com (Michael J. Ferrador N2KRA)
How do Modern (300Mhz Alpha, and 100Mhz 601,R4X00,Pentium) machines
compare with Classic Lisp Workstations?
Around 1992, a VLSI design system we were developing would get roughly
the same performance on a 40 MHz TI Explorer 2+ and on a Sparc-2 running
Harlequin Lispworks or Franz Allegro. The Explorer's superior GC
performance made up for smaller memory (24 to 32 MB compared to 64 to
128 MB on the Suns), and the code was heavily object-oriented.
Last week, a little search program I wrote, heavy on list and hash-table
operations, ran in 60 seconds on both my 16 MB Explorer 2+ and on a 128
MB Sparc-10/41 running CMU CL -- generational GC vs stop-and-copy. (It
took much longer in our slightly old version of Allegro, because it
tickled a hash-table performance bug that has since been fixed; I don't
have a Harlequin license for this machine.)
However, Suns continue to get faster, and Explorers don't.
I just happen to have two plaques on my desk, one with our Lisp chip,
one with a TMS320C80 (MVP). The Lisp chip is 1.25-micron, two-level
metal. The MVP is 0.6-micron, four-level metal (I think); it's four
times the area and has eight times as many transistors.
What other Lisp Workstations were there, I'm only heard of Symbolics
and TI Explorer.
Lisp Machines Inc also made machines; TI actually licensed some of the
technology from them, and they sold Explorers as one part of their
product line. Xerox sold D-machines (Dorado? Dolphin?) that were
lispms of West-Coast lineage, running Interlisp.
What was a typical configuration?
We sold Explorers with two 140-MB disks initially, but they did make
bigger ones later. Mine has four 140s, which is plenty of room for the
37 MB Lisp image (OS, utilities, environment; TCP, NFS, X), 128 MB
swap, 60 MB file space, and random little microcode and other
partitions. The boot menu offers a "network" option, but that was never
implemented.
By the way, the other week I pulled a memory board that appeared to be
getting flaky -- it's a 2-MB board, about 10x14 inches with 288 little
surface-mount 4164 memory chips and a bunch of assorted latches, and
it had been in almost continuous operation since November 1985.
Paul Fuqua
Texas Instruments, Dallas, Texas ··@hc.ti.com
"You get what you settle for." -- Thelma & Louise
In article <··········@news2.new-york.net>,
Michael J. Ferrador N2KRA <·····@starcomm.overleaf.com> wrote:
> How do Modern (300Mhz Alpha, and 100Mhz 601,R4X00,Pentium) machines
>compare with Classic Lisp Workstations? Are there any modern 0.5 micron
>100Mhz revamps of Lisp Processors?
>
> What other Lisp Workstations were there, I'm only heard of Symbolics
>and TI Explorer.
Someone else has already mentioned LMI, so I won't repeat it...
There was also the Xerox 1100 series (1108 Dandelion, 1109 Dandetiger
(don't ask), 1132 Dorado and later 1186 Daybreak). These machines ran
Interlisp-D; the 1108 and later 1186 were substantially cheaper than
the Symbolics/TI/LMI machines, and consequently sold rather better at
least this side of the pond.
The 1108 had 4Mb core (don't know what it could be expanded to), 80Mb
disk, and a megapixel mono head about 20" diagonal. It was claimed to
run 2 MIPS, and my memory of it was that you got the same sort of
reponsiveness as from roughly a 386dx40 running Windows. Until it
started to swap, that it. When it started to swap it ran like treacle.
The 1186 fundamentally replaced the 1108. It was about twice as fast,
had a slightly bigger (and much better) screen, and was a lot cheaper.
It could in principle be expanded to 16Mb core, and would have run a
lot better with that, but I never used one which had been expanded.
Both machines were seriously nice development tools, because the
Interlisp-D system was incredibly easy to work with and to develop
complex systems quickly. In particular, the LOOPS system really did
make software re-use trivial.
I believe the Dorado was a machine to drool over, but I guess they
liked to keep them clean because I never got near one :-}
A derivative of Interlisp-D (now called Medley) is available for Intel
architecture, including a Common LISP compiler integrated with the
Interlisp system. On a good 486 with plenty of core this should be a
marvelous tool; I confess I haven't tried it, because I don't use
Intel machines at home and I haven't persuaded my employers that it
would be of value to them :-(.
--
------- ·····@rheged.dircon.co.uk (Simon Brooke)
'there are no solutions, only precipitates'
From: Jeff Dalton
Subject: Re: Modern Workstations -vs- Old Lisp Machines
Date:
Message-ID: <D5Dzoq.C0@cogsci.ed.ac.uk>
In article <··········@rheged.dircon.co.uk> ·····@rheged.dircon.co.uk (Simon Brooke) writes:
>In article <··········@news2.new-york.net>,
>Michael J. Ferrador N2KRA <·····@starcomm.overleaf.com> wrote:
>> How do Modern (300Mhz Alpha, and 100Mhz 601,R4X00,Pentium) machines
>>compare with Classic Lisp Workstations? Are there any modern 0.5 micron
>>100Mhz revamps of Lisp Processors?
>>
>> What other Lisp Workstations were there, I'm only heard of Symbolics
>>and TI Explorer.
>
>Someone else has already mentioned LMI, so I won't repeat it...
>
>There was also the Xerox 1100 series [...]
You might also count the Perq/Accent/SPICE Lisp combination.
-- jd