schmidt@SUMEX-AIM.STANFORD.EDU (Christopher Schmidt) (09/01/89)
Here is a brief summary of items I found interesting at AAAI-89 in
Detroit.
Expertelligence showed a British product, Procyon Common Lisp, a
Common Lisp for the Macintosh. I was particularly excited by it because it
included many Interlisp-inspired features. The most appealing to me was
DEdit, but it also included HistMenu, Inspector, EditBM, a debugger very much
like Apple Common LispUs, BreakAround and BreakWhen, a user-extensible
pretty-printer, on-line help, and BreakDown. The Macintosh-oriented features
were very much like Apple Common LispUs, but maybe a little bit better Eg.,
like ACL, you donUt need Inside Macintosh to draw graphics on the screen, but
Procyon takes it a step further and you donUt have to do any extra work to
make sure your windows get redrawn when they get exposed. The product comes
in three levels. The "Personal" configuration consists of the features
mentioned above, and costs $620 ($445 academic). The "Professional"
configuration adds CLOS, full color support, faster garbage collection, and
costs $1250 ($900 academic). The advantage of Procyon CLOS over PCL is that
it is much faster and the classes of built-in objects like the window system
arenUt hidden from the user. The "Developer" version adds a Foreign Function
Interface, Runtime Generator, and Interface Designer, and costs $2000 ($1450
academic). The minimal application is 512K in size; a fair amount less than,
say, HypercardUs overhead. They claim that their lisp is faster and more
compact than ACL. The Interface Designer was very good and similar to ACLUs.
A press release said that Action! would also run on top of Procyon. (Action
is an Expertelligence interface designer for the Explorer and Microexplorer.)
I thought that this product had everything going for it except the lack of
lightweight processes and the price. Eg., an academic site licence for Coral
Common Lisp cost us the same as only 3.4 copies of Procyon would. I tried to
talk Expertelligence down on price, but they didnUt budge. They do have a
promotion, though. If you trade-in Experlisp or ACL before October 15th, you
can get the Professional version for only $700. Also, the manuals look *very*
good.
The Apple booth had 40 MHz 68030 Daystar Digital accelerators in every
Macintosh. Daystar was also offering a promotion on 50 MHz accelerators such
that if you order by midnight tonight, you can get them for the price of the
40 MHz accelerator. The top-of-the-line 50 MHz accelerator lists for $6000;
or $7000 with a 40 MHz 68882. They were a little vague on academic pricing,
but said maybe 20%.
A company called Cognitive Software was selling a Neural Net system
for the macintosh. The appealing feature that distinguished it from other NN
systems in my mind was the Common Lisp (or rather a subset) programmability of
the nodes. A software-only version costs $695. Software for the Levco
Transputer co-processors costs $3995. Levco Transputers have an incremental
cost of about $1200 and you can fit up to 12 of them (on 3 boards) in a
Macintosh. As an Apple VAR they will sell you a 120 MIP "NeuroComputer" (a
loaded Mac II plus 12 Transputer processors) for $52,588.31. The mac itself
includes [the equivalent of?] the 12 MIP Daystar Digital accelerator mentioned
above. There is a 20% academic discount available.
Apple had flyers out for X11R3 under both A/UX and the Mac OS.
Apple Common Lisp 1.3 (which they persist in calling Allegro Common
Lisp) was announced. Since they didnUt have any flyers for it, I had to look
out for improvements. The most marked one I saw was that the Dialog Designer
(previously distributed in half-baked form) has been cleaned up and is really
quite useful now. You can re-design already coded dialogs, for example.
Regrettably, they still donUt have lightweight processes, either. They did
say that they will have IAC ready when Release 7.0 comes out, for what thatUs
worth.
A company called Gensym and Hewlett-Packard had the best booth
(entertainmentwise). They had a 1:96 scale model of the Biosphere II project
being built in Arizona. The model had heaters, sensors, and fans connected up
to 4 H-P computers running GensymUs G2 "real-time expert system." The
computers were running software developed for the real Biosphere, running a
combination of simulation and the "real" data/actuators on the model. Very
slick. G2 runs on a number of computers, including the mac (at $9000) and
costs as much as $18,000 per copy. It supports lightweight processes, but has
no escape to lisp; instead they offer foreign function interfaces for C and
Pascal.
SunUs SPE continues to look more and more like Interlisp-D. The
person who demonstrated it to me even said that he thought they might have a
structure editor in a couple of years. Most exciting for academic sites is
the new low price of $500-750 (down from $3000). Sun Common Lisp is up to
$4000, but I believe it is somewhat discountable.
Symbolics showed the MacIvory 2, which I was told doubled the speed of
last yearUs MacIvory. They introduced an Ivory 2-based VMEbus card that will
run in either a Sun-3 or Sun-4. The memory (10 or 20 Mb) is on the board
itself, so it can run a little better than 40/32 times faster than the
MacIvory 2. Symbolics also showed a new Common Lisp window standard called
CLIM, endorsed by them, International Lisp Associates, and another vendor
whose name escapes me. They showed the same application running on a
MacIvory, IBM PC, and Sun (?).
NTT showed the ELIS lisp machine. I think this was the first time
this lisp machine has been shown at AAAI, and would thus be the only *new*
machine introduced this year. (I think.)
TI had the biggest booth by far and gave away CLOS reference cards.
Their booth was dedicated primarily to showing-off customer applications.
The Info-1100 usersU group dinner was quite congenial. I was pleased
to meet in person many whom I have known only from this list.
Naturally, I welcome any corrections, criticism, and supplements to
the above. I have only only my faulty eyes, ears, and memory to rely on and
would like to hear the reactions of others!
--Christopher
-------lgm@cbnewsc.ATT.COM (lawrence.g.mayka) (09/05/89)
In article <MacMM.25595.12767.schmidt@sumex-aim.stanford.edu> schmidt@SUMEX-AIM.STANFORD.EDU (Christopher Schmidt) writes: > NTT showed the ELIS lisp machine. I think this was the first time >this lisp machine has been shown at AAAI, and would thus be the only *new* >machine introduced this year. (I think.) Does anyone have any further information about this machine? The very short writeup I saw referred to it as a multiuser workstation, a product of a joint venture between Nippon Tel & Tel and Matsushita. When will it be shipping, etc.? Send me email if you think the matter is too product-centered for a newsgroup. Lawrence G. Mayka AT&T Bell Laboratories lgm@ihlpf.att.com
"Frank_Shih.mvenvos"@XEROX.COM (09/08/89)
I read a short paper at their booth (I didn't keep the paper, so I'm doing this from memory), it described a microcoded machine which implements a Lisp interpreter (!?), directly. I don't recall the technology underlying the hardware as being unusual (I seem to recall something like 2 micron CMOS). It included a very few benchmarks comparing other (compiled) Lisp implementations (in particular, D-machines), versus their microcoded interpreter. I recall some of their benchmarks were "relatively good" (some were better than compiled Interlisp-on-D-machine). Since the paper didn't include compiled benchmarks for their machine, could it be that there IS NO low-level instruction set to compile into? So assuming the paper is describing the ELIS, it seems to be a rather unusual lisp machine. I seem to recall pricing in the very low 10s of thousands.... -Frank.