about LISP performance

For friendly off topic discussion not covered in a forum above.
Forum rules
No politics, please.
robespierre
Posts: 1570
Joined: Mon Sep 12, 2011 2:28 pm
Location: Boston

Re: about LISP performance

Unread postby robespierre » Fri Feb 28, 2014 1:09 am

Do all of the color demos (Finger Of God, Spheres) work on a NuVista? I thought some of them were closely tied into the L-Graphics hw, because they did pan and zoom effects at 1280x1024. (I seem to remember that 1440x1152 was also a setting.)
:PI: :O2: :Indigo2IMP: :Indigo2IMP:

User avatar
kjaer
Posts: 426
Joined: Wed May 07, 2008 7:47 pm
Location: Seattle, WA
Contact:

Re: about LISP performance

Unread postby kjaer » Fri Feb 28, 2014 3:50 pm

Finger Of God and Spheres definitely do. The hardware pan and zoom works, but is not very smooth. I'm not sure if this is a limitation of the NuVista or of the MacIvory (or NuBus for that matter). The only thing that doesn't seem to work is color map animation. But it might. I'm not sure what it's supposed to look like. I have a 3650 with full color, but it needs attention before I can play with it and compare to NuVista.

The NuVista and NuVista+ really only do broadcast resolution.
:OnyxR: :IRIS3130: :IRIS2400: :Onyx: :ChallengeL: :4D220VGX: :Indigo: :Octane: :Cube: :Indigo2IMP: :Indigo2: :Indy:

User avatar
kjaer
Posts: 426
Joined: Wed May 07, 2008 7:47 pm
Location: Seattle, WA
Contact:

Re: about LISP performance

Unread postby kjaer » Fri Feb 28, 2014 3:59 pm

I actually took some photos of the other demos too, just didn't upload them.
Attachments
image.jpg
NuVista+ color spheres & "wandering window" demo
:OnyxR: :IRIS3130: :IRIS2400: :Onyx: :ChallengeL: :4D220VGX: :Indigo: :Octane: :Cube: :Indigo2IMP: :Indigo2: :Indy:

robespierre
Posts: 1570
Joined: Mon Sep 12, 2011 2:28 pm
Location: Boston

Re: about LISP performance

Unread postby robespierre » Fri Feb 28, 2014 4:23 pm

Looks good!
The 3650 is a sweet machine.
:PI: :O2: :Indigo2IMP: :Indigo2IMP:

User avatar
geo
Posts: 527
Joined: Tue Jul 21, 2009 5:07 am
Location: Nagano-ken, Shiojiri-shi, Japan

Re: about LISP performance

Unread postby geo » Mon Mar 03, 2014 12:26 am

thanks kjaer!! very interesting!! hope there's more to come ;)
still cant imagine why lispm went cold :(
:Octane: (Sakura) :O2: (Sasuke) :1600SW: (Naruto) ... lil Jesse! (O2 laptop)
“Imagination is more important than knowledge.“ – A. Einstein

kubatyszko
Donor
Donor
Posts: 458
Joined: Sat Nov 12, 2011 3:18 am
Location: Los Angeles
Contact:

Re: about LISP performance

Unread postby kubatyszko » Mon Apr 28, 2014 11:08 pm

if you mean lisp machines - well they were great and I heard some are still in use, but "generic" CPU's picked up so fast and they became obsolete...
if you mean lispm as in Rainer Joswig who is probably one of the very few surviving LISP users - he's alive and well - just active in his own "world".

I can't remember but I think I have tried compiling sbcl or at least ccl a while back - I found the latest version that compiled just fine on IRIX (it was still *very* old), and that's the last thing I recall - since then I had to remove that hard disk because I suspect that was causing my Fuel to crash, might plug it in someday to see what did I get stuck on.

I would also like to have some more recent implementation available.

Cheers
Designer of Fuel ATX PSU adapter board : viewtopic.php?f=3&t=16732341
(now ships from USA), In stock!

robespierre
Posts: 1570
Joined: Mon Sep 12, 2011 2:28 pm
Location: Boston

Re: about LISP performance

Unread postby robespierre » Tue Oct 03, 2017 9:44 pm

kjaer wrote:I have compiled results to share, and there's no word for them but "hilarious". I ended up having to use n=50000000 (as in the original examples) for the compiled results, because n=50000 was too fast for the resolution of the timer.

MacIvory 2: 28
MacIvory 1: 54
MicroExplorer: 339


Necropost:
The reason for the poor showing from the MicroExplorer is probably that the counter overflows its fixnum size, forcing bignums to be allocated. As its words are only 32 bits, it has a smaller range for fixnums than the Ivory machines. I expect that if you use n= 16 million instead of 50 million you will see a dramatic improvement.
You can compare CL:MOST-POSITIVE-FIXNUM, which will be (1- (expt 2 31)) on the Ivory machines, but only (1- (expt 2 24)) on the Explorer.
:PI: :O2: :Indigo2IMP: :Indigo2IMP:

User avatar
vishnu
Donor
Donor
Posts: 3183
Joined: Sun Mar 18, 2007 3:25 pm
Location: Minneapolis, Minnesota USA

Re: about LISP performance

Unread postby vishnu » Tue Oct 03, 2017 10:49 pm

Speaking of which, with regard to the Explorer, if I'm remembering right symbolics was the first dot com ever to be registered. Could be wrong about that as my memory from that era is notoriously faulty. At the time I was working for FMC Corporation when we were (approximately) the 50th dot com to be registered. We thought we were soooo cool... 8-)
Project:
Temporarily lost at sea...
Plan:
World domination! Or something...

:Tezro: :Octane2:

robespierre
Posts: 1570
Joined: Mon Sep 12, 2011 2:28 pm
Location: Boston

Re: about LISP performance

Unread postby robespierre » Wed Oct 04, 2017 1:39 am

vishnu wrote:Speaking of which, with regard to the Explorer, if I'm remembering right symbolics was the first dot com ever to be registered.

The Internet certainly thinks so. In 1985 apparently.
Note that domain registration depends on the DNS system, which was codified in November 1983 by RFC 882. Symbolics was incorporated in April 1980, and issued its first public shares in November 1984 (15 million at $6).
:PI: :O2: :Indigo2IMP: :Indigo2IMP:

User avatar
kjaer
Posts: 426
Joined: Wed May 07, 2008 7:47 pm
Location: Seattle, WA
Contact:

Re: about LISP performance

Unread postby kjaer » Wed Oct 04, 2017 11:11 am

robespierre wrote:Necropost:
The reason for the poor showing from the MicroExplorer is probably that the counter overflows its fixnum size, forcing bignums to be allocated. As its words are only 32 bits, it has a smaller range for fixnums than the Ivory machines. I expect that if you use n= 16 million instead of 50 million you will see a dramatic improvement.
You can compare CL:MOST-POSITIVE-FIXNUM, which will be (1- (expt 2 31)) on the Ivory machines, but only (1- (expt 2 24)) on the Explorer.


Could be.

I started a project to benchmark 8-bit home micros, finding the first n primes via trial division (the point wasn't to find primes efficiently but rather to have some fixed amount of work that could be reasonably expressed with a limited BASIC interpreter). At one point I extended it to APL to compare that to BASIC on my IBM 5100, then LISP/80 and Laxen & Perry Forth on the Kaypro 10, and, well... it's gone out of control from there.

http://www.typewritten.org/Articles/Ben ... rimes.html

Anyway, back to my point: I'll excerpt the (n=1000) LISP results since they're all mixed in with the others:

Code: Select all

Apple   Mac Mini        1250 MHz PowerPC G4    1048576 KB  Macintosh Common LISP 5.0            0.3 variant B • OS X 10.3.9
IBM     PC Server 320    133 MHz Pentium        262144 KB  Allegro Common LISP Lite 5.0.1       3   variant B • compiled
ALR     Evolution IV/e    66 MHz 80486 DX2       65536 KB  TI PC Scheme 3.03                   16   conventional memory model
Apple   Macintosh II             microExplorer   12288 KB  microExplorer System Release 6.1    22   variant B • compiled
ALR     Evolution IV/e    66 MHz 80486 DX2       65536 KB  TI PC Scheme 3.03                   28   extended memory model
Apple   Macintosh IIfx           MacIvory 2      49152 KB  Genera 8.3                          30   variant B • compiled
ALR     Evolution IV/e    66 MHz 80486 DX2       65536 KB  TI PC Scheme 3.03                   39   expanded memory model
Apple   Macintosh IIfx    40 MHz 68030           32768 KB  Pearl LISP 1.1                      73   24-bit addressing
Apple   Macintosh IIfx           MacIvory 2      49152 KB  Genera 8.3                         242   variant B • interpreted
Xerox   1186                     Mesa             3712 KB  enVoS LISP Medley 1.0              283   XCL • variant B • compiled
DEC     VAXstation 2000    5 MHz MicroVAX II      6144 KB  VAX LISP 2.0                       336   variant B • compiled
Apple   Macintosh II             microExplorer   12288 KB  microExplorer System Release 6.1   343   variant B • interpreted


And for giggles, n=100:

Code: Select all

Apple   Macintosh IIfx   40 MHz 68030        32768 KB  Pearl LISP 1.1                      0.7 24-bit addressing
Xerox   1186                    Mesa          3712 KB  enVoS LISP Medley 1.0               2.1 XCL • variant B • compiled
DEC     VAXstation 2000   5 MHz MicroVAX II   6144 KB  VAX LISP 2.0                        3   variant B • compiled
Kaypro  10                4 MHz Z80A            64 KB  Software Toolworks LISP/80 1.00   183   variant B • 21 GCs


I'm actually not entirely convinced the microExplorer is faster than the MacIvory, generally. It certainly is on this benchmark, though, probably because of the MacIvory 1 & 2 having their memory on the NuBus rather than local to the processor, as on the microExplorer.
:OnyxR: :IRIS3130: :IRIS2400: :Onyx: :ChallengeL: :4D220VGX: :Indigo: :Octane: :Cube: :Indigo2IMP: :Indigo2: :Indy:

User avatar
kjaer
Posts: 426
Joined: Wed May 07, 2008 7:47 pm
Location: Seattle, WA
Contact:

Re: about LISP performance

Unread postby kjaer » Wed Oct 04, 2017 11:27 am

kjaer wrote:Finger Of God and Spheres definitely do. The hardware pan and zoom works, but is not very smooth. I'm not sure if this is a limitation of the NuVista or of the MacIvory (or NuBus for that matter). The only thing that doesn't seem to work is color map animation. But it might. I'm not sure what it's supposed to look like. I have a 3650 with full color, but it needs attention before I can play with it and compare to NuVista.


As long as we're reviving this thread... I was able to verify that the NuVista support for MacIvory doesn't entirely work with NuVista+. Color map animation for example does work on the NuVista, but not NuVista+. There were a couple other things that looked different, but I didn't do a close side-by-side comparison at the time. It's still on my list of projects.

Also learned that non-RGB (composite, Y/C) output from the NuVista requires a separate box, that isn't required with NuVista+. I suspect this has something to do with why those outputs are greyscale only with NuVista+, on Genera. This is something else that's still on my list of projects.
:OnyxR: :IRIS3130: :IRIS2400: :Onyx: :ChallengeL: :4D220VGX: :Indigo: :Octane: :Cube: :Indigo2IMP: :Indigo2: :Indy:


Return to “Everything Else”

Who is online

Users browsing this forum: No registered users and 2 guests