Page 1 of 1

How much LISP fits on the parallellas ?

PostPosted: Wed Dec 19, 2012 3:11 pm
by theover
I'd find it interesting to have some sort of parallel lisp on the board, especially to run "Maxima" on (symbolic algebraic math program http://maxima.sourceforge.net/), but isn't the memory bandwidth going to be in the way? I mean does a list interpeter kernel run on 32 kB? I doubt it, so there might be a lot of work to make sense of such idea in practice.

Of course the idea of a new Connection Machine (http://en.wikipedia.org/wiki/Connection_Machine) may help Finck and Reese do some good work, and I would find it fun to work on.

Theo V.

Re: How much LISP fits on the parallellas ?

PostPosted: Thu Dec 20, 2012 1:54 am
by 8l

Re: How much LISP fits on the parallellas ?

PostPosted: Thu Dec 20, 2012 4:53 pm
by lispm
theover wrote:I'd find it interesting to have some sort of parallel lisp on the board, especially to run "Maxima" on (symbolic algebraic math program http://maxima.sourceforge.net/), but isn't the memory bandwidth going to be in the way? I mean does a list interpeter kernel run on 32 kB? I doubt it, so there might be a lot of work to make sense of such idea in practice.

Of course the idea of a new Connection Machine (http://en.wikipedia.org/wiki/Connection_Machine) may help Finck and Reese do some good work, and I would find it fun to work on.


One of the possibilities would be to use some kind of data-parallel Lisp extension (similar to what the Connection Machine had). There would be a main Lisp running on the ARM processor which can compute over graphs and vectors using the parallel cpus. The model for this would be *Lisp, Connection Machine Lisp, Paralation Lisp or others.

Re: How much LISP fits on the parallellas ?

PostPosted: Fri Jan 11, 2013 5:23 pm
by peter.klicman
Star Lisp was also the fist thing I thought of. So maybe you find Steve Omohundro's "Essential Star Lisp Manual"
- describing PVARS - of interest.

http://omohundro.files.wordpress.com/2009/03/omohundro86_the_essential_starlisp_manual.pdf

Peter

Re: How much LISP fits on the parallellas ?

PostPosted: Wed Feb 13, 2013 4:22 pm
by theover
IIRC the CM had "xectors": associative vectors.

Re: How much LISP fits on the parallellas ?

PostPosted: Thu Feb 14, 2013 11:47 pm
by janeTA
Yes, xectors are nicely explained in Danny Hillis' book "The Connection Machine" http://www.amazon.com/Connection-Machin ... 0262081571 (and elsewhere)? I suspect Lisp on Parallella could learn a lot from TM's CmLisp for SIMD abstraction/s alone?

Re: How much LISP fits on the parallellas ?

PostPosted: Mon Apr 29, 2013 4:04 pm
by Eismcc
Starlisp PVARs are an interesting choice. I'd love to help get that going.

Re: How much LISP fits on the parallellas ?

PostPosted: Wed Sep 10, 2014 7:33 pm
by Madlisper
:D please do not forget that cm used to have a frontend. The idea here is that we could implement some parallelLisp DIRECTLY! Just imagine the powerbof some emacs running a psetq with evaluated second forms directly!
I dream of this .. Hope you too.
At the first sight It seems feasible within a ┬┤correct┬┤ community :oops: correct does not mean linux-like; once a month a small piece of code would bring enough. Even for busy people knowing that enthusiasm will hopefully be enough strong to biring up (finally) those libraries that we all hope that some others would build :mrgreen:
Whatvdo you think?

Re: How much LISP fits on the parallellas ?

PostPosted: Wed Sep 10, 2014 7:40 pm
by Madlisper
One addition: I'dlike to call this "aquarella" like successive stratuses over a white paper. Each one with its own core mask... Sharing memory and grouping results like series and generators.

Re: How much LISP fits on the parallellas ?

PostPosted: Sun Sep 14, 2014 7:58 pm
by xilman
theover wrote:I mean does a list interpeter kernel run on 32 kB? I doubt it, so there might be a lot of work to make sense of such idea in practice.
Some of us old farts remember when usable LISP implementations ran on Z80 machines with 32K memory...

Incidentally, I've been toying with the idea of recreating CP/M for the Epiphany, purely for the fun of it. Some of my first paid emplyoment was writing "scientific" software for Z80 machines with 32-48K RAM, CP/M and floppy drives which held 240K bytes per side.

We had it tough. And you try telling the youth of today that, and they won't believe you.