Pythoniac: Thoughts on a hardware Python processor

[David LeBlanc]

> -----Original Message-----
> From: python-list-admin at python.org
> [mailto:python-list-admin at python.org]On Behalf Of Paul Rubin
> Sent: Wednesday, June 26, 2002 18:49
> To: python-list at python.org
> Subject: Re: Pythoniac: Thoughts on a hardware Python processor
>
>
> I don't think it makes sense to stick with existing Python bytecodes
> for hardware execution.  Take a look at the work done on Lisp machines
> in the 70's and 80's instead.

Ah yes, thanks for reminding me. The Self project claimed to have gotten
Self to run in the mid 90's percentile of "C speed", largely by making as
many primatives as possible actual machine ops of the host processor - not
too good for portability. OTOH, Self has many of the complexities of Python,
and if this is true, this is blazing speed. For anyone that's interested,
there is still a Self website (and by darned, a new release this year!) at
http://research.sun.com/self/index.html and there are some good papers there
on their compiler techniques.
http://research.sun.com/self/release/optionalFiles.html offers the sources
to the Self VM (in C++ and "Sparc-asm")

Of course, some of the Self guys went on to another project at Sun: Java.
I've read somewhere that Java HotSpot JIT compiler tech came out of the Self
work (oops, there it is on one of the Self pages).

Actually, Java and this idea of hosting Python on Forth brings up an
interesting question: If Sun can't make a VM that will run everywhere, why
would a Forth based Python implementation do any better?

Dave LeBlanc
Seattle, WA USA