True, that quote refers to CLOS, but to me the implication is that the Java
implementation is defective with respect to on-line redefining of lisp
programs (e.g. compiling and running something just created on the fly) and
linking them into the running program.
This is actually what I expect as a consequence of the Java VM. So if you
don't use such facilities, I think you could probably be ok. It would be
like forbidding the use of defun except in previously compiled files.
For some applications this may be OK. After all, Fortran and C operate that
way. And if your view of lisp is that it is just Fortran or C with a
different syntax, you might not notice the difference. Batch compiled
benchmarks may therefore be misleading.
But even so, benchmarks posted at
http://www.cliki.net/Performance%20Benchmarks suggest that ABCL can be a few
hundred to a few thousand times slower on various ordinary computations,
though comparable on some others..
Is there a real compelling reason to use this for Maxima? If not I would
just put it in the category of curiosities.
RJF
> -----Original Message-----
> From: Raymond Toy [mailto:raymond.toy at ericsson.com]
> Sent: Friday, June 22, 2007 8:22 AM
> To: fateman at cs.berkeley.edu
> Cc: 'Reinhard Oldenburg'; maxima at math.utexas.edu
> Subject: Re: [Maxima] Maxima running on Java VM?
>
> Richard Fateman wrote:
> >
> > My understanding of the Java VM is that it is deficient in terms of
> > providing facilities for Lisp, and that a full
> implementation of Lisp would
> > be very inefficient in Java. (to quote their own project
> page, "intolerably
> > slow").
>
> I read that as saying abcl's CLOS was intolerably slow.
>
> For some benchmarks, see http://www.cliki.net/Performance%20Benchmarks
>
> From the benchmarks from that site, while performance varies
> a lot, I
> would venture to say that it is comparable to clisp in speed.
> I have no
> problems with running maxima with clisp.
>
> Ray
>