[Maxima] octave, matlab, maxima
pip at iszf.irk.ru
Mon May 26 01:03:07 CDT 2008
It's a good idea to connect maxima with one of a free Matlab's clones. I'm
not sure that octave is the best one.
There are several choices : octave, scilab, and python (with numpy&scipy
extensions). The last one is potentially much more powerful of two because
python covers much wider area for numerical and enginering applications than
octave and scilab. Also, it is much easer to run in parallel. This is right,
that the ordinary numerics and array operations may be slower in python among
the others. However when you run things using the python extensions, say,
solving ODE and etc, the speed is comparable because all of them takes the
help some of fortran/c math library. The whole task can be solved with
cl-python. However not all of python features are implemented in cl-python
yet. Also it uses asdf system which has an additional advantage.
03:21:50 Richard Fateman wrote:
> I've been playing with octave a bit (octave is essentially a free clone of
> matlab), after not using Matlab for 15 years or more.
> Why might it interesting to hook up to maxima?
> 1. Many people know and love Matlab (maybe Octave too). Its conventions,
> while crummy from a technical programming language perspective -- bad scope
> rules, no real packages, and a terrible convention that redefines an array
> A as a function if there is a file named A.m in accessible directories --
> making compilation very iffy...
> yet it has fostered many applications: many people contributing useful
> programs. Octave comes with (free) code tuned on a per-cpu basis. (e.g.
> when I installed Octave, it knew I had a Pentium D and loaded appropriate
> assembler). For some people it might be nice to know that someone who
> really cares about Lapack, parallel Lapack, etc, is on the job, rather than
> hoping the Fortran-to-Lisp stuff is working, and up to the latest revision,
> etc. There are also packages for gnuplot as well as a java plotting
> 2. Its programming tool are sometimes neat and/or bizarre. To reverse a
> vector V of n items, instead of doing
> reverse(V), one does something like V([n:-1:1]). Things like this can be
> added to Maxima, and maybe that should be revisited.
> 3. Maxima could provide the same utility for Octave as Maple does for
> Matlab; this is, I think, of only slight interest to people who know how
> to use Maxima. It could pique interest in CAS among of the (much larger, I
> think) community of Octave users. Octave/Matlab is, (I think) rather
> widely used in courses, and having a free symbolic toolkit for Octave would
> introduce those students to a few features of a CAS.
> 4. What else could be done? Well it would be really neat if one could have
> Octave++ where all the numerics were done in arbitrary precision. For
> now, even the Lapack subset is probably not up for that, much less stuff
> like numerical quadrature. Some of this could probably be taken from MPFR
> if someone cared. (Any MPFR experts who read this mailing list?)
> Now if we could get some Octave enthusiast to pick this up, instead of
> diverting a Maxima expert, that would be even better.
> Maxima mailing list
> Maxima at math.utexas.edu
More information about the Maxima