Hi,
On Monday 31 March 2003 10:58, Madhusudan Singh wrote:
> Hi
> I recently came across maxima.
What a good meeting! ;-)
>My knowledge of lisp is
> zero.
Using maxima don't need the lisp knowledge at all. Though you
could gain very much from leaning of the maxima's sources.
>My platform is Linux.
>
> I have noticed that a relatively simple function
what do you call "simple"??
> (defined as a function of 3-4 variables, in terms of some
> other defined functions (using := notation) and atomic
> variables (using : notation) ), plotted against one of the
> variables, keeping the rest fixed, takes an unexpected amount
> of time (I have been exposed to Mathematica and Matlab
> earlier) to plot. So far, on a 700 MHz Celeron machine running
> imaxima+maxima(5.9.0)+gcl(2.5.2),
I'm not sure whether it is the influence of imaxima?? On my
mashine the plot appears less then in second (for 3d plot it is
typically 0.1 sec for gcl-maxima on linux athlonxp1500+).
I also prefer to start maxima within emacs but only in using
Maxima-mode.
So I would suggest to check whether the function has
singularities in the domain to plot or try to plot from console
to exclude the influence of imaxima interface.
> It is possible to compile maxima source files into *.o
> and *.LISP files. I tried to create an executable (hoping
> thereby that the executable would be executed faster) using ld
> -o test.exe test.o <maxima-lib>, but the resulting executable
> exits with a Segmentation fault.
Please, explain more clearly. The way to compile the maxima' file
to lisp is compile_file("file") and then :lisp(compile-file
"file") to *.o
> At another level, how does maxima performance compare with
> Mathematica / Matlab ?
Actually here we have 2 questions. Mathematica (and Maxima) is
designed for the purpose different than Matlab. Mathlab solves
the problem essentially in numerical way. Well, both
Mathematica and Maxima have abilities to treat the problem
numerically. However the speed of the numerical code obtained
with Mathematica and Maxima will be less than mathlab.
So if you try to compare speed of the symbolic computation in
Mathematica and Maxima you find maxima much faster ( for my codes
I found 10 and more times).
If you compare the speed of numerical code formulated entirely
(without translating to the lisp) in maxima with analogous and
in mathlab, you likely find the mathlab code faster.
rgds,
v