[Israel.pm] PDL

Shlomi Fish shlomif at iglu.org.il
Sun Jul 22 01:26:02 PDT 2007


Hi!

On Monday 09 July 2007, Yosef Meller wrote:
> ציטוט Amir E. Aharoni:
> > Hi,
> >
> > I'm trying to move away from Windows as much as i can. I don't need it
> > for anything, but my wife is doing her second degree in Physics (she
> > is a lot more smarter than i am) and she is a heavy user of scientific
> > software, in particular MATLAB.
> >
> > I don't know almost anything about this kind of software, but some
> > people claim that PDL is a worthy and free competitor to MATLAB.
>
> If MATLAB is her preferred environment, I would suggest GNU-Octave as a
> free replacement. It lags behind in graphics, but other than that it's
> very useful, and Octave 3.0 is nearing its release and promises to close
> most of the gap.
>

I should note that I overheard that Octave is often slower than Matlab (which 
shouldn't matter much) and also sometimes generates different results. But I 
didn't experience it first hand.

> > Does anyone on this list have positive experience with it? If you can
> > compare it to MATLAB, it would be nice too.
>
> I have a negative experience with PDL - after I wrote some program with
> it, when I upgraded PDL I suddenly found that I have no eigenvalues
> because of a bug, so my program wouldn't work at all.

Have you reported this bug, preferably with a regression test included? If 
not, then I'd appreciate if you can supply more information about it. There 
can always be bugs, but the developers depend on the users to report them.

I have some experience with PDL as well. I've used it for 
http://www.shlomifish.org/lecture/Perl/Lightning/Opt-Multi-Task-in-PDL/ and 
also created the following custom extension for it: (in C).

http://search.cpan.org/dist/PDL-Lib-Linear-Solve/

It worked from what I tried, but note that I did not try to push it to the 
limit.

>
> The only thing I can really say about it is that if you must use Perl
> for physics work, PDL is probably better than plain Perl.
>

It would be faster and less memory hungry, at least. I should note that the 
default pdl() constructor is implemented inefficiently. It is faster to use 
the PDL::IO::FastRaw module.

> > I know i can search for such comparisons on the web, but it will be
> > particularly useful to know how does PDL fit into the Israeli academic
> > environment (my wife studies in the Hebrew University) - for example
> > would it possible to prepare a thesis in Physics in HUJI or Technion
> > (solid state, semiconductors, etc.) using PDL instead of MATLAB.
> >
> > Many thanks in advance to anyone who can help.
>
> Also there's numpy for Pythoneers, scilab (not exactly MATLAB
> compatible, but close enough), 

There's a program that ships with Matlab provided to convert Matlab code to 
Scilab (but not the other way around).

> and possibly others. 
>
> And on the non-free front, MATLAB is available for Linux too.

That's right.

Regards,

	Shlomi Fish

---------------------------------------------------------------------
Shlomi Fish      shlomif at iglu.org.il
Homepage:        http://www.shlomifish.org/

If it's not in my E-mail it doesn't happen. And if my E-mail is saying
one thing, and everything else says something else - E-mail will conquer.
    -- An Israeli Linuxer



More information about the Perl mailing list