Re: Octave on Reddit

[GFotios]

On Sep 29, 2010, at 3:11 PM, Michael Creel wrote:

> 2010/9/29 Jaroslav Hajek <address@hidden>:
> > 2010/9/29 Jordi Gutiérrez Hermoso <address@hidden>:
> > > It seems a popular news site has picked up Octave's homepage:
> > >
> > >     
> > > http://www.reddit.com/r/math/comments/dk61b/free_alternative_to_matlab_check_octave_out/
> > >
> > > A few months ago, something similar happened:
> > >
> > >     
> > > http://www.reddit.com/r/math/comments/c3pqb/ever_wished_you_had_a_matlab_license_octave_is/
> > >
> > > It's really sad to read the comments, though. Many suggest to use
> > > other things instead of Octave, and many others say that Octave is
> > > irreparably inferior as compared to Matlab.
> >
> > I've only read the first one, but it's not that bad. NumPy/SciPy are
> > becoming popular, but I suspect those people saying that it has an
> > "almost identical syntax" to Matlab have never really used Matlab.
> > Anyone who used both knows that the syntax is *way* different  
> > compared
> > to Octave; and not just the syntax, but the semantics as well (COW  
> > vs.
> > implicit pointers).
> >
> > Yes, plotting is slow and not as nice as Matlab's, but it's about the
> > best that can be squeezed out of Gnuplot. Regarding computing
> > functions, someone only mentioned contourc, but I wonder exactly how
> > much slower it is. Etc.
> >
> > As a summary, there's nothing new, we know all that. It seems to be a
> > rule on the web that negative comments are always in broad supply,
> > regarding almost anything even if many people like it. In Czech, we
> > have a proverb saying that "An empty pot thuds the loudest." ;-)
> >
> > regards
> >
> > --
> > RNDr. Jaroslav Hajek, PhD
> > computing expert & GNU Octave developer
> > Aeronautical Research and Test Institute (VZLU)
> > Prague, Czech Republic
> > url: www.highegg.matfyz.cz
>
> At my university we have a site licence for Matlab, and last year I
> got curious and decided to try it out to see if I was missing
> anything. Suffice it to say that my experiment lasted for a week or
> two, and after that I happily went back to Octave. Octave with MPI and
> a little C++ for bottleneck computations is a winning combination, in
> my opinion. I don't think that popularity is necessarily a good
> measure of quality - there are a lot of Windows users out there, after
> all.
> Michael

I agree with the combo Octave + C++ + MPI. There is really nothing new  
to Reddit, people that like to argue about something never tried (i  
guess), Jaroslav is correct. I implemented and measured the time for  
the assembly of FEM matrices in Python, Python+f2py and Matlab (once  
upon a time when i was choosing my dev environment), and the results  
were disappointing for Python!!! Now that i use Octave a lot i ll show  
sth similar for a shape optimization problem (later in 2010 for those  
interested). Regarding Octave, i would like the interpreter to be  
faster (say like Scilab's; the same simple loop takes 3\times or so  
longer in Octave unfortunately) (and when i m able to do sth about it,  
i ll).

/Fotis

results.pdf
Description: Adobe PDF document