[IPOL discuss] matlab to C/C++ translation tools

Lionel Moisan Lionel.Moisan at parisdescartes.fr
Sat Mar 23 22:26:46 CET 2013


Hi,

An interesting alternative could be to use Scilab, which is free and open-source,
and shares a syntax very similar to Matlab (they both originated from the same
INRIA project long ago). By the way, our team at MAP5 is currently involved in a project
with Scilab to develop a research-oriented image toolbox. We plan to distribute a first
version as an atoms package this summer, so it could be an interesting occasion to
invite more people to develop Scilab image processing codes. Of course, Scilab (like Matlab)
is not as efficent as C for programming, but for many applications the loss
of speed can be limited to a reasonable factor, so I guess it could be an interesting
feature to allow Scilab in IPOL codes. For intensive and/or matrix-unfriendly
programs we use C subroutines that can be mixed with Scilab (like Matlab with mex files).
The readability cost can be limited if the algorithmic and the I/O parts are well
separated.
There is another open-source environment called octave, that shares (exactly ?) the
same syntax as Matlab, but I don't know much about it.

Best,
Lionel


Le 23/03/2013 07:38, Jean-Michel Morel a écrit :
>
> Dear all,
> This is potentially very interesting, but I am concerned by two issues :
> -will such a C++ code be readable? I gather that still it will be better to review the original
> Matlab code
> -what about the functions in Matlab whose code is not public? How could they possibly be translated?
> Best,
> Jean-Michel
>
>
> On Sat, Mar 23, 2013 at 6:24 AM, Nicolas Limare <nicolas.limare at cmla.ens-cachan.fr
> <mailto:nicolas.limare at cmla.ens-cachan.fr>> wrote:
>
>     Hi,
>
>     The matter of publishing MATLAB implementations in IPOL is often
>     discussed. I recently heard about two projects of a MATLAB-to-C++
>     translator, and this could solve some of the objections against MATLAB
>     (black box, not mixable with other languages, requires a license).
>
>     These MATLAB-to-C++ translators are commercial products developped by
>     the Silkan[1] and companies. They are not more open or free than
>     MATLAB, but if I understand correctly they produce a C++
>     implementation equivalent to a MATLAB code, and this C++ code can be
>     open and free (to be verified, I do not know the details). This could
>     be an option for IPOL, for example receiving the MATLAB code but
>     reviewing and publishing the C++ code converted from MATLAB (with or
>     without co-author, etc... to be discussed later).
>
>     I read that both companies are interested in academic partnerships and
>     could possibly offer the use of their tools for academic purposes as
>     it would help them improve their products.
>
>     Do you think this option should be explored? Is someone interested in
>     contacting these companies to know more about their MATLAB-to-C++
>     translators?
>
>     [1] http://www.silkan.com/
>     [2] http://www.metascale.com/
>
>     --
>     Nicolas LIMARE - CMLA - ENS Cachan http://limare.perso.math.cnrs.fr/
>     IPOL - image processing on line http://www.ipol.im/
>
>     _______________________________________________
>     discuss mailing list
>     discuss at list.ipol.im <mailto:discuss at list.ipol.im>
>     https://tools.ipol.im/mailman/listinfo/discuss
>
>
>
>
> _______________________________________________
> discuss mailing list
> discuss at list.ipol.im
> https://tools.ipol.im/mailman/listinfo/discuss


-- 
Lionel Moisan
Université Paris Descartes
Laboratoire de Mathématiques Appliquées (MAP5)
45 rue des Saints-Pères F-75006 Paris
tel: +33 1 83 94 58 48
http://www.mi.parisdescartes.fr/~moisan


More information about the discuss mailing list