Question about NSP
Hazem
Greetings everyone,
I have just become aware of the existence of NSP, which appears to be
a ScilabGTK deriviative. See http://cermics.enpc.fr/~jpc/nsp-tiddly/mine.html
My question is, what are the main differences and what is the
motivation of the NSP fork? Does NSP use the same programming language
as Scilab?
Regards,
Hazem
Leland C Scott
jpc
> I have just become aware of the existence of NSP, which appears to be
> My question is, what are the main differences and what is the
> motivation of the NSP fork? Does NSP use the same programming language
> as Scilab?
Hi Hazem,
First of all Nsp is not exactly a fork since forking is taking the
code
of a project and start a new project from that code, changing it.
Nsp was started from scratch and its aim was to provide a new
interpreter
and data-types written in C for a Scilab-like langage.
Thus, Nsp was at the beginning for me a parallel activity to Scilab (a
project I
was involved in). I have then inserted in Nsp Scilab code for which I
was
the main developer in Scilab (gtk code, old Scilab graphics etc....).
Bruno Pinçon who was also involved in Scilab also switched
to Nsp for which he has developed new code and also have ported
to Nsp some Scilab code he was copyright owner.
My main motivation when starting Nsp was to get rid of Scilab
interpreter
and data manipulation which is hacked Fortran code which is almost
not maintainable.
An other motivation was to be able to use gtk widgets in Nsp as a
complete library
as it is for example done in pygtk.
If you run the Gtk demos in Nsp you will obtain the full gtk-demos but
fully written at
Nsp level (i.e similar to Scilab script).
A last motivation was to provide an easier and much more complete
interface language.
Merging Nsp interpreter to Scilab was proposed during year 2006 and a
working group
have explored the possibility. The conclusion was that Scilab will
follow an other path
highlighted with sexy keywords: java, C++, visitors.
It is not easy to summarize the main differences.
Nsp is very similar to Scilab concerning the grammar, but slight
differences exist.
For the semantics of operators, it is also similar except for some
functions were the
Matlab semantics is preferred in Nsp.
As an example a large Scilab toolbox called scicos was ported to nsp
(not the last
version but an old one).
You won't find in Nsp all the functions available
in Scilab, but you will also find in Nsp functions unavailable in
Scilab. A set of
Nsp toolboxes exists but are not already available on the web (openmi,
glpk,lpsolve,
Sedumi, SDPT3,...) it should be done in a near future.
Regards,
jpc
Sylvestre Ledru
Hello Jean-Philippe,
> A set of
> Nsp toolboxes exists but are not already available on the web (openmi,
> glpk,lpsolve,
> Sedumi, SDPT3,...) it should be done in a near future.
I would be interested to see your work on NSP/OpenMPI.
Do you have any roadmap or schedule when they will be available ?
Sylvestre
Sylvestre Ledru
> Le Wed, 01 Oct 2008 04:13:37 -0700, jpc a écrit:
>
> Hello Jean-Philippe,
>
>> A set of
>> Nsp toolboxes exists but are not already available on the web (openmi,
>> glpk,lpsolve,
>> Sedumi, SDPT3,...) it should be done in a near future.
>
> I would be interested to see your work on NSP/OpenMPI.
Goups, I read MPI, not MI. Professional deformation ;)
Sorry for the mistake.
S
jpc
Hi Sylvestre, In fact you have corrected a mistake that was mine.
It was of course OpenMPI.
I originaly started with mpitb which is a mpi Matlab/Octave based on
mexlib for Matlab. Then for efficiency and better control data packing
and transmission I have
rewriten the whole stuff as native nsp interfaces/classes.
mpitb, then contains a huge file that can be used as a tutorial or as
a test
file for nsp. Since all the standard objects in Nsp (i.e all the ones
that can be saved/loaded)
can be serialized, it is possible to transmit them with mpi using
their serialized version.
I still need to finish testing the tutorial script.... It could be on
the web this month.
jpc
Hazem
Thank you, jpc. best wishes with your work.
Hazem