Re: Parallel Octave

[Jean Michel Sellier]

Hi Julio,

thanks for replying.

Basically, let us say that for every linear solver and eigensolver implemented in Octave there is always the parallel correspective in PETSC or SLEPC. I was wondering if we could do something like that (for example):

If one has :

[V,E]=eig(M);

then Octave call the usual 1-cpu solver. But if one has:

[V,E]=eig(M,'parallel');

then Octave call the parallel implementation of the solver (in this particular case a SLEPC solver).

I think that including PETSC/SLEPC in Octave should not be a big deal at all, as one can easily see from the documentation.

That could be EXTREMELY useful.

Does that make sense to you?

Thanks,

JM

2012/4/4 Júlio Hoffimann <address@hidden>

Hi Jean,

What kind of solver you're trying to implement? How exactly Octave is related to it? As for most people here, available time is the biggest issue. We all (at least me) have interest in doing parallel numerical solvers of any kind. It's joyful!

I have reasonable experience with FEM, parallel programming with MPI and beginning to play with FVM this first trimester of my master degree. Our research group used PETSC in a project, but i never had the opportunity myself to work with it.

Just curious if i have the necessary skills...

Regards,

Júlio.

2012/4/3 Jean Michel Sellier <address@hidden>

Hello,

I was wondering if someone would be interested in implementing/using parallel solvers into Octave. There are several good libraries around that could be used right away. PETSC and SLEPC, for example, are very good libraries that do all the job (no need to know anything about parallelization) and they are both released under GPL. Furthermore, the libraries work perfectly on clusters, but also on single CPU machines.

I would like to implement them in Octave and use this new version of Octave on clusters that I can access, but I need help with the implementation.

Is anyone interested?

Thanks,

JM

--

Jean Michel Sellier, Ph.D., M.Sc.

Research Assistant Professor
Network for Computational Nanotechnology
207 S. Martin Jischke Drive
Purdue University
West Lafayette, IN, USA

Personal Web Pages:

http://nanohub.org/members/28836

https://engineering.purdue.edu/gekcogrp/research-group/JeanMichelSellier/

--

Jean Michel Sellier, Ph.D., M.Sc.

Research Assistant Professor
Network for Computational Nanotechnology
207 S. Martin Jischke Drive
Purdue University
West Lafayette, IN, USA

Personal Web Pages:

http://nanohub.org/members/28836

https://engineering.purdue.edu/gekcogrp/research-group/JeanMichelSellier/