[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Getfem-users] Outsourcing calculation of stiffness matrix
From: |
Yves Renard |
Subject: |
Re: [Getfem-users] Outsourcing calculation of stiffness matrix |
Date: |
Tue, 15 Mar 2011 16:21:25 +0100 |
User-agent: |
KMail/1.13.5 (Linux/2.6.32-29-server; KDE/4.4.5; x86_64; ; ) |
Dear Roman,
This is a nice idea. Nonlinear term is the right tool, I think.
I also began to reflect on the possibility of adding terms with a syntax easier
than the one of Getfem generic assembly and also to the possibility of a
language for describing non-linear terms. But it remained in its infancy at
the moment.
Yves.
On mardi 15 mars 2011, Roman Putanowicz wrote:
> Dear All,
>
> I am working on the following problem:
>
> I would like to be able to "outsource" calculation of
> element matrices and vectors outside GetFEM++.
> In other words I would like to keep all GetFEM++ machinery
> but during the assembly force GetFEM++ to call external
> function that will provide the matrices to be assembled.
> The whole point is to be able to write the function in
> a scripting language for an interpreter embedded in the
> application (I aim at Python).
>
> I __do not__ want to pre compute element matrices and
> pass them as data to assembly routines.
>
> The solution I see is to write a special class
> that inherits getfem::nonlinear_elem_term and use
> the ability to perform arbitrary calculations during
> assembly via NonLin term.
>
> In fact I envisage writing a special brick that
> will initialize the external routines (either
> load shared library or load the scripts)
> and during the assembly will use mentioned
> special nonlinear term.
>
> The first attempts are promising but I would like
> to ask for your opinions on this idea. Have anybody
> tried this? Are there any subtle points to watch out?
> I will appreciate any comments.
>
> Regards
>
> Roman
--
Yves Renard (address@hidden) tel : (33) 04.72.43.87.08
Pole de Mathematiques, INSA-Lyon fax : (33) 04.72.43.85.29
20, rue Albert Einstein
69621 Villeurbanne Cedex, FRANCE
http://math.univ-lyon1.fr/~renard
---------