octave-maintainers
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [fem-fenics] MPI parallelisation


From: Eugenio Gianniti
Subject: Re: [fem-fenics] MPI parallelisation
Date: Sun, 27 Jul 2014 23:13:19 +0000

On 16 Jul 2014, at 14:46, Eugenio Gianniti <address@hidden> wrote:

> Dear all,
> 
> I am currently stuck in the implementation of the MPI parallelisation in 
> fem-fenics. I wrote some code for the parallel assembly of matrices, but if I 
> try to run the examples it crashes with DOLFIN internal errors even before 
> this new code is executed. I quickly recall what I know about the issue:
> 
>  1. MPI in itself does not seem to be the cause, since running with mpirun -n 
> 1 everything works (basically I understand that DOLFIN internally checks the 
> number of processes, not the MPI initialisation, to decide if it needs to 
> “act parallel”);
> 
>  2. just to be sure, I ran the equivalent C++ implementations of these 
> examples, and everything works fine in parallel, too;
> 
>  3. trying different examples (in Octave) leads to different errors, spread 
> all around DOLFIN’s code base, sometimes even before the call to assemble;
> 
>  4. whilst different examples cause different errors, the same example leads 
> always to the same error.
> 
> From the information I could extract, my guess is that the issue may arise 
> from the dynamic loading of the Octave functions that reference DOLFIN code. 
> However, I do not know how to further troubleshoot the problem. So I wonder 
> if someone has already dealt with a similar issue and has any advice to share 
> or if someone more familiar with Octave’s internals can provide me with any 
> helpful insight.
> 
> Thanks,
> Eugenio

I further worked on this issue, and now have an implementation that allows for 
solving a system on a mesh loaded from file and without DirichletBCs. In this 
simplified setting, seemingly the parallel solution is the serial one divided 
by the number of processes.

Moreover, I tried to fix also the problem of meshes defined with the msh 
package. DOLFIN provides a handy method to distribute the mesh, but marked 
subdomains are not supported in version 1.3.0: indeed, Mesh.oct does mark a 
subdomain, but I cannot get any clue on why it does, so I need Marco to explain 
some details of that, please.

Eugenio


reply via email to

[Prev in Thread] Current Thread [Next in Thread]