Hello Kostas,

Thank you! GetFEM's higher-dimensional capabilities were actually one of the 
major reasons I chose the software package. Collisionless plasma dynamics 
requires six dimensions for full simulation since there is no requirement for 
thermodynamic equilibrium in each position cell, so the velocity distribution 
also has to be modelled.

I made an account at the site you linked - should I just make a new topic with 
the same title or should I copy the email chain in as well?

That makes sense - I am curious how it should be compiled, though, since I have 
seen separate source files for the compilations in Windows and Linux and am 
wondering if these produce the same .pyd source files for the Python interface. 
My Python interface is in Windows, but the instructions on the website seem to 
be much easier from Linux.

Sincerely,
Eric A. Comstock
________________________________
From: Konstantinos Poulios <[email protected]>
Sent: Friday, March 28, 2025 11:20 AM
To: Comstock, Eric A <[email protected]>
Cc: [email protected] <[email protected]>
Subject: Re: Looking for help - Unable to edit MUMPS parameters in Python 
interface

You don't often get email from [email protected]. Learn why this is 
important<https://aka.ms/LearnAboutSenderIdentification>
Hello Eric

Great to hear from a user actually using GetFEM in more than 3 dimensions!!!

I would suggest to move the discussion to  https://getfem.discourse.group

The short answer to your question is that our mumps interface is very 
rudimentary, we have not implemented editing of MUMPS parameters through the 
scripting interface. It has been on my todo list for quite some time. For the 
time being, you need to edit the C++ code (file: gmm_MUMPS_interface.h) and 
recompile GetFEM.

All the best
Kostas



On Fri, Mar 28, 2025 at 1:00 PM Comstock, Eric A 
<[email protected]<mailto:[email protected]>> wrote:
Hello All,

I am encountering an issue using the getFEM 5.4 interface with Python 3.7 on 
Anaconda.

I am using getFEM to simulate collisionless plasma dynamics, and recently 
bumped my simulations from 4D (2 dimensions of space, 2 of velocity), to 6D (3 
dimensions of space, 3 of velocity), to increase fidelity for a research paper 
I am writing. This has increased coupling on my stiffness matrices, and is 
causing MUMPS to throw the "Solve with MUMPS failed: error -9, increase 
ICNTL(14)" error. I would like to do this, and increase the allocated memory, 
but I have not been able to figure out how to from the Python interface.

My main issue is that there does not appear to be a way to change ICNTL(14) of 
the MUMPS solver, since getfem.py just calls the _getfem.pyd file, which is a 
binary. I am wondering if I have either missed a way of changing it in the 
program itself, or what part of the .pyd file needs to be edited, if that is 
the only way of doing it.

I have used getFEM successfully before in 4D spaces, so I am certain that the 
additional coupling from the fifth and sixth dimensions is what is causing the 
problem.

Sincerely,
Eric A. Comstock

Reply via email to