A (reasonably comprehensive) build method for Elmer

Discussion about building and installing Elmer
Post Reply
greenlinux
Posts: 9
Joined: 15 Aug 2020, 18:11
Antispam: Yes

A (reasonably comprehensive) build method for Elmer

Post by greenlinux »

I have written a build method for Elmer which includes a number of the optional packages.

https://bearinstruments.co.uk/building- ... om-source/

It seems well to point out that the fastest way to get going with Elmer is to use a pre-compiled binary or the official virtual machine.

James
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: A (reasonably comprehensive) build method for Elmer

Post by raback »

Hi James,

This is impressive work! I wonder if Elmer has often been built so inclusively.

As you say, there are many libraries that are often not needed, even I don't know where some are used. Most of the libraries have to do with parallel computation in some way. Adding Hypre and MUMPS libraries probably gives you sufficient tools for >90% of the parallel problems.

Some strategies are rather seldom used. One of these is FETI4I that you could not quite trace. This is related to FETI development in IT4Innovations National Supercomputing Center, Czech Republic. FETI may be attractive for massively parallel elasticity, for example. The library was called FETI4I some years ago, now it is known as ESPRESO. See,

https://github.com/It4innovations/espreso

As is stated there: "The library contains a simple C API that allows to utilize FETI based solvers in third party softwares. The API has been successfully tested with an open source multiphysical simulation software Elmer FEM developed by CSC - IT Center for Science. The interface is available via the feti4i.h header in the include directory.".

-Peter
greenlinux
Posts: 9
Joined: 15 Aug 2020, 18:11
Antispam: Yes

Re: A (reasonably comprehensive) build method for Elmer

Post by greenlinux »

Hi Peter,

Thanks for looking through it. I'm glad you approve. I'll try to add the pieces that are still missing over time.

It all started because I wanted to compile the small change you made to calc fields to split up the magnetic field energy from the electric field energy.

I'm planning to simulate an axial flux switched reluctance motor for a niche electric vehicle traction application when I have a moment. For this I will need a good deal of processing power as it has to be 3D and there is no universal machine theory to start from (stator currents are not sinusoidal). Hence my interest in setting up Elmer on Amazon web services' cloud system. If I manage to set Elmer up on AWS I'll write something to say how specifically it was done.

All the best,

James
Post Reply