nonlinear domain decomposition?

[alena]

Hi all, 

I am wondering, is it currently possible to use nonlinear domain decompositions solvers, such as NASM or ASPIN, from PETSc? 

If yes, is there any example available?

If no, are there any future plans to incorporate this into MOOSE/Libmesh framework? 

[Fande Kong]

There are a few efforts in libMesh, but not sure how far we already have. 

I personally love to have nonlinear preconditioning techniques such as NASM, ASPIN, nonlinear elimination in MOOSE. Unfortunately, there are no resources in our side to support me have such work.

But it is not hard at all, the thing we need to do is to have the right DM interfaces and then PETSc will pick up the things automatically. If you like to contribute something, I am excited to help you out here.

Thanks,

Fande,

--
You received this message because you are subscribed to the Google Groups "moose-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moose-users...@googlegroups.com.
Visit this group at https://groups.google.com/group/moose-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/9379b215-a707-4a55-bd90-91dba42987d6%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[alena]

Hi Fande, 

I would like to contribute, but I am afraid, that it requires way much more work, than I can actually manage...

Do you have any concrete idea on what exactly and where needs to be done? 

[Fande Kong]

The good start point should be a PETSc example, where you will learn how DMDA interacts with the nonlinear preconditioner, and then the same techniques can be implemented in moose as well.

Fande,

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/7c0a148c-5025-4600-93d2-74b2b6db476d%40googlegroups.com.