Make MultiAppInterpolationTransfer/MultiAppUserObjectTransfer work with distributed meshes

[vincentlaboure]

Reason

Many large multiphysics problems make use of LayeredAverage/MultiAppUserObjectTransfer and/or MultiAppInterpolationTransfer to transfer data between apps. If the mesh becomes too large, a replicated mesh is too memory intensive but, unfortunately, distributed meshes do not currently work with MultiAppUserObjectTransfer/MultiAppInterpolationTransfer.

• [x] MultiAppUserObjectTransfer (use general field)
• [ ] MultiAppGeometricnterpolationTransfer

Design

Make MultiAppUserObjectTransfer/MultiAppInterpolationTransfer work with distributed meshes.

Impact

Bigger problems can be considered.

[lindsayad]

@fdkong @roystgnr are we aware of this issue? If so, are there any important technical hurdles or just developer time?

[fdkong]

@friedmud intended to make MultiAppInterpolationTransfer work with only replicated meshes, IIRC.

[fdkong]

It is possible to make MultiAppInterpolationTransfer work with distributed meshes, but you might use other distributed transfers. I am not 100% convinced we should make MultiAppInterpolationTransfer work with distributed meshes

[lindsayad]

What should be used instead?

[fdkong]

MeshFunction and NearestNode should do that

[vincentlaboure]

Two comments:

1. MultiAppUserObjectTransfer is definitely a higher priority
2. Can you explain why you think MultiAppInterpolationTransfer shouldn't be supported ?

[friedmud]

As mentioned in #20616 I am waiting to do this until the general transfer framework ( #17417 ) is merged to make this happen.

[rui-hu]

We have similar needs to have MultiAppUserObjectTransfer working with distributed mesh, and find this issue. Is there any progress on this since the last update?

@aprilnovak @lingzou-anl

[lindsayad]

Have you tried MultiAppGeneralFieldUserObjectTransfer?

[rui-hu]

not yet. we will give it a try.