[MOAB-dev] Help for migration of cells between MPI subdomains with moab

[Olivier Jamond]

Hello Iulian

> Congrats for France's win yesterday!
Thanks :) I though that in the US our popular soccer was of little 
interest, but I see that it actually is!
> I will work on that example, I promise I will have something after the weekend
Thanks for that, but do not spoil your weekend, I can wait for some more 
days!
> In the meantime, I was thinking at your problem, and what types of elements do you migrate?
> Does it have to be general, any type, or for your initial application you have a limited set ?
> If elements are about the same type, it is possible to use crystal router efficiently; or at least if your element types do not have very different number of nodes.

Our code is intended to deal with industrial problems involving various 
kinds of numerical methods/discretizations/models/... in the same 
calculation. For example we do fluid structure interaction between shell 
finite-elements and finite-volumes, or coupling between volume elements 
and 1D elements when modeling steel-reinforced concrete.

So yes, we must be able to migrate several kind of elements within a 
given computation. But maybe we can migrate them type by type?

> Is local memory per task of concern or not really ? pack/unpack methods are pretty efficient, they do use memory well; Crystal router is more of a memory hog, but it may be fine if you do not have to send many cells from one particular  task to another.
As long as the memory consumption scales well as the number of MPI 
processes increases, the memory per task is not of major concern (to 
some reasonable extents of course!)
> If you use Zoltan, you need to have a global id sort of identifier for all cells and vertices, and that should help ; Also, I know that Zoltan has a pretty decent migrate capability; Did you try using it or not ?
No, we didn't try Zoltan migration capability yet. But I think that 
Zoltan can handle the data communication, but the application still have 
to pack/unpack data, no?
> what Zoltan do you use? Zoltan or Zoltan2 ?

We use Zoltan.


Many thanks,
Olivier