[MOAB-dev] MOAB & periodicity

[Tim Tautges]

IIRC, the way we did this for structured mesh was to set the global ids properly on vertices on the periodic boundary, 
then let the parallel find shared vertices work as it already does/did. So effectively, in parallel the mesh thinks it 
is connected across the periodic boundary. There are probably two caveats to this working properly for unstructured mesh:

1) That wouldn't work in serial, so you would probably need to manage that yourself explicitly (easily done if you know 
the matching across the periodic boundary, which you'll need anyway to set the global ids there properly).

2) I don't recall whether all the parallel shared entity stuff works properly on polyhedra, given the connectivity of a 
polyhedron represents faces rather than vertices.

- tim

On 04/12/2016 11:57 AM, Vijay S. Mahadevan wrote:
> Xavier,
>
> The unstructured mesh representation in MOAB doesn't have native
> support for periodicity and if you really want to implement that, you
> will have to manage this at your solver level through either Tags or
> your custom data cache.
>
> There was some recent discussion about a similar topic. Please read
> through it to see if some of that discussion is useful for you.
>
> http://lists.mcs.anl.gov/pipermail/moab-dev/2016/007189.html
>
> In parallel, using Tags to maintain the representation should provide
> the necessary framework to get this working seamlessly.
>
> Vijay
>
> PS: I also noticed that I have a reply to Alexander in my draft, that
> never got sent out. Oops.
>
> On Tue, Apr 12, 2016 at 11:19 AM, Xavier Garnaud <xgarnaud at gmail.com> wrote:
>> Dear all,
>>
>> I am thinking about using MOAB as the basis for a cell-centered
>> (unstructured)  finite volume solver. An important feature for me is to be
>> able to handle periodicity (translational & rotational). As far as I
>> understand, this is not directly possible with MOAB but I am wondering how
>> difficult it would be.
>>
>> What I would like to do is have "periodic ghost" cells that I would handle
>> as the standard ghost cells (except I would possibly apply a rotation to the
>> vectors after the data is exchanged for rotational periodicity)
>>
>> How difficult do you think it would be to implement it? Do you have any
>> example that implement similar operations?
>>
>> Best regards
>>
>> Xavier
>>

-- 
Timothy J. Tautges
Manager, Directed Meshing, CD-adapco
Phone: (631) 629-3139 (internal: x29130)
Mobile/Gvoice: (608) 354-1459
timothy.tautges at cd-adapco.com