[Dune] Features of Ghost-Elements

Martin Nolte nolte at mathematik.uni-freiburg.de
Sun Jun 12 23:16:12 CEST 2011


Hi Jö,

reading your statement again, Andreas answer is the correct one: You get two
different indices (which is the expected behavior for copies).

Best,

Martin

PS: Sorry for answering in private. Somehow it's really hard for me to keep the
mails on the list (simple reply does not work). The "private" answer was:

"Yes, I was referring to such a situation. The answer to all your other
questions should be positive, too."


On 06/12/2011 02:15 PM, Jö Fahlke wrote:
> Am Sat, 11. Jun 2011, 23:00:45 +0100 schrieb Andreas Dedner:
>> Do you mean a situation like this:
>>>   +--------+--------+
>>>   |        |        |
>>>   | I_2    | ghost  |
>>>   |        |        |
>>>   +--------+--------+
>>>   |        |        |
>>>   | I_1    |  I_0   |
>>>   |        |        |
>>>   +--------+--------+
>>>
>>> In this case
>>>
>>>  * Would I visit both copies of the ghost when iterating over the elements of
>>>    the gridview?
>>>
>> ibegin() on both I_0 and I_1 would give you a ghost both with the
>                            ^^^
> 
> I suppose you mean I_2 instead of I_1 there.
> 
>> same geometry but with different index. So you have a G_0 and a
>> G_2 which are geometrically identical but topologically different.
>> ibegin on G_2 for example would give you only I_2.
>> The communication method works, i.e., the data from the corresponding
>> interior element would be communicated to both G_0 and G_2.
> 
>>>  * Would gv.indexSet().index(ghost) yield the same index for both copies?
> 
> Your answer actually contradicts the answer I got from Martin privately
> regarding the indices.  However, since you agree on the other two answers
> ("yes" both times), the answer to the indices question doesn't matter in my
> case, I should be OK either way.
> 
>>>  * When doing an Interior_All communication, would both copies of the ghost
>>>    receive data from the corresponding interior element of the other process?
> 
> Thanks,
> Jö.
> 
> 
> 
> 
> _______________________________________________
> Dune mailing list
> Dune at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune

-- 
Martin Nolte <nolte at mathematik.uni-freiburg.de>

Universität Freiburg                                   phone: +49-761-203-5642
Abteilung für angewandte Mathematik                    fax:   +49-761-203-5632
Hermann-Herder-Straße 10
79104 Freiburg, Germany



More information about the Dune mailing list