[Dune] difference between OwnerOverlapCopyAttributeSet::copy and OwnerOverlapCopyAttributeSet::overlap
Rebecca Neumann
rebecca.neumann at iwr.uni-heidelberg.de
Thu Sep 6 18:54:37 CEST 2012
Hey Andreas,
the idea of this distribution is, that only one process is owner, all
the others have the copy attribute.
In the pdelab istlhelper the proecss with the lowest rank will be the owner.
Best Regards,
Rebecca
--
Rebecca Neumann
Universität Heidelberg
Interdisziplinäres Zentrum für Wissenschaftliches Rechnen
Im Neuenheimer Feld 368, 69120 Heidelberg
Tel: 06221/548890 Fax: 06221/548884
e-mail: <rebecca.neumann at iwr.uni-heidelberg.de>
On 06.09.2012 18:45, Andreas Lauser wrote:
> Hi Rebecca,
>
> Thank you for the quick and helpful answer. I have to annoy you with an
> additional question, though: For entities on the border, is there only one
> owner or are all processes which are adjacent to a border entity its owners? I
> tried to figure that out using PDELab's ParallelIstlHelper, but that code
> defeated me...
>
> cheers
> Andreas
>
> Am Donnerstag, 6. September 2012, 18:36:44 schrieb Rebecca Neumann:
>> Hi Andreas,
>> that is right the attribute overlap is not used. For every degree of
>> freedom that is shared with other processes, the owner attribute is used
>> if the process is the owner, and else the copy attribute.
>> I think the overlap attribute is just there for historical reasons?
>> When I implemented nonoverlapping AMG I used the overlap attribute for
>> the ghost degrees of freedom that do not lie on the processor border.
>> Best Regards,
>> Rebecca
>>
>>> Hi Dune,
>>>
>>> Today I was wrestling with my and dune-istl's parallelization code once
>>> again. During that, I started to wonder what the difference between
>>> OwnerOverlapCopyAttributeSet::copy and
>>> OwnerOverlapCopyAttributeSet::overlap is. I could not find anything in
>>> the doxygen documentation, and the corrosponding lines in
>>> dune/istl/owneroverlapcopy.hh read:
>>>
>>> struct OwnerOverlapCopyAttributeSet
>>>
>>> {
>>>
>>> enum AttributeSet {
>>>
>>> owner=1, overlap=2, copy=3 };
>>>
>>> };
>>>
>>> what made me wonder even more, is that the code in
>>> dune/istl/paamg/test/anisotropic.hh (which creates the matrix for the
>>> parallel AMG test) does not use OwnerOverlapCopyAttributeSet::overlap
>>> even though the matrix which it creates clearly corresponds to one which
>>> stems from an overlapping grid. (if I understood that code correctly.)
>>>
>>> can anyone with deeper insight on ISTL tell the difference?
>>>
>>> cheers
>>>
>>> Andreas
>> _______________________________________________
>> Dune mailing list
>> Dune at dune-project.org
>> http://lists.dune-project.org/mailman/listinfo/dune
More information about the Dune
mailing list