itaps-parallel definition for 'neighbors of a part'

[Onkar Sahni]

In my mind, "objects in the partition" are the "ones that were used to
define the division into parts", where partitioning algorithm will assign
them to one and only one part and the real-object-copy will exist only on
a single part (if ghosting is requested then some objects might be copied
to "neighboring parts" and ghost-object-copies may exist on multiple
parts).

Regarding covering of all the entities, at this point I am thinking that
at least lower-dimensional entities of an object are include irrespective
of the fact whether a lower-dimensional entity might have multiple copies
on different parts and a copy among multiple copies will be assigned owner
(I do not not necessarily see owner only with right-to-modify as it could
be just one role in certain applications, other role could be in counting
see comment on getNumOfTypePar).

For getNumOfTypePar, I think if entities exist on inter-part boundaries
and have copies on multiple parts then such entities should be counted
only once or a copy (owner, other role owner could play) should be
counted.

Thanks,
Onkar


> I'm personally okay with either variation of wording on the bits that
> differ; if pressed, I'd lean towards the second, at least at present.
>
> The thing that I still can't get past --- and I know I keep coming back
> to this --- is the implication that there might be objects that -aren't-
> in the partition.  This may well be just a semantic thing, where
> "objects in the partition" is being used as an abbreviation for "objects
> that were used to define the division into parts", and I'm
> misinterpreting that as "objects for which ownership (right to modify)
> is clearly defined".  The latter interpretation has to be, in my
> opinion, a covering of all the entities.  I suppose I'd be willing to
> relax that to "a covering of all the entities that might be modified",
> but then queries like getNumOfTypePar are either going to return
> something other than the total number of entities for which ownership is
> clearly defined, or have behavior that is counter-intuitive (my
> intuition, at least, expects that call to give a count for the global
> parallel mesh).
>
> Carl
>
> --
> ------------------------------------------------------------------------
> Dr. Carl Ollivier-Gooch, P.Eng.                   Voice: +1-604-822-1854
> Associate Professor                                 Fax: +1-604-822-2403
> Department of Mechanical Engineering             email: cfog at mech.ubc.ca
> University of British Columbia              http://www.mech.ubc.ca/~cfog
> Vancouver, BC  V6T 1Z4                  http://tetra.mech.ubc.ca/ANSLab/
> ------------------------------------------------------------------------
>
>