itaps-parallel ITAPS for 2D meshes

[Mark Shephard]

They are faces bounded by edges and vertices.

Note that we solve non-manifold problems where the "elements" (the 
entities that have to be partitioned) can be combinations of regions, 
faces, etc. In addition, we have reasons (defined by the analysis 
procedures) where small groups of elements need to be kept together (be 
a single object to the partitioning process).

This is why from the beginning I stated I liked to not have a one-to-one 
correspondence between a single mesh entity type and objects to be 
partitioned (something it is my understanding the abstractions of Zoltan 
supports).

Mark


Devine, Karen D wrote:
> In 2D meshes (e.g., triangular or quadrilateral elements), what is the
> entity type for an element?  Is it a region?  Or must regions be 3D?
> 
> Here's the reason I ask.  The function iMeshP_createGhostEnts takes as
> integer arguments a ghost dimension and bridge dimension.  At first, I
> thought of these arguments as entity types, e.g., regions and faces.  But
> when I started to draw my 2D picture for the SC08 slides, I realized that
> what I thought of as regions were actually 2D....so what should the ghost
> dimension be?
> 
> Would the ghost function be clearer taking entity types rather than numeric
> dimensions?
> 
> Thanks.
> Karen
> 
>