dolfin team mailing list archive

[Kent-Andre Mardal <kent-and@xxxxxxxxx>]

tir, 23,.01.2007 kl. 11.05 +0100, skrev Anders Logg:
> I'm about to write something down for FFC/DOLFIN. My idea is that the
> numbering scheme is the UFC numbering scheme + something else.
> 
> The UFC numbering is only concerned with the numbering of mesh
> entitites with respect to a cell, based on the vertices of the cell.
> 
> My suggestion is that the numbering is based on a lexicographical
> ordering of entities and that the vertices of the cell are ordered
> lexicographically based on their global index.
> 
> Here's an example.
> 
>     Triangle with vertices v0 = 10, v1 = 15, v2 = 95
> 
>     Cell-edge:  e0 non-incident to v0
>                 e1 non-incident to v1
>                 e2 non-incident to v2
> 
>     (This is based on a lexicographical ordering of non-incident
>      vertices as in the UFC manual.)
> 
>     Edge-vertex: e0 = (v1, v2)
>                  e1 = (v0, v2) 
>                  e2 = (v0, v1)
> 
>     (This is based on a lexicographical ordering of the entities
>      within the entity (edge) in question.)
> 

I think we will go for this ordering. 

> So there are two steps: First number all entities based on a
> lexicographical ordering as in the UFC manual, then number everything
> else based on the global index.
> 
> UFC is only concerned with the first step.
> 
> We also need to define something we can call "alignment", which is how
> two cells meet at a common facet. For two triangles, the alignment is
> 0 or 1. For two tetrahedra, the alignment is 0, 1, 2, 3, 4 or 5.
> This can also be based on a lexicographical ordering of matching
> vertices.
> 

We might need this also. But up until now we have tried an alternative 
approach in which we defined the global facet as a ordered sequence 
of the vertices on facet. 

> The text in the UFC manual is written so that it can be copied to
> other manuals (SyFi, DOLFIN, FFC) and extended.
> 
> The DOLFIN mesh library should be updated so that each DOLFIN mesh
> follows this numbering convention.
> 

Good. 

Kent