Mesh entity indices are ambigious

[Garth Wells]

UFC is ambiguous with respect to local or global mesh entity indices. To make DOLFIN 'work' in parallel a confusing mix of local and global indices are used. This issue should be clarified.

[Anders Logg (logg)]

As far as UFC is concerned, shouldn't 'local' refer to local wrt a
cell and 'global' refer to global wrt the mesh known to UFC which is
the mesh on a local processor?

Anything global beyond that should be handled by DOLFIN.

--
Anders

On Thu, Oct 04, 2012 at 04:21:08PM -0000, Garth Wells wrote:
> Public bug reported:
>
> UFC is ambiguous with respect to local or global mesh entity indices. To
> make DOLFIN 'work' in parallel a confusing mix of local and global
> indices are used. This issue should be clarified.
>
> ** Affects: ufc
> Importance: Undecided
> Status: New
>

[Garth Wells (garth-wells)]

On Thu, Oct 4, 2012 at 6:16 PM, Anders Logg <email address hidden> wrote:
> As far as UFC is concerned, shouldn't 'local' refer to local wrt a
> cell and 'global' refer to global wrt the mesh known to UFC which is
> the mesh on a local processor?
>

What we have now is a confusing mixture of local and global, and UFC
says nothing about local and global.

Garth

> Anything global beyond that should be handled by DOLFIN.
>
> --
> Anders
>
>
> On Thu, Oct 04, 2012 at 04:21:08PM -0000, Garth Wells wrote:
>> Public bug reported:
>>
>> UFC is ambiguous with respect to local or global mesh entity indices. To
>> make DOLFIN 'work' in parallel a confusing mix of local and global
>> indices are used. This issue should be clarified.
>>
>> ** Affects: ufc
>> Importance: Undecided
>> Status: New
>>
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1061721
>
> Title:
> Mesh entity indices are ambigious
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/ufc/+bug/1061721/+subscriptions

[Anders Logg (logg)]

On Thu, Oct 04, 2012 at 05:35:49PM -0000, Garth Wells wrote:
> On Thu, Oct 4, 2012 at 6:16 PM, Anders Logg <email address hidden> wrote:
> > As far as UFC is concerned, shouldn't 'local' refer to local wrt a
> > cell and 'global' refer to global wrt the mesh known to UFC which is
> > the mesh on a local processor?
> >
>
> What we have now is a confusing mixture of local and global, and UFC
> says nothing about local and global.

Is it just the terminology that is confusing?

I think UFC is clear in that local means cell and global means mesh
(the part of the mesh it knows which would be the local mesh).

--
Anders

[Garth Wells (garth-wells)]

OK, UFC may be clear but then DOLFIN is wrong. DOLFI N provides a local/global option for UFCCell, and DOLFIN breaks if the wrong option is used. There are demos that use ufc::cell::index to index into a MeshFunction, in which case it must be the local index, but UFC says that it's the global index.