On May 14, 2011, at 3:51 PM, Jed Brown wrote:
> On Sat, May 14, 2011 at 22:39, Barry Smith <bsmith at mcs.anl.gov> wrote:
> and get things like MatGetVecs() to properly pass local to global
> information to the created vectors.
>
> This currently happens.
Good.
>
> So for example we could have something like
>
> typedef struct _n_PetscLayout* PetscLayout;
> struct _n_PetscLayout{
> MPI_Comm comm;
> PetscInt n,N; /* local, global vector size */
> PetscInt rstart,rend; /* local start, local end + 1 */
> PetscInt *range; /* the offset of each processor */
> PetscInt bs; /* number of elements in each block (generally for
> multi-component problems) Do NOT multiply above numbers by bs */
> PetscInt refcnt; /* MPI Vecs obtained with VecDuplicate() and from
> MatGetVecs() reuse map of input object */
>
> PetscInt nfields; IS *fields;
> IS localtoglobalmapping,blocklocaltoglobalmapping.
> USER INFO LIKE boundary condition locations;
> };
>
> My only concern here is that there is a bit of duplicated information between
> the IS (or ISLocalToGlobalMapping?) and the PetscLayout.
What is that duplicate information? Can it be removed or can we provide
automatic consistency checks so that the "duplication" is not a problem?
> I would like to see the local space become a more central property of Mat and
> Vec (this would simplify DMComposite and allow fusing the scatters which
> should be good for performance), so putting it in PetscLayout makes sense to
> me. For (rare) memory reasons, I think it should be possible for the
> ISLocalToGlobalMapping to be NULL.
Of course it can be null now and would be able to be null in the future.
Barry
