[petsc-dev] Grouping vector dofs in multi-component problems

[Jed Brown]

On Tue, 26 Jan 2010 11:05:27 -0600, Matthew Knepley <knepley at gmail.com> wrote:
> This still only applies to the very simple schemes which have the same
> # dofs at each site. I have a more complicated thing for unstructured
> FEM. It would be nice if both had the same interface.

Structured vs. unstructured has nothing to do with it.  A mixed
discretization is definitely different, but I use DMComposite for such
things so that each DM still has a natural block size.  Vector elements
are special and would require the user to manually specify the
low-energy spaces (or the special mixed DM to understand it).

Note that I'm suggesting this as something that goes in DM, PC would
remain purely algebraic and always speak in terms of low-energy
subspaces.

Jed