[petsc-dev] Mat/DM dependency

[Matthew Knepley]

On Fri, Jan 1, 2010 at 1:58 PM, Jed Brown <jed at 59a2.org> wrote:

> On Fri, 1 Jan 2010 13:32:08 -0600, Matthew Knepley <knepley at gmail.com>
> wrote:
> > I would rather see then hierarchy change.
>
> Maybe, but many things in the DM interface depend on Mat and Vec.
> Where would DMGetInterpolation and DMGetMatrix end up?
>

That is definitely a problem. It makes me think we need an R^N part just
like we have now, and then a more general part with operators on fiber
bundles.


> > I think it is natural for the operator (Mat) to depend on the space
> > (DM) on which it is discretized.
>
> Agreed.
>
> > Just because shortsightedness in the past has confined us to really
> > simple spaces (R^N) does not mean we can't change that.
>
> KSP operates in a finite dimensional linear space, i.e. isometric to
> R^n, so at least up to that level, we're not really exploiting the more
> general interpretation.
>

It depends how you see it. KSP can operate just as well on a fiber bundle.

   Matt


> Jed
>
-- 
What most experimenters take for granted before they begin their experiments
is infinitely more interesting than any results to which their experiments
lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20100101/ed5b376c/attachment.html>