[petsc-dev] FieldSplit fixes in 3.3

[Matthew Knepley]

On Tue, Aug 21, 2012 at 8:59 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:

> On Tue, Aug 21, 2012 at 8:51 PM, Matthew Knepley <knepley at gmail.com>wrote:
>
>> I thought we agreed in this thread that we were (for now) going with
>>> Matt's bastardized model of attaching the Schur null space to A11. Doesn't
>>> that mean that this hunk should also be reverted (and have a comment
>>> explaining this indirect effect)?
>>>
>>
>> For future reference, this was my bastardized model in 3.3, but in
>> petsc-dev I either
>>
>>   a) attach them to IS on input, which works beautifully
>>
>> or
>>
>>   b) Tell the DM about them
>>
>
> In either case, what happens when you switch back and forth between Schur
> and, e.g. multiplicative? Does that cause there to be a different IS or a
> different DM?
>

You are right, it has the same conceptual flaw.


> I fear that by including the physics in the DM, we may be obligated to
> have a DMGetSchurComplement() (or, in the more general nonlinear language,
> DMEliminate()). Note that some mixed discretizations have sparse Schur
> complements and it could even make sense to implement a nonlinear smoother
> in the reduced space.
>

I really do not want to do that, at least not now since we have not even
shaken out all the FS implications or have a library of common split PCs,
nor
have we gotten nonlinear FS working. After that, we can do DMEliminate(),
if DM still exists.

   Matt

-- 
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/20120821/d966569e/attachment.html>