[petsc-dev] DMMG without libpetscsnes?

[Barry Smith]

On Jan 7, 2010, at 2:16 PM, Jed Brown wrote:

> On Thu, 7 Jan 2010 14:09:21 -0600, Barry Smith <bsmith at mcs.anl.gov>  
> wrote:
>>    Yes we would lose that functionality. But maybe ok since the
>> Cascadic multigrid is pretty much mesh sequencing.
>> We could avoid supporting adaptive mesh refinement for linear  
>> problems
>> (linear problems are silly anyways :-)).
>
> Yes, but putting GMG inside a Schur fieldsplit preconditioner is
> something I expect to be doing pretty frequently, and wrapping the
> linear operation inside a SNES feels goofy in this context.

    There would be no SNES in this context!  One would be able to  
cleanly and simply (more so then today) put geometric multigrid inside  
a Schure fieldsplit preconditioner. If one can't then we screwed up  
our new design.

    Barry

>
>> Yes this is a known problem (at least known by me :-) that will need
>> to be resolved with the next approach.
>
> Great, I just wanted to make sure it's on the radar.
>
> Jed