[petsc-dev] Use of dmmg->computefunction ?

[Barry Smith]

On Jan 14, 2010, at 4:23 PM, Jed Brown wrote:

> On Thu, 14 Jan 2010 15:50:25 -0600, Barry Smith <bsmith at mcs.anl.gov>  
> wrote:
>>
>>   How many times am I going to have to pound it into your head that
>> DMMG is a misconceived, poorly designed piece of junk :-(
>
> Okay, okay, I get it [1].  But DMMG still does cool stuff that is
> currently an unreasonable amount of work to do without using it.  Of
> course doing this without DMMG is very interesting to me because I'd
> like some of these things to happen inside of a time integrator, and  
> TS
> has no business knowing about DMMG.  I can currently make this happen
> with some funky patching of the SNES so that SNESSolve calls my  
> function
> that picks up the DMMG and calls DMMGSolve, but that's not ideal.
> Anyway, while thinking about this, I end up browsing DMMG source and  
> get
> curious when I can't find an algorithmic reason for something.

    This browsing is good. You can help us when we move the "cool  
stuff" from DMMG into SNES/KSP where it belongs and where it can be  
used automatically by TS :-)

    Barry

>
> Jed
>
>
> [1] You have high standards.  :-)