[petsc-dev] DMViewFromOptions()

[Barry Smith]

On Dec 15, 2013, at 10:17 AM, Matthew Knepley <knepley at gmail.com> wrote:

> On Sun, Dec 15, 2013 at 8:54 AM, Barry Smith <bsmith at mcs.anl.gov> wrote:
> 
>   Matt,
> 
>     WHEN do you want the viewing to be done? That is where the DMViewFromOptions() should be called, and you should move it there. Surely you don’t wish to view the DM in DMSetFromOptions() where presumably nothing is setup?   Note that in analogy, MatViewFromOptions() is called in MatAssemblyEnd().
> 
>     Since we are going to allow XXXViewFromOptions() to be turned off with a master switch for performance, we can just put it were we want it and don’t need to call it one place and then use the resulting viewer somewhere else.
> 
> For now, I would like to remove it from DMSetUp() where it has no meaning, and just make the user put it in manually.
> Is that alright?

   Yes, it can also go into the end of DMSetUp_DA(). 

    Is there some concept of a DMAssemblyEnd() stage where one knows that the DM truly is ready to go? Should there be?

   Barry

> 
>    Thanks,
> 
>       Matt
>  
> 
>    Barry
> 
> On Dec 14, 2013, at 11:15 PM, Matthew Knepley <knepley at gmail.com> wrote:
> 
> > The move to DMSetUp() has broken every DMPlex test. The full mesh
> > is usually not constructed at DMSetUp(), which just allocates memory.
> > This is analogous to MatSetUp(), for which we do not expect values to
> > be set.
> >
> > I don't think the ViewFromOptions call makes sense in DMSetUp() and
> > think it should be moved back to SetFromOptions.
> >
> >    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
> 
> 
> 
> 
> -- 
> 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