On Jan 27, 2012, at 5:40 PM, Matthew Knepley wrote:

> 
> We can take this to dev, but would you advocate doing this by making new 
> comms?
> 
> > I want blocks of MatNest to be able to reside on subcomms. If that 
> > happened, inner KSPs for PCFieldSplit could also move to subcomms. PCMG 
> > should be able to run coarse levels on subcomms.
> 
> This seems like an excellent source of new and confusing parallel errors. How 
> do we control organization of
> subcomms, compatibility (checking without deadlock), and reporting? Should we 
> maintain a relation stored
> in the communicators (comm graph), or outside?
> 
> However, I think this is the right move. 

   Is now the right time. Shouldn't we wait until MPI's replacement is working 
and do things with that model?

   Barry

> 
>    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


Reply via email to