On Tue, Oct 11, 2016 at 4:10 PM, Roy Stogner <royst...@ices.utexas.edu>
wrote:
>
>
> On Sun, 9 Oct 2016, David Knezevic wrote:
>
> On Thu, Oct 6, 2016 at 10:34 PM, Roy Stogner <royst...@ices.utexas.edu>
>> wrote:
>>
>> I think the proper fix is to call the coupling functors in
>> scatter_constraints(), then query the processors who own the
>> elements
>> which are to be coupled for any constraints on those elements' dofs.
>>
>
> Just about done.
>
> Except now I'm paranoid, because I ran across my years-old comment,
>
> // Next we need to push constraints to processors which don't own
> // the constrained dof, don't own the constraining dof, but own an
> // element supporting the constraining dof.
> ...
> // Getting distributed adaptive sparsity patterns right is hard.
>
> And I don't recall *why* we had that need! When we're doing an
> enforce_constraints_exactly() it's only important to have all our own
> dofs' constraints and their dependencies. When we're constraining an
> element stiffness matrix we need to build a constraint matrix C, but
> that constraint matrix only requires us to know about constrained dofs
> on the local element, not about constraining dofs on the local
> element.
>
> So why the heck did I think processors needed to know about everywhere
> a locally-supported dof constrain*ing*? If I can't figure out what
> the reason was then I can't decide whether or not it will be
> applicable to coupled-ghosted elements too.
>
hmm, I agree with you, I can't see why the constraining dofs are
required... maybe we can proceed on the assumption that they aren't
required, and do some tests to see if we hit a case where that assumption
is wrong?
David
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel
