I would agree, but I'll defer the design to Roy, Paul, and anyone else who 
might like to weigh in. I don't usually use these aspects of the API. 

-Ben



On Jun 11, 2013, at 5:29 PM, "David Knezevic" <[email protected]> 
wrote:

> It seems DiffContext and FEMContext would have to be modified to make 
> the below possible.
> 
> How about adding elem_neighbor_jacobian, neighbor_elem_jacobian, 
> neighbor_neighbor_jacobian, and neighbor_residual to DiffContext, and 
> providing an API for optionally initializing these and adding them to 
> the global matrix, rhs?
> 
> I believe something like this would be required for DG in FEMSystem in 
> general, so it seems worth adding, no?
> 
> 
> 
> 
> On 06/11/2013 06:02 PM, David Knezevic wrote:
>> I recently posted about assembling "DG-like" terms on an interface. 
>> With some help from Ben re augmenting the sparsity pattern this now 
>> all works fine, as in the pull request for miscellaneous_ex9.
>> 
>> But my next step is to get this working with reduced basis code, and 
>> the libMesh reduced basis code uses FEMContext for assembly.
>> 
>> Roy, or other FEMSystem folks: Any thoughts on how to assemble these 
>> "DG-like term" using a FEMContext? It's not obvious to me how to deal 
>> with the dof indices from the "other side" of the interface in a 
>> FEMContext? (i.e. the Ken, Kne, Knn matrices in the example)
>> 
>> Thanks,
>> David
> 
> 
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by Windows:
> 
> Build for Windows Store.
> 
> http://p.sf.net/sfu/windows-dev2dev
> _______________________________________________
> Libmesh-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/libmesh-devel

------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:

Build for Windows Store.

http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
Libmesh-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-devel

Reply via email to