On 11/14/25 11:20, Nihar Bhardwaj Darbhamulla wrote:
I have added the code snippet initializing and running the CG+AMG
preconditioner below. As it happens, the inner (1,1) block arises out of
the symmetric discretization of a Stokes system (as in Step-22), and
using BoomerAMG as is does not allow for the recognition of multiple
DoFs associated with each node, while this is an option available in
using BoomerAMG through PETSc ( /*-pc_hypre_boomeramg_nodal_coarsen*/ ),
but is not exposed through the dealii interface. The solutions I can
think of are:
1. adding the option to the PETScWrappers source code and testing it.
I'm not sure if it has been tried before with the PETSc Wrappers,
and whether there is a certain re-ordering that is to be kept in
mind to execute this successfully.
2. Switch to using Trilinos wrappers. However, my research group has
more background on using PETSc, and the switch would come at a loss
of support available from a linear algebra library POV.
3. Modifying the formulation to go from a symmetric gradient weak-form
to a vector laplacian weak-form. This comes with modeling
justifications and would be problem specific.
For now, I would prefer option 1, but would like to know if it has been
tried before, and if any caveats are associated with using it.
I don't know whether it has been tried before, but in practice all you'd
have to do is add the necessary information to the AdditionalData
object, and in the place where we set up the AMG preconditioner in
deal.II, i.e., the place where we currently already use the
AdditionalData information, extend the code to now also support the
nodal coarsening. You'll want to look up how the existing
PreconditionAMG class is implemented.
Of course, we would be excited to merge such a patch into deal.II!
Best
W.
--
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see
https://groups.google.com/d/forum/dealii?hl=en
---
You received this message because you are subscribed to the Google Groups "deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion visit
https://groups.google.com/d/msgid/dealii/549d0eb0-b5e4-4c23-931f-bdc647a55b76%40colostate.edu.
