Solving a TrilinosWrappers::BlockSparseMatrix using Trilinos iterative solver (step-57)

[Bruno Blais]

Hello,

I am currently trying to work on a "HPC-ready with Trilinos" version of step-57 for steady-state solution of the Navier-Stokes equation since I am curious about compairing it to our stabilized solver. I think that it  could give much better results for steady-state.

The approach detailed in step-57  uses Block matrices instead of regular matrices since it preconditions the pressure block differently than the momentum block.

From my understanding, this requires eventually the solution of a BlockSparseMatrix. However, looking at the documentation for TrilinosWrappers::SolverBase, I see that there are no call to solve that can take as an argument a BlockSparseMatrix, they all take a regular matrix. Consequently, I am not sure how I should proceed. Is this something that is lacking in current implementation or there is a work around that I am unaware of?

What I am trying to do initially is to port step-57 (https://www.dealii.org/current/doxygen/deal.II/step_57.html) to use Trilinos for the linear algebra. Consequently, what would be changing from step-57 is mostly related to changes in the BlockSchurPreconditioner class of step-57. Hope that is clear :).

Thanks!

Bruno

[Daniel Arndt]

Bruno,

You can just use the deal.II solvers instead. Have a look, e.g., at step-32.

Best,

Daniel