Hello all together, I'm currently working on a modified version of step-37 dealing with a matrix-free implementation of a Laplace problem. As a postprocessing step, I need to solve an interface problem: the function is essentially already implemented in VectorTools::project_boundary_values() <https://www.dealii.org/current/doxygen/deal.II/namespaceVectorTools.html#a5aa5117bc6324d18df70c9cb86442b88>. However, I cannot use the default implementation in the library, because I don't have a boundary function to project, but I have already a nodal vector (the RHS) describing the boundary. Thus, I copied more or less the implementation of 'VectorTools::project_boundary_values' into my program and this works out. However, the implementation relies on a matrix-based approach and there are lots of data structures I need to setup for this small and simple system. I was wondering if I can re-use the MatrixFree object for this purpose, as it has already all the preconditioners and operators implemented.
The idea would be to pass a global vector into the cg.solve() function and (instead of looping over all cells) looping over all boundary cells and only perform operations on the desired faces belonging to the relevant interface. There are mainly three questions I was wondering about: (1) will the SolverControl emit convergence although I operate only on some of the DoFs (do I need to reset the remaining ones) ? (2) are there other options to achieve this (solving a problem on a subdomain such as an interface) using matrix-free (3) do you think the approach makes sense or should I just go with the sparse matrix-vector approach since the overall system will be anyway relatively small (dim - 1)? Thanks in advance and kind regards, David -- 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 on the web visit https://groups.google.com/d/msgid/dealii/5055feff-a2c4-4679-9cf5-8ed2fa6dc73cn%40googlegroups.com.
