Thank you for your help Timo, that worked well. Best wishes,
Tom On Wednesday, July 23, 2025 at 3:40:25 PM UTC+1 Timo Heister wrote: > Hi Tom, > > What we do in ASPECT is the following: > 1. evaluate the coefficient on the active cells and store in a Table > (for each Q point) > 2. Project into a FE_DGQ(k) on each cell and store in a finite element > coefficient vector (you can pick k, k=0 would be a constant per cell) > 3. interpolate_to_mg() to get a coefficient vector on each multigrid level > 4. convert to a Table (interpolate to quadrature points or store a > constant per cell) > > You can see this in action here: > > https://github.com/geodynamics/aspect/blob/0a1784a421d555240c7710b77fe157aea12bb66c/source/simulator/solver/stokes_matrix_free_local_smoothing.cc#L371 > > We do this because coefficient data is not available on coarser levels > and because, even if it was, it is unlikely to improve the convergence > to have the real coefficient on the coarser levels. > > > On Fri, Jul 11, 2025 at 11:26 AM Thomas Cheetham > <[email protected]> wrote: > > > > Hello everyone, > > > > I am currently developing a matrix-free code which has a non-analytical > spatially varying coefficient. As in step-37 I am using a Table<2, > VectorizedArray<number>> to store evaluations of the coefficient at all the > quadrature points for a given cell batch. > > > > I read my coefficient in from a CSV file corresponding to the values at > DOF in a structured mesh which I am using to generate a FEFieldFunction > object which i call for all quadrature points to initialize the values in > the matrix-free coefficient table. This approach is fairly slow but works > okay for the finest level (where the parallel distributed triangulation is > the same as the structured mesh). However for coarser multigrid levels I > run into issues where the cell within which I am trying to evaluate points > is no longer owned by the current processor. > > > > I was wondering if anyone has any suggestions as to how to get around > this, or perhaps has any ideas for a faster approach? > > > > Many thanks, > > > > Tom > > > > -- > > 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/3f0dd574-5f3a-4300-8803-9a80dd2f318bn%40googlegroups.com > . > -- 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/ef4af5b2-90ac-4328-b5ff-4b717a3f19c9n%40googlegroups.com.
