Hi all, >> I just want to point out one aspect to all who are interested in this: Only >> the mesh refinement is distributed, the original mesh itself is still held >> on all processes. > > That is true -- the coarse mesh is stored on all processors.
Note that also only isotropic refinement is possible with the distributed::Triangulation. This is a limitation from the library p4est that manages the distributed mesh. We also haven't implemented support for hp-adaptivity (which should be doable though). By the way, the distributed Triangulation is only the most prominent part of the recent developments. Other projects might profit from various other features and performance enhancement we had to work to make it work, i.e.: - support for ghost elements in Trilinos and PETSc Vectors - fast Indexing class IndexSet, which is used in many parts of the library - more efficient mesh smoothing - memory savings in parallel computations in e.g. CompressedSimpleSparsityPattern, ConstraintMatrix and probably a few other things I forgot. :-) -- Timo Heister http://num.math.uni-goettingen.de/~heister _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
