Hello all, I am using deal.ii to solve a system of Hamilton Jacobi equations. I need to use FE_Nothing element in a part of my domain.
I read a couple of posts in this forum regarding this but I have some questions. I will appreciate if anyone can answer these. Since I need to use MPI, and I need to use hp::DofHandler so I cannot use p::d::triangulation. So, I need to use Triangulation and then use metis to partition that. In one of the posts I read "you can always partition regular triangulation by GridTools::partition_triangulation (n_mpi_processes, triangulation); and adjust your assembly loops to if (cell->subdomain_id == this_mpi_process) to parallelize with MPI. The downside is that all your processes will now own the **complete mesh**. There are also some other minor downsides like serial SolutionTransfer and not straight-forward serialization/restart." 1. So, I wanted to know what are the minor downsides with solution transfer and restart. I was using Parallel Distributed Solution transfer upto now and need to have checkpoint/restart option available. 2. Is there no concept of locally owned or locally ghost cells if I use the GridTools::partition_triangulation()? Uptil now, some calculations I was doing were for both (loc. owned + ghost) cells like moving the mesh. What happens here? 3. Will there be any change regarding locally owned dofs and locally relevant dofs? 4. Even with p::d::tria, the coarsest mesh was owned by all processors. So, what exactly does this mean -- "The downside is that all your processes will now own the **complete mesh**." Thanks. -- 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 dealii+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
