Known issues and wishlist - vonkarmaninstitute/pantera-pic-dsmc GitHub Wiki

Known issues

Users please use the "issues" menu of GitLab and open a new issue.

Performance improvements

• While gathering moments on the grid (in postprocess.f90) currently, each MPI process stores a copy of the entire grid. This can become very expensive for large 2D domains. Additionally, a MPI reduce operation is required at the end. We could instead make each MPI process own only its cells, and just an MPI gather would be required at the end.
• UMFPACK works on the root process only. We should use a parallel solver such as MUMPS. We should investigate interfacing with more popular libraries like PETSc.
• In the pantera-implicit branch, the mass matrix stores all nodes, but only a N-by-3 matrix is in theory required.
• Multiple initializations in different domain regions.
• Optionally write output data in VTK binary files.