Locally-Refined Grids on Parallel Computers

If the evolution of small-scale structures in large domains is to be computed, a reasonable efficiency can only be achived by using locally-refined grids on parallel computers. We perform our 3d MHD simulations on computers with distributed memory (at the moment an IBM SP) using MPI. We use a tetrahedral grid which is stored in a tree structure. The grid concept was developed and implemented by Bernhard Schupp for his PhD thesis.

(Click to see a 2.6 MB movie)

Related Publications: