Chris Walshaw and Martin Berzins
Modern PDE solvers increasingly use adaptive unstructured meshes in order to discretise complex geometries and control numerical error. The problem of dividing up the domain equally for a distributed memory parallel computer whilst minimising the inter-subdomain dependencies can be tackled with graph-based methods such as Recursive Spectral Bisection. This paper describes an extension to such methods which renders them more suitable for time-dependent problems where frequent remeshing may occur, possibly with only relatively small changes to the mesh. Numerical testing shows that this new approach gives a good speed-up for mesh partitioning when used to enhance Recursive Spectral Bisection. It is also shown that the overall computational time can be less than that needed when cheaper but more naive load-balancing algorithms are used.