Improving Load Balancing and Parallel Partitioning in Fluidity

Paul Woodhams
HECToR CSE Team, The Numerical Algorithms Group Ltd., Oxford

Jon Hill, Patrick Farrell
Earth Systems and Engineering,Imperial College London, London


Abstract:

Fluidity is a non-hydrostatic, finite element/control volume CFD numerical forward model used in a number of scientific areas; geodynamics, ocean modelling, renewable energy, and geophysical fluid dynamics. The applications cover a range of scales from laboratory-scale problems, through to whole earth mantle simulations. One of the unique features of Fluidity is its ability to adapt the computational mesh to the current simulated state: dynamic adaptive remeshing. When running on multiple processors the adapted mesh must be load balanced, which involves swapping elements from processor to processor. Zoltan is a library that performs such dynamic load balancing. Here, we document the steps taken to integrate Zoltan within Fluidity which extends the current functionality of Fluidity, thereby extending the range of science areas that it can be applied to. Although there is a small performance hit over the current customised load balancing library, this is only noticeable when the number of elements per process is small.





Jon Hill 2012-03-02