A parallel iteration algorithm for greedy selection based IDW mesh deformation in OpenFOAM

Inverse Distance Weighting (IDW) interpolation based mesh deformation is an explicit technique adopted by large-scale mesh computation for Computational Fluid Dynamic (CFD). To obtain the optimum interpolating points set, greedy algorithm is utilized widely in the interpolation process. However, greedy algorithm will bring additional time cost because of its recycling point selection. This paper proposes a parallel iteration algorithm for greedy selection based IDW mesh deformation in OpenFOAM (a widely used software of CFD).The iteration method is proposed based on that some computing values are recycling in greedy loop and it make the computational complexity reduce from O(n3) to O(n2). The parallel algorithm is based on the re-decompositon method which can greatly keep the load balancing. The experiments are operated on a three-dimensional undulating fish model with 3.7 × 105 mesh cells. Results show that our parallel iteration method could obtain a more than 18.2 times performance increase, compared with the conventional greedy algorithm applied in IDW interpolation.