An Efficient Parallelization Strategy For The Adaptive Integral Method Based On Graph Partitioning

Abstract

The adaptive integral method (AIM) is frequently used to efficiently solve scattering problems involving conducting objects immersed in a uniform or stratified medium. We propose an efficient parallelization strategy for AIM suitable for distributed-memory computers. In the proposed approach, a weighted graph is associated to the given mesh, and partitioned in a way that results in an efficient distribution of computations across different processes. Numerical tests show that the proposed algorithm compares favorably to existing techniques to parallelize the AIM.