Distributed and Adaptive Fast Multipole Method in Three Dimensions

Abstract

We develop a general distributed implementation of an adaptive fast multipole method in three space dimensions. We rely on a balanced type of adaptive space discretization which supports a highly transparent and fully distributed implementation. A complexity analysis indicates favorable scaling properties and numerical experiments on up to 512 cores and 1 billion source points verify them. The parameters controlling the algorithm are subject to in-depth experiments and the performance response to the input parameters implies that the overall implementation is well-suited to automated tuning.