A Scalable Parallel Fast Marching Method

The fast marching method is an efficient strategy to solve the Eikonal equation with broad applications in computational fluid dynamics. However, the traditional fast marching method is a purely sequential algorithm and thus not straightforward to parallelize. In this presentation, four parallel fast marching methods are discussed: the non-blocking parallel fast marching method (NB-PFMM), the blocking parallel fast marching method (B-PFMM), the extended domain-blocking parallel fast marching method (DB-PFMM), and the layered-blocking parallel fast marching method (LB-PFMM). When combined with proper domain decomposition approaches, these methods are not only efficient but can scale over a wide range of processor counts. The applicability and performance of the different parallel fast marching methods are presented and compared as applied to a variety of test cases.