The effect of parallelization on a tetrahedral mesh optimization method

Abstract

A parallel algorithm for simultaneous untangling and smoothing of tetrahedral meshes is proposed in this paper. This algorithm is derived from a sequential mesh optimization method. We provide a detailed analysis of its parallel scalability and efficiency, load balancing, and parallelism bottlenecks using six benchmark meshes. In addition, the influence of three previously-published graph coloring techniques on the performance of our parallel algorithm is evaluated. We demonstrate that the proposed algorithm is highly scalable when run on a shared-memory computer with up to 128 Itanium 2 processors. However, some parallel deterioration is observed. Here, we analyze its main causes using a theoretical performance model and experimental results.