A particle-based parallel scheme for material point method (MPM) using message passing interface (MPI)

Material point method (MPM) requires a high computational cost associated with mapping information between a background grid and material particles even though it has a great advantage of solving large deformation problems. This paper presents a particle-based parallel scheme for MPM using message passing interface to effectively reduce the computational cost of MPM. Unlike grid-based parallel schemes, computational workloads are evenly distributed over the computing cores by partitioning material particles into balanced particle subsets. The particle-based parallel scheme guarantees an optimal load balancing between the computing cores because the same number of material particles are assigned to each core. The performance of the present particle-based parallel scheme is compared with that of a grid-based parallel scheme.