抄録
In this research, a new distributed memory parallel algorithm of the explicit MPS (Moving Particle Simulation) method is presented. The analysis region is divided for a distributed memory parallel computation using ParMETIS. Two communication techniques of an overlapping method and a non-overlapping method are estimated by parallel scalability tests. Since we find that load balance is most important for the distributed memory parallel algorithm of the explicit MPS method, we find that the non-overlapping method is more effective than the overlapping method. As a result, we have been able to do the MPS analysis of 268 million particles in 38 seconds per one time step. Performance during large scale simulation is examined by computing tsunami wave run-up on a virtual gulf area using up to 58 million particles.
