Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : November 23, 2018 - November 25, 2018
We have implemented the orthogonal recursive bisection method to parallelize the non-ordinary state-based peridynamics aiming for massively parallel supercomputers. The fracture phenomena expected to be explored by the peridynamics could be extremely complicated and possibly require a vast number of particles. Hence, it is believed that the parallelization is one of the mandatory tasks in the development of the peridynamics. In this study, we examined the parallel performance of our peridynamics code based on the orthogonal recursive bisection method. Using this domain decomposition method with 512 computational cores, we achieved a peridynamic simulation with more than 6 millions particles. It is also shown that the obtained weak scaling performance was close to the ideal one.
