Abstract
In this study, a dynamic load-balancing parallel computation method is investigated to predict the motion of multiple solid objects in fluids, taking account of their collisions and fluid-solid interactions. The parallel computation method is based on the 3D domain decomposition method using flat MPI. The “slice-grid method” enables us to dissolve three-dimensional deviation of the number of blocks included in the “block-computation domain”. On the other hand, the “fluid-computation domain” is fixed throughout the computation. Numerical experiments have been conducted for comparisons of the computational efficiencies between the present and previous methods. As a result, we found 256-process simulation of water mass collapse, including 1,000,000 objects in the water, exert 129.90 speed up compared to sequential execution.
