DYNAMIC LOAD-BALANCING PARALLEL COMPUTATION METHOD FOR MULTIPLE ARBITRARILY-SHAPED OBJECTS IN FLUIDS

It is important to predict the movements of multiple solid objects driven by free surface flows in case that the flood and Tsunami occur. In this study, essential improvement was performed for our multiphase parallel computational method, in which multiple complicated-shaped solid objects are transported by the pressure and viscous effects of the predicted multiphase model. Our parallel computational method is based on the domain decomposition method. Thus, the improvement was achieved by dynamically adjusting only the ``block-computational domain'' according to the number of blocks included in them. On the other hand, the ``fluid-computation domain'' is fixed throughout the computation. As a result, we found 64-process simulation of water column collapse, including 12,800 objects in the water, exert 24.52 speedup compared to sequential execution.