Acceleration of large-scale multi-physics simulation for biomedical EMC with manycore architecture based computing

Recently, computer simulation scale for biomedical EMC becomes extremely huge, because numerical models, which represent shape, structure, configuration, etc., of biological bodies, increase these preciseness. Parallelizing simulation code is one of the effective solutions to deal with such a high definition and massive scale problem. Finite difference methods, frequently used for solving biomedical EMC problem, are classified as the stencil scheme which has the nature of memory-bound feature. Hence, it is desirable to employ many-core architecture with the high-speed memory bandwidth (approximately 200-300GB/s) such as the graphics processing unit (GPU) or the many integrated core (MIC) hardware accelerators. The purpose of this study is to accelerate the multiphysics simulation system for numerical dosimetries by the many-core architecture based computing technique.