Consideration of optimum algorithm for acceleration of a multi-phase flow analysis with GPU

Nagatake, Taku ; Kunugi, Tomoaki*; Takase, Kazuyuki

GPU (Graphics Processing Unit) has been developed for drawing system in computer. In order to achieve the high-performance computing for rendering high-quality 3D graphics, the GPU has many processing units and the wide memory bandwidth. Nowadays, the performance of GPU becomes much higher than that of CPU (Central Processing Unit), so that the attempt to make use of GPU for the science calculation and general purpose is performed: This attempt is called "GPGPU" (General Purpose GPU) or "GPU computing". Now we are developing MARS (Multi-interfaces Advection and Reconstruction Solver) (1) with GPU. The MARS is one of the surface volume tracking methods for multi-phase flows. In this paper, the performance of GPU is compared to the CPU. And the parallelized method for the multi-GPU system is considered.