Summary:
In recent year, the performance of Central Processing Unit (CPU) for a computer system has been increased drastically, and it has been required a large scale numerical calculation. While at same time, the limitation of the integration degree and the operating frequency of CPU is approached in past ten years. To cover above limitations, CPU developed having a plurality of processing cores, and the concept of the accelerator such as Graphics Processing Unit (GPU) and Many Integrated Core (MIC) appears. In addition, multi-core CPU, GPU and MIC becoming cheaper and easy to construct high-performance computer system. In order to derive the performance of these devices the parallelization techniques are essential technology. Thus, the parallelization technique should be applied for not only high-performance computing but also huge-scale calculation.
General Purpose computing on GPU (GPGPU) is an area that has attracted much attention, researches about GPGPU have been investigated actively in recent years. In effect, many GPU cluster supercomputers are included in TOP 500 of supercomputer.
In this presentation, including actual case studies, we discuss how to apply parallelization technique of GPU and MIC on the numerical calculation.