Gpu Based On Particle Simulation High-performance Computing Systems

Computational science, parallel with the theory and physical experiments, has become the third pillar of scientific research. Because of its computing needs are endless, so high-performance computing in computational science to become a critical driving force. Improving the calculation speed of the method is generally to use a faster processor and other hardware devices, or use a more optimized method and library programming. In recent years, CPU has been predicted in accordance with the pace of Moore's Law, but using the method of increasing the frequency of CPU has encountered the technology bottlenecks, being unable to break through. On the other hand, using this approach led to excessive power consumption and heat, but also restricted the rapid development of CPU. GPU is a processor, originally processing graphics tasks. Because of its unique architecture make it well suited for high efficiency and low cost, high-performance parallel computation. Rapid development in recent years, GPU high-performance computing has gradually become an important branch.Molecular dynamics and cosmology, as two important areas of application for computational science, usually need to solve complex scientific calculations and data processing problems. Particle simulation is often encountered in these two areas. But the current CPU-based high-performance computing technology can not meet the above areas to quickly solve the problem of particle simulation requirements, thus greatly restricting the research process. Therefore, this article proposes a GPU-based high performance computing solutions to solve the computational bottlenecks on the particle simulation problems. This article will first design and construct of high-performance computing based on GPU hardware platform, and build the software environment for GPU computing, and then the two issues in the field of particle simulation applications ported to the platform. Finally, we use these two key areas of test cases for calculation and validation on the GPU high-performance computing systems. Experimental results show that the new technology based on high-performance GPU computing solution will greatly enhance the speed of computer simulation, and computing systems can greatly reduce the costs. Today in promoting energy saving, this advantage not only inspires the construction of supercomputers, but also makes the desktop supercomputer possible.