Font Size: a A A

High Performance Parallel Optimization Of Hybrid Simulation Algorithms For CPU,GPU And Its Heterogeneous Forms

Posted on:2021-03-23Degree:MasterType:Thesis
Country:ChinaCandidate:Y JinFull Text:PDF
GTID:2428330602478882Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
High performance computing(HPC)usually uses a certain number of processors or several computers in the network management cluster to calculate the problem.Parallel computing is one of the ways to solve the problem of high performance computing.It can make multiple instructions run at the same time,which can be divided into time parallel and space parallel.At present,high-performance parallel computing is one of the important research methods in physics,biology,chemistry and other fields.Nowadays,more and more fields begin to solve practical problems in scientific research with the help of high-performance parallel computing technology,such as the numerical simulation method widely used in space science.Researchers use one-dimensional,two-dimensional and three-dimensional numerical simulation programs for scientific research and have achieved rich results.With the development of scientific research,these simulation methods need more and more accurate calculation,which requires very high performance of computer.For this reason,parallel computing has become a hot research topic.This paper focuses on high performance parallel computing.Firstly,from the development background of numerical simulation,this paper introduces the application and development status of high performance parallel computing in space science;secondly,it introduces the numerical simulation method used in this paper,describes the physical model and principle of the method,and realizes the series sequence of the method,introduces the program flow and basic structure;secondly,it uses the parallel computer based on GPU and MPI respectively The method optimizes the hybrid simulation program and tests the performance of the program.The results show that using GPU or large-scale cluster to do parallel computing in spatial numerical simulation has a significant performance improvement compared with using a single CPU,and has a higher cost performance.Finally,the influence of the cavity ion density on the formation of the electrostatic shock wave in the lunar wake is analyzed by using the optimized hybrid simulation program.The evolution of the density structure and the formation of the electrostatic shock wave in the lunar wake are simulated.It is found that the initial cavity ion density plays an important role in the evolution of the density structure/electrostatic shock wave in the process of plasma filling the lunar wake The practical application of the self-developed hybrid simulation program in scientific research,and the correctness and reliability of the program are proved by the experimental results consistent with the theory.
Keywords/Search Tags:high performance computing, parallel computing, spatial numerical simulation, hybrid simulation, GPU, MPI
PDF Full Text Request
Related items