Research On Parallel Efficiency Optimization In 3D Particle Simulation

With the rapid development of computer technology in the past few decades,the electromagnetic particle simulation method has become more and more popular in the research of electromagnetic radiation,particle motion and other physical problems.However,the simulation of complex electrical vacuum devices often consumes a lot of time.In order to solve the problem of slow computation,parallel computation has been applied to the simulation of electromagnetic particles.In recent years,the development of parallel computing technology has brought a feasible scheme to optimize the parallel efficiency.The work of this paper is mainly to conduct the thorough research to the related the development of the parallel computing technology,the CHIPIC-3 d particle simulation software on the basis of Shared memory put forward by the application of the MPI-3 features,design the electromagnetic particle simulation parallel efficiency optimization and dynamic load balancing algorithm,finally the realization of parallel computing in original CHIPIC-3 d on the basis of improve the efficiency of parallel computing.The main work includes the following aspects:1.In-depth study of MPI standard,mainly on the shared memory feature proposed by MPI-3.The point to point communication efficiency of MSMPI is 4.4 times higher than that of MPICH2 by comparing the point to point communication efficiency of MSMPI and MPICH2.The simulation results show that MSMPI is compatible with MPICH2 in code.2.The parallel efficiency optimization scheme of electromagnetic particle simulation based on shared memory is mainly researched and designed.First,a parallel optimization scheme for electromagnetic field update is designed for explicit and implicit recurrence of electromagnetic field.The field array FLDS is stored in a shared memory window,and multiple processes can directly manipulate the data from the window.The proposed scheme is implemented in CHIPIC-3D,and the parallel tests of the waveguide model and the MITL petal cold cavity are carried out.The results show that the parallel efficiency can be up to 46% higher than that of the original parallel calculation.At the same time,in 6 processes,the optimized memory ratio is only 23.6% of that before optimization.Second,a parallel solution of current density is designed to solve the related errors caused by the shared field array.Thirdly,the EM-PIC parallel optimization scheme was designed and implemented in CHIPIC-3D.Simulation tests were carried out on the thermal cavity backwave tube,klystron and magnetic insulated transmission line.The test results show that the EM-PIC parallel optimization scheme is designed correctly and the parallel efficiency can be improved by 44% at the highest.3.A dynamic load balancing algorithm is designed based on the shared memory feature.This algorithm mainly takes the number of particles as the load and dynamically allocates the number of particles that each process needs to process in the process of operation.Through the simulation calculation of the magnetic insulated transmission line,the test results show that the parallel efficiency of the dynamic load balancing algorithm can be improved by up to 38% based on the parallel optimization algorithm.