A Research Of EM-PIC Parallel Computing Based On GPU

Electromagnetic particle-in-cell(PIC) simulation method is a method that simulation the movement of particles and the interaction between particles and Electromagnetic environment.By this method,we can simulation Vacuum Electronic Devices and prediction the operative mode of them.As the wide application of computer technology,simulation through computer becomes the main part of developing advance Vacuum Electronic Devices.With computer simulation,we can prediction the parameter of devices before experimentation,and prediction movement of particles visualized.While the program of simulating EM-PIC is time consuming,it becomes more and more important to exploit more effective programs and compute devices.Add GPU Parallelism to personal computer can be a available way to accelerate programs.The main research is as follows:1.Study the technology of GPU Parallelism,realize the program of GPU Parallelism based on Fortran.Apply OpenACC and CUDA in a serial matrix multiplication program,and compere the result and efficiency between them.Analyze the advantages and disadvantages of the two methods and optimization of the programs.2.More study of EM-PIC simulation algorithms based on CUDA.Build a rational grid model,allocate the arithmetic to threads of GPU.and analyze the overhead of accessing cache.3.Applications of the CUDA program take part in CHIPIC-3D, which is a soft used for simulating high power microwave sources. Based on GPU,the parallel computing capability of CHIPIC-3D is implemented.Optimizing the electromagnetic parallel algorithms part, acceleration of the program is implemented.The simulation of rectangular waveguide and magnetically insulated transmission line(MITL) by accelerated program is competed.4.The problems that implementing the simulation of PIC based on CUDA is Analyzed and summarized.Come up with the possibly solutions,and prepare the foundation of future research.