GPU Acceleration Of Two Dimensional Particle-in-cell In Vacuum Electronic Device

Numerical simulation technology has been playing a key role in the development of vacuum electronic device, but with the new structure, new materials and new technology of new vacuum electronic device in the urgent need and rapid development,the applicability of the numerical simulation technology has been confronted with more and more challenges. At the same time, with the rapid development of computer,because of its particularly suitable for solving the problem which the physical laws are not very clear and difficult to analyze in the mathematical way, Particle-in-Cell (PIC)simulation technology has been increasingly showing its advantages. .But PIC simulation method costs much computer resources and time, it takes dozens of hours or even weeks to complete the simulation of the complex structure of the vacuum electronic devices, this greatly limits the application of PIC technology in engineering design. However, with the enhancement of algorithmic ability and the continuous improvement of computer computing power, this restriction is gradually being eliminated.GPU is a heterogeneous multiprocessor chip, which has a massively parallel processor core. Using it for parallel computing can greatly shorten the running time of the program. The appearance of CUDA allows programmers to use the GPU for parallel programming without the need to grasp the complex theory, so that the GPU can be developed rapidly. PIC technology has natural parallelism, so based on CUDA C/C++language and its library functions, it is one of the feasible ways to improve the simulation efficiency by accelerating the existing PIC simulation program by GPU.In this paper, based on the application of vacuum electronic device, the 2D3V electromagnetic PIC simulation algorithm is deduced, programmed and verified in the x-y coordinate system, and then GPU is used to optimize the program. The work mainly includes:1. Introduced the importance of PIC simulation technology in the research of vacuum electronic device and illustrated the excellent performance of GPU, then stated the advantage of parallel implementation of PIC simulation algorithm by GPU;2. Summarized and outlined GPU architecture and CUDA programming;3. Derived the 2D3V electromagnetic PIC simulation algorithm based on the application of vacuum electronic device in x-y coordinate system. The algorithm includes the updating of the grid's electromagnetic field, the updating of the position of the particle velocity and the density of the current.4. Developed the 2D3V electromagnetic PIC simulation program and simulated two kinds of Beam-wave interaction. Then compared the simulation results with the results of the particle simulation of PIC software MAGIC and verified the accuracy of the program.5. Established a GPU acceleration algorithm in view of the above 2D3V electromagnetic PIC simulation. Then used CUDA to achieve the parallelism and optimization of the program. Then compared and verified the serial results and parallel results, and finally gave the speed-up ratio.