GPU Acceleration Of PIC/MCC Simulation In Laser Plasma Interation

Energy is playing an important role in our daily production and life. With the increasing depletion of fossil fuels, humanity is facing an unprecedented energy crisis. The problem of energy development and innovation is an issue concerned by our government and all the world. Our social and economic development must focus on solving this problem.Fusion reaction can release large amounts of energy and less pollution to the environment, is an ideal substitute for fossil fuels. So the inertial confinement fusion has become the main object studied by all the scientists. Fast ignition is an important way to realize the inertial confinement fusion. It has attracted many scholars to study it, in which theory and numerical simulation is an important means for fast ignition research now. However, due to the ultrashort and ultrastrong laser, high density and across the large magnitude charged particles in fast ignition process, this produce a huge computational burden and extend the study period in fast ignition key physical problems simulation. With limited resources of CPU, people began to look to another processor--GPU. Because GPU has an unique structure to achieve large-scale parallel computing, therefore, GPU is much favored with researchers in the field of high-performance computing.This paper is based on relativistic laser plasma interactions simulation software LPICMCC ++ which contains collision effects developed by UESTC, studied the parallel technology of laser-plasma PIC/MCC methods, developed a GPU-based parallel simulation of laser-plasma interaction algorithm. At the same time, we completed the parallel LPICMCC++ code, including writing, debugging and testing. The main contents are:(1) Introduce the background and significance of this paper work. Summarize the basic principles of inertial confinement fusion, briefly introduced dynamic and current situation of inertial confinement fusion and the fast ignition at domestic and abroad. Explain GPU-based parallel computing, hardware architecture and programming model in detail.(2) Describe the numerical simulation algorithm of relativistic laser plasma interactions simulation software LPICMCC++ including collision effect in detail, including: the solving of particles moving, the solving of charge and current density, the solving of field, laser ionization, electron collisional ionization, binary collision. Above algorithm and serial software based on above algorithm are the foundation of parallesim LPICMCC++ software.(3) In LPICMCC ++ software parallelization example, detailed explanation PIC/MCC methods approach ported to the GPU, including hotspots analysis, improving original LPICMCC++ serial code and the solution of each module in particle simulation of the application in CUDA.(4) In the physical model of ultrashort and ultrastrong laser plasma interaction as an example, compare the two results calculated by serial LPICMCC++ code and parallel LPICMCC++ code, verify the correcthess of the LPICMCC++ program parallelization, calculate the error of these two program. Finally, calculate the GPU acceleration parallelism LPICMCC++ software speedup based on different physical models.