Parallelization And Optimization Of Laser-Plasma-Interaction Simulation

In recent years,the progress in generating intense ultra-short laser pulse demand more and more for kinetic descriptions of the interaction of such laser pulse with plasmas.Particle-In-Cell(PIC)algorithm is a widely-used method in plasma physics to study the trajectories of charged particles under electromagnetic fields.We should make good use of the modern rapidly developing HPC hardware and software technique,especially utilizing CUDA on GPU,if we want to conduct the simulations involving more particles.Though there have been some implementations of PIC algorithm on GPU,some important issues still need to be clarified in detail,based on the characteristic of the Laser-Plasma-Interaction simulation.Our three major contributions are:(1)Introducing a way to change the original CPU Laser-Plasma-Interaction code into a parameterized adaptive GPU implementation with the whole algorithm ported.(2)Developing a series of methods to speed up the particle scatter phase: Dynamic Duplication Algorithm,mix-precision computing and a parameterized particle sorting algorithm.(3)Utilizing the GPUDirect RDMA technique in a Kepler cluster and evaluate how it can benefit the MPI communication performance.The results from our numerical experiment show that these optimizations produce 6.1x speed-up compared with the initial GPU version using the same number of GPUs for the key “scatter” phase.The speed-up for the MPI communication part is 2.8x when the message size is over 3KB.Overall,we accelerate the whole GPU simulation code by a factor of 3.5x.Our findings demonstrate that particular optimizations based on the features of the simulation and modern GPU cluster are essential for achieving significantly improved performance.To the best of our knowledge,we are the first one to do the evaluation of using GPUDirect RDMA technique in this kind of Laser-Plasma-Interaction Simulation.