A Parallel Current Deposition Method For Pic Simulation On Gpu

This paper offers a GPU-based implementation of the Particle-In-Cell (PIC) algorithm for dense plasma simulations. In particular we focus on fast algorithms for current depositing and proposed a parallel segmented reduction method on GPU. The presented method makes full use of the advantages of zigzag scheme and can greatly avoid writing-conflict caused by depositing currents. We obtained performance increases of about 4x over thread racing implementations in our experiments. Our method is expected to get higher speedups in denser plasma simulations due to its high degree of parallelism.