MPI Parallelization of PIC Simulation with Adaptive Mesh Refinement

With the prevalence of massively parallel computer architecture, MPI parallelization of existing simulation codes for stand-alone system and its parallel optimization to achieve feasible scalability are in critical need. Of many numerical approaches, adaptive mesh refinement(AMR) is known to be one of the particular cases in which MPI parallelization is challenging. In this manuscript, our ongoing project and roadmap to port the AMR-enabled Particle-in-Cell code onto distributed environments is exhibited. Our present technique is characterized by two main features. First, remote memory access is employed for inter-process data transfer in attempt to access the data with less intermediate processes. Secondly, in order to achieve better load balance, the domain decomposition according to the modified Morton ordering is introduced.