Parallelization of an implicit algorithm for multi-dimensional particle-in-cell simulations

摘要

The implicit 2D3V particle-in-cell (PIC) code developed to study theinteraction of ultrashort pulse lasers with matter [G. M. Petrov and J. Davis,Computer Phys. Comm. 179, 868 (2008); Phys. Plasmas 18, 073102 (2011)] has beenparallelized using MPI (Message Passing Interface). Details on the algorithmimplementation are given with emphasis on code optimization by overlappingcomputations with communications. Performance evaluation has been made on asmall Linux cluster with 32 processors for two typical regimes of PICoperation: "particle dominated", for which the bulk of the computation time isspent on pushing particles, and "field dominated", for which computing thefields is prevalent. We found that the MPI implementation of the code offers asignificant numerical speedup. In the "particle dominated" regime it is closeto the maximum theoretical one, while in the other regime it is about 75-80 %of the maximum speed-up. The code parallelization will allow futureimplementation of atomic physics and extension to three dimensions.