Optimizing Code Parallelization for Molecular Dynamics Simulations

Abstract

Plasma, the 4th matter can be simulated using fluid models, kinetic models and hybrid models. Out of these kinetic models are most accurate but simultaneously computation- ally most expensive too. Kinetic modelling involves study of plasma using molecular dynamics simulations. This method track the evolution of millions of charged particles based on its interactions. As time duration for which simulation is to be carried out and number of charged particles are huge serial code takes tremendous time to complete simulation. Multi-level and multi-tire level for simultaneous execution for varied param- eters so simulation time can be reduced extensively by parallelizing the code using these and reducing memory intensive operations. The speed performance can be evaluated by comparing simulation time of the parallel code and serial code.