Optimization of the Direct Simulation Monte Carlo code (DSMC3D) for rarefied gas flow simulations

The Direct Simulation Monte Carlo DSMC method has become the method of choice for modeling rarefied gas dynamics in a variety of scenarios after its introduction 50 years ago. Several parallel DSMC codes have been developed over the last 30 years to leverage the increasing computational power of parallel machines. In this work we are going to focus on the DSMC3D code developed at Eindhoven University of Technology. The code is based on the message passing interface paradigm (MPI) using a cartesian domain decomposition. However, at large scales the code presents some load balancing issues that impact the performance. The main objective of the project is to introduce another level of parallelization using a shared memory approach with OpenMP that allows us to reduce the number of MPI’s processes.