DSMC Performance Frontiers: Interplay of Physics and Computation

The Direct Simulation Monte Carlo (DSMC) method is a widely accepted technique for simulating rarefied gas flows, renowned for its high accuracy in capturing non-equilibrium phenomena. However, its significant computational costs often hinder its application to complex problems. This study aims to comprehensively evaluate the computational demands and performance of the DSMC method across various scenarios, with a focus on optimization strategies to improve its efficiency.

We investigate the multi-threading performance, revealing a non-linear dependence of simulation time on processor count. Our results show that increasing processor count reduces simulation time up to a threshold, beyond which a bottleneck effect emerges due to inter-thread communication overhead, limiting the efficiency of multi-threaded DSMC simulations. To further accelerate the DSMC method, we also explore the utilization of Graphical Processing Units (GPUs).

Furthermore, we examine the impact of temperature, Knudsen number, and the corresponding Maxwell-Boltzmann distribution on the performance of DSMC method. This systematic analysis provides valuable insights into optimizing DSMC simulations for diverse physical conditions, including low-speed and high-speed flows, and sheds light on the complex interplay between computational efficiency and physical accuracy.