Algorithms of Two-Level Parallelization for DSMC of Unsteady Flows in Molecular Gasdynamics

TL;DR

This paper presents a two-level parallelization scheme for direct simulation Monte Carlo of unsteady gas flows, combining independent runs and data parallelization, with load balancing strategies to improve efficiency on shared memory multiprocessors.

Contribution

It introduces a novel two-level parallelization algorithm with static and dynamic load balancing for DSMC simulations of unsteady flows.

Findings

Achieved significant speedup in unsteady flow simulations.

Identified conditions for optimal application of the algorithms.

Demonstrated effectiveness of dynamic load balancing in DSMC.

Abstract

The general scheme of two-level parallelization (TLP) for direct simulation Monte Carlo of unsteady gas flows on shared memory multiprocessor computers has been described. The high efficient algorithm of parallel independent runs is used on the first level. The data parallelization is employed for the second one. Two versions of TLP algorithm are elaborated with static and dynamic load balancing. The method of dynamic processor reallocation is used for dynamic load balancing. Two gasdynamic unsteady problems were used to study speedup and efficiency of the algorithms. The conditions of efficient application field for the algorithms have been determined.