Dynamic load balance of chemical source term evaluation in high-fidelity combustion simulations

TL;DR

This paper introduces a dynamic load balancing strategy using the DLB library to improve the efficiency of chemical source term evaluations in high-fidelity combustion simulations, effectively addressing load imbalance caused by stiffness and scale disparities.

Contribution

The paper presents a novel runtime load balancing approach that automatically redistributes computational resources during combustion simulations without explicit data transfer.

Findings

Achieved speedups of 2.3x for detailed chemistry and 7x for reduced chemistry on single nodes.

DLB improves performance in multi-node parallel runs similar to single-node results.

The strategy effectively mitigates load imbalance in stiff, scale-disparate combustion simulations.

Abstract

This paper presents a load balancing strategy for reaction rate evaluation and chemistry integration in reacting flow simulations. The large disparity in scales during combustion introduces stiffness in the numerical integration of the PDEs and generates load imbalance during the parallel execution. The strategy is based on the use of the DLB library to redistribute the computing resources at node level, lending additional CPU-cores to higher loaded MPI processes. This approach does not require explicit data transfer and is activated automatically at runtime. Two chemistry descriptions, detailed and reduced, are evaluated on two different configurations: laminar counterflow flame and a turbulent swirl-stabilized flame. For single-node calculations, speedups of 2.3x and 7x are obtained for the detailed and reduced chemistry, respectively. Results on multi-node runs also show that DLB improves the performance of the pure-MPI code similar to single node runs. It is shown DLB can get performance improvements in both detailed and reduced chemistry calculations.