An asynchronous discontinuous Galerkin method for combustion simulations

Abstract

The discontinuous Galerkin (DG) method has been widely considered in recent years to develop scalable flow solvers for its ability to handle discontinuities, such as shocks and detonations, with greater accuracy and high arithmetic intensity. However, its scalability is severely affected by communication bottlenecks that arise from data movement and synchronization at extreme scales. Recently, an asynchronous discontinuous Galerkin (ADG) method was proposed to reduce communication overhead by either relaxing synchronization or avoiding communication between processing elements (PEs). The numerical properties of the ADG method were verified by solving simple one-dimensional partial differential equations. In this study, the application of the ADG method is extended to complex chemically reacting flows, particularly to evaluate its efficacy in capturing flame fronts and detonations. New asynchrony-tolerant weighted essentially non-oscillatory (AT-WENO) limiters are derived to accurately capture flow discontinuities with communication delays at the PE boundaries. The numerical accuracy of the ADG method is verified on the premixed spontaneous ignition of a fuel-lean mixture with detailed chemistry. The progression of the flame front is accurately represented by the ADG method, and the numerical errors incurred at the PE boundaries turn out to be insignificant. The efficacy of the AT-WENO limiters in capturing discontinuities is demonstrated by the propagation of a detonation wave. The findings of this study serve as a basis for the development of highly scalable discontinuous Galerkin-based solvers for combustion simulations on massively parallel supercomputers.