A task-based parallel framework for ensemble simulations of rocket ignition: Verification and performance assessment*

We have previously reported an integrated computational framework for exascale-oriented ensemble simulations of laser-induced ignition in a methane-oxygen rocket combustor (Maeda et al, arXiv. 2022). In this study, we perform continued performance verification and assessment of Hypersonic Task-based Research Solver (HTR) (Di Renzo et al., Comp. Phys. Comm. 2020), the reacting flow solver in the framework for high-fidelity simulations of turbulent combustion. HTR solves the compressible, multi-species Navier-Stokes equations with finite-rate chemistry on curvilinear grids, employing the task-based programming model built on the Legion runtime system for scalable and portable simulation on supercomputers with heterogeneous architectures. The accuracy and performance of the solver are analyzed using canonical reacting flow problems which  are relevant to laser-induced ignition and combustion. High-order shock capturing schemes and sub-grid scale models are tested on various curvilinear meshes. Various strategies of task-mapping on GPUs and CPUs are compared. Lastly, we discuss the influence of the choice of numerical schemes and models on the accuracy of the prediction of ignition success when applied to ensemble simulations of a three-dimensional model combustor.