Large-Scale Simulations of Turbulent Flows using Lattice Boltzmann Methods on Heterogeneous High Performance Computers

Abstract

Current GPU-accelerated supercomputers promise to enable large-scale simulations of turbulent flows. Lattice Boltzmann Methods (LBM) are particularly well-suited to fulfilling this promise due to their intrinsic compatibility with highly parallel execution on both SIMD CPUs and GPUs. A novel LBM scheme for wall-modeled LES in complex geometries is described with a special focus on the efficient implementation in the open source LBM framework OpenLB. Detailed scalability results are provided for all HoreKa partitions, utilizing up to 128 nodes and covering problem sizes up to 18 billion cells.