lbmpy: Automatic code generation for efficient parallel lattice Boltzmann methods

Lattice Boltzmann methods are a popular mesoscopic alternative to classical computational fluid dynamics based on the macroscopic equations of continuum mechanics. Many variants lattice have been developed that vary in complexity, accuracy, and cost. Extensions available simulate multi-phase, multi-component, turbulent, non-Newtonian flows. In this work we present lbmpy, code generation package supports wide variety different methods. Additionally, lbmpy provides generic development environment for new schemes. A high-level domain-specific language allows user formulate, extend test various all cases, method can be specified symbolic form. Transformations operate representation yield highly efficient compute kernels. This is achieved by automatically parallelizing methods, application-specific automatized steps optimize resulting code. pipeline transformations applied range variants, including single- two-relaxation-time schemes, multi-relaxation-time as well more advanced cumulant entropically stabilized integrated into high-performance computing frameworks enable massively parallel, distributed simulations. demonstrated using waLBerla multiphysics conduct scaling experiments SuperMUC-NG supercomputing system up 147 456 cores.