TAU Hyperflex Report

Abstract

Computational meshes used in practical aerodynamic applications may possess various properties. For example, structured or unstructured ordering may be used to utilize efficient data structures or simplify mesh generation; high-aspect-ratio cells may be used to resolve boundary layers of high-Reynoldsnumber flows; the cell orientation may align with specific flow directions to maximize benefits of certain discretization techniques, such as approximated Riemann solvers. The HyperFlex version of the DLR TAU-code provides a suit of computational algorithms and solution techniques that can be canonically switched depending on mesh characteristics to optimize accuracy, efficiency, and robustness of flow solutions. The current HyperFlex implementation derives from experiences gained during previous efforts concerned with enabling structured data and algorithms in an unstructured code environment, developing an hierarchy of preconditioning techniques embedded into a multistage Runge-Kutta method, and implementing alternative multigrid techniques. The practical aspects of code design, such as maintainability and expendability, are also taken into account.