SciCADE 2013
International Conference on Scientific Computation and Differential Equations
September 16-20, 2013, Valladolid (Spain)

Invited Talk

Show full list of Invited Contributions to Minisymposia Show talk context (MS02)


Real world CFD applications on GPGPU based LBM with local grid-refinement

M. Geier, M. Schönherr and M. Krafczyk

Abstract
In order for the lattice Boltzmann method to be applicable to real world CFD applications the code needs to be flexible and efficient at the same time. It has to support off-grid boundary conditions to fit arbitrary geometrical constrains with a Cartesian mesh and local grid-refinement is necessary to enhance the resolution in areas of higher gradients. It is essential that grid-refinement and off-grid boundaries work seamlessly and efficiently together. In this contribution we present an LBM implementation for GPGPUs based on the sparse EsoTwist data structure. EsoTwist combines streaming and collision into a single step and requires only one set of variables. It allows for a random access order to the lattice notes and is hence completely thread-safe and suitable for many-core implementations. Grid-refinement and boundary conditions are designed to be as compact as possible in order to work efficiently with the EsoTwist data structures without compromising accuracy.

Organized by         Universidad de Valladolid     IMUVA