Accelerating solid-fluid interaction using Lattice-Boltzmann and Immersed Boundary coupled simulations on heterogeneous platforms

摘要

We propose a numerical approach based on the Lattice-Boltzmann (LBM) and Immersed Boundary (IB) methods to tackle the problem of the interaction of solids with an incompressible fluid flow. The proposed method uses a Cartesian uniform grid that incorporates both the fluid and the solid domain. This is a very optimum and novel method to solve this problem and is a growing research topic in Computational Fluid Dynamics. We explain in detail the parallelization of the whole method on both GPUs and an heterogeneous GPU-Multicore platform and describe different optimizations, focusing on memory management and CPU-GPU communication. Our performance evaluation consists of a series of numerical experiments that simulate situations of industrial and research interest. Based on these tests, we have shown that the baseline LBM implementation achieves satisfactory results on GPUs. Unfortunately, when coupling LBM and IB methods on GPUs, the overheads of IB degrade the overall performance. As an alternative we have explored an heterogeneous implementation that is able to hide such overheads and allows us to exploit both Multicore and GPU resources in a cooperative way.