GPU implementation for high-order numerical flow simulation

Modern Graphics Processing Units (GPUs) with many-core architectures have emerged asgeneral-purpose parallel computing platforms that can accelerate simulation applicationstremendously. In this study, GPU is implemented in order to accelerate flow simulation with ahigh-order method. A two-dimensional inviscid Compressible flow and two-dimensionalincompressible viscous flow are chosen as the test cases. The Beam and Warming linearizationscheme is used to discretize the Euler equation for the compressible flow and stream functionvorticityformulation for incompressible flow simulation. For both test cases the fourth-ordercompact finite-difference scheme is applied to spatial discretization. Moreover, a highperformanceGPU-implemented block-tridiagonal solver based on Block Cyclic Reduction(BCR) algorithm is utilized. The efficiency of GPU for high-order flow simulation is investigatedusing NVIDIA GTX480 graphics card. Speedups between 2.5x–4.5x are achieved in comparisonto the Intel Core 2 Duo E7400 2.8GHz CPU for different grid sizes.