基于GPU并行计算的浅水波运动数值模拟

利用有限体积法求解描述水流运动的二维浅水方程组,模拟洪水波运动传播过程,并通过GPU并行计算技术对程序进行加速,建立了浅水运动高效模拟方法。数值模拟结果表明,基于本文提出的GPU并行策略以及通用并行计算架构(CUDA)支持,能够实现相比CPU单核心最高112倍的加速比,为利用单机实现快速洪水预测以及防灾减灾决策提供有效支撑。此外,对基于GPU并行计算的浅水模拟计算精度进行了论证,并对并行性能优化进行了分析。利用所建模型模拟了溃坝洪水在三维障碍物间的传播过程。

Numerical simulation of shallow water motion based on parallel computation using GPU

The propagation of flood waves was simulated by solving the two-dimensional shallow water equations for flow motion using finite volume method.GPU parallel computing was adopted for acceleration and thus high efficiency methods for shallow water simulation were developed.Numerical simulation results show that with the GPU parallelization strategy described in this paper, as well as the support of CUDA (Compute Unified Device Architecture), a computing acceleration ratio as high as 112 against single CPU was achieved.This offers a potential of using a single desktop computer for real-time simulation of flooding and support the rapid decision for disaster prevention and mitigation.Besides, the accuracy of GPU parallel computing in shallow water simulation was also analysed, as well as the performance optimization.A simulation was carried out for dam-break flow among three-dimensional obstacles.