基于GPU的多重网格Navier-Stokes解算器并行优化方法研究

Implementation and Optimization of a Navier-Stokes Solver with Multigrid Methodology Based on GPU

随着工业计算需求的激增,计算流体力学(Computational Fluid Dynamics,CFD)学科对计算效率问题越来越重视。作者基于自行开发的Navier-Stokes解算器,引入多重网格加速收敛算法,并结合NVIDIA GPU计算平台,从数值方法和高性能计算两个方面为CFD实现加速。数值加速算例测试结果表明,基于多重网格算法的GPU解算器相对CPU版本代码双精度可获得45倍以上的加速。

With rapid growth of the computing demands in industry, the computational efficiency gets more and more attention in the field of computational fluid dynamics（CFD）. In this paper, we present the parallelization and acceleration techniques based on NVIDIA GPU for CFD, using our self-developed Euler solver with multigrid methodology（MGM）. Numerical experiments results showed that for aerodynamics, the GPU accelerated Euler Solver with MGM achieved more than 45 times speedup over the CPU code in double precision computation.