【摘要】 随着图形处理器(GPU)性能的大幅度提高以及可编程特性的发展,将通用数值算法的某些处理阶段从CPU迁移到GPU上已成为可能,从而达到加速计算的目的。首先简要介绍了一种常见的数值计算方法:交变方向隐式时域有限差分法(ADI-FDTD);然后详细论述了利用GPU加速ADI-FDTD计算的基本原理与关键技术,并给出了在GPU上求解ADI-FDTD线性方程组的共轭梯度法实现框架;最后,通过具体的计算实例和相关的性能比较验证了这种加速算法的精确性与效率特点。更多还原

【Abstract】 With the advancement of Graphics Processing Unit(GPU) and the creation of its new feature of programmability,it has come possible to transfer some of the processing stages in general numerical algorithms from CPU to GPU in order to accelerate the computation.In this paper,starting from a brief introduction to Alternative Direction Implicit Finite Difference Time Domain(ADI-FDTD) algorithm,detailed introduction and analysis were given to the fundamentals and the key technique of GPU for accelerating ADI-FDTD computation,in combination with the implementation frame of the conjugate gradient method for solving linear equations system on GPU.Finally,some computed examples were presented,and various comparisons were made to prove the efficiency and accuracy of this acceleration approach.更多还原