张波, 薛正辉, 任武, 李伟明, 盛新庆. 基于图形处理器的时域有限差分算法硬件加速[J]. 电波科学学报, 2011, 26(5): 864-868.
      引用本文: 张波, 薛正辉, 任武, 李伟明, 盛新庆. 基于图形处理器的时域有限差分算法硬件加速[J]. 电波科学学报, 2011, 26(5): 864-868.
      ZHANG Bo, XUE Zheng-hui, REN Wu, LI Wei-ming, SHENG Xin-qing. Accelerating 3D FDTD algorithm using GPU computing[J]. CHINESE JOURNAL OF RADIO SCIENCE, 2011, 26(5): 864-868.
      Citation: ZHANG Bo, XUE Zheng-hui, REN Wu, LI Wei-ming, SHENG Xin-qing. Accelerating 3D FDTD algorithm using GPU computing[J]. CHINESE JOURNAL OF RADIO SCIENCE, 2011, 26(5): 864-868.

      基于图形处理器的时域有限差分算法硬件加速

      Accelerating 3D FDTD algorithm using GPU computing

      • 摘要: 为利用图形处理器(GPU)的运算能力进行时域有限差分算法(FDTD)硬件加速,针对GPU运算的特点对FDTD算法进行了优化处理,并提出一套采用"XY平面各点并行,Z向循环推进更新"等方法的计算流程,用以解决三维FDTD算法加速过程中所面临的网格划分及边界条件引入等问题。通过与同期中央处理器(CPU)运算平台比较,GPU加速运算具有良好的数值精度并可实现174倍以上的加速比,体现出优异的工程价值与应用前景。

         

        Abstract: In order to accelerate 3D finite-difference time-domain(FDTD)algorithm using graphics processing unit(GPU),with the implement of "XY plane parallel,Z direction loop update" and other techniques,a GPU FDTD programming model is proposed which is both efficient and accurate.A project is simulated using the proposed model, and the result shows good agreement with CPU computing while simulation speed is 174 times faster, thus proves the effectiveness and great potential of GPU acceleration in FDTD research.

         

      /

      返回文章
      返回