高级检索
杨明, 喻文健. 悬浮随机行走电容提取中多介质格林函数表的快速生成[J]. 计算机辅助设计与图形学学报, 2021, 33(3): 333-339. DOI: 10.3724/SP.J.1089.2021.18508
引用本文: 杨明, 喻文健. 悬浮随机行走电容提取中多介质格林函数表的快速生成[J]. 计算机辅助设计与图形学学报, 2021, 33(3): 333-339. DOI: 10.3724/SP.J.1089.2021.18508
Yang Ming, Yu Wenjian. Accelerated Generation of Multi-Dielectric Green’s Function Table[J]. Journal of Computer-Aided Design & Computer Graphics, 2021, 33(3): 333-339. DOI: 10.3724/SP.J.1089.2021.18508
Citation: Yang Ming, Yu Wenjian. Accelerated Generation of Multi-Dielectric Green’s Function Table[J]. Journal of Computer-Aided Design & Computer Graphics, 2021, 33(3): 333-339. DOI: 10.3724/SP.J.1089.2021.18508

悬浮随机行走电容提取中多介质格林函数表的快速生成

Accelerated Generation of Multi-Dielectric Green’s Function Table

  • 摘要: 为加速悬浮随机行走预刻画,利用了多介质格林函数本身的对称性,通过增加虚拟齐次纽曼边界,将有限差分空间离散区域缩减至原本的1/4,从而显著减小了差分矩阵的规模,加速了多介质格林函数表的生成.一般多介质格林函数表的对称性在理论上得到了证明.所提出的数值计算方法基于C++编程实现,其效率和正确性在一台32核服务器上通过数值实验得到了验证.实验结果表明,文中快速生成方法在达到超过8倍加速比的同时,取得了与原始有限差分方法结果完全一致的多介质格林函数表,显著减少了悬浮随机行走电容提取方法预处理的时间.

     

    Abstract: A new method for fast generation of multi-dielectric Green’s function table is proposed.The new method utilizes the symmetric property of multi-dielectric Green’s function table.By inserting virtual homogeneous Neumann boundary inside the transition cube,the original cube with FDM grid imposed is cut down to one fourth of itself,which leads to smaller FDM matrices and faster generation of Green’s function tables.The symmetry of multi-dielectric Green’s function table is proved in theory and furthermore a new method for its fast computation is proposed who is implemented in C++and whose efficiency and accuracy are validated through numerical experiments on a server with 32 cores.The experimental results show that the proposed method brings more than 8X speed-up compared to original FDM.Meanwhile,the Green function tables obtained are just as same as the original FDM.With the proposed method,pre-characterization time cost can be significantly reduced in FRW based capacitance extraction for multi-dielectric VLSI structures.

     

/

返回文章
返回