高级检索
骆祖莹, 赵国兴, 周金和. ECO布局中的电源线/地线网络局部SOR分析方法[J]. 计算机辅助设计与图形学学报, 2010, 22(6): 921-926.
引用本文: 骆祖莹, 赵国兴, 周金和. ECO布局中的电源线/地线网络局部SOR分析方法[J]. 计算机辅助设计与图形学学报, 2010, 22(6): 921-926.
Luo Zuying, Zhao Guoxing, Zhou Jinhe. Localized SOR Analysis Method for Power/Ground Networks in ECO Placements[J]. Journal of Computer-Aided Design & Computer Graphics, 2010, 22(6): 921-926.
Citation: Luo Zuying, Zhao Guoxing, Zhou Jinhe. Localized SOR Analysis Method for Power/Ground Networks in ECO Placements[J]. Journal of Computer-Aided Design & Computer Graphics, 2010, 22(6): 921-926.

ECO布局中的电源线/地线网络局部SOR分析方法

Localized SOR Analysis Method for Power/Ground Networks in ECO Placements

  • 摘要: 供电电压直接决定芯片性能, 在IC设计的各个阶段考虑供电电压约束具有重要的意义.受制于电源线/地线 (P/G) 网络分析的高复杂性, 尽管供电电压已成为布图规划设计中的一个设计约束, 但目前在布局设计中还未考虑供电电压约束.有别于ICCG, SOR等经典的全局分析算法, 提出了一种局部的连续过松弛方法 (SORPECO) , 并在ECO布局过程中对P/G网电压约束进行高效的分析.基于前一个布局的P/G网电压分布, 针对ECO试探布局中某些轻微设计变动, SORPECO只需对这些设计变动的局部变化周边区域进行松弛, 以更新P/G网电压分布.受益于P/G网络分析的局部性, SORPECO拥有局部、高效和高精度等优点.实验结果表明, 与通常用于布图规划的传统高效的ICCG算法相比, SORPECO不仅精度损耗几乎可以忽略 (最大误差<0.062%) , 而且可以加速2个数量级.

     

    Abstract: Since supply voltage directly determines the chip performance, it is important to take the supply voltage as the design constraint at all IC design stage.Owing to the mammoth complexity of power /ground (P/G) analysis, the supply voltage has not been considered in the placement though it became a constraint of the floor-planning.Different from the traditional global analysis algorithms such as ICCG and SOR, localized successive over-relaxation method (SORPECO) is proposed in the paper.SORPECO is the first try to efficiently analyze P/G voltage distribution in ECO placements.Based on the P/G voltage distribution of previous placement solution, SORPECO can obtain a new P/G voltage distribution only through relaxing small local area surrounding some slight changes in a tentative solution.Thanks to the locality of P G network analysis, SORPECO is of advantages such as locality, efficiency, and high accuracy.In comparison with the traditional efficient ICCG algorithm usually used in floor-planning, experimental results show that SORPECO is two orders of magnitudes faster with ignored accuracy loss (maximum errors <0.062%) .

     

/

返回文章
返回