一种基于敏感度的时延驱动快速布局算法

摘要

This paper presents a sensitivity-based timing-driven fast placement algorithm, which considers the critical net delay optimization in both the equation solving stage and the cell shifting stage. A sensitivity net-based weighting model is used in e-quation solving stage and a heuristic critical path based algorithm is used in cell shifting stage.Experimental results show that worst negative slack (WNS) and total negative slack (TMS) are improved by 25.82% and 20.53% respectively using the static timing analysis tool Astro.Meanwhile, average wire length is only increased by 3.14% and the runtime is not increased significantly in comparison with FastPlace.%本文提出了一种基于敏感度的时延驱动的快速布局算法.该算法在解方程和单元扩散这两个阶段同时优化关键线网时延.其中基于敏感度的线网权重模型用于解方程阶段,基于关键路径的启发式单元优化算法用于单元扩散阶段.静态时延分析工具Astro分析结果表明:与FastPlace相比,最小负松弛量(worst negative slack)平均提高了25.82％,负松弛量总和(total negative slack)平均提高了20.53％,同时总线长仅平均增加3.14％且运行时间没有明显增加.