大规模混合模式初始详细布局算法

以大规模混合模式布局问题为背景,提出了有效的初始详细布局算法.在大规模混合模式布局问题中,由于受到计算复杂性的限制,有效的初始布局算法显得非常重要.该算法采用网络流方法来满足行容量约束,采用线性布局策略解决单元重叠问题.同时,为解决大规模设计问题,整体上采用分治策略和简化策略,有效地控制问题的规模,以时间开销的少量增加换取线长的明显改善.实验结果表明该算法能够取得比较好的效果,平均比PAFLO算法有16%的线长改善,而CPU计算时间只有少量增加.

Better Initial Placement Algorithm for Large-Scale Mixed-Mode Detailed Placement

An algorithm is presented for better legal solution in detailed placement of large scale mixed macros and standard cells IC design.Due to the limitation of computing complexity,an effective and efficient initial placement is very important for detailed placement.Novelty of this algorithm lies in a better solution at initial stage by using network-flow method to satisfy row capacity constraint and the thought of linear placement problem(LPP) to resolve overlaps.Moreover,divide-and-conquer strategy and other simplified methods are adopted to minimize complexity.Experimental results show that the algorithm can get an average of 16% wire length improvement on PAFLO in reasonable CPU time.