An Optimized Cost Flow Algorithm to Spread Cells in Detailed Placement
ACM Transactions on Design Automation of Electronic Systems(2019)
摘要
Placement is an important and challenging step in VLSI physical design. The placement solution can significantly impact timing and routability. In sub-nanometric technology nodes, several restrictions have been imposed on the placement solutions. These restrictions make designing an optimized and legal solution very hard. Achieving optimized placement solutions is especially challenging in regions with high-density utilization. The quality of placement solution can significantly impact the final circuit implementation. In this work, we present a cell spreading algorithm to move cells out from high-density utilization regions. Our algorithm opens up new spaces in regions with high cell concentration. These spaces can then be exploited by detailed placement algorithms to further optimize the placement solution. The objective of our technique is to reduce area density utilization while considering cell displacement and circuit delay. The outcome of the proposed algorithm is to obtain a uniform distribution of cells in the placement area while having minimal effects on the delay. To achieve this goal, our proposed algorithm uses branch and cut, and network flow techniques. Experimental results on industrial and academic circuits illustrate that our proposed algorithm can minimize circuit delay (up to 25%), cell displacement (up to 17μ m), dynamic power consumption (up to 5.3%), and leakage power (up to 15%).
更多查看译文
关键词
EDA, branch and cut, cell spreading, network flow, placement
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络