An Efficient Cooperation between Routing and Placement with Technology Node Constraints
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Traditionally, the placement and routing stages of the physical design are performed separately. However, because of the additional complexities arising in advanced technology nodes, they have become more interdependent. As a result, finding ways to improve the cooperation between routing and placement has become an important topic in Electronic Design Automation (EDA). In this thesis, a framework that incorporates cooperation between routing and placement with the existence of technology node constraints is proposed and developed. The core of this framework is Cooperation Routing and Placement (CRP) engine, which includes techniques to combine routing and placement. The proposed engine is tested on the ACM/IEEE International Symposium on Physical Design (ISPD) 2018 and 2019 contest benchmarks. In the proposed engine, to generate candidate placements, an Integer Linear Programming (ILP)-based Detailed Placement (ILP-DP) engine is developed, that can generate multiple high-quality legalized positions for each cell. Also, due to the complexity of routing topology with technology node constraints, a net classification technique is used to route nets according to net characteristics. Two routing algorithms including AStar and PatternRoute are used to route and estimate the cost of each cell's movement. In addition to that, since runtime is one of the main challenges in the cooperation between routing and placement, two caching techniques called Cost and Net caching are proposed. Thanks to the Cost caching technique, the global routing runtime compared with state-of-the-art improved by 28.56% on average. Also, by using the Net caching technique a parallel approach is developed that speedup 2.6 times compared to a sequential approach. Numerical results show that the proposed framework by moving 0.7% of cells can improve the detailed routing score by 0.3% on average in the presence of technology node constraints. The proposed engine also shows how only improving single objective wirelength estimation techniques like Half-Perimeter WireLength (HPWL) in the placement step can degrade the quality of the detailed routing solution. This shows the importance of using 3D routing feedback during the placement. The proposed engine can be employed as an add-on to the physical design flow between the global routing and detailed routing steps to improve the quality of the detailed routing solution.