Estimating Pareto Optimum Fronts to Determine Knob Settings in Electronic Design Automation Tools

Billy Huggins, W. Rhett Davis, Paul Franzon
North Carolina State University


The ability to configure physical design tools is often dependent on the experience and knowledge of the physical designer (PD). Technology node sizes are ever decreasing, digital design sizes vary drastically, and design constraints change based on the needs of the application. As these changes occur frequently and physical design times can be extensive, the need for accurate quality of design results early in the design process is crucial. Collecting these metrics is computationally expensive, creating a need to determine how to best create and extract information as design flows change. This paper describes the use of estimated Pareto optimal trade-off sets to provide designers with the capability of visualizing the results of Electronic Design Automation (EDA) tool configuration settings, or “knobs”, that will offer an optimal post detail route design based on two design metrics, critical path length and core area. We will show that when given a set of design constraints create a point that occurs along the Pareto front, the knob settings used are optimal. With only 38 samples per design, we were able to produce estimated routed design metrics with a worst case error (WCE) of less than 10%.