Machine Learning Based Parameter Tuning for Performance and Power optimization of Multisource Clock Tree Synthesis

Abstract

Many schemes for driving synchronous digital circuits have been devised over the years. The Clocking Architecture that is considered for evaluation is H – Tree based clock distribution (MSCTS (Multisource Clock Tree Synthesis). All the Combinations of the parameters that effect the Timing and some additional parameters that may improve timing cannot be tested in a short amount of time using EDA physical implementation tool. To improve Timing performance, TUNA (Machine learning based tool) is used. TUNA is a machine learning based tool, allowing a tool to explore all possibilities within the predefined user inputs, and achieve best QoR. Taking Timing and Power as priority, Experiments are performed for best QoR results. The comparative analysis of conventical physical Implementation and ML (Machine Learning based approached TUNA point) is performed. This paper summarizes the improvements that have been made in Multisource CTS using ML based approach for better QoR.