An evolutionary search for energy recovery opportunities in partially reversible FCN circuits

Power dissipation is more important than ever when designing digital systems. Moreover, evaluating emerging technologies’ boundaries is essential as conventional technologies approach their energy limits. One way to push those boundaries forward, especially regarding energy loss reduction, is by applying post-synthesis methods to look for energy recovery opportunities. In this work, we propose a Genetic Algorithm (GA) based search to identify exploitable fan-outs and use them to significantly reduce energy losses in partially reversible Field-Coupled Nanocomputing (FCN) circuits when dealing with circuit depth (delay) degradation. We found a new spectrum of circuit configurations and analyzed them regarding the tradeoff between these two metrics: energy, and depth. The results show an average increase of 4× in the number of points within the Pareto frontier. Based on an in-depth analysis of information loss circumstances in FCN circuits, local changes in the circuit netlist can reduce fundamental energy limits without any resynthesis process.