Lazy Man’s Resynthesis For Glitching-Aware Power Minimization

This paper presents a novel resynthesis engine for minimizing the dynamic power of digital circuits. Traditional logic synthesis methods primarily focus on zero-delay toggles—logic state changes occurring between the start and end of a clock cycle. In contrast, our engine targets both zero-delay toggles and glitches, unintended transitions within a clock cycle caused by path imbalances. Glitches significantly contribute to power consumption in arithmetic circuits, making their minimization a critical challenge in electronic design. The proposed method uses a database of Pareto-optimal netlists to replace sub-networks in the target circuit with power-efficient alternatives. These replacements are guided by a simulation-driven cost function that evaluates workload-independent switching activity and penalizes gates with high fan-out. We call our approach Lazy Man’s Resynthesis because it builds on an algorithm named Lazy Man’s Synthesis, extending it from technology-independent delay optimization to post-mapping power optimization. Applied to the ISCAS and EPFL benchmarks, our method reduces glitching activity by 4.72% and dynamic power by 9.44%, achieving a 7.61% improvement over the state-of-the-art.