Adaptive Verifiability-Driven Strategy for Evolutionary Approximation of Arithmetic Circuits
This work addresses the need for automated, provably-correct circuit approximations in energy-aware computing, representing a novel method for a known bottleneck.
The paper tackles the problem of designing approximate arithmetic circuits for energy-aware applications by integrating formal methods into an evolutionary algorithm, resulting in a strategy that provides superior scalability and versatility across various scenarios.
We present a novel approach for designing complex approximate arithmetic circuits that trade correctness for power consumption and play important role in many energy-aware applications. Our approach integrates in a unique way formal methods providing formal guarantees on the approximation error into an evolutionary circuit optimisation algorithm. The key idea is to employ a novel adaptive search strategy that drives the evolution towards promptly verifiable approximate circuits. As demonstrated in an extensive experimental evaluation including several structurally different arithmetic circuits and target precisions, the search strategy provides superior scalability and versatility with respect to various approximation scenarios. Our approach significantly improves capabilities of the existing methods and paves a way towards an automated design process of provably-correct circuit approximations.