Reliability Evaluation of Approximate Arithmetic Circuits Based on Signal Probability

Recently, approximate arithmetic circuits (AACs) have been widely used in applications with inherent error tolerance. To guarantee the usability of AACs in the approximation applications, it is necessary to analyze AAC reliability. In this paper, we propose two accurate AAC reliability evaluation methods based on signal reliability analysis, where subtraction correlation coefficient and division correlation coefficient are applied respectively to solve the correlation problem caused by fanout reconvergence. The accuracy and scalability of these two methods are verified by randomly selected circuits from the EvoApprox8b library. The experimental results show that, compared with Monte Carlo (MC) simulation, our two proposed methods have average error rates of 0.34% and 0.65% and the run-time costs are 0.012% and 0.019% of MC’s simulation time. Moreover, the proposed methods are superior to two recently reliability analysis based methods in terms of efficiency and accuracy.