Locating Critical-Reliability Gates for Sequential Circuits based on the Time Window Graph Model

Abstract

The dependability of systems gains extensive attention as they are vulnerable to unexpected events, such as soft errors in circuits. Many works proposed to enhance the reliability of circuits at a low cost by hardening the reliability-critical gates (RCGs). However, it is challenging and necessary to determine the gates that need to be improved, especially for sequential circuits. Currently, methods used to locate RCGs for sequential circuits are struggling to balance the accuracy and computational overhead. This paper presents a method for locating RCGs of sequential circuits, the states of gates are aggregated into slices based on the time window graph model, then a probability-based calculation method is proposed for quantifying the criticality of gate reliability oriented to slices, finally, locating RCGs for sequential circuits which takes the accumulation of faults into consideration. Experiments on fifty ISCAS-89 and six ITC-99 benchmark circuits show that, on average, the accuracy of our method is 0.9417 of that of Monte Carlo (MC) method, and the speed and memory cost are 1725 times faster and 5.57 times lower than those of MC method. In addition, the proposed method is much faster and more accurate than stochastic method for large-scale circuits.