Title:A Design of Reliable Linear FSMs with Equivalent States in Stochastic Computing

Authors: Hideyuki Ichihara, Takayuki FUKUDA, Tomoo Inoue

Jounrnals:IEEE Proc. of International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems

Published Month: 10

Published Year: 2021

Type: inproceedings

Stochastic computing (SC) has attractive characteristics such as smaller area and lower power than deterministic (or general binary) computing. A SC circuit design scheme based on linear finite state machines (linear FSMs) has been proposed for realizing arithmetic functions. In this study, we discuss the transient fault tolerance of linear-FMS-based SC circuits. In general, when a linear FSM is implemented as a sequential circuit, several extra states, which are not included in the original linear FSM, are generated. The utilization of such extra states can improve the reliability of the linear-FMS-based SC circuit without any extra hardware. We present a design method, in which each extra state becomes equivalent to an original state. The proposed design method includes an extended state assignment algorithm and a rule for selecting equivalent states. Experimental results clarify that our design methodology can improve the transient fault tolerance of the linear-FSM-based SC circuits.