Variation-aware optimisation for reconfigurable cyber-physical systems

Abstract

Cyber-Physical Systems are present in many industries such as aerospace, automotive, health-care and transportation, and over time they have become critical and require high levels of resiliency and fault tolerance. Often they are implemented on reconfigurable logic due to IP design reutilisation, highperformance, and low-cost. Nevertheless, the continuous technology shrinking and the increasing demand for systems that operate under different power profiles with high-performance has led to implementations operating below the maximum performance offered by a particular technology. Design tools are conservative in the estimation of the maximum performance that can be achieved by a design when placed on a device, accounting for any variability in the fabrication process of the device. This work takes a new view on the performance improvement of circuit designs by pushing them into the error prone regime, as defined by the synthesis tools, and by investigating methodologies that reduce the impact of timing errors at the output of the system. In this work two novel error reduction techniques are proposed to address this problem. One is based on reduced-precision redundancy and the other on an error optimisation framework that uses information from a prior characterisation of the device. Both of these methods allow to achieve graceful degradation in performance whilst variation increases.