Evaluating selective redundancy in data-flow software-based techniques

Abstract

This paper presents an analysis of the efficiency of using selective redundancy applied to registers in software-based techniques. The proposed selective redundancy chooses a set of allocated registers to be duplicated in software in order to provide detection of upsets that occur in the processor hardware and provokes data-flow errors. The selective redundancy is implemented over miniMIPS microprocessor software. A fault injection campaign is performed by injecting single event effect upsets in the miniMIPS hardware. Results show error detection capability, performance degradation and program memory footprint for many case studies. With that, designers can find the best trade-off in using selective redundancy in software.