Mitigation of Single-and Multiple-Cycle-Duration SETs using Double-Mode-Redundancy (DMR) in Time

TMR-in-hardware is a traditional architectural solution to radiation hardening. Simple in concept, TMR-in-hardware incurs at least 3times power and area penalties over the unhardened singlet computation. In this paper we focus upon compute-intensive pipelined computations and present an efficient alternative - dual mode redundancy (DMR). DMR-in-time utilizes dual computation threads to detect an error. To correct an error DMR-in-time utilizes relatively small input and output queues for rolling back the computation to a known correct state. DMR-in-time incurs only a 2times decrease in throughput, requires no downtime for error correction, and has power requirements only marginally greater than the unhardened singlet computation