Silent Data Corruption Estimation and Mitigation Without Fault Injection Estimation et attenuation de la corruption silencieuse des donnees sans injection de faute

Silent data corruptions (SDCs) have been always regarded as the serious effect of radiation-induced faults. Traditional solutions based on redundancies are very expensive in terms of chip area, energy consumption, and performance. Consequently, providing low-cost and efficient approaches to cope with SDCs has received researchers' attention more than ever. On the other hand, identifying SDC-prone data and instruction in a program is a very challenging issue, as it requires time-consuming fault injection processes into different parts of a program. In this article, we present a cost-efficient approach to detecting and mitigating the rate of SDCs in the whole program with the presence of multibit faults without a fault injection process. This approach uses a combination of machine learning and a metaheuristic algorithm that predicts the SDC event rate of each instruction. The evaluation results show that the proposed approach provides a high level of detection accuracy of 99% while offering a low-performance overhead of 58%.