SAFFIRA: a Framework for Assessing the Reliability of Systolic-Array-Based DNN Accelerators

arXiv:2403.02946v1 Announce Type: cross
Abstract: Systolic array has emerged as a prominent architecture for Deep Neural Network (DNN) hardware accelerators, providing high-throughput and low-latency performance essential for deploying DNNs across diverse applications. However, when used in safety-critical applications, reliability assessment is mandatory to guarantee the correct behavior of DNN accelerators. While fault injection stands out as a well-established practical and robust method for reliability assessment, it is still a very time-consuming process. This paper addresses the time efficiency issue by …

abstract accelerators applications architecture array arxiv assessment behavior cs.ai cs.ar cs.lg deep neural network diverse diverse applications dnn dnn accelerators framework hardware latency low network neural network performance reliability safety safety-critical type