Many-Core Real-Time Network-on-Chip I/O Systems for Reducing Contention and Enhancing Predictability

Jiang, Zhe and Dai, Xiaotian and Zhao, Shuai and Wei, Ran and Gray, Ian (2023) Many-Core Real-Time Network-on-Chip I/O Systems for Reducing Contention and Enhancing Predictability. In: CPS-IoT Week '23: Proceedings of Cyber-Physical Systems and Internet of Things Week 2023eek 2023 :. ACM, New York, pp. 227-233. ISBN 9798400700491

Abstract

In safety-critical and high-integrity computing, it is important to guarantee both performance and time-predictability of I/O operations. However, with the continued growth of hardware and architectural complexity, satisfying such real-time requirements has become increasingly challenging because of complex I/O transaction paths and extensive hardware contention. This paper proposes a systematic framework with a novel I/O controller and a reconfigurable NoC, effectively optimising I/O transaction paths to encounter reduced contention. Moreover, we present a theoretical model and optimisation process to further improve real-time performance. The evaluations show that our approach outperforms state-of-the-art I/O processing techniques on a range of metrics.