This paper presents the design and analysis of on-chip interconnect architectures for real time Multimedia Systems-on-Chip (MSoC) targeting Internet of Things (IoT) applications. The interconnect architecture provides high flexibility in connection for hardware implementation of reconfigurable neural network. Due to technology’s miniaturization in ultra-deep submicron technology, the on-chip interconnect performance and power consumption become a bottle-neck. In this paper, the hybrid optimization technique is proposed to address these challenges using schmitt trigger as a repeater and tapering. Here, the proposed optimization technique is incorporated with a dedicated point to point based interconnection (PTP-BI) configuration. A comparative study with others without optimization technique (Model–I) shows the effectiveness of the proposed optimization technique (Model–II). The technology node scaling impacts are also analyzed for both techniques. Finally, the percentage reduction of latency and power consumption are evaluated in two different cases to observe the impacts of varying the interconnect length.
Comparative Study of Hybrid Optimizations Technique for On-Chip Interconnect in Multimedia SoCs
