A NOVEL COMMUNICATION-INFRASTRUCTURE-BASED DYNAMIC ROUTING ALGORITHM FOR NETWORK-ON-CHIP TO IMPROVE LOCAL AND GLOBAL CONGESTION AWARENESS

Abstract

Network-on-Chip (NoC) architectures integrate all essential components of a computing system onto a single chip, enabling efficient communication among various processing units. These architectures typically encompass processors, memory modules, and input/output interfaces, facilitating seamless interaction among heterogeneous components. NoCs can incorporate a diverse range of functional units, including central processing units (CPUs), graphics processing units (GPUs), Bluetooth, Wi-Fi, and other peripheral modules, thereby enhancing system scalability and performance. The efficiency of data transmission in NoCs heavily depends on routing algorithms, which play a crucial role in managing communication latency, congestion, and overall network reliability. Routing algorithms play a fundamental role in managing data flow across various components within a Network-on-Chip (NoC) architecture, ensuring efficient and reliable communication. These algorithms are responsible for determining optimal data transmission paths while addressing key challenges such as congestion, latency, and resource utilization. Given the high density of interconnections in NoC-based systems, an effective routing strategy is essential to minimize communication interference, reduce power consumption, and enhance overall data transfer efficiency. By optimizing these factors, routing algorithms contribute significantly to improving system performance, increasing energy efficiency, and ensuring seamless on-chip communication. These algorithms play a pivotal role in enhancing the overall system performance and optimizing the efficiency of various on-chip components. In Network-on-Chip (NoC) architectures, incorporating global congestion awareness into the routing decision-making process is crucial for selecting the most efficient path to the destination. In this study, a novel communication subsystem is introduced to facilitate the acquisition of global congestion information, alongside the development of an intelligent routing algorithm that utilizes both local and global congestion data to optimize path selection. The proposed approach aims to improve system performance by enhancing throughput and reducing latency. Simulation and evaluation results obtained using the Noxim simulator demonstrate the superior efficiency of the proposed method compared to existing routing algorithms, highlighting its effectiveness in optimizing data transmission and network reliability.