Abstract: Today, a potential alternative known as network-on-chip (NoC) is being used in multicore systems to circumvent the limitations of traditional on-chip networks. NoC designs enable the use of high-capacity wireless connections to significantly decrease the latency associated with multi-hop communications. This article discusses a new technique for increasing the performance and power usage of NoCs. The suggested method makes use of a customized version of the RED algorithm's queue length concerns. A constantly increasing set of cores on the device need a scalable architecture; this is the most essential option. Mesh-based NoCs are the most commonly available topologies in many-core processors today as a scalable alternative to the conventional shared bus. To ensure long-term scalability and performance, the low-latency connection between the cores becomes more important. The delay between endpoints in an ideal network is roughly equal to that of a single cycle. The suggested method was tested using a variety of simulated traffic patterns and the SPLASH-2 trace-driven benchmark suite. The testing findings show that the algorithm significantly lowers latency and power usage when compared to a traditional NoC.
Keywords: Routing algorithm, Power efficiency, Latency, Topology, Interconnection Network, Network-on-chip
|
DOI:
10.17148/IJARCCE.2021.101239
