Learning-Based Routing Algorithms for On-Chip Networks

摘要: In this chapter, we investigate highly adaptive routing algorithms for balancing the traffic over network based on learning approaches. The proposed methods aim to provide up-to-dated local and global congestion information at each switch. At first, method is applied a utilizing minimal routing. low loads, can achieve optimal performance, while they are inefficient in avoiding hotspots when load increases. reason of inefficiency that propagate messages through most two directions When shortest paths congested, sending more them deteriorate condition considerably. order address issue, present non-minimal algorithm on-chip networks provides wide range alternative between pair source destination switches. Initially, determines all permitted turns including 180-degree single channel without creating cycles. implementation best usage allowable route adaptively network. On top that, selecting less congested path, an optimized scalable utilized. estimate latency from output region.