Hybrid adaptive routing algorithm for 2D mesh on-chip networks

The congestion in on-chip networks is a major factor that degrades the performance due to increased message latency. In this paper, we present a hybrid routing scheme based on the reinforcement learning method, Q-leaning and odd-even turn model for 2-D mesh topology. This approach restricts the locations where some turns can be taken so that deadlock is avoided and also avoids congestion by considering the latency related information stored in the routing table. This hybrid Odd even Q routing (HOEQ) approach results in better routing decision and turns out to be more reliable. Experimental results show that the proposed approach performs better for given traffic patterns.