Reinforcement Learning Enabled Routing for High-Performance Networks-on-Chip

Md Farhadur Reza, Tung Thanh Le
IEEE ISCAS, May 2021, Daegu, South Korea.
(Equal Contribution)
Publication

For citation, we encourage you to cite our work if you used our code or mentioned our paper work. Thank you!

M. F. Reza and T. T. Le, "Reinforcement Learning Enabled Routing for High-Performance Networks-on-Chip," 2021 IEEE International Symposium on Circuits and Systems (ISCAS), 2021, pp. 1-5, doi: 10.1109/ISCAS51556.2021.9401790.

Abstract

With the increase in cores in the multi-core architectures, the probability of congestion increases because of longer paths among sources and destinations in the network-on-chip (NoC) and because of running multiple applications in a chip. Reactive detection and/or single fixed routing algorithm are not effective to prevent congestion from happening for different traffic patterns in NoC. Therefore, we propose reinforcement learning based proactive routing technique that selects the best routing algorithm from multiple available routing algorithms using NoC utilization and congestion information to improve communication performance. Simulation results demonstrate latency performance improvement while providing robust NoC performance for different NoC states.

How it works

From the terminal:

• Run all

   ./run_all.sh
  

• Run individual algorithm

   python interconnect-routing-gym/example/rl_sarsa_example.py
  

Code

The codes are located in the example folder.

Outputs

The results will be stored in the Inter_Connect_Networks folder.

References

1. Reinforcement Learning based Interconnection Routing for Adaptive Traffic Optimization.
   Note that we were inspired by this paper's implementation.
   

2. Garnet2.0: An On-Chip Network Model for Heterogeneous SoCs