Design And Simulation Of DyXY-Based Linear-Programming Routing Algorithm For NoC

In the NoC(Network on Chip),the routing algorithm is an important factor that affects the network performance.In the on-chip network,the communication between the IP(Integrated Protocal)need for the corresponding routing algorithm,routing algorithm has a very good impact on the performance of the on-chip network.At present,the general routing algorithms used in NoC include XY dimensional order routing algorithm,turn model routing algorithm and DyXY routing algorithm and so on.The DyXY routing algorithm providing adaptive routing based on network congestion information,and achieve the deadlock and livelock free at the same time'.DyXY routing algorithm can only use local congestion information to calculate the routing,when network congestion is serious,the network congestion situation will be further aggravated.Because the DyXY routing algorithm has some shortcomings in dealing with congestion control problems,the performance of network such as delay and throughput is poor.Therefore,this paper proposes a routing algorithm named DyXY-LP(DyXY-Based Linear-Programming).Under the condition of low load and low congestion,the DyXY routing algorithm is used to route packets.After the network congestion reaches a certain level,the network information is collectedby the main control node,and then the linear programming model is used to calculate the route according to the information,so as to avoid the increase of congestion.The main control node is similar to brain,which realizes the overall optimization and adjustment of the global traffic.In this thesis,the simulation of the routing algorithm is carried out on the network simulator Booksim2.The performance evaluation is carried out by using uniform synthetic traffic and transpose synthetic traffic.The simulation results show that in the transpose traffic distribution,DyXY-LP compared to XY and DyXY,the performance of average delay were increased by 27%and 7.6%,the throughput increased by 11.6%and 8.1%,respectively,which meet the design requirements.