The Mapping Methodology With Network Coding For NoC Applications

Wireless Network-on-Chip(NoC) is a data interactive infrastructure by which multi-processing units can cooperate with each other to provide high-speed, high-bandwidth and Qo S guaranteed inter-cores connection. The primary step of fulfilling applications on wireless NoC is mapping the application tasks to improve the performance of wireless NoC. Due to the natural broadcasting feature of wireless NoC, network coding can be adopted and significantly reduce the network latency, improve the network throughput and reduce the extra communication overhead in the case of network congestion. Therefore, wireless NoC can achieve a better performance with network coding after task mapping.This thesis mainly studies effective task mapping and network coding mapping to achieve optimal performance on wireless NoC. The main contributions are listed as follows:(1) A task mapping algorithm is proposed for three optimization objectives: the latency, power consumption and heat balance. The algorithm is an improved NSGA-Ⅱ algorithm that can optimize latency, power consumption and heat balance simultaneously, while effectively preventing premature convergence phenomenon in the genetic process. It is a feasible method for three-objective optimization for task mapping on wireless No C.(2) A regional network coding model is proposed for alleviating the congestion of multiple hotspots. The model divides the entire wireless NoC into multiple non-overlapping sub-regions, and then it constructs network coding independently in each sub-regions to optimize network performance. A wireless NoC architecture based on distributed wireless butterfly network coding is also be put forward. This architecture can reduce latency and the number of hotspots on wireless NoC by multiple disjoint butterfly network coding regions. It is an efficient and parallel multi-core co-processing platform for complex systems.(3) This thesis proposes a minimum hotspots algorithm for the optimized design with distributed butterfly network coding on wireless NoC. It maximizes the network coding gain by maximizing the number of butterfly network coding.(4) Simulation enviroment is built with Matlab for the mapping on wireless No C. Simulation results show that the proposed three-objective optimization task mapping algorithm can further optimize the performance than the existed three-objective optimization task mapping algorithm, and can get a better compromise performance in the latency, power consumption and heat balance relative to the two-objective optimization task mapping algorithm. Simulation results also show that the minimum hotspot algorithm with the distributed butterfly network coding can improve the transmission performance, effectively reduce the number of hotspots in the network, and release congestions of the network.