Research On Mapping Algorithm Based On Swarm Intelligence In Network On Chip

With more and more IP (Intellectual Property) cores integrated on a single chip, the existing shared bus based System-on-Chip (SoC) faces enormous challenges in several aspects, such as latency, throughput and scalability. As an effective technology of the multi-core interconnection, Network-on-Chip (NoC) becomes the focus in current research. Mapping optimization is a key step in the NoC design. The main task is to map IP cores onto nodes in the network efficiently in order that the network has better performance in terms of energy consumption, latency and throughput. The performance of mapping algorithm decides the result of mapping and it has a critical influence on the performance of practical system finally.In this paper, the significant issues in NoC design are summarized firstly, and then we give the development of NoC and the flow of mapping optimization in current research. In addition, the swarm intelligence algorithms and the developments of the mapping algorithm in current research are summarized, respectively. On one hand, we propose bee colony intelligence based mapping algorithm according to the mapping from IP cores onto regular NoC platform. The proposed algorithm translates the low energy mapping problem into the single objective mathematics optimization model by constructing the model of energy consumption and mapping constraint. The algorithm achieves low energy consumption mapping results by simulating the different operations of bees. On the other hand, a fish swarm intelligence based multi-objective mapping algorithm is proposed to solve the low energy consumption and traffic balance mapping model. The multi-objective mapping model is constructed by considering the energy consumption and hotspot problem due to the traffic aggregation in the center of the mesh based network, which affects the performance of system seriously. In the meantime, the fish swarm intelligence based high performance mapping algorithm is proposed to finish the mapping from the IP core to the node in the network. Finally, simulations are run with the software, MATLAB, and the communication graphs are used as the basic input parameters which are popular in the application specific NoC simulation. A series of simulations using the application specific mapping algorithm show that the proposed algorithms have better performance in energy consumption and traffic balance than the current algorithms.