Research Of NoC Mapping Problem Based On Particle Swarm Optimization Algorithm

For more than half a century, integrated circuit (IC) manufacturing technology has been developing quickly by following Moore's law. But there is an increasingly deeper gap between manufacture capacity and design capacity. To narrow this scissors difference, there will be an important revolution in design technology every ten years. Thus, Soc(System on Chip) technology was proposed at the beginning of 1990s. But with the further development in techniques, SoC based on the shared bus network confronts with many problems such as interference, low reusability and low expansibility. Therefore, a new architecture NoC (Network on Chip) was proposed around 2000, and it breaks the bottlenecks that SoC faces.The research of NoC foucses on network topology, routing, switching technology, QoS, flow control, resource-network interface, performance evaluation, mapping ect. This paper takes NoC mapping problem as research subject, which is currently one of the hot spots at home and broad. Its mapping result will have a direct impact on the power consumption, latency and other performances of the system. The research on this topic is of theoretical and practical significance.This paper firstly introduces the related concepts of NoC, and then presents the single-objective optimization model of communication power consumption and latency respectively and the multi-objective optimization model including these two parts. Then by improving the discrete particle swarm optimization(PSO) algorithms proposed in the literatures, we try to optimize these models and find a best solution. Simulation results show that we can obtain more better optimal results.