NoC Mapping Based On Multi-Objective Micro-Genetic Algorithm

According to forecast of ITRS, with the development of deep micro-nanotechnology, by 2015, the feature size of semiconductor devices will get close to its limit,which is 25-35 nm. At the same time, single-chip will be able to integrate several thousands of IP cores and the clock frequency will reach lOGHz. At this point, with bus technology as the main features, SoC faces huge bottleneck in the address space expansion, communication bandwidth, the system clock. In addition, SoC will not be able to satisfy energy consumption, latency, QoS and other performance requirements of the Applications on chip. Around 2001, Research institution proposed the concept of NoC, which apply the computer network interconnection technology and parallel computing technology on chip design space, and solve the problems of SoC. NoC has been one of the hot topics of current chip technology and the trend of the development of SoC.Currently, NoC mapping is one of the research hot topics of NoC design space. In the NoC application configuration model, the mapping is the key link. On NoC, different mapping scenarios could make a great affection on IP cores communication performance, such as energy consumption and delay et al. In particular, as the scale of application becomes larger and heterogeneous characterization of NoC IP cores become prominent, the adaptability and performance of the mapping algorithms will seriously affect the mapping efficiency. Ultimately, these will affect the performance of NoC. Therefore, it is has high research significance to design and implementation high adaptability, excellent performance and efficiency NoC mapping algorithms.This paper describes the research background and research focus of NoC. It studies the points of NoC topologies, communication protocols, routing exchange algorithms, application configuration model whit research method of comparative analysis. And then, it gives the relevant characterization maps and several mapping optimization index function models. This paper designs and implements a multi-objective optimization Algorithm for NoC Mapping optimization meeting the constraints. The main research contents are as follows:1. Design NoC application configuration model and the mapping process;2. Design and implement XY routing algorithm based on the bandwidth guide factor; 3. Propose and implement a two-phase multi-objective micro-genetic optimization Algorithm TP-MGA for four objectives optimization of NoC Mapping. And we use parellel computing model to promote the performance and efficiency of the TP-MGA. Then we compare our mapping algorithm with other multi-objective optimization methods by experiment.Comparing with other multi-objective optimization methods, this parallel TP-MGA algorithm has high efficiency and performance.