On-chip Network Region Segmentation Based On Genetic Algorithm Mapping Algorithm Research

With the continuous development of semiconductor technology and process, the level of IC integration continues to increase. There are some bottlenecks in the development of System-on-Chip (SOC) which proposed in the middle of20th century. Some academics and research units draw the computer network technology into IC design and propose the Network-on-chip (NoC) concept of IC design.Network-on-Chip separates the calculations with the communications, based on packet communication, with the computer network technology, using global asynchronous local synchronous as its clock solution to solve the problems of the System-on-Chip in the development process.This article briefly introduces the common model and algorithm of the platform topology and routing of the Network-on-Chip at the beginning, then introduces Network-on-Chip mapping which is one of the core issues of the Network-on-chip.This article advances a kind of Network-on-Chip region segmentation mapping algorithm based genetic algorithm combined with the existing Network-on-Chip algorithm that divides the application characteristics graph of Network-on-Chip which contains some cut points into some subgraph using the graph cutting method of graph theory, and use mapping algorithm to optimize each subgraph. By dividing the application characteristics graph, the totalsearch space is reduced. After region Segmentation, this algorithm can achieve parallel computing. Then we amend the genetic algorithm process at the NoC mapping application of GA and propose a improved genetic algorithm to conserve the genetic diversity against the defects of the genetic algorithms combined with the improved methods of genetic algorithm and the ideas of protecting the population diversity of immune algorithm. This improved algorithm introduces the elite first strategy and protects the genetic diversity among the population while take full advantage of the genetic diversity to optimize the global search consoles of Network-on-Chip mapping. Simulation results gived at the end of each chapter show that the performance of the improved algorithm gets better result than the traditional genetic algotirhm.