The Research And Development Of Placement Algorithm For Network On Chip

As semiconductor process technology into the nano-stage, how to effectively use the transistor becomes a new problem in multi-core architecture. Only following the single-core development, chip design will face the performance bottlenecks in the interconnect delay, memory bandwidth, power limits and others.Therefore, it is necessary to study the new type of chip architecture to accommodate the seemingly contradictory demands between performance improvement and power consumption decreasement simultaneously. The multi-core can use multiple low-frequency nuclear units generate more than high-frequency single-core processing power to get better cost performance. Multi-core series of technical problems the chip industry has become a recent focus of the study and future major development direction. Multi-core technology is an important contemporary IC design technology, which has low power consumption and strong parallel processing, as well as excellent computing performance to satisfy people's demand for integrated circuit performance.A complete NoC design methodology includes many aspects which are critical to the development of NoC has caused extensive academic research. NoC design space of NoC research concepts and key issues grouped into three categories: infrastructure, communication mechanisms and mapping optimization. This paper is based on the applications of 256 IP core are mapped to a 16 x 16 blocks of the 2D-mesh topology process.In the usual NoC architecture design, high performance and low power consumption have become two important metrics to evaluate whether the structure is efficient enough to manage applications. In this paper, simulated annealing, neural networks, ant colony algorithm, genetic algorithm and so on intelligent algorithms are proposed and analyzed for placement. The largest space matrix as an important constraint is introduced for the new placement based on genetic idea, which is inspired by biological evolution derived from a global optimization algorithm. The better solution is selected out of a sudden mutation, the laws of genetic inheritance and other fine species to adapt to environmental changes. The essence is a direct searching method not depending on the specific issues of. It uses a biological point of view of genetics, with natural selection, genetic variation and other mechanisms to achieve the improvement of each individual adaptation. This algorithm is designed primarily to improve the NoC platform to run applications to avoid conflicts, optimize the layout so to reduce the communication in the router's running multiple applications compete for network resources. According to the experiment, we conclude that the results show lower latency and better performance using our intelligent placement algorithm than before.