Optimization Of Network On-chip Mapping Based On Dynamic Variation Of Task Interval In The Range

With the rapid development of the chip industry today,VLSI(Very Large Scale Integrated Circuit Design)technology has also been developed,and multiple processing units can be integrated on a single chip.The traditional SoC bus form has certain dimensions for today's transmission data.limitation.So in 2000,scholars proposed the concept of network on-chip(NoC),effectively solving the bottleneck encountered in the current development of traditional SoC.The structure of the on-chip network has good scalability and supports parallel communication.With the continuous optimization of the performance of the IP core itself,its processing capability is not the main factor that restricts the development of the system.How to realize the real-time performance of massive data transmission has become an urgent problem to be solved in the development of the electronic industry.Therefore,the network on-chip has become the famous field of the academic today.The network on-chip design mainly determines the topology,determines the routing mechanism,and determines the network nodes to the selected topologies.Among them,the mapping scheme of the network on-chip has the greatest impact on the performance and stability of the entire network on-chip.Therefore,the mapping optimization of the network on-chip is an important field of network on-chip research.The mapping of the network on-chip refers to the optimization goal of performance or special conditions that are required according to the system,given the communication task map and IP core set and topology.The IP core can be reasonably allocated to specific locations in the grid,so that the performance of the on-chip network meets the requirements.This paper discusses the research done by scholars on mapping optimization problems.There are algorithms that use mathematical methods to obtain exact solutions,and heuristic algorithms that speed up the efficiency of solving them.For the purpose of mapping optimization,some are based on single target mapping optimization,and some scholars consider multiple aspects for optimization research.In the study of network on-chip mapping optimization,there are different evaluation indicators,such as energy consumption,delay throughput,etc.,where the energy consumption problem is the main aspect of optimization research,reducing energy consumption can not prevent the reliability of the chip from being lowered due to local overheating.For mobile terminals,it can also improve the battery life.At present,most of the research is based on the fact that the communication task is the transmission amount between the given data,and the actual situation may be fluctuating within a certain range.The author uses this as the entry point and adds the basic energy consumption model.In the case of data transmission fluctuations,an improved energy consumption model is established using minmax(minimum maximum load mechanism),and an improved tabu search algorithm is used to solve the model,considering some of the existence of heuristic algorithms such as premature Convergence,falling into local optimization and other defects,adding greedy thought to the original algorithm improves the choice of initial solution and optimizes related parameters.Finally,12 sets of simulation experiments were carried out with specific classical data and random data.The paper compared the advantages and disadvantages of the algorithm with the genetic algorithm and particle swarm optimization algorithm.The results show that when the experimental data scale is small,the results of the three algorithms are the same.However,when the data size is large,the results of the tabu search algorithm are better than those obtained by the other two.At the same time,three sets of horizontal comparison experiments under the same data scale were carried out,which were compared from horizontal to vertical.The experimental results show that for the same scale problem,the algorithm fluctuates between 3%and 4%in the running time of the conclusion,which can guarantee the stability of the algorithm in operation.