Research On Low-Power Mapping Of Heterogeneous Multi-Core Network-on-Chip

With the rapid development of network on chip, more processing units (also called IP core) can be integrated into the network on chip. The structure of the system on chip is gradually developing towards the direction of multi cores and heterogeneous. Although heterogeneous multi-core network-on-chip can better meet the demand, speed up the execution speed and improve the performance, but with the cores increasing on network-on-chip, the system power consumption has become a problem that can’t be ignored, because it limits the further improve the system performance and chip integration, and at the same time, it reduces the lifespan of the chip and influences the stability of the chip. In addition, it is a key point to the mobile devices with higher energy consumption performance requirements.Different processing units have different execution performance and power consumption of the task on heterogeneous multi-core systems. As long as assign the task to the processing unit which good at dealing with and mapping to the appropriate communication topology structure, it can be very good to improve system performance and reduce system execution and communication power consumption. But the task allocation and mapping is a NP problems, how to achieve low power consumption mapping is the focus and difficulty of the current research. For this purpose, this paper presents a mapping of quantum ant colony algorithm based on heterogeneous multi-core on-chip network system in the application of task allocation and IP cores mapping problem and realize the application execution and communication task power minimization problem. The algorithm by changing the basic ant colony algorithm in pheromone release, using quantum optimization algorithm quantum probability amplitudes instead of pheromone, and pheromone updating is through the use of quantum optimization algorithm quantum phase rotation way, realize the algorithm of the ant pheromone adaptive updates. This method can effectively reduce the premature convergence of the algorithm when using ant colony algorithm, and increase the search space of the algorithm, which is conducive to the local optimal solution.The application of task communication on random map generated by using the proposed quantum ant colony algorithm and ant colony algorithm mapping, particle swarm algorithm and genetic mapping algorithm for simulation of comparative studies, the algorithm mapping results the minimum total power consumption, the implementation of the same number of the average power consumption, compared to many aspects of node the lowest power consumption power and algorithm of task execution time, the implementation of the algorithm convergence, the topological structure of the execution of the same time the lowest power consumption, comparative mapping algorithm based on the experimental results can be obtained in this algorithm, fast search, global optimization, to achieve low power consumption and time performance and other indicators are better than the. In the simulation experiment, it can be concluded that the proposed algorithm can achieve more than 24% of the total power consumption in the case of fewer iterations.