Research On Low Power Algorithm For Heterogeneous Multi-core System

With the popularity of digital technology and multimedia technology,portable and mobility strong products obtained the unprecedented development.Among them, the products of battery life greatly influenced consumers’ shoppingtendencies. Facing with the situation of limited battery development, the problemof power consumption has become a difficulty and a focus of attention in thedesign of embedded system. With the development of multi-core processor,design for low power consumption becomes more complex, given the effection ofthe power consumption from hardware/software partitioning and schedulingalgorithm, this paper based on solving the low power algorithm research underthe heterogeneous multi-core processor system.In the analysis and summary on the basis of the existing algorithm of lowpower consumption, this paper uses the two phase heuristic algorithm to solve theheterogeneous multi-core processor system under low power design problem.First of all, modeling of system structure and constructing a directed acyclicgraph. Then, distribute the tasks to the corresponding processing unit. Taskdivision effects to a certain extent, affect the later the potential of reducingenergy consumption, the result of task division affect the potential of the lowerenergy consumption. At the same time, the scheduling algorithm after thedivision determines the overall system energy consumption level.In hardware/software partitioning phase, this paper apply the powerfulparallelism of quantum computing combine quantum algorithm with geneticalgorithm for quantum genetic algorithm. Its unique way of qubits encodedmakes a quantum chromosome can also characterize a variety of matchingcondition of traditional chromosome, using quantum revolving door instead ofthe traditional Operation of selection, crossover and mutation updates, in theensure the diversity algorithm at the same time make population with large probability to good pattern evolution. In addition, quantum computing with astrong parallelism to handle huge amounts of data quickly in a short time, thusgreatly reduces the time complexity of the algorithm.After the method to determine the distribution of the task, the key of taskscan completed lies in whether to adopt efficient task scheduling algorithm.Combining with the current popular dynamic voltage scaling technology, thispaper puts forward a scheduling algorithm based on DVS, this algorithm priorityarrangement key task nodes, execution time variance as the priority principle fornon-critical task nodes, by dynamic voltage scaling technology graduallyapproaching the deadline task overtime, maximum reduce the power consumptionof the system as a whole.To verify the algorithm performance, this paper designs the simulationsystem, using TGFF tools generates data as input parameters, C language is usedto program experiment. For objective shows the effectiveness of the algorithm,each stage sets up experiments respectively, mainly from two aspects of loweringpower consumption rate and the time complexity of the algorithm to evaluatealgorithm. Verification results shows that the quantum genetic algorithm in thecourse of evolution stability, fast convergence speed and global optimizationability is strong; With the combination of dynamic voltage scaling priority listscheduling algorithm significantly reduces the system power consumption,shortens the time complexity of this algorithm, effectively achieves the purposeof reducing power consumption in a heterogeneous multi-core processor system.