论文部分内容阅读
与同构多核处理器相比,单指令集异构多核处理器能够更好的匹配程序行为的多样性,从而具有更好的性能功耗比.异构多核处理器的能效优势依赖于操作系统合理而有效的调度,追求性能与功耗的统一,是典型的多目标优化问题.提出将多目标优化遗传算法应用于寻找异构多核环境下最优的静态任务调度方案,提出表征任务相对顺序的染色体编码结构,使种群初始化时的有效个体所占比例变为100%.提出使用先序关系矩阵来确定任务的执行顺序,克服了高度值方法存在的严重不足.仿真结果表明,先序关系矩阵方法能扩大搜索范围,在种群规模足够大时,可以找到高度值方法漏掉的部分最优解.
Compared with isomorphic multi-core processors, single instruction set heterogeneous multi-core processors can better match the diversity of program behaviors and thus have better performance-to-power ratio.The energy efficiency of heterogeneous multi-core processors depends on the operating system Reasonable and effective scheduling, the pursuit of the unity of performance and power consumption, is a typical multi-objective optimization problem.Multi-objective optimization genetic algorithm is proposed to find the optimal static task scheduling scheme in heterogeneous multicore environment, and the relative order of characterization tasks , The proportion of effective individuals in population initialization is changed to 100% .Priority relation matrix is proposed to determine the execution order of tasks and overcome the serious shortage of height value methods. The simulation results show that the order of precedence The matrix method can expand the search range. When the population size is large enough, we can find some optimal solutions that the height value method missed.