Abstract: Combined with Dynamic Voltage Scaling (DVS) and (m,k)-firm model, a completion ratio guaranteed dynamic energy efficiency algorithm named VAP_DY for the execution of dependent tasks on soft real-time multiprocessor systems with multiple supply voltages was proposed. VAP_DY leveraged application's performance requirements, uncertainties in execution time, and tolerance for reasonable execution failures to scale each processor's supply voltage at run-time to reduce multiprocessor system's total energy consumption. Analyses and experiments show that VAP_DY can effectively reduce energy consumption, while guaranteeing both timing constraint and completion ratio.

Key words: energy consumption, completion ratio, timing constraint, multiple voltage, Dynamic Voltage Scaling (DVS)

摘要： 结合DVS技术和(m,k)-firm模型，提出一个保证完成率、适用于软实时多电压多处理器系统中有依赖关系任务集的动态低功耗算法VAP_DY。该算法权衡应用的性能需求、执行时间的不确定性和系统对合理执行失败的容忍来动态调整每个处理器运行时的供电电压,以降低多处理器系统的总功耗。分析和实验结果表明，VAP_DY能够在保证时间和完成率约束的条件下有效降低系统功耗。

关键词: 功耗, 完成率, 时间约束, 多电压, 动态电压调整