Scholarlink

Energy Aware Scheduler (EAS) is a main method for power management in mobile devices, and its effectiveness is largely influenced by its energy model. The energy model with EAS is overly simplistic, leading to accuracy issues that subsequently impact scheduling efficiency. This is because the energy model neglects static power consumption and fails to address the dynamic power differences across different workload types. Static power consumption and workload types must be integrated into the energy model of EAS to enhance its scheduling efficiency. We established a Dynamic Power Model for EAS (DM-EAS) based on the Performance Monitoring Unit (PMU) power model. DM-EAS utilizes PMU information to characterize dynamic power consumption, identify the impacts of temperature and workload types on actual power consumption, and adjust power consumption in real time. Compared to the previous EAS energy model, the established model showed an average power optimization of 8.21% in tested applications.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Performance Monitoring Unit-Based Power Modeling for Mobile Energy-Aware Scheduler

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yuzhao Liang,
  • Jintao Ge,
  • Xing Gao

摘要

Energy Aware Scheduler (EAS) is a main method for power management in mobile devices, and its effectiveness is largely influenced by its energy model. The energy model with EAS is overly simplistic, leading to accuracy issues that subsequently impact scheduling efficiency. This is because the energy model neglects static power consumption and fails to address the dynamic power differences across different workload types. Static power consumption and workload types must be integrated into the energy model of EAS to enhance its scheduling efficiency. We established a Dynamic Power Model for EAS (DM-EAS) based on the Performance Monitoring Unit (PMU) power model. DM-EAS utilizes PMU information to characterize dynamic power consumption, identify the impacts of temperature and workload types on actual power consumption, and adjust power consumption in real time. Compared to the previous EAS energy model, the established model showed an average power optimization of 8.21% in tested applications.