透過您的圖書館登入
IP:18.116.13.113
  • 期刊

記憶體功耗管理之ESL模擬平台建置

ESL-based Simulation Platform for Memory Power Management

摘要


在行動裝置中,記憶體功耗占系統功耗的比重逐漸地增加。然而,針對降低記憶體功耗,有關於Linux記憶體管理與應用程式執行時之記憶體功耗動態分析的研究數量仍屬少數。此外,由於傳統的記憶體功耗分析技術需要透過電路板以及儀器去量測電流數據,因此無法在設計初期去分析作業系統記憶體管理對記憶體功耗的影響。因此,本論文提出一個整合LPDDR2控制器模型與QEMU的整合模擬平台,用以分析執行Android應用程式時記憶體的存取行為和功耗。實驗結果顯示當使用兩種不同的記憶體管理策略時,其平均功耗差異為1.21mW以及效能差異為119.94秒。

並列摘要


For mobile devices, the ratio of memory power consumption to system power consumption is continuously increasing. However, for reducing memory power consumption, related work for Linux memory management and dynamic memory power analysis during running applications is still scarce. Moreover, because conventional memory power analysis techniques measure current values via circuit board and instrument, it is unable to analyze the impact of operating system memory management on memory power at early design stage. Therefore, this paper proposes an integrated simulation platform with a LPDDR2 controller model and QEMU to analyze memory access behavior and power consumption when executing an Android application. Experimental results show that power difference is 1.21mW and time difference is 119.94 seconds when adopting two different memory management policies.

延伸閱讀