洪士灝臺灣大學:資訊工程學研究所許文昌Hsu, Wen-ChangWen-ChangHsu2010-05-172018-07-052010-05-172018-07-052009U0001-2008200916055100http://ntur.lib.ntu.edu.tw//handle/246246/183387本論文主要是在虛擬平台上建立一個軟體剖析(profiling)機制來協助在嵌式系統上的軟體開發工作，此機制主要是建立在QEMU 這個開放原始碼的模擬器。QEMU 模擬器廣泛被使用在開發嵌入式軟體的計畫中，如包含在Android 軟體發工具組中用來作為Android 開放原始碼平台中之虛擬平台，用以加速並協助ndroid 系統上的軟體開發。然而，在現有的QEMU 設計中並沒有估量軟體執行時的相關機制，因此我們嘗試在該虛擬平台上建立一套效能分析的架構，我們的標是提供快速並且準確的時間評估模型來幫助程式設計師在系統開發的早期估程式效能。我們的實驗展示數個不同的時間模型，提供較高精準度的模型需要長的模擬時間，使用者可以跟據自己的需求，在模擬時間與準確率之間折衷選。In this paper, we present a performance analysis framework to support applicationevelopment on heterogeneous multicore systems with virtual platforms. Our works based on the open source QEMU simulator which has been widely used to developoftware on embedded systems and was included in the Android Software Developmentit. However, to support the software development on multicore platfroms,iming information is critical and thus we attempt to build virtual platforms with theapacity of performance evaluation. Our goal is to provide a framework that givesuick, accurate timing estimation to help multicore application development duringarly design phases. In this paper, we discuss the construction of various timingodels for our virtual platform to simulate program execution.Acknowledgements iibstract (Chinese) iiibstract ivist of Tables viiist of Figures viii Introduction 1.1 Scenarios of Using Virtualization Techniques . . . . . . . . . . . . . 2.2 Approaches for Pro ling . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Pro ling with a Virtual Platform . . . . . . . . . . . . . . . . . . . . 5.4 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Background and Related Work 7.1 Cycle-Accurate Simulation . . . . . . . . . . . . . . . . . . . . . . . . 8.2 Functional Emulation . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3 Pro ling Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4 Performance Prediction with simpli ed Hardware Models . . . . . . . 12.5 Virtual Platform with Timing Models . . . . . . . . . . . . . . . . . . 14 Implementation of Timing Models in QEMU 15.1 QEMU Internal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15.2 Overview of our Virtual Platform . . . . . . . . . . . . . . . . . . . . 17.3 Timing Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.3.1 Regression Timing Model A, B and C . . . . . . . . . . . . . . 19.3.2 Cache Simulation for Model C and Datasheet model . . . . . . 20.3.3 Datasheet Model . . . . . . . . . . . . . . . . . . . . . . . . . 21.3.4 OS Scheduler In uence . . . . . . . . . . . . . . . . . . . . . . 21 Experiment Results 24.1 Experiment Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25.2 Experiment Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27.2.1 Results and Discussions for Model A and B . . . . . . . . . . 27.2.2 Model C and Datasheet Model . . . . . . . . . . . . . . . . . . 31 Conclusion and Future Work 38ibliography 40application/pdf4222235 bytesapplication/pdfen-US虛擬化技術嵌入式系統效能分析virtual platformembedded systemsperformance analysisthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/183387/1/ntu-98-R96922119-1.pdf