物聯網系統應用的開發逐漸變得困難而且耗時。物聯網的仿真模擬被期望成為一種可以節省時間以及金錢的系統評估優化方法。然而目前現有的仿真模擬工具，多半只關注模擬功能的正確性，缺乏了提供系統優化設計的程式執行軌跡訊息。特別是對於識別瓶頸來說，能夠在原始程序中識別包含程式中不同段執行指令的函式/物件是至關重要。然而，隨著執行訊息剖析能力的增加，也會導致模擬速度的降低，因此需要可以在函式/物件層面能支持多種仿真模擬粒度，以在仿真模擬的時間以及真實度之間作權衡。此外，上述的工具也缺乏了支持與環境互動的功能，因此需要將可隨著時間和位置方面變換的環境模型添加進入仿真模擬平台中，以將環境數據動態地注入到模擬的物聯網設備中。在本論文中，我們提出了一個針對物聯網終端設備具備多粒度與環境建模的仿真模擬平台。我們提出一個源級別剖析機制，可以生成程式執行指令的軌跡追蹤，並將每個執行的指令鏈接到源級別函式。我們提出的系統是可配置的，允許動態地更改函式級別的仿真模擬粒度，並通過環境建模支持外部輸入訊息的注入模擬，以影響針對物聯網終端設備的仿真。

The development of Internet of Things (IoT) systems becomes difficult and time consuming. IoT simulation is desirable to provide a way to evaluate the system to save time and money. Current simulation tools only focus on the functional correctness but lack of execution information for design optimization. Particularly, it needs to identify the functions/objects in the original program that contain the executed instructions, which is essential in identifying bottlenecks. However, addition of the profiling ability slows down the simulation speed, it needs to support multiple simulation granularity at the functions/objects level to trade off simulation fidelity with simulation time. Furthermore, those tools also lack of supports for interactions with the environment. It needs to add environmental models in terms of time and location to dynamically inject environmental data to the simulated IoT devices. In this thesis, we present a multi-granular environment-modeling simulation platform for IoT end devices. We propose a source level profiling mechanism to allow generation of trace of executed instructions and link each executed instruction to the source level function. Our work is configurable to allow change of simulation granularity at the function level dynamically and support environmental modeling with external inputs to influence IoT simulation for the IoT end devices.

1 Introduction 1

2 Background 6
2.1 The QEMU Emulator 6
2.1.1 Dynamic Binary Translation 7
2.1.2 Block Chaining 8
2.2 Instrumentation and Profiling 9
2.3 Executable and Linkable Format 11

3 System Design 12

4 Implementation 20

5 Experiments 26
5.1 Experimental Environment 26
5.2 Validation 28
5.3 Results of Simulation Granularity 29
5.4 Effects of Environment Modeling 30
5.5 Performance Analysis 30

6 Conclusion 38

[1] Tran Van Dung, Ittetsu Taniguchi, and Hiroyuki Tomiyama, “Cache simulation for instruction set simulator qemu”, in Dependable, Autonomic and Secure Computing (DASC), 2014 IEEE 12th International Conference on. IEEE, 2014, pp. 441–446.

[2] “Arm mbed”, https://docs.mbed.com.

[3] “Windows 10 iot core”, https://developer.microsoft.com/zh-tw/windows/iot/Docs.

[4] “Zephyr project”, https://www.zephyrproject.org/doc/index.html.

[5] Fabrice Bellard, “Qemu, a fast and portable dynamic translator.”.

[6] David Thach, Yutaka Tamiya, Shin’ya Kuwamura, and Atsushi Ike, “Fast cycle estimation methodology for instruction-level emulator”, in 2012 Design, Automation & Test in Europe Conference & Exhibition (DATE). IEEE, 2012, pp. 248–251.

[7] N. Nethercote, Dynamic Binary Analysis and Instrumentation, PhD thesis, University of Cambridge, United Kingdom, 2004.

[8] UNIX System Laboratories, System V Application Binary Interface, Prentice-Hall, 1993.

[9] Andreas Ermedahl Jan Gustafsson, Adam Betts and Björn Lisper, “The Mälardalen WCET benchmarks – past, present and future”, Brussels, Belgium, July 2010, pp. 137–147, OCG.

[10] D. Ernst T. M. Austin T. Mudge M. R. Guthaus, J. S. Ringenberg and R. B. Brown, “Mibench: A free, commercially representative embedded benchmark suite”, in WWC ’01 Proceedings of the Workload Characterization, 2001. WWC-4. 2001 IEEE Interna- tional Workshop. Dec 2001, pp. 3–14, IEEE Computer Society Washington, DC, USA.