在論文中，我們提出了一種基於時間同步的確定性執行技術有效且易於使用的平行程式的偵錯方法。平行程式由於不確定性(nondeterministic)的性質和非再造性(non-reproducible)的錯誤知名難以偵錯。為了解決這個問題，我們提出了一個平行程式偵錯器，命名為GDB-Parallel。GDB-Parallel基於確定性多核心指令集仿真（Multi-Core Instruction Set Simulation, MCISS）技術，能夠有效控管平行程式以及監督多執行序之間的通信和交錯次序。基本上，我們的方法，根據來自使用者選擇的時間模型得出的時間順序來構建平行程式的執行路徑，有效地消除了執行重放(replay)技術有關的問題和限制。我們通過修改QEMU將它和GDB做整合，進而實作出GDB-Parallel用以證明我們的想法。我們的偵錯器的使用和操作幾乎和傳統的GDB偵錯器一樣，因此我們保持了原本GDB的功能並使它在平行程式偵錯時能夠發揮功效，並額外設計了一些新的功能讓平行程式偵錯更有效率。我們已經測試了許多實際的例子，並且我們在論文的最後展示了幾個快照來證明我們工具的有效性。

In this thesis we propose an effective and easy-to-use parallel program debugging approach based on timing-synchronized deterministic execution technology. Parallel programs are known difficult to debug because of the nondeterministic nature and non-reproducible bugs. To resolve this issue, we propose a parallel programs debugger, named GDB-Parallel, which is capable of handling multi-thread programs and monitoring thread communications and interleavings based upon deterministic Multi-Core Instruction Set Simulation (MCISS) technology. Basically, our approach constructs execution path according to the chronological order derived from a user selected timing model and effectively eliminates the replay-related issues and concerns. We have actually implemented the idea by modifying QEMU and integrating with GDB. The usage and operation of our proposed debugger is almost the same as the traditional sequential GDB debugger as we maintain all sequential debugging features while having some new parallel debugging features. We have tested on many real examples and we show a few snapshots at the end of this thesis to demonstrate the effectiveness of our tool.

1. Introduction 2
2. Existing Debugging Techniques 7
2.1 The Heuristic Approach 7
2.2 The Manual Tracing Approach 8
3. Performing Deterministic Executions 11
3.1 Identify Sync Points 11
3.2 Timing Model 12
3.3 Synchronization 13
4. Debugging Through Deterministic Executions 14
4.1 Adapt GDB for Parallel Debugging 14
4.1.1 Breakpoint 16
4.1.2 Step-Tracing 18
4.1.3 Runtime State Modification 19
4.1.4 Sync-Step Function 20
4.2 Timing Models 22
4.3 Constructing Debugger Using MCISS 25
5. Implementation and Testing 26
5.1 Modify QEM into MCISS 27
5.2 Testing Setup 28
5.3 Testing Results 29
5.3.1 Determinism 29
5.3.2 Performance 30
6. Walk Through a Real Example 30
7. CONCLUSIONS 35
8. REFERENCES 36

[1] LEE, Edward A. The problem with threads. Computer, 2006, 39.5: 33-42.
[2] SUTTER, Herb; LARUS, James. Software and the concurrency revolution. Queue, 2005, 3.7: 54-62.
[3] TAN, M. A Minimal GDB Stub for Embedded Remote Debugging, 2002.Columbia University.
[4] STALLMAN, Richard M.; PESCH, Roland H. Using GDB: A guide to the GNU source-level debugger. Free software foundation, 1991.
[5] ZHU, Jianwen; GAJSKI, Daniel D. A retargetable, ultra-fast instruction set simulator. In: Design, Automation and Test in Europe Conference and Exhibition 1999. Proceedings. IEEE, 1999. p. 298-302.
[6] LIN, P.-CP; DU, Evason; TSAY, Ren-Song. A fast and accurate instruction-oriented processor simulation approach. In: VLSI Design, Automation, and Test (VLSI-DAT), 2013 International Symposium on. IEEE, 2013. p. 1-5.
[7] LIN, Kai-Li; LO, Chen-Kang; TSAY, Ren-Song. Source-level timing annotation for fast and accurate TLM computation model generation. In: Design Automation Conference (ASP-DAC), 2010 15th Asia and South Pacific. IEEE, 2010. p. 235-240.
[8] CHEN, Shu-Yung; CHEN, Chien-Hao; TSAY, Ren-Song. An activity-sensitive contention delay model for highly efficient deterministic full-system simulations. In: Design, Automation and Test in Europe Conference and Exhibition (DATE), 2014. IEEE, 2014. p. 1-6.
[9] WU, Meng-Huan, et al. A high-parallelism distributed scheduling mechanism for multi-core instruction-set simulation. In: Proceedings of the 48th Design Automation Conference. ACM, 2011. p. 339-344.
[10] WU, Meng-Huan, et al. An effective synchronization approach for fast and accurate multi-core instruction-set simulation. In: Proceedings of the seventh ACM international conference on Embedded software. ACM, 2009. p. 197-204
[11] WU, Meng-Huan, et al. An effective synchronization approach for fast and accurate multi-core instruction-set simulation. In: Proceedings of the seventh ACM international conference on Embedded software. ACM, 2009. p. 197-204.
[12] YU, Fan-Wei, et al. A critical-section-level timing synchronization approach for deterministic multi-core instruction set simulations. In: Proceedings of the Conference on Design, Automation and Test in Europe. EDA Consortium, 2013. p. 643-648.
[13] ZENG, Bo-Han; TSAY, Ren-Song; WANG, Ting-Chi. An efficient hybrid synchronization technique for scalable multi-core instruction set simulations. In: Design Automation Conference (ASP-DAC), 2013 18th Asia and South Pacific. IEEE, 2013. p. 588-593.
[14] CARLSON, Trevor E.; HEIRMAN, Wim; EECKHOUT, Lieven. Sniper: exploring the level of abstraction for scalable and accurate parallel multi-core simulation. In: Proceedings of 2011 International Conference for High Performance Computing, Networking, Storage and Analysis. ACM, 2011. p. 52.
[15] MOLLOY, Michael Karl. On the integration of delay and throughput measures in distributed processing models. 1981.
[16] AJMONE MARSAN, M., et al. Modeling bus contention and memory interference in a multiprocessor system. Computers, IEEE Transactions on, 1983, 100.1: 60-72.
[17] LUCIA, Brandon; WOOD, Benjamin P.; CEZE, Luis. Isolating and understanding concurrency errors using reconstructed execution fragments. In:ACM SIGPLAN Notices. ACM, 2011. p. 378-388.
[18] ZHANG, Wei, et al. ConSeq: detecting concurrency bugs through sequential errors. In: ACM SIGPLAN Notices. ACM, 2011. p. 251-264.
[19] ERICKSON, John, et al. Effective Data-Race Detection for the Kernel. In: OSDI. 2010. p. 1-16.
[20] MUSUVATHI, Madanlal, et al. Finding and Reproducing Heisenbugs in Concurrent Programs. In: OSDI. 2008. p. 267-280.
[21] SAVAGE, Stefan, et al. Eraser: A dynamic data race detector for multithreaded programs. ACM Transactions on Computer Systems (TOCS), 1997, 15.4: 391-411.
[22] MOZILLA, rr-Project. [Website]. 2014. Available from: http://rr-project.org/.
[23] ECLIPSE: Parallel Tools Platform (PTP) User Guide. [Website]. 2014. Available from: http://help.eclipse.org/juno/index.jsp?topic=%2Forg.eclipse.ptp.doc.user%2Fhtml%2F06parDebugging.html
[24] ALTEKAR, Gautam; STOICA, Ion. ODR: output-deterministic replay for multicore debugging. In: Proceedings of the ACM SIGOPS 22nd symposium on Operating systems principles. ACM, 2009. p. 193-206.
[25] ZAMFIR, Cristian, et al. Debug determinism: the sweet spot for replay-based debugging. In: Workshop on Hot Topics in Operating Systems. 2011.
[26] NARAYANASAMY, Satish; POKAM, Gilles; CALDER, Brad. Bugnet: Continuously recording program execution for deterministic replay debugging. In: ACM SIGARCH Computer Architecture News. IEEE Computer Society, 2005. p. 284-295.
[27] WANG, Yan, et al. DrDebug: Deterministic Replay based Cyclic Debugging with Dynamic Slicing. In: Proceedings of Annual IEEE/ACM International Symposium on Code Generation and Optimization. ACM, 2014. p. 98.
[28] GOTTSCHLICH, Justin E., et al. Concurrent predicates: A debugging technique for every parallel programmer. In: Proceedings of the 22nd international conference on Parallel architectures and compilation techniques. IEEE Press, 2013. p. 331-340.
[29] BERGAN, Tom, et al. CoreDet: a compiler and runtime system for deterministic multithreaded execution. In: ACM SIGARCH Computer Architecture News. ACM, 2010. p. 53-64.
[30] DEVIETTI, Joseph, et al. DMP: deterministic shared memory multiprocessing. In: ACM SIGARCH Computer Architecture News. ACM, 2009. p. 85-96.
[31] LIU, Tongping; CURTSINGER, Charlie; BERGER, Emery D.Dthreads: efficient deterministic multithreading. In: Proceedings of the Twenty-Third ACM Symposium on Operating Systems Principles. ACM, 2011. p. 327-336.
[32] OLSZEWSKI, Marek; ANSEL, Jason; AMARASINGHE, Saman. Kendo: efficient deterministic multithreading in software. ACM Sigplan Notices, 2009, 44.3: 97-108.
[33] AVIRAM, Amittai, et al. Efficient system-enforced deterministic parallelism. Communications of the ACM, 2012, 55.5: 111-119.
[34] BELLARD, Fabrice. QEMU, a Fast and Portable Dynamic Translator. In: USENIX Annual Technical Conference, FREENIX Track. 2005. p. 41-46.
[35] WOO, Steven Cameron, et al. The SPLASH-2 programs: Characterization and methodological considerations. In: ACM SIGARCH Computer Architecture News. ACM, 1995. p. 24-36.
[36] MARK, D. Hill and Min Xu. Racey: A stress test for determinis tic execution. In http://www.cs.wisc.edu/~markhill/racey.html
[37] DOWNEY, Allen B. The Little Book of Semaphores. Version, 2005, 2.5: 11-15.
[38] SHAMS, Ramtin; KENNEDY, R. A. Efficient histogram algorithms for NVIDIA CUDA compatible devices. In: Proc. Int. Conf. on Signal Processing and Communications Systems (ICSPCS). 2007. p. 418-422.
[39] HUANG, Jeff; ZHANG, Charles; DOLBY, Julian. CLAP: recording local executions to reproduce concurrency failures. In: ACM SIGPLAN Notices. ACM, 2013. p. 141-152.