臺灣博碩士論文加值系統

隨著科技的發展，嵌入式系統的應用越來越廣泛，在某些運算量龐大的系統中，單一核心處理器已漸漸無法負擔。基於對速度的要求，多核心系統便成為研究的目標。在硬體設計方面，軟核心的系統具有便於修改的特性，在系統的開發中，可以降低開發成本。而在軟體方面，要使多核心系統能夠有效發揮其功能，除了硬體架構上的規劃外，平行化的程式設計成為一個重要的課題。現行的編譯器對於程式的平行化難以有效的轉化，程式設計師必須在寫程式之初便對程式的分工做考慮，本論文提供一個平行化的程式架構和函式庫，使程式設計者在設計多核心嵌入式系統的軟體時，有方便的平行化程式設計的環境。藉由程式的測試，本論文所提出的軟體架構能有效的將程式平行化，增加執行的效率。而便於使用的功能函式，也能加快程式的設計時間。

As the development of technology, embedded systems have been widely used in all kinds of applications. In some systems, which need to support a huge amount of computing, uni-core processor is not enough. For this reason, multi-core system development has become more important. In the hardware, soft-core based systems have the characteristic of flexibility. When we develop the system, soft-core based system can reduce the developmental cost. In the software, if we want to make the multi-core system with good performance, in addition to the hardware architecture and task scheduling, parallel programming is an important issue. The current parallelizing compilers are hard to parallelize the programs effectively. The programmer must think about how to allot the task to each processor in the beginning. In this thesis, we present a software framework for designing parallel program. The proposed framework provides a convenient parallel programming environment for programmers to design the multi-core system’s software. From the experiments, the proposed framework can parallelize the program effectively by applying the provided functions.

ACKNOWLEDGMENTS i
中文摘要 ii
ABSTRACT iii
TABLE OF CONTENTS iv
LIST OF FIGURES v
LIST OF TABLE vi
CHAPTER 1 INTRODUCTION 1
CHAPTER 2 BACKGROUND 2
2.1 Multi-core Architectures 2
2.2 Soft-core Based Multi-core System 3
2.3 Parallel Programming 3
CHAPTER 3 PARALLEL PROGRAMMING ENVIRONMENT 6
3.1 Hardware Architecture 6
3.2 Software Environment 7
3.2.1 Partition Function 8
3.2.2 Synchronization Function 11
3.2.3 Collective Communication Function 13
3.2.4 Memory Management Function 18
CHAPTER 4 EXPERIMENT AND DISCUSSION 20
4.1 Experimental Environment 20
4.2 Matrix Multiplication 20
4.3 Livermore Loop 23
4.4 Experimental Results 24
CHAPTER 5 CONCLUSIONS AND FUTURE WORK 27
REFERENCES 28
APPENDIX 29
A. The source code of each function 29

[1]Qihang Huang, Zhiyi Huang, Paul Werstein, and Martin Purvis, “GPU as a General Purpose Computing Resource,” Proceeding of Ninth International Conference on Parallel and Distributed Computing, Applications and Technologies, pages 151 – 158, Dec 2008.
[2]Jonqhan Park and Soonhoi Ha, “Performance Analysis of Parallel Execution of H.264 Encoder on the Cell Processor,” Proceedings of IEEE/ACM/IFIP Workshop on Embedded Systems for Real-Time Multimedia, pages 27 -32. Oct. 2007.
[3]Jian-Jia Chen, Chuan-Yue Yang, Tei-Wei Kuo, and Chi-Sheng Shih, “Energy-Efficient Real-Time Task Scheduling in Multiprocessor DVS Systems,” Proceedings of Design Automation Conference 2007, ASP-DAC’07. Asia and South Pacific. Pages 342 – 349, Jan. 2007.
[4]Message Passing Interface Forum. http://www.mpi-forum.org/.
[5]OpenMP: Simple, Portable, Scalable SMP Programming. http://openmp.org/wp.
[6]NVIDIA, CUDA Programming Guide, Version 2.2.1. http://www.nvidia.com/object/cuda_develop.html, May 2009.
[7]OpenCL: The open standard for parallel programming of heterogeneous systems. http://www.khronos.org/opencl/.
[8]David A. Bader, Varun Kanade, and Kamesh Madduri, “SWARM: A Parallel Programming Framework for Multicore Processors,” Proceeding of International Conference on Parallel and Distributed Processing Symposium 2007, pages 1-8, March 2007.
[9]Altera Co., “Creating Multiprocessor NiosII System,” San Jose, 2007.