基于PVM on Win32的网络并行数值计算研究
发布时间：2022-02-23 22:35
高速信息网络的飞速发展以及网络并行计算支撑软件如PVM、MPI、Express、Linda,P4等的出现,使得网络并行计算的投资少、见效快、灵活性强、性能价格比高等诸多优点显得更加突出,受到国内外越来越多的大学和科研单位的青睐,已成为并行计算和分布式计算技术的重要发展方向。本研究结合具体的数值试验,主要完成了以下工作:（1）探讨了PVM on Win32的网络并行数值计算平台的构架,并行环境变量的设置,PVM的启动及虚拟机的配置等;（2）较详细地讨论了基于PVM on Win32的网络并行数值计算程序设计的方法及注意事项,包括PVM基本编程模式的研究、任务的分解、通信的设计及PVM并行数值算法的数值稳定性等;（3）重点研究了影响PVM并行程序性能的几个重要因素,包括负载平衡、通信开销、网络性能、任务粒度、处理机个数以及处理机内存容量问题等,并提出了相应的策略,以最大限度地提高PVM并行程序的性能;（4）针对基于PVM on Win32的由桌面PC机联网而成的网络并行计算环境中,处理机的运算速度较快而处理机间的通信相对较慢的实际情况,将求解线性方程组的Gauss-Jordan消去法与Ga...
【文章来源】：贵州师范大学贵州省
【文章页数】：45 页
【学位级别】：硕士
【文章目录】：
Abstract in Chinese
Abstract
1 Introduction
1.1 The Prospect of Network Parallel Computing
1.2 PVM and Its Characteristics
1.3 The Current Research Status
1.4 Main Work in This Study
1.5 The Significance of This Study
2 Building PVM on Wi1132 Network Parallel Environments
2.1 Installing and Setting PVM on Wi1132 System
2.2 Starting PVM and Configuring Virtual Machine
3 Parallel Programming Under PVM on Win32 Network Parallel Environments
3.1 Decomposing Task
3.2 Designing Communication
3.3 Choosing Programming Models
3.4 Writing, Compiling and Executing PVM Parallel Programs
3.5 Debugging PVM Parallel Programs
3.6 Optimizing PVM Parallel Numerical Programs
4 Factors Affecting the Performance of PVM Parallel Programs
4.1 Load-Balancing
4.1.1 The Importance of Load-Balancing
4.1.2 Methods for Load-Balancing
4.2 Communication Costs
4.2.1 Communication Costs Are the Key Factor Affecting the Performance of PVM Parallel Programs
4.2.2 Techniques of Decreasing Communication Costs
4.3 Network Performance
4.3.1 Network Performance Affects the Performance of PVM Parallel Programs
4.3.2 How to Improve Network Performance
4.4 Task Granularity and the Number of Processors
4.4.1 Task Granularity and the Number of Processors Affect Speedup
4.4.2 A Strategy for Optimizing the Number of Processors
4.5 Memory Problems
5 Network Parallel Computing Practices
5.1 A Parallel Algorithm for Solving Dense Linear Equations
5.1.1 Basic Idea
5.1.2 Numerical Experiment
5.1.3 Analysis and Enlightenment of the Experimental Results
5.2 A Piecewise Gauss-Seidel Parallel Iterative Algorithm for Solving Linear Equations
5.2.1 Basic Idea
5.2.2 Numerical Experiment
5.2.3 Analysis and Enlightenment of the Experimental Results
5.3 A Parallel Subspace Iterative Algorithm for Eigenvalue Problems and Its Implementation on LAN
5.3.1 Serial Subspace Iterative Algorithm
5.3.2 Basic Idea of Parallel Algorithm
5.3.3 Numerical Experiment
5.3.4 Analysis and Enlightenment of the Experimental Results
6 Conclusion
Bibliography
Appendix: Published Papers of the Author
Acknowledgements
原创性声明
关于学位论文使用授权的声明
【参考文献】：
期刊论文
[1]微机网络环境下提高PVM并行程序性能的策略[J]. 尚月强. 计算机工程与设计. 2007(13)
[2]局域网上求解三角形方程组的一种并行算法[J]. 尚月强. 计算机工程与应用. 2007(19)
硕士论文
[1]基于主机负载预测的机群动态任务调度策略研究[D]. 陈荣征.广东工业大学 2008
本文编号：3641436
【文章来源】：贵州师范大学贵州省
【文章页数】：45 页
【学位级别】：硕士
【文章目录】：
Abstract in Chinese
Abstract
1 Introduction
1.1 The Prospect of Network Parallel Computing
1.2 PVM and Its Characteristics
1.3 The Current Research Status
1.4 Main Work in This Study
1.5 The Significance of This Study
2 Building PVM on Wi1132 Network Parallel Environments
2.1 Installing and Setting PVM on Wi1132 System
2.2 Starting PVM and Configuring Virtual Machine
3 Parallel Programming Under PVM on Win32 Network Parallel Environments
3.1 Decomposing Task
3.2 Designing Communication
3.3 Choosing Programming Models
3.4 Writing, Compiling and Executing PVM Parallel Programs
3.5 Debugging PVM Parallel Programs
3.6 Optimizing PVM Parallel Numerical Programs
4 Factors Affecting the Performance of PVM Parallel Programs
4.1 Load-Balancing
4.1.1 The Importance of Load-Balancing
4.1.2 Methods for Load-Balancing
4.2 Communication Costs
4.2.1 Communication Costs Are the Key Factor Affecting the Performance of PVM Parallel Programs
4.2.2 Techniques of Decreasing Communication Costs
4.3 Network Performance
4.3.1 Network Performance Affects the Performance of PVM Parallel Programs
4.3.2 How to Improve Network Performance
4.4 Task Granularity and the Number of Processors
4.4.1 Task Granularity and the Number of Processors Affect Speedup
4.4.2 A Strategy for Optimizing the Number of Processors
4.5 Memory Problems
5 Network Parallel Computing Practices
5.1 A Parallel Algorithm for Solving Dense Linear Equations
5.1.1 Basic Idea
5.1.2 Numerical Experiment
5.1.3 Analysis and Enlightenment of the Experimental Results
5.2 A Piecewise Gauss-Seidel Parallel Iterative Algorithm for Solving Linear Equations
5.2.1 Basic Idea
5.2.2 Numerical Experiment
5.2.3 Analysis and Enlightenment of the Experimental Results
5.3 A Parallel Subspace Iterative Algorithm for Eigenvalue Problems and Its Implementation on LAN
5.3.1 Serial Subspace Iterative Algorithm
5.3.2 Basic Idea of Parallel Algorithm
5.3.3 Numerical Experiment
5.3.4 Analysis and Enlightenment of the Experimental Results
6 Conclusion
Bibliography
Appendix: Published Papers of the Author
Acknowledgements
原创性声明
关于学位论文使用授权的声明
【参考文献】：
期刊论文
[1]微机网络环境下提高PVM并行程序性能的策略[J]. 尚月强. 计算机工程与设计. 2007(13)
[2]局域网上求解三角形方程组的一种并行算法[J]. 尚月强. 计算机工程与应用. 2007(19)
硕士论文
[1]基于主机负载预测的机群动态任务调度策略研究[D]. 陈荣征.广东工业大学 2008
本文编号：3641436
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/3641436.html
