基于PVM的GAMMA程序执行环境研究

由Ｂａｎａｔｒｅ和Ｍｅｔａｙｅｒ提出的并行计算模型ＧＡＭＭＡ,通过使用多重集转化技术,允许在料高层次开发求解问题的逻辑并行性,目前已被广泛作并行程序构造工具。本文探讨如何基于网络并行计算环境－ＰＶＭ有效地实现ＧＡＭＭＡ模型。实验结果表明ＧＡＭＭＡ作为高层、隐式并行模型适合开发大规模并行程序。     

Reusing Test Cases Based on the Function Point

Reusing test cases from existing test case library is quite common in the software testing field. Testing practice tells us that there is a strong relationship between the granularity of a function unit under testing and that of the test case. A function unit with small granularity usually results in the test cases with the same small granularity. Therefore a test case defined as the function point,i. e.,the smallest size function unit,was provided for the first time.Though test cases with smaller granularity usually have better reusability,the cost of accurately reusing and integrating such test cases is also higher. In order to balance the test case reusability and the cost of test case reuse,a novel test case reuse model based on the function point was proposed in this paper. In this model,a reusable test case for specification-based testing was defined and some reuse strategies and three formal reuse methods were given. Finally,the complete automatic software process was realized by a reusing generation tool. The new method has improved reuse accuracy,while greatly enhances the software productivity.     

网络并行计算系统性能评估工具

本文通过分析网络并行计算系统性能评估的重要性，阐述了网络并行计算系统性能评估工具的种类及其特点，在此基础上探讨并比较了目前流行的几种性能评估工具的结构、功能和特点．     

面向互联网应用的云操作系统的架构设计

云计算的核心理念是按需提供IT服务, 这对基础设施提供商带来巨大的影响. 介绍以中小企业为主要服务群体的云操作系统INACloud的架构设计, 并且综合考虑实际运维问题、自主创新和开源软件等因素以及面向互联网应用动态多变的资源需求, 重点探讨调度策略、监控设计和计费管理等方面.     

CVL的分析与改进

ＣＶＬ是美国卡内基·梅隆大学的研究人员用ＡＮＳＩＣ语言编写的向量函数库，它包括许多向量并行操作原语供编写并行程序时使用．本文详细描述了ＣＶＬ的主要特点及对其作的一些有益改进，包括增加了向量动态分段和数据动态分配功能等．通过这些改进，明显地简化了编写并行程序的过程，提高了ＣＶＬ并行操作原语执行的效率．     

The Web Application Test Based on Page Coverage Criteria

Software testing coverage criteria play an important role in the whole testing process. The current coverage criteria for web applications are based on program or URL. They are not suitable for black-box test or intuitional to use. This paper defines a kind of test criteria based on page coverage sequences only navigated by web application, including Page_Single, Page_Post, Page_Pre, Page_Seq2, Page_SeqK. The test criteria based on page coverage sequences made by interactions between web application and browser are being under consideration after that. In order to avoid ambiguity of natural language, these coverage criteria are depicted using Z formal language. The empirical result shows that the criteria complement traditional coverage and fault detection capability criteria.     

并行多层快速多极子算法最细层数据的建立

分布树最细层数据的建立是并行多层快速多极子算法(MLFMA)的基础.最细层数据分布均匀与否,会影响到后续整个算法的负载平衡及其并行效率.研究了最细层数据的建立方法,提出采用并行正则采样排序算法来代替众多文献所推荐的并行桶排序算法, 以此来建立分布树的最细层数据.针对多种常见的散射体在不同处理器数下的实验结果表明, 在大多数情况下,改进后的算法较原算法性能有较为明显的提高.