Performance of an embedded optical vector matrix multiplication processor architecture

An embedded architecture of optical vector matrix multiplier (OVMM) is presented. The embedded architecture is aimed at optimising the data flow of vector matrix multiplier (VMM) to promote its performance. Data dependence is discussed when the OVMM is connected to a cluster system. A simulator is built to analyse the performance according to the architecture. According to the simulation, Amdahl's law is used to analyse the hybrid opto-electronic system. It is found that the electronic part and its interaction with optical part form the bottleneck of system.

嵌入式DSP系统中SDF模型资源优化技术研究

随着嵌入式DSP系统硬件的飞速发展，各种数据和控制流被应用到嵌入式DSP应用程序的计算机辅助设计中去。其中同步数据流(SDF)被广泛用于图形化的DSP设计环境。同步数据流的特点在于能够在编译时刻静态地确定一个有效的调度序列，从而减少运行时的开销。而对于资源短缺的DSP系统，如何优化同步数据流编译生成程序的内存和计算资源具有重要的意义。 对于内存资源优化，研究热点计算体单一出现调度(SAS)算法对于存在反馈环和数据密集处理的应用不可解或内存优化效果很差。文中提出了将SAS和Non-SAS类型调度算法相结合的层次化的存储优化方法。该方法定义了数据密集分量和强连通分量来描述环和数据密集处理结构，并依据数据优先消耗原则设计了启发式的Non-SAS调度算法对分量进行存储优化。该方法适用于任意SDF模型，并有良好的存储优化效果。实验结果证实了其有效性。 计算资源优化即SDF模型的并行计算。本文针对简单SDF模型，提出一种基于优先权的多处理调度方法。它为程序模块分级，采用高优先级优先执行的策略实现了SDF图的多处理器并行静态调度。相对于其他方法，它不需要把SDF图转换为先序图，所以具有更好的时间与空间复杂度。实验结果证实了该方法的有效性。 针对计算体数目多，输入输出速率大的复杂SDF模型，一种基于通信的快速多水平优化方法被提出。它把SDF模型转化为基于通信的数据流模型，并应用快速多水平优化算法对其进行分区，最后把分区映射到原SDF模型中。通过分析，它有良好的时间和空间复杂度，并且优化后的并行程序有较短的程序执行时间和较高的吞吐率。 综上，本文为嵌入式SDF模型的资源优化提供了完善的优化方法。

同步数据流模型调度序列的空间优化

提出了一种嵌入式DSP系统的存储优化方法。该方法基于同步数据流模型SDF(Synchronous Data Flow)。针对其他优化算法不适用于存在反馈环的同步数据流模型的问题,该方法为反馈环的空间优化设计实现了启发式的调度算法,并提出了将SAS(Single Appearance Schedules)和Non-SAS类型调度序列相结合的层次化的空间优化方案,为同步数据流模型调度序列的空间优化提供一个通用的解决方案。实验结果证实了该方案的有效性。

同步数据流可视化建模和调度优化的实现与研究

数据流是为解决数字信号处理领域应用程序设计、开发难度大等问题而提出的，和传统的控制流相比，数据流能够更加自然地描述信号处理系统，更加清晰地表达系统的并发性。应用数据流设计的信号处理系统具有较高的性能，采用数据流语言能够大大加速DSP应用程序的设计和开发。同步数据流（Synchronous Data Flow，简称SDF）基于数据流，它和数据流最大的不同在于：SDF的计算单元在编译时刻消耗数据和产生数据的数目固定，这一特点决定了采用SDF设计和建立的模型能够在编译时刻确定调度序列，使得SDF适合用于多速率信号处理系统的建模。模型化多处理器系统Modex是一个面向同步数据流的可视化建模系统，它支持开发人员运用SDF进行可视化建模、为建立的模型生成调度序列、对调度序列进行空间优化、为建立的模型生成面向目标平台的C语言代码、对模型仿真验证、对模型进行资源消耗评估、将模型映射到指定的处理器执行等。 本文介绍Modex系统的两个关键技术：可视化建模和SDF调度序列空间优化的实现和研究。文章着重介绍SDF模型描述语言的定义，图元的构建和组织，图元的图形显示，控制图元的控制器，直接操作图元的命令对象，图元和视图之间的同步机制以及图形编辑器等部分的设计和实现。Modex系统的可视化建模基于GEF(Graphical Editing Framework)，为同步数据流建模提供了丰富的可视化编辑操作，同时图元和视图之间松散耦合，可视化建模具有良好的互操作性以及扩展性。文章针对调度序列的空间优化提出了将SAS(Single Appearance Sequence）和非SAS类型调度序列相结合的思想，并基于该思想设计了生成空间优化的非SAS类型调度序列算法IAO（Increase Available Output），并结合EA(Evolutionary Approach)算法实现了面向通用,特别是存在反馈环的SDF模型的空间优化方案SGUTS（Solution for General Graph Using Two Kinds of Sequences）。SGUTS是一个层次化的优化框架，它通过聚集将一个存在反馈的模型分为上层模型和下层模型，然后采用EA、IAO算法分别为这两层模型进行优化，从而得到整个模型的优化结果。SGUTS不仅解决了存在反馈环的SDF模型空间优化问题，而且相比较其他算法，SGUTS取得了较好的优化结果。

JUTA:一个Java自动化单元测试工具

描述了一个Java自动化的单元测试工具JUTA.JUTA首先调用工具Soot解析单个Java方法的源码,并将源码解析成一个控制流图.在此基础上,采用符号执行的方法分析控制流图上的路径.工具能够自动地产生满足覆盖率标准的程序的测试用例.这种方法产生的所有测试用例都是可执行的,并且一般来说具有较小的测试用例数.如果用户能够合理地给出描述程序错误的断言,框架JUTA能够自动地检查源码中部分特定类型的错误.实验结果表明工具对Java单元代码的动态测试和静态测试均能在可接受的时间内给出有效的结果.

incremental learning of triadic plsa for collaborative filtering

Beijing University of Technology (BJUT); Beijing Municipal Lab of Brain Informatics; Chinese Society of Radiology; National Natural Science Foundation of China (NSFC); State Administration of Foreign Experts Affairs

一种五自由度机电测控系统的设计与实现

文章提出了一种基于五自由度机电系统的测控系统设计方法,并从机械系统构成、测控系统结构及软硬件设计方面论述了系统实现技术。针对系统特定的功能要求,文章详细介绍了位姿组合测量和各位姿自由度控制的测控方法,并应用模块化设计和数据流分析方法进行软硬件设计。通过实验进行系统特性分析,得到运行参数指标。实验证明这种设计方法和实现技术合理可行。

载人潜器半物理虚拟仿真系统及其性能分析

为了缩短深海载人潜器控制系统的开发周期,测试系统性能,根据载人潜器的自身特点,构建和实现了载人潜器所有传感器、执行器设备模拟,以及潜器在海底实景的运动状态的半物理虚拟仿真系统。首先,以仿真系统的硬件架构和数据流程为基础,重点介绍了潜器设备模拟的软件体系结构;其次,在仿真平台上测试了各种网络数据的传输延迟,分析了其整体性能;最后,得出了潜器控制器设计的时间约束。现阶段整个系统已经开发成功,并投入运行到潜器控制系统的安装调试中。

胜利勘探数据中心及勘探辅助决策系统研究

In the paper through extensive study and design, the technical plan for establishing the exploration database center is made to combine imported and self developed techniques. By research and repeated experiment a modern database center has been set up with its hardware and network having advanced performance, its system well configured, its data store and management complete, and its data support being fast and direct. Through study on the theory, method and model of decision an exploration decision assistant schema is designed with one decision plan of well location decision support system being evaluated and put into action. 1. Study on the establishment of Shengli exploration database center Research is made on the hardware configuration of the database center including its workstations and all connected hardware and system. The hardware of the database center is formed by connecting workstations, microcomputer workstations, disk arrays, and those equipments used for seismic processing and interpretation. Research on the data store and management includes the analysis of the contents to be managed, data flow, data standard, data QC, data backup and restore policy, optimization of database system. A reasonable data management regulation and workflow is made and the scientific exploration data management system is created. Data load is done by working out a schedule firstly and at last 200 more projects of seismic surveys has been loaded amount to 25TB. 2. Exploration work support system and its application Seismic data processing system support has the following features, automatic extraction of seismic attributes, GIS navigation, data order, extraction of any sized data cube, pseudo huge capacity disk array, standard output exchange format etc. The prestack data can be accessed by the processing system or data can be transferred to other processing system through standard exchange format. For supporting seismic interpretation system the following features exist such as auto scan and store of interpretation result, internal data quality control etc. the interpretation system is connected directly with database center to get real time support of seismic data, formation data and well data. Comprehensive geological study support is done through intranet with the ability to query or display data graphically on the navigation system under some geological constraints. Production management support system is mainly used to collect, analyze and display production data with its core technology on the controlled data collection and creation of multiple standard forms. 3. exploration decision support system design By classification of workflow and data flow of all the exploration stages and study on decision theory and method, target of each decision step, decision model and requirement, three concept models has been formed for the Shengli exploration decision support system including the exploration distribution support system, the well location support system and production management support system. the well location decision support system has passed evaluation and been put into action. 4. Technical advance Hardware and software match with high performance for the database center. By combining parallel computer system, database server, huge capacity ATL, disk array, network and firewall together to create the first exploration database center in China with reasonable configuration, high performance and able to manage the whole data sets of exploration. Huge exploration data management technology is formed where exploration data standards and management regulations are made to guarantee data quality, safety and security. Multifunction query and support system for comprehensive exploration information support. It includes support system for geological study, seismic processing and interpretation and production management. In the system a lot of new database and computer technology have been used to provide real time information support for exploration work. Finally is the design of Shengli exploration decision support system. 5. Application and benefit Data storage has reached the amount of 25TB with thousand of users in Shengli oil field to access data to improve work efficiency multiple times. The technology has also been applied by many other units of SINOPEC. Its application of providing data to a project named Exploration achievements and Evaluation of Favorable Targets in Hekou Area shortened the data preparation period from 30 days to 2 days, enriching data abundance 15 percent and getting information support from the database center perfectly. Its application to provide former processed result for a project named Pre-stack depth migration in Guxi fracture zone reduced the amount of repeated process and shortened work period of one month and improved processing precision and quality, saving capital investment of data processing of 30 million yuan. It application by providing project database automatically in project named Geological and seismic study of southern slope zone of Dongying Sag shortened data preparation time so that researchers have more time to do research, thus to improve interpretation precision and quality.