一种新的非冯·诺依曼计算机体系结构TriBA

针对复杂问题的分解方式,提出一种基三计算机体系结构,该体系结构易于扩展,具有分形特征.设计了一种面向对象多核CPU,支持对象并行运行,在一定程度上实现了软件结构与计算机系统结构的统一.片上多核间采用基三分层互连网络互连,连接简单,体现计算局部性特点,易于硬件实现. 基于该互连方式讨论了面向对象多核CPU上的4种基本软件流水模型.     

Realistic efficiency evaluations for parallel computations under workstation cluster

In recent years high performance scientific computing under workstation cluster connected by local area network is becoming a hot point. Owing to both the longer latency and the higher overhead for protocol processing compared with the powerful single workstation capacity, it is becoming severe important to keep balance not only for numerical load but also for communication load, and to overlap communications with computations while parallel computing. Hence, our efficiency evaluation rules must discover these capacities of a given parallel algorithm in order to optimize, the existed algorithm to attain its highest parallel efficiency. The traditional efficiency evaluation rules can not succeed in this work any more. Fortunately, thanks to Culler's detail discuss in LogP model about interconnection networks for MPP systems, we present a system of efficiency evaluation rules for parallel computations under workstation cluster with PVm3.0 parallel software framework in this paper. These rules can satisfy above acquirements successfully. At last, two typical synchronous and asynchronous applications are designed to verify the validity of these rules under 4 SGIs workstations cluster connected by Ethernet.     

Image Edge Detection Based on Oscillation

IntroductionArtificial neural networks can be dividedintotwo classes :rate code neural networks and pulse-based neural networks .Inrate code neural networks the outputs of the neurons arecontinuous , whereas in pulse-based neural networks theneurons emit pulses . Although rate code neural models havebeen well studied, researchers continue to find the evidencerecently that in many cases the brain may accomplishcomputations on pulses(action potential) . Great effort hasbeen devoted towards the a…     

基于ARM和μC/OS-Ⅱ以太网接口系统设计

TCP／IP及互联网技术在工业领域得到广泛应用,然而大量工业现场设备尚不具备网络接口,主要通过串行口进行通讯,因而无法直接连接到Internet。笔者设计了一种基于ARM处理器的嵌入式以太网接口系统,用来解决这些传统设备与现今的网络设备之间的互联问题。     

通信时延下多智能体系统的静态一致性研究

假设每个个体接收信息产生的时延相同,且处理自身信息产生的时延不同,针对通信时延下具有有向对称通信网络的多智能体系统,提出了一个自时延比例-微分控制协议,使系统的状态最终趋于一致.基于频域分析法和奈氏判据,给出了系统状态收敛到一致的充分条件,并计算出了系统静态一致时平衡状态的具体表达式.在Matlab环境下进行了仿真实验,其结果验证了理论的正确性和有效性.     

用多级衍射全息光栅实现神经网络的光学互连

用Mach Zehnder干涉仪分两步记录制作了具有多级衍射的全息光栅 .再利用其多级衍射效应 ,实现了 1 D神经网络的光学互连 .结果表明 ,应用该方法不仅能得到满意的光学互连 ,且具有简单易行的优点 .     

P2P网络中资源搜索机制的研究

基于当前Gnutella网络统计特性的分析,重点分析了三个影响搜索性能的因素:超级结点,历史搜索记录以及延迟特性.分析结果表明上述三个指标对于设计一个新的搜索策略非常重要,它们构成了判断节点是否转发搜索请求的重要指标.在据此判断指标所提出的新型搜索机制中,每个节点就可以在选择邻居转发搜索请求时更加有目的性,搜索成功率会更高.实验仿真数据表明,与基于洪泛策略和基于随机行走策略的传统搜索机制相比较,此搜索策略在保持高搜索成功率的前提下能够大大减少消息通信量,从而减少系统开销,提高系统效率.     

A Research on Internal Mechanism of Evaluation System

Evaluation is that a subject evaluates an object with certain value criterion, which is the premise and foundation of decision. The evaluating problem is becoming one of the popular issues in academe. In this paper, the essential meaning of evaluation is defined firstly, and evaluation process analyzed: putting forward the problem of evaluation; defining evaluation subject, evaluation organizer and evaluation criterion and method; processing and feeding back evaluation results. According to the essential meaning and process, the evaluation support system (ESS) is established, consisting of four function modules: problem definition module, data extraction module,mold database module and man-machine communication module. Based on the complexity of factual problem, the group evaluation support system(GESS) is established, which is an integration system of many valuators‘ wisdom, experience and corresponding ESS. GESS is used to support group valuators to solve semi-construction evaluation problem on the basis of computer and network in order to acquire scientificity and objectivity in the evaluation of big system.     

Efficient statistical analysis method of power/ground （P/G） network

In this paper, we propose an incremental statistical analysis method with complexity reduction as a pre-process for on-chip power/ ground （P/G） networks. The new method exploits locality of P/G network analyses and aims at P/G networks with a large number of strongly connected subcircuits （called strong connects） such as trees and chains. The method consists of three steps. First it compresses P/ G circuits by removing strong connects. As a result, current variations （CVs） of nodes in strong connects are transferred to some remaining nodes. Then based on the locality of power grid voltage responses to its current inputs, it efficiently calculates the correlative resistor （CR） matrix in a local way to directly compute the voltage variations by using small parts of the remaining circuit. Last it statistically recovers voltage variations of the suppressed nodes inside strong connects. This new method for statistically compressing and expanding strong connects in terms of current or voltage variations in a closed form is very efficient owning to its property of incremental analysis. Experimental results demonstrate that the method can efficiently compute low-bounds of voltage variations for P/G networks and it has two or three orders of magnitudes speedup over the traditional Monte-Carlo-based simulation method, with only 2.0% accuracy loss.     

Efficient statistical analysis method of power/ground(P/G) network

In this paper, we propose an incremental statistical analysis method with complexity reduction as a pre-process for on-chip power/ground (P/G) networks. The new method exploits locality of P/G network analyses and aims at P/G networks with a large number of strongly connected subcircuits(called strong connects) such as trees and chains. The method consists of three steps. First it compresses P/G circuits by removing strong connects. As a result, current variations (CV) of nodes in strong connects are transferred to some remaining nodes. Then based on the locality of power grid voltage responses to its current inputs, it efficiently calculates the correlative resistor (CR) matrix in a local way to directly compute the voltage variations by using small parts of the remaining circuit. Last it statistically recovers voltage variations of the suppressed nodes inside strong connects. This new method for statistically compressing and expanding strong connects in terms of current or voltage variations in a closed form is very efficient owning to its property of incremental analysis. Experimental results demonstrate that the method can efficiently compute low-bounds of voltage variations for P/G networks and it has two or three orders of magnitudes speedup over the traditional Monte-Carlo-based simulation method, with only 2.0% accuracy loss.     

基于自组织特征映射网络的综合评价模型

首先根据人脑思维的模糊性特点 ,以模糊集合论为基础 ,构造一种用于指导自组织特征映射网络学习过程的模糊熵准则 ,可以在较大范围内有效地解决自组织特征映射网络 (SOFM )的学习问题 .然后提出一种基于模糊自组织特征映射网络的综合评价模型 .该模型通过确定标准对象 ,数据的标准化处理 ,网络自适应学习和评价结果输出等环节 ,可以有效地解决一类综合评价问题 .最后通过一个实例进一步说明其灵活性与实用性 .参 6 .     

Systolic—专用硬件的通用设计方法

为了获得较高的处理速度,我们经常需要为一个计算机系统设计一些专用的高速并行处理器。Systolic结构是一种结构规整、控制简单的并行结构。它是由一组简单的处理单元构成一个网络。网络中的节点是处理单元,节点的互连是规则的,网络中数据有节奏地流动。Systolic阵列中内部节点的结构是一样的,只有少量的与外界有I/O操作的边界节点是特殊的。Systolic结构的专用硬件可以提高速度而保持专用硬件处理速度与I/O带宽的平衡。由于Systolic结构的规整、简单的结构,它很适合VLSI实现。     

基于多个 DSP 的图像处理 SIMD 并行结构

通过分析低层图像处理算法特点，指出标准ＳＩＭＤ结构不能完成某些低层图像处理算法的数据综合任务，提出的改进型ＳＩＭＤ结构增加了多指令控制、共享存储器单元，能执行更多种类图像处理算法，讨论了图像在ＳＩＭＤ结构计算机上的一维、二维分块方法，结果表明一维图像分块在互连网络硬件量、算法执行速度两方面更具优越性．     

基于紧致编码的高阶神经网络光计算

提出一种能压缩三阶神经网络的互连张量规模近一半的紧致编码方法。设计两种光电混和系统-非相干多重成象系统和非相干相关系统,并用它实现了三阶互连的二维神经网络,构成图象平方关联存贮器。在光学实现中,借助于一种多重矩阵排列方式,作者用一块二维掩模板表示六维的互连张量,用二维的发光二极管(LED)阵列表示四维的输入张量,实验结果与数值模拟的结果吻合。     

具有不同通信时延的多个体系统的一致性

针对具有不同通信时延的多个体系统的一致性问题,提出了一种引入自身时延的一致性协议.根据广义Nyquist判据与Gerschgorin圆盘定理,得到了系统渐近收敛一致的充分条件.在多个体系统的有向连接拓扑含有一个全局可达节点的条件下,一致收敛的充分条件仅与个体的自身时延、以及个体与邻居个体的连接权值有关.在提出的协议中,引入自身时延能够加速具有通信时延的多个体系统的收敛速度,而通信时延并不影响系统的收敛,但其可以延长收敛时间.此外,得到了系统最终收敛的一致状态的求解方法.仿真结果证明了结论的正确性.     

大规模并行处理机的互连网络初探

介绍了互连网络在大规模并行处理机中的重要地位、发展特点 ,分析了各种互连网络的性能 ,最后提出了互连网络有待开发的问题 .     

水驱砂岩油藏孔喉结构变化的三维网络模拟

以长期水驱实验为基础,建立了等效水驱砂岩储层孔喉结构变化的三维网络模拟模型,结合三维微粒运移机制和有限差分求解方法,得到了长期水驱砂岩油藏孔喉结构变化规律：（1） 冲刷后喉道半径呈增加趋势,喉道半径变化范围变大,极小喉道半径呈微弱减小趋势;（2） 孔隙网络模型中冲刷半径扩大的孔道分布形式与原始孔隙网络结构密切相关,并非所有的大孔道都串联起来贯穿岩芯孔隙网络的两个端面,但入口端和出口端部分大孔道相互连通,形成端面上的大孔道网络群。网络模拟注水结果结合采油井测试,可为注水剖面的调整提供更加可靠的依据。     

Linux下以太网的IPv6隧道技术的实现

IPv6是一种新的互联网协议的规范,由于与IPv4不直接兼容,限制了其应用和发展.为此,提出一种在Linux下网络层隧道技术的配置与实现方案,该方案的实现使得各个IPv6的信息孤岛能通过IPv4网络进行通信,解决了IPv6到IPv4的转换问题.     

直连网络在太比特路由器中的应用

概述了直连网络的性能参数及常见拓扑结构，重点研究了太比特路由器交换网络设计中的2项关键技术：交换机制和路由算法。分析表明虚切通比其他交换机制更适合应用于大容量路由器中。防死锁、增加容错性以及保证服务质量是路由算法设计中值得注意的几个关键问题。在此基础上提出了一种新的适用于太比特路由器的拓扑结构XD网络，该网络与现有的几种拓扑相比在扩展性、容错能力以及实现上具有优势。研究表明XD网络是一种极具潜力的太比特路由器交换网络候选拓扑。