共查询到20条相似文献，搜索用时 31 毫秒
1.
With the wide acceptance of the computing concept of cluster-based or cluster-of-clusters computing, network-based high
performance computing is gaining increasing popularity among parallel computing community. Current parallel environments such as MPI
(Message Passing Interface) or BSP (Bulk Synchronous Parallel) can not apply readily to heterogeneous computing platform. In order to
exploit the maximum performance potential that a heterogeneous environment brings, we have made some experiments on implementing
BSP in pure Java, a fully platform-independent language. The experimental results show that implementation of parallel libraries in Java is
feasible and the performance is moderately acceptable. The paper describes our experiences in implementation of BSP in pure Java.
Implementation and efficiency issues such as data transport, network communication, support thread, barrier synchronization and dynamic
class loading mechanism are discussed in detail. 相似文献
2.
TANG Zhi-bo LI Guo-Ping XU Wei-min School of Computer Science Engineering. Shanghai University Shanghai China 《上海大学学报(英文版)》2000,(Z1)
1.1 BackgroundThe SU'HPCS (Shanghai Ufllverslty HighPerfor-mance Computing System) Is a 144-processorLlnux cluster constricted from commodity personalcomputer teChnology it is designed tO be dedicated tOhigh performance computing for computationalenglneenng and scientific problems We have nowconstructed many parallel environmentS m SUHPCS,such as MPll'], PVMI"'], BSP14], and DSMI'], as well asHPFI61, tO help users enroll their aPPlications intOParallel comPUting system But o… 相似文献
3.
Parallel virtual machine(PVM) has been a prevalent parallel computing environment for heterogeneous systems. The generic
portion of PVM is implemented on the basis of TCP/IP in result of high cost and low performance in communication. This paper
describes the design and implementation of GMPVM(parallel virtual machine based on Myrinet message passing system), a PVM
portion based on Myrinet's GM API(message passing system on Myrinet). The goal of this work is to make the maximum use of the high
bandwidth and low latency of Myrinet; to provide PVM portion with a user-level, light weighted communication subsystem. GMPVM is
developed on ZiQiang-2000 (SUHPCS). The difference between PVM and GMPVM is figured out here, and also some optimization
technique discussed. 相似文献
4.
Cell BE为一个异构多核并行处理架构,具有超高速通信能力,能够有效地提供超级计算机的性能。在分析与研究Cell BE的硬件架构、并行编程环境与模型的基础上,选取PS3为主机,构建一个CellBE高性能计算实验平台。通过选取相关的硬件设备,在构建好硬件平台的基础上,构建系统与软件平台。通过在实验平台上运行高性能计算应用程序,来对平台进行实验测试。以求解多体问题的计算为平台实验应用程序,将四叉树计算初始化过程算法在Cell BE实验平台上实现,并进行运行测试。在不同的粒子数实验规模下的计算时间结果表明,在Cell BE平台上获得了较高的对CPU的加速效果。 相似文献
5.
通过分析网格技术在校园网建设中的应用现状和OA系统的功能,利用教育网关键网络节点,构建了一种基于网格技术的OA系统扩展应用平台,以期对网格技术的推广和高校的信息化、数字化建设起到积极的促进作用。 相似文献
6.
胡师彦 《湖南城市学院学报》2001,18(6):57-60
介绍了并行计算的基本概念,利用局域网和PVM构成并行计算环境,实现了对电荷量变化的并行计算.改进了传统的隐式龙格库塔方法,在实际计算中取得良好效果.最后讨论了与并行计算效率有关的因素. 相似文献
7.
随着并行计算的发展,MPI作为一种重要的消息传递方式标准,被广泛地应用到科学计算中。将MPI技术与电磁场时域有限差分法结合,构建了用于高性能并行计算的PC集群环境,通过算例验证了这套系统具有强大的计算功能、准确的计算精度和很高的计算效率。 相似文献
8.
DONG Jing-yi DING Jun MENG Rui TONG Wei-qin School of Computer Engineering Science Shanghai University Shanghai China 《上海大学学报(英文版)》2000,(Z1)
1 Introduction1.1 MOderationWith the development of netWorldng technology,cluster-based Supercomputer is widely accepted becauseof its high ratio of perfonnance/cost in the sense of peakpeifonnance. And SMP (shared memory Processor)nodes will improve the overall pdrirmance. OneaPProach to Programming SMP clusterS is tO use acombination of shared memory pro~g at the levelof each SMP unit and explicit message passing betweenSMPs. This mixed-mode style of coding would requirethe Progral… 相似文献
9.
10.
Using commodity SMPs (shared memory processors) to build cluster-based supercomputer has become a mainstream trend. Yet
programming this kind of supercomputer system requires an environment support both message passing and shared memory programming.
This paper describes our preliminary work in an effort to target BSP library for cluster of SMPs. In order to exploit the maximum
performance potential that a cluster of SMPs brings, we adopt thread technique to reduce system overhead and to exploit the capacity of
SMPs. A three-layer synchronization mechanism is proposed to support barrier synchronization within an SMP node, a group of SMP nodes
and the whole cluster respectively. A comparison is made between our BSP library and the currently available BSP libraries such as PUB. 相似文献
11.
BP神经网络是在数据挖掘、语音识别和文本语言转换等领域最为广泛使用的网络之一。但也有其不足之处，主要表现在收敛速度慢和学习时间长。并行是解决这一问题的可行方案。本文提出了一种利用微机机群来实现并行处理，在并行编程环境PVM中实现BP神经网络的并行学习算法。即主要采用将大量计算由主控节点转移到从结点上、设置逻辑锁和粗粒度任务划分等方法以减少计算和通信时间，有效实现学习过程的并行处理。在微机机群系统下的PVM环境中实现该算法。结果表明改进后的算法不仅大大减少了学习时间，提高了学习效率，而且能够很好面对大规模BP神经网络的学习。 相似文献
12.
随着INTERNET的进一步发展,分布式应用的开发越来越受到人们的重视.而现有的几种分布计算技术的不足也逐渐显露.新的分布计算技术-基于Mobile Agent的分布计算的出现给人们带来了新的发展.本文以三菱公司基于Java语言的Mobile Agent系统Concordia为原型,讨论了其体系结构及其Agent的迁移实现问题.* 相似文献
13.
As a parallel programming model, Map-Reduce is used for distributed computing of massive data. Map-Reduce model encapsulates
the details of parallel implementation, fault-tolerant processing, local computing and load balancing, etc., provides a simple but powerful interface. In case of having no clear idea about distributed and parallel programming, this
interface can be utilized to save development time. This paper introduces the method of using Hadoop, the open-source Map-Reduce
software platform, to combine PCs to carry out scalable parallel computing. Our experiment using 12 PCs to compute N-body problem based on Map-Reduce model shows that we can get a 9.8x speedup ratio. This work indicates that the Map-Reduce can be applied in scalable parallel computing. 相似文献
14.
Java Web 是计算机专业核心实践课程。传统课程管理方法需要教师手动逐个部署、运行并检查学生项目程序,效率极低且极易发生错误,实现 Java Web 课程项目自动管理和分析势在必行。近年来飞速发展的在线教育、云计算及程序在线评判系统为该类需要复杂分布式部署和运行环境的课程项目管理提供了新思路。因此,设计并实现一个基于云环境的 Java Web 课程管理系统。该系统可实现学生提交课程项目的云端在线自动部署和运行,解决其中负载均衡、命名冲突等问题,并提供项目相似性检测等分析功能。与人工评阅方法相比,该系统可节省至少 57%的评阅时间,同时可提供准确性达 97%的项目相似性检测功能,大幅提高 Java Web课程管理自动化程度和教师工作效率。 相似文献
15.
In this paper, we propose a parallel computing technique for content-based image retrieval （CBIR） system. This technique is mainly used for single node with multi-core processor, which is different from those based on cluster or network computing architecture. Due to its specific applications （such as medical image processing） and the harsh terms of hardware resource requirement, the CBIR system has been prevented from being widely used. With the increasing volume of the image database, the widespread use of multi-core processors, and the requirement of the retrieval accuracy and speed, we need to achieve a retrieval strategy which is based on multi-core processor to make the retrieval faster and more convenient than before. Experimental results demonstrate that this parallel architecture can significantly improve the performance of retrieval system. In addition, we also propose an efficient parallel technique with the combinations of the cluster and the multi-core techniques, which is supposed to gear to the new trend of the cloud computing. 相似文献
16.
This paper proposes a design optimization method for the multi-objective orbit design of earth observation satellites, for which the optimality of orbit performance indices with different units, such as: total coverage time, the frequency of coverage, average time per coverage and maximum coverage gap, etc. is required simultaneously. By introducing index normalization method to convert performance indices into dimensionless variables within the range of [0, 1], a design optimization method based on the principal component analysis and cluster analysis is proposed, which consists of index normalization method, principal component analysis, multiple-level cluster analysis and weighted evaluation method. The results of orbit optimization for earth observation satellites show that the optimal orbit can be obtained by using the proposed method. The principal component analysis can reduce the total number of indices with a non-independent relationship to save computing time. Similarly, the multiple-level cluster analysis with parallel computing could save computing time. 相似文献
17.
在高性能并行分布式网格计算环境下,计算节点具有动态、异构、分布和多管理域特性,这使得网格资源的发现及管理变得比较困难。在对网格调度模型分析的基础上,使用模糊聚类的相关理论提出在任务调度之前对资源进行预处理,依据用户QoS需求权重对机器性能进行聚类,得到和用户QoS需求最匹配的较优资源子集。这样既合理地减少了映射的目标资源,又实现了目标资源子集与用户QoS需求的最优匹配。在一定程度上提高了整个系统的性能,满足了不同用户的特殊需求。 相似文献
18.
Taisir Subhi 《High Ability Studies》1999,10(1):69-84
One of the prerequisites for acceptance and implementation of computers in an educational system is a positive attitude of both teachers and students toward their use. This research, therefore, focuses on differences in attitudes toward computers between male and female teachers, female and male gifted students, and teachers and gifted students. Samples consisted of 217 gifted students (115 males and 102 females) and 125 teachers (30 males and 95 females) in the Amman Educational District, Jordan. Computer attitudes were measured by means of the Attitudes of Both Teachers and Students Toward Computers Questionnaire, and compared statistically to variables such as performance in mathematics, IQ, gender, teachers’ experience in teaching and computing, and teachers’ training. Results showed that participants are generally positive towards computers, and there were no significant differences in attitude between gifted students and teachers or between males and females. Attitudes towards computers, however, may be predicted by performance in mathematics and by IQ scores. 相似文献
19.
本文提出了在Ｔｒａｎｓｐｕｔｅｒ网络上计算均匀双３次Ｂ样条曲面的一种并行实现．这个工作对构造基于Ｔｒａｎｓｐｕｔｅｒ上的ＣＡＤ和图形系统是有效的 相似文献
20.
Dynamic distribution model is one of the best schemes for parallel volume rendering. However, in homogeneous cluster system, since the granularity is traditionally identical, all processors communicate almost simultaneously and computation load may lose balance. Due to problems above, a dynamic distribution model with prime granularity for parallel computing is presented. Granularities of each processor are relatively prime, and related theories are introduced, A high parallel performance can be achieved by minimizing network competition and using a load balancing strategy that ensures all processors finish almost simultaneously, Based on Master-Slave-Gleaner （ MSG） scheme, the parallel Splatting Algorithm for volume rendering is used to test the model on IBM Cluster 1350 system. The experimental results show that the model can bring a considerable improvement in performance, including computation efficiency, total execution time, speed, and load balancing. 相似文献