水质预报系统的MPI+OpenMP并行优化研究

OpenMP(Open Multi-Processing)能够在共享存储环境中获得理想的并行性能,MPI＋OpenMP混合并行编程模型更贴近于多核心集群的体系结构。在多核心集群上进行实验,混合模型相对串行程序可缩短95%的运算时间,加速比超过20。实验结果表明,该混合模型比纯MPI并行的系统具有更优的加速比和扩展性,是对纯MPI并行系统的一种改进。     

On stochastic behaviors of load-sharing parallel systems

Abstract

This paper mainly investigates a general load-sharing parallel system having two units. First, we construct some comparisons among a load standby system, a warm standby system, a hot standby system and a cold standby system. Moreover, some stochastic comparisons between the load-sharing parallel system and one of its two components are obtained in the sense of the usual stochastic order. Finally, the residual life of this system and its properties are examined.     

Robust Bayesian analysis for parallel system with masked data under inverse Weibull lifetime distribution

Abstract

This paper considers the statistical analysis of masked data in a parallel system with inverse Weibull distributed components under type II censoring. Based on Gamma conjugate prior, the Bayesian estimation as well as the hierarchical Bayesian estimation for the parameters and the reliability function of system are obtained by using the Bayesian theory and the hierarchical Bayesian method. Finally, Monte Carlo simulations are provided to compare the performances of the estimates under different masking probabilities and effective sample sizes.     

Incomplete block designs for symmetrical parallel line assays

New series of incomplete block designs for symmetrical parallel lines are proposed. From these designs important contrasts like L_p, L₁ and L′₁ are estimated free from block effects. In addition to these, other odd order contrasts are also estimated orthogonal to block effects. The designs are shown to have simple analysis.     

并行组合扩频系统的抗干扰性能分析

分析了该系统抗单频干扰、窄带干扰及宽带干扰的性能，并与传统扩频系统的抗干扰性能进行了比较，在码片速率与信息速率相同的情况下，并行组合扩频系统的抗干扰性能明显优于传统扩频系统。     

电路测试生成的一种并行处理方法

基于消息传递的多指令流多数据流（ＭＩＭＤ）多机并行系统，提出一种组合电路测试图形生成的新的并行处理方法。该方法首先定义基本门电路的特征函数，通过特征函数的迭加，得到与被测电路对应的一约束网络的特征函数ＣＡＴＰＧ，用遗传算法计算ＣＡＴＰＧ的零点获得故障的测试。结果表明在电路的描述及操作、数据传送、算法实现等方面较其他方法有明显优势     

Markov chain Monte Carlo methods for family trees using a parallel processor

A 1024 CPU parallel computer is used to obtain simulated genotypes in the Tristan da Cunha pedigree using random local updating methods. A four-colour theorem is invoked to justify simultaneous updating. Multiple copies of the program are run simultaneously. These results are used to infer the source of the B allele of the ABO blood group that is present in the population.     

联络及其在现代物理中的解释

本文从曲线坐标、曲面上向量平移入手,导入了联络,继而引入协变微分、短程线及曲率张量,最后指明联络在广义相对论中的意义。     

流分类技术的研究

介绍了流分类算法的概念以及对流分类算法的要求;把目前存在的流分类算法分成三类:多维查找转换为一维查找算法、相关区域查找算法、独立区域查找算法,并对各类算法的性能进行了讨论;通过引入并行流分类算法说明了流分类算法的研究重点是减小存储空间和提高更新速度。     

MUSIC算法在高速并行处理机上的实现

利用FPGA的特点,提出了一种实现MUSIC算法的新的高速并行硬件解决方案。给出了对MUSIC算法进行预处理的结果,为硬件实现加速提供可能;导出了MUSIC算法适合于FPGA实现的并行分解方法;介绍了用FPGA来实现谱峰搜索的新方法。结果表明,该设计方法处理速度理想,能够满足实时性要求。