摘要:

This paper describes a method that allows the speed up of parallel processes in distributed arbitration schemes as used in Futurebus + . It is based on special arbitration codes that decrease the maximal arbitration time to a specified value. Such codes can be applied with few, if any, minor changes of the hardware. The general structure of these codes is given.

ISSN:1063-7192    

DOI:10.1080/10637199508915494

出版商:Taylor & Francis Group    

年代:1995

数据来源: Taylor

摘要:

Simple parallel algorithm for fixed-dimensional Linear Programming is developed. It provides a straightforward implementation on MIMD architectures with high efficiency. The implementation on transputer arrays is described and numerical results of performance are presented.

ISSN:1063-7192    

DOI:10.1080/10637199508915495

出版商:Taylor & Francis Group    

年代:1995

数据来源: Taylor

摘要:

In this paper we extend thearithmetic mean methodfor large, sparse systems of linear equations to the case where the coefficient matrix is a singular, irreducibleM-matrix. Matrices of this kind arise in the computation of the stationary distribution vector for an irreducible Markov chain, as well as in other applications. We report on some numerical experiments on a simple reliability model, where two forms of the arithmetic mean method are compared. The method is well-suited for parallel implementation on a multi-processor.

ISSN:1063-7192    

DOI:10.1080/10637199508915496

出版商:Taylor & Francis Group    

年代:1995

数据来源: Taylor

摘要:

In this paper, we present a new algorithm based on Modified Gram-Schmidt (MGS) orthogonal matrix factorization for parallel solution of linear equations Ax = b. Unlike the existing methods, the proposed algorithm using a new technique called two-sided elimination unifies both the triangularization and back-substitution phases to produce the complete solution vector x. The new algorithm replaces the back-substitution phase in the existing methods, which requires O(N) steps using O(N) processors or O(log22N) steps using O(N3) processors, by only one step division. Being based on the MGS method, the new algorithm is numerically stable. Finally, we study the performance of the algorithm on hypercube multiprocessor systems.

ISSN:1063-7192    

DOI:10.1080/10637199508915497

出版商:Taylor & Francis Group    

年代:1995

数据来源: Taylor

摘要:

Time dependent partial differential equations are often solved using algorithms which parallelize the solution process in the spatial domain. However, as the number of processors increases, the parallel efficiency is limited by the increasing communication/computation ratio. Recently, several researchers have proposed algorithms incorporating time domain parallelism in order to increase efficiency. In this paper we discuss a class of such algorithms and analyze it rigorously.

ISSN:1063-7192    

DOI:10.1080/10637199508915498

出版商:Taylor & Francis Group    

年代:1995

数据来源: Taylor

摘要:

In this paper, a divide-and-inner product parallel algorithm for evaluating a polynomial of degreeN (N + 1 = KL)on a MIMD computer is presented. It needs 2K + log2Lsteps to evaluate a polynomial of degreeNin parallel onL + 1 processors(L ≤ 2K− 2log2K)which is a decrease of log2Lsteps as compared with theL-order Homer's method [1], and which is a decrease of (2log2L)1/2steps as compared with the some MIMD algorithms [3, 4]. The new algorithm is simple in structure and easy to be realized.

ISSN:1063-7192    

DOI:10.1080/10637199508915499

出版商:Taylor & Francis Group    

年代:1995

数据来源: Taylor

摘要:

As an alternative to Choleski factorisation, in 1965 Martin et al. [3] suggested a root-free symmetric decomposition for a positive definite matrix. Since around 1970, root-free form has been favoured to Choleski form. In the present paper, by introducing symmetric elimination transformations we first obtain an elimination-variant of the root-free symmetric decomposition method of Martin et al. We then extend the idea lo partitioned systems. By introducing (block) symmetric elimination transformations we obtain a root-free parallel symmetric elimination algorithm for positive definite linear systems. With the given system of sizeN, partitioned intop2blocks, we discuss an implementation of the parallel algorithm; usingpprocessors, which exhibits good load-balancing in view of symmetry of the updates across the partitioning blocks. The parallel algorithm has an operations count ofO(N3/3p), indicating that the algorithm could achieveefficiencyclose to 1 if implemented on ap-processor machine.

ISSN:1063-7192    

DOI:10.1080/10637199508915500

出版商:Taylor & Francis Group    

年代:1995

数据来源: Taylor

摘要:

This paper is concerned with the numerical solution of quasi-linear singularly perturbed boundary value problems. We assume that the solution exhibits one boundary layer and no turning points. Recent results about this class of problems are used to define a numerical method based on piecewise-uniform meshes. This method is naturally suitable for parallel implementation and several decompositions of the computational work into independent processes are possible. Numerical results obtained on a network of transputers T805-20 show the reliability and the efficiency of the proposed method.

ISSN:1063-7192    

DOI:10.1080/10637199508915501

出版商:Taylor & Francis Group    

年代:1995

数据来源: Taylor