摘要:

This is the first in a series of papers about asynchronous parallel algorithms based on domain decomposition method (DDM) for solving partial differential equations on MIMD machines, especially on CM-5. These algorithms are all called Schwarz-Chaotic Over Relaxation (S-COR), the generalization of the Schwarz Alternating Method. In this paper, the linear elliptic partial differential equations is discussed. Numerical results are presented to demonstrate the effectiveness of these decomposition algorithms on the Thinking Machines CM-5 parallel supercomputer.

ISSN:1063-7192    

DOI:10.1080/10637199608915572

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

In this paper, a new monotone-Schwarz parallel algorithm for solving a class of semilinear elliptic systems is proposed. In the case of overlapping subdomains, the detailed procedures for constructing iterative sequences and the convergence proofs are investigated.

ISSN:1063-7192    

DOI:10.1080/10637199608915573

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

The concept of localization strategy for massively parallel algorithm is presented in the paper, A massively parallel algorithm of the numerical solution of some tridiagonal systems is considered by analyzing the possibility of localization computing. The algorithm can be used in the practical computation of cubic spline approximation as well as for solving some finite difference equations.

ISSN:1063-7192    

DOI:10.1080/10637199608915574

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

Multigrid methods combine a number of standard sparse matrix techniques. Usual implementations separate the individual components (e.g., an iterative methods, residual computation, and interpolation between grids) into nicely structured routines. However, many computers today employ quite sophisticated and potentially large caches whose correct use are instrumental in gaining much of the peak performance of the processors.

ISSN:1063-7192    

DOI:10.1080/10637199608915575

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

One of the most intellectual steps in compiling for distributed memory parallel machines is to determine a suitable data partitioning scheme for a particular program. Most of the parallelizing compilers for these machines provide no or little support to the user in this difficult task. We have developed DPART, an automatic data partitioning system for Fortran 77 procedures. This paper describes the partitioning strategics of alignment, distribution, and processor layout in DPART. Finally we present experimental results for TRED2, DGEFA, and JACOBI procedures to demonstrate the effectiveness of this system.

ISSN:1063-7192    

DOI:10.1080/10637199608915576

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

High Performance Fortran is a language designed to support efficient data parallel programming on a variety of parallel machines. This kind of parallel programming has been proven to be very user-friendly, easy to debug and easy to use. In this programming model, the programmer explicitly specifies the layout of data in a global space, relying on a compiler to generate a parallel program including all the communication. While this frees the programmers from the tedium of thinking about local name spaces and message-passing, no assistance is provided in determining an efficient data layout scheme on the target machine.

ISSN:1063-7192    

DOI:10.1080/10637199608915577

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

Nondeterminacy is a main obstacle of parallel debugging. Current debugging strategies either detect non-dcterminacy separately, or control the execution in a two-phase manner. Here we present a strategy called “state freezing”. Based on the strategy, one-phase (we call one-line) debugging of parallel programs is available. Basic algorithms and an example are given in the paper.

ISSN:1063-7192    

DOI:10.1080/10637199608915578

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

Shape optimization based electronic prototyping is part of any future scenario for intelligent manufacturing. Unfortunately, the resulting optimization problems can be prohibitively large. Moreover, it is known that most of the optimization problems belong to the class of "hard" problems. Thus, seeking parallel methods for their solution is well justified. The idea of divide and conquer is already used to approximate the solution of large optimization problems sequentially. This suggests that it might be more efficient to decouple the system into several smaller subsystems, optimize them locally and in parallel, and then approximate the global solution by coordinating some form of global optimization on the interfaces of the subsystems. This approach is referred to as a two-level scheme. In general, an optimization problem involving many variables and constraints cannot be decomposed into independent subproblems that can be independently optimized. However, the above described problem decomposition approach yields good approximations to the global minimum while allowing the parallel application of shape optimization on local subsystems. The effectiveness of the two-level scheme comes from the inherent parallelism in modeling physical objects. The analysis and shape optimization are implemented using the parallel mesh and mesh splitting tool and its algorithmic infrastructure. For the shape optimization problem, we are developing two-level semi-optimal algorithms based on local and global mesh and decomposition data.

ISSN:1063-7192    

DOI:10.1080/10637199608915579

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

A new class of methods of building lattice models have been developed recently in numerical simulation of fluid dynamics. Their essential idea is model-rebuilding and direct simulating of mathematical physical problems based on the kinetic description of physical systems. Following the work in fluid dynamics, extensions of the methods to some typical mathematical physical equations have been made. In this paper, a theoretical framework for building lattice models of general partial differential equations is proposed.

ISSN:1063-7192    

DOI:10.1080/10637199608915580

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

In this paper we present benchmark results from the parallel implementation of the three-dimensional Navier-Stokes solverPrism[1] on the Cray T3D, We compare the single processor performance with other Cray computers, namely the Cray C90, J90 and EL98, as well as Digital Equipment Corporation's DEC 3000/500 (which uses the same processor as the T3D) and AlphaServer 8400 5/300 (which uses the same processor as the Cray T3E, the successor system to the T3D). The numerical method used in the solver is based on mixed spectral element-Fourier expansions in (x−y) andz-directions, respectively. Each (or a group) of Fourier modes can be computed on a separate processor as the linear contributions (Helmholtz solves) in Navier-Stokes equations are completely uncoupled. Coupling is obtained via the nonlinear contributions (adveaion terms) and requires a global transpose of the data and one-dimensional multiple-point FFTs.

ISSN:1063-7192    

DOI:10.1080/10637199608915581

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor