Downloads & Free Reading Options - Results

An investigation is conducted into the viability of using a generalized Conjugate Gradient-like method as an iterative solver to obtain steady-state solutions of very low-speed fluid flow problems. Low-speed flow at Mach 0.1 over a backward-facing step is chosen as a representative test problem. The unsteady form of the two dimensional, compressible Navier-Stokes equations is integrated in time using discrete time-steps. The Navier-Stokes equations are cast in an implicit, upwind finite-volume, flux split formulation. The new iterative solver is used to solve a linear system of equations at each step of the time-integration. Preconditioning techniques are used with the new solver to enhance the stability and convergence rate of the solver and are found to be critical to the overall success of the solver. A study of various preconditioners reveals that a preconditioner based on the Lower-Upper Successive Symmetric Over-Relaxation iterative scheme is more efficient than a preconditioner based on Incomplete L-U factorizations of the iteration matrix. The performance of the new preconditioned solver is compared with a conventional Line Gauss-Seidel Relaxation (LGSR) solver. Overall speed-up factors of 28 (in terms of global time-steps required to converge to a steady-state solution) and 20 (in terms of total CPU time on one processor of a CRAY-YMP) are found in favor of the new preconditioned solver, when compared with the LGSR solver.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

Preconditioned conjugate gradient methods for solving sparse symmetric and positive finite systems of linear equations are described. Necessary and sufficient conditions are given for when these preconditioners can be used and an analysis of their effectiveness is given. Efficient computer implementations of these methods are discussed and results on the CYBER 203 and the Finite Element Machine under construction at NASA Langley Research Center are included.

The authors considers large, sparse linear systems which result from the discretization of partial differential equations on regular and irregular domains, and he focuses on the application of the preconditioned conjugate gradient (PCCG) method to the solution of such systems. More specifically, the goal of this paper is the efficient implementation of the PCCG method on vector supercomputers. The contribution to the above goal is made by 1) the introduction of a data structure which way be effectively manipulated on vector machines, 2) the utilization of preconditioning matricies which are obtained by incomplete factorization with diagonal update sets, and 3) the introduction of new numbering schemes for both regular and irregular grids. (Author)

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

The solution of nonlinear, transient finite element problems may be achieved by using step-by-step integration of the equations of motion combined with a Newton solution of the resulting nonlinear algebraic equations. The use of Newton type methods leads to a set of linear simultaneous algebraic equations whose solution gives the iterate. For very large problems the solution of the large set of linearized equations may be a formidable task - often consuming more than half of the computing effort when performed by a direct method based upon Gauss elimination. Accordingly, it is of considerable importance to investigate alternative methods to solve the problem. The present study presents results obtained by using a Preconditioned Conjugate Gradient Method (PCG) described in (7) and a Preconditioned Lanczos Method (PLM) described in (6) to solve a variety of numerical examples. Based upon results obtained it is evident that a significant reduction in overall effort, compared to direct solutions, may be achieved using the preconditioned methods.

The preconditioned conjugate gradient (PCG) method is an effective means for solving systems of linear equations where the coefficient matrix is symmetric and positive definite. The incomplete LDL(t) factorizations are a widely used class of preconditionings, including the SSOR, Dupont-Kendall-Rachford, Generalized SSOR, ICCG(0), and MICCG(0) preconditionings. The efficient implementation of PCG with a preconditioning from this class is discussed. (Author)

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

We describe the basis of a matrix ordering heuristic for improving the incomplete factorization used in preconditioned conjugate gradient techniques applied to anisotropic PDE's. Several new matrix ordering techniques, derived from well-known algorithms in combinatorial graph theory, which attempt to implement this heuristic, are described. These ordering techniques are tested against a number of matrices arising from linear anisotropic PDE's, and compared with other matrix ordering techniques. A variation of RCM is shown to generally improve the quality of incomplete factorization preconditioners.

In this paper, we present a class of iterative descent methods for solving large, sparse, nonsymmetric systems of linear equations whose coefficient matrices have positive-definite symmetric parts. Such problems commonly arise from the discretization of non-self-adjoint elliptic partial differential equations. The methods we consider are modelled after the conjugate gradient method. They require no estimation of parameters and their rate of convergence appears to depend on the spectrum of A rather than ATA. Their convergence can also be accelerated by preconditioning techniques.

This document considers a few different implementations of classical iterative methods on parallel processors with the purpose of studying how multiprocessor architecture affects performance. The framework is that of general nonsymmetric linear systems that arise from the discretization of partial differential equations and the authors concentrate on the solution methods based GMRES, a conjugate gradient-like method, combined with well-known preconditionings. The computer architectures considered are shared memory machines and loosely coupled linear or mesh connected arrays. Keywords: GMRES(A Generalized Minimal Residual Algorithm).