Accession Number : ADA290615

Title :   Linear Algebraic Computation on Distributed Memory Parallel Machines.

Descriptive Note : Final rept. 1 Feb 91-30 Jun 94,

Corporate Author : YALE UNIV NEW HAVEN CT DEPT OF COMPUTER SCIENCE

Personal Author(s) : Eisenstat, Stanley C.

PDF Url : ADA290615

Report Date : SEP 1994

Pagination or Media Count : 4

Abstract : The project entailed investigating several problems in the parallel solution of sparse systems of linear equations and eigenproblems, including: algorithms for factoring sparse matrices; techniques for the solution of sparse triangular systems; iterative methods for sparse systems, focusing mainly on preconditioning techniques for conjugate-direction methods; the solution of the symmetric tridiagonal eigenproblem. While this may seem to be an eclectic group of topics, there are, in fact, close relationships among them. As one example, a common technique for preconditioning iterative methods depends crucially on efficient solution of triangular systems. As another, it should be possible to construct an effective Lanczos-type algorithm for sparse, symmetric eigenproblems by combining the techniques required for conjugate-direction iterations for linear systems with those required for the solution of symmetric tridiagonal eigenproblems. (AN)

Descriptors :   *ALGORITHMS, *PARALLEL PROCESSING, *MATHEMATICAL PROGRAMMING, COMPUTATIONS, DISTRIBUTED DATA PROCESSING, MATRICES(MATHEMATICS), EFFICIENCY, EIGENVALUES, PARALLEL PROCESSORS, PROBLEM SOLVING, SOLUTIONS(GENERAL), PERTURBATIONS, SPARSE MATRIX, ITERATIONS, LINEAR ALGEBRAIC EQUATIONS, STRUCTURED PROGRAMMING.

Subject Categories : Operations Research
      Computer Programming and Software

Distribution Statement : APPROVED FOR PUBLIC RELEASE