Vector-sparse solver for unsymmetrical matrices

Hakizumwami Birali Runesha

doi:10.1016/S0965-9978(00)00024-7

Vector-sparse solver for unsymmetrical matrices

2000-08-01

6

PRE

AI

H

Hakizumwami Birali Runesha

DOI：10.1016/S0965-9978(00)00024-7

原文链接

原文求助

摘要

En 中文

In this work, efficient algorithms for sparse computations (reordering algorithms, storage schemes, symbolic factorization, master degree-of-freedom, L*D*U numerical factorization, forward and backward solutions) are developed and integrated into the proposed procedures. In order to exploit fast saxpy operations offered by many vector computers, take advantage of available cache in many workstations, and minimize data movements into fast memory, special storage schemes are designed to store the coefficient (unsymmetrical) matrix. Thus, the upper triangular portion of the coefficient matrix is stored in a compressed sparse row format, while the lower triangular portion of the same matrix is stored in a compressed sparse column format. A reordering algorithm is applied on one portion of the matrix to minimize fill-ins. Unsymmetrical matrix-vector multiplication has also been sparsely computed for error-norm check purpose. The entire sparse procedures have been coded in standard FORTRAN language. (C) 2000 Elsevier Science Ltd. All rights reserved.

Keyword:

Fortran language

modified minimum degree

TRANSA

sparse solver

unsymmetrical matrices

Vector-sparse solver for unsymmetrical matrices

摘要

AI总结

期刊

机构

引用论文