Direct methods for solving large sparse equations based on the block two by two decomposition of the matrix
Matematičeskoe modelirovanie, Tome 5 (1993) no. 2, pp. 66-81
Cet article a éte moissonné depuis la source Math-Net.Ru
Several algorithms for reordering sparse symmetric positive definite matrix to a block two by two form are considered; a task of finding a permutation such that filling is at minimum in a block $(1,1)$ and is located mainly in blocks $(2,1)$, $(2,2)$ is posed. In this respect two algorithms from widely known sparse matrix package SPARSPAK are analyzed: QMD – a Quotient Minimum Degree and ND – nested dissection algorithms; a new one is proposed which is called $\mathrm{BND}+\mathrm{qmd}$ – Balanced ND with internal (influencing block $(1,1)$) qmd-ordering. The results of numerical experiments for a set of grid problems containing 10000–25000 unknown values are presented. These results show the usage of implicit solution scheme may provide up to 25–30% reduction of primary storage without visible increasing the number of operations required to solve triangular system.
@article{MM_1993_5_2_a5,
author = {A. B. Kycherov and E. J. Oleinik},
title = {Direct methods for solving large sparse equations based on the block two by two decomposition of the matrix},
journal = {Matemati\v{c}eskoe modelirovanie},
pages = {66--81},
year = {1993},
volume = {5},
number = {2},
language = {ru},
url = {http://geodesic.mathdoc.fr/item/MM_1993_5_2_a5/}
}
TY - JOUR AU - A. B. Kycherov AU - E. J. Oleinik TI - Direct methods for solving large sparse equations based on the block two by two decomposition of the matrix JO - Matematičeskoe modelirovanie PY - 1993 SP - 66 EP - 81 VL - 5 IS - 2 UR - http://geodesic.mathdoc.fr/item/MM_1993_5_2_a5/ LA - ru ID - MM_1993_5_2_a5 ER -
A. B. Kycherov; E. J. Oleinik. Direct methods for solving large sparse equations based on the block two by two decomposition of the matrix. Matematičeskoe modelirovanie, Tome 5 (1993) no. 2, pp. 66-81. http://geodesic.mathdoc.fr/item/MM_1993_5_2_a5/