Geodesic
Parallel versions of some methods with factorized preconditioning matrices for solving elliptic equations on unstructed triangular grids
Matematičeskoe modelirovanie, Tome 17 (2005) no. 2, pp. 65-80 
O. Yu. Milyukova. Parallel versions of some methods with factorized preconditioning matrices for solving elliptic equations on unstructed triangular grids. Matematičeskoe modelirovanie, Tome 17 (2005) no. 2, pp. 65-80. http://geodesic.mathdoc.fr/item/MM_2005_17_2_a6/
@article{MM_2005_17_2_a6,
     author = {O. Yu. Milyukova},
     title = {Parallel versions of some methods with factorized preconditioning matrices for solving elliptic equations on unstructed triangular grids},
     journal = {Matemati\v{c}eskoe modelirovanie},
     pages = {65--80},
     year = {2005},
     volume = {17},
     number = {2},
     language = {ru},
     url = {http://geodesic.mathdoc.fr/item/MM_2005_17_2_a6/}
}
TY  - JOUR
AU  - O. Yu. Milyukova
TI  - Parallel versions of some methods with factorized preconditioning matrices for solving elliptic equations on unstructed triangular grids
JO  - Matematičeskoe modelirovanie
PY  - 2005
SP  - 65
EP  - 80
VL  - 17
IS  - 2
UR  - http://geodesic.mathdoc.fr/item/MM_2005_17_2_a6/
LA  - ru
ID  - MM_2005_17_2_a6
ER  - 
%0 Journal Article
%A O. Yu. Milyukova
%T Parallel versions of some methods with factorized preconditioning matrices for solving elliptic equations on unstructed triangular grids
%J Matematičeskoe modelirovanie
%D 2005
%P 65-80
%V 17
%N 2
%U http://geodesic.mathdoc.fr/item/MM_2005_17_2_a6/
%G ru
%F MM_2005_17_2_a6

Voir la notice de l'article provenant de la source Math-Net.Ru

Parallel versions of the conjugate gradient methods with incomplete factorization type preconditioning and modified incomplete factorization type preconditioning are proposed for solving elliptic equations on unstructed triangular grid on distributed-memory parallel computers. The construction of the parallel versions of the methods is based on the special orderings of nodes of a grid. The main attention devote to the methods of computational domain decomposition, the hierarchical algorithm for partitioning graphs is used. The calculations of model problems demonstrate the admited increase of iteration number with the number of processors for moderate number of processors.