RUS  ENG
Full version
JOURNALS // Preprints of the Keldysh Institute of Applied Mathematics // Archive

Keldysh Institute preprints, 2021 048, 32 pp. (Mi ipmp2965)

This article is cited in 3 papers

MPI+OpenMP parallel implementation of conjugate gradient method with preconditioner of block partial inverse triangular decomposition of IC2S and IC1

O. Yu. Milyukova


Abstract: The paper proposes a new preconditioner for solving systems of linear algebraic equations with a symmetric positively defined matrix by the method of conjugate gradients – Block Incomplete Inverse Cholesky BIIC preconditioner in combination with a triangular first-order decomposition 'by value' - BIIC-IC1. The algorithm based on MPI+OpenMP techniques is proposed for the construction and application of the BIIC preconditioner combined with stabilized triangular decomposition of the second order 'by value' (BIIC-IS2S). In this case, the BIIC-IC2S preconditioner uses the number of blocks multiple of the number of processors used and the number of threads used. Two algorithms based on MPI+OpenMP techniques are proposed for the construction and application of the BIIC-IC1 preconditioner. Comparative timing results for the MPI+OpenMP and MPI implementations of the proposed preconditioning used with the conjugate gradient method for a model problem and the sparse matrix collections SuiteSparse are presented.

Keywords: sparse matrices, implicit block preconditioning, incomplete Cholesky factorization, parallel preconditioning, conjugate gradient method.

DOI: 10.20948/prepr-2021-48



© Steklov Math. Inst. of RAS, 2026