Iterative methods for solving large-scale problems of structural mechanics using multi-core computers

The paper studies the conjugate gradient method for solving systems of linear algebraic equations with symmetric sparse matrices that arise when the finite-element method is applied to the problems of structural mechanics. The main focus is on designing effective preconditioning and parallelizing the method for multi-core desktop computers. Preconditioning is based on the incomplete Cholesky “by value” factorization method and implemented based on the technique of sparse matrices, which allows increasing convergence considerably without a significant increase of the computer's resources. Parallelization is implemented for the incomplete factorization as well as for iterative process stages. The method is integrated into the SCAD software package (www.scadsoft.com). The paper includes a discussion of the results of calculations done with direct and iterative methods for large-scale design models of tall buildings, originally from the SCAD Soft1 problem collection.