Abstract. This paper is concerned with proving theoretical results related to the convergence of the Conjugate Gradient method for solving positive definite symmetric linear systems. New relations for ratios of the A-norm of the error and the norm of the residual are provided starting from some earlier results of Sadok [13]. These results use the well-known correspondence between the Conjugate Gradient method and the Lanczos algorithm.

In this paper, a general-purpose block LU preconditioner for saddle point problems is presented. The main difference between the approach presented here and that of other studies is that an explicit, accurate approximation of the Schur complement matrix is efficiently computed. This is used to compute a preconditioner to the Schur complement matrix, which in turn defines a preconditioner for the global system. A number of different variants are developed and results are reported for a few linear systems arising from CFD applications.

10 figures 10th IMACS International Symposium on Iterative Methods in Scientific ComputingInternational audienceThis paper addresses the classical and discrete Euler-Lagrange equations for systems of $n$ particles interacting quadratically in $\mathbb{R}^d$. By highlighting the role played by the center of mass of the particles, we solve the previous systems via the classical quadratic eigenvalue problem (QEP) and its discrete transcendental generalization. The roots of classical and discrete QEP being given, we state some conditional convergence results. Next, we focus especially on periodic and choreographic solutions and we provide some numerical experiments which confirm the convergence