This paper presents the behavior of general parallelsynchronous and asynchronous multisplitting and two-stage methods for the numerical simulation of steel solidification in continuous casting. Thanks to the mathemat...
详细信息
This paper presents the behavior of general parallelsynchronous and asynchronous multisplitting and two-stage methods for the numerical simulation of steel solidification in continuous casting. Thanks to the mathematical analysis and the implementation of these methods one can show the results of parallel experiments for the target application. The mathematical model is constituted by coupled nonlinear boundary value problems, namely the heat equation taking into account, on part of the boundary, a radiation phenomenon described by the Stefan law. For the numerical solution of such partial differential equations we consider, depending on whether the coefficient of thermal conductivity is constant or temperature-dependent, both an implicit or a semi-implicit discretization with respect to the time of the studied evolution problem, while the spatial discretization is carried out by adapted finite difference schemes. Then large scale discretized algebraic systems are solved by sequential and synchronous or asynchronous iterative algorithms;comparison of these various previous methods implemented on clusters and grid are achieved in both cases when the thermal conductivity is constant and more generally dependent of the temperature.
The present study deals with pseudo-linear problems solving using parallel asynchronous multisplitting methods combined with Krylov methods. With appropriate and realistic assumptions, the behavior of such parallel it...
详细信息
The present study deals with pseudo-linear problems solving using parallel asynchronous multisplitting methods combined with Krylov methods. With appropriate and realistic assumptions, the behavior of such parallel iterative algorithms will be analyzed by partial ordering techniques in relation with the discrete maximum principle. Applications to discretized boundary value problems are presented, the implementation of the algorithms is described and parallel experiments are analyzed.
The paper improves a preliminary experimental study on a cluster by adding both theoretical results and experimental tests on a grid platform. These algorithms solve univalued and multivalued pseudo-linear problems us...
详细信息
The paper improves a preliminary experimental study on a cluster by adding both theoretical results and experimental tests on a grid platform. These algorithms solve univalued and multivalued pseudo-linear problems using parallel asynchronous multisplitting methods combined with Krylov's methods. This paper also analyses these algorithms using contraction techniques. Two distinct applications, with discretized boundary value problems, are analyzed and simulated. First, a univalued convection-diffusion problem perturbed by an increasing diagonal operator is presented. Then, follows the description of a diffusion problem whose solution is constrained. This situation classically leads to the solution of a multivalued pseudo-linear problem in which the linear part is perturbed by an increasing diagonal multivalued operator. parallel asynchronous and synchronousalgorithms were implemented and tested on a grid platform composed of physically adjacent or geographically distant machines. In addition, the simulation results are detailed and show that the elapsed times obtained for the asynchronousalgorithms are significantly less than those obtained for the synchronousalgorithms.
暂无评论