Parallel Asynchronous Algorithms III. Application, Performances

Add to my library

AF1387 V1 Article

Parallel Asynchronous Algorithms III. Application, Performances

Authors : Pierre SPITERI, Jean-Claude MIELLOU

Publication date: October 10, 2021 | Lire en français

Add to my library Add to my library

Logo Techniques de l'Ingenieur You do not have access to this resource.
Request your free trial access! Free trial

Already subscribed?

Overview

ABSTRACT

This article is essentially devoted to the applicative aspects of the parallel asynchronous iterative methods to solve numerically large scale problems resulting from the discretization of pseudo-linear boundary value problems and also from optimisation problems or from algebro-differential equations but also from non-numerical problems. On the other hand numerical tests allow to show under which conditions the parallel iterative asynchronous methods compared to the synchronous ones are efficient.

Read this article from a comprehensive knowledge base, updated and supplemented with articles reviewed by scientific committees.

Read the article

AUTHORS

  • Pierre SPITERI : Professor Emeritus - University of Toulouse, INP-ENSEEIHT – IRIT, Toulouse, France

  • Jean-Claude MIELLOU : Honorary Professor - University of Bourgogne Franche-Comté, Department of Mathematics, Besançon, France

 INTRODUCTION

In our first article [AF 1 385] , we presented the mathematical model describing asynchronous parallel iterative methods for solving univariate and multivariate pseudo-linear problems, the latter corresponding to the situation where the solution is subject to inequality constraints. We have also presented three separate methods for analyzing the behavior of these iterative methods, under suitable assumptions. The convergence of these methods can be analyzed either by contraction techniques, or by partial-order techniques linked to the use of the discrete maximum principle, or because successive iterates belong to nested sets centered on the solution U * of the problem to be solved. It should be noted that this last approach, based on nested sets, unifies the two previous approaches based on contraction or partial order techniques, but does not provide practical convergence criteria. A comparison of these three analysis techniques has revealed the advantages of each; in particular, under suitable assumptions, contraction techniques give an estimate of the asymptotic speed of convergence of these iterative methods, and obtain a convergence property whatever the block decomposition of the problem to be solved. What's more, this type of contraction analysis can be used to solve univocal and multivocal pseudo-linear problems, whereas partial-order techniques can only be used to solve univocal pseudo-linear problems, since the continuity assumption required in the analysis criterion for partial-order techniques is not satisfied.

Also in this first article [AF 1 385] , various methods for decomposing the problem to be solved into blocks were indicated. These include sub-domain methods and, more generally, multi-decomposition methods. Thanks to the results established above, the behavior of these methods has been analyzed. Finally, this article presents a number of examples of classical problems where the analysis criteria have been practically applied.

The following article

You do not have access to this resource.
Logo Techniques de l'Ingenieur

Exclusive to subscribers. 97% yet to be discovered!

You do not have access to this resource. Click here to request your free trial access!

Already subscribed?


KEYWORDS

high performance computing   |   discretized pseudo-linear problems   |   efficiency of algorithms   |   large scale problems

Ongoing reading
Parallel Asynchronous Algorithms III. Application, Performances

Article included in this offer

"Mathematics"

( 167 articles )

Complete knowledge base

Updated and enriched with articles validated by our scientific committees

Services

A set of exclusive tools to complement the resources

View offer details

Dans les ressources documentaires

Algorithmes parallèles asynchrones I - Modélisation et analyse

Les algorithmes itératifs parallèles asynchrones et leurs extensions constituées par les méthodes de sous...

Algorithmes parallèles asynchrones II - Implémentation

L’implémentation des algorithmes itératifs parallèles asynchrones est l'objet du présent article. On abor...

Calcul de fonctions de matrices

On rappelle les définitions d’une fonction f ( A ) d’une matrice carrée à coefficients réels ou complexe...

Comportement dynamique des systèmes à événements discrets dans l’algèbre des dioïdes

Cet article s’intéresse au comportement dynamique des systèmes à événements discrets, dans une structure ...

WhitePaper Les géants du web
2 November 2016
Les géants du web

Avec Internet, tout va plus vite. Les grandes firmes du web le savent et adaptent leur stratégie pour continuer à se développer : Recherche et développement, ra...

Tous les livres blancs
Article Pasqal, le champion français du quantique
12 July 2024
Pasqal, le champion français du quantique

Qubits supraconducteurs, photons, atomes neutres, ions piégés, silicium… Il existe différentes technologies pour développer un ordinateur quantique. En misant s...

Toutes les actualités
Toutes les conférences en ligne
Contact us