Skip to Main content Skip to Navigation

Ordonnancement avec communications pour systèmes multiprocesseurs dans divers modèles d'exécution

Abstract : In few years, parallel computers have been designed and have been widely developped. The main technical problems have been solved and today, the crucial goal is to provide a framework, a parallel programming environment, in order to program these machines. This thesis is part of the IMAG APACHE project which aims at building such an environment. The graphs are DAGs (Directed Acyclic Graphs). The process leading to the parallelization of an application consists of three main steps, with the scheduling and mapping as the central phase. Within this context, we focus our study on the designing of efficient and adaptable scheduling strategies, in the case of various granularity and within various execution models. From an algorithm that provided optimal schedules for the case of UECT intrees (unit execution and communication time tasks) on two processors, we have shown that it was possible to yield near-optimal schedules for intrees with a ration computation/communications lower or greater than one. Moreover, the same algorithm provides, for given families of intrees, optimal schedules within a totally different execution model. This study on two processors is extended for the case of m (greater than two) identical processors and for two uniform processors. Finally, part of this work is dedicated to the study of scheduling strategies for the special graphs generated by Athapascan (APACHE project) for which the determination of the granularity can be done during the scheduling phase.
Complete list of metadata

Cited literature [50 references]  Display  Hide  Download
Contributor : Thèses Imag <>
Submitted on : Tuesday, February 24, 2004 - 11:34:47 AM
Last modification on : Wednesday, March 10, 2021 - 1:50:03 PM
Long-term archiving on: : Thursday, September 13, 2012 - 12:00:17 PM


  • HAL Id : tel-00005049, version 1




Frédéric Guinand. Ordonnancement avec communications pour systèmes multiprocesseurs dans divers modèles d'exécution. Réseaux et télécommunications [cs.NI]. Institut National Polytechnique de Grenoble - INPG, 1995. Français. ⟨tel-00005049⟩



Record views


Files downloads