One-Step Algorithm for Mixed Data and Task Parallel Scheduling Without Data Replication

Vincent Boudet 1 Frédéric Desprez 2 Frédéric Suter 2
1 APR - Algorithmes et Performance des Réseaux
LIRMM - Laboratoire d'Informatique de Robotique et de Microélectronique de Montpellier
Abstract : In this paper we propose an original algorithm for mixed data and task parallel scheduling. The main specificities of this algorithm are to simultaneously perform the allocation and scheduling processes, and avoid the data replication. The idea is to base the scheduling on an accurate evaluation of each task of the application depending on the processor grid. Then no assumption is made with regard to the homogeneity of the execution platform. The complexity of our algorithm are given. Performance achieved by our schedules both in homogeneous and heterogeneous worlds, are compared to data-parallel executions for two applications: the complex matrix multiplication and the Strassen decomposition.
Type de document :
Communication dans un congrès
IPDPS'03: 17th International Parallel and Distributed Processing Symposium, Apr 2003, Nice, France. 2003, 〈http://www.ipdps.org/ipdps2003/〉
Liste complète des métadonnées

https://hal-lirmm.ccsd.cnrs.fr/lirmm-00269808
Contributeur : Christine Carvalho de Matos <>
Soumis le : vendredi 10 juillet 2015 - 11:31:36
Dernière modification le : jeudi 24 mai 2018 - 15:59:21
Document(s) archivé(s) le : lundi 12 octobre 2015 - 11:31:18

Fichier

LIP-RR2002-34.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : lirmm-00269808, version 1

Citation

Vincent Boudet, Frédéric Desprez, Frédéric Suter. One-Step Algorithm for Mixed Data and Task Parallel Scheduling Without Data Replication. IPDPS'03: 17th International Parallel and Distributed Processing Symposium, Apr 2003, Nice, France. 2003, 〈http://www.ipdps.org/ipdps2003/〉. 〈lirmm-00269808〉

Partager

Métriques

Consultations de la notice

644

Téléchargements de fichiers

213