Skip to Main content Skip to Navigation
Theses

Algorithmes parallèles auto-adaptatifs et applications

Daouda Traoré 1
1 MOAIS - PrograMming and scheduling design fOr Applications in Interactive Simulation
Inria Grenoble - Rhône-Alpes, LIG - Laboratoire d'Informatique de Grenoble
Abstract : This thesis focuses on the building of algorithms and parallel programs that obliviously adapts the execution platform (number of processors, speeds of processors, ...). The building that we propose is based on the technology developed in the Moais team; it consists on the recursive and dynamic coupling of two algorithms: one sequential that minimizes the work (the number of operations), but not the parallel time; one parallel that minimizes the depth (parallel time) on an unbounded number of resources, but not the work. The two algorithms are interleaved on-line based on a work-stealing schedule. The contribution of this thesis consists in a theoretical analysis of coupling (lower bound, asymptotic optimality in some cases), and a "generic" implementation of the coupling scheme witch is applied to several examples (new on-line adaptive parallel prefix computation, adaptive merging and sorting, adaptive parallelization of several STL algorithms). In this thesis, we also propose a new strict optimal parallel static algorithm for prefix computation.
Document type :
Theses
Complete list of metadata

Cited literature [92 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00353274
Contributor : Daouda Traore <>
Submitted on : Thursday, January 15, 2009 - 11:32:04 AM
Last modification on : Tuesday, February 9, 2021 - 3:24:21 PM
Long-term archiving on: : Tuesday, June 8, 2010 - 8:13:13 PM

Identifiers

  • HAL Id : tel-00353274, version 1

Collections

Citation

Daouda Traoré. Algorithmes parallèles auto-adaptatifs et applications. Informatique [cs]. Institut National Polytechnique de Grenoble - INPG, 2008. Français. ⟨tel-00353274⟩

Share

Metrics

Record views

974

Files downloads

6605