Parallelism-High Performance Computing-Grid
- Domaine
- Parallelism-High Performance Computing-Grid
- Domain - extra
- Physics
- Année
- 2012
- Starting
- September 2012
- État
- Open
- Sujet
- Beyond polyhedral model
- Thesis advisor
- EISENBEIS Christine
- Co-advisors
- Cédric Bastoul
- Laboratory
- Collaborations
- LAL Laboratoire de l'Accélérateur Linéaire (Paris-Sud)
LPT Laboratoire de Physique Théorique (Paris-Sud)
Partenaires d'un projet ANR. - Abstract
- There are a number of numerical transformations that are applied in HPC scientific computing in order to match multiprocessors features, namely, parallelism, communication and memory hierarchy. Here, we refer to algorithm transformations that may change the semantics of corresponding programs and thus are not encompassed in today's optimizing tools. These transformations have important impact on the numerical stability and convergence of algorithms. The topic of this PhD thesis is to understand how to formalize these transformations and whether they can be directly included in the polyhedral model used in compiler optimization, or if the polyhedral model must be extended. We plan then to implement these transformations in an existing meta-language tool. Targeted architectures are clusters of GPU, supercomputers such as IBM Bluegene or very large clusters.
- Context
- Objectives
- Work program
- Extra information
- Prerequisite
- Détails
thesis.pdf- Expected funding
- Institutional funding
- Status of funding
- Expected
- Candidates
- Utilisateur
- cedric.bastoul
- Créé
- Jeudi 31 mai 2012 23:19:37 CEST
- dernière modif.
- Jeudi 31 mai 2012 23:20:15 CEST
Connexion
Ecole Doctorale Informatique Paris-Sud
Directrice
Nicole Bidoit
Assistante
Stéphanie Druetta
Conseiller aux thèses
Dominique Gouyou-Beauchamps
ED 427 - Université Paris-Sud
UFR Sciences Orsay
Bat 650 - aile nord - 417
Tel : 01 69 15 63 19
Fax : 01 69 15 63 87
courriel: ed-info à lri.fr