Bienvenue aux Mines Paristech
Bienvenue à MINES ParisTech
Newsletter International
Website
Théorie & Pratique
Vous êtes

webTV

Lecture

Aftermovie du H4ck@mines 2019 MINES ParisTech

Lecture

MIG 2019 - Systèmes embarqués

Lecture

Teaser Journées Portes Ouvertes

Lecture

Season's Greetings - Meilleurs voeux 2019

Lecture

Plan stratégique de l'Ecole

+ Toutes les vidéos

Partager

Synthèse de code hybride à partir d'un programme à mémoire purement distribuée

Synthèse de code hybride à partir d'un programme à mémoire purement distribuée

Hybrid code synthesis from a pure distributed memory program

Proposition de thèse

Spécialité

Informatique temps réel, robotique et automatique - Fontainebleau

Ecole doctorale

Ingénierie des Systèmes, Matériaux, Mécanique, Énergétique

Directeur de thèse

TADONKI Claude

Unité de recherche

Mathématiques et Systèmes

ContactClaude TADONKI
Date de validité

01/10/2019

Site Web
Mots-clés

parallélisme, mémoire

parallelism, memory

Résumé

With the advent of multicore architectures we need us to reconsider the way we organize the tasks of our parallel code and plan their cooperation. Current (and future) supercomputers are built up with multicore processors probably coupled with accelerators such as GPUs. Regarding the connectivity, latency and network bandwidth have certainly evolved significantly, but the gap between the virtual topology and physical topology greatly increases the effective cost of interprocessor communications, especially for applications involving a high connectivity like stencil computation. Therefore, it becomes crucial to really consider hybrid implementations, which consider distributed memory model on top and shared memory model on the nodes. However, several factors are likely to discourage this effort: the required skill to design and implement hybrid codes is not common; in addition, number of quite complex codes have already been written based on the distributed memory model exclusively, and it is hard to consider modifying them; moreover, the belief of a clear effective reward from a hybrid code is moderate, because getting a benefit as close as expected is indeed not trivial. Therefore, designing a quasi-systematic methodology and a corresponding framework, that allow to easily switching from a pure MPI code to an efficient corresponding MPI + OpenMP version would be a valuable contribution. In this thesis, we propose to tackle the problem quite rigorously, on the basis of a tasks graph model and a clearly quantified accounting of major hardware and network aspects. The first goal will be to understand the criteria for an efficient and scalable hybrid code, followed by a methodology for a quick and pragmatic implementation. Applications envisaged for conducting this study are: Lattice QCD; pattern matching on a large sequence of high resolution images; partial differential equations; shortest paths computation in a graph.

With the advent of multicore architectures we need us to reconsider the way we organize the tasks of our parallel code and plan their cooperation. Current (and future) supercomputers are built up with multicore processors probably coupled with accelerators such as GPUs. Regarding the connectivity, latency and network bandwidth have certainly evolved significantly, but the gap between the virtual topology and physical topology greatly increases the effective cost of interprocessor communications, especially for applications involving a high connectivity like stencil computation. Therefore, it becomes crucial to really consider hybrid implementations, which consider distributed memory model on top and shared memory model on the nodes. However, several factors are likely to discourage this effort: the required skill to design and implement hybrid codes is not common; in addition, number of quite complex codes have already been written based on the distributed memory model exclusively, and it is hard to consider modifying them; moreover, the belief of a clear effective reward from a hybrid code is moderate, because getting a benefit as close as expected is indeed not trivial. Therefore, designing a quasi-systematic methodology and a corresponding framework, that allow to easily switching from a pure MPI code to an efficient corresponding MPI + OpenMP version would be a valuable contribution. In this thesis, we propose to tackle the problem quite rigorously, on the basis of a tasks graph model and a clearly quantified accounting of major hardware and network aspects. The first goal will be to understand the criteria for an efficient and scalable hybrid code, followed by a methodology for a quick and pragmatic implementation. Applications envisaged for conducting this study are: Lattice QCD; pattern matching on a large sequence of high resolution images; partial differential equations; shortest paths computation in a graph.

Contexte

Parallélisme

Encadrement

Claude Tadonki, CRI, Centre de recherche en informatique

Profil candidat

Please send a CV, a motivation letter, the list of grades for your last academic year and 1 or 2 recommendation letters.

Please see description in French

Objectif

Innovation

Références

-

Type financement

Concours pour un contrat doctoral

Retour à la liste des propositions

actualité

Prix Maurice Allais de Science économique 2019

Formation Prix Maurice Allais de Science économique 2019 Deux articles primés Le Jury a unanimement souhaité cette…
> En savoir +

Des recherches prometteuses en sciences des matériaux

Formation Des recherches prometteuses en sciences des matériaux Manon Isard , doctorante au Cemef MINES ParisTech , a reçu…
> En savoir +

Thierry Coupez en mission auprès des icebergs

Formation Thierry Coupez en mission auprès des icebergs Thierry Coupez , numéricien expert en calcul intensif et…
> En savoir +

Des prix pour Stimshop

Formation Des prix pour Stimshop Lors de la 5 e édition de son événement "Global Innovation & Enterprise", le 21 février 2019,…
> En savoir +

@W_AT_World révèle ses

Formation @W_AT_World révèle ses "40 in their 40s" Félicitations à Margaret Kyle , professeure d'économie au…
> En savoir +

+ Toutes les actualités

contact

Régine MOLINS
Direction de l'Enseignement
Service du Doctorat
> envoyer un mail

Plan du site
MINES
ParisTech

60, Boulevard Saint-Michel
75272 PARIS Cedex 06
Tél. : +33 1 40 51 90 00

Implantations
Mentions légales | efil.fr | ©2012 MINES ParisTech | +33 1 40 51 90 00 |