MPI detach — Towards automatic asynchronous local completion - CEA - Commissariat à l’énergie atomique et aux énergies alternatives Accéder directement au contenu
Article Dans Une Revue Parallel Computing Année : 2022

MPI detach — Towards automatic asynchronous local completion

Résumé

When aiming for large scale parallel computing, waiting time due to network latency, synchronization, and load imbalance are the primary opponents of high parallel efficiency. A common approach to hide latency with computation is the use of non-blocking communication. In the presence of a consistent load imbalance, synchronization cost is just the visible symptom of the load imbalance. Tasking approaches as in OpenMP, TBB, OmpSs, or C ++20 coroutines promise to expose a higher degree of concurrency, which can be distributed on available execution units and significantly increase load balance. Available MPI non-blocking functionality does not integrate seamlessly into such tasking parallelization. In this work, we present a slim extension of the MPI interface to allow seamless integration of non-blocking communication with available concepts of asynchronous execution in OpenMP and C ++. We furthermore investigate compile-time analysis necessary to transform an application using blocking MPI communication into an application integrating OpenMP tasks with our proposed MPI interface extension.
Fichier principal
Vignette du fichier
mpi-detach.pdf (195.23 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

cea-03537990 , version 1 (20-01-2022)

Identifiants

Citer

Joachim Protze, Marc-André Hermanns, Matthias S Müller, Van Man Nguyen, Julien Jaeger, et al.. MPI detach — Towards automatic asynchronous local completion. Parallel Computing, 2022, 109, pp.102859. ⟨10.1016/j.parco.2021.102859⟩. ⟨cea-03537990⟩
131 Consultations
390 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More