3D Cartesian Transport Sweep for Massively Parallel Architectures with PARSEC

Salli Moustafa (1, 2) , Mathieu Faverge (3, 2) , Laurent Plagne (1) , Pierre Ramet (3, 2)

1 EDF R&D - EDF R&D
2 HiePACS - High-End Parallel Algorithms for Challenging Numerical Simulations
3 LaBRI - Laboratoire Bordelais de Recherche en Informatique

High-fidelity nuclear power plant core simulations require solving the Boltzmann transport equation. In discrete ordinates methods, the most computationally demanding oper-ation of this equation is the sweep operation. Considering the evolution of computer architectures, we propose in this paper, as a first step toward heterogeneous distributed architectures, a hybrid parallel implementation of the sweep operation on top of the generic task-based runtime system: PARSEC. Such an implementation targets three nested levels of parallelism: mes-sage passing, multi-threading, and vectorization. A theoretical performance model was designed to validate the approach and help the tuning of the multiple parameters involved in such an approach. The proposed parallel implementation of the Sweep achieves a sustained performance of 6.1 Tflop/s, corresponding to 33.9% of the peak performance of the targeted supercom-puter. This implementation compares favorably with state-of-art solvers such as PARTISN; and it can therefore serve as a building block for a massively parallel version of the neutron transport solver DOMINO developed at EDF.

Mots clés: Computational model; Distributed computing; Hybrid parallelism; 3D Sweep; Task-Based Programing Model

Domaines: Calcul parallèle, distribué et partagé [cs.DC]

Salli Moustafa, Mathieu Faverge, Laurent Plagne, Pierre Ramet. 3D Cartesian Transport Sweep for Massively Parallel Architectures with PARSEC. IEEE International Parallel & Distributed Processing Symposium (IPDPS 2015), May 2015, Hyderabad, India. pp.581-590, ⟨10.1109/IPDPS.2015.75⟩. ⟨hal-01078362⟩