https://doi.org/10.1051/epjn/2025053
Regular Article
Accelerating split-exponential track length estimator on GPU for Monte Carlo simulations
1
CEA, DES, IRESNE, DER, Service de Physique des Reacteurs et du Cycle, Cadarache, F-13108, Saint-Paul-lez-Durance, France
2
CEA, Université Paris-Saclay, Service d’Etudes de Réacteurs et de Mathématiques Appliquées, 91191 Gif-sur-Yvette, France
* e-mail: henri.hutinet@cea.fr
Received:
31
May
2025
Received in final form:
5
August
2025
Accepted:
7
August
2025
Published online: 8 October 2025
In the context of computing 3D volumetric tallies for nuclear applications, the combination of Monte Carlo methods and high-performance computing is essential to achieve accurate yet computationally feasible simulations that meet industrial time constraints. The next-event Split Exponential Track-Length Estimator (seTLE) is particularly well suited for estimating tallies on meshes. To alleviate the computational burden associated with seTLE, such as sampling numerous outgoing pseudoparticles at each collision, estimating cross sections, performing ray tracing through complex geometries, and accumulating scores across the geometry, we leverage the parallel computing capabilities of Graphics Processing Units (GPUs). We assess the performance of our implementation using two shielding configurations and one criticality benchmark. Both photon and neutron transport simulations are considered. Scores are evaluated over Cartesian meshes, material volumes, and energy group structures. In all cases, acceleration factors greater than unity are observed in the detectors, reaching several hundred in selected regions of the phase space. In a final experiment, we demonstrate that our GPU-based implementation achieves a net energy gain (in Watt) even when compared to a conventional CPU-based TLE, despite the additional computational cost of GPU use.
© H. Hutinet et al., Published by EDP Sciences, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
