Sébastien MORAIS
Sébastien Morais était Ingénieur-Chercheur en Informatique au CEA, établissement qu’il a quitté en 2023.
SIAM International Meshing Roundtable 2023, Springer Nature Switzerland, p. 43-63, 2023
abstract
Abstract
This paper presents Coupe, a mesh partitioning platform. It provides solutions to solve different variants of the mesh partitioning problem, mainly in the context of load-balancing parallel mesh-based applications. From partitioning weights ensuring balance to topological partitioning that minimizes communication metrics through geometric methods, Coupe offers a large panel of algorithms to fit user-specific problems. Coupe exploits shared memory parallelism, is written in Rust, and consists of an open-source library and command line tools. Experimenting with different algorithms and parameters is easy. The code is available on Github.
Euro-Par 2022 International Workshops, Glasgow, UK, August 22–26, 2022, Revised Selected Papers, Glasgow, United Kingdom, 2023
abstract
Abstract
Mesh partitioning used for load balancing in distributed numerical simulations is typically managed with tools that are good enough but not optimal. Their use scope is not explicitly dedicated to load balancing, and they cannot make use of all available information. In this paper, the mesh partitioning problem and the context for its use are precisely defined. Then, existing tools are presented, along with their characteristics and features that are missing. Finally, a new partitioning platform – the subject of my PhD thesis – is presented: its architecture, software engineering choices made along the way, and how it can be the best fit for load balancing distributed simulations. The platform is open-source and is hosted on GitHub: https://github.com/LIHPC-Computational-Geometry/coupe .
Euro-Par 2022: Parallel Processing - 28th International Conference on Parallel and Distributed Computing, Glasgow, UK, August 22-26, 2022, Proceedings, Springer, p. 136–151, 2022
CEA, LIHPC Computational Geometry group, 2022
2020 Proceedings of the SIAM Workshop on Combinatorial Scientific Computing, p. 85-95, 2020
abstract
Abstract
Running numerical simulations on HPC architectures requires distributing data to be processed over the various available processing units. This task is usually done by partitioning tools, whose primary goal is to balance the workload while minimizing inter-process communication. However, they do not take the memory load and memory capacity of the processing units into account. As this can lead to memory overflow, we propose a new approach to address mesh partitioning by including ghost cells in the memory usage and by considering memory capacity as a strong constraint to abide. We model the problem using a bipartite graph and present a new greedy algorithm that aims at producing a partition according to the memory capacity. This algorithm focuses on memory consumption, and we use it in a multi-level approach to improving the quality of the returned solutions during the refinement phase. The experimental results obtained from our benchmarks show that our approach can yield solutions respecting memory constraints for instances where traditional partitioning tools fail.
ISAV 18: Proceedings of the Workshop on In Situ Infrastructures for Enabling Extreme-Scale Analysis and Visualization, p. 7-12, 2018
Conférence d'informatique en Parallélisme, Architecture et Système (Compas'2016), 2016
Euro-Par 2016: Parallel Processing - 22nd International European Conference on Parallel and Distributed Computing, Grenoble, FR, August 24-26, 2016, Proceedings, p. 196-208, 2016
Thèse de doctorat, spécialité informatique, CEA, Université d'Evry-Val-d'Essonne, 2016
ROADEF - 15ème Congrès Annuel de La Société Française de Recherche Opérationnelle et d'aide à La Décision, Société française de recherche opérationnelle et d'aide à la décision, 2014
CSC 14, p. 2, 2014