Professional Summary
Pierre Ramet is a Full Professor of Computer Science at Bordeaux University and a researcher at Inria. His research interests span high-performance computing, focusing on sparse linear algebra and parallel algorithms. He leads the team in charge of developing PaStiX, a high-performance sparse direct solver.
Education
HdR in Computer Science
2017
Bordeaux University
PhD in Computer Science
2000
Bordeaux University
Interests
High-Performance Computing
Sparse Linear Algebra
Parallel Algorithms
Scientific Computing
🔬 PaStiX Highlights
Featured publications and algorithmic contributions
Sparse Linear Algebra • PaStiX
Low Rank compression for sparse matrices
From BLR to HODLR
Exploiting low-rank approximations within sparse direct solvers to achieve dramatic reductions in memory consumption and computational complexity on large-scale problems.
Experience
Full Professor
Bordeaux University
Parallel algorithms in scientific computing
Head of Department
SATANAS department of LaBRI
Head of the SATANAS (Supports and Algorithms for High Performance Numerical Applications) department
Scientific Advisor
CEA (French Atomic Energy Commission)
Scientific advisor in the field of high-performance computing for numerical simulations
Deputy Head of Department
SATANAS department of LaBRI
Deputy head of the SATANAS (Supports and Algorithms for High Performance Numerical Applications) department
Scientific Expert
GENCI (French National High-Performance Computing Agency)
Scientific expert on the Mathematics and Computer Science evaluation board
Assistant Professor
Bordeaux University
Parallel algorithms in scientific computing
Education
HdR in Computer Science
Bordeaux University
Habilitation à Diriger des Recherches
PhD in Computer Science
Bordeaux University
Doctorat en Informatique
🔬 Research & Activities
Pierre Ramet is a Full Professor of Computer Science at Bordeaux University and a researcher at Inria. His research interests span high-performance computing, focusing on sparse linear algebra and parallel algorithms.
He leads the team in charge of developing PaStiX, a high-performance sparse direct solver.
- Research team: TOPAL (Tools and Optimization for High-Performance Applications and Learning)
European & National Initiatives
DARE
2025 – present
HPC Digital Autonomy with RISC-V in Europe
EuroHPC project aiming at sovereign European computing, focused on high-performance linear algebra software stacks optimized for emerging RISC-V architectures.
PEPR NumPEx
2021 – present
French Exascale Software Initiative
Design of next-generation software methods and scalable numerical algorithms to harness upcoming European and French exascale supercomputers.
HYBQUANT
2024 – present
Maison du Quantique en Nouvelle-Aquitaine (HQI)
France Hybrid HPC Quantum Initiative (HQI): co-design and governance of hybrid high-performance computing and quantum simulation environments.
EoCoE
2019 – 2021
Energy Oriented Center of Excellence: Toward Exascale for Energy
Advanced high-performance mathematical modeling and parallel numerical methods targeting clean energy transition and fusion simulation.
ANR Research Grants
SOLHARIS (CE46)
2019 – 2024
SOLvers for Heterogeneous Architectures over Runtime systems, Investigating Scalability
Scalable direct methods, runtime systems (StarPU) and scheduling algorithms for extreme-scale heterogeneous architectures. Collaboration: Inria, CNRS, IRIT, CEA.
SASHIMI (JCJC18)
2018 – 2023
Sparse Direct Solver using Hierarchical Matrices
Integration of hierarchical low-rank compression (BLR / HODLR) in PaStiX over runtime systems, optimizing asymptotic complexity and memory consumption.
SOLHAR (MN13)
2013 – 2017
Solvers for Heterogeneous Architectures on Top of Task-Based Runtime Systems
Design and implementation of direct sparse linear solver algorithms on emerging platforms equipped with hardware accelerators.
ANEMOS (MN11)
Deputy Head • 2011 – 2015
Advanced Numeric for ELMs: Modeling and Optimized Schemes
Modeling Edge Localized Modes (ELMs) in Tokamak plasmas (ITER) and adapting direct sparse solvers to heterogeneous multicore/multi-GPU nodes.
PETAL & PETALH
2008 – 2012
Preconditioning Scientific Applications on Petascale Heterogeneous Machines
ANR COSINUS08 / MN10: Block incomplete factorizations (ILU) and Schur complement memory reduction for large-scale scientific applications.
ASTER (CIS06)
Deputy Head • 2006 – 2010
Adaptive MHD Simulation of Tokamak ELMs for ITER
Development of magnetohydrodynamic simulation codes, adaptive mesh refinement and parallel solvers for fusion plasma instabilities.
SOLSTICE (CIS06)
2006 – 2010
SOLvers et SimulaTIons in Extreme Computing
High-performance parallel linear solvers for complex multi-physics and multi-scale grand challenge simulations.
NUMASIS (CIS05)
2005 – 2009
HPC on NUMA Architectures for Seismology Simulations
Parallel algorithms and runtime memory models for NUMA multicore architectures, applied to seismic wave propagation and sparse solvers.
🎓 Teaching & Academic Responsibilities
Pierre Ramet has an extensive teaching activity at Bordeaux University, teaching approximately 300 hours (HETD) each year across undergraduate, graduate, and engineering curricula.
Main Teaching Domains & Programs
- IUT de Bordeaux (Computer Science Dept. / BUT Informatique):
- Operating systems, computer networks, computer architecture, and systems programming.
- Cryptography and security fundamentals.
- DevOps (virtualization and CI/CD automation).
- Unsupervised machine learning in the context of the DU IA (Diplôme d’Université Intelligence Artificielle).
- SAE (Situations d’Apprentissage et d’Évaluation): Project-based learning, problem-solving, and software engineering integration.
- Enseirb-Matmeca & Master in Computer Science:
- Numerical algorithms.
- High-Performance Computing (HPC), parallel & distributed algorithms, and scientific computing.
- Cryptography and security.
- Introduction to Quantum Computing.
- International Teaching:
- Graduate-level courses and summer schools abroad (Gabon, Vietnam).
Pedagogical Responsibilities
- Head of Licences Professionnelles (IUT de Bordeaux, 2007–2014):
- DAGPI (Développement d’Applications et Gestion de Projet Informatique).
- DAWIN (Développement d’Applications Web et Images Numériques).
- Coordinator for further education / graduate studies placement (2017–2025) for the Computer Science Department.
- Active member of university pedagogical committees and exam boards.
Publications
Recent publications are also available on HAL, and the complete bibliography can be browsed in the BibTeX Archive (bibtex2html).
(2024).
Comparative Study of Mixed-Precision and Low-Rank Compression Techniques in Sparse Direct Solvers.
2024 SIAM Conference on Applied Linear Algebra.
(2024).
Toward an algebraic multigrid method for the indefinite Helmholtz equation.
18th Copper Mountain Conference On Iterative Methods.
(2024).
Enhancing sparse direct solver scalability through runtime system automatic data partition.
Lecture Notes in Computer Science.
Projects
News & Releases
Participation au projet européen EuroHPC DARE
Participation au projet européen DARE (EuroHPC JU) axé sur le développement de bibliothèques d'algèbre linéaire pour les architectures européennes RISC-V.
Soutenance de thèse : Clément Richefort
Soutenance de la thèse de doctorat de Clément Richefort à l'Université de Bordeaux / LaBRI (bourse CEA).
Initiative HybQuant : Maison du quantique en Nouvelle-Aquitaine
Sélection et montage du projet HybQuant dans le cadre de l'appel national HQI, reliant calcul intensif classique et algorithmes quantiques.
Build PaStiX with Guix-HPC
See [Gitlab Inria](https://gitlab.inria.fr/guix-hpc/guix-hpc)
Contact
Pierre Ramet
Professeur des Universités, Université de Bordeaux
Inria Bordeaux - Sud-Ouest
351 cours de la Libération, 33405 Talence Cedex, France
- Office: Inria Bordeaux / LaBRI
- Phone: +33 5 24 57 40 39
- Email: pierre.ramet@u-bordeaux.fr / ramet@labri.fr
- PGP Public Key: ramet.asc