Categories
seminaire passé seminar

[SEMINAR] May 30 2023 – DENERGIUM đŸ§‘â€đŸ« HervĂ© MATHIEU

đŸ§‘â€đŸ« HervĂ© MATHIEU – cofunder of DENERGIUM
🌎 May 30 2023
☕ 4:00 PM
🏱 Maison de la Simulation, Batiment Digiteo Saclay, Salle Mandelbrot
🔗 https://www.linkedin.com/company/denergium/ 
🔗 https://www.denergium.com

Abstract

DENERGIUM est une startup créée dĂ©but 2023. DENERGIUM est une entreprise technologique, un Ă©diteur de logiciels et un acteur environnemental positif. DENERGIUM a pour mission de rendre plus efficace l’utilisation des infrastructures informatiques. Pour cela DENERGIUM s’appuie sur une technologie issue d’Inria, qui Ă  partir de donnĂ©es Ă©nergĂ©tiques acquises dans un datacenter et d’algorithmes avancĂ©s (modĂšles mathĂ©matiques, IA) permet d’apporter des optimisations aux gestionnaires de datacenter mais aussi aux utilisateurs des datacenters. Notre marchĂ© est l’ensemble des infrastructures informatiques opĂ©rant du calcul massif (HPC, IA, BigData). DENERGIUM est accompagnĂ©e par Inria Startup Studio, Unitec et Ovhcloud startup program.

A partir d’une dĂ©monstration d’EnergyScopium optimize, nous pourrons Ă©changer sur la problĂ©matique de l’optimisation Ă©nergĂ©tique dans l’usage des clusters de calcul : configuration matĂ©rielle et logicielle des serveurs, lancement des calculs, choix des bibliothĂšques logicielles, dĂ©veloppement logiciel.

Categories
seminaire passé seminar

[SEMINAR] May 16 2023 – Development and GPU porting efforts for a high-performance low-Mach CFD solver based on unstructured adaptive grids đŸ§‘â€đŸ« Patrick BEGOU and Vincent MOUREAU

đŸ§‘â€đŸ« Patrick BEGOU – Research Engineer at LEGI, UMR 5519
đŸ§‘â€đŸ« Vincent MOUREAU, CNRS research fellow at CORIA, UMR6614
🌎 May 16 2023
☕ 9:30 AM
🏱 Maison de la Simulation, Batiment Digiteo Saclay, Salle Mandelbrot

Abstract

With the steady increase of the power of parallel super-computers, 3D unsteady simulations offer a great potential to study turbulent flows. However, the simulation of highly non-linear phenomena such as turbulence, primary atomization or premixed flames requires very accurate numerical methods and high resolution to capture vortex and interface dynamics. While most direct numerical simulations of turbulent flows are carried out with structured grids or Cartesian-based Adaptive Mesh Refinement, recent advances in numerical methods for tetrahedron-based meshes and parallel mesh adaptation strategies raise the attractiveness of unstructured grids. The use of tetrahedra has two advantages for practical configurations: complex geometries are easily meshed and the mesh is locally more isotropic than Cartesian grids. The first part of the presentation will be focused on the development of highly-efficient dynamic adaptation of tetrahedron-based unstructured grids and on projection methods for low-Mach number flows which can cope with adaptive grids. The proposed methodology, which heavily relies on the remeshing library MMG (www.mmgtools.org) has been thoroughly optimized to reach good performances with grids of several billion cells on more than 10 000 cores. This dynamic mesh adaptation strategy has been implemented in the YALES2 code (www.coria-cfd.fr) and applied to the modeling of turbulent flows in many configurations. In these academic and industrial applications, the local mesh adaptation enabled a drastic reduction of the CPU cost compared to the fixed-grid approach and enabled to reach unprecedented mesh resolutions. The second part of the presentation will be dedicated to the efforts to port these methodologies to GPUs while continuing promoting fast development and innovation from a community of non-GPU experts.

Categories
seminaire passé

[SEMINAR] April 18 2023 – Quantum computers: What are they? What are they supposed to be good at? Will they work? đŸ§‘â€đŸ« by Xavier Waintal 

đŸ§‘â€đŸ« Xavier Waintal, CEA Grenoble, PHELIQS
🌎 April 18 2023
☕ 9:30 AM
🏱 Maison de la Simulation, Batiment Digiteo Saclay, Salle Mandelbrot

Abstract

In the last three decades, our ability to build and control quantum states has improved dramatically and could become the basis of a new form of numerical calculation. While these machines are still in their infancy  (no existing quantum computer can multiply 5 by 3), the hope is that for a very specific class of problem, they could be exponentially faster than their classical counterpart. In this talk, I will give an introduction to quantum computing from the point of view of a physicist. In particular I will emphasize how the main quantum computing ressource, entanglement, is also its biggest problem, decoherence.

Categories
seminaire passé seminar

[SEMINAR] March 2023 – Data-driven modelling of hypersonic flows in non-equilibriumÂ đŸ‘©â€đŸ« Taraneh Sayadi

đŸ‘©â€đŸ« Taraneh Sayadi – d’Alembert, Sorbonne University ITV, RWTH-Aachen University
🌎 March 9 2023

Abstract:

At very large Mach numbers, fluid flows are strongly influenced by non-equilibrium gas effects such as finite-rate chemical reactions or internal mode excitation arising from extreme temperatures. These effects have an order-one influence on quantities of interest, such as stability properties, transition and heating and must be taken into account to achieve effective designs, reliable predictions, and successful flow control.  Accurate simulations of these flows rely on detailed thermochemical gas models (look-up libraries), which dramatically increase the cost of the underlying calculations. In this talk I will first present state-of-the-art detailed simulations of such complex flows and the incurring cost, motivating the second part of the talk where I will present a novel model-agnostic data-driven technique to extract a surrogate of the thermochemical models, reducing the cost of the simulations considerably while maintaining accuracy. 

Slides:

Categories
Non classé seminaire passé seminar

[SEMINAR] January 2023 – Simulations numĂ©riques ab initio de l’irradiation ionisante de la matiĂšreÂ đŸ§‘â€đŸ« AurĂ©llien de la Lande

đŸ§‘â€đŸ« AurĂ©lien de la Lande, CNRS
🌎 January 25 2023

Nous avons dĂ©veloppĂ© Ă  l’Institut de Chimie Physique d’Orsay des approches de simulation ab initio originales pour simuler le dĂ©pĂŽt d’énergie par des ions rapides ou des photons XUV dans des systĂšmes molĂ©culaires de grandes tailles, tels que ceux rencontrĂ©s en biologie1. Durant ce sĂ©minaire, nous intro- duirons les Ă©quations du mouvement des Ă©lectrons dans le cadre de la thĂ©orie de la fonctionnelle de la densitĂ©2. 

Nos codes de simulation reposent sur de nouveaux algorithmes de la thĂ©orie de la fonctionnelle de la densitĂ© dĂ©pendant du temps permettant de simuler des systĂšmes de taille nanomĂ©trique et inhomo- gĂšnes3,4. L’une des astuces principales est de recourir Ă  des densitĂ©s Ă©lectroniques auxiliaires permettant de calculer la rĂ©pulsion coulombienne et les effets liĂ©s Ă  la nature quantique des Ă©lectrons (Ă©change et corrĂ©lation). Le couplage avec la librairie ScaLapack permet une rĂ©duction importante du cout de calcul du propagateur3. Pour aller plus loin une interface avec la libraire Magma a rĂ©cemment Ă©tĂ© rĂ©alisĂ©e. La rĂ©duction du coĂ»t de calcul est remarquable et permet d’entrevoir des applications sans prĂ©cĂ©dent en terme de taille de systĂšmes simulĂ©s. 

J’illustrerai l’apport de ces approches par diverses Ă©tudes rĂ©centes du groupe. Un premier exemple a trait Ă  l’irradiation comparĂ©e d’oligomĂšres d’ADN solvatĂ©s par des protons, de noyaux d’hĂ©lium ou de carbone (Fig. 1)5. L’étude a permis de mettre en Ă©vidence des processus clĂ©s de l’étape physique de l’irradiation ; par exemple le mĂ©canisme d’ionisation par flux-et-reflux du nuage Ă©lectronique, la localisation des Ă©lectrons secondaires ou encore les probabilitĂ©s d’ionisation des bases d’ADN ou du solvant5. Dans un second exemple je mettrai en Ă©vidence un effet de taille remarquable lors le processus d’ionisation d’acide aminĂ©s, de peptides et de protĂ©ines par des photons ionisant XUV. Cette dĂ©couverte permet de faire des hypothĂšses sur les sites d’ionisation primaires possibles du milieu cellulaire par ce type de rayonnement. 

Categories
Non classé seminaire passé

[SEMINAR] January 2023 – Vers une accĂ©lĂ©ration de la mise Ă  l’équilibre des modĂšles de climat đŸ‘©â€đŸ« by Julie DESHAYES

đŸ‘©â€đŸ« Julie Deshayes, CNRS
🌎 January 2023

Abstract:

La modĂ©lisation du climat terrestre est incontournable pour planifier l’adaptation et la mitigation du changement climatique. La rĂ©alisation de scĂ©narios du climat futur a un coĂ»t numĂ©rique consĂ©quent, mĂȘme Ă  basse rĂ©solution spatiale (de l’ordre de 1°). L’essentiel de la consommation en calcul et stockage est consacrĂ©e Ă  la mise Ă  l’équilibre du modĂšle (en particulier sa composante ocĂ©anique) et Ă  la calibration des paramĂštres. En collaboration avec Martial Mancip, et grĂące au soutien du programme PNRIA du CNRS, nous avons commencĂ© Ă  Ă©laborer une solution, basĂ©e sur des mĂ©thodes innovantes (statistiques avancĂ©es et issues de l’intelligence artificielle), pour accĂ©lĂ©rer la mise Ă  l’équilibre de l’ocĂ©an du modĂšle de climat de l’IPSL. L’idĂ©e est de disposer d’un modĂšle d’infĂ©rence qui extrapole une sĂ©rie de pas de temps de simulation (en mois/annĂ©es), puis rĂ©injecte la solution extrapolĂ©e dans le modĂšle climatique pour effectuer de nouvelles Ă©tapes de simulation. Ces deux Ă©tapes seraient rĂ©pĂ©tĂ©es autant de fois que nĂ©cessaire pour obtenir un algorithme stable qui converge vers une solution physiquement admissible comparable aux Ă©quilibres complets, tout en rĂ©duisant considĂ©rablement le nombre de pas de temps calculĂ©s explicitement avec le modĂšle. La rĂ©duction du temps de calcul explicite par le modĂšle climatique (opĂ©rĂ© sous CPU), qui est plus coĂ»teux que l’infĂ©rence par les techniques de Data Science (rĂ©alisĂ©e sous GPU), conduit Ă  une amĂ©lioration de la frugalitĂ© du calcul numĂ©rique de la modĂ©lisation climatique.

Categories
seminaire passé

[SEMINAR] March 2020 – Two-level Coarse Corrected Optimized Schwarz Methods using PETSc đŸ§‘â€đŸ« Serge Van Criekingen

đŸ§‘â€đŸ« Serge Van Criekingen, IDRIS, France
🌎 March 2020

Abstract

Parallel Schwarz-type domain decomposition methods are based on an iterative process where, at each iteration, a local solve is simultaneously performed on each of the (possibly overlapping) subdomains, using interface values previously computed on neighboring subdomains. The reference method in this framework is the Restricted Additive Schwarz (RAS) method, implemented as a preconditioner in the PETSc library. Using existing PETSc tools, we here implement two improvements to this method: a new coarse correction to obtain a two-level scalable method, as well as optimized transmission conditions, resulting in an Optimized 2-level Restricted Additive Schwarz method.

The first improvement, namely the introduction of a coarse correction to insure scalability, is wellknown and due to that fact that, in the case of elliptic problems, information is only transferred from each subdomain to its direct neighbors at each iteration of a 1-level method such as RAS. This makes the number of iterations grow with the number of subdomains. Scalability is achieved by introducing a coarse grid on which a reduced-size calculation is performed, yielding a coarse correction at each iteration of the solution process. Such a 2-level method permits global propagation of the iterative corrections throughout the entire domain, leading to the scalability of the method. Many choices for the coarse grid point locations are possible, and we here follow a method introduced by M.J. Gander et al. yielding a reduced number of iterations.

The second improvement, namely optimized transmission conditions, stems from the idea that the transmission conditions used in the iterative process at subdomain interfaces can also be chosen such as to reduce the number of iterations. In our case, we consider Robin transmission conditions instead of the classical Dirichlet ones, i.e. a well-chosen combination of Dirichlet and Neumann values at subdomain interfaces. A good choice of the Robin coefficient representing the relative weight of Dirichlet and Neumann values permits minimizing the number of iterations, which led to the name Optimized Schwarz Methods.

We combine these two improvements and apply them to a 2D Laplace test case up to 16,384 CPU cores. We obtain substantially improved computation times, comparable to the ones obtained with the multigrid library HYPRE interfaced by PETSc. This is significant in that Schwarz-type domain decomposition methods were up to now not considered competitive with multigrid methods on this type of problem. Furthermore, we extend the method to non-symmetric problems, adding an advection term to the Laplacian, and investigate various ways of adapting the coarse space.

Categories
seminaire passé

[SEMINAR] March 2019 – Emulating quantum computers đŸ§‘â€đŸ« Victor Alessandrini

đŸ§‘â€đŸ« Victor Alessandrini, Maison de la Simulation, France
🌎 March 2019

Abstract:

There is today increasing consensus on the fact that quantum computing – an emerging data processing technology – may in the future play a significant role (with not yet fully understood boundaries) in high performance scientific computing. The simulation of quantum computers on standard computing platforms is today a necessary step to understand, assess, and develop quantum algorithms for computation, paving the way for the eventual future adoption of this disruptive technology. Emulation software will remain useful for some time to validate results of the earlier quantum computing platforms, and to help tuning application quantum codes. We will present a fully portable C++ emulation library under development for more than one year. The library disposes of three interfaces: a shared memory version running on SMP nodes, a MPI extension enabling access to a larger number of qubits, and a Python wrapper of the C++ library. A significant pedagogical effort is implemented in the documentation. Besides the traditional C++ class documentation, we are producing strategic higher level documentation and pedagogical papers on applications, including Jupyter notebooks. This seminar will propose first an introduction to quantum computing, explaining the fundamental differences with classical computing, underlining the strong features as well as the limiting bottlenecks of the technology. We will present next a short and very high level overview of the emulation library, focusing on a few major strategic choices. Then, the issues involved in quantum algorithms will be illustrated with a quantum chemistry example used in the validation tests of the software. Finally, we will conclude with a rapid overview of different areas in which quantum algorithm research is evolving (quantum chemistry, condensed matter physics, combinatorial optimization, fault tolerant quantum computing,
).

Categories
Non classé seminaire passé

[SEMINAR] February 2019 – Hercules đŸ§‘â€đŸ« Olivier Bressand

đŸ§‘â€đŸ« Olivier Bressand, CEA-DAM, France
🌎 February 2019

Abstract

Hercules is a CEA-DAM platform for managing data produced by simulation codes. It integrates different I/O services to read, write in parallel database in the framework of protection / recovery, intercode (coupling or code sequence) and post-processing (visualization and analysis). It is based on a data model that covers many domains of simulation (structured, unstructured, AMR-block, AMR-tree based, multi-fluid, laser, atom, Euler, Lagrange, ale, 1D, 2D, 3D) and provides services to produce, filter, and disaggregate data in the sequential or parallel HPC application.

Categories
Non classé seminaire passé

[SEMINAR] February 2019 – MPC, The Multi-Processor Computing Framework đŸ§‘â€đŸ« Julien Jaeger

đŸ§‘â€đŸ« Julien Jaeger, CEA-DAM, France
🌎 February 2019

Abstract

The MPC (Multi-Processor Computing) framework provides a unified parallel runtime designed to improve the scalability and performances of applications running on clusters of (very) large multiprocessor/multicore NUMA nodes. Thanks to its design, MPC allows mixed-mode programming models and efficient interaction with the HPC software stack. MPC provides implementations for the MPI, OpenMP and POSIX Threads standards. All these standards can be mixed together in an efficient way, thanks to process virtualization, and the sharing of information and resources