DALI: Digits, architectures et logiciels informatiques

L’équipe DALI développe une thématique de recherche unifiée afin d’améliorer la qualité numérique et la haute performance des calculs.  DALI permet l’interaction, rare en France au sein d’une même équipe, d’experts en micro-architecture et en arithmétique des ordinateurs.
Côté performances, nos travaux portent sur l’exploitation du potentiel de calcul toujours croissant des processeurs : élargissement des chemins (micro-architecture vectorielle), multiplication des cœurs (parallélisme de tâches), augmentation du parallélisme d’instructions. Côté arithmétique, la qualité numérique des applications de calcul  scientifique et la sûreté de fonctionnement d’applications embarquées dépendent crucialement de la maîtrise de la précision finie et de l’arithmétique flottante en particulier. Il s’agit de contrôler et certifier les calculs (algorithmes, codes) mais aussi d’optimiser la précision des résultats. De nombreux logiciels, scientifiques ou embarqués, nécessitent d’améliorer la qualité numérique sans pour autant sacrifier la rapidité d’exécution.  Ainsi se rejoignent amélioration de la performance et de la qualité numérique.

Membres

Permanents

Non permanents

Thématiques de recherche

La cohérence thématique des travaux de recherche sur l’amélioration de la qualité et de la performance des calculs est une des forces de l’équipe DALI.

L’amélioration de la performance des calculs est étroitement liée aux améliorations apportées aux micro-architectures. Celle-ci est réalisée suivant plusieurs directions, par élargissement des chemins (micro-architecture vectorielle), multiplication des cœurs (parallélisme de tâches), ou encore augmentation du parallélisme d’instructions (ILP). La qualité numérique des applications de calcul scientifique ou la sûreté de fonctionnement d’applications embarquées critiques dépendent crucialement de la maîtrise des effets de la précision finie des calculs — et de l’arithmétique flottante en particulier. Il s’agit alors de contrôler et valider les calculs (algorithmes, codes) mais aussi d’améliorer et optimiser la précision des calculs et des résultats.

Les travaux développés sur la période 2008-2013 sont organisés autour de 4 actions de recherche :

  • Action 1. Mesure reproductible et analyse du potentiel de parallélisme et des performances.
  • Action 2. Meilleure exploitation des nouvelles architectures multicœurs.
  • Action 3. Implantation sûre et efficace de protocoles cryptographiques.
  • Action 4. Synthèse de code pour l’implémentation de calculs précis, rapides et certifiés.

Faits marquants

  • Intégration de l’équipe DALI au LIRMM le 1er janvier 2011.
  • Organisation de 4 manifestations scientifiques : 17th Static Analysis Symposium (SAS’10), Ecole thématique Archi’11, Rencontres Arithmétique et Informatique Mathématique (RAIM’11), Ecole jeunes chercheurs du GDR Informatique Mathématique (EJCIM’13).
  • Obtention du Prix « chercheur d’avenir » de la région Languedoc-Roussillon en 2010.
  • Best Poster Award DASIP’2012, Karlsruhe, Germany.
  • Inclusion d’une bibliothèque de calcul par intervalles au SDK CUDA de NVidia.

Partenariats et collaborations académiques

Partenaires industriels : Actility, Airbus, Inpixal, Prover, Rockwell-Collins, Thales, Total.
Projets collaboratifs : ANR Blanc EvaFlo (2006-2010), Projet DPAC MASSANE (2007-2010), ANR ARPEGE (2009-2011), Projet FNRAE SARDANES (2009-2012), Projet Chercheur d’avenir Compil’HD (2010-2012), ANR INS DEFIS (2011-2014), ANR INS CAFEIN (2012-2015), PEPS QUARENUM (2013).

Collaborations nationales : CEA-LIST, ENSTA, EXASCALE, IRISA, LIENS, LIP, LIP6, LRI, LSIS, ONERA.
Collaborations internationales : Universitat de Girona, Technical University Hamburg, University of Malaysia Sabah, Mississippi State University, Microsoft Research Redmond, Rice University, Tokyo WCU, University of Waterloo (Canada), University of Wollongong, University of Waseda.

Publications majeures

  • Bernard Goossens and David Parello, Limits of Instruction-Level Parallelism Capture, International Conference on Computational Science (ALCHEMY Workshop), 2013, to appear.
  • Stef Graillat, Philippe Langlois, and Nicolas Louvet. Algorithms for accurate, validated and fast computations with polynomials. Japan Journal of Industrial and Applied Mathematics, 26(2,3):191-214, 2009.
  • Sylvain Collange, Marc Daumas and David Defour, Interval Arithmetic, in GPU Computing Gems, Jade Edition, ISBN 978-0-12-385963-1, 2011.
  • Anwar Hasan, Nicolas Meloni, Ashkan Namin, and Christophe Negre,Block Recombination Approach for Subquadratic Space Complexity Binary Field Multiplication based on Toeplitz Matrix-Vector Product ,IEEE Transactions on Computers, Vol 61(2), pages 151-163, 2012.
  • Arnault Ioualalen and Matthieu Martel, A New Abstract Domain for the Representation of Mathematically Equivalent Expressions, Static Analysis Symposium, SAS'12, Lecture Notes in Computer Science, Volume 7460, pages 75-93, Springer-Verlag, 2012.
  • Claude-Pierre Jeannerod, Hervé Knochel, Christophe Monat, and Guillaume Revy, Computing floating-point square roots via bivariate polynomial evaluation, IEEE Transactions on Computers, Vol. 60(2), pages 214-227, February 2011.

Publications depuis 2014 - Evaluation 2019

Articles de revues internationales

2019

  1. Efficient Fixed Base Exponentiation and Scalar Multiplication based on a Multiplicative Splitting Exponent Recoding
    Jean-Marc Robert, Christophe Negre, Thomas Plantard
    Journal of Cryptographic Engineering, Springer, 2019, 9 (2), pp.115-136.
  2. Hierarchical approach for deriving a reproducible unblocked LU factorization
    Roman Iakymchuk, Stef Graillat, David Defour, Enrique Quintana-Ortí
    International Journal of High Performance Computing Applications, SAGE Publications, 2019, pp.#1094342019832968.

2017

  1. Midpoint-Radius Interval-based Method to Deal with Uncertainty in Power Flow Analysis
    Manuel Marin, Federico Milano, David Defour
    Electric Power Systems Research, Elsevier, 2017, 147, pp.81-87.
  2. Trade-offs of certified fixed-point code synthesis for linear algebra basic blocks
    Matthieu Martel, Mohamed Amine Najahi, Guillaume Revy
    Journal of Systems Architecture, Elsevier, 2017, 76, pp.133-148.
  3. Computing On Many Cores
    Bernard Goossens, David Parello, Katarzyna Porada, Djallal Rahmoune
    Concurrency and Computation: Practice and Experience, Wiley, 2017, 29 (15), pp.e4120.
  4. Exact Lookup Tables for the Evaluation of Trigonometric and Hyperbolic Functions
    Hugues de Lassus Saint-Geniès, David Defour, Guillaume Revy
    IEEE Transactions on Computers, Institute of Electrical and Electronics Engineers, 2017, 66 (12), pp.2058-2071.
  5. Automatic source-to-source error compensation of floating-point programs: code synthesis to optimize accuracy and time
    Laurent Thévenoux, Philippe Langlois, Matthieu Martel
    Concurrency and Computation: Practice and Experience, Wiley, 2017, Concurrency and Computation: Practice and Experience, 29 (7), pp.e3953.
  6. Efficient Regular Modular Exponentiation Using Multiplicative Half-Size Splitting
    Christophe Negre, Thomas Plantard
    Journal of Cryptographic Engineering, Springer, 2017, 7 (3), pp.245-253.

2016

  1. An efficient representation format for fuzzy intervals based on symmetric membership functions
    Manuel Marin, David Defour, Federico Milano
    ACM Transactions on Mathematical Software, Association for Computing Machinery, 2016, 43 (3), pp.23:1--23:22.
  2. A software scheduling solution to avoid corrupted units on GPUs
    David Defour, Eric Petit
    Journal of Parallel and Distributed Computing, Elsevier, 2016, 90-91, pp.1--8.

2015

  1. Numerical Reproducibility for the Parallel Reduction on Multi- and Many-Core Architectures
    Sylvain Collange, David Defour, Stef Graillat, Roman Iakymchuk
    Parallel Computing, Elsevier, 2015, 49, pp.83-97.
  2. Transformation of a PID Controller for Numerical Accuracy
    Nasrine Damouche, Matthieu Martel, Alexandre Chapoutot
    Electronic Notes in Theoretical Computer Science, Elsevier, 2015, 317, pp.47-54.
  3. New Parallel Approaches for Scalar Multiplication in Elliptic Curve over Fields of Small Characteristic
    Christophe Negre, Jean-Marc Robert
    IEEE Transactions on Computers, Institute of Electrical and Electronics Engineers, 2015, 64 (10), pp.2875-2890.

2014

  1. Efficient Subquadratic Space Complexity Binary Polynomial Multipliers Based On Block Recombination
    Murat Cenk, Anwar Hasan, Christophe Negre
    IEEE Transactions on Computers, Institute of Electrical and Electronics Engineers, 2014, 63 (9), pp.2273-2287.
  2. First steps towards more numerical reproducibility
    Fabienne Jézéquel, Philippe Langlois, Nathalie Revol
    ESAIM: Proceedings and Surveys, EDP Sciences, 2014, ESAIM: Proceedings and Surveys, 45, pp.229-238.
  3. Efficient Binary Polynomial Multiplication Based on Optimized Karatsuba Reconstruction
    Christophe Negre
    Journal of Cryptographic Engineering, Springer, 2014, 4 (2), pp.91--106.
  4. A Fast Chaos-Based Pseudo-Random Bit Generator Using Binary64 Floating-Point Arithmetic
    Michael François, David Defour, Christophe Negre
    Informatica, Slovene Society Informatika, Ljubljana, 2014, 38 (2), pp.115-124.
  5. Pseudo-random number generator based on mixing of three chaotic maps
    Michael François, Thomas Grosges, Dominique Barchiesi, Robert Erra
    Communications in Nonlinear Science and Numerical Simulation, Elsevier, 2014, 19 (4), pp.887--895.

Communications internationales

2018

  1. Performance optimization of the air shower simulation program for the Cherenkov Telescope Array
    Luisa Arrabito, Konrad Bernlöhr, Johan Bregeon, Gernot Maier, Philippe Langlois, David Parello, Guillaume Revy
    CHEP: Computing in High Energy and Nuclear Physics, Jul 2018, Sofia, Bulgaria. <http://chep2018.org>
  2. Meta-implementation of vectorized logarithm function in binary floating-point arithmetic
    Hugues de Lassus Saint-Geniès, Nicolas Brunie, Guillaume Revy
    ASAP: Application-specific Systems, Architectures and Processors, Jul 2018, Milan, Italy. <https://asap18.necst.it>
  3. Numerical Accuracy Stuff: Tools. . . and Prerequisites
    Philippe Langlois
    CTAOptSim General Workshop, Dec 2018, Montpellier, France.

2017

  1. Asynchronous Power Flow on Graphic Processing Units
    Manuel Marin, David Defour, Federico Milano
    PDP: Parallel, Distributed and network-Based Processing, Mar 2017, St Petersburg, Russia.
  2. Efficient Leak Resistant Modular Exponentiation in RNS
    Andrea Lesavourey, Christophe Negre, Thomas Plantard
    ARITH: Computer Arithmetic, Jul 2017, London, United Kingdom. pp.156-163.
  3. Reproducible Parallel Simulations in HPC
    Chemseddine Chohra, Philippe Langlois, Rafife Nheilli, David Parello
    CSE: Computational Science and Engineering, Feb 2017, Altanta, Georgia, United States. <https://archive.siam.org/meetings/cse17/>

2016

  1. Hierarchical Approach for Deriving a Reproducible LU factorization on GPUs
    Roman Iakymchuk, Stef Graillat, David Defour, Enrique Quintana-Ortí
    The Numerical Reproducibility at Exascale (NRE16) workshop held as part of the Supercomputing Conference (SC16), Nov 2016, Salt Lake City, UT, United States.
  2. First improvements toward a reproducible Telemac-2D
    Rafife Nheili, Philippe Langlois, Christophe Denis
    XXIIIrd TELEMAC-MASCARET User Conference , Oct 2016, Paris, France. <http://www.opentelemac.org/index.php/user-conference26>
  3. Reproducible and Accurate Algorithms for Numerical Linear Algebra
    Roman Iakymchuk, David Defour, Sylvain Collange, Stef Graillat
    PP: Parallel Processing for Scientific Computing, Apr 2016, Paris, France.
  4. Reproducible and Accurate Matrix Multiplication
    Roman Iakymchuk, David Defour, Sylvain Collange, Stef Graillat
    SCAN: Scientific Computing, Computer Arithmetic and Validated Numerics, Sep 2014, Wurzburg, Germany. pp.126-137.
  5. Recovering numerical reproducibility in hydrodynamic simulations
    Philippe Langlois, Rafife Nheili, Christophe Denis
    ARITH: Computer Arithmetic, Jul 2016, Silicon Valley, Santa Clara, CA, United States. pp.63-70.
  6. Reproducible, Accurately Rounded and Efficient BLAS
    Chemseddine Chohra, Philippe Langlois, David Parello
    Euro-Par: Parallel Processing Workshops., Aug 2016, Grenoble, France. pp.609-620.
  7. Enhanced Digital Signature using RNS Digit Exponent Representation
    Thomas Plantard, Jean-Marc Robert
    WAIFI: Workshop on the Arithmetic of Finite Fields, Department of Mathematics of Ghent University, Jul 2016, Gand, Belgium. pp.177-192.
  8. Parallel Locality and Parallelization Quality
    Bernard Goossens, David Parello, Katarzyna Porada, Djallal Rahmoune
    PMAM: Programming Models and Applications for Multicores and Manycores, Mar 2016, Barcelona, Spain. pp.59-68.
  9. Automated design of floating-point logarithm functions on integer processors
    Guillaume Revy
    ARITH: Computer Arithmetic, Jul 2016, Silicon Valley, Santa Clara, CA, United States. pp.172-180.
  10. Efficient Randomized Regular Modular Exponentiation using Combined Montgomery and Barrett Multiplications
    Andrea Lesavourey, Christophe Negre, Thomas Plantard
    ICETE: International Joint Conference on e-Business and Telecommunications, Jul 2016, Lisbon, Portugal. pp.368-375.
  11. Parallel experiments with RARE-BLAS
    Chemseddine Chohra, Philippe Langlois, David Parello
    SYNASC: Symbolic and Numeric Algorithms for Scientific Computing, Sep 2016, Timisoara, Romania. pp.135-138.
  12. Towards Fast, Accurate and Reproducible LU Factorization
    Roman Iakymchuk, David Defour, Stef Graillat
    SCAN 2016, 17th international symposium on Scientific Computing, Computer Arithmetic and Validated Numerics, Sep 2016, Uppsala, Sweden. pp.59-60.

2015

  1. ExBLAS: Reproducible and Accurate BLAS Library
    Roman Iakymchuk, Sylvain Collange, David Defour, Stef Graillat
    NRE: Numerical Reproducibility at Exascale, Nov 2015, Austin, TX, United States.
  2. Automatic Source-to-Source Error Compensation of Floating-Point Programs
    Laurent Thévenoux, Philippe Langlois, Matthieu Martel
    Computational Science and Engineering (CSE), Oct 2015, Porto, Portugal. pp.9--16.
  3. Measuring predictability of Nvidia’s GPU warp and block schedulers: Application to the summation problem
    David Defour
    MCSoC: Embedded Multicore/Many-core Systems-on-Chip, Sep 2015, Turin, Italy. pp.17-24.
  4. Reproducible floating-point atomic addition in data-parallel environment
    David Defour, Sylvain Collange
    ACSIS, Sep 2015, Lodz, Poland. pp.721-728.
  5. Toward a Core Design to Distribute an Execution on a Many-Core Processor
    Bernard Goossens, David Parello, Katarzyna Porada, Djallal Rahmoune
    PaCT: Parallel Computing Technologies, Aug 2015, Petrozavodsk, Russia. pp.390-404.
  6. Range Reduction Based on Pythagorean Triples for Trigonometric Function Evaluation
    Hugues de Lassus Saint-Geniès, David Defour, Guillaume Revy
    ASAP: Application-specific Systems, Architectures and Processors, Jul 2015, Toronto, Canada. pp.74-81.
  7. Numerical Reproducibility: Feasibility Issues
    Philippe Langlois, Rafife Nheili, Christophe Denis
    NTMS: New Technologies, Mobility and Security, Jul 2015, Paris, France. pp.1-5.
  8. Parallel Approaches for Efficient Scalar Multiplication over Elliptic Curve
    Christophe Negre, Jean-Marc Robert
    Multiplication over Elliptic Curve. SECRYPT, Jul 2015, Colmar, France. pp.202-209.
  9. Numerical Reproducibility in open TELEMAC: A Case Study within the Tomawac Library
    Rafife Nheili, Philippe Langlois, Christophe Denis
    HPCSET: High Performance Computing Simulation in Energy/Transport Domains, Jul 2015, Frankfurt, Germany.
  10. Trade-off Approaches for Leak Resistant Modular Arithmetic in RNS
    Christophe Negre, Guilherme Perin
    ACISP: Australasian Conference on Information Security and Privacy, Jun 2015, Brisbane, Australia. pp.107-124.
  11. Intra-procedural Optimization of the Numerical Accuracy of Programs
    Nasrine Damouche, Matthieu Martel, Alexandre Chapoutot
    FMICS: Formal Methods for Industrial Critical Systems, Jun 2015, Oslo, Norway. pp.31-46.
  12. Efficient Modular Exponentiation Based on Multiple Multiplications by a Common Operand
    Christophe Negre, Thomas Plantard, Jean-Marc Robert
    ARITH: Computer Arithmetic, INRIA, Jun 2015, Lyon, France. pp.144-151.
  13. Optimizing the Accuracy of a Rocket Trajectory Simulation by Program Transformation
    Nasrine Damouche, Matthieu Martel, Alexandre Chapoutot
    CF: Computing Frontiers, May 2015, Ischia, Italy. pp.40.
  14. Reproducible Triangular Solvers for High-Performance Computing
    Roman Iakymchuk, David Defour, Sylvain Collange, Stef Graillat
    ITNG: Information Technology - New Generations, Apr 2015, Las Vegas, NV, United States. pp.353-358.
  15. An efficient midpoint-radius implementation to handle symmetric fuzzy intervals
    Manuel Marin, David Defour, Federico Milano
    RAIM: Rencontres Arithmétiques de l’Informatique Mathématique, Apr 2015, Rennes, France.

2014

  1. Software Implementation of Parallelized ECSM over Binary and Prime Fields
    Jean-Marc Robert
    Inscrypt: Information Security and Cryptology, Dec 2014, Beijing, China. pp.445-462.
  2. Toward the synthesis of fixed-point code for matrix inversion based on Cholesky decomposition
    Matthieu Martel, Mohamed Amine Najahi, Guillaume Revy
    DASIP: Design and Architectures for Signal and Image Processing, Oct 2014, Madrid, Spain. pp.1-8.
  3. Automated Synthesis of Target-Dependent Programs for Polynomial Evaluation in Fixed-Point Arithmetic
    Christophe Mouilleron, Mohamed Amine Najahi, Guillaume Revy
    SYNASC: Symposium on Symbolic and Numeric Algorithms for Scientific Computing, Sep 2014, Timisoara, Romania. pp.141-148.
  4. Level 1 Parallel RTN-BLAS: Implementation and Efficiency Analysis
    Chemseddine Chohra, Philippe Langlois, David Parello
    SCAN: Scientific Computing, Computer Arithmetic and Validated Numerics, Sep 2014, Wurzburg, Germany. <http://www.scan2014.uni-wuerzburg.de/talks/>
  5. Efficiency of Reproducible Level 1 BLAS
    Chemseddine Chohra, Philippe Langlois, David Parello
    SCAN: Scientific Computing, Computer Arithmetic, and Validated Numerics, Sep 2014, Würzburg, Germany. pp.99-108.
  6. Power Flow Analysis under Uncertainty using Symmetric Fuzzy Arithmetic
    Manuel Marin, David Defour, Federico Milano
    PES General Meeting 2014 | Conference & Exposition, Jul 2014, National Harbor, MD, United States. pp.1-5.
  7. A Pseudo-Random Bit Generator Based on Three Chaotic Logistic Maps and IEEE 754-2008 Floating-Point Arithmetic
    Michael François, David Defour, Pascal Berthomé
    Theory and Applications of Models of Computation, Apr 2014, Chennai, India. pp.229-247.
  8. FuzzyGPU : a fuzzy arithmetic library for GPU
    Manuel Marin, David Defour
    PDP: Parallel, Distributed and Network-Based Processing, Feb 2014, Torino, Italy. pp.624-631.
  9. Code Size and Accuracy-Aware Synthesis of Fixed-Point Programs for Matrix Multiplication
    Matthieu Martel, Mohamed Amine Najahi, Guillaume Revy
    PECCS: Pervasive and Embedded Computing and Communication Systems, Jan 2014, Lisbonne, Portugal. <10.5220/0004884802040214>
  10. Reproducible and Accurate Matrix Multiplication for High-Performance Computing
    Sylvain Collange, David Defour, Stef Graillat, Roman Iakymchuk
    SCAN: Scientific Computing, Computer Arithmetic and Validated Numerics, Sep 2014, Wuerzburg, Germany. pp.42-43.
  11. A Reproducible Accurate Summation Algorithm for High-Performance Computing
    Sylvain Collange, David Defour, Stef Graillat, Roman Iakymchuk
    EX: Exascale Applied Mathematics Challenges and Opportunities, Jul 2014, Chicago, United States. <http://www.siam.org/meetings/ex14/>

Mots-clés

Micro-architecture des processeurs, Simulation d’unités de calcul, Arithmetique des ordinateurs, Calcul certifié, Synthèse de code, Operateurs cryptographiques, Logiciel numerique, Calcul haute performance

Dernière mise à jour le 09/01/2019