Menu

Tristan CABEL

Sophia Antipolis

En résumé

I am an engineer specialized in high performance computing using new specific hardware such as GPUs. I self taught and mastered GPU programming in February 2009 and since then, I continued to learn and stay abreast of new technologies and software regarding HPC.

Programming : Fortran, C/C++, Python, VHDL,
Tools/Librairies: CUDA, OpenCL, OpenMP, MPI, StarPU, Metis, Doxygen
Software : GCC, Emacs , GDB, Allinea DDT, Valegrind, CMake, Git,

Mes compétences :
MPI
Fortran
Cuda
C
Simulation numérique
HPC
OpenCL
C++

Entreprises

  • Amadeus Nice - Data Science Engineer

    Sophia Antipolis 2016 - maintenant

  • INRIA - Ingénieur de recherche expert

    Le Chesnay 2014 - 2016 I joined INRIA under the large-scale initiative "Computer and Computational Sciences at Exascale" ( http://www-sop.inria.fr/c2s_at_exa/ ) with the aim of improving a software developped by ANDRA for the simulation of radioactive waste management named TRACES. To this end, I used latest HPC developments and inria technologies including :
    - numerical simulation platform num3sis ( http://num3sis.inria.fr/ )
    - MPI 3.0 RMA
    - visualisation with VTK, live visualisation with Paraview Catalyst
    - IO with parallel HDF5
    - Sparse linear solver (HYPRE, Maphys)

  • CINES - Research Engineer

    2014 - 2014 - User support (level 2) on the cines main Supercomputer
    - Participation in some of the PRACE research projects ):
    -- WP5 with the SHAPE project (help an Irish SME doing bio-informatics on Intel MIC)
    -- WP6 testing iRODS and gridftp on PRACE network
    -- WP7 benchmarking HPC applications on different clusters (BullX, CRAY,..)
    -- Prace Preparatory access expert to help researchers in the parallelization and optimization of their code

  • National University of Singapore - Associate Scientist

    2011 - 2013 My work at TemasekLab@NUS consists of improving a discontinuous Galerkin solver applied to Maxwell equations. In order to do it, I uppgraded the code to use OpenCL 1.1 on a Nvidia GPUs and then optimized it.
    After that, I worked on the integration of StarPU in order to use a CPU and a GPU concurrently.
    Finally, I adapted this code to the modelisation of reflectarray antennas.

    I was also involved in the purchase of a GPU workstation: I took part into the creation of the specifications, the review of the different offers , the OS (CentOS 6) and software installation, and finally the administrative responsibilities (installation of software, creation and management of users’ accounts and answering queries).

  • Alyotech - Développeur CUDA

    Antony Cedex. 2009 - 2009 Stage de fin d'études. Le sujet était : "Portages d'applications scientifiques sur GPU". Pour cela, je me suis formé à l'API CUDA de NVIDIA et j'ai développé un algorithme de clipping(calcul de faces vues) fonctionnant en parallèle. Ensuite, j'ai retenu une implémentation hybride : certaines opérations se font sur le GPU(Graphics Processor Unit) et d'autres sur le CPU.

  • INRIA Sophia Antipolis - Ingénieur de recherche

    Le Chesnay 2009 - 2011 I joined INRIA team Nachos in the aim of adapting a software developed in Fortran 77 to use GPU in order to improve performances. The software is a finite element solver using a discontinuous Galerkin method applied to electromagnetic Maxwell equations in time domain.

    I started by improving a first parallel version of an initial software developed in Fortran 90 to use CUDA and MPI.

    Results of this project were presented in the INRIA booth at SC10 (International Conference for High Performance Computing, Networking, Storage and Analysis , New Orleans, Louisiana , November 2010) and a research report about an application of my work is available under the title : "Multi-GPU acceleration of a DGTD method for modeling human exposure to electromagnetic waves" ( http://hal.inria.fr/CNRS/inria-00583617 ) .

    Then, I continued the development of the software to use a new generation of NVIDIA's graphic cards architecture “Fermi” and CUDA 4.0. At the same time I was involved in the development of a C++ discontinuous Galerkin library using GPUs in a numerical simulation software called num3sis ( http://num3sis.inria.fr ).

  • Nokor-it - Développeur

    2008 - 2008 Développement de tests unitaires en environnement JAVA au Cambodge. J'ai utiliser les frameworks suivants : Hibernate, Junit, EasyMock, Selenium. Ce stage ayant été effectué au Cambodge, il m'a permis de découvrir une culture totalement différente et m'a appris à m'intégrer dans un environnement nouveau.

Formations

Réseau

Annuaire des membres :