Le but de ce site est de rassembler des informations pour un public cherchant à porter et optimiser des applications
HPC (High Performance Computing) sur GPU (
Graphics Processing Unit).
GPU@EPFL
NVIDIA's June/July 2010 GPU Computing Webinars are now open for registration.
The new webinars series features many topics presented by invited guest speakers in addition to NVIDIA staff.
New webinars include an overview of MAGMA by Prof Jack Dongarra, DirectCompute presentation by Matt Sandy from Microsoft, and presentations
by Accelereyes, EM Photonics, MainConcept and others. See below for details.
- DirectCompute for GPU Computing - an Introduction by Microsoft 1 Hour
Friday, June 4, 2010 9:00 AM PDT https://www2.gotomeeting.com/register/293947915
- An Introduction to the MAGMA project - acceleration of dense linear algebra, 1.5 hrs
Monday, June 14, 2010 9:00 AM PDT https://www2.gotomeeting.com/register/565103979
- An Introduction to CULA GPU Accelerated Linear Algebra, 1.5 hrs
Tuesday, June 15, 2010 9:00 AM https://www2.gotomeeting.com/register/357212371
- Rapid Application Development platform for GPGPUs - Jacket with MATLAB, 1.5 hrs
Tuesday, June 22, 2010 8:00 AM https://www2.gotomeeting.com/register/502467706
- Introduction to MainConcept's CUDA H.264/AVC Encoder, 1.5 hrs
Tuesday, June 29, 2010 9:00 AM https://www2.gotomeeting.com/register/882669387
- An Introduction to OpenCL Application Development with gDEBugger , 1.5 hrs
Tuesday, July 6, 2010 9:00 AM https://www2.gotomeeting.com/register/784334826
- Rapid Prototyping and Visualization with OpenCL Studio, 1.5 hrs
Wednesday, July 7, 2010 9:00 AM https://www2.gotomeeting.com/register/616979594
- Thrust, A C++ Standard Template Library for CUDA C - An Introduction, 1.5 hrs
Thursday, July 1, 2010 9:00 AM PDT https://www2.gotomeeting.com/register/870778442
- PGI Fortran - An Introduction and Overview, 2 Hours
Tuesday June 8, 2010 9:00 PM https://www2.gotomeeting.com/register/751486762
Wednesday June 9, 2010 9:00 AM https://www2.gotomeeting.com/register/794118451
- GPU Computing using CUDA C - An Introduction, 2 Hours
Wednesday, June 30, 2010 9:00 AM - 10:30 AM PDT https://www2.gotomeeting.com/register/820921322
- GPU Computing using CUDA C - Advanced 1, 2 hours
Wednesday, July 14, 2010 9:00 AM PDT https://www2.gotomeeting.com/register/601855626
- GPU Computing using CUDA C - Advanced 2, 1.5 hours
Wednesday, July 21, 2010 9:00 AM PDT https://www2.gotomeeting.com/register/317429427
- GPU Computing and OpenCL\u2122 - An Introduction, 1.5 hours
Tuesday, July 13, 2010 9:00 AM PDT https://www2.gotomeeting.com/register/594355834
- GPU Computing and OpenCL\u2122 - Advanced 1, 2 hours
Thursday, July 15, 2010 9:00 AM PDT https://www2.gotomeeting.com/register/679215722
LMMM - Laboratoire de modélisation multi-échelles des matériaux:
Parallel Multi-Scale Hemodynamics
We outline a computational method for clinical cardiovascular diagnosis based on massive simulations running on the IBM Blue Gene architecture and Graphical Processing Units (GPU) commodity hardware. Our approach is based on the combination of microscopic Molecular Dynamics (MD) with a hydrokinetic Lattice Boltzmann (LB) method and takes into explicit account the hydrodynamic interactions among the suspended bodies and the surrounding fluid. The end result is a multi-physics/scale parallel code, called MUPHY. The software exhibits excellent scalability on the Blue Gene/L platform and multi-GPUs. By developing a series of ad hoc optimization techniques, MUPHY proves extremely efficient in simulating irregular domains, such as for blood flows in extended coronary arteries, with the capability of accessing physical information at micrometric scale.
DIT makes available to the EPFL community two experimental HPC platforms: one is based on the Cell processor technology and the other on the GPU as used for general purpose computations. The goal of these experimental clusters is to offer the possibility to evaluate and test these new technologies.
NEWS
8/12/09At the SC09 conference and exhibition in Portland, NVIDIA® showcased the latest research and trends in GPU computing in a
full-day CUDA tutorial, the GPU computing poster showcase, and in its GPU Computing Theater.
Presentations and recordings presented by NVIDIA are now available for download at www.nvidia.com/sc09 including:
- Presentation for the full-day tutorial on "High Performance Computing with CUDA" (http://www.nvidia.com/object/SC09_Tutorial.html)
- Presentations from the GPU Computing Theater by computational researchers and industry luminaries such as
- o Jack Dongarra, University of Tennessee
- Jeff Vetter, Oak Ridge National Laboratory
- Satoshi Matsuoka, Tokyo Institute of Technology
- Pat McCornick, Los Alamos National Laboratory
- Paul Crozier, Sandia National Laboratories
- Ross Walker, San Diego Supercomputing Center / UC San Diego
- Mike Clark, Harvard University
- Posters showcasing "what's next" for GPU computing out of the world of research and academia
11/18/09 New GPU Computing Collaboration Network for Scientists, Researchers, and Developers Launches at SC09
.
The Coordinated Science Laboratory of the University of Illinois at Urbana-Champaign, home of the first CUDA Center of Excellence, announced the
launch of gpucomputing.net a network of research communities designed to foster collaboration among scientists, researchers, and developers utilizing GPU computing.
GPU Gems 3