Programming Graphic Cards for Scientific Applications
(Séminaire du 3e cycle romand d'informatique
Location:
University of Geneva, CUI
Battelle Bat. A
7 route de Drize
1227 Carouge
http://cui.unige.ch/~chopard/direction.html
Date: Nov 16, 2007, 9:45am-5pm
Room: 406 (3nd floor, right, Battelle, Buidling A)
Lecturer:
Dr. Thomas Sangild
Centre for Medical Image Computing,
Department of Medical Physics and Bioengineering,
University College London, United Kingdom.
http://www.daimi.au.dk/~sangild/
Thomas Sangild Sřrensen obtained his MS in computer science from the
University of Aarhus, Denmark in 2000 and a PhD in medicine in
2004. From 2004-2007 he was assistant professor at the Centre for
Advanced Visualization and Interaction, University of Aarhus,
2006-2007 honorary lecturer at King's College London, and currently
holds a research position at the Centre for Medical Image Computing
at University College London. He has previsouly tought two graduate
courses on general purpuse computation on GPUs at the Department of
Computer Science, University of Aarhus.
Abstract:
Graphics card vendor Nvidia has recently
released its Compute Unified Device Architecture (CUDA), a new
computing architecture for solving complex computational problems on
commodity graphics hardware (GPUs). By designing appropriate parallel
algorithms in standard C, the performance gain is significant
compared to conventional CPU-based implementations. This course is
intended to give an introduction to Cuda through the lecturer's own
research and examples contained in the Cuda SDK. Upon completion of
the course the participants will be able to start their own research
using Cuda.
Tentative Program:
- 09:45-10:00 - Registration, wecome
- 10:00-11:00 - An introduction to parallel computing on GPUs
examplified by surgical simulation and MRI reconstruction
- 11:00-11.30 - A first look at Cuda
- Lunch
- 13:00-14:30 - Cuda in more detail
- Performance issues
- Examples from the Cuda SDK
- 14:30-15:00 -Coffee break
- 15:00-16:30 - (Simple) Lattice Boltzmann and cellular automata
models in Cuda
Lecture notes
Participation is free, registration at the conference.
Members from UNIGE, UNIL, UNINE, UNIFR and EPFL can have their travel
expenses reimbursed.
Organization:
Bastien Chopard,
Computer Science Department,
University of Geneva
Bastien.Chopard@cui.unige.ch