GPGPU 2013
29th April-3rd May 2013
Department
of Computer Science
University
of Cape Town
Although originally designed to handle computation for computer graphics, fully programmable Graphics Processing Units (GPUs) now offer a powerful alternative to general purpose computational applications traditionally handled by Central Processing Units (CPUs). As GPUs have overtaken CPUs in sheer processing power, GPU solutions have become particularly attractive for computationally intensive problems, such as data processing and physical simulations.
This course covered the
practical development of algorithms and associated coding in NVIDIA’s CUDA
programming environment for effective general purpose computing with GPUs.
The course was attended by UCT Computer Science Honours (4th year) students and a number of MSc students from UCT and other South African universities.
 |
|
Speakers:
International:
John Stone
Associate Director, CUDA
Center of Excellence at University of Illinois at Urbana-Champaign
NVIDIA CUDA Fellow, 2010 to Present
Manuel Ujaldón
Associate Professor, University of
Malaga
NVIDIA CUDA Fellow, 2012 to Present
Local:
Michelle Kuttel
Associate Professor, University of Cape Town
James Gain
Associate Professor, University of Cape Town
Simon Ratcliffe
"The Barefoot Astronomer", SKA South Africa
2013 Programme:
| Monday 29 April | Tuesday 30 April | Wednesday 1 May | Thursday 2 May | Friday 3 May | |
|---|---|---|---|---|---|
| 9-10:30 am | Overview and introduction | CUDA applications I. John Stone (UIUC) | A brief OpenACC intro plus other general approaches to GPU computing. Manuel Ujaldón | Future Nvidia developments: Echelon project, Dragonfly interconnect, Maxwell and Volta GPUs. Manuel Ujaldón | Supercomputers and GPUs: Presence in the top500, an overview of Titan supercomputer. Manuel Ujaldón |
| 10:30-11am | TEA | TEA | TEA | TEA | TEA |
| 11-12:30am | Programming in CUDA: the essentials : J. Stone | CUDA applications II. John Stone (UIUC) | The Kepler architecture and six ways to enhance CUDA programs using its new capabilities. Manuel Ujaldón | Programming for hybrid architectures. J. Stone (UIUC) | Many core and the SKA. Simon Ratcliff (SKA) |
| 12:30-2pm | LUNCH | LUNCH | LUNCH | LUNCH | LUNCH |
| 2-5pm | Prac 01 - Introduction to cluster computing - Hello World on the cluster - CUDA Runtime API - Vector | Prac 02 - Parallel Reduction | Prac 03 - Numeric Integration | Prac 04 - N-body Simulation |
