The Georgia Institute of Technology and the experimental GPU-focused Keeneland project will host a two-day tutorial conference later this month to examine developer issues relating to GPU clusters and high-performance computing. Held from April 14 to 15, the Georgia Tech NVIDIA CUDA Center of Excellence will present speakers with expertise in GPU heterogeneous processing for computational science.
The sessions at this event are targeted at developer/users of Georgia Tech's new GPU cluster called Keeneland, as well as any interested GPU computing developer. Research being conducted on the National Science Foundation-funded Keeneland cluster is described as an A-Z of sciences, including: astrophysics, behavioral and neural science, cryptography, fluid dynamics, molecular dynamics, nuclear physics, protein folding, turbulence simulation, and more.
The full agenda is as follows:
- Keeneland architectural overview and motivation
- Introduction to CUDA
- Introduction to OpenCL
- Application case studies
- Overviews of GPU enabled libraries, correctness and performance tools and compilers
- High-level language interfaces, such as MATLAB, R and Accelereyes, to GPUs
- Advanced optimization techniques for GPUs
- Hands-on time with the Keeneland Initial Delivery System