OpenCL for researchers
- Parallel Programming in Science with OpenCL

Location and Date: Copenhagen, Denmark, March 4th-8th, 2013

This five day course gives a fundamental introducton to OpenCL programming and moves on to cover advanced topics such as computational mapping/rethinking of classical algorithms in physical simulaton.
With OpenCL one can have heterogeneous computing within ones grasp and can easily move from one hardware platform to the other. Further, OpenCL is open standard and there are no proprietary issues from applying OpenCL. Most implementatons of OpenCL use the open source LLVM JIT compiler to compile the OpenCL kernels.
Unlike CUDA, OpenCL is backed by many manufacturers, such as AMD, NVIDIA, Intel, IBM, Qualcomm and many more. The course will cover practical aspects; from getting ones compiler working and writing ones frst hello world program, to studying advanced programming language constructs in OpenCL. The programming model of OpenCL is carefully introduced and many problems and algorithms are demonstrated with real life examples of source code from the Bullet Physics engine. After this course, participants should have reached a knowledge level and skills where they can start using OpenCL in their own research.
The course is for researchers in any discipline of science. We do assume some programming experience with C/C++ or similar languages. However, the course has a broad scope and participants need not be Computer Scientist or experts in programming.

Lecturers

• Erwin Coumans, Bullet Physics.
  Homepage: http://bulletphysics.org/
• Stephen Frye, Electronic Arts, UK.
  Homepage: http://www.ea.com/

Course Credit

5 ECTS
To obtain course credit, students must participate in all programming exercises.

Registration

Registration is closed. 

The course fees are:

• Free of charge for all in academia
• 300€ for others

Program and course material

Tentative

Sourse code