The increasing need of resources in video games has been driving the market to design and produce high performance video boards. Cuda is a technology allowing to program these video boards (GPU) almost like conventional processors (CPU). PyCuda is an extension which allows to use this programming model in Python directly.
This marriage between Python dynamism and GPU computation power gives life to interesting and fascinating applications.
The talk is divided into:
- Introduction to GPU programming
- What cuda is
- How PyCuda is structured on this technology (boost-python, numpy)
- Gpu-Metaprogramming
- Practical examples
- Scientific examples
- Questions and answers
This talk is mainly addresses to people using Python for scientific computation or high graphic applications and assumes some knowledge of numpy. Nonetheless, all the examples are very visual and can be valued from all the audience.