Mikael Kalms is a Swedish game developer. He joined the industry in 1998.
Since then, he has worked at EA DICE, Nixxes Software and Twilight 3D Oy.
For the last three months he was a visitor at the MPI working on his master thesis project.
CUDA is an accessible solution to achieve parallelism in C++ programs on desktop computers. But how well does it work in practice? And does it really speed up your algorithms hundred-fold? And how long does it take to implement an algorithm in CUDA?
The talk will discuss when to use CUDA (and when not to), how to design your algorithms to take advantage of the CUDA architecture, and how to keep down the implementation time. It will conclude with a case study of a real-time particle simulation.