Karlsruhe 2011 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 76: Grid-Computing I

T 76.7: Talk

Wednesday, March 30, 2011, 18:20–18:35, 30.23: 2-1

Scientific Computing on Graphics Processing Units for the Example Case of IceCube — •Marius Wallraff, David Boersma, Benjamin Hoffmann, and Christopher Wiebusch — III. Physikalisches Institut, RWTH Aachen, D-52056 Aachen

General-Purpose computing on Graphics Processing Units (GPGPU) is a relatively new technique to solve certain computationally intensive problems in a cost-, time-, and energy-efficient manner. Internally, GPUs consist of hundreds of simple processors which run in parallel. Graphics cards are inexpensive and thus provide highly available computing power. Depending on the parallelizability of the specific algorithms, a single graphics card can substitute up to about 100 normal CPUs.

This talk will give a short overview of the different architectures and programming languages. Opportunities and limitations will be explained, and two example cases for applications in simulation and reconstruction for the IceCube Neutrino Observatory will be shown.