
Splinter Meeting of the AG2011, September
20-23, 2011, Heidelberg
"From Star Clusters to Galaxy Formation - The Virtual Universe"
Our Splinter Meeting includes a scientific part, and a special day on programming and exhibition of GPU architectures.
The presentation of science with GPUs is supported financially and logistically by NVIDIA and IBM.
SOC: Reinhold Bien, Jose Fiestas, Stefan Harfst, Andreas Just, Volker Springel, Rainer Spurzem (Contact:
reinhold@ari.uni-heidelberg.de)
Scientific Rationale:
Galactic astrophysical research is facing the challenge of
a wealth of new observational data and the emergence of ever more
powerful modelling techniques. For example, the launch of ESA's space
mission Gaia is scheduled for the year 2012, which is expected to
yield an unprecedented amount of accurate astrometric and physical
data for a huge number of stars. In addition, we mention RAVE, SDSS,
PanSTARRS, and the Herschel Space Observatory. Fully harvesting this
information will also require novel theoretical approaches. It
therefore is timely to bring together scientists interested in
simulations that cover the broad range from star clusters, disc
dynamics, central black holes, or sinking satellites, to galaxy
formation in a cosmological context. The numerical methods discussed
in the splinter meeting may include N-body integration, SPH and
hydrodynamical codes, particle-mesh techniques, and other
procedures, as well as hardware and software co-evolution.
Special emphasis is placed in this workshop to present cutting
edge results obtained on novel computing architectures,
used for computational astrophysics and dynamics.
In recent years general purpose computing on many-core
accelerator boards (in practise in most cases GPU cards,
graphical processing units) has changed the supercomputing
landscape. A single GPU connected with a standard personal
computer yields Teraflop scale parallel computing capacity,
which few years ago was the exclusive domain of big and
expensive supercomputers. The fastest computer in the world
according to the Top500 list is a cluster using thousands
of GPU for accelerating computations. New parallel programming
methods have to be employed. Computational astrophysics is
one of the pioneers in the field; in our workshop we want to
bring together theoretical astronomers, special supercomputing
experts, and those who are already working in the crossroads
between the fields.