Institute for Particle and Nuclear Physics, Wigner RCP, HAS

Serious toys - GPU Day 2010


These days research fields with the demand of huge numerical calculation
capacity focus on programming of Graphical Processing Units (GPU) - well
known from toy industry and games on personal computers.
The enhanced interest on the GPUs is based on their computational capacity
increase, which is faster than the well known Moore's Law - valid for the
ordinary CPU's capacity. In parallel, general purpose graphical languages
reached such a high level where it is easy to learn them and implement them
into the usual high-level object-oriented programming languages. Last, but
not least a "personal supercomputer" with a 7 Teraflop computational
capacity (including e.g. 4 graphical cards) can be built from a $2000
investment - this means 30 dollarcent for 1 Gigaflop (1 billion numerical
calculation per second).

On 4 June 2010 Gergely Gábor Barnaföldi and Gergő Debreczeni organized a
one-day-workshop at the KFKI Particle and Nuclear Physics Institute of the
HAS (MTA KFKI RMKI) in Budapest.
Research scientists, university professors, and lots of students
participated and exchanged their views on the topic of programming
graphical processors. Several institutes of the HAS have been represented
by their researchers including RMKI, SzFKI, MFA, and the Eötvös University.
Furthermore 40 students from universities were presented also.

The main aim of the program was to open a forum of GPU programmers for
discussions
on various research fields applying GPUs.
A parallel aim was to increase the publicity of GPU programming for
university students. The morning of the "GPU Day 2010" was an introduction section with tutorials for "GPU newcomers", including the test possibility of the
local high-performance GPU machines from the RMKI and the MFA.  In the
afternoon sections the technology leaders of the fields (mostly Hungarian
research groups working on physics) displayed their results obtained on GPU
resources.

All presented materials and practical informations were collected and
archived at http://gpu.kfki.hu web page.


Péter Lévai, MTA KFKI RMKI