17-19 October 2012
National Library of Serbia
Europe/Belgrade timezone

Evolution of an gravitotional bound n-body system with DLA (Diffusion Limited Aggregation) initial mass distribution using GPU parralel computing

18 Oct 2012, 15:45
1h 15m
National Library of Serbia

National Library of Serbia

Skerliceva 1 11000 Belgrade Serbia
Poster Computational Physics Poster session

Speakers

Mr Bogdan Alexandru Dumitru (Institute of Space Science)Mr Ciprian Mitu (Institute of Space Science)

Description

In this paper we present a collection of algorithms to generate power law mass distribution as initial conditions for gravitational N-body simulations. This type of algorithms are highly parallizable and can analyze large data customary on clusters. The input data will be high resolution images in the most used format in astrophysics, FITS and diffusion limited aggregation generated clusters. The application will algorithm running on GPU (with CUDA and OpenCL), for the generation of initial mass distribution and fort the evolution of the n-body sysytem. Data analysis and plotting is done in ROOT Framework. General scientific image on the formation and evolution of the galaxies is that they formed by gravitational collapse of matter. The study of the mass distribution of spiral galaxies will shade new light on the formation and evolution of galaxies. The application will provide new data of spiral arms and their morphology. The method is name all-pairs that is a brute-force technique that evaluates all pair-wire interactions among the n bodies. The all-pairs method is combined with a faster method based on a far-field approximation of longer-range forces, which is valid only between parts of the system that are well separated. Fast N-body algorithms of this form include the Barnes-Hut method (BH) (Barnes and Hut 1986), the fast multipole method (FMM) (Greengard 1987), and the particle-mesh methods (Hockney and Eastwood 1981, Darden et al. 1993). The benchmarks was done on an server with 4 GPU (Nvidia Tesla) with 480 cores each. The tests was been maid on 480, 960, 1440, 1920 cores on CUDA and OpenCL and the difference was 5-10% in CUDA favor.

Primary authors

Mr Bogdan Alexandru Dumitru (Institute of Space Science) Mr Ciprian Mitu (Institute of Space Science) Dr Ion Sorin Zgura (Institute of Space Science) Mr Mihai Niculescu (Institute of Space Science)

Presentation Materials

There are no materials yet.