N-body simulation using CUDA

Abstract

The N-body problem is an astrophysical phenomenon which is studied in the attempt to understand the movement of stars and galaxies in space. Using today's processors, one can simulate such a problem through an N-body simulation, which can use Newtonian gravity applied between a number of particles. It is simple to perform such a simulation with a few particles, but it becomes increasingly difficult and expensive as the number of particles increases. This project investigates the improvement of N-body simulation performance by using parallel algorithms on CUDA, as opposed to using the standard sequential methods, as well as the advantage of using optimized algorithms as opposed to the standard brute. force algorithms. It provides results which show the discrepancy between the algorithms and the different implementations, as well as a visualisation tool which allows for better studying of the N-body simulation.