high performance computing on graphics processing units: hgpu.org

As hardware has become increasingly powerful, the doors have been opened for a wide range of more computationally intensive simulation procedures. In particular, agent-based modelling has seen a recent surge of interest in the fields of Biology and Economics. For this project we propose using an agent-based model to create an implementation of the classic physics problem of cosmological N-body simulation, in order to investigate the process of galaxy formation. Our simulation should be displayed as an evolving 3D graphic, to allow for easy interpretation of the data. The GPU, with its massively-parallel architecture is well suited to the computation of N-body systems, and allows us to create high-fidelity simulations using mid-level consumer hardware. We therefore also investigate methods for performing our model’s calculations on the GPU. The primary aims of the project are to determine whether agent-based models are suitable for the investigation of physical phenomena, and to produce a realistic model of galaxy formation in doing so.