Contents

General purpose computing on graphics processing units, known as GPGPU but now often referred to as GPU computing, is the approach of performing computation on the GPU instead of the CPU. GPU computing has been made possible by the increasing pro-grammability and performance of GPUs. The programmability of GPUs is accessed via shader programs, typically written in a C like language. Until recently shader programs and shader languages have been targeted (naturally) towards graphics applications. How-ever there has been increasing interest in using GPUs for non graphics computationally intensive tasks. This has lead to new languages and tools for GPU development which are more general than shaders. For example, Nvidia released its compute unified device archi-tecture (CUDA) in 2007. CUDA allows what is essentially general purpose C code to be executed on GPU processors, without the graphics orientation found in shader languages. More recently Open Compute Language (OpenCL) has been proposed for GPU computing, and more generally for CPU-GPU hybrid computing. This research aims to compare performance of particle system computation betwee