high performance computing on graphics processing units: hgpu.org

Driven by the gaming industry and the great emphasis placed on the visual sense, graphics processing units (GPUs) have improved their performances in recent years, even outperforming the computational capacity of single core CPUs. In fact multi-core architectures are nowadays common for both CPUs and GPUs in order to exploit parallelism in computing. In this paper we will have an overview about computational intensive tasks in multimedia signal processing including audio and video which can benefit from an implementation on hardware with parallel computing capability. The architecture and performances of modern CPUs and GPUs are described firstly. Then three fundamental algorithms which are used to execute advanced audio signal processing are presented and their implementation on GPU is discussed. The third section deals with some basic tasks involved in video compression. Finally we get some conclusions, it turns out that the massive parallel processing capability of modern CPU and GPU can be used to facilitate multimedia applications with computational intensive requirements.