Speaker
Scott
Rostrup
University
of
Waterloo
Abstract
Though originally designed with video games and graphics, accelerator architectures such as the Cell Processor (PS3) and GPUs (Graphics Processing Unit) are beginning to be adopted by the scientific, financial, and engineering communities. They are highly parallel architectures suitable for a wide array of data-parallel computing tasks and though they represent two very different hardware implementations many parallel programming principles apply equally well to both architectures. Then I'll use the sparse matrix-vector product as a template problem to introduce these ideas in the context of using an Nvidia GPU. The aim of this talk is to introduce accelerator architectures and demonstrate that GPUs are accessible and easy to get started with provided you are already familiar with coding in C.