Riccardo Gatti

riccardo.gatti@unipv.it

The Future of Parallel Computing: GPU vs CELL
General Purpose Planning against fast graphical computation architectures, which is the best solution for General Purposes Computation?

 

Luca Bianchi, Riccardo Gatti, Luca Lombardi

Keywords: Real-time Visual Simulations, Parallel Computing, HPC, CELL, GPU, GPGPU, General Purposes, Fast Computing

Abstract: Complex models require high performance computing (HPC) which means Parallel Computing. That is a fact. The question we try to address in this paper is "which is the best suitable solution for HPC contexts such as rendering? Will it be possible to use it in General Purpose elaborations?" We start from these questions and analyze two different approaches, IBM CELL and the well known GPGPU, showing how changing our minds and breaking some assumptions can lead to unexpected results and open a whole set of new possibilities. We talk about rendering, but quickly move slightly towards general purpose computation, because many algorithms used in Visual Simulations are not only referred to rendering issues but to a wider range of problems.