The Role of Power Dissipation and Locality of Reference in the Specification of High Performance Graphics Algorithms

Abstract

The amount of power dissipated by the implementation of an algorithm, for instance in the form of a dedicatedchip-set, is considered to be one of the most important constraints for the selection of a high performance graphicsalgorithm. This is due to the fact that the realization of computational capability within the reach of one Teraoperations per second is non-practical with general purpose CPU-chips. The case study of a high performancesurface visualization engine is used to introduce the reader with the aspect of power dissipation in relation tocomputational power. We introduce a low-power' parallel datapath' RISe processor, based on a highly efficientmapping of locality of reference in the algorithm onto silicon. A subsequent classification is made for varioushigh performance graphics algorithms.