The advent of GPU ray tracers

Abstract

Ever since it's introduction in 1980 the ray tracing algorithm has been a tricky puzzle. On the one hand it offers photo-realistic rendering superior to the more common rasterization algorithm, but on the other it's much too slow to serve as a replacement. Much of the rasterization algorithm's performance advantage lies in the use of special purpose hardware. This hardware, known as graphics cards, has evolved rapidly over the past decades, driven by demands for ever higher polygon counts. More recently, this pressure has lead to the introduction of programmable shader units to the graphics card architecture. These units became ever more programmable as researchers realized their potential as general parallel processing units. Eventually this trend lead to the exploration of the graphics card architecture as a ray tracing platform. This paper aims to explore the advent of this type of ray tracer and explain the developments that have been made in this field in order to see what the current state of the art is and where further research is required. It will also explore the defects that can be identified in the current literature and discuss how to address them.