Sampling-based Performance Prediction of Raytracing

S. Juhász and A. Csikvári (Hungary)

Keywords

raytracing, antialiasing, cost function, samplingbased performance prediction

Abstract

Raytracing, a rendering technique that excels at calculating shadows, reflections, and refractions, is often held up as a standard against which to judge all other 3D rendering. This method is well-known about its extremely high computation power requirements as well. Owe to the spectacular evolution of the microprocessor technology, raytracing plays an increasingly important role in creating high quality images and animations. Raytracing has the advantage of being more easily parallelizable than other methods of comparable quality using different lighting models. The right choice and the right sizing of the parallel execution environment call for an accurate prediction of the rendering time. The previously proposed estimation methods strongly rely on the data representation and the internal structure of the raytracer engines, thus their usability is limited by the continuous evolution of the applied algorithms, or simply by the lack of knowledge on the application generating the images. This paper suggests another prediction method making no presumption about the rendering engine, and not requiring the knowledge of the source file structures either. Unfortunately the price of the ultimate predicting ability is paid by the relative slowness of the approach.

Important Links:



Go Back