Analyzing Deferred Rendering Techniques

Abstract

In this article various deferred rendering algorithms are investigated and a classification that formalizes the comparison between these popular rendering techniques is introduced. This classification consists of measuring functions that can be used to determine the expected algorithm performance in various situations. Multiple analysis spaces are defined that better express the strengths and weaknesses of each algorithm.

Given the abundance of deferred rendering methods and the performance tradeoffs implied by different hardware targets, rendered objects complexity and light setup complexity, our framework makes choosing or modifying an algorithm out of this collection a simpler process.

The following spaces are used for algorithm examination and comparison: GPU commands, processing cost, allocated memory and expected bandwidth consumption. Furthermore, the analysis spaces are independent of the illumination model and are suitable for a decoupled examination, in which each stage of rendering process is usually executed at a different sampling rate.