Graphics Reference
In-Depth Information
2
IV
Deferred Rendering Techniques
on Mobile Devices
Ashley Vaughan Smith
2.1 Introduction
With ongoing advances in GPU technology and the introduction of OpenGL
ES 3.0, it has now become possible to utilize rendering techniques that were
previously only usable on desktop GPUs. Applications that use dynamic lighting
require a technique to shade the scene in an ecient way in real time. On mobile
devices with a low power requirement, this is dicult to achieve.
This chapter discusses the available techniques for deferred rendering [Harg-
reaves and Harris 04] on mobile devices and implementation details that allow
optimized rendering of these techniques. The purpose of this chapter is to give the
reader an overview of the available deferred rendering techniques using OpenGL
ES 2.0 and 3.0, but not an in-depth tutorial on implementing each technique.
Readers unfamiliar with deferred rendering may wish to review previous litera-
ture about deferred rendering, e.g., [Calver 03,Thibieroz 04]. Also included are
new opportunities to use OpenGL ES extensions to optimize these techniques
even further.
2.2 Review
In review, there are two mainstream approaches to going about lighting a scene:
forward rendering and deferred rendering. Forward rendering was common in
fixed function pipelines, such as OpenGL ES 1.0, where there was no option to
use deferred rendering. It is also possible to use forward rendering with pro-
grammable pipelines like OpenGL ES 2.0. An application that uses forward
rendering requires information for a set number of lights that will be used to af-
fect the geometry in the first pass. This usually requires some way of discarding
lights when there are too many affecting a single piece of geometry or rendering
