Graphics Reference
In-Depth Information
10
Implementing Deferred
Rendering
In this chapter, we will cover the following topics:
F
Filling the G-Buffer
F
Implementing a screen-aligned quad renderer
F
Reading the G-Buffer
F
Adding multiple lights
F
Incorporating multisample anti-aliasing
Introduction
Deferred rendering refers to any rendering technique that defers the calculation of one or
more components of a rendered frame to a screen-space operation using information that
has been collected during one or more
attribute collection
passes.
One of the benefits of deferred rendering over forward rendering is that it can simplify
and improve the performance of implementations using many lights and support better
GPU utilization. It does, however, require extra texture memory for a G-Buffer, and can
introduce complexities around the implementation of multisample anti-aliasing (MSAA)
and transparencies.
There are a number of deferred rendering techniques with their own pros and cons. We will
present a classic deferred rendering technique in this chapter. The same basic concepts
can be applied across techniques although the number of rendering passes and/or the
information recorded at each stage may vary. Other techniques such as tiled deferred
rendering may also make use of compute shaders.
