Dynamic Tessellation - Practical Rendering and Computation with Direct3D 11 - page 408

Dynamic Tessellation
Earlier in the topic, the new tessellation-specific stages of the pipeline were introduced
(Chapter 3) and the key parameters and implementation details were discussed in depth
(Chapter 4). However, neither of these sections provided real-world demonstrations of this
key new Direct3D 11 technology.
In the introduction to Chapter 4, two common approaches to tessellation were intro-
duced, namely subdivision and higher-order surfaces. This chapter offers demonstrations
of both, to allow comparison between the two methods, and to demonstrate how the vari-
ous parameters and new shader units work together to produce a final output.
Subdivision is a form of tessellation based on refinement, in which we progressively
add more detail (in the form of additional triangles) until the final mesh better represents an
ideal shape. The first section of this chapter, "Terrain Tessellation," follows this approach.
Thinking about higher order is closer to the understanding that tessellation is about
representing curved or smooth surfaces. The ideal mesh will usually be described in terms
of a parametric mathematical equation, typically quadratic or cubic; the second section of
this chapter demonstrates this.
Broadly speaking, using subdivision is more algorithmic, while using higher-order
surfaces is more mathematical. We will explore the differences between these two ap-
proaches throughout this chapter, and will also build on the earlier introduction to the tes-
sellation stages to provide a well-rounded understanding of the technology.
9.1 Terrain Tessellation
Terrain rendering is probably one of the most common forms of procedural or run-time
generated graphics in the real-time rendering space. Unlike the majority of other art assets
413