Modern Graphics Hardware - Computer Graphics: Principles and Practice - page 1119

parameters that specify the image position to be evaluated. (Regardless of the
image size, parameter values between 0 and 1 specify positions from one edge of
the image to the other. Parameter values outside [ 0, 1 ] may wrap, reflect, or clamp,
as specified ahead of time.) Thus, a shader can specify coordinates that have been
interpolated across the primitive during rasterization (i.e., they can implement tra-
ditional texture correspondence) or they can choose to specify coordinates that
have been computed arbitrarily, perhaps as functions of the results of other texture
image interpolations. Such dependent texture image interpolation is a very pow-
erful feature of modern GPUs, but, as we will see in Section 38.7.2, it complicates
their implementation.
Listing 38.2 replaces the constant-based color attenuation of the example frag-
ment shader in Listing 38.1 with attenuation defined by a 1D texture image:
brightness_table . Like the global variable brightness , brightness_table is
specified by the application prior to rendering and remains constant throughout
rendering. Unlike brightness , however, which was stored in a GPU register (e.g.,
within a core in Figure 38.4), brightness_table is bulk memory state (see Sec-
tion 38.3), stored in the off-chip GPU memory system (e.g., GDDR3 memory in
Figure 38.4). Sampler s is a simple data structure that specifies a texture image
( brightness_table ) and the technique to be used to evaluate it (linear interpola-
tion between texels).

Listing 38.2: Another simple HLSL fragment shader.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

texture1D brightness_table;

sampler1D s = sampler_State {
texture = brightness_table;
filter = LINEAR;

};

struct v2f {
float4 Position : POSITION;
float4 Color : COLOR0;
float TexCoord : TEXCOORD0;

};

struct f2p {
float Depth : DEPTH0;
float4 Color : COLOR0;

}

// attenuate fragment brightness with a 1-D texture
void main(in v2f Input, out f2p Output) {
Output.Depth = Input.Position.z;
Output.Color = tex1D(s, Input.TexCoord) * Input.Color;

}

The input structure v2f is extended with a third component, TexCoord , with
semantic TEXCOORD0 , which specifies the texture location to be evaluated. Within
main , the 1D texture interpolation function tex1D is called with sampler s (spec-
ifying that brightness_table is to be interpolated linearly) and coordinate
Input.TexCoord (indicating where brightness_table is to be evaluated), and
it returns the floating-point result. The four-component vector Input.Color is
scaled by this value, and the result, an attenuated, four-component color vector, is
assigned to Output.Color .