The GLSL Shader Language - Graphics Shaders - page 98

• There are no char , char * , or string data types, and GLSL has no string-
manipulation functions.
• There is no sizeof( ) operator, because there is no need to deal with data
in various sizes. There are standard constructors for arrays and matrices
of all needed sizes.
• No implicit type conversions are allowed in GLSL. Conversions are sup-
ported by explicit type constructors.
Instead of implicit conversions, or type casts, there are three explicit con-
structors for simple types, as follows:
• int(arg): converts the argument to an int; the argument may be a float
or a bool.
• float(arg): converts the argument to a float; the argument may be an
int or a bool.
• bool(arg): converts the argument to a boolean; the argument may be a
float or an int.
The usual conversion operations are used: conversions from float to int
simply drop the fractional part, nonzero floats or ints convert to the Boolean
true , etc. This is a different syntax from the familiar cast operations, but it gives
you the same functionality if you need it.
New Matrix and Vector Types
GLSL supports a number of predefined data types for vectors and matri-
ces. Vectors may have a real, integer, or Boolean base type, but matrices
must be real. Many familiar vector and matrix operations and functions can
be applied to variables of these types, and a number of useful new func-
tions are also provided. These are discussed in several sections later in this
chapter.
GLSL's built-in floating-point scalar and vector types are loat , vec2 ,
vec3 , and vec4 . The storage for a variable of type vecN is simply that of a tra-
ditional array, but you want to use the built-in type rather than the traditional
array type. Using the vecN types explicitly makes a much larger number of
operations available for the data, and these operations can then take advan-
tage of graphics card parallelism to work at a much higher speed.
GLSL's built-in integer, scalar, and vector types are int , ivec2 , ivec3 , and
ivec4 . Again, the storage for an ivec variable is the same as that for a tradi-
tional array, but the explicit ivec type can take a much larger set of operations.