Attribute


Description:

public Attribute (AttributeBuffer attribute_buffer, string name, ulong stride, ulong offset, int components, AttributeType type)

Describes the layout for a list of vertex attribute values (For example, a list of texture coordinates or colors).

The name is used to access the attribute inside a GLSL vertex shader and there are some special names you should use if they are applicable: <itemizedlist> <listitem>"cogl_position_in" (used for vertex positions)</listitem> <listitem> "cogl_color_in" (used for vertex colors)</listitem> <listitem>"cogl_tex_coord0_in", "cogl_tex_coord1", ... (used for vertex texture coordinates)</listitem> <listitem>"cogl_normal_in" (used for vertex normals)</listitem> <listitem> "cogl_point_size_in" (used to set the size of points per-vertex. Note this can only be used if cogl_feature_id_point_size_attribute is advertised and set_per_vertex_point_size is called on the pipeline. </listitem> </itemizedlist>

The attribute values corresponding to different vertices can either be tightly packed or interleaved with other attribute values. For example it's common to define a structure for a single vertex like:

typedef struct
{
float x, y, z; /<!-- -->* position attribute *<!-- -->/
float s, t; /<!-- -->* texture coordinate attribute *<!-- -->/
} MyVertex;

And then create an array of vertex data something like:

MyVertex vertices[100] = { .... }

In this case, to describe either the position or texture coordinate attribute you have to move <literal>sizeof (MyVertex)< /literal> bytes to move from one vertex to the next. This is called the attribute stride. If you weren't interleving attributes and you instead had a packed array of float x, y pairs then the attribute stride would be <literal>(2 * sizeof (float)) </literal>. So the stride is the number of bytes to move to find the attribute value of the next vertex.

Normally a list of attributes starts at the beginning of an array. So for the <literal>MyVertex</literal> example above the offset is the offset inside the <literal>MyVertex</literal> structure to the first component of the attribute. For the texture coordinate attribute the offset would be <literal>offsetof (MyVertex, s)</literal> or instead of using the offsetof macro you could use <literal>sizeof (float) * 3</literal>. If you've divided your array into blocks of non-interleved attributes then you will need to calculate the offset as the number of bytes in blocks preceding the attribute you're describing.

An attribute often has more than one component. For example a color is often comprised of 4 red, green, blue and alpha components , and a position may be comprised of 2 x and y components. You should aim to keep the number of components to a minimum as more components means more data needs to be mapped into the GPU which can be a bottlneck when dealing with a large number of vertices.

Finally you need to specify the component data type. Here you should aim to use the smallest type that meets your precision requirements. Again the larger the type then more data needs to be mapped into the GPU which can be a bottlneck when dealing with a large number of vertices.

Parameters:

attribute_buffer

The AttributeBuffer containing the actual attribute data

name

The name of the attribute (used to reference it from GLSL)

stride

The number of bytes to jump to get to the next attribute value for the next vertex. (Usually <literal>sizeof (MyVertex)< /literal>)

offset

The byte offset from the start of attribute_buffer for the first attribute value. (Usually <literal>offsetof ( MyVertex, component0)</literal>

components

The number of components (e.g. 4 for an rgba color or 3 for and (x,y,z) position)

type

FIXME

Returns:

A newly allocated Attribute describing the layout for a list of attribute values stored in array.