mid's site

Until now we pushed vertices one-by-one using the glVertex* and its associated functions. This means 1 function call for every single vertex! That's clearly not ideal. On a computer with OpenGL 1.1 and up, there exists a way to push more vertices at once, and this method should also be more efficient.

To use this feature, we use the glEnableClientState and glDisableClientState pair with each vertex attribute we wish to compress. Enabling client states tells OpenGL to read vertex attributes from an array, which we shall pass via any of glVertexPointer, glColorPointer, etc.

The vertex drawing operation can then be done with a quick glDrawArrays. Applying this to the half-cube example,

// GL_VERTEX_ARRAY and glVertexPointer have unfortunate naming.
// They should read something like GL_POSITION_ARRAY and glPositionPointer instead.

// Setup vertex information
static float hsz = 0.2;
float array[] = {
	// -Z square
	-hsz, -hsz, -hsz,
	+hsz, -hsz, -hsz,
	-hsz, +hsz, -hsz,
	
	-hsz, +hsz, -hsz,
	+hsz, -hsz, -hsz,
	+hsz, +hsz, -hsz,
	
	// -X square
	-hsz, -hsz, -hsz,
	-hsz, -hsz, +hsz,
	-hsz, +hsz, -hsz,
	
	-hsz, +hsz, -hsz,
	-hsz, -hsz, +hsz,
	-hsz, +hsz, +hsz,
	
	// -Y square
	-hsz, -hsz, -hsz,
	+hsz, -hsz, -hsz,
	-hsz, -hsz, +hsz,
	
	-hsz, -hsz, +hsz,
	+hsz, -hsz, -hsz,
	+hsz, -hsz, +hsz,
};

glColor3f(1, 1, 1);

glEnableClientState(GL_VERTEX_ARRAY);

// Our position array
// has `3` components,
// is composed of `float`s,
// spaced `sizeof(float) * 3` bytes apart,
// starting at `array`
glVertexPointer(3, GL_FLOAT, sizeof(float) * 3, array);

// Draw triangles, using 18 vertices starting index 0
glDrawArrays(GL_TRIANGLES, 0, 18);

glDisableClientState(GL_VERTEX_ARRAY);

To reiterate, disabling and/or resetting OpenGL state at the end of these snippets is done for cleanliness. If you are aware of how state is modified by your application, you can and should omit state changes when possible.

Each vertex attribute has its own array form. Here is the same example using per-vertex colors in addition:

// Setup vertex information
static float hsz = 0.2;
float positions[] = {
	// -Z square
	-hsz, -hsz, -hsz,
	+hsz, -hsz, -hsz,
	-hsz, +hsz, -hsz,
	
	-hsz, +hsz, -hsz,
	+hsz, -hsz, -hsz,
	+hsz, +hsz, -hsz,
	
	// -X square
	-hsz, -hsz, -hsz,
	-hsz, -hsz, +hsz,
	-hsz, +hsz, -hsz,
	
	-hsz, +hsz, -hsz,
	-hsz, -hsz, +hsz,
	-hsz, +hsz, +hsz,
	
	// -Y square
	-hsz, -hsz, -hsz,
	+hsz, -hsz, -hsz,
	-hsz, -hsz, +hsz,
	
	-hsz, -hsz, +hsz,
	+hsz, -hsz, -hsz,
	+hsz, -hsz, +hsz,
};

uint8_t colors[] = {
	// -Z square
	255, 0, 0,
	255, 0, 0,
	255, 0, 0,
	
	255, 0, 0,
	255, 0, 0,
	255, 0, 0,
	
	// -X square
	0, 255, 0,
	0, 255, 0,
	0, 255, 0,
	
	0, 255, 0,
	0, 255, 0,
	0, 255, 0,
	
	// -Y square
	0, 0, 255,
	0, 0, 255,
	0, 0, 255,
	
	0, 0, 255,
	0, 0, 255,
	0, 0, 255,
};

glEnableClientState(GL_VERTEX_ARRAY);
glEnableClientState(GL_COLOR_ARRAY);

// Our position array
// has 3 components,
// is composed of floats
// spaced `sizeof(float) * 3` bytes apart,
// starting at `positions`
glVertexPointer(3, GL_FLOAT, sizeof(float) * 3, positions);

// Our color array
// has 3 components,
// is composed of unsigned bytes
// spaced 3 bytes apart,
// starting at `colors`
glColorPointer(3, GL_UNSIGNED_BYTE, 3, colors);

// Draw triangles, using 18 vertices starting index 0
glDrawArrays(GL_TRIANGLES, 0, 18);

glDisableClientState(GL_COLOR_ARRAY);
glDisableClientState(GL_VERTEX_ARRAY);

It is possible and is typical to interleave these vertex attributes for simplicity in most cases, and C makes this easy:

struct Vertex {
	float px, py, pz;
	uint8_t cx, cy, cz;
};

// Setup vertex information
static float hsz = 0.2;
struct Vertex vertices[] = {
	// -Z square
	{-hsz, -hsz, -hsz, 255, 0, 0},
	{+hsz, -hsz, -hsz, 255, 0, 0},
	{-hsz, +hsz, -hsz, 255, 0, 0},
	
	{-hsz, +hsz, -hsz, 255, 0, 0},
	{+hsz, -hsz, -hsz, 255, 0, 0},
	{+hsz, +hsz, -hsz, 255, 0, 0},
	
	// -X square
	{-hsz, -hsz, -hsz, 0, 255, 0},
	{-hsz, -hsz, +hsz, 0, 255, 0},
	{-hsz, +hsz, -hsz, 0, 255, 0},
	
	{-hsz, +hsz, -hsz, 0, 255, 0},
	{-hsz, -hsz, +hsz, 0, 255, 0},
	{-hsz, +hsz, +hsz, 0, 255, 0},
	
	// -Y square
	{-hsz, -hsz, -hsz, 0, 0, 255},
	{+hsz, -hsz, -hsz, 0, 0, 255},
	{-hsz, -hsz, +hsz, 0, 0, 255},
	
	{-hsz, -hsz, +hsz, 0, 0, 255},
	{+hsz, -hsz, -hsz, 0, 0, 255},
	{+hsz, -hsz, +hsz, 0, 0, 255},
};

glEnableClientState(GL_VERTEX_ARRAY);
glEnableClientState(GL_COLOR_ARRAY);

// Our position array
// has 3 components,
// is composed of floats
// spaced `sizeof(struct Vertex)` bytes apart,
// starting at `vertices->px`
glVertexPointer(3, GL_FLOAT, sizeof(struct Vertex), &vertices->px);

// Our color array
// has 3 components,
// is composed of unsigned bytes
// spaced `sizeof(struct Vertex)` bytes apart,
// starting at `vertices->cx`
glColorPointer(3, GL_UNSIGNED_BYTE, sizeof(struct Vertex), &vertices->cx);

// Draw triangles, using 18 vertices starting index 0
glDrawArrays(GL_TRIANGLES, 0, 18);

glDisableClientState(GL_COLOR_ARRAY);
glDisableClientState(GL_VERTEX_ARRAY);

Be careful not to get caught by padding bugs.

There's no reason you cannot interleave only some attributes and keep others separate, or maybe in completely different arrays. This could be more optimal for cases where certain attributes must be processed on the CPU each frame (such as positions in CPU-side model animations).