Vertex Arrays

You may have noticed that OpenGL requires many function calls to render geometric primitives. Drawing a 20-sided polygon requires 22 function calls: one call to glBegin(), one call for each of the vertices, and a final call to glEnd(). In the two previous code examples, additional information (polygon boundary edge flags or surface normals) added function calls for each vertex. This can quickly double or triple the number of function calls required for one geometric object. For some systems, function calls have a great deal of overhead and can hinder performance.

An additional problem is the redundant processing of vertices that are shared between adjacent polygons. For example, the cube in Figure 2-14 has six faces and eight shared vertices. Unfortunately, using the standard method of describing this object, each vertex would have to be specified three times: once for every face that uses it. So 24 vertices would be processed, even though eight would be enough.

Figure 2-14 : Six Sides; Eight Shared Vertices

OpenGL has vertex array routines that allow you to specify a lot of vertex-related data with just a few arrays and to access that data with equally few function calls. Using vertex array routines, all 20 vertices in a 20-sided polygon could be put into one array and called with one function. If each vertex also had a surface normal, all 20 surface normals could be put into another array and also called with one function.

Arranging data in vertex arrays may increase the performance of your application. Using vertex arrays reduces the number of function calls, which improves performance. Also, using vertex arrays may allow non-redundant processing of shared vertices. (Vertex sharing is not supported on all implementations of OpenGL.)

Note: Vertex arrays are standard in version 1.1 of OpenGL but were not part of the OpenGL 1.0 specification. With OpenGL 1.0, some vendors have implemented vertex arrays as an extension.

There are three steps to using vertex arrays to render geometry.

1. Activate (enable) up to six arrays, each to store a different type of data: vertex coordinates, RGBA colors, color indices, surface normals, texture coordinates, or polygon edge flags.
2. Put data into the array or arrays. The arrays are accessed by the addresses of (that is, pointers to) their memory locations. In the client-server model, this data is stored in the client's address space.
3. Draw geometry with the data. OpenGL obtains the data from all activated arrays by dereferencing the pointers. In the client-server model, the data is transferred to the server's address space. There are three ways to do this:

1. Accessing individual array elements (randomly hopping around)
2. Creating a list of individual array elements (methodically hopping around)
3. Processing sequential array elements

The dereferencing method you choose may depend upon the type of problem you encounter.

Interleaved vertex array data is another common method of organization. Instead of having up to six different arrays, each maintaining a different type of data (color, surface normal, coordinate, and so on), you might have the different types of data mixed into a single array. (See "Interleaved Arrays" for two methods of solving this.)

Step 1: Enabling Arrays

The first step is to call glEnableClientState() with an enumerated parameter, which activates the chosen array. In theory, you may need to call this up to six times to activate the six available arrays. In practice, you'll probably activate only between one to four arrays. For example, it is unlikely that you would activate both GL_COLOR_ARRAY and GL_INDEX_ARRAY, since your program's display mode supports either RGBA mode or color-index mode, but probably not both simultaneously.

void glEnableClientState(GLenum array)

Specifies the array to enable. Symbolic constants GL_VERTEX_ARRAY, GL_COLOR_ARRAY, GL_INDEX_ARRAY, GL_NORMAL_ARRAY, GL_TEXTURE_COORD_ARRAY, and GL_EDGE_FLAG_ARRAY are acceptable parameters.

If you use lighting, you may want to define a surface normal for every vertex. (See "Normal Vectors.") To use vertex arrays for that case, you activate both the surface normal and vertex coordinate arrays:

glEnableClientState(GL_NORMAL_ARRAY);

glEnableClientState(GL_VERTEX_ARRAY);

Suppose that you want to turn off lighting at some point and just draw the geometry using a single color. You want to call glDisable() to turn off lighting states (see Chapter 5). Now that lighting has been deactivated, you also want to stop changing the values of the surface normal state, which is wasted effort. To do that, you call

glDisableClientState(GL_NORMAL_ARRAY);

void glDisableClientState(GLenum array);

Specifies the array to disable. Accepts the same symbolic constants as glEnableClientState().

You might be asking yourself why the architects of OpenGL created these new (and long!) command names, gl*ClientState(). Why can't you just call glEnable() and glDisable()? One reason is that glEnable() and glDisable() can be stored in a display list, but the specification of vertex arrays cannot, because the data remains on the client's side.

Step 2: Specifying Data for the Arrays

There is a straightforward way by which a single command specifies a single array in the client space. There are six different routines to specify arrays - one routine for each kind of array. There is also a command that can specify several client-space arrays at once, all originating from a single interleaved array.

void glVertexPointer(GLint size, GLenum type, GLsizei stride,
const GLvoid *pointer);

Specifies where spatial coordinate data can be accessed. pointer is the memory address of the first coordinate of the first vertex in the array. type specifies the data type (GL_SHORT, GL_INT, GL_FLOAT, or GL_DOUBLE) of each coordinate in the array. size is the number of coordinates per vertex, which must be 2, 3, or 4. stride is the byte offset between consecutive vertexes. If stride is 0, the vertices are understood to be tightly packed in the array.

To access the other five arrays, there are five similar routines:

void glColorPointer(GLint size, GLenum type, GLsizei stride,
const GLvoid *pointer);
void glIndexPointer(GLenum type, GLsizei stride, const GLvoid *pointer);
void glNormalPointer(GLenum type, GLsizei stride,
const GLvoid *pointer);
void glTexCoordPointer(GLint size, GLenum type, GLsizei stride,
const GLvoid *pointer);
void glEdgeFlagPointer(GLsizei stride, const GLvoid *pointer);

The main differences among the routines are whether size and type are unique or must be specified. For example, a surface normal always has three components, so it is redundant to specify its size. An edge flag is always a single Boolean, so neither size nor type needs to be mentioned. Table 2-4 displays legal values for size and data types.

Example 2-9 uses vertex arrays for both RGBA colors and vertex coordinates. RGB floating-point values and their corresponding (x, y) integer coordinates are loaded into the GL_COLOR_ARRAY and GL_VERTEX_ARRAY.

Example 2-9 : Enabling and Loading Vertex Arrays: varray.c

static GLint vertices[] = {25, 25,

                          100, 325,

                          175, 25,

                          175, 325,

                          250, 25,

                          325, 325};

static GLfloat colors[] = {1.0, 0.2, 0.2,

                          0.2, 0.2, 1.0,

                          0.8, 1.0, 0.2,

                          0.75, 0.75, 0.75,

                          0.35, 0.35, 0.35,

                          0.5, 0.5, 0.5};

glEnableClientState (GL_COLOR_ARRAY);

glEnableClientState (GL_VERTEX_ARRAY);

glColorPointer (3, GL_FLOAT, 0, colors);

glVertexPointer (2, GL_INT, 0, vertices);

Stride

With a stride of zero, each type of vertex array (RGB color, color index, vertex coordinate, and so on) must be tightly packed. The data in the array must be homogeneous; that is, the data must be all RGB color values, all vertex coordinates, or all some other data similar in some fashion.

Using a stride of other than zero can be useful, especially when dealing with interleaved arrays. In the following array of GLfloats, there are six vertices. For each vertex, there are three RGB color values, which alternate with the (x, y, z) vertex coordinates.

static GLfloat intertwined[] =

      {1.0, 0.2, 1.0, 100.0, 100.0, 0.0,

       1.0, 0.2, 0.2, 0.0, 200.0, 0.0,

       1.0, 1.0, 0.2, 100.0, 300.0, 0.0,

       0.2, 1.0, 0.2, 200.0, 300.0, 0.0,

       0.2, 1.0, 1.0, 300.0, 200.0, 0.0,

       0.2, 0.2, 1.0, 200.0, 100.0, 0.0};

Stride allows a vertex array to access its desired data at regular intervals in the array. For example, to reference only the color values in the intertwined array, the following call starts from the beginning of the array (which could also be passed as &intertwined[0]) and jumps ahead 6 * sizeof(GLfloat) bytes, which is the size of both the color and vertex coordinate values. This jump is enough to get to the beginning of the data for the next vertex.

glColorPointer (3, GL_FLOAT, 6 * sizeof(GLfloat), intertwined);

For the vertex coordinate pointer, you need to start from further in the array, at the fourth element of intertwined (remember that C programmers start counting at zero).

glVertexPointer(3, GL_FLOAT,6*sizeof(GLfloat), &intertwined[3]);

Step 3: Dereferencing and Rendering

Until the contents of the vertex arrays are dereferenced, the arrays remain on the client side, and their contents are easily changed. In Step 3, contents of the arrays are obtained, sent down to the server, and then sent down the graphics processing pipeline for rendering.

There are three ways to obtain data: from a single array element (indexed location), from a sequence of array elements, and from an ordered list of array elements.

Dereference a Single Array Element

void glArrayElement(GLint ith)

Obtains the data of one (the ith) vertex for all currently enabled arrays. For the vertex coordinate array, the corresponding command would be glVertex[size][type]v(), where size is one of [2,3,4], and type is one of [s,i,f,d] for GLshort, GLint, GLfloat, and GLdouble respectively. Both size and type were defined by glVertexPointer(). For other enabled arrays, glArrayElement() calls glEdgeFlagv(), glTexCoord[size][type]v(), glColor[size][type]v(), glIndex[type]v(), and glNormal[type]v(). If the vertex coordinate array is enabled, the glVertex*v() routine is executed last, after the execution (if enabled) of up to five corresponding array values.

glArrayElement() is usually called between glBegin() and glEnd(). (If called outside, glArrayElement() sets the current state for all enabled arrays, except for vertex, which has no current state.) In Example 2-10, a triangle is drawn using the third, fourth, and sixth vertices from enabled vertex arrays (again, remember that C programmers begin counting array locations with zero).

Example 2-10 : Using glArrayElement() to Define Colors and Vertices

glEnableClientState (GL_COLOR_ARRAY);

glEnableClientState (GL_VERTEX_ARRAY);

glColorPointer (3, GL_FLOAT, 0, colors);

glVertexPointer (2, GL_INT, 0, vertices);

glBegin(GL_TRIANGLES);

glArrayElement (2);

glArrayElement (3);

glArrayElement (5);

glEnd();

When executed, the latter five lines of code has the same effect as

glBegin(GL_TRIANGLES);

glColor3fv(colors+(2*3*sizeof(GLfloat));

glVertex3fv(vertices+(2*2*sizeof(GLint));

glColor3fv(colors+(3*3*sizeof(GLfloat));

glVertex3fv(vertices+(3*2*sizeof(GLint));

glColor3fv(colors+(5*3*sizeof(GLfloat));

glVertex3fv(vertices+(5*2*sizeof(GLint));

glEnd();

Since glArrayElement() is only a single function call per vertex, it may reduce the number of function calls, which increases overall performance.

Be warned that if the contents of the array are changed between glBegin() and glEnd(), there is no guarantee that you will receive original data or changed data for your requested element. To be safe, don't change the contents of any array element which might be accessed until the primitive is completed.

Dereference a List of Array Elements

glArrayElement() is good for randomly "hopping around" your data arrays. A similar routine, glDrawElements(), is good for hopping around your data arrays in a more orderly manner.

void glDrawElements(GLenum mode, GLsizei count, GLenum type,
void *indices);

Defines a sequence of geometric primitives using count number of elements, whose indices are stored in the array indices. type must be one of GL_UNSIGNED_BYTE, GL_UNSIGNED_SHORT, or GL_UNSIGNED_INT, indicating the data type of the indices array. mode specifies what kind of primitives are constructed and is one of the same values that is accepted by glBegin(); for example, GL_POLYGON, GL_LINE_LOOP, GL_LINES, GL_POINTS, and so on.

The effect of glDrawElements() is almost the same as this command sequence:

int i;

glBegin (mode);

for (i = 0; i < count; i++)

   glArrayElement(indices[i]);

glEnd();

glDrawElements() additionally checks to make sure mode, count, and type are valid. Also, unlike the preceding sequence, executing glDrawElements() leaves several states indeterminate. After execution of glDrawElements(), current RGB color, color index, normal coordinates, texture coordinates, and edge flag are indeterminate if the corresponding array has been enabled.

With glDrawElements(), the vertices for each face of the cube can be placed in an array of indices. Example 2-11 shows two ways to use glDrawElements() to render the cube. Figure 2-15 shows the numbering of the vertices used in Example 2-11.

Figure 2-15 : Cube with Numbered Vertices

Example 2-11 : Two Ways to Use glDrawElements()

static GLubyte frontIndices = {4, 5, 6, 7};

static GLubyte rightIndices = {1, 2, 6, 5};

static GLubyte bottomIndices = {0, 1, 5, 4};

static GLubyte backIndices = {0, 3, 2, 1};

static GLubyte leftIndices = {0, 4, 7, 3};

static GLubyte topIndices = {2, 3, 7, 6};

glDrawElements(GL_QUADS, 4, GL_UNSIGNED_BYTE, frontIndices);

glDrawElements(GL_QUADS, 4, GL_UNSIGNED_BYTE, rightIndices);

glDrawElements(GL_QUADS, 4, GL_UNSIGNED_BYTE, bottomIndices);

glDrawElements(GL_QUADS, 4, GL_UNSIGNED_BYTE, backIndices);

glDrawElements(GL_QUADS, 4, GL_UNSIGNED_BYTE, leftIndices);

glDrawElements(GL_QUADS, 4, GL_UNSIGNED_BYTE, topIndices);

Or better still, crunch all the indices together:

static GLubyte allIndices = {4, 5, 6, 7, 1, 2, 6, 5, 

 0, 1, 5, 4, 0, 3, 2, 1,

 0, 4, 7, 3, 2, 3, 7, 6};

glDrawElements(GL_QUADS, 24, GL_UNSIGNED_BYTE, allIndices);

Note: It is an error to encapsulate glDrawElements() between a glBegin()/glEnd() pair.

With both glArrayElement() and glDrawElements(), it is also possible that your OpenGL implementation caches recently processed vertices, allowing your application to "share" or "reuse" vertices. Take the aforementioned cube, for example, which has six faces (polygons) but only eight vertices. Each vertex is used by exactly three faces. Without glArrayElement() or glDrawElements(), rendering all six faces would require processing twenty-four vertices, even though sixteen vertices would be redundant. Your implementation of OpenGL may be able to minimize redundancy and process as few as eight vertices. (Reuse of vertices may be limited to all vertices within a single glDrawElements() call or, for glArrayElement(), within one glBegin()/glEnd() pair.)

Dereference a Sequence of Array Elements

While glArrayElement() and glDrawElements() "hop around" your data arrays, glDrawArrays() plows straight through them.

void glDrawArrays(GLenum mode, GLint first, GLsizei count);

Constructs a sequence of geometric primitives using array elements starting at first and ending at first+count-1 of each enabled array. mode specifies what kinds of primitives are constructed and is one of the same values accepted by glBegin(); for example, GL_POLYGON, GL_LINE_LOOP, GL_LINES, GL_POINTS, and so on.

The effect of glDrawArrays() is almost the same as this command sequence:

int i;

glBegin (mode);

for (i = 0; i < count; i++)

   glArrayElement(first + i);

glEnd();

As is the case with glDrawElements(), glDrawArrays() also performs error checking on its parameter values and leaves the current RGB color, color index, normal coordinates, texture coordinates, and edge flag with indeterminate values if the corresponding array has been enabled.

Try This

• Change the icosahedron drawing routine in Example 2-13 to use vertex arrays.

Interleaved Arrays

Advanced

Earlier in this chapter (in "Stride"), the special case of interleaved arrays was examined. In that section, the array intertwined, which interleaves RGB color and 3D vertex coordinates, was accessed by calls to glColorPointer() and glVertexPointer(). Careful use of stride helped properly specify the arrays.

static GLfloat intertwined[] =

      {1.0, 0.2, 1.0, 100.0, 100.0, 0.0,

       1.0, 0.2, 0.2, 0.0, 200.0, 0.0,

       1.0, 1.0, 0.2, 100.0, 300.0, 0.0,

       0.2, 1.0, 0.2, 200.0, 300.0, 0.0,

       0.2, 1.0, 1.0, 300.0, 200.0, 0.0,

       0.2, 0.2, 1.0, 200.0, 100.0, 0.0};

There is also a behemoth routine, glInterleavedArrays(), that can specify several vertex arrays at once. glInterleavedArrays() also enables and disables the appropriate arrays (so it combines both Steps 1 and 2). The array intertwined exactly fits one of the fourteen data interleaving configurations supported by glInterleavedArrays(). So to specify the contents of the array intertwined into the RGB color and vertex arrays and enable both arrays, call

glInterleavedArrays (GL_C3F_V3F, 0, intertwined);

This call to glInterleavedArrays() enables the GL_COLOR_ARRAY and GL_VERTEX_ARRAY arrays. It disables the GL_INDEX_ARRAY, GL_TEXTURE_COORD_ARRAY, GL_NORMAL_ARRAY, and GL_EDGE_FLAG_ARRAY.

This call also has the same effect as calling glColorPointer() and glVertexPointer() to specify the values for six vertices into each array. Now you are ready for Step 3: Calling glArrayElement(), glDrawElements(), or glDrawArrays() to dereference array elements.

void glInterleavedArrays(GLenum format, GLsizei stride, void *pointer)

Initializes all six arrays, disabling arrays that are not specified in format, and enabling the arrays that are specified. format is one of 14 symbolic constants, which represent 14 data configurations; Table 2-5 displays format values. stride specifies the byte offset between consecutive vertexes. If stride is 0, the vertexes are understood to be tightly packed in the array. pointer is the memory address of the first coordinate of the first vertex in the array.

Note that glInterleavedArrays() does not support edge flags.

The mechanics of glInterleavedArrays() are intricate and require reference to Example 2-12 and Table 2-5. In that example and table, you'll see et, ec, and en, which are the boolean values for the enabled or disabled texture coordinate, color, and normal arrays, and you'll see st, sc, and sv, which are the sizes (number of components) for the texture coordinate, color, and vertex arrays. tc is the data type for RGBA color, which is the only array that can have non-float interleaved values. pc, pn, and pv are the calculated strides for jumping over individual color, normal, and vertex values, and s is the stride (if one is not specified by the user) to jump from one array element to the next.

The effect of glInterleavedArrays() is the same as calling the command sequence in Example 2-12 with many values defined in Table 2-5. All pointer arithmetic is performed in units of sizeof(GL_UNSIGNED_BYTE).

Example 2-12 : Effect of glInterleavedArrays(format, stride, pointer)

int str;

/*  set et, ec, en, st, sc, sv, tc, pc, pn, pv, and s

 *  as a function of Table 2-5 and the value of format

 */

str = stride;

if (str == 0)

   str = s;

glDisableClientState(GL_EDGE_FLAG_ARRAY);

glDisableClientState(GL_INDEX_ARRAY);

if (et) {

   glEnableClientState(GL_TEXTURE_COORD_ARRAY);

   glTexCoordPointer(st, GL_FLOAT, str, pointer);

}

else

   glDisableClientState(GL_TEXTURE_COORD_ARRAY);

if (ec) {

   glEnableClientState(GL_COLOR_ARRAY);

   glColorPointer(sc, tc, str, pointer+pc);

}

else

   glDisableClientState(GL_COLOR_ARRAY);

if (en) {

   glEnableClientState(GL_NORMAL_ARRAY);

   glNormalPointer(GL_FLOAT, str, pointer+pn);

}

else

   glDisableClientState(GL_NORMAL_ARRAY);

glEnableClientState(GL_VERTEX_ARRAY);

glVertexPointer(sv, GL_FLOAT, str, pointer+pv);

In Table 2-5, T and F are True and False. f is sizeof(GL_FLOAT). c is 4 times sizeof(GL_UNSIGNED_BYTE), rounded up to the nearest multiple of f.

Start by learning the simpler formats, GL_V2F, GL_V3F, and GL_C3F_V3F. If you use any of the formats with C4UB, you may have to use a struct data type or do some delicate type casting and pointer math to pack four unsigned bytes into a single 32-bit word.

For some OpenGL implementations, use of interleaved arrays may increase application performance. With an interleaved array, the exact layout of your data is known. You know your data is tightly packed and may be accessed in one chunk. If interleaved arrays are not used, the stride and size information has to be examined to detect whether data is tightly packed.

Note: glInterleavedArrays() only enables and disables vertex arrays and specifies values for the vertex-array data. It does not render anything. You must still complete Step 3 and call glArrayElement(), glDrawElements(), or glDrawArrays() to dereference the pointers and render graphics.