Creating spherical meshes with Direct x? - c++

Question

How do you go about creating a sphere with meshes in Direct-x? I'm using C++ and the program will be run on windows, only.
Everything is currently rendered through an IDiRECT3DDEVICE9 object.

Answer 1

You could use the D3DXCreateSphere function.

Answer 2

There are lots of ways to create a sphere.
One is to use polar coordinates to generate slices of the sphere.
struct Vertex
{
float x, y, z;
float nx, ny, nz;
};
Given that struct you'd generate the sphere as follows (I haven't tested this so I may have got it slightly wrong).
std::vector< Vertex > verts;
int count = 0;
while( count < numSlices )
{
const float phi = M_PI / numSlices;
int count2 = 0;
while( count2 < numSegments )
{
const float theta = M_2PI / numSegments
const float xzRadius = fabsf( sphereRadius * cosf( phi ) );
Vertex v;
v.x = xzRadius * cosf( theta );
v.y = sphereRadius * sinf( phi );
v.z = xzRadius * sinf( theta );
const float fRcpLen = 1.0f / sqrtf( (v.x * v.x) + (v.y * v.y) + (v.z * v.z) );
v.nx = v.x * fRcpLen;
v.ny = v.y * fRcpLen;
v.nz = v.z * fRcpLen;
verts.push_back( v );
count2++;
}
count++;
}
This is how D3DXCreateSphere does it i believe. Of course the code above does not form the faces but thats not a particularly complex bit of code if you set your mind to it :)
The other, and more interesting in my opinion, way is through surface subdivision.
If you start with a cube that has normals defined the same way as the above code you can recursively subdivide each side. Basically you find the center of the face. Generate a vector from the center to the new point. Normalise it. Push the vert out to the radius of the sphere as follows (Assuming v.n* is the normalised normal):
v.x = v.nx * sphereRadius;
v.y = v.ny * sphereRadius;
v.z = v.nz * sphereRadius;
You then repeat this process for the mid point of each edge of the face you are subdividing.
Now you can split each face into 4 new quadrilateral faces. You can then subdivide each of those quads into 4 new quads and so on until you get to the refinement level you require.
Personally I find this process provides a nicer vertex distribution on the sphere than the first method.