LAR Gaming

For Games, Perfomance is EVERYTHING!

The biggest the FPS, the better the game is to the player.
And, because of this, I have decided to implement a new function in LarX’s rendering system: Culling, in my case, Frustum Culling.
Well, you may thing: WTF?
Culling is not rendering the objects that are not being showed in screen. It saves performance, because we don’t need to do unnecessary calcs, either in the CPU(Matrix, several loops…) and in the GPU…
It’s extremely simple to do this in XNA:
In your Camera class, we only have to create a BoundingFrustum, and initialize it with the camera’s View and Projection Matrixes:

Frustum = new BoundingFrustum(view * projection);

And in the Camera’s Update Method, Update the Frustum’s Matrix:

Frustum.Matrix = View * Projection;

Then, in the method that renders everything, we only need to test, foreach bounding sphere of the models, if they are on screen:

//Foreach object to be rendered
for (int j = 0; j < Objects.Count; j++)
{
// Starts a ContainmentType, of type Disjoint
ContainmentType currentContainmentType = ContainmentType.Disjoint;

//Gets the Sphere from the obj…
BoundingSphere meshBoundingSphere = Objects[j].Sphere;
//…and test it with the camera frustum
currentContainmentType = Camera.Frustum.Contains(meshBoundingSphere);
//If the sphere is showed on screen…
if (currentContainmentType != ContainmentType.Disjoint)
{
// …Render
DrawModel(…);
}
}

Remember: You have to use your own way to calculate the BoundingSphere of the model. Soon as I get my notebook back, I’ll post my way here.

ps.: This is a very good way, saves a lot of perfomance, but, remember: Some models, faaaar away, will be rendered anyway, because they are on camera’s view, either if there is a wall in front of it.

That’s it