Sign in to follow this  
Lil_Lloyd

INSANE quadtree bugs/artefacts

Recommended Posts

Hi people

I've been working on a quad tree demo for some time and used this gentleman's code as a study to learn: http://www.youtube.com/watch?v=9N-kgCqy2xs

 

However my results for the island scene look like this:

 

http://www.youtube.com/watch?v=pCvvwMWGLDA&list=UUtyzxNh2dw1ru0P2761QA7g&index=1

 

I started following the base code loosely, but to try and get the rendering of the quadtree correct I adhered quite closely to the terrain files as I was being driven slowly crazy. Even now when I see very little difference between the sets of code there must be some faults. 

 

The important code is as follows:

 

int halfWidth  = m_width/2;
	int halfHeight = m_height/2;

	//init verts
	vec3* vertData = new vec3[m_numVertices];

	for(int z = 0; z < m_height; z++){
		for(int x = 0; x < m_width; x++){

			int index = z * m_width + x;

			int imgIndexZ = z;
			int imgIndexX = x;
			if(z == imgHeight){
				imgIndexZ--;
			}
			if(x == imgWidth){
				imgIndexX--;
			}

			int imgIndex = imgIndexZ * m_width + imgIndexX;


			vertData[index].x = (x - halfWidth) * LANDSCAPE_SCALE;
			vertData[index].y = (float)data[imgIndex] * HEIGHT_SCALE;
			vertData[index].z = (z - halfHeight) * LANDSCAPE_SCALE;
		}
	}

 

The above is how I init my vertices values, with data being an image file holding the height values. The land mass itself renders fine if I render it as a whole block without a quadtree using an indexed VAO and GL_TRIANGLES.

 

The indices generating function of the terrain chunks if I go the alternative route, however, is as follows:

 

int heightmapDataPosX = posX;
int heightmapDataPosY = posY;
unsigned int offset = (unsigned int)pow(2.0f, (float)(lod));
unsigned int div = (unsigned int)pow(2.0f, (float)(depth+lod));
	
int heightmapDataSizeX = HMsizeX/(div) + 1;
int heightmapDataSizeY = HMsizeY/(div) + 1;

int nHMWidth   = heightmapDataSizeX;
int nHMHeight  = heightmapDataSizeY;
int nHMOffsetX = heightmapDataPosX;
int nHMOffsetY = heightmapDataPosY;

GLuint nHMTotalWidth = HMsizeX;
GLuint nHMTotalHeight = HMsizeY;

m_indOffsetW[lod] = nHMWidth*2;
m_indOffsetH[lod] = nHMHeight-1;
GLuint numIndices = (nHMWidth-3)*(nHMHeight-4)*2;
m_indices[lod] = new GLuint[numIndices]; 
m_numIndices[lod] = numIndices;

int index = 0;

for(GLuint y=1; y< nHMHeight-3; y++)
{
  for(GLuint x=1; x< nHMWidth-2; x++)
  {
    GLuint id0 = COORD(x*(offset) + nHMOffsetX, y*(offset) + nHMOffsetY);
    GLuint id1 = COORD(x*(offset) + nHMOffsetX, y*(offset) + nHMOffsetY+(offset));
    m_indices[lod][index++] = id0;
    m_indices[lod][index++] = id1;
  }
}

 

I have to alter the termination conditions for y and x and the numIndices because for some reason I get a memory access error, even though the numberOfVertices should be ok...

 

I know its masochistic but if anyone can help me I'd be so grateful! I've been going crazy for days... My github repo is here: https://github.com/LloydGhettoWolf/TitanicTerrain/tree/workingversion

Share this post


Link to post
Share on other sites

Your lack of replies is likely because it is hard to tell which bugs are related to your quadtree (which should be an octree, yes?) and which are related to something else.

For example you have missing triangles.  That would not be related to quadtrees unless you have done something horribly wrong and are in need of a major rewrite.

 

Besides, you already have  memory-access errors.

 

Fix everything else so we can see problems that are only related to the quadtree.

 

 

L. Spiro

Share this post


Link to post
Share on other sites

I fixed the big issues using a different approach found on the rastertek site, but thankyou for replying. However, now I have bizarre frustum culling issues where nearby quads are being culled. I'll post a separate thread for that.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this