I've implemented the first stage of my Chunked LOD terrain engine, which simply decides on when to split a tile into 4 sub-tiles based on the distance from the camera to the center of the tile, i.e:
void TerrainRenderer::RenderTile(const float& left, const float& right, const float& top, const float& bottom, const XMFLOAT4& colour, int& tilesRendered, int& tilesFrustumCulled)
{
// Calculate the centre of the quad
XMVECTOR centre = XMVectorSet((left + right) / 2.0f, 0.0f, (top + bottom) / 2.0f, 0.0f);
// We'll need the width and height of the cells
float width = right - left;
float height = bottom - top;
// Frustum culling removed for brevity
// Calculate the distance from the camera to the centre of this quad
XMVECTOR diff = Renderer::GetInstance()->GetCameraPosVec() - centre;
float distance = XMVector3Length(diff).m128_f32[0];
// Should we render here or split?
if (distance < (width * 2.0f))
{
// Too close, split into 4 sub-tiles
RenderTile(left, left + (width/2.0f), top, top + (height/2.0f), constColourRed, tilesRendered, tilesFrustumCulled); // Top left
RenderTile(left + (width/2.0f), right, top, top + (height/2.0f), constColourGreen, tilesRendered, tilesFrustumCulled); // Top right
RenderTile(left + (width/2.0f), right, top + (height/2.0f), bottom, constColourBlue, tilesRendered, tilesFrustumCulled); // Bottom right
RenderTile(left, left + (width/2.0f), top + (height/2.0f), bottom, constColourYellow, tilesRendered, tilesFrustumCulled); // Bottom left
}
else
{
// No need to split, render here
// Actual drawing code removed for brevity
tilesRendered++;
}
}
This works fairly well, but the next step is to get the vertices to warp between levels so that there is no popping when a switch between levels occurs.
The original article by Thatcher Ulrich says:
At rendering time, we will compute a chunk's morph parameter, ensuring that the morph parameter is always 0 when the chunk is about to split and 1 when the chunk is about to merge.[/quote]
The problem is how do I calculate this?
Its simple to calculate when a tile is going to split, so that's fine. But how does a tile know when its close to merging?
Let me explain using a diagram:
The Green circle is the eye\camera, the red dots are the center of the tile, the numbers are the distance etc.
On the left image we can see a tile just before it splits. The distance from the eye to the center of the tile is 100.
On the right image the camera has moved one unit closer, which causes the tile to split into 4. The distance from the camera to the center of the bigger original tile (Shown in purple) is now 99, however the distances to the centers of the new smaller tiles are 86, 88, 105 and 108 (These are rough guesses, not actual calculated values).
So this means that the morph factor for each of the children should now be at (Or very close to) 1, but given that they all have different distances to the camera how do I make it so that they all have a value of 1 at this point?
If I were to calculate the morph factors using simple distance to the center of the tile based math then we'd get a pop in the morph value when the tiles split, as just before the split the distance will be 100, and just after we'll have 86, 88, 105 and 108, so there is a jump of -14, -12, 5 and 8 respectively.
Another way of explaining it is that whilst a tile knows how close it is to splitting, it doesn't know when it will merge (Only the parent tile knows that). Does this mean we have to pass this info down from the parent tile to its children?
It was a bit difficult to explain the problem so I hope it made sense?
Any help would be appreciated.
Thanks
Ben
