I am calculating my specular highlights in world space, in HLSL.
There is an XZ plane and a light slightly above the plane.
The problem is, the farther the light is from the origin of world space, the farther the highlight on the plane is from the light (and closer to the origin). (If the light moves right, the highlight also moves right but slower, so it is somewhere between the origin and the light).
When the camera moves, although the highlights change shape, they stay in the same place.
What could cause this problem?
Specular Highlight in the wrong place?
1. Is your view vector more than just float3(0, 0, 1)?
2. Are you world space normals calculated from float3 object space normals? If you would use float4's they would get translated (position) into world space and not only rotated.
3. Is everything properly normalized?
4. Are your dot products properly clamped between 0 and 1?
5. Are you sure your view vector is in world space (also check for 2. here)?
If this doesn't solve your problems, I'd recommend posting your code.
2. Are you world space normals calculated from float3 object space normals? If you would use float4's they would get translated (position) into world space and not only rotated.
3. Is everything properly normalized?
4. Are your dot products properly clamped between 0 and 1?
5. Are you sure your view vector is in world space (also check for 2. here)?
If this doesn't solve your problems, I'd recommend posting your code.
Author
Quote:Original post by DarkChris
1. Is your view vector more than just float3(0, 0, 1)?
2. Are you world space normals calculated from float3 object space normals? If you would use float4's they would get translated (position) into world space and not only rotated.
3. Is everything properly normalized?
4. Are your dot products properly clamped between 0 and 1?
5. Are you sure your view vector is in world space (also check for 2. here)?
If this doesn't solve your problems, I'd recommend posting your code.
No, I don't think those are the problem...
[Edited by - xytor on October 15, 2010 7:21:08 PM]
Well I've read through your first post again and also found no mistakes in your shader... But the only thing that could cause this problems and explains why there's no mistake in the shader would be, if the cameras position would always be (0,0,0) in world space...
Author
Quote:Original post by DarkChris
Well I've read through your first post again and also found no mistakes in your shader... But the only thing that could cause this problems and explains why there's no mistake in the shader would be, if the cameras position would always be (0,0,0) in world space...
No, it's not. However, in the vertex shader, doing Vert.posWorld.xz*=0.5; makes it better. What's going on?
Ok than you should either try debugging it by using PIX or visualizing every line of the code step by step, by returning the calculated values as the resulting color... You could at least try if the ViewVector colors change when moving the camera...
A wild guess: Maybe posCam in
float3 view = normalize(posCam-IN.posWorld);
is not properly initialized?
float3 view = normalize(posCam-IN.posWorld);
is not properly initialized?
Author
Quote:Original post by knighty
A wild guess: Maybe posCam in
float3 view = normalize(posCam-IN.posWorld);
is not properly initialized?
posCam is indeed properly initialized, I checked.
Actually, the problem seems to be IN.posWorld. This is a TEXCOORD1 that is passed from the vertex shader. For some reason, shader model 3 doesn't like POSITION to be used in the pixel shader.
Anyway, is it possible that the shader is interpolating the posWorld in a weird way?
Author
Quote:Original post by Rubicon
Have you tried labelling all your vs out and ps in as TEXCOORDn ?
Doesn't the Vertex Shader need to output at least POSITION?
I tried labeling everything else as TEXCOORDn, but that changed nothing.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement