(This is actually a crosspost from gamedev , but I thought you guys and girls might be helpful too!)
Edit: I elaborated about what I mean with overlay. After some sleep I have found that my call to glGenerateMipmaps is useless, since I use null as pointer to glTexImage2D.
On OpenGL 3.3, Ubuntu 11.04, nVidia drivers 280.13, GeForce GT240.
I’m making a deferred shader and wants to display my G-buffer in small overlays in the lower part of the screen. Overlays in this case meaning no z-test, textured screen aligned quads, and rendered last.
When rendering (one at a time) the G-buffers (FBO textures ) in full window they shows up correctly.
When rendering to these overlays I get trash (seemingly random GPU-memory as static interference, but different between runs, sometimes clear picture).
If I create a different texture, say from file, it shows without trash in the overlays, so that code should be OK.
Can anyone explain why? I have used glGenerateMipmap for the textures in the G-buffer.
I know this might be not enough information, but starts with this. Please ask about any relevant information.
Can the problem be related to FBO-rendering and mipmapping? Surely my code wants to show unreserved/uninitialized memory, but I don’t know why.
The full-window green picture is my current result from the lighting stage, in resolution (256*256) to save GPU memory because of fat G-buffer.
Also, in the material shader (generating the G-buffer) I use “out vec4 fragment[3]”, and the textures are of different internal formats. Can that be a problem?
PS.
The lights are at (-30,0,0) and (30,0,0) but seems to be “nearer”. Can that be a precision problem? I use GL_RGBA16F for the normals, GL_RGBA32F for the positions, GL_DEPTH_COMPONENT24 for depth and GL_RGBA for the albedo (Yes, an enormous waste, but I will optimize later).
OK, I found the problem. I generated mipmaps before any meaningful data was in the memory. Turning off mipmapping for the relevant textures solved the problem. Thank you anyway for your attention!
