hey everybody 2 (simple) questions for the advanced ogl programmer: If I have a rendering context I am bound to the pixelformat. So I give in Floating Point values the are clamped to 0.1 So If I give in bytes everythin is fine. But how does the gl handle these values internally ? If I put in 0xFE for flat shaded polygon, and I read out the back buffer. Is it assured that I get the same values back ???? Or are there some roundings ? 2. Has somebody a simple example source sniipped for doing offscreen image processing with shaders. I only need the projection modelview and the coordinates for the screen aligned quad. and the appropriate shader to make texels = pixels. Or does somebody know an integratation / summerize shader ? I mean you put in a texture and get at the end a texture with size 1 and the sum of all pixel values from the source target. thanks!!

Ok, I assumed that for interpolation. But If I am using flat shading with
only one color with 8bit each channel?

I implemented a shader program for dense matrix multiplication some years ago. It can speed up the calculation for about 20 times comparing with the CPU matrix multiplication. Therefore, I can tell you that it is feasible to do integratation using shader.

And, there are even more advanced research that using GPU to do matrix decomposition ( http://www.cs.unc.edu/~geom/Numeric/svd/ ). I remembered that they posted their paper together with the source code on their website. Try to search with google, maybe you can find it.