ggxcc

GGX cube map convolver for ioquake3's OpenGL2 renderer

So what's this for?

The OpenGL2 renderer in ioquake3 uses cube maps for specular reflection, and its internal shaders assume that the mipmaps in each cube map correspond to a linear roughness scale. However, the renderer uses a simple downsampling method for these mipmaps, causing reflections to look pixelated and too sharp.

This program does offline GGX convolution for a cubemap, which can then be reinserted into the renderer for proper reflections.

Why isn't this part of OpenGL2?

At the moment, this operation is too slow to be done at level load. On my machine, a i5-2500k, it takes about 16 seconds to convolve a single 256x256 cubemap, and levels usually contain 8 or more cubemaps to process. If in the development of this tool the processing becomes much faster, it will be integrated into the OpenGL2 renderer, but not before.