KHR_materials_anisotropy extension branch.

[UX3D-nopper]

No description provided.

[flashk]

The schema does not mention which channel anisotropyTexture is sampled from, but the example implementation shows it being sampled from the red channel. Should it be sampled from the alpha channel instead? This would allow combining it with anisotropyDirectionTexture.

[UX3D-nopper]

@UX3D-eckerlein Can you please cross check, if this packing is possible?
If yes, please adapt to code and schema.

[emackey]

@elalish Can you double-check my GLSL math in e0333e7? I added the mat2 thing where anisotropyDirection is now allowed to rotate the vector from the texture, but I don't have an easy way to test it.

[elalish]

@emackey Your math looks pretty good to me - sorry, I put a couple of comments in your commit, which makes them hard to find. But we definitely need to update the spec a bit more.

[emackey]

Per today's meeting, the anisotropy strength formula has been set.

alphaTangent = mix(alphaRoughness, 1.0, anisotropy * anisotropy);
alphaBitangent = alphaRoughness;

("alpha" is roughness squared here.)

Using the material's roughness along the bitangent (or along the direction perpendicular to the chosen anisotropy direction) allows rasterizers to continue using material roughness as a base lookup into the IBL, unmodified. The tangent direction then needs an approximation such as stretching or multisampling, but only along one axis.

[emackey]

This was approved today for "Release Candidate" status by the PBR TSG.