Input to a lighting function for a single layer (either the base layer or the clearcoat layer).

Input to a lighting function (point_light, spot_light, directional_light).

Values derived from the LightingInput for both diffuse and specular lights.

light radius is a non-physical construct for efficiency purposes, because otherwise every light affects every fragment in the scene

Falloff without the distance attenuation, for lights that have it baked-in (eg LTC lights)

Simple implementation, has precision problems when using fp16 instead of fp32 see https://google.github.io/filament/Filament.md.html#listing_speculardfp16

This is from the KHR_materials_anisotropy spec: https://github.com/KhronosGroup/glTF/blob/main/extensions/2.0/Khronos/KHR_materials_anisotropy/README.md#individual-lights

Visibility function (Specular G) V(v,l,a) = G(v,l,α) / { 4 (n⋅v) (n⋅l) } such that f_r becomes f_r(v,l) = D(h,α) V(v,l,α) F(v,h,f0) where V(v,l,α) = 0.5 / { n⋅l sqrt((n⋅v)^2 (1−α2) + α2) + n⋅v sqrt((n⋅l)^2 (1−α2) + α2) } Note the two sqrt’s, that may be slow on mobile, see https://google.github.io/filament/Filament.md.html#listing_approximatedspecularv

The visibility function, anisotropic variant.

Probability-density function that matches the bounded VNDF sampler https://gpuopen.com/download/Bounded_VNDF_Sampling_for_Smith-GGX_Reflections.pdf (Listing 2)

https://gpuopen.com/download/Bounded_VNDF_Sampling_for_Smith-GGX_Reflections.pdf (Listing 1)

Smith geometric shadowing function

https://google.github.io/filament/Filament.md.html#materialsystem/clearcoatmodel

Fresnel function see https://google.github.io/filament/Filament.md.html#citation-schlick94 F_Schlick(v,h,f_0,f_90) = f_0 + (f_90 − f_0) (1 − v⋅h)^5

Multiscattering approximation: https://google.github.io/filament/Filament.md.html#listing_energycompensationimpl

N, V, and L must all be normalized.

Returns L in the xyz components and the modified roughness in the w component.

Cook-Torrance approximation of the microfacet model integration using Fresnel law F to model f_m f_r(v,l) = { D(h,α) G(v,l,α) F(v,h,f0) } / { 4 (n⋅v) (n⋅l) }

https://google.github.io/filament/Filament.md.html#listing_clearcoatbrdf

Disney approximation See https://google.github.io/filament/Filament.md.html#citation-burley12 minimal quality difference

Scale/bias approximation

The ideal solution is to switch to Linearly Transformed Cosines, which is more accurate in all cases, and a popular choice for realtime.

Integrate one edge of the spherical polygon formed by the LTC-transformed quad. Implements Eq. 11 using a polynomial approximation based on https://advances.realtimerendering.com/s2016/s2016_ltc_rnd.pdf Using the equation as-is produces visible artifacts.