[question] Are brdfLUT sampling coordinates correct?

Hi!

I was going through several PBR implementation and noticed that this implementation samples the BRDF LUT texture using the (1.0 - roughness) as a second texture coordinate: https://github.com/KhronosGroup/glTF-WebGL-PBR/blob/master/shaders/pbr-frag.glsl#L151 texture2D(u_brdfLUT, vec2(pbrInputs.NdotV, 1.0 - pbrInputs.perceptualRoughness))

while all other implementations use just roughness (no “1.0 - roughness” flip):

1. Google’s Filament: https://github.com/google/filament/blob/master/shaders/src/light_indirect.fs#L58
2. LearnOpenGL: https://github.com/JoeyDeVries/LearnOpenGL/blob/master/src/6.pbr/2.2.2.ibl_specular_textured/2.2.2.pbr.fs#L162
3. McNopper’s OpenGL: https://github.com/McNopper/OpenGL/blob/master/Example33/shader/brdf.frag.glsl#L46
4. SaschaWillems’ Vulkan-glTF: https://github.com/SaschaWillems/Vulkan-glTF-PBR/blob/master/data/shaders/pbr.frag#L175
5. SaschaWillems’ Vulkan: https://github.com/SaschaWillems/Vulkan/blob/master/data/shaders/pbribl/pbribl.frag#L136

LUT files across the different implementation look similar (see below). This fact makes me wonder who has it right? Our LUT: Filament’s LUT: LearnOpenGL’s LUT:

[javagl]

As far as I understood, the reference viewer is about to be developed in this fork/branch: https://github.com/ux3d/glTF-WebGL-PBR/tree/reference-viewer , but it only just started. Correct me if I’m wrong.

[lhog]

Guess this one is sorted out. Thanks a lot for your inputs and closing it for now.