PBR spotlight falloff #31
Conversation
There was a problem hiding this comment.
So, if I understand this correctly, spotExponent is just used as a way to pipe data in and no attempt is made to convert exponent values so a user can get reasonable responses to adjusting this value? I would prefer to avoid simplepbr-specifc trickery so it remains working as a drop-in replacement for Panda's auto shader.
| float shadowSpot = smoothstep(spotcutoff-SPOTSMOOTH, spotcutoff+SPOTSMOOTH, spotcos); | ||
|
|
||
| // https://www.pbr-book.org/3ed-2018/Light_Sources/Point_Lights#SpotLight::Falloff | ||
| float cosTheta = dot(normalize(p3d_LightSource[i].spotDirection), -l); |
There was a problem hiding this comment.
I would prefer that these didn't get renamed.
|
Also see Moguri/panda3d-gltf#85 for how the glTF converter converts the outer distance into an exponent. If you're going to use a linear falloff instead of an exponent, it should probably use the inverse of that calculation. That said, is there a good reason not to just apply the exponent the way Panda does it here? |
|
The awkward answer is "the systems I'm using Panda3d to prototype for all use inner-angle" Longer answer: before switching to simplePBR, I'd done similar not-entirely-right work to set an exponent to approximate inner-angle. Later I accidentally discovered a side-effect where, in the regular pipeline, setting the scale of a spotlight had an influence on the inner angle Is there a way to add uniforms to a light instance? If so, I can dispense with the dirty hack and make pbr-inner-angle cleanly opt-in. (and maybe even figure out how to do area lights) |
|
Closing due to inactivity |
implements spotlight falloff as defined at pbr-book.org. It's juuust slightly dirty in that it's smuggling the falloff angle (ok. the cos of the falloff angle) in by way of the existing panda3d spotlight exponent property, which... wasn't supported by simplePBR anyway.