Contact Shadows #22382
Contact Shadows #22382
Conversation
tychedelia
left a comment
tychedelia
left a comment
There was a problem hiding this comment.
Would you mind adding or updating an existing example?
|
Sure, I was going to ask the crowd where it would make sense to add to an example. |
alice-i-cecile
added
C-Feature
|
It looks like your PR has been selected for a highlight in the next release blog post, but you didn't provide a release note. Please review the instructions for writing release notes, then expand or revise the content in the release notes directory to showcase your changes. |
|
Closes #16164 |
|
Does it make sense to do the raymarching directly in the material shader? It might give better perf to pull it out to its own pass. Also MeshletMesh needs testing with this feature (and in the various forward/deferref/prepass modes). |
|
I have an example for contact shadows that I made in my old branch: https://github.com/pcwalton/bevy/blob/contact-shadows/examples/3d/contact_shadows.rs Feel free to steal it if you want, or not (it's rather old and would need to be updated). Up to you :) |
|
@JMS55 I don't think the raymarching can really be done anywhere else, because in forward you need to know what's in shadow right then. You can't really subtract a light after the fact. (Well, maybe you technically could make it work, but that'd be a huge can of worms.) |
| use_secant: u32, | ||
| }; | ||
|
|
||
| // Settings for contact shadows. |
There was a problem hiding this comment.
Mention the Rust module where the Rust-side equivalent of this can be found in the comment, so folks can easily cross-reference to find out what the settings mean.
| let L = normalize(view_bindings::clusterable_objects.data[light_id].position_radius.xyz - in.world_position.xyz); | ||
|
|
||
| let noise = utils::interleaved_gradient_noise(in.frag_coord.xy, view_bindings::globals.frame_count); | ||
| let normal_bias = in.world_normal * 0.005; |
There was a problem hiding this comment.
Can you pull this 0.005 out into a const somewhere? Ideally it'd be adjustable in a uniform, but at least we should put it at the top of a file somewhere so that the tunable settings are in one place.
| shadow *= smoothstep(0.5, 1.0, rm_result.hit_penetration_frac); | ||
| } | ||
| } | ||
| #endif |
There was a problem hiding this comment.
Is there no way you can avoid the duplication by pulling this out into a separate function?
|
|
||
| let rm_result = raymarch::depth_ray_march_march(&rm); | ||
| if rm_result.hit { | ||
| shadow *= smoothstep(0.5, 1.0, rm_result.hit_penetration_frac); |
There was a problem hiding this comment.
Do you actually need zero derivatives at 0.5 and 1 or would linstep be more appropriate? I'm not sure what motivates the use of smoothstep here.
5 participants
Objective
Solution
Showcase
Screen.Recording.2026-01-04.at.7.19.25.PM.mp4
Screen.Recording.2026-01-04.at.7.36.03.PM.mp4