Fix meshlet shaders for bindless mode. #16825
Conversation
We have to extract the material ID from the mesh and stuff it in the vertex during visibility buffer resolution.
pcwalton
added
A-Rendering
| @@ -5,8 +5,7 @@ | |||
| mesh_bindings::mesh | |||
| } | |||
|
|
|||
| fn sample_depth_map(uv: vec2<f32>, instance_index: u32) -> f32 { | |||
| let slot = mesh[instance_index].material_bind_group_slot; | |||
| fn sample_depth_map(uv: vec2<f32>, material_bind_group_slot: u32) -> f32 { | |||
There was a problem hiding this comment.
Why did this change? Was it missed in an earlier PR?
There was a problem hiding this comment.
Not quite sure what you mean. It changed because we don't use that mesh array in visbuffer resolution. If you're asking whether I didn't fix meshlets in an earlier PR, yes, I broke them (that's the whole point of this PR).
There was a problem hiding this comment.
Ah nvm I see that it used to get the slot in the function itself, and all you did was move it outside the function so that the slot could be obtained differently for meshlet meshes.
| @@ -12,6 +12,7 @@ | |||
| } | |||
| #import bevy_render::maths::{affine3_to_square, mat2x4_f32_to_mat3x3_unpack} | |||
|
|
|||
| #ifndef MESHLET_MESH_MATERIAL_PASS | |||
There was a problem hiding this comment.
Why were the #ifndef MESHLET_MESH_MATERIAL_PASS added here? Did they cause compilation issues?
There was a problem hiding this comment.
The mesh array that's used in those functions isn't present during visbuffer resolution.
There was a problem hiding this comment.
Right, but I don't think that causes any actual issues, unless you try and use them? It was never an issue in 0.15. Unless a recent PR changed that. Anyways I'm fine with the ifdef, I was just wondering why it was added.
|
I can't test meshlets atm (don't have access to my desktop), but approving on the assumption that it was tested and this works. Code looks fine. |
We have to extract the material ID from the mesh and stuff it in the vertex during visibility buffer resolution.