Meshlet rendering (initial feature) (#10164)

# Objective
- Implements a more efficient, GPU-driven
(https://github.com/bevyengine/bevy/issues/1342) rendering pipeline
based on meshlets.
- Meshes are split into small clusters of triangles called meshlets,
each of which acts as a mini index buffer into the larger mesh data.
Meshlets can be compressed, streamed, culled, and batched much more
efficiently than monolithic meshes.


![image](https://github.com/bevyengine/bevy/assets/47158642/cb2aaad0-7a9a-4e14-93b0-15d4e895b26a)

![image](https://github.com/bevyengine/bevy/assets/47158642/7534035b-1eb7-4278-9b99-5322e4401715)

# Misc
* Future work: https://github.com/bevyengine/bevy/issues/11518
* Nanite reference:
https://advances.realtimerendering.com/s2021/Karis_Nanite_SIGGRAPH_Advances_2021_final.pdf
Two pass occlusion culling explained very well:
https://medium.com/@mil_kru/two-pass-occlusion-culling-4100edcad501

---------

Co-authored-by: Ricky Taylor <rickytaylor26@gmail.com>
Co-authored-by: vero <email@atlasdostal.com>
Co-authored-by: François <mockersf@gmail.com>
Co-authored-by: atlas dostal <rodol@rivalrebels.com>

Cargo.toml

@@ -321,6 +321,12 @@ embedded_watcher = ["bevy_internal/embedded_watcher"]
 # Enable stepping-based debugging of Bevy systems
 bevy_debug_stepping = ["bevy_internal/bevy_debug_stepping"]
 
+# Enables the meshlet renderer for dense high-poly scenes (experimental)
+meshlet = ["bevy_internal/meshlet"]
+
+# Enables processing meshes into meshlet meshes for bevy_pbr
+meshlet_processor = ["bevy_internal/meshlet_processor"]
+
 # Enable support for the ios_simulator by downgrading some rendering capabilities
 ios_simulator = ["bevy_internal/ios_simulator"]
 
@@ -950,6 +956,18 @@ description = "Demonstrates irradiance volumes"
 category = "3D Rendering"
 wasm = false
 
+[[example]]
+name = "meshlet"
+path = "examples/3d/meshlet.rs"
+doc-scrape-examples = true
+required-features = ["meshlet"]
+
+[package.metadata.example.meshlet]
+name = "Meshlet"
+description = "Meshlet rendering for dense high-poly scenes (experimental)"
+category = "3D Rendering"
+wasm = false
+
 [[example]]
 name = "lightmaps"
 path = "examples/3d/lightmaps.rs"

crates/bevy_core_pipeline/src/lib.rs

@@ -24,6 +24,8 @@ pub mod upscaling;
 pub use skybox::Skybox;
 
 /// Experimental features that are not yet finished. Please report any issues you encounter!
+///
+/// Expect bugs, missing features, compatibility issues, low performance, and/or future breaking changes.
 pub mod experimental {
     pub mod taa {
         pub use crate::taa::{

crates/bevy_internal/Cargo.toml

@@ -159,6 +159,12 @@ bevy_debug_stepping = [
   "bevy_app/bevy_debug_stepping",
 ]
 
+# Enables the meshlet renderer for dense high-poly scenes (experimental)
+meshlet = ["bevy_pbr?/meshlet"]
+
+# Enables processing meshes into meshlet meshes for bevy_pbr
+meshlet_processor = ["bevy_pbr?/meshlet_processor"]
+
 # Provides a collection of developer tools
 bevy_dev_tools = ["dep:bevy_dev_tools"]
 

crates/bevy_pbr/Cargo.toml

@@ -15,6 +15,10 @@ pbr_transmission_textures = []
 shader_format_glsl = ["bevy_render/shader_format_glsl"]
 trace = ["bevy_render/trace"]
 ios_simulator = ["bevy_render/ios_simulator"]
+# Enables the meshlet renderer for dense high-poly scenes (experimental)
+meshlet = []
+# Enables processing meshes into meshlet meshes
+meshlet_processor = ["dep:meshopt", "dep:thiserror"]
 
 [dependencies]
 # bevy
@@ -34,12 +38,17 @@ bevy_window = { path = "../bevy_window", version = "0.14.0-dev" }
 bevy_derive = { path = "../bevy_derive", version = "0.14.0-dev" }
 
 # other
+meshopt = { version = "0.2", optional = true }
+thiserror = { version = "1", optional = true }
 bitflags = "2.3"
 fixedbitset = "0.5"
 # direct dependency required for derive macro
 bytemuck = { version = "1", features = ["derive"] }
 radsort = "0.1"
 smallvec = "1.6"
+serde = { version = "1", features = ["derive", "rc"] }
+bincode = "1"
+range-alloc = "0.1"
 nonmax = "0.5"
 
 [lints]

crates/bevy_pbr/src/deferred/pbr_deferred_functions.wgsl

@@ -7,10 +7,16 @@
     rgb9e5,
     mesh_view_bindings::view,
     utils::{octahedral_encode, octahedral_decode},
-    prepass_io::{VertexOutput, FragmentOutput},
+    prepass_io::FragmentOutput,
     view_transformations::{position_ndc_to_world, frag_coord_to_ndc},
 }
 
+#ifdef MESHLET_MESH_MATERIAL_PASS
+#import bevy_pbr::meshlet_visibility_buffer_resolve::VertexOutput
+#else
+#import bevy_pbr::prepass_io::VertexOutput
+#endif
+
 #ifdef MOTION_VECTOR_PREPASS
     #import bevy_pbr::pbr_prepass_functions::calculate_motion_vector
 #endif
@@ -116,7 +122,11 @@ fn deferred_output(in: VertexOutput, pbr_input: PbrInput) -> FragmentOutput {
 #endif
     // motion vectors if required
 #ifdef MOTION_VECTOR_PREPASS
+#ifdef MESHLET_MESH_MATERIAL_PASS
+    out.motion_vector = in.motion_vector;
+#else
     out.motion_vector = calculate_motion_vector(in.world_position, in.previous_world_position);
+#endif
 #endif
 
     return out;

crates/bevy_pbr/src/extended_material.rs

@@ -67,6 +67,30 @@ pub trait MaterialExtension: Asset + AsBindGroup + Clone + Sized {
         ShaderRef::Default
     }
 
+    /// Returns this material's [`crate::meshlet::MeshletMesh`] fragment shader. If [`ShaderRef::Default`] is returned,
+    /// the default meshlet mesh fragment shader will be used.
+    #[allow(unused_variables)]
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_fragment_shader() -> ShaderRef {
+        ShaderRef::Default
+    }
+
+    /// Returns this material's [`crate::meshlet::MeshletMesh`] prepass fragment shader. If [`ShaderRef::Default`] is returned,
+    /// the default meshlet mesh prepass fragment shader will be used.
+    #[allow(unused_variables)]
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_prepass_fragment_shader() -> ShaderRef {
+        ShaderRef::Default
+    }
+
+    /// Returns this material's [`crate::meshlet::MeshletMesh`] deferred fragment shader. If [`ShaderRef::Default`] is returned,
+    /// the default meshlet mesh deferred fragment shader will be used.
+    #[allow(unused_variables)]
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_deferred_fragment_shader() -> ShaderRef {
+        ShaderRef::Default
+    }
+
     /// Customizes the default [`RenderPipelineDescriptor`] for a specific entity using the entity's
     /// [`MaterialPipelineKey`] and [`MeshVertexBufferLayoutRef`] as input.
     /// Specialization for the base material is applied before this function is called.
@@ -211,6 +235,30 @@ impl<B: Material, E: MaterialExtension> Material for ExtendedMaterial<B, E> {
         }
     }
 
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_fragment_shader() -> ShaderRef {
+        match E::meshlet_mesh_fragment_shader() {
+            ShaderRef::Default => B::meshlet_mesh_fragment_shader(),
+            specified => specified,
+        }
+    }
+
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_prepass_fragment_shader() -> ShaderRef {
+        match E::meshlet_mesh_prepass_fragment_shader() {
+            ShaderRef::Default => B::meshlet_mesh_prepass_fragment_shader(),
+            specified => specified,
+        }
+    }
+
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_deferred_fragment_shader() -> ShaderRef {
+        match E::meshlet_mesh_deferred_fragment_shader() {
+            ShaderRef::Default => B::meshlet_mesh_deferred_fragment_shader(),
+            specified => specified,
+        }
+    }
+
     fn specialize(
         pipeline: &MaterialPipeline<Self>,
         descriptor: &mut RenderPipelineDescriptor,

crates/bevy_pbr/src/lib.rs

@@ -6,8 +6,20 @@
     html_favicon_url = "https://bevyengine.org/assets/icon.png"
 )]
 
+#[cfg(feature = "meshlet")]
+mod meshlet;
 pub mod wireframe;
 
+/// Experimental features that are not yet finished. Please report any issues you encounter!
+///
+/// Expect bugs, missing features, compatibility issues, low performance, and/or future breaking changes.
+#[cfg(feature = "meshlet")]
+pub mod experimental {
+    pub mod meshlet {
+        pub use crate::meshlet::*;
+    }
+}
+
 mod bundle;
 pub mod deferred;
 mod extended_material;
@@ -111,6 +123,8 @@ pub const PBR_PREPASS_FUNCTIONS_SHADER_HANDLE: Handle<Shader> =
 pub const PBR_DEFERRED_TYPES_HANDLE: Handle<Shader> = Handle::weak_from_u128(3221241127431430599);
 pub const PBR_DEFERRED_FUNCTIONS_HANDLE: Handle<Shader> = Handle::weak_from_u128(72019026415438599);
 pub const RGB9E5_FUNCTIONS_HANDLE: Handle<Shader> = Handle::weak_from_u128(2659010996143919192);
+const MESHLET_VISIBILITY_BUFFER_RESOLVE_SHADER_HANDLE: Handle<Shader> =
+    Handle::weak_from_u128(2325134235233421);
 
 /// Sets up the entire PBR infrastructure of bevy.
 pub struct PbrPlugin {
@@ -236,6 +250,13 @@ impl Plugin for PbrPlugin {
             "render/view_transformations.wgsl",
             Shader::from_wgsl
         );
+        // Setup dummy shaders for when MeshletPlugin is not used to prevent shader import errors.
+        load_internal_asset!(
+            app,
+            MESHLET_VISIBILITY_BUFFER_RESOLVE_SHADER_HANDLE,
+            "meshlet/dummy_visibility_buffer_resolve.wgsl",
+            Shader::from_wgsl
+        );
 
         app.register_asset_reflect::<StandardMaterial>()
             .register_type::<AmbientLight>()

crates/bevy_pbr/src/material.rs

@@ -1,3 +1,8 @@
+#[cfg(feature = "meshlet")]
+use crate::meshlet::{
+    prepare_material_meshlet_meshes_main_opaque_pass, queue_material_meshlet_meshes,
+    MeshletGpuScene,
+};
 use crate::*;
 use bevy_asset::{Asset, AssetEvent, AssetId, AssetServer};
 use bevy_core_pipeline::{
@@ -170,6 +175,36 @@ pub trait Material: Asset + AsBindGroup + Clone + Sized {
         ShaderRef::Default
     }
 
+    /// Returns this material's [`crate::meshlet::MeshletMesh`] fragment shader. If [`ShaderRef::Default`] is returned,
+    /// the default meshlet mesh fragment shader will be used.
+    ///
+    /// This is part of an experimental feature, and is unnecessary to implement unless you are using `MeshletMesh`'s.
+    #[allow(unused_variables)]
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_fragment_shader() -> ShaderRef {
+        ShaderRef::Default
+    }
+
+    /// Returns this material's [`crate::meshlet::MeshletMesh`] prepass fragment shader. If [`ShaderRef::Default`] is returned,
+    /// the default meshlet mesh prepass fragment shader will be used.
+    ///
+    /// This is part of an experimental feature, and is unnecessary to implement unless you are using `MeshletMesh`'s.
+    #[allow(unused_variables)]
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_prepass_fragment_shader() -> ShaderRef {
+        ShaderRef::Default
+    }
+
+    /// Returns this material's [`crate::meshlet::MeshletMesh`] deferred fragment shader. If [`ShaderRef::Default`] is returned,
+    /// the default meshlet mesh deferred fragment shader will be used.
+    ///
+    /// This is part of an experimental feature, and is unnecessary to implement unless you are using `MeshletMesh`'s.
+    #[allow(unused_variables)]
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_deferred_fragment_shader() -> ShaderRef {
+        ShaderRef::Default
+    }
+
     /// Customizes the default [`RenderPipelineDescriptor`] for a specific entity using the entity's
     /// [`MaterialPipelineKey`] and [`MeshVertexBufferLayoutRef`] as input.
     #[allow(unused_variables)]
@@ -248,6 +283,18 @@ where
                         .after(prepare_materials::<M>),),
                 );
             }
+
+            #[cfg(feature = "meshlet")]
+            render_app.add_systems(
+                Render,
+                (
+                    prepare_material_meshlet_meshes_main_opaque_pass::<M>,
+                    queue_material_meshlet_meshes::<M>,
+                )
+                    .chain()
+                    .in_set(RenderSet::Queue)
+                    .run_if(resource_exists::<MeshletGpuScene>),
+            );
         }
 
         if self.shadows_enabled || self.prepass_enabled {

crates/bevy_pbr/src/meshlet/asset.rs

@@ -0,0 +1,102 @@
+use bevy_asset::{
+    io::{Reader, Writer},
+    saver::{AssetSaver, SavedAsset},
+    Asset, AssetLoader, AsyncReadExt, AsyncWriteExt, LoadContext,
+};
+use bevy_math::Vec3;
+use bevy_reflect::TypePath;
+use bytemuck::{Pod, Zeroable};
+use serde::{Deserialize, Serialize};
+use std::sync::Arc;
+
+/// A mesh that has been pre-processed into multiple small clusters of triangles called meshlets.
+///
+/// A [`bevy_render::mesh::Mesh`] can be converted to a [`MeshletMesh`] using `MeshletMesh::from_mesh` when the `meshlet_processor` cargo feature is enabled.
+/// The conversion step is very slow, and is meant to be ran once ahead of time, and not during runtime. This type of mesh is not suitable for
+/// dynamically generated geometry.
+///
+/// There are restrictions on the [`crate::Material`] functionality that can be used with this type of mesh.
+/// * Materials have no control over the vertex shader or vertex attributes.
+/// * Materials must be opaque. Transparent, alpha masked, and transmissive materials are not supported.
+/// * Materials must use the [`crate::Material::meshlet_mesh_fragment_shader`] method (and similar variants for prepass/deferred shaders)
+///   which requires certain shader patterns that differ from the regular material shaders.
+/// * Limited control over [`bevy_render::render_resource::RenderPipelineDescriptor`] attributes.
+///
+/// See also [`super::MaterialMeshletMeshBundle`] and [`super::MeshletPlugin`].
+#[derive(Asset, TypePath, Serialize, Deserialize, Clone)]
+pub struct MeshletMesh {
+    /// The total amount of triangles summed across all meshlets in the mesh.
+    pub total_meshlet_triangles: u64,
+    /// Raw vertex data bytes for the overall mesh.
+    pub vertex_data: Arc<[u8]>,
+    /// Indices into `vertex_data`.
+    pub vertex_ids: Arc<[u32]>,
+    /// Indices into `vertex_ids`.
+    pub indices: Arc<[u8]>,
+    /// The list of meshlets making up this mesh.
+    pub meshlets: Arc<[Meshlet]>,
+    /// A list of spherical bounding volumes, 1 per meshlet.
+    pub meshlet_bounding_spheres: Arc<[MeshletBoundingSphere]>,
+}
+
+/// A single meshlet within a [`MeshletMesh`].
+#[derive(Serialize, Deserialize, Copy, Clone, Pod, Zeroable)]
+#[repr(C)]
+pub struct Meshlet {
+    /// The offset within the parent mesh's [`MeshletMesh::vertex_ids`] buffer where the indices for this meshlet begin.
+    pub start_vertex_id: u32,
+    /// The offset within the parent mesh's [`MeshletMesh::indices`] buffer where the indices for this meshlet begin.
+    pub start_index_id: u32,
+    /// The amount of triangles in this meshlet.
+    pub triangle_count: u32,
+}
+
+/// A spherical bounding volume used for culling a [`Meshlet`].
+#[derive(Serialize, Deserialize, Copy, Clone, Pod, Zeroable)]
+#[repr(C)]
+pub struct MeshletBoundingSphere {
+    pub center: Vec3,
+    pub radius: f32,
+}
+
+/// An [`AssetLoader`] and [`AssetSaver`] for `.meshlet_mesh` [`MeshletMesh`] assets.
+pub struct MeshletMeshSaverLoad;
+
+impl AssetLoader for MeshletMeshSaverLoad {
+    type Asset = MeshletMesh;
+    type Settings = ();
+    type Error = bincode::Error;
+
+    async fn load<'a>(
+        &'a self,
+        reader: &'a mut Reader<'_>,
+        _settings: &'a Self::Settings,
+        _load_context: &'a mut LoadContext<'_>,
+    ) -> Result<Self::Asset, Self::Error> {
+        let mut bytes = Vec::new();
+        reader.read_to_end(&mut bytes).await?;
+        bincode::deserialize(&bytes)
+    }
+
+    fn extensions(&self) -> &[&str] {
+        &["meshlet_mesh"]
+    }
+}
+
+impl AssetSaver for MeshletMeshSaverLoad {
+    type Asset = MeshletMesh;
+    type Settings = ();
+    type OutputLoader = Self;
+    type Error = bincode::Error;
+
+    async fn save<'a>(
+        &'a self,
+        writer: &'a mut Writer,
+        asset: SavedAsset<'a, Self::Asset>,
+        _settings: &'a Self::Settings,
+    ) -> Result<(), Self::Error> {
+        let bytes = bincode::serialize(asset.get())?;
+        writer.write_all(&bytes).await?;
+        Ok(())
+    }
+}

crates/bevy_pbr/src/meshlet/copy_material_depth.wgsl

@@ -0,0 +1,10 @@
+#import bevy_core_pipeline::fullscreen_vertex_shader::FullscreenVertexOutput
+
+@group(0) @binding(0) var material_depth: texture_2d<u32>;
+
+/// This pass copies the R16Uint material depth texture to an actual Depth16Unorm depth texture.
+
+@fragment
+fn copy_material_depth(in: FullscreenVertexOutput) -> @builtin(frag_depth) f32 {
+    return f32(textureLoad(material_depth, vec2<i32>(in.position.xy), 0).r) / 65535.0;
+}

crates/bevy_pbr/src/meshlet/cull_meshlets.wgsl

@@ -0,0 +1,118 @@
+#import bevy_pbr::meshlet_bindings::{
+    meshlet_thread_meshlet_ids,
+    meshlet_bounding_spheres,
+    meshlet_thread_instance_ids,
+    meshlet_instance_uniforms,
+    meshlet_occlusion,
+    view,
+    should_cull_instance,
+    get_meshlet_previous_occlusion,
+}
+#ifdef MESHLET_SECOND_CULLING_PASS
+#import bevy_pbr::meshlet_bindings::depth_pyramid
+#endif
+#import bevy_render::maths::affine3_to_square
+
+/// Culls individual clusters (1 per thread) in two passes (two pass occlusion culling), and outputs a bitmask of which clusters survived.
+/// 1. The first pass is only frustum culling, on only the clusters that were visible last frame.
+/// 2. The second pass performs both frustum and occlusion culling (using the depth buffer generated from the first pass), on all clusters.
+
+@compute
+@workgroup_size(128, 1, 1) // 128 threads per workgroup, 1 instanced meshlet per thread
+fn cull_meshlets(@builtin(global_invocation_id) cluster_id: vec3<u32>) {
+    // Fetch the instanced meshlet data
+    if cluster_id.x >= arrayLength(&meshlet_thread_meshlet_ids) { return; }
+    let instance_id = meshlet_thread_instance_ids[cluster_id.x];
+    if should_cull_instance(instance_id) {
+        return;
+    }
+    let meshlet_id = meshlet_thread_meshlet_ids[cluster_id.x];
+    let bounding_sphere = meshlet_bounding_spheres[meshlet_id];
+    let instance_uniform = meshlet_instance_uniforms[instance_id];
+    let model = affine3_to_square(instance_uniform.model);
+    let model_scale = max(length(model[0]), max(length(model[1]), length(model[2])));
+    let bounding_sphere_center = model * vec4(bounding_sphere.center, 1.0);
+    let bounding_sphere_radius = model_scale * bounding_sphere.radius;
+
+    // In the first pass, operate only on the clusters visible last frame. In the second pass, operate on all clusters.
+#ifdef MESHLET_SECOND_CULLING_PASS
+    var meshlet_visible = true;
+#else
+    var meshlet_visible = get_meshlet_previous_occlusion(cluster_id.x);
+    if !meshlet_visible { return; }
+#endif
+
+    // Frustum culling
+    // TODO: Faster method from https://vkguide.dev/docs/gpudriven/compute_culling/#frustum-culling-function
+    for (var i = 0u; i < 6u; i++) {
+        if !meshlet_visible { break; }
+        meshlet_visible &= dot(view.frustum[i], bounding_sphere_center) > -bounding_sphere_radius;
+    }
+
+#ifdef MESHLET_SECOND_CULLING_PASS
+    // In the second culling pass, cull against the depth pyramid generated from the first pass
+    if meshlet_visible {
+        let bounding_sphere_center_view_space = (view.inverse_view * vec4(bounding_sphere_center.xyz, 1.0)).xyz;
+        let aabb = project_view_space_sphere_to_screen_space_aabb(bounding_sphere_center_view_space, bounding_sphere_radius);
+
+        // Halve the AABB size because the first depth mip resampling pass cut the full screen resolution into a power of two conservatively
+        let depth_pyramid_size_mip_0 = vec2<f32>(textureDimensions(depth_pyramid, 0)) * 0.5;
+        let width = (aabb.z - aabb.x) * depth_pyramid_size_mip_0.x;
+        let height = (aabb.w - aabb.y) * depth_pyramid_size_mip_0.y;
+        let depth_level = max(0, i32(ceil(log2(max(width, height))))); // TODO: Naga doesn't like this being a u32
+        let depth_pyramid_size = vec2<f32>(textureDimensions(depth_pyramid, depth_level));
+        let aabb_top_left = vec2<u32>(aabb.xy * depth_pyramid_size);
+
+        let depth_quad_a = textureLoad(depth_pyramid, aabb_top_left, depth_level).x;
+        let depth_quad_b = textureLoad(depth_pyramid, aabb_top_left + vec2(1u, 0u), depth_level).x;
+        let depth_quad_c = textureLoad(depth_pyramid, aabb_top_left + vec2(0u, 1u), depth_level).x;
+        let depth_quad_d = textureLoad(depth_pyramid, aabb_top_left + vec2(1u, 1u), depth_level).x;
+
+        let occluder_depth = min(min(depth_quad_a, depth_quad_b), min(depth_quad_c, depth_quad_d));
+        if view.projection[3][3] == 1.0 {
+            // Orthographic
+            let sphere_depth = view.projection[3][2] + (bounding_sphere_center_view_space.z + bounding_sphere_radius) * view.projection[2][2];
+            meshlet_visible &= sphere_depth >= occluder_depth;
+        } else {
+            // Perspective
+            let sphere_depth = -view.projection[3][2] / (bounding_sphere_center_view_space.z + bounding_sphere_radius);
+            meshlet_visible &= sphere_depth >= occluder_depth;
+        }
+    }
+#endif
+
+    // Write the bitmask of whether or not the cluster was culled
+    let occlusion_bit = u32(meshlet_visible) << (cluster_id.x % 32u);
+    atomicOr(&meshlet_occlusion[cluster_id.x / 32u], occlusion_bit);
+}
+
+// https://zeux.io/2023/01/12/approximate-projected-bounds
+fn project_view_space_sphere_to_screen_space_aabb(cp: vec3<f32>, r: f32) -> vec4<f32> {
+    let inv_width = view.projection[0][0] * 0.5;
+    let inv_height = view.projection[1][1] * 0.5;
+    if view.projection[3][3] == 1.0 {
+        // Orthographic
+        let min_x = cp.x - r;
+        let max_x = cp.x + r;
+
+        let min_y = cp.y - r;
+        let max_y = cp.y + r;
+
+        return vec4(min_x * inv_width, 1.0 - max_y * inv_height, max_x * inv_width, 1.0 - min_y * inv_height);
+    } else {
+        // Perspective
+        let c = vec3(cp.xy, -cp.z);
+        let cr = c * r;
+        let czr2 = c.z * c.z - r * r;
+
+        let vx = sqrt(c.x * c.x + czr2);
+        let min_x = (vx * c.x - cr.z) / (vx * c.z + cr.x);
+        let max_x = (vx * c.x + cr.z) / (vx * c.z - cr.x);
+
+        let vy = sqrt(c.y * c.y + czr2);
+        let min_y = (vy * c.y - cr.z) / (vy * c.z + cr.y);
+        let max_y = (vy * c.y + cr.z) / (vy * c.z - cr.y);
+
+        return vec4(min_x * inv_width, -max_y * inv_height, max_x * inv_width, -min_y * inv_height) + vec4(0.5);
+    }
+}

crates/bevy_pbr/src/meshlet/downsample_depth.wgsl

@@ -0,0 +1,16 @@
+#import bevy_core_pipeline::fullscreen_vertex_shader::FullscreenVertexOutput
+
+@group(0) @binding(0) var input_depth: texture_2d<f32>;
+@group(0) @binding(1) var samplr: sampler;
+
+/// Performs a 2x2 downsample on a depth texture to generate the next mip level of a hierarchical depth buffer.
+
+@fragment
+fn downsample_depth(in: FullscreenVertexOutput) -> @location(0) vec4<f32> {
+    let depth_quad = textureGather(0, input_depth, samplr, in.uv);
+    let downsampled_depth = min(
+        min(depth_quad.x, depth_quad.y),
+        min(depth_quad.z, depth_quad.w),
+    );
+    return vec4(downsampled_depth, 0.0, 0.0, 0.0);
+}

crates/bevy_pbr/src/meshlet/dummy_visibility_buffer_resolve.wgsl

@@ -0,0 +1,4 @@
+#define_import_path bevy_pbr::meshlet_visibility_buffer_resolve
+
+/// Dummy shader to prevent naga_oil from complaining about missing imports when the MeshletPlugin is not loaded,
+/// as naga_oil tries to resolve imports even if they're behind an #ifdef.

crates/bevy_pbr/src/meshlet/from_mesh.rs

@@ -0,0 +1,98 @@
+use super::asset::{Meshlet, MeshletBoundingSphere, MeshletMesh};
+use bevy_render::{
+    mesh::{Indices, Mesh},
+    render_resource::PrimitiveTopology,
+};
+use meshopt::{build_meshlets, compute_meshlet_bounds_decoder, VertexDataAdapter};
+use std::borrow::Cow;
+
+impl MeshletMesh {
+    /// Process a [`Mesh`] to generate a [`MeshletMesh`].
+    ///
+    /// This process is very slow, and should be done ahead of time, and not at runtime.
+    ///
+    /// This function requires the `meshlet_processor` cargo feature.
+    ///
+    /// The input mesh must:
+    /// 1. Use [`PrimitiveTopology::TriangleList`]
+    /// 2. Use indices
+    /// 3. Have the exact following set of vertex attributes: `{POSITION, NORMAL, UV_0, TANGENT}`
+    pub fn from_mesh(mesh: &Mesh) -> Result<Self, MeshToMeshletMeshConversionError> {
+        // Validate mesh format
+        if mesh.primitive_topology() != PrimitiveTopology::TriangleList {
+            return Err(MeshToMeshletMeshConversionError::WrongMeshPrimitiveTopology);
+        }
+        if mesh.attributes().map(|(id, _)| id).ne([
+            Mesh::ATTRIBUTE_POSITION.id,
+            Mesh::ATTRIBUTE_NORMAL.id,
+            Mesh::ATTRIBUTE_UV_0.id,
+            Mesh::ATTRIBUTE_TANGENT.id,
+        ]) {
+            return Err(MeshToMeshletMeshConversionError::WrongMeshVertexAttributes);
+        }
+        let indices = match mesh.indices() {
+            Some(Indices::U32(indices)) => Cow::Borrowed(indices.as_slice()),
+            Some(Indices::U16(indices)) => indices.iter().map(|i| *i as u32).collect(),
+            _ => return Err(MeshToMeshletMeshConversionError::MeshMissingIndices),
+        };
+        let vertex_buffer = mesh.get_vertex_buffer_data();
+        let vertices =
+            VertexDataAdapter::new(&vertex_buffer, mesh.get_vertex_size() as usize, 0).unwrap();
+
+        // Split the mesh into meshlets
+        let meshopt_meshlets = build_meshlets(&indices, &vertices, 64, 64, 0.0);
+
+        // Calculate meshlet bounding spheres
+        let meshlet_bounding_spheres = meshopt_meshlets
+            .iter()
+            .map(|meshlet| {
+                compute_meshlet_bounds_decoder(
+                    meshlet,
+                    mesh.attribute(Mesh::ATTRIBUTE_POSITION)
+                        .unwrap()
+                        .as_float3()
+                        .unwrap(),
+                )
+            })
+            .map(|bounds| MeshletBoundingSphere {
+                center: bounds.center.into(),
+                radius: bounds.radius,
+            })
+            .collect();
+
+        // Assemble into the final asset
+        let mut total_meshlet_triangles = 0;
+        let meshlets = meshopt_meshlets
+            .meshlets
+            .into_iter()
+            .map(|m| {
+                total_meshlet_triangles += m.triangle_count as u64;
+                Meshlet {
+                    start_vertex_id: m.vertex_offset,
+                    start_index_id: m.triangle_offset,
+                    triangle_count: m.triangle_count,
+                }
+            })
+            .collect();
+
+        Ok(Self {
+            total_meshlet_triangles,
+            vertex_data: vertex_buffer.into(),
+            vertex_ids: meshopt_meshlets.vertices.into(),
+            indices: meshopt_meshlets.triangles.into(),
+            meshlets,
+            meshlet_bounding_spheres,
+        })
+    }
+}
+
+/// An error produced by [`MeshletMesh::from_mesh`].
+#[derive(thiserror::Error, Debug)]
+pub enum MeshToMeshletMeshConversionError {
+    #[error("Mesh primitive topology was not TriangleList")]
+    WrongMeshPrimitiveTopology,
+    #[error("Mesh attributes were not {{POSITION, NORMAL, UV_0, TANGENT}}")]
+    WrongMeshVertexAttributes,
+    #[error("Mesh had no indices")]
+    MeshMissingIndices,
+}

crates/bevy_pbr/src/meshlet/gpu_scene.rs

@@ -0,0 +1,977 @@
+use super::{persistent_buffer::PersistentGpuBuffer, Meshlet, MeshletBoundingSphere, MeshletMesh};
+use crate::{
+    Material, MeshFlags, MeshTransforms, MeshUniform, NotShadowCaster, NotShadowReceiver,
+    PreviousGlobalTransform, RenderMaterialInstances, ShadowView,
+};
+use bevy_asset::{AssetEvent, AssetId, AssetServer, Assets, Handle, UntypedAssetId};
+use bevy_core_pipeline::core_3d::Camera3d;
+use bevy_ecs::{
+    component::Component,
+    entity::{Entity, EntityHashMap},
+    event::EventReader,
+    query::{AnyOf, Has},
+    system::{Commands, Query, Res, ResMut, Resource, SystemState},
+    world::{FromWorld, World},
+};
+use bevy_render::{
+    render_resource::{binding_types::*, *},
+    renderer::{RenderDevice, RenderQueue},
+    texture::{CachedTexture, TextureCache},
+    view::{ExtractedView, RenderLayers, ViewDepthTexture, ViewUniform, ViewUniforms},
+    MainWorld,
+};
+use bevy_transform::components::GlobalTransform;
+use bevy_utils::{default, HashMap, HashSet};
+use encase::internal::WriteInto;
+use std::{
+    iter,
+    mem::size_of,
+    ops::{DerefMut, Range},
+    sync::Arc,
+};
+
+/// Create and queue for uploading to the GPU [`MeshUniform`] components for
+/// [`MeshletMesh`] entities, as well as queuing uploads for any new meshlet mesh
+/// assets that have not already been uploaded to the GPU.
+pub fn extract_meshlet_meshes(
+    // TODO: Replace main_world when Extract<ResMut<Assets<MeshletMesh>>> is possible
+    mut main_world: ResMut<MainWorld>,
+    mut gpu_scene: ResMut<MeshletGpuScene>,
+) {
+    let mut system_state: SystemState<(
+        Query<(
+            Entity,
+            &Handle<MeshletMesh>,
+            &GlobalTransform,
+            Option<&PreviousGlobalTransform>,
+            Option<&RenderLayers>,
+            Has<NotShadowReceiver>,
+            Has<NotShadowCaster>,
+        )>,
+        Res<AssetServer>,
+        ResMut<Assets<MeshletMesh>>,
+        EventReader<AssetEvent<MeshletMesh>>,
+    )> = SystemState::new(&mut main_world);
+    let (instances_query, asset_server, mut assets, mut asset_events) =
+        system_state.get_mut(&mut main_world);
+
+    // Reset all temporary data for MeshletGpuScene
+    gpu_scene.reset();
+
+    // Free GPU buffer space for any modified or dropped MeshletMesh assets
+    for asset_event in asset_events.read() {
+        if let AssetEvent::Unused { id } | AssetEvent::Modified { id } = asset_event {
+            if let Some((
+                [vertex_data_slice, vertex_ids_slice, indices_slice, meshlets_slice, meshlet_bounding_spheres_slice],
+                _,
+            )) = gpu_scene.meshlet_mesh_slices.remove(id)
+            {
+                gpu_scene.vertex_data.mark_slice_unused(vertex_data_slice);
+                gpu_scene.vertex_ids.mark_slice_unused(vertex_ids_slice);
+                gpu_scene.indices.mark_slice_unused(indices_slice);
+                gpu_scene.meshlets.mark_slice_unused(meshlets_slice);
+                gpu_scene
+                    .meshlet_bounding_spheres
+                    .mark_slice_unused(meshlet_bounding_spheres_slice);
+            }
+        }
+    }
+
+    for (
+        instance_index,
+        (
+            instance,
+            handle,
+            transform,
+            previous_transform,
+            render_layers,
+            not_shadow_receiver,
+            not_shadow_caster,
+        ),
+    ) in instances_query.iter().enumerate()
+    {
+        // Skip instances with an unloaded MeshletMesh asset
+        if asset_server.is_managed(handle.id())
+            && !asset_server.is_loaded_with_dependencies(handle.id())
+        {
+            continue;
+        }
+
+        // Upload the instance's MeshletMesh asset data, if not done already, along with other per-frame per-instance data.
+        gpu_scene.queue_meshlet_mesh_upload(
+            instance,
+            render_layers.cloned().unwrap_or(default()),
+            not_shadow_caster,
+            handle,
+            &mut assets,
+            instance_index as u32,
+        );
+
+        // Build a MeshUniform for each instance
+        let transform = transform.affine();
+        let previous_transform = previous_transform.map(|t| t.0).unwrap_or(transform);
+        let mut flags = if not_shadow_receiver {
+            MeshFlags::empty()
+        } else {
+            MeshFlags::SHADOW_RECEIVER
+        };
+        if transform.matrix3.determinant().is_sign_positive() {
+            flags |= MeshFlags::SIGN_DETERMINANT_MODEL_3X3;
+        }
+        let transforms = MeshTransforms {
+            transform: (&transform).into(),
+            previous_transform: (&previous_transform).into(),
+            flags: flags.bits(),
+        };
+        gpu_scene
+            .instance_uniforms
+            .get_mut()
+            .push(MeshUniform::new(&transforms, None));
+    }
+}
+
+/// Upload all newly queued [`MeshletMesh`] asset data from [`extract_meshlet_meshes`] to the GPU.
+pub fn perform_pending_meshlet_mesh_writes(
+    mut gpu_scene: ResMut<MeshletGpuScene>,
+    render_queue: Res<RenderQueue>,
+    render_device: Res<RenderDevice>,
+) {
+    gpu_scene
+        .vertex_data
+        .perform_writes(&render_queue, &render_device);
+    gpu_scene
+        .vertex_ids
+        .perform_writes(&render_queue, &render_device);
+    gpu_scene
+        .indices
+        .perform_writes(&render_queue, &render_device);
+    gpu_scene
+        .meshlets
+        .perform_writes(&render_queue, &render_device);
+    gpu_scene
+        .meshlet_bounding_spheres
+        .perform_writes(&render_queue, &render_device);
+}
+
+/// For each entity in the scene, record what material ID (for use with depth testing during the meshlet mesh material draw nodes)
+/// its material was assigned in the `prepare_material_meshlet_meshes` systems, and note that the material is used by at least one entity in the scene.
+pub fn queue_material_meshlet_meshes<M: Material>(
+    mut gpu_scene: ResMut<MeshletGpuScene>,
+    render_material_instances: Res<RenderMaterialInstances<M>>,
+) {
+    // TODO: Ideally we could parallelize this system, both between different materials, and the loop over instances
+    let gpu_scene = gpu_scene.deref_mut();
+
+    for (i, (instance, _, _)) in gpu_scene.instances.iter().enumerate() {
+        if let Some(material_asset_id) = render_material_instances.get(instance) {
+            let material_asset_id = material_asset_id.untyped();
+            if let Some(material_id) = gpu_scene.material_id_lookup.get(&material_asset_id) {
+                gpu_scene.material_ids_present_in_scene.insert(*material_id);
+                gpu_scene.instance_material_ids.get_mut()[i] = *material_id;
+            }
+        }
+    }
+}
+
+// TODO: Try using Queue::write_buffer_with() in queue_meshlet_mesh_upload() to reduce copies
+fn upload_storage_buffer<T: ShaderSize + bytemuck::Pod>(
+    buffer: &mut StorageBuffer<Vec<T>>,
+    render_device: &RenderDevice,
+    render_queue: &RenderQueue,
+) where
+    Vec<T>: WriteInto,
+{
+    let inner = buffer.buffer();
+    let capacity = inner.map_or(0, |b| b.size());
+    let size = buffer.get().size().get() as BufferAddress;
+
+    if capacity >= size {
+        let inner = inner.unwrap();
+        let bytes = bytemuck::cast_slice(buffer.get().as_slice());
+        render_queue.write_buffer(inner, 0, bytes);
+    } else {
+        buffer.write_buffer(render_device, render_queue);
+    }
+}
+
+pub fn prepare_meshlet_per_frame_resources(
+    mut gpu_scene: ResMut<MeshletGpuScene>,
+    views: Query<(
+        Entity,
+        &ExtractedView,
+        Option<&RenderLayers>,
+        AnyOf<(&Camera3d, &ShadowView)>,
+    )>,
+    mut texture_cache: ResMut<TextureCache>,
+    render_queue: Res<RenderQueue>,
+    render_device: Res<RenderDevice>,
+    mut commands: Commands,
+) {
+    gpu_scene
+        .previous_cluster_id_starts
+        .retain(|_, (_, active)| *active);
+
+    if gpu_scene.scene_meshlet_count == 0 {
+        return;
+    }
+
+    let gpu_scene = gpu_scene.as_mut();
+
+    gpu_scene
+        .instance_uniforms
+        .write_buffer(&render_device, &render_queue);
+    upload_storage_buffer(
+        &mut gpu_scene.instance_material_ids,
+        &render_device,
+        &render_queue,
+    );
+    upload_storage_buffer(
+        &mut gpu_scene.thread_instance_ids,
+        &render_device,
+        &render_queue,
+    );
+    upload_storage_buffer(
+        &mut gpu_scene.thread_meshlet_ids,
+        &render_device,
+        &render_queue,
+    );
+    upload_storage_buffer(
+        &mut gpu_scene.previous_cluster_ids,
+        &render_device,
+        &render_queue,
+    );
+
+    let needed_buffer_size = 4 * gpu_scene.scene_triangle_count;
+    let visibility_buffer_draw_index_buffer =
+        match &mut gpu_scene.visibility_buffer_draw_index_buffer {
+            Some(buffer) if buffer.size() >= needed_buffer_size => buffer.clone(),
+            slot => {
+                let buffer = render_device.create_buffer(&BufferDescriptor {
+                    label: Some("meshlet_visibility_buffer_draw_index_buffer"),
+                    size: needed_buffer_size,
+                    usage: BufferUsages::STORAGE | BufferUsages::INDEX,
+                    mapped_at_creation: false,
+                });
+                *slot = Some(buffer.clone());
+                buffer
+            }
+        };
+
+    let needed_buffer_size = gpu_scene.scene_meshlet_count.div_ceil(32) as u64 * 4;
+    for (view_entity, view, render_layers, (_, shadow_view)) in &views {
+        let instance_visibility = gpu_scene
+            .view_instance_visibility
+            .entry(view_entity)
+            .or_insert_with(|| {
+                let mut buffer = StorageBuffer::default();
+                buffer.set_label(Some("meshlet_view_instance_visibility"));
+                buffer
+            });
+        for (instance_index, (_, layers, not_shadow_caster)) in
+            gpu_scene.instances.iter().enumerate()
+        {
+            // If either the layers don't match the view's layers or this is a shadow view
+            // and the instance is not a shadow caster, hide the instance for this view
+            if !render_layers.unwrap_or(&default()).intersects(layers)
+                || (shadow_view.is_some() && *not_shadow_caster)
+            {
+                let vec = instance_visibility.get_mut();
+                let index = instance_index / 32;
+                let bit = instance_index - index * 32;
+                if vec.len() <= index {
+                    vec.extend(iter::repeat(0).take(index - vec.len() + 1));
+                }
+                vec[index] |= 1 << bit;
+            }
+        }
+        upload_storage_buffer(instance_visibility, &render_device, &render_queue);
+        let instance_visibility = instance_visibility.buffer().unwrap().clone();
+
+        // Early submission for GPU data uploads to start while the render graph records commands
+        render_queue.submit([]);
+
+        let create_occlusion_buffer = || {
+            render_device.create_buffer(&BufferDescriptor {
+                label: Some("meshlet_occlusion_buffer"),
+                size: needed_buffer_size,
+                usage: BufferUsages::STORAGE | BufferUsages::COPY_DST,
+                mapped_at_creation: false,
+            })
+        };
+        let (previous_occlusion_buffer, occlusion_buffer, occlusion_buffer_needs_clearing) =
+            match gpu_scene.previous_occlusion_buffers.get(&view_entity) {
+                Some((buffer_a, buffer_b)) if buffer_b.size() >= needed_buffer_size => {
+                    (buffer_a.clone(), buffer_b.clone(), true)
+                }
+                Some((buffer_a, _)) => (buffer_a.clone(), create_occlusion_buffer(), false),
+                None => (create_occlusion_buffer(), create_occlusion_buffer(), false),
+            };
+        gpu_scene.previous_occlusion_buffers.insert(
+            view_entity,
+            (occlusion_buffer.clone(), previous_occlusion_buffer.clone()),
+        );
+
+        let visibility_buffer = TextureDescriptor {
+            label: Some("meshlet_visibility_buffer"),
+            size: Extent3d {
+                width: view.viewport.z,
+                height: view.viewport.w,
+                depth_or_array_layers: 1,
+            },
+            mip_level_count: 1,
+            sample_count: 1,
+            dimension: TextureDimension::D2,
+            format: TextureFormat::R32Uint,
+            usage: TextureUsages::RENDER_ATTACHMENT | TextureUsages::TEXTURE_BINDING,
+            view_formats: &[],
+        };
+
+        let visibility_buffer_draw_indirect_args_first =
+            render_device.create_buffer_with_data(&BufferInitDescriptor {
+                label: Some("meshlet_visibility_buffer_draw_indirect_args_first"),
+                contents: DrawIndirectArgs {
+                    vertex_count: 0,
+                    instance_count: 1,
+                    first_vertex: 0,
+                    first_instance: 0,
+                }
+                .as_bytes(),
+                usage: BufferUsages::STORAGE | BufferUsages::INDIRECT,
+            });
+        let visibility_buffer_draw_indirect_args_second =
+            render_device.create_buffer_with_data(&BufferInitDescriptor {
+                label: Some("meshlet_visibility_buffer_draw_indirect_args_second"),
+                contents: DrawIndirectArgs {
+                    vertex_count: 0,
+                    instance_count: 1,
+                    first_vertex: 0,
+                    first_instance: 0,
+                }
+                .as_bytes(),
+                usage: BufferUsages::STORAGE | BufferUsages::INDIRECT,
+            });
+
+        let depth_size = Extent3d {
+            // If not a power of 2, round down to the nearest power of 2 to ensure depth is conservative
+            width: previous_power_of_2(view.viewport.z),
+            height: previous_power_of_2(view.viewport.w),
+            depth_or_array_layers: 1,
+        };
+        let depth_mip_count = depth_size.width.max(depth_size.height).ilog2() + 1;
+        let depth_pyramid = texture_cache.get(
+            &render_device,
+            TextureDescriptor {
+                label: Some("meshlet_depth_pyramid"),
+                size: depth_size,
+                mip_level_count: depth_mip_count,
+                sample_count: 1,
+                dimension: TextureDimension::D2,
+                format: TextureFormat::R32Float,
+                usage: TextureUsages::RENDER_ATTACHMENT | TextureUsages::TEXTURE_BINDING,
+                view_formats: &[],
+            },
+        );
+        let depth_pyramid_mips = (0..depth_mip_count)
+            .map(|i| {
+                depth_pyramid.texture.create_view(&TextureViewDescriptor {
+                    label: Some("meshlet_depth_pyramid_texture_view"),
+                    format: Some(TextureFormat::R32Float),
+                    dimension: Some(TextureViewDimension::D2),
+                    aspect: TextureAspect::All,
+                    base_mip_level: i,
+                    mip_level_count: Some(1),
+                    base_array_layer: 0,
+                    array_layer_count: None,
+                })
+            })
+            .collect();
+
+        let material_depth_color = TextureDescriptor {
+            label: Some("meshlet_material_depth_color"),
+            size: Extent3d {
+                width: view.viewport.z,
+                height: view.viewport.w,
+                depth_or_array_layers: 1,
+            },
+            mip_level_count: 1,
+            sample_count: 1,
+            dimension: TextureDimension::D2,
+            format: TextureFormat::R16Uint,
+            usage: TextureUsages::RENDER_ATTACHMENT | TextureUsages::TEXTURE_BINDING,
+            view_formats: &[],
+        };
+
+        let material_depth = TextureDescriptor {
+            label: Some("meshlet_material_depth"),
+            size: Extent3d {
+                width: view.viewport.z,
+                height: view.viewport.w,
+                depth_or_array_layers: 1,
+            },
+            mip_level_count: 1,
+            sample_count: 1,
+            dimension: TextureDimension::D2,
+            format: TextureFormat::Depth16Unorm,
+            usage: TextureUsages::RENDER_ATTACHMENT,
+            view_formats: &[],
+        };
+
+        let not_shadow_view = shadow_view.is_none();
+        commands.entity(view_entity).insert(MeshletViewResources {
+            scene_meshlet_count: gpu_scene.scene_meshlet_count,
+            previous_occlusion_buffer,
+            occlusion_buffer,
+            occlusion_buffer_needs_clearing,
+            instance_visibility,
+            visibility_buffer: not_shadow_view
+                .then(|| texture_cache.get(&render_device, visibility_buffer)),
+            visibility_buffer_draw_indirect_args_first,
+            visibility_buffer_draw_indirect_args_second,
+            visibility_buffer_draw_index_buffer: visibility_buffer_draw_index_buffer.clone(),
+            depth_pyramid,
+            depth_pyramid_mips,
+            material_depth_color: not_shadow_view
+                .then(|| texture_cache.get(&render_device, material_depth_color)),
+            material_depth: not_shadow_view
+                .then(|| texture_cache.get(&render_device, material_depth)),
+        });
+    }
+}
+
+pub fn prepare_meshlet_view_bind_groups(
+    gpu_scene: Res<MeshletGpuScene>,
+    views: Query<(
+        Entity,
+        &MeshletViewResources,
+        AnyOf<(&ViewDepthTexture, &ShadowView)>,
+    )>,
+    view_uniforms: Res<ViewUniforms>,
+    render_device: Res<RenderDevice>,
+    mut commands: Commands,
+) {
+    let Some(view_uniforms) = view_uniforms.uniforms.binding() else {
+        return;
+    };
+
+    for (view_entity, view_resources, view_depth) in &views {
+        let entries = BindGroupEntries::sequential((
+            gpu_scene.thread_meshlet_ids.binding().unwrap(),
+            gpu_scene.meshlet_bounding_spheres.binding(),
+            gpu_scene.thread_instance_ids.binding().unwrap(),
+            gpu_scene.instance_uniforms.binding().unwrap(),
+            gpu_scene.view_instance_visibility[&view_entity]
+                .binding()
+                .unwrap(),
+            view_resources.occlusion_buffer.as_entire_binding(),
+            gpu_scene.previous_cluster_ids.binding().unwrap(),
+            view_resources.previous_occlusion_buffer.as_entire_binding(),
+            view_uniforms.clone(),
+            &view_resources.depth_pyramid.default_view,
+        ));
+        let culling = render_device.create_bind_group(
+            "meshlet_culling_bind_group",
+            &gpu_scene.culling_bind_group_layout,
+            &entries,
+        );
+
+        let entries = BindGroupEntries::sequential((
+            view_resources.occlusion_buffer.as_entire_binding(),
+            gpu_scene.thread_meshlet_ids.binding().unwrap(),
+            gpu_scene.previous_cluster_ids.binding().unwrap(),
+            view_resources.previous_occlusion_buffer.as_entire_binding(),
+            gpu_scene.meshlets.binding(),
+            view_resources
+                .visibility_buffer_draw_indirect_args_first
+                .as_entire_binding(),
+            view_resources
+                .visibility_buffer_draw_index_buffer
+                .as_entire_binding(),
+        ));
+        let write_index_buffer_first = render_device.create_bind_group(
+            "meshlet_write_index_buffer_first_bind_group",
+            &gpu_scene.write_index_buffer_bind_group_layout,
+            &entries,
+        );
+
+        let entries = BindGroupEntries::sequential((
+            view_resources.occlusion_buffer.as_entire_binding(),
+            gpu_scene.thread_meshlet_ids.binding().unwrap(),
+            gpu_scene.previous_cluster_ids.binding().unwrap(),
+            view_resources.previous_occlusion_buffer.as_entire_binding(),
+            gpu_scene.meshlets.binding(),
+            view_resources
+                .visibility_buffer_draw_indirect_args_second
+                .as_entire_binding(),
+            view_resources
+                .visibility_buffer_draw_index_buffer
+                .as_entire_binding(),
+        ));
+        let write_index_buffer_second = render_device.create_bind_group(
+            "meshlet_write_index_buffer_second_bind_group",
+            &gpu_scene.write_index_buffer_bind_group_layout,
+            &entries,
+        );
+
+        let view_depth_texture = match view_depth {
+            (Some(view_depth), None) => view_depth.view(),
+            (None, Some(shadow_view)) => &shadow_view.depth_attachment.view,
+            _ => unreachable!(),
+        };
+        let downsample_depth = (0..view_resources.depth_pyramid_mips.len())
+            .map(|i| {
+                render_device.create_bind_group(
+                    "meshlet_downsample_depth_bind_group",
+                    &gpu_scene.downsample_depth_bind_group_layout,
+                    &BindGroupEntries::sequential((
+                        if i == 0 {
+                            view_depth_texture
+                        } else {
+                            &view_resources.depth_pyramid_mips[i - 1]
+                        },
+                        &gpu_scene.depth_pyramid_sampler,
+                    )),
+                )
+            })
+            .collect();
+
+        let entries = BindGroupEntries::sequential((
+            gpu_scene.thread_meshlet_ids.binding().unwrap(),
+            gpu_scene.meshlets.binding(),
+            gpu_scene.indices.binding(),
+            gpu_scene.vertex_ids.binding(),
+            gpu_scene.vertex_data.binding(),
+            gpu_scene.thread_instance_ids.binding().unwrap(),
+            gpu_scene.instance_uniforms.binding().unwrap(),
+            gpu_scene.instance_material_ids.binding().unwrap(),
+            view_resources
+                .visibility_buffer_draw_index_buffer
+                .as_entire_binding(),
+            view_uniforms.clone(),
+        ));
+        let visibility_buffer_raster = render_device.create_bind_group(
+            "meshlet_visibility_raster_buffer_bind_group",
+            &gpu_scene.visibility_buffer_raster_bind_group_layout,
+            &entries,
+        );
+
+        let copy_material_depth =
+            view_resources
+                .material_depth_color
+                .as_ref()
+                .map(|material_depth_color| {
+                    render_device.create_bind_group(
+                        "meshlet_copy_material_depth_bind_group",
+                        &gpu_scene.copy_material_depth_bind_group_layout,
+                        &[BindGroupEntry {
+                            binding: 0,
+                            resource: BindingResource::TextureView(
+                                &material_depth_color.default_view,
+                            ),
+                        }],
+                    )
+                });
+
+        let material_draw = view_resources
+            .visibility_buffer
+            .as_ref()
+            .map(|visibility_buffer| {
+                let entries = BindGroupEntries::sequential((
+                    &visibility_buffer.default_view,
+                    gpu_scene.thread_meshlet_ids.binding().unwrap(),
+                    gpu_scene.meshlets.binding(),
+                    gpu_scene.indices.binding(),
+                    gpu_scene.vertex_ids.binding(),
+                    gpu_scene.vertex_data.binding(),
+                    gpu_scene.thread_instance_ids.binding().unwrap(),
+                    gpu_scene.instance_uniforms.binding().unwrap(),
+                ));
+                render_device.create_bind_group(
+                    "meshlet_mesh_material_draw_bind_group",
+                    &gpu_scene.material_draw_bind_group_layout,
+                    &entries,
+                )
+            });
+
+        commands.entity(view_entity).insert(MeshletViewBindGroups {
+            culling,
+            write_index_buffer_first,
+            write_index_buffer_second,
+            downsample_depth,
+            visibility_buffer_raster,
+            copy_material_depth,
+            material_draw,
+        });
+    }
+}
+
+/// A resource that manages GPU data for rendering [`MeshletMesh`]'s.
+#[derive(Resource)]
+pub struct MeshletGpuScene {
+    vertex_data: PersistentGpuBuffer<Arc<[u8]>>,
+    vertex_ids: PersistentGpuBuffer<Arc<[u32]>>,
+    indices: PersistentGpuBuffer<Arc<[u8]>>,
+    meshlets: PersistentGpuBuffer<Arc<[Meshlet]>>,
+    meshlet_bounding_spheres: PersistentGpuBuffer<Arc<[MeshletBoundingSphere]>>,
+    meshlet_mesh_slices: HashMap<AssetId<MeshletMesh>, ([Range<BufferAddress>; 5], u64)>,
+
+    scene_meshlet_count: u32,
+    scene_triangle_count: u64,
+    next_material_id: u32,
+    material_id_lookup: HashMap<UntypedAssetId, u32>,
+    material_ids_present_in_scene: HashSet<u32>,
+    /// Per-instance Entity, RenderLayers, and NotShadowCaster
+    instances: Vec<(Entity, RenderLayers, bool)>,
+    /// Per-instance transforms, model matrices, and render flags
+    instance_uniforms: StorageBuffer<Vec<MeshUniform>>,
+    /// Per-view per-instance visibility bit. Used for RenderLayer and NotShadowCaster support.
+    view_instance_visibility: EntityHashMap<StorageBuffer<Vec<u32>>>,
+    instance_material_ids: StorageBuffer<Vec<u32>>,
+    thread_instance_ids: StorageBuffer<Vec<u32>>,
+    thread_meshlet_ids: StorageBuffer<Vec<u32>>,
+    previous_cluster_ids: StorageBuffer<Vec<u32>>,
+    previous_cluster_id_starts: HashMap<(Entity, AssetId<MeshletMesh>), (u32, bool)>,
+    previous_occlusion_buffers: EntityHashMap<(Buffer, Buffer)>,
+    visibility_buffer_draw_index_buffer: Option<Buffer>,
+
+    culling_bind_group_layout: BindGroupLayout,
+    write_index_buffer_bind_group_layout: BindGroupLayout,
+    visibility_buffer_raster_bind_group_layout: BindGroupLayout,
+    downsample_depth_bind_group_layout: BindGroupLayout,
+    copy_material_depth_bind_group_layout: BindGroupLayout,
+    material_draw_bind_group_layout: BindGroupLayout,
+    depth_pyramid_sampler: Sampler,
+}
+
+impl FromWorld for MeshletGpuScene {
+    fn from_world(world: &mut World) -> Self {
+        let render_device = world.resource::<RenderDevice>();
+
+        Self {
+            vertex_data: PersistentGpuBuffer::new("meshlet_vertex_data", render_device),
+            vertex_ids: PersistentGpuBuffer::new("meshlet_vertex_ids", render_device),
+            indices: PersistentGpuBuffer::new("meshlet_indices", render_device),
+            meshlets: PersistentGpuBuffer::new("meshlets", render_device),
+            meshlet_bounding_spheres: PersistentGpuBuffer::new(
+                "meshlet_bounding_spheres",
+                render_device,
+            ),
+            meshlet_mesh_slices: HashMap::new(),
+
+            scene_meshlet_count: 0,
+            scene_triangle_count: 0,
+            next_material_id: 0,
+            material_id_lookup: HashMap::new(),
+            material_ids_present_in_scene: HashSet::new(),
+            instances: Vec::new(),
+            instance_uniforms: {
+                let mut buffer = StorageBuffer::default();
+                buffer.set_label(Some("meshlet_instance_uniforms"));
+                buffer
+            },
+            view_instance_visibility: EntityHashMap::default(),
+            instance_material_ids: {
+                let mut buffer = StorageBuffer::default();
+                buffer.set_label(Some("meshlet_instance_material_ids"));
+                buffer
+            },
+            thread_instance_ids: {
+                let mut buffer = StorageBuffer::default();
+                buffer.set_label(Some("meshlet_thread_instance_ids"));
+                buffer
+            },
+            thread_meshlet_ids: {
+                let mut buffer = StorageBuffer::default();
+                buffer.set_label(Some("meshlet_thread_meshlet_ids"));
+                buffer
+            },
+            previous_cluster_ids: {
+                let mut buffer = StorageBuffer::default();
+                buffer.set_label(Some("meshlet_previous_cluster_ids"));
+                buffer
+            },
+            previous_cluster_id_starts: HashMap::new(),
+            previous_occlusion_buffers: EntityHashMap::default(),
+            visibility_buffer_draw_index_buffer: None,
+
+            // TODO: Buffer min sizes
+            culling_bind_group_layout: render_device.create_bind_group_layout(
+                "meshlet_culling_bind_group_layout",
+                &BindGroupLayoutEntries::sequential(
+                    ShaderStages::COMPUTE,
+                    (
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        uniform_buffer::<ViewUniform>(true),
+                        texture_2d(TextureSampleType::Float { filterable: false }),
+                    ),
+                ),
+            ),
+            write_index_buffer_bind_group_layout: render_device.create_bind_group_layout(
+                "meshlet_write_index_buffer_bind_group_layout",
+                &BindGroupLayoutEntries::sequential(
+                    ShaderStages::COMPUTE,
+                    (
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_sized(false, None),
+                        storage_buffer_sized(false, None),
+                    ),
+                ),
+            ),
+            downsample_depth_bind_group_layout: render_device.create_bind_group_layout(
+                "meshlet_downsample_depth_bind_group_layout",
+                &BindGroupLayoutEntries::sequential(
+                    ShaderStages::FRAGMENT,
+                    (
+                        texture_2d(TextureSampleType::Float { filterable: false }),
+                        sampler(SamplerBindingType::NonFiltering),
+                    ),
+                ),
+            ),
+            visibility_buffer_raster_bind_group_layout: render_device.create_bind_group_layout(
+                "meshlet_visibility_buffer_raster_bind_group_layout",
+                &BindGroupLayoutEntries::sequential(
+                    ShaderStages::VERTEX,
+                    (
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        uniform_buffer::<ViewUniform>(true),
+                    ),
+                ),
+            ),
+            copy_material_depth_bind_group_layout: render_device.create_bind_group_layout(
+                "meshlet_copy_material_depth_bind_group_layout",
+                &BindGroupLayoutEntries::single(
+                    ShaderStages::FRAGMENT,
+                    texture_2d(TextureSampleType::Uint),
+                ),
+            ),
+            material_draw_bind_group_layout: render_device.create_bind_group_layout(
+                "meshlet_mesh_material_draw_bind_group_layout",
+                &BindGroupLayoutEntries::sequential(
+                    ShaderStages::FRAGMENT,
+                    (
+                        texture_2d(TextureSampleType::Uint),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                        storage_buffer_read_only_sized(false, None),
+                    ),
+                ),
+            ),
+            depth_pyramid_sampler: render_device.create_sampler(&SamplerDescriptor {
+                label: Some("meshlet_depth_pyramid_sampler"),
+                ..default()
+            }),
+        }
+    }
+}
+
+impl MeshletGpuScene {
+    /// Clear per-frame CPU->GPU upload buffers and reset all per-frame data.
+    fn reset(&mut self) {
+        // TODO: Shrink capacity if saturation is low
+        self.scene_meshlet_count = 0;
+        self.scene_triangle_count = 0;
+        self.next_material_id = 0;
+        self.material_id_lookup.clear();
+        self.material_ids_present_in_scene.clear();
+        self.instances.clear();
+        self.view_instance_visibility
+            .values_mut()
+            .for_each(|b| b.get_mut().clear());
+        self.instance_uniforms.get_mut().clear();
+        self.instance_material_ids.get_mut().clear();
+        self.thread_instance_ids.get_mut().clear();
+        self.thread_meshlet_ids.get_mut().clear();
+        self.previous_cluster_ids.get_mut().clear();
+        self.previous_cluster_id_starts
+            .values_mut()
+            .for_each(|(_, active)| *active = false);
+        // TODO: Remove unused entries for previous_occlusion_buffers
+    }
+
+    fn queue_meshlet_mesh_upload(
+        &mut self,
+        instance: Entity,
+        render_layers: RenderLayers,
+        not_shadow_caster: bool,
+        handle: &Handle<MeshletMesh>,
+        assets: &mut Assets<MeshletMesh>,
+        instance_index: u32,
+    ) {
+        let queue_meshlet_mesh = |asset_id: &AssetId<MeshletMesh>| {
+            let meshlet_mesh = assets.remove_untracked(*asset_id).expect(
+                "MeshletMesh asset was already unloaded but is not registered with MeshletGpuScene",
+            );
+
+            let vertex_data_slice = self
+                .vertex_data
+                .queue_write(Arc::clone(&meshlet_mesh.vertex_data), ());
+            let vertex_ids_slice = self.vertex_ids.queue_write(
+                Arc::clone(&meshlet_mesh.vertex_ids),
+                vertex_data_slice.start,
+            );
+            let indices_slice = self
+                .indices
+                .queue_write(Arc::clone(&meshlet_mesh.indices), ());
+            let meshlets_slice = self.meshlets.queue_write(
+                Arc::clone(&meshlet_mesh.meshlets),
+                (vertex_ids_slice.start, indices_slice.start),
+            );
+            let meshlet_bounding_spheres_slice = self
+                .meshlet_bounding_spheres
+                .queue_write(Arc::clone(&meshlet_mesh.meshlet_bounding_spheres), ());
+
+            (
+                [
+                    vertex_data_slice,
+                    vertex_ids_slice,
+                    indices_slice,
+                    meshlets_slice,
+                    meshlet_bounding_spheres_slice,
+                ],
+                meshlet_mesh.total_meshlet_triangles,
+            )
+        };
+
+        // Append instance data for this frame
+        self.instances
+            .push((instance, render_layers, not_shadow_caster));
+        self.instance_material_ids.get_mut().push(0);
+
+        // If the MeshletMesh asset has not been uploaded to the GPU yet, queue it for uploading
+        let ([_, _, _, meshlets_slice, _], triangle_count) = self
+            .meshlet_mesh_slices
+            .entry(handle.id())
+            .or_insert_with_key(queue_meshlet_mesh)
+            .clone();
+
+        let meshlets_slice = (meshlets_slice.start as u32 / size_of::<Meshlet>() as u32)
+            ..(meshlets_slice.end as u32 / size_of::<Meshlet>() as u32);
+
+        let current_cluster_id_start = self.scene_meshlet_count;
+
+        self.scene_meshlet_count += meshlets_slice.end - meshlets_slice.start;
+        self.scene_triangle_count += triangle_count;
+
+        // Calculate the previous cluster IDs for each meshlet for this instance
+        let previous_cluster_id_start = self
+            .previous_cluster_id_starts
+            .entry((instance, handle.id()))
+            .or_insert((0, true));
+        let previous_cluster_ids = if previous_cluster_id_start.1 {
+            0..(meshlets_slice.len() as u32)
+        } else {
+            let start = previous_cluster_id_start.0;
+            start..(meshlets_slice.len() as u32 + start)
+        };
+
+        // Append per-cluster data for this frame
+        self.thread_instance_ids
+            .get_mut()
+            .extend(std::iter::repeat(instance_index).take(meshlets_slice.len()));
+        self.thread_meshlet_ids.get_mut().extend(meshlets_slice);
+        self.previous_cluster_ids
+            .get_mut()
+            .extend(previous_cluster_ids);
+
+        *previous_cluster_id_start = (current_cluster_id_start, true);
+    }
+
+    /// Get the depth value for use with the material depth texture for a given [`Material`] asset.
+    pub fn get_material_id(&mut self, material_id: UntypedAssetId) -> u32 {
+        *self
+            .material_id_lookup
+            .entry(material_id)
+            .or_insert_with(|| {
+                self.next_material_id += 1;
+                self.next_material_id
+            })
+    }
+
+    pub fn material_present_in_scene(&self, material_id: &u32) -> bool {
+        self.material_ids_present_in_scene.contains(material_id)
+    }
+
+    pub fn culling_bind_group_layout(&self) -> BindGroupLayout {
+        self.culling_bind_group_layout.clone()
+    }
+
+    pub fn write_index_buffer_bind_group_layout(&self) -> BindGroupLayout {
+        self.write_index_buffer_bind_group_layout.clone()
+    }
+
+    pub fn downsample_depth_bind_group_layout(&self) -> BindGroupLayout {
+        self.downsample_depth_bind_group_layout.clone()
+    }
+
+    pub fn visibility_buffer_raster_bind_group_layout(&self) -> BindGroupLayout {
+        self.visibility_buffer_raster_bind_group_layout.clone()
+    }
+
+    pub fn copy_material_depth_bind_group_layout(&self) -> BindGroupLayout {
+        self.copy_material_depth_bind_group_layout.clone()
+    }
+
+    pub fn material_draw_bind_group_layout(&self) -> BindGroupLayout {
+        self.material_draw_bind_group_layout.clone()
+    }
+}
+
+#[derive(Component)]
+pub struct MeshletViewResources {
+    pub scene_meshlet_count: u32,
+    previous_occlusion_buffer: Buffer,
+    pub occlusion_buffer: Buffer,
+    pub occlusion_buffer_needs_clearing: bool,
+    pub instance_visibility: Buffer,
+    pub visibility_buffer: Option<CachedTexture>,
+    pub visibility_buffer_draw_indirect_args_first: Buffer,
+    pub visibility_buffer_draw_indirect_args_second: Buffer,
+    visibility_buffer_draw_index_buffer: Buffer,
+    pub depth_pyramid: CachedTexture,
+    pub depth_pyramid_mips: Box<[TextureView]>,
+    pub material_depth_color: Option<CachedTexture>,
+    pub material_depth: Option<CachedTexture>,
+}
+
+#[derive(Component)]
+pub struct MeshletViewBindGroups {
+    pub culling: BindGroup,
+    pub write_index_buffer_first: BindGroup,
+    pub write_index_buffer_second: BindGroup,
+    pub downsample_depth: Box<[BindGroup]>,
+    pub visibility_buffer_raster: BindGroup,
+    pub copy_material_depth: Option<BindGroup>,
+    pub material_draw: Option<BindGroup>,
+}
+
+fn previous_power_of_2(x: u32) -> u32 {
+    // If x is a power of 2, halve it
+    if x.count_ones() == 1 {
+        x / 2
+    } else {
+        // Else calculate the largest power of 2 that is less than x
+        1 << (31 - x.leading_zeros())
+    }
+}

crates/bevy_pbr/src/meshlet/material_draw_nodes.rs

@@ -0,0 +1,379 @@
+use super::{
+    gpu_scene::{MeshletViewBindGroups, MeshletViewResources},
+    material_draw_prepare::{
+        MeshletViewMaterialsDeferredGBufferPrepass, MeshletViewMaterialsMainOpaquePass,
+        MeshletViewMaterialsPrepass,
+    },
+    MeshletGpuScene,
+};
+use crate::{
+    MeshViewBindGroup, PrepassViewBindGroup, PreviousViewProjectionUniformOffset,
+    ViewFogUniformOffset, ViewLightProbesUniformOffset, ViewLightsUniformOffset,
+};
+use bevy_core_pipeline::prepass::ViewPrepassTextures;
+use bevy_ecs::{query::QueryItem, world::World};
+use bevy_render::{
+    camera::ExtractedCamera,
+    render_graph::{NodeRunError, RenderGraphContext, ViewNode},
+    render_resource::{
+        LoadOp, Operations, PipelineCache, RenderPassDepthStencilAttachment, RenderPassDescriptor,
+        StoreOp,
+    },
+    renderer::RenderContext,
+    view::{ViewTarget, ViewUniformOffset},
+};
+
+/// Fullscreen shading pass based on the visibility buffer generated from rasterizing meshlets.
+#[derive(Default)]
+pub struct MeshletMainOpaquePass3dNode;
+impl ViewNode for MeshletMainOpaquePass3dNode {
+    type ViewQuery = (
+        &'static ExtractedCamera,
+        &'static ViewTarget,
+        &'static MeshViewBindGroup,
+        &'static ViewUniformOffset,
+        &'static ViewLightsUniformOffset,
+        &'static ViewFogUniformOffset,
+        &'static ViewLightProbesUniformOffset,
+        &'static MeshletViewMaterialsMainOpaquePass,
+        &'static MeshletViewBindGroups,
+        &'static MeshletViewResources,
+    );
+
+    fn run(
+        &self,
+        _graph: &mut RenderGraphContext,
+        render_context: &mut RenderContext,
+        (
+            camera,
+            target,
+            mesh_view_bind_group,
+            view_uniform_offset,
+            view_lights_offset,
+            view_fog_offset,
+            view_light_probes_offset,
+            meshlet_view_materials,
+            meshlet_view_bind_groups,
+            meshlet_view_resources,
+        ): QueryItem<Self::ViewQuery>,
+        world: &World,
+    ) -> Result<(), NodeRunError> {
+        if meshlet_view_materials.is_empty() {
+            return Ok(());
+        }
+
+        let (
+            Some(meshlet_gpu_scene),
+            Some(pipeline_cache),
+            Some(meshlet_material_depth),
+            Some(meshlet_material_draw_bind_group),
+        ) = (
+            world.get_resource::<MeshletGpuScene>(),
+            world.get_resource::<PipelineCache>(),
+            meshlet_view_resources.material_depth.as_ref(),
+            meshlet_view_bind_groups.material_draw.as_ref(),
+        )
+        else {
+            return Ok(());
+        };
+
+        let mut render_pass = render_context.begin_tracked_render_pass(RenderPassDescriptor {
+            label: Some("meshlet_main_opaque_pass_3d"),
+            color_attachments: &[Some(target.get_color_attachment())],
+            depth_stencil_attachment: Some(RenderPassDepthStencilAttachment {
+                view: &meshlet_material_depth.default_view,
+                depth_ops: Some(Operations {
+                    load: LoadOp::Load,
+                    store: StoreOp::Store,
+                }),
+                stencil_ops: None,
+            }),
+            timestamp_writes: None,
+            occlusion_query_set: None,
+        });
+        if let Some(viewport) = camera.viewport.as_ref() {
+            render_pass.set_camera_viewport(viewport);
+        }
+
+        render_pass.set_bind_group(
+            0,
+            &mesh_view_bind_group.value,
+            &[
+                view_uniform_offset.offset,
+                view_lights_offset.offset,
+                view_fog_offset.offset,
+                **view_light_probes_offset,
+            ],
+        );
+        render_pass.set_bind_group(1, meshlet_material_draw_bind_group, &[]);
+
+        // 1 fullscreen triangle draw per material
+        for (material_id, material_pipeline_id, material_bind_group) in
+            meshlet_view_materials.iter()
+        {
+            if meshlet_gpu_scene.material_present_in_scene(material_id) {
+                if let Some(material_pipeline) =
+                    pipeline_cache.get_render_pipeline(*material_pipeline_id)
+                {
+                    let x = *material_id * 3;
+                    render_pass.set_bind_group(2, material_bind_group, &[]);
+                    render_pass.set_render_pipeline(material_pipeline);
+                    render_pass.draw(x..(x + 3), 0..1);
+                }
+            }
+        }
+
+        Ok(())
+    }
+}
+
+/// Fullscreen pass to generate prepass textures based on the visibility buffer generated from rasterizing meshlets.
+#[derive(Default)]
+pub struct MeshletPrepassNode;
+impl ViewNode for MeshletPrepassNode {
+    type ViewQuery = (
+        &'static ExtractedCamera,
+        &'static ViewPrepassTextures,
+        &'static ViewUniformOffset,
+        Option<&'static PreviousViewProjectionUniformOffset>,
+        &'static MeshletViewMaterialsPrepass,
+        &'static MeshletViewBindGroups,
+        &'static MeshletViewResources,
+    );
+
+    fn run(
+        &self,
+        _graph: &mut RenderGraphContext,
+        render_context: &mut RenderContext,
+        (
+            camera,
+            view_prepass_textures,
+            view_uniform_offset,
+            previous_view_projection_uniform_offset,
+            meshlet_view_materials,
+            meshlet_view_bind_groups,
+            meshlet_view_resources,
+        ): QueryItem<Self::ViewQuery>,
+        world: &World,
+    ) -> Result<(), NodeRunError> {
+        if meshlet_view_materials.is_empty() {
+            return Ok(());
+        }
+
+        let (
+            Some(prepass_view_bind_group),
+            Some(meshlet_gpu_scene),
+            Some(pipeline_cache),
+            Some(meshlet_material_depth),
+            Some(meshlet_material_draw_bind_group),
+        ) = (
+            world.get_resource::<PrepassViewBindGroup>(),
+            world.get_resource::<MeshletGpuScene>(),
+            world.get_resource::<PipelineCache>(),
+            meshlet_view_resources.material_depth.as_ref(),
+            meshlet_view_bind_groups.material_draw.as_ref(),
+        )
+        else {
+            return Ok(());
+        };
+
+        let color_attachments = vec![
+            view_prepass_textures
+                .normal
+                .as_ref()
+                .map(|normals_texture| normals_texture.get_attachment()),
+            view_prepass_textures
+                .motion_vectors
+                .as_ref()
+                .map(|motion_vectors_texture| motion_vectors_texture.get_attachment()),
+            // Use None in place of Deferred attachments
+            None,
+            None,
+        ];
+
+        let mut render_pass = render_context.begin_tracked_render_pass(RenderPassDescriptor {
+            label: Some("meshlet_prepass"),
+            color_attachments: &color_attachments,
+            depth_stencil_attachment: Some(RenderPassDepthStencilAttachment {
+                view: &meshlet_material_depth.default_view,
+                depth_ops: Some(Operations {
+                    load: LoadOp::Load,
+                    store: StoreOp::Store,
+                }),
+                stencil_ops: None,
+            }),
+            timestamp_writes: None,
+            occlusion_query_set: None,
+        });
+        if let Some(viewport) = camera.viewport.as_ref() {
+            render_pass.set_camera_viewport(viewport);
+        }
+
+        if let Some(previous_view_projection_uniform_offset) =
+            previous_view_projection_uniform_offset
+        {
+            render_pass.set_bind_group(
+                0,
+                prepass_view_bind_group.motion_vectors.as_ref().unwrap(),
+                &[
+                    view_uniform_offset.offset,
+                    previous_view_projection_uniform_offset.offset,
+                ],
+            );
+        } else {
+            render_pass.set_bind_group(
+                0,
+                prepass_view_bind_group.no_motion_vectors.as_ref().unwrap(),
+                &[view_uniform_offset.offset],
+            );
+        }
+
+        render_pass.set_bind_group(1, meshlet_material_draw_bind_group, &[]);
+
+        // 1 fullscreen triangle draw per material
+        for (material_id, material_pipeline_id, material_bind_group) in
+            meshlet_view_materials.iter()
+        {
+            if meshlet_gpu_scene.material_present_in_scene(material_id) {
+                if let Some(material_pipeline) =
+                    pipeline_cache.get_render_pipeline(*material_pipeline_id)
+                {
+                    let x = *material_id * 3;
+                    render_pass.set_bind_group(2, material_bind_group, &[]);
+                    render_pass.set_render_pipeline(material_pipeline);
+                    render_pass.draw(x..(x + 3), 0..1);
+                }
+            }
+        }
+
+        Ok(())
+    }
+}
+
+/// Fullscreen pass to generate a gbuffer based on the visibility buffer generated from rasterizing meshlets.
+#[derive(Default)]
+pub struct MeshletDeferredGBufferPrepassNode;
+impl ViewNode for MeshletDeferredGBufferPrepassNode {
+    type ViewQuery = (
+        &'static ExtractedCamera,
+        &'static ViewPrepassTextures,
+        &'static ViewUniformOffset,
+        Option<&'static PreviousViewProjectionUniformOffset>,
+        &'static MeshletViewMaterialsDeferredGBufferPrepass,
+        &'static MeshletViewBindGroups,
+        &'static MeshletViewResources,
+    );
+
+    fn run(
+        &self,
+        _graph: &mut RenderGraphContext,
+        render_context: &mut RenderContext,
+        (
+            camera,
+            view_prepass_textures,
+            view_uniform_offset,
+            previous_view_projection_uniform_offset,
+            meshlet_view_materials,
+            meshlet_view_bind_groups,
+            meshlet_view_resources,
+        ): QueryItem<Self::ViewQuery>,
+        world: &World,
+    ) -> Result<(), NodeRunError> {
+        if meshlet_view_materials.is_empty() {
+            return Ok(());
+        }
+
+        let (
+            Some(prepass_view_bind_group),
+            Some(meshlet_gpu_scene),
+            Some(pipeline_cache),
+            Some(meshlet_material_depth),
+            Some(meshlet_material_draw_bind_group),
+        ) = (
+            world.get_resource::<PrepassViewBindGroup>(),
+            world.get_resource::<MeshletGpuScene>(),
+            world.get_resource::<PipelineCache>(),
+            meshlet_view_resources.material_depth.as_ref(),
+            meshlet_view_bind_groups.material_draw.as_ref(),
+        )
+        else {
+            return Ok(());
+        };
+
+        let color_attachments = vec![
+            view_prepass_textures
+                .normal
+                .as_ref()
+                .map(|normals_texture| normals_texture.get_attachment()),
+            view_prepass_textures
+                .motion_vectors
+                .as_ref()
+                .map(|motion_vectors_texture| motion_vectors_texture.get_attachment()),
+            view_prepass_textures
+                .deferred
+                .as_ref()
+                .map(|deferred_texture| deferred_texture.get_attachment()),
+            view_prepass_textures
+                .deferred_lighting_pass_id
+                .as_ref()
+                .map(|deferred_lighting_pass_id| deferred_lighting_pass_id.get_attachment()),
+        ];
+
+        let mut render_pass = render_context.begin_tracked_render_pass(RenderPassDescriptor {
+            label: Some("meshlet_deferred_prepass"),
+            color_attachments: &color_attachments,
+            depth_stencil_attachment: Some(RenderPassDepthStencilAttachment {
+                view: &meshlet_material_depth.default_view,
+                depth_ops: Some(Operations {
+                    load: LoadOp::Load,
+                    store: StoreOp::Store,
+                }),
+                stencil_ops: None,
+            }),
+            timestamp_writes: None,
+            occlusion_query_set: None,
+        });
+        if let Some(viewport) = camera.viewport.as_ref() {
+            render_pass.set_camera_viewport(viewport);
+        }
+
+        if let Some(previous_view_projection_uniform_offset) =
+            previous_view_projection_uniform_offset
+        {
+            render_pass.set_bind_group(
+                0,
+                prepass_view_bind_group.motion_vectors.as_ref().unwrap(),
+                &[
+                    view_uniform_offset.offset,
+                    previous_view_projection_uniform_offset.offset,
+                ],
+            );
+        } else {
+            render_pass.set_bind_group(
+                0,
+                prepass_view_bind_group.no_motion_vectors.as_ref().unwrap(),
+                &[view_uniform_offset.offset],
+            );
+        }
+
+        render_pass.set_bind_group(1, meshlet_material_draw_bind_group, &[]);
+
+        // 1 fullscreen triangle draw per material
+        for (material_id, material_pipeline_id, material_bind_group) in
+            meshlet_view_materials.iter()
+        {
+            if meshlet_gpu_scene.material_present_in_scene(material_id) {
+                if let Some(material_pipeline) =
+                    pipeline_cache.get_render_pipeline(*material_pipeline_id)
+                {
+                    let x = *material_id * 3;
+                    render_pass.set_bind_group(2, material_bind_group, &[]);
+                    render_pass.set_render_pipeline(material_pipeline);
+                    render_pass.draw(x..(x + 3), 0..1);
+                }
+            }
+        }
+
+        Ok(())
+    }
+}

crates/bevy_pbr/src/meshlet/material_draw_prepare.rs

@@ -0,0 +1,405 @@
+use super::{MeshletGpuScene, MESHLET_MESH_MATERIAL_SHADER_HANDLE};
+use crate::{environment_map::EnvironmentMapLight, irradiance_volume::IrradianceVolume, *};
+use bevy_asset::AssetServer;
+use bevy_core_pipeline::{
+    core_3d::Camera3d,
+    prepass::{DeferredPrepass, DepthPrepass, MotionVectorPrepass, NormalPrepass},
+    tonemapping::{DebandDither, Tonemapping},
+};
+use bevy_derive::{Deref, DerefMut};
+use bevy_render::{
+    camera::TemporalJitter,
+    mesh::{Mesh, MeshVertexBufferLayout, MeshVertexBufferLayoutRef, MeshVertexBufferLayouts},
+    render_resource::*,
+    view::ExtractedView,
+};
+use bevy_utils::HashMap;
+use std::hash::Hash;
+
+/// A list of `(Material ID, Pipeline, BindGroup)` for a view for use in [`super::MeshletMainOpaquePass3dNode`].
+#[derive(Component, Deref, DerefMut, Default)]
+pub struct MeshletViewMaterialsMainOpaquePass(pub Vec<(u32, CachedRenderPipelineId, BindGroup)>);
+
+/// Prepare [`Material`] pipelines for [`super::MeshletMesh`] entities for use in [`super::MeshletMainOpaquePass3dNode`],
+/// and register the material with [`MeshletGpuScene`].
+#[allow(clippy::too_many_arguments)]
+pub fn prepare_material_meshlet_meshes_main_opaque_pass<M: Material>(
+    mut gpu_scene: ResMut<MeshletGpuScene>,
+    mut cache: Local<HashMap<MeshPipelineKey, CachedRenderPipelineId>>,
+    pipeline_cache: Res<PipelineCache>,
+    material_pipeline: Res<MaterialPipeline<M>>,
+    mesh_pipeline: Res<MeshPipeline>,
+    render_materials: Res<RenderMaterials<M>>,
+    render_material_instances: Res<RenderMaterialInstances<M>>,
+    asset_server: Res<AssetServer>,
+    mut mesh_vertex_buffer_layouts: ResMut<MeshVertexBufferLayouts>,
+    mut views: Query<
+        (
+            &mut MeshletViewMaterialsMainOpaquePass,
+            &ExtractedView,
+            Option<&Tonemapping>,
+            Option<&DebandDither>,
+            Option<&ShadowFilteringMethod>,
+            Has<ScreenSpaceAmbientOcclusionSettings>,
+            (
+                Has<NormalPrepass>,
+                Has<DepthPrepass>,
+                Has<MotionVectorPrepass>,
+                Has<DeferredPrepass>,
+            ),
+            Has<TemporalJitter>,
+            Option<&Projection>,
+            Has<RenderViewLightProbes<EnvironmentMapLight>>,
+            Has<RenderViewLightProbes<IrradianceVolume>>,
+        ),
+        With<Camera3d>,
+    >,
+) where
+    M::Data: PartialEq + Eq + Hash + Clone,
+{
+    let fake_vertex_buffer_layout = &fake_vertex_buffer_layout(&mut mesh_vertex_buffer_layouts);
+
+    for (
+        mut materials,
+        view,
+        tonemapping,
+        dither,
+        shadow_filter_method,
+        ssao,
+        (normal_prepass, depth_prepass, motion_vector_prepass, deferred_prepass),
+        temporal_jitter,
+        projection,
+        has_environment_maps,
+        has_irradiance_volumes,
+    ) in &mut views
+    {
+        let mut view_key =
+            MeshPipelineKey::from_msaa_samples(1) | MeshPipelineKey::from_hdr(view.hdr);
+
+        if normal_prepass {
+            view_key |= MeshPipelineKey::NORMAL_PREPASS;
+        }
+        if depth_prepass {
+            view_key |= MeshPipelineKey::DEPTH_PREPASS;
+        }
+        if motion_vector_prepass {
+            view_key |= MeshPipelineKey::MOTION_VECTOR_PREPASS;
+        }
+        if deferred_prepass {
+            view_key |= MeshPipelineKey::DEFERRED_PREPASS;
+        }
+
+        if temporal_jitter {
+            view_key |= MeshPipelineKey::TEMPORAL_JITTER;
+        }
+
+        if has_environment_maps {
+            view_key |= MeshPipelineKey::ENVIRONMENT_MAP;
+        }
+
+        if has_irradiance_volumes {
+            view_key |= MeshPipelineKey::IRRADIANCE_VOLUME;
+        }
+
+        if let Some(projection) = projection {
+            view_key |= match projection {
+                Projection::Perspective(_) => MeshPipelineKey::VIEW_PROJECTION_PERSPECTIVE,
+                Projection::Orthographic(_) => MeshPipelineKey::VIEW_PROJECTION_ORTHOGRAPHIC,
+            };
+        }
+
+        match shadow_filter_method.unwrap_or(&ShadowFilteringMethod::default()) {
+            ShadowFilteringMethod::Hardware2x2 => {
+                view_key |= MeshPipelineKey::SHADOW_FILTER_METHOD_HARDWARE_2X2;
+            }
+            ShadowFilteringMethod::Castano13 => {
+                view_key |= MeshPipelineKey::SHADOW_FILTER_METHOD_CASTANO_13;
+            }
+            ShadowFilteringMethod::Jimenez14 => {
+                view_key |= MeshPipelineKey::SHADOW_FILTER_METHOD_JIMENEZ_14;
+            }
+        }
+
+        if !view.hdr {
+            if let Some(tonemapping) = tonemapping {
+                view_key |= MeshPipelineKey::TONEMAP_IN_SHADER;
+                view_key |= tonemapping_pipeline_key(*tonemapping);
+            }
+            if let Some(DebandDither::Enabled) = dither {
+                view_key |= MeshPipelineKey::DEBAND_DITHER;
+            }
+        }
+
+        if ssao {
+            view_key |= MeshPipelineKey::SCREEN_SPACE_AMBIENT_OCCLUSION;
+        }
+
+        // TODO: Lightmaps
+
+        view_key |= MeshPipelineKey::from_primitive_topology(PrimitiveTopology::TriangleList);
+
+        for material_id in render_material_instances.values() {
+            let Some(material) = render_materials.get(material_id) else {
+                continue;
+            };
+
+            if material.properties.alpha_mode != AlphaMode::Opaque
+                || material.properties.reads_view_transmission_texture
+            {
+                continue;
+            }
+
+            let Ok(material_pipeline_descriptor) = material_pipeline.specialize(
+                MaterialPipelineKey {
+                    mesh_key: view_key,
+                    bind_group_data: material.key.clone(),
+                },
+                fake_vertex_buffer_layout,
+            ) else {
+                continue;
+            };
+            let material_fragment = material_pipeline_descriptor.fragment.unwrap();
+
+            let mut shader_defs = material_fragment.shader_defs;
+            shader_defs.push("MESHLET_MESH_MATERIAL_PASS".into());
+
+            let pipeline_descriptor = RenderPipelineDescriptor {
+                label: material_pipeline_descriptor.label,
+                layout: vec![
+                    mesh_pipeline.get_view_layout(view_key.into()).clone(),
+                    gpu_scene.material_draw_bind_group_layout(),
+                    material_pipeline.material_layout.clone(),
+                ],
+                push_constant_ranges: vec![],
+                vertex: VertexState {
+                    shader: MESHLET_MESH_MATERIAL_SHADER_HANDLE,
+                    shader_defs: shader_defs.clone(),
+                    entry_point: material_pipeline_descriptor.vertex.entry_point,
+                    buffers: Vec::new(),
+                },
+                primitive: PrimitiveState::default(),
+                depth_stencil: Some(DepthStencilState {
+                    format: TextureFormat::Depth16Unorm,
+                    depth_write_enabled: false,
+                    depth_compare: CompareFunction::Equal,
+                    stencil: StencilState::default(),
+                    bias: DepthBiasState::default(),
+                }),
+                multisample: MultisampleState::default(),
+                fragment: Some(FragmentState {
+                    shader: match M::meshlet_mesh_fragment_shader() {
+                        ShaderRef::Default => MESHLET_MESH_MATERIAL_SHADER_HANDLE,
+                        ShaderRef::Handle(handle) => handle,
+                        ShaderRef::Path(path) => asset_server.load(path),
+                    },
+                    shader_defs,
+                    entry_point: material_fragment.entry_point,
+                    targets: material_fragment.targets,
+                }),
+            };
+
+            let material_id = gpu_scene.get_material_id(material_id.untyped());
+
+            let pipeline_id = *cache.entry(view_key).or_insert_with(|| {
+                pipeline_cache.queue_render_pipeline(pipeline_descriptor.clone())
+            });
+            materials.push((material_id, pipeline_id, material.bind_group.clone()));
+        }
+    }
+}
+
+/// A list of `(Material ID, Pipeline, BindGroup)` for a view for use in [`super::MeshletPrepassNode`].
+#[derive(Component, Deref, DerefMut, Default)]
+pub struct MeshletViewMaterialsPrepass(pub Vec<(u32, CachedRenderPipelineId, BindGroup)>);
+
+/// A list of `(Material ID, Pipeline, BindGroup)` for a view for use in [`super::MeshletDeferredGBufferPrepassNode`].
+#[derive(Component, Deref, DerefMut, Default)]
+pub struct MeshletViewMaterialsDeferredGBufferPrepass(
+    pub Vec<(u32, CachedRenderPipelineId, BindGroup)>,
+);
+
+/// Prepare [`Material`] pipelines for [`super::MeshletMesh`] entities for use in [`super::MeshletPrepassNode`],
+/// and [`super::MeshletDeferredGBufferPrepassNode`] and register the material with [`MeshletGpuScene`].
+#[allow(clippy::too_many_arguments)]
+pub fn prepare_material_meshlet_meshes_prepass<M: Material>(
+    mut gpu_scene: ResMut<MeshletGpuScene>,
+    mut cache: Local<HashMap<MeshPipelineKey, CachedRenderPipelineId>>,
+    pipeline_cache: Res<PipelineCache>,
+    prepass_pipeline: Res<PrepassPipeline<M>>,
+    render_materials: Res<RenderMaterials<M>>,
+    render_material_instances: Res<RenderMaterialInstances<M>>,
+    mut mesh_vertex_buffer_layouts: ResMut<MeshVertexBufferLayouts>,
+    asset_server: Res<AssetServer>,
+    mut views: Query<
+        (
+            &mut MeshletViewMaterialsPrepass,
+            &mut MeshletViewMaterialsDeferredGBufferPrepass,
+            &ExtractedView,
+            AnyOf<(&NormalPrepass, &MotionVectorPrepass, &DeferredPrepass)>,
+        ),
+        With<Camera3d>,
+    >,
+) where
+    M::Data: PartialEq + Eq + Hash + Clone,
+{
+    let fake_vertex_buffer_layout = &fake_vertex_buffer_layout(&mut mesh_vertex_buffer_layouts);
+
+    for (
+        mut materials,
+        mut deferred_materials,
+        view,
+        (normal_prepass, motion_vector_prepass, deferred_prepass),
+    ) in &mut views
+    {
+        let mut view_key =
+            MeshPipelineKey::from_msaa_samples(1) | MeshPipelineKey::from_hdr(view.hdr);
+
+        if normal_prepass.is_some() {
+            view_key |= MeshPipelineKey::NORMAL_PREPASS;
+        }
+        if motion_vector_prepass.is_some() {
+            view_key |= MeshPipelineKey::MOTION_VECTOR_PREPASS;
+        }
+
+        view_key |= MeshPipelineKey::from_primitive_topology(PrimitiveTopology::TriangleList);
+
+        for material_id in render_material_instances.values() {
+            let Some(material) = render_materials.get(material_id) else {
+                continue;
+            };
+
+            if material.properties.alpha_mode != AlphaMode::Opaque
+                || material.properties.reads_view_transmission_texture
+            {
+                continue;
+            }
+
+            let material_wants_deferred = matches!(
+                material.properties.render_method,
+                OpaqueRendererMethod::Deferred
+            );
+            if deferred_prepass.is_some() && material_wants_deferred {
+                view_key |= MeshPipelineKey::DEFERRED_PREPASS;
+            } else if normal_prepass.is_none() && motion_vector_prepass.is_none() {
+                continue;
+            }
+
+            let Ok(material_pipeline_descriptor) = prepass_pipeline.specialize(
+                MaterialPipelineKey {
+                    mesh_key: view_key,
+                    bind_group_data: material.key.clone(),
+                },
+                fake_vertex_buffer_layout,
+            ) else {
+                continue;
+            };
+            let material_fragment = material_pipeline_descriptor.fragment.unwrap();
+
+            let mut shader_defs = material_fragment.shader_defs;
+            shader_defs.push("MESHLET_MESH_MATERIAL_PASS".into());
+
+            let view_layout = if view_key.contains(MeshPipelineKey::MOTION_VECTOR_PREPASS) {
+                prepass_pipeline.view_layout_motion_vectors.clone()
+            } else {
+                prepass_pipeline.view_layout_no_motion_vectors.clone()
+            };
+
+            let fragment_shader = if view_key.contains(MeshPipelineKey::DEFERRED_PREPASS) {
+                M::meshlet_mesh_deferred_fragment_shader()
+            } else {
+                M::meshlet_mesh_prepass_fragment_shader()
+            };
+
+            let entry_point = match fragment_shader {
+                ShaderRef::Default => "prepass_fragment".into(),
+                _ => material_fragment.entry_point,
+            };
+
+            let pipeline_descriptor = RenderPipelineDescriptor {
+                label: material_pipeline_descriptor.label,
+                layout: vec![
+                    view_layout,
+                    gpu_scene.material_draw_bind_group_layout(),
+                    prepass_pipeline.material_layout.clone(),
+                ],
+                push_constant_ranges: vec![],
+                vertex: VertexState {
+                    shader: MESHLET_MESH_MATERIAL_SHADER_HANDLE,
+                    shader_defs: shader_defs.clone(),
+                    entry_point: material_pipeline_descriptor.vertex.entry_point,
+                    buffers: Vec::new(),
+                },
+                primitive: PrimitiveState::default(),
+                depth_stencil: Some(DepthStencilState {
+                    format: TextureFormat::Depth16Unorm,
+                    depth_write_enabled: false,
+                    depth_compare: CompareFunction::Equal,
+                    stencil: StencilState::default(),
+                    bias: DepthBiasState::default(),
+                }),
+                multisample: MultisampleState::default(),
+                fragment: Some(FragmentState {
+                    shader: match fragment_shader {
+                        ShaderRef::Default => MESHLET_MESH_MATERIAL_SHADER_HANDLE,
+                        ShaderRef::Handle(handle) => handle,
+                        ShaderRef::Path(path) => asset_server.load(path),
+                    },
+                    shader_defs,
+                    entry_point,
+                    targets: material_fragment.targets,
+                }),
+            };
+
+            let material_id = gpu_scene.get_material_id(material_id.untyped());
+
+            let pipeline_id = *cache.entry(view_key).or_insert_with(|| {
+                pipeline_cache.queue_render_pipeline(pipeline_descriptor.clone())
+            });
+
+            let item = (material_id, pipeline_id, material.bind_group.clone());
+            if view_key.contains(MeshPipelineKey::DEFERRED_PREPASS) {
+                deferred_materials.push(item);
+            } else {
+                materials.push(item);
+            }
+        }
+    }
+}
+
+// Meshlet materials don't use a traditional vertex buffer, but the material specialization requires one.
+fn fake_vertex_buffer_layout(layouts: &mut MeshVertexBufferLayouts) -> MeshVertexBufferLayoutRef {
+    layouts.insert(MeshVertexBufferLayout::new(
+        vec![
+            Mesh::ATTRIBUTE_POSITION.id,
+            Mesh::ATTRIBUTE_NORMAL.id,
+            Mesh::ATTRIBUTE_UV_0.id,
+            Mesh::ATTRIBUTE_TANGENT.id,
+        ],
+        VertexBufferLayout {
+            array_stride: 48,
+            step_mode: VertexStepMode::Vertex,
+            attributes: vec![
+                VertexAttribute {
+                    format: Mesh::ATTRIBUTE_POSITION.format,
+                    offset: 0,
+                    shader_location: 0,
+                },
+                VertexAttribute {
+                    format: Mesh::ATTRIBUTE_NORMAL.format,
+                    offset: 12,
+                    shader_location: 1,
+                },
+                VertexAttribute {
+                    format: Mesh::ATTRIBUTE_UV_0.format,
+                    offset: 24,
+                    shader_location: 2,
+                },
+                VertexAttribute {
+                    format: Mesh::ATTRIBUTE_TANGENT.format,
+                    offset: 32,
+                    shader_location: 3,
+                },
+            ],
+        },
+    ))
+}

crates/bevy_pbr/src/meshlet/meshlet_bindings.wgsl

@@ -0,0 +1,130 @@
+#define_import_path bevy_pbr::meshlet_bindings
+
+#import bevy_pbr::mesh_types::Mesh
+#import bevy_render::view::View
+
+struct PackedMeshletVertex {
+    a: vec4<f32>,
+    b: vec4<f32>,
+    tangent: vec4<f32>,
+}
+
+// TODO: Octahedral encode normal, remove tangent and derive from UV derivatives
+struct MeshletVertex {
+    position: vec3<f32>,
+    normal: vec3<f32>,
+    uv: vec2<f32>,
+    tangent: vec4<f32>,
+}
+
+fn unpack_meshlet_vertex(packed: PackedMeshletVertex) -> MeshletVertex {
+    var vertex: MeshletVertex;
+    vertex.position = packed.a.xyz;
+    vertex.normal = vec3(packed.a.w, packed.b.xy);
+    vertex.uv = packed.b.zw;
+    vertex.tangent = packed.tangent;
+    return vertex;
+}
+
+struct Meshlet {
+    start_vertex_id: u32,
+    start_index_id: u32,
+    triangle_count: u32,
+}
+
+struct MeshletBoundingSphere {
+    center: vec3<f32>,
+    radius: f32,
+}
+
+struct DrawIndirectArgs {
+    vertex_count: atomic<u32>,
+    instance_count: u32,
+    first_vertex: u32,
+    first_instance: u32,
+}
+
+#ifdef MESHLET_CULLING_PASS
+@group(0) @binding(0) var<storage, read> meshlet_thread_meshlet_ids: array<u32>; // Per cluster (instance of a meshlet)
+@group(0) @binding(1) var<storage, read> meshlet_bounding_spheres: array<MeshletBoundingSphere>; // Per asset meshlet
+@group(0) @binding(2) var<storage, read> meshlet_thread_instance_ids: array<u32>; // Per cluster (instance of a meshlet)
+@group(0) @binding(3) var<storage, read> meshlet_instance_uniforms: array<Mesh>; // Per entity instance
+@group(0) @binding(4) var<storage, read> meshlet_view_instance_visibility: array<u32>; // 1 bit per entity instance, packed as a bitmask
+@group(0) @binding(5) var<storage, read_write> meshlet_occlusion: array<atomic<u32>>; // 1 bit per cluster (instance of a meshlet), packed as a bitmask
+@group(0) @binding(6) var<storage, read> meshlet_previous_cluster_ids: array<u32>; // Per cluster (instance of a meshlet)
+@group(0) @binding(7) var<storage, read> meshlet_previous_occlusion: array<u32>; // 1 bit per cluster (instance of a meshlet), packed as a bitmask
+@group(0) @binding(8) var<uniform> view: View;
+@group(0) @binding(9) var depth_pyramid: texture_2d<f32>; // Generated from the first raster pass (unused in the first pass but still bound)
+
+fn should_cull_instance(instance_id: u32) -> bool {
+    let bit_offset = instance_id % 32u;
+    let packed_visibility = meshlet_view_instance_visibility[instance_id / 32u];
+    return bool(extractBits(packed_visibility, bit_offset, 1u));
+}
+
+fn get_meshlet_previous_occlusion(cluster_id: u32) -> bool {
+    let previous_cluster_id = meshlet_previous_cluster_ids[cluster_id];
+    let packed_occlusion = meshlet_previous_occlusion[previous_cluster_id / 32u];
+    let bit_offset = previous_cluster_id % 32u;
+    return bool(extractBits(packed_occlusion, bit_offset, 1u));
+}
+#endif
+
+#ifdef MESHLET_WRITE_INDEX_BUFFER_PASS
+@group(0) @binding(0) var<storage, read> meshlet_occlusion: array<u32>; // 1 bit per cluster (instance of a meshlet), packed as a bitmask
+@group(0) @binding(1) var<storage, read> meshlet_thread_meshlet_ids: array<u32>; // Per cluster (instance of a meshlet)
+@group(0) @binding(2) var<storage, read> meshlet_previous_cluster_ids: array<u32>; // Per cluster (instance of a meshlet)
+@group(0) @binding(3) var<storage, read> meshlet_previous_occlusion: array<u32>; // 1 bit per cluster (instance of a meshlet), packed as a bitmask
+@group(0) @binding(4) var<storage, read> meshlets: array<Meshlet>; // Per asset meshlet
+@group(0) @binding(5) var<storage, read_write> draw_indirect_args: DrawIndirectArgs; // Single object shared between all workgroups/meshlets/triangles
+@group(0) @binding(6) var<storage, read_write> draw_index_buffer: array<u32>; // Single object shared between all workgroups/meshlets/triangles
+
+fn get_meshlet_occlusion(cluster_id: u32) -> bool {
+    let packed_occlusion = meshlet_occlusion[cluster_id / 32u];
+    let bit_offset = cluster_id % 32u;
+    return bool(extractBits(packed_occlusion, bit_offset, 1u));
+}
+
+fn get_meshlet_previous_occlusion(cluster_id: u32) -> bool {
+    let previous_cluster_id = meshlet_previous_cluster_ids[cluster_id];
+    let packed_occlusion = meshlet_previous_occlusion[previous_cluster_id / 32u];
+    let bit_offset = previous_cluster_id % 32u;
+    return bool(extractBits(packed_occlusion, bit_offset, 1u));
+}
+#endif
+
+#ifdef MESHLET_VISIBILITY_BUFFER_RASTER_PASS
+@group(0) @binding(0) var<storage, read> meshlet_thread_meshlet_ids: array<u32>; // Per cluster (instance of a meshlet)
+@group(0) @binding(1) var<storage, read> meshlets: array<Meshlet>; // Per asset meshlet
+@group(0) @binding(2) var<storage, read> meshlet_indices: array<u32>; // Many per asset meshlet
+@group(0) @binding(3) var<storage, read> meshlet_vertex_ids: array<u32>; // Many per asset meshlet
+@group(0) @binding(4) var<storage, read> meshlet_vertex_data: array<PackedMeshletVertex>; // Many per asset meshlet
+@group(0) @binding(5) var<storage, read> meshlet_thread_instance_ids: array<u32>; // Per cluster (instance of a meshlet)
+@group(0) @binding(6) var<storage, read> meshlet_instance_uniforms: array<Mesh>; // Per entity instance
+@group(0) @binding(7) var<storage, read> meshlet_instance_material_ids: array<u32>; // Per entity instance
+@group(0) @binding(8) var<storage, read> draw_index_buffer: array<u32>; // Single object shared between all workgroups/meshlets/triangles
+@group(0) @binding(9) var<uniform> view: View;
+
+fn get_meshlet_index(index_id: u32) -> u32 {
+    let packed_index = meshlet_indices[index_id / 4u];
+    let bit_offset = (index_id % 4u) * 8u;
+    return extractBits(packed_index, bit_offset, 8u);
+}
+#endif
+
+#ifdef MESHLET_MESH_MATERIAL_PASS
+@group(1) @binding(0) var meshlet_visibility_buffer: texture_2d<u32>; // Generated from the meshlet raster passes
+@group(1) @binding(1) var<storage, read> meshlet_thread_meshlet_ids: array<u32>; // Per cluster (instance of a meshlet)
+@group(1) @binding(2) var<storage, read> meshlets: array<Meshlet>; // Per asset meshlet
+@group(1) @binding(3) var<storage, read> meshlet_indices: array<u32>; // Many per asset meshlet
+@group(1) @binding(4) var<storage, read> meshlet_vertex_ids: array<u32>; // Many per asset meshlet
+@group(1) @binding(5) var<storage, read> meshlet_vertex_data: array<PackedMeshletVertex>; // Many per asset meshlet
+@group(1) @binding(6) var<storage, read> meshlet_thread_instance_ids: array<u32>; // Per cluster (instance of a meshlet)
+@group(1) @binding(7) var<storage, read> meshlet_instance_uniforms: array<Mesh>; // Per entity instance
+
+fn get_meshlet_index(index_id: u32) -> u32 {
+    let packed_index = meshlet_indices[index_id / 4u];
+    let bit_offset = (index_id % 4u) * 8u;
+    return extractBits(packed_index, bit_offset, 8u);
+}
+#endif

crates/bevy_pbr/src/meshlet/meshlet_mesh_material.wgsl

@@ -0,0 +1,52 @@
+#import bevy_pbr::{
+    meshlet_visibility_buffer_resolve::resolve_vertex_output,
+    view_transformations::uv_to_ndc,
+    prepass_io,
+    pbr_prepass_functions,
+    utils::rand_f,
+}
+
+@vertex
+fn vertex(@builtin(vertex_index) vertex_input: u32) -> @builtin(position) vec4<f32> {
+    let vertex_index = vertex_input % 3u;
+    let material_id = vertex_input / 3u;
+    let material_depth = f32(material_id) / 65535.0;
+    let uv = vec2<f32>(vec2(vertex_index >> 1u, vertex_index & 1u)) * 2.0;
+    return vec4(uv_to_ndc(uv), material_depth, 1.0);
+}
+
+@fragment
+fn fragment(@builtin(position) frag_coord: vec4<f32>) -> @location(0) vec4<f32> {
+    let vertex_output = resolve_vertex_output(frag_coord);
+    var rng = vertex_output.meshlet_id;
+    let color = vec3(rand_f(&rng), rand_f(&rng), rand_f(&rng));
+    return vec4(color, 1.0);
+}
+
+#ifdef PREPASS_FRAGMENT
+@fragment
+fn prepass_fragment(@builtin(position) frag_coord: vec4<f32>) -> prepass_io::FragmentOutput {
+    let vertex_output = resolve_vertex_output(frag_coord);
+
+    var out: prepass_io::FragmentOutput;
+
+#ifdef NORMAL_PREPASS
+    out.normal = vec4(vertex_output.world_normal * 0.5 + vec3(0.5), 1.0);
+#endif
+
+#ifdef MOTION_VECTOR_PREPASS
+    out.motion_vector = vertex_output.motion_vector;
+#endif
+
+#ifdef DEFERRED_PREPASS
+    // There isn't any material info available for this default prepass shader so we are just writing 
+    // emissive magenta out to the deferred gbuffer to be rendered by the first deferred lighting pass layer.
+    // This is here so if the default prepass fragment is used for deferred magenta will be rendered, and also
+    // as an example to show that a user could write to the deferred gbuffer if they were to start from this shader.
+    out.deferred = vec4(0u, bevy_pbr::rgb9e5::vec3_to_rgb9e5_(vec3(1.0, 0.0, 1.0)), 0u, 0u);
+    out.deferred_lighting_pass_id = 1u;
+#endif
+
+    return out;
+}
+#endif

crates/bevy_pbr/src/meshlet/mod.rs

@@ -0,0 +1,280 @@
+//! Render high-poly 3d meshes using an efficient GPU-driven method. See [`MeshletPlugin`] and [`MeshletMesh`] for details.
+
+mod asset;
+#[cfg(feature = "meshlet_processor")]
+mod from_mesh;
+mod gpu_scene;
+mod material_draw_nodes;
+mod material_draw_prepare;
+mod persistent_buffer;
+mod persistent_buffer_impls;
+mod pipelines;
+mod visibility_buffer_raster_node;
+
+pub mod graph {
+    use bevy_render::render_graph::RenderLabel;
+
+    #[derive(Debug, Hash, PartialEq, Eq, Clone, RenderLabel)]
+    pub enum NodeMeshlet {
+        VisibilityBufferRasterPass,
+        Prepass,
+        DeferredPrepass,
+        MainOpaquePass,
+    }
+}
+
+pub(crate) use self::{
+    gpu_scene::{queue_material_meshlet_meshes, MeshletGpuScene},
+    material_draw_prepare::{
+        prepare_material_meshlet_meshes_main_opaque_pass, prepare_material_meshlet_meshes_prepass,
+    },
+};
+
+pub use self::asset::{Meshlet, MeshletBoundingSphere, MeshletMesh};
+#[cfg(feature = "meshlet_processor")]
+pub use self::from_mesh::MeshToMeshletMeshConversionError;
+
+use self::{
+    asset::MeshletMeshSaverLoad,
+    gpu_scene::{
+        extract_meshlet_meshes, perform_pending_meshlet_mesh_writes,
+        prepare_meshlet_per_frame_resources, prepare_meshlet_view_bind_groups,
+    },
+    graph::NodeMeshlet,
+    material_draw_nodes::{
+        MeshletDeferredGBufferPrepassNode, MeshletMainOpaquePass3dNode, MeshletPrepassNode,
+    },
+    material_draw_prepare::{
+        MeshletViewMaterialsDeferredGBufferPrepass, MeshletViewMaterialsMainOpaquePass,
+        MeshletViewMaterialsPrepass,
+    },
+    pipelines::{
+        MeshletPipelines, MESHLET_COPY_MATERIAL_DEPTH_SHADER_HANDLE, MESHLET_CULLING_SHADER_HANDLE,
+        MESHLET_DOWNSAMPLE_DEPTH_SHADER_HANDLE, MESHLET_VISIBILITY_BUFFER_RASTER_SHADER_HANDLE,
+        MESHLET_WRITE_INDEX_BUFFER_SHADER_HANDLE,
+    },
+    visibility_buffer_raster_node::MeshletVisibilityBufferRasterPassNode,
+};
+use crate::{graph::NodePbr, Material};
+use bevy_app::{App, Plugin};
+use bevy_asset::{load_internal_asset, AssetApp, Handle};
+use bevy_core_pipeline::{
+    core_3d::{
+        graph::{Core3d, Node3d},
+        Camera3d,
+    },
+    prepass::{DeferredPrepass, MotionVectorPrepass, NormalPrepass},
+};
+use bevy_ecs::{
+    bundle::Bundle,
+    entity::Entity,
+    query::Has,
+    schedule::IntoSystemConfigs,
+    system::{Commands, Query},
+};
+use bevy_render::{
+    render_graph::{RenderGraphApp, ViewNodeRunner},
+    render_resource::{Shader, TextureUsages},
+    view::{prepare_view_targets, InheritedVisibility, Msaa, ViewVisibility, Visibility},
+    ExtractSchedule, Render, RenderApp, RenderSet,
+};
+use bevy_transform::components::{GlobalTransform, Transform};
+
+const MESHLET_BINDINGS_SHADER_HANDLE: Handle<Shader> = Handle::weak_from_u128(1325134235233421);
+const MESHLET_MESH_MATERIAL_SHADER_HANDLE: Handle<Shader> =
+    Handle::weak_from_u128(3325134235233421);
+
+/// Provides a plugin for rendering large amounts of high-poly 3d meshes using an efficient GPU-driven method. See also [`MeshletMesh`].
+///
+/// Rendering dense scenes made of high-poly meshes with thousands or millions of triangles is extremely expensive in Bevy's standard renderer.
+/// Once meshes are pre-processed into a [`MeshletMesh`], this plugin can render these kinds of scenes very efficiently.
+///
+/// In comparison to Bevy's standard renderer:
+/// * Minimal rendering work is done on the CPU. All rendering is GPU-driven.
+/// * Much more efficient culling. Meshlets can be culled individually, instead of all or nothing culling for entire meshes at a time.
+/// Additionally, occlusion culling can eliminate meshlets that would cause overdraw.
+/// * Much more efficient batching. All geometry can be rasterized in a single indirect draw.
+/// * Scales better with large amounts of dense geometry and overdraw. Bevy's standard renderer will bottleneck sooner.
+/// * Much greater base overhead. Rendering will be slower than Bevy's standard renderer with small amounts of geometry and overdraw.
+/// * Much greater memory usage.
+/// * Requires preprocessing meshes. See [`MeshletMesh`] for details.
+/// * More limitations on the kinds of materials you can use. See [`MeshletMesh`] for details.
+///
+/// This plugin is not compatible with [`Msaa`], and adding this plugin will disable it.
+///
+/// This plugin does not work on the WebGL2 backend.
+///
+/// ![A render of the Stanford dragon as a `MeshletMesh`](https://raw.githubusercontent.com/bevyengine/bevy/meshlet/crates/bevy_pbr/src/meshlet/meshlet_preview.png)
+pub struct MeshletPlugin;
+
+impl Plugin for MeshletPlugin {
+    fn build(&self, app: &mut App) {
+        load_internal_asset!(
+            app,
+            MESHLET_BINDINGS_SHADER_HANDLE,
+            "meshlet_bindings.wgsl",
+            Shader::from_wgsl
+        );
+        load_internal_asset!(
+            app,
+            super::MESHLET_VISIBILITY_BUFFER_RESOLVE_SHADER_HANDLE,
+            "visibility_buffer_resolve.wgsl",
+            Shader::from_wgsl
+        );
+        load_internal_asset!(
+            app,
+            MESHLET_CULLING_SHADER_HANDLE,
+            "cull_meshlets.wgsl",
+            Shader::from_wgsl
+        );
+        load_internal_asset!(
+            app,
+            MESHLET_WRITE_INDEX_BUFFER_SHADER_HANDLE,
+            "write_index_buffer.wgsl",
+            Shader::from_wgsl
+        );
+        load_internal_asset!(
+            app,
+            MESHLET_DOWNSAMPLE_DEPTH_SHADER_HANDLE,
+            "downsample_depth.wgsl",
+            Shader::from_wgsl
+        );
+        load_internal_asset!(
+            app,
+            MESHLET_VISIBILITY_BUFFER_RASTER_SHADER_HANDLE,
+            "visibility_buffer_raster.wgsl",
+            Shader::from_wgsl
+        );
+        load_internal_asset!(
+            app,
+            MESHLET_MESH_MATERIAL_SHADER_HANDLE,
+            "meshlet_mesh_material.wgsl",
+            Shader::from_wgsl
+        );
+        load_internal_asset!(
+            app,
+            MESHLET_COPY_MATERIAL_DEPTH_SHADER_HANDLE,
+            "copy_material_depth.wgsl",
+            Shader::from_wgsl
+        );
+
+        app.init_asset::<MeshletMesh>()
+            .register_asset_loader(MeshletMeshSaverLoad)
+            .insert_resource(Msaa::Off);
+    }
+
+    fn finish(&self, app: &mut App) {
+        let Ok(render_app) = app.get_sub_app_mut(RenderApp) else {
+            return;
+        };
+
+        render_app
+            .add_render_graph_node::<MeshletVisibilityBufferRasterPassNode>(
+                Core3d,
+                NodeMeshlet::VisibilityBufferRasterPass,
+            )
+            .add_render_graph_node::<ViewNodeRunner<MeshletPrepassNode>>(
+                Core3d,
+                NodeMeshlet::Prepass,
+            )
+            .add_render_graph_node::<ViewNodeRunner<MeshletDeferredGBufferPrepassNode>>(
+                Core3d,
+                NodeMeshlet::DeferredPrepass,
+            )
+            .add_render_graph_node::<ViewNodeRunner<MeshletMainOpaquePass3dNode>>(
+                Core3d,
+                NodeMeshlet::MainOpaquePass,
+            )
+            .add_render_graph_edges(
+                Core3d,
+                (
+                    NodeMeshlet::VisibilityBufferRasterPass,
+                    NodePbr::ShadowPass,
+                    NodeMeshlet::Prepass,
+                    NodeMeshlet::DeferredPrepass,
+                    Node3d::Prepass,
+                    Node3d::DeferredPrepass,
+                    Node3d::CopyDeferredLightingId,
+                    Node3d::EndPrepasses,
+                    Node3d::StartMainPass,
+                    NodeMeshlet::MainOpaquePass,
+                    Node3d::MainOpaquePass,
+                    Node3d::EndMainPass,
+                ),
+            )
+            .init_resource::<MeshletGpuScene>()
+            .init_resource::<MeshletPipelines>()
+            .add_systems(ExtractSchedule, extract_meshlet_meshes)
+            .add_systems(
+                Render,
+                (
+                    perform_pending_meshlet_mesh_writes.in_set(RenderSet::PrepareAssets),
+                    configure_meshlet_views
+                        .after(prepare_view_targets)
+                        .in_set(RenderSet::ManageViews),
+                    prepare_meshlet_per_frame_resources.in_set(RenderSet::PrepareResources),
+                    prepare_meshlet_view_bind_groups.in_set(RenderSet::PrepareBindGroups),
+                ),
+            );
+    }
+}
+
+/// A component bundle for entities with a [`MeshletMesh`] and a [`Material`].
+#[derive(Bundle, Clone)]
+pub struct MaterialMeshletMeshBundle<M: Material> {
+    pub meshlet_mesh: Handle<MeshletMesh>,
+    pub material: Handle<M>,
+    pub transform: Transform,
+    pub global_transform: GlobalTransform,
+    /// User indication of whether an entity is visible
+    pub visibility: Visibility,
+    /// Inherited visibility of an entity.
+    pub inherited_visibility: InheritedVisibility,
+    /// Algorithmically-computed indication of whether an entity is visible and should be extracted for rendering
+    pub view_visibility: ViewVisibility,
+}
+
+impl<M: Material> Default for MaterialMeshletMeshBundle<M> {
+    fn default() -> Self {
+        Self {
+            meshlet_mesh: Default::default(),
+            material: Default::default(),
+            transform: Default::default(),
+            global_transform: Default::default(),
+            visibility: Default::default(),
+            inherited_visibility: Default::default(),
+            view_visibility: Default::default(),
+        }
+    }
+}
+
+fn configure_meshlet_views(
+    mut views_3d: Query<(
+        Entity,
+        &mut Camera3d,
+        Has<NormalPrepass>,
+        Has<MotionVectorPrepass>,
+        Has<DeferredPrepass>,
+    )>,
+    mut commands: Commands,
+) {
+    for (entity, mut camera_3d, normal_prepass, motion_vector_prepass, deferred_prepass) in
+        &mut views_3d
+    {
+        let mut usages: TextureUsages = camera_3d.depth_texture_usages.into();
+        usages |= TextureUsages::TEXTURE_BINDING;
+        camera_3d.depth_texture_usages = usages.into();
+
+        if !(normal_prepass || motion_vector_prepass || deferred_prepass) {
+            commands
+                .entity(entity)
+                .insert(MeshletViewMaterialsMainOpaquePass::default());
+        } else {
+            commands.entity(entity).insert((
+                MeshletViewMaterialsMainOpaquePass::default(),
+                MeshletViewMaterialsPrepass::default(),
+                MeshletViewMaterialsDeferredGBufferPrepass::default(),
+            ));
+        }
+    }
+}

crates/bevy_pbr/src/meshlet/persistent_buffer.rs

@@ -0,0 +1,124 @@
+use bevy_render::{
+    render_resource::{
+        BindingResource, Buffer, BufferAddress, BufferDescriptor, BufferUsages,
+        CommandEncoderDescriptor,
+    },
+    renderer::{RenderDevice, RenderQueue},
+};
+use range_alloc::RangeAllocator;
+use std::{num::NonZeroU64, ops::Range};
+
+/// Wrapper for a GPU buffer holding a large amount of data that persists across frames.
+pub struct PersistentGpuBuffer<T: PersistentGpuBufferable> {
+    /// Debug label for the buffer.
+    label: &'static str,
+    /// Handle to the GPU buffer.
+    buffer: Buffer,
+    /// Tracks free slices of the buffer.
+    allocation_planner: RangeAllocator<BufferAddress>,
+    /// Queue of pending writes, and associated metadata.
+    write_queue: Vec<(T, T::Metadata, Range<BufferAddress>)>,
+}
+
+impl<T: PersistentGpuBufferable> PersistentGpuBuffer<T> {
+    /// Create a new persistent buffer.
+    pub fn new(label: &'static str, render_device: &RenderDevice) -> Self {
+        Self {
+            label,
+            buffer: render_device.create_buffer(&BufferDescriptor {
+                label: Some(label),
+                size: 0,
+                usage: BufferUsages::STORAGE | BufferUsages::COPY_DST | BufferUsages::COPY_SRC,
+                mapped_at_creation: false,
+            }),
+            allocation_planner: RangeAllocator::new(0..0),
+            write_queue: Vec::new(),
+        }
+    }
+
+    /// Queue an item of type T to be added to the buffer, returning the byte range within the buffer that it will be located at.
+    pub fn queue_write(&mut self, data: T, metadata: T::Metadata) -> Range<BufferAddress> {
+        let data_size = data.size_in_bytes() as u64;
+        if let Ok(buffer_slice) = self.allocation_planner.allocate_range(data_size) {
+            self.write_queue
+                .push((data, metadata, buffer_slice.clone()));
+            return buffer_slice;
+        }
+
+        let buffer_size = self.allocation_planner.initial_range();
+        let double_buffer_size = (buffer_size.end - buffer_size.start) * 2;
+        let new_size = double_buffer_size.max(data_size);
+        self.allocation_planner.grow_to(buffer_size.end + new_size);
+
+        let buffer_slice = self.allocation_planner.allocate_range(data_size).unwrap();
+        self.write_queue
+            .push((data, metadata, buffer_slice.clone()));
+        buffer_slice
+    }
+
+    /// Upload all pending data to the GPU buffer.
+    pub fn perform_writes(&mut self, render_queue: &RenderQueue, render_device: &RenderDevice) {
+        if self.allocation_planner.initial_range().end > self.buffer.size() {
+            self.expand_buffer(render_device, render_queue);
+        }
+
+        let queue_count = self.write_queue.len();
+
+        for (data, metadata, buffer_slice) in self.write_queue.drain(..) {
+            let buffer_slice_size = NonZeroU64::new(buffer_slice.end - buffer_slice.start).unwrap();
+            let mut buffer_view = render_queue
+                .write_buffer_with(&self.buffer, buffer_slice.start, buffer_slice_size)
+                .unwrap();
+            data.write_bytes_le(metadata, &mut buffer_view);
+        }
+
+        let queue_saturation = queue_count as f32 / self.write_queue.capacity() as f32;
+        if queue_saturation < 0.3 {
+            self.write_queue = Vec::new();
+        }
+    }
+
+    /// Mark a section of the GPU buffer as no longer needed.
+    pub fn mark_slice_unused(&mut self, buffer_slice: Range<BufferAddress>) {
+        self.allocation_planner.free_range(buffer_slice);
+    }
+
+    pub fn binding(&self) -> BindingResource<'_> {
+        self.buffer.as_entire_binding()
+    }
+
+    /// Expand the buffer by creating a new buffer and copying old data over.
+    fn expand_buffer(&mut self, render_device: &RenderDevice, render_queue: &RenderQueue) {
+        let size = self.allocation_planner.initial_range();
+        let new_buffer = render_device.create_buffer(&BufferDescriptor {
+            label: Some(self.label),
+            size: size.end - size.start,
+            usage: BufferUsages::STORAGE | BufferUsages::COPY_DST | BufferUsages::COPY_SRC,
+            mapped_at_creation: false,
+        });
+
+        let mut command_encoder = render_device.create_command_encoder(&CommandEncoderDescriptor {
+            label: Some("persistent_gpu_buffer_expand"),
+        });
+        command_encoder.copy_buffer_to_buffer(&self.buffer, 0, &new_buffer, 0, self.buffer.size());
+        render_queue.submit([command_encoder.finish()]);
+
+        self.buffer = new_buffer;
+    }
+}
+
+/// A trait representing data that can be written to a [`PersistentGpuBuffer`].
+///
+/// # Safety
+/// * All data must be a multiple of `wgpu::COPY_BUFFER_ALIGNMENT` bytes.
+/// * The amount of bytes written to `buffer` in `write_bytes_le()` must match `size_in_bytes()`.
+pub unsafe trait PersistentGpuBufferable {
+    /// Additional metadata associated with each item, made available during `write_bytes_le`.
+    type Metadata;
+
+    /// The size in bytes of `self`.
+    fn size_in_bytes(&self) -> usize;
+
+    /// Convert `self` + `metadata` into bytes (little-endian), and write to the provided buffer slice.
+    fn write_bytes_le(&self, metadata: Self::Metadata, buffer_slice: &mut [u8]);
+}

crates/bevy_pbr/src/meshlet/persistent_buffer_impls.rs

@@ -0,0 +1,77 @@
+#![allow(clippy::undocumented_unsafe_blocks)]
+
+use super::{persistent_buffer::PersistentGpuBufferable, Meshlet, MeshletBoundingSphere};
+use std::{mem::size_of, sync::Arc};
+
+const MESHLET_VERTEX_SIZE_IN_BYTES: u32 = 48;
+
+unsafe impl PersistentGpuBufferable for Arc<[u8]> {
+    type Metadata = ();
+
+    fn size_in_bytes(&self) -> usize {
+        self.len()
+    }
+
+    fn write_bytes_le(&self, _: Self::Metadata, buffer_slice: &mut [u8]) {
+        buffer_slice.clone_from_slice(self);
+    }
+}
+
+unsafe impl PersistentGpuBufferable for Arc<[u32]> {
+    type Metadata = u64;
+
+    fn size_in_bytes(&self) -> usize {
+        self.len() * size_of::<u32>()
+    }
+
+    fn write_bytes_le(&self, offset: Self::Metadata, buffer_slice: &mut [u8]) {
+        let offset = offset as u32 / MESHLET_VERTEX_SIZE_IN_BYTES;
+
+        for (i, index) in self.iter().enumerate() {
+            let size = size_of::<u32>();
+            let i = i * size;
+            let bytes = (*index + offset).to_le_bytes();
+            buffer_slice[i..(i + size)].clone_from_slice(&bytes);
+        }
+    }
+}
+
+unsafe impl PersistentGpuBufferable for Arc<[Meshlet]> {
+    type Metadata = (u64, u64);
+
+    fn size_in_bytes(&self) -> usize {
+        self.len() * size_of::<Meshlet>()
+    }
+
+    fn write_bytes_le(
+        &self,
+        (vertex_offset, index_offset): Self::Metadata,
+        buffer_slice: &mut [u8],
+    ) {
+        let vertex_offset = (vertex_offset as usize / size_of::<u32>()) as u32;
+        let index_offset = index_offset as u32;
+
+        for (i, meshlet) in self.iter().enumerate() {
+            let size = size_of::<Meshlet>();
+            let i = i * size;
+            let bytes = bytemuck::cast::<_, [u8; size_of::<Meshlet>()]>(Meshlet {
+                start_vertex_id: meshlet.start_vertex_id + vertex_offset,
+                start_index_id: meshlet.start_index_id + index_offset,
+                triangle_count: meshlet.triangle_count,
+            });
+            buffer_slice[i..(i + size)].clone_from_slice(&bytes);
+        }
+    }
+}
+
+unsafe impl PersistentGpuBufferable for Arc<[MeshletBoundingSphere]> {
+    type Metadata = ();
+
+    fn size_in_bytes(&self) -> usize {
+        self.len() * size_of::<MeshletBoundingSphere>()
+    }
+
+    fn write_bytes_le(&self, _: Self::Metadata, buffer_slice: &mut [u8]) {
+        buffer_slice.clone_from_slice(bytemuck::cast_slice(self));
+    }
+}

crates/bevy_pbr/src/meshlet/pipelines.rs

@@ -0,0 +1,295 @@
+use super::gpu_scene::MeshletGpuScene;
+use bevy_asset::Handle;
+use bevy_core_pipeline::{
+    core_3d::CORE_3D_DEPTH_FORMAT, fullscreen_vertex_shader::fullscreen_shader_vertex_state,
+};
+use bevy_ecs::{
+    system::Resource,
+    world::{FromWorld, World},
+};
+use bevy_render::render_resource::*;
+
+pub const MESHLET_CULLING_SHADER_HANDLE: Handle<Shader> = Handle::weak_from_u128(4325134235233421);
+pub const MESHLET_WRITE_INDEX_BUFFER_SHADER_HANDLE: Handle<Shader> =
+    Handle::weak_from_u128(5325134235233421);
+pub const MESHLET_DOWNSAMPLE_DEPTH_SHADER_HANDLE: Handle<Shader> =
+    Handle::weak_from_u128(6325134235233421);
+pub const MESHLET_VISIBILITY_BUFFER_RASTER_SHADER_HANDLE: Handle<Shader> =
+    Handle::weak_from_u128(7325134235233421);
+pub const MESHLET_COPY_MATERIAL_DEPTH_SHADER_HANDLE: Handle<Shader> =
+    Handle::weak_from_u128(8325134235233421);
+
+#[derive(Resource)]
+pub struct MeshletPipelines {
+    cull_first: CachedComputePipelineId,
+    cull_second: CachedComputePipelineId,
+    write_index_buffer_first: CachedComputePipelineId,
+    write_index_buffer_second: CachedComputePipelineId,
+    downsample_depth: CachedRenderPipelineId,
+    visibility_buffer_raster: CachedRenderPipelineId,
+    visibility_buffer_raster_depth_only: CachedRenderPipelineId,
+    visibility_buffer_raster_depth_only_clamp_ortho: CachedRenderPipelineId,
+    copy_material_depth: CachedRenderPipelineId,
+}
+
+impl FromWorld for MeshletPipelines {
+    fn from_world(world: &mut World) -> Self {
+        let gpu_scene = world.resource::<MeshletGpuScene>();
+        let cull_layout = gpu_scene.culling_bind_group_layout();
+        let write_index_buffer_layout = gpu_scene.write_index_buffer_bind_group_layout();
+        let downsample_depth_layout = gpu_scene.downsample_depth_bind_group_layout();
+        let visibility_buffer_layout = gpu_scene.visibility_buffer_raster_bind_group_layout();
+        let copy_material_depth_layout = gpu_scene.copy_material_depth_bind_group_layout();
+        let pipeline_cache = world.resource_mut::<PipelineCache>();
+
+        Self {
+            cull_first: pipeline_cache.queue_compute_pipeline(ComputePipelineDescriptor {
+                label: Some("meshlet_culling_first_pipeline".into()),
+                layout: vec![cull_layout.clone()],
+                push_constant_ranges: vec![],
+                shader: MESHLET_CULLING_SHADER_HANDLE,
+                shader_defs: vec!["MESHLET_CULLING_PASS".into()],
+                entry_point: "cull_meshlets".into(),
+            }),
+
+            cull_second: pipeline_cache.queue_compute_pipeline(ComputePipelineDescriptor {
+                label: Some("meshlet_culling_second_pipeline".into()),
+                layout: vec![cull_layout],
+                push_constant_ranges: vec![],
+                shader: MESHLET_CULLING_SHADER_HANDLE,
+                shader_defs: vec![
+                    "MESHLET_CULLING_PASS".into(),
+                    "MESHLET_SECOND_CULLING_PASS".into(),
+                ],
+                entry_point: "cull_meshlets".into(),
+            }),
+
+            write_index_buffer_first: pipeline_cache.queue_compute_pipeline(
+                ComputePipelineDescriptor {
+                    label: Some("meshlet_write_index_buffer_first_pipeline".into()),
+                    layout: vec![write_index_buffer_layout.clone()],
+                    push_constant_ranges: vec![],
+                    shader: MESHLET_WRITE_INDEX_BUFFER_SHADER_HANDLE,
+                    shader_defs: vec!["MESHLET_WRITE_INDEX_BUFFER_PASS".into()],
+                    entry_point: "write_index_buffer".into(),
+                },
+            ),
+
+            write_index_buffer_second: pipeline_cache.queue_compute_pipeline(
+                ComputePipelineDescriptor {
+                    label: Some("meshlet_write_index_buffer_second_pipeline".into()),
+                    layout: vec![write_index_buffer_layout],
+                    push_constant_ranges: vec![],
+                    shader: MESHLET_WRITE_INDEX_BUFFER_SHADER_HANDLE,
+                    shader_defs: vec![
+                        "MESHLET_WRITE_INDEX_BUFFER_PASS".into(),
+                        "MESHLET_SECOND_WRITE_INDEX_BUFFER_PASS".into(),
+                    ],
+                    entry_point: "write_index_buffer".into(),
+                },
+            ),
+
+            downsample_depth: pipeline_cache.queue_render_pipeline(RenderPipelineDescriptor {
+                label: Some("meshlet_downsample_depth".into()),
+                layout: vec![downsample_depth_layout],
+                push_constant_ranges: vec![],
+                vertex: fullscreen_shader_vertex_state(),
+                primitive: PrimitiveState::default(),
+                depth_stencil: None,
+                multisample: MultisampleState::default(),
+                fragment: Some(FragmentState {
+                    shader: MESHLET_DOWNSAMPLE_DEPTH_SHADER_HANDLE,
+                    shader_defs: vec![],
+                    entry_point: "downsample_depth".into(),
+                    targets: vec![Some(ColorTargetState {
+                        format: TextureFormat::R32Float,
+                        blend: None,
+                        write_mask: ColorWrites::ALL,
+                    })],
+                }),
+            }),
+
+            visibility_buffer_raster: pipeline_cache.queue_render_pipeline(
+                RenderPipelineDescriptor {
+                    label: Some("meshlet_visibility_buffer_raster_pipeline".into()),
+                    layout: vec![visibility_buffer_layout.clone()],
+                    push_constant_ranges: vec![],
+                    vertex: VertexState {
+                        shader: MESHLET_VISIBILITY_BUFFER_RASTER_SHADER_HANDLE,
+                        shader_defs: vec![
+                            "MESHLET_VISIBILITY_BUFFER_RASTER_PASS".into(),
+                            "MESHLET_VISIBILITY_BUFFER_RASTER_PASS_OUTPUT".into(),
+                        ],
+                        entry_point: "vertex".into(),
+                        buffers: vec![],
+                    },
+                    primitive: PrimitiveState {
+                        topology: PrimitiveTopology::TriangleList,
+                        strip_index_format: None,
+                        front_face: FrontFace::Ccw,
+                        cull_mode: None,
+                        unclipped_depth: false,
+                        polygon_mode: PolygonMode::Fill,
+                        conservative: false,
+                    },
+                    depth_stencil: Some(DepthStencilState {
+                        format: CORE_3D_DEPTH_FORMAT,
+                        depth_write_enabled: true,
+                        depth_compare: CompareFunction::GreaterEqual,
+                        stencil: StencilState::default(),
+                        bias: DepthBiasState::default(),
+                    }),
+                    multisample: MultisampleState::default(),
+                    fragment: Some(FragmentState {
+                        shader: MESHLET_VISIBILITY_BUFFER_RASTER_SHADER_HANDLE,
+                        shader_defs: vec![
+                            "MESHLET_VISIBILITY_BUFFER_RASTER_PASS".into(),
+                            "MESHLET_VISIBILITY_BUFFER_RASTER_PASS_OUTPUT".into(),
+                        ],
+                        entry_point: "fragment".into(),
+                        targets: vec![
+                            Some(ColorTargetState {
+                                format: TextureFormat::R32Uint,
+                                blend: None,
+                                write_mask: ColorWrites::ALL,
+                            }),
+                            Some(ColorTargetState {
+                                format: TextureFormat::R16Uint,
+                                blend: None,
+                                write_mask: ColorWrites::ALL,
+                            }),
+                        ],
+                    }),
+                },
+            ),
+
+            visibility_buffer_raster_depth_only: pipeline_cache.queue_render_pipeline(
+                RenderPipelineDescriptor {
+                    label: Some("meshlet_visibility_buffer_raster_depth_only_pipeline".into()),
+                    layout: vec![visibility_buffer_layout.clone()],
+                    push_constant_ranges: vec![],
+                    vertex: VertexState {
+                        shader: MESHLET_VISIBILITY_BUFFER_RASTER_SHADER_HANDLE,
+                        shader_defs: vec!["MESHLET_VISIBILITY_BUFFER_RASTER_PASS".into()],
+                        entry_point: "vertex".into(),
+                        buffers: vec![],
+                    },
+                    primitive: PrimitiveState {
+                        topology: PrimitiveTopology::TriangleList,
+                        strip_index_format: None,
+                        front_face: FrontFace::Ccw,
+                        cull_mode: None,
+                        unclipped_depth: false,
+                        polygon_mode: PolygonMode::Fill,
+                        conservative: false,
+                    },
+                    depth_stencil: Some(DepthStencilState {
+                        format: CORE_3D_DEPTH_FORMAT,
+                        depth_write_enabled: true,
+                        depth_compare: CompareFunction::GreaterEqual,
+                        stencil: StencilState::default(),
+                        bias: DepthBiasState::default(),
+                    }),
+                    multisample: MultisampleState::default(),
+                    fragment: None,
+                },
+            ),
+
+            visibility_buffer_raster_depth_only_clamp_ortho: pipeline_cache.queue_render_pipeline(
+                RenderPipelineDescriptor {
+                    label: Some("visibility_buffer_raster_depth_only_clamp_ortho_pipeline".into()),
+                    layout: vec![visibility_buffer_layout],
+                    push_constant_ranges: vec![],
+                    vertex: VertexState {
+                        shader: MESHLET_VISIBILITY_BUFFER_RASTER_SHADER_HANDLE,
+                        shader_defs: vec![
+                            "MESHLET_VISIBILITY_BUFFER_RASTER_PASS".into(),
+                            "DEPTH_CLAMP_ORTHO".into(),
+                        ],
+                        entry_point: "vertex".into(),
+                        buffers: vec![],
+                    },
+                    primitive: PrimitiveState {
+                        topology: PrimitiveTopology::TriangleList,
+                        strip_index_format: None,
+                        front_face: FrontFace::Ccw,
+                        cull_mode: None,
+                        unclipped_depth: false,
+                        polygon_mode: PolygonMode::Fill,
+                        conservative: false,
+                    },
+                    depth_stencil: Some(DepthStencilState {
+                        format: CORE_3D_DEPTH_FORMAT,
+                        depth_write_enabled: true,
+                        depth_compare: CompareFunction::GreaterEqual,
+                        stencil: StencilState::default(),
+                        bias: DepthBiasState::default(),
+                    }),
+                    multisample: MultisampleState::default(),
+                    fragment: Some(FragmentState {
+                        shader: MESHLET_VISIBILITY_BUFFER_RASTER_SHADER_HANDLE,
+                        shader_defs: vec![
+                            "MESHLET_VISIBILITY_BUFFER_RASTER_PASS".into(),
+                            "DEPTH_CLAMP_ORTHO".into(),
+                        ],
+                        entry_point: "fragment".into(),
+                        targets: vec![],
+                    }),
+                },
+            ),
+
+            copy_material_depth: pipeline_cache.queue_render_pipeline(RenderPipelineDescriptor {
+                label: Some("meshlet_copy_material_depth".into()),
+                layout: vec![copy_material_depth_layout],
+                push_constant_ranges: vec![],
+                vertex: fullscreen_shader_vertex_state(),
+                primitive: PrimitiveState::default(),
+                depth_stencil: Some(DepthStencilState {
+                    format: TextureFormat::Depth16Unorm,
+                    depth_write_enabled: true,
+                    depth_compare: CompareFunction::Always,
+                    stencil: StencilState::default(),
+                    bias: DepthBiasState::default(),
+                }),
+                multisample: MultisampleState::default(),
+                fragment: Some(FragmentState {
+                    shader: MESHLET_COPY_MATERIAL_DEPTH_SHADER_HANDLE,
+                    shader_defs: vec![],
+                    entry_point: "copy_material_depth".into(),
+                    targets: vec![],
+                }),
+            }),
+        }
+    }
+}
+
+impl MeshletPipelines {
+    pub fn get(
+        world: &World,
+    ) -> Option<(
+        &ComputePipeline,
+        &ComputePipeline,
+        &ComputePipeline,
+        &ComputePipeline,
+        &RenderPipeline,
+        &RenderPipeline,
+        &RenderPipeline,
+        &RenderPipeline,
+        &RenderPipeline,
+    )> {
+        let pipeline_cache = world.get_resource::<PipelineCache>()?;
+        let pipeline = world.get_resource::<Self>()?;
+        Some((
+            pipeline_cache.get_compute_pipeline(pipeline.cull_first)?,
+            pipeline_cache.get_compute_pipeline(pipeline.cull_second)?,
+            pipeline_cache.get_compute_pipeline(pipeline.write_index_buffer_first)?,
+            pipeline_cache.get_compute_pipeline(pipeline.write_index_buffer_second)?,
+            pipeline_cache.get_render_pipeline(pipeline.downsample_depth)?,
+            pipeline_cache.get_render_pipeline(pipeline.visibility_buffer_raster)?,
+            pipeline_cache.get_render_pipeline(pipeline.visibility_buffer_raster_depth_only)?,
+            pipeline_cache
+                .get_render_pipeline(pipeline.visibility_buffer_raster_depth_only_clamp_ortho)?,
+            pipeline_cache.get_render_pipeline(pipeline.copy_material_depth)?,
+        ))
+    }
+}

crates/bevy_pbr/src/meshlet/visibility_buffer_raster.wgsl

@@ -0,0 +1,88 @@
+#import bevy_pbr::{
+    meshlet_bindings::{
+        meshlet_thread_meshlet_ids,
+        meshlets,
+        meshlet_vertex_ids,
+        meshlet_vertex_data,
+        meshlet_thread_instance_ids,
+        meshlet_instance_uniforms,
+        meshlet_instance_material_ids,
+        draw_index_buffer,
+        view,
+        get_meshlet_index,
+        unpack_meshlet_vertex,
+    },
+    mesh_functions::mesh_position_local_to_world,
+}
+#import bevy_render::maths::affine3_to_square
+
+/// Vertex/fragment shader for rasterizing meshlets into a visibility buffer.
+
+struct VertexOutput {
+    @builtin(position) clip_position: vec4<f32>,
+#ifdef MESHLET_VISIBILITY_BUFFER_RASTER_PASS_OUTPUT
+    @location(0) @interpolate(flat) visibility: u32,
+    @location(1) @interpolate(flat) material_depth: u32,
+#endif
+#ifdef DEPTH_CLAMP_ORTHO
+    @location(0) unclamped_clip_depth: f32,
+#endif
+}
+
+#ifdef MESHLET_VISIBILITY_BUFFER_RASTER_PASS_OUTPUT
+struct FragmentOutput {
+    @location(0) visibility: vec4<u32>,
+    @location(1) material_depth: vec4<u32>,
+}
+#endif
+
+@vertex
+fn vertex(@builtin(vertex_index) vertex_index: u32) -> VertexOutput {
+    let packed_ids = draw_index_buffer[vertex_index / 3u];
+    let cluster_id = packed_ids >> 8u;
+    let triangle_id = extractBits(packed_ids, 0u, 8u);
+    let index_id = (triangle_id * 3u) + (vertex_index % 3u);
+    let meshlet_id = meshlet_thread_meshlet_ids[cluster_id];
+    let meshlet = meshlets[meshlet_id];
+    let index = get_meshlet_index(meshlet.start_index_id + index_id);
+    let vertex_id = meshlet_vertex_ids[meshlet.start_vertex_id + index];
+    let vertex = unpack_meshlet_vertex(meshlet_vertex_data[vertex_id]);
+    let instance_id = meshlet_thread_instance_ids[cluster_id];
+    let instance_uniform = meshlet_instance_uniforms[instance_id];
+
+    let model = affine3_to_square(instance_uniform.model);
+    let world_position = mesh_position_local_to_world(model, vec4(vertex.position, 1.0));
+    var clip_position = view.view_proj * vec4(world_position.xyz, 1.0);
+#ifdef DEPTH_CLAMP_ORTHO
+    let unclamped_clip_depth = clip_position.z;
+    clip_position.z = min(clip_position.z, 1.0);
+#endif
+
+    return VertexOutput(
+        clip_position,
+#ifdef MESHLET_VISIBILITY_BUFFER_RASTER_PASS_OUTPUT
+        packed_ids,
+        meshlet_instance_material_ids[instance_id],
+#endif
+#ifdef DEPTH_CLAMP_ORTHO
+        unclamped_clip_depth,
+#endif
+    );
+}
+
+#ifdef MESHLET_VISIBILITY_BUFFER_RASTER_PASS_OUTPUT
+@fragment
+fn fragment(vertex_output: VertexOutput) -> FragmentOutput {
+    return FragmentOutput(
+        vec4(vertex_output.visibility, 0u, 0u, 0u),
+        vec4(vertex_output.material_depth, 0u, 0u, 0u),
+    );
+}
+#endif
+
+#ifdef DEPTH_CLAMP_ORTHO
+@fragment
+fn fragment(vertex_output: VertexOutput) -> @builtin(frag_depth) f32 {
+    return vertex_output.unclamped_clip_depth;
+}
+#endif

crates/bevy_pbr/src/meshlet/visibility_buffer_raster_node.rs

@@ -0,0 +1,448 @@
+use super::{
+    gpu_scene::{MeshletViewBindGroups, MeshletViewResources},
+    pipelines::MeshletPipelines,
+};
+use crate::{LightEntity, ShadowView, ViewLightEntities};
+use bevy_color::LinearRgba;
+use bevy_ecs::{
+    query::QueryState,
+    world::{FromWorld, World},
+};
+use bevy_render::{
+    camera::ExtractedCamera,
+    render_graph::{Node, NodeRunError, RenderGraphContext},
+    render_resource::*,
+    renderer::RenderContext,
+    view::{ViewDepthTexture, ViewUniformOffset},
+};
+
+/// Rasterize meshlets into a depth buffer, and optional visibility buffer + material depth buffer for shading passes.
+pub struct MeshletVisibilityBufferRasterPassNode {
+    main_view_query: QueryState<(
+        &'static ExtractedCamera,
+        &'static ViewDepthTexture,
+        &'static ViewUniformOffset,
+        &'static MeshletViewBindGroups,
+        &'static MeshletViewResources,
+        &'static ViewLightEntities,
+    )>,
+    view_light_query: QueryState<(
+        &'static ShadowView,
+        &'static LightEntity,
+        &'static ViewUniformOffset,
+        &'static MeshletViewBindGroups,
+        &'static MeshletViewResources,
+    )>,
+}
+
+impl FromWorld for MeshletVisibilityBufferRasterPassNode {
+    fn from_world(world: &mut World) -> Self {
+        Self {
+            main_view_query: QueryState::new(world),
+            view_light_query: QueryState::new(world),
+        }
+    }
+}
+
+impl Node for MeshletVisibilityBufferRasterPassNode {
+    fn update(&mut self, world: &mut World) {
+        self.main_view_query.update_archetypes(world);
+        self.view_light_query.update_archetypes(world);
+    }
+
+    fn run(
+        &self,
+        graph: &mut RenderGraphContext,
+        render_context: &mut RenderContext,
+        world: &World,
+    ) -> Result<(), NodeRunError> {
+        let Ok((
+            camera,
+            view_depth,
+            view_offset,
+            meshlet_view_bind_groups,
+            meshlet_view_resources,
+            lights,
+        )) = self.main_view_query.get_manual(world, graph.view_entity())
+        else {
+            return Ok(());
+        };
+
+        let Some((
+            culling_first_pipeline,
+            culling_second_pipeline,
+            write_index_buffer_first_pipeline,
+            write_index_buffer_second_pipeline,
+            downsample_depth_pipeline,
+            visibility_buffer_raster_pipeline,
+            visibility_buffer_raster_depth_only_pipeline,
+            visibility_buffer_raster_depth_only_clamp_ortho,
+            copy_material_depth_pipeline,
+        )) = MeshletPipelines::get(world)
+        else {
+            return Ok(());
+        };
+
+        let culling_workgroups = meshlet_view_resources.scene_meshlet_count.div_ceil(128);
+        let write_index_buffer_workgroups = (meshlet_view_resources.scene_meshlet_count as f32)
+            .cbrt()
+            .ceil() as u32;
+
+        render_context
+            .command_encoder()
+            .push_debug_group("meshlet_visibility_buffer_raster_pass");
+        if meshlet_view_resources.occlusion_buffer_needs_clearing {
+            render_context.command_encoder().clear_buffer(
+                &meshlet_view_resources.occlusion_buffer,
+                0,
+                None,
+            );
+        }
+        cull_pass(
+            "meshlet_culling_first_pass",
+            render_context,
+            meshlet_view_bind_groups,
+            view_offset,
+            culling_first_pipeline,
+            culling_workgroups,
+        );
+        write_index_buffer_pass(
+            "meshlet_write_index_buffer_first_pass",
+            render_context,
+            &meshlet_view_bind_groups.write_index_buffer_first,
+            write_index_buffer_first_pipeline,
+            write_index_buffer_workgroups,
+        );
+        render_context.command_encoder().clear_buffer(
+            &meshlet_view_resources.occlusion_buffer,
+            0,
+            None,
+        );
+        raster_pass(
+            true,
+            render_context,
+            meshlet_view_resources,
+            &meshlet_view_resources.visibility_buffer_draw_indirect_args_first,
+            view_depth.get_attachment(StoreOp::Store),
+            meshlet_view_bind_groups,
+            view_offset,
+            visibility_buffer_raster_pipeline,
+            Some(camera),
+        );
+        downsample_depth(
+            render_context,
+            meshlet_view_resources,
+            meshlet_view_bind_groups,
+            downsample_depth_pipeline,
+        );
+        cull_pass(
+            "meshlet_culling_second_pass",
+            render_context,
+            meshlet_view_bind_groups,
+            view_offset,
+            culling_second_pipeline,
+            culling_workgroups,
+        );
+        write_index_buffer_pass(
+            "meshlet_write_index_buffer_second_pass",
+            render_context,
+            &meshlet_view_bind_groups.write_index_buffer_second,
+            write_index_buffer_second_pipeline,
+            write_index_buffer_workgroups,
+        );
+        raster_pass(
+            false,
+            render_context,
+            meshlet_view_resources,
+            &meshlet_view_resources.visibility_buffer_draw_indirect_args_second,
+            view_depth.get_attachment(StoreOp::Store),
+            meshlet_view_bind_groups,
+            view_offset,
+            visibility_buffer_raster_pipeline,
+            Some(camera),
+        );
+        copy_material_depth_pass(
+            render_context,
+            meshlet_view_resources,
+            meshlet_view_bind_groups,
+            copy_material_depth_pipeline,
+            camera,
+        );
+        render_context.command_encoder().pop_debug_group();
+
+        for light_entity in &lights.lights {
+            let Ok((
+                shadow_view,
+                light_type,
+                view_offset,
+                meshlet_view_bind_groups,
+                meshlet_view_resources,
+            )) = self.view_light_query.get_manual(world, *light_entity)
+            else {
+                continue;
+            };
+
+            let shadow_visibility_buffer_pipeline = match light_type {
+                LightEntity::Directional { .. } => visibility_buffer_raster_depth_only_clamp_ortho,
+                _ => visibility_buffer_raster_depth_only_pipeline,
+            };
+
+            render_context.command_encoder().push_debug_group(&format!(
+                "meshlet_visibility_buffer_raster_pass: {}",
+                shadow_view.pass_name
+            ));
+            if meshlet_view_resources.occlusion_buffer_needs_clearing {
+                render_context.command_encoder().clear_buffer(
+                    &meshlet_view_resources.occlusion_buffer,
+                    0,
+                    None,
+                );
+            }
+            cull_pass(
+                "meshlet_culling_first_pass",
+                render_context,
+                meshlet_view_bind_groups,
+                view_offset,
+                culling_first_pipeline,
+                culling_workgroups,
+            );
+            write_index_buffer_pass(
+                "meshlet_write_index_buffer_first_pass",
+                render_context,
+                &meshlet_view_bind_groups.write_index_buffer_first,
+                write_index_buffer_first_pipeline,
+                write_index_buffer_workgroups,
+            );
+            render_context.command_encoder().clear_buffer(
+                &meshlet_view_resources.occlusion_buffer,
+                0,
+                None,
+            );
+            raster_pass(
+                true,
+                render_context,
+                meshlet_view_resources,
+                &meshlet_view_resources.visibility_buffer_draw_indirect_args_first,
+                shadow_view.depth_attachment.get_attachment(StoreOp::Store),
+                meshlet_view_bind_groups,
+                view_offset,
+                shadow_visibility_buffer_pipeline,
+                None,
+            );
+            downsample_depth(
+                render_context,
+                meshlet_view_resources,
+                meshlet_view_bind_groups,
+                downsample_depth_pipeline,
+            );
+            cull_pass(
+                "meshlet_culling_second_pass",
+                render_context,
+                meshlet_view_bind_groups,
+                view_offset,
+                culling_second_pipeline,
+                culling_workgroups,
+            );
+            write_index_buffer_pass(
+                "meshlet_write_index_buffer_second_pass",
+                render_context,
+                &meshlet_view_bind_groups.write_index_buffer_second,
+                write_index_buffer_second_pipeline,
+                write_index_buffer_workgroups,
+            );
+            raster_pass(
+                false,
+                render_context,
+                meshlet_view_resources,
+                &meshlet_view_resources.visibility_buffer_draw_indirect_args_second,
+                shadow_view.depth_attachment.get_attachment(StoreOp::Store),
+                meshlet_view_bind_groups,
+                view_offset,
+                shadow_visibility_buffer_pipeline,
+                None,
+            );
+            render_context.command_encoder().pop_debug_group();
+        }
+
+        Ok(())
+    }
+}
+
+fn cull_pass(
+    label: &'static str,
+    render_context: &mut RenderContext,
+    meshlet_view_bind_groups: &MeshletViewBindGroups,
+    view_offset: &ViewUniformOffset,
+    culling_pipeline: &ComputePipeline,
+    culling_workgroups: u32,
+) {
+    let command_encoder = render_context.command_encoder();
+    let mut cull_pass = command_encoder.begin_compute_pass(&ComputePassDescriptor {
+        label: Some(label),
+        timestamp_writes: None,
+    });
+    cull_pass.set_bind_group(0, &meshlet_view_bind_groups.culling, &[view_offset.offset]);
+    cull_pass.set_pipeline(culling_pipeline);
+    cull_pass.dispatch_workgroups(culling_workgroups, 1, 1);
+}
+
+fn write_index_buffer_pass(
+    label: &'static str,
+    render_context: &mut RenderContext,
+    write_index_buffer_bind_group: &BindGroup,
+    write_index_buffer_pipeline: &ComputePipeline,
+    write_index_buffer_workgroups: u32,
+) {
+    let command_encoder = render_context.command_encoder();
+    let mut cull_pass = command_encoder.begin_compute_pass(&ComputePassDescriptor {
+        label: Some(label),
+        timestamp_writes: None,
+    });
+    cull_pass.set_bind_group(0, write_index_buffer_bind_group, &[]);
+    cull_pass.set_pipeline(write_index_buffer_pipeline);
+    cull_pass.dispatch_workgroups(
+        write_index_buffer_workgroups,
+        write_index_buffer_workgroups,
+        write_index_buffer_workgroups,
+    );
+}
+
+#[allow(clippy::too_many_arguments)]
+fn raster_pass(
+    first_pass: bool,
+    render_context: &mut RenderContext,
+    meshlet_view_resources: &MeshletViewResources,
+    visibility_buffer_draw_indirect_args: &Buffer,
+    depth_stencil_attachment: RenderPassDepthStencilAttachment,
+    meshlet_view_bind_groups: &MeshletViewBindGroups,
+    view_offset: &ViewUniformOffset,
+    visibility_buffer_raster_pipeline: &RenderPipeline,
+    camera: Option<&ExtractedCamera>,
+) {
+    let mut color_attachments_filled = [None, None];
+    if let (Some(visibility_buffer), Some(material_depth_color)) = (
+        meshlet_view_resources.visibility_buffer.as_ref(),
+        meshlet_view_resources.material_depth_color.as_ref(),
+    ) {
+        let load = if first_pass {
+            LoadOp::Clear(LinearRgba::BLACK.into())
+        } else {
+            LoadOp::Load
+        };
+        color_attachments_filled = [
+            Some(RenderPassColorAttachment {
+                view: &visibility_buffer.default_view,
+                resolve_target: None,
+                ops: Operations {
+                    load,
+                    store: StoreOp::Store,
+                },
+            }),
+            Some(RenderPassColorAttachment {
+                view: &material_depth_color.default_view,
+                resolve_target: None,
+                ops: Operations {
+                    load,
+                    store: StoreOp::Store,
+                },
+            }),
+        ];
+    }
+
+    let mut draw_pass = render_context.begin_tracked_render_pass(RenderPassDescriptor {
+        label: Some(if first_pass {
+            "meshlet_visibility_buffer_raster_first_pass"
+        } else {
+            "meshlet_visibility_buffer_raster_second_pass"
+        }),
+        color_attachments: if color_attachments_filled[0].is_none() {
+            &[]
+        } else {
+            &color_attachments_filled
+        },
+        depth_stencil_attachment: Some(depth_stencil_attachment),
+        timestamp_writes: None,
+        occlusion_query_set: None,
+    });
+    if let Some(viewport) = camera.and_then(|camera| camera.viewport.as_ref()) {
+        draw_pass.set_camera_viewport(viewport);
+    }
+
+    draw_pass.set_bind_group(
+        0,
+        &meshlet_view_bind_groups.visibility_buffer_raster,
+        &[view_offset.offset],
+    );
+    draw_pass.set_render_pipeline(visibility_buffer_raster_pipeline);
+    draw_pass.draw_indirect(visibility_buffer_draw_indirect_args, 0);
+}
+
+fn downsample_depth(
+    render_context: &mut RenderContext,
+    meshlet_view_resources: &MeshletViewResources,
+    meshlet_view_bind_groups: &MeshletViewBindGroups,
+    downsample_depth_pipeline: &RenderPipeline,
+) {
+    render_context
+        .command_encoder()
+        .push_debug_group("meshlet_downsample_depth");
+
+    for i in 0..meshlet_view_resources.depth_pyramid_mips.len() {
+        let downsample_pass = RenderPassDescriptor {
+            label: Some("meshlet_downsample_depth_pass"),
+            color_attachments: &[Some(RenderPassColorAttachment {
+                view: &meshlet_view_resources.depth_pyramid_mips[i],
+                resolve_target: None,
+                ops: Operations {
+                    load: LoadOp::Clear(LinearRgba::BLACK.into()),
+                    store: StoreOp::Store,
+                },
+            })],
+            depth_stencil_attachment: None,
+            timestamp_writes: None,
+            occlusion_query_set: None,
+        };
+
+        let mut downsample_pass = render_context.begin_tracked_render_pass(downsample_pass);
+        downsample_pass.set_bind_group(0, &meshlet_view_bind_groups.downsample_depth[i], &[]);
+        downsample_pass.set_render_pipeline(downsample_depth_pipeline);
+        downsample_pass.draw(0..3, 0..1);
+    }
+
+    render_context.command_encoder().pop_debug_group();
+}
+
+fn copy_material_depth_pass(
+    render_context: &mut RenderContext,
+    meshlet_view_resources: &MeshletViewResources,
+    meshlet_view_bind_groups: &MeshletViewBindGroups,
+    copy_material_depth_pipeline: &RenderPipeline,
+    camera: &ExtractedCamera,
+) {
+    if let (Some(material_depth), Some(copy_material_depth_bind_group)) = (
+        meshlet_view_resources.material_depth.as_ref(),
+        meshlet_view_bind_groups.copy_material_depth.as_ref(),
+    ) {
+        let mut copy_pass = render_context.begin_tracked_render_pass(RenderPassDescriptor {
+            label: Some("meshlet_copy_material_depth_pass"),
+            color_attachments: &[],
+            depth_stencil_attachment: Some(RenderPassDepthStencilAttachment {
+                view: &material_depth.default_view,
+                depth_ops: Some(Operations {
+                    load: LoadOp::Clear(0.0),
+                    store: StoreOp::Store,
+                }),
+                stencil_ops: None,
+            }),
+            timestamp_writes: None,
+            occlusion_query_set: None,
+        });
+        if let Some(viewport) = &camera.viewport {
+            copy_pass.set_camera_viewport(viewport);
+        }
+
+        copy_pass.set_bind_group(0, copy_material_depth_bind_group, &[]);
+        copy_pass.set_render_pipeline(copy_material_depth_pipeline);
+        copy_pass.draw(0..3, 0..1);
+    }
+}

crates/bevy_pbr/src/meshlet/visibility_buffer_resolve.wgsl

@@ -0,0 +1,186 @@
+#define_import_path bevy_pbr::meshlet_visibility_buffer_resolve
+
+#import bevy_pbr::{
+    meshlet_bindings::{
+        meshlet_visibility_buffer,
+        meshlet_thread_meshlet_ids,
+        meshlets,
+        meshlet_vertex_ids,
+        meshlet_vertex_data,
+        meshlet_thread_instance_ids,
+        meshlet_instance_uniforms,
+        get_meshlet_index,
+        unpack_meshlet_vertex,
+    },
+    mesh_view_bindings::view,
+    mesh_functions::mesh_position_local_to_world,
+    mesh_types::MESH_FLAGS_SIGN_DETERMINANT_MODEL_3X3_BIT,
+    view_transformations::{position_world_to_clip, frag_coord_to_ndc},
+}
+#import bevy_render::maths::{affine3_to_square, mat2x4_f32_to_mat3x3_unpack}
+
+#ifdef PREPASS_FRAGMENT
+#ifdef MOTION_VECTOR_PREPASS
+#import bevy_pbr::{
+    prepass_bindings::previous_view_proj,
+    pbr_prepass_functions::calculate_motion_vector,
+}
+#endif
+#endif
+
+/// Functions to be used by materials for reading from a meshlet visibility buffer texture.
+
+#ifdef MESHLET_MESH_MATERIAL_PASS
+struct PartialDerivatives {
+    barycentrics: vec3<f32>,
+    ddx: vec3<f32>,
+    ddy: vec3<f32>,
+}
+
+// https://github.com/ConfettiFX/The-Forge/blob/2d453f376ef278f66f97cbaf36c0d12e4361e275/Examples_3/Visibility_Buffer/src/Shaders/FSL/visibilityBuffer_shade.frag.fsl#L83-L139
+fn compute_partial_derivatives(vertex_clip_positions: array<vec4<f32>, 3>, ndc_uv: vec2<f32>, screen_size: vec2<f32>) -> PartialDerivatives {
+    var result: PartialDerivatives;
+
+    let inv_w = 1.0 / vec3(vertex_clip_positions[0].w, vertex_clip_positions[1].w, vertex_clip_positions[2].w);
+    let ndc_0 = vertex_clip_positions[0].xy * inv_w[0];
+    let ndc_1 = vertex_clip_positions[1].xy * inv_w[1];
+    let ndc_2 = vertex_clip_positions[2].xy * inv_w[2];
+
+    let inv_det = 1.0 / determinant(mat2x2(ndc_2 - ndc_1, ndc_0 - ndc_1));
+    result.ddx = vec3(ndc_1.y - ndc_2.y, ndc_2.y - ndc_0.y, ndc_0.y - ndc_1.y) * inv_det * inv_w;
+    result.ddy = vec3(ndc_2.x - ndc_1.x, ndc_0.x - ndc_2.x, ndc_1.x - ndc_0.x) * inv_det * inv_w;
+
+    var ddx_sum = dot(result.ddx, vec3(1.0));
+    var ddy_sum = dot(result.ddy, vec3(1.0));
+
+    let delta_v = ndc_uv - ndc_0;
+    let interp_inv_w = inv_w.x + delta_v.x * ddx_sum + delta_v.y * ddy_sum;
+    let interp_w = 1.0 / interp_inv_w;
+
+    result.barycentrics = vec3(
+        interp_w * (delta_v.x * result.ddx.x + delta_v.y * result.ddy.x + inv_w.x),
+        interp_w * (delta_v.x * result.ddx.y + delta_v.y * result.ddy.y),
+        interp_w * (delta_v.x * result.ddx.z + delta_v.y * result.ddy.z),
+    );
+
+    result.ddx *= 2.0 / screen_size.x;
+    result.ddy *= 2.0 / screen_size.y;
+    ddx_sum *= 2.0 / screen_size.x;
+    ddy_sum *= 2.0 / screen_size.y;
+
+    let interp_ddx_w = 1.0 / (interp_inv_w + ddx_sum);
+    let interp_ddy_w = 1.0 / (interp_inv_w + ddy_sum);
+
+    result.ddx = interp_ddx_w * (result.barycentrics * interp_inv_w + result.ddx) - result.barycentrics;
+    result.ddy = interp_ddy_w * (result.barycentrics * interp_inv_w + result.ddy) - result.barycentrics;
+    return result;
+}
+
+struct VertexOutput {
+    position: vec4<f32>,
+    world_position: vec4<f32>,
+    world_normal: vec3<f32>,
+    uv: vec2<f32>,
+    ddx_uv: vec2<f32>,
+    ddy_uv: vec2<f32>,
+    world_tangent: vec4<f32>,
+    mesh_flags: u32,
+    meshlet_id: u32,
+#ifdef PREPASS_FRAGMENT
+#ifdef MOTION_VECTOR_PREPASS
+    motion_vector: vec2<f32>,
+#endif
+#endif
+}
+
+/// Load the visibility buffer texture and resolve it into a VertexOutput.
+fn resolve_vertex_output(frag_coord: vec4<f32>) -> VertexOutput {
+    let vbuffer = textureLoad(meshlet_visibility_buffer, vec2<i32>(frag_coord.xy), 0).r;
+    let cluster_id = vbuffer >> 8u;
+    let meshlet_id = meshlet_thread_meshlet_ids[cluster_id];
+    let meshlet = meshlets[meshlet_id];
+    let triangle_id = extractBits(vbuffer, 0u, 8u);
+    let index_ids = meshlet.start_index_id + vec3(triangle_id * 3u) + vec3(0u, 1u, 2u);
+    let indices = meshlet.start_vertex_id + vec3(get_meshlet_index(index_ids.x), get_meshlet_index(index_ids.y), get_meshlet_index(index_ids.z));
+    let vertex_ids = vec3(meshlet_vertex_ids[indices.x], meshlet_vertex_ids[indices.y], meshlet_vertex_ids[indices.z]);
+    let vertex_1 = unpack_meshlet_vertex(meshlet_vertex_data[vertex_ids.x]);
+    let vertex_2 = unpack_meshlet_vertex(meshlet_vertex_data[vertex_ids.y]);
+    let vertex_3 = unpack_meshlet_vertex(meshlet_vertex_data[vertex_ids.z]);
+
+    let instance_id = meshlet_thread_instance_ids[cluster_id];
+    let instance_uniform = meshlet_instance_uniforms[instance_id];
+    let model = affine3_to_square(instance_uniform.model);
+
+    let world_position_1 = mesh_position_local_to_world(model, vec4(vertex_1.position, 1.0));
+    let world_position_2 = mesh_position_local_to_world(model, vec4(vertex_2.position, 1.0));
+    let world_position_3 = mesh_position_local_to_world(model, vec4(vertex_3.position, 1.0));
+    let clip_position_1 = position_world_to_clip(world_position_1.xyz);
+    let clip_position_2 = position_world_to_clip(world_position_2.xyz);
+    let clip_position_3 = position_world_to_clip(world_position_3.xyz);
+    let frag_coord_ndc = frag_coord_to_ndc(frag_coord).xy;
+    let partial_derivatives = compute_partial_derivatives(
+        array(clip_position_1, clip_position_2, clip_position_3),
+        frag_coord_ndc,
+        view.viewport.zw,
+    );
+
+    let world_position = mat3x4(world_position_1, world_position_2, world_position_3) * partial_derivatives.barycentrics;
+    let vertex_normal = mat3x3(vertex_1.normal, vertex_2.normal, vertex_3.normal) * partial_derivatives.barycentrics;
+    let world_normal = normalize(
+        mat2x4_f32_to_mat3x3_unpack(
+            instance_uniform.inverse_transpose_model_a,
+            instance_uniform.inverse_transpose_model_b,
+        ) * vertex_normal
+    );
+    let uv = mat3x2(vertex_1.uv, vertex_2.uv, vertex_3.uv) * partial_derivatives.barycentrics;
+    let ddx_uv = mat3x2(vertex_1.uv, vertex_2.uv, vertex_3.uv) * partial_derivatives.ddx;
+    let ddy_uv = mat3x2(vertex_1.uv, vertex_2.uv, vertex_3.uv) * partial_derivatives.ddy;
+    let vertex_tangent = mat3x4(vertex_1.tangent, vertex_2.tangent, vertex_3.tangent) * partial_derivatives.barycentrics;
+    let world_tangent = vec4(
+        normalize(
+            mat3x3(
+                model[0].xyz,
+                model[1].xyz,
+                model[2].xyz
+            ) * vertex_tangent.xyz
+        ),
+        vertex_tangent.w * (f32(bool(instance_uniform.flags & MESH_FLAGS_SIGN_DETERMINANT_MODEL_3X3_BIT)) * 2.0 - 1.0)
+    );
+
+#ifdef PREPASS_FRAGMENT
+#ifdef MOTION_VECTOR_PREPASS
+    let previous_model = affine3_to_square(instance_uniform.previous_model);
+    let previous_world_position_1 = mesh_position_local_to_world(previous_model, vec4(vertex_1.position, 1.0));
+    let previous_world_position_2 = mesh_position_local_to_world(previous_model, vec4(vertex_2.position, 1.0));
+    let previous_world_position_3 = mesh_position_local_to_world(previous_model, vec4(vertex_3.position, 1.0));
+    let previous_clip_position_1 = previous_view_proj * vec4(previous_world_position_1.xyz, 1.0);
+    let previous_clip_position_2 = previous_view_proj * vec4(previous_world_position_2.xyz, 1.0);
+    let previous_clip_position_3 = previous_view_proj * vec4(previous_world_position_3.xyz, 1.0);
+    let previous_partial_derivatives = compute_partial_derivatives(
+        array(previous_clip_position_1, previous_clip_position_2, previous_clip_position_3),
+        frag_coord_ndc,
+        view.viewport.zw,
+    );
+    let previous_world_position = mat3x4(previous_world_position_1, previous_world_position_2, previous_world_position_3) * previous_partial_derivatives.barycentrics;
+    let motion_vector = calculate_motion_vector(world_position, previous_world_position);
+#endif
+#endif
+
+    return VertexOutput(
+        frag_coord,
+        world_position,
+        world_normal,
+        uv,
+        ddx_uv,
+        ddy_uv,
+        world_tangent,
+        instance_uniform.flags,
+        meshlet_id,
+#ifdef PREPASS_FRAGMENT
+#ifdef MOTION_VECTOR_PREPASS
+        motion_vector,
+#endif
+#endif
+    );
+}
+#endif

crates/bevy_pbr/src/meshlet/write_index_buffer.wgsl

@@ -0,0 +1,43 @@
+#import bevy_pbr::meshlet_bindings::{
+    meshlet_thread_meshlet_ids,
+    meshlets,
+    draw_indirect_args,
+    draw_index_buffer,
+    get_meshlet_occlusion,
+    get_meshlet_previous_occlusion,
+}
+
+var<workgroup> draw_index_buffer_start_workgroup: u32;
+
+/// This pass writes out a buffer of cluster + triangle IDs for the draw_indirect() call to rasterize each visible meshlet.
+
+@compute
+@workgroup_size(64, 1, 1) // 64 threads per workgroup, 1 workgroup per cluster, 1 thread per triangle
+fn write_index_buffer(@builtin(workgroup_id) workgroup_id: vec3<u32>, @builtin(num_workgroups) num_workgroups: vec3<u32>, @builtin(local_invocation_index) triangle_id: u32) {
+    // Calculate the cluster ID for this workgroup
+    let cluster_id = dot(workgroup_id, vec3(num_workgroups.x * num_workgroups.x, num_workgroups.x, 1u));
+    if cluster_id >= arrayLength(&meshlet_thread_meshlet_ids) { return; }
+
+    // If the meshlet was culled, then we don't need to draw it
+    if !get_meshlet_occlusion(cluster_id) { return; }
+
+    // If the meshlet was drawn in the first pass, and this is the second pass, then we don't need to draw it
+#ifdef MESHLET_SECOND_WRITE_INDEX_BUFFER_PASS
+    if get_meshlet_previous_occlusion(cluster_id) { return; }
+#endif
+
+    let meshlet_id = meshlet_thread_meshlet_ids[cluster_id];
+    let meshlet = meshlets[meshlet_id];
+
+    // Reserve space in the buffer for this meshlet's triangles, and broadcast the start of that slice to all threads
+    if triangle_id == 0u {
+        draw_index_buffer_start_workgroup = atomicAdd(&draw_indirect_args.vertex_count, meshlet.triangle_count * 3u);
+        draw_index_buffer_start_workgroup /= 3u;
+    }
+    workgroupBarrier();
+
+    // Each thread writes one triangle of the meshlet to the buffer slice reserved for the meshlet
+    if triangle_id < meshlet.triangle_count {
+        draw_index_buffer[draw_index_buffer_start_workgroup + triangle_id] = (cluster_id << 8u) | triangle_id;
+    }
+}

crates/bevy_pbr/src/pbr_material.rs

@@ -823,6 +823,21 @@ impl Material for StandardMaterial {
         PBR_SHADER_HANDLE.into()
     }
 
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_fragment_shader() -> ShaderRef {
+        Self::fragment_shader()
+    }
+
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_prepass_fragment_shader() -> ShaderRef {
+        Self::prepass_fragment_shader()
+    }
+
+    #[cfg(feature = "meshlet")]
+    fn meshlet_mesh_deferred_fragment_shader() -> ShaderRef {
+        Self::deferred_fragment_shader()
+    }
+
     fn specialize(
         _pipeline: &MaterialPipeline<Self>,
         descriptor: &mut RenderPipelineDescriptor,

crates/bevy_pbr/src/prepass/mod.rs

@@ -29,6 +29,10 @@ use bevy_render::{
 use bevy_transform::prelude::GlobalTransform;
 use bevy_utils::tracing::error;
 
+#[cfg(feature = "meshlet")]
+use crate::meshlet::{
+    prepare_material_meshlet_meshes_prepass, queue_material_meshlet_meshes, MeshletGpuScene,
+};
 use crate::*;
 
 use std::{hash::Hash, marker::PhantomData};
@@ -172,6 +176,15 @@ where
                     // queue_material_meshes only writes to `material_bind_group_id`, which `queue_prepass_material_meshes` doesn't read
                     .ambiguous_with(queue_material_meshes::<StandardMaterial>),
             );
+
+        #[cfg(feature = "meshlet")]
+        render_app.add_systems(
+            Render,
+            prepare_material_meshlet_meshes_prepass::<M>
+                .in_set(RenderSet::Queue)
+                .before(queue_material_meshlet_meshes::<M>)
+                .run_if(resource_exists::<MeshletGpuScene>),
+        );
     }
 }
 

crates/bevy_pbr/src/render/pbr.wgsl

@@ -16,11 +16,24 @@
 }
 #endif
 
+#ifdef MESHLET_MESH_MATERIAL_PASS
+#import bevy_pbr::meshlet_visibility_buffer_resolve::resolve_vertex_output
+#endif
+
 @fragment
 fn fragment(
+#ifdef MESHLET_MESH_MATERIAL_PASS
+    @builtin(position) frag_coord: vec4<f32>,
+#else
     in: VertexOutput,
     @builtin(front_facing) is_front: bool,
+#endif
 ) -> FragmentOutput {
+#ifdef MESHLET_MESH_MATERIAL_PASS
+    let in = resolve_vertex_output(frag_coord);
+    let is_front = true;
+#endif
+
     // generate a PbrInput struct from the StandardMaterial bindings
     var pbr_input = pbr_input_from_standard_material(in, is_front);
 

crates/bevy_pbr/src/render/pbr_fragment.wgsl

@@ -17,7 +17,9 @@
 #import bevy_pbr::gtao_utils::gtao_multibounce
 #endif
 
-#ifdef PREPASS_PIPELINE
+#ifdef MESHLET_MESH_MATERIAL_PASS
+#import bevy_pbr::meshlet_visibility_buffer_resolve::VertexOutput
+#else ifdef PREPASS_PIPELINE
 #import bevy_pbr::prepass_io::VertexOutput
 #else
 #import bevy_pbr::forward_io::VertexOutput
@@ -31,7 +33,12 @@ fn pbr_input_from_vertex_output(
 ) -> pbr_types::PbrInput {
     var pbr_input: pbr_types::PbrInput = pbr_types::pbr_input_new();
 
+#ifdef MESHLET_MESH_MATERIAL_PASS
+    pbr_input.flags = in.mesh_flags;
+#else
     pbr_input.flags = mesh[in.instance_index].flags;
+#endif
+
     pbr_input.is_orthographic = view.projection[3].w == 1.0;
     pbr_input.V = pbr_functions::calculate_view(in.world_position, pbr_input.is_orthographic);
     pbr_input.frag_coord = in.position;
@@ -98,7 +105,11 @@ fn pbr_input_from_standard_material(
 #endif // VERTEX_TANGENTS
 
     if ((pbr_bindings::material.flags & pbr_types::STANDARD_MATERIAL_FLAGS_BASE_COLOR_TEXTURE_BIT) != 0u) {
+#ifdef MESHLET_MESH_MATERIAL_PASS
+        pbr_input.material.base_color *= textureSampleGrad(pbr_bindings::base_color_texture, pbr_bindings::base_color_sampler, uv, in.ddx_uv, in.ddy_uv);
+#else
         pbr_input.material.base_color *= textureSampleBias(pbr_bindings::base_color_texture, pbr_bindings::base_color_sampler, uv, view.mip_bias);
+#endif
     }
 #endif // VERTEX_UVS
 
@@ -117,7 +128,11 @@ fn pbr_input_from_standard_material(
         var emissive: vec4<f32> = pbr_bindings::material.emissive;
 #ifdef VERTEX_UVS
         if ((pbr_bindings::material.flags & pbr_types::STANDARD_MATERIAL_FLAGS_EMISSIVE_TEXTURE_BIT) != 0u) {
+#ifdef MESHLET_MESH_MATERIAL_PASS
+            emissive = vec4<f32>(emissive.rgb * textureSampleGrad(pbr_bindings::emissive_texture, pbr_bindings::emissive_sampler, uv, in.ddx_uv, in.ddy_uv).rgb, 1.0);
+#else
             emissive = vec4<f32>(emissive.rgb * textureSampleBias(pbr_bindings::emissive_texture, pbr_bindings::emissive_sampler, uv, view.mip_bias).rgb, 1.0);
+#endif
         }
 #endif
         pbr_input.material.emissive = emissive;
@@ -128,7 +143,11 @@ fn pbr_input_from_standard_material(
         let roughness = lighting::perceptualRoughnessToRoughness(perceptual_roughness);
 #ifdef VERTEX_UVS
         if ((pbr_bindings::material.flags & pbr_types::STANDARD_MATERIAL_FLAGS_METALLIC_ROUGHNESS_TEXTURE_BIT) != 0u) {
+#ifdef MESHLET_MESH_MATERIAL_PASS
+            let metallic_roughness = textureSampleGrad(pbr_bindings::metallic_roughness_texture, pbr_bindings::metallic_roughness_sampler, uv, in.ddx_uv, in.ddy_uv);
+#else
             let metallic_roughness = textureSampleBias(pbr_bindings::metallic_roughness_texture, pbr_bindings::metallic_roughness_sampler, uv, view.mip_bias);
+#endif
             // Sampling from GLTF standard channels for now
             metallic *= metallic_roughness.b;
             perceptual_roughness *= metallic_roughness.g;
@@ -140,7 +159,11 @@ fn pbr_input_from_standard_material(
         var specular_transmission: f32 = pbr_bindings::material.specular_transmission;
 #ifdef PBR_TRANSMISSION_TEXTURES_SUPPORTED
         if ((pbr_bindings::material.flags & pbr_types::STANDARD_MATERIAL_FLAGS_SPECULAR_TRANSMISSION_TEXTURE_BIT) != 0u) {
-            specular_transmission *= textureSample(pbr_bindings::specular_transmission_texture, pbr_bindings::specular_transmission_sampler, uv).r;
+#ifdef MESHLET_MESH_MATERIAL_PASS
+            specular_transmission *= textureSampleGrad(pbr_bindings::specular_transmission_texture, pbr_bindings::specular_transmission_sampler, uv, in.ddx_uv, in.ddy_uv).r;
+#else
+            specular_transmission *= textureSampleBias(pbr_bindings::specular_transmission_texture, pbr_bindings::specular_transmission_sampler, uv, view.mip_bias).r;
+#endif
         }
 #endif
         pbr_input.material.specular_transmission = specular_transmission;
@@ -148,19 +171,30 @@ fn pbr_input_from_standard_material(
         var thickness: f32 = pbr_bindings::material.thickness;
 #ifdef PBR_TRANSMISSION_TEXTURES_SUPPORTED
         if ((pbr_bindings::material.flags & pbr_types::STANDARD_MATERIAL_FLAGS_THICKNESS_TEXTURE_BIT) != 0u) {
-            thickness *= textureSample(pbr_bindings::thickness_texture, pbr_bindings::thickness_sampler, uv).g;
+#ifdef MESHLET_MESH_MATERIAL_PASS
+            thickness *= textureSampleGrad(pbr_bindings::thickness_texture, pbr_bindings::thickness_sampler, uv, in.ddx_uv, in.ddy_uv).g;
+#else
+            thickness *= textureSampleBias(pbr_bindings::thickness_texture, pbr_bindings::thickness_sampler, uv, view.mip_bias).g;
+#endif
         }
 #endif
         // scale thickness, accounting for non-uniform scaling (e.g. a “squished” mesh)
+        // TODO: Meshlet support
+#ifndef MESHLET_MESH_MATERIAL_PASS
         thickness *= length(
             (transpose(mesh[in.instance_index].model) * vec4(pbr_input.N, 0.0)).xyz
         );
+#endif
         pbr_input.material.thickness = thickness;
 
         var diffuse_transmission = pbr_bindings::material.diffuse_transmission;
 #ifdef PBR_TRANSMISSION_TEXTURES_SUPPORTED
         if ((pbr_bindings::material.flags & pbr_types::STANDARD_MATERIAL_FLAGS_DIFFUSE_TRANSMISSION_TEXTURE_BIT) != 0u) {
-            diffuse_transmission *= textureSample(pbr_bindings::diffuse_transmission_texture, pbr_bindings::diffuse_transmission_sampler, uv).a;
+#ifdef MESHLET_MESH_MATERIAL_PASS
+            diffuse_transmission *= textureSampleGrad(pbr_bindings::diffuse_transmission_texture, pbr_bindings::diffuse_transmission_sampler, uv, in.ddx_uv, in.ddy_uv).a;
+#else
+            diffuse_transmission *= textureSampleBias(pbr_bindings::diffuse_transmission_texture, pbr_bindings::diffuse_transmission_sampler, uv, view.mip_bias).a;
+#endif
         }
 #endif
         pbr_input.material.diffuse_transmission = diffuse_transmission;
@@ -169,7 +203,11 @@ fn pbr_input_from_standard_material(
         var specular_occlusion: f32 = 1.0;
 #ifdef VERTEX_UVS
         if ((pbr_bindings::material.flags & pbr_types::STANDARD_MATERIAL_FLAGS_OCCLUSION_TEXTURE_BIT) != 0u) {
+#ifdef MESHLET_MESH_MATERIAL_PASS
+            diffuse_occlusion = vec3(textureSampleGrad(pbr_bindings::occlusion_texture, pbr_bindings::occlusion_sampler, uv, in.ddx_uv, in.ddy_uv).r);
+#else
             diffuse_occlusion = vec3(textureSampleBias(pbr_bindings::occlusion_texture, pbr_bindings::occlusion_sampler, uv, view.mip_bias).r);
+#endif
         }
 #endif
 #ifdef SCREEN_SPACE_AMBIENT_OCCLUSION
@@ -199,9 +237,14 @@ fn pbr_input_from_standard_material(
             uv,
 #endif
             view.mip_bias,
+#ifdef MESHLET_MESH_MATERIAL_PASS
+            in.ddx_uv,
+            in.ddy_uv,
+#endif
         );
 #endif
 
+// TODO: Meshlet support
 #ifdef LIGHTMAP
         pbr_input.lightmap_light = lightmap(
             in.uv_b,

crates/bevy_pbr/src/render/pbr_functions.wgsl

@@ -74,6 +74,10 @@ fn apply_normal_mapping(
     uv: vec2<f32>,
 #endif
     mip_bias: f32,
+#ifdef MESHLET_MESH_MATERIAL_PASS
+    ddx_uv: vec2<f32>,
+    ddy_uv: vec2<f32>,
+#endif
 ) -> vec3<f32> {
     // NOTE: The mikktspace method of normal mapping explicitly requires that the world normal NOT
     // be re-normalized in the fragment shader. This is primarily to match the way mikktspace
@@ -98,7 +102,11 @@ fn apply_normal_mapping(
 #ifdef VERTEX_UVS
 #ifdef STANDARD_MATERIAL_NORMAL_MAP
     // Nt is the tangent-space normal.
+#ifdef MESHLET_MESH_MATERIAL_PASS
+    var Nt = textureSampleGrad(pbr_bindings::normal_map_texture, pbr_bindings::normal_map_sampler, uv, ddx_uv, ddy_uv).rgb;
+#else
     var Nt = textureSampleBias(pbr_bindings::normal_map_texture, pbr_bindings::normal_map_sampler, uv, mip_bias).rgb;
+#endif
     if (standard_material_flags & pbr_types::STANDARD_MATERIAL_FLAGS_TWO_COMPONENT_NORMAL_MAP) != 0u {
         // Only use the xy components and derive z for 2-component normal maps.
         Nt = vec3<f32>(Nt.rg * 2.0 - 1.0, 0.0);

crates/bevy_pbr/src/render/pbr_prepass.wgsl

@@ -7,13 +7,26 @@
     mesh_view_bindings::view,
 }
 
+#ifdef MESHLET_MESH_MATERIAL_PASS
+#import bevy_pbr::meshlet_visibility_buffer_resolve::resolve_vertex_output
+#endif
+
 #ifdef PREPASS_FRAGMENT
 @fragment
 fn fragment(
+#ifdef MESHLET_MESH_MATERIAL_PASS
+    @builtin(position) frag_coord: vec4<f32>,
+#else
     in: prepass_io::VertexOutput,
     @builtin(front_facing) is_front: bool,
+#endif
 ) -> prepass_io::FragmentOutput {
+#ifdef MESHLET_MESH_MATERIAL_PASS
+    let in = resolve_vertex_output(frag_coord);
+    let is_front = true;
+#else
     pbr_prepass_functions::prepass_alpha_discard(in);
+#endif
 
     var out: prepass_io::FragmentOutput;
 
@@ -46,6 +59,10 @@ fn fragment(
             in.uv,
 #endif // VERTEX_UVS
             view.mip_bias,
+#ifdef MESHLET_MESH_MATERIAL_PASS
+            in.ddx_uv,
+            in.ddy_uv,
+#endif // MESHLET_MESH_MATERIAL_PASS
         );
 
         out.normal = vec4(normal * 0.5 + vec3(0.5), 1.0);
@@ -55,7 +72,11 @@ fn fragment(
 #endif // NORMAL_PREPASS
 
 #ifdef MOTION_VECTOR_PREPASS
+#ifdef MESHLET_MESH_MATERIAL_PASS
+    out.motion_vector = in.motion_vector;
+#else
     out.motion_vector = pbr_prepass_functions::calculate_motion_vector(in.world_position, in.previous_world_position);
+#endif
 #endif
 
     return out;

crates/bevy_render/src/mesh/mesh/mod.rs

@@ -369,6 +369,14 @@ impl Mesh {
         self
     }
 
+    /// Returns the size of a vertex in bytes.
+    pub fn get_vertex_size(&self) -> u64 {
+        self.attributes
+            .values()
+            .map(|data| data.attribute.format.get_size())
+            .sum()
+    }
+
     /// Computes and returns the index data of the mesh as bytes.
     /// This is used to transform the index data into a GPU friendly format.
     pub fn get_index_buffer_bytes(&self) -> Option<&[u8]> {

crates/bevy_render/src/mesh/mod.rs

@@ -57,7 +57,7 @@ impl MeshVertexBufferLayouts {
     /// Inserts a new mesh vertex buffer layout in the store and returns a
     /// reference to it, reusing the existing reference if this mesh vertex
     /// buffer layout was already in the store.
-    pub(crate) fn insert(&mut self, layout: MeshVertexBufferLayout) -> MeshVertexBufferLayoutRef {
+    pub fn insert(&mut self, layout: MeshVertexBufferLayout) -> MeshVertexBufferLayoutRef {
         // Because the special `PartialEq` and `Hash` implementations that
         // compare by pointer are on `MeshVertexBufferLayoutRef`, not on
         // `Arc<MeshVertexBufferLayout>`, this compares the mesh vertex buffer

docs/cargo_features.md

@@ -64,6 +64,8 @@ The default feature set enables most of the expected features of a game engine,
 |glam_assert|Enable assertions to check the validity of parameters passed to glam|
 |ios_simulator|Enable support for the ios_simulator by downgrading some rendering capabilities|
 |jpeg|JPEG image format support|
+|meshlet|Enables the meshlet renderer for dense high-poly scenes (experimental)|
+|meshlet_processor|Enables processing meshes into meshlet meshes for bevy_pbr|
 |minimp3|MP3 audio format support (through minimp3)|
 |mp3|MP3 audio format support|
 |pbr_transmission_textures|Enable support for transmission-related textures in the `StandardMaterial`, at the risk of blowing past the global, per-shader texture limit on older/lower-end GPUs|

examples/3d/meshlet.rs

@@ -0,0 +1,180 @@
+//! Meshlet rendering for dense high-poly scenes (experimental).
+
+#[path = "../helpers/camera_controller.rs"]
+mod camera_controller;
+
+use bevy::{
+    pbr::{
+        experimental::meshlet::{MaterialMeshletMeshBundle, MeshletMesh, MeshletPlugin},
+        CascadeShadowConfigBuilder, DirectionalLightShadowMap,
+    },
+    prelude::*,
+    render::render_resource::AsBindGroup,
+};
+use camera_controller::{CameraController, CameraControllerPlugin};
+use std::f32::consts::PI;
+
+// Note: This example showcases the meshlet API, but is not the type of scene that would benefit from using meshlets.
+
+fn main() {
+    App::new()
+        .insert_resource(DirectionalLightShadowMap { size: 4096 })
+        .add_plugins((
+            DefaultPlugins,
+            MeshletPlugin,
+            MaterialPlugin::<MeshletDebugMaterial>::default(),
+            CameraControllerPlugin,
+        ))
+        .add_systems(Startup, setup)
+        .add_systems(Update, draw_bounding_spheres)
+        .run();
+}
+
+fn setup(
+    mut commands: Commands,
+    asset_server: Res<AssetServer>,
+    mut standard_materials: ResMut<Assets<StandardMaterial>>,
+    mut debug_materials: ResMut<Assets<MeshletDebugMaterial>>,
+    mut meshes: ResMut<Assets<Mesh>>,
+) {
+    info!("\nMeshlet Controls:\n    Space - Toggle bounding spheres");
+
+    commands.spawn((
+        Camera3dBundle {
+            transform: Transform::from_translation(Vec3::new(1.8, 0.4, -0.1))
+                .looking_at(Vec3::ZERO, Vec3::Y),
+            ..default()
+        },
+        EnvironmentMapLight {
+            diffuse_map: asset_server.load("environment_maps/pisa_diffuse_rgb9e5_zstd.ktx2"),
+            specular_map: asset_server.load("environment_maps/pisa_specular_rgb9e5_zstd.ktx2"),
+            intensity: 150.0,
+        },
+        CameraController::default(),
+    ));
+
+    commands.spawn(DirectionalLightBundle {
+        directional_light: DirectionalLight {
+            illuminance: light_consts::lux::FULL_DAYLIGHT,
+            shadows_enabled: true,
+            ..default()
+        },
+        cascade_shadow_config: CascadeShadowConfigBuilder {
+            num_cascades: 1,
+            maximum_distance: 5.0,
+            ..default()
+        }
+        .build(),
+        transform: Transform::from_rotation(Quat::from_euler(
+            EulerRot::ZYX,
+            0.0,
+            PI * -0.15,
+            PI * -0.15,
+        )),
+        ..default()
+    });
+
+    // A custom file format storing a [`bevy_render::mesh::Mesh`]
+    // that has been converted to a [`bevy_pbr::meshlet::MeshletMesh`]
+    // using [`bevy_pbr::meshlet::MeshletMesh::from_mesh`], which is
+    // a function only available when the `meshlet_processor` cargo feature is enabled.
+    let meshlet_mesh_handle = asset_server.load("models/bunny.meshlet_mesh");
+    let debug_material = debug_materials.add(MeshletDebugMaterial::default());
+
+    for x in -2..=2 {
+        commands.spawn(MaterialMeshletMeshBundle {
+            meshlet_mesh: meshlet_mesh_handle.clone(),
+            material: standard_materials.add(StandardMaterial {
+                base_color: match x {
+                    -2 => Srgba::hex("#dc2626").unwrap().into(),
+                    -1 => Srgba::hex("#ea580c").unwrap().into(),
+                    0 => Srgba::hex("#facc15").unwrap().into(),
+                    1 => Srgba::hex("#16a34a").unwrap().into(),
+                    2 => Srgba::hex("#0284c7").unwrap().into(),
+                    _ => unreachable!(),
+                },
+                perceptual_roughness: (x + 2) as f32 / 4.0,
+                ..default()
+            }),
+            transform: Transform::default()
+                .with_scale(Vec3::splat(0.2))
+                .with_translation(Vec3::new(x as f32 / 2.0, 0.0, -0.3)),
+            ..default()
+        });
+    }
+    for x in -2..=2 {
+        commands.spawn(MaterialMeshletMeshBundle {
+            meshlet_mesh: meshlet_mesh_handle.clone(),
+            material: debug_material.clone(),
+            transform: Transform::default()
+                .with_scale(Vec3::splat(0.2))
+                .with_rotation(Quat::from_rotation_y(PI))
+                .with_translation(Vec3::new(x as f32 / 2.0, 0.0, 0.3)),
+            ..default()
+        });
+    }
+
+    commands.spawn(PbrBundle {
+        mesh: meshes.add(Plane3d::default().mesh().size(5.0, 5.0)),
+        material: standard_materials.add(StandardMaterial {
+            base_color: Color::WHITE,
+            perceptual_roughness: 1.0,
+            ..default()
+        }),
+        ..default()
+    });
+}
+
+#[allow(clippy::too_many_arguments)]
+fn draw_bounding_spheres(
+    query: Query<(&Handle<MeshletMesh>, &Transform), With<Handle<MeshletDebugMaterial>>>,
+    debug: Query<&MeshletBoundingSpheresDebug>,
+    camera: Query<&Transform, With<Camera>>,
+    mut commands: Commands,
+    meshlets: Res<Assets<MeshletMesh>>,
+    mut gizmos: Gizmos,
+    keys: Res<ButtonInput<KeyCode>>,
+    mut should_draw: Local<bool>,
+) {
+    if keys.just_pressed(KeyCode::Space) {
+        *should_draw = !*should_draw;
+    }
+
+    match debug.get_single() {
+        Ok(meshlet_debug) if *should_draw => {
+            let camera_pos = camera.single().translation;
+            for circle in &meshlet_debug.circles {
+                gizmos.circle(
+                    circle.0,
+                    Dir3::new(camera_pos - circle.0).unwrap(),
+                    circle.1,
+                    Color::BLACK,
+                );
+            }
+        }
+        Err(_) => {
+            if let Some((handle, transform)) = query.iter().last() {
+                if let Some(meshlets) = meshlets.get(handle) {
+                    let mut circles = Vec::new();
+                    for bounding_sphere in meshlets.meshlet_bounding_spheres.iter() {
+                        let center = transform.transform_point(bounding_sphere.center);
+                        circles.push((center, transform.scale.x * bounding_sphere.radius));
+                    }
+                    commands.spawn(MeshletBoundingSpheresDebug { circles });
+                }
+            }
+        }
+        _ => {}
+    }
+}
+
+#[derive(Component)]
+struct MeshletBoundingSpheresDebug {
+    circles: Vec<(Vec3, f32)>,
+}
+
+#[derive(Asset, TypePath, AsBindGroup, Clone, Default)]
+struct MeshletDebugMaterial {
+    _dummy: (),
+}
+impl Material for MeshletDebugMaterial {}

examples/README.md

@@ -136,6 +136,7 @@ Example | Description
 [Lightmaps](../examples/3d/lightmaps.rs) | Rendering a scene with baked lightmaps
 [Lines](../examples/3d/lines.rs) | Create a custom material to draw 3d lines
 [Load glTF](../examples/3d/load_gltf.rs) | Loads and renders a glTF file as a scene
+[Meshlet](../examples/3d/meshlet.rs) | Meshlet rendering for dense high-poly scenes (experimental)
 [Orthographic View](../examples/3d/orthographic.rs) | Shows how to create a 3D orthographic view (for isometric-look in games or CAD applications)
 [Parallax Mapping](../examples/3d/parallax_mapping.rs) | Demonstrates use of a normal map and depth map for parallax mapping
 [Parenting](../examples/3d/parenting.rs) | Demonstrates parent->child relationships and relative transformations