initial instancing support

Cargo.toml

@@ -8,7 +8,7 @@ edition = "2018"
 legion = { git = "https://github.com/TomGillen/legion.git", rev = "8628b227bcbe57582fffb5e80e73c634ec4eebd9" }
 legion_transform = { path = "src/transform" }
 wgpu = { git = "https://github.com/gfx-rs/wgpu-rs.git", rev = "44fa1bc2fa208fa92f80944253e0da56cb7ac1fe"}
-glam = "0.8.3"
+glam = "0.8.4"
 winit = "0.20.0-alpha4"
 glsl-to-spirv = "0.1"
 zerocopy = "0.2"

examples/simple.rs

@@ -179,19 +179,19 @@ fn main() {
     scheduler.add_system(ApplicationStage::Update, build_wander_system());
     scheduler.add_system(ApplicationStage::Update, build_navigate_system());
     scheduler.add_system(ApplicationStage::Update, build_move_system());
-    scheduler.add_system(ApplicationStage::Update, build_light_rotator_system());
-    scheduler.add_system(ApplicationStage::Update, build_spawner_system(&mut world));
+    // scheduler.add_system(ApplicationStage::Update, build_light_rotator_system());
+    // scheduler.add_system(ApplicationStage::Update, build_spawner_system(&mut world));
     scheduler.add_system(ApplicationStage::Update, build_print_status_system());
 
-    world.insert((), vec![
-        // plane
-        (
-            Material::new(math::vec4(0.1, 0.2, 0.1, 1.0)),
-            plane_handle.clone(),
-            LocalToWorld::identity(),
-            Translation::new(0.0, 0.0, 0.0)
-        ),
-    ]);
+    // world.insert((), vec![
+    //     // plane
+    //     (
+    //         Material::new(math::vec4(0.1, 0.2, 0.1, 1.0)),
+    //         plane_handle.clone(),
+    //         LocalToWorld::identity(),
+    //         Translation::new(0.0, 0.0, 0.0)
+    //     ),
+    // ]);
 
     let x = *world.insert((), vec![
         // lights
@@ -232,17 +232,17 @@ fn main() {
         // ),
     ]).first().unwrap();
 
-    world.insert((), vec![
-        (
-            Material::new(math::vec4(1.0, 1.0, 1.0, 1.0)),
-            _cube_handle.clone(),
-            LocalToWorld::identity(),
-            Translation::new(0.0, 0.0, 3.0),
-            Scale(1.0),
-            Parent(x),
-            LocalToParent::identity(),
-        )
-    ]);
+    // world.insert((), vec![
+    //     (
+    //         Material::new(math::vec4(1.0, 1.0, 1.0, 1.0)),
+    //         _cube_handle.clone(),
+    //         LocalToWorld::identity(),
+    //         Translation::new(0.0, 0.0, 3.0),
+    //         Scale(1.0),
+    //         Parent(x),
+    //         LocalToParent::identity(),
+    //     )
+    // ]);
 
     world.insert((), vec![
         
@@ -262,6 +262,12 @@ fn main() {
         )
     ]);
 
+    let mut rng = StdRng::from_entropy();
+    for _ in 0 .. 70000 {
+        create_person(&mut world, _cube_handle.clone(),
+            Translation::new(rng.gen_range(-50.0, 50.0), 0.0, rng.gen_range(-50.0, 50.0)));
+    }
+
     Application::run(universe, world, scheduler);
 }
 
@@ -281,6 +287,7 @@ fn spawn_person(command_buffer: &mut CommandBuffer, mesh_handle: Handle<Mesh>) {
             Velocity {
                 value: math::vec3(0.0, 0.0, 0.0),
             },
+            Instanced,
             Material::new(math::vec4(0.5, 0.3, 0.3, 1.0) * random::<f32>()),
             mesh_handle.clone(),
             LocalToWorld::identity(),
@@ -288,3 +295,28 @@ fn spawn_person(command_buffer: &mut CommandBuffer, mesh_handle: Handle<Mesh>) {
         ),
     ]);
 }
+
+fn create_person(world: &mut World, mesh_handle: Handle<Mesh>, translation: Translation) {
+    world.insert((), vec![
+        (
+            Person{},
+            Wander {
+                duration_bounds: math::vec2(3.0, 10.0),
+                distance_bounds: math::vec2(-50.0, 50.0),
+                elapsed: 0.0,
+                duration: 0.0,
+            },
+            NavigationPoint {
+                target: math::vec3(0.0, 0.0, 0.0),
+            },
+            Velocity {
+                value: math::vec3(0.0, 0.0, 0.0),
+            },
+            Instanced,
+            Material::new(math::vec4(0.5, 0.3, 0.3, 1.0) * random::<f32>()),
+            mesh_handle,
+            LocalToWorld::identity(),
+            translation
+        ),
+    ]);
+} 

src/application.rs

@@ -51,10 +51,12 @@ impl Application {
 
         // let shadow_pass = ShadowPass::new(&mut self.device, &mut self.world, &self.render_resources, vertex_buffer_descriptor.clone());
         // let forward_shadow_pass = ForwardShadowPass::new(&mut self.device, &self.world, &self.render_resources, &shadow_pass, vertex_buffer_descriptor.clone(), &self.swap_chain_descriptor);
-        let forward_pass = ForwardPass::new(&mut self.device, &self.world, &self.render_resources, vertex_buffer_descriptor, &self.swap_chain_descriptor);
+        // let forward_pass = ForwardPass::new(&mut self.device, &self.world, &self.render_resources, vertex_buffer_descriptor.clone(), &self.swap_chain_descriptor);
+        let forward_instanced_pass = ForwardInstancedPass::new(&mut self.device, &self.world, &self.render_resources, vertex_buffer_descriptor, &self.swap_chain_descriptor);
         // self.render_passes.push(Box::new(shadow_pass));
         // self.render_passes.push(Box::new(forward_shadow_pass));
-        self.render_passes.push(Box::new(forward_pass));
+        // self.render_passes.push(Box::new(forward_pass));
+        self.render_passes.push(Box::new(forward_instanced_pass));
     }
 
     fn update(&mut self) {
@@ -113,7 +115,8 @@ impl Application {
         let mut encoder =
             self.device.create_command_encoder(&wgpu::CommandEncoderDescriptor { todo: 0 });
 
-        let mut entities = <(Write<Material>, Read<LocalToWorld>)>::query();
+        let mut entities = <(Write<Material>, Read<LocalToWorld>)>::query()
+            .filter(!component::<Instanced>());
         let entities_count = entities.iter(&mut self.world).count();
         let size = mem::size_of::<MaterialUniforms>();
         let temp_buf_data = self.device
@@ -133,7 +136,7 @@ impl Application {
 
         self.render_resources.update_lights(&self.device, &mut encoder, &mut self.world);
 
-        for mut material in <Write<Material>>::query().iter(&mut self.world) {
+        for mut material in <Write<Material>>::query().filter(!component::<Instanced>()).iter(&mut self.world) {
             if let None = material.bind_group {
                 let material_uniform_size = mem::size_of::<MaterialUniforms>() as wgpu::BufferAddress;
                 let uniform_buf = self.device.create_buffer(&wgpu::BufferDescriptor {

src/render/forward/mod.rs

@@ -20,7 +20,9 @@ pub struct ForwardPass {
 
 impl Pass for ForwardPass {
     fn render(&mut self, device: &Device, frame: &SwapChainOutput, encoder: &mut CommandEncoder, world: &mut World, _: &RenderResources) { 
-        let mut mesh_query = <(Read<Material>, Read<Handle<Mesh>>)>::query();
+        let mut mesh_query =
+            <(Read<Material>, Read<Handle<Mesh>>)>::query()
+            .filter(!component::<Instanced>());
         let mut pass = encoder.begin_render_pass(&wgpu::RenderPassDescriptor {
             color_attachments: &[wgpu::RenderPassColorAttachmentDescriptor {
                 attachment: &frame.view,

src/render/forward_instanced/forward_instanced.frag

@@ -0,0 +1,40 @@
+#version 450
+
+const int MAX_LIGHTS = 10;
+
+layout(location = 0) in vec3 v_Normal;
+layout(location = 1) in vec4 v_Position;
+layout(location = 2) in vec4 v_Color;
+
+layout(location = 0) out vec4 o_Target;
+
+struct Light {
+    mat4 proj;
+    vec4 pos;
+    vec4 color;
+};
+
+layout(set = 0, binding = 0) uniform Globals {
+    mat4 u_ViewProj;
+    uvec4 u_NumLights;
+};
+layout(set = 0, binding = 1) uniform Lights {
+    Light u_Lights[MAX_LIGHTS];
+};
+
+void main() {
+    vec3 normal = normalize(v_Normal);
+    vec3 ambient = vec3(0.05, 0.05, 0.05);
+    // accumulate color
+    vec3 color = ambient;
+    for (int i=0; i<int(u_NumLights.x) && i<MAX_LIGHTS; ++i) {
+        Light light = u_Lights[i];
+        // compute Lambertian diffuse term
+        vec3 light_dir = normalize(light.pos.xyz - v_Position.xyz);
+        float diffuse = max(0.0, dot(normal, light_dir));
+        // add light contribution
+        color += diffuse * light.color.xyz;
+    }
+    // multiply the light by material color
+    o_Target = vec4(color, 1.0) * v_Color;
+}

src/render/forward_instanced/forward_instanced.vert

@@ -0,0 +1,26 @@
+#version 450
+
+// vertex attributes
+layout(location = 0) in vec4 a_Pos;
+layout(location = 1) in vec4 a_Normal;
+
+// Instanced attributes
+layout (location = 2) in vec3 a_instancePos;
+layout (location = 3) in vec4 a_instanceColor;
+
+
+layout(location = 0) out vec3 v_Normal;
+layout(location = 1) out vec4 v_Position;
+layout(location = 2) out vec4 v_Color;
+
+layout(set = 0, binding = 0) uniform Globals {
+    mat4 u_ViewProj;
+    uvec4 u_NumLights;
+};
+
+void main() {
+    v_Normal = vec3(a_Normal.xyz);
+    v_Position = vec4(a_Pos) + vec4(a_instancePos, 1.0);
+    v_Color = a_instanceColor;
+    gl_Position = u_ViewProj * v_Position;
+}

src/render/forward_instanced/mod.rs

@@ -0,0 +1,267 @@
+use crate::{render::*, asset::*, render::mesh::*, math, LocalToWorld};
+use legion::prelude::*;
+use std::mem;
+use zerocopy::AsBytes;
+use wgpu::{Buffer, CommandEncoder, Device, VertexBufferDescriptor, SwapChainDescriptor, SwapChainOutput};
+
+
+pub struct InstanceBufferInfo {
+    pub buffer: wgpu::Buffer,
+    pub instance_count: usize,
+    pub mesh_id: usize,
+}
+
+pub struct ForwardInstancedPass {
+    pub pipeline: wgpu::RenderPipeline,
+    pub bind_group: wgpu::BindGroup,
+    pub forward_uniform_buffer: wgpu::Buffer,
+    pub depth_texture: wgpu::TextureView,
+    pub instance_buffer_infos: Vec<InstanceBufferInfo>,
+}
+
+impl Pass for ForwardInstancedPass {
+    fn render(&mut self, device: &Device, frame: &SwapChainOutput, encoder: &mut CommandEncoder, world: &mut World, _: &RenderResources) { 
+        self.instance_buffer_infos = ForwardInstancedPass::create_instance_buffer_infos(device, world);
+        let mut pass = encoder.begin_render_pass(&wgpu::RenderPassDescriptor {
+            color_attachments: &[wgpu::RenderPassColorAttachmentDescriptor {
+                attachment: &frame.view,
+                resolve_target: None,
+                load_op: wgpu::LoadOp::Clear,
+                store_op: wgpu::StoreOp::Store,
+                clear_color: wgpu::Color {
+                    r: 0.3,
+                    g: 0.4,
+                    b: 0.5,
+                    a: 1.0,
+                },
+            }],
+            depth_stencil_attachment: Some(wgpu::RenderPassDepthStencilAttachmentDescriptor {
+                attachment: &self.depth_texture,
+                depth_load_op: wgpu::LoadOp::Clear,
+                depth_store_op: wgpu::StoreOp::Store,
+                stencil_load_op: wgpu::LoadOp::Clear,
+                stencil_store_op: wgpu::StoreOp::Store,
+                clear_depth: 1.0,
+                clear_stencil: 0,
+            }),
+        });
+        pass.set_pipeline(&self.pipeline);
+        pass.set_bind_group(0, &self.bind_group, &[]);
+
+        let mut mesh_storage = world.resources.get_mut::<AssetStorage<Mesh, MeshType>>().unwrap();
+        for instance_buffer_info in self.instance_buffer_infos.iter() {
+            if let Some(mesh_asset) = mesh_storage.get(instance_buffer_info.mesh_id) {
+                mesh_asset.setup_buffers(device);
+                pass.set_index_buffer(mesh_asset.index_buffer.as_ref().unwrap(), 0);
+                pass.set_vertex_buffers(0, &[(&mesh_asset.vertex_buffer.as_ref().unwrap(), 0)]);
+                pass.set_vertex_buffers(1, &[(&instance_buffer_info.buffer, 0)]);
+                pass.draw_indexed(0 .. mesh_asset.indices.len() as u32, 0, 0 .. instance_buffer_info.instance_count as u32);
+            };
+        }
+    }
+    
+    fn resize(&mut self, device: &Device, frame: &SwapChainDescriptor) {
+        self.depth_texture = Self::get_depth_texture(device, frame);
+    }
+
+    fn get_camera_uniform_buffer(&self) -> Option<&Buffer> { 
+        Some(&self.forward_uniform_buffer)
+    }
+}
+
+impl ForwardInstancedPass {
+    pub const DEPTH_FORMAT: wgpu::TextureFormat = wgpu::TextureFormat::Depth32Float;
+ 
+    pub fn new(device: &Device, world: &World, render_resources: &RenderResources, vertex_buffer_descriptor: VertexBufferDescriptor, swap_chain_descriptor: &SwapChainDescriptor) -> Self {
+        let vs_bytes = shader::load_glsl(
+            include_str!("forward_instanced.vert"),
+            shader::ShaderStage::Vertex,
+        );
+        let fs_bytes = shader::load_glsl(
+            include_str!("forward_instanced.frag"),
+            shader::ShaderStage::Fragment,
+        );
+
+        let bind_group_layout =
+        device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
+            bindings: &[
+                wgpu::BindGroupLayoutBinding {
+                    binding: 0, // global
+                    visibility: wgpu::ShaderStage::VERTEX | wgpu::ShaderStage::FRAGMENT,
+                    ty: wgpu::BindingType::UniformBuffer { dynamic: false },
+                },
+                wgpu::BindGroupLayoutBinding {
+                    binding: 1, // lights
+                    visibility: wgpu::ShaderStage::VERTEX | wgpu::ShaderStage::FRAGMENT,
+                    ty: wgpu::BindingType::UniformBuffer { dynamic: false },
+                }
+            ],
+        });
+
+        let light_count = <Read<Light>>::query().iter_immutable(world).count();
+        let forward_uniforms = ForwardUniforms {
+            proj: math::Mat4::identity().to_cols_array_2d(),
+            num_lights: [light_count as u32, 0, 0, 0],
+        };
+
+        let uniform_size = mem::size_of::<ForwardUniforms>() as wgpu::BufferAddress;
+        let forward_uniform_buffer = device.create_buffer_with_data(
+            forward_uniforms.as_bytes(),
+            wgpu::BufferUsage::UNIFORM | wgpu::BufferUsage::COPY_DST,
+        );
+
+        // Create bind group
+        let bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
+            layout: &bind_group_layout,
+            bindings: &[
+                wgpu::Binding {
+                    binding: 0,
+                    resource: wgpu::BindingResource::Buffer {
+                        buffer: &forward_uniform_buffer,
+                        range: 0 .. uniform_size,
+                    },
+                },
+                wgpu::Binding {
+                    binding: 1,
+                    resource: wgpu::BindingResource::Buffer {
+                        buffer: &render_resources.light_uniform_buffer.buffer,
+                        range: 0 .. render_resources.light_uniform_buffer.size,
+                    },
+                }
+            ],
+        });
+
+        let simple_material_uniforms_size = mem::size_of::<SimpleMaterialUniforms>();
+        let instance_buffer_descriptor = wgpu::VertexBufferDescriptor {
+            stride: simple_material_uniforms_size as wgpu::BufferAddress,
+            step_mode: wgpu::InputStepMode::Instance,
+            attributes: &[
+                wgpu::VertexAttributeDescriptor {
+                    format: wgpu::VertexFormat::Float3,
+                    offset: 0,
+                    shader_location: 2,
+                },
+                wgpu::VertexAttributeDescriptor {
+                    format: wgpu::VertexFormat::Float4,
+                    offset: 3 * 4,
+                    shader_location: 3,
+                },
+            ],
+        };
+
+        let pipeline_layout = device.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
+            bind_group_layouts: &[&bind_group_layout],
+        });
+
+        let vs_module = device.create_shader_module(&vs_bytes);
+        let fs_module = device.create_shader_module(&fs_bytes);
+
+        let pipeline = device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
+            layout: &pipeline_layout,
+            vertex_stage: wgpu::ProgrammableStageDescriptor {
+                module: &vs_module,
+                entry_point: "main",
+            },
+            fragment_stage: Some(wgpu::ProgrammableStageDescriptor {
+                module: &fs_module,
+                entry_point: "main",
+            }),
+            rasterization_state: Some(wgpu::RasterizationStateDescriptor {
+                front_face: wgpu::FrontFace::Ccw,
+                cull_mode: wgpu::CullMode::Back,
+                depth_bias: 0,
+                depth_bias_slope_scale: 0.0,
+                depth_bias_clamp: 0.0,
+            }),
+            primitive_topology: wgpu::PrimitiveTopology::TriangleList,
+            color_states: &[
+                wgpu::ColorStateDescriptor {
+                    format: swap_chain_descriptor.format,
+                    color_blend: wgpu::BlendDescriptor::REPLACE,
+                    alpha_blend: wgpu::BlendDescriptor::REPLACE,
+                    write_mask: wgpu::ColorWrite::ALL,
+                },
+            ],
+            depth_stencil_state: Some(wgpu::DepthStencilStateDescriptor {
+                format: Self::DEPTH_FORMAT,
+                depth_write_enabled: true,
+                depth_compare: wgpu::CompareFunction::Less,
+                stencil_front: wgpu::StencilStateFaceDescriptor::IGNORE,
+                stencil_back: wgpu::StencilStateFaceDescriptor::IGNORE,
+                stencil_read_mask: 0,
+                stencil_write_mask: 0,
+            }),
+            index_format: wgpu::IndexFormat::Uint16,
+            vertex_buffers: &[vertex_buffer_descriptor, instance_buffer_descriptor],
+            sample_count: 1,
+            sample_mask: !0,
+            alpha_to_coverage_enabled: false,
+        });
+
+        let instance_buffer_infos = ForwardInstancedPass::create_instance_buffer_infos(device, world);
+
+        ForwardInstancedPass {
+            pipeline,
+            bind_group,
+            forward_uniform_buffer,
+            depth_texture: Self::get_depth_texture(device, swap_chain_descriptor),
+            instance_buffer_infos
+        }
+    }
+
+    fn create_instance_buffer_infos(device: &Device, world: &World) -> Vec<InstanceBufferInfo> {
+        let mut entities = <(Read<Material>, Read<LocalToWorld>, Read<Handle<Mesh>>, Read<Instanced>)>::query();
+        let entities_count = entities.iter_immutable(world).count();
+        let size = mem::size_of::<SimpleMaterialUniforms>();
+
+        // TODO: use a staging buffer for more efficient gpu reads
+        let temp_buf_data = device
+            .create_buffer_mapped(entities_count * size, wgpu::BufferUsage::COPY_SRC | wgpu::BufferUsage::VERTEX);
+
+        // TODO: generate these buffers for multiple meshes
+        
+        let mut last_mesh_id = None;
+        for ((material, transform, mesh, _), slot) in entities.iter_immutable(world)
+            .zip(temp_buf_data.data.chunks_exact_mut(size))
+        {
+
+            last_mesh_id = Some(*mesh.id.read().unwrap());
+            let (_, _, translation) = transform.0.to_scale_rotation_translation();
+            slot.copy_from_slice(
+                SimpleMaterialUniforms {
+                    position: translation.into(),
+                    color: material.color.into(),
+                }
+                .as_bytes(),
+            );
+        }
+        
+        let mut instance_buffer_infos = Vec::new();
+        instance_buffer_infos.push(InstanceBufferInfo {
+            mesh_id: last_mesh_id.unwrap(),
+            buffer: temp_buf_data.finish(),
+            instance_count: entities_count,
+        });
+
+        instance_buffer_infos
+    }
+    
+    fn get_depth_texture(device: &Device, swap_chain_descriptor: &SwapChainDescriptor) -> wgpu::TextureView {
+        let texture = device.create_texture(&wgpu::TextureDescriptor {
+            size: wgpu::Extent3d {
+                width: swap_chain_descriptor.width,
+                height: swap_chain_descriptor.height,
+                depth: 1,
+            },
+            array_layer_count: 1,
+            mip_level_count: 1,
+            sample_count: 1,
+            dimension: wgpu::TextureDimension::D2,
+            format: Self::DEPTH_FORMAT,
+            usage: wgpu::TextureUsage::OUTPUT_ATTACHMENT,
+        });
+
+        texture.create_default_view()
+    }
+}
+

src/render/material.rs

@@ -7,6 +7,8 @@ pub struct Material {
     pub uniform_buf: Option<wgpu::Buffer>,
 }
 
+pub struct Instanced;
+
 impl Material {
     pub fn new(color: math::Vec4) -> Self {
         Material {
@@ -29,3 +31,10 @@ pub struct MaterialUniforms {
     pub model: [[f32; 4]; 4],
     pub color: [f32; 4],
 }
+
+#[repr(C)]
+#[derive(Clone, Copy, AsBytes, FromBytes)]
+pub struct SimpleMaterialUniforms {
+    pub position: [f32; 3],
+    pub color: [f32; 4],
+}

src/render/mod.rs

@@ -3,14 +3,16 @@ pub mod shader;
 pub mod mesh;
 mod forward;
 mod forward_shadow;
+mod forward_instanced;
 mod shadow;
 mod light;
 mod pass;
 mod material;
 mod render_resources;
 
-pub use forward_shadow::{ForwardShadowPass};
 pub use forward::{ForwardPass, ForwardUniforms};
+pub use forward_shadow::{ForwardShadowPass};
+pub use forward_instanced::ForwardInstancedPass;
 pub use shadow::ShadowPass;
 pub use light::*;
 pub use shader::*;

src/render/shadow/mod.rs

@@ -21,7 +21,9 @@ pub struct ShadowUniforms {
 impl Pass for ShadowPass {
     fn render(&mut self, device: &Device, _: &SwapChainOutput, encoder: &mut CommandEncoder, world: &mut World, render_resources: &RenderResources) {
         let mut light_query = <(Read<Light>, Read<LocalToWorld>, Read<Translation>)>::query();
-        let mut mesh_query = <(Read<Material>, Read<Handle<Mesh>>)>::query();
+        let mut mesh_query =
+            <(Read<Material>, Read<Handle<Mesh>>)>::query()
+                .filter(!component::<Instanced>());
 
         for (i, (mut light, _)) in <(Write<Light>, Read<LocalToWorld>)>::query().iter(world).enumerate() {
             if let None = light.target_view {