bevy/crates/bevy_sprite/src/mesh2d/material.rs

394 lines
16 KiB
Rust
Raw Normal View History

use bevy_app::{App, Plugin};
use bevy_asset::{AddAsset, Asset, AssetServer, Handle};
use bevy_core_pipeline::Transparent2d;
use bevy_ecs::{
entity::Entity,
prelude::{Bundle, World},
system::{
lifetimeless::{Read, SQuery, SRes},
Query, Res, ResMut, SystemParamItem,
},
world::FromWorld,
};
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
use bevy_log::error;
use bevy_render::{
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
mesh::{Mesh, MeshVertexBufferLayout},
render_asset::{RenderAsset, RenderAssetPlugin, RenderAssets},
render_component::ExtractComponentPlugin,
render_phase::{
AddRenderCommand, DrawFunctions, EntityRenderCommand, RenderCommandResult, RenderPhase,
SetItemPipeline, TrackedRenderPass,
},
render_resource::{
BindGroup, BindGroupLayout, PipelineCache, RenderPipelineDescriptor, Shader,
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
SpecializedMeshPipeline, SpecializedMeshPipelineError, SpecializedMeshPipelines,
},
renderer::RenderDevice,
view::{ComputedVisibility, Msaa, Visibility, VisibleEntities},
RenderApp, RenderStage,
};
use bevy_transform::components::{GlobalTransform, Transform};
use bevy_utils::FloatOrd;
use std::hash::Hash;
use std::marker::PhantomData;
use crate::{
DrawMesh2d, Mesh2dHandle, Mesh2dPipeline, Mesh2dPipelineKey, Mesh2dUniform, SetMesh2dBindGroup,
SetMesh2dViewBindGroup,
};
/// Materials are used alongside [`Material2dPlugin`] and [`MaterialMesh2dBundle`]
/// to spawn entities that are rendered with a specific [`Material2d`] type. They serve as an easy to use high level
/// way to render [`Mesh2dHandle`] entities with custom shader logic. For materials that can specialize their [`RenderPipelineDescriptor`]
/// based on specific material values, see [`SpecializedMaterial2d`]. [`Material2d`] automatically implements [`SpecializedMaterial2d`]
/// and can be used anywhere that type is used (such as [`Material2dPlugin`]).
pub trait Material2d: Asset + RenderAsset {
/// Returns this material's [`BindGroup`]. This should match the layout returned by [`Material2d::bind_group_layout`].
fn bind_group(material: &<Self as RenderAsset>::PreparedAsset) -> &BindGroup;
/// Returns this material's [`BindGroupLayout`]. This should match the [`BindGroup`] returned by [`Material2d::bind_group`].
fn bind_group_layout(render_device: &RenderDevice) -> BindGroupLayout;
/// Returns this material's vertex shader. If [`None`] is returned, the default mesh vertex shader will be used.
/// Defaults to [`None`].
#[allow(unused_variables)]
fn vertex_shader(asset_server: &AssetServer) -> Option<Handle<Shader>> {
None
}
/// Returns this material's fragment shader. If [`None`] is returned, the default mesh fragment shader will be used.
/// Defaults to [`None`].
#[allow(unused_variables)]
fn fragment_shader(asset_server: &AssetServer) -> Option<Handle<Shader>> {
None
}
/// The dynamic uniform indices to set for the given `material`'s [`BindGroup`].
/// Defaults to an empty array / no dynamic uniform indices.
#[allow(unused_variables)]
#[inline]
fn dynamic_uniform_indices(material: &<Self as RenderAsset>::PreparedAsset) -> &[u32] {
&[]
}
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
/// Customizes the default [`RenderPipelineDescriptor`].
#[allow(unused_variables)]
#[inline]
fn specialize(
descriptor: &mut RenderPipelineDescriptor,
layout: &MeshVertexBufferLayout,
) -> Result<(), SpecializedMeshPipelineError> {
Ok(())
}
}
impl<M: Material2d> SpecializedMaterial2d for M {
type Key = ();
#[inline]
Use storage buffers for clustered forward point lights (#3989) # Objective - Make use of storage buffers, where they are available, for clustered forward bindings to support far more point lights in a scene - Fixes #3605 - Based on top of #4079 This branch on an M1 Max can keep 60fps with about 2150 point lights of radius 1m in the Sponza scene where I've been testing. The bottleneck is mostly assigning lights to clusters which grows faster than linearly (I think 1000 lights was about 1.5ms and 5000 was 7.5ms). I have seen papers and presentations leveraging compute shaders that can get this up to over 1 million. That said, I think any further optimisations should probably be done in a separate PR. ## Solution - Add `RenderDevice` to the `Material` and `SpecializedMaterial` trait `::key()` functions to allow setting flags on the keys depending on feature/limit availability - Make `GpuPointLights` and `ViewClusterBuffers` into enums containing `UniformVec` and `StorageBuffer` variants. Implement the necessary API on them to make usage the same for both cases, and the only difference is at initialisation time. - Appropriate shader defs in the shader code to handle the two cases ## Context on some decisions / open questions - I'm using `max_storage_buffers_per_shader_stage >= 3` as a check to see if storage buffers are supported. I was thinking about diving into 'binding resource management' but it feels like we don't have enough use cases to understand the problem yet, and it is mostly a separate concern to this PR, so I think it should be handled separately. - Should `ViewClusterBuffers` and `ViewClusterBindings` be merged, duplicating the count variables into the enum variants? Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-04-07 16:16:35 +00:00
fn key(
_render_device: &RenderDevice,
_material: &<Self as RenderAsset>::PreparedAsset,
) -> Self::Key {
}
#[inline]
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
fn specialize(
_key: Self::Key,
descriptor: &mut RenderPipelineDescriptor,
layout: &MeshVertexBufferLayout,
) -> Result<(), SpecializedMeshPipelineError> {
<M as Material2d>::specialize(descriptor, layout)
}
#[inline]
fn bind_group(material: &<Self as RenderAsset>::PreparedAsset) -> &BindGroup {
<M as Material2d>::bind_group(material)
}
#[inline]
fn bind_group_layout(render_device: &RenderDevice) -> BindGroupLayout {
<M as Material2d>::bind_group_layout(render_device)
}
#[inline]
fn vertex_shader(asset_server: &AssetServer) -> Option<Handle<Shader>> {
<M as Material2d>::vertex_shader(asset_server)
}
#[inline]
fn fragment_shader(asset_server: &AssetServer) -> Option<Handle<Shader>> {
<M as Material2d>::fragment_shader(asset_server)
}
#[allow(unused_variables)]
#[inline]
fn dynamic_uniform_indices(material: &<Self as RenderAsset>::PreparedAsset) -> &[u32] {
<M as Material2d>::dynamic_uniform_indices(material)
}
}
/// Materials are used alongside [`Material2dPlugin`] and [`MaterialMesh2dBundle`](crate::MaterialMesh2dBundle)
/// to spawn entities that are rendered with a specific [`SpecializedMaterial2d`] type. They serve as an easy to use high level
/// way to render [`Mesh2dHandle`] entities with custom shader logic. [`SpecializedMaterial2d`s](SpecializedMaterial2d) use their [`SpecializedMaterial2d::Key`]
/// to customize their [`RenderPipelineDescriptor`] based on specific material values. The slightly simpler [`Material2d`] trait
/// should be used for materials that do not need specialization. [`Material2d`] types automatically implement [`SpecializedMaterial2d`].
pub trait SpecializedMaterial2d: Asset + RenderAsset {
/// The key used to specialize this material's [`RenderPipelineDescriptor`].
type Key: PartialEq + Eq + Hash + Clone + Send + Sync;
/// Extract the [`SpecializedMaterial2d::Key`] for the "prepared" version of this material. This key will be
/// passed in to the [`SpecializedMaterial2d::specialize`] function when compiling the [`RenderPipeline`](bevy_render::render_resource::RenderPipeline)
/// for a given entity's material.
Use storage buffers for clustered forward point lights (#3989) # Objective - Make use of storage buffers, where they are available, for clustered forward bindings to support far more point lights in a scene - Fixes #3605 - Based on top of #4079 This branch on an M1 Max can keep 60fps with about 2150 point lights of radius 1m in the Sponza scene where I've been testing. The bottleneck is mostly assigning lights to clusters which grows faster than linearly (I think 1000 lights was about 1.5ms and 5000 was 7.5ms). I have seen papers and presentations leveraging compute shaders that can get this up to over 1 million. That said, I think any further optimisations should probably be done in a separate PR. ## Solution - Add `RenderDevice` to the `Material` and `SpecializedMaterial` trait `::key()` functions to allow setting flags on the keys depending on feature/limit availability - Make `GpuPointLights` and `ViewClusterBuffers` into enums containing `UniformVec` and `StorageBuffer` variants. Implement the necessary API on them to make usage the same for both cases, and the only difference is at initialisation time. - Appropriate shader defs in the shader code to handle the two cases ## Context on some decisions / open questions - I'm using `max_storage_buffers_per_shader_stage >= 3` as a check to see if storage buffers are supported. I was thinking about diving into 'binding resource management' but it feels like we don't have enough use cases to understand the problem yet, and it is mostly a separate concern to this PR, so I think it should be handled separately. - Should `ViewClusterBuffers` and `ViewClusterBindings` be merged, duplicating the count variables into the enum variants? Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-04-07 16:16:35 +00:00
fn key(
render_device: &RenderDevice,
material: &<Self as RenderAsset>::PreparedAsset,
) -> Self::Key;
/// Specializes the given `descriptor` according to the given `key`.
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
fn specialize(
key: Self::Key,
descriptor: &mut RenderPipelineDescriptor,
layout: &MeshVertexBufferLayout,
) -> Result<(), SpecializedMeshPipelineError>;
/// Returns this material's [`BindGroup`]. This should match the layout returned by [`SpecializedMaterial2d::bind_group_layout`].
fn bind_group(material: &<Self as RenderAsset>::PreparedAsset) -> &BindGroup;
/// Returns this material's [`BindGroupLayout`]. This should match the [`BindGroup`] returned by [`SpecializedMaterial2d::bind_group`].
fn bind_group_layout(render_device: &RenderDevice) -> BindGroupLayout;
/// Returns this material's vertex shader. If [`None`] is returned, the default mesh vertex shader will be used.
/// Defaults to [`None`].
#[allow(unused_variables)]
fn vertex_shader(asset_server: &AssetServer) -> Option<Handle<Shader>> {
None
}
/// Returns this material's fragment shader. If [`None`] is returned, the default mesh fragment shader will be used.
/// Defaults to [`None`].
#[allow(unused_variables)]
fn fragment_shader(asset_server: &AssetServer) -> Option<Handle<Shader>> {
None
}
/// The dynamic uniform indices to set for the given `material`'s [`BindGroup`].
/// Defaults to an empty array / no dynamic uniform indices.
#[allow(unused_variables)]
#[inline]
fn dynamic_uniform_indices(material: &<Self as RenderAsset>::PreparedAsset) -> &[u32] {
&[]
}
}
/// Adds the necessary ECS resources and render logic to enable rendering entities using the given [`SpecializedMaterial2d`]
/// asset type (which includes [`Material2d`] types).
pub struct Material2dPlugin<M: SpecializedMaterial2d>(PhantomData<M>);
impl<M: SpecializedMaterial2d> Default for Material2dPlugin<M> {
fn default() -> Self {
Self(Default::default())
}
}
impl<M: SpecializedMaterial2d> Plugin for Material2dPlugin<M> {
fn build(&self, app: &mut App) {
app.add_asset::<M>()
.add_plugin(ExtractComponentPlugin::<Handle<M>>::extract_visible())
.add_plugin(RenderAssetPlugin::<M>::default());
if let Ok(render_app) = app.get_sub_app_mut(RenderApp) {
render_app
.add_render_command::<Transparent2d, DrawMaterial2d<M>>()
.init_resource::<Material2dPipeline<M>>()
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
.init_resource::<SpecializedMeshPipelines<Material2dPipeline<M>>>()
.add_system_to_stage(RenderStage::Queue, queue_material2d_meshes::<M>);
}
}
}
pub struct Material2dPipeline<M: SpecializedMaterial2d> {
pub mesh2d_pipeline: Mesh2dPipeline,
pub material2d_layout: BindGroupLayout,
pub vertex_shader: Option<Handle<Shader>>,
pub fragment_shader: Option<Handle<Shader>>,
marker: PhantomData<M>,
}
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
#[derive(Eq, PartialEq, Clone, Hash)]
pub struct Material2dKey<T> {
mesh_key: Mesh2dPipelineKey,
material_key: T,
}
impl<M: SpecializedMaterial2d> SpecializedMeshPipeline for Material2dPipeline<M> {
type Key = Material2dKey<M::Key>;
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
fn specialize(
&self,
key: Self::Key,
layout: &MeshVertexBufferLayout,
) -> Result<RenderPipelineDescriptor, SpecializedMeshPipelineError> {
let mut descriptor = self.mesh2d_pipeline.specialize(key.mesh_key, layout)?;
if let Some(vertex_shader) = &self.vertex_shader {
descriptor.vertex.shader = vertex_shader.clone();
}
if let Some(fragment_shader) = &self.fragment_shader {
descriptor.fragment.as_mut().unwrap().shader = fragment_shader.clone();
}
descriptor.layout = Some(vec![
self.mesh2d_pipeline.view_layout.clone(),
self.material2d_layout.clone(),
self.mesh2d_pipeline.mesh_layout.clone(),
]);
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
M::specialize(key.material_key, &mut descriptor, layout)?;
Ok(descriptor)
}
}
impl<M: SpecializedMaterial2d> FromWorld for Material2dPipeline<M> {
fn from_world(world: &mut World) -> Self {
let asset_server = world.resource::<AssetServer>();
let render_device = world.resource::<RenderDevice>();
let material2d_layout = M::bind_group_layout(render_device);
Material2dPipeline {
mesh2d_pipeline: world.resource::<Mesh2dPipeline>().clone(),
material2d_layout,
vertex_shader: M::vertex_shader(asset_server),
fragment_shader: M::fragment_shader(asset_server),
marker: PhantomData,
}
}
}
type DrawMaterial2d<M> = (
SetItemPipeline,
SetMesh2dViewBindGroup<0>,
SetMaterial2dBindGroup<M, 1>,
SetMesh2dBindGroup<2>,
DrawMesh2d,
);
pub struct SetMaterial2dBindGroup<M: SpecializedMaterial2d, const I: usize>(PhantomData<M>);
impl<M: SpecializedMaterial2d, const I: usize> EntityRenderCommand
for SetMaterial2dBindGroup<M, I>
{
type Param = (SRes<RenderAssets<M>>, SQuery<Read<Handle<M>>>);
fn render<'w>(
_view: Entity,
item: Entity,
(materials, query): SystemParamItem<'w, '_, Self::Param>,
pass: &mut TrackedRenderPass<'w>,
) -> RenderCommandResult {
let material2d_handle = query.get(item).unwrap();
let material2d = materials.into_inner().get(material2d_handle).unwrap();
pass.set_bind_group(
I,
M::bind_group(material2d),
M::dynamic_uniform_indices(material2d),
);
RenderCommandResult::Success
}
}
#[allow(clippy::too_many_arguments)]
pub fn queue_material2d_meshes<M: SpecializedMaterial2d>(
transparent_draw_functions: Res<DrawFunctions<Transparent2d>>,
material2d_pipeline: Res<Material2dPipeline<M>>,
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
mut pipelines: ResMut<SpecializedMeshPipelines<Material2dPipeline<M>>>,
mut pipeline_cache: ResMut<PipelineCache>,
Use storage buffers for clustered forward point lights (#3989) # Objective - Make use of storage buffers, where they are available, for clustered forward bindings to support far more point lights in a scene - Fixes #3605 - Based on top of #4079 This branch on an M1 Max can keep 60fps with about 2150 point lights of radius 1m in the Sponza scene where I've been testing. The bottleneck is mostly assigning lights to clusters which grows faster than linearly (I think 1000 lights was about 1.5ms and 5000 was 7.5ms). I have seen papers and presentations leveraging compute shaders that can get this up to over 1 million. That said, I think any further optimisations should probably be done in a separate PR. ## Solution - Add `RenderDevice` to the `Material` and `SpecializedMaterial` trait `::key()` functions to allow setting flags on the keys depending on feature/limit availability - Make `GpuPointLights` and `ViewClusterBuffers` into enums containing `UniformVec` and `StorageBuffer` variants. Implement the necessary API on them to make usage the same for both cases, and the only difference is at initialisation time. - Appropriate shader defs in the shader code to handle the two cases ## Context on some decisions / open questions - I'm using `max_storage_buffers_per_shader_stage >= 3` as a check to see if storage buffers are supported. I was thinking about diving into 'binding resource management' but it feels like we don't have enough use cases to understand the problem yet, and it is mostly a separate concern to this PR, so I think it should be handled separately. - Should `ViewClusterBuffers` and `ViewClusterBindings` be merged, duplicating the count variables into the enum variants? Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-04-07 16:16:35 +00:00
render_device: Res<RenderDevice>,
msaa: Res<Msaa>,
render_meshes: Res<RenderAssets<Mesh>>,
render_materials: Res<RenderAssets<M>>,
material2d_meshes: Query<(&Handle<M>, &Mesh2dHandle, &Mesh2dUniform)>,
mut views: Query<(&VisibleEntities, &mut RenderPhase<Transparent2d>)>,
) {
if material2d_meshes.is_empty() {
return;
}
Use storage buffers for clustered forward point lights (#3989) # Objective - Make use of storage buffers, where they are available, for clustered forward bindings to support far more point lights in a scene - Fixes #3605 - Based on top of #4079 This branch on an M1 Max can keep 60fps with about 2150 point lights of radius 1m in the Sponza scene where I've been testing. The bottleneck is mostly assigning lights to clusters which grows faster than linearly (I think 1000 lights was about 1.5ms and 5000 was 7.5ms). I have seen papers and presentations leveraging compute shaders that can get this up to over 1 million. That said, I think any further optimisations should probably be done in a separate PR. ## Solution - Add `RenderDevice` to the `Material` and `SpecializedMaterial` trait `::key()` functions to allow setting flags on the keys depending on feature/limit availability - Make `GpuPointLights` and `ViewClusterBuffers` into enums containing `UniformVec` and `StorageBuffer` variants. Implement the necessary API on them to make usage the same for both cases, and the only difference is at initialisation time. - Appropriate shader defs in the shader code to handle the two cases ## Context on some decisions / open questions - I'm using `max_storage_buffers_per_shader_stage >= 3` as a check to see if storage buffers are supported. I was thinking about diving into 'binding resource management' but it feels like we don't have enough use cases to understand the problem yet, and it is mostly a separate concern to this PR, so I think it should be handled separately. - Should `ViewClusterBuffers` and `ViewClusterBindings` be merged, duplicating the count variables into the enum variants? Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-04-07 16:16:35 +00:00
let render_device = render_device.into_inner();
for (visible_entities, mut transparent_phase) in views.iter_mut() {
let draw_transparent_pbr = transparent_draw_functions
.read()
.get_id::<DrawMaterial2d<M>>()
.unwrap();
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
let msaa_key = Mesh2dPipelineKey::from_msaa_samples(msaa.samples);
for visible_entity in &visible_entities.entities {
if let Ok((material2d_handle, mesh2d_handle, mesh2d_uniform)) =
material2d_meshes.get(*visible_entity)
{
if let Some(material2d) = render_materials.get(material2d_handle) {
if let Some(mesh) = render_meshes.get(&mesh2d_handle.0) {
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
let mesh_key = msaa_key
| Mesh2dPipelineKey::from_primitive_topology(mesh.primitive_topology);
Use storage buffers for clustered forward point lights (#3989) # Objective - Make use of storage buffers, where they are available, for clustered forward bindings to support far more point lights in a scene - Fixes #3605 - Based on top of #4079 This branch on an M1 Max can keep 60fps with about 2150 point lights of radius 1m in the Sponza scene where I've been testing. The bottleneck is mostly assigning lights to clusters which grows faster than linearly (I think 1000 lights was about 1.5ms and 5000 was 7.5ms). I have seen papers and presentations leveraging compute shaders that can get this up to over 1 million. That said, I think any further optimisations should probably be done in a separate PR. ## Solution - Add `RenderDevice` to the `Material` and `SpecializedMaterial` trait `::key()` functions to allow setting flags on the keys depending on feature/limit availability - Make `GpuPointLights` and `ViewClusterBuffers` into enums containing `UniformVec` and `StorageBuffer` variants. Implement the necessary API on them to make usage the same for both cases, and the only difference is at initialisation time. - Appropriate shader defs in the shader code to handle the two cases ## Context on some decisions / open questions - I'm using `max_storage_buffers_per_shader_stage >= 3` as a check to see if storage buffers are supported. I was thinking about diving into 'binding resource management' but it feels like we don't have enough use cases to understand the problem yet, and it is mostly a separate concern to this PR, so I think it should be handled separately. - Should `ViewClusterBuffers` and `ViewClusterBindings` be merged, duplicating the count variables into the enum variants? Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-04-07 16:16:35 +00:00
let material_key = M::key(render_device, material2d);
Mesh vertex buffer layouts (#3959) This PR makes a number of changes to how meshes and vertex attributes are handled, which the goal of enabling easy and flexible custom vertex attributes: * Reworks the `Mesh` type to use the newly added `VertexAttribute` internally * `VertexAttribute` defines the name, a unique `VertexAttributeId`, and a `VertexFormat` * `VertexAttributeId` is used to produce consistent sort orders for vertex buffer generation, replacing the more expensive and often surprising "name based sorting" * Meshes can be used to generate a `MeshVertexBufferLayout`, which defines the layout of the gpu buffer produced by the mesh. `MeshVertexBufferLayouts` can then be used to generate actual `VertexBufferLayouts` according to the requirements of a specific pipeline. This decoupling of "mesh layout" vs "pipeline vertex buffer layout" is what enables custom attributes. We don't need to standardize _mesh layouts_ or contort meshes to meet the needs of a specific pipeline. As long as the mesh has what the pipeline needs, it will work transparently. * Mesh-based pipelines now specialize on `&MeshVertexBufferLayout` via the new `SpecializedMeshPipeline` trait (which behaves like `SpecializedPipeline`, but adds `&MeshVertexBufferLayout`). The integrity of the pipeline cache is maintained because the `MeshVertexBufferLayout` is treated as part of the key (which is fully abstracted from implementers of the trait ... no need to add any additional info to the specialization key). * Hashing `MeshVertexBufferLayout` is too expensive to do for every entity, every frame. To make this scalable, I added a generalized "pre-hashing" solution to `bevy_utils`: `Hashed<T>` keys and `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . Why didn't I just do the quick and dirty in-place "pre-compute hash and use that u64 as a key in a hashmap" that we've done in the past? Because its wrong! Hashes by themselves aren't enough because two different values can produce the same hash. Re-hashing a hash is even worse! I decided to build a generalized solution because this pattern has come up in the past and we've chosen to do the wrong thing. Now we can do the right thing! This did unfortunately require pulling in `hashbrown` and using that in `bevy_utils`, because avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. Note that adding `hashbrown` doesn't increase our dependency count because it was already in our tree. I will probably break these changes out into their own PR. * Specializing on `MeshVertexBufferLayout` has one non-obvious behavior: it can produce identical pipelines for two different MeshVertexBufferLayouts. To optimize the number of active pipelines / reduce re-binds while drawing, I de-duplicate pipelines post-specialization using the final `VertexBufferLayout` as the key. For example, consider a pipeline that needs the layout `(position, normal)` and is specialized using two meshes: `(position, normal, uv)` and `(position, normal, other_vec2)`. If both of these meshes result in `(position, normal)` specializations, we can use the same pipeline! Now we do. Cool! To briefly illustrate, this is what the relevant section of `MeshPipeline`'s specialization code looks like now: ```rust impl SpecializedMeshPipeline for MeshPipeline { type Key = MeshPipelineKey; fn specialize( &self, key: Self::Key, layout: &MeshVertexBufferLayout, ) -> RenderPipelineDescriptor { let mut vertex_attributes = vec![ Mesh::ATTRIBUTE_POSITION.at_shader_location(0), Mesh::ATTRIBUTE_NORMAL.at_shader_location(1), Mesh::ATTRIBUTE_UV_0.at_shader_location(2), ]; let mut shader_defs = Vec::new(); if layout.contains(Mesh::ATTRIBUTE_TANGENT) { shader_defs.push(String::from("VERTEX_TANGENTS")); vertex_attributes.push(Mesh::ATTRIBUTE_TANGENT.at_shader_location(3)); } let vertex_buffer_layout = layout .get_layout(&vertex_attributes) .expect("Mesh is missing a vertex attribute"); ``` Notice that this is _much_ simpler than it was before. And now any mesh with any layout can be used with this pipeline, provided it has vertex postions, normals, and uvs. We even got to remove `HAS_TANGENTS` from MeshPipelineKey and `has_tangents` from `GpuMesh`, because that information is redundant with `MeshVertexBufferLayout`. This is still a draft because I still need to: * Add more docs * Experiment with adding error handling to mesh pipeline specialization (which would print errors at runtime when a mesh is missing a vertex attribute required by a pipeline). If it doesn't tank perf, we'll keep it. * Consider breaking out the PreHash / hashbrown changes into a separate PR. * Add an example illustrating this change * Verify that the "mesh-specialized pipeline de-duplication code" works properly Please dont yell at me for not doing these things yet :) Just trying to get this in peoples' hands asap. Alternative to #3120 Fixes #3030 Co-authored-by: Carter Anderson <mcanders1@gmail.com>
2022-02-23 23:21:13 +00:00
let pipeline_id = pipelines.specialize(
&mut pipeline_cache,
&material2d_pipeline,
Material2dKey {
mesh_key,
material_key,
},
&mesh.layout,
);
let pipeline_id = match pipeline_id {
Ok(id) => id,
Err(err) => {
error!("{}", err);
continue;
}
};
let mesh_z = mesh2d_uniform.transform.w_axis.z;
transparent_phase.add(Transparent2d {
entity: *visible_entity,
draw_function: draw_transparent_pbr,
pipeline: pipeline_id,
// NOTE: Back-to-front ordering for transparent with ascending sort means far should have the
// lowest sort key and getting closer should increase. As we have
// -z in front of the camera, the largest distance is -far with values increasing toward the
// camera. As such we can just use mesh_z as the distance
sort_key: FloatOrd(mesh_z),
// This material is not batched
batch_range: None,
});
}
}
}
}
}
}
/// A component bundle for entities with a [`Mesh2dHandle`] and a [`SpecializedMaterial2d`].
#[derive(Bundle, Clone)]
pub struct MaterialMesh2dBundle<M: SpecializedMaterial2d> {
pub mesh: Mesh2dHandle,
pub material: Handle<M>,
pub transform: Transform,
pub global_transform: GlobalTransform,
/// User indication of whether an entity is visible
pub visibility: Visibility,
/// Algorithmically-computed indication of whether an entity is visible and should be extracted for rendering
pub computed_visibility: ComputedVisibility,
}
impl<M: SpecializedMaterial2d> Default for MaterialMesh2dBundle<M> {
fn default() -> Self {
Self {
mesh: Default::default(),
material: Default::default(),
transform: Default::default(),
global_transform: Default::default(),
visibility: Default::default(),
computed_visibility: Default::default(),
}
}
}