bevy/pipelined/bevy_render2/src/texture/image.rs

400 lines
14 KiB
Rust
Raw Normal View History

2021-06-21 23:28:52 +00:00
use super::image_texture_conversion::image_to_texture;
2021-06-25 03:04:28 +00:00
use crate::{
Modular Rendering (#2831) This changes how render logic is composed to make it much more modular. Previously, all extraction logic was centralized for a given "type" of rendered thing. For example, we extracted meshes into a vector of ExtractedMesh, which contained the mesh and material asset handles, the transform, etc. We looked up bindings for "drawn things" using their index in the `Vec<ExtractedMesh>`. This worked fine for built in rendering, but made it hard to reuse logic for "custom" rendering. It also prevented us from reusing things like "extracted transforms" across contexts. To make rendering more modular, I made a number of changes: * Entities now drive rendering: * We extract "render components" from "app components" and store them _on_ entities. No more centralized uber lists! We now have true "ECS-driven rendering" * To make this perform well, I implemented #2673 in upstream Bevy for fast batch insertions into specific entities. This was merged into the `pipelined-rendering` branch here: #2815 * Reworked the `Draw` abstraction: * Generic `PhaseItems`: each draw phase can define its own type of "rendered thing", which can define its own "sort key" * Ported the 2d, 3d, and shadow phases to the new PhaseItem impl (currently Transparent2d, Transparent3d, and Shadow PhaseItems) * `Draw` trait and and `DrawFunctions` are now generic on PhaseItem * Modular / Ergonomic `DrawFunctions` via `RenderCommands` * RenderCommand is a trait that runs an ECS query and produces one or more RenderPass calls. Types implementing this trait can be composed to create a final DrawFunction. For example the DrawPbr DrawFunction is created from the following DrawCommand tuple. Const generics are used to set specific bind group locations: ```rust pub type DrawPbr = ( SetPbrPipeline, SetMeshViewBindGroup<0>, SetStandardMaterialBindGroup<1>, SetTransformBindGroup<2>, DrawMesh, ); ``` * The new `custom_shader_pipelined` example illustrates how the commands above can be reused to create a custom draw function: ```rust type DrawCustom = ( SetCustomMaterialPipeline, SetMeshViewBindGroup<0>, SetTransformBindGroup<2>, DrawMesh, ); ``` * ExtractComponentPlugin and UniformComponentPlugin: * Simple, standardized ways to easily extract individual components and write them to GPU buffers * Ported PBR and Sprite rendering to the new primitives above. * Removed staging buffer from UniformVec in favor of direct Queue usage * Makes UniformVec much easier to use and more ergonomic. Completely removes the need for custom render graph nodes in these contexts (see the PbrNode and view Node removals and the much simpler call patterns in the relevant Prepare systems). * Added a many_cubes_pipelined example to benchmark baseline 3d rendering performance and ensure there were no major regressions during this port. Avoiding regressions was challenging given that the old approach of extracting into centralized vectors is basically the "optimal" approach. However thanks to a various ECS optimizations and render logic rephrasing, we pretty much break even on this benchmark! * Lifetimeless SystemParams: this will be a bit divisive, but as we continue to embrace "trait driven systems" (ex: ExtractComponentPlugin, UniformComponentPlugin, DrawCommand), the ergonomics of `(Query<'static, 'static, (&'static A, &'static B, &'static)>, Res<'static, C>)` were getting very hard to bear. As a compromise, I added "static type aliases" for the relevant SystemParams. The previous example can now be expressed like this: `(SQuery<(Read<A>, Read<B>)>, SRes<C>)`. If anyone has better ideas / conflicting opinions, please let me know! * RunSystem trait: a way to define Systems via a trait with a SystemParam associated type. This is used to implement the various plugins mentioned above. I also added SystemParamItem and QueryItem type aliases to make "trait stye" ecs interactions nicer on the eyes (and fingers). * RenderAsset retrying: ensures that render assets are only created when they are "ready" and allows us to create bind groups directly inside render assets (which significantly simplified the StandardMaterial code). I think ultimately we should swap this out on "asset dependency" events to wait for dependencies to load, but this will require significant asset system changes. * Updated some built in shaders to account for missing MeshUniform fields
2021-09-23 06:16:11 +00:00
render_asset::{PrepareAssetError, RenderAsset},
2021-06-25 03:04:28 +00:00
render_resource::{Sampler, Texture, TextureView},
renderer::{RenderDevice, RenderQueue},
texture::BevyDefault,
2021-06-25 03:04:28 +00:00
};
Modular Rendering (#2831) This changes how render logic is composed to make it much more modular. Previously, all extraction logic was centralized for a given "type" of rendered thing. For example, we extracted meshes into a vector of ExtractedMesh, which contained the mesh and material asset handles, the transform, etc. We looked up bindings for "drawn things" using their index in the `Vec<ExtractedMesh>`. This worked fine for built in rendering, but made it hard to reuse logic for "custom" rendering. It also prevented us from reusing things like "extracted transforms" across contexts. To make rendering more modular, I made a number of changes: * Entities now drive rendering: * We extract "render components" from "app components" and store them _on_ entities. No more centralized uber lists! We now have true "ECS-driven rendering" * To make this perform well, I implemented #2673 in upstream Bevy for fast batch insertions into specific entities. This was merged into the `pipelined-rendering` branch here: #2815 * Reworked the `Draw` abstraction: * Generic `PhaseItems`: each draw phase can define its own type of "rendered thing", which can define its own "sort key" * Ported the 2d, 3d, and shadow phases to the new PhaseItem impl (currently Transparent2d, Transparent3d, and Shadow PhaseItems) * `Draw` trait and and `DrawFunctions` are now generic on PhaseItem * Modular / Ergonomic `DrawFunctions` via `RenderCommands` * RenderCommand is a trait that runs an ECS query and produces one or more RenderPass calls. Types implementing this trait can be composed to create a final DrawFunction. For example the DrawPbr DrawFunction is created from the following DrawCommand tuple. Const generics are used to set specific bind group locations: ```rust pub type DrawPbr = ( SetPbrPipeline, SetMeshViewBindGroup<0>, SetStandardMaterialBindGroup<1>, SetTransformBindGroup<2>, DrawMesh, ); ``` * The new `custom_shader_pipelined` example illustrates how the commands above can be reused to create a custom draw function: ```rust type DrawCustom = ( SetCustomMaterialPipeline, SetMeshViewBindGroup<0>, SetTransformBindGroup<2>, DrawMesh, ); ``` * ExtractComponentPlugin and UniformComponentPlugin: * Simple, standardized ways to easily extract individual components and write them to GPU buffers * Ported PBR and Sprite rendering to the new primitives above. * Removed staging buffer from UniformVec in favor of direct Queue usage * Makes UniformVec much easier to use and more ergonomic. Completely removes the need for custom render graph nodes in these contexts (see the PbrNode and view Node removals and the much simpler call patterns in the relevant Prepare systems). * Added a many_cubes_pipelined example to benchmark baseline 3d rendering performance and ensure there were no major regressions during this port. Avoiding regressions was challenging given that the old approach of extracting into centralized vectors is basically the "optimal" approach. However thanks to a various ECS optimizations and render logic rephrasing, we pretty much break even on this benchmark! * Lifetimeless SystemParams: this will be a bit divisive, but as we continue to embrace "trait driven systems" (ex: ExtractComponentPlugin, UniformComponentPlugin, DrawCommand), the ergonomics of `(Query<'static, 'static, (&'static A, &'static B, &'static)>, Res<'static, C>)` were getting very hard to bear. As a compromise, I added "static type aliases" for the relevant SystemParams. The previous example can now be expressed like this: `(SQuery<(Read<A>, Read<B>)>, SRes<C>)`. If anyone has better ideas / conflicting opinions, please let me know! * RunSystem trait: a way to define Systems via a trait with a SystemParam associated type. This is used to implement the various plugins mentioned above. I also added SystemParamItem and QueryItem type aliases to make "trait stye" ecs interactions nicer on the eyes (and fingers). * RenderAsset retrying: ensures that render assets are only created when they are "ready" and allows us to create bind groups directly inside render assets (which significantly simplified the StandardMaterial code). I think ultimately we should swap this out on "asset dependency" events to wait for dependencies to load, but this will require significant asset system changes. * Updated some built in shaders to account for missing MeshUniform fields
2021-09-23 06:16:11 +00:00
use bevy_ecs::system::{lifetimeless::SRes, SystemParamItem};
2021-04-11 20:13:07 +00:00
use bevy_reflect::TypeUuid;
use thiserror::Error;
2021-06-25 03:04:28 +00:00
use wgpu::{
Extent3d, ImageCopyTexture, ImageDataLayout, Origin3d, TextureDimension, TextureFormat,
TextureViewDescriptor,
};
2021-04-11 20:13:07 +00:00
pub const TEXTURE_ASSET_INDEX: u64 = 0;
pub const SAMPLER_ASSET_INDEX: u64 = 1;
#[derive(Debug, Clone, TypeUuid)]
#[uuid = "6ea26da6-6cf8-4ea2-9986-1d7bf6c17d6f"]
2021-06-21 23:28:52 +00:00
pub struct Image {
2021-04-11 20:13:07 +00:00
pub data: Vec<u8>,
2021-06-21 23:28:52 +00:00
// TODO: this nesting makes accessing Image metadata verbose. Either flatten out descriptor or add accessors
pub texture_descriptor: wgpu::TextureDescriptor<'static>,
pub sampler_descriptor: wgpu::SamplerDescriptor<'static>,
2021-04-11 20:13:07 +00:00
}
2021-06-21 23:28:52 +00:00
impl Default for Image {
2021-04-11 20:13:07 +00:00
fn default() -> Self {
let format = wgpu::TextureFormat::bevy_default();
let data = vec![1; format.pixel_size() as usize];
2021-06-21 23:28:52 +00:00
Image {
data,
2021-06-21 23:28:52 +00:00
texture_descriptor: wgpu::TextureDescriptor {
size: wgpu::Extent3d {
width: 1,
height: 1,
depth_or_array_layers: 1,
},
format,
2021-06-21 23:28:52 +00:00
dimension: wgpu::TextureDimension::D2,
label: None,
mip_level_count: 1,
sample_count: 1,
usage: wgpu::TextureUsages::TEXTURE_BINDING | wgpu::TextureUsages::COPY_DST,
2021-04-11 20:13:07 +00:00
},
2021-06-21 23:28:52 +00:00
sampler_descriptor: wgpu::SamplerDescriptor::default(),
2021-04-11 20:13:07 +00:00
}
}
}
2021-06-21 23:28:52 +00:00
impl Image {
2021-04-11 20:13:07 +00:00
pub fn new(
size: Extent3d,
dimension: TextureDimension,
data: Vec<u8>,
format: TextureFormat,
) -> Self {
debug_assert_eq!(
size.volume() * format.pixel_size(),
data.len(),
"Pixel data, size and format have to match",
);
2021-07-02 01:05:20 +00:00
let mut image = Self {
data,
..Default::default()
};
2021-06-21 23:28:52 +00:00
image.texture_descriptor.dimension = dimension;
image.texture_descriptor.size = size;
image.texture_descriptor.format = format;
image
2021-04-11 20:13:07 +00:00
}
pub fn new_fill(
size: Extent3d,
dimension: TextureDimension,
pixel: &[u8],
format: TextureFormat,
) -> Self {
2021-06-21 23:28:52 +00:00
let mut value = Image::default();
value.texture_descriptor.format = format;
value.texture_descriptor.dimension = dimension;
2021-04-11 20:13:07 +00:00
value.resize(size);
debug_assert_eq!(
pixel.len() % format.pixel_size(),
0,
"Must not have incomplete pixel data."
);
debug_assert!(
pixel.len() <= value.data.len(),
"Fill data must fit within pixel buffer."
);
for current_pixel in value.data.chunks_exact_mut(pixel.len()) {
current_pixel.copy_from_slice(pixel);
2021-04-11 20:13:07 +00:00
}
value
}
pub fn aspect_2d(&self) -> f32 {
2021-06-21 23:28:52 +00:00
self.texture_descriptor.size.height as f32 / self.texture_descriptor.size.width as f32
2021-04-11 20:13:07 +00:00
}
pub fn resize(&mut self, size: Extent3d) {
2021-06-21 23:28:52 +00:00
self.texture_descriptor.size = size;
self.data.resize(
size.volume() * self.texture_descriptor.format.pixel_size(),
0,
);
2021-04-11 20:13:07 +00:00
}
/// Changes the `size`, asserting that the total number of data elements (pixels) remains the
/// same.
pub fn reinterpret_size(&mut self, new_size: Extent3d) {
assert!(
2021-06-21 23:28:52 +00:00
new_size.volume() == self.texture_descriptor.size.volume(),
2021-04-11 20:13:07 +00:00
"Incompatible sizes: old = {:?} new = {:?}",
2021-06-21 23:28:52 +00:00
self.texture_descriptor.size,
2021-04-11 20:13:07 +00:00
new_size
);
2021-06-21 23:28:52 +00:00
self.texture_descriptor.size = new_size;
2021-04-11 20:13:07 +00:00
}
/// Takes a 2D texture containing vertically stacked images of the same size, and reinterprets
/// it as a 2D array texture, where each of the stacked images becomes one layer of the
/// array. This is primarily for use with the `texture2DArray` shader uniform type.
pub fn reinterpret_stacked_2d_as_array(&mut self, layers: u32) {
// Must be a stacked image, and the height must be divisible by layers.
2021-06-21 23:28:52 +00:00
assert!(self.texture_descriptor.dimension == TextureDimension::D2);
assert!(self.texture_descriptor.size.depth_or_array_layers == 1);
assert_eq!(self.texture_descriptor.size.height % layers, 0);
2021-04-11 20:13:07 +00:00
self.reinterpret_size(Extent3d {
2021-06-21 23:28:52 +00:00
width: self.texture_descriptor.size.width,
height: self.texture_descriptor.size.height / layers,
2021-04-11 20:13:07 +00:00
depth_or_array_layers: layers,
});
}
/// Convert a texture from a format to another
/// Only a few formats are supported as input and output:
/// - `TextureFormat::R8Unorm`
/// - `TextureFormat::Rg8Unorm`
/// - `TextureFormat::Rgba8UnormSrgb`
/// - `TextureFormat::Bgra8UnormSrgb`
pub fn convert(&self, new_format: TextureFormat) -> Option<Self> {
super::image_texture_conversion::texture_to_image(self)
.and_then(|img| match new_format {
TextureFormat::R8Unorm => Some(image::DynamicImage::ImageLuma8(img.into_luma8())),
TextureFormat::Rg8Unorm => {
Some(image::DynamicImage::ImageLumaA8(img.into_luma_alpha8()))
}
TextureFormat::Rgba8UnormSrgb => {
Some(image::DynamicImage::ImageRgba8(img.into_rgba8()))
}
TextureFormat::Bgra8UnormSrgb => {
Some(image::DynamicImage::ImageBgra8(img.into_bgra8()))
}
_ => None,
})
.map(super::image_texture_conversion::image_to_texture)
}
/// Load a bytes buffer in a [`Texture`], according to type `image_type`, using the `image`
/// crate`
2021-06-21 23:28:52 +00:00
pub fn from_buffer(buffer: &[u8], image_type: ImageType) -> Result<Image, TextureError> {
2021-04-11 20:13:07 +00:00
let format = match image_type {
ImageType::MimeType(mime_type) => match mime_type {
"image/png" => Ok(image::ImageFormat::Png),
"image/vnd-ms.dds" => Ok(image::ImageFormat::Dds),
"image/x-targa" => Ok(image::ImageFormat::Tga),
"image/x-tga" => Ok(image::ImageFormat::Tga),
"image/jpeg" => Ok(image::ImageFormat::Jpeg),
"image/bmp" => Ok(image::ImageFormat::Bmp),
"image/x-bmp" => Ok(image::ImageFormat::Bmp),
_ => Err(TextureError::InvalidImageMimeType(mime_type.to_string())),
},
ImageType::Extension(extension) => image::ImageFormat::from_extension(extension)
.ok_or_else(|| TextureError::InvalidImageMimeType(extension.to_string())),
}?;
// Load the image in the expected format.
// Some formats like PNG allow for R or RG textures too, so the texture
// format needs to be determined. For RGB textures an alpha channel
// needs to be added, so the image data needs to be converted in those
// cases.
let dyn_img = image::load_from_memory_with_format(buffer, format)?;
Ok(image_to_texture(dyn_img))
}
}
/// An error that occurs when loading a texture
#[derive(Error, Debug)]
pub enum TextureError {
#[error("invalid image mime type")]
InvalidImageMimeType(String),
#[error("invalid image extension")]
InvalidImageExtension(String),
#[error("failed to load an image: {0}")]
ImageError(#[from] image::ImageError),
}
/// Type of a raw image buffer
pub enum ImageType<'a> {
/// Mime type of an image, for example `"image/png"`
MimeType(&'a str),
/// Extension of an image file, for example `"png"`
Extension(&'a str),
}
2021-06-21 23:28:52 +00:00
pub trait Volume {
fn volume(&self) -> usize;
}
impl Volume for Extent3d {
fn volume(&self) -> usize {
(self.width * self.height * self.depth_or_array_layers) as usize
}
}
pub struct PixelInfo {
pub type_size: usize,
pub num_components: usize,
}
pub trait TextureFormatPixelInfo {
fn pixel_info(&self) -> PixelInfo;
fn pixel_size(&self) -> usize {
let info = self.pixel_info();
info.type_size * info.num_components
}
}
impl TextureFormatPixelInfo for TextureFormat {
fn pixel_info(&self) -> PixelInfo {
let type_size = match self {
// 8bit
TextureFormat::R8Unorm
| TextureFormat::R8Snorm
| TextureFormat::R8Uint
| TextureFormat::R8Sint
| TextureFormat::Rg8Unorm
| TextureFormat::Rg8Snorm
| TextureFormat::Rg8Uint
| TextureFormat::Rg8Sint
| TextureFormat::Rgba8Unorm
| TextureFormat::Rgba8UnormSrgb
| TextureFormat::Rgba8Snorm
| TextureFormat::Rgba8Uint
| TextureFormat::Rgba8Sint
| TextureFormat::Bgra8Unorm
| TextureFormat::Bgra8UnormSrgb => 1,
// 16bit
TextureFormat::R16Uint
| TextureFormat::R16Sint
| TextureFormat::R16Float
| TextureFormat::Rg16Uint
| TextureFormat::Rg16Sint
| TextureFormat::Rg16Float
| TextureFormat::Rgba16Uint
| TextureFormat::Rgba16Sint
| TextureFormat::Rgba16Float => 2,
// 32bit
TextureFormat::R32Uint
| TextureFormat::R32Sint
| TextureFormat::R32Float
| TextureFormat::Rg32Uint
| TextureFormat::Rg32Sint
| TextureFormat::Rg32Float
| TextureFormat::Rgba32Uint
| TextureFormat::Rgba32Sint
| TextureFormat::Rgba32Float
| TextureFormat::Depth32Float => 4,
// special cases
TextureFormat::Rgb10a2Unorm => 4,
TextureFormat::Rg11b10Float => 4,
TextureFormat::Depth24Plus => 3, // FIXME is this correct?
TextureFormat::Depth24PlusStencil8 => 4,
// TODO: this is not good! this is a temporary step while porting bevy_render to direct wgpu usage
_ => panic!("cannot get pixel info for type"),
};
let components = match self {
TextureFormat::R8Unorm
| TextureFormat::R8Snorm
| TextureFormat::R8Uint
| TextureFormat::R8Sint
| TextureFormat::R16Uint
| TextureFormat::R16Sint
| TextureFormat::R16Float
| TextureFormat::R32Uint
| TextureFormat::R32Sint
| TextureFormat::R32Float => 1,
TextureFormat::Rg8Unorm
| TextureFormat::Rg8Snorm
| TextureFormat::Rg8Uint
| TextureFormat::Rg8Sint
| TextureFormat::Rg16Uint
| TextureFormat::Rg16Sint
| TextureFormat::Rg16Float
| TextureFormat::Rg32Uint
| TextureFormat::Rg32Sint
| TextureFormat::Rg32Float => 2,
TextureFormat::Rgba8Unorm
| TextureFormat::Rgba8UnormSrgb
| TextureFormat::Rgba8Snorm
| TextureFormat::Rgba8Uint
| TextureFormat::Rgba8Sint
| TextureFormat::Bgra8Unorm
| TextureFormat::Bgra8UnormSrgb
| TextureFormat::Rgba16Uint
| TextureFormat::Rgba16Sint
| TextureFormat::Rgba16Float
| TextureFormat::Rgba32Uint
| TextureFormat::Rgba32Sint
| TextureFormat::Rgba32Float => 4,
// special cases
TextureFormat::Rgb10a2Unorm
| TextureFormat::Rg11b10Float
| TextureFormat::Depth32Float
| TextureFormat::Depth24Plus
| TextureFormat::Depth24PlusStencil8 => 1,
// TODO: this is not good! this is a temporary step while porting bevy_render to direct wgpu usage
_ => panic!("cannot get pixel info for type"),
};
PixelInfo {
type_size,
num_components: components,
}
}
}
2021-06-25 03:04:28 +00:00
#[derive(Debug, Clone)]
pub struct GpuImage {
pub texture: Texture,
pub texture_view: TextureView,
pub sampler: Sampler,
}
impl RenderAsset for Image {
type ExtractedAsset = Image;
type PreparedAsset = GpuImage;
Modular Rendering (#2831) This changes how render logic is composed to make it much more modular. Previously, all extraction logic was centralized for a given "type" of rendered thing. For example, we extracted meshes into a vector of ExtractedMesh, which contained the mesh and material asset handles, the transform, etc. We looked up bindings for "drawn things" using their index in the `Vec<ExtractedMesh>`. This worked fine for built in rendering, but made it hard to reuse logic for "custom" rendering. It also prevented us from reusing things like "extracted transforms" across contexts. To make rendering more modular, I made a number of changes: * Entities now drive rendering: * We extract "render components" from "app components" and store them _on_ entities. No more centralized uber lists! We now have true "ECS-driven rendering" * To make this perform well, I implemented #2673 in upstream Bevy for fast batch insertions into specific entities. This was merged into the `pipelined-rendering` branch here: #2815 * Reworked the `Draw` abstraction: * Generic `PhaseItems`: each draw phase can define its own type of "rendered thing", which can define its own "sort key" * Ported the 2d, 3d, and shadow phases to the new PhaseItem impl (currently Transparent2d, Transparent3d, and Shadow PhaseItems) * `Draw` trait and and `DrawFunctions` are now generic on PhaseItem * Modular / Ergonomic `DrawFunctions` via `RenderCommands` * RenderCommand is a trait that runs an ECS query and produces one or more RenderPass calls. Types implementing this trait can be composed to create a final DrawFunction. For example the DrawPbr DrawFunction is created from the following DrawCommand tuple. Const generics are used to set specific bind group locations: ```rust pub type DrawPbr = ( SetPbrPipeline, SetMeshViewBindGroup<0>, SetStandardMaterialBindGroup<1>, SetTransformBindGroup<2>, DrawMesh, ); ``` * The new `custom_shader_pipelined` example illustrates how the commands above can be reused to create a custom draw function: ```rust type DrawCustom = ( SetCustomMaterialPipeline, SetMeshViewBindGroup<0>, SetTransformBindGroup<2>, DrawMesh, ); ``` * ExtractComponentPlugin and UniformComponentPlugin: * Simple, standardized ways to easily extract individual components and write them to GPU buffers * Ported PBR and Sprite rendering to the new primitives above. * Removed staging buffer from UniformVec in favor of direct Queue usage * Makes UniformVec much easier to use and more ergonomic. Completely removes the need for custom render graph nodes in these contexts (see the PbrNode and view Node removals and the much simpler call patterns in the relevant Prepare systems). * Added a many_cubes_pipelined example to benchmark baseline 3d rendering performance and ensure there were no major regressions during this port. Avoiding regressions was challenging given that the old approach of extracting into centralized vectors is basically the "optimal" approach. However thanks to a various ECS optimizations and render logic rephrasing, we pretty much break even on this benchmark! * Lifetimeless SystemParams: this will be a bit divisive, but as we continue to embrace "trait driven systems" (ex: ExtractComponentPlugin, UniformComponentPlugin, DrawCommand), the ergonomics of `(Query<'static, 'static, (&'static A, &'static B, &'static)>, Res<'static, C>)` were getting very hard to bear. As a compromise, I added "static type aliases" for the relevant SystemParams. The previous example can now be expressed like this: `(SQuery<(Read<A>, Read<B>)>, SRes<C>)`. If anyone has better ideas / conflicting opinions, please let me know! * RunSystem trait: a way to define Systems via a trait with a SystemParam associated type. This is used to implement the various plugins mentioned above. I also added SystemParamItem and QueryItem type aliases to make "trait stye" ecs interactions nicer on the eyes (and fingers). * RenderAsset retrying: ensures that render assets are only created when they are "ready" and allows us to create bind groups directly inside render assets (which significantly simplified the StandardMaterial code). I think ultimately we should swap this out on "asset dependency" events to wait for dependencies to load, but this will require significant asset system changes. * Updated some built in shaders to account for missing MeshUniform fields
2021-09-23 06:16:11 +00:00
type Param = (SRes<RenderDevice>, SRes<RenderQueue>);
2021-06-25 03:04:28 +00:00
fn extract_asset(&self) -> Self::ExtractedAsset {
self.clone()
}
fn prepare_asset(
image: Self::ExtractedAsset,
Modular Rendering (#2831) This changes how render logic is composed to make it much more modular. Previously, all extraction logic was centralized for a given "type" of rendered thing. For example, we extracted meshes into a vector of ExtractedMesh, which contained the mesh and material asset handles, the transform, etc. We looked up bindings for "drawn things" using their index in the `Vec<ExtractedMesh>`. This worked fine for built in rendering, but made it hard to reuse logic for "custom" rendering. It also prevented us from reusing things like "extracted transforms" across contexts. To make rendering more modular, I made a number of changes: * Entities now drive rendering: * We extract "render components" from "app components" and store them _on_ entities. No more centralized uber lists! We now have true "ECS-driven rendering" * To make this perform well, I implemented #2673 in upstream Bevy for fast batch insertions into specific entities. This was merged into the `pipelined-rendering` branch here: #2815 * Reworked the `Draw` abstraction: * Generic `PhaseItems`: each draw phase can define its own type of "rendered thing", which can define its own "sort key" * Ported the 2d, 3d, and shadow phases to the new PhaseItem impl (currently Transparent2d, Transparent3d, and Shadow PhaseItems) * `Draw` trait and and `DrawFunctions` are now generic on PhaseItem * Modular / Ergonomic `DrawFunctions` via `RenderCommands` * RenderCommand is a trait that runs an ECS query and produces one or more RenderPass calls. Types implementing this trait can be composed to create a final DrawFunction. For example the DrawPbr DrawFunction is created from the following DrawCommand tuple. Const generics are used to set specific bind group locations: ```rust pub type DrawPbr = ( SetPbrPipeline, SetMeshViewBindGroup<0>, SetStandardMaterialBindGroup<1>, SetTransformBindGroup<2>, DrawMesh, ); ``` * The new `custom_shader_pipelined` example illustrates how the commands above can be reused to create a custom draw function: ```rust type DrawCustom = ( SetCustomMaterialPipeline, SetMeshViewBindGroup<0>, SetTransformBindGroup<2>, DrawMesh, ); ``` * ExtractComponentPlugin and UniformComponentPlugin: * Simple, standardized ways to easily extract individual components and write them to GPU buffers * Ported PBR and Sprite rendering to the new primitives above. * Removed staging buffer from UniformVec in favor of direct Queue usage * Makes UniformVec much easier to use and more ergonomic. Completely removes the need for custom render graph nodes in these contexts (see the PbrNode and view Node removals and the much simpler call patterns in the relevant Prepare systems). * Added a many_cubes_pipelined example to benchmark baseline 3d rendering performance and ensure there were no major regressions during this port. Avoiding regressions was challenging given that the old approach of extracting into centralized vectors is basically the "optimal" approach. However thanks to a various ECS optimizations and render logic rephrasing, we pretty much break even on this benchmark! * Lifetimeless SystemParams: this will be a bit divisive, but as we continue to embrace "trait driven systems" (ex: ExtractComponentPlugin, UniformComponentPlugin, DrawCommand), the ergonomics of `(Query<'static, 'static, (&'static A, &'static B, &'static)>, Res<'static, C>)` were getting very hard to bear. As a compromise, I added "static type aliases" for the relevant SystemParams. The previous example can now be expressed like this: `(SQuery<(Read<A>, Read<B>)>, SRes<C>)`. If anyone has better ideas / conflicting opinions, please let me know! * RunSystem trait: a way to define Systems via a trait with a SystemParam associated type. This is used to implement the various plugins mentioned above. I also added SystemParamItem and QueryItem type aliases to make "trait stye" ecs interactions nicer on the eyes (and fingers). * RenderAsset retrying: ensures that render assets are only created when they are "ready" and allows us to create bind groups directly inside render assets (which significantly simplified the StandardMaterial code). I think ultimately we should swap this out on "asset dependency" events to wait for dependencies to load, but this will require significant asset system changes. * Updated some built in shaders to account for missing MeshUniform fields
2021-09-23 06:16:11 +00:00
(render_device, render_queue): &mut SystemParamItem<Self::Param>,
) -> Result<Self::PreparedAsset, PrepareAssetError<Self::ExtractedAsset>> {
2021-06-25 03:04:28 +00:00
let texture = render_device.create_texture(&image.texture_descriptor);
let sampler = render_device.create_sampler(&image.sampler_descriptor);
let format_size = image.texture_descriptor.format.pixel_size();
render_queue.write_texture(
ImageCopyTexture {
texture: &texture,
mip_level: 0,
origin: Origin3d::ZERO,
aspect: wgpu::TextureAspect::All,
2021-06-25 03:04:28 +00:00
},
&image.data,
ImageDataLayout {
offset: 0,
bytes_per_row: Some(
std::num::NonZeroU32::new(
image.texture_descriptor.size.width * format_size as u32,
)
.unwrap(),
),
rows_per_image: if image.texture_descriptor.size.depth_or_array_layers > 1 {
std::num::NonZeroU32::new(image.texture_descriptor.size.height)
} else {
None
},
2021-06-25 03:04:28 +00:00
},
image.texture_descriptor.size,
);
let texture_view = texture.create_view(&TextureViewDescriptor::default());
Modular Rendering (#2831) This changes how render logic is composed to make it much more modular. Previously, all extraction logic was centralized for a given "type" of rendered thing. For example, we extracted meshes into a vector of ExtractedMesh, which contained the mesh and material asset handles, the transform, etc. We looked up bindings for "drawn things" using their index in the `Vec<ExtractedMesh>`. This worked fine for built in rendering, but made it hard to reuse logic for "custom" rendering. It also prevented us from reusing things like "extracted transforms" across contexts. To make rendering more modular, I made a number of changes: * Entities now drive rendering: * We extract "render components" from "app components" and store them _on_ entities. No more centralized uber lists! We now have true "ECS-driven rendering" * To make this perform well, I implemented #2673 in upstream Bevy for fast batch insertions into specific entities. This was merged into the `pipelined-rendering` branch here: #2815 * Reworked the `Draw` abstraction: * Generic `PhaseItems`: each draw phase can define its own type of "rendered thing", which can define its own "sort key" * Ported the 2d, 3d, and shadow phases to the new PhaseItem impl (currently Transparent2d, Transparent3d, and Shadow PhaseItems) * `Draw` trait and and `DrawFunctions` are now generic on PhaseItem * Modular / Ergonomic `DrawFunctions` via `RenderCommands` * RenderCommand is a trait that runs an ECS query and produces one or more RenderPass calls. Types implementing this trait can be composed to create a final DrawFunction. For example the DrawPbr DrawFunction is created from the following DrawCommand tuple. Const generics are used to set specific bind group locations: ```rust pub type DrawPbr = ( SetPbrPipeline, SetMeshViewBindGroup<0>, SetStandardMaterialBindGroup<1>, SetTransformBindGroup<2>, DrawMesh, ); ``` * The new `custom_shader_pipelined` example illustrates how the commands above can be reused to create a custom draw function: ```rust type DrawCustom = ( SetCustomMaterialPipeline, SetMeshViewBindGroup<0>, SetTransformBindGroup<2>, DrawMesh, ); ``` * ExtractComponentPlugin and UniformComponentPlugin: * Simple, standardized ways to easily extract individual components and write them to GPU buffers * Ported PBR and Sprite rendering to the new primitives above. * Removed staging buffer from UniformVec in favor of direct Queue usage * Makes UniformVec much easier to use and more ergonomic. Completely removes the need for custom render graph nodes in these contexts (see the PbrNode and view Node removals and the much simpler call patterns in the relevant Prepare systems). * Added a many_cubes_pipelined example to benchmark baseline 3d rendering performance and ensure there were no major regressions during this port. Avoiding regressions was challenging given that the old approach of extracting into centralized vectors is basically the "optimal" approach. However thanks to a various ECS optimizations and render logic rephrasing, we pretty much break even on this benchmark! * Lifetimeless SystemParams: this will be a bit divisive, but as we continue to embrace "trait driven systems" (ex: ExtractComponentPlugin, UniformComponentPlugin, DrawCommand), the ergonomics of `(Query<'static, 'static, (&'static A, &'static B, &'static)>, Res<'static, C>)` were getting very hard to bear. As a compromise, I added "static type aliases" for the relevant SystemParams. The previous example can now be expressed like this: `(SQuery<(Read<A>, Read<B>)>, SRes<C>)`. If anyone has better ideas / conflicting opinions, please let me know! * RunSystem trait: a way to define Systems via a trait with a SystemParam associated type. This is used to implement the various plugins mentioned above. I also added SystemParamItem and QueryItem type aliases to make "trait stye" ecs interactions nicer on the eyes (and fingers). * RenderAsset retrying: ensures that render assets are only created when they are "ready" and allows us to create bind groups directly inside render assets (which significantly simplified the StandardMaterial code). I think ultimately we should swap this out on "asset dependency" events to wait for dependencies to load, but this will require significant asset system changes. * Updated some built in shaders to account for missing MeshUniform fields
2021-09-23 06:16:11 +00:00
Ok(GpuImage {
2021-06-25 03:04:28 +00:00
texture,
texture_view,
sampler,
Modular Rendering (#2831) This changes how render logic is composed to make it much more modular. Previously, all extraction logic was centralized for a given "type" of rendered thing. For example, we extracted meshes into a vector of ExtractedMesh, which contained the mesh and material asset handles, the transform, etc. We looked up bindings for "drawn things" using their index in the `Vec<ExtractedMesh>`. This worked fine for built in rendering, but made it hard to reuse logic for "custom" rendering. It also prevented us from reusing things like "extracted transforms" across contexts. To make rendering more modular, I made a number of changes: * Entities now drive rendering: * We extract "render components" from "app components" and store them _on_ entities. No more centralized uber lists! We now have true "ECS-driven rendering" * To make this perform well, I implemented #2673 in upstream Bevy for fast batch insertions into specific entities. This was merged into the `pipelined-rendering` branch here: #2815 * Reworked the `Draw` abstraction: * Generic `PhaseItems`: each draw phase can define its own type of "rendered thing", which can define its own "sort key" * Ported the 2d, 3d, and shadow phases to the new PhaseItem impl (currently Transparent2d, Transparent3d, and Shadow PhaseItems) * `Draw` trait and and `DrawFunctions` are now generic on PhaseItem * Modular / Ergonomic `DrawFunctions` via `RenderCommands` * RenderCommand is a trait that runs an ECS query and produces one or more RenderPass calls. Types implementing this trait can be composed to create a final DrawFunction. For example the DrawPbr DrawFunction is created from the following DrawCommand tuple. Const generics are used to set specific bind group locations: ```rust pub type DrawPbr = ( SetPbrPipeline, SetMeshViewBindGroup<0>, SetStandardMaterialBindGroup<1>, SetTransformBindGroup<2>, DrawMesh, ); ``` * The new `custom_shader_pipelined` example illustrates how the commands above can be reused to create a custom draw function: ```rust type DrawCustom = ( SetCustomMaterialPipeline, SetMeshViewBindGroup<0>, SetTransformBindGroup<2>, DrawMesh, ); ``` * ExtractComponentPlugin and UniformComponentPlugin: * Simple, standardized ways to easily extract individual components and write them to GPU buffers * Ported PBR and Sprite rendering to the new primitives above. * Removed staging buffer from UniformVec in favor of direct Queue usage * Makes UniformVec much easier to use and more ergonomic. Completely removes the need for custom render graph nodes in these contexts (see the PbrNode and view Node removals and the much simpler call patterns in the relevant Prepare systems). * Added a many_cubes_pipelined example to benchmark baseline 3d rendering performance and ensure there were no major regressions during this port. Avoiding regressions was challenging given that the old approach of extracting into centralized vectors is basically the "optimal" approach. However thanks to a various ECS optimizations and render logic rephrasing, we pretty much break even on this benchmark! * Lifetimeless SystemParams: this will be a bit divisive, but as we continue to embrace "trait driven systems" (ex: ExtractComponentPlugin, UniformComponentPlugin, DrawCommand), the ergonomics of `(Query<'static, 'static, (&'static A, &'static B, &'static)>, Res<'static, C>)` were getting very hard to bear. As a compromise, I added "static type aliases" for the relevant SystemParams. The previous example can now be expressed like this: `(SQuery<(Read<A>, Read<B>)>, SRes<C>)`. If anyone has better ideas / conflicting opinions, please let me know! * RunSystem trait: a way to define Systems via a trait with a SystemParam associated type. This is used to implement the various plugins mentioned above. I also added SystemParamItem and QueryItem type aliases to make "trait stye" ecs interactions nicer on the eyes (and fingers). * RenderAsset retrying: ensures that render assets are only created when they are "ready" and allows us to create bind groups directly inside render assets (which significantly simplified the StandardMaterial code). I think ultimately we should swap this out on "asset dependency" events to wait for dependencies to load, but this will require significant asset system changes. * Updated some built in shaders to account for missing MeshUniform fields
2021-09-23 06:16:11 +00:00
})
2021-06-25 03:04:28 +00:00
}
}