Struct ComputeBundle 

pub struct ComputeBundle<B = BindGroup> { /* private fields */ }

A bundle of wgpu::ComputePipeline, its wgpu::BindGroupLayout and optionally wgpu::BindGroup.

Overview

This is an abstraction of a compute pipeline with its associated resources, so that any compute operations can be easily setup and dispatched.

It is recommended to use ComputeBundleBuilder to create a compute bundle

Shader Format

The compute shader is suggested to be in the following form:

override workgroup_size: u32;

@compute @workgroup_size(workgroup_size)
fn main(@builtin(global_invocation_id) id: vec3<u32>) {
    let index = id.x;

    if index >= arrayLength(&data) {
        return;
    }

    // Do something with `data[index]`
}

• workgroup_size is an overridable variable of type u32.
• The entry point function (here main) must have the @compute attribute and a @workgroup_size(workgroup_size) attribute.
• The entry point function is suggested to have a parameter with @builtin(global_invocation_id) attribute to get the global invocation ID for indexing into the data.

Implementations

impl<B> ComputeBundle<B>

pub fn create_bind_group<'a>( &self, device: &Device, index: usize, resources: impl IntoIterator<Item = BindingResource<'a>>, ) -> Option<BindGroup>

Create the bind group at the given index.

index refers to the index in ComputeBundle::bind_group_layouts.

Returns None if the index is out of bounds.

As a good practice, if you are designing API for others to use, do not let the user create bind groups manually as they will have to make sure the binding resources match the layout.

pub fn workgroup_size(&self) -> u32

Get the number of invocations in one workgroup.

pub fn label(&self) -> Option<&str>

Get the label.

pub fn bind_group_layouts(&self) -> &[BindGroupLayout]

Get the bind group layouts.

This corresponds to the bind_group_layout_descriptors provided when creating the compute bundle.

pub fn pipeline(&self) -> &ComputePipeline

Get the compute pipeline.

pub fn dispatch_with_bind_groups<'a>( &self, encoder: &mut CommandEncoder, bind_groups: impl IntoIterator<Item = &'a BindGroup>, count: u32, )

Dispatch the compute bundle for count instances with provided bind group.

Each bind group in bind_groups corresponds to the bind group layout at the same index in ComputeBundle::bind_group_layouts.

impl ComputeBundle

pub fn new<'a, 'b>( label: Option<&str>, device: &Device, bind_group_layout_descriptors: impl IntoIterator<Item = &'a BindGroupLayoutDescriptor<'a>>, resources: impl IntoIterator<Item = impl IntoIterator<Item = BindingResource<'a>>>, compilation_options: PipelineCompilationOptions<'_>, shader_source: ShaderSource<'_>, entry_point: &str, workgroup_size: Option<u32>, ) -> Result<Self, ComputeBundleCreateError>

Create a new compute bundle.

shader_source requires an overridable variable workgroup_size of u32, see docs of ComputeBundle.

pub fn bind_groups(&self) -> &[BindGroup]

Get the bind groups.

pub fn dispatch(&self, encoder: &mut CommandEncoder, count: u32)

Dispatch the compute bundle for count instances.

pub fn update_bind_group( &mut self, index: usize, bind_group: BindGroup, ) -> Option<BindGroup>

Update the bind group at index.

Returns Some of the previous bind group if it was updated, or None if the index is out of bounds.

pub fn update_bind_group_with_binding_resources<'a>( &mut self, device: &Device, index: usize, resources: impl IntoIterator<Item = BindingResource<'a>>, ) -> Option<BindGroup>

Update the bind group at index with the provided binding resources.

Returns Some of the previous bind group if it was updated, or None if the index is out of bounds.

impl ComputeBundle<()>

pub fn new_without_bind_groups<'a>( label: Option<&str>, device: &Device, bind_group_layout_descriptors: impl IntoIterator<Item = &'a BindGroupLayoutDescriptor<'a>>, compilation_options: PipelineCompilationOptions<'_>, shader_source: ShaderSource<'_>, entry_point: &str, workgroup_size: Option<u32>, ) -> Result<Self, ComputeBundleCreateError>

Create a new compute bundle without internally managed bind group.

To create a bind group with layout matched to one of the layout in this compute bundle, use the ComputeBundle::create_bind_group method.

pub fn dispatch<'a>( &self, encoder: &mut CommandEncoder, count: u32, bind_groups: impl IntoIterator<Item = &'a BindGroup>, )

Dispatch the compute bundle for count instances.

Trait Implementations

impl<B: Clone> Clone for ComputeBundle<B>

fn clone(&self) -> ComputeBundle<B>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<B: Debug> Debug for ComputeBundle<B>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.