Struct Compute 

pub struct Compute { /* private fields */ }

A compute pipeline asset, plus the bind group values (buffers/textures) it dispatches with — WGSL shader source and its bind group layout compile into a wgpu::ComputePipeline; many Computes sharing the same shader automatically share one compiled pipeline (see ComputePipelineCache). Dispatch it via Backend::dispatch_compute.

Implementations

impl Compute

pub fn new(shader_source: &'static str) -> Self

pub fn with_label(self, label: &'static str) -> Self

pub fn with_entry_point(self, entry: &'static str) -> Self

pub fn with_entry(self, name: &'static str, kind: BindingKind) -> Self

Declares one of this pass’s own (group 0) bind group entries, at the next auto-assigned binding index — the low-level counterpart to the streamlined .texture(...)/.buffer(...)/etc. calls, for a kind one of those doesn’t produce (a dynamic-offset buffer, a non-default sample type). Pair it with the matching value-only .with_texture(...)/etc.

pub fn with_entry_at( self, name: &'static str, binding: u32, kind: BindingKind, ) -> Self

Same as with_entry, pinning an explicit binding index instead of auto-assigning the next one.

pub fn with_extra_group(self, group: GroupEntry) -> Self

Appends a bind group beyond this pass’s own (group 0) — typically GroupEntry::Global, a layout shared with other materials/computes via GlobalLayoutPool. Groups append in call order, starting at group 1.

pub fn with_texture(self, name: &'static str, handle: Handle<Texture>) -> Self

Value-only counterpart to .texture(...) — see texture.

pub fn with_texture_array( self, name: &'static str, handle: Handle<TextureArray>, ) -> Self

Value-only counterpart to .texture_array(...) — see texture.

pub fn with_cubemap(self, name: &'static str, handle: Handle<Cubemap>) -> Self

Value-only counterpart to .cubemap(...) — see texture.

pub fn with_texture_view(self, name: &'static str, view: TextureView) -> Self

Binds an already-built TextureView directly — e.g. one mip level from GPUTexture::get_view, or a standalone render target from Texture::empty. Unlike .with_texture/.with_texture_array/.with_cubemap, no Handle lookup happens at upload time — view must already exist.

pub fn with_sampler(self, name: &'static str, kind: SamplerKind) -> Self

Value-only counterpart to .sampler(...) — see texture.

pub fn with_uniform(self, name: &'static str, data: Vec<u8>) -> Self

Value-only counterpart to .uniform(...) — see texture.

pub fn with_storage(self, name: &'static str, data: Vec<u8>) -> Self

Value-only counterpart to .storage(...) — see texture.

pub fn with_uniform_value<T>(self, name: &'static str, value: &T) -> Self

where T: ShaderType + WriteInto,

Same as with_uniform, but takes a typed value instead of pre-packed bytes — uses encase to lay it out with correct WGSL uniform (std140) alignment. Value-only counterpart to .uniform_value(...).

pub fn with_storage_value<T>(self, name: &'static str, value: &T) -> Self

where T: ShaderType + WriteInto,

Same as with_storage, but takes a typed value instead of pre-packed bytes — uses encase to lay it out with correct WGSL storage (std430) alignment. Value-only counterpart to .storage_value(...).

pub fn texture(self, name: &'static str, handle: Handle<Texture>) -> Self

Declares a compute-visible texture_2d<f32> entry at the next auto-assigned binding index and binds handle to it — the streamlined one-call form of .with_entry(name, BindingKind::texture_2d(COMPUTE)) followed by .with_texture(name, handle). Reach for those two directly for a non-default sample type or an explicit binding index — visibility is always COMPUTE for a compute pass, so there’s no visibility to override here.

pub fn texture_array( self, name: &'static str, handle: Handle<TextureArray>, ) -> Self

Streamlined form of .texture_array(...) — see texture.

pub fn cubemap(self, name: &'static str, handle: Handle<Cubemap>) -> Self

Streamlined form of .cubemap(...) — see texture.

pub fn sampler(self, name: &'static str, kind: SamplerKind) -> Self

Streamlined form of .sampler(...) — see texture.

pub fn uniform(self, name: &'static str, data: Vec<u8>) -> Self

Streamlined form of .uniform(...) — see texture.

pub fn storage(self, name: &'static str, data: Vec<u8>) -> Self

Streamlined form of .storage(...) (read-write — a compute pass binding a storage buffer usually means to write it) — see texture. Use .with_entry(...) + .with_storage(...) directly for a read-only one.

pub fn uniform_value<T>(self, name: &'static str, value: &T) -> Self

where T: ShaderType + WriteInto,

Streamlined, typed form of .uniform(...) — declares the entry and binds an encase-laid-out value in one call. See texture.

pub fn storage_value<T>(self, name: &'static str, value: &T) -> Self

where T: ShaderType + WriteInto,

Streamlined, typed form of .storage(...) (read-write) — see storage.

pub fn with_buffer(self, name: &'static str, buffer: Buffer) -> Self

Binds an existing Buffer instead of uploading raw bytes — for a buffer another pass already wrote to. Unlike .with_uniform/ .with_storage, no buffer is created here; buffer must already carry the usage flags this binding needs.

pub fn with_dynamic_buffer( self, name: &'static str, buffer: DynamicBuffer, ) -> Self

Binds an existing DynamicBuffer — the dynamic-offset counterpart to .with_buffer.

pub fn with_param(self, name: &'static str, entry: BindingValue) -> Self

pub fn build_asset( self, name: &str, assets: &mut Assets<Compute>, ) -> Handle<Compute>

Trait Implementations

impl Asset<Backend> for Compute

type Deps<'a> = (Read<'a, GlobalLayoutPool>, Read<'a, ComputePipelineCache>, Read<'a, Assets<Texture>>, Read<'a, Assets<TextureArray>>, Read<'a, Assets<Cubemap>>, Read<'a, GlobalSamplers>)

fn upload<'a>( &self, backend: &Backend, deps: &Self::Deps<'a>, ) -> Option<GPUCompute>

impl AssetSource for Compute

type Processed = GPUCompute

impl Default for Compute

fn default() -> Self

Returns the “default value” for a type.