Struct WgpuDispatcher 

pub struct WgpuDispatcher { /* private fields */ }

Dispatcher that compiles engawa render graphs to wgpu commands. Construct once; call dispatch_with per frame.

Implementations

impl WgpuDispatcher

pub fn new(device: &Device, queue: &Queue, target_format: TextureFormat) -> Self

Construct a dispatcher. The device/queue handles are cloned (wgpu handles are internally reference-counted) — the dispatcher holds no lifetime borrow, so a consumer that does not own its device (mado’s TerminalRenderer) can still own a dispatcher.

pub fn cached_pipeline_count(&self) -> usize

Number of Materials with a compiled pipeline in the cache. Pipelines compile once per Material name; a second dispatch_with over the same graph must not grow this count.

pub fn invalidate_material(&mut self, name: &str)

Drop the cached pipeline for one Material name. The cache is keyed by name only — a hot-reload that swaps a Material’s shader under the same name MUST call this or the stale pipeline keeps dispatching.

pub fn dispatch_with( &mut self, graph: &CompiledGraph, bindings: &ResourceBindings, bound: BoundResources, frame: &FrameUniforms, ) -> Result<CommandBuffer, WgpuDispatcherError>

Canonical per-call dispatch: write the per-frame uniforms, compile any uncached Materials, walk the graph, return the recorded CommandBuffer ready to submit.

Trait Implementations

impl Dispatcher for WgpuDispatcher

fn dispatch_node( &mut self, node: &Node, _bindings: &ResourceBindings, ) -> Result<(), DispatchError>

Dispatch a single node. Called by dispatch_graph in execution order. Implementations record / encode / dispatch as appropriate for their backend.

fn dispatch_graph( &mut self, graph: &CompiledGraph, bindings: &ResourceBindings, ) -> Result<(), DispatchError>

Walk graph.execution_order, validating bindings + then calling dispatch_node for each. The default impl handles all engawa-side concerns; backends typically don’t override this.