Struct WgpuPipelineSystem 

pub struct WgpuPipelineSystem { /* private fields */ }

The wgpu/naga_oil [PipelineSystem] backend. Injected via runtime.backends as Arc<dyn PipelineSystem>.

Implementations

impl WgpuPipelineSystem

pub fn new() -> Result<Self, RenderError>

Builds the system: registers every lib module with the composer (fails fast on a compose error). No graphics device needed at construction.

Trait Implementations

impl PipelineSystem for WgpuPipelineSystem

fn layout( &self, device: &dyn GraphicsDevice, key: LayoutKey, variant: &ShaderVariantKey, ) -> Result<BindGroupLayoutId, RenderError>

Returns the canonical bind-group layout for key under variant, creating + caching it on first request. Both lit lanes (pipeline slot) and the material projection (group-2 bind group) obtain layouts here, so they share the same id.

fn inline_layout( &self, device: &dyn GraphicsDevice, label: &'static str, entries: &[BindGroupLayoutEntry], ) -> Result<BindGroupLayoutId, RenderError>

Returns a bespoke (inline) bind-group layout, identified + cached by its stable label, creating it on first request. Lanes with non-canonical layouts (shadow / UI / overlay / unlit) obtain their layout ids here so they can build ring buffers / bind groups against the same id the pipeline was built with.

fn pipeline( &self, device: &dyn GraphicsDevice, spec: &PipelineSpec, ) -> Result<RenderPipelineId, RenderError>

Returns the render pipeline for spec, creating + caching it on first request (keyed by (shader, variant, color_format)). Resolves the spec’s bind-group layouts, compiles the shader for the variant, and builds the pipeline. Lanes call this every frame; it is a cache hit after the first.

fn compute_pipeline( &self, device: &dyn GraphicsDevice, spec: &ComputePipelineSpec, ) -> Result<ComputePipelineId, RenderError>

Returns the compute pipeline for spec, creating + caching it on first request (keyed by (shader, variant)). Resolves the spec’s bind-group layouts, compiles the shader for the variant, and builds the pipeline. Used by Forward+ light culling.

fn set_overlay_source(&self, logical_path: &str, source: String)

Hot-reload: override a logical shader source (lib or pipeline) with new text, marking it (and dependent pipelines) for recompose. Default no-op until the hot-reload phase wires it.

fn recompose_dirty( &self, device: &dyn GraphicsDevice, ) -> Result<(), RenderError>

Hot-reload: recompose any modules marked dirty by set_overlay_source and rebuild the affected cached pipelines in place. Default no-op.