Struct GpuCompiler

pub struct GpuCompiler<E> { /* private fields */ }

Graph-compiler provider backed by one initialized GPU executor.

Implementations

impl<E> GpuCompiler<E>

pub const fn new(executor: E) -> GpuCompiler<E>

Wrap one initialized concrete executor.

pub fn device(&self) -> &GpuDeviceDescriptor

where E: GpuGraphExecutor,

Underlying device facts.

pub const fn executor(&self) -> &E

Borrow the API-specific executor for advanced operations.

Trait Implementations

impl<E> Backend for GpuCompiler<E>

where E: GpuGraphExecutor,

fn descriptor(&self) -> BackendDescriptor

Runtime descriptor and capability inventory.

impl<E> Debug for GpuCompiler<E>

where E: Debug,

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

Formats the value using the given formatter.

impl<E> MeshCompiler for GpuCompiler<E>

where E: GpuGraphExecutor,

fn compile_mesh_batch_into( &self, graph: &GeometryGraph, roots: &[NodeId], options: &ExecutionOptions, destination: &mut Vec<TriMesh>, ) -> Result<(), GeomError>

Overriding the _into seam keeps both batch call shapes on the single-dispatch GPU path; overriding only compile_mesh_batch would leave compile_mesh_batch_into silently falling back to one submission per root.

fn compile_mesh( &self, graph: &GeometryGraph, root: NodeId, options: &ExecutionOptions, ) -> Result<TriMesh, GeomError>

Compile one root.

fn scratch_requirement(&self) -> ScratchRequirement

Scratch this compiler needs beyond the graph and the produced meshes.

fn output_bound(&self) -> OutputBound

Outputs produced per requested root.

fn compile_mesh_reported( &self, graph: &GeometryGraph, root: NodeId, options: &ExecutionOptions, ) -> Result<CompileOutcome, GeomError>

Compile one root and report what happened to each attribute channel.

fn compile_mesh_batch( &self, graph: &GeometryGraph, roots: &[NodeId], options: &ExecutionOptions, ) -> Result<Vec<TriMesh>, GeomError>

Compile roots as a batch.