Struct HostIngressStream 

pub struct HostIngressStream { /* private fields */ }

Host-ingress adapter for one in-flight chunked dispatch stream.

This is a compatibility adapter for environments where input bytes arrive through host memory. It is intentionally named after ingress, not execution: the device-resident execution model lives in vyre-runtime::megakernel.

Implementations

impl HostIngressStream

pub fn new(pipeline: WgpuPipeline, config: DispatchConfig) -> Self

Create a host-ingress stream from a compiled wgpu pipeline.

pub fn from_runner<F>(runner: F, config: DispatchConfig) -> Self

where F: Fn(Vec<u8>, DispatchConfig) -> Result<Vec<Vec<u8>>, BackendError> + Send + Sync + 'static,

Create a host-ingress stream from a custom chunk runner.

pub fn push_chunk( &mut self, bytes: Vec<u8>, ) -> Result<Option<Vec<Vec<u8>>>, BackendError>

Push a host-memory chunk and return the previous chunk’s output when one exists.

Errors

Returns a backend error if the previous chunk failed or the worker thread panicked before reporting a backend result.

pub fn finish(&mut self) -> Result<Option<Vec<Vec<u8>>>, BackendError>

Wait for the final in-flight chunk and return its output.

Errors

Returns a backend error if the final chunk failed or the worker panicked.