Struct DispatchFuture 

pub struct DispatchFuture { /* private fields */ }

A future representing an in-flight GPU dispatch.

On the no-CUDA stub path this resolves immediately. On CUDA hosts the future polls a cust::event::Event recorded on the launch stream: event.query() returns Ok(()) once the GPU has finished the associated work. Until then we re-schedule via the waker so the wasmtime fiber can continue to be suspended.

The future carries a tracing::Span captured at construction time (the active span belonging to whichever host function created it, typically wasi_cuda.launch). Every poll call enters that span, so any work — including the cust::event::Event::query poll path — is attributed to the originating dispatch in the trace tree even though poll is invoked from the Tokio runtime’s reactor, not from inside the host function. Without this the dispatch’s poll events would be parented to the runtime worker’s empty context and would disappear from the distributed trace.

Implementations

impl DispatchFuture

pub fn ready(permit: DispatchPermit) -> Self

Build a future bound to the given back-pressure permit. The future resolves the next time it is polled (no-CUDA path) or once the CUDA event has fired (CUDA path with bind_event).

Trait Implementations

impl Future for DispatchFuture

type Output = ()

The type of value produced on completion.

fn poll(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<()>

Attempts to resolve the future to a final value, registering the current task for wakeup if the value is not yet available.