Trait PipelineStage 

pub trait PipelineStage: Send + 'static {
    type In: Send + 'static;
    type Out: Send + 'static;

    // Required method
    fn enqueue(
        &mut self,
        stream: &Arc<CudaStream>,
        wait_for: Option<&CudaEvent>,
        input: Self::In,
    ) -> Result<(CudaEvent, Self::Out), GpuError>;
}

One stage in a multi-stream GPU pipeline.

Implementations enqueue their kernel onto stream synchronously (no host wait) and return a CudaEvent marking the completion of that stage’s GPU work, plus the typed output. The executor arranges that the next stage’s wait_for is the previous stage’s returned event, so cross-stage synchronization is on-device only.

Required Associated Types

type In: Send + 'static

type Out: Send + 'static

Required Methods

fn enqueue( &mut self, stream: &Arc<CudaStream>, wait_for: Option<&CudaEvent>, input: Self::In, ) -> Result<(CudaEvent, Self::Out), GpuError>

Implementors