Struct GpuStream 

pub struct GpuStream { /* private fields */ }

A CUDA stream obtained from the libtorch stream pool.

RAII: the stream is returned to the pool on drop.

Implementations

impl GpuStream

pub fn new(device: Device, high_priority: bool) -> Result<Self>

Create a new CUDA stream from the pool on the given device.

high_priority: if true, uses a high-priority stream that preempts normal-priority work at SM boundaries.

pub fn synchronize(&self) -> Result<()>

Block the CPU thread until all work on this stream completes.

pub fn wait_event(&self, event: &GpuEvent) -> Result<()>

Make this stream wait for a recorded event before executing any further work. Does not block the CPU.

pub fn is_complete(&self) -> bool

Non-blocking check: has all work on this stream completed?

pub fn current(device: Device) -> Result<Self>

The calling thread’s current stream on device (the default stream unless a StreamGuard is active). The returned value owns only the wrapper — dropping it never destroys the underlying stream.

pub fn device(&self) -> Device

The device this stream belongs to.

Trait Implementations

impl Drop for GpuStream

fn drop(&mut self)

Executes the destructor for this type.

fn pin_drop(self: Pin<&mut Self>)

🔬This is a nightly-only experimental API. (pin_ergonomics)

Execute the destructor for this type, but different to Drop::drop, it requires self to be pinned.

impl Send for GpuStream