Struct CudaEventWrapper 

pub struct CudaEventWrapper { /* private fields */ }

Safe wrapper around a cudarc CudaEvent.

Records a point in a stream’s execution timeline and allows the host or other streams to wait until that point is reached.

All methods return GpuResult rather than panicking on CUDA errors.

Implementations

impl CudaEventWrapper

pub fn new(ctx: &Arc<CudaContext>) -> GpuResult<Self>

Create a new event associated with the given device’s context.

The event is created with CU_EVENT_DISABLE_TIMING (the cudarc default when None is passed for flags). Use [new_with_timing] if you need elapsed-time queries.

pub fn new_with_timing(ctx: &Arc<CudaContext>) -> GpuResult<Self>

Create a new event with timing enabled.

Required if you want to call elapsed_ms. Timing events are slightly more expensive than non-timing events.

pub fn record(&self, stream: &CudaStream) -> GpuResult<()>

Record the current point in stream’s execution into this event.

After recording, synchronize will block until all work submitted to stream before this call has completed.

Errors

Returns Err if the stream belongs to a different CUDA context than the event, or if the CUDA driver reports an error.

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

Block the calling CPU thread until all work recorded in this event has completed on the GPU.

Errors

Returns Err if the CUDA driver reports an error (e.g., a previous async kernel launch failed).

pub fn query(&self) -> GpuResult<bool>

Query whether all work recorded in this event has completed.

Returns Ok(true) if complete, Ok(false) if still in progress. This is a non-blocking check.

pub fn wait_on(&self, stream: &CudaStream) -> GpuResult<()>

Make stream wait for all work recorded in this event to complete before executing any subsequent operations.

This is a GPU-side wait — it does not block the CPU.

Errors

Returns Err if the stream and event belong to different CUDA contexts.

pub fn inner(&self) -> &CudaEvent

Borrow the underlying cudarc CudaEvent.

Trait Implementations

impl Debug for CudaEventWrapper

Available on crate feature cuda only.

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

Formats the value using the given formatter.