Struct CompletionEvent 

pub struct CompletionEvent { /* private fields */ }

Synchronization event for asynchronous accelerator operations.

Tracks completion of asynchronous operations on accelerator devices, enabling operation chaining without CPU synchronization.

Examples

use tenferro_tensor::CompletionEvent;

// CompletionEvent is typically created by GPU backends.
let _event: Option<CompletionEvent> = None;

Implementations

impl CompletionEvent

pub fn noop() -> Self

Create a no-op completion event.

This is useful for testing and for cases where no synchronization is needed.

Trait Implementations

impl Clone for CompletionEvent

fn clone(&self) -> CompletionEvent

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Send for CompletionEvent

Safety

CompletionEvent can be safely sent across threads because:

• The raw pointer in CompletionEventInner is only used as an opaque handle
• CUDA/ROCm events are internally thread-safe when used with proper synchronization
• The event is only used for synchronization and does not expose mutable state
• The pointer is never dereferenced directly by this crate

impl Sync for CompletionEvent

Safety

CompletionEvent can be safely shared across threads because:

• The raw pointer is an opaque handle to a GPU event object
• CUDA/ROCm event APIs are thread-safe for query operations
• The event does not contain any Rust-managed mutable state
• Concurrent access to the event is managed by the GPU driver