cuda::experimental::event_ref

Defined in /home/runner/work/cccl/cccl/cudax/include/cuda/experimental/__event/event_ref.cuh

class event_ref

An non-owning wrapper for an untimed cudaEvent_t.

Subclassed by cuda::experimental::event

Public Types

using value_type = ::cudaEvent_t

Public Functions

inline constexpr event_ref(::cudaEvent_t __evnt) noexcept

Construct a new event_ref object from a cudaEvent_t

This constructor provides an implicit conversion from cudaEvent_t

Note

: It is the callers responsibilty to ensure the event_ref does not outlive the event denoted by the cudaEvent_t handle.

Post

get() == __evnt

event_ref(int) = delete

Disallow construction from an int, e.g., 0.

event_ref(cuda::std::nullptr_t) = delete

Disallow construction from nullptr.

inline void record(stream_ref __stream) const

Records an event on the specified stream.

Parameters

__stream –

Throws

cuda_error – if the event record fails

inline void wait() const

Waits until all the work in the stream prior to the record of the event has completed.

Throws

cuda_error – if waiting for the event fails

inline bool is_done() const

Checks if all the work in the stream prior to the record of the event has completed.

If is_done returns true, calling wait() on this event will return immediately

Throws

cuda_error – if the event query fails

inline constexpr ::cudaEvent_t get() const noexcept

Retrieve the native cudaEvent_t handle.

Returns

cudaEvent_t The native handle being held by the event_ref object.

inline explicit constexpr operator bool() const noexcept

Checks if the event_ref is valid.

Returns

true if the event_ref is valid, false otherwise.

inline constexpr bool operator==(event_ref __lhs, event_ref __rhs) noexcept

Compares two event_refs for equality.

Note

Allows comparison with cudaEvent_t due to implicit conversion to event_ref.

Parameters
Returns

true if lhs and rhs refer to the same cudaEvent_t object.

inline constexpr bool operator!=(event_ref __lhs, event_ref __rhs) noexcept

Compares two event_refs for inequality.

Note

Allows comparison with cudaEvent_t due to implicit conversion to event_ref.

Parameters
Returns

true if lhs and rhs refer to different cudaEvent_t objects.