rapidsmpf::CudaEvent Class Reference

RAII wrapper for a CUDA event with convenience methods. More...

#include <cuda_event.hpp>

Public Member Functions
	CudaEvent (unsigned flags=cudaEventDisableTiming)
	Construct a CUDA event. More...
	~CudaEvent () noexcept
	Destroy the CUDA event. More...
	CudaEvent (CudaEvent const &)=delete
	Non-copyable.
CudaEvent &	operator= (CudaEvent const &)=delete
	Non-copy-assignable.
	CudaEvent (CudaEvent &&other) noexcept
	Move constructor. More...
CudaEvent &	operator= (CudaEvent &&other)
	Move assignment operator. More...
void	record (rmm::cuda_stream_view stream)
	Record the event on a CUDA stream. More...
bool	is_ready () const
	Check if the CUDA event has been completed. More...
void	host_wait () const
	Wait for the event to be completed (blocking). More...
void	stream_wait (rmm::cuda_stream_view stream) const
	Make a CUDA stream wait on this event (non-blocking). More...
cudaEvent_t const &	value () const noexcept
	Access the underlying CUDA event handle. More...
	operator cudaEvent_t const & () const noexcept
	Implicit conversion operator to the CUDA event handle. More...

Static Public Member Functions
static std::shared_ptr< CudaEvent >	make_shared_record (rmm::cuda_stream_view stream, unsigned flags=cudaEventDisableTiming)
	Create and record a CUDA event on a given stream. More...

Detailed Description

RAII wrapper for a CUDA event with convenience methods.

Creates a CUDA event on construction and destroys it on destruction.

Note

To prevent undefined behavior due to unfinished memory operations, events should be used in the following cases if any of the operations below were performed asynchronously with respect to the host:

1. Before addressing a device buffer's allocation.
2. Before accessing a device buffer's data that has been copied from any location, or processed by a CUDA kernel.
3. Before accessing a host buffer's data that has been copied from device or processed by a CUDA kernel.

CudaEvent objects must not have static storage duration, since CUDA resources are not guaranteed to be valid during program initialization or shutdown.

Definition at line 35 of file cuda_event.hpp.

Constructor & Destructor Documentation

CudaEvent() [1/2]

rapidsmpf::CudaEvent::CudaEvent

(

unsigned

flags = cudaEventDisableTiming

)

Construct a CUDA event.

Parameters

flags

CUDA event creation flags.

Exceptions

rapidsmpf::cuda_error

if cudaEventCreateWithFlags fails.

~CudaEvent()

rapidsmpf::CudaEvent::~CudaEvent ( )

noexcept

Destroy the CUDA event.

Automatically releases the underlying CUDA event resource.

CudaEvent() [2/2]

rapidsmpf::CudaEvent::CudaEvent ( CudaEvent &&  other )

noexcept

Move constructor.

Parameters

other

Source CudaEvent to move from.

Member Function Documentation

host_wait()

void rapidsmpf::CudaEvent::host_wait

(

)

const

Wait for the event to be completed (blocking).

Exceptions

rapidsmpf::cuda_error

if cudaEventSynchronize fails.

is_ready()

bool rapidsmpf::CudaEvent::is_ready

(

)

const

Check if the CUDA event has been completed.

Returns

true if the event has been completed, false otherwise.

Exceptions

rapidsmpf::cuda_error

if cudaEventQuery fails.

make_shared_record()

static std::shared_ptr<CudaEvent> rapidsmpf::CudaEvent::make_shared_record ( rmm::cuda_stream_view  stream, unsigned  flags = cudaEventDisableTiming  )

static

Create and record a CUDA event on a given stream.

Convenience factory that constructs a shared CudaEvent with the specified creation flags, immediately records it on the provided stream, and returns it as a std::shared_ptr.

Parameters

stream	CUDA stream on which to record the event.
flags	CUDA event creation flags.

Returns

A shared pointer to the newly created and recorded CudaEvent.

Exceptions

rapidsmpf::cuda_error

if event creation or recording fails.

operator cudaEvent_t const &()

rapidsmpf::CudaEvent::operator cudaEvent_t const & ( ) const

inlinenoexcept

Implicit conversion operator to the CUDA event handle.

Returns

Const reference to the underlying cudaEvent_t.

Definition at line 139 of file cuda_event.hpp.

operator=()

CudaEvent& rapidsmpf::CudaEvent::operator=

(

CudaEvent &&

other

)

Move assignment operator.

Parameters

other

Source CudaEvent to move from.

Returns

Reference to this object.

record()

void rapidsmpf::CudaEvent::record

(

rmm::cuda_stream_view

stream

)

Record the event on a CUDA stream.

Marks the event as occurring after all prior operations on the given stream.

Parameters

stream

The CUDA stream to record the event on.

Exceptions

rapidsmpf::cuda_error

if cudaEventRecord fails.

stream_wait()

void rapidsmpf::CudaEvent::stream_wait

(

rmm::cuda_stream_view

stream

)

const

Make a CUDA stream wait on this event (non-blocking).

Ensures that all operations submitted to the given stream after this call will not begin execution until this event has completed.

Parameters

stream

CUDA stream that should wait for the event.

Exceptions

rapidsmpf::cuda_error

if cudaStreamWaitEvent fails.

value()

cudaEvent_t const& rapidsmpf::CudaEvent::value ( ) const

noexcept

Access the underlying CUDA event handle.

Returns

Const reference to the underlying cudaEvent_t.

---

The documentation for this class was generated from the following file:

• /__w/rapidsmpf/rapidsmpf/cpp/include/rapidsmpf/cuda_event.hpp