Wrapper class for cudaStream_t. More...

Public Member Functions
VECMEM_CUDA_EXPORT	stream_wrapper (int device=INVALID_DEVICE)
	Construct a new stream (for the specified device)

VECMEM_CUDA_EXPORT	stream_wrapper (void *stream)
	Wrap an existing `cudaStream_t` object.

VECMEM_CUDA_EXPORT	stream_wrapper (const stream_wrapper &parent)
	Copy constructor.

VECMEM_CUDA_EXPORT	stream_wrapper (stream_wrapper &&parent)
	Move constructor.

VECMEM_CUDA_EXPORT	~stream_wrapper ()
	Destructor.

VECMEM_CUDA_EXPORT stream_wrapper &	operator= (const stream_wrapper &rhs)
	Copy assignment.

VECMEM_CUDA_EXPORT stream_wrapper &	operator= (stream_wrapper &&rhs)
	Move assignment.

VECMEM_CUDA_EXPORT void *	stream () const
	Access a typeless pointer to the managed `cudaStream_t` object.

VECMEM_CUDA_EXPORT void	synchronize ()
	Wait for all queued tasks from the stream to complete.

VECMEM_CUDA_EXPORT std::string	device_name () const
	Get the name of the device that the stream operates on.

Static Public Attributes
static constexpr int	INVALID_DEVICE = -1
	Invalid/default device identifier.

Detailed Description

Wrapper class for cudaStream_t.

It is necessary for passing around CUDA stream objects in code that should not be directly exposed to the CUDA header(s).

Constructor & Destructor Documentation

vecmem::cuda::stream_wrapper::stream_wrapper ( void * stream )

Wrap an existing cudaStream_t object.

Without taking ownership of it!

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