cuda::experimental::device_memory_resource

class device_memory_resource : public cuda::experimental::__memory_resource_base

Stream ordered memory resource

device_memory_resource uses cudaMallocFromPoolAsync / cudaFreeAsync for allocation/deallocation. A device_memory_resource is a thin wrapper around a c cudaMemPool_t.

Warning

device_memory_resource does not own the pool and it is the responsibility of the user to ensure that the lifetime of the pool exceeds the lifetime of the device_memory_resource.

Public Types

using default_queries = properties_list<device_accessible>

Public Functions

inline explicit device_memory_resource(::cuda::device_ref __device)

Constructs a device_memory_resource using the default cudaMemPool_t of a given device.

Throws:

cuda_error – if retrieving the default cudaMemPool_t fails.

device_memory_resource(int) = delete

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

inline explicit device_memory_resource(

::cudaMemPool_t __pool,

) noexcept

Constructs the device_memory_resource from a cudaMemPool_t.

Parameters:

__pool – The cudaMemPool_t used to allocate memory.

inline explicit device_memory_resource(

device_memory_pool &__pool,

) noexcept

Constructs the device_memory_resource from a device_memory_pool by calling get().

Parameters:

__pool – The device_memory_pool used to allocate memory.

Related

inline friend constexpr void get_property(

device_memory_resource const&,

device_accessible,

) noexcept

Enables the device_accessible property for device_memory_resource.