cuda::experimental::stf::data_place

class data_place

Designates where data will be stored (CPU memory vs.

on device 0 (first GPU), device 1 (second GPU), …)

This typed enum is aligned with CUDA device ordinals but does not implicitly convert to int. See device_ordinal below.

Public Functions

data_place() = default

Default constructor.

The object is intialized as invalid.

inline bool is_composite() const: checks if this data place is a composite data place

inline bool is_green_ctx() const: checks if this data place is a green context data place

inline bool is_device() const: checks if this data place corresponds to a specific device

inline decorated_stream getDataStream(async_resources_handle &async_resources) const

Public Static Functions

static inline data_place device(int dev_id = 0)

Data is placed on device with index dev_id.

Two relaxations are allowed: -1 can be passed to create a placeholder for the host, and -2 can be used to create a placeholder for a managed device.

static inline data_place current_device(): Select the embedded memory of the current device as data_place.

template<typename partitioner_t> static data_place composite(partitioner_t p, const exec_place_grid &g)

static inline data_place composite(get_executor_func_t f, const exec_place_grid &grid)

Public Static Attributes

static const data_place invalid: Constant representing an invalid data_place object.

static const data_place host: Constant representing the host CPU as the data_place.

static const data_place managed: Constant representing a managed memory location as the data_place.

static const data_place affine: This actually does not define a data_place, but means that we should use the data place affine to the execution place.

static const data_place device_auto: Constant representing a placeholder that lets the library automatically select a GPU device as the data_place.

Friends

inline friend size_t to_index(const data_place &p)

Returns an index guaranteed to be >= 0 (0 for managed CPU, 1 for pinned CPU, 2 for device 0, 3 for device 1, …).

Requires that p is initialized and different from data_place::invalid.

inline friend int device_ordinal(const data_place &p)

Returns the device ordinal (0 = first GPU, 1 = second GPU, …

and by convention the CPU is -1) Requires that p is initialized.

class composite_state

Public Functions

inline composite_state(exec_place_grid grid, get_executor_func_t partitioner_func)