cuda::device::peer_to_peer Namespace Reference

API functions and definitions relating to communications among peer CUDA GPU devices on the same system. More...

Typedefs
using	attribute_t = CUdevice_P2PAttribute
	While Individual CUDA devices have individual "attributes" (attribute_t), there are also attributes characterizing pairs; this type is used for identifying/indexing them.

Functions
bool	can_access (const device_t &accessor, const device_t &peer)
	Determine whether one CUDA device can access the global memory of another CUDA device. More...
void	enable_access (const device_t &accessor, const device_t &peer)
	Enable access by one CUDA device to the global memory of another. More...
void	disable_access (const device_t &accessor, const device_t &peer)
	Disable access by one CUDA device to the global memory of another. More...
bool	can_access_each_other (const device_t &first, const device_t &second)
	Determine whether two CUDA devices can currently access each other.
void	enable_bidirectional_access (const device_t &first, const device_t &second)
	Enable access both by the first to the second device and the other way around.
void	disable_bidirectional_access (const device_t &first, const device_t &second)
	Disable access both by the first to the second device and the other way around.
attribute_value_t	get_attribute (attribute_t attribute, const device_t &first, const device_t &second)
	Get one of the numeric attributes for a(n ordered) pair of devices, relating to their interaction. More...

Variables
constexpr const attribute_t	link_performance_rank = CU_DEVICE_P2P_ATTRIBUTE_PERFORMANCE_RANK
	Aliases for CUDA driver GPU attribute codes. More...
constexpr const attribute_t	access_support = CU_DEVICE_P2P_ATTRIBUTE_ACCESS_SUPPORTED
	1 if access is supported, 0 otherwise
constexpr const attribute_t	native_atomics_support = CU_DEVICE_P2P_ATTRIBUTE_NATIVE_ATOMIC_SUPPORTED
	1 if the first device can perform native atomic operations on the second device, 0 otherwise

Detailed Description

API functions and definitions relating to communications among peer CUDA GPU devices on the same system.

Function Documentation

can_access()

bool cuda::device::peer_to_peer::can_access ( const device_t &  accessor, const device_t &  peer  )

inline

Determine whether one CUDA device can access the global memory of another CUDA device.

Parameters

accessor	device interested in making a remote access
peer	device to be accessed

Returns

true iff acess is possible

disable_access()

void cuda::device::peer_to_peer::disable_access ( const device_t &  accessor, const device_t &  peer  )

inline

Disable access by one CUDA device to the global memory of another.

Parameters

accessor	device interested in making a remote access
peer	device to be accessed

Todo:

Consider disabling this, given that access is context-specific

enable_access()

void cuda::device::peer_to_peer::enable_access ( const device_t &  accessor, const device_t &  peer  )

inline

Enable access by one CUDA device to the global memory of another.

Parameters

accessor	device interested in making a remote access
peer	device to be accessed

Todo:

Consider disabling this, given that access is context-specific

get_attribute()

attribute_value_t cuda::device::peer_to_peer::get_attribute ( attribute_t  attribute, const device_t &  first, const device_t &  second  )

inline

Get one of the numeric attributes for a(n ordered) pair of devices, relating to their interaction.

Note

This is the device-pair equivalent of device_t::get_attribute()

Parameters

attribute	identifier of the attribute of interest
first	the device accessing an att
second	destination device

Returns

the numeric attribute value

Variable Documentation

link_performance_rank

constexpr const attribute_t cuda::device::peer_to_peer::link_performance_rank = CU_DEVICE_P2P_ATTRIBUTE_PERFORMANCE_RANK

Aliases for CUDA driver GPU attribute codes.

A relative value indicating the performance of the link between two devices