A structure holding a collection various properties of a device. More...

Inheritance diagram for cuda::device::properties_t:

Collaboration diagram for cuda::device::properties_t:

Public Member Functions
	properties_t (const cudaDeviceProp &cdp) noexcept

	properties_t (cudaDeviceProp &&cdp) noexcept


bool	usable_for_compute () const noexcept
	Convenience methods for accessing fields in the raw parent class, or combining or converting them into a more useful type.

compute_capability_t	compute_capability () const noexcept

compute_architecture_t	compute_architecture () const noexcept

pci_location_t	pci_id () const noexcept

unsigned	max_in_flight_threads_on_device () const

grid::block_dimension_t	max_threads_per_block () const noexcept

grid::block_dimension_t	max_warps_per_block () const noexcept

size_t	max_shared_memory_per_block () const noexcept

size_t	global_memory_size () const noexcept

bool	can_map_host_memory () const noexcept

Detailed Description

A structure holding a collection various properties of a device.

Note: Somewhat annoyingly, CUDA devices have all of attributes, properties and flags. Attributes have integral number values; properties have all sorts of values, including arrays and limited-length strings, and flags are actually associated with a device's primary context, as it is actually contexts which have flags (which are either binary or small-finite-domain type fitting into an overall flags value: {context::flags_t}). Flags and properties are obtained all at once (the latter, using the runtime API), attributes are more one-at-a-time.

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

Public Member Functions