Struct Device 

pub struct Device { /* private fields */ }

Where a program runs.

Cloning is cheap: the GPU handle is shared, so two clones name the same adapter and the same uploaded buffers.

Implementations

impl Device

pub fn cpu() -> Device

The processor everything already ran on.

pub fn default_gpu() -> Option<Device>

The machine’s preferred adapter, or None where there is none.

The adapter is opened once per process and shared; asking twice costs nothing and hands back the same device.

pub fn with_precision(&self, p: Precision) -> Device

The same device, computing in p.

Precision::F32 is an explicit request to compute a f64 program in single precision. It is the only way a machine whose shaders have no f64 runs anything at all on its GPU.

pub fn precision(&self) -> Precision

pub fn is_gpu(&self) -> bool

pub fn info(&self) -> Option<&DeviceInfo>

What this device is, or None for the CPU.

pub fn upload(&self, y: &Array) -> Result<Array, DeviceError>

y with its elements resident on this device.

The result is an ordinary array — same shape, same values, readable by anything — that additionally holds the device allocation, so a run that reaches the device with it uploads nothing. Uploading to the CPU is the identity.

pub fn holds(&self, y: &Array) -> bool

Is this array already resident on this device, at this precision?

Trait Implementations

impl Clone for Device

fn clone(&self) -> Device

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Device

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.