Struct Array3D 

pub struct Array3D { /* private fields */ }

A 3-D CUDA array. Element format chosen at creation; channels are 1/2/4.

Implementations

impl Array3D

pub fn new( context: &Context, width: usize, height: usize, depth: usize, format: ArrayFormat, num_channels: u32, ) -> Result<Self>

Allocate a plain width × height × depth 3-D array.

pub fn with_flags( context: &Context, width: usize, height: usize, depth: usize, format: ArrayFormat, num_channels: u32, flags: u32, ) -> Result<Self>

Allocate with custom flags (layered, cubemap, surface-ldst, etc. — see baracuda_cuda_sys::types::CUarray3D_flags).

pub unsafe fn from_borrowed( context: &Context, handle: CUarray, width: usize, height: usize, depth: usize, format: ArrayFormat, num_channels: u32, ) -> Self

Wrap an existing array handle (e.g. a mipmap level) without taking ownership. Drop is a no-op; the parent owns lifetime.

Safety

handle must be a valid CUarray that outlives this wrapper.

pub fn as_raw(&self) -> CUarray

pub fn width(&self) -> usize

pub fn height(&self) -> usize

pub fn depth(&self) -> usize

pub fn bytes_per_element(&self) -> usize

Element width in bytes (channel size × channel count).

pub fn copy_from_host<T: DeviceRepr>(&self, host: &[T]) -> Result<()>

Synchronous host → 3-D array copy. host.len() must equal width × max(height, 1) × max(depth, 1).

pub fn copy_to_host<T: DeviceRepr>(&self, host: &mut [T]) -> Result<()>

Synchronous 3-D array → host copy.

pub fn copy_from_host_async<T: DeviceRepr>( &self, host: &[T], stream: &Stream, ) -> Result<()>

Async variant of Array3D::copy_from_host on stream. Note: CUDA requires the host buffer to be pinned for real async behavior — otherwise the driver falls back to a sync copy.

Trait Implementations

impl Clone for Array3D

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Array3D

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

Formats the value using the given formatter.