Struct GpuImage 

pub struct GpuImage { /* private fields */ }

An image living in GPU storage buffers (STORAGE | COPY_SRC, plus MAP_READ on unified-memory adapters), on the device of the GpuContext that produced it.

Most images are a single buffer: use GpuImage::buffer to bind it in your own wgpu/WebGPU pipelines, GpuImage::to_texture for an Rgba8Unorm texture, or download it. Images larger than the adapter’s storage-binding limit are tiled (GpuImage::is_tiled): GpuImage::tiles lists one buffer per band with its rectangle in the image (full-width row bands, or full-height column bands for RGBA8 outputs rotated by 90/270 degrees). download and to_texture reassemble tiled images.

Buffers come from FlashRaw’s buffer pool and are returned to the pool when the GpuImage is dropped (large allocations are expensive to recreate). Keep the GpuImage alive while the GPU may still use them, or take ownership with GpuImage::into_buffer / GpuImage::into_tiles. A buffer may be larger than the data it holds.

Implementations

impl GpuImage

pub fn width(&self) -> u32

pub fn height(&self) -> u32

pub fn channels(&self) -> u32

Samples per pixel (1 for CFA/mono, 3 for RGB, 4 for RGBA8).

pub fn format(&self) -> GpuPixelFormat

pub fn len(&self) -> usize

Number of samples (width * height * channels; bytes for RGBA8).

pub fn is_empty(&self) -> bool

pub fn byte_len(&self) -> usize

Bytes of image data (buffers may be larger).

pub fn buffer(&self) -> Option<&Buffer>

The storage buffer holding the whole image (row-major, from byte 0), or None for a tiled image (see GpuImage::tiles). See the type docs for lifetimes.

pub fn is_tiled(&self) -> bool

The image is split into several buffers (it is larger than one storage binding).

pub fn tiles(&self) -> &[GpuTile]

The buffers and the rectangle each one holds (one tile covering the image when untiled). Tiles are disjoint, cover the image, and are ordered by position.

pub fn into_buffer(self) -> Result<Buffer, GpuImage>

Take ownership of the buffer of an untiled image (it will not be recycled). A tiled image is returned unchanged as the error.

pub fn into_tiles(self) -> Vec<GpuTile>

Take ownership of all buffers (they will not be recycled).

pub fn device(&self) -> &Device

pub fn queue(&self) -> &Queue

pub fn context(&self) -> &Arc<GpuContext>

pub async fn download_async(&self) -> Result<(HostPixels, f64)>

Download to host memory (tiled images are reassembled). Returns the pixels and the wall time in ms.

pub async fn download_rgb_async(&self) -> Result<(HostPixels, f64)>

Download a 4-channel 16-bit image (the 16-bit develop outputs) as 3-channel RGB: the fourth sample of every pixel is dropped while copying out of the device buffer, so no 4-channel host copy is made. Other images download unchanged.

pub fn download_rgb(&self) -> Result<(HostPixels, f64)>

Blocking GpuImage::download_rgb_async (native only).

pub fn download(&self) -> Result<(HostPixels, f64)>

Blocking download (native only).

pub fn to_texture(&self) -> Result<Texture>

Copy an RGBA8 image (tiled or not) into a new Rgba8Unorm texture (usage STORAGE_BINDING | TEXTURE_BINDING | COPY_SRC) for sampling/rendering. The copy is queued; it completes before any later submission on the queue.

Trait Implementations

impl Debug for GpuImage

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

Formats the value using the given formatter.

impl Drop for GpuImage

fn drop(&mut self)

Executes the destructor for this type.

fn pin_drop(self: Pin<&mut Self>)

🔬This is a nightly-only experimental API. (pin_ergonomics)

Execute the destructor for this type, but different to Drop::drop, it requires self to be pinned.

impl PartialEq for GpuImage

Identity semantics: two GpuImages are equal when they are the same buffers.

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.