Struct StagingBelt

pub struct StagingBelt { /* private fields */ }

Staging belt is a machine that uploads data.

Internally it uses a ring-buffer of staging buffers that are sub-allocated. It has an advantage over Queue.write_buffer in a way that it returns a mutable slice, which you can fill to avoid an extra data copy.

Implementations

impl StagingBelt

pub fn new(chunk_size: BufferAddress, device: &Device) -> Self

Create a new staging belt.

The chunk_size is the unit of internal buffer allocation. It’s better when it’s big, but ideally still 1-4 times less than the total amount of data uploaded per submission.

pub fn write_buffer( &mut self, target: &Buffer, offset: BufferAddress, size: BufferSize, device: &Device, ) -> BufferViewMut<'_>

Allocate the staging belt slice of size to be uploaded into the target buffer at the specified offset.

The upload will only really be scheduled at the next StagingBelt::flush call.

pub fn flush(&mut self, device: &Device) -> CommandBuffer

Produce a command buffer with all the accumulated transfers.

At this point, all the partially used staging buffers are closed until the GPU is done copying the data from them.

pub fn recall(&mut self) -> impl Future<Output = ()> + Send

Recall all of the closed buffers back for re-usal.

This has to be called after the command buffer produced by flush is submitted!