Struct bevy_render::batching::gpu_preprocessing::BatchedInstanceBuffers

pub struct BatchedInstanceBuffers<BD, BDI>
where
    BD: GpuArrayBufferable + Sync + Send + 'static,
    BDI: Pod,
{
    pub data_buffer: UninitBufferVec<BD>,
    pub work_item_buffers: EntityHashMap<Entity, PreprocessWorkItemBuffer>,
    pub current_input_buffer: RawBufferVec<BDI>,
    pub previous_input_buffer: RawBufferVec<BDI>,
}

The GPU buffers holding the data needed to render batches.

For example, in the 3D PBR pipeline this holds MeshUniforms, which are the BD type parameter in that mode.

We have a separate buffer data input type (BDI) here, which a compute shader is expected to expand to the full buffer data (BD) type. GPU uniform building is generally faster and uses less system RAM to VRAM bus bandwidth, but only implemented for some pipelines (for example, not in the 2D pipeline at present) and only when compute shader is available.

Fields

data_buffer: UninitBufferVec<BD>

A storage area for the buffer data that the GPU compute shader is expected to write to.

There will be one entry for each index.

work_item_buffers: EntityHashMap<Entity, PreprocessWorkItemBuffer>

The index of the buffer data in the current input buffer that corresponds to each instance.

This is keyed off each view. Each view has a separate buffer.

current_input_buffer: RawBufferVec<BDI>

The uniform data inputs for the current frame.

These are uploaded during the extraction phase.

previous_input_buffer: RawBufferVec<BDI>

The uniform data inputs for the previous frame.

The indices don’t generally line up between current_input_buffer and previous_input_buffer, because, among other reasons, entities can spawn or despawn between frames. Instead, each current buffer data input uniform is expected to contain the index of the corresponding buffer data input uniform in this list.

Implementations

impl<BD, BDI> BatchedInstanceBuffers<BD, BDI>

where BD: GpuArrayBufferable + Sync + Send + 'static, BDI: Pod,

pub fn new() -> Self

Creates new buffers.

pub fn instance_data_binding(&self) -> Option<BindingResource<'_>>

Returns the binding of the buffer that contains the per-instance data.

This buffer needs to be filled in via a compute shader.

pub fn clear(&mut self)

Clears out the buffers in preparation for a new frame.

Trait Implementations

impl<BD, BDI> Default for BatchedInstanceBuffers<BD, BDI>

where BD: GpuArrayBufferable + Sync + Send + 'static, BDI: Pod,

fn default() -> Self

Returns the “default value” for a type.

impl<BD, BDI> Resource for BatchedInstanceBuffers<BD, BDI>

where BD: GpuArrayBufferable + Sync + Send + 'static, BDI: Pod, Self: Send + Sync + 'static,