Struct MaterialStore 

pub struct MaterialStore { /* private fields */ }

Render-thread GPU storage for packed material bytes, one GrowableBuffer per stride (materials whose packed size differs get separate buffers - see ursus_materials::pack::aos::AosLayout, which this store mirrors on the GPU side).

For V1, exactly one stride is expected to be in use at a time (mesh.frag / depth_prepass.frag / shadow.frag all declare a single set = 1, binding = 0 MaterialBuffer), so descriptor_set binds the first stride’s buffer to that slot. Supporting multiple simultaneous strides on the GPU side (multiple bindings, or a technique-driven choice of which to bind) is future work - see the architecture doc’s SoA/ multi-strategy discussion.

Implementations

impl MaterialStore

pub fn new( device: Device, physical_device: PhysicalDevice, instance: Instance, descriptors: &mut DescriptorAllocator, ) -> Result<Self>

pub fn descriptor_set(&self) -> DescriptorSetId

pub fn upload( &mut self, handle: MaterialHandle, bytes: &[u8], descriptors: &DescriptorAllocator, ) -> Result<()>

Writes bytes (this material’s full packed data, bytes.len() is its stride) for handle, growing/creating the stride’s buffer as needed, re-uploading that stride’s whole bucket to the GPU, and - if this is the first stride seen, or the bound stride’s buffer just reallocated - (re)binding descriptor_set to point at it.

pub fn index_of(&self, handle: MaterialHandle) -> Option<(usize, usize)>

pub fn buffer_for_stride(&self, stride: usize) -> Option<Buffer>

pub fn strides(&self) -> impl Iterator<Item = usize> + '_