Struct GpuObjectData 

#[repr(C)]
pub struct GpuObjectData {

    pub model: [[f32; 4]; 4],
    pub tint: [f32; 3],
    pub roughness: f32,
    pub emissive: [f32; 3],
    pub metallic: f32,
    pub albedo_index: u32,
    pub normal_index: u32,
    pub emissive_map_index: u32,
    pub orm_map_index: u32,
    pub bb_min: [f32; 3],
    pub cull_distance: f32,
    pub bb_max: [f32; 3],
    pub alpha_cutoff: f32,
}

Per-object record consumed by the Metal “bindless” static main pass.

The static pipeline no longer rebinds model/material/textures per draw call: every object’s transform, material scalars, and texture-pool indices live in one GPU buffer, and the vertex/fragment shaders index it by the object id delivered through [[base_instance]]. That makes each draw call stateless, which is the prerequisite for the GPU-driven compute-cull + indirect-command-buffer pass.

The renderer rebuilds the buffer each frame from its DrawObject list, so update_model / update_visibility changes are picked up automatically. bb_min / bb_max / cull_distance mirror the DrawObject AABB and are unused by the main-pass shaders; they are carried so a GPU-driven compute cull kernel can read object bounds straight from this buffer without a layout change.

Layout (144 bytes) must stay in sync with the shader-side record, declared once in shaders/object_common.slang and spliced into every pass that strides the buffer; shader_layout in concinnity-device pins this struct against slangc’s reflection of it on every target.

albedo_index / normal_index (and the emissive / ORM indices) are indices into each backend’s single handle-indexed texture pool: albedo, normal, and every optional map share one dense table, so a texture used in more than one role resolves to one descriptor. The pool’s two fallbacks occupy reserved slots past the real textures: an object with no normal map addresses the flat-normal entry and one with no albedo addresses the white entry, so neither index needs a shader branch.

Fields

model: [[f32; 4]; 4]

Column-major model-to-world matrix.

tint: [f32; 3]

Linear-space RGB multiplier on the albedo sample.

roughness: f32

Perceptual roughness [0, 1].

emissive: [f32; 3]

Additive emission colour in linear space.

metallic: f32

Metallic factor [0, 1].

albedo_index: u32

Index into the bindless texture pool for this object’s albedo map.

normal_index: u32

Index into the bindless texture pool for this object’s normal map.

emissive_map_index: u32

Bindless-pool index for the emissive map (0 = none). Mirrors MaterialUniforms::emissive_map_index.

orm_map_index: u32

Bindless-pool index for the packed occlusion/roughness/metalness map (0 = none). Mirrors MaterialUniforms::orm_map_index.

bb_min: [f32; 3]

World-space AABB minimum (compute-cull bounds input).

cull_distance: f32

View-distance cutoff; 0 = no cutoff (compute-cull bounds input).

bb_max: [f32; 3]

World-space AABB maximum (compute-cull bounds input).

alpha_cutoff: f32

Alpha-cutout threshold in [0, 1]; 0 disables the test. Mirrors MaterialUniforms::alpha_cutoff.

Trait Implementations

impl Clone for GpuObjectData

fn clone(&self) -> GpuObjectData

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for GpuObjectData

impl NoUninit for GpuObjectData