Module view 

The camera blocks: what the forward pass and the G-buffer pre-pass need to place a fragment, and the per-draw model pair the pre-pass differentiates for motion vectors.

Structs

GBufferModel

Per-draw model matrices for the G-buffer pre-pass. Matches GbModel in shaders/gbuffer_prepass.slang, which Metal and DirectX bind as a constant buffer; Vulkan carries the same pair plus the roughness in one push-constant block (vulkan::uniforms::GbModelPush), because a pipeline layout may declare only one. For a static or skinned object with no motion the caller sets prev == cur.

GBufferView

Per-frame view inputs to the unified G-buffer pre-pass. The jittered current VP drives the rasterised position (matching the main pass); view takes the normal + position into view space (where SSR / SSAO / SSGI / RT work); the un-jittered cur/prev VPs derive a jitter-free motion vector. Matches GbView in shaders/gbuffer_prepass.slang. 256 bytes (four float4x4, all naturally 16-aligned, no padding).

ViewUniforms

Per-frame view-projection uniforms, uploaded once per frame and shared across every draw in it. view is the standalone view matrix the vertex shader uses to compute view-space depth for cascade selection in the fragment shader.