菜单
Activating compact storage formats (R16Float/Rg16Float) for denoiser AOVs requires coordinated BGL + WGSL + texture-format updates across writer and reader passes — change all in lockstep or the bind group will fail validation at runtime
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when working on 3D-specific rendering — deferred chain, G-buffer, clustered shading, Stratum GI, ReGIR, SVGF/A-SVGF, TAA, shadow techniques, depth/HZB, anything 3D-specific. ALWAYS load titan-rendering-core first; this skill assumes core conventions. Loads Titan-specific overrides for realtime-rendering. NOT for 2D rendering or editor panels.
Use when diagnosing a Titan renderer artifact — flicker, ghosting, fireflies, banding, light leaks, seams, boil, NaN/Inf, aliasing, smear/trails, dark voids, shadow artifacts, GI instability, denoiser/TAA issues. The frame-isolation-first debugging method: capture 60Hz, diff frames / A-B vs classical, consult literature, isolate via systematic-debugging, propose (and prove) a fix, record empirically. NOT for landing the production fix (hand to titan-rendering-3d-agent), 2D rendering, or editor panels.
Use when working in titan-rendering-frame-graph, titan-rendering-core, or titan-rendering-post-process-core — frame graph DAG, RenderTarget, DevicePool, PostProcessHook, CameraFrameData, HDR intermediate, tonemap, color management. Foundational layer shared between current 3D and future 2D rendering. Loads Titan-specific overrides for realtime-rendering. NOT for 3D-specific concerns (deferred chain, GI, shadows) — load titan-rendering-3d after this.
When a Titan bench tool using Renderer3d lacks on_resize → renderer.resize(), depth_prepass_view and HZB occlusion resources are never created — DepthPrepass, HzbBuild, and SceneCull passes silently skip, and taps produce no output
When adding a camera UBO shared between Rust and WGSL, verify field-by-field that the CPU layout matches the WGSL struct — especially packed vec4 fields where reordering silently breaks the shader
When building a depth prepass for reverse-Z HZB occlusion culling, use Greater compare (not Always) — the closest fragment must win so the HZB stores the nearest occluding surface, not the farthest
| name | compact-aov-coordinated-activation |
| description | Activating compact storage formats (R16Float/Rg16Float) for denoiser AOVs requires coordinated BGL + WGSL + texture-format updates across writer and reader passes — change all in lockstep or the bind group will fail validation at runtime |
| source | auto-skill |
| extracted_at | 2026-06-15T05:05:48.015Z |
When activating compact storage formats for denoiser AOVs (shading confidence, variance, moments, history confidence), every pass that writes to or reads from the texture must agree on the format. A mismatch panics at bind group creation.
For each AOV storage texture being compacted (e.g., R32Float → R16Float):
TextureFormat::R32Float to
TextureFormat::R16Float in the BindGroupLayoutEntry for the storage
binding.texture_storage_2d<r32float, write> to texture_storage_2d<r16float, write>.render_hook.rs) must create the texture with the matching format.Sampled (non-storage) readers do NOT need BGL format changes — they bind with
TextureSampleType::Float regardless of the underlying format. Only storage
texture writers/readers that declare an explicit format need updating.
| AOV | Writer | WGSL | BGL |
|---|---|---|---|
| shading_confidence | stratum_shading | stratum_shading.wgsl | stratum_shading.rs binding 5 |
| variance_out | svgf_temporal | svgf_temporal.wgsl | svgf_temporal.rs binding 5 |
| moments_curr | svgf_temporal | svgf_temporal.wgsl | svgf_temporal.rs binding 4 |
| history_conf_curr | svgf_temporal | svgf_temporal.wgsl | svgf_temporal.rs binding 10 |
| specular_moments | specular_temporal | stratum_specular_temporal.wgsl | specular_temporal.rs binding 4 |
| specular_variance | specular_temporal | stratum_specular_temporal.wgsl | specular_temporal.rs binding 5 |
After changing all sites, rebuild and run the probe bench. A format mismatch
produces a panic at Device::create_bind_group:
Storage texture binding N expects format = R32Float, but given a view with format = R16Float