New L3 crate effectcraft-gpu: a wgpu compute backend for the per-frame
compositing walk (layer transforms with the CPU's nearest/bilinear/bicubic
sampling and minification pre-filter, motion-blur accumulation, track
mattes, Preserve Transparency, layer style passes and knockout, all 38
blend modes as a WGSL port of color::blend, 8/16 bpc clamp+quantise,
colour-space conversions). 3D runs, adjustment and wireframe layers run on
the CPU between GPU steps. GPU effects with the CPU semantics: Gaussian,
Fast Box and Directional Blur, Glow, Levels, Curves, Hue/Saturation, Tint,
Fill, Gradient Ramp, Fractal Noise, Drop Shadow, Brightness & Contrast,
Exposure, Invert, Transform (registry gpu badge from effects::GPU_EFFECTS).
render: Backend (Cpu/Gpu/Auto) in RenderOpts, the Accelerator trait and
Renderer::accel, GPU effect chains in the effect stack, hooks for the GPU
walk. Project setting gpu_acceleration (Mercury GPU Acceleration / Software
Only), render.backend command, file.projectSettings renderer. Session and
Render Queue exports carry the accelerator. CLI --gpu and bench --gpu.
Tests compare GPU vs CPU renders (blend modes x bit depths, transforms,
masks, mattes, motion blur, colour management, styles, collapse,
adjustment fallback, every GPU effect) and skip without an adapter.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Resolve conflicts with perf (layer cache / Arc<Buf>), export (RenderOpts.view: None
for output), timeline (separate dimensions + one-node POI visibility), mask/pen
viewer gestures (alongside camera tools and view-aware layer matrices), key
dialogs (alongside Camera/Light Settings), render queue + MCP command modules.
3D path tangent honours separated position; DOF blurs a per-frame copy so cached
layer buffers stay shared. New test: cached renders match uncached after camera,
focus, light and 3D view changes. Fix a clone-on-copy lint from main.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- render: LayerCache (content-keyed processed-layer cache: source -> masks ->
effects, keyed by a hash of the layer's non-transform props evaluated at the
frame time, solid item, switches, scale; time folded in for clock-driven
effects / wiggle paths). Static and transform-only layers render once;
edits only re-render the edited layer. Session + UI frame jobs share it.
- effects: TIME_DEPENDENT list + test that every clock-reading effect is listed.
- raster: affine composite fast path (per-row span clipping, unchecked interior
bilinear, integer-translation copies, inlined Normal blend), checked against
composite_warp_reference; box/Gaussian blur without transposes, two
ping-pong buffers; accumulate_warp so motion blur sums into one buffer
instead of one full-frame image per sub-sample.
- ui: viewer frame is urgent and jumps the prefetch queue; stale-revision jobs
are dropped; paused prefetch waits for the viewer frame; "Frame Render Time"
reports viewer frames only; parallel f32 -> Color32 conversion.
- cli: `frame --bench N` (per-layer/per-effect timings, wall + CPU time) and
`frame --bench-play N` (consecutive frames with/without the layer cache).
- tests: cache transparency (scrub, edits of values/keys/effects/masks/
solids/text/timing), hit/miss accounting, time-dependent effects, demo
cached == uncached, fast warp/blur/accumulate == reference.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Brandon Thomas