occlusion computes its three channels as one vector instead of writing a
vector component through a loop index, which FXC cannot address and tried
to unroll every enclosing loop for (X3550/X3511). classic3d now also
builds like the effect kernels: if a compiler rejects it, 3D runs render
on the CPU instead of the whole compositor failing to start. The ignored
FXC diagnostic test now checks classic3d instead of the stale pointwise
failure.
Resolve the conflict in crates/gpu/src/fx_stylize.rs: main moved ops::warp to
the ops::WarpOptions parameter struct; the branch's per-sample transform loop
(motion-blur samples, now passing the sample by reference) uses the new
signature with the branch's reference argument kept.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Accelerator::comp_frame_yuv420 (default None) renders a top-level comp frame and
converts it over the background to the H.264 export's 8-bit BT.709 limited-range
4:2:0, every byte within one level of the CPU conversion (the GPU kernel follows
it operation for operation, including NaN and half-away-from-zero ties, but a
driver may contract its arithmetic). A 4K frame then reads back 12 MB of planes
instead of 133 MB of RGBA f32.
The H.264 export asks the accelerator for planes when a frame needs nothing after
the render (progressive, project depth, no crop/resize, rendered at the output
size) and otherwise renders and converts on the CPU as before. Deferred
readbacks (a browser worker) keep the pass-based path. Sizes the kernel can't
take (width not a multiple of 8, odd height) are declined before any GPU work.
Like Cx::frames, h264_frames caps GPU frames in flight only while the
accelerator is up (map_frames_while with the same liveness check): once the GPU
stops for good, the remaining frames render on the CPU side by side instead of
one at a time.
rgba_to_yuv420 moves to effectcraft_raster::yuv so the export and the GPU tests
share one reference. The GPU tests skip without an adapter (tests::gpu()), as
the crate's others do.
The WGSL grammar test scanner indexed token characters directly, so a malformed shader could panic the test instead of failing it. Checked access turns that into a reported violation.
Poison recovery, OsStr::to_str and char predicates only; S4325 casts are
f64 to f32 and stay.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Content r from Center smears out to r / (1 - s), so the buffer is padded by
the farthest corner's reach (capped at 4096, skipped on adjustment layers),
on the CPU and GPU paths alike.
* cargo fmt
* GPU tests: llvmpipe on Vulkan gets the boundary-flip allowance; skip the Reshape boundary cases that crash lavapipe (#2238)
* GPU tests: the llvmpipe boundary-flip allowance is an opt-in of the two tests that need it
Shadow maps use the full scene-derived size (point-light faces capped
at 2048) under a 26M-texel scene budget halved largest-first; the PCF
filter measures its reach in output pixels with a two-texel minimum,
samples texel centres on a jittered adaptive 5..10 grid, and compares
each texel against the receiver plane's own depth so wide kernels never
self-shadow lit casters. The WGSL mirror is parenthesised (Tint rejects
the old form) and guarded by a shader-wide grammar test; the CPU
reference and shader remain step-for-step parity-tested.
Both devices EffectCraft creates now ask for `MemoryHints::MemoryUsage`:
- the window device (`apps/effectcraft/src/main.rs`, through a new `device_descriptor` helper);
- the headless compositor device (`crates/gpu/src/context.rs`).
The limits, labels, features and error paths are unchanged. The devices tests create stay as they were.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The gradient antialias could leave a half-covered pixel where an opaque source reaches h = 1. Add an endpoint guard on CPU and GPU and cover an opaque edge in the regression test.
Use the measured Softness kernel (sigma = 0.42 * Softness), threshold the blurred height with a ~1 px antialias, and allow Cut Away 100 to empty the matte. Split the source-matte displacement from the height-field RGB refraction: the matte uses a z=1 bevel normal fitted to the T2 step edges, while RGB follows the measured z=height refraction. Align the documented AE defaults and add CPU regression tests for the cut threshold, Cut Away 100, and the inward matte warp. Update the GPU kernel and parameters to match, keeping CPU/GPU parity.
On an adjustment layer Motion Tile kept the comp-sized buffer whatever the
Output Width / Height, and the adjustment draw dropped any stack result whose
size differed from the canvas. After Effects grows (or crops) the comp
rectangle there as on any layer, so a Transform after it brings the extra
tiles into view, as in the frames attached to #1727 and #1409.
- motion_tile_frame: one output frame for the CPU and GPU effect; on an
adjustment it grows about the comp rectangle on whole pixels, keeps the
buffer at 100 % and past 16 384 pixels a side.
- adjustment_crop: the CPU and GPU adjustment draws put a stack result whose
frame changed back onto the canvas instead of dropping it (Enc::crop on
the GPU).
Tests: the #1727 After Effects frames (910 x 340 % and 90 x 340 % then
Scale 12 %), tiles past the frame and cropping on CPU, GPU vs CPU for
adjustment and full-frame layers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kernels.wgsl repeats Glow's bright-pass gain 8.0 and knee 0.0611 as literals; they now
name effects::GLOW_GAIN / GLOW_KNEE (exported) and the test
glow_constants_match_the_cpu (no GPU needed) fails if they drift. docs/effects-parity.md:
the summary row no longer says no effect is compared with After Effects while the
Blur & Sharpen row reports Radial Blur's comparison (CodeRabbit on #1873), and the
Radial Blur changelog row says CPU only (it has no GPU path).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
WGSL cannot read Rust constants, so the Rec. 601 luma (Tint, Tritone, Luma Key /
Extract), Black & White's tint weights and Exposure's 2.2 power stay literals in the
kernels; they now name their Rust source (effects::LUMA_601, BW_TINT_LUMA,
EXPOSURE_LINEAR_GAMMA, newly exported) and the test shader_constants_match_the_cpu
(no GPU needed) fails if a literal drifts. gpu_tint_defaults_match_after_effects_greys
says "skipped: no GPU adapter" when it skips (EFFECTCRAFT_REQUIRE_GPU=1 still makes it
fail). docs/effects-parity.md: Brightness & Contrast is close to, not the same as,
After Effects (up to 6.2/255), and the summary row lists what is measured.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After Effects' Curves effect interpolates its points with a natural cubic
spline (zero curvature at the ends), clamped to 0..1: AE 2026 renders of an
RGB curve (0,0) (125,199) (255,255) and a Red curve (0,0) (128,44)
(255,255) (cases 962 / 963, ROUND4) give exactly the natural spline's values
on every grey step (0, 30, 59, 115, 164, 202, 227, 242, 250, 255 and 0, 0, 1,
6, 19, 44, 83, 135, 194, 255). EffectCraft used a monotone (Fritsch-Carlson)
spline: 34 / 67 / 125 / 171 ... on the RGB curve, up to 10/255 off (ROUND4
measured up to 6 levels on real projects); now within 1/255, CPU and GPU (the
same Hermite tables carry the spline's knot slopes). Lumetri's curves and
Glow's Arbitrary Map keep the monotone spline (not measured).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After Effects draws the Transform effect as position + K * R * S * (p - anchor)
with the skew K = rot(-axis) * shearY(tan skew) * rot(axis) applied after the
rotation, and keeps what lands outside the layer's bounds. EffectCraft skewed
before rotating (shearX about the axis) and clipped to the layer.
Measured in AE 2026 on a grey 960x540 solid (cases 723, 819, 820): 723 (anchor,
position, scale 120/80, skew 10 axis 20, rotation 30, opacity 70) differed in
11.25 % of the frame and now in 0.065 % (edges, max 16/255), CPU and GPU alike;
819 / 820 (skew 15 axis 60, rotation -25, scale 130/70, anchor) match to the
edge. The buffer grows on each side by that side's overshoot (also for the
motion-blur samples), capped at 16 384 px a side and 16 Mpx in all so a layer
moved far away costs a bounded allocation (regression test); adjustment layers
keep their size.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After Effects 2026 draws Fractal Noise's Block noise at Scale 100 and
Complexity 1 in cells of 66.67 px (the value changes every 66.67 px along a
row), not 100 px; Soft Linear noise has the matching correlation length.
EffectCraft's lattice is now Scale x 2/3 px (NOISE_CELL), CPU and GPU. Case
904 E03 (Soft Linear, Complexity 4, Scale 100): the noise's autocorrelation
at 32 px was 0.68 (AE 0.41), now 0.54; pixel values cannot match (other
seeds). Still open: AE's noise spreads about 20 % wider around its mean (std
48 vs 40/255 in E03). Existing projects and the Backgrounds presets with
Fractal Noise get finer noise (two thirds of the former cell size).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After Effects feathers both edges of the wiped wedge (the start ray too) with
a Gaussian whose width is constant in pixels: Completion 40, Start Angle 30,
Counterclockwise, Feather 20 (AE 2026, case 904 E12) fits sigma = 0.4 x
Feather = 8.0 px at every radius (mean alpha error 0.0002 over the frame).
EffectCraft ramped only the moving edge, linearly over Feather degrees, so
the edge was 14 px wide near the centre and 140 px wide at 400 px out, and
the start edge was hard: 378 of 5184 grid points off (max 111/255, 6.6 % of
pixels > 8), now 1 (max 3), 0.009 % of pixels > 8 (near the centre). CPU and
GPU (the kernel now measures distances to the two rays). 0 % and 100 % are
unchanged, so the three Transitions presets keep their end frames; their
Feather (6 / 10) now means pixels.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After Effects measures the blinds' phase from the layer's centre and softens
both edges of the transparent band with a Gaussian of sigma = Feather / 2; the
band runs from the origin's phase to Completion x Width. EffectCraft measured
from the buffer corner and ramped only one edge linearly over Feather (a hard
edge on the other side). Fitted to AE 2026 renders of Completion 40,
Direction 30, Width 50, Feather 5 on a 960 x 540 and a 480 x 270 layer (the
origin is the centre + 0.6 px): case 715 was 499 of 5184 grid points off
(max 91/255, 7.1 % of pixels > 8), now 28 (0.056 %), all within ~6 px of the
layer edge where AE differs (up to 48/255, not modelled); away from it within
6/255 (test: one AE row of the 480 x 270 render). CPU and GPU (new erf-based
kernel, same term count). Completion 0 / 100 are unchanged, so the
Transitions presets keep their start and end frames.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After Effects' Radial Blur keeps a layer's edges: Zoom 20 on a flat grey 960x540
solid over red (case 713) stays flat and opaque to its edge, and a zoomed
full-frame ramp (case 829) keeps the frame opaque (2/255 at x = 2, 253/255 at
x = 1917 on row 540). EffectCraft sampled transparency outside the layer, so its
edges faded and the red below showed through (713: 3.9 % of the frame off, max
64/255). Now 713 is identical to AE (max 0/255), CPU and GPU. CC Radial Blur
(already matching AE) is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With Glow Colors = Original Colors, After Effects thresholds each channel of
the straight colour separately and the glow is 8 x (value - threshold +
0.061) per channel (times Intensity), nothing below: fitted to AE 2026
renders of grey 0.65 / 0.8 / 0.9 / 1, orange (1, 0.5, 0) and (0.2, 0.6, 1)
solids at Threshold 40 / 60 / 80 % and Radius 10 / 30 (36 channel values,
exact to the 8-bit capture; an orange solid glows red only). EffectCraft
scaled the colour by a luminance threshold (k = (L - thr) / (1 - thr)), about
3.7 x too dim on white. Cases 806 / 810 / 811 (32 bpc): outside the layer
now within 3/255 of AE (806 white at Intensity 1: 119 = AE, was 32); grid
0 points off for 806-810. The glow's alpha (the blurred layer alpha) and the
Arbitrary Map / A & B colours and alpha-based glows are unchanged.
Stacked on fix-glow-halo (same lines). Not changed: Composite Original's
semantics are swapped against AE's names (EffectCraft's "On Top" adds the
glow over the original, which AE calls "Behind"; AE's "On Top" keeps the
original over the glow), so imported AE Glows at the default Behind show
no glow inside the layer (811 interior 204 vs AE 255).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With Glow Based On Color Channels, After Effects gives the glow the layer's whole
alpha (blurred), not the thresholded part, and Intensity scales only its colour:
a layer below the threshold darkens what lies under its edge. Fitted to AE 2026
renders of a grey solid over red (cases 524, 807-809): the halo is the same at
Intensity 1 and 3 and with Composite Original Behind or On Top, and its edge is
an erfc step of sigma = 0.384 x Glow Radius at radius 10 and 30 (was radius / 2).
524 (Glow defaults on a 50 % grey solid): EffectCraft had no halo (max 116/255
off at the edge); now red 248/234/208/165/140 at 7.5/5.5/3.5/1.5/0.5 px outside
vs AE 249/235/206/164/140, max 2/255 over the frame, CPU and GPU. The blur
width applies to every glow; alpha-based glows keep their coverage rule
(thresholded alpha, scaled by Intensity). Not matched yet: the brightness of glows above the
threshold (AE about 3.7 x brighter for a white solid, cases 806, 810, 811).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Without a Fill Mask, After Effects turns the layer into the fill colour and
applies Opacity to its alpha; nothing of the original shows through. AE 2026
renders Fill (0, 0.5, 1) at Opacity 50 % on a Gradient Ramp over a red solid
as premultiplied (0, 0.251, 0.502) at alpha 0.502 at every pixel; EffectCraft
mixed the fill with the ramp (mean difference 95.5/255), now within 1/255,
CPU and GPU. Invert fills the transparent part the same way. With a Fill
Mask the mix with the original is unchanged (not measured). Lottie's fill
effect (colour, alpha x opacity) now plays the same in EffectCraft.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After Effects weighs each of the four points by its inverse squared
distance softened by Blend squared, 1 / (d^2 + Blend^2) in pixels: fitted to
AE 2026 renders of the default gradient at Blend 10, 100 and 300 and case
904 E16 at Blend 50 (power 2.00 at every Blend, softening 0 / 2367 / 9817 /
90127 = about Blend^2; mean error 0.27/255, max 0.9). EffectCraft used
1 / (d^2 / (100 Blend))^1.5, in which Blend cancels out: E16 was 34/255 off
on 99 % of the grid, now 0 grid points off, max 1/255, CPU and GPU.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After Effects 2026 keys Luma Key and Extract (Channel Luminance) on Rec. 601
weights of the working-space values: Key Out Darker / Extract Black Point at
65, 166 and 24 on pure red, green and blue keep red (601: 76, 709: 54) and
blue (29 / 18) and key out green (150 / 182) - Rec. 709, which EffectCraft
used, decides all three the other way. Case 904 E18 (Extract 30 / 220 /
softness 10 on grey and colour steps): 513 of 5184 grid points off (pure blue
kept at 231 in AE, keyed out in EC), now 0, max 0. CPU and GPU.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bflatastic