A mitered corner reaches up to miter_limit widths past the path, past the
path bounds grown by the width; take the bounds from the compiled stroke.
An empty frame's effects now count its stroke too, as layer_bounds and
composite_bounds already did.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A soft shape's transform frame, snap rectangle and Properties X/Y/W/H came from
the path plus the full stroke width on every side, so a centred or inside stroke
grew the frame when Density or Feather was set. The frame is now the pixels the
same shape renders at full Density without Feather (edge strips only, clipped to
the canvas), identical to the hard shape's for every stroke alignment.
feather_matches_photoshops_ramp now enforces the 1-level claim the scorecard
makes (measured max deviation 1 at 8, 16 and 32 bit).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Baseline x86-64 has no SSE4.1 roundss, so f32::floor is a software routine there, called once per sample of the backdrop under every adjustment layer of an 8- or 16-bit document. branchless_floor rounds through 2^23 and steps down where that went up: bit-identical to floor for every f32 (a new test sweeps every 4099th bit pattern and the edges around whole numbers and 2^23), and with no branch the loop vectorizes.
An adjustment layer's result was blended back into the backdrop pixel by pixel through blend_rgb, which looks up the HDR flag and dispatches on the mode, three channel calls deep, for every pixel. Normal (nearly every adjustment layer) gives the source exactly, so blend_back skips the dispatch for it, and walks the backdrop, the result and the coverage in step instead of indexing by coordinates. Bit-identical.
Baseline x86-64 has no SSE4.1 roundss, so f32::floor is an out-of-line libm call there. lut, which every Levels, Curves, Hue/Saturation range and modern Brightness/Contrast sample goes through, floored its non-negative table position. Truncating is the same there (NaN casts to 0 either way) and inlines. Bit-identical.
color_balance_channel recomputed the folded black point, white point and midtone gamma, and four of the five powf calls of the gamma > 1 Hermite join, for every sample. Those depend only on the sliders, so BalanceCurve works them out once per channel and the per-sample step is the clamp and one powf (or the Hermite polynomial). The CPU compositor builds the three curves once per buffer and the GPU plan once per LUT (it sampled color_balance_channel 3 x 4096 times). Results are bit-identical: a new test compares the curve against the per-sample formula it replaced, to the bit, over 123 slider settings, both Preserve Luminosity modes and a 1/1024 ramp.
docs/file-format-parity.md: main added three 2026-10-15 revision rows
(DDS save, PNG sTER, PSD guides) where the PR added its own; keep all four.
Co-authored-by: varanus-salvator <262257003+varanus-salvator@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Reuse tile scratch and tone tables across rendering work. Skip fully masked
adjustments and identity float curves. Previous-path tests preserve output
across sample depths, masks and clipping.
A Gradient Overlay (and gradient strokes, linked patterns, linked bevel
textures) is laid out over the layer's bounds. Painting grows those
bounds, but the stroke commands reported only the stroke as damage, and
the canvas's committed-stroke path repainted only what the live stroke
had not shown. The canvas therefore kept the old gradient around earlier
strokes: one gradient per stroke instead of one over the whole layer.
Session dispatch now grows a command's reported damage by
photocraft_compose::relayout_damage(before, after): the whole composite
of any visible layer whose effects span its frame, when that frame moved.
The canvas adds the same area to a committed live stroke. Strokes inside
the existing bounds and frame-independent effects keep tight damage.
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
`bake` rendered each layer through `render_layer`, which has no document context (default Global
Light, no patterns, the render rect as canvas). It now uses the new doc-aware `render_layer_in`.
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A pass-through group that is the base of a clipping group renders isolated while it holds a
visible adjustment (`clip_base_adjusts`), so showing, hiding, adding or removing a clipped
adjustment inside it can switch the group's other children (a Multiply layer beside the base)
between isolated and pass-through rendering. `change_bounds_in` now takes the list holding the
parent group and reports that group's change bounds in this case; `layers_damage`, `step_damage`
and `one_sided_bounds` pass it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Parallel bands each collected their finished tiles until the band's last
one was done, so with work stealing many bands could hold most of their
tiles at once, close to a second copy of the output. Each tile is now put
into its band behind a lock as soon as it is rendered; only the tiles in
flight are held, as before the bands ran in parallel.
The compositor draws a clipped adjustment that has no unclipped layer
beneath it as an ordinary adjustment, but change_bounds counted it as
contained in a pass-through group. Hiding the layer clipped to such a
group (which ends its isolation) or the group itself then recomposited
only the group's pixels and left the rest of the canvas stale.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The gloss contour was applied to the highlight and shadow amounts after
the lighting split, so a contour that starts high (Rolling Slope -
Descending) turned every unlit pixel into a full highlight: a bevelled
glyph's flat face painted (255, 40, 42) over its (151, 22, 23) fill. It
now maps the shading before the split, with the flat face's mapped
shading as the neutral, on the CPU and the GPU. Linear is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Port the clipped-adjustment coverage path from #4711 and verify depth, alpha, masks, and document-coordinate GPU parity.
Co-authored-by: Vikram Bhaskaran <vikram.bhaskaran@outlook.com>
A pixel layer with nothing clipped to it is read from its tiles a row at a
time and blended straight onto the backdrop, instead of being copied into a
tile-sized buffer first (a megabyte per layer per 256² render tile). Normal
RGB rows go through a branch-free kernel over blocks of 8 pixels that the
compiler vectorizes; few-band renders also run their bands in parallel.
Bit for bit the same: the row kernel is tested against the scalar loop (NaN,
infinities, subnormals, signed zeros), the row path against the tile copy at
8/16/32-bit with every kind of mode, fill and mask, and the whole GPU parity
suite's CPU references (1620 renders) are unchanged.
P60 on the CPU compositor: 627 -> 530 ms (the render without effects
300 -> 199 ms).
A layer's stand-in surface converted every stored tile to floats before
writing any, a transient several times the layer's stored size (twice,
once per stand-in). Convert about BAND_PIXELS of tiles at a time instead,
and leave hidden layers unconverted, since the compositor never draws them.
An isolated group in one of the special eight blend modes below 100% Fill
blended with Fill as plain coverage on the exact ink path, while the
compositor (and so the split route) keeps Fill apart from opacity, so its
inks differed from the canvas. Blend groups through blend_into_fill like
other layers, and drop the group exclusion from the test reference.
The ink compositor skipped a layer with no tiles in the rect whose default
pixel was all zeros, reading the last value as alpha. A CMYK layer without
an alpha channel defaults to no ink, which is opaque blank paper, so under
an adjustment (Threshold, Invert, ...) its untouched pixels composited as
transparent. Skip only surfaces with an alpha channel and a transparent
default, as `empty_in` does.