Plan transcribed #3C4646 as rgb(60,70,90)=#3C465A (20 off on blue);
spec §1 specifies rgb(60,70,70). Also add module-level egui-0.35
version-drift note and hex comments on inline widget-state colors
for grep-able spec↔code comparison.
Palette constants + sci_fi_visuals() + setup(). Hand-applied over
Visuals::dark(); no custom font, no new deps. Not yet wired into
ui_system (Task 4 does that via setup_egui_style).
Higher default turbulence frequency gives the volumetric disk more
trackable texture features at startup, making the Keplerian rotation
(rot ∝ 1/r^1.5) visually readable without needing to open the egui
panel. The old 0.8 default produced a near-featureless smooth slab at
default settings — the disk was rotating but the eye had nothing to lock
onto. 2.0 sits in the upper half of the slider range (0.2–3.0) while
still leaving room to push higher for debugging.
Planets were landing on/in the accretion disk. Root cause: orbit radius was
r = radius_factor · kerr_isco(χ), keyed off the disk's INNER edge. The disk
extends from ISCO out to disk_outer (default 25), so for every spin the
planets' radii fell inside the disk's radial span:
χ=0: planets r ∈ [4.5, 17.25], disk [3, 25] — all inside
χ=1: planets r ∈ [0.75, 2.88], disk [0.5, 25] — all inside, near hole
Re-keyed to the disk OUTER edge: r = radius_factor · disk_outer. With the
factor now meaning "multiple of disk_outer" (default 1.3, slider 1.1..=2.0,
per-planet jitter ±0.25 floored at 1.05), planets stay at r ≈ 32.5 —
reliably outside the disk regardless of spin, while still close enough for
lensing to curve their images.
Changes:
- orbit_position: new disk_outer param, r = radius_factor · disk_outer
- orbit_system: passes params.disk_outer
- spawn_planet_system: jitter range tightened to ±0.25 around the factor,
floor 1.05 (was ±0.75 around an ISCO-keyed k, floor 1.0)
- params default planet_radius_factor 2.5 → 1.3 (new semantics)
- UI: slider 1.5..=5.0 → 1.1..=2.0, label "Radius (× disk outer)",
readout shows orbit r and disk_outer instead of ISCO
- tests updated for the new signature and r = factor · disk_outer invariant
The "Radius factor k" UI slider wrote params.planet_radius_factor and
triggered respawn via the dirty flag, but spawn_planet_system ignored it —
every planet's radius_factor came from rng.gen_range(2.0..4.0), so the
slider was a silent no-op (it changed the ISCO readout label and respawned,
but the actual orbit radii never reflected its value).
Now samples radius_factor as planet_radius_factor ± 0.75, so the slider
controls the orbital scale while preserving per-planet radius diversity
(no fully-identical radii). Clamped to >=1.0 so jitter near the slider's
low end can't drop a planet inside the ISCO.
Caught by the Phase 3.4 final code review.
Adds kerr_nodal_precession(r, chi) = Ω_φ − Ω_θ using the Okazaki (1987)
vertical epicyclic frequency:
Ω_θ² = Ω_φ² · (1 − 4a√M/r^1.5 + 3a²/r²)
χ=0 → a=0 → Ω_θ = Ω_φ → zero precession (exact Schwarzschild degeneracy).
Note on the cross term: it is a√M/r^1.5 (radius to the −1.5), NOT a·Ω_φ/r.
The latter makes Ω_θ too large and breaks the "precession grows monotonically
with spin" property — caught by the grows_with_spin test.
Three tests: vanishes at zero spin, grows monotonically with spin at fixed r,
and exceeds the weak-field 2aM/r³ approximation in the strong-field (r<6)
region. Full suite green (14 tests).
Two distinct sampling artifacts on the accretion disk, found with a
CPU pixel-mirror of the integrator:
1. Moiré on the disk. integrate_disk_segment placed its N sub-samples
on the deterministic (i+0.5)/N grid. That lattice aliases against
the 2-D pixel grid and the per-ray RK45 step lattice → a concentric
banding. Fix: jitter the sub-sample position within stratum i using
a hash of the pixel coordinate and the sample index. The jitter is
pixel-seeded only (no time), so the noise is spatially stable (no
per-frame flicker) and stays inside the stratum, preserving the
quadrature order. Verified by CPU mirror: banding transitions drop
sharply and the noise is unstructured.
2. Higher-order lensed-image rings read as an artifact. Rays near the
critical impact parameter wrap around the hole; each wrap crosses
the disk plane once, projecting to a ring (a physical Einstein-ring
structure, the same one Gargantua shows). The integrator renders
each wrap as a sharp discrete band, so at high `steps` the rings
look like a bug. Fix: per-pixel stochastic supersampling — fire
several jittered sub-rays per pixel and average them, which
antialiases the band edges into a smooth gradient while keeping the
underlying lensing physics.
The two fixes share a `pixel_seed` (var<private>, set from
@builtin(position)) so the jitter is deterministic per pixel.
Shader refactor: the single-ray march is extracted into march(dir);
fragment() now loops it MAX_AA_SAMPLES times (compile-time cap for
WGSL's static-loop-bound requirement, runtime count from a new
aa_samples uniform via early break — the fbm3/MAX_OCTAVES pattern).
Uniform layout: added aa_samples: u32 + two f32 pads to round the
struct to 16-byte alignment, mirrored byte-for-byte across the WGSL
struct and BlackHoleUniforms (ShaderType).
UI/params: new AaQuality enum (Off/Low/High = 1/2/4 samples) in the
Quality panel. Defaults are tiered per AGENTS.md: Off on web, Low (2×)
on desktop — the RK45 march is already ~10× Phase 1's cost, so desktop
stays at 2× with High (4×) selectable for a fully smooth look.
cargo test: 20 passed. cargo build --release: clean.
sample_jets suppresses jets for χ < 0.05 (previous commit), but the egui
panel's checkbox still flipped freely and Strength stayed editable at low
spin, so enabling jets there did nothing with no explanation — the toggle
looked broken.
Mirror the shader's gate in the UI: keep the Enabled checkbox interactive
(it expresses user intent), gray out Strength when jets won't render
(matching the existing pattern used for disk quality, Doppler, color
model, and bloom), and show a one-line hint ("Spin (χ) too low — jets
need χ ≥ 0.05.") so the no-op is self-documenting.
Two independent sources of noise in the web console, both from the
default plugin set:
1. 'Failed to deserialize meta for asset shaders/*.wgsl'
DefaultPlugins adds AssetPlugin with meta_check = Always, so bevy
fetches <path>.meta for every shader. The shaders ship raw (no .meta
files); on the trunk dev server those requests don't return a clean
404 that bevy maps to NotFound, so bevy receives bytes and tries to
RON-deserialize them as AssetMetaMinimal, logging a SpannedError per
shader. Set meta_check = Never: it skips the meta lookup and uses the
loader's default meta — correct for processor-free raw assets.
2. 'AudioContext was not allowed to start'
The default AudioPlugin opens a WebAudio sink at startup; browsers
block that until a user gesture and log it as a JS error. This app
has no audio, so disable AudioPlugin entirely — removes the noise and
shrinks the wasm binary.
Defaults retuned to match the Interstellar look now that the density
model drives appearance:
- disk_outer 15 → 25: Gargantua-like extent.
- disk_half_thickness semantics → H/R ratio, default 0.15 (was an
absolute 0.3/0.2 world-space band). Standard thin-disk scale height.
- disk_color_mode default → Blackbody (was Gradient): the Novikov-Thorne
radial temperature gradient × Kerr Doppler gives the smooth white-hot
inner → deep-orange outer look the new density model is built around.
- disk_temp 10000 → 6500 K: tuned so the NT profile yields warm-white
inner fading to deep orange at the edge.
- density_strength 1.0 → 1.2: the new model has no floor and decays to 0
at the edges, so a slightly higher multiplier keeps the bulk opaque
after per-step integration.
UI: "Half thickness" → "Thickness (H/R)" with a 0.02..=0.3 range
(thin-disk regime); "Outer radius" range widened to 6..=50.
BlackHoleUniforms::Default (material.rs) mirrors the same values so the
GPU default matches params.rs before the first mirror_params tick. Field
order/types/layout unchanged; no uniform-struct migration needed.
- params.rs: collapse `if cfg!(wasm) { false } else { true }` for
star_aa into `!cfg!(wasm)` (clippy::needless_bool).
- render/plugin.rs: `resize_offscreen` is a Bevy system whose argument
count is fixed by its SystemParam set; allow clippy::too_many_arguments
alongside the existing type_complexity allow on the same function.
The control panel grew past one screen after the disk/jet additions
(Color model, Temperature, Jets section). egui::Window without a height
constraint just clips overflowing content. Give it a default width/height,
enable resizing, and wrap the body in ScrollArea::vertical so all sections
remain reachable.
Bring three accretion-disk strengths from others/blackhole-simulation into
the WGSL renderer, behind opt-in controls. The physics.rs CPU mirror is
untouched — these are all color/shading functions, outside the deriv/rk45/
integration-loop lockstep surface.
Color model (DiskColorMode uniform, default Gradient to preserve the prior
look):
- Gradient (mode 0): existing white-hot → orange ramp + Newtonian
apply_doppler, unchanged.
- Blackbody (mode 1): Novikov-Thorne radial temperature profile
(isco_r^0.75 · (1−√isco_r)^0.25) shifted by the Kerr equatorial
4-velocity Doppler factor δ, fed through a Tanner-Helland blackbody
color function, times δ^3.5 beaming. The approaching side therefore
turns hotter/bluer and the receding side cooler/redder — the physical
signature the gradient mode lacks. kerr_doppler() solves g_μν u^μ u^ν=-1
for circular orbits exactly; the max(0.01,…) floor guards pow-of-negative
NaN. A unified disk_emission() now assembles color for both the flat and
volumetric paths, replacing two duplicated tcol/falloff/doppler chains.
Relativistic jets (jets_enabled, default on): bipolar cones along the spin
axis with 0.92c outflow beaming (δ^3.5, no floor needed — β fixed keeps the
denominator positive), Gaussian radial falloff, exponential length decay,
and outward-flowing value-noise turbulence. Front-to-back composited into
the same accumulators as the disk. Ported to ptr<function,…> accumulators
(WGSL has no inout) and the existing value_noise3.
UI: Color model dropdown + temperature slider under Accretion Disk; new
Jets section with enable + strength. Four uniform fields added as one
16-byte row, Rust/WGSL order kept identical.
Verified: cargo build --release, cargo test (17 tests, mirror untouched),
naga WGSL validation.
Set the OrbitCamera default to the framing the UI screenshot shows:
yaw -1.065, pitch -0.335, distance 30, fov 1.0 (the latter two were
already the defaults).
The old default (yaw 0, pitch +0.7) was a 3/4 view from above the disk
plane. The new pitch sits the eye just below the disk plane, so the
gravitationally lensed far side of the disk arcs over the top of the
shadow — the iconic Interstellar framing. yaw -1.065 (~-61°) picks a
viewing azimuth that keeps the lensed arc and photon ring balanced in
frame.
Live-tunable volumetric disk params. Quality ComboBox (Off/Low/Medium/
High) mirrors the Bloom quality pattern; the 8 sliders disable when Off.
Off reverts to the flat disk for perf escape + visual A/B.
Extends mirror_params with the 9 new field assignments. DiskQuality enum
gains an as_u32() accessor for the WGSL tier selector. GPU now receives
live values; shader does not consume them yet.
Data layer for the volumetric disk. DiskQuality gates octave counts
(Off/Low/Medium/High); Off is a full revert to the flat disk. Eight new
tunable f32 fields for thickness, filament, density, and spiral-arm
control. Web defaults to Low tier + 0.2 half-thickness; desktop to
High + 0.3. Nothing reads these yet.
Two runtime validation errors crashed the app on startup:
1. BrightPassMaterial used #[uniform(0)] threshold: f32 (bare f32 = 4-byte
binding), but the WGSL BrightPassUniform struct is 16 bytes (threshold +
3 pads). The pipeline-layout mismatch ("Buffer structure size 16...
greater than min_binding_size, which is 4") killed the opaque_mesh2d
pipeline. Fix: wrap threshold in a BrightPassUniform struct matching the
WGSL layout, same pattern as BlurUniform/CompositeUniform.
2. bevy_egui 0.41's auto-context assigns PrimaryEguiContext to the FIRST
Added<Camera> — with the 7-camera bloom pipeline, that's the offscreen
camera (renders to Rgba16Float Image). egui's pipeline expects the
window's Rgba8UnormSrgb surface → format mismatch crash. Fix: disable
auto_create_primary_context (PreStartup system) and explicitly tag the
composite (window) camera with PrimaryEguiContext.
Both fixes verified: the app now runs without validation errors for 8+
seconds. Physics tests (20/20) unaffected.
Adds a dedicated Quality collapsible section: bloom quality dropdown
(Off/Low/Medium/High), bloom threshold/strength, exposure, resolution
scale, star AA toggle. Changing bloom quality at runtime triggers a
pipeline rebuild (despawn bloom entities + respawn via spawn_bloom_pipeline
helper). MSAA honestly labeled as decorative.
Note: the rebuild currently re-spawns the full 3-level pyramid whenever
bloom is enabled (Off vs On). Partial pyramids for Low/Medium are a
future enhancement.
BloomQuality { Off, Low, Medium, High } controls pyramid depth (0-3).
Web defaults to Low (1 level, minimal float textures), desktop to High
(3 levels, full cinematic). Off skips all bloom passes and composites
scene-only with ACES — the fallback path for WebGPU browsers without
float-filterable support. The composite samples a 1x1 black texture
when bloom is off.
Note: this task gates Off vs full-On (3-level pyramid always runs when
enabled). Partial pyramid for Low/Medium is deferred to the runtime
rebuild in Task 8.
Adds CompositeMaterial + composite.wgsl: combines the HDR scene with the
bloom pyramid output, applies ACES (Narkowicz) tone mapping, and writes
LDR to the window. Removes UpscaleMaterial + upscale.wgsl.
mirror_params now updates composite (bloom_strength, exposure) and
brightpass (threshold) uniforms each frame. resize_offscreen rebuilds
the full bloom pyramid targets on window resize.
Adds BlurMaterial + blur.wgsl with a 13-tap weighted Gaussian kernel
(var<private> arrays — naga rejects dynamic indexing of const arrays),
run in downsample then upsample passes. For the 3-level pyramid:
bloom_0 (half, from brightpass) → bloom_1 (quarter) → bloom_2 (eighth),
then upsample to bloom_final (half). Camera orders -18/-17/-16/-15.
The bloom output is not yet composited (Task 6); this commit only wires
the passes and confirms they run in sequence without crashing.
Adds BrightPassMaterial + brightpass.wgsl: extracts luminance above a
soft-knee threshold from the HDR offscreen into a half-res Rgba16Float
target (bloom_0). Camera order -19 (after offscreen at -20). The brightpass
output is not yet composited (Task 6); this commit only adds the pass and
confirms it runs without crashing.
star_color's smoothstep(0.5,0.0,d) with scale=80 produced blocky rectangles
at low render_scale. Replace with a gaussian speck: distance to cell center
with exp(-d²/r²) falloff, producing a round 2-3 pixel anti-aliased disk.
star_aa uniform toggles between the gaussian path (desktop default) and the
old square-cell fast path (web default).
Also adds bloom_threshold/bloom_strength/exposure uniform fields to
BlackHoleUniforms (used by later bloom tasks) with safe defaults.
Switch the offscreen target from Bgra8UnormSrgb to Rgba16Float for HDR
headroom (disk brightness can exceed 1.0 without clamping). Set offscreen
camera order to -20 (the bloom pipeline, added next, needs cameras ordered
so offscreen renders first). Introduce three marker components:
- Nudgable: all render cameras carry it; nudge_camera now queries this
instead of a hardcoded Or<(OffscreenCamera, UpscaleCamera)>.
- QuadScaleFactor(fx, fy): fraction of offscreen res each quad fills;
resize_offscreen rescales each quad independently via this.
- CompositeQuad: the final quad renders to the window (rescaled against
window res, not offscreen).
The upscale path still works (textureSample reads float textures);
ACES tone-mapping is added in a later commit.
At 1.318 rad (~75.5 deg) the disk is nearly vertical: its normal points
almost along +Z while the camera yaw axis is world-Y, so the two axes
sit ~75 deg apart. Orbiting via yaw (a horizontal circle around world-Y)
therefore shears the disk through view -- full circle, squashed ellipse,
edge-on sliver, back circle -- which reads as the disk swinging or the
scene rolling.
0.45 rad (~25.8 deg) brings the disk near-horizontal (normal approx
(0, 0.90, 0.43)), shrinking the yaw-axis / disk-normal misalignment
from ~75 deg to ~26 deg. The swing is correspondingly muted; orbiting
feels like turning around the hole rather than shearing across a wall.
Both defaults are updated in lockstep: BlackHoleParams (the live runtime
default) and BlackHoleUniforms (the bootstrap uniform filled before the
first mirror_params frame).
basis() derived `right` from `forward × world_up`, which degenerated to
a zero vector when forward aligned with world-Y (pitch = ±π/2) — the
cross product vanished, normalization produced garbage, and crossing
the pole flipped the image 180°.
Right now depends on yaw alone: Vec3::new(cos yaw, 0, -sin yaw). It is
a unit vector in the y=0 plane for any yaw, so it never goes to zero
regardless of pitch. In the original upper hemisphere (cp > 0) this is
exactly the old value after normalization, so existing views are
unchanged; it additionally stays perfectly horizontal for all pitch,
fixing a subtle roll the cross-product form introduced below the equator.
With the singularity gone, pitch is unclamped from [0.05, π-0.05] to
[-π+0.05, π-0.05]: the camera can now cross both poles and reach below
the disk plane without flipping. The ±0.05 margin only avoids the exact
origin-axis points where forward = ∓Y and yaw becomes meaningless.
yaw is also wrapped to (-π, π] on each drag so it cannot drift out of
the UI slider range (which would snap the view when the slider is later
touched); cos/sin periodicity makes the wrap visually lossless.
The UI pitch slider range is widened to match. Stale "gimbal pole"
comments that described the old restriction are corrected.
basis() still returns (eye, forward, right, up), so the uniform bridge
in plugin.rs and both shaders are untouched.
Phase 2 shipped the Kerr deriv() and the adaptive RK45 integration loop in
the shader (black_hole.wgsl:260, 320-390), but the CPU mirror in physics.rs
only covered the single-step Kerr derivative (kerr_bending_accel), not the
loop. That left the AGENTS.md 'CPU <-> shader mirror' contract broken for
Phase 2: the spin-dependent capture radius, the adaptive step control, and
the budget = accepted-steps-only / dt_min forced-accept semantics existed
only on the GPU, with nothing testable on the CPU side.
Add a faithful CPU mirror and the loop-level tests that close the gap:
- rk45_step(): Dormand-Prince step mirroring black_hole.wgsl:260-285, using
kerr_bending_accel so it is also the Phase 1 step at chi=0.
- is_captured_rk45(): the full adaptive loop mirroring black_hole.wgsl:320-390
— same seeding (total_path / steps), same reject/retry at dt_min (the
forced-accept floor that prevents infinite retry), same accept-then-refine,
and the spin-dependent capture radius r+(chi) via kerr_horizon.
Tests (tests/physics_test.rs):
- spin=0 RK45 loop captures below bcrit and escapes above (degeneracy).
- spin=0.9 still captures a b~2.0 ray (horizon shrinks but not past b<bcrit).
- capture set does not grow as spin increases across a bcrit-straddling sweep.
- rk45_step error shrinks monotonically with dt (the property the adaptive
loop's reject/accept decision depends on).
Note: probing the error scaling showed it falls between 2nd and 4th order in
dt, not a clean 5th, because of the per-stage normalize() projection the
shader (and now the mirror) applies. The loop only relies on monotonicity, so
this is correct-as-shipped; documented in a doc-comment on rk45_step so a
future reader is not misled by the '45' in the name.
cargo test: 17 passed (3 Phase 1 inline + 14 integration, up from 12).
Fixes the following clippy warnings in plugin.rs:
1. clippy::too_many_arguments: Suppressed on spawn_fullscreen_quad since Bevy system functions naturally take many arguments as resources/queries.
2. clippy::field_reassign_with_default: Refactored instantiation of BlackHoleMaterial to set fields in the struct literal instead of post-creation assignment.
3. clippy::type_complexity: Suppressed on resize_offscreen and nudge_camera systems as Bevy queries with Filters/Disjunctions are inherently complex.
Clippy warned that the manual Default implementation for BlackHoleMaterial
could be derived (clippy::derivable_impls). We now derive Default directly
on the struct and remove the manual implementation block.
Fixes clippy::approx_constant errors caused by using the hardcoded value 3.14
for camera yaw and accretion disk tilt sliders. Using std::f32::consts::PI
resolves the errors and uses the precise constant value.
Clippy warned about excessive precision in the BCRIT float literal (2.5980762).
Since f32 has 24 bits of mantissa (approx 7 decimal digits of precision), the
excessive decimal digits are truncated to 2.598_076 to compile cleanly under
clippy::excessive_precision.
The integration test still passes as the difference is well within the 1e-5 limit.
resize_offscreen declared two Query<&mut Transform> params (FullscreenQuad
and UpscaleQuad filters). Bevy's conflict checker does not treat With<T>
filters as disjoint access — both queries declare write access to the same
component, so the app panicked at startup with error[B0001].
Merge the two queries into a ParamSet (borrowed one at a time) and move
them to the end of the parameter list per the Params convention.
- Set default pitch to 0.7 rad to keep the camera away from the π/2 gimbal pole, preventing horizontal drag from degenerating into an in-plane roll.
- Adjust scroll zoom logic to divide instead of multiply by the zoom factor, so that scrolling up zooms in (reduces distance) rather than out.
