Span stages in C#, HLSL and GLSL: SpanPath/SpanStage inputs, float-rounding allowances on curvature and fold, zero-length flat spans, pinned shader guards, exact noise pins, compile-hlsl.ps1 with -WX, struct-aware shader mirror. Soul verdicts cut-stage-functions.s2-s4 and cut-stage-shaders.s1 hold; main's interim changes are cultcache-rs and docs only, disjoint from the verified C#/shader tree.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The additive 1e-9 absorbed the whole exact departure at scales 1e-4 and 1e-2 (1e-11 to 1e-9),
so dropping the value allowance's absolute term 2^-14 survived the test. The tolerance is 1e-13.
Test only; builds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
dxc runs with -WX; a ShaderPath that is not a file throws naming the parameter and entry index,
never the path; the GLSL wrapper builds a struct parameter from its declaration's field zeros.
design.md and README state the struct-argument twin rule, the span twins and the missing shader identity.
No C# change; the compiles run on Starfire after the last commit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CultMath.glsl gains the three structs and cultmath_span_stage/fbm/ridged by the lowering; the fixture
gains the two span families after the last one, earlier families byte-identical (0 deleted lines).
Builds; CultMath.Tests 452 passed at this tree.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
cultmath_span_stage, cultmath_span_fbm and cultmath_span_ridged with SpanPath, SpanStage and
SpanNoise, statement for statement from math.span_stage (no shader identity). They follow the
includes, not cultmath_ridged_grad: the bounds read the SNOISE_* constants the includes define.
The mirror comparison matches every mirror function to the math function of the same name,
public or private, with struct parameters mapped by name and fields; MirrorOnly is gone.
New tests: identity and realistic-span mirror sweeps, a large-octave-step fold test; the
near-origin fold test gains scales 1e-4 and 1e-2; the integral_clamp_linear tests also read the
committed fixture's cases.
The GLSL, the fixture families and the docs follow: CommittedGlslEqualsLowering and
GoldenFixtureMatchesCSharp are red at this commit. It builds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The spec's negative check greps the script for arm64; the comment matched. Comment
only; script rerun on Starfire (compiles both copies, exit 0; arm64 dxc control throws). Builds.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Stryker survivor on the previous head: direction t1 <= t0 changes a zero-length
sloped span from +0 to -0, which no test observed. Builds; test-only.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
glsl: line 307 only (the flat-branch guard). Fixture: one family appended
('zero-length spans and flat infinite clamps'), earlier families byte-identical.
Builds. Together with the previous commit the lowering and fixture tests pass.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The b == 0 branch now guards t1 != t0 like the sloped branch (C# and both HLSL
copies), so an infinite bound times a zero length no longer forms NaN.
Edge-grid mirror test and a counting-clamp shadow test replace the text-order
check; the four pin tests assert equality with their measured expression;
tools/compile-hlsl.ps1 is the one dxc compile verdict for the HLSL copies.
Builds. CultMath.glsl and glsl-parity.json are not yet regenerated, so
CommittedGlslEqualsLowering and GoldenFixtureMatchesCSharp are red at this
commit; the next commit regenerates them.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Two samples of the probe: the cut's own (seed not recorded) and an audit's (seed 0x501) give 0.057/0.046/0.075 and 0.064/0.0485/0.057 of e_n/e_g/e_h; a new seed moves them. Comment only in math.Span.cs: builds.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The fold test's float arm no longer re-spells the octave coordinate (OneOctave): each octave's float value is the difference of span_fbm / span_ridged at o + 1 and o octaves, with a stated bound on what the accumulating float sum loses.
New fixed span at halfLength 6 pins the Gl reach operator (nu / halfLength fails); a near-zero midpoint on a stationary path (halfLength 0, and halfLength 1 under a loose speed bound) pins the halfLength-0 fold rule and S_o's reach term.
Builds: yes (to be confirmed by the Yggdrasil job).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Kills the Stryker survivors of the first allowance commit: the velocity stretch sign, the norm at
a zero axis, gain-weight and acceleration-factor arithmetic, the octaveStep term of S_o. Not yet
built or run.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
span_fbm/span_ridged/span_stage lose the 14-argument signatures for SpanPath, SpanStage,
octaves, lacunarity, gain. Per octave S_o and e_n, e_g, e_h (af_snoise's derivation plus the
Hessian's 2^-10) widen the local Hessian tier, the gradient bound and the fold travel. Tests
compare against the exact double kernel in coordinate bands, add the pinned rank-one span,
widen the stage parameter ranges, and state the lab-form mapping. Not built or run yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The decode_stack Windows run is at 10a39dac (tests unchanged since 4b55c7d3);
the head after this commit awaits a Soul rerun.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
SpanFoldCoversAccelerationAlone draws spans with zero midpoint velocity and an acceleration along the gradient
that carries n across zero, so the fold's acceleration term and its gradient bound are the only thing holding
the branch departure under fold; two hand mutants survived without it.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Legacy and v1 reads on 64 KiB, redb pull_all and compare-and-swap on 104 KiB,
and one Largest level on 88 KiB, the sizes where a plain from_slice or a missing
outer maybe_grow dies in release as well as debug.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The README and decode_payload claimed the calling thread needs only its own
frames. Measured, the door has a payload-independent floor (release 56 KiB vs
16 for a same-signature non-decoding function; debug 104 vs 88). Say so, and
replace "Windows not yet verified" with the one by-hand Starfire run of
decode_stack on x86_64-pc-windows-msvc.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
SpanStageGradientMatchesDifferences now also compares field.w and field.xyz with an octave sum assembled
in the test from Stage.Point and snoise_grad, so shift, offset, octaveStep, M, lacunarity and gain are pinned
numerically, which a value/gradient self-consistency check cannot do.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The octave stages of a warp chain over a span, in C#: stretched, shifted, offset octave sums whose
field equals fbm_grad and ridged_grad bit for bit at the identity transform and whose curvature
and fold bounds derive from SNOISE_LIPSCHITZ, SNOISE_HESSIAN and SNOISE_THIRD only (local tier from
the midpoint Hessian's Frobenius norm plus the third-derivative term, capped by the global pin).
SpanStageTests encloses the Hessian norm and the branch departure over seeded spans in both regimes
and checks the gradient against differences and the stretch/axis conventions. README and design.md
state the span, the transform and the bound rules. No shader twin: stage-shaders is part 2.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
lo = min(lo, hi) now precedes the b == 0 branch in C# and HLSL (Unity copy byte-identical, GLSL
regenerated), so no clamp ever sees lo above hi; GLSL leaves that undefined. The b == 0 path used to
call clamp(a, lo, hi) with the raw bounds. New tests: the flat and sloped lo > hi values, the
statement-order check on the committed HLSL and GLSL, and an appended glsl-parity.json family of lo > hi
and infinite-bound edge cases (every earlier case byte-identical).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Each held stretch contributes only when its length is positive, so an infinite bound times a zero
length never forms: relu (hi = +inf) and its mirrors give the finite integral instead of NaN, in C#,
HLSL (Unity copy byte-identical) and the generated GLSL. Finite cases are unchanged.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
ThirdTensor is now snoise_hessian's per-corner derivation differentiated once more, in double, on the
kernel's lattice, tie order and hash; central differences of the float Hessian are gone from the
measurement and survive only as the oracle in ThirdTensorMatchesDifferencesOfTheHessian (Richardson
extrapolated differences of the double Hessian). MeasureThird on the closed form: refined 487.0917,
x 1.10 = 535.8009 (the differenced tensor read 488.1047, its roundoff). SNOISE_THIRD, ThirdWitness,
CULTMATH_SNOISE_THIRD (both HLSL copies), the generated GLSL constant and the design.md provenance
paragraph follow.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The compiled mirror declares its static consts without public access; the first version of
EveryShaderConstantEqualsItsCSharpTwin found none and failed. Reflect NonPublic too.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
EveryShaderConstantEqualsItsCSharpTwin reflects each CULTMATH_* literal of the compiled HLSL
mirror and asserts it is bit-equal to the math constant of the same name, so CULTMATH_SNOISE_THIRD
(and every other constant) edited in one runtime fails.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Measured on Yggdrasil: the decode needs 56 KiB (Wide32) and 80 KiB (Largest) in
debug on a thread that returns a bool; returning the record through
std::thread's own frames adds several copies of it, so the tests box it and
size threads at 96 KiB and from 160 KiB.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
decode_grown checks the stack before any frame that holds T, then serde_stacker
at every container; decode_payload walks the nesting bound and calls it; the
legacy envelope array, redb reads and v1 snapshots decode through it too.
Red zone 512 KiB against MAX_RECORD_INLINE_BYTES = 16 KiB, which the
DatabaseEntry, CultSlots and CultUnion derives assert at compile time.
cultcache-rs-derive 0.2.0.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Their results carry the curvature constant; the platform-bound families stay as generated on Windows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
