Commit graph CultLib/packages/cultmath
Author SHA1 Message Date
Meta Vens
00a477e387 cultmath tests: snoise second-difference check tolerates kernel-edge points by a stated fraction
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-14 06:51:42 +02:00
Meta Vens
276484e15b cultmath tests: second differences of snoise compared against a two-level Richardson extrapolation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-14 06:45:41 +02:00
Meta Vens
b8a5a9cc14 cultmath tests: ThirdNorm ascends from the six strongest starts only (cost of MeasureThird)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-14 06:43:57 +02:00
Meta Vens
38473f3fa5 cultmath: snoise_hessian, integral_clamp_linear and SNOISE_THIRD scaffold (not yet measured or regenerated)
math.Span.cs adds the analytic Hessian of snoise(float3) on snoise_grad's tied kernel and the closed-form
integral of a clamped linear ramp; HLSL twins and the Unity copies follow, HessianNorm now reads the
library Hessian, the GLSL lowering gains float3x3 to mat3, and MeasureThird/ThirdConstantPinsSampled and
SpanPrimitiveTests land. SNOISE_THIRD is a placeholder (1) until MeasureThird runs; CultMath.glsl and the
fixture are not yet regenerated. Does not pass its tests.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-14 06:31:43 +02:00
Meta Vens
53f4c21311 CultMath: infinite.json written by oracle.py on Yggdrasil (Blender 5.2.2, SSE4.2); CultMath.Cycles.glsl regenerated for the power parity
infinite.json holds 47 cases of infinite operands for cycles_power, cycles_modulo and Map Range Smooth Step, made inside Blender's kernel; only that family was regenerated. CultMath.Cycles.glsl changed by the generator alone (CULTMATH_WRITE_GLSL=1); glsl-parity.json and CultMath.glsl are byte-identical. This commit builds and its tests are expected green.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-14 04:46:11 +02:00
Meta Vens
77a60d2910 CultMath Cycles: an infinite exponent is even in cycles_power; the oracle's infinite family and the pins for infinite operands
cycles_power read b - 2 floor(0.5 b) != 0 as odd, which is NaN for b = +-inf, so a negative base got a negative result where Cycles' powf(a, +-inf) is never negative. Odd now means the remainder is exactly 1 (C#, HLSL, Unity copy). oracle.py gains the infinite family (operands made in-kernel as +-exp(100), results read through comparisons with 0) for power, modulo and Map Range Smooth Step; the infinite.json fixture and the regenerated GLSL follow in the next commit. This commit builds; InfiniteOperandsFollowBlender and CommittedGlslEqualsLowering fail until it lands.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-14 04:42:39 +02:00
Meta Vens
d8f38057a1 CultMath: the six cycles-oracle fixtures regenerated by oracle.py on Yggdrasil (per-function seeds, arcsine/arccosine/arctan2 rows, overflowing smoothstep rows); CultMath.Cycles.glsl and glsl-parity.json regenerated
Fixtures are the oracle's output (SSE4.2 kernel, Blender 5.2.2 Linux), two runs cmp-identical; CultMath.glsl line 1 (the
Cycles function list in the generated header) carries the three new names.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-14 03:00:32 +02:00
Meta Vens
c65725da8f CultMath Cycles: arcsine, arccosine and arctan2 with Cycles' guards; per-function oracle seeds; the ruled unfused reference, empty Tolerances and the class table
Fixtures and the regenerated GLSL follow in the next commit, so the tests that read the new rows
(InverseTrigonometry..., OperandStreams...) are red at this commit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-14 02:45:51 +02:00
Meta Vens
e48f948892 CultMath: mix.json regenerated with Burn rows that put a colour above 1 in every channel
CyclesOracleTests and the mirror tests pass (29 of 29) on it. Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 10:02:37 +02:00
Meta Vens
3ea9c32382 CultMath oracle: the exact-zero Burn rows put a colour above 1 in every channel
The z channel's boundary mutant survived because 0/0 reads as 0 after the clamp. mix.json follows. Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 10:00:57 +02:00
Meta Vens
e9f6de2164 CultMath: mix.json regenerated with the exact-zero Burn divisor cases
Byte-identical regeneration for the other five families. CyclesOracleTests and the mirror tests pass (29 of 29).
Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 09:51:51 +02:00
Meta Vens
3a83ee361a CultMath: Burn cases with a divisor of exactly 0, and tests for what the film hides (guards returning 0) and for rgb_to_hsv's primaries
Stryker's survivors in math.Cycles.cs that were not equivalent: Burn's divisor <= 0 boundary, Logarithm's
a <= 0 guard, and rgb_to_hsv's h < 0 wrap. mix.json is regenerated in a later commit, so
FamilyMatchesBlender is unchanged here and the new Burn rows are absent. Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 09:50:14 +02:00
Meta Vens
cf6354584f CultMath: CultMath.Cycles.glsl regenerated for the power change
CultMath.Tests passes 386 of 386 at this commit's tree. Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 09:11:09 +02:00
Meta Vens
7d42eb3f20 CultMath: cycles_power has no int-range quirk (Blender returns 1 for (-1)^(2^31)); math.json regenerated with the -1 base and Compare epsilon cases
The oracle showed Cycles treats a huge exponent as integral for a negative base, so the speculative
b >= 2^31 rule is gone. CultMath.Cycles.glsl is stale at this commit (the HLSL changed), so
CommittedGlslEqualsLowering fails here; the regenerated file follows. Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 09:07:53 +02:00
Meta Vens
602c88444d CultMath oracle: power cases with a base of -1 and exponents past the int range, and compare cases inside FLT_EPSILON
The film zeroes an infinite result, which had hidden both the power quirk and Compare's epsilon floor
from the fixtures (two hand mutants of the HLSL survived). The fixtures are regenerated in a later commit.

Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 09:04:28 +02:00
Meta Vens
6196dcde17 CultMath: CyclesOracleTests also holds the HLSL mirror (the shader text compiled as C#) to Blender's results at the fixture operands
A mutant of CultMath.Cycles.hlsl that changes a Cycles result the fixtures reach now fails a test on its own;
before, only the random-input mirror comparison reached the HLSL. Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 08:46:41 +02:00
Meta Vens
abe98fc40a CultMath: the oracle test names its one tolerance (power and logarithm, one ulp for another OS's libm)
Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 06:20:31 +02:00
Meta Vens
8ec0446e33 CultMath: the Cycles oracle fixtures (Blender 5.2.2 LTS d13f752e3b9c, Linux, SSE4.2 kernel), CultMath.Cycles.glsl and the golden-fixture families
Written by tools/cycles-oracle/oracle.py and GlslLowering on Yggdrasil; the tool's second run was
byte-identical. CultMath.Tests passes 379 of 379 at this commit's tree. Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 06:19:42 +02:00
Meta Vens
ddca0252e6 CultMath: a Blender ramp table has 257 entries (size + 1 at i / 256); the oracle samples it so, and the oracle test leaves unread operands unread
Builds. Fixtures and CultMath.Cycles.glsl are still to come.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 06:14:44 +02:00
Meta Vens
892cfbaa6a CultMath: rgb_to_hsv divides by 6 the way Cycles' kernel does (a multiply by 1/6); the ramp is checked by a constant ramp (index) and a linear one (weight); Map Range sockets by their identifiers; the mirror pins 115 functions
Builds. Fixtures and CultMath.Cycles.glsl are still to come.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 05:51:30 +02:00
Meta Vens
51c5da11c4 CultMath: cycles_modulo is exact (power-of-two subtraction, no quotient form); RGB to BW coefficients are Blender's; oracle.py pins the SSE4.2 CPU kernel
The oracle's first fixtures showed Cycles' AVX2 kernel fuses multiply-adds and its film zeroes
non-finite results; the tool now exits 4 off the SSE4.2 kernel (CYCLES_CPU_NO_AVX2=1) and the
test skips cases Cycles cannot report. The Unity copy of CultMath.hlsl carries the new include.

Builds. Fixtures and CultMath.Cycles.glsl are still to come.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 05:28:03 +02:00
Meta Vens
e3c94f8edc CultMath: oracle.py reads Mix and Map Range sockets by identifier; GlslMirrorTests compiles; compile-glsl.ps1 takes -Cycles
Builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 04:41:18 +02:00
Meta Vens
355cfa2a3d CultMath: cycles_ramp_index casts its int operands explicitly (GLSL ES 3.00 has no implicit int to float); README and GLSL file-contract tests name the Cycles files
Builds. The oracle fixtures and CultMath.Cycles.glsl are still to come, so the oracle and golden tests fail here.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 04:07:08 +02:00
Meta Vens
99884cc34c CultMath: the Cycles oracle tool (oracle.py) and CyclesOracleTests
oracle.py bakes each Cycles node under Blender 5.2.2 LTS (build d13f752e3b9c, refusing any other)
into float images; CyclesOracleTests holds math.cycles_* to the fixtures it writes. No fixtures
exist at this commit, so the tests fail here; builds.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 04:04:39 +02:00
Meta Vens
559800f494 CultMath: Cycles node semantics (Math, Map Range, Clamp, Mix, RGB to BW, ramp index) as math.Cycles.cs and CultMath.Cycles.hlsl under Apache-2.0
Ports of Blender v5.2.2's intern/cycles kernel functions, one function per node
operation, with the HLSL mirror, the Unity copy, the lowering table entry for a
separate Apache-2.0 GLSL file, the golden-fixture families, the package licence
expressions and the notices section.

Builds. Does not pass yet: CultMath.Cycles.glsl and the regenerated golden
fixture arrive in a later commit, so CommittedGlslEqualsLowering and
GoldenFixtureMatchesCSharp fail at this commit.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-12 03:59:59 +02:00
Meta Vens
ae3af9feb8 CultMath 0.4.1: Unity plugin DLL and pdb rebuilt on Windows at 0.4.1
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-11 04:48:31 +02:00
Meta Vens
61c2d76de6 CultMath 0.4.1: version, changelog (BoundedLeastSquares stopping fix, Validate), cultlib dependency pin
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-11 04:31:21 +02:00
Meta Vens
b941b68085 CultMath tests: pin the FreeViolated record of the real solves, column by column
The 30 trace cases' FreeViolated lines (column, gradient magnitude, threshold) are pinned by exact equality beside the base-derived decisions they feed, so a free-set test that is mis-called from SolveTraced (a shifted threshold, a short free count) fails here and not only in the hand-built FreeSetViolated tests.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-11 03:04:25 +02:00
Meta Vens
4c1a53f493 CultMath tests: the new trace tests' newline escapes were written as raw newlines; restore them
The previous commit's BoundedLeastSquaresTraceTests.cs did not compile (raw newlines inside string and char literals). The source file compiled; the CultMath.Tests project did not. This commit builds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-11 02:42:25 +02:00
Meta Vens
f8253837ed CultMath BoundedLeastSquares: the free-set stationarity test is a traced decision; the stall boundary is pinned to the ulp
FreeSetViolated is the one owner of the free-set test and records each violating column with its gradient magnitude and the threshold it was judged against (SolveEvent.FreeViolated). Tests pin the column and the threshold by exact equality, one ulp each side of the tolerance at four scales, and the stall comparison f >= mark - allowance at mark == allowance and one ulp each way.

Builds are unverified until the batch's single verify job; this commit has not been compiled.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-11 02:37:05 +02:00
Meta Vens
04c5f870d7 CultMath tests: a trace case for the cap reached right after a release
Stryker found the post-pass cap record unobserved: the only capped case ended on the step site.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-11 00:07:36 +02:00
Meta Vens
08f05dede8 CultMath BoundedLeastSquares: the stopping rule's decisions are traced and pinned; Validate is the one owner of n < 0
The release choice and the pass judgement are internal functions (SelectRelease, JudgePass) and Solve reports each
decision, in order and with the values it compared, to an optional internal ISolveTrace. The tests assert those decisions
with exact equality against the traces the base commit dbcb2a8e produced on the same problems, and call the two functions
directly on states a solve may never reach. Solve's own n < 0 check is deleted (Validate throws the same exception), and the
NonFiniteTarget member doc names the n = 0 case.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-10 22:24:02 +02:00
Meta Vens
a42d7684fb CultMath tests: the rescaled noise-gate problems converge in a few passes, not four
A rescaled problem takes five passes on its own merits; the assertion is that the solve does not spin to the cap.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-10 19:59:50 +02:00
Meta Vens
300fe188d1 CultMath tests: pin the BLS release stop (bls-stop-pins)
W5 with its columns rotated so every column is the released one; the noise gate across column scales; two small problems
whose release is confirmed by a second pass. Untested at this commit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-10 19:47:57 +02:00
Meta Vens
004d8bb78a CultMath BoundedLeastSquares: Validate names the invalid input; Solve's validation has one owner
Adds BoundedLeastSquaresInput and BoundedLeastSquares.Validate (tolerance, bounds, matrix, target; first failure).
Solve calls it and its inline tolerance, bounds and A^T b finiteness checks are deleted. Untested at this commit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-10 19:24:59 +02:00
Meta Vens
dbcb2a8ea8 CultMath BoundedLeastSquares: release on the gradient against its rounding, never by the cost
A bound is released only when its violation exceeds both kktTol and the rounding
bound of its double gradient (Higham gamma_(n+1)); the cost-progress rule now judges
only passes that release nothing, and a release resets fMark. Fixes a cold start
stopping Converged with a violated bound (Soul W5, finding
aetheria-release:finding:cut-thrust-allocator-core.s1.solver-converged-early).
Tests: the W5 problem (cold == re-solve at 1e-12 and 1e-9) and the seed-141 noise
case. Builds; verified on Yggdrasil in the following report.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-10 08:51:18 +02:00
Meta Vens
2855963ca1 CultMath docs: state the float32 tie band in ulps, as measured (Soul s4)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-10 00:12:23 +02:00
Meta Vens
299a4c5018 CultMath docs: magnitude-scaled float32 band; the frustum balls read no noise constant
Docs only; builds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-10 00:04:07 +02:00
Meta Vens
104838ce10 CultMath tests: SnoiseTieScopeTests comment matches its set and the float32 band
Comment only; builds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-10 00:04:07 +02:00
Meta Vens
5384d476ce CultMath smoke: fail unless Unity exited 0 and this run wrote the passing results
Stale logs are deleted before launch; the results file must be newer than the launch. All commits build (no source change).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-10 00:04:06 +02:00
Meta Vens
b88fdccb75 CultMath: commit the Unity import and dispatch smoke, with its broken control
packages/cultmath/scripts/unity-smoke/ holds the good compute shader, the
deliberately broken control, the EditMode test (FindKernel, dispatch, readback),
and run-unity-smoke.ps1, which builds a scratch project outside the repo with a
file: path to the package, takes the Unity path as a parameter, requires all
three tests to pass, and deletes the project. The package README's Unity section
names it as the pre-tag step. Not -nographics: GetComputeShaderMessageCount
reports 0 for a broken shader until FindKernel or Dispatch compiles it. Ran once
on Starfire with Unity 6000.3.24f1: 3 passed, control rejected. Builds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 23:41:04 +02:00
Meta Vens
ab0eacd1c6 CultMath tests: SnoiseTieScopeTests pins each tie-rule site on its own
Asserts per output (value, and each snoise_grad component) that it differs
from 0.3.0 on the tie set, and pins the 0.4.0 values as goldens beside the
0.3.0 ones, so reverting either math.cs:614 or :720 alone fails the test. The
comment says inputs within a few float32 ulps of the set can tie too. Not yet
run when committed; the next commit-free verification run follows.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 23:37:24 +02:00
Meta Vens
d6a31416fd CultMath 0.4.0: name all three empirical constants; float32 ulp neighbours of the snoise tie set
CHANGELOG and package README: the noise enclosures rest on SNOISE_LIPSCHITZ,
SNOISE2_LIPSCHITZ and SNOISE_HESSIAN, all empirical with a margin; the README
bullet no longer says "with a proof" or "every point". Breaking: inputs within a
few float32 ulps of the tie set can also change (probe: none at 1e-5 or more,
a handful at 1e-6 near magnitude 30). Text only; builds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 23:37:24 +02:00
Meta Vens
f9118e1a66 CultMath 0.4.0: the noise-ball enclosures are empirical, not proven
CHANGELOG and README no longer claim iv_snoise_ball and iv_fbm_ball enclose
every input unconditionally: they rest on SNOISE_LIPSCHITZ, an empirical bound
with a margin, and the tests sample. Docs only; builds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 23:07:48 +02:00
Meta Vens
7bd34a96ce CultMath 0.4.0: state the true scope of the snoise tie change
The Breaking entry now names the whole tie set (inputs whose y-x and z-x are
whole numbers, every integer point included, change by up to 0.776), with
examples and a consumer warning; design.md and the NoiseGradTests comment drop
the narrower origin-and-(t,t,t) wording. Docs and test comment only; builds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 23:07:48 +02:00
Meta Vens
96ffb97bbe CultMath tests: pin the snoise tie-rule scope against the 0.3.0 DLL's values
Inputs whose coordinate differences y-x and z-x are whole numbers move from the
0.3.0 value; inputs off that set keep its bits. The goldens were computed by
the cultmath-unity-v0.3.0 tag's tracked CultMath.dll. Builds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 23:07:42 +02:00
Meta Vens
476a5fa491 CultMath: tracked DLL and pdb rebuilt at 0.4.0 on Windows
Copied from artifacts/cultmath-unity-build of the release script's first run
(netstandard2.1, non-incremental, Source Link off, ContinuousIntegrationBuild).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 22:07:05 +02:00
Meta Vens
84c4bb14e3 CultMath 0.4.0: version, changelog, NuGet packs CultMath.Affine.hlsl
Bumps CultMath.csproj, the Unity package.json, the org.gamecult.cultlib
dependency and the semver-policy paragraph to 0.4.0. CHANGELOG states the snoise
tie rule under Breaking and the additions since 0.3.0. The tracked DLL and pdb
still read 0.3.0 at this commit, so the release script's parity check fails here.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 22:03:43 +02:00
Meta Vens
9df0a82faf CultMath tests: pin the KKT tolerance threshold, both decision sites and the forwarded default
The release flips at 1.05x and 0.95x of the computed relative gradient at four
weights either side of 1e-6; all-free conditioned solves change between the
default and 1e-12; the omitted tolerance releases at 1.1e-6 and not at 0.9e-6.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 20:12:08 +02:00
Meta Vens
885b90c355 CultMath: BoundedLeastSquares crefs no longer name the overloaded Solve
Both crefs were ambiguous (CS0419) once Solve had two overloads.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 20:12:08 +02:00