mirror of
https://github.com/mihakralj/QuanTAlib.git
synced 2026-08-12 23:58:04 +00:00
Miha Kralj
67ad6f0cba
Comprehensive refactor across all indicators replacing the periodic ResyncInterval-based drift correction (every 1000 ticks recalculate from scratch) with Kahan compensated summation for running sums. Key changes: - Remove ResyncInterval constants and TickCount fields from all State records - Add Kahan compensation fields (SumComp, SumSqComp, etc.) to State records - Replace naive sum += val - removed with Kahan delta pattern - Remove Resync()/RecalculateSum() methods that did O(N) recalculation - Update batch/SIMD paths to use Kahan compensation instead of resync loops - IIR filters (EMA, REMA, RGMA) simplified: inherently self-correcting - Version bump to 0.8.7 - Build system: README version stamping via Directory.Build.props - Minor doc/test tolerance adjustments for new numerical characteristics Affected modules: channels, core, cycles, dynamics, errors, momentum, oscillators, statistics, trends_FIR, trends_IIR, volatility, volume
702 lines
24 KiB
C#
702 lines
24 KiB
C#
using System.Runtime.CompilerServices;
|
|
using System.Runtime.InteropServices;
|
|
using System.Runtime.Intrinsics;
|
|
using System.Runtime.Intrinsics.X86;
|
|
|
|
namespace QuanTAlib;
|
|
|
|
/// <summary>
|
|
/// Kurtosis: Measures the tailedness (heaviness of tails) of the probability distribution
|
|
/// of a real-valued random variable.
|
|
/// </summary>
|
|
/// <remarks>
|
|
/// This implementation calculates excess kurtosis (kurtosis - 3), so a normal distribution
|
|
/// has excess kurtosis of 0.
|
|
///
|
|
/// Interpretation:
|
|
/// - Positive (leptokurtic): Heavier tails than normal, more extreme events
|
|
/// - Zero (mesokurtic): Normal distribution tail behavior
|
|
/// - Negative (platykurtic): Lighter tails than normal, fewer extreme events
|
|
///
|
|
/// Formula (population excess kurtosis):
|
|
/// g₂ = m₄ / m₂² - 3
|
|
///
|
|
/// where:
|
|
/// m₂ = (1/n) Σ(xᵢ - μ)² (second central moment)
|
|
/// m₄ = (1/n) Σ(xᵢ - μ)⁴ (fourth central moment)
|
|
///
|
|
/// Sample excess kurtosis applies Fisher's correction:
|
|
/// G₂ = ((n-1)/((n-2)(n-3))) * ((n+1)*g₂ + 6)
|
|
///
|
|
/// Implementation uses O(1) running sums of powers (x, x², x³, x⁴) with Kahan
|
|
/// compensated summation for numerical stability over long streams, eliminating
|
|
/// the need for periodic resynchronization.
|
|
/// </remarks>
|
|
[SkipLocalsInit]
|
|
public sealed class Kurtosis : AbstractBase
|
|
{
|
|
private readonly int _period;
|
|
private readonly RingBuffer _buffer;
|
|
private readonly bool _isPopulation;
|
|
private double _sum;
|
|
private double _sumSq;
|
|
private double _sumCu;
|
|
private double _sumQu;
|
|
private double _sumComp;
|
|
private double _sumSqComp;
|
|
private double _sumCuComp;
|
|
private double _sumQuComp;
|
|
private double _p_sumComp;
|
|
private double _p_sumSqComp;
|
|
private double _p_sumCuComp;
|
|
private double _p_sumQuComp;
|
|
private const double Epsilon = 1e-10;
|
|
|
|
public override bool IsHot => _buffer.IsFull;
|
|
|
|
/// <summary>
|
|
/// Creates a new Kurtosis indicator.
|
|
/// </summary>
|
|
/// <param name="period">The lookback period (must be >= 4).</param>
|
|
/// <param name="isPopulation">If true, calculates Population Kurtosis. If false, Sample Kurtosis (default).</param>
|
|
public Kurtosis(int period, bool isPopulation = false)
|
|
{
|
|
if (period < 4)
|
|
{
|
|
throw new ArgumentOutOfRangeException(nameof(period), "Period must be greater than or equal to 4 for Kurtosis.");
|
|
}
|
|
_period = period;
|
|
_isPopulation = isPopulation;
|
|
_buffer = new RingBuffer(period);
|
|
Name = $"Kurtosis({period})";
|
|
WarmupPeriod = period;
|
|
}
|
|
|
|
/// <summary>
|
|
/// Creates a chained Kurtosis indicator.
|
|
/// </summary>
|
|
/// <param name="source">The source indicator to chain from.</param>
|
|
/// <param name="period">The lookback period.</param>
|
|
/// <param name="isPopulation">If true, calculates Population Kurtosis.</param>
|
|
public Kurtosis(ITValuePublisher source, int period, bool isPopulation = false) : this(period, isPopulation)
|
|
{
|
|
ArgumentNullException.ThrowIfNull(source);
|
|
source.Pub += HandleInput;
|
|
}
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
private void HandleInput(object? sender, in TValueEventArgs e)
|
|
{
|
|
Update(e.Value);
|
|
}
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
public override TValue Update(TValue input, bool isNew = true)
|
|
{
|
|
// Snapshot current state for rollback
|
|
double p_sum = _sum;
|
|
double p_sumSq = _sumSq;
|
|
double p_sumCu = _sumCu;
|
|
double p_sumQu = _sumQu;
|
|
_p_sumComp = _sumComp;
|
|
_p_sumSqComp = _sumSqComp;
|
|
_p_sumCuComp = _sumCuComp;
|
|
_p_sumQuComp = _sumQuComp;
|
|
|
|
if (isNew)
|
|
{
|
|
if (_buffer.IsFull)
|
|
{
|
|
double oldVal = _buffer.Oldest;
|
|
double oldSq = oldVal * oldVal;
|
|
|
|
// Kahan subtract oldVal from _sum
|
|
double y = -oldVal - _sumComp;
|
|
double t = _sum + y;
|
|
_sumComp = (t - _sum) - y;
|
|
_sum = t;
|
|
|
|
// Kahan subtract oldSq from _sumSq
|
|
y = -oldSq - _sumSqComp;
|
|
t = _sumSq + y;
|
|
_sumSqComp = (t - _sumSq) - y;
|
|
_sumSq = t;
|
|
|
|
// Kahan subtract oldCu from _sumCu
|
|
y = -(oldSq * oldVal) - _sumCuComp;
|
|
t = _sumCu + y;
|
|
_sumCuComp = (t - _sumCu) - y;
|
|
_sumCu = t;
|
|
|
|
// Kahan subtract oldQu from _sumQu
|
|
y = -(oldSq * oldSq) - _sumQuComp;
|
|
t = _sumQu + y;
|
|
_sumQuComp = (t - _sumQu) - y;
|
|
_sumQu = t;
|
|
}
|
|
|
|
double val = input.Value;
|
|
if (!double.IsFinite(val))
|
|
{
|
|
val = _buffer.Count > 0 ? _buffer.Newest : 0;
|
|
}
|
|
_buffer.Add(val);
|
|
double valSq = val * val;
|
|
|
|
// Kahan add val to _sum
|
|
{
|
|
double y = val - _sumComp;
|
|
double t = _sum + y;
|
|
_sumComp = (t - _sum) - y;
|
|
_sum = t;
|
|
}
|
|
|
|
// Kahan add valSq to _sumSq
|
|
{
|
|
double y = valSq - _sumSqComp;
|
|
double t = _sumSq + y;
|
|
_sumSqComp = (t - _sumSq) - y;
|
|
_sumSq = t;
|
|
}
|
|
|
|
// Kahan add valCu to _sumCu
|
|
{
|
|
double y = (valSq * val) - _sumCuComp;
|
|
double t = _sumCu + y;
|
|
_sumCuComp = (t - _sumCu) - y;
|
|
_sumCu = t;
|
|
}
|
|
|
|
// Kahan add valQu to _sumQu
|
|
{
|
|
double y = (valSq * valSq) - _sumQuComp;
|
|
double t = _sumQu + y;
|
|
_sumQuComp = (t - _sumQu) - y;
|
|
_sumQu = t;
|
|
}
|
|
}
|
|
else
|
|
{
|
|
// Restore previous state before applying correction
|
|
_sum = p_sum;
|
|
_sumSq = p_sumSq;
|
|
_sumCu = p_sumCu;
|
|
_sumQu = p_sumQu;
|
|
_sumComp = _p_sumComp;
|
|
_sumSqComp = _p_sumSqComp;
|
|
_sumCuComp = _p_sumCuComp;
|
|
_sumQuComp = _p_sumQuComp;
|
|
|
|
double oldNewest = _buffer.Newest;
|
|
_buffer.UpdateNewest(input.Value);
|
|
|
|
double val = input.Value;
|
|
double valSq = val * val;
|
|
double oldSq = oldNewest * oldNewest;
|
|
|
|
// Kahan subtract old + add new for _sum
|
|
{
|
|
double y = (-oldNewest + val) - _sumComp;
|
|
double t = _sum + y;
|
|
_sumComp = (t - _sum) - y;
|
|
_sum = t;
|
|
}
|
|
|
|
// Kahan subtract old + add new for _sumSq
|
|
{
|
|
double y = (-oldSq + valSq) - _sumSqComp;
|
|
double t = _sumSq + y;
|
|
_sumSqComp = (t - _sumSq) - y;
|
|
_sumSq = t;
|
|
}
|
|
|
|
// Kahan subtract old + add new for _sumCu
|
|
{
|
|
double y = (-(oldSq * oldNewest) + (valSq * val)) - _sumCuComp;
|
|
double t = _sumCu + y;
|
|
_sumCuComp = (t - _sumCu) - y;
|
|
_sumCu = t;
|
|
}
|
|
|
|
// Kahan subtract old + add new for _sumQu
|
|
{
|
|
double y = (-(oldSq * oldSq) + (valSq * valSq)) - _sumQuComp;
|
|
double t = _sumQu + y;
|
|
_sumQuComp = (t - _sumQu) - y;
|
|
_sumQu = t;
|
|
}
|
|
}
|
|
|
|
double kurtosis = 0;
|
|
if (_buffer.Count >= 4)
|
|
{
|
|
double n = _buffer.Count;
|
|
double mean = _sum / n;
|
|
|
|
// Second central moment (variance): m₂ = Σ(x-μ)²/n
|
|
// = (SumSq - Sum²/n) / n
|
|
double m2Numerator = _sumSq - ((_sum * _sum) / n);
|
|
if (m2Numerator < Epsilon)
|
|
{
|
|
m2Numerator = 0;
|
|
}
|
|
|
|
double m2 = m2Numerator / n;
|
|
|
|
if (m2 > Epsilon)
|
|
{
|
|
// Fourth central moment: m₄ = Σ(x-μ)⁴/n
|
|
// Expanding (x-μ)⁴ = x⁴ - 4x³μ + 6x²μ² - 4xμ³ + μ⁴
|
|
// m₄ = SumQu/n - 4·mean·SumCu/n + 6·mean²·SumSq/n - 3·mean⁴
|
|
// Note: last term -4·mean³·Sum/n + mean⁴ = -4·mean⁴ + mean⁴ = -3·mean⁴
|
|
double meanSq = mean * mean;
|
|
double m4 = (_sumQu / n)
|
|
- (4.0 * mean * _sumCu / n)
|
|
+ (6.0 * meanSq * _sumSq / n)
|
|
- (3.0 * meanSq * meanSq);
|
|
|
|
// Population excess kurtosis: g₂ = m₄/m₂² - 3
|
|
double g2 = (m4 / (m2 * m2)) - 3.0;
|
|
|
|
if (_isPopulation)
|
|
{
|
|
kurtosis = g2;
|
|
}
|
|
else
|
|
{
|
|
// Sample excess kurtosis (Fisher's correction):
|
|
// G₂ = ((n-1)/((n-2)(n-3))) · ((n+1)·g₂ + 6)
|
|
double denom = (n - 2.0) * (n - 3.0);
|
|
if (Math.Abs(denom) > Epsilon)
|
|
{
|
|
kurtosis = ((n - 1.0) / denom) * (((n + 1.0) * g2) + 6.0);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
Last = new TValue(input.Time, kurtosis);
|
|
PubEvent(Last);
|
|
return Last;
|
|
}
|
|
|
|
public override TSeries Update(TSeries source)
|
|
{
|
|
if (source.Count == 0)
|
|
{
|
|
return [];
|
|
}
|
|
|
|
int len = source.Count;
|
|
var t = new List<long>(len);
|
|
var v = new List<double>(len);
|
|
CollectionsMarshal.SetCount(t, len);
|
|
CollectionsMarshal.SetCount(v, len);
|
|
|
|
var tSpan = CollectionsMarshal.AsSpan(t);
|
|
var vSpan = CollectionsMarshal.AsSpan(v);
|
|
|
|
Batch(source.Values, vSpan, _period, _isPopulation);
|
|
source.Times.CopyTo(tSpan);
|
|
|
|
// Reset running state before priming
|
|
_buffer.Clear();
|
|
_sum = 0;
|
|
_sumSq = 0;
|
|
_sumCu = 0;
|
|
_sumQu = 0;
|
|
_sumComp = 0;
|
|
_sumSqComp = 0;
|
|
_sumCuComp = 0;
|
|
_sumQuComp = 0;
|
|
|
|
// Prime the state
|
|
int primeStart = Math.Max(0, len - _period);
|
|
for (int i = primeStart; i < len; i++)
|
|
{
|
|
Update(source[i]);
|
|
}
|
|
|
|
return new TSeries(t, v);
|
|
}
|
|
|
|
public override void Reset()
|
|
{
|
|
_buffer.Clear();
|
|
_sum = 0;
|
|
_sumSq = 0;
|
|
_sumCu = 0;
|
|
_sumQu = 0;
|
|
_sumComp = 0;
|
|
_sumSqComp = 0;
|
|
_sumCuComp = 0;
|
|
_sumQuComp = 0;
|
|
Last = default;
|
|
}
|
|
|
|
public override void Prime(ReadOnlySpan<double> source, TimeSpan? step = null)
|
|
{
|
|
DateTime ts = DateTime.MinValue;
|
|
foreach (double value in source)
|
|
{
|
|
Update(new TValue(ts, value));
|
|
if (step.HasValue)
|
|
{
|
|
ts = ts.Add(step.Value);
|
|
}
|
|
}
|
|
}
|
|
|
|
public static TSeries Batch(TSeries source, int period, bool isPopulation = false)
|
|
{
|
|
var kurtosis = new Kurtosis(period, isPopulation);
|
|
return kurtosis.Update(source);
|
|
}
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
public static void Batch(ReadOnlySpan<double> source, Span<double> output, int period, bool isPopulation = false)
|
|
{
|
|
if (source.Length != output.Length)
|
|
{
|
|
throw new ArgumentException("Source and output must have the same length", nameof(output));
|
|
}
|
|
|
|
if (period < 4)
|
|
{
|
|
throw new ArgumentException("Period must be greater than or equal to 4", nameof(period));
|
|
}
|
|
|
|
int len = source.Length;
|
|
if (len == 0)
|
|
{
|
|
return;
|
|
}
|
|
|
|
// SIMD path for large, clean datasets
|
|
const int SimdThreshold = 256;
|
|
if (len >= SimdThreshold && Avx2.IsSupported && !source.ContainsNonFinite())
|
|
{
|
|
CalculateAvx2Core(source, output, period, isPopulation);
|
|
return;
|
|
}
|
|
|
|
// Scalar path
|
|
CalculateScalarCore(source, output, period, isPopulation);
|
|
}
|
|
|
|
public static (TSeries Results, Kurtosis Indicator) Calculate(TSeries source, int period, bool isPopulation = false)
|
|
{
|
|
var indicator = new Kurtosis(period, isPopulation);
|
|
TSeries results = indicator.Update(source);
|
|
return (results, indicator);
|
|
}
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
private static double CalculateKurtosisFromSums(double sum, double sumSq, double sumCu, double sumQu, double n, bool isPopulation)
|
|
{
|
|
double mean = sum / n;
|
|
|
|
double m2Numerator = sumSq - ((sum * sum) / n);
|
|
if (m2Numerator < Epsilon)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
double m2 = m2Numerator / n;
|
|
|
|
if (m2 <= Epsilon)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
// Fourth central moment via raw moments
|
|
double meanSq = mean * mean;
|
|
double m4 = (sumQu / n)
|
|
- (4.0 * mean * sumCu / n)
|
|
+ (6.0 * meanSq * sumSq / n)
|
|
- (3.0 * meanSq * meanSq);
|
|
|
|
double g2 = (m4 / (m2 * m2)) - 3.0;
|
|
|
|
if (isPopulation)
|
|
{
|
|
return g2;
|
|
}
|
|
|
|
// Fisher's correction for sample excess kurtosis
|
|
double denom = (n - 2.0) * (n - 3.0);
|
|
if (Math.Abs(denom) < Epsilon)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
return ((n - 1.0) / denom) * (((n + 1.0) * g2) + 6.0);
|
|
}
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
private static void CalculateScalarCore(ReadOnlySpan<double> source, Span<double> output, int period, bool isPopulation)
|
|
{
|
|
int len = source.Length;
|
|
double sum = 0;
|
|
double sumSq = 0;
|
|
double sumCu = 0;
|
|
double sumQu = 0;
|
|
double sumC = 0, sqC = 0, cuC = 0, quC = 0; // Kahan compensation
|
|
|
|
int i = 0;
|
|
|
|
// Warmup phase
|
|
int warmupEnd = Math.Min(period, len);
|
|
for (; i < warmupEnd; i++)
|
|
{
|
|
double val = source[i];
|
|
if (!double.IsFinite(val))
|
|
{
|
|
val = 0;
|
|
}
|
|
|
|
double valSq = val * val;
|
|
sum += val;
|
|
sumSq += valSq;
|
|
sumCu += valSq * val;
|
|
sumQu += valSq * valSq;
|
|
|
|
double n = i + 1;
|
|
output[i] = (n >= 4) ? CalculateKurtosisFromSums(sum, sumSq, sumCu, sumQu, n, isPopulation) : 0;
|
|
}
|
|
|
|
// Sliding window phase — Kahan compensated
|
|
for (; i < len; i++)
|
|
{
|
|
double val = source[i];
|
|
if (!double.IsFinite(val))
|
|
{
|
|
val = 0;
|
|
}
|
|
|
|
double oldVal = source[i - period];
|
|
if (!double.IsFinite(oldVal))
|
|
{
|
|
oldVal = 0;
|
|
}
|
|
|
|
double valSq = val * val;
|
|
double oldSq = oldVal * oldVal;
|
|
|
|
// Kahan sum
|
|
{
|
|
double y = (val - oldVal) - sumC;
|
|
double t = sum + y;
|
|
sumC = (t - sum) - y;
|
|
sum = t;
|
|
}
|
|
// Kahan sumSq
|
|
{
|
|
double y = (valSq - oldSq) - sqC;
|
|
double t = sumSq + y;
|
|
sqC = (t - sumSq) - y;
|
|
sumSq = t;
|
|
}
|
|
// Kahan sumCu
|
|
{
|
|
double y = ((valSq * val) - (oldSq * oldVal)) - cuC;
|
|
double t = sumCu + y;
|
|
cuC = (t - sumCu) - y;
|
|
sumCu = t;
|
|
}
|
|
// Kahan sumQu
|
|
{
|
|
double y = ((valSq * valSq) - (oldSq * oldSq)) - quC;
|
|
double t = sumQu + y;
|
|
quC = (t - sumQu) - y;
|
|
sumQu = t;
|
|
}
|
|
|
|
output[i] = CalculateKurtosisFromSums(sum, sumSq, sumCu, sumQu, period, isPopulation);
|
|
}
|
|
}
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveInlining)]
|
|
private static void WarmupKurtosis(int period, bool isPopulation, ref double srcRef, ref double outRef, out double sum, out double sumSq, out double sumCu, out double sumQu)
|
|
{
|
|
sum = 0;
|
|
sumSq = 0;
|
|
sumCu = 0;
|
|
sumQu = 0;
|
|
for (int i = 0; i < period; i++)
|
|
{
|
|
double val = Unsafe.Add(ref srcRef, i);
|
|
double valSq = val * val;
|
|
sum += val;
|
|
sumSq += valSq;
|
|
sumCu += valSq * val;
|
|
sumQu += valSq * valSq;
|
|
|
|
double n = i + 1;
|
|
Unsafe.Add(ref outRef, i) = (n >= 4) ? CalculateKurtosisFromSums(sum, sumSq, sumCu, sumQu, n, isPopulation) : 0;
|
|
}
|
|
}
|
|
|
|
[MethodImpl(MethodImplOptions.AggressiveOptimization)]
|
|
private static void CalculateAvx2Core(ReadOnlySpan<double> source, Span<double> output, int period, bool isPopulation)
|
|
{
|
|
int len = source.Length;
|
|
const int VectorWidth = 4;
|
|
|
|
ref double srcRef = ref MemoryMarshal.GetReference(source);
|
|
ref double outRef = ref MemoryMarshal.GetReference(output);
|
|
|
|
double invN = 1.0 / period;
|
|
double n = period;
|
|
|
|
WarmupKurtosis(period, isPopulation, ref srcRef, ref outRef, out double sum, out double sumSq, out double sumCu, out double sumQu);
|
|
|
|
if (len <= period)
|
|
{
|
|
return;
|
|
}
|
|
|
|
var vInvN = Vector256.Create(invN);
|
|
var vThree = Vector256.Create(3.0);
|
|
var vFour = Vector256.Create(4.0);
|
|
var vSix = Vector256.Create(6.0);
|
|
var vEpsilon = Vector256.Create(Epsilon);
|
|
var vZero = Vector256<double>.Zero;
|
|
|
|
// Fisher's correction constants
|
|
double fisherNum = isPopulation ? 1.0 : (n - 1.0);
|
|
double fisherDenom = isPopulation ? 1.0 : ((n - 2.0) * (n - 3.0));
|
|
double fisherNp1 = isPopulation ? 1.0 : (n + 1.0);
|
|
double fisherAdd = isPopulation ? 0.0 : 6.0;
|
|
var vFisherScale = Vector256.Create(isPopulation ? 1.0 : fisherNum / fisherDenom);
|
|
var vFisherNp1 = Vector256.Create(isPopulation ? 1.0 : fisherNp1);
|
|
var vFisherAdd = Vector256.Create(fisherAdd);
|
|
|
|
int simdEnd = period + (((len - period) / VectorWidth) * VectorWidth);
|
|
|
|
for (int i = period; i < simdEnd; i += VectorWidth)
|
|
{
|
|
var vNew = Vector256.LoadUnsafe(ref Unsafe.Add(ref srcRef, i));
|
|
var vOld = Vector256.LoadUnsafe(ref Unsafe.Add(ref srcRef, i - period));
|
|
|
|
// Deltas for Sum
|
|
var vDelta = Avx.Subtract(vNew, vOld);
|
|
|
|
// Deltas for SumSq
|
|
var vNewSq = Avx.Multiply(vNew, vNew);
|
|
var vOldSq = Avx.Multiply(vOld, vOld);
|
|
var vDeltaSq = Avx.Subtract(vNewSq, vOldSq);
|
|
|
|
// Deltas for SumCu
|
|
var vNewCu = Avx.Multiply(vNewSq, vNew);
|
|
var vOldCu = Avx.Multiply(vOldSq, vOld);
|
|
var vDeltaCu = Avx.Subtract(vNewCu, vOldCu);
|
|
|
|
// Deltas for SumQu
|
|
var vNewQu = Avx.Multiply(vNewSq, vNewSq);
|
|
var vOldQu = Avx.Multiply(vOldSq, vOldSq);
|
|
var vDeltaQu = Avx.Subtract(vNewQu, vOldQu);
|
|
|
|
// Prefix sum for Sum
|
|
var vShift1 = Avx2.Permute4x64(vDelta.AsUInt64(), 0b_10_01_00_00).AsDouble(); // skipcq: CS-R1131 - SIMD prefix sum pattern requires specific permutation
|
|
vShift1 = Avx.Blend(vZero, vShift1, 0b_1110);
|
|
var vP1 = Avx.Add(vDelta, vShift1);
|
|
var vShift2 = Avx2.Permute4x64(vP1.AsUInt64(), 0b_01_00_00_00).AsDouble(); // skipcq: CS-R1131 - SIMD prefix sum pattern requires specific permutation
|
|
vShift2 = Avx.Blend(vZero, vShift2, 0b_1100);
|
|
var vP2 = Avx.Add(vP1, vShift2);
|
|
var vSums = Avx.Add(Vector256.Create(sum), vP2);
|
|
|
|
// Prefix sum for SumSq
|
|
var vShiftSq1 = Avx2.Permute4x64(vDeltaSq.AsUInt64(), 0b_10_01_00_00).AsDouble(); // skipcq: CS-R1131 - SIMD prefix sum pattern requires specific permutation
|
|
vShiftSq1 = Avx.Blend(vZero, vShiftSq1, 0b_1110);
|
|
var vP1Sq = Avx.Add(vDeltaSq, vShiftSq1);
|
|
var vShiftSq2 = Avx2.Permute4x64(vP1Sq.AsUInt64(), 0b_01_00_00_00).AsDouble(); // skipcq: CS-R1131 - SIMD prefix sum pattern requires specific permutation
|
|
vShiftSq2 = Avx.Blend(vZero, vShiftSq2, 0b_1100);
|
|
var vP2Sq = Avx.Add(vP1Sq, vShiftSq2);
|
|
var vSumSqs = Avx.Add(Vector256.Create(sumSq), vP2Sq);
|
|
|
|
// Prefix sum for SumCu
|
|
var vShiftCu1 = Avx2.Permute4x64(vDeltaCu.AsUInt64(), 0b_10_01_00_00).AsDouble(); // skipcq: CS-R1131 - SIMD prefix sum pattern requires specific permutation
|
|
vShiftCu1 = Avx.Blend(vZero, vShiftCu1, 0b_1110);
|
|
var vP1Cu = Avx.Add(vDeltaCu, vShiftCu1);
|
|
var vShiftCu2 = Avx2.Permute4x64(vP1Cu.AsUInt64(), 0b_01_00_00_00).AsDouble(); // skipcq: CS-R1131 - SIMD prefix sum pattern requires specific permutation
|
|
vShiftCu2 = Avx.Blend(vZero, vShiftCu2, 0b_1100);
|
|
var vP2Cu = Avx.Add(vP1Cu, vShiftCu2);
|
|
var vSumCus = Avx.Add(Vector256.Create(sumCu), vP2Cu);
|
|
|
|
// Prefix sum for SumQu
|
|
var vShiftQu1 = Avx2.Permute4x64(vDeltaQu.AsUInt64(), 0b_10_01_00_00).AsDouble(); // skipcq: CS-R1131 - SIMD prefix sum pattern requires specific permutation
|
|
vShiftQu1 = Avx.Blend(vZero, vShiftQu1, 0b_1110);
|
|
var vP1Qu = Avx.Add(vDeltaQu, vShiftQu1);
|
|
var vShiftQu2 = Avx2.Permute4x64(vP1Qu.AsUInt64(), 0b_01_00_00_00).AsDouble(); // skipcq: CS-R1131 - SIMD prefix sum pattern requires specific permutation
|
|
vShiftQu2 = Avx.Blend(vZero, vShiftQu2, 0b_1100);
|
|
var vP2Qu = Avx.Add(vP1Qu, vShiftQu2);
|
|
var vSumQus = Avx.Add(Vector256.Create(sumQu), vP2Qu);
|
|
|
|
// Calculate Kurtosis
|
|
var vMean = Avx.Multiply(vSums, vInvN);
|
|
var vMeanSq = Avx.Multiply(vMean, vMean);
|
|
|
|
// m2 = (SumSq - Sum²/n) / n
|
|
var vSumSquared = Avx.Multiply(vSums, vSums);
|
|
var vM2Num = Fma.IsSupported
|
|
? Fma.MultiplyAddNegated(vSumSquared, vInvN, vSumSqs)
|
|
: Avx.Subtract(vSumSqs, Avx.Multiply(vSumSquared, vInvN));
|
|
vM2Num = Avx.Max(vZero, vM2Num);
|
|
var vM2 = Avx.Multiply(vM2Num, vInvN);
|
|
|
|
// m4 = SumQu/n - 4·mean·SumCu/n + 6·mean²·SumSq/n - 3·mean⁴
|
|
var vTerm1 = Avx.Multiply(vSumQus, vInvN);
|
|
var vTerm2 = Avx.Multiply(vFour, Avx.Multiply(vMean, Avx.Multiply(vSumCus, vInvN)));
|
|
var vTerm3 = Avx.Multiply(vSix, Avx.Multiply(vMeanSq, Avx.Multiply(vSumSqs, vInvN)));
|
|
var vTerm4 = Avx.Multiply(vThree, Avx.Multiply(vMeanSq, vMeanSq));
|
|
|
|
var vM4 = Avx.Subtract(Avx.Add(Avx.Subtract(vTerm1, vTerm2), vTerm3), vTerm4);
|
|
|
|
// g2 = m4 / m2² - 3
|
|
var vM2Sq = Avx.Multiply(vM2, vM2);
|
|
var vG2 = Avx.Subtract(Avx.Divide(vM4, vM2Sq), vThree);
|
|
|
|
// Apply Fisher's correction: scale * (np1 * g2 + 6)
|
|
Vector256<double> vResult;
|
|
if (isPopulation)
|
|
{
|
|
vResult = vG2;
|
|
}
|
|
else
|
|
{
|
|
var vCorrected = Fma.IsSupported
|
|
? Fma.MultiplyAdd(vFisherNp1, vG2, vFisherAdd)
|
|
: Avx.Add(Avx.Multiply(vFisherNp1, vG2), vFisherAdd);
|
|
vResult = Avx.Multiply(vFisherScale, vCorrected);
|
|
}
|
|
|
|
// Mask: zero out where m2 <= epsilon
|
|
var vMask = Avx.Compare(vM2, vEpsilon, FloatComparisonMode.OrderedGreaterThanNonSignaling);
|
|
vResult = Avx.BlendVariable(vZero, vResult, vMask);
|
|
|
|
vResult.StoreUnsafe(ref Unsafe.Add(ref outRef, i));
|
|
|
|
sum = vSums.GetElement(3);
|
|
sumSq = vSumSqs.GetElement(3);
|
|
sumCu = vSumCus.GetElement(3);
|
|
sumQu = vSumQus.GetElement(3);
|
|
}
|
|
|
|
for (int i = simdEnd; i < len; i++)
|
|
{
|
|
double val = Unsafe.Add(ref srcRef, i);
|
|
double oldVal = Unsafe.Add(ref srcRef, i - period);
|
|
|
|
double valSq = val * val;
|
|
double oldSq = oldVal * oldVal;
|
|
sum = sum - oldVal + val;
|
|
sumSq = sumSq - oldSq + valSq;
|
|
sumCu = Math.FusedMultiplyAdd(valSq, val, Math.FusedMultiplyAdd(-oldSq, oldVal, sumCu));
|
|
sumQu = Math.FusedMultiplyAdd(valSq, valSq, Math.FusedMultiplyAdd(-oldSq, oldSq, sumQu));
|
|
|
|
Unsafe.Add(ref outRef, i) = CalculateKurtosisFromSums(sum, sumSq, sumCu, sumQu, n, isPopulation);
|
|
}
|
|
}
|
|
}
|