public static Vector128 <float> MaskExp(Vector128 <float> power, int exponentMask)
{
Vector128 <float> restrictedPower = Avx.Blend(power, AvxExtensions.BroadcastScalarToVector128(1.0F), (byte)exponentMask);
Vector128 <float> exponent = Avx.Blend(MathV.Exp(restrictedPower), Vector128 <float> .Zero, (byte)exponentMask);
return(exponent);
}
public static unsafe RtMatrix Transpose(RtMatrix matrix)
{
RtMatrix result = new RtMatrix();
if (Avx2.IsSupported && useIntrinsics)
{
var row1 = Avx.LoadVector256(&matrix.M11);
var row2 = Avx.LoadVector256(&matrix.M21);
var row3 = Avx.LoadVector256(&matrix.M31);
var row4 = Avx.LoadVector256(&matrix.M41);
var l12 = Avx.UnpackLow(row1, row2);
var l34 = Avx.UnpackLow(row3, row4);
var h12 = Avx.UnpackHigh(row1, row2);
var h34 = Avx.UnpackHigh(row3, row4);
Avx.Store(&result.M11, Avx.Blend(l12, Avx2.Permute4x64(l34, 0x4E), 0x0C));
Avx.Store(&result.M21, Avx.Blend(h12, Avx2.Permute4x64(h34, 0x4E), 0x0C));
Avx.Store(&result.M31, Avx.Blend(Avx2.Permute4x64(l12, 0x4E), l34, 0x0c));
Avx.Store(&result.M41, Avx.Blend(Avx2.Permute4x64(h12, 0x4E), h34, 0x0c));
return(result);
}
result.M11 = matrix.M11;
result.M12 = matrix.M21;
result.M13 = matrix.M31;
result.M14 = matrix.M41;
result.M21 = matrix.M12;
result.M22 = matrix.M22;
result.M23 = matrix.M32;
result.M24 = matrix.M42;
result.M31 = matrix.M13;
result.M32 = matrix.M23;
result.M33 = matrix.M33;
result.M34 = matrix.M43;
result.M41 = matrix.M14;
result.M42 = matrix.M24;
result.M43 = matrix.M34;
result.M44 = matrix.M44;
return(result);
}
public unsafe static Vector128 <float> Exp2(Vector128 <float> power)
{
Debug.Assert(Avx.MoveMask(Avx.And(Avx.CompareGreaterThan(power, AvxExtensions.BroadcastScalarToVector128(MathV.FloatMaximumPower)), Avx.CompareOrdered(power, power))) == 0);
byte zeroMask = (byte)Avx.MoveMask(Avx.CompareLessThan(power, AvxExtensions.BroadcastScalarToVector128(-MathV.FloatMaximumPower)));
Vector128 <float> integerPart = Avx.RoundToNearestInteger(power);
Vector128 <float> integerExponent = Avx.ShiftLeftLogical(Avx.Add(Avx.ConvertToVector128Int32(integerPart), MathV.FloatMantissaZero128), MathV.FloatMantissaBits).AsSingle();
// evaluate polynomial
Vector128 <float> beta1 = AvxExtensions.BroadcastScalarToVector128(MathV.Exp2Beta1);
Vector128 <float> beta2 = AvxExtensions.BroadcastScalarToVector128(MathV.Exp2Beta2);
Vector128 <float> beta3 = AvxExtensions.BroadcastScalarToVector128(MathV.Exp2Beta3);
Vector128 <float> beta4 = AvxExtensions.BroadcastScalarToVector128(MathV.Exp2Beta4);
Vector128 <float> x = Avx.Subtract(power, integerPart); // fractional part
Vector128 <float> fractionalExponent = AvxExtensions.BroadcastScalarToVector128(MathV.One);
fractionalExponent = Avx.Add(fractionalExponent, Avx.Multiply(beta1, x));
Vector128 <float> x2 = Avx.Multiply(x, x);
fractionalExponent = Avx.Add(fractionalExponent, Avx.Multiply(beta2, x2));
Vector128 <float> x3 = Avx.Multiply(x2, x);
fractionalExponent = Avx.Add(fractionalExponent, Avx.Multiply(beta3, x3));
Vector128 <float> x4 = Avx.Multiply(x3, x);
fractionalExponent = Avx.Add(fractionalExponent, Avx.Multiply(beta4, x4));
// form exponent
Vector128 <float> exponent = Avx.Multiply(integerExponent, fractionalExponent);
// suppress exponent overflows by truncating values less than 2^-127 to zero
if (zeroMask != 0)
{
exponent = Avx.Blend(exponent, Vector128 <float> .Zero, zeroMask);
}
return(exponent);
}
static unsafe int Main(string[] args)
{
int testResult = Pass;
if (Avx.IsSupported)
{
using (TestTable <float> floatTable = new TestTable <float>(new float[8] {
1, -5, 100, 0, 1, -5, 100, 0
}, new float[8] {
22, -1, -50, 0, 22, -1, -50, 0
}, new float[8]))
{
var vf1 = Unsafe.Read <Vector256 <float> >(floatTable.inArray1Ptr);
var vf2 = Unsafe.Read <Vector256 <float> >(floatTable.inArray2Ptr);
// SDDD SDDD
var vf3 = Avx.Blend(vf1, vf2, 1);
Unsafe.Write(floatTable.outArrayPtr, vf3);
if (!floatTable.CheckResult((x, y, z) => (z[0] == y[0]) && (z[1] == x[1]) &&
(z[2] == x[2]) && (z[3] == x[3]) &&
(z[4] == x[4]) && (z[5] == x[5]) &&
(z[6] == x[6]) && (z[7] == x[7])))
{
Console.WriteLine("0Avx Blend failed on float:");
foreach (var item in floatTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
// DSDD DDDD
vf3 = Avx.Blend(vf1, vf2, 2);
Unsafe.Write(floatTable.outArrayPtr, vf3);
if (!floatTable.CheckResult((x, y, z) => (z[0] == x[0]) && (z[1] == y[1]) &&
(z[2] == x[2]) && (z[3] == x[3]) &&
(z[4] == x[4]) && (z[5] == x[5]) &&
(z[6] == x[6]) && (z[7] == x[7])))
{
Console.WriteLine("Avx Blend failed on float:");
foreach (var item in floatTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
// DDSD DDDD
vf3 = Avx.Blend(vf1, vf2, 4);
Unsafe.Write(floatTable.outArrayPtr, vf3);
if (!floatTable.CheckResult((x, y, z) => (z[0] == x[0]) && (z[1] == x[1]) &&
(z[2] == y[2]) && (z[3] == x[3]) &&
(z[4] == x[4]) && (z[5] == x[5]) &&
(z[6] == x[6]) && (z[7] == x[7])))
{
Console.WriteLine("Avx Blend failed on float:");
foreach (var item in floatTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
// SDSD SDSD
vf3 = Avx.Blend(vf1, vf2, 85);
Unsafe.Write(floatTable.outArrayPtr, vf3);
if (!floatTable.CheckResult((x, y, z) => (z[0] == y[0]) && (z[1] == x[1]) &&
(z[2] == y[2]) && (z[3] == x[3]) &&
(z[4] == y[4]) && (z[5] == x[5]) &&
(z[6] == y[6]) && (z[7] == x[7])))
{
Console.WriteLine("Avx Blend failed on float:");
foreach (var item in floatTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
// SDDD DDDD
vf3 = (Vector256 <float>) typeof(Avx).GetMethod(nameof(Avx.Blend), new Type[] { vf1.GetType(), vf2.GetType(), typeof(byte) }).Invoke(null, new object[] { vf1, vf2, (byte)(1) });
Unsafe.Write(floatTable.outArrayPtr, vf3);
if (!floatTable.CheckResult((x, y, z) => (z[0] == y[0]) && (z[1] == x[1]) &&
(z[2] == x[2]) && (z[3] == x[3]) &&
(z[4] == x[4]) && (z[5] == x[5]) &&
(z[6] == x[6]) && (z[7] == x[7])))
{
Console.WriteLine("Avx Blend failed on float:");
foreach (var item in floatTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
}
using (TestTable <double> doubleTable = new TestTable <double>(new double[4] {
1, -5, 100, 0
}, new double[4] {
22, -1, -50, 0
}, new double[4]))
{
var vf1 = Unsafe.Read <Vector256 <double> >(doubleTable.inArray1Ptr);
var vf2 = Unsafe.Read <Vector256 <double> >(doubleTable.inArray2Ptr);
// DD DD
var vf3 = Avx.Blend(vf1, vf2, 0);
Unsafe.Write(doubleTable.outArrayPtr, vf3);
if (!doubleTable.CheckResult((x, y, z) => (z[0] == x[0]) && (z[1] == x[1]) &&
(z[2] == x[2]) && (z[3] == x[3])))
{
Console.WriteLine("Avx Blend failed on double:");
foreach (var item in doubleTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
// SD DD
vf3 = Avx.Blend(vf1, vf2, 1);
Unsafe.Write(doubleTable.outArrayPtr, vf3);
if (!doubleTable.CheckResult((x, y, z) => (z[0] == y[0]) && (z[1] == x[1]) &&
(z[2] == x[2]) && (z[3] == x[3])))
{
Console.WriteLine("Avx Blend failed on double:");
foreach (var item in doubleTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
// DS DD
vf3 = Avx.Blend(vf1, vf2, 2);
Unsafe.Write(doubleTable.outArrayPtr, vf3);
if (!doubleTable.CheckResult((x, y, z) => (z[0] == x[0]) && (z[1] == y[1]) &&
(z[2] == x[2]) && (z[3] == x[3])))
{
Console.WriteLine("Avx Blend failed on double:");
foreach (var item in doubleTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
// SS DD
vf3 = Avx.Blend(vf1, vf2, 51);
Unsafe.Write(doubleTable.outArrayPtr, vf3);
if (!doubleTable.CheckResult((x, y, z) => (z[0] == y[0]) && (z[1] == y[1]) &&
(z[2] == x[2]) && (z[3] == x[3])))
{
Console.WriteLine("Avx Blend failed on double:");
foreach (var item in doubleTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
// DD DD
vf3 = (Vector256 <double>) typeof(Avx).GetMethod(nameof(Avx.Blend), new Type[] { vf1.GetType(), vf2.GetType(), typeof(byte) }).Invoke(null, new object[] { vf1, vf2, (byte)(0) });
Unsafe.Write(doubleTable.outArrayPtr, vf3);
if (!doubleTable.CheckResult((x, y, z) => (z[0] == x[0]) && (z[1] == x[1]) &&
(z[2] == x[2]) && (z[3] == x[3])))
{
Console.WriteLine("Avx Blend failed on double:");
foreach (var item in doubleTable.outArray)
{
Console.Write(item + ", ");
}
Console.WriteLine();
testResult = Fail;
}
}
}
return(testResult);
}
unsafe void IConversionProcessor.ConvertLine(byte *ipstart, byte *opstart, int cb)
{
float *ip = (float *)ipstart, ipe = (float *)(ipstart + cb);
byte * op = opstart;
#if HWINTRINSICS
if (Avx2.IsSupported)
{
var vzero = Vector256 <float> .Zero;
var vmin = Vector256.Create(0.5f / byte.MaxValue);
var vscale = Vector256.Create((float)byte.MaxValue);
var vmaskp = Avx.LoadVector256((int *)Unsafe.AsPointer(ref MemoryMarshal.GetReference(HWIntrinsics.PermuteMaskDeinterleave8x32)));
ipe -= Vector256 <byte> .Count;
while (ip <= ipe)
{
var vf0 = Avx.LoadVector256(ip);
var vf1 = Avx.LoadVector256(ip + Vector256 <float> .Count);
var vf2 = Avx.LoadVector256(ip + Vector256 <float> .Count * 2);
var vf3 = Avx.LoadVector256(ip + Vector256 <float> .Count * 3);
ip += Vector256 <byte> .Count;
var vfa0 = Avx.Shuffle(vf0, vf0, HWIntrinsics.ShuffleMaskAlpha);
var vfa1 = Avx.Shuffle(vf1, vf1, HWIntrinsics.ShuffleMaskAlpha);
var vfa2 = Avx.Shuffle(vf2, vf2, HWIntrinsics.ShuffleMaskAlpha);
var vfa3 = Avx.Shuffle(vf3, vf3, HWIntrinsics.ShuffleMaskAlpha);
vfa0 = Avx.Max(vfa0, vmin);
vfa1 = Avx.Max(vfa1, vmin);
vfa2 = Avx.Max(vfa2, vmin);
vfa3 = Avx.Max(vfa3, vmin);
vf0 = Avx.Multiply(vf0, Avx.Reciprocal(vfa0));
vf1 = Avx.Multiply(vf1, Avx.Reciprocal(vfa1));
vf2 = Avx.Multiply(vf2, Avx.Reciprocal(vfa2));
vf3 = Avx.Multiply(vf3, Avx.Reciprocal(vfa3));
vf0 = Avx.Blend(vf0, vfa0, HWIntrinsics.BlendMaskAlpha);
vf1 = Avx.Blend(vf1, vfa1, HWIntrinsics.BlendMaskAlpha);
vf2 = Avx.Blend(vf2, vfa2, HWIntrinsics.BlendMaskAlpha);
vf3 = Avx.Blend(vf3, vfa3, HWIntrinsics.BlendMaskAlpha);
vf0 = Avx.BlendVariable(vf0, vzero, HWIntrinsics.AvxCompareEqual(vfa0, vmin));
vf1 = Avx.BlendVariable(vf1, vzero, HWIntrinsics.AvxCompareEqual(vfa1, vmin));
vf2 = Avx.BlendVariable(vf2, vzero, HWIntrinsics.AvxCompareEqual(vfa2, vmin));
vf3 = Avx.BlendVariable(vf3, vzero, HWIntrinsics.AvxCompareEqual(vfa3, vmin));
vf0 = Avx.Multiply(vf0, vscale);
vf1 = Avx.Multiply(vf1, vscale);
vf2 = Avx.Multiply(vf2, vscale);
vf3 = Avx.Multiply(vf3, vscale);
var vi0 = Avx.ConvertToVector256Int32(vf0);
var vi1 = Avx.ConvertToVector256Int32(vf1);
var vi2 = Avx.ConvertToVector256Int32(vf2);
var vi3 = Avx.ConvertToVector256Int32(vf3);
var vs0 = Avx2.PackSignedSaturate(vi0, vi1);
var vs1 = Avx2.PackSignedSaturate(vi2, vi3);
var vb0 = Avx2.PackUnsignedSaturate(vs0, vs1);
vb0 = Avx2.PermuteVar8x32(vb0.AsInt32(), vmaskp).AsByte();
Avx.Store(op, vb0);
op += Vector256 <byte> .Count;
}
ipe += Vector256 <byte> .Count;
}
else if (Sse41.IsSupported)
{
var vzero = Vector128 <float> .Zero;
var vmin = Vector128.Create(0.5f / byte.MaxValue);
var vscale = Vector128.Create((float)byte.MaxValue);
ipe -= Vector128 <byte> .Count;
while (ip <= ipe)
{
var vf0 = Sse.LoadVector128(ip);
var vf1 = Sse.LoadVector128(ip + Vector128 <float> .Count);
var vf2 = Sse.LoadVector128(ip + Vector128 <float> .Count * 2);
var vf3 = Sse.LoadVector128(ip + Vector128 <float> .Count * 3);
ip += Vector128 <byte> .Count;
var vfa0 = Sse.Shuffle(vf0, vf0, HWIntrinsics.ShuffleMaskAlpha);
var vfa1 = Sse.Shuffle(vf1, vf1, HWIntrinsics.ShuffleMaskAlpha);
var vfa2 = Sse.Shuffle(vf2, vf2, HWIntrinsics.ShuffleMaskAlpha);
var vfa3 = Sse.Shuffle(vf3, vf3, HWIntrinsics.ShuffleMaskAlpha);
vfa0 = Sse.Max(vfa0, vmin);
vfa1 = Sse.Max(vfa1, vmin);
vfa2 = Sse.Max(vfa2, vmin);
vfa3 = Sse.Max(vfa3, vmin);
vf0 = Sse.Multiply(vf0, Sse.Reciprocal(vfa0));
vf1 = Sse.Multiply(vf1, Sse.Reciprocal(vfa1));
vf2 = Sse.Multiply(vf2, Sse.Reciprocal(vfa2));
vf3 = Sse.Multiply(vf3, Sse.Reciprocal(vfa3));
vf0 = Sse41.Blend(vf0, vfa0, HWIntrinsics.BlendMaskAlpha);
vf1 = Sse41.Blend(vf1, vfa1, HWIntrinsics.BlendMaskAlpha);
vf2 = Sse41.Blend(vf2, vfa2, HWIntrinsics.BlendMaskAlpha);
vf3 = Sse41.Blend(vf3, vfa3, HWIntrinsics.BlendMaskAlpha);
vf0 = Sse41.BlendVariable(vf0, vzero, Sse.CompareEqual(vfa0, vmin));
vf1 = Sse41.BlendVariable(vf1, vzero, Sse.CompareEqual(vfa1, vmin));
vf2 = Sse41.BlendVariable(vf2, vzero, Sse.CompareEqual(vfa2, vmin));
vf3 = Sse41.BlendVariable(vf3, vzero, Sse.CompareEqual(vfa3, vmin));
vf0 = Sse.Multiply(vf0, vscale);
vf1 = Sse.Multiply(vf1, vscale);
vf2 = Sse.Multiply(vf2, vscale);
vf3 = Sse.Multiply(vf3, vscale);
var vi0 = Sse2.ConvertToVector128Int32(vf0);
var vi1 = Sse2.ConvertToVector128Int32(vf1);
var vi2 = Sse2.ConvertToVector128Int32(vf2);
var vi3 = Sse2.ConvertToVector128Int32(vf3);
var vs0 = Sse2.PackSignedSaturate(vi0, vi1);
var vs1 = Sse2.PackSignedSaturate(vi2, vi3);
var vb0 = Sse2.PackUnsignedSaturate(vs0, vs1);
Sse2.Store(op, vb0);
op += Vector128 <byte> .Count;
}
ipe += Vector128 <byte> .Count;
}
#endif
float fmax = new Vector4(byte.MaxValue).X, fround = new Vector4(0.5f).X, fmin = fround / fmax;
while (ip < ipe)
{
float f3 = ip[3];
if (f3 < fmin)
{
*(uint *)op = 0;
}
else
{
float f3i = fmax / f3;
byte o0 = ClampToByte((int)(ip[0] * f3i + fround));
byte o1 = ClampToByte((int)(ip[1] * f3i + fround));
byte o2 = ClampToByte((int)(ip[2] * f3i + fround));
byte o3 = ClampToByte((int)(f3 * fmax + fround));
op[0] = o0;
op[1] = o1;
op[2] = o2;
op[3] = o3;
}
ip += 4;
op += 4;
}
}
unsafe void IConversionProcessor.ConvertLine(byte *ipstart, byte *opstart, int cb)
{
fixed(float *atstart = &LookupTables.Alpha[0])
{
byte * ip = ipstart, ipe = ipstart + cb;
float *op = (float *)opstart, at = atstart;
#if HWINTRINSICS
if (Avx2.IsSupported)
{
var vscale = Vector256.Create(1f / byte.MaxValue);
ipe -= Vector256 <byte> .Count;
while (ip <= ipe)
{
var vi0 = Avx2.ConvertToVector256Int32(ip);
var vi1 = Avx2.ConvertToVector256Int32(ip + Vector256 <int> .Count);
var vi2 = Avx2.ConvertToVector256Int32(ip + Vector256 <int> .Count * 2);
var vi3 = Avx2.ConvertToVector256Int32(ip + Vector256 <int> .Count * 3);
ip += Vector256 <byte> .Count;
var vf0 = Avx.ConvertToVector256Single(vi0);
var vf1 = Avx.ConvertToVector256Single(vi1);
var vf2 = Avx.ConvertToVector256Single(vi2);
var vf3 = Avx.ConvertToVector256Single(vi3);
vf0 = Avx.Multiply(vf0, vscale);
vf1 = Avx.Multiply(vf1, vscale);
vf2 = Avx.Multiply(vf2, vscale);
vf3 = Avx.Multiply(vf3, vscale);
var vfa0 = Avx.Shuffle(vf0, vf0, HWIntrinsics.ShuffleMaskAlpha);
var vfa1 = Avx.Shuffle(vf1, vf1, HWIntrinsics.ShuffleMaskAlpha);
var vfa2 = Avx.Shuffle(vf2, vf2, HWIntrinsics.ShuffleMaskAlpha);
var vfa3 = Avx.Shuffle(vf3, vf3, HWIntrinsics.ShuffleMaskAlpha);
vf0 = Avx.Multiply(vf0, vfa0);
vf1 = Avx.Multiply(vf1, vfa1);
vf2 = Avx.Multiply(vf2, vfa2);
vf3 = Avx.Multiply(vf3, vfa3);
vf0 = Avx.Blend(vf0, vfa0, HWIntrinsics.BlendMaskAlpha);
vf1 = Avx.Blend(vf1, vfa1, HWIntrinsics.BlendMaskAlpha);
vf2 = Avx.Blend(vf2, vfa2, HWIntrinsics.BlendMaskAlpha);
vf3 = Avx.Blend(vf3, vfa3, HWIntrinsics.BlendMaskAlpha);
Avx.Store(op, vf0);
Avx.Store(op + Vector256 <int> .Count, vf1);
Avx.Store(op + Vector256 <int> .Count * 2, vf2);
Avx.Store(op + Vector256 <int> .Count * 3, vf3);
op += Vector256 <byte> .Count;
}
ipe += Vector256 <byte> .Count;
}
else if (Sse41.IsSupported)
{
var vscale = Vector128.Create(1f / byte.MaxValue);
ipe -= Vector128 <byte> .Count;
while (ip <= ipe)
{
var vi0 = Sse41.ConvertToVector128Int32(ip);
var vi1 = Sse41.ConvertToVector128Int32(ip + Vector128 <int> .Count);
var vi2 = Sse41.ConvertToVector128Int32(ip + Vector128 <int> .Count * 2);
var vi3 = Sse41.ConvertToVector128Int32(ip + Vector128 <int> .Count * 3);
ip += Vector128 <byte> .Count;
var vf0 = Sse2.ConvertToVector128Single(vi0);
var vf1 = Sse2.ConvertToVector128Single(vi1);
var vf2 = Sse2.ConvertToVector128Single(vi2);
var vf3 = Sse2.ConvertToVector128Single(vi3);
vf0 = Sse.Multiply(vf0, vscale);
vf1 = Sse.Multiply(vf1, vscale);
vf2 = Sse.Multiply(vf2, vscale);
vf3 = Sse.Multiply(vf3, vscale);
var vfa0 = Sse.Shuffle(vf0, vf0, HWIntrinsics.ShuffleMaskAlpha);
var vfa1 = Sse.Shuffle(vf1, vf1, HWIntrinsics.ShuffleMaskAlpha);
var vfa2 = Sse.Shuffle(vf2, vf2, HWIntrinsics.ShuffleMaskAlpha);
var vfa3 = Sse.Shuffle(vf3, vf3, HWIntrinsics.ShuffleMaskAlpha);
vf0 = Sse.Multiply(vf0, vfa0);
vf1 = Sse.Multiply(vf1, vfa1);
vf2 = Sse.Multiply(vf2, vfa2);
vf3 = Sse.Multiply(vf3, vfa3);
vf0 = Sse41.Blend(vf0, vfa0, HWIntrinsics.BlendMaskAlpha);
vf1 = Sse41.Blend(vf1, vfa1, HWIntrinsics.BlendMaskAlpha);
vf2 = Sse41.Blend(vf2, vfa2, HWIntrinsics.BlendMaskAlpha);
vf3 = Sse41.Blend(vf3, vfa3, HWIntrinsics.BlendMaskAlpha);
Sse.Store(op, vf0);
Sse.Store(op + Vector128 <int> .Count, vf1);
Sse.Store(op + Vector128 <int> .Count * 2, vf2);
Sse.Store(op + Vector128 <int> .Count * 3, vf3);
op += Vector128 <byte> .Count;
}
ipe += Vector128 <byte> .Count;
}
#endif
while (ip < ipe)
{
float o0 = at[(uint)ip[0]];
float o1 = at[(uint)ip[1]];
float o2 = at[(uint)ip[2]];
float o3 = at[(uint)ip[3]];
ip += 4;
op[0] = o0 * o3;
op[1] = o1 * o3;
op[2] = o2 * o3;
op[3] = o3;
op += 4;
}
}
}