public byte8(byte3 x012, byte2 x34, byte3 x567)
{
if (Sse2.IsSse2Supported)
{
v128 hi = Sse2.bslli_si128(x567, 2 * sizeof(byte));
if (Sse4_1.IsSse41Supported)
{
hi = Sse4_1.blend_epi16(x34, hi, 0b0110);
}
else
{
hi = Mask.BlendEpi16_SSE2(x34, hi, 0b0110);
}
hi = Sse2.bslli_si128(hi, 3 * sizeof(byte));
this = Mask.BlendV(x012, hi, new byte8(0, 0, 0, 255, 255, 255, 255, 255));
}
else
{
this.x0 = x012.x;
this.x1 = x012.y;
this.x2 = x012.z;
this.x3 = x34.x;
this.x4 = x34.y;
this.x5 = x567.x;
this.x6 = x567.y;
this.x7 = x567.z;
}
}
public byte2x4(byte v)
{
this.c0 = v;
this.c1 = v;
this.c2 = v;
this.c3 = v;
}
public static byte2 tzcnt(byte2 x)
{
if (Ssse3.IsSsse3Supported)
{
v128 NIBBLE_MASK = new v128(0x0F0F_0F0F);
v128 SHUFFLE_MASK_LO = new v128(8, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0);
v128 SHUFFLE_MASK_HI = new v128(8, 4, 5, 4, 6, 4, 5, 4, 7, 4, 5, 4, 6, 4, 5, 4);
return(Sse2.min_epu8(Ssse3.shuffle_epi8(SHUFFLE_MASK_LO, Sse2.and_si128(NIBBLE_MASK, x)),
Ssse3.shuffle_epi8(SHUFFLE_MASK_HI, Sse2.and_si128(NIBBLE_MASK, Sse2.srli_epi16(x, 4)))));
}
else if (Sse2.IsSse2Supported)
{
v128 compareMask = x & (byte2)(-(sbyte2)x);
byte2 first = Mask.BlendV(default(v128), new byte4(1), Sse2.cmpeq_epi8(compareMask, default(v128)));
byte2 second = Mask.BlendV(default(v128), new byte4(4), Sse2.cmpeq_epi8(compareMask & (byte4)0x0F, default(v128)));
byte2 third = Mask.BlendV(default(v128), new byte4(2), Sse2.cmpeq_epi8(compareMask & (byte4)0x33, default(v128)));
byte2 fourth = Mask.BlendV(default(v128), new byte4(1), Sse2.cmpeq_epi8(compareMask & (byte4)0x55, default(v128)));
return((first + second) + (third + fourth));
}
else
{
return(new byte2(tzcnt(x.x), tzcnt(x.y)));
}
}
public static byte2 reversebits(byte2 x)
{
x = ((x >> 1) & 0x55) | ((x & 0x55) << 1);
x = ((x >> 2) & 0x33) | ((x & 0x33) << 2);
return((x >> 4) | (x << 4));
}
public byte2x3(byte m00, byte m01, byte m02,
byte m10, byte m11, byte m12)
{
this.c0 = new byte2(m00, m10);
this.c1 = new byte2(m01, m11);
this.c2 = new byte2(m02, m12);
}
public byte8(byte2 x01, byte3 x234, byte3 x567)
{
if (Sse2.IsSse2Supported)
{
v128 mid = Sse2.bslli_si128(x234, 2 * sizeof(byte));
v128 hi = Sse2.bslli_si128(x567, 5 * sizeof(byte));
hi = Mask.BlendV(mid, hi, new byte8(0, 0, 0, 0, 0, 255, 255, 255));
if (Sse4_1.IsSse41Supported)
{
this = Sse4_1.blend_epi16(x01, hi, 0b1110);
}
else
{
this = Mask.BlendEpi16_SSE2(x01, hi, 0b1110);
}
}
else
{
this.x0 = x01.x;
this.x1 = x01.y;
this.x2 = x234.x;
this.x3 = x234.y;
this.x4 = x234.z;
this.x5 = x567.x;
this.x6 = x567.y;
this.x7 = x567.z;
}
}
public byte2x4(byte2 c0, byte2 c1, byte2 c2, byte2 c3)
{
this.c0 = c0;
this.c1 = c1;
this.c2 = c2;
this.c3 = c3;
}
public static byte2 lcm(sbyte2 x, sbyte2 y)
{
byte2 absX = (byte2)abs(x);
byte2 absY = (byte2)abs(y);
return((absX / gcd(absX, absY)) * absY);
}
public byte8(byte3 x012, byte3 x345, byte2 x67)
{
if (Sse2.IsSse2Supported)
{
v128 mid = Sse2.bslli_si128(x345, 3 * sizeof(byte));
v128 hi = Sse2.bslli_si128(x67, 6 * sizeof(byte));
mid = Mask.BlendV(x012, mid, new byte8(0, 0, 0, 255, 255, 255, 0, 0));
if (Sse4_1.IsSse41Supported)
{
this = Sse4_1.blend_epi16(mid, hi, 0b1000);
}
else
{
this = Mask.BlendEpi16_SSE2(mid, hi, 0b1000);
}
}
else
{
this.x0 = x012.x;
this.x1 = x012.y;
this.x2 = x012.z;
this.x3 = x345.x;
this.x4 = x345.y;
this.x5 = x345.z;
this.x6 = x67.x;
this.x7 = x67.y;
}
}
public static byte2 floorpow2(byte2 x)
{
x |= x >> 1;
x |= x >> 2;
x |= x >> 4;
return(x - (x >> 1));
}
public byte2x4(byte m00, byte m01, byte m02, byte m03,
byte m10, byte m11, byte m12, byte m13)
{
this.c0 = new byte2(m00, m10);
this.c1 = new byte2(m01, m11);
this.c2 = new byte2(m02, m12);
this.c3 = new byte2(m03, m13);
}
public static byte2 ceilpow2(byte2 x)
{
x -= 1;
x |= x >> 1;
x |= x >> 2;
x |= x >> 4;
return(x + 1);
}
internal static byte2 vdiv_byte(byte2 dividend, byte2 divisor)
{
Assert.AreNotEqual(divisor.x, 0);
Assert.AreNotEqual(divisor.y, 0);
v128 floatResult = vdiv_byte_quotient((int2)dividend, (int2)divisor);
return((byte2)(*(float2 *)&floatResult));
}
public static byte2 andnot(byte2 left, byte2 right)
{
if (Sse2.IsSse2Supported)
{
return(Sse2.andnot_si128(right, left));
}
else
{
return(left & ~right);
}
}
public static bool any(byte2 x)
{
if (Sse2.IsSse2Supported)
{
return(0 != Sse2.extract_epi16(x, 0));
}
else
{
return(math.any(x != 0));
}
}
public static quarter2 asquarter(byte2 x)
{
if (Sse.IsSseSupported)
{
return((v128)x);
}
else
{
return(*(quarter2 *)&x);
}
}
public static byte2 avg(byte2 x, byte2 y)
{
if (Sse2.IsSse2Supported)
{
return(Sse2.avg_epu8(x, y));
}
else
{
return(new byte2((byte)((x.x + y.x + 1) >> 1), (byte)((x.y + y.y + 1) >> 1)));
}
}
public static byte avg(byte2 c)
{
if (Sse2.IsSse2Supported)
{
return(Sse2.avg_epu8(c, Sse2.bsrli_si128(c, 1 * sizeof(byte))).Byte0);
}
else
{
return((byte)((1u + csum(c)) / 2u));
}
}
public static byte2 max(byte2 a, byte2 b)
{
if (Sse2.IsSse2Supported)
{
return(Sse2.max_epu8(a, b));
}
else
{
return(new byte2((byte)math.max((uint)a.x, (uint)b.x), (byte)math.max((uint)a.y, (uint)b.y)));
}
}
public static byte cmin(byte2 x)
{
if (Ssse3.IsSsse3Supported)
{
return(min(x, x.yy).x);
}
else
{
return((byte)math.min((uint)x.x, (uint)x.y));
}
}
public static bool all(byte2 x)
{
if (Sse2.IsSse2Supported)
{
return(0 == Sse2.extract_epi16(Sse2.cmpeq_epi8(x, default(v128)), 0));
}
else
{
return(math.all(x != 0));
}
}
public static byte2 subadd(byte2 a, byte2 b)
{
if (Ssse3.IsSsse3Supported)
{
return(a + Ssse3.sign_epi8(b, new byte2(255, 1)));
}
else
{
return(a - select(b, (byte2)(-(sbyte2)b), new bool2(false, true)));
}
}
internal static byte2 vrem_byte(byte2 dividend, byte2 divisor)
{
Assert.AreNotEqual(divisor.x, 0);
Assert.AreNotEqual(divisor.y, 0);
int2 castDividend = dividend;
int2 castDivisor = divisor;
v128 floatResult = vdiv_byte_quotient(castDividend, castDivisor);
return((byte2)(castDividend - ((int2)(*(float2 *)&floatResult) * castDivisor)));
}
public static byte2 divrem(byte2 dividend, byte divisor, out byte2 remainder)
{
if (Constant.IsConstantExpression(divisor))
{
remainder = dividend % divisor;
return(dividend / divisor);
}
else
{
return(divrem(dividend, (byte2)divisor, out remainder));
}
}
public static byte2 divrem(byte2 dividend, byte2 divisor, out byte2 remainder)
{
if (Sse2.IsSse2Supported)
{
return(Operator.vdivrem_byte(dividend, divisor, out remainder));
}
else
{
remainder = dividend % divisor;
return(dividend / divisor);
}
}
internal static byte2 vdivrem_byte(byte2 dividend, byte2 divisor, out byte2 remainder)
{
Assert.AreNotEqual(divisor.x, 0);
Assert.AreNotEqual(divisor.y, 0);
int2 castDividend = dividend;
int2 castDivisor = divisor;
v128 floatResult = vdiv_byte_quotient(castDividend, castDivisor);
int2 quotientCast = (int2)(*(float2 *)&floatResult);
remainder = (byte2)(castDividend - quotientCast * castDivisor);
return((byte2)quotientCast);
}
public byte2 NextByte(byte2 max)
{
if (Ssse3.IsSsse3Supported)
{
short2 temp = (short2)max * new short2(NextState(), NextState());
return(Ssse3.shuffle_epi8(temp, new byte4(1, 3, 0, 0)));
}
else
{
return((byte2)(((short2)max * new short2(NextState(), NextState())) >> 8));
}
}
public static bool2 ispow2(byte2 x)
{
if (Sse2.IsSse2Supported)
{
v128 result = Sse2.and_si128(Sse2.and_si128(Operator.greater_mask_byte(x, default(v128)),
Sse2.cmpeq_epi8(default(v128), x & (x - 1))),
new byte16(1));
return(*(bool2 *)&result);
}
else
{
return(new bool2(math.ispow2((uint)x.x), math.ispow2((uint)x.y)));
}
}
public static uint sad(byte2 a, byte2 b)
{
if (Sse2.IsSse2Supported)
{
v128 maskedA = Sse2.and_si128(a, new v128(maxmath.bitmask32(16), 0, 0, 0));
v128 maskedB = Sse2.and_si128(b, new v128(maxmath.bitmask32(16), 0, 0, 0));
return(Sse2.sad_epu8(maskedA, maskedB).UShort0);
}
else
{
return((uint)(math.abs(a.x - b.x) + math.abs(a.y - b.y)));
}
}
public static byte2 gcd(byte2 x, byte2 y)
{
if (Sse2.IsSse2Supported)
{
v128 ZERO = default(v128);
v128 result = ZERO;
v128 result_if_zero_any = ZERO;
v128 x_is_zero = Sse2.cmpeq_epi8(x, ZERO);
v128 y_is_zero = Sse2.cmpeq_epi8(y, ZERO);
v128 any_zero = Sse2.or_si128(x_is_zero, y_is_zero);
result_if_zero_any = Mask.BlendV(result_if_zero_any, y, x_is_zero);
result_if_zero_any = Mask.BlendV(result_if_zero_any, x, y_is_zero);
v128 doneMask = any_zero;
byte2 shift = tzcnt(x | y);
x = shrl(x, tzcnt(x));
do
{
y = shrl(y, tzcnt(y));
v128 tempX = x;
x = Sse2.min_epu8(x, y);
y = Sse2.max_epu8(y, tempX);
y -= x;
v128 loopCheck = Sse2.andnot_si128(doneMask, Sse2.cmpeq_epi8(y, ZERO));
result = Mask.BlendV(result, x, loopCheck);
doneMask = Sse2.or_si128(doneMask, loopCheck);
} while (-1 != doneMask.SShort0);
result = shl(result, shift);
result = Mask.BlendV(result, result_if_zero_any, any_zero);
return(result);
}
else
{
return(new byte2((byte)gcd((uint)x.x, (uint)y.x), (byte)gcd((uint)x.y, (uint)y.y)));
}
}
