public static byte8 tzcnt(byte8 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 & (byte8)(-(sbyte8)x);
byte8 first = Mask.BlendV(default(v128), new byte8(1), Sse2.cmpeq_epi8(compareMask, default(v128)));
byte8 second = Mask.BlendV(default(v128), new byte8(4), Sse2.cmpeq_epi8(compareMask & (byte8)0x0F, default(v128)));
byte8 third = Mask.BlendV(default(v128), new byte8(2), Sse2.cmpeq_epi8(compareMask & (byte8)0x33, default(v128)));
byte8 fourth = Mask.BlendV(default(v128), new byte8(1), Sse2.cmpeq_epi8(compareMask & (byte8)0x55, default(v128)));
return((first + second) + (third + fourth));
}
else
{
return(new byte8(tzcnt(x.x0), tzcnt(x.x1), tzcnt(x.x2), tzcnt(x.x3), tzcnt(x.x4), tzcnt(x.x5), tzcnt(x.x6), tzcnt(x.x7)));
}
}
internal static byte8 vrem_byte(byte8 dividend, byte8 divisor)
{
int8 castDividend = dividend;
int8 castDivisor = divisor;
return((byte8)(castDividend - ((int8)vdiv_byte_quotient(castDividend, castDivisor) * castDivisor)));
}
internal static byte8 vdivrem_byte_SSE_FALLBACK(byte8 dividend, byte8 divisor, out byte8 remainder)
{
Assert.AreNotEqual(divisor.x0, 0);
Assert.AreNotEqual(divisor.x1, 0);
Assert.AreNotEqual(divisor.x2, 0);
Assert.AreNotEqual(divisor.x3, 0);
Assert.AreNotEqual(divisor.x4, 0);
Assert.AreNotEqual(divisor.x5, 0);
Assert.AreNotEqual(divisor.x6, 0);
Assert.AreNotEqual(divisor.x7, 0);
if (Sse2.IsSse2Supported)
{
ushort8 quotients = ushort8.zero;
ushort8 remainders = ushort8.zero;
ushort8 divisorCast = divisor;
ushort8 dividendCast = dividend;
remainders |= (new ushort8(1) & (dividendCast >> 7));
v128 subtractDivisorFromRemainder = Sse2.cmpeq_epi16(maxmath.min(divisorCast, remainders), divisorCast);
remainders -= Mask.BlendV(default(v128), divisorCast, subtractDivisorFromRemainder);
quotients |= new ushort8(1) & subtractDivisorFromRemainder;
for (int i = 6; i > 0; i--)
{
quotients <<= 1;
remainders <<= 1;
remainders |= (new ushort8(1) & (dividendCast >> i));
subtractDivisorFromRemainder = Sse2.cmpeq_epi8(maxmath.min(divisorCast, remainders), divisorCast);
remainders -= Mask.BlendV(default(v128), divisorCast, subtractDivisorFromRemainder);
quotients |= new ushort8(1) & subtractDivisorFromRemainder;
}
remainders <<= 1;
quotients <<= 1;
remainders |= new ushort8(1) & dividendCast;
subtractDivisorFromRemainder = Sse2.cmpeq_epi16(maxmath.min(divisorCast, remainders), divisorCast);
remainders -= Mask.BlendV(default(v128), divisorCast, subtractDivisorFromRemainder);
quotients |= new ushort8(1) & subtractDivisorFromRemainder;
byte16 temp = Sse2.packus_epi16(remainders, quotients);
remainder = temp.v8_0;
return(temp.v8_8);
}
else
{
throw new CPUFeatureCheckException();
}
}
public static byte8 reversebits(byte8 x)
{
x = ((x >> 1) & 0x55) | ((x & 0x55) << 1);
x = ((x >> 2) & 0x33) | ((x & 0x33) << 2);
return((x >> 4) | (x << 4));
}
public static byte2x3 operator %(byte2x3 left, byte2x3 right)
{
if (Sse2.IsSse2Supported)
{
#if DEBUG
byte8 packed_LHS = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(left.c0, left.c1),
Sse2.unpacklo_epi16(left.c2, new byte2(1)));
byte8 packed_RHS = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(right.c0, right.c1),
Sse2.unpacklo_epi16(right.c2, new byte2(1)));
byte8 rem = packed_LHS % packed_RHS;
return(new byte2x3(rem.v2_0, rem.v2_2, rem.v2_4));
#else
byte8 packed_LHS = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(left.c0, left.c1),
left.c2);
byte8 packed_RHS = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(right.c0, right.c1),
right.c2);
byte8 rem = packed_LHS % packed_RHS;
return(new byte2x3(rem.v2_0, rem.v2_2, rem.v2_4));
#endif
}
else
{
return(new byte2x3(left.c0 % right.c0, left.c1 % right.c1, left.c2 % right.c2));
}
}
public static byte2x3 operator %(byte2x3 left, byte right)
{
if (Sse2.IsSse2Supported)
{
if (!Constant.IsConstantExpression(right))
{
#if DEBUG
byte8 packed = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(left.c0, left.c1),
Sse2.unpacklo_epi16(left.c2, new byte2(1)));
byte8 rem = packed % right;
return(new byte2x3(rem.v2_0, rem.v2_2, rem.v2_4));
#else
byte8 packed = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(left.c0, left.c1),
left.c2);
byte8 rem = packed % right;
return(new byte2x3(rem.v2_0, rem.v2_2, rem.v2_4));
#endif
}
}
return(new byte2x3(left.c0 % right, left.c1 % right, left.c2 % right));
}
public static byte2x3 operator /(byte2x3 left, byte2x3 right)
{
if (Sse2.IsSse2Supported)
{
#if DEBUG
byte8 packed_LHS = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(left.c0, left.c1),
Sse2.unpacklo_epi16(left.c2, new byte2(1)));
byte8 packed_RHS = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(right.c0, right.c1),
Sse2.unpacklo_epi16(right.c2, new byte2(1)));
byte8 div = packed_LHS / packed_RHS;
return(new byte2x3(div.v2_0, div.v2_2, div.v2_4));
#else
byte8 packed_LHS = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(left.c0, left.c1),
left.c2);
byte8 packed_RHS = Sse2.unpacklo_epi32(Sse2.unpacklo_epi16(right.c0, right.c1),
right.c2);
byte8 div = packed_LHS / packed_RHS;
return(new byte2x3(div.v2_0, div.v2_2, div.v2_4));
#endif
}
else
{
return(new byte2x3(left.c0 / right.c0, left.c1 / right.c1, left.c2 / right.c2));
}
}
public static byte8 lcm(sbyte8 x, sbyte8 y)
{
byte8 absX = (byte8)abs(x);
byte8 absY = (byte8)abs(y);
return((absX / gcd(absX, absY)) * absY);
}
public static byte8 bitmask8(byte8 numBits, byte8 index = default(byte8))
{
Assert.IsBetween(index.x0, 0u, 8u);
Assert.IsBetween(index.x1, 0u, 8u);
Assert.IsBetween(index.x2, 0u, 8u);
Assert.IsBetween(index.x3, 0u, 8u);
Assert.IsBetween(index.x4, 0u, 8u);
Assert.IsBetween(index.x5, 0u, 8u);
Assert.IsBetween(index.x6, 0u, 8u);
Assert.IsBetween(index.x7, 0u, 8u);
Assert.IsBetween(numBits.x0, 0u, 8u - index.x0);
Assert.IsBetween(numBits.x1, 0u, 8u - index.x1);
Assert.IsBetween(numBits.x2, 0u, 8u - index.x2);
Assert.IsBetween(numBits.x3, 0u, 8u - index.x3);
Assert.IsBetween(numBits.x4, 0u, 8u - index.x4);
Assert.IsBetween(numBits.x5, 0u, 8u - index.x5);
Assert.IsBetween(numBits.x6, 0u, 8u - index.x6);
Assert.IsBetween(numBits.x7, 0u, 8u - index.x7);
if (Sse2.IsSse2Supported)
{
// mask
index = shl(byte.MaxValue, index);
v128 isMaxBitsMask = Sse2.cmpeq_epi8(numBits, new byte8(8));
return(isMaxBitsMask | andnot(index, shl(index, numBits)));
}
else
{
return((byte8)(-toint16(numBits == 16)) | andnot(index, shl(index, numBits)));
}
}
public static byte8 floorpow2(byte8 x)
{
x |= x >> 1;
x |= x >> 2;
x |= x >> 4;
return(x - (x >> 1));
}
public static byte8 ceilpow2(byte8 x)
{
x -= 1;
x |= x >> 1;
x |= x >> 2;
x |= x >> 4;
return(x + 1);
}
internal static byte8 vdivrem_byte(byte8 dividend, byte8 divisor, out byte8 remainder)
{
int8 castDividend = dividend;
int8 castDivisor = divisor;
int8 quotientCast = (int8)vdiv_byte_quotient(castDividend, castDivisor);
remainder = (byte8)(castDividend - quotientCast * castDivisor);
return((byte8)quotientCast);
}
public static byte8 gcd(byte8 x, byte8 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;
byte8 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.SLong0);
result = shl(result, shift);
result = Mask.BlendV(result, result_if_zero_any, any_zero);
return(result);
}
else
{
return(new byte8((byte)gcd((uint)x.x0, (uint)y.x0),
(byte)gcd((uint)x.x1, (uint)y.x1),
(byte)gcd((uint)x.x2, (uint)y.x2),
(byte)gcd((uint)x.x3, (uint)y.x3),
(byte)gcd((uint)x.x4, (uint)y.x4),
(byte)gcd((uint)x.x5, (uint)y.x5),
(byte)gcd((uint)x.x6, (uint)y.x6),
(byte)gcd((uint)x.x7, (uint)y.x7)));
}
}
public static byte8 andnot(byte8 left, byte8 right)
{
if (Sse2.IsSse2Supported)
{
return(Sse2.andnot_si128(right, left));
}
else
{
return(left & ~right);
}
}
public DebuggerProxy(byte8 v)
{
x0 = v.x0;
x1 = v.x1;
x2 = v.x2;
x3 = v.x3;
x4 = v.x4;
x5 = v.x5;
x6 = v.x6;
x7 = v.x7;
}
public static uint sad(byte8 a, byte8 b)
{
if (Sse2.IsSse2Supported)
{
return(Sse2.sad_epu8(a, b).UShort0);
}
else
{
return((uint)(((math.abs(a.x0 - b.x0) + math.abs(a.x1 - b.x1)) + (math.abs(a.x2 - b.x2) + math.abs(a.x3 - b.x3))) + ((math.abs(a.x4 - b.x4) + math.abs(a.x5 - b.x5)) + (math.abs(a.x6 - b.x6) + math.abs(a.x7 - b.x7)))));
}
}
public static bool any(byte8 x)
{
if (Sse2.IsSse2Supported)
{
return(0 != ((v128)x).ULong0);
}
else
{
return(any(x != 0));
}
}
public static bool all_eq(byte8 c)
{
if (Ssse3.IsSsse3Supported)
{
return(((byte8)Ssse3.shuffle_epi8(c, default(v128))).Equals(c));
}
else
{
return(((c.x0 == c.x1 & c.x0 == c.x2) & (c.x0 == c.x3 & c.x0 == c.x4)) & ((c.x0 == c.x5 & c.x0 == c.x6) & c.x0 == c.x7));
}
}
public static byte8 avg(byte8 x, byte8 y)
{
if (Sse2.IsSse2Supported)
{
return(Sse2.avg_epu8(x, y));
}
else
{
return(new byte8((byte)((x.x0 + y.x0 + 1) >> 1), (byte)((x.x1 + y.x1 + 1) >> 1), (byte)((x.x2 + y.x2 + 1) >> 1), (byte)((x.x3 + y.x3 + 1) >> 1), (byte)((x.x4 + y.x4 + 1) >> 1), (byte)((x.x5 + y.x5 + 1) >> 1), (byte)((x.x6 + y.x6 + 1) >> 1), (byte)((x.x7 + y.x7 + 1) >> 1)));
}
}
public static quarter8 asquarter(byte8 x)
{
if (Sse.IsSseSupported)
{
return((v128)x);
}
else
{
return(*(quarter8 *)&x);
}
}
public static byte8 max(byte8 a, byte8 b)
{
if (Sse2.IsSse2Supported)
{
return(Sse2.max_epu8(a, b));
}
else
{
return(new byte8((byte)math.max((uint)a.x0, (uint)b.x0), (byte)math.max((uint)a.x1, (uint)b.x1), (byte)math.max((uint)a.x2, (uint)b.x2), (byte)math.max((uint)a.x3, (uint)b.x3), (byte)math.max((uint)a.x4, (uint)b.x4), (byte)math.max((uint)a.x5, (uint)b.x5), (byte)math.max((uint)a.x6, (uint)b.x6), (byte)math.max((uint)a.x7, (uint)b.x7)));
}
}
public static uint csum(byte8 x)
{
if (Sse2.IsSse2Supported)
{
return(sad(x, byte8.zero));
}
else
{
return((uint)(((x.x0 + x.x1) + (x.x2 + x.x3)) + ((x.x4 + x.x5) + (x.x6 + x.x7))));
}
}
public static byte8 subadd(byte8 a, byte8 b)
{
if (Ssse3.IsSsse3Supported)
{
return(a + Ssse3.sign_epi8(b, new byte8(255, 1, 255, 1, 255, 1, 255, 1)));
}
else
{
return(a - select(b, (byte8)(-(sbyte8)b), new bool8(false, true, false, true, false, true, false, true)));
}
}
public static byte8 countbits(byte8 x)
{
if (Ssse3.IsSsse3Supported)
{
return((v128)countbits((byte16)(v128)x));
}
else
{
return(new byte8((byte)math.countbits((uint)x.x0), (byte)math.countbits((uint)x.x1), (byte)math.countbits((uint)x.x2), (byte)math.countbits((uint)x.x3), (byte)math.countbits((uint)x.x4), (byte)math.countbits((uint)x.x5), (byte)math.countbits((uint)x.x6), (byte)math.countbits((uint)x.x7)));
}
}
public static bool all(byte8 x)
{
if (Sse2.IsSse2Supported)
{
return(0 == Sse2.cmpeq_epi8(x, default(v128)).ULong0);
}
else
{
return(all(x != 0));
}
}
public static byte8 divrem(byte8 dividend, byte divisor, out byte8 remainder)
{
if (Constant.IsConstantExpression(divisor))
{
remainder = dividend % divisor;
return(dividend / divisor);
}
else
{
return(divrem(dividend, (byte8)divisor, out remainder));
}
}
public static byte2x4 operator %(byte2x4 left, byte2x4 right)
{
if (Sse2.IsSse2Supported)
{
byte8 rem = new byte8(left.c0, left.c1, left.c2, left.c3) % new byte8(right.c0, right.c1, right.c2, right.c3);
return(new byte2x4(rem.v2_0, rem.v2_2, rem.v2_4, rem.v2_6));
}
else
{
return(new byte2x4(left.c0 % right.c0, left.c1 % right.c1, left.c2 % right.c2, left.c3 % right.c3));
}
}
public static byte2x4 operator /(byte2x4 left, byte2x4 right)
{
if (Sse2.IsSse2Supported)
{
byte8 div = new byte8(left.c0, left.c1, left.c2, left.c3) / new byte8(right.c0, right.c1, right.c2, right.c3);
return(new byte2x4(div.v2_0, div.v2_2, div.v2_4, div.v2_6));
}
else
{
return(new byte2x4(left.c0 / right.c0, left.c1 / right.c1, left.c2 / right.c2, left.c3 / right.c3));
}
}
public static bool8 ispow2(byte8 x)
{
if (Sse2.IsSse2Supported)
{
return(Sse2.and_si128(Sse2.and_si128(Operator.greater_mask_byte(x, byte8.zero),
Sse2.cmpeq_epi8(default(v128), x & (x - 1))),
new byte16(1)));
}
else
{
return(new bool8(math.ispow2((uint)x.x0), math.ispow2((uint)x.x1), math.ispow2((uint)x.x2), math.ispow2((uint)x.x3), math.ispow2((uint)x.x4), math.ispow2((uint)x.x5), math.ispow2((uint)x.x6), math.ispow2((uint)x.x7)));
}
}
public byte8 NextByte8(byte8 max)
{
if (Ssse3.IsSsse3Supported)
{
short8 temp = (short8)max * new short8(NextState(), NextState(), NextState(), NextState(), NextState(), NextState(), NextState(), NextState());
return(Ssse3.shuffle_epi8(temp, new byte8(1, 3, 5, 7, 9, 11, 13, 15)));
}
else
{
return((byte8)(((short8)max * new short8(NextState(), NextState(), NextState(), NextState(), NextState(), NextState(), NextState(), NextState())) >> 8));
}
}
