public static void ulong4()
{
Random64 rng = new Random64(135);
for (ulong i = 0; i < 64; i++)
{
ulong4 x = rng.NextULong4();
ulong4 y = rng.NextULong4();
Assert.AreEqual(new ulong4((ulong)_gcd(x.x, y.x), (ulong)_gcd(x.y, y.y), (ulong)_gcd(x.z, y.z), (ulong)_gcd(x.w, y.w)), maxmath.gcd(x, y));
}
}
public static void long4()
{
Random64 rng = new Random64(135);
for (long i = 0; i < 64; i++)
{
long4 x = rng.NextLong4();
ulong4 n = rng.NextULong4();
Assert.AreEqual(new long4((long)_intpow(x.x, n.x), (long)_intpow(x.y, n.y), (long)_intpow(x.z, n.z), (long)_intpow(x.w, n.w)), maxmath.intpow(x, n));
}
}
public static bool4 ispow2(ulong4 x)
{
if (Avx2.IsAvx2Supported)
{
v128 result = new byte4(1) & ((byte4)(long4)Avx2.mm256_and_si256(Operator.greater_mask_ulong(x, default(v256)),
Avx2.mm256_cmpeq_epi64(default(v256), x & (x - 1))));
return(*(bool4 *)&result);
}
else
{
return(new bool4(ispow2(x.xy), ispow2(x.zw)));
}
}
public static ulong cmin(ulong4 x)
{
if (Sse4_2.IsSse42Supported)
{
ulong2 temp = min(x.xy, x.zw);
return(min(temp, temp.yy).x);
}
else
{
return(math.min(x.x, math.min(x.y, math.min(x.z, x.w))));
}
}
internal static v256 greater_mask_ulong(v256 left, v256 right)
{
if (Avx2.IsAvx2Supported)
{
ulong4 mask = 1ul << 63;
return(Avx2.mm256_cmpgt_epi64(Avx2.mm256_xor_si256(left, mask),
Avx2.mm256_xor_si256(right, mask)));
}
else
{
throw new CPUFeatureCheckException();
}
}
public static ulong4 gcd(ulong4 x, ulong4 y)
{
if (Avx2.IsAvx2Supported)
{
v256 ZERO = default(v256);
v256 result = ZERO;
v256 result_if_zero_any = ZERO;
v256 x_is_zero = Avx2.mm256_cmpeq_epi64(x, ZERO);
v256 y_is_zero = Avx2.mm256_cmpeq_epi64(y, ZERO);
v256 any_zero = Avx2.mm256_or_si256(x_is_zero, y_is_zero);
result_if_zero_any = Avx2.mm256_blendv_epi8(result_if_zero_any, y, x_is_zero);
result_if_zero_any = Avx2.mm256_blendv_epi8(result_if_zero_any, x, y_is_zero);
v256 doneMask = any_zero;
v256 shift = tzcnt(x | y);
x = Avx2.mm256_srlv_epi64(x, tzcnt(x));
do
{
y = Avx2.mm256_srlv_epi64(y, tzcnt(y));
v256 tempX = x;
v256 x_greater_y = Operator.greater_mask_ulong(x, y);
x = Avx2.mm256_blendv_epi8(x, y, x_greater_y);
y = Avx2.mm256_blendv_epi8(y, tempX, x_greater_y);
y -= x;
v256 loopCheck = Avx2.mm256_andnot_si256(doneMask, Avx2.mm256_cmpeq_epi64(y, ZERO));
result = Avx2.mm256_blendv_epi8(result, x, loopCheck);
doneMask = Avx2.mm256_or_si256(doneMask, loopCheck);
} while (bitmask32(4 * sizeof(ulong)) != Avx2.mm256_movemask_epi8(doneMask));
result = Avx2.mm256_sllv_epi64(result, shift);
result = Avx2.mm256_blendv_epi8(result, result_if_zero_any, any_zero);
return(result);
}
else
{
return(new ulong4(gcd(x._xy, y._xy), gcd(x._zw, y._zw)));
}
}
public static long4 compareto(ulong4 x, ulong4 y)
{
if (Avx2.IsAvx2Supported)
{
long4 xGreatery = Operator.greater_mask_ulong(x, y);
long4 yGreaterx = Operator.greater_mask_ulong(y, x);
return((0 - xGreatery) + yGreaterx);
}
else
{
return(new long4(compareto(x.xy, y.xy),
compareto(x.zw, y.zw)));
}
}
public static void ulong4()
{
for (int i = 0; i < ULong4.TestData_LHS.Length; i++)
{
for (int j = 1; j < 4; j++)
{
ulong4 rol = maxmath.vrol(ULong4.TestData_LHS[i], j);
for (int k = 0; k < 4; k++)
{
Assert.AreEqual(rol[k], ULong4.TestData_LHS[i][((4 - j) + k) % 4]);
}
}
}
}
public static void ulong4()
{
for (int i = 0; i < ULong4.TestData_LHS.Length; i++)
{
for (int j = 1; j < 4; j++)
{
ulong4 shl = maxmath.vshl(ULong4.TestData_LHS[i], j);
for (int k = 0; k < 4; k++)
{
Assert.IsTrue(shl[k] == ((k - j < 0) ? 0 : ULong4.TestData_LHS[i][k - j]));
}
}
}
}
public static void ulong4()
{
for (int i = 0; i < ULong4.TestData_LHS.Length; i++)
{
for (int j = 1; j < 4; j++)
{
ulong4 shr = maxmath.vshr(ULong4.TestData_LHS[i], j);
for (int k = 0; k < 4; k++)
{
Assert.IsTrue(shr[k] == ((j + k >= 4) ? 0 : ULong4.TestData_LHS[i][j + k]));
}
}
}
}
public static void Cast_ToULong()
{
bool result = true;
for (int i = 0; i < NUM_TESTS; i++)
{
ulong4 x = (ulong4)TestData_LHS[i];
result &= x.x == (ulong)TestData_LHS[i].x &
x.y == (ulong)TestData_LHS[i].y &
x.z == (ulong)TestData_LHS[i].z &
x.w == (ulong)TestData_LHS[i].w;
}
Assert.AreEqual(true, result);
}
public ulong4 NextULong4(ulong4 min, ulong4 max)
{
Assert.IsNotSmaller(max.x, min.x);
Assert.IsNotSmaller(max.y, min.y);
Assert.IsNotSmaller(max.z, min.z);
Assert.IsNotSmaller(max.w, min.w);
max -= min;
ulong4 result = ulong4.zero;
Common.umul128(NextState(), max.x, out result.x);
Common.umul128(NextState(), max.y, out result.y);
Common.umul128(NextState(), max.z, out result.z);
Common.umul128(NextState(), max.w, out result.w);
return(min + result);
}
public static bool4 isdivisible(uint4 dividend, uint4 divisor)
{
Assert.AreNotEqual(0u, divisor.x);
Assert.AreNotEqual(0u, divisor.y);
Assert.AreNotEqual(0u, divisor.z);
Assert.AreNotEqual(0u, divisor.w);
if (Constant.IsConstantExpression(divisor))
{
ulong4 compile = (new ulong4(ulong.MaxValue) / divisor) + 1;
return(dividend * compile <= compile - 1);
}
else
{
return(dividend % divisor == 0);
}
}
public static ulong4 floorpow2(ulong4 x)
{
if (Avx2.IsAvx2Supported)
{
return(shrl(0x8000_0000_0000_0000, lzcnt(x)));
}
else
{
x |= x >> 1;
x |= x >> 2;
x |= x >> 4;
x |= x >> 8;
x |= x >> 16;
x |= x >> 32;
return(x - (x >> 1));
}
}
public long4 NextLong4(long4 min, long4 max)
{
Assert.IsNotSmaller(max.x, min.x);
Assert.IsNotSmaller(max.y, min.y);
Assert.IsNotSmaller(max.z, min.z);
Assert.IsNotSmaller(max.w, min.w);
ulong4 result = ulong4.zero;
max -= min;
Common.umul128(NextState(), (ulong)(max.x), out result.x);
Common.umul128(NextState(), (ulong)(max.y), out result.y);
Common.umul128(NextState(), (ulong)(max.z), out result.z);
Common.umul128(NextState(), (ulong)(max.w), out result.w);
return(min + (long4)result);
}
public static void ror_ulong4()
{
bool result = true;
for (int i = 0; i < ULong4.NUM_TESTS; i++)
{
for (int j = 0; j < NUM_ROTATION_TESTS; j++)
{
ulong4 test = maxmath.ror(ULong4.TestData_LHS[i], ULong4.TestData_RHS[i]);
result &= test.x == math.ror(ULong4.TestData_LHS[i].x, (int)ULong4.TestData_RHS[i].x);
result &= test.y == math.ror(ULong4.TestData_LHS[i].y, (int)ULong4.TestData_RHS[i].y);
result &= test.z == math.ror(ULong4.TestData_LHS[i].z, (int)ULong4.TestData_RHS[i].z);
result &= test.w == math.ror(ULong4.TestData_LHS[i].w, (int)ULong4.TestData_RHS[i].w);
}
}
Assert.AreEqual(true, result);
}
public static ulong4 ceilpow2(ulong4 x)
{
if (Avx2.IsAvx2Supported)
{
return(shrl(0x8000_0000_0000_0000, lzcnt(x - 1) - 1));
}
else
{
x -= 1;
x |= x >> 1;
x |= x >> 2;
x |= x >> 4;
x |= x >> 8;
x |= x >> 16;
x |= x >> 32;
return(x + 1);
}
}
public static long4 intpow(long4 x, ulong4 n)
{
if (Avx2.IsAvx2Supported)
{
v256 ZERO = long4.zero;
v256 ONE = new long4(1);
v256 doneMask = ZERO;
v256 result = ZERO;
v256 p = x;
v256 y = ONE;
Loop:
v256 y_times_p = Operator.mul_long(y, p);
y = Avx2.mm256_blendv_epi8(y, y_times_p, Avx2.mm256_cmpeq_epi64(ONE, ONE & n));
n >>= 1;
v256 n_is_zero = Avx2.mm256_cmpeq_epi64(ZERO, n);
result = Avx2.mm256_blendv_epi8(result, y, Avx2.mm256_andnot_si256(doneMask, n_is_zero));
doneMask = n_is_zero;
if (-1 != Avx2.mm256_movemask_epi8(doneMask))
{
p = Operator.mul_long(p, p);
goto Loop;
}
else
{
return(result);
}
}
else
{
return(new long4(intpow(x.xy, n.xy), intpow(x.zw, n.zw)));
}
}
public static ulong4 lzcnt(ulong4 x)
{
if (Avx2.IsAvx2Supported)
{
ulong4 y = x >> 32;
ulong4 cmp = Avx2.mm256_cmpeq_epi64(y, default(v256));
ulong4 bits = Avx2.mm256_blendv_epi8(y, 0x0000_0000_FFFF_FFFF & x, cmp);
ulong4 offset = Avx2.mm256_blendv_epi8((ulong4)0x041E, (ulong4)0x043E, cmp);
bits += 0x4330_0000_0000_0000ul;
bits = Avx.mm256_sub_pd(bits, new v256(4503599627370496d));
bits = offset - (bits >> 52);
return(Avx2.mm256_blendv_epi8(bits, new ulong4(64), Avx2.mm256_cmpeq_epi64(x, default(v256))));
}
else
{
return(new ulong4((ulong)math.lzcnt(x.x), (ulong)math.lzcnt(x.y), (ulong)math.lzcnt(x.z), (ulong)math.lzcnt(x.w)));
}
}
public static void rol_ulong4()
{
bool result = true;
Random32 rng = new Random32(RNG_SEED);
for (int i = 0; i < ULong4.NUM_TESTS; i++)
{
for (int j = 0; j < NUM_ROTATION_TESTS; j++)
{
int n = rng.NextInt();
ulong4 test = maxmath.rol(ULong4.TestData_LHS[i], n);
result &= test.x == math.rol(ULong4.TestData_LHS[i].x, n);
result &= test.y == math.rol(ULong4.TestData_LHS[i].y, n);
result &= test.z == math.rol(ULong4.TestData_LHS[i].z, n);
result &= test.w == math.rol(ULong4.TestData_LHS[i].w, n);
}
}
Assert.AreEqual(true, result);
}
public static bool4 isdivisible(uint4 dividend, uint divisor)
{
Assert.AreNotEqual(0u, divisor);
if (Constant.IsConstantExpression(divisor))
{
if (math.ispow2(divisor))
{
return((dividend & (divisor - 1)) == 0);
}
else
{
ulong4 compile = (new ulong4(ulong.MaxValue) / divisor) + 1;
return(dividend * compile <= compile - 1);
}
}
else
{
return(dividend % divisor == 0);
}
}
internal static double4 ULong4ToDouble4(ulong4 x)
{
if (Avx2.IsAvx2Supported)
{
ulong4 magic_lo = 0x4330_0000_0000_0000;
ulong4 magic_hi = 0x4530_0000_0000_0000;
ulong4 magic_dbl = 0x4530_0000_0010_0000;
ulong4 lo = Avx2.mm256_blend_epi32(magic_lo, x, 0b0101_0101);
ulong4 hi = magic_hi ^ (x >> 32);
v256 hi_dbl = Avx.mm256_sub_pd(hi, magic_dbl);
v256 result = Avx.mm256_add_pd(hi_dbl, lo);
return(*(double4 *)&result);
}
else
{
throw new CPUFeatureCheckException();
}
}
public static bool8 isdivisible(uint8 dividend, uint8 divisor)
{
Assert.AreNotEqual(0u, divisor.x0);
Assert.AreNotEqual(0u, divisor.x1);
Assert.AreNotEqual(0u, divisor.x2);
Assert.AreNotEqual(0u, divisor.x3);
Assert.AreNotEqual(0u, divisor.x4);
Assert.AreNotEqual(0u, divisor.x5);
Assert.AreNotEqual(0u, divisor.x6);
Assert.AreNotEqual(0u, divisor.x7);
if (Constant.IsConstantExpression(divisor))
{
ulong4 compile_lo = (new ulong4(ulong.MaxValue) / divisor.v4_0) + 1;
ulong4 compile_hi = (new ulong4(ulong.MaxValue) / divisor.v4_4) + 1;
return(new bool8(dividend.v4_0 * compile_lo <= compile_lo - 1,
dividend.v4_4 * compile_lo <= compile_lo - 1));
}
else
{
return(dividend % divisor == 0);
}
}
public static ulong4 clamp(ulong4 x, ulong4 a, ulong4 b)
{
return(max(a, min(x, b)));
}
public static ulong4 countbits(ulong4 x)
{
return(new ulong4((uint)math.countbits(x.x), (uint)math.countbits(x.y), (uint)math.countbits(x.z), (uint)math.countbits(x.w)));
}
public static bool4 isinrange(ulong4 x, ulong4 min, ulong4 max)
{
return(maxmath.min(maxmath.max(x, min), max) == x);
}
public ulong4x3(ulong4 c0, ulong4 c1, ulong4 c2)
{
this.c0 = c0;
this.c1 = c1;
this.c2 = c2;
}
public static extern CUResult cuMemcpyDtoH_v2(ref ulong4 dstHost, CUdeviceptr srcDevice, SizeT ByteCount);
public static bool all(ulong4 x)
{
return(all((long4)x));
}
public ulong4x3(ulong v)
{
this.c0 = v;
this.c1 = v;
this.c2 = v;
}
public static ulong4 andnot(ulong4 left, ulong4 right)
{
return((ulong4)andnot((long4)left, (long4)right));
}