classiq 0.99.0__py3-none-any.whl → 0.102.0__py3-none-any.whl

classiq/open_library/functions/modular_arithmetics.py

@@ -0,0 +1,597 @@
+import warnings
+
+from classiq.interface.exceptions import ClassiqDeprecationWarning
+
+from classiq.open_library.functions.bit_operations import (
+    bitwise_negate,
+    cyclic_shift_left,
+)
+from classiq.open_library.functions.qft_functions import qft
+from classiq.open_library.functions.qft_space_arithmetics import modular_add_qft_space
+from classiq.open_library.functions.utility_functions import multiswap
+from classiq.qmod.builtins.constants import SIGNED
+from classiq.qmod.builtins.functions.allocation import free
+from classiq.qmod.builtins.functions.standard_gates import X
+from classiq.qmod.builtins.operations import (
+    allocate,
+    bind,
+    control,
+    if_,
+    inplace_add,
+    inplace_xor,
+    invert,
+    repeat,
+    within_apply,
+)
+from classiq.qmod.cparam import CInt
+from classiq.qmod.qfunc import qperm
+from classiq.qmod.qmod_variable import Const, Output, QArray, QBit, QNum
+from classiq.qmod.symbolic import mod_inverse
+
+# Modular Adding and Subtraction
+
+
+@qperm
+def modular_add_inplace(modulus: CInt, x: Const[QNum], y: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+
+    Performs the transformation |x>|y> -> |x>|(x + y mod modulus)>.
+    Note:
+    |x> and |y> should have values smaller than `modulus`.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x> and |y>.
+
+    Implementation based on: https://arxiv.org/pdf/1706.06752 Chapter 3.2 Fig 3
+
+    Args:
+        modulus: Classical number modulus (CInt)
+        x: 1st quantum number input (unsigned).
+        y: 2nd quantum number input (unsigned). Will hold the result after the operation.
+    """
+    # Use a carry qubit to detect a negative result after subtracting modulus (underflow)
+    carry = QBit()
+    allocate(carry)
+    temp: QNum = QNum("temp", y.size + 1, SIGNED, 0)
+    within_apply(
+        lambda: bind([y, carry], temp),
+        lambda: (
+            inplace_add(x, temp),
+            inplace_add(-modulus, temp),
+        ),
+    )
+    # If carry is set (negative result), add modulus back to y
+    control(carry, lambda: inplace_add(modulus, y))
+    # Update carry qubit based on comparison (y >= x after operation)
+    carry ^= y >= x
+    free(carry)
+
+
+@qperm
+def modular_negate_inplace(modulus: CInt, x: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+
+    Performs the transformation |x> -> |(-x mod modulus)>.
+    Note:
+    |x> should have values smaller than `modulus`.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x>.
+
+    Args:
+        modulus: Classical number modulus
+        x: Quantum number input (unsigned). Will hold the result after the operation.
+    """
+    n = x.size
+    neg_modulus = 2**n - modulus - 1
+    is_all_zeros = QBit()
+    allocate(is_all_zeros)
+    is_all_zeros ^= x == 0
+    control(is_all_zeros, lambda: inplace_add(modulus, x))
+    inplace_add(neg_modulus, x)
+    # If x=0, then we have neg_modulus + modulus = all ones
+    is_all_zeros ^= x == (2**n - 1)
+    bitwise_negate(x)
+    free(is_all_zeros)
+
+
+@qperm
+def modular_subtract_inplace(modulus: CInt, x: Const[QNum], y: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+
+    Performs the transformation |x>|y> -> |x>|(x - y mod modulus)>.
+    Note:
+    |x> and |y> should have values smaller than `modulus`.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x> and |y>.
+
+    Args:
+        modulus: Classical number modulus
+        x: 1st quantum number input (unsigned). Const.
+        y: 2nd quantum number input (unsigned). In-place target, will hold the result after the operation.
+    """
+    modular_negate_inplace(modulus, y)
+    modular_add_inplace(modulus, x, y)
+
+
+@qperm
+def modular_double_inplace(modulus: CInt, x: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+
+    Performs the transformation |x> -> |(2x mod modulus)>.
+    Note:
+    |x> should have a value smaller than `modulus`.
+    The modulus must be a constant odd integer.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x>.
+
+    Implementation based on: https://arxiv.org/pdf/1706.06752 Chapter 3.2 Fig 4
+
+    Args:
+        modulus: Classical number modulus
+        x: Quantum number input (unsigned). Will hold the result after the operation.
+    """
+    carry = QBit()
+    allocate(carry)
+    res_and_carry: QNum = QNum("res_and_carry", x.size + 1, SIGNED, 0)
+    within_apply(
+        lambda: bind([x, carry], res_and_carry),
+        lambda: (
+            cyclic_shift_left(res_and_carry),  # holds 2*x
+            inplace_add(-modulus, res_and_carry),
+        ),
+    )
+    control(carry, lambda: inplace_add(modulus, x))
+    # Post-fix carry
+    carry ^= (x % 2) == 0
+    free(carry)
+
+
+@qperm
+def modular_add_constant_inplace(modulus: CInt, a: CInt, x: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+
+    Performs the transformation |x> -> |(x + a mod modulus)>.
+    Note:
+    |x> and `a` should have values smaller than `modulus`.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x>.
+
+    Implementation is based on the logic in: https://arxiv.org/pdf/1706.06752 Chapter 3.2 Fig 3
+
+    Args:
+        modulus: Classical number modulus
+        a: constant unsigned number input for the addition.
+        x: Quantum number input (unsigned). Will hold the result after the operation.
+    """
+    carry = QBit()
+    allocate(carry)
+    temp: QNum = QNum("temp", x.size + 1, SIGNED, 0)
+    within_apply(
+        lambda: bind([x, carry], temp),
+        lambda: (
+            inplace_add(a, temp),
+            inplace_add(-modulus, temp),
+        ),
+    )
+    # If carry is set, we need to add modulus back
+    control(carry, lambda: inplace_add(modulus, x))
+    carry ^= x >= a
+    free(carry)
+
+
+# Modular Multiplication
+
+
+@qperm
+def modular_multiply(
+    modulus: CInt,
+    x: Const[QArray[QBit]],
+    y: Const[QArray[QBit]],
+    z: QArray[QBit],
+) -> None:
+    """
+    [Qmod Classiq-library function]
+    Performs the transformation |x>|y>|0> -> |x>|y>|(x*y mod modulus)>
+    Note:
+    |x>, |y> should have the same size and have values smaller than `modulus`.
+    The modulus must be a constant odd integer.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x> and |y>.
+    The output register z must be pre-allocated with the same size as x and y.
+    Implementation is based on the logic in: https://arxiv.org/pdf/1706.06752 Chapter 3.2 Fig 5
+    Args:
+        modulus: Classical number modulus
+        x: Quantum number input (unsigned), multiplicand.
+        y: Quantum number input (unsigned), multiplier.
+        z: Quantum number (unsigned), pre-allocated output variable that will hold the result.
+    """
+    n = x.len
+    repeat(
+        n,
+        lambda idx: [
+            control(x[n - idx - 1], lambda: modular_add_inplace(modulus, y, z)),
+            if_(idx != (n - 1), lambda: modular_double_inplace(modulus, z)),
+        ],
+    )
+
+
+@qperm
+def modular_square(modulus: CInt, x: Const[QArray[QBit]], z: QArray[QBit]) -> None:
+    """
+    [Qmod Classiq-library function]
+    Performs the transformation |x>|0> -> |x>|(x^2 mod modulus)>.
+    Note:
+    |x> should have the same size and have values smaller than `modulus`.
+    The modulus must be a constant odd integer.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x>.
+    The output register z must be pre-allocated with the same size as x.
+    Implementation is based on: https://arxiv.org/pdf/1706.06752 Chapter 3.2 Fig 6
+    Args:
+        modulus: Classical number modulus
+        x: Quantum number input (unsigned), the input to square.
+        z: Quantum number (unsigned), pre-allocated output variable to hold the result.
+    """
+    n = x.len
+    repeat(
+        n - 1,
+        lambda i: [
+            control(x[n - i - 1], lambda: modular_add_inplace(modulus, x, z)),
+            modular_double_inplace(modulus, z),
+        ],
+    )
+    control(x[0], lambda: modular_add_inplace(modulus, x, z))
+
+
+@qperm(disable_perm_check=True)
+def modular_multiply_constant(modulus: CInt, x: Const[QNum], a: CInt, y: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+    Performs the transformation |x>|y> -> |x>|(x * a mod modulus)>.
+    Note:
+    |x> and |y> should have values smaller than `modulus`.
+    The modulus must be a constant odd integer.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x> and |y>.
+
+    Args:
+        modulus: Classical number modulus
+        x: Quantum number (unsigned), input variable.
+        a: Classical number constant
+        y: Quantum number (unsigned), output variable that will hold the result.
+    """
+    x_arr: QArray[QBit] = QArray()
+    within_apply(
+        lambda: [bind(x, x_arr), qft(y)],
+        lambda: repeat(
+            count=x_arr.len,
+            iteration=lambda index: control(
+                x_arr[index],
+                lambda: modular_add_qft_space(modulus, (a * (2**index)) % modulus, y),
+            ),
+        ),
+    )
+
+
+@qperm
+def modular_multiply_constant_inplace(modulus: CInt, a: CInt, x: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+    In-place modular multiplication of x by a classical constant modulo a symbolic modulus.
+    Performs |x> -> |(x * a mod modulus)>.
+    Note:
+    |x> should have values smaller than `modulus`.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x>.
+    The constant `a` should have an inverse modulo `modulus`, i.e. gcd(a, modulus) = 1.
+    The constant `a` should satisfy 0 <= a < modulus.
+
+    Args:
+        modulus: Classical number modulus
+        a: Classical number constant
+        x: Quantum number (unsigned), in-place input/output.
+    """
+    y: QNum = QNum("y", x.size + 1)
+    allocate(y)
+    modular_multiply_constant(modulus, x, a, y)
+    multiswap(x, y)
+    invert(lambda: modular_multiply_constant(modulus, x, mod_inverse(a, modulus), y))
+    free(y)
+
+
+@qperm
+def inplace_modular_multiply(n: CInt, a: CInt, x: QArray[QBit]) -> None:
+    """
+    [Qmod Classiq-library function]
+
+    Performs multiplication of a quantum number `x` by a classical number `a` modulo classical number `n`
+    (Applies $x=xa \\mod n$).
+
+    Args:
+        n: The modulo number. Should be non-negative.
+        a: The classical factor. Should be non-negative.
+        x: The quantum factor.
+
+    Comment: It is assumed that `a` has an inverse modulo `n`
+    """
+    warnings.warn(
+        "Function 'inplace_modular_multiply' is deprecated. Use 'modular_multiply_constant_inplace' instead.",
+        ClassiqDeprecationWarning,
+        stacklevel=1,
+    )
+    modular_multiply_constant_inplace(n, a, x)
+
+
+@qperm
+def modular_exp(n: CInt, a: CInt, x: QArray[QBit], power: QArray[QBit]) -> None:
+    """
+    [Qmod Classiq-library function]
+    Raises a classical integer `a` to the power of a quantum number `power` modulo classical integer `n`
+    times a quantum number `x`. Performs $x=(a^{power} \\mod n)*x$ in-place.
+    (and specifically if at the input $x=1$, at the output $x=a^{power} \\mod n$).
+    Args:
+        n: The modulus number. Should be non-negative.
+        a: The base of the exponentiation. Should be non-negative.
+        x: A quantum number that multiplies the modular exponentiation and holds the output. It should be at least the size of $\\lceil \\log(n) \rceil$.
+        power: The power of the exponentiation.
+    """
+    warnings.warn(
+        "Function 'modular_exp' is deprecated. Use 'modular_exponentiate_inplace' instead.",
+        ClassiqDeprecationWarning,
+        stacklevel=1,
+    )
+    repeat(
+        count=power.len,
+        iteration=lambda index: control(
+            power[index],
+            lambda: modular_multiply_constant_inplace(n, (a ** (2**index)) % n, x),
+        ),
+    )
+
+
+@qperm
+def modular_exponentiate(modulus: CInt, a: CInt, x: QNum, p: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+
+    Raises a classical integer `a` to the power of a quantum number `p` modulo classical integer `modulus`
+    times a quantum number `x`. Performs $x=(a^{p} \\mod modulus)*x$ in-place.
+    (and specifically if at the input $x=1$, at the output $x=a^{p} \\mod modulus$).
+
+    Args:
+        modulus: The modulus number. Should be non-negative.
+        a: The base of the exponentiation. Should be non-negative.
+        x: A quantum number that multiplies the modular exponentiation and holds the output. It should be at least the size of $\\lceil \\log(modulus) \rceil$.
+        p: The power of the exponentiation.
+    """
+    p_arr: QArray[QBit] = QArray(length=p.size)
+    within_apply(
+        lambda: bind(p, p_arr),
+        lambda: repeat(
+            count=p_arr.len,
+            iteration=lambda index: control(
+                p_arr[index],
+                lambda: modular_multiply_constant_inplace(
+                    modulus, (a ** (2**index)) % modulus, x
+                ),
+            ),
+        ),
+    )
+
+
+# Helper Functions
+
+
+def get_bit(number: int, index: int) -> int:
+    """
+    Returns the value (0 or 1) of the bit at the specified index in a non-negative integer number.
+    Index 0 is the least significant bit (LSB).
+    """
+    return (number >> index) & 1
+
+
+# Montgomery representation
+
+
+@qperm
+def modular_to_montgomery_inplace(modulus: CInt, x: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+    Converts a quantum integer |x> into its Montgomery representation modulo modulus in place.
+    The Montgomery factor is R = 2**n, where n = x.size (the number of qubits in |x>).
+    This function performs the transformation |x> -> |(x * R mod modulus)>.
+    Note:
+    |x> should have values smaller than `modulus`.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x>.
+    The modulus must be odd so that R = 2**n is invertible modulo modulus (gcd(R, modulus) = 1).
+
+    Args:
+        modulus: Classical number modulus
+        x: Quantum number, in-place operand to convert to Montgomery form.
+    """
+    n = x.size
+    mont_factor = 2**n % modulus
+    modular_multiply_constant_inplace(modulus, mont_factor, x)
+
+
+@qperm
+def modular_montgomery_to_standard_inplace(modulus: CInt, x: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+    Converts quantum integer |x> from Montgomery representation to standard form in place modulo modulus.
+    The Montgomery factor is R = 2**n, where n = x.size (the number of qubits in |x>).
+    This function performs the transformation |x> -> |(x * R^-1 mod modulus)>.
+    Note:
+    |x> should have values smaller than `modulus`.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x>.
+    The modulus must be odd so that R = 2**n is invertible modulo modulus (gcd(R, modulus) = 1).
+
+    Args:
+        modulus: Classical number modulus
+        x: Quantum number, in-place operand to convert from Montgomery form.
+    """
+    n = x.size
+    modular_multiply_constant_inplace(modulus, mod_inverse(2**n % modulus, modulus), x)
+
+
+# Modular Inverse
+
+
+@qperm
+def modular_inverse_inplace(modulus: CInt, v: QNum, m: Output[QArray[QBit]]) -> None:
+    """
+    [Qmod Classiq-library function]
+    Computes the modular inverse of a quantum number |v> modulo modulus in place, using the Kaliski algorithm.
+    Performs the transformation |v> -> |(v^-1 mod modulus)>.
+
+    Based on: https://arxiv.org/pdf/2302.06639 Chapter 5
+
+    Note:
+    |v> should have values smaller than `modulus`.
+    If |v> = 0, the output will be 0 (although 0 does not have an inverse modulo `modulus`).
+    The modulus should be prime OR at least gcd(v, modulus) = 1.
+    The modulus must be a constant odd integer.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |v>.
+    The ancilla qubits m are provided as Output, will be allocated to length 2*n.
+
+    Args:
+        modulus: Classical number modulus
+        v: Quantum number, in-place operand to compute the modular inverse.
+        m: Output quantum array (QArray[QBit]) allocated to length 2*n (n = v.size) and used as ancilla during the algorithm.
+    """
+    n = v.size
+    allocate(2 * n, m)
+    # Convert v to Montgomery form
+    modular_to_montgomery_inplace(modulus, v)
+    u: QNum = QNum("u", n)
+    r: QNum = QNum("r", n + 1)
+    s: QNum = QNum("s", n + 1)
+    a: QBit = QBit()
+    b: QBit = QBit()
+    f: QBit = QBit()
+    allocate(u)
+    allocate(r)
+    allocate(s)
+    allocate(a)
+    allocate(b)
+    allocate(f)
+    # Initialize
+    u ^= modulus
+    r ^= 0
+    s ^= 1
+    f ^= 1
+    # Main loop (2*n iterations)
+    repeat(2 * n, lambda i: kaliski_iteration(modulus, i, v, m, u, r, s, a, b, f))
+    # Finalization steps
+    modular_rsub_inplace(2 * modulus, modulus, r)
+    multiswap(v, r)
+    m_num: QNum = QNum("m_num", 2 * n)
+    bind(m, m_num)
+    control(
+        m_num == 1,
+        stmt_block=lambda: [inplace_xor(1, s), inplace_xor(modulus, u)],  # type: ignore[arg-type]
+        else_block=lambda: [inplace_xor(1, u), inplace_xor(modulus, s)],  # type: ignore[arg-type]
+    )
+    bind(m_num, m)
+    modular_montgomery_to_standard_inplace(modulus, v)
+    # Free variables
+    free(u)
+    free(r)
+    free(s)
+    free(a)
+    free(b)
+    free(f)
+
+
+@qperm
+def kaliski_iteration(
+    modulus: CInt,
+    i: CInt,
+    v: QNum,
+    m: QArray[QBit],
+    u: QNum,
+    r: QNum,
+    s: QNum,
+    a: QBit,
+    b: QBit,
+    f: QBit,
+) -> None:
+    """
+    Single iteration of the Kaliski modular inverse algorithm main loop.
+    Based on: https://arxiv.org/pdf/2302.06639 Figure 15
+
+    Note:
+    Assumes the global inversion constraints (odd modulus, 1 < modulus < 2**n).
+    Called with 0 <= v < modulus; per-iteration ancilla bit is m[i].
+
+    Args:
+        modulus: Classical number modulus (CInt)
+        i: Loop iteration index.
+        v: The QNum to invert (quantum number, will be mutated).
+        m: Quantum array of ancilla qubits (QArray[QBit]).
+        u: QNum (quantum number, auxiliary for algorithm).
+        r: QNum (quantum number, auxiliary).
+        s: QNum (quantum number, auxiliary).
+        a: QBit (ancilla qubit)
+        b: QBit (ancilla qubit)
+        f: QBit (ancilla qubit)
+    """
+    # Step 1: Update f, m[i]
+    control((v == 0) & f, lambda: X(m[i]))
+    f ^= m[i]
+    # Step 2: Update a, b, m[i]
+    control(f & ((u % 2) == 0), lambda: X(a))
+    control(f & ~a & ((v % 2) == 0), lambda: X(m[i]))
+    b ^= a
+    b ^= m[i]
+    # Step 3: Update a, m[i]
+    control(
+        (u > v) & f & ~b,
+        lambda: (
+            X(a),
+            X(m[i]),
+        ),
+    )
+    # Step 4: Update u, v, r, s
+    control(
+        a,
+        lambda: (
+            multiswap(u, v),
+            multiswap(r, s),
+        ),
+    )
+    # Step 5: Update u, v, r, s
+    control(f & ~b, lambda: (inplace_add(-u, v), inplace_add(r, s)))
+    # Step 6: Update a, b, v, r
+    b ^= m[i]
+    b ^= a
+    control(f, lambda: invert(lambda: cyclic_shift_left(v)))
+    modular_double_inplace(2 * modulus, r)
+    larger_than_modulus = QBit()
+    allocate(larger_than_modulus)
+    larger_than_modulus ^= r > modulus
+    control(larger_than_modulus, lambda: inplace_add(-modulus, r))
+    control(((r % 2) == 1), lambda: X(larger_than_modulus))
+    free(larger_than_modulus)
+    control(
+        a,
+        lambda: (
+            multiswap(u, v),
+            multiswap(r, s),
+        ),
+    )
+    control(((s % 2) == 0), lambda: X(a))
+
+
+@qperm
+def modular_rsub_inplace(modulus: CInt, a: CInt, x: QNum) -> None:
+    """
+    [Qmod Classiq-library function]
+    Performs the in-place modular right-subtraction |x> -> |(a - x mod modulus)>.
+    Note:
+    |x> should have values smaller than `modulus`.
+    The modulus should satisfy 1 < modulus < 2**n, where n is the size of |x>.
+    The classical constant `a` should be in the range 0 <= a < modulus.
+
+    Args:
+        modulus: Classical number modulus
+        a: Classical constant to subtract from
+        x: Quantum number, in-place operand to perform the modular right-subtraction.
+    """
+    modular_negate_inplace(modulus, x)
+    modular_add_constant_inplace(modulus, a, x)

classiq/open_library/functions/qft_space_arithmetics.py

@@ -0,0 +1,81 @@
+from classiq.open_library.functions.qft_functions import qft_no_swap
+from classiq.qmod.builtins.classical_functions import qft_const_adder_phase
+from classiq.qmod.builtins.functions.allocation import free
+from classiq.qmod.builtins.functions.standard_gates import PHASE, X
+from classiq.qmod.builtins.operations import (
+    allocate,
+    control,
+    invert,
+    repeat,
+    skip_control,
+    within_apply,
+)
+from classiq.qmod.cparam import CInt
+from classiq.qmod.qfunc import qfunc, qperm
+from classiq.qmod.qmod_variable import QArray, QBit
+
+
+@qfunc
+def _check_msb(ref: CInt, x: QArray[QBit], aux: QBit) -> None:
+    within_apply(
+        lambda: invert(lambda: qft_no_swap(x)),
+        lambda: control(x[0] == ref, lambda: X(aux)),
+    )
+
+
+@qfunc
+def qft_space_add_const(value: CInt, phi_b: QArray[QBit]) -> None:
+    """
+    [Qmod Classiq-library function]
+
+    Adds a constant to a quantum number (in the Fourier space) using the Quantum Fourier Transform (QFT) Adder algorithm.
+    Assuming that the input `phi_b` has `n` qubits, the result will be $\\phi_b+=value \\mod 2^n$.
+
+    To perform the full algorithm, use:
+    within_apply(lambda: QFT(phi_b), qft_space_add_const(value, phi_b))
+
+    Args:
+        value: The constant to add to the quantum number.
+        phi_b: The quantum number (at the aft space) to which the constant is added.
+
+    """
+    repeat(
+        count=phi_b.len,
+        iteration=lambda index: PHASE(
+            theta=qft_const_adder_phase(
+                index, value, phi_b.len  # type:ignore[arg-type]
+            ),
+            target=phi_b[index],
+        ),
+    )
+
+
+@qperm(disable_perm_check=True)
+def modular_add_qft_space(n: CInt, a: CInt, phi_b: QArray[QBit]) -> None:
+    """
+    [Qmod Classiq-library function]
+
+    Adds a constant `a` to a quantum number `phi_b` modulo the constant `n`.
+    The quantum number `phi_b` is assumed to be in the QFT space.
+
+    Args:
+        n: The modulo number.
+        a: The constant to add to the quantum number.
+        phi_b: The quantum number to which the constant is added.
+
+    """
+    aux = QBit()
+
+    allocate(aux)
+    qft_space_add_const(a, phi_b),
+    skip_control(
+        lambda: (
+            invert(lambda: qft_space_add_const(n, phi_b)),
+            _check_msb(1, phi_b, aux),
+            control(aux, lambda: qft_space_add_const(n, phi_b)),
+        )
+    )
+    invert(lambda: qft_space_add_const(a, phi_b))
+    skip_control(lambda: _check_msb(0, phi_b, aux))
+    qft_space_add_const(a, phi_b)
+    free(aux)