Arithmetic circuit library: multipliers (Qiskit#6470)

* Implement initial draft of ripple-carry adder circuit * Switch to ancilla register for ancilla qubits * Remove padding for simpler implementation * Add small fixes from review * Add initial unit tests Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Switch carry out from ancilla to usual register * Update __init__ docstring with raises * Add suggestions from review; Clean-up Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Add additional test for ripple-carry adder * Implement initial QFT adder circuit Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Fix QFT adder carry out; Switch to non-modular as default * Expand adder tests with QFT adder * Rename test_adder.py -> test_adders.py * Include additional QFT adder documentation * Fix typos * Implement ripple-carry adder circuit (#11) * Implement initial draft of ripple-carry adder circuit * Switch to ancilla register for ancilla qubits * Remove padding for simpler implementation * Add small fixes from review * Add initial unit tests Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Switch carry out from ancilla to usual register * Update __init__ docstring with raises * Add suggestions from review; Clean-up Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Add additional test for ripple-carry adder Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Fix documentation math Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Add non-modular test as default Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Remove duplicate file after merge * Implement QFT adder (#22) * Implement initial draft of ripple-carry adder circuit * Switch to ancilla register for ancilla qubits * Remove padding for simpler implementation * Add small fixes from review * Add initial unit tests Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Switch carry out from ancilla to usual register * Update __init__ docstring with raises * Add suggestions from review; Clean-up Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Add additional test for ripple-carry adder * Implement initial QFT adder circuit Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Fix QFT adder carry out; Switch to non-modular as default * Expand adder tests with QFT adder * Rename test_adder.py -> test_adders.py * Include additional QFT adder documentation * Fix typos * Fix documentation math Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Add non-modular test as default Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Remove duplicate file after merge Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * Implemented classical adder in QFT paper (#13) * empty * empty * Implemented classical adder in QFT paper * Implemented classical adder in QFT paper * fix classicaladd imports * fix missing imports * Added the tests part for ClassicalAdd * move cin to bottom * Updated the documentation and test parts for ClassicalAdder * Updated init files * Deleted adder folder from the git repository * Deleted test_classicaladd.py * Deleted old classicaladder file * Deleted __init__.py file from repository * Updated classical_adder file Co-authored-by: Cryoris <jules.gacon@googlemail.com> Co-authored-by: Julien Gacon <gaconju@gmail.com> * unify tests * inject base class ``Adder`` * add reno * fix lint * Implement initial classical multiplier * Implement test for multipliers * Update docstring * fix trailing whitespace in line beginning * rework docstrings for less duplication * Change instructions to gates in adder circuits * Generalize multiplier for any adder * Expand multiplier test with all adders * Fix pylint errors * Implement review suggestions Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> * several updates * modular RippleCarryAdder * test that ancillas are uncomputed * fix ClassicalAdder ancilla uncompute * replace all to_instruction by to_gate * lint * rename classical to plain, make plain modular * Implement initial QFT multiplier draft * Update QFT multiplier docstring * Create common multiplier class * Renamed PlainAdder to VBERippleCarryAdder (#29) Co-authored-by: Cryoris <jules.gacon@googlemail.com> * rename modular->fixed point and fix lint * add author names to adders * add comment on how the test works * black * Fix common multiplier errors * Update classical multiplier default adder name * Fix QFT multiplier registers * Implement tests for multipliers * Fix lint * Add black formatter changes * Switch to adder object instead of type * Update classical multiplier drawing * Switch to PhaseGate in QFT multiplier * Rename multipliers * Implement modulo in QFT multiplier * Update to half adder in cumulative multiplier * Make number of output qubits optional * Implement modular cumulative multiplier * Add test for modular multiplication * Fix wrong parameter name * Expand common multiplier class with checks; Add more tests for multipliers * Add black and lint changes * Add releasenote * Attempt cyclic import fix * Attempt to fix cyclic import once more * Update qiskit/circuit/library/arithmetic/multipliers/hrs_cumulative_multiplier.py Co-authored-by: Julien Gacon <gaconju@gmail.com> * Update qiskit/circuit/library/arithmetic/multipliers/hrs_cumulative_multiplier.py Co-authored-by: Julien Gacon <gaconju@gmail.com> * Update qiskit/circuit/library/arithmetic/multipliers/hrs_cumulative_multiplier.py Co-authored-by: Julien Gacon <gaconju@gmail.com> * Update qiskit/circuit/library/arithmetic/multipliers/multiplier.py Co-authored-by: Julien Gacon <gaconju@gmail.com> * Update qiskit/circuit/library/arithmetic/multipliers/hrs_cumulative_multiplier.py Co-authored-by: Julien Gacon <gaconju@gmail.com> * Add additional test cases Co-authored-by: Julien Gacon <gaconju@gmail.com> * Include circuit diagram for QFT multiplier Co-authored-by: Julien Gacon <gaconju@gmail.com> * Update docstring as suggested Co-authored-by: Julien Gacon <gaconju@gmail.com> * Move num_result_qubits logic to parent class Co-authored-by: Julien Gacon <gaconju@gmail.com> * Switch to NotImplementedError for cumulative multiplier Co-authored-by: Julien Gacon <gaconju@gmail.com> * Add black changes * Update Multiplier class docstring Co-authored-by: Julien Gacon <gaconju@gmail.com> * Add lint changes * Update qiskit/circuit/library/arithmetic/multipliers/multiplier.py Co-authored-by: Julien Gacon <gaconju@gmail.com> * Update qiskit/circuit/library/arithmetic/multipliers/multiplier.py Co-authored-by: Julien Gacon <gaconju@gmail.com> * Add black and pylint fixes Co-authored-by: Julien Gacon <jules.gacon@googlemail.com> Co-authored-by: Julien Gacon <gaconju@gmail.com> Co-authored-by: Manjula <64902594+ManjulaGandhi@users.noreply.github.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
chriseclectic · Jun 1, 2021 · 57d6f9d · 57d6f9d
1 parent a2193a5
commit 57d6f9d
Show file tree

Hide file tree

Showing 8 changed files with 535 additions and 0 deletions.
diff --git a/qiskit/circuit/library/__init__.py b/qiskit/circuit/library/__init__.py
@@ -151,6 +151,15 @@
    VBERippleCarryAdder
    WeightedAdder
 
+Multipliers
++++++++++++
+
+.. autosummary::
+   :toctree: ../stubs/
+
+   HRSCumulativeMultiplier
+   RGQFTMultiplier
+
 Comparators
 +++++++++++
 
@@ -362,6 +371,8 @@
     CDKMRippleCarryAdder,
     DraperQFTAdder,
     PiecewiseChebyshev,
+    HRSCumulativeMultiplier,
+    RGQFTMultiplier,
 )
 
 from .n_local import (

diff --git a/qiskit/circuit/library/arithmetic/__init__.py b/qiskit/circuit/library/arithmetic/__init__.py
@@ -23,3 +23,4 @@
 from .linear_amplitude_function import LinearAmplitudeFunction
 from .adders import VBERippleCarryAdder, CDKMRippleCarryAdder, DraperQFTAdder
 from .piecewise_chebyshev import PiecewiseChebyshev
+from .multipliers import HRSCumulativeMultiplier, RGQFTMultiplier
diff --git a/qiskit/circuit/library/arithmetic/multipliers/__init__.py b/qiskit/circuit/library/arithmetic/multipliers/__init__.py
@@ -0,0 +1,16 @@
+# This code is part of Qiskit.
+#
+# (C) Copyright IBM 2017, 2021.
+#
+# This code is licensed under the Apache License, Version 2.0. You may
+# obtain a copy of this license in the LICENSE.txt file in the root directory
+# of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
+#
+# Any modifications or derivative works of this code must retain this
+# copyright notice, and modified files need to carry a notice indicating
+# that they have been altered from the originals.
+
+"""The multiplier circuit library."""
+
+from .hrs_cumulative_multiplier import HRSCumulativeMultiplier
+from .rg_qft_multiplier import RGQFTMultiplier
diff --git a/qiskit/circuit/library/arithmetic/multipliers/hrs_cumulative_multiplier.py b/qiskit/circuit/library/arithmetic/multipliers/hrs_cumulative_multiplier.py
@@ -0,0 +1,134 @@
+# This code is part of Qiskit.
+#
+# (C) Copyright IBM 2017, 2021.
+#
+# This code is licensed under the Apache License, Version 2.0. You may
+# obtain a copy of this license in the LICENSE.txt file in the root directory
+# of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
+#
+# Any modifications or derivative works of this code must retain this
+# copyright notice, and modified files need to carry a notice indicating
+# that they have been altered from the originals.
+
+"""Compute the product of two qubit registers using classical multiplication approach."""
+
+from typing import Optional
+from qiskit.circuit import QuantumRegister, AncillaRegister, QuantumCircuit
+
+from .multiplier import Multiplier
+
+
+class HRSCumulativeMultiplier(Multiplier):
+    r"""A multiplication circuit to store product of two input registers out-of-place.
+
+    Circuit uses the approach from [1]. As an example, a multiplier circuit that
+    performs a non-modular multiplication on two 3-qubit sized registers with
+    the default adder is as follows (where ``Adder`` denotes the
+    ``CDKMRippleCarryAdder``):
+
+    .. parsed-literal::
+
+          a_0: ────■─────────────────────────
+                   │
+          a_1: ────┼─────────■───────────────
+                   │         │
+          a_2: ────┼─────────┼─────────■─────
+               ┌───┴────┐┌───┴────┐┌───┴────┐
+          b_0: ┤0       ├┤0       ├┤0       ├
+               │        ││        ││        │
+          b_1: ┤1       ├┤1       ├┤1       ├
+               │        ││        ││        │
+          b_2: ┤2       ├┤2       ├┤2       ├
+               │        ││        ││        │
+        out_0: ┤3       ├┤        ├┤        ├
+               │        ││        ││        │
+        out_1: ┤4       ├┤3       ├┤        ├
+               │  Adder ││  Adder ││  Adder │
+        out_2: ┤5       ├┤4       ├┤3       ├
+               │        ││        ││        │
+        out_3: ┤6       ├┤5       ├┤4       ├
+               │        ││        ││        │
+        out_4: ┤        ├┤6       ├┤5       ├
+               │        ││        ││        │
+        out_5: ┤        ├┤        ├┤6       ├
+               │        ││        ││        │
+        aux_0: ┤7       ├┤7       ├┤7       ├
+               └────────┘└────────┘└────────┘
+
+    Multiplication in this circuit is implemented in a classical approach by performing
+    a series of shifted additions using one of the input registers while the qubits
+    from the other input register act as control qubits for the adders.
+
+    **References:**
+
+    [1] Häner et al., Optimizing Quantum Circuits for Arithmetic, 2018.
+    `arXiv:1805.12445 <https://arxiv.org/pdf/1805.12445.pdf>`_
+
+    """
+
+    def __init__(
+        self,
+        num_state_qubits: int,
+        num_result_qubits: Optional[int] = None,
+        adder: Optional[QuantumCircuit] = None,
+        name: str = "HRSCumulativeMultiplier",
+    ) -> None:
+        r"""
+        Args:
+            num_state_qubits: The number of qubits in either input register for
+                state :math:`|a\rangle` or :math:`|b\rangle`. The two input
+                registers must have the same number of qubits.
+            num_result_qubits: The number of result qubits to limit the output to.
+                If number of result qubits is :math:`n`, multiplication modulo :math:`2^n` is performed
+                to limit the output to the specified number of qubits. Default
+                value is ``2 * num_state_qubits`` to represent any possible
+                result from the multiplication of the two inputs.
+            adder: Half adder circuit to be used for performing multiplication. The
+                CDKMRippleCarryAdder is used as default if no adder is provided.
+            name: The name of the circuit object.
+        Raises:
+            NotImplementedError: If ``num_result_qubits`` is not default and a custom adder is provided.
+        """
+        super().__init__(num_state_qubits, num_result_qubits, name=name)
+
+        if self.num_result_qubits != 2 * num_state_qubits and adder is not None:
+            raise NotImplementedError("Only default adder is supported for modular multiplication.")
+
+        # define the registers
+        qr_a = QuantumRegister(num_state_qubits, name="a")
+        qr_b = QuantumRegister(num_state_qubits, name="b")
+        qr_out = QuantumRegister(self.num_result_qubits, name="out")
+        self.add_register(qr_a, qr_b, qr_out)
+
+        # prepare adder as controlled gate
+        if adder is None:
+            from qiskit.circuit.library.arithmetic.adders import CDKMRippleCarryAdder
+
+            adder = CDKMRippleCarryAdder(num_state_qubits, kind="half")
+
+        # get the number of helper qubits needed
+        num_helper_qubits = adder.num_ancillas
+
+        # add helper qubits if required
+        if num_helper_qubits > 0:
+            qr_h = AncillaRegister(num_helper_qubits, name="helper")  # helper/ancilla qubits
+            self.add_register(qr_h)
+
+        # build multiplication circuit
+        for i in range(num_state_qubits):
+            excess_qubits = max(0, num_state_qubits + i + 1 - self.num_result_qubits)
+            if excess_qubits == 0:
+                num_adder_qubits = num_state_qubits
+                adder_for_current_step = adder
+            else:
+                num_adder_qubits = num_state_qubits - excess_qubits + 1
+                adder_for_current_step = CDKMRippleCarryAdder(num_adder_qubits, kind="fixed")
+            controlled_adder = adder_for_current_step.to_gate().control(1)
+            qr_list = (
+                [qr_a[i]]
+                + qr_b[:num_adder_qubits]
+                + qr_out[i : num_state_qubits + i + 1 - excess_qubits]
+            )
+            if num_helper_qubits > 0:
+                qr_list.extend(qr_h[:])
+            self.append(controlled_adder, qr_list)
diff --git a/qiskit/circuit/library/arithmetic/multipliers/multiplier.py b/qiskit/circuit/library/arithmetic/multipliers/multiplier.py
@@ -0,0 +1,101 @@
+# This code is part of Qiskit.
+#
+# (C) Copyright IBM 2017, 2021.
+#
+# This code is licensed under the Apache License, Version 2.0. You may
+# obtain a copy of this license in the LICENSE.txt file in the root directory
+# of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
+#
+# Any modifications or derivative works of this code must retain this
+# copyright notice, and modified files need to carry a notice indicating
+# that they have been altered from the originals.
+
+"""Compute the product of two equally sized qubit registers."""
+
+from typing import Optional
+
+from qiskit.circuit import QuantumCircuit
+
+
+class Multiplier(QuantumCircuit):
+    r"""Compute the product of two equally sized qubit registers into a new register.
+
+    For two input registers :math:`|a\rangle_n`, :math:`|b\rangle_n` with :math:`n` qubits each
+    and an output register with :math:`2n` qubits, a multiplier performs the following operation
+
+    .. math::
+
+        |a\rangle_n |b\rangle_n |0\rangle_{t} \mapsto |a\rangle_n |b\rangle_n |a \cdot b\rangle_t
+
+    where :math:`t` is the number of bits used to represent the result. To completely store the result
+    of the multiplication without overflow we need :math:`t = 2n` bits.
+
+    The quantum register :math:`|a\rangle_n` (analogously :math:`|b\rangle_n` and
+    output register)
+
+    .. math::
+
+        |a\rangle_n = |a_0\rangle \otimes \cdots \otimes |a_{n - 1}\rangle,
+
+    for :math:`a_i \in \{0, 1\}`, is associated with the integer value
+
+    .. math::
+
+        a = 2^{0}a_{0} + 2^{1}a_{1} + \cdots + 2^{n - 1}a_{n - 1}.
+
+    """
+
+    def __init__(
+        self,
+        num_state_qubits: int,
+        num_result_qubits: Optional[int] = None,
+        name: str = "Multiplier",
+    ) -> None:
+        """
+        Args:
+            num_state_qubits: The number of qubits in each of the input registers.
+            num_result_qubits: The number of result qubits to limit the output to.
+                Default value is ``2 * num_state_qubits`` to represent any possible
+                result from the multiplication of the two inputs.
+            name: The name of the circuit.
+        Raises:
+            ValueError: If ``num_state_qubits`` is smaller than 1.
+            ValueError: If ``num_result_qubits`` is smaller than ``num_state_qubits``.
+            ValueError: If ``num_result_qubits`` is larger than ``2 * num_state_qubits``.
+        """
+        if num_state_qubits < 1:
+            raise ValueError("The number of qubits must be at least 1.")
+
+        if num_result_qubits is None:
+            num_result_qubits = 2 * num_state_qubits
+
+        if num_result_qubits < num_state_qubits:
+            raise ValueError(
+                "Number of result qubits is smaller than number of input state qubits."
+            )
+        if num_result_qubits > 2 * num_state_qubits:
+            raise ValueError(
+                "Number of result qubits is larger than twice the number of input state qubits."
+            )
+
+        super().__init__(name=name)
+        self._num_state_qubits = num_state_qubits
+        self._num_result_qubits = num_result_qubits
+
+    @property
+    def num_state_qubits(self) -> int:
+        """The number of state qubits, i.e. the number of bits in each input register.
+
+        Returns:
+            The number of state qubits.
+        """
+        return self._num_state_qubits
+
+    @property
+    def num_result_qubits(self) -> int:
+        """The number of result qubits to limit the output to.
+
+        Returns:
+            The number of result qubits.
+        """
+        return self._num_result_qubits
diff --git a/qiskit/circuit/library/arithmetic/multipliers/rg_qft_multiplier.py b/qiskit/circuit/library/arithmetic/multipliers/rg_qft_multiplier.py
@@ -0,0 +1,97 @@
+# This code is part of Qiskit.
+#
+# (C) Copyright IBM 2017, 2021.
+#
+# This code is licensed under the Apache License, Version 2.0. You may
+# obtain a copy of this license in the LICENSE.txt file in the root directory
+# of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
+#
+# Any modifications or derivative works of this code must retain this
+# copyright notice, and modified files need to carry a notice indicating
+# that they have been altered from the originals.
+
+"""Compute the product of two qubit registers using QFT."""
+
+from typing import Optional
+import numpy as np
+
+from qiskit.circuit import QuantumRegister
+from qiskit.circuit.library.standard_gates import PhaseGate
+from qiskit.circuit.library.basis_change import QFT
+
+from .multiplier import Multiplier
+
+
+class RGQFTMultiplier(Multiplier):
+    r"""A QFT multiplication circuit to store product of two input registers out-of-place.
+
+    Multiplication in this circuit is implemented using the procedure of Fig. 3 in [1], where
+    weighted sum rotations are implemented as given in Fig. 5 in [1]. QFT is used on the output
+    register and is followed by rotations controlled by input registers. The rotations
+    transform the state into the product of two input registers in QFT base, which is
+    reverted from QFT base using inverse QFT.
+    As an example, a circuit that performs a modular QFT multiplication on two 2-qubit
+    sized input registers with an output register of 2 qubits, is as follows:
+
+    .. parsed-literal::
+
+          a_0: ────────────────────────────────────────■───────■──────■──────■────────────────
+                                                       │       │      │      │
+          a_1: ─────────■───────■───────■───────■──────┼───────┼──────┼──────┼────────────────
+                        │       │       │       │      │       │      │      │
+          b_0: ─────────┼───────┼───────■───────■──────┼───────┼──────■──────■────────────────
+                        │       │       │       │      │       │      │      │
+          b_1: ─────────■───────■───────┼───────┼──────■───────■──────┼──────┼────────────────
+               ┌──────┐ │P(4π)  │       │P(2π)  │      │P(2π)  │      │P(π)  │       ┌───────┐
+        out_0: ┤0     ├─■───────┼───────■───────┼──────■───────┼──────■──────┼───────┤0      ├
+               │  qft │         │P(2π)          │P(π)          │P(π)         │P(π/2) │  iqft │
+        out_1: ┤1     ├─────────■───────────────■──────────────■─────────────■───────┤1      ├
+               └──────┘                                                              └───────┘
+
+    **References:**
+
+    [1] Ruiz-Perez et al., Quantum arithmetic with the Quantum Fourier Transform, 2017.
+    `arXiv:1411.5949 <https://arxiv.org/pdf/1411.5949.pdf>`_
+
+    """
+
+    def __init__(
+        self,
+        num_state_qubits: int,
+        num_result_qubits: Optional[int] = None,
+        name: str = "RGQFTMultiplier",
+    ) -> None:
+        r"""
+        Args:
+            num_state_qubits: The number of qubits in either input register for
+                state :math:`|a\rangle` or :math:`|b\rangle`. The two input
+                registers must have the same number of qubits.
+            num_result_qubits: The number of result qubits to limit the output to.
+                If number of result qubits is :math:`n`, multiplication modulo :math:`2^n` is performed
+                to limit the output to the specified number of qubits. Default
+                value is ``2 * num_state_qubits`` to represent any possible
+                result from the multiplication of the two inputs.
+            name: The name of the circuit object.
+
+        """
+        super().__init__(num_state_qubits, num_result_qubits, name=name)
+
+        # define the registers
+        qr_a = QuantumRegister(num_state_qubits, name="a")
+        qr_b = QuantumRegister(num_state_qubits, name="b")
+        qr_out = QuantumRegister(self.num_result_qubits, name="out")
+        self.add_register(qr_a, qr_b, qr_out)
+
+        # build multiplication circuit
+        self.append(QFT(self.num_result_qubits, do_swaps=False).to_gate(), qr_out[:])
+
+        for j in range(1, num_state_qubits + 1):
+            for i in range(1, num_state_qubits + 1):
+                for k in range(1, self.num_result_qubits + 1):
+                    lam = (2 * np.pi) / (2 ** (i + j + k - 2 * num_state_qubits))
+                    self.append(
+                        PhaseGate(lam).control(2),
+                        [qr_a[num_state_qubits - j], qr_b[num_state_qubits - i], qr_out[k - 1]],
+                    )
+
+        self.append(QFT(self.num_result_qubits, do_swaps=False).inverse().to_gate(), qr_out[:])