pyqrack-cpu-complex128 1.58.0__py3-none-manylinux_2_35_x86_64.whl

pyqrack/qrack_ace_backend.py

@@ -0,0 +1,1357 @@
+# (C) Daniel Strano and the Qrack contributors 2017-2025. All rights reserved.
+#
+# Use of this source code is governed by an MIT-style license that can be
+# found in the LICENSE file or at https://opensource.org/licenses/MIT.
+import math
+import os
+import random
+import sys
+import time
+
+from .qrack_simulator import QrackSimulator
+from .pauli import Pauli
+
+
+_IS_QISKIT_AVAILABLE = True
+try:
+    from qiskit.circuit.quantumcircuit import QuantumCircuit
+    from qiskit.compiler import transpile
+    from qiskit.quantum_info.operators.symplectic.clifford import Clifford
+except ImportError:
+    _IS_QISKIT_AVAILABLE = False
+
+_IS_QISKIT_AER_AVAILABLE = True
+try:
+    from qiskit_aer.noise import NoiseModel, depolarizing_error
+except ImportError:
+    _IS_QISKIT_AER_AVAILABLE = False
+
+
+# Initial stub and concept produced through conversation with Elara
+# (the custom OpenAI GPT)
+class LHVQubit:
+    def __init__(self, toClone=None):
+        # Initial state in "Bloch vector" terms, defaults to |0⟩
+        if toClone:
+            self.bloch = toClone.bloch.copy()
+        else:
+            self.reset()
+
+    def reset(self):
+        self.bloch = [0.0, 0.0, 1.0]
+
+    def h(self):
+        # Hadamard: rotate around Y-axis then X-axis (simplified for LHV)
+        x, y, z = self.bloch
+        self.bloch = [(x + z) / math.sqrt(2), y, (z - x) / math.sqrt(2)]
+
+    def x(self):
+        x, y, z = self.bloch
+        self.bloch = [x, y, -z]
+
+    def y(self):
+        x, y, z = self.bloch
+        self.bloch = [-x, y, z]
+
+    def z(self):
+        x, y, z = self.bloch
+        self.bloch = [x, -y, z]
+
+    def rx(self, theta):
+        # Rotate Bloch vector around X-axis by angle theta
+        x, y, z = self.bloch
+        cos_theta = math.cos(theta)
+        sin_theta = math.sin(theta)
+        new_y = cos_theta * y - sin_theta * z
+        new_z = sin_theta * y + cos_theta * z
+        self.bloch = [x, new_y, new_z]
+
+    def ry(self, theta):
+        # Rotate Bloch vector around Y-axis by angle theta
+        x, y, z = self.bloch
+        cos_theta = math.cos(theta)
+        sin_theta = math.sin(theta)
+        new_x = cos_theta * x + sin_theta * z
+        new_z = -sin_theta * x + cos_theta * z
+        self.bloch = [new_x, y, new_z]
+
+    def rz(self, theta):
+        # Rotate Bloch vector around Z-axis by angle theta (in radians)
+        x, y, z = self.bloch
+        cos_theta = math.cos(theta)
+        sin_theta = math.sin(theta)
+        new_x = cos_theta * x - sin_theta * y
+        new_y = sin_theta * x + cos_theta * y
+        self.bloch = [new_x, new_y, z]
+
+    def s(self):
+        self.rz(math.pi / 2)
+
+    def adjs(self):
+        self.rz(-math.pi / 2)
+
+    def t(self):
+        self.rz(math.pi / 4)
+
+    def adjt(self):
+        self.rz(-math.pi / 4)
+
+    def u(self, theta, phi, lam):
+        # Apply general single-qubit unitary gate
+        self.rz(lam)
+        self.ry(theta)
+        self.rz(phi)
+
+    def prob(self, basis=Pauli.PauliZ):
+        """Sample a classical outcome from the current 'quantum' state"""
+        if basis == Pauli.PauliZ:
+            prob_1 = (1 - self.bloch[2]) / 2
+        elif basis == Pauli.PauliX:
+            prob_1 = (1 - self.bloch[0]) / 2
+        elif basis == Pauli.PauliY:
+            prob_1 = (1 - self.bloch[1]) / 2
+        else:
+            raise ValueError(f"Unsupported basis: {basis}")
+        return prob_1
+
+    def m(self):
+        result = random.random() < self.prob()
+        self.reset()
+        if result:
+            self.x()
+        return result
+
+
+# Provided by Elara (the custom OpenAI GPT)
+def _cpauli_lhv(prob, targ, axis, anti, theta=math.pi):
+    """
+    Apply a 'soft' controlled-Pauli gate: rotate target qubit
+    proportionally to control's Z expectation value.
+
+    theta: full rotation angle if control in |1⟩
+    """
+    # Control influence is (1 - ctrl.bloch[2]) / 2 = P(|1⟩)
+    # BUT we avoid collapse by using the expectation value:
+    control_influence = (1 - prob) if anti else prob
+
+    effective_theta = control_influence * theta
+
+    # Apply partial rotation to target qubit:
+    if axis == Pauli.PauliX:
+        targ.rx(effective_theta)
+    elif axis == Pauli.PauliY:
+        targ.ry(effective_theta)
+    elif axis == Pauli.PauliZ:
+        targ.rz(effective_theta)
+
+
+class QrackAceBackend:
+    """A back end for elided quantum error correction
+
+    This back end uses elided repetition code on a nearest-neighbor topology to emulate
+    a utility-scale superconducting chip quantum computer in very little memory.4
+
+    The backend was originally designed assuming an (orbifolded) 2D qubit grid like 2019 Sycamore.
+    However, it quickly became apparent that users can basically design their own connectivity topologies,
+    without breaking the concept. (Not all will work equally well.)
+
+    Consider distributing the different "patches" to different GPUs with self.sim[sim_id].set_device(gpu_id)!
+    (If you have 3+ patches, maybe your discrete GPU can do multiple patches in the time it takes an Intel HD
+    to do one patch worth of work!)
+
+    Attributes:
+        sim(QrackSimulator): Array of simulators corresponding to "patches" between boundary rows.
+        long_range_columns(int): How many ideal rows between QEC boundary rows?
+        is_transpose(bool): Rows are long if False, columns are long if True
+    """
+
+    def __init__(
+        self,
+        qubit_count=1,
+        long_range_columns=5,
+        long_range_rows=2,
+        is_transpose=False,
+        isTensorNetwork=False,
+        isSchmidtDecomposeMulti=False,
+        isSchmidtDecompose=True,
+        isStabilizerHybrid=False,
+        isBinaryDecisionTree=False,
+        isPaged=True,
+        isCpuGpuHybrid=True,
+        isOpenCL=True,
+        isHostPointer=(
+            True if os.environ.get("PYQRACK_HOST_POINTER_DEFAULT_ON") else False
+        ),
+        noise=0,
+        toClone=None,
+    ):
+        if toClone:
+            qubit_count = toClone.num_qubits()
+            long_range_columns = toClone.long_range_columns
+            long_range_rows = toClone.long_range_rows
+            is_transpose = toClone.is_transpose
+        if qubit_count < 0:
+            qubit_count = 0
+        if long_range_columns < 0:
+            long_range_columns = 0
+
+        self._factor_width(qubit_count, is_transpose)
+        self.long_range_columns = long_range_columns
+        self.long_range_rows = long_range_rows
+        self.is_transpose = is_transpose
+
+        fppow = 5
+        if "QRACK_FPPOW" in os.environ:
+            fppow = int(os.environ.get("QRACK_FPPOW"))
+        if fppow < 5:
+            self._epsilon = 2**-9
+        elif fppow > 5:
+            self._epsilon = 2**-51
+        else:
+            self._epsilon = 2**-22
+
+        self._coupling_map = None
+
+        # If there's only one or zero "False" columns or rows,
+        # the entire simulator is connected, anyway.
+        len_col_seq = long_range_columns + 1
+        col_patch_count = (self._row_length + len_col_seq - 1) // len_col_seq
+        if (self._row_length < 3) or ((long_range_columns + 1) >= self._row_length):
+            self._is_col_long_range = [True] * self._row_length
+        else:
+            col_seq = [True] * long_range_columns + [False]
+            self._is_col_long_range = (col_seq * col_patch_count)[: self._row_length]
+            if long_range_columns < self._row_length:
+                self._is_col_long_range[-1] = False
+        len_row_seq = long_range_rows + 1
+        row_patch_count = (self._col_length + len_row_seq - 1) // len_row_seq
+        if (self._col_length < 3) or ((long_range_rows + 1) >= self._col_length):
+            self._is_row_long_range = [True] * self._col_length
+        else:
+            row_seq = [True] * long_range_rows + [False]
+            self._is_row_long_range = (row_seq * row_patch_count)[: self._col_length]
+            if long_range_rows < self._col_length:
+                self._is_row_long_range[-1] = False
+        sim_count = col_patch_count * row_patch_count
+        self._row_offset = 0
+        for r in range(self._col_length):
+            for c in self._is_col_long_range:
+                self._row_offset += 1 if c else 3
+
+        self._qubit_dict = {}
+        sim_counts = [0] * sim_count
+        sim_id = 0
+        tot_qubits = 0
+        for r in self._is_row_long_range:
+            for c in self._is_col_long_range:
+                qubit = [(sim_id, sim_counts[sim_id])]
+                sim_counts[sim_id] += 1
+
+                if (not c) or (not r):
+                    t_sim_id = (sim_id + 1) % sim_count
+                    qubit.append((t_sim_id, sim_counts[t_sim_id]))
+                    sim_counts[t_sim_id] += 1
+
+                    qubit.append(
+                        LHVQubit(
+                            toClone=(
+                                toClone._qubit_dict[tot_qubits][2] if toClone else None
+                            )
+                        )
+                    )
+
+                if (not c) and (not r):
+                    t_sim_id = (sim_id + col_patch_count) % sim_count
+                    qubit.append((t_sim_id, sim_counts[t_sim_id]))
+                    sim_counts[t_sim_id] += 1
+
+                    t_sim_id = (t_sim_id + 1) % sim_count
+                    qubit.append((t_sim_id, sim_counts[t_sim_id]))
+                    sim_counts[t_sim_id] += 1
+
+                if not c:
+                    sim_id = (sim_id + 1) % sim_count
+
+                self._qubit_dict[tot_qubits] = qubit
+                tot_qubits += 1
+
+        self.sim = []
+        for i in range(sim_count):
+            self.sim.append(
+                toClone.sim[i].clone()
+                if toClone
+                else QrackSimulator(
+                    sim_counts[i],
+                    isTensorNetwork=isTensorNetwork,
+                    isSchmidtDecomposeMulti=isSchmidtDecomposeMulti,
+                    isSchmidtDecompose=isSchmidtDecompose,
+                    isStabilizerHybrid=isStabilizerHybrid,
+                    isBinaryDecisionTree=isBinaryDecisionTree,
+                    isPaged=isPaged,
+                    isCpuGpuHybrid=isCpuGpuHybrid,
+                    isOpenCL=isOpenCL,
+                    isHostPointer=isHostPointer,
+                    noise=noise,
+                )
+            )
+
+            # You can still "monkey-patch" this, after the constructor.
+            if "QRACK_QUNIT_SEPARABILITY_THRESHOLD" not in os.environ:
+                self.sim[i].set_sdrp(0.02375)
+
+    def clone(self):
+        return QrackAceBackend(toClone=self)
+
+    def num_qubits(self):
+        return self._row_length * self._col_length
+
+    def get_row_length(self):
+        return self._row_length
+
+    def get_column_length(self):
+        return self._col_length
+
+    def _factor_width(self, width, is_transpose=False):
+        col_len = math.floor(math.sqrt(width))
+        while ((width // col_len) * col_len) != width:
+            col_len -= 1
+        row_len = width // col_len
+
+        self._col_length, self._row_length = (
+            (row_len, col_len) if is_transpose else (col_len, row_len)
+        )
+
+    def _ct_pair_prob(self, q1, q2):
+        p1 = self.sim[q1[0]].prob(q1[1]) if isinstance(q1, tuple) else q1.prob()
+        p2 = self.sim[q2[0]].prob(q2[1]) if isinstance(q2, tuple) else q2.prob()
+
+        if p1 < p2:
+            return p2, q1
+
+        return p1, q2
+
+    def _cz_shadow(self, q1, q2):
+        prob_max, t = self._ct_pair_prob(q1, q2)
+        if prob_max > 0.5:
+            if isinstance(t, tuple):
+                self.sim[t[0]].z(t[1])
+            else:
+                t.z()
+
+    def _qec_x(self, c):
+        if isinstance(c, tuple):
+            self.sim[c[0]].x(c[1])
+        else:
+            c.x()
+
+    def _qec_h(self, t):
+        if isinstance(t, tuple):
+            self.sim[t[0]].h(t[1])
+        else:
+            t.h()
+
+    def _qec_s(self, t):
+        if isinstance(t, tuple):
+            self.sim[t[0]].s(t[1])
+        else:
+            t.s()
+
+    def _qec_adjs(self, t):
+        if isinstance(t, tuple):
+            self.sim[t[0]].adjs(t[1])
+        else:
+            t.adjs()
+
+    def _anti_cz_shadow(self, c, t):
+        self._qec_x(c)
+        self._cz_shadow(c, t)
+        self._qec_x(c)
+
+    def _cx_shadow(self, c, t):
+        self._qec_h(t)
+        self._cz_shadow(c, t)
+        self._qec_h(t)
+
+    def _anti_cx_shadow(self, c, t):
+        self._qec_x(c)
+        self._cx_shadow(c, t)
+        self._qec_x(c)
+
+    def _cy_shadow(self, c, t):
+        self._qec_adjs(t)
+        self._cx_shadow(c, t)
+        self._qec_s(t)
+
+    def _anti_cy_shadow(self, c, t):
+        self._qec_x(c)
+        self._cy_shadow(c, t)
+        self._qec_x(c)
+
+    def _unpack(self, lq):
+        return self._qubit_dict[lq]
+
+    def _get_qb_lhv_indices(self, hq):
+        qb = []
+        if len(hq) < 2:
+            qb = [0]
+            lhv = -1
+        elif len(hq) < 4:
+            qb = [0, 1]
+            lhv = 2
+        else:
+            qb = [0, 1, 3, 4]
+            lhv = 2
+
+        return qb, lhv
+
+    def _correct(self, lq, phase=False):
+        hq = self._unpack(lq)
+
+        if len(hq) == 1:
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        if phase:
+            for q in qb:
+                b = hq[q]
+                self.sim[b[0]].h(b[1])
+            b = hq[lhv]
+            b.h()
+
+        if len(hq) == 5:
+            # RMS
+            p = [
+                self.sim[hq[0][0]].prob(hq[0][1]),
+                self.sim[hq[1][0]].prob(hq[1][1]),
+                hq[2].prob(),
+                self.sim[hq[3][0]].prob(hq[3][1]),
+                self.sim[hq[4][0]].prob(hq[4][1]),
+            ]
+            # Balancing suggestion from Elara (the custom OpenAI GPT)
+            prms = math.sqrt(
+                (p[0] ** 2 + p[1] ** 2 + 3 * (p[2] ** 2) + p[3] ** 2 + p[4] ** 2) / 7
+            )
+            qrms = math.sqrt(
+                (
+                    (1 - p[0]) ** 2
+                    + (1 - p[1]) ** 2
+                    + 3 * ((1 - p[2]) ** 2)
+                    + (1 - p[3]) ** 2
+                    + (1 - p[4]) ** 2
+                )
+                / 7
+            )
+            result = ((prms + (1 - qrms)) / 2) >= 0.5
+            syndrome = (
+                [1 - p[0], 1 - p[1], 1 - p[2], 1 - p[3], 1 - p[4]]
+                if result
+                else [p[0], p[1], p[2], p[3], p[4]]
+            )
+            for q in range(5):
+                if syndrome[q] > (0.5 + self._epsilon):
+                    if q == 2:
+                        hq[q].x()
+                    else:
+                        self.sim[hq[q][0]].x(hq[q][1])
+        else:
+            # RMS
+            p = [
+                self.sim[hq[0][0]].prob(hq[0][1]),
+                self.sim[hq[1][0]].prob(hq[1][1]),
+                hq[2].prob(),
+            ]
+            # Balancing suggestion from Elara (the custom OpenAI GPT)
+            prms = math.sqrt((p[0] ** 2 + p[1] ** 2 + p[2] ** 2) / 3)
+            qrms = math.sqrt(((1 - p[0]) ** 2 + (1 - p[1]) ** 2 + (1 - p[2]) ** 2) / 3)
+            result = ((prms + (1 - qrms)) / 2) >= 0.5
+            syndrome = [1 - p[0], 1 - p[1], 1 - p[2]] if result else [p[0], p[1], p[2]]
+            for q in range(3):
+                if syndrome[q] > (0.5 + self._epsilon):
+                    if q == 2:
+                        hq[q].x()
+                    else:
+                        self.sim[hq[q][0]].x(hq[q][1])
+
+        if phase:
+            for q in qb:
+                b = hq[q]
+                self.sim[b[0]].h(b[1])
+            b = hq[lhv]
+            b.h()
+
+    def u(self, lq, th, ph, lm):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].u(b[1], th, ph, lm)
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].u(b[1], th, ph, lm)
+
+        b = hq[lhv]
+        b.u(th, ph, lm)
+
+        self._correct(lq, False)
+        self._correct(lq, True)
+
+    def r(self, p, th, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].r(p, th, b[1])
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].r(p, th, b[1])
+
+        b = hq[lhv]
+        if p == Pauli.PauliX:
+            b.rx(th)
+        elif p == Pauli.PauliY:
+            b.ry(th)
+        elif p == Pauli.PauliZ:
+            b.rz(th)
+
+        if p != Pauli.PauliZ:
+            self._correct(lq, False)
+        if p != Pauli.PauliX:
+            self._correct(lq, True)
+
+    def h(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].h(b[1])
+            return
+
+        self._correct(lq)
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].h(b[1])
+
+        b = hq[lhv]
+        b.h()
+
+        self._correct(lq)
+
+    def s(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].s(b[1])
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].s(b[1])
+
+        b = hq[lhv]
+        b.s()
+
+    def adjs(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].adjs(b[1])
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].adjs(b[1])
+
+        b = hq[lhv]
+        b.adjs()
+
+    def x(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].x(b[1])
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].x(b[1])
+
+        b = hq[lhv]
+        b.x()
+
+    def y(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].y(b[1])
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].y(b[1])
+
+        b = hq[lhv]
+        b.y()
+
+    def z(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].z(b[1])
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].z(b[1])
+
+        b = hq[lhv]
+        b.z()
+
+    def t(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].t(b[1])
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].t(b[1])
+
+        b = hq[lhv]
+        b.t()
+
+    def adjt(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            self.sim[b[0]].adjt(b[1])
+            return
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            self.sim[b[0]].adjt(b[1])
+
+        b = hq[lhv]
+        b.adjt()
+
+    def _get_gate(self, pauli, anti, sim_id):
+        gate = None
+        shadow = None
+        if pauli == Pauli.PauliX:
+            gate = self.sim[sim_id].macx if anti else self.sim[sim_id].mcx
+            shadow = self._anti_cx_shadow if anti else self._cx_shadow
+        elif pauli == Pauli.PauliY:
+            gate = self.sim[sim_id].macy if anti else self.sim[sim_id].mcy
+            shadow = self._anti_cy_shadow if anti else self._cy_shadow
+        elif pauli == Pauli.PauliZ:
+            gate = self.sim[sim_id].macz if anti else self.sim[sim_id].mcz
+            shadow = self._anti_cz_shadow if anti else self._cz_shadow
+        else:
+            raise RuntimeError(
+                "QrackAceBackend._get_gate() should never return identity!"
+            )
+
+        return gate, shadow
+
+    def _get_connected(self, i, is_row):
+        long_range = self._is_row_long_range if is_row else self._is_col_long_range
+        length = self._col_length if is_row else self._row_length
+
+        connected = [i]
+        c = (i - 1) % length
+        while long_range[c] and (len(connected) < length):
+            connected.append(c)
+            c = (c - 1) % length
+        if len(connected) < length:
+            connected.append(c)
+        boundary = len(connected)
+        c = (i + 1) % length
+        while long_range[c] and (len(connected) < length):
+            connected.append(c)
+            c = (c + 1) % length
+        if len(connected) < length:
+            connected.append(c)
+
+        return connected, boundary
+
+    def _apply_coupling(self, pauli, anti, qb1, lhv1, hq1, qb2, lhv2, hq2, lq1_lr):
+        for q1 in qb1:
+            if q1 == lhv1:
+                continue
+            b1 = hq1[q1]
+            gate_fn, shadow_fn = self._get_gate(pauli, anti, b1[0])
+            for q2 in qb2:
+                if q2 == lhv2:
+                    continue
+                b2 = hq2[q2]
+                if b1[0] == b2[0]:
+                    gate_fn([b1[1]], b2[1])
+                elif (
+                    lq1_lr
+                    or (b1[1] == b2[1])
+                    or ((len(qb1) == 2) and (b1[1] == (b2[1] & 1)))
+                ):
+                    shadow_fn(b1, b2)
+
+    def _cpauli(self, lq1, lq2, anti, pauli):
+        lq1_row = lq1 // self._row_length
+        lq1_col = lq1 % self._row_length
+        lq2_row = lq2 // self._row_length
+        lq2_col = lq2 % self._row_length
+
+        hq1 = self._unpack(lq1)
+        hq2 = self._unpack(lq2)
+
+        lq1_lr = len(hq1) == 1
+        lq2_lr = len(hq2) == 1
+
+        self._correct(lq1)
+
+        qb1, lhv1 = self._get_qb_lhv_indices(hq1)
+        qb2, lhv2 = self._get_qb_lhv_indices(hq2)
+        # Apply cross coupling on hardware qubits first
+        self._apply_coupling(pauli, anti, qb1, lhv1, hq1, qb2, lhv2, hq2, lq1_lr)
+        # Apply coupling to the local-hidden-variable target
+        if lhv2 >= 0:
+            _cpauli_lhv(
+                hq1[lhv1].prob() if lhv1 >= 0 else self.sim[hq1[0][0]].prob(hq1[0][1]),
+                hq2[lhv2],
+                pauli,
+                anti,
+            )
+
+        self._correct(lq1, True)
+        if pauli != Pauli.PauliZ:
+            self._correct(lq2, False)
+        if pauli != Pauli.PauliX:
+            self._correct(lq2, True)
+
+    def cx(self, lq1, lq2):
+        self._cpauli(lq1, lq2, False, Pauli.PauliX)
+
+    def cy(self, lq1, lq2):
+        self._cpauli(lq1, lq2, False, Pauli.PauliY)
+
+    def cz(self, lq1, lq2):
+        self._cpauli(lq1, lq2, False, Pauli.PauliZ)
+
+    def acx(self, lq1, lq2):
+        self._cpauli(lq1, lq2, True, Pauli.PauliX)
+
+    def acy(self, lq1, lq2):
+        self._cpauli(lq1, lq2, True, Pauli.PauliY)
+
+    def acz(self, lq1, lq2):
+        self._cpauli(lq1, lq2, True, Pauli.PauliZ)
+
+    def mcx(self, lq1, lq2):
+        if len(lq1) > 1:
+            raise RuntimeError(
+                "QrackAceBackend.mcx() is provided for syntax convenience and only supports 1 control qubit!"
+            )
+        self._cpauli(lq1[0], lq2, False, Pauli.PauliX)
+
+    def mcy(self, lq1, lq2):
+        if len(lq1) > 1:
+            raise RuntimeError(
+                "QrackAceBackend.mcy() is provided for syntax convenience and only supports 1 control qubit!"
+            )
+        self._cpauli(lq1[0], lq2, False, Pauli.PauliY)
+
+    def mcz(self, lq1, lq2):
+        if len(lq1) > 1:
+            raise RuntimeError(
+                "QrackAceBackend.mcz() is provided for syntax convenience and only supports 1 control qubit!"
+            )
+        self._cpauli(lq1[0], lq2, False, Pauli.PauliZ)
+
+    def macx(self, lq1, lq2):
+        if len(lq1) > 1:
+            raise RuntimeError(
+                "QrackAceBackend.macx() is provided for syntax convenience and only supports 1 control qubit!"
+            )
+        self._cpauli(lq1[0], lq2, True, Pauli.PauliX)
+
+    def macy(self, lq1, lq2):
+        if len(lq1) > 1:
+            raise RuntimeError(
+                "QrackAceBackend.macy() is provided for syntax convenience and only supports 1 control qubit!"
+            )
+        self._cpauli(lq1[0], lq2, True, Pauli.PauliY)
+
+    def macz(self, lq1, lq2):
+        if len(lq1) > 1:
+            raise RuntimeError(
+                "QrackAceBackend.macz() is provided for syntax convenience and only supports 1 control qubit!"
+            )
+        self._cpauli(lq1[0], lq2, True, Pauli.PauliZ)
+
+    def swap(self, lq1, lq2):
+        self.cx(lq1, lq2)
+        self.cx(lq2, lq1)
+        self.cx(lq1, lq2)
+
+    def iswap(self, lq1, lq2):
+        self.swap(lq1, lq2)
+        self.cz(lq1, lq2)
+        self.s(lq1)
+        self.s(lq2)
+
+    def adjiswap(self, lq1, lq2):
+        self.adjs(lq2)
+        self.adjs(lq1)
+        self.cz(lq1, lq2)
+        self.swap(lq1, lq2)
+
+    def prob(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            return self.sim[b[0]].prob(b[1])
+
+        self._correct(lq)
+        if len(hq) == 5:
+            # RMS
+            p = [
+                self.sim[hq[0][0]].prob(hq[0][1]),
+                self.sim[hq[1][0]].prob(hq[1][1]),
+                hq[2].prob(),
+                self.sim[hq[3][0]].prob(hq[3][1]),
+                self.sim[hq[4][0]].prob(hq[4][1]),
+            ]
+            # Balancing suggestion from Elara (the custom OpenAI GPT)
+            prms = math.sqrt(
+                (p[0] ** 2 + p[1] ** 2 + 3 * (p[2] ** 2) + p[3] ** 2 + p[4] ** 2) / 7
+            )
+            qrms = math.sqrt(
+                (
+                    (1 - p[0]) ** 2
+                    + (1 - p[1]) ** 2
+                    + 3 * ((1 - p[2]) ** 2)
+                    + (1 - p[3]) ** 2
+                    + (1 - p[4]) ** 2
+                )
+                / 7
+            )
+        else:
+            # RMS
+            p = [
+                self.sim[hq[0][0]].prob(hq[0][1]),
+                self.sim[hq[1][0]].prob(hq[1][1]),
+                hq[2].prob(),
+            ]
+            # Balancing suggestion from Elara (the custom OpenAI GPT)
+            prms = math.sqrt((p[0] ** 2 + p[1] ** 2 + p[2] ** 2) / 3)
+            qrms = math.sqrt(((1 - p[0]) ** 2 + (1 - p[1]) ** 2 + (1 - p[2]) ** 2) / 3)
+
+        return (prms + (1 - qrms)) / 2
+
+    def m(self, lq):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            return self.sim[b[0]].m(b[1])
+
+        p = self.prob(lq)
+        result = ((p + self._epsilon) >= 1) or (random.random() < p)
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            p = self.sim[b[0]].prob(b[1]) if result else (1 - self.sim[b[0]].prob(b[1]))
+            if p < self._epsilon:
+                if self.sim[b[0]].m(b[1]) != result:
+                    self.sim[b[0]].x(b[1])
+            else:
+                self.sim[b[0]].force_m(b[1], result)
+
+        b = hq[lhv]
+        b.reset()
+        if result:
+            b.x()
+
+        return result
+
+    def force_m(self, lq, result):
+        hq = self._unpack(lq)
+        if len(hq) < 2:
+            b = hq[0]
+            return self.sim[b[0]].force_m(b[1], result)
+
+        self._correct(lq)
+
+        qb, lhv = self._get_qb_lhv_indices(hq)
+
+        for q in qb:
+            b = hq[q]
+            p = self.sim[b[0]].prob(b[1]) if result else (1 - self.sim[b[0]].prob(b[1]))
+            if p < self._epsilon:
+                if self.sim[b[0]].m(b[1]) != result:
+                    self.sim[b[0]].x(b[1])
+            else:
+                self.sim[b[0]].force_m(b[1], result)
+
+        b = hq[1]
+        b.reset()
+        if result:
+            b.x()
+
+        return c
+
+    def m_all(self):
+        # Randomize the order of measurement to amortize error.
+        result = 0
+        rows = list(range(self._col_length))
+        random.shuffle(rows)
+        for lq_row in rows:
+            row_offset = lq_row * self._row_length
+            cols = list(range(self._row_length))
+            random.shuffle(cols)
+            for lq_col in cols:
+                lq = row_offset + lq_col
+                if self.m(lq):
+                    result |= 1 << lq
+
+        return result
+
+    def measure_shots(self, q, s):
+        samples = []
+        for _ in range(s):
+            clone = self.clone()
+            _sample = clone.m_all()
+            sample = 0
+            for i in range(len(q)):
+                if (_sample >> q[i]) & 1:
+                    sample |= 1 << i
+            samples.append(sample)
+
+        return samples
+
+    def _apply_op(self, operation):
+        name = operation.name
+
+        if (name == "id") or (name == "barrier"):
+            # Skip measurement logic
+            return
+
+        conditional = getattr(operation, "conditional", None)
+        if isinstance(conditional, int):
+            conditional_bit_set = (self._classical_register >> conditional) & 1
+            if not conditional_bit_set:
+                return
+        elif conditional is not None:
+            mask = int(conditional.mask, 16)
+            if mask > 0:
+                value = self._classical_memory & mask
+                while (mask & 0x1) == 0:
+                    mask >>= 1
+                    value >>= 1
+                if value != int(conditional.val, 16):
+                    return
+
+        if (name == "u1") or (name == "p"):
+            self._sim.u(operation.qubits[0]._index, 0, 0, float(operation.params[0]))
+        elif name == "u2":
+            self._sim.u(
+                operation.qubits[0]._index,
+                math.pi / 2,
+                float(operation.params[0]),
+                float(operation.params[1]),
+            )
+        elif (name == "u3") or (name == "u"):
+            self._sim.u(
+                operation.qubits[0]._index,
+                float(operation.params[0]),
+                float(operation.params[1]),
+                float(operation.params[2]),
+            )
+        elif name == "r":
+            self._sim.u(
+                operation.qubits[0]._index,
+                float(operation.params[0]),
+                float(operation.params[1]) - math.pi / 2,
+                (-1 * float(operation.params[1])) + math.pi / 2,
+            )
+        elif name == "rx":
+            self._sim.r(
+                Pauli.PauliX, float(operation.params[0]), operation.qubits[0]._index
+            )
+        elif name == "ry":
+            self._sim.r(
+                Pauli.PauliY, float(operation.params[0]), operation.qubits[0]._index
+            )
+        elif name == "rz":
+            self._sim.r(
+                Pauli.PauliZ, float(operation.params[0]), operation.qubits[0]._index
+            )
+        elif name == "h":
+            self._sim.h(operation.qubits[0]._index)
+        elif name == "x":
+            self._sim.x(operation.qubits[0]._index)
+        elif name == "y":
+            self._sim.y(operation.qubits[0]._index)
+        elif name == "z":
+            self._sim.z(operation.qubits[0]._index)
+        elif name == "s":
+            self._sim.s(operation.qubits[0]._index)
+        elif name == "sdg":
+            self._sim.adjs(operation.qubits[0]._index)
+        elif name == "t":
+            self._sim.t(operation.qubits[0]._index)
+        elif name == "tdg":
+            self._sim.adjt(operation.qubits[0]._index)
+        elif name == "cx":
+            self._sim.cx(operation.qubits[0]._index, operation.qubits[1]._index)
+        elif name == "cy":
+            self._sim.cy(operation.qubits[0]._index, operation.qubits[1]._index)
+        elif name == "cz":
+            self._sim.cz(operation.qubits[0]._index, operation.qubits[1]._index)
+        elif name == "dcx":
+            self._sim.mcx(operation.qubits[0]._index, operation.qubits[1]._index)
+            self._sim.mcx(operation.qubits[1]._index, operation.qubits[0]._index)
+        elif name == "swap":
+            self._sim.swap(operation.qubits[0]._index, operation.qubits[1]._index)
+        elif name == "iswap":
+            self._sim.iswap(operation.qubits[0]._index, operation.qubits[1]._index)
+        elif name == "iswap_dg":
+            self._sim.adjiswap(operation.qubits[0]._index, operation.qubits[1]._index)
+        elif name == "reset":
+            qubits = operation.qubits
+            for qubit in qubits:
+                if self._sim.m(qubit._index):
+                    self._sim.x(qubit._index)
+        elif name == "measure":
+            qubits = operation.qubits
+            clbits = operation.clbits
+            cregbits = (
+                operation.register
+                if hasattr(operation, "register")
+                else len(operation.qubits) * [-1]
+            )
+
+            self._sample_qubits += qubits
+            self._sample_clbits += clbits
+            self._sample_cregbits += cregbits
+
+            if not self._sample_measure:
+                for index in range(len(qubits)):
+                    qubit_outcome = self._sim.m(qubits[index]._index)
+
+                    clbit = clbits[index]
+                    clmask = 1 << clbit
+                    self._classical_memory = (self._classical_memory & (~clmask)) | (
+                        qubit_outcome << clbit
+                    )
+
+                    cregbit = cregbits[index]
+                    if cregbit < 0:
+                        cregbit = clbit
+
+                    regbit = 1 << cregbit
+                    self._classical_register = (
+                        self._classical_register & (~regbit)
+                    ) | (qubit_outcome << cregbit)
+
+        elif name == "bfunc":
+            mask = int(operation.mask, 16)
+            relation = operation.relation
+            val = int(operation.val, 16)
+
+            cregbit = operation.register
+            cmembit = operation.memory if hasattr(operation, "memory") else None
+
+            compared = (self._classical_register & mask) - val
+
+            if relation == "==":
+                outcome = compared == 0
+            elif relation == "!=":
+                outcome = compared != 0
+            elif relation == "<":
+                outcome = compared < 0
+            elif relation == "<=":
+                outcome = compared <= 0
+            elif relation == ">":
+                outcome = compared > 0
+            elif relation == ">=":
+                outcome = compared >= 0
+            else:
+                raise QrackError("Invalid boolean function relation.")
+
+            # Store outcome in register and optionally memory slot
+            regbit = 1 << cregbit
+            self._classical_register = (self._classical_register & (~regbit)) | (
+                int(outcome) << cregbit
+            )
+            if cmembit is not None:
+                membit = 1 << cmembit
+                self._classical_memory = (self._classical_memory & (~membit)) | (
+                    int(outcome) << cmembit
+                )
+        else:
+            err_msg = 'QrackAceBackend encountered unrecognized operation "{0}"'
+            raise RuntimeError(err_msg.format(operation))
+
+    def _add_sample_measure(self, sample_qubits, sample_clbits, num_samples):
+        """Generate data samples from current statevector.
+
+        Taken almost straight from the terra source code.
+
+        Args:
+            measure_params (list): List of (qubit, clbit) values for
+                                   measure instructions to sample.
+            num_samples (int): The number of data samples to generate.
+
+        Returns:
+            list: A list of data values in hex format.
+        """
+        # Get unique qubits that are actually measured
+        measure_qubit = [qubit for qubit in sample_qubits]
+        measure_clbit = [clbit for clbit in sample_clbits]
+
+        # Sample and convert to bit-strings
+        if num_samples == 1:
+            sample = self._sim.m_all()
+            result = 0
+            for index in range(len(measure_qubit)):
+                qubit = measure_qubit[index]._index
+                qubit_outcome = (sample >> qubit) & 1
+                result |= qubit_outcome << index
+            measure_results = [result]
+        else:
+            measure_results = self._sim.measure_shots(
+                [q._index for q in measure_qubit], num_samples
+            )
+
+        data = []
+        for sample in measure_results:
+            for index in range(len(measure_qubit)):
+                qubit_outcome = (sample >> index) & 1
+                clbit = measure_clbit[index]._index
+                clmask = 1 << clbit
+                self._classical_memory = (self._classical_memory & (~clmask)) | (
+                    qubit_outcome << clbit
+                )
+
+            data.append(bin(self._classical_memory)[2:].zfill(self.num_qubits()))
+
+        return data
+
+    def run_qiskit_circuit(self, experiment, shots=1):
+        if not _IS_QISKIT_AVAILABLE:
+            raise RuntimeError(
+                "Before trying to run_qiskit_circuit() with QrackAceBackend, you must install Qiskit!"
+            )
+
+        instructions = []
+        if isinstance(experiment, QuantumCircuit):
+            instructions = experiment.data
+        else:
+            raise RuntimeError('Unrecognized "run_input" argument specified for run().')
+
+        self._shots = shots
+        self._sample_qubits = []
+        self._sample_clbits = []
+        self._sample_cregbits = []
+        self._sample_measure = True
+        _data = []
+        shotLoopMax = 1
+
+        is_initializing = True
+        boundary_start = -1
+
+        for opcount in range(len(instructions)):
+            operation = instructions[opcount]
+
+            if operation.name == "id" or operation.name == "barrier":
+                continue
+
+            if is_initializing and (
+                (operation.name == "measure") or (operation.name == "reset")
+            ):
+                continue
+
+            is_initializing = False
+
+            if (operation.name == "measure") or (operation.name == "reset"):
+                if boundary_start == -1:
+                    boundary_start = opcount
+
+            if (boundary_start != -1) and (operation.name != "measure"):
+                shotsPerLoop = 1
+                shotLoopMax = self._shots
+                self._sample_measure = False
+                break
+
+        preamble_memory = 0
+        preamble_register = 0
+        preamble_sim = None
+
+        if self._sample_measure or boundary_start <= 0:
+            boundary_start = 0
+            self._sample_measure = True
+            shotsPerLoop = self._shots
+            shotLoopMax = 1
+        else:
+            boundary_start -= 1
+            if boundary_start > 0:
+                self._sim = self
+                self._classical_memory = 0
+                self._classical_register = 0
+
+                for operation in instructions[:boundary_start]:
+                    self._apply_op(operation)
+
+                preamble_memory = self._classical_memory
+                preamble_register = self._classical_register
+                preamble_sim = self._sim
+
+        for shot in range(shotLoopMax):
+            if preamble_sim is None:
+                self._sim = self
+                self._classical_memory = 0
+                self._classical_register = 0
+            else:
+                self._sim = QrackAceBackend(toClone=preamble_sim)
+                self._classical_memory = preamble_memory
+                self._classical_register = preamble_register
+
+            for operation in instructions[boundary_start:]:
+                self._apply_op(operation)
+
+            if not self._sample_measure and (len(self._sample_qubits) > 0):
+                _data += [bin(self._classical_memory)[2:].zfill(self.num_qubits())]
+                self._sample_qubits = []
+                self._sample_clbits = []
+                self._sample_cregbits = []
+
+        if self._sample_measure and (len(self._sample_qubits) > 0):
+            _data = self._add_sample_measure(
+                self._sample_qubits, self._sample_clbits, self._shots
+            )
+
+        del self._sim
+
+        return _data
+
+    def get_qiskit_basis_gates():
+        return [
+            "id",
+            "u",
+            "u1",
+            "u2",
+            "u3",
+            "r",
+            "rx",
+            "ry",
+            "rz",
+            "h",
+            "x",
+            "y",
+            "z",
+            "s",
+            "sdg",
+            "sx",
+            "sxdg",
+            "p",
+            "t",
+            "tdg",
+            "cx",
+            "cy",
+            "cz",
+            "swap",
+            "iswap",
+            "reset",
+            "measure",
+        ]
+
+    # Mostly written by Dan, but with a little help from Elara (custom OpenAI GPT)
+    def get_logical_coupling_map(self):
+        if self._coupling_map:
+            return self._coupling_map
+
+        coupling_map = set()
+        rows, cols = self._row_length, self._col_length
+
+        # Map each column index to its full list of logical qubit indices
+        def logical_index(row, col):
+            return row * cols + col
+
+        for col in range(cols):
+            connected_cols, _ = self._get_connected(col, False)
+            for row in range(rows):
+                connected_rows, _ = self._get_connected(row, False)
+                a = logical_index(row, col)
+                for c in connected_cols:
+                    for r in connected_rows:
+                        b = logical_index(r, c)
+                        if a != b:
+                            coupling_map.add((a, b))
+
+        self._coupling_map = sorted(coupling_map)
+
+        return self._coupling_map
+
+    # Designed by Dan, and implemented by Elara:
+    def create_noise_model(self, x=0.25, y=0.25):
+        if not _IS_QISKIT_AER_AVAILABLE:
+            raise RuntimeError(
+                "Before trying to run_qiskit_circuit() with QrackAceBackend, you must install Qiskit Aer!"
+            )
+        noise_model = NoiseModel()
+
+        for a, b in self.get_logical_coupling_map():
+            col_a, col_b = a % self._row_length, b % self._row_length
+            row_a, row_b = a // self._row_length, b // self._row_length
+            is_long_a = self._is_col_long_range[col_a]
+            is_long_b = self._is_col_long_range[col_b]
+
+            if is_long_a and is_long_b:
+                continue  # No noise on long-to-long
+
+            if (col_a == col_b) or (row_a == row_b):
+                continue  # No noise for same column
+
+            if is_long_a or is_long_b:
+                y_cy = 1 - (1 - y) ** 2
+                y_swap = 1 - (1 - y) ** 3
+                noise_model.add_quantum_error(depolarizing_error(y, 2), "cx", [a, b])
+                noise_model.add_quantum_error(depolarizing_error(y_cy, 2), "cy", [a, b])
+                noise_model.add_quantum_error(depolarizing_error(y_cy, 2), "cz", [a, b])
+                noise_model.add_quantum_error(
+                    depolarizing_error(y_swap, 2), "swap", [a, b]
+                )
+            else:
+                y_cy = 1 - (1 - y) ** 2
+                y_swap = 1 - (1 - y) ** 3
+                noise_model.add_quantum_error(depolarizing_error(y_cy, 2), "cx", [a, b])
+                noise_model.add_quantum_error(depolarizing_error(y_cy, 2), "cy", [a, b])
+                noise_model.add_quantum_error(depolarizing_error(y_cy, 2), "cz", [a, b])
+                noise_model.add_quantum_error(
+                    depolarizing_error(y_swap, 2), "swap", [a, b]
+                )
+
+        return noise_model