"""
\********************************************************************************
* Copyright (c) 2023 the Qrisp authors
*
* This program and the accompanying materials are made available under the
* terms of the Eclipse Public License 2.0 which is available at
* http://www.eclipse.org/legal/epl-2.0.
*
* This Source Code may also be made available under the following Secondary
* Licenses when the conditions for such availability set forth in the Eclipse
* Public License, v. 2.0 are satisfied: GNU General Public License, version 2
* with the GNU Classpath Exception which is
* available at https://www.gnu.org/software/classpath/license.html.
*
* SPDX-License-Identifier: EPL-2.0 OR GPL-2.0 WITH Classpath-exception-2.0
********************************************************************************/
"""
import copy
import numpy as np
from sympy.core.expr import Expr
from sympy import lambdify
def adaptive_substitution(expr, subs_dic, precision=10):
if isinstance(expr, Expr):
return float(expr.evalf(precision, subs_dic))
elif isinstance(expr, (float, int)):
# TO-DO
return float(expr)
else:
# TO-DO
return expr
# Class that describes an operation which can be performed on a quantum computer
# Example would be an X gate or a measurement
[docs]
class Operation:
"""
This class describes operations like quantum gates, measurements or classical logic
gates. Operation objects do not carry information about which Qubit/Clbits they are
applied to. This can be found in the Instruction class, which is a combination of an
Operation object together with its operands.
Operation objects consist of five basic attributes:
* ``.name`` : A string identifying the operation
* ``.num_qubits`` : An integer specifying the amount qubits on which to operate
* ``.num_clbits`` : An integer specifying the amount of classical bits on which to
operate.
* ``.params`` : A list of floats specifying the parameters of the Operation
* ``.definition`` : A :ref:`QuantumCircuit`. For synthesized (i.e. non-elementary)
operations, this QuantumCircuit specifies the operation.
Operation objects together with their Operands can be appended to QuantumCircuits by
using the :meth:`append <qrisp.QuantumCircuit.append>` method.
QuantumCircuits can be turned into Operations by using the
:meth:`to_gate <qrisp.QuantumCircuit.to_gate>` method.
Examples
--------
We create a QuantumCircuit and append a couple of operations
>>> from qrisp import QuantumCircuit, XGate, CXGate, PGate
>>> qc = QuantumCircuit(2)
>>> qc.append(XGate(), 0)
>>> qc.append(CXGate(), [0,1])
>>> qc.append(PGate(0.5), 1)
>>> synthed_op = qc.to_op()
>>> qc.append(synthed_op, qc.qubits)
"""
# If only given the operation init_op (which can be portable) a copied instance of
# this operation will be returned
def __init__(
self,
name=None,
num_qubits=0,
num_clbits=0,
definition=None,
params=[],
init_op=None,
):
if init_op is not None:
name = init_op.name
num_qubits = init_op.num_qubits
num_clbits = init_op.num_clbits
params = init_op.params
definition = init_op.definition
elif not isinstance(name, str):
raise Exception("Tried to create a Operation with name of type({type(name)} (required is str)")
# Name of the operation - this is how the backend behind the interface will
# identify the operation
self.name = name
# Amount of qubits
self.num_qubits = num_qubits
# Amount of classical bits
self.num_clbits = num_clbits
# List of parameters (also available behind the interface)
self.params = []
# If a definition circuit is given, this means we are supposed to create a
# non-elementary operation
if definition is not None:
# Copy circuit in order to prevent modification
# self.definition = QuantumCircuit(init_qc = definition)
self.definition = definition
self.abstract_params = set(definition.abstract_params)
else:
self.definition = None
self.abstract_params = set()
# Find abstract parameters (ie. sympy expressions and log them)
for par in params:
if isinstance(par, (float, int, np.floating, np.int32)):
pass
elif isinstance(par, Expr):
if len(par.free_symbols):
self.abstract_params = self.abstract_params.union(par.free_symbols)
else:
par = float(par)
else:
raise Exception(
f"Tried to create operation with parameters of type {type(par)}"
)
self.params.append(par)
# These attributes store some information for the uncomputation algorithm
# Qfree basically means that the unitary is a permutation matrix
# (up to local phase shifts). Permeability means that this gate commutes with
# the z operator on a given qubit
self.is_qfree = None
self.permeability = {i: None for i in range(self.num_qubits)}
[docs]
def copy(self):
"""
Returns a copy of the Operation object.
Returns
-------
Operation
The copied operation.
"""
res = copy.copy(self)
if self.definition:
copied_definition = self.definition.copy()
else:
copied_definition = None
res.definition = copied_definition
return res
# Method to get the unitary matrix of the operation
# The parameter decimals has no influence on what is calculated
# Rounding is usefull here because the floating point errors
# sometimes make it hard to read the unitary
[docs]
def get_unitary(self, decimals=-1):
"""
Returns the unitary matrix (if applicable) of the Operation as a numpy array.
Parameters
----------
decimals : int, optional
Amount of decimals to return. By default, the full precision is returned.
Raises
------
Exception
Could not calculate the unitary.
Returns
-------
numpy.ndarray
The unitary matrix of the Operation.
Examples
--------
>>> from qrisp import CPGate
>>> import numpy as np
>>> CPGate(np.pi/2).get_unitary(decimals = 3)
array([[1.+0.j, 0.+0.j, 0.+0.j, 0.+0.j],
[0.+0.j, 1.+0.j, 0.+0.j, 0.+0.j],
[0.+0.j, 0.+0.j, 1.+0.j, 0.+0.j],
[0.+0.j, 0.+0.j, 0.+0.j, 0.+1.j]], dtype=complex64)
"""
if self.name == "barrier":
from qrisp.simulator.unitary_management import np_dtype
return np.eye(2**self.num_qubits, dtype = np_dtype)
# Check if the unitary is already available
if hasattr(self, "unitary"):
if decimals == -1:
return self.unitary
else:
return np.round(self.unitary, decimals)
# If we are dealing with a non-elementary gate, calculate the unitary from
# the definition circuit
else:
if not isinstance(self.definition, type(None)):
self.unitary = self.definition.get_unitary()
return self.get_unitary()
# If no definition circuit is known, raise an error.
# Note that the get_unitary methods of more specific gate families specified
# by the inheritors of this class
else:
raise Exception("Unitary of operation " + self.name + " not defined.")
# Method to return the inverse of the given operation. Again, the methods of more
# specific gate families are specified by the inheritors of this class
[docs]
def inverse(self):
"""
Returns the inverse of this Operation (if applicable).
Raises
------
Exception
Tried to invert non-unitary operation.
Returns
-------
Operation
The daggered operation.
Examples
--------
We invert a phase gate and inspect it's parameters
>>> from qrisp import PGate
>>> phase_gate = PGate(0.8)
>>> phase_gate.inverse().params
[-0.8]
"""
# Check if the instruction contains classical bits => operation is not
# invertible
if self.num_clbits:
raise Exception("Tried to invert non-unitary operation")
# Check if a definition is available and invert it, if so
if self.definition is not None:
inverse_circ = self.definition.inverse()
if self.name[-3:] == "_dg":
res = inverse_circ.to_op(name=self.name[:-3])
else:
res = inverse_circ.to_op(name=self.name + "_dg")
elif self.name == "qb_alloc":
from qrisp.circuit import QubitDealloc
res = QubitDealloc()
elif self.name == "qb_dealloc":
from qrisp.circuit import QubitAlloc
res = QubitAlloc()
elif self.name == "barrier":
res = self.copy()
# Otherwise raise an error
else:
raise Exception("Don't know how to invert Operation " + self.name)
res.is_qfree = self.is_qfree
res.permeability = dict(self.permeability)
return res
# Method to create a controlled gate
[docs]
def control(self, num_ctrl_qubits=1, ctrl_state=-1, method=None):
"""
Returns the controlled version of this Operation (if applicable).
Parameters
----------
num_ctrl_qubits : int, optional
The amount of control qubits. The default is 1.
ctrl_state : int or str, optional
The state on which to activate the basis gate. The default is "1111...".
method : str, optional
The method for synthesizing the required multi-controlled X gates.
Available are ``gray`` and ``gray_pt`` and ``auto``.
Note that "gray_pt" introduces an extra phase (which needs to be uncomputed)
but is more resource efficient. ``auto`` will be transformed into a more
efficient ancilla supported version at compile time if used in a
QuantumSession. The default is ``gray``.
Raises
------
AttributeError
Tried to control non-unitary operation.
Returns
-------
Operation
The controlled operation.
Examples
--------
We control a parametrized X Rotation.
>>> from qrisp import QuantumCircuit, RXGate
>>> mcrx_gate = RXGate(0.5).control(3)
>>> qc = QuantumCircuit(4)
>>> qc.append(mcrx_gate, qc.qubits)
>>> print(qc)
::
qb_4: ─────■─────
│
qb_5: ─────■─────
│
qb_6: ─────■─────
┌────┴────┐
qb_7: ┤ Rx(0.5) ├
└─────────┘
"""
if method is None:
method = "auto"
if self.num_clbits != 0:
raise AttributeError("Tried to control non-unitary operation")
res_num_ctrl_qubits = num_ctrl_qubits
if isinstance(self, PTControlledOperation):
res_num_ctrl_qubits += len(self.controls)
# Check if the method is phase tolerant
if (method.find("pt") != -1 or method.find("gidney") != -1) and res_num_ctrl_qubits != 1:
return PTControlledOperation(
self, num_ctrl_qubits, ctrl_state=ctrl_state, method=method
)
else:
return ControlledOperation(
self, num_ctrl_qubits, ctrl_state=ctrl_state, method=method
)
# TO-DO implement more robust hashing method
def __hash__(self):
return hash(hash(self.name) + hash(tuple(self.params)))
def is_permeable(self, indices):
from qrisp.uncomputation import is_permeable
return is_permeable(self, indices)
[docs]
def bind_parameters(self, subs_dic):
"""
Binds abstract parameters to specified values.
Parameters
----------
subs_dic : dict
A dictionary containing the parameters as keys.
Returns
-------
res : Operation
The Operation with bound parameters.
Examples
--------
We create a phase gate with an abstract parameter and bind it to a specified
value.
>>> from qrisp import PGate
>>> from sympy import Symbol
>>> phi = Symbol("phi")
>>> abstract_p_gate = PGate(phi)
>>> abstract_p_gate.params
[phi]
>>> bound_p_gate = abstract_p_gate.bind_parameters({phi : 1.5})
>>> bound_p_gate.params
[1.5]
"""
new_params = []
repl_args = [subs_dic[symb] for symb in self.abstract_params]
if not hasattr(self, "lambdified_params"):
self.lambdified_params = []
args = list(self.abstract_params)
for par in self.params:
self.lambdified_params.append(lambdify(args, par, modules = "numpy"))
for l_par in self.lambdified_params:
new_params.append(l_par(*repl_args))
res = self.copy()
res.params = new_params
if res.definition is not None:
res.definition = res.definition.bind_parameters(subs_dic)
return res
def c_if(self, num_control=1, ctrl_state=-1):
if ctrl_state == -1:
ctrl_state = 2**num_control - 1
return ClControlledOperation(self, num_control, ctrl_state)
# Class to describe 1-Qubit gates as unitaries of the form U(theta, phi, lam) =
# = exp(-1j*phi/2*sigma_z) exp(-1j*theta/2*sigma_y) exp(-1j*lam/2*sigma_z)
# See https://qiskit.org/documentation/stubs/qiskit.circuit.library.U3Gate.html
# for more information
class U3Gate(Operation):
def __init__(self, theta, phi, lam, name="u3", global_phase=0):
# Initialize Operation instance
super().__init__(
name=name,
num_qubits=1,
num_clbits=0,
definition=None,
params=[theta, phi, lam],
)
if isinstance(global_phase, Expr):
if len(global_phase.free_symbols):
self.abstract_params = self.abstract_params.union(global_phase.free_symbols)
else:
global_phase = float(global_phase)
self.global_phase = global_phase
# Set parameters
self.theta = self.params[0]
self.phi = self.params[1]
self.lam = self.params[2]
if self.name in ["rx", "ry", "rz", "p"]:
self.params = [sum(self.params)]
if self.name in ["rz", "p"]:
self.permeability[0] = True
self.is_qfree = True
else:
self.permeability[0] = False
self.is_qfree = False
elif self.name in ["h"]:
self.params = []
self.permeability[0] = False
self.is_qfree = False
# Specify inversion method
def inverse(self):
# The inverse of a product of matrices if the reverted product of the inverses,
# i.e. (A*B*C)^(-1) = C^-1 * B^-1 * A^-1
if self.name[-3:] == "_dg":
new_name = self.name[:-3]
else:
new_name = self.name + "_dg"
# For exponentials of hermitian matrices, the inverse is the hermitean
# conjugate, which implies that we simply have to negate the parameters
# theta, phi, lambda
res = U3Gate(
-self.theta,
-self.lam,
-self.phi,
name=new_name,
global_phase=-self.global_phase,
)
if self.name == "u3":
res.name = "u3"
# These are special gates that require only a single parameter
if self.name in ["rx", "ry", "rz", "p", "h", "gphase"]:
res.name = self.name
res.params = [-par for par in self.params]
if self.name in ["s", "t", "s_dg", "t_dg"]:
res.params = []
if res.is_qfree is not None:
res.is_qfree = bool(self.is_qfree)
res.permeability = dict(self.permeability)
res.is_qfree = bool(self.is_qfree)
return res
# Method to calculate the unitary matrix
def get_unitary(self, decimals=-1):
if hasattr(self, "unitary"):
if decimals != -1:
return np.around(self.unitary, decimals)
else:
return self.unitary
else:
from qrisp.simulator.unitary_management import u3matrix
# Generate unitary
self.unitary = u3matrix(self.theta, self.phi, self.lam, self.global_phase, use_sympy = bool(len(self.abstract_params)))
return self.get_unitary(decimals)
def bind_parameters(self, subs_dic):
new_params = []
repl_args = [subs_dic[symb] for symb in self.abstract_params]
if not hasattr(self, "lambdified_params"):
self.lambdified_params = []
args = list(self.abstract_params)
for par in [self.theta, self.phi, self.lam, self.global_phase]:
self.lambdified_params.append(lambdify(args, par, modules = "numpy"))
for l_par in self.lambdified_params:
new_params.append(l_par(*repl_args))
return U3Gate(new_params[0], new_params[1], new_params[2], self.name, new_params[3])
# return U3Gate(
# adaptive_substitution(self.theta, subs_dic),
# adaptive_substitution(self.phi, subs_dic),
# adaptive_substitution(self.lam, subs_dic),
# self.name,
# adaptive_substitution(self.global_phase, subs_dic),
# )
pi = float(np.pi)
# This class is special for pauli gates. In principle, we could also use the U3Gate
# class, but this could lead to unneccessary floating point errors
class PauliGate(U3Gate):
def __init__(self, name):
from qrisp.simulator.unitary_management import pauli_x, pauli_y, pauli_z
if name == "x":
super().__init__(pi, 0, pi)
self.unitary = pauli_x
self.is_qfree = True
self.permeability[0] = False
elif name == "y":
super().__init__(pi, pi / 2, pi / 2)
self.unitary = pauli_y
self.is_qfree = True
self.permeability[0] = False
elif name == "z":
super().__init__(0, 0, pi)
self.unitary = pauli_z
self.is_qfree = True
self.permeability[0] = True
else:
raise Exception("Gate " + name + " is not a Pauli gate")
self.name = name
self.params = []
def inverse(self):
return PauliGate(self.name)
def __repr__(self):
return self.name
# This class describes phase tolerant controlled operations
# Phase tolerant means that the unitary can take the form
# [D_0, 0, 0, 0]
# [0, D_1,0, 0]
# [0, 0, D_2, 0]
# [0, 0, 0, U]
#Where U is the operation to be controlled and D_i are diagonal operators
#For a regular controlled gate, all D_i have to be identity matrices.
#Phase tolerant controlled operations are usally more efficient than their controlled equivalent.
#In many cases, they can replace the controlled version without changing the semantics,
#because the phases introduced by the D_i are uncomputed at some later point.
class PTControlledOperation(Operation):
def __init__(self, base_operation, num_ctrl_qubits=1, ctrl_state=-1, method="auto"):
# Object which describes the method. Can be a string lr a callable function
self.method = method
# QuantumOperation object which describes which action is controlled,
# i.e. for a CX gate, this would be an X gate
self.base_operation = base_operation
# List of control qubits
self.controls = list(range(num_ctrl_qubits))
# The control state - can be specified either as a string or as an int
if ctrl_state == -1:
self.ctrl_state = num_ctrl_qubits * "1"
elif isinstance(ctrl_state, int):
from qrisp.misc import bin_rep
self.ctrl_state = bin_rep(ctrl_state, num_ctrl_qubits)[::-1]
else:
self.ctrl_state = str(ctrl_state)
# Check if control state specification matches control qubit amount
if len(self.ctrl_state) != num_ctrl_qubits:
raise Exception(
"Specified control state incompatible with given control qubit amount"
)
# Now we generate the definition circuit. Note that most of the generation
# process also applies to the ControlledOperation class, however this class has
# more specific method (.inverse() for instance),
# which is why we make this distinction
# If the base operation is a ControlledOperation, the result
# PTControlledOperation can be generated by applying the phase tolerant control
# algorithm to the base gate of the controlled operation in question
elif isinstance(base_operation, PTControlledOperation):
if method == "gray":
method = self.method
self.__init__(
base_operation.base_operation,
num_ctrl_qubits=num_ctrl_qubits + len(base_operation.controls),
ctrl_state=self.ctrl_state + base_operation.ctrl_state,
method=method,
)
return
if base_operation.name == "gphase":
from qrisp import PGate, QuantumCircuit, bin_rep
definition_circ = QuantumCircuit(num_ctrl_qubits + 1)
if num_ctrl_qubits > 1:
temp_gate = PGate(base_operation.params[0]).control(
num_ctrl_qubits - 1, ctrl_state=self.ctrl_state[1:], method=method
)
else:
temp_gate = PGate(base_operation.params[0])
if self.ctrl_state[0] == "0":
definition_circ.x(-2)
definition_circ.append(temp_gate, definition_circ.qubits[:num_ctrl_qubits])
if self.ctrl_state[0] == "0":
definition_circ.x(-2)
elif self.base_operation.name == "gray_phase_gate":
raise
from qrisp.circuit import multi_controlled_gray_circ
definition_circ, target_phases = multi_controlled_gray_circ(
self.base_operation, num_ctrl_qubits, self.ctrl_state
)
self.target_phases = target_phases
self.phase_tolerant = False
elif self.base_operation.name == "swap":
from qrisp.core import fredkin_qc
definition_circ = fredkin_qc(num_ctrl_qubits, ctrl_state, method)
# For the case of a pauli gate with a single control, we insert an extra case
# since here is no need for any advanced algorithm here and we do not need
# to apply the phase tolerant naming convention
elif (
isinstance(base_operation, PauliGate)
and num_ctrl_qubits == 1
and self.ctrl_state == "1"
):
if base_operation.name == "x":
super().__init__(
name="cx", num_qubits=2, num_clbits=0, params=[]
)
self.permeability = {0: True, 1: False}
elif base_operation.name == "y":
super().__init__(
name="cy", num_qubits=2, num_clbits=0, params=[]
)
self.permeability = {0: True, 1: False}
elif base_operation.name == "z":
super().__init__(
name="cz", num_qubits=2, num_clbits=0, params=[]
)
self.permeability = {0: True, 1: True}
self.is_qfree = True
return
# In the case of an u3gate as a base operation, we first check, if the method
# object is callable, we generate the definition circuit from this function.
# Otherwise we call the algorithm for generating controlled u3 gates
elif isinstance(base_operation, U3Gate):
if callable(method):
definition_circ = method(self, num_ctrl_qubits, self.ctrl_state)
else:
from qrisp.circuit.controlled_operations import multi_controlled_u3_circ
definition_circ = multi_controlled_u3_circ(
base_operation,
num_ctrl_qubits,
ctrl_state=self.ctrl_state,
method=method,
)
# If the base operation has a definition, we can simply apply the phase tolerant
# control algorithm to every gate contained in this defintion.
# This is done in the function multi_controlled_circuit
elif base_operation.definition:
from qrisp.circuit import multi_controlled_circuit
definition_circ = multi_controlled_circuit(
base_operation.definition,
num_ctrl_qubits,
ctrl_state=self.ctrl_state,
method=method,
)
# Raise exception if no possility of synthesizing a controlled game is known
else:
raise Exception(
"Control method for gate " + base_operation.name + " not implemented"
)
# Generate gate name
if num_ctrl_qubits == 1:
name_prefix = "ptc"
else:
name_prefix = "pt" + str(num_ctrl_qubits) + "c"
Operation.__init__(
self,
name=name_prefix + base_operation.name,
num_qubits=base_operation.num_qubits + num_ctrl_qubits,
num_clbits=0,
definition=definition_circ,
params=base_operation.params,
)
for i in range(self.num_qubits):
if i < num_ctrl_qubits:
self.permeability[i] = True
else:
self.permeability[i] = base_operation.permeability[i - num_ctrl_qubits]
if base_operation.is_qfree is not None:
self.is_qfree = bool(base_operation.is_qfree)
def inverse(self):
# Generate inverse operation by applying the constructor
# to the inverse base_operation to get the meta-data right
res = PTControlledOperation(
self.base_operation.inverse(),
len(self.controls),
ctrl_state=self.ctrl_state,
method=self.method,
)
# In order to make sure the definition circuit is also correct,
# we invert the circuit if its invertible
try:
res.definition = self.definition.inverse()
except AttributeError:
pass
if res.is_qfree is not None:
res.is_qfree = bool(self.is_qfree)
res.permeability = dict(self.permeability)
return res
def bind_parameters(self, subs_dic):
from copy import copy
res = copy(self)
if not isinstance(self.definition, type(None)):
res.definition = self.definition.bind_parameters(subs_dic)
res.base_operation = self.base_operation.bind_parameters(subs_dic)
res.params = res.base_operation.params
res.abstract_params = set(self.base_operation.params) - set(subs_dic.keys())
return res
# Class to describe controlled operation
# Very simlar to phase tolerant operations but with a more specifix naming
# convention, inversion algorithm and unitary generation algorithm
class ControlledOperation(PTControlledOperation):
def __init__(self, base_operation, num_ctrl_qubits=1, ctrl_state=-1, method="gray"):
super().__init__(
base_operation,
num_ctrl_qubits=num_ctrl_qubits,
ctrl_state=ctrl_state,
method=method,
)
if num_ctrl_qubits == 1:
name_prefix = "c"
else:
name_prefix = str(len(self.controls)) + "c"
self.name = name_prefix + base_operation.name
def inverse(self):
return ControlledOperation(
self.base_operation.inverse(),
len(self.controls),
ctrl_state=self.ctrl_state,
method=self.method,
)
# For generating the unitary we don't have to generate the unitary by multiplying
# the gates of the definition circuit.
# Instead, we can use that the unitary of a controlled operation is
# the identity matrix apart from the bottom right block matrix, where we find
# the unitary of the base operation.
def get_unitary(self, decimals=-1):
if hasattr(self, "unitary"):
if decimals != -1:
return np.around(self.unitary, decimals)
else:
return self.unitary
else:
from qrisp.simulator.unitary_management import controlled_unitary
self.unitary = controlled_unitary(self)
return self.get_unitary(decimals)
class ClControlledOperation(Operation):
def __init__(self, base_op, num_control = 1, ctrl_state = -1):
if ctrl_state == -1:
ctrl_state = num_control*"1"
if isinstance(ctrl_state, int):
from qrisp.misc import bin_rep
ctrl_state = bin_rep(ctrl_state, num_control)[::-1]
self.base_op = base_op
self.num_control = num_control
self.ctrl_state = ctrl_state
Operation.__init__(self,
name = "c_if_" + base_op.name,
num_qubits = base_op.num_qubits,
num_clbits = base_op.num_clbits + num_control,
params = list(base_op.params))
self.permeability = dict(base_op.permeability)