Source code for qrisp.environments.quantum_conditionals

"""
\********************************************************************************
* 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
********************************************************************************/
"""

from qrisp.core.quantum_variable import QuantumVariable
from qrisp.core.gate_application_functions import p, rz, x, z
from qrisp.core.session_merging_tools import merge
from qrisp.environments.quantum_environments import QuantumEnvironment
from qrisp.environments.quantum_inversion import invert
from qrisp.misc import (
    find_calling_line,
    perm_lock,
    perm_unlock,
    redirect_qfunction,
    unlock,
)
from qrisp.circuit import fast_append

def quantum_condition(function):
    def q_env_generator(*args, **kwargs):
        return ConditionEnvironment(function, args, kwargs=kwargs)

    return q_env_generator


# Class to describe if-environments
# For more information on how Environments work, check the QuantumEnvironment class.
# This class is instantiated with a function returning a qubit,
# which indicates wther the condition is met.
# On entering this environment, the function will be evaluated and the content
# (i.e. the quantum operations) will be collected.
# When compiled, this environment will check for parent conditional environments and
# include their truth value into the compilation process. Subsequently, every operation
# that happened inside this environment is turned into it's controlled version.
# Finally, the function evaluating the truth value will be uncomputed and
# the environment is reset for it's next use.
[docs] class ConditionEnvironment(QuantumEnvironment): r""" This class enables the usage of *if*-conditionals as we are used to from classical programming: :: from qrisp import QuantumChar, QuantumFloat, h, multi_measurement q_ch = QuantumChar() qf = QuantumFloat(3, signed = True) h(q_ch[0]) with q_ch == "a": qf += 2 >>> print(multi_measurement([q_ch,qf])) {('a', 2): 0.5, ('b', 0): 0.5} In this code snippet, we first bring the :ref:`QuantumChar` ``q_ch`` into the superposition .. math:: \ket{\text{q_ch}} = \frac{1}{\sqrt{2}} \left( \ket{\text{"a"}} + \ket{\text{"b"}} \right) After that, we enter the ConditionEnvironment, which controls the operations on the condition that ``q_ch`` is in state $\ket{a}$. Finally, we simultaneously measure both :ref:`QuantumVariables <QuantumVariable>`. We see that the incrementation of ``qf`` only occured on the branch, where ``q_ch`` is equal to the character ``"a"``. The resulting quantum state is: .. math:: \ket{\psi} = \frac{1}{\sqrt{2}} \left( \ket{\text{"a"}}\ket{2} + \ket{\text{"b"}}\ket{0} \right) It is furthermore possible to invert the condition truth value or apply phases. For this we acquire the :ref:`QuantumBool` containing the truth value using the *as* statement :: from qrisp import x, p import numpy as np with q_ch == "a" as cond_bool: qf += 2 cond_bool.flip() qf -= 2 p(np.pi/4, cond_bool) >>> qf.qs.statevector() sqrt(2)*(|a>*|4> + exp(I*pi/4)*|b>*|-2>)/2 **Constructing custom conditional environments** Apart from infix notation like the equality operator, Qrisp also provides an interface for creating custom conditonal environments. Parameters ---------- cond_eval_function : function A function which evaluates the truth value. Must return a :ref:`QuantumBool`. Intermediate results do not have to be uncomputed or deleted, as this is automatically performed, when the condition truth value is uncomputed. args : list The arguments on which to evaluate. kwargs : dict, optional A dictionary of keyword arguments for ``cond_eval_function``. The default is {}. Examples -------- We will now demonstrate how a ConditionEnvironment, that evaluates the equality of two :ref:`QuantumVariables <QuantumVariable>` can be constructed: :: from qrisp import QuantumBool, QuantumVariable, x, cx, mcx def quantum_eq(qv_0, qv_1): if qv_0.size != qv_1.size: raise Exception("Tried to evaluate equality condition for QuantumVariables of differing size") temp_qv = QuantumVariable(qv_0.size) cx(qv_0, temp_qv) cx(qv_1, temp_qv) x(temp_qv) res = QuantumBool() mcx(temp_qv, res) return res In this function, we create a temporary variable where we apply CX gates controlled on the inputs onto. The qubits where ``qv_0`` and ``qv_1`` agree, will then be in state $\ket{0}$. After this, we apply regular X gates onto ``temp_qv`` such that the qubits where the inputs agree, are in state $\ket{1}$. Finally, we apply a multi-controlled X gate onto the result to flip the result, if all of qubits of the qubits in ``temp_qv`` are in the $\ket{0}$ state. We inspect the resulting QuantumCircuit >>> from qrisp import QuantumChar >>> q_ch_0 = QuantumChar() >>> q_ch_1 = QuantumChar() >>> res_bool = quantum_eq(q_ch_0, q_ch_1) >>> print(q_ch_0.qs) :: QuantumCircuit: -------------- q_ch_0.0: ──■───────────────────────────────────────────────────────── q_ch_0.1: ──┼────■──────────────────────────────────────────────────── │ │ q_ch_0.2: ──┼────┼────■─────────────────────────────────────────────── │ │ │ q_ch_0.3: ──┼────┼────┼────■────────────────────────────────────────── │ │ │ │ q_ch_0.4: ──┼────┼────┼────┼────■───────────────────────────────────── │ │ │ │ │ q_ch_1.0: ──┼────┼────┼────┼────┼────■──────────────────────────────── │ │ │ │ │ │ q_ch_1.1: ──┼────┼────┼────┼────┼────┼────■─────────────────────────── │ │ │ │ │ │ │ q_ch_1.2: ──┼────┼────┼────┼────┼────┼────┼────■────────────────────── │ │ │ │ │ │ │ │ q_ch_1.3: ──┼────┼────┼────┼────┼────┼────┼────┼────■───────────────── │ │ │ │ │ │ │ │ │ q_ch_1.4: ──┼────┼────┼────┼────┼────┼────┼────┼────┼────■──────────── ┌─┴─┐ │ │ │ │ ┌─┴─┐ │ │ │ │ ┌───┐ temp_qv.0: ┤ X ├──┼────┼────┼────┼──┤ X ├──┼────┼────┼────┼──┤ X ├──■── └───┘┌─┴─┐ │ │ │ └───┘┌─┴─┐ │ │ │ ├───┤ │ temp_qv.1: ─────┤ X ├──┼────┼────┼───────┤ X ├──┼────┼────┼──┤ X ├──■── └───┘┌─┴─┐ │ │ └───┘┌─┴─┐ │ │ ├───┤ │ temp_qv.2: ──────────┤ X ├──┼────┼────────────┤ X ├──┼────┼──┤ X ├──■── └───┘┌─┴─┐ │ └───┘┌─┴─┐ │ ├───┤ │ temp_qv.3: ───────────────┤ X ├──┼─────────────────┤ X ├──┼──┤ X ├──■── └───┘┌─┴─┐ └───┘┌─┴─┐├───┤ │ temp_qv.4: ────────────────────┤ X ├────────────────────┤ X ├┤ X ├──■── └───┘ └───┘└───┘┌─┴─┐ res.0: ───────────────────────────────────────────────────────┤ X ├ └───┘ Live QuantumVariables: --------------------- QuantumChar q_ch_0 QuantumChar q_ch_1 QuantumVariable temp_qv QuantumBool res We can now construct the conditional environment from this function :: from qrisp import ConditionEnvironment, multi_measurement, h #Create some sample arguments on which to evaluate the condition q_bool_0 = QuantumBool() q_bool_1 = QuantumBool() q_bool_2 = QuantumBool() h(q_bool_0) with ConditionEnvironment(cond_eval_function = quantum_eq, args = [q_bool_0, q_bool_1]): q_bool_2.flip() >>> print(multi_measurement([q_bool_0, q_bool_1, q_bool_2])) {(False, False, True): 0.5, (True, False, False): 0.5} This agrees with our expectation, that ``q_bool_2`` is only ``True`` if the other two agree. **The quantum_condition decorator** Creating quantum conditions like this seems a bit unwieldy. For a more convenient solution, we provide the ``quantum_condition`` decorator. This decorator can be applied to a function returning a :ref:`QuantumBool`, which is then returning the corresponding ConditionEnvironment instead. To demonstrate, we construct a "less than" condition for QuantumFloats :: from qrisp import quantum_condition, cx @quantum_condition def less_than(qf_0, qf_1): temp_qf = qf_0 - qf_1 res = QuantumBool() cx(temp_qf.sign(), res) return res qf_0 = QuantumFloat(3) qf_1 = qf_0.duplicate() qf_0[:] = 2 h(qf_1[:2]) res_q_bool = QuantumBool() with less_than(qf_0, qf_1): res_q_bool.flip() >>> print(multi_measurement([qf_0, qf_1, res_q_bool])) {(2, 0, False): 0.25, (2, 1, False): 0.25, (2, 2, False): 0.25, (2, 3, True): 0.25} **Quantum-Loops** An interesting application of conditional environments is the ``qRange`` iterator. Using this construct, we can mimic a loop as we are used from classical computing, where the end of the loop is determined by a quantum state: :: from qrisp import QuantumFloat, qRange, h n = QuantumFloat(3) qf = QuantumFloat(5) n[:] = 4 h(n[0]) n_results = n.get_measurement() for i in qRange(n): qf += i >>> print(qf) {10: 0.5, 15: 0.5} This script calculates the sum of all integers up to a certain threshold. The threshold (n) is a :ref:`QuantumFloat` in superposition, implying the result of the sum is also in a superposition. The expected results can be quickly determined by using Gauß's formula: .. math:: \sum_{i = 0}^n i = \frac{n(n+1)}{2} >>> print("Excpected outcomes:", [n*(n+1)/2 for n in n_results.keys()]) Excpected outcomes: [10.0, 15.0] """ # Constructor of the class def __init__(self, cond_eval_function, args, kwargs={}): # The function which evaluates the condition - should return a QuantumBool def save_cond_eval(*args, **kwargs): with fast_append(0): res = cond_eval_function(*args, **kwargs) return res self.cond_eval_function = save_cond_eval # Save the arguments on which the function should be evaluated self.args = args # Save the keyword arguments self.kwargs = kwargs # Note the QuantumSession of the arguments of the arguments self.arg_qs = merge(args) self.manual_allocation_management = True # Method to enter the environment def __enter__(self): from qrisp.qtypes.quantum_bool import QuantumBool # For more information on why this attribute is neccessary check the comment # on the line containing subcondition_truth_values = [] self.sub_condition_envs = [] self.qbool = QuantumBool(name="cond_env*", qs=self.arg_qs) self.condition_truth_value = self.qbool[0] super().__enter__() merge(self.env_qs, self.arg_qs) return self.qbool def __exit__(self, exception_type, exception_value, traceback): from qrisp.environments import ControlEnvironment, InversionEnvironment, ConjugationEnvironment # We determine the parent condition environment self.parent_cond_env = None QuantumEnvironment.__exit__(self, exception_type, exception_value, traceback) for env in self.env_qs.env_stack[::-1]: if isinstance(env, (ConditionEnvironment, ControlEnvironment)): self.parent_cond_env = env break if not isinstance(env, (InversionEnvironment, ConjugationEnvironment)): if not type(env) == QuantumEnvironment: break # Compile method def compile(self): from qrisp.qtypes.quantum_bool import QuantumBool from qrisp.environments.control_environment import ControlEnvironment, convert_to_custom_control # Create the quantum variable where the condition truth value should be saved # Incase we have a parent environment we create two qubits because # we use the second qubit to compute the toffoli of this one and the parent # environments truth value in order to not have the environment operations # controlled on two qubits if len(self.env_data) == 0: self.qbool.delete() else: # The first step we have to perform is calculating the truth value of # this environments quantum condition. For this we differentiate between # the case that this condition is embedded in another condition or not if self.parent_cond_env is not None: # In the parent case we also need to make sure that the code is executed # if the parent environment is executed. For this a possible approach # would be to control the content on both, the parent and # the chield truth value. However, for each nesting level the gate count # to generate the controlled-controlled-controlled... version # of the gates inside the environment increases exponentially. # Because of this we compute the toffoli of the parent and # child truth value and control the environment gates on this qubit. cond_eval_bool = self.cond_eval_function(*self.args, **self.kwargs) if not isinstance(cond_eval_bool, QuantumBool): raise Exception( "Tried to compile QuantumCondition environment with" "a condition evaluation function not returning a QuantumBool" ) # Create and execute phase tolerant toffoli gate toffoli_qb_list = [ self.parent_cond_env.condition_truth_value, cond_eval_bool[0], ] from qrisp import mcx mcx(toffoli_qb_list, self.condition_truth_value, method="gray_pt") else: # Without any parent environment we can simply synhesize the # truth value of the quantum condition into it's qubit redirect_qfunction(self.cond_eval_function)( *self.args, target=self.qbool, **self.kwargs ) if isinstance(self.env_qs.data[-1], QuantumEnvironment): env = self.env_qs.data.pop(-1) env.compile() cond_eval_bool = self.qbool from qrisp import recursive_qv_search perm_lock(recursive_qv_search(self.args)) unlock(self.condition_truth_value) inversion_tracker = 1 # This list will contain the qubits holding the truth values of # conditional/control environments within this environment. The instruction # from the subcondition environments do not need to be controlled, # since their compile method compiles their condition truth value based # on the truth value of the parent environment. subcondition_truth_values = [ env.condition_truth_value for env in self.sub_condition_envs ] # Now we need to recover the instructions from the data list and # perform their controlled version on the condition_truth_value qubit while self.env_data: instruction = self.env_data.pop(0) # If the instruction == conditional environment, compile the environment if isinstance(instruction, (ControlEnvironment, ConditionEnvironment)): instruction.compile() subcondition_truth_values = [ env.condition_truth_value for env in self.sub_condition_envs ] continue # If the instruction == general environment, compile the instruction and # add the compilation result to the list of instructions that need to be # conditionally executed elif issubclass(instruction.__class__, QuantumEnvironment): temp_data_list = list(self.env_qs.data) self.env_qs.clear_data() instruction.compile() self.env_data = list(self.env_qs.data) + self.env_data self.env_qs.clear_data() self.env_qs.data.extend(temp_data_list) subcondition_truth_values = [ env.condition_truth_value for env in self.sub_condition_envs ] continue if instruction.op.name in ["qb_alloc", "qb_dealloc"]: self.env_qs.append(instruction) continue if set(instruction.qubits).intersection(subcondition_truth_values): self.env_qs.append(instruction) continue # Support for inversion of a condition without opening a new environment if set(instruction.qubits).issubset([self.condition_truth_value]): if instruction.op.name == "x": inversion_tracker *= -1 perm_unlock(self.condition_truth_value) x(self.condition_truth_value) perm_lock(self.condition_truth_value) elif instruction.op.name == "p": p(instruction.op.params[0], self.condition_truth_value) elif instruction.op.name == "rz": rz(instruction.op.params[0], self.condition_truth_value) elif instruction.op.name == "z": z(self.condition_truth_value) else: raise Exception( "Tried to perform invalid operations" "on condition truth value (allowed are x, p, z, rz)" ) continue if self.condition_truth_value in instruction.qubits: self.env_qs.append(convert_to_custom_control(instruction, self.condition_truth_value)) continue else: # Create controlled instruction instruction.op = instruction.op.control( num_ctrl_qubits=1) # Add condition truth value qubit to the instruction qubit list instruction.qubits = [self.condition_truth_value] + list( instruction.qubits ) # Append instruction self.env_qs.append(instruction) unlock(self.condition_truth_value) perm_unlock(self.condition_truth_value) if inversion_tracker == -1: x(self.condition_truth_value) perm_unlock(recursive_qv_search(self.args)) # Uncompute truth values # If we had a parent environment, we first uncompute the # "actual truth value", i.e. the mcx of the parent and # this environments truth value if self.parent_cond_env is not None: self.parent_cond_env.sub_condition_envs.extend( self.sub_condition_envs + [self] ) mcx(toffoli_qb_list, self.qbool, method="gray_pt_inv") self.qbool.delete() # We now uncompute the environments' truth value # For this we can use the uncompute method, which will however not # recompute any intermediate values, therefore blocking alot of qubits # during execution. Especially in nested environments this can quickly # become a problem, because the blocked ancillae can not be reused # for further condition evaluations. # For the condition environment examples we saw an increase from 36 to 44 # qubits without recomputation while only lowering the depth by about 25% # There might be cases where recomputation based uncomputation # is not worth it but for now, we leave it recompute = True # if not recompute: # try: # cond_eval_bool.uncompute() # except: # recompute = True if recompute: with invert(): redirected_qfunction = redirect_qfunction(self.cond_eval_function) redirected_qfunction( *self.args, target=cond_eval_bool, **self.kwargs ) if isinstance(self.env_qs.data[-1], QuantumEnvironment): env = self.env_qs.data.pop(-1) env.compile() cond_eval_bool.delete()
# This decorator allows to have conditional evaluations to return condition environments # when called after a "with" statement but QuantumBools otherwise. # from qrisp import QuantumFloat # a = QuantumFloat(2) # b = QuantumFloat(2) # temp = (a == b) # this is a QuantumBool # with temp: # QuantumBools have an enter method, # but they use the less efficient control environment # pass # with a == b: # This is a ConditionEnvironment entered # pass # We do this mainly for performance reasons. If a QuantumBool is returned, # it can also be entered, but it uses the less efficient (almost a factor 2) # control Environment. # The detection is based on the inspect module. # This implies that the detection only works if the with statements is exactly # two levels above the cond_eval_function call. # That means, a class like QuantumFloat can call __eq__ and then the result # of this decorator in order for this to work def adaptive_condition(cond_eval_function): def new_cond_eval_function(*args, **kwargs): from qrisp import auto_uncompute calling_line = find_calling_line(2) uncomputed_function = auto_uncompute(cond_eval_function) if calling_line.split(" ")[0] == "with" and "&" not in calling_line and "|" not in calling_line and "~" not in calling_line: return quantum_condition(uncomputed_function)(*args, **kwargs) else: return uncomputed_function(*args, **kwargs) return new_cond_eval_function @adaptive_condition def q_eq(input_0, input_1, invert = False): from qrisp import mcx, cx, conjugate, QuantumBool res = QuantumBool(name="eq_cond*", qs = input_0[0].qs()) if isinstance(input_1, QuantumVariable): if input_0.size != input_1.size: raise Exception( "Tried to evaluate equality conditional" "for QuantumVariables of differing size" ) def multi_cx(input_0, input_1): for i in range(len(input_0)): cx(input_0[i], input_1[i]) with conjugate(multi_cx)(input_0, input_1): mcx(input_1, res, ctrl_state=0) else: label_int = input_0.encoder(input_1) mcx(input_0, res, ctrl_state=label_int) if invert: res.flip() return res