TranspileLayout

class qiskit.transpiler.TranspileLayout(initial_layout, input_qubit_mapping, final_layout=None, _input_qubit_count=None, _output_qubit_list=None)

GitHub

Bases: object

Layout attributes for the output circuit from transpiler.

The transpiler is unitary-preserving up to the “initial layout” and “final layout” permutations. The initial layout permutation is caused by setting and applying the initial layout during the Layout stage. The final layout permutation is caused by SwapGate insertion during the Routing stage. This class provides an interface to reason about these permutations using a variety of helper methods.

During the layout stage, the transpiler can potentially remap the order of the qubits in the circuit as it fits the circuit to the target backend. For example, let the input circuit be:

from qiskit.circuit import QuantumCircuit, QuantumRegister
 
qr = QuantumRegister(3, name="MyReg")
qc = QuantumCircuit(qr)
qc.h(0)
qc.cx(0, 1)
qc.cx(0, 2)
qc.draw("mpl")

Circuit diagram output by the previous code.

Suppose that during the layout stage the transpiler reorders the qubits to be:

from qiskit import QuantumCircuit
 
qc = QuantumCircuit(3)
qc.h(2)
qc.cx(2, 1)
qc.cx(2, 0)
qc.draw("mpl")

Then the output of the initial_virtual_layout() method is equivalent to:

Layout({
    qr[0]: 2,
    qr[1]: 1,
    qr[2]: 0,
})

(it is also this attribute in the QuantumCircuit.draw() and circuit_drawer() which is used to display the mapping of qubits to positions in circuit visualizations post-transpilation).

Building on the above example, suppose that during the routing stage the transpiler needs to insert swap gates, and the output circuit becomes:

from qiskit import QuantumCircuit
 
qc = QuantumCircuit(3)
qc.h(2)
qc.cx(2, 1)
qc.swap(0, 1)
qc.cx(2, 1)
qc.draw("mpl")

Then the output of the routing_permutation() method is:

[1, 0, 2]

which maps positions of qubits before routing to their final positions after routing.

There are three public attributes associated with the class, however these are mostly provided for backwards compatibility and represent the internal state from the transpiler. They are defined as:

initial_layout - This attribute is used to model the permutation caused by the Layout stage. It is a Layout object that maps the input QuantumCircuits Qubit objects to the position in the output QuantumCircuit.qubits list.

input_qubit_mapping - This attribute is used to retain input ordering of the original QuantumCircuit object. It maps the virtual Qubit object from the original circuit (and initial_layout) to its corresponding position in QuantumCircuit.qubits in the original circuit. This is needed when computing the permutation of the Operator of the circuit (and used by Operator.from_circuit()).

final_layout - This attribute is used to model the permutation caused by the Routing stage. It is a Layout object that maps the output circuit’s qubits from QuantumCircuit.qubits in the output circuit to their final positions after routing. Importantly, this only represents the permutation caused by inserting SwapGates into the QuantumCircuit during the Routing stage. It is not a mapping from the original input circuit’s position to the final position at the end of the transpiled circuit. If you need this, you can use the final_index_layout() to generate this. If final_layout is set to None, this indicates that routing was not run, and can be considered equivalent to a trivial layout with the qubits from the output circuit’s qubits list.

Attributes

Parameters

initial_layout (Layout) –
input_qubit_mapping (dict[circuit.Qubit, int]) –
final_layout (Layout | None) –
_input_qubit_count (int | None) –
_output_qubit_list (List[Qubit] | None) –

final_layout

Type: Layout | None

Default value: None

initial_layout

Type: Layout

input_qubit_mapping

Type: dict[circuit.Qubit, int]

Methods

final_index_layout

final_index_layout(filter_ancillas=True)

GitHub

Generate the final layout as an array of integers.

This method will generate an array of final positions for each qubit in the input circuit. For example, if you had an input circuit like:

qc = QuantumCircuit(3)
qc.h(0)
qc.cx(0, 1)
qc.cx(0, 2)

and the output from the transpiler was:

tqc = QuantumCircuit(3)
tqc.h(2)
tqc.cx(2, 1)
tqc.swap(0, 1)
tqc.cx(2, 1)

then the final_index_layout() method returns:

[2, 0, 1]

This can be seen as follows. Qubit 0 in the original circuit is mapped to qubit 2 in the output circuit during the layout stage, which is mapped to qubit 2 during the routing stage. Qubit 1 in the original circuit is mapped to qubit 1 in the output circuit during the layout stage, which is mapped to qubit 0 during the routing stage. Qubit 2 in the original circuit is mapped to qubit 0 in the output circuit during the layout stage, which is mapped to qubit 1 during the routing stage. The output list length will be as wide as the input circuit’s number of qubits, as the output list from this method is for tracking the permutation of qubits in the original circuit caused by the transpiler.

Parameters

filter_ancillas (bool) – If set to False any ancillas allocated in the output circuit will be included in the layout.

Returns

A list of final positions for each input circuit qubit.

Return type

List[int]

final_virtual_layout

final_virtual_layout(filter_ancillas=True)

GitHub

Generate the final layout as a Layout object.

This method will generate an array of final positions for each qubit in the input circuit. For example, if you had an input circuit like:

qc = QuantumCircuit(3)
qc.h(0)
qc.cx(0, 1)
qc.cx(0, 2)

and the output from the transpiler was:

tqc = QuantumCircuit(3)
tqc.h(2)
tqc.cx(2, 1)
tqc.swap(0, 1)
tqc.cx(2, 1)

then the return from this function would be a layout object:

Layout({
    qc.qubits[0]: 2,
    qc.qubits[1]: 0,
    qc.qubits[2]: 1,
})

Parameters

filter_ancillas (bool) – If set to False any ancillas allocated in the output circuit will be included in the layout.

Returns

A layout object mapping to the final positions for each qubit.

Return type

Layout

from_property_set

classmethod from_property_set(dag, property_set)

GitHub

Construct the TranspileLayout by reading out the fields from the given PropertySet. Returns None if there are no layout-setting keys present.

This includes combining the different keys of the property set into the full set of initial and final layouts, including virtual permutations.

This does not invalidate or in any way mutate the given property set. In order to “canonicalize” the property set afterwards, call write_into_property_set().

This reads the following property-set keys:

layout

Required. The Layout object mapping virtual qubits (potentially expanded with ancillas) to physical-qubit indices. This corresponds directly to initial_layout.

Note

In all standard use, this is a required field. However, if virtual_permutation_layout is set, then a “trivial” layout will be inferred, even if the circuit is not actually laid out to hardware. This is an unfortunate limitation of this class’s data model, where it is not possible to specify a final permutation without also having an initial layout. This deficiency will be corrected in Qiskit 3.0.

original_qubit_indices

Required (but automatically set by the PassManager). The mapping {virtual: index} that indicates which relative index each incoming virtual qubit was, in the input circuit. This can be expanded with ancillas too (in which case the ancilla indices don’t mean much, since they weren’t in the incoming circuit).

num_input_qubits

Required (but automatically set by the PassManager). The number of explicit virtual qubits in the input circuit (i.e. not including implicit ancillas).

final_layout

Optional. The effective final permutation, in terms of the current qubits of the DAGCircuit. This corresponds directly to final_layout.

virtual_permutation_layout

Optional. This is set by certain optimization passes that run before layout selection, such as ElidePermutations. It is similar in spirit to final_layout, but typically only applies to the input virtual qubits.

Warning

This object uses the opposite permutation convention to final_layout due to an oversight in Qiskit during its introduction. In other words, virtual_permutation_layout maps a Qubit instance at the end of the circuit to its integer index at the start of the circuit.

Parameters

dag (DAGCircuit) – the current state of the DAGCircuit.
property_set (PropertySet) – the current transpiler’s property set. This must at least have the layout key set.

Return type

TranspileLayout | None

initial_index_layout

initial_index_layout(filter_ancillas=False)

GitHub

Generate an initial layout as an array of integers.

Parameters

filter_ancillas (bool) – If set to True any ancilla qubits added to the transpiler will not be included in the output.

Returns

A layout array that maps a position in the array to its new position in the output circuit.

Return type

List[int]

initial_virtual_layout

initial_virtual_layout(filter_ancillas=False)

GitHub

Return a Layout object for the initial layout.

This returns a mapping of virtual Qubit objects in the input circuit to the positions of the physical qubits selected during layout. This is analogous to the initial_layout attribute.

Parameters

filter_ancillas (bool) – If set to True only qubits in the input circuit will be in the returned layout. Any ancilla qubits added to the output circuit will be filtered from the returned object.

Returns

A layout object mapping the input circuit’s Qubit objects to the positions of the selected physical qubits.

Return type

Layout

routing_permutation

routing_permutation()

GitHub

Generate a final layout as an array of integers.

If there is no final_layout attribute present then that indicates there was no output permutation caused by routing or other transpiler transforms. In this case the function will return a list of [0, 1, 2, .., n].

Returns

A layout array that maps a position in the array to its new position in the output circuit.

Return type

List[int]

write_into_property_set

write_into_property_set(property_set)

GitHub

‘Unpack’ this layout into the loose-constraints form of the property_set.

This is the inverse method of from_property_set().

This always writes the follow property-set keys, overwriting them if they were already set:

layout

Directly corresponds to initial_layout.

original_qubit_indices

Directly corresponds to input_qubit_mapping.

final_layout

Directly corresponds to final_layout. Note that this might not be identical to the final_layout from before a call to from_property_set(), because the effects of virtual_permutation_layout will have been combined into it.

virtual_permutation_layout

Deleted from the property set; TranspileLayout “finalizes” the multiple separate permutations into one single permutation, to retain the canonical form.

In addition, the following keys are updated, if this TranspileLayout has a known value for them. They are left as-is if not, to handle cases where this class was manually constructed without setting certain optional fields.

num_input_qubits

The number of non-ancilla virtual qubits in the input circuit.

Parameters

property_set (dict[str, object]) – the PropertySet (or general dict) that the output should be written into. This mutates the input in place.