Conveying Initial Layout and Output Permutations

[burgholzer]

At the moment, our mappers track the initial layout (initial logical to physical qubit mapping) and output permutation (which physical qubit has to be measured to obtain which logical qubit result) of the mapped circuits and, upon dumping the respective circuits to .qasm, emit them as comments of the form

// i 0 1 ... n
// o 0 1 ... m

This is not standardized in any way and might even change in our own toolset in the future.

At least for the output permutation a possible fix would be to add appropriate measurements to the mapped circuits. For example:

// o 1 0
OPENQASM 2.0;
include "qelib1.inc";

qreg q[2];
creg c[2];

[... circuit ...]

could become

OPENQASM 2.0;
include "qelib1.inc";

qreg q[2];
creg c[2];

[... circuit ...]
measure q[1] -> c[0];
measure q[0] -> c[1];

At the moment there does not seem to exist an apparent/standardized solution for conveying the initial layout in .qasm other than the solution available right now.

At least for the Python bindings, circumventing the .qasm dump and directly constructing a Qiskit QuantumCircuit should be possible. At the moment, Qiskit hides the initial layout in a private member _layout of QuantumCircuit.
Setting this member appropriately and adding measurements to convey the output permutation should allow to use the resulting circuits, e.g., for our verification tool QCEC.

If anyone is interested in working on this, feel free to reach out!

[burgholzer]

#50 introduces measurements at the end of mapped circuits according to the tracked output permutation.

As for the initial layout, cda-tum/mqt-core#35 tracks the corresponding feature.