Qubit stop time

[ewoutvandenberg]

Environment

Latest Qiskit Terra version (v.0.21.2)

What is happening?

In the qubit_stop_time function, the maximum stop time is returned when all qubits are flagged as completed (current line 4971). If one of the provided qubits only has delays on it, this condition is never satisfied and zero will be returned at the end of the function. The stop time of two qubits, one with gates and one with only delays, will therefore be zero.

How can we reproduce the issue?

See source code.

What should happen?

Not sure if the behavior described in the "What is happening" section is correct, the docstring says the function should "Return the stop time of the last instruction, excluding delays, over the supplied qubits."

Any suggestions?

Perhaps return the maximum stop time of all qubits marked as done at the end of the instruction loop, and zero if none of the qubits is marked as done. Using a set of completed qubits rather than a dictionary may make it easier to determine the length of completed qubits. A similar change may be needed in determining the qubit start time.

[jakelishman]

Please can you provide a minimal code example that shows this function not working as you expect, and what output you would expect to receive from it?

[ewoutvandenberg]

import qiskit

circuit = qiskit.QuantumCircuit(2)
circuit.x(0)

durations = qiskit.transpiler.InstructionDurations([('x',(0,), 80, 'dt')])
transpiled = qiskit.transpile(circuit,
                              initial_layout=[0,1],
                              basis_gates=['x'],
                              optimization_level=0,
                              scheduling_method='asap',
                              instruction_durations = durations)

print(transpiled.qubit_stop_time(0))   # 80
print(transpiled.qubit_stop_time(1))   # 0
print(transpiled.qubit_stop_time(0,1)) # Should be 80?

[JustinWoodring]

Here is a PR that addresses this issue, #11097.