Incorrect simulation of nested classical controls on the AerBackend

[dandanua]

I was trying a fairly simple test of nested classical controls in quantinuum jupyter lab and ran into an issue.

Consider a setup with 3 qubits and 3 classical registers. On the first qubit we use the H gate and then prepare a Bell state on the other two qubits using a standard circuit controlled by the state of the first qubit. However, when the Bell state preparation is done using an equivalent classically controlled operation (well, not quite equivalent, but the simulation result should be the same), the simulation result of the whole setup is incorrect. Here is the code:

from pytket.circuit import *
from pytket.circuit.display import render_circuit_jupyter as draw
from pytket.extensions.qiskit import AerBackend, AerStateBackend
from pytket.utils import probs_from_counts

U0 = Circuit(2, 2)
U0.H(0)
# ------ 
# The use of CX version below works fine and returns the correct result in the end
# U0.CX(0, 1)
# U0.measure_all()
# ---------
# The classical control version fails
U0.Measure(0, 0)
U0.X(1, condition_bits=[0], condition_value=1)
U0.Measure(1, 1)

g_U0 = CustomGateDef.define("u0", U0, [])

U = Circuit()
q = U.add_q_register("q", 3)
c = U.add_c_register("c", 3)
U.H(0)
U.Measure(0, 0)
U.add_custom_gate(g_U0, [], [1, 2, 1, 2], condition=if_not_bit(c[0]))

backend = AerBackend()
c_U = backend.get_compiled_circuit(U)

handle = backend.process_circuit(c_U, n_shots=1000)
counts = backend.get_result(handle).get_counts()
print(counts)
print(probs_from_counts(counts))

which returns

{(0, 0, 1): 0.258, (0, 1, 1): 0.243, (1, 0, 0): 0.499}

while it should return

{(0, 0, 0): 0.256, (0, 1, 1): 0.253, (1, 0, 0): 0.491}

as in the case of using CX in the custom nested gate.

[CalMacCQ]

Hi @dandanua, Thank you for raising an issue. We'll take a look and get back to you.

Can I ask which version of pytket/pytket-qiskit you are using?

[dandanua]

@CalMacCQ

Hi, it's the one that Quantinuum Nexus is using in their Jupyter Lab, within the standard Python kernel. I don't know how to check the version number from there.

[CalMacCQ]

Ah thanks, good to know. thats enough for me. I'll dig into the issue soon.

I think you should just be able to do a ! followed by the shell command in a python notebook cell to find package verions.

! pip show pytket-qiskit 

[dandanua]

Version: 0.56.0

BTW, I've tried to run the same circuit against the H1-Emulator using the "getting-started" notebook workflow. The compilation job ran successfully and produced some transformed circuit but the execution job has failed with the error

JobError: Job errored with detail: <class 'pytket.backends.backend_exceptions.CircuitNotValidError'>: Circuit with index 0 in submitted does not satisfy GateSetPredicate:{ TK2 ExplicitPredicate CopyBits ClExpr WASM ClassicalExpBox ZZMax ZZPhase Barrier MultiBit PhasedX ExplicitModifier Rz RangePredicate Reset SetBits Measure } (try compiling with backend.get_compiled_circuits first).

I'm not sure how to fix this, since I already execute the compiled circuit as in the guide.

[CalMacCQ]

BTW, I've tried to run the same circuit against the H1-Emulator using the "getting-started" notebook workflow. The compilation job ran successfully and produced some transformed circuit but the execution job has failed with the error

hmm strange, I'll make a note to look into that as well as the issue above. The transformed circuit link seems not to work for me.

[dandanua]

You must have blocked access to imgur. Here is another link