ZPowGate to QASM false conversion #6726
Comments
|
For the time being, I'd appreciate guidance on how I can implement |
NoureldinYosri
added
the
triage/discuss
verult
added
triage/accepted
|
So, I got some decompositions for XPow, YPow, and ZPow, but having some trouble with their controlled versions. I am for instance trying the decomposition for a controlled XPow gate, and not quite sure what the decomposition would be. For XPow itself I did self.RX(exponent * np.pi, qubit_indices)
self.GlobalPhase(exponent * np.pi/2)
if global_shift != 0:
self.GlobalPhase(global_shift * np.pi/2)The tricky part is that the controlled version wouldn't be self.CRX(exponent * np.pi, control_index, target_index)
self.GlobalPhase(exponent * np.pi/2)
if global_shift != 0:
self.GlobalPhase(global_shift * np.pi/2)It would need a controlled global phase gate too, which I don't think exists, as global phase is not really a gate, but more of a global correction. I don't really know how a conditional global phase gate could be implemented. IIRC qiskit just calculates the unitary when it creates the conditional version of it, so I'd really try to stay away from gates that use conditional global phase. |
|
The only solution I can immediately think of is implementing its underlying unitary matrix instead, and decomposing it to a qubit qubit sequence of gates just like any other two qubit gates. This is not ideal, as the explicit decomposition could be more efficient. |
|
I'll try to clean up and make a PR on the |
|
I'm playing around with the idea of applying the unitary of the global phase to an actual qubit. So, it sounds very stupid as global phase is a property that's applied to the entire system, but from some testing, if we apply the global phase as a single qubit gate (its unitary representation) to any of the qubits, it behaves the same way as applying a global phase gate. This way, we can account for global phase I think a tad easier. I did question myself ALOT about this idea, and still am, but I made sure the code was correct and the equality was properly checked. I also did the check for gradual accumulation of the global phases and it basically behaves the same way, you'd apply another and another gate. from qiskit import QuantumCircuit
from qiskit.quantum_info import Operator
qc_1 = QuantumCircuit(3)
qc_2 = QuantumCircuit(3)
qc_3 = QuantumCircuit(3)
qc_4 = QuantumCircuit(3)
# Define GHZ state
qc.h(0)
qc.cx(0, 1)
qc.cx(0, 2)
# Apply global phase
global_phase_angle = 1/3
global_phase = np.exp(1j * global_phase_angle) * np.identity(2)
qc_1.global_phase = global_phase_angle
qc_2.unitary(global_phase, [0])
qc_3.unitary(global_phase, [1])
qc_4.unitary(global_phase, [2])
# Define the unitary operators for each circuit
u1 = Operator(qc_1).data
u2 = Operator(qc_2).data
u3 = Operator(qc_3).data
u4 = Operator(qc_4).data
# Ensure all are equal
assert np.all([np.all(u1 == u2), np.all(u1 == u3), np.all(u1 == u4)])P.S : If I have done some really stupid mistake, I apologize in advance. P.S : I think I can account for the controlled global phase by fixing the relative phase of the controlled indices using a MCPhase. |
mhucka
added this to the 2025-T7: Ensure interoperability with other circuit representations milestone
dstrain115
added
triage/discuss
|
Raising this on cirq discuss again. It looks like all of these are due to global phase. For instance, changing the last line from It is true that the unitaries will be different if you use a controlled version of that unitary, but it looks like the decomposition of the controlled XPowGate into qasm actually creates a decomposition, it doesn't do a 'ctrl' operation on 'rz' (which could be off). For example, adding ".controlled_by" in the above code to generate a CZPowGate creates this qasm: So, the qiskit and cirq gates are only off by a global phase. I don't think we want to try to support keeping the global phases in sync, as that will be difficult, clutter up the resulting qasm with a bunch of superfluous 'gphase' commands, and be functionally equivalent. |
|
sync meeting - TODO - verify that a to/from qasm roundtrip produces the same unitary for cirq. |
NoureldinYosri
added
triage/accepted
Description of the issue
Greetings there,
Hope all are well. I am trying to implement the
quantum_shannon_decomposition.pyin qiskit, and to do that I need to representZPowGateusing gates such as Ry, Rz, global phase, and such. To see what the conversion is I used thecirq.qasm()function, however, the conversion is wrong as can be seen below.How to reproduce the issue
Cirq version
The text was updated successfully, but these errors were encountered: