Is QFT/inverse QFT available on Qulacs?

[pat-phattharaporn]

Hi! I'm wondering is QFT/inverse QFT available on Qulacs? If they are, would you mind pointing me to the documentation? Otherwise, are you planning to include it in the package or we need to manually construct the circuit? Thank you very much!

[corryvrequan]

Hi @pat-phattharaporn
Qulacs does not have an API to generate QFT / Inverse QFT circuit, and currently, we have no plan to include a template for constructing QFT circuit in the library. (It is possible to include it in the future, but we have currently working on speed-up of simulation rather than improving usability.)
Thus, yes, we need to manually construct the circuit.

[ghost]

Hello! On a related note, I'm trying to construct the QFT circuit from scratch yardstiq/quantum-benchmarks#30 but I was unable to find a CUz gate because of which I'm decomposing the CUz into dense matrix and CNOT gates. Am I missing the API for CUz gate?

[corryvrequan]

Hi @Codewithsk .

Sorry for very late reply.
Qulacs does not have controled-RZ gate as a basic gate.
(There are CUz gate in qft modules inside C-language codes, but they are not exported to python.)
Instead, you can use the following.

gate = DenseMatrix([target], matrix2x2)
gate.add_control_qubit(control, 1)

or

gate = DenseMatrix([target, control], matrix4x4)

I guess which is faster depends on the environment.

[ghost]

Hi @corryvrequan!

We're using the following:

circuit.add_dense_matrix_gate(wire, [[1,0],[0, np.exp(1j * (np.pi/(2**(wire-i)))/2)]])
circuit.add_CNOT_gate(i, wire)
circuit.add_dense_matrix_gate(wire, [[1,0],[0, np.exp(-1j * (np.pi/(2**(wire-i)))/2)]])
circuit.add_CNOT_gate(i, wire)

Is there a performance difference between these two implementations?

[KowerKoint]

Sorry for very late reply.
I have no idea that which implementation is better, but I think there is just a little difference.
You can measure the performance.
You can use U1 gate to implement circuit.add_dense_matrix_gate(wire, [[1,0],[0, np.exp(1j * (np.pi/(2**(wire-i)))/2)]]) circuit.add_CNOT_gate(i, wire) and it is better.