Simplifying Qiskit circuit with c_if()

Question

I'm trying to simplify the inner for loop of this implementation of the Mosca-Ekert semi-classical variant of Shor's algorithm. The inner for loop should have only linear length, but this implementation is exponential.

From https://github.com/ttlion/ShorAlgQiskit/blob/master/Shor_Sequential_QFT.py

    """ Cycle to create the Sequential QFT, measuring qubits and applying the right gates according to measurements """
    for i in range(0, 2*n):
        """reset the top qubit to 0 if the previous measurement was 1"""
        circuit.x(up_reg).c_if(c_aux, 1)
        circuit.h(up_reg)
        cMULTmodN(circuit, up_reg[0], down_reg, aux, a**(2**(2*n-1-i)), N, n)
        """cycle through all possible values of the classical register and apply the corresponding conditional phase shift"""
        for j in range(0, 2**i):
            """the phase shift is applied if the value of the classical register matches j exactly"""
            circuit.u1(getAngle(j, i), up_reg[0]).c_if(up_classic, j)
        circuit.h(up_reg)
        circuit.measure(up_reg[0], up_classic[i])

The obvious way of using c_if(up_classic[j],j) doesn't work. I've tried using an if statement as in

for j in range(0, i):
    """the phase shift is applied if the value of the classical register matches j exactly"""
    if up_classic[j]:
        circuit.u1(getAngle(2**j,i), up_reg[0])

but test runs fail to find the correct factor for small semiprimes like 21 or 33.

Answer 1

The conditional c_if on Qiskit checks a condition on a classical register, not a classical bit. Therefore, up_classic[j] is not valid.

Doing the if at construction time (like on your second case) makes no sense, since up_classic[j] is unknown at that time. Notice that, in the original code, up_classic is set at the end of the loop, with circuit.measure(up_reg[0], up_classic[i]).