Introduction to dynamic circuits


Click here to view this tutorial as an interactive Jupyter notebook in IBM Quantum Lab (requires sign-in).

This tutorial introduces the current dynamic circuit support through Qiskit on IBM Quantum hardware.

Dynamic circuits are quantum circuits that contain mid-circuit measurements that affect the control flow of gate execution later in the circuit, otherwise known as feed-forward operations. The canonical and most simple example is conditional reset, where we perform a projective measurement of the qubit, and if the result is a 1, we flip the state using an X-gate from the |1> to the |0> state. If the result is a 0 we do nothing, as we are already in |0> .

Dynamic circuit support requires significant changes to the hardware and software backends that execute quantum circuits, compared to traditional ‘static’ circuits. Many of IBM Quantum’s hardware backends have been upgraded to support dynamic circuits. Support for dynamic circuits will continue to grow and improve in ease of use and efficiency. We aim to keep the tutorials in this documentation up to date with what IBM’s hardware and software backends provide.

import warnings

Run a dynamic circuit


Before running this notebook, make sure to the follow the steps in the Getting started with Dynamic Circuits guide, to verify you are set up to run dynamic circuit programs on IBM hardware.

Step 1: Load backend

Backends provide a handle to the quantum hardware instance. The main approach for running circuits will be to call<your circuit>, dynamic=True). Backends are obtained from the IBMProvider imported from qiskit_ibm_provider, and are specified by a system name along with the hub/group/project string that gives you access to that system.

# Note: This can be any hub/group/project that has access to the required device and Qiskit Runtime.
hub = "<hub>"
group = "<group>"
project = "<project>"
backend_name = "<your backend>"
hgp = f"{hub}/{group}/{project}"
from qiskit_ibm_provider import IBMProvider

provider = IBMProvider()
backend = provider.get_backend(backend_name, instance=hgp)

Loading a simulator backend

We can also use a simulator to simulate the behavior of dynamic circuits. We obtain a simulator backend similarly to a hardware backend.

from qiskit import Aer
backend_sim = Aer.get_backend('aer_simulator')

Step 2: Write a dynamic circuit using Qiskit

We start with a conditional reset circuit on a single qubit. We first try to excite the qubit with an X-gate followed by a measurement, and another X-gate if the measurement returns a 1.

In older versions of Qiskit, limited conditional execution was supported using the c_if() construct. This was a function that all gate objects exposed, which allowed gate execution to be conditioned on the value of a classical register. A significant limitation was that only a single gate could be conditioned on the classical register value.

In contrast, the new style of classical control flow with the if_test() allows significantly more flexibility in the values that can be conditioned on and the sub-circuits that can be defined within the body of the if_test(). It also enables else conditions. More documentation on the if_test() is available here.

from qiskit import QuantumCircuit, QuantumRegister, ClassicalRegister
from qiskit import transpile

qr = QuantumRegister(1)
crx = ClassicalRegister(1, name="xresult")
crm = ClassicalRegister(1, name="measureresult")
qc_reset = QuantumCircuit(qr, crx, crm, name="Reset")
qc_reset.measure(0, crx)
# qc_reset.x(0).c_if(crx, 1) # old style conditional gate
with qc_reset.if_test((crx, 1)): # new style conditional block (more flexible)
qc_reset.measure(0, crm)

qc_reset = transpile(qc_reset, backend, initial_layout=[0])
qc_reset.draw(output="mpl", idle_wires=False)

Step 3: Run conditional reset in a simulator

Running on the simulator is very similar to running on the hardware.

reset_sim_job =
reset_sim_result = reset_sim_job.result()
reset_sim_counts = reset_sim_result.get_counts(0)
print(f"Full counts including reset: {reset_sim_counts}")
Full counts including reset: {'0 1': 1024}

Step 4: Run conditional reset on hardware

Now we run our conditional reset program on the backend. We turn off the automatic qubit initialization provided by the backend so that we can examine the results of our reset program directly.

reset_job =
    qc_reset, dynamic=True, init_qubits=False
)  # Turn off automatic init

Let’s understand the results:

from qiskit.result import marginal_counts

reset_result = reset_job.result()
reset_counts = reset_result.get_counts(0)
mitigated_reset_results = marginal_counts(reset_counts, indices=[1])
print(f"Full counts including reset: {reset_counts}")
    f"Results from our reset - |0>s prepared {mitigated_reset_results.get('0')}, |1>s prepared {mitigated_reset_results['1']}"
Full counts including reset: {'0 0': 108, '0 1': 3190, '1 0': 12, '1 1': 690}
Results from our reset - |0>s prepared 3298, |1>s prepared 702

Step 5: Conditional Reset with OpenQASM 3

Another way to write the same circuit above is to use OpenQASM 3, a quantum assembly language well-suited to describing dynamic circuits. The live specification provides full details. Qiskit provides support to directly convert QuantumCircuit objects into OpenQASM 3.0 strings.

    from qiskit import qasm3
except ImportError:
        "You likely do not have the latest version of Qiskit installed. Please install it to run this code"

def dump_qasm3(circuit, backend=backend):
    return qasm3.Exporter(

qc_reset_qasm3 = dump_qasm3(qc_reset)
bit[1] xresult;
bit[1] measureresult;
x $0;
xresult[0] = measure $0;
if (xresult == 1) {
    x $0;
measureresult[0] = measure $0;


The names of the classical registers (xresult and measureresult) in the qasm3 string above are the same as those provided when the circuit was first constructed. These make the OpenQASM 3 string much more human-readable.

Step 6: Submit an OpenQASM 3 string directly

It is also possible to submit an OpenQASM 3 string directly instead of a QuantumCircuit object.

This is useful when the QuantumCircuit does not yet support all of the semantics of OpenQASM 3. To see what features of OpenQASM 3 are currently supported in hardware, please see the feature support table here.

See Dynamic-Circuits-basics-with-OpenQASM3 for more information.


We cannot use circuit merging when OpenQASM 3 source strings are submitted. In this case each OpenQASM 3 source will run individually.

qasm3_job =, dynamic=True)
{'0': 76, '1': 3358, '10': 44, '11': 522}