IBM Dynamic Circuits: The Utility-Scale Quantum Efficiency
Every IBM User Gets Utility-Scale Dynamic Circuits for Groundbreaking Quantum Efficiency
A major Qiskit Runtime upgrade from IBM Dynamic Circuits allows all users to access utility-scale dynamic circuits. This unique implementation allows for strong economies and the investigation of complex difficulties and application scenarios that were previously unreachable with conventional “static” circuits. IBM dynamic circuits were added to Qiskit Runtime in 2022, but customers had trouble scaling them. IBM has removed these barriers, allowing all Qiskit Runtime users to fully explore utility-size dynamic circuits.
New dynamic circuits offer significant performance advantages over their predecessors, already advancing the topics and applications that quantum computers can examine.
Understanding IBM Dynamic Circuits' Power Conventional quantum circuits, sometimes called static circuits, use predetermined quantum logic gates on qubits that are configured and tested.
Dynamic circuits use mid-circuit measurements to identify a qubit's value before circuit execution. Importantly, they use classical computation and conditional logic, commonly known as classical feedforward, to identify which quantum operations to do in the next segment of the circuit based on measurement data.
This lets sophisticated quantum protocols be implemented in shallow or constant circuit depth. Dynamic circuits can run larger problems with more qubits quicker than feedforward when scaling is favourable. They are attractive for utility-scale challenges with short-term advantage.
Unleash Performance and Parallelism
The initial 2022 dynamic circuit implementation's global control flow required conditional operations to affect many circuit components one after another, which was a major shortcoming. Because this method was slow, qubit decoherence happened before it was done.
IBM built the new dynamic circuits implementation from scratch, adding parallel conditional operations. The new infrastructure detects and performs independent conditional activities simultaneously. Parallel execution increases circuit depth, execution time, noise reduction, and result fidelity.
Time reductions in traditional techniques enabled these advances. Utilization-scale dynamic circuits yield measurable benefits:
Two-qubit gates decreased 28% each Trotter step, a little simulation time increase.
Performance can be 24% better than unitary circuits.
The new MidCircuitMeasure command captures qubit findings about a microsecond (940 ns) faster than the old version, improving dynamic circuit duration by 65%.
To 600 ns, feedforward latency has decreased. Improved sequence translator increases payload generation time 20 times.
Improvements allow a utility-scale quantum computer to scale to full device utilisation by using all 100+ qubits.
Improvements to Control and Debugging The complexity of classical feedforward makes circuit scheduling difficult. Qiskit lacked a reliable technique to replicate classical operation execution time, therefore users had to manually incorporate fixed delays, which was wasteful.
Stretch duration was invented by IBM to remedy this. Stretch lets users express temporal purpose without specifying delay durations. This simplifies scheduling and allows precise dynamical decoupling error suppression approaches to minimise error buildup during the extended mid-circuit measurement phase when qubits are idle.
In Sampler work results, Qiskit Runtime may now give precise circuit timing for troubleshooting and optimisation. The new visualisation tool draw_circuit_schedule_timing reduces idle time enabling more exact scheduling and performance adjustments, improving circuit quality.
Optimising Mid-Circuit Measurements
Utility-scale dynamic circuits introduce a MidCircuitMeasure instruction for IBM QPU mid-circuit measurements. Before, mid-circuit and terminal measurements used the same generic measure command. Terminal measurements, which finish execution without affecting quantum operations, made mid-circuit operations noisier since the initial instruction was not speed-optimized. Optimised instructions improve calibration and performance.
Since this optimised measurement instruction is not available everywhere, users should use service.backends(filters=lambda b: "measure_2" in b.supported_instructions) to discover backends that support it.
Current research opportunities and limitations
A robust new implementation from redesigned dynamic circuits lets clients research novel utility-scale applications. Researchers demonstrated the possibility by replicating a 46-site kicking Ising Hamiltonian on 106 qubits using the novel circuits. Its scalability allows common users to explore with exciting theoretical ideas like constant- or shallow-depth quantum state preparation. Another intriguing application for dynamic circuits is quantum advantage at constant depth, which static circuits cannot achieve.
Utility-scale dynamic circuits are powerful but being improved. Only the conditional if statement is supported (excluding for loops and switch statements); nested conditionals are not allowed; and resets and measurements are not supported. Please consult the docs for full limitations.