Efficient Quantum Control via Automatic Control Skips

We are excited to share our latest paper Efficient Quantum Control via Automatic Control Skips now available on arXiv. This work tackles one of the core challenges in quantum programming: efficiently implementing conditional logic without overwhelming today’s limited quantum hardware. Our method automatically identifies operations where control gates can be safely skipped. This significantly reduces circuit complexity without compromising correctness. It's a step forward in making scalable, high-level quantum programming more practical and resource-efficient.

Conditional statements (If “condition”, do “something”) serve as fundamental building blocks of modern programming and are vital for anyone implementing classical algorithms. It makes sense that they would also be essential for implementing quantum algorithms. Here’s the thing: generalizing for the quantum regime isn’t straightforward. This is because evaluating the condition generally requires performing a measurement that would collapse the wavefunction—one of the big advantages in quantum computing. The solution is to use control gates, which allow us to maintain control qubit coherence and with it the many benefits of quantum computation. The best-known example of this is the CX gate, the controlled version of the X gate.

The downside is that control gates are costly. These gates act on qubits that represent both the condition and the operation target. As a result, they’re inherently multiqubit and generally exhibit lower fidelities—making the circuits more complex and more fragile. Moreover, controlled versions of more complicated gates are often not native to quantum hardware. They require additional decompositions of the resulting circuit, which increases gate counts and depth measures and degrades the overall circuit quality. In short, control gates are crucial building blocks of quantum algorithms, but using too many of them can make the algorithm impractical—especially given the limited resources available on today’s quantum hardware.

The good news, as we discovered, is that it’s often possible to skip the execution of some control operations. Suppose we want to apply the control of a code block defined by a sequence of operations ABCDE. A straightforward implementation would be to apply the controlled version of A, then that of B, that of C, and so on. But, under certain conditions, it’s possible to skip controlling some of them without altering the circuit functionality. For example, we can apply the controlled A, the controlled B, then the uncontrolled C, the controlled D, and the uncontrolled E.

Control skips have been used to improve numerous quantum algorithms. But, until now, finding patterns of skippable operations has been done manually, tailored explicitly by the programmer for each algorithm. This approach is impractical because quantum algorithms increase in size and complexity. What’s more, high-level programming languages that are crucial for implementing quantum algorithms at scale, can hide these patterns, ruling out manual optimization.

In our recent paper, we show how skippable operations can be identified generically and automatically, oblivious to functional intent, at any hierarchical description level. We prove that finding the best set of skippable operations is difficult, but even suboptimal solutions can still yield major performance improvements, such as a 50% reduction in gate counts. Our approach fits best in the context of a quantum program compiler and has already been integrated into the Qmod compiler offered by Classiq.

Because this method is algorithm-agnostic and works across all levels of circuit representation, it allows quantum developers to write clear, high-level code without sacrificing performance. By automatically identifying and applying control skips, this approach brings hardware-level efficiency to abstract program descriptions. This aligns perfectly with Classiq’s mission to make quantum algorithm development scalable, efficient, and practical—without compromising efficiency.