Quantum-Classical Programming in Guppy and Brat 

Abstract

Practical applications of quantum computers involve a mixture of classical and quantum operations across quite different time scales. At the large scale, we might interleave complete subroutines, such as estimating the value of a quantum observable and calculating its classical parameter updates, as seen in the VQE algorithm. At the small scale, real-time classical logic may decide on the fly, which quantum operations should be applied next while the qubits are still coherent, as needed for example for quantum error correction schemes.

In this talk, I will present two languages that address hybrid quantum-classical computation from opposite ends of this spectrum. First, I will introduce Brat, a purely functional, dependently typed language that enforces a strict distinction between its quantum and classical fragment on the type level. This design allows for staged compilation and makes Brat well suited for the large time scale quantum-classical interaction described above. On the other end of this spectrum, I will introduce Guppy, an imperative quantum programming language embedded in Python that focusses on real-time classical logic and control-flow. Guppy was designed for maximum expressivity as well as ease of use, and I will touch on some of the design trade-offs required for achieving this goal while still providing useful static safety guarantees.

Finally, we will take a closer look at the shared compiler infrastructure underlying Brat and Guppy. Both languages lower to HUGR, Quantinuum’s novel quantum–classical intermediate representation, allowing them to benefit from a common optimisation pipeline. Time permitting, I may also briefly touch on some ongoing research in the area of quantum-classical program transformation and optimisation.