Towards Supporting QIR: Steps for Adopting the Quantum Intermediate Representation

TL;DR

This paper explores how to adopt the Quantum Intermediate Representation (QIR) in quantum computing, demonstrating a practical transition from a quantum circuit simulator to a QIR runtime without performance loss and enabling support for classical control flow.

Contribution

It provides a practical approach to adopting QIR, showing how existing simulators can be transformed into QIR runtimes with minimal effort and no performance degradation.

Findings

Transitioning to QIR is less complex than expected.

QIR support does not impair simulator performance.

Support for arbitrary classical control flow can be added easily.

Abstract

Intermediate representations (IRs) play a crucial role in the software stack of a quantum computer to facilitate efficient optimizations for executing an application on hardware. One of those IRs is the Quantum Intermediate Representation (QIR), which builds on the classical LLVM compiler infrastructure. In this article, we outline different approaches to how QIR can be adopted. This exploration culminates in a demonstration of what it takes to turn an existing quantum circuit simulator into a QIR runtime and that such a transition is less daunting than it might seem at first. We further show that switching to QIR does not entail any performance deficits compared to the original simulator. On the contrary, the presented steps effortlessly allow adding support for arbitrary classical control flow to any classical simulator. We conclude with an outlook on future directions using QIR. The implemented QIR runtime is available under https://github.com/munich-quantum-toolkit/core.