Qubit Mapping Tailored to Advanced Quantum ISAs Reduces Routing Overhead, Enabling Circuits with up to 35% Fidelity

Compiling instructions for quantum computers presents a significant challenge, as efficiently assigning and connecting quantum bits, known as qubit mapping and routing, often introduces substantial delays and complexity. Zhaohui Yang, Kai Zhang, and Xinyang Tian, along with colleagues at their respective institutions, demonstrate that this process need not be as costly as previously thought. Their research introduces Canopus, a new framework that optimises qubit mapping and routing by directly considering the specific capabilities of modern quantum hardware. By moving beyond simplified models and embracing the diverse instruction sets now available, including gates beyond standard operations, Canopus achieves a 15%-35% reduction in routing overhead compared to existing methods. This advancement represents a crucial step towards realising the full potential of near-term and fault-tolerant quantum computers, paving the way for more complex and efficient quantum algorithms.

This inefficiency arises from a fundamental disconnect, as compilers often rely on abstract routing models that fail to account for the specific capabilities of physical quantum devices. Recent advances in hardware have enabled the implementation of diverse instruction set architectures beyond standard gate operations, creating an opportunity to address this limitation.

Quantum Algorithm Performance Across Benchmarks

Analysis of performance data reveals variations in execution characteristics across several quantum algorithms when run on different quantum benchmarks including ISA, ZZPhase, SQiSW, and Bench. Data also includes performance metrics for specific quantum circuits such as bigadd and wstate, and averages across multiple runs, providing valuable insights into the relative performance of different algorithms and benchmarks. This data is organized in a structured matrix format with numerical values representing measurable performance metrics, and can be used for comparing algorithm performance, benchmarking quantum computers, optimizing algorithms, identifying performance trends, and conducting statistical analysis.

Canopus Optimizes Qubit Mapping and Routing

Scientists have developed Canopus, a new framework for optimizing qubit mapping and routing in quantum computing, achieving significant reductions in circuit complexity. This work addresses a critical bottleneck in both near-term and fault-tolerant quantum computation, where efficiently assigning qubits and arranging operations is essential for successful execution. The team’s approach centers on a novel method for co-optimizing routing and synthesis, minimizing the overall cost of quantum circuits given any quantum instruction set architecture. The core of Canopus lies in its use of a canonical representation of two-qubit gates and the monodromy polytope, which allows for accurate quantification of gate costs and guides synthesis-routing co-optimization.

By formalizing the analysis of commutation relations between two-qubit gates, the researchers created a generalized optimization mechanism extending beyond traditional gate-focused methods. Experiments demonstrate that Canopus consistently reduces routing overhead by 15% to 35% compared to state-of-the-art methods across diverse quantum instruction sets and hardware topologies. This reduction in overhead translates to fewer operations needed to prepare and execute quantum algorithms. Further investigation revealed that theoretically expressive instruction sets consistently outperform the conventional gate instruction set, challenging prior conclusions and providing guidelines for hardware-software co-design.

Case studies, including the execution of a quantum Fourier transform kernel and end-to-end quantum error correction circuit simulations, demonstrate Canopus’ superiority in both near-term and fault-tolerant applications. Specifically, on a one-dimensional chain topology, Canopus discovered a provably optimal routing scheme, surpassing previously reported optimal solutions, and experiments on IBM’s quantum processing units suggest an average 2. 1x fidelity improvement compared to standard methods. These results highlight the potential of Canopus to significantly improve the performance and scalability of quantum computers.

Canopus Optimizes Qubit Routing and Reduces Overhead

Scientists have developed Canopus, a new framework for qubit mapping and routing that significantly reduces inefficiencies in quantum computation. Existing methods often impose substantial overhead in circuit duration due to a disconnect between compiler models and the capabilities of modern quantum hardware. Canopus addresses this by centering qubit routing on deep co-optimization, tailored to diverse instruction set architectures beyond standard gate operations, and leverages a canonical representation of two-qubit gates and the monodromy polytope to model synthesis costs, enabling more intelligent insertion of swap operations during routing. Experimental results demonstrate that Canopus consistently reduces routing overhead by 15% to 35% compared to state-of-the-art methods, across a range of instruction sets and hardware configurations.

This improvement stems from a formalized approach to commutativity-based optimizations, achieved through the use of a generalized canonical form for two-qubit gates. Furthermore, the research yields guidelines for the co-design of quantum programs, instruction sets, and hardware, suggesting that theoretically expressive instruction sets outperform conventional approaches. The team successfully demonstrated Canopus’ superiority in both near-term and fault-tolerant applications, including a provably optimal routing scheme for the Quantum Fourier Transform and a 2. 1x fidelity improvement on IBM’s quantum processing units. Future work will focus on extending the framework to handle more complex quantum algorithms.