A Resource-Virtualized and Hardware-Aware Quantum Compilation Framework for Real Quantum Computing Processors

Hong-Ze Xu, Xudan Chai, Mengjun Hu, Zheng‐An Wang, Yulong Feng, Yu Chen, Xinpeng Zhang, Jingbo Wang, Wei-Feng Zhuang, Yu‐Xin Jin, Yirong Jin, Haifeng Yu, Heng Fan, Dong E. LiuResearch 2025 · DOI

TL;DR

This paper introduces QSteed, a novel quantum compilation framework that tackles the challenge of efficiently compiling quantum programs for noisy, real-world quantum hardware. QSteed achieves this by virtualizing quantum processors into smaller, manageable "virtual QPUs" (VQPUs) and then applying hardware-aware optimizations to these VQPU subregions, leading to improved compilation speed and circuit fidelity.

What Problem Does This Paper Address?

Quantum computers are rapidly advancing, but they are still scarce and prone to errors (noise). Compiling user-written quantum programs into instructions that can actually run on this hardware is a complex task. Current compilers often struggle with the inherent noise and varying architectures of different quantum processors. They tend to map programs to the entire chip, which can be inefficient and lead to lower-quality results due to the non-uniform distribution of noise. This paper addresses the need for a more intelligent and adaptable compilation approach that can efficiently utilize the available quantum resources while mitigating the impact of hardware noise.

Methods

QSteed introduces a two-pronged approach:

  1. Resource Virtualization: Instead of treating the entire quantum chip as a single unit, QSteed virtualizes it into a hierarchy of abstractions:
  2. QPU: A digital twin of the physical processor.
  3. StdQPU: An idealized representation to handle chip defects.
  4. SubQPU: High-quality substructures identified within a QPU using heuristic strategies (fidelity-first, degree-first, random selection).
  5. VQPU: The final, compiler-facing representation of a SubQPU, containing essential information for compilation. These VQPUs are stored in a database, allowing for unified management of multiple quantum backends.

  6. Select-Then-Compile Workflow: When a user submits a quantum circuit, QSteed first selects the most suitable VQPU from its database based on circuit structure or fidelity metrics. Then, it performs hardware-aware transpilation (optimization and mapping) specifically on that smaller VQPU subregion. This targeted approach avoids the overhead of optimizing for the entire, potentially noisy, chip.

Key components of the transpiler include:

  • Standardized Circuit: Preprocessing to clean up user code.
  • VQPU Selector: Chooses the best VQPU based on fidelity or structural similarity.
  • Quantum Circuit Transpiler: Optimizes the circuit for the selected VQPU, including noise-aware qubit mapping and routing.
  • Hardware Constraint Validation: Final checks to ensure the compiled circuit can run on the hardware.

Datasets

The paper uses calibration data from the Quafu superconducting cluster's Baihua quantum processor and also simulation parameters from Google's Willow processor. They also utilize benchmark circuits from the QASM-Bench suite [52] and generate random circuits for testing.

Key Insights and Ideas

  • Virtualizing Quantum Hardware is Key: The core idea is that by breaking down a large, noisy quantum chip into smaller, high-quality "virtual" regions (VQPUs), compilation can become significantly more efficient and effective.
  • "Select-Then-Compile" Outperforms "Compile-All": Targeting a specific, optimized subregion (VQPU) for compilation, rather than the entire chip, leads to faster compilation times and often better circuit fidelity.
  • Noise-Awareness is Crucial: The framework explicitly incorporates strategies to identify and utilize the best-performing parts of the hardware, directly addressing the impact of noise on quantum computations.
  • A Unified Framework for Heterogeneous Hardware: The VQPU database and abstraction layers provide a promising path towards managing and compiling for diverse quantum hardware architectures in a unified way.

Key Results

  • Faster Compilation: QSteed consistently outperforms leading compilers like Qiskit and Pytket in compilation speed for small to medium-scale benchmark circuits.
  • Comparable or Higher Fidelity: QSteed achieves comparable or higher circuit fidelity compared to other compilers, especially when using its noise-aware VQPU selection strategies.
  • Robustness Demonstrated: The framework's performance was validated on both the Baihua (superconducting) and Willow (superconducting) processors, indicating its architectural robustness.
  • Simulation of Large Circuits: For larger circuits where direct hardware execution is unreliable, QSteed's circuit cost function (a proxy for fidelity) showed competitive results.

Limitations and Future Work

  • Scalability for Future Quantum Computers: The current offline precomputation of the VQPU database might become a bottleneck for future fault-tolerant quantum computers with millions of qubits. Future work may involve more scalable heuristics and parallel search strategies.
  • Calibration Drift: The system relies on up-to-date calibration data. While it can refresh the VQPU database, it cannot detect hardware drift if the hardware itself isn't recalibrated. Future work aims to integrate with hardware benchmarking and automated calibration systems.
  • Cross-Platform Applicability: While the architecture is promising for other platforms (ion traps, neutral atoms), adapting it requires developing specific VQPU identification and selection strategies tailored to those architectures.
  • Optimization Depth: The current version of QSteed focuses on core compilation and virtualization. More advanced optimization passes could be integrated to further improve circuit performance.

Summary updated

Abstract

As quantum computing systems continue to scale up and become more clustered, efficiently compiling user quantum programs into high-fidelity executable sequences on real hardware remains a key challenge for current quantum compilation systems. In this study, we introduce a system-software framework that integrates resource virtualization and hardware-aware compilation for real quantum computing processors, termed QSteed. QSteed virtualizes quantum processors through a 4-layer abstraction hierarchy comprising the real quantum processing unit (QPU), standard QPU (StdQPU), substructure of the QPU (SubQPU), and virtual QPU (VQPU). These abstractions, together with calibration data, device topology, and noise descriptors, are maintained in a dedicated database to enable unified and fine-grained management across superconducting quantum platforms. At run time, the modular compiler queries the database to match each incoming circuit with the most suitable VQPU, after which it confines layout, routing, gate resynthesis, and noise-adaptive optimizations to that virtual subregion. The complete stack has been deployed on the Quafu superconducting cluster, where experimental runs confirm the correctness of the virtualization model and the efficacy of the compiler without requiring modifications to user code. By integrating resource virtualization with a select-then-compile workflow, QSteed demonstrates a robust architecture for compiling programs on noisy superconducting processors. This architectural approach offers a promising path toward efficient compilation needs across various superconducting quantum computing platforms in the noisy intermediate-scale quantum era.
BibTeX
@inproceedings{xu2025resource,
  author = {Hong-Ze Xu and Xudan Chai and Mengjun Hu and Zheng‐An Wang and Yulong Feng and Yu Chen and Xinpeng Zhang and Jingbo Wang and Wei-Feng Zhuang and Yu‐Xin Jin and Yirong Jin and Haifeng Yu and Heng Fan and Dong E. Liu},
  title = {{A Resource-Virtualized and Hardware-Aware Quantum Compilation Framework for Real Quantum Computing Processors}},
  booktitle = {Research},
  year = {2025},
  doi = {10.34133/research.0947},
}

Loading related content...

Figure