🧭 Canopus: Canonical-Optimized Placement Utility Suite for Quantum Compilation

Canopus (Canonical-Optimized Placement Utility Suite) is a unified qubit mapping/routing framework tailored to diverse quantum ISAs, i.e., the implementation of the paper "Unifying Qubit Routing Across Diverse Quantum ISAs via Canonical Representation". Its main function is to optimize the layout and routing of qubits on quantum hardware, handling optimal synthesis with diverse ISAs in a unified approach through two-qubit canonical gate representation, providing guidance for hardware-software co-design.

Canopus evokes the name of the second-brightest star in the sky, symbolizing its role as a "navigational" guide for routing qubits through the complex constraints of quantum hardware.

Installation

Prerequisite: monodromy (+ lrslib)

Canopus depends on the monodromy library for Weyl-chamber coverage computation. monodromy is not on PyPI and must be installed manually before pip install canopus-quantum:

pip install git+https://github.com/Youngcius/monodromy

monodromy itself needs the lrs binary from lrslib. Platform-specific setup:

brew install gmp
cd lrslib-073 && make
sudo make install

sudo apt update && sudo apt install -y build-essential libgmp-dev
cd lrslib-073 && make
sudo make install

conda install -c conda-forge gmp
cd lrslib-073
make INCLUDEDIR="$CONDA_PREFIX/include" LIBDIR="$CONDA_PREFIX/lib"
make prefix=$HOME/.local install
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc && source ~/.bashrc

From PyPI (recommended)

pip install canopus-quantum

Pre-compiled wheels are published for Linux and macOS on Python 3.10 – 3.13. No Rust toolchain is required for this path.

The distribution name on PyPI is canopus-quantum; the import name remains canopus:

import canopus

From source

git clone https://github.com/Youngcius/canopus.git
cd canopus
pip install .

pip install . runs the maturin build backend in an isolated environment, which:

1. Compiles the Rust extension (canopus.utils._accel) for the active Python version,
2. Packages the Python sources, and
3. Installs the resulting wheel into the current environment.

Prerequisite for source builds: a working Rust toolchain. Install via rustup:

curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

Usage

More end-to-end examples live under ./examples/:

• routing.ipynb — Canopus routing in detail
• rebasing.ipynb — Optimal ISA rebase (B gate, √iSWAP, arbitrary)
• python route_demo.py to test the routing effect by Sabre and Canopus on a demo circuit
• python route_qft.py <n> — Compare Canopus vs Sabre routing on an n-qubit QFT
• python rebase_xxx.py to test the rebase passes for arbitrary ISAs

Package layout

canopus/                            # Python package
├── __init__.py
├── backends.py                     # Backend / ISA / cost-estimator definitions
├── basics.py                       # CanonicalGate, BGate, SQiSWGate
├── mapping.py                      # CanopusMapping & SabreMapping transpiler passes
├── synthesis.py                    # rebase_to_{canonical, sqisw, zzphase, custom, ...}
├── decomposition/                  # Two-qubit decomposition kernels
├── extensions/                     # Optional integrations (bqskit, ...)
└── utils/
    ├── _core.py                    # Pure-Python utilities
    ├── _accel.cpython-*.so         # Rust-compiled accelerator (built by maturin)
    └── _accel.pyi                  # Type stubs for the Rust extension

src/                                # Rust source for the accelerator
└── lib.rs                          # PyO3 bindings

Development

Install the dev extras and build the Rust extension in editable mode:

pip install -e ".[dev]"
poe dev          # equivalent to: maturin develop --release

Common tasks (defined in pyproject.toml):

See CONTRIBUTING.md for the full contribution workflow.

Evaluation artifact

./experiments/ contains the full evaluation suite that accompanies the paper.

Case studies

• ./experiments/eval_qft/ — QFT kernel
• ./experiments/eval_qldpc/ — QLDPC stabilizer circuits

Benchmark suite

Evaluation commands are managed via ./experiments/Makefile. First prepare prerequisite files (coupling files, coverage sets, logical-level optimized circuits):

cd experiments && make

Then:

• make canopus — evaluate Canopus
• make baselines — evaluate baseline compilers (Sabre, TOQM, BQSKit)
• make sum_result — summarize results once routing evaluation completes
• make disp_result — display summarized routing overheads across compilers, topologies, and ISAs

For fine-grained evaluation, you may also run bench_all.py, bench_all_toqm.py, etc. directly.

Citation

If you use Canopus in your work, please cite:

@article{yang2025unifying,
  title   = {Unifying Qubit Routing Across Diverse Quantum ISAs via Canonical Representation},
  author  = {Yang, Zhaohui and Zhang, Kai and Tian, Xinyang and Ren, Xiangyu and
             Liu, Yingjian and Li, Yunfeng and Ding, Dawei and Chen, Jianxin and Xie, Yuan},
  journal = {arXiv preprint arXiv:2511.04608},
  year    = {2025}
}

License

This project is licensed under the Apache License 2.0 — see the LICENSE file for details.