fpm-qsim 0.2.0

C++ accelerated FPM quantum simulation primitives with Python bindings

fpm-qsim

C++ accelerated FPM quantum-simulation primitives exposed as a Python extension module.

Install from PyPI:

pip install fpm-qsim

Use it from Python:

import fpm_cpp as fpm

Layout

.
├── src/
│   ├── fpm_core.hpp          # Header-only C++17 FPM core
│   └── fpm_cpp_bindings.cpp  # pybind11 extension bindings
├── scripts/
│   └── build.sh              # Manual Linux/macOS/WSL build helper
├── tests/
│   ├── smoke_test.py         # Fast installed-wheel smoke test
│   └── equivalence_test.py   # Reference equivalence checks
├── pyproject.toml            # PyPI build metadata
├── setup.py                  # Extension build configuration
├── MANIFEST.in               # Source distribution contents
└── README.md

Example

import numpy as np
import fpm_cpp as fpm

rho0 = np.array([[0.5, 0.5], [0.5, 0.5]], dtype=np.complex128)

rho1 = fpm.lindblad_step(
    rho0,
    gamma=0.1,
    dt=1.0,
    method="exact",
    use_omp=False,
)

traj = fpm.simulate(
    rho0,
    gamma=0.05,
    dt=1.0,
    n_steps=1000,
    method="exact",
    use_omp=True,
)

fpm.bounded_gamma(29.3)

Build From Source

PyPI installs prebuilt wheels when available. If no wheel matches your platform, pip builds from source and requires:

• Python 3.9+
• NumPy
• pybind11
• A C++17 compiler

Manual local build:

python -m pip install pybind11 numpy
./scripts/build.sh
python tests/smoke_test.py

On Windows, normal pip install fpm-qsim builds with MSVC when a matching wheel is unavailable.

API Surface

The extension module is fpm_cpp.

Core functions:

• lindblad_step(rho, gamma, dt=1.0, method="exact", use_omp=True)
• simulate(rho0, gamma, dt=1.0, n_steps=1, method="exact", use_omp=True)
• kappa_from_gamma(gamma, dt=1.0)
• kappa_exact(gamma, dt=1.0)
• gamma_from_kappa(kappa, dt=1.0)
• bounded_gamma(gamma_raw, gamma_max=GAMMA_MAX)
• gamma_from_energy(daemon, gate_power, load=None, dt=1.0, C_N=1.0, bounded=True)

State and accounting:

• DaemonState
• ConservationLedger
• exp_route_cost(taylor_order=8)
• bill_exp_route_cost(ledger, daemon, taylor_order=8, cost_per_op=1e-5)

Constants:

• GAMMA_MAX
• FALSIFICATION_THRESHOLD
• ENERGY_FLOOR_FRACTION
• ISOTROPIC_WEIGHT_LIMIT

Verification

After installing or building locally:

python tests/smoke_test.py

tests/equivalence_test.py is retained for reference comparison workflows that also have the old Python implementation available.

License

MIT