scpn-quantum-engine 0.2.0

Rust acceleration engine for scpn-quantum-control (PyO3 + rayon)

scpn-quantum-engine

Rust acceleration engine for scpn-quantum-control, built with PyO3 and rayon.

The engine is an optional native extension. Every routine it exports has a pure-Python (NumPy/Qiskit) fallback in scpn-quantum-control, so the library runs correctly whether or not this extension is installed. When the extension is present, the Python dispatchers prefer the native path and fall back silently if a given export is unavailable.

Installation

The wheel is built from this directory with maturin:

# Build a release wheel
maturin build --release --out dist

# Or build and install into the active environment in one step
maturin develop --release

The native module imports as scpn_quantum_engine:

import scpn_quantum_engine as engine

engine.order_parameter(...)          # Kuramoto order parameter
engine.feedback_policy_batch(...)    # real-time feedback policy
engine.all_xy_expectations(...)      # transverse Pauli expectations

Accelerated routines

The extension exports native kernels grouped by subsystem; each mirrors a Python reference and is checked for parity in the scpn-quantum-control test suite (tests/test_rust_ffi_validation.py, tests/test_rust_new_functions.py):

• Kuramoto dynamics — Euler/trajectory integration, order parameter.
• Coupling graphs — K_nm construction, analog and hybrid coupling terms.
• Spectral and sector analysis — Koopman generator, magnetisation sectors, correlation matrices, OTOC, Lanczos coefficients, symmetry-decay fits.
• Pauli and Lindblad operators — fast expectation values, sparse order parameter, jump-operator assembly.
• Real-time feedback — batched feedback policy, sub-microsecond jitter tracking, closed-loop replay.
• Error mitigation and QEC — PEC coefficients and sampling, DLA protected subspace metrics.
• Cryptography and entropy — ML-DSA NTT/INTT, NIST SP 800-22 statistics (monobit, runs, longest-run, Berlekamp–Massey).
• Compiler autodiff primitives — value/JVP/VJP kernels for the linear algebra used by the differentiable compiler.

The canonical, generated catalogue is published at https://anulum.github.io/scpn-quantum-control/rust_engine/.

Parity and fallback contract

The native kernels are bit-for-bit or tolerance-matched against their Python references. The Python side never requires the extension: a missing build, a missing export, or a native error all resolve to the NumPy/Qiskit fallback.

Licence

AGPL-3.0-or-later; a commercial licence is available. See the repository root for licence terms and contact details.