ruvector/crates/ruqu-core

ruqu-core

Quantum Execution Intelligence Engine in pure Rust — 5 simulation backends with automatic routing, noise models, error mitigation, OpenQASM 3.0 export, and cryptographic witness logging.

Features

• 5 Simulation Backends — StateVector (exact, up to 32 qubits), Stabilizer (millions of qubits), Clifford+T (moderate T-count), TensorNetwork (MPS-based), Hardware (device profiles)
• Cost-Model Planner — Automatically routes circuits to the optimal backend based on qubit count, gate mix, and T-count
• Universal Gate Set — H, X, Y, Z, CNOT, CZ, Toffoli, Rx, Ry, Rz, Phase, SWAP, and custom unitaries
• QEC Control Plane — Union-find decoder with O(n*a(n)) amortized time, sub-polynomial decoders, QEC scheduling, control theory integration
• OpenQASM 3.0 — Full circuit export to standard quantum assembly format
• Noise & Mitigation — Depolarizing, amplitude/phase damping, custom Kraus operators, zero-noise extrapolation, probabilistic error cancellation
• SIMD Acceleration — AVX2/NEON vectorized gate application for 2-4x speedup
• Multi-Threading — Rayon-based parallelism for large qubit counts
• Cryptographic Witnesses — Tamper-evident execution logs for reproducibility and verification
• Transpiler — Gate decomposition, routing, and hardware-aware optimization
• Mixed Precision — Configurable f32/f64 simulation for speed vs accuracy tradeoff

Installation

cargo add ruqu-core

With optional features:

cargo add ruqu-core --features parallel,simd

Quick Start

use ruqu_core::prelude::*;

// Create a Bell state |00> + |11>
let mut circuit = QuantumCircuit::new(2);
circuit.h(0).cnot(0, 1);

let result = Simulator::run(&circuit)?;
let probs = result.state.probabilities();
// probs ~= [0.5, 0.0, 0.0, 0.5]

Simulation Backends

Backend	Qubits	Best For
StateVector	Up to 32	Exact simulation, small circuits
Stabilizer	Millions	Clifford-only circuits (Gottesman-Knill)
Clifford+T	Moderate	Circuits with low T-count
TensorNetwork	Variable	Shallow/structured circuits (MPS)
Hardware	Device-dependent	Real device profiles and constraints

The cost-model planner automatically selects the best backend:

use ruqu_core::planner::CostModelPlanner;

let planner = CostModelPlanner::new();
let backend = planner.select(&circuit); // Auto-routes to optimal backend

OpenQASM 3.0 Export

use ruqu_core::qasm::to_qasm3;

let qasm = to_qasm3(&circuit);
println!("{}", qasm);
// OPENQASM 3.0;
// qubit[2] q;
// h q[0];
// cx q[0], q[1];

Quantum Gates

Gate	Description	Matrix
H	Hadamard	Creates superposition
X	Pauli-X (NOT)	Bit flip
Y	Pauli-Y	Bit + phase flip
Z	Pauli-Z	Phase flip
CNOT	Controlled-NOT	Two-qubit entanglement
CZ	Controlled-Z	Controlled phase
Rx(θ)	X-rotation	Rotate around X-axis
Ry(θ)	Y-rotation	Rotate around Y-axis
Rz(θ)	Z-rotation	Rotate around Z-axis
SWAP	Swap qubits	Exchange qubit states
Toffoli	CCX	Three-qubit AND gate

Performance

Benchmarks on Apple M2 (single-threaded):

Qubits	Gates	Time
10	100	0.3ms
15	100	8ms
20	100	250ms
25	100	8s

With --features parallel on 8 cores, 20+ qubits see 3-5x speedup.

Noise Simulation

use ruqu_core::noise::{NoiseModel, Depolarizing};

let noise = NoiseModel::new()
    .add_single_qubit(Depolarizing::new(0.01))  // 1% error rate
    .add_two_qubit(Depolarizing::new(0.02));    // 2% for CNOT

let noisy_state = simulator.run_noisy(&circuit, &noise)?;

Related Crates

• ruqu-algorithms — VQE, Grover, QAOA, Surface Code
• ruqu-exotic — Quantum-classical hybrid algorithms
• ruqu-wasm — WebAssembly bindings

Architecture

Part of the RuVector quantum ecosystem. See ADR-QE-001 for core architecture and ADR-QE-015 for the execution intelligence engine design.

License

MIT OR Apache-2.0