quantrs2-sim 0.1.3

//! Benchmarking utilities for quantum simulators
//!
//! This module provides utilities for benchmarking different simulator
//! implementations and comparing their performance.

use quantrs2_circuit::builder::{Circuit, Simulator};
use quantrs2_core::{error::QuantRS2Result, qubit::QubitId};
use scirs2_core::random::ChaCha8Rng;
use scirs2_core::random::{RngExt, SeedableRng};
use std::time::{Duration, Instant};

use crate::optimized_simulator::OptimizedSimulator;
use crate::optimized_simulator_chunked::OptimizedSimulatorChunked;
use crate::optimized_simulator_simple::OptimizedSimulatorSimple;
use crate::statevector::StateVectorSimulator;

/// Benchmark results for a quantum simulator
#[derive(Debug, Clone)]
pub struct BenchmarkResult {
    /// Name of the simulator
    pub simulator_name: String,
    /// Number of qubits
    pub num_qubits: usize,
    /// Total number of gates
    pub num_gates: usize,
    /// Execution time
    pub duration: Duration,
    /// Additional notes or context
    pub notes: Option<String>,
    /// Peak memory usage (if available)
    pub peak_memory: Option<usize>,
}

impl BenchmarkResult {
    /// Create a new benchmark result
    #[must_use]
    pub fn new(
        simulator_name: &str,
        num_qubits: usize,
        num_gates: usize,
        duration: Duration,
        notes: Option<String>,
        peak_memory: Option<usize>,
    ) -> Self {
        Self {
            simulator_name: simulator_name.to_string(),
            num_qubits,
            num_gates,
            duration,
            notes,
            peak_memory,
        }
    }

    /// Format the result as a string
    #[must_use]
    pub fn format(&self) -> String {
        let duration_ms = self.duration.as_millis();
        let rate = if duration_ms > 0 {
            self.num_gates as f64 / (duration_ms as f64 / 1000.0)
        } else {
            f64::INFINITY
        };

        let notes = if let Some(ref notes) = self.notes {
            format!(" ({notes})")
        } else {
            String::new()
        };

        let memory_str = if let Some(mem) = self.peak_memory {
            format!(", Memory: {:.2} MB", mem as f64 / (1024.0 * 1024.0))
        } else {
            String::new()
        };

        format!(
            "Simulator: {}{}\n  Qubits: {}\n  Gates: {}\n  Time: {} ms\n  Rate: {:.2} gates/s{}",
            self.simulator_name,
            notes,
            self.num_qubits,
            self.num_gates,
            duration_ms,
            rate,
            memory_str
        )
    }
}

/// Run a benchmark on a simulator with a given circuit
pub fn run_benchmark<S, const N: usize>(
    name: &str,
    simulator: &S,
    circuit: &Circuit<N>,
    notes: Option<String>,
) -> QuantRS2Result<BenchmarkResult>
where
    S: Simulator<N>,
{
    let num_qubits = N;
    let num_gates = circuit.gates().len();

    let start = Instant::now();
    let _result = simulator.run(circuit)?;
    let duration = start.elapsed();

    Ok(BenchmarkResult::new(
        name, num_qubits, num_gates, duration, notes, None,
    ))
}

/// Compare different simulator implementations on the same circuit
#[must_use]
pub fn compare_simulators<const N: usize>(circuit: &Circuit<N>) -> Vec<BenchmarkResult> {
    let mut results = Vec::new();

    // For very large circuits, only test optimized implementations
    if N <= 24 {
        // Run standard simulator
        let standard_sim = StateVectorSimulator::new();
        if let Ok(result) = run_benchmark("Standard Simulator", &standard_sim, circuit, None) {
            results.push(result);
        }
    }

    // Run simple optimized simulator
    let simple_opt_sim = OptimizedSimulatorSimple::new();
    if let Ok(result) = run_benchmark("Simple Optimized", &simple_opt_sim, circuit, None) {
        results.push(result);
    }

    // Run chunked optimized simulator
    let chunked_sim = OptimizedSimulatorChunked::new();
    if let Ok(result) = run_benchmark("Chunked Optimized", &chunked_sim, circuit, None) {
        results.push(result);
    }

    // Run full optimized simulator with automatic selection
    let optimized_sim = OptimizedSimulator::new();
    if let Ok(result) = run_benchmark(
        "Full Optimized",
        &optimized_sim,
        circuit,
        Some("Auto-selection".to_string()),
    ) {
        results.push(result);
    }

    // For smaller circuits, also test memory-efficient configuration
    if N <= 20 {
        let memory_efficient_sim = OptimizedSimulator::memory_efficient();
        if let Ok(result) = run_benchmark(
            "Memory Efficient",
            &memory_efficient_sim,
            circuit,
            Some("Low memory".to_string()),
        ) {
            results.push(result);
        }
    }

    results
}

/// Generate a random quantum circuit for benchmarking
#[must_use]
pub fn generate_benchmark_circuit<const N: usize>(
    num_gates: usize,
    two_qubit_ratio: f64,
) -> Circuit<N> {
    let mut circuit = Circuit::new();
    let mut rng = ChaCha8Rng::seed_from_u64(42); // Use fixed seed for reproducibility

    for _ in 0..num_gates {
        // Decide if this is a single-qubit or two-qubit gate
        let is_two_qubit = rng.random::<f64>() < two_qubit_ratio;

        if is_two_qubit && N > 1 {
            // Select two different qubits
            let qubit1 = QubitId::new(rng.random_range(0..N as u32));
            let mut qubit2 = QubitId::new(rng.random_range(0..N as u32));
            while qubit2 == qubit1 {
                qubit2 = QubitId::new(rng.random_range(0..N as u32));
            }

            // Choose a two-qubit gate
            let gate_type = rng.random_range(0..3);
            match gate_type {
                0 => {
                    let _ = circuit.cnot(qubit1, qubit2);
                }
                1 => {
                    let _ = circuit.cz(qubit1, qubit2);
                }
                _ => {
                    let _ = circuit.swap(qubit1, qubit2);
                }
            }
        } else {
            // Select a qubit
            let qubit = QubitId::new(rng.random_range(0..N as u32));

            // Choose a single-qubit gate
            let gate_type = rng.random_range(0..7);
            match gate_type {
                0 => {
                    let _ = circuit.h(qubit);
                }
                1 => {
                    let _ = circuit.x(qubit);
                }
                2 => {
                    let _ = circuit.y(qubit);
                }
                3 => {
                    let _ = circuit.z(qubit);
                }
                4 => {
                    let _ = circuit.s(qubit);
                }
                5 => {
                    let _ = circuit.t(qubit);
                }
                _ => {
                    // Random rotation
                    let angle = rng.random_range(0.0..std::f64::consts::TAU);
                    let rotation_type = rng.random_range(0..3);
                    match rotation_type {
                        0 => {
                            let _ = circuit.rx(qubit, angle);
                        }
                        1 => {
                            let _ = circuit.ry(qubit, angle);
                        }
                        _ => {
                            let _ = circuit.rz(qubit, angle);
                        }
                    }
                }
            }
        }
    }

    circuit
}

/// Run a comprehensive benchmark suite for circuits of different sizes
pub fn run_benchmark_suite() {
    println!("=== Quantrs Simulator Benchmark Suite ===");
    println!("Testing various circuit sizes with different simulator implementations");
    println!();

    // Test small circuits (up to 16 qubits)
    benchmark_circuit_size::<16>("Small", 16, 100, 0.3);

    // Test medium circuits (up to 20 qubits)
    benchmark_circuit_size::<20>("Medium", 20, 50, 0.3);

    // Test large circuits (up to 25 qubits)
    benchmark_circuit_size::<25>("Large", 25, 20, 0.2);

    // Test very large circuits for memory-efficient implementation
    benchmark_large_circuit::<28>("Very Large", 28, 10, 0.1);
}

// Helper function to benchmark a specific circuit size
fn benchmark_circuit_size<const N: usize>(
    size_name: &str,
    max_qubits: usize,
    gates_per_qubit: usize,
    two_qubit_ratio: f64,
) {
    println!("\n=== {size_name} Circuit Tests (up to {max_qubits} qubits) ===");

    // Only proceed if the template parameter is big enough
    if N < max_qubits {
        println!("Cannot benchmark this size - const generic N is too small");
        return;
    }

    // Benchmark increasing qubit counts
    for qubits in [
        max_qubits / 4,
        max_qubits / 2,
        (3 * max_qubits) / 4,
        max_qubits,
    ] {
        let num_gates = qubits * gates_per_qubit;
        println!("\nCircuit with {qubits} qubits and {num_gates} gates:");

        // Generate the circuit
        let circuit = generate_benchmark_circuit::<N>(num_gates, two_qubit_ratio);

        // Run all benchmarks
        let results = compare_simulators(&circuit);

        // Print results
        for result in &results {
            println!("{}\n", result.format());
        }
    }
}

// Special benchmark for very large circuits
fn benchmark_large_circuit<const N: usize>(
    size_name: &str,
    max_qubits: usize,
    gates_per_qubit: usize,
    two_qubit_ratio: f64,
) {
    println!("\n=== {size_name} Circuit Tests ({max_qubits} qubits) ===");

    // Only proceed if the template parameter is big enough
    if N < max_qubits {
        println!("Cannot benchmark this size - const generic N is too small");
        return;
    }

    let num_gates = max_qubits * gates_per_qubit;
    println!("\nCircuit with {max_qubits} qubits and {num_gates} gates:");

    // Generate the circuit
    let circuit = generate_benchmark_circuit::<N>(num_gates, two_qubit_ratio);

    // Only run with chunked and full optimized simulators
    let mut results = Vec::new();

    // Run chunked optimized simulator
    let chunked_sim = OptimizedSimulatorChunked::new();
    if let Ok(result) = run_benchmark("Chunked Optimized", &chunked_sim, &circuit, None) {
        results.push(result);
    }

    // Run full optimized simulator
    let optimized_sim = OptimizedSimulator::new();
    if let Ok(result) = run_benchmark(
        "Full Optimized",
        &optimized_sim,
        &circuit,
        Some("Auto-selection".to_string()),
    ) {
        results.push(result);
    }

    // Print results
    for result in &results {
        println!("{}\n", result.format());
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    #[ignore] // Run only when explicitly requested, not during regular testing
    fn benchmark_small_circuit() {
        const QUBITS: usize = 16;
        let circuit = generate_benchmark_circuit::<QUBITS>(100, 0.3);

        let results = compare_simulators(&circuit);

        for result in &results {
            println!("{}\n", result.format());
        }
    }

    #[test]
    #[ignore] // Run only when explicitly requested, not during regular testing
    fn benchmark_medium_circuit() {
        const QUBITS: usize = 20;
        let circuit = generate_benchmark_circuit::<QUBITS>(50, 0.3);

        let results = compare_simulators(&circuit);

        for result in &results {
            println!("{}\n", result.format());
        }
    }

    #[test]
    #[ignore] // Run only when explicitly requested, not during regular testing
    fn benchmark_large_circuit() {
        const QUBITS: usize = 25;
        let circuit = generate_benchmark_circuit::<QUBITS>(20, 0.2);

        // For large circuits, only compare the optimized implementations
        let mut results = Vec::new();

        // Run optimized simulators
        let chunked_sim = OptimizedSimulatorChunked::new();
        if let Ok(result) = run_benchmark("Chunked Optimized", &chunked_sim, &circuit, None) {
            results.push(result);
        }

        let optimized_sim = OptimizedSimulator::new();
        if let Ok(result) = run_benchmark(
            "Full Optimized",
            &optimized_sim,
            &circuit,
            Some("Auto-selection".to_string()),
        ) {
            results.push(result);
        }

        for result in &results {
            println!("{}\n", result.format());
        }
    }

    #[test]
    #[ignore] // Run only when explicitly requested, not during regular testing
    fn run_full_benchmark_suite() {
        run_benchmark_suite();
    }
}