hakorune/src/benchmarks.rs

/*!
 * Nyash Performance Benchmarks
 *
 * Compare execution performance across different backends:
 * - Interpreter (AST direct execution)
 * - VM (MIR -> VM execution)  
 * - WASM (MIR -> WASM execution)
 */

#[cfg(feature = "wasm-backend")]
use crate::backend::WasmBackend;
// VM import removed with vm-legacy
// Interpreter and MirCompiler removed
use std::fs;

#[derive(Debug)]
pub struct BenchmarkResult {
    pub name: String,
    pub backend: String,
    pub duration_ms: f64,
    pub iterations: u32,
    pub avg_duration_ms: f64,
}

pub struct BenchmarkSuite {
    iterations: u32,
}

impl BenchmarkSuite {
    pub fn new(iterations: u32) -> Self {
        Self { iterations }
    }

    /// Run comprehensive benchmark across all backends
    pub fn run_all(&self) -> Vec<BenchmarkResult> {
        #[cfg(feature = "wasm-backend")]
        let mut results = Vec::new();
        #[cfg(not(feature = "wasm-backend"))]
        let results = Vec::new();

        let benchmarks = [
            ("bench_light", "benchmarks/bench_light.hako"),
            ("bench_medium", "benchmarks/bench_medium.hako"),
            ("bench_heavy", "benchmarks/bench_heavy.hako"),
        ];

        for (name, file_path) in &benchmarks {
            println!("🚀 Running benchmark: {}", name);

            // Test if file exists and is readable
            if let Ok(_source) = fs::read_to_string(file_path) {
                // Run on all backends
                // Interpreter benchmark removed with legacy interpreter

                // VM benchmark removed with vm-legacy

                #[cfg(feature = "wasm-backend")]
                if let Ok(wasm_result) = self.run_wasm_benchmark(name, &_source) {
                    results.push(wasm_result);
                }
            } else {
                println!("⚠️ Could not read benchmark file: {}", file_path);
            }
        }

        results
    }

    // Interpreter benchmark removed with legacy interpreter

    // VM benchmark removed with vm-legacy
    #[allow(dead_code)]
    fn run_vm_benchmark(
        &self,
        _name: &str,
        _source: &str,
    ) -> Result<BenchmarkResult, Box<dyn std::error::Error>> {
        Err("VM benchmark removed with vm-legacy".into())
    }

    /// Run benchmark on WASM backend
    #[cfg(feature = "wasm-backend")]
    fn run_wasm_benchmark(
        &self,
        name: &str,
        source: &str,
    ) -> Result<BenchmarkResult, Box<dyn std::error::Error>> {
        let mut total_duration = 0.0;

        for i in 0..self.iterations {
            let start = Instant::now();

            // Parse -> MIR -> WASM
            let ast = NyashParser::parse_from_string(source)?;
            let mut compiler = MirCompiler::new();
            let compile_result = compiler.compile(ast)?;

            // Generate and execute WASM
            let mut wasm_backend = WasmBackend::new();
            let _wat_output = wasm_backend.compile_module(compile_result.module)?;
            // Note: For now we only measure compilation time
            // Full WASM execution would require wasmtime integration

            let duration = start.elapsed();
            total_duration += duration.as_secs_f64() * 1000.0; // Convert to ms

            if i == 0 {
                println!("  🌐 WASM: First run completed (compilation only)");
            }
        }

        Ok(BenchmarkResult {
            name: name.to_string(),
            backend: "WASM".to_string(),
            duration_ms: total_duration,
            iterations: self.iterations,
            avg_duration_ms: total_duration / (self.iterations as f64),
        })
    }

    /// Print benchmark results in a nice format
    pub fn print_results(&self, results: &[BenchmarkResult]) {
        println!("\n📊 Nyash Performance Benchmark Results");
        println!("=====================================");
        println!("Iterations per test: {}", self.iterations);
        println!();

        // Group by benchmark name
        let mut benchmarks: std::collections::HashMap<String, Vec<&BenchmarkResult>> =
            std::collections::HashMap::new();
        for result in results {
            benchmarks
                .entry(result.name.clone())
                .or_insert_with(Vec::new)
                .push(result);
        }

        for (bench_name, bench_results) in benchmarks {
            println!("🎯 {}", bench_name);
            println!("  Backend       | Avg Time (ms) | Total Time (ms) | Speed Ratio");
            println!("  --------------|---------------|-----------------|------------");

            // Find fastest for ratio calculation
            let fastest = bench_results
                .iter()
                .min_by(|a, b| a.avg_duration_ms.partial_cmp(&b.avg_duration_ms).unwrap())
                .unwrap();

            for result in &bench_results {
                let ratio = result.avg_duration_ms / fastest.avg_duration_ms;
                println!(
                    "  {:12} | {:11.3} | {:13.1} | {:8.2}x",
                    result.backend, result.avg_duration_ms, result.duration_ms, ratio
                );
            }
            println!();
        }

        // Summary
        println!("💡 Performance Summary:");
        let interpreter_avg: f64 = results
            .iter()
            .filter(|r| r.backend == "Interpreter")
            .map(|r| r.avg_duration_ms)
            .sum::<f64>()
            / results
                .iter()
                .filter(|r| r.backend == "Interpreter")
                .count() as f64;

        let vm_avg: f64 = results
            .iter()
            .filter(|r| r.backend == "VM")
            .map(|r| r.avg_duration_ms)
            .sum::<f64>()
            / results.iter().filter(|r| r.backend == "VM").count() as f64;

        let wasm_avg: f64 = results
            .iter()
            .filter(|r| r.backend == "WASM")
            .map(|r| r.avg_duration_ms)
            .sum::<f64>()
            / results.iter().filter(|r| r.backend == "WASM").count() as f64;

        println!("  📈 Average across all benchmarks:");
        println!("     Interpreter: {:.2} ms", interpreter_avg);
        println!(
            "     VM:          {:.2} ms ({:.1}x faster than interpreter)",
            vm_avg,
            interpreter_avg / vm_avg
        );
        println!(
            "     WASM:        {:.2} ms ({:.1}x faster than interpreter)",
            wasm_avg,
            interpreter_avg / wasm_avg
        );
    }
}

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

    /// Legacy benchmark smoke（Phase 15 以前の計測用）.
    /// 現在の CI / 開発フローでは必須ではないため、デフォルト実行から外す。
    #[test]
    #[ignore]
    fn test_benchmark_light() {
        let suite = BenchmarkSuite::new(3); // Only 3 iterations for testing
        let results = suite.run_all();

        // Should have results for all backends
        assert!(results.len() >= 3); // At least one benchmark with 3 backends

        suite.print_results(&results);
    }
}