trueno 0.17.4

use super::super::super::*;

// ========================================================================
// Phase 11: High-Performance Profiling Patterns (E.9) - F150-F155
// ========================================================================

/// F150: RDTSCP overhead < 15ns
#[test]
#[ignore = "Environment-dependent: timing varies on CI runners under load"]
fn test_f150_cpu_cycles_overhead() {
    // Warm up
    for _ in 0..100 {
        let _ = cpu_cycles();
    }

    // Measure overhead
    let start = std::time::Instant::now();
    for _ in 0..10000 {
        let _ = cpu_cycles();
    }
    let elapsed = start.elapsed();
    let avg_ns = elapsed.as_nanos() as f64 / 10000.0;

    // Should be < 15ns on most platforms
    // On unsupported platforms, cpu_cycles() returns 0 and is essentially free
    assert!(avg_ns < 50.0, "cpu_cycles() overhead should be < 50ns, got {:.1}ns", avg_ns);
}

/// F151: Cycle count monotonic
#[test]
fn test_f151_cpu_cycles_monotonic() {
    let c1 = cpu_cycles();
    // Do some work
    let mut sum = 0u64;
    for i in 0..1000 {
        sum = sum.wrapping_add(i);
    }
    let _ = sum; // Prevent optimization
    let c2 = cpu_cycles();

    // On platforms that support cycle counting, should be monotonic
    // On unsupported platforms, both will be 0
    assert!(c2 >= c1, "Cycle count should be monotonic: {} >= {}", c2, c1);
}

/// F152: Cached time precision < 200us drift
#[test]
fn test_f152_cached_time_precision() {
    // Initialize time service
    init_time_service();

    // Wait for it to warm up
    std::thread::sleep(std::time::Duration::from_millis(2));

    // Compare cached vs actual using Instant::now() as reference
    let cached = cached_nanos();
    let reference_start = std::time::Instant::now();
    std::thread::sleep(std::time::Duration::from_micros(100));
    let cached_after = cached_nanos();
    let elapsed_real = reference_start.elapsed().as_nanos() as u64;

    if cached > 0 && cached_after > 0 {
        let cached_elapsed = cached_after.saturating_sub(cached);
        let drift = elapsed_real.abs_diff(cached_elapsed);

        // Should be within 2ms (2_000_000ns)
        // CI runners (especially macOS shared) can have scheduling jitter
        assert!(
            drift < 2_000_000, // 2ms tolerance for CI stability
            "Cached time drift should be < 2ms, got {}us",
            drift / 1000
        );
    }
}

/// F153: Cached time overhead < 2ns
#[test]
#[ignore = "Environment-dependent: timing varies on CI runners under load"]
fn test_f153_cached_time_overhead() {
    // Initialize time service
    init_time_service();
    std::thread::sleep(std::time::Duration::from_millis(1));

    // Warm up
    for _ in 0..100 {
        let _ = cached_nanos();
    }

    // Measure overhead
    let start = std::time::Instant::now();
    for _ in 0..100000 {
        let _ = cached_nanos();
    }
    let elapsed = start.elapsed();
    let avg_ns = elapsed.as_nanos() as f64 / 100000.0;

    // Should be very fast (atomic load)
    assert!(avg_ns < 20.0, "cached_nanos() overhead should be < 20ns, got {:.1}ns", avg_ns);
}

/// F154: Poll count accuracy
#[test]
fn test_f154_poll_count_accuracy() {
    let mut profiler = AsyncTaskProfiler::new("test_task");

    // Simulate 5 polls with 3 yields
    for i in 0..5 {
        profiler.on_poll_start();
        let is_ready = i == 4; // Ready on last poll
        profiler.on_poll_end(is_ready);
    }

    assert_eq!(profiler.poll_count, 5, "Should have 5 polls");
    assert_eq!(profiler.yield_count, 4, "Should have 4 yields (Pending)");
    assert!((profiler.efficiency() - 0.2).abs() < 0.01, "Efficiency should be 1/5 = 0.2");
    assert!((profiler.yield_ratio() - 0.8).abs() < 0.01, "Yield ratio should be 4/5 = 0.8");
}

/// F155: Page fault detection (Linux only)
#[test]
fn test_f155_page_fault_detection() {
    // Get initial page fault count
    let (minor1, major1) = get_page_faults();

    // Do something that might cause page faults
    let v: Vec<u8> = vec![0u8; 4096 * 10]; // Allocate 10 pages
    let _ = v.iter().sum::<u8>(); // Touch pages

    let (minor2, major2) = get_page_faults();

    // On Linux, we should see page faults
    // On other platforms, both will be 0
    #[cfg(target_os = "linux")]
    {
        // Should have at least some minor faults from allocation
        assert!(minor2 >= minor1, "Minor faults should not decrease: {} >= {}", minor2, minor1);
    }

    // Major faults should be rare (no swapping in this test)
    assert!(major2 - major1 < 10, "Should have minimal major faults: {} - {} < 10", major2, major1);
}

/// F150+: BrickStats cycle tracking
#[test]
fn test_brick_stats_cycle_tracking() {
    let mut stats = BrickStats::new("test_brick");

    // Add samples with cycles
    stats.add_sample_with_cycles(1000, 100, 3000); // 1us, 100 elem, 3000 cycles
    stats.add_sample_with_cycles(2000, 200, 6000); // 2us, 200 elem, 6000 cycles

    assert_eq!(stats.total_cycles, 9000);
    assert_eq!(stats.min_cycles, 3000);
    assert_eq!(stats.max_cycles, 6000);
    assert!((stats.cycles_per_element() - 30.0).abs() < 0.1); // 9000/300 = 30
    assert!((stats.avg_cycles() - 4500.0).abs() < 0.1); // 9000/2 = 4500

    // IPC should be elements/cycles = 300/9000 = 0.033
    let ipc = stats.estimated_ipc();
    assert!(ipc > 0.0 && ipc < 1.0, "IPC should be low (memory bound)");

    let diagnosis = stats.diagnose_from_cycles();
    assert!(
        diagnosis.contains("memory") || diagnosis.contains("insufficient"),
        "Low IPC should indicate memory bound"
    );
}

/// F150+: AsyncTaskProfiler ExecutionNode conversion
#[test]
fn test_async_task_profiler_to_execution_node() {
    let mut profiler = AsyncTaskProfiler::new("request_handler");
    profiler.poll_count = 3;
    profiler.yield_count = 2;
    profiler.total_poll_ns = 1500;

    let node = profiler.to_execution_node();

    if let ExecutionNode::AsyncTask { name, poll_count, yield_count, total_poll_ns } = node {
        assert_eq!(name, "request_handler");
        assert_eq!(poll_count, 3);
        assert_eq!(yield_count, 2);
        assert_eq!(total_poll_ns, 1500);
    } else {
        panic!("Expected AsyncTask node");
    }
}

/// F150+: ExecutionGraph with AsyncTask node
#[test]
fn test_execution_graph_async_task() {
    let mut graph = ExecutionGraph::new();

    graph.add_node(ExecutionNode::AsyncTask {
        name: "inference".into(),
        poll_count: 5,
        yield_count: 4,
        total_poll_ns: 2500,
    });

    // Test ASCII tree
    let tree = graph.to_ascii_tree();
    assert!(tree.contains("inference"), "Should contain task name");
    assert!(tree.contains("polls:5"), "Should contain poll count");

    // Test DOT export
    let dot = graph.to_dot();
    assert!(dot.contains("inference"), "DOT should contain task name");
    assert!(dot.contains("lightcyan"), "AsyncTask should have cyan color");
}

/// F150+: with_page_fault_tracking helper
#[test]
fn test_with_page_fault_tracking() {
    let (result, minor, major) = with_page_fault_tracking("test_alloc", || {
        let v: Vec<u8> = vec![42u8; 100];
        v.len() // Just return the length instead of summing
    });

    assert_eq!(result, 100);
    // Just verify it doesn't panic and returns reasonable values
    assert!(minor < 1_000_000, "Minor faults should be bounded");
    assert!(major < 100, "Major faults should be minimal");
}