#pragma once
#include <condition_variable>
#include <functional>
#include <mutex>
#include <queue>
#include <thread>
#include <vector>
#include <cstdint>

// Fixed-size thread pool for parallel operations.
// Minimizes thread creation overhead for batch operations.
class ThreadPool {
    std::vector<std::thread> workers_;
    std::queue<std::function<void()>> tasks_;
    std::mutex queue_mutex_;
    std::condition_variable condition_;
    bool stop_ = false;

public:
    explicit ThreadPool(size_t num_threads);
    ~ThreadPool();
    
    // Add task to queue. Thread-safe.
    void enqueue(std::function<void()> task);
    
    // Wait for all queued tasks to complete
    void wait_all();
    
    // Get optimal thread count (capped at 8 for I/O bound work)
    static uint32_t default_thread_count();
};

// Synchronization primitive: wait for N completions
class CompletionLatch {
    std::atomic<size_t> count_{0};
    std::mutex mutex_;
    std::condition_variable cv_;
    
public:
    explicit CompletionLatch(size_t total) : count_(total) {}
    
    void arrive() {
        if (--count_ == 0) {
            std::lock_guard<std::mutex> lock(mutex_);
            cv_.notify_all();
        }
    }
    
    void wait() {
        std::unique_lock<std::mutex> lock(mutex_);
        cv_.wait(lock, [&] { return count_.load() == 0; });
    }
};