#include <iostream>
#include <vector>
#include <iomanip>
#include <algorithm>
#include <random>
#include <pthread.h>
#include <sys/time.h>
#include <unistd.h>
#include <sstream>

constexpr int NIL = -1;

struct Node {
    int value;
    int next;
};

struct SortResult {
    int head;
    long long comparisons;
};

struct Args {
    Node* pool;
    int head;
    int depth;
    int size;
    int threshold;
    SortResult* res_out;
};

int active_threads = 0;
int max_threads = 4;
pthread_mutex_t counter_mutex = PTHREAD_MUTEX_INITIALIZER;
pthread_mutex_t log_mutex = PTHREAD_MUTEX_INITIALIZER;
bool enable_logging = false;

// Высокоточное логирование для exporter.py
void log_event(const std::string& type, int head, int depth) {
    if (!enable_logging) return;
    pthread_mutex_lock(&log_mutex);

    timeval tv{};
    gettimeofday(&tv, nullptr);
    double current_time = tv.tv_sec + tv.tv_usec * 1e-6;

    std::cout << type << " TID=" << pthread_self()
              << " depth=" << depth
              << " range=[" << head << "," << head << "]"
              << " time=" << std::fixed << std::setprecision(6) << current_time << std::endl;

    pthread_mutex_unlock(&log_mutex);
}

void split_list(Node* pool, int head, int& left, int& right) {
    if (head == NIL || pool[head].next == NIL) {
        left = head; right = NIL; return;
    }
    int slow = head, fast = pool[head].next;
    while (fast != NIL) {
        fast = pool[fast].next;
        if (fast != NIL) { slow = pool[slow].next; fast = pool[fast].next; }
    }
    left = head;
    right = pool[slow].next;
    pool[slow].next = NIL;
}

SortResult merge_lists(Node* pool, int l, int r) {
    long long comps = 0;
    if (l == NIL) return {r, 0};
    if (r == NIL) return {l, 0};
    int res_head = NIL, tail = NIL;
    auto append = [&](int idx) {
        if (res_head == NIL) res_head = tail = idx;
        else { pool[tail].next = idx; tail = idx; }
    };
    while (l != NIL && r != NIL) {
        comps++;
        if (pool[l].value <= pool[r].value) { int n = pool[l].next; append(l); l = n; }
        else { int n = pool[r].next; append(r); r = n; }
    }
    if (l != NIL) pool[tail].next = l;
    if (r != NIL) pool[tail].next = r;
    return {res_head, comps};
}

SortResult parallel_list_sort(Node* pool, int head, int depth, int size, int threshold);

void* thread_func(void* arg) {
    Args* a = (Args*)arg;
    *(a->res_out) = parallel_list_sort(a->pool, a->head, a->depth, a->size, a->threshold);
    return nullptr;
}

SortResult parallel_list_sort(Node* pool, int head, int depth, int size, int threshold) {
    log_event("START", head, depth);
    if (head == NIL || pool[head].next == NIL) {
        log_event("END", head, depth);
        return {head, 0};
    }

    int left_p, right_p;
    split_list(pool, head, left_p, right_p);
    int new_size = size / 2;

    pthread_t tid;
    bool spawned = false;
    SortResult res_right = {NIL, 0};

    if (size > threshold) {
        pthread_mutex_lock(&counter_mutex);
        if (active_threads < max_threads) {
            active_threads++;
            spawned = true;
        }
        pthread_mutex_unlock(&counter_mutex);
    }

    if (spawned) {
        Args* args = new Args{pool, right_p, depth + 1, new_size, threshold, &res_right};
        pthread_create(&tid, nullptr, thread_func, args);
        SortResult res_left = parallel_list_sort(pool, left_p, depth + 1, new_size, threshold);
        pthread_join(tid, nullptr);

        pthread_mutex_lock(&counter_mutex);
        active_threads--;
        pthread_mutex_unlock(&counter_mutex);

        SortResult m = merge_lists(pool, res_left.head, res_right.head);
        delete args;
        log_event("END", m.head, depth);
        return {m.head, res_left.comparisons + res_right.comparisons + m.comparisons};
    } else {
        SortResult res_left = parallel_list_sort(pool, left_p, depth + 1, new_size, threshold);
        SortResult res_r_seq = parallel_list_sort(pool, right_p, depth + 1, new_size, threshold);
        SortResult m = merge_lists(pool, res_left.head, res_r_seq.head);
        log_event("END", m.head, depth);
        return {m.head, res_left.comparisons + res_r_seq.comparisons + m.comparisons};
    }
}

int main(int argc, char* argv[]) {
    int n = 100000;
    int threshold = 5000;
    if (argc >= 2) n = std::atoi(argv[1]);
    if (argc >= 3) max_threads = std::atoi(argv[2]);
    if (argc >= 4) threshold = std::atoi(argv[3]);

    // Включаем логи только для маленьких N
    if (n < 5000) enable_logging = true;

    std::vector<Node> pool(n);
    std::mt19937 rng(1337);
    for (int i = 0; i < n; i++) {
        pool[i].value = rng() % 100000;
        pool[i].next = (i == n - 1) ? NIL : i + 1;
    }

    timeval t1, t2;
    gettimeofday(&t1, nullptr);
    SortResult final_res = parallel_list_sort(pool.data(), 0, 0, n, threshold);
    gettimeofday(&t2, nullptr);

    double elapsed = (t2.tv_sec - t1.tv_sec) + (t2.tv_usec - t1.tv_usec) * 1e-6;
    std::cerr << "STAT: threads=" << max_threads << " size=" << n << " threshold=" << threshold << " time=" << elapsed << std::endl;

    return 0;
}