#include <pthread.h>
#include <atomic>
#include <chrono>
#include <algorithm>
#include "row.hpp"
#include "sort_algorithms.hpp"

struct MatrixSortContext {
    uint8_t* A;
    int F, H, W;
};

struct ChunkArgs {
    MatrixSortContext* ctx;
    std::atomic<int>* next_row;
    int total_rows;
    int fixed_chunk;
};

void* chunk_worker(void* arg) {
    ChunkArgs* ca = (ChunkArgs*)arg;
    int start;

    while ((start = ca->next_row->fetch_add(ca->fixed_chunk)) < ca->total_rows) {
        int end = std::min(start + ca->fixed_chunk, ca->total_rows);
        for (int i = start; i < end; ++i) {
            uint8_t* row_ptr = ca->ctx->A + (static_cast<size_t>(i) * ca->ctx->W);
            row<uint8_t> r(row_ptr, (uint32_t)ca->ctx->W, false);
            quick_sort(r);
        }
    }
    return nullptr;
}

long long run_chunk_scheduler(MatrixSortContext* ctx, int K, int chunk_size) {
    int total_rows = ctx->F * ctx->H;
    std::atomic<int> next_row(0);
    pthread_t threads[K];
    ChunkArgs args[K];

    auto start_bench = std::chrono::high_resolution_clock::now();
    for (int i = 0; i < K; ++i) {
        args[i] = {ctx, &next_row, total_rows, chunk_size};
        pthread_create(&threads[i], nullptr, chunk_worker, &args[i]);
    }
    for (int i = 0; i < K; ++i) pthread_join(threads[i], nullptr);
    auto end_bench = std::chrono::high_resolution_clock::now();

    return std::chrono::duration_cast<std::chrono::milliseconds>(end_bench - start_bench).count();
}