/**
 * \file
 * \brief   PDS HW2 tests
 *
 * To run these test execute:
 *      make tests
 *      mpirun -np <N> ./out/tests
 *
 * Note:
 *  Yes each process runs the entire test suite!!
 *
 * \author
 *    Christos Choutouridis AEM:8997
 *    <cchoutou@ece.auth.gr>
 */

#include <gtest/gtest.h>
#include <mpi.h>
#include <random>
#include "distsort.hpp"
/*
 * Global fixtures
 */

// MPI handler for the test session
MPI_t<>         ts_mpi;

// Mersenne seeded from hw if possible. range: [type_min, type_max]
std::random_device rd;
std::mt19937 gen(rd());

class TMPIdistSort : public ::testing::Test {
protected:
    static void SetUpTestSuite() {
        int argc = 0;
        char** argv = nullptr;
        ts_mpi.init(&argc, &argv);
    }

    static void TearDownTestSuite() {
        ts_mpi.finalize();
    }
};


/*
 * MPI: SysTest (acceptance)
 * Each process executes distBubbletonic for uin8_t [16]
 */
TEST_F(TMPIdistSort, distBubbletonic_test1) {
    // Create and fill vector
    using tsValue_t = uint8_t;      // Test parameters
    size_t ts_buffer_size = 16;

    ShadowedVec_t<tsValue_t> ts_Data;
    std::uniform_int_distribution<tsValue_t > dis(
            std::numeric_limits<tsValue_t>::min(),
            std::numeric_limits<tsValue_t>::max()
    );
    ts_Data.resize(ts_buffer_size);
    std::generate(ts_Data.begin(), ts_Data.end(), [&]() { return dis(gen); });

    // Execute function under test in all processes
    distBubbletonic(ts_Data, ts_mpi.size(), ts_mpi.rank());

    // Local min and max
    auto local_min = *std::min_element(ts_Data.begin(), ts_Data.end());
    auto local_max = *std::max_element(ts_Data.begin(), ts_Data.end());

    // Gather min/max to rank 0
    std::vector<tsValue_t> global_mins(ts_mpi.size());
    std::vector<tsValue_t> global_maxes(ts_mpi.size());
    MPI_Datatype datatype = MPI_TypeMapper<tsValue_t>::getType();

    MPI_Gather(&local_min, 1, datatype, global_mins.data(), 1, datatype, 0, MPI_COMM_WORLD);
    MPI_Gather(&local_max, 1, datatype, global_maxes.data(), 1, datatype, 0, MPI_COMM_WORLD);

    // Check results
    EXPECT_EQ(std::is_sorted(ts_Data.begin(), ts_Data.end()), true);
    if (ts_mpi.rank() == 0) {
        for (size_t i = 1; i < global_mins.size(); ++i) {
            EXPECT_LE(global_maxes[i - 1], global_mins[i]);
        }
    }
}

/*
 * MPI: SysTest (acceptance)
 * Each process executes distBubbletonic for uin32_t [1 << 16]
 */
TEST_F(TMPIdistSort, distBubbletonic_test2) {
    // Create and fill vector
    using tsValue_t = uint32_t;      // Test parameters
    size_t ts_buffer_size = 1 << 16;

    ShadowedVec_t<tsValue_t> ts_Data;
    std::uniform_int_distribution<tsValue_t > dis(
            std::numeric_limits<tsValue_t>::min(),
            std::numeric_limits<tsValue_t>::max()
    );
    ts_Data.resize(ts_buffer_size);
    std::generate(ts_Data.begin(), ts_Data.end(), [&]() { return dis(gen); });

    // Execute function under test in all processes
    distBubbletonic(ts_Data, ts_mpi.size(), ts_mpi.rank());

    // Local min and max
    auto local_min = *std::min_element(ts_Data.begin(), ts_Data.end());
    auto local_max = *std::max_element(ts_Data.begin(), ts_Data.end());

    // Gather min/max to rank 0
    std::vector<tsValue_t> global_mins(ts_mpi.size());
    std::vector<tsValue_t> global_maxes(ts_mpi.size());
    MPI_Datatype datatype = MPI_TypeMapper<tsValue_t>::getType();

    MPI_Gather(&local_min, 1, datatype, global_mins.data(), 1, datatype, 0, MPI_COMM_WORLD);
    MPI_Gather(&local_max, 1, datatype, global_maxes.data(), 1, datatype, 0, MPI_COMM_WORLD);

    // Check results
    EXPECT_EQ(std::is_sorted(ts_Data.begin(), ts_Data.end()), true);
    if (ts_mpi.rank() == 0) {
        for (size_t i = 1; i < global_mins.size(); ++i) {
            EXPECT_LE(global_maxes[i - 1], global_mins[i]);
        }
    }
}


/*
 * MPI: SysTest (acceptance)
 * Each process executes distBitonic for uin8_t [16]
 */
TEST_F(TMPIdistSort, distBitonic_test1) {
    // Create and fill vector
    using tsValue_t = uint8_t;      // Test parameters
    size_t ts_buffer_size = 16;

    ShadowedVec_t<tsValue_t> ts_Data;
    std::uniform_int_distribution<tsValue_t > dis(
            std::numeric_limits<tsValue_t>::min(),
            std::numeric_limits<tsValue_t>::max()
    );
    ts_Data.resize(ts_buffer_size);
    std::generate(ts_Data.begin(), ts_Data.end(), [&]() { return dis(gen); });

    // Execute function under test in all processes
    distBitonic(ts_Data, ts_mpi.size(), ts_mpi.rank());

    // Local min and max
    auto local_min = *std::min_element(ts_Data.begin(), ts_Data.end());
    auto local_max = *std::max_element(ts_Data.begin(), ts_Data.end());

    // Gather min/max to rank 0
    std::vector<tsValue_t> global_mins(ts_mpi.size());
    std::vector<tsValue_t> global_maxes(ts_mpi.size());
    MPI_Datatype datatype = MPI_TypeMapper<tsValue_t>::getType();

    MPI_Gather(&local_min, 1, datatype, global_mins.data(), 1, datatype, 0, MPI_COMM_WORLD);
    MPI_Gather(&local_max, 1, datatype, global_maxes.data(), 1, datatype, 0, MPI_COMM_WORLD);

    // Check results
    EXPECT_EQ(std::is_sorted(ts_Data.begin(), ts_Data.end()), true);
    if (ts_mpi.rank() == 0) {
        for (size_t i = 1; i < global_mins.size(); ++i) {
            EXPECT_LE(global_maxes[i - 1], global_mins[i]);
        }
    }
}

/*
 * MPI: SysTest (acceptance)
 * Each process executes distBitonic for uin32_t [1 << 16]
 */
TEST_F(TMPIdistSort, distBitonic_test2) {
    // Create and fill vector
    using tsValue_t = uint32_t;      // Test parameters
    size_t ts_buffer_size = 1 << 16;

    ShadowedVec_t<tsValue_t> ts_Data;
    std::uniform_int_distribution<tsValue_t > dis(
            std::numeric_limits<tsValue_t>::min(),
            std::numeric_limits<tsValue_t>::max()
    );
    ts_Data.resize(ts_buffer_size);
    std::generate(ts_Data.begin(), ts_Data.end(), [&]() { return dis(gen); });

    // Execute function under test in all processes
    distBitonic(ts_Data, ts_mpi.size(), ts_mpi.rank());

    // Local min and max
    auto local_min = *std::min_element(ts_Data.begin(), ts_Data.end());
    auto local_max = *std::max_element(ts_Data.begin(), ts_Data.end());

    // Gather min/max to rank 0
    std::vector<tsValue_t> global_mins(ts_mpi.size());
    std::vector<tsValue_t> global_maxes(ts_mpi.size());
    MPI_Datatype datatype = MPI_TypeMapper<tsValue_t>::getType();

    MPI_Gather(&local_min, 1, datatype, global_mins.data(), 1, datatype, 0, MPI_COMM_WORLD);
    MPI_Gather(&local_max, 1, datatype, global_maxes.data(), 1, datatype, 0, MPI_COMM_WORLD);

    // Check results
    EXPECT_EQ(std::is_sorted(ts_Data.begin(), ts_Data.end()), true);
    if (ts_mpi.rank() == 0) {
        for (size_t i = 1; i < global_mins.size(); ++i) {
            EXPECT_LE(global_maxes[i - 1], global_mins[i]);
        }
    }
}