/* ************************************************************************
 * Copyright 2020-2021 Advanced Micro Devices, Inc.
 * ************************************************************************ */

#pragma once

#include "bytes.hpp"
#include "cblas_interface.hpp"
#include "flops.hpp"
#include "near.hpp"
#include "norm.hpp"
#include "rocblas.hpp"
#include "rocblas_datatype2string.hpp"
#include "rocblas_init.hpp"
#include "rocblas_math.hpp"
#include "rocblas_random.hpp"
#include "rocblas_test.hpp"
#include "rocblas_vector.hpp"
#include "unit.hpp"
#include "utility.hpp"

template <typename T>
void testing_syrk_batched_bad_arg(const Arguments& arg)
{
    auto rocblas_syrk_batched_fn
        = arg.fortran ? rocblas_syrk_batched<T, true> : rocblas_syrk_batched<T, false>;

    rocblas_local_handle    handle{arg};
    const rocblas_fill      uplo        = rocblas_fill_upper;
    const rocblas_operation transA      = rocblas_operation_none;
    const rocblas_int       N           = 100;
    const rocblas_int       K           = 100;
    const rocblas_int       lda         = 100;
    const rocblas_int       ldc         = 100;
    const T                 alpha       = 1.0;
    const T                 beta        = 1.0;
    rocblas_int             batch_count = 2;

    const size_t safe_size = 100;
    // allocate memory on device
    device_batch_vector<T> dA(safe_size, 1, batch_count);
    device_batch_vector<T> dC(safe_size, 1, batch_count);
    CHECK_DEVICE_ALLOCATION(dA.memcheck());
    CHECK_DEVICE_ALLOCATION(dC.memcheck());

    EXPECT_ROCBLAS_STATUS(
        rocblas_syrk_batched_fn(
            nullptr, uplo, transA, N, K, &alpha, dA, lda, &beta, dC, ldc, batch_count),
        rocblas_status_invalid_handle);

    EXPECT_ROCBLAS_STATUS(
        rocblas_syrk_batched_fn(
            handle, rocblas_fill_full, transA, N, K, &alpha, dA, lda, &beta, dC, ldc, batch_count),
        rocblas_status_invalid_value);

    EXPECT_ROCBLAS_STATUS(
        rocblas_syrk_batched_fn(
            handle, uplo, transA, N, K, nullptr, dA, lda, &beta, dC, ldc, batch_count),
        rocblas_status_invalid_pointer);

    EXPECT_ROCBLAS_STATUS(
        rocblas_syrk_batched_fn(
            handle, uplo, transA, N, K, &alpha, nullptr, lda, &beta, dC, ldc, batch_count),
        rocblas_status_invalid_pointer);

    EXPECT_ROCBLAS_STATUS(
        rocblas_syrk_batched_fn(
            handle, uplo, transA, N, K, &alpha, dA, lda, nullptr, dC, ldc, batch_count),
        rocblas_status_invalid_pointer);

    EXPECT_ROCBLAS_STATUS(
        rocblas_syrk_batched_fn(
            handle, uplo, transA, N, K, &alpha, dA, lda, &beta, nullptr, ldc, batch_count),
        rocblas_status_invalid_pointer);

    // quick return with invalid pointers
    EXPECT_ROCBLAS_STATUS(
        rocblas_syrk_batched_fn(
            handle, uplo, transA, 0, K, nullptr, nullptr, lda, nullptr, nullptr, ldc, batch_count),
        rocblas_status_success);

    if(is_complex<T>)
    {
        EXPECT_ROCBLAS_STATUS(rocblas_syrk_batched_fn(handle,
                                                      uplo,
                                                      rocblas_operation_conjugate_transpose,
                                                      N,
                                                      K,
                                                      &alpha,
                                                      dA,
                                                      lda,
                                                      &beta,
                                                      dC,
                                                      ldc,
                                                      batch_count),
                              rocblas_status_invalid_value);
    }
}

template <typename T>
void testing_syrk_batched(const Arguments& arg)
{
    auto rocblas_syrk_batched_fn
        = arg.fortran ? rocblas_syrk_batched<T, true> : rocblas_syrk_batched<T, false>;

    rocblas_local_handle handle{arg};
    rocblas_fill         uplo        = char2rocblas_fill(arg.uplo);
    rocblas_operation    transA      = char2rocblas_operation(arg.transA);
    rocblas_int          N           = arg.N;
    rocblas_int          K           = arg.K;
    rocblas_int          lda         = arg.lda;
    rocblas_int          ldc         = arg.ldc;
    T                    alpha       = arg.get_alpha<T>();
    T                    beta        = arg.get_beta<T>();
    rocblas_int          batch_count = arg.batch_count;

    double gpu_time_used, cpu_time_used;
    double rocblas_error = 0.0;

    // Note: K==0 is not an early exit, since C still needs to be multiplied by beta
    bool invalid_size = N < 0 || K < 0 || ldc < N || (transA == rocblas_operation_none && lda < N)
                        || (transA != rocblas_operation_none && lda < K) || batch_count < 0;
    if(N == 0 || batch_count == 0 || invalid_size)
    {
        // ensure invalid sizes checked before pointer check

        EXPECT_ROCBLAS_STATUS(rocblas_syrk_batched_fn(handle,
                                                      uplo,
                                                      transA,
                                                      N,
                                                      K,
                                                      nullptr,
                                                      nullptr,
                                                      lda,
                                                      nullptr,
                                                      nullptr,
                                                      ldc,
                                                      batch_count),
                              invalid_size ? rocblas_status_invalid_size : rocblas_status_success);

        return;
    }

    const auto size_A = size_t(lda) * (transA == rocblas_operation_none ? K : N);
    const auto size_C = size_t(ldc) * N;

    // allocate memory on device
    device_batch_vector<T> dA(size_A, 1, batch_count);
    device_batch_vector<T> dC(size_C, 1, batch_count);
    device_vector<T>       d_alpha(1);
    device_vector<T>       d_beta(1);
    CHECK_DEVICE_ALLOCATION(dA.memcheck());
    CHECK_DEVICE_ALLOCATION(dC.memcheck());
    CHECK_DEVICE_ALLOCATION(d_alpha.memcheck());
    CHECK_DEVICE_ALLOCATION(d_beta.memcheck());

    // Naming: dX is in GPU (device) memory. hK is in CPU (host) memory
    host_vector<T>       h_alpha(1);
    host_vector<T>       h_beta(1);
    host_batch_vector<T> hA(size_A, 1, batch_count);
    host_batch_vector<T> hC_1(size_C, 1, batch_count);
    host_batch_vector<T> hC_2(size_C, 1, batch_count);
    host_batch_vector<T> hC_gold(size_C, 1, batch_count);

    CHECK_HIP_ERROR(h_alpha.memcheck());
    CHECK_HIP_ERROR(h_beta.memcheck());
    CHECK_HIP_ERROR(hA.memcheck());
    CHECK_HIP_ERROR(hC_1.memcheck());
    CHECK_HIP_ERROR(hC_2.memcheck());
    CHECK_HIP_ERROR(hC_gold.memcheck());

    // Initial Data on CPU
    h_alpha[0] = alpha;
    h_beta[0]  = beta;

    // Initialize data on host memory
    rocblas_init_vector(hA, arg, rocblas_client_alpha_sets_nan, true);
    rocblas_init_vector(hC_1, arg, rocblas_client_beta_sets_nan);

    hC_2.copy_from(hC_1);
    hC_gold.copy_from(hC_1);

    // copy data from CPU to device
    CHECK_HIP_ERROR(dA.transfer_from(hA));

    if(arg.unit_check || arg.norm_check)
    {
        // host alpha/beta
        CHECK_ROCBLAS_ERROR(rocblas_set_pointer_mode(handle, rocblas_pointer_mode_host));
        CHECK_HIP_ERROR(dC.transfer_from(hC_1));

        CHECK_ROCBLAS_ERROR(rocblas_syrk_batched_fn(handle,
                                                    uplo,
                                                    transA,
                                                    N,
                                                    K,
                                                    &h_alpha[0],
                                                    dA.ptr_on_device(),
                                                    lda,
                                                    &h_beta[0],
                                                    dC.ptr_on_device(),
                                                    ldc,
                                                    batch_count));

        // copy output from device to CPU
        CHECK_HIP_ERROR(hC_1.transfer_from(dC));

        // device alpha/beta
        CHECK_ROCBLAS_ERROR(rocblas_set_pointer_mode(handle, rocblas_pointer_mode_device));
        CHECK_HIP_ERROR(dC.transfer_from(hC_2));
        CHECK_HIP_ERROR(d_alpha.transfer_from(h_alpha));
        CHECK_HIP_ERROR(d_beta.transfer_from(h_beta));

        CHECK_ROCBLAS_ERROR(rocblas_syrk_batched_fn(handle,
                                                    uplo,
                                                    transA,
                                                    N,
                                                    K,
                                                    d_alpha,
                                                    dA.ptr_on_device(),
                                                    lda,
                                                    d_beta,
                                                    dC.ptr_on_device(),
                                                    ldc,
                                                    batch_count));

        // CPU BLAS
        cpu_time_used = get_time_us_no_sync();

        // cpu reference
        for(int i = 0; i < batch_count; i++)
        {
            cblas_syrk<T>(uplo, transA, N, K, h_alpha[0], hA[i], lda, h_beta[0], hC_gold[i], ldc);
        }

        cpu_time_used = get_time_us_no_sync() - cpu_time_used;

        // copy output from device to CPU
        CHECK_HIP_ERROR(hC_2.transfer_from(dC));

        if(arg.unit_check)
        {
            if(std::is_same<T, rocblas_float_complex>{}
               || std::is_same<T, rocblas_double_complex>{})
            {
                const double tol = K * sum_error_tolerance<T>;
                near_check_general<T>(N, N, ldc, hC_gold, hC_1, batch_count, tol);
                near_check_general<T>(N, N, ldc, hC_gold, hC_2, batch_count, tol);
            }
            else
            {
                unit_check_general<T>(N, N, ldc, hC_gold, hC_1, batch_count);
                unit_check_general<T>(N, N, ldc, hC_gold, hC_2, batch_count);
            }
        }

        if(arg.norm_check)
        {
            auto err1 = std::abs(norm_check_general<T>('F', N, N, ldc, hC_gold, hC_1, batch_count));
            auto err2 = std::abs(norm_check_general<T>('F', N, N, ldc, hC_gold, hC_2, batch_count));
            rocblas_error = err1 > err2 ? err1 : err2;
        }
    }

    if(arg.timing)
    {
        int number_cold_calls = arg.cold_iters;
        int number_hot_calls  = arg.iters;

        CHECK_ROCBLAS_ERROR(rocblas_set_pointer_mode(handle, rocblas_pointer_mode_host));

        for(int i = 0; i < number_cold_calls; i++)
        {
            rocblas_syrk_batched_fn(handle,
                                    uplo,
                                    transA,
                                    N,
                                    K,
                                    h_alpha,
                                    dA.ptr_on_device(),
                                    lda,
                                    h_beta,
                                    dC.ptr_on_device(),
                                    ldc,
                                    batch_count);
        }

        hipStream_t stream;
        CHECK_ROCBLAS_ERROR(rocblas_get_stream(handle, &stream));
        gpu_time_used = get_time_us_sync(stream); // in microseconds
        for(int i = 0; i < number_hot_calls; i++)
        {
            rocblas_syrk_batched_fn(handle,
                                    uplo,
                                    transA,
                                    N,
                                    K,
                                    h_alpha,
                                    dA.ptr_on_device(),
                                    lda,
                                    h_beta,
                                    dC.ptr_on_device(),
                                    ldc,
                                    batch_count);
        }
        gpu_time_used = get_time_us_sync(stream) - gpu_time_used;

        ArgumentModel<e_uplo, e_transA, e_N, e_K, e_alpha, e_lda, e_beta, e_ldc, e_batch_count>{}
            .log_args<T>(rocblas_cout,
                         arg,
                         gpu_time_used,
                         syrk_gflop_count<T>(N, K),
                         ArgumentLogging::NA_value,
                         cpu_time_used,
                         rocblas_error);
    }
}