/* ************************************************************************
 * 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"

//
// herkx_batched when TWOK = false
//

template <typename T, bool TWOK = true>
void testing_her2k_batched_bad_arg(const Arguments& arg)
{
    // clang-format off
    auto rocblas_herXX_batched_fn = arg.fortran
                                        ? (TWOK ? rocblas_her2k_batched<T, real_t<T>, true>
                                                : rocblas_herkx_batched<T, real_t<T>, true>)
                                        : (TWOK ? rocblas_her2k_batched<T, real_t<T>, false>
                                                : rocblas_herkx_batched<T, real_t<T>, false>);
    // clang-format on

    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       ldb    = 100;
    const rocblas_int       ldc    = 100;
    const T                 alpha  = 1.0;
    using U                        = real_t<T>;
    const U     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> dB(safe_size, 1, batch_count);
    device_batch_vector<T> dC(safe_size, 1, batch_count);
    CHECK_DEVICE_ALLOCATION(dA.memcheck());
    CHECK_DEVICE_ALLOCATION(dB.memcheck());
    CHECK_DEVICE_ALLOCATION(dC.memcheck());

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

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

    EXPECT_ROCBLAS_STATUS(rocblas_herXX_batched_fn(handle,
                                                   uplo,
                                                   rocblas_operation_transpose,
                                                   N,
                                                   K,
                                                   &alpha,
                                                   dA,
                                                   lda,
                                                   dB,
                                                   ldb,
                                                   &beta,
                                                   dC,
                                                   ldc,
                                                   batch_count),
                          rocblas_status_invalid_value);

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

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

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

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

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

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

template <typename T, bool TWOK = true>
void testing_her2k_batched(const Arguments& arg)
{
    // clang-format off
    auto rocblas_herXX_batched_fn = arg.fortran
                                        ? (TWOK ? rocblas_her2k_batched<T, real_t<T>, true>
                                                : rocblas_herkx_batched<T, real_t<T>, true>)
                                        : (TWOK ? rocblas_her2k_batched<T, real_t<T>, false>
                                                : rocblas_herkx_batched<T, real_t<T>, false>);
    // clang-format on

    auto herXX_gflop_count_fn = TWOK ? her2k_gflop_count<T> : herkx_gflop_count<T>;
    auto herXX_ref_fn         = TWOK ? cblas_her2k<T> : cblas_herkx<T>;

    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          ldb    = arg.ldb;
    rocblas_int          ldc    = arg.ldc;
    T                    alpha  = arg.get_alpha<T>();
    using U                     = real_t<T>;
    U           beta            = arg.get_beta<U>();
    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 = batch_count < 0 || N < 0 || K < 0 || ldc < N
                        || (transA == rocblas_operation_none && (lda < N || ldb < N))
                        || (transA != rocblas_operation_none && (lda < K || ldb < K));
    if(N == 0 || batch_count == 0 || invalid_size)
    {
        // ensure invalid sizes checked before pointer check
        EXPECT_ROCBLAS_STATUS(rocblas_herXX_batched_fn(handle,
                                                       uplo,
                                                       transA,
                                                       N,
                                                       K,
                                                       nullptr,
                                                       nullptr,
                                                       lda,
                                                       nullptr,
                                                       ldb,
                                                       nullptr,
                                                       nullptr,
                                                       ldc,
                                                       batch_count),
                              invalid_size ? rocblas_status_invalid_size : rocblas_status_success);

        return;
    }

    size_t     cols   = (transA == rocblas_operation_none ? std::max(K, 1) : N);
    size_t     rows   = (transA != rocblas_operation_none ? std::max(K, 1) : N);
    const auto size_A = lda * cols;
    const auto size_B = ldb * cols;
    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> dB(size_B, 1, batch_count);
    device_batch_vector<T> dC(size_C, 1, batch_count);
    device_vector<T>       d_alpha(1);
    device_vector<U>       d_beta(1);
    CHECK_DEVICE_ALLOCATION(dA.memcheck());
    CHECK_DEVICE_ALLOCATION(dB.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<U>       h_beta(1);
    host_batch_vector<T> hA(size_A, 1, batch_count);
    host_batch_vector<T> hB(size_B, 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(hB.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);
    if(TWOK)
    {
        rocblas_init_vector(hB, arg, rocblas_client_never_set_nan, false, true);
    }
    else
    { // require symmetric A*B^H so testing with B = A
        hB.copy_from(hA);
    }
    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));
    CHECK_HIP_ERROR(dB.transfer_from(hB));

    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_herXX_batched_fn(handle,
                                                     uplo,
                                                     transA,
                                                     N,
                                                     K,
                                                     &h_alpha[0],
                                                     dA.ptr_on_device(),
                                                     lda,
                                                     dB.ptr_on_device(),
                                                     ldb,
                                                     &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_herXX_batched_fn(handle,
                                                     uplo,
                                                     transA,
                                                     N,
                                                     K,
                                                     d_alpha,
                                                     dA.ptr_on_device(),
                                                     lda,
                                                     dB.ptr_on_device(),
                                                     ldb,
                                                     d_beta,
                                                     dC.ptr_on_device(),
                                                     ldc,
                                                     batch_count));

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

        // CPU BLAS
        if(arg.timing)
        {
            cpu_time_used = get_time_us_no_sync();
        }

        // cpu reference
        for(int i = 0; i < batch_count; i++)
        {
            // herkx: B equals A to ensure a symmetric result
            herXX_ref_fn(uplo,
                         transA,
                         N,
                         K,
                         &h_alpha[0],
                         hA[i],
                         lda,
                         hB[i],
                         ldb,
                         &h_beta[0],
                         hC_gold[i],
                         ldc);
        }

        if(arg.timing)
        {
            cpu_time_used = get_time_us_no_sync() - cpu_time_used;
        }

        if(arg.unit_check)
        {
            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);
        }

        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_herXX_batched_fn(handle,
                                     uplo,
                                     transA,
                                     N,
                                     K,
                                     h_alpha,
                                     dA.ptr_on_device(),
                                     lda,
                                     dB.ptr_on_device(),
                                     ldb,
                                     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_herXX_batched_fn(handle,
                                     uplo,
                                     transA,
                                     N,
                                     K,
                                     h_alpha,
                                     dA.ptr_on_device(),
                                     lda,
                                     dB.ptr_on_device(),
                                     ldb,
                                     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_ldb,
                      e_beta,
                      e_ldc,
                      e_batch_count>{}
            .log_args<T>(rocblas_cout,
                         arg,
                         gpu_time_used,
                         herXX_gflop_count_fn(N, K),
                         ArgumentLogging::NA_value,
                         cpu_time_used,
                         rocblas_error);
    }
}