/* -------------------------------------------------------------------------
 * This file is part of the MindStudio project.
 * Copyright (c) 2025 Huawei Technologies Co.,Ltd.
 *
 * MindStudio is licensed under Mulan PSL v2.
 * You can use this software according to the terms and conditions of the Mulan PSL v2.
 * You may obtain a copy of Mulan PSL v2 at:
 *
 *          http://license.coscl.org.cn/MulanPSL2
 *
 * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
 * EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
 * MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
 * See the Mulan PSL v2 for more details.
 * ------------------------------------------------------------------------- */
#include <algorithm>
#include <cstdint>
#include <cstdlib>
#include <iostream>
#include <vector>

#include "acl/acl.h"
#include "aclnn_add_custom.h"

const int SUCCESS = 0;
const int FAILED = 1;

#define CHECK_RET(cond, return_expr) \
    do {                             \
        if (!(cond)) {               \
            return_expr;             \
        }                            \
    } while (0)

#define LOG_PRINT(message, ...)         \
    do {                                \
        printf(message, ##__VA_ARGS__); \
    } while (0)

int64_t GetShapeSize(const std::vector<int64_t> &shape)
{
    int64_t shapeSize = 1;
    for (auto i : shape) {
        shapeSize *= i;
    }
    return shapeSize;
}

int Init(int32_t deviceId, aclrtStream *stream)
{
    // Fixed code, acl initialization
    auto ret = aclInit(nullptr);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return FAILED);
    ret = aclrtSetDevice(deviceId);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSetDevice failed. ERROR: %d\n", ret); return FAILED);
    ret = aclrtCreateStream(stream);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtCreateStream failed. ERROR: %d\n", ret); return FAILED);

    return SUCCESS;
}

template <typename T>
int CreateAclTensor(const std::vector<T> &hostData, const std::vector<int64_t> &shape, void **deviceAddr,
                    aclDataType dataType, aclTensor **tensor)
{
    auto size = GetShapeSize(shape) * sizeof(T);
    // Call aclrtMalloc to allocate device memory
    auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret); return FAILED);

    // Call aclrtMemcpy to copy host data to device memory
    ret = aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return FAILED);

    // Call aclCreateTensor to create a aclTensor object
    *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, nullptr, 0, aclFormat::ACL_FORMAT_ND, shape.data(),
                              shape.size(), *deviceAddr);
    return SUCCESS;
}

void DestroyResources(std::vector<void *> tensors, std::vector<void *> deviceAddrs, aclrtStream stream,
                      int32_t deviceId, void *workspaceAddr = nullptr)
{
    // Release aclTensor and device
    for (uint32_t i = 0; i < tensors.size(); i++) {
        if (tensors[i] != nullptr) {
            aclDestroyTensor(reinterpret_cast<aclTensor *>(tensors[i]));
        }
        if (deviceAddrs[i] != nullptr) {
            aclrtFree(deviceAddrs[i]);
        }
    }
    if (workspaceAddr != nullptr) {
        aclrtFree(workspaceAddr);
    }
    // Destroy stream and reset device
    aclrtDestroyStream(stream);
    aclrtResetDevice(deviceId);
    aclFinalize();
}

int main(int argc, char **argv)
{
    // 1. (Fixed code) Initialize device / stream, refer to the list of external interfaces of acl
    int32_t deviceId = 0;
    if (argc > 1) {
        deviceId = static_cast<int32_t>(strtol(argv[1], nullptr, 10));
    }
    LOG_PRINT(
        "Running on NPU [%d]. If this is the first run on this card, scheduling may take a few seconds; please wait...\n",
        deviceId);

    aclrtStream stream;
    auto ret = Init(deviceId, &stream);
    CHECK_RET(ret == 0, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return FAILED);

    // 2. Create input and output, need to customize according to the interface of the API
    std::vector<int64_t> inputXShape = {8, 2048};
    std::vector<int64_t> inputYShape = {8, 2048};
    std::vector<int64_t> outputZShape = {8, 2048};
    void *inputXDeviceAddr = nullptr;
    void *inputYDeviceAddr = nullptr;
    void *outputZDeviceAddr = nullptr;
    aclTensor *inputX = nullptr;
    aclTensor *inputY = nullptr;
    aclTensor *outputZ = nullptr;
    std::vector<aclFloat16> inputXHostData(inputXShape[0] * inputXShape[1]);
    std::vector<aclFloat16> inputYHostData(inputYShape[0] * inputYShape[1]);
    std::vector<aclFloat16> outputZHostData(outputZShape[0] * outputZShape[1]);
    for (int i = 0; i < inputXShape[0] * inputXShape[1]; ++i) {
        inputXHostData[i] = aclFloatToFloat16(1.0);
        inputYHostData[i] = aclFloatToFloat16(2.0);
        outputZHostData[i] = aclFloatToFloat16(0.0);
    }
    std::vector<void *> tensors = {inputX, inputY, outputZ};
    std::vector<void *> deviceAddrs = {inputXDeviceAddr, inputYDeviceAddr, outputZDeviceAddr};
    // Create inputX aclTensor
    ret = CreateAclTensor(inputXHostData, inputXShape, &inputXDeviceAddr, aclDataType::ACL_FLOAT16, &inputX);
    CHECK_RET(ret == ACL_SUCCESS, DestroyResources(tensors, deviceAddrs, stream, deviceId); return FAILED);
    // Create inputY aclTensor
    ret = CreateAclTensor(inputYHostData, inputYShape, &inputYDeviceAddr, aclDataType::ACL_FLOAT16, &inputY);
    CHECK_RET(ret == ACL_SUCCESS, DestroyResources(tensors, deviceAddrs, stream, deviceId); return FAILED);
    // Create outputZ aclTensor
    ret = CreateAclTensor(outputZHostData, outputZShape, &outputZDeviceAddr, aclDataType::ACL_FLOAT16, &outputZ);
    CHECK_RET(ret == ACL_SUCCESS, DestroyResources(tensors, deviceAddrs, stream, deviceId); return FAILED);

    // 3. Call the API of the custom operator library
    uint64_t workspaceSize = 0;
    aclOpExecutor *executor;
    // Calculate the workspace size and allocate memory for it
    ret = aclnnAddCustomGetWorkspaceSize(inputX, inputY, outputZ, &workspaceSize, &executor);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAddCustomGetWorkspaceSize failed. ERROR: %d\n", ret);
              DestroyResources(tensors, deviceAddrs, stream, deviceId); return FAILED);

    void *workspaceAddr = nullptr;
    if (workspaceSize > 0) {
        ret = aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
        CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret);
                  DestroyResources(tensors, deviceAddrs, stream, deviceId, workspaceAddr); return FAILED);
    }
    // Execute the custom operator
    ret = aclnnAddCustom(workspaceAddr, workspaceSize, executor, stream);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAdd failed. ERROR: %d\n", ret);
              DestroyResources(tensors, deviceAddrs, stream, deviceId, workspaceAddr); return FAILED);

    // 4. (Fixed code) Synchronize and wait for the task to complete
    ret = aclrtSynchronizeStream(stream);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret);
              DestroyResources(tensors, deviceAddrs, stream, deviceId, workspaceAddr); return FAILED);

    // 5. Get the output value, copy the result from device memory to host memory, need to modify according to the
    // interface of the API
    auto size = GetShapeSize(outputZShape);
    std::vector<aclFloat16> resultData(size, 0);
    ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outputZDeviceAddr,
                      size * sizeof(aclFloat16), ACL_MEMCPY_DEVICE_TO_HOST);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("copy result from device to host failed. ERROR: %d\n", ret);
              DestroyResources(tensors, deviceAddrs, stream, deviceId, workspaceAddr); return FAILED);

    // 6. Detroy resources, need to modify according to the interface of the API
    DestroyResources(tensors, deviceAddrs, stream, deviceId, workspaceAddr);

    // print the output result
    std::vector<aclFloat16> goldenData(size, aclFloatToFloat16(3.0));

    LOG_PRINT("The input consists of two arrays, whose values are all 1.0 and 2.0\n");
    LOG_PRINT("result is:\n");
    for (int64_t i = 0; i < 10; i++) {
        LOG_PRINT("%.1f ", aclFloat16ToFloat(resultData[i]));
    }
    LOG_PRINT("\n");
    if (std::equal(resultData.begin(), resultData.end(), goldenData.begin())) {
        LOG_PRINT("test pass\n");
    } else {
        LOG_PRINT("test failed\n");
        return FAILED;
    }
    return SUCCESS;
}