/**
 * Copyright (c) 2025 Huawei Technologies Co., Ltd.
 * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
 * CANN Open Software License Agreement Version 2.0 (the "License").
 * Please refer to the License for details. You may not use this file except in compliance with the License.
 * 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 FITNESS FOR A PARTICULAR PURPOSE.
 * See LICENSE in the root of the software repository for the full text of the License.
 */

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_kv_rms_norm_rope_cache.h"

#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;
}

void PrintOutResult(std::vector<int64_t> &shape, void** deviceAddr) {
  auto size = GetShapeSize(shape);
  std::vector<int8_t> resultData(size, 0);
  auto ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]),
                         *deviceAddr, size * sizeof(resultData[0]), ACL_MEMCPY_DEVICE_TO_HOST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("copy result from device to host failed. ERROR: %d\n", ret); return);
  auto batchKv = shape[0];
  auto batchSeq = shape[2];
  // Print by [B, S]
  uint64_t bIdx = 0;
  uint64_t sIdx = 0;
  for (uint64_t i = 0; i < size; i++) {
    LOG_PRINT("result[%lu] is: %d\n", i, resultData[i]);
    // Fast diagonal traverse
    bIdx = (bIdx >= batchKv) ? batchKv : (bIdx + 1);
    sIdx = (sIdx >= batchSeq) ? batchSeq : (sIdx + 1);
    i += bIdx * batchKv * batchSeq + sIdx * batchSeq;
  }
}

int Init(int32_t deviceId, aclrtStream* stream) {
  // 固定写法，资源初始化
  auto ret = aclInit(nullptr);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return ret);
  ret = aclrtSetDevice(deviceId);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSetDevice failed. ERROR: %d\n", ret); return ret);
  ret = aclrtCreateStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtCreateStream failed. ERROR: %d\n", ret); return ret);
  return 0;
}

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);
  // 调用aclrtMalloc申请device侧内存
  auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret); return ret);
  // 调用aclrtMemcpy将host侧数据拷贝到device侧内存上
  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 ret);

  // 计算连续tensor的strides
  std::vector<int64_t> strides(shape.size(), 1);
  for (int64_t i = shape.size() - 2; i >= 0; i--) {
    strides[i] = shape[i + 1] * strides[i + 1];
  }

  // 调用aclCreateTensor接口创建aclTensor
  *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND,
                            shape.data(), shape.size(), *deviceAddr);
  return 0;
}

int main() {
  // 1. 固定写法，device/stream初始化, 参考acl API
  // 根据自己的实际device填写deviceId
  int32_t deviceId = 0;
  aclrtStream stream;
  auto ret = Init(deviceId, &stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);

  // 2. 构造输入与输出，需要根据API的接口定义构造
  uint32_t batchKv = 32;
  std::vector<int64_t> kvShape = {batchKv,1,1,576};
  std::vector<int64_t> gammaShape = {512,};
  std::vector<int64_t> cosShape = {batchKv,1,1,64};
  std::vector<int64_t> sinShape = {batchKv,1,1,64};
  std::vector<int64_t> indexShape = {batchKv,1};
  std::vector<int64_t> kpeCacheShape = {batchKv,1,1,64};
  std::vector<int64_t> ckvCacheShape = {batchKv,1,1,512};
  std::vector<int64_t> kRopeShape = {batchKv,1,1,64};
  std::vector<int64_t> cKvShape = {batchKv,1,1,512};

  std::vector<int16_t> kvHostData(batchKv*1*1*576,0);
  std::vector<int16_t> gammaHostData(512,0);
  std::vector<int16_t> cosHostData(batchKv*1*1*64,0);
  std::vector<int16_t> sinHostData(batchKv*1*1*64,0);
  std::vector<int64_t> indexHostData(batchKv*1,0);
  std::vector<int16_t> kpeCacheHostData(batchKv*1*1*64,0);
  std::vector<int16_t> ckvCacheHostData(batchKv*1*1*512,0);
  std::vector<int16_t> kRopeHostData(batchKv*1*1*64,0);
  std::vector<int16_t> cKvHostData(batchKv*1*1*512,0);

  void* kvDeviceAddr = nullptr;
  void* gammaDeviceAddr = nullptr;
  void* cosDeviceAddr = nullptr;
  void* sinDeviceAddr = nullptr;
  void* indexDeviceAddr = nullptr;
  void* kpeCacheDeviceAddr = nullptr;
  void* ckvCacheDeviceAddr = nullptr;
  void* kRopeDeviceAddr = nullptr;
  void* cKvDeviceAddr = nullptr;

  aclTensor* kv = nullptr;
  aclTensor* gamma = nullptr;
  aclTensor* cos = nullptr;
  aclTensor* sin = nullptr;
  aclTensor* index = nullptr;
  aclTensor* kpeCache = nullptr;
  aclTensor* ckvCache = nullptr;
  aclTensor* kRope = nullptr;
  aclTensor* cKv = nullptr;


  double epsilon = 1e-5;
  char cacheMode[] = "Norm";
  bool isOutputKv = false;

  ret = CreateAclTensor(kvHostData, kvShape, &kvDeviceAddr, aclDataType::ACL_FLOAT16, &kv);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  ret = CreateAclTensor(gammaHostData, gammaShape, &gammaDeviceAddr, aclDataType::ACL_FLOAT16, &gamma);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  ret = CreateAclTensor(cosHostData, cosShape, &cosDeviceAddr, aclDataType::ACL_FLOAT16, &cos);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  ret = CreateAclTensor(sinHostData, sinShape, &sinDeviceAddr, aclDataType::ACL_FLOAT16, &sin);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  ret = CreateAclTensor(indexHostData, indexShape, &indexDeviceAddr, aclDataType::ACL_INT64, &index);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  ret = CreateAclTensor(kpeCacheHostData, kpeCacheShape, &kpeCacheDeviceAddr, aclDataType::ACL_FLOAT16, &kpeCache);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  ret = CreateAclTensor(ckvCacheHostData, ckvCacheShape, &ckvCacheDeviceAddr, aclDataType::ACL_FLOAT16, &ckvCache);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  ret = CreateAclTensor(kRopeHostData, kRopeShape, &kRopeDeviceAddr, aclDataType::ACL_FLOAT16, &kRope);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  ret = CreateAclTensor(cKvHostData, cKvShape, &cKvDeviceAddr, aclDataType::ACL_FLOAT16, &cKv);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  // 3. 调用CANN算子库API，需要修改为具体的API
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;

  // 调用aclnnKvRmsNormRopeCache第一段接口
  ret = aclnnKvRmsNormRopeCacheGetWorkspaceSize(kv,gamma,cos,sin,index,
                                                kpeCache,ckvCache,nullptr,nullptr,nullptr,nullptr,epsilon,cacheMode,isOutputKv,kRope,cKv,&workspaceSize,&executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnKvRmsNormRopeCacheGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);

  // 根据第一段接口计算出的workspaceSize申请device内存
  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); return ret);
  }

  // 调用aclnnKvRmsNormRopeCache第二段接口
  ret = aclnnKvRmsNormRopeCache(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnKvRmsNormRopeCache failed. ERROR: %d\n", ret); return ret);

  // 4. 固定写法，同步等待任务执行结束
  ret = aclrtSynchronizeStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);

  // 5. 获取输出的值，将device侧内存上的结果拷贝至host侧，需要根据具体API的接口定义修改
  PrintOutResult(kpeCacheShape, &kpeCacheDeviceAddr);
  PrintOutResult(ckvCacheShape, &ckvCacheDeviceAddr);

  // 6. 释放aclTensor和aclScalar，需要根据具体API的接口定义修改
  aclDestroyTensor(kv);
  aclDestroyTensor(gamma);
  aclDestroyTensor(cos);
  aclDestroyTensor(sin);
  aclDestroyTensor(index);
  aclDestroyTensor(kpeCache);
  aclDestroyTensor(ckvCache);
  aclDestroyTensor(kRope);
  aclDestroyTensor(cKv);

  // 7. 释放device 资源
  aclrtFree(kvDeviceAddr);
  aclrtFree(gammaDeviceAddr);
  aclrtFree(cosDeviceAddr);
  aclrtFree(sinDeviceAddr);
  aclrtFree(indexDeviceAddr);
  aclrtFree(kpeCacheDeviceAddr);
  aclrtFree(ckvCacheDeviceAddr);
  aclrtFree(kRopeDeviceAddr);
  aclrtFree(cKvDeviceAddr);


  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();

  return 0;
}