/**
 * 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 "aclnn_quant_scatter.h"
#include "aclnn_kernels/cast.h"
#include "aclnn_kernels/contiguous.h"
#include "quant_update_scatter.h"
#include "aclnn_kernels/common/op_error_check.h"
#include "aclnn/aclnn_base.h"
#include "opdev/common_types.h"
#include "opdev/data_type_utils.h"
#include "opdev/format_utils.h"
#include "opdev/op_dfx.h"
#include "opdev/op_executor.h"
#include "opdev/op_log.h"
#include "opdev/shape_utils.h"
#include "opdev/tensor_view_utils.h"
#include "opdev/platform.h"

using namespace op;
#ifdef __cplusplus
extern "C" {
#endif

static const std::initializer_list<DataType> DTYPE_SUPPORT_LIST_SELFREF = {DataType::DT_INT8};

static const std::initializer_list<DataType> DTYPE_SUPPORT_LIST_INDICES = {DataType::DT_INT32, DataType::DT_INT64};

static const std::initializer_list<DataType> DTYPE_SUPPORT_LIST_UPDATES = {DataType::DT_BF16, DataType::DT_FLOAT16};

static const std::initializer_list<DataType> DTYPE_SUPPORT_LIST_QUANT_SCALES = {DataType::DT_BF16, DataType::DT_FLOAT};

static const std::initializer_list<DataType> DTYPE_SUPPORT_LIST_QUANT_ZERO_POINTS = {DataType::DT_BF16,
                                                                                     DataType::DT_INT32};

static inline bool CheckNotNull(const aclTensor* selfRef, const aclTensor* indices, const aclTensor* updates,
                                const aclTensor* quantScales) {
  OP_CHECK_NULL(selfRef, return false);
  OP_CHECK_NULL(indices, return false);
  OP_CHECK_NULL(updates, return false);
  OP_CHECK_NULL(quantScales, return false);
  return true;
}

static inline bool CheckDtypeCombineValid(const aclTensor* updates, const aclTensor* quantScales,
                                          const aclTensor* quantZeroPoints) {
  DataType updatesDtype = updates->GetDataType();
  DataType quantScalesDtype = quantScales->GetDataType();

  if (quantZeroPoints) {
    DataType quantZeroPointsDtype = quantZeroPoints->GetDataType();
    if (updatesDtype == DataType::DT_FLOAT16 && quantScalesDtype == DataType::DT_FLOAT &&
        quantZeroPointsDtype == DataType::DT_INT32) {
      return true;
    } else if (updatesDtype == DataType::DT_BF16 && quantScalesDtype == DataType::DT_BF16 &&
               quantZeroPointsDtype == DataType::DT_BF16) {
      return true;
    } else {
      OP_LOGE(ACLNN_ERR_PARAM_INVALID,
              "(updates,quantScales,quantZeroPoints) Dtype should be (float16, float, int32) or (bfloat16, bfloat16, "
              "bfloat16), but is (%s, %s, %s).",
              op::ToString(updates->GetDataType()).GetString(), op::ToString(quantScales->GetDataType()).GetString(),
              op::ToString(quantZeroPoints->GetDataType()).GetString());
      return false;
    }
  } else {
    if (updatesDtype == DataType::DT_FLOAT16 && quantScalesDtype == DataType::DT_FLOAT) {
      return true;
    } else if (updatesDtype == DataType::DT_BF16 && quantScalesDtype == DataType::DT_BF16) {
      return true;
    } else {
      OP_LOGE(ACLNN_ERR_PARAM_INVALID,
              "(updates,quantScales) Dtype should be (float16, float) or (bfloat16, bfloat16), but is (%s, %s).",
              op::ToString(updates->GetDataType()).GetString(), op::ToString(quantScales->GetDataType()).GetString());
      return false;
    }
  }
}

static inline bool CheckDtypeValid(const aclTensor* selfRef, const aclTensor* indices, const aclTensor* updates,
                                   const aclTensor* quantScales, const aclTensor* quantZeroPoints) {
  OP_CHECK_DTYPE_NOT_SUPPORT(selfRef, DTYPE_SUPPORT_LIST_SELFREF, return false);
  OP_CHECK_DTYPE_NOT_SUPPORT(indices, DTYPE_SUPPORT_LIST_INDICES, return false);
  OP_CHECK_DTYPE_NOT_SUPPORT(updates, DTYPE_SUPPORT_LIST_UPDATES, return false);
  OP_CHECK_DTYPE_NOT_SUPPORT(quantScales, DTYPE_SUPPORT_LIST_QUANT_SCALES, return false);

  if (quantZeroPoints) {
    OP_CHECK_DTYPE_NOT_SUPPORT(quantZeroPoints, DTYPE_SUPPORT_LIST_QUANT_ZERO_POINTS, return false);
  }

  if (!CheckDtypeCombineValid(updates, quantScales, quantZeroPoints)) {
    return false;
  }
  return true;
}

static inline bool CheckShape(const aclTensor* selfRef, const aclTensor* updates) {
  if (selfRef->GetViewShape().GetDimNum() != updates->GetViewShape().GetDimNum()) {
    OP_LOGE(ACLNN_ERR_PARAM_INVALID, "dimension of selfRef and updates should equal");
    return false;
  }
  return true;
}

static void CheckFormatNZ(const aclTensor* x)
{
    if (x == nullptr) { return; }
    op::Format format = x->GetStorageFormat();
    if (format == Format::FORMAT_FRACTAL_NZ) {
        OP_LOGW("Format of input gets [%s], this format may lead to precision failure",
        op::ToString(format).GetString());
    }
}

static aclnnStatus CheckParams(const aclTensor* selfRef, const aclTensor* indices, const aclTensor* updates,
                               const aclTensor* quantScales, const aclTensor* quantZeroPoints) {
  // 1. 检查参数是否为空指针
  CHECK_RET(CheckNotNull(selfRef, indices, updates, quantScales), ACLNN_ERR_PARAM_NULLPTR);

  // 2. 检查参数的数据类型是否符合预期
  CHECK_RET(CheckDtypeValid(selfRef, indices, updates, quantScales, quantZeroPoints), ACLNN_ERR_PARAM_INVALID);

  // 3. 检查输入tensor的shape
  CHECK_RET(CheckShape(selfRef, updates), ACLNN_ERR_PARAM_INVALID);

  // 4. 检查输入是否为NZ，若为NZ提示warning
  CheckFormatNZ(selfRef);
  CheckFormatNZ(indices);
  CheckFormatNZ(updates);
  CheckFormatNZ(quantScales);
  CheckFormatNZ(quantZeroPoints);

  return ACLNN_SUCCESS;
}

aclnnStatus aclnnInplaceQuantScatterGetWorkspaceSize(aclTensor* selfRef, const aclTensor* indices,
                                                     const aclTensor* updates, const aclTensor* quantScales,
                                                     const aclTensor* quantZeroPoints, int64_t axis, int64_t quantAxis,
                                                     int64_t reduction, uint64_t* workspaceSize,
                                                     aclOpExecutor** executor) {
  L2_DFX_PHASE_1(aclnnInplaceQuantScatter,
                 DFX_IN(selfRef, indices, updates, quantScales, quantZeroPoints, axis, quantAxis, reduction),
                 DFX_OUT(selfRef));

  // 固定写法，创建OpExecutor
  auto uniqueExecutor = CREATE_EXECUTOR();
  CHECK_RET(uniqueExecutor.get() != nullptr, ACLNN_ERR_INNER_CREATE_EXECUTOR);

  // 参数检查
  auto ret = CheckParams(selfRef, indices, updates, quantScales, quantZeroPoints);
  CHECK_RET(ret == ACLNN_SUCCESS, ret);

  if (selfRef->IsEmpty() || indices->IsEmpty() || updates->IsEmpty() || quantScales->IsEmpty()) {
    *workspaceSize = 0;
    uniqueExecutor.ReleaseTo(executor);
    return ACLNN_SUCCESS;
  }

  // 将输入selfRef转换成连续的tensor
  auto selfRefContiguous = l0op::Contiguous(selfRef, uniqueExecutor.get());
  CHECK_RET(selfRefContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);

  // 将输入indices转换成连续的tensor
  auto indicesContiguous = l0op::Contiguous(indices, uniqueExecutor.get());
  CHECK_RET(indicesContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);

  // 将输入updates转换成连续的tensor
  auto updatesContiguous = l0op::Contiguous(updates, uniqueExecutor.get());
  CHECK_RET(updatesContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);

  // 将输入quantScales转换成连续的tensor
  auto quantScalesContiguous = l0op::Contiguous(quantScales, uniqueExecutor.get());
  CHECK_RET(quantScalesContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);

  const aclTensor* quantZeroPointsContiguous = nullptr;
  if (quantZeroPoints) {
    // 将输入quantZeroPoints转换成连续的tensor
    quantZeroPointsContiguous = l0op::Contiguous(quantZeroPoints, uniqueExecutor.get());
    CHECK_RET(quantZeroPointsContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);
  }

  const std::string reduce = "update";

  // 执行L0算子
  auto quantUpdateScatterResult =
      l0op::QuantUpdateScatter(selfRefContiguous, indicesContiguous, updatesContiguous, quantScalesContiguous,
                               quantZeroPointsContiguous, reduce, axis, quantAxis, uniqueExecutor.get());
  CHECK_RET(quantUpdateScatterResult != nullptr, ACLNN_ERR_INNER_NULLPTR);

  // 将计算结果拷贝到输出data上
  auto viewCopyResult = l0op::ViewCopy(quantUpdateScatterResult, selfRef, uniqueExecutor.get());
  CHECK_RET(viewCopyResult != nullptr, ACLNN_ERR_INNER_NULLPTR);

  // 获取计算过程中需要使用的workspace大小
  *workspaceSize = uniqueExecutor->GetWorkspaceSize();
  uniqueExecutor.ReleaseTo(executor);
  return ACLNN_SUCCESS;
}

aclnnStatus aclnnInplaceQuantScatter(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor,
                                     aclrtStream stream) {
  L2_DFX_PHASE_2(aclnnInplaceQuantScatter);
  return CommonOpExecutorRun(workspace, workspaceSize, executor, stream);
}

#ifdef __cplusplus
}
#endif