/**
 * 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 "signbit.h"
#include "aclnn_signbit.h"
#include "math/less/op_api/less.h"
#include "aclnn_kernels/cast.h"
#include "conversion/fill/op_api/fill.h"
#include "aclnn_kernels/contiguous.h"
#include "aclnn/aclnn_base.h"
#include "aclnn_kernels/common/op_error_check.h"
#include "opdev/common_types.h"
#include "opdev/data_type_utils.h"
#include "opdev/shape_utils.h"
#include "opdev/format_utils.h"
#include "opdev/op_dfx.h"
#include "opdev/op_executor.h"
#include "opdev/op_log.h"
#include "opdev/platform.h"
#include "opdev/tensor_view_utils.h"
#include "op_api/aclnn_check.h"

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

constexpr size_t MAX_DIM_LEN = 8;

// 根据API定义，需要列出所能支持的所有dtype
static const std::initializer_list<op::DataType> ASCEND910_DTYPE_SUPPORT_LIST = {
    op::DataType::DT_FLOAT,  op::DataType::DT_INT32,  op::DataType::DT_INT64, op::DataType::DT_FLOAT16,
    op::DataType::DT_INT16,  op::DataType::DT_INT8,   op::DataType::DT_UINT8, op::DataType::DT_DOUBLE,
    op::DataType::DT_UINT32, op::DataType::DT_UINT64, op::DataType::DT_BOOL,  op::DataType::DT_UINT16};

static const std::initializer_list<op::DataType> ASCEND910B_DTYPE_SUPPORT_LIST = {
    op::DataType::DT_FLOAT,  op::DataType::DT_INT32,  op::DataType::DT_INT64, op::DataType::DT_FLOAT16,
    op::DataType::DT_INT16,  op::DataType::DT_INT8,   op::DataType::DT_UINT8, op::DataType::DT_DOUBLE,
    op::DataType::DT_UINT32, op::DataType::DT_UINT64, op::DataType::DT_BOOL,  op::DataType::DT_UINT16,
    op::DataType::DT_BF16};

static const std::initializer_list<op::DataType> SIGNBIT_DTYPE_SUPPORT_LIST = {
    op::DataType::DT_FLOAT,  op::DataType::DT_INT32,  op::DataType::DT_INT64, op::DataType::DT_FLOAT16,
    op::DataType::DT_INT8,   op::DataType::DT_UINT8, op::DataType::DT_DOUBLE, op::DataType::DT_UINT64,
    op::DataType::DT_BOOL, op::DataType::DT_BF16};

static bool CanUseSignbit(const aclTensor* self)
{
    return IsRegBase() && CheckType(self->GetDataType(), SIGNBIT_DTYPE_SUPPORT_LIST);
}

static bool CheckNotNull(const aclTensor* self, const aclTensor* out)
{
    OP_CHECK_NULL(self, return false);
    OP_CHECK_NULL(out, return false);
    return true;
}

static inline const std::initializer_list<op::DataType>& GetDtypeSupportList()
{
    if (GetCurrentPlatformInfo().GetSocVersion() >= SocVersion::ASCEND910B &&
        GetCurrentPlatformInfo().GetSocVersion() <= SocVersion::ASCEND910E) {
        return ASCEND910B_DTYPE_SUPPORT_LIST;
    } else {
        return ASCEND910_DTYPE_SUPPORT_LIST;
    }
}

static bool CheckDtypeValid(const aclTensor* self, const aclTensor* out)
{
    auto supportList = GetDtypeSupportList();
    // 检查self的数据类型是否在Less算子的支持列表内
    OP_CHECK_DTYPE_NOT_SUPPORT(self, supportList, return false);

    // 检查out的数据类型是否在Less算子的支持列表内
    OP_CHECK_DTYPE_NOT_SUPPORT(out, supportList, return false);
    return true;
}

static bool CheckPromoteType(const aclTensor* out)
{
    // 检查BOOL类型能否转换为输出的数据类型(算子返回的都是BOOL类型)
    OP_CHECK_RESULT_DTYPE_CAST_FAILED(DataType::DT_BOOL, out->GetDataType(), return false);
    return true;
}

static bool CheckShape(const aclTensor* self, const aclTensor* out)
{
    OP_CHECK_MAX_DIM(self, MAX_DIM_LEN, return false);
    OP_CHECK_MAX_DIM(out, MAX_DIM_LEN, return false);
    OP_CHECK_SHAPE_NOT_EQUAL(self, out, return false);

    return true;
}

static aclnnStatus CheckParams(const aclTensor* self, const aclTensor* out)
{
    // 1. 检查参数是否为空指针
    CHECK_RET(CheckNotNull(self, out), ACLNN_ERR_PARAM_NULLPTR);

    // 2. 检查输入的数据类型是否在API支持的数据类型范围之内，需要根据api定义校验
    CHECK_RET(CheckDtypeValid(self, out), ACLNN_ERR_PARAM_INVALID);

    // 3. 检查self的数据类型能否转换为输出数据类型
    CHECK_RET(CheckPromoteType(out), ACLNN_ERR_PARAM_INVALID);

    // 4. 检查shape
    CHECK_RET(CheckShape(self, out), ACLNN_ERR_PARAM_INVALID);

    return ACLNN_SUCCESS;
}

static aclnnStatus FillScalar(aclTensor* out, bool val, aclOpExecutor* executor)
{
    FVector<int64_t> shape;
    size_t dimNum = out->GetViewShape().GetDimNum();
    for (size_t idx = 0; idx < dimNum; idx++) {
        int64_t tmpVal = out->GetViewShape().GetDim(idx);
        shape.push_back(tmpVal);
    }
    auto dims = executor->ConvertToTensor(shape.data(), shape.size(), DataType::DT_INT64);
    auto shapeArray = executor->AllocIntArray(shape.data(), shape.size());

    FVector<bool> valVector = {val};
    auto valTensor = executor->ConvertToTensor(valVector.data(), valVector.size(), out->GetDataType());
    auto fillOut = l0op::Fill(dims, valTensor, shapeArray, executor);
    CHECK_RET(fillOut != nullptr, ACLNN_ERR_INNER_NULLPTR);
    auto viewCopyResult = l0op::ViewCopy(fillOut, out, executor);
    CHECK_RET(viewCopyResult != nullptr, ACLNN_ERR_INNER_NULLPTR);
    return ACLNN_SUCCESS;
}

aclnnStatus aclnnSignbitGetWorkspaceSize(
    const aclTensor* self, aclTensor* out, uint64_t* workspaceSize, aclOpExecutor** executor)
{
    L2_DFX_PHASE_1(aclnnSignbit, DFX_IN(self), DFX_OUT(out));

    // 固定写法，创建OpExecutor
    auto uniqueExecutor = CREATE_EXECUTOR();
    CHECK_RET(uniqueExecutor.get() != nullptr, ACLNN_ERR_INNER_CREATE_EXECUTOR);
    // 固定写法，参数检查
    auto ret = CheckParams(self, out);
    CHECK_RET(ret == ACLNN_SUCCESS, ret);

    // Less算子的空tensor在kernel中支持，对标竞品根据算子实际情况补充
    if (self->IsEmpty()) {
        *workspaceSize = 0;
        uniqueExecutor.ReleaseTo(executor);
        return ACLNN_SUCCESS;
    }

    // 固定写法，将输入self转换成连续的tensor
    auto selfContiguous = l0op::Contiguous(self, uniqueExecutor.get());
    CHECK_RET(selfContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);

    if (self->GetDataType() == op::DataType::DT_BOOL) {
        FillScalar(out, false, uniqueExecutor.get());
        *workspaceSize = uniqueExecutor->GetWorkspaceSize();
        uniqueExecutor.ReleaseTo(executor);
        return ret;
    }

    const aclTensor* signBitOpOut = nullptr;
    if (CanUseSignbit(self)) {
        // 调用Signbit可以解决-0精度问题
        signBitOpOut = l0op::Signbit(selfContiguous, uniqueExecutor.get());
    } else {
        // 创建数据为0的tensor
        FVector<float> zeroVector = {0};
        auto zeroTensor = uniqueExecutor.get()->ConvertToTensor(zeroVector.data(), zeroVector.size(), self->GetDataType());
        CHECK_RET(zeroTensor != nullptr, ACLNN_ERR_INNER_NULLPTR);
        signBitOpOut = l0op::Less(selfContiguous, zeroTensor, uniqueExecutor.get());
    }
    CHECK_RET(signBitOpOut != nullptr, ACLNN_ERR_INNER_NULLPTR);

    // 固定写法，将计算结果(BOOL)转换成输出out的数据类型
    auto castOut = l0op::Cast(signBitOpOut, out->GetDataType(), uniqueExecutor.get());
    CHECK_RET(castOut != nullptr, ACLNN_ERR_INNER_NULLPTR);

    // 固定写法，将计算结果拷贝到输出out上，out可能是非连续的tensor
    auto viewCopyResult = l0op::ViewCopy(castOut, out, uniqueExecutor.get());
    CHECK_RET(viewCopyResult != nullptr, ACLNN_ERR_INNER_NULLPTR);

    // 固定写法，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 aclnnSignbit(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, aclrtStream stream)
{
    L2_DFX_PHASE_2(aclnnSignbit);
    // 固定写法，调用框架能力，完成计算
    return CommonOpExecutorRun(workspace, workspaceSize, executor, stream);
}

#ifdef __cplusplus
}
#endif