/**
 * 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_addr.h"
#include "../../add/op_api/add.h"
#include "addr.h"
#include "../../mul/op_api/mul.h"
#include "../../logical_or/op_api/logical_or.h"
#include "../../logical_and/op_api/logical_and.h"
#include "aclnn_kernels/cast.h"
#include "conversion/unsqueeze/op_host/op_api/unsqueeze.h"
#include "aclnn_kernels/contiguous.h"
#include "opdev/op_log.h"
#include "opdev/op_dfx.h"
#include "opdev/common_types.h"
#include "opdev/data_type_utils.h"
#include "opdev/make_op_executor.h"
#include "aclnn_kernels/common/op_error_check.h"
#include "op_api/aclnn_check.h"

using namespace op;

#ifdef __cplusplus
extern "C" {
#endif

#define MAX_ADDR_INPUT_DIMS_NUMS 2
#define DEFAULT_OUTER_VEC_DIMS_NUMS 1

enum OuterExpandMode
{
    All,
    IgnoreInput,
    IgnoreInputScaling
};

static const std::initializer_list<DataType> ASCEND910_DTYPE_SUPPORT_LIST = {
    DataType::DT_DOUBLE, DataType::DT_FLOAT, DataType::DT_FLOAT16, DataType::DT_INT64, DataType::DT_INT32,
    DataType::DT_INT16,  DataType::DT_INT8,  DataType::DT_UINT8,   DataType::DT_BOOL};

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

static const std::initializer_list<DataType> ASCEND950_ADDR_DTYPE_SUPPORT_LIST = {
    DataType::DT_FLOAT, DataType::DT_FLOAT16, DataType::DT_BF16,
    DataType::DT_INT8,  DataType::DT_UINT8,   DataType::DT_BOOL};

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

static bool IsSupportAddr(op::DataType hightDtype)
{
    if (IsRegBase()) {
        return CheckType(hightDtype, ASCEND950_ADDR_DTYPE_SUPPORT_LIST);
    }
    return false;
}

static bool CheckBetaAndAlphaDtyeValid(
    const aclScalar* beta, const aclScalar* alpha, DataType highDtype, bool allIntInputs)
{
    // beta和alpha为bool类型值时，其他入参只能是bool类型
    if (beta && highDtype != DataType::DT_BOOL && beta->GetDataType() == DataType::DT_BOOL) {
        OP_LOGE(ACLNN_ERR_PARAM_INVALID, "Boolean beta only supported for Boolean results.");
        return false;
    }

    if (alpha && highDtype != DataType::DT_BOOL && alpha->GetDataType() == DataType::DT_BOOL) {
        OP_LOGE(ACLNN_ERR_PARAM_INVALID, "Boolean alpha only supported for Boolean results.");
        return false;
    }

    // 如果输入都是整型tensor，那beta和alpha不能是浮点型数
    if (allIntInputs && beta && !IsIntegralType(beta->GetDataType(), true)) {
        OP_LOGE(
            ACLNN_ERR_PARAM_INVALID, "For integral input tensors, argument beta must not be a floating point number");
        return false;
    }

    if (allIntInputs && alpha && !IsIntegralType(alpha->GetDataType(), true)) {
        OP_LOGE(
            ACLNN_ERR_PARAM_INVALID, "For integral input tensors, argument alpha must not be a floating point number");
        return false;
    }

    return true;
}

static inline const aclTensor* vecUnsqueezeWithDim(const aclTensor* vec, int64_t dim, aclOpExecutor* executor)
{
    const int64_t appendDim[] = {dim};
    auto dims = executor->AllocIntArray(appendDim, 1);
    auto vecMul = l0op::UnsqueezeNd(vec, dims, executor);
    CHECK_RET(vecMul != nullptr, nullptr);

    return vecMul;
}

static bool CheckNotNull(const aclTensor* self, const aclTensor* vec1, const aclTensor* vec2, aclTensor* out)
{
    OP_CHECK_NULL(self, return false);
    OP_CHECK_NULL(vec1, return false);
    OP_CHECK_NULL(vec2, return false);
    OP_CHECK_NULL(out, return false);

    return true;
}

static bool CheckDtypeValid(
    const aclTensor* self, const aclTensor* vec1, const aclTensor* vec2, const aclScalar* beta, const aclScalar* alpha,
    aclTensor* out)
{
    const auto& supportList = GetDtypeSupportList();
    // 检查self的数据类型是否在支持列表内
    OP_CHECK_DTYPE_NOT_SUPPORT(self, supportList, return false);

    // 检查vec1的数据类型是否在支持列表内
    OP_CHECK_DTYPE_NOT_SUPPORT(vec1, supportList, return false);

    // 检查vec2的数据类型是否在支持列表内
    OP_CHECK_DTYPE_NOT_SUPPORT(vec2, supportList, return false);

    // 检查out的数据类型是否在支持列表内
    OP_CHECK_DTYPE_NOT_SUPPORT(out, supportList, return false);

    // 检查beta的数据类型是否在支持列表内，beta为空例外
    if (beta != nullptr) {
        OP_CHECK_DTYPE_NOT_SUPPORT(beta, supportList, return false);
    }

    // 检查alpha的数据类型是否在支持列表内，alpha为空例外
    if (alpha != nullptr) {
        OP_CHECK_DTYPE_NOT_SUPPORT(alpha, supportList, return false);
    }

    // 检查beta和alpha dtype是否符合要求
    auto hightDtype = op::PromoteType(self->GetDataType(), op::PromoteType(vec2->GetDataType(), vec1->GetDataType()));
    bool allIntInputs =
        (IsIntegralType(self->GetDataType(), true) && IsIntegralType(vec1->GetDataType(), true) &&
         IsIntegralType(vec2->GetDataType(), true));
    auto optValid = CheckBetaAndAlphaDtyeValid(beta, alpha, hightDtype, allIntInputs);
    CHECK_RET(optValid, false);

    return true;
}

static bool CheckShape(const aclTensor* self, const aclTensor* vec1, const aclTensor* vec2, aclTensor* out)
{
    // vec1和vec2只能是1维
    OP_CHECK_WRONG_DIMENSION(vec1, 1, return false);
    OP_CHECK_WRONG_DIMENSION(vec2, 1, return false);

    // self的维度不能超过2
    OP_CHECK_MAX_DIM(self, MAX_ADDR_INPUT_DIMS_NUMS, return false);

    if (self == out) { // 如果out就是self，则self只能是2维, for addr_
        OP_CHECK_WRONG_DIMENSION(self, MAX_ADDR_INPUT_DIMS_NUMS, return false);
    }

    // 检查self与vec1和vec2外积结果之间是否能broadcast；add_校验更严格，不需要进行broadcast
    op::Shape outerShape = {(vec1->GetViewShape())[0], (vec2->GetViewShape())[0]};
    op::Shape broadcastShape = self->GetViewShape();
    if (self != out && !BroadcastInferShape(self->GetViewShape(), outerShape, broadcastShape)) {
        OP_LOGE(
            ACLNN_ERR_PARAM_INVALID, "the size of tensor self %s must match the size of tensor outer %s.",
            op::ToString(self->GetViewShape()).GetString(), op::ToString(outerShape).GetString());
        return false;
    }

    // broadcast之后的tensor size必须要与外积相同
    if (broadcastShape != outerShape) {
        OP_LOGE(
            ACLNN_ERR_PARAM_INVALID, "size mismatch, input: %s, v1: %s, v2: %s",
            op::ToString(broadcastShape).GetString(), op::ToString(vec1->GetViewShape()).GetString(),
            op::ToString(vec2->GetViewShape()).GetString());
        return false;
    }

    return true;
}

static aclnnStatus CheckParams(
    const aclTensor* self, const aclTensor* vec1, const aclTensor* vec2, const aclScalar* beta, const aclScalar* alpha,
    aclTensor* out)
{
    // 1. 检查参数是否为空指针
    CHECK_RET(CheckNotNull(self, vec1, vec2, 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内，需要根据api定义校验
    CHECK_RET(CheckDtypeValid(self, vec1, vec2, beta, alpha, out), ACLNN_ERR_PARAM_INVALID);

    // 3. 检查shape是否满足约束
    CHECK_RET(CheckShape(self, vec1, vec2, out), ACLNN_ERR_PARAM_INVALID);

    return ACLNN_SUCCESS;
}

static const aclTensor* addrStubMul(
    const aclTensor* left, const aclTensor* right, DataType highDtype, aclOpExecutor* executor)
{
    const aclTensor* mulOut = nullptr;
    if (highDtype == DataType::DT_BOOL) {
        mulOut = l0op::LogicalAnd(left, right, executor);
    } else {
        mulOut = l0op::Mul(left, right, executor);
    }

    return mulOut;
}

static const aclTensor* addrStubAdd(
    const aclTensor* left, const aclTensor* right, DataType highDtype, aclOpExecutor* executor)
{
    const aclTensor* addOut = nullptr;
    if (highDtype == DataType::DT_BOOL) {
        addOut = l0op::LogicalOr(left, right, executor);
    } else {
        addOut = l0op::Add(left, right, executor);
    }

    return addOut;
}

static aclnnStatus addrOutHandle(
    const aclTensor* outCalcuResult, aclTensor* out, aclOpExecutor* executor, DataType hightDtype)
{
    OP_CHECK_SHAPE_NOT_EQUAL_WITH_EXPECTED_SIZE(out, outCalcuResult->GetViewShape(), return ACLNN_ERR_PARAM_INVALID);
    // 输出类型转换成out dtype类型
    auto temp = outCalcuResult;
    if (hightDtype == DataType::DT_BOOL) {
        temp = l0op::Cast(temp, DataType::DT_INT8, executor);
        CHECK_RET(temp != nullptr, ACLNN_ERR_INNER_NULLPTR);
        temp = l0op::Cast(temp, DataType::DT_BOOL, executor);
        CHECK_RET(temp != nullptr, ACLNN_ERR_INNER_NULLPTR);
    }
    auto addrOutCast = l0op::Cast(temp, out->GetDataType(), executor);
    CHECK_RET(addrOutCast != nullptr, ACLNN_ERR_INNER_NULLPTR);

    // 计算结果拷贝，支持非连续
    auto viewCopyResult = l0op::ViewCopy(addrOutCast, out, executor);
    CHECK_RET(viewCopyResult != nullptr, ACLNN_ERR_INNER_NULLPTR);

    return ACLNN_SUCCESS;
}

static aclnnStatus addrStub(
    const aclTensor* selfCast, const aclTensor* vec1Cast, const aclTensor* vec2Cast, const aclScalar* beta,
    const aclScalar* alpha, aclTensor* out, aclOpExecutor* executor, const op::DataType& hightDtype)
{
    // vec1要转置下，从1行m列扩展成m行1列，才能进行外积运算；vec2相应的也扩展成1行n列的二维tensor
    auto vecMul1 = vecUnsqueezeWithDim(vec1Cast, 1, executor);
    CHECK_RET(vecMul1 != nullptr, ACLNN_ERR_INNER_NULLPTR);
    auto vecMul2 = vecUnsqueezeWithDim(vec2Cast, 0, executor);
    CHECK_RET(vecMul2 != nullptr, ACLNN_ERR_INNER_NULLPTR);

    // 计算vec1和vec2外积
    auto outerVec = addrStubMul(vecMul1, vecMul2, hightDtype, executor);
    CHECK_RET(outerVec != nullptr, ACLNN_ERR_INNER_NULLPTR);

    // 处理外积比例因子, alpha * outer(vec1, vec2)
    // alpha为nullptr，以默认值做处理
    auto mulOuterVec = (const aclTensor*)outerVec;
    if (alpha != nullptr && alpha->ToFloat() != 1) {
        auto alphaTensor = executor->ConvertToTensor(alpha, hightDtype);
        mulOuterVec = addrStubMul(outerVec, alphaTensor, hightDtype, executor);
        CHECK_RET(mulOuterVec != nullptr, ACLNN_ERR_INNER_NULLPTR);
    }

    // 处理扩展矩阵, beta * self + alpha * outer(vec1, vec2)
    OuterExpandMode expandMode = All;
    if (beta != nullptr) {
        if (beta->ToFloat() == 0) { // beta为false或者0，则忽略self值
            expandMode = IgnoreInput;
        } else if (beta->ToFloat() == 1) { // beta为true或者1，直接进行矩阵扩展，无需考虑beta比例因子
            expandMode = IgnoreInputScaling;
        }
    } else { // beta为nullptr，以默认值做处理，不考虑beta比例因子
        expandMode = IgnoreInputScaling;
    }

    if (expandMode == IgnoreInput) {
        return addrOutHandle(mulOuterVec, out, executor, DataType::DT_MAX);
    }

    auto selfScaling = selfCast;
    if (expandMode == All) {
        auto betaTensor = executor->ConvertToTensor(beta, hightDtype);
        selfScaling = addrStubMul(selfCast, betaTensor, hightDtype, executor);
        CHECK_RET(selfScaling != nullptr, ACLNN_ERR_INNER_NULLPTR);
    }

    auto addrOut = addrStubAdd(selfScaling, mulOuterVec, hightDtype, executor);
    CHECK_RET(addrOut != nullptr, ACLNN_ERR_INNER_NULLPTR);
    return addrOutHandle(addrOut, out, executor, DataType::DT_MAX);
}

static aclnnStatus addrProc(
    const aclTensor* self, const aclTensor* vec1, const aclTensor* vec2, const aclScalar* beta, const aclScalar* alpha,
    aclTensor* out, aclOpExecutor* executor)
{
    // 计算输入最高类型
    auto hightDtype = op::PromoteType(self->GetDataType(), op::PromoteType(vec2->GetDataType(), vec1->GetDataType()));

    // 连续性转换
    auto selfContiguous = l0op::Contiguous(self, executor);
    CHECK_RET(selfContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);
    auto vec1Contiguous = l0op::Contiguous(vec1, executor);
    CHECK_RET(vec1Contiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);
    auto vec2Contiguous = l0op::Contiguous(vec2, executor);
    CHECK_RET(vec2Contiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);

    // 类型转换，都转换成最高类型进行计算
    auto selfCast = l0op::Cast(selfContiguous, hightDtype, executor);
    CHECK_RET(selfCast != nullptr, ACLNN_ERR_INNER_NULLPTR);
    auto vec1Cast = l0op::Cast(vec1Contiguous, hightDtype, executor);
    CHECK_RET(vec1Cast != nullptr, ACLNN_ERR_INNER_NULLPTR);
    auto vec2Cast = l0op::Cast(vec2Contiguous, hightDtype, executor);
    CHECK_RET(vec2Cast != nullptr, ACLNN_ERR_INNER_NULLPTR);

    const aclTensor* alphaTensor;
    const aclTensor* betaTensor;
    if (alpha != nullptr) {
        alphaTensor = executor->ConvertToTensor(alpha, hightDtype);
    } else {
        const aclScalar* valueScalar = executor->AllocScalar(1);
        alphaTensor = executor->ConvertToTensor(valueScalar, hightDtype);
    }
    if (beta != nullptr) {
        betaTensor = executor->ConvertToTensor(beta, hightDtype);
    } else {
        const aclScalar* valueScalar = executor->AllocScalar(1);
        betaTensor = executor->ConvertToTensor(valueScalar, hightDtype);
    }

    if (IsSupportAddr(hightDtype)) {
        auto addrOut = l0op::Addr(selfCast, vec1Cast, vec2Cast, betaTensor, alphaTensor, hightDtype, executor);
        CHECK_RET(addrOut != nullptr, ACLNN_ERR_INNER_NULLPTR);
        return addrOutHandle(addrOut, out, executor, hightDtype);
    } else {
        return addrStub(selfCast, vec1Cast, vec2Cast, beta, alpha, out, executor, hightDtype);
    }
}

aclnnStatus aclnnAddrGetWorkspaceSize(
    const aclTensor* self, const aclTensor* vec1, const aclTensor* vec2, const aclScalar* betaOptional,
    const aclScalar* alphaOptional, aclTensor* out, uint64_t* workspaceSize, aclOpExecutor** executor)
{
    L2_DFX_PHASE_1(aclnnAddr, DFX_IN(self, vec1, vec2, betaOptional, alphaOptional), 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, vec1, vec2, betaOptional, alphaOptional, out);
    CHECK_RET(ret == ACLNN_SUCCESS, ret);

    // 空Tensor处理
    if (vec1->IsEmpty() || vec2->IsEmpty()) {
        *workspaceSize = 0;
        uniqueExecutor.ReleaseTo(executor);
        return ACLNN_SUCCESS;
    }

    // 开始addr 计算
    auto addrRet = addrProc(self, vec1, vec2, betaOptional, alphaOptional, out, uniqueExecutor.get());
    CHECK_RET(addrRet == ACLNN_SUCCESS, addrRet);

    // 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 aclnnAddr(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, const aclrtStream stream)
{
    L2_DFX_PHASE_2(aclnnAddr);

    return CommonOpExecutorRun(workspace, workspaceSize, executor, stream);
}

aclnnStatus aclnnInplaceAddrGetWorkspaceSize(
    aclTensor* selfRef, const aclTensor* vec1, const aclTensor* vec2, const aclScalar* betaOptional,
    const aclScalar* alphaOptional, uint64_t* workspaceSize, aclOpExecutor** executor)
{
    return aclnnAddrGetWorkspaceSize(
        selfRef, vec1, vec2, betaOptional, alphaOptional, selfRef, workspaceSize, executor);
}

aclnnStatus aclnnInplaceAddr(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, const aclrtStream stream)
{
    L2_DFX_PHASE_2(aclnnInplaceAddr);

    return CommonOpExecutorRun(workspace, workspaceSize, executor, stream);
}

#ifdef __cplusplus
}
#endif