/**
 * This program is free software, you can redistribute it and/or modify it.
 * Copyright (c) 2025 Huawei Technologies Co., Ltd.
 * This file is a part of the CANN Open Software.
 * Licensed under 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_max_unpool2d.h"
#include "aclnn_kernels/cast.h"
#include "aclnn_kernels/contiguous.h"
#include "index/scatter_elements_v2/op_api/scatter_elements.h"
#include "level0/zero_op.h"
#include "aclnn_kernels/reshape.h"
#include "opdev/common_types.h"
#include "opdev/data_type_utils.h"
#include "opdev/format_utils.h"
#include "opdev/op_dfx.h"
#include "opdev/make_op_executor.h"
#include "opdev/op_executor.h"
#include "opdev/op_log.h"
#include "opdev/shape_utils.h"
#include "opdev/tensor_view_utils.h"
#include "aclnn_kernels/common/op_error_check.h"

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

static constexpr size_t CHW_DIM_NUM = 3;
static constexpr size_t NCHW_DIM_NUM = 4;
static constexpr size_t DIM_ZERO = 0;
static constexpr size_t DIM_ONE = 1;
static constexpr size_t DIM_TWO = 2;
static constexpr size_t DIM_THREE = 3;
static constexpr int64_t AXIS = 2;
static constexpr int64_t EXPECT_SIZE = 2;

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

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

static inline bool CheckNotNull(const aclTensor* self, const aclTensor* indices, const aclIntArray* outputSize,
                                const aclTensor* outRef)
{
    OP_CHECK_NULL(self, return false);
    OP_CHECK_NULL(indices, return false);
    OP_CHECK_NULL(outputSize, return false);
    OP_CHECK_NULL(outRef, return false);
    return true;
}

static inline bool CheckDtypeValid(const aclTensor* self, const aclTensor* indices, const aclTensor* outRef)
{
    OP_CHECK_DTYPE_NOT_SUPPORT(self, DTYPE_SUPPORT_LIST, return false);
    OP_CHECK_DTYPE_NOT_MATCH(outRef, self->GetDataType(), return false);
    OP_CHECK_DTYPE_NOT_SUPPORT(indices, INDEX_DTYPE_SUPPORT_LIST, return false);
    return true;
}

static bool CheckShape(const aclTensor* self, const aclTensor* indices)
{
    size_t selfDimNum = self->GetViewShape().GetDimNum();
    OP_CHECK(selfDimNum == CHW_DIM_NUM || selfDimNum == NCHW_DIM_NUM,
             OP_LOGE(ACLNN_ERR_PARAM_INVALID, "Expected self to be a 3d or 4d Tensor, instead got: %zu", selfDimNum),
             return false);

    OP_CHECK(selfDimNum == indices->GetViewShape().GetDimNum(),
             OP_LOGE(ACLNN_ERR_PARAM_INVALID,
                     "Expected dim of self and indices to be equal, instead got self: %zu, gradOutput: %zu", selfDimNum,
                     indices->GetViewShape().GetDimNum()),
             return false);

    OP_CHECK_SHAPE_NOT_EQUAL(self, indices, return false);
    return true;
}

static bool CheckInpuNullTensor(const aclTensor* self)
{
    auto inputShape = self->GetViewShape();
    size_t dimNum = inputShape.GetDimNum();

    for (size_t i = 1; i < dimNum; ++i) {
        if (inputShape.GetDim(i) <= 0) {
            OP_LOGE(ACLNN_ERR_PARAM_INVALID,
                    "max_unpool2d():expected input to have non-empty spatiak dimensions, "
                    "but input has sizes %zu with dimension %zu being empty.",
                    dimNum, i);
            return false;
        }
    }
    return true;
}

static bool CheckOutputSize(const aclTensor* self, const aclIntArray* outputSize)
{
    uint64_t size = outputSize->Size();
    if (size != EXPECT_SIZE) {
        OP_LOGE(ACLNN_ERR_PARAM_INVALID, "outputSize length should be 2, but now is %lu.", size);
        return false;
    }
    for (size_t i = 0; i < size; ++i) {
        if ((*outputSize)[i] <= 0) {
            OP_LOGE(ACLNN_ERR_PARAM_INVALID, "outputSize value should greater than 0, but the sizes of %zu is %ld.", i,
                    (*outputSize)[i]);
            return false;
        }
    }
    int64_t dimH = self->GetViewShape().GetDim(DIM_ONE);
    int64_t dimW = self->GetViewShape().GetDim(DIM_TWO);
    if (self->GetViewShape().GetDimNum() == NCHW_DIM_NUM) {
        dimH = self->GetViewShape().GetDim(DIM_TWO);
        dimW = self->GetViewShape().GetDim(DIM_THREE);
    }
    OP_CHECK(((*outputSize)[0] * (*outputSize)[1]) >= (dimH * dimW),
             OP_LOGE(ACLNN_ERR_PARAM_INVALID,
                     "The output volums are of size %ld x %ld, should greater than or equal to "
                     "self of size %ld x %ld.",
                     (*outputSize)[0], (*outputSize)[1], dimH, dimW),
             return false);
    return true;
}

static inline aclnnStatus CheckParams(const aclTensor* self, const aclTensor* indices, const aclIntArray* outputSize,
                                      const aclTensor* outRef)
{
    // 1. 检查参数是否为空指针
    CHECK_RET(CheckNotNull(self, indices, outputSize, outRef), 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, indices, outRef), ACLNN_ERR_PARAM_INVALID);

    // 3. 检查输出shape
    CHECK_RET(CheckShape(self, indices), ACLNN_ERR_PARAM_INVALID);

    // 4. 校验输入tensor是否为空
    CHECK_RET(CheckInpuNullTensor(self), ACLNN_ERR_PARAM_INVALID);

    // 5. 校验output_size是否合法
    CHECK_RET(CheckOutputSize(self, outputSize), ACLNN_ERR_PARAM_INVALID);
    return ACLNN_SUCCESS;
}

const aclIntArray* CalcMaxUnpool2dInputNewShape(int64_t dimN, int64_t dimC, int64_t dimH, int64_t dimW,
                                                aclOpExecutor* executor)
{
    FVector<int64_t> newShape;
    newShape.emplace_back(dimN);
    newShape.emplace_back(dimC);
    newShape.emplace_back(dimH * dimW);
    return executor->AllocIntArray(newShape.data(), newShape.size());
}

aclnnStatus aclnnMaxUnpool2dGetWorkspaceSize(const aclTensor* self, const aclTensor* indices,
                                             const aclIntArray* outputSize, aclTensor* outRef, uint64_t* workspaceSize,
                                             aclOpExecutor** executor)
{
    OP_CHECK_COMM_INPUT(workspaceSize, executor);

    L2_DFX_PHASE_1(aclnnMaxUnpool2d, DFX_IN(self, indices, outputSize), DFX_OUT(outRef));

    // 固定写法，创建OpExecutor
    auto uniqueExecutor = CREATE_EXECUTOR();
    OP_CHECK(uniqueExecutor.get() != nullptr, OP_LOGE(ACLNN_ERR_INNER_CREATE_EXECUTOR, "Create executor error."),
             return ACLNN_ERR_INNER_CREATE_EXECUTOR);

    // 固定写法，参数检查
    auto ret = CheckParams(self, indices, outputSize, outRef);
    CHECK_RET(ret == ACLNN_SUCCESS, ret);

    if (self->IsEmpty()) {
        *workspaceSize = 0;
        uniqueExecutor.ReleaseTo(executor);
        return ACLNN_SUCCESS;
    }

    size_t selfDimNum = self->GetViewShape().GetDimNum();
    int64_t outH = (*outputSize)[0];
    int64_t outW = (*outputSize)[1];
    int64_t dimN = 1;
    int64_t dimC = self->GetViewShape().GetDim(DIM_ZERO);
    int64_t dimH = self->GetViewShape().GetDim(DIM_ONE);
    int64_t dimW = self->GetViewShape().GetDim(DIM_TWO);

    if (selfDimNum == NCHW_DIM_NUM) {
        dimN = self->GetViewShape().GetDim(DIM_ZERO);
        dimC = self->GetViewShape().GetDim(DIM_ONE);
        dimH = self->GetViewShape().GetDim(DIM_TWO);
        dimW = self->GetViewShape().GetDim(DIM_THREE);
    }

    auto selfContiguous = l0op::Contiguous(self, uniqueExecutor.get());
    CHECK_RET(selfContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);

    auto indicesContiguous = l0op::Contiguous(indices, uniqueExecutor.get());
    CHECK_RET(indicesContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);

    auto outRefContiguous = l0op::Contiguous(outRef, uniqueExecutor.get());
    CHECK_RET(outRefContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);

    const aclIntArray* inputNewShapeArray = CalcMaxUnpool2dInputNewShape(dimN, dimC, dimH, dimW, uniqueExecutor.get());
    CHECK_RET(inputNewShapeArray != nullptr, ACLNN_ERR_INNER_NULLPTR);

    const aclIntArray* outNewShapeArray = CalcMaxUnpool2dInputNewShape(dimN, dimC, outH, outW, uniqueExecutor.get());
    CHECK_RET(outNewShapeArray != nullptr, ACLNN_ERR_INNER_NULLPTR);

    auto selfReshape = l0op::Reshape(selfContiguous, inputNewShapeArray, uniqueExecutor.get());
    CHECK_RET(selfReshape != nullptr, ACLNN_ERR_INNER_NULLPTR);

    auto indicesReshape = l0op::Reshape(indicesContiguous, inputNewShapeArray, uniqueExecutor.get());
    CHECK_RET(indicesReshape != nullptr, ACLNN_ERR_INNER_NULLPTR);

    auto outRefReshape = l0op::Reshape(outRefContiguous, outNewShapeArray, uniqueExecutor.get());
    CHECK_RET(outRefReshape != nullptr, ACLNN_ERR_INNER_NULLPTR);

    auto zeroOut = l0op::ZerosLike(outRefReshape, uniqueExecutor.get());
    CHECK_RET(zeroOut != nullptr, ACLNN_ERR_INNER_NULLPTR);

    static const std::string reductionCurr = "none";
    auto scatterRes = l0op::ScatterElements(zeroOut, indicesReshape, selfReshape, AXIS, reductionCurr,
                                            uniqueExecutor.get());
    CHECK_RET(scatterRes != nullptr, ACLNN_ERR_INNER_NULLPTR);

    auto outReshape2 = l0op::Reshape(scatterRes, outRef->GetViewShape(), uniqueExecutor.get());
    CHECK_RET(outReshape2 != nullptr, ACLNN_ERR_INNER_NULLPTR);

    auto viewCopyResult = l0op::ViewCopy(outReshape2, outRef, uniqueExecutor.get());
    CHECK_RET(viewCopyResult != nullptr, ACLNN_ERR_INNER_NULLPTR);

    *workspaceSize = uniqueExecutor->GetWorkspaceSize();
    uniqueExecutor.ReleaseTo(executor);
    return ACLNN_SUCCESS;
}

aclnnStatus aclnnMaxUnpool2d(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, aclrtStream stream)
{
    L2_DFX_PHASE_2(aclnnMaxUnpool2d);
    return CommonOpExecutorRun(workspace, workspaceSize, executor, stream);
}

#ifdef __cplusplus
}
#endif