/**
 * 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 "mul.h"
#include "aclnn_kernels/common/op_error_check.h"
#include "opdev/aicpu/aicpu_task.h"
#include "opdev/make_op_executor.h"
#include "opdev/op_dfx.h"
#include "opdev/op_def.h"
#include "opdev/op_log.h"
#include "opdev/op_executor.h"
#include "opdev/shape_utils.h"
#include "opdev/platform.h"
#include "op_api/aclnn_check.h"

using namespace op;

namespace l0op {

OP_TYPE_REGISTER(Mul);

static const std::initializer_list<DataType> ASCEND910_AICORE_DTYPE_SUPPORT_LIST = {
    DataType::DT_FLOAT, DataType::DT_FLOAT16, DataType::DT_INT32, DataType::DT_INT8,
    DataType::DT_UINT8, DataType::DT_INT64,   DataType::DT_BOOL};

static const std::initializer_list<DataType> ASCEND910B_AICORE_DTYPE_SUPPORT_LIST = {
    DataType::DT_FLOAT, DataType::DT_FLOAT16, DataType::DT_INT32,     DataType::DT_INT8, DataType::DT_UINT8,
    DataType::DT_INT64, DataType::DT_BF16,    DataType::DT_COMPLEX64, DataType::DT_BOOL};

static const std::initializer_list<DataType> REGBASE_AICORE_DTYPE_SUPPORT_LIST = {
    DataType::DT_FLOAT,     DataType::DT_FLOAT16,   DataType::DT_BF16,  DataType::DT_INT32,
    DataType::DT_INT8,      DataType::DT_UINT8,     DataType::DT_INT64, ge::DT_INT16,
    DataType::DT_COMPLEX32, DataType::DT_COMPLEX64, DataType::DT_BOOL};

static const std::initializer_list<DataType> ASCEND610LITE_AICORE_DTYPE_SUPPORT_LIST = {
    DataType::DT_FLOAT, DataType::DT_FLOAT16, DataType::DT_INT32, DataType::DT_INT8, DataType::DT_UINT8};

static constexpr int64_t DIM_FOUR = 4;

static inline const std::initializer_list<DataType>& GetAiCoreDtypeSupportListBySocVersion()
{
    auto curArch = GetCurrentPlatformInfo().GetCurNpuArch();
    if (IsRegBase(curArch)) {
        return REGBASE_AICORE_DTYPE_SUPPORT_LIST;
    }
    switch (curArch) {
        case NpuArch::DAV_2201: {
            return ASCEND910B_AICORE_DTYPE_SUPPORT_LIST;
        }
        case NpuArch::DAV_1001: {
            return ASCEND910_AICORE_DTYPE_SUPPORT_LIST;
        }
        case NpuArch::DAV_3102: {
            return ASCEND610LITE_AICORE_DTYPE_SUPPORT_LIST;
        }
        default: {
            return ASCEND910_AICORE_DTYPE_SUPPORT_LIST;
        }
    }
}

bool IsDoubleSupport(const aclTensor* self, const aclTensor* other)
{
    if (IsRegBase() && self->GetDataType() == DataType::DT_DOUBLE && other->GetDataType() == DataType::DT_DOUBLE) {
        return true;
    }
    return false;
}

static bool isBroadcastTemplateNonContiguousSupport(const aclTensor* input)
{
    auto viewShape = input->GetViewShape();
    size_t shapeDim = viewShape.GetDimNum();

    // dim > 4 不支持
    if (shapeDim > DIM_FOUR) {
        OP_LOGI("Broadcast Template NonContiguous Unsupported. shapeDim: %d > 4", shapeDim);
        return false;
    }

    // 仅regbase类的芯片支持
    if (!IsRegBase()) {
        OP_LOGI("Broadcast Template NonContiguous Unsupported. not RegBase");
        return false;
    }

    return true;
}

// 根据芯片类型、dtype判断算子是否支持走AiCore
inline static bool IsAiCoreSupport(const aclTensor* self)
{
    return CheckType(self->GetDataType(), GetAiCoreDtypeSupportListBySocVersion());
}

bool IsMulSupportNonContiguous(const aclTensor* self, const aclTensor* other)
{
    bool isSupportNonContiguous = isBroadcastTemplateNonContiguousSupport(self) &&
                                  isBroadcastTemplateNonContiguousSupport(other);
    return isSupportNonContiguous &&
           ((IsAiCoreSupport(self) && IsAiCoreSupport(other)) || IsDoubleSupport(self, other));
}

// AICORE算子kernel
static const aclTensor* MulAiCore(const aclTensor* self, const aclTensor* other, const aclTensor* mulOut,
                                  aclOpExecutor* executor)
{
    L0_DFX(MulAiCore, self, other, mulOut);
    // 使用框架宏ADD_TO_LAUNCHER_LIST_AICORE，将AiCore Mul算子加入任务队列
    // Mul是算子的OpType，self、other是算子的输入，mulOut是算子的输出
    ADD_TO_LAUNCHER_LIST_AICORE(Mul, OP_INPUT(self, other), OP_OUTPUT(mulOut));
    return mulOut;
}

// AICPU算子kernel
static const aclTensor* MulAiCpu(const aclTensor* self, const aclTensor* other, const aclTensor* mulOut,
                                 aclOpExecutor* executor)
{
    L0_DFX(MulAiCpu, self, other);
    // 使用框架宏ADD_TO_LAUNCHER_LIST_AICPU，将AiCpu Mul算子加入任务队列
    // Mul是算子的OpType，self、other是算子的输入，mulOut是算子的输出
    static internal::AicpuTaskSpace space("Mul");
    auto ret = ADD_TO_LAUNCHER_LIST_AICPU(Mul, OP_ATTR_NAMES(), OP_INPUT(self, other), OP_OUTPUT(mulOut));
    CHECK_RET(ret == ACLNN_SUCCESS, nullptr);
    return mulOut;
}

const aclTensor* Mul(const aclTensor* self, const aclTensor* other, aclOpExecutor* executor)
{
    Shape broadcastShape;
    OP_CHECK_BROADCAST_AND_INFER_SHAPE(self, other, broadcastShape, return nullptr);

    bool isMixDataType = (self->GetDataType() == DataType::DT_FLOAT16 && other->GetDataType() == DataType::DT_FLOAT) ||
                         (self->GetDataType() == DataType::DT_FLOAT && other->GetDataType() == DataType::DT_FLOAT16) ||
                         (self->GetDataType() == DataType::DT_BF16 && other->GetDataType() == DataType::DT_FLOAT) ||
                         (self->GetDataType() == DataType::DT_FLOAT && other->GetDataType() == DataType::DT_BF16);

    auto mulOut = isMixDataType ? executor->AllocTensor(broadcastShape, DataType::DT_FLOAT) :
                                  executor->AllocTensor(broadcastShape, self->GetDataType());
    if (isMixDataType || (IsAiCoreSupport(self) && IsAiCoreSupport(other)) || IsDoubleSupport(self, other)) {
        return MulAiCore(self, other, mulOut, executor);
    }

    return MulAiCpu(self, other, mulOut, executor);
}

const aclTensor* MulInplace(const aclTensor* self, const aclTensor* rfRes, aclOpExecutor* executor)
{
    Shape broadcastShape;
    OP_CHECK_BROADCAST_AND_INFER_SHAPE(self, rfRes, broadcastShape, return nullptr);

    // 校验输出tensor的shape和rfRes tensor一致
    if (broadcastShape != rfRes->GetViewShape()) {
        OP_LOGE(ACLNN_ERR_PARAM_INVALID,
                "self and rfRes broadcastShape [%s] not equal to rfRes shape [%s], do not support inplace from the "
                "'rfRes' tensor!",
                op::ToString(broadcastShape).GetString(), op::ToString(rfRes->GetViewShape()).GetString());
        return nullptr;
    }

    // 校验输出tensor的dtype和rfRes tensor一致
    bool isMixDataType = (self->GetDataType() == DataType::DT_FLOAT16 && rfRes->GetDataType() == DataType::DT_FLOAT) ||
                         (self->GetDataType() == DataType::DT_FLOAT && rfRes->GetDataType() == DataType::DT_FLOAT16) ||
                         (self->GetDataType() == DataType::DT_BF16 && rfRes->GetDataType() == DataType::DT_FLOAT) ||
                         (self->GetDataType() == DataType::DT_FLOAT && rfRes->GetDataType() == DataType::DT_BF16);
    if (isMixDataType && (rfRes->GetDataType() == DataType::DT_FLOAT16 || rfRes->GetDataType() == DataType::DT_BF16)) {
        OP_LOGE(ACLNN_ERR_PARAM_INVALID,
                "out dtype DataType::DT_FLOAT not equal to rfRes dtype [%s], do not support inplace from the 'rfRes' "
                "tensor!",
                op::ToString(rfRes->GetDataType()).GetString());
        return nullptr;
    }

    if (isMixDataType || (IsAiCoreSupport(self) && IsAiCoreSupport(rfRes)) || IsDoubleSupport(self, rfRes)) {
        return MulAiCore(self, rfRes, rfRes, executor);
    }

    return MulAiCpu(self, rfRes, rfRes, executor);
}

} // namespace l0op