/*
 * SPDX-FileCopyrightText: Copyright (c) 1993-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 * SPDX-License-Identifier: Apache-2.0
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#ifndef TRT_DISENTANGLED_ATTENTION_PLUGIN_H
#define TRT_DISENTANGLED_ATTENTION_PLUGIN_H

#include "NvInferPlugin.h"
#include "common/plugin.h"
#include "common/serialize.hpp"
#include "disentangledAttentionCommon.h"
#include <cstdint>
#include <iostream>
#include <string>
#include <vector>

// One of the preferred ways of making TensorRT to be able to see
// our custom layer requires extending IPluginV2 and IPluginCreator classes.
// For requirements for overriden functions, check TensorRT API docs.
namespace nvinfer1
{
namespace plugin
{

// using namespace nvinfer1;

class DisentangledAttentionPlugin : public nvinfer1::IPluginV3,
                                    public nvinfer1::IPluginV3OneCore,
                                    public nvinfer1::IPluginV3OneBuild,
                                    public nvinfer1::IPluginV3OneRuntime
{
public:
    DisentangledAttentionPlugin();

    DisentangledAttentionPlugin(int32_t span, float factor);

    DisentangledAttentionPlugin(void const* serialData, size_t serialLength);

    // Destructor
    virtual ~DisentangledAttentionPlugin() {}

    // IPluginV3OneCore methods
    int32_t getNbOutputs() const noexcept override;
    void setPluginNamespace(char const* pluginNamespace) noexcept;
    char const* getPluginNamespace() const noexcept override;
    char const* getPluginName() const noexcept override;
    char const* getPluginVersion() const noexcept override;
    nvinfer1::IPluginV3* clone() noexcept override;
    nvinfer1::PluginFieldCollection const* getFieldsToSerialize() noexcept override;
    nvinfer1::IPluginCapability* getCapabilityInterface(nvinfer1::PluginCapabilityType type) noexcept override;

    // IPluginV3OneBuild methods
    int32_t getOutputShapes(nvinfer1::DimsExprs const* inputs, int32_t nbInputs, nvinfer1::DimsExprs const* shapeInputs,
        int32_t nbShapeInputs, nvinfer1::DimsExprs* outputs, int32_t nbOutputs,
        nvinfer1::IExprBuilder& exprBuilder) noexcept override;
    int32_t configurePlugin(nvinfer1::DynamicPluginTensorDesc const* in, int32_t nbInputs,
        nvinfer1::DynamicPluginTensorDesc const* out, int32_t nbOutputs) noexcept override;
    int32_t getOutputDataTypes(nvinfer1::DataType* outputTypes, int32_t nbOutputs, nvinfer1::DataType const* inputTypes,
        int32_t nbInputs) const noexcept override;
    bool supportsFormatCombination(int32_t pos, nvinfer1::DynamicPluginTensorDesc const* inOut, int32_t nbInputs,
        int32_t nbOutputs) noexcept override;

    // IPluginV3OneRuntime methods
    int32_t enqueue(nvinfer1::PluginTensorDesc const* inputDesc, nvinfer1::PluginTensorDesc const* outputDesc,
        void const* const* inputs, void* const* outputs, void* workspace, cudaStream_t stream) noexcept override;
    size_t getWorkspaceSize(nvinfer1::DynamicPluginTensorDesc const* inputs, int32_t nbInputs,
        nvinfer1::DynamicPluginTensorDesc const* outputs, int32_t nbOutputs) const noexcept override;
    int32_t onShapeChange(nvinfer1::PluginTensorDesc const* inputs, int32_t nbInputs,
        nvinfer1::PluginTensorDesc const* outputs, int32_t nbOutputs) noexcept override;
    nvinfer1::IPluginV3* attachToContext(nvinfer1::IPluginResourceContext* context) noexcept override;

private:
    // Helper method for enqueue()
    template <typename TDataType>
    void enqueueType(nvinfer1::PluginTensorDesc const* inputDesc, nvinfer1::PluginTensorDesc const* outputDesc,
        void const* const* inputs, void* const* outputs, cudaStream_t stream, TDataType factor);

    std::string mNamespace;

    // attributes
    int32_t mSpan;
    float mFactor;

    // Field serialization storage
    std::vector<nvinfer1::PluginField> mDataToSerialize;
    nvinfer1::PluginFieldCollection mFCToSerialize;
};

class DisentangledAttentionPluginCreator : public nvinfer1::IPluginCreatorV3One
{
public:
    DisentangledAttentionPluginCreator();

    ~DisentangledAttentionPluginCreator() override = default;

    char const* getPluginName() const noexcept override;

    char const* getPluginVersion() const noexcept override;

    nvinfer1::PluginFieldCollection const* getFieldNames() noexcept override;

    nvinfer1::IPluginV3* createPlugin(
        char const* name, nvinfer1::PluginFieldCollection const* fc, nvinfer1::TensorRTPhase phase) noexcept override;

    void setPluginNamespace(char const* pluginNamespace) noexcept;

    char const* getPluginNamespace() const noexcept override;

private:
    nvinfer1::PluginFieldCollection mFC;
    std::vector<nvinfer1::PluginField> mPluginAttributes;
    std::string mNamespace;
};

} // namespace plugin
} // namespace nvinfer1

#endif // TRT_DISENTANGLED_ATTENTION_PLUGIN_H