#include <stdint.h>

#include <vector>
#include <queue>
#include <mutex>

#include "UnityPluginAPI/IUnityInterface.h"
#include "UnityPluginAPI/IUnityLog.h"
#include "UnityPluginAPI/IUnityGraphics.h"
#include "UnityPluginAPI/IUnityRenderingExtensions.h"

#include <d3d12.h>
#include <dxgi.h>
#include "UnityPluginAPI/IUnityGraphicsD3D12.h"

#include "ffx-fsr2-api/ffx_fsr2.h"
#include "ffx-fsr2-api/dx12/ffx_fsr2_dx12.h"

#include "FSR3UnityTypes.h"

static const int32_t BaseEventId = 0;

static IUnityInterfaces* s_UnityInterfaces = nullptr;
static IUnityLog* s_Log = nullptr;
static IUnityGraphics* s_Graphics = nullptr;
static IUnityGraphicsD3D12v7* s_GraphicsD3D12 = nullptr;
static UnityGfxRenderer s_RendererType = kUnityGfxRendererNull;

static FfxDevice s_Device = nullptr;
static FfxFsr2Interface s_Fsr2Interface;

static void UNITY_INTERFACE_API OnGraphicsDeviceEvent(UnityGfxDeviceEventType eventType);
static void UNITY_INTERFACE_API OnRenderEventAndData(int eventID, void* data);
static void UNITY_INTERFACE_API OnSetTextureEvent(int eventID, void* data);

struct FSR3Feature
{
    FfxFsr2Context upscalingContext;
    uint32_t flags;
    bool isValid;

    FSR3TextureTable textureTable;
};

static std::vector<FSR3Feature> s_Features;
static std::queue<uint32_t> s_FeatureSlots;
static std::mutex s_FeatureMutex;

// Unity plugin load event
extern "C" void UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API UnityPluginLoad(IUnityInterfaces* unityInterfaces)
{
    s_UnityInterfaces = unityInterfaces;
    s_Log = unityInterfaces->Get<IUnityLog>();
    s_Graphics = unityInterfaces->Get<IUnityGraphics>();

    s_Graphics->RegisterDeviceEventCallback(OnGraphicsDeviceEvent);

    // Run OnGraphicsDeviceEvent(initialize) manually on plugin load
    // to not miss the event in case the graphics device is already initialized
    OnGraphicsDeviceEvent(kUnityGfxDeviceEventInitialize);
}

// Unity plugin unload event
extern "C" void UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API UnityPluginUnload()
{
    s_Graphics->UnregisterDeviceEventCallback(OnGraphicsDeviceEvent);
}

static void UNITY_INTERFACE_API OnGraphicsDeviceEvent(UnityGfxDeviceEventType eventType)
{
    switch (eventType)
    {
    case kUnityGfxDeviceEventInitialize:
    {
        s_RendererType = s_Graphics->GetRenderer();
        if (s_RendererType != kUnityGfxRendererD3D12)
            return;

        s_GraphicsD3D12 = s_UnityInterfaces->Get<IUnityGraphicsD3D12v7>();
        if (s_GraphicsD3D12 == nullptr)
        {
            UNITY_LOG_ERROR(s_Log, "Could not obtain D3D12 Graphics interface!");
            return;
        }

        break;
    }
    case kUnityGfxDeviceEventShutdown:
    {
        s_GraphicsD3D12 = nullptr;
        s_RendererType = kUnityGfxRendererNull;
        break;
    }
    case kUnityGfxDeviceEventBeforeReset:
    {
        break;
    }
    case kUnityGfxDeviceEventAfterReset:
    {
        break;
    }
    };
}

// Thread-safe allocation of feature slot
static uint32_t AllocateFeatureSlot()
{
    std::lock_guard<std::mutex> lock(s_FeatureMutex);

    if (s_FeatureSlots.empty())
    {
        // Create a new feature if there are no free slots
        uint32_t featureSlot = (uint32_t)s_Features.size();
        s_Features.push_back(std::move(FSR3Feature()));
        return featureSlot;
    }

    // Reallocate an existing free slot
    uint32_t featureSlot = s_FeatureSlots.front();
    s_FeatureSlots.pop();
    return featureSlot;
}

// Thread-safe freeing and clearing of feature slot
static void FreeFeatureSlot(uint32_t featureSlot)
{
    std::lock_guard<std::mutex> lock(s_FeatureMutex);

    s_FeatureSlots.push(featureSlot);
    memset(&s_Features[featureSlot], 0, sizeof(FSR3Feature));
}

extern "C" bool UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API AMDUP_InitApi()
{
    if (s_GraphicsD3D12 == nullptr)
        return false;

    ID3D12Device* device = s_GraphicsD3D12->GetDevice();
    if (device == nullptr)
        return false;

    s_Device = ffxGetDeviceDX12(device);

    size_t scratchBufferSize = ffxFsr2GetScratchMemorySizeDX12();
    void* scratchBuffer = malloc(scratchBufferSize);
    ffxFsr2GetInterfaceDX12(&s_Fsr2Interface, device, scratchBuffer, scratchBufferSize);

    return true;
}

extern "C" void UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API AMDUP_ShutdownApi()
{
    for (uint32_t slot = 0; slot < s_Features.size(); ++slot)
    {
        auto& feature = s_Features[slot];
        if (feature.isValid)
        {
            ffxFsr2ContextDestroy(&feature.upscalingContext);
            FreeFeatureSlot(slot);
        }
    }

    if (s_Fsr2Interface.scratchBuffer != nullptr)
    {
        free(s_Fsr2Interface.scratchBuffer);
        s_Fsr2Interface.scratchBuffer = nullptr;
    }
    
    s_Device = nullptr;
}

extern "C" uint32_t UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API AMDUP_GetDeviceVersion()
{
    return 0x0u;
}

extern "C" UnityRenderingEventAndData UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API AMDUP_GetRenderEventCallback()
{
    return OnRenderEventAndData;
}

extern "C" UnityRenderingEventAndData UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API AMDUP_GetSetTextureEventCallback()
{
    return OnSetTextureEvent;
}

extern "C" uint32_t UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API AMDUP_CreateFeatureSlot()
{
    return AllocateFeatureSlot();
}

extern "C" float UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API AMDUP_GetUpscaleRatioFromQualityMode(FSR3Quality qualityMode)
{
    switch (qualityMode)
    {
    case FSR3Quality::qQuality:
        return 1.5f;
    case FSR3Quality::qBalanced:
        return 1.7f;
    case FSR3Quality::qPerformance:
        return 2.0f;
    case FSR3Quality::qUltraPerformance:
        return 3.0f;
    default:
        return 1.0f;
    }
}

extern "C" bool UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API AMDUP_GetRenderResolutionFromQualityMode(FSR3Quality qualityMode, uint32_t displayWidth, uint32_t displayHeight, uint32_t* renderWidth, uint32_t* renderHeight)
{
    if (renderWidth == nullptr || renderHeight == nullptr)
        return false;

    float ratio = AMDUP_GetUpscaleRatioFromQualityMode(qualityMode);
    *renderWidth = (uint32_t)roundf(displayWidth / ratio);
    *renderHeight = (uint32_t)roundf(displayHeight / ratio);
    return true;
}

extern "C" int32_t UNITY_INTERFACE_EXPORT UNITY_INTERFACE_API AMDUP_GetBaseEventId()
{
    return BaseEventId;
}

// Plugin function to handle a specific rendering event
static void UNITY_INTERFACE_API OnRenderEventAndData(int eventID, void* data)
{
    if (s_GraphicsD3D12 == nullptr || s_Device == nullptr)
        return;

    // User rendering code
    switch (eventID)
    {
    case BaseEventId + FSR3PluginEvent::eDestroyFeature:
    {
        uint32_t featureSlot = (uint32_t)(int64_t)data;
        if (featureSlot < 0 || featureSlot >= s_Features.size())
            return;

        auto& feature = s_Features[featureSlot];
        if (feature.isValid)
        {
            ffxFsr2ContextDestroy(&feature.upscalingContext);
            FreeFeatureSlot(featureSlot);
        }
        break;
    }
    case BaseEventId + FSR3PluginEvent::eExecute:
    {
        auto* params = (FSR3CommandExecutionData*)data;
        if (params->featureSlot < 0 || params->featureSlot >= s_Features.size())
            return;

        auto& feature = s_Features[params->featureSlot];

        FfxFsr2DispatchDescription dispatchDescription{};
        
        UnityGraphicsD3D12RecordingState state;
        s_GraphicsD3D12->CommandRecordingState(&state);
        dispatchDescription.commandList = ffxGetCommandListDX12(state.commandList);

        dispatchDescription.color = ffxGetResourceDX12(&feature.upscalingContext, (ID3D12Resource*)feature.textureTable.colorInput);
        dispatchDescription.depth = ffxGetResourceDX12(&feature.upscalingContext, (ID3D12Resource*)feature.textureTable.depth);
        dispatchDescription.motionVectors = ffxGetResourceDX12(&feature.upscalingContext, (ID3D12Resource*)feature.textureTable.motionVectors);
        dispatchDescription.exposure = ffxGetResourceDX12(&feature.upscalingContext, (ID3D12Resource*)feature.textureTable.exposureTexture);
        dispatchDescription.reactive = ffxGetResourceDX12(&feature.upscalingContext, (ID3D12Resource*)feature.textureTable.reactiveMask);
        dispatchDescription.transparencyAndComposition = ffxGetResourceDX12(&feature.upscalingContext, (ID3D12Resource*)feature.textureTable.transparencyMask);
        dispatchDescription.output = ffxGetResourceDX12(&feature.upscalingContext, (ID3D12Resource*)feature.textureTable.colorOutput, nullptr, FFX_RESOURCE_STATE_UNORDERED_ACCESS);

        dispatchDescription.jitterOffset.x = params->jitterOffsetX;
        dispatchDescription.jitterOffset.y = params->jitterOffsetY;
        dispatchDescription.motionVectorScale.x = params->MVScaleX;
        dispatchDescription.motionVectorScale.y = params->MVScaleY;
        dispatchDescription.reset = params->reset;
        dispatchDescription.enableSharpening = params->enableSharpening;
        dispatchDescription.sharpness = params->sharpness;
        dispatchDescription.frameTimeDelta = params->frameTimeDelta;
        dispatchDescription.preExposure = params->preExposure;
        dispatchDescription.renderSize.width = params->renderSizeWidth;
        dispatchDescription.renderSize.height = params->renderSizeHeight;
        dispatchDescription.cameraFovAngleVertical = params->cameraFovAngleVertical;

        if (feature.flags & FFX_FSR2_ENABLE_DEPTH_INVERTED)
        {
            dispatchDescription.cameraFar = params->cameraNear;
            dispatchDescription.cameraNear = params->cameraFar;
        }
        else
        {
            dispatchDescription.cameraFar = params->cameraFar;
            dispatchDescription.cameraNear = params->cameraNear;
        }

        ffxFsr2ContextDispatch(&feature.upscalingContext, &dispatchDescription);
        break;
    }
    case BaseEventId + FSR3PluginEvent::ePostExecute:
    {
        auto* params = (FSR3CommandExecutionData*)data;
        break;
    }
    case BaseEventId + FSR3PluginEvent::eInit:
    {
        auto* params = (FSR3CommandInitializationData*)data;
        if (params->featureSlot < 0 || params->featureSlot >= s_Features.size())
            return;

        auto& feature = s_Features[params->featureSlot];
        feature.flags = params->flags;

        FfxFsr2ContextDescription contextDescription{};
        contextDescription.callbacks = s_Fsr2Interface;
        contextDescription.device = s_Device;
        contextDescription.displaySize = { params->displaySizeWidth, params->displaySizeHeight };
        contextDescription.maxRenderSize = { params->maxRenderSizeWidth, params->maxRenderSizeHeight };
        contextDescription.flags = params->flags;

        feature.isValid = ffxFsr2ContextCreate(&feature.upscalingContext, &contextDescription) == FFX_OK;
        break;
    }
    }
}

static void UNITY_INTERFACE_API OnSetTextureEvent(int eventID, void* data)
{
    if (s_GraphicsD3D12 == nullptr)
        return;

    auto* params = (UnityRenderingExtTextureUpdateParamsV2*)data;

    // userData = (featureId & (int) ushort.MaxValue) << 16 | (textureSlot & (int) short.MaxValue) << 1 | (clearTextureTable ? 1 : 0);
    uint32_t featureSlot = (params->userData >> 16) & 0xFFFF;
    uint32_t textureSlot = (params->userData >> 1) & 0x7FFF;
    uint32_t clearTextureTable = params->userData & 0x1;

    if (featureSlot < 0 || featureSlot >= s_Features.size())
        return;

    auto& feature = s_Features[featureSlot];

    // User rendering code
    switch (eventID)
    {
    case kUnityRenderingExtEventUpdateTextureBeginV2:
    {
        if (clearTextureTable)
        {
            memset(&feature.textureTable, 0, sizeof(FSR3TextureTable));
        }
        break;
    }
    case kUnityRenderingExtEventUpdateTextureEndV2:
    {
        // We organized the texture table struct to be ordered the same as the texture slot enum
        // This way we can use the texture slot value simply as a pointer offset into the texture table
        *(((intptr_t*)&feature.textureTable) + textureSlot) = (intptr_t)s_GraphicsD3D12->TextureFromNativeTexture((UnityTextureID)params->textureID);
        break;
    }
    }
}