#pragma once

#include "WorldMesher.h"
#include "StandaloneWorkerThread.h"
#include "VoxelForgeCore/VoxelForgeCore.h"

#include <condition_variable>
#include <algorithm>
#include <array>
#include <atomic>
#include <bit>
#include <chrono>
#include <cmath>
#include <cstdint>
#include <cstdio>
#include <cstring>
#include <deque>
#include <functional>
#include <fstream>
#include <limits>
#include <unordered_set>
#include <utility>

struct FStandaloneTileKeyHash
{
    std::size_t operator()(const FVoxelTileKey& Key) const noexcept
    { return static_cast<std::size_t>(GetTypeHash(Key)); }
};

struct FStandaloneDesiredTileSet
{
    std::vector<FVoxelTileKey> Tiles;
    FIntVector CenterChunk=FIntVector::ZeroValue;
    int32 MinimumChunkZ=MIN_int32;
    int32 MaximumChunkZ=MAX_int32;
    int32 StrateTopChunkZ=0;
    int32 StrateBottomChunkZ=0;
    bool bHasStrate=false;
};

struct FStandaloneCpuTile
{
    FVoxelTileKey Key;
    int32 WorkerId=-1;
    uint64 RequestedCrossing=0;
    std::vector<FStandaloneCpuVertex> Vertices;
    std::vector<uint32> Indices;
    std::vector<uint8> TriangleSkirtFaceMasks;
    uint32 SurfaceVertexCount=0;
    uint32 SkirtTriangleCount=0;
    uint64 DensitySamples=0;
    double WorkerSeconds=0.0;
    std::chrono::steady_clock::time_point QueuedAt{};
    std::chrono::steady_clock::time_point StartedAt{};
    std::chrono::steady_clock::time_point CompletedAt{};
    uint64 DensityHash=14695981039346656037ull;
    bool bHasDensityHash=false;
    uint64 MeshHash=14695981039346656037ull;
    bool bHasMeshHash=false;
    uint64 PayloadHash=14695981039346656037ull;
    bool bHasPayloadHash=false;
    uint64 TopologyHash=14695981039346656037ull;
    bool bHasTopologyHash=false;
    uint64 MaterialHash=14695981039346656037ull;
    bool bHasMaterialHash=false;
    bool bGenerationFailed=false;
    uint64 MeshPayloadAllocatedBytes=0;
    VoxelForgeCore::TileGenerationStats CoreDiagnostics{};
    std::array<uint64,VoxelDensityProfile::BucketCount> ProfileCycles{};

};

inline uint64 HashStandaloneMeshPayload(const FVoxelMeshData& Mesh)
{
    uint64 Hash=14695981039346656037ull;
    const auto AddWord=[&](uint64 Word,int32 Bytes)
    {
        for(int32 I=0;I<Bytes;++I)
        {
            Hash^=static_cast<uint8>(Word>>(I*8));
            Hash*=1099511628211ull;
        }
    };
    const auto AddDouble=[&](double Value)
    { AddWord(std::bit_cast<uint64>(Value),8); };
    AddWord(static_cast<uint64>(Mesh.Vertices.Num()),8);
    for(const FVector& Position:Mesh.Vertices)
    { AddDouble(Position.X);AddDouble(Position.Y);AddDouble(Position.Z); }
    AddWord(static_cast<uint64>(Mesh.Normals.Num()),8);
    for(const FVector& Normal:Mesh.Normals)
    { AddDouble(Normal.X);AddDouble(Normal.Y);AddDouble(Normal.Z); }
    AddWord(static_cast<uint64>(Mesh.UVs.Num()),8);
    for(const auto& UV:Mesh.UVs)
    { AddDouble(static_cast<double>(UV.X));AddDouble(static_cast<double>(UV.Y)); }
    AddWord(static_cast<uint64>(Mesh.Colors.Num()),8);
    for(const FColor& Color:Mesh.Colors)
    { AddWord(Color.R,1);AddWord(Color.G,1);AddWord(Color.B,1);AddWord(Color.A,1); }
    AddWord(static_cast<uint64>(Mesh.Triangles.Num()),8);
    for(const int32 Index:Mesh.Triangles)AddWord(static_cast<uint32>(Index),4);
    AddWord(static_cast<uint64>(Mesh.TriangleSkirtFaceMasks.Num()),8);
    for(const uint8 Faces:Mesh.TriangleSkirtFaceMasks)AddWord(Faces,1);
    AddWord(static_cast<uint32>(Mesh.NumSurfaceVertices),4);
    AddWord(static_cast<uint32>(Mesh.NumSkirtTriangles),4);
    AddWord(static_cast<uint32>(Mesh.NumCeilingTriangles),4);
    return Hash;
}

inline uint64 HashStandaloneMeshTopology(const FVoxelMeshData& Mesh)
{
    uint64 Hash=14695981039346656037ull;
    const auto AddWord=[&](uint64 Word,int32 Bytes)
    {
        for(int32 I=0;I<Bytes;++I)
        {
            Hash^=static_cast<uint8>(Word>>(I*8));
            Hash*=1099511628211ull;
        }
    };
    using FTriangleKey=std::array<uint64,3>;
    const auto VertexKey=[&](int32 Index)
    {
        uint64 VertexHash=14695981039346656037ull;
        const auto AddVertexWord=[&](uint64 Word)
        {
            for(int32 Byte=0;Byte<8;++Byte)
            {
                VertexHash^=static_cast<uint8>(Word>>(Byte*8));
                VertexHash*=1099511628211ull;
            }
        };
        if(Index<0||Index>=Mesh.Vertices.Num()||Index>=Mesh.Normals.Num())
        { AddVertexWord(UINT64_MAX);return VertexHash; }
        const FVector& Position=Mesh.Vertices[Index];
        const FVector& Normal=Mesh.Normals[Index];
        AddVertexWord(std::bit_cast<uint64>(static_cast<double>(Position.X)));
        AddVertexWord(std::bit_cast<uint64>(static_cast<double>(Position.Y)));
        AddVertexWord(std::bit_cast<uint64>(static_cast<double>(Position.Z)));
        AddVertexWord(std::bit_cast<uint64>(static_cast<double>(Normal.X)));
        AddVertexWord(std::bit_cast<uint64>(static_cast<double>(Normal.Y)));
        AddVertexWord(std::bit_cast<uint64>(static_cast<double>(Normal.Z)));
        return VertexHash;
    };
    std::vector<FTriangleKey> Triangles;
    Triangles.reserve(static_cast<size_t>(Mesh.Triangles.Num()/3));
    for(int32 I=0;I+2<Mesh.Triangles.Num();I+=3)
        Triangles.push_back({VertexKey(Mesh.Triangles[I]),
            VertexKey(Mesh.Triangles[I+1]),VertexKey(Mesh.Triangles[I+2])});
    std::sort(Triangles.begin(),Triangles.end());
    AddWord(static_cast<uint64>(Triangles.size()),8);
    for(const FTriangleKey& Triangle:Triangles)
        for(const uint64 VertexHash:Triangle)AddWord(VertexHash,8);
    return Hash;
}

inline uint64 HashStandaloneCpuTileMesh(const FStandaloneCpuTile& Tile)
{
    uint64 Hash=14695981039346656037ull;
    const auto AddWord=[&](uint64 Word,int32 Bytes)
    {
        for(int32 I=0;I<Bytes;++I)
        {
            Hash^=static_cast<uint8>(Word>>(I*8));
            Hash*=1099511628211ull;
        }
    };
    AddWord(Tile.Vertices.size(),8);
    for(const FStandaloneCpuVertex& Vertex:Tile.Vertices)
    {
        for(const float Value:Vertex.Position)AddWord(std::bit_cast<uint32>(Value),4);
        for(const float Value:Vertex.Normal)AddWord(std::bit_cast<uint32>(Value),4);
    }
    AddWord(Tile.Indices.size(),8);
    for(const uint32 Index:Tile.Indices)AddWord(Index,4);
    return Hash;
}

// Far parent repairs and shell stitching are implemented by VoxelForgeCore.

struct FStandaloneMeshCacheMetadata
{
    uint64 SourceDataHash=0;
    int32 Seed=0;
    int32 ClipRadius=3;
    int32 FullResClipLevels=2;
    int32 CoarseTileCells=16;
    int32 StrateContentCutMinLevel=9;
    int32 SkirtCells=0;
    int32 LODOctaveDrop=0;
    int32 bGenerateSkirts=0;
    uint32 AlgorithmVersion=1;
    uint32 TunnelMode=0;
    uint32 VoxelSizeCm=25;
};

class FStandaloneMeshCache
{
public:
    static constexpr uint32 CurrentVersion=1;

    void Clear(){Tiles.clear();bLoaded=false;BytesLoaded=0;}
    bool IsLoaded()const{return bLoaded;}
    size_t Num()const{return Tiles.size();}
    uint64 GetBytesLoaded()const{return BytesLoaded;}
    const FStandaloneCpuTile* Find(const FVoxelTileKey& Key)const
    {
        const auto It=Tiles.find(Key);
        return It==Tiles.end()?nullptr:&It->second;
    }

    static bool Save(const std::string& Path,const FStandaloneMeshCacheMetadata& Metadata,
                     std::vector<FStandaloneCpuTile>& InTiles,std::string& Error)
    {
        std::vector<FStandaloneCpuTile*> Ordered;
        Ordered.reserve(InTiles.size());
        for(FStandaloneCpuTile& Tile:InTiles)Ordered.push_back(&Tile);
        std::sort(Ordered.begin(),Ordered.end(),[](const FStandaloneCpuTile* A,const FStandaloneCpuTile* B)
        {
            if(A->Key.Level!=B->Key.Level)return A->Key.Level<B->Key.Level;
            if(A->Key.Coord.Z!=B->Key.Coord.Z)return A->Key.Coord.Z<B->Key.Coord.Z;
            if(A->Key.Coord.Y!=B->Key.Coord.Y)return A->Key.Coord.Y<B->Key.Coord.Y;
            return A->Key.Coord.X<B->Key.Coord.X;
        });
        std::ofstream File(Path,std::ios::binary|std::ios::trunc);
        if(!File){Error="could not open mesh cache for writing: "+Path;return false;}
        struct FHeader
        {
            char Magic[8];uint32 Version;uint32 Endian;uint64 SourceDataHash;
            int32 Seed,ClipRadius,FullResClipLevels,CoarseTileCells,StrateContentCutMinLevel;
            int32 SkirtCells,LODOctaveDrop,bGenerateSkirts;
            uint32 AlgorithmVersion,TunnelMode,VoxelSizeCm;
            uint64 TileCount;
        };
        static_assert(sizeof(FHeader)==80);
        const FHeader Header{{'V','F','M','E','S','H','1','\0'},1,0x01020304u,
            Metadata.SourceDataHash,Metadata.Seed,Metadata.ClipRadius,Metadata.FullResClipLevels,
            Metadata.CoarseTileCells,Metadata.StrateContentCutMinLevel,Metadata.SkirtCells,
            Metadata.LODOctaveDrop,Metadata.bGenerateSkirts,Metadata.AlgorithmVersion,
            Metadata.TunnelMode,Metadata.VoxelSizeCm,static_cast<uint64>(Ordered.size())};
        File.write(reinterpret_cast<const char*>(&Header),sizeof(Header));
        struct FRecord
        {
            int32 Level,X,Y,Z;uint64 Samples,DensityHash;uint32 VertexCount,IndexCount;
        };
        static_assert(sizeof(FRecord)==40);
        for(const FStandaloneCpuTile* Tile:Ordered)
        {
            if(Tile->Vertices.size()>UINT32_MAX||Tile->Indices.size()>UINT32_MAX
                ||Tile->Indices.size()%3!=0)
            {Error="mesh cache tile exceeds the 32-bit record limits";return false;}
            const FRecord Record{Tile->Key.Level,Tile->Key.Coord.X,Tile->Key.Coord.Y,
                Tile->Key.Coord.Z,Tile->DensitySamples,Tile->DensityHash,
                static_cast<uint32>(Tile->Vertices.size()),static_cast<uint32>(Tile->Indices.size())};
            File.write(reinterpret_cast<const char*>(&Record),sizeof(Record));
            if(!Tile->Vertices.empty())File.write(reinterpret_cast<const char*>(Tile->Vertices.data()),
                static_cast<std::streamsize>(Tile->Vertices.size()*sizeof(FStandaloneCpuVertex)));
            if(!Tile->Indices.empty())File.write(reinterpret_cast<const char*>(Tile->Indices.data()),
                static_cast<std::streamsize>(Tile->Indices.size()*sizeof(uint32)));
            if(!File){Error="mesh cache write failed: "+Path;return false;}
        }
        File.flush();
        if(!File){Error="mesh cache flush failed: "+Path;return false;}
        return true;
    }

    bool Load(const std::string& Path,const FStandaloneMeshCacheMetadata& Expected,std::string& Error)
    {
        Clear();
        std::ifstream File(Path,std::ios::binary);
        if(!File){Error="cache file is missing or unreadable";return false;}
        struct FHeader
        {
            char Magic[8];uint32 Version;uint32 Endian;uint64 SourceDataHash;
            int32 Seed,ClipRadius,FullResClipLevels,CoarseTileCells,StrateContentCutMinLevel;
            int32 SkirtCells,LODOctaveDrop,bGenerateSkirts;
            uint32 AlgorithmVersion,TunnelMode,VoxelSizeCm;
            uint64 TileCount;
        };
        static_assert(sizeof(FHeader)==80);
        FHeader H{};File.read(reinterpret_cast<char*>(&H),sizeof(H));
        const char ExpectedMagic[8]={'V','F','M','E','S','H','1','\0'};
        if(!File||std::memcmp(H.Magic,ExpectedMagic,sizeof(ExpectedMagic))!=0
            ||H.Version!=1||H.Endian!=0x01020304u||H.TileCount>1000000)
        {Error="mesh cache header is invalid";return false;}
        if(H.SourceDataHash!=Expected.SourceDataHash||H.Seed!=Expected.Seed
            ||H.ClipRadius!=Expected.ClipRadius||H.FullResClipLevels!=Expected.FullResClipLevels
            ||H.CoarseTileCells!=Expected.CoarseTileCells
            ||H.StrateContentCutMinLevel!=Expected.StrateContentCutMinLevel
            ||H.SkirtCells!=Expected.SkirtCells||H.LODOctaveDrop!=Expected.LODOctaveDrop
            ||H.bGenerateSkirts!=Expected.bGenerateSkirts
            ||H.AlgorithmVersion!=Expected.AlgorithmVersion||H.TunnelMode!=Expected.TunnelMode
            ||H.VoxelSizeCm!=Expected.VoxelSizeCm)
        {Error="mesh cache metadata does not match this world configuration";return false;}
        struct FRecord
        {
            int32 Level,X,Y,Z;uint64 Samples,DensityHash;uint32 VertexCount,IndexCount;
        };
        static_assert(sizeof(FRecord)==40);
        std::unordered_map<FVoxelTileKey,FStandaloneCpuTile,FStandaloneTileKeyHash> Loaded;
        Loaded.reserve(static_cast<size_t>(H.TileCount)*2+1);
        for(uint64 I=0;I<H.TileCount;++I)
        {
            FRecord Record{};File.read(reinterpret_cast<char*>(&Record),sizeof(Record));
            if(!File||Record.Level<0||Record.Level>8||Record.VertexCount>10000000
                ||Record.IndexCount>60000000)
            {Error="mesh cache contains an invalid tile record";return false;}
            FStandaloneCpuTile Tile;Tile.Key=FVoxelTileKey(FIntVector(Record.X,Record.Y,Record.Z),Record.Level);
            Tile.DensitySamples=Record.Samples;Tile.DensityHash=Record.DensityHash;Tile.bHasDensityHash=true;
            Tile.Vertices.resize(Record.VertexCount);Tile.Indices.resize(Record.IndexCount);
            if(!Tile.Vertices.empty())File.read(reinterpret_cast<char*>(Tile.Vertices.data()),
                static_cast<std::streamsize>(Tile.Vertices.size()*sizeof(FStandaloneCpuVertex)));
            if(!Tile.Indices.empty())File.read(reinterpret_cast<char*>(Tile.Indices.data()),
                static_cast<std::streamsize>(Tile.Indices.size()*sizeof(uint32)));
            if(!File){Error="mesh cache ended inside tile geometry";return false;}
            for(const uint32 Index:Tile.Indices)
                if(Index>=Tile.Vertices.size())
                {Error="mesh cache contains an out-of-range mesh index";return false;}
            for(const FStandaloneCpuVertex& Vertex:Tile.Vertices)
                for(const float Value:Vertex.Position)
                    if(!std::isfinite(Value))
                    {Error="mesh cache contains a non-finite vertex";return false;}
            for(const FStandaloneCpuVertex& Vertex:Tile.Vertices)
                for(const float Value:Vertex.Normal)
                    if(!std::isfinite(Value))
                    {Error="mesh cache contains a non-finite normal";return false;}
            if(!Loaded.emplace(Tile.Key,std::move(Tile)).second)
            {Error="mesh cache contains duplicate tile keys";return false;}
        }
        const std::streamoff End=File.tellg();
        if(End<0){Error="could not determine mesh cache size";return false;}
        BytesLoaded=static_cast<uint64>(End);Tiles.swap(Loaded);bLoaded=true;
        return true;
    }

private:
    std::unordered_map<FVoxelTileKey,FStandaloneCpuTile,FStandaloneTileKeyHash> Tiles;
    bool bLoaded=false;
    uint64 BytesLoaded=0;
};

struct FStandaloneStreamingUpdate
{
    FIntVector CenterChunk=FIntVector::ZeroValue;
    int32 DesiredTiles=0;
    int32 RetainedResidentTiles=0;
    int32 NewJobs=0;
    int32 CancelledQueuedJobs=0;
    int32 FullResTiles=0;
    std::vector<std::pair<uint64,uint64>> CancelledByCrossing;
};

struct FStandaloneTileJob
{
    FVoxelTileKey Key;
    int32 BandZMin=MIN_int32;
    int32 BandZMax=MAX_int32;
    uint64 RequestedCrossing=0;
    size_t PriorityRank=0;
    std::chrono::steady_clock::time_point QueuedAt{};
    std::chrono::steady_clock::time_point StartedAt{};
};

struct FStandaloneCpuTileTimelineEvent
{
    enum class EType : uint8 { Queued, Dispatch, Finish };
    EType Type=EType::Queued;
    FVoxelTileKey Key;
    int32 WorkerId=-1;
    size_t QueueDepth=0;
    size_t ActiveWorkers=0;
    size_t PriorityRank=0;
    std::chrono::steady_clock::time_point At{};
    std::chrono::steady_clock::time_point QueuedAt{};
    std::chrono::steady_clock::time_point StartedAt{};
    std::chrono::steady_clock::time_point FinishedAt{};
    double WorkerMilliseconds=-1.0;
    uint64 DensitySamples=0;
    uint32 SkirtTriangles=0;
    VoxelForgeCore::TileGenerationStats Diagnostics{};
};

struct FStandaloneTileJobHeapOrder
{
    bool operator()(const FStandaloneTileJob& A,const FStandaloneTileJob& B)const
    {
        if(A.PriorityRank!=B.PriorityRank)return A.PriorityRank>B.PriorityRank;
        if(A.Key.Level!=B.Key.Level)return A.Key.Level>B.Key.Level;
        if(A.Key.Coord.Z!=B.Key.Coord.Z)return A.Key.Coord.Z>B.Key.Coord.Z;
        if(A.Key.Coord.Y!=B.Key.Coord.Y)return A.Key.Coord.Y>B.Key.Coord.Y;
        return A.Key.Coord.X>B.Key.Coord.X;
    }
};

namespace StandaloneWorldMesher
{
    inline double TileDistanceSquaredChunks(const FVoxelTileKey& Tile,const FIntVector& Center)
    {
        const double Scale=static_cast<double>(1<<Tile.Level);
        const double DX=(static_cast<double>(Tile.Coord.X)+0.5)*Scale-Center.X;
        const double DY=(static_cast<double>(Tile.Coord.Y)+0.5)*Scale-Center.Y;
        const double DZ=(static_cast<double>(Tile.Coord.Z)+0.5)*Scale-Center.Z;
        return DX*DX+DY*DY+DZ*DZ;
    }

    inline FStandaloneDesiredTileSet BuildDesiredSet(
        const FStandaloneOwnerWorld& World,const FVector& CameraPositionVoxels)
    {
        FStandaloneDesiredTileSet Desired;
        Desired.CenterChunk=CameraChunk(CameraPositionVoxels);
        if(World.Settings->bClampViewToStrate)
            Desired.bHasStrate=World.Manager->GetStrateChunkZBounds(
                Desired.CenterChunk.Z,Desired.StrateTopChunkZ,Desired.StrateBottomChunkZ);
        VoxelClipmapDesiredTiles::ResolveVerticalBounds(
            Desired.CenterChunk.Z,World.Settings->ViewDistanceUp,World.Settings->ViewDistanceDown,
            World.Settings->StrateViewMarginChunks,World.Settings->bClampViewToStrate,
            Desired.bHasStrate,Desired.StrateTopChunkZ,Desired.StrateBottomChunkZ,
            Desired.MinimumChunkZ,Desired.MaximumChunkZ);
        VoxelClipmapDesiredTiles::FParameters Parameters;
        Parameters.ClipRadius=World.ClipRadius;
        Parameters.MaxClipLevel=FMath::Clamp(World.MaxClipLevel,0,8);
        Parameters.RenderDistanceChunks=World.RenderDistanceChunks;
        TArray<FVoxelTileKey> Tiles;
        VoxelClipmapDesiredTiles::Build(Desired.CenterChunk,Desired.MinimumChunkZ,
            Desired.MaximumChunkZ,Parameters,Tiles);
        Desired.Tiles.reserve(static_cast<size_t>(Tiles.Num()));
        for(const FVoxelTileKey& Tile:Tiles)Desired.Tiles.push_back(Tile);
        std::sort(Desired.Tiles.begin(),Desired.Tiles.end(),[&](const FVoxelTileKey& A,const FVoxelTileKey& B)
        {
            const double DA=TileDistanceSquaredChunks(A,Desired.CenterChunk);
            const double DB=TileDistanceSquaredChunks(B,Desired.CenterChunk);
            return DA==DB?A.Level<B.Level:DA<DB;
        });
        return Desired;
    }
}

class FStandaloneStreamingGenerator
{
public:
    static void SetCoarseFirstSchedulingEnabled(bool bEnabled)
    {bCoarseFirstSchedulingEnabled.store(bEnabled,std::memory_order_relaxed);}
    static bool IsCoarseFirstSchedulingEnabled()
    {return bCoarseFirstSchedulingEnabled.load(std::memory_order_relaxed);}

    explicit FStandaloneStreamingGenerator(FStandaloneOwnerWorld& InWorld,int32 InWorkerCount=8,
                                           bool bInHashDensity=false,
                                           const FStandaloneMeshCache* InMeshCache=nullptr,
                                           int32 InMaxLevelFilter=-1,
                                           bool bInSkipUvAndColors=false,
                                           bool bInSkipMaterialGroups=false,
                                           bool bInSkipUV=false,
                                           bool bInSkipColors=false,
                                           VoxelForgeCore::World* InCoreWorld=nullptr,
                                           bool bInCaptureColumnCoordinates=false,
                                           int32 InLod8HorizonInteriorStepVoxels=0,
                                           bool bInSharedColumnCache=true)
        :World(&InWorld),WorkerCount(FMath::Clamp(InWorkerCount,1,16)),bHashDensity(bInHashDensity),
         MeshCache(InMeshCache),MaxLevelFilter(InMaxLevelFilter),
         CoreWorld(InCoreWorld),bSkipUvAndColors(bInSkipUvAndColors),bSkipMaterialGroups(bInSkipMaterialGroups),
         bSkipUV(bInSkipUV),bSkipColors(bInSkipColors),
         bCaptureColumnCoordinates(bInCaptureColumnCoordinates),
         bSharedColumnCache(bInSharedColumnCache),
         Lod8HorizonInteriorStepVoxels(std::max(0,InLod8HorizonInteriorStepVoxels))
    {
        VoxelGenLOD::InitializeConsoleSwitches();
        UVoxelMarchingCubesMesher::InitializeConsoleSwitches();
    }
    FStandaloneStreamingGenerator(const FStandaloneStreamingGenerator&)=delete;
    FStandaloneStreamingGenerator& operator=(const FStandaloneStreamingGenerator&)=delete;
    ~FStandaloneStreamingGenerator(){Stop();}

    void EnableColdTimelineTrace(std::chrono::steady_clock::time_point Origin,
                                 const std::atomic<uint64>* FrameOrdinal)
    {
        TimelineOrigin=Origin;
        TimelineFrameOrdinal=FrameOrdinal;
        TimelineQueueEvents.clear();
        TimelineWorkerEvents.clear();
        TimelineWorkerEvents.resize(static_cast<size_t>(WorkerCount));
        bColdTimelineTrace=true;
    }

    void DumpColdTimelineTrace() const
    {
        if(!bColdTimelineTrace)return;
        std::vector<FStandaloneCpuTileTimelineEvent> Events=TimelineQueueEvents;
        size_t Total=Events.size();
        for(const auto& WorkerEvents:TimelineWorkerEvents)Total+=WorkerEvents.size();
        Events.reserve(Total);
        for(const auto& WorkerEvents:TimelineWorkerEvents)
            Events.insert(Events.end(),WorkerEvents.begin(),WorkerEvents.end());
        std::stable_sort(Events.begin(),Events.end(),[](const auto& A,const auto& B)
            {return A.At<B.At;});
        std::printf("[CpuTileTimeline] mode=instrumented events=%zu workers=%d flush=after_run\n",
            Events.size(),WorkerCount);
        for(const auto& Event:Events)
        {
            const char* Type=Event.Type==FStandaloneCpuTileTimelineEvent::EType::Queued
                ?"queued":Event.Type==FStandaloneCpuTileTimelineEvent::EType::Dispatch
                    ?"dispatch":"finish";
            std::printf("[CpuTileTimeline] event=%s key=(%d,%d,%d,lod%d) at_ms=%.3f frame=%llu worker_id=%d queue_depth=%zu active_workers=%zu priority_rank=%zu queued_ms=%.3f dispatch_ms=%.3f finish_ms=%.3f service_ms=%.3f worker_ms=%.3f\n",
                Type,Event.Key.Coord.X,Event.Key.Coord.Y,Event.Key.Coord.Z,Event.Key.Level,
                TimelineMilliseconds(Event.At),static_cast<unsigned long long>(TimelineFrame()),
                Event.WorkerId,Event.QueueDepth,Event.ActiveWorkers,Event.PriorityRank,
                TimelineMilliseconds(Event.QueuedAt),TimelineMilliseconds(Event.StartedAt),
                TimelineMilliseconds(Event.FinishedAt),
                Event.StartedAt.time_since_epoch().count()!=0
                    ?std::chrono::duration<double,std::milli>(Event.At-Event.StartedAt).count():-1.0,
                Event.WorkerMilliseconds);
            if(Event.Type==FStandaloneCpuTileTimelineEvent::EType::Finish)
                PrintTileStageEvent(Event);
        }
        std::fflush(stdout);
    }

    FStandaloneStreamingUpdate Start(const FVector& CameraPositionVoxels,
                                     const std::vector<FVoxelTileKey>& ResidentKeys={},
                                     uint64 CrossingIndex=0)
    {
        Stop();
        {
            std::lock_guard<std::mutex> Lock(Mutex);
            bStop=false;
        }
        for(int32 I=0;I<WorkerCount;++I)Workers.emplace_back([this,I](){WorkerLoop(I);});
        bHasLastCameraPosition=false;TravelDirection=FVector::ZeroVector;
        return UpdateDesired(CameraPositionVoxels,ResidentKeys,CrossingIndex);
    }

    FStandaloneStreamingUpdate StartWithDesired(const FStandaloneDesiredTileSet& InitialDesired,
                                     const std::vector<FVoxelTileKey>& ResidentKeys={},
                                     uint64 CrossingIndex=0)
    {
        Stop();
        {
            std::lock_guard<std::mutex> Lock(Mutex);
            bStop=false;
        }
        for(int32 I=0;I<WorkerCount;++I)Workers.emplace_back([this,I](){WorkerLoop(I);});
        bHasLastCameraPosition=false;TravelDirection=FVector::ZeroVector;
        return UpdateDesiredSet(InitialDesired,ResidentKeys,CrossingIndex);
    }

    void SetDesiredTileFilter(const std::vector<FVoxelTileKey>& Keys)
    {
        TileFilter.clear();TileFilter.reserve(Keys.size()*2+1);
        for(const FVoxelTileKey& Key:Keys)TileFilter.insert(Key);
        bUseTileFilter=true;
    }

    bool QueuePriorityTasks(std::vector<std::function<void()>> Tasks)
    {
        if(Tasks.empty())return true;
        {
            std::lock_guard<std::mutex> Lock(Mutex);
            if(bStop||Workers.empty())return false;
            for(auto& Task:Tasks)PriorityTasks.push_back(std::move(Task));
        }
        Wake.notify_all();
        return true;
    }

    void ClearPriorityTasks()
    {
        std::lock_guard<std::mutex> Lock(Mutex);
        PriorityTasks.clear();
    }

    void Stop()
    {
        {
            std::lock_guard<std::mutex> Lock(Mutex);
            bStop=true;
            Queue.clear();PriorityTasks.clear();
        }
        Wake.notify_all();
        for(FStandaloneWorkerThread& Worker:Workers)if(Worker.joinable())Worker.join();
        Workers.clear();
        std::lock_guard<std::mutex> Lock(Mutex);
        Ready.clear();Outstanding.clear();Desired.clear();Running.clear();
    }

    FStandaloneStreamingUpdate UpdateDesired(
        const FVector& CameraPositionVoxels,const std::vector<FVoxelTileKey>& ResidentKeys,
        uint64 CrossingIndex=0)
    {
        if(bHasLastCameraPosition)
        {
            const FVector Delta=CameraPositionVoxels-LastCameraPositionVoxels;
            const double LengthSquared=static_cast<double>(Delta.X)*Delta.X
                +static_cast<double>(Delta.Y)*Delta.Y;
            if(LengthSquared>1.0e-12)
            {
                const double InverseLength=1.0/std::sqrt(LengthSquared);
                TravelDirection=FVector(static_cast<float>(Delta.X*InverseLength),
                    static_cast<float>(Delta.Y*InverseLength),0.0f);
            }
            else TravelDirection=FVector::ZeroVector;
        }
        else
            TravelDirection=FVector::ZeroVector;
        LastCameraPositionVoxels=CameraPositionVoxels;
        bHasLastCameraPosition=true;
        FStandaloneDesiredTileSet NewDesired=StandaloneWorldMesher::BuildDesiredSet(*World,CameraPositionVoxels);
        if(CoreWorld)
        {
            const auto CoreTiles=CoreWorld->desiredTiles({CameraPositionVoxels.X,CameraPositionVoxels.Y,CameraPositionVoxels.Z});
            NewDesired.Tiles.clear();NewDesired.Tiles.reserve(CoreTiles.size());
            for(const auto& Key:CoreTiles)
                NewDesired.Tiles.emplace_back(FIntVector(Key.coordinate.x,Key.coordinate.y,Key.coordinate.z),Key.lod);
        }
        return UpdateDesiredSet(std::move(NewDesired),ResidentKeys,CrossingIndex);
    }

    FStandaloneStreamingUpdate UpdateDesiredWithSet(
        FStandaloneDesiredTileSet NewDesired,const std::vector<FVoxelTileKey>& ResidentKeys,
        uint64 CrossingIndex=0)
    {
        return UpdateDesiredSet(std::move(NewDesired),ResidentKeys,CrossingIndex);
    }

    std::vector<FStandaloneCpuTile> TakeReady()
    {
        std::vector<FStandaloneCpuTile> Out;
        std::lock_guard<std::mutex> Lock(Mutex);
        Out.swap(Ready);
        for(const FStandaloneCpuTile& Tile:Out)Outstanding.erase(Tile.Key);
        return Out;
    }

    FStandaloneStreamingUpdate RestrictDesiredTileFilterKeepingStarted(
        const std::vector<FVoxelTileKey>& BaseFilter,
        const std::vector<FVoxelTileKey>& CandidateKeys,
        const std::vector<FVoxelTileKey>& ResidentKeys,
        std::vector<FVoxelTileKey>& RetainedCandidateKeys)
    {
        std::unordered_set<FVoxelTileKey,FStandaloneTileKeyHash> Candidates;
        Candidates.reserve(CandidateKeys.size()*2+1);
        for(const FVoxelTileKey& Key:CandidateKeys)Candidates.insert(Key);

        FStandaloneStreamingUpdate Stats;
        std::lock_guard<std::mutex> Lock(Mutex);
        std::unordered_set<FVoxelTileKey,FStandaloneTileKeyHash> Retained;
        Retained.reserve(CandidateKeys.size()*2+1);
        for(const FVoxelTileKey& Key:Running)
            if(Candidates.contains(Key))Retained.insert(Key);
        for(const FStandaloneCpuTile& Tile:Ready)
            if(Candidates.contains(Tile.Key))Retained.insert(Tile.Key);
        for(const FVoxelTileKey& Key:ResidentKeys)
            if(Candidates.contains(Key))Retained.insert(Key);

        TileFilter.clear();
        TileFilter.reserve((BaseFilter.size()+Retained.size())*2+1);
        for(const FVoxelTileKey& Key:BaseFilter)TileFilter.insert(Key);
        for(const FVoxelTileKey& Key:Retained)TileFilter.insert(Key);
        bUseTileFilter=true;

        FStandaloneDesiredTileSet NewDesired=CurrentDesired;
        NewDesired.Tiles.erase(std::remove_if(NewDesired.Tiles.begin(),NewDesired.Tiles.end(),
            [&](const FVoxelTileKey& Key){return !TileFilter.contains(Key);}),NewDesired.Tiles.end());
        Stats.CenterChunk=NewDesired.CenterChunk;
        Stats.DesiredTiles=static_cast<int32>(NewDesired.Tiles.size());
        std::unordered_set<FVoxelTileKey,FStandaloneTileKeyHash> ResidentSet;
        ResidentSet.reserve(ResidentKeys.size()*2+1);
        for(const FVoxelTileKey& Key:ResidentKeys)ResidentSet.insert(Key);
        for(const FVoxelTileKey& Key:NewDesired.Tiles)
        {
            if(Key.Level<World->FullResClipLevels)++Stats.FullResTiles;
            if(ResidentSet.contains(Key))++Stats.RetainedResidentTiles;
        }

        Desired.clear();Desired.reserve(NewDesired.Tiles.size()*2+1);
        for(const FVoxelTileKey& Key:NewDesired.Tiles)Desired.insert(Key);
        FullResTileCount=Stats.FullResTiles;
        std::unordered_map<uint64,uint64> CancelledByCrossing;
        auto QueueEnd=std::remove_if(Queue.begin(),Queue.end(),[&](const FStandaloneTileJob& Job)
        {
            if(Desired.contains(Job.Key))return false;
            Outstanding.erase(Job.Key);++Stats.CancelledQueuedJobs;
            ++CancelledByCrossing[Job.RequestedCrossing];return true;
        });
        Queue.erase(QueueEnd,Queue.end());
        std::make_heap(Queue.begin(),Queue.end(),FStandaloneTileJobHeapOrder{});
        Stats.CancelledByCrossing.reserve(CancelledByCrossing.size());
        for(const auto& Pair:CancelledByCrossing)Stats.CancelledByCrossing.push_back(Pair);
        CurrentDesired=std::move(NewDesired);
        RetainedCandidateKeys.assign(Retained.begin(),Retained.end());
        std::sort(RetainedCandidateKeys.begin(),RetainedCandidateKeys.end(),
            [](const FVoxelTileKey& A,const FVoxelTileKey& B)
            {
                if(A.Level!=B.Level)return A.Level<B.Level;
                if(A.Coord.Z!=B.Coord.Z)return A.Coord.Z<B.Coord.Z;
                if(A.Coord.Y!=B.Coord.Y)return A.Coord.Y<B.Coord.Y;
                return A.Coord.X<B.Coord.X;
            });
        return Stats;
    }

    bool IsDesired(const FVoxelTileKey& Key)const
    {std::lock_guard<std::mutex> Lock(Mutex);return Desired.contains(Key);}
    int32 DesiredCount()const
    {std::lock_guard<std::mutex> Lock(Mutex);return static_cast<int32>(Desired.size());}
    int32 FullResDesiredCount()const
    {std::lock_guard<std::mutex> Lock(Mutex);return FullResTileCount;}
    FIntVector CenterChunk()const
    {std::lock_guard<std::mutex> Lock(Mutex);return CurrentDesired.CenterChunk;}
    int32 GetWorkerCount()const{return WorkerCount;}
    uint64 GetCacheHits()const{return CacheHitCount.load(std::memory_order_relaxed);}

    void SetMesherStageProfilingEnabled(bool Enabled)
    {bProfileMesherStages=Enabled;}
    void SetTileDrainInstrumentationEnabled(bool Enabled)
    {bInstrumentTileDrain=Enabled;}
    void SetPassageComposerProfilingEnabled(bool Enabled)
    {bProfilePassageComposer=Enabled;}
    void SetExactMidpointDensityMemoEnabled(bool Enabled)
    {bExactMidpointDensityMemo=Enabled;}
    void SetActiveLaneBatchingEnabled(bool Enabled)
    {bActiveLaneBatching=Enabled;}
    void SetProofBlockCells(int32 Cells)
    {ProofBlockCells=std::clamp(Cells,1,32);}
    void SetUniformBlockProofAuditEnabled(bool Enabled)
    {bAuditUniformBlockProofs=Enabled;}
    FStandaloneCpuTile GenerateDiagnostic(const FVoxelTileKey& Key,
        std::uint32_t GlobalOctaveDrop=0,std::uint32_t CellStride=1,
        bool ForceGenerate=false)
    {
        FStandaloneTileJob Job;Job.Key=Key;Job.StartedAt=FClock::now();Job.QueuedAt=Job.StartedAt;
        return Generate(Job,GlobalOctaveDrop,CellStride,ForceGenerate);
    }

private:
    FStandaloneStreamingUpdate UpdateDesiredSet(FStandaloneDesiredTileSet NewDesired,
        const std::vector<FVoxelTileKey>& ResidentKeys,uint64 CrossingIndex)
    {
        if(MaxLevelFilter>=0)
            NewDesired.Tiles.erase(std::remove_if(NewDesired.Tiles.begin(),NewDesired.Tiles.end(),
                [&](const FVoxelTileKey& Key){return Key.Level>MaxLevelFilter;}),NewDesired.Tiles.end());
        if(bUseTileFilter)
            NewDesired.Tiles.erase(std::remove_if(NewDesired.Tiles.begin(),NewDesired.Tiles.end(),
                [&](const FVoxelTileKey& Key){return !TileFilter.contains(Key);}),NewDesired.Tiles.end());
        FStandaloneStreamingUpdate Stats;
        Stats.CenterChunk=NewDesired.CenterChunk;
        Stats.DesiredTiles=static_cast<int32>(NewDesired.Tiles.size());
        std::unordered_set<FVoxelTileKey,FStandaloneTileKeyHash> Resident;
        Resident.reserve(ResidentKeys.size()*2+1);
        for(const FVoxelTileKey& Key:ResidentKeys)Resident.insert(Key);
        for(const FVoxelTileKey& Key:NewDesired.Tiles)
        {
            if(Key.Level<World->FullResClipLevels)++Stats.FullResTiles;
            if(Resident.contains(Key))++Stats.RetainedResidentTiles;
        }

        std::lock_guard<std::mutex> Lock(Mutex);
        Desired.clear();Desired.reserve(NewDesired.Tiles.size()*2+1);
        for(const FVoxelTileKey& Key:NewDesired.Tiles)Desired.insert(Key);
        FullResTileCount=Stats.FullResTiles;
        std::unordered_map<uint64,uint64> CancelledByCrossing;
        auto QueueEnd=std::remove_if(Queue.begin(),Queue.end(),[&](const FStandaloneTileJob& Job)
        {
            if(Desired.contains(Job.Key))return false;
            Outstanding.erase(Job.Key);++Stats.CancelledQueuedJobs;
            ++CancelledByCrossing[Job.RequestedCrossing];return true;
        });
        Queue.erase(QueueEnd,Queue.end());
        Stats.CancelledByCrossing.reserve(CancelledByCrossing.size());
        for(const auto& Pair:CancelledByCrossing)Stats.CancelledByCrossing.push_back(Pair);

        std::vector<FVoxelTileKey> Prioritized=NewDesired.Tiles;
        if(IsCoarseFirstSchedulingEnabled())
        {
            // The local walkable ring leads, the coarse enclosure follows, then target-detail work.
            // L'anneau praticable local passe d'abord, puis l'enveloppe grossiere, puis le detail cible.
            const int32 FirstCoarseLevel=FMath::Max(2,World->FullResClipLevels);
            auto Tier=[&](const FVoxelTileKey& Key)
            {
                if(Key.Level==0&&std::abs(Key.Coord.X-NewDesired.CenterChunk.X)<=1
                    &&std::abs(Key.Coord.Y-NewDesired.CenterChunk.Y)<=1
                    &&std::abs(Key.Coord.Z-NewDesired.CenterChunk.Z)<=1)return 0;
                return Key.Level>=FirstCoarseLevel?1:2;
            };
            auto IsAhead=[&](const FVoxelTileKey& Key)
            {
                if(TravelDirection.X*TravelDirection.X+TravelDirection.Y*TravelDirection.Y<=1.0e-12f)
                    return false;
                const double Scale=static_cast<double>(1<<Key.Level);
                const double OffsetX=(static_cast<double>(Key.Coord.X)+0.5)*Scale-NewDesired.CenterChunk.X;
                const double OffsetY=(static_cast<double>(Key.Coord.Y)+0.5)*Scale-NewDesired.CenterChunk.Y;
                return OffsetX*TravelDirection.X+OffsetY*TravelDirection.Y>0.0;
            };
            std::sort(Prioritized.begin(),Prioritized.end(),[&](const FVoxelTileKey& A,const FVoxelTileKey& B)
            {
                const int32 TierA=Tier(A),TierB=Tier(B);
                if(TierA!=TierB)return TierA<TierB;
                if(TierA==1&&A.Level!=B.Level)return A.Level>B.Level;
                const bool bAheadA=IsAhead(A),bAheadB=IsAhead(B);
                if(bAheadA!=bAheadB)return bAheadA;
                const double DA=StandaloneWorldMesher::TileDistanceSquaredChunks(A,NewDesired.CenterChunk);
                const double DB=StandaloneWorldMesher::TileDistanceSquaredChunks(B,NewDesired.CenterChunk);
                if(DA!=DB)return DA<DB;
                if(A.Level!=B.Level)return A.Level<B.Level;
                if(A.Coord.Z!=B.Coord.Z)return A.Coord.Z<B.Coord.Z;
                if(A.Coord.Y!=B.Coord.Y)return A.Coord.Y<B.Coord.Y;
                return A.Coord.X<B.Coord.X;
            });
        }
        else
        {
            std::vector<VoxelForgeCore::TileKey> Unordered;Unordered.reserve(NewDesired.Tiles.size());
            for(const FVoxelTileKey& Key:NewDesired.Tiles)
                Unordered.push_back({{Key.Coord.X,Key.Coord.Y,Key.Coord.Z},Key.Level});
            const auto CorePrioritized=VoxelForgeCore::World::prioritizeTiles(Unordered,
                {NewDesired.CenterChunk.X,NewDesired.CenterChunk.Y,NewDesired.CenterChunk.Z});
            Prioritized.clear();Prioritized.reserve(CorePrioritized.size());
            for(const auto& Key:CorePrioritized)
                Prioritized.emplace_back(FIntVector(Key.coordinate.x,Key.coordinate.y,Key.coordinate.z),Key.lod);
        }
        std::unordered_map<FVoxelTileKey,size_t,FStandaloneTileKeyHash> PriorityRanks;
        PriorityRanks.reserve(Prioritized.size()*2+1);
        for(size_t Rank=0;Rank<Prioritized.size();++Rank)
        {
            PriorityRanks.emplace(Prioritized[Rank],Rank);
        }
        for(FStandaloneTileJob& Job:Queue)Job.PriorityRank=PriorityRanks.at(Job.Key);

        const int32 BandMin=World->StrateContentCutMinLevel<9&&NewDesired.bHasStrate
            ? NewDesired.StrateBottomChunkZ*CHUNK_SIZE:MIN_int32;
        const int32 BandMax=World->StrateContentCutMinLevel<9&&NewDesired.bHasStrate
            ? (NewDesired.StrateTopChunkZ+1)*CHUNK_SIZE:MAX_int32;
        for(const FVoxelTileKey& Key:NewDesired.Tiles)
        {
            if(Resident.contains(Key)||Outstanding.contains(Key))continue;
            FStandaloneTileJob Job;
            Job.Key=Key;Job.BandZMin=BandMin;Job.BandZMax=BandMax;
            Job.RequestedCrossing=CrossingIndex;Job.PriorityRank=PriorityRanks.at(Key);
            Job.QueuedAt=FClock::now();
            Queue.push_back(Job);
            if(bColdTimelineTrace)
                RecordTimelineEvent(FStandaloneCpuTileTimelineEvent::EType::Queued,Key,-1,
                    Queue.size(),Running.size(),Job.PriorityRank,Job.QueuedAt,Job.QueuedAt);
            Outstanding.insert(Key);++Stats.NewJobs;
        }
        CurrentDesired=std::move(NewDesired);
        std::make_heap(Queue.begin(),Queue.end(),FStandaloneTileJobHeapOrder{});
        Wake.notify_all();
        return Stats;
    }
    FStandaloneCpuTile Generate(const FStandaloneTileJob& Job,
        std::uint32_t DiagnosticGlobalOctaveDrop=0,std::uint32_t DiagnosticCellStride=1,
        bool ForceGenerate=false)
    {
        const auto StartTime=Job.StartedAt;
        FStandaloneCpuTile Out;Out.Key=Job.Key;Out.RequestedCrossing=Job.RequestedCrossing;
        Out.QueuedAt=Job.QueuedAt;Out.StartedAt=Job.StartedAt;
        if(MeshCache&&!ForceGenerate)
        {
            if(const FStandaloneCpuTile* Cached=MeshCache->Find(Job.Key))
            {
                Out=*Cached;Out.RequestedCrossing=Job.RequestedCrossing;
                Out.WorkerSeconds=std::chrono::duration<double>(FClock::now()-StartTime).count();
                Out.CompletedAt=FClock::now();
                if(bHashDensity)
                {Out.MeshHash=HashStandaloneCpuTileMesh(Out);Out.bHasMeshHash=true;}
                CacheHitCount.fetch_add(1,std::memory_order_relaxed);
                return Out;
            }
        }
        UVoxelMarchingCubesMesher Mesher;
        Mesher.SetGenerator(World->Generator.get());
        Mesher.bStandaloneSkipUvAndColors=bSkipUvAndColors;
        Mesher.bStandaloneSkipUV=bSkipUV;
        Mesher.bStandaloneSkipColors=bSkipColors;
        Mesher.bStandaloneSkipMaterialGroups=bSkipMaterialGroups;
        Mesher.bGenerateSkirts=World->bGenerateSkirts;
        Mesher.SkirtCells=static_cast<float>(World->SkirtCells);
        Mesher.LODOctaveDrop=World->LODOctaveDrop;
        const int32 Level=Job.Key.Level;
        const int32 Extent=Job.Key.ExtentVoxels();
        const int32 Cells=Level<World->FullResClipLevels?CHUNK_SIZE:FMath::Max(1,World->CoarseTileCells);
        const int32 Step=FMath::Max(1,Extent/Cells);
        int64 SampleCount=0;
        const bool bProfile=VoxelDensityProfile::GetMode()!=VoxelDensityProfile::EMode::Disabled;
        const VoxelDensityProfile::FSnapshot Before=bProfile
            ? VoxelDensityProfile::SnapshotCurrentThread():VoxelDensityProfile::FSnapshot{};
        if(CoreWorld)
        {
            VoxelForgeCore::TileGenerationOptions Options;
            Options.bandMinZVoxel=Job.BandZMin;Options.bandMaxZVoxel=Job.BandZMax;
            Options.hashDensity=bHashDensity;
            Options.skipUvAndColors=bSkipUvAndColors;
            Options.skipMaterialGroups=bSkipMaterialGroups;Options.skipUv=bSkipUV;Options.skipColors=bSkipColors;
            Options.captureColumnCoordinates=bCaptureColumnCoordinates;
            Options.useSharedColumnCache=bSharedColumnCache;
            Options.lod8HorizonInteriorStepVoxels=Lod8HorizonInteriorStepVoxels;
            Options.profileMesherStages=(bProfileMesherStages||bInstrumentTileDrain)&&!ForceGenerate;
            Options.profilePassageComposer=bProfilePassageComposer;
            Options.instrumentTileDrain=bInstrumentTileDrain&&!ForceGenerate;
            Options.exactMidpointDensityMemo=bExactMidpointDensityMemo&&!ForceGenerate;
            Options.activeLaneBatching=bActiveLaneBatching&&!ForceGenerate;
            Options.proofBlockCells=static_cast<std::uint32_t>(ProofBlockCells);
            Options.auditUniformBlockProofs=bAuditUniformBlockProofs&&!ForceGenerate;
            Options.diagnosticGlobalOctaveDrop=DiagnosticGlobalOctaveDrop;
            Options.diagnosticCellStride=std::max<std::uint32_t>(1,DiagnosticCellStride);
            Options.diagnosticColumnInterpolationStride=
                std::max<std::uint32_t>(1,DiagnosticCellStride);
            std::string Error;
            auto Result=CoreWorld->generateTile(
                {{Job.Key.Coord.X,Job.Key.Coord.Y,Job.Key.Coord.Z},Job.Key.Level},Options,Error);
            if(!Result)
            {
                std::fprintf(stderr,"[VoxelForgeCore] tile generation failed: %s\n",Error.c_str());
                Out.bGenerationFailed=true;
                Out.WorkerSeconds=std::chrono::duration<double>(FClock::now()-StartTime).count();
                Out.CompletedAt=FClock::now();return Out;
            }
            Out.DensitySamples=Result->densitySampleCount;
            Out.CoreDiagnostics=Result->diagnostics;
            Out.SurfaceVertexCount=Result->groups.surfaceVertexCount;
            Out.SkirtTriangleCount=Result->groups.skirtTriangleCount;
            Out.MeshPayloadAllocatedBytes=Result->meshPayloadAllocatedBytes;
            Out.DensityHash=Result->densityHash;Out.MeshHash=Result->meshHash;
            Out.PayloadHash=Result->payloadHash;Out.TopologyHash=Result->topologyHash;
            Out.MaterialHash=Result->materialHash;
            Out.bHasDensityHash=Result->hasDensityHash;Out.bHasMeshHash=Result->hasMeshHash;
            Out.bHasPayloadHash=Result->hasPayloadHash;Out.bHasTopologyHash=Result->hasTopologyHash;
            Out.bHasMaterialHash=Result->hasMaterialHash;
            Out.Vertices.reserve(Result->vertices.size());
            for(const auto& Vertex:Result->vertices)
                Out.Vertices.push_back({
                    {static_cast<float>(Vertex.position.x*0.01),static_cast<float>(Vertex.position.y*0.01),
                     static_cast<float>(Vertex.position.z*0.01)},
                    {static_cast<float>(Vertex.normal.x),static_cast<float>(Vertex.normal.y),
                     static_cast<float>(Vertex.normal.z)},
                    Vertex.hostMaterialId,Vertex.resourceMaterialIds,Vertex.resourceWeightsQ16});
            Out.Indices=std::move(Result->indices);
            Out.TriangleSkirtFaceMasks=std::move(Result->skirtFaceMasks);
            if(bProfile)
            {
                const VoxelDensityProfile::FSnapshot After=VoxelDensityProfile::SnapshotCurrentThread();
                for(size_t I=0;I<Out.ProfileCycles.size();++I)
                    Out.ProfileCycles[I]=After.Cycles[I]>=Before.Cycles[I]?After.Cycles[I]-Before.Cycles[I]:0;
            }
            Out.WorkerSeconds=std::chrono::duration<double>(FClock::now()-StartTime).count();
            Out.CompletedAt=FClock::now();return Out;
        }
        std::vector<float> DensityValues;
        FVoxelMeshData Mesh=Mesher.GenerateMesh(Job.Key.OriginVoxels(),Step,Cells,nullptr,
            Job.BandZMin,Job.BandZMax,&SampleCount,nullptr,nullptr,
            bHashDensity?&DensityValues:nullptr);
        if(bHashDensity)
        {
            Out.bHasDensityHash=true;
            for(const float Density:DensityValues)
            {
                Out.DensityHash^=std::bit_cast<uint32>(Density);
                Out.DensityHash*=1099511628211ull;
            }
        }
        if(bProfile)
        {
            const VoxelDensityProfile::FSnapshot After=VoxelDensityProfile::SnapshotCurrentThread();
            for(size_t I=0;I<Out.ProfileCycles.size();++I)
            {
                const size_t Bucket=I;
                Out.ProfileCycles[I]=After.Cycles[Bucket]>=Before.Cycles[Bucket]
                    ?After.Cycles[Bucket]-Before.Cycles[Bucket]:0;
            }
        }
        Out.DensitySamples=static_cast<uint64>(FMath::Max<int64>(0,SampleCount));
        Out.SurfaceVertexCount=static_cast<uint32>(std::clamp(Mesh.NumSurfaceVertices,0,Mesh.Vertices.Num()));
        Out.SkirtTriangleCount=static_cast<uint32>(std::max(0,Mesh.NumSkirtTriangles));
        Out.MeshPayloadAllocatedBytes=static_cast<uint64>(Mesh.Vertices.GetAllocatedSize())
            +static_cast<uint64>(Mesh.Normals.GetAllocatedSize())
            +static_cast<uint64>(Mesh.UVs.GetAllocatedSize())
            +static_cast<uint64>(Mesh.Colors.GetAllocatedSize())
            +static_cast<uint64>(Mesh.TriangleSkirtFaceMasks.GetAllocatedSize())
            +static_cast<uint64>(Mesh.Triangles.GetAllocatedSize());
        if(bHashDensity)
        {
            Out.PayloadHash=HashStandaloneMeshPayload(Mesh);Out.bHasPayloadHash=true;
            Out.TopologyHash=HashStandaloneMeshTopology(Mesh);Out.bHasTopologyHash=true;
        }
        Out.Vertices.reserve(static_cast<size_t>(Mesh.Vertices.Num()));
        for(int32 I=0;I<Mesh.Vertices.Num();++I)
        {
            const FVector& P=Mesh.Vertices[I];const FVector& N=Mesh.Normals[I];
            Out.Vertices.push_back({
                {static_cast<float>(P.X*0.01),static_cast<float>(P.Y*0.01),static_cast<float>(P.Z*0.01)},
                {static_cast<float>(N.X),static_cast<float>(N.Y),static_cast<float>(N.Z)}});
        }
        Out.Indices.reserve(static_cast<size_t>(Mesh.Triangles.Num()));
        for(const int32 Index:Mesh.Triangles)Out.Indices.push_back(static_cast<uint32>(Index));
        Out.TriangleSkirtFaceMasks.assign(Mesh.TriangleSkirtFaceMasks.begin(),
            Mesh.TriangleSkirtFaceMasks.end());
        Out.TriangleSkirtFaceMasks.resize(Out.Indices.size()/3,0);
        if(bHashDensity)
        {Out.MeshHash=HashStandaloneCpuTileMesh(Out);Out.bHasMeshHash=true;}
        Out.WorkerSeconds=std::chrono::duration<double>(FClock::now()-StartTime).count();
        Out.CompletedAt=FClock::now();
        return Out;
    }

    void WorkerLoop(int32 WorkerId)
    {
        constexpr uint32 PriorityTaskBurst=3;
        uint32 ConsecutivePriorityTasks=0;
        for(;;)
        {
            FStandaloneTileJob Job;
            std::function<void()> PriorityTask;
            size_t DispatchQueueDepth=0,DispatchActiveWorkers=0;
            {
                std::unique_lock<std::mutex> Lock(Mutex);
                Wake.wait(Lock,[&](){return bStop||!PriorityTasks.empty()||!Queue.empty();});
                if(bStop)return;
                const bool bPreferPriority=!PriorityTasks.empty()
                    &&(Queue.empty()||ConsecutivePriorityTasks<PriorityTaskBurst);
                if(bPreferPriority)
                {
                    PriorityTask=std::move(PriorityTasks.front());PriorityTasks.pop_front();
                    ++ConsecutivePriorityTasks;
                }
                else
                {
                    std::pop_heap(Queue.begin(),Queue.end(),FStandaloneTileJobHeapOrder{});
                    Job=Queue.back();Queue.pop_back();Job.StartedAt=FClock::now();
                    Running.insert(Job.Key);ConsecutivePriorityTasks=0;
                    DispatchQueueDepth=Queue.size();DispatchActiveWorkers=Running.size();
                }
            }
            if(PriorityTask)
            {
                PriorityTask();
                continue;
            }
            if(bColdTimelineTrace)
                RecordTimelineEvent(FStandaloneCpuTileTimelineEvent::EType::Dispatch,Job.Key,
                    WorkerId,DispatchQueueDepth,DispatchActiveWorkers,Job.PriorityRank,
                    Job.StartedAt,Job.QueuedAt,Job.StartedAt);
            FStandaloneCpuTile Tile=Generate(Job);
            Tile.WorkerId=WorkerId;
            {
                std::lock_guard<std::mutex> Lock(Mutex);
                if(bStop)return;
                Running.erase(Tile.Key);
                if(bColdTimelineTrace)
                    RecordTimelineEvent(FStandaloneCpuTileTimelineEvent::EType::Finish,Tile.Key,
                        WorkerId,Queue.size(),Running.size(),Job.PriorityRank,Tile.CompletedAt,
                        Tile.QueuedAt,Tile.StartedAt,Tile.CompletedAt,Tile.WorkerSeconds*1000.0,
                        &Tile.CoreDiagnostics,Tile.DensitySamples,Tile.SkirtTriangleCount);
                Ready.push_back(std::move(Tile));
            }
        }
    }

    using FClock=std::chrono::steady_clock;
    double TimelineMilliseconds(FClock::time_point At)const
    {
        return At.time_since_epoch().count()==0?-1.0:
            std::chrono::duration<double,std::milli>(At-TimelineOrigin).count();
    }
    uint64 TimelineFrame()const
    {
        return TimelineFrameOrdinal?TimelineFrameOrdinal->load(std::memory_order_relaxed):0;
    }
    static void PrintTileStageEvent(const FStandaloneCpuTileTimelineEvent& Event)
    {
        const auto& D=Event.Diagnostics;
        std::printf("[CpuTileStages] key=(%d,%d,%d,lod%d) worker_id=%d worker_ms=%.3f samples=%llu classify_ms=%.3f column_ms=%.3f column_exclusive_ms=%.3f mc_ms=%.3f mc_grid_ms=%.3f mc_cell_ms=%.3f mc_gradient_ms=%.3f mc_interpolation_ms=%.3f mc_dedup_ms=%.3f mc_skirt_ms=%.3f mc_other_ms=%.3f connector_ms=%.3f ground_ms=%.3f cap_ms=%.3f horizon_ms=%.3f mesh_assembly_ms=%.3f candidate_cells=%llu classified_cells=%llu surface_cells=%llu proof_solid_cells=%llu proof_air_cells=%llu proof_density_samples=%llu zband_reject_cells=%llu x8_calls=%llu x8_active_lanes=%llu grad_calls=%llu grad_memo_hits=%llu grad_fallback_calls=%llu midpoint_attempts=%llu midpoint_memo_hits=%llu skirt_edges=%llu skirt_boundary_edges=%llu skirt_density_probes=%llu skirt_triangles=%u fi_local_hit=%llu fi_local_miss=%llu fi_local_conflicts=%llu fi_shared_hit=%llu fi_shared_miss=%llu fi_shared_computes=%llu fi_shared_conflicts=%llu fi_lock_wait_us=%.3f\n",
            Event.Key.Coord.X,Event.Key.Coord.Y,Event.Key.Coord.Z,Event.Key.Level,
            Event.WorkerId,Event.WorkerMilliseconds,
            static_cast<unsigned long long>(Event.DensitySamples),D.classificationMilliseconds,
            D.columnComputeMilliseconds,D.columnEvaluationExclusiveMilliseconds,
            D.marchingCubesExclusiveMilliseconds+D.marchingCubesSkirtMilliseconds,
            D.marchingCubesDensityGridMilliseconds,D.marchingCubesCellClassificationMilliseconds,
            D.marchingCubesGradientNormalsMilliseconds,D.marchingCubesEdgeInterpolationMilliseconds,
            D.marchingCubesVertexDedupStreamMilliseconds,D.marchingCubesSkirtMilliseconds,
            D.marchingCubesOtherMilliseconds,D.farConnectorShellMilliseconds,
            D.groundRefinementExclusiveMilliseconds,D.capRefinementExclusiveMilliseconds,
            D.farHorizonMilliseconds,D.meshAssemblyMilliseconds,
            static_cast<unsigned long long>(D.marchingCubesCandidateCells),
            static_cast<unsigned long long>(D.marchingCubesCellClassificationCalls),
            static_cast<unsigned long long>(D.marchingCubesSurfaceCellCalls),
            static_cast<unsigned long long>(D.marchingCubesProofSkippedSolidCells),
            static_cast<unsigned long long>(D.marchingCubesProofSkippedAirCells),
            static_cast<unsigned long long>(D.marchingCubesProofSkippedDensitySamples),
            static_cast<unsigned long long>(D.marchingCubesZBandRejectedCells),
            static_cast<unsigned long long>(D.marchingCubesX8DensityCalls),
            static_cast<unsigned long long>(D.marchingCubesX8ActiveLanes),
            static_cast<unsigned long long>(D.marchingCubesGradientCalls),
            static_cast<unsigned long long>(D.marchingCubesGradientMemoHits),
            static_cast<unsigned long long>(D.marchingCubesGradientFallbackDensityCalls),
            static_cast<unsigned long long>(D.marchingCubesEdgeMidpointDensityCalls),
            static_cast<unsigned long long>(D.marchingCubesEdgeMidpointDensityMemoHits),
            static_cast<unsigned long long>(D.marchingCubesSkirtEdgeCandidates),
            static_cast<unsigned long long>(D.marchingCubesSkirtBoundaryCandidates),
            static_cast<unsigned long long>(D.marchingCubesSkirtDensityProbes),
            Event.SkirtTriangles,
            static_cast<unsigned long long>(D.islandLocalCacheHits),
            static_cast<unsigned long long>(D.islandLocalCacheMisses),
            static_cast<unsigned long long>(D.islandLocalCacheConflictReplacements),
            static_cast<unsigned long long>(D.islandSharedCacheHits),
            static_cast<unsigned long long>(D.islandSharedCacheMisses),
            static_cast<unsigned long long>(D.islandSharedCacheComputes),
            static_cast<unsigned long long>(D.islandSharedCacheConflictReplacements),
            static_cast<double>(D.islandSharedCacheLockWaitNanoseconds)/1000.0);
        if(D.passageComposerProfiled)
            std::printf("[PassageComposerTile] key=(%d,%d,%d,lod%d) worker_ms=%.3f samples=%llu native_field_calls=%llu features=%llu route_candidates=%llu mouth_candidates=%llu queries=%llu passage_iterations=%llu upper_mouth_checks=%llu column_candidates=%llu column_distance_calls=%llu negative_bounds=%llu negative_rejects=%llu positive_bounds=%llu positive_rejects=%llu carve_segment_visits=%llu smooth_carve_joins=%llu support_segment_visits=%llu smooth_support_joins=%llu support_capsule_checks=%llu later_bore_checks=%llu other_capsule_checks=%llu mouth_support_calls=%llu mouth_route_segment_checks=%llu relief_field_calls=%llu smoothstep_evaluations=%llu\n",
                Event.Key.Coord.X,Event.Key.Coord.Y,Event.Key.Coord.Z,Event.Key.Level,
                Event.WorkerMilliseconds,static_cast<unsigned long long>(Event.DensitySamples),
                static_cast<unsigned long long>(D.fieldCalls),
                static_cast<unsigned long long>(D.passageTileFeatureCount),
                static_cast<unsigned long long>(D.passageTileRouteSegmentCandidates),
                static_cast<unsigned long long>(D.passageTileMouthSegmentCandidates),
                static_cast<unsigned long long>(D.passageComposerQueries),
                static_cast<unsigned long long>(D.passageIterations),
                static_cast<unsigned long long>(D.passageUpperMouthFilterChecks),
                static_cast<unsigned long long>(D.passageColumnCandidates),
                static_cast<unsigned long long>(D.passageColumnDistanceCalls),
                static_cast<unsigned long long>(D.passageNegativeBoundChecks),
                static_cast<unsigned long long>(D.passageNegativeBoundRejects),
                static_cast<unsigned long long>(D.passagePositiveBoundChecks),
                static_cast<unsigned long long>(D.passagePositiveBoundRejects),
                static_cast<unsigned long long>(D.passageCarveSegmentVisits),
                static_cast<unsigned long long>(D.passageSmoothCarveJoins),
                static_cast<unsigned long long>(D.passageSupportSegmentVisits),
                static_cast<unsigned long long>(D.passageSmoothSupportJoins),
                static_cast<unsigned long long>(D.passageSupportCapsuleChecks),
                static_cast<unsigned long long>(D.passageLaterBoreChecks),
                static_cast<unsigned long long>(D.passageOtherCapsuleChecks),
                static_cast<unsigned long long>(D.passageMouthSupportCalls),
                static_cast<unsigned long long>(D.passageMouthRouteSegmentChecks),
                static_cast<unsigned long long>(D.passageReliefFieldCalls),
                static_cast<unsigned long long>(D.passageSmoothstepEvaluations));
    }
    void RecordTimelineEvent(FStandaloneCpuTileTimelineEvent::EType Type,
        const FVoxelTileKey& Key,int32 WorkerId,size_t QueueDepth,size_t ActiveWorkers,
        size_t PriorityRank,FClock::time_point At,FClock::time_point Queued={},
        FClock::time_point Started={},FClock::time_point Finished={},double WorkerMs=-1.0,
        const VoxelForgeCore::TileGenerationStats* Diagnostics=nullptr,
        uint64 DensitySamples=0,uint32 SkirtTriangles=0)
    {
        FStandaloneCpuTileTimelineEvent Event;
        Event.Type=Type;Event.Key=Key;Event.WorkerId=WorkerId;
        Event.QueueDepth=QueueDepth;Event.ActiveWorkers=ActiveWorkers;
        Event.PriorityRank=PriorityRank;Event.At=At;Event.QueuedAt=Queued;
        Event.StartedAt=Started;Event.FinishedAt=Finished;Event.WorkerMilliseconds=WorkerMs;
        Event.DensitySamples=DensitySamples;Event.SkirtTriangles=SkirtTriangles;
        if(Diagnostics)Event.Diagnostics=*Diagnostics;
        if(WorkerId>=0&&static_cast<size_t>(WorkerId)<TimelineWorkerEvents.size())
            TimelineWorkerEvents[static_cast<size_t>(WorkerId)].push_back(Event);
        else TimelineQueueEvents.push_back(Event);
    }
    FStandaloneOwnerWorld* World=nullptr;
    inline static std::atomic<bool> bCoarseFirstSchedulingEnabled{true};
    FVector LastCameraPositionVoxels=FVector::ZeroVector;
    FVector TravelDirection=FVector::ZeroVector;
    bool bHasLastCameraPosition=false;
    int32 WorkerCount=8;
    bool bHashDensity=false;
    const FStandaloneMeshCache* MeshCache=nullptr;
    int32 MaxLevelFilter=-1;
    VoxelForgeCore::World* CoreWorld=nullptr;
    bool bSkipUvAndColors=false;
    bool bSkipMaterialGroups=false;
    bool bSkipUV=false;
    bool bSkipColors=false;
    bool bCaptureColumnCoordinates=false;
    bool bSharedColumnCache=true;
    int32 Lod8HorizonInteriorStepVoxels=0;
    bool bProfileMesherStages=false;
    bool bInstrumentTileDrain=false;
    bool bProfilePassageComposer=false;
    bool bExactMidpointDensityMemo=false;
    bool bActiveLaneBatching=false;
    int32 ProofBlockCells=8;
    bool bAuditUniformBlockProofs=false;
    std::atomic<uint64> CacheHitCount{0};
    mutable std::mutex Mutex;
    std::condition_variable Wake;
    bool bStop=true;
    std::vector<FStandaloneWorkerThread> Workers;
    std::deque<std::function<void()>> PriorityTasks;
    std::vector<FStandaloneTileJob> Queue;
    std::vector<FStandaloneCpuTile> Ready;
    std::unordered_set<FVoxelTileKey,FStandaloneTileKeyHash> Running;
    std::unordered_set<FVoxelTileKey,FStandaloneTileKeyHash> Desired;
    std::unordered_set<FVoxelTileKey,FStandaloneTileKeyHash> Outstanding;
    std::unordered_set<FVoxelTileKey,FStandaloneTileKeyHash> TileFilter;
    bool bUseTileFilter=false;
    FStandaloneDesiredTileSet CurrentDesired;
    int32 FullResTileCount=0;
    bool bColdTimelineTrace=false;
    FClock::time_point TimelineOrigin{};
    const std::atomic<uint64>* TimelineFrameOrdinal=nullptr;
    std::vector<FStandaloneCpuTileTimelineEvent> TimelineQueueEvents;
    std::vector<std::vector<FStandaloneCpuTileTimelineEvent>> TimelineWorkerEvents;
};