#pragma once

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

#include <algorithm>
#include <atomic>
#include <chrono>
#include <condition_variable>
#include <cstdio>
#include <cstdint>
#include <fstream>
#include <mutex>
#include <numeric>
#include <string>
#include <string_view>
#include <thread>
#include <vector>

#if defined(__linux__)
#include <unistd.h>
#include <sys/resource.h>
#endif

namespace StandaloneTlsProbe
{
#if defined(__linux__)
    struct FMemorySample
    {
        std::uint64_t virtualKiB=0;
        std::uint64_t residentKiB=0;
    };

    inline FMemorySample ReadMemory()
    {
        std::ifstream statm("/proc/self/statm");
        std::uint64_t virtualPages=0,residentPages=0;
        statm>>virtualPages>>residentPages;
        const long pageSize=sysconf(_SC_PAGESIZE);
        if(!statm||pageSize<=0)return {};
        return {virtualPages*static_cast<std::uint64_t>(pageSize)/1024,
            residentPages*static_cast<std::uint64_t>(pageSize)/1024};
    }

    inline int Run(int workerCount,std::string_view touchKind="none")
    {
        workerCount=std::clamp(workerCount,0,12);
        VoxelForgeCore::WorldAssemblyRecipeV1 surfaceRecipe;
        VoxelForgeCore::WorldAssemblyStratumPlanV1 surfacePlan;
        VoxelForgeCore::Heightfield2D surfaceField(surfaceRecipe,surfacePlan);
        if(touchKind=="surface-world")
        {
            VoxelForgeCore::WorldAssemblyBiomeProfileV1 biome;
            biome.stableId=1;
            surfaceRecipe.biomes.push_back(biome);
            surfacePlan.strateIndex=0;
            surfacePlan.topChunkZ=1;
            surfacePlan.bottomChunkZ=-1;
            surfaceField=VoxelForgeCore::Heightfield2D(surfaceRecipe,surfacePlan);
            if(!surfaceField.isValid())
            {
                std::fprintf(stderr,"[TlsProbe] invalid synthetic SurfaceWorld cache probe input\n");
                return 2;
            }
        }
        VoxelForgeCore::FloatingIslandRecipeV1 islandRecipe;
        VoxelForgeCore::FloatingIslandPlanV1 islandPlan;
        for(std::uint32_t index=0;index<islandRecipe.primaryIslandCount;++index)
        {
            VoxelForgeCore::FloatingIslandAnchorV1 island;
            island.stableId=index+1;
            island.ordinal=index;
            island.centerSubvoxels={static_cast<std::int64_t>(index)*64*256,0,0};
            island.radiusSubvoxels=20*256;
            island.topHalfSubvoxels=20*256;
            island.topZSubvoxels=0;
            island.bottomZSubvoxels=-80*256;
            island.taperEndQ16=islandRecipe.taperEndMinimumQ16;
            islandPlan.islands.push_back(island);
        }
        VoxelForgeCore::FloatingIslandFieldV1 islandField(islandRecipe,islandPlan);
        if(touchKind=="floating-islands"&&!islandField.isValid())
        {
            std::fprintf(stderr,"[TlsProbe] invalid synthetic FloatingIslands cache probe input\n");
            return 2;
        }
        std::mutex mutex;
        std::condition_variable changed;
        bool release=false;
        bool touchRequested=false;
        std::atomic<int> ready{0};
        std::atomic<int> touchReady{0};
        std::vector<FStandaloneWorkerThread> workers;
        std::vector<long> startupMinorFaults(static_cast<std::size_t>(workerCount),0);
        std::vector<long> firstTouchMinorFaults(static_cast<std::size_t>(workerCount),0);
        std::vector<std::int64_t> checksums(static_cast<std::size_t>(workerCount),0);
        workers.reserve(static_cast<std::size_t>(workerCount));
        FMemorySample previous=ReadMemory();
        std::printf("[TlsProbe] workers=0 create_ms=0.000 rss_kib=%llu rss_delta_kib=0 vms_kib=%llu stack_bytes=%zu\n",
            static_cast<unsigned long long>(previous.residentKiB),
            static_cast<unsigned long long>(previous.virtualKiB),
            FStandaloneWorkerThread::StackSizeBytes);
        for(int index=1;index<=workerCount;++index)
        {
            const auto start=std::chrono::steady_clock::now();
            workers.emplace_back([&,index]
            {
#if defined(__linux__)
                rusage startupUsage{};
                getrusage(RUSAGE_THREAD,&startupUsage);
                startupMinorFaults[static_cast<std::size_t>(index-1)]=startupUsage.ru_minflt;
#endif
                ready.fetch_add(1,std::memory_order_release);
                changed.notify_all();
                {
                    std::unique_lock lock(mutex);
                    changed.wait(lock,[&]{return touchRequested||release;});
                }
                if(touchRequested)
                {
#if defined(__linux__)
                    rusage beforeTouch{};
                    rusage afterTouch{};
                    getrusage(RUSAGE_THREAD,&beforeTouch);
#endif
                    std::int64_t checksum=0;
                    if(touchKind=="surface-world")
                    {
                        for(int sample=0;sample<4;++sample)
                            checksum+=surfaceField.point((index*4+sample)*256,0,0).densitySubvoxels;
                    }
                    else if(touchKind=="floating-islands")
                    {
                        for(int sample=0;sample<4;++sample)
                            checksum+=islandField.densitySubvoxels(
                                (index*4+sample)*256,0,-20*256);
                    }
                    checksums[static_cast<std::size_t>(index-1)]=checksum;
#if defined(__linux__)
                    getrusage(RUSAGE_THREAD,&afterTouch);
                    firstTouchMinorFaults[static_cast<std::size_t>(index-1)]=
                        afterTouch.ru_minflt-beforeTouch.ru_minflt;
#endif
                    touchReady.fetch_add(1,std::memory_order_release);
                    changed.notify_all();
                }
                std::unique_lock lock(mutex);
                changed.wait(lock,[&]{return release;});
            });
            {
                std::unique_lock lock(mutex);
                changed.wait(lock,[&]{return ready.load(std::memory_order_acquire)>=index;});
            }
            const double createMs=std::chrono::duration<double,std::milli>(
                std::chrono::steady_clock::now()-start).count();
            const FMemorySample current=ReadMemory();
            const std::int64_t rssDelta=static_cast<std::int64_t>(current.residentKiB)
                -static_cast<std::int64_t>(previous.residentKiB);
            std::printf("[TlsProbe] workers=%d create_ms=%.3f rss_kib=%llu rss_delta_kib=%lld vms_kib=%llu\n",
                index,createMs,static_cast<unsigned long long>(current.residentKiB),
                static_cast<long long>(rssDelta),static_cast<unsigned long long>(current.virtualKiB));
            previous=current;
        }
        const FMemorySample beforeTouch=ReadMemory();
        std::printf("[TlsProbeTouch] kind=%.*s rss_before_kib=%llu\n",
            static_cast<int>(touchKind.size()),touchKind.data(),
            static_cast<unsigned long long>(beforeTouch.residentKiB));
        if(touchKind!="none")
        {
            {
                std::lock_guard lock(mutex);
                touchRequested=true;
            }
            changed.notify_all();
            {
                std::unique_lock lock(mutex);
                changed.wait(lock,[&]{return touchReady.load(std::memory_order_acquire)==workerCount;});
            }
            const FMemorySample afterTouch=ReadMemory();
            std::printf("[TlsProbeTouch] kind=%.*s rss_after_kib=%llu rss_delta_kib=%lld checksum_sum=%lld\n",
                static_cast<int>(touchKind.size()),touchKind.data(),
                static_cast<unsigned long long>(afterTouch.residentKiB),
                static_cast<long long>(static_cast<std::int64_t>(afterTouch.residentKiB)
                    -static_cast<std::int64_t>(beforeTouch.residentKiB)),
                static_cast<long long>(std::accumulate(checksums.begin(),checksums.end(),std::int64_t{0})));
            for(int index=0;index<workerCount;++index)
                std::printf("[TlsProbeTouch] worker=%d startup_minor_faults=%ld first_touch_minor_faults=%ld\n",
                    index+1,startupMinorFaults[static_cast<std::size_t>(index)],
                    firstTouchMinorFaults[static_cast<std::size_t>(index)]);
        }
        {
            std::lock_guard lock(mutex);
            release=true;
        }
        changed.notify_all();
        for(FStandaloneWorkerThread& worker:workers)worker.join();
        std::printf("[TlsProbe] status=PASS threads_joined=%d\n",workerCount);
        return 0;
    }
#else
    inline int Run(int,std::string_view="none")
    {
        std::fprintf(stderr,"TLS probe is Linux-only.\n");
        return 2;
    }
#endif
}