acdream/src/AcDream.App/Rendering/TerrainModernRenderer.cs

using System.Numerics;
using AcDream.App.Rendering.Gpu;
using AcDream.App.Rendering.Wb;
using AcDream.Core.Terrain;

namespace AcDream.App.Rendering;

/// <summary>
/// Phase N.5b modern terrain dispatcher. Single global vertex/index arena with
/// a slot allocator (one slot per landblock, 384 verts × 40 bytes = 15,360
/// bytes per slot). Per-frame: build a DrawElementsIndirectCommand array from
/// visible slots and dispatch via one multi-draw-indirect call. Atlas
/// textures bound via the device's global texture table.
///
/// <para>Campaign V slice V11 deleted the raw-GL submission arm
/// (<c>TerrainModernRenderer.cs</c>'s former GL fields/constructor/Draw/Dispose
/// bodies); the RHI arm this file now exclusively hosts the shared allocator
/// and visibility logic for is defined in <c>TerrainModernRenderer.Rhi.cs</c>.</para>
/// </summary>
public sealed partial class TerrainModernRenderer : IDisposable
{
    // VertsPerLandblock MUST stay divisible by 6 — terrain_modern.vert uses
    // `gl_VertexID % 6` to pick the cell-corner index (BL/BR/TR/TL), and
    // because we bake `slot * VertsPerLandblock` into indices CPU-side and
    // pass BaseVertex=0 to MultiDrawElementsIndirect, gl_VertexID becomes
    // `slot * VertsPerLandblock + local_index`. The shader's modulo-6 only
    // reduces to `local_index % 6` because 384 is a multiple of 6. Changing
    // either constant without auditing the shader will silently mis-render.
    private const int VertsPerLandblock = LandblockMesh.VerticesPerLandblock;  // 384 (= 64 cells * 6 verts)
    private const int IndicesPerLandblock = VertsPerLandblock;
    private const int VertexSize = 40;  // sizeof(TerrainVertex)
    private const int IndexSize = sizeof(uint);
    private const float LandblockSize = LandblockMesh.LandblockSize;  // 192

    private readonly TerrainAtlas _atlas;

    /// <summary>A.5 T22.5: exposes the terrain atlas so callers can update
    /// anisotropic level mid-session via <see cref="TerrainAtlas.SetAnisotropic"/>.</summary>
    public TerrainAtlas Atlas => _atlas;

    private readonly GpuRetiredTerrainSlotAllocator _alloc;
    private readonly GpuRetirementLedger _retirementLedger;
    private bool _disposed;

    // Per-slot live data (index by slot integer; null entries are unused slots).
    private SlotData?[] _slots;

    // Reverse map: landblockId -> slot, for RemoveLandblock and replacement.
    private readonly Dictionary<uint, int> _idToSlot = new();

    // Backing-store capacity bookkeeping, shared with the RHI arm's arena.
    private long _globalVboCapacityBytes;
    private long _globalEboCapacityBytes;

    // Per-GPU-fenced-frame-slot draw bookkeeping, shared with the RHI arm.
    private int _dynamicFrameSlot;
    private bool _dynamicFrameStarted;

    /// <summary>
    /// The dynamic per-frame-slot indirect-command buffer pool this used to
    /// report was raw-GL-only bookkeeping, deleted with that arm at Campaign V
    /// slice V11. The RHI arm allocates its indirect-command storage from the
    /// GPU frame's own upload ring instead, so there is no separate pool to
    /// count.
    /// </summary>
    internal int DynamicIndirectBufferCount => 0;

    // Reusable per-frame buffers.
    private readonly List<int> _visibleSlots = new();
    private readonly HashSet<uint> _walkVisibleLandblocks = new();
    private DrawElementsIndirectCommand[] _deicScratch = Array.Empty<DrawElementsIndirectCommand>();

    // S3 chunk 3 (§9.2 B1): DrawLandCells' reusable per-entry index-run
    // scratch — repopulated per cell by AppendCellIndexRuns, one
    // (Start, Count) pair per contiguous run relative to THAT entry's own
    // slot (side 8: one run/cell; side 1: one run total). Converted to
    // absolute (FirstIndex, Count) pairs in _batchRunScratch below because a
    // fix-round-1 batch (F2) can span several landblocks/slots.
    private readonly List<(int Start, int Count)> _cellRunScratch = new();

    // S3 chunk 3 fix round 1 (F2): DrawLandCells' reusable ABSOLUTE run
    // scratch — one (FirstIndex, Count) pair per contiguous index run across
    // every entry of one batch, in entry order, cleared per call.
    private readonly List<(uint FirstIndex, int Count)> _batchRunScratch = new();

    // S3 chunk 3 fix round 1 (F3): the walk path's own per-FRAME submission
    // stats — cleared in BeginFrame, populated by DrawLandCells. Distinct
    // from _visibleSlots (Draw()'s own per-call list, the non-walk fallback
    // path) because the walk submits many small batches across a frame
    // rather than one Draw() call; TerrainDrawDiagnosticsController reports
    // whichever path actually ran this publication window.
    private readonly HashSet<int> _walkSlotsThisFrame = new();
    private int _walkDrawsThisFrame;

    // Diag.
    public int LoadedSlots   => _alloc.LoadedCount;
    public int VisibleSlots  => _visibleSlots.Count;
    public int CapacitySlots => _alloc.Capacity;

    /// <summary>S3 chunk 3 fix round 1 (F3): distinct slots <see
    /// cref="DrawLandCells"/> actually submitted THIS FRAME (cleared at
    /// <see cref="BeginFrame"/>) — the walk path's own per-frame
    /// "VisibleSlots" answer, since the walk never calls <see cref="Draw"/>.</summary>
    internal int WalkVisibleSlotCount => _walkSlotsThisFrame.Count;

    /// <summary>S3 chunk 3 fix round 1 (F3): the number of <see
    /// cref="DrawLandCells"/> batches (indirect draws) submitted THIS FRAME.</summary>
    internal int WalkDrawCount => _walkDrawsThisFrame;

    /// <summary>
    /// Resets the per-GPU-fenced-frame-slot draw state. A retail outside view
    /// may draw terrain more than once in a frame.
    /// </summary>
    public void BeginFrame(int frameSlot)
    {
        ArgumentOutOfRangeException.ThrowIfNegative(frameSlot);
        if (_directionalShadowFrameSequence == long.MaxValue)
            throw new InvalidOperationException(
                "Directional-shadow terrain frame identity was exhausted.");
        _directionalShadowFrameSequence++;
        _retirementLedger.RetryPendingPublications();
        _dynamicFrameSlot = frameSlot;
        _dynamicFrameStarted = true;
        // S3 chunk 3 fix round 1 (F3): reset the walk path's per-frame
        // submission stats — one presented frame, one answer.
        _walkSlotsThisFrame.Clear();
        _walkDrawsThisFrame = 0;
    }

    /// <summary>
    /// Two-tier streaming entry point. Accepts a prebuilt mesh from
    /// <see cref="LandblockStreamResult.Loaded.MeshData"/> built on the worker
    /// thread, together with the world-space origin computed by the caller
    /// (render-thread GameWindow derives it from landblockId + liveCenterX/Y).
    ///
    /// Delegates to <see cref="AddLandblock(uint,LandblockMeshData,Vector3)"/>
    /// so both paths share one upload path. Per Phase A.5 spec T15.
    /// </summary>
    public void AddLandblockWithMesh(uint landblockId, LandblockMeshData meshData, Vector3 worldOrigin)
        => AddLandblock(landblockId, meshData, worldOrigin);

    public void AddLandblock(uint landblockId, LandblockMeshData meshData, Vector3 worldOrigin)
    {
        ArgumentNullException.ThrowIfNull(meshData);
        if (meshData.Vertices.Length != VertsPerLandblock)
            throw new ArgumentException(
                $"Expected {VertsPerLandblock} vertices, got {meshData.Vertices.Length}",
                nameof(meshData));
        if (meshData.Indices.Length != IndicesPerLandblock)
            throw new ArgumentException(
                $"Expected {IndicesPerLandblock} indices, got {meshData.Indices.Length}",
                nameof(meshData));

        // A prior replacement may have committed the logical slot switch
        // before queue publication failed. Retry those retained physical-slot
        // transactions before allocating more terrain storage.
        _alloc.RetryPendingPublications();

        bool replacing = _idToSlot.TryGetValue(landblockId, out int replacedSlot);
        int slot = _alloc.Allocate(out var needsGrow);
        bool published = false;
        try
        {
            if (needsGrow)
            {
                int newCap = Math.Max(_alloc.Capacity * 2, slot + 1);
                EnsureCapacity(newCap);
            }

            // Bake worldOrigin into vertex positions; capture min/max Z for AABB.
            var bakedVerts = new TerrainVertex[VertsPerLandblock];
            float zMin = float.MaxValue, zMax = float.MinValue;
            for (int i = 0; i < VertsPerLandblock; i++)
            {
                var v = meshData.Vertices[i];
                var worldPos = v.Position + worldOrigin;
                bakedVerts[i] = new TerrainVertex(worldPos, v.Normal, v.Data0, v.Data1, v.Data2, v.Data3);
                if (worldPos.Z < zMin) zMin = worldPos.Z;
                if (worldPos.Z > zMax) zMax = worldPos.Z;
            }
            if (zMin == float.MaxValue) { zMin = 0f; zMax = 0f; }

            // Bake baseVertex into indices on the CPU side (driver-portable pattern).
            uint baseVertex = (uint)(slot * VertsPerLandblock);
            var bakedIndices = new uint[IndicesPerLandblock];
            for (int i = 0; i < IndicesPerLandblock; i++)
                bakedIndices[i] = meshData.Indices[i] + baseVertex;

            UploadRhiLandblock(slot, bakedVerts, bakedIndices);

            _slots[slot] = new SlotData
            {
                LandblockId = landblockId,
                WorldOrigin = worldOrigin,
                FirstIndex  = (uint)(slot * IndicesPerLandblock),
                IndexCount  = IndicesPerLandblock,
                AabbMin = new Vector3(worldOrigin.X, worldOrigin.Y, zMin),
                AabbMax = new Vector3(worldOrigin.X + LandblockSize, worldOrigin.Y + LandblockSize, zMax),
            };
            _idToSlot[landblockId] = slot;
            published = true;

            if (replacing)
            {
                _slots[replacedSlot] = null;
                _alloc.FreeAfterGpuUse(replacedSlot);
            }
        }
        finally
        {
            if (!published)
                _alloc.ReleaseUnsubmitted(slot);
        }
    }

    public void RemoveLandblock(uint landblockId)
    {
        // Removal clears the logical lookup before retirement publication. A
        // retry therefore has to advance retained publications even when the
        // landblock is no longer present in the map.
        _alloc.RetryPendingPublications();
        if (!_idToSlot.TryGetValue(landblockId, out var slot))
            return;
        _idToSlot.Remove(landblockId);
        _slots[slot] = null;
        _alloc.FreeAfterGpuUse(slot);
        // No GPU clear: the per-frame DEIC array won't reference this slot.
    }

    public void Draw(
        ICamera camera,
        FrustumPlanes? frustum = null,
        uint? neverCullLandblockId = null,
        IReadOnlySet<uint>? inViewLandcells = null)
    {
        if (_alloc.LoadedCount == 0) return;

        Matrix4x4 viewProjection = camera.View * camera.Projection;

        // Campaign FW4: RetailFrameWalk is the sole visibility authority.
        // Terrain remains one full-landblock MDI draw, but only landblocks
        // containing a walk-admitted landcell participate. The walk publishes
        // its exact in_view set directly to typed consumers; this draw path
        // does not reconstruct or retain a second visibility answer.
        _walkVisibleLandblocks.Clear();
        if (inViewLandcells is not null)
        {
            foreach (uint cellId in inViewLandcells)
            {
                _walkVisibleLandblocks.Add(cellId & 0xFFFF0000u);
            }
        }

        // Build visible slot list with per-slot frustum cull.
        _visibleSlots.Clear();
        for (int slot = 0; slot < _slots.Length; slot++)
        {
            var data = _slots[slot];
            if (data is null) continue;
            if (inViewLandcells is not null
                && !_walkVisibleLandblocks.Contains(data.LandblockId & 0xFFFF0000u))
            {
                continue;
            }
            if (frustum is not null && data.LandblockId != neverCullLandblockId)
            {
                if (!FrustumCuller.IsAabbVisible(frustum.Value, data.AabbMin, data.AabbMax))
                    continue;
            }
            _visibleSlots.Add(slot);
        }
        if (_visibleSlots.Count == 0) return;

        BuildIndirectCommands();
        if (!_dynamicFrameStarted)
            throw new InvalidOperationException("BeginFrame must be called before drawing terrain.");
        DrawRhi(viewProjection, _visibleSlots.Count);
    }

    /// <summary>
    /// Builds this frame's <c>DrawElementsIndirectCommand</c> array from the
    /// visible slot list. Pure CPU.
    /// </summary>
    private void BuildIndirectCommands()
    {
        if (_deicScratch.Length < _visibleSlots.Count)
            _deicScratch = new DrawElementsIndirectCommand[Math.Max(_visibleSlots.Count, 64)];
        for (int i = 0; i < _visibleSlots.Count; i++)
        {
            var data = _slots[_visibleSlots[i]]!;
            _deicScratch[i] = new DrawElementsIndirectCommand
            {
                Count         = (uint)data.IndexCount,
                InstanceCount = 1u,
                FirstIndex    = data.FirstIndex,
                BaseVertex    = 0,            // baked into indices on upload
                BaseInstance  = 0,
            };
        }
    }

    /// <summary>
    /// S3 chunk 3 (§9.2 B1/B2), fix round 1 (F1/F2): the walk's per-land-cell
    /// terrain draw — retail <c>RenderDeviceD3D::DrawLandCell</c>
    /// @0x0059f120, batched. Submits <see cref="AppendCellIndexRuns"/>'s
    /// index runs for every entry of <paramref name="cells"/> as ONE
    /// indirect draw spanning however many DISTINCT landblocks the batch
    /// covers (fix round 1's F2: the terrain index buffer is one buffer with
    /// a per-slot <c>FirstIndex</c> offset, so a batch is never required to
    /// stay within one landblock — <see cref="Walk.WalkFrameDriver.Replay"/>
    /// only splits it at a genuine intervening GPU submission), UNCLIPPED:
    /// retail never view-clips terrain — the walk's own per-cell
    /// <c>CellInView</c> admission (<c>RetailFrameWalk.DrawLandscape</c>'s
    /// <c>DrawLandCell</c> gate) already decided which cells reach here, so
    /// there is no frustum test and no <c>inViewLandcells</c> filter here.
    /// Fix round 1 F6: this is not a missing culling step — retail has no
    /// separate terrain frustum test beyond <c>LScape::draw_check_blocks</c>/
    /// <c>landcell_check</c> (ported as <c>WalkLandscape.CheckBlocks</c>),
    /// so the walk's own admission IS the sole culling authority for
    /// terrain, matching retail exactly. Unlike the non-walk <see cref="Draw"/> entry point this supersedes
    /// for the walk path only (<see cref="Draw"/> keeps serving its one
    /// remaining non-walk caller — fix round 1 F6 corrects the prior "flat/
    /// directional-shadow paths" wording: the only caller left is the flat-
    /// terrain fallback, <c>WorldScenePassExecutor.DrawFlatTerrain</c>; a
    /// directional-shadow-receiving surface selects its receiver pipeline
    /// inside the SAME <c>DrawRhi</c> call this makes, not through a second
    /// caller).
    ///
    /// <para><b>F1 — the slot key.</b> <paramref name="cells"/> carries each
    /// entry's landblock id in the WALK's own <c>bx&lt;&lt;24 | by&lt;&lt;16</c>
    /// encoding (<c>WalkLandBlock.LandblockId</c>, low word 0x0000), while
    /// <see cref="_idToSlot"/> is keyed by the DAT landblock id
    /// (<c>LandblockRenderPublisher.LandblockId</c>, low word 0xFFFF — the
    /// same convention <see cref="AddLandblock"/>/<see cref="RemoveLandblock"/>
    /// already use and must keep using, since those callers already pass the
    /// DAT id). Normalized HERE, at this one walk entry point, rather than at
    /// every walk call site.</para>
    ///
    /// <para>An entry whose normalized id has no uploaded slot (a
    /// streaming-timing race between the walk's own block graph and this
    /// renderer's landblock upload, OR a genuinely unknown block) is a
    /// silent per-entry no-op — the walk's visited-cell bookkeeping is the
    /// timing authority, not this call; the other entries in the same batch
    /// still submit.</para>
    /// </summary>
    public void DrawLandCells(
        Matrix4x4 viewProjection,
        IReadOnlyList<(uint LandblockId, int SideCellCount, int CellIndex)> cells)
    {
        ArgumentNullException.ThrowIfNull(cells);
        if (cells.Count == 0) return;
        if (!_dynamicFrameStarted)
            throw new InvalidOperationException("BeginFrame must be called before drawing terrain.");

        _batchRunScratch.Clear();
        for (int i = 0; i < cells.Count; i++)
        {
            (uint landblockId, int sideCellCount, int cellIndex) = cells[i];
            // F1: the walk hands 0xXXYY0000 (WalkLandBlock.LandblockId);
            // AddLandblock stores under the DAT id 0xXXYYFFFF
            // (LandblockRenderPublisher.LandblockId). Normalize the lookup,
            // not the storage — every non-walk caller still stores/removes
            // under the DAT id unchanged.
            uint slotKey = (landblockId & 0xFFFF0000u) | 0xFFFFu;
            if (!_idToSlot.TryGetValue(slotKey, out int slot))
                continue;
            uint baseFirstIndex = (uint)(slot * IndicesPerLandblock);

            _cellRunScratch.Clear();
            AppendCellIndexRuns(sideCellCount, cellIndex, _cellRunScratch);
            for (int r = 0; r < _cellRunScratch.Count; r++)
            {
                (int start, int count) = _cellRunScratch[r];
                _batchRunScratch.Add((baseFirstIndex + (uint)start, count));
            }
            _walkSlotsThisFrame.Add(slot);
        }

        int commandCount = _batchRunScratch.Count;
        if (commandCount == 0) return;

        if (_deicScratch.Length < commandCount)
            _deicScratch = new DrawElementsIndirectCommand[Math.Max(commandCount, 64)];
        for (int i = 0; i < commandCount; i++)
        {
            (uint firstIndex, int count) = _batchRunScratch[i];
            _deicScratch[i] = new DrawElementsIndirectCommand
            {
                Count         = (uint)count,
                InstanceCount = 1u,
                FirstIndex    = firstIndex,
                BaseVertex    = 0,            // baked into indices on upload
                BaseInstance  = 0,
            };
        }
        _walkDrawsThisFrame++;
        DrawRhi(viewProjection, commandCount);
    }

    /// <summary>
    /// S3 chunk 3 (§9.1 R3): retail LOD cell (side <paramref
    /// name="sideCellCount"/>, LOD coords <c>(X, Y) = (cellIndex / side,
    /// cellIndex % side)</c>) covers <c>cx ∈ [X·8/n, (X+1)·8/n)</c>,
    /// <c>cy ∈ [Y·8/n, (Y+1)·8/n)</c> of the 8×8 cell-major mesh index
    /// buffer (<c>LandblockMesh.Build</c>: <c>cy</c> outer, <c>cx</c>
    /// inner, 6 indices per cell, <c>indices[i] = i</c>). Appends one
    /// contiguous run per covered <c>cy</c> row — side 8 emits one run of
    /// 6 indices per cell, side 4 two runs of 12, side 2 four runs of 24;
    /// side 1's single coarse cell covers EVERY row contiguously (the
    /// mesh's rows sit back-to-back with no gap), so its 8 per-row runs
    /// concatenate into ONE run of all 384 indices. Pure arithmetic — no
    /// baked range table, no per-frame rebuild.
    /// </summary>
    internal static void AppendCellIndexRuns(
        int sideCellCount, int cellIndex, List<(int Start, int Count)> runs)
    {
        ArgumentNullException.ThrowIfNull(runs);
        if (sideCellCount is not (1 or 2 or 4 or 8))
        {
            throw new ArgumentOutOfRangeException(
                nameof(sideCellCount),
                sideCellCount,
                "A landscape LOD grid must be 1, 2, 4, or 8 cells per side.");
        }
        if ((uint)cellIndex >= (uint)(sideCellCount * sideCellCount))
            throw new ArgumentOutOfRangeException(nameof(cellIndex));

        int span = LandblockMesh.CellsPerSide / sideCellCount;
        int coarseX = cellIndex / sideCellCount;
        int coarseY = cellIndex % sideCellCount;
        int firstCx = coarseX * span;
        int firstCy = coarseY * span;
        if (span == LandblockMesh.CellsPerSide)
        {
            // The block's only coarse cell (side 1) covers cx AND cy 0..7 —
            // every row's run then abuts the next (row cy's run ends at
            // (cy*8+8)*6 = (cy+1)*8*6, exactly row cy+1's start), so the 8
            // per-row runs are one contiguous 384-index range.
            runs.Add((0, VertsPerLandblock));
            return;
        }
        int runLength = span * LandblockMesh.VerticesPerCell;
        for (int cy = firstCy; cy < firstCy + span; cy++)
        {
            int start = (cy * LandblockMesh.CellsPerSide + firstCx) * LandblockMesh.VerticesPerCell;
            runs.Add((start, runLength));
        }
    }

    public void Dispose()
    {
        if (_disposed)
            return;
        _retirementLedger.RetryPendingPublications();
        DisposeRhi();
    }

    // ----------------------------------------------------------------
    // Private helpers
    // ----------------------------------------------------------------

    private void EnsureCapacity(int newCapacity)
    {
        if (newCapacity <= _alloc.Capacity)
            return;
        EnsureRhiCapacity(newCapacity);
    }

    private sealed class SlotData
    {
        public uint    LandblockId;
        public Vector3 WorldOrigin;
        public uint    FirstIndex;
        public int     IndexCount;
        public Vector3 AabbMin;
        public Vector3 AabbMax;
    }
}