ModernUO/Projects/UOContent/Engines/Pathing/Cache/StepCache.cs
Kamron Batman 7c9215d97c
feat(pathfinding): lazy .swb backing store for the step cache (#2448)
## Summary

Adds a binary disk format + lazy reader so the step cache can warm-start from a precomputed file without paying chunk-build cost on the first pathfind through a region. **Resident memory stays bounded by `MaxResidentChunks` regardless of file size** — opening a `.swb` reads only the header + chunk-offset index (~16 bytes per indexed chunk), and individual chunks are seeked + deserialized only when `ResolveMissingChunk` asks for them.

The lazy design (vs. an eager bulk load): a 250 MB bake on a RAM-constrained shard never materializes more than the LRU cap (~40 MB at the default 8192-chunk cap), and unwanted regions never enter memory at all.

Builds on PR #2447.

## What changed

- **`StepCacheFile`** — binary reader/writer module. Writer emits header → chunks (offsets recorded) → index trailer, then patches the header's `IndexOffset` field. Reader is `OpenForLazy(path)` returning a `LazyReader` that holds an open `FileStream` + offset dictionary.
- **`StepCacheFile.LazyReader`** — `TryReadChunk(chunkX, chunkY)` does a single seek + bulk read for one record. `Dispose` releases the underlying stream. Files are opened with `FileShare.Read | FileShare.Delete` so admin tooling can replace them.
- **TileData fingerprint via XxHash3.** The `.swb` header carries a hash of `LandTable + ItemTable` flags. Load rejects any file whose hash doesn't match the running server. Computed via `HashUtility.ComputeHash64` (engine-blessed hasher) — adds a `ReadOnlySpan<byte>` overload alongside the existing `ReadOnlySpan<char>` one for parity.
- **`StepCache.SaveToFile(path, mapId)`** — writes resident chunks for the given map.
- **`StepCache.TryOpenLazyReader(path, mapId)`** — opens the file, validates header, holds the reader for the map's lifetime.
- **`StepCache.ResolveMissingChunk`** — now consults the lazy reader before invoking the runtime baker. A loaded chunk whose `BuiltMultisVersion` doesn't match the live sector falls through to the baker (snapshot was made before a multi was added/removed in that sector).
- **`StepCache.Clear` closes lazy readers.** Test cleanup can delete `.swb` files cleanly.
- **Auto-load at startup.** `PathCacheCommands.Configure()` opens `Data/Pathfinding/<mapId>.swb` as a lazy reader for every map.
- **`[PathCacheSave`** / **`[PathCacheLoad`** — admin commands for the same workflow.
- **`pathfinding.maxResidentChunks` shard-tunable.** Read from `server.cfg` at boot via `ServerConfiguration.GetOrUpdateSetting` (default 8192 ≈ 40 MB). Small shards can tune down; large shards with substantial bakes can tune up to reduce eviction churn. Default is written back to `server.cfg` on first boot, matching the engine pattern used by other settings.

## File layout (v1)

```
Header (48 bytes):
  u32  Magic           = 0x42575300 ('SWB\0')
  u32  Version         = 1
  u32  MapId
  u64  TileDataHash    XxHash3 over LandTable + ItemTable flags (HashUtility)
  u64  BakeTimestamp   informational
  u32  ChunkCount
  u64  IndexOffset     file position where the chunk index begins

Chunk records (fixed size, ~5,393 bytes each, +32 if multi-Z):
  u16  ChunkX
  u16  ChunkY
  u32  BuiltMultisVersion
  u8   HasMultiZ
  byte WalkMask[256], WetMask[256]
  sbyte SourceZ[256], WalkZN..WalkZNW[256], SwimZN..SwimZNW[256]
  [byte MultiZCells[32] when HasMultiZ == 1]

Index trailer (16 × ChunkCount bytes):
  (u64 chunkKey, u64 fileOffset)
```

## Memory math

| Scenario | Disk file | RAM at boot | Notes |
|---|---|---|---|
| Empty / no `.swb` files | — | 0 | Silent; cache builds on demand. |
| Admin-curated towns (5K chunks) | 25 MB | 0 + per-query | Index ≈ 80 KB. Resident grows to the configured cap under steady-state queries. |
| Full-map bake (50K chunks) | 250 MB | 0 + per-query | Index ≈ 800 KB. Same configured cap. Cold areas never load. |
| All 5 maps fully baked | 1.25 GB | 0 + per-query | Index ≈ 4 MB total. Same configured cap. |

## Hash choice (FNV-1a → XxHash3)

The original draft used inlined FNV-1a-64. Switched to XxHash3 via `HashUtility`:

- ~30× faster on this workload (~30 GB/s SIMD vs ~2 GB/s byte-by-byte). Boot-time only, so absolute saving is microseconds — the real wins are elsewhere.
- Stronger collision resistance and distribution.
- Drops ~25 lines of inlined hash code; matches the rest of the codebase's hashing pattern.
- Hash is stable as long as `HashUtility`'s `xxHash3Seed` constant doesn't change (already marked `// DO NOT CHANGE THIS NUMBER`).
2026-05-06 01:32:44 -07:00

474 lines
18 KiB
C#

using System;
using System.Collections.Generic;
using Server.Logging;
namespace Server.Engines.Pathing.Cache;
/// <summary>
/// Singleton store of per-chunk static walkability data. Chunks correspond to
/// Map.SectorSize = 16; key encoding packs (mapId, chunkX, chunkY) into a long.
/// Lazily built on first query; invalidated by version-check vs Sector.MultisVersion;
/// memory bounded by MaxResidentChunks via probabilistic LRU eviction.
///
/// Default-walker scope only. Cells with multi-Z surfaces and queries for non-default
/// walkers route to the MovementImpl slow path via the Fallthrough_* hit kinds.
/// </summary>
public sealed class StepCache
{
private static readonly ILogger logger = LogFactory.GetLogger(typeof(StepCache));
public static StepCache Instance { get; } = new();
private readonly Dictionary<long, StepChunk> _chunks = new();
// Parallel list of keys for O(1) random sampling during eviction. Kept in lockstep
// with _chunks: append on Miss_NotBuilt, swap-and-pop on eviction.
private readonly List<long> _keysList = new();
// Telemetry counters
private long _hits;
private long _missesNotBuilt;
private long _missesDirtyRebuild;
private long _fallthroughMultiZ;
private long _fallthroughOffMap;
private long _fallthroughSourceZMismatch;
private long _evictionsByLruCap;
private long _buildsTotal;
private StepCache() { }
/// <summary>Hard cap on resident chunk count. Default 8192. Override for tests / ops.</summary>
public int MaxResidentChunks { get; set; } = 8192;
/// <summary>
/// Pack (mapId, chunkX, chunkY) into a single long key.
/// Layout: [reserved 16][mapId 16][chunkX 16][chunkY 16].
/// </summary>
internal static long EncodeKey(int mapId, int chunkX, int chunkY) =>
((long)(mapId & 0xFFFF) << 32) | ((long)(chunkX & 0xFFFF) << 16) | (long)(chunkY & 0xFFFF);
public CacheStats GetStats() => new CacheStats(
residentChunks: _chunks.Count,
hits: _hits,
missesNotBuilt: _missesNotBuilt,
missesDirtyRebuild: _missesDirtyRebuild,
fallthroughMultiZ: _fallthroughMultiZ,
fallthroughOffMap: _fallthroughOffMap,
fallthroughSourceZMismatch: _fallthroughSourceZMismatch,
evictionsByLruCap: _evictionsByLruCap,
buildsTotal: _buildsTotal
);
/// <summary>
/// Drop all cached chunks AND zero every telemetry counter. Used by tests and
/// benchmarks that need a known cold-start state. Counter reset is intentional —
/// counters are since-last-clear, not since-startup.
/// </summary>
public void Clear()
{
_chunks.Clear();
_keysList.Clear();
CloseLazyReaders();
_hits = 0;
_missesNotBuilt = 0;
_missesDirtyRebuild = 0;
_fallthroughMultiZ = 0;
_fallthroughOffMap = 0;
_fallthroughSourceZMismatch = 0;
_evictionsByLruCap = 0;
_buildsTotal = 0;
}
// Per-map open .swb readers, populated by TryOpenLazyReader at startup. Chunks are
// fetched on demand from the file when ResolveMissingChunk fires; resident memory
// stays bounded by MaxResidentChunks regardless of file size.
private readonly Dictionary<int, StepCacheFile.LazyReader> _lazyReaders = new();
/// <summary>
/// Persist all resident chunks for <paramref name="mapId"/> to a .swb file. Returns
/// the number of chunks written. The file embeds a TileData fingerprint so a stale
/// file (built before a client patch) can be detected and rejected at open time.
/// </summary>
public int SaveToFile(string path, int mapId)
{
var matching = 0;
foreach (var key in _keysList)
{
DecodeKey(key, out var keyMapId, out _, out _);
if (keyMapId == mapId)
{
matching++;
}
}
var enumerator = _keysList.GetEnumerator();
StepCacheFile.Write(path, (uint)mapId, (uint)matching, EmitChunk);
enumerator.Dispose();
return matching;
bool EmitChunk(out int chunkX, out int chunkY, out StepChunk chunk)
{
while (enumerator.MoveNext())
{
var key = enumerator.Current;
DecodeKey(key, out var emittedMapId, out chunkX, out chunkY);
if (emittedMapId == mapId)
{
chunk = _chunks[key];
return true;
}
}
chunkX = chunkY = 0;
chunk = null!;
return false;
}
}
/// <summary>
/// Open a .swb file as a lazy backing store for <paramref name="mapId"/>. Reads only
/// header + chunk-offset index (~16 bytes per chunk); individual records are fetched
/// on demand by <see cref="ResolveMissingChunk"/>. Returns false on missing file,
/// magic / version mismatch, or TileData hash mismatch (stale bake).
/// </summary>
public bool TryOpenLazyReader(string path, int mapId)
{
var reader = StepCacheFile.OpenForLazy(path);
if (reader == null)
{
return false;
}
if (reader.MapId != (uint)mapId)
{
logger.Warning(
"StepCache: {Path} declares mapId {FileMapId} but caller requested {RequestedMapId}; ignoring",
path, reader.MapId, mapId
);
reader.Dispose();
return false;
}
if (_lazyReaders.TryGetValue(mapId, out var existing))
{
existing.Dispose();
}
_lazyReaders[mapId] = reader;
logger.Information(
"StepCache: opened {Path} ({ChunkCount} chunks indexed) for map {MapId}",
path, reader.IndexedChunkCount, mapId
);
return true;
}
/// <summary>
/// Number of .swb readers currently open. Mostly for tests / telemetry.
/// </summary>
public int OpenLazyReaderCount => _lazyReaders.Count;
/// <summary>Test-only diagnostic: does the lazy reader for <paramref name="mapId"/> hold an offset for (chunkX, chunkY)?</summary>
internal bool LazyReaderHasChunk(int mapId, int chunkX, int chunkY) =>
_lazyReaders.TryGetValue(mapId, out var r) && r.Has(chunkX, chunkY);
/// <summary>
/// Closes all open lazy readers, releasing their underlying file streams. Called from
/// <see cref="Clear"/> so test cleanup can delete .swb files (they're held with
/// FileShare.Read | FileShare.Delete, so this is mostly belt-and-suspenders).
/// </summary>
public void CloseLazyReaders()
{
foreach (var reader in _lazyReaders.Values)
{
reader.Dispose();
}
_lazyReaders.Clear();
}
/// <summary>
/// Probabilistic LRU sample size — picks SampleSize random resident chunks per
/// eviction and evicts the oldest of that sample. Approximates true LRU at a tiny
/// fraction of the cost (no full sort). Redis uses the same approach (`maxmemory-samples`).
/// 5 yields ~quality-of-true-LRU for cache eviction; higher values trade speed for accuracy.
/// </summary>
private const int LruSampleSize = 5;
/// <summary>
/// If resident chunk count exceeds MaxResidentChunks, evict via probabilistic LRU
/// until the count is at or below the cap. Per-eviction cost is O(LruSampleSize),
/// independent of resident count — sustained cap pressure has no perpetual perf hit.
/// Called from CacheEvictionTimer; also callable directly from tests.
/// </summary>
public void EnforceLruCap()
{
var overflow = _chunks.Count - MaxResidentChunks;
if (overflow <= 0)
{
return;
}
while (overflow-- > 0 && _keysList.Count > 0)
{
var oldestIdx = -1;
long oldestTouched = long.MaxValue;
long oldestKey = 0;
// Sample LruSampleSize random keys; track the oldest by LastTouchedTicks.
// With replacement is fine — collisions are rare and don't break correctness.
var samples = Math.Min(LruSampleSize, _keysList.Count);
for (var s = 0; s < samples; s++)
{
var idx = Utility.Random(_keysList.Count);
var k = _keysList[idx];
var touched = _chunks[k].LastTouchedTicks;
if (touched < oldestTouched)
{
oldestTouched = touched;
oldestKey = k;
oldestIdx = idx;
}
}
_chunks.Remove(oldestKey);
// Swap-and-pop _keysList[oldestIdx] with the tail; O(1) regardless of position.
var last = _keysList.Count - 1;
if (oldestIdx != last)
{
_keysList[oldestIdx] = _keysList[last];
}
_keysList.RemoveAt(last);
_evictionsByLruCap++;
}
}
internal static void DecodeKey(long key, out int mapId, out int chunkX, out int chunkY)
{
mapId = (int)((key >> 32) & 0xFFFF);
chunkX = (int)((key >> 16) & 0xFFFF);
chunkY = (int)(key & 0xFFFF);
}
private const int ChunkSize = 16;
/// <summary>
/// Hot-path query. Returns the cached mask + 8 destination Z values + hit kind.
/// Inspect <see cref="StepMask.IsHit"/> to decide whether to use the result or fall
/// back to the slow path.
/// </summary>
public StepMask TryGetMask(Map map, int x, int y, sbyte sourceZ)
{
if (map == null || map == Map.Internal || x < 0 || y < 0 || x >= map.Width || y >= map.Height)
{
_fallthroughOffMap++;
return new StepMask(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, CacheHitKind.Fallthrough_OffMap);
}
var chunkX = x >> 4;
var chunkY = y >> 4;
var key = EncodeKey(map.MapID, chunkX, chunkY);
var hitKindResult = CacheHitKind.Hit;
if (!_chunks.TryGetValue(key, out var chunk))
{
chunk = ResolveMissingChunk(map, chunkX, chunkY);
_chunks[key] = chunk;
_keysList.Add(key);
hitKindResult = CacheHitKind.Miss_NotBuilt;
}
else
{
var sector = map.GetRealSector(chunkX, chunkY);
if (chunk.BuiltMultisVersion != sector.MultisVersion)
{
chunk = BuildChunk(map, chunkX, chunkY);
_chunks[key] = chunk;
hitKindResult = CacheHitKind.Miss_DirtyRebuild;
// _missesDirtyRebuild++ deferred to the outcome switch below so a
// multi-Z fallthrough on a freshly dirty-rebuilt chunk doesn't double-count.
}
}
chunk.LastTouchedTicks = Core.TickCount;
var cellIndex = ((y - (chunkY << 4)) << 4) | (x - (chunkX << 4));
if (chunk.IsCellMultiZ(cellIndex))
{
_fallthroughMultiZ++;
return new StepMask(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, CacheHitKind.Fallthrough_MultiZ);
}
// Source-Z guard: the cache stores one answer per cell baked at SourceZ.
// StepHeight tolerance accepts incremental Z jitter; loosening it breaks parity
// because tile reachability shifts at step-height boundaries.
if (Math.Abs(sourceZ - chunk.SourceZ[cellIndex]) > StepHeight)
{
_fallthroughSourceZMismatch++;
return new StepMask(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, CacheHitKind.Fallthrough_SourceZMismatch);
}
switch (hitKindResult)
{
case CacheHitKind.Miss_NotBuilt: { _missesNotBuilt++; break; }
case CacheHitKind.Miss_DirtyRebuild: { _missesDirtyRebuild++; break; }
case CacheHitKind.Hit: { _hits++; break; }
}
return new StepMask(
chunk.WalkMask[cellIndex],
chunk.WetMask[cellIndex],
chunk.WalkZN[cellIndex],
chunk.WalkZNE[cellIndex],
chunk.WalkZE[cellIndex],
chunk.WalkZSE[cellIndex],
chunk.WalkZS[cellIndex],
chunk.WalkZSW[cellIndex],
chunk.WalkZW[cellIndex],
chunk.WalkZNW[cellIndex],
chunk.SwimZN[cellIndex],
chunk.SwimZNE[cellIndex],
chunk.SwimZE[cellIndex],
chunk.SwimZSE[cellIndex],
chunk.SwimZS[cellIndex],
chunk.SwimZSW[cellIndex],
chunk.SwimZW[cellIndex],
chunk.SwimZNW[cellIndex],
hitKindResult
);
}
/// <summary>
/// Chunk-miss resolution: try the lazy file reader for this map first; if there's no
/// file or no record at this (chunkX, chunkY), fall back to the runtime baker. The
/// file path validates each loaded chunk's MultisVersion against the live sector — a
/// stale snapshot triggers a rebuild rather than serving a wrong answer.
/// </summary>
private StepChunk ResolveMissingChunk(Map map, int chunkX, int chunkY)
{
if (_lazyReaders.TryGetValue(map.MapID, out var reader))
{
var loaded = reader.TryReadChunk(chunkX, chunkY);
if (loaded != null)
{
var sector = map.GetRealSector(chunkX, chunkY);
if (loaded.BuiltMultisVersion == sector.MultisVersion)
{
return loaded;
}
// Snapshot is stale (multis added/removed since the bake). Fall through
// to the runtime baker; a future SaveToFile will overwrite the entry.
}
}
return BuildChunk(map, chunkX, chunkY);
}
private StepChunk BuildChunk(Map map, int chunkX, int chunkY)
{
var chunk = new StepChunk();
var sector = map.GetRealSector(chunkX, chunkY);
chunk.BuiltMultisVersion = sector.MultisVersion;
var baseX = chunkX << 4;
var baseY = chunkY << 4;
for (var dy = 0; dy < ChunkSize; dy++)
{
for (var dx = 0; dx < ChunkSize; dx++)
{
var x = baseX + dx;
var y = baseY + dy;
var cell = (dy << 4) | dx;
map.GetAverageZ(x, y, out _, out var avgZ, out _);
// Bake from the slow path's "standing Z" (the surface Z a creature actually
// stands at, not the ground avg). A* tracks newZ as standing Z, so SourceZ
// must match for the source-Z guard not to over-fire.
var standingZ = (sbyte)StepProbe.ComputeStandingZ(map, x, y, avgZ);
var result = StepProbe.ComputeMaskAt(map, x, y, standingZ);
chunk.WalkMask[cell] = result.WalkMask;
chunk.WetMask[cell] = result.WetMask;
chunk.SourceZ[cell] = standingZ;
chunk.WalkZN[cell] = result.WalkZ_N;
chunk.WalkZNE[cell] = result.WalkZ_NE;
chunk.WalkZE[cell] = result.WalkZ_E;
chunk.WalkZSE[cell] = result.WalkZ_SE;
chunk.WalkZS[cell] = result.WalkZ_S;
chunk.WalkZSW[cell] = result.WalkZ_SW;
chunk.WalkZW[cell] = result.WalkZ_W;
chunk.WalkZNW[cell] = result.WalkZ_NW;
chunk.SwimZN[cell] = result.SwimZ_N;
chunk.SwimZNE[cell] = result.SwimZ_NE;
chunk.SwimZE[cell] = result.SwimZ_E;
chunk.SwimZSE[cell] = result.SwimZ_SE;
chunk.SwimZS[cell] = result.SwimZ_S;
chunk.SwimZSW[cell] = result.SwimZ_SW;
chunk.SwimZW[cell] = result.SwimZ_W;
chunk.SwimZNW[cell] = result.SwimZ_NW;
// Multi-Z = ≥2 surfaces reachable from standingZ. Mirrors the baker's
// CheckStaticStep filter so we don't over-mark.
if (CountReachableSurfaces(map, x, y, standingZ) > 1)
{
chunk.MarkCellMultiZ(cell);
}
}
}
_buildsTotal++;
return chunk;
}
private const int PersonHeight = 16;
private const int StepHeight = 2;
/// <summary>
/// Counts walkable surfaces actually reachable from a creature standing at sourceZ.
/// Mirrors <see cref="StepProbe"/>.CheckStaticStep so cells flagged multi-Z
/// here are exactly those where the baker would have multiple candidate destinations.
/// Reachable when: surface and !impassable; stepTop ≥ itemTop; vertical overlap with
/// the creature's PersonHeight envelope.
/// </summary>
internal static int CountReachableSurfaces(Map map, int x, int y, sbyte sourceZ)
{
var startTop = sourceZ + PersonHeight;
var stepTop = startTop + StepHeight;
var count = 0;
foreach (var tile in map.Tiles.GetStaticAndMultiTiles(x, y))
{
var data = TileData.ItemTable[tile.ID & TileData.MaxItemValue];
if (!data.Surface || data.Impassable)
{
continue;
}
var itemZ = tile.Z;
var itemTop = data.Bridge ? itemZ : itemZ + data.Height;
if (stepTop < itemTop)
{
continue;
}
if (sourceZ + PersonHeight > itemZ && itemZ + data.Height > sourceZ)
{
count++;
}
}
// Land surface check — same shape, but use GetAverageZ for the land's effective top.
var landTile = map.Tiles.GetLandTile(x, y);
var landFlags = TileData.LandTable[landTile.ID & TileData.MaxLandValue].Flags;
if (!landTile.Ignored && (landFlags & TileFlag.Impassable) == 0)
{
map.GetAverageZ(x, y, out var landZ, out _, out var landTop);
if (stepTop >= landZ && sourceZ + PersonHeight > landZ && landTop > sourceZ)
{
count++;
}
}
return count;
}
}