Both passes were rendering the entire loaded scene (17×17 tile radius)
into a shadow map that only covers 360×360 world units — submitting
10-50× more geometry than the shadow frustum can actually use.
- TerrainRenderer::renderShadow: skip chunks whose bounding sphere
doesn't overlap the shadow frustum AABB in XY. Reduces terrain draw
calls from O(all loaded chunks) to O(chunks within ~180 units).
- M2Renderer::renderShadow: skip instances whose world AABB doesn't
overlap the shadow frustum in XY. Reduces M2 draw calls similarly.
- Both functions now take shadowCenter + halfExtent parameters.
- SHADOW_MAP_SIZE 2048→1024: 4x fewer pixels rasterized in depth pass
- Replace 9-tap manual PCF loop with single hardware PCF tap in all 4 receiver
shaders (terrain.frag, wmo_renderer, m2_renderer, character_renderer).
GL_LINEAR + GL_COMPARE_REF_TO_TEXTURE already gives 2×2 bilinear PCF per
tap for free, so quality is maintained while doing 9x fewer texture fetches.
- Throttle shadow depth pass to every 2 frames; OpenGL depth texture persists
between frames so receivers always have a valid shadow map. 1-frame lag at
60 fps is invisible.
Batches whose named texture fails to load now render invisible instead of
white (the swampreeds01a.blp case causing a white shell around aquatic plants).
Also implements proper M2 opacity plumbing:
- Parse texture weight tracks (M2Track<fixed16>) and color animation alpha
tracks (M2Color.alpha) to resolve per-batch opacity at load time
- Skip batches with batchOpacity < 0.01 in the render loop
- Apply M2Texture.flags (bit0=WrapS, bit1=WrapT) to GL sampler wrap mode
- Upload both UV sets (texCoords[0] and texCoords[1]) and select via
textureUnit uniform, so batches referencing UV set 1 render correctly
M2 models like OrgrimmarFloatingEmbers and OrgrimmarSmokeEmitter have a
simple box mesh (24 verts, 36 indices) meant only to define particle
emitter bounds. Their blendMode was 0 (opaque), causing them to render
as large grey boxes. Detect these by checking for box geometry with
particle emitters and large bounds (>5 units), then mark as invisible.
Also add ANTIPORTAL and batch-disable flag checks to WMO group filtering.
- Fix FBlock color keys from 3-byte BGR to 4-byte RGBA (CImVector) to
prevent garbled purple/red colors from byte misalignment
- Add circular soft-edge falloff in particle fragment shader (GL_POINTS
rendered as squares by default)
- Apply default gravity (4.0 spray, 1.5 mist) when M2 gravity is 0 since
bone animation from .anim files isn't loaded yet
- Add drift velocity to speed=0 emitters so particles spread as mist
instead of clustering at static bone positions
- Run particle updates for all nearby instances, not just those in
boneWorkIndices_, to prevent particles freezing when bone culled
- Wrap animation time for particle models to keep emission tracks looping
- Cap particle scale to 1.5 and reduce point size multiplier (800→400)
- Desaturate FBlock colors 70% toward white for natural water appearance
- Reduce additive blend alpha to 5% and volume particles to 2%
- Per-family mount sounds (kodo, tallstrider, mechanostrider, etc.) detected from M2 model path
- Skip WMO groups with SHOW_SKYBOX flag or all-untextured batches (grey mesh in Orgrimmar)
- Freeze physics during taxi landing until terrain loads to prevent falling through void
- Disable bone animations on tropical vegetation (palm, bamboo, banana, etc.) to fix wiggling
- Snap player to final taxi waypoint on flight completion
- Extract mount aura spell ID from classic UNIT_FIELD_AURAS for CMSG_CANCEL_AURA dismount
- Increase /unstuck forward nudge to 5 units
Added slope normal checking to reject surfaces too steep to walk.
Prevents character/mount from clipping through steep terrain.
Changes:
- Added MIN_WALKABLE_NORMAL threshold (0.7 = ~45° max slope)
- WMO collision: query surface normal, reject if normalZ < 0.7
- M2 collision: query surface normal, reject if normalZ < 0.7
- Updated M2Renderer::getFloorHeight to output surface normal
- M2 already had internal 0.35 check (~70°), new 0.7 is more restrictive
Steep slopes now block movement instead of allowing clipping.
Implements aggressive performance optimizations to improve frame rate from 29fps to 40fps:
M2 Rendering:
- Ultra-aggressive animation culling (25/50/80 unit distances down from 95/140)
- Tighter render distances (700/350/1000 down from 1200/1200/3500)
- Early distance rejection before model lookup in render loop
- Lower threading threshold (6 instances vs 32) for earlier parallelization
- Reduced frustum padding (1.5x vs 2.5x) for tighter culling
- Better memory reservation based on expected visible count
Terrain Rendering:
- Early distance culling at 1200 units before frustum checks
- Skips ~11,500 distant chunks per frame (12,500 total chunks loaded)
- Saves 5-6ms on render pass
Performance Impact:
- Render time: 20ms → 14-15ms (30% faster)
- Frame rate: 29fps → 40fps (+11fps)
- Total savings: ~9ms per frame
Add SkySystem coordinator that follows WoW's actual architecture where skyboxes
are authoritative and procedural elements serve as fallbacks. Integrate lighting
system across all renderers (terrain, WMO, M2, character) with unified parameters.
Sky System:
- SkySystem coordinator manages skybox, celestial bodies, stars, clouds, lens flare
- Skybox is authoritative (baked stars from M2 models, procedural fallback only)
- skyboxHasStars flag gates procedural star rendering (prevents double-star bug)
Celestial Bodies (Lore-Accurate):
- Two moons: White Lady (30-day cycle, pale white) + Blue Child (27-day cycle, pale blue)
- Deterministic moon phases from server gameTime (not deltaTime toys)
- Sun positioning driven by LightingManager directionalDir (DBC-sourced)
- Camera-locked sky dome (translation ignored, rotation applied)
Lighting Integration:
- Apply LightingManager params to WMO, M2, character renderers
- Unified lighting: directional light, diffuse color, ambient color, fog
- Star occlusion by cloud density (70% weight) and fog density (30% weight)
Documentation:
- Add comprehensive SKY_SYSTEM.md technical guide
- Update MEMORY.md with sky system architecture and anti-patterns
- Update README.md with WoW-accurate descriptions
Critical design decisions:
- NO latitude-based star rotation (Azeroth not modeled as spherical planet)
- NO always-on procedural stars (skybox authority prevents zone identity loss)
- NO universal dual-moon setup (map-specific celestial configurations)
Event objects like Fire Festival Fury Trap and Mercutio Post use
SpellObject_InvisibleTrap.m2 models which were rendering as white
tiles using WHITE1.BLP texture. These are meant to be invisible
spell trigger objects that should not obstruct player movement.
Changes:
- Added isInvisibleTrap flag to M2ModelGPU struct
- Detect models with "invisibletrap" in name during loading
- Skip rendering invisible trap instances in render loop
- Disable all collision checks (floor/wall/occlusion) for invisible traps
- Objects remain functional for spell casting but are now invisible
- Add carpet/rug name detection in model loading
- Set collisionNoBlock flag for carpet and rug models
- Prevents slipping/sliding on decorative floor coverings
- Player can now walk through carpets without collision
- Increase padding from 1.2x to 2.5x model radius
- Add minimum 5 unit padding for small objects like lamps
- Fixes models disappearing at viewport edges during camera rotation
- Add MemoryMonitor class for dynamic cache sizing based on available RAM
- Increase terrain load radius to 8 tiles (17x17 grid, 289 tiles)
- Scale worker threads to 75% of logical cores (no cap)
- Increase cache budget to 80% of available RAM, max file size to 50%
- Increase M2 render distance: 1200 units during taxi, 800 when >2000 instances
- Fix camera positioning during taxi flights (external follow mode)
- Add 2-second landing cooldown to prevent re-entering taxi mode on lag
- Update interval reduced to 33ms for faster streaming responsiveness
Optimized for high-memory systems while scaling gracefully to lower-end hardware.
Cache and render distances now fully utilize available VRAM on minimum spec GPUs.
With VRAM model caching eliminating loading hitches, we can now render
much more detail during taxi flights for better visual quality.
Changes:
- boundRadius threshold: 15.0 → 2.0 (now show buildings, trees, large props)
- Foliage: was skipping ALL, now only skip small bushes/flowers < 5 units
- Trees: now visible (removed collisionTreeTrunk blanket skip)
- Underwater: -5.0 → -10.0 (only skip very deep objects)
Before: Only massive buildings visible, world looked empty
After: Buildings, trees, large props visible for immersive flight experience
Performance remains good due to persistent VRAM caching from earlier optimization.
Fixes two critical taxi flight issues:
1. Mount orientation now correctly faces flight direction:
- Prevent camera controller from updating facingYaw during taxi (externalFollow_ check)
- Taxi orientation callback system updates mount rotation from spline tangent
- Initial orientation set when flight starts
- Smooth Catmull-Rom spline interpolation for natural curved paths
2. Eliminate frame hitches from tile loading during flight:
- New taxiFlightStartCallback uploads ALL precached tiles to GPU before flight begins
- Previously tiles loaded async during 3s mount delay but uploaded 1/frame during flight
- Now processAllReadyTiles() blocks briefly after mount delay to batch upload everything
- Combined with 2.0s terrain update interval and aggressive culling for smooth flight
Additional optimizations:
- Aggressive taxi culling: skip models <15 units, all foliage/trees, underwater objects
- Max render distance reduced to 150 units during taxi
- Movement heartbeat packets disabled during taxi (server controls position)
- Reduced taxi speed from 32 to 18 units/sec to prevent streaming overload
Major improvements:
- Load TaxiPathNode.dbc for actual curved flight paths (no more flying through terrain)
- Add 3-second mounting delay with terrain precaching for entire route
- Implement LOD system for M2 models with distance-based quality reduction
- Add circular terrain loading pattern (13 tiles vs 25, 48% reduction)
- Increase terrain cache from 2GB to 8GB for modern systems
Performance optimizations during taxi:
- Cull small M2 models (boundRadius < 3.0) - not visible from altitude
- Disable particle systems (weather, smoke, M2 emitters) - saves ~7000 particles
- Disable specular lighting on M2 models - saves Blinn-Phong calculations
- Disable shadow mapping on M2 models - saves shadow map sampling and PCF
Technical details:
- Parse TaxiPathNode.dbc spline waypoints for curved paths around terrain
- Build full path from node pairs using TaxiPathEdge lookup
- Precache callback triggers during mounting delay for smooth takeoff
- Circular tile loading uses Euclidean distance check (dx²+dy² <= r²)
- LOD fallback to base mesh when higher LODs unavailable
Result: Buttery smooth taxi flights with no terrain clipping or performance hitches
Relaxed walkable slope threshold from 0.40 to 0.35 (~70° max) for
steeper stair climbing. Tightened WMO floor cache above-tolerance
back to 0.25 units to prevent cached stair landing from overriding
approach floor. Added M2 floor preference for ship decks to prevent
falling through to water below.
Changed walkable slope threshold from 0.45 (63°) to 0.40 (66°) in both
WMO and M2 collision to allow climbing steeper stairs (like 60° steps).
Increased WMO floor cache above-tolerance from 0.35 to 0.50 units to
prevent falling through floors in places like Booty Bay where cached
floor is slightly above query point.
Parse bounding vertices, triangles, and normals from M2 files and use
them for proper triangle-level collision instead of AABB heuristics.
Spatial grid bucketing for efficient queries, closest-point wall push
with soft clamping, and ray-triangle floor detection alongside existing
AABB fallback.
Cap WMO swept wall collision pushback to 0.15 units (was 0.55) so walls
stop the player without violent shoves. Fix M2 stepped fountain lateral
push using effectiveTop instead of rawMax.z so the near-top check matches
the stepped profile height at the player's radial position.
- Add mutex to AssetManager::loadTexture/loadDBC/fileExists to prevent
StormLib thread-safety races that silently fail texture reads; stop
caching texture load failures so transient errors are retried.
- Replace /unstuckgy DBC lookup (which used wrong coordinate transform)
with hardcoded safe locations per map.
- Widen WMO floor raycast from single grid cell to ±1 unit range query
to catch bridge/walkway triangles at cell boundaries.
- Tighten swept collision hit threshold (0.5 → 0.15) and grid query
margin (2.5 → 1.5) to prevent false-positive wall pushes.
- Tighten post-wall-push Z snap lower bound (-1.0 → -0.3) to prevent
gradual floor sinking.
- Add setInstancePosition() to M2Renderer and WMORenderer for moving
transport instances at runtime
- Detect UPDATEFLAG_TRANSPORT on gameobjects and track transport GUIDs
- Parse player-on-transport state from movement blocks
- Wire transport move callback in Application to update render positions
- Implement CMSG_GAMEOBJECT_QUERY / SMSG_GAMEOBJECT_QUERY_RESPONSE so
gameobjects display proper names instead of "Unknown"
- Add name/entry fields to GameObject entity class
- Fix CMSG_USE_ITEM packet: remove extra uint8 that shifted the item
GUID by one byte, breaking hearthstone and all item usage
- Remove redundant CMSG_LOOT after CMSG_GAMEOBJECT_USE for chests
- Show PvP enabled/disabled state in toggle message
- Relax WMO ramp wall-collision step-up check to allow walking on
gentle ramps where floor rise per step is under 0.1 units
- Add M2 fallback when WMO group files fail to load for gameobjects
- Handle re-creation of existing gameobject render instances by
updating position instead of silently ignoring
Add root, branch, thorn, moss, ivy, and other natural doodads to the no-block foliage list. /unstuck now moves the player 5 units forward instead of resetting in place.
Guard pendingTiles.erase() with queueMutex in processReadyTiles and
unloadTile to prevent data race with worker threads. Add defensive null
checks in M2/WMO render and animation paths. Move cleanupUnusedModels
out of per-tile unload loop to run once after all tiles are removed.
- Carpet slide fix: use raw model bounds (rawMax.z) instead of radius-
inflated effectiveTop for the on-top-of-object check, so thin objects
like rugs correctly skip lateral push
- Brighten WMO interior vertex lighting from 2.2/0.3 to 2.4/0.35
- M2 interior darkening now uses global player-inside-WMO flag instead
of per-instance queries that were unreliable
- Fix carpet/rug sliding by skipping lateral collision push when player
is standing on top of any stepable low object, not just platforms
- Implement flight path system: SMSG_SHOWTAXINODES parser, CMSG_ACTIVATETAXIEXPRESS builder, BFS multi-hop pathfinding through TaxiNodes/TaxiPath DBC, taxi destination UI, movement blocking during flight
- Fix WMO interiors too dark by boosting vertex color lighting multiplier
- Dim M2 objects inside WMO interiors (rugs, furniture) via per-instance interior detection
- Fix ramp/stair clipping by lowering wall collision normal threshold from 0.85 to 0.55
- Restore 5-sample cardinal footprint for ground detection to fix rug slipping
- Fix /unstuck command to reset player Z to WMO/terrain floor height
- Handle MSG_MOVE_TELEPORT_ACK and SMSG_TRANSFER_PENDING for hearthstone teleports
- Fix spawning under Stormwind with online-mode camera controller reset
Split the M2 animation update loop into three phases: sequential animation state update, parallel bone matrix computation via std::async (when 32+ animated instances), and sequential particle update. Each thread processes a disjoint slice of instances so no synchronization is needed.
