Fr0zka/VoxelForge
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

perf(generation): per-tile reach test skips landing and structural posts -- ~12% less worker CPU

Browse Source 
The remove-one ablation put landing posts at ~10% and passage structural posts at ~6% of worker
time, though landings exist only near passage ends and the origin. The parked exact-cull round's
per-SAMPLE post bounds had been a net loss. This round decides reach once per TILE (Codex;
reviewed by me).

- VoxelTilePostReach.h: one per-tile reach mask (thread-local byte), set when the tile is set
  up. The origin landing, passage landing SDF/floors, native floors, tunnel air and passage
  structural posts return their input unchanged on tiles no instance can reach. Per sample this
  costs an inline flag read.
- Reach proofs come from each post's geometry (origin rooms, landing SDF and floors, native floor
  band, structural support spheres), not from comments.
- voxel.TilePostReachDebug=1 evaluates the skipped posts anyway and compares the FINAL field;
  voxel.TilePostReachScale=0.5 deliberately shrinks the reach and must fail.

Default build: 694 tiles, landing reachable on 4 (skipped on 690), structural posts skipped on 169.
Soundness: 123,958,673 comparisons, 0 differences; shrunk reach: 13,929 differences, FAIL as
required.

Final DEFAULT build vs the committed-tip DLL (same command line, two interleaved runs each):
  static  worker 34.80 -> 30.52 s (-12%)   LOD0 generation p50/p95 67.0/107.6 -> 55.8/95.8 ms
  moving  worker 75.96 -> 66.66 s (-12%)   LOD0 generation p50/p95 66.3/106.0 -> 56.0/95.8 ms
Field exact: 841/841 keyed tiles vs a fresh committed-tip trace (669,902 triangles, which includes
9260775's gap floors), export b3e5f4c3..9b377, capability 20,830 / 10,909; strate-crossing and
sealed-proof tests pass.
Ledger: all launches passed or failed where designed to; early raw-SDF debug attempts failed
until the comparator was corrected to compare the final field. No crash folders.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TaFdXhTdYMkW1xrV6ApT6Y
This commit is contained in:
Fr0zkaandClaude Opus 5 committed 2026-09-15 07:42:13 +02:00
1 parent 92607751b1
commit ce4d8783e9
7 files changed
+790 -20
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

Source/VoxelForge/Public/VoxelPassageGeometry.h
+9 -2
View File
@@ -471,13 +471,20 @@ namespace VoxelPassageGeometry
FORCEINLINE bool OriginLandingRoomTouchesLattice(
const FBox& VoxelBox, const FIntVector& LatticeOrigin, int32 Step,
float StrateTopZ, float StrateBottomZ, float SealThickness, float OriginRadius)
float StrateTopZ, float StrateBottomZ, float SealThickness, float OriginRadius,
float ReachScale = 1.0f)
{
if (!VoxelBox.IsValid) return false;
if (!VoxelMath::IsFinite(ReachScale) || ReachScale <= 0.0f)
{
// This is an invalid proof request, not an empty room. The manager treats a
// malformed scale as conservative/unknown before calling this helper.
return true;
}
const FOriginLandingGeometry Geometry = BuildOriginLandingGeometry(
StrateTopZ, StrateBottomZ, SealThickness, OriginRadius);
if (!Geometry.bValid) return false;
const float Pad = LandingCarveBlendVoxels;
const float Pad = LandingCarveBlendVoxels * ReachScale;
return LatticeAxisHasSampleInInterval(
FMath::Max((float)VoxelBox.Min.X, -Geometry.HalfWidth - Pad),
FMath::Min((float)VoxelBox.Max.X, Geometry.HalfWidth + Pad),