promotional bannermobile promotional banner

Three Dash Structure Optimizer

Up to 8x faster chunk generation in structure-heavy packs when used alongside C2ME & Lithium, and ~70% shorter autosave lag spikes for smoother server TPS. Also adds extra TPS optimizations beyond structures. Works with other optimizers or standalone.
Back to Files

three-dash-structure-optimizer-2.0.2.jar

File namethree-dash-structure-optimizer-2.0.2.jar
Uploader
DlVl4ikDlVl4ik
Uploaded
Nov 5, 2026
Downloads
266
Size
1.1 MB
Mod Loaders
ForgeFabricNeoForge
File ID
9452354
Type
R
Release
Supported game versions
  • 26.3
  • 26.2

Curse Maven Snippet

Forge

implementation "curse.maven:three-dash-structure-optimizer-1741418:9452354"

Fabric

modImplementation "curse.maven:three-dash-structure-optimizer-1741418:9452354"

NeoForge

implementation "curse.maven:three-dash-structure-optimizer-1741418:9452354"

Learn more about Curse Maven

What's new

2.0.2


1. What's New in 2.0.2

  1. Sky Cell Culling & Density Truncation (terrain.sky_cells, terrain.sky_truncate):
    • Skips expensive 3D noise router and density function evaluations for high-altitude air volumes above the max terrain surface.
    • Saves up to 57% of density branch evaluations on terrain generation threads with strict mathematical equivalence.
    • Enforced aquifer barrier safety: safe truncation boundary is mathematically bounded to skipSamplingAboveY + 1, completely preserving high-altitude mountain lake barriers and preventing fluid leakage.
  2. Noodle Cave Pre-Bounding (terrain.noodle_bound):
    • Bounds noodle cave noise queries early to skip irrelevant underground sub-evaluations.
  3. Thread-Safe Structure Jigsaw List Caching (Minecraft 26.3):
    • Caches immutable template pool target lists during heavy jigsaw structure placement using thread-safe AtomicReferenceArray publication, eliminating repetitive heap allocations and completely eliminating data races on multi-threaded worker pools (C2ME).
    • Invalidation on reload: automatically flushes cache on /reload or template reload.
  4. Dimension-Isolated Structure Layout Slicing (Roguelike Dungeons):
    • Partitions time-sliced dungeon generation by dimension (ResourceKey<Level>), preventing cross-dimensional session pruning and infinite room planning resets in the Nether and the End.
  5. Zero-Stall Structure Context Guard (O(1) Gate 0 Priority):
    • StructureCallerContext fast thread-local check elevated to top priority in UniversalStructureDetector, bypassing burst counts and 20-tick cooldowns. Eliminates 700–2500 ms server thread freezes and deadlocks with C2ME workers.
  6. Optimized NBT Chunk Palettes & Boundary Safety (Minecraft 26.2):
    • Direct palette serialization to NBT bypassing slow DataFixerUpper codec chains on generation worker threads.
    • Region boundary guard queries getLatestStatus() in memory, preventing erroneous AIR block and Y = -64 replacement for freshly generated in-memory chunks.
  7. Live In-Mod Referee (Zero World Divergence Guarantee):
    • Continuous live self-check compares optimized routines against vanilla logic:
      • Chunk palette direct tags vs vanilla codec: 20,000 containers compared, 0 differences.
      • Biome climate column cache vs game samples: 1,000,448 samples compared, 0 differences.
      • Jigsaw element lists vs game lists: 8,218 lists compared, 0 differences.
      • Sky truncation voxel fills vs vanilla fills: 49+ fills compared, 0 differing voxels.
      • Fail-open fallback: automatically reverts to vanilla behavior if any anomaly is encountered.

2. Speed Benchmarks: 2.0.2 vs 2.0.1

All benchmarks were run on dedicated Oracle Cloud VM stands (VM1 & VM2, A1 2 ARM cores, dedicated pinning, seed 20260923, identical modpacks).

Minecraft 26.2 (Fabric Stand VM1, 110 mods, 961 chunks)

Metric 2.0.1 Base 2.0.2 Final Difference (%) Absolute Gain
Generation Time (gen_s) 59.30 s 51.05 s -13.9% -8.25 s faster
Total Process CPU (gen_cpu_s) 84.7 s 74.0 s -12.6% -10.7 s CPU saved
C2ME Worker CPU (c2me_workers) 56.2 s 47.8 s -14.9% -8.4 s CPU saved

Minecraft 26.3 (Fabric Stand VM2, 110 mods, 961 chunks)

Metric 2.0.1 Base 2.0.2 Round 1 2.0.2 Round 2 (Final) Difference vs 2.0.1 (%)
Generation Time (gen_s) 43.4 s 40.5 s 40.2 s -7.4% faster
Total Process CPU (gen_cpu_s) 70.0 s 66.0 s 65.3 s -6.7% CPU saved
C2ME Worker CPU (c2me_workers) 40.0 s 36.0 s 36.5 s -8.8% CPU saved

3. Multi-Loader Stability: Round 2 Verification (Final Jar)

Every supported mod loader was verified in live generation runs using the final release jar (e018a9f6).

Loader Minecraft Mods Loaded Generation Time (gen_s) Loaded Chunks Errors / Status
Fabric 26.3 110 mods 40.20 s 986 chunks 0 errors, Exit 0
NeoForge 26.3 100 mods 46.45 s 986 chunks 0 errors, Exit 0
Forge 26.2 21 mods 56.95 s 2,809 chunks 0 errors, Exit 0

4. World Equivalence & Divergence Audit (Round 2)

Target System Samples Audited Divergences Verdict
Direct Chunk Palettes vs Codec 20,000 containers 0 differences Bit-for-bit identical
Climate Column Sampling 1,000,448 samples 0 differences Bit-for-bit identical
Structure Jigsaw Lists 8,218 queries 0 differences Bit-for-bit identical
Sky Truncation Voxels 28,959+ volumes 0 differing voxels Bit-for-bit identical
PreBake Template Upgrades 32 / 32 templates 0 differences Bit-for-bit identical
World Save Sizes Across 3 loaders Identical size Bit-for-bit identical

This mod has no related projects