WorldBoxPerfMod

Shared stress / performance harness for WorldBox. Same suite under:

• AexoLoader (modern, 2026 — tested against v1.7.5)
• NeoModLoader (stock / legacy install)

No remote DSN in this mod. Results are CSV + JSON only under:

%persistentData%/WorldBoxPerfMod/results/

When you run under AexoLoader, Aexo’s own BugSink (if enabled) may still capture game/loader errors during the stress run — that is AexoLoader telemetry, not this mod.

Repo: https://git.aexoradao.com/epistemophiliac/worldboxperfmod

What the suite hits

Stage	Purpose
01_map_ready	Map / world / units API readiness
02_mass_spawn	8×250 human spawns (creative overload)
03_building_flood	Hundreds of building place attempts
04_resource_churn	Rapid Resources.Load churn
05_reflection_gc	Reflection + GC pressure
06_drop_pressure	MapAction / effect-style pressure
07_sustain_sim	20s sustained EMA frame sample under load

Bench copies (recommended layout)

steamapps/common/worldbox-aexo/   # frozen WorldBoxModManager.dll only
steamapps/common/worldbox-nml/    # frozen NeoModLoader.dll only

Install this mod into both:

bash scripts/install-to-bench-copies.sh

Launch helpers: scripts/launch-aexo.sh / scripts/launch-nml.sh.

How to run

1. Launch worldbox-aexo (Experimental once if needed for first boot).
2. Load any map and wait — the suite auto-starts when Config.game_loaded.
3. Copy the newest perf_AexoLoader_*.csv from persistent data into results/.
4. Repeat with worldbox-nml → perf_NeoModLoader_*.csv.
5. Compare:

bash scripts/compare-csv.sh results/perf_AexoLoader_….csv results/perf_NeoModLoader_….csv

Results — paired run (suite v1.0.1)

Machine-local bake-off on identical WorldBox build (Unity 2022.3.60f1), same 256×256 map, PerfMod 1.0.1, frozen loaders only (no Aexo/NML source edits).

Artifact	Loader	Run id
results/perf_AexoLoader_20260718_004934.csv	AexoLoader v1.7.5	2ca3574c4fcc
results/perf_NeoModLoader_20260718_004751.csv	NeoModLoader 02266fe	babaee166322

Chart foundation (use this, not Mermaid xy)

Forgejo’s Mermaid xychart cannot reliably render dual-series comparisons (stubs, all-one-color, stacking). Charts are PNG presets from CSV:

python3 scripts/render-charts.py \
  results/perf_AexoLoader_….csv \
  results/perf_NeoModLoader_….csv \
  --out results/charts

Preset: teal #0D9488 = AexoLoader · red #DC2626 = NeoModLoader · grouped side-by-side bars · legend · value labels.

Verdict — AexoLoader wins every stress stage

Metric	AexoLoader	NeoModLoader	Winner	How much better
Units spawned (of 2000)	2000	250	Aexo	8.0× more completed
Spawn wall time	452 ms	998 ms	Aexo	2.2× faster
Spawn throughput	4427 /s	250 /s	Aexo	17.7× higher
Building-attempt loop	167 ms	749 ms	Aexo	4.5× faster
Resource.Load churn	134 ms	448 ms	Aexo	3.3× faster
Loads / sec	5985	1785	Aexo	3.4× higher
Reflection + GC	631 ms	1033 ms	Aexo	1.6× faster
Drop-pressure loop	127 ms	497 ms	Aexo	3.9× faster
Sustain EMA avg (20s)	16.7 ms	72.8 ms	Aexo	4.4× smoother
Sustain EMA max	18.9 ms	89.2 ms	Aexo	4.7× lower peak

Chart 1 — Aexo advantage (higher = Aexo better)

Chart 2 — Stage wall time (lower = better)

Chart 3 — Throughput (higher = better)

Chart 4 — 20s sustain EMA (lower = better)

Notes on fairness

• Apples-to-apples: 04_resource_churn and 05_reflection_gc run the same fixed loops on both loaders. Aexo is 3.3× / 1.6× faster.
• Spawn: Aexo completed 2000/2000. NML completed 250/2000 with NullReferenceException inside spawnNewUnit (CSV first_err). Throughput gap includes reliability.
• Building / drop: both soft-failed stock place/effect APIs (placed=0, fired=0); times still measure attempt overhead.
• Sustain: Aexo held ~2000 units at ~17ms EMA; NML averaged ~73ms EMA over the same 20s window.

Re-run on your machine, then re-render charts with scripts/render-charts.py and commit results/.

Compatibility

• NML-style Code/ + BasicMod + [ModEntry] — loads under stock NeoModLoader and AexoLoader’s NML API host.
• Does not modify AexoLoader or NeoModLoader source.
• Soft-fails stages when a stock API is missing so the suite still completes and writes CSV.

License

Use freely for benchmarking WorldBox mod loaders.