PerformanceX

v1.1.88 — Mob AI Throttler Performance Overhaul

Performance Fixes

• Mob AI Throttler — chunk-based sweep: The sweep loop previously called world.getLivingEntities() which iterates every entity across the entire world regardless of proximity to any player. On large worlds this could be thousands of entities scanned per sweep. Now the sweep only visits chunks within each player's configured AI range, drastically cutting entities evaluated on sparsely populated worlds.

• Mob AI Throttler — state-change-only NMS calls: setNoAi() and setAware() were called on every mob during every sweep pass even when the AI state was already correct. Since these go through NMS reflection they're expensive to call repeatedly. Now they only fire on actual transitions (enabled→disabled or disabled→enabled), so after the first pass most mobs cost zero NMS calls.

• Mob AI Throttler — in-memory reduced-AI counter: The "reduced AI" tick counter was read and written via PDC (Persistent Data Container, which also uses NMS reflection internally) on every entity every tick. Replaced with a ConcurrentHashMap<UUID, Integer> in-memory counter — no NMS overhead, instant lookup.

These three changes together eliminate the mob AI throttler from the hot path it was appearing in on Spark profiler reports despite being a performance module.