Brian Neumann-Fopiano
38cf4ac7c7
Optimize world generation
...
- Limit terrain cleanup to affected columns and batch cave boundary reads.
- Reuse offline generation workers and native air states.
- Gather history writes and defer unpublished append synchronization until atomic publication.
- Cover terrain parity, worker lifecycle, history durability, and publication failures.
2026-09-23 21:32:17 -04:00
Brian Neumann-Fopiano
ca9f28c9b0
Expanded terrain generation tooling
...
- Standalone Minecraft 26.3 generation uses native registries, biome planning and structure planning with bounded concurrency
- Offline export publishes fresh TERRAIN checkpoints for native server completion
- Checkpoints retain mantle data, immutable generation history and captured server configuration
- Output publication uses staged directories, durability checks and atomic moves
- Offline distribution tasks provide an installed launcher and reusable native fixtures
- Dimension settings control terrain, biome-bound and cave-density sampling with unchanged defaults
- Pack validation and schemas enforce the supported sampling intervals
- Biome layer caches isolate data, engine and seed identities
- Floating biome palettes keep stable thickness patterns across generation orders
- Hydrology batches terrain samples within the owning generation runtime
- Hydrology planners share bounded basis caches and reuse raster geometry and distance samples
- Demand planning admits new tiles as active plans finish
- Volumetric height queries avoid unnecessary density evaluation and bound cached anchors
- Cave generation reuses chunk-scoped inputs and semantic capture
- Object replay indexes mutations by destination and reuses resolved records
- Source-plan caches coordinate concurrent builds and reduce overlay allocation
- Prerequisite carving reads preserve the underlying hydrology before object placement
- Tree material decoding preserves stored native block identities
- Engine shutdown drains generation outside the lifecycle lock and coordinates runtime retirement
- Pregeneration sizes temporary worker boosts from processor and heap capacity
- Pregeneration restores native worker counts after drain and failure paths
- Native structure queries retain per-origin coordination without broad window locks
- Saved biome writes drain pending batches under their write stripe
- Boundary column lookups avoid repeated voxel decoding
- Modded registry consumers use explicit VolmLib providers and version-specific native source overlays
- Regression tests cover generation parity, cache isolation, concurrency, shutdown, persistence and offline checkpoints
2026-09-23 07:00:55 -04:00
Brian Neumann-Fopiano