nx_worldpipe_core.nx

nx_worldpipe_core.nx -- THE PURE STAGED-TERRAIN CORE. Extracted from nx_worldpipe.nx on 2026-08-28 for gameengine rung GE14a, so the SHIPPING WASM ENGINE and the NAS-side pipeline cannot disagree about the ground they respectively walk on and bake. WHY THE SPLIT EXISTS, MEASURED (the rung's original wording said "the engine imports nx_worldpipe" and that is NOT IMPLEMENTABLE): nx_wasm_craft carries ZERO sys_mmap / sys_read_file / sys_write -- it is allocator-free and syscall-free BY CONSTRUCTION, which is exactly why it compiles to wasm ("wasm has no mmap; every buffer is an arena offset" is written in its own arena map) -- while nx_worldpipe.wp_hydro_bake allocates 819,200 B and pulls nx_trimesh and nx_terrain_erode_lib into the closure. So the pipeline splits along the line the TARGET draws: everything here is PURE ARITHMETIC over (x,z) with no allocation and no syscall, and every allocating stage (hydrology bake, erosion bake, features, lighting) stays in nx_worldpipe. That is the boundary, and it is a property of the code, not a preference. ONE RULER, TWO CONSUMERS: the function bodies below are the SAME BYTES nx_worldpipe carried, moved not copied -- nx_worldpipe now imports this lib and keeps its call sites unchanged, so a copy cannot drift into a second answer for the same ground. The single source change is that the moved code calls rlf_noise3 DIRECTLY instead of through nx_trimesh's one-line delegate (tm_noise3 is literally `return rlf_noise3(...)`), because nx_relief_lib is 7,310 B of pure arithmetic while nx_trimesh is a 45 KB mesh library the wasm target must not carry. That change makes byte-identity of the rebuilt ELF unavailable as the neutrality proof, so the proof is an EXTERNAL CONSTANT instead: nx_worldpipe_core_gate asserts a KAT of heights and parameters harvested from the PRE-EXTRACTION organ. STAGE 1 params: 0 continentalness, 1 temperature, 2 humidity, 3 erosion, 4 weirdness, 5 river. license_tier: ORIGINAL No hw writes (Rule 26).