wip(assembler): round S3 -- far ground from the native heightfield; LOD7/8 tiles 36-66% cheaper

Far-shell repair (81-96% of S2's LOD7/8 tile time) is skipped for certified
SurfaceWorld tiles; ground and cap come from the shared planar heightfield.
Near LOD0-3 hashes equal S2; far recipe version 1 salts LOD5+ identities.
Cold first view 6.6 -> 3.4 s (aim 1.57 s). Not a visual sign-off: distant cap
shows floating pieces, sprint streaming arrives late.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MmSoF64BbRFmk9C3p1VsBJ
This commit is contained in:
Fr0zkaandClaude Opus 5.5 committed 2026-10-18 21:29:52 +02:00
1 parent 6ddb834b1c
commit d0f97bf79a
26 files changed
+976 -30

No files matched your search

@@ -1194,3 +1194,27 @@ and no link, commit, push or stash was created.
[Full S2 report](../../WorldAssemblerM3/Round6/REPORT.md).
— Sol
## World assembler S3 / M3 Round 7 — 2026-10-18
Far SurfaceWorld ground and cap now come from the native heightfield at LOD5+, and certified
LOD7/8 SurfaceWorld tiles skip the derived shell repair. The S2 shell stage used about 81% of
SurfaceWorld LOD7/8 tile p95 and 96% of FloatingIslands LOD8. S3 warmed p95 fell 39–66% across
those owner LOD7/8 rows. A falsifiable LOD7/8 ground audit passes 128 shared points within 0.043 cm
of exact near-field roots, and its 100 cm datum-offset control fails. LOD0–LOD3 density, mesh,
payload and topology hashes match S2 exactly; SurfaceWorld and FloatingIslands W1/W8/W16 pass.
The owner horizon is not ready for visual sign-off. Cold whole-view presentation improved from
6.622 s to 3.420 s, still above the 1.57 s target. At 60 m/s the clipmap arrived after the player
crossed the route. The approach/retreat captures expose gaps and large faces, and the new distant
cap has isolated-looking pieces. I report those as failures rather than claiming a pop-free sequence.
No new source file was added. Core, editor audit and standalone viewer compile spots are listed with
the run ledger and capture links in [WorldAssemblerM3/Round7/REPORT.md](../../WorldAssemblerM3/Round7/REPORT.md).
CODEMAP section 3 and ARCHITECTURE section 8.24 are updated. Core-off, M2 fixtures, staged 25 cm
SurfaceWorld/FloatingIslands, edits and W1/W8/W16 pass. No new UECC folder, timeout or 0xC0000142
occurred. I did not open the owner's editor, build his project or edit assets, and I did not commit,
push or stash.
— Luna