Fr0zkaandClaude Opus 5.5
78641da0bb
Editor-only, default off (voxel.ResearchRedefinedDensityField). The redefined field is a seeded, periodic cross-corridor pattern per strate plus the authored passages, landings and a shaft tree, with four value-noise detail bands at every near-surface 25 cm sample and sound 8^3 brick proofs. 259^3 ring, six workers: 308 ms median (467 ms without skips) vs 8.9 s for the live field. Detail, boundary-artefact, determinism (1 vs 6 workers, two runs) and bound-audit checks pass. Why it is not a pass: the slices show a mostly-solid field with a few straight corridors, where the owner's world is mostly open space; rooms, the tunnel network, composer recipes and archetypes are gone. Much of the speed comes from 60% of bricks being proven solid in that simpler world. What it does show: detail at full resolution plus brick skipping costs ~80-130 ns/sample; the cost of the live field is how rooms/tunnels/passages are represented, not the detail or the lattice. The two walkability tests were redirected to the research field in the working tree; that change is NOT committed (saved as scratchpad step2d-test-redirect.patch) so they keep testing the live world. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MmSoF64BbRFmk9C3p1VsBJ
7.6 KiB
7.6 KiB
Step 2d launch ledger
All Unreal processes were launched through Tools/VoxelForgeTest.ps1 and ran serially. The
guarded launches used below-normal priority, the shared persistent DDC where recorded,
-MaxParallelActions=4, a prelaunch 6 GiB memory check, an 8 GiB per-process limit, and a
30-minute timeout. The retained ledgers show no crash reporters, no process above 8 GiB, and
no timeout. Exit 255 below means automation assertions failed; it was not a native crash.
Benchmark attempts
| Label | Build | Unreal process | Result |
|---|---|---|---|
step2d_rebuild_01 |
UBT PID 25780, exit 6, peak 2.71 GiB | Not launched | Compile failed on the ApplicationContextMask automation flag; corrected in the next source pass. |
step2d_rebuild_02 |
UBT PID 19420, passed, peak 1.57 GiB | PID 13996, exit 2, peak 1.83 GiB | 3.834 s optimized ring; brick/sample skipping was too low for the speed gate. |
step2d_rebuild_03 |
Passed; PID not retained in the compact log | PID 15480, exit 2, peak 1.82 GiB | 3.918 s; detail spectrum failed. The smoothing negative control worked; the proof negative control did not. |
step2d_rebuild_04 |
Not launched | Not launched | Harness rejected a relative -Out path before Unreal startup. |
step2d_rebuild_05 |
UBT PID 23904, passed | PID 24708, exit 2, peak 1.82 GiB | Scratch reuse gave 233.27 density mismatches and hash/boundary failures; scratch ownership was corrected. |
step2d_rebuild_06 |
UBT PID 7540, passed | PID 25828, exit 2, peak 1.82 GiB | Detail, smoothing, boundary, and hash checks passed, but 3.70 s and 1.46% sample skipping missed the speed gate. |
step2d_rebuild_07 |
UBT PID 23800, passed | PID 5208, exit 2, peak 1.81 GiB | 3.66 s; the cached XY macro did not improve the timing. |
step2d_rebuild_08 |
UBT PID 25304, passed | PID 7108, exit 2, peak 1.81 GiB | 2.62 s; 22.20% of bricks and 21.18% of samples skipped, still over target. |
step2d_rebuild_09 |
UBT PID 25612, passed, peak 1.51 GiB | PID 24256, exit 0, peak 1.82 GiB | Superficially under 600 ms, but invalid as evidence: classifier gap/default/chunk semantics had not yet been fixed. Excluded. |
step2d_harness_10 |
UBT PID 2544, passed, peak 1.49 GiB | PID 23204, exit 0, peak 1.84 GiB | 503.026 ms and hash/detail/boundary checks passed. Superseded because its deliberately undersized-bound control did not catch a wrong sign. |
step2d_harness_11 |
UBT PID 2900, passed, peak 1.50 GiB | PID 10688, exit 0, peak 1.83 GiB | 497.518 ms; the scale-0 versus scale-4 undersized-bound control caught 35,847 wrong-sign samples. |
step2d_harness_12 |
UBT PID 13868, passed, peak 1.87 GiB | PID 16872, exit 0, peak 1.81 GiB | Final dynamically scheduled run: 308.177 ms median, all hashes and research gates passed. |
Functional runs
| Label | Build | Unreal process | Result |
|---|---|---|---|
step2d_functional_01 |
UBT PID 21184, passed | PID 3444, exit 255, peak 3.26 GiB | Base connectivity passed; refinement failed on an endpoint self-query. The launcher detected a write under C:\Users\Jahni\AppData\Local\UnrealEngine; later runs used isolated per-run local data. |
step2d_functional_02 |
UBT PID 11844, passed, peak 1.87 GiB | PID 5048, exit 255, peak 3.13 GiB | Same endpoint-snap failure; isolated LocalAppData was clean. |
step2d_functional_03 |
UBT PID 2060, passed, peak 1.90 GiB | PID 25648, exit 255, peak 3.19 GiB | Origin probe compared raw density 8 with a coarse center at -3.534534 near Z=-1012. |
step2d_functional_04 |
Failed launch; retained PID 23996, stage split not retained | Exit 255; peak not retained | Range-normalization fix did not resolve the origin probe. No crash was recorded. |
step2d_redefined_seeds_01 |
Build passed; PID not retained | PID 21888, exit 255 | Initial multi-seed endpoint-snap expectation was stricter than the route's full-density segment check; expectation was revised. |
step2d_functional_05 |
UBT PID 17000, passed | PID 7540, exit 255 | 12/16 fitted VerticalShafts seeds passed. |
step2d_functional_06 |
Build passed; PID not retained | PID 21948, exit 255, peak 3.43 GiB | 12/16; seed 1337's physical arrival/departure audit found 13 bad samples of 15. |
step2d_redefined_seeds_02 |
UBT PID 14672, exit 0, peak 1.63 GiB | PID 2220, exit 0, peak 2.61 GiB | Passed an earlier three-seed traversal version. |
step2d_functional_07 |
Build/Unreal stage PID 15712 in retained summary | Exit 255, peak 3.42 GiB | 12/16; no native crash. |
| Relative-output-path prelaunch attempt | Not launched | Not launched | Harness stopped before Unreal startup; corrected to an absolute output directory. |
step2d_functional_08 |
UBT PID 12596, passed, peak 1.87 GiB | PID 20624, exit 255, peak 3.42 GiB | 13/16; origin path passed, three measurement limits remained. |
step2d_functional_09 |
UBT PID 10848, passed, peak 1.50 GiB | PID 16764, exit 255, peak 3.46 GiB | 13/16 after the cache-key fix. |
step2d_functional_10 |
UBT PID 24704, passed, peak 1.44 GiB | PID 7512, exit 255, peak 3.56 GiB | 13/16 after connector-radius changes. |
step2d_functional_11 |
UBT PID 17088, exit 6, peak 1.96 GiB | Not launched | Compile error C4456 from a duplicated XYCount; fixed before the next build. |
step2d_functional_12 |
UBT PID 4760, passed, peak 1.95 GiB | PID 26520, exit 255, peak 3.49 GiB | 13/16 after explicit bridges. |
step2d_functional_13 |
UBT PID 19416, passed, peak 1.50 GiB | PID 26444, exit 255, peak 3.50 GiB | 13/16; diagnostics showed 16 retries and both endpoints in the same large component. |
step2d_functional_14 |
UBT PID 3156, passed, peak 1.50 GiB | PID 17892, exit 255, peak 3.49 GiB | 13/16 after widening the landing bridge. |
step2d_functional_15 |
UBT PID 5648, passed, peak 1.49 GiB | PID 23788, exit 255, peak 3.53 GiB | 14/16 after adding the vertical cavern band. |
step2d_functional_16 |
UBT PID 20028, passed, peak 1.13 GiB | PID 4916, exit 255, peak 3.60 GiB | Controlled-wall assumptions and the 14/16 refinement audit both failed. |
step2d_functional_17 |
UBT PID 14952, passed, peak 1.50 GiB | PID 2604, exit 0, peak 2.69 GiB | Three-seed traversal test passed: 21 landing brackets and 18 routes. |
step2d_functional_18 |
UBT PID 10496, passed, peak 1.49 GiB | PID 25636, exit 0, peak 2.70 GiB | Passage-open-space test passed. |
step2d_functional_19 |
UBT PID 22588, passed, peak 1.48 GiB | PID 7316, exit 255, peak 3.71 GiB | Controlled wall returned conservative unknown after zero retries; refinement remained 14/16. |
step2d_functional_20 |
UBT PID 14076, passed, peak 1.88 GiB | PID 23120, exit 255, peak 3.60 GiB | Wall samples were -2/-102/-2, so the chosen location was not an air/solid/air control; refinement remained 14/16. |
step2d_functional_21 |
UBT PID 8224, passed, peak 1.58 GiB | PID 25652, exit 255, peak 3.67 GiB | StrateConnectivity passed; refinement remained 14/16 because the fallback had been placed in a skipped diagnostic branch. |
step2d_functional_22 |
UBT PID 26528, passed, peak 1.62 GiB | PID 24396, exit 0, peak 3.68 GiB | Both required connectivity tests passed; the fitted audit directly confirmed 16/16 routes at step 2. |
step2d_functional_23 |
Build not requested | PID 15176, exit 0, peak 2.44 GiB | PassageLandsInOpenSpace passed. |
step2d_functional_24 |
Build not requested | PID 26336, exit 0, peak 2.42 GiB | RedefinedFieldTraversalSeeds passed across three seeds. |
The manual Win32_OperatingSystem memory probe was denied once before a launch; no Unreal process
was started by that probe. The harness's own native memory preflight succeeded on subsequent
launches. No files outside the workspace were deleted.