Fr0zkaandClaude Opus 5.5
044cd6d9c5
Reach = 0.75 shell + 0.21 relief + 0.50/6 smooth-max support + 0.01 = 1.0533 m, used by proof, feature, fallback, off-grid bounds and the sweep cutoff. Removes the dead 4 m term (v225 added metres to core-voxel coordinates). Captures and mouth close-ups byte-identical; proof audit 0 misses over 25,728 shell/crest samples; floor 7.884 mm; seams, walkability, fast tier, Clang/GCC W1/W8/W12 green; timing equal within noise. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016dhxM43dnWvUu2B5Ph1iG1
1.8 KiB
Linux130 — passage reach
On wip/l130-reach-units from b96dd73, passage influence is now 1.0533 m: 0.75 m shell + 0.21 m relief + 0.50 m smooth-max support (width/6) + 0.01 m margin. One metres-to-core-voxels conversion feeds proof, feature, fallback and off-grid bounds; the sweep cutoff uses the same metre value. The dead 4 m cutoff is removed. Evaluator version is 227.
The debug shell/crest audit sampled 25,728 points, including 5 mm inside and outside the reach; 18,993 had sign-relevant passage contribution. Proof and feature bounds missed none. The nine-station floor probe still peaks at 7.884 mm, matching L125. The 23 L125 views and six L129 mouth close-ups are byte-identical. Fast verification, walkability and shell audits pass; the seam audit found 0 mismatches across 4 pairs / 4,900 shared samples.
Clang/GCC W1/W8/W12 hashes agree across all 18 runs. Versus v225, hashes changed for owner4 full-detail, whole-view and LOD0–3; owner2 surface-world and floating-islands whole-view and LOD2–3. Exact-query hashes stayed unchanged. The v227 canonical baselines are recorded; changed hashes reflect the shorter density influence halo, while sampled sign-relevant bounds and captured surfaces remain covered.
L80 timing used 10 interleaved pairs per stratum/worker count. Across the eight case medians, the weighted deltas were +0.33% for LOD0–2 and +0.31% whole-view. Across 80 paired runs, mean deltas were +0.26% (95% interval −0.44% to +0.96%) and +0.47% (−0.06% to +1.00%); both intervals include zero. No measurable timing change; UE movement was skipped. The Linux UE host build succeeded in 76.30 s with -NoUBA -MaxParallelActions=4, without prlimit. ARCHITECTURE.md and CODEMAP.md had no note for this optimized influence reach, so they were unchanged.