diff options
| author | Valentin Popov <valentin@popov.link> | 2026-07-18 18:17:38 +0300 |
|---|---|---|
| committer | Valentin Popov <valentin@popov.link> | 2026-07-18 18:17:38 +0300 |
| commit | a339a3a6359b27352b5d2f5e93df5b073190e16b (patch) | |
| tree | cc1f6650c4d6c25340ff7897e32e2b4b3875630e /docs/tomes | |
| parent | 273bd9e477fc5a3a87a38b20e53c46187c666def (diff) | |
| download | fparkan-a339a3a6359b27352b5d2f5e93df5b073190e16b.tar.xz fparkan-a339a3a6359b27352b5d2f5e93df5b073190e16b.zip | |
feat(tools): summarize terrain shade profiles
Diffstat (limited to 'docs/tomes')
| -rw-r--r-- | docs/tomes/05-render.md | 12 |
1 files changed, 12 insertions, 0 deletions
diff --git a/docs/tomes/05-render.md b/docs/tomes/05-render.md index e0970c7..faf9dfc 100644 --- a/docs/tomes/05-render.md +++ b/docs/tomes/05-render.md @@ -1750,6 +1750,18 @@ state, not a persistent cache identity or a WEAR row. The observation narrows the next terrain task to profiling the cache's generated entries and their consumer; it still does not establish a Land1 blend equation. +The probe now also reads the bounded eight-byte entry table and only hashes +the fixed 212-byte profile banks it observes; raw process bytes never enter +the repository. The same AutoDemo cache had 3,426 materialized record pointers +among 3,462 entries and referenced banks `0..4`. Two consecutive captures +kept that count, bank set, and all five FNV-1a-64 bank hashes stable: +`9426539192111566946`, `2915543853261903601`, `13156371703608727216`, +`11280405573172155968`, and `18234288329145104612`. Their result views changed +from `0x008A565C` to `0x008A55A8`. Thus the generated cache/profile state is +stable over these render samples while the shared result view is not. This +still identifies neither the record-field semantics nor the final terrain +blend/pass operation. + ```powershell powershell -NoProfile -ExecutionPolicy Bypass -File ` tools\capture-terrain-shade-cache.ps1 -ProcessId <iron_3d-pid> |
