mapspinner 0.1.78 → 0.1.79
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/AGENTS.md +17 -18
- package/package.json +1 -1
package/AGENTS.md
CHANGED
|
@@ -5,17 +5,18 @@
|
|
|
5
5
|
SDK changes must be validated against the test suite and verified in both the dev demo (planet.html) and external consumer examples. No changes ship without passing tests.
|
|
6
6
|
|
|
7
7
|
## Architecture (GPU one-fractal)
|
|
8
|
-
-
|
|
9
|
-
|
|
10
|
-
|
|
11
|
-
(`
|
|
12
|
-
|
|
13
|
-
|
|
14
|
-
|
|
15
|
-
|
|
16
|
-
|
|
17
|
-
|
|
18
|
-
|
|
8
|
+
- LIVE height path = **procedural `composeHeight` per-vertex in the VS** (`terrain.glsl`): every vertex
|
|
9
|
+
evaluates broadShapeM + carves directly. There is NO baked height pool on the default path.
|
|
10
|
+
- THC height pool (OPTIONAL, **default-OFF** behind `window.__thc`): **R32F**, **130-texel tiles**,
|
|
11
|
+
**512 layers** (`THC_BAKE_RES`/`THC_POOL_LAYERS`, `gl-render.js`). An O(1) baked-tile fetch that
|
|
12
|
+
replaces composeHeight when enabled. MEASURED NET-NEGATIVE at the deck (cuts VS but pushes the cost to
|
|
13
|
+
FS, full-frame flat-to-worse — 2026-06-15) so it stays off; keep it as a lever, do not default it on.
|
|
14
|
+
(Any AGENTS history claiming "R16F/1024 live" was aspirational doc-drift, corrected 2026-06-15.)
|
|
15
|
+
- Shader precision: **global default `highp float`** (`gl-render.js` hdr = `precision highp float`). The
|
|
16
|
+
earlier mediump-default experiment was REVERTED — any world-scale noise UV (freq up to ~9000) evaluated
|
|
17
|
+
in fp16 scrambled the lattice at close range, and chasing per-site highp islands kept missing sites.
|
|
18
|
+
highp-default kills the whole class in one line; the explicit highp islands left in the shader are now
|
|
19
|
+
redundant-but-harmless. int + sampler2DArray stay highp. fp16 on a planet-scale value collapses geometry.
|
|
19
20
|
- Reduced octaves: `broadShapeM` 14→12, `vtxDisplace` 9→6, single-octave rock detail; tanh ceiling
|
|
20
21
|
`tanh(x/8000)` gives pointed peaks (CLI `shapeReport allGatesPass`). Distance-gated cheap FS far path.
|
|
21
22
|
Tightened LOD (`planet-orchestrator` splitFactor + near-radius). Dead atlas apparatus deleted.
|
|
@@ -33,13 +34,11 @@ camera you can't aim. Use the DATA diagnostics — `__diag.pxPerPoly()` (on-scre
|
|
|
33
34
|
## The terrain pipeline in one page (read this before touching terrain)
|
|
34
35
|
|
|
35
36
|
Earth-scale terrain SDK, WebGL2, served at `http://localhost:8080/` (entry `planet.html`, `server.js`).
|
|
36
|
-
GPU one-fractal: no tile producer. A finer
|
|
37
|
-
LOD is a denser sample of the SAME field. The procedural broadShapeM fractal is the
|
|
38
|
-
path (
|
|
39
|
-
|
|
40
|
-
|
|
41
|
-
relief. Do NOT delete the procedural path while the atlas is opt-in. Recall
|
|
42
|
-
"tv8-atlas-flat-is-flat-normals-root-2026-06-07" + "tv8-reliable-visual-witness-method-2026-06-03".
|
|
37
|
+
GPU one-fractal: no tile producer, no atlas. A finer
|
|
38
|
+
LOD is a denser sample of the SAME field. The procedural broadShapeM fractal is the ONLY render
|
|
39
|
+
path. (The baked-atlas apparatus was REMOVED in prior work — only historical comments remain across
|
|
40
|
+
src/; any AGENTS history describing an opt-in atlas / `__toggleAtlas` / `atlas-bake.mjs` is stale
|
|
41
|
+
doc-drift, corrected 2026-06-15. Recall "tv8-reliable-visual-witness-method-2026-06-03".)
|
|
43
42
|
|
|
44
43
|
THE SPOOL `browser` VERB IS A FULL BROWSER, NOT HEADLESS (user correction 2026-06-11): it drives a
|
|
45
44
|
locally-profiled Chromium with the REAL GPU + real ANGLE backend (witnessed: ANGLE AMD D3D11 -- the
|