Commit graph

7 commits

Author SHA1 Message Date
54e267dfa6 render3d: ground_fill's conformance test and the generator of its golden
tests/ground_fill_test.ludic compares gg_json (gf_hash over fixed inputs, Math.lerp cases whose order
matters, the yaw as model.vert turns by it, a hand-written density, one cover chunk with a clearing and
one solid chunk) line for line with tests/ground_fill_golden.json, and checks ground_fill draws exactly
the candidates and a solid layer's far bands a subset of band 0. The golden is written by
tests/gen/ground_fill_golden.ludic, run from the repository root; it is not committed here.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 15:24:23 +03:00
c18579f777 tiles_open_test: "nothing on a device" asks what terrain_from_baked would have made - the coarse height and normal textures and the page pool - not gvk_frame_no, whose default is 1 (env.ludic), not 0: the open path ticks no frame
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 03:23:36 +03:00
42147937b2 render3d: terrain_tiles_open_file(path, key, version, half, ox, oz) - a map's baked tiles opened for the CPU questions alone (terrain_height, _file, _smooth, terrain_ortho), with no device and no renderer booted: terrain_from_baked's opening without its coarse level on the GPU. False and nothing changed for another bake's key or version, a missing file, or a whole copy already kept (it answers first). tests/tiles_open_test: a bake written with LBAK's header answers its heights and photograph at a map origin of (300, -120), nothing on a device; the refusals leave no file open, and a bake opened after them answers. The hand-written tiles move to tests/fakes/tiles_file.ludic (payload, raw file, bake, the expected answer at an origin, the cache's snapshot), shared with height_file_test
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 03:13:18 +03:00
130b9091d1 render3d: the terrain's questions never page - ter_h and ter_o read a resident tile in place and any other through the file scratch (heights and now the photograph, four tiles each, sized when the tiles open, forgotten when they close), so terrain_height, _smooth, terrain_ortho, the chunk heightfields, ground_fill, grass, water and shadow no longer take slots or move the clock. Only the loaders page: tp_fill reads through tt_h_load / tt_n_load / tt_o_load, and terrain_tiles_prefetch as before. terrain_height_file is now terrain_height, kept as the name callers wrote against. No answer changes: every read was already the whole copy's texel. tests/height_file_test: heights and photograph equal the texels written, resident or not; a whole-map sweep of the questions, and a ground layer's fill over the map (a one-layer LGD2 written in the test), leave the slots, their contents, the clock and the read counts unchanged
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 02:43:17 +03:00
c2fc8f78f2 render3d: terrain_height_file - terrain_height's exact bilinear that never pages: a resident tile is read in place, any other from tt_file into four tiles of scratch of its own (in the state's defaults, forgotten when the file closes), so no slot is taken and no recency bumped - a sweep at boot no longer changes which tiles the cache holds, and so what is drawn from it. With the whole copy it is terrain_height. tests/height_file_test: equal to terrain_height over a grid across many tiles, resident and not, at tile corners and past the edge; a whole-map sweep leaves slot_of, tile_in, ref, the heights, the hand and the read counts unchanged
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 02:29:55 +03:00
9f07fc8a62 render3d: heightgen's plain and SMOOTH variants compiled (61 programs) - a smoothed or noise-only terrain had no height field on Vulkan; run-time defines as pure functions (program_defs.ludic) and manifest_test holds every program the source can ask for, literal or computed, to the manifest
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 22:30:35 +03:00
c3fc896dea render3d: a renderer that cannot draw says so and stops - LUDIC_HOME's leading ~ expanded (an unexpanded one left a run without shaders, logging 27 missing variants and drawing on), the SPIR-V manifest ends with E <count> and a cut or miscounted one is refused, and every such failure is a fault: r3d_open / r3d_load_step fail on it, r3d_fault() / r3d_fault_exit(code) for the game; tests/manifest_test
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 22:12:39 +03:00