Symptoms: stretched slabs over the ground in the Depths from a VRAM mirror copied from the shadow too early

This commit is contained in:
ipavlin98 committed 2026-11-18 02:23:21 +03:00
1 parent 0f2cc1071a
commit 07f9760971
1 file changed
+1
+1
View File
@@ -18,6 +18,7 @@ GPU semantics behind the graphics rows: [maxwell.md](../maxwell.md). The method
| dark or magenta 16x16 blocks, missing characters | a pass samples its own render targets (G-buffer feedback) | feedback-loop layout and barriers on that pass |
| tile rectangles, wrong data in tiles | storage buffers bound in the wrong slot order | `passes.txt`: storage buffers with their c0 slots |
| streaks on cliffs, characters stretched into streaks | primitive restart ignored; integer vertex inputs against a float format | validation run; the draw's topology and vertex formats in the dump |
| huge stretched slabs over the ground, a ceiling over Link, the ground up to his chest, from the load on for the whole session (the Depths); some runs only, most at a high frame rate; with `sync_ahead`, `replay_workers` and a discrete GPU | a VRAM mirror copied from the shadow before the CPU's write reached it, and kept the old vertices while marked synced (fixed: `vk_memory.c` `vk_mem_mirror_seal`) | `TOTK_MEMORY=verify` mismatch lines; `memory = import` or `replay_workers = 0` makes it go; `fps_limit = 144` shows it in most runs |
| garbage geometry on title or load screens | 8-bit indices widened on the CPU before a copy in the same submission wrote them | whether a copy to the index buffer precedes the draw in one submission |
| black tiles, masks not cleared | clears with a partial channel or stencil mask skipped or done in full | `passes.txt` clears and their masks |
| garbage in feedback effects at first use | new GPU-only images start undefined; the console starts them at zero | clear on creation and compare |