Merge the 0.6.1 fixes from the 0.6.1.x test builds.

This brings in the test branch up to the shader reads, not the texture eviction by disuse still in progress there:
- black frames: a released or taken-back texture keeps its content, and a new texture over a live one of another size gets the region both cover;
- the menu font atlas uploaded through the runtime's checked upload batch;
- the game held to 240 frames a second, which stops the quick menu's crash;
- the FSR 3.1 contexts evicted instead of turning FSR off after a few render scales;
- AMD frame generation kept off while NVIDIA's holds the device;
- less device memory held that only costs the system's commit;
- shaders read from the disk cache off the render thread past a small budget;
- the exit code 70 text;
- pull requests #260, #261, #270 and #271.
The tool lists in make_kit.py and make_launcher.py and the setting names in cfg.c take both sides: romfs_check.py and pack_shader_cache.py, FOG_CULL, SCENE_SHARE and GPA_TWIN_MAX_KB.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
chrg6
2026-11-09 14:39:42 -05:00
29 changed files with 637 additions and 135 deletions

View File

@@ -101,6 +101,11 @@ keeps per-core data without locks and relies on that, so the host models it (`ru
translates, translates every stored program again for the next start, then deletes the old translations (a pruned
copy of `shader-cache.pack` replaces the pack at the next start). A program the new translator refuses keeps its
old translation. `--prepare-shaders` still translates everything with the current translator.
- Shader disk reads in play: a draw's shader that is in the disk cache but not in memory is read on the render thread
within a budget of 1/8 of the wall time (2 ms at once), and past it by three loader threads while its draws are
skipped as for a translation (`runtime/gfx_shader.c` `load_budget`). Opening a loose file is what costs (5.6 ms a
pair the first time on Windows), so an update that carries the shaders over packs the cache once it holds 1,000
loose translations (`tools/make_game.py` `PACK_LOOSE_MIN`).
- Mesh and task shaders run as `VK_EXT_mesh_shader`; indirect mesh draws read NVN's `{count, first}` record through a
ring of Vulkan `{x, y, z}` records filled on the GPU.
- Pipelines: `VK_EXT_graphics_pipeline_library` where the driver has it: stage libraries compiled ahead, a fast link

View File

@@ -311,7 +311,8 @@ pruning, unsafe archives, the background update (a release staged beside the run
`staged.json` marker, a staged release installed after a restart, no install while the game runs, the install once
it exited, nothing by itself with Install Updates Automatically off, a failed staging tried again after the next
check), the font decoding on a system font wrapped like the game's, `make_game.py` on a synthetic
`.nup` (events, skipped steps, `--rebuild` without the dump, step keys), the RomFS step beside the lift and compile
`.nup` (events, skipped steps, `--rebuild` without the dump, step keys, the shaders key following the translator the
build uses), the RomFS step beside the lift and compile
and its failure stopping the build, the prompts step with a stand-in prompt tool (when it runs again, and that it
makes no other step run), the button art a release copies and the packages and tools the prompts need in a release,
the compile step's object cache with the system's C compiler and its

View File

@@ -42,11 +42,15 @@ cards).
- The device-lost crash at area loads on GeForce RTX cards (seen on the 4090, 5090, 5080 and 5070 Ti), most often when
the system's memory ran low: a released texture could be read after it was freed.
- A crash at start with code 71 when RivaTuner Statistics Server runs.
- A crash when Frame generation was switched from NVIDIA DLSS to AMD FSR 3.1 in the menu: that change now takes effect
at the next start, as the menu says.
- A crash in Intel's Windows driver while the pipeline cache was saved during play.
- Long sessions running out of memory: idle textures are now freed when the system's memory runs low too, not only the
video memory.
- Shader preparation no longer crashes the RADV driver (AMD GPUs on Linux) on a texture size query.
- Things that moved no longer leave a trail when neither DLSS nor FSR 3.1 is on.
- FSR 3.1 no longer turns itself off for the rest of the session once the render scale has been changed a few times
in the menu.
- With frame generation on, the mouse turned the camera as fast as if the game drew every frame on screen: twice as
fast with FSR 3.1 frame generation.
- The game's own Restart no longer brings the launcher back over the game, where Play could start a second copy.