build_game.py used to reuse build/<ver>'s generated objects only as a whole and only for generated/<ver>, so a private
lift of a patch change (gen.py --out generated/<ver>-prN, build_game.py --gen) compiled all 421 chunks, about 5 min on
12 jobs, for one changed chunk. main_objects now decides per object: same compiler, flags and rule in the main
build.ninja (with the main runtime/ and generated dir in place of ours), main build up to date, and every file the main
compile read according to its Ninja deps log byte-identical on our side and unchanged since the object was written
(NTFS ChangeTime, st_ctime elsewhere), the compiler binaries too. A file new on our side whose name those compiles
included or probed blocks the reuse. Anything unknown compiles. The build prints "reused N of M generated objects",
--verbose lists the rest with the reason; obj/reused.txt lets a later build delete a copied object before compiling it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The game thread hands its sound commands to the audio thread in lists: 16, of which 4 are always held, so at most 12
wait in the command ring. The game thread submits at most one per update (1.2.1 main+0x9bd144), the audio thread
takes one per 5 ms audio frame (after a list, main+0x883780 goes on past the ring's check). Above about 200 updates a
second the ring stays full. The game's guard (main+0x9bcdc4) counts the updates that find the ring of free lists
empty and exactly 12 lists in the command ring (0x9bce1c cmp w9, #0xc); the count returns to 0 only when a free list
is there. At every multiple of 300 (10 s at 30 fps, 1.25 s at 240) it clears the sound command switch
(main+0x2268bfc) and stops every active voice (main+0x226f29c). Only the switch's constructor sets it to 1
(main+0xf31ce8, from the engine's init main+0xf31530), so the game is silent until it is restarted.
Measured on 1.2.1 (Linux, RX 6750 XT), same save and scripted input, mix dumped without a device: with fps_limit 60,
144, 180, 200 and 201 the mix has sound; with 205, 210, 220, 230, 239 and 240 it goes silent after the boot logo and
stays silent, also at the 130-148 fps of the scene that follows. The guard is read from the code; the switch going
to 0 has not been observed in a run.
The set audiodrain (TOTK_AUDIO_DRAIN_FIX, audio_drain_fix, on by default) sends the audio thread back to the ring's
check after each list, so it takes every queued list in the pass: one site per version (1.2.1 main+0x883780, 1.4.3
main+0x149244), each going to a target the game's own code already branches to in the same way (1.2.1 main+0x883c18,
1.4.3 main+0x149700). The setting is in runtime/cfg.c, totk.ini.example and docs/settings.md.
At fps_limit 240, 120 s runs of one build per game version with the set switched, seconds with sound in the mix: on
1.2.1 5 of 118 with the set off and 108 of 118 with it on; on 1.4.3 2 of 118 off and 110 of 119 on. Not run:
Windows. Not covered: 1.1.0, whose main we do not have, so its table has no site and it keeps the game's code.
Co-Authored-By: Claude <noreply@anthropic.com>
plugin_init.c reads TOTK_PLUGIN_TRACE through cfg_bool, which stops the game when the key is unset and metas[] has no
row. Every test run set the key, so only a plain start hit it. tests/settings_doc_check.py now fails on any
cfg_value/cfg_int/cfg_double/cfg_bool reading without a metas[] row that is not guarded by cfg_get of the same key.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
tools/build_game.py --portable writes build.ninja paths on the tools' drive relative to the kit, also through '..'
(the generated C and objects in build-cache/work), names the compiler, windres and ar the same way (backslashes:
CreateProcess does not find a relative program with forward slashes), and passes clang -resource-dir and --sysroot
relative to the kit, because clang names its own headers by its absolute folder whatever argv[0] says. The Ninja deps
log then holds no absolute path. installer/steps.py build_args passes --portable for the build kit.
Measured on an install moved after a portable build: the rebuild compiled modules.c only, 3.5 s; before, the move cost
a full compile of 430 objects (335 s on 12 cores). Without --portable build.ninja is byte-identical to before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The rebuild for ExeFS patches, code cheats and mod plugins ran the setup window from the setup folder, whose absolute
path build-cache/rebuild.txt held, with the generated C and objects in that folder's temp/. A moved or deleted setup
folder left code mods without a rebuild.
- A new Kit step after the check copies the release folder without TotK-Setup.exe into <game>/build-kit
(installer/layout.py KIT_ITEMS), and every later step runs its tools from there. A newer setup replaces the kit,
the kit's own rebuild keeps it.
- The release's temp folder is <game>/build-cache/work, so the lift cache and the objects stay with the game.
- rebuild.txt names the kit relative to the game folder; runtime/cheats.c and runtime/romfs.c resolve relative
entries against the exe's folder.
- runtime_key drops the lines of clang --version that name its folder, so a rebuild from the kit or a moved game
folder keeps the compile and prepare steps the setup recorded.
- An update over an install of an older setup moves that setup's temp folder into build-cache/work when both are on
one disk.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
tools/exefs_plugins.py ran every plugin's init in one service run and refused them all when one was refused. Now a run
without inits lists the plugin modules, then each plugin runs in service runs of its own (TOTK_PLUGIN_ONLY, the others
stay loaded): a refused svc, a crash (the FATAL line is the reason), a timeout or a failed check refuses that plugin
alone, with its reason in plugins.status; the rest are built in. 4 plugins take 2.3 s.
At start, plugins_start compares each plugin right after its own init and takes its branches out of main's .text before
the next one, so each init sees the game's bytes, as in its service run. Only that init's aliases are flushed: the
start stays at 113 ms for Full Inventory Expansion + Quick Weapon Durability (107 ms before).
The installer's second lift keeps every plugin of the service runs in the layout when one is built: the records hold
that layout's addresses. A refused plugin stays loaded and never starts. A mod with a built plugin and a refused one,
used without its code, runs the built one (mods.c; --probe-mods case T).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After the design in #845 by itaskaev2. tools/exefs_plugins.py writes an accepted plugin's patch set
(recomp/patches.py plugin lines and hook sites, one ordered set mod_<hash> per plugin, several plugins
in one record told apart by their hooks) and a line per plugin of build-cache/plugins.status: built, or
refused with the first reason. It also checks the code the hooks reach (direct branches and calls from
the stubs, impls, callbacks and targets): no svc but OutputDebugString, imports only from a reviewed
list. The init's imports may now include malloc, free, memcpy, memmove and memset, because the init runs
again at every start and its heap blocks exist while the game runs.
The setup's lift step runs the plugin stage when a mod has exefs/subsdk*: a lift with the plugins, a
totk.exe of it in the compile step's folder, the service run, then the lift with the hook sites; a failed
stage warns (warn.plugins_skipped) and installs without the plugins. docs/mods.md describes what works,
what is refused and why, and the rebuild.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The plugin service run loads the modules as boot does, runs sdk's DT_INIT and then each plugin's from nn::init::Start's
module-init callback (rtld initialises the last loaded module first), writes the record and exits before any game
constructor, graphics or audio. The record holds every word the init wrote into a game module, the plugin's image
before and after its init, the process-memory maps, and the svcs and imports it called.
exlaunch's init needs a Switch-like address space: QueryMemory shows the game's own rtld image below main and the
host's ranges as Inaccessible (it takes Code RX runs for modules), and a 64 GiB free pool for its random alias picks.
New svcs: GetSystemTick, GetInfo 65001 (current process handle), and Map/UnmapProcessMemory, which only the service
run serves: an alias is a copy whose changed words go back into the source when it is unmapped. Any other svc, such
as the CreateSession fallback, ends the run with a reason. nn::oe::GetDisplayVersion is answered with the build's game
version. The game calls none of these svcs; the SVC trace test is still one mask test per SVC.
After the design in #845 by itaskaev2, with the init run natively instead of under Unicorn.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Recognize the verified per-frame NVN wait site while retaining replay and shadow-memory guards. Document support and cover wait selection, polls, unrelated calls, fence completion, timeouts, and disabled modes with a deterministic backend.
Validated 1.6 million wait selections, runtime probes, and a two-minute gameplay smoke test in the v0.4.1 test build. The smoke test retained DLAA and display settings on a disposable profile copy; no FPS improvement is claimed.
Load order: mods_exefs_apply ranks each mod's patch sets by its position (rt_patch_rank) and runtime/patches.c fills
the pick byte of every contested site from the sets that are on, lowest rank first; a tie (two files of one mod)
goes to the later file, as emulators apply a mod's files in name order. Reorder and restart, no rebuild. The pick is
redone when an ordered set is switched at run time. Nothing changes for sites one set patches.
Conflicts: one table (mods_conflict) of winner, loser, part (files, data bytes, code sites) and count. romfs.c fills
it for files, mods_exefs_apply for the data bytes two mods stored and for the contested code sites. The log has both
directions per entry and one line per contested code site; each row of the Mods tab lists what the mod overrides and
what overrides it.
Use without its code: a per-mod switch under the row of a mod with a code module or an ExeFS patch file the port
cannot apply (mods_no_code in totk.ini, off by default). On, that code is left out and the rest of the mod is used;
the row says what is left out and why. mods_without_code = 1 holds it on for every mod and now covers the
unsupported patch files too.
--probe-mods: order to winner at contested sites in both orders, the same-mod tie, a switched-off mod, a run-time
switch of the winner, the conflict table for data and code, the per-mod switch.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Taken from PR #790, which carries Eric Dallo's work from his fork (ericdallo/totk-pc-recomp, 6241358f): only its
mods_without_code part, adapted to dev, which has the ExeFS data writes already (#835) and no plugin runner. A mod with
an exefs/subsdk* or a replaced exefs/main stays off as a whole; mods_without_code = 1 under [mods] uses the rest of it
(RomFS files, cheats, ExeFS patch files) without that code. ModInfo.code names the modules that do not load; the
scan logs it, the Mods tab says it on the mod's row and shows the switch once such a mod is there.
Co-Authored-By: itaskaev2 <165683571+itaskaev2@users.noreply.github.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
F6 asks the game for its own autosave, as the sequence node OneShotAutoSaveMgrRequestAutoSave does, after the game's
own check (its wait between autosaves skipped), from the pad read: the game itself makes that request from several
threads. The slot the commit wrote is copied to <profile>/quicksave; a request the game holds back for 5 s is taken
back. F7 holds the window on its last picture, runs the game's load list underneath and, from the confirmation until
the load is over, opens the quick save's files for every read of a slot folder. A toast at the top of the picture says
what happened. The two keys are set on the Controls tab by pressing them. Keyboard and mouse keys only, game 1.4.3.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
reflex_apply_locked read TOTK_FPS_LIMIT with cfg_int, which turns "display" into the setting's default, 60: with
Reflex on, a 144 Hz monitor stayed at 60 fps. It now takes frame_step_limit_hz, the rate the pacer holds (the
setting, the monitor's rate for display, 60 without the real-time step), and checks it on every present, as the
window measures the monitor's rate after the swapchain is made.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The runtime ended the Vulkan pass at every game Barrier. A Barrier that asks only for fragment ordering and texture
invalidation (ORDER_FRAGMENTS | INVALIDATE_TEXTURE) now leaves the pass open when none of the pass's draws can write a
storage buffer or image, and the next draw sees the writes through a by-region barrier in the feedback layout. Off by
default (barrier_merge = 0, TOTK_BARRIER_MERGE). TOTK_TRACE=gfx also prints how many passes ended by a change of targets
and by the game's Barrier, how many Barrier calls were merged and which NVN barrier bits they carry.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
With VK_EXT_device_address_binding_report enabled, AMD's Windows driver (26.10.1) tracks every bound range in a
chained hash map that it searches on nearly every recorded command. The chains grow as textures stream in and out, so
after 5-10 minutes in a busy area recording a submission cost 3-4 times more and an RX 9070 XT fell from 60 to 20-30
fps. The map only serves the fault report after a device lost, so TOTK_ADDR_REPORT now defaults to 0 and =1 turns it
on for a run that chases a fault. ADDR_REPORT is a known key now, so the ini line no longer logs "unknown key".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With it on, a GTX 1050 Ti lost the device at one place in the world, after the validation layer reported a descriptor
pool reset while a command buffer still used it and binds of that pool's sets. The cause is not found and the slower
diagnostic runs did not bring the loss back. bind_inc saves about 1% of the frame, so it waits until the cause is known;
bind_inc = 1 in totk.ini still turns it on.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PAD_TOUCHPAD follows the paddles in input_map.h and reads SDL_GAMEPAD_BUTTON_TOUCHPAD in input.c pad_physical, so
input_capture sees the click when a Controls cell waits for a press (#852 had no pad id to report it). It is unbound by
default; pads without a touchpad never report it held.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
TOTK_BIND_INC had no settings row, so unset meant off. It now has one with default 1, like TOTK_MEMX: 0 or off turns
it off, check keeps the comparison mode. The check mode found 0 differences on 13 scenes (town, fight2, Kakariko,
Great Sky Island, Depths lit and dark, a shrine, rain, a cave, evening, a field, the menus, entering a building).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The ten conflicts in runtime/vk_exec_draw.c were dev's memcmp/memcpy -> MX_EQ/MX_NE/MX_CPY swap on lines task/bindinc
had rewritten. Each keeps the task/bindinc code with the swap applied to its hot-path compares and copies (push and set
signature compares, the shader memo's pipe compare, the pipeline memo and recorded-state compares, the vertex
attribute and stream lists). The check-mode compares that call bi_mismatch keep memcmp as the reference.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The setup wrote the path of its Python inside the AppImage mount into
build-cache/rebuild.txt. The mount is gone once the setup exits, so the
game could not start the rebuild for code cheats and ExeFS patches, kept
its old code, and rebuild.skip then stopped every later attempt. The
command now starts $APPIMAGE --rebuild <install folder>, and a rebuild
that does not start leaves no rebuild.skip (#560, #403)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The conflict in recomp/gen.py was the Usage line and the options next to --v-locals and --v-trust; both are kept, and
v_locals and v_trust are in the Emitter options the cache key holds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
From task/fastrep 18eab597 (TOTK_PIPE_GEN), now under TOTK_BIND_INC: the pipeline memo keeps the pipe_gen its state
was built under; a draw with that generation and the same shaders, primitive, restart, feedback flags and depth mode
takes the memo's state and pipeline and skips the three 360-byte compares (memo, cache entry, recorded state). The
shader memo skips its compare of pipe the same way. check builds and compares as before and counts a state the
generation called unchanged that differs (bi_mismatch logs the first of each kind with its marks and generations).
The census runs only in check mode. Lookout Landing check, 60 s: 67% of draws keep the state, 0 mismatches.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
vk_exec_bind compares before it writes: VkExec.dirt gets a bit per program, uniform, storage, texture, image and
vertex buffer slot that changed, pipe_gen moves with every change of the pipeline state (also depth clamp, raster
discard and the target formats), fbl_serial with the feedback-layout images, OV.gen with the uniform overlays.
vk_exec_draw_new_cb, copy_state and state_reset mark everything. With TOTK_BIND_INC=on or check, bi_census sorts each
prepared draw by what changed since the draw before (nothing, uniform slots only, other slots, another pipeline) and
counts the descriptors to rebuild, the generation moves and the shader, pipeline, vertex buffer and feedback keeps;
a bind_inc line every 300 frames. No draw is prepared differently yet.
Lookout Landing, 4500 draws a frame: nothing changed 8.7%, uniforms only 55.5%, other slots 22.9%, another pipeline
12.8%; 3.4 of 11.2 descriptors a draw to rebuild; uniform overlays move on 2.3% of draws, snapshots 0.1%.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
K = 8 cut guest time 3.9-6.2% in the town on a 7900X against K = 0. func.py tests the same default; --v-locals 0
gives the previous output.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
recomp/vlocals.py rewrites each emitted function: c->v forms become the RT_VLOCALS accessors of runtime/simd.h, the
K registers the function names most get bits in VMASK, they are loaded at entry, stored where the dirty mask may hold
them before calls and exits, and reloaded after calls; --v-trust skips the reload of V8-V15. A function with a form the
rewrite does not know keeps every register in the Cpu. K = 0, the default, emits the same C as before. The manifest
counts the rewritten functions and syncs. tests/diff/func.py takes --v-locals, --v-trust and --work (one work
directory per concurrent run).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The runtime reads TOTK_SHADOW through cfg_get, so `shadow = 1` in totk.ini already turns shadow storage on, but the
key was missing from cfg.c known[]: every such totk.ini logged "unknown key 'SHADOW'", which reads as if the line did
nothing. The docs only named the environment variable.
known[] now has SHADOW, totk.ini.example describes `shadow` next to sync_ahead (which needs it), and the sync_ahead
text, its log line, docs/building.md and docs/architecture.md name the totk.ini key next to TOTK_SHADOW. Behaviour is
unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The ability guard (runtime/ov_ability.c) checks the holder, the actor, its component and the reference word the game
writes, so it catches every bad target seen in the reports: -1, 0x80, an empty bounding box (FLT_MAX) and a vtable in
main's data. The optref site at main+0x1bd01c8 (1.2.1: 0x1622bf4) is dropped: set ability patches the same address and
must win there.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The setup wrote 64 patch folders: every prompts style for all 15 shaped game resolutions. The new shapes step
(installer/screens.py) writes romfs-patches-<w>x<h> only for the shapes of the connected displays, the output_res the
game defaults to on each and the totk.ini output_res, and the prompts' shaped folder only for the style totk.ini
selects. One folder per shape: the runtime maps the game resolution to its shape's size (cfg.c cfg_shape_size, the
closest ratio within 1%), and an exact-size folder from an older install still wins. The settings menu marks a game
resolution whose shape has no folder and can start the generator for it in the background, for the next start.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Pull request #1136 with three fixes: the Russian and Portuguese rows name Ultrahand as the catalogs do, smooth map
zoom takes the wheel only on the map menu_cursor.c finds, and the weapon bar opens only where the camera hooks saw
free roam, a lock-on or a mount. Off by default (gameplay_tweaks_enabled = 0).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The holder of the PlayerVirtualKey request can be freed while the request still points at it; its first word then reads
as a small number and the game dereferences it as the target actor. No guest object lives below 2^32, so such a target
now takes the game's own branch for a null target. Same TOTK_OPTREF_GUARD switch as the existing optref site.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>