TOTK_REC_CHECK=1 searches the packets of the part being recorded for each image, view and buffer vk_memory.c defers or
destroys. A deferral stamped below the part's value, or a destruction of an object a packet names, is logged with the
packet op, both stamps and the GPU's value, and counted by kind and op in a summary at exit. put() zeroes each packet's
room while the check is on: packers reserve room for optional arrays they may leave unwritten, and those words held
handles from the ring's previous lap. TOTK_REC_DELAY_US=N makes the recorder wait N us before each batch, as on a slow
machine. rec_begin takes the part's value for the check. Both are off by default; off, the recorder pays one test per
batch and per packet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
vk_memory.c defer stamped every deferred image, view and buffer with the last submitted timeline value. An object the
replay lets go of while it records a submission can be named by that submission too: vk_texture.c slot_resolve hands a
draw a released entry, and entry_of destroys released entries at that address later in the same submission. The
completion thread then destroyed the object once the GPU reached the earlier value, before the submission naming it
ran, and with the recorder thread behind, before the recorder made the call. The stamp is now the later of the last
submitted value and vk_exec_recording_fence, as LAS Recomp and Savage Wind Recomp stamp their deferrals
(runtime/nvn/vulkan/textures.c defer). TOTK_DEFER_STAMP=0 brings back the old stamp.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
-march=x86-64-v2 does not reach inline assembly, target attributes or intrinsics, which is how kern_ap_stub's YMM
saves got into libtotkrt-v2.a (issue #454). tools/check_isa.py disassembles objects and lists every function with
VEX/EVEX, BMI1/2, LZCNT, MOVBE, ADX, SHA, AES, PCLMUL or RDRAND/RDSEED code or YMM/ZMM/mask registers, except
functions in its DISPATCHED list (sha256.c blocks_ni, behind __builtin_cpu_supports). build_game.py --cpu v2 runs it
over the runtime objects after compiling and before linking or archiving, so make_release.py's v2 library build fails
too. Objects unchanged since the last clean scan are skipped; a full scan of the runtime takes about a second. The
release ships the script, and the installer's runtime fingerprint includes it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The asynchronous preemption stub saved and restored YMM0-15 with vmovdqa and ran vzeroupper in inline asm, which
-march=x86-64-v2 does not reach. The v2 runtime library therefore carried VEX code, and a CPU without AVX stopped
with an illegal instruction the first time a core owner was interrupted, a few seconds into boot (issue #454,
Pentium Gold G6400, totk+0xe7a2b50 = kern_ap_stub+0x20). Without __AVX__ the stub now saves XMM0-15 with movdqa into
the same slots; v2 game code keeps nothing in the upper YMM halves. The v3 stub is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
AMD buffer marks at every replay pass and dispatch, on by default (fight2 off/on x2: 89.45 vs 89.6 fps, no new slow
frames); NVIDIA diagnostic checkpoints with TOTK_PASS_CRUMBS=1, compiled but not run here.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The replay marks the start and the end of every render pass and dispatch
(vk_exec.c vk_exec_pass_begin and vk_exec_pass_end, vk_exec_draw.c
op_dispatch and op_dispatch_ind), outside the pass so no pass is broken.
With VK_AMD_buffer_marker the GPU writes the last started and finished mark
into words 2 and 3 of the breadcrumb buffer; with
VK_NV_device_diagnostic_checkpoints the same points are checkpoints, read
with vkGetQueueCheckpointData2NV after the loss. A ring of 4096 records maps
a mark to its submission, index, target formats and size, draw count and the
input ids of the first and last draw's shaders. vk_device_fault_report logs
both marks with their records. The calls go through the recorder thread.
TOTK_PASS_CRUMBS (menu: GPU pass marks): unset, AMD buffer marks only; 1,
also NVIDIA checkpoints, whose extension is enabled only then; 0, none.
The design follows Savage Wind Recomp 3fae40c; the code is ours.
--probe-vkexec gets a pass marks part: pass, dispatch, pass in one
submission; the marks read back after the fence name the last pass and the
report text says so.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Zero-byte NVN pools get no Vulkan record, WaitSynchronization keeps a pending cancel for the next blocking wait,
SignalToAddress subtracts 2 when it wakes all waiters, and the 1.2.1 texture loader widens pool offsets unsigned.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The loader keeps a texture's place in its pool as a 32-bit word (main+0xad4294) and widens it with a sign extend for
nvnTextureBuilderSetStorage's offset (main+0xad43f0) and for SetPackagedTextureData's source (main+0xad440c). Past
2 GiB the offset turns negative and the storage lands 4 GiB below the pool (MacNativeGames ca311e2d, seen with a 4K
texture pack's heap table in a 3.58 GiB pool). The shipped heap table keeps every pool far below 2 GiB, but
TOTK_APP_POOL raised with such a pack's heap table (RomFS mods replace it) can reach it. The new patch set pooloff
(TOTK_POOL_OFFSET_PATCH, on by default) widens unsigned: the same value below 2 GiB, right up to 4 GiB.
1.4.3 needs no patch: both of its loaders (main+0x583c34, main+0x27ed214) keep the offset in 64 bits (str x25 / mov
x2, x25 / add x1, x25, x8). gen.py accepts both original words, and the lifted C at the sites reads the offset as
uint32_t with the set on and as int32_t with it off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
SignalAndModifyByWaitingCountIfEqual with count <= 0 and at least one waiter stored value - 1; Horizon stores
value - 2 (mesosphere KAddressArbiter::SignalAndModifyByWaitingCountIfEqual; yuzu, LAS Recomp and nxnative agree).
The word was already updated with compare-and-swap, as LAS Recomp 1144aea made theirs; the count > 0 values were
right.
--probe-kernel arbiter part expected the old value: it now wants 4 for value 6 with one waiter and count -1, and
4 for two waiters and count 0. Before: 5 for both. After: ok.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Ours took a pending CancelSynchronization first, before the signalled objects and before the timeout-0 return, so a
poll or a wait on a signalled object used it up and the blocking wait it was meant for ran on. Horizon checks a
signalled object, then timeout 0, then the cancel (mesosphere KSynchronizationObject::Wait; LAS Recomp 1144aea moved
the poll the same way): both keep it pending. A cancel that wakes a blocked wait still ends it cancelled:
CancelSynchronization now wakes it with R_CANCELLED, which the loop tells apart from a signal's -1.
--probe-kernel events part: a cancel sent before a poll and before a wait on a signalled event; each is followed by a
blocking wait with a 500 ms timeout. Before: the poll and the signalled wait returned 0xec01 (cancelled) and the
blocking waits timed out (0xea01). After: timeout, success, then cancelled both times.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
vk_pool_init gave a size-0 pool with storage a record: the host import of 0 bytes failed (the driver refuses the
memory, VK_ERROR_INVALID_EXTERNAL_HANDLE) and the record went into the pool index as the empty range [mem, mem]. A
lookup at that address then found the empty pool before the pool that starts there and reported it unbacked. TotK
makes such pools now and then (MacNativeGames d3407e29, 1.2.1). Such a pool now gets no record, as a GPU_NO_ACCESS
pool; vk_pool_fini finds none and does nothing. Removal by key already kept a neighbour's range safe at Finalize.
--probe-vkexec gains an empty pool part: a zero-byte pool at the start of another, lookups there before and after
its Finalize. It failed before ("a lookup at the start of a pool finds the empty pool there").
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
LASRecomp, MacNativeGames, d2r-mobile, nxnative and SavageWindRecomp moved; SwitchRecomp and deko3d did not.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The CPU fills these chunks and the zero buffer and never flushes them. On Intel integrated GPUs (ANV on Meteor
Lake) the host-cached memory type that VMA picks for random host access is not coherent, so the GPU read what
an earlier submission had left there. A lighting fragment shader got another draw's UpdateUniformBuffer payload
as its step count (0x3f800000 instead of 96 or 128), looped about a billion times and lost the device; other
stale payloads drew the black frames. Requiring HOST_COHERENT keeps the same memory types on GPUs where host
memory is already coherent.
TotK's terrain shaders fill the gradient registers of a texture-with-derivatives instruction per coordinate (du/dx, du/dy, dv/dx, dv/dy), while the translator takes the first two registers as dPdx and the last two as dPdy. The mip level then came from the wrong derivatives and was too sharp on terrain faces turned across the screen, so patches that followed the faces crawled when the camera moved.
After a shader loads, every two-dimensional gradient pair has its inner components swapped by two new composite constructs, so the shader cache is untouched. TOTK_TXD_TRANSPOSE=0 turns it off. The finding is row 2.13 of docs/maxwell.md.
With dlaa = 0 and dlss = quality left from the menu, the scene still rendered at 67 % and the plain upscaler
stretched it, so the picture looked blurrier with DLSS switched off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
advance_voice computed fraction + seconds * rate * pitch while voice_render adds rate * pitch / 48000 per output sample. The two rounded apart on most frames at 32 kHz and 44.1 kHz, so the voice jumped one sample ahead of what was rendered. audio_voice_consumed now runs the mixer's own loop.
TOTK_AUDIO_DUMP=<path> writes the sink mix before the clamp and logs how many frames were rendered under 1 ms after the previous one; a one-time log line reports the first enabled AUX effect.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The loader asked for E1, the container byte of a plain NDEF tag, so it refused every real amiibo dump as
"Not an amiibo dump". --probe-services now loads a dump with an amiibo header and refuses the same file with E1.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
plat_fs_kind served every path through FindFirstFileW, which fails on a path with a trailing separator and reads
wildcards. GetFileAttributesW decides again, as before; only reparse points need the find data's tag.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The game destroys the nfp activate and deactivate events and attaches the same SystemEventType again before each
detection. kern_bind_system_event kept the first writable handle per guest address and returned early, so the second
Attach left the event destroyed and the game's next TryWaitSystemEvent aborted (issue #398, equipping the Amiibo
power). A kept handle now counts only while the SDK's state byte of the event is nonzero; otherwise the event is
created again and its entry takes the new handle, and a signal to a destroyed event is dropped. probe-services
destroys and attaches both nfp events again and checks that they work.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mods were described only in one dense list in the tester guide. docs/mods.md walks a player through installing a mod,
its order and switches, TKMM merges, code patches and their rebuild, what renaming a mod folder does, the reasons the
Mods tab gives for an unused mod, and what to check when something looks wrong. docs/cheats.md gains a folder example
and the rename note.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The lift key hashed each cheat and ExeFS patch file by its name and content only, but tools/cheats.py and
tools/exefs_patches.py name every patch set after "<mod folder>/<cheat or file>". After a mod folder was renamed,
totk.exe saw none of its code cheats in the build and started the rebuild, the setup found the lift current and skipped
it, and the cheats stayed "after the rebuild" for good. The key now takes each file's path under mods/ as well.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Linux release builds against the Steam Runtime's glibc 2.31, which lacks that 2.34 call: plat_restart and
plat_spawn stopped the runtime library from compiling. A weak reference uses it when the running glibc has it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
_putenv_s in cfg_probe.c and _stricmp in the menu's key binding row stopped the Linux runtime library from compiling.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A 0.2.8 install's romfs/index.json had two flipped bytes, and the update stopped with an unexpected JSONDecodeError
instead of extracting again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>