Commit Graph
26 Commits
Author SHA1 Message Date
Valiantsin DzerakhandClaude Opus 5.5 41f201ad39 Skip full-screen draws while their pipeline builds by default (TOTK_PIPE_SKIP=fullscreen): in a fight a full-screen pass whose stage libraries were cold in the driver cache held three frames in a row at 64-72 ms, each spending the 50 ms wait budget, and some draws were skipped after all
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-28 15:02:18 +03:00
Valiantsin DzerakhandClaude Opus 5.5 bac4d6cf91 The shader memo skips its 1 KiB handle compare while no texture or image was bound
memo_match compared the vertex and fragment stages' 64 texture and 64 image handles on every draw. Every OP_BIND_TEX and now every OP_BIND_IMG bumps tex_serial; a key whose serial is current has the same handles, and a key that matched by the compare takes the current serial. memo_match went from 0.95% to 0.27% of the fight replay.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 15:54:00 +03:00
Valiantsin DzerakhandClaude Opus 5.5 c9c06f2501 Render target views of registered textures live as long as their image
vk_exec_pass_begin created every attachment view again in each submission and freed them at its fence: 2% of the fight replay on the 7900X, more on drivers with a slower vkCreateImageView. view_of now takes the view from vk_exec_target_view, which keeps it in the sampled-view tables under the texture lock, and vk_exec_image_gone releases it with the image. Images that are not registered textures keep the per-submission views.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 15:53:59 +03:00
Valiantsin DzerakhandClaude Opus 5.5 97d3762ba1 Strip milestone and task labels from code comments and drop the dead gfx_stubs.c
Comments named the agent task (T1-T7, M2-M5, Task 2.x) that once owned each file or block, and pointed at local
notes the repository does not carry (backlog items, milestone plans, research notes). They now say what the code
does. gfx_stubs.c held weak stand-ins for functions that every build has defined for a long time.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 13:12:50 +03:00
Valiantsin DzerakhandClaude Opus 5.5 2bcba6bc3f Keep which shader of a program serves each stage per program serial instead of reading every stage's control block several times a draw
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 01:01:54 +03:00
Valiantsin DzerakhandClaude Opus 5.5 bcbbcb0b1a Keep texture lookups per slot instead of per global texture generation, and vertex buffer spans while their ranges repeat
TotK finalizes and initializes transient textures every frame: the global tex_gen and slots_gen moved about every 16
draws in the monsters fight, so every cache keyed on them (the slot cache in vk_texture_by_id_locked, the texture types
of answers_for, the texture descriptor memo) lived for a few draws. A slot resolves through the entries at its own
storage address only, so each address hash now has its own generation and each slot a token
(vk_texture_slot_token: registration serial and address generation, lock-free). The memos check the global generation
first and the slot's token only when it moved.

The descriptor memo no longer drops an entry at each submission (the first binding still makes the storage check; a
re-upload writes the same image) or pass (the feedback layout is set on each hit from the kept image), and is twice
the size. feedback_scan keeps each handle's sampled subresource across passes instead of looking up every fragment
texture after every texture bind. bind_vertex_buffers keeps each stream's resolved span while its range and
vk_mem_gen stay the same.

Monsters fight on the RX 9070: replay 12.0 -> 10.8 ms per frame; field 8.1 -> 7.4 ms. TOTK_TEX_MEMO=check and
TOTK_SHADER_MEMO=check runs found no difference.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 00:55:17 +03:00
Valiantsin DzerakhandClaude Opus 5.5 6b65fb6c0e Keep sampled-texture descriptors between draws while the slot, sampler, submission and feedback pass stay the same
tex_info resolved the slot, built a view key, probed the view table and looked up the sampler for every texture
binding of every draw. A direct-mapped memo keyed by handle, resource type and separate returns the descriptor while
vk_texture_slots_gen, vk_sampler_gen, the submission epoch and (with feedback-loop images) the pass are unchanged;
storage images are not kept. TOTK_TEX_MEMO=0 turns it off, =check recomputes every hit (a fight run with it and
TOTK_SHADER_MEMO=check stopped on nothing). Against main, profiler on, no bind stats: fight replay 13.9 to 11.9 ms per
frame, render thread busy 16.3 to 14.6 ms, CPU-bound windows 15 to 9; field replay 9.5 to 8.2 ms.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 22:50:22 +03:00
Valiantsin DzerakhandClaude Opus 5.5 1836d5c5d0 Keep texture image views for the image's lifetime instead of creating them again in every command buffer
Views made per command buffer were created anew every frame: vkCreateImageView under get_view took about 8% of the fight
replay on the 7900X. They now live in one table keyed as before, used under the texture lock, and destroy_tex hands an
image's views to the deferred release with the image (vk_exec_image_gone). The buffer range limits ubo_info and
ssbo_info read on every binding are read once. Fight replay 14.0 to 12.5 ms per frame (15.7 before the branch), field
9.5 to 9.2 ms; a fight run under the validation layer shows only the known cp_t_tcb_20 message.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 22:42:47 +03:00
Valiantsin DzerakhandClaude Opus 5.5 b273005901 Look up texture types only for the slots a shader's translation reads, and reuse the previous draw's pipeline while its state repeats
answers_for looked up the type of all 64 texture and image handles of a stage whenever one changed; the use mask of the
first shader of each program stage now limits it to the slots that select the variant, with a generation per slot and
the texture lock taken only on a miss (texture type recomputations per draw 0.39 to 0.05 in fight). prepare_draw keeps
the previous draw's pipeline info while the pipeline state is equal (TOTK_PIPE_MEMO), and reads the clock only for
draws that may wait. Fight on the 7900X: replay 15.7 to 14.0 ms per frame, 53.3 to 58.7 fps; field 11.1 to 9.5 ms.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 22:31:53 +03:00
Valiantsin DzerakhandClaude Opus 5.5 c4108c49a8 Map NVN's 16-bit packed formats with their components counted from the least significant bit, as NVN does for RGB10A2: RGB565 is B5G6R5, RGB5A1 is A1B5G5R5, and the old table, which took LAS Recomp's order, read the Depths terrain's normal and blend maps scrambled and stretched the ground textures along one axis
A1B5G5R5 needs VK_KHR_maintenance5, so RGB5A1 stands in as A1R5G5B5 with R and B swapped by the view swizzle (TexVkFormat.chan); RGBA4 and RGB565 fall back the same way to mandatory formats on devices that lack A4B4G4R4 or B5G6R5. RGB5 and BGR5 now sample alpha as 1.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 16:48:05 +03:00
Valiantsin DzerakhandClaude Opus 5.5 6029719713 Read depth textures sampled without a compare as one texel, and round the reference of D16 array shadow compares to 16 bits, as MacNativeGames 6283599 does: the game binds linear samplers to depth it reads as data, and the Switch compares D16 against a reference at the texture's precision
The point form of a sampler slot is built once (vk_sampler_point_by_id_locked). Translator patch 0009 rounds the 2DArrayShadow reference in Ryujinx's SPIR-V backend.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 16:25:56 +03:00
Valiantsin DzerakhandClaude Opus 5.5 05f3643dc0 Bind storage buffers whose address the shader loads from a user uniform block: the Depths terrain reads its vertices through one (about 500 draws a frame in the G-buffer, the light probes and the shadow cascade) and drew nothing with the zero buffer, so the ground was missing and rocks lost their bounce light
ubo_ptr reads the address at bind time from the latest UpdateUniformBuffer of the submission that covers it, else from memory, since the terrain rewrites one block before every draw and guest memory gets it only at the write-back. bind_sig compares the address on every draw. MacNativeGames resolves these the same way (NVN_RB_SBUFU, pend_read).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 16:25:46 +03:00
Valiantsin DzerakhandClaude Opus 5.5 7ceda9475b Size a pass without attachments by the far corner of viewport 0, else of the scissor, and reopen it when a draw needs more area, as MacNativeGames 62da927 does; an offset viewport no longer falls outside the scissor
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 14:40:00 +03:00
Valiantsin DzerakhandClaude Opus 5.5 b0805112ed Merge main into task/menu: the docs and comment cleanup, with the Dear ImGui set-up of menu.cpp kept in the LAS Recomp notices row
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 10:16:23 +03:00
Valiantsin DzerakhandClaude Opus 5.5 8b514f6ead Keep every default in the settings table: cfg_value, cfg_int, cfg_double and cfg_bool read it, and the consumers of table keys drop their own defaults
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 09:32:30 +03:00
Valiantsin DzerakhandClaude Opus 5.5 111f748c51 Remove citations of local build/ notes from code comments and correct the TOTK_REPLAY_ASYNC default in vk_backend.c
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-26 09:23:21 +03:00
Valiantsin DzerakhandClaude Opus 5.5 469c2be1e7 Wait for the texture's newest writer instead of draining the replay queue in ReadTexels, with TOTK_REPLAY_CHECK comparing against a full drain
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-25 23:27:49 +03:00
Valiantsin DzerakhandClaude Opus 5.5 d161a8aefe Build a waiting draw's stage libraries on idle workers in parallel and skip draws into passes with a per-frame target, with TOTK_PIPE_WAIT and TOTK_PIPE_SKIP knobs
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-25 18:48:19 +03:00
Valiantsin DzerakhandClaude Opus 5.5 00d7c6689f Write descriptor sets through update templates made with each set layout, with TOTK_DESC_TEMPLATE=0 keeping the write structs
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-25 18:03:30 +03:00
Valiantsin DzerakhandClaude Opus 5.5 d36034b1ec Skip building a graphics draw's descriptor sets when its inputs equal the previous draw's in the command buffer
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-25 17:42:54 +03:00
Valiantsin DzerakhandClaude Opus 5.5 b7c59adcd9 Reuse the previous draw's shaders while its program, bound state and textures repeat, and skip pushes equal to the last one
About 65% of field draws repeat the previous draw's program, state and texture handles; shader_for falls from
10.8% to 6.4% of the RX 9070 render thread and field replay from 9.7 to 9.2 ms. answers_for looks up only the
slots whose handle changed. TOTK_SHADER_MEMO=check verifies both against full lookups; TOTK_BIND_STATS counts
repeats per draw.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-25 13:40:37 +03:00
Valiantsin DzerakhandClaude Opus 5.5 bf41b9cfef Make the program-to-shader cache 4-way set-associative so conflict misses stop hashing whole shaders with SHA-256 every frame
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-25 13:02:21 +03:00
Valiantsin DzerakhandClaude Opus 5.5 1fe552ae0a Mirror buffer-bound pools in device-local memory on discrete GPUs
Imported pools put vertices, indices, constants and storage buffers in system memory, so discrete GPUs read them over
PCIe on every draw. A pool now gets a device-local mirror on its first use as a vertex, index, uniform, storage or
indirect buffer; bindings, buffer copies, fills and uniform updates use it at the same offsets. The first access of a
submission syncs the pages written since (write-watch snapshots shared with the texture checks, plus
nvnMemoryPoolFlushMappedRange ranges) in a prologue command buffer before the part; what the GPU writes into a mirror
goes back to guest memory at the end of the submission. TOTK_MEMORY=import|mirror|verify overrides the device class;
verify compares sampled mirror pages with guest memory at the fence. The vkexec probe gains a mirror part.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-25 00:45:56 +03:00
Valiantsin DzerakhandClaude Opus 5.5 04314ef476 Count what a VRAM mirror of buffer-bound pools would upload under TOTK_TRACE=perf
At the start of every submit the dirty census queries write watch over the imported pools the replay bound as vertex,
index, uniform, storage or indirect buffers, stamps the shared page generations and logs eager, lazy and first-use
bytes per frame and per submit by pool class as mem: dirty every 60 frames, with FlushMappedRange calls and bytes and
the live and cumulative mirror size. The device class line (uma or discrete, heaps and memory types) is logged once at
device selection and the pool census also at frame 600. perf_summary.py summarises the new lines.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-25 00:12:47 +03:00
Valiantsin DzerakhandClaude Opus 5.5 94d5367040 Log GPU time per frame, VRAM, process RAM and a bound verdict in the perf line, and a pool census
Each submission part writes a timestamp pair under TOTK_TRACE=perf; the completion thread reads it and sums the busy
time without double-counting overlap. A second 60-frame perf line carries GPU ms per frame, fence waits of the render
thread and of other threads, VRAM usage and budget (VK_EXT_memory_budget, else the allocator's own blocks against the
heap size), working set and private bytes, and bound: gpu, cpu or pace. Devices without timestamp bits or a timestamp
period log one line and show gpu n/a.

vk_mem_resolve now reports the pool class of a span. The replay counts bytes bound or copied per class and use, and
vk_mem_census logs the live pools by kind and flags with that traffic at frame 1800 (every 1800 frames with perf).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-24 19:17:24 +03:00
Valiantsin DzerakhandClaude Opus 5.5 19c9b648d1 Add the recompiler, Windows runtime, tools and docs
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-23 22:27:54 +03:00