[Filip722]

What

Shader-info collection walks all instructions looking for a 64-bit shared atomic even when the fault-buffer flag is already true. Short-circuit that pure discovery scan once its result cannot affect the flag.

• Keep the identical predicate and order when the incoming flag is false.
• Preserve all earlier validation and later metadata collection.
• Reuse the existing shader-memory and LDS atomic fixtures; no new test infrastructure or diagnostic enters the patch.

cb28abeb89b added the atomic fault requirement; cd329c060aa made it unconditional on sharedMemoryBytes. That requirement remains, including workgroup-backed LDS. SharedAccessOf is a total pure switch, and the scan is discovery, not validation. Normal Recompile and the diagnostic deserializer receive the same complete ShaderInfo. No guest instruction, FP, EXEC, fault or memory-ordering semantics change.

Tested

Debian 13 / Linux 7.2.12-x64v3-xanmod1, Ryzen 5 7640HS, AMD Radeon 760M, RADV Mesa 25.0.7-2+deb13u1. Main cde94eb55a65 and candidate 7fa759dc64ec pass the same five focused tests: agc_shader_memory, agc_spirv_lds_atomic_lock, agc_driver_buffer_descriptor_runtime, agc_driver_lds_mixed_width_atomics and agc_driver_lds_lock_budget. Actual hardware includes mixed-width locking and misalignment faults on device-backed/workgroup LDS. Both arms require Vulkan, disable shader caching and insert Khronos validation on all three Radeon fixtures: zero VUID errors, missing layers or skips. Initial skipped build stubs were detected; the real LDS source was enabled before both final matched checks. These are local results; CI is not claimed.

Private same-process checks run the actual original/candidate Collect bodies on the same programs after restoring input ShaderInfo/completion, with a separate baseline pixel-input copy. All 392 complete ShaderInfo, completion flags and pixel-coverage results match. ShaderMemory has 382 calls: 61 already-true inputs and 321 false-to-false. SpirvLdsAtomicLock has 10: six false-to-true and four false-to-false. Atomic-predicate visits 15,666→10,258, with all required false-flag searches unchanged. No new production counter or fixture is added.

Native tradeoff: identical GCC 15.2 Release flags emit the expected flag branch bypassing the search loop. Main grows 1,670→1,674 instructions while bytes fall 8,446→8,442; branches 262→264, static calls 108 unchanged. Cold stays 221 instructions/40 branches/40 calls, bytes 1,136→1,144. Combined bytes 9,582→9,586. Both retain 872 local stack bytes and six saved-GPR pushes; spills were not measured. Production text sections match the uninstrumented private candidate. This is eliminated executed CPU work, with no static code-size, native GPU instruction or elapsed compilation gain claim.

Verified timing-only evidence includes all four original traces, original summaries, upstream digests, protocol, per-run audits, work/native statistics, validation and exact source/library provenance. Raw objects, ISA, cache files, game/save data, screenshots and console logs remain local. The overlap checker could not merge stale PRs #9748/#12619; their exact patches were manually reviewed: target-merge layered image metadata and flat pixel-input aliases respectively, with no shared changed lines or fault-scan mechanism. No dependency is required.

Sons of Sparta (PPSA28997 v01.008.001), restored camp, 1920x1080 native / focused 1152x648 Wayland client, mailbox, 25:12s windows. Both initial warmups use the same retained same-driver RADV cache and empty source shader caches, then independently warm and restart. All four A1/B1/B2/A2 captures pass with no exclusions or replacements: 21.75316→21.74201 FPS (observed -0.051%; pairs +0.173%/-0.275%). GPU busy 42.8125→42.7600 ms/present. The few warmed scene runs do not establish a significant/general gameplay gain or measure cold compilation latency.

Timing used main 4bc9e779aa7a and candidate 3815c275a08f. The full candidate runtime set is independently verified: shader library differs; VideoOut differs only in ten bytes of embedded revision text, all other bytes identical. All 50 native/9 guest loaded-library hashes are pinned. Source-revisions.json distinguishes timed builds from submission revisions after any upstream refresh; the collector source is byte-identical through that refresh. The submission refresh changes only the release workflow timeout; runtime sources and build flags are unchanged. Refreshed heads are not claimed to have been timed.

Original summaries in A1/B1/B2/A2 order:

{"commit":"4bc9e779aa7a","dirty":true,"env":{"APS5_NO_PAD_INPUT":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_DIR":"<path>","APS5_TRACE_SECONDS":"25:12"},"gpu":"AMD Radeon 760M (RADV PHOENIX)","driver":"radv Mesa 25.0.7-2+deb13u1","os":"Linux 7.2.12-x64v3-xanmod1","cpu":"AMD Ryzen 5 7640HS w/ Radeon 760M Graphics","cpu_threads":12,"memory_mib":23820,"title_id":"PPSA28997","title":"God of War Sons of Sparta","content_version":"01.008.001","resolution":"1920x1080","window":{"spec":"seconds 25:12","first_present":549,"ms":12015.108},"presents":261,"frame_ms":{"p50":45.965,"p90":46.452,"p99":47.622,"max":50.127},"gpu_busy_ms_per_present":42.830,"worker_busy_pct":{"queue worker 0x0":78.428},"top_phases_ms_per_frame":{"Driver.Packet":35.978,"Driver.Packet.draw":33.471,"Driver.Draw":33.431,"Driver.Draw.draw_enqueued":29.765,"Recorder.Sync":28.711,"VideoOut.Flip":9.063,"Driver.Present":9.030,"VideoOut.Flip.present":9.026},"trace_dropped":false,"trace":"aps5-trace-4bc9e779aa7a-3906520.json.gz"}
{"commit":"3815c275a08f","dirty":true,"env":{"APS5_NO_PAD_INPUT":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_DIR":"<path>","APS5_TRACE_SECONDS":"25:12"},"gpu":"AMD Radeon 760M (RADV PHOENIX)","driver":"radv Mesa 25.0.7-2+deb13u1","os":"Linux 7.2.12-x64v3-xanmod1","cpu":"AMD Ryzen 5 7640HS w/ Radeon 760M Graphics","cpu_threads":12,"memory_mib":23820,"title_id":"PPSA28997","title":"God of War Sons of Sparta","content_version":"01.008.001","resolution":"1920x1080","window":{"spec":"seconds 25:12","first_present":541,"ms":12040.281},"presents":262,"frame_ms":{"p50":45.936,"p90":46.329,"p99":46.792,"max":47.791},"gpu_busy_ms_per_present":42.747,"worker_busy_pct":{"queue worker 0x0":78.377},"top_phases_ms_per_frame":{"Driver.Packet":35.883,"Driver.Packet.draw":33.374,"Driver.Draw":33.334,"Driver.Draw.draw_enqueued":29.821,"Recorder.Sync":28.574,"VideoOut.Flip":9.075,"Driver.Present":9.064,"VideoOut.Flip.present":9.037},"trace_dropped":false,"trace":"aps5-trace-3815c275a08f-3907543.json.gz"}
{"commit":"3815c275a08f","dirty":true,"env":{"APS5_NO_PAD_INPUT":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_DIR":"<path>","APS5_TRACE_SECONDS":"25:12"},"gpu":"AMD Radeon 760M (RADV PHOENIX)","driver":"radv Mesa 25.0.7-2+deb13u1","os":"Linux 7.2.12-x64v3-xanmod1","cpu":"AMD Ryzen 5 7640HS w/ Radeon 760M Graphics","cpu_threads":12,"memory_mib":23820,"title_id":"PPSA28997","title":"God of War Sons of Sparta","content_version":"01.008.001","resolution":"1920x1080","window":{"spec":"seconds 25:12","first_present":546,"ms":12014.509},"presents":261,"frame_ms":{"p50":45.942,"p90":46.601,"p99":47.733,"max":49.419},"gpu_busy_ms_per_present":42.773,"worker_busy_pct":{"queue worker 0x0":78.443},"top_phases_ms_per_frame":{"Driver.Packet":35.974,"Driver.Packet.draw":33.407,"Driver.Draw":33.365,"Driver.Draw.draw_enqueued":29.544,"Recorder.Sync":28.440,"VideoOut.Flip":9.066,"Driver.Present":9.054,"VideoOut.Flip.present":9.027},"trace_dropped":false,"trace":"aps5-trace-3815c275a08f-3908557.json.gz"}
{"commit":"4bc9e779aa7a","dirty":true,"env":{"APS5_NO_PAD_INPUT":"1","APS5_PAD_INPUT_FILE":"<path>","APS5_TRACE_DIR":"<path>","APS5_TRACE_SECONDS":"25:12"},"gpu":"AMD Radeon 760M (RADV PHOENIX)","driver":"radv Mesa 25.0.7-2+deb13u1","os":"Linux 7.2.12-x64v3-xanmod1","cpu":"AMD Ryzen 5 7640HS w/ Radeon 760M Graphics","cpu_threads":12,"memory_mib":23820,"title_id":"PPSA28997","title":"God of War Sons of Sparta","content_version":"01.008.001","resolution":"1920x1080","window":{"spec":"seconds 25:12","first_present":562,"ms":12027.354},"presents":262,"frame_ms":{"p50":45.836,"p90":46.373,"p99":47.437,"max":48.209},"gpu_busy_ms_per_present":42.795,"worker_busy_pct":{"queue worker 0x0":78.435},"top_phases_ms_per_frame":{"Driver.Packet":35.875,"Driver.Packet.draw":33.379,"Driver.Draw":33.338,"Driver.Draw.draw_enqueued":29.490,"Recorder.Sync":28.448,"VideoOut.Flip":9.075,"Driver.Present":9.065,"VideoOut.Flip.present":9.037},"trace_dropped":false,"trace":"aps5-trace-4bc9e779aa7a-3909485.json.gz"}

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: none
• Unblocks: none
• GPUs tested: AMD Radeon 760M (RADV PHOENIX), Mesa 25.0.7-2+deb13u1, Debian GNU/Linux 13 / Linux 7.2.12-x64v3-xanmod1
• AI-assisted: yes

[github-actions]

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 9748	core/shader/recompiler/Optimization/src/ShaderInfoCollector.cpp
PR 12619	core/shader/recompiler/Optimization/src/ShaderInfoCollector.cpp

[Noxy-HNR]

Windows data point for 7fa759dc64: Windows 11 (build 26300), Intel Core Ultra 7 255HX, NVIDIA GeForce RTX 5070 Ti Laptop GPU (driver 617.14), WinLibs MinGW-w64 GCC 15.2.0, Release, ANYPS5_REQUIRE_VULKAN=1, with the RTX selected as the Vulkan device.

On this PR and on its base cde94eb55a, built the same way, the five tests from the description pass on both: agc_spirv_lds_atomic_lock, agc_shader_memory, agc_driver_buffer_descriptor_runtime, agc_driver_lds_mixed_width_atomics, agc_driver_lds_lock_budget.

The change is a straight || short-circuit of the same predicate, and the scan has no side effects, so the flag ends up the same either way. Looks good to me.

[TechLevelZero]

✅ NVIDIA GPU test: passed

PR head 7fa759d merged onto main 6b48aee · tested 2026-10-31 02:29 UTC · 22 min

GPU	NVIDIA RTX 2000 Ada Generation, driver 595.99.02, Vulkan 1.4.329
Host	Ubuntu 26.04.1 LTS, kernel 7.0.0-38-generic, container ubuntu:24.04
Logs	build and test logs for this run
Build	✅ passed (Release, -DANYPS5_ENABLE_SPIRV_TOOLS=ON)
Tests	✅ 2,314 passed · 4 failed (0 new, 3 also fail on main, 1 unstable on main) · 41 skipped (ANYPS5_REQUIRE_VULKAN=1)
Test setup	CPU-only tests without a GPU; GPU tests (and any CPU-only test that failed or skipped without one) once on the GPU with VK_LAYER_KHRONOS_validation and ANYPS5_NO_SHADER_CACHE=1; main ran the same way
Test selection	2,359 of 3,041 tests run: the ones this change can reach, chosen by tools/select_tests.py from main's build graph
Vulkan validation	✅ no new validation IDs vs main (863 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_graphics 33.4 s, ⚠️ agc_driver_graphics_maps_fallback 31.8 s, ⚠️ agc_driver_bda_device 31.3 s

Unstable on main (1, these also failed in recent main runs on this machine; not counted against this PR)

agc_submit_progress: failed in 6 of the last 10 main runs

Failing on main too (3, not counted against this PR)

agc_submit_progress_no_timeline
guest_mmap_executable
guest_socket_options_multicast

Validation IDs already present on main (3 IDs in 380 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (380 tests): agc_compute_pipeline_replay, agc_driver, agc_driver_address_passes, agc_driver_address_passes_draw_transitions, agc_driver_address_passes_targets_split, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, and 372 more
VUID-vkCmdCopyImage-srcImage-01548 (12 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_cpu_write_after_memo, agc_driver_depth_plane_cpu_write_after_page_unwatch, agc_driver_depth_plane_cpu_write_after_partial_unwatch, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_partial_store_unimported, agc_driver_depth_plane_sampling, and 4 more
VUID-vkCmdCopyImage-srcImage-01551 (12 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_cpu_write_after_memo, agc_driver_depth_plane_cpu_write_after_page_unwatch, agc_driver_depth_plane_cpu_write_after_partial_unwatch, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_partial_store_unimported, agc_driver_depth_plane_sampling, and 4 more

Skipped (41)

agc_driver_bda_reads_shadowed_results: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_bda_reads_shadowed_results_per_build: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_msaa_dcc_compared_import_2: skipped, the device reports host imports that keep write protection, so the write-watch probe compares none of them
agc_driver_msaa_dcc_compared_import_4: skipped, the device reports host imports that keep write protection, so the write-watch probe compares none of them
agc_driver_msaa_dcc_compared_import_8: skipped, the device reports host imports that keep write protection, so the write-watch probe compares none of them
agc_driver_sample_order_grid_reference: skipped, the device does not keep depth drawn at custom sample locations at those locations, even after a transition that names them: the composite of uniform renders holds the guest's depth in 0 of 3005 sloped samples, so no path can be compared with it bit for bit
agc_driver_stencil_ref_export: skipped, the device lacks VK_EXT_shader_stencil_export
agc_driver_vertex_elected_atomic_max: skipped, the device runs vertex programs in subgroups, so they keep their cross-lane reductions
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_import_settle_initial_dcc: skipped, the device has no deferred settle of host imports: imports under 1 MiB wait for one only on Windows
agc_unit_shadow_completion_store: skipped, the device reports watched host memory written when it imports or reads it
guest_http_uri_locale: 
guest_net_ctl: skipped, the host has no IPv4 address
guest_video_out_color_space: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_hdr: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_high_frame_rate: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_output_mode: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_resolution_status: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_wqhd_detection: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
video_out_flip_controls: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_cpuflip: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_cpuflip_in_order: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_eop: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_eop_aftergpu: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_eop_main: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_fresh: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_labels: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is eithe

Reproduce

git fetch origin 7fa759dc64ec9b1a26011019e194ac9ac726487e 6b48aee5d2ea9fce6155163d78f8ce8dc84f7827
git checkout 6b48aee5d2ea9fce6155163d78f8ce8dc84f7827 && git merge --no-ff 7fa759dc64ec9b1a26011019e194ac9ac726487e
git submodule update --init --recursive
cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=ON -DANYPS5_ENABLE_SPIRV_TOOLS=ON
cmake --build build && cmake --build build --target libs
VK_DRIVER_FILES=/nonexistent/icd.json ctest --test-dir build -E '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 32 --output-on-failure
ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation \
  ctest --test-dir build -R '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 4 --output-on-failure
# failures are then re-run alone: --timeout 360 --parallel 2; new validation IDs: the test alone, twice, --parallel 1

_{Automated build and test on a contributor's NVIDIA machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs.}

[adriadam10]

Checked the diff: usesFaultBuffer || any_of(...) keeps the same predicate and only skips the pure scan when the flag is already true; no other state depends on the scan. Windows run by @Noxy-HNR on the five listed tests and the NVIDIA run by @TechLevelZero also pass. The one red Windows test, check_compile_preprocessor, timed out at 20 s and is unrelated to this diff; merged.