[imellag]

What

StorageTexture::FlushPending syncs when an unfinished batch writes the flushed range (#6364, 9fffcea): a writer that stamps guest memory only when its batch completes (an address-based dispatch) must have its marks land before the write-back decides which 64 KiB blocks to keep. When the writer is in the open batch, that is a full Recorder::Sync.

Some writers need no such wait: storage image write-backs (writeBackLayers, writeBackWindows) call GuestMemory::MarkWritten for what they store while they are recorded, and a unit shadow publish copies bytes those write-backs already stamped. The next write-back of the same range is recorded after them in the same queue, so it lands after them on the GPU anyway. Today the flush still waits for them. Example: an image is written back at one label, written again by a dispatch, then flushed again at the next label; the second flush finds the first one's publish in the open batch and drains the queue.

• Recorder::NoteStampedWrites notes such ranges like NotePendingWrites and also keeps their write notes in the batch (stampedNotes). The two write-back paths and recordCopies in UnitShadow.cpp note their writes this way. The staged publish (the copy into a host buffer that writes guest memory at completion) does not.
• When a flush stores a single image, FlushPending keeps the stamped notes that flush recorded (the image's write-backs and its publish), from Recorder::StampedNotesSince, in the image's flushNotes (the last 32). A flush that stores several images keeps none, so no image takes another's notes.
• The next flush adds the image's flushNotes to the existing ownNotes exception of fix(agc): store resident linear color targets at completion points, limited to their drawn rectangles #9685, so it applies only where that one does: the flush stores a single image and that image has a host import. Otherwise ownNotes is cleared as on main. Any other writer over the range (another image's write-back or publish, a dispatch, draw, fill, label, CPU-resolved copy, address-based lease) still makes the flush wait, and CPU readers still sync on these writes through the pending-write snapshot as before.

tests/execution/StorageFlushOrdering.cpp gains a stored-again case: images are stored, written again by a dispatch and flushed again in the same open batch, and the newer texels must land. With one image and a host import the second flush must not submit the open batch. It must wait with two adjacent images flushed together, without a host import, when another stamped write that is not the image's own (noted with NoteStampedWrites over the image) is in the open batch, or when the image is flushed alone after a joint flush with its neighbour. A second registration, agc_driver_storage_flush_ordering_no_import, runs the stored-again cases with APS5_NO_HOST_IMPORT=1. main fails the one-image case (it waits), the first head of this PR failed the two-image case, the second head failed the unrelated stamped write case and the third the alone-after-joint case (they did not wait), the current head passes all of them. The two existing cases (writer in flight, writer in the open batch) still pass.

Overlaps:

• perf(agc): wait on a flush only for writers over the image layers it stores #8887 (SleglJan) narrows the same check to the blocks of the stored layers. Complementary: here the writer is over the image's own blocks, so perf(agc): wait on a flush only for writers over the image layers it stores #8887 still waits for it. Same lines in FlushPending and in this test. perf(agc): wait on a flush only for writers over the image layers it stores #8887 merged on 2026-11-02 and this branch contains it (main merge 41f3e3c20e).
• fix(agc): keep CPU stores made after a deferred DCC key store was recorded #6186 (ee6aba54) tags the publish note in recordCopies with a write kind (ShadowPublishWrite) and conflicts with this PR in Recorder.hpp, UnitShadow.cpp, the AGC CMakeLists.txt and the AGC debt file. Order agreed with ee6aba54 on fix(agc): keep CPU stores made after a deferred DCC key store was recorded #6186 (2026-10-31): this PR first, then fix(agc): keep CPU stores made after a deferred DCC key store was recorded #6186 rebases onto it and carries the stamped flag through its write kind. This PR does not need fix(agc): keep CPU stores made after a deferred DCC key store was recorded #6186, so it is not in Depends on.
• fix(agc): publish small tiled writes before labels #11694 (droid2299, merged) sends small tiled write-backs to the import instead of the unit shadow; that path (writeBackWindows' imported pieces) is noted as stamped here too.

Tested

No performance claim: main cannot run titles or the AGC execution tests on macOS yet, and I have no other platform for a main-vs-change capture.

Current head a8306d56f5: 41f3e3c20e merged main 88e804b0a7 with #8887 (combined results in this comment), 2b211c7231 merged main 5b1d381d5a, and a8306d56f5 fixes the test's expectation on hosts that compare host imports instead of watching them (details). The runs below are of the earlier heads.

Head e7e78cff0e, rebased onto main 98ddd9c179. One conflict, in the AGC CMakeLists.txt: main registers agc_driver_registration_prepare_budget where this PR registers agc_driver_storage_flush_ordering_no_import; both are kept. The other files merged without conflicts, and what this PR builds on is unchanged on main: noteWrite and NotePendingWrites still give each range its own write note, and PendingWriterSerial with ownNotes still skips a write by its note. With #8887's head 1f15e200f7 the only conflict is in StorageFlushOrdering.cpp, where both add cases.

macOS 27.2, Apple M4 Pro, MoltenVK 1.4.2 (Vulkan SDK 1.4.363.0), x86_64 Release under Rosetta. Support branch, not main: fork branch support/macos-boot-1030 (6747733e89) = main 98ddd9c179 + only the open macOS PRs #11447 #11554 #10552 #7631 #7804 #4630 and #13220 (the duplicate write_watch test registration that stops main's CMake on macOS), plus one local line that reads environ through _NSGetEnviron() in HostReportSources.cpp (main doesn't compile it on macOS; no open PR covers it yet). The conflicts of #11554, #10552, #7631, #7804 and #4630 with main are resolved in the merge commits. Change: fork branch support/macos-boot-1030-pr12070 (f426fc662b) = that branch + the four commits of this PR.

Focused tests on both with ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation VK_LOADER_DEBUG=layer, ctest -R 'storage_flush_ordering|linear_readback_label|linear_target|agc_driver_storage_(cache_budget|one_slice_alias|read_during_idle_store|write_back)|storage_cache_retained|depth_storage_reuse|flip_stores_small' -V:

• base: 25 of 25 pass;
• change: 26 of 26 pass, the extra one being this PR's agc_driver_storage_flush_ordering_no_import. Both storage_flush_ordering registrations pass (storage flush ordering tests passed, storage flush ordering stored-again tests passed);
• in both logs the loader lists VK_LAYER_KHRONOS_validation as enabled by VK_INSTANCE_LAYERS, there is no was not found but was requested by env var VK_INSTANCE_LAYERS, and no VUID- line.

The 10 failures of the previous run (the "two different device resolvers" abort) are gone: #11705 has merged. Logs attached.

Previous head 3b3cd7377f (support branch support/macos-boot-1027, main d13b3d7c55): the same tests failed the same 10 tests on both sides for the resolver abort and passed otherwise, both storage_flush_ordering registrations passed on the change, no VUID-. God of War (PPSA28997) ran through the Cyclops fight on the second head with no crash or device loss; not rerun since.

python3 tools/check_conventions.py --base upstream/main (98ddd9c179): passed. tools/pr_overlap.py: conflicting lines only with #8887 (StorageFlushOrdering.cpp).

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: Apple M4 Pro, MoltenVK 1.4.2, macOS 27.2
• AI-assisted: yes

[github-actions]

This pull request is labeled perf, so it needs the summary line of a capture on main and one with the change, from the same title, scene and machine, as the Performance report form asks. Please paste both aps5-summary.jsonl lines, attach the two traces and add the output of tools/trace_digest.py.

[github-actions]

Other open pull requests touch the same code. Merging one will break the other, or both implement the same thing. If this one builds on them, list them in Depends on.

Pull request	Conflicting files (lines on main)	Same exports	Same new files
PR 7642			core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake
PR 9508	core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake		core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake
PR 10193			core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake
PR 13928			core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake
PR 13929			core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake

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 1368	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 1395	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 5609	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 6186	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/UnitShadow.cpp
PR 6187	core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 6188	core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp core/libs/prx/libSceAgcDriver/Graphics/src/UnitShadow.cpp
PR 6192	core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 6197	core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 6209	core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 6280	core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 6481	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 7085	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 7449	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp
PR 7514	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 7604	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 7642	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp core/libs/prx/libSceAgcDriver/Graphics/src/UnitShadow.cpp core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake core/libs/prx/libSceAgcDriver/tests/execution/StorageFlushOrdering.cpp
PR 8484	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 9289	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp
PR 9356	core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp
PR 9387	core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp
PR 9884	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 10193	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp core/libs/prx/libSceAgcDriver/Graphics/src/UnitShadow.cpp core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake core/libs/prx/libSceAgcDriver/tests/execution/StorageFlushOrdering.cpp
PR 10249	core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 10442	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp
PR 11444	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 12746	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 13251	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 13486	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 13793	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp
PR 13796	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 13809	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp
PR 13863	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp core/libs/prx/libSceAgcDriver/Graphics/src/UnitShadow.cpp core/libs/prx/libSceAgcDriver/tests/execution/StorageFlushOrdering.cpp
PR 13920	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp core/libs/prx/libSceAgcDriver/Graphics/src/UnitShadow.cpp core/libs/prx/libSceAgcDriver/tests/execution/StorageFlushOrdering.cpp
PR 13928	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp core/libs/prx/libSceAgcDriver/Graphics/src/UnitShadow.cpp core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake core/libs/prx/libSceAgcDriver/tests/execution/StorageFlushOrdering.cpp
PR 13929	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp core/libs/prx/libSceAgcDriver/Graphics/src/UnitShadow.cpp core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_storage_flush_ordering_tests.cmake core/libs/prx/libSceAgcDriver/tests/execution/StorageFlushOrdering.cpp
PR 14344	core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp
PR 15136	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/include/Texture.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 15529	core/libs/prx/libSceAgcDriver/Graphics/src/Texture.cpp
PR 15548	core/libs/prx/libSceAgcDriver/Graphics/include/Recorder.hpp core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp
PR 15570	core/libs/prx/libSceAgcDriver/Graphics/src/Recorder.cpp

[oneandonlydean]

@imellag thanks, the record-time stamping itself looks sound. One question before this can go in: #9685 (c7f9eaa) made FlushPending skip a target's own earlier stores only when the flush stores that target alone and it has a host import; otherwise ownNotes is cleared and the flush waits. Stamping writeBackLayers' note makes PendingUnstampedWriterSerial skip those same stores in the cases #9685 excluded (several images flushed, or no import), and it also skips other images' write-backs and every unit-shadow publish. Could you either apply the stamped skip only where the own-store skip is allowed today, or explain why #9685's restriction does not apply to stamped writes, and add a case that flushes two adjacent images (one without an import) together after a stamped store? A same-machine main-vs-head measurement on current main would also help, since the current numbers are from one scene on a non-main tree.

[MarcZX10]

I reviewed the recorder changes and the new ordering test, and I agree with the concern already raised by the collaborator: PendingUnstampedWriterSerial skips every stamped note before FlushPending can apply the existing ownNotes exception, which is deliberately restricted to a single image with an existing host import. The new test covers a repeated store for one image, but not multiple flushed images or an image without a host import. Please explain why the #9685 restriction does not apply to those stamped writes, or constrain the skip to the same safe cases and add the requested multi-image/no-import regression case. I am not treating the current test as evidence for those cases.

The performance report also does not use the repository main baseline: it names a custom tree containing open macOS boot PRs and #11705. Please rerun the matched capture against current main and this exact change, and refresh the attached digest/traces as requested by the performance-report check. Finally, coordinate the same-line/test overlap with open #8887 and record the agreed ordering in Depends on per CONTRIBUTING; the overlap section currently says either PR may land second while the checklist says none.

[imellag]

@oneandonlydean @MarcZX10 thanks, you are right about #9685's restriction. A flush that stores several images, or an image without a host import, can store on the CPU or over a neighbour, and must wait for every pending GPU write over the range, stamped or not.

Pushed 156bbaa (branch rebased on current main d13b3d7):

• The stamped skip now applies only where the own-store skip applies today: the flush stores one image and that image has a host import. Otherwise FlushPending uses PendingWriterSerial exactly as on main.
• StorageFlushOrdering.cpp: the stored-again case now runs for one image (must not wait with an import, must wait without one) and for two adjacent images flushed together (must wait). A second registration, agc_driver_storage_flush_ordering_no_import, runs the stored-again cases with APS5_NO_HOST_IMPORT=1. Checked on the three versions: main fails the one-image case (it waits), the first head of this PR fails the two-image case (it does not wait), this head passes both. I could not make only one of two adjacent images lack an import inside one test block (imports cover the block), so the no-import case runs alone under APS5_NO_HOST_IMPORT=1.

Baseline. main cannot run titles or the AGC execution tests on macOS yet. Both arms are now main plus only the boot PRs, nothing else:

• base: fork branch support/macos-boot-1027 (76d41ef) = main d13b3d7 + feat(libkernel): direct memory mappings on macOS hosts #7631 fix(libkernel): build the file, module and rtld calls on macOS hosts #10552 build(libkernel): compile Exception on macOS #11249 build(agc): compile Lifecycle, GuestMemory and GuestBufferMemory on macOS #11447 fix(agc): load the Vulkan loader and list MoltenVK on macOS #11554 build(libSceAudioIn): keep --exclude-libs off Apple's linker #10669 feat(libkernel): process, thread and clock queries on macOS hosts #7804 (feat(libkernel): process, thread and clock queries on macOS hosts #7804 conflicts resolved locally in Process.cpp and Time.cpp);
• change: fork branch support/macos-boot-1027-pr12070 = base + the two commits of this PR.

Why each one: #7631, #7804, #10552, #11249 libkernel/libc on macOS; #11447 and #11554 AGC driver on macOS; #10669 libSceAudioIn link. The earlier numbers in the description came from a tree that also had #11551 and #11705; they are replaced below.

Same focused tests on both arms with ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation: the same 10 tests fail on both arms with the same messages (the aborts are "deferred destruction found two different device resolvers", fixed by #11705, which is not in this base), all the others pass, both storage_flush_ordering registrations pass on the change, no VUID- in either log.

God of War (PPSA28997), Cyclops fight, one pair on these two branches, same machine, APS5_TRACE_SUMMARY_EVERY=10:

	fight frame time per 10 s	scene before the fight	[recorder] pending-write syncs per 160 s run
base	23.6-23.7 ms	16.9-17.3 ms	5,427
change	22.1-22.3 ms	16.7 ms (vsync)	0

This is one pair only; I will add a matched trace capture of these two branches with trace_digest output and update the description.

#8887. I measured it on macOS and proposed there that it lands first; once SleglJan agrees I will set Depends on: #8887 and rebase this PR on it. Until then the checklist stays at none.

[MarcZX10]

Re-reviewed exact head 156bbaa. The new guard and the no-import/multi-image regression cases address the broad skip outside the existing single-image, host-import case. I found three remaining blockers:

1. When ownStoresSkipped is true, PendingUnstampedWriterSerial ignores every overlapping note in batch.stampedNotes, not only this flush's ownNotes. NoteStampedWrites is used both for storage write-backs and UnitShadow::recordCopies. The single-image/host-import guard does not establish that every overlapping stamped note belongs to the image being flushed; a pending unit-shadow publish or another overlapping stamped writer can still be skipped. Please narrow the exception to the flush's own writes, or document the invariant that rules out unrelated stamped writers and add a regression with an unrelated stamped publish overlapping a single imported image.

2. The current heads of #8887 and #6186 both conflict with this head in shared AGC code. Pairwise merge-tree reports conflicts for #8887 in Recorder.hpp, Recorder.cpp, Texture.cpp and StorageFlushOrdering.cpp; for #6186 it reports conflicts in CMakeLists.txt, Recorder.hpp, UnitShadow.cpp and the debt file. The PR describes these overlaps but still says Depends on: none. CONTRIBUTING requires agreement on order and a declared dependency for same-line overlaps. Please coordinate and declare the agreed order. I checked #11694 as well; its current head merges cleanly with this PR, so I do not see a required ordering there.

3. The updated report includes matched traces and digests, but its baseline is e5d3b00 from a custom tree with macOS boot PRs and #11705, rather than current upstream main d13b3d7. The perf contribution rule requires a current-main/change comparison on the same machine and scene. Please provide that comparison on a supported platform or remove the performance claim and label.

The current-main merge preview for this head is clean. I am requesting these specific corrections; this is not an approval of the performance or ordering claims.

[imellag]

@MarcZX10 thanks. Updated what I can without new test runs:

1. Narrowing the skip to the flush's own writes: agreed. I will replace the batch-wide stampedNotes with the image's own notes (its write-backs and the publishes of its own flushes, through the existing ownNotes path), which also drops the UnitShadow.cpp change, and add a regression with an unrelated stamped publish over a single imported image. That needs a test run, so it comes in a later commit; the description lists it as open.
2. Order: agreed with SleglJan on perf(agc): wait on a flush only for writers over the image layers it stores #8887: perf(agc): wait on a flush only for writers over the image layers it stores #8887 lands first, and I will rebase this PR on it (Depends on: #8887 is set). With item 1 this PR no longer touches UnitShadow.cpp or the recorder's batch fields, which removes most of the overlap with fix(agc): keep CPU stores made after a deferred DCC key store was recorded #6186; I will check the remaining overlap against fix(agc): keep CPU stores made after a deferred DCC key store was recorded #6186 then.
3. Performance: I removed the performance claim, the numbers and the perf title; the PR is now fix(agc). main cannot run titles on macOS, so there is no main-vs-change capture from me. The Tested section names the support branch the tests ran on.

[maskym]

Windows + NVIDIA result for head 156bbaaaba, merged into main f759e165d9 (the PR's GPUs tested line names no NVIDIA run on Windows).

This PR's own tests: agc_driver_storage_flush_ordering (0.54 s) and the new agc_driver_storage_flush_ordering_no_import (0.69 s), all passed.

Host: Windows 11 26300, Intel i7-12700K, NVIDIA GeForce RTX 2070 SUPER (vendor 0x10de, device 0x1e84; Vulkan loader 1.4.357 from the NVIDIA driver 621.01), Khronos Validation Layers nightly 1.4.366 (b2b44c0d37) through VK_ADD_LAYER_PATH, WinLibs GCC 15.2.0 (x86_64-ucrt-posix-seh, r6), Release, BUILD_TESTING=ON, Ninja.

Same focused run on main f759e165d9 and on this tree, same machine and GPU: VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation, ANYPS5_NO_SHADER_CACHE=1, ANYPS5_REQUIRE_VULKAN=1, VK_LOADER_DEBUG=error,warn,layer, ctest --test-dir build -R '^(agc|guest_video_out|video_out)' --parallel 4 -V.

• main: 1127/1128 passed (9 skipped), the failure being agc_driver_mesh_barycentric_emulation (AGC: agc_driver_mesh_barycentric_emulation fails since #9193 on devices with mesh shaders and VK_KHR_fragment_shader_barycentric #12121).
• This tree: every pre-existing test passes, including that one (the tree carries test(agc): disable barycentric built-ins in the mesh barycentric emulation test #12062, its fix); the only tests not run are feat(libSceVideoOut): limit the guest display to the host resolution #11948's three, whose executable fails to link on Windows (reported on that PR). The loader inserted VK_LAYER_KHRONOS_validation in every test (998 and 999 insertions), neither log contains was not found but was requested by env var VK_INSTANCE_LAYERS, and both report exactly one VUID, VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 18 times (pre-existing on main, AGC: pipelines use differing per-attachment blend states while independentBlend is never enabled (VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605) #11990 / fix(agc): enable independentBlend for color targets with different blend states #6782): this PR adds no new VUID- error.

The tree also carried the open PRs #11948, #12062, #12132, #11680, #11679, #11678, #11677, #11675, #11674, #11605, #11604, #11600, all merged cleanly, so the run is a joint one. Not run on a PS5.

AI-assisted: yes (Claude Code), the runs and the numbers above are from this machine.

[imellag]

@MarcZX10 a preview of the combined stack you asked for, before #8887 lands. This branch is unchanged.

I merged #8887's current head ec47dad254 into this PR's head e7e78cff0e, locally only. The one conflict is the test list at the end of StorageFlushOrdering.cpp's main: both PRs add cases after Run(..., "writer in the open batch"). The resolution keeps all of them, this PR's RunStoredAgain and RunStoredAloneAfterJoint first, then #8887's RunWriterOutsideLayers, RunWriterInsideLayer and RunShadowedUnitUnderWriter. Texture.cpp merges cleanly, as SleglJan said.

macOS run of that stack: Apple M4 Pro, macOS 27.2, MoltenVK 1.4.2, x86-64 Release under Rosetta 2, on our support branch support/macos-boot-1030c (05bd71a66b, main 21a556f567 + the open macOS PRs). Same pattern as in Tested plus storage_flush_wait_mip_chain, with ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation VK_LOADER_DEBUG=layer:

• Support branch alone: 25 of 25.
• Support branch + perf(agc): wait on a flush only for writers over the image layers it stores #8887: 26 of 26 (posted on perf(agc): wait on a flush only for writers over the image layers it stores #8887).
• Support branch + this PR on perf(agc): wait on a flush only for writers over the image layers it stores #8887: 27 of 27, including both storage_flush_ordering registrations (stored-again and the main one) and flush wait over a DCC mip chain: ok.
• The validation layer is loaded in every log. The only VUID- is VUID-VkDeviceCreateInfo-pProperties-04451 from agc_storage_flush_wait_mip_chain. That comes from agc_driver_recorder_tests' existing device creation and appears 150 times on the base too (details on perf(agc): wait on a flush only for writers over the image layers it stores #8887).

Once #8887 lands, I'll rebase onto main with this resolution and rerun the same set.

AI-assisted: yes

[milescoviello]

@imellag a data point from Ghost of Tsushima (PPSA02225), on the ownership question @MarcZX10 raised. It isn't a request to widen the skip.

I cherry-picked this PR's four commits (a70ae428a6..e7e78cff0e) onto my local GoT development tree, which is main from 10-29 plus local diagnostics. With it, GoT's FlushPending waits during storage-image refreshes stay as they were. Linux, RTX 4070 Ti, in-game runs, average per 10 s after the first 200 s: with this PR, the refresh flushes waited 2.6 s and 2.3 s in two runs; without it, 1.3 s and 2.5 s. That is 6 to 7 ms per flush either way.

A local diagnostic classified each overlapping pending write in the recorder's batches whenever FlushPending synced. Over the first 4,000 syncs of one of the runs with this PR:

Flush reason	Syncs	Only another image's stamped notes	Only pending address-based reads	Own stamped notes	Unstamped writes
storage refresh	2,317	2,227	90	0	0
readback before label	1,645	650	995	0	0
storage image creation	20	19	1	0	0

In this title, nearly every refresh sync waits for stamped notes of images other than the one being flushed, and none waits for an unstamped (address-based) writer. So the single-image, own-notes exception doesn't apply to any of them. Lifting it would need the invariant the review asked for, namely that an unrelated stamped write-back or unit-shadow publish can't change what this flush keeps. This is only how one title's waits split. My diagnostic doesn't tell write-backs from unit-shadow publishes.

AI-assisted: yes (Claude Code)

[TechLevelZero]

✅ NVIDIA GPU test: passed

PR head 485a312 merged onto main 3741ca9 · tested 2026-11-01 10:11 UTC · 19 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	✅ 2,473 passed · 5 failed (0 new, 3 also fail on main, 2 unstable on main) · 44 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	2,522 of 3,121 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 (930 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_lookup_log_fits_build 91.0 s, ⚠️ agc_driver_pipeline_journal 60.2 s, ⚠️ agc_driver_interpolation_modes 49.1 s, ⚠️ agc_driver_lazy_resolved_abi 34.6 s, ⚠️ agc_driver_graphics_maps_fallback 31.2 s

Unstable on main (2, 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
agc_submit_progress_no_timeline: failed in 8 of the last 10 main runs

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

agc_driver_image_table
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (1, not in the main baseline's validation pass)

agc_driver_storage_flush_ordering_no_import

Validation IDs already present on main (5 IDs in 480 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (480 tests): agc_driver, agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, and 472 more
VUID-vkCmdCopyImage-srcImage-01548 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-vkCmdCopyImage-srcImage-01551 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (1 test): agc_driver_prepared_exports
VUID-VkShaderModuleCreateInfo-pCode-08737 (1 test): agc_driver_image_sample_derivatives

Skipped (44)

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_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_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_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_backtoback: 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_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_dcc: 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_deferred: 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_f

Reproduce

git fetch origin 485a3126a1aac5eac6f222ece5f2249e1f05647e 3741ca9327246d0df81638726a986cef862b6db5
git checkout 3741ca9327246d0df81638726a986cef862b6db5 && git merge --no-ff 485a3126a1aac5eac6f222ece5f2249e1f05647e
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.}

[imellag]

@milescoviello thanks for the Ghost of Tsushima data. It matches what this PR does since the review: it skips only the flushed image's own stamped notes, and only when the flush stores that single image. Your table has 0 syncs in the "own stamped notes" column, so unchanged refresh waits in GoT are what I'd expect, and the PR claims nothing for that title.

The waits on other images' stamped notes are the case the invariant MarcZX10 asked for would cover. I'm not widening the skip without it.

AI-assisted: yes

[imellag]

@MarcZX10 #8887 has landed, so here is the combined-stack result you asked for.

Head 41f3e3c20e merges main 88e804b0a7 (which has #8887) into the branch. The only conflict was the one I described above: both PRs add cases at the end of StorageFlushOrdering.cpp's main. The resolution keeps all of them, this PR's RunStoredAgain / RunStoredAloneAfterJoint first, then #8887's RunWriterOutsideLayers, RunWriterInsideLayer and RunShadowedUnitUnderWriter. The diff against main is still only this PR's 7 files.

macOS run: Apple M4 Pro, macOS 27.2, MoltenVK 1.4.2 (Vulkan SDK 1.4.363.0), x86-64 Release under Rosetta 2. main doesn't build the driver on macOS by itself, so both sides are a local support branch: main 88e804b0a7 + the open macOS PRs #11554, #7631, #7804 and #4630, + #14883 (FreeTypeOl NID patching), + two build fixes for main's dma-buf run list import that I'm preparing as a PR (the __APPLE__ udmabuf stand-ins lacked udmabuf_create_list, and ScatteredRuns.cpp compiled its Linux body). The change side is that branch + this head, and nothing else differs. Command: ctest -R 'storage_flush_ordering|storage_flush_wait_mip_chain|linear_readback_label|linear_target|agc_driver_storage_(cache_budget|one_slice_alias|read_during_idle_store|write_back)|storage_cache_retained|depth_storage_reuse|flip_stores_small' -V, with ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation VK_LOADER_DEBUG=layer:

• base (main with perf(agc): wait on a flush only for writers over the image layers it stores #8887): 26 of 26 pass, including agc_driver_storage_flush_ordering and agc_storage_flush_wait_mip_chain;
• this head: 27 of 27 pass, the same 26 plus this PR's agc_driver_storage_flush_ordering_no_import (the stored-again cases);
• both logs show the loader inserting VK_LAYER_KHRONOS_validation, and neither has was not found but was requested by env var VK_INSTANCE_LAYERS. The only VUID- in either log is VUID-VkDeviceCreateInfo-pProperties-04451, twice in each, from agc_driver_recorder_tests' device creation (test(agc): enable the portability subset with its features in the recorder tests' device #13942 removes it). So there are no new IDs.

This head isn't measured in titles yet. The GoW numbers in the description are from the earlier stack.

AI-assisted: yes

[TechLevelZero]

✅ NVIDIA GPU test: passed

PR head 41f3e3c merged onto main 88e804b · tested 2026-11-02 10:18 UTC · 26 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	✅ 2,550 passed · 5 failed (0 new, 3 also fail on main, 2 unstable on main) · 45 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	2,600 of 3,234 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 (971 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 120.1 s, ⚠️ agc_driver_graphics_maps_fallback 30.2 s

Unstable on main (2, these also failed in recent main runs on this machine; not counted against this PR)

agc_driver_pipeline_journal: failed in 2 of the last 10 main runs
agc_submit_progress: failed in 8 of the last 10 main runs

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

agc_driver_image_table
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (1, not in the main baseline's validation pass)

agc_driver_storage_flush_ordering_no_import

Validation IDs already present on main (4 IDs in 495 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (495 tests): agc_driver, agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, and 487 more
VUID-vkCmdCopyImage-srcImage-01548 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-vkCmdCopyImage-srcImage-01551 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-VkShaderModuleCreateInfo-pCode-08737 (1 test): agc_driver_image_sample_derivatives

Skipped (45)

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_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_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_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_backtoback: 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_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_dcc: 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_deferred: 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_fps_limit: 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_o

Reproduce

git fetch origin 41f3e3c20e807aff689f44486566948e0d358c61 88e804b0a7df1f1ac373a9e2f72fb378d925fdd4
git checkout 88e804b0a7df1f1ac373a9e2f72fb378d925fdd4 && git merge --no-ff 41f3e3c20e807aff689f44486566948e0d358c61
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.}

[imellag]

The red windows check on 41f3e3c20e isn't a test failure. The job was cancelled at its 45-minute timeout-minutes while still in "Build and run changed guest core/libs tests" (09:16 to 10:02 UTC); the log has no failed test or compile error. Could a maintainer re-run it? I can't trigger a re-run from the fork.

AI-assisted: yes

[TechLevelZero]

❌ AMD GPU test: failed

PR head 41f3e3c merged onto main 7feb37e · tested 2026-11-03 08:27 UTC · 49 min

GPU	AMD Radeon RX 6300 (RADV NAVI24), driver Mesa 25.2.8-0ubuntu0.24.04.4, Vulkan 1.4.318
Host	Ubuntu 26.04.1 LTS, kernel 7.0.0-42-generic, container ubuntu:24.04
Logs	build and test logs for this run
Build	✅ passed (Release, -DANYPS5_ENABLE_SPIRV_TOOLS=ON)
Tests	❌ 2,723 passed · 10 failed (1 new: 1 not compared, 8 also fail on main, 1 unstable on main) · 58 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	2,791 of 3,743 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 (935 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 120.1 s, ⚠️ public_headers 49.1 s, ⚠️ agc_driver_image_sample_unnormalized 41.6 s, ⚠️ agc_driver_lazy_resolved_abi 35.5 s

Failing tests

This PR changes test files, so each failing test was also built and run on main with only the PR's test changes applied. ✅ there means the failure comes from the PR's other changes; ❌ means the test fails on main too, so this run does not show the PR caused it.

Test	With this PR	Same test on main 7feb37e (main + this PR's test changes only)
agc_driver_storage_flush_ordering	❌ failed	not compared: the PR's tests do not build without its other changes

agc_driver_storage_flush_ordering: Failed (also failed when re-run alone), assertion and diagnostics

ctest: Failed

[write-watch] dma-buf imports keep write protection: unknown (probe failed at open /dev/udmabuf, -1000072003); imported ranges are compared
AGC graphics: storage flush ordering, 2 images stored again in the open batch: the second store did not wait for the earlier store, which only a single image with a host import may skip

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

agc_driver_pipeline_journal: failed in 5 of the last 7 main runs

Failing on main too (8, not counted against this PR)

agc_driver_image_table
agc_driver_native_depth_bda_clear
agc_driver_native_multisample_depth
agc_idle_publish_wait
agc_submit_progress
agc_submit_progress_no_timeline
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (1, not in the main baseline's validation pass)

agc_driver_storage_flush_ordering_no_import

Validation IDs already present on main (4 IDs in 525 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (525 tests): agc_driver, agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, and 517 more
VUID-vkCmdCopyImage-srcImage-01548 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth, agc_driver_stencil_null_feedback
VUID-vkCmdCopyImage-srcImage-01551 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth, agc_driver_stencil_null_feedback
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (1 test): agc_driver_srgb8_color_target

Skipped (58)

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_copied_writer_label_order: skipped, the device reports a minStorageBufferOffsetAlignment of 4, so the dispatch binds its record at +0x1004 in place and nothing is written back; the test needs an alignment above 4
agc_driver_float16_unused_nan_payloads: skipped, the device has no native f16 conversions with RTE rounding and kept f16 denormals
agc_driver_image_table_vouched_lookups: skipped, the device reports 4 of 4 host-imported pages written after a GPU read, so a watched imported table cannot be vouched for
agc_driver_linear_target_direct_write_completion: skipped, the device does not keep host imports write-watched, so no completion stamp over the target is tracked
agc_driver_ngg_vertex_wave_reduction: skipped, s3 describes a 32-lane wave only on 32-lane subgroups, and the device has 64 lanes
agc_driver_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_unit_shadow: skipped, the device does not import watched host memory for unit shadows
agc_driver_unit_shadow_budget: skipped, the device does not import watched host memory for unit shadows
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_shadowed_key_stores: skipped, the device imports no watched host memory under a unit shadow
agc_small_tiled_shadow_label: skipped, host imports are unavailable: the test block is not imported and watched
agc_storage_clear_hand_over: skipped, the device imports no watched host memory for the clear-key hand-over
agc_unit_shadow: skipped, the device does not import watched host memory for unit shadows
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_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_backtoback: 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_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 co

Reproduce

git fetch origin 41f3e3c20e807aff689f44486566948e0d358c61 7feb37ee3e30e10139b483fb4df689d5b5b69ab8
git checkout 7feb37ee3e30e10139b483fb4df689d5b5b69ab8 && git merge --no-ff 41f3e3c20e807aff689f44486566948e0d358c61
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 12 --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 16 --output-on-failure
# failures are then re-run alone: --timeout 360 --parallel 2; new validation IDs: the test alone, twice, --parallel 1
# the failing tests, on this merge and on main:
ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation \
  ctest --test-dir build -R '^(agc_driver_storage_flush_ordering)$' --parallel 1 -V

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

[oneandonlydean]

Hi imellag,

Thanks for the work on this. The AMD GPU bot reported one new failure at this exact head (41f3e3c, merged onto main 7feb37e, 3 Nov, 08:27 UTC): agc_driver_storage_flush_ordering fails, and it also failed when re-run on its own. The bot could not compare it against main, because the test changes in this PR do not build without the production changes, so it is counted against this PR.

The message it prints is: "storage flush ordering, 2 images stored again in the open batch: the second store did not wait for the earlier store, which only a single image with a host import may skip". That is the check at StorageFlushOrdering.cpp:271: the text has no "without a host import" suffix, so on that machine the test saw a host import for both images, and the second FlushPending with two images did not advance OpenSerial(). In the production change, the new flush-note recording in StorageTexture::FlushPending is limited to flush.size() == 1, so the two-image case relies on the earlier paths to make the second store wait.

Could you check that case on an AMD/RADV setup (or any host where the two-image import path is taken), and either make the second store wait there or explain why the test's expectation does not hold on that driver? The bot's log is the gist linked in its comment on the PR. Separately, this branch now conflicts with main in Graphics/include/Texture.hpp.

[TechLevelZero]

✅ NVIDIA GPU test: passed

PR head 2b211c7 merged onto main 5b1d381 · tested 2026-11-04 10:38 UTC · 13 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	✅ 2,831 passed · 5 failed (0 new, 3 also fail on main, 2 unstable on main) · 48 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	2,884 of 3,883 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 (975 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 106.6 s

Unstable on main (2, these also failed in recent main runs on this machine; not counted against this PR)

agc_submit_progress: failed in 4 of the last 10 main runs
agc_submit_progress_no_timeline: failed in 3 of the last 10 main runs

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

agc_driver_image_table
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (1, not in the main baseline's validation pass)

agc_driver_storage_flush_ordering_no_import

Validation IDs already present on main (13 IDs in 544 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (544 tests): agc_driver, agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, and 536 more
VUID-VkShaderModuleCreateInfo-pCode-08737 (8 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_float16_fma_rounding, agc_driver_float16_unused_nan_payloads, agc_driver_packed_alu, agc_driver_packed_f16_denormal_modes, agc_driver_packed_float16_native
VUID-VkShaderModuleCreateInfo-pCode-08739 (8 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_float16_fma_rounding, agc_driver_float16_unused_nan_payloads, agc_driver_packed_alu, agc_driver_packed_f16_denormal_modes, agc_driver_packed_float16_native
VUID-VkShaderModuleCreateInfo-pCode-08741 (8 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_float16_fma_rounding, agc_driver_float16_unused_nan_payloads, agc_driver_packed_alu, agc_driver_packed_f16_denormal_modes, agc_driver_packed_float16_native
VUID-vkCmdCopyImage-srcImage-01548 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-vkCmdCopyImage-srcImage-01551 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (2 tests): agc_driver_prepared_exports, agc_driver_srgb8_color_target
VUID-VkPipelineShaderStageCreateInfo-module-parameter (1 test): agc_driver_float16_unused_nan_payloads
VUID-VkPipelineShaderStageCreateInfo-pSpecializationInfo-06849 (1 test): agc_driver_float16_unused_nan_payloads
VUID-VkSamplerCreateInfo-addressModeU-01079 (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeU-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeV-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeW-parameter (1 test): agc_driver_image_gather_volume

Skipped (48)

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_gcn_display_present: skipped, no display: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
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_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_stencil_ref_export: skipped, the device lacks VK_EXT_shader_stencil_export
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
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_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_backtoback: 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_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_dcc: 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_deferred: 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: Vu

Reproduce

git fetch origin 2b211c7231100e492394f2e6b5bb4aa5c5e04a5d 5b1d381d5ad5d9f923326b18479c3bd889cdd588
git checkout 5b1d381d5ad5d9f923326b18479c3bd889cdd588 && git merge --no-ff 2b211c7231100e492394f2e6b5bb4aa5c5e04a5d
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.}

[TechLevelZero]

✅ AMD GPU test: passed

PR head 2b211c7 merged onto main 5b1d381 · tested 2026-11-04 12:08 UTC · 36 min

GPU	AMD Radeon RX 6300 (RADV NAVI24), driver Mesa 25.2.8-0ubuntu0.24.04.4, Vulkan 1.4.318
Host	Ubuntu 26.04.1 LTS, kernel 7.0.0-42-generic, container ubuntu:24.04
Logs	build and test logs for this run
Build	✅ passed (Release, -DANYPS5_ENABLE_SPIRV_TOOLS=ON)
Tests	✅ 2,817 passed · 8 failed (0 new, 7 also fail on main, 1 unstable on main) · 59 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	2,884 of 3,883 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 (975 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 120.3 s, ⚠️ public_headers 87.6 s, ⚠️ agc_shader_sampler_limits 30.7 s

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

agc_driver_pipeline_journal: failed in 8 of the last 10 main runs

Failing on main too (7, not counted against this PR)

agc_driver_image_table
agc_driver_native_depth_bda_clear
agc_driver_native_multisample_depth
agc_submit_progress
agc_submit_progress_no_timeline
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (1, not in the main baseline's validation pass)

agc_driver_storage_flush_ordering_no_import

Validation IDs already present on main (8 IDs in 544 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (544 tests): agc_driver, agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, and 536 more
VUID-vkCmdCopyImage-srcImage-01548 (4 tests): agc_driver_depth_clear_hle, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-vkCmdCopyImage-srcImage-01551 (4 tests): agc_driver_depth_clear_hle, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (2 tests): agc_driver_prepared_exports, agc_driver_srgb8_color_target
VUID-VkSamplerCreateInfo-addressModeU-01079 (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeU-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeV-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeW-parameter (1 test): agc_driver_image_gather_volume

Skipped (59)

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_copied_writer_label_order: skipped, the device reports a minStorageBufferOffsetAlignment of 4, so the dispatch binds its record at +0x1004 in place and nothing is written back; the test needs an alignment above 4
agc_driver_float16_unused_nan_payloads: skipped, the device has no native f16 conversions with RTE rounding and kept f16 denormals
agc_driver_gcn_display_present: skipped, no display: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
agc_driver_image_table_vouched_lookups: skipped, the device reports 4 of 4 host-imported pages written after a GPU read, so a watched imported table cannot be vouched for
agc_driver_linear_target_direct_write_completion: skipped, the device does not keep host imports write-watched, so no completion stamp over the target is tracked
agc_driver_ngg_vertex_wave_reduction: skipped, s3 describes a 32-lane wave only on 32-lane subgroups, and the device has 64 lanes
agc_driver_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_unit_shadow: skipped, the device does not import watched host memory for unit shadows
agc_driver_unit_shadow_budget: skipped, the device does not import watched host memory for unit shadows
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_idle_publish_wait: skipped, the device reports 4 of 4 host-imported pages written after a GPU read, so a forced watch of the import reads its submissions as CPU stores
agc_shadowed_key_stores: skipped, the device imports no watched host memory under a unit shadow
agc_small_tiled_shadow_label: skipped, host imports are unavailable: the test block is not imported and watched
agc_storage_clear_hand_over: skipped, the device imports no watched host memory for the clear-key hand-over
agc_storage_pending_key_refresh: skipped, refreshes over a pending key store are not testable here
agc_unit_shadow: skipped, the device does not import watched host memory for unit shadows
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_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_backtoback: 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_controls: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not avai

Reproduce

git fetch origin 2b211c7231100e492394f2e6b5bb4aa5c5e04a5d 5b1d381d5ad5d9f923326b18479c3bd889cdd588
git checkout 5b1d381d5ad5d9f923326b18479c3bd889cdd588 && git merge --no-ff 2b211c7231100e492394f2e6b5bb4aa5c5e04a5d
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 12 --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 16 --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 AMD machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs.}

[imellag]

@oneandonlydean thanks, and thanks for the merge of main. The test's expectation was wrong on hosts where host imports are compared instead of watched, which is what the AMD log shows ([write-watch] host imports resolve write protection: yes ... imported ranges are compared). It is not something this PR's flush changes cause; fixed in a8306d5 (test only).

I reproduced it on macOS with APS5_WRITE_WATCH_IMPORTS=unwatch, which forces compared imports: the PR's head then fails with the same message as the bot. To see what main does there, I ran the same stored-again cases on main's driver (support branch + main b11c7bf84f, with the test's one NoteStampedWrites call swapped for NotePendingWrites so it builds) and logged whether the second store submitted the open batch:

case, compared imports	main waits
1 image stored again	no
2 images stored again	no
1 image under another stamped write	yes
1 image alone after a joint store of 2	no

So over a compared import the second store doesn't wait on main either, except for a stamped write that isn't the image's own. The test now checks the waits only over watched imports (ComparedHostImportCovers), keeps the "another stamped write" check everywhere, and still checks every texel after the stores in all cases.

macOS 27.2, Apple M4 Pro, MoltenVK 1.4.2, with ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation, the PR's focused set (ctest -R 'storage_flush_ordering|linear_readback_label|linear_target|agc_driver_storage_(cache_budget|one_slice_alias|read_during_idle_store|write_back)|storage_cache_retained|depth_storage_reuse|flip_stores_small'):

• default (watched imports): main 27 of 27, this head 28 of 28;
• APS5_WRITE_WATCH_IMPORTS=unwatch: main 27 of 27, this head 28 of 28;
• no VUID- in any log, and none has was not found but was requested by env var VK_INSTANCE_LAYERS.

I don't have an AMD/RADV machine, so a rerun of the bot would confirm it there.

AI-assisted: yes

[TechLevelZero]

✅ NVIDIA GPU test: passed

PR head a8306d5 merged onto main 5b1d381 · tested 2026-11-04 12:36 UTC · 12 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	✅ 2,833 passed · 3 failed (0 new, 3 also fail on main) · 48 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	2,884 of 3,883 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 (975 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 103.1 s

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

agc_driver_image_table
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (1, not in the main baseline's validation pass)

agc_driver_storage_flush_ordering_no_import

Validation IDs already present on main (13 IDs in 544 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (544 tests): agc_driver, agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, and 536 more
VUID-VkShaderModuleCreateInfo-pCode-08737 (8 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_float16_fma_rounding, agc_driver_float16_unused_nan_payloads, agc_driver_packed_alu, agc_driver_packed_f16_denormal_modes, agc_driver_packed_float16_native
VUID-VkShaderModuleCreateInfo-pCode-08739 (8 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_float16_fma_rounding, agc_driver_float16_unused_nan_payloads, agc_driver_packed_alu, agc_driver_packed_f16_denormal_modes, agc_driver_packed_float16_native
VUID-VkShaderModuleCreateInfo-pCode-08741 (8 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_float16_fma_rounding, agc_driver_float16_unused_nan_payloads, agc_driver_packed_alu, agc_driver_packed_f16_denormal_modes, agc_driver_packed_float16_native
VUID-vkCmdCopyImage-srcImage-01548 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-vkCmdCopyImage-srcImage-01551 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (2 tests): agc_driver_prepared_exports, agc_driver_srgb8_color_target
VUID-VkPipelineShaderStageCreateInfo-module-parameter (1 test): agc_driver_float16_unused_nan_payloads
VUID-VkPipelineShaderStageCreateInfo-pSpecializationInfo-06849 (1 test): agc_driver_float16_unused_nan_payloads
VUID-VkSamplerCreateInfo-addressModeU-01079 (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeU-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeV-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeW-parameter (1 test): agc_driver_image_gather_volume

Skipped (48)

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_gcn_display_present: skipped, no display: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
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_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_stencil_ref_export: skipped, the device lacks VK_EXT_shader_stencil_export
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
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_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_backtoback: 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_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_dcc: 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_deferred: 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: Vu

Reproduce

git fetch origin a8306d56f5c04286f1ff44e85ac453cfd023c3ad 5b1d381d5ad5d9f923326b18479c3bd889cdd588
git checkout 5b1d381d5ad5d9f923326b18479c3bd889cdd588 && git merge --no-ff a8306d56f5c04286f1ff44e85ac453cfd023c3ad
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.}

[TechLevelZero]

✅ AMD GPU test: passed

PR head a8306d5 merged onto main 5b1d381 · tested 2026-11-04 13:00 UTC · 37 min

GPU	AMD Radeon RX 6300 (RADV NAVI24), driver Mesa 25.2.8-0ubuntu0.24.04.4, Vulkan 1.4.318
Host	Ubuntu 26.04.1 LTS, kernel 7.0.0-42-generic, container ubuntu:24.04
Logs	build and test logs for this run
Build	✅ passed (Release, -DANYPS5_ENABLE_SPIRV_TOOLS=ON)
Tests	✅ 2,817 passed · 8 failed (0 new, 7 also fail on main, 1 unstable on main) · 59 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	2,884 of 3,883 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 (975 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 120.3 s, ⚠️ public_headers 83.6 s, ⚠️ agc_driver_image_sample_unnormalized 49.4 s, ⚠️ agc_driver_lazy_resolved_abi 43.2 s, ⚠️ agc_driver_image_gather_volume 30.6 s

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

agc_driver_pipeline_journal: failed in 8 of the last 10 main runs

Failing on main too (7, not counted against this PR)

agc_driver_image_table
agc_driver_native_depth_bda_clear
agc_driver_native_multisample_depth
agc_submit_progress
agc_submit_progress_no_timeline
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (1, not in the main baseline's validation pass)

agc_driver_storage_flush_ordering_no_import

Validation IDs already present on main (8 IDs in 544 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (544 tests): agc_driver, agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, and 536 more
VUID-vkCmdCopyImage-srcImage-01548 (4 tests): agc_driver_depth_clear_hle, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-vkCmdCopyImage-srcImage-01551 (4 tests): agc_driver_depth_clear_hle, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (2 tests): agc_driver_prepared_exports, agc_driver_srgb8_color_target
VUID-VkSamplerCreateInfo-addressModeU-01079 (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeU-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeV-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeW-parameter (1 test): agc_driver_image_gather_volume

Skipped (59)

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_copied_writer_label_order: skipped, the device reports a minStorageBufferOffsetAlignment of 4, so the dispatch binds its record at +0x1004 in place and nothing is written back; the test needs an alignment above 4
agc_driver_float16_unused_nan_payloads: skipped, the device has no native f16 conversions with RTE rounding and kept f16 denormals
agc_driver_gcn_display_present: skipped, no display: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
agc_driver_image_table_vouched_lookups: skipped, the device reports 4 of 4 host-imported pages written after a GPU read, so a watched imported table cannot be vouched for
agc_driver_linear_target_direct_write_completion: skipped, the device does not keep host imports write-watched, so no completion stamp over the target is tracked
agc_driver_ngg_vertex_wave_reduction: skipped, s3 describes a 32-lane wave only on 32-lane subgroups, and the device has 64 lanes
agc_driver_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_unit_shadow: skipped, the device does not import watched host memory for unit shadows
agc_driver_unit_shadow_budget: skipped, the device does not import watched host memory for unit shadows
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_idle_publish_wait: skipped, the device reports 4 of 4 host-imported pages written after a GPU read, so a forced watch of the import reads its submissions as CPU stores
agc_shadowed_key_stores: skipped, the device imports no watched host memory under a unit shadow
agc_small_tiled_shadow_label: skipped, host imports are unavailable: the test block is not imported and watched
agc_storage_clear_hand_over: skipped, the device imports no watched host memory for the clear-key hand-over
agc_storage_pending_key_refresh: skipped, refreshes over a pending key store are not testable here
agc_unit_shadow: skipped, the device does not import watched host memory for unit shadows
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_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_backtoback: 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_controls: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not avai

Reproduce

git fetch origin a8306d56f5c04286f1ff44e85ac453cfd023c3ad 5b1d381d5ad5d9f923326b18479c3bd889cdd588
git checkout 5b1d381d5ad5d9f923326b18479c3bd889cdd588 && git merge --no-ff a8306d56f5c04286f1ff44e85ac453cfd023c3ad
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 12 --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 16 --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 AMD machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs.}

[adam25136]

Windows and NVIDIA Blackwell run, for head a8306d56f5 merged into main b98c4af193 (git merge --no-ff, clean merge). This device keeps host imports watched rather than compared:

[write-watch] host imports of write-watched memory keep the write watch: yes (4096 of 4096 scratch pages reported written at the import, 0 by its 4-byte settle read, 0 after it, 0 after a full GPU read, 0 after a second, 0 after a GPU write, 1 after a CPU store); direct host imports of private memory are settled and stay watched, separate shared aliases retain guest write tracking

So this run covers the watched side of the test fix in a8306d56f5.

GPU	NVIDIA GeForce RTX 5080 (10de:2c02), driver 621.14 (32.0.16.2114), Vulkan 1.4.357
Host	Windows 10 Pro 22H2 (19045), i9-12900KF, WinLibs GCC 15.2.0 r7, Release, BUILD_TESTING=ON
Selection	2,405 tests, chosen by tools/select_tests.py select for the 7 changed files
Run	select_tests.py ctest build --parallel 4 --timeout 120 with ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1; main b98c4af193 ran its full suite (4,172 tests) the same way on this machine
Result	2,326 passed, 76 skipped, 3 failed, 0 new
Validation	The 1,361 agc_* tests of the selection again with VK_LAYER_KHRONOS_validation 1.4.367 (the Khronos nightly built from 3baa2b8), ctest -V --parallel 4, implicit layers disabled; main ran its 1,363 the same way. This change logs the same 151 messages in the same 23 tests as main, and no ID that main doesn't have for the same test. The new agc_driver_storage_flush_ordering_no_import logs none

agc_driver_storage_flush_ordering (storage flush ordering tests passed) and the new agc_driver_storage_flush_ordering_no_import (storage flush ordering stored-again tests passed) pass.

The 3 failures fail on main b98c4af193 here too:

• public_headers: five WKSecurityOrigin* declarations are missing on main since feat(libSceNKWebKit): implement WKHitTestResult property getters #15171 (see the comments there). Its 64 public_headers_compile_* tests are not run on either side, and they are 64 of the 76 skips.
• agc_driver_direct_write_completion is AGC: agc_driver_direct_write_completion fails on Windows since #13159 #15564.
• agc_driver_quad_pixel_mask is AGC: agc_driver_quad_pixel_mask samples the wrong mip level under a partial quad mask on Windows (RTX 5060 Ti and Radeon 860M) #14237; it fails on main with driver 621.14 and passed with 617.42.

The other 12 skips also skip on main here. The slowest selected test, agc_driver_pipeline_journal, took 24.3 s.

No title was run.

AI-assisted: yes. Claude Code ran the builds and tests and drafted this comment; the numbers are from the runs on this machine.

[imellag]

@adam25136 thanks for the Windows / RTX 5080 run. It covers the watched side of the test fix in a8306d56f5, which I couldn't check on macOS.

@MarcZX10 both open points from your last review are settled:

• perf(agc): wait on a flush only for writers over the image layers it stores #8887 merged on 2026-11-02, and this branch already contains it (main merge 41f3e3c20e, combined results in my comment of 11-02).
• On fix(agc): keep CPU stores made after a deferred DCC key store was recorded #6186, ee6aba54 agreed to perf(agc): wait on a flush only for writers over the image layers it stores #8887 → this PR → fix(agc): keep CPU stores made after a deferred DCC key store was recorded #6186 (comment) and will rebase onto this PR after it lands.

The description is updated to match (Depends on: none). Could you take another look?

The windows check was cancelled at the 45-minute job limit (job 126571891342), not failed; a re-run would be appreciated.

AI-assisted: yes