Repository navigation
Conversation
MarcZX10
left a comment
MarcZX10
left a comment
There was a problem hiding this comment.
The RTX 4070 Ti SUPER run is detailed and the new alias regression is meaningful. Before approval, the AGC validation rule still needs the same focused tests on both current main and this head, using ANYPS5_REQUIRE_VULKAN=1, ANYPS5_NO_SHADER_CACHE=1, and VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation, with logs compared for newly introduced VUID- errors. The description reports the red main behavior and the PR-side 51 checks, confirms the layer loaded and no final VUIDs, but does not show the equivalent main-side focused run/log or establish that the comparison used the same test set and environment. Please add those baseline results (including confirmation the layer was found and the VUID comparison), or explain the exact equivalent baseline procedure. I could not run this locally because the required GPU is unavailable here.
|
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.
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.
|
|
Intel run, a second vendor next to the RTX 4070 Ti SUPER. Setup: Intel Arc Graphics (Meteor Lake,
AI-assisted: yes (Claude Code) |
|
@MarcZX10 I added the requested baseline comparison to the Tested section. On this RTX 4070 Ti SUPER Windows host, current |
Dismissing a stale review: the requested changes were addressed after it (new head and/or updated description with the asked-for runs) and the merge queue blocks on it. Please request changes again if anything is still open.
What
Binding a new depth attachment whose plane or HTILE overlapped an active native multisample depth owner was rejected as a layout/metadata change. Even an unwritten owner could not be replaced by another valid layout over the same guest memory.
Use the existing native-depth guest-access transition before constructing the new attachment: publish supported written contents, retire intersecting owners, then validate/create the new layout. Keep the existing attachment unchanged when its identity matches. Release the surface-cache lock before publication; the GPU lock continues to serialize the transition. Unsupported publication paths still throw.
Tested
Windows 11 Pro build 26300, WinLibs GCC 15.2.0, Release, RTX 4070 Ti SUPER, NVIDIA 617.42.
native multisample depth attachment layout or metadata changed.external GPU writes over native multisample depth are not implemented. Menu/gameplay remain unconfirmed; no speedup claim. Full CTest suite not run.Overlap review: 28 shared-file PR diffs inspected. #6808 adds overlap diagnostics, #7256 changes creation-lock placement, #9533/#9717/#9742 add separate stencil handling, and #9950 adds written D32 publication. Their nearby edits need normal merge coordination; they do not provide this attachment-alias transition.
Independent validation of the published commit: Published head
e393705239a3104846c1a91548e7c1fc9a2f9a38: full Windows Release build and native multisample depth, depth-storage dispatch, native-depth BDA clear/retirement/clear/refusal checks and patched-library validation pass on a clean tracked tree without overlays (7/7, zero skips). Vulkan is required, shader caches are disabled, and loader logs confirm actual Khronos instance/device validation layers; no VUIDs. Hardware and toolchain are the Windows/NVIDIA configuration listed above. This independently validates the published commit in addition to the earlier integration runs. The complete CTest suite was not run.Review baseline comparison (2026-10-24): I ran the same seven focused CTest names on current
main7966e590ef7cb1054500eef513f99c59e05547b0and compared its log with the clean, previously run published heade393705239a3104846c1a91548e7c1fc9a2f9a38. Both runs used this Windows 11 / RTX 4070 Ti SUPER host, Release / WinLibs GCC 15.2.0,ANYPS5_REQUIRE_VULKAN=1,ANYPS5_NO_SHADER_CACHE=1,ANYPS5_GPU=RTX 4070 Ti SUPER,VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation, the same Khronos layer path andVK_LOADER_DEBUG=error,warn,layer. The main-side build was limited to the four targets required by the selected tests; the PR head also passed the full Release build reported above.ctest --test-dir <build> -R '^(agc_driver_native_multisample_depth|agc_driver_depth_storage_dispatch|agc_driver_native_depth_(bda_clear|retirement|clears|clear_refusals)|patched_libraries)$' -j1 -VmainVK_LAYER_KHRONOS_validationat instance and device levelVK_LAYER_KHRONOS_validationat instance and device levelVUID-identifiers / validation warnings or errors / missing requested layeragc_driver_native_multisample_depthnative depth attachment alias changes layout while retaining held GPU viewsBoth logs report
Selected GPU name=NVIDIA GeForce RTX 4070 Ti SUPER. The matched JUnit test names and results, loader insertion lines and VUID scans were compared directly. The earlier red control onmainwith the new fixture overlaid is reported above; this baseline comparison uses each revision's own clean source.Checklist
main; no other open PR implements the same functionsNotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file.mdfiles or images added to the repository (attach them to this PR)