feat(agc): implement unnormalized S# clamp modes 4 and 5 (half border) by shahulmubeen-a · Pull Request #14964 · boykopovar/AnyPS5

shahulmubeen-a · 2026-11-02T05:36:50Z

What

A FORCE_UNNORMALIZED S# with clamp mode 4 (clamp-half-border) or 5 (mirror-once-half-border) on X or Y is refused today (after #14963):

guest sampler descriptor uses unnormalized coordinates with clamp mode 4 on X; only wrap, mirror, clamp-to-last-texel, mirror-once, clamp-to-border and mirror-once-border are implemented

These are the last two clamp modes an unnormalized S# refuses. The measurement is the one from #14963 (RDNA2 gfx1036 through tools/hw-oracle, image_sample_lz and image_sample_lz_o, 64x4 linear 32_FLOAT with texel 100 + x + 1000 * y, transparent black and opaque white borders, point and bilinear, a run with unequal filters and a run on a view from mip 1).

What the hardware does:

Point filter: mode 4 samples like mode 2 (clamp-to-last-texel) and mode 5 like mode 3 (mirror-once). It never reads the border.
Bilinear filter: mode 4 clamps the coordinate to [0, size of the level] and then reads the border past the edge. At or beyond an edge the result is half the edge texel and half the border colour. Mode 5 first mirrors (texel index i < 0 reads -i - 1, each tap on its own, as mode 7), then applies the same clamp at the far edge.
Filter choice: with unequal filters the magnification filter decides, as main already assumes for unnormalized samplers: magnification point with minification bilinear never reads the border, the reverse reads half of it.
Level size: on a view from mip 1 (32x2) the clamp is at 32, the size of the sampled level.
The texel offset is added first. Y behaves like X.

Measured rows (excerpt)

X, 64x4, v = 1.5 (row 1, texel = 1100 + x), transparent black border:

u               -64.5  -63.5   -1.0  -0.25    0.0   0.25    0.5  63.75   64.0   64.5  128.5
clamp 4 point    1100   1100   1100   1100   1100   1100   1100   1163   1163   1163   1163
clamp 4 linear    550    550    550    550    550    825   1100 872.25  581.5  581.5  581.5
clamp 5 point    1163   1163   1100   1100   1100   1100   1100   1163   1163   1163   1163
clamp 5 linear  581.5   1163 1100.5   1100   1100   1100   1100 872.25  581.5  581.5  581.5

With the white border: 550.5, 825.25, 872.5 and 582 in place of 550, 825, 872.25 and 581.5.

Y (u = 10.5, texel = 110 + 1000 * y), linear, transparent black: clamp 4 -9.5:55 0.0:55 0.5:110 3.75:2332.5 4.0:1555 8.5:1555; clamp 5 -9.5:1555 -3.5:3110 -1.0:610 0.0:110 4.5:1555.

Offsets (image_sample_lz_o, linear, transparent black): clamp 4 1.5 + (-3):550 -1.5 + 3:1101 62.5 + 3:581.5; clamp 5 2.5 + (-5):1102 63.5 + 1:581.5.

Unequal filters, clamp 4, transparent black, -1.0 / 0.25 / 64.0: magnification point, minification bilinear 1100 / 1100 / 1163; magnification bilinear, minification point 550 / 825 / 581.5.

View from mip 1 (32x2, v = 0.5, texel 5000 + x), linear, transparent black: clamp 4 -0.25:2500 0.25:3750 31.75:3773.25 32.0:2515.5 64.0:2515.5; clamp 5 -32.5:2515.5 -31.5:5031 0.0:5000 32.5:2515.5.

How the change implements this:

DecodeSamplerResource accepts modes 4 and 5 with unnormalized coordinates. It binds clamp-to-border for that axis when the magnification filter is bilinear and clamp-to-edge when it is point (the host sampler already uses the magnification filter for both).
ProveUnnormalized sets the mirror axis bit for mode 5, as for 1, 3 and 7, and a new per-sampler constant HalfBorderUnnormalizedBase (57792, the next free 64-aligned base after fix(agc): bind a T# minimum LOD through an S# that holds the LOD at 0 #12980's MinLodBase 57600 and VolumeGatherBase 57664 and fix(agc): apply a LOD bias beyond the device limit in the shader #14621's SamplerLodBiasBase 57728) carries the axes in mode 4 or 5.
In the sample branch that feat(agc): sample unnormalized S#s with mirror and mirror-once clamp modes #11126 added, the shader clamps the coordinate on those axes, after the texel offset and the mirroring, to [0, size], where size comes from OpImageQuerySizeLod at level 0 of the bound view, the level an unnormalized sample reads. With a point filter the clamp changes nothing, since clamp-to-edge reads the last texel at size anyway.

The model check in #14963's first comment (fold to the float below -c, at least 0.5, this clamp, then the host sampler with 8-bit bilinear weights) matches all 1660 measured values, including every mode 4 and 5 row and the mip-1 view.

The debt/libSceAgcDriver.md entry now covers modes 4 and 5; with this change no clamp mode throws for unnormalized coordinates. Anisotropy and MC_COORD_TRUNC still do.

No title I know of uses these modes, so Unblocks is none.

Tested

Windows 11 Pro 26300, NVIDIA GeForce RTX 4070 Ti SUPER (610.88), WinLibs GCC 15.2.0, Release. Tested head: 51aaebd on c1baa5b (#14963) on #11126's head 1c9c1fa.

agc_driver_image_sample_unnormalized gains six mode 4 and 5 samplers, checked like the others (all sample instructions, 1-level and 4-level views, and the view from mip 1, whose 6x3 size the clamp has to use):
- bilinear clamp-half-border on X and Y, opaque white;
- point clamp-half-border on X with mirror-once-half-border on Y, opaque white;
- bilinear mirror-once-half-border on X with clamp-half-border on Y, transparent black;
- point mirror-once-half-border on X with clamp-half-border on Y, opaque black;
- bilinear mirror on X with mirror-once-half-border on Y, opaque black;
- point magnification and bilinear minification with clamp-half-border on X and Y, which must sample as point.
The reference clamps a bilinear coordinate on a mode 4 axis to [0, size] and on a mode 5 axis to [-size, size] before the footprint, mirrors index i < 0 on mode 5, and returns the border colour past the edge for a bilinear mode 4 or 5.
- With the sources of c1baa5b, it fails with guest sampler descriptor uses unnormalized coordinates with clamp mode 4 on X; only wrap, mirror, clamp-to-last-texel, mirror-once, clamp-to-border and mirror-once-border are implemented.
- With SpirvImageEmitter.cpp from c1baa5b (no clamp in the shader), it fails with image_sample_lz, bilinear clamp-half-border on X and Y, opaque white, 1-level image: (-1.000000, 2.500000) offset (0, 0) component 0 is 255.000000, expected 127.500000.
agc_driver_graphics (the sampler decode test) expects modes 4 and 5 to decode to clamp-to-border with a bilinear magnification filter and clamp-to-edge with a point one, each with the other minification filter. With the sources of c1baa5b it fails with the same mode 4 refusal.
I ran 84 sampling and shader tests with ctest -V, VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation, ANYPS5_NO_SHADER_CACHE=1 and ANYPS5_REQUIRE_VULKAN=1 on main 926f7dd and on c1baa5b and 51aaebd. The regex was sample|gather|sampler|shader_memory|offset|unnormalized|shader_disk_cache|on_line|on_plane|on_volume|lod|^agc_driver_graphics$.
- The layer was loaded in every run: there is no was not found but was requested, and each log has the same 9 validation warnings (Undefined-Value-ShaderOutputNotConsumed-DynamicRendering).
- No log has a VUID- error.
- Every run passes 83 and skips agc_driver_sample_order_grid_reference, which skips itself on this device (it does not keep depth at custom sample locations).
Retested on the head with main merged in (05700f8, main ef0d852): the same regex now matches 104 tests. Run with the validation layer, which VK_LOADER_DEBUG=layer shows inserted as an instance and a device layer: 103 pass and agc_driver_sample_order_grid_reference skips itself. No VUID- lines.
tools/check_conventions.py --base origin/main passes.

Checklist

🤖 Generated with Claude Code

github-actions · 2026-11-02T06:02:38Z

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 12662	`core/shader/recompiler/PipelineSpecialization.hpp` L51
PR 12980	`core/shader/recompiler/PipelineSpecialization.hpp` L51

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 907	`core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 2434	`docs/dev/debt/libSceAgcDriver.md`
PR 5021	`core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 5371	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 5609	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp`
PR 5925	`docs/dev/debt/libSceAgcDriver.md`
PR 6186	`docs/dev/debt/libSceAgcDriver.md`
PR 6194	`docs/dev/debt/libSceAgcDriver.md`
PR 6223	`docs/dev/debt/libSceAgcDriver.md`
PR 6273	`docs/dev/debt/libSceAgcDriver.md`
PR 6280	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 6412	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 6481	`docs/dev/debt/libSceAgcDriver.md`
PR 6488	`core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 6548	`docs/dev/debt/libSceAgcDriver.md`
PR 6804	`core/libs/prx/libSceAgcDriver/Graphics/src/GuestSamplerResource.cpp` `core/libs/prx/libSceAgcDriver/tests/GuestSamplerResource.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 6808	`docs/dev/debt/libSceAgcDriver.md`
PR 6987	`docs/dev/debt/libSceAgcDriver.md`
PR 6989	`docs/dev/debt/libSceAgcDriver.md`
PR 6995	`docs/dev/debt/libSceAgcDriver.md`
PR 7002	`docs/dev/debt/libSceAgcDriver.md`
PR 7085	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 7223	`core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 7256	`docs/dev/debt/libSceAgcDriver.md`
PR 7449	`docs/dev/debt/libSceAgcDriver.md`
PR 7604	`docs/dev/debt/libSceAgcDriver.md`
PR 7642	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 7668	`docs/dev/debt/libSceAgcDriver.md`
PR 7689	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp`
PR 8592	`docs/dev/debt/libSceAgcDriver.md`
PR 8707	`docs/dev/debt/libSceAgcDriver.md`
PR 9020	`docs/dev/debt/libSceAgcDriver.md`
PR 9183	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 9318	`docs/dev/debt/libSceAgcDriver.md`
PR 9335	`docs/dev/debt/libSceAgcDriver.md`
PR 9356	`docs/dev/debt/libSceAgcDriver.md`
PR 9387	`docs/dev/debt/libSceAgcDriver.md`
PR 9592	`docs/dev/debt/libSceAgcDriver.md`
PR 9672	`docs/dev/debt/libSceAgcDriver.md`
PR 9673	`docs/dev/debt/libSceAgcDriver.md`
PR 9675	`docs/dev/debt/libSceAgcDriver.md`
PR 9700	`docs/dev/debt/libSceAgcDriver.md`
PR 9748	`docs/dev/debt/libSceAgcDriver.md`
PR 9852	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp`
PR 9889	`core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 9895	`docs/dev/debt/libSceAgcDriver.md`
PR 9897	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp`
PR 9898	`docs/dev/debt/libSceAgcDriver.md`
PR 9915	`docs/dev/debt/libSceAgcDriver.md`
PR 9950	`docs/dev/debt/libSceAgcDriver.md`
PR 10049	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 10193	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 10233	`docs/dev/debt/libSceAgcDriver.md`
PR 10256	`docs/dev/debt/libSceAgcDriver.md`
PR 10359	`docs/dev/debt/libSceAgcDriver.md`
PR 10743	`docs/dev/debt/libSceAgcDriver.md`
PR 10867	`core/libs/prx/libSceAgcDriver/tests/execution/ImageSampleUnnormalized.cpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 11825	`docs/dev/debt/libSceAgcDriver.md`
PR 11998	`docs/dev/debt/libSceAgcDriver.md`
PR 12120	`docs/dev/debt/libSceAgcDriver.md`
PR 12565	`docs/dev/debt/libSceAgcDriver.md`
PR 12567	`docs/dev/debt/libSceAgcDriver.md`
PR 12790	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 13599	`docs/dev/debt/libSceAgcDriver.md`
PR 13707	`docs/dev/debt/libSceAgcDriver.md`
PR 13863	`docs/dev/debt/libSceAgcDriver.md`
PR 13920	`docs/dev/debt/libSceAgcDriver.md`
PR 13928	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 13929	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 13963	`docs/dev/debt/libSceAgcDriver.md`
PR 13994	`docs/dev/debt/libSceAgcDriver.md`
PR 14005	`docs/dev/debt/libSceAgcDriver.md`
PR 14032	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp`
PR 14212	`docs/dev/debt/libSceAgcDriver.md`
PR 14302	`docs/dev/debt/libSceAgcDriver.md`
PR 14311	`docs/dev/debt/libSceAgcDriver.md`
PR 14344	`docs/dev/debt/libSceAgcDriver.md`
PR 14561	`docs/dev/debt/libSceAgcDriver.md`
PR 14621	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `core/shader/recompiler/PipelineSpecialization.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 14702	`core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 14853	`core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 14943	`docs/dev/debt/libSceAgcDriver.md`
PR 14944	`core/shader/recompiler/Optimization/src/DescriptorBindingBuilder.cpp` `docs/dev/debt/libSceAgcDriver.md`
PR 14976	`docs/dev/debt/libSceAgcDriver.md`

TechLevelZero · 2026-11-02T06:17:48Z

✅ NVIDIA GPU test: passed

PR head 51aaebd merged onto main 93dbd0f · tested 2026-11-02 06:17 UTC · 22 min


GPU	NVIDIA RTX 2000 Ada Generation, driver 595.99.02, Vulkan 1.4.329
Host	Ubuntu 26.04.1 LTS, kernel 7.0.0-38-generic, container ubuntu:24.04
Logs	build and test logs for this run
Build	✅ passed (Release, `-DANYPS5_ENABLE_SPIRV_TOOLS=ON`)
Tests	✅ 2,451 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	2,501 of 3,165 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 (932 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ `agc_driver_pipeline_journal` 120.1 s, ⚠️ `agc_driver_lookup_log_fits_build` 79.7 s, ⚠️ `agc_driver_image_sample_unnormalized` 52.9 s, ⚠️ `agc_driver_graphics_maps_fallback` 30.9 s, ⚠️ `agc_driver_graphics` 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 1 of the last 10 main runs
agc_submit_progress: failed in 6 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

Validation IDs already present on main (5 IDs in 484 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (484 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 476 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 (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 51aaebd7d658689b4f325fec012ff35ce1fae53a 93dbd0f82aea684e95f15f06e7d42c0064af18cf
git checkout 93dbd0f82aea684e95f15f06e7d42c0064af18cf && git merge --no-ff 51aaebd7d658689b4f325fec012ff35ce1fae53a
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.}

JJenkx · 2026-11-02T16:07:01Z

AMD check of this stack, since it was tested on NVIDIA only: Radeon RX 6700 XT (RDNA 2), AMD proprietary driver 26.10.1, Windows 11 Enterprise 10.0.22621, WinLibs GCC 15.2.0 UCRT r7, Release.

Exact head 51aaebd7d6, which includes #14963's c1baa5b783, against its base 1c9c1faf4b, #11126's head and already in main. Each was built in its own tree, and runs went through the shared GPU queue with VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation, ANYPS5_NO_SHADER_CACHE=1 and ANYPS5_REQUIRE_VULKAN=1.

ctest -R '(image_sample|sampler|border|clamp)' --repeat until-fail:5:

Arm	Tests	Result	VUIDs
base `1c9c1faf4b`	41	205/205 pass	none
this stack	41	205/205 pass	none

agc_driver_image_sample_unnormalized runs every case on this driver. It skips only when there is no Vulkan device, and it printed image sample unnormalized tests passed in all 5 runs. That covers this PR's mode 4 and 5 and feat(agc): implement unnormalized S# clamp mode 7 (mirror-once-border) #14963's mode 7 samplers, across the sample instructions, the 1- and 4-level views and the view from mip 1.
The other 40 sampler, border and clamp tests pass unchanged.
Full suite on this head: 3,266 tests: 3,219 pass and 19 skip; the validation layer was inserted 1,096 times. The 28 failures are exactly the set main fails on this machine: 26 video_out_flip_* that need a display (it was off), agc_driver_sample_order_grid_reference (fix(agc): bind the sample order depth set after its pipeline #9503) and ngs2_pitchshift (fix(libSceNgs2): compute the PitchShift ratio with libc's msun exp2f #12864). The VUIDs are the same three main reports here. The stack adds no failure.

This covers the AMD proprietary driver on RDNA 2. RADV and a PS5 were not run.

🤖 Generated with Claude Code

BunnyFl

An RDNA 4 run, to go with JJenkx's RDNA 2 one. Windows 11, AMD Radeon RX 9070 XT (RDNA 4), AMD proprietary driver 32.0.32015.7029, WinLibs GCC, BUILD_TESTING=ON, with VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation, ANYPS5_REQUIRE_VULKAN=1 and ANYPS5_NO_SHADER_CACHE=1.

Head 51aaebd7d6 (which carries #14963's c1baa5b783), merged into current main:

agc_driver_image_sample_unnormalized and agc_driver_graphics pass, with 0 VUIDs.
With main's GuestSamplerResource.cpp and the four recompiler files restored, both fail, with 0 VUIDs. image_sample_unnormalized stops on guest sampler descriptor uses unnormalized coordinates with clamp mode 7 on X.

So the emitted modes 4, 5 and 7 give the measured gfx1036 values on RDNA 4 as well. The half-border bilinear result and the per-tap mirroring hold on this generation too.

RDNA2 (gfx1036) mirrors texel index i < 0 to -i - 1 under clamp mode 7, as under modes 1 and 3, and reads the border colour past the edge. Bind clamp-to-border for a mode 7 axis and fold its coordinate with the existing mirror specialization. Raise a folded coordinate to at least 0.5: each bilinear tap is mirrored on its own, so the tap at index -1 reads texel 0 again where clamp-to-border would read the border. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

RDNA2 (gfx1036) samples clamp mode 4 like clamp-to-last-texel and mode 5 like mirror-once with a point filter. With a bilinear filter both clamp the coordinate, mode 5 after mirroring, to [0, size of the level] and read the border colour past the edge, so an edge sample is half texel and half border. The magnification filter decides, as for the other unnormalized samplers. Bind clamp-to-border for a mode 4 or 5 axis when the magnification filter is bilinear and clamp-to-edge otherwise, set the mirror axis bit for mode 5, and carry the mode 4 and 5 axes in a new specialization constant (HalfBorderUnnormalizedBase, 57728). The shader clamps those axes to the size of level 0 of the bound view, the level an unnormalized sample reads. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

RNK123 · 2026-11-04T04:33:56Z

RADV/RX 9060 XT: agc_driver_image_sample_unnormalized passes 3/3 on this head merged onto main, including the mode 4 and 5 half-border cases. The previous AMD report explicitly left RADV untested.

Head 9b4cca8cb6 merged onto main c2eafe4461. The existing unnormalized-sampling test passed 3/3 on main and 3/3 on the merged tree; the test fixture differs because this PR adds the half-border cases, so the baseline result does not cover those cases. No skipped case was reported. The Radeon and Khronos validation were confirmed in each run, without VUID- or SYNC-HAZARD- diagnostics. A pre-existing unrelated LSFG implicit-layer loading failure appeared on both arms; the requested Khronos layer loaded.

Sensitivity check: on the merged tree, forcing the half-border axis mask to zero at the shader-emission call made the test fail 3/3 for the bilinear X/Y clamp-half-border sample: component 0 was 255 instead of 127.5. Restoring the shader emitter passed 3/3. This is a targeted mutation, not an unmodified-main comparison for the new cases.

Host: CachyOS Linux 7.3.1-1-cachyos, GCC 16.2.1, Release, RX 9060 XT (RADV GFX1200), Mesa 26.2.5-arch3.1, Vulkan 1.4.354, Khronos validation 1.4.363.

Reproduction

Build agc_driver_image_sample_unnormalized_tests on main and the PR merged onto it, then run three times per tree:

export ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1
export VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation VK_LOADER_DEBUG=layer
export VK_LAYER_PATH="$HOME/.local/share/anyps5-tester/vk-layers/usr/share/vulkan/explicit_layer.d"
ctest --test-dir build-main -V -R '^agc_driver_image_sample_unnormalized$'

Limits: one focused synthetic sampling test, no title, performance measurement or PS5 hardware. The mutation demonstrates sensitivity for the stated bilinear case, not independent confirmation of console behavior. AI-assisted independent hardware test run (GPT-6.1 Terra).

TechLevelZero · 2026-11-04T04:37:10Z

✅ NVIDIA GPU test: passed

PR head 9b4cca8 merged onto main 1c8c0ef · tested 2026-11-04 04:37 UTC · 31 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	✅ 2,732 passed · 5 failed (0 new, 3 also fail on main, 2 unstable on main) · 50 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	2,787 of 3,842 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 (952 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ `agc_driver_pipeline_journal` 120.5 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 4 of the last 10 main runs
agc_submit_progress_no_timeline: failed in 4 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

Validation IDs already present on main (8 IDs in 531 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (531 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 523 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_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 (50)

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_driver_vertex_elected_atomic_max: skipped, the device runs vertex programs in subgroups, so they keep their cross-lane reductions
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_import_settle_initial_dcc: skipped, the device has no deferred settle of host imports: imports under 1 MiB wait for one only on Windows
agc_unit_shadow_completion_store: skipped, the device reports watched host memory written when it imports or reads it
guest_http_uri_locale: 
guest_net_ctl: skipped, the host has no IPv4 address
guest_video_out_color_space: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_hdr: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_high_frame_rate: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_output_mode: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_resolution_status: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_wqhd_detection: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
video_out_flip_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
vide

Reproduce

git fetch origin 9b4cca8cb664e64c3b3e13e82e5df71a801f4363 1c8c0efc8c41314402e6966da5bef6ea3d639d10
git checkout 1c8c0efc8c41314402e6966da5bef6ea3d639d10 && git merge --no-ff 9b4cca8cb664e64c3b3e13e82e5df71a801f4363
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 · 2026-11-04T05:33:16Z

✅ AMD GPU test: passed

PR head 9b4cca8 merged onto main 1c8c0ef · tested 2026-11-04 05:33 UTC · 40 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	✅ 2,718 passed · 8 failed (0 new, 7 also fail on main, 1 unstable on main) · 61 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	2,787 of 3,842 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 (952 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ `agc_driver_pipeline_journal` 120.4 s, ⚠️ `agc_driver_image_sample_unnormalized` 70.6 s, ⚠️ `public_headers` 69.7 s, ⚠️ `agc_shader_sampler_limits` 38.8 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

Validation IDs already present on main (8 IDs in 531 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (531 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 523 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 (1 test): 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 (61)

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_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_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_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_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 d

Reproduce

git fetch origin 9b4cca8cb664e64c3b3e13e82e5df71a801f4363 1c8c0efc8c41314402e6966da5bef6ea3d639d10
git checkout 1c8c0efc8c41314402e6966da5bef6ea3d639d10 && git merge --no-ff 9b4cca8cb664e64c3b3e13e82e5df71a801f4363
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.}

BunnyFl · 2026-11-04T15:15:52Z

@shahulmubeen-a @danielhorngt a heads-up: this PR and #14621 now take the same specialization constant base.

This PR, 9b4cca8cb6: core/shader/recompiler/PipelineSpecialization.hpp:54 is HalfBorderUnnormalizedBase = 57728u.
fix(agc): apply a LOD bias beyond the device limit in the shader #14621, d52b3c3184: core/shader/recompiler/PipelineSpecialization.hpp:54 is SamplerLodBiasBase = 57728u.

On main b11c7bf84f the last base is VolumeGatherBase = 57664u (:53). Both PRs add their line right after it. I merged #14964 and then #14621 onto that main, and git stops with a conflict in PipelineSpecialization.hpp, as well as DescriptorBindingBuilder.cpp and docs/dev/debt/libSceAgcDriver.md. So the IDs won't be shared silently, but whichever merges second has to resolve it. Its author then needs the next free base (57792 if nothing else has taken it) and a rerun of its tests. Could you two agree which goes first, and record it with Depends on on the other?

BunnyFl

@shahulmubeen-a re-checked at 9b4cca8cb6. git range-diff of my approved 51aaebd7d6 against this head shows one change to this PR's own patch: HalfBorderUnnormalizedBase moves from 57600 to 57728 (core/shader/recompiler/PipelineSpecialization.hpp:54), because main now uses 57600 for MinLodBase (:52). #14963's commit underneath is unchanged.

What I ran: head 9b4cca8cb6 merged onto main b11c7bf84f (merge bd34bba607), Windows 11, WinLibs GCC, Release, AMD Radeon RX 9070 XT, driver 32.0.32015.12012, with ANYPS5_REQUIRE_VULKAN=1, ANYPS5_NO_SHADER_CACHE=1 and VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation:

agc_driver_image_sample_unnormalized: passes, including the mode 4 and 5 half-border cases. The log has no VUID-, SYNC-HAZARD or missing-layer line.
check_conventions.py --base upstream/main passes.
pr_overlap.py --branch: 63 open PRs touch the same files, none the same exports. The 57728 clash with #14621 is in my note above. Whichever of you lands second takes the next free base.

shahulmubeen-a · 2026-11-04T17:04:19Z

@BunnyFl thanks for catching this. @danielhorngt, a proposal for the order: #14964 first, then #14621 on top of it.

feat(agc): implement unnormalized S# clamp modes 4 and 5 (half border) #14964 is approved with no open change requests, and 9b4cca8cb6 passes on NVIDIA, AMD and RADV. It already waits on feat(agc): implement unnormalized S# clamp mode 7 (mirror-once-border) #14963 (same stack), so it won't land before that one.
fix(agc): apply a LOD bias beyond the device limit in the shader #14621 still has an open change request from 11-03.

If that works for you, #14621 would move SamplerLodBiasBase to 57792 and record Depends on: #14964. If #14621 is ready first, I'm happy to swap: I'll move HalfBorderUnnormalizedBase to 57792, resolve the DescriptorBindingBuilder.cpp and debt-doc conflicts, rerun the unnormalized sampler tests with the validation layer and add Depends on: #14621 here.

Written by Claude Code, verified by me.

BunnyFl · 2026-11-05T15:47:02Z

@shahulmubeen-a a heads-up on numbering, no change asked of this PR as it stands: HalfBorderUnnormalizedBase = 57728u (core/shader/recompiler/PipelineSpecialization.hpp:54 at 9b4cca8cb6) is the same base #14621 takes for SamplerLodBiasBase at the same line, and #10489 also moves FmaskFormatBase to 57728u. All three are free on main, where the last base is VolumeGatherBase = 57664u. Whichever lands second will need a free base. main has no static_assert that these 64-wide ranges don't overlap (core/shader/recompiler/PipelineSpecialization.hpp:48-53 on main): a merge resolution that keeps both lines would still build. cc @danielhorngt (#14621), @EagleBearerYS (#10489).

AI-assisted: yes (Claude Code)

shahulmubeen-a · 2026-11-06T04:11:29Z

Thanks man

TechLevelZero · 2026-11-06T05:07:55Z

✅ NVIDIA GPU test: passed

PR head 05700f8 merged onto main f8167f3 · tested 2026-11-06 05:07 UTC · 21 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	✅ 2,926 passed · 4 failed (0 new, 3 also fail on main, 1 unstable on main) · 56 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	2,986 of 4,158 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 (987 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ `agc_driver_pipeline_journal` 120.1 s, ⚠️ `agc_driver_image_sample_unnormalized` 33.2 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 1 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

Validation IDs already present on main (9 IDs in 595 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (595 tests): agc_driver, agc_driver_address_draw_space_reuse, 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, and 587 more
VUID-VkShaderModuleCreateInfo-pCode-08737 (14 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_f16_fma_denormal_modes, agc_driver_f16_fma_invalid_nan, agc_driver_float16_fma_rounding, agc_driver_float16_modifiers, agc_driver_float16_native, and 6 more
VUID-VkShaderModuleCreateInfo-pCode-08739 (14 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_f16_fma_denormal_modes, agc_driver_f16_fma_invalid_nan, agc_driver_float16_fma_rounding, agc_driver_float16_modifiers, agc_driver_float16_native, and 6 more
VUID-VkShaderModuleCreateInfo-pCode-08741 (14 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_f16_fma_denormal_modes, agc_driver_f16_fma_invalid_nan, agc_driver_float16_fma_rounding, agc_driver_float16_modifiers, agc_driver_float16_native, and 6 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 (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

Skipped (56)

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_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_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_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_net_ctl_disconnected: 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_device_capability: 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_host_resolution_disabled: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_host_resolution_four_k: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_host_resolution_height_only: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_host_resolution_limit: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_host_resolution_malformed: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_host_resolution_width_only: 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 de

Reproduce

git fetch origin 05700f8da77e15b9000cdf80d22f5b54f529d9d8 f8167f32cd3795a776b955aa10e8284d4de67aef
git checkout f8167f32cd3795a776b955aa10e8284d4de67aef && git merge --no-ff 05700f8da77e15b9000cdf80d22f5b54f529d9d8
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.}

github-actions Bot added shader core/shader docs agc libSceAgc* labels Nov 2, 2026

JJenkx mentioned this pull request Nov 2, 2026

feat(agc): implement unnormalized S# clamp mode 7 (mirror-once-border) #14963

Open

12 tasks done

BunnyFl approved these changes Nov 2, 2026

View reviewed changes

github-actions Bot added the needs rebase Conflicts with main, rebase needed label Nov 2, 2026

shahulmubeen-a and others added 2 commits November 4, 2026 08:55

shahulmubeen-a force-pushed the feat/agc-unnormalized-half-border branch from 51aaebd to 9b4cca8 Compare November 4, 2026 03:53

github-actions Bot removed the needs rebase Conflicts with main, rebase needed label Nov 4, 2026

BunnyFl mentioned this pull request Nov 4, 2026

fix(agc): apply a LOD bias beyond the device limit in the shader #14621

Merged

12 tasks done

BunnyFl approved these changes Nov 4, 2026

View reviewed changes

EagleBearerYS mentioned this pull request Nov 5, 2026

feat(shader): raise the image limit to 128 and name the counts in its refusals #10489

Open

12 tasks done

github-actions Bot added the needs rebase Conflicts with main, rebase needed label Nov 5, 2026

oneandonlydean self-assigned this Nov 5, 2026

oneandonlydean added 2 commits November 6, 2026 13:08

Merge branch 'main'

d8ba7cb

Merge branch 'main'

05700f8

github-actions Bot removed the needs rebase Conflicts with main, rebase needed label Nov 6, 2026

Conversation

shahulmubeen-a commented Nov 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Tested

Checklist

Uh oh!

github-actions Bot commented Nov 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

TechLevelZero commented Nov 2, 2026

✅ NVIDIA GPU test: passed

Uh oh!

JJenkx commented Nov 2, 2026

Uh oh!

BunnyFl left a comment

Choose a reason for hiding this comment

Uh oh!

RNK123 commented Nov 4, 2026

Uh oh!

TechLevelZero commented Nov 4, 2026

✅ NVIDIA GPU test: passed

Uh oh!

TechLevelZero commented Nov 4, 2026

✅ AMD GPU test: passed

Uh oh!

BunnyFl commented Nov 4, 2026

Uh oh!

BunnyFl left a comment

Choose a reason for hiding this comment

Uh oh!

shahulmubeen-a commented Nov 4, 2026

Uh oh!

BunnyFl commented Nov 5, 2026

Uh oh!

shahulmubeen-a commented Nov 6, 2026

Uh oh!

TechLevelZero commented Nov 6, 2026

✅ NVIDIA GPU test: passed

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

shahulmubeen-a commented Nov 2, 2026 •

edited

Loading

github-actions Bot commented Nov 2, 2026 •

edited

Loading