fix(shader): sample a 3D T# through DIM 2D_ARRAY as a volume by thuanquang · Pull Request #7381 · boykopovar/AnyPS5

thuanquang · 2026-10-18T03:48:15Z

What

Yakuza 0 Director's Cut (PPSA31327) stops a few seconds into its first cutscene (discussion #6722):

shader 0xed00a68f00 stage 5: image descriptor is incompatible with the static runtime image interface: descriptor format 56 decodes to {class 1, dim 4, conversion 0, packed 0, cube 0, depthBits 0, unorm16 0, srgb 0}

The pixel shader samples a 3D RGBA8 T# with DIM 2D_ARRAY image instructions. #5555 accepts 2D-array loads and stores on a 3D T#, and #5461 accepts biased DIM 2D samples of a one-slice volume, but a 2D-array sample of a volume has no runtime mode.

I measured what RDNA2 does with the hardware oracle template, run through the Windows driver's comgr and HIP runtime on an RX 6700 XT (gfx1031). On 16x16x4 R32_FLOAT 3D T#s with one and three levels, image_sample, image_sample_l and image_sample_lz with DIM 2D_ARRAY returned the same value as the same instruction with DIM 3D in all 256 lanes of 66 cases:

point, bilinear and z-linear filters, wrap addressing;
r from -0.125 to 3;
quad steps of 1 to 4 texels in u and v and of 0 to 4 slices in r (so implicit LOD uses r's derivative too);
LOD 1 for image_sample_l.

So the third address VGPR is the normalized r coordinate, addressed by the sampler's Z mode, not a layer index. These cases used wrap only: r = 1.0 reads slice 0 and r = 0.375 slice 1.

This change adds a Dim3D runtime mode for a 2D-array sampled image whose accesses are all image_sample, image_sample_l or image_sample_lz (new tracker flag volumeArraySampleCompatible, the same instruction set as cubeFaceCompatible). The emitter already samples a Dim3D mode with the instruction's three coordinates, and the driver already binds a 3D T# as a 3D view with addressModeW from CLAMP_Z, so neither needed a change.

These keep throwing, because they are not measured:

bias, LOD clamp, derivatives, offsets and comparison;
gathers and image_get_lod;
an image that is both loaded and sampled.

ImageResource has a new field, so the shader disk cache format goes from 40 to 41 and the ImageResource field count from 39 to 40. (main took 38, 39 and 40 for other format changes, and field count 39 for #5958's field, after this PR was opened.)

Overlap:

fix(agc): keep texture images inside their extent's mip chain #6280, feat(agc): back extra guest mip levels with storage images #7085 and fix(shader): answer image_get_resinfo on any T# dimension #7550 each add an ImageResource field and edit the same fieldCount<ImageResource>() line. As @klosixgaming-arch found, two PRs that make the same count edit merge without a conflict, and the result fails to build. Whichever lands second has to raise the count again.
fix(shader): store through DIM 2D image_store to a cube T#'s BASE_ARRAY face #6799 rewrites the cubeFaceCompatible line that the new flag's line follows, and the debt entry after the new one.
fix(agc): keep the guest's sample order where pixel (0,0)'s locations stand in for a 2x2 grid #6683 changes FormatVersion too.

Tested

Windows 11 26200, AMD Radeon RX 6700 XT (gfx1031), driver 32.0.21045.5002, WinLibs GCC 15.2.0 r7, Release, BUILD_TESTING=ON, ANYPS5_REQUIRE_VULKAN=1:

New agc_driver_image_array_sample_on_volume: a compute shader on a 16x16x4 RGBA8 3D T# with a repeat point sampler.
- image_sample_lz with DIM 3D and with DIM 2D_ARRAY, and image_sample_l with DIM 2D_ARRAY, are given r = (z + 0.5) / 4 + k for k from -1 to 2, and must return slice z.
- On a two-level volume, image_sample_l with LOD 1 must return slice z / 2 of level 1, with DIM 3D and with DIM 2D_ARRAY. Each level has its own markers, so reading the LOD from the wrong register would fail.
- image_sample_b with DIM 2D_ARRAY must still be refused.
- On main 2626e5ee7 it fails with the title's message: image descriptor is incompatible with the static runtime image interface: descriptor format 56 decodes to {class 1, dim 4, conversion 0, packed 0, cube 0, depthBits 0, unorm16 0, srgb 0}.
agc_shader_memory (verifyVolumeArrayLoadMode): the flag selects the Dim3D mode for a 3D T#, a 2D-array T# still selects the 2D-array mode, and without the flag a sample on a 3D T# is still refused.
agc_shader_disk_cache: the new field round-trips.
Rebased onto main 7f596df48 on 2026-10-19, head 1b32568. main had taken FormatVersion 40 and fieldCount<ImageResource>() 39 (feat(shader): sample converted unorm images in the shader #5958) for other fields, so this now raises them to 41 and 40. Before this rebase the count edit merged silently with main's identical one and the merge failed to build, as @klosixgaming-arch and @johandavidcm found. agc_driver_image_array_sample_on_volume, agc_shader_memory and agc_shader_disk_cache pass 2 of 2 on 1b32568 on the RX 6700 XT.
Full ctest --parallel 4 --timeout 120 on main 7f596df48 with this change: 1604 tests, 17 fail, none from this PR:
- 7 Python-based tests that time out on this machine's slow Python startup: conventions_* and hardware_oracle_compare (CI's conventions job passes on this head);
- guest_pthread_rwlock and guest_file_lock_process, under load;
- main's three known AMD failures: agc_driver_lds_mixed_width_atomics (fix(shader): release the LDS lock with an exchange and drop fences inside it #6976), agc_driver_lds_lock_budget (AGC: AMD device-backed LDS lock-budget case returns half of expected sum #7344), agc_driver_layered_volume_vertex_no_gpl (AMD Windows: GPL-mode cache failures (alpha/wave32 pass after local driver update) #4768);
- agc_driver_layered_volume_draw, agc_driver_layered_volume_vertex, agc_driver_draw_cache_reuse_off and agc_driver_stencil_replay_no_gpl (VK_ERROR_DEVICE_LOST at the first submit). All four pass from renamed copies of their executables, so they hit stale entries in AMD's per-executable pipeline cache;
- agc_driver_store_idle_results failed once and passed when rerun.
Results from others, on this PR's earlier heads:
- @johandavidcm ran Yakuza 0 Director's Cut (PPSA31327 01.015.000) on an RTX 4070 with this branch on top of fix(agc): compare the depth write base only when the depth plane is written #6725 and fix(agc): accept PA_CL_CLIP_CNTL CLIP_DISABLE as depth clamp #6572. The stop is gone, and the opening sequence plays through to gameplay. Without the PR, the title stops at the first cutscene with the message above.
- @wachi35676: the three tests pass on an RTX 5070 Ti and on an AMD RDNA2 iGPU (1002:13c0, 32.0.21042.62). With ResourceMaterializer.cpp and ResourceTracker.cpp set back to main, agc_shader_memory and the volume test fail on both. Nothing else in the 716-test agc_|shader|recompiler set changes status.
- @screamm (RTX 4070) and @klosixgaming-arch (RTX 5090): the three tests pass.
Full ctest --parallel 4 --timeout 120 on main 1e2d9b369 with this change: 1463 tests. The only failures are three AMD GPU tests that fail without it on this machine too: agc_driver_lds_mixed_width_atomics (fix(shader): release the LDS lock with an exchange and drop fences inside it #6976), agc_driver_lds_lock_budget (AGC: AMD device-backed LDS lock-budget case returns half of expected sum #7344) and agc_driver_layered_volume_vertex_no_gpl (AMD Windows: GPL-mode cache failures (alpha/wave32 pass after local driver update) #4768).
python tools/check_conventions.py --base origin/main with this title and body passes.
Not tested:
- the PS5's gfx1013;
- an implicit-LOD sample in a pixel shader through the emulator (measured on hardware only);
- other GPUs and drivers.

Hardware measurement (RX 6700 XT, gfx1031, driver 32.0.21045.5002): DIM 2D_ARRAY vs DIM 3D, lane by lane

Each case runs the same instruction twice in separate processes, once with DIM 2D_ARRAY and once with DIM 3D, over 256 lanes in 2x2 quads, and compares all four returned components of every lane. The T# is a 16x16x4 R32_FLOAT volume (SW_MODE 9) over memory where every dword holds its own index, so a value names the dword (or blend) the hardware read. The sampler wraps in x, y and z, with MAX_LOD 0xfff.

Instruction	Levels	Filter	r / quad steps	Cases	Lanes differing
`image_sample_lz`	1	point	r = -0.125, 0 ... 3 (12 values)	12	0 of 3072
`image_sample_lz`	1	bilinear XY, `Z_FILTER` 0	r = 0 ... 2 (6 values)	6	0 of 1536
`image_sample_lz`	1	bilinear XY and Z	r = 0 ... 2 (6 values)	6	0 of 1536
`image_sample`	3	point; trilinear	u/v steps of 1, 2, 4 texels, r steps of 0 to 4 slices	14	0 of 3584
`image_sample_lz`	3	point; trilinear	same	14	0 of 3584
`image_sample_l` (LOD 1)	3	point; trilinear	same	14	0 of 3584

With point filtering and one level, r = 0, 0.125 read slice 0, 0.375 slice 1, 0.5 and 0.625 slice 2, 0.875 slice 3, 1.0, 1.125, 2.0 and 3.0 slice 0, 1.5 slice 2 and -0.125 slice 3, for both DIM values. image_load with DIM 2D_ARRAY takes the integer slice instead, as #5555 implements. With point filtering and three levels, lane 0 of a plain image_sample whose r steps by 2 slices across its quad reads dword 19456, as when u and v step by 2 texels, instead of the 23552 that a 1-texel step with constant r reads, with either DIM value. So r's derivative takes part in the LOD.

Kernel core (the T# is built in s[24:31] from the input buffer address, the sampler in s[32:35]):

  s_mov_b32 s27, 0xa0920fac    ; dst_sel xyzw, SW_MODE 9, LAST_LEVEL 2, TYPE 3D
  s_mov_b32 s28, 3             ; DEPTH 4
  s_mov_b32 s29, 0x20          ; MAX_MIP 2
  s_mov_b32 s33, 0xfff000      ; MAX_LOD
  image_sample v[10:13], v[4:6], s[24:31], s[32:35] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY   ; or dim:SQ_RSRC_IMG_3D

Checklist

🤖 Generated with Claude Code

github-actions · 2026-10-18T03:59:41Z

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 6412	`core/shader/recompiler/ShaderDiskCache.cpp` L141-143
PR 6858	`core/shader/recompiler/ShaderDiskCache.hpp` L16
PR 6924	`core/shader/recompiler/ShaderDiskCache.hpp` L16
PR 7689	`core/shader/recompiler/ShaderDiskCache.hpp` L16

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/Optimization/src/ResourceMaterializer.cpp`
PR 5021	`docs/dev/debt/shader.md`
PR 5371	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/IntermediateRepresentation/include/IntermediateRepresentation/IrMetadata/Resources.hpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `docs/dev/debt/shader.md`
PR 5426	`docs/dev/debt/shader.md`
PR 5500	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `docs/dev/debt/shader.md`
PR 5609	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.cpp`
PR 5656	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.cpp`
PR 5925	`core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 5941	`docs/dev/debt/shader.md`
PR 6009	`docs/dev/debt/shader.md`
PR 6177	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp`
PR 6235	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `docs/dev/debt/shader.md`
PR 6280	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/IntermediateRepresentation/include/IntermediateRepresentation/IrMetadata/Resources.hpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 6290	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp`
PR 6539	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.cpp`
PR 6604	`docs/dev/debt/shader.md`
PR 6645	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp`
PR 6683	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 6879	`core/shader/recompiler/ShaderDiskCache.cpp`
PR 6946	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.cpp`
PR 6947	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 6970	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.cpp`
PR 6977	`core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `docs/dev/debt/shader.md`
PR 7223	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 7332	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `docs/dev/debt/shader.md`
PR 7392	`docs/dev/debt/shader.md`
PR 7395	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 7484	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `docs/dev/debt/shader.md`
PR 7550	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/IntermediateRepresentation/include/IntermediateRepresentation/IrMetadata/Resources.hpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 7642	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/IntermediateRepresentation/include/IntermediateRepresentation/IrMetadata/Resources.hpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`

screamm · 2026-10-18T14:57:54Z

Second-GPU run on head f405b7cae, Windows 11 Pro 26220, NVIDIA GeForce RTX 4070 (596.21), WinLibs MinGW-w64 GCC 15.2, Release. Stacked on main a62c724c3 together with #7413 (d622a55dc); merges cleanly, no shared file. Targeted build (relinker, libs and the shader and agc_driver_* image/sample/shader/table/volume test executables), not a full build.

agc_driver_image_array_sample_on_volume, agc_shader_memory and agc_shader_disk_cache pass 3 of 3. The execution test ran on the RTX 4070 (no skip): DIM 3D and DIM 2D_ARRAY image_sample_lz, DIM 2D_ARRAY image_sample_l, and LOD 1 on the two-level volume all return the expected slice, so this GPU's sampler wraps r like the RX 6700 XT measurement, and image_sample_b is still refused.
Negative control: ResourceMaterializer.cpp and ResourceTracker.cpp set back to main, tests kept. The execution test fails with the title's message (image descriptor is incompatible with the static runtime image interface: descriptor format 56 decodes to {class 1, dim 4, ...}) and agc_shader_memory fails. Restored, both pass.
ctest -j1 --timeout 300 -R '^(agc_driver_(shader|table|descriptor|sample|image|volume|3d)|shader_|agc_shader_|scalar_call)': 81 of 81 pass, including the neighbouring volume and image-sample tests.
check_conventions.py --base origin/main passes and git diff --check is clean.
No regression elsewhere: PPSA99005 and PPSA01153, 60 s each, main (3e3e8b905) against the stack: no stop line, no driver failure, same dump content. They do not sample a 3D T# through 2D_ARRAY, so that says nothing about the change. PPSA31327 is not available to me.
Review note: the extra Dim3D mode is now added to every direct 2D-array sampled image that only uses image_sample/_l/_lz, so the runtime image interface of those shaders gets one more variant even when no volume is bound. I did not measure a cost.

Not checked: PPSA31327, the full ^agc_driver_ pass and the non-GPU suite, SPIR-V validation with spirv-val, AMD/Intel, Linux.

Written with Claude Code assistance (validation by a subagent, reviewed in this session).

wachi35676 · 2026-10-19T00:13:09Z

Tested at f405b7cae on Windows 11 (GCC 15.2.0, Release), merged onto main 9f25d0bf5, on an RTX 5070 Ti (617.42) and an AMD Radeon iGPU, RDNA2 (1002:13c0, 32.0.21042.62, ANYPS5_GPU=AMD).

agc_driver_image_array_sample_on_volume, agc_shader_memory and agc_shader_disk_cache pass on both, and the volume test actually runs rather than skipping.
In the 716-test agc_|shader|recompiler set, no test changes status against main on either GPU.
With only ResourceMaterializer.cpp and ResourceTracker.cpp set back to main, agc_shader_memory and the volume test fail on both GPUs with descriptor format 56 decodes to {class 1, dim 4, ...}.

Small wording point: the debt entry says the Z mode "wraps or clamps", but the measurement used wrap only.

AI-assisted: yes (Claude Code); results above were run on this machine.

klosixgaming-arch · 2026-10-19T00:51:05Z

Tested on Windows 11, MinGW-w64 GCC 15.2.0, NVIDIA GeForce RTX 5090 (driver 617.42), Ryzen 9 9950X3D. Head 15882eaea merged onto current main (a373e85). The merge is textually clean, but it does not build:

core/shader/recompiler/ShaderDiskCache.cpp:143:43: error: static assertion failed: ImageResource changed: update the info encoder

Cause: this PR adds volumeArraySampleCompatible to ImageResource and changes the assertion on line 143 from fieldCount<ImageResource>() == 38 to == 39. main has since changed the same line to 39 for a different field (8545c23), so git merges the two identical edits without a conflict and the result asserts 39 while the struct now has 40 fields. It needs a rebase and the count bumped to 40; it won't show a conflict or the "needs rebase" label.

With only that assertion changed to == 40 on top of the merge, it builds and agc_shader_memory, agc_shader_bindless and agc_driver_image_array_sample_on_volume pass here. I did not run Yakuza 0 (no dump). I did not review the new Dim3D sampling code line by line, so I am not approving yet.

AI-assisted: yes. Written with AI assistance and reviewed by me.

johandavidcm · 2026-10-19T02:35:24Z

PPSA31327 (Yakuza 0 Director's Cut 01.015.000) with this branch, as asked. Windows 11 Pro 26200, WinLibs GCC 15.2.0 r7, Release, RTX 4070 (617.42). main 40393b0 with #6725 (c8f7fc0), #6572 (fd301cc) and 15882ea, plus the bring-up discussion's local PlayGo file; #7199 is already in main.

Without this PR (main 66702f1 with fix(agc): compare the depth write base only when the depth plane is written #6725, fix(agc): accept PA_CL_CLIP_CNTL CLIP_DISABLE as depth clamp #6572 and feat(libSceAgcDriver): support CMASK fast clears on base mip 0 of mipmapped 2D targets #7199): the title stops a few seconds into the first cutscene with image descriptor is incompatible with the static runtime image interface: descriptor format 56 decodes to {class 1, dim 4, ...}.
With it: the whole opening sequence plays, past that point and through to gameplay with the character under control. Textures look right; gameplay runs at about 2 FPS. The title wrote autosave0000.sav. I closed it after about 5 minutes; the log has no stop line or driver failure, and the process exits with 0.

To build on that main I had to set fieldCount<ImageResource>() == 40, as @klosixgaming-arch found. FormatVersion has the same problem: main already went from 38 to 39 for another field, and this PR's 38 -> 39 merges silently with it, so after the rebase it stays at 39 although ImageResource gained a field again. It needs 40 too, or caches written by main's format 39 are read with the new field order.

AI-assisted: yes

thuanquang · 2026-10-19T07:22:10Z

Thanks @klosixgaming-arch and @johandavidcm for catching the silent count merge, and @johandavidcm for the Yakuza 0 run. Rebased onto main 7f596df48 (1b32568):

fieldCount<ImageResource>() is now 40: main's 39 (feat(shader): sample converted unorm images in the shader #5958's field) plus this PR's.
FormatVersion is now 41, because main already uses 40.
The debt entry no longer says "wraps or clamps" (@wachi35676): only wrap was measured.

On the RX 6700 XT the volume test, agc_shader_memory and agc_shader_disk_cache pass 2 of 2 on 1b32568. The full suite on main 7f596df48 with it shows no failure from this PR (details in the description).

The body now lists your runs.

blawesom · 2026-10-19T08:44:17Z

Rebase handoff — rebased onto current main cleanly (authorship unchanged).

Verified on Linux (CachyOS, kernel 7.2.9, GCC 16.2.1, Ninja, Release, BUILD_TESTING=ON, AMD RX 9070 XT / RADV): builds; conventions pass; full suite green apart from the four known main regressions (#5444 Mesa, #7584 BDA, #7766, #7784), none from this PR.

To pick it up:

git fetch https://github.com/blawesom/AnyPS5 rebase/7381
git checkout <your branch>
git merge --ff-only FETCH_HEAD

Head is 31890290d.

blawesom · 2026-10-19T08:44:31Z

Correction: the pushed rebase/7381 was mid-rebase (a later PR's conflicted state). The correct clean branch is being re-pushed now — your PR rebased cleanly with no conflicts; head 1b3256848 content, builds and full suite pending the re-verification. Apologies for the noise.

blawesom · 2026-10-19T08:47:35Z

The clean branch is re-pushed: head f3ab91fcc, builds, conventions pass, full suite green apart from the four known main regressions (agc_driver_tessellation — Mesa #5444; agc_driver_bda_device — #7584; agc_driver_image_table_vouched_lookups — #7766; agc_driver_guest_memory_lifetime — #7784), none from this PR. To pick up: git fetch https://github.com/blawesom/AnyPS5 rebase/7381 then git merge --ff-only FETCH_HEAD on your branch.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

adriadam10 · 2026-10-19T19:40:24Z

Checked the tracker and materializer changes (loads, gathers, bias, comparison and mixed load+sample all clear the flag; a real 2D-array T# still picks the 2D-array mode), the execution test (it fails on main with the title's message and pins the slice and LOD markers), the cache version/field count bumps and the debt entry. main moved under the PR: the cubeFaceStore line in ResourceTracker.cpp and the extended cube entry in docs/dev/debt/shader.md conflicted with it, so I rebased keeping both. The Build run on the rebased head is the gate; merged.

Rewrite the debt entry added by #7381 so it opens with what is unverified (RDNA2 gfx1031 only, wrap addressing only), keeps the Z_FILTER difference, and lists everything that still throws: the other sample forms, gathers, image_get_lod and image_get_resinfo, and the three samples of a T# selected from an image table, of a T# also loaded with DIM 2D_ARRAY, and of an 11_11_10_UNORM, 11_11_10_UINT or depth-bits T#. The entry goes from 1286 to 898 characters. Co-Authored-By: Claude <noreply@anthropic.com>

github-actions Bot added shader core/shader docs agc libSceAgc* labels Oct 18, 2026

EagleBearerYS mentioned this pull request Oct 18, 2026

fix(shader): keep a shader's selected image-table records within the image and pair limits #7527

Merged

12 tasks done

screamm mentioned this pull request Oct 18, 2026

fix(shader): link a table call whose descriptor load sits before its loop #7413

Merged

12 tasks done

github-actions Bot added the needs rebase Conflicts with main, rebase needed label Oct 18, 2026

EagleBearerYS mentioned this pull request Oct 18, 2026

feat(agc): look up table images at run time through a device image table on the GPU heap (APS5_GPU_PATH) #7642

Draft

11 tasks done

github-actions Bot added the unblocks title Names a title ID on the Unblocks line label Oct 18, 2026

RINNECODER mentioned this pull request Oct 18, 2026

sync(upstream): integrate Batch 38 ADX lowering, macOS exports, clear-key hand-over, and BDA fault masking RINNECODER/AnyPS5#562

Merged

thuanquang force-pushed the fix/volume-array-sample branch from f405b7c to 15882ea Compare October 18, 2026 21:45

github-actions Bot removed the needs rebase Conflicts with main, rebase needed label Oct 18, 2026

github-actions Bot added the needs rebase Conflicts with main, rebase needed label Oct 19, 2026

thuanquang force-pushed the fix/volume-array-sample branch from 15882ea to 1b32568 Compare October 19, 2026 07:02

github-actions Bot removed the needs rebase Conflicts with main, rebase needed label Oct 19, 2026

thuanquang mentioned this pull request Oct 19, 2026

fix(shader): answer image_get_resinfo on any T# dimension #7550

Merged

12 tasks done

github-actions Bot added the needs rebase Conflicts with main, rebase needed label Oct 19, 2026

adriadam10 self-assigned this Oct 19, 2026

johandavidcm mentioned this pull request Oct 19, 2026

fix(audioout2): keep each port's grain when ports share a buffer #7719

Closed

12 tasks done

fix(shader): sample a 3D T# through DIM 2D_ARRAY as a volume

1ed8eb6

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

adriadam10 force-pushed the fix/volume-array-sample branch from 1b32568 to 1ed8eb6 Compare October 19, 2026 14:30

github-actions Bot removed the needs rebase Conflicts with main, rebase needed label Oct 19, 2026

johandavidcm mentioned this pull request Oct 19, 2026

perf: PPSA31327 gameplay at 3.3 FPS: the draw committer rescans sampled textures' DCC keys on every lookup #8068

Open

3 tasks done

adriadam10 merged commit d10ef4d into boykopovar:main Oct 19, 2026
10 checks passed

johandavidcm mentioned this pull request Oct 19, 2026

feat(agc): sample the draw committer with APS5_SAMPLE_COMMITTER #8075

Merged

12 tasks done

Amoenus mentioned this pull request Oct 19, 2026

feat(agc): back extra guest mip levels with storage images #7085

Open

12 tasks done

thuanquang mentioned this pull request Oct 20, 2026

docs(shader): state the gap first for 2D-array samples through a 3D T# #8276

Merged

12 tasks done

Navigation Menu

Repository navigation

fix(shader): sample a 3D T# through DIM 2D_ARRAY as a volume - #7381

fix(shader): sample a 3D T# through DIM 2D_ARRAY as a volume#7381
adriadam10 merged 1 commit into
boykopovar:mainfrom
thuanquang:fix/volume-array-sample

thuanquang commented Oct 18, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 18, 2026 •

edited

Loading

Uh oh!

screamm commented Oct 18, 2026

Uh oh!

wachi35676 commented Oct 19, 2026

Uh oh!

klosixgaming-arch commented Oct 19, 2026

Uh oh!

johandavidcm commented Oct 19, 2026

Uh oh!

thuanquang commented Oct 19, 2026

Uh oh!

blawesom commented Oct 19, 2026

Uh oh!

blawesom commented Oct 19, 2026

Uh oh!

blawesom commented Oct 19, 2026

Uh oh!

Uh oh!

adriadam10 commented Oct 19, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

Conversation

thuanquang commented Oct 18, 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 Oct 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

screamm commented Oct 18, 2026

Uh oh!

wachi35676 commented Oct 19, 2026

Uh oh!

klosixgaming-arch commented Oct 19, 2026

Uh oh!

johandavidcm commented Oct 19, 2026

Uh oh!

thuanquang commented Oct 19, 2026

Uh oh!

blawesom commented Oct 19, 2026

Uh oh!

blawesom commented Oct 19, 2026

Uh oh!

blawesom commented Oct 19, 2026

Uh oh!

Uh oh!

adriadam10 commented Oct 19, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

thuanquang commented Oct 18, 2026 •

edited

Loading

github-actions Bot commented Oct 18, 2026 •

edited

Loading