fix(shader): take lane-mask arms and a merged high word in a pointer-keyed record pointer by EagleBearerYS · Pull Request #8939 · boykopovar/AnyPS5

EagleBearerYS · 2026-10-21T06:56:08Z

What

Depends on: #7395. #7223, which #7395 was stacked on, has merged. Until #7395 merges, this PR's diff shows its commits too.

#7395 lets a T# or S# be read through a pointer that a scalar-buffer record holds, with the pointer's low word merged by Phis with constants, lone Ballot words and copies of the record word. Two titles' chains still refuse on the pointer's words:

Little Nightmares Enhanced Edition (PPSA10737), run by klosixgaming-arch (comment). Windows 11, RTX 5090, version 01.004.000, on feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's head of 2026-10-19 plus fix(shader): link a table call whose descriptor load sits before its loop #7413; it stops 2 min 46 s in:
```
[dispatch] failed: compute shader 0x2010fa0000, stage 0: ShaderRecompiler::Recompile: GetImageResource dword 0 is not a valid runtime value (value=LoadAddressU32[…] … GetAddressResource@pc=0x3384(arg0=Phi(arg0=Phi(arg0=Phi(arg0=BitwiseAnd32@pc=0x1600(arg0=CompositeExtractU32x4@pc=0x15d4(arg0=Ballot@pc=0x15d4(…)), arg1=imm=0x0), arg1=BitwiseNot32@pc=0x15fc(…)) …
```
One arm of the low word is the waterfall loop's Ballot mask ANDed with the NOT of the loop's lane bit, the s_andn2 that clears the current lane.
Ratchet & Clank: Rift Apart (PPSA01474), run by ps54all (comment), on main 82f0c9a plus open PRs and feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's earlier commits:
```
compute shader 0x116bb67000: … GetImageResource dword 0 is not a valid runtime value; chain: LoadAddressU32@pc=0x00009a30(GetAddressResource(Phi(… BitwiseAnd32(…, CompositeExtractU32x4(Ballot(Immediate), Immediate))) …
```
A Ballot word is ANDed into the low word, and the high word is merged through a Phi. One arm of that Phi reads a second buffer, and on that edge the low word is a lane mask: the waterfall loop reusing the SGPR pair, as oneandonlydean read it from the IR (comment).

The change, in ResourceTracker.cpp, for pointer-keyed columns only:

fix(shader): take lane-mask ANDs and a high-word Phi in a pointer-keyed record pointer.
- Lane-mask ANDs. An AND with a lane-mask operand is taken. A lane-mask operand is a Ballot word, or an AND or Phi made only of those and immediates. The other operand is walked on, so a record read in it must still match the column. An AND with no lane-mask operand still refuses.
- A merged high word. The high word now goes through the same walk as the low word. Its record reads must be reads of the same column at the low word's offset + 4. When it has no single dominating read, the table comes from the first record read whose V# words dominate the descriptor load. If none does, the column is refused.
- A high-word arm read from a second buffer (a V# loaded through a pointer, then read at offset 0) stays refused, since it isn't a read of the record. Ratchet & Clank's chain has such an arm too (Tamalero's data on #6191). The fourth commit takes it where its edge's low word is a lane mask.
fix(shader): skip lane-mask arms built from a waterfall's lane bit in a pointer-keyed record pointer.
- A lane-mask operand also takes the loop's lane bit, ShiftLeftLogical32(1, index), and ORs and NOTs of lane-mask operands.
  - The index can come from FindILsb32, FindUMsb32 or ReadFirstLane, through BitwiseAnd32, IAdd32, ISub32, immediates and the arms of SelectU32.
  - That covers s_ff1_i32_b32, s_ff1_i32_b64, s_bitset0_b32 and the shift's own AND 31.
- An arm made only of lane-mask operands is skipped. Little Nightmares' arm is one.
docs(shader): the image-table debt entry lists the new arms and the merged high word, and says that a high word read from a second buffer stays refused.
fix(shader): pair a pointer-keyed record pointer's two words by Phi edge, from oneandonlydean's reading of Ratchet & Clank: Rift Apart's IR (comment).
- The low and high words are walked together. Where both are Phis of one block, each low-word edge takes the high word that arrives from the same predecessor.
- An edge whose low word is an immediate or a lane mask is skipped with its high word. Rift Apart's second-buffer high word reaches the load only on such an edge, as oneandonlydean read it, so the walk no longer refuses that edge.
- A record low word still needs every high word it can meet to be an immediate, a lane mask or the same column at offset + 4. A second-buffer high word paired with the record's low word still refuses.
- One shape that planned before now refuses: both Phis in one block, with the record's low word on one edge (its high word an immediate) and the record's high word on another (its low word a lane mask). No edge carries a whole record pointer, so its old plan could only fault at run time.
docs(shader): the debt entry says the two words are walked edge by edge, and that a second-buffer high word stays refused on an edge whose low word is the record word.
fix(shader): refuse a pointer-keyed record arm whose two words are both immediates (9e50a74). Where the pointer's two words are Phis of one block, an edge on which both are immediates is a constant pointer that no record holds, such as a null pointer, so the walk refuses the column there instead of planning it.

At run time, a pointer built from a lane mask faults as outside the snapshot, as #7395's lane-mask arms do.

Disk cache: no plan field changes, but so that a cache written by #7395's recompiler isn't read as this one's, this PR has its own format, 66 (see Tested). main is at 56 since #7223. Ours take one number each: #10489 57, #7642 62, #7395 63, #8939 66, #9457 67, #9633 70, #9790 71, #9749 72 and #9969 73. The other open PRs seen changing it hold 51 (#14944), 53 (#12944), 56 (#9508, the same as main), 65 (#14032), 68 (#14636), 69 (#14871) and 74 (#14853). Whichever lands second on a number takes the next free one.

A shader that planned before plans the same column, apart from the fourth commit's one newly refused shape and the sixth commit's constant-pointer edge.
Shapes that refused before now plan. A refusal is never cached.

Marvel's Spider-Man 2 (PPSA08338), run by spiff637 (comment), stops on a record → pointer → V# → T# chain keyed by ReadFirstLane(SelectU32(LogicalAnd, IAdd32, Phi)). The second commit's agc_driver_pointer_vsharp_select_table builds a record → pointer → V# → T# chain with the T# read under a second key, and it passes without this PR's tracker changes. The title doesn't take it yet: hugodearaujo's run of the US release, Marvel's Spider-Man 2 (PPSA03016), still stops on the #6191 chain with this PR and #7679 (comment). There the T# is read at a fixed offset of the pointer-loaded V#, with no second key, and the same shader then reads its next block through a linked pointer. The test stays as coverage of the keyed form.

Tested

New tests in tests/execution/ImageTable.cpp, on lavapipe. They are synthetic wave64 waterfall loops, assembled with llvm-mc -mcpu=gfx1030 -mattr=+wavefrontsize64. The fourth commit's is the second-buffer shader with one instruction added (s_mov_b32 s44, exec_lo, 0xbeac037e) and its three crossing branch offsets adjusted, checked by disassembly.
- The first commit's: agc_driver_pointer_phi_lane_and_table (lane-mask ANDs), agc_driver_pointer_phi_high_table (a merged high word) and agc_driver_pointer_phi_high_second_table (the second-buffer arm, which refused; 4750c85 later removed this test, see below).
- The second commit's: agc_driver_pointer_phi_lane_not_table (Little Nightmares' BitwiseNot32 and s_andn2 arms) and agc_driver_pointer_vsharp_select_table (Spider-Man 2's chain).
- The fourth commit's: agc_driver_pointer_phi_high_second_lane_table (Rift Apart's edge: the second-buffer high word with a lane-mask low word, which must plan).
- What they check: each lane samples the T# its key's record selects. Each arm whose pointer wasn't captured faults as outside the snapshot, publishes no output and keeps the artifact id.
Red, with only ResourceTracker.cpp reverted and the tests kept, on Linux with CI's configuration:
- On the first commit's base, lane_and and high fail with the dword-0 refusal, and high_second passes (refused). Each half of the commit fixes only its own test.
- On the second commit's base, lane_not fails: GetImageResource dword 0 is not a valid runtime value; chain: LoadAddressU32@pc=0x000000c0(GetAddressResource(Phi(Immediate, Phi(ReadConstBuffer@pc=0x00000074(…), Phi(BitwiseAnd32(CompositeExtractU32x4(Ballot(LogicalAnd), Immediate), BitwiseNot32(ShiftLeftLogical32(Immediate, BitwiseAnd32))), …. vsharp_select passes (coverage).
- On the fourth commit's base (e9d4626), high_second_lane fails: GetImageResource dword 0 is not a valid runtime value; chain: LoadAddressU32@pc=0x00000108(GetAddressResource(Phi(Phi(ReadConstBuffer@pc=0x0000007c(…. The other 10 image-table cases pass, high_second among them (refused).
Green at the head, Linux, lavapipe (Mesa 25.2.8, llvmpipe LLVM 20.1.2), CI's configuration, on a cloud worker:
- 17 tests pass 10 runs each: the five new tests, feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's pointer-Phi ballot and copies tests, the image-table family and agc_shader_disk_cache.
- A mirror of CI's Linux job: 1,718 of 1,721 pass, against 1,716 of 1,719 on the base. The 3 failures on both are that machine's IPv6 socket tests.
- The Khronos validation layer reports no new VUIDs on the focused tests.
GitHub's runners (run, at 2788b8f, the head before the docs commit): the five new tests, feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's pointer-Phi and nested sampler tests and agc_shader_disk_cache pass 5 runs in a row on Linux. On Windows the GPU ones skip and the rest pass. The libc and libkernel suites pass on both.
At the head, a7d1e62 (the fourth and fifth commits):
- A cloud worker, Linux with CI's configuration: the 11 image-table cases pass 20 runs each. A mirror of CI's Linux job: 1,807 of 1,813 pass. The six failures are network tests that depend on that machine (no IPv6, and its socket buffers); the five beyond its known IPv6 failure fail the same way on e9d4626.
- GitHub's runners (run): on Linux the 12 image-table cases and agc_shader_disk_cache_large_file pass 5 runs each. Windows: builds; the Vulkan cases skip on the runner.
- Linux ^agc (run): 813 of 874 pass, 60 skip and 1 fails. The one failure is agc_driver_sgpr_pattern_fill_keys, a timing test, which passed in the worker's full run of the same head.
Ratchet & Clank: Rift Apart (PPSA01474), retail, by @Tamalero (comment): version 01.005.003, Linux, RTX 4090 (NVIDIA 615.78.08), cold shader cache, one unattended run per build.
- feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's head 6c1371b stops at 275 s on compute 0x116bb67500: GetImageResource dword 0, LoadAddressU32@pc=0x9a30.
- This PR at a7d1e62 gets past it, with no regression up to the next stop, as they report. At 270 s it stops on a different shader, compute 0x116bb5ad00: GetImageResource dword 0, LoadAddressU32@pc=0x10b84.
- As they read the new chain, it has the same shape one level deeper. The low word is a Phi of the record read at pc=0xfd94 and a Ballot lane-mask word. The high word's other arm is ReadConstBuffer@pc=0x10350, through a V# whose four dwords are loaded at pc=0x10338 from another pointer.
9162d9d, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 4661f4b, which merges main 0db47b2. The only conflict was ShaderDiskCache.hpp's FormatVersion line: main moved to 45 (fix(shader): normalize offset perspective barycentrics after interpolation #7831), and this branch keeps its 46. On GitHub's runners (run), the pointer, nested, image-table and disk-cache sets (22 tests) pass 10 runs in a row on Linux. On Windows the set has 24, with two Windows-only disk-cache tests: they build, 5 pass 10 runs in a row, and the other 19, which need a Vulkan device, skip. The libc and libkernel suites pass on both, run with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners. An earlier run linked here didn't run this set, so it's replaced.
Little Nightmares Enhanced Edition (PPSA10737) has since run on this branch, in @zaryabzip's run below.
Ghostrunner (PPSA03685), a replay by @frankea (comment): compute shader 0x200c520000, agc_shader_replay --prepare with ANYPS5_NO_SHADER_CACHE=1 on the request the driver dumped, RTX 4090, Windows 11; replay only, no title run. On main 513ca7e with fix(shader): link table callees that together exceed s_call_b64 range #9068, fix(shader): link a table call whose V# is a slice of an s_load_dwordx8 or s_load_dwordx16 #8070 and fix(shader): find strongly connected components without recursion #9701 to refactor(shader): walk each construct when checking the structured constructs #9703, it stops on Shader recompiler: non-invariant descriptor Phi causes GetImageResource validation abort #7459's GetImageResource dword 0 is not a valid runtime value: a T# behind a readfirstlane-keyed record in a waterfall loop whose pointer Phis carry Ballot AND-NOT words. With this PR (9162d9d) it gets past that and stops on sampler resource limit exceeded, which feat(shader): raise the sampler and sampled-pair limits past the 32-bit sampler masks #10000 raises.
Ghostrunner (PPSA03685), a second replay by @TheSovietWaffle (comment): the same compute shader 0x200c520000, agc_shader_replay --prepare on the title's dumped request with ANYPS5_NO_SHADER_CACHE=1, Windows 10 22H2, WinLibs GCC, Release; replay only, no title or GPU run. The base is main 0db47b2 with fix(shader): link table callees that together exceed s_call_b64 range #9068, fix(shader): link a table call whose V# is a slice of an s_load_dwordx8 or s_load_dwordx16 #8070 and refactor(shader): walk each construct when checking the structured constructs #9703.
- The base stops on GetImageResource dword 0 is not a valid runtime value at LoadAddressU32@pc=0x000517dc, and so does the base with fix(shader): plan an image table for a T# read at a first-lane offset #10228.
- With this PR (9162d9d), the 0x517dc load is accepted, and analysis stops on sampler resource limit exceeded.
- With this PR and feat(shader): raise the sampler and sampled-pair limits past the 32-bit sampler masks #10000, it gets past the sampler limit and stops on another image load, LoadAddressU32@pc=0x0002ef78. Its pointer Phi has, besides the record reads with lane-mask arms, an arm read at a fixed offset of a buffer whose V# is loaded from memory. fix(shader): follow pointer hops and fixed-offset V# reads behind a record pointer #9457 with feat(shader): raise the sampler and sampled-pair limits past the 32-bit sampler masks #10000 stops at the same 0x2ef78.
0858667, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 319fa1f (which merges main 379b1dc), on top of the new commit 9e50a74 (6. above). The conflict in ResourceTracker.cpp is between this PR's paired walk and fix(shader): resolve table column pointers through Phi nodes #7880's unwrap of the pointer's words: pointer-keyed columns keep the paired walk, the others take fix(shader): resolve table column pointers through Phi nodes #7880's UnwrapTableRead. FormatVersion stays at feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 47.
- The planner behaviour changes in two places. In the merge, an unmerged record pointer's low word is unwrapped as fix(shader): resolve table column pointers through Phi nodes #7880 does, and the column is refused when it isn't a table read; before, the word was taken as it resolved. Without that, the merge fails fix(shader): resolve table column pointers through Phi nodes #7880's agc_shader_memory and agc_shader_bindless on both runners with GetImageResource dword 0 is not a valid runtime value (run). And 9e50a74 refuses a pointer-keyed Phi edge whose two words are both immediates (a constant pointer no record holds, such as a null pointer), where the walk planned it before.
- 9e50a74's red and green (run): with 9162d9d's code, agc_driver_pointer_phi_high_table fails 3 of 3 on Linux with "a record pointer arm whose two words are both immediates was planned". With the commit, ×10: Linux 24 run, 24 pass, 0 skip, 0 fail; Windows 26 run, 7 pass, 19 skip, 0 fail.
- Focused, ×10 (run): agc_shader_bindless, agc_shader_memory and the ^agc_driver_pointer_, ^agc_driver_nested_, ^agc_driver_image_table and ^agc_shader_disk_cache tests. Linux: 24 run, 24 pass, 0 skip, 0 fail. Windows: 26 run, 7 pass, 19 skip (Vulkan device tests; the runner has none), 0 fail.
- The libc and libkernel suites pass on both runners, run with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners. The full ^agc set at fix(shader): read a chased pointer's high word at its own offset #9749's and fix(shader): plan several T#s read through one pointer-loaded V#, keeping the V# real for their other uses #9790's heads, which carry this merge, shows only main's three known failures (see those PRs).
Ghostrunner (PPSA03685) title runs, by @xXPBSXx (comment): Windows 10 Pro, WinLibs GCC 15.2.0, Release, NVIDIA GeForce RTX 4080 SUPER, runs of up to 600 s with no input, all libraries rebuilt for each tree (pass/stop only). Both states are main 5d3ab77 with fix(shader): structurize a 4,750-block graph before Unreal's render thread watchdog fires #9702 merged in, so compute shader 0x200c520000 gets past structurization.
- Without this PR, both runs stop in 0x200c520000 on GetImageResource dword 0 is not a valid runtime value (LoadAddressU32@pc=0x00051884 and pc=0x0005192c).
- With this PR's 0858667 merged in as well (it carries feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's changes and an earlier version of feat(shader): resolve nested image tables through runtime pointers #7223's beneath it, so these runs don't separate the three), neither run has that refusal. Both stop later on storing storage image results pending over memory outside the write watch without the guest bytes saved before them is not implemented, the check fix(agc): save the guest bytes of results that become pending over a surface that is no longer fully write-watched #7048 addresses. sampler resource limit exceeded, the replays' next stop, doesn't appear in these runs.
232a8e4, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 2525821 (which merges main c9b2660). Clean: FormatVersion follows feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 50, and this PR's own diff is unchanged. On GitHub's runners, 10 runs (run) of the pointer, nested, image-table and disk-cache sets (agc_shader_bindless, agc_shader_memory, ^agc_driver_pointer_, ^agc_driver_nested_, ^agc_driver_image_table, ^agc_shader_disk_cache): Linux: 24 run, 24 pass, 0 skip, 0 fail. Windows: 26 run, 7 pass, 19 skip (no Vulkan device), 0 fail. The libc and libkernel suites pass on both, run with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
e2af03b, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 71741a7, which merges feat(shader): resolve nested image tables through runtime pointers #7223's f1cc524 (yabanana's merge of main c8d34aa). Clean: FormatVersion follows feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 50, and this PR's own diff is unchanged: the merge brings in exactly what feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's merge did (249 files). On GitHub's runners, 10 runs (run) of the pointer, nested, image-table and disk-cache sets (agc_shader_bindless, agc_shader_memory, ^agc_driver_pointer_, ^agc_driver_nested_, ^agc_driver_image_table, ^agc_shader_disk_cache): Linux: 24 run, 24 pass, 0 skip, 0 fail. Windows: 26 run, 7 pass, 19 skip (no Vulkan device), 0 fail. The libc and libkernel suites pass on both (143 and 206 tests on Linux, 170 and 222 on Windows), run with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
Ghostrunner (PPSA03685) replays, by @TheSovietWaffle (comment, IR): Windows 10 22H2, WinLibs GCC 15.2.0, Release, agc_shader_replay --prepare with ANYPS5_NO_SHADER_CACHE=1 on a fresh shader_200c520000.prepare.req from a run on main 74acfad.
- main 74acfad with this PR (e2af03b): a title run on an RTX 3080 (617.42) stops after about 152 s on sampler resource limit exceeded in compute shader 0x200c520000, and the replay stops the same way. feat(shader): raise the sampler and sampled-pair limits past the 32-bit sampler masks #10000 raises that limit.
- With feat(shader): raise the sampler and sampled-pair limits past the 32-bit sampler masks #10000 and fix(shader): index a pointer-loaded V#'s T# array by a word read from the same buffer #9969, fix(shader): plan several T#s read through one pointer-loaded V#, keeping the V# real for their other uses #9790 or fix(shader): follow a conditional pointer chase through a V# behind a pointer-keyed record pointer #9633 on top: the request stops at LoadAddressU32@pc=0x0002ef78 (GetImageResource dword 0 is not a valid runtime value). The IR shows the T# pointer is a Phi of the record pointer and a pointer read at offset 0 through a V# loaded from the record at +0x30, with that V#'s words edited before use (& 0xffff on the base's high word, << 4 on the record count, | 0x30000000 on word 3). No PR in this stack plans that arm; it's the title's next stop.
752620e: 29365ff, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 9343a17, the format commit cfb72cb, the test-only commit 4750c85, and 752620e, a merge of main 0cd07dc (run).
- 29365ff takes feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's merge of feat(shader): resolve nested image tables through runtime pointers #7223's 153400b (yabanana's merge of main 0d62324, which moves feat(shader): resolve nested image tables through runtime pointers #7223 to FormatVersion 49) and feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's test-only commit. Clean, and this PR's own diff is unchanged.
- cfb72cb gives this PR its own disk-cache format, 51, one above feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 50, so a cache written by feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's recompiler isn't read as this one's. The open bumps take one number each: feat(shader): resolve nested image tables through runtime pointers #7223 49 (as do perf(agc): GPU descriptor heap on VK_EXT_descriptor_heap (APS5_GPU_HEAP) #9508, feat(agc): blend into 10_11_11 unorm color targets in the pixel shader #9748 and other open PRs), feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395 50, fix(shader): take lane-mask arms and a merged high word in a pointer-keyed record pointer #8939 51, feat(shader): raise the sampler and sampled-pair limits past the 32-bit sampler masks #10000 52, fix(shader): follow pointer hops and fixed-offset V# reads behind a record pointer #9457 53, feat(shader): raise the image limit to 128 and name the counts in its refusals #10489 54, fix(shader): follow a conditional pointer chase through a V# behind a pointer-keyed record pointer #9633 55, fix(shader): read a chased pointer's high word at its own offset #9749 56, fix(shader): plan several T#s read through one pointer-loaded V#, keeping the V# real for their other uses #9790 57, fix(shader): index a pointer-loaded V#'s T# array by a word read from the same buffer #9969 58, fix(agc): span table-reading wave64 programs by default on Intel drivers that share handles across fused threads #10846 59. Whichever lands second on a number takes the next free one.
- 4750c85 drops two checks that expected planning to refuse a shape with not a valid runtime value, two of the assertions @oneandonlydean listed on fix(shader): index a pointer-loaded V#'s T# array by a word read from the same buffer #9969, under the rule that tests don't assert that a shape the code doesn't support yet throws: a high-word arm read from a buffer whose V# is loaded through a pointer, and a record-pointer arm whose two words are both immediates. The first was the whole of agc_driver_pointer_phi_high_second_table's mode, so that test and its registration go; the lane-mask, high-word and second-lane cases stay.
- 752620e merges main 0cd07dc. The one conflict was ResourceTracker.cpp's include list, where mrfelek's fix(shader): plan an image table for a T# read at a first-lane offset #10228 adds <numeric> beside this PR's <set>; both stay. This PR's changed lines are the same.
- The pointer-Phi, V#-select and SGPR-fill tests with the disk-cache, bindless and shader-memory tests ran 10 times with no failure. Linux: 10 run, 10 pass, 0 skip, 0 fail. Windows: 12 run, 6 pass, 6 skip (the Vulkan device tests; the runner has none), 0 fail.
- Suites: libc and libkernel pass on Linux (166 and 264). On Windows libc passes (194), and libkernel has 4 failures of 286. guest_raise_exception failed after "address waking round 47: the handler did not run within 5 s", as main 9fc181d does in the same verify setup (run); that is libkernel (Windows): guest_raise_exception slowed about 3x and sometimes aborts at "address waking" since 540364a24 (#10666) #11813, and this PR doesn't touch libkernel. exec_arguments, guest_timed_join_import_libkernel_sys and guest_timed_join_import_libkernel_web fail at their setup in the verify's Windows build (FileNotFoundError starting the title, or copying the libraries they load), and pass in main's own Windows job at 1829def (job). The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
24d5cbb, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's ca14648 (run): feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's test-only commit, which drops the nested-sampler test's check that an S# with another consumer is refused. Clean, and this PR's changed lines are the same. The pointer-Phi, V#-select, SGPR-fill and nested-sampler tests with the disk-cache, bindless and shader-memory tests ran 10 times with no failure. Linux: 11 run, 11 pass, 0 skip, 0 fail. Windows: 13 run, 6 pass, 7 skip (the Vulkan device tests), 0 fail. Suites: libc and libkernel pass on Linux (166 and 264). On Windows libc passes (194), and libkernel passes 283 of 286: exec_arguments, guest_timed_join_import_libkernel_sys and guest_timed_join_import_libkernel_web fail at their setup in the verify's Windows build (FileNotFoundError starting the title, or copying the libraries they load), and pass in main's own Windows job at 1829def (job). The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
408401b, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's ca5ba62, which carries main 38ccd50 and with it fix(agc): sample cube descriptors with fewer than six slices as six faces #11507's FormatVersion 49 (run). Clean: FormatVersion stays 51 (see What), and this PR's diff over feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395 is the same. The pointer-Phi, V#-select, SGPR-fill and nested-sampler tests with the disk-cache, bindless and shader-memory tests and fix(agc): sample cube descriptors with fewer than six slices as six faces #11507's agc_driver_image_sample_short_cube ran 10 times with no failure. Linux: 12 run, 12 pass, 0 skip, 0 fail. Windows: 14 run, 6 pass, 8 skip (the Vulkan device tests), 0 fail. Suites: libc and libkernel pass on Linux (166 and 265). On Windows libc passes (194), and libkernel passes 284 of 287: exec_arguments, guest_timed_join_import_libkernel_sys and guest_timed_join_import_libkernel_web fail at their setup in the verify's Windows build (FileNotFoundError starting the title, or copying the libraries they load), with main's code too, and pass in main's own Windows job at 1829def (job). The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
3e2ebbb: 238806d, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's e04a705, and 3e2ebbb, the format commit (run).
- 238806d takes feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's test-only commit d908758, which drops agc_driver_nested_image_table's two checks that a valid shader is refused (a V# with another buffer consumer and a T# with another consumer), and feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's merge of feat(shader): resolve nested image tables through runtime pointers #7223's 0dbf547 (yabanana's merge of main f7e84c4, which moves feat(shader): resolve nested image tables through runtime pointers #7223 to FormatVersion 50). Clean.
- 3e2ebbb gives this PR its own disk-cache format, 53 (52 is feat(shader): raise the sampler and sampled-pair limits past the 32-bit sampler masks #10000's), so a cache written by feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's recompiler isn't read as this one's. Apart from that line, this PR's diff over feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395 is unchanged.
- The pointer, nested, image-table and disk-cache tests with agc_shader_bindless and agc_shader_memory ran 10 times with no failure. Linux: 24 run, 24 pass, 0 skip, 0 fail. Windows: 26 run, 8 pass, 18 skip (the Vulkan device tests; the Windows runner has no Vulkan device), 0 fail. The libc and libkernel suites run 191 and 291 tests on Linux and 220 and 314 on Windows, with 0 failed. They ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
ffa34ec, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 4769d5a, which drops "and refusals" from two nested-table test summary lines (run). It's clean and brings only that change. The image-table, nested, pointer, disk-cache, bindless and shader-memory tests ran 10 times with no failure. Linux: 24 run, 24 pass, 0 skip, 0 fail. Windows: 26 run, 8 pass, 18 skip, 0 fail; the 18 skip because the runner has no Vulkan device. The libc and libkernel suites run 191 and 291 tests on Linux and 220 and 314 on Windows, with 0 failed. The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
Little Nightmares Enhanced Edition (PPSA10737), 01.004.000, @zaryabzip's run (comment, retest on #6275): Windows 11, RTX 3060 Laptop GPU (driver 617.42), this PR's 13c1955 and feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 967f0db merged into main 0836dfd. Compute shader 0x2010fa0000 gets past its stop, and agc_shader_replay --prepare on its dumped request prepares it (481,271 SPIR-V words). On main, and on main with feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395 alone, it fails at pc 0x3300 with GetImageResource dword 0 is not a valid runtime value. The title then stops at compute shader 0x200c340000 with the same message, a pointer chase that fix(shader): follow a conditional pointer chase through a V# behind a pointer-keyed record pointer #9633 and fix(shader): read a chased pointer's high word at its own offset #9749, stacked on this PR, plan when the chase isn't a loop (from reading the code, not run).
3d14230, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's merge of main 0836dfd (1f158eb) (run).
- CMakeLists.txt conflicted from build(agc): register each AGC test executable in its own CMake fragment #12862's split. tools/agc_test_split.py resolve moves this PR's registrations into tests/cmake/agc_driver_image_table_tests.cmake: +7 −2 against feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's merge, the 5 new add_test lines and feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's two set_tests_properties lines extended with this PR's names.
- ShaderDiskCache::FormatVersion conflicted too, since feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's merge brings 63. This PR's own format moves from 53 to 66, the lowest number above 63 that no open PR uses, so its recompiler change keeps a disk cache format of its own.
  - That leaves it above fix(shader): follow pointer hops and fixed-offset V# reads behind a record pointer #9457's 55, which sits on top of this PR. fix(shader): follow pointer hops and fixed-offset V# reads behind a record pointer #9457's head is oneandonlydean's merge of main, left as he made it, so the stack's numbers are uneven until one of them merges again.
- Nothing else conflicted, and apart from the move and the number this PR's diff is unchanged.
- The count: with -DBUILD_TESTING=ON, ctest -N gives 3089 tests: main's 3075, feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 9 and this PR's 5, each listed once.
- ×10 of this PR's verify set. Linux: 14 run, 14 pass, 0 skip, 0 fail. Windows: 14 run, 1 pass, 13 skip (no Vulkan device), 0 fail. The libc and libkernel suites run 199 and 308 tests on Linux and 226 and 330 on Windows, with 0 failed. The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
- Validation, VK_LAYER_KHRONOS_validation on GitHub's Linux runner (run), main 0836dfd against this merge, for the same tests. The loader inserted the layer 1 time on main and 13 with this merge. The only VUID in either arm is VUID-VkDeviceCreateInfo-pNext-pNext, so this merge adds none. agc_driver_image_table fails in that mode in both arms, because the mode treats that main message as a failure.
- Upstream CI's Windows job at this head (job) fails one test of 39: feat(shader): resolve nested image tables through runtime pointers #7223's agc_shader_nested_snapshot_retry, which this PR now carries through feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's merge. As on feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395 and fix(agc): span table-reading wave64 programs by default on Intel drivers that share handles across fused threads #10846, it exits with 0xc0000135 before it starts, because its registration doesn't put libc's directory on PATH on Windows. df80d9b, below, fixes it.
NVIDIA RTX 2000 Ada Generation at 3d14230, from TechLevelZero's NVIDIA bot (report): driver 595.99.02, Ubuntu 26.04.1, 2,509 of 3,087 tests selected.
- 2,459 passed, 6 failed and 44 skipped. None of the failures is new: 4 also fail on main, and 2 are unstable on main.
- With VK_LAYER_KHRONOS_validation, there are no new validation IDs against main, over 939 tests on the GPU.
df80d9b, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 5fdf20f, which gives feat(shader): resolve nested image tables through runtime pointers #7223's agc_shader_nested_snapshot_retry libc's directory on PATH on Windows, beside feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's sampler retry, as oneandonlydean asked on fix(agc): span table-reading wave64 programs by default on Intel drivers that share handles across fused threads #10846 (comment). It brings in only that line, so the upstream Windows failure noted above is gone.
- ×10 (run): this PR's set with both retry tests. Linux: 16 run, 16 pass, 0 skip, 0 fail. Windows: 16 run, 3 pass, 13 skip (no Vulkan device), 0 fail. The libc and libkernel suites run 199 and 308 tests on Linux and 226 and 330 on Windows, with 0 failed. The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
- Upstream CI's Windows job at this head (job), merged onto main d262000: 39 tests, 39 passed, 0 skipped, 0 failed, both retry tests included.
NVIDIA RTX 2000 Ada Generation at df80d9b, from TechLevelZero's NVIDIA bot (report): merged onto main e6f5730, driver 595.99.02, Ubuntu 26.04.1, 2,548 of 3,148 tests selected.
- 2,497 passed, 6 failed and 45 skipped. None of the failures is new: 3 also fail on main, and 3 are unstable on main.
- With VK_LAYER_KHRONOS_validation, there are no new validation IDs against main, over 946 tests on the GPU. 14 tests ran with the layer but aren't in main's baseline, so they weren't compared. Both nested snapshot retry tests passed on the GPU.
87984fe, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 3283b23, which merges main 88e804b (run). Clean: FormatVersion stays 66, and this PR's own diff is unchanged; the plan-encoder numbers come from feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's merge (see feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395). The shader-memory, nested, pointer, image-table, disk-cache and bindless tests, with fix(shader): split scalar-address descriptor phis #7694's agc_shader_descriptor_phis, pass 10 runs in a row. Linux: 27 run, 27 pass, 0 skip, 0 fail. Windows: 29 run, 11 pass, 18 skip, 0 fail. Suites: libc and libkernel pass on Linux (210 and 310) and on Windows (237 and 331). The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners. TechLevelZero's NVIDIA bot ran this head merged onto main 88e804b on an RTX 2000 Ada (report): 2,563 passed, 7 failed, none new (3 also fail on main, 4 unstable on main), 45 skipped, with no new validation IDs (985 tests on the GPU).
afb123a, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 6cb37cc, which merges main 4f64f81 (run). Clean: FormatVersion stays 66, and this PR's own diff is unchanged. The ResourceMaterializer.cpp resolution, with feat(agc): sample 1D arrays past maxImageArrayLayers as rows of a 2D image #12246's view loop moved back into Snapshot, comes from feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's merge (see feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395). The shader-memory, nested, pointer, image-table, disk-cache and bindless tests, with fix(shader): split scalar-address descriptor phis #7694's agc_shader_descriptor_phis, pass 10 runs in a row. Linux: 27 run, 27 pass, 0 skip, 0 fail. Windows: 29 run, 11 pass, 18 skip (no Vulkan device), 0 fail. Suites: libc and libkernel pass on Linux (217 and 318) and on Windows (244 and 339). The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
58ebe4b, a merge of main 4f41f4c (×10, complete ×10).
- The conflicts: ShaderDiskCache.hpp's FormatVersion, where main is now at 52 (feat(shader): raise the sampler and sampled-pair limits past the 32-bit sampler masks #10000) and this PR keeps its 66; and docs/dev/debt/shader.md, where this PR's pointer lane-mask entry and main's new sampler-limit entry are adjacent and both stay. This PR's own lines are unchanged apart from the FormatVersion line's old value.
- ×10: agc_driver_image_table, agc_driver_many_direct_samplers and every test matching ^agc_shader_memory, ^agc_shader_disk_cache or ^agc_shader_sampler_limits. Linux: 7 run, 7 pass, 0 skip, 0 fail. Windows: 9 run, 7 pass, 2 skip (no Vulkan device), 0 fail. The libc and libkernel suites run 300 and 322 tests on Linux and 327 and 342 on Windows, with 0 failed. The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.
- Complete ×10, adding every ^agc_driver_pointer_, ^agc_driver_nested_, ^agc_shader_nested and ^agc_shader_table test, so every test in this PR's diff over main runs. Linux: 22 run, 22 pass, 0 skip, 0 fail. Windows: 24 run, 10 pass, 14 skip (no Vulkan device), 0 fail.
- GitHub's windows check is red at this head on one test, avplayer_http_patched_modules, at AvPlayerHttpModules.cmake:155, where the owner executable avplayer_http_patched_modules_tests.exe wasn't built for the selected test (job). fix(ci): build owners of wrappers around selected guest tests #15375 is the open fix. This PR changes no avplayer or HTTP file, and the check passed at afb123a.
9eab037, a merge of feat(shader): read samplers through nested and pointer-keyed descriptor tables #7395's 014d1e3, which carries main c2eafe4 with yabanana's feat(shader): resolve nested image tables through runtime pointers #7223 (×10).
- The conflicts: ShaderDiskCache.hpp's FormatVersion, which stays 66 (main is at 56 since feat(shader): resolve nested image tables through runtime pointers #7223), and docs/dev/debt/shader.md, where this PR's wording of the pointer-keyed image-table entry stays. This PR's own lines are unchanged apart from the FormatVersion line's old value.
- ×10: agc_driver_many_direct_samplers and every test matching ^agc_driver_pointer_, ^agc_driver_nested_, ^agc_shader_nested_, ^agc_driver_image_table, ^agc_shader_memory, ^agc_shader_disk_cache or ^agc_shader_sampler_limits, which takes in every test in this PR's diff over main. Linux: 27 run, 27 pass, 0 skip, 0 fail. Windows: 29 run, 10 pass, 19 skip (no Vulkan device), 0 fail. The libc and libkernel suites run 301 and 325 tests on Linux and 331 and 343 on Windows, with 0 failed. The suites ran with APS5_SKIP_TIMING_TESTS=1, which relaxes precise_wait's median bounds on GitHub's hosted runners.

Checklist

An S# read from a V# that a scalar record selects through a pointer was refused, so recompiling failed with "GetSamplerResource dword 0 is not a valid runtime value". The nested branch now takes sampler columns: they get child tables per distinct V# like nested T#s, share equal children, count against the nested key budget and the S# entry budget, and fault on an unmapped, non-buffer or uncaptured parent. The plan decoder accepts a nested sampler column; the plan encoding is unchanged.

An S# or T# loaded through a pointer that a scalar record holds was refused when a Phi merged the pointer's low word with constants, so recompiling failed with "GetSamplerResource dword 0 is not a valid runtime value". Such a column is now pointer-keyed: the snapshot captures every record's pointer and the descriptor behind it, the map holds boykopovar#7223's sorted pointer dictionary with descriptor records as its codes, and the GPU looks up the actual 64-bit pointer. A pointer that was not captured faults as outside the snapshot, and a null one as unmapped. The plan gains TableColumn::pointerKeyed and the snapshot gains ImageTableColumnSnapshot::pointerKeyed, so the disk-cache format moves to 43 and the image-table ABI to 4.

A T# loaded through a record pointer whose low word passes nested Phis was refused when one Phi arm held a word of a Ballot lane mask, as in a waterfall loop that reuses the pointer's SGPR for an s_and_saveexec or v_cmp mask, so recompiling failed with "GetImageResource dword 0 is not a valid runtime value". A pointer-keyed column now accepts such a word as an arm that is not a record pointer, as it does for constants, and still refuses any other foreign arm. At run time a pointer built from a lane mask is not a captured pointer and faults as outside the snapshot. A pointer-keyed column also accepts low-word reads, and a high-word read, of the same record word under different keys: same table, record offset, stride and addend. The GPU looks the pointer up by its value, so the key does not matter there. The column takes its table and record offset from the high-word read, which dominates the descriptor load. A shader that planned before plans the same way, so the disk-cache format stays at 43.

…ed record pointer A T# loaded through a record pointer was still refused with "GetImageResource dword 0 is not a valid runtime value" when the pointer's low word passed ANDs with lane masks, as a waterfall loop that copies EXEC into the pointer's SGPR builds them, or when its high word was a Phi of the record's high word read under several keys. For a pointer-keyed column, the walk of the low word now takes an AND with a lane-mask operand (a Ballot word, or an AND or Phi of only Ballot words and immediates) and keeps walking its other operand, so its record reads must still match the column and anything else still refuses. The high word is walked the same way: its record reads must be the record word at the low word's offset plus 4. A high word that is one read still gives the column its table and record offset, as before. Otherwise the table comes from a record read whose V# words are defined before the descriptor load, and the record offset from the high word, which the GPU does not read for a pointer-keyed column. A high-word arm read from another buffer, such as one whose V# is loaded through a pointer, stays refused. A shader that planned before plans the same way, so the disk-cache format does not change.

… a pointer-keyed record pointer A T# loaded through a record pointer was still refused with "GetImageResource dword 0 is not a valid runtime value" when a Phi of the pointer's low word merged the record read, a zero, and the waterfall's lane mask with the handled lane's bit cleared, BitwiseAnd32(CompositeExtractU32x4(Ballot(...), 0), BitwiseNot32(ShiftLeftLogical32(1, index))), as s_ff1, s_lshl and s_not with s_and, or s_andn2, build it in the SGPR the pointer reuses. The walk took only Ballot words and ANDs of them as lane masks, so it walked into the NOT and refused. A lane-mask operand is now a Ballot word or the loop's lane bit, a shift of 1 by an index from FindILsb32, FindUMsb32 or ReadFirstLane (through the AND, add, subtract and select that s_ff1_i32_b64, s_bitset0 and the shift's own mask put around it), or an AND, OR, NOT or Phi of only those and immediates. An arm of the pointer that is entirely such an operand is skipped. An AND with a record read in it still has the read matched against the column, and anything else still refuses. agc_driver_pointer_phi_lane_not_table builds that shape in a wave64 waterfall: the low word's Phis merge the record read at +0, a zero and the lane mask without the handled lane's bit, cleared by s_not and s_and and by s_andn2, and the high word is one read at +4. Each lane samples the T# of its record, and every masked word faults outside the snapshot. agc_driver_pointer_vsharp_select_table reads a record keyed by ReadFirstLane(SelectU32(LogicalAnd, IAdd32, Phi)), a pointer from it, a V# through the pointer and the T# through the V#; it passes without this change and keeps that chain checked, including a record pointer moved to the other V# without a recompile. Shaders that planned before plan the same way, and a refusal is never cached, so the disk-cache format does not change.

…keyed records A pointer-keyed record pointer now also takes lane-mask ANDs, the waterfall loop's lane bit and its OR and NOT forms, and a high word merged from the record's next word. A high word read from a second buffer stays refused.

github-actions · 2026-10-21T07:16:30Z

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 6416	`core/shader/recompiler/ShaderDiskCache.hpp` L16
PR 10351	`core/shader/recompiler/ShaderDiskCache.hpp` L16
PR 11551	`core/shader/recompiler/ShaderDiskCache.hpp` L16
PR 14636	`core/shader/recompiler/ShaderDiskCache.hpp` L16
PR 14871	`core/shader/recompiler/ShaderDiskCache.hpp` L16
PR 14904	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` L1875

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` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 5021	`core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/shader.md`
PR 5371	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/shader.md`
PR 5609	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp`
PR 5925	`core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
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/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/shader.md`
PR 6412	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/IntermediateRepresentation/include/IntermediateRepresentation/IrMetadata/ResourcePlan.hpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 6481	`core/libs/prx/libSceAgcDriver/tests/execution/ImageTable.cpp`
PR 6488	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/ImageTableAbi.hpp` `core/shader/recompiler/IntermediateRepresentation/include/IntermediateRepresentation/IrMetadata/ControlFlowInfo.hpp` `core/shader/recompiler/IntermediateRepresentation/include/IntermediateRepresentation/IrMetadata/ResourcePlan.hpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/shader.md`
PR 6858	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 6924	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 6970	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.cpp`
PR 7085	`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` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/shader.md`
PR 7392	`docs/dev/debt/shader.md`
PR 7410	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp` `docs/dev/debt/shader.md`
PR 7684	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/libs/prx/libSceAgcDriver/tests/execution/ImageTable.cpp` `core/shader/recompiler/IntermediateRepresentation/include/IntermediateRepresentation/IrMetadata/ControlFlowInfo.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 7689	`core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/ShaderDiskCache.cpp`
PR 8004	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp` `docs/dev/debt/shader.md`
PR 9020	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 9183	`core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/shader.md`
PR 9335	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 9356	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 9387	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 9748	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 9852	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 9889	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.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` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/shader.md`
PR 9897	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/libs/prx/libSceAgcDriver/tests/cmake/agc_shader_memory_tests.cmake` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 10049	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 10193	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/libs/prx/libSceAgcDriver/tests/cmake/agc_shader_memory_tests.cmake` `core/shader/recompiler/ImageTableAbi.hpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 10233	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `docs/dev/debt/shader.md`
PR 10359	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp`
PR 10757	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp`
PR 10867	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/shader.md`
PR 11351	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp`
PR 11483	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp` `docs/dev/debt/shader.md`
PR 12324	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/IntermediateRepresentation/include/IntermediateRepresentation/IrMetadata/ControlFlowInfo.hpp` `core/shader/recompiler/Recompiler.cpp`
PR 12548	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp` `docs/dev/debt/shader.md`
PR 12790	`core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp` `docs/dev/debt/shader.md`
PR 12828	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/shader/recompiler/Recompiler.cpp`
PR 12944	`core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `docs/dev/debt/shader.md`
PR 13928	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/libs/prx/libSceAgcDriver/tests/cmake/agc_shader_memory_tests.cmake` `core/shader/recompiler/ImageTableAbi.hpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 13929	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/libs/prx/libSceAgcDriver/tests/cmake/agc_shader_memory_tests.cmake` `core/shader/recompiler/ImageTableAbi.hpp` `core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 14032	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/libs/prx/libSceAgcDriver/tests/cmake/agc_shader_memory_tests.cmake` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`
PR 14212	`docs/dev/debt/shader.md`
PR 14217	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/Recompiler.cpp` `docs/dev/debt/shader.md`
PR 14218	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/shader/recompiler/Recompiler.cpp` `docs/dev/debt/shader.md`
PR 14561	`core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp`
PR 14621	`core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 14702	`core/shader/recompiler/Optimization/src/ResourceMaterializer.cpp` `core/shader/recompiler/SpirvBackend/src/SpirvImageEmitter.cpp`
PR 14944	`core/libs/prx/libSceAgcDriver/tests/ShaderDiskCache.cpp` `core/libs/prx/libSceAgcDriver/tests/ShaderMemory.cpp` `core/libs/prx/libSceAgcDriver/tests/cmake/agc_shader_memory_tests.cmake` `core/shader/recompiler/Optimization/src/ResourceTracker.cpp` `core/shader/recompiler/Recompiler.cpp` `core/shader/recompiler/ShaderDiskCache.cpp` `core/shader/recompiler/ShaderDiskCache.hpp`

…ane-masks

oneandonlydean · 2026-10-21T13:40:30Z

@EagleBearerYS thanks for #8939. We ran PPSA01474 on main plus #7395 and #8939 (f877d5e; the pointer walk is the same at e9d4626) and dumped the IR of the shader that stops. It still stops at LoadAddressU32@pc=0x9a30 with GetImageResource dword 0 is not a valid runtime value. Its address changes from run to run (0x116bb5d300, 0x116bb66700, 0x116bb5c500 and 0x116bb5c900 are the same program).

The IR shows which check refuses it, and why that arm isn't a pointer:

The descriptor at 0x9a30 is loaded through GetAddressResource(low, high), and both words are Phis that merge at pc 0x5a48.
Low word: immediates, lane masks, and the record column read at pc 0x1aa8. Your lane-mask handling accepts all of these.
High word: the same record column's high word, plus one arm from the second buffer (ReadConstBuffer at offset 0 of a buffer whose V# is loaded through the record pointer, pc 0x900c/0x908c). PointerRecordReads refuses that arm on the high-word walk, because it isn't the same record column.
Pairing the Phi edges: the second-buffer high word reaches the load on exactly one incoming edge. On that edge the low word is BitwiseAnd32(Phi(lane masks), CompositeExtractU32x4(Ballot(true), 0)), a pure lane mask. Every edge that carries a record-column low word carries the record-column high word.

So on the one path where the high word comes from the second buffer, the "pointer" is (a lane mask, a T#-table dword). That's the SGPR pair being reused by the waterfall loop, not an address. Tamalero's dump on #6191 (issuecomment-6255548496) agrees: the record's high word is 0x00000012 in all 27,396 populated records, and no dword of that T# table equals 0x12.

Could the walk pair the two words by incoming edge? Concretely:

Walk the low- and high-word Phis together, edge by edge.
On an edge where the low word is lane-mask-only (or an immediate), skip both words, as the low-word walk already skips lane-mask arms on its own.
Refuse as now on any edge that pairs a record-column low word with a high word from anywhere other than the same column at offset + 4.
agc_driver_pointer_phi_high_second_table keeps refusing when the second-buffer high word is paired with a record-column low word. A new case pairs it with a lane-mask low word, which should plan, and its record lanes should sample the T# their record selects.

This decides it at plan time from the IR alone, with no run-time fault path and no new plan field. If you'd rather keep the walks separate, a run-time fault for pointers built from that arm (as lane-mask pointers already do) would also work for us. We can share the IR excerpt around 0x9a30, and we're happy to run the title on the result.

The low and high words of a pointer-keyed record pointer were walked separately, so a high-word arm on a path whose low word is a lane mask refused the plan even though that pair is no pointer. Ratchet & Clank: Rift Apart (PPSA01474) stops this way at LoadAddressU32@pc=0x9a30: one high-word arm reads a second buffer, and on that edge the low word is a lane mask, the waterfall loop reusing the SGPR pair. Walk the two words together. Where both are Phis of one block, each low-word edge takes the high word of the same predecessor. An edge whose low word is a constant or a lane mask is skipped with its high word. A record low word still needs every high word it can meet to be a constant, a lane mask or the same record column at offset + 4, so a second-buffer high word paired with the record word stays refused. agc_driver_pointer_phi_high_second_lane_table adds the lane-mask low word to the second-buffer arm: it plans, each group samples the T# its record selects, and that arm faults outside the snapshot.

…nter The two words of the pointer are now walked edge by edge, and an edge whose low word is a constant or a lane mask is skipped with its high word. A high word read from a second buffer stays refused on an edge whose low word is the record word.

EagleBearerYS · 2026-10-21T14:07:27Z

@oneandonlydean, thank you for reading the IR this closely. The edge-by-edge pairing is the better fix, so I've written it as you describe, with both cases: agc_driver_pointer_phi_high_second_table keeps refusing the second-buffer high word paired with the record-column low word, and a new case pairs it with a lane-mask low word and checks that it plans and that each record lane samples the T# its record selects. It's on GitHub's runners now. I'll post here once it's pushed to this branch, so you can run Ratchet & Clank: Rift Apart (PPSA01474) on it.

…walk red/green and Linux CI mirror

EagleBearerYS · 2026-10-21T14:50:05Z

@oneandonlydean, thank you for the IR dump and the steps. a7d1e62 on this branch pairs the two words as you describe:

PointerRecordPairs walks the low- and high-word Phis together. Where both are Phis of one block, each low-word edge takes the high word that arrives from the same predecessor. A word that is no Phi there stays the same on every edge.
An edge whose low word is an immediate or a lane mask is skipped with its high word.
A record-column low word still needs every high word it can meet to be an immediate, a lane mask, or the same record column at offset + 4. Anything else refuses, as before.
agc_driver_pointer_phi_high_second_table still refuses its second-buffer high word, which it pairs with a record-column low word. The new agc_driver_pointer_phi_high_second_lane_table is the same shader with s_mov_b32 s44, exec_lo added to the second-buffer arm. That read is BitwiseAnd32(exec_lo, CompositeExtractU32x4(Ballot(true), 0)) in the IR, the form you list. It plans, each group's lanes sample the T# their record selects, and that arm faults outside the snapshot when a group takes it.

There's no new plan field and no new run-time path.

Tested on GitHub's runners, on main b6fe513 (run 39613056959):

Linux (lavapipe): the 12 image-table cases, including both pointer_phi_high_second cases, plus agc_shader_disk_cache_large_file, 5 runs each, all pass.
Windows: builds; the Vulkan cases skip on the runner.
Linux ^agc (run 39613110357): 873 of 874 pass, 60 skip. The one failure is agc_driver_sgpr_pattern_fill_keys, a timing test, which passed in the full run of the same head below.
Linux with CI's flags on a separate machine:
- the 11 image-table cases, 20 of 20 each;
- red: with only ResourceTracker.cpp from e9d4626, the new case fails with "GetImageResource dword 0 is not a valid runtime value", and agc_driver_pointer_phi_high_second_table still refuses as before;
- the full ctest: 1807 of 1813. The six failures are network tests that depend on that machine (no IPv6, and its socket buffers); the five beyond its known IPv6 failure fail the same way on e9d4626.

One shape that planned before is now refused. Both Phis are in one block, one edge carries the record's low word with an immediate high word, and the other carries a lane mask with the record's high word. No edge carries a whole record pointer, so the old plan could only fault at run time there. If you'd rather keep it planned, falling back to the old separate high-word walk when the paired walk finds no high read keeps it as before.

Could you run Ratchet & Clank: Rift Apart (PPSA01474) on a7d1e62? If it stops again, the IR excerpt around the new stop would help.

Tamalero · 2026-10-21T19:20:35Z

Retail check of a7d1e62 on Ratchet & Clank: Rift Apart (PPSA01474, 01.005.003), Linux, RTX 4090 (NVIDIA 615.78.08), cold shader cache, one unattended run per build:

Build	Result
#7395 head `6c1371b` (contains #7223 `8716815` and `main` `b6fe513`)	275 s, stops on compute 0x116bb67500: `GetImageResource dword 0`, `LoadAddressU32@pc=0x9a30`, the stop you dumped
this PR, `a7d1e62`	270 s, gets past 0x9a30 and stops on a different shader, compute 0x116bb5ad00: `GetImageResource dword 0`, `LoadAddressU32@pc=0x10b84`

The new chain has the same overall shape one level deeper. The low word is a Phi of the record read at pc=0xfd94 and a Ballot lane-mask word (CompositeExtractU32x4). The high word arm is ReadConstBuffer@pc=0x10350 through a V# whose four dwords are loaded at pc=0x10338 from another pointer:

LoadAddressU32@pc=0x00010b84(GetAddressResource(Phi(...ReadConstBuffer@pc=0x0000fd94(GetBufferResource, Immediate)..., CompositeExtractU32x4...),
  Phi(...ReadConstBuffer@pc=0x0000fd94..., ReadConstBuffer@pc=0x00010350(GetBufferResource(LoadAddressU32@pc=0x00010338(GetAddressResource, Immediate, Immediate, Immediate) x4), Immediate))), ...)

So a7d1e62 fixes the 0x9a30 stop on the retail title. No regression up to the new stop. I can dump this shader's IR or the runtime values of the pc=0x10338/0x10350 reads if that helps.

Tested and written by Claude (AI assistant) on a contributor's machine.

EagleBearerYS · 2026-10-21T19:26:30Z

@Tamalero, thank you for the retail run. It's now in the description: this PR gets Ratchet & Clank: Rift Apart (PPSA01474) past the 0x9a30 stop.

We'd like to look at the 0x10b84 stop next. If you have time, the IR dump would help most:

Run with APS5_DUMP_IR=116bb5ad00: the shader's code address, in lowercase hex without 0x.
It prints ==== IR 0x116bb5ad00 after <pass> blocks to stderr. Only the last block before the stop is needed.
The address can change from run to run (dean saw this PR's first stop at several, comment). If the stop names another address, use that one.
That run skips the shader disk cache's plans, which is expected.

If it's easy, the runtime values would help too: on the dispatch that stops, the dword the pc=0x10350 read returns, and the record's high word, the dword after the pc=0xfd94 read.

We'll read the dump before saying what the fix is.

Brings boykopovar#8991 (match image-table T# halves offset before and after readfirstlane). ResourceTracker.cpp: main's FirstLaneOffset, ParseFirstLaneOffset and SameFirstLaneOffset go before TryMakeTableColumn, which keeps this branch's signature with the buffer parameter. Main's first-lane base check in the column loop merged on its own.

…ane-masks Brings boykopovar#7395's merge of main bd4d801, with boykopovar#8991. It merged without conflicts.

TechLevelZero · 2026-11-02T11:47:26Z

✅ NVIDIA GPU test: passed

PR head 87984fe merged onto main 88e804b · tested 2026-11-02 11:47 UTC · 27 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	✅ 2,563 passed · 7 failed (0 new, 3 also fail on main, 4 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	2,615 of 3,234 tests run: the ones this change can reach, chosen by `tools/select_tests.py` from main's build graph
Vulkan validation	✅ no new validation IDs vs main (985 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ `agc_driver_pipeline_journal` 120.1 s, ⚠️ `agc_driver_graphics_maps_fallback` 30.9 s, ⚠️ `agc_driver_graphics` 30.4 s

Unstable on main (4, these also failed in recent main runs on this machine; not counted against this PR)

agc_driver_job_manager_compute_capacity: failed in 1 of the last 10 main runs
agc_driver_pipeline_journal: failed in 2 of the last 10 main runs
agc_submit_progress: failed in 8 of the last 10 main runs
agc_submit_progress_no_timeline: failed in 8 of the last 10 main runs

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

agc_driver_image_table
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (14, not in the main baseline's validation pass)

agc_driver_nested_image_table
agc_driver_nested_image_table_budget
agc_driver_nested_image_table_large
agc_driver_nested_sampler_table
agc_driver_pointer_phi_ballot_table
agc_driver_pointer_phi_copies_table
agc_driver_pointer_phi_high_second_lane_table
agc_driver_pointer_phi_high_table
agc_driver_pointer_phi_lane_and_table
agc_driver_pointer_phi_lane_not_table
agc_driver_pointer_phi_sampler_table
agc_driver_pointer_vsharp_select_table
agc_shader_nested_sampler_snapshot_retry
agc_shader_nested_snapshot_retry

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

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

Skipped (45)

agc_driver_bda_reads_shadowed_results: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_bda_reads_shadowed_results_per_build: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_sample_order_grid_reference: skipped, the device does not keep depth drawn at custom sample locations at those locations, even after a transition that names them: the composite of uniform renders holds the guest's depth in 0 of 3005 sloped samples, so no path can be compared with it bit for bit
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_import_settle_initial_dcc: skipped, the device has no deferred settle of host imports: imports under 1 MiB wait for one only on Windows
agc_unit_shadow_completion_store: skipped, the device reports watched host memory written when it imports or reads it
guest_http_uri_locale: 
guest_net_ctl: skipped, the host has no IPv4 address
guest_video_out_color_space: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_hdr: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_high_frame_rate: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_output_mode: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_resolution_status: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_wqhd_detection: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
video_out_flip_aftergpu: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_backtoback: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_controls: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_cpuflip: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_cpuflip_in_order: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_dcc: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_deferred: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_eop: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_eop_aftergpu: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_eop_main: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_fps_limit: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_o

Reproduce

git fetch origin 87984fe7282bd635d2a565935e05bc0920671a3f 88e804b0a7df1f1ac373a9e2f72fb378d925fdd4
git checkout 88e804b0a7df1f1ac373a9e2f72fb378d925fdd4 && git merge --no-ff 87984fe7282bd635d2a565935e05bc0920671a3f
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.}

Main's 1D array layer rows add a loop to the table snapshot that records each table's registered image views, keyed by the table's index. This branch copies a column's records in a helper that also serves nested child columns, which have no table index, so the loop stays in the snapshot of the top-level table after the copy.

…asks

TechLevelZero · 2026-11-02T18:25:29Z

✅ NVIDIA GPU test: passed

PR head afb123a merged onto main 1d91f29 · tested 2026-11-02 18:25 UTC · 35 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	✅ 2,692 passed · 6 failed (0 new, 3 also fail on main, 3 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	2,743 of 3,544 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 (998 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` 30.8 s, ⚠️ `agc_driver_graphics_maps_fallback` 30.6 s, ⚠️ `agc_driver_native_depth_clears` 30.1 s

Unstable on main (3, these also failed in recent main runs on this machine; not counted against this PR)

agc_driver_pipeline_journal: failed in 3 of the last 10 main runs
agc_submit_progress: failed in 8 of the last 10 main runs
agc_submit_progress_no_timeline: failed in 7 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

Passed on retry (1, failed in the parallel run, passed when re-run with a 360 s timeout; not counted against this PR)

agc_driver_native_depth_clears: Timeout in the parallel run; also passed only on retry in 12 other runs in the last 7 days

Validated but not compared (14, not in the main baseline's validation pass)

agc_driver_nested_image_table
agc_driver_nested_image_table_budget
agc_driver_nested_image_table_large
agc_driver_nested_sampler_table
agc_driver_pointer_phi_ballot_table
agc_driver_pointer_phi_copies_table
agc_driver_pointer_phi_high_second_lane_table
agc_driver_pointer_phi_high_table
agc_driver_pointer_phi_lane_and_table
agc_driver_pointer_phi_lane_not_table
agc_driver_pointer_phi_sampler_table
agc_driver_pointer_vsharp_select_table
agc_shader_nested_sampler_snapshot_retry
agc_shader_nested_snapshot_retry

Validation IDs already present on main (4 IDs in 510 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (510 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 502 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

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 afb123a49ec3f7555c274c769abfa5bbd4f87b81 1d91f2966d0eef82a021628a7aa8e78e615ac807
git checkout 1d91f2966d0eef82a021628a7aa8e78e615ac807 && git merge --no-ff afb123a49ec3f7555c274c769abfa5bbd4f87b81
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.}

# Conflicts:

BunnyFl

Reviewed this PR's own change on top of #7395's head (git diff of the two heads: ResourceTracker.cpp, the new agc_driver_pointer_phi_* cases, the cache format and the debt entry). #7395 itself I haven't re-reviewed here.

Reads right. PointerRecordPairs walks the low and high words together, pairing a Phi's arms by predecessor block (PhiIncoming) when both words are Phis in the same block. It skips an edge only when its low word is a constant or a lane mask, and, for paired Phis, refuses an edge where both words are immediates. A record word ANDed with a lane mask is unwrapped only when exactly one AND operand is a lane mask (MaskedRecordWord), and every accepted read must keep the same offset and record column (PointerRecordRead). The skipped edges can't bind a wrong resource silently: per the debt entry (docs/dev/debt/shader.md), the snapshot holds the captured records' pointers and the GPU looks up the actual 64-bit pointer, so a lane-mask value that reaches the lookup faults instead.

Tested on Windows 11, WinLibs GCC 15.2.0, with this head merged onto current main: 21 tests (agc_driver_pointer_*, agc_driver_image_table*, agc_shader_memory*, agc_shader_disk_cache*) pass, nothing skipped. Negative control: with #7395's ResourceTracker.cpp in the same merged tree, the four new cases (..._lane_and_table, ..._high_table, ..._high_second_lane_table, ..._lane_not_table) fail and #7395's three pass. agc_driver_pointer_phi_lane_and_table fails with GetImageResource dword 0 is not a valid runtime value; chain: ... BitwiseAnd32(Phi(CompositeExtractU32x4(Ballot(... (the shape the titles stop on).

XairRrR

Reviewed — approved

The edge-paired walk of a pointer-keyed record pointer's two words is sound. It merges and builds on current main, the focused tests pass with no new VUIDs, and four of the five new tests fail without this PR's ResourceTracker.cpp.

What I ran

Head afb123a49ec3f7555c274c769abfa5bbd4f87b81, main 9090972b7e91fd266819cbd77a8b058e177b5682, merged tree 811a3c5c010226d737ab32ee801a80c9d77bb65d (Release, GCC 14.2.0, Ninja, lavapipe Mesa 26.1.6 / LLVM 19, Debian 13 VM).

git diff --check 9090972b7e91fd266819cbd77a8b058e177b5682...afb123a49ec3f7555c274c769abfa5bbd4f87b81
python3 tools/check_conventions.py --base 9090972b7e91fd266819cbd77a8b058e177b5682
ctest --test-dir build -j8 --test-load 32 --output-on-failure -R "^(agc_driver_pointer_|agc_driver_nested_|agc_driver_image_table|agc_shader_disk_cache|agc_shader_nested|agc_shader_memory$|agc_shader_bindless$)"
# negative control
git show 6cb37cc9eda523074d5fd844537c28ce2bbe6907:core/shader/recompiler/Optimization/src/ResourceTracker.cpp > core/shader/recompiler/Optimization/src/ResourceTracker.cpp
ninja -C build -j32 -l32
ctest --test-dir build -j8 --test-load 32 --output-on-failure -R "^agc_driver_pointer_"
git show 811a3c5c010226d737ab32ee801a80c9d77bb65d:core/shader/recompiler/Optimization/src/ResourceTracker.cpp > core/shader/recompiler/Optimization/src/ResourceTracker.cpp
ninja -C build -j32 -l32
ctest --test-dir build -j8 --test-load 32 -R "^agc_driver_pointer_"
git status --porcelain

Results

Check	Result
`git diff --check`	clean
`check_conventions.py`	All conventions checks passed
Configure / build / libs (script)	ok
Selected tests (script)	2113 ran, 0 new failures, 0 timeouts. 7 known failures also fail on `main` (`agc_driver_division`, the denormal-mode tests, `unwrap_self`, `vector_constant_factor_tininess`)
Validation layer (script)	769 tests, layer loaded in 768, none missing, 0 tests with new VUIDs against `main`
Focused set (26 tests: 8 `agc_driver_pointer_*`, nested, image-table, disk-cache, nested snapshot retry, `agc_shader_memory`, `agc_shader_bindless`)	26/26 pass. The slowest is `agc_driver_image_table` at 9.8 s
Negative control	4 of 5 new tests fail. `vsharp_select` passes, as the description says it would
Worktree after restore	clean (`git status --porcelain` empty), 8/8 `agc_driver_pointer_*` pass again

Negative control output (truncated)

1/8 Test #1657: agc_driver_pointer_phi_lane_not_table ...........***Failed    0.12 sec
ShaderRecompiler::Recompile: GetImageResource dword 0 is not a valid runtime value; chain: LoadAddressU32@pc=0x000000c0(GetAddressResource(Phi(Immediate, Phi(ReadConstBuffer@pc=0x00000074(...
2/8 Test #1656: agc_driver_pointer_phi_high_second_lane_table ...***Failed    0.12 sec
ShaderRecompiler::Recompile: GetImageResource dword 0 is not a valid runtime value; chain: LoadAddressU32@pc=0x00000108(GetAddressResource(Phi(Phi(ReadConstBuffer@pc=0x0000007c(...
3/8 Test #1655: agc_driver_pointer_phi_high_table ...............***Failed    0.12 sec
ShaderRecompiler::Recompile: GetImageResource dword 0 is not a valid runtime value; chain: LoadAddressU32@pc=0x000000e0(GetAddressResource(Phi(ReadConstBuffer@pc=0x0000007c(...
4/8 Test #1654: agc_driver_pointer_phi_lane_and_table ...........***Failed    0.12 sec
ShaderRecompiler::Recompile: GetImageResource dword 0 is not a valid runtime value; chain: LoadAddressU32@pc=0x000000ec(GetAddressResource(Phi(Immediate, Phi(BitwiseAnd32(Phi(CompositeExtractU32x4(Ballot(LogicalAnd), Immediate), ...
5/8 Test #1652: agc_driver_pointer_phi_ballot_table .............   Passed    0.36 sec
6/8 Test #1658: agc_driver_pointer_vsharp_select_table ..........   Passed    0.29 sec
7/8 Test #1653: agc_driver_pointer_phi_copies_table .............   Passed    0.41 sec
8/8 Test #1651: agc_driver_pointer_phi_sampler_table ............   Passed    0.94 sec
50% tests passed, 4 tests failed out of 8

Negative control

This PR's own production change over #7395 is ResourceTracker.cpp. It is identical at the head and in the merged tree, and main has not changed it since #7395's head 6cb37cc9ed. So I replaced only that file with #7395's version (6cb37cc9eda523074d5fd844537c28ce2bbe6907) in the merged tree, kept the tests, and rebuilt.

lane_and, high, high_second_lane and lane_not fail with the GetImageResource dword 0 refusal the titles stop on.
#7395's three pointer-Phi tests still pass.
agc_driver_pointer_vsharp_select_table passes without the change, as the description says (it is coverage of #7395's chain).

This matches BunnyFl's Windows negative control. I then restored the merged file, rebuilt, and confirmed the worktree is clean.

Findings

Blocking

None.

Non-blocking

Dead test mode left behind by 4750c85. The commit removed the agc_driver_pointer_phi_high_second_table registration and its --pointer-phi-high-second dispatch. The test code for it is still there:
- PointerHighSecondCode (80 words, ImageTable.cpp#L395)
- PointerMask::HighSecond (L2072)
- its fall-through arm in the code/name selection (L2090-L2091)
Nothing can reach them. Suggest deleting the array and the enum value, and making the last arm of each ternary explicit.
The rule that refuses an edge where both words are immediates is inconsistent with how other non-record edges are handled. See ResourceTracker.cpp#L1035-L1043.
- When both words are Phis of one block, an edge on which both are immediates refuses the whole column.
- An edge with a lane-mask low word is skipped instead. So is an immediate low word whose high word is a record read. In both cases the run-time lookup faults on a pointer that was not captured.
- A null pointer also faults at run time ("unmapped", per #7395's commit). Skipping that edge would therefore be just as safe.
- Refusing it may block the common shape where a loop header merges an SGPR pair zeroed before the loop with the record pointer.
This is not a correctness problem, because the refusal throws. If the refusal is intended, the debt entry could say so. Otherwise, consider skipping the edge like the other pointer-free edges. Since 4750c85 removed the assertion, no test covers this branch.
Stale description items.
- Under "What" (item 1) and "Tested", agc_driver_pointer_phi_high_second_table is still described as a test that "must refuse". 4750c85 removed that test.
- The checklist's "Based on … merged with #7395's head e04a705" is out of date (the head now merges #7395's 6cb37cc9ed).
Reviewers who read the description to see what the tests cover will be misled.
The merge brings the whole stack. Merging this PR before #7395 (and #7223) lands would bring them in with it: about 1,500 of the 1,695 added lines are theirs (ResourceMaterializer.cpp, SpirvImageEmitter.cpp, the ABI to 4, ShaderDiskCache 66). The Depends on: #7395 line covers this. I reviewed this PR's own change over #7395 in detail, and their parts only as far as this change uses them.

Not verified

The title claims (Ratchet & Clank: Rift Apart, Little Nightmares Enhanced Edition, Ghostrunner). I launched no titles.
Hardware other than lavapipe. The NVIDIA RTX 2000 Ada results come from TechLevelZero's bot, not from me.
Windows behaviour, including the libc PATH lines for the nested retry tests.
How the pair walk performs on very large shaders. seen holds (low, high) pairs, which grows quadratically when the two words' Phis are not paired. I have no shader large enough to measure it.
I did not re-review #7223's and #7395's materializer, SPIR-V emitter and disk-cache changes as a whole.

TechLevelZero · 2026-11-02T23:06:02Z

✅ AMD GPU test: passed

PR head afb123a merged onto main a4eca2a · tested 2026-11-02 23:05 UTC · 42 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	✅ 2,687 passed · 13 failed (0 new, 12 also fail on main, 1 unstable on main) · 54 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	2,754 of 3,591 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 (992 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ `agc_driver_pipeline_journal` 120.1 s, ⚠️ `agc_driver_graphics` 99.6 s, ⚠️ `public_headers` 65.8 s, ⚠️ `unwrap_self` 58.7 s, ⚠️ `agc_driver_image_sample_unnormalized` 40.9 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 4 of the last 6 main runs

Failing on main too (12, not counted against this PR)

agc_driver_guest_memory_lifetime_direct
agc_driver_image_table
agc_driver_job_manager_compute_capacity
agc_driver_native_depth_bda_clear
agc_driver_native_depth_buffer_clear
agc_driver_native_depth_clears
agc_driver_native_multisample_depth
agc_idle_publish_wait
agc_submit_progress
agc_submit_progress_no_timeline
guest_mmap_executable
guest_socket_options_multicast

Passed on retry (1, failed in the parallel run, passed when re-run with a 360 s timeout; not counted against this PR)

trace_digest_TraceDigestTests_test_cli_rejects_invalid_event_fields_without_partial_output: Timeout in the parallel run; also passed only on retry in 19 other runs in the last 7 days

Validated but not compared (15, not in the main baseline's validation pass)

agc_driver_nested_image_table
agc_driver_nested_image_table_budget
agc_driver_nested_image_table_large
agc_driver_nested_sampler_table
agc_driver_pointer_phi_ballot_table
agc_driver_pointer_phi_copies_table
agc_driver_pointer_phi_high_second_lane_table
agc_driver_pointer_phi_high_table
agc_driver_pointer_phi_lane_and_table
agc_driver_pointer_phi_lane_not_table
agc_driver_pointer_phi_sampler_table
agc_driver_pointer_vsharp_select_table
agc_shader_nested_sampler_snapshot_retry
agc_shader_nested_snapshot_retry
trace_digest_TraceDigestTests_test_cli_rejects_invalid_event_fields_without_partial_output

Validation IDs already present on main (4 IDs in 510 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (510 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 502 more
VUID-vkCmdCopyImage-srcImage-01548 (6 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-vkCmdCopyImage-srcImage-01551 (6 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (1 test): agc_driver_prepared_exports

Skipped (54)

agc_driver_bda_reads_shadowed_results: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_bda_reads_shadowed_results_per_build: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_copied_writer_label_order: skipped, the device reports a minStorageBufferOffsetAlignment of 4, so the dispatch binds its record at +0x1004 in place and nothing is written back; the test needs an alignment above 4
agc_driver_float16_unused_nan_payloads: skipped, the device has no native f16 conversions with RTE rounding and kept f16 denormals
agc_driver_image_table_vouched_lookups: skipped, the device reports 4 of 4 host-imported pages written after a GPU read, so a watched imported table cannot be vouched for
agc_driver_ngg_vertex_wave_reduction: skipped, s3 describes a 32-lane wave only on 32-lane subgroups, and the device has 64 lanes
agc_driver_unit_shadow: skipped, the device does not import watched host memory for unit shadows
agc_driver_unit_shadow_budget: skipped, the device does not import watched host memory for unit shadows
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_import_settle_initial_dcc: skipped, the device has no deferred settle of host imports: imports under 1 MiB wait for one only on Windows
agc_shadowed_key_stores: skipped, the device imports no watched host memory under a unit shadow
agc_small_tiled_shadow_label: skipped, host imports are unavailable: the test block is not imported and watched
agc_storage_clear_hand_over: skipped, the device imports no watched host memory for the clear-key hand-over
agc_unit_shadow: skipped, the device does not import watched host memory for unit shadows
agc_unit_shadow_completion_store: skipped, the device reports watched host memory written when it imports or reads it
guest_http_uri_locale: 
guest_net_ctl: skipped, the host has no IPv4 address
guest_video_out_color_space: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_hdr: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_high_frame_rate: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_output_mode: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_resolution_status: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
guest_video_out_wqhd_detection: skipped, no display or Vulkan device: SDL_InitSubSystem(VIDEO | GAMECONTROLLER) failed: wayland,x11 not available
video_out_flip_aftergpu: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_backtoback: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_controls: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
video_out_flip_cpuflip: skipped, no display or Vulkan device: DisplayWindow: Vulkan support is either not 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_deferr

Reproduce

git fetch origin afb123a49ec3f7555c274c769abfa5bbd4f87b81 a4eca2a056bf3bbbec98ab42e9d28163e4ea9cb8
git checkout a4eca2a056bf3bbbec98ab42e9d28163e4ea9cb8 && git merge --no-ff afb123a49ec3f7555c274c769abfa5bbd4f87b81
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.}

XairRrR

Re-checked — approved

The only new commit is 58ebe4b, a clean merge of main 4f41f4c (it brings #10000's sampler limits, among others). This PR's own change is the same apart from context lines. Checks and the focused tests pass. The one new VUID the script reports comes from a main test that this PR does not reach.

What changed since my last review

One new commit, 58ebe4b ("Merge main into fix/shader-pointer-key-lane-masks"). Its parents are afb123a, which I reviewed last time, and main 4f41f4c. The commit message lists no conflicts. The range-diff in delta.patch only reorders the cherry-picked commits; the patches are the same.
The PR's own diff is the same as before, apart from context lines. I compared git diff 9090972b7e...afb123a49e with git diff 4f41f4c375...58ebe4b440, ignoring index and hunk lines. Three things differ, and all three are context that comes from main:
- main's FormatVersion is now 52, and this PR still replaces it with 66, which is still above main's.
- A set_tests_properties context line now includes agc_shader_sampler_limits.
- The docs/dev/debt/shader.md context now has #10000's sampler-limit entry.
No new author comments or replies. The description is unchanged.

What I ran

Head 58ebe4b44099cd04f81411a07d2c4d25d95a7a87, main 4f41f4c375f5abb487dbd184fbc150f31994d762, merged tree bad6f1b67d11d12ed7ed1d1a76b7dbba23e6eee1. Release build with GCC 14.2.0 and Ninja, on lavapipe (Mesa 26.1.6, LLVM 19), in a Debian 13 VM.

git diff 9090972b7e91fd266819cbd77a8b058e177b5682...afb123a49ec3f7555c274c769abfa5bbd4f87b81   # compared with the next diff
git diff 4f41f4c375f5abb487dbd184fbc150f31994d762...58ebe4b44099cd04f81411a07d2c4d25d95a7a87
git diff --check 4f41f4c375f5abb487dbd184fbc150f31994d762...58ebe4b44099cd04f81411a07d2c4d25d95a7a87
python3 tools/check_conventions.py --base 4f41f4c375f5abb487dbd184fbc150f31994d762
ctest --test-dir build -j8 --test-load 32 --output-on-failure -R "^(agc_driver_pointer_|agc_driver_nested_|agc_driver_image_table|agc_shader_disk_cache|agc_shader_nested|agc_shader_memory$|agc_shader_bindless$|agc_shader_sampler_limits$)"
env VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation ANYPS5_NO_SHADER_CACHE=1 ANYPS5_REQUIRE_VULKAN=1 VK_LOADER_DEBUG=error,warn,layer ctest --test-dir build -V -R "^agc_driver_native_depth_uniform_planes$"
git log -S native_depth_uniform_planes 4f41f4c375 -- core/libs/prx/libSceAgcDriver
git merge-base --is-ancestor c9987271e3 d1ae628bc2f62459cce1e9872ba5632c938c71f2

Check	Result
`git diff --check`	clean
`check_conventions.py`	All conventions checks passed
Configure, build and libs (run by the script)	ok
Selected tests (run by the script)	2244 ran: 0 new failures, 0 timeouts. 7 known failures also fail on `main`.
Focused set, 27 tests (the 7 `agc_driver_pointer_*` tests, the nested, image-table, disk-cache and nested snapshot retry tests, `agc_shader_memory`, `agc_shader_bindless` and `main`'s new `agc_shader_sampler_limits`)	27/27 pass. The slowest is `agc_shader_sampler_limits` at 14.6 s.
Validation (run by the script)	782 tests, layer loaded in 781, missing in none. One test has new VUIDs: `agc_driver_native_depth_uniform_planes` (see below).

The new VUID comes from main, not from this PR.

The VUID is VUID-vkCmdDispatch-groupCountX-00386 in agc_driver_native_depth_uniform_planes. The script marks that test "(test not run on main)".
The test comes from main's c998727 (2026-10-31). That commit is not in the validation baseline d1ae628, so the baseline has no run of it to compare with.
The test exercises Graphics/src/DepthSurface.cpp's import of a native multisample D32S8 target. This PR changes no libSceAgcDriver production file.
I reran the test with the layer in the merged tree. It reports groupCountX (69144) exceeds device limit maxComputeWorkGroupCount[0] (65535), which belongs to that depth-import dispatch, not to an image-table path. I did not run the test on a main-only build.

Validation excerpt

Validation Error: [ VUID-vkCmdDispatch-groupCountX-00386 ] | MessageID = 0x8e927036
vkCmdDispatch(): groupCountX (69144) exceeds device limit maxComputeWorkGroupCount[0] (65535).

Negative control: not rerun. The tests and this PR's ResourceTracker.cpp lines are the same as at afb123a, where four of the five new tests failed without the change. Rerunning it here would also be harder: main has since changed ResourceTracker.cpp (#10000, #12132), so dropping in #7395's version of the file would also revert main's changes.

Previous findings

Dead test mode (PointerHighSecondCode, PointerMask::HighSecond). Still open, and unchanged:
- PointerHighSecondCode is still at ImageTable.cpp#L395.
- PointerMask::HighSecond is still at L2072.
- Nothing can reach either of them.
An edge where both words are immediates refuses the whole column, while other pointer-free edges are skipped. Still open, and unchanged. It is now at ResourceTracker.cpp#L1039-L1047. It is not a correctness problem, because the refusal throws.
Stale description items. Still open:
- "What" item 1 and "Tested" still describe agc_driver_pointer_phi_high_second_table as a test that "must refuse", but 4750c85 removed that test.
- The checklist still says the branch is "merged with #7395's head e04a705".
The merge brings in the whole stack (#7395, #7223). Still applies, and is covered by Depends on: #7395.

New findings

Blocking

None.

Non-blocking

None.

Not verified

The title claims (Ratchet & Clank: Rift Apart, Little Nightmares Enhanced Edition, Ghostrunner). I launched no titles.
Hardware other than lavapipe, and Windows.
agc_driver_native_depth_uniform_planes with the validation layer on a main-only build. I attribute its VUID to main from where the test comes from and from the code it exercises.
I did not reread how #10000's sampler limits, now on main, interact with the stack's nested S# budget beyond the passing tests. agc_shader_sampler_limits and the nested sampler tests pass together.

TechLevelZero · 2026-11-03T15:24:03Z

✅ NVIDIA GPU test: passed

PR head 58ebe4b merged onto main 4f41f4c · tested 2026-11-03 15:23 UTC · 30 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	✅ 2,776 passed · 6 failed (0 new, 3 also fail on main, 3 unstable on main) · 48 skipped (`ANYPS5_REQUIRE_VULKAN=1`)
Test setup	CPU-only tests without a GPU; GPU tests (and any CPU-only test that failed or skipped without one) once on the GPU with `VK_LAYER_KHRONOS_validation` and `ANYPS5_NO_SHADER_CACHE=1`; main ran the same way
Test selection	2,830 of 3,811 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 (945 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ `agc_driver_pipeline_journal` 120.2 s

Unstable on main (3, 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: failed in 6 of the last 10 main runs
agc_submit_progress_no_timeline: failed in 5 of the last 10 main runs

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

agc_driver_image_table
guest_mmap_executable
guest_socket_options_multicast

Validated but not compared (2, not in the main baseline's validation pass)

agc_shader_nested_sampler_snapshot_retry
agc_shader_nested_snapshot_retry

Validation IDs already present on main (4 IDs in 524 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (524 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 516 more
VUID-vkCmdCopyImage-srcImage-01548 (9 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, agc_driver_stencil_null_feedback, and 1 more
VUID-vkCmdCopyImage-srcImage-01551 (9 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, agc_driver_stencil_null_feedback, and 1 more
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (1 test): agc_driver_srgb8_color_target

Skipped (48)

agc_driver_bda_reads_shadowed_results: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_bda_reads_shadowed_results_per_build: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_sample_order_grid_reference: skipped, the device does not keep depth drawn at custom sample locations at those locations, even after a transition that names them: the composite of uniform renders holds the guest's depth in 0 of 3005 sloped samples, so no path can be compared with it bit for bit
agc_driver_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_stencil_ref_export: skipped, the device lacks VK_EXT_shader_stencil_export
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_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 co

Reproduce

git fetch origin 58ebe4b44099cd04f81411a07d2c4d25d95a7a87 4f41f4c375f5abb487dbd184fbc150f31994d762
git checkout 4f41f4c375f5abb487dbd184fbc150f31994d762 && git merge --no-ff 58ebe4b44099cd04f81411a07d2c4d25d95a7a87
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.}

boykopovar#7223 has merged. The conflicts are this branch's additions on top of it: its tests in the image-table test lists, pointerKeyed in the image-table column and its equality check, the plan encoder's field count and list version, and FormatVersion 63. agc_shader_memory_tests.cmake merged without a conflict but registered agc_shader_nested_snapshot_retry twice: boykopovar#7223's block after agc_shader_descriptor_phis and this branch's before it. The merge keeps this branch's block, which adds agc_shader_nested_sampler_snapshot_retry, at boykopovar#7223's place.

e94fb3f is boykopovar#7395 with its merge of main after boykopovar#7223. The two conflicts are this branch's own changes over boykopovar#7395: FormatVersion 66, and its wording of the pointer-keyed image-table debt entry.

TechLevelZero · 2026-11-04T09:29:46Z

✅ NVIDIA GPU test: passed

PR head 9eab037 merged onto main 5b1d381 · tested 2026-11-04 09:29 UTC · 27 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	✅ 2,849 passed · 6 failed (0 new, 3 also fail on main, 3 unstable on main) · 49 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	2,904 of 3,883 tests run: the ones this change can reach, chosen by `tools/select_tests.py` from main's build graph
Vulkan validation	✅ no new validation IDs vs main (985 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ `agc_driver_pipeline_journal` 120.1 s

Unstable on main (3, these also failed in recent main runs on this machine; not counted against this PR)

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

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

agc_driver_image_table
guest_mmap_executable
guest_socket_options_multicast

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

agc_shader_nested_sampler_snapshot_retry

Validation IDs already present on main (13 IDs in 547 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (547 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 539 more
VUID-VkShaderModuleCreateInfo-pCode-08737 (8 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_float16_fma_rounding, agc_driver_float16_unused_nan_payloads, agc_driver_packed_alu, agc_driver_packed_f16_denormal_modes, agc_driver_packed_float16_native
VUID-VkShaderModuleCreateInfo-pCode-08739 (8 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_float16_fma_rounding, agc_driver_float16_unused_nan_payloads, agc_driver_packed_alu, agc_driver_packed_f16_denormal_modes, agc_driver_packed_float16_native
VUID-VkShaderModuleCreateInfo-pCode-08741 (8 tests): agc_driver_dpp16_operands, agc_driver_dpp8_pk_fmac, agc_driver_dx10_clamp, agc_driver_float16_fma_rounding, agc_driver_float16_unused_nan_payloads, agc_driver_packed_alu, agc_driver_packed_f16_denormal_modes, agc_driver_packed_float16_native
VUID-vkCmdCopyImage-srcImage-01548 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-vkCmdCopyImage-srcImage-01551 (7 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (2 tests): agc_driver_prepared_exports, agc_driver_srgb8_color_target
VUID-VkPipelineShaderStageCreateInfo-module-parameter (1 test): agc_driver_float16_unused_nan_payloads
VUID-VkPipelineShaderStageCreateInfo-pSpecializationInfo-06849 (1 test): agc_driver_float16_unused_nan_payloads
VUID-VkSamplerCreateInfo-addressModeU-01079 (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeU-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeV-parameter (1 test): agc_driver_image_gather_volume
VUID-VkSamplerCreateInfo-addressModeW-parameter (1 test): agc_driver_image_gather_volume

Skipped (49)

agc_driver_bda_reads_shadowed_results: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_bda_reads_shadowed_results_per_build: skipped, the device reports watched host memory written when it imports or reads it
agc_driver_gcn_display_present: skipped, no display: Vulkan support is either not configured in SDL or not available in current SDL video driver (offscreen) or platform
agc_driver_sample_order_grid_reference: skipped, the device does not keep depth drawn at custom sample locations at those locations, even after a transition that names them: the composite of uniform renders holds the guest's depth in 0 of 3005 sloped samples, so no path can be compared with it bit for bit
agc_driver_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_stencil_ref_export: skipped, the device lacks VK_EXT_shader_stencil_export
agc_host_import_sparse: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_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 av

Reproduce

git fetch origin 9eab0374897613a391d0d6cd43d39d4ebbace6d5 5b1d381d5ad5d9f923326b18479c3bd889cdd588
git checkout 5b1d381d5ad5d9f923326b18479c3bd889cdd588 && git merge --no-ff 9eab0374897613a391d0d6cd43d39d4ebbace6d5
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.}

4769d4a is boykopovar#8939 with boykopovar#7395's merge of main after boykopovar#7223. The one conflict is this branch's FormatVersion 67.

EagleBearerYS added 6 commits October 21, 2026 04:37

github-actions Bot added build shader core/shader docs agc libSceAgc* needs rebase Conflicts with main, rebase needed labels Oct 21, 2026

EagleBearerYS added 3 commits October 21, 2026 03:09

Merge boykopovar#7223's 8716815 into feat/shader-nested-sampler-tables

92200f9

Merge main into feat/shader-nested-sampler-tables

6c1371b

Merge feat/shader-nested-sampler-tables into fix/shader-pointer-key-l…

e9d4626

…ane-masks

github-actions Bot removed the needs rebase Conflicts with main, rebase needed label Oct 21, 2026

EagleBearerYS added 2 commits October 21, 2026 06:59

EagleBearerYS added a commit to EagleBearerYS/AnyPS5 that referenced this pull request Oct 21, 2026

docs(cloud-2): PRM-051 result, boykopovar#8939's edge-paired pointer …

1a63e13

…walk red/green and Linux CI mirror

frankea mentioned this pull request Oct 21, 2026

Recompiler: a T# reached through two loads after a dynamically indexed read (pointer → pointer → T#, pointer → V# → buffer → T#) stops Ratchet & Clank (PPSA01474) and Guardians of the Galaxy (PPSA01748) #6191

Open

oneandonlydean mentioned this pull request Oct 21, 2026

feat(shader): resolve nested image tables through runtime pointers #7223

Merged

12 tasks done

github-actions Bot added the needs rebase Conflicts with main, rebase needed label Oct 21, 2026

EagleBearerYS added 2 commits October 21, 2026 13:03

Merge feat/shader-nested-sampler-tables into fix/shader-pointer-key-l…

8704f73

…ane-masks Brings boykopovar#7395's merge of main bd4d801, with boykopovar#8991. It merged without conflicts.

github-actions Bot removed the needs rebase Conflicts with main, rebase needed label Oct 21, 2026

github-actions Bot removed the review queue Next in the maintainer review queue: checks pass, no open change request, no conflict label Nov 2, 2026

EagleBearerYS added 2 commits November 2, 2026 06:16

Merge boykopovar#7395's main merge into fix/shader-pointer-key-lane-m…

afb123a

…asks

EagleBearerYS added a commit to EagleBearerYS/AnyPS5 that referenced this pull request Nov 2, 2026

Merge boykopovar#8939's main merge into fix/shader-table-family-6191

0a4f7cd

github-actions Bot added needs rebase Conflicts with main, rebase needed review queue Next in the maintainer review queue: checks pass, no open change request, no conflict and removed needs rebase Conflicts with main, rebase needed labels Nov 2, 2026

liskeee pushed a commit to liskeee/AnyPS5 that referenced this pull request Nov 2, 2026

Merge boykopovar#8939's main merge into fix/shader-table-family-6191

04d09ee

# Conflicts:

BunnyFl approved these changes Nov 2, 2026

View reviewed changes

XairRrR approved these changes Nov 2, 2026

View reviewed changes

github-actions Bot added the needs rebase Conflicts with main, rebase needed label Nov 3, 2026

Merge main into fix/shader-pointer-key-lane-masks

58ebe4b

github-actions Bot removed the review queue Next in the maintainer review queue: checks pass, no open change request, no conflict label Nov 3, 2026

Merge main into feat/shader-nested-sampler-tables

63b5c09

XairRrR approved these changes Nov 3, 2026

View reviewed changes

github-actions Bot removed the needs rebase Conflicts with main, rebase needed label Nov 3, 2026

This was referenced Nov 3, 2026

fix(shader): accept a nested S# whose V# gives the key of a sibling V#'s T# array #15555

Open

fix(libkernel): give each APR submit priority its own queue #15649

Merged

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

EagleBearerYS added 3 commits November 3, 2026 20:45

docs(shader): note that a nested child's T# skips registered views

014d1e3

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

hugodearaujo mentioned this pull request Nov 4, 2026

perf(agc): bind a one-layer 2D array table entry to the 2D texture of its surface #15710

Open

12 tasks done

Conversation

EagleBearerYS commented Oct 21, 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 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

oneandonlydean commented Oct 21, 2026

Uh oh!

EagleBearerYS commented Oct 21, 2026

Uh oh!

EagleBearerYS commented Oct 21, 2026

Uh oh!

Tamalero commented Oct 21, 2026

Uh oh!

EagleBearerYS commented Oct 21, 2026

Uh oh!

TechLevelZero commented Nov 2, 2026

✅ NVIDIA GPU test: passed

Uh oh!

TechLevelZero commented Nov 2, 2026

✅ NVIDIA GPU test: passed

Uh oh!

BunnyFl left a comment

Choose a reason for hiding this comment

Uh oh!

XairRrR left a comment

Choose a reason for hiding this comment

Reviewed — approved

What I ran

Results

Negative control

Findings

Blocking

Non-blocking

Not verified

Uh oh!

TechLevelZero commented Nov 2, 2026

✅ AMD GPU test: passed

Uh oh!

XairRrR left a comment

Choose a reason for hiding this comment

Re-checked — approved

What changed since my last review

What I ran

Previous findings

New findings

Blocking

Non-blocking

Not verified

Uh oh!

TechLevelZero commented Nov 3, 2026

✅ NVIDIA GPU test: passed

Uh oh!

TechLevelZero commented Nov 4, 2026

✅ NVIDIA GPU test: passed

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

10 participants

EagleBearerYS commented Oct 21, 2026 •

edited

Loading

github-actions Bot commented Oct 21, 2026 •

edited

Loading