[TheSovietWaffle]

What

Tessellation lowering accepted a computed control-point index only for hull LDS reads (#4727, extended to selects and phis by 9f429e65a). Two gaps, reported in #8051:

1. Domain reads were refused. Onimusha: Way of the Sword (PPSA27836) picks one of its patch's three control points from data in its first scene (v_cmp results feed v_cndmask/v_add_co_ci, then v_lshl_add_u32 scales the index by the 32-byte stride before a buffer_load from the off-chip ring), and main stops with tessellation lowering: the tessellation ring or LDS address is not affine in the control point and relative patch at pc 0x354.
2. A computed hull read left its patch when the relative patch index is also data. The read keeps the program's own address, which contains relative * inputs * localStride; when the program also uses its relative patch index as data, the bases are replaced by the grouped patch instead of 0, so that term stayed and the read landed in another patch's slots.

Changes:

• Off-chip EvaluationInput reads get the same dynamic term as hull LDS reads. It must be a whole number of control points, and the backend clamps the domain gl_in index to the last output control point, as it already does for hull reads (the hull input control points entry in docs/dev/debt/shader.md now covers both).
• A value read from the patch's own ring or LDS no longer counts as depending on the relative patch index: the lowered read takes its patch implicitly.
• When the relative patch index is used as data, computed reads (hull and domain) subtract their relative patch term before the bases are replaced. Programs that do not use it as data keep the program's own address, as 9f429e65a's tests require.

Tested

Windows 10 Pro 22H2 (19045), WinLibs GCC 15.2.0 r7, Release, on main 3aa35dc01.

• New GPU case DomainComputedPoint in agc_driver_tessellation: a copy of DomainCode that reads control points 1 and 2 at cndmask(0, 1, FLT_MAX < x0) * 16 + base, always the static address at run time; it draws the same patches as DomainCode. Without the change it fails with the "not affine" refusal.
• New unit case in agc_tessellation_lowering: a hull read at a computed index in a program that also references its relative patch gets address - patch * 96; without the subtraction it fails. 9f429e65a's cases pass unchanged.
• agc_driver_tessellation, agc_tessellation_lowering and agc_tessellation_factor_ring pass on both GPUs below. On an earlier base (8c9161eac, with fix(shader): preserve read lane dominance across branches #8089's change merged in) all 806 agc tests ran on both: 805 pass on each, and the one failure on each (agc_driver_pass_input_hazard on the RTX 3080, agc_trimmed_import on the Radeon) fails the same way on unchanged main.
• The retained Onimusha domain request prepares (22,268 SPIR-V words); with it the title ran past the first scene for over 4 hours, with local workarounds for AGC: pixel shaders that export a stencil reference value abort on devices without VK_EXT_shader_stencil_export (NVIDIA; Demon's Souls, 1080p renderconfig) #4276 and AGC: compute buffer read of a native 4-sample D32 depth target needs sample-preserving publication (Onimusha: Way of the Sword, PPSA27836) #6570, so this is not a claim that the title works.
• tools/check_conventions.py --base origin/main passes. tools/pr_overlap.py was not run (no gh on this host).

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: Fixes shader: a domain program reading a control point at an index computed from data is refused as "not affine", and a computed hull read leaves its patch when the relative patch is also data; fix branch ready #8051
• Unblocks: PPSA27836 Onimusha: Way of the Sword (PPSA27836) #5769 (the first-scene tessellation stop; the title reaches it only with AGC: pixel shaders that export a stencil reference value abort on devices without VK_EXT_shader_stencil_export (NVIDIA; Demon's Souls, 1080p renderconfig) #4276 and AGC: compute buffer read of a native 4-sample D32 depth target needs sample-preserving publication (Onimusha: Way of the Sword, PPSA27836) #6570 worked around)
• GPUs tested: NVIDIA RTX 3080, 617.42, Windows 10 22H2; AMD Radeon RDNA 2 iGPU (1002:164e), 32.0.21045.5002, Windows 10 22H2
• AI-assisted: yes

🤖 Generated with Claude Code

[github-actions]

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 5021	docs/dev/debt/shader.md
PR 5371	docs/dev/debt/shader.md
PR 5941	docs/dev/debt/shader.md
PR 6009	docs/dev/debt/shader.md
PR 6235	core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp docs/dev/debt/shader.md
PR 6280	docs/dev/debt/shader.md
PR 6412	docs/dev/debt/shader.md
PR 6488	docs/dev/debt/shader.md
PR 6657	core/libs/prx/libSceAgcDriver/tests/execution/Tessellation.cpp core/shader/recompiler/Optimization/src/TessellationLowering.cpp
PR 6683	core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 6879	docs/dev/debt/shader.md
PR 6924	docs/dev/debt/shader.md
PR 6947	core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 6977	docs/dev/debt/shader.md
PR 7085	docs/dev/debt/shader.md
PR 7201	core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 7395	docs/dev/debt/shader.md
PR 7484	docs/dev/debt/shader.md
PR 7642	core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 7684	docs/dev/debt/shader.md
PR 7888	core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp docs/dev/debt/shader.md
PR 7922	docs/dev/debt/shader.md
PR 8004	docs/dev/debt/shader.md
PR 8077	docs/dev/debt/shader.md

[wachi35676]

Results on current main (02fb1ce53, 395 commits after this PR's base) with f95471ec6 merged (clean). Tested on an RTX 5070 Ti (NVIDIA 617.42) and an AMD Radeon iGPU (AMD proprietary 2.0.353). Windows 11 (26300), WinLibs GCC 15.2.0, Release, ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1.

• agc_tessellation_lowering, agc_tessellation_factor_ring and agc_driver_tessellation pass 3 of 3 on each GPU.
• Negative control: this PR's two test files alone on main fail on both GPUs. agc_tessellation_lowering fails with a hull read at a computed control-point index stays in its patch when the relative patch is also data, and agc_driver_tessellation fails with tessellation lowering: the tessellation ring or LDS address is not affine in the control point and relative patch at pc 0x34.
• The 9 spirv tests pass on each GPU.
• Related set of 98 tests (tessellation, lds, subgroup, shader, stencil, msaa, multisample, label, write_tracking, resident and so on, plus agc_driver_graphics): 98 of 98 pass on each GPU. agc_driver_stencil_ref_export skips on NVIDIA because the GPU has no stencil export.

Not run: the validation layer and a title.

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

[DavidAmsellem]

First Linux run of this PR, on AMD with RADV. Head f95471ec6 merged onto main 2fee39f1b (clean merge), built in full and compared with the same main, on the same machine.

AMD Radeon RX 7900 XT (RADV NAVI31, gfx1100), Mesa 25.2.8, Linux Mint 22.3 (Ubuntu 24.04 base), kernel 7.0.0-38, GCC 13.3, Release.

Full suite (ANYPS5_REQUIRE_VULKAN=1 ctest -j8) at 46d1d2db3: 1715/1716 with this PR, against 1715/1716 on main. The one failure is the same on both, so this PR adds none: agc_driver_guest_memory_lifetime (#7784, which #7849 addresses).

Comparing the JUnit output with main: this PR adds no ctest entry, and none is skipped or removed. The tests it extends, agc_tessellation_lowering and agc_driver_tessellation, run and pass.

GPUs tested: AMD Radeon RX 7900 XT (RADV NAVI31, gfx1100), Mesa 25.2.8, Linux Mint 22.3 / kernel 7.0.0-38

AI-assisted: yes