[milescoviello]

What

A vertex-path program that touches LDS or GDS or writes memory refuses any s_barrier:

SPIR-V emission failed: ... opcode=Barrier reason=s_barrier in a vertex program that accesses LDS or GDS or writes memory is not implemented

NGG passthrough programs (VGT_SHADER_STAGES_EN PRIMGEN_EN + PRIMGEN_PASSTHRU_EN) start with the s_barrier that orders GS_ALLOC_REQ. Crimson Desert (PPSA25286) registers one, routing 0x2402000, 2133 instructions. Its only barrier is at pc 0x20, after a few scalar instructions, and it writes a buffer about 7 KB later (buffer_atomic_add at 0x1e78, then three buffer_store_dwordx4). Its registration preparation fails with the line above, and the draw that uses it rethrows it.

A barrier that runs before every memory access of the program has nothing to order: no load, store, atomic, image or LDS access can happen before it in any wave. Such a barrier now emits nothing, the same as the barrier of a program with no memory access. The check walks the entry block and its straight-line successors. A barrier behind a memory access, a branch or a merge still throws.

Tested

Linux (Ubuntu 24.04, GCC 13.3, Release), RTX 4070 Ti, NVIDIA 595.99.

• agc_vertex_barrier: a vertex program with s_barrier before a ds_write_b32, and one with a scalar instruction ahead of the barrier, compile without an OpControlBarrier. The existing case with the LDS write before the barrier is still refused. With main's SpirvFlowEmitter.cpp the new cases fail with the refusal above. 3 of 3 runs pass.
• ctest (^agc_vertex_barrier|^agc_driver_(ngg|vertex|mesh|graphics|prepared|tessellation)): 30 of 30 pass.

Crimson Desert (PPSA25286) with #6612, which gets it past its first stop:

Build	Result
main 5d9dcc4 + #6612	stops 123 s in on the refusal above
main 5d9dcc4 + #6612 + this PR	reaches the 150 s limit with no stop

Validated on Windows 11 with an AMD Radeon 860M (RDNA 3.5, driver 32.0.13058.8001) by @LeandroLP23 (comment), in one build of main 6efb726 with this head and 13 other pull requests: agc_vertex_barrier passes 3 runs of 3. 713 of 716 agc_* tests pass; the 3 that fail fail the same way on main there.

Not checked on Intel or on a console.

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 (the title runs need feat(agc): draw tessellation followed by an NGG geometry stage #6612)
• Unblocks: none
• GPUs tested: NVIDIA GeForce RTX 4070 Ti, driver 595.99, Ubuntu 24.04; AMD Radeon 860M, driver 32.0.13058.8001, Windows 11 (tests, by @LeandroLP23)
• 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 4072	core/shader/recompiler/SpirvBackend/src/SpirvFlowEmitter.cpp
PR 5021	core/shader/recompiler/SpirvBackend/src/SpirvFlowEmitter.cpp docs/dev/debt/shader.md
PR 5371	docs/dev/debt/shader.md
PR 5426	docs/dev/debt/shader.md
PR 5500	docs/dev/debt/shader.md
PR 5941	docs/dev/debt/shader.md
PR 6009	docs/dev/debt/shader.md
PR 6235	docs/dev/debt/shader.md
PR 6674	docs/dev/debt/shader.md
PR 6767	docs/dev/debt/shader.md
PR 6924	docs/dev/debt/shader.md
PR 6947	core/shader/recompiler/SpirvBackend/src/SpirvFlowEmitter.cpp
PR 6977	docs/dev/debt/shader.md
PR 7085	docs/dev/debt/shader.md
PR 7223	docs/dev/debt/shader.md
PR 7332	core/shader/recompiler/SpirvBackend/src/SpirvFlowEmitter.cpp docs/dev/debt/shader.md
PR 7392	core/shader/recompiler/SpirvBackend/src/SpirvFlowEmitter.cpp docs/dev/debt/shader.md
PR 7395	docs/dev/debt/shader.md
PR 7429	core/shader/recompiler/SpirvBackend/src/SpirvFlowEmitter.cpp docs/dev/debt/shader.md
PR 7484	docs/dev/debt/shader.md
PR 7642	core/shader/recompiler/SpirvBackend/src/SpirvFlowEmitter.cpp

[LeandroLP23]

AMD on Windows: the tests this PR names pass on an RDNA 3.5 GPU.

Radeon 860M (RDNA 3.5, driver 32.0.13058.8001), Windows 11 26300, WinLibs GCC 15.2.0 r7, Release. main 6efb7261b with this PR (1ce582ebc) merged, together with #7296, #7446, #7377, #7445, #6708, #6748, #6785, #6706, #7449, #7391, #7489, #7484, #7514. They were built once, so the result is for the combination; each merge was clean.

• agc_vertex_barrier: pass, 3 runs of 3.
• Every agc_* test on the combined build: 713 of 716 pass. The 3 that fail, agc_driver_host_import_lifetime, agc_driver_quad_pixel_mask and agc_driver_layered_volume_vertex_no_gpl, fail the same way on unchanged main here, so none comes from these PRs.

AI-assisted

[milescoviello]

Thanks @LeandroLP23 for the AMD run on Windows. I added it to the description.

AI-assisted: yes (Claude Code)

[milescoviello]

Rebased onto main 0c25997 (7092dde). The only conflict was in docs/dev/debt/shader.md, where main had updated the clip/cull distance entry next to this PR's s_barrier entry. Both are kept: this PR's s_barrier text and main's clip/cull text. The 35 tests matching ^agc_vertex_barrier|^agc_driver_(ngg|vertex|mesh|graphics|prepared|tessellation) pass.

AI-assisted: yes (Claude Code)

[adriadam10]

Checked the entry-block walk (entry has no predecessors, any branch/merge stops it, memory-access set covers const, LDS, buffer, address and image ops), the new tests (fail on main without the change; the LDS-before-barrier case still throws) and the TechnicalDebt line. Merging through the queue.