[XairRrR]

What

docs/dev/ARCHITECTURE.md lists the draw pipelines that are created whole instead of linked from pipeline libraries: per-face depth bias, a split alpha blend into an 8_8 target with the alternate component order, rect lists, geometry shaders and "a stage without a variant id". libraryKeys in core/libs/prx/libSceAgcDriver/Graphics/src/Pipeline.cpp builds more pipelines whole than that. Each of these cases was added after the sentence was written:

• barycentrics emulated through generated tessellation stages (BarycentricPatches);
• tessellation on a driver without a usable tessellation pipeline cache, which today means MoltenVK (TessellationPipelineCacheUsable);
• a pixel shader with a fragment shader interlock (FragmentInterlock);
• a pixel shader whose LDS array is kept in a storage buffer (RecompileResult::functionLdsBuffer);
• every pipeline when APS5_PIPELINE_STATS=1 is set (Context::pipelineExecutableInfo).

The stage test is also not a "variant id" any more: it is an empty RecompileResult::Identity. So a reader who checks, for example, a fragment shader interlock draw or a tessellated draw on MoltenVK against the doc expects pipeline libraries, but the code creates the pipeline whole.

The sentence now lists every case libraryKeys rejects and points to libraryKeys in Pipeline.cpp. Only the document changes.

Tested

Arch Linux 7.2.11-zen, GCC 16.2.1, Ninja, Release, Intel Core i9-14900KF, NVIDIA GeForce RTX 4070 Ti SUPER (driver 615.78.08).

This is a documentation-only change, so nothing was built or run. I checked each listed case against the condition in libraryKeys (Pipeline.cpp) and its helpers (BarycentricPatches, FragmentInterlock, TessellationPipelineCacheUsable, the APS5_PIPELINE_STATS setup in VulkanDevice.cpp, pixelFunctionLdsBuffer in SpirvEmitter.cpp) on current main. The new relative link resolves. python3 tools/check_conventions.py --base origin/main reports "All conventions checks passed.", and git diff --check origin/main is clean. tools/pr_overlap.py --branch finds no overlap with the open pull requests it could compare. For the 11 that conflict with main and share ARCHITECTURE.md (#5925, #6412, #6743, #7642, #7684, #9356, #9387, #9508, #10069, #10193, #11351), I checked their diffs by hand: none of them edits this sentence. #11454 shows it only as context.

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: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

[adriadam10]

Checked each listed case against the condition in libraryKeys (Pipeline.cpp:333-337), the TessellationPipelineCacheUsable and APS5_PIPELINE_STATS call sites, and the link; docs-only, Conventions green; merged.

[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 5925	docs/dev/ARCHITECTURE.md
PR 6412	docs/dev/ARCHITECTURE.md
PR 6743	docs/dev/ARCHITECTURE.md
PR 7642	docs/dev/ARCHITECTURE.md
PR 7684	docs/dev/ARCHITECTURE.md
PR 9356	docs/dev/ARCHITECTURE.md
PR 9387	docs/dev/ARCHITECTURE.md
PR 9508	docs/dev/ARCHITECTURE.md
PR 10069	docs/dev/ARCHITECTURE.md
PR 10193	docs/dev/ARCHITECTURE.md
PR 11351	docs/dev/ARCHITECTURE.md