feat(agc): implement 10_11_11 unorm color targets through 32-bit uint attachments

[Celegans12]

What

Color targets with CB_COLOR_INFO format 6 (COLOR_10_11_11) and NUMBER_UNORM throw unsupported color format 6 number type 0 component swap 0; Vulkan has no unorm 11_11_10 format.

• Such a target becomes an R32_UINT attachment in the layout of image format 30 (unorm11 R and G in bits 0-21, unorm10 B in 22-31; Mesa ac_formats.c, gfx10-rsrc.json), as feat(agc): implement 10_11_11 unorm images through 32-bit uint views #805 reads it.
• The registered pixel shader (perf(agc): baseline shader recompilation on registration #639) packs each float MRT export like format-30 image stores (EmitFormatStoreComponent: nearest even, NaN and negatives to 0) behind spec constant ExportPackingBase + target; SpecializeFragmentOutputs makes a packed target's output the uint4 ValidateShaders requires. TechnicalDebt: the CB is assumed to round alike.
• targetExportPacking is draw state: outside the static ABI key, inside the module key, the result memo, the draw key and the serialized request (version 14, after feat(shader): carry the shader float mode into the recompile request #1816's 13). Specialization failures are now rethrown instead of terminating inside std::call_once on MinGW.
• Still throws: blending, partial RGB writes, UNORM16/SNORM16 exports, DCC, ROUND_MODE, component swaps, other number types.
• PPSA26344 v1.008: the G-buffer pass writes slot 2 (0x5065bd0000, 3840x2160) this way; 79 skips in one run.
• feat(agc): accept UINT16_ABGR color exports into unsigned integer targets #1073 and feat(agc): accept 8_8_8_8 UINT color targets #1736 edit the same ValidateShaders check; feat(agc): vertex clip and cull distances #423, feat(agc): single-sample sample masks as a per-quad-pixel discard #448 and test(shader): cover stage-specific float modes and legacy captures #1442 change the request version lines; fix(agc): support alternate channel order in 16-bit color targets #1349 edits DecodeColorFormat and DecodeState, with no conflict beyond its own with main.

Tested

• Windows 11, MinGW-w64 GCC 15.2.0, based on main f145139: configure -DBUILD_TESTING=ON, full build, --target libs, ctest:
• 489/489 tests pass (ctest --output-on-failure)
• Linux (WSL Ubuntu 24.04, GCC 13.3, Mesa llvmpipe, upstream's job configuration), merged with main f145139: 498/498 pass
• New agc_driver_packed_unorm_color_target draws through the registered pixel shader (tiled and linear; 32_ABGR and FP16_ABGR; ties, NaN, infinities) and checks exact words. It fails with the packing in the static key (unregistered program is not a compute shader) or out of the module key.
• unorm10_11_11TargetTests (agc_driver_graphics): one artifact, two specializations. agc_shader_memory: version 1-13 requests decode unpacked. None of the three builds on main.
• Not tested: console behaviour.

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 in TechnicalDebt
• 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
• AI-assisted: yes

[adriadam10]

Checked on current main: the only conflict was with #1073 in ShaderValidation.cpp (uintExport vs packed in the same check). I merged both conditions (a target expects u32x4 if either applies, each with its own message) and rebased. Build + ctest of libSceAgcDriver green (267/267); the new tests don't compile without the change and agc_driver_packed_unorm_color_target passes.

Unmodelled cases still throw and the CB-rounding assumption is recorded in TechnicalDebt. Merging once CI is green on the rebased head.

[github-actions]

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
#1607	docs/dev/TechnicalDebt.md L332-333

[adriadam10]

Checked on current main: build + ctest of libSceAgcDriver/libSceAgc green, the new tests don't compile without the change, agc_driver_packed_unorm_color_target draws through the registered pixel shader on lavapipe, unmodelled cases still throw, the CB-rounding assumption is in TechnicalDebt. Merging once the full Linux/Windows build (dispatched manually) is green.