[Tamalero]

What

DecodeState throws blend bypass conflicts with enabled blending when CB_BLENDn_CONTROL enables blending for a target whose CB_COLORn_INFO has BLEND_BYPASS set. That is bit 16 in Mesa's src/amd/registers/gfx10.json.

The bit makes the color block skip the blender for that target. Mesa sets it "according to docs" for UINT/SINT and 8_24/24_8/X24_8_32_FLOAT targets, which the blender cannot process (src/amd/common/ac_descriptors.c, shared by radeonsi and RADV). So a set bit with blending enabled is a valid state, not a conflict: the target is written unblended.

Blending is now enabled only when it is requested and not bypassed.

Tested

Linux (CachyOS), GCC 16, RTX 4090, Release, on main 1c5e628:

• agc_driver_graphics: the new BlendBypassTests check that blending stays off for a bypassed target with blending enabled, and stays on for an identical target without the bit. With main's State.cpp the test fails with blend bypass conflicts with enabled blending; with the change it passes.
• Guardians of the Galaxy (PPSA01748): the title issues these draws right after its loading screen. Without this change it stops there with the error above; with it, it reaches its first UI. That run also carries feat(agc): implement image and sampler tables indexed at runtime offsets #757, feat(agc): implement image tables loaded through 64-bit pointers #758 and the metadata pass PR.
• tools/check_conventions.py --base origin/main: passes.

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
• Unblocks: PPSA01748 (no bring-up discussion yet)
• AI-assisted: yes

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

🤖 Generated with Claude Code

[kr4y4]

A Windows and Intel run, since the description covers Linux on an RTX 4090.

Setup: Windows 11 (26200), WinLibs GCC 15.2.0 r7, Release, CI configure flags, an RTX 4070 Laptop GPU and an Intel Raptor Lake-S iGPU. The head baaad78 merged into main 495bf91, which passes 1023 of 1023 on both GPUs.

• agc_driver_graphics passes on both GPUs with --repeat until-fail:3. The full suite (ctest --timeout 120 --parallel 4) passes 1023 of 1023 on both.
• With State.cpp as on main, agc_driver_graphics fails on both GPUs with blend bypass conflicts with enabled blending, as in the description.

Written with AI assistance.

[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
#5200	core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp L896-897, L900-917