[adriadam10]

What

agc_driver_flat_apertures (added in #1190) has timed out on every Linux CI run of main since it was merged (Timeout 120.09 sec, main run 38097151664).

The test isn't hung; it's slow on lavapipe. Each of its 126 rows compiles its own shader, which takes 0.4–0.95 s per row on CI's lavapipe (Mesa 25.2.8, /dev/udmabuf denied as on CI). The 120 rows that run there take 59 s on an idle machine and more than 150 s on a loaded one, against a 120 s CI timeout and the 30 s rule in CONTRIBUTING. On RADV the whole test takes 2 s.

It's now registered as six tests, agc_driver_flat_apertures_0 … _5. Each runs --part K/6, every sixth row starting at row K, so they share the slow and fast rows evenly. This follows the pattern agc_driver_constant_store already uses for its variants. Run with no argument, the executable still runs every row. A malformed --part throws.

Tested

Linux (CachyOS), GCC 16, Release.

• CI's lavapipe with /dev/udmabuf denied, ctest --parallel 4 as on CI: the six parts pass in 10.3–13.3 s each. The single test took 59–150+ s before.
• RADV: the six parts together check all 126 rows, the same as the single test.
• --part 6/6 exits with 1.
• tools/check_test_registration.py reports nothing built and not run, and tools/check_conventions.py 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 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]

Merged: in this PR's full build (run 38103030261) all six agc_driver_flat_apertures_* parts pass on Linux. The only red test is agc_unit_shadow, which main has failed since #2864; it's reported in #3215.

[Silverash-i8]

Windows run, since the PR has no Intel or Windows results yet: Windows 11, WinLibs GCC 15.2.0 r7, Release, merged onto main d31b57a in one build with #3217, #3202, #3167, #3137, #2863, #2849 and #2847, on an RTX 5060 Laptop and the Arrow Lake iGPU (ANYPS5_GPU=Intel).

• All six agc_driver_flat_apertures_0 to _5 pass on both GPUs (about 7 s each on the RTX 5060, under 1 s each on the iGPU).

Full ctest of that build on both GPUs has no failure that isn't also on plain main here (guest_filesystem needs the symlink privilege, #3169; the Intel ones are listed on #2114 and #3207).

AI-assisted (Claude Code); I ran this on my machine.