test(agc): run agc_driver_bda_device_tests under ctest

[johnmaia]

What

agc_driver_bda_device_tests was EXCLUDE_FROM_ALL and never registered, so it ran nowhere (#22, #23 left it out pending #14 and #20). It now runs under ctest with the other AGC driver tests.

• The test picked devices.front(). It now ranks devices as VulkanDevice does (discrete, integrated, virtual, then the rest; Vulkan 1.1 or newer), so on a host with an iGPU it runs on the same GPU as the runtime.
• Without a usable Vulkan device it exits 77 (skipped), as OpenVulkanTestDevice does, unless ANYPS5_REQUIRE_VULKAN is set. The Windows job has no Vulkan driver.
• On a CPU device (lavapipe) the BDA execution tests are not run and the test says so. The contract, allocation and color transfer tests still run there.

The BDA execution tests segfault on lavapipe on main too: the third dispatch, a 32-bit read at guest + 1 that spans both ranges, faults in llvmpipe's JIT code on vpbroadcastd (%rcx) with rcx = 0. The probe returns 0 for that read, and the only load through the probe result sits behind OpBranchConditional mapped && aligned. llvmpipe executes it anyway. With APS5_BDA_BYTE_READS=1, which removes that branch, all 22 dispatches pass on lavapipe.

A 20-line GLSL shader reproduces it outside the project (if (address != 0) value = Word(address).value; with address = 0): it crashes lavapipe at every GALLIUM_OVERRIDE_CPU_CAPS level (sse2, sse4_2, avx, avx2, native on an Alder Lake CPU; Mesa 25.2.8, LLVM 20.1.2), and returns the right value on RADV and ANV. Reported upstream as Mesa #16515 (same AnyPS5 shader), same class as #16217.

agc_driver_recorder_tests has the same devices.front() and is also unregistered, but fails on main on every device (an unchanged draw input was copied again); that is a separate PR.

Tested

Linux, Ubuntu 24.04.5, GCC 13.3, Mesa 25.2.8; i3-12100 with RX 7600 (RADV) and UHD 730 (ANV).

• ctest (Release, BUILD_TESTING=ON, ANYPS5_REQUIRE_VULKAN=1): 400/400 on the RX 7600; 400/400 on lavapipe alone with the CI settings (GALLIUM_OVERRIDE_CPU_CAPS=avx, --repeat until-fail:3), 21 skipped as before.
• agc_driver_bda_device_tests alone:

Vulkan devices	selected	result
all, iGPU enumerated first	RX 7600	passes
all, default order	RX 7600	passes
UHD 730 + lavapipe	UHD 730	passes
lavapipe only	lavapipe	passes, BDA execution not tested
none	-	exit 77 (skipped)
none, ANYPS5_REQUIRE_VULKAN=1	-	exit 1

• Windows: not tested locally; the change on that job is the skip above.
• No title was run.

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

🤖 Generated with Claude Code

[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
#1298	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526
#1309	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526
#1311	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526
#1312	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526
#1313	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526
#1316	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526
#1318	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526
#1341	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526
#1343	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526
#1372	core/libs/prx/libSceAgcDriver/CMakeLists.txt L526