[SandstromPL]

What

ANYPS5_GPU is a case-insensitive substring of the device name: that is how the driver matches it (VulkanDevice.cpp), and how docs/dev/ENVIRONMENT.md and docs/user/USAGE.md describe it. Two fixtures that open their own Vulkan device match it case-sensitively instead:

• PooledBufferRange uses std::strstr, from test(agc): honor GPU selection in pooled buffer checks #7104. With a request it matches nothing, it keeps the first device and then refuses it: selected device does not match ANYPS5_GPU.
• NativeMultisampleDepth uses std::string::find. With a request it matches nothing, it finds no device: no Vulkan 1.1 device.

So a request the driver accepts, such as ANYPS5_GPU=nvidia on an NVIDIA GPU, makes both report a skip, and ctest still prints 100% tests passed. Under ANYPS5_REQUIRE_VULKAN they fail instead.

Both now pick their device with #9253's SelectFixturePhysicalDevice, which follows the driver's rule. This depends on #9253, which adds that helper; until #9253 merges, its commit shows here too.

On machines with more than one device, this also changes two choices:

• Without ANYPS5_GPU, PooledBufferRange now ranks devices by type (discrete, integrated, virtual, then the rest) and needs Vulkan 1.1. Before, it took the first discrete device, or else the first device.
• With a request, NativeMultisampleDepth now takes the first matching device instead of the best-ranked match. This is what the driver does, and since fix(agc): retire an unwritten native multisample depth surface on guest access #6727 the fixture also uses the driver's own device, so the two now pick the same GPU.
• A new case in the selection CPU test covers the second point: a request that matches an integrated GPU and a later discrete one takes the integrated one. With the old rank-among-matches rule, this check fails.

As in #9253, the helper does not skip a device that lacks a graphics and compute queue, as the driver does. Neither fixture checked queues before.

The other four fixtures that select their own device (Recorder, PerformanceOverlay, SampledCacheBudget, BdaDevice) already match ignoring case, so this PR leaves them as they are.

Tested

Linux (Ubuntu 24.04, GCC 13.3), NVIDIA GeForce RTX 3050 Ti Laptop GPU (driver 580.178.04). The main runs are on main 0db47b2. The PR runs are on the same main with #9253 and this PR merged (a clean merge). All runs use ANYPS5_REQUIRE_VULKAN=1 and ANYPS5_NO_SHADER_CACHE=1.

The 14 tests are:

• agc_driver_pooled_buffer_range;
• the four NativeMultisampleDepth tests (agc_driver_native_multisample_depth, agc_driver_native_depth_clears, agc_driver_native_depth_clear_refusals, agc_driver_native_depth_retirement);
• test(agc): honor ANYPS5_GPU in the fixtures that open their own Vulkan device #9253's nine.

ANYPS5_GPU	main	this PR
NVIDIA	14 pass	14 pass
nvidia, geforce rtx	5 fail: agc_driver_pooled_buffer_range (selected device does not match ANYPS5_GPU) and the four NativeMultisampleDepth tests (no Vulkan 1.1 device); without ANYPS5_REQUIRE_VULKAN the same 5 are skipped	14 pass
no such device 1d3d154f	the same 5 fail; #9253's nine pass, since they ignore the request on main	14 fail with no Vulkan 1.1 device whose name contains ANYPS5_GPU="no such device 1d3d154f"; devices: NVIDIA GeForce RTX 3050 Ti Laptop GPU

• With VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation (ctest -V), the 14 pass on main and with this PR. The layer loads in both logs.
  • Both logs have the same five VUID-VkDeviceCreateInfo-pNext-pNext lines, all in the four NativeMultisampleDepth tests and already on main.
• agc_driver_fixture_device_selection passes, with the new case.

This machine has one GPU, so the choices between several devices are covered only by the CPU test.

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: test(agc): honor ANYPS5_GPU in the fixtures that open their own Vulkan device #9253
• Fixes: none
• Unblocks: none
• GPUs tested: NVIDIA GeForce RTX 3050 Ti Laptop GPU (driver 580.178.04), Ubuntu 24.04
• AI-assisted: yes (Claude Code)

🤖 Generated with Claude Code

[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 6300	core/libs/prx/libSceAgcDriver/tests/execution/NativeMultisampleDepth.cpp
PR 6808	core/libs/prx/libSceAgcDriver/tests/execution/NativeMultisampleDepth.cpp