A PC whose GPUs were installed from different driver packages of one vendor
(an ASUS G513QY: the integrated Radeon on the 2021 package, the RX 6800M on a
2026 one) gives the Vulkan loader two versions of that vendor's driver, and
the RX 6800M is listed once through each. wgpu's request_adapter takes the
first discrete entry, the 2021 driver's, and the window drawn through it dies
inside amdvlk64.dll (0xc0000005) before it paints: the main thread is in the
2021 driver, the fault on a thread of the 2026 one, no Rust frame anywhere.
Left to the #820 ladder, two crashed launches end in DirectX 12 only.
filmcraft_gpu::adapter::choose is wgpu's own rule (the adapters that can draw
to the surface, by device type for WGPU_POWER_PREF, then as enumerated) with
one addition: when the adapter that rule picks is listed again by the same
driver (same backend, device type, PCI vendor and device and driver name; the
same bus id when both report one), the entry exposing the most wgpu features
is used, i.e. the newer version (68 against 60 here, a strict superset).
Without such a duplicate the choice is wgpu's own, and entries of one GPU from
different drivers (RADV and AMDVLK, a vendor's driver and Dozen) are left
alone.
graphics::apply installs it as egui-wgpu's native_adapter_selector in the
default mode on Windows, as #609 was asked to be rebased onto the #827
start-up. DirectX 12 only, software mode and the other platforms keep their
adapter choice, and the ladder is untouched. The adapters and the choice go
to the log.
Tests: the laptop's listing (wgpu's rule picks the entry that crashed, choose
the other one), listings without two versions of one driver over every power
preference (choose equals wgpu's rule), surface, bus id and driver cases, and
one against the adapters of the machine running the tests (the rule is what
request_adapter returns there).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
#827 rewrote apps/filmcraft/src/graphics.rs and the start-up in main.rs, so
the selector of this branch no longer applies. main's tree is taken whole
(graphics.rs, main.rs and the README note about WGPU_ADAPTER_NAME included);
the next commit hooks the selector into the GraphicsStart flow.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A PC whose GPUs were installed from different driver packages (an ASUS G513QY:
the integrated Radeon on the 2021 package, the RX 6800M on the 2026 one, each with
its own Vulkan ICD) lists the RX 6800M twice. wgpu's default takes the first
discrete adapter, the stale ICD, and drawing through it dies with 0xc0000005
inside amdvlk64.dll before any Rust frame can catch it.
apps/filmcraft/src/graphics.rs installs an egui-wgpu native_adapter_selector that
keeps wgpu's ranking (device type for WGPU_POWER_PREF, then enumeration order)
between different GPUs but, among entries that are one physical GPU (same backend,
vendor and device id; bus id when both drivers report one), prefers the entry
exposing the most wgpu features, i.e. the newer driver. Adapters that cannot draw
to the window are skipped. With one driver per GPU the rule has nothing to merge
and the choice is wgpu's own. WGPU_ADAPTER_NAME (part of the name, backend or
driver text) picks an adapter by hand; candidates, duplicates and the choice are
logged.
The choice is a pure function over plain candidate data, with regression tests
built from this machine's vulkaninfo listing (wgpu's rule picked the crashing
entry 0; the selector picks entry 1) plus wgpu-order, surface, two-cards and
override cases. Verified on the affected laptop: the release build now opens,
answers the control channel, screenshots and quits (exit 0); it crashed on every
launch before.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
