listGpuDevices abstract method
Lists the GPU-class devices the backend exposes.
With an empty probeBackends only already-registered backends are
inspected — no backend module is loaded, so an unsupported GPU runtime
cannot crash the process during enumeration. Pass specific backends in
probeBackends to opt into loading just those modules (each guarded)
before enumerating. Backends with no offload devices (and web/WebGPU)
return an empty list.
Implementation
Future<List<GpuDeviceInfo>> listGpuDevices({
List<GpuBackend> probeBackends = const [],
});