Function run_persistent_kernel 

pub fn run_persistent_kernel(
    backend: &WgpuBackend,
    program: &Program,
    config: &DispatchConfig,
    queue: PersistentQueue,
) -> Result<PersistentKernelReport, BackendError>

Compile a persistent pipeline and drain queue through the GPU.

The pipeline and bind-group layout are created once and reused for every work item. Only the input/output buffer contents change between launches. If the backend reports device loss before or during drainage, recovery is attempted and the pipeline is recompiled automatically.

Errors

Returns BackendError when queue validation, pipeline compilation, device recovery, or GPU dispatch fails.