Struct WgpuMegakernelDispatcher 

pub struct WgpuMegakernelDispatcher<'a> { /* private fields */ }

Runtime wrapper for persistent megakernel dispatch.

Implementations

impl<'a> WgpuMegakernelDispatcher<'a>

pub fn new(backend: &'a dyn VyreBackend) -> Self

Create a new dispatcher.

pub fn dispatch_megakernel_bytes( &self, work_queue_bytes: &[u8], config: &MegakernelConfig, ) -> Result<MegakernelReport, BackendError>

Decode a raw little-endian MegakernelWorkItem queue and launch the megakernel.

Errors

Returns a backend error when work_queue_bytes is not exactly aligned to MegakernelWorkItem records or when backend dispatch fails.

pub fn dispatch_megakernel( &self, work_items: &[MegakernelWorkItem], config: &MegakernelConfig, ) -> Result<MegakernelReport, BackendError>

Launch the megakernel.

pub fn dispatch_megakernel_with_io_queue( &self, work_items: &[MegakernelWorkItem], config: &MegakernelConfig, io_queue_bytes: Vec<u8>, ) -> Result<MegakernelReport, BackendError>

Launch the megakernel with a caller-supplied IO queue.

The queue is validated against the megakernel ABI before any backend work starts, so malformed queue views fail before compilation or GPU submission.

Trait Implementations

impl MegakernelDispatch for WgpuMegakernelDispatcher<'_>

fn dispatch_megakernel( &self, work_queue: &[MegakernelWorkItem], config: &MegakernelConfig, ) -> Result<MegakernelReport, BackendError>

Drain the requested megakernel dispatch.