Struct BatchedOperationInvocation 

pub struct BatchedOperationInvocation<'a, B> { /* private fields */ }

Borrowed provider view for exactly one physical command. Participant-local resources remain separate projections while invocation/step/plan resources may be shared by every projection.

Implementations

impl<'a, B> BatchedOperationInvocation<'a, B>

pub fn batch_identity(&self) -> &BatchOperationIdentity

pub fn participants(&self) -> &[OperationInvocation<'a, B>]

pub fn operation(&self) -> &OperationDescriptor

pub fn node_id(&self) -> &NodeId

pub fn provider_id(&self) -> &ProviderId

pub fn work_shape(&self) -> &BatchWorkShape

pub fn work_contract(&self) -> &NodeWorkContract

pub fn program_binding(&self) -> Option<&ProgramBindingNodeBinding>

pub fn attach_binding_command<C>( &self, operation: EncodedDeviceOperation<C>, command: C, ) -> EncodedDeviceOperation<C>

Attaches the command that refreshes this provider’s invocation-scoped binding workspace. A compiled slot moves the command into the non-aliasing wave prelude; without that authority it remains adjacent to the eager compute command.

pub fn participant_token_ranges(&self) -> &[BatchParticipantTokenRange]

pub fn binding_uses_packed_batch_coordinates( &self, role: ResolvedValueRole, ordinal: u32, ) -> Result<bool, VNextError>

Returns whether one exact resolved binding uses the packed token coordinate space of this physical batch. This is deliberately distinct from physical sharing: Request backing can be shared by child sequences while still using source-token coordinates.