Struct StepResourceLease 

pub struct StepResourceLease<R>
where
    R: DeviceRuntime,
{ /* private fields */ }

Resources whose lifetime is one exact continuous-batch execution frame. Child invocation leases retain this scope through Arc, so shared frame capacity and every participant authority outlive asynchronous device work.

Implementations

impl<R> StepResourceLease<R>

where R: DeviceRuntime,

pub const fn batch_step_id(&self) -> BatchStepId

pub fn execution_lane(&self) -> &Arc<ExecutionLane<R>>

pub fn reusable_execution_bucket(&self) -> Option<&ReusableExecutionBucketSpec>

pub fn try_retire_normal( self: Arc<Self>, ) -> Result<StepRetirementReceipt, StepFinalizationFailure<R>>

pub fn try_abort( self: Arc<Self>, ) -> Result<StepRetirementReceipt, StepFinalizationFailure<R>>

pub fn try_rollback_unsubmitted( self: Arc<Self>, ) -> Result<StepRetirementReceipt, StepFinalizationFailure<R>>

Releases a capacity-deferred step before any invocation was prepared or submitted. Unlike try_abort, this leaves every participant session open so the scheduler can retry after the exact capacity source advances.

pub fn participant_count(&self) -> u32

pub fn coordinator_id(&self) -> LogicalAdmissionCoordinatorId

pub fn participants( &self, ) -> impl ExactSizeIterator<Item = &Arc<AdmittedSequenceResources<R>>>

pub fn participant_frames( &self, ) -> impl ExactSizeIterator<Item = StepParticipantFrameAssignment> + '_

pub fn backing_slices(&self) -> &[LogicalBackingSliceAuthority]

pub fn logical_capacity(&self) -> Option<&LogicalBatchCapacityLease>

pub fn work_shape(&self) -> &BatchWorkShape

pub fn bind_invocation_work_shape( &self, participant_tokens: Vec<(BatchParticipantAuthority, TokenSpanWork)>, ) -> Result<BatchWorkShape, VNextError>

pub fn bind_all_invocation_work_shape( &self, token_spans: Vec<TokenSpanWork>, ) -> Result<BatchWorkShape, VNextError>

pub fn shared_all_invocation_work_shape( &self, token_spans: &[TokenSpanWork], ) -> Result<Arc<BatchWorkShape>, VNextError>

Reuses the exact immutable work authority admitted for this step when a submission wave binds the same participants and token spans.

pub fn claimed_backing(&self) -> &ClaimedBackingTransaction

pub fn static_provisioning(&self) -> Option<&StaticProvisioningLease<R>>

pub fn plan_evidence(&self) -> TrustedPlanRuntimeEvidence

pub fn try_admit_invocation( self: &Arc<Self>, request: InvocationResourceAdmissionRequest, ) -> Result<InvocationResourceAdmissionDecision<R>, VNextError>

pub fn try_prepare_submission_wave( self: &Arc<Self>, requests: Vec<InvocationResourceAdmissionRequest>, ) -> Result<StepSubmissionWaveAdmissionDecision<R>, VNextError>

pub fn try_prepare_determinism_submission_wave( self: &Arc<Self>, requests: Vec<InvocationResourceAdmissionRequest>, ) -> Result<StepSubmissionWaveAdmissionDecision<R>, VNextError>

Prepares a canonical immutable-plan node subset that can only be submitted through the exact determinism restore/readback path.

pub fn try_prepare_full_plan_submission_wave( self: &Arc<Self>, work_shape: Arc<BatchWorkShape>, fit_policy: AdmissionFitPolicy, pressure_action: AdmissionPressureAction, ) -> Result<StepSubmissionWaveAdmissionDecision<R>, VNextError>

Prepares the only production wave topology: every immutable-plan node in order over one shared participant/work authority. Callers do not rebuild a per-node request vector for topology already committed by the plan hash.

Trait Implementations

impl<R> Drop for StepResourceLease<R>

where R: DeviceRuntime,

fn drop(&mut self)

Executes the destructor for this type.

fn pin_drop(self: Pin<&mut Self>)

🔬This is a nightly-only experimental API. (pin_ergonomics)

Execute the destructor for this type, but different to Drop::drop, it requires self to be pinned.