Struct TieredBackend 

pub struct TieredBackend { /* private fields */ }

A SimBackend that interprets immediately and promotes to generated code as soon as background compilation completes.

Implementations

impl TieredBackend

pub fn new(laid_out: &LaidOutProgram, options: &SimulatorOptions) -> Self

Build a tiered simulation targeting this host’s default compiled tier: ready to run immediately on the interpreter while the compiled tier is prepared in the background.

The native tier observes cancellation (see TieredBackend::cancel_background_compilation); the Cranelift tier runs to completion once started.

pub fn is_compiled(&self) -> bool

Whether the compiled tier has been adopted.

pub fn execution_stats(&self) -> TieredExecutionStats

Return deterministic lifecycle and execution measurements.

pub fn start_execution_timing(&mut self)

Start measuring time to promotion for a benchmark workload.

pub fn finish_execution_timing(&mut self) -> Option<TieredExecutionTiming>

Stop timing and return the observed promotion interval.

pub fn promotion_error(&self) -> Option<&SimulatorError>

Why promotion has not happened yet, for diagnostics.

Returns None once running fully compiled or while background compilation is still in progress.

pub fn cancel_background_compilation(&mut self) -> bool

Request cancellation of background compilation.

The native-tier worker unwinds at its next task boundary and the simulation stays on the interpreter permanently; the reason becomes retrievable through TieredBackend::promotion_error. A result that already reached the channel (or arrives afterwards, for example from a tier that ignores the token) is rejected so cancellation is honored regardless of compiler timing. Returns whether a background compilation was still pending, so callers that only want to reclaim a finished worker can ignore the call cheaply.

Trait Implementations

impl Drop for TieredBackend

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.

impl SimBackend for TieredBackend

type Event = TieredEventRef

The event handle type produced by this backend.

fn eval_comb(&mut self) -> Result<(), SimulatorErrorCode>

fn eval_apply_ff_at( &mut self, event: TieredEventRef, ) -> Result<(), SimulatorErrorCode>

Evaluate and apply a flip-flop domain for the given event.

fn eval_comb_apply_ff_at( &mut self, event: TieredEventRef, ) -> Result<(), SimulatorErrorCode>

Evaluate combinational logic and then evaluate/apply one flip-flop domain. Backends may override this to compile the two phases as one function; the default preserves the same ordering with two calls.

fn eval_comb_apply_ff_many_at( &mut self, event: TieredEventRef, count: u64, ) -> (u64, Result<(), SimulatorErrorCode>)

Execute up to count identical fused ticks. The returned count is the number of iterations completed before a runtime event or error forced a return to the host. Backends without an in-function loop execute one iteration so the caller can preserve per-tick observation semantics.

fn eval_only_ff_at( &mut self, event: TieredEventRef, ) -> Result<(), SimulatorErrorCode>

Evaluate FF domain without applying (for cascaded clocks).

fn apply_ff_at( &mut self, event: TieredEventRef, ) -> Result<(), SimulatorErrorCode>

Apply (commit) an already-evaluated FF domain.

fn resolve_signal(&self, addr: &AbsoluteAddr) -> SignalRef

fn resolve_event(&self, addr: &AbsoluteAddr) -> TieredEventRef

fn resolve_event_opt(&self, addr: &AbsoluteAddr) -> Option<TieredEventRef>

fn resolve_eval_only_event(&self, addr: &AbsoluteAddr) -> Option<TieredEventRef>

fn resolve_apply_event(&self, addr: &AbsoluteAddr) -> Option<TieredEventRef>

fn set<T: Copy>(&mut self, signal: SignalRef, value: T)

fn set_wide(&mut self, signal: SignalRef, value: BigUint)

fn set_four_state(&mut self, signal: SignalRef, value: BigUint, mask: BigUint)

fn get(&self, signal: SignalRef) -> BigUint

fn get_as<T: Default + Copy>(&self, signal: SignalRef) -> T

fn get_four_state(&self, signal: SignalRef) -> (BigUint, BigUint)

fn memory_as_ptr(&self) -> (*const u8, usize)

fn memory_as_mut_ptr(&mut self) -> (*mut u8, usize)

Exposing a retained writable view must permanently disable sparse VCD tracking in backend-owned state outside the returned memory range.

fn memory_owner(&self) -> Option<Arc<dyn Any + Send + Sync>>

Return an opaque owner that keeps the memory allocation alive.

fn vcd_tracking_enabled(&self) -> bool

Opt into generated write notifications only while no raw mutable view has escaped. The default keeps other backend implementations correct.

fn runtime_event_buffer_as_ptr(&self) -> (*const u8, usize)

fn runtime_event_buffer(&self) -> Option<Arc<RuntimeEventBuffer>>

fn set_comb_capture_event_enabled(&mut self, active_sites: &[bool])

fn stable_region_size(&self) -> usize

fn layout(&self) -> &MemoryLayout

fn id_to_addr_slice(&self) -> &[AbsoluteAddr]

fn id_to_event_slice(&self) -> &[TieredEventRef]

fn num_events(&self) -> usize

fn clear_triggered_bits(&mut self)

fn mark_triggered_bit(&mut self, id: usize)

fn get_triggered_bits(&self) -> BitSet

fn take_vcd_activity(&mut self, groups: &mut Vec<usize>) -> bool

Consume waveform activity independently of clock trigger processing. Returns false when full scanning is required (including raw host views).

fn mark_vcd_signal(&mut self, signal: SignalRef)

Called by host setters, which share the generated-store notification ABI.