Struct RunMetadata 

pub struct RunMetadata {
    pub backend: ResolvedBackend,
    pub engine: Option<Engine>,
    pub exactness: Exactness,
    pub placement: Placement,
    pub shots: Option<usize>,
}

How a result was produced, attached to every Simulate result type.

A GPU-attached run below the device crossover reports Placement::Host, and automatic dispatch onto an approximate backend reports Exactness::Approximate.

Fields

backend: ResolvedBackend

engine: Option<Engine>

Which sampler ran, when backend is a label several share. None when the backend is the whole answer, which is every backend other than ResolvedBackend::CompiledStabilizer.

exactness: Exactness

placement: Placement

shots: Option<usize>

Shots drawn, for a result estimated by sampling. None for an analytic result.

Implementations

impl RunMetadata

pub fn is_exact(&self) -> bool

pub fn fidelity_lower_bound(&self) -> Option<f64>

Reported fidelity floor for the produced state, None when the result is exact or the engine reports none.

For the MPS backend this is 1 minus the summed per-SVD relative discarded weights: a first-order truncation estimate, not a certificate. Errors compound across SVDs, and summing the weights understates the compounded error, since the strict bound on the infidelity is the square of the summed square roots. The two agree only when a single SVD truncates. The sparse backend at a raised pruning threshold reports 1 minus the absolute squared weight it dropped, the same first-order estimate; it renormalizes what it keeps, so the dropped weight is the whole of the error it reports. The budgeted Pauli engines leave this None and report their additive observable bound on the result types instead.

Trait Implementations

impl Clone for RunMetadata

fn clone(&self) -> RunMetadata

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for RunMetadata

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.