Enum CostCalibration 

pub enum CostCalibration {
    Measured,
    ArchDefault,
    Uncalibrated,
}

Where a cost model’s numbers actually came from.

The distinction is load-bearing and rlx used to erase it. RocmCostModel falls back to sgemm_gflops: 10_000.0 when ROCm is absent; MetalHwModel falls back to AppleGpuFamily::Unknown => (300e9, 180e9, 60e9). Both then feed fastest_device_for, which ranks devices against each other — so on an unrecognized GPU rlx would confidently place a graph using invented throughput, and say nothing.

CAKE B.5 states the rule this implements: “Timing-model coverage is separately evidence-gated: B200 is the measured baseline, H100 is calibrated independently, and other targets report a coverage limitation instead of inheriting estimates.” Reporting a coverage limitation is strictly more useful than a confident guess, because a caller can act on it.

Variants

Measured

Measured on this machine (a calibration cache exists, or the model probed the device). Comparisons across such models are meaningful.

ArchDefault

Compile-time constants chosen for this specific detected architecture. Reasonable, but not measured here — treat cross-device comparisons as indicative rather than authoritative.

Uncalibrated

The architecture was not recognized, so the numbers are a generic guess. Not a basis for ranking one device against another.

Implementations

impl CostCalibration

pub fn is_rankable(&self) -> bool

True when this model’s numbers may be compared against another device’s.

pub fn as_str(&self) -> &'static str

Trait Implementations

impl Clone for CostCalibration

fn clone(&self) -> CostCalibration

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for CostCalibration

impl Debug for CostCalibration

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

Formats the value using the given formatter.

impl Eq for CostCalibration

impl Ord for CostCalibration

fn cmp(&self, other: &CostCalibration) -> Ordering

This method returns an Ordering between self and other.

fn max(self, other: Self) -> Self

where Self: Sized,

Compares and returns the maximum of two values.

fn min(self, other: Self) -> Self

where Self: Sized,

Compares and returns the minimum of two values.

fn clamp(self, min: Self, max: Self) -> Self

where Self: Sized,

Restrict a value to a certain interval.

fn clamp_to<R>(self, range: R) -> Self

where Self: Sized, R: ClampBounds<Self>,

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

Restrict a value to a certain range.

impl PartialEq for CostCalibration

fn eq(&self, other: &CostCalibration) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl PartialOrd for CostCalibration

fn partial_cmp(&self, other: &CostCalibration) -> Option<Ordering>

This method returns an ordering between self and other values if one exists.

fn lt(&self, other: &Rhs) -> bool

Tests less than (for self and other) and is used by the < operator.

fn le(&self, other: &Rhs) -> bool

Tests less than or equal to (for self and other) and is used by the <= operator.

fn gt(&self, other: &Rhs) -> bool

Tests greater than (for self and other) and is used by the > operator.

fn ge(&self, other: &Rhs) -> bool

Tests greater than or equal to (for self and other) and is used by the >= operator.

impl StructuralPartialEq for CostCalibration