Struct HardwareProfile 

pub struct HardwareProfile {
    pub peak_gflops: f64,
    pub memory_bandwidth_gbs: f64,
    pub name: String,
}

Hardware profile for roofline analysis.

Fields

peak_gflops: f64

Peak throughput in GFLOPS.

memory_bandwidth_gbs: f64

Memory bandwidth in GB/s.

name: String

Human-readable profile name.

Implementations

impl HardwareProfile

pub fn ridge_point(&self) -> f64

Ridge point: arithmetic intensity where compute and memory are balanced.

pub fn predict_latency_secs(&self, cost: &OpCost) -> f64

Predict execution latency in seconds using the roofline model.

latency = max(flops / peak_flops_per_sec, bytes / bandwidth_per_sec)

pub fn bottleneck(&self, cost: &OpCost) -> Bottleneck

Determine the bottleneck for a given operation cost.

Trait Implementations

impl Clone for HardwareProfile

fn clone(&self) -> HardwareProfile

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for HardwareProfile

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

Formats the value using the given formatter.