Trait Benchmark

pub trait Benchmark {
    type Args: Clone;

    // Required methods
    fn prepare(&self) -> Self::Args;
    fn execute(&self, args: Self::Args);
    fn name(&self) -> String;
    fn sync(&self);

    // Provided methods
    fn num_samples(&self) -> usize { ... }
    fn options(&self) -> Option<String> { ... }
    fn shapes(&self) -> Vec<Vec<usize>> { ... }
    fn sync_elapsed(&self) -> Result<Duration, TimestampsError> { ... }
    fn run(&self, timing_method: TimingMethod) -> BenchmarkDurations { ... }
    fn run_one_full(&self, args: Self::Args) -> Duration { ... }
    fn run_one_device_only(&self, args: Self::Args) -> Duration { ... }
}

Benchmark trait.

Required Associated Types

type Args: Clone

Benchmark arguments.

Required Methods

fn prepare(&self) -> Self::Args

Prepare the benchmark, run anything that is essential for the benchmark, but shouldn’t count as included in the duration.

Notes

This should not include warmup, the benchmark will be run at least one time without measuring the execution time.

fn execute(&self, args: Self::Args)

Execute the benchmark and returns the time it took to complete.

fn name(&self) -> String

Name of the benchmark, should be short and it should match the name defined in the crate Cargo.toml

fn sync(&self)

Wait for computation to complete.

Provided Methods

fn num_samples(&self) -> usize

Number of samples per run required to have a statistical significance.

fn options(&self) -> Option<String>

The options passed to the benchmark.

fn shapes(&self) -> Vec<Vec<usize>>

Shapes dimensions

fn sync_elapsed(&self) -> Result<Duration, TimestampsError>

Wait for computation to complete and return hardware reported computation duration.

fn run(&self, timing_method: TimingMethod) -> BenchmarkDurations

Run the benchmark a number of times.

fn run_one_full(&self, args: Self::Args) -> Duration

Collect one sample directly measuring the full execute + sync step.

fn run_one_device_only(&self, args: Self::Args) -> Duration

Collect one sample using timing measurements reported by the device.

Implementors