12 stable releases (3 major)

4.0.1 Dec 29, 2020
4.0.0 Jul 8, 2020
3.0.3 Jul 8, 2020
3.0.2 Jun 18, 2020
1.1.2 Nov 27, 2019

#16 in Profiling

Download history 11/week @ 2021-04-01 19/week @ 2021-04-08 125/week @ 2021-04-15 102/week @ 2021-04-22 25/week @ 2021-04-29 45/week @ 2021-05-06 53/week @ 2021-05-13 20/week @ 2021-05-20 24/week @ 2021-05-27 7/week @ 2021-06-03 29/week @ 2021-06-10 110/week @ 2021-06-17 147/week @ 2021-06-24 159/week @ 2021-07-01 102/week @ 2021-07-08 118/week @ 2021-07-15

233 downloads per month
Used in rstats

Apache-2.0

22KB
248 lines

Devtimer Build Status Crates.io Crates.io Crates.io Crates.io

The compact yet complete benchmarking suite for Rust. Period.

Rationale

I've seen many, many benchmarking tools. However, no one realizes that we need simplicity to simplify development and increase productivity. devtimer provides a very compact yet complete benchmarking suite for code written in Rust. It makes use of the standard library only to provide benchmark operations. You can either use it for benchmarking a single operation or you can use it for running an operation multiple times and finding the min, max and average execution times. Since this crate has no external dependencies, it is small, fast and does exactly what it claims to. Happy benchmarking!

Need help migrating from an older version? See the changelog.

Usage

Add this to your cargo.toml :

devtimer = "*"

Then add this line to your source file (i.e main.rs or lib.rs or where you need to use it):

use devtimer::DevTime;

Example usage

Simple usage

Let's say there are two functions called very_long_operation() and another_op() that take a very long time to execute. Then we can time it's execution as shown below:

fn main() {
    let mut timer = DevTime::new_simple();
    timer.start();
    very_long_operation();
    timer.stop();
    println!("The operation took: {} ns", timer.time_in_nanos().unwrap());
    // You can keep re-using the timer for other operations
    timer.start(); // this resets the timer and starts it again
    another_op();
    timer.stop();
    println!("The operation took: {} secs", timer.time_in_secs().unwrap());
    println!("The operation took: {} milliseconds", timer.time_in_millis().unwrap());
    println!("The operation took: {} microseconds", timer.time_in_micros().unwrap());
    println!("The operation took: {} ns", timer.time_in_nanos().unwrap());

    // With version 1.1.0 and upwards
    timer.start_after(&std::time::Duration::from_secs(2));
    // The timer will start after two seconds
    // Do some huge operation now
    timer.stop();
    println!("The operation took: {} nanoseconds", devtimer.time_in_nanos().unwrap());
}

Example: Benchmarking

use devtimer::run_benchmark;
fn main() {
  // We will simulate a long operation by std::thread::sleep()
  // Run 10 iterations for the test
  let bench_result = run_benchmark(10, |_| {
    // Fake a long running operation
    std::thread::sleep(std::time::Duration::from_secs(1);
  });
  bench_result.print_stats();
}

Advanced Benchmarking

The run_benchmark() function also provides a usize that can be used, say if you want to get something from an array to do the test. For example:

run_benchmark(100, |n| {
  do_action(data_source[n]);
});

Example: Tagged timers

use devtimer::DevTime;
fn main() {
  let mut cmplx = DevTime::new_complex();
  // Create a timer with tag `timer-1` 
  cmplx.create_timer("timer-1").unwrap();
  cmplx.start_timer("timer-1").unwrap();
  // Simulate a slow operation
  std::thread::sleep(std::time::Duration::from_secs(1));
  cmplx.stop_timer("timer-1").unwrap();
  
  // Create a timer with tag `cool-timer` 
  cmplx.create_timer("cool-timer").unwrap();
  cmplx.start_timer("cool-timer").unwrap();
  // Simulate a slow operation
  std::thread::sleep(std::time::Duration::from_secs(2));
  cmplx.stop_timer("cool-timer").unwrap();

  // We can output a benchmark in this way
  println!(" `cool-timer` took: {}", cmplx.time_in_micros("cool-timer").unwrap());

  // Or we can iterate through all timers
  for (tname, timer) in cmplx.iter() {
    println!("{} - {} ns", tname, timer.time_in_micros().unwrap());
  }

  // Or we can print results in the default '{timername} - {time} ns' format
  cmplx.print_stats();
}

Timing functions available (names are self explanatory):

  • time_in_secs() -> Returns the number of seconds the operation took
  • time_in_millis() -> Returns the number of milliseconds the operation took
  • time_in_micros() -> Returns the number of microseconds the operation took
  • time_in_nanos() -> Return the number of nanoseconds the operation took

See the full docs here.

License

This project is licensed under the Apache-2.0 License. Keep coding and benchmarking!

No runtime deps