Crates.io | mbench |
lib.rs | mbench |
version | 0.1.0 |
source | src |
created_at | 2022-10-09 16:30:07.850175 |
updated_at | 2022-10-09 16:30:07.850175 |
description | Simple and easy to use microbenchmark macro. Zero deps. |
homepage | |
repository | https://github.com/vglinka/mbench |
max_upload_size | |
id | 684078 |
size | 17,044 |
Measures the average time for a given number of iterations. Easy to use, zero deps.
Cargo.toml
[dependencies]
mbench = "0.1"
main.rs
#[macro_use]
extern crate mbench;
fn my_func(x: u32) { vec![x; u16::MAX as usize].sort(); }
fn main() {
let mut text = String::new();
microbench!(
text = "Hello world!".to_string();
my_func(42);
// ... code ....
);
println!("{text}");
}
Output
Elapsed / 10 = 3.70ms
Hello world!
The default is 10 iterations, but you can set an arbitrary number microbench!(4, {/* code */});
.
microbench!(1, {
text = "Hello world!".to_string();
my_func(42);
});
Elapsed / 1 = 3.34ms
Hello world
In addition, there is a simplified syntax for setting the number of iterations (1,5,10,50,100,500,1000):
microbench1!(/* code */);
microbench5!(/* code */);
microbench10!(/* code */);
microbench50!(/* code */);
microbench100!(/* code */);
microbench500!(/* code */);
microbench1000!(/* code */);
With macro fixedbench!([0..=2], 10, {/* code */});
you can set
the range in milliseconds (and optionally 10 iterations in this example).
If the code runs longer or faster than the interval specified,
then there will be a panic.
This can be used in tests. The test fails if the code becomes slower after the change. The default is 1 iteration.
#[macro_use]
extern crate mbench;
fn my_func(ms: u64) {
let ms = std::time::Duration::from_millis(ms);
std::thread::sleep(ms);
}
#[cfg(test)]
mod tests {
use super::my_func;
#[test]
fn simple() {
fixedbench!([..=2], {
my_func(3);
});
}
}
running 1 test
test tests::simple ... FAILED
failures:
---- tests::simple stdout ----
Elapsed / 1 = 3.36ms
thread 'tests::simple' panicked at 'expected [..=2]ms, finished in 3ms.', main.rs:15:9
Keep in mind that the running time of the code may vary on different machines. After completing work on a piece of code, such a test should be removed or commented out.