Crates.io | lagoon |
lib.rs | lagoon |
version | 0.1.3 |
source | src |
created_at | 2021-07-08 11:12:52.541622 |
updated_at | 2021-07-14 10:04:46.655198 |
description | A thread pool crate with an array of features |
homepage | |
repository | https://github.com/zesterer/lagoon |
max_upload_size | |
id | 420274 |
size | 114,262 |
Lagoon is a thread pool crate that aims to address many of the problems with existing thread pool crates.
Lagoon's scoped jobs can be used for simple rayon-like data parallelism.
// A vector of the numbers 0 to 99
let mut data = (0..100).collect::<Vec<u32>>();
lagoon::ThreadPool::default().scoped(|s| {
// For each element in the vector...
for x in data.iter_mut() {
// ...spawn a job that squares that element
s.run(move || *x *= *x);
}
});
// Demonstrate that the elements have indeed been squared
assert!((0..100)
.map(|x| x * x)
.zip(data.into_iter())
.all(|(x, y)| x == y));
Lagoon has very competitive performance. Below are timings required for each thread pool crate to spawn a new pool, execute 100,000 trivial jobs, and then finish executing (i.e: smaller is better) compared to common alternative crates.
Spawning 100000 trivial tasks/lagoon time: [15.124 ms 16.437 ms 17.871 ms]
Spawning 100000 trivial tasks/threadpool time: [59.108 ms 59.549 ms 59.989 ms]
Spawning 100000 trivial tasks/uvth time: [11.494 ms 12.598 ms 13.750 ms]
Spawning 100000 trivial tasks/rusty_pool time: [40.203 ms 44.778 ms 49.612 ms]
Benchmarks were run on an AMD Ryzen 7 3700x with 16 threads.
Lagoon is licensed under the MIT license (see LICENSE
) in the main repository.