#thread-pool #latency #minimize #optimized #worker #job #async

fast-threadpool

Thread-pool implementation optimized to minimize latency

8 releases

0.3.0 Jan 8, 2021
0.2.3 Dec 17, 2020
0.2.1 Oct 27, 2020
0.1.2 Oct 25, 2020

#908 in Concurrency

31 downloads per month

MIT license

22KB
425 lines

Fast Thread-pool

This thread-pool implementation is optimized to minimize latency. In particular, you are guaranteed to never pay the cost of thread spawn before your task is executed. New threads are spawned only during the "useless time" of the workers (for example, after returning the result of a job).

The only possible case of latency is the lack of "available" workers. To minimize the probability of this case, this thread-pool constantly keeps a certain number of workers available (configurable).

This implementation allows you to wait for the result of a job asynchronously, so you can use it as a replacement for the spawn_blocking function of your async runtime.

Use

use fast_threadpool::{ThreadPool, ThreadPoolConfig};

let threadpool = ThreadPool::start(ThreadPoolConfig::default(), ()).into_sync_handler();

assert_eq!(4, threadpool.execute(|_| { 2 + 2 })?);

Use from async task

use fast_threadpool::{ThreadPool, ThreadPoolConfig};

let threadpool = ThreadPool::start(ThreadPoolConfig::default(), ()).into_async_handler();

assert_eq!(4, threadpool.execute(|_| { 2 + 2 }).await?);

Dependencies

~265–700KB
~13K SLoC