#thread #executor


A single-threaded executor with some tricks up its sleeve

11 releases

0.1.0-alpha.11 Apr 23, 2021
0.1.0-alpha.9 Apr 12, 2021
0.1.0-alpha.3 Apr 7, 2021

#176 in Concurrency

Download history 4/week @ 2021-06-30 3/week @ 2021-07-07 46/week @ 2021-07-14 2/week @ 2021-07-21 24/week @ 2021-07-28 16/week @ 2021-08-04 23/week @ 2021-08-11 2/week @ 2021-08-18 3/week @ 2021-08-25 2/week @ 2021-09-08 4/week @ 2021-09-15 11/week @ 2021-09-22 1/week @ 2021-09-29 1/week @ 2021-10-06 5/week @ 2021-10-13

78 downloads per month
Used in audio-device


671 lines


Documentation Crates Actions Status

A single-threaded executor with some tricks up its sleeve.

This was primarily written for use in audio as a low-latency way of interacting with a single background thread for audio-related purposes, but is otherwise a general purpose library that can be used to do anything.

Soundness Warning: This crate uses a fair bit of unsafe. Some of the tricks employed needs to be rigirously sanity checked for safety before you can rely on this for production uses.

The default way to access the underlying thread is through the submit method. This blocks the current thread for the duration of the task allowing the background thread to access variables which are in scope. Like n below.

let thread = ste::spawn();

let mut n = 10;
thread.submit(|| n += 10)?;
assert_eq!(20, n);


Restricting thread access using tags

This library provides the ability to construct a Tag which is uniquely associated with the thread that created it. This can then be used to ensure that data is only accessible by one thread.

This is useful, because many APIs requires thread-locality where instances can only safely be used by the thread that created them. This is a low-level tool we provide which allows the safe implementation of Send for types which are otherwise !Send.

Note that correctly using a Tag is hard, and incorrect use has severe safety implications. Make sure to study its documentation closely before use.

struct Foo {
    tag: ste::Tag,

impl Foo {
    fn new() -> Self {
        Self {
            tag: ste::Tag::current_thread(),

    fn say_hello(&self) {
        println!("Hello World!");

let thread = ste::spawn();

let foo = thread.submit(|| Foo::new())?;

thread.submit(|| {
    foo.say_hello(); // <- OK!


Using say_hello outside of the thread that created it is not fine and will panic to prevent racy access:

let thread = ste::spawn();

let foo = thread.submit(|| Foo::new())?;

foo.say_hello(); // <- Oops, panics!


Known unsafety and soundness issues

Below you can find a list of unsafe use and known soundness issues this library currently has. The soundness issues must be fixed before this library goes out of alpha.

Pointers to stack-local addresses

In order to efficiently share data between a thread calling submit and the background thread, the background thread references a fair bit of stack-local data from the calling thread which involves a fair bit of unsafe.

While it should be possible to make this use safe (as is the hope of this library), it carries a risk that if the background thread were to proceed executing a task that is no longer synchronized properly with a caller of submit it might end up referencing data which is either no longer valid (use after free), or contains something else (dirty).

Tag re-use

Tag containers currently use a tag based on the address of a slab of allocated memory that is associated with each Thread. If however a Thread is shut down, and a new later recreated, there is a slight risk that this might re-use an existing memory address.

Memory addresses are quite thankful to use, because they're cheap and quite easy to access. Due to this it might however be desirable to use a generated ID per thread instead which can for example abort a program in case it can't guarantee uniqueness.

License: MIT/Apache-2.0


~30K SLoC