3 //! ## The threading model
5 //! An executing Rust program consists of a collection of native OS threads,
6 //! each with their own stack and local state. Threads can be named, and
7 //! provide some built-in support for low-level synchronization.
9 //! Communication between threads can be done through
10 //! [channels], Rust's message-passing types, along with [other forms of thread
11 //! synchronization](../../std/sync/index.html) and shared-memory data
12 //! structures. In particular, types that are guaranteed to be
13 //! threadsafe are easily shared between threads using the
14 //! atomically-reference-counted container, [`Arc`].
16 //! Fatal logic errors in Rust cause *thread panic*, during which
17 //! a thread will unwind the stack, running destructors and freeing
18 //! owned resources. While not meant as a 'try/catch' mechanism, panics
19 //! in Rust can nonetheless be caught (unless compiling with `panic=abort`) with
20 //! [`catch_unwind`](../../std/panic/fn.catch_unwind.html) and recovered
21 //! from, or alternatively be resumed with
22 //! [`resume_unwind`](../../std/panic/fn.resume_unwind.html). If the panic
23 //! is not caught the thread will exit, but the panic may optionally be
24 //! detected from a different thread with [`join`]. If the main thread panics
25 //! without the panic being caught, the application will exit with a
26 //! non-zero exit code.
28 //! When the main thread of a Rust program terminates, the entire program shuts
29 //! down, even if other threads are still running. However, this module provides
30 //! convenient facilities for automatically waiting for the termination of a
31 //! thread (i.e., join).
33 //! ## Spawning a thread
35 //! A new thread can be spawned using the [`thread::spawn`][`spawn`] function:
40 //! thread::spawn(move || {
45 //! In this example, the spawned thread is "detached," which means that there is
46 //! no way for the program to learn when the spawned thread completes or otherwise
49 //! To learn when a thread completes, it is necessary to capture the [`JoinHandle`]
50 //! object that is returned by the call to [`spawn`], which provides
51 //! a `join` method that allows the caller to wait for the completion of the
57 //! let thread_join_handle = thread::spawn(move || {
61 //! let res = thread_join_handle.join();
64 //! The [`join`] method returns a [`thread::Result`] containing [`Ok`] of the final
65 //! value produced by the spawned thread, or [`Err`] of the value given to
66 //! a call to [`panic!`] if the thread panicked.
68 //! Note that there is no parent/child relationship between a thread that spawns a
69 //! new thread and the thread being spawned. In particular, the spawned thread may or
70 //! may not outlive the spawning thread, unless the spawning thread is the main thread.
72 //! ## Configuring threads
74 //! A new thread can be configured before it is spawned via the [`Builder`] type,
75 //! which currently allows you to set the name and stack size for the thread:
78 //! # #![allow(unused_must_use)]
81 //! thread::Builder::new().name("thread1".to_string()).spawn(move || {
82 //! println!("Hello, world!");
86 //! ## The `Thread` type
88 //! Threads are represented via the [`Thread`] type, which you can get in one of
91 //! * By spawning a new thread, e.g., using the [`thread::spawn`][`spawn`]
92 //! function, and calling [`thread`][`JoinHandle::thread`] on the [`JoinHandle`].
93 //! * By requesting the current thread, using the [`thread::current`] function.
95 //! The [`thread::current`] function is available even for threads not spawned
96 //! by the APIs of this module.
98 //! ## Thread-local storage
100 //! This module also provides an implementation of thread-local storage for Rust
101 //! programs. Thread-local storage is a method of storing data into a global
102 //! variable that each thread in the program will have its own copy of.
103 //! Threads do not share this data, so accesses do not need to be synchronized.
105 //! A thread-local key owns the value it contains and will destroy the value when the
106 //! thread exits. It is created with the [`thread_local!`] macro and can contain any
107 //! value that is `'static` (no borrowed pointers). It provides an accessor function,
108 //! [`with`], that yields a shared reference to the value to the specified
109 //! closure. Thread-local keys allow only shared access to values, as there would be no
110 //! way to guarantee uniqueness if mutable borrows were allowed. Most values
111 //! will want to make use of some form of **interior mutability** through the
112 //! [`Cell`] or [`RefCell`] types.
114 //! ## Naming threads
116 //! Threads are able to have associated names for identification purposes. By default, spawned
117 //! threads are unnamed. To specify a name for a thread, build the thread with [`Builder`] and pass
118 //! the desired thread name to [`Builder::name`]. To retrieve the thread name from within the
119 //! thread, use [`Thread::name`]. A couple examples of where the name of a thread gets used:
121 //! * If a panic occurs in a named thread, the thread name will be printed in the panic message.
122 //! * The thread name is provided to the OS where applicable (e.g., `pthread_setname_np` in
123 //! unix-like platforms).
127 //! The default stack size for spawned threads is 2 MiB, though this particular stack size is
128 //! subject to change in the future. There are two ways to manually specify the stack size for
131 //! * Build the thread with [`Builder`] and pass the desired stack size to [`Builder::stack_size`].
132 //! * Set the `RUST_MIN_STACK` environment variable to an integer representing the desired stack
133 //! size (in bytes). Note that setting [`Builder::stack_size`] will override this.
135 //! Note that the stack size of the main thread is *not* determined by Rust.
137 //! [channels]: crate::sync::mpsc
138 //! [`join`]: JoinHandle::join
139 //! [`Result`]: crate::result::Result
140 //! [`Ok`]: crate::result::Result::Ok
141 //! [`Err`]: crate::result::Result::Err
142 //! [`thread::current`]: current
143 //! [`thread::Result`]: Result
144 //! [`unpark`]: Thread::unpark
145 //! [`thread::park_timeout`]: park_timeout
146 //! [`Cell`]: crate::cell::Cell
147 //! [`RefCell`]: crate::cell::RefCell
148 //! [`with`]: LocalKey::with
149 //! [`thread_local!`]: crate::thread_local
151 #![stable(feature = "rust1", since = "1.0.0")]
152 #![deny(unsafe_op_in_unsafe_fn)]
154 #[cfg(all(test, not(target_os = "emscripten")))]
158 use crate::cell::UnsafeCell;
159 use crate::ffi::{CStr, CString};
162 use crate::marker::PhantomData;
164 use crate::num::NonZeroU64;
165 use crate::num::NonZeroUsize;
167 use crate::panicking;
169 use crate::ptr::addr_of_mut;
171 use crate::sync::Arc;
172 use crate::sys::thread as imp;
173 use crate::sys_common::thread;
174 use crate::sys_common::thread_info;
175 use crate::sys_common::thread_parker::Parker;
176 use crate::sys_common::{AsInner, IntoInner};
177 use crate::time::Duration;
179 ////////////////////////////////////////////////////////////////////////////////
180 // Thread-local storage
181 ////////////////////////////////////////////////////////////////////////////////
186 #[stable(feature = "scoped_threads", since = "1.63.0")]
189 #[stable(feature = "scoped_threads", since = "1.63.0")]
190 pub use scoped::{scope, Scope, ScopedJoinHandle};
192 #[stable(feature = "rust1", since = "1.0.0")]
193 pub use self::local::{AccessError, LocalKey};
195 // The types used by the thread_local! macro to access TLS keys. Note that there
196 // are two types, the "OS" type and the "fast" type. The OS thread local key
197 // type is accessed via platform-specific API calls and is slow, while the fast
198 // key type is accessed via code generated via LLVM, where TLS keys are set up
199 // by the elf linker. Note that the OS TLS type is always available: on macOS
200 // the standard library is compiled with support for older platform versions
201 // where fast TLS was not available; end-user code is compiled with fast TLS
202 // where available, but both are needed.
204 #[unstable(feature = "libstd_thread_internals", issue = "none")]
205 #[cfg(target_thread_local)]
207 pub use self::local::fast::Key as __FastLocalKeyInner;
208 #[unstable(feature = "libstd_thread_internals", issue = "none")]
210 pub use self::local::os::Key as __OsLocalKeyInner;
211 #[unstable(feature = "libstd_thread_internals", issue = "none")]
212 #[cfg(all(target_family = "wasm", not(target_feature = "atomics")))]
214 pub use self::local::statik::Key as __StaticLocalKeyInner;
216 ////////////////////////////////////////////////////////////////////////////////
218 ////////////////////////////////////////////////////////////////////////////////
220 /// Thread factory, which can be used in order to configure the properties of
223 /// Methods can be chained on it in order to configure it.
225 /// The two configurations available are:
227 /// - [`name`]: specifies an [associated name for the thread][naming-threads]
228 /// - [`stack_size`]: specifies the [desired stack size for the thread][stack-size]
230 /// The [`spawn`] method will take ownership of the builder and create an
231 /// [`io::Result`] to the thread handle with the given configuration.
233 /// The [`thread::spawn`] free function uses a `Builder` with default
234 /// configuration and [`unwrap`]s its return value.
236 /// You may want to use [`spawn`] instead of [`thread::spawn`], when you want
237 /// to recover from a failure to launch a thread, indeed the free function will
238 /// panic where the `Builder` method will return a [`io::Result`].
245 /// let builder = thread::Builder::new();
247 /// let handler = builder.spawn(|| {
251 /// handler.join().unwrap();
254 /// [`stack_size`]: Builder::stack_size
255 /// [`name`]: Builder::name
256 /// [`spawn`]: Builder::spawn
257 /// [`thread::spawn`]: spawn
258 /// [`io::Result`]: crate::io::Result
259 /// [`unwrap`]: crate::result::Result::unwrap
260 /// [naming-threads]: ./index.html#naming-threads
261 /// [stack-size]: ./index.html#stack-size
262 #[must_use = "must eventually spawn the thread"]
263 #[stable(feature = "rust1", since = "1.0.0")]
266 // A name for the thread-to-be, for identification in panic messages
267 name: Option<String>,
268 // The size of the stack for the spawned thread in bytes
269 stack_size: Option<usize>,
273 /// Generates the base configuration for spawning a thread, from which
274 /// configuration methods can be chained.
281 /// let builder = thread::Builder::new()
282 /// .name("foo".into())
283 /// .stack_size(32 * 1024);
285 /// let handler = builder.spawn(|| {
289 /// handler.join().unwrap();
291 #[stable(feature = "rust1", since = "1.0.0")]
292 pub fn new() -> Builder {
293 Builder { name: None, stack_size: None }
296 /// Names the thread-to-be. Currently the name is used for identification
297 /// only in panic messages.
299 /// The name must not contain null bytes (`\0`).
301 /// For more information about named threads, see
302 /// [this module-level documentation][naming-threads].
309 /// let builder = thread::Builder::new()
310 /// .name("foo".into());
312 /// let handler = builder.spawn(|| {
313 /// assert_eq!(thread::current().name(), Some("foo"))
316 /// handler.join().unwrap();
319 /// [naming-threads]: ./index.html#naming-threads
320 #[stable(feature = "rust1", since = "1.0.0")]
321 pub fn name(mut self, name: String) -> Builder {
322 self.name = Some(name);
326 /// Sets the size of the stack (in bytes) for the new thread.
328 /// The actual stack size may be greater than this value if
329 /// the platform specifies a minimal stack size.
331 /// For more information about the stack size for threads, see
332 /// [this module-level documentation][stack-size].
339 /// let builder = thread::Builder::new().stack_size(32 * 1024);
342 /// [stack-size]: ./index.html#stack-size
343 #[stable(feature = "rust1", since = "1.0.0")]
344 pub fn stack_size(mut self, size: usize) -> Builder {
345 self.stack_size = Some(size);
349 /// Spawns a new thread by taking ownership of the `Builder`, and returns an
350 /// [`io::Result`] to its [`JoinHandle`].
352 /// The spawned thread may outlive the caller (unless the caller thread
353 /// is the main thread; the whole process is terminated when the main
354 /// thread finishes). The join handle can be used to block on
355 /// termination of the spawned thread, including recovering its panics.
357 /// For a more complete documentation see [`thread::spawn`][`spawn`].
361 /// Unlike the [`spawn`] free function, this method yields an
362 /// [`io::Result`] to capture any failure to create the thread at
365 /// [`io::Result`]: crate::io::Result
369 /// Panics if a thread name was set and it contained null bytes.
376 /// let builder = thread::Builder::new();
378 /// let handler = builder.spawn(|| {
382 /// handler.join().unwrap();
384 #[stable(feature = "rust1", since = "1.0.0")]
385 pub fn spawn<F, T>(self, f: F) -> io::Result<JoinHandle<T>>
391 unsafe { self.spawn_unchecked(f) }
394 /// Spawns a new thread without any lifetime restrictions by taking ownership
395 /// of the `Builder`, and returns an [`io::Result`] to its [`JoinHandle`].
397 /// The spawned thread may outlive the caller (unless the caller thread
398 /// is the main thread; the whole process is terminated when the main
399 /// thread finishes). The join handle can be used to block on
400 /// termination of the spawned thread, including recovering its panics.
402 /// This method is identical to [`thread::Builder::spawn`][`Builder::spawn`],
403 /// except for the relaxed lifetime bounds, which render it unsafe.
404 /// For a more complete documentation see [`thread::spawn`][`spawn`].
408 /// Unlike the [`spawn`] free function, this method yields an
409 /// [`io::Result`] to capture any failure to create the thread at
414 /// Panics if a thread name was set and it contained null bytes.
418 /// The caller has to ensure that the spawned thread does not outlive any
419 /// references in the supplied thread closure and its return type.
420 /// This can be guaranteed in two ways:
422 /// - ensure that [`join`][`JoinHandle::join`] is called before any referenced
424 /// - use only types with `'static` lifetime bounds, i.e., those with no or only
425 /// `'static` references (both [`thread::Builder::spawn`][`Builder::spawn`]
426 /// and [`thread::spawn`][`spawn`] enforce this property statically)
431 /// #![feature(thread_spawn_unchecked)]
434 /// let builder = thread::Builder::new();
437 /// let thread_x = &x;
439 /// let handler = unsafe {
440 /// builder.spawn_unchecked(move || {
441 /// println!("x = {}", *thread_x);
445 /// // caller has to ensure `join()` is called, otherwise
446 /// // it is possible to access freed memory if `x` gets
447 /// // dropped before the thread closure is executed!
448 /// handler.join().unwrap();
451 /// [`io::Result`]: crate::io::Result
452 #[unstable(feature = "thread_spawn_unchecked", issue = "55132")]
453 pub unsafe fn spawn_unchecked<'a, F, T>(self, f: F) -> io::Result<JoinHandle<T>>
459 Ok(JoinHandle(unsafe { self.spawn_unchecked_(f, None) }?))
462 unsafe fn spawn_unchecked_<'a, 'scope, F, T>(
465 scope_data: Option<Arc<scoped::ScopeData>>,
466 ) -> io::Result<JoinInner<'scope, T>>
473 let Builder { name, stack_size } = self;
475 let stack_size = stack_size.unwrap_or_else(thread::min_stack);
477 let my_thread = Thread::new(name.map(|name| {
478 CString::new(name).expect("thread name may not contain interior null bytes")
480 let their_thread = my_thread.clone();
482 let my_packet: Arc<Packet<'scope, T>> = Arc::new(Packet {
484 result: UnsafeCell::new(None),
485 _marker: PhantomData,
487 let their_packet = my_packet.clone();
489 let output_capture = crate::io::set_output_capture(None);
490 crate::io::set_output_capture(output_capture.clone());
493 if let Some(name) = their_thread.cname() {
494 imp::Thread::set_name(name);
497 crate::io::set_output_capture(output_capture);
499 // SAFETY: the stack guard passed is the one for the current thread.
500 // This means the current thread's stack and the new thread's stack
501 // are properly set and protected from each other.
502 thread_info::set(unsafe { imp::guard::current() }, their_thread);
503 let try_result = panic::catch_unwind(panic::AssertUnwindSafe(|| {
504 crate::sys_common::backtrace::__rust_begin_short_backtrace(f)
506 // SAFETY: `their_packet` as been built just above and moved by the
507 // closure (it is an Arc<...>) and `my_packet` will be stored in the
508 // same `JoinInner` as this closure meaning the mutation will be
509 // safe (not modify it and affect a value far away).
510 unsafe { *their_packet.result.get() = Some(try_result) };
513 if let Some(scope_data) = &my_packet.scope {
514 scope_data.increment_num_running_threads();
520 // `imp::Thread::new` takes a closure with a `'static` lifetime, since it's passed
521 // through FFI or otherwise used with low-level threading primitives that have no
522 // notion of or way to enforce lifetimes.
524 // As mentioned in the `Safety` section of this function's documentation, the caller of
525 // this function needs to guarantee that the passed-in lifetime is sufficiently long
526 // for the lifetime of the thread.
528 // Similarly, the `sys` implementation must guarantee that no references to the closure
529 // exist after the thread has terminated, which is signaled by `Thread::join`
534 mem::transmute::<Box<dyn FnOnce() + 'a>, Box<dyn FnOnce() + 'static>>(
545 ////////////////////////////////////////////////////////////////////////////////
547 ////////////////////////////////////////////////////////////////////////////////
549 /// Spawns a new thread, returning a [`JoinHandle`] for it.
551 /// The join handle provides a [`join`] method that can be used to join the spawned
552 /// thread. If the spawned thread panics, [`join`] will return an [`Err`] containing
553 /// the argument given to [`panic!`].
555 /// If the join handle is dropped, the spawned thread will implicitly be *detached*.
556 /// In this case, the spawned thread may no longer be joined.
557 /// (It is the responsibility of the program to either eventually join threads it
558 /// creates or detach them; otherwise, a resource leak will result.)
560 /// This call will create a thread using default parameters of [`Builder`], if you
561 /// want to specify the stack size or the name of the thread, use this API
564 /// As you can see in the signature of `spawn` there are two constraints on
565 /// both the closure given to `spawn` and its return value, let's explain them:
567 /// - The `'static` constraint means that the closure and its return value
568 /// must have a lifetime of the whole program execution. The reason for this
569 /// is that threads can outlive the lifetime they have been created in.
571 /// Indeed if the thread, and by extension its return value, can outlive their
572 /// caller, we need to make sure that they will be valid afterwards, and since
573 /// we *can't* know when it will return we need to have them valid as long as
574 /// possible, that is until the end of the program, hence the `'static`
576 /// - The [`Send`] constraint is because the closure will need to be passed
577 /// *by value* from the thread where it is spawned to the new thread. Its
578 /// return value will need to be passed from the new thread to the thread
579 /// where it is `join`ed.
580 /// As a reminder, the [`Send`] marker trait expresses that it is safe to be
581 /// passed from thread to thread. [`Sync`] expresses that it is safe to have a
582 /// reference be passed from thread to thread.
586 /// Panics if the OS fails to create a thread; use [`Builder::spawn`]
587 /// to recover from such errors.
591 /// Creating a thread.
596 /// let handler = thread::spawn(|| {
600 /// handler.join().unwrap();
603 /// As mentioned in the module documentation, threads are usually made to
604 /// communicate using [`channels`], here is how it usually looks.
606 /// This example also shows how to use `move`, in order to give ownership
607 /// of values to a thread.
611 /// use std::sync::mpsc::channel;
613 /// let (tx, rx) = channel();
615 /// let sender = thread::spawn(move || {
616 /// tx.send("Hello, thread".to_owned())
617 /// .expect("Unable to send on channel");
620 /// let receiver = thread::spawn(move || {
621 /// let value = rx.recv().expect("Unable to receive from channel");
622 /// println!("{value}");
625 /// sender.join().expect("The sender thread has panicked");
626 /// receiver.join().expect("The receiver thread has panicked");
629 /// A thread can also return a value through its [`JoinHandle`], you can use
630 /// this to make asynchronous computations (futures might be more appropriate
636 /// let computation = thread::spawn(|| {
637 /// // Some expensive computation.
641 /// let result = computation.join().unwrap();
642 /// println!("{result}");
645 /// [`channels`]: crate::sync::mpsc
646 /// [`join`]: JoinHandle::join
647 /// [`Err`]: crate::result::Result::Err
648 #[stable(feature = "rust1", since = "1.0.0")]
649 pub fn spawn<F, T>(f: F) -> JoinHandle<T>
655 Builder::new().spawn(f).expect("failed to spawn thread")
658 /// Gets a handle to the thread that invokes it.
662 /// Getting a handle to the current thread with `thread::current()`:
667 /// let handler = thread::Builder::new()
668 /// .name("named thread".into())
670 /// let handle = thread::current();
671 /// assert_eq!(handle.name(), Some("named thread"));
675 /// handler.join().unwrap();
678 #[stable(feature = "rust1", since = "1.0.0")]
679 pub fn current() -> Thread {
680 thread_info::current_thread().expect(
681 "use of std::thread::current() is not possible \
682 after the thread's local data has been destroyed",
686 /// Cooperatively gives up a timeslice to the OS scheduler.
688 /// This calls the underlying OS scheduler's yield primitive, signaling
689 /// that the calling thread is willing to give up its remaining timeslice
690 /// so that the OS may schedule other threads on the CPU.
692 /// A drawback of yielding in a loop is that if the OS does not have any
693 /// other ready threads to run on the current CPU, the thread will effectively
694 /// busy-wait, which wastes CPU time and energy.
696 /// Therefore, when waiting for events of interest, a programmer's first
697 /// choice should be to use synchronization devices such as [`channel`]s,
698 /// [`Condvar`]s, [`Mutex`]es or [`join`] since these primitives are
699 /// implemented in a blocking manner, giving up the CPU until the event
700 /// of interest has occurred which avoids repeated yielding.
702 /// `yield_now` should thus be used only rarely, mostly in situations where
703 /// repeated polling is required because there is no other suitable way to
704 /// learn when an event of interest has occurred.
711 /// thread::yield_now();
714 /// [`channel`]: crate::sync::mpsc
715 /// [`join`]: JoinHandle::join
716 /// [`Condvar`]: crate::sync::Condvar
717 /// [`Mutex`]: crate::sync::Mutex
718 #[stable(feature = "rust1", since = "1.0.0")]
720 imp::Thread::yield_now()
723 /// Determines whether the current thread is unwinding because of panic.
725 /// A common use of this feature is to poison shared resources when writing
726 /// unsafe code, by checking `panicking` when the `drop` is called.
728 /// This is usually not needed when writing safe code, as [`Mutex`es][Mutex]
729 /// already poison themselves when a thread panics while holding the lock.
731 /// This can also be used in multithreaded applications, in order to send a
732 /// message to other threads warning that a thread has panicked (e.g., for
733 /// monitoring purposes).
740 /// struct SomeStruct;
742 /// impl Drop for SomeStruct {
743 /// fn drop(&mut self) {
744 /// if thread::panicking() {
745 /// println!("dropped while unwinding");
747 /// println!("dropped while not unwinding");
754 /// let a = SomeStruct;
759 /// let b = SomeStruct;
764 /// [Mutex]: crate::sync::Mutex
767 #[stable(feature = "rust1", since = "1.0.0")]
768 pub fn panicking() -> bool {
769 panicking::panicking()
772 /// Puts the current thread to sleep for at least the specified amount of time.
774 /// The thread may sleep longer than the duration specified due to scheduling
775 /// specifics or platform-dependent functionality. It will never sleep less.
777 /// This function is blocking, and should not be used in `async` functions.
779 /// # Platform-specific behavior
781 /// On Unix platforms, the underlying syscall may be interrupted by a
782 /// spurious wakeup or signal handler. To ensure the sleep occurs for at least
783 /// the specified duration, this function may invoke that system call multiple
791 /// // Let's sleep for 2 seconds:
792 /// thread::sleep_ms(2000);
794 #[stable(feature = "rust1", since = "1.0.0")]
795 #[deprecated(since = "1.6.0", note = "replaced by `std::thread::sleep`")]
796 pub fn sleep_ms(ms: u32) {
797 sleep(Duration::from_millis(ms as u64))
800 /// Puts the current thread to sleep for at least the specified amount of time.
802 /// The thread may sleep longer than the duration specified due to scheduling
803 /// specifics or platform-dependent functionality. It will never sleep less.
805 /// This function is blocking, and should not be used in `async` functions.
807 /// # Platform-specific behavior
809 /// On Unix platforms, the underlying syscall may be interrupted by a
810 /// spurious wakeup or signal handler. To ensure the sleep occurs for at least
811 /// the specified duration, this function may invoke that system call multiple
813 /// Platforms which do not support nanosecond precision for sleeping will
814 /// have `dur` rounded up to the nearest granularity of time they can sleep for.
816 /// Currently, specifying a zero duration on Unix platforms returns immediately
817 /// without invoking the underlying [`nanosleep`] syscall, whereas on Windows
818 /// platforms the underlying [`Sleep`] syscall is always invoked.
819 /// If the intention is to yield the current time-slice you may want to use
820 /// [`yield_now`] instead.
822 /// [`nanosleep`]: https://linux.die.net/man/2/nanosleep
823 /// [`Sleep`]: https://docs.microsoft.com/en-us/windows/win32/api/synchapi/nf-synchapi-sleep
828 /// use std::{thread, time};
830 /// let ten_millis = time::Duration::from_millis(10);
831 /// let now = time::Instant::now();
833 /// thread::sleep(ten_millis);
835 /// assert!(now.elapsed() >= ten_millis);
837 #[stable(feature = "thread_sleep", since = "1.4.0")]
838 pub fn sleep(dur: Duration) {
839 imp::Thread::sleep(dur)
842 /// Blocks unless or until the current thread's token is made available.
844 /// A call to `park` does not guarantee that the thread will remain parked
845 /// forever, and callers should be prepared for this possibility.
847 /// # park and unpark
849 /// Every thread is equipped with some basic low-level blocking support, via the
850 /// [`thread::park`][`park`] function and [`thread::Thread::unpark`][`unpark`]
851 /// method. [`park`] blocks the current thread, which can then be resumed from
852 /// another thread by calling the [`unpark`] method on the blocked thread's
855 /// Conceptually, each [`Thread`] handle has an associated token, which is
856 /// initially not present:
858 /// * The [`thread::park`][`park`] function blocks the current thread unless or
859 /// until the token is available for its thread handle, at which point it
860 /// atomically consumes the token. It may also return *spuriously*, without
861 /// consuming the token. [`thread::park_timeout`] does the same, but allows
862 /// specifying a maximum time to block the thread for.
864 /// * The [`unpark`] method on a [`Thread`] atomically makes the token available
865 /// if it wasn't already. Because the token is initially absent, [`unpark`]
866 /// followed by [`park`] will result in the second call returning immediately.
868 /// In other words, each [`Thread`] acts a bit like a spinlock that can be
869 /// locked and unlocked using `park` and `unpark`.
871 /// Notice that being unblocked does not imply any synchronization with someone
872 /// that unparked this thread, it could also be spurious.
873 /// For example, it would be a valid, but inefficient, implementation to make both [`park`] and
874 /// [`unpark`] return immediately without doing anything.
876 /// The API is typically used by acquiring a handle to the current thread,
877 /// placing that handle in a shared data structure so that other threads can
878 /// find it, and then `park`ing in a loop. When some desired condition is met, another
879 /// thread calls [`unpark`] on the handle.
881 /// The motivation for this design is twofold:
883 /// * It avoids the need to allocate mutexes and condvars when building new
884 /// synchronization primitives; the threads already provide basic
885 /// blocking/signaling.
887 /// * It can be implemented very efficiently on many platforms.
893 /// use std::sync::{Arc, atomic::{Ordering, AtomicBool}};
894 /// use std::time::Duration;
896 /// let flag = Arc::new(AtomicBool::new(false));
897 /// let flag2 = Arc::clone(&flag);
899 /// let parked_thread = thread::spawn(move || {
900 /// // We want to wait until the flag is set. We *could* just spin, but using
901 /// // park/unpark is more efficient.
902 /// while !flag2.load(Ordering::Acquire) {
903 /// println!("Parking thread");
905 /// // We *could* get here spuriously, i.e., way before the 10ms below are over!
906 /// // But that is no problem, we are in a loop until the flag is set anyway.
907 /// println!("Thread unparked");
909 /// println!("Flag received");
912 /// // Let some time pass for the thread to be spawned.
913 /// thread::sleep(Duration::from_millis(10));
915 /// // Set the flag, and let the thread wake up.
916 /// // There is no race condition here, if `unpark`
917 /// // happens first, `park` will return immediately.
918 /// // Hence there is no risk of a deadlock.
919 /// flag.store(true, Ordering::Release);
920 /// println!("Unpark the thread");
921 /// parked_thread.thread().unpark();
923 /// parked_thread.join().unwrap();
926 /// [`unpark`]: Thread::unpark
927 /// [`thread::park_timeout`]: park_timeout
928 #[stable(feature = "rust1", since = "1.0.0")]
930 // SAFETY: park_timeout is called on the parker owned by this thread.
932 current().inner.as_ref().parker().park();
936 /// Use [`park_timeout`].
938 /// Blocks unless or until the current thread's token is made available or
939 /// the specified duration has been reached (may wake spuriously).
941 /// The semantics of this function are equivalent to [`park`] except
942 /// that the thread will be blocked for roughly no longer than `dur`. This
943 /// method should not be used for precise timing due to anomalies such as
944 /// preemption or platform differences that might not cause the maximum
945 /// amount of time waited to be precisely `ms` long.
947 /// See the [park documentation][`park`] for more detail.
948 #[stable(feature = "rust1", since = "1.0.0")]
949 #[deprecated(since = "1.6.0", note = "replaced by `std::thread::park_timeout`")]
950 pub fn park_timeout_ms(ms: u32) {
951 park_timeout(Duration::from_millis(ms as u64))
954 /// Blocks unless or until the current thread's token is made available or
955 /// the specified duration has been reached (may wake spuriously).
957 /// The semantics of this function are equivalent to [`park`][park] except
958 /// that the thread will be blocked for roughly no longer than `dur`. This
959 /// method should not be used for precise timing due to anomalies such as
960 /// preemption or platform differences that might not cause the maximum
961 /// amount of time waited to be precisely `dur` long.
963 /// See the [park documentation][park] for more details.
965 /// # Platform-specific behavior
967 /// Platforms which do not support nanosecond precision for sleeping will have
968 /// `dur` rounded up to the nearest granularity of time they can sleep for.
972 /// Waiting for the complete expiration of the timeout:
975 /// use std::thread::park_timeout;
976 /// use std::time::{Instant, Duration};
978 /// let timeout = Duration::from_secs(2);
979 /// let beginning_park = Instant::now();
981 /// let mut timeout_remaining = timeout;
983 /// park_timeout(timeout_remaining);
984 /// let elapsed = beginning_park.elapsed();
985 /// if elapsed >= timeout {
988 /// println!("restarting park_timeout after {elapsed:?}");
989 /// timeout_remaining = timeout - elapsed;
992 #[stable(feature = "park_timeout", since = "1.4.0")]
993 pub fn park_timeout(dur: Duration) {
994 // SAFETY: park_timeout is called on the parker owned by this thread.
996 current().inner.as_ref().parker().park_timeout(dur);
1000 ////////////////////////////////////////////////////////////////////////////////
1002 ////////////////////////////////////////////////////////////////////////////////
1004 /// A unique identifier for a running thread.
1006 /// A `ThreadId` is an opaque object that uniquely identifies each thread
1007 /// created during the lifetime of a process. `ThreadId`s are guaranteed not to
1008 /// be reused, even when a thread terminates. `ThreadId`s are under the control
1009 /// of Rust's standard library and there may not be any relationship between
1010 /// `ThreadId` and the underlying platform's notion of a thread identifier --
1011 /// the two concepts cannot, therefore, be used interchangeably. A `ThreadId`
1012 /// can be retrieved from the [`id`] method on a [`Thread`].
1017 /// use std::thread;
1019 /// let other_thread = thread::spawn(|| {
1020 /// thread::current().id()
1023 /// let other_thread_id = other_thread.join().unwrap();
1024 /// assert!(thread::current().id() != other_thread_id);
1027 /// [`id`]: Thread::id
1028 #[stable(feature = "thread_id", since = "1.19.0")]
1029 #[derive(Eq, PartialEq, Clone, Copy, Hash, Debug)]
1030 pub struct ThreadId(NonZeroU64);
1033 // Generate a new unique thread ID.
1034 fn new() -> ThreadId {
1036 fn exhausted() -> ! {
1037 panic!("failed to generate unique thread ID: bitspace exhausted")
1041 if #[cfg(target_has_atomic = "64")] {
1042 use crate::sync::atomic::{AtomicU64, Ordering::Relaxed};
1044 static COUNTER: AtomicU64 = AtomicU64::new(0);
1046 let mut last = COUNTER.load(Relaxed);
1048 let Some(id) = last.checked_add(1) else {
1052 match COUNTER.compare_exchange_weak(last, id, Relaxed, Relaxed) {
1053 Ok(_) => return ThreadId(NonZeroU64::new(id).unwrap()),
1054 Err(id) => last = id,
1058 use crate::sys_common::mutex::StaticMutex;
1060 // It is UB to attempt to acquire this mutex reentrantly!
1061 static GUARD: StaticMutex = StaticMutex::new();
1062 static mut COUNTER: u64 = 0;
1065 let guard = GUARD.lock();
1067 let Some(id) = COUNTER.checked_add(1) else {
1068 drop(guard); // in case the panic handler ends up calling `ThreadId::new()`, avoid reentrant lock acquire.
1074 ThreadId(NonZeroU64::new(id).unwrap())
1080 /// This returns a numeric identifier for the thread identified by this
1083 /// As noted in the documentation for the type itself, it is essentially an
1084 /// opaque ID, but is guaranteed to be unique for each thread. The returned
1085 /// value is entirely opaque -- only equality testing is stable. Note that
1086 /// it is not guaranteed which values new threads will return, and this may
1087 /// change across Rust versions.
1089 #[unstable(feature = "thread_id_value", issue = "67939")]
1090 pub fn as_u64(&self) -> NonZeroU64 {
1095 ////////////////////////////////////////////////////////////////////////////////
1097 ////////////////////////////////////////////////////////////////////////////////
1099 /// The internal representation of a `Thread` handle
1101 name: Option<CString>, // Guaranteed to be UTF-8
1107 fn parker(self: Pin<&Self>) -> Pin<&Parker> {
1108 unsafe { Pin::map_unchecked(self, |inner| &inner.parker) }
1113 #[stable(feature = "rust1", since = "1.0.0")]
1114 /// A handle to a thread.
1116 /// Threads are represented via the `Thread` type, which you can get in one of
1119 /// * By spawning a new thread, e.g., using the [`thread::spawn`][`spawn`]
1120 /// function, and calling [`thread`][`JoinHandle::thread`] on the
1122 /// * By requesting the current thread, using the [`thread::current`] function.
1124 /// The [`thread::current`] function is available even for threads not spawned
1125 /// by the APIs of this module.
1127 /// There is usually no need to create a `Thread` struct yourself, one
1128 /// should instead use a function like `spawn` to create new threads, see the
1129 /// docs of [`Builder`] and [`spawn`] for more details.
1131 /// [`thread::current`]: current
1133 inner: Pin<Arc<Inner>>,
1137 // Used only internally to construct a thread object without spawning
1138 // Panics if the name contains nuls.
1139 pub(crate) fn new(name: Option<CString>) -> Thread {
1140 // We have to use `unsafe` here to construct the `Parker` in-place,
1141 // which is required for the UNIX implementation.
1143 // SAFETY: We pin the Arc immediately after creation, so its address never
1145 let inner = unsafe {
1146 let mut arc = Arc::<Inner>::new_uninit();
1147 let ptr = Arc::get_mut_unchecked(&mut arc).as_mut_ptr();
1148 addr_of_mut!((*ptr).name).write(name);
1149 addr_of_mut!((*ptr).id).write(ThreadId::new());
1150 Parker::new(addr_of_mut!((*ptr).parker));
1151 Pin::new_unchecked(arc.assume_init())
1157 /// Atomically makes the handle's token available if it is not already.
1159 /// Every thread is equipped with some basic low-level blocking support, via
1160 /// the [`park`][park] function and the `unpark()` method. These can be
1161 /// used as a more CPU-efficient implementation of a spinlock.
1163 /// See the [park documentation][park] for more details.
1168 /// use std::thread;
1169 /// use std::time::Duration;
1171 /// let parked_thread = thread::Builder::new()
1173 /// println!("Parking thread");
1175 /// println!("Thread unparked");
1179 /// // Let some time pass for the thread to be spawned.
1180 /// thread::sleep(Duration::from_millis(10));
1182 /// println!("Unpark the thread");
1183 /// parked_thread.thread().unpark();
1185 /// parked_thread.join().unwrap();
1187 #[stable(feature = "rust1", since = "1.0.0")]
1189 pub fn unpark(&self) {
1190 self.inner.as_ref().parker().unpark();
1193 /// Gets the thread's unique identifier.
1198 /// use std::thread;
1200 /// let other_thread = thread::spawn(|| {
1201 /// thread::current().id()
1204 /// let other_thread_id = other_thread.join().unwrap();
1205 /// assert!(thread::current().id() != other_thread_id);
1207 #[stable(feature = "thread_id", since = "1.19.0")]
1209 pub fn id(&self) -> ThreadId {
1213 /// Gets the thread's name.
1215 /// For more information about named threads, see
1216 /// [this module-level documentation][naming-threads].
1220 /// Threads by default have no name specified:
1223 /// use std::thread;
1225 /// let builder = thread::Builder::new();
1227 /// let handler = builder.spawn(|| {
1228 /// assert!(thread::current().name().is_none());
1231 /// handler.join().unwrap();
1234 /// Thread with a specified name:
1237 /// use std::thread;
1239 /// let builder = thread::Builder::new()
1240 /// .name("foo".into());
1242 /// let handler = builder.spawn(|| {
1243 /// assert_eq!(thread::current().name(), Some("foo"))
1246 /// handler.join().unwrap();
1249 /// [naming-threads]: ./index.html#naming-threads
1250 #[stable(feature = "rust1", since = "1.0.0")]
1252 pub fn name(&self) -> Option<&str> {
1253 self.cname().map(|s| unsafe { str::from_utf8_unchecked(s.to_bytes()) })
1256 fn cname(&self) -> Option<&CStr> {
1257 self.inner.name.as_deref()
1261 #[stable(feature = "rust1", since = "1.0.0")]
1262 impl fmt::Debug for Thread {
1263 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1264 f.debug_struct("Thread")
1265 .field("id", &self.id())
1266 .field("name", &self.name())
1267 .finish_non_exhaustive()
1271 ////////////////////////////////////////////////////////////////////////////////
1273 ////////////////////////////////////////////////////////////////////////////////
1275 /// A specialized [`Result`] type for threads.
1277 /// Indicates the manner in which a thread exited.
1279 /// The value contained in the `Result::Err` variant
1280 /// is the value the thread panicked with;
1281 /// that is, the argument the `panic!` macro was called with.
1282 /// Unlike with normal errors, this value doesn't implement
1283 /// the [`Error`](crate::error::Error) trait.
1285 /// Thus, a sensible way to handle a thread panic is to either:
1287 /// 1. propagate the panic with [`std::panic::resume_unwind`]
1288 /// 2. or in case the thread is intended to be a subsystem boundary
1289 /// that is supposed to isolate system-level failures,
1290 /// match on the `Err` variant and handle the panic in an appropriate way
1292 /// A thread that completes without panicking is considered to exit successfully.
1296 /// Matching on the result of a joined thread:
1299 /// use std::{fs, thread, panic};
1301 /// fn copy_in_thread() -> thread::Result<()> {
1302 /// thread::spawn(|| {
1303 /// fs::copy("foo.txt", "bar.txt").unwrap();
1308 /// match copy_in_thread() {
1309 /// Ok(_) => println!("copy succeeded"),
1310 /// Err(e) => panic::resume_unwind(e),
1315 /// [`Result`]: crate::result::Result
1316 /// [`std::panic::resume_unwind`]: crate::panic::resume_unwind
1317 #[stable(feature = "rust1", since = "1.0.0")]
1318 pub type Result<T> = crate::result::Result<T, Box<dyn Any + Send + 'static>>;
1320 // This packet is used to communicate the return value between the spawned
1321 // thread and the rest of the program. It is shared through an `Arc` and
1322 // there's no need for a mutex here because synchronization happens with `join()`
1323 // (the caller will never read this packet until the thread has exited).
1325 // An Arc to the packet is stored into a `JoinInner` which in turns is placed
1327 struct Packet<'scope, T> {
1328 scope: Option<Arc<scoped::ScopeData>>,
1329 result: UnsafeCell<Option<Result<T>>>,
1330 _marker: PhantomData<Option<&'scope scoped::ScopeData>>,
1333 // Due to the usage of `UnsafeCell` we need to manually implement Sync.
1334 // The type `T` should already always be Send (otherwise the thread could not
1335 // have been created) and the Packet is Sync because all access to the
1336 // `UnsafeCell` synchronized (by the `join()` boundary), and `ScopeData` is Sync.
1337 unsafe impl<'scope, T: Sync> Sync for Packet<'scope, T> {}
1339 impl<'scope, T> Drop for Packet<'scope, T> {
1340 fn drop(&mut self) {
1341 // If this packet was for a thread that ran in a scope, the thread
1342 // panicked, and nobody consumed the panic payload, we make sure
1343 // the scope function will panic.
1344 let unhandled_panic = matches!(self.result.get_mut(), Some(Err(_)));
1345 // Drop the result without causing unwinding.
1346 // This is only relevant for threads that aren't join()ed, as
1347 // join() will take the `result` and set it to None, such that
1348 // there is nothing left to drop here.
1349 // If this panics, we should handle that, because we're outside the
1350 // outermost `catch_unwind` of our thread.
1351 // We just abort in that case, since there's nothing else we can do.
1352 // (And even if we tried to handle it somehow, we'd also need to handle
1353 // the case where the panic payload we get out of it also panics on
1354 // drop, and so on. See issue #86027.)
1355 if let Err(_) = panic::catch_unwind(panic::AssertUnwindSafe(|| {
1356 *self.result.get_mut() = None;
1358 rtabort!("thread result panicked on drop");
1360 // Book-keeping so the scope knows when it's done.
1361 if let Some(scope) = &self.scope {
1362 // Now that there will be no more user code running on this thread
1363 // that can use 'scope, mark the thread as 'finished'.
1364 // It's important we only do this after the `result` has been dropped,
1365 // since dropping it might still use things it borrowed from 'scope.
1366 scope.decrement_num_running_threads(unhandled_panic);
1371 /// Inner representation for JoinHandle
1372 struct JoinInner<'scope, T> {
1373 native: imp::Thread,
1375 packet: Arc<Packet<'scope, T>>,
1378 impl<'scope, T> JoinInner<'scope, T> {
1379 fn join(mut self) -> Result<T> {
1381 Arc::get_mut(&mut self.packet).unwrap().result.get_mut().take().unwrap()
1385 /// An owned permission to join on a thread (block on its termination).
1387 /// A `JoinHandle` *detaches* the associated thread when it is dropped, which
1388 /// means that there is no longer any handle to the thread and no way to `join`
1391 /// Due to platform restrictions, it is not possible to [`Clone`] this
1392 /// handle: the ability to join a thread is a uniquely-owned permission.
1394 /// This `struct` is created by the [`thread::spawn`] function and the
1395 /// [`thread::Builder::spawn`] method.
1399 /// Creation from [`thread::spawn`]:
1402 /// use std::thread;
1404 /// let join_handle: thread::JoinHandle<_> = thread::spawn(|| {
1405 /// // some work here
1409 /// Creation from [`thread::Builder::spawn`]:
1412 /// use std::thread;
1414 /// let builder = thread::Builder::new();
1416 /// let join_handle: thread::JoinHandle<_> = builder.spawn(|| {
1417 /// // some work here
1421 /// A thread being detached and outliving the thread that spawned it:
1424 /// use std::thread;
1425 /// use std::time::Duration;
1427 /// let original_thread = thread::spawn(|| {
1428 /// let _detached_thread = thread::spawn(|| {
1429 /// // Here we sleep to make sure that the first thread returns before.
1430 /// thread::sleep(Duration::from_millis(10));
1431 /// // This will be called, even though the JoinHandle is dropped.
1432 /// println!("♫ Still alive ♫");
1436 /// original_thread.join().expect("The thread being joined has panicked");
1437 /// println!("Original thread is joined.");
1439 /// // We make sure that the new thread has time to run, before the main
1440 /// // thread returns.
1442 /// thread::sleep(Duration::from_millis(1000));
1445 /// [`thread::Builder::spawn`]: Builder::spawn
1446 /// [`thread::spawn`]: spawn
1447 #[stable(feature = "rust1", since = "1.0.0")]
1448 pub struct JoinHandle<T>(JoinInner<'static, T>);
1450 #[stable(feature = "joinhandle_impl_send_sync", since = "1.29.0")]
1451 unsafe impl<T> Send for JoinHandle<T> {}
1452 #[stable(feature = "joinhandle_impl_send_sync", since = "1.29.0")]
1453 unsafe impl<T> Sync for JoinHandle<T> {}
1455 impl<T> JoinHandle<T> {
1456 /// Extracts a handle to the underlying thread.
1461 /// use std::thread;
1463 /// let builder = thread::Builder::new();
1465 /// let join_handle: thread::JoinHandle<_> = builder.spawn(|| {
1466 /// // some work here
1469 /// let thread = join_handle.thread();
1470 /// println!("thread id: {:?}", thread.id());
1472 #[stable(feature = "rust1", since = "1.0.0")]
1474 pub fn thread(&self) -> &Thread {
1478 /// Waits for the associated thread to finish.
1480 /// This function will return immediately if the associated thread has already finished.
1482 /// In terms of [atomic memory orderings], the completion of the associated
1483 /// thread synchronizes with this function returning. In other words, all
1484 /// operations performed by that thread [happen
1485 /// before](https://doc.rust-lang.org/nomicon/atomics.html#data-accesses) all
1486 /// operations that happen after `join` returns.
1488 /// If the associated thread panics, [`Err`] is returned with the parameter given
1491 /// [`Err`]: crate::result::Result::Err
1492 /// [atomic memory orderings]: crate::sync::atomic
1496 /// This function may panic on some platforms if a thread attempts to join
1497 /// itself or otherwise may create a deadlock with joining threads.
1502 /// use std::thread;
1504 /// let builder = thread::Builder::new();
1506 /// let join_handle: thread::JoinHandle<_> = builder.spawn(|| {
1507 /// // some work here
1509 /// join_handle.join().expect("Couldn't join on the associated thread");
1511 #[stable(feature = "rust1", since = "1.0.0")]
1512 pub fn join(self) -> Result<T> {
1516 /// Checks if the associated thread has finished running its main function.
1518 /// `is_finished` supports implementing a non-blocking join operation, by checking
1519 /// `is_finished`, and calling `join` if it returns `true`. This function does not block. To
1520 /// block while waiting on the thread to finish, use [`join`][Self::join].
1522 /// This might return `true` for a brief moment after the thread's main
1523 /// function has returned, but before the thread itself has stopped running.
1524 /// However, once this returns `true`, [`join`][Self::join] can be expected
1525 /// to return quickly, without blocking for any significant amount of time.
1526 #[stable(feature = "thread_is_running", since = "1.61.0")]
1527 pub fn is_finished(&self) -> bool {
1528 Arc::strong_count(&self.0.packet) == 1
1532 impl<T> AsInner<imp::Thread> for JoinHandle<T> {
1533 fn as_inner(&self) -> &imp::Thread {
1538 impl<T> IntoInner<imp::Thread> for JoinHandle<T> {
1539 fn into_inner(self) -> imp::Thread {
1544 #[stable(feature = "std_debug", since = "1.16.0")]
1545 impl<T> fmt::Debug for JoinHandle<T> {
1546 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1547 f.debug_struct("JoinHandle").finish_non_exhaustive()
1551 fn _assert_sync_and_send() {
1552 fn _assert_both<T: Send + Sync>() {}
1553 _assert_both::<JoinHandle<()>>();
1554 _assert_both::<Thread>();
1557 /// Returns an estimate of the default amount of parallelism a program should use.
1559 /// Parallelism is a resource. A given machine provides a certain capacity for
1560 /// parallelism, i.e., a bound on the number of computations it can perform
1561 /// simultaneously. This number often corresponds to the amount of CPUs a
1562 /// computer has, but it may diverge in various cases.
1564 /// Host environments such as VMs or container orchestrators may want to
1565 /// restrict the amount of parallelism made available to programs in them. This
1566 /// is often done to limit the potential impact of (unintentionally)
1567 /// resource-intensive programs on other programs running on the same machine.
1571 /// The purpose of this API is to provide an easy and portable way to query
1572 /// the default amount of parallelism the program should use. Among other things it
1573 /// does not expose information on NUMA regions, does not account for
1574 /// differences in (co)processor capabilities or current system load,
1575 /// and will not modify the program's global state in order to more accurately
1576 /// query the amount of available parallelism.
1578 /// Where both fixed steady-state and burst limits are available the steady-state
1579 /// capacity will be used to ensure more predictable latencies.
1581 /// Resource limits can be changed during the runtime of a program, therefore the value is
1582 /// not cached and instead recomputed every time this function is called. It should not be
1583 /// called from hot code.
1585 /// The value returned by this function should be considered a simplified
1586 /// approximation of the actual amount of parallelism available at any given
1587 /// time. To get a more detailed or precise overview of the amount of
1588 /// parallelism available to the program, you may wish to use
1589 /// platform-specific APIs as well. The following platform limitations currently
1590 /// apply to `available_parallelism`:
1593 /// - It may undercount the amount of parallelism available on systems with more
1594 /// than 64 logical CPUs. However, programs typically need specific support to
1595 /// take advantage of more than 64 logical CPUs, and in the absence of such
1596 /// support, the number returned by this function accurately reflects the
1597 /// number of logical CPUs the program can use by default.
1598 /// - It may overcount the amount of parallelism available on systems limited by
1599 /// process-wide affinity masks, or job object limitations.
1602 /// - It may overcount the amount of parallelism available when limited by a
1603 /// process-wide affinity mask or cgroup quotas and `sched_getaffinity()` or cgroup fs can't be
1604 /// queried, e.g. due to sandboxing.
1605 /// - It may undercount the amount of parallelism if the current thread's affinity mask
1606 /// does not reflect the process' cpuset, e.g. due to pinned threads.
1607 /// - If the process is in a cgroup v1 cpu controller, this may need to
1608 /// scan mountpoints to find the corresponding cgroup v1 controller,
1609 /// which may take time on systems with large numbers of mountpoints.
1610 /// (This does not apply to cgroup v2, or to processes not in a
1614 /// - It may overcount the amount of parallelism available when running in a VM
1615 /// with CPU usage limits (e.g. an overcommitted host).
1619 /// This function will, but is not limited to, return errors in the following
1622 /// - If the amount of parallelism is not known for the target platform.
1623 /// - If the program lacks permission to query the amount of parallelism made
1624 /// available to it.
1629 /// # #![allow(dead_code)]
1630 /// use std::{io, thread};
1632 /// fn main() -> io::Result<()> {
1633 /// let count = thread::available_parallelism()?.get();
1634 /// assert!(count >= 1_usize);
1638 #[doc(alias = "available_concurrency")] // Alias for a previous name we gave this API on unstable.
1639 #[doc(alias = "hardware_concurrency")] // Alias for C++ `std::thread::hardware_concurrency`.
1640 #[doc(alias = "num_cpus")] // Alias for a popular ecosystem crate which provides similar functionality.
1641 #[stable(feature = "available_parallelism", since = "1.59.0")]
1642 pub fn available_parallelism() -> io::Result<NonZeroUsize> {
1643 imp::available_parallelism()