1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261
//! Lagoon is a thread pool crate that aims to address many of the problems with existing thread pool crates. //! //! ## Features //! //! - **Scoped jobs**: Safely spawn jobs that have access to their parent scope! //! - **Job handles**: Receive the result of a job when it finishes, or wait on it to finish! //! - **Global pool**: A pay-for-what-you-use global thread pool that avoids dependencies fighting over resources! //! - **Customise thread attributes**: Specify thread name, stack size, etc. //! //! ```ignore //! let pool = lagoon::ThreadPool::default(); //! //! // Spawn some jobs that notify us when they're finished //! let jobs = (0..10) //! .map(|i| pool.run_recv(move || { //! println!("Hello! i = {}", i); //! })) //! .collect::<Vec<_>>(); //! //! // Wait for all jobs to finish //! for job in jobs { //! job.join().unwrap(); //! } //! ``` #![cfg_attr(docsrs, feature(doc_cfg))] #![deny(missing_docs)] #[cfg(feature = "scope")] mod scope; #[cfg(feature = "recv")] mod recv; #[cfg(feature = "scope")] #[cfg_attr(docsrs, doc(cfg(feature = "scope")))] pub use scope::Scope; #[cfg(feature = "recv")] #[cfg_attr(docsrs, doc(cfg(feature = "recv")))] pub use recv::JobHandle; use std::{ thread::{self, JoinHandle}, error, fmt, io, }; // use flume::{Sender, unbounded}; use crossbeam_channel::{unbounded, Sender}; /// Attempt to determine the available concurrency of the host system. /// /// In most cases, this corresponds to the number of CPU cores that are available to the program. If the `num_cpus` /// feature is enabled (it is by default) the [`num_cpus`](https://crates.io/crates/num_cpus) crate will be used to /// determine this value. Otherwise, the nightly-only `std::thread::available_concurrency` function will be used. #[cfg(feature = "num_cpus")] pub fn available_concurrency() -> Option<usize> { Some(num_cpus::get()) } #[cfg(not(feature = "num_cpus"))] pub fn available_concurrency() -> Option<usize> { std::thread::available_concurrency().map(|n| n.get()) } /// An error that may be produced when creating a [`ThreadPool`]. #[derive(Debug)] pub enum Error { /// An IO error occurred when attempting to spawn a thread. Io(io::Error), /// The thread pool has no threads. NoThreads, /// A timeout occurred when attempting to join a job. Timeout, } impl fmt::Display for Error { fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { match self { Self::Io(err) => write!(f, "{}", err), Self::NoThreads => write!(f, "thread pool has no threads"), Self::Timeout => write!(f, "a timeout occurred"), } } } impl error::Error for Error {} struct Job { f: Box<dyn FnOnce() + Send>, } // TODO: Use when stable, see https://github.com/rust-lang/rust/issues/74465 //static GLOBAL: std::lazy::SyncLazy<ThreadPool> = std::lazy::SyncLazy::new(|| ThreadPool::default()); // I'll try spinning, that's a good trick! Actually, this isn't so bad: instead of spinning in a hot loop, we just // yield to the scheduler every time we fail to access the global pool. This prevents priority inversion. static GLOBAL: spin::once::Once<ThreadPool, spin::Yield> = spin::once::Once::new(); /// A pool of threads that may be used to execute jobs. pub struct ThreadPool { tx: Sender<Job>, handles: Vec<JoinHandle<()>>, } impl Default for ThreadPool { fn default() -> Self { Self::build().finish().unwrap() } } impl ThreadPool { /// The default number of threads that will be used if the available concurrency of the environment cannot be /// determined automatically. pub const DEFAULT_THREAD_COUNT: usize = 8; /// Returns a reference to the global [`ThreadPool`], instantiating as with [`ThreadPool::default`] if it is not /// already initialized. /// /// This should be used when you don't require any specific thread pool configuration to avoid multiple thread /// pools fighting for scheduler time. pub fn global() -> &'static Self { Self::global_with_builder(ThreadPoolBuilder::default()) } /// Returns a reference to the global [`ThreadPool`], initializing it with the given [`ThreadPoolBuilder`] if it /// is not already initialized. /// /// Do *not* use this function if you are writing a library. Use [`ThreadPool::global`] instead so that the final /// binary crate is the one to determine the thread pool configuration (the end developer likely knows more about /// their specific threading requirements than you do and can use this information to optimise thread pool /// performance and minimise resource contention). /// /// If your application has specific pool requirements (for example, most games require thread pools to use N - 1 /// threads to ensure that at least one core is free at any given time to keep the main thread running smoothly /// without stuttering) you should use this function *as early as possible* in the program's execution (i.e: at the /// top of the `main` function) to avoid dependencies initializing it first. /// /// Note additionally that the configuration you choose might interfere with dependencies that also use the global /// thread pool. Choose sensible, accomodating defaults where possible. pub fn global_with_builder(builder: ThreadPoolBuilder) -> &'static Self { GLOBAL.call_once(|| builder.finish().expect("Failed to initialise global thread pool")) } /// Begin building a new [`ThreadPool`]. pub fn build() -> ThreadPoolBuilder { ThreadPoolBuilder::default() } /// Returns the number of threads in this pool. pub fn thread_count(&self) -> usize { self.handles.len() } /// Returns the number of jobs waiting to be executed. pub fn queue_len(&self) -> usize { self.tx.len() } /// Enqueue a function to be executed as a job when a thread is free to do so. /// /// ``` /// let pool = lagoon::ThreadPool::default(); /// /// for i in 0..10 { /// pool.run(move || println!("I am the {}th job!", i)); /// } /// ``` pub fn run<F: FnOnce() + Send + 'static>(&self, f: F) { self.tx.send(Job { f: Box::new(f) }).unwrap() } /// Enqueue a function to be executed as a job when a thread is free to do so, returning a handle that allows /// retrieval of the return value of the function. #[cfg(feature = "recv")] pub fn run_recv<F: FnOnce() -> R + Send + 'static, R: Send + 'static>(&self, f: F) -> recv::JobHandle<R> { let (tx, rx) = oneshot::channel(); self.run(move || { let _ = tx.send(f()); }); recv::JobHandle::new(rx) } /// Signal to threads (not jobs) that they should stop, then wait for them to finish processing jobs. /// /// All outstanding jobs will be executed before this function returns. pub fn join_all(self) -> thread::Result<()> { let Self { tx, handles } = self; drop(tx); for handle in handles { handle.join()?; } Ok(()) } /// Create a scope that allows the spawning of threads with safe access to the current scope. /// /// This function will wait for all jobs created in the scope to finish before continuing. See [`Scope`] for more /// information about scoped jobs. #[cfg(feature = "scope")] pub fn scoped<'pool, 'scope, F: FnOnce(scope::Scope<'pool, 'scope>) -> R, R>(&'pool self, f: F) -> R { scope::run(self, f) } } /// A type used to configure a [`ThreadPool`] prior to its creation. #[derive(Clone, Default)] pub struct ThreadPoolBuilder { thread_count: Option<usize>, thread_name: Option<String>, thread_stack_size: Option<usize>, } impl ThreadPoolBuilder { /// Configure the [`ThreadPool`] with the given number of threads. If unspecified, the thread pool will attempt to /// detect the number of hardware threads available to the process and use that. If this also fails, /// [`ThreadPool::DEFAULT_THREAD_COUNT`] will be used. pub fn with_thread_count(self, thread_count: usize) -> Self { Self { thread_count: Some(thread_count), ..self } } /// Give the threads owned by this [`ThreadPool`] the given name. If unspecified, the default name will be the same /// as those created by [`std::thread::spawn`]. pub fn with_thread_name(self, name: String) -> Self { Self { thread_name: Some(name), ..self } } /// Give the threads owned by this [`ThreadPool`] a specific stack size. If unspecified, the default stack size /// will be the same as those created by [`std::thread::spawn`]. pub fn with_thread_stack_size(self, size: usize) -> Self { Self { thread_stack_size: Some(size), ..self } } /// Finish configuration, returning a [`ThreadPool`]. pub fn finish(self) -> Result<ThreadPool, Error> { let thread_count = self.thread_count .or_else(|| available_concurrency()) .unwrap_or(ThreadPool::DEFAULT_THREAD_COUNT); if thread_count == 0 { return Err(Error::NoThreads); } let (tx, rx) = unbounded(); Ok(ThreadPool { tx, handles: (0..thread_count) .map(|_| { let rx = rx.clone(); let builder = thread::Builder::new(); let builder = match self.thread_name.clone() { Some(name) => builder.name(name), None => builder, }; let builder = match self.thread_stack_size { Some(size) => builder.stack_size(size), None => builder, }; builder.spawn(move || { while let Ok(job) = rx.recv() { let job = std::panic::AssertUnwindSafe(job); let _ = std::panic::catch_unwind(move || { (job.0.f)(); }); } }).map_err(Error::Io) }) .collect::<Result<_, _>>()?, }) } }