mirror of
https://github.com/zed-industries/zed.git
synced 2026-05-31 19:05:00 +07:00
Adds `scheduler::spawn_dedicated_thread` (and inherent `spawn_dedicated` methods on `PlatformScheduler` and `TestScheduler`) so single-threaded actors that own `!Send` state can run on their own OS thread and freely do blocking I/O without disturbing any other executor. ### Why A single-threaded actor that needs to do blocking syscalls is currently stuck: it can't run on the shared foreground executor (blocking would stall every other foreground session), and it can't move to the background pool because its state isn't `Send`. `spawn_dedicated` gives each such actor its own thread and its own `LocalExecutor`, while still participating in the same testable scheduler infrastructure as everything else. ### Shape - `pub fn spawn_dedicated_thread(session_id, scheduler, f) -> Task<_>` in `scheduler`. Owns the OS thread, the per-session runnable channel, and the `LocalExecutor` setup. - Inherent `spawn_dedicated` on `PlatformScheduler` (allocates its own `SessionId`, delegates to the free function). - Inherent `spawn_dedicated` on `TestScheduler` (no real thread — runs as a fresh local session driven by the test scheduler's run loop, so determinism under `many` is preserved). - Renames `Scheduler::schedule_foreground` → `schedule_local` and `scheduler::ForegroundExecutor` → `scheduler::LocalExecutor` to reflect that these are session-pinned queues rather than "the main thread" (a dedicated session runs on its own thread). GPUI's wrapper `gpui::ForegroundExecutor` and the `foreground_executor` field/method names are unchanged to keep blast radius small. - `LocalExecutor::new` now takes an explicit dispatch closure, so the routing decision (default session, dedicated thread, or something else) lives at the construction site. ### Tests - `TestScheduler` side: round-trip, `!Send` future, `Send` closure capturing shared state, inner `executor.spawn`, determinism under `many` seeds, drop-cancels-future, detached child runs after root completes. - `PlatformScheduler` side: real separate thread (blocking syscalls don't stall the test), `!Send` future output, drop-cancels-future, thread tears down after work completes, detached child outlives root. cc @as-cii Release Notes: - N/A --------- Co-authored-by: Antonio Scandurra <me@as-cii.com> Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
498 lines
17 KiB
Rust
498 lines
17 KiB
Rust
use crate::{App, PlatformDispatcher, PlatformScheduler};
|
|
use futures::channel::mpsc;
|
|
use futures::prelude::*;
|
|
use gpui_util::{TryFutureExt, TryFutureExtBacktrace};
|
|
use scheduler::Instant;
|
|
use scheduler::Scheduler;
|
|
use std::{future::Future, marker::PhantomData, mem, pin::Pin, rc::Rc, sync::Arc, time::Duration};
|
|
|
|
pub use scheduler::{FallibleTask, LocalExecutor as SchedulerLocalExecutor, Priority, Task};
|
|
|
|
/// A pointer to the executor that is currently running,
|
|
/// for spawning background tasks.
|
|
#[derive(Clone)]
|
|
pub struct BackgroundExecutor {
|
|
inner: scheduler::BackgroundExecutor,
|
|
dispatcher: Arc<dyn PlatformDispatcher>,
|
|
}
|
|
|
|
/// A pointer to the executor that is currently running,
|
|
/// for spawning tasks on the main thread.
|
|
#[derive(Clone)]
|
|
pub struct ForegroundExecutor {
|
|
inner: scheduler::LocalExecutor,
|
|
dispatcher: Arc<dyn PlatformDispatcher>,
|
|
not_send: PhantomData<Rc<()>>,
|
|
}
|
|
|
|
/// Extension trait for `Task<Result<T, E>>` that adds `detach_and_log_err` with an `&App` context.
|
|
///
|
|
/// This trait is automatically implemented for all `Task<Result<T, E>>` types.
|
|
pub trait TaskExt<T, E> {
|
|
/// Run the task to completion in the background and log any errors that occur.
|
|
fn detach_and_log_err(self, cx: &App);
|
|
/// Like [`Self::detach_and_log_err`], but uses `{:?}` formatting on failure so `anyhow::Error`
|
|
/// values emit their full backtrace. Prefer `detach_and_log_err` unless a backtrace is wanted.
|
|
fn detach_and_log_err_with_backtrace(self, cx: &App);
|
|
}
|
|
|
|
impl<T, E> TaskExt<T, E> for Task<Result<T, E>>
|
|
where
|
|
T: 'static,
|
|
E: 'static + std::fmt::Display + std::fmt::Debug,
|
|
{
|
|
#[track_caller]
|
|
fn detach_and_log_err(self, cx: &App) {
|
|
let location = core::panic::Location::caller();
|
|
cx.foreground_executor()
|
|
.spawn(self.log_tracked_err(*location))
|
|
.detach();
|
|
}
|
|
|
|
#[track_caller]
|
|
fn detach_and_log_err_with_backtrace(self, cx: &App) {
|
|
let location = *core::panic::Location::caller();
|
|
cx.foreground_executor()
|
|
.spawn(self.log_tracked_err_with_backtrace(location))
|
|
.detach();
|
|
}
|
|
}
|
|
|
|
impl BackgroundExecutor {
|
|
/// Creates a new BackgroundExecutor from the given PlatformDispatcher.
|
|
pub fn new(dispatcher: Arc<dyn PlatformDispatcher>) -> Self {
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
let scheduler: Arc<dyn Scheduler> = if let Some(test_dispatcher) = dispatcher.as_test() {
|
|
test_dispatcher.scheduler().clone()
|
|
} else {
|
|
Arc::new(PlatformScheduler::new(dispatcher.clone()))
|
|
};
|
|
|
|
#[cfg(not(any(test, feature = "test-support")))]
|
|
let scheduler: Arc<dyn Scheduler> = Arc::new(PlatformScheduler::new(dispatcher.clone()));
|
|
|
|
Self {
|
|
inner: scheduler::BackgroundExecutor::new(scheduler),
|
|
dispatcher,
|
|
}
|
|
}
|
|
|
|
/// Returns the underlying scheduler::BackgroundExecutor.
|
|
///
|
|
/// This is used by Ex to pass the executor to thread/worktree code.
|
|
pub fn scheduler_executor(&self) -> scheduler::BackgroundExecutor {
|
|
self.inner.clone()
|
|
}
|
|
|
|
/// Enqueues the given future to be run to completion on a background thread.
|
|
#[track_caller]
|
|
pub fn spawn<R>(&self, future: impl Future<Output = R> + Send + 'static) -> Task<R>
|
|
where
|
|
R: Send + 'static,
|
|
{
|
|
self.spawn_with_priority(Priority::default(), future.boxed())
|
|
}
|
|
|
|
/// Enqueues the given future to be run to completion on a background thread with the given priority.
|
|
///
|
|
/// When `Priority::RealtimeAudio` is used, the task runs on a dedicated thread with
|
|
/// realtime scheduling priority, suitable for audio processing.
|
|
#[track_caller]
|
|
pub fn spawn_with_priority<R>(
|
|
&self,
|
|
priority: Priority,
|
|
future: impl Future<Output = R> + Send + 'static,
|
|
) -> Task<R>
|
|
where
|
|
R: Send + 'static,
|
|
{
|
|
if priority == Priority::RealtimeAudio {
|
|
self.inner.spawn_realtime(future)
|
|
} else {
|
|
self.inner.spawn_with_priority(priority, future)
|
|
}
|
|
}
|
|
|
|
/// Scoped lets you start a number of tasks and waits
|
|
/// for all of them to complete before returning.
|
|
pub async fn scoped<'scope, F>(&self, scheduler: F)
|
|
where
|
|
F: FnOnce(&mut Scope<'scope>),
|
|
{
|
|
let mut scope = Scope::new(self.clone(), Priority::default());
|
|
(scheduler)(&mut scope);
|
|
let spawned = mem::take(&mut scope.futures)
|
|
.into_iter()
|
|
.map(|f| self.spawn_with_priority(scope.priority, f))
|
|
.collect::<Vec<_>>();
|
|
for task in spawned {
|
|
task.await;
|
|
}
|
|
}
|
|
|
|
/// Scoped lets you start a number of tasks and waits
|
|
/// for all of them to complete before returning.
|
|
pub async fn scoped_priority<'scope, F>(&self, priority: Priority, scheduler: F)
|
|
where
|
|
F: FnOnce(&mut Scope<'scope>),
|
|
{
|
|
let mut scope = Scope::new(self.clone(), priority);
|
|
(scheduler)(&mut scope);
|
|
let spawned = mem::take(&mut scope.futures)
|
|
.into_iter()
|
|
.map(|f| self.spawn_with_priority(scope.priority, f))
|
|
.collect::<Vec<_>>();
|
|
for task in spawned {
|
|
task.await;
|
|
}
|
|
}
|
|
|
|
/// Get the current time.
|
|
///
|
|
/// Calling this instead of `std::time::Instant::now` allows the use
|
|
/// of fake timers in tests.
|
|
pub fn now(&self) -> Instant {
|
|
self.inner.scheduler().clock().now()
|
|
}
|
|
|
|
/// Returns a task that will complete after the given duration.
|
|
/// Depending on other concurrent tasks the elapsed duration may be longer
|
|
/// than requested.
|
|
#[track_caller]
|
|
pub fn timer(&self, duration: Duration) -> Task<()> {
|
|
if duration.is_zero() {
|
|
return Task::ready(());
|
|
}
|
|
self.spawn(self.inner.scheduler().timer(duration))
|
|
}
|
|
|
|
/// In tests, run an arbitrary number of tasks (determined by the SEED environment variable)
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
pub fn simulate_random_delay(&self) -> impl Future<Output = ()> + use<> {
|
|
self.dispatcher.as_test().unwrap().simulate_random_delay()
|
|
}
|
|
|
|
/// In tests, move time forward. This does not run any tasks, but does make `timer`s ready.
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
pub fn advance_clock(&self, duration: Duration) {
|
|
self.dispatcher.as_test().unwrap().advance_clock(duration)
|
|
}
|
|
|
|
/// In tests, run one task.
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
pub fn tick(&self) -> bool {
|
|
self.dispatcher.as_test().unwrap().scheduler().tick()
|
|
}
|
|
|
|
/// In tests, run tasks until the scheduler would park.
|
|
///
|
|
/// Under the scheduler-backed test dispatcher, `tick()` will not advance the clock, so a pending
|
|
/// timer can keep `has_pending_tasks()` true even after all currently-runnable tasks have been
|
|
/// drained. To preserve the historical semantics that tests relied on (drain all work that can
|
|
/// make progress), we advance the clock to the next timer when no runnable tasks remain.
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
pub fn run_until_parked(&self) {
|
|
let scheduler = self.dispatcher.as_test().unwrap().scheduler();
|
|
scheduler.run();
|
|
}
|
|
|
|
/// In tests, prevents `run_until_parked` from panicking if there are outstanding tasks.
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
pub fn allow_parking(&self) {
|
|
self.dispatcher
|
|
.as_test()
|
|
.unwrap()
|
|
.scheduler()
|
|
.allow_parking();
|
|
|
|
if std::env::var("GPUI_RUN_UNTIL_PARKED_LOG").ok().as_deref() == Some("1") {
|
|
log::warn!("[gpui::executor] allow_parking: enabled");
|
|
}
|
|
}
|
|
|
|
/// Sets the range of ticks to run before timing out in block_on.
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
pub fn set_block_on_ticks(&self, range: std::ops::RangeInclusive<usize>) {
|
|
self.dispatcher
|
|
.as_test()
|
|
.unwrap()
|
|
.scheduler()
|
|
.set_timeout_ticks(range);
|
|
}
|
|
|
|
/// Undoes the effect of [`Self::allow_parking`].
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
pub fn forbid_parking(&self) {
|
|
self.dispatcher
|
|
.as_test()
|
|
.unwrap()
|
|
.scheduler()
|
|
.forbid_parking();
|
|
}
|
|
|
|
/// In tests, returns the rng used by the dispatcher.
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
pub fn rng(&self) -> scheduler::SharedRng {
|
|
self.dispatcher.as_test().unwrap().scheduler().rng()
|
|
}
|
|
|
|
/// How many CPUs are available to the dispatcher.
|
|
pub fn num_cpus(&self) -> usize {
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
if let Some(test) = self.dispatcher.as_test() {
|
|
return test.num_cpus_override().unwrap_or(4);
|
|
}
|
|
num_cpus::get()
|
|
}
|
|
|
|
/// Override the number of CPUs reported by this executor in tests.
|
|
/// Panics if not called on a test executor.
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
pub fn set_num_cpus(&self, count: usize) {
|
|
self.dispatcher
|
|
.as_test()
|
|
.expect("set_num_cpus can only be called on a test executor")
|
|
.set_num_cpus(count);
|
|
}
|
|
|
|
/// Whether we're on the main thread.
|
|
pub fn is_main_thread(&self) -> bool {
|
|
self.dispatcher.is_main_thread()
|
|
}
|
|
|
|
#[doc(hidden)]
|
|
pub fn dispatcher(&self) -> &Arc<dyn PlatformDispatcher> {
|
|
&self.dispatcher
|
|
}
|
|
}
|
|
|
|
impl ForegroundExecutor {
|
|
/// Creates a new ForegroundExecutor from the given PlatformDispatcher.
|
|
pub fn new(dispatcher: Arc<dyn PlatformDispatcher>) -> Self {
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
let (scheduler, session_id): (Arc<dyn Scheduler>, _) =
|
|
if let Some(test_dispatcher) = dispatcher.as_test() {
|
|
(
|
|
test_dispatcher.scheduler().clone(),
|
|
test_dispatcher.session_id(),
|
|
)
|
|
} else {
|
|
let platform_scheduler = Arc::new(PlatformScheduler::new(dispatcher.clone()));
|
|
let inner = platform_scheduler.foreground_executor();
|
|
return Self {
|
|
inner,
|
|
dispatcher,
|
|
not_send: PhantomData,
|
|
};
|
|
};
|
|
|
|
#[cfg(not(any(test, feature = "test-support")))]
|
|
let inner = {
|
|
let platform_scheduler = Arc::new(PlatformScheduler::new(dispatcher.clone()));
|
|
platform_scheduler.foreground_executor()
|
|
};
|
|
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
let inner = {
|
|
let scheduler_for_dispatch = Arc::downgrade(&scheduler);
|
|
scheduler::LocalExecutor::new(session_id, scheduler, move |runnable| {
|
|
if let Some(scheduler) = scheduler_for_dispatch.upgrade() {
|
|
scheduler.schedule_local(session_id, runnable);
|
|
}
|
|
})
|
|
};
|
|
|
|
Self {
|
|
inner,
|
|
dispatcher,
|
|
not_send: PhantomData,
|
|
}
|
|
}
|
|
|
|
/// Enqueues the given Task to run on the main thread.
|
|
#[track_caller]
|
|
pub fn spawn<R>(&self, future: impl Future<Output = R> + 'static) -> Task<R>
|
|
where
|
|
R: 'static,
|
|
{
|
|
self.inner.spawn(future.boxed_local())
|
|
}
|
|
|
|
/// Enqueues the given Task to run on the main thread with the given priority.
|
|
#[track_caller]
|
|
pub fn spawn_with_priority<R>(
|
|
&self,
|
|
_priority: Priority,
|
|
future: impl Future<Output = R> + 'static,
|
|
) -> Task<R>
|
|
where
|
|
R: 'static,
|
|
{
|
|
// Priority is ignored for foreground tasks - they run in order on the main thread
|
|
self.inner.spawn(future)
|
|
}
|
|
|
|
/// Used by the test harness to run an async test in a synchronous fashion.
|
|
#[cfg(any(test, feature = "test-support"))]
|
|
#[track_caller]
|
|
pub fn block_test<R>(&self, future: impl Future<Output = R>) -> R {
|
|
use std::cell::Cell;
|
|
|
|
let scheduler = self.inner.scheduler();
|
|
|
|
let output = Cell::new(None);
|
|
let future = async {
|
|
output.set(Some(future.await));
|
|
};
|
|
let mut future = std::pin::pin!(future);
|
|
|
|
// In async GPUI tests, we must allow foreground tasks scheduled by the test itself
|
|
// (which are associated with the test session) to make progress while we block.
|
|
// Otherwise, awaiting futures that depend on same-session foreground work can deadlock.
|
|
scheduler.block(None, future.as_mut(), None);
|
|
|
|
output.take().expect("block_test future did not complete")
|
|
}
|
|
|
|
/// Block the current thread until the given future resolves.
|
|
/// Consider using `block_with_timeout` instead.
|
|
pub fn block_on<R>(&self, future: impl Future<Output = R>) -> R {
|
|
self.inner.block_on(future)
|
|
}
|
|
|
|
/// Block the current thread until the given future resolves or the timeout elapses.
|
|
pub fn block_with_timeout<R, Fut: Future<Output = R>>(
|
|
&self,
|
|
duration: Duration,
|
|
future: Fut,
|
|
) -> Result<R, impl Future<Output = R> + use<R, Fut>> {
|
|
self.inner.block_with_timeout(duration, future)
|
|
}
|
|
|
|
#[doc(hidden)]
|
|
pub fn dispatcher(&self) -> &Arc<dyn PlatformDispatcher> {
|
|
&self.dispatcher
|
|
}
|
|
|
|
#[doc(hidden)]
|
|
pub fn scheduler_executor(&self) -> SchedulerLocalExecutor {
|
|
self.inner.clone()
|
|
}
|
|
}
|
|
|
|
/// Scope manages a set of tasks that are enqueued and waited on together. See [`BackgroundExecutor::scoped`].
|
|
pub struct Scope<'a> {
|
|
executor: BackgroundExecutor,
|
|
priority: Priority,
|
|
futures: Vec<Pin<Box<dyn Future<Output = ()> + Send + 'static>>>,
|
|
tx: Option<mpsc::Sender<()>>,
|
|
rx: mpsc::Receiver<()>,
|
|
lifetime: PhantomData<&'a ()>,
|
|
}
|
|
|
|
impl<'a> Scope<'a> {
|
|
fn new(executor: BackgroundExecutor, priority: Priority) -> Self {
|
|
let (tx, rx) = mpsc::channel(1);
|
|
Self {
|
|
executor,
|
|
priority,
|
|
tx: Some(tx),
|
|
rx,
|
|
futures: Default::default(),
|
|
lifetime: PhantomData,
|
|
}
|
|
}
|
|
|
|
/// How many CPUs are available to the dispatcher.
|
|
pub fn num_cpus(&self) -> usize {
|
|
self.executor.num_cpus()
|
|
}
|
|
|
|
/// Spawn a future into this scope.
|
|
#[track_caller]
|
|
pub fn spawn<F>(&mut self, f: F)
|
|
where
|
|
F: Future<Output = ()> + Send + 'a,
|
|
{
|
|
let tx = self.tx.clone().unwrap();
|
|
|
|
// SAFETY: The 'a lifetime is guaranteed to outlive any of these futures because
|
|
// dropping this `Scope` blocks until all of the futures have resolved.
|
|
let f = unsafe {
|
|
mem::transmute::<
|
|
Pin<Box<dyn Future<Output = ()> + Send + 'a>>,
|
|
Pin<Box<dyn Future<Output = ()> + Send + 'static>>,
|
|
>(Box::pin(async move {
|
|
f.await;
|
|
drop(tx);
|
|
}))
|
|
};
|
|
self.futures.push(f);
|
|
}
|
|
}
|
|
|
|
impl Drop for Scope<'_> {
|
|
fn drop(&mut self) {
|
|
self.tx.take().unwrap();
|
|
|
|
// Wait until the channel is closed, which means that all of the spawned
|
|
// futures have resolved.
|
|
let future = async {
|
|
self.rx.next().await;
|
|
};
|
|
let mut future = std::pin::pin!(future);
|
|
self.executor
|
|
.inner
|
|
.scheduler()
|
|
.block(None, future.as_mut(), None);
|
|
}
|
|
}
|
|
|
|
#[cfg(test)]
|
|
mod test {
|
|
use super::*;
|
|
use crate::{App, TestDispatcher, TestPlatform};
|
|
use std::cell::RefCell;
|
|
|
|
/// Helper to create test infrastructure.
|
|
/// Returns (dispatcher, background_executor, app).
|
|
fn create_test_app() -> (TestDispatcher, BackgroundExecutor, Rc<crate::AppCell>) {
|
|
let dispatcher = TestDispatcher::new(0);
|
|
let arc_dispatcher = Arc::new(dispatcher.clone());
|
|
let background_executor = BackgroundExecutor::new(arc_dispatcher.clone());
|
|
let foreground_executor = ForegroundExecutor::new(arc_dispatcher);
|
|
|
|
let platform = TestPlatform::new(background_executor.clone(), foreground_executor);
|
|
let asset_source = Arc::new(());
|
|
let http_client = http_client::FakeHttpClient::with_404_response();
|
|
|
|
let app = App::new_app(platform, asset_source, http_client);
|
|
(dispatcher, background_executor, app)
|
|
}
|
|
|
|
#[test]
|
|
fn sanity_test_tasks_run() {
|
|
let (dispatcher, _background_executor, app) = create_test_app();
|
|
let foreground_executor = app.borrow().foreground_executor.clone();
|
|
|
|
let task_ran = Rc::new(RefCell::new(false));
|
|
|
|
foreground_executor
|
|
.spawn({
|
|
let task_ran = Rc::clone(&task_ran);
|
|
async move {
|
|
*task_ran.borrow_mut() = true;
|
|
}
|
|
})
|
|
.detach();
|
|
|
|
// Run dispatcher while app is still alive
|
|
dispatcher.run_until_parked();
|
|
|
|
// Task should have run
|
|
assert!(
|
|
*task_ran.borrow(),
|
|
"Task should run normally when app is alive"
|
|
);
|
|
}
|
|
}
|