move join stuff to context, but should be moved to workerthread

This commit is contained in:
Janis 2025-06-20 13:03:02 +02:00
parent f6f8095440
commit a2112b9ef5

View file

@ -1081,6 +1081,7 @@ struct WorkerThread {
index: usize,
heartbeat: Arc<CachePadded<AtomicBool>>,
queue: UnsafeCell<JobList>,
join_count: Cell<u8>,
}
pub struct Scope<'scope> {
@ -1115,7 +1116,7 @@ impl WorkerThread {
index,
heartbeat,
queue: UnsafeCell::new(JobList::new()),
// join_count: Cell::new(0),
join_count: Cell::new(0),
// _pd: PhantomData,
}
}
@ -1505,109 +1506,28 @@ impl<'scope> Scope<'scope> {
A: FnOnce(&Self) -> RA + Send,
B: FnOnce(&Self) -> RB + Send,
{
self.join_heartbeat_every::<_, _, _, _, 64>(a, b)
// self.join_heartbeat(a, b)
}
pub fn join_seq<A, B, RA, RB>(&self, a: A, b: B) -> (RA, RB)
where
RA: Send,
RB: Send,
A: FnOnce(&Self) -> RA + Send,
B: FnOnce(&Self) -> RB + Send,
{
let rb = b(&self);
let ra = a(&self);
(ra, rb)
}
pub fn join_heartbeat_every<A, B, RA, RB, const TIMES: usize>(&self, a: A, b: B) -> (RA, RB)
where
RA: Send,
RB: Send,
A: FnOnce(&Self) -> RA + Send,
B: FnOnce(&Self) -> RB + Send,
{
thread_local! {
static JOIN_COUNT: Cell<usize> = Cell::new(0);
#[inline(always)]
fn make_scope_closure<'scope, A, RA>(
this: SendPtr<Scope<'scope>>,
a: A,
) -> impl FnOnce() -> RA + use<'scope, RA, A>
where
A: FnOnce(&Scope<'scope>) -> RA + Send,
RA: Send,
{
let scope = unsafe { this.as_ref() };
move || a(scope)
}
// a threadlocal counter is much faster than a sync atomic counter
let count = JOIN_COUNT.with(|count| {
count.set(count.get().wrapping_add(1) % TIMES);
count.get()
});
// let count = self.join_count.load(Ordering::Relaxed);
// self.join_count
// .store(count.wrapping_add(1) % TIMES, Ordering::Relaxed);
// let count = self
// .join_count
// .update(Ordering::Relaxed, Ordering::Relaxed, |n| {
// n.wrapping_add(1) % TIMES
// });
if count == 1 {
self.join_heartbeat(a, b)
} else {
self.join_seq(a, b)
}
}
pub fn join_heartbeat<A, B, RA, RB>(&self, a: A, b: B) -> (RA, RB)
where
RA: Send,
RB: Send,
A: FnOnce(&Self) -> RA + Send,
B: FnOnce(&Self) -> RB + Send,
{
let worker = WorkerThread::current_ref().expect("join is run in workerthread.");
let this = SendPtr::new_const(self).unwrap();
let a = StackJob::new(
move || unsafe {
WorkerThread::current_ref()
.expect("stackjob is run in workerthread.")
.tick();
let scope = this.as_ref();
a(scope)
},
NopLatch,
);
let job = a.as_job();
worker.push_front(&job);
use std::panic::{catch_unwind, resume_unwind, AssertUnwindSafe};
let rb = match catch_unwind(AssertUnwindSafe(|| b(self))) {
Ok(val) => val,
Err(payload) => {
cold_path();
// if b panicked, we need to wait for a to finish
worker.wait_until_job::<RA>(unsafe { job.transmute_ref::<RA>() });
resume_unwind(payload);
}
};
let ra = if job.state() == JobState::Empty as u8 {
unsafe {
job.unlink();
}
// a is allowed to panic here, because we already finished b.
unsafe { a.unwrap()() }
} else {
match worker.wait_until_job::<RA>(unsafe { job.transmute_ref::<RA>() }) {
Some(t) => t.into_result(), // propagate panic here
None => unsafe { a.unwrap()() },
}
};
drop(a);
(ra, rb)
unsafe {
worker.context.join_heartbeat_every::<_, _, _, _, 64>(
make_scope_closure(this, a),
make_scope_closure(this, b),
)
}
}
fn from_context(ctx: Arc<Context>) -> Self {
@ -1629,7 +1549,19 @@ where
A: FnOnce() -> RA + Send,
B: FnOnce() -> RB + Send,
{
Scope::scope(|scope| scope.join(|_| a(), |_| b()))
join_in(Context::global().clone(), a, b)
}
/// run two closures potentially in parallel, in the global threadpool.
#[allow(dead_code)]
pub fn join_in<A, B, RA, RB>(context: Arc<Context>, a: A, b: B) -> (RA, RB)
where
RA: Send,
RB: Send,
A: FnOnce() -> RA + Send,
B: FnOnce() -> RB + Send,
{
context.join(a, b)
}
pub struct ThreadPool {
@ -1649,6 +1581,16 @@ impl ThreadPool {
}
}
pub fn join<A, B, RA, RB>(&self, a: A, b: B) -> (RA, RB)
where
RA: Send,
RB: Send,
A: FnOnce() -> RA + Send,
B: FnOnce() -> RB + Send,
{
self.context.join(a, b)
}
pub fn scope<'scope, R, F>(&self, f: F) -> R
where
F: FnOnce(&Scope<'scope>) -> R + Send,
@ -1757,6 +1699,112 @@ impl Context {
self.shared_job.notify_one();
}
#[inline]
pub fn join<A, B, RA, RB>(self: &Arc<Self>, a: A, b: B) -> (RA, RB)
where
RA: Send,
RB: Send,
A: FnOnce() -> RA + Send,
B: FnOnce() -> RB + Send,
{
// SAFETY: join_heartbeat_every is safe to call from a worker thread.
self.run_in_worker(move |_| unsafe { self.join_heartbeat_every::<_, _, _, _, 64>(a, b) })
}
#[inline]
fn join_seq<A, B, RA, RB>(&self, a: A, b: B) -> (RA, RB)
where
RA: Send,
RB: Send,
A: FnOnce() -> RA + Send,
B: FnOnce() -> RB + Send,
{
let rb = b();
let ra = a();
(ra, rb)
}
/// This function must be called from a worker thread.
#[inline]
unsafe fn join_heartbeat_every<A, B, RA, RB, const TIMES: usize>(&self, a: A, b: B) -> (RA, RB)
where
RA: Send,
RB: Send,
A: FnOnce() -> RA + Send,
B: FnOnce() -> RB + Send,
{
let worker = WorkerThread::current_ref().expect("join is run in workerthread.");
// SAFETY: each worker is only ever used by one thread, so this is safe.
worker
.join_count
.set(worker.join_count.get().wrapping_add(1) % TIMES as u8);
// TODO: add counter to job queue, check for low job count to decide whether to use heartbeat or seq.
// see: chili
if worker.join_count.get() == 0 {
self.join_heartbeat(a, b)
} else {
self.join_seq(a, b)
}
}
/// This function must be called from a worker thread.
unsafe fn join_heartbeat<A, B, RA, RB>(&self, a: A, b: B) -> (RA, RB)
where
RA: Send,
RB: Send,
A: FnOnce() -> RA + Send,
B: FnOnce() -> RB + Send,
{
let worker = WorkerThread::current_ref().expect("join is run in workerthread.");
let a = StackJob::new(
move || {
// TODO: bench whether tick'ing here is good.
// turns out this actually costs a lot of time, likely because of the thread local check.
// WorkerThread::current_ref()
// .expect("stackjob is run in workerthread.")
// .tick();
a()
},
NopLatch,
);
let job = a.as_job();
worker.push_front(&job);
use std::panic::{catch_unwind, resume_unwind, AssertUnwindSafe};
let rb = match catch_unwind(AssertUnwindSafe(|| b())) {
Ok(val) => val,
Err(payload) => {
cold_path();
// if b panicked, we need to wait for a to finish
worker.wait_until_job::<RA>(unsafe { job.transmute_ref::<RA>() });
resume_unwind(payload);
}
};
let ra = if job.state() == JobState::Empty as u8 {
unsafe {
job.unlink();
}
// a is allowed to panic here, because we already finished b.
unsafe { a.unwrap()() }
} else {
match worker.wait_until_job::<RA>(unsafe { job.transmute_ref::<RA>() }) {
Some(t) => t.into_result(), // propagate panic here
None => unsafe { a.unwrap()() },
}
};
drop(a);
(ra, rb)
}
/// Runs closure in this context, processing the other context's worker's jobs while waiting for the result.
fn run_in_worker_cross<T, F>(self: &Arc<Self>, worker: &WorkerThread, f: F) -> T
where