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
use crossbeam_channel::{bounded, Receiver};
use lazy_static::lazy_static;
use log::{error, trace};
use std::env;
use std::sync::atomic::{AtomicUsize, Ordering};
static WORKER_SPAWN_COUNTER: AtomicUsize = AtomicUsize::new(0);
lazy_static! {
static ref NUM_CPUS: usize = if let Ok(num) = env::var("BELLMAN_NUM_CPUS") {
if let Ok(num) = num.parse() {
num
} else {
num_cpus::get()
}
} else {
num_cpus::get()
};
static ref WORKER_SPAWN_MAX_COUNT: usize = *NUM_CPUS * 4;
pub static ref THREAD_POOL: rayon::ThreadPool = rayon::ThreadPoolBuilder::new()
.num_threads(*NUM_CPUS)
.build()
.unwrap();
pub static ref VERIFIER_POOL: rayon::ThreadPool = rayon::ThreadPoolBuilder::new()
.num_threads(NUM_CPUS.max(6))
.build()
.unwrap();
}
#[derive(Clone)]
pub struct Worker {}
impl Worker {
pub fn new() -> Worker {
Worker {}
}
pub fn log_num_cpus(&self) -> u32 {
log2_floor(*NUM_CPUS)
}
pub fn compute<F, R>(&self, f: F) -> Waiter<R>
where
F: FnOnce() -> R + Send + 'static,
R: Send + 'static,
{
let (sender, receiver) = bounded(1);
let thread_index = if THREAD_POOL.current_thread_index().is_some() {
THREAD_POOL.current_thread_index().unwrap()
} else {
0
};
let previous_count = WORKER_SPAWN_COUNTER.fetch_add(1, Ordering::SeqCst);
if previous_count > *WORKER_SPAWN_MAX_COUNT {
THREAD_POOL.install(move || {
trace!("[{}] switching to install to help clear backlog[current threads {}, threads requested {}]",
thread_index,
THREAD_POOL.current_num_threads(),
WORKER_SPAWN_COUNTER.load(Ordering::SeqCst));
let res = f();
sender.send(res).unwrap();
WORKER_SPAWN_COUNTER.fetch_sub(1, Ordering::SeqCst);
});
} else {
THREAD_POOL.spawn(move || {
let res = f();
sender.send(res).unwrap();
WORKER_SPAWN_COUNTER.fetch_sub(1, Ordering::SeqCst);
});
}
Waiter { receiver }
}
pub fn scope<'a, F, R>(&self, elements: usize, f: F) -> R
where
F: FnOnce(&rayon::Scope<'a>, usize) -> R + Send,
R: Send,
{
let chunk_size = if elements < *NUM_CPUS {
1
} else {
elements / *NUM_CPUS
};
THREAD_POOL.scope(|scope| f(scope, chunk_size))
}
}
pub struct Waiter<T> {
receiver: Receiver<T>,
}
impl<T> Waiter<T> {
pub fn wait(&self) -> T {
if THREAD_POOL.current_thread_index().is_some() {
error!("The wait call should never be done inside the worker thread pool");
debug_assert!(false);
}
self.receiver.recv().unwrap()
}
pub fn done(val: T) -> Self {
let (sender, receiver) = bounded(1);
sender.send(val).unwrap();
Waiter { receiver }
}
}
fn log2_floor(num: usize) -> u32 {
assert!(num > 0);
let mut pow = 0;
while (1 << (pow + 1)) <= num {
pow += 1;
}
pow
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_log2_floor() {
assert_eq!(log2_floor(1), 0);
assert_eq!(log2_floor(3), 1);
assert_eq!(log2_floor(4), 2);
assert_eq!(log2_floor(5), 2);
assert_eq!(log2_floor(6), 2);
assert_eq!(log2_floor(7), 2);
assert_eq!(log2_floor(8), 3);
}
}