mmtk/util/rust_util/
mod.rs

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
//! This module works around limitations of the Rust programming language, and provides missing
//! functionalities that we may expect the Rust programming language and its standard libraries
//! to provide.

pub mod atomic_box;
pub mod rev_group;
pub mod zeroed_alloc;

/// Const function for min value of two usize numbers.
pub const fn min_of_usize(a: usize, b: usize) -> usize {
    if a > b {
        b
    } else {
        a
    }
}

#[cfg(feature = "nightly")]
pub use core::intrinsics::{likely, unlikely};

// likely() and unlikely() compiler hints in stable Rust
// [1]: https://github.com/rust-lang/hashbrown/blob/a41bd76de0a53838725b997c6085e024c47a0455/src/raw/mod.rs#L48-L70
// [2]: https://users.rust-lang.org/t/compiler-hint-for-unlikely-likely-for-if-branches/62102/3
#[cfg(not(feature = "nightly"))]
#[inline]
#[cold]
fn cold() {}

#[cfg(not(feature = "nightly"))]
#[inline]
pub fn likely(b: bool) -> bool {
    if !b {
        cold();
    }
    b
}
#[cfg(not(feature = "nightly"))]
#[inline]
pub fn unlikely(b: bool) -> bool {
    if b {
        cold();
    }
    b
}

use std::cell::UnsafeCell;
use std::mem::MaybeUninit;
use std::sync::Once;

/// InitializeOnce creates an uninitialized value that needs to be manually initialized later. InitializeOnce
/// guarantees the value is only initialized once. This type is used to allow more efficient reads.
/// Unlike the `lazy_static!` which checks whether the static is initialized
/// in every read, InitializeOnce has no extra check for reads.
pub struct InitializeOnce<T: 'static> {
    v: UnsafeCell<MaybeUninit<T>>,
    /// This is used to guarantee `init_fn` is only called once.
    once: Once,
}

impl<T> InitializeOnce<T> {
    pub const fn new() -> Self {
        InitializeOnce {
            v: UnsafeCell::new(MaybeUninit::uninit()),
            once: Once::new(),
        }
    }

    /// Initialize the value. This should be called before ever using the struct.
    /// If this method is called by multiple threads, the first thread will
    /// initialize the value, and the other threads will be blocked until the
    /// initialization is done (`Once` returns).
    pub fn initialize_once(&self, init_fn: &'static dyn Fn() -> T) {
        self.once.call_once(|| {
            unsafe { &mut *self.v.get() }.write(init_fn());
        });
        debug_assert!(self.once.is_completed());
    }

    /// Get the value. This should only be used after initialize_once()
    pub fn get_ref(&self) -> &T {
        // We only assert in debug builds.
        debug_assert!(self.once.is_completed());
        unsafe { (*self.v.get()).assume_init_ref() }
    }

    /// Get a mutable reference to the value.
    /// This is currently only used for SFTMap during plan creation (single threaded),
    /// and before the plan creation is done, the binding cannot use MMTK at all.
    ///
    /// # Safety
    /// The caller needs to make sure there is no race when mutating the value.
    #[allow(clippy::mut_from_ref)]
    pub unsafe fn get_mut(&self) -> &mut T {
        // We only assert in debug builds.
        debug_assert!(self.once.is_completed());
        unsafe { (*self.v.get()).assume_init_mut() }
    }
}

impl<T> std::ops::Deref for InitializeOnce<T> {
    type Target = T;
    fn deref(&self) -> &Self::Target {
        self.get_ref()
    }
}

unsafe impl<T> Sync for InitializeOnce<T> {}

/// Create a formatted string that makes the best effort idenfying the current process and thread.
pub fn debug_process_thread_id() -> String {
    let pid = unsafe { libc::getpid() };
    #[cfg(target_os = "linux")]
    {
        // `gettid()` is Linux-specific.
        let tid = unsafe { libc::gettid() };
        format!("PID: {}, TID: {}", pid, tid)
    }
    #[cfg(not(target_os = "linux"))]
    {
        // TODO: When we support other platforms, use platform-specific methods to get thread
        // identifiers.
        format!("PID: {}", pid)
    }
}

#[cfg(test)]
mod initialize_once_tests {
    use super::*;

    #[test]
    fn test_threads_compete_initialize() {
        use std::sync::atomic::AtomicUsize;
        use std::sync::atomic::Ordering;
        use std::thread;

        // Create multiple threads to initialize the same `InitializeOnce` value
        const N_THREADS: usize = 1000;
        // The test value
        static I: InitializeOnce<usize> = InitializeOnce::new();
        // Count how many times the function is called
        static INITIALIZE_COUNT: AtomicUsize = AtomicUsize::new(0);
        // The function to create initial value
        fn initialize_usize() -> usize {
            INITIALIZE_COUNT.fetch_add(1, Ordering::SeqCst);
            42
        }

        let mut threads = vec![];
        for _ in 1..N_THREADS {
            threads.push(thread::spawn(|| {
                I.initialize_once(&initialize_usize);
                // Every thread should see the value correctly initialized.
                assert_eq!(*I, 42);
            }));
        }
        threads.into_iter().for_each(|t| t.join().unwrap());

        // The initialize_usize should only be called once
        assert_eq!(INITIALIZE_COUNT.load(Ordering::SeqCst), 1);
    }
}