sync_cell_slice/
lib.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
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
262
263
264
265
266
267
268
269
/*
 * SPDX-FileCopyrightText: 2024 Sebastiano Vigna
 *
 * SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later
 */

#![doc = include_str!("../README.md")]

use std::cell::Cell;

/// A mutable memory location that is [`Sync`].
///
/// # Memory layout
///
/// `SyncCell<T>` has the same memory layout and caveats as [`Cell<T>`], but it
/// is [`Sync`] if `T` is. In particular, if [`Cell<T>`] has the same in-memory
/// representation as its inner type `T`, then `SyncCell<T>` has the same
///  in-memory representation as its inner type `T` (but the code does not rely
/// on this). `SyncCell<T>` is also [`Send`] if [`Cell<T>`] is [`Send`].
///
/// `SyncCell<T>` is useful when you need to share a mutable memory location
/// across threads, and you rely on the fact that the intended behavior will not
/// cause data races. For example, the content will be written once and then
/// read many times, in this order.
///
/// The main usage of `SyncCell<T>` is to be to able to write to different
/// locations of a slice in parallel, leaving the control of data races to the
/// user, without the access cost of an atomic variable. For this purpose,
/// `SyncCell` implements the [`as_slice_of_cells`](SyncCell::as_slice_of_cells)
/// method, which turns a mutable reference to `SyncCell<[T]>` into a reference
/// to `[SyncCell<T>]`, similarly to the [analogous method of
/// `Cell`](Cell::as_slice_of_cells).
///
/// Since this is the most common usage, the extension trait [`SyncSlice`] adds
/// to slices a method [`as_sync_slice`](SyncSlice::as_sync_slice) that turns a
/// mutable reference to a slice of `T` into a reference to a slice of
/// `SyncCell<T>`.
///
/// # Methods
///
/// `SyncCell<T>` painstakingly reimplements the methods of `Cell<T>` as unsafe,
/// since they rely on external synchronization mechanisms to avoid undefined
/// behavior.
///
/// `SyncCell` implements a few traits implemented by [`Cell`] by delegation for
/// convenience, but some, as [`Clone`] or [`PartialOrd`], cannot be implemented
/// because they would use unsafe methods.
///
/// # Safety
///
/// Multiple thread can read from and write to the same `SyncCell` at the same
/// time. It is responsibility of the user to ensure that there are no data
/// races, which would cause undefined behavior.
///
/// # Examples
///
/// In this example, you can see that `SyncCell` enables mutation across
/// threads:
///
/// ```
/// use sync_cell_slice::SyncCell;
/// use sync_cell_Slice::SyncSlice;
///
/// let mut x = 0;
/// let c = SyncCell::new(x);
///
/// let mut v = vec![1, 2, 3, 4];
/// let s = v.as_sync_slice();
///
/// std::thread::scope(|scope| {
///     scope.spawn(|| {
///         // You can use interior mutability in another thread
///         unsafe { c.set(5) };
///     });
///
///     scope.spawn(|| {
///         // You can use interior mutability in another thread
///         unsafe { s[0].set(5) };
///     });
///     scope.spawn(|| {
///         // You can use interior mutability in another thread
///         // on the same slice
///         unsafe { s[1].set(10) };
///     });
/// });
/// ```
///
/// In this example, we invert a permutation in parallel:
///
/// ```
/// use sync_cell_slice::SyncCell;
/// use sync_cell_slice::SyncSlice;
///
/// let mut perm = vec![0, 2, 3, 1];
/// let mut inv = vec![0; perm.len()];
/// let inv_sync = inv.as_sync_slice();
///
/// std::thread::scope(|scope| {
///     scope.spawn(|| { // Invert first half
///         for i in 0..2 {
///             unsafe { inv_sync[perm[i]].set(i) };
///         }
///     });
///
///     scope.spawn(|| { // Invert second half
///         for i in 2..perm.len() {
///             unsafe { inv_sync[perm[i]].set(i) };
///        }
///     });
/// });
///
/// assert_eq!(inv, vec![0, 3, 1, 2]);

#[repr(transparent)]
pub struct SyncCell<T: ?Sized>(Cell<T>);

// This is where we depart from Cell.
unsafe impl<T: ?Sized> Send for SyncCell<T> where Cell<T>: Send {}
unsafe impl<T: ?Sized + Sync> Sync for SyncCell<T> {}

impl<T: Default> Default for SyncCell<T> {
    /// Creates a `SyncCell<T>`, with the `Default` value for `T`.
    #[inline]
    fn default() -> SyncCell<T> {
        SyncCell::new(Default::default())
    }
}

impl<T> From<T> for SyncCell<T> {
    /// Creates a new `SyncCell` containing the given value.
    fn from(value: T) -> SyncCell<T> {
        SyncCell::new(value)
    }
}

impl<T> SyncCell<T> {
    /// Creates a new `SyncCell` containing the given value.
    #[inline]
    pub fn new(value: T) -> Self {
        Self(Cell::new(value))
    }

    /// Sets the contained value by delegation to [`Cell::set`]
    ///
    /// # Safety
    ///
    /// Multiple thread can read from and write to the same `SyncCell` at the
    /// same time. It is responsibility of the user to ensure that there are no
    /// data races, which would cause undefined behavior.
    #[inline]
    pub unsafe fn set(&self, val: T) {
        self.0.set(val);
    }

    /// Swaps the values of two `SyncCell`s by delegation to [`Cell::swap`].
    ///
    /// # Safety
    ///
    /// Multiple thread can read from and write to the same `SyncCell` at the
    /// same time. It is responsibility of the user to ensure that there are no
    /// data races, which would cause undefined behavior.
    #[inline]
    pub unsafe fn swap(&self, other: &SyncCell<T>) {
        self.0.swap(&other.0);
    }

    /// Replaces the contained value with `val`, and returns the old contained
    /// value by delegation to [`Cell::replace`].
    ///
    /// # Safety
    ///
    /// Multiple thread can read from and write to the same `SyncCell` at the
    /// same time. It is responsibility of the user to ensure that there are no
    /// data races, which would cause undefined behavior.
    #[inline]
    pub unsafe fn replace(&self, val: T) -> T {
        self.0.replace(val)
    }

    /// Unwraps the value, consuming the cell.
    #[inline]
    pub fn into_inner(self) -> T {
        self.0.into_inner()
    }
}

impl<T: Copy> SyncCell<T> {
    /// Returns a copy of the contained value by delegation to [`Cell::get`].
    ///
    /// # Safety
    ///
    /// Multiple thread can read from and write to the same `SyncCell` at the
    /// same time. It is responsibility of the user to ensure that there are no
    /// data races, which would cause undefined behavior.
    #[inline]
    pub unsafe fn get(&self) -> T {
        self.0.get()
    }
}

impl<T: ?Sized> SyncCell<T> {
    /// Returns a raw pointer to the underlying data in this cell
    /// by delegation to [`Cell::as_ptr`].
    ///
    /// # Safety
    ///
    /// Multiple thread can read from and write to the same `SyncCell` at the
    /// same time. It is responsibility of the user to ensure that there are no
    /// data races, which would cause undefined behavior.
    #[inline]
    pub const unsafe fn as_ptr(&self) -> *mut T {
        self.0.as_ptr()
    }

    /// Returns a mutable reference to the underlying data by delegation to
    /// [`Cell::get_mut`].
    #[inline]
    pub fn get_mut(&mut self) -> &mut T {
        self.0.get_mut()
    }

    /// Returns a `&SyncCell<T>` from a `&mut T`.
    #[allow(trivial_casts)]
    #[inline]
    pub fn from_mut(value: &mut T) -> &Self {
        // SAFETY: `SyncCell<T>` has the same memory layout as `Cell<T>`.
        unsafe { &*(Cell::from_mut(value) as *const Cell<T> as *const Self) }
    }
}

impl<T: Default> SyncCell<T> {
    /// Takes the value of the cell, leaving [`Default::default`] in its place.
    ///
    /// # Safety
    ///
    /// Multiple thread can read from and write to the same `SyncCell` at the
    /// same time. It is responsibility of the user to ensure that there are no
    /// data races, which would cause undefined behavior.
    #[inline]
    pub unsafe fn take(&self) -> T {
        self.0.replace(Default::default())
    }
}

#[allow(trivial_casts)]
impl<T> SyncCell<[T]> {
    /// Returns a `&[SyncCell<T>]` from a `&SyncCell<[T]>`.
    #[inline]
    pub fn as_slice_of_cells(&self) -> &[SyncCell<T>] {
        let slice_of_cells = self.0.as_slice_of_cells();
        // SAFETY: `SyncCell<T>` has the same memory layout as `Cell<T>`
        unsafe { &*(slice_of_cells as *const [Cell<T>] as *const [SyncCell<T>]) }
    }
}

/// Extension trait turning a mutable reference to a slice of `T` into a
/// reference to a slice of `SyncCell<T>`.
///
/// The resulting slice is `Sync` if `T` is `Sync`.
pub trait SyncSlice<T> {
    /// Returns a `&[SyncCell<T>]` from a `&mut [T]`.
    fn as_sync_slice(&mut self) -> &[SyncCell<T>];
}

impl<T> SyncSlice<T> for [T] {
    fn as_sync_slice(&mut self) -> &[SyncCell<T>] {
        SyncCell::from_mut(self).as_slice_of_cells()
    }
}