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
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
|
/** \file boost/atomic.hpp */
// Copyright (c) 2009 Helge Bahmann
//
// Distributed under the Boost Software License, Version 1.0.
// See accompanying file LICENSE_1_0.txt or copy at
// http://www.boost.org/LICENSE_1_0.txt)
/* this is just a pseudo-header file fed to doxygen
to more easily generate the class documentation; will
be replaced by proper documentation down the road */
namespace boost {
/**
\brief Memory ordering constraints
This defines the relative order of one atomic operation
and other memory operations (loads, stores, other atomic operations)
executed by the same thread.
The order of operations specified by the programmer in the
source code ("program order") does not necessarily match
the order in which they are actually executed on the target system:
Both compiler as well as processor may reorder operations
quite arbitrarily. <B>Specifying the wrong ordering
constraint will therefore generally result in an incorrect program.</B>
*/
enum memory_order {
/**
\brief No constraint
Atomic operation and other memory operations may be reordered freely.
*/
memory_order_relaxed,
/**
\brief Data dependence constraint
Atomic operation must strictly precede any memory operation that
computationally depends on the outcome of the atomic operation.
*/
memory_order_consume,
/**
\brief Acquire memory
Atomic operation must strictly precede all memory operations that
follow in program order.
*/
memory_order_acquire,
/**
\brief Release memory
Atomic operation must strictly follow all memory operations that precede
in program order.
*/
memory_order_release,
/**
\brief Acquire and release memory
Combines the effects of \ref memory_order_acquire and \ref memory_order_release
*/
memory_order_acq_rel,
/**
\brief Sequentially consistent
Produces the same result \ref memory_order_acq_rel, but additionally
enforces globally sequential consistent execution
*/
memory_order_seq_cst
};
/**
\brief Atomic datatype
An atomic variable. Provides methods to modify this variable atomically.
Valid template parameters are:
- integral data types (char, short, int, ...)
- pointer data types
- any other data type that has a non-throwing default
constructor and that can be copied via <TT>memcpy</TT>
Unless specified otherwise, any memory ordering constraint can be used
with any of the atomic operations.
*/
template<typename Type>
class atomic {
public:
/**
\brief Create uninitialized atomic variable
Creates an atomic variable. Its initial value is undefined.
*/
atomic();
/**
\brief Create an initialize atomic variable
\param value Initial value
Creates and initializes an atomic variable.
*/
explicit atomic(Type value);
/**
\brief Read the current value of the atomic variable
\param order Memory ordering constraint, see \ref memory_order
\return Current value of the variable
Valid memory ordering constraints are:
- @c memory_order_relaxed
- @c memory_order_consume
- @c memory_order_acquire
- @c memory_order_seq_cst
*/
Type load(memory_order order=memory_order_seq_cst) const;
/**
\brief Write new value to atomic variable
\param value New value
\param order Memory ordering constraint, see \ref memory_order
Valid memory ordering constraints are:
- @c memory_order_relaxed
- @c memory_order_release
- @c memory_order_seq_cst
*/
void store(Type value, memory_order order=memory_order_seq_cst);
/**
\brief Atomically compare and exchange variable
\param expected Expected old value
\param desired Desired new value
\param order Memory ordering constraint, see \ref memory_order
\return @c true if value was changed
Atomically performs the following operation
\code
if (variable==expected) {
variable=desired;
return true;
} else {
expected=variable;
return false;
}
\endcode
This operation may fail "spuriously", i.e. the state of the variable
is unchanged even though the expected value was found (this is the
case on architectures using "load-linked"/"store conditional" to
implement the operation).
The established memory order will be @c order if the operation
is successful. If the operation is unsuccessful, the
memory order will be
- @c memory_order_relaxed if @c order is @c memory_order_acquire ,
@c memory_order_relaxed or @c memory_order_consume
- @c memory_order_release if @c order is @c memory_order_acq_release
or @c memory_order_release
- @c memory_order_seq_cst if @c order is @c memory_order_seq_cst
*/
bool compare_exchange_weak(
Type &expected,
Type desired,
memory_order order=memory_order_seq_cst);
/**
\brief Atomically compare and exchange variable
\param expected Expected old value
\param desired Desired new value
\param success_order Memory ordering constraint if operation
is successful
\param failure_order Memory ordering constraint if operation is unsuccessful
\return @c true if value was changed
Atomically performs the following operation
\code
if (variable==expected) {
variable=desired;
return true;
} else {
expected=variable;
return false;
}
\endcode
This operation may fail "spuriously", i.e. the state of the variable
is unchanged even though the expected value was found (this is the
case on architectures using "load-linked"/"store conditional" to
implement the operation).
The constraint imposed by @c success_order may not be
weaker than the constraint imposed by @c failure_order.
*/
bool compare_exchange_weak(
Type &expected,
Type desired,
memory_order success_order,
memory_order failure_order);
/**
\brief Atomically compare and exchange variable
\param expected Expected old value
\param desired Desired new value
\param order Memory ordering constraint, see \ref memory_order
\return @c true if value was changed
Atomically performs the following operation
\code
if (variable==expected) {
variable=desired;
return true;
} else {
expected=variable;
return false;
}
\endcode
In contrast to \ref compare_exchange_weak, this operation will never
fail spuriously. Since compare-and-swap must generally be retried
in a loop, implementors are advised to prefer \ref compare_exchange_weak
where feasible.
The established memory order will be @c order if the operation
is successful. If the operation is unsuccessful, the
memory order will be
- @c memory_order_relaxed if @c order is @c memory_order_acquire ,
@c memory_order_relaxed or @c memory_order_consume
- @c memory_order_release if @c order is @c memory_order_acq_release
or @c memory_order_release
- @c memory_order_seq_cst if @c order is @c memory_order_seq_cst
*/
bool compare_exchange_strong(
Type &expected,
Type desired,
memory_order order=memory_order_seq_cst);
/**
\brief Atomically compare and exchange variable
\param expected Expected old value
\param desired Desired new value
\param success_order Memory ordering constraint if operation
is successful
\param failure_order Memory ordering constraint if operation is unsuccessful
\return @c true if value was changed
Atomically performs the following operation
\code
if (variable==expected) {
variable=desired;
return true;
} else {
expected=variable;
return false;
}
\endcode
In contrast to \ref compare_exchange_weak, this operation will never
fail spuriously. Since compare-and-swap must generally be retried
in a loop, implementors are advised to prefer \ref compare_exchange_weak
where feasible.
The constraint imposed by @c success_order may not be
weaker than the constraint imposed by @c failure_order.
*/
bool compare_exchange_strong(
Type &expected,
Type desired,
memory_order success_order,
memory_order failure_order);
/**
\brief Atomically exchange variable
\param value New value
\param order Memory ordering constraint, see \ref memory_order
\return Old value of the variable
Atomically exchanges the value of the variable with the new
value and returns its old value.
*/
Type exchange(Type value, memory_order order=memory_order_seq_cst);
/**
\brief Atomically add and return old value
\param operand Operand
\param order Memory ordering constraint, see \ref memory_order
\return Old value of the variable
Atomically adds operand to the variable and returns its
old value.
*/
Type fetch_add(Type operand, memory_order order=memory_order_seq_cst);
/**
\brief Atomically subtract and return old value
\param operand Operand
\param order Memory ordering constraint, see \ref memory_order
\return Old value of the variable
Atomically subtracts operand from the variable and returns its
old value.
This method is available only if \c Type is an integral type
or a non-void pointer type. If it is a pointer type,
@c operand is of type @c ptrdiff_t and the operation
is performed following the rules for pointer arithmetic
in C++.
*/
Type fetch_sub(Type operand, memory_order order=memory_order_seq_cst);
/**
\brief Atomically perform bitwise "AND" and return old value
\param operand Operand
\param order Memory ordering constraint, see \ref memory_order
\return Old value of the variable
Atomically performs bitwise "AND" with the variable and returns its
old value.
This method is available only if \c Type is an integral type
or a non-void pointer type. If it is a pointer type,
@c operand is of type @c ptrdiff_t and the operation
is performed following the rules for pointer arithmetic
in C++.
*/
Type fetch_and(Type operand, memory_order order=memory_order_seq_cst);
/**
\brief Atomically perform bitwise "OR" and return old value
\param operand Operand
\param order Memory ordering constraint, see \ref memory_order
\return Old value of the variable
Atomically performs bitwise "OR" with the variable and returns its
old value.
This method is available only if \c Type is an integral type.
*/
Type fetch_or(Type operand, memory_order order=memory_order_seq_cst);
/**
\brief Atomically perform bitwise "XOR" and return old value
\param operand Operand
\param order Memory ordering constraint, see \ref memory_order
\return Old value of the variable
Atomically performs bitwise "XOR" with the variable and returns its
old value.
This method is available only if \c Type is an integral type.
*/
Type fetch_xor(Type operand, memory_order order=memory_order_seq_cst);
/**
\brief Implicit load
\return Current value of the variable
The same as <tt>load(memory_order_seq_cst)</tt>. Avoid using
the implicit conversion operator, use \ref load with
an explicit memory ordering constraint.
*/
operator Type(void) const;
/**
\brief Implicit store
\param value New value
\return Copy of @c value
The same as <tt>store(value, memory_order_seq_cst)</tt>. Avoid using
the implicit conversion operator, use \ref store with
an explicit memory ordering constraint.
*/
Type operator=(Type v);
/**
\brief Atomically perform bitwise "AND" and return new value
\param operand Operand
\return New value of the variable
The same as <tt>fetch_and(operand, memory_order_seq_cst)&operand</tt>.
Avoid using the implicit bitwise "AND" operator, use \ref fetch_and
with an explicit memory ordering constraint.
*/
Type operator&=(Type operand);
/**
\brief Atomically perform bitwise "OR" and return new value
\param operand Operand
\return New value of the variable
The same as <tt>fetch_or(operand, memory_order_seq_cst)|operand</tt>.
Avoid using the implicit bitwise "OR" operator, use \ref fetch_or
with an explicit memory ordering constraint.
This method is available only if \c Type is an integral type.
*/
Type operator|=(Type operand);
/**
\brief Atomically perform bitwise "XOR" and return new value
\param operand Operand
\return New value of the variable
The same as <tt>fetch_xor(operand, memory_order_seq_cst)^operand</tt>.
Avoid using the implicit bitwise "XOR" operator, use \ref fetch_xor
with an explicit memory ordering constraint.
This method is available only if \c Type is an integral type.
*/
Type operator^=(Type operand);
/**
\brief Atomically add and return new value
\param operand Operand
\return New value of the variable
The same as <tt>fetch_add(operand, memory_order_seq_cst)+operand</tt>.
Avoid using the implicit add operator, use \ref fetch_add
with an explicit memory ordering constraint.
This method is available only if \c Type is an integral type
or a non-void pointer type. If it is a pointer type,
@c operand is of type @c ptrdiff_t and the operation
is performed following the rules for pointer arithmetic
in C++.
*/
Type operator+=(Type operand);
/**
\brief Atomically subtract and return new value
\param operand Operand
\return New value of the variable
The same as <tt>fetch_sub(operand, memory_order_seq_cst)-operand</tt>.
Avoid using the implicit subtract operator, use \ref fetch_sub
with an explicit memory ordering constraint.
This method is available only if \c Type is an integral type
or a non-void pointer type. If it is a pointer type,
@c operand is of type @c ptrdiff_t and the operation
is performed following the rules for pointer arithmetic
in C++.
*/
Type operator-=(Type operand);
/**
\brief Atomically increment and return new value
\return New value of the variable
The same as <tt>fetch_add(1, memory_order_seq_cst)+1</tt>.
Avoid using the implicit increment operator, use \ref fetch_add
with an explicit memory ordering constraint.
This method is available only if \c Type is an integral type
or a non-void pointer type. If it is a pointer type,
the operation
is performed following the rules for pointer arithmetic
in C++.
*/
Type operator++(void);
/**
\brief Atomically increment and return old value
\return Old value of the variable
The same as <tt>fetch_add(1, memory_order_seq_cst)</tt>.
Avoid using the implicit increment operator, use \ref fetch_add
with an explicit memory ordering constraint.
This method is available only if \c Type is an integral type
or a non-void pointer type. If it is a pointer type,
the operation
is performed following the rules for pointer arithmetic
in C++.
*/
Type operator++(int);
/**
\brief Atomically subtract and return new value
\return New value of the variable
The same as <tt>fetch_sub(1, memory_order_seq_cst)-1</tt>.
Avoid using the implicit increment operator, use \ref fetch_sub
with an explicit memory ordering constraint.
This method is available only if \c Type is an integral type
or a non-void pointer type. If it is a pointer type,
the operation
is performed following the rules for pointer arithmetic
in C++.
*/
Type operator--(void);
/**
\brief Atomically subtract and return old value
\return Old value of the variable
The same as <tt>fetch_sub(1, memory_order_seq_cst)</tt>.
Avoid using the implicit increment operator, use \ref fetch_sub
with an explicit memory ordering constraint.
This method is available only if \c Type is an integral type
or a non-void pointer type. If it is a pointer type,
the operation
is performed following the rules for pointer arithmetic
in C++.
*/
Type operator--(int);
/** \brief Deleted copy constructor */
atomic(const atomic &) = delete;
/** \brief Deleted copy assignment */
const atomic & operator=(const atomic &) = delete;
};
/**
\brief Insert explicit fence for thread synchronization
\param order Memory ordering constraint
Inserts an explicit fence. The exact semantic depends on the
type of fence inserted:
- \c memory_order_relaxed: No operation
- \c memory_order_release: Performs a "release" operation
- \c memory_order_acquire or \c memory_order_consume: Performs an
"acquire" operation
- \c memory_order_acq_rel: Performs both an "acquire" and a "release"
operation
- \c memory_order_seq_cst: Performs both an "acquire" and a "release"
operation and in addition there exists a global total order of
all \c memory_order_seq_cst operations
*/
void atomic_thread_fence(memory_order order);
/**
\brief Insert explicit fence for synchronization with a signal handler
\param order Memory ordering constraint
Inserts an explicit fence to synchronize with a signal handler called within
the context of the same thread. The fence ensures the corresponding operations
around it are complete and/or not started. The exact semantic depends on the
type of fence inserted:
- \c memory_order_relaxed: No operation
- \c memory_order_release: Ensures the operations before the fence are complete
- \c memory_order_acquire or \c memory_order_consume: Ensures the operations
after the fence are not started.
- \c memory_order_acq_rel or \c memory_order_seq_cst: Ensures the operations
around the fence do not cross it.
Note that this call does not affect visibility order of the memory operations
to other threads. It is functionally similar to \c atomic_thread_fence, only
it does not generate any instructions to synchronize hardware threads.
*/
void atomic_signal_fence(memory_order order);
}
|