liuxiaolong
2021-07-20 58d904a328c0d849769b483e901a0be9426b8209
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
/*
 * 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)
 *
 * Copyright (c) 2011 Helge Bahmann
 * Copyright (c) 2013 Tim Blechmann
 * Copyright (c) 2014 Andrey Semashev
 */
/*!
 * \file   atomic/detail/core_ops_gcc_sync.hpp
 *
 * This header contains implementation of the \c core_operations template.
 */
 
#ifndef BOOST_ATOMIC_DETAIL_CORE_OPS_GCC_SYNC_HPP_INCLUDED_
#define BOOST_ATOMIC_DETAIL_CORE_OPS_GCC_SYNC_HPP_INCLUDED_
 
#include <cstddef>
#include <boost/memory_order.hpp>
#include <boost/atomic/detail/config.hpp>
#include <boost/atomic/detail/storage_traits.hpp>
#include <boost/atomic/detail/core_operations_fwd.hpp>
#include <boost/atomic/detail/extending_cas_based_arithmetic.hpp>
#include <boost/atomic/detail/type_traits/integral_constant.hpp>
#include <boost/atomic/detail/capabilities.hpp>
#include <boost/atomic/detail/header.hpp>
 
#ifdef BOOST_HAS_PRAGMA_ONCE
#pragma once
#endif
 
namespace boost {
namespace atomics {
namespace detail {
 
struct core_operations_gcc_sync_base
{
    static BOOST_CONSTEXPR_OR_CONST bool full_cas_based = false;
    static BOOST_CONSTEXPR_OR_CONST bool is_always_lock_free = true;
 
    static BOOST_FORCEINLINE void fence_before_store(memory_order order) BOOST_NOEXCEPT
    {
        if ((static_cast< unsigned int >(order) & static_cast< unsigned int >(memory_order_release)) != 0u)
            __sync_synchronize();
    }
 
    static BOOST_FORCEINLINE void fence_after_store(memory_order order) BOOST_NOEXCEPT
    {
        if (order == memory_order_seq_cst)
            __sync_synchronize();
    }
 
    static BOOST_FORCEINLINE void fence_after_load(memory_order order) BOOST_NOEXCEPT
    {
        if ((static_cast< unsigned int >(order) & (static_cast< unsigned int >(memory_order_acquire) | static_cast< unsigned int >(memory_order_consume))) != 0u)
            __sync_synchronize();
    }
};
 
template< std::size_t Size, bool Signed, bool Interprocess >
struct core_operations_gcc_sync :
    public core_operations_gcc_sync_base
{
    typedef typename storage_traits< Size >::type storage_type;
 
    static BOOST_CONSTEXPR_OR_CONST std::size_t storage_size = Size;
    static BOOST_CONSTEXPR_OR_CONST std::size_t storage_alignment = storage_traits< storage_size >::alignment;
    static BOOST_CONSTEXPR_OR_CONST bool is_signed = Signed;
    static BOOST_CONSTEXPR_OR_CONST bool is_interprocess = Interprocess;
 
    // In general, we cannot guarantee atomicity of plain loads and stores of anything larger than a single byte on
    // an arbitrary CPU architecture. However, all modern architectures seem to guarantee atomic loads and stores of
    // suitably aligned objects of up to a pointer size. For larger objects we should probably use intrinsics to guarantee
    // atomicity. If there appears an architecture where this doesn't hold, this threshold needs to be updated (patches are welcome).
    typedef atomics::detail::integral_constant< bool, storage_size <= sizeof(void*) > plain_stores_loads_are_atomic;
 
    static BOOST_FORCEINLINE void store(storage_type volatile& storage, storage_type v, memory_order order) BOOST_NOEXCEPT
    {
        store(storage, v, order, plain_stores_loads_are_atomic());
    }
 
    static BOOST_FORCEINLINE void store(storage_type volatile& storage, storage_type v, memory_order order, atomics::detail::true_type) BOOST_NOEXCEPT
    {
        fence_before_store(order);
        storage = v;
        fence_after_store(order);
    }
 
    static BOOST_FORCEINLINE void store(storage_type volatile& storage, storage_type v, memory_order order, atomics::detail::false_type) BOOST_NOEXCEPT
    {
        exchange(storage, v, order);
    }
 
    static BOOST_FORCEINLINE storage_type load(storage_type const volatile& storage, memory_order order) BOOST_NOEXCEPT
    {
        return load(storage, order, plain_stores_loads_are_atomic());
    }
 
    static BOOST_FORCEINLINE storage_type load(storage_type const volatile& storage, memory_order order, atomics::detail::true_type) BOOST_NOEXCEPT
    {
        storage_type v = storage;
        fence_after_load(order);
        return v;
    }
 
    static BOOST_FORCEINLINE storage_type load(storage_type const volatile& storage, memory_order, atomics::detail::false_type) BOOST_NOEXCEPT
    {
        // Note: don't use fetch_add or other arithmetics here since storage_type may not be an arithmetic type.
        storage_type expected = storage_type();
        storage_type desired = expected;
        // We don't care if CAS succeeds or not. If it does, it will just write the same value there was before.
        return __sync_val_compare_and_swap(const_cast< storage_type volatile* >(&storage), expected, desired);
    }
 
    static BOOST_FORCEINLINE storage_type fetch_add(storage_type volatile& storage, storage_type v, memory_order) BOOST_NOEXCEPT
    {
        return __sync_fetch_and_add(&storage, v);
    }
 
    static BOOST_FORCEINLINE storage_type fetch_sub(storage_type volatile& storage, storage_type v, memory_order) BOOST_NOEXCEPT
    {
        return __sync_fetch_and_sub(&storage, v);
    }
 
    static BOOST_FORCEINLINE storage_type exchange(storage_type volatile& storage, storage_type v, memory_order order) BOOST_NOEXCEPT
    {
        // GCC docs mention that not all architectures may support full exchange semantics for this intrinsic. However, GCC's implementation of
        // std::atomic<> uses this intrinsic unconditionally. We do so as well. In case if some architectures actually don't support this, we can always
        // add a check here and fall back to a CAS loop.
        if ((static_cast< unsigned int >(order) & static_cast< unsigned int >(memory_order_release)) != 0u)
            __sync_synchronize();
        return __sync_lock_test_and_set(&storage, v);
    }
 
    static BOOST_FORCEINLINE bool compare_exchange_strong(
        storage_type volatile& storage, storage_type& expected, storage_type desired, memory_order, memory_order) BOOST_NOEXCEPT
    {
        storage_type expected2 = expected;
        storage_type old_val = __sync_val_compare_and_swap(&storage, expected2, desired);
 
        if (old_val == expected2)
        {
            return true;
        }
        else
        {
            expected = old_val;
            return false;
        }
    }
 
    static BOOST_FORCEINLINE bool compare_exchange_weak(
        storage_type volatile& storage, storage_type& expected, storage_type desired, memory_order success_order, memory_order failure_order) BOOST_NOEXCEPT
    {
        return compare_exchange_strong(storage, expected, desired, success_order, failure_order);
    }
 
    static BOOST_FORCEINLINE storage_type fetch_and(storage_type volatile& storage, storage_type v, memory_order) BOOST_NOEXCEPT
    {
        return __sync_fetch_and_and(&storage, v);
    }
 
    static BOOST_FORCEINLINE storage_type fetch_or(storage_type volatile& storage, storage_type v, memory_order) BOOST_NOEXCEPT
    {
        return __sync_fetch_and_or(&storage, v);
    }
 
    static BOOST_FORCEINLINE storage_type fetch_xor(storage_type volatile& storage, storage_type v, memory_order) BOOST_NOEXCEPT
    {
        return __sync_fetch_and_xor(&storage, v);
    }
 
    static BOOST_FORCEINLINE bool test_and_set(storage_type volatile& storage, memory_order order) BOOST_NOEXCEPT
    {
        if ((static_cast< unsigned int >(order) & static_cast< unsigned int >(memory_order_release)) != 0u)
            __sync_synchronize();
        return !!__sync_lock_test_and_set(&storage, 1);
    }
 
    static BOOST_FORCEINLINE void clear(storage_type volatile& storage, memory_order order) BOOST_NOEXCEPT
    {
        __sync_lock_release(&storage);
        if (order == memory_order_seq_cst)
            __sync_synchronize();
    }
};
 
#if BOOST_ATOMIC_INT8_LOCK_FREE > 0
template< bool Signed, bool Interprocess >
struct core_operations< 1u, Signed, Interprocess > :
#if defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_1)
    public core_operations_gcc_sync< 1u, Signed, Interprocess >
#elif defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_2)
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 2u, Signed, Interprocess >, 1u, Signed >
#elif defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4)
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 4u, Signed, Interprocess >, 1u, Signed >
#elif defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_8)
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 8u, Signed, Interprocess >, 1u, Signed >
#else
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 16u, Signed, Interprocess >, 1u, Signed >
#endif
{
};
#endif
 
#if BOOST_ATOMIC_INT16_LOCK_FREE > 0
template< bool Signed, bool Interprocess >
struct core_operations< 2u, Signed, Interprocess > :
#if defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_2)
    public core_operations_gcc_sync< 2u, Signed, Interprocess >
#elif defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4)
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 4u, Signed, Interprocess >, 2u, Signed >
#elif defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_8)
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 8u, Signed, Interprocess >, 2u, Signed >
#else
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 16u, Signed, Interprocess >, 2u, Signed >
#endif
{
};
#endif
 
#if BOOST_ATOMIC_INT32_LOCK_FREE > 0
template< bool Signed, bool Interprocess >
struct core_operations< 4u, Signed, Interprocess > :
#if defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4)
    public core_operations_gcc_sync< 4u, Signed, Interprocess >
#elif defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_8)
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 8u, Signed, Interprocess >, 4u, Signed >
#else
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 16u, Signed, Interprocess >, 4u, Signed >
#endif
{
};
#endif
 
#if BOOST_ATOMIC_INT64_LOCK_FREE > 0
template< bool Signed, bool Interprocess >
struct core_operations< 8u, Signed, Interprocess > :
#if defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_8)
    public core_operations_gcc_sync< 8u, Signed, Interprocess >
#else
    public extending_cas_based_arithmetic< core_operations_gcc_sync< 16u, Signed, Interprocess >, 8u, Signed >
#endif
{
};
#endif
 
#if BOOST_ATOMIC_INT128_LOCK_FREE > 0
template< bool Signed, bool Interprocess >
struct core_operations< 16u, Signed, Interprocess > :
    public core_operations_gcc_sync< 16u, Signed, Interprocess >
{
};
#endif
 
} // namespace detail
} // namespace atomics
} // namespace boost
 
#include <boost/atomic/detail/footer.hpp>
 
#endif // BOOST_ATOMIC_DETAIL_CORE_OPS_GCC_SYNC_HPP_INCLUDED_