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
|
/* Copyright (c) 2010 - 2021 Advanced Micro Devices, Inc.
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE. */
#ifndef CONCURRENT_HPP_
#define CONCURRENT_HPP_
#include "top.hpp"
#include "os/alloc.hpp"
#include <atomic>
#include <new>
//! \addtogroup Utils
namespace amd { /*@{*/
namespace details {
template <typename T, int N> struct TaggedPointerHelper {
static constexpr uintptr_t TagMask = (1u << N) - 1;
private:
TaggedPointerHelper(); // Cannot instantiate
void* operator new(size_t); // allocate or
void operator delete(void*); // delete a TaggedPointerHelper.
public:
//! Create a tagged pointer.
static TaggedPointerHelper* make(T* ptr, size_t tag) {
return reinterpret_cast<TaggedPointerHelper*>((reinterpret_cast<uintptr_t>(ptr) & ~TagMask) |
(tag & TagMask));
}
//! Return the pointer value.
T* ptr() { return reinterpret_cast<T*>(reinterpret_cast<uintptr_t>(this) & ~TagMask); }
//! Return the tag value.
size_t tag() const { return reinterpret_cast<uintptr_t>(this) & TagMask; }
};
} // namespace details
/*! \brief An unbounded thread-safe queue.
*
* This queue orders elements first-in-first-out. It is based on the algorithm
* "Simple, Fast, and Practical Non-Blocking and Blocking Concurrent Queue
* Algorithms by Maged M. Michael and Michael L. Scott.".
*
* FIXME_lmoriche: Implement the new/delete operators for SimplyLinkedNode
* using thread-local allocation buffers.
*/
template <typename T, int N = 5> class ConcurrentLinkedQueue : public HeapObject {
//! A simply-linked node
struct Node {
typedef details::TaggedPointerHelper<Node, N> TaggedPointerHelper;
typedef TaggedPointerHelper* Ptr;
T value_; //!< The value stored in that node.
std::atomic<Ptr> next_; //!< Pointer to the next node
//! Create a Node::Ptr
static inline Ptr ptr(Node* ptr, size_t counter = 0) {
return TaggedPointerHelper::make(ptr, counter);
}
};
private:
std::atomic<typename Node::Ptr> head_; //! Pointer to the oldest element.
std::atomic<typename Node::Ptr> tail_; //! Pointer to the most recent element.
private:
//! \brief Allocate a free node.
static inline Node* allocNode() {
return new (AlignedMemory::allocate(sizeof(Node), 1 << N)) Node();
}
//! \brief Return a node to the free list.
static inline void reclaimNode(Node* node) { AlignedMemory::deallocate(node); }
public:
//! \brief Initialize a new concurrent linked queue.
ConcurrentLinkedQueue();
//! \brief Destroy this concurrent linked queue.
~ConcurrentLinkedQueue();
//! \brief Enqueue an element to this queue.
inline void enqueue(T elem);
//! \brief Dequeue an element from this queue.
inline T dequeue();
//! \brief Check if queue is empty
inline bool empty();
};
/*@}*/
template <typename T, int N> inline ConcurrentLinkedQueue<T, N>::ConcurrentLinkedQueue() {
// Create the first "dummy" node.
Node* dummy = allocNode();
dummy->next_ = NULL;
DEBUG_ONLY(dummy->value_ = NULL);
// Head and tail should now point to it (empty list).
head_ = tail_ = Node::ptr(dummy);
// Make sure the instance is fully initialized before it becomes
// globally visible.
std::atomic_thread_fence(std::memory_order_release);
}
template <typename T, int N> inline ConcurrentLinkedQueue<T, N>::~ConcurrentLinkedQueue() {
typename Node::Ptr head = head_;
typename Node::Ptr tail = tail_;
while (head->ptr() != tail->ptr()) {
Node* node = head->ptr();
head = head->ptr()->next_;
reclaimNode(node);
}
reclaimNode(head->ptr());
}
template <typename T, int N> inline void ConcurrentLinkedQueue<T, N>::enqueue(T elem) {
Node* node = allocNode();
node->value_ = elem;
node->next_ = NULL;
for (;;) {
typename Node::Ptr tail = tail_.load(std::memory_order_acquire);
typename Node::Ptr next = tail->ptr()->next_.load(std::memory_order_acquire);
if (likely(tail == tail_.load(std::memory_order_acquire))) {
if (next->ptr() == NULL) {
if (tail->ptr()->next_.compare_exchange_weak(next, Node::ptr(node, next->tag() + 1),
std::memory_order_acq_rel,
std::memory_order_acquire)) {
tail_.compare_exchange_strong(tail, Node::ptr(node, tail->tag() + 1),
std::memory_order_acq_rel, std::memory_order_acquire);
return;
}
} else {
tail_.compare_exchange_strong(tail, Node::ptr(next->ptr(), tail->tag() + 1),
std::memory_order_acq_rel, std::memory_order_acquire);
}
}
}
}
template <typename T, int N> inline T ConcurrentLinkedQueue<T, N>::dequeue() {
for (;;) {
typename Node::Ptr head = head_.load(std::memory_order_acquire);
typename Node::Ptr tail = tail_.load(std::memory_order_acquire);
typename Node::Ptr next = head->ptr()->next_.load(std::memory_order_acquire);
if (likely(head == head_.load(std::memory_order_acquire))) {
if (head->ptr() == tail->ptr()) {
if (next->ptr() == NULL) {
return NULL;
}
tail_.compare_exchange_strong(tail, Node::ptr(next->ptr(), tail->tag() + 1),
std::memory_order_acq_rel, std::memory_order_acquire);
} else {
T value = next->ptr()->value_;
if (head_.compare_exchange_weak(head, Node::ptr(next->ptr(), head->tag() + 1),
std::memory_order_acq_rel, std::memory_order_acquire)) {
// we can reclaim head now
reclaimNode(head->ptr());
return value;
}
}
}
}
}
template <typename T, int N> inline bool ConcurrentLinkedQueue<T, N>::empty() {
for (;;) {
typename Node::Ptr head = head_.load(std::memory_order_acquire);
typename Node::Ptr tail = tail_.load(std::memory_order_acquire);
typename Node::Ptr next = head->ptr()->next_.load(std::memory_order_acquire);
if (likely(head == head_.load(std::memory_order_acquire))) {
if (head->ptr() == tail->ptr()) {
if (next->ptr() == NULL) {
return true;
}
}
return false;
}
}
}
} // namespace amd
#endif /*CONCURRENT_HPP_*/
|
