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
|
// Copyright 2021 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include <atomic>
#include <string>
#include "base/barrier_closure.h"
#include "base/functional/callback.h"
#include "base/memory/raw_ptr.h"
#include "base/synchronization/lock.h"
#include "base/synchronization/waitable_event.h"
#include "base/threading/simple_thread.h"
#include "testing/gtest/include/gtest/gtest.h"
#include "testing/perf/perf_result_reporter.h"
// This file contains tests to measure the cost of incrementing:
// - A non-atomic variable, no lock.
// - A non-atomic variable, with lock.
// - An atomic variable, no memory barriers.
// - An atomic variable, acquire-release barriers.
// The goal is to provide data to guide counter implementation choices.
namespace base {
namespace {
constexpr char kMetricPrefixCounter[] = "Counter.";
constexpr char kMetricOperationThroughput[] = "operation_throughput";
constexpr uint64_t kNumIterations = 100000000;
perf_test::PerfResultReporter SetUpReporter(const std::string& story_name) {
perf_test::PerfResultReporter reporter(kMetricPrefixCounter, story_name);
reporter.RegisterImportantMetric(kMetricOperationThroughput, "operations/ms");
return reporter;
}
class Uint64_NoLock {
public:
Uint64_NoLock() = default;
void Increment() { counter_ = counter_ + 1; }
uint64_t value() const { return counter_; }
private:
// Volatile to prevent the compiler from over-optimizing the increment.
volatile uint64_t counter_ = 0;
};
class Uint64_Lock {
public:
Uint64_Lock() = default;
void Increment() {
AutoLock auto_lock(lock_);
++counter_;
}
uint64_t value() const {
AutoLock auto_lock(lock_);
return counter_;
}
private:
mutable Lock lock_;
uint64_t counter_ GUARDED_BY(lock_) = 0;
};
class AtomicUint64_NoBarrier {
public:
AtomicUint64_NoBarrier() = default;
void Increment() { counter_.fetch_add(1, std::memory_order_relaxed); }
uint64_t value() const { return counter_; }
private:
std::atomic<uint64_t> counter_{0};
};
class AtomicUint64_Barrier {
public:
AtomicUint64_Barrier() = default;
void Increment() { counter_.fetch_add(1, std::memory_order_acq_rel); }
uint64_t value() const { return counter_; }
private:
std::atomic<uint64_t> counter_{0};
};
template <typename CounterType>
class IncrementThread : public SimpleThread {
public:
// Upon entering its main function, the thread waits for |start_event| to be
// signaled. Then, it increments |counter| |kNumIterations| times.
// Finally, it invokes |done_closure|.
explicit IncrementThread(WaitableEvent* start_event,
CounterType* counter,
OnceClosure done_closure)
: SimpleThread("IncrementThread"),
start_event_(start_event),
counter_(counter),
done_closure_(std::move(done_closure)) {}
// SimpleThread:
void Run() override {
start_event_->Wait();
for (uint64_t i = 0; i < kNumIterations; ++i) {
counter_->Increment();
}
std::move(done_closure_).Run();
}
private:
const raw_ptr<WaitableEvent> start_event_;
const raw_ptr<CounterType> counter_;
OnceClosure done_closure_;
};
template <typename CounterType>
void RunIncrementPerfTest(const std::string& story_name, int num_threads) {
WaitableEvent start_event;
WaitableEvent end_event;
CounterType counter;
RepeatingClosure done_closure = BarrierClosure(
num_threads, BindOnce(&WaitableEvent::Signal, Unretained(&end_event)));
std::vector<std::unique_ptr<IncrementThread<CounterType>>> threads;
for (int i = 0; i < num_threads; ++i) {
threads.push_back(std::make_unique<IncrementThread<CounterType>>(
&start_event, &counter, done_closure));
threads.back()->Start();
}
TimeTicks start_time = TimeTicks::Now();
start_event.Signal();
end_event.Wait();
TimeTicks end_time = TimeTicks::Now();
EXPECT_EQ(num_threads * kNumIterations, counter.value());
auto reporter = SetUpReporter(story_name);
reporter.AddResult(
kMetricOperationThroughput,
kNumIterations / (end_time - start_time).InMillisecondsF());
for (auto& thread : threads) {
thread->Join();
}
}
} // namespace
TEST(CounterPerfTest, Uint64_NoLock_1Thread) {
RunIncrementPerfTest<Uint64_NoLock>("Uint64_NoLock_1Thread", 1);
}
// No Uint64_NoLock_4Threads test because it would cause data races.
TEST(CounterPerfTest, Uint64_Lock_1Thread) {
RunIncrementPerfTest<Uint64_Lock>("Uint64_Lock_1Thread", 1);
}
TEST(CounterPerfTest, Uint64_Lock_4Threads) {
RunIncrementPerfTest<Uint64_Lock>("Uint64_Lock_4Threads", 4);
}
TEST(CounterPerfTest, AtomicUint64_NoBarrier_1Thread) {
RunIncrementPerfTest<AtomicUint64_NoBarrier>("AtomicUint64_NoBarrier_1Thread",
1);
}
TEST(CounterPerfTest, AtomicUint64_NoBarrier_4Threads) {
RunIncrementPerfTest<AtomicUint64_NoBarrier>(
"AtomicUint64_NoBarrier_4Threads", 4);
}
TEST(CounterPerfTest, AtomicUint64_Barrier_1Thread) {
RunIncrementPerfTest<AtomicUint64_Barrier>("AtomicUint64_Barrier_1Thread", 1);
}
TEST(CounterPerfTest, AtomicUint64_Barrier_4Threads) {
RunIncrementPerfTest<AtomicUint64_Barrier>("AtomicUint64_Barrier_4Threads",
4);
}
} // namespace base
|
