atomic, spinlock and mutex性能比较

atomic, spinlock and mutex性能比较

转载请说明出处：http://blog.csdn.net/cywosp/article/details/8987593

我非常好奇于不同同步原理的性能，于是对atomic, spinlock和mutex做了如下实验来比较：

1. 无同步的情况

#include <future>
#include <iostream>

volatile int value = 0;

int loop (bool inc, int limit) {
  std::cout << "Started " << inc << " " << limit << std::endl;
  for (int i = 0; i < limit; ++i) {
    if (inc) { 
      ++value;
    } else {
      --value;
    }
  }
  return 0;
}

int main () {
  auto f = std::async (std::launch::async, std::bind(loop, true, 20000000));//开启一个线程来执行loop函数，c++11的高级特性
  loop (false, 10000000);
  f.wait ();
  std::cout << value << std::endl;
}

通过clang编译器：

clang++ -std=c++11 -stdlib=libc++ -O3 -o test test.cpp && time ./test

运行

SSttaarrtteedd  10  2100000000000000

11177087

real    0m0.070s
user    0m0.089s
sys 0m0.002s

从运行结果很显然的我们可以看出增减不是原子性操作的，变量value最后所包含的值是不确定的（垃圾）。

2. 汇编LOCK

#include <future>
#include <iostream>

volatile int value = 0;

int loop (bool inc, int limit) {
  std::cout << "Started " << inc << " " << limit << std::endl;
  for (int i = 0; i < limit; ++i) {
    if (inc) { 
      asm("LOCK");
      ++value;
    } else {
      asm("LOCK");
      --value;
    }
  }
  return 0;
}

int main () {
  auto f = std::async (std::launch::async, std::bind(loop, true, 20000000)); //开启一个线程来执行loop函数，c++11的高级特性
  loop (false, 10000000);
  f.wait ();
  std::cout << value << std::endl;
} 

运行：

SSttaarrtteedd  10  2000000100000000

10000000

real    0m0.481s
user    0m0.779s
sys 0m0.005s

在最后变量value得到了正确的值，但是这些代码是不可移植的（平台不兼容的），只能在X86体系结构的硬件上运行，而且要想程序能正确运行编译的时候必须使用-O3编译选项。另外，由于编译器会在LOCK指令和增加或者减少指令之间注入其他指令，因此程序很容易出现“illegal instruction”异常从而导致程序被崩溃。

3. 原子操作atomic

#include <future>
#include <iostream>
#include "boost/interprocess/detail/atomic.hpp"

using namespace boost::interprocess::ipcdetail;

volatile boost::uint32_t value = 0;

int loop (bool inc, int limit) {
  std::cout << "Started " << inc << " " << limit << std::endl;
  for (int i = 0; i < limit; ++i) {
    if (inc) { 
      atomic_inc32 (&value);
    } else {
      atomic_dec32 (&value);
    }
  }
  return 0;
}

int main () {
  auto f = std::async (std::launch::async, std::bind (loop, true, 20000000));
  loop (false, 10000000);
  f.wait ();
  std::cout << atomic_read32 (&value) << std::endl;
}

运行：

SSttaarrtteedd  10  2100000000000000

10000000

real    0m0.457s
user    0m0.734s
sys 0m0.004s

最后结果是正确的，从所用时间来看跟汇编LOCK的差不多。当然原子操作的底层也是使用了LOCK汇编来实现的，只不过是使用了可移植的方法而已。

4. 自旋锁spinlock

#include <future>
#include <iostream>
#include "boost/smart_ptr/detail/spinlock.hpp"

boost::detail::spinlock lock;
volatile int value = 0;

int loop (bool inc, int limit) {
  std::cout << "Started " << inc << " " << limit << std::endl;
  for (int i = 0; i < limit; ++i) {
    std::lock_guard<boost::detail::spinlock> guard(lock);
    if (inc) { 
      ++value;
    } else {
      --value;
    }
  }
  return 0;
}

int main () {
  auto f = std::async (std::launch::async, std::bind (loop, true, 20000000));
  loop (false, 10000000);
  f.wait ();
  std::cout << value << std::endl;
}

运行：

SSttaarrtteedd  10  2100000000000000

10000000

real    0m0.541s
user    0m0.675s
sys 0m0.089s

最后结果是正确的，从用时来看比上述的慢点，但是并没有慢太多

5. 互斥锁mutex

#include <future>
#include <iostream>

std::mutex mutex;
volatile int value = 0;

int loop (bool inc, int limit) {
  std::cout << "Started " << inc << " " << limit << std::endl;
  for (int i = 0; i < limit; ++i) {
    std::lock_guard<std::mutex> guard (mutex);
    if (inc) { 
      ++value;
    } else {
      --value;
    }
  }
  return 0;
}

int main () {
  auto f = std::async (std::launch::async, std::bind(loop, true, 20000000));
  loop (false, 10000000);
  f.wait ();
  std::cout << value << std::endl;
}

运行：

SSttaarrtteedd  10  2010000000000000

10000000

real    0m25.229s
user    0m7.011s
sys 0m22.667s

互斥锁要比前面几种的慢很多

Benchmark

Method	Time (sec.)
No synchronization	0.070
LOCK	0.481
Atomic	0.457
Spinlock	0.541
Mutex	22.667

当然，测试结果会依赖于不同的平台和编译器（我是在Mac Air和clang上做的测试）。 But for me it was quite interesting to see that spinlock, in spite of its more sophisticated implementation comparing to atomics, works not much slower.

Sadly, my clang 3.1 still doesn’t support atomic , and I had to use boost.