Fast way : Synchronization with mutex and use of Atomic Boolean in C++

Asked Sep 28 '15 at 10:19

Active Sep 28 '15 at 10:25

Viewed 74 times

I have two threads as follows:

Method A

FunctionA()              // Thread 1
{
    mutexA.lock()
    // MapA modification
    mutexA.unlock();
}

FunctionB()               // Thread 2
{
    mutexA.lock()
    // MapA swapped
    mutexA.unlock()
}

Does the above mentioned mutex lock is fast for synchronization or the below logic is fast ? Let me know if the below logic will cause undefined behaviour

Method B

atomic_bool toggleBool = true;
FunctionA()                   // Thread 1
{
  if(GetLockStatus())
  {
  // MapA modified

  toggleBool = true;
  }
}

FunctionB()                 // Thread 2
{
  if(GetLockStatus())
  {
    //MapA Swap
    toggleBool = true;
  }
}

GetLockStatus()               //Common function to get status
{
   // Critical Section or Mutex Lock
   if(toggleBool)
   {
      toggleBool = false;
      return true;
   }
   else
   {
      return false;  
   }
}

Method A or Method B..? Which is faster with less overhead ?

edited Sep 28 '15 at 10:25

asked Sep 28 '15 at 10:19

Sel_va

1

Method one's broken (mismatched mutex lock/unlock). Method two is broken (TOC/TOU race). – Mat Sep 28 '15 at 10:22
2

Only care about speed if you notice you have a bottleneck and profiled to know what the bottleneck is. Otherwise go for readability and maintainability. – Some programmer dude Sep 28 '15 at 10:22
In **Method A**, did you mean `mutexA.unlock()` in `FunctionB`? – Paolo M Sep 28 '15 at 10:24
@Paolo M: Ya... Updated the question – Sel_va Sep 28 '15 at 10:25
5

btw: better to use lock_guard or some other RAII-wrapper instead of raw mutex in A. Imagine what happens when an exception is thrown in the middle of functionA or functionB. – alagner Sep 28 '15 at 10:31
You could always time them to see which is faster. – Galik Sep 28 '15 at 10:43

Fast way : Synchronization with mutex and use of Atomic Boolean in C++

0 Answers0