28
votes

I have an application, where some STL containers are read in 3 threads, and written in 2. I know there is TBB for multi-threaded containers, but it is not an option in my application.

So I want to make the program thread-safe using std::mutex and my bare hands. This is a simple version of what I did:

int readers = 0;
std::mutex write;

// One write, no reads.
void write_fun()
{
    write.lock();// We lock the resource
    while(readers > 0){}// We wait till everyone finishes read.
    // DO WRITE
    write.unlock();// Release
}

// Multiple reads, no write
void read_fun()
{
    // We wait if it is being written.
    while(!write.try_lock()){}
    write.unlock();

    readers++;
    // do read
    readers--;
}

Is this the correct way to do this in C++11?

2

2 Answers

45
votes

Pretty close, couple things to note, in c++ for exception safety and readability, IMO, it is good to use RAII locks. What you really need is a shared_mutex like in boost or coming in c++14.

std::shared_mutex write; //use boost's or c++14 

// One write, no reads.
void write_fun()
{
    std::lock_guard<std::shared_mutex> lock(write);
    // DO WRITE
}

// Multiple reads, no write
void read_fun()
{
    std::shared_lock<std::shared_mutex> lock(write);
    // do read
}

If you don't want to use boost @howardhinmant was do kind as to give a link to a reference implementation

8
votes

This is safe, but still likely not fair or performant:

std::atomic<int> readers;
std::mutex write;

// One write, no reads.
void write_fun()
{
    write.lock();// We lock the resource
    while(readers > 0){}// We wait till everyone finishes read.
    // DO WRITE
    write.unlock();// Release
}

// Multiple reads, no write
void read_fun()
{
    // We wait if it is being written.
    write.lock();
    readers++;
    write.unlock();

    // do read
    readers--;
}

A solution with condition variables could avoid busy waiting for readers to fall to 0, left as an exercise for the reader.