Question

anyone has any idea how is Barrier Synchronization implemented internally in scala? I am guessing it is using Semaphores, but I am wondering how exactly it works, does anyone know?

I guess one naive way to think of this might be to create an array of Semaphores one for each process. Any better / more efficient implementations?

Answer 1

Ok, sorry for this guys, I have found it after all. here it is in case anyone is interested.

class Barrier(n:Int){
  assert(n>1);
  private var waiting = 0; // number of processes currently waiting
  private val waitSem = new Semaphore; waitSem.down
  private val mutex = new Semaphore;

  def sync = {
    mutex.down;
    if(waiting==n-1){ waitSem.up; }
    else{ 
      waiting+=1; mutex.up; waitSem.down; 
      // Wait until woken
      waiting-=1; 
      if(waiting==0) mutex.up; else waitSem.up;
    }
  }
}