// thread_loop.c  run 4 threads  n_thread,
//                each runs twice  nloop,
//                each has synchronized start  B1
//                and three computations  B2,B3,B4 for data transfer to main
//                np is number of cores needed, n_thread+1

#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#define n_thread  4
#define np        5
#define nloop     2
static int nb = 0;
static int results[n_thread];
static pthread_cond_t go;
static pthread_mutex_t bar;

void barrier(void)
{
  printf("in barrier \n");
  fflush(stdout);

  pthread_mutex_lock(&bar);
    nb++;
    printf("nb=%d of np=%d \n", nb, np);
    fflush(stdout);

    if(nb == np)
    {
      nb = 0;
      printf("about to broadcast \n");
      fflush(stdout);
      pthread_cond_broadcast(&go);
    }
    else
    {
      printf("about to cond wait \n");
      fflush(stdout);
      pthread_cond_wait(&go, &bar);      
    }
  pthread_mutex_unlock(&bar);
} // end barrier 

void * thread_compute(void * threadID)
{
  int i, iloop=0; // local variables in each thread 

  i = (int)threadID;
  printf("   thread %d at first barrier, iloop %d \n", i, iloop);
  fflush(stdout);
  barrier(); // B1 

  for(iloop=0; iloop<nloop; iloop++)
  {
    printf("   thread %d running after first barrier, iloop %d \n", i, iloop);
    fflush(stdout);
    results[i] = 10*(i+1)+iloop; // computation1 goes here
    // done this far 
    printf("   thread %d at second barrier, iloop %d \n", i, iloop);
    fflush(stdout);
    barrier(); // B2 

    printf("   thread %d running after second barrier, iloop %d \n", i, iloop);
    fflush(stdout);
    results[i] += 100*(i+1); // computation2 goes here
    // done this far 
    printf("   thread %d at third barrier, iloop= \n", i, iloop);
    fflush(stdout);
    barrier(); // B3 

    printf("   thread %d running after third barrier, iloop %d \n", i, iloop);
    fflush(stdout);
    results[i] += 1000*(i+1); // computation3 goes here
    // done 
    printf("   thread %d at fourth barrier, iloop %d \n", i, iloop);
    fflush(stdout);
    barrier(); // B4 
  } // end iloop 
  // pthread_exit(NULL); 
  return threadID;
}

int main(int argc, char * argv[])
{
  int rc, t;
  int errcode;
  int status;
  int iloop;
  pthread_t threads[n_thread];

  printf("pthread_loop.c running \n");
  fflush(stdout);

  pthread_mutex_init(&bar, NULL);
  pthread_mutex_unlock(&bar);
  pthread_cond_init(&go, NULL);

  for(t=0; t<n_thread; t++)
  {
    printf("creating thread %d \n", t);
    fflush(stdout);
    results[t] = 0;
    rc = pthread_create(&threads[t], NULL, thread_compute, (void *)t);
    if(rc)
    {
      printf("error, thread not created, code=%d, at t=%d \n", rc, t);
      fflush(stdout);
      exit(-1);
    }
  }
  printf("waiting for threads to reach first barrier \n");
  fflush(stdout);
  barrier(); // B1 wait for all threads to run 

  for(iloop=0; iloop<nloop; iloop++)
  {
    printf("all threads finished first task \n");
    fflush(stdout);
    for(t=0; t<n_thread; t++)
    {
      printf("results[%d]=%d \n", t, results[t]);
      fflush(stdout);
    }

    printf("wait for all threads to finish second task \n");
    fflush(stdout);
    barrier(); // B2 

    printf("all threads finished second task \n");
    fflush(stdout);
    for(t=0; t<n_thread; t++)
    {
      printf("final results[%d]=%d \n", t, results[t]);
      fflush(stdout);
    }

    printf("wait for all threads to finish third task \n");
    fflush(stdout);
    barrier(); // B3 

    printf("all threads finished third task \n");
    fflush(stdout);
    barrier(); // B4 wait for all at "done" 
    for(t=0; t<n_thread; t++)
    {
      printf("final results[%d]=%d \n", t, results[t]);
      fflush(stdout);
    }
  } // end iloop 

  // wait for all threads to exit and join
  for(t=0; t<n_thread; t++)
  {
    if (errcode=pthread_join(threads[t],(void *)&status)) 
      printf("compute &d failed join: %s\n", t, strerror(errcode));
    // check thread's exit status and release its resources 
    if (status != t)
      printf("compute %d terminated abnormally, status=%d \n", t, status);
  }

  printf("master exiting \n");
  fflush(stdout);
  return 0;
} // end thread_loop.c