#include <pthread.h>
#include <iostream>

/*
  Programm zur parallelen Berechung des Skalarproduktes zweier Vektoren mit Hilfe von pthreads

  2) verbesserte Version mit lokalen Variablen für die Summer
 */

const int N = 1000;
const int P = 2;

double x[N], y[N];
double sum;

pthread_mutex_t lock;

void* parallel_sum (void * p_address)
{
    double local_sum = 0.0;
    int p = * (int*)p_address;
    for (int i=p*N/P; i<(p+1)*N/P; i++)
    {
       // update local sum
       local_sum += x[i] * y[i];
    }    

    // enter critical section
    pthread_mutex_lock(&lock);
    // update global sum
    sum += local_sum;
    // leave critical section
    pthread_mutex_unlock(&lock);
}

int main()
{
    // init vectors
    for (int i=0; i<N; i++)
    {
        x[i] = i;
        y[i] = N-i;
    }

    // initialize sum
    sum = 0.0;

    // thread variables
    pthread_t threads[P];
    int       thread_numbers[P];

    // mutex
    pthread_mutex_init(&lock, NULL);
    
    // start threads
    for (int i=0; i<P; i++)
    {
        thread_numbers[i] = i;
        pthread_create(& threads[i], NULL,
            parallel_sum, (void*) & thread_numbers[i]);
    }

    // wait for threads
    for (int i=0; i<P; i++)
    {
        pthread_join(threads[i], NULL);
    }

    // cleanup mutex
    pthread_mutex_destroy(&lock);

    std::cout << "Scalar Product: " << sum << std::endl;

    return 0;
}