// OpenMP Scheduling example SJ

#include <stdlib.h>
#include <stdio.h>
#include <sys/time.h>
#include <time.h>
#include <omp.h>

int dummy_seq(int N);
int dummy_par1(int N);
int dummy_par2(int N);
int dummy_par3(int N);
double ltime();

int main(int argc, char *argv[]) {
	int N = 100000;
	int P = 1;
	int res = 0;
    double starttime, endtime, seqtime, partime;

    if (argc > 1)
        N = atoi(argv[1]);

    P = omp_get_num_procs();

	// wake up processors to full speed
	res = dummy_seq(N);
	res += dummy_seq(N);
	res += dummy_seq(N);
	printf("%d\n", res);

	starttime = ltime();
	res = dummy_seq(N);
	endtime = ltime();
    seqtime = endtime-starttime;
	printf("Sequential %d, N=%d, time = %.6lf s, %6.3lf Giter/s\n", res, N,
           seqtime, ((double)N*N/(200*seqtime))/1000000000);

	starttime = ltime();
	res = dummy_par1(N);
	endtime = ltime();
    partime = endtime-starttime;
	printf("Parallel1  %d, N=%d, time = %.6lf s, %6.3lf Giter/s\n", res, N,
		   partime, ((double)N*N/(200*partime))/1000000000);
	printf("Speedup %6.2f, P = %d, efficiency = %4.2f\n", seqtime/partime, P,
		   (seqtime/partime)/P);

	starttime = ltime();
	res = dummy_par2(N);
	endtime = ltime();
    partime = endtime-starttime;
	printf("Parallel2  %d, N=%d, time = %.6lf s, %6.3lf Giter/s\n", res, N,
		   partime, ((double)N*N/(200*partime))/1000000000);
	printf("Speedup %6.2f, P = %d, efficiency = %4.2f\n", seqtime/partime, P,
		   (seqtime/partime)/P);

	starttime = ltime();
	res = dummy_par3(N);
	endtime = ltime();
    partime = endtime-starttime;
	printf("Parallel3  %d, N=%d, time = %.6lf s, %6.3lf Giter/s\n", res, N,
		   partime, ((double)N*N/(200*partime))/1000000000);
	printf("Speedup %6.2f, P = %d, efficiency = %4.2f\n", seqtime/partime, P,
		   (seqtime/partime)/P);

}


int dummy_seq(int N) {
	int i, j, s = 0;
	for (i = 0; i < N; i++) {
		for (j = 0; j < i/100; j++) {
			if (j%2 || i%2) s++;	// some extra logic to avoid too
		}							// powerful optimizer
	}
	return s;
}

int dummy_par1(int N) {
	int i, j, s = 0, s2 = 0;
	#pragma omp parallel private(s2)
	  {
		s2 = 0;
		#pragma omp for
		for (i = 0; i < N; i++) {
			for (j = 0; j < i/100; j++) {
				if (j%2 || i%2) s2++;
			}
		}

		#pragma omp critical
		s += s2;
	  }
	return s;
}

int dummy_par2(int N) {
	int i, j, s = 0, s2 = 0;
	#pragma omp parallel private(s2)
	  {
		s2 = 0;
		#pragma omp for schedule(static,10)
		for (i = 0; i < N; i++) {
			for (j = 0; j < i/100; j++) {
				if (j%2 || i%2) s2++;
			}
		}

		#pragma omp critical
		s += s2;
	  }

	return s;
}

int dummy_par3(int N) {
	int i, j, s = 0, s2 = 0;
	#pragma omp parallel private(s2)
	  {
		s2 = 0;
		#pragma omp for schedule(guided)
		for (i = 0; i < N; i++) {
			for (j = 0; j < i/100; j++) {
				if (j%2 || i%2) s2++;
			}
		}

		#pragma omp critical
		s += s2;
	  }

	return s;
}

double ltime()
{
    struct timeval tv;
    struct timezone tz;
    gettimeofday(&tv, &tz);
    return tv.tv_sec + (double)tv.tv_usec/1000000;
}