#include <stdio.h>
#include "stdlib.h"
#include "string.h"
#include "sys/time.h"

#define SIZE	(128*2048)
unsigned long long A[SIZE];
unsigned long long B[SIZE];
unsigned long long C[SIZE];
unsigned long long D[SIZE];


#pragma omp target device(fpga)  implements(simple_code)  copy_in(a,b,c)  copy_out(d)
extern void sw_compute(unsigned long long a[SIZE], unsigned long long b[SIZE], unsigned long long c[SIZE], unsigned long long d[SIZE])

int main(int argc, char **argv)
{
	unsigned char   *p1, *p2, *p3, *p4;
	int             i;

	p1 = (char *)A;
	p2 = (char *)B;
	p3 = (char *)C;
	p4 = (char *)D;
	for (i = 0; i < SIZE * sizeof(unsigned long long); i++) {
		*p1++ = i;
		*p2++ = i-1;
		*p3++ = i-2;
		*p4++ = i-3;
	}

#pragma omp task label(simple_code) input(A[0:SIZE-1],B[0:SIZE-1],C[0:SIZE-1]) output(D[0:SIZE-1])
	for (i=0;i<SIZE;i++) {
		D[i] = A[i] & B[i] | C[i];
	}

/* Alternativa:
   Este codigo generaria el anterior en un paso intermedio, por lo que
   es lo mismo.
	for (i=0;i<SIZE;i++) {
#pragma omp block nest(1) factor(SIZE)
#pragma omp task label(simple_code) input(A[i],B[i],C[i])  output(D[i])
		D[i] = A[i] & B[i] | C[i];
	}


*/

	return 0;
}