#include <stdio.h>
#include <offload.h>
#ifdef _OPENMP
#include <omp.h>
#endif

#define REAL float


///////////////////axpy

/*#pragma offload_attribute(push, target(mic))
static REAL x[SIZE];
static REAL y[SIZE];
#pragma offload_attribute(pop)*/

//__attribute__((target(mic))) REAL *x;
//__attribute__((target(mic))) REAL *y;
REAL *x;
REAL *y;
//
static void init(REAL *A, long n);

void f(REAL *x, REAL *y, int nn, int s, int a){

int n,i;
#pragma offload target(mic:0) nocopy (x: length(s) alloc_if(0) free_if(0)) \
                                nocopy (y: length(s) alloc_if(0) free_if(0)) 
#pragma omp parallel for
    for(n=0; n<nn; n++)
    {
    for (i=0; i<s; i++)
        {
            y[i] = x[i] * a + y[i];
        }
    }
}

void axpy()
{

    int NTIMES 2000 
    int SIZE 500000
    int FACTOR 1 
    int i,n;
    int array_sent = 0;
    //REAL *x_host, *y_host;
    //init(x_host, SIZE);
    //init(y_host, SIZE);
    init(x, SIZE);
    init(y, SIZE);
    REAL c = y[0];
    double start_timer = omp_get_wtime();
    
    double alloc_time = omp_get_wtime();
    #pragma offload target(mic:0) nocopy (x: length(SIZE) alloc_if(1) free_if(0)) \
    				nocopy (y: length(SIZE) alloc_if(1) free_if(0)) 
    {
    }
    alloc_time = omp_get_wtime() - alloc_time;

    double copy_time = omp_get_wtime();
    #pragma offload target(mic:0) in (x: length(SIZE) alloc_if(0) free_if(0)) \
                                in (y: length(SIZE) alloc_if(0) free_if(0)) 
    {
    }
    copy_time = omp_get_wtime() - copy_time;
  
    double kernel_time = omp_get_wtime();
    /*#pragma offload target(mic:0) nocopy (x: length(SIZE) alloc_if(0) free_if(0)) \
    				nocopy (y: length(SIZE) alloc_if(0) free_if(0)) 
    //#pragma omp parallel for
    for(n=0; n<NTIMES; n++)
    {
    for (i=0; i<SIZE; i++)
	{
	    y[i] = x[i] * FACTOR + y[i];
	}
    }
	*/

	f(x,y,NTIMES,SIZE,FACTOR);
    kernel_time = omp_get_wtime() - kernel_time;
    double free_time = omp_get_wtime();
    #pragma offload target(mic:0) nocopy (x:  alloc_if(0) free_if(1)) \
    				out (y: length(SIZE) alloc_if(0) free_if(1)) 
    {
    }
    free_time = omp_get_wtime() - free_time;
    
    

    double walltime = omp_get_wtime() - start_timer;
    if(c != y[0]) printf("Copy back to host successfully!\n\n");
    printf("PASS axpy\n\n");
    printf("Alloc time = %.8f sec\n\n", alloc_time);
    printf("Copy time = %.8f sec\n\n", copy_time);
    printf("Kernel time = %.8f sec\n\n", kernel_time);
    printf("Free time = %.8f sec\n\n", free_time);
    printf("Total time = %.8f sec\n\n", walltime);
}

static void init(REAL *A, long n) {
    long i;
    for (i = 0; i < n; i++) {
        A[i] = ((REAL) (drand48()) + 13);
    }
}




int main(void)
{
    x = (REAL*)malloc(sizeof(REAL)*SIZE);
    y = (REAL*)malloc(sizeof(REAL)*SIZE);
    axpy();
    free(x);
    free(y);
}