1.     #ifndef MIC_DEV
2.     #define MIC_DEV 0
3.     #endif
4.      
5.     #include <stdio.h>
6.     #include <stdlib.h>
7.     #include <omp.h>
8.     #include <math.h>
9.     #include <sys/time.h>
10.    
11.    double getTime(void) {
12.    struct timeval t;
13.    gettimeofday(&t,0);
14.    return ((double)t.tv_sec + ((double)t.tv_usec / 1000000.0));
15.    }
16.    
17.    // ################################ 
18.    // openMP multiply matrices
19.    void doMult(int size, float (* restrict A)[size],
20.         float (* restrict B)[size], float (* restrict C)[size]
21.        ) {
22.    
23.    #pragma offload target(mic:MIC_DEV) \
24.        in(A:length(size*size)) in( B:length(size*size))    \
25.        out(C:length(size*size))
26.        {
27.    #pragma omp parallel for default(none) shared(C,size)
28.            for (int i = 0; i < size; ++i)
29.                for (int j = 0; j < size; ++j)
30.                    C[i][j] =0.f;
31.            // matrix multiplication.
32.    #pragma omp parallel for default(none) shared(A,B,C,size)
33.            for (int i = 0; i < size; ++i)
34.                for (int k = 0; k < size; ++k)
35.                    for (int j = 0; j < size; ++j)
36.                      C[i][j] += A[i][k] * B[k][j];
37.        }
38.    }
39.     
40.    // ################################ 
41.    int main(int argc, char *argv[]) {
42.        if(argc != 4) {
43.            fprintf(stderr,"Use: %s size nThreads iter\n",argv[0]);
44.            return -1;
45.        }
46.        int i,j,k;
47.        int size=atoi(argv[1]);
48.        int nThreads=atoi(argv[2]);
49.        int iter=atoi(argv[3]);
50.        omp_set_num_threads(nThreads);
51.        float (*restrict A)[size] = malloc(sizeof(float)*size*size);
52.        float (*restrict B)[size] = malloc(sizeof(float)*size*size);
53.        float (*restrict C)[size] = malloc(sizeof(float)*size*size);
54.        // init natrices A and B 
55.    #pragma omp parallel for default(none) shared(A,B,size) private(i,j,k)
56.        for (i = 0; i < size; ++i) {
57.            for (j = 0; j < size; ++j) {
58.                A[i][j] = (float)((rand() % 3)-1);
59.                B[i][j] = (float)((rand() % 3)-1);
60.            }
61.        }
62.        // warmup
63.        doMult(size, A,B,C);
64.        double startTime = getTime();
65.        for (int i=0; i < iter; i++) {
66.            doMult(size, A,B,C);
67.        }
68.        double endTime = getTime();
69.        double lapsed = endTime-startTime;
70.        lapsed /= iter;
71.      
72.        printf("%s nThrds %d size %d elapsedtime %g GFlop/s %g\n",
73.             argv[0], nThreads, size, lapsed, 2e-9*size*size*size/lapsed);
74.     
75.        free(A); free(B); free(C);
76.        return 0;
77.    }
78.