microblas/mmm.c at master · threadmaster/microblas · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
#ifdef __cplusplus
extern "C" {
#endif
    void mmm_( int *len,  double *a, double *b, double*c );
#ifdef __cplusplus
    }
#endif

/*  S E R I A L   C O D E  */

void mmm_( int *len,  double *a, double *b, double *c ){

    int i, j, k;
    int veclen = *len;
    int mod;

#ifdef STRIP8
    const int stride = 8;

    mod = veclen % stride;

    for (i=0; i<veclen; i++) {
        for (j=0; j<veclen; j++) {
            *(c+(i*veclen+j)) = 0.0;
            for (k=0;k<mod;k++){
                *(c+(i*veclen+j)) += *(a+(i*veclen+k)) * *(b+(k*veclen+j));
            }
            for (k=mod;k<veclen;k+=stride) {
                *(c+(i*veclen+j)) += *(a+(i*veclen+k  )) * *(b+( k   *veclen+j))
                                   + *(a+(i*veclen+k+1)) * *(b+((k+1)*veclen+j))
                                   + *(a+(i*veclen+k+2)) * *(b+((k+2)*veclen+j))
                                   + *(a+(i*veclen+k+3)) * *(b+((k+3)*veclen+j))
                                   + *(a+(i*veclen+k+4)) * *(b+((k+4)*veclen+j))
                                   + *(a+(i*veclen+k+5)) * *(b+((k+5)*veclen+j))
                                   + *(a+(i*veclen+k+6)) * *(b+((k+6)*veclen+j))
                                   + *(a+(i*veclen+k+7)) * *(b+((k+7)*veclen+j));
            }
        }
    }
#elif STRIP4
const int stride = 4;

mod = veclen % stride;

for (i=0; i<veclen; i++) {
    for (j=0; j<veclen; j++) {
       *(c+(i*veclen+j)) = 0.0;
       for (k=0;k<mod;k++){
          *(c+(i*veclen+j)) += *(a+(i*veclen+k)) * *(b+(k*veclen+j));
       }
       for (k=mod;k<veclen;k+=stride) {
          *(c+(i*veclen+j)) += *(a+(i*veclen+k  )) * *(b+( k   *veclen+j))
                             + *(a+(i*veclen+k+1)) * *(b+((k+1)*veclen+j))
                             + *(a+(i*veclen+k+2)) * *(b+((k+2)*veclen+j))
                             + *(a+(i*veclen+k+3)) * *(b+((k+3)*veclen+j));
       }
       }
    }

#else

// Normal Matrix Multiplication

    for (i=0; i<veclen; i++) {
        for (j=0; j<veclen; j++) {
            *(c+(i*veclen+j)) = 0.0;
            for (k=0;k<veclen;k++){
                *(c+(i*veclen+j)) += *(a+(i*veclen+k)) * *(b+(k*veclen+j));
            }
        }
    }
#endif
}