GEMM_EVAL/matmul.c at main · mpgemm/GEMM_EVAL · 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
//------------------------------------------------------------------------------------
// Matrix multiplication code adapted from an SME programming example provided
// by Arm Ltd.
//------------------------------------------------------------------------------------

#include "matmul.h"

void matmul_ref(const uint64_t rows_l, const uint64_t cols_l,
                const uint64_t cols_r, const float *restrict input_left,
                const float *restrict input_right, float *restrict output) {
  for (uint64_t x = 0; x < rows_l; ++x) {
    for (uint64_t y = 0; y < cols_r; ++y) {
      float acc = 0.0f;

      for (uint64_t z = 0; z < cols_l; ++z) {
        acc += input_left[(x * cols_l) + z] * input_right[(z * cols_r) + y];
      }

      output[(x * cols_r) + y] = acc;
    }
  }
}


// uint64_t sme_cntw() {
//   uint64_t cnt;
//   asm volatile("  smstart sm         \n" // smstart sm
//                "  cntw %[res]\n"
//                "  smstop sm          \n" // smstop sm
//                : [res] "=r"(cnt)
//                :
//                : "p0", "p1", "p2", "p3", "p4", "p5", "p6", "p7", "p8", "p9",
//                  "p10", "p11", "p12", "p13", "p14", "p15", "z0", "z1", "z2",
//                  "z3", "z4", "z5", "z6", "z7", "z8", "z9", "z10", "z11", "z12",
//                  "z13", "z14", "z15", "z16", "z17", "z18", "z19", "z20", "z21",
//                  "z22", "z23", "z24", "z25", "z26", "z27", "z28", "z29", "z30",
//                  "z31");
//   return cnt;
// }

// void matmul_opt(const uint64_t rows_l, const uint64_t cols_l,
//                 const uint64_t cols_r, const float *restrict input_left,
//                 const float *restrict input_right, float *restrict output) {
//   asm volatile(
//       "  smstart sm         \n" // smstart sm
//       "  smstop sm          \n" // smstop sm
//       :
//       :
//       :
//   );
// }

// void preprocess_l(const uint64_t rows, const uint64_t cols,
//                   const float *restrict a, float *restrict a_mod) {
//   asm volatile(
//       "  smstart sm         \n" // smstart sm
//       "  smstop sm          \n" // smstop sm
//       :
//       :
//       :
//   );
// }