Clean up benchmarks build
[riscv-tests.git] / mt / ay_matmul.c
1 #include "stdlib.h"
2
3 #include "util.h"
4
5 #include "dataset.h"
6 void __attribute__((noinline)) matmul(const int coreid, const int ncores, const int lda, const data_t A[], const data_t B[], data_t C[] )
7 {
8 if(coreid > 1) return;
9 static __thread int i, j, k;
10 static __thread data_t tempA0, tempA1, tempA2, tempA3, tempA4, tempA5, tempA6, tempA7;
11 static __thread data_t tempC0, tempC1, tempC2, tempC3, tempC4, tempC5, tempC6, tempC7, tempC8, tempC9, tempC10, tempC11, tempC12, tempC13, tempC14, tempC15;
12
13 static __thread int start, end, jStride, jToRow, jToCol;
14
15 start = coreid << 9;
16 end = ((ncores == 1) ? 2 : (coreid+1) ) << 9;
17 jStride = 8;
18
19 for (j=start; j < end; j+=jStride) {
20 jToRow = (j>>5)<<5;
21 jToCol = j%32;
22 tempC0 = 0;
23 tempC1 = 0;
24 tempC2 = 0;
25 tempC3 = 0;
26 tempC4 = 0;
27 tempC5 = 0;
28 tempC6 = 0;
29 tempC7 = 0;
30 for ( i=0; i < lda; i+=2 ) {
31 tempA0 = A[i + jToRow];
32 tempA1 = A[i+1 + jToRow];
33 tempC0 += tempA0 * B[(jToCol ) + (i<<5)];
34 tempC1 += tempA0 * B[(jToCol+1 ) + (i<<5)];
35 tempC2 += tempA0 * B[(jToCol+2 ) + (i<<5)];
36 tempC3 += tempA0 * B[(jToCol+3 ) + (i<<5)];
37 tempC4 += tempA0 * B[(jToCol+4 ) + (i<<5)];
38 tempC5 += tempA0 * B[(jToCol+5 ) + (i<<5)];
39 tempC6 += tempA0 * B[(jToCol+6 ) + (i<<5)];
40 tempC7 += tempA0 * B[(jToCol+7 ) + (i<<5)];
41 tempC0 += tempA1 * B[(jToCol ) + ((i+1)<<5)];
42 tempC1 += tempA1 * B[(jToCol+1 ) + ((i+1)<<5)];
43 tempC2 += tempA1 * B[(jToCol+2 ) + ((i+1)<<5)];
44 tempC3 += tempA1 * B[(jToCol+3 ) + ((i+1)<<5)];
45 tempC4 += tempA1 * B[(jToCol+4 ) + ((i+1)<<5)];
46 tempC5 += tempA1 * B[(jToCol+5 ) + ((i+1)<<5)];
47 tempC6 += tempA1 * B[(jToCol+6 ) + ((i+1)<<5)];
48 tempC7 += tempA1 * B[(jToCol+7 ) + ((i+1)<<5)];
49 }
50 C[j] =tempC0;
51 C[j + 1 ]=tempC1;
52 C[j + 2 ]=tempC2;
53 C[j + 3 ]=tempC3;
54 C[j + 4 ]=tempC4;
55 C[j + 5 ]=tempC5;
56 C[j + 6 ]=tempC6;
57 C[j + 7 ]=tempC7;
58 }
59
60 }