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
|
/******************************************************************************
* Copyright (c) Intel Corporation - All rights reserved. *
* This file is part of the LIBXSMM library. *
* *
* For information on the license, see the LICENSE file. *
* Further information: https://github.com/hfp/libxsmm/ *
* SPDX-License-Identifier: BSD-3-Clause *
******************************************************************************/
/* Hans Pabst (Intel Corp.)
******************************************************************************/
#include <libxsmm_source.h>
int main(/*int argc, char* argv[]*/) {
const size_t t = sizeof(double);
int flags = LIBXSMM_GEMM_FLAG_NONE, batchsize = 1000, m = 13, n = 5, k = 7, ki, i, j;
double *a = malloc(t * batchsize * m * k), *b = malloc(t * batchsize * k * n);
double *c = malloc(t * m * n), alpha = 1, beta = 1;
/* generates and dispatches a matrix multiplication kernel */
libxsmm_dmmfunction kernel = libxsmm_dmmdispatch(
m, n, k, NULL /*lda*/, NULL /*ldb*/, NULL /*ldc*/, &alpha, &beta, &flags, NULL /*prefetch*/);
assert(kernel);
for (i = 0; i < batchsize; ++i) { /* initialize input */
for (ki = 0; ki < k; ++ki) {
for (j = 0; j < m; ++j) a[i * j * ki] = ((double)1) / ((i + j + ki) % 25);
for (j = 0; j < n; ++j) b[i * j * ki] = ((double)7) / ((i + j + ki) % 75);
}
}
memset(c, 0, t * m * n);
/* kernel multiplies and accumulates matrices: C += Ai * Bi */
for (i = 0; i < batchsize; ++i) kernel(a + i * m * k, b + i * k * n, c);
free(a), free(b), free(c);
return 0;
}
|
