blob: 2ecb225a81abb53ee6deab3fc6a8186db2e421a8 (
plain)
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
|
/* { dg-require-effective-target vect_int } */
#include <stdarg.h>
#include "tree-vect.h"
#define N 64
#define DOT2 -21856
signed char X[N] __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__)));
signed char Y[N] __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__)));
/* char->short->short dot product.
The dot-product pattern should be detected.
The reduction is currently not vectorized becaus of the signed->unsigned->signed
casts, since this patch:
2005-12-26 Kazu Hirata <kazu@codesourcery.com>
PR tree-optimization/25125
When the dot-product is detected, the loop should be vectorized on vect_sdot_qi
targets (targets that support dot-product of signed char).
This test would currently fail to vectorize on targets that support
dot-product of chars into an int accumulator.
Alternatively, the loop could also be vectorized as widening-mult + summation,
or with type-conversion support.
*/
__attribute__ ((noinline)) short
foo2(int len) {
int i;
short result = 0;
for (i=0; i<len; i++) {
result += (X[i] * Y[i]);
}
return result;
}
int main (void)
{
int i;
short dot2;
check_vect ();
for (i=0; i<N; i++) {
X[i] = i;
Y[i] = 64-i;
__asm__ volatile ("");
}
dot2 = foo2 (N);
if (dot2 != DOT2)
abort ();
return 0;
}
/* { dg-final { scan-tree-dump-times "vect_recog_dot_prod_pattern: detected" 1 "vect" { xfail *-*-* } } } */
/* { dg-final { scan-tree-dump-times "vect_recog_widen_mult_pattern: detected" 1 "vect" { target { ! vect_multiple_sizes } } } } */
/* { dg-final { scan-tree-dump-times "vect_recog_widen_mult_pattern: detected" 2 "vect" { target vect_multiple_sizes } } } */
/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { xfail *-*-* } } } */
/* { dg-final { cleanup-tree-dump "vect" } } */
|
