1 /* { dg-require-effective-target vect_float } */
2 /* { dg-additional-options "--param vect-max-peeling-for-alignment=0" } */
3
4 #include <stdarg.h>
5 #include "tree-vect.h"
6
7 #define N1 3001
8
9 #if VECTOR_BITS > 256
10 #define N2 (VECTOR_BITS / 32 + 2)
11 #else
12 #define N2 10
13 #endif
14
15 __attribute__ ((noinline)) int
16 main1 (float *pa)
17 {
18 int i;
19
20 for (i = 0; i < N1; i++)
21 {
22 pa[i] = 2.0;
23 }
24
25 /* check results: */
26 for (i = 0; i < N1; i++)
27 {
28 if (pa[i] != 2.0)
29 abort ();
30 }
31
32 #pragma GCC unroll 0
33 for (i = 1; i <= N2; i++)
34 {
35 pa[i] = 3.0;
36 }
37
38 /* check results: */
39 for (i = 1; i <= N2; i++)
40 {
41 if (pa[i] != 3.0)
42 abort ();
43 }
44
45 return 0;
46 }
47
48 int main (void)
49 {
50 int i;
51 float a[N1] __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__)));
52 float b[N1] __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__)));
53
54 check_vect ();
55
56 /* from bzip2: */
57 for (i = 0; i < N1; i++)
58 b[i] = i;
59 a[0] = 0;
60 for (i = 1; i <= 256; i++) a[i] = b[i-1];
61
62 /* check results: */
63 for (i = 1; i <= 256; i++)
64 {
65 if (a[i] != i-1)
66 abort ();
67 }
68 if (a[0] != 0)
69 abort ();
70
71 main1 (a);
72
73 return 0;
74 }
75
76 /* 2 loops vectorized in main1, 2 loops vectorized in main:
77 the first loop in main requires vectorization of conversions,
78 the second loop in main requires vectorization of misaligned load. */
79
80 /* main && main1 together: */
81 /* { dg-final { scan-tree-dump-times "vectorized 2 loops" 2 "vect" { target powerpc*-*-* i?86-*-* x86_64-*-* } } } */
82
83 /* in main1: */
84 /* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" { target !powerpc*-*-* !i?86-*-* !x86_64-*-* } } } */
85 /* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" { target { vect_no_align && { ! vect_hw_misalign } } } } } */
86
87 /* in main: */
88 /* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { target { vect_no_align && { ! vect_hw_misalign } } } } } */
89