|
145
|
1 /* Disabling epilogues until we find a better way to deal with scans. */
|
|
|
2 /* { dg-additional-options "--param vect-epilogues-nomask=0" } */
|
|
111
|
3 /* { dg-require-effective-target vect_int } */
|
|
|
4
|
|
|
5 #include <stdarg.h>
|
|
|
6 #include "tree-vect.h"
|
|
|
7
|
|
|
8 #define N 128
|
|
|
9
|
|
|
10 typedef struct {
|
|
|
11 int a;
|
|
|
12 int b;
|
|
|
13 int c;
|
|
|
14 int d;
|
|
|
15 int e;
|
|
|
16 int f;
|
|
|
17 int g;
|
|
|
18 int h;
|
|
|
19 } s;
|
|
|
20
|
|
|
21 int
|
|
|
22 main1 (s *arr)
|
|
|
23 {
|
|
|
24 int i;
|
|
|
25 s *ptr = arr;
|
|
|
26 s res[N];
|
|
|
27
|
|
|
28 for (i = 0; i < N; i++)
|
|
|
29 {
|
|
|
30 res[i].c = ptr->c + ptr->c;
|
|
|
31 res[i].a = ptr->a + ptr->a;
|
|
|
32 res[i].d = ptr->d + ptr->d;
|
|
|
33 res[i].b = ptr->b + ptr->b;
|
|
|
34 res[i].f = ptr->f + ptr->f;
|
|
|
35 res[i].e = ptr->e + ptr->e;
|
|
|
36 res[i].h = ptr->h + ptr->h;
|
|
|
37 res[i].g = ptr->g + ptr->g;
|
|
|
38 ptr++;
|
|
|
39 }
|
|
|
40
|
|
|
41 /* check results: */
|
|
|
42 for (i = 0; i < N; i++)
|
|
|
43 {
|
|
|
44 if (res[i].c != arr[i].c + arr[i].c
|
|
|
45 || res[i].a != arr[i].a + arr[i].a
|
|
|
46 || res[i].d != arr[i].d + arr[i].d
|
|
|
47 || res[i].b != arr[i].b + arr[i].b
|
|
|
48 || res[i].f != arr[i].f + arr[i].f
|
|
|
49 || res[i].e != arr[i].e + arr[i].e
|
|
|
50 || res[i].h != arr[i].h + arr[i].h
|
|
|
51 || res[i].g != arr[i].g + arr[i].g)
|
|
|
52 abort();
|
|
|
53 }
|
|
|
54
|
|
|
55 ptr = arr;
|
|
|
56 for (i = 0; i < N; i++)
|
|
|
57 {
|
|
|
58 res[i].c = ptr->c + ptr->c;
|
|
|
59 res[i].a = ptr->a + ptr->a;
|
|
|
60 res[i].d = ptr->d + ptr->d;
|
|
|
61 res[i].b = ptr->b + ptr->b;
|
|
|
62 res[i].f = ptr->f + ptr->f;
|
|
|
63 res[i].e = ptr->e + ptr->e;
|
|
|
64 res[i].h = ptr->e + ptr->e;
|
|
|
65 res[i].g = ptr->g + ptr->g;
|
|
|
66 ptr++;
|
|
|
67 }
|
|
|
68
|
|
|
69 /* check results: */
|
|
|
70 for (i = 0; i < N; i++)
|
|
|
71 {
|
|
|
72 if (res[i].c != arr[i].c + arr[i].c
|
|
|
73 || res[i].a != arr[i].a + arr[i].a
|
|
|
74 || res[i].d != arr[i].d + arr[i].d
|
|
|
75 || res[i].b != arr[i].b + arr[i].b
|
|
|
76 || res[i].f != arr[i].f + arr[i].f
|
|
|
77 || res[i].e != arr[i].e + arr[i].e
|
|
|
78 || res[i].h != arr[i].e + arr[i].e
|
|
|
79 || res[i].g != arr[i].g + arr[i].g)
|
|
|
80 abort();
|
|
|
81 }
|
|
|
82
|
|
|
83 }
|
|
|
84
|
|
|
85 int main (void)
|
|
|
86 {
|
|
|
87 int i;
|
|
|
88 s arr[N];
|
|
|
89
|
|
|
90 check_vect ();
|
|
|
91
|
|
|
92 for (i = 0; i < N; i++)
|
|
|
93 {
|
|
|
94 arr[i].a = i;
|
|
|
95 arr[i].b = i * 2;
|
|
|
96 arr[i].c = 17;
|
|
|
97 arr[i].d = i+34;
|
|
|
98 arr[i].e = i * 3 + 5;
|
|
|
99 arr[i].f = i * 5;
|
|
|
100 arr[i].g = i - 3;
|
|
|
101 arr[i].h = 56;
|
|
131
|
102 asm volatile ("" ::: "memory");
|
|
111
|
103 }
|
|
|
104
|
|
|
105 main1 (arr);
|
|
|
106
|
|
|
107 return 0;
|
|
|
108 }
|
|
|
109
|
|
|
110 /* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" { target { vect_strided8 && { ! { vect_no_align} } } } } } */
|
|
|
111 /* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { target { ! { vect_strided8 || vect_no_align } } } } } */
|
|
|
112 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "vect" { target { ! vect_perm } } } } */
|
|
131
|
113 /* SLP fails for the second loop with variable-length SVE because
|
|
|
114 the load size is greater than the minimum vector size. */
|
|
|
115 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 2 "vect" { target vect_perm xfail { aarch64_sve && vect_variable_length } } } } */
|
|
111
|
116
|
