|
111
|
1 /* { dg-require-effective-target vect_int } */
|
|
|
2
|
|
|
3 #include <stdarg.h>
|
|
|
4 #include "tree-vect.h"
|
|
|
5
|
|
131
|
6 #if VECTOR_BITS > 128
|
|
|
7 #define N (VECTOR_BITS * 2 / 32)
|
|
|
8 #else
|
|
111
|
9 #define N 16
|
|
131
|
10 #endif
|
|
111
|
11
|
|
|
12 int
|
|
|
13 main1 ()
|
|
|
14 {
|
|
|
15 unsigned int i;
|
|
|
16 unsigned int out[N*8];
|
|
131
|
17 #if N == 16
|
|
111
|
18 unsigned int in[N*8] = {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63};
|
|
131
|
19 #else
|
|
|
20 unsigned int in[N * 8];
|
|
|
21 for (unsigned int i = 0; i < N * 8; ++i)
|
|
|
22 {
|
|
|
23 in[i] = i & 63;
|
|
|
24 asm volatile ("" ::: "memory");
|
|
|
25 }
|
|
|
26 #endif
|
|
111
|
27 unsigned int ia[N*2], a0, a1, a2, a3;
|
|
|
28
|
|
|
29 /* The last stmt requires interleaving of not power of 2 size - not
|
|
|
30 vectorizable. */
|
|
|
31 for (i = 0; i < N/2; i++)
|
|
|
32 {
|
|
|
33 out[i*12] = in[i*12];
|
|
|
34 out[i*12 + 1] = in[i*12 + 1];
|
|
|
35 out[i*12 + 2] = in[i*12 + 2];
|
|
|
36 out[i*12 + 3] = in[i*12 + 3];
|
|
|
37 out[i*12 + 4] = in[i*12 + 4];
|
|
|
38 out[i*12 + 5] = in[i*12 + 5];
|
|
|
39 out[i*12 + 6] = in[i*12 + 6];
|
|
|
40 out[i*12 + 7] = in[i*12 + 7];
|
|
|
41 out[i*12 + 8] = in[i*12 + 8];
|
|
|
42 out[i*12 + 9] = in[i*12 + 9];
|
|
|
43 out[i*12 + 10] = in[i*12 + 10];
|
|
|
44 out[i*12 + 11] = in[i*12 + 11];
|
|
|
45
|
|
|
46 ia[i] = in[i*12 + 7];
|
|
|
47 }
|
|
|
48
|
|
|
49 /* check results: */
|
|
|
50 for (i = 0; i < N/2; i++)
|
|
|
51 {
|
|
|
52 if (out[i*12] != in[i*12]
|
|
|
53 || out[i*12 + 1] != in[i*12 + 1]
|
|
|
54 || out[i*12 + 2] != in[i*12 + 2]
|
|
|
55 || out[i*12 + 3] != in[i*12 + 3]
|
|
|
56 || out[i*12 + 4] != in[i*12 + 4]
|
|
|
57 || out[i*12 + 5] != in[i*12 + 5]
|
|
|
58 || out[i*12 + 6] != in[i*12 + 6]
|
|
|
59 || out[i*12 + 7] != in[i*12 + 7]
|
|
|
60 || out[i*12 + 8] != in[i*12 + 8]
|
|
|
61 || out[i*12 + 9] != in[i*12 + 9]
|
|
|
62 || out[i*12 + 10] != in[i*12 + 10]
|
|
|
63 || out[i*12 + 11] != in[i*12 + 11]
|
|
|
64 || ia[i] != in[i*12 + 7])
|
|
|
65 abort ();
|
|
|
66 }
|
|
|
67
|
|
|
68 /* Hybrid SLP with unrolling by 2. */
|
|
|
69 for (i = 0; i < N; i++)
|
|
|
70 {
|
|
|
71 out[i*6] = in[i*6];
|
|
|
72 out[i*6 + 1] = in[i*6 + 1];
|
|
|
73 out[i*6 + 2] = in[i*6 + 2];
|
|
|
74 out[i*6 + 3] = in[i*6 + 3];
|
|
|
75 out[i*6 + 4] = in[i*6 + 4];
|
|
|
76 out[i*6 + 5] = in[i*6 + 5];
|
|
|
77
|
|
|
78 ia[i] = i;
|
|
|
79 }
|
|
|
80
|
|
|
81 /* check results: */
|
|
|
82 for (i = 0; i < N/2; i++)
|
|
|
83 {
|
|
|
84 if (out[i*6] != in[i*6]
|
|
|
85 || out[i*6 + 1] != in[i*6 + 1]
|
|
|
86 || out[i*6 + 2] != in[i*6 + 2]
|
|
|
87 || out[i*6 + 3] != in[i*6 + 3]
|
|
|
88 || out[i*6 + 4] != in[i*6 + 4]
|
|
|
89 || out[i*6 + 5] != in[i*6 + 5]
|
|
|
90 || ia[i] != i)
|
|
|
91 abort ();
|
|
|
92 }
|
|
|
93
|
|
|
94 return 0;
|
|
|
95 }
|
|
|
96
|
|
|
97 int main (void)
|
|
|
98 {
|
|
|
99 check_vect ();
|
|
|
100
|
|
|
101 main1 ();
|
|
|
102
|
|
|
103 return 0;
|
|
|
104 }
|
|
|
105
|
|
|
106 /* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" } } */
|
|
|
107 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "vect" } } */
|
