1 /* { dg-do compile } */
2 /* { dg-options "-mavx512bw -mavx512vbmi -O2" } */
3 /* { dg-final { scan-assembler-times {(?n)vmov[a-z0-9]*[ \t]*%ymm[0-9]} 12} } */
4 /* { dg-final { scan-assembler-not "pxor" } } */
5
6
7 typedef float v16sf __attribute__((vector_size(64)));
8 typedef double v8df __attribute__ ((vector_size (64)));
9 typedef long long v8di __attribute__((vector_size(64)));
10 typedef int v16si __attribute__((vector_size(64)));
11 typedef short v32hi __attribute__ ((vector_size (64)));
12 typedef char v64qi __attribute__ ((vector_size (64)));
13
14 v8df
15 foo_v8df (v8df x)
16 {
17 return __builtin_shuffle (x, (v8df) { 0, 0, 0, 0, 0, 0, 0, 0 },
18 (v8di) { 0, 1, 2, 3, 15, 14, 10, 11 });
19 }
20
21 v8df
22 foo_v8df_l (v8df x)
23 {
24 return __builtin_shuffle ((v8df) { 0, 0, 0, 0, 0, 0, 0, 0 }, x,
25 (v8di) { 8, 9, 10, 11, 0, 1, 2, 3 });
26 }
27
28 v8di
29 foo_v8di (v8di x)
30 {
31 return __builtin_shuffle (x, (v8di) { 0, 0, 0, 0, 0, 0, 0, 0 },
32 (v8di) { 0, 1, 2, 3, 8, 9, 10, 11 });
33 }
34
35 v8di
36 foo_v8di_l (v8di x)
37 {
38 return __builtin_shuffle ((v8di) { 0, 0, 0, 0, 0, 0, 0, 0 }, x,
39 (v8di) { 8, 9, 10, 11, 7, 6, 5, 4 });
40 }
41
42 v16sf
43 foo_v16sf (v16sf x)
44 {
45 return __builtin_shuffle (x, (v16sf) { 0, 0, 0, 0, 0, 0, 0, 0,
46 0, 0, 0, 0, 0, 0, 0, 0 },
47 (v16si) { 0, 1, 2, 3, 4, 5, 6, 7,
48 16, 17, 18, 19, 20, 21, 22, 23 });
49 }
50
51 v16sf
52 foo_v16sf_l (v16sf x)
53 {
54 return __builtin_shuffle ((v16sf) { 0, 0, 0, 0, 0, 0, 0, 0,
55 0, 0, 0, 0, 0, 0, 0, 0 }, x,
56 (v16si) { 16, 17, 18, 19, 20, 21, 22, 23,
57 0, 15, 2, 13, 4, 11, 6, 9 });
58 }
59
60 v16si
61 foo_v16si (v16si x)
62 {
63 return __builtin_shuffle (x, (v16si) { 0, 0, 0, 0, 0, 0, 0, 0,
64 0, 0, 0, 0, 0, 0, 0, 0 },
65 (v16si) { 0, 1, 2, 3, 4, 5, 6, 7,
66 31, 30, 29, 28, 20, 21, 22, 23 });
67 }
68
69 v16si
70 foo_v16si_l (v16si x)
71 {
72 return __builtin_shuffle ((v16si) { 0, 0, 0, 0, 0, 0, 0, 0,
73 0, 0, 0, 0, 0, 0, 0, 0 }, x,
74 (v16si) { 16, 17, 18, 19, 20, 21, 22, 23,
75 15, 0, 13, 2, 11, 4, 9, 6 });
76 }
77
78 v32hi
79 foo_v32hi (v32hi x)
80 {
81 return __builtin_shuffle (x, (v32hi) { 0, 0, 0, 0, 0, 0, 0, 0,
82 0, 0, 0, 0, 0, 0, 0, 0,
83 0, 0, 0, 0, 0, 0, 0, 0,
84 0, 0, 0, 0, 0, 0, 0, 0 },
85 (v32hi) { 0, 1, 2, 3, 4, 5, 6, 7,
86 8, 9, 10, 11, 12, 13, 14, 15,
87 63, 33, 61, 35, 59, 37, 57, 39,
88 55, 41, 53, 43, 51, 45, 49, 47 });
89 }
90
91 v32hi
92 foo_v32hi_l (v32hi x)
93 {
94 return __builtin_shuffle ((v32hi) { 0, 0, 0, 0, 0, 0, 0, 0,
95 0, 0, 0, 0, 0, 0, 0, 0,
96 0, 0, 0, 0, 0, 0, 0, 0,
97 0, 0, 0, 0, 0, 0, 0, 0 }, x,
98 (v32hi) { 32, 33, 34, 35, 36, 37, 38, 39,
99 40, 41, 42, 43, 44, 45, 46, 47,
100 31, 0, 29, 2, 27, 4, 25, 6,
101 23, 8, 21, 10, 19, 12, 17, 14 });
102 }
103
104 v64qi
105 foo_v64qi (v64qi x)
106 {
107 return __builtin_shuffle (x, (v64qi) { 0, 0, 0, 0, 0, 0, 0, 0,
108 0, 0, 0, 0, 0, 0, 0, 0,
109 0, 0, 0, 0, 0, 0, 0, 0,
110 0, 0, 0, 0, 0, 0, 0, 0,
111 0, 0, 0, 0, 0, 0, 0, 0,
112 0, 0, 0, 0, 0, 0, 0, 0,
113 0, 0, 0, 0, 0, 0, 0, 0,
114 0, 0, 0, 0, 0, 0, 0, 0 },
115 (v64qi) {0, 1, 2, 3, 4, 5, 6, 7,
116 8, 9, 10, 11, 12, 13, 14, 15,
117 16, 17, 18, 19, 20, 21, 22, 23,
118 24, 25, 26, 27, 28, 29, 30, 31,
119 64, 127, 66, 125, 68, 123, 70, 121,
120 72, 119, 74, 117, 76, 115, 78, 113,
121 80, 111, 82, 109, 84, 107, 86, 105,
122 88, 103, 90, 101, 92, 99, 94, 97 });
123 }
124
125 v64qi
126 foo_v64qi_l (v64qi x)
127 {
128 return __builtin_shuffle ((v64qi) { 0, 0, 0, 0, 0, 0, 0, 0,
129 0, 0, 0, 0, 0, 0, 0, 0,
130 0, 0, 0, 0, 0, 0, 0, 0,
131 0, 0, 0, 0, 0, 0, 0, 0,
132 0, 0, 0, 0, 0, 0, 0, 0,
133 0, 0, 0, 0, 0, 0, 0, 0,
134 0, 0, 0, 0, 0, 0, 0, 0,
135 0, 0, 0, 0, 0, 0, 0, 0 }, x,
136 (v64qi) { 64, 65, 66, 67, 68, 69, 70, 71,
137 72, 73, 74, 75, 76, 77, 78, 79,
138 80, 81, 82, 83, 84, 85, 86, 87,
139 88, 89, 90, 91, 92, 93, 94, 95,
140 0, 63, 2, 61, 4, 59, 6, 57,
141 8, 55, 10, 53, 12, 51, 14, 49,
142 16, 47, 18, 45, 20, 43, 22, 41,
143 24, 39, 26, 37, 28, 35, 30, 33 });
144 }