| 1 |
/* { dg-options "-O2 -fno-schedule-insns -fno-schedule-insns2" } */ |
| 2 |
/* { dg-final { check-function-bodies "**" "" "" { target aarch64_little_endian } } } */ |
| 3 |
|
| 4 |
#include <arm_neon.h> |
| 5 |
|
| 6 |
/* |
| 7 |
** s32x2_1: |
| 8 |
** dup v0\.2s, w0 |
| 9 |
** ret |
| 10 |
*/ |
| 11 |
int32x2_t |
| 12 |
s32x2_1 (int32_t x) |
| 13 |
{ |
| 14 |
int32_t arr[] = { x, x }; |
| 15 |
return vld1_s32 (arr); |
| 16 |
} |
| 17 |
|
| 18 |
/* |
| 19 |
** s32x2_2: |
| 20 |
** fmov s0, w0 |
| 21 |
** ret |
| 22 |
*/ |
| 23 |
int32x2_t |
| 24 |
s32x2_2 (int32_t x) |
| 25 |
{ |
| 26 |
int32_t arr[] = { x, 0 }; |
| 27 |
return vld1_s32 (arr); |
| 28 |
} |
| 29 |
|
| 30 |
/* |
| 31 |
** s32x2_3: |
| 32 |
** fmov s0, w0 |
| 33 |
** ins v0\.s\[1\], w1 |
| 34 |
** ret |
| 35 |
*/ |
| 36 |
int32x2_t |
| 37 |
s32x2_3 (int32_t x, int32_t y) |
| 38 |
{ |
| 39 |
int32_t arr[] = { x, y }; |
| 40 |
return vld1_s32 (arr); |
| 41 |
} |
| 42 |
|
| 43 |
/* |
| 44 |
** f32x2_1: |
| 45 |
** dup v0\.2s, v0.s\[0\] |
| 46 |
** ret |
| 47 |
*/ |
| 48 |
float32x2_t |
| 49 |
f32x2_1 (float32_t x) |
| 50 |
{ |
| 51 |
float32_t arr[] = { x, x }; |
| 52 |
return vld1_f32 (arr); |
| 53 |
} |
| 54 |
|
| 55 |
/* |
| 56 |
** f32x2_2: |
| 57 |
** uzp1 v0\.2s, v0\.2s, v1\.2s |
| 58 |
** ret |
| 59 |
*/ |
| 60 |
float32x2_t |
| 61 |
f32x2_2 (float32_t x, float32_t y) |
| 62 |
{ |
| 63 |
float32_t arr[] = { x, y }; |
| 64 |
return vld1_f32 (arr); |
| 65 |
} |
| 66 |
|
| 67 |
/* |
| 68 |
** s16x4_1: |
| 69 |
** dup v0\.4h, w0 |
| 70 |
** ret |
| 71 |
*/ |
| 72 |
int16x4_t |
| 73 |
s16x4_1 (int16_t x) |
| 74 |
{ |
| 75 |
int16_t arr[] = { x, x, x, x }; |
| 76 |
return vld1_s16 (arr); |
| 77 |
} |
| 78 |
|
| 79 |
/* |
| 80 |
PR target/117092 |
| 81 |
** s16x4_2: { xfail *-*-* } |
| 82 |
** ... |
| 83 |
** fmov [dsh]0, [wx][0-9]+ |
| 84 |
** ret |
| 85 |
*/ |
| 86 |
int16x4_t |
| 87 |
s16x4_2 (int16_t x) |
| 88 |
{ |
| 89 |
int16_t arr[] = { x, 0, 0, 0 }; |
| 90 |
return vld1_s16 (arr); |
| 91 |
} |
| 92 |
|
| 93 |
/* |
| 94 |
** s16x4_3: |
| 95 |
** dup v0\.4h, w1 |
| 96 |
** ins v0.h\[0\], w0 |
| 97 |
** ret |
| 98 |
*/ |
| 99 |
int16x4_t |
| 100 |
s16x4_3 (int16_t x, int16_t y) |
| 101 |
{ |
| 102 |
int16_t arr[] = { x, y, y, y }; |
| 103 |
return vld1_s16 (arr); |
| 104 |
} |
| 105 |
|
| 106 |
/* |
| 107 |
** f16x4_1: |
| 108 |
** dup v0\.4h, v0.h\[0\] |
| 109 |
** ret |
| 110 |
*/ |
| 111 |
float16x4_t |
| 112 |
f16x4_1 (float16_t x) |
| 113 |
{ |
| 114 |
float16_t arr[] = { x, x, x, x }; |
| 115 |
return vld1_f16 (arr); |
| 116 |
} |
| 117 |
|
| 118 |
/* |
| 119 |
** s64x2_1: |
| 120 |
** dup v0\.2d, x0 |
| 121 |
** ret |
| 122 |
*/ |
| 123 |
int64x2_t |
| 124 |
s64x2_1 (int64_t x) |
| 125 |
{ |
| 126 |
int64_t arr[] = { x, x }; |
| 127 |
return vld1q_s64 (arr); |
| 128 |
} |
| 129 |
|
| 130 |
/* |
| 131 |
** s64x2_2: |
| 132 |
** fmov d0, x0 |
| 133 |
** ret |
| 134 |
*/ |
| 135 |
int64x2_t |
| 136 |
s64x2_2 (int64_t x) |
| 137 |
{ |
| 138 |
int64_t arr[] = { x, 0 }; |
| 139 |
return vld1q_s64 (arr); |
| 140 |
} |
| 141 |
|
| 142 |
/* |
| 143 |
** s64x2_3: |
| 144 |
** fmov d0, x0 |
| 145 |
** ins v0\.d\[1\], x1 |
| 146 |
** ret |
| 147 |
*/ |
| 148 |
int64x2_t |
| 149 |
s64x2_3 (int64_t x, int64_t y) |
| 150 |
{ |
| 151 |
int64_t arr[] = { x, y }; |
| 152 |
return vld1q_s64 (arr); |
| 153 |
} |
| 154 |
|
| 155 |
/* |
| 156 |
** f64x2_1: |
| 157 |
** dup v0\.2d, v0.d\[0\] |
| 158 |
** ret |
| 159 |
*/ |
| 160 |
float64x2_t |
| 161 |
f64x2_1 (float64_t x) |
| 162 |
{ |
| 163 |
float64_t arr[] = { x, x }; |
| 164 |
return vld1q_f64 (arr); |
| 165 |
} |
| 166 |
|
| 167 |
/* |
| 168 |
** f64x2_2: |
| 169 |
** uzp1 v0\.2d, v0\.2d, v1\.2d |
| 170 |
** ret |
| 171 |
*/ |
| 172 |
float64x2_t |
| 173 |
f64x2_2 (float64_t x, float64_t y) |
| 174 |
{ |
| 175 |
float64_t arr[] = { x, y }; |
| 176 |
return vld1q_f64 (arr); |
| 177 |
} |
| 178 |
|
| 179 |
/* |
| 180 |
** s32x4_1: |
| 181 |
** dup v0\.4s, w0 |
| 182 |
** ret |
| 183 |
*/ |
| 184 |
int32x4_t |
| 185 |
s32x4_1 (int32_t x) |
| 186 |
{ |
| 187 |
int32_t arr[] = { x, x, x, x }; |
| 188 |
return vld1q_s32 (arr); |
| 189 |
} |
| 190 |
|
| 191 |
/* |
| 192 |
** s32x4_2: { xfail *-*-* } |
| 193 |
** fmov s0, w0 |
| 194 |
** ret |
| 195 |
*/ |
| 196 |
int32x4_t |
| 197 |
s32x4_2 (int32_t x) |
| 198 |
{ |
| 199 |
int32_t arr[] = { x, 0, 0, 0 }; |
| 200 |
return vld1q_s32 (arr); |
| 201 |
} |
| 202 |
|
| 203 |
/* |
| 204 |
** s32x4_3: |
| 205 |
** dup v0\.4s, w1 |
| 206 |
** ins v0.s\[0\], w0 |
| 207 |
** ret |
| 208 |
*/ |
| 209 |
int32x4_t |
| 210 |
s32x4_3 (int32_t x, int32_t y) |
| 211 |
{ |
| 212 |
int32_t arr[] = { x, y, y, y }; |
| 213 |
return vld1q_s32 (arr); |
| 214 |
} |
| 215 |
|
| 216 |
/* |
| 217 |
** f32x4_1: |
| 218 |
** dup v0\.4s, v0.s\[0\] |
| 219 |
** ret |
| 220 |
*/ |
| 221 |
float32x4_t |
| 222 |
f32x4_1 (float32_t x) |
| 223 |
{ |
| 224 |
float32_t arr[] = { x, x, x, x }; |
| 225 |
return vld1q_f32 (arr); |
| 226 |
} |
| 227 |
|
| 228 |
void consume (float32x4_t, float32x4_t, float32x4_t, float32x4_t); |
| 229 |
|
| 230 |
/* |
| 231 |
** produce_1: |
| 232 |
** ( |
| 233 |
** dup v0\.4s, v0\.s\[0\] |
| 234 |
** dup v1\.4s, v1\.s\[0\] |
| 235 |
** dup v2\.4s, v2\.s\[0\] |
| 236 |
** dup v3\.4s, v3\.s\[0\] |
| 237 |
** | |
| 238 |
** dup v3\.4s, v3\.s\[0\] |
| 239 |
** dup v2\.4s, v2\.s\[0\] |
| 240 |
** dup v1\.4s, v1\.s\[0\] |
| 241 |
** dup v0\.4s, v0\.s\[0\] |
| 242 |
** ) |
| 243 |
** b consume |
| 244 |
*/ |
| 245 |
void |
| 246 |
produce_1 (float32_t a, float32_t b, float32_t c, float32_t d) |
| 247 |
{ |
| 248 |
float arr[4][4] = { |
| 249 |
{ a, a, a, a }, |
| 250 |
{ b, b, b, b }, |
| 251 |
{ c, c, c, c }, |
| 252 |
{ d, d, d, d } |
| 253 |
}; |
| 254 |
consume (vld1q_f32 (arr[0]), vld1q_f32 (arr[1]), |
| 255 |
vld1q_f32 (arr[2]), vld1q_f32 (arr[3])); |
| 256 |
} |
| 257 |
|
| 258 |
/* |
| 259 |
** produce_2: |
| 260 |
** ( |
| 261 |
** dup v0\.4s, v0\.s\[0\] |
| 262 |
** dup v1\.4s, v1\.s\[0\] |
| 263 |
** dup v2\.4s, v2\.s\[0\] |
| 264 |
** dup v3\.4s, v3\.s\[0\] |
| 265 |
** | |
| 266 |
** dup v3\.4s, v3\.s\[0\] |
| 267 |
** dup v2\.4s, v2\.s\[0\] |
| 268 |
** dup v1\.4s, v1\.s\[0\] |
| 269 |
** dup v0\.4s, v0\.s\[0\] |
| 270 |
** ) |
| 271 |
** b consume |
| 272 |
*/ |
| 273 |
void |
| 274 |
produce_2 (float32_t a, float32_t b, float32_t c, float32_t d) |
| 275 |
{ |
| 276 |
float arr0[] = { a, a, a, a }; |
| 277 |
float arr1[] = { b, b, b, b }; |
| 278 |
float arr2[] = { c, c, c, c }; |
| 279 |
float arr3[] = { d, d, d, d }; |
| 280 |
consume (vld1q_f32 (arr0), vld1q_f32 (arr1), |
| 281 |
vld1q_f32 (arr2), vld1q_f32 (arr3)); |
| 282 |
} |