| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-additional-options "-O3 -ffast-math -march=armv9-a" } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 4 |
|
| 5 |
#include "arm_neon.h" |
| 6 |
|
| 7 |
#pragma GCC target "+nosve" |
| 8 |
|
| 9 |
/* |
| 10 |
** test_vamax_f16: |
| 11 |
** fabs v1.4h, v1.4h |
| 12 |
** fabs v0.4h, v0.4h |
| 13 |
** fmaxnm v0.4h, v0.4h, v1.4h |
| 14 |
** ret |
| 15 |
*/ |
| 16 |
float16x4_t |
| 17 |
test_vamax_f16 (float16x4_t a, float16x4_t b) |
| 18 |
{ |
| 19 |
int i; |
| 20 |
float16x4_t c; |
| 21 |
|
| 22 |
for (i = 0; i < 4; ++i) { |
| 23 |
a[i] = __builtin_fabsf16 (a[i]); |
| 24 |
b[i] = __builtin_fabsf16 (b[i]); |
| 25 |
c[i] = __builtin_fmaxf16 (a[i], b[i]); |
| 26 |
} |
| 27 |
return c; |
| 28 |
} |
| 29 |
|
| 30 |
/* |
| 31 |
** test_vamaxq_f16: |
| 32 |
** fabs v1.8h, v1.8h |
| 33 |
** fabs v0.8h, v0.8h |
| 34 |
** fmaxnm v0.8h, v0.8h, v1.8h |
| 35 |
** ret |
| 36 |
*/ |
| 37 |
float16x8_t |
| 38 |
test_vamaxq_f16 (float16x8_t a, float16x8_t b) |
| 39 |
{ |
| 40 |
int i; |
| 41 |
float16x8_t c; |
| 42 |
|
| 43 |
for (i = 0; i < 8; ++i) { |
| 44 |
a[i] = __builtin_fabsf16 (a[i]); |
| 45 |
b[i] = __builtin_fabsf16 (b[i]); |
| 46 |
c[i] = __builtin_fmaxf16 (a[i], b[i]); |
| 47 |
} |
| 48 |
return c; |
| 49 |
} |
| 50 |
|
| 51 |
/* |
| 52 |
** test_vamax_f32: |
| 53 |
** fabs v1.2s, v1.2s |
| 54 |
** fabs v0.2s, v0.2s |
| 55 |
** fmaxnm v0.2s, v0.2s, v1.2s |
| 56 |
** ret |
| 57 |
*/ |
| 58 |
float32x2_t |
| 59 |
test_vamax_f32 (float32x2_t a, float32x2_t b) |
| 60 |
{ |
| 61 |
int i; |
| 62 |
float32x2_t c; |
| 63 |
|
| 64 |
for (i = 0; i < 2; ++i) { |
| 65 |
a[i] = __builtin_fabsf32 (a[i]); |
| 66 |
b[i] = __builtin_fabsf32 (b[i]); |
| 67 |
c[i] = __builtin_fmaxf32 (a[i], b[i]); |
| 68 |
} |
| 69 |
return c; |
| 70 |
} |
| 71 |
|
| 72 |
/* |
| 73 |
** test_vamaxq_f32: |
| 74 |
** fabs v1.4s, v1.4s |
| 75 |
** fabs v0.4s, v0.4s |
| 76 |
** fmaxnm v0.4s, v0.4s, v1.4s |
| 77 |
** ret |
| 78 |
*/ |
| 79 |
float32x4_t |
| 80 |
test_vamaxq_f32 (float32x4_t a, float32x4_t b) |
| 81 |
{ |
| 82 |
int i; |
| 83 |
float32x4_t c; |
| 84 |
|
| 85 |
for (i = 0; i < 4; ++i) { |
| 86 |
a[i] = __builtin_fabsf32 (a[i]); |
| 87 |
b[i] = __builtin_fabsf32 (b[i]); |
| 88 |
c[i] = __builtin_fmaxf32 (a[i], b[i]); |
| 89 |
} |
| 90 |
return c; |
| 91 |
} |
| 92 |
|
| 93 |
/* |
| 94 |
** test_vamaxq_f64: |
| 95 |
** fabs v1.2d, v1.2d |
| 96 |
** fabs v0.2d, v0.2d |
| 97 |
** fmaxnm v0.2d, v0.2d, v1.2d |
| 98 |
** ret |
| 99 |
*/ |
| 100 |
float64x2_t |
| 101 |
test_vamaxq_f64 (float64x2_t a, float64x2_t b) |
| 102 |
{ |
| 103 |
int i; |
| 104 |
float64x2_t c; |
| 105 |
|
| 106 |
for (i = 0; i < 2; ++i) { |
| 107 |
a[i] = __builtin_fabsf64 (a[i]); |
| 108 |
b[i] = __builtin_fabsf64 (b[i]); |
| 109 |
c[i] = __builtin_fmaxf64 (a[i], b[i]); |
| 110 |
} |
| 111 |
return c; |
| 112 |
} |
| 113 |
|
| 114 |
/* |
| 115 |
** test_vamin_f16: |
| 116 |
** fabs v1.4h, v1.4h |
| 117 |
** fabs v0.4h, v0.4h |
| 118 |
** fminnm v0.4h, v0.4h, v1.4h |
| 119 |
** ret |
| 120 |
*/ |
| 121 |
float16x4_t |
| 122 |
test_vamin_f16 (float16x4_t a, float16x4_t b) |
| 123 |
{ |
| 124 |
int i; |
| 125 |
float16x4_t c; |
| 126 |
|
| 127 |
for (i = 0; i < 4; ++i) { |
| 128 |
a[i] = __builtin_fabsf16 (a[i]); |
| 129 |
b[i] = __builtin_fabsf16 (b[i]); |
| 130 |
c[i] = __builtin_fminf16 (a[i], b[i]); |
| 131 |
} |
| 132 |
return c; |
| 133 |
} |
| 134 |
|
| 135 |
/* |
| 136 |
** test_vaminq_f16: |
| 137 |
** fabs v1.8h, v1.8h |
| 138 |
** fabs v0.8h, v0.8h |
| 139 |
** fminnm v0.8h, v0.8h, v1.8h |
| 140 |
** ret |
| 141 |
*/ |
| 142 |
float16x8_t |
| 143 |
test_vaminq_f16 (float16x8_t a, float16x8_t b) |
| 144 |
{ |
| 145 |
int i; |
| 146 |
float16x8_t c; |
| 147 |
|
| 148 |
for (i = 0; i < 8; ++i) { |
| 149 |
a[i] = __builtin_fabsf16 (a[i]); |
| 150 |
b[i] = __builtin_fabsf16 (b[i]); |
| 151 |
c[i] = __builtin_fminf16 (a[i], b[i]); |
| 152 |
} |
| 153 |
return c; |
| 154 |
} |
| 155 |
|
| 156 |
/* |
| 157 |
** test_vamin_f32: |
| 158 |
** fabs v1.2s, v1.2s |
| 159 |
** fabs v0.2s, v0.2s |
| 160 |
** fminnm v0.2s, v0.2s, v1.2s |
| 161 |
** ret |
| 162 |
*/ |
| 163 |
float32x2_t |
| 164 |
test_vamin_f32 (float32x2_t a, float32x2_t b) |
| 165 |
{ |
| 166 |
int i; |
| 167 |
float32x2_t c; |
| 168 |
|
| 169 |
for (i = 0; i < 2; ++i) { |
| 170 |
a[i] = __builtin_fabsf32 (a[i]); |
| 171 |
b[i] = __builtin_fabsf32 (b[i]); |
| 172 |
c[i] = __builtin_fminf32 (a[i], b[i]); |
| 173 |
} |
| 174 |
return c; |
| 175 |
} |
| 176 |
|
| 177 |
/* |
| 178 |
** test_vaminq_f32: |
| 179 |
** fabs v1.4s, v1.4s |
| 180 |
** fabs v0.4s, v0.4s |
| 181 |
** fminnm v0.4s, v0.4s, v1.4s |
| 182 |
** ret |
| 183 |
*/ |
| 184 |
float32x4_t |
| 185 |
test_vaminq_f32 (float32x4_t a, float32x4_t b) |
| 186 |
{ |
| 187 |
int i; |
| 188 |
float32x4_t c; |
| 189 |
|
| 190 |
for (i = 0; i < 4; ++i) { |
| 191 |
a[i] = __builtin_fabsf32 (a[i]); |
| 192 |
b[i] = __builtin_fabsf32 (b[i]); |
| 193 |
c[i] = __builtin_fminf32 (a[i], b[i]); |
| 194 |
} |
| 195 |
return c; |
| 196 |
} |
| 197 |
|
| 198 |
/* |
| 199 |
** test_vaminq_f64: |
| 200 |
** fabs v1.2d, v1.2d |
| 201 |
** fabs v0.2d, v0.2d |
| 202 |
** fminnm v0.2d, v0.2d, v1.2d |
| 203 |
** ret |
| 204 |
*/ |
| 205 |
float64x2_t |
| 206 |
test_vaminq_f64 (float64x2_t a, float64x2_t b) |
| 207 |
{ |
| 208 |
int i; |
| 209 |
float64x2_t c; |
| 210 |
|
| 211 |
for (i = 0; i < 2; ++i) { |
| 212 |
a[i] = __builtin_fabsf64 (a[i]); |
| 213 |
b[i] = __builtin_fabsf64 (b[i]); |
| 214 |
c[i] = __builtin_fminf64 (a[i], b[i]); |
| 215 |
} |
| 216 |
return c; |
| 217 |
} |