| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-additional-options "-O3 -march=armv9-a+fp8 -moverride=tune=cheap_fpmr_write" } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 4 |
|
| 5 |
#include "arm_neon.h" |
| 6 |
|
| 7 |
/* |
| 8 |
** test_vcvt1_bf16: |
| 9 |
** msr fpmr, x0 |
| 10 |
** bf1cvtl v0.8h, v0.8b |
| 11 |
** ret |
| 12 |
*/ |
| 13 |
bfloat16x8_t |
| 14 |
test_vcvt1_bf16 (mfloat8x8_t a, fpm_t b) |
| 15 |
{ |
| 16 |
return vcvt1_bf16_mf8_fpm(a, b); |
| 17 |
} |
| 18 |
|
| 19 |
/* |
| 20 |
** test_high_vcvt1_bf16: |
| 21 |
** msr fpmr, x0 |
| 22 |
** bf1cvtl2 v0.8h, v0.16b |
| 23 |
** ret |
| 24 |
*/ |
| 25 |
bfloat16x8_t |
| 26 |
test_high_vcvt1_bf16 (mfloat8x16_t a, fpm_t b) |
| 27 |
{ |
| 28 |
return vcvt1_high_bf16_mf8_fpm(a, b); |
| 29 |
} |
| 30 |
|
| 31 |
/* |
| 32 |
** test_low_vcvt1_bf16: |
| 33 |
** msr fpmr, x0 |
| 34 |
** bf1cvtl v0.8h, v0.8b |
| 35 |
** ret |
| 36 |
*/ |
| 37 |
bfloat16x8_t |
| 38 |
test_low_vcvt1_bf16 (mfloat8x16_t a, fpm_t b) |
| 39 |
{ |
| 40 |
return vcvt1_low_bf16_mf8_fpm(a, b); |
| 41 |
} |
| 42 |
|
| 43 |
/* |
| 44 |
** test_vcvt1_f16: |
| 45 |
** msr fpmr, x0 |
| 46 |
** f1cvtl v0.8h, v0.8b |
| 47 |
** ret |
| 48 |
*/ |
| 49 |
float16x8_t |
| 50 |
test_vcvt1_f16 (mfloat8x8_t a, fpm_t b) |
| 51 |
{ |
| 52 |
return vcvt1_f16_mf8_fpm(a, b); |
| 53 |
} |
| 54 |
|
| 55 |
/* |
| 56 |
** test_high_vcvt1_f16: |
| 57 |
** msr fpmr, x0 |
| 58 |
** f1cvtl2 v0.8h, v0.16b |
| 59 |
** ret |
| 60 |
*/ |
| 61 |
float16x8_t |
| 62 |
test_high_vcvt1_f16 (mfloat8x16_t a, fpm_t b) |
| 63 |
{ |
| 64 |
return vcvt1_high_f16_mf8_fpm(a, b); |
| 65 |
} |
| 66 |
|
| 67 |
/* |
| 68 |
** test_low_vcvt1_f16: |
| 69 |
** msr fpmr, x0 |
| 70 |
** f1cvtl v0.8h, v0.8b |
| 71 |
** ret |
| 72 |
*/ |
| 73 |
float16x8_t |
| 74 |
test_low_vcvt1_f16 (mfloat8x16_t a, fpm_t b) |
| 75 |
{ |
| 76 |
return vcvt1_low_f16_mf8_fpm(a, b); |
| 77 |
} |
| 78 |
|
| 79 |
/* |
| 80 |
** test_vcvt2_bf16: |
| 81 |
** msr fpmr, x0 |
| 82 |
** bf2cvtl v0.8h, v0.8b |
| 83 |
** ret |
| 84 |
*/ |
| 85 |
bfloat16x8_t |
| 86 |
test_vcvt2_bf16 (mfloat8x8_t a, fpm_t b) |
| 87 |
{ |
| 88 |
return vcvt2_bf16_mf8_fpm(a, b); |
| 89 |
} |
| 90 |
|
| 91 |
/* |
| 92 |
** test_high_vcvt2_bf16: |
| 93 |
** msr fpmr, x0 |
| 94 |
** bf2cvtl2 v0.8h, v0.16b |
| 95 |
** ret |
| 96 |
*/ |
| 97 |
bfloat16x8_t |
| 98 |
test_high_vcvt2_bf16 (mfloat8x16_t a, fpm_t b) |
| 99 |
{ |
| 100 |
return vcvt2_high_bf16_mf8_fpm(a, b); |
| 101 |
} |
| 102 |
|
| 103 |
/* |
| 104 |
** test_low_vcvt2_bf16: |
| 105 |
** msr fpmr, x0 |
| 106 |
** bf1cvtl v0.8h, v0.8b |
| 107 |
** ret |
| 108 |
*/ |
| 109 |
bfloat16x8_t |
| 110 |
test_low_vcvt2_bf16 (mfloat8x16_t a, fpm_t b) |
| 111 |
{ |
| 112 |
return vcvt1_low_bf16_mf8_fpm(a, b); |
| 113 |
} |
| 114 |
|
| 115 |
/* |
| 116 |
** test_vcvt2_f16: |
| 117 |
** msr fpmr, x0 |
| 118 |
** f2cvtl v0.8h, v0.8b |
| 119 |
** ret |
| 120 |
*/ |
| 121 |
float16x8_t |
| 122 |
test_vcvt2_f16 (mfloat8x8_t a, fpm_t b) |
| 123 |
{ |
| 124 |
return vcvt2_f16_mf8_fpm(a, b); |
| 125 |
} |
| 126 |
|
| 127 |
/* |
| 128 |
** test_high_vcvt2_f16: |
| 129 |
** msr fpmr, x0 |
| 130 |
** f2cvtl2 v0.8h, v0.16b |
| 131 |
** ret |
| 132 |
*/ |
| 133 |
float16x8_t |
| 134 |
test_high_vcvt2_f16 (mfloat8x16_t a, fpm_t b) |
| 135 |
{ |
| 136 |
return vcvt2_high_f16_mf8_fpm(a, b); |
| 137 |
} |
| 138 |
|
| 139 |
/* |
| 140 |
** test_low_vcvt2_f16: |
| 141 |
** msr fpmr, x0 |
| 142 |
** f1cvtl v0.8h, v0.8b |
| 143 |
** ret |
| 144 |
*/ |
| 145 |
float16x8_t |
| 146 |
test_low_vcvt2_f16 (mfloat8x16_t a, fpm_t b) |
| 147 |
{ |
| 148 |
return vcvt1_low_f16_mf8_fpm(a, b); |
| 149 |
} |
| 150 |
|
| 151 |
/* |
| 152 |
** test_vcvt_f16: |
| 153 |
** msr fpmr, x0 |
| 154 |
** fcvtn v0.8b, v0.4h, v1.4h |
| 155 |
** ret |
| 156 |
*/ |
| 157 |
mfloat8x8_t |
| 158 |
test_vcvt_f16 (float16x4_t a, float16x4_t b, fpm_t c) |
| 159 |
{ |
| 160 |
return vcvt_mf8_f16_fpm(a, b, c); |
| 161 |
} |
| 162 |
|
| 163 |
/* |
| 164 |
** test_vcvtq_f16: |
| 165 |
** msr fpmr, x0 |
| 166 |
** fcvtn v0.16b, v0.8h, v1.8h |
| 167 |
** ret |
| 168 |
*/ |
| 169 |
mfloat8x16_t |
| 170 |
test_vcvtq_f16 (float16x8_t a, float16x8_t b, fpm_t c) |
| 171 |
{ |
| 172 |
return vcvtq_mf8_f16_fpm(a, b, c); |
| 173 |
} |
| 174 |
|
| 175 |
/* |
| 176 |
** test_vcvt_f32: |
| 177 |
** msr fpmr, x0 |
| 178 |
** fcvtn v0.8b, v0.4s, v1.4s |
| 179 |
** ret |
| 180 |
*/ |
| 181 |
mfloat8x8_t |
| 182 |
test_vcvt_f32 (float32x4_t a, float32x4_t b, fpm_t c) |
| 183 |
{ |
| 184 |
return vcvt_mf8_f32_fpm(a, b, c); |
| 185 |
} |
| 186 |
|
| 187 |
/* |
| 188 |
** test_vcvt_high_f32: |
| 189 |
** msr fpmr, x0 |
| 190 |
** fcvtn2 v0.16b, v1.4s, v2.4s |
| 191 |
** ret |
| 192 |
*/ |
| 193 |
mfloat8x16_t |
| 194 |
test_vcvt_high_f32 (mfloat8x8_t a, float32x4_t b, float32x4_t c, fpm_t d) |
| 195 |
{ |
| 196 |
return vcvt_high_mf8_f32_fpm(a, b, c, d); |
| 197 |
} |