| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-additional-options "-O3 -march=armv9-a+fp8dot4 -moverride=tune=cheap_fpmr_write" } */ |
| 3 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 4 |
|
| 5 |
#include "arm_neon.h" |
| 6 |
|
| 7 |
/* |
| 8 |
** test_vdot_f32_fpm: |
| 9 |
** msr fpmr, x0 |
| 10 |
** fdot v0.2s, v1.8b, v2.8b |
| 11 |
** ret |
| 12 |
*/ |
| 13 |
float32x2_t |
| 14 |
test_vdot_f32_fpm (float32x2_t a, mfloat8x8_t b, mfloat8x8_t c, fpm_t d) |
| 15 |
{ |
| 16 |
return vdot_f32_mf8_fpm (a, b, c, d); |
| 17 |
} |
| 18 |
|
| 19 |
/* |
| 20 |
** test_vdotq_f32_fpm: |
| 21 |
** msr fpmr, x0 |
| 22 |
** fdot v0.4s, v1.16b, v2.16b |
| 23 |
** ret |
| 24 |
*/ |
| 25 |
float32x4_t |
| 26 |
test_vdotq_f32_fpm (float32x4_t a, mfloat8x16_t b, mfloat8x16_t c, fpm_t d) |
| 27 |
{ |
| 28 |
return vdotq_f32_mf8_fpm (a, b, c, d); |
| 29 |
} |
| 30 |
|
| 31 |
/* |
| 32 |
** test_vdot_lane_f32_fpm_0: |
| 33 |
** msr fpmr, x0 |
| 34 |
** fdot v0.2s, v1.8b, v2.4b\[0\] |
| 35 |
** ret |
| 36 |
*/ |
| 37 |
float32x2_t |
| 38 |
test_vdot_lane_f32_fpm_0 (float32x2_t a, mfloat8x8_t b, mfloat8x8_t c, fpm_t d) |
| 39 |
{ |
| 40 |
return vdot_lane_f32_mf8_fpm (a, b, c, 0, d); |
| 41 |
} |
| 42 |
|
| 43 |
/* |
| 44 |
** test_vdot_lane_f32_fpm_1: |
| 45 |
** msr fpmr, x0 |
| 46 |
** fdot v0.2s, v1.8b, v2.4b\[1\] |
| 47 |
** ret |
| 48 |
*/ |
| 49 |
float32x2_t |
| 50 |
test_vdot_lane_f32_fpm_1 (float32x2_t a, mfloat8x8_t b, mfloat8x8_t c, fpm_t d) |
| 51 |
{ |
| 52 |
return vdot_lane_f32_mf8_fpm (a, b, c, 1, d); |
| 53 |
} |
| 54 |
|
| 55 |
/* |
| 56 |
** test_vdot_laneq_f32_fpm_0: |
| 57 |
** msr fpmr, x0 |
| 58 |
** fdot v0.2s, v1.8b, v2.4b\[0\] |
| 59 |
** ret |
| 60 |
*/ |
| 61 |
float32x2_t |
| 62 |
test_vdot_laneq_f32_fpm_0 (float32x2_t a, mfloat8x8_t b, mfloat8x16_t c, fpm_t d) |
| 63 |
{ |
| 64 |
return vdot_laneq_f32_mf8_fpm (a, b, c, 0, d); |
| 65 |
} |
| 66 |
|
| 67 |
/* |
| 68 |
** test_vdot_laneq_f32_fpm_3: |
| 69 |
** msr fpmr, x0 |
| 70 |
** fdot v0.2s, v1.8b, v2.4b\[3\] |
| 71 |
** ret |
| 72 |
*/ |
| 73 |
float32x2_t |
| 74 |
test_vdot_laneq_f32_fpm_3 (float32x2_t a, mfloat8x8_t b, mfloat8x16_t c, fpm_t d) |
| 75 |
{ |
| 76 |
return vdot_laneq_f32_mf8_fpm (a, b, c, 3, d); |
| 77 |
} |
| 78 |
|
| 79 |
/* |
| 80 |
** test_vdotq_lane_f32_fpm_0: |
| 81 |
** msr fpmr, x0 |
| 82 |
** fdot v0.4s, v1.16b, v2.4b\[0\] |
| 83 |
** ret |
| 84 |
*/ |
| 85 |
float32x4_t |
| 86 |
test_vdotq_lane_f32_fpm_0 (float32x4_t a, mfloat8x16_t b, mfloat8x8_t c, fpm_t d) |
| 87 |
{ |
| 88 |
return vdotq_lane_f32_mf8_fpm (a, b, c, 0, d); |
| 89 |
} |
| 90 |
|
| 91 |
/* |
| 92 |
** test_vdotq_lane_f32_fpm_1: |
| 93 |
** msr fpmr, x0 |
| 94 |
** fdot v0.4s, v1.16b, v2.4b\[1\] |
| 95 |
** ret |
| 96 |
*/ |
| 97 |
float32x4_t |
| 98 |
test_vdotq_lane_f32_fpm_1 (float32x4_t a, mfloat8x16_t b, mfloat8x8_t c, fpm_t d) |
| 99 |
{ |
| 100 |
return vdotq_lane_f32_mf8_fpm (a, b, c, 1, d); |
| 101 |
} |
| 102 |
|
| 103 |
/* |
| 104 |
** test_vdotq_laneq_f32_fpm_0: |
| 105 |
** msr fpmr, x0 |
| 106 |
** fdot v0.4s, v1.16b, v2.4b\[0\] |
| 107 |
** ret |
| 108 |
*/ |
| 109 |
float32x4_t |
| 110 |
test_vdotq_laneq_f32_fpm_0 (float32x4_t a, mfloat8x16_t b, mfloat8x16_t c, fpm_t d) |
| 111 |
{ |
| 112 |
return vdotq_laneq_f32_mf8_fpm (a, b, c, 0, d); |
| 113 |
} |
| 114 |
|
| 115 |
/* |
| 116 |
** test_vdotq_laneq_f32_fpm_3: |
| 117 |
** msr fpmr, x0 |
| 118 |
** fdot v0.4s, v1.16b, v2.4b\[3\] |
| 119 |
** ret |
| 120 |
*/ |
| 121 |
float32x4_t |
| 122 |
test_vdotq_laneq_f32_fpm_3 (float32x4_t a, mfloat8x16_t b, mfloat8x16_t c, fpm_t d) |
| 123 |
{ |
| 124 |
return vdotq_laneq_f32_mf8_fpm (a, b, c, 3, d); |
| 125 |
} |