| 1 |
/* { dg-do assemble } */ |
| 2 |
/* { dg-require-effective-target arm_hard_ok } */ |
| 3 |
/* { dg-require-effective-target arm_v8_2a_bf16_neon_ok } */ |
| 4 |
/* { dg-add-options arm_v8_2a_bf16_neon } */ |
| 5 |
/* { dg-additional-options "-save-temps -O2 -mfloat-abi=hard" } */ |
| 6 |
/* { dg-final { check-function-bodies "**" "" } } */ |
| 7 |
|
| 8 |
#include "arm_neon.h" |
| 9 |
|
| 10 |
/* |
| 11 |
**test_vfmabq_f32: |
| 12 |
** ... |
| 13 |
** vfmab.bf16 q0, q1, q2 |
| 14 |
** bx lr |
| 15 |
*/ |
| 16 |
float32x4_t |
| 17 |
test_vfmabq_f32 (float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) |
| 18 |
{ |
| 19 |
return vbfmlalbq_f32 (r, a, b); |
| 20 |
} |
| 21 |
|
| 22 |
/* |
| 23 |
**test_vfmatq_f32: |
| 24 |
** ... |
| 25 |
** vfmat.bf16 q0, q1, q2 |
| 26 |
** bx lr |
| 27 |
*/ |
| 28 |
float32x4_t |
| 29 |
test_vfmatq_f32 (float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) |
| 30 |
{ |
| 31 |
return vbfmlaltq_f32 (r, a, b); |
| 32 |
} |
| 33 |
|
| 34 |
/* |
| 35 |
**test_vfmabq_lane_f32: |
| 36 |
** ... |
| 37 |
** vfmab.bf16 q0, q1, d4\[0\] |
| 38 |
** bx lr |
| 39 |
*/ |
| 40 |
float32x4_t |
| 41 |
test_vfmabq_lane_f32 (float32x4_t r, bfloat16x8_t a, bfloat16x4_t b) |
| 42 |
{ |
| 43 |
return vbfmlalbq_lane_f32 (r, a, b, 0); |
| 44 |
} |
| 45 |
|
| 46 |
/* |
| 47 |
**test_vfmatq_lane_f32: |
| 48 |
** ... |
| 49 |
** vfmat.bf16 q0, q1, d4\[2\] |
| 50 |
** bx lr |
| 51 |
*/ |
| 52 |
float32x4_t |
| 53 |
test_vfmatq_lane_f32 (float32x4_t r, bfloat16x8_t a, bfloat16x4_t b) |
| 54 |
{ |
| 55 |
return vbfmlaltq_lane_f32 (r, a, b, 2); |
| 56 |
} |
| 57 |
|
| 58 |
/* |
| 59 |
**test_vfmabq_laneq_f32: |
| 60 |
** ... |
| 61 |
** vfmab.bf16 q0, q1, d5\[1\] |
| 62 |
** bx lr |
| 63 |
*/ |
| 64 |
float32x4_t |
| 65 |
test_vfmabq_laneq_f32 (float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) |
| 66 |
{ |
| 67 |
return vbfmlalbq_laneq_f32 (r, a, b, 5); |
| 68 |
} |
| 69 |
|
| 70 |
/* |
| 71 |
**test_vfmatq_laneq_f32: |
| 72 |
** ... |
| 73 |
** vfmat.bf16 q0, q1, d5\[3\] |
| 74 |
** bx lr |
| 75 |
*/ |
| 76 |
float32x4_t |
| 77 |
test_vfmatq_laneq_f32 (float32x4_t r, bfloat16x8_t a, bfloat16x8_t b) |
| 78 |
{ |
| 79 |
return vbfmlaltq_laneq_f32 (r, a, b, 7); |
| 80 |
} |