| 1 |
/* { dg-do assemble { target { aarch64*-*-* } } } */ |
| 2 |
/* { dg-require-effective-target arm_v8_2a_bf16_neon_ok } */ |
| 3 |
/* { dg-add-options arm_v8_2a_bf16_neon } */ |
| 4 |
/* { dg-additional-options "-O2 --save-temps" } */ |
| 5 |
|
| 6 |
#include <arm_neon.h> |
| 7 |
|
| 8 |
bfloat16x4x2_t |
| 9 |
test_vld2_lane_bf16 (const bfloat16_t *ptr, bfloat16x4x2_t b) |
| 10 |
{ |
| 11 |
return vld2_lane_bf16 (ptr, b, 2); |
| 12 |
} |
| 13 |
|
| 14 |
bfloat16x8x2_t |
| 15 |
test_vld2q_lane_bf16 (const bfloat16_t *ptr, bfloat16x8x2_t b) |
| 16 |
{ |
| 17 |
return vld2q_lane_bf16 (ptr, b, 2); |
| 18 |
} |
| 19 |
|
| 20 |
/* { dg-final { scan-assembler-times "ld2\\t{v\[0-9\]+.h - v\[0-9\]+.h}\\\[2\\\], \\\[x0\\\]" 2 } } */ |
| 21 |
|
| 22 |
bfloat16x4x3_t |
| 23 |
test_vld3_lane_bf16 (const bfloat16_t *ptr, bfloat16x4x3_t b) |
| 24 |
{ |
| 25 |
return vld3_lane_bf16 (ptr, b, 2); |
| 26 |
} |
| 27 |
|
| 28 |
bfloat16x8x3_t |
| 29 |
test_vld3q_lane_bf16 (const bfloat16_t *ptr, bfloat16x8x3_t b) |
| 30 |
{ |
| 31 |
return vld3q_lane_bf16 (ptr, b, 2); |
| 32 |
} |
| 33 |
|
| 34 |
/* { dg-final { scan-assembler-times "ld3\t{v\[0-9\]+.h - v\[0-9\]+.h}\\\[2\\\], \\\[x0\\\]" 2 } } */ |
| 35 |
|
| 36 |
bfloat16x4x4_t |
| 37 |
test_vld4_lane_bf16 (const bfloat16_t *ptr, bfloat16x4x4_t b) |
| 38 |
{ |
| 39 |
return vld4_lane_bf16 (ptr, b, 2); |
| 40 |
} |
| 41 |
|
| 42 |
bfloat16x8x4_t |
| 43 |
test_vld4q_lane_bf16 (const bfloat16_t *ptr, bfloat16x8x4_t b) |
| 44 |
{ |
| 45 |
return vld4q_lane_bf16 (ptr, b, 2); |
| 46 |
} |
| 47 |
|
| 48 |
/* { dg-final { scan-assembler-times "ld4\t{v\[0-9\]+.h - v\[0-9\]+.h}\\\[2\\\], \\\[x0\\\]" 2 } } */ |