| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=x86-64-v3 -mavx10.2 -O2" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vfmadd132bf16\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vfmsub132bf16\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vfnmadd132bf16\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 6 |
/* { dg-final { scan-assembler-times "vfnmsub132bf16\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 7 |
/* { dg-final { scan-assembler-times "vfmadd132bf16\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 8 |
/* { dg-final { scan-assembler-times "vfmsub132bf16\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 9 |
/* { dg-final { scan-assembler-times "vfnmadd132bf16\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 10 |
/* { dg-final { scan-assembler-times "vfnmsub132bf16\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 11 |
|
| 12 |
#include <immintrin.h> |
| 13 |
|
| 14 |
typedef __bf16 v16bf __attribute__ ((__vector_size__ (32))); |
| 15 |
typedef __bf16 v8bf __attribute__ ((__vector_size__ (16))); |
| 16 |
|
| 17 |
v16bf |
| 18 |
foo_madd_256 (v16bf a, v16bf b, v16bf c) |
| 19 |
{ |
| 20 |
return a * b + c; |
| 21 |
} |
| 22 |
|
| 23 |
v16bf |
| 24 |
foo_msub_256 (v16bf a, v16bf b, v16bf c) |
| 25 |
{ |
| 26 |
return a * b - c; |
| 27 |
} |
| 28 |
|
| 29 |
v16bf |
| 30 |
foo_nmadd_256 (v16bf a, v16bf b, v16bf c) |
| 31 |
{ |
| 32 |
return -a * b + c; |
| 33 |
} |
| 34 |
|
| 35 |
v16bf |
| 36 |
foo_nmsub_256 (v16bf a, v16bf b, v16bf c) |
| 37 |
{ |
| 38 |
return -a * b - c; |
| 39 |
} |
| 40 |
|
| 41 |
v8bf |
| 42 |
foo_madd_128 (v8bf a, v8bf b, v8bf c) |
| 43 |
{ |
| 44 |
return a * b + c; |
| 45 |
} |
| 46 |
|
| 47 |
v8bf |
| 48 |
foo_msub_128 (v8bf a, v8bf b, v8bf c) |
| 49 |
{ |
| 50 |
return a * b - c; |
| 51 |
} |
| 52 |
|
| 53 |
v8bf |
| 54 |
foo_nmadd_128 (v8bf a, v8bf b, v8bf c) |
| 55 |
{ |
| 56 |
return -a * b + c; |
| 57 |
} |
| 58 |
|
| 59 |
v8bf |
| 60 |
foo_nmsub_128 (v8bf a, v8bf b, v8bf c) |
| 61 |
{ |
| 62 |
return -a * b - c; |
| 63 |
} |