| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=x86-64-v3 -mavx10.2 -O2" } */ |
| 3 |
/* { dg-final { scan-assembler-times "vmulbf16\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */ |
| 4 |
/* { dg-final { scan-assembler-times "vaddbf16\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 5 |
/* { dg-final { scan-assembler-times "vdivbf16\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 6 |
/* { dg-final { scan-assembler-times "vsubbf16\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 7 |
/* { dg-final { scan-assembler-times "vmulbf16\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */ |
| 8 |
/* { dg-final { scan-assembler-times "vaddbf16\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 9 |
/* { dg-final { scan-assembler-times "vdivbf16\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 10 |
/* { dg-final { scan-assembler-times "vsubbf16\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 11 |
/* { dg-final { scan-assembler-times "vrcpbf16\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 12 |
/* { dg-final { scan-assembler-times "vrcpbf16\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ |
| 13 |
|
| 14 |
#include <immintrin.h> |
| 15 |
|
| 16 |
typedef __bf16 v16bf __attribute__ ((__vector_size__ (32))); |
| 17 |
typedef __bf16 v8bf __attribute__ ((__vector_size__ (16))); |
| 18 |
|
| 19 |
v16bf |
| 20 |
foo_mul_256 (v16bf a, v16bf b) |
| 21 |
{ |
| 22 |
return a * b; |
| 23 |
} |
| 24 |
|
| 25 |
v16bf |
| 26 |
foo_add_256 (v16bf a, v16bf b) |
| 27 |
{ |
| 28 |
return a + b; |
| 29 |
} |
| 30 |
|
| 31 |
v16bf |
| 32 |
foo_div_256 (v16bf a, v16bf b) |
| 33 |
{ |
| 34 |
return a / b; |
| 35 |
} |
| 36 |
|
| 37 |
v16bf |
| 38 |
foo_sub_256 (v16bf a, v16bf b) |
| 39 |
{ |
| 40 |
return a - b; |
| 41 |
} |
| 42 |
|
| 43 |
__attribute__((optimize("fast-math"))) |
| 44 |
v16bf |
| 45 |
foo_div_fast_math_256 (v16bf a, v16bf b) |
| 46 |
{ |
| 47 |
return a / b; |
| 48 |
} |
| 49 |
|
| 50 |
v8bf |
| 51 |
foo_mul_128 (v8bf a, v8bf b) |
| 52 |
{ |
| 53 |
return a * b; |
| 54 |
} |
| 55 |
|
| 56 |
v8bf |
| 57 |
foo_add_128 (v8bf a, v8bf b) |
| 58 |
{ |
| 59 |
return a + b; |
| 60 |
} |
| 61 |
|
| 62 |
v8bf |
| 63 |
foo_div_128 (v8bf a, v8bf b) |
| 64 |
{ |
| 65 |
return a / b; |
| 66 |
} |
| 67 |
|
| 68 |
v8bf |
| 69 |
foo_sub_128 (v8bf a, v8bf b) |
| 70 |
{ |
| 71 |
return a - b; |
| 72 |
} |
| 73 |
|
| 74 |
__attribute__((optimize("fast-math"))) |
| 75 |
v8bf |
| 76 |
foo_div_fast_math_128 (v8bf a, v8bf b) |
| 77 |
{ |
| 78 |
return a / b; |
| 79 |
} |