| 1 |
/* { dg-do compile { target { ! ia32 } } } */ |
| 2 |
/* { dg-require-effective-target masm_intel } */ |
| 3 |
/* { dg-options "-O2 -mamx-transpose -mamx-bf16 -mamx-complex -mamx-fp16 -mamx-tf32 -masm=intel" } */ |
| 4 |
/* { dg-final { scan-assembler "ttdpbf16ps\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2+\[^\n\]*%tmm3" } } */ |
| 5 |
/* { dg-final { scan-assembler "ttdpfp16ps\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2+\[^\n\]*%tmm3" } } */ |
| 6 |
/* { dg-final { scan-assembler "ttransposed\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2" } } */ |
| 7 |
/* { dg-final { scan-assembler "t2rpntlvwz0\[ \\t]%tmm\[0-9\]" } } */ |
| 8 |
/* { dg-final { scan-assembler "t2rpntlvwz0t1\[ \\t]%tmm\[0-9\]" } } */ |
| 9 |
/* { dg-final { scan-assembler "t2rpntlvwz1\[ \\t]%tmm\[0-9\]" } } */ |
| 10 |
/* { dg-final { scan-assembler "t2rpntlvwz1t1\[ \\t]%tmm\[0-9\]" } } */ |
| 11 |
/* { dg-final { scan-assembler "tconjtcmmimfp16ps\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2+\[^\n\]*%tmm3" } } */ |
| 12 |
/* { dg-final { scan-assembler "tconjtfp16\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2" } } */ |
| 13 |
/* { dg-final { scan-assembler "ttcmmimfp16ps\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2+\[^\n\]*%tmm3" } } */ |
| 14 |
/* { dg-final { scan-assembler "ttcmmrlfp16ps\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2+\[^\n\]*%tmm3" } } */ |
| 15 |
/* { dg-final { scan-assembler "ttmmultf32ps\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2+\[^\n\]*%tmm3" } } */ |
| 16 |
#include <immintrin.h> |
| 17 |
|
| 18 |
extern const void* base; |
| 19 |
extern const int stride; |
| 20 |
|
| 21 |
void TEST() |
| 22 |
{ |
| 23 |
_tile_tdpbf16ps (1, 2, 3); |
| 24 |
_tile_tdpfp16ps (1, 2, 3); |
| 25 |
_tile_transposed (1, 2); |
| 26 |
_tile_2rpntlvwz0 (5, base, stride); |
| 27 |
_tile_2rpntlvwz0t1 (4, base, stride); |
| 28 |
_tile_2rpntlvwz1 (3, base, stride); |
| 29 |
_tile_2rpntlvwz1t1 (2, base, stride); |
| 30 |
_tile_conjtcmmimfp16ps (1, 2, 3); |
| 31 |
_tile_conjtfp16 (1, 2); |
| 32 |
_tile_tcmmimfp16ps (1, 2, 3); |
| 33 |
_tile_tcmmrlfp16ps (1, 2, 3); |
| 34 |
_tile_tmmultf32ps (1, 2, 3); |
| 35 |
} |