| 1 |
/* { dg-do compile { target { ! ia32 } } } */ |
| 2 |
/* { dg-options "-O2 -mavx512vl -mavx512bw -masm=att" } */ |
| 3 |
|
| 4 |
typedef char V1 __attribute__((vector_size (16))); |
| 5 |
|
| 6 |
void |
| 7 |
f1 (V1 x) |
| 8 |
{ |
| 9 |
register V1 a __asm ("xmm16"); |
| 10 |
a = x; |
| 11 |
asm volatile ("" : "+v" (a)); |
| 12 |
a = __builtin_shuffle (a, (V1) { 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5 }); |
| 13 |
asm volatile ("" : "+v" (a)); |
| 14 |
} |
| 15 |
|
| 16 |
/* { dg-final { scan-assembler-times "vpalignr\[^\n\r]*\\\$6\[^\n\r]*%xmm16\[^\n\r]*%xmm16\[^\n\r]*%xmm16" 1 } } */ |
| 17 |
|
| 18 |
typedef short V2 __attribute__((vector_size (16))); |
| 19 |
|
| 20 |
void |
| 21 |
f2 (V2 x) |
| 22 |
{ |
| 23 |
register V2 a __asm ("xmm16"); |
| 24 |
a = x; |
| 25 |
asm volatile ("" : "+v" (a)); |
| 26 |
a = __builtin_shuffle (a, (V2) { 5, 6, 7, 0, 1, 2, 3, 4 }); |
| 27 |
asm volatile ("" : "+v" (a)); |
| 28 |
} |
| 29 |
|
| 30 |
/* { dg-final { scan-assembler-times "vpalignr\[^\n\r]*\\\$10\[^\n\r]*%xmm16\[^\n\r]*%xmm16\[^\n\r]*%xmm16" 1 } } */ |
| 31 |
|
| 32 |
typedef int V3 __attribute__((vector_size (16))); |
| 33 |
|
| 34 |
void |
| 35 |
f3 (V3 x) |
| 36 |
{ |
| 37 |
register V3 a __asm ("xmm16"); |
| 38 |
a = x; |
| 39 |
asm volatile ("" : "+v" (a)); |
| 40 |
a = __builtin_shuffle (a, (V3) { 3, 0, 1, 2 }); |
| 41 |
asm volatile ("" : "+v" (a)); |
| 42 |
} |
| 43 |
|
| 44 |
/* { dg-final { scan-assembler-times "vpshufd\[^\n\r]*\\\$147\[^\n\r]*%xmm16\[^\n\r]*%xmm16" 1 } } */ |
| 45 |
|
| 46 |
typedef long long V4 __attribute__((vector_size (16))); |
| 47 |
|
| 48 |
void |
| 49 |
f4 (V4 x) |
| 50 |
{ |
| 51 |
register V4 a __asm ("xmm16"); |
| 52 |
a = x; |
| 53 |
asm volatile ("" : "+v" (a)); |
| 54 |
a = __builtin_shuffle (a, (V4) { 1, 0 }); |
| 55 |
asm volatile ("" : "+v" (a)); |
| 56 |
} |
| 57 |
|
| 58 |
/* { dg-final { scan-assembler-times "vpalignr\[^\n\r]*\\\$8\[^\n\r]*%xmm16\[^\n\r]*%xmm16\[^\n\r]*%xmm16" 1 } } */ |
| 59 |
|
| 60 |
typedef float V5 __attribute__((vector_size (16))); |
| 61 |
|
| 62 |
void |
| 63 |
f5 (V5 x) |
| 64 |
{ |
| 65 |
register V5 a __asm ("xmm16"); |
| 66 |
a = x; |
| 67 |
asm volatile ("" : "+v" (a)); |
| 68 |
a = __builtin_shuffle (a, (V3) { 3, 0, 1, 2 }); |
| 69 |
asm volatile ("" : "+v" (a)); |
| 70 |
} |
| 71 |
|
| 72 |
/* { dg-final { scan-assembler-times "vpermilps\[^\n\r]*\\\$147\[^\n\r]*%xmm16\[^\n\r]*%xmm16" 1 } } */ |
| 73 |
|
| 74 |
typedef double V6 __attribute__((vector_size (16))); |
| 75 |
|
| 76 |
void |
| 77 |
f6 (V6 x) |
| 78 |
{ |
| 79 |
register V6 a __asm ("xmm16"); |
| 80 |
a = x; |
| 81 |
asm volatile ("" : "+v" (a)); |
| 82 |
a = __builtin_shuffle (a, (V4) { 1, 0 }); |
| 83 |
asm volatile ("" : "+v" (a)); |
| 84 |
} |
| 85 |
|
| 86 |
/* { dg-final { scan-assembler-times "vpermilpd\[^\n\r]*\\\$1\[^\n\r]*%xmm16\[^\n\r]*%xmm16" 1 } } */ |