| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-require-effective-target arm_neon_ok } */ |
| 3 |
/* { dg-require-effective-target arm_little_endian } */ |
| 4 |
/* { dg-options "-O2" } */ |
| 5 |
/* { dg-add-options arm_neon } */ |
| 6 |
|
| 7 |
#include <arm_neon.h> |
| 8 |
|
| 9 |
uint8x8_t |
| 10 |
tst_vext_u8 (uint8x8_t __a, uint8x8_t __b) |
| 11 |
{ |
| 12 |
uint8x8_t __mask1 = {2, 3, 4, 5, 6, 7, 8, 9}; |
| 13 |
|
| 14 |
return __builtin_shuffle ( __a, __b, __mask1) ; |
| 15 |
} |
| 16 |
|
| 17 |
uint8x8_t |
| 18 |
tst_vext_u8_rotate (uint8x8_t __a) |
| 19 |
{ |
| 20 |
uint8x8_t __mask1 = {2, 3, 4, 5, 6, 7, 0, 1}; |
| 21 |
return __builtin_shuffle ( __a, __mask1) ; |
| 22 |
} |
| 23 |
|
| 24 |
uint16x4_t |
| 25 |
tst_vext_u16 (uint16x4_t __a, uint16x4_t __b) |
| 26 |
{ |
| 27 |
uint16x4_t __mask1 = {2, 3, 4, 5}; |
| 28 |
return __builtin_shuffle ( __a, __b, __mask1) ; |
| 29 |
} |
| 30 |
|
| 31 |
uint16x4_t |
| 32 |
tst_vext_u16_rotate (uint16x4_t __a) |
| 33 |
{ |
| 34 |
uint16x4_t __mask1 = {2, 3, 0, 1}; |
| 35 |
return __builtin_shuffle ( __a, __mask1) ; |
| 36 |
} |
| 37 |
|
| 38 |
uint32x2_t |
| 39 |
tst_vext_u32 (uint32x2_t __a, uint32x2_t __b) |
| 40 |
{ |
| 41 |
uint32x2_t __mask1 = {1, 2}; |
| 42 |
return __builtin_shuffle ( __a, __b, __mask1) ; |
| 43 |
} |
| 44 |
|
| 45 |
/* This one is mapped into vrev64.32. */ |
| 46 |
uint32x2_t |
| 47 |
tst_vext_u32_rotate (uint32x2_t __a) |
| 48 |
{ |
| 49 |
uint32x2_t __mask1 = {1, 0}; |
| 50 |
return __builtin_shuffle ( __a, __mask1) ; |
| 51 |
} |
| 52 |
|
| 53 |
uint8x16_t |
| 54 |
tst_vextq_u8 (uint8x16_t __a, uint8x16_t __b) |
| 55 |
{ |
| 56 |
uint8x16_t __mask1 = {4, 5, 6, 7, 8, 9, 10, 11, |
| 57 |
12, 13, 14, 15, 16, 17, 18, 19}; |
| 58 |
return __builtin_shuffle ( __a, __b, __mask1) ; |
| 59 |
} |
| 60 |
|
| 61 |
uint8x16_t |
| 62 |
tst_vextq_u8_rotate (uint8x16_t __a) |
| 63 |
{ |
| 64 |
uint8x16_t __mask1 = {4, 5, 6, 7, 8, 9, 10, 11, |
| 65 |
12, 13, 14, 15, 0, 1, 2, 3}; |
| 66 |
return __builtin_shuffle ( __a, __mask1) ; |
| 67 |
} |
| 68 |
|
| 69 |
uint16x8_t |
| 70 |
tst_vextq_u16 (uint16x8_t __a, uint16x8_t __b) |
| 71 |
{ |
| 72 |
uint16x8_t __mask1 = {2, 3, 4, 5, 6, 7, 8, 9}; |
| 73 |
return __builtin_shuffle ( __a, __b, __mask1) ; |
| 74 |
} |
| 75 |
|
| 76 |
uint16x8_t |
| 77 |
tst_vextq_u16_rotate (uint16x8_t __a) |
| 78 |
{ |
| 79 |
uint16x8_t __mask1 = {2, 3, 4, 5, 6, 7, 0, 1}; |
| 80 |
return __builtin_shuffle ( __a, __mask1) ; |
| 81 |
} |
| 82 |
|
| 83 |
uint32x4_t |
| 84 |
tst_vextq_u32 (uint32x4_t __a, uint32x4_t __b) |
| 85 |
{ |
| 86 |
uint32x4_t __mask1 = {1, 2, 3, 4}; |
| 87 |
return __builtin_shuffle ( __a, __b, __mask1) ; |
| 88 |
} |
| 89 |
|
| 90 |
uint32x4_t |
| 91 |
tst_vextq_u32_rotate (uint32x4_t __a) |
| 92 |
{ |
| 93 |
uint32x4_t __mask1 = {1, 2, 3, 0}; |
| 94 |
return __builtin_shuffle ( __a, __mask1) ; |
| 95 |
} |
| 96 |
|
| 97 |
uint64x2_t |
| 98 |
tst_vextq_u64 (uint64x2_t __a, uint64x2_t __b) |
| 99 |
{ |
| 100 |
uint64x2_t __mask1 = {1, 2}; |
| 101 |
return __builtin_shuffle ( __a, __b, __mask1) ; |
| 102 |
} |
| 103 |
|
| 104 |
uint64x2_t |
| 105 |
tst_vextq_u64_rotate (uint64x2_t __a) |
| 106 |
{ |
| 107 |
uint64x2_t __mask1 = {1, 0}; |
| 108 |
return __builtin_shuffle ( __a, __mask1) ; |
| 109 |
} |
| 110 |
|
| 111 |
/* { dg-final {scan-assembler-times "vext\.8\\t" 4} } */ |
| 112 |
/* { dg-final {scan-assembler-times "vext\.16\\t" 4} } */ |
| 113 |
/* { dg-final {scan-assembler-times "vext\.32\\t" 3} } */ |
| 114 |
/* { dg-final {scan-assembler-times "vrev64\.32\\t" 1} } */ |
| 115 |
/* { dg-final {scan-assembler-times "vext\.64\\t" 2} } */ |