| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv64gcv_zvfhmin -mabi=lp64d -O3 -mrvv-vector-bits=zvl" } */ |
| 3 |
|
| 4 |
#include <stdint-gcc.h> |
| 5 |
|
| 6 |
typedef int8_t vnx16qi __attribute__((vector_size (16))); |
| 7 |
typedef int16_t vnx8hi __attribute__((vector_size (16))); |
| 8 |
typedef int32_t vnx4si __attribute__((vector_size (16))); |
| 9 |
typedef int64_t vnx2di __attribute__((vector_size (16))); |
| 10 |
typedef uint8_t vnx16uqi __attribute__((vector_size (16))); |
| 11 |
typedef uint16_t vnx8uhi __attribute__((vector_size (16))); |
| 12 |
typedef uint32_t vnx4usi __attribute__((vector_size (16))); |
| 13 |
typedef uint64_t vnx2udi __attribute__((vector_size (16))); |
| 14 |
|
| 15 |
typedef _Float16 vnx8hf __attribute__((vector_size (16))); |
| 16 |
typedef float vnx4sf __attribute__((vector_size (16))); |
| 17 |
typedef double vnx2df __attribute__((vector_size (16))); |
| 18 |
|
| 19 |
#define MASK_16 0, 17, 2, 19, 4, 21, 6, 23, 8, 25, 10, 27, 12, 29, 14, 31 |
| 20 |
#define MASK_8 0, 9, 2, 11, 4, 13, 6, 15 |
| 21 |
#define MASK_4 0, 5, 2, 7 |
| 22 |
#define MASK_2 0, 3 |
| 23 |
|
| 24 |
void __attribute__ ((noipa)) |
| 25 |
merge0 (vnx16qi x, vnx16qi y, vnx16qi *out) |
| 26 |
{ |
| 27 |
vnx16qi v = __builtin_shufflevector ((vnx16qi) x, (vnx16qi) y, MASK_16); |
| 28 |
*(vnx16qi*)out = v; |
| 29 |
} |
| 30 |
|
| 31 |
void __attribute__ ((noipa)) |
| 32 |
merge1 (vnx16uqi x, vnx16uqi y, vnx16uqi *out) |
| 33 |
{ |
| 34 |
vnx16uqi v = __builtin_shufflevector ((vnx16uqi) x, (vnx16uqi) y, MASK_16); |
| 35 |
*(vnx16uqi*)out = v; |
| 36 |
} |
| 37 |
|
| 38 |
void __attribute__ ((noipa)) |
| 39 |
merge2 (vnx8hi x, vnx8hi y, vnx8hi *out) |
| 40 |
{ |
| 41 |
vnx8hi v = __builtin_shufflevector ((vnx8hi) x, (vnx8hi) y, MASK_8); |
| 42 |
*(vnx8hi*)out = v; |
| 43 |
} |
| 44 |
|
| 45 |
void __attribute__ ((noipa)) |
| 46 |
merge3 (vnx8uhi x, vnx8uhi y, vnx8uhi *out) |
| 47 |
{ |
| 48 |
vnx8uhi v = __builtin_shufflevector ((vnx8uhi) x, (vnx8uhi) y, MASK_8); |
| 49 |
*(vnx8uhi*)out = v; |
| 50 |
} |
| 51 |
|
| 52 |
void __attribute__ ((noipa)) |
| 53 |
merge4 (vnx4si x, vnx4si y, vnx4si *out) |
| 54 |
{ |
| 55 |
vnx4si v = __builtin_shufflevector ((vnx4si) x, (vnx4si) y, MASK_4); |
| 56 |
*(vnx4si*)out = v; |
| 57 |
} |
| 58 |
|
| 59 |
void __attribute__ ((noipa)) |
| 60 |
merge5 (vnx4usi x, vnx4usi y, vnx4usi *out) |
| 61 |
{ |
| 62 |
vnx4usi v = __builtin_shufflevector ((vnx4usi) x, (vnx4usi) y, MASK_4); |
| 63 |
*(vnx4usi*)out = v; |
| 64 |
} |
| 65 |
|
| 66 |
void __attribute__ ((noipa)) |
| 67 |
merge6 (vnx2di x, vnx2di y, vnx2di *out) |
| 68 |
{ |
| 69 |
vnx2di v = __builtin_shufflevector ((vnx2di) x, (vnx2di) y, MASK_2); |
| 70 |
*(vnx2di*)out = v; |
| 71 |
} |
| 72 |
|
| 73 |
void __attribute__ ((noipa)) |
| 74 |
merge7 (vnx2udi x, vnx2udi y, vnx2udi *out) |
| 75 |
{ |
| 76 |
vnx2udi v = __builtin_shufflevector ((vnx2udi) x, (vnx2udi) y, MASK_2); |
| 77 |
*(vnx2udi*)out = v; |
| 78 |
} |
| 79 |
|
| 80 |
void __attribute__ ((noipa)) |
| 81 |
merge8 (vnx8hf x, vnx8hf y, vnx8hf *out) |
| 82 |
{ |
| 83 |
vnx8hf v = __builtin_shufflevector ((vnx8hf) x, (vnx8hf) y, MASK_8); |
| 84 |
*(vnx8hf*)out = v; |
| 85 |
} |
| 86 |
|
| 87 |
void __attribute__ ((noipa)) |
| 88 |
merge9 (vnx4sf x, vnx4sf y, vnx4sf *out) |
| 89 |
{ |
| 90 |
vnx4sf v = __builtin_shufflevector ((vnx4sf) x, (vnx4sf) y, MASK_4); |
| 91 |
*(vnx4sf*)out = v; |
| 92 |
} |
| 93 |
|
| 94 |
void __attribute__ ((noipa)) |
| 95 |
merge10 (vnx2df x, vnx2df y, vnx2df *out) |
| 96 |
{ |
| 97 |
vnx2df v = __builtin_shufflevector ((vnx2df) x, (vnx2df) y, MASK_2); |
| 98 |
*(vnx2df*)out = v; |
| 99 |
} |
| 100 |
|
| 101 |
/* { dg-final { scan-assembler-times {\tvmerge.vvm} 11 } } */ |