| 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 vnx64qi __attribute__((vector_size (64))); |
| 7 |
typedef int16_t vnx32hi __attribute__((vector_size (64))); |
| 8 |
typedef int32_t vnx16si __attribute__((vector_size (64))); |
| 9 |
typedef int64_t vnx8di __attribute__((vector_size (64))); |
| 10 |
typedef uint8_t vnx64uqi __attribute__((vector_size (64))); |
| 11 |
typedef uint16_t vnx32uhi __attribute__((vector_size (64))); |
| 12 |
typedef uint32_t vnx16usi __attribute__((vector_size (64))); |
| 13 |
typedef uint64_t vnx8udi __attribute__((vector_size (64))); |
| 14 |
|
| 15 |
typedef _Float16 vnx32hf __attribute__((vector_size (64))); |
| 16 |
typedef float vnx16sf __attribute__((vector_size (64))); |
| 17 |
typedef double vnx8df __attribute__((vector_size (64))); |
| 18 |
|
| 19 |
#define MASK_64 0, 65, 2, 67, 4, 69, 6, 71, 8, 73, \ |
| 20 |
10, 75, 12, 77, 14, 79, 16, 81, 18, 83, \ |
| 21 |
20, 85, 22, 87, 24, 89, 26, 91, 28, 93, 30, 95, \ |
| 22 |
32, 97, 34, 99, 36, 101, 38, 103, 40, 105, \ |
| 23 |
42, 107, 44, 109, 46, 111, 48, 113, 50, 115, \ |
| 24 |
52, 117, 54, 119, 56, 121, 58, 123, 60, 125, \ |
| 25 |
62, 127 |
| 26 |
#define MASK_32 0, 33, 2, 35, 4, 37, 6, 39, 8, 41, \ |
| 27 |
10, 43, 12, 45, 14, 47, 16, 49, 18, 51, \ |
| 28 |
20, 53, 22, 55, 24, 57, 26, 59, 28, 61, 30, 63 |
| 29 |
#define MASK_16 0, 17, 2, 19, 4, 21, 6, 23, 8, 25, 10, 27, 12, 29, 14, 31 |
| 30 |
#define MASK_8 0, 9, 2, 11, 4, 13, 6, 15 |
| 31 |
|
| 32 |
void __attribute__ ((noipa)) |
| 33 |
merge0 (vnx64qi x, vnx64qi y, vnx64qi *out) |
| 34 |
{ |
| 35 |
vnx64qi v = __builtin_shufflevector ((vnx64qi) x, (vnx64qi) y, MASK_64); |
| 36 |
*(vnx64qi*)out = v; |
| 37 |
} |
| 38 |
|
| 39 |
void __attribute__ ((noipa)) |
| 40 |
merge1 (vnx64uqi x, vnx64uqi y, vnx64uqi *out) |
| 41 |
{ |
| 42 |
vnx64uqi v = __builtin_shufflevector ((vnx64uqi) x, (vnx64uqi) y, MASK_64); |
| 43 |
*(vnx64uqi*)out = v; |
| 44 |
} |
| 45 |
|
| 46 |
void __attribute__ ((noipa)) |
| 47 |
merge2 (vnx32hi x, vnx32hi y, vnx32hi *out) |
| 48 |
{ |
| 49 |
vnx32hi v = __builtin_shufflevector ((vnx32hi) x, (vnx32hi) y, MASK_32); |
| 50 |
*(vnx32hi*)out = v; |
| 51 |
} |
| 52 |
|
| 53 |
void __attribute__ ((noipa)) |
| 54 |
merge3 (vnx32uhi x, vnx32uhi y, vnx32uhi *out) |
| 55 |
{ |
| 56 |
vnx32uhi v = __builtin_shufflevector ((vnx32uhi) x, (vnx32uhi) y, MASK_32); |
| 57 |
*(vnx32uhi*)out = v; |
| 58 |
} |
| 59 |
|
| 60 |
void __attribute__ ((noipa)) |
| 61 |
merge4 (vnx16si x, vnx16si y, vnx16si *out) |
| 62 |
{ |
| 63 |
vnx16si v = __builtin_shufflevector ((vnx16si) x, (vnx16si) y, MASK_16); |
| 64 |
*(vnx16si*)out = v; |
| 65 |
} |
| 66 |
|
| 67 |
void __attribute__ ((noipa)) |
| 68 |
merge5 (vnx16usi x, vnx16usi y, vnx16usi *out) |
| 69 |
{ |
| 70 |
vnx16usi v = __builtin_shufflevector ((vnx16usi) x, (vnx16usi) y, MASK_16); |
| 71 |
*(vnx16usi*)out = v; |
| 72 |
} |
| 73 |
|
| 74 |
void __attribute__ ((noipa)) |
| 75 |
merge6 (vnx8di x, vnx8di y, vnx8di *out) |
| 76 |
{ |
| 77 |
vnx8di v = __builtin_shufflevector ((vnx8di) x, (vnx8di) y, MASK_8); |
| 78 |
*(vnx8di*)out = v; |
| 79 |
} |
| 80 |
|
| 81 |
void __attribute__ ((noipa)) |
| 82 |
merge7 (vnx8udi x, vnx8udi y, vnx8udi *out) |
| 83 |
{ |
| 84 |
vnx8udi v = __builtin_shufflevector ((vnx8udi) x, (vnx8udi) y, MASK_8); |
| 85 |
*(vnx8udi*)out = v; |
| 86 |
} |
| 87 |
|
| 88 |
void __attribute__ ((noipa)) |
| 89 |
merge8 (vnx32hf x, vnx32hf y, vnx32hf *out) |
| 90 |
{ |
| 91 |
vnx32hf v = __builtin_shufflevector ((vnx32hf) x, (vnx32hf) y, MASK_32); |
| 92 |
*(vnx32hf*)out = v; |
| 93 |
} |
| 94 |
|
| 95 |
void __attribute__ ((noipa)) |
| 96 |
merge9 (vnx16sf x, vnx16sf y, vnx16sf *out) |
| 97 |
{ |
| 98 |
vnx16sf v = __builtin_shufflevector ((vnx16sf) x, (vnx16sf) y, MASK_16); |
| 99 |
*(vnx16sf*)out = v; |
| 100 |
} |
| 101 |
|
| 102 |
void __attribute__ ((noipa)) |
| 103 |
merge10 (vnx8df x, vnx8df y, vnx8df *out) |
| 104 |
{ |
| 105 |
vnx8df v = __builtin_shufflevector ((vnx8df) x, (vnx8df) y, MASK_8); |
| 106 |
*(vnx8df*)out = v; |
| 107 |
} |
| 108 |
|
| 109 |
/* { dg-final { scan-assembler-times {\tvmerge.vvm} 11 } } */ |