| 1 |
// 20240508 djb: split out vec256_gf.h |
| 2 |
// 20240507 djb: added vec256_load, vec256_set8x, vec256_8x_shl, vec256_8x_shr, vec256_4x_shr, vec256_2x_swap64, vec256_extractbot |
| 3 |
|
| 4 |
#ifndef VEC256_H |
| 5 |
#define VEC256_H |
| 6 |
|
| 7 |
#include "vec128.h" |
| 8 |
|
| 9 |
#include <immintrin.h> |
| 10 |
|
| 11 |
typedef __m256i vec256; |
| 12 |
|
| 13 |
static inline vec256 vec256_load(const void *x) |
| 14 |
{ |
| 15 |
return _mm256_loadu_si256((const __m256i *) x); |
| 16 |
} |
| 17 |
|
| 18 |
static inline vec256 vec256_set1_16b(uint16_t a) |
| 19 |
{ |
| 20 |
return _mm256_set1_epi16(a); |
| 21 |
} |
| 22 |
|
| 23 |
static inline vec256 vec256_set1_32b(uint64_t a) |
| 24 |
{ |
| 25 |
return _mm256_set1_epi32(a); |
| 26 |
} |
| 27 |
|
| 28 |
static inline vec256 vec256_setzero() |
| 29 |
{ |
| 30 |
return _mm256_setzero_si256(); |
| 31 |
} |
| 32 |
|
| 33 |
static inline vec256 vec256_set4x(uint64_t a0, uint64_t a1, uint64_t a2, uint64_t a3) |
| 34 |
{ |
| 35 |
return _mm256_set_epi64x(a3, a2, a1, a0); |
| 36 |
} |
| 37 |
|
| 38 |
static inline vec256 vec256_set8x(uint32_t a0, uint32_t a1, uint32_t a2, uint32_t a3, uint32_t a4, uint32_t a5, uint32_t a6, uint32_t a7) |
| 39 |
{ |
| 40 |
return _mm256_set_epi32(a7, a6, a5, a4, a3, a2, a1, a0); |
| 41 |
} |
| 42 |
|
| 43 |
#define vec256_extract2x(a,i) ((vec128) _mm256_extractf128_si256((vec256) (a),(i))) |
| 44 |
#define vec256_extract(a,i) ((uint64_t) _mm256_extract_epi64((vec256) (a),(i))) |
| 45 |
#define vec256_extractbot _mm256_castsi256_si128 |
| 46 |
|
| 47 |
static inline vec256 vec256_8x_shl(vec256 a, int n) |
| 48 |
{ |
| 49 |
return _mm256_slli_epi32(a, n); |
| 50 |
} |
| 51 |
|
| 52 |
static inline vec256 vec256_8x_shr(vec256 a, int n) |
| 53 |
{ |
| 54 |
return _mm256_srli_epi32(a, n); |
| 55 |
} |
| 56 |
|
| 57 |
static inline vec256 vec256_4x_shr(vec256 a, int n) |
| 58 |
{ |
| 59 |
return _mm256_srli_epi64(a, n); |
| 60 |
} |
| 61 |
|
| 62 |
static inline vec256 vec256_2x_swap64(vec256 a) |
| 63 |
{ |
| 64 |
return _mm256_shuffle_epi32(a,0x4e); |
| 65 |
} |
| 66 |
|
| 67 |
static inline int vec256_testz(vec256 a) |
| 68 |
{ |
| 69 |
return _mm256_testz_si256(a, a); |
| 70 |
} |
| 71 |
|
| 72 |
static inline vec256 vec256_and(vec256 a, vec256 b) |
| 73 |
{ |
| 74 |
return _mm256_and_si256(a, b); |
| 75 |
} |
| 76 |
|
| 77 |
static inline vec256 vec256_xor(vec256 a, vec256 b) |
| 78 |
{ |
| 79 |
return _mm256_xor_si256(a, b); |
| 80 |
} |
| 81 |
|
| 82 |
static inline vec256 vec256_or(vec256 a, vec256 b) |
| 83 |
{ |
| 84 |
return _mm256_or_si256(a, b); |
| 85 |
} |
| 86 |
|
| 87 |
#define vec256_sll_4x(a, s) ((vec256) _mm256_slli_epi64((vec256) (a), (s))) |
| 88 |
#define vec256_srl_4x(a, s) ((vec256) _mm256_srli_epi64((vec256) (a), (s))) |
| 89 |
|
| 90 |
static inline vec256 vec256_unpack_low(vec256 a, vec256 b) |
| 91 |
{ |
| 92 |
return _mm256_permute2x128_si256 (a, b, 0x20); |
| 93 |
} |
| 94 |
|
| 95 |
static inline vec256 vec256_unpack_high(vec256 a, vec256 b) |
| 96 |
{ |
| 97 |
return _mm256_permute2x128_si256 (a, b, 0x31); |
| 98 |
} |
| 99 |
|
| 100 |
static inline vec256 vec256_unpack_low_2x(vec256 a, vec256 b) |
| 101 |
{ |
| 102 |
return _mm256_unpacklo_epi64 (a, b); |
| 103 |
} |
| 104 |
|
| 105 |
static inline vec256 vec256_unpack_high_2x(vec256 a, vec256 b) |
| 106 |
{ |
| 107 |
return _mm256_unpackhi_epi64 (a, b); |
| 108 |
} |
| 109 |
|
| 110 |
#endif |