-rw-r--r-- 2108 libmceliece-20230612/crypto_kem/348864/avx/vec128.h raw
/* This file is for functions related to 128-bit vectors including functions for bitsliced field operations */ #ifndef VEC128_H #define VEC128_H #define vec128_mul_asm CRYPTO_SHARED_NAMESPACE(vec128_mul_asm) #define vec128_mul_GF CRYPTO_NAMESPACE(vec128_mul_GF) #include "params.h" #include <stdint.h> #include <smmintrin.h> typedef __m128i vec128; static inline vec128 vec128_set1_16b(uint16_t a) { return _mm_set1_epi16(a); } static inline vec128 vec128_setzero() { return _mm_setzero_si128(); } #define vec128_extract(a, i) ((uint64_t) _mm_extract_epi64((vec128) (a), (i))) static inline int vec128_testz(vec128 a) { return _mm_testz_si128(a, a); } static inline vec128 vec128_and(vec128 a, vec128 b) { return _mm_and_si128(a, b); } static inline vec128 vec128_xor(vec128 a, vec128 b) { return _mm_xor_si128(a, b); } static inline vec128 vec128_or(vec128 a, vec128 b) { return _mm_or_si128(a, b); } #define vec128_sll_2x(a, s) ((vec128) _mm_slli_epi64((vec128) (a), (s))) #define vec128_srl_2x(a, s) ((vec128) _mm_srli_epi64((vec128) (a), (s))) static inline vec128 vec128_set2x(uint64_t a0, uint64_t a1) { return _mm_set_epi64x(a1, a0); } static inline vec128 vec128_unpack_low(vec128 a, vec128 b) { return _mm_unpacklo_epi64(a, b); } static inline vec128 vec128_unpack_high(vec128 a, vec128 b) { return _mm_unpackhi_epi64(a, b); } static inline vec128 vec128_setbits(uint64_t a) { return _mm_set1_epi64x(-a); } static inline void vec128_copy(vec128 *dest, vec128 *src) { int i; for (i = 0; i < GFBITS; i++) dest[i] = src[i]; } static inline void vec128_add(vec128 *c, vec128 *a, vec128 *b) { int i; for (i = 0; i < GFBITS; i++) c[i] = vec128_xor(a[i], b[i]); } static inline vec128 vec128_or_reduce(vec128 * a) { int i; vec128 ret; ret = a[0]; for (i = 1; i < GFBITS; i++) ret = vec128_or(ret, a[i]); return ret; } extern void vec128_mul_asm(vec128 *, vec128 *, const vec128 *, int); /* bitsliced field multiplications */ static inline void vec128_mul(vec128 *h, vec128 *f, const vec128 *g) { vec128_mul_asm(h, f, g, 16); } #endif