uint64_t
siphash13(const unsigned char key[16], const unsigned char *m, size_t len) {
xmmi k,v02,v20,v13,v11,v33,mi;
uint64_t last7;
uint32_t lo, hi;
size_t i, blocks;
k = _mm_loadu_si128((xmmi *)(key + 0));
v02 = siphash_init[0].v;
v13 = siphash_init[1].v;
v02 = _mm_xor_si128(v02, _mm_unpacklo_epi64(k, k));
v13 = _mm_xor_si128(v13, _mm_unpackhi_epi64(k, k));
last7 = (uint64_t)(len & 0xff) << 56;
for (i = 0, blocks = (len & ~7); i < blocks; i += 8) {
mi = _mm_loadl_epi64((xmmi *)(m + i));
v13 = _mm_xor_si128(v13, _mm_slli_si128(mi, 8));
sipcompress()
v02 = _mm_xor_si128(v02, mi);
}
switch (len - blocks) {
case 7: last7 |= (uint64_t)m[i + 6] << 48;
case 6: last7 |= (uint64_t)m[i + 5] << 40;
case 5: last7 |= (uint64_t)m[i + 4] << 32;
case 4: last7 |= (uint64_t)m[i + 3] << 24;
case 3: last7 |= (uint64_t)m[i + 2] << 16;
case 2: last7 |= (uint64_t)m[i + 1] << 8;
case 1: last7 |= (uint64_t)m[i + 0] ;
case 0:
default:;
};
mi = _mm_unpacklo_epi32(_mm_cvtsi32_si128((uint32_t)last7),_mm_cvtsi32_si128((uint32_t)(last7 >> 32)));
v13 = _mm_xor_si128(v13, _mm_slli_si128(mi, 8));
sipcompress()
v02 = _mm_xor_si128(v02, mi);
v02 = _mm_xor_si128(v02, siphash_final.v);
sipcompress()
sipcompress()
sipcompress()
v02 = _mm_xor_si128(v02, v13);
v02 = _mm_xor_si128(v02, _mm_shuffle_epi32(v02, _MM_SHUFFLE(1,0,3,2)));
lo = _mm_cvtsi128_si32(v02);
hi = _mm_cvtsi128_si32(_mm_srli_si128(v02, 4));
return ((uint64_t)hi << 32) | lo;
}
uint64_t
siphash(const unsigned char key[16], const unsigned char *m, size_t len) {
uint64_t v0, v1, v2, v3;
uint64_t mi, k0, k1;
uint64_t last7;
size_t i, blocks;
k0 = U8TO64_LE(key + 0);
k1 = U8TO64_LE(key + 8);
v0 = k0 ^ 0x736f6d6570736575ull;
v1 = k1 ^ 0x646f72616e646f6dull;
v2 = k0 ^ 0x6c7967656e657261ull;
v3 = k1 ^ 0x7465646279746573ull;
last7 = (uint64_t)(len & 0xff) << 56;
#define sipcompress() \
v0 += v1; v2 += v3; \
v1 = ROTL64(v1,13); v3 = ROTL64(v3,16); \
v1 ^= v0; v3 ^= v2; \
v0 = ROTL64(v0,32); \
v2 += v1; v0 += v3; \
v1 = ROTL64(v1,17); v3 = ROTL64(v3,21); \
v1 ^= v2; v3 ^= v0; \
v2 = ROTL64(v2,32);
for (i = 0, blocks = (len & ~7); i < blocks; i += 8) {
mi = U8TO64_LE(m + i);
v3 ^= mi;
sipcompress()
sipcompress()
v0 ^= mi;
}
switch (len - blocks) {
case 7: last7 |= (uint64_t)m[i + 6] << 48;
case 6: last7 |= (uint64_t)m[i + 5] << 40;
case 5: last7 |= (uint64_t)m[i + 4] << 32;
case 4: last7 |= (uint64_t)m[i + 3] << 24;
case 3: last7 |= (uint64_t)m[i + 2] << 16;
case 2: last7 |= (uint64_t)m[i + 1] << 8;
case 1: last7 |= (uint64_t)m[i + 0] ;
case 0:
default:;
};
v3 ^= last7;
sipcompress()
sipcompress()
v0 ^= last7;
v2 ^= 0xff;
sipcompress()
sipcompress()
sipcompress()
sipcompress()
return v0 ^ v1 ^ v2 ^ v3;
}
/* The 64bit 1-3 variant */
uint64_t
siphash13(const unsigned char key[16], const unsigned char *m, size_t len) {
uint64_t v0, v1, v2, v3;
uint64_t mi, k0, k1;
uint64_t last7;
size_t i, blocks;
k0 = U8TO64_LE(key + 0);
k1 = U8TO64_LE(key + 8);
v0 = k0 ^ 0x736f6d6570736575ull;
v1 = k1 ^ 0x646f72616e646f6dull;
v2 = k0 ^ 0x6c7967656e657261ull;
v3 = k1 ^ 0x7465646279746573ull;
last7 = (uint64_t)(len & 0xff) << 56;
for (i = 0, blocks = (len & ~7); i < blocks; i += 8) {
mi = U8TO64_LE(m + i);
v3 ^= mi;
sipcompress() /* 1 c round */
v0 ^= mi;
}
switch (len - blocks) {
case 7: last7 |= (uint64_t)m[i + 6] << 48;
case 6: last7 |= (uint64_t)m[i + 5] << 40;
case 5: last7 |= (uint64_t)m[i + 4] << 32;
case 4: last7 |= (uint64_t)m[i + 3] << 24;
case 3: last7 |= (uint64_t)m[i + 2] << 16;
case 2: last7 |= (uint64_t)m[i + 1] << 8;
case 1: last7 |= (uint64_t)m[i + 0] ;
case 0:
default:;
};
v3 ^= last7;
sipcompress() /* 1 more c round */
v0 ^= last7;
v2 ^= 0xff;
sipcompress() /* and 3 final d rounds */
sipcompress()
sipcompress()
return v0 ^ v1 ^ v2 ^ v3;
#undef sipcompress
}
