void Threefish_512::decrypt_n(const byte in[], byte out[], size_t blocks) const
{
BOTAN_ASSERT(m_K.size() == 9, "Key was set");
BOTAN_ASSERT(m_T.size() == 3, "Tweak was set");
#define THREEFISH_ROUND(X0,X1,X2,X3,X4,X5,X6,X7,ROT1,ROT2,ROT3,ROT4) \
do { \
X4 ^= X0; \
X5 ^= X1; \
X6 ^= X2; \
X7 ^= X3; \
X4 = rotate_right(X4, ROT1); \
X5 = rotate_right(X5, ROT2); \
X6 = rotate_right(X6, ROT3); \
X7 = rotate_right(X7, ROT4); \
X0 -= X4; \
X1 -= X5; \
X2 -= X6; \
X3 -= X7; \
} while(0)
#define THREEFISH_INJECT_KEY(r) \
do { \
X0 -= m_K[(r ) % 9]; \
X1 -= m_K[(r+1) % 9]; \
X2 -= m_K[(r+2) % 9]; \
X3 -= m_K[(r+3) % 9]; \
X4 -= m_K[(r+4) % 9]; \
X5 -= m_K[(r+5) % 9] + m_T[(r ) % 3]; \
X6 -= m_K[(r+6) % 9] + m_T[(r+1) % 3]; \
X7 -= m_K[(r+7) % 9] + (r); \
} while(0)
#define THREEFISH_DEC_8_ROUNDS(R1,R2) \
do { \
THREEFISH_ROUND(X6,X0,X2,X4, X1,X7,X5,X3, 8,35,56,22); \
THREEFISH_ROUND(X4,X6,X0,X2, X1,X3,X5,X7, 25,29,39,43); \
THREEFISH_ROUND(X2,X4,X6,X0, X1,X7,X5,X3, 13,50,10,17); \
THREEFISH_ROUND(X0,X2,X4,X6, X1,X3,X5,X7, 39,30,34,24); \
THREEFISH_INJECT_KEY(R1); \
\
THREEFISH_ROUND(X6,X0,X2,X4, X1,X7,X5,X3, 44, 9,54,56); \
THREEFISH_ROUND(X4,X6,X0,X2, X1,X3,X5,X7, 17,49,36,39); \
THREEFISH_ROUND(X2,X4,X6,X0, X1,X7,X5,X3, 33,27,14,42); \
THREEFISH_ROUND(X0,X2,X4,X6, X1,X3,X5,X7, 46,36,19,37); \
THREEFISH_INJECT_KEY(R2); \
} while(0)
for(size_t i = 0; i != blocks; ++i)
{
u64bit X0 = load_le<u64bit>(in, 0);
u64bit X1 = load_le<u64bit>(in, 1);
u64bit X2 = load_le<u64bit>(in, 2);
u64bit X3 = load_le<u64bit>(in, 3);
u64bit X4 = load_le<u64bit>(in, 4);
u64bit X5 = load_le<u64bit>(in, 5);
u64bit X6 = load_le<u64bit>(in, 6);
u64bit X7 = load_le<u64bit>(in, 7);
THREEFISH_INJECT_KEY(18);
THREEFISH_DEC_8_ROUNDS(17,16);
THREEFISH_DEC_8_ROUNDS(15,14);
THREEFISH_DEC_8_ROUNDS(13,12);
THREEFISH_DEC_8_ROUNDS(11,10);
THREEFISH_DEC_8_ROUNDS(9,8);
THREEFISH_DEC_8_ROUNDS(7,6);
THREEFISH_DEC_8_ROUNDS(5,4);
THREEFISH_DEC_8_ROUNDS(3,2);
THREEFISH_DEC_8_ROUNDS(1,0);
store_le(out, X0, X1, X2, X3, X4, X5, X6, X7);
in += 64;
out += 64;
}
#undef THREEFISH_DEC_8_ROUNDS
#undef THREEFISH_INJECT_KEY
#undef THREEFISH_ROUND
}
void Threefish_512_AVX2::decrypt_n(const byte in[], byte out[], size_t blocks) const
{
const u64bit* K = &get_K()[0];
const u64bit* T_64 = &get_T()[0];
const __m256i ROTATE_1 = _mm256_set_epi64x(37,19,36,46);
const __m256i ROTATE_2 = _mm256_set_epi64x(42,14,27,33);
const __m256i ROTATE_3 = _mm256_set_epi64x(39,36,49,17);
const __m256i ROTATE_4 = _mm256_set_epi64x(56,54, 9,44);
const __m256i ROTATE_5 = _mm256_set_epi64x(24,34,30,39);
const __m256i ROTATE_6 = _mm256_set_epi64x(17,10,50,13);
const __m256i ROTATE_7 = _mm256_set_epi64x(43,39,29,25);
const __m256i ROTATE_8 = _mm256_set_epi64x(22,56,35, 8);
#define THREEFISH_ROUND(X0, X1, SHR) \
do { \
const __m256i SHL = _mm256_sub_epi64(_mm256_set1_epi64x(64), SHR); \
X0 = _mm256_permute4x64_epi64(X0, _MM_SHUFFLE(2, 1, 0, 3)); \
X1 = _mm256_permute4x64_epi64(X1, _MM_SHUFFLE(1, 2, 3, 0)); \
X1 = _mm256_xor_si256(X1, X0); \
X1 = _mm256_or_si256(_mm256_sllv_epi64(X1, SHL), _mm256_srlv_epi64(X1, SHR)); \
X0 = _mm256_sub_epi64(X0, X1); \
} while(0)
#define THREEFISH_INJECT_KEY(X0, X1, R, K0, K1, T0I, T1I) \
do { \
const __m256i T0 = _mm256_permute4x64_epi64(T, _MM_SHUFFLE(T0I, 0, 0, 0)); \
const __m256i T1 = _mm256_permute4x64_epi64(T, _MM_SHUFFLE(0, T1I, 0, 0)); \
X0 = _mm256_sub_epi64(X0, K0); \
X1 = _mm256_sub_epi64(X1, K1); \
X1 = _mm256_sub_epi64(X1, R); \
R = _mm256_sub_epi64(R, ONE); \
X0 = _mm256_sub_epi64(X0, T0); \
X1 = _mm256_sub_epi64(X1, T1); \
} while(0)
#define THREEFISH_DEC_8_ROUNDS(X0, X1, R, K1, K2, K3, T0, T1, T2) \
do { \
THREEFISH_INJECT_KEY(X0, X1, R, K2, K3, T2, T0); \
THREEFISH_ROUND(X0, X1, ROTATE_8); \
THREEFISH_ROUND(X0, X1, ROTATE_7); \
THREEFISH_ROUND(X0, X1, ROTATE_6); \
THREEFISH_ROUND(X0, X1, ROTATE_5); \
\
THREEFISH_INJECT_KEY(X0, X1, R, K1, K2, T0, T1); \
THREEFISH_ROUND(X0, X1, ROTATE_4); \
THREEFISH_ROUND(X0, X1, ROTATE_3); \
THREEFISH_ROUND(X0, X1, ROTATE_2); \
THREEFISH_ROUND(X0, X1, ROTATE_1); \
} while(0)
/*
v1.0 key schedule: 9 ymm registers (only need 2 or 3)
(0,1,2,3),(4,5,6,7) [8]
then mutating with vpermq
*/
const __m256i K0 = _mm256_set_epi64x(K[6], K[4], K[2], K[0]);
const __m256i K1 = _mm256_set_epi64x(K[7], K[5], K[3], K[1]);
const __m256i K2 = _mm256_set_epi64x(K[8], K[6], K[4], K[2]);
const __m256i K3 = _mm256_set_epi64x(K[0], K[7], K[5], K[3]);
const __m256i K4 = _mm256_set_epi64x(K[1], K[8], K[6], K[4]);
const __m256i K5 = _mm256_set_epi64x(K[2], K[0], K[7], K[5]);
const __m256i K6 = _mm256_set_epi64x(K[3], K[1], K[8], K[6]);
const __m256i K7 = _mm256_set_epi64x(K[4], K[2], K[0], K[7]);
const __m256i K8 = _mm256_set_epi64x(K[5], K[3], K[1], K[8]);
const __m256i ONE = _mm256_set_epi64x(1, 0, 0, 0);
const __m256i* in_mm = reinterpret_cast<const __m256i*>(in);
__m256i* out_mm = reinterpret_cast<__m256i*>(out);
for(size_t i = 0; i != blocks; ++i)
{
__m256i X0 = _mm256_loadu_si256(in_mm++);
__m256i X1 = _mm256_loadu_si256(in_mm++);
const __m256i T = _mm256_set_epi64x(T_64[0], T_64[1], T_64[2], 0);
__m256i R = _mm256_set_epi64x(18, 0, 0, 0);
interleave_epi64(X0, X1);
THREEFISH_DEC_8_ROUNDS(X0, X1, R, K8,K0,K1, 3, 1, 2);
THREEFISH_DEC_8_ROUNDS(X0, X1, R, K6,K7,K8, 2, 3, 1);
THREEFISH_DEC_8_ROUNDS(X0, X1, R, K4,K5,K6, 1, 2, 3);
THREEFISH_DEC_8_ROUNDS(X0, X1, R, K2,K3,K4, 3, 1, 2);
THREEFISH_DEC_8_ROUNDS(X0, X1, R, K0,K1,K2, 2, 3, 1);
THREEFISH_DEC_8_ROUNDS(X0, X1, R, K7,K8,K0, 1, 2, 3);
THREEFISH_DEC_8_ROUNDS(X0, X1, R, K5,K6,K7, 3, 1, 2);
THREEFISH_DEC_8_ROUNDS(X0, X1, R, K3,K4,K5, 2, 3, 1);
THREEFISH_DEC_8_ROUNDS(X0, X1, R, K1,K2,K3, 1, 2, 3);
THREEFISH_INJECT_KEY(X0, X1, R, K0, K1, 2, 3);
deinterleave_epi64(X0, X1);
_mm256_storeu_si256(out_mm++, X0);
_mm256_storeu_si256(out_mm++, X1);
}
#undef THREEFISH_DEC_8_ROUNDS
#undef THREEFISH_ROUND
#undef THREEFISH_INJECT_KEY
#undef THREEFISH_DEC_2_8_ROUNDS
#undef THREEFISH_ROUND_2
#undef THREEFISH_INJECT_KEY_2
}
void Threefish_512::decrypt_n(const uint8_t in[], uint8_t out[], size_t blocks) const
{
BOTAN_ASSERT(m_K.size() == 9, "Key was set");
BOTAN_ASSERT(m_T.size() == 3, "Tweak was set");
#if defined(BOTAN_HAS_THREEFISH_512_AVX2)
if(CPUID::has_avx2())
{
return avx2_decrypt_n(in, out, blocks);
}
#endif
#define THREEFISH_ROUND(X0,X1,X2,X3,X4,X5,X6,X7,ROT1,ROT2,ROT3,ROT4) \
do { \
X4 ^= X0; \
X5 ^= X1; \
X6 ^= X2; \
X7 ^= X3; \
X4 = rotate_right(X4, ROT1); \
X5 = rotate_right(X5, ROT2); \
X6 = rotate_right(X6, ROT3); \
X7 = rotate_right(X7, ROT4); \
X0 -= X4; \
X1 -= X5; \
X2 -= X6; \
X3 -= X7; \
} while(0)
#define THREEFISH_INJECT_KEY(r) \
do { \
X0 -= m_K[(r ) % 9]; \
X1 -= m_K[(r+1) % 9]; \
X2 -= m_K[(r+2) % 9]; \
X3 -= m_K[(r+3) % 9]; \
X4 -= m_K[(r+4) % 9]; \
X5 -= m_K[(r+5) % 9] + m_T[(r ) % 3]; \
X6 -= m_K[(r+6) % 9] + m_T[(r+1) % 3]; \
X7 -= m_K[(r+7) % 9] + (r); \
} while(0)
#define THREEFISH_DEC_8_ROUNDS(R1,R2) \
do { \
THREEFISH_ROUND(X6,X0,X2,X4, X1,X7,X5,X3, 8,35,56,22); \
THREEFISH_ROUND(X4,X6,X0,X2, X1,X3,X5,X7, 25,29,39,43); \
THREEFISH_ROUND(X2,X4,X6,X0, X1,X7,X5,X3, 13,50,10,17); \
THREEFISH_ROUND(X0,X2,X4,X6, X1,X3,X5,X7, 39,30,34,24); \
THREEFISH_INJECT_KEY(R1); \
\
THREEFISH_ROUND(X6,X0,X2,X4, X1,X7,X5,X3, 44, 9,54,56); \
THREEFISH_ROUND(X4,X6,X0,X2, X1,X3,X5,X7, 17,49,36,39); \
THREEFISH_ROUND(X2,X4,X6,X0, X1,X7,X5,X3, 33,27,14,42); \
THREEFISH_ROUND(X0,X2,X4,X6, X1,X3,X5,X7, 46,36,19,37); \
THREEFISH_INJECT_KEY(R2); \
} while(0)
BOTAN_PARALLEL_FOR(size_t i = 0; i < blocks; ++i)
{
uint64_t X0, X1, X2, X3, X4, X5, X6, X7;
load_le(in + BLOCK_SIZE*i, X0, X1, X2, X3, X4, X5, X6, X7);
THREEFISH_INJECT_KEY(18);
THREEFISH_DEC_8_ROUNDS(17,16);
THREEFISH_DEC_8_ROUNDS(15,14);
THREEFISH_DEC_8_ROUNDS(13,12);
THREEFISH_DEC_8_ROUNDS(11,10);
THREEFISH_DEC_8_ROUNDS(9,8);
THREEFISH_DEC_8_ROUNDS(7,6);
THREEFISH_DEC_8_ROUNDS(5,4);
THREEFISH_DEC_8_ROUNDS(3,2);
THREEFISH_DEC_8_ROUNDS(1,0);
store_le(out + BLOCK_SIZE*i, X0, X1, X2, X3, X4, X5, X6, X7);
}
#undef THREEFISH_DEC_8_ROUNDS
#undef THREEFISH_INJECT_KEY
#undef THREEFISH_ROUND
}