{ \
constexpr int L0 = Constants::rotate::value[0][(N - 1) % 8]; \
constexpr int R0 = 32 - L0; \
xmms0 = _mm_add_epi32(xmms0, xmms1); \
xmms2 = _mm_add_epi32(xmms2, xmms3); \
xmms4 = _mm_add_epi32(xmms4, xmms5); \
xmms6 = _mm_add_epi32(xmms6, xmms7); \
xmmt1 = _mm_slli_epi32(xmms1, L0); \
xmmt3 = _mm_slli_epi32(xmms3, L0); \
xmmt5 = _mm_slli_epi32(xmms5, L0); \
xmmt7 = _mm_slli_epi32(xmms7, L0); \
xmms1 = _mm_srli_epi32(xmms1, R0); \
xmms3 = _mm_srli_epi32(xmms3, R0); \
xmms5 = _mm_srli_epi32(xmms5, R0); \
xmms7 = _mm_srli_epi32(xmms7, R0); \
xmmt1 = _mm_or_si128(xmms1, xmmt1); \
xmmt3 = _mm_or_si128(xmms3, xmmt3); \
xmmt5 = _mm_or_si128(xmms5, xmmt5); \
xmmt7 = _mm_or_si128(xmms7, xmmt7); \
xmms1 = _mm_xor_si128(xmms0, xmmt1); \
xmms3 = _mm_xor_si128(xmms2, xmmt3); \
xmms5 = _mm_xor_si128(xmms4, xmmt5); \
xmms7 = _mm_xor_si128(xmms6, xmmt7); \
}