void SPECK128_Dec_6_Blocks(uint32x4_p &block0, uint32x4_p &block1,
uint32x4_p &block2, uint32x4_p &block3, uint32x4_p &block4,
uint32x4_p &block5, const word64 *subkeys, unsigned int rounds)
{
#if (CRYPTOPP_BIG_ENDIAN)
const uint8x16_p m1 = {31,30,29,28,27,26,25,24, 15,14,13,12,11,10,9,8};
const uint8x16_p m2 = {23,22,21,20,19,18,17,16, 7,6,5,4,3,2,1,0};
#else
const uint8x16_p m1 = {7,6,5,4,3,2,1,0, 23,22,21,20,19,18,17,16};
const uint8x16_p m2 = {15,14,13,12,11,10,9,8, 31,30,29,28,27,26,25,24};
#endif
// [A1 A2][B1 B2] ... => [A1 B1][A2 B2] ...
uint64x2_p x1 = (uint64x2_p)VecPermute(block0, block1, m1);
uint64x2_p y1 = (uint64x2_p)VecPermute(block0, block1, m2);
uint64x2_p x2 = (uint64x2_p)VecPermute(block2, block3, m1);
uint64x2_p y2 = (uint64x2_p)VecPermute(block2, block3, m2);
uint64x2_p x3 = (uint64x2_p)VecPermute(block4, block5, m1);
uint64x2_p y3 = (uint64x2_p)VecPermute(block4, block5, m2);
for (int i = static_cast<int>(rounds-1); i >= 0; --i)
{
const uint64x2_p rk = vec_splats((unsigned long long)subkeys[i]);
y1 = VecXor(y1, x1);
y2 = VecXor(y2, x2);
y3 = VecXor(y3, x3);
y1 = RotateRight64<3>(y1);
y2 = RotateRight64<3>(y2);
y3 = RotateRight64<3>(y3);
x1 = VecXor(x1, rk);
x2 = VecXor(x2, rk);
x3 = VecXor(x3, rk);
x1 = VecSub(x1, y1);
x2 = VecSub(x2, y2);
x3 = VecSub(x3, y3);
x1 = RotateLeft64<8>(x1);
x2 = RotateLeft64<8>(x2);
x3 = RotateLeft64<8>(x3);
}
#if (CRYPTOPP_BIG_ENDIAN)
const uint8x16_p m3 = {31,30,29,28,27,26,25,24, 15,14,13,12,11,10,9,8};
const uint8x16_p m4 = {23,22,21,20,19,18,17,16, 7,6,5,4,3,2,1,0};
#else
const uint8x16_p m3 = {7,6,5,4,3,2,1,0, 23,22,21,20,19,18,17,16};
const uint8x16_p m4 = {15,14,13,12,11,10,9,8, 31,30,29,28,27,26,25,24};
#endif
// [A1 B1][A2 B2] ... => [A1 A2][B1 B2] ...
block0 = (uint32x4_p)VecPermute(x1, y1, m3);
block1 = (uint32x4_p)VecPermute(x1, y1, m4);
block2 = (uint32x4_p)VecPermute(x2, y2, m3);
block3 = (uint32x4_p)VecPermute(x2, y2, m4);
block4 = (uint32x4_p)VecPermute(x3, y3, m3);
block5 = (uint32x4_p)VecPermute(x3, y3, m4);
}
void SPECK64_Dec_Block(uint32x4_p &block0, uint32x4_p &block1,
const word32 *subkeys, unsigned int rounds)
{
#if (CRYPTOPP_BIG_ENDIAN)
const uint8x16_p m1 = {7,6,5,4, 15,14,13,12, 23,22,21,20, 31,30,29,28};
const uint8x16_p m2 = {3,2,1,0, 11,10,9,8, 19,18,17,16, 27,26,25,24};
#else
const uint8x16_p m1 = {3,2,1,0, 11,10,9,8, 19,18,17,16, 27,26,25,24};
const uint8x16_p m2 = {7,6,5,4, 15,14,13,12, 23,22,21,20, 31,30,29,28};
#endif
// [A1 A2 A3 A4][B1 B2 B3 B4] ... => [A1 A3 B1 B3][A2 A4 B2 B4] ...
uint32x4_p x1 = VecPermute(block0, block1, m1);
uint32x4_p y1 = VecPermute(block0, block1, m2);
for (int i = static_cast<int>(rounds-1); i >= 0; --i)
{
#if CRYPTOPP_POWER7_AVAILABLE
const uint32x4_p rk = vec_splats(subkeys[i]);
#else
// subkeys has extra elements so memory backs the last subkey
const uint8x16_p m = {0,1,2,3, 0,1,2,3, 0,1,2,3, 0,1,2,3};
uint32x4_p rk = VecLoad(subkeys+i);
rk = VecPermute(rk, rk, m);
#endif
y1 = VecXor(y1, x1);
y1 = RotateRight32<3>(y1);
x1 = VecXor(x1, rk);
x1 = VecSub(x1, y1);
x1 = RotateLeft32<8>(x1);
}
#if (CRYPTOPP_BIG_ENDIAN)
const uint8x16_p m3 = {19,18,17,16, 3,2,1,0, 23,22,21,20, 7,6,5,4};
const uint8x16_p m4 = {27,26,25,24, 11,10,9,8, 31,30,29,28, 15,14,13,12};
#else
const uint8x16_p m3 = {3,2,1,0, 19,18,17,16, 7,6,5,4, 23,22,21,20};
const uint8x16_p m4 = {11,10,9,8, 27,26,25,24, 15,14,13,12, 31,30,29,28};
#endif
// [A1 A3 B1 B3][A2 A4 B2 B4] => [A1 A2 A3 A4][B1 B2 B3 B4]
block0 = (uint32x4_p)VecPermute(x1, y1, m3);
block1 = (uint32x4_p)VecPermute(x1, y1, m4);
}