| poly1305_avx2.cpp |
4x26 values packed in a 256-bit register
The 26 bit is somewhat a lie; we actually use the full 64 bit width
but assume that after a 32x32->64 multiply there is still enough
space to store sums into 64 bits. We could pack slightly more bits,
but 26x5 = 130 is enough.
|
9384 |
- |