Daniel Lemire's blog

, 1 min read

Transcoding Unicode strings at crazy speeds with AVX-512

2 thoughts on “Transcoding Unicode strings at crazy speeds with AVX-512”

  1. KWillets says:

    I think I understand what you’re doing in expand_and_identify, you’re trying every alignment of 4 bytes and compressing the ones that coincide with initials?

    1. I see you have been reading the code. This was the original approach for UTF-8 to UTF-16 but it has been replaced with something more efficient.