This is somewhat common in branchless parsers. You need a few more bytes at the end of the sequence in order to allow efficient reads to take place. This isn't

This is somewhat common in branchless parsers. You need a few more bytes at the end of the sequence in order to allow efficient reads to take place. This isn't uncommon, and is why it's pretty important that performance-critical software also RTFM.