Hey, OP here. Code not written by me, but by a friend (iscgar) with no HN account.

On my late 2019 MacBook Pro, using clang 12.

    clang -O3 -march=native fizz4.c
    ./a.out | pv > /dev/null

Starts at around 4GiB/s and drops down to 3.8 after a while, probably due to battery/heating.