BQN & CBQN are absolutely wonderful pieces of code. L is mac/lin but linux is avx512 only specifically to try to deal with that problem. The compute on compressed algos helps fit more in those cache lines! https://lv1.sh/blog/compute-on-compressed/