Sparse-ternary-fma – 23x faster AVX-512 kernel for FHE&AI, from my ePrint paper | Not Hacker News!