[jedisct1]

Poly1305 is straightforward to implement and also scales with the width of SIMD vectors, just like other polynomial-based MACs. BLAKE3 is nice but tricky to implement and optimize; a textbook implementation performs very poorly.

Avoid using Rust implementations for serious benchmarks, especially in the context of symmetric cryptography.

For really fast ciphers on general-purpose hardware, check out AEGIS and HiAE. https://github.com/hiae-aead/benchmarks https://datatracker.ietf.org/doc/draft-irtf-cfrg-aegis-aead/ https://datatracker.ietf.org/doc/draft-pham-cfrg-hiae/