I was reading the BLAKE2 paper and noticed that when you use BLAKE2 in keyed mode the key is padded to a multiple of the block length, and then the message you hash is placed after. So keyed BLAKE2 always does at least two iterations. BLAKE2 and ChaCha20 have similar cpb, but when considering short (<= block length) inputs, ChaCha20 might significantly outperform keyed BLAKE2.
Benchmarks should be implemented first, and then it should be pretty easy to test this (see #2).
I was reading the BLAKE2 paper and noticed that when you use BLAKE2 in keyed mode the key is padded to a multiple of the block length, and then the message you hash is placed after. So keyed BLAKE2 always does at least two iterations. BLAKE2 and ChaCha20 have similar cpb, but when considering short (<= block length) inputs, ChaCha20 might significantly outperform keyed BLAKE2.
Benchmarks should be implemented first, and then it should be pretty easy to test this (see #2).