[Feature Request:] Blake2sp, Blake2bp and 2 more.

[lelik107]

Hello! I'd like to ask you for adding these 2 hash functions.

Blake2sp-256, Blake2bp-512 as they are. This is a reference implementation as it was presented for NIST hash function competition aka SHA-3.
https://github.com/BLAKE2/BLAKE2

p variants are designed to run in parallel: 8 threads for Blake2sp-256 and 4 threads for Blake2bp-512 but even in the SW that doesn't use multi threading they usually twice faster than Blake2s-256 and Blake2b-512.

The author of any official Blake2 С implementation is Samuel Neves, a member of BLAKE3 team too. https://github.com/sneves
AVX2 didn't exist when SHA-3 competition was and later he did this implementation: https://github.com/sneves/blake2-avx2
It marked as experimental but in this context this isn't like alpha or beta, this just means it's not the very code from SHA-3 competition. As I know AVX2 variant by itself has no fallback code (if you have no AVX2 - you have no hash), though it's used in 7-zip for example (I don't know where Igor Pavlov got AVX512 optimization) - just for Blake2sp-256, because RAR5 format uses exactly it.
https://github.com/ip7z/7zip/blob/main/C/Blake2s.c
https://github.com/ip7z/7zip/blob/main/C/Blake2.h
I'm not sure which variant is more appropriate for RHash. Thus maybe it's better to go with the reference one.

The feature request ends here, the other 2 are the suggestions for Aleksey if he finds them interesting.

Thank you.

[lelik107]

KangarooTwelve aka K-12, basically Keccak (SHA-3) with the rounds reduced by half 24/2=12 and some bells and whistles.
https://keccak.team/kangarootwelve.html
https://github.com/XKCP/K12
https://github.com/XKCP/XKCP
It usually performs better then any standardized Keccak Team designs FIPS 202 (SHA3, SHAKE) or NIST SP.800-185 (cSHAKE, KMAC, TupleHash, and ParallelHash). And there is no evidence it's insecure in any way.
https://keccak.team/sw_performance.html

LSH is a current South Korean national standard KS X 3262, https://en.wikipedia.org/wiki/LSH_(hash_function)
and according to this with a decent performance:
http://bench.cr.yp.to/results-hash/amd64-hertz.html (LSH256 and LSH512)
The problem is that the docs are in Korean. The sources are here, KISA is Korea Internet & Security Agency:
https://seed.kisa.or.kr/async/MultiFile/download.do?FS_KEYNO=FS_0000000364&MNK=MN_0000001247
And a part of CryptoPP: https://github.com/weidai11/cryptopp (everything with lsh).