I helped design Meow Hash using AES-NI. It is not general purpose crypto strength, but ridiculously fast, targeting a theoretical performance of 16 bytes per cycle on some processors, too fast for memory to keep up. https://github.com/cmuratori/meow_hash
Regarding the speed, meow_hash is certainly performant, but when hashing a single file, b3sum actually seems to be faster, perhaps due to using more threads. Could meow_hash be made even faster by running more operations in parallel?
time ~/.cargo/bin/b3sum ubuntu-19.10-beta-desktop-amd64.iso
0899b731b6b57d75a65273fe29d54802653b3bbe3bae6732140c487c4f0ece71 ubuntu-19.10-beta-desktop-amd64.iso
real 0m0,180s
user 0m1,735s
sys 0m0,092s
time meow_example ubuntu-19.10-beta-desktop-amd64.iso
meow_example 0.5/calico - basic usage example of the Meow hash
(C) Copyright 2018-2019 by Molly Rocket, Inc. (https://mollyrocket.com)
See https://mollyrocket.com/meowhash for details.
Hash of "ubuntu-19.10-beta-desktop-amd64.iso":
E5CC524A-522BDE7E-ED34A277-5C7D73AF
This made me curious. Is it because at this stage it is a proposal that has not yet been verified/analysed or are there actual reasons that you know of that make this not "general purpose strong"?
I don't actually have proof that it isn't crypto strength. But comparing it to other algorithms that have been broken, it seems unlikely that it would hold given the rather modest amount of computation done.
I do believe that it meets the requirements for being a MAC function, and I'm completely certain that it is a great non-cryptographic hash function.
