Taken from google-code
This is WIP... everything quite messy structure/code etc.. its just ment for testing for now
Just noticed rurban's extensions of smhasher on github along with a hash-bench for perl5
Copied over some hash tests from there
In the author of the fast FNV hashes explains that his he has optimized the functions for the Intel Sandy-Bridge architecture. However in 2016 we have Skylake and formerly Hashwell which feature 1 ALU more then Sandy bridge used to have. Therefore the FNV algorithm might be much faster on newer HW if the loop was unrolled 4 times and not 3. Referring to FNV32a_YoshimitsuTRIAD