[boost] Re: Re: Re: [review] hash functions

From: Thorsten Ottosen (nesotto_at_[hidden])
Date: 2005-03-10 13:53:33

"Peter Dimov" <pdimov_at_[hidden]> wrote in message 
news:000901c52580$359f96a0$6601a8c0_at_pdimov...

| Simplistic benchmarks can't be used to evaluate the performance of a
| component; it should be benchmarked in its usual role.

right. I just stumpled about this parag. :

" Update(1): David C. wrote: I tested your hash function against all of the 
popular ones, including Bob Burtles. It turns out it was not only the quickest 
but had the best distribution (I created histograms of the chain lengths). The 
architecture I work on is IBM Z/OS (S/390 mainframes). Well done mate, will be 
using your code from now on! Ok, not exactly RISC, but at least this 
demonstrates that this function is good beyond the x86 architecture. "

from http://www.azillionmonkeys.com/qed/hash.html

-Thorsten