hi,

would you mind sharing the performance improvement you achieved using a
better suited data structure (tree vs linked list) on your machine once you
fully implemented the SBOXES for both cyphers?
On my machine (2010 netbook with Intel Atom core using PIL version (3 1 8
0)) i get the following numbers from bench:
8.627 sec '(nth)
5.331 sec '(lup)

freeemint_

2016-10-01 14:41 GMT+02:00 Mike Pechkin <mike.pech...@gmail.com>:

> hi,
>
> Current implementation of bcrypt is slow (8 rounds == >4secs).
> Its ok, nth is not for free.
> o) blowfish encrypt one block of input data takes 32K nth calls.
> o) 8 round's bcrypt takes 16M nth calls.
>
> I will try re-implement blowfish's sboxes access and modifications via
> balance+lup
> Check this out:
> http://pastebin.com/SQ6FtdCx
>
> Note: balance+lup gives more profit under pil64 on the same hardware.
>
> Mike
>
>

Reply via email to