Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 28 Jun 2013 22:39:37 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: Parallella: bcrypt

On Fri, Jun 28, 2013 at 05:29:32PM +0200, Katja Malvoni wrote:
> On Fri, Jun 28, 2013 at 11:16 AM, Solar Designer <solar@...nwall.com> wrote:
> > On Fri, Jun 28, 2013 at 10:56:20AM +0200, Katja Malvoni wrote:
> > > With other version of BF_ROUND macro and with "#if 1" execution speed on
> > > one core is 39.131000 ms and 288.555000 ms on all cores.
> >
> > The relative speedup from the two changes mentioned above is as I had
> > expected it, but these are still rather poor speeds.  Anyway, please
> > proceed to integrate this into JtR, and we'll benchmark from there.
> 
> Can you please give me some tips on how to start integration?

You may either create a new JtR format on top of your existing full
bcrypt on Epiphany code, or you may start with the approach shown in
that john-1.7.8-fpga-hooks.diff.gz patch, but then you need to transfer
the 4 KB blocks back and forth.  It is non-obvious which of these two
ways is easier and which will result in better performance.

To learn the JtR formats interface, take a look at formats.h (it has
comments on the various format methods), BF_fmt.c, and maybe dummy.c and
c3_fmt.c.

> I was working on enabling all cores to work with different key and setting.

As I mentioned before, you don't need different setting per core.  Only
different key per core.

Alexander

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ