Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 29 Jun 2011 21:44:49 +0400
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: Parallel cracking

On Tue, Jun 28, 2011 at 06:37:08PM -0600, RB wrote:
> On 2011-06-28, Rich Rumble <richrumble@...il.com> wrote:
> > 0:08:00:00 - Switching to length 8
> > 0:08:00:00 - Expanding tables for length 8 to character count 26
> > 0:08:00:00 - Trying length 8, fixed @1, character count 26
> >
> > Have the next thread start at
> >
> > 0:12:00:00 - Switching to length 8
> > 0:12:00:00 - Expanding tables for length 8 to character count 28
> > 0:12:00:00 - Trying length 8, fixed @5, character count 28
> > and so on...
> 
> That is precisely what the MPI patch does

Not precisely.

With the MPI patch, different nodes get adjacent "cracking order"
entries - see the comment on order[] in charset.h.  Each node can
continue running until it hits the end of the order[] array (which never
happens with typical settings, except for all-digit runs and the like),
skipping over other nodes' entries.

In the example above, we'd have a problem after 4 hours, but with the
MPI patch we would not.

Alexander

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.