[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20120130233602.GA426@openwall.com>
Date: Tue, 31 Jan 2012 03:36:02 +0400
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: DES with OpenMP
On Mon, Jan 30, 2012 at 01:44:35PM +0000, Alex Sicamiotis wrote:
> I was looking now at the john.log, and for openMP 2 threads it was
>
> 0:00:00:00 - Candidate passwords will be buffered and tried in chunks of 8192
>
> for openMP 1 thread it was
>
> 0:00:11:32 - Candidate passwords will be buffered and tried in chunks of 4096
>
> and for non-openMP it was
>
> 0:00:11:38 - Candidate passwords will be buffered and tried in chunks of 128
>
> ...and I remembered what you said about this code difference. So.. supposing this can give, say +200k c/s which I'm seeing between the openmp and non-openmp, where do I tweak the code to test various values?
This is DES_bs_cpt in DES_bs.h, and this setting is only used in OpenMP
builds. By default it's 32, but you can try any value starting with 1.
BTW, I guess those 11 minutes seen in the log above are wasted counting
and validating your preprocessed rules, which you have a ridiculous
number of (600 million, IIRC). Originally, this was not an intended use
for the rules, but I guess I'll need to introduce some better support
for that, which would avoid the waste of time at startup. Would you
trade saving the 11 minutes for lack of a progress indicator maybe (so
that we don't need to count the rules)?
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.
