[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20120229210758.GA29448@openwall.com>
Date: Thu, 1 Mar 2012 01:07:58 +0400
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: rule: >NN , format=cuda-mscash, filters, --make-charset segfault
Lukas,
On Wed, Feb 29, 2012 at 09:11:26PM +0100, Lukas Odzioba wrote:
> 2012/2/29 <buawig@...il.com>:
> >c) I'm running john 1.7.9-jumbo-5 (mpi) + cuda patch [1]
> >but when starting john:
> >mpiexec -np 8 ./john --format=cuda-mscash -w=... --rules hashes.txt
> ...
> First of all you should not use mpi with cuda. For now it will not give you
> any speedup.
Actually, it might: in my testing of some of your CUDA code, running two
instances at once (without MPI, though) provided slightly better
combined performance, probably due to inefficiencies/stalls in the code.
So "mpiexec -np 2" might make sense in some special cases.
> There was bug in cuda-1 version I fixed it already, and now I am trying to
> upload it to wiki (I am having problems with ISP), so it will be available
> on wiki ASAP.
I see that this is now uploaded - thanks! However, interdiff'ing the
patches I don't see what bug you're referring to, and I don't see
anything relevant to loading of MSCash hashes (something that would
cause them to be loaded where they previously were not).
Can you please clarify?
Thanks,
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.
